MA,TSA (7H( SETTS INSTIT(-TE OF TE~H.

VOLOG1

RADIATION L.ABOR.iTORY SERIES

~oar(l of l<klitors

LorIs X. RI!) EXOIR, IldI/oJ-zP1-Clzicf

GEORGE 13. COLT.IXS, I)epv(y I?dllor-iw(hiej

13RITIY)X CF{.4XCE, S. .4. (20[TDSMIT, R. G. HERB, HUBERT XI. J.ANES, JI-LIAN K. Kxlrr,
JAMES I,. L.A}VSON,LEON B. LISFORO, CAROL G. l[oxrco~lERy, C. \TEWTON, .41.BER~
hl. STOXE,LOCTIS
A. TCRNER,GEORGE13. VALLEY, JR., IIERBERT H. JVHE+TON

1, RADAR SYSTEM ~NG1xEERINGI?idenOILT

2. IL.* DAR.41Ds TO NAVIGATIONHdt
3. RADAR BEAcoNsRobe?is
4. LOIZAxPierce, McKenzie, and lfood trar<i
5. ~~1.sE GE NERAroRs-G/asOe and ~.i{mc(lz
6. LIICROWAVE NIAGxETRoxsCOl/l /(,s
7. KLYSTRONS AND MICROIV,+VETRIOIIES Ilm Ilf/[on, Knipp, and Kt[per
8. PRINCIPLES OF MICROWAVE CIII(L lrfi .Ifo,dqo), tei-y, Ddie, and luwell
9. hIICROJVAVETRAXSMISSIOX CIRCI Im - 12[Lyan
10, \~AVEGUIDE HAsDBOOK3~arc( (7i/z
11. TECHNIQUE OF iWCROWAVE JIE.iSLmMEWY ,IIOII!{IOUIWY

12. klICRO\VAVEANTEINXA TIIEOFCY.+x I) L)ESIGX --,iiller

13. PROPAGATION OF SHORT RAI)IO IV.\vEs -Keru
14. IIflCROWAVE ~uPI.ExERs-S)) L7(/lLrL
un<l .lIot<([yo!ltery
15. CRYSTAL RECTIFIERS-TrMY and 11k~t))Ler
16. }lICROWAVE hfIxERs~ounft
17. COMPONENTS HANI)BOOK~/Uck/)7L~n
18. VACUUM TUBE AMPLIFIERS- T-(i//c!/ aml Tla//,rLan
19. \TAvEFOW.ls(;hanceJ Hfighes, .l/ac.\zcho/, Sayle, and l\tllia JfLs
20, ~LECTRONIC TIME hIEAScREMENTs-(/mnce, Hu/sizer, ~[fw.~lcho~,
and tFilliams
21. ~1.ECTROXIC Instruments-Greenlrood, Hofakl))], un~{ .~~uc]iae
22. CATHODE RAY TUBE D1spl,AysSo//er, Starr, and talkg
23, llICROWAVE RECEIYERS-Van Ioorhis
24. THRESHOLI) SIG NALSLa?OSOn a!Ld ( hlenheck
25. THEORY OF SERYOMECIIAXISMS -Ja?,L&!, .I-lchols, and Phillips
26. RADAR SCAXXERS AXD R.41)oA1~s-Cad!/, Kareltlz, and Turner
27. COMPUTIXG hlECH.+XISMS AND I,lxK.iGEs-Swboda
28. INDEXHelLnQj

\ ,.
MICROWAVE ANTENNA
THEORY AND DESIGN
Ediied by
SAMUEL SILVER
ASSOCIATE PROFESSOR OF ELECTRICAL ENGINEERING
UNNEB.SITY OF CALIFORNIA, i3EP.KELEY

OFFICE OF SCIENTIFIC RESEARCH AND DEVELOPMENT

NATIONAL DEFENSE RESEARCH COMMITTEE

FIRST EDITION

NEW YORK, TORONTO LONDON

McGRAW-HILL BOOK CO,MPANY, INC.
1949
 .,
,.,

MICROWAVE .$xTEN\-.$ THEC!R Y .ISD DESIGN

(hPYRIGH,T, 1949, B>- THE

hlC~RA W-HILL BOOK ~(IIIP.I.NY, lKC.

P31XTEI) lx THE U>-lTEI) STATES OF AMERICA

.111 rights Testwed. This book, or

parts thereof, HI(IY not be reproduced
in any form (rilho?(l prr)rlission of
/he ,L)(//)
/ishers,

THE MAPLE PRESS COMPANY, YORK, PA,

*ienCe

m
 ,,
)/
/y,,
,, . \

lf[CRO JV.4 VE A NTE.VNA TfZEOR Y AND DESIGN

EDITORIAL STAFF

SAMUEL SILVER

HUBERT M. JAMES

CO.VTRIB LTI.VG A PTHORS

J. E. llATON R. hf. R IZDHEFFER

L. J. I;YGES J. R. RISSER

T. J. KEARY S. SILVER

H. KRUTTER O. A. TYSON

(2. G. hlAcFARL.4NE L. C. \AN ATTA

 Foreword

HE tremendous research and development effort that ~vent into the

T development of radar and related techniques during }Vorld IJ ar II
resulted not only in hundreds of radar sets for military (and some for
possible peacetime) use but also in a great body of information and ncm
techniques in the electronics and high-frequency fields. 13ecause this
basic material may be of great value to science and engineering, it seemed
most important to publish it as soon as security permitted.
The Radiation Laboratory of 311T, ~vhich operated under the super-
vision of the National Defense Research (ommittec, undertook the great
task of preparing these volumes. The ~vorl{ described berein, ho\\-eyer,is
the collective result of ~vork done at many laboratories, Army, Xavy,
university, and industrial, both in this country and in JZngland, (<anada,
and other Dominions.
The Radiation Laboratory, once its proposals ]vere approved and
finances provided by the Office of Scientific Research and l)evelopment,
chose Louis N. Ridenour as Fklitor-in-(bief to led and direct tbe entire
-, project. An editorial staff ]vas then selected of those best quulificd for
this type of task. Finally the authors for the various volumes or chapters
or sections were chosen from among those experts ~vho ~t-ere intimately
familiar with the various fields, and ]vbo \vere able and willing to ]vrite
the summaries of them. This entire staff agreed to remain at ~vork at
MIT for six months or more after the \\-orkof the Radiation I.aboratory
was complete. These volumes stand as a monument to this group.
These volumes serve as a memorial to tbe unnamed hundreds and
thousands of other scientists, engineers, and others ]vho actually carried
on the research, development, and engineering work tbe results of which
are herein described. There ~vere so many involved in this ~vork and they
worked so closely together even though often in \\-idelyseparated labora-
tories that it is impossible to name or even to know those ]vho contributed
to a particular idea or development, (My certain ones ~vho~u-ote reports
w- or articles have even been mentioned, But to all those ~vho contributed
~ in any way to this great cooperative development enterprise, both in this
~ country and in England, these volumes are dedicated,
a
L. A. DLTBRIDGE.
z 1,!
 Preface

HE need that arose during the ]var for utilizing the microwave region
T of the radio frequency spectrum for communications and radar stimu-
lated the development of nelv types of antennas. lhe problems and
design techniques, lying as they do in the domain of both applied electro-
magnetic theory and optics, are quite distinct from those of long-wave
antennas. It is the aim of the present volume to make available to the
antenna engineer a systematic treatment of the basic principles and the
fundamental microwave antenna types and techniques. The elements
of electromagnetic theory and physical optics that are needed as a basis
for design techniques are developed quite fully. Critical attention is
paid to the assumptions and approximations that are commonly made
in the theoretical developments to emphasize the domain of applicability
of the results. The subject of geometrical optics has been treated only
to the extent necessary to formulate its basic principles and to sho~v its
relation as a short wavelength approximation to the more exact methods
of field theory. The brevity of treatment should not be taken as an
index of the relative importance of geometrical optics to that of electro-
magnetic theory and physical optics. It is in fact true that the former
is generally the starting point in the design of the optical elements
(reflectors and lenses) of an antenna. However, the use of ray theory
for microwave systems presents no new problems over those encountered
in opticson which there are a number of excellent treatisesexcept
that perhaps the law of the optical path appears more prominently in
micro~vave applications.
In the original planning of the book it was the intention of the editors
to integrate all of the major wQrk done in this country and in Great
Britoin and Canada. This proved, however, to be too ambitious an
undertaking. Nfany subjects have regrettably been omitted completely,
and others have had to be treated in a purely cursory manner. It \vas
unfortunately necessary to omit two chapters on rapid scanning antennas
prepared by Dr. C. V. Robinson. The time required to revise the
material to conform ~vith the requirements of military security and yet
to represent an adequate exposition of the subject would have unduly
delayed the publication of the hook. Certain sections of Dr. Robinsons
material have been incorporated into Chaps. 6 and 12.
ix
x PREFACE

I take pleasure in expressing here my appreciation to Prof. Hubei-t

M. James who, as Technical Editor, shared with me much of the
editorial work and the attendant responsibilities. The scope of the book,
the order of presentation of the material, and the sectional division within
chapters were arrived at by us jointly in consultation with the authors.
I am personally indebted to Professor .James for his editorial Ivork on
my own chapters.
The responsibility for the final form of the book, the errors of omission
and commission, is mine. A word of explanation to the authors of the
various chapters is in order. After the close of the Office of Iublications
and the dispersal of the group, I have on occasions made use of my
editorial prerogative to revise their presentations. I hope that the results
meet ~vith their approval. The policy of assignment of credit also needs
explanation. The interpretation of both Professor James and myself of
the policy on credit assignment formulated by the Editorial Board for
the Technical Series has been to the effect that no piece of work discussed
in the text would be associated with an individual or individuals. Radi-
ation Laboratory reports are referred to in the sense that they represent
source material for the chapter rather than individual acknowledgements.
References to unpublished material of the Radiation Laboratory note-
books have been assiduously a~oided, although such material has been
dramm upon extensively by all of us. In defense of this policy it may be
stated that the ]vorlc at the Radiation Laboratory was truly a cooperative
effort, and in only a few instances would it have been possible to assign
individual credit unequivocally.
The completion of the book was made possible through the efforis of
a number of people; in behalf of the editorial staff and the authors I wish
to acknowledge their assistance and contributions. Mrs. Barbara Vogel
and Mrs. Ellen Fine of the Radiation Laboratory served as technical
assistants; the production of figures and photographs \vas expedited by
hlrs. Frances Bourget and Mrs. nary Sheats. It proved impossible to
finish the ]t-orl<by the closing date of the Office of lublications; the haval
Research Laboratory accepted the ~vork as one of the projects of the
newly formed Antenna Research Section and contributed generously in
personnel and facilities. Special thanks are due to A. S. Dunbar,
1, Katz, and Dr. I. Maddaus for their editorial assistance; to Queenie
Parigian and Louise Beltramini for preparation of the manuscript;
and to Betty Hodgkins who prepared almost all of the figures.
The editors are indebted to Dr. G. G. Macfarlane of the Tele-
communications Research Establishment, Great Britainj for his
critical review of several of the theoretical chapters and his contribution
on the theory of slot radiators in Chap. 9. John Powell of the
Radiation Laboratory prepared material on lenses that was used in
Ch:lp. 11. The S:1( iomd Rcsc:wch (ouncil of Can:&~ :md the llrit isll
(entnd Radio 13urw~u h~~vc ~rwiously granted us permission to ti~li(.
m:ltcrial from ( unudi:m :md I;ritish reports in accord:mcc ~~ith mlrrrnt
security U3glllotioms. l>hc I?wII Telephone I.abora,twy supplied the
photographs of mct:d lens antennas.

S.4 MUEL khLVlil{.

K:\v\T, lll)sl.\1i1lI T. WIMWIIY,
!flslllxlm)x, l). (.,

:lprd, 1947.
 .

Contents

FORE WORD BY L. A. DUBRmGE . . . . . . . . . . . . . . . . . . vii

PRE1744CE
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix

CHAP.1. SURVEY OF MICROWAVE ANT~~NNADESIGN PROBLEMS 1

1.1. The WavelcngthRegion. . . . . . 1
1.2. .Lntenna Patterns . . . . . . . . . . . . . . . . . . . ..2
1.3. Types of }Iicrowave Beams. 6
1.4. lIicrowave Transmission I,ines . 7
1.5. Radiating llernents . . . . . . . . . ..8
16. .4 Survey of kllcrowavc }.ntenna Types 9
1.7. Impedance Specifications. 13
1.8. Program of the Present Volume 14

CH.4P. 2. CIRCUIT RJ31JATIOIW, Rf3CIPR0CiTY THW3RF~!>fS. 16

21. Introduction . . . . . . . . . . . . . . . . . . . ...16
22. The Four-terminal ~etwork. 17
23. The Rayleigh Reciprocity Theorcnl 19
2.4. Th6venins Theorem and the Nfzximum-power Theorem 20
2.5. The Two-wire Transmission I,ine 21
2.6. The Homogeneous Transmission I,ine 23
2.7. The LosslessLin e..... ..26
2.8. Transformation Charts. 29
2.9. The Four-terminal Network l:quivalent of a Sectio]l of Trans-
missiOn Line.... . . . . . 36

TRANSMI~EVG ANn RECEIVING ANTENNAS. . 37

2.10. The Antema as a Terminating Impedance 37
2.11. The Receiving Antenna System 40
2.12. The Transmitter and Receiver as a Coupled System 45
2.13. Reciprocity between the Transmitting and Itecei\,ing Patterns of
an Antenna . . . . . . . . . . . . . . . . . . . ...48
2.14. The .kverage Cross Section for a Matched System 50
2.15. Dependence of the Cross Section on Antenna Mismatch 51
2.16. The Four-terminal Network Representation. 53
2.17. l)evelopment of the Network Equations 56
2.18. The Reciprocity Relation between the Transfer Impedance
Coetlicient s, . . . . . . . . . . . . . . . . . . ...59
X111
xiv CO.YYfil.VTS

CHAP. 3. R.\ III.iTIOS FR031 CUIUWXT I) ISTRII)IT1OSS. (iI

31. Tllc Field Ilquations. 6]

32. The (;onstit,ltive Paramctrrs; I.inc:wity mid SIlperpo sition . 6,5
33. Ilou])dary Contlitions. 66
3 ~. The Field ~q~latio]ls for H:mnonic TIInc ])(, p[,lld(>I1[W 68
3.5. Iayntings Thmreln . . . . M)
36. The ll-a\,c k;qll:ltions. 71
3.7. Simple l~avc Sollltions. 73
38. General Sollttion of the Field I}q,,~tiol,s in Tcr],,s of tl)r fk),,rtcs,
for a Ti]llc-pcriodir lri(ld. 8(I
3.9, Field ]),,e to Sollrtt,s in an U]IhoIuI(ltYi Ii(,gion 84
3.10. Field in a licgion Rotmdcd hy S(lrfa((,s of I]lfillltcl}- (OI)(l\l(tIvr
llc,clia . . . . . . . . . . . . . . . . . . . . . ..8fi
311. Tl](, Far-zone Fields 87
312. Iolarization. !)1
3~13. The I;lcctric l)ipolc !)2
314. Tllc IItignetir lh]mlc 95
3.15. The F:lr-zonr Firl(ls of I,i]l(,-rurr(,nt l)istrilllltif!)ls !)(i
316. The H:df-~1 IV(, l)ilmlt, !)8
317. Sllpcrpm]tion of l~icl{ls !)!1
31S The 1)0111>1(,-(111)01(
System 101
319. I{cgldur Space .Irr:iys 104

C][.IP. 4. ll-.l;l; FI{O>-TS :~~1) RAYS 107

41. TIIC II\lygrns-Crccn Forn~lll:L for thr Ill((,tr(l]]]:Lg]l(ti[, litl(l 107
42, Gcol]lctrical ()~]tirs: l~avefronts and l{:lys 110
43. C1lr~:itllrr of the Ilmys in an Inllo]]Iogc,l](,[~!ls Ilcdiunl 111
4+. Energy FlOIv in (lcometrical Optics 112
45. (;comctrical optics :is :L Zero-~ra~clcngth I,in]it 114
46. The H~lygens-Frrsnrl Principle and Gconlctriral Optics: The Far-
zonc .kpproximation 116
47. The Principle of Stationm-y Ihasc 11{)
48. Ft=rnlats Irincip]e. 122
4.9, The I,a,v of the optiral Path 125

(]~ \l,, 5, S(.LTTERISC, .kN-T) DIFFR.ACTIOX. 12!)

51. (;cncral (onsidcrntions 129

52. Bol]ndary (o]lditions 130
53. Iieflection hy an Infillite Plane S(lrtarc; the lrinril)lv of ItII:igrs 132

APPROXIMATE JIETHODS FOR REFLECTOI+S OF ;lILBITMARY S] J.\PE ,.. 137

54, The Geometrical-optics IIcthorl 138
55, Calculation of the Scattrrrd Firl[l 139
56, Superposition of the So~lrcc l~icl(l :IIId tl)(, Sc:\ttercd l~icld. 143
57, The Current-distril, ~ltion 31cthod 144
58. Calc(llation of the Scattrrrd Field 146
59. Application to Point-source md IJirl{J-sollrre l;ecds. 149
510. Reaction of a Reflector on a Point-source Feed 155
511. The .Aperture-fielci Ifethod . . 158
512. The Fraunhofer Region. ,.. 160
 .

coYTAl
7s xv

l) IF1.ll.\(r
10N. . .. . . . ...,,162
S.IS, (i,r,crtLI (~)])si,l,,r:,iit)])s mI tht, .ipproxim:ltc Ilt,tllods 162
514. l{ttlll~,tltJJl to :1 S[alilr l)illr:~(tlf)n IIo I)lcIN 164
515. l)~~lli]lcts lrl]l{ipl~i for tll( l<;lc,~tro]ll:tg])f,ti(, Ir](,ld 167

(ti~l,. 6. .lIKI{TURF; ll.LLllll-ATIOX AX]) .l>-TI,;X-XA IATTERNS 169

61. lril]l:~ry and S[,co]ltlxry ltlttrms 169

62. Tll(, l)iffr:trtion Fieltl 169
(i 3. Iouri(r Integral li(,l)rc,s(,]ltati{jl~ of the Fraunhofcr lie~ion 174
64, (+CJI(M1 I:caturts of tht, Sccolld:wy l~ttvm 175
6,5. TIIC l{rctmlg~ll:~r .fporture ]80
6.ti. Tl!o-(lilllt,nsio]lal Prol)lcms 182
67. Il]:i.w-error kYfects. . 186
08. TIIC (irc(llm .ipcrturc 192
&9. Th(, Field o]] the Axis in thr Frcsnrl ]tcgion ] 96

(l[,\],, 7. l[I(:ROIV,.fV}; TRAA-S~f ISSIOA I,IN-ES 200

71 llicro~j :ivc nnd I,ong-~vave Trimsmission I,illm 200

72. lrop:l~atio]l in ~f:~vcgllidcs of lniform (;ross Swtion 201
73. orthogo]lallty Rcl:ltions and Power Flow. 207
74. Transnlissiun I,inr (onsidcrations in l~:lvrguidrs 209
75. XctJt ork Kqllivalents of Junrtions and ohstaclcs 214
76. 7/l.l/-modc Trallsmission I,irlcs 216
77. (ozxis,l I,incs: ?~.!f-rnodc 217
7.8. (oaxial I.ines: T.If - and T]i-nlodcs 219
7.!), (:Is,.acIc Tmnsformcrs: TJ~.lf-mode 221
71o. Iarallel St~lhs and Series ILeactancm. 223
711. licctang~llar }Vavcguidcs: IA- and ?,lf-modes 226
712. Impcdanrc Transformers for Iiectangular (;uidcs 229
713. Circular ll-aveguide: T~- and TJf-modcw. 233
7.14. Ivindows for LTSCin Circular Guides 235
715. Iarallel-plate i~aveguide. . 235
716.11esignN Totes . . . . . . . . . . . . . .. 238

CHAF. 8. lfICROWAVE DIPOLIl A3JTE~~AS ANI) F13f?DS 239

81. Characteristics of Antenna Feeds 239

8.2, Coaxial I,ine Terminations: The Skirt Dipole 240
83. Asymmetric Dipole Termination. 242
84. Symmetrically 13nergizcri Dipoles: Slot-fed Systems 245
85. Shape and Size of the Dipole . 248
86. lVaveguidc-line-fcd Dipoles. . 250
87. Directive Dipole Feeds, . . . 250
88. Dipole-disk Fords . . . . . . . . . . . . . . . . . . ...251
89, Double-dipole Feces, .,.... . . . . . . . . . . . . .253
8.10. Lfulti-dipole Systems, . . . 256

CHAP. 9. LINI?AR ARRAY AXTEiYNAS AND FF23DS 257

9.1. (kmcral Considerations. . 257

XVI CON TEJVTS

PA~EEN THEORY . . . . . . . . . . . . . . . . . . . . . . ...256

92. General Array Formula. . . 258

93. The Associated Polynomial 261
9.4. U1liformArrays . . . . . ...264
9.5, Broadside 13e:~nls . . . . ...267
9.6. Erl(l-tire I~ean]s . . . . . . . . . ...274
9.7. 13ca1u Synthesis . . . . . . . . . . . . . .. 279

RADIATING EI.EMMNTS . . . . . . . . . . . . . . . . . .. 284

9.8. llipole Radiators . . . . . . . . . . . . . . 284

9.9. Slots in Jvaveguide }Talk. . 286
9.10. Theory of Slot Radiators. . 287
9.11. Slots in Rectangular J$aveguide; 1~,,-mode 291
912. Experimental Data on Slot l{adi~tors 295
913. Probe-fedS lots . . . . . . . . . . . . . 299
9.14. fVaveguide Radiators 301
9.15. Axially Symmetrical Radiators . 303
9.16. Streamlined Radiators . . . 310

ARRAYS . . . . . . . . . . . . . . . . . . . . .. 312

9.17. Loaded-line Analysis. . 313

9.18. End-fire .4rray . . . . . . . . . . . . . . . . . . . . . . 316

BROADSIDE AREAYS. . . . . . . . . . . . . . . . . . . . . . . .. 318

9.19, Suppression of Extraneous Major I,ohcs . 318
920. ResonantArrays . . . . . . . . . . . . . . . .. 321
921. Beacon Antenna Systems. 327
922. T$onresonant Arrays . . . . . 328
923, Broadband Systems with >Tormal Beams 331

CHAP. 10. WAVEGUIDE AND HORN FE~;~S. . 334

10.1. Radiation from Waveguide of Arhitrmy Cross Scrtion 334
10.2. Radiation from Circular ~~av(guide 336
103. Radiation from Rectwwlar Guide. 341
10.4. Waveguide Antenna Feeds 347
105. The Double-slot Feed . 348
10.6. Electromagnetic Horns. . . . . . 349
10.7. hrodes in lplane Sectoral Horns 35o
108. Jfodes in If-plane Sectoral Horns 355
109. Vector Diffraction Theory Applied to Srctoral Horns. 357
10.10. Characteristics of Observed Itadiation ltittmns from Horns of
Rectangular Cross Section 358
10.11. Admittance of Waveguidc and Horns . 366
10.12. Transformation of the L-plaI~c HorII .idulittan cc f mm the Throat
tothe Uniform Guide . . . . . . . . . . . . . . . . ...369
10.13, Admittance Characteristics of H-plane Sectoral Horms 374
1014. CompoundHorns. .,.. , . . . . . . . . . . . . ...376
10.15 .The Box Horn... . . . . . . . . ...377
1016. Beam Sllaping hy ~lcans of Obstaclrs in HOHI :md \VaveK{lidv
.4pertmcs. . . . . . . . . . . 380
10.17, Prcssllrizing and hfatch]ng 383
CHAP. 11. DII:I,IcCTILIC i~X1) lIJ:T:lI,-lI.. fTI<: I. I<:XSES 388
11.1. Uses of l,[>lls(>sin 31icro~v~vc ,~ntcmms. 388

I)ll:l,l:C,rRICI
JEXSF;S . . . . . . . . . . . . . . . . . . . . . . . 389

112. lrill{ilJlrso fI>rsign, . 389

113, Sinlplc I,cllses IVltllo(it Zoning 390
114. Zoned l)lclrctric I,cnscs . . 395
11,5. Usc of lIatcri:tls ~v]th Hi~ll l{cfmctivc Indcxrs 398
11.6. I)lclcctric I,osscs mnd Tolcranccs on Irons Iaramctcrs. 399
11.7. Itcflections from L)iclcctric Surfaccw 401

hfETAL-I,L~TELE
XSES . . . . . . . . . . . . . . . . . . . . . ...402

11.8. Parallel-plate I,ensm, . . . . 402

11.!). Other 31eta-lcr,s Structures. 406
11.10. l[cta-plate I.cns Tolrrancrs 407
11.11. Band,vi(lth of l[etal-plztc I,cnscs; Achromatic Doublets 408
11.12. Iteflections from Surfaces of Parallel-plate I,enses 410

CHAP. 12. PENCILBFAhf AND SIhfPLE FANAl;D-BEAN1 ANTEKAAS 413

PENCIL-BEAM
ANTENNAS
. . . . . . . . .,.........,...413
12.1. Pencil-beam Requirements and Tcchniqlles 413
12.2. Gcomctriral Parameters 415
123. The Surface-current and Aperture-firldI)istrih(ltions. 417
12.4. The ILadiation Field of the Reflector 420
12.5. The Antenna Gain., . . . . . . . . . . . 423
126. Primary Pattern Designs for hI:mimizing (lain 433
127. Experimental Itcsldts on %condm-y lattmns 433
128. Impedance Characteristics 439
12.9. The Vertex-plate lf:~tching Trchniquc 443
12.10. Itotation of Iohu-ization Technique 447
12.11. Structural Design Problems. 448

SIMFLE FANNED-BEAM ANTENNAS. 45o

12.12. Applications of Fanned Beams and Nfcthods of Pmdllrtion 45o

1213. Symmetrically Cut Parahololds 451
12.14. Feed Offset and Contour Cutting of Reflectors 453
1215. The Parabolic Cylinder and Line Source 457
12.16. Parallel-plate Systems 459
12.17. Pdlbox Design Problems 460

CHAP. 13. SHAPED-BEAM ANTENNAS. . 465

131. Shaped-beam Applications and Requirements 465

132. Effect of a Directional Target Response 468
133. Survey of Beam-shaping Techniques. 471
134. Design of Extended Feeds. . . . 487
135. Cylindrical Reflector Antennas . 494
136. Reflector Design on the Basis of Ray Theory . 497
137. Radiation Pattern Analysis. 500
138. Double Curvature Reflector Antennas 502
13.9. Variable Beam Shape. 508
 ,..
XV1ll COA71{.V 7s

CHAP. 14. ANTk;XNA IXSTALLAT1ON IROBI,E31S 510

GE~~It.4L SIKVl~Y OF I~Scr.4LLArIO~ I~OBLEMS. 510

14.1. (:rolm(l .lllt(,llnas 510

14.2.SllilJ.illterlI]as,,, . . . . . . . . ... ..511
143, ~lir(,r:ift<intellnzs 512
14,4. Scanning Antennas on Aircraft 513
14.5. Beacon Antennas on Aircraft 521

522

523
528
537
540

543
543

544
544
545
547
550
552
556

557
557
,.561
564
570
572
573

574
,574
578
580
580
581
,582
,58!5
58A
587

593
593
5{)4
601
604
60!)

ISl)lI;X . . . . . . . . . . . . . 6 I.5
 .

C11.IPTI;R 1

SURVEY OF MICROWAVE ANTENNA DESIGN PROBLEMS

]]Y s. SII,V1;ll

1.1. The Wavelength Region.Ihe designation of the boundaries of

the micro!mve region of tIw rlcct romagmct ic spectrum is pllrcly arbitrary,
Tile Iong-]vavelcngth limit IIas Ixxm set v:~riously at 25 (Jr 40 cm, even
at 100 cm. From the point, of vie\v of antenna theory and design techn-
iques, the 25-cm val~le is the most appropriate choice, The short-
wa~elength limit to )ihich it is possible to extend the present terhniq(les
ll:~snot ~etl)ec>r~r
caclle(i; it isinthcnciglll)or}loo(lof lmm. Accordingly
\veshall cunsi(lcr the microlvave region to extend in wavelength from 0.1
to 25 cm, in frcqllcncy from 3 X 105to 1200 31c/see,
This is the transition region bet\\-eenthe or(linary radio region, in
}vhich the \\-avelengtllis very k~rge comparwl with the dimensions of all
the components of the system (cxccpt perhops for theh~rge and cumber-
some antennas), and the optical region, in ]t-hich the \vavclengths arc
excessil-ely small. I.ong-\vavc concepts rm(l techniques continue to be
useful in the micro \vave region, and at the same time certain devices
used in the optical regionsllr haslense sandn~irror sarcemployeci. From
the point of vie}v of the antenna designer the most important character-
istic of this fre(~ucncy region is that the wa~~elengths are of the order of
magnitude of the dimcmsilmsof conventional and easily handled mechan-
ical devices. This leads to radical modification of earlier antenna
techniques and to the appearance of nefv and striking possibilities,
especially in the construction and use of complex antenna structures.
It follows from elementary diffraction theory that if D is the maximum
dimension of an antenna in a given plane and k the ivavrlength of the
radiation, then the minimum angle }vithin which the radiation can be
concentrated in that plane is

(1)

With microwaves one can thus produce highly directive antennas such
as have no parallel in long-wave practice; if agivendirectivity is desired,
it can be obtained \vith a microwave antenna ]vhich is smaller than the
equivalent long-!vave antenna. The ease with which these small antennas
can be installed and manipulated inarestricted space contributes greatly
to the potential uses of microwaves. In addition, the convenient size of
1
2 SURVEY OF MICROIV. t JE ANTEXAVA DESIG.V PROBLEMS [SEC, 1.2

microwave antenna elements and of the complete antenna structure makes

it feasible to construct and use antennas of elaborate structure for special
purposes; in particular, it is possible to introduce mechanical motions of
parts of the antenna with respect to other parts, with consequent rapid
motion of the antenna beam.
The microwave region is a transition region also as regards theoretical
methods. The techniques required range from lumped-constant circuit
theory, on the low-frequency side, through transmission-line theory, field
theory, and diffraction theory to geometrical optics, on the high-fre-
quency side. There is frequent need for using several of these theories
in parallelcombining field theory and transmission-line theory, sup-
plementing geometrical optics by diffraction theory, and so on. Optical
problems in the microwave antenna field are relatively complex, and
some are of quite novel character: For instance, the optics of a curved
two-dimensional domain finds practical application in the design of
rapid-scanning antennas.
1.2. Antenna Patterns.-Before undertaking a survey of the more
important types of microwave antenna, it will be necessary to state
precisely the terms in which the performance of an antenna will be
described.
The Antenna as a Radiating Device: The Gain Function.The field
set up by any radiating system can be dirided into two components:
the induction field and the radiation field. The induction field is impor-
tant only in the immediate vicinity of the radiating system; the energy
associated with it pulsates back and forth between the radiator and
near-by space. At large distances the radiation field is dominant; it
represents a continual flow of energy directly outward from the radiator,
with a density that varies inversely with the sq~iarc of the distance and,
in general, depends on the direction from the source.
In evaluating the performance of an antenna as a radiating system
one considers only the field at a large distance, where the induction field
can be neglected. The antenna is then treated as an effective point
source, radiating power that, per unit solid angle, is a function of direc-
tion only. The directive properties of an antenna are most con~eniently
expressed in terms of the gain function G(6,O). I/et 6and @ be respec-
tively the colatitude and azimuth angles in a set of polar coordinates
centered at the antenna. Let F(O,@) be the power radiated per unit
solid angle in direction 0, @ and P~ the total power radiated. The gain
function is defined as the ratio of the power radiated in a given direction
per unit solid angle to the average power radia~ed per unit solid angle:

(2)

47r
SEC. 1.2] ANTENNA PATTERNS 3

Thus G(L9,~) expresses the increase in power radiated in a given direction

by the antenna over that from an isotropic radiator emitting the same
total power; it is independent of the actual power level. The gain
function is conveniently visualized as the surface

r = G(f3,@) (3)

distant from origin in each direction by an amount equal to the gain

function for that direction. Typical gain-function surfaces for micro-
wave antennas are illustrated in Fig. 1.1.
The maximum value of the gain function is called the gain; it
will be denoted by GM. The gain of an antenna is the greatest factor
by which the power transmitted in a given direction can be increased
by using that antenna instead of an isotropic radiator.
The transmitting pattern of an antenna is the surface

(4)

it is thus the gain-function surface normalized to unit maximum radius.

A cross section of this surface in any plane that includes the origin is
called the polar diagram of the antenna in this plane. The polar
diagram is sometimes renormalized to unit maximum radius.
W-hen the pattern of an antenna has a single principal lobe, this is
usually referred to as the antenna beam. This beam may have a
wide variety of forms, as is shown in Fig. 1.1.
The Antenna as a Receiving Dwice: The Receiving Cross Section .The
performance of an antenna as a receiving device can be described in
terms of a receiving cross section or receiring pattern.
A receiving antenna will pick up energy from an incident plane wave
and will feed it into a transmission line which terminates in an absorbing
load, the detector. The amount of energy absorbed in the load will
depend on the orientation of the antenna, the polarization of the wave,
and the impedance match in the receiving system. In specifying the
performance of the antenna, we shall suppose that the polarization of
the wave and the impedance characteristics of the detector are such that
maximum power is absorbed. The absorbed power can then be expressed
as the power incident on an effecti~-c absorbing area, called the receiving
cross section, or absorption cross section A, of the antenna. If S is
the power flux density in the incident wave, the absorbed power is

P, = ASA, (5)

The receiving cross section will depend on the direction in which the
plane wave is incident on the antenna. We shall write it as A, = A,(d,I$),
where o and @ are the spherical angles, already defined, of the direction
4 SLRJE1 OF JIIClK)IV.4 JE .4.V7E.\.VA DI<,SIG.I 11{01$1.1<.11.V [SW. 12

of incidence of the lvave, This function, like the gain function, is repre-
sented conveniently as the surface
? = .4, (0,0). (6 J
The receiving pattern of an antenna is drfincd, :malogolls]y t(
the transmitting pattern, as the above surface normalized to unit maxi-
mum radius:
(7)

It is a consequence of the reciprocity theorem to be discllssed in

Chap. 2 that the receiving and transmitting patterns of an antenna are
identical:
~(~,o) = ~r(g:y),
(8)
G.,, A,.,,
It will also be shown that the ratio .4, u0 v is a constant for all matched
antennas:
.-lr,f ~ ~,
(9)
G 41r
Thus for any matched receiving system
,
A,((l, @l) = :; G(e,l+). (lo)

Coverage Pattern, One Way.-The characteristics of an antenna may

also be described in terms of the performance of a radio or radar system
of which it is a part. It is necessary to distinguish between the case of
one-way transmission, in which a given antenna serves for transmission
or for reception only, and the case of radar or two-way transmission, in
which a single antenna performs both functions.
We consider first a transmitting antenna and a receiving antenna
separated by a large distance R. Let G, and G, be the respective gain
functions of the two antennas for the direction of transmission. If the
total power transmitted is P, the power radiated in the direction of the
receiver, per unit solid angle, will be (1/4m)PG~. The receiving antenna
will present a receiving cross section (1/47r)G,x2 to the incident wave; it
will, in effect, subtend a solid angle G,A2/47rRzat the transmitter. The
power absorbed at the receiver will thus be

(11)

The maximum operating range is determined by the signal-to-noise

ratio of the detector system. If P,m is the minimum detectable signal
for the receiver, the maximum operating range is

P$i A
R .,., = ~; (G,G,) ~ (12:
P
(-) ,m
 SEC. 12] ANTENNA PATTERNS 5

Thus, if it is possible to ignore the effect of the earth on the propagation

of the wave and if G, is constant, it will be possible to operate the receiving
system satisfactorily everywhere within the surface

(13)

where the transmitter is taken to be at the origin. This surface will be

called the free-space coverage pattern for one-way transmission.
Coverage Pattern, Two Ways. - -In most radar applications the same
antenna is used for transmission and reception. One is here interested
in detecting a target, which may be characterized by its ( scattering
cross section u. This is the actual cross section of a sphere that in the
same position as the target would scatter back to the receiver the same
amount of energy as is returned by the target. For this fictitious iso-
tropic scatterer, the effective angle subtended at the transmitter is U/R2
and the total power intercepted is

(14)

Scattered isotropically, this power would appear back at the transmitter

as a power flux, per unit area,

(15)

Actually, the scattering of most targets is not uniform. The scattering

cross section of the target will in any case-be defined by Eq. (15), but it
will usually be a function of the orientation of the target.
The power absorbed b:- the receiver from the scattered wave will be

P,= A+S=R (16)

since here G, = G,. If the effect of the earth cm transmission of the

waves can be neglected, it will be possible to detect the target only when
it lies within the surface

(17)

about the transmitter as an origin. This surface will be called the free-
space coverage pattern for twe-way transmissi,m.
The extent of the coverage patterns is determined by characteristics
of the system and targetoutput power, receiver sensitivity, target size
that are not under the control of the antenna designer. The form of
the coverage patterns is determined by but is not the same as the form
of the antenna transmitting a,nd receiving patterns; in the coverage
patterns, r is proportional to [G,(o, r#J)]Jfi
rather than to G,(o, +). The
6 SURVEY OF MICROWA FE AIV7EiVA44 DESIG.V PROBLEWS [SEC. 13

desired form of the coverage pattern is largely determined by the use to

be made of the system. From it, one can derive the required form of the
transmitting or receiving pattern of the antenna; it is usually in terms of
this type of pattern that antenna performance is measured and specified.
It is to be emphasized that the discussion of coverage patterns gi~en

(b)

(c) (d)
FIG.I.I.Typicalgain-functionsurfacesformicrowaveantennas. (a) Toroidal(omni-
directional)pattern;(b) pencil-beampattern;(c) flat-topflaredbeam; (d) asymmetrically
flaredbeam.

here assumes free-space conditions. In many important applications,

coverage is affected by interference and diffraction phenomena due to
the earth, by meteorological conditions, and by other factors. A detailed
account of these factors, which may be of considerable importance in
determining the antenna transmitting pattern required tora given appli-
cation, will be found in Vol. 13 of the Radiation I,aboratory Series.
103. Types of Microwave Beams.The most important types of
microwave beams are illustrated in Fig. 1.1.
The least directive beam is the toroidal beam, 1 which is uniform in
1Such a beam is also referredto as omnidirectional. (IRE Standards and
Definitions,1946.)
SEC. 1.4] MICRO WAVE TRANSMISSION LINES 7

azimuth but directive in elevation. Such a beam is desirable as a marker

for an airfield because it can be detected from all directions.
The most directive type of antenna gives a pencil beam, in which
the major portion of the energy is confined to a small cone of nearly
circular cross section. With the high directivity of this beam goes a
very high gain, often as great as 1000. In radar applications such a
beam may be used like a searchlight beam in determining the angular
position of a target.
Although the pencil beam is useful for precise determination of radar
target positions, it is difficult to use in locating random targets. For
the latter purpose it is better to use a fanned beam, which extends
through a greater angle in one plane than it does in a plane perpendicular
to that plane. The greater part of the energy is then directed into a cone
of roughly elliptical cross section, with the long axis, for example, ver-
tical. By sweeping this beam in azimuth, one can scan the sky more
rapidly than with a pencil beam, decreasing the time during which a
target may go undetected. Such a fanned beam still permits precise
location of targets in azimuth, at the expense of loss of information
concerning target elevation.
Other applications of microwave beams require the use of beams with
carefully shaped polar diagrams. These include one-sided flares, such
as is illustrated in Fig. 1Id, in which the polar diagram in the flare
plane is roughly an obtuse triangle, whereas in transverse planes the beam
remains narrow. In radar use, such a beam at the same time permits
precise location of targets in azimuth and assures most effective distribu-
tion of radiation within the vertical plane of the beam. Toroidal beams
with a one-sided flare in elevation have also been developed.
No theoretical factors limit any of the above beam types to the micro-
wave region, but many practical limitations are imposed on long-wave
antennas by the necessary relationship between the dimensions of the
antenna elements and the wavelengths.
104. Microwave Transmission Lines.-The form of microwave
antennas depends upon the nature of the available radiating elements,
and this in turn depends upon the nature of the transmission lines that
feed energy to these elements. We therefore preface a survey of the
main types of microwave antennas with a brief description of microwave
transmission lines; a detailed discussion of these lines will be found in
Chap. 7.
Unshielded parallel-wire transmission lines are not suitable for micro-
wave use; if they are not to radiate excessive y, the spacing of the wires
must be so small that the power-carrying capacity of the line is severely
limited.
Use of the self-shielding coaxial line is possible in the microwa~ t~
8 S( I<il{~- 01 .ifl(:l{() WA }1< .1 i< 7liA.YA DKSIG.V PIIOI{LE.IIS [SE<. 1,5

region but is generally restricted to lfa~-elengths of approximately 10 cm

or more. IJor proper action as a transmission line, a coaxial line slLoulcf
transmit electromagnetic Ir:lves in only
a single mode; other\\ise the generator
100!{s into an indetermirmte impedance
and tends to be erratic in operation.

7 on this account it is necessary to keep

the at-erage circumference of inner and
outer condllctors less than the frce-
space wavelength of the transmitted
~ravcs, .\t ~vavelengths shorter than
10 cm this limitation on the dimensions
of c~mxial lines begins to limit their
(b) polvcr-carrying capacity to a (Ir=gree
that m~kes them lmsatisfactory for
most purposes.
The most ~lseful transmission line
in the miclotvave region is the hollo\\-
pipe. Sllrll pipes \vill sllpport the
propagatiorr of :lrleiect,rom:~g~lrti(,!j-:~~e
(c) only it-hen they are sufficiently large
comp:we(l \\ith its free-space \vave-
length. As g~lides for long-lrave
radi:Ltl(Jn, ]nt,oleral)ly large pipes are
reql[ir(,(l, I)llt in the microlrave region
it lxw)mes pf)ssit)le to mse pipesof rmn-
vcnlcnt, SIZC. I,ike the coaxi:d gllide,
there is :Llsfjan llpper limit imp(w(lon
the crow-sectional dimension of the pipe
if it, is to tr:msmit the \v:ive in only N
single mo(le. II()\\-e\-er,in theal)scnce
of :ln inner con(lllctor, this size limit:L-
tion (l(wnot :Iflect the Ix)li-cr r:llxwily
so seril}llsly :Wit does in tllc c(mi:~l line.
1.5. Radiating Elements.T h e
natllre of t hc ra(li:~ting elements
trrmin:~ting :L transmission line is to
:L (l) USi(l(>I:Ll)l( (Xt Cllt (1(t(llllill((1 })~
tile n:~tllre of the li]le itwlt. 1y])ieal
l~,ng-lv:~le r:l(li:~tillg clenwnt + :Irr the
{lil)ole :lrl((~]l]l:ls, sll~,ll :1s tllc (,t~lltcr-
(Iril.en i]:llf-\\:l\-e (Iil)tlle, nll(l loop
 coaxial lines lend themselves to sllch terminations. Many long-wave
antenna ideas have lwen rarr-ied uver into the micro\rave region, par-
tic~~hwlythose connected with thehalf-]rave dipde; the tramsitiorr, ho\v-
ever, is riot rnereiy a mattrr of wovelcmgth scaling. In a microl}ave
antenna tl~e cross-sertional dimensions of the transmission line are com-
partihlc to the dimensions of the half-~vavc dipole, and consequently, the
coupling lmtween the radiator and tile line becomes a more significant
prol)lem tlian in a corresp(jnclin~ Iong-ivave system. The cross-sectional
dimensions of the dipole element are dso comparable to its length. A
typi~al microwave dipole is shown in Fig. 12c; the analysis and undt=r-
stancling of S1lC}Imicro}vave dipoles is at best still in a qualitative stage.
The ose of hollow ~vaveyuide lines leads to the employment of entirely
(Lffc,rent radiating systems. The simplest radiating termination for such
a line is j~lst the open end of the g~lirle, through which the energy passes
into space. The dimensions of the mouth aperture are then comparable
to the wavelength; as a result of diffraction, the energy does not continue
in a lwam corresponding to the cross section of the pipe but spreads out
considerably about, the direction of propagation defined by the guide.
The degree of spreading depends on the ratio of aperture dimensions to
wa~ekmgth. On flaring or constricting the terminal region of the guide
in order to control the directivity of the radiated energy, one arrives at
electromagnetic horns based on the same fundamental principles as
acoustic horns (Fig. 1.20!).
Another type of element that appears in microwave antennas is the
radiating slot (Fig. 1.2r). There is a distribution of current over the
inside wall of a waveguide associated with the wave that is propagated
in the interior. If a slot is milled in the wall of the guide so as to cut
across the lines of current flow, the interior of the guide is coupled to
space and energy is radiated through the slot. (If the slot is milled along
the line of current flow, the space coupling and radiation are negligible. )
I slot will radiate most effectively if it is resonant at the frequency in
question. The long dimension of a resonant slot is nearly a half \\-ave-
iength, and the transverse dimension a small fraction of this; the perim-
eter rJtthe slot is thus closely a wavelength.
1.6. A Survey of Microwave Antenna Types.We are now in a posi-
tion to mention briefly the principal types of antennas to be considered
in this book.
Antennas jo~ Toroidal Beams.A toroidal beam may be produced
by an isolated half-wave antenna. This is a useful antenna over a large
frequency range, the iimit being set by the mechanical problems of sup-
porting the antenna and achieving the required isolation. The beam
thus produced, however, is too broad in elevation for many purposes.
A simple system that maintains azimuthal symmetry but permits
control of directivity in elevation is the biconical horn, illustrated in
10 ASURVEY OF MICROW.41E .4,17fl,V.VAI)EL7[G.N-
PRO13LJY.tf,9 [Sm. 16

Fig, 13. The primary driving element between the apexes of the coues
is a stub fed from a coaxial line. The spread of the energy is determined
by the flare angle and the ratio of mouth dimension to wavelength.
Although this antenna is useful ov~r a
large freq~lency range, maximum di-
rectivity for given antenna ~veight and
size is obtairmble in the microwave
region, where the largest ratio of
aperture to wavelength can be
realized.
Increased directivity in a toroidal
beam can also be obtained with an
array of radiating elements such as
dipoles, dots, or bimnical horns built
up along the symmetry axis of the
beam. The directivity of the array is
determined by its length measured in
~vavelengtbs; high directivities arc
conveniently obtained by this method only in the microlvave region. .1
typical microwave array of this type is shoum in Fig. 1.4.
Pt,ncil-brum A nfrnnas.-Bearr~s that hare direr tivit y both in eleva-
tion and azimuth may be pr(xlllccd by a pair of dipole elements or by a
dipole with a reflecting plate. The major portion of the energy is con-
tained in a cone ~rith apex angle somewhat less than 180.

FIG. 14 -.4. mirmwa~c lmaron array.

Similar beams arc prodllced by horn antennas that permit control
of the directivity throllgh choice of the flare tingle and the n~{)lltl] dimen-
sions. Horns are useful at lo\ver frequencies as JVC1las in the rnicrolrave
region; indeed, the early work on horns Ivas done for \~ti\-elengthsranging
from 50 to 100 cm.
More directive healns-trlic pencil bemns-can be prtd~lced b,v
building up space aryays of the almw systems. T\\-,)-tiimensiorlalarrays
(mattress arrays) an{i mldt,i,mit horn systems arc IISC(Iat l,,,-er frequen-
cies. Their dircctivity is severely limited, ho\\-ever,hy tl~e ]nrtll:mical
problems occasioned by the rcc(llired ratio of (Iimrnsions to }f :L, t,-
Iengths. Such arrays have not been employe(l in tlie micro~~-averegif)n.
$.Ec. 1.6] A SURVEY OF MICROWAVE ANTENNA TYPES 11

At these wavelengths it becomes feasible, and indeed very convenient,

to replace the two-dimensional array technique by the use of reflectors
and lenses.

(a)

(b)
FIG. 1.5.Pencil-beam antennas. (a) ParaboloidaI mirror; (b) metal-plate lens. (Metcd-
plate lens photo~aph courtes~ of the Bell Telephone Labordorie8.)

Highly directive pencil beams are produced by placing a partially

directive system such as the double-dipole unit, dipole-reflector unit, or
horn at the focus of a paraboloidal reflector or x ccntrosymrnetric lens,.
The use t)f these devices is based (JII the r{)ncrI>ts of ray optics, a(cor(linx
to lvhich thr reflector or Ims takes the dilcrgrt]l ra~s fr<)m tlm pt~int
source at the ftwl[s and converts tlwrn into :L beam of par:dlcl rays.
Despite the diffraction r[fects which limit thr npplicati{)rr of ray optics
and are very important in the micro \~ave regi(m, it is pr:w(i{,nllle to
make the apert[~rm so large that extremely sh:lrp lxmms can lx pr(xlllcrd.
Conversely, it is possihlc to ol~tain gwd (lirecti~-it.v Ii-ith :m antcnn:l s()
snmll that :~ircroft installations are prLwtiral. Paralj(jl(Jill:Gl:m(l Imra -
bolic reflectors arc Imrd at lower frrilllencies in s,,rne spcri:ll r:Lsrs, }),lf
in the rcqllircd Iargc sizes they tend to be less S:LtlSfU; t[)I~tll:Ln nl:Lttrrss
arrays,
Plastic lenses arc used in the mirro\va~c rrgion in prrcisely the same
\vay as g]:Lss lrnses in the optical region. In ad(lition, a neir device,
the metal lens, has been de~-elopxl for micro!vaves. Lhe fwl~-elcngtl~
of an clertromagnctic ~va~-ein an air-tilled )~avcg~lide is grratrr than that,
in free space; from the optical point of vic)v the ]raveg~lide is a region
of index of rcfractiorr less t}lan unity. A stack of ]!-aregtlidcs thins c~,n-
stitutes a refractive medium analogous to dielectric material, from lvhich
a metal lens can be fashioned, Figlu-e 1.5 shoivs micro~wi~~e prmcil-
}]eam an{ cnnas employing, respectively, a paralmloidal mirror and a
metal lens as directive devicm.
.4ntennas ,for Flared i3cams.Simple flared beams and one-sided
flares arc Iikc\vise prod~lced by means of reflectors and lenses and by
arrays of dipole-reflector units or radiating slots. S\lch arrays by t}lem-
selves give beams that are highly directive in planes containing the array
axis but are fairly broad in the transverse plane. In order to gain greater
directivity in the transverse plane the array may be used as a line so~u-ce
along the focal line of a parabolic cylindrical reflector; this focuses radia-
tion from a line source in the same \vay that a reflector in the form of a
paraboloid of revol~ltion focuses radiation from a point source. By
suitable shaping of the cross section of the cylinder, one can produce
beams with carefully controlled one-sided flares and other useful special
characteristics. Typical rnicrotrave antennas of this type are sholvn in
Fig. 1+.
Except for a few types of linear array, all micro]vave antennas use
primary sources of radiation together \vith reflectors and lenses. The
radiating element, wiich extracts po\ver directly from the transmission
line, is spoken of as the primary feed, the antenna feed, or simply
the feed; its radiation pattern as an isol~ted unit is kno]vn as the
primary pat tern of the antenna. In combination with the optical
elements of the ~ntenna, the feccf produces the o~-er-all pattern , { thn
antenna, often referred to as the secondary pattern of the antenna.
SEC, 1.7] IMPEDANCE 8PECIFICA TIONS 13

One of our major problems will be to establish the relationships among the
primary pattern of the antenna feed, the properties of the optical ele-
ments, and the secondary pattern,

(a)

(b)
Fm. 1.6.Antennas for pmaucmg ifared beams. (a) Simple flared-beam antenna; (b)
one-sided flared-ham system.

1.7. Impedance Specifications.-The achievement of a satisfactory

antenna pattern is by no means the only problem to be considered by the
antenna designer. It is important that the antenna pick up maximum
power from an incident wave and that it radiate the power delivered to
it by a transmission line without reflecting an appreciable portion of it
back into the transmitter. In other words, it is important that the
antenna have satisfactory impedance characteristics.
 The impedance problem in micro~vave antenna design takes on a
some~vhat special character because of the characteristics of other ele-
ments of the system, particularly the transmitting tubes. Conventional
triode-tube oscillators are not generally useful in the microwave region.
This is due to inherent limitations in the tube itself and to the fact that,
elements in the tank circuit no longer behave like lumped impedances.
The self-resonant freq~lency of the ordinary tube is considerably below
the microwave range, and it is therefore impossible to design a practical
circuit, that will oscillate at the required high fre<lucmcy. A modified
triode has been designed for use down to 10 cm. It has limited polver
capacity and is used ~vhere 101vpo\ver is acceptahlc. More ~enerally,
magnetrons and klystrons are used, the former for very hi~h poller levels.
The operating characteristics of these tubes are ~ery sensitive to the
impedance into which they are required to operate, thr freqlteucy ~arying
rapidly with changes in this impedance. JItJre serio[w than this fre-
quency pulling is the fact that the mmgnetron \vill cease to oscillale
withmlt too much provocation. (loser tolerances are, thereforr, imposed
on the impedance of a microwave antenna than those lvhich w~ld{l be
dictated by polver considerations. Ntany tllbfw ran be tuned over a fre-
ql]ency band, but at any frequency setting they must operate into the
proper impedance. Th(ls it is cllstornary to specify that a mimo,ravc
antenna be satisfactorily matched to the transmission line within close
tolerances, not simply at an intended operating freqllency, b~lt over a
band of frequencies.
In rapid-scanning antennas the impedance prt)l)lem is even more
complex. The arrangement of the mechanical parts varies d~lring :L
scan; it is necessary to make sllre th:~t the impedanrc properties of tllr
antenna remain satisfactory in all p,arts of thr scan, as Ivell as for a gi~.en
range of wavelengths. This element of the problem has an important
bearing on the choice of scheme+ for rapi(l-sr:uming antennas.
Throughout this volume the imped:mm ch:umrteristics of antennas
will be considered in parallel with their radiation pat terns.
1.8. Program of the Present Volume .This I)(mk falls into f(mu main
divisions: basic theory, theory i~nd dcsi~n of fewls, tl~rory and drsign of
complete antenna systems, and :~lltenn:L-m(,:~sLlling tc(,llni(llles and
equipment.
The fol]o!ving chapter summarizes certain parts of ct)nventi<}nd cir-
cuit theory that are pertinent to antenna prol)lcrns. In partimllar, it is
shown that the antenna drsigncr need make nu distinction bet\\ccn trans-
mitting :Lnd receiving antennas. Chfipter 3 states the basic principles
of field theor.v and applies them to the disc~lwion of mlrrcnt distrilllltions
as sources of radiation fields. Clmpters 4 to 6 then disc~lss clrrtronmg-
netic waves }rithout regard to t}lcir sollrces. CluLptcr 4 gives a brief
treatment of wavefronts and rays. Chapter 5 deals with the interaction
between electromagnetic ~vaves and obstacles; the general theory of
reflectors is here developed as a boundary-condition problem, and a
discussion is given of the relation bet~veen this theory and conventional
diffraction theory, which also finds application to microwave antenna
problems. Finally, Chap. 6 applies this theory in treating one of the
fundamental problems of antenna designthe relation between the field
distribution over the aperture of an antenna (such as a lens or reflector)
and its secondary pattern.
Chapter 7, on microwave transmission lines, serves as introduction
to the chapters on antenna feeds: dipole feeds, linear arrays, and horns.
Of these types all but the first have found applications also as complete
antennas; these applications will be indicated in these chapters.
A chapter on lenses precedes the treatment of more complex antenna
systems which is organized according to the type of beam to be produced:
pencil beams, simple fanned beams, and more complexly shaped beams.
When an antenna is installed on ground or a ship or airplanegenerally,
enclosed in a housingits performance is modified from that in free
space by its enclosure and neighboring objects. The subject of antenna-
iustallation problems is discussed briefly to acquaint the engineer with
the phenomena that may be expected to occur and some of the currently
known solutions of the problems.
The concluding chapters provide a statement of the basic techniques
of antenna measurements and a description of certain types of measur-
ing equipment that have given satisfactory service in the Radiation
Laboratory.
 CHAPTER 2

CIRCUIT RELATIONS, RECIPROCITY THEOREMS

BY s. SII.VER
2.1. Introduction. hchc circuit theory considerations and techniques
characteristic of low-f req~lency radio vwrk do not carry over in a simple
manner to the microlrave region. Thus, for example, in treating a cir-
cuit element as a lumped impedance, it is assumed that the current
(and voltage) at any given instant has the same value at every point in
the clement. This assumption is valid if the dimensions of the circuit
element are small compared with the wavelength, with the result that
the phase differences between separated points in the element are negligi-
ble. If, ho~vever, the wavelength becomes comparable to the dimensions
of the element, these phase differences become significant; at a given
instant the current at one point in the element may be passing through
its maximum value, ]vhile at another point it is zero. In such cases the
circuit element must be regarded as a system of distributed impedances.
The extension of conventional circuit theory to microl$-ave systems
is further complicated by the use of circuit elements such as waveguides,
in which voltages and currents are not uniquely defined. The analysis
of these elements must be approached from the point of view that they
serve to g~lide electromagnetic \vaves; attention is centered on electric
and magnetic fields rather than on voltage and current. The final result
of the field theory analysis is that under s~litable conditions~~hich are
generally encountered in practicea lvaveguide can be set into equiva-
lence with a two-wire transmission line in ~vhich the fundamental quan-
tities are voltage and current. The latter are directly related to the
waveguides electric and magnetic fields, respecti~-ely. 1 By means of
this equivalence the concepts of impedance, impedance matching, and
loaded lines are carkied over to ~vavegllides.
A waveguide can itself be treated as a system of distributed imped-
ances. i)istribut,ed impedances are treated in the same ~vay as lumped
impedances, by use of Kirchhoffs current and voltage lalm for networks.
A system of distributed impedance can, in fact, be replaced by a netlv(jrk
of lumped-impedance elements. The latter differ from the conventional
radio-circuit elements in that their impedance is a transcendental func-

I The subject is trmted in ( hap. 7. A fl[ll trratmmt of the ext[~lwi[,]]

of riwIIIt
thwry to w:lv(gui(lc+\ri11LL!foun{i i,] }-:)1 8 of this .wui{s,
16

I
tion of freqlleucy rather than an algeljraie funrti{)n. By means of tllesc
equivalent lllrrlpr{i-elerrlellt net \\orks, the net\\(Jrfi theorems that are
applicable tu lolr-frefl~lenry l~llll])ecl-elelll[tlt netlvorfis are carlied over
to systems \\ith distril)uted impedance 1he frost ptirt of this chapter
\vill review several nct}i-ork theorems LLnd t}le t]vo-}rir.e tr:Lr]sr]lissi(jI1-liIle
theory that are Ilsed in micrtjj! :~vc circuit theory. 1he s(lljjects \\illbe
treated briefly, the reader I)eing referred tt) stand:lrd texts for more
complete discussions :~n(l proofs of tile results qlloted hem.
The relation Ix%fveeu a transmitting and u mceiring antenna also
can be expresse(l in terms of an eflllivtilcnt netll ork. In this \ray one
can arrive at a reciprocity theorem JIhich rel:~tes the transmission char-
acteristics of an antenna to its receiling cll:tr~lcttlisti(s. of particular
importance to antenna([esigrl is tile fact, proved I)y IIscef the reciprocity
theorem, that the transmitting pattern of an antenna is the same as its
receiving pattern. The reciprocity theorem \vill be discussed in the
Iatter part of this chapter.
2.2. The Four-terminal Network. I.et usconsider an arl)itrary net-
nwrk, free from generators, made 11P of linear bilateral elements. A
linear bilateral element is one for
\~hich the relation bet~veen ~oltage ~. ~1 i2
Oc
and current is linear:
V = IZ, (1) ~ j a2
OD
1
where the value of the impedance Z FIO.21.-IJour-tcr]rlird network.
is independent of the direction of the
voltage drop across the element.3 For convenience the net~~ork \vill be
pictured as enclosed in a box and presenting to the outside only a pair
of input and a pair of output terminals. This is illustrated schematically
in Fig. 2.1. A boxed net]~ork of this type is referred to as a four-terminal
or two-terminal-pair network.
The network as a unit involves four quantities: the current i,, the
voltage drop VI from :4 to n, the mlrrent L, and the voltage drop Vz
from C to D. In conse(luence of the linear property [F;q. (l)] of each
component element of the net\vork, the relations between the, voltages
Vl, Vz and the currents il, i, are linear:

VI = Z1lil Z12i2j
(2)
V2 = Z21i1 Z22i2. }

1W. L. Everitt, Communicdfon Enginperirq,llfcCr:Lw-Hill,New York, 1937;

E. A. Guillcmin,CowLmunicaIion .Ve~mrks,Vols. 1, 11,l~ilry, Xm- York, 1931;T. E.
Shea, Transmission Vdw;orks md IT-W. Filler!, Y:in Nostrand,New lork, 1!129,
zSee Chap. 1 for the definition.sof thesepatterns
3It is assumedthat we are dealingwith a singlefrequency,that both the voltage
and currentdependon time throughthe samefactor e~u~.
18 CIRCUIT RELA 7IONJS RECIPROCITY 7HhOEEMS [SE(. 22

The impedance coefficient Z,, is the input impedance at All when CD

is open-circuited (zZ = O); similarly ZZZ is the input impedance at CD
when Al? is open-circuited. The quantities Z,, and Z21 arc known as
the transfer impedance coefficients of the network. As a result of the
bilateral property of the component elements of the netwwrk, the transfer
impedance coefficients satisfy the reciprocity relationl
Z,2 = Z21, (3)

As an al~ernative to the relations expressed by 13q. (2), the currents

may be expressed as linear functions of the voltage:

iI = Y1lV1 F12V2,
(4)
iz = Y21V1 Y2212. }

The admittance coefficient Y,, is the input admittance at All when the
terminals CD are short-circuited; Yz2 is the admittance at CD \vhen A B
is short-circuited; and Ylj, YZ1 are the transfer admittance coefficients.
The latter coefficients satisfy a reciprocity relation

Y,2 = Y2, (5)

in the case of bilateral elements. The impedance and admittance coeffi-
cients of the netwmi-k are related:

Y,, = y; y,2 = :;1; yz, = ylz = ~, (6)

where
A = Z,IZ22 Z,ZZZ,. (7)
By virtue of the reciprocityy relations, [Eqs. (3) and (5)], the network
has only three independent parameters, Consequently it can be replaced
by a network of three lumped-im-
pedance elements arranged in the
form of either a T- or ~-section as
shown in Fig. 22. The imped-
TYA:DZ ance elements of the T-section are
B T-sect,on DB r- sect,..
D designated by Z,, Z2, Z~. In the
I]o. 2.2.h-and m-sectionequivalentsof :L case nf the H-section it is more con-
four-terminalnetwork.
venient to (Ise admittances; the
elements are designated by YA = l/ZA, YE = l/Z3, Yc = l/ZC. The
relations between the elements of the reduced networks and the coefficients
of Eqs. (2) and (4) are
a. T-section:
ZI == Zll Z,2,
Z2 = Zn z,,, (8)
Z3 = Z12, 1
1~. ,1. Guillerllill,op. cd., Vols. I, II, Wiley, New York, 1331,particularly
Vol. 1, ~haD. IV.
SEC.2.3] THE RA YLEICH I{ ECIIIKWITI THJ?OREibl 19

b. II-section:
Y,4 = YI1 Y12,
Yc = Y22 Ylz, (9)
Y, = Y12. /
The relations between the T- and II-section elements for one and the
same four-terminal network are gi~en by

(10)

where the quantity A is that defined in Eq. (7).

The network can also be characterized by any three of the following
measurable quantities: the input impedance at A B \vhen CD is short-
circuited, the input impedance at A B \vhen CD is open-circuited, the
input impedances at C~ when AB
is open-circuited or short-circuited,
The relations between these quan-
tities and the impedance coeffi-
cients or the I- and II-section
elements can easily be derived from
Eqs. (2) and (8) or (9); they are (a)
given explicitly by Everitt. 1
2.3. The Rayleigh Reciprocity

a32
Theorem.The reciprocity relation
between the transfer impedance co-
efficients given in Eq. (3) is funda-
mental to the various reciprocity
theorems pertaining to net]vorks.
Ml of these theorems are variants (b)
of the general theorem derived by l:l<;.2:{. - l{wipt ovity ttleow!vt
forthefour-
terrmnal netwul,k.
Rayleigh. The particular form of
the theorem as it applies to a four-terminal net\vork will be discussed here.
In Fig. 2.3, i, and it are the currents in the network terminals when a
generator of emf V. is applied to the terminals AB through an impedance
Z, to feed a load Z. across the terminals CD; i{ and i; are the correspond-
ing currents at the terminals \vhena generator of emf VAis applied to the
terminals CD through an impedance Z. to feed a load Z. across A B.
The generator in each case is assumed to have zero internal impedance.
The reciprocity theorem states that

~Gi; = ~,j~. (11)

1 }V. 1,. F;vcritt, op. cit., (h~p. 11,

2ltayleigh,Thwryof ,Sou72d, Vol. 1, ,Sws, 10.5111,}Iacmillan, Yew York, reprinted
by I)ovcr Publications,.K{,IvYork, 1945.
20 CIR~l 17 RELA T1OV,S, RI?(lPR()([71 TIIl<ORk;\l,T [k. 2.4

Using Eqs. (2), we find for Case a of Fig. 23

Z,, T7,,
22 = (z,, + Z,)(z,, + z,,) Z,2Z,,

For Case b, remembering that the role of input and outpllt terminals
must be interchanged in Eqs. (2), we have
Z,2V;

( = (z,, + Z.)(z,, + z,,) Zl,z,,
Jfultiplying the first of these by 17:,and the second by ~o, one finds tlmt
the reciprocity theorem in I;q, (1 1) }Iolds pr~)vide(l that Z,* = Z,,.
Conversely, if a four-terminal netwmrk is linear in the sense of l?q. (2 ~
_2
z,, - ZL2 Z22-Z,2

mzL-t!E3zL
B D D
FIG.2.4.Th&enins theorem and the maximum-power transfer condition.

and if the reciprocity theorem [Eq. (1 1)] holds for the network, then the
transfer impedance coefficients satisfy the reciprocity relation of Eq. (3).
2.4. Th6venins Theorem and the Maximum-power Theorem.Con-
sider a network made up of linear bilateral elements and containing a
system of generators. Th6venins theorem states that the current
through any impedance Z. across a pair of terminals C, D of the network
is the same as the current in an impedance Z~ connected across a generator
whose emf is the open-circuit voltage across CD (the voltage Ivith Z1.
removed) and whose internal impedance is tbe input impedance meas-
ured at CD looking into the passive net\vork (the network with generators
replaced by their respective internal impedances). The theorem is illus-
trated diagrammatically in Fig. 2.4.
Tht%enins theorem is useful in discussing the conditions for maxi-
mum-power transfer from a generator through a network to a load
impedance Z.. As is well kno]vn, ~vhen a load impedance is connected
directly to a generator of internal impedance Z., maximum-power trans-
fer is effected with a load impedance that is the complex conj~lg:~tc of the
generator impedance:
z. = z;.

1W. L. Evcritt, 0p. ant., p. 47.

SEC.25] THE TWO-WIRE TRANSMISS1O.V LINE 21

Consider then the case in which the load Z. is fed by the generator through
a four-terminal net}vork, the generator emf being l~ and ts internal
impedance Z. (Fig. 2.4). The four-terminal network may be replaced
by its T-section equivalent as shown. By Th6venins theorem the SYS-
tem is equivalent to a generator of emf VCZIJ(ZI, + Z.) and internal
impedance 212 Z~J(Zli + ZG) v(z+r12)
V(z
feeding the load impedance Z. di-
rectly. It follows then that maxi- ~
mum-power transfer will be achieved
with a load that is the complex con- ; ; ,(z)~ +Z
jugate of the internal impedance of ,_7(z +(/2)
the effective generator: 1

(2:2)
z. = 2;2 ~+ z: (12)

2.5. The Two-wire Transmission

Line.-0ne of the most important
distributed-impedance systems from
the point of view of antenna theory
is the two-wire transmission line. 1
For the present the line ~vill be con- - ~,
-7I I
sidered in its conventional form, as a I:lG.25.-rwwirere line,
pair of linear conductors in a plane,
which support the propagation of a wave of wavelength small compared
with the length of the lines The problem of interest is the distribution
of voltage and current along the line for a wa~~eof single frequency, in
which the voltage and current vary with eiui.
The line is shown schematically in Fig. 2.5 as a pair of parallel \vires.
In general, however, the spacing betlveen tbe }vires may vary along the
line; the only restriction imposed is that the line have an axis of sym-
metry. Position along the line is specified by the coordinate z along
the symmetry axis. It is further assumed that the ]ille is isolated from
perturbing objects, so that at any position along the line the currents
at every instant may be eqlml and opposite in the t]vo component, lines.
The properties of the line are specified by its distributed parameters:
(1) the series impedance per unit length,

3(2) = Ii(z) + jd(z), (13a)

where R(z) is the series resistance and L(z) the series inductance per
unit length, taking both component lines together, and (2) the shunt

W. L. Evcritt, op. cit. For a very complete treatmentthe madcr is rcfmrx,d

to It. WT.King, H. It. Nfimno,.\. H. \Vin,q,Transmission Li7Les, .Irttenrzas, ad 11we
C(,ifk.s, MrGrav-Hill, \TcwYork, 1945, (hap. 1.
22 CIRCU17 Ifl<I>/1TIO.YS, 1{1{611l{UC1 IY lHEOli8.%l,? [s);,, 2.5

admittance per unit length,

71(z) = G(z) + juC(.z), (13b)

where G(z) is the transverse conductance and C(z) the capacitance per
unit length between the component members of the line. These param-
eters may be functions of position because of variations in the cond~lctors,
in the spacing betlveen the latter, or in the structlu-e of the surrounding
dielectric medium.
Taking either conductor for reference, let i(z) be the current at the
point 2 and V(z) the voltage drop from the reference conductor to the
other member at the same point. To obtain the space dependence of
i(z) and V(z), consider a section of line of lcngt h dz about the point z.
Applying Ohms law, \ve have

V(2 + dz) J(z) = i(2)m3(z) dz

and
i(Z + dZ) i(Z) = ~(Z)y?(Z) dz

for, respectively, the series and shunt relations across the element of
line. The terms on the left-hand side, by use of Taylors theorem,
become (dV/dz) dz and (di/dz) ck respectively. Thus the differential
equations of the line are found to be
dV =
(~(z)i(z), (14G)
z
di
m(z) v(z). (14b)
dz =

Second-order differential equations for voltage and current alone are

obtained by eliminating voltage or current from one or the other of
these equations:
a:y _ [
(15a)

(15b)

From a generalized p~~intof view, Eqs. (14) ran be regarded as the

definition of a two-wire transmission line. That is, given a physical
system supporting a wave with time dependence ew~,the propagation
of which is expressible in terms of a single coordinate z and two quan-
tities (i, V) related by equations of the form of Eqs. (14), it is possible to
set up a t~vo-wire line representation for the system. The voltage and
current of the equivalent line are direct!y proportional to the wa~e quan-
tities entering the differential equations, and the series impedance and
shunt admittance per unit length of the equivalent line are proportional
to the coefficients of the wave quantities in the differential equations.
Sm. 26] 1HE HO MOGIlNE(31i,V 71{.4 .V,V,!ll,~,~IO.V I,l:vl< 23

The generalized concept of a transmission line \villbe made use of in the

discllssion of wavegllidcs in Chap. i, \rhere it \vill be seen th~t the elec-
tric and magnetic field vectors satisfy transmission-line rxluations.
2.6. The Homogeneous Transmission Line .-lhluat ions (15) are the
general equations for a line ~vhose parameters ~ and 9/ are functions of
position. \tTeshall be concerned mainly \vith lines for Jvbich the param-
eters are independent of position, and the subseqllcnt disc~msion \villbe
confined to the so-called homogeneo~ls line. For such a line the coeffi-
cients of dV/dz and dz/dz in Eqs. (15) vanish; consequently, voltage and
current satisfy the same differential equation. The voltage eqllat,ion
becomes
(16]

Defining the complex number ~ by

y = ~ + j~ = (~fi);5 (17)

with the square root taken to be such that both a and p are positive
quantities, ~vefind the solution of Eq. (16) to be

v(z) = A lez + .42CZ (18)

nr
-.
V(z) = A Ie-e-ip + A ,eei@. (18a)

The current i(z) has the same form but is not independent of the voltage.
The relation between them is established by Eq. (14a). On inserting
Eq. (18) into this equation, it is found that
1
i(z) = ~0 (A Ief, ~ *~7.). (19)

The constant ZO is known as the characteristic impedance of the line; it

is given by
}5
z,= ; . (20)
()
If Eq. (18a) is multiplied through by the time factor e@, it will be seen
that the right-hand side is the sum of two waves: The term eid represents
a wave traveling in the positive z-direction, whereas ei~z represents a
wave traveling in the negative z-direction. The wavelength of propaga-
tion is related to the phase constant L3by

(21)

The amplitude of each component wave undergoes attenuation along the

direction of propagation as represented by the factors e-z and e respec-
tively; a is known as the voltage attenuation constant. It is seen from
24 CIRC lJIT RELATIONS, RL?CIPROCI T Y THEORIIM,S [SW. 2.6

Eq. (17) that a may be different from zero, that is, the line may be 10SSY
if one or both of the distributed parameters ~ and !Jl are complex, and
that the line is nonlossy, a = O, if the distributed parameters are both
pure imaginary quantities of the same sign. In the case of the t\vo-\vire
line for which the distributed parameters are given by Eqs. (13) this
means that the line is nonlossy if the series resistance and shunt conduc-
tance are zero, that is, if the distributed impedance along the line is
purely reactive.
The amplitudes .4, and .4 z of the component waves are determined by
the excitation conditions at the input end of the line and the nature of
the termination of the line. Consider a line of total length L, fed by a
generator of emf V. and internal impedance Zc, and terminated in a load
impedance Z~ as shown in Fig. 2.5. In this case the component \vaves
are interpreted simply as a wave of amplitude A 1 incident on the load
Z. and a wave of amplitude .i2 reflected by it. Let the origin z = O be
taken at the termination; the generator is thus located at z = L.
The impedance at any point z along the Iine looking tov-m-d the termina-
tion is the ratio Z(z) = V(Z)/~(Z), which is, by Eqs. (18) and (19),

Z(2) = Zo
(
.4 ~eyz + .4 ze~z
;,1,C-,z _ .427.
)
(22)

At the terminal point, z = O, this must be equal to the terminating

impedance ZL; Ire have then
A1+Az = ~.,
(23)
A, .42 ZO

Thus the ratio of the amplitudes AJA, is determined solely by the

termination. This shows also the significance of the charactc~ristic
impedance: If i?L = ZO, then A 2 = O; there is no reflected \vave. .4 line
terminated in an impedance equal to its characteristic impedance thus
behaves as though it extended to infinity.
A second relation Letll een the amplitudes is obtained from the con-
ditions at the input end of the line. The input impedance Z,,, to the
line is obtained from Eq. (22) l)y setting z = L, and the current at
the point is obtained from Eq. (19) by the same substitution. We ha~e
then
VC = iz=-~.) (ZG + Zi7),
whence we obtain

(1+%)(7+(1
-%)2o =
(24)

From Eqs. (23) and (24) II-efinally get

VCZO(Z,, + z,,)
A1= -- (25a)
(Za + Z(,) (Z,, + ZJe~J (Z<, ---Z,]) (Z,, ZO)e-~
m, 2.6] ?HE HO.WOGE.VI?O 1S Tliil iVSill I,%Y1O.V LI,V.V 25

and
V.zo(z,. z)
2 = (Z. +z)(z. +zo)o (z. zo)(zf. zo)i-~ 25b)

It should benotecf that these expressions give the amplitudesof theinci-

dent and reflected ~vaves at the termination, or more specifically at
z = O. Therespective amplitudes .A~(z) ancl. ~j(z)at an arbitrary point
z are given in terms of the above by

A;(2) = Ale; A;(z) = A2P. (26)

The ratio of the amplitudesof the wavesat any given point isknovm as
the voltage reflection coefficient r(z) at that point. Wehave

(27)

or
~(z) = l(0)ez~, (27a)

where r(0) is the reflection r~efficient at the point z = 0, On making

use of l?qs. (26) and (27) together Irith (22), \ve find that the relation
between r(z) and the impedance Z(z) is

(28)

It is convenient for many purposes to introduce the normalized impedancr

((z) :
(29)

The relations between the reflection coefficient and the normalized

imnedance are then
(30)

Equation (27a) expresses the transformation property of a transmis-

sion line. It is readily seen that l?q. (27a) can be generalized to the
form
r(z t 1) = r(z)e~~~ = r(.2)e*j=+*~fl. (31)
The phase of the reflection coefficient has a space periodicity of A/2.
The amplitude of the reflection coefficient is independent of position
in a nonlossy line. In a lossy line it decreases as we move along the line
toward the generator from the load, corresponding to the increase in the
amplitude of the incident wave and the attenuation of the reflected
wave. The transformation property of the line applies to the impedance
likewise. From Eqs. (28) and (31) it follows that the impedance at a
point z 1 is related to the impedance at the point z by
26 CIRCUIT RIILA TIONS, RECIPROCITY THEOREMS [SEC,27

Z(z) + Z tanh (-yl) .

2(2 t) = 20 (32)
[ Z, + Z(z) tanh (-yl)1
or, in terms 01 the normalized impedance,
((z) + tanh (-A.
W 1, = I + { tad (W (32a)

A section of line of length 1 thus serves as an impedance-transformation

device, converting an impedance Z(z) at the output end into an imped-
ance Z(z 1) at the input end. The impedance transformation is asso-
ciated with the reflected wave; if the terminal impedance is equal to the
characteristic impedance, the reflection coefficient vanishes and the input
impedance at any point on the line (looking toward the termination)
is eaual
. to 2,. If the reflection coefficient is zero, the termination is said
to be matched to the line; otherwise, it is said to be mismatched.
The properties of the line can be discussed in terms of admittance as
well as impedance. The corresponding relations are obtained by replac-
ing Z by 1/Y. The admittance transformation effected by a section of
line is

(z
~)
=1 . ~ ~(z) + Y,, tanh (71)
LY, + Y@) tanh (Tl) 1 (33)

where the characteristic admittance is defined to be

l, = ;;,. (34)

A normalized admittance q(z) is defined in a similar manner as the

normalized impedance
Y(z)
7(2) = y; . (35)

and the relations bet~veen it and the reflection coefficient are

17 lr
r . i~q; (36)
q = l-+-r
2.7. The Lossless Line.The further discussion of the transmission
line will be partimllarized to the case of a lossless line. The microwave
lines to be treated in Chap. 7 can be considered to be lossless over the
length of line that enters into the problem of the design af an antenna.
If the line is lossless, a = O and the propagation constant ~ is a pure
imaginary,
y = jp,
The voltage and current, relations in this case are
l(z) = .4 ,fl~ + A 2(,8, (37a)

(37b)
SEC.2.7! 7HE L(),?,5I,I<SX LINE 27

and the impedance and admittance transformation formulas become

~(z) +~tanf?l
r(z1) ==+jrtanpl, (38)

q(z) +j tan @
7(2 1) = ~tan~i (39)

The transformations have a space periodicity of a half ~vavelength:

the impedance and admittance take on the same values at intervals of a

half wavelength. The reflection coefficient is likewise periodic; if in
Eq. (31) a is set equal to zero, we get

Since r passes through a complete cycle of phase over a half-wavelel,gth

section of line, there are two points ~vithin every such interval at which
r is a real number. It follolvs from Eq. (30) that at these points the
impedance and admittance are real numbers. The magnitude of r does
not vary along the line. Consequently, at every point the reflection
coefficient is a measure of the po\ver loss arising from the impedance
mismatch at the termination. The power carried by the incident wave
is proportional to 1A,[2, and that carried by the reflected \vave is propor-
tional to 1AZI2. The magnitude of r, is given by

(41)

hence Ir I~is the fraction of the incident po~ver reflected by the termina-
tion, and 1 Ir ~ is the fraction of the incident po\ver extracted by the
termination.
In measurements on a transmission line the significant quantity is
the square of the magnitude of the voltage averaged over a time cycle,
given directly by ~1I(z) 12. In computing this from Eq. (37a) it must be
remembered that the amplit{ldes A I and A j are in general complex.
tVriting
Al = j.411eJ4, AZ = 1.4,1e~~, (42)

in F.q. (37a), \\-eobtain

~11(z)] = ~VV* = ; [~.4,~ + \.4,1 + 21Au42) cOS (262 0,+ 0,)]. (43)

Thlls the time average I~rl2 takes the form of a st:in(iirlg-\\-:L\,e

pattern
28 (: Ili?clIT I{ I<I,A71(1.VS, I{ ECIPI<(X17I lhi$ol<l!.liis [Sm. 27

along the line. Ihe maxima and minima occur at those points for \vhich

2,R2 f#I,+I#12 = i2n~

and
262 01 + 02 = *(2n + l)rr, ?l = 0,1 ,2,....

respectively, the distance bet\veen a maximum and adjacent minimllm

being a quarter wavelength. The maximum and minimum vallles are

Ivl:lax = (1.4,1 + 1.4,1),

IvI:,,n = (1.4,1 1.4,1).

The ratio of the maximum to minimum vallle m kno}yn as the power

standing-wave ratio, designated here by rz:

The square root of power standing-lva~e ratio r is knolrn as the voltag~

standing-wave ratio. It follo\vs from lk~. (41) that

1 + Irl
(44a)
=llrl
~vith the inverse relation
r1
[rl . . (44b)
T+l
The magnitude of r may be determined from the measured standing-
wave ratio by means of Eq. (44 b). The phase of r can be deduced from
the positions of the maxima and minima. On inserting Eq. (42) into
Eq. (27a) and replacing Y by jp, we have
r(z) = lr~e~<z@-$+@. (45)

Thus r takes on real values at the points where the standing-wave pat-
tern takes on maximum and minimum values. The phase of r may be
taken to be zero at a maximum point, ~vith

r=lrl=~; (46a)

then at a minimum point the phase of r will correspondingly be rr, and

lr
r = Irld = (46b)
l+r

The phase of J7at any other point, taking a maximum or minimum posi-
tion as a reference point, is then readily deduced by means of Eq. (40).
The impedance at any point can likewise be deduced from measured
values of the standing-wave ratio and the positions of maxima or minima.
It was noted previously that the impedance and admittance is real et the
points ~vhere r is a real number; hence the impwlancc is real at t,he
maximum and minimum points of the standing-lvave pattern. 31aking
use of Eq, (30) together with Eqs. (46), ]ve find in fact that the impedance
takes on the following values at those points:

1 + Irl
(max); (47a)
=llrl=r
~ = ~..lr ~ (rein).
l+lrl=~

Given the value of ~ at any one such point, the value at any other point
is obtained by means of the transformation formula [Ilq. (38)]. Similar
considerations apply to the admittance ~~alues, ~lt the points of maxima
and minima the admittance is a pure conductance ivitb the values

~=: (max), (48a)

r
q=r. (rein). (48/,)
2.8. Transformation Charts.The impedance and reelection cocfh-
cient transformations along a line can be presented graphically in forms
that are very useful in experi-
mental work. There are many ~~.
types of charts, of which two, the 1
so-called circle diagrams, will be
discussed here. They are especi-
Irl=l
ally suited to lossless lines.
The Reelection Coejicient
(&rzitk) Charts.Consider first ----- _ ! _ _ -
~he reflection coefficient transfor- rRe
mation along a lossless line as ex- = cons:
pressed by Eq. (45):
~(z) = rr]e(~z-~,+%). (245) \
@)
Let us set up a complex plane, as
shown in Fig. 2.6, with the real I
1:1~,2.6.011the wflcctioll<wfi.lent i-llatt.
and imaginary axes associated
!lith corresponding components of r, designated by r,t, and r,,,,. r is
then represented by a vector from the origin. The magnitude of r can
never exceed ~mity beca~lse the amplitude of the reflected Ivave must be
less than that of the incident ~vave; ct)nseq~lently we are confined
to the portion of the complex plane cirmlmscrilwd by the unit circle.
It is evident that polar coordinates in the complex plcne are more
appropriate than the cartesian cot)r{linates rlt,,, r,,,, for ~lisc(lssirlg
1P. H. Snlith, Elect ronj{s, ,JInU:Lry,
1!)41,
30 (IIK1lT RIILA TIO.Y-S, II YK11RQCI T 1- YHI<ORI<MIS [%c. 28

the line transformation of Eq. (45). The family of circles centered about
the origin correspond to c~u-ves 1~I = constant or, by virtue of Eq. (44a),
to curves of constant voltage standing-wave ratio. The curves of con-
stant phase of r m-e the family of radial lines from the origin. The line
transformation given by Eq. (45) corresponds to a rotation of r about the
origin without change in length: displacement along the Iine in the direc-
tion of increasing z, that is, away from the generator, produces an increase
in the phase of r, thus rotating r in the positive sense (counterclock~vise),
whereas a displacement along the line toward the generator rotates r
in the negative sense.
The polar coordinate curves are of such simple form that usuarly they
are not drawn in explicitly on the chart. Instead, another pair of families
of curves are introduced, the circles of constant resistive and reactive
components of the impedance, R and X respectively. Writing

=(3+(:)
and r = rR. + j~ImJin Eq. (30),

(230)

and separating rear and imaginary parts, one finds

(49)

These can be ~vritten as

[
r.. ~

-)
20

l+ZO
+ r;,,, =

()
----
~4~2
Zo
y (50a)

respectively.
+(r=a (w
r+)
It is seen from Eq. (50a) that the curve R/Z~ = constant
(50b)

is a circle with its center on the positive real axis at (R/zo)/ (1 + R/ZO)
and radius 1/(1 + R/ZO). Every such circle is tangent to the [ine
I~. = 1 at its point of intersection with the real axis. The circle cor-
responding to R/ZO = 1 passes through the origin and enc[oses a[l the
circles for which R/ZO > 1.
Similar[y Eq. (50b) sho~vs that the curves X/ZO = constant are a
family of circles. For a given value of X/Z,, the center of the circle is
SEC. 28] TRANSFORMATION CHARTS 31

at the point (1, ZO/X) and its radius is lZO/Xl. Every such circ~ is
tangent to the real axis at the point FW = 1. The curves lying in the
upper half of the plane correspond to positive (inductive) reactance, and
those in the lower half plane to negative (capacitive) reactance. It
can be shown that the circles X/ZO = constant are orthogonal to the
circles R/ZO = constant.

FIG.2.7.The Snlith chart.

The Smith chart consists of the circles just described. A typical
chart is shown in Fig. 2.7, the circles being labeled with the corresponding
values of the parameters R/ZO, X/ZO. These curves serve as a system
of coordinate lines. The terminal puint of the vector r associated with
the complex number { = (R/ZO) + j(X/Z,) is located at the intersection
of the circles R/ZO and X/ZO. The distance from the origin to the inter-
section of the circle R/ZO ~vith the real axis is equal to the magnitude of
the vector T that corresponds to a standing-!vave ratio
R
if ;0 > 1,
=20 /
~=~o if 1 <1.
R 2,
32 CIRCUIT RELA TIOiVS, RJ7CIPROCITY THEOREMS [SEC. 2.8

This follows from the fact that r is real when ~ is real and from the rela-
tions of Eq. (47) between the value of ~ when it is a real number and the
standing-wave ratio.
To illustrate these relationships let us suppose that the standing-wave
ratio r has been measured on a given line, together with the position of
a voltage minimum; the reflection coefficient and impedance are desired
at a point a distance 1 from the minimum position away from the gen-
erator. It ~rill be recalled [Eq. (46 b)] that at a maximum position the
phase of r is equal to m; r is then directed along the negative real axis
The impedance at this point is real, being R/Z~ = l/r. The vector ~
thus extencls from the origin to the circle corresponding to R/ZO = l/T-.
Counterclockwise rotation of this vector through an angle 2@ carries us
to the desired point on the chart; the components of f at that point are
read off from the pair of intersecting circles. It will be noted on Fig.
27 that the periphery of the chart carries a phase angle scale with the
phase designated by the ratio of line length to wavelength.
The Smith chart can also be used to study the admittance transforma-
tion. First it should be noted that there are two conventions for the
definition of admittance. The convention adopted in this book defines
the normalized admittance q = (G/ Yo) + j(B/ I0) to be the reciprocal
of the normalized impedance { = (R/ZO) + j(X/Zo); positive susceptance
B thus corresponds to negative (capacitive) reactance. The other
convention defines the admittance to be the conjugate of th~ reciprocal
impedance, in order that positive susceptance (like positive reactance)
should be inductive. The use of the latter convention changes the use
of the chart in ways which the reader can easily develop.
Eqllation (36) gives the relation betlreen the admittance and the
voltage reflection coeiiicient:
lr
=l+r
I,et us define a ne]r coefficient
T=r (51)

and associate Irith it a complex plane ~vith axes ~Reand 11~. (Actually
the same complex plane serves for both I_ and T, the two vectors making
an angle of 180 ~~-itheach other. ) The vector T is, in fact, the current
rejection coefficient, expressing the ratio of the amplitude of the reflected
current ~va~-eto the amplitude of the inriclent current wave. The law
of transformation of 1 alon~ the line is !)wciwly the same as that given
for r by Eq. (49). On substituting 11~. (51) into the i-elation between q
and r, we obtain
1+r
(52)
=lT

This is the same as the relation })etnmn f and J7;it follows that the curves
G/Y, = constant are a family of circles that coincide with the constant
R/ZO family in the ~-~ trtiusfornmtion and that tllc clu~mB/}-O = con-
stant coincide ~rith the .YjZO circles. Ii-ith rwpcct ([) tl]e l:~tter it
should be noted (in using the cll:wt for admittance) that the curves lying
in the upper half plane represent capacitive slwceptam,e. he dis-
tinction that need be made l)et~wxm the use of tile chart for impml:mce
and admittance can he m:de clear by considering tile pr(]l)lem of fin[ling
the admittance at a point dist:mt 1 from a volta~c nlinin~~l]nin tile (lirm-
tion alvay from the gener:ltor, the standin~-i~:hvc mtio :L*ain Iwing r.
At av()ltage nlininlllr~l llliesalc~rlg t}lencg:~tilel e:ll:~\i~;lle1l(e ~extends
along the positi~-e red axis to the cirrle
G
= r.
l,

The starting point thus lies on the positire rral axis, instrad of on tlw
negative axis, lIoving along the line :L\Yayfrom tllc gcnr);ltol :L~:lin
rotates T in the positive sense (collrltc,r(,lti(k}tisr) tllrollgl~ an :L]lglr 21j~.
The admittance at the ne~r point is (lcternliuml from tile ll:~ir t)i int(,r-
secting coordinate curves, jllst as in tile c:wc of tlw inllmi:lncr. It
should be clear that the admit- A
tance and impedance points on the ~~, (%)
Smith chart for one and the same
pnint on the line are diametri-
cally opposite to one another,
The Smith chart is p:llt ic[llarly
suited to the stlldy of :m imlwd-
ance mismatch that arises frt)rn
the superposition of rcflrctions. &
Irl
Forexample, thereln:~yl)(,fi~(,lirs
.,
of dlscontmu ities on a tr:~nsnlis-
sim line; the overall refler(i(}n cc)-
efficient at a given point) is, to a
good approximatif)n, tl~e lert(jr
sum of the reflw,tion c(mffi(irnts 11<,.?~ ll\<.r(,l:lti(jl, I)ct,j,,rll j]l,,j~<[:,,),.,,
:11,(1
:,(1,,,)([:1,,(.(..
that wmdd })e pro(lllce(l fit the
point bytheindiri(lllal (lis(onti[l(li(i(s:l~,iill gscl):ll:lt(l~-ill t{,(,:ll~s[.ll(,r,,f
all the others. The vector :~(l(lition of (llc ([,llll)[)ll(,llt (,twffi(i(nls yi(,lfl,
good resldts if the components arr snmll. 1l]r ~llljjrtl iii]] lx ctjmi(lcl(,tl
further in later chapters in connection ~t-i(ll fil)c(,ili( l)r)hlrnw.
The Bipolar C))arf.s-A cornl)lcx pl:tnc c:ln Iw wt 11]) of which tllr
real and imaginary axes are ass(x,i:ltrd lli(ll col,l(~sl]{~llflil]g(I[)ll~lN~]l(~tlis
of the norm:dizwl in~pml:~n(,e~ (or n(ll,n~:llizc(l :I(llllit[:lll((l ~) jllsi :\::ill
the r~se of the rcfle(,tion r(wfficivuts. jiill(,(~ 11)(J1,(:111):111<1~ Z,l ot 111(,
34 CIRCCI T RELA TIO.VS, RECIPROCITY THEOREMS [SEC2 t3

half plane containing the positive real axis comes under consideration.
The impedance (admittance) is represented in this plane by a vector
from the origin. With reference to the admittance we note again that it
is taken here to be the reciprocal of the impedance. One and the same
plane serves for both impedance and admittance; Fig. 2.8 shows the
relation between the impedance and admittance points in the plane
for a given point on a transmission line.
The impedance transformation
(2.30)

does not take so simple a form in the ~-plane as did the reflection coeffi-
cient transformation in the T-plane. Displacement along the line pro-
duces a change in both the magnitude and phase of the impedance.
The geometrical transformation is simplified by introducing two
families of circles: the curves ]rl = constant and the curves r-phase =
constant. These curves are obtained from the 17-~ transformation
r = (~ 1)/({ + 1) of Eq. (30). Writing r = {r]e@, we find that

,r,2=(:-
)2+EY
and
(E+)2+(9
()
an(a:h
These can be rewritten as

respectively. It will be seen that the curves II = constant and@ = con-

stant are circles. The circle for a given II1 has its center on the real
axis at a distance (1 + Irl 2)/(1 Ir] 2, from the origin; its radius is
21rl/(1 Irlz). Curves of constant Ir I are also curves of constant
standing-wave ratio. By Eq. (43 b), we find that the center of the circle
is at (T2 + 1)/2r and that its radius is (rz 1)/2r. The circle intersects
the real axis at the points l/r and r, corresponding to the values that we
obtained previously [Eq. (47)] for the impedance at these points on the
line where it is real. These two points on the chart thus correspond to
points on the line at which the voltage minima and voltage maxima,
%c. 2.8] 71{.1 .Y,SI,OR.\fA 11O.1- (HA 1<7S 35

respectively, ounlr. The family of circles Irl = constant is showm in

Fig. 2.9, .Myherethey are labeled according to the power standing-wave ratio
r-z.
A circle of constant phase, Q = constant, has its center on the imagi-
nary axis at the point (0, cot o), and has a radius Icsc *I. This second
family of circles is orthogonal to the first, just as in the 17-plane the curves
of constant, Irl and constant phase are orthogonal. In the {-plane all

\ I ,---T I /\ I {

J2,6
Realcomponent
IJIG.2.9.Thebipolari,],pcdar,ccchart,

the constant-phase circles intersect in a point (1, O), corresponding to

the intersection of all the constant phase lines at the origin in the r-plane.
The two families of curves in the ~-plane, taken together with their
image families in the left-hand portion of the plane, constitute a system
of curvilinear coordinates known as the bipolar coordinates; hence the
name of the chart.
The constant-phase curves are labeled in Fig. 2.9 so as to give directly
the change in the phase of r corresponding to a displacement along the
line from a voltage-nlinimllrn point. All voltage-minimum points must
be on the segment of the real axis between zero and unity; this is there-
fore taken as the zero-phase line. The separation bet~veen a voltage
minimum and the adjacent maximum on a line is 1 = k/4, which cor-
responds to a phase shift 2D1 = 180. All voltage-maximum points must
36 CIRCCIT REI].4 TIO.J-,S, RI<CIIRCWI T1 THEORE.%fS [SEC. 29

lie on the real axis bet\veen 1 to cc ; hence this segment of the real axiti
is taken as the phase line @ = 180.
By means of the I]ipolar curves the line transformation can be fol-
loJ~ecl easily. It is clear that displacement along a git-en transmission
line causes the impedance point in the {-plane to move around a circle
of constant standing-irave ratio. Displacement in the counterclockwise
sense corresponds to the same sense of rotation in the r-plane. A half
lvavelength of line produces a phase shift of 2131= 360 and hence a
complete revolution around the r = constant circle. This periodic
property of the impedance transformation ]vas noted pre~-iously (Sec.
2.7). To illustrate the use of the chart, consider again a line in which a
standing-~vave ratio r has been measured and a voltage minimum point
has been located. It is desired to find the impedance at a distance
1 from the minimum point a!vay from the generator. The starting point
is the interscrtion bet!veen the r-z-circle and the real axis on the segment
(O, 1). lVe then move counterclockwise on the P-circle until it inter-
sects the constant phase circle @ = 2~1; this is the desired impedance
point.
The same families of bipolar curves serve for the admittance diagram
likeJvise. In using the chart for admittance it m~lst be noted that volt-
age minimum points are on the segment of the real axis (1, cc) while
voltage maximum points lie on the segment (O, 1). If the voltage mini-
mum is retained as a zero-phase reference point, the real axis segment
(1, cc) must be taken as the zero-phase line and the segment (O, 1) as
the 180 line. The sense of rotation about a circle r = constant remains
the same.
It should be kept in mind that the normalized impedance is dis-
continuous across a junction between lines of different characteristic
impedances; the impedance itself is continuous. On moving across such
a junction the point in both the reflection coefficient and the bipolar
charts in general will moire from one circle r = constant to another. If
we ~ass from a line of characteristic impedance ZO, to a line of charac-
teristic impedance ZO,, the normalized impedance undergoes a change
given by

2.9. The Four-terminal Network Equivalent of a Section of Trans-

mission Line.-For many purposes, in the analysis of systems involving
transmission lines it is convenient to replace a section of line by its
equivalent four-terminal network. The elements of the network will
be derived here for the case of the 10SSYhomogeneous line. Consider
a section of line of length 1, and take the origin z = O at the input end;
let V,, il be the voltage and current at this end, and let Vz, iz be the volt-
age and current at the output end z = 1. From theline equaticmslI;qs.
(18) anci (19)] l~e have then
2=(): v, = A+. 42,

Zl =
Ao
,1-
(.41 AZ);
2=1: T2 = .4,(7~ +A2c7~,
1
i2 = r-- (41@ ~2(j7/),
k

Using the two current equations to solve for (I, and Aj in terms of i,
and iz and substituting into the voltage eqlmtions, ~veobtain

VI = Z,lil Z,2i2,
(55)
V2 = Z21i~ Z22i2, 1
with
Z,, = Z,, = Z,coth(~l), (56a)
Z12 = 22, = Z, csch (71). (56b)

We thus find directly that the network is linear and that the transfer
impedance coefficients satisfy the reciprocity relations. Since the line
is homogeneous, the network is symmetrical ~vith respect to its t Iro ends;
hence Z,, = Z,z. For a nonlossy line ~ = j~; on substitution into the
above, the network parameters are found to be

Z,, = ZZ2 = jZO cot fil, (57a)

Z12 = Z,, = jZO csc fit. (57b)

TRANSMITTING AND RECEIVING ANTENNAS

2.10. The Antenna as a Terminating Impedance.The impedance
relations between a transmitting or receiving antenna and its transmis-
sion line are of particular interest. In the follo~ying sections several
general ideas that are associated with the analysis of these relations will
be discussed. I,et us consider first the case of a line feeding a transmit-
ting antenna. It will be assllmed for the presenk that the antenna is
isolatedin particular, that it is removed from all other antennasso
that interactions with other systems need not be considered. The
antenna fllnctions like a dissipative load on the line in that it extracts
power from it; part of this energy is radiated into space, and part is
dissipated into heat in the antenna str{lct~n-e. In general, the antenna
does not absorb all of the pofver incident on it from the line b~lt gives rise
to a reflected Irave in the line; in effect tlw line is tm-minatml bv an
imped:mce diffrrent from its cli:~l:~(tcl.istitimpedanre. Iloivever, the
definition of the trrminal imped:mre reprrwnting the antenna is not free
from ambigllity and reqllires some Consi(lrmtion.
It is to be notrd first th:lt the definition of a terminfil impwlanrr
 ,

38 CIRCC71T REI,A TIO.VS, RI? CIPROCIT1 THEOREMS [SEC. 2.10

implies the identification of a driving point, or set of input terminals, for

the antenna. In some cases, such as the half-wave dipole or rhombic
loop antennas fed from a two-wire line as illustrated in Fig. 1.2, the
structural discontinuity bet}f-een the line and the radiator sllggests a
driving point. lhis, however, is not enough; it is necessary that the
current distribution in the line be that characteristic of a transmission
line up to the assigned driving point. At long wavelengths this condi-
tion is realized with the antennas cited above: the interaction between
the antenna and line can be represented by a lumped reactive impedance
across the driving terminals in parallel with the impedance charac-
teristic of the antenna itself. At short wavelengths, however, the inter-
action between the radiating system and the line causes a perturbation
of the current distribution on the latter that may extend back over an
appreciable distance; electrically there is no point of transition from trans-
mission-line currents to antenna currents. This is a particularly cogent
point in the case of microwave syste,ms that make use of waveguide
lines, in which the electromagnetic fields exist in the form of a number of
modes.1 A waveguide is equivalent to a two-wire line only when it is
supporting propagation of a wave in a single mode. Microwave lines
are, in fact, generally so designed that they can support free propagation
of only one mode. Nevertheless, though a single mode is incident on the
antenna, the antenna itself excites other modes, in addition to giving rise
to a reflected wave in the incident mode. It is only at points so far
from the antenna that the other modes have been attenuated to negligible
amplitudes that a waveguide is equivalent to a two-wire line. Attention
should also be called to the absence of a unique driving point in cases
where the transition from the line to the radiator is effected by a con-
tinuous structural transition. An example of this is a waveguide flaring
gradually into a horn without structural discontinuities in the walls.
In these cases, again, the transition from transmission-line currents to
antenna currents cannot be localized to a point.
The action of an arbitrary antenna as a terminal load on the line can
be specified in terms of the reflection coefficient r measured in the trans-
mission line, at a point so far from the antenna that its only effect is the
production of the reflected transmission-line wave. At any point in the
transmission-line region an impedance (or admittance) can be determined
from the measured r, by means of Eq. (30); this can be taken as the load
impedance terminating the line at that point. Furthermore, any such
point may be regarded as the junction between the line and the input
terminals to the antenna in so far as the practical analysis of the system
is concerned.
This raises the question of the representation of an antenna by an
equivalent net~vork. There is no unique network associated with a
1See Chap. 7.
SE<.2.10] THE ANTENNA AS A TERMI. VA TIN(2 IMPEDANCE ~g

given value of r. The load to be associated with r at a point taken

arbitrarily as the input to the antenna may be represented by an imped-
ance in the form of a series combination of a resistance and reactance or
equally well by an admittance made up of a resistance and reactance in
parallel. In either case the resistance measures the power dissipated in
the region beyond the input terminals to the antenna; this, if the line is
lossless, is the power dissipated by the antenna in radiation and ohmic
losses. If ~ is the total power (averaged over a cycle) dissipated by
the antenna and t and ~ are respectively the effective current and voltage
at the input terminals, the resistance of the impedance representation is
given by
~ = i2R (58a)
and that of the admittance representation is gi~-en by

(581))

It is tempting to carry over the concept of radiation resistance, ~lsed

so extensively in the long-wavelength region. The total power dissipated
by the antenna is the sum of the radiated power ~, and the power PO
dissipated in ohmic losses in the antenna structure. Correspondingly,
the resistive component of the impedance representing the antenna would
be taken as the sum of two elements: an ohmic resistance Ro and a radia-
tion resistance R,. I~ach element ~vould be given in terms of the power
component by :L defining relation sllch as 13q. (58), for example, the radia-
tion resistance by
~r = il?,. (59)

In the long-] wavelengthregion this resolution is possible because one can

define lmiquely a driving point at Ivhich the antenna network can be
dissociated from the line and because it is possible, on the basis of field
theory, to set up an unambiguous netlvork that is characteristic of the
antenna itself. In the case of the dipole and loop antennas referred to
earlier the net]vork is a series combination of a resistance and reactance.
However, in the general case, where the driving point is merely an arbi-
trary reference point on the line, the antenna netl\-orkcannot be dissociated
from the line, and either an impedance or an admittance representation
can be used. In the admittance representation the resolution of R into
an ohmic component Ro and a radiation component R, (if it is to be
made at all) must place the t]J-o components in parallel. In vietv of the
transformation properties of the line, it is evident that these resistances
IVillIw f~mctions of the position of toe reference point.
Tile practictil si~nificax~rc of tl~e rcferencc point and of the antenna
impedance Z~ (Ieterminc(l from tlm nwabllrcx1 raluc of ~ at that point
ln:~y I)e ill(lstralc(l I)y refcrelltc to tllc matching problem. I,ct 1 be the
length of line from the reference point (regar(lcd nojv as the terminal
40 CIRCUIT 12ELA TIONS, RECIPROCITY THEOREMS [SEC. 211

point of the line) to the generator. It follows from Sees. 24 and 2.9,
the line will transfer maximum power from the generator to a terminal
load of impedance
z; Csc @
ZL = jZO cot @ + z: + jzo cot@ (60)

where Zc is the internal impedance of the generator and 20 is the char-

acteristic impedance of the line. If the antenna impedance Z. is
different from Z., it is possible to introduce a reactive network between
the input terminals of Z. and the line, which (at one frequency at least)
transforms Z. into Z.; this network will effect maximum-power transfer
to the antenna.
It is to be noted that in microwave systems another matching prob-
lem exists: The characteristics of the generator are such that the reflected
wave in the line must be eliminated. This requires that the antenna
impedance Z~ be transformed into ZOin general a different transforma-
tion from that required by the maximum-power-transfer condition. In
these systems the generator must be independently matched to the line;
the generator internal impedance Z. is transformed into ZO with the
result Icf. Eq. (60)] that the maximum-power condition then coincides
with the condition for eliminating the reflected wave in the line.
2.11. The Receiving Antenna System.The equivalent circuit repre-
sentations used in discussing receiving antennas also need examination.
Consider an arbitrary antennait maybe a cer~ter;driven dipole, a horn,
or a combination of such elements with reflectors and lensesfeeding
into one end of a transmission line that at the other end is terminated
in a passive load impedance. (That is, the receiving circuit is free from
generators.) When an external electromagnetic field falls on the receiv-
ing-antenna system, the interaction between the antenna and the field
gives rise to a wave in the line. The antenna may be regarded as a
device that transforms energy carried by a free wave in space into energy
carried by a guided wave on the transmission line. From the point of
view of the terminal load, however, the antenna functions as a generator,
and it is customary. to replace it by a generator in discussing the efficiency
of the receiving system as it depends on the antenna, line, and load
impedances. It is our purpose to discuss the nature of the equivalent
generator. In this connection the problem of modes in microwave sys-
tems again arises. The field excited in the line by the antenna always
consists of a number of the modes that are possible in the given line.
It will be assumed that the line is designed to support free propagation
of a single mode and that the length of line between the antenna and
load is more than sufficient to attenuate the other modes to negligible
amplitudes; there will then be an appreciable region over which the guide
is equivalent to a two-wire line.
sm. 2.11] THE REC117 VING A .V7l?.V.TA .9 JS71Eilf 41

Before discussing the equivalent generator representation, it will be

well tc consider briefly the physical processes of the interaction between
the receiving system and the external field. For this purpose it will be
assumed that an essentially plane wave from a very distant source is
falling on the receiving antenna. In the neighborhood of the receiving
antenna the incident wavefront may be regarded as a plane surface,
over which the electric and magnetic field intensities are sensibly con-
stant in magnitude; furthermore, the electric and magnetic field vectors
lie in the plane, normal to the direction of propagation of the wavefront. 1
We shall assume for the moment that the load impedance terminating the
line is equal to the characteristic impedance of the line. Under the action
of the incident wave a distribution of currents and charges is excited in
the antenna structure; the currents are commlmicated to the transmission
line and give rise to a wave in this ~vhich proceeds toward the load.
Since the load is matched to the line, this wave is completely absorbed by
the load. The current and charge distribution existing on the antenna
under this matched-load condition will be designated as the primary
induced distribution.
Consider now an arbitrary load impedance. This will absorb only
part of the wave excited by the primary indllced distribution and will
give rise to a reflected wave, which ~villproceed to the antenna and excite
there a charge and current distribution, as if the system w-ere a trans-
mitting system. This new distribution of charges and currents will be
termed a secondary induced distribution. The reaction of the antenna
to the reflected wave depends on the impedance of the antenna relative
to the line, as discussed in the preceding section. If the antenna imped-
ance is equal to the line characteristic impedance, there will exist in the
line only the two component \vaves already mentioned. On the other
hand, if the antenna is mismatched, there will occur a process of multiple
reflection between the antenna and the load. The resultant secondary
induced distribution on the antenna is the s~lm of the component distribu-
tions arising from the multiple reflections bet]reen the antenna and load;
its magnitude and phase relati~-e to the primary distribution are deter-
mined by the antenna and load impedances and the length of line between
them. It will be recognized that since the component waves are all of
the same frequency, the net result inside the line is tivo waves, onethe
resultant incident ~va~-etra~cling to\~ard the load, and the second
the resultant reflected wa},e- traveling away from it. Their relative
~.mplitudes are given by the reflection coefficient corresponding to the
Impedance mismatch between the load and the line.
Since the primary and secondary induced distributions on the antenna
both vary with time (with a frequency equal to that of the incident
wave), they radiate and set up an electromagnetic wave in space. This
1A generaltreatmentof [,l[,[,trt)fnagnf:tif,
fil,llls is givvn in ( hap. 3.
42 CIRCUIT RELATIONS, RECIPROCITY THEOREMS [SEC. 2.11

wave is known as the scattered wave. The interaction between the

receiving system and the incident wave is completely expressed in the
relation between the scattered wave and the incident wave fields. There
are two interaction effects: (1) energy is taken from the incident wave and
dissipated in heat in the antenna, the line and the load, being thus com-
pletely lost to the field in space, and (2) energy is taken from the incident
wave and reradiated into all directions about the antenna. The first
effect is known as absorption; the second as scattering. If the dimensions
of the antenna are large compared with the wavelength, the interaction
between the scattered wave field and the incident wave is such as to
give rise to a rather sharply defined shadow region behind the antenna,
that is, on the side of the antenna away from the source of the incident
wave. In this direction the scattered wave set up by the induced dis-
tribution on the antenna is out of phase with the incident wave; the
destructive interference between the tJro fields results in the removal of
energy from the incident watie. This energy includes both the absorbed
and scattered energy. 1 If the dimensions of the antenna are of the order
of magnitude of the wavelength or are small compared with it, there is
no sharply defined shadow region. The fundamental process is the same,
however, in that destructive interference between the scattered wave and
the incident wave in vario(ls directions removes energy from the latter
wave; this energy is in part absorbed and in part scattered by the antenna.
The interaction bet~veen the antenna and the incident wave may be
visualized by ti~inking of the antenna as presenting a certain interception
area or cross section to the incident wave and removing from it all the
energy incident on the cross section. The total interception area is
resolved into two parts: the absorption cross section and the scattering
cross section. Reference ~vas made to cross sections in Sec. 12. To
repeat: Let S be the power intensity, that is, power flow per unit area
of the incident wave, Pab, and P,,a~ the absorbed and scattered powers,
and .4, and A, the corresponding cross sections; then

Pab, = .4,8, (61a)

P .,., = ASS. (61b)

The cross sections are functions of the aspect presented by the antenna
to the incident wave. The reader is referred to Sm. 12 for the definition
of the receiving pattern.
The definition and measurement of the absorbed power is unambiguous
in principle. In microwave systems the poJver dissipation in the antenna
anti line is generally small compared Tvith that iu the load; hence the
1The significanceof the shado~vhas lmcndisc~lsscdin greatdctall f,~ri11~cxw of :1
plane wave incidmt on a sphereby 1.. IIrlllouin,On I.ight Sc:~tterillg hy Sp -es,
A pplted Ilfalh. la,,e~ Reporls, >-111{ (, ColumbiaUniversity,87.1, December 1943, and
87.2, .kpril 1944.
absorption cross sectionor receiving cross sectioncan be evaluated
with small error from the power absorbed by the load. The scattered
power,however, isnotdirectlym easurab]e, andits theoretical evaluation
is subject to ambiguities. Although electromagnetic fields are additive,
their energies are not additive, the resultant energy being modified by
the interaction between the fields. Consequently the energy flow com-
putedfor thescattered wave field, regarded asisolated from the incident
wave field, does not necessarily
represent the energy removed
from the latter and reradiated in
all other directions. This is par- ~
titularly true when the antenna
dimensions arecomparable to the -
wavelength and the interaction G$m
1 . L . I
between the scattered and inci-
11~. 2.10.Cirruit reK>resentatlOn
of ttle
dent waves, ~vhich results in re- recei~,ing
antennasystem.
moval of energy - from the latter,
cannot be localized to a ~vell-defined shadow region.
The equivalent circuit representation of the receiving system is based
on the fact that the antenna functions like a ... generator in so far as the
load is concerned. In replacing the antenna byan equivalent generator
it is generally assumed (1) that for a given aspect of the antenna to~vard
the incident wave, the emf of the generator isproporticmal to the field
intensity of the wave and (2) that the generator has an internal impedance
equal to the input impedance which the antenna presents to the line ~vhen
used as a transmitter. lhe complete circuit is sho~vn in Fig. 2.10, where
the line, assumed to be nonlossy, is replaced by its equivalent T-section;
Z. and Z, are the antenna and the load impedance respectively. It is
evident that this circuit representation involves the same difficulties as
the representation of the antenna by a load impedancethe definition of
Z. and of the input terminals to the antenna. When a driving point can
be localized in the transmission problem, the same point also serves for
the output terminals of the generator feeding the line in the receiver
problem. 310re generally, ~vhen the input terminals to the antenna
can be defined only as an arbitrary reference point on the line, the gen-
erator voltage must be a function of the position of that point; it is not
a prior-i evident that the po~ver relations bet]veen the antenna and load
calculated on the basis of the equivalent circuit are independent of the
choice of antenna terminals. It will bc sho~vn in a later section that the
results for the absorption cross section are independent of that choice.
It ~villbe noted that in I;ig. 2.10 power is dissipated both in the load
impedance ZI, and in the internal impedance of the generator. The
power dissipatr(1 in ~J, is int crprctml :LS the polvrr absorbed from the
incident ]j-a~c I)y tllc antenna and delivered to the load, The po\ver
 .

44 CIRCl!IT RELATIONS, RECIPROCITY THEOREMS [SEC. 2.11

dissipated in Z. is frequently interpreted as the scattered power-the

power absorbed by the antenna (dissipated in its ohmic resistance) plus
the powerreradiated. Neglecting theohmic losses, the powerdissipated
in 24 would thus measure the scattering cross section. It will, however,
be seen in Sec. 212 that the power dissipated in the internal impedance
of the equivalent generator has no direct relation to the energy reradiated
by the antenna and in general cannot be used in discussing the scattering
cross section. Two important cases in which the above interpretation
is valid are that of the dipole antenna and the small (compared with
wavelength) loop antenna. In these antennas, the current distributions
induced by the incident wave under conditions of matched load termina-
tions are the same as the currents excited on the antennas when they
are driven by the line in transmission.
The equivalent circuit representation can thus in general be used only
for the treatment of absorption. It is readily found that the power
delivered to the load by the generator is given by

where
2

~ (6%)

The condition for maximu~(,-po-iver transfer from the generator to the

load in the equivalent circuit gives the impedance relations required for
maximum absorption cross section: the load impedance Z~ must be
such that its impedance, transformed through the T-network of Fig. 210,
is equal to the complex conjugate of 2A. It was noted before that if a con-
jugate impedance relationship exists across any point in the line, it
exists at all points on the line; consequently the load impedance deter-
mined by the conjugate condition is independent of the arbitrary point
taken to be the input terminals of the antenna.
It follows from Eq. (62) that the absorption cross section is zero when
the line is terminated in either a short circuit l,.TL = O) or an open cir-
cuit (ZL = ~). In each case the reflection coefficient of the termination
has the magnitude unity, and all po~ver incident on the termination is
reflected. It is of interest for these cases to compute the power dissipated
in 2A on the basis of the circuit representation. We find

IVGI
pop = 212A jzo cot
fv Re 2., (ZL = m), (63aj
SEC. 212] 7HE YRA.VSMI1YJY1< AIVD RELfiI VER 45

IVGI
,. Rc ZA, (ZL = o). (63b)
1 = 21Z. + jZO tan DIIA

In both cases there are certain lengths of line

nk
~for Z.= co, for Z. = O, n being an integer
[ t;) G) 1
for which the power given by these eqllations is equal to zero. For
cases in which the dissipation in Z~ may be interpreted as scattered
power, this means that the scattering cross section vanishes for the
stated terminations and associated line lengths. This can be under-
stood readily from physical considerations. Since the reHcctiun coeffi-
cient of the load is unity, the voltage impressed across the driving point
of the antenna by the reflected wave in the line is equal in magnitude
to that impressed by the external incident ~vave. The current distribu-
tions excited on the antenna by the t~ro waves are the same except for
phase; hence, by suitable adjustment of the line length, the primary
and secondary induced distributions on the antenna can be put 180 out
of phase, with the result that they give rise to no resultant scattered
wave. The absorption and scattering cross sections are then both equal
to zero. Similar phenomena can be observed ~vith more genre-al types
of antennas. The phase betJ~een the primary and secondary induced
antenna distributions is determined by the load impedance and the line
length. If the load reflection coefficient is unity, the component dis-
tributions on the antenna w-ill be comparable in magnitude, and by suit-
able adjustment of the line length their relative phase can be adjusted
to give a minimum scattering muss section.
2.12. The Transmitter and Receiver as a Coupled System.The
preceding sections treat the transmitting and receiving antennas as iso-
lated systems and neglect the significant featlu-e of the interxtion het~veen
them. Any discussion of a transmitting pattern implies the presence
of a receiving antenna to explore the field; conversely, a discussion of a
receiving antenna assumes the existence of a radiating systcm. The
interaction between the transmitter and receiver is a res~dt of scattering.
Consider a transmitting antenna that, when completely isolated, is
matched to its line. When a receiving antenna is introdlwcd into the
field of this transmitting antenna, it gi~es rise to a scattered ware.
This, when intercepted by the transmitting antenna, in turn gives rise
to a wave transmitted do~vn the feed line of that antenna. The net effect
is that the transmitting antenna no longer presents a matched impcdancc
to its line. The transmitting antenna also in turn gives rise to a scnt-
tered wave that is partly absorbed I)y the recei~-ing systcm and part]y
rescattered. The interaction betwmm the two antennas is thus dl~e to
multiple scattering and absorption.
46 CIRCUIT RELATIONS, RECIPROCITY THEOR&lMS [SEC, 212

From the point of view of the transmission lines, the antennas and
the external space form a network that couples the lines together. In
Fig. 2.11, .4 and B represent the transmitter and receiver respectively,
and O and O are arbitrary but fixed reference points on the respective
lines. It will be assumed that there is no activated generator other
than the cme feeding the transmitter A; the network between O and O

Z,= Z*l-Z,2 Z2=Z22-Z,2

fox
()

$+
~1
v, I

I J
1:1~.2 1I.lr)ur-tertnlndnet~vork
representation
of the coupledtransmitter-receiver
is passire. It will also be assumed that the network is a four-terminal
network in the sense of Sec. 2.2. Thus the voltages and currents VI,
ZI at O and VZ, i~ at O, are linearly related:
v, = Zllil 2A;
V2 = Z21i1 Zzzic;
and the transfer impedance coefficients obey the reciprocity condition
Z,n = 2,,. The transfer impedance expresses the coupling between the
antennas. The basis for these assumptions concerning the properties
of the net~vork is discussed in Sees. 2.16 to 2.18.
The network may be replaced by an equivalent T-section in the
manner discussed in Sec. 2.2. This has been indicated in Fig. 2.11.
The impedance coefficients are functions of the antennas, their relative
configurations, the properties of the external medium and of the trans-
mission lines, and the distance bet~veen the antennas. In the case of
waveguide lines, the reference points O and O defining the network
terminals must be at such distances from the antennas that all modes
other than that for which the line is designed have negligible amplitudes.
As the distance R., between the antennas increases, the im~ortance of
multiple scattering diminishes. The amplitude of the wave returning
to a given antenna as a result of a single scattering process is attenuated
by a distance factor (R~.)-2; that due to stage multiple scattering process
is attenuated by a factor (RAB)4. In the limit R., = cc the coupling
between the two antennas vanishes-the terminals O and O are isolated
from each other. In this limit the impedance arm Zlz of the T-section
becomes a short circuit:
lim 2,, = 0. (64)
RAB+ cc
SEC. 212] THE TRA.VSMI TTER AND RECEIVER 47

Also, in this limiting case, Z, and Z, reduce to the input impedances

Z! and Z; (referred to O and O respectively) of antennas A and B in
their isolated states. When R.. is large but not infinite and A is trans-
mitting, the scattered wave from B has a small amplitude when it reaches
A; the input impedance of A is but slightly different from Z!. If the
impedance at O is sensibly independent of the position and orientation
of antenna B, we have one of the requisite conditions under which B,
acting as a receiver, may be considered to be measuring the transmission
pattern of A. In this situation the antennas are weakly coupled; the
transfer impedance is negligible in its effect on the transmitting antenna.
As concerns the receiver, however, the transfer impedance is not negligible,
for it represents the transfer of energy from the transmitting antenna to
the receiving system. The same considerations apply when B is trans-
fl f)
I
z; z;
1 (ZG)equiv
I I
Zc I
I
I ZL ~ ZL
I (Vc ) equiv
I
I
I I
m
I
/
l~lG.212,-On the receiving sy~tem circuit.
mitting and A is the receiver. For the Iveakly coupled case we may then
set Zl and Z, equal to the respective values at RAB = cc and to a first
approximation write
7J1l = zy + 212, (65a)
222 = z; + 212. (65b)

This coupled network representation provides the correct approach

to the equivalent circuit of the receiving system discussed in Sec. 2.11.
That case was actually one of a weakly ~oupled transmitter-receiver
system. Without loss of generality we may consider a generator of emf
Va and internal impedance Z. to be applied directly to the terminals at
O and a load impedance Z. to be applied directly at O (Fig. 212). By
Th&enins theorem (Sec. 24) the system is equivalent to one in ~vhich
the load is connected to a generator producing an emf

(66)

and having an internal impedance

Z,,(Z: + z.)
(Z.) .xuiv = ! + ~~,+-~~~ (67)

In obtaining these reslllts the lveak-coupling approx~maticms for Z,, and

48 CIRCUIT RELATIONS, RECIPROCITY THEOREMS [SEC. 2.13

2,, given by Eqs. (65) have been used. The receiving antenna is thus
represented by an equivalent generator; the emf of the generator is
proportional to the amplitude of the incident wave (which is propor-
tional to l.). The effect of the orientation of the antenna with respect
to the wave is contained in the functional dependence of the transfer
coefficient 212 on orientation. The internal impedance differs from Z;
by the small quantity Zlz; neglecting the latter, we have the result
(assumed previously) that the equivalent generator impedance is equal
to the input impedance of the antenna when it is transmitting. The
present analysis shows explicitly that the equivalent circuit applies only
to absorption, for Th6venins theorem is applicable only to the treatment
of the power transferred to ZL. In general the power dissipation com-
puted for the equivalent generator impedance is not equal to the power
dissipated in the network bet~veen V. and the load; hence it cannot be
interpreted as scattered po~rer.
2.13. Reciprocity between the Transmitting and Receiving Patterns
of an Antenna.The four-terminal net~vork analysis lea is to the very
important theorem that the transmitting and receiving patterns of an
antenna are the same. In this connection the meaning of a pattern must
be understood from the practical standpoint of the coupled system.
One condition has already been stated: In the case of the transmission
pattern, the distance from the transmitter to the receiver must be so
large that the former is not affected (~vithin the limits of measurements)
by the wave scattered from the latter. In addition, however, one must
consider the interactions between the receiving antenna and objects in
its immediate neighborhood. Multiple reflection and scattering will take
place bet~veen the receiver and such objects; the receiving antenna con-
sists, in fact, of the antenna proper together with all neighboring objects
with }vhich its interactions are significant. If the receiving antenna is to
measure the field at a point, its directive properties must be such that all
such interactions are negligible. These interactions at the receiving
antenna are similar to but are to be distinguished from the interactions
between the transmitter and surrounding objects such as ground. The
receiver measures the resultant of the field produced by the transmitter
and any neighboring objects that interact with it; these together form,
in fact, an extended radiating system.
In Fig, 213, .4 represents the antenna under consideration. In taking
a transmitting pattern a receiver B is, in principle, moved over a large
sphere about .-1, and the relative amounts of po~ver absorbed by the load
terminating the line B in successive positions give the transmitting
pattern of .4. Con~ersely, the receiving pattern of A is obtained as the
relative amolmts of po~yer absorbed I)y a load terminating A lJhen it is
receiving from the antenna B at sllrcessi~e positions on the sphere. In
%c. 2.13] It ECIPIK)L17Y B1<TWE13AIPA TTBR.VS 49

accordance with the usual experimental conditions, no restrictions are

made as to the generator impedance or load impedance; the only require-
ment is that they remain constant in the course of taking a given pattern.
The load in the receiving system will again be taken to be applied directly
to the reference point O or O.
There is an equivalent four-terminal network between O and O
every position of B. Consider the <
transmitting pattern. If Z. is the
\
load impedance at O, the network \ \
equations give (without approxi- 2B
@
mations) \
\
&zL = ~72 = ZIZil Ztziz \
or \
o \
z,, .
2 = 222 + z. ((8) :@

The currents have the usual sig- 1

I
nificance, indicated in Fig. 2.11.
1~1~.~.1~.on the pattern reciprocity
The power absorbed in the load is theorem

p.,, = ~ ]i,12 Z2,Z; z. 2 Re Z.. (69)

Since the coupling is weak, the dependence of the input current i, on the
position of antenna B is negligible. In the denominator of Eq. (69),
the coefficient 222 may be replaced by Z!, for it follows from the weak-
coupling approximation of Eq. (65) that this introduces an error of the
magnitude (Re ZIZ) 3. For two successive positions of B the ratio of the
absorbed powers is given by

(P.,,), _ Iz,,!;
(70)
(P.hs) 2 lz121j
The transmitting pattern is thus determined by the transfer impedance
coefficient alone.
If now B is transmitting and the potver absorbed by a fixed load
terminating A at the point O is measured, the result should be the same
as in Eq. (69) except that il is replaced by the input current i at O
and 2,, is replaced by 2,1. The variation in power with the position of
B (assuming again weak-coupling conditions) is then likewise given by
the transfer impedance alonein fact, by Eq. (70). Hence, subject to
the condition that the transfer impedance corficients ob~y the reciprocit~]
relation, it is jound that the transmitting and rcc~iving patterns of an antenna
are the same. If then G(6, O) is the gain function of the antenna as a
transmitter in the direction 8, 0, the absorption cross section A,(L9, O)
50 CIRCUIT RELA TIONSJ RECIPROCITY THEOREMS [i%C.2.14

presented by the antenna to a plane wave incident from the direction O,

4 is
A,(o,o) = G(tI,@)& (71)
where

(72)

is the average cross section over all aspects. The practical result of the
reciprocity theorem is that no distinction need be made between the
transmitting and receiving functions of an antenna in the analysis of
design problems.
2.14. The Average Cross Section for a Matched System.In conse-
quence of the reciprocal relation between the transfer impedance coeffi-
cients 212 = 221, the four-terminal network representation of the
transmitter-receiver system obeys the Rayleigh theorem. of Eq. (11).
This, taken together with the pattern reciprocity theorem established in the
preceding section, leads to a further important result: The average absorp-
tion cross section o.f receiving system in which the load is matched to the
antenna impedance is a universal constant. The demonstration given here
applies strictly to the case in which the ohmic losses in the antenna and
line are negligible.
Consider again a weakly coupled transmitter-receiver system made
up of antennas A and B, ~vith input terminals at assigned reference
points O and O as in Fig. 2.11. Let the input impedances of the respec-
tive antennas be
z? = I<. + jx.,
Z! = R, + jXm

For a weakly coupled system these are but negligibly different from the
input impedances at O and O when the respective antennas are trans-
mitting. Let us apply a generator of emf V. and internal impedance
Z?, equal to the conjugate of the impedance of antenna A, across the
terminals at (1. The receiving system is assumed to be so matched that
the load impedance across O is Z!. If ij is the current at O, the po~ver
absorbed by the receiver is

This power can be computed in another ~~ay. Let PA be the total

polver radiated by the antenna A; the power radiated per unit solid angle
in the direction of B is (PA/4iT)G~,j GAB being the gain function of A
in the direction AB. The absorption cross section presented by B to
the wa}re from A is by I?q. (71) equal to GM.~,R, G~Abeing the gain func-
tion of B in the direction of A. The solid angle subtended by the cross
section at A is G~.l.~,B/Rj~, ~vhence the power absorbed by B is
sEc. 2.15] DEPE.VDE.VCE OF, THE (ROSS S1<CTION 51

(74)

However, P* is equal to the po~ver supplied to antenna A by the generator:

p _ IPC12.
(75)
8RA

Collecting these results, ~ve obtain

(76)

If the situation is reversed so that B transmits and .4 receives, with a

generator of emf ~G and internal impedance Z! applied across () and
a load impedance ZIo* across 0, ~Veobt:~in by the sar,le ca)culatiOn as
before:
16~li~12R~~RARB
GABG..ii,. = ,~;,, - (77)

In this case i; is the current at the terminals at 0. By the Rayleigh

theorem we have
i; = i2; (78)

hence, on comparing Eqs. (76) and (77), we find

~v.4 = ~TB; (79)

The average cross sections of the two antennas are equal. Since the
antennas are purely arbitrary, this means that the average cross secti(m
of a matched system is a universal constant.
The evaluation of the constant requires at least one detailed analysis
of the interaction between an antenna and a plane \vave on the basis of
electromagnetic field theory. The reader is referred to Slater for such
a treatment of the electric dipole antenna. It is shown there that, the
value of the constant is
x, = :. (80)

The cross section A,(6, O) presented by an antenna to a plane wave inci-

dent from the direction O, 0 is therefore

(81)

2.16. Dependence of the Cross Section on Antenna Mismatch.The

matched-impedance condition between the antenna and the loadthat
the load impedance be the conjugate of that of the antennais the same
I J. C: Slater,MicrowaoeTransmission, McGraw-Hill, A-cwYork, 1942.Chap. tI.
52 CIRClJIT RIII,A TIOATS, RECIPROCITY lHEORE,ifS [SF.(=,
2.15

as the condition for maximum-po~ver transfer from a generator to a load.

This condition can be realized by separately matrbing the antenna and
load to the characteristic impedance of the transmission line if the char-
acteristic impedance is real, as it is for a nonlossy line. The line-matched
system is of partimllar interest in the study of microwave antennas and
is generally taken as a reference system, since transmitting antennas are
required to be matched to the line. Consequently, it is of interest to
, determine the effect of a mismatch
I bet~veen the antenna and the line
Al Z.
1 on the absorption cross section.
The functional dependence of
A
the cross section on line mismatch
~1 is of considerable importance in
,:%
LO I
Z. I Z. the measurement of the gain of
z$=~-[~~ Z* gg
EC / micro~vave antennas. It may be
desired, for example, to study the
1,
h A dependence of the gain on configu-
(a) (b) 0 rational parameters, such as the
l:l(+. 2.14.(3I1 the dependence of the relative positions of a radiator and
a]lsil!ptlonrrohsscltlon 011Illislllatrll:(a] the
T]llsnlatched ; (l,) tl,e li],c-,,l:ltvhed a reflector in a scanning antenna.
Sy$tel,,
system in which a network transfor,nsZ,4 It is impractical in such investiga-
i]lto Z,
tions to match the antenna to the
line in each configuration; rather, a line-matched detector is used through-
out, and the results are corrected for the antenna mismatch of the given
configurations.
Consider the receiving system in Fig. 2.14, composed of an antenna
A feeding a line terminated in a load equal to the characteristic impedance
Z, of line. I.et r be the reflection coefficient of the antenna (in trans-
mission) at a gi~,en reference point O and Z~ = RA + jX~ the associated
impedance. lVe may replace the antenna by an equivalent generator
of internal impcdancc Z.; the emf of the generator will be designated by
V.. Consider no]v t!vo cases: (1) Fig. 2.14a, in which the antenna is
mismatched and feeds directly to the line at O, and (2) Fig. 2. 14bthe
line-matched systemin which a Iossless net~vork has been introduced
bctwmm the antenna terminals at O and the line to transform the antenna
impedance into Z,) at the output trrminals O. It is readily verified that
such a netlvork }Jhich transforms the impedance ZA at O into ZOat O trans-
forms the impcdancc Z, at O into the complex conjugate Z; at O. Case b
therefore meets the conditions of Sec. 2.14. The power absorbed in the
load in the t}~o cases is
Case a:

(~dc = YJ
2
-z _,
lZo + ZAi
(82)
SEC.2.16] THE FOLR-7l~li.JII.~-AL .~-LTliO1ili R1<PRILSE.Y TA TION 53

Case b:

The ratio of the power absorbed in the t]vo cases is the ratio of the respec-
tive absorption cross sections:

(84)

Here (A,) 0 designates the cross section of the matched s~stcnl.

The antenn~ impedance can be evaluated in terms of the reflection
coefficient r. Thus
l+r

and
2.=20
()
~r!

Substituting into Eq. (83), we obtain the cfcsired result:

(A,)mi, = (.1,)0(1 1~1). (85)

The decrease in cross sectionor reception efficiency-is prcciscly tlw

same as the reflection loss intr~ducd by the mismatch on transmission.
Also it will be noted that the mismatch depends only on I~ I; hence the
result is independent of the choice of the reference point O taken as the
input terminals to the antenna.
2.16. The Four-terminal Network Representation.This and the
following sections summarize the considerations underlying the postulate
(Sec. 2 12) that the transmitter-receiver systcm is equivalent to a four-
terminal network between the respective transmission lines. Use will
be made of results proved later in Chaps. 3 and 7. The treatment is
formulated primarily for microwave systems in which the transmission
lines are waveguides. The systems are assumed to be ideal, in the sense
that ohmic losses in the lines and the antennas are negligible.
Consider a pair of antennas A and B, each of \vhich is fed from a
waveguide, as shown in Fig. 2.15. It is assumed that the guides arc
designed to support free propagation of a single mode only. The refer-
ence planes O and O which serve as the input terminals to the antennas
are perpendicular to the respective guide axes and are taken in the trans-
mission-line region of the guides, where only the freely propagated mode
has an amplitude significantly different from zero, WC shall consider
the closed surface S made up of the surface O inside the guide .4, the
 .

54 cIH(1 1 T 111iI..l 710.YS, I{ E(111{()(171 7H1<01!l<,II. S [s1,(. 21(i

interior surface of the guide, the surfaces of the condllctors comprising

the antenna, and finally the extelior s~lrface of g~lide .1 ; this encloses the
A-system completely .i similar s~lrface S encloses the I]-s}-stem. lye
shall be concerned \vith the electromagnetic field in the region J bound-d
by a sphere of infinite racfiIis and by tile surfaces S :Lnd S.

i)

(@) (i&JO)
o
(a) (b)
FIG.215.-oI1 the four-terminal net>vorkanalysis of the tra,lb!,]itter-rcreiver
system.
It will be assumed that there are no generators in the region V. As
regards antennas A and B, either \vemay have the one transmitting and
the other receiving or generators may be applied to both antennas simul-
taneously. However, the particular case involved is of no concern, since
we are interested in the general nature of the relation betiveen the
tangential components E,, HI of the field over the plane O in guide A
and the tangential components Ez, Hz over the plane 0 in B.
The magnitudes of the tangential electric and magnetic fields are
determined by voltage and clwrent parameters V and i, respectively,
which are analogous to the voltage and current in a balanced tivo-~virc
line. Inorder toset~lp afo~lr-terminal netiv()rk re~~reserltati(Jn,]vernust
show that the relation betfvcen the voltage and current parameters Vl,
il at the plane O and the parameters Vz, iz at O is linear:

11 = Z,,il
T, = z2,i,
z,2i2;
Zzziz.
)
)
(86)

To validate the various reciprocity theorems developed in Sees. 2.13 to

2.15 we must then sho}v that the transfer impedance coefficients satisfy
the reciprocity relation
Z12 ==Zzl. (87)

The remainder of this section will concern itself \\it,hthe definition

of the. voltage and current parameters and an exposition of certain of
SEC. 2. 16] THE FOUR-TERMINAL NET WORK RfiPRESEiVTA T?ON 55

their properties that are needed in developing the proof of the four.
terminal network representation. The latter subject proper will be
treated in the followings ection, andin Sec. 218tk~e reciprocity relation
between the transfer impedance coefficients will be established.
The fields in a waveguide are functions of position both over the
cross section of the guide andalong its axis. It will be shown in Chap. 7
that the tangential components of the field over any cross section of a
guide, for a given mode, have the form

E,... = Vg(z,y),
. (88)
H tmg = dl(zjy), I
where the coordinates x, y refer to position on the cross-section plane.
The functions g(z,y) andh(z,y) arecharacteristic of thegiven mode and
satisfy the relation

i=. [g(z,y) xh(z,y)]dS = 1. (88a)

/ crosssection
The quantities V and ithe voltage and current parameters, respectively
are functions of position along the guide axis. If position along the
latter is designated by z, the voltage and current parameters for a general
field of a given mode take the form

V = V+e@ + Ve+@, (89a)

i = TO( V+eI& V_e+jlJz); (89b)

that is, the general field is made up of two waves traveling in opposite
directions along the guide axis, the subscript + in Eqs. (89) referring to
the direction of propagation of the component wave with respect to the
positive z-direction. The quantity TOis a constant, characteristic of the
given mode. Thus the voltage and current parameters obey the same
equations as do the voltage and current in a two-wire line, of charac-
teristic admittance TO. As in the case of the two-wire line the amplitudes
V+ and V_ are determined by the boundary conditions at the input and
terminal points in the guide.
If V. and VP are the voltag~ p~,vameters of two fields of the same
mode, for different boundary conditions on the line, and ia and id are
the respective current parameters, it follows from Eqs. (89) that the
field with a voltage parameter

V, = m=V. + m~V~ (90a)

has a current parameter
i~ = m=im + mfli~. (90b)

This leads at once, by virtue of Eqs. (88), to the corresponding property

of the electric and magnetic fields: I.et E-, Ha and Ed, HL be two Iinearl:y
56 CIRCUIT RELATIONS, RECIPROCITY THEOREMS [SEC, 2.1;

independent fields, of the same mode; then, if we construct the field

E, = maEa + mBE@, (91a)

where ma and m~ are both different from zero, tl.emagnetic field H,
associated with EY is correspondingly
H, = mtiHa +vzdHB. (91b)
This relation between the fields is of fundamental importance to the
discussion in the following section.
217. Development of the Network Equations.We may now pass to
th~ details of the four-terminal network problem. The procedure is to
consider the relation between the fields within the respective guides
and the fields in the external space, thereby arriving at a relaticm between
the fields in the two guides A and B. For this purpose the interior regions
of the guides are thought of as connected with external space to form a
composite region V bounded by the surfaces S and S, as was outlined
in the previous section and illustrated schematically in Fig. 2.15.
Every set of values of electric and magnetic fields El, H, over O and
Ez, H, over O (and hence voltage and current parameters 1,, il, V,, i,)
is associated with a field E, H in the region V. Consider three such fields
that are not simple multiples of one another:
(E,., H,.; E,., H,.; % H.),
(E,~, H,P ; E,d, H,fl; EB, H@),
(E,y, H,,; E,,, H,,; E,, H,).
It follows from Eq. (88) that over the planes O and O the successive
fields differ from each other only in their voltage parameters. (Only
a single mode exists in each guide in the regions of the reference planes. )
Any one of the three fields can be obtained as a linear combination of the
other two, with coefficients m. and m~ which satisfy the relations

VIY = mallm + m~Vl~,

(92a)
VZ7 = m.Vz= + m6V26. 1
By virtue of Eq. (88) the voltage parameters can be replaced by the elec-
tric fields El~ . . E2T. By Eqs. (91), the associated magnetic fields
follow the same law of resolution:

HI, = m.H,a + m.jH,8,

(92b)
HZT = maHzm+ mpHz8. }
T!lis resolution can be effected regardless of the behavior of the fields
througllollt the region 1. Ho!vever, it is meaningful only if the field
E,, H, is the same linear combination of the fields E., H. and E~, HB
throughout V as it is over the reference planes, that is, if
E, = m.Em + mpEfl. (92C)
s~c. 217] DEVELOPMENT OF ?HE NET WORK EQ UA TIO,VS 57

Proof of Eq. (92c) follows from the uniqueness theorem of the elec-
tromagnetic field. 1 The application of the theorem, however, involves
restrictions on the fields. The medium in the region V is characterized
by three constitutive parameters: the conductivity u, the electric induc-
tive capacity c, and the magnetic permeability p. These in general vary
from point to point and are functions of frequencies. In special cases
(such as ferromagnetic media) they are functions of the field intensities;
such nonlinear regions are excluded in the formulation of the uniqueness
theorem. Since the region V includes virtually all space, ferromagnetic
media cannot be simply excluded; we must instead impose the restric-
tion that the fields set up by the antennas be such that their amplitudes
are negligible in the neighborhood of such media. Subject to this
condition, the uniqueness theorem states that in a region V which is free
from generators the field is determined completely by the values of
n x E over the boundary surfaces S and S. The reader is referred to
Stratton for the proof. The same technique that is employed in the
development of the uniqueness theorem leads to the following superposi-
tion principle: If E. is the field in V corresponding to the boundary condi-
tion n x E = F= over IS and S and Eb the field ~vith the boundary
condition n x E = F6, then the field E. associated with the boundary
conditions
n x E = maFa + ?nbFh,

ma and mb being constants, is

E, = m.E. + m~Eb,

It will be noted that since the wavegllides and antennas are all ideal
conductors, all fields E, H, with ~vhich }ve are concerned in the region V,
satisfy the same boundary conditions
nxE=O
over the surfaces S and S exclusi~-c of the truss sections O and O. over
the regions O and O the tangential component of E asslmnes prescribed
values El and EZ respectively. Hence the rcsollltion of E,., and Ez, in
Eq. (92) becomes, in fact, a resolution of the tangential components of
the field E, over S and S in terms of a pair of linearly independent fields:
n x Ey = m.(n x EJ + md(n x Ed).
From the superposition theorem \ve hal-e then that everywhere in 1

which ~vas the desired result stated in I1{. (92c). Thlw given any pair
I SWfor example,,J..1. Stratton,E[d)o///~#Jtc(icThco~y, 31cC,r:llv-Hdl,.\-c\ll-ork,
1941,sec. 9.2.
1Cf. Chap. 3, Sec. 3.2.
 ,

58 LIR67JI IT RELA TIO.VS, RIXIPROCI 1 Y TII.YOREMS [SEC, 2.17

of linearly independent fields over the reference planes O and O, all other
fields may be expressed as a linear combination of the two, the law of
combination holding for all points ir. the region V.
It is convenient to take as the basic set of linearly independent fields
the two fields corresponding to short-circuit terminations over the plane
O and the plane O respectively. Consider first the short-circuit termina-
tion over O, and let the fields over O be designated by E,., H,a, the fields
over O by EZ~, Hz=; let Vla . . . iza be the corresponding voltage and
current parameters. Since the short-circuit means that O is the surface
of the perfect conductor, we must have E2a = O, and hence VzO = O.
Of the three remaining quantities, one maybe regarded as an independent
variable, being adjustable, for example, by a generator applied over the
surface O. Let V,. be the independent variable. From Eqs. (89) it
follows that for fixed conditions in V, that is, a prescribed termination in
antenna B and hence a fixed terminal condition in guide .4, the current
parameter ila varies directly with the voltage parameter VI.:

VI= = allil., (93)

where all is a constant independent of the field amplitude. Furthermore

since O is short-circuited, the field in V must satisfy the condition
n x E = O over all of S, for all values of El.. From the superposition
principle it follows then that the field at all points in V is proportional
to the magnitude of E1ti; in particular, then, the current iz= is proportional
to VI.:
iza = b.Vl. = b.a,llilmj (94)

with ba also a constant independent of the field amplitude.

Similar relations are obtained for the case of a short-circuit termina-
tion over O. Letting V,p, i,d, VW, iz~ be the voltages and currents over
O and O respectively, we have in this case

The general field can be ~~ritten as a linear combination of this basic set:

~Olution of ~qs, (97) for ?nmi,aan[l ~nji2d and substitution into I~q. (~~)
give the linear relation between the voltages and currents in the tll-o
sm. 2.18] TIIli Itl(1l[iOCIl l liEI>. t 710,1 59

guides:

(98)

z,, = !+ Z22 = ~,

allazzbll a11a22ba (98a)

z,, = r> Z2, =7,

.1 = I b=b@,,a,,. I

It is necessary to ol)servc sign conventions in ~ising Ilqs. (98) to relate

the fields over O to the fields over O. The convention will be adopted
here to correspond to that used in Sec. 22: regarding O as the input
terminals to the four-terminal nctworkj the positive z-direction in guide A
is toward the antenna, and il is the positive current entering the network;
at the input terminals O, the positive current leaves the network, the
positive z-direction in the second guide being away from the antenna.
2.18. The Reciprocity Relation between the Transfer Impedance
Coefficients. -Equations (!18) establish a four-terminal network repre-
sentation for the Collpled transmitter-receiver system. The final prob-
lem to be considered is the justification of the assumption that the
transfer impedance coefficients satisfy the reciprocity relation

Z12 = Z2,

}Ve sh:dl make use of the I,orentz reciprocity theorem: I,et E., Ha and
Ed, H@ be tl~o linearly independent fields in the region 1; then

(99)

The cojdit,ions for the validity of the I.orentz theorem are the same as
those stiplllated for tile llniqllcness theorem an(l sllpcrposition principle
in the preceding section.
I,et us apply the theorcrn to the tJ1-o basic fields employed in the
preceding section, The relation (99) in this case reduces to

J o
(nx Ea). H8dS=
/ 0
(n x Ee) . H. cM. (loo)

llal:ing use of Jtq, (M) and tfiking into accolmt the sign conventions on
the c~lrrcnt parameters, JJ-eobtain

1See the zrticlc hy .$. %mmerfekl in Frank and V. 31ism, Die Diflerentid- und
11, p 933, repri!~tcdby Alary S.
[ ateyralq(ct (IL(JrLgfIL der Jlccha nic urld Ph ysik, Vol.
ltmr,)berg,N-W I-ork, 1943.
60 CIR(:{JIT REL.4 TIONS, RECIIROC1 TY 1HEOREMS ~SEC.218

V1.ilp
\ o g(zy) x (zy)] s

= V2~i2a ~, L. [dw) x hJx,y)] d~. (101)

/
By virtlw of the property of the functions g, h of Eq. (88a) itfollolvs that

V1ail~ = V2S&=. (lOlaj

lf m)!v the cllrrents are expressed in terms of the voltages by means of

lLIS. (94) and (95c), it is seen that the coefficients ba and b~ of the pre-
vif)lls section are related:
ba = b6.

It then follows from I?qs. (98a) that the transfer impedance coefficients
obey the reciprocity relation
Zlz = Z21.
 CHAPTER 3

RADIATION FROM CURRENT DISTRIBUTIONS

BY S. SILVER
The fundamental approach to an understanding of microwave
antennas is necessarily based on electromagnetic theory. This chapter
therefore begins with a discussion of the field equations and the general
properties of an electromagnetic field; the treatment is necessarily cursory,
being intended as a summary of material that is familiar to the reader. 1
This theory is then applied to the simplest problem of antenna theory,
the calculation of the radiation fields due to known current distributions.
A discussion of certain idealized current distributions illustrates the
principles of superposition and interference and furnishes a theoretical
guide to the design of various antenna feeds.
3.1. The Field Equations.The field equations relate the electric
field vectors E and D and the magnetic field vectors B and H to each
other and to the sources of the field, the electric charges and currents.
Sources of the Field.The sources will be specified in terms of density
functions.
The excess of positive over negative charge in a volume V is

(1)

where p is the charge density per unit volume.

The rate of transport of charge across a surface S, that is, the net
current passing through S, is

I= J.ndS, (2)
/ s
where J is the current density and n is the unit normal to the surface S
in the direction defined as positive. The current J has the direction of
flow of positive charge, a negative charge moving in one direction being
equivalent to a positive charge moving in the opposite direction.
In the rationalized meter-kilogram-second (inks) system of units, 2

1The reader is referredto .J. .4. Stratton, Elecfro~r~agw/w Theory, lfcGraw-Hill,

New York, 1941,for a inoredetailedtreatmentof many of the subjectscoveredin this
chapter.
2Stratton, op. cit., pp. 16, 602.
61
which M used in this book, the char~e density 1s measure d in ct)lllf)miy.
per cubic meter and the current density in ampei-cs per sqlmre mete?.
As a conseqllenre of the conservation of charge, the charge density
and current dcnsit y are subject to an important rclat ion. Tllr total
current passing wlt of a closed s~lrface ISmust rqual the rate of decrease
of positive charge in the enclosed volume. That is,

(3)

where n is the unit vector normal to the sllrfacc and directed o~lt tj-orn
the region V. By the di~ergence theorcnl

(4)

Substitution of this into Eq. (3) gives

HvJ+$)(L=O (5)

This must hold for any arbitrary vol~lmej no matter how small; c(~nse-
quently the integrand itself must be zero:

vcJ+~=O, (O)

This is the so-called equation of continllity.

Finite charges and rurrents are sometimes limited to s~~rfacesof dis-
continuity. In surh cases the excess of positive over ncgati~e char~e
on a surface S is

where v is the charge density per unit area. Similarly if we let C be a

curve on the surface of discontinuity and nl a unit vector normal tu C
in the tangent plane, then the total current crossing C, that is, the rate
of transport of charge ~cross C, is

I= K . n, ds, (8)
/ c
where K is the surface-current density. The surface-current density K
and the charge distribution q on the boundary of an infinitely conducting
medium must satisfy an equation of continuity analogo~ls to the volume

I A treatmentof the CIivcrgrncr throrcm WI(I,Stokcssand C,rct,t)sthcoren]s,lvhl{1)

areusedsubsequently,m<tybc follrld ill any text on vt,[,tur an:ll}-..is. Ser for cx:ullpl{,,
H. B Phillips, Jcrkw.lnnlysis, l~ilc.v, Xcw }ork, 1!)33.
SW. 3.1] YHli FIEI>D liQ7JA ?IO,V.S 63

distributions, This eq~mtion of contin~lity, in iutcgral form, is

(9)

where C is any ciosed curve enclosing an area S.

Another form of this relation is

VSOK+~t=O, (lo)

where the surface divergence of K, V.! c K, is defined by

(11)

A being the area circumscribed by the cllr\-e C.

Dcjnitzons of the Field Quantrtie,s-The field vectors E and B measure
the forces exerted on charges and c~lrrcnts respectively. The force on a
stationary charge q at any point in the field is

F = Eq, (12)

1he total force on a c~lrrcnt distril)lltif)n t,hrough a volume T of space is

F= J X B dv, (13)
I T
the integrand being the vector prod~lct of J and B. The vmtor E is meas-
ured in volts per meter and B in ~vehers per square meter.
The field vectors D and H are determined by the field sour{:cs and
are independent of the medium. The net o~lr]vard fl(m of D thro~lg}~ a
closed surface S is a direct measure of the enclosed charge Q:

D.ndS=Q, (14)
+s
ivhere n is the unit vector normal (J~ltlvardfrom the enclosed region. The
magnetic field H is related to the c~lrreut. If Z is the net current passing
through a surface S bounded by a curve C, then

$c H.ds=l. (15)

The integral on the left is the line integral of the tangential component
of H along the curve C; the direction of integration is s~lch that an obscr~,cr
traversing the curve in that dirertion Jrill hare on his left the positi~,c
normal n used in dcfirliug the clwrent 1,
The licld Equa(20ns.-rllle field cqllatious expressing the rclalious
betJveen the ticlcf vcrtors and the sources m:~y be set ~lp either in (lifie)-
ential or integral form.
 ,

64 1(A 1)1,1 Ilo,v iJIM),\l ((1< 1{1<.17 l)l.sl 1{1131YIoh.s lsrx. 31

The differential relationships, 31ax\vc11seqllaticmsj are

VXE+dj=O, (16a)

VxH=J+~d+, (16b)

v. B=O, (16C)
V. D=P. (1(id)

Equation (lfk) may he drrivcd from lk~. (1(kz) I)y taking the clivcrgen[e
of the latter, Similarly, ]tq. (1 (;d) may he derivml I)y taking the {ii\.er-
gence of Eq, (16b) and comparing the reslllt ~rith the eq(lation of continllity

(l (jr)

Equations (16a) to (lGe) must be obeyed simultaneo~dy hy t}w ficl(l

components and sources of any elect romfiglletic firlti.
The corresponding integral rclatiorw am i}le folloli-ing, let C lx a
closed curve spanned by an arbitrary s~lrface S; then

$c E.ds= -~
al / ,y
B .11 ds, (17fl)

$ c ds=NJ+ands (17[,)

the positive direction of integration armlnd the c~lrvr C lwin~ that

defined previollsly. The first of these relations is I?aradays la\J of elec-
tromagnetic induction, and the second is the generalization of .imphs
law in \vhich the cllrrent density J dlle to charge is s~lpplcmented hy the
displacement-current (lcnsity dD/dt. These e(lll:~ti{)ns r:m IW drrivcd
from I;qs. (1 (is) and (1(ib) l)y the Ilse of Stok(ws tllet)renl. By :~l)plica-
tion of the divergence theorem to Eqs. (1(k) and (16d) one obtains tlfo
more integral relations:

B.nd,S =(), (17C)

48
(17(/)

\vhere the integrals extend over tl~c clf)srd sllrf:we ,$ of a w)l~lme V,

Equivalent .Ifaqnetic C)tnrfjc atui C~~rrc/~t.ll(~~l:~liorls ( I(k) and (Iic)
express the fact that there exist no f rce magne( ic charges an(l c(Jmespond-
ing magnetic cllrrents, IIt)\Ic~cr, it is at times (Jc)n!-elliellt t{) illl rodllcc
equivalent distril)l~tions of sllch cll:~r~cs :u](I cllnwlt<, .~ sinll)lc exan]plc
is provided by the in finitwimal cllrrrllt l(mp. 171is i< e(llliv:llcnt to N
magnetic dipole normal to the plane of the loop, lf (Ile c(lrmnt in tile
SEC. 3.2] THE CCLVSTITUTIVE PARAMETERS 65

loop varies with time, the dipole strength varies likewise; the effect is
that of a magnetic-current element.
In diffraction theory, equivalent magnetic-charge and magnetic-cur-
rent distributions are introduced in a more general way. In the presence
of a magnetic-charge distribution of density P,. and a magnetic-current
distribution of density J-, Nfaxwells equations assume the more sym-
metrical form

vxE= J.~, (18a)

vxH=J+~, (18b)

v. B=p~, (18C)
v. D=P, (18d)
with two equations of continuity

VOJ+~=O, (18e)

V. Jm+~=O. (18f)

It is to be emphasized that the magnetic-source densities are mere

formalisms. We introduce them here to avoid later repetition of certain
mathematical developments. They will be different from zero only under
very special circumstances.
3.2. The Constitutive Parameters; Linearity and Superposition.
There exist between the various field vectors further relations that depend
on the medium.
In isotropic media the vectors D and E have the same direction
at any given point, as do the vectors B and H. The ratios of their
magnitudes are constitutive parameters of the medium:

~=,D (19a)
E

the electric inductive capacity, and

~=, B (19b)
H

the magnetic inductive capacity. These quantities may be functions of

the field intensities and the frequencies. They depend on the field
intensities only for a small group of substances ~vhich we shall exclude
from our discussion. The frequency dependence is a very general prop-
erty. In vacuo these parameters are constants and have the values

co = 8.85 X 1012 farad/meter,

#0 = 47r X 107 henry/meter
(;(i Ii.tI)lA 771(IX FIiO.JV C( I{ I{I<,Y i I) IS TI<l BITTION,? [s,:(, 33

The constitutiye parameters are mole commonly specified in terms

of the specific inductive capacities

k= = :, (20a)
~o
km = ~. (20b)
Po
The quantity k, is known as the dielectric constant; k,mas the magnetic
permeability. These ratios are dimensionless and independent of the
lmits. For practically all materials o interest in antenna ]vork km is
but negligibly different from unity and will be talicn equal to unity unless
othertJise indicated.
It is important to note that, although D and E are in the same direc-
tion, they are not necessarily in phase. Such phase differences depend
on the molec~dar st,rllet[lrc of the medi(lm and arc connected ]vith dis-
sipation of clcctrornagnetic energy in i hc medi~lm. They w-e c(~nvcnient]y
taken into account I)y expressing t as a complex number,

t ==c, ]6,. (21)

l}~e energy losses associated with the imaginary part of c are to I)e dis-
tinguished from the ronrf[lction loss associfitecf }Vith ctmdllction currents.
T}\o types of clu-rents may contribllte to the solwce f~mrtion J : mm-
vection currents and conduction currents. In the present volunle \re
shall be concerned only !I-ith conduction currents, for ~vhich the current
density is proportional to the electric field vector E :

J = aE. (22)

The constant u is the conductivity of the medium. IJike the other con-
stitutivc parameters it rna,v bc fre(lllency dependent. .i conducting
medium cannot support a free vulunw-charge derlsity p; if the conduc-
tivity is at all appreciable, p may be tdmu to be zero at till times.
If the constitutive parameters are indepcndcut of the field ~twugth,
all relations bet]reen the field ~c(:tors-31:~xl!clls e~lllfitions an{l the
constitulive relations [J]qs. (l!la), ( l!)b), and (22)] are lillcar, (-ndrr
such circumstances the superpmsitiun princil)lc applies. lllis st:~tes th:lt
if a set of field vectors E,, . , H, and sollrcc fllncti(uls p, and J,
mtisfies the field equations and a second set, of field vectors E*, . , H2
and source functions p~ tind JZ d[ws so also, then tllr sllm of these t\\o
solutions EL + E*, . . . , PI + P2. J1 + J? :LIs() satis(ics tile field :u~d
constitutive equations and dcsrrilx>s a pt)ssil)lc elect rt)nlagneti(, field.
343. Boundary Conditions. In ad(lit itm to the ficl,l c(lll:~titms, ]vhich
give the relations lmtlvmu the (Jl(,lllcnts f)t II]c fi(,l(l in a nlc(li(lnl \fitl\
continuo(ls]y varying l)r(}perti(,s, \\(,nlllst liIlo\\-tile rrlalit)us t]l:lt exist
 I
at a boundary where the properties of the medium change discontinu-
ously. The derivation of i hew boundary conditions starts from the
integral forms of the field collations; the procedure is standard and \rill
he found in any text on electromagnetic theory; we shall simply state
the results.
I,et us consider the boundary surface between two media with
const,itutive parameters cl, p,, uI, and (z, Pzj uZ, respectively. Let the
positive unit vector n normal to the boundaly sllrface be directed from
medium 1 into medium 2. If E,, En, . . , H,, Hz are the field vectors
at contiguous points on either side of ti-.e bour,dary, the boundary ctin-
ditions are the follolring: I
1. The tangential component of the electric field intensity is con-
tinuous across the boundary:
nx(Ez El)=O. (23)
It can he sholrn that a field penetrates into a conducting medium
a distance invmscly proportional to the sqlmre root of the con-
ductivity. Thus if al = m, E, must be zero; this boundary con-
dition then reduces to
nxEz=O (m, = m). (24)
2. There is a discontinuity in the norma,l component of D at the
boundary if there exists a surface layer of charge:
n . (D2 D1) = n. (E2E2 clE1) = q, (25)

the charge density per unit area being q. Such layers of charge
occur, in general, only when one of the media has infinite
conductivity.
3. The normal component of B varies continuollsly across a boundary:

n. (Bz B,) = n. (P2H2 ~, H,) == o. (26)

4, ii discontin(lity in the tangential component, of H occurs only

~vhen there is a surface-c llrrent sheet on the boundary

nx(Hz H, )=K, (27a)

K being the surface-current density. Such current sheets exist

only if one of the media, say the tirst, is infinitely conducting.
In this WSC, !~oivc~er, tIlc firlcf cannot penetrate the rnediurn; HI
must be zero. lye have then

nxHz=K (0, = m) (27b)

and likewise
n. BZ=O (0, = m). (28)
 Under all other conditions K is zcu), and the tan~cntial component
of H as well as the normal component of B is continuous.

These boundary conditions apply to fields th:it satisfy A1ax\rells

equations [12cIs.( 11)] e~wryllhcre. Ile sh:~ll }la~-e occa,sion in diffraction
problems to consider a }xnmdary s~lrf:~cr lwt t}cm t!\o regions of the
same medium. From solutions of llaxnells cq~iati{ms in these tlvo
regions we shall form functions that are sol(ltions of 31ax\vc11sequations
every tvhere except on thi:s s~lrfare, !vhere they are discontinuous. These
discontinuities can hc fornmlly associated ~~ith distrihlltions of magnetic
charges and currents on tl~c hol~n(lary sllrf:wc lJY equations that can be
obtained from the 31ax~c!l cqlmti~)ns [1;(1s, (13)] in ~vhich magnetic
sources have been intr(xlller (l:

n x (EZ -- El) = K., (29)

and
n . (BZ BJ = ~m, (30)

respectively, where Km is the density of the fictiti{nls magnetic-current

sheet over the boundary and q~ is the dmlsity of the fictitious s~lrface
layer of magnetic charge, As in the (I:LSC
of clr(.t ri(. cllrrcnt and charge, the
magnetic-source functions mllst satisfy a slwface equation of contin~lity,

aq,n
VS . K,,, + ~1 = O, (31)

]Vhere as before V.S is the s(lrf:~r[~-(livcl,gc~l(eoperator.

3.4. The Field Equations for Harmonic Time Dependence. It ~~ill
be sufh(ient for most of our pur[msw to c(~nsidcr flclds havirlg a harnlonic
time dependence. In such cases we shall take W fic~d and source dis-
trib~ltions to depend on time throllgh thr samr factor CM. The real
and imagimwy parts of these complex sollltions of the field equxtiorrs
l~ill themselves be solutions of the field wllmtit)ns and }Yill describe real
fields, ~he assumption of harmoni(, time dependence \\-illnot grratly
affect tllc Xenrrality of OU~ resldts bccallsc an :Lrl)itrary field and solute
distribution c:tn Iw rfwjlved into h:Lrmr(,nicc(,ulponents.
IYith tllc restriction of tllc time cfrpendcrlcc to the tirrle factor c~f,
the field equatifms may be trritten as

v x E + juPH = J,,,, (32a)

v x H = (u + jox)E, (32b)
v . (pH) = P.,, (32c)
V o (cE) = P, (32d)
v. J+jup=O, (32( )
V . J,,, + jmp,,i = O. (32,f)
SEC. 35] PO YIVT1,Y(J,5 7HI<Ol{E,\f 69

These equations apply equally to the field quantities and their space-
dependent factors. Equations (19a), (19 b), and (22) have been applied
in this formulation. Equations (32c) and (32d) have been \vritten in
the general form, for inhomogeneous media in which c and p are func-
tions of position. It should be noted that the equation of continuity
determines the charge density directly from the current.
3-6. Poyntings Theorem. Discussions of the energy relations in an
electromagnetic field are usually based on Poyntings theorem. From
the first two of hlaxfvells equations, (16a) and (l W), we obtain

H.vx EE. vxH=H. ~E .: E.J. (33)

The quantity on the left is equal to V . ~E x H). On use of the con-

stitlltive relations [Ikluations. (19a) and IW)], Eq. (33) becomes

(EXH+EJ=-W+W (34)

This is Poyntings theorem. Formally, Poyntings theorem resembles the

equations of continuity previously considered; it expresses the conserva-
tion of energy, rather than that of charge. The Poynting vector

S=EXH (35)
is interpreted as the intensity of flow of energy, that is, the rate of flow
of energy per unit area normal to the direction of S. The quantities
CE2/2 and ~Hz/2 represent the densities of electric and magnetic energy,
respectively. The term E . J measures the rate of dissipation or produc-
tion of electromagnetic energy per unit volume. If E o J is positive, it
is a dissipation term; if it is negative, it represents production of electro-
[:]agnetic energy.
The analogy of Poyntings theorem to the equation of continuity
is brought out more clearly in the corresponding integral form. I,et us
integrate Fkl. (34) over a volume V enclosed I)y a surface S:

\ v
v.(Ex H) do+
\

Jd=-:/v(~+$)dv
36)
lIaking use of the divergcmce theorem, we can transform the first integral
into a surface integwd over the boundary, obt:~ining

$s S.ndS+ / v J= -:, L(:+$)V 37)

JVith the interpretations of the intcgrancls given above, Eq. (37) states
that the net rate of ffoiv of energy o~lt through the boundary surface
pl(is the rate of dissipfition of electromagnetic energy within the volume
(or minus the rate of production) is equal to the rate of decrease of elec-
tromagnet ic energy stored in the volume V.
Equation (34) is quite general in its applications. lVe have now to
express Poyntings theorem in a form applicable to fields varying peri-
odically with time. In this connection it must be noted that the complex
exponential representation of periodic fields can be carried through all
linear operations but that in nonlinear operations (such as formation of
the products occurring in the Poynting theorem) the real expressions
for the field quantities must be used. The complex field vectors may be
expressed as
E = (E@) = (E, +jE,)@t, (38a)
H = (H@) = (H, +jH,)@. (381,)
The corresponding real fields are
ReE = (E, coscd E,sinut), (39a)
Re H = (H, cus cd H, sin ut). (39b)

The Poynting vector is thus

S= ReEx ReH (40)

= [E, x H, COSd + E, x I-L sin of (E, x H, + E, x H,) sin d cm d]

In general we are not interested in the instant: meollsffotv but in the

energy flolv averaged over a cycle. That is, \ve i~ish to know

=WSJ(7=3 (41)

the over-line denoting the time-average val~le. A-ow the time average
of sin w~cm d vanishes, and the time average of both cos~ cd and sin! d
is ~. Hence
~ = ~(E, xH, + E,x H,). (42)

It will be observed that except for the fartor ~, the right-h:md side of
Eq. (42) is the real part of E x H*, ]rhere H * ueprcsents the ct)nlplex
conjugate of H. We hfive then
~ = ~Re (E x ~). (43)

JYe shall seldom be concerned v-ith the instantaneous Il)ynting \ector.

Unless explicitly stated obherwise, all future reference to the l)oynting
vector ~vill be to the time-average value given by l;q. (43); the overline
\vill be omitted hereafter except where a distinction must be made.
It is of interest to formulate Poyntings theorem in terms of time-
averaged quantities. Since the divergence is a linear operator, in\wlving
space derivatives only,
v. S=v. ~= V. Re~(Ex H*) =~Rev. (Ex H*). (44)
 .

Sw. 3.6] THE WA VIi EQUATIONS 71

In the absence of magnetic charges or currents one has, for a field with
harmonic time dependence,

1
v.(Ex H*)=: (H*ovx EE oVx H*)
5

;( u j(Jc*)E. E* +H. H*. (45)

Taking the real Dart of Eq. (45), ]vith due regard

.. for the complex form
of 6* [13q. (21)], we obtain the modified Poyntings theorem
V.s= ~(u + UC,)E. E,* (46)

or, in integral form,

S.ndS (47)
$s

Since the unit normal n is directed outward from the region enclosed by
the surface S, the term on the left of Eq. (47) is the net average polver
flow across S into the region V. In view of the harmonic time depend-
ence of the field] there can be no average increase in the energy stored;
the terms on the right must be interpreted as electromagnetic energy
dissipated }Vithin the region V. Thus, the imaginary component of the
electric indllctive capacity, like conductivityy of the material, results
in energy dissipation. A material ~vith a complex dielectric constant is
called a lossy dielectric. By lq. (47), if a medium is rmither a con-
ductor nor a 10SSYdielectric, the net po}ver flow across a closed surface
S into the region enclosed by it is zero.
3.6. The Wave Equations.We turn no~v to a consideration of the
wave eqlwtions satisfied by electromagnetic fields. \Ve begin ]rith
Mamvells eqllations in the form [Eqs. (18)] that inclu(les magnetic
sources but confirm our discussion to lineau homogeneous media; c and
Y are constants independent of position.
Taking the cud of I;q. (18a), eliminating the magmctic vector B I)y
means of Eqs. (18b) and (196), JVeobtain

VXVXE+pC;;E=g$V xJ.. (48)

Similarly, intmchanging the roles of l~qs. (18a) and (186), ~veget

(49)

IV(>noiv make Ilse of the vector identity

vxvx P=v(v. P)v~P. (50)

72 RADIATION FROM CURRENT DISTRIBUTIONS [SEC. 36

On application of this to both the previous equations and replacement

of V o E and V . H by p/c and pJp respectively, Eqs. (48) and (49) become
a2E
dJ
V2E /.Lt~ =PZ+VXJ. +1 VP, (51a)
e
v2HpE~ d2H f= r3Jm
VXJ+; Vpm. (51b)

On the left sides of these equations are the familiar differential terms of
the wave equation; the terms on the right represent the effects of dis-
tributions of sources. In a source-free medium these equations reduce
to the homogeneous wave equations

(52a)

(522))

with the speed of propagation of the wave given by

~= 1
(53)
Vz

The speed of propagation in free space is a constant, independent of

frequency:
c= = 3 X 108 meters/see. (54)
dk
The index of refraction of a medium is defined as

n=:=~k=. (55a)

For most media the magnetic permeability km is unity, and

The wave equations simplify for fields with time dependence e]ui, in
that the time can be totally eliminated from the equations. There result
the so-called vector Helmholtz equations for the space dependence
of the fields:

VXVXEkzE=jupJ VX J.,, (56a)

VXVXH PH=-jucJm+V xJ, (56b)
where
~z = ~zpe. (57)

The constant k is known as the propagation constant. In nonlossy

media it is real and is related to the ~yavelength by

(58)
 .

SEC. 37] SIMPLE WAVE ,SOLUTIOKS 73

If c is complex, both the speed of propagation defined byEq. (53) and

the propagation constant are complex. The attenuation of a wave as
it propagatesin a lossy medillm indirectly connected ]vith the imaginary
part of the propagation constant.
Apply ing Eq. (50) to Eqs. (56) yields

(59b)

Tn a source-free medium these redllce to the homogeneous equations

v~E + k2E = O, (60a)

vH + kH = O. (60b)
It should be emphasized that ills. (60) imply that each rectangular
comp(ment of the held vectors F..,
F ,U, , H, satisfies the scalar Helm-
Itoltz equation
V2+ + kz+ = (), (61)
Though all fields that satisfy IIaxwell]s equations necessarily satisfy
the wave equations, the converse is not true. A set of field \ectors E
and H that satisfy the lm~-e equations constitute an admissible electro-
magnetic field only if at the same time they satisfy llax\rells equations.
F~[rthermore, the fields must behave properly at the boundaries of the
region concerned in accordance 11-iththe boundary conditions formulated
in Sec. 3.3. If the region is infinite in extent, separate attention must be
paid to the behavior at infinity,
3.7. Simple Wave Solutions.Gen~ral considerations relative to
Jrave propagation \vill be developed in the next chapter. We shall con-
sider here several simple ~J-aveforms, solutions of Eqs. (60), that recur
frequently in general antenna theory. These are (1) the homogeneous
plane wave, (2) the circularly symmetrical cylindrical \vaves, and (3)
the isotropic spherical ~vave. In each case the medium is assumed to be
homogeneous, nonconducting, and free from sources.
Plane Waws.-hehe plane wa~e is mathematically the simplest type
of electromagnetic ~va~-e;its propagation is essentially one-dimensional.
Let us attempt to find a field such that the directions and magnitudes of
the field vectors are constant o~er any plane normal to the direction of a
vector s (Fig. 31) bllt vary periodically along lines parallel to S. In the
case of tbe electric field vector E, the conditions stated above ;~ill be
satisfied if the field has the form

E(x,v,2,L) = EOc@~-~r.s), (62)

tvhcn r is the positiun vector from the origin to tbe field point (z,u,z)
74 RADIATION FROM CUR RJ7,VT L)IS TRIBUTIO,VS [SEC. 3.7

and k is the propagation constant defined by Eq. (57); the amplitude EO

is independent of position and time. Since the planes normal to the
unit vector s are defined by r . s = constant, this field must be uniform
over every such plane. These
z planes are equiphase surfaces for
the wave, and its propagation can be
visualized as a continuous progres-
sion of one equiphase surface into
the contiguous one. It is seen
further that at any instant the field
.
has the same magnitude over each
of the family of parallel planes
-Y
~s=c+25=c+nA
.
z
k
FIG. 3.1.The plane wave. n=o, l,2j . . . . (63)
It is readily verified that the electric field vector defined by Eq. (62)
satisfies the wave equation [Eq. (52a)]. Obviously a similar expression
for the magnetic field vector H(z,y,z,t) is a solution of Eq. (52b). How-
ever, if these field vectors are to describe an electromagnetic field, they
must be so related as to satisfy Maxwells equations. The require~
relation is Eq. (32a):
H.~vx EO (64)
UP
On introduction of Eq. (62) this becomes
,
-
H= $ (S x E) = + (S X E,)@u-k@. (65)

The space-time dependence of H is the same as that of E, but the direc-

tion of H is normal to that of both s and E,. Equation (32d) requires
V . E to be zero in a source-free medium. Thus

v. E=jks. E=O; (66)

that is, E is normal to s. To satisfy 31ax\ve11sequations, the electric
and magnetic field vectors must thus lie in the plane normal to s. It
follows at once that the energy flow, that is, the Poynting vector,

S = ~Re (E xH*) = ~ ~ ?51E,I%, (67)

()
is in the direction of propagation of the wave, normal to the eqlliphase
surfaces.
It is of interest to determine Ivhether or not there ran exist a Jane
wave of the form of that in Eq. (62) if the magnitluie of Eo is an arbitrary
SEC. 37] SIMPLE WA Vh SOLUTIONS 75

function of position over an equiphase plane. Without 10SSof generality

we can take the direction of propagation along the z-axis and the direc-
tion of EOalong the z-axis. We are thus considering the field

E= = EO(z,g)ei~z, EV=EZ=O (68)

(omitting the time factor e~). If this is to be a possible field, E= must

satisfy the scalar Helmholt z equation [Eq. (61)]. This will be true only
if EO(z,y) is a solution of the two-dlrnensional Laplace equation:

d2E0 + d2E0
=0.
@2
(69)
ad

Since there are no sources, EO(Z,y) must be finite and continuous over the
infinite z,y-domain. However, being a solution to Laplaces qquation,
EOcan have no maxima or minima in this infinite region. Consequently,
EO(Z,y) must be a constant; arbitrary amplitude distributions and infinde
plane equiphase surfaces are incompatible.
It should be noted that the infinite plane wave is impossible physically
because the total energy transported across an equiphase surface is
infinite. The practical importance of the plane wave lies in its use in the
analysis cf other waves. There are two parameters characterizing the
plane \vave: its angular frequency o and the direction of propagation s.
By superposing time-periodic plane waves, all traveling in the same
direction but with various values of u and amplitudes E,(w), it is pos-
sible to build up a plane wave of more general time dependencea pulse-
modulated or ot her\\i se modulated wave. By superposing plane waves
with the same frequency u but with various directions of propagation
and amplitudes EO(s), it is possible to synthesize a time-periodic wave
with a more general type of equiphase surface. Because each component
\vave satisfies 31ax~ve11sequations, the resultant obtained by superposi-
tion like~vise satisfies the field equations.
Cylindrical Waoes.-Circularly symmetrical cylindrical waves are the
l~mentary forms of two-dimensional propagation. The equiphase sur-
faces ~f these waves are coaxial circular cylinders; the wave is propagated
along the radii of the phase surfaces.
Cylindrical coordinates, as defined in Fig. 32, are appropriate for the
analysis. The z-axis is taken as the axis of symmetry, and r and 0 are
polar coordinates in a plane normal to the z-axis. At each point we
define unit vectors i,, i~, i, in the direction of increasing r, 8, and z,
respecti\-ely; the field vectors may, cm occasion, be resolved into com-
ponents in these directions.
}Ve shall no}v seek sol~ltions of the field equations in Ivhich the field
i.ectors are everywhere tangential to the cylindrical cquiphase surfaces
and have constant amplit~l(lc o~cr c:wh slich surface (that is, t}le ampli-
76 RA I) IATION FROM C(JRhWNT IIISTRIBCTIO.VS [SEC. 3,7

tudes are functions of r only). We shall seek solutions of two different,

z types, distinguished ~vith reference to the
1 directions of the field vectors:
\
\ Case a:
/ Hz = O, E,(r) # o.
I Case b:
p, Iiz
E, = O, Hz(r) # o.
1 In each solution, -of course, H, = E, = O.
we begin by determining the form of
~Y the z-component of the field as a solution
of the ~vave equation; later we shall deter-
mine the remaining field components by
means of the field equations.
x Since E,(r) and H.(r) are components
l~lt;. :12-Cylindrical coordinates.
of the res~ective field vectors in a rectan-
gular coordinate system, they must satisfy the scalar Helmholtz equation
[Iiq. (61 )], In cylindrical coordinates this becomes

(70)

\vhere~ may r~~presenteither h, or H,. since ~ is independent of band

z, this red~~resto

(71 )

On introduction of
f = kr, (72)
this becumes

(73)

This is the differential e(~uation satisfied by the Bessel functions or, more
generally speaking, by the cylinder functions or order zero.1 of the
many solutions of this equation Irhich \ve might identify \vith the func-
tions E, or H,, those of immediate interest here are the Hankel functions
Hi) (g) and H~2)(~). The nature of these functions is most evident in
their asymptotic behavior for large values of ~ = Iir:

I
 .

AEC. 3.7] SIMPLE WAVE SOLUTIONS 77

The second of these functions, multiplied by the time factor w, repre-

sents a wave traveling in the positi~e r-direction; the phase quantity
tit kr is the analogue of the quantity tit k.r for a plane wave traveling
in the positive x-direction. Thus H~z)(kr) represents a cylindrical tvavc
diverging from a line source on the z-axis. Similarly, 11~)(h) represents
a wave traveling in the negative r--direction, that is, a ~vave from infinity
converging to a line focus along the z-axis.
Restricting attention to the diverging ~vavc function H~2)(kr), ~~e
consider first Case a. We assume
E = H~2)(kr)@wi. (75)

and use the field equations to determine the associated magnetic tield.
The curl of a vector P, expressed in cylindrical coordinates, is

p=(+%-%)i+(+-% )i+:[l(rp@)-%li ()
Taking the cud of the vector E and making use of Eq. (32a), ~veobtain

H= -i[:H)kr)l[i,
(77)

It is left to the reader to verify that the field vectors E and H defined
by Eqs. (75) and (77) satisfy the other field equations. over the cylin-
drical surfaces of constant r, E and H are perpendicular to each othe[-
at every point and lie in the tangent plane to the surface; as in the case
of a plane ~vave, E and H are normal to the direction of propagation, and
the Poynting vector is normal to the equiphase surface. As the radius
of the equiphase surface becomes large, it becomes sensibly plane in
the neighborhood of any point. We must, therefore, expect that as
r + cc , the relationship betlveen E and H approaches that existing in a
plane ~vave. The asymptotic form for H may })e obtained l)v int rt)duc-
ing l?q. (74b) into F,q. (77). Aeglecting terms of higher order in 1/r,
}re find

(78)

whence

(79)

Thus, in the limit as r ~ m,

(80\

as was to be expected.
78 l<A1)lA I10A FROM CURRENT I)ISTRIBU1710NS [SEC, 37

The derivation of the field for Case b proceeds in a similar manner.

We assume
H = H$) (kr)e,u~iZ; (81)

the associated electric fiel(i follows by application of Eq. (32b), which, in

the case at hand, becomes

E=3~6vx H. (82)

It follows that

E i% Hd~-
(83)

The general remarks concerning Case a apply to the present case also.
It is easily verified that here too the relationships approach those in a
plane ~vave as the radius of the cylindrical phase surface becomes ~ery
large; that is,

lim H = ~ (i, xE). (84)

,+ . J
We have thus obtained t~vo independent field distributions ]~ith
cylindrical equiphase surfaces. Ire shall refer to these as cylindrical
modes of free-space propagation. The first field, Case a, can arise from
a linear distribution of electric current along the z-axis and \villbe spoken
of as a field of the electric type; Case b can be associated ~vith a linear
distribution of magnetic current along the z-axis and is correspondingly
referred to as a field of the magnetic type. These are the simplest cylin-
drical modes of free-space propagation. A treatment of the general
theory of cylindrical waves will be found in Stratton. 1
Isotropic Spherical Waws.-Next ~ve shall consider the isotropic
spherical wave with equiphase ~urfaces that are concentric spheres and
field amplitudes that are constant in magnitude over each equiphasc
surface.
The spherical coordinates r, 0, and O, illustrated in Fig. 33, arc
appropriate for this discussion. With the spherical coordinates are
associated a set of orthogonal unit vectors i,, i~, id at each point in space,
in the directions of increasing r, 9, and @ respectively.
Let the center of the family of equiphase spheres be at the origin (f
the coordinate system. An attempt to conetruct a tielcl that is a func-
tion of r alone, as in the case of cylind ical \vaves, Ivill fail. For example,
suppose that Ire try to construct a field in ~~hich the field vectors have
only the components
E = E(r)ie,
(85)
H ==H(r)iO. }
1J. .4. Stratton, A/rcf~ov[flgnetfc
Ihrorjy, l~c(jraw-lllll, Xcw lork, 1941,C}\:Ip\],
 .

SEC. 37] SIMPLE WAVE SOL UTIOAIS 79

It will be seen that there is an essential ambiguity in the directions of these

vectors at all points for which 19= O or ~. The ambiguity can be ren-
dered trivial only by making the magnitudes of the fields vanish for
@ = O and m; the field can then be independent of d only if it vanishes
identically.
The isotropic spherical wave is, in general, a possible waveform only
for scalar fields such as are encountered in acoustics. However, it is
often useful for reference and comparison with electromagnetic waves.
z

/ \
/ \
/

z
FIG.3.3,Sphericalrmrdinates.

Accordingly we shall note briefly the spherically symmetrical solutions

of the scalar Helmholtz equation [Eq. (61)]. In spherical coordinates,
the Laplacian V2 is
dz

()
72=;2: ~z; +
la
sin 0 & + ~~ (86)
rz sin 0 --(
do ) r- smz e W

When ~ is a function of r only, the Helmholtz equation becomes

(87)

It is readily verified that

(88)

are solutions of this equation. The solution $_, multiplied by the time
factor @, represents a wave diverging from a source at the origin, while
++e~~ represents a spherical wave converging to a point foc{ls at the
origin.
80 RADIATION FROM CURRENT DISTR1BUTIONS [SEC, 343

3.8. General Solution of the Field Equations in Terms of the Sources,

for a Time-periodic Fieid.The plane and cylindrical waves discussed in
the preceding section are solutions of the homogeneous field equations
which apply in regions of space free from charge and current distributions.
In deriving the form of these fields, no attention was paid to their ulti-
mate sources, which lay outside the domain of validity of the solution.
Our present task is the more exacting one of determining what fields
will arise from a prescribed set of
sources in a homogeneous medium.
For reference the complete set of
field equations is repeated here.
Magnetic charge and current distri-
butions are included for later use.

V X E + jawH = J., (332a)

VXHjucE=J, (332b)
v. H=& (332c)

v. E=~, (332d)
E
V. J+@P=O, (332e)
FIG. 3.4.Notation for Greens theorem. V. Jm+jup. =O, (332f)

also the pair of vector Helmholtz equations,

VxVx Ek2E=jupJ V xJm, (89)
VxVx Hk2H=ja,Jm+V xJ. (90)

The integration of these equations is based on a vector Greens

theorem : Consider the region V, illustrated in Fig. 3.4, bounded by the
surfaces S1, . . , h~. Let F and G be two vector functions of position
in this region, each continuous and having continuous first and second
derivatives everywhere within V and on the boundary surfaces. Then,
if n is the unit vector normal to a bounding surface, directed into the
region 1,

(F. vxvx GG. vxvx F)rh

/ v
. (Gx VxF FxVx G). nclS. (91)
/ S,+s, +... i-s.
As indicated, the surface integral extends over all boundary surfaces.
Let us suppose that there exists in a volume V, such as that consid-
ered above, an electromagnetic field such that E and H meet the condi-
I The procedureadopted here is due to J. A. Stratton and L. J. Chu, Phys. Rev.,
66, 99 (1939). A proof of the Greens theorem is givwr in this paper.
sm. WI G17NfilML SOLUTION (M !lHJ? 1IJ2LD MJUA!S1ONS 81

tions of continuity required of the vector function F of the Greens

theorem. lVe shall no}~ see, ~vith the aid of this theorem, ho~v one can
express the field at an arbitrary point P in the volume V in terms of the
field sources \vithin this volume and the values of the field itself over the
boundaries of the region.
Wedefine the vectorf unction of position

G=e~a=+a, (92)

\vhere r is the clistancc from Z to any other point h the region and a is
an arbitrary but otherlvise constant vector. This \rill satisfy the con-
tinuity conditions required of the function G in the C,rcens theorem
everywhere, except at P, where it has a sing~darity. Accordingly, Ire
surround P by a sphere > of radius rO and consider that portion V of V
which is Imunded by the surfaces S,, . . . , S. and Z; in this restricted
region, G as defined by l;q. (92) and F = E of the electromagnetic field
satisfy the conditions required for application of the Greens theorem.
\Ve have then

(#a. VxVx EE. VxVx#a)d.

/ v

.4s the first step in the maniplllat,ion of this r(l~ahtion, II-e shall trans-
form the volllme integral involving the electric ficl(l into an equivalent
integral involving only the field sources. Intrwl(lction of the vector +a
into the vector identity [Erl. (W)] and lwc of ihc facts that ~ satisfies
the scalar Helmholtz equation and a is a constant vector \vill suffice to
sho\v that

Takiug this in conjlluctitjn itith I;([. (8!)), jlc t)l)t:lin

A fe\v additional transformations me nccrssary:

E. V(a. V+)= V.[E(a. V~)] (a. v~)v.E (96)

=V. [E(a. V~)]~a. V+,

and
82 RAI)IA7ION PROM CURXi?N17 DISTRIBUTIO.VA [SEC. 38

By use of these, l;q. (93) can be given the desired form:

a.
/( v $
+~xv+-:v$)dv+a~vv$
.
+ ~, V . [E(a . V+)] dv
\
. [(Ex Vx+a). n(+ax VxE). n]dS. (98)
/ S,+. .+x
We can now bring each term in Eq. (98) into the form of a scalar
product with the vector a and then completely eliminate this vector
from the problem. The second and third volume integrals can be trans-
formed into surface integrals:

a. Vx#J. dv=a. @ x J~ dS, (99)

/ v 1 St+. ..+z

V . [E(a . V*)] dv = (n. E) (a . V+) dS

I v 1 S,+... +x
r
a . (n. E) Vi dS. (100)
1 S,+. ..+z

To the surface integrals cm the right-hand side of Eq. (W) \veapply the
following transformations:

[Ex(Vx+a)l .n=[Ex(V+ xa)]. n= [(n xE)x V4]. a, (101)

+(ax VxE). n=.@jqb(ax H). n ~(ax JJ. n
= jwp+a o (n xH) + ~a. (n xJJ. (102)

Collecting these results, we obtain finally

a. ju~+ + m X V+ : v+ dv
Nv )

=a. s,+, +,, [~~~~(n XW -1- (n xE) x V# + (n. E) v+]ds. (103)

/
Since Eq. (103) must hold for every vector a, the integrals themselves
must be equal. That is,

,[jw#(nxH) + (n xE) xvi+ (n. E) VY]dS

/

H v
jG.wtJ + J~ x v+ ~ W
)/ dv s,+,,

+ (n x E) x V+ + (n. E) V+] dS,

+s,, [j~~$(n x H)

(104)

where for convenience ~re have split off the integral over the sphere 2.
In the limit as z shrinks down on P, this integral will depend only on the
 .

S~r. 3 8] (7E.VERAI. ,$OI,[JTIO,V OF TIIE FI171.D EQIJ.4TIOAS 83

field at P. Thus we have a relation bet\veen the field at P and a volume

integral over the sources of the field, plus surface integrals involving the
fiqld itself,
Next let us consider the integral over 2. On the surface of this
sphere we have

The normal n is directed along the radius mlt from P, Jet dfl be the
solid angle slll)ten(led at P I)y an element of sllrface dS on 2; the surface
integral can then be written

[ ] dti = j-Oe-~~ {mP(n x H) + Ii[(n x E) x n + (n o E)n]} dfl

\x / >

e-ik,

1 x
[(nx

j4irTof?-ik0
E)xn+(n.

(wn
E)n]dQ

x H + k%) 4me-ikroE, (106)

where the overline denotes the mean value of the function over the sur-
face of the sphere. If now w-e let the sphere shrink to zero, the term
containing r-nvanishes because by hypothesis the fieid vectors are finite
in the neighborhood of P. At the same time ~ approaches E,, the value
of the field vector at P. Thus

Iim [ ] d,q = 47TEP. (107)

To-o \ ~
In this limit the region T comes to inclllde the whole of the region T7,
and Eq. (104) becomes

The analysis follows the same course for the magnetic vector H, with
the corresponding result:

The fields at the observation point, P have thlls been expressed as the
s~im of contrib~ltions from the s(mrces distribllted through the region T
 .

84 RADIATION FROM CIJRRENT DISTRIBUTIONS [SEC. 39

and from fields existing on the bounding surfaces. These latter surface
integrals represent contributions to the field from sources lying outside V;
specifically, the slu-face integral over a surface S, enclosing an exterior
volume V, represents the effect of sources within Vi.
Each of the three terms in the surface integral can be correlated with
a corresponding term in the volume integral according to the way in
which the function
~jk,
+= (110)
r

is involved. In Eq. (108), for example, (n x H), (n x E), and (n o E)

enter the surface integral exactly as the electric-current density J, mag-
netic-current density J,., and the charge density p, respectively, enter the
volume integral; a similar correspondence will be observed in Eq. (109).
Thus the effects of sources lying in an exterior region Vi, bounded by
the surface St, are represented formally as arising from a surface distribu-
tion of charges and currents on the boundary St, with surface densities

K=(nx H),
Km = (n xE),
(111)
~ = ~(n . E),
q~ = p(n H), 1

E and H being the fields existing over that surface.

3.9. Field Due to Sources in an Unbounded Region.We have now to
consider the case in ]vhich the region V is unbounded and the sources
of the field are confined to a region of finite extent. There is then only
one boundary surface S,,, which we shall at first take to be a sphere of
large radius R about the point P, enclosing all sources of the field. 13qua-
tions (108) and (109) then reduce to a single surface integral over this
large sphere S(R).
Let R, be a unit vector directed out along the radius of this sphere;
that is, let R, = n. On introduction of this vector and the explicit
form of +, the sllrfare integral of Eq. (108) beromes

&
1
,(,)
[~w+(n

1
x H) + (n x E) xv+ + (n. E) v+] CM

jcop(R, x H) jk + ~ [R, x (R, x E)
47r ,~(~) { ()

-/

(R1 . E) R,]
1:;:ds = Him + {4RXH)+(YEI
@kR
~

I
E
~ ~ &S. (112)

I
SEC.3.9] FIELD DUE 70 SOURCES IN AN UNBOUNDED REGION 85

If we now let the radius R become infinite, the surface of the sphere
increases as R. The surface integral will vanish as R ~ cc if the fields
satisfy the conditions
lim RE is finite, (l13a)
R+ m
)!
lim R
R+ . [
(R, xH)+
01
~ E =0. (113 Z))

In the case of Eq. (109), the surface integral \vill vanish if

lim RH is finite, (113C)
}+Ii+ m

lim
R+ .
R
[() c
;
(R, xE) H
1
=0. (113d)

Conditions (113a) and (113c) require that at large distances from the
sources, the magnitudes of the field vectors decrease at least as rapidly
as R-. Conditions (113b) and (113d), the so-called radiation condi-
tions, ensure that all radiation across the bounding sphere consist of
waves diverging to infinity. This may be seen as follows: Taking the
scalar product of Eq. (113b) with Rl, we obtain
lim (RE) . R, = O. (114)
E- cc

The component of E in the direction R, thus diminishes more rapidly

than R; we may say that E is perpendicular to R,, to terms of the order
of R-I. On the other hand, Eq. (113d) states that

(115)

It follows that to terms of the order of R-, H is perpendicular to both

E and RI and E and H are related in the same manner as in a plane wave
progressing away from the center of the sphere S(R).
If (as will be shown in Sec. 3.11 to be the case) the fields arising from
sources confined to a finite region of space satisfy ~qs. (113) at infinity,
then the surface integrals over the infinite spheres vanish and the field
vectors in the unbounded region are given by

The fields are expressed here entirely in terms of the sources.

These fields can be expressed in terms of the current distributions
alone by llse of the eql[ations of continuity [Eqs. (32c) and (32j)], which
 .

86 RADIATION FROM CURRENT DIS1RIBUTIONS [SEC.310

relate the charge densities to the current distributions, Thus Eq. (l16a)
becomes

pk,
Ep=~ k2J ~ + (V .J)ve; +j.cJ. X V ~ dv, (117)
47r6x
/[ v r ( )1

Let i., a = 1, 2, 3 be unit vectors in the z-, y-, and z-directions, respec-
tively. Then

(l18a)

(1186)

fVR)v(J~%)dv=-/.)nJ
1)
~)ds=O=O
as soon as R is taken so large that S(R) lies outside the region to which the
current distribution is confined. It follows that the first terms on the
right of Eq. (118b) contribute nothing to the integral in Eq. (1 17). Thus
we obtain

E,= r::
/v [(J )V + kJ @cJm X V] ~ dv. (120)

Similarly, for the magnetic field \veobtain

H,= &p ~ [(J. o )V + k2Jm+ j.wJ X V] ; dV. (121)

/

3S10. Field in a Region Bounded by Surfaces of Infinitely Conductive

Media.A second case of importance is that in which the region V is
bounded by surfaces Si which are the surfaces of bodies of infinite con-
ductivity and by the surface S~ at infinity. We again assume that the
fields at infinity satisfy the condition of Eq. (1 13). The integrals over
S-in Eqs. (108) and (109) then vanish, and we have to consider only the
integrals over the surfates of the conductors. At the surface of an
infinitely conducting body the boundary conditions of Sec. 3.3 are

nxE=O, n. H=O,
n.E=q (122)
nxH=K,
, 1
~ and K being the surface distributions of electric charge and currrnt
Thus Eqs, (108) and (109) become
SEC. 3.11] THE FAR-ZONE FIELDS 87

E,= &v
H jcqJJ-:V+J.x V
)
~dv

+:/s.
vmds24)
It will be observed that the expressions for the fields due to surface
currents and charges could have been obtained from the volume integrals
as limiting forms, on considering that the volume distribution passes
into a surface-layer distribution.
The results of this section will form the basis for the general theory
of reflectors to be developed in Chap. 5.
3.11. The Far-zone Fields.Let us now return to the case of the
unbounded region and examine in more detail the relations between the
field solutions

Ep= &e ~ [(J - V)v + k2J j.MJm XV] ~ dv, (3.120)

/

H,= + ~ [(J. - V)v + kJ. +ALJ X v]~dv (3121)

/

and the radiation conditions developed in Sec. 39.

These solutions are based on the assumption that the sources are
confined to a finite region of space. Let us choose an origin in the neigh-
borhood of these sources, and let p be the vector from the origin to the
source element at the point z, y, z (Fig. 3.5). The vector from the origin
to the field point P we shall write as RRI, RI being a unit vector; similarly,
rrl will be the vector from the source element to the point P.
In the integrands of Eqs. (120) and (121), the operator V acts on the
coordinates of the source element, whereas the point P is treated as a
fixed origin. For example,

and
(3 (+:)=1 (125)

(Jvvta=[- (-l)(JrJr k(J. r,)r, + ~ jh + ;

-J ;jk+F 1 :. (126)
( )1

Thus the integranrls in these eqllations are power series in r-; for the
88 RADIATION FROM CURRENT DISTRIBUTIONS [SEC. 311

first-degree terms in Eq. (120) we have

In evaluating these integra!s we must take into account the variation of

T and of the unit vector rl with the position of the source element.
In general this offers serious diffi-
culties, but simplifications can be
effected if the field point is at a
very great distance from the cur-
rent distribution and the origin.
First, the angle between the vet- .
(a) tors rl and RI, which decreases
/
x with Rl, can be neglected; rl can
be replaced by R, in the integrals.
Next, the factor r- in the inte-
grand can be replaced by the con-
stant Rl, from which it differs by
the terms of the second order in
R1. The variation of r cannot
be neglected wholly in the phase
factor. Here, making use of the
x fact that r, and R, are effectively
FIG. 3.5.On the far-zone field: (a) arb,- parallel, we write
trary field pojnt P; (b) simplifying x-alation-
shlps for a point in the far zone, r= Re. Rl. (127)

With these approximations, Eqs. (120) and (121) take on forms valid
for the far-zone fields:

.-v
~, =
4.R
@R
1[
~ J(J. RJRI+
(YJ-Rle+ikeR
+O();2, (128)

and

H, = ~~ e-j~ ~ Jm (Jm . R,)R,

/[ (YJxRleikRdv

+ O & (129)
()

The calculation of the terms of order R-2 is tedious but straightforward

and will be left to the reader.
The integrals in Eqs. (128) and (129) are independent of r. Thus it
is evident that REP and RHP remain finite as R + m, as required by
SEC. 311] THE PA R-ZONE FIELDS 89

the boundary conditions [Eqs. (113a) and (113c) ]. It is further evident

that the field vectors are transverse to the unit vector R,; the J term in
the integrand has a component in the direction of Rl, but this is always
canceled by subtraction of the second term (J o RJR1. A simple cal-
culation shows that the radiation conditions [Eqs. (113b) and (113d)]
are satisfied; for example,

rhus, E and Hare related as in a plane wave, being mutually perpendicular

and in a plane normal to R1.
We must now examine the integrals of Eqs, (128) and (129) in a little
more detail. We introduce the system of spherical coordinates R, d,
+, defined in Fig. 3.5, with polar axis along the z-axis. Let iOand i+
be unit vectors having the directions of increasing 0 and @ at the point
P; RI is, of course, the unit vector in the radial direction. In terms of
Cartesian components

p = xi. + Vi, + ziZ, (131)

RI = sin o cos @i, + sin o sin @iV+ cos Oiz; (132)
thus
@. Rl=(zcos @+ysin@)sin O +ZCOSO. (133)

The components of the electric field vector along i~ and it are easily
found to be

= - *e-ik/v[J+t) J~i+l+keRk~~
~~ e-ik~ F,(O,1$) (13k)
47rR
and

As indicated, the integrals are functions of only the angular coordinates

0 and & The components of the electric field and the resultant far-
field vector have the form to be expected for a source located at the
origin. However, the far field is only a quasi-point-source field; the
equiphase surfaces are not the family of spheres of constant R because
the space factors F, and FZ are in general complex. This is to be expected
because the choice of origin was purely arbitrary.
90 RADIATION FROM CURRENT DISTRIBUTIONS [SEC. 311

The point-source character of the far field becomes more evident on

considering the power flow in the far zone. The Poynting vector is

S=~Re(Ex H*)=~
()
lCYJ
~ +Re[Ex(R, xE*)]

2L () (IE,[ + IEJ)R, (135)

or

(136)

where
l(e,@) = IF1(8,4)12 + ]F2(e,@)]. (137)
The power flow is radially outward from the origin, with an intensity of
flow that falls off with the square of R and depends also upon 8 and ~;
with respect to power flow the current distribution is, in effect, a directive
point source at the origin.
In discussing the power flow it is convenient to use, instead of the
Poynting vector, the power P(L9,c$) radiated per unit solid angle in the
direction O, 0. This is given by

(138)

whiph is independent of the radial distance R. The angular distribution

of the power flow may be represented graphically by a three-dimensional
plot in spherical coordinates, in which the angular coordinates o and @
are those of the direction of observation and the radial coordinate is
proportional to P(19,~). It is customary to normalize the maximum of
the power pattern to unity. The resulting figure is spoken of as the
polar diagram or radiation pattern of the current distribution.
The power distribution is also specified in terms of a gain function
G(8, +) with respect to an isotropic radiator, as defined in Chap. 1,; in
terms of P(o, o) we have

G(e,l#l) =
1
,2. . (04)
F(L9,1#J)
sin OdO do
Goo//
%k(e,f$)
. (139)
2= x

H
00
Y(f?,@) sin @de d+

The maximum value of the gain function is termed the absolute gain.
In design specifications this is generally quoted in decibels above the gain
of an isotropic radiator (which is unity):

Gain in db = 10 log,, [G(6,@)]_. (140)

 .

SEC. 3.12] POLARIZATION 91

3.12. Polarization.In the preceding section we have considered the

separate components Eo and E4 of the electric field vector in the far zone;
we have now to note some properties of the resultant field vector.
The factors Fl(o, I#J)and Ft(dj O) in the expressions for Eo and E+ are
in general complex quantities, which we may write thus:

Here the As and ~s are real, and YI and -yl are in general not equal.
The vector E, is thus the resu!tant of a pair of time-periodic vectors
E,i8 and E@i@at right angles to each other, with relative amplitude and

L- 2., A (b) (c)

(a)
FIG. 3.6.Elliptical polarization: (a) orientation of the ellipse; (b) right-handed polariza-
tion; (c) left-handed polarization, with direction of propagation toward the reader.

phase which vary with .9 and @ This resultant vector E, simultane-

ously rotates in space and varies in magnitude in such a way that its
terminal point describes an ellipse; the radiation field is elliptically
polarized. To show this we note that the real parts of E8e+~Wand
E$e~o, as given by Eq. (134), are the real Ed- and Et-components of the
electric field. These become, on use of Eqs. (141),

W4 1(8,0)
EO = sin (tit kR -yl) = cw sin (d kR TJ, (142a)
4.R
~ = @w42(d, @)
* sin (cd kR ~.J = a+ sin (cd kR T1 6), (142b)
4.R

where 6 = 72 -y, is the phase of E+ with respect to EO. Expanding

the sine term in E+ and eliminating the terms involving d kR -yI,
we obtain a relation bet]veen E@ and E~ that holds at all times:

(143)

This is the equation of an ellipse traced out by the terminus of the

vector E~. The relation of the ellipse to the component vectors is shown
in Fig. 3.6. The sense of polarization is defined for an observer watching
the oncoming wave: The polarization is termed right-handed or left-
92 RADIATION FROM CURRENT DISTRIBUTIONS [SEC. 3.13

handed according as the terminus of the vector E, traces out the ellipse
in the clockwise or counterclockwise sense, respectively.
If the phase difference 6 is an odd multiple of 7r/2 and the amplitudes
are equal, the ellipse becomes a circle; right-handed and left-handed
circular polarization are defined in the same manner as for elliptical
polarization. If the phase 6 is an integral multiple of m, the ellipse degen-
erates into a straight line traced out by a linearly polarized resultant.
As o and @ are varied, both 6(0, O) and EJE$ will vary; the polariza-
tion of the radiation from an extended source may change from linear
to elliptical to circular and back again as one changes the direction of the
observation.
3.13. The Electric Dipole. -In the preceding sections we have seen
how a radiation field arises from a
z
distribution of time-varying currents.
We now turn to a discussion of some
special idealized current distribu-
~ q yt
M 1 tions and their associated electromag-
k L-q netic fields, leaving aside the question
ni of their physical realizability.
k
x The most elementary form of
(a) (b)
idealized radiator is the oscillating
FIG. 3.7.The electric dipole: (a)
mathematicaldipole; (b) antennarePre- electric dipole (Fig. 3.7). A dipole
sentation of a dipole, 1 < < h. consists mathematically of a pair of
equal and opposite charges, each of magnitude q, separated by an infi-
nitesimal distance ~. If the vector 5 is directed from q to +q, the
dipole moment of the dipole is defined to be the vector

p = qb. (144)

An antenna equivalent to a dipole is shown in Fig. 3.7. It consists

of thin wires terminated in small spheres, the over-all dimensions of the
structure being very small compared with a wavelength. The spheres
form the capacitive element of the structure, and the charge at any
instant can be considered to be localized on them. If the antenna is
energized by a harmonic emf applied across the gap at the center, the
charges on the spheres are given by
q = q&jul;
(145)

the magnitude of the dipole moment of the antenna is

p = qoleid = poe~uf) (146)
with amplitude
po = qol.

Since 1 <<h, the current at any instant, may be taken to be the same at
SEC. 3.13] THE ELECTRIC DIPOLE 93

all points along the wings of the antenna. The current 1 is related to
the charge q by I = dq/dt = juq and to the magnitude of the dipole
moment by

(147)

The electromagnetic field set up by a dipole is best described in

spherical coordinates with the origin at the center and the polar axis
along the axis of the dipole (Fig. 3.8). The derivation of the field will
be found in any text on electromagnetic theory; we shall simply state
the results:

Asa consequence of theaxial symmetry of the radiator, the field isinde-

pendent of O. It can be resolved into three partial fields according to
the dependence on r: (1) the
static field varying inversely
with r3, (2) the induction field
varying inversely withrz, and (3)
the radiation field varying in-
versely with r. The static field
is, in fact, that which would be
computed for a static dipole with
fixed moment p~e~f-~rj. The in-
duction field is the quasi-station-
ary-state field commonly observed
in the neighborhood of circuit ele-
ments at low frequencies; the
magnetic component of the induc- 1Ic;.:3-8.1ield of an electric dipole oriented
along the z-axis.
tion field is that which would be
calculated on the basis of the Biot-Savart la~v for stationary currents.
At small distances from the dipole the static and induction fields predom-
inate. At a distance,
1A
r>=,
k 27r
the radiation field becomes the leading term, and at sufficiently large
distances the static and induction fields become negligible relative to
For exarnplc,,1..1. Stratton, ElectromagneticTheory, McGraw-Hill, New York,
1941, Chap. VIII,
94 RADIATION FROM CURRENT DISTRIBUTIONS [SEC. 3.13

the radiation field. However, It is only at distances much greater than

r = A/27r that one can entirely neglect the static and induction fields.
The radiation field represents a flow of energy away from the dipole.
There is no corresponding energy loss in the static and induction fields;
the energy associated with these fields pulsates periodically back and forth
between space and the antenna and its associated circuit just as do the
energies in capacitances and inductances at low frequencies. The far-
field Poynting
. vector computed by Eq. (43) arises entirely from the r-l
70 30@ w K@ lllY terms in the fields. - It is

s= ~~ lpol~ i,, (149)

where i, is the unit vector in the

120 outward radial direction.
The dipole is a true point
source because the equiphase sur-
130
faces are spheres with centers at
l~o the origin; it is directive because
the intensity of the field varies
with the direction of observation.
lW The power pattern of the dipole is
independent of azimuth o and is
1s0 sufficiently represented by a cut
FIG. 3.9.Meridional polar diagram of the in any one meridian plane, like
powerpattern of an electric dipole. that shown in Fig. 3.9. In design
specifications it is customary to characterize such cuts in the
t~ee-dimensional polar diagram by two widths if they exist: (1) the
half-power width Q, which is the full angle in that cut between
the two directions in which the power radiated is one-half the maximum
value, and (2) the tenth-power width @ (i%), the angle between the
directions in which the power radiated is one-tenth of the maximum.
These widths for the meridional polar diagram of an electric dipole are

@ = 90,
@(~) = 146.

Since the pattern is uniform in azimuth, the polar diagram in a cut taken
normally to the dipole axis is a circle. The gain function of the dipole
[Eq. (139)] is
G(o, o) = ~ sinz 0, (150)

and the absolute gain is

G~ = (:) = 1.76 db. (1,51)

SEC. 314] THE MAGNETIC DIPOLE 95

The impedance presented by the dipole to its feed line consists of a

resistive component and a reactive component. We shall here consider
only the resistive component, which corresponds to the power dissi-
pated by the dipole. There are two elements in the power dissipation: (1)
ohmic losses in the conductors of the dipole structure and (2) power
radiated to space. In the idealized case, to which we restrict ourseives,
the dipole consists of perfectly conducting elements. There is then only
radiation loss to consider; the resistive component of the impedance is its
radiation resistance. Let P be the average power radiated per unit time.
The radiation resistance Rr is defined by
~ = ~11,12R,, (152)

where 10 is the maximum value of the current. The radiated power ~

is computed by integrating the Poynting vector [Eq. (149)] over a com-
plete sphere. By use of Eqs. (147) and (152) the radiation resistance is
then found to be

(153)

3.14. The Magnetic Dipole.The magnetic counterpart of the electric

dipole antenna is a current loop with radius
small compared with the wavelength (Fig. Z
3.10). Such a current loop is equivalent to a
t
magnetic dipole along the axis normal to the
plane of the loop; this axis has been taken to
be the z-axis in the figure. If 1 is the current
in the loop and A is a vector normal to the
locp, with magnitude equai to its area, the
magnetic moment at any instant is
FIG. 3.10.Magnetic di-
m = IA. (154) pole antenna: cu:rent loop and
equivalent magnetic dipole,
If 10 is the amplitude of the time-periodic cur-
rent and m~ the corresponding amplitude of the magnetic moment, the
magnitude of the magnetic moment is given by

m = IOAe@ = mOejut. (155)

The direction of the dipole in relation to the direction of the current is

shown in Fig. 3.10.
The field of the magnetic i oole, like that of the electric dipole, is
most conveniently described in spherical coordinates. The field com-
ponents are

+ H9Y:-iwmoe(-k) (156a)
96 RADIATION FROM C1RRE.VT D1,YTIUII 1I1O.VS [sEC. 315

As with the electric dipole, the field is independent of the azimuth

angle O. Comparison of Eqs. (15(;) \rith the electric dipole field [I;qs.
(148)] will show that the roles of E and H are interchanged. lVith minor
revisions required by this interchange, the discussion of the electric
dipole as a directive point source can be carried o~-er to the magnetic
dipole. The power patterns are identical, and the absol~ltc gain of the
magnetic dipole, like that of the electric dipole, is 1.76 db. The radiation
resistance of the loop is found to be

(157)

The reader shoulcl note that the far-zone fields of the electrir and
magnetic dipoles sho!v the general properties mentioned in Sec. 3 ] ].
In particular, he should note that in the far zone (and there only)
1,

H=
()~ (i, xE); (158)

E and H are mutually perpendicular and lie in a plane transverse to the

direction of propagation.
z 3.15. The Far-zone Fields of Line-
A
(
current Distributions. ~Ye shall next,
o comp{ltc the far-zone fields due to a linlr-
pcriodic c~lrrent in a thin straight \\ire
\\ estending along the z-axis from z = 1,2
to z = +1/2, that is, along the polar axis
of the r, 6, @ coordinate s~stem, .lVc
shall allo~v the length of the I\-ireto be
comp:lrable to a wavelength or even rq(uil
x
to a number of ~.wavelengths, The pllasc
I ditferenccs lJetJ1-eenthe mlrrents :~t sepa-
ILLtN~points on the \\ire !I-ill then lje sig-
nificant, and \\-eshall need to consider tile
current to be a function of position :alc~ngthe \~ire:

I = I(z)rji., (159)

Since the properties of the ficlcl in the far zone arc those c)f 3 pl:iu]e
wave, it will be sllfficic,nt to calc~llate the elr(,tric fic,l(l intclbi(l-. In
Eqs. (134) ,ve can first of all discard the ,~l:~grletil-,llll(llt clensity J,,,.
 SEC.3.15] LINE-C [JRRENT DISTRIBUTIONS 97

We note further that J . i+ is zero; consequently,

E+ = O. (160)

By Eq. (133), with z = v = O, p . RI is simply 2 cos 0, and the volume

integral for Ee degenerates into a line integral:
1/2
Eo = ~ e-i l(z)iZ . i~e~~a
m8dz, (161a)
/ -1/2

or

(161b)

(As usual, the time factor e~ is understood implicitly.)

Again, because of the axial symmetry of the radiator, the field is
independent of the azimuth angle O. As with an electric dipole, the
electric-field vector lies in the meridian plane; the magnetic-field vector
is at right angles to this, parallel to io. The function F(tI), known as the
form factor of the field pattern, will in general be complex; the equi-
phase surfaces are not spheres of constant R.
The integral expression in Eq. (161 b) admits of an interesting inter-
pretation. On comparing the integrand with the far field of an electric
dipole [Eq. (148)] it will be noted that the integral can be interpreted
as a sum of the fields of a distribution of dipoles along the wire, the dipole
moment dp associated with the element of conductor dz at the point z
(Fig. 3.11) being given by

(162)

In superposing the component fields at the field point one must, of course,
take account of the phase differences bet\veen the contributions from
different dipole elements, due to the differences in path length to the
field point. If A is the path difference between two elements, the phase
difference is 2rA/A = kA. Taking the origin as a reference point for
path length, the path difference corresponding to a point z on the wire
is A(Z) = z cos 8; hence the phase factor e~~O6in the integrand. It will
be noted that Eq. (162) is essentially the relation between the current
and dipole moment set down in Eq. (147).
The precise form of the current function 1(z) can be controlled by
changing the point at v-hich the driving voltage is applied to the wire
and the way in which the wire is terminated. We shall now consider
the case in which the wire is driven at the center, for example, by a
98 RADIATION FROM CURRENT DISTRIBUTIONS [SEC.3.16

parallel-wire line feeding across a small gap at the origin, and there is
no load at the ends of the \vire. In this case the current is necessarily
zero at the ends of the wire; its distribution along the ~vire can always
be expressed as a sum of standing \vaves, each of which vanishes at the
ends. Such standing waves have the form

I*(2) = I,(m) Cos Uyj m=l,3,5, ...,

(163)
1.,(2) = 10(nr) sin fl~~, m = 2,4, 6,...,
I

where 10(m) is the value of the current at a cur-rent antinode. In general

the current will consist of a sllpcrposition of standing waves. It \rill,
however, consist of a single standing tvave if 1 = m~2; this is the case
which we shall treat. Substituting the corresponding I(z) into Eq. (161 b),
one finds with little difficulty

35Io(m) sin ~
E, = j
()
~ -
27rR
,_ik[cos@&],

m = 1, 3, 5, - . . , (164a)

w~=2,4,.... (164b)

The term form factor is here applied to the terms in brackets. The
surfaces of constant R are equiphase surfaces; the far-zone field of a
standing-~va~e current is that of a true point sourer at the center of the
current distribution. The field intensity in the eqlmtt)rial plane 0 = 7r/2
is zero when m is an even integer bcca{lse the c~urent distribution is
antisymmetrical with respect to the origin; the contributions to the fielcl
from current elements at +Z and z are 180 out of phase at points in the
equatorial plane and there annul each other.
3.16. The Half-wave Dipole.- the most important line-current
distribution in microwa~-e tintenrm theory is that ~~ith 1 = ~/2. This is
usually called the half-1~-~vc dipole a misnomer due, perhaps, to its
diminutive structure at microlmve frequencies and here retained lwrause
of its convenience, On setting m = 1 in lc~. (1(W) ~~eobtain thc field I
I
pattern of this radiator:

T I
Cos Cos 0
2T.[;
10 ~,k1([()] 2
Eo=j ~ (I (i,5)
() sill d
SEC.3.17] SUPERPOSITION OF FIELDS 99

The corresponding power pattern is

P(e,o) = ; ~, (166)
f) 3 [Cos!::e)]o
The pattern differs only slightly from that of the electric dipole; it is
uniform in azimuth and has its single maximum in the equatorial plane.
00 20 40 50
~
700

\ 80
\
\
\
J I ~(z)= 10COS(~) 900
/
/

// 100J
1
(a)

1100

180 160 140 1300 120

(b)
F1&3.12.-The half-wave dipole:(a) current distribution along the wire;(b) meridional
A.
polar diagram compared with that of the electric dipole: -~ &pole, --- infinitesimal
dipole.

Figure 3.12 shows the meridional polar diagram in comparison with that
of the dipole. The gain of the half-wave dipole is

Gm= (1.65) =2.17db. (167)

The slight increase in directivity over that of the electric dipole arises
from the fact that at points off the equatorial plane there is partial
destructive interference between contributions from different portions
of the wire, which lie at different distances from the point of observation;
this leaves the radiation in the equatorial plane relatively stronger.
3.17. Superposition of Fields.We shall often have occasion to deal
with sources that consist of a number of separate current distributions.
As long as the total system is confined to a finite region of spac~the
only practical casethis problem is in principle covered adequately by
the general theory of Sees. 3.9 to 3.11. It will, however, be useful to
reconsider it from the point of view of the superposition principle stated
in Sec. 3.2. The total field is the sum of the component fields due to
100 RADIA TIOiV FFW.}1 CURRE.VT DISTliII~ LT7710.VS [Six 317

each component current system. We shall confine our attention to the

far-zone fields, existing at field points far removed from e~-ery source in
the total system.
The notation to be employed is illustrated in Fig. 3.13. We choose
an origin O within the neighborhood of the sources; a primary system
of rectangular coordinates z, U, and z; and an associated spherical system
r, O,and @ The distance from the origin to the field point \vill,as before,
he R; ROis a unit vector in that direction. In connection ]rith any of the
component radiating systems, say the ith, v-e use a secondary coordinate

/
x
FJQ.3.lt3.-S1lr1erJ>oaitinrl
of fields.
system, with axes parallel to those of the primary system and origin Oi
within that source distribution at the vector position R, 1}ith respect to
0. The polar coordinates of the field point P in this secondary coordi-
nate system will be denoted by Ti, t?,,&
As in the general discussion of far-zone fields, we may consider all
the OY to be parallel to OP and all the 6,, +i to be equal to 0 and @ respec-
tively. Furthermore, the field due to the ith radiating system can be
expressed in terms of an equivalent quasi-point source at 0,. That is,
the component fields are, by Eqs. (134),

= -*e-klt[JLi+( i)i$lelkeROdodo
(168a)
SEC. 318] THE DOUBLE-DIPOLE SYSTEM 101

and

E,,= -~e-iki/v[Ji.i4 -(~)%Jmi.i,

. ~+ &+ri~2i(@,@) .
47rr,

The total field is obtained by summing the component fields. We

note, however, as in the discussion in Sec. 3.11, that we can replace r;l by
R, with an error of the order of Rq; in the phase factors we can simi-
larly write
ri=R RO. Ri. (169)

The total field is, therefore, given by

(170a)

where
(170b)

and
(171a)

where

%(e,f$) =
z F2i(0, @)eikR0R. (171b)

The space factors $, and $2 are complex, and the discussion of polariza-
tion in Sec. 312 apulies without change.
The problem ~sthus reduced to the superposition of quasi-point-
source fields arising from sources 0~ and described by the space factors Fl{
and F2i. The composition of the over-all space factors $, and & in
terms of these and the phase differences arising from the relative positions
of the sources is a procedure useful in many other fields-for example,
the theory of X-ray diffraction.
3.18. The Double-dipole System.The radiation patterns of com-
pound systems are usually more directive than the patterns of the
component systems; destructive interference between the fields of the com-
ponent systems takes place in certain directions, constructive interference
!n others, with the consequence that the total power density changes
more rapidly with angle and reaches more extreme values than does the
power density for any component system.
An important compound system with wide application to microwave
102 RADIATION FROM CURRENT DISTRIBUTIONS [SEC. 3.18

antennas is obtained by superposing two half-wave dipoles. We shall

here restrict ourselves to the case in which the dipole axes are parallel
and the currents are of equal strength, though of arbitrary relative phase.
We consider, then, two half-wave dipoles with centers at (O, a/2, O)
and (O, a/2, O) and axes parallel to the z-axis, carrying currents of ampli-
tudes 10 and 10e* (Fig- 3 ~14). Since neither source gives rise to an

xJ
~lG. 3.14.The double-dipole system.

E+-component in the far field, the total field can have no such component.
The space factors of the dipoles are alike, except for the current phase
term e-i$. Combining Eqs. (165) and (170b), we find the resultant field:

~,=~(:)=[%%e)l
172)
On making the substitutions

R,. R,=~sint9sin$,
(173)
R, . R, = ~ sin 0 sin O,

we obtain tinally

This is a dipole field modified by the presence of the last factor. The
spheres of constant R are the equiphase surfaces; at large R the field is
that of a directive point source at the origin midway between the dipoles.
The pattern is symmetrical in 1#1
about o = 7r/2 and in d about e = z/2;
SEC. 3.18] THE DOUBLE-DIPOLE SYSTEM 103

that is, it is symmetrical with respect to the yz-plane, which contains the
dipoles, and the zy-plane, to which they are perpendicular. These
planes of symmetry are known, respectively, as the principal -E-plane
and the principal H-plane of the radiation pattern. Since the pattern
is a function of both o and ~, a three-dimensional polar diagram is required

+900 +800 +7P +600 +70 +W +9W

/, --
,8m~+w0+~~o.8

titA+lO
0.4

ID%vv
-900 -703 -50 40 -30 -409 -5(P -70 -*
H=plane E=plane
FIG 3 15.E- and H-plane polar diagrams in the power pattern of the double-dipole
system,

for a complete presentation of its properties. However, in practice it is

usually sufficient to consider the principal E- and H-plane cuts.
The details of the pattern depend on the precise values of a and ~.
We shall here consider one special case, in which a = A/4, and $ = u/2.
The form factor is then (except for constant terms)

3(0,0)
Cos
= ()
T
5
sin @
Cos e
Cos : ( lsin
[
Osin@)
1
~ (175)

The principal E- and H-plane cuts of the power pattern (proportional to

the square of the form factor) are shown in Fig. 3.15. only a small
fraction of the power is radiated in the hemisphere to the left of the
104 12A.!)IA TION FRoM CUIiRENT DISTRIBUTIONS [SEC. 3.19

u-plane; no power is radiated in the negative y-direction; maximum

power in the positive y-direction. In the negative y-direction the radia-
tion from the dipole at y = +a/2 must travel a distance greater by a
quarter wavelength than the radiation from y = a/2 with resulting
phase retardation of 90. Since the current in the first dipole is 90
behind that in the other dipole, the fields from the two dipoles are 180
out of phase and annul each other. In the positive y-direction, the phase
retardation in the field from the dipole at y = a/2, due to the additional
path length traversed, is just compensated by the 90 phase lead of the
radiating current; the fields from the t\ro dipoles are in phase and rein-
force each other. Since each dipole has maximum field intensity in the
xy-plane, this has the consequence that the maximum in the total field
intensity lies in the + y-direction.
As a measure of the directivity of the power pattern, we may take the
half- and tenth-power ~vidths of the polar diagram in each of the principal
planes. These are designated by@ Eand @L(~u) for the E-plane half- and
tenth-power widths respectively; corresponding notation applies in the
H-plane. For the system under consideration

H. = 76, @~ = 180,
(176)
E).(;V) = 130, @.(fi) = 252. 1

3.19. Regular Space Arrays.The double-dipole system is the sim-

plest possible example of an important class of directive systems: regular
space arrays of similar radiators. I,et us consider a system of current
distributions, identical in structure but perhaps differing from one another
in over-ail amplitude and phase. The radiating units need not be simple
dipoles; they may be double-dipole systems or more complex current
systems, but all must have the same orientation in space and be described
by similar space factors F,(d, ~) and F,(6, ~), with respect to similarly
situated origins Oi. Now let these radiating units and their origins 0,
be arranged into a space array at the intersection points of a threc-
dimensional rectangular lattice (see Fig. 316). Let a,, a,, and a, be
the basis vectors c)f the lattice in the r-, y-, and z-directions, respectively,
and let the extents of the lattice in these directions be Nlal, N2% and
.V,a,. Choosing one of the corner elements of this lattice as a reference
point, fre can specify the position of an arbitrary lattice point Oi by the
relative-position vector
R~ = n,al + ?lgaz + ?zsas, (177)

where nl, n.z, and ns are integers Irss than or e(lual to .Vl, .172,and .Vs,
respectively, !Vr shall let the amplitll(le of the ith system he An,~,n,
and shall admit the possibility of a progressive phase delay in each of
the three basis directions of the lattice: the phase of the ith radiating
SEC. 3.19] REGULAR SPACE ARRAYS 105

system, relative to the reference system, shall be

* TL,7L,TL, = %+1 + n2#2 + ?L3*3. (178)

We need consider only the space factors $, and % defined in Eqs. (170b)
and (171 b). The space factors Fli and F2i are independent of i except for

FIQ. 3.16.The apace array.

the constant multipliers An,~,~,e~~lx Accordingly, the space factors

for the system as a whole are given by

5,(8,+) = II(19,c$)A(0,+),
(179)
34 L9,1$)= F,(O,+)A(O,+), 1

where

Here
A(e)+) =
m N,

na=On*=Ona=O
N,

A ~,n,naexp j
3

[Z i=l
n;(kRO . ai ~i) .
1
(180)

RO. al = al sin O cos @ = alul, (181a)

RO. az = az sin 0 sin @ = azuz, (181b)
RO. aS=a3cos0 = a3u3. (181c)

The total space factor is thus a product of the space factor for a radiating
unit by a lattice factor. The lattice factor, it will be noted, is itself the
space factor of a ~attice array of isotropic radiators with relative ampli-
tudes An,m~, and relatlve phases t~,~,~,.
If the radiating units all have the same amplitude, say equal to unity,
the sums in Eq. (180) can be evaluated. The term on the right becomes
a product of three factors:
106 RADIATION FROM CURRENT DISTRIBUTIONS [SEC. 319

where
N.
A, = @! (kalult,
) (183)
z
71,=0
This geometric series is easily summed. One finds

~i(hr;+l/ 2)(k.,u,$, )
sin ~ (ka(ui vi)
Ai = ~,?,i(La,u,-+J
(184)
[ sin + (k.;?Ji *i) 1
The power pattern of the space array is proportional to
)$,1 + j~,l
and is consequently given by
~(o,o) = Po(@,@) lA(L9,
d)12 (185)

except for the multiplicative constants. The second factor is the product
of three factors,

each of the form sin Nx/sin z. Such a function has principal maxima
at z = km, h being an integer; if N is large, the maxima are very sharp,
the function being only slightly different from zero between successive
peaks. The composite lattice factor will then have its principal maxima
only for those values of u, for which the three factors simultaneously
achieve their maximum values, that is, when

(187.)

u~=sin Osin$=
()h,+~ A
2ir ;2
(187b)

U3 = Cos e = &+& A, (187c)

() 2X .3

hl, hz, hs being positive or negative integers. These conditions cannot

be satisfied simultaneously by any choice of 0 and @ for arbitrarily chosen
hl, h~, hs; the possibility of simultaneously satisfying the three conditions
is determined by the values of the phases +i and the lattice dimensions
ai/i. Except when F,(o, O) has a zero in direction 0, 0 determined by
the above conditions, the lattice space factor of a very large lattice deter-
mines, essentially completely, the direction of the principal maxima in Lie
total radiation pattern.
 CHAPTER 4

WAVEFRONTS AND RAYS

BY S. SILVER
The preceding chapter dealt with radiation fields in their direct rela-
tion to the sources. It was found that the field represents a flow of
energy outward from the region of the sources; also it was demonstrated
separately that the energy flow in a time-varying field is a wave phe-
nomenon. We now turn our attention to the study of wave propagation
and the associated energy flow, without direct reference to the sources.
Several simple waveforms have already been dkcussed: plane, cylin-
drical, and spherical waves. In each case the wave was described by a
family of equiphase surfaces or wavefronts, and the propagation of the
wave was visualized as a progression of each wavefront into a contiguous
one; furthermore, the energy flow at every point was in a direction normal
to the wavefront. The main subject of this chapter is the extension of
these ideas to general waveforms,
4.1. The Huygens-Green Formula for the Electromagnetic Field.
We have now to consider the following problem: Given the values of the
electric and magnetic field vectors
over an equiphase surface, how can
we determine the field vectors at a
specified field point ?
The solution to this problem is,
in fact, contained in the general
integral of the field equations ob-
tained in Sec. 38. Let the fields be
specified Over an ewphase surface ~ FIG.4. 1.on the Huygens-Green relation.
(Fig. 4.1) which encloses all sources
of the field, and let P be the field point at which the vectors E and H are
to be determined. We now apply the general relations of Eqs. (3.108)
and (3. 109) to the region bounded by S and the sphere at infinity, Since
the sources of the field lie outside this region, the volume integrals vanish

xH)#+(n xE)x V#+(n, E) Wldl (la)

107
108 W.1lliPlioiVT,~ A A]) 1(A }S [SEC.4.1

and

H, = & ~[j~,(n x E)# + (n x H) x VIJ + (no H) v+] dAS, (lb)

/
)vhere + = e~ir/r and n is the unit vector normal to S indicated in Fig.
41. These equations provide the solution of the stated problem.
Equations (1) may be regarded as an analytical formulation of the
lluygens-Frrsnel principle, \vhich serves generally as a basis for the study
of wave propagation. The Huygens-Fresnel principle states that each
point on a given wavefront can be regarded as a secondary source ~rhich
gives rise to a spherical wavelet; the wave at a field point is to be obtained
by superposition of these elementary ~vavelets, with due regard to their
phase differences when they reach the point in question. Equations (1)
specify the nature of the wavelets arising at the various points on the
equiphase surface;1 as was pointed out in connection with Eq. (3.1 11),
the sources of the wavelets can be regarded as surface layers of electric
and magnetic currents and charges.
For the further purposes of this chapter it is desirable to write the
surface integrals in somewhat different form. By means of a rather
Laborious calculation they can be transformed into

respectively. z Relations of the same form must, of course, hold for the

I A comprehensivetreatmentof Huygens principle and its applicationto scalar

and vector waves has been given by Bakerand Copson, The Mathematical Theory of
Huygens Principle, Oxford, h-ew York, 1939. It should be noted that the integral
expression for the fields, and hence the interpretation of the sources, is not unique;
it is possible to adcl to ~qs. (1) any surface integral that is equal to zero. This is
actually done in making the transformation from ~qs. (1) to Eqs. (2) in this chapter.
2 This transformation can be effectedonly if the surfaceS is completely closed;
otherwiseadditionaltermsappear. The resultscan be obtainedby a simplerandfor
our purposesmoreusefulprocedurethanby directtransformationof Eqs. (l). It was
shown in Sec. 36that in a source-freeregioneach rectangularcomponentof a field
vector satisfiesthe scalar Helmholtzequation
V2U + k2u = O.

The integrationof this equationcan be performedby meansof Greenstheoremin a

manner analogousto that by which we integratedthe field equations. The scalar
Greens theorem states that given two continuous scalarfunctions F and G .ving
continuousfirstand secondderivativesin a region V suchas wasillustratedin Fig. 34,
then
SEC. 4.1] THE HU YGENS-GREEN FORM ULA 109

components of E and H in any rectangular coordinate system. We can

therefore develop most of our considerations in terms of the scalar
relation

(3)

where u will stand for any one of the rectangular components of E or H.

Equation (3) can be regarded as the mathematical expression of
Huygens principle for a scalar wave; the resultant wave amplitude at P
is again expressed as a sum of contributions for the elements of surface
dtl. The first part of the integral is a summation of terms of the form
(e-J/r) (&L/dn) dSa summation of the amplitudes of isotropic sp~erical
wavelets arising from sources of strength proportional to (du/dn) dS on
the surface elements dS, The second part of the integral can be inter..
preted similarly. We note that
d_ ~,k.
d+
-( ) ()
e,k, 1
~ cos(n,r)= jk+- cos (n,r), (4)
dn dr rr

because the field point P is the origin in the integral formulation. The
second part of the integral is thus a summation of anisotropic wavelets
from sources of strength proportional to u dS on the surface elements dS.
The directivity of the sources is expressed by the factor cos (n)r); each
wavelet includes a term for which the amplitude falls off with T2, like
the induction field of a dipole source. Substituting this result into Eq.
(3), we obtain

p= -:/s%
[(k+:)cos(nr) +Hds
Despite the arbitrary feature of the integral formulations pointed out
5)

in the footnote on page 108, we shall consider the Huygens-Green rela-

tions [Eqs. (1) and (5)] as the analytical formulations of the Huygens.
Fresnel principle for electromagnetic and scalar waves respectively. It

/v FV-GV2FV=
-L...SW+%
The co vention as to the directionof n is the sameas shownin the figure. Let F be
the sphericalwa- function4 = e- Ik/r andG the functionu satisfyingthe Helmholtz
equationfor the samevalueof k. The fieldpointP is againsurroundedby a sphereZ,
the radius of which later is allowed to approach zero. In the region bounded bY
5
~1... S. and x the volume integralvanishes, The detailsof the limiting process
that is then appliedto z follow very closelythosefor the vector case; the resultgives
the value of u at the fieldpoint P, namely,

=-LL.S(%-%)S
110 tt7A VEIRONTIS A.VD RAYS [SEC,42

should be emphasized that according to the Huygens-Fresnel principle,

there is no one-to-one correspondence between the field at the point P
and the field at any point on the wave surface; the field at P is an inte-
grated effect of contributions from every point on the wave surface.
4.2. Geometrical Optics: Wavefronts and Rays.The Huygens-
Fresnel principle, as expressed by the Greens theorem integrals, gives a
rigorous solution of the wave equation. It is frequently convenient,
however, to approach the subject of wave propagation from the less
rigorous point of view of geometrical optics, in which attention is focused
on the successive positions of equiphase surfaces, or wavefronts, and an
associated system of rays.
Let the wavefront at time t,
be the surface L(z,g,z) = l., of Fig.
4.2 and the new wavefront after
passage of a very short time & be
the surface L(z,y,z) = L, + 6L.
Geometrical optics is then con-
cerned not only with the form of
these surfaces but also with a
point-to-point t ran sf ormation
from one wavefront into the suc-
ceeding one. This is, of course, in
fundamental contrast to the point
of view of the Huygens-Fresnel
FIG. 4.2.On the propagation of a wavefront principle. 1 The point-to-point
in geometrical optics.
correlation of the wavefronts is
established by the rays, a family of curves having at each point the
direction of the energy flow in the field. In the case of electromagnetic
waves, a ray can be traced out by proceeding at each point in the direc-
tion of the Poynting \ector at that point. The rays are nearly normal to
the wavefront-exactly normal in the wave systems to be discussed in this
volumeand pass through corresponding points in successive wavefronts.
In an arbitrary medium the wave field is characterized by a ray
velocity and a wave velocity at every point. The ray velocity is the
velocity of energy propagation; it is represented at each point by a vector
that is tangent to the ray passing through that point. The wave velocity,
on the other hand, is always normal to the wavefront; it is the rate of
displacement of the wavefront in the direction normal to that surface.
Thus if V(z,yjz) is the wave velocity at a point (z,v,z) of the first wave-

1Treatmentsof geometricaloptics as a self-containedtheory are given by .J. L.

Synge,Geornetriccd Optics,Cambridge,London, 1937,and by ph. Frankand v. ~ises,
der Physik, Vol. II, Chap. 1, reprintedby Mary S. Rosenberg,
Differential-gleichungen
New York. 1943.
SEC. 4.3] RAYS IN AN IKHOMOOENEOUS MEDIUM 111

front in Fig. 4.2, tne vector vdt will extend from that point to the cor-
responding point on the second wavefront. The case illustrated is that
of an inhomogeneous medium in which v(z,y,z) is a function of position.
In an anisotropic medium the ray velocity and ~vave velocity differ,
in general, both in magnitude and direction; in isotropic media the
ray velocity and wave velocity are identical. We shall here restrict our
attention to isotropic but possibly inhomogeneous media; more gen-
eral discussions will be found in the references of the footnote on page 110.
As a result of the identity of the ray velocity and wave velocity, the rays
in an isotropic medium make up a family of curves orthogonal to the
family of wavefronts; the energy flow at any point is normal to the wavc-
front passing through that point.
The form of the wavefrcmts and rays can be determined as soon as the
function L(z,y,z) is given. This function is not uniquely determined by
the foregoing remarks. We shall, in addition, require that it be chosen so
that the wavefront L(zjyjz) = LO shall be one of constant phase (a/c)LO
relative to the phase at some chosen point, The function L(z,y,z) thus
defined is of basic importance in the analytic theory of geometrical optics.
It satisfies a differential equation which we shall now derive.
The phase increment between the two successive surfaces of Fig. 4.2
is (u/c) 8L. Moreover, since the wave proceeds from one surface to the
next in time at while the phase at any fixed position changes at the rate
u, this phase difference must be U6L. Finally if 6s. is the distance betl~,een
the surfaces at (z,y,d) and v is the wave velocity at that point, we have

(6)

However, we must also have

~L = IvL] &Sn. (7)

It follows that

IvLI = ~ = n, (8)

where n is the index of refractionin general a function of position in the

medium. The function L must therefore satisfy the differential equation

(9)

4.3. Curvature of the Rays in an Inhomogeneous Medium.In a

homogeneous medium the rays are straight lines; in an inhomogeneous
medium they have a curvature that ~ve shall now compute. I,et s be a
unit vector in the direction of the ray at a chosen point. This is normal
to the wavefront and must have the direction of VI,; so by Eq. (8) we
112 WA VIIFRON T,~ A NI) RA S,$ [SEC. 44

have then
~=.VL (lo)
n
Let N be a unit vector in the direction of the radius of curvature of the
ray at the same point and P the radius of curvature; the vector curvature
of the ray is then N/p. This curvature, ho~vever, is also given by ds/ds,
where s is distance measured along the ray. By the vector identity

g=(s. v)s=sx(vxs) (11)

we have then
N
s X(v X5). (12)
;
On taking the scalar product with N, Eq. (12) becomes
1
N. (sxvxs)=(Nxs). (vx S). (13)
P

Using Eq. (10) to compute V x s, and replacing V ~

we obtain finally
n o by (1/n)V(ln n)

1
N. V(ln n). (14)
i
Since the radius of curvature is an essentially positive quantity, it
follows from Eq. (14] that the rate of change of the refractive index in
the direction of the radius of curvature is positi~-e; that is, the ray bends
toward the region of higher index of refraction. In a homogeneous
medium where n is independent of position, the right-hand side of Eq.
(14) is zero, the radius of corvature is infinite, and the rays are straight
lines, From Eq. (12) it follows also that in a homogeneous medium the
vector field of the rays satisfies the condition

Vxs=o, (15)

This is a sufficient condition for the existence of a family of surfaces

orthogonal to the field of vectors s.
4.4. Energy Flow in Geometrical Optics.Consideration of the rays
leads to a simple hydrodynamic picture of the energy flow. It was
pointed out previously that the rays are lines of flolv of energy. Let us
consider the two wave surfaces LI and L2 of Fig. 4.3 and a tube of rays
that cuts out elements d.4 1and dA j on the respective surfaces. No power
will flow across the sides of the tube; the flow acrms any section normal to
the tubes will be constant. If S is the rate of flow per unit area, the con-
dition of constant power flow through the tube is

S,dA, = S, dA2. (16)

SEC. 4.4] hNIIRGY FLOW IN GhOMETIiICAL OPTICS 113

In the case of electromagnetic waves the quantity S is the magnitude of

the Poynting vector; we shall assume that as in the case of plane and
cylindrical waves (Sec. 3.7)

s = ;
()2
; pq. (17)

If the permeability p is independent of position, the relation between the

electric amplitudes at dA I and d.42 is

In terms of the refractive index n = (c/~0) ~~we have

nIlE112 cZAl = njlllzlz dAz. (19)

Unlike the Huygens-Fresnel principle, geometrical optics sets up a one-

to-one correspondence between the amplitude at one field point and the
amplitude at another.
~, L2

x
.. B
/. L,
~
,:/;, dA2

L2 B
Oz
zr7 : r A
dA ,
Oy --
CA -
c
@j L, j $ ~

(a) (b)
?4
Y
FIQ. 4,3.Energy relations in geometric optirs: (a) tube of rays in an inhomogeneous
mediuln; (b) relations between wavefronts in a homogeneous medium.

It will be of interest to apply Eq. (19) to the case of a homogeneous

medium in which the rays consist of straight lines. The segments of
rays bet~veen the \vavefronts L1 and Lz, as sho\vn in Fig. 4.3b, will have
equal lengths p. Let the ray through the point A on surface L, be the
z-axis, and let the u- and yz-planes coincide \vith the principal planes
of Ll atA. .4 ray through an adjacent point B lying in L1 and the u-plane
\villintersect the ray through A at the point 0., at a distance RI which is
one of the principal radii of curvature of LI at point A; a ray through an
adjacent point C in the yz-plane ~vill similarly intersect the ray through
.4 at the point 0,, at a distance Rz which is the second principal radius
of curvature of Ll at A, The radii of rurvature ~vill be considered to be
positive if the centers of curvature lie on the negative z-axis, as shown.
114 WA VEFRONTS AND RAYS [SEC.4.5

The point A on the surface Lz lies on the ray through A. It can be

shown that the principal planes of L2 at A are coincident with those of
L,; through A we ran pass coordinate axes z, y which correspond to the
axes of. r, y, respectively. It is obvious that the principal radii of
cur\ature of the surface Lz at the point A are R L + p and R2 + p.
Let us now consider an element of area d.?l, ~vhicb includes A and is
bounded by the curve r. The rays through the curve intersect L2 in
the curve I, which bounds an element of area dA z around the point A.
These areas are given by

(20)

It is evident from the figure that the mordinatcs of corresponding points

(xjy) and (x,y), near A and A respectively, are thus related:
R, + p
x = (21a)
R, x

~f = &R~ y- (21b)

Substitution of these relations into Eq. (20) gives the relation between
the cross sections of the tube of rays at L1 and Lz:

(22)

Inserting this result into Eq. (19) and recalling that in the present case
nl = n2.,we obtain the relation
I!
]E2] = !El; R:R2- (23)
(1 (RI +P)(~2 +P) 1)
When R, and Rz are both finite and the surface L, is so far from L, that
and Rz, this reduces to
p >> R1

(24)

This last relation will be of use to us in the discussion of scattering of

radiation by curved surfaces.
4.5. Geometrical Optics as a Zero-wavelength Limit .We shall no~~
investigate the relation betfveen geometrical optics and the field equa-
tions, taking up in the succeeding section its connection with the
I The subject is trmted from the point of vimr of the scalar \ravrequ~tionby 1.
Debye, Polar .Mole.ules,( hap. 8, reprint by I)ovcr Iul)li(atio,,s, Xew York, 1945;
also in the articleby .k.Sommerfcldin 1%. I:rankan(i l:. llism, I)tflrre?~tialgfeichungeT~
der Physik, Chap. 20, reprintby Mary Rosmherg, Now York, 1943.
Huygens-Fresnel principle; the analysisw.11 reconfined to homogeneous
media.
A careful review of the ideas of the preceding sections will make it
evident that geometrical optics is based on the idea that a ~ravefront
behaves locally like a plane wave. The corresponding solution to the
scalar wave equation is
u = A(z,y,z)~J[~~~oL(r,v,z)l (25)
where kO = 27r/kO,il(x,y,z) is the amplitude of the wave (usually a func-
tion of position) and L(z, y,z) is the characteristic function defining sur-
faces of equal phase. We are here concerned ~rith aliuearly polarized
electromagnetic field and must consider the vector counterparts of this
solution. Let us then investigate the possibility of satisfying the field
equations by electric- and magetic-field vectors having the form

Thearnplitud evectcrs aand~maybecomplex, buttheir phases in that

case must be independent of position.
On substituting these expressions into the homogeneous forms of the
field equations [Eqs. (332)], it ~vill I)e found that the amplitude and
phase functions must satisfy the relations

(27a)

(27b)

On eliminating ~ from Eq. (27b) by means of Eq. (27a) and replacing

k~ by U(WOCO)
~ it will be found that a must satisfy the equation

a = + [vL(rI . VL) m[vqq + ~+ [VL x (v x (Y)

+ v x (VL x rr)] + ~lr$ [v x (v x a)]; (28)

n is again the index of refraction. Similarly, on eliminating a from F,q.

(27a) we find that Q must also satisfy Eq. (28).
If VL and the derivatives of a and D are finite, the last two terms on
the right are of the orders 1/~0 and l/k~, respectively, as compared with
the first. As x goes to zero, knapproaches infinity and the last two terms
approach zero. For Eq. (28) and the analogous equation in Q to be
satisfied under these conditions we must have

a.vL=O, (29a)
Q. VI, =O, (29b)
lv~.~ = n. (29c)
The last of these conditions is the differential equation for the charac-
teristic function that ~vasdeveloped in Sec. 4.2. The first two conditions
state that a and @ must be transverse to VL; it follows that a and ~ lie
in a plane transverse to the direction of propagation. Furthermore,
Eq. (27a) can be written as

g = (;)(: x .) - ,+ (;) (v x a), (30)

the second term being of order l/ko compared with the first. In the
limit k ~ O the second term vanishes. Since VL/n is a unit vector in
the direction of propagation, we see that in the limit x ~ O, ~ must be
perpendicular to a as well as to the direction of propagation. It follows
that the Poynting vector is normal to the wavefront and that its magni-
tude is

(31)

to terms of the order of I/k,.

V/e have thus seen that the field vectors of geometrical optics [Eqs,
(26)] possess the properties which were shown in Sec. 311 to be possessed
by the far-zone fields. In this region, at least, we may expect geometrical
optics to serve as a reasonable approximation to the exact theory.
It should be emphasized that the terms of order l/ko and l/k{ in
Eqs. (28) and (30) may be considered negligible for short ~vavelengths
only if the derivatives entering into these terms are tinite. In the
neighborhood of a geometrical focal point the function L varies rapidly
and its derivatives assume large values; at the boundary of a geometrical
shadow the amplitude varies rapidly. In these regions the geometrical-
optics approximation fails, and phenomena are observed that are not,
covered by the simple theory of wavefronts and rays.
4.6. The Huygens-Fresnel Principle and Geometrical Optics: The
Far-zone Approximation.It will be instructive to investigate the rela-
tion between the Huygens-Fresnel principle and geometrical optics to
see under what conditions the point-to-point amplitude relation [Eq. (23)]
that was obtained in Sec. 4.4 on the basis of the geometrical-optics con-
cept of the flow of energy in tubes of rays can be derived from the
Huygens-Fresnel principle in the limit of zero wavelength. The discus-
sion will be restricted again to homogeneous media.
For our present purposes it is sufficient to consider any one scalar
component of a field vector; we therefore take as our starting point the
scalar integral formula [Eq. (5)]:

p= -+~z:[u(k+:)cosnr)+
:ld~ (45)
SEC. 46] FAR-ZONE APPROXIMATION 117

where the surface S encloses all the sources of the field. In view of the
results obtained in the previous section we confine our attention to
the field far from the sources; the present section isdirected toward the
development of an approximation to Eq. (5) suitable forequiphasesur-
faces in this region.
Inthefar zone the field isaquasi-point-source field (Sec,3.11); that
is, the amplitude function takes the form

(32)

where p, 0, @ are the spherical coordinates of a point in the far zone with
respect to an arbitrary origin in the neighborhood of the sources. If n
is the unit normal to b directed out from the region containing the sources
(Fig. 4.1) and pl, 01, +1 are unit vectors in the directions of increasing p,
o and O, respectively, at a point on S, the normal derivative of u on this
surface is

By use of Eq. (32) we obtain

1 aF
Vu = jkuel + ~ Pl+~#l+
( ~+%)+($) 34)

In the far zone p >> k; consequently

vu = jkue,, (35)
and
13u
= jku cos (n,p, ), (36)
an

providing also that the variation of the amplitude in the 0 and @ direc-
tions is small compared to that in the radial direction. The integral
relation thus becomes

1
uP= ~~ jku[cos (n,@,) cos (n,r)] ~ cos (n,r) dS. (37)
4us/{ T }
Finally, if we consider only field points P such that r >> A for all points
on S, the last term in the integrand is negligible with respect to the first.
We then have, as an approximation valid in the far zone,

uP=~ UICOS(n,pl) cos (n,r)] ~ dS, T>> L (38)

2A
1 s
In the limit x ~ O this equation can be applied with virtually no restric-
tion as to the location of the field point P.
118 WA VEFRONTS AND RAYS [SEC. 4.6

Equation (38) applies to any surface in the far-zone region that

encloses the sources of the field. Let us now consider the surface S to
be an equiphase surface and assume on the basis of the preceding section
that the field can be expressed in the form of Eq. (25),
~ = /j (z,y,z)eI~oL(z,~,z). (425)

It was seen that in the limit x -0 this leads to a solution of the field
equations such that the Poynting vector is normal to the equiphase sur-
face. On the other hand, in the investigation of the far-zone fields in
Sec. 3.11 it was found that neglecting terms of order l/p3 the Poynting
vector is in the direction of CIIindependent of the choice of the origin in
the neighborhood of the sources. Consequently, if ISis an equiphase sur-
face, we have as an approximation valid for short wavelengths
cos (n,pl) = 1. (39)

The integral relation [Eq. (38)] in this case reduces to

(40)

It will be recognized that Eq. (39) is tantamount to assuming that the

equiphase surfaces do not differ widely from spheres about the source
distribution. Also in view of the condition associated with Eq. (35)
that p>> k, the assumption is implied that the radii of curvature of the
equiphase surfaces are large compared with the wavelength.
A consideration of the normal derivative &A/dn in terms of field
expression of Eq. (25) shows an additional assumption, concerning the
amplitude A (z,y,z), which underlies the use of the far-zone integral [Eq.
(40)]. Taking Eq. (25), we have for the normal derivative of u on an
equiphase surface

au 1 8A
=U jk+xz (41)
an ( )

In obtaining this result use is made of Eq. (29c). Substituting Eq. (41)
into Eq. (5) shows that we pass from the latter to Eq. (40) under the
condition that

(42)

This is satisfied, of course, in the limit x -+ O provided that (l/A) (13A/&t)

is finite everywhere. In the practical case, where the wavelength is
small but not equal to zero, the contribution of (1/A) (dA/dn) can be
neglected to a good approximation if the fractional change in amplitude
over a distance equal to the wavelength is small compared with unity.
SEC.4.7] THE PRINCIPLE OF STATIONARY PHALSE 119

4.7. The Pficiple of Stationa~Phase.-Equation (40) still expresses

the field at point lasa superposition of spherical wavelets arising from
every point on the equiphase surface. The transition to the geometrical
optics result of Sec. 4.4 is carried through on the basis of the principle of
stationary phase which we shall discuss in this section.
Let the surfaceS of Fig. 44betheequiphase surface and Pbe the
field point. There are at least two points on Sat whichthe normal to
the surface lies along the line passing through P. Let N be the nearest
of such points to the latter, and let NP be the z-axis of our coordinate
system; the x- and y-axes are taken in the principal planes of curvature

P(O,0,P)

FIG. 4.4.On the principle of stationary phase.

of the surface at the point N. The surface is divided into segments Sl

and f% by the curve r along which the tangent planes to the surface are
parallel tothe z-axis. Weshall assume that ineachsegment there is only
one point at which the line of the normal passes through P. This condi-
tion implies that P is not a focal point of the rays associated with the
surface S.
Denoting the distance NP by p, the integral of Eq. (40) can be
rewritten as

Theintegral isasurnof vector elements and can be treated graphically

bythecustomary procedures of vector addition. The magnitude of the
vector element contributed by an arbitrary element of surface cLSis
(u/r)[l -cos(n,r)] dS, andtheangle bet~,eeni tandthev ectorfromthe
element of area at the point Nis (27r/x)(r p). Consider now the con-
tribution from an arbitrary portion of the surface as a function of the
wavelength. If the wavelength is large, the angle between vectors
from adjacent surface elements issmall; thevector diagram in this case
takes the form of a gradual curve, as is illustrated in Fig. 4.5aj and we
may in general expect a resultant vector u~ significantly different from
120 WA VEFRONTS AND RA ls [SEC. 47

zero. On the other hand, if the wavelength is small, the angle between
adjacent elements is large and the vector diagram takes the form of a
tightly wound curve as is shown in Fig. 4.5b. In the latter case the result-
ant vector UEmay in general be expected to be virtually zero, the more
so as k + O. Thus, as a result of the rapid variation in the phase of the
integrand of Eq. (43), we have de-
structive interference and virtually
complete cancellation between the
spherical wavelets from an arbitrary
portion of the phase surface.
(a) (b)
The situation is different, how-
FIG. 4.5.The vector representation
of the Huygens-lJresnelintegral: (n) the ever, for those portions of the surface
long-wavelength case; (b) the short- in the neighborhood of the point N
wavelength case.
on the segment S, and the correspond-
ing point on SZ. It is observed that the phase function
+(z,v)= rp=[z+ y+(pz)]~ p (44)
is stationary in the neighborhood of these points; at these points

(45)

Consequently, in the neighborhood of these points the phase varies

SIOW1 y, despite the short wavelength, and the vector diagrams represent-
ing the contributions of the areas around these points take the form of
Fig. 4.5a rather than that of Fig. 4.5b. The stationary phase areas yield
contributions to the integral [Eq. (43)] compared with which the con-
tributions of other portions of the surface are negligible. We are thus
led to the principle of stationary phase: For short wavelengths, the
integral of Eq. (43) representing the effect of the whole surface s is
negligibly different from the sum of the contributions of the areas about
those points on S at which the phase has a stationary value.
It will be observed further that at the stationary point on the seg-
ment S2, cos (n,r) = 1 and that the cos (n,r) will not be very different
from unity over the area in the neighborhood of the stationary point in
view of our earlier assumptions (Sec. 4.6) as to the nature of the surface.
The contribution from this area is again zero, since 1 cos (n,r) vanishes,
and we are left then solely with the contribution of the area around the
point N. The amplitude of the integrand of Eq. (43) may be considered
constantequal to its value at the point N over this area, and Eq. (43)
then reduces to
~jk p
j
uP=uv ~i~+dz dy, (46)
Ap // bN
where ~Mis a small area around the point N.
SEC. 4.7] THE PRINCIPLE OF STATIONARY PHASE 121

The equation of the slwface in the neighborhood of N is

=-(&+k)+ (47)

RI and R2 being the principal radii of curvature. Inserting this into Eq.
(44), we find that to second-order terms the phase function over the
area 6Nis

= XR+)X+X%92
this is to be inserted into Eq. (46). We may now, however, reverse the
(48)

application of the stationary phase principle and argue that the integral
of Eq. (46) may be extended over the infinite (z,y)-domain with negligible
error. We thus obtain
~ jkp . k

with
up = ~uN
A /!
p -me
- z[=+@ dz dy, (49)

a=(%+)(%+) (50)

The integral of Eq. (49) can be transformed to Fresnel integrals,

m. .
e~( d~ = 42 ez,
\ .
with the final resultl
1The argumentmay be appliedin generalto integrals

over a region R in which F(u,u)has bounded variationin each variable. If (uo,oo)is

a stationarypoint of the function o in the region R, and if the coefficientsa and i3
of the canonicalform of dz+ at that point,

are both differentfrom zero, the asymptotic value of the integralfor large k is
. .

~ z F(uO,uO)eiWC%.%)
/i e3$(RE+8)d( dq;
-. .
or,

If @ has morethan one stationarypoint in the region,the total valueof the integralis
~Mainedby summingthe latter expressionover the stationarypoints.
The prin.;ple was formulatedby Lord Kelvin, Math, Phys. Papers IV, 303-306
(1910), for one-dimensionalintegrals; the latter has been discussed recently in a
 122 WA VEFRON1S AND RAYS [SEC. 4.8

R,RZ $5
up ~jkp
(51)
= N (R, +P)(F?2 +P)

The amplitude relation is seen to be identical with that obtained on

the energy flow basis. The factor ejp simply represents the phase
change corresponding to the displacement of the wavefront S along the
optical rays to the wavefront containing the point P.
4.8. Fermats Principle.-We shall now return to the discussion of the
methods of geometrical optics and shall consider several principles that
underlie the design of reflectors and lenses. The first of these is Fermats
principle, which is often taken as the basic postulate in the development
of the general theory. 1
Before stating Fermats principle we must introduce the idea of
optical path length. The optical path length AL along a curve r
between points P, and P2 is defined by a line integral along this curve:

AL = nldsl, (52)
/ r
where n is the refractive index at the line element ds.
This concept is intimately connected with the ideas discussed in
Sec. 4.2. Between t\vo adjacent phase surfaces L(z,y,z) = LO and
L(z,y,z) = L, + ~L, there is an increment in the value of the char~cter-
istic function L which is, by Eq. (6),

(53)

The distance 6s. between the two surfaces is a function of position, but
the quantity 6L = n 8s. is a constant; this, it will be noted, is the optical
path length along any ray between the two surfaces. It follows imme-
diately that the optical path length, as given by Eq. (51), is the same for
every ray bet\veen any two wavefronts L(x,y,~) = LO and L(z,~,z) = LI;
it is, in fact,
AL = ILI Lol. (54)

Thus the characteristic function L(zjy,z) can be interpreted as the optical

path length along a ray from the wavefront L(z,y,z) = O to the wave-
front in which the point (z,y,z) lies.
i
rigorous mannerby A. Wintner, J. Math. Phys,, 24, 127 (1945). As yet, rigorous
extensionto the two-dimensionalcase has not been made. The convergenceof the
integralsthat is requiredfor the process outlinedhere to be valid is assuredin the
case that a and B are both positive and k has a small negative imaginary com-
ponent k,; the finalresultis then tp be interpretedas the limit (afterintegration)as
k; -O.
1 Cf. J. L. Synge,Geometr{rat
O~!ics, Cambridge,London, 1937.

I
SEC. 4.8] F17RMA T,~ PRINCIPLE 123

The idea of optical path length is not restricted in its application to

rays. One can determine the optical path length between two points
P, and Pz along any curve r whatever; its value will in general vary with
the choice of r. Fermats principle provides a method for using these
values in selecting possible ray paths from PI to P2 from all other paths.
It may be stated thus:
Fermats principle: The optical ray or rays from a source at a point
PI to a point of observation Pe is the curve along which the optical path fength.
is stationary with respect to infinitesimal variations in path.
Usually the optical path along a ray has a maximum or minimum
value with respect to neighboring paths. The inclusion of the plural
possibilityy rays in the above formulation of Fermats principle is
,W
M
nl
P, _p ~+ J:,)(S1+ & , n2 -a,
--< P,
0 -----0 T131
P,s, / 761 P, m,
/ \
~m - o
@ / o -\m
*
K
\\*> \
~Q%, ~ s P232 \
~, 22
<v%,
/ P2
2? K
P2

(a) (b)
FIG, 4.6.Notation for the derivation of Snells laws: (a) reflection; (b) refraction.
required to cover situations in which the point Pz may be reached by
rays from PI by a direct path or by reflection from surfaces at which
there are discontinuities in the index of refraction.
It follows directly from Fermats principle that in a homogeneous
medium (n = constant) the rays are straight lines. The optical path
length is in this case proportional to the geometrical path length, and a
straight line gives a minimum value for both.
Fermats principle can also be used in deriving Snells laws of reflec-
tion and refraction at the interface between two homogeneous media.
Let us consider first the laws of reflection. Let the point O of Fig. 4.6a
be the point on the reflecting surface M for which the optical path length
from PI to O to PI has a stationary value. The optical path must con-
sist of straight line segments from PI to O and O to P2, since these paths
are in a homogeneous medium. The optical path length is then certainly
stationary with respect to neighboring curved paths from PI to P2 by
way of M which leave the point O unchanged; but by our postulate it is
stationary also with respect to straight-line paths with near-by reflection
points 0. Let us then consider a neighboring point O, displaced with
124 WA VEFRONTS ANI) RA Y.! [Sm. 48

respect to O by ~ 61, ~ being a unit vector in the tangent plane. We shali

now compute the variation in optical path length as O is changed.
Let s, and St (Fig. 4.6) be unit vectors in the direction P,O and 0P2,
respectively, andmthe unit vector normal tothe surface at O. Then we
may write the vector PIO as plsl and the vector OPZ as pZS.Z. Similarly
let sI+ 8s1 and s,+ &beu nit vectors along thelines PIO and OPa,
respectively, and pl + 3P1 and P2+ 3P2the lengths of these lines. The
variation in optical path by way of O with respect to the path by \vay
of the point O is

6L = rz(bp, + 3PJ; (55)

by our postulate this must vanish to terms of the first order in al. From
Fig. 4.6 it is clear that

(Pl + JP1)(S1+ as,) = p,s, + t N, (56a)

(P2 + aP2)(s2 + 6s2) = P2S2 ? 61. (56b)

To terms of the first order we have then

13p,sl + p, 8s1 = r 81, (57a)

tipzsz + pt & = r 61, (57b)
whence
apl = SI . c N, (58a)
dpz = s2. f61, (58b)

since sl . 6s1 = O. By Fermats principle, then,

iSL=n(s, -s.J. ~61=0 (59)

for all variations 62; hence

(S, s.z). r=o (60)

for every unit vector ? in the tangent plane. This gives immediately
the two laws of reflection:

1. The incident ray, the reflected ray, and the normal to the reflecting
surface all lie in the same plane. (The plane defined by s, and
SZis normal to the tangent plane.)
2. The incident and reflected rays make equal angles with the normal.
[COS(sl,~) = cos (s2,7); that is, the angles (sl,~) and (sa,r) are
equal. ]

The law of refraction is derived in a similar manner in Fig. 4.6b. The

variations in actual length of the paths P1O and OP2 are given again by
Eqs. (58); the optical path variation is, however,

6L = nl 6P1 + nz 6P2 = (n,sl n2s2). T 61. (61)

SEC. 49] THE LAW OF THE OPTICAL PATH 125

By I?ermats principle this again vanishes for every 61,whence

(?2,s, n,s,) 7 = o (62)

for every vector z in the tangent plane. This implies the two laws of
refraction:

1. The incident ray, reflected ray, and the normal lie in a plane.
2. nl cos (s1,7) = n2 cos (s2,z), (63a)

or in terms of the angles between the rays and the normal

nl sin (m,s,) = nz sin (m,sz). (63b)
Snells laws of reflection and refraction are again an expression of the
fundamental assumption of geometrical optics that the wavefront
behaves locally like a plane wave; in addition, they assume that the
boundary surface can be treated locally like its tangent plane. In field
theory Snells laws derive rigorously from application of the boundary
conditions of Sec. 3.3 only for the case of an infinite plane wave incident
upon an infinite plane boundary. 1 They follow in a good approximation
from these boundary relations if the radii of curvature of the two wave-
fronts (incident and reflected or refracted) and of the boundary are large
compared with the wavelength.
4.9. The Law of the Optical Path.Fermats principle provides an
independent formulation of optical rays from the method of the charac-
teristic function L(z,y,n) and equiphase surfaces developed in Sec. 42.
It was shown in the latter section that the rays are orthogonal to the
equiphase surfaces, and it was observed further in the preceding section
that the optical path along the rays between a pair of equiphase surfaces
is a constant. The treatment of Sec. 4.2 applies, however, only to
media in which the index of refraction is a continuous function of
position. We shall now show that the system of rays arising by refrac-
tion or reflection (in accordance with Snells laws) at a boundary of
discontinuity in the refractive index have associated with them a family
of equiphase surfaces, so that the law of constant optical path holds for
any pair of wavefronts, one a member of the incident system and one
of the refracted (reflected) system.
It was seen that in a homogeneous medium the rays are straight lines.
A family of straight lines for which there exists a family of orthogonal
surfaces is said to constitute a normal congruence. Thus, the rays
defined in Sec. 4.2the normals to the surfaces L(z,y,z) = constant
constitute a normal congruence. I.et us now consider the problem of
refraction or reflection, it being given that the incident system of rays
form a normal congruence associated with a family of equiphase surfaces

1See M. Born, OPtzk, p. 15, reprintby F;dwardsBros., Ann Arbor, Mich., 1943.
 126 WA VEFRONTS A,VL> RA 1,9 [SEC. 4.9

L(z,y,z) = constant. The first question is whether the refracted (or

reflected) system of rays is likewise a normal congruence. This is
answered in the affirmative by the theorem of Malus which we state with-
out proof:l
Theoremof .Vlalus:A normal congruence aftmany number ojre$ections
and refractions is again a normal congruence. The system of refracted
rays thus has associated therewith a family of orthogonal surfaces.
We shall now investigate the optical path along the rays between a
member of the incident ~ravefronts and a member of the surfaces ortho-
gonal to the refracted system of rays. I.et L,of Fig. 4.7 beawa,vefront

L,
FIG. 47.-On the law of the optiral path.

in the incident system and La one of the orthogonal surfaces of the

refracted system of rays and consider an incident tube of rays passing
through the closed curve I1 onLl; let r~ be the curve of intersection
of the tube with the refracting surface M and rz the curve of intersection
of the refracted tube of rays with the surface L2.
We shall evaluate the optical path from L, to Lz along any pair of
rays, say the paths ABC and A BC shown in Fig. 4.7. Let us consider
first the integrals

+ n,s, odl, (64)

\ A
(r,)

I See for example R. K. Luneberg, MafhernakcalTheory of Optics, Lectures in

Applied Mathematics,Brown University,1944.
SEC. 4.9] THE LAW OF THE OPTICAL PATH 127

and
c c B
~11) n& . d = nzsa . dl + m& . d + nzsz . df
+ / B / / c
[rj~
B

+ nzsz ~dl, (65)

\
(rm?
where SI and sz are unit vectors along the inc~dent and refracted rays,
respectively. Since s, and S2 are both normal congruences, they each
satisfy the equation [see Eq. (15)]

Vxs=o;

therefore, the line integrals around the closed paths

$(1) and $(II) are
zero. Furthermore the integrals over r I and Iz are zero, since s, and
S2 are normal to these respective curves. Adding the above integrals
(I) and (II) and transposing suitable terms, we th& obtain
B c B c
nlsl odl + n& . dl = ?zIsl . dl + 922s2. d
/ A / B \ A . \ B
B

+ (nlsl n2s2) dl. (66)

/
(r:)

The last integral of Eq. (66) vanishes as a result of Snells law of

refraction. The left-hand side is the optical path ABC, while the first
two integrals on the right-hand side constitute the optical path ABC.
We have, therefore,

(67)
~~Bcnd = ~ABcnd
The optical path and hence the phase increment are constant along all
rays from the equiphase surface L 1in the incident system to the surface
L2 in the refracted region. The family of surfaces orthogonal to the
refracted rays thus constitutes the refracted system of equiphase surfaces.
The law of the optical path often provides a simpler approach to
the determination of reflecting or refracting surfaces than do Snells
laws. As an example, let us design a reflector that transforms a spherical
wave into a plane wave. It is evident that the surface is a surface of
revolution and that it is sufficient to consider a plane section containing
the axis of revolution. In Fig. 4.8 let F be a point source, the center of
curvature of the spherical wave; M the reflecting surface; and LOany one
ot the family of plane wavefronts into which the spherical waves are to
be transformed. The optical path from F to the wavefront LO is
FP+ AP = const. = j +d. (O&V
128 WA VEFRONTS AND RAYS [SEC. 4.9

The constant may be evaluated by considering the path along the axis;
if the distance OF = j, the optical path is equal to f + d. ~Then FP
and AP are evaluated in terms of p and ~, Eq. (68) becomes the equation

0. ., +
d

Lo

FIG. 4X.-Appliration of the law of the optical path.

of the surface in polar form

2f
(69)
P=l+COS*

This is the equation of a parabola of focal length ~.

In contrast to the above calculation, application of Snells laws would
lead to the setting up of the differential equation of the surface; it ~vould
then be necessary to integrate this eqllation. Further examples of the
~pplication of the law of the optical path will be discussed in later chap-
ters in the design of mirrors and lenses.
 CHAPTER 5

SCATTERING AND DIFFRACTION

BY S. SILVER
The introduction of an obstacle into the path of a wave gives rise to
phenomena that are not covered by the geometrical theory of wave-
fronts and rays developed in the preceding chapter. These phenomena
scattering and diffraction-are of fundamental importance in micro-
wave antennas, for they underlie the formation of antenna patterns by
reflectors and lenses. In the present chapter the theory of scattering
and diffraction is developed with reference to general techniques; the
specific problems associated with antenna patterns will be taken up in
Chap. 6.
5.1. General Considerations.-The discussion of the scattering
problem will be restricted to the case of an obstacle of infinite conduc-
tivity. The problem with which we are concerned is the following:
Given a primary system of sources that produces an electromagnetic
field E,, HO; an infinitely conducting body is introduced into the field,
and it is required to find the new field E, H.
In practice the primary sources are distributions of currents and
charges over a system of conductors activated by generators. We shall
refer to the latter system of conductors and generators as the source
system, in distinction to the currents and charges over the obstacles.
The solut~on to our problem is based on the superposition principle
of Sec. 3.2. On introducing the body into the field of the sources a dis-
tribution of current and charge is induced over its surface. We then
have two component fields: one arising from the induced distribution
over the body and the second arising from the currents and charges in
the source system. The total field E, H results from the superposition
of the component fields. It should be noted, however, that the field of
the body reacts on the source system with a resulting perturbation of its
current distribution, so that the component field of the latter differs
from the original field EO, Ho.
The interaction between the body and the source systemand the
total field E, Hcan be analyzed as a superposition of multiple scattering
processes. First we consider the interaction of the body with the original
field Eo, HO, assuming no change in the source currents. The body sets
up a scattered wave E:, Hl, arising from an induced distribution over its
surface. The scattered wave falling on the source-system conductors
129
130 SCA TTERINQ AND DIFFRACTION [SEC,52

induces a current distribution in the latter that gives rise to a secondary

scattered wave E;, H:. The interaction of the secondary wave with the
body is again a scattering process leading to an incluced distribution over
the body and a scattered wave E,, H;, and so on. The total induced
distribution over the body is the sum of the distributions associated with
the component scattered waves l?:, E:, . . , and the resultant distribu-
tion in the source system is the sum of t!le distributions associated with
EO, E& , respectively,
If the distance R between the source system and the body is large
compared with the dimensions of either, the scattering processes of order
higher t!lan the first car, generally be neglected; for
example, in general the ratio E(/EU evaluated at the
body is of order l/R2 and the ratio E~/Eo is of the
order l/R. Also, in special cases, where, although
the distance R is not large, the geometry of the body
is such that the amplitude of the scattereci wave E,,
n
H, at the source system is small, multiple scattering
FIQ. 5.1
may be neglected in the analysis of the total field E, H.
These conditions are usually met in microlvave antennas, and the multiple
scattering will be neglected in the study of the antenna pattern.
6.2. Boundary Conditions. With attention restricted to a single
scattering process, our problem is that of finding the scattered field E,,
HI set up by an infinitely conducting body Ivhen it is introduced into an
initial field EO, HO; the total field is then
E= EO+ E,, (la)
H= H,+ H,. (lb)
It is assumed that the initial field is prescribed for all space.
Let V in Fig. 5.1 be the region occupied by the body; n is a unit vector
normal to the boundary surface S of V, directed outward into the sur-
rounding space. Since the conductivity of the body is infinite, the total
field E, H is zero everywhere inside the regiun 1; according tu the bound-
ary conditions of Sec. 3.3 there is a distribution of charge and current
over the surface S:
q = c(n . E), (2a)
K=nx H, (2b)

respectively. E and H are the total fields j~lst o~ltside T, and c and p
are the constitutive parameters of the s~lrrollntling medillm at the })t)~lncf-
ary surface, These chargp and current distrib(ltions are the s{]~lrres
of the scattered Ivave El, H,.
From Eqs. (]) it is seen immediately that at all points in the interior
of the body the scattere(l \v:~veis o~lt of pllas~ \Jith the original tield:
El = E,, H, == H,, (3)
SEC. 5.2] BOUNDARY CONDI1IONS 131

since the total field is zero. Accordingly we need concern ourselves only
with the region exterior to V. Here the scattered field must be deter-
mined as a solution of Maxwells equations that satisfies appropriate
boundary conditionsat infinity and over the surfaces. The boundary
conditions to be imposed at infinity are the radiation conditions [Eqs.
(3.113)], since the field arises froma current distribution confined toa
finite region of s~ace. Over the surface S, the scattered field must be
such th~t the to~al field satisfies the boundary conditions [Eqs. (3.24)
and (3.28)]:
nxE=O, (4a)
n. H=O. (4b)

From Eqs. (1) we have that the corresponding boundary conditions on

El and HI are

nxEl=nx EO, (5a)

n.Hl= n. HO. (5b)

Since the field E,, H, is known, Eqs. (5) prescribe the tangential compo-
nent of El and the normal component of HI as known functions over AS.
The boundary conditions [Eqs. (4a) and (4b) or (5)] are not independ-
ent. If the field satisfies Maxwells equations and one of the boundary
conditions, it necessarily satisfies the other. Let us assume, for example,
that condition (4a) is satisfied by the total field. Applying the integral
relation between the field vectors [Eq. (3.17a)] to any area on S bounded
by an arbitrary curve I, we have

d
E.ds=~ ~ B.ndS=O, (6)
$ r /

since E o ds = O by virtue of the boundary condition (4a). The result

holds for an arbitrary area, no matter how small; consequently n o B = O
over the surface. Therefore only one of the boundary conditions need be
considered in selecting the appropriate solutions of Maxwells equations.
The problem can be approached from another point of view. We
shall restrict ourselves at this point to an ewttime dependence and to
homogeneous media. It is evident that if the surface distributions
[Eqs. (2)] are known, the scattered field is obtained directly by the meth-
ods of Sees. 3.9 and 3.10. It can be verified readily that the surface
distributions [Eqs. (2)], satisfy the equation of continuity [Eq. (3.9)],
ove~ the surface (E, H being required to satisfy Maxwells equations);
as a result the field vectors E,, HI can be expressed in terms of the current
distribution alone, as was done in Sec. 3.9. In fact, the appropriate
expressions are obtained from Eqs. (3.120) and (3.121) by passing from
volhme to surface integrals. The scattered wave is then
132 SCATTERING AN~ DIFFRACYION [SEC.541

E,= &e
/ ~ [(K . v)v + lcK] ~dS, (7)

H,=: ~(Kx V)e~d& (8)

/

where r is the distance from the field point to the element of surface dS.
The fields given by Eqs. (7) and (8) necessarily satisfy Maxwells
equations and the radiation conditions at infinity. To determine the
current density K on the boundary surface S we must use condition (5a)
or (56). Letting n denote the unit vector normal to S at the point of
observation, we have

nx EO= & ~ n x [(K . V)V + k2K] ~ dS. (9)

/

The left-hand side is a known function, and Eq. (9) is an integral equa-
tion for the determination of the unknown current distribution K. The
scattering problem is thus transformed to the problem of solving the
integral equation rather than Maxwells equations.
It will be observed that the current distribution which satisfies the
integral equation leads through Eq. (7) to an electric field that satisfies
the requisite boundary conditions over S and at infinity. It was pointed
out earlier that the electric and magnetic fields [Eqs. (7) and (8)] satisfy
Maxwells equations. Solution of the integral equation (9) thus yields
the unique solution of the problem. 1
6.3. Reflection by an Infinite Plane Surface: the Principle of Images.
The simplest obstacle problem is that of an infinite plane conductor.
Here the solution can be obtained on the basis of geometrical considera-
tions. Two cases will be discussed: (1) the initial field is a plane wave,
and (2) the initial field arises from a dipole source.
Re$ection oj a Plane Wave.Although the reflection of a plane wave
by a plane surface has been treated frequently elsewhere, it will be of
interest to treat the problem here in terms of the general ideas set forth
in the preceding section.
Let us consider a plane wave, of the type discussed in Sec. 3.7, travel-
ing in the direction defined by the unit vector Sm The initial field is
then [Eq. (3.62)]

(10)

An infinite plane conducting sheet is now introduced into the field. For
convenience the conductor will be taken to lie in the xy-plane (Fig. 5.2).
The unit vector n, normal to the sheet, is taken to be in the positive

I For z discussion of the uniquenesstheorem see J Stratton, Eledromu.gnetic

TiMory, McGraw-Hill, New York, 1941,Chap. 9, Sec. 2.
SEC. 5.3] REFLECTION BY AN INFINITE PLANE SURFACE 133

z-direction, and the angle of incidence, which is the acute angle between
the lines of direction of sOand n, is designated as 8.
The field set up by the current and charge distribution over the sur.
face of the conductor must be such as to produce zero resultant field in
the negative z-region. The scattered field in this region is therefore a

/)= I
4-/
/1,
,0

)%
I/
$/ Eiz
.
e -%
Ety / -.
-2

Eti
4 /
/
/

;/
I
~
E rz

/
/

Y .._v

FIG. 52 Reflection of a plane wave.

plane wave traveling in the same direction as E, but 180 out of phase
with it; denoting the former by EL, we have then
E, = E,e~@-iJ. (11)

It is evident, however, that the infinite plane current sheet sets up in

the positive z-region a field that is the mirror image of that in the nega-
tive z-region. Hence the scattered field in the region of interest is a
plane wave
E, = Ele~<u@@, (12)

traveling in the direction s, which is the mirror image of SO,with an ampli-

tude E, bearing the following relations to the amplitude of E, and thereby
to the incident wave amplitude EO: (1) Their magnitudes are equal,
IE,I= IEOI; (13)
134 SCATTERING AND DIFFRACTION [SEC. 53

(2) their respective components parallel to the zy-plane are equal in

magnitude and direction

nxE, =nx ( E,) = n xE,; (14)

(3) their components normal to the zy-plane are equal in magnitude but
opposite in direction,

n. El= n. (EO) = n.EO. (15)

It is seen that as a resldt of Eq. (14) the boundary conditions [Eqs. (5)]
are satisfied.
It follows from the image relation betlveen sOand s, that the vectors
sO,n, and sl all lie in the same plane and that
sO. n= sl. n. (16)

The relations between these vectors can also be expressed as

sl = so 2(n . so)n, (17a)

so = .51 2(n o sl)n. (17Z))

From the point of view of geometrical optics the unit vectors so and s,
define the directions of the rays in the \vaves E, and E,, respectively.
It will be recognized that the relations among s,, n, and s, are just the
lalvs of reflection derived in Chap. 4. It is thus seen that in this case
the scattering reduces to geometrical reflection of the initial \vave.
The magnetic-field vectors are obtained from the respective elertric-
field vectors by the plane wave relation of I<;q. (3.65). I.etting H, and
H, be the magnetic vectors of the incident and reflected waves, respec-
tively, we have

(18a)

(18b)

The total magnetic field is H = H, + H,, whence by Eq. (2b) the surface
current density on the reflector is

K=nx(H, +H,). (19)

Either by symmetry considerations or by direct calculation, it can be

shown that
nxH, =nx H,; (20)

-onsequently, F.q. ( I9) l)ecornes

72
K=2(nx H,)=2 ~ [nx(sox E,)], (21)
()
SEC. 53] REFLECTION BY AN INFINITE PLANE SURFACE 135

or, alternatively,

K=2(nx H,)=2 ; [nx(s,x E,)]. (22)

()
In the case of a linearly polarized wave it is convenient for some pur-
poses to express the field amplitudes in another way. Let &be the ampli-
tude of E, in magnitude and phase at any given point on the surface.
The vector amplitude is
E, = &eO, (23)

where eO is a unit vector that is constant over the reflecting surface.

Similarly the vector amplitude of E, at the same given point on the sur-
face is
E, = &el, (24)

with el likewise a unit vector. The unit vectors e~ and e, are related
by Eqs. (14) and (15):

nx(eO+el)=O, (25a)
n.eo=n. el. (25b)

In terms of these the current density expressions [Eqs. (21) and (22)]
become

(26)

and
),

respective y.
K = 2
()
~ [s,(n o e,) e,(n s s,)]&, (27)

Dipole Sources.Let us now consider the case where the initial field
is due to an infinitesimal electric dipole. The infinite plane reflector will
again be taken to be the zy-plane, and the dipole is located on the z-axis
at a distance a from the reflector as shown in Fig. 5.3. The orientation
of the dipole axis with respect to the reflector is arbitrary.
The current on the dipole is, of course, changed by the presence of
the reflector. In this case, however, the reaction of the reflector merely
produces a new dipole moment M in the source. This is due to the fact
that the current induced in the source by the reflector is necessarily that
of an infinitesimal dipole of, say, moment M,. The latter is along the
same line as the original dipole moment .WO,and the superposition of these
two is, therefore, again a simple dipole. The resultant moment of the
source will be designated by M; the field of the dipole is given in Sec.
3.13.
As in the case of the plane wave, the current distribution over the
surface of the conductor must be such that the total field is zero in the
136 SCATTERING AND DIFFRACTION [SEC!.
53

hemisphere of space of the negative z-axis. In so far as this region is

concerned, the reflector is, therefore, equivalent to a dipole M coinci-
dent with the source. By symmetry, however, the reflector produces
a field in the region of the positive z-axis that is the mirror image of its
field in the negative z-region; with respect to the positive z-region the
reflector is equivalent to a dipole located at a distance a on the negative

J-
.l+~

?4
a
<M

(a)
a

-M
M
/ z

(b)
M
-.

5.3.Dipolc ilr,agc>: (u) arbitrar.yorientation; (h)

4-(c)
M

d,pole parallel to the reflector; (c)

d]polc nor]])al to the mflec-tor.

z-axis. The sense of thedipole \vith respect tothesource is easily deter-

mined from the requirement that the fields of the image and the source
must combine to give a zero resultant tangential electric field over the
reflector. This leads at once to the result that the image dipole is
obtained by reflection of M in the plane. The total field in the positive
z-region is that of a double-dipole system made up of the source and the
image dipole; the field is obtained by the methods discussed in Sees.
3.18 and 3.19,
The arbitrarily oriented dipole can a!~~ays be resolved into a com-
ponent parallel to the plane (Fig. 5.3/1) and a component normal to the
plane (Fig. 5.3c). The images for these t\~o cases with respect to the
source M are an antiphase dipole and a synphase dipole, respectively.
SEC. 5.3] REFLECTION BY AN INFINITE PLANE SURFACE 137

By considering the fields for these two cases, the reader can verify that
the image sources correspond to geometrical reflection of the spherical
wave from the source by the conducting plane; at each point on the
latter the reflection takes place as though the incident wave were an
infinite plane wave.
The image sources for magnetic dipoles are easily arrived at either
by direct consideration of magnetic dipole fields (Sec. 314) or by con-
sidering the image of a small rectangular current loop, which is equivalent
to a magnetic dipole normal to its plane. The image of a current loop
can be obtained by regarding it as an array of electric dipoles. It is
then found that the image of a magnetic dipole is obtained by direct
reflection of the source in the plane: images for dipoles parallel and normal
to the plane are synphase and antiphase, respectively.
The method of images can be applied to any source distribution. If
only the radiation field is desired, the source distribution can be consid-
ered as a system of electric dipoles, the dipole moment distribution being
given in terms of the current density J by

(28)

dv being an element of volume in the source distribution [cj. Eq. (3162)].

Every dipole moment is resolved into a parallel and a normal component
with respect to the reflector, and the total field is the sum of the com-
ponent fields of the dipole elements and their images. With arbitrary
current distributions, however, it must be kept in mind that the reflector
plays an important part in determining the distribution. Only in special
cases, such as a half-wave dipole radiator of negligible thickness, does
the reaction of the reflector produce a change in the magnitude and phase
of the amplitude of the distribution as a whole without affecting the
relative magnitude and phase throughout the entire distribution. The
half-wave dipole can be treated on the same basis as the infinitesimal
dipole, substituting for the field of the latter the field of the half-wave
radiator given in Sec. 3.16.

APPROXIMATE METHODS FOR REFLECTORS OF ARBITRARY SHAPE

Exact solutions of the scattering problem have been obtained for only
a limited number of cases involving simple primary fields and reflectors
of simple geometry, such as spheres and cylinders. These problems are
treated in standard works on electromagnetic theory, to which the reader
is referred for the results. 1 In treating reflectors of arbitrary shape it is
necessary to resort to approximation techniques. Several such methods,

I See, for example, J. A. Stratton, E~ectromagnetic

Theory, McGraw-Hill, New
York, 1941, Chap. 9.
138 SCATTERING AND DIFFRACTION [SEC. 54

which yield very good results at high frequencies, are discussed in the
following sections.
6.4. The Geometrical-optics Method.The first method to be con-
sidered belongs more properly to the field of geometrical optics than to
that of electromagnetic theory. It is applicable to the case of a point
source, which has a broad radiation pattern in the absence of a reflector,
together with a defocusing reflector. A reflector of this type renders
every divergent pencil of rays incident on it more divergent on reflection,
as is illustrated below in Fig. 5.4. The scattering pattern of the reflector
is, therefore, very broad, energy being scattered in almost every direction
in space. In such a system the salient features of the total field, such

Y
(a) (b)
FIG. 54.-On the geometrical-optics method.

as the directions of zero and maximum amplitude, arise from the inter-
action between the scattered field and the primary source field. The
finer details of the structure of the scattered field are of secondary interest,
and therefore an analysis of the scattering on the basis of geometrical
optics suffices.
Illustrative of the type of problem to which the method can be applied
successfully is the analysis of the effects of the fuselage or wing structure
of an airplane on the radiation pattern of a microwave beacon antenna
mounted on it. The primary interest is in the lobe structure introduced
into the beacon pattern by interaction with the scattered field from the
aircraft structure, whereas the fine structure of the scattered field arising
from deviations from geometrical optics is of negligible significance.
Let the primary source be located at the point O in Fig. 54a. The
assumption that the source is a point radiator is justified in the practical
case of a more general source system if the reflertor is in the far-zone
field of the former. It was shown in Sec. 3.11 that in so far as the far-
zone field is concerned any current distribution reduces to a directive
point source, and in Chap. 4 it ~vas found that the far-zone field can Le
described adequately in terms of wavefronts and rays. We shall assume
SEC.5.5] CALCULATION OF THE SCATTERED FIELD 139

further that the wavefronts from the source differ negligibly from spheres
about the point O.
The geometrical-optics analysis of the scattering assumes that at each
point on the reflector th,: incident ray from the source is reflected by the
tangent plane according to the laws of reflection developed in Sec. 4.7.
The intensity of the scattered radiation in a given direction is obtained
by applying the principle of conservation of energy to the total power
contained in an incident cone of rays and the total power contained in
the associated reflected pencil of rays. The use of the laws of reflection
assumes that the reflector can be regarded locally asa plane surface and
the incident wavefront can be regarded locally as a plane wave. It is,
therefore, necessary to require that the radii of curvature of the reflector
and of the incident wavefront be large compared with the wavelength.
The latter condition, however, has already been assured by the fact that
the reflector is in the far-zone field of the sources.
5.5. Calculation of the Scattered Field.The procedure followed
here to determine the scattered power in a given direction is to consider
the local transformation from the incident to the reflected wavefront at
every point on the surfaceof the reflector. This determines the principal
radii of curvature l? land Raof the reflected wavefront, together with the
value of the field amplitude &,at the point of reflection. The magnitude
of the field amplitude 8P at a distance p along the reflected ray from a
given point onthereflector isthenobtained bymeansof Eq. (4.23):

(423)

We shall first investigate the wmplitude transformation from the inci-

dent to the reflected wavefront. Let us consider an infinitesimal cone
of rays from O incident on the reflector as shown in Fig. 54a; the cone
intersects the reflector in an element of surface dAS. The cone will be
taken to have a circular cross section; the ray along the axis of symmetry
is referred to as the central or principal ray. The vector n is a unit
vector normal to dSat the point of incidence of the central ray; let i be
the angle of incidence between the central ray and the normal. If &~
and &, are the magnitudes at the surface of the reflector of the field ampli-
tudes in the incident and reflected tubes of rays, respectively, and dS,
and dSa are the cross-sectional areas of the respective tubes at the same
point, the relation

expresses the conservation of power in passing from the incident to the

1.Alternativetechniqueshave been developedby R. C. Spencer,Reflections from
Smooth Curved Surfaces, RL Report No. 661, January 1945; C. B. Barker and
H. J. Riblet, Reflections from CurvedSurfaces,RL Report No. 976, February1946.
140 SCATTERING AND DIFFRACTION [SEC. 5.5

reflected tubes of rays. From the law of reflection we have that the
angle between the reflected principal ray and the normal is likewise i, so
that
d~l = d~z = dS GCISi, (30)
whence
l&,l = l&7]. (31)

The transformation of the polarization on reflection is obtained

directly from the results of the plane wave problem of Sec. 53. Let
E, be the incident electric-field vector at the surface and E, the reflected-
field vector; we have then from Eqs. (14) and (15)

nx(E, +E,)=O, (32a)

n. E,=n. E~, (32b)
or,
E,=(n. EJn-(nx EJxn. (32c)

The determination of the radii of curvature of the reflected wavefront

is a somewhat more difficult task. It will be necessary to make slight
changes in notation: The point of incidence of the central ray on the
reflector, at which the transformation of the wavefront is desired, will
be designated by P, and the unit vector normal to the surface at that point
by n,; the unit vector normal to the surface at any other point is n
The point P is taken as the origin of the coordinate system (Fig. 5.4b)
with the z-axis along np and the xy-plane tangent to the surface; the
yz-plane is the plane of incidence (containing the central ray and n,).
The axes ~, q are the lines of intersection of the principal planes of curva-
ture of the surface with zy-plane; the principal radii of curvature of the
reflector at P will be designated by R~ and R,, respectively. The plane
of incidence makes an angle u with one of the principal planes, say the
plane containing the q-axis. Let TObe the distance OP; the distance from
O to an arbitrary point z, y, z on the reflector is
r = [zz + yz + 22 + rj 2ro(y sin i + 2 cos i)]~i. (33)

Consider now the member of the family of reflected wavefronts that

intersects the reflected central ray at a distance p from the reflector.
Let u, v, w be the coordinates of a point on the wave front, and let x, y, z
be the coordinates of the point on the reflector for which the reflected
ray passes through a given point (u,o,w) on the wavefront. By the
law of the optical path (Sec. 4.8) the equation of the reflected wavefront
is then

T+[(u Z)z + (v Y)* + (w z)q~~ = rO+ p. (34)

Now let s, be a unit vector in the direction of an arbitrary incltiint

ray and s a unit vector along the associated reflected ray. From the
SEC.5,5] CALCULATION OF THE SCATTERED FIELD 141

law of reflection [Eq. (517a)] we have

sl = so 2(s0 . n)n. (35)

If U,V,W are the coordinates of any point on the reflected ray, the unit
vector SI is given in component form by

ux
= = [(u $)2 + (u 7J)2+ (Z/J z)]% (36)

wth corresponding expressions for SIu, sIS. Similarly, the components

of so are
x sOU=y ro sini .z-r Ocosi
so, = ;; 9 Soz= (37)
r r
Let
z = F(z,y) (38)
denote the equation of the reflector surface. The components of the
normal n at an arbitrary point on the surface are then

1 dF 1 dF
n.= ; ny= ;
A ax A au z = 2;

= [1 +(9+(%)7 39)
Substitution of Eqs. (36), (37), and (39) into Eq. (35) gives

u = x + Gl(x,y) U(U,V,W; Z,Y,Z),

V = y + G2(x, Y) U@,V,W; W,Z),
w = z + Gj(x,y) U(U,V,W; X,Y,Z), (40)
u = [(u x) + (v y) + (w z) ]~, 1
where

(41)

and

So.n= r.! za~(y rOsini)aX+ (z rocosi) . (42)

A [ au
1
Equations (4o) give the coordinates of arbitrary points on the system
of reflected rays. If, in particular, we consider the family of points
lying on the reflected wavefront that is defined by Eq. (34), the coordinates
u, v, w of the system of Eqs. (40) must satisfy Eq. (34); in particular
lJ(u.v,w; x,y,z) must satisfy the latter equation. Substitution for U
142 SCA TTERIAG AND DIFFRACTION [Snc. 55

into Eq. (40) then leads to

u = z + G1(z,y)(ro T + p),
V = y + G2(z,y)(r0 r + P), (43)
W = z + Gs(~,~)(rO r + p). 1

The coordinate z is eliminated from these equations by means of the

equation z = F(x, y) for the surface of the reflector. Equhtions (43)
then become a set of parametric equations (zJy being the parameters)
for the reflected wavefront that intersects the central reflected ray at a
distance p from the reflector. On setting p = O, we obtain
~ = x + G1(z, Y)(~o ),
v = y + G2(x,Y) (7-o ), (44)
W = Z + G3(.zjy)(r0 r), )

the parametric equations ioi- a surface that intersects the reflector at the
point P and there represents the reflected wavefront arising from the
segment of the incident wavefront defined by a small cone of rays about
the central ray.
The procedure for finding the principal radii of curvature of a surface
from its parametric equations is straightforward and can be found in any
book on differential geometry; 1 the details of the calculation will not
be reproduced here. We are interested in the radii of curvature of the
wave surface at the point u = v = w = O; i.e., z = y = 0. In elimi-
nating z from Eqs. (44) it is, therefore, necessary to use only the equation
for the reflector surface in the neighborhood of the point P. Referred
to the principal axes ~, q, the equation of the surface of the reflector is

=-(&+&) (45)

By a simple transformation, the equation of the surface with respect to

the z, y-axes is then found to be

1
~=__ COS2u sinz co sin z ~ COS2u
2 [( Rf R, 2+
)( RC+R, ),Y2

-2sinoc0s4bw+ )
radii of curvature Rt, Rn are considered to be positive if the surface is
convex with respect to the positive z-axis.
We are chiefly concerned with the over-all pattern produced by the
reflector and the source system; hence we are interested in the scattered
field at large distances from the reflector. Provided that neither one
1For example, L. P, Eisenhart, A Treatise on the Dzflerential Geometry of Curves and
Surfaces,Ginn, Boston, 1909.
SEC.56] SUPERPOSITION OF THE SOURCE FIELD 143

of the radii of curvature RI, R2 of the reflected wavefront is infinite, it is

physically possible to consider distancesp so large that p>> R,, p>>Rz.
In that case the calculation of the scattered field intensity is somewhat
simplified ;instead of Eq. (4.23) we can use Eq. (4.24),

The product of the radii of curvature of the reflected wave surface at

the point P is

cos i
R,R, = ~ Cos ~

-+ Cy + : SJ;2: + C* + C*+ * Coszi

RtR, [ v ( v )1
(47)

The result can be put into a more symmetrical form by introducing the
angles 01 and f?zbetween the incident ray and the principal axes of the
reflector f and q, respectively. The scattered field at a distance p from
the reflector in the direction defined by the reflected ray is then given by

R(R7 cos i
18P] = ~ IGI [(4,; + RFR,) cos i + 2r,,(R: sin 0, + R, Sinz %) i. (48)
1
The bracketed term is known as the divergence factor of the surface,

RERq cos i
(49)
D = (47-: + RtRn) cos i + 2r,(Rf sin 0, + R, sin 9,);

it is the ratio of scattered power per unit solid angle in the direction of
the reflected ray to the incident power per unit solid angle. By use of
Eqs. (31) and (32c), together with Eq. (48), the scattered field can be
obtained in magnitude, phase, and direction:

EF = ~ {(n . E,)n (n x E,) x n ]D~~e-~~P. (50)

5.6. Superposition of the Source Field and the Scattered Field.

The method of superposing the scattered field on the original field of the
sources is fundamentally the same as that used in Chap. 3 in treating the
far-zone fields of current distributions. It will be assumed that the
source field is linearly polarized. The fundamental elements are illus-
trated in Fig, 5.5. I.et P, be the total power radintwf by the source,
and let G(s) he the gain f~lnction in the direction defined by the unit
vector s. The field of the so[lrce alone over a sphere of radius R is

(51)
144 SCATTERING AND DIFFRACTION [SEC.5.7

The unit vector eo describes the polarization of the field over a sphere
with radius such that lcR = 2nr, n = O, 1, 2, . . . . The total field
in a given direction s is the sum of the scattered field produced in that
direction bythereflector andthe source field. Inthefar-zone treatment
the ray from the source to the given field point is taken to be parallel
to the ray from the reflector.

o
FIG. 5.5.Superposition of the scattered field on tl,e source field.

The field intensity incident on the reflector is

i=[2(9%GsOreos% (52)

The scattered field is then, by Eq. (50~,

)i
Ep=~2;
[() ~ DP,
~G(so)
1
[(n o eo)n - (n x eO) x n]e-~o+p; (53)

the distance p has been set equal to R in the expression for the amplitude.
In so far as the phase is concerned, it is seen from the figure that

Rsl = ?_oSo + PSI, (54)

whence
To + ~ = R + ro(l + COS 2i). (55)

The total field in the direction s, is, therefore,

where
e, = [n o eO(sO)]n [n x eO(sO)]x n. (56a)

5.7. The Current-distribution Method.The geometrical-optics

method discussed in the preceding sections can furnish no information
on the structure of the scattered field that results from deviations from
geometrical propagation of the reflected wavefront. By geometrical
optics this wave is discontinuous (geometrical shadow behind the reflec-
SEC, 5.7] THE CURRENT-DISTRIBUTION METHOD 145

tor), and it was pointed out in Sec. 45 that in the presence of a discon-
tinuity geometrical optics does not give accurate results. The deviations
decrease in significance as the wavelength goes to zero; the geometrical-
optics method is to be regarded as a zero-wavelength approximation to
the scattered field.
The current-distribution method which wi [1 be formulated in this
section leads to a better approximation for the scattered field and also
makes possible the analysis of secondary effects such as the reaction of
the reflector on the sources. The cardinal feature of the method is that
it attempts to approximate the current distribution over the surface of
the reflector; the scattered field is obtained from the current distribution
by Eqs. (7) and (8) and is thus an electromagnetic field that satisfies
Maxtvells equations. We shall be interested primarily in the far-zone
field of the current distribution in ubtain-
ing the composite pattern of the reflector
and the sources.
The current distribution over the re-
flector is obtained on the basis of geomet~
rical optics, which can be expected to
yield good results only if the reflector is
far enough from the sources for the field
of these to be described adequately in
terms of wavefronts and rays. On the l:IG. 56-011 the rurrent-distrihu-
basis of ray optics there is a sharply de- tion method.
fined shadow region behind the reflector
in which the total field is equal to zero. In Fig. 5.6, S is the reflecting
surface and r is the boundary curve between the geometrically illumi-
nated area and the shadow area. According to the boundary condition
[Eq. (2b)], since the total jicldis zero, the current distribution over the
shadow area is zero. It isa matter of experience that the shadow region
is more sharply defined the smaller the wavelength and the larger the
ratio of thereflector dimensions tothewavelcngth. The first assumption
of our approximation technique, then, is that there is no current over the
shadow area of the reflector. The current distribution over the illumi-
nated region of S is obtained on the assumption that at every point the
incident field is reflected w thou2h an infinite plane \vave were incident
on the infinite tangent pl:me. 1A E,, H, again be the initial field; let
s~~be a unit vector in the direction of the Poynting vector, tlmt is, along
the incident ray. If n is the unit vector normal to the surface at the
point of incidence and S, a unit vector in the direetifm of the reflected
ray, the surface current density, according to Ills. (21) and (22), is

K = 2(n x H,) = 2 ~ ~ [n x (s,, x E,)], (57a)

()
146 SCATTERING AND DIFFRACTION [SEC. 58

or, in terms of the reflected field E,, H, afi the surface,

K=2(nx H,)=2
()
.; [nx(s, xE)]. (57b)

The surface charge density is obtained from the total field E,+ E,by
means of Eq. (2a); making use of the plane ]vave relations [Eq. (]5)],
we then find that the charge density is

q = 2c(n. E,) = 2~(n. E,). (5s)

From the discussion of Sec. 54 it is seen that the procedure for obtain-
ing the current and charge distributions is based on the assumption that
the radii of curvature of the incident ~vavefront are large compared ~~itl~
thewavelength asarealso the radii of curvatureof the reflector. Ontllc
other hand, in the present case there are no conditions imposed on t
focusing or defocusing characteristics of the reflector. It is clear tha,
Eqs. (57) and (58) represent high-frequency approximations to the actual
currents and charges and may be expected to approach the latter in the
limit of zero wavelength. The current method differs from the previous
wavefront procedure in that a frequency dependence of the scattered field
is introduced into the subsequent calculation of the field arising from the
current and charge distributions. Also, the field at a given point in
space is the resultant of contributions from all points on the illuminated
area S0 rather than from the point of geometrical reflection alone.
5.8. Calculation of the Scattered Field.The expressions for the
electric and magnetic fields in terms of the currents and charges \vere
derived in Sees. 3.9 and 3.10. It will be recalled that the fields thus
obtained satisfy Lfaxwells equations only if the source-density functions
satisfy the equation of continuity [Eqs. (3.6) and (3.9)]. The reader can
readily verify that if the initial field E,, Hi satisfies Maxwell]s equations,
the current and charge distributions given by Eqs. (57a) and (58) do,
in fact, satisfy the surface equation of continuity given in integral form
by Eq. (3.9). The situation is different, however, at the boundary line
r between the illuminated and shadow regions. The current distribll-
tion is discontinuous across the boundary, being zero over the shado\v
area; compatibility y with the equation of continuity can be achieved on] y
by introducing a line distribution of charge along the curve r.1
In Fig. 57 ~ is a unit vector along the boundary curve r; nl is a unit
vector in the tangent plane normal to ~. The linear charge density along
r will be denoted by u. Considering a small area of sides ds and al
(the latter normal to I) and expressing the condition that the net c~lrrent

1The discussionthat follows parallelsth:~t~ivcn hy Str:Ltt{),,:,,1,1(<1,,,i,, their

trratment of diffraction; sec J. A. Stratton, ElectromagneticTheory, Nfc[;raw-Hill,
Sew York, 1941,Sec. 8.15.
SEC. 58] CALCLJLA TION OF Tffll SCA TTI?RED F] ELII 147

flowfrom the area is equal to the rate of decrease of the charge enclosed,
we obtain

n, .Kds= ~ds; (59)

the contributions from the sides N vanish as 61~ O while ds remains

fixed. We have then
au
nl. K=. (59a)
at
Substituting from Kqs. (57a) and (57b) for the current distribution, we
find
au
=2n, .(nx H,)= 2r. Hi (60a)
ix
and
au .
27. H,. (60b)
at
For time periodic fields these give tbe charge distribution dirertly; for
we have du/dt = jwu, \vhence
Shadowregion
2 2
~=Tt.Hi=~ ~.H,.
Ju ]W
(61)
:{e Illuminated
region
The scattered field is thus the
sum of the contributions of three $
2 /;:kkK n
source distributions: (1) the surface r
currents over the illuminated area,
1
(2) the surface charges over the o
same area, and (3) the line dist ribu - l:[c. 5.7.Calculation
of theelectriccharge
tion of charge along the boundary on theshadowImund.ary curve1.
c,urve r. JVe shall now restrict our analysis to time-periodic fields. Ap-
plying the results of Sees. 39 and 310, we find the scattered field to be

E.s =-~ Zm s, [~w(n x H,)y (no E,) v+] dS

\
- &$, V*(T . H,) ds, (62a)

H.=: , (n x H,) x V+ dS, (62b)

/ ,,,
j,,

where ~ = e-~ , with r the distance from the field point to the element
of area dS on the reflector; S,, designates the geometri(:dly illuminated
area; the sense of the line integral around r is s~lrh that the tmtwtird
normal to S,, is on the left. The fields can also l)e expressed in the same
\va.yin terms of the reflected fields E,, H, at i}~e s~lrface S().
148 SCA TTERI,VG AND DIFFRACTION [SEC, 58

It was shown in Sec. 39 that if the current and charge distributions

satisfy the equation of continuity, the fields can be expressed in terms of
integrals involving the currents alone. In view of the introduction of
the boundary line distribution it will be well to carry through the details
of the transformation for this special case. It will be recalled (cj. Sec.
3.8) that the gradient operations in the integrands of Eqs. (62a) and
(62b) are referred to the field point as an origin. Taking a rectangular
system of coordinates with the origin at the field point, let the coordinates
of a point on ASO be x. (zI = z, X2 = y, X3 = z), and let i. be unit vectors
along the x=-axes. The line integral of Eq. (62a) is then

v+fTFLjds = ~i.$r. (M#)ds. 163

$ r

By Stokes theorem each integral on the right-hand side transforms into

a surface integral:

and
() vx H,:
.
=V~xHi+~VxHij
a a
(65)

(v#xH) n=(nx Hi). v~, (66)

n = jut ~ (n . Ei). (67)

(% VXH) .
In the last of these use has been made of the field equation [Eq. (3.23b)].
Collecting these together, ~ve have

or

~~ (n x H,) . VV~ dS. (68)

1
Substituting into FkI. (62a), ~ve then obtain

E., = ~ (n x H,) . V(V~) + k(n x H.)+ ds (69)

21r3&le/[ ~ 1
SEC, 5.9] APPLICATION TO POINT-SOURCE FEEDS 149

It will be recognized that this is obtainable directly from Eq. (5.7)by

inserting the value for the surface current density given by Eq. (57a).
It wasshownin Sec. 3.11 that the field integral, taking the forrnof Eq.
(69), leads to a far-zone field in which the field vectors are transverse
to the direction of propagation. The effect of the boundary line distri-
bution is therefore to cancel the longitudinal field component introduced
by the surface charge and current -
distributions. Subsequent calcula-
tions can be made on the basis of
Eq. (69); the contributions of the
charge distributions need not be
evaluated explicitly.
Let p be the vector from a given
origin (see Fig. 5.8) to the element
of surface LLS;let RI be a unit vec-
tor from the origin to the field
point, the distance bet~veen them / Y
FIG. 5.S,Calculation of the total field.
being R. The scattered field inten-
sit y in the far-zone is then, according to Eq. (3.128),

Es=~eJ
jw ~~ {n x H, [(n x H,) . R,]R, }e~@R dS. (70)
/ so

The magnetic field need not be calculated separately but is given by the
far-zone relation

H., =
()
; (Rl X Es). (71)

6.9. Application to Point-source and Line-source Feeds.Two cases

of major interest are those in which the initial field E~, Hi arises from a
point-source system and a line-source system, respectively. Where the
reflector enters into the problem by intention as a component of the
antenna, the source system will be referred to as the jeed; this term is
used extensively in later chapters.
The Point-source Feed.It was noted previously that at sufficiently
large distances from any radiating system, the latter is equivalent to a
directive point source. Microwave point-source feeds are specially
designed so that the required distances are within practical ranges for
use with a reflector.
Let the point O in Fig. 5.8 be the point-source equivalent of the feed;
it will be assumed again that within the cone of illumination falling on
the reflector the incident wavefronts differ negligibly from spheres about
the point O. The reference system of coordinates will be taken with the
origin at the source system. Spherical coordinates will be designated
150 SCATTERING AND DIFFRACTION [SEC. 5.9

generally by p, d, d; the coordinates of a field point in the far-zone region

of the system as a wholereflector and feedwill be R, 8, *. If GJ(o, o)
is the gain function of the feed and P is the total radiated power, the
primary radiation field-of the feed aloneis

H, =
()
: 6 (@, xE,), (72b)

where pl is a unit vector along p and et(e, O) is a unit vector defining the
polarization of the electric field intensity. The current density K at a
point p, O, I#Jon the reflector is then

Substitution of the expression for n x H, from Eq. (73) into Eq. (70)
gives the scattered field in the far-zone. Equation (70) shows explicitly
that there is no field component in the R1-direction. Let ie and io
be unit vectors in the direction of increasing (3 and 0, respectively. The
transverse components of the scattered field are then

(74a)

(74b)

I= G(~@)]>4 [n x (p, x e,)]e-i(@R) dAS; (74C)

/ so

the vector p = ppl is the radius vector from O to the element of surface
dS. The total field at the point R, 6, @ is

I
E@ = E~@ + Ese =
%l:kyl~@@) (,5)

F,(@,@) = [GJ(@,@)]~~i@. et(@,@) ~ ie . I ;

{ I

Et = Ei@ + E.+ = R z=
-k [yYTi~
(@@ (7,,

F.J@@) = [G~(@3)]~+i~ o e; ~ i* I 1
{ }

The magnetic field is obtained by means of Eq. (71). The Poynting

vector of the total field is S = ~ Re (E x H*), and the power per unit
solid angle P(@@), radiated by the system as a whole in the direction
SEC. 5.9] APPLICATION TO POINT-SOURCE FEEDS 151

(@,@), is R21SI; hence, the gain function of the composite system is

or
G(@)@) = IF,I + IF,12. (77b)

The Line-source Feed.Line-source feeds are generally used with

a cylindrical reflector, the generating element of which is parallel to the
line source. The following analysis will be confined to such systems.

FIG.59.-The cylindrical reflector with a line-source feed.

The line source may be a system of point-scurce radiators distributed
along a line, such as the linear-array antennas discussed in Chap. 9,
or it may take the form of a long, narrow, rectangular aperture through
which energy is being radiated into space. It \vill be assumed that the
length 1 of the source is large compared \vith the ~vavelength.
The reflector and source system are illustrated in Fig. 5.9, with the
source along the x-axis. We shall assume that the maximum distance
from the latter to the reflector does not exceed 12/k and that the minimum
distance is large compared with the wavelength. Within such distances
from the source its field is essentially in the form of a cylindrical wave
(cj. Sec. 37). The wave incident on the reflector is, therefore, best dis-
cussed in terms of cylindrical coordinates. The z-axis in Fig. 5.9 serves
as the axis for the cylindrical coordinate system, the polar coordinates
of which, in the planes normal to the z-axis, are denoted by p and ~.
The cylindrical-wave zone can he divided into two general regions:
(1) a near-zone lwgion in the immediate vicinity of the source and (2)
a q{iasi-radiation zone at distances large compared \vith the wavelength
but less than l/k. In the latter region the predominant components of
Lhe field lie in the tangent plane of the cylindrical \vavefront and are
152 SCATlERiNG AND DIFFRACTION [SEC.59

mutually perpendicular as in the case of the isotropic cylindrical wave

discussed in Sec. 3.7. With a general line source, the field intensity is not
uniform over the wavefront but varies both along the x-direction and
about the cylinder axis. The radiation-zone field can be written

(78a)

and

(78b)

where PI is a unit vector in the direction of increasing p. The radiation

zone of the cylindrical wave field is to be distinguished from the general
far-zone field with respect to which the line source behaves like a point
source.
We shall confine our attention to fields in which the polarization is
uniform over the wavefront. Two fundamental cases are to be consid-
ered: (1) longitudinal polarization in which the E-vector is parallel to
the x-axis, so that
A(~,z) = A (~,z)i=, (79a)
and (2) transverse polarization in which the electric vector lies in the
planes transverse to the x-axis, that is,

A(~,z) = A(~,z)i$; (792))

the vectors i. and ir are the basis vectors of the cy~indrical coordinate
system. In most cases of interest the amplitude function A (~,z) is sepa-
rable in its dependence on the two coordinates. Referring to the power
flow rather than the amplitude, we shall introduce a twodlmensional
gain function G(V). Let P be the total power radiated by the source;
let (P/l)F(x) dx be the total power in the cylindrical wave field between
the planes x = consta.lt and z + dx = constant. The power radiated
per radian between thtse planes in the direction ~ k then

[80)

The gain function G(t) must obviously satisfy the cond;tion

The function F(x) expresses the distribution of intensity along the y-direc-
tion; it must satisfy th~ condition
1/2
F(Z) dx = 2, (82)
,/ 1/
SEC.5.9] APPLICATION TO POINT-SOURCE FEEDS 153

it being assumed that the origin of the coordinate system is at the center
of the line source. The amplitude A (z,*) may be evaluated in terms of
the power-distribution functions as in the case of the point-source feed.
The magnitude of the Poynting vector is

We ha.ve then

(83a)
or

(83b)

The current densities, or rather n x H,,, for the two types of polariza-
tion are the following:
1. Longitudinal polarization:

or

(84b)

2. Transverse polarization:

The angle i is the angle of incidence, and ~ is a unit vector tangent to

the cylinder in the cross-section plane. The positive directions of the
angles and vectors arc shown in Fig. 5.9.
The far-zone field is expressed in terms of spherical coordinates.
Because of the geometry of the system it is convenient to use a set of
spherical coordinates somewhat different from that used in the treatment
of the point-source field, the z-axis being taken as the polar axis; the
definition of the coordinates is given in Fig. 5.9. Let i@ and iz again be
unit vectors in the increasing 8- and @directions. For the case of the
longitudinally polarized source, the scattered field is
154 SCATTERING AND DIFFRACTION [SEC. 59

dsis the element of arclength along the cross section, or

s= [+(%)T- (87)

The scattered field has no @-component, the electric vector lying entirely
in the meridional plane passing through the z-axis. The scattered field
for the transversely polarized source is

In this case there is also a @-component proportional to sin o in

magnitude. However, if 1>> X, the beam is confined to the neighborhood
of @ = O and the cross-polarization component is small.
On expressing the vectors p and RI in rectangular components, one
finds the phase factor of the integrands of Eqs. (86) and (88) to be

P (e + ~i.) RI = P[l + cos @ COS(II + @)] x sin @. (89)

It is apparent that the integrands are factorable into a function of z
and a function of ~. Considering the central plane @ = O, we see that
the ~-dependence of the field arises entirely from the integral over ~
the field distribution is determined by the angular characteristic G(y)
of the source and the cross-section contour of the reflector. As regards
the planes 6) = constant, both aspects of the reflector contribute to
some degree. However, it will be shovm in Chap. 6 that if the length
1 is large compared with the wavelength, the major portion of the field
is confined within a small angular region @ about the central plane. Over
this region the variation of cos @ in Eq. (89) is of second order compared
with sin @; on setting cos @ = 1, separability is obtained, the field dis-
tribution in the planes @ = constant being determined entirely by the
linear characteristic of the source F(x). The transverse distribution of
the field is thus virtually the same for all transverse planes.
The primary field of the source alone must be added, of course, to
the scattered field to obtain the total field. Here the far-zone field
of the source (for which it is effectively a point source) must be used
instead of the cylindrical wave field of the radiation zone. It will be
assumed that this is known and expressed in a form similar to Eq. (72a),
in terms, of course, of the spherical coordinates showm in Fig. 5.9. It is
also assumed that the equivalent point so~u-ce is located at the origin
of the coordinate system, since the phase terms entering into the reflector
field have been referred to that origin. The procedure for superposing
the fields is exactly the same as that delineated in the previous case and
need not be discussed further here.
SEC.5.10] REACTION OF A REFLECTOR 155

Attention should be called to one point in the procedures discussed in

this section that has been the cause of some concern in the past. It wili
be found in general that the radiation field of a current distribution
such as is given by Eqs. (72) and (78)does not satisfy Maxwells
equations exactly. Consequently, except in special cases, the current
and charge distributions on the surface of the reflector, as found by the
methods already outlined, do not satisfy the equation of continuity
exactly. However, the terms that are neglected, which would result in
satisfying the required conditions exactly, are smaller in order of magni-
tude than the radiation field components and are in general in time
quadrature with the latter; they therefore introduce a nonessential con-
tribution to thescattered field and the scattered power pattern.
5.10. Reaction of a Reflector on a Point-source Feed.One of the
fundamental problems in the design of an antenna employing a reflector
is the effect of the latter on the impedance characteristics of the antenna.
The problem can be treated on the basis of re-radiation from the current
distribution on the reflector;1 the analysis will be carried out here for
the case of a point-source feed.
The radiating system that constitutes the feed must be considered
in its relation to a transmission line. The basic idea of the following
analysis is that the interaction bet\veen the feed and the field of the cur-
rent distribution on the reflector gives rise to a reflected wave in the
transmission line and thus an impedance mismatch from the point of
view of the line. Our object is to calculate the reflection coefficient
ratio of the reflected to incident wave amplitudesin the transmission
line due to the reflector. The field of the reflector, which is given in
general by 13q. (69), is regarded as a superposition of spherical wavelets
arising from every element of surface dil. The total reflected wave in
the transmission line is then considered to be the sum of component
waves arising from the interaction between the feed and the separate
wavelets. The current element K dS is regarded as a dipole source, and
only the radiation terms are retained for the individual wavelets. The
interaction between one of these and the feed is evaluated on the assump-
tion that the distance from the reflector to the feed is so large that the
wavelet can be regarded M a plane wave over the effective area of the
feed. This assumption is consistent ~vith our previous condition that
the reflector be in the far-zone of the feed system. Multiple scattering
between the feed and the reflector is neglected; this is likewise consis-
tent with the previous assumptions.
It will be assumed that in the absence of the reflector the feed is
matched to the transmission line; there is then only an incident wave
within the line. Let V, be the voltage at some reference cross section
1S. Silver, .<nal\-sisand ( orrwtion of the Impedance kfismatch Due to a
Reflector, RI. Report No. 810, Scptemhcr1945.
156 SCATTERING AND DIFFRACTION [SEC. 5.10

of the line. The total power transported across thereference cross see-
tion is
P = alv,l, (90)

where a is a constant characteristic of the line and the field distribution

over the cross section of the line. If the dielectric and ohmic losses in
the line and the radiating system are negligible, P is the total power
radiated by the feed. By Eq. (73), the current density at a point p,
O, @on the reflector, with its phase referred tot hat of the voltage at the
reference cross section in the line, is

K 2(nxHi) ?[:(;)GJ(e@)lnx (@,x

e,e-kp
g)
Expanding the vector term, we have

n x (PI x e,) = (no e%)p~+ et cos ~, (92)

where i is the angle of incidence. The current is here resolved into one
component along the incident ray and one component parallel to the
polarization of the primary field.
The field of the wavelet arising from the current element K CMis the
integrand of Eq. (69):

& [(n x HJ s V(V~) + Ic(n x H,)+] dS.

Applying the results of Eqs. (3.125) and (3.126), one sees that the radia-
tion field components arise only from the component of the current that
is transverse to the direction of propagation of the wavelet. Conse-
quently, to the order of approximation that all the other terms are
neglected, the component of the current in the direction pl contributes
nothing to the reaction on the feed. As regards the component in the
direction e, it is observed that this coincides with the polarization of the
feed and therefore no polarization obliquity factor enters into the inter-
action with the feed. The field intensity of the spherical wavelet with
which we are concerned is then

dE, = ~(nx H,). e~dS

or

Er= *v[~(;)Gf(@lcOsie
3)
The magnitude of the Poynting vector of the spherical walrelet is

ISrl = ; (;) ldE,12.

SEC. 5.10] REACTION OF A REFLECTOR 157

If the wavelet may be considered plane over the receiving cross section
of the feed [see Eq. (280)], the power that would be extracted from the
wavelet acting alone is

(94)

The voltage cZV, of the reflected wave set up thereby in the line, at the
given reference point, is

Vr= (Ye-i(z
=[L(;YGJ(@r~E
5)
The phase term 6 is a constant determined by the feed and the choice
of the reference point; we need not be concerned with its precise value.
Substituting Eq. (93) into (95), we obtain the reflection coefficient con-
tributed by the element of surface CMof the reflector:
~r_dv, _ 1
Gj(d, @) cos i e-~(kp+b)cM, (96)
vi 47rp
The phase term 6 absorbs the j of Eq. (93). The reflection coefficient
due to the entire reflector is, therefore,

(97)

Use will be made of this result in Chap. 12 to devise a method for elimi-
nating the mismatch.
The principle of stationary phase (Sec. 4.7) may be applied to the
integral of Eq. (97) to obtain an estimate of the mismatch for the case
of short wavelengths. It will be recognized that the phase is stationary
at those points on the reflector at which the rays from the feed strike at
normal incidence. The essential contribution to r arises from the area
in the immediate neighborhood of the stationary point. The calculation
is hardly different from that used for Eqs. (4.46) to (4.51) and will not
be given here. If G. is the gain of the feed in the direction of normal
incidence, P. the feed-to-reflector distance, R( and Rv the principal radii
of curvature of the reflector at the stationary point, we have
x
G.A R(R, N2kpn+ 6)
r.= (98)
87rp. [ (Ri + P.)(R, + P.) 1
If there is more than one point of normal incidence, the total effect is
obtained by summing the separate values of r..
The same result [Eq. (98)] can be obtained directly on the basis of
geometrical optics. The reflected field intensity at the feed is deter-
] S. Silver, Contribution of the Dish to the Impedance of an Antenna, RL
Report No. 442, September1943.
158 SCATTERING AND DIFFRACTION [SEC. 5.11

mined by considering the dispersion of a small incident cone of rays by

the reflector, making use of the techniques of Sec. 5.5.
5.11. The Aperture-field Method.It was shown in Sec. 3.8 that the
field at a point in space lying outside a surface that encloses all the sources
of the field can be expressed in terms of integrals of the field vectors over
the surf ace. Thus, if the scattered field E,, H. is known over any surface
z that surrounds the reflector completely, the scattered field at an
external point P in space is given by Eqs. (3.108) and (3109):

H,(P) = ~ , [jmc(n x E,)+

/
+ (n x H,) x Vi + (n. H,) v1] dtl. (99b)

In applications to antenna problems, the field over Z may not be

known. The aperture-field method formulates a high-frequency approxi-
mation to the field. The surface z is taken in the immediate vicinity
of the reflector, and it is assumed that energy passes to z from the
reflector by propagation along the reflected rays. The field over z is
then calculated by the methods of Sec. 55 in conjunction with l?q. (423);
the same conditions must therefore be imposed on the radii of curvature
of the incident wavefront and the reflecting surface.
The present method has no special advantages over the current-dis-
tribution method for the treatment of an arbitrary reflector. Ho~vever,
there is one class of reflectors for which it has decided advantages, both
in ease of application and in establishing relations with other phe-
nomena. The reflectors to ~vhich the method is particularly suited
and to which the subsequent discussion is restrictedhave the property
that the entire family of rays reflected from the illuminated area SO lie
in one hemisphere of space, as shown in Fig. 5.10; also, in the neighbor-
hood of the reflector it is possible in general to draw a finite curve r,
circumscribing the entire family of reflected rays. The shadow boundary
r on the reflector then defines an aperture and serves as an exit pupil for
the reflected rays, which can be regarded as arising from a distribution of
image sources behind the reflector.
On the basis of the ray diagram it is to be ~xpected that the scattered
field will be concentrated largely in the hemisphere of space containing
the reflected rays. Our discussion will pertain to points in this region,
and the surface 2 will, therefore, be taken to be made up of an infinite
plane containing a curve such as r., plus the hemispherical cap of infinite
radius. The aperture of the system may be defined as the area A on
the infinite plane circumscribed by the curve r. obtained by projection
SEC.5.11] TIIE APE RTFRE-FIELI) .IIETI{OD 159

of the shadow boundary r along the reflected rays. It may be noted at

this point that since the scattered field must satisfy the radiation condi-
tions [Eqs. (31 13) and (3114)] at infinity, the hemispherical cap will
make no contribution to the field integrals in Eqs. (99a) and (99 b).
It is evident that the determination of the scattered field over the
plane 2 by reference to the reflected rays leads to a discontinuous dis-
tribution, with a nonzero field over the area circumscribed by I~ and
zero field over the area of 2 outside r.i. This introduces into the problem
a featluv that is equivalent to the discontinuity in the current distribu-
tion over the reflector at the shadow boundary in the previous method.

x
P
r

-2

the aperture-field method.

It was pointed out at the close of Sec. 38 that the terms entering into the
integrands of l~qs. (99a) and (99/)) can be set into correspondence with
s~wface distributions of electric currents and charges and magnetic cur-
rents and charges, The electric and magnetic fields over the surface
cannot be assigned arbitrarily; they must be assigned in such a way that
the equi~alent current and charge distributions satisfy the surfac~ equa-
tion of continuity ~Eq. (3,9)] if the integrals are to give field components
that satisfy hIax\vells equations. In order to make the distributions
over 2 compatible ~vith the equation of continuity it is necessary to
introduce line distributions of electric and magnetic charges along the
boundary curve rA.
The computation of the boundary charge distriblltions proceeds along
exactly the same lines as in the case of the current-distribution methods.1
With E,, H, denoting components of the scattered field over 2, the
density u. of the boundary line distribution of electric charge and the

1SPCalso J, A. Stratton, l?leclro)~tagrte[ic Theory, LfcGra\v-Hill,New York, 1941,

Sec. 815.
160 SCA T1ERI.VG AND DIPFRACTION [SEC. 512

density u~of the magnetic charge are

(100)

The unit vector mandthe positive normal nto the surface 2 are defined
in Fig. 5.10. The amended expressions for the fields are then

+: 1~ [jti~(n x E,)+ + (n x H,) x V+ + (n. H,) vi] dS, (lOlb)

where A is the area enclosed by rA.

The integrals over the boundary r. can be transformed into surface
integrals by the same process used in transforming RI. (5.62a). It is
then found that the field expressions are

E,(P) = #& -, [k(n x H,)* + (n x H,) . V(V~)

i
+ jtic(n x E,) x Vi] dS, (102a)

H,(P) = ~~p -4 [k(n x E,)+ + (n x E,) . V(V~)

/
,iw(n x H,) x v4] dS. (102b)

The boundary line charges have the same effect here as in the case of
the current distribution: They cancel the longitudinal field component
of the far-zone field that arises from the surface current and charge
distributions.
The discussion has been developed with reference to a plane area.
This is not necessary for the application of Eqs. (99a) and (99 b); the
surface 2 may be any curved surface of infinite extent that divides the
space into two regionsone of the reflector and one of the scattered field.
The aperture area in that case will be a curved s~lrface bounded by a
curve r~ that is the projection of the shadow boundary along the reflected
rays. There is no change in the final result; the integrals ( 102a) and
(102b) apply to the curved aperture surface A.
5.12. The Fraunhofer Region.We shall now carry through the
reduction of the integrals for the far-zone field. The latter ~rillbe referred
to henceforth as the Fraunhofer region because of the relation of the
SEC.5.12] THE FRA UNIIOFER REGIO,V 161

problem to optical diffraction problems. The latter will be discussed in

a later section.
Let p again be the vector from the origin of the coordinate system to
the element dS of the aperture area and R, a unit vector from the origin
to the field point in the direction 0, 0. Applying the results of Sec.
3.11 and inserting therein the expressions for the electric and magnetic
currents given in terms of the fields by Eqs. (3.111), we find that Eq.
(102a) reduces to

~k nxE,
;~
,kRR1x
E,(P) = ~Re-
/[ .4
!
()
6
R, x(n xH,)
1
e~kQRldS. (103)

Let sbe a unit vwtoralong araythrough the aperture Inthegeomet-

rical-optics approi.mation, the electric and magnetic fields over the
aperture are related by

H = a(S xE). (104)

[In!rw spacea = (~/p) i. However, asweshallseelaterj the reflector

is only one special case of an aperture problem; the theory can be applied
to problems such as the radiation from hornsin which aw-ill have some
other wduc.] Substituting for H, in Eq. (103), ~veobtain

E(P) =
~k
~
~,jk.? RI x

H ~
nxE, a
()
~ [R,. (s xE,)n

(s x E,)(n . R,)]
I e~@RdS. (lo5)

For some purposes it may prow convenient to take as the aperture

area A the wavefront of the system of rays. In that case the unit
vectors s and n are identical, since the rays are normal to the wavefront.
More generally, ho~vever, it is convenient to use a plane aperture; the
vector n is then constant o~-er the surface and in the direction of the polar
axis of the spherical coordinate system. The field E, over the aperture
is gcnm-ally specified in terms of the polarization, magnit~ldcj and phase
distribution T(.r,y). If tlw \ravcfronts associated \vith the rays through
the aperture are the surf:wes l,(.r, y,z) = constant (cf. Sec. 4.2), the phase
distribution is
T(.I,7J) = L@L(l,y,o), (106)

where k~ = 27r/AOis the free-space propagation constant,. From Eq.

(4.10) it follows that the components of the rector s over the aperture
plane are
162 SCA TTERI.VG AND DIFFRA CTIO.V SEC. 5.13

1 al 1 a?,
s: == (1 s: S;)fl. (107)
z = ZaY; =lidy
The total power passing through the aperture is the integral of the normal
component of the Poynting vector:

The gain function for the aperture is therefore

(109)

It is overlooked in many treatments of aperture problems that if

there is a phase distribution over the aperture other than a constant
phase, the field vectors E,, H, do not lie in the aperture plane and the
Poynting vector is not normal to the plane. In cases where the p}lase
distribution v(z,y) represents small de~iations from constant phase,
these factors can be neglected without too serious an error. Subject
to this approximation, Eq, (105) simplifies to
_j,&,kR
E,(P) = (110)
=~R1 [( +WR)XN1)
where the vector N is

N= E.o,Q+!/,> .8.101(?,s,
Ere,h i,,,, (lloa)
/ A

The expression for p o RI for the plane apcrt~we has bwn inserted. In
using these relations it must bc kept in mind that the flr!d ~ertors are
assumed to lie in the aperture plane. Tle & ancl ~-components of Eq.
(110) are

DIFFRACTION
6.13. General Considerations on the Approximate Methods.-Both
the current-distribution and apert~uwfield meth{ds led t<) a calculation
of the srattered field as arising from a distril~llti~)il of s(nlrces over an
open surface, the boundary of J\hich is ddinml I)y Ihr system of reflected
rays. In contrast, to the ~et}mc,t]i(:~l-tll>ti[s mtthtx~, the field at anv
point was found as the sllperposition ui (~]ntril]~[tions {lore all elements of
SEC,5.13! G1l.YERAL CO.YLTIDERA710.YS 163

thesource clistribution. Ingeneralj therefore, the last two methods will

lead to nonzero field intensities in the region of space not covered by
the system of rays; also, in the region of the rays, the fields will differ from
those obtained on the basis of geometrical scattering. These deviations
from geometrical propagation of the scattered field are known as diffrac-
tion phenomena.
The diffraction effects are due fundamentally to the fact that the
sources are distributed over an open surface; that is, the reflected wave-
front is not a closed surface. The same effects will arise no matter by
what process a field distribution is generated over a finite open area in
space. Thus a lens illuminated hy a point-source or line-source feed
likewise defines an exit pupil for the system of rays incident in it and
produces a segment of a warefront in its aperture plane. Lenses and
reflectors that have aperture areas of the same size and shape and produce
the same field distributions over the apertures have field patterns that
differ in no essential detail. The same phenomena are observed when a
wave passes through an aperture in an infinite opaque screen or through
the mouth of a horn into free space.
Experiment shows that whenever the dimensions of the aperture are
large compared with the Ivavelcngth, the diffraction effects are small and
the major portion of the field pattern is concentrated in the region covered
by the rays from the aperture. On the basis of this fact a common high-
frequency approximation technique is used for all problems of the type
mentioned above. The mathematical details have already been devel-
oped in Sees. 5 11 to 5.12, and we need only summarize here the general
ideas in the application of the results to the various types of problems. In
each case the aperture area is associated with a surface 2 of infinite extent
\vhich divides all space into two separate regions. The problcm is then
equivalent to that of an aperture in an infmitc screen on the surface Z, It
is assumed that the field o~-cr 2 is zero every\~here except over the aper-
ture area; in cffert, it is assumed that diffraction effects at wide angles
with respect to the aperture-ray system are negligible. In the case of the
reflector and lens it is assumed that the aperture field is produced by geo-
metrical reflection or refraction of the rays from the primary feed. In the
case of a horn the aperture field is taken to be that which would exist over
the aperture area in a horn of infinite extent-possibly after correction is
made for reflection from the opening. In the infinite screen problem,
the aperture field is taken to he that which exists over the area in the
unperturbed [rave in the ahsrnce of the screen.
As was pointed out in Sec. .511, the calculation of the diffraction field
is based on the integrals of the field equations obtained in Sec. 3.8 by
means of Greens theorem. Ho~~ever, the application of C,reens theorem
was predicated on certain assumptions concerning the continuity of the
164 SCATTERING AND DIFFRACTION [SEC.514

distribution over the surfaceassumptions that arenotfulfilled by the

distribution over Zinthe approximation technique. The method should
rather be considered to be based on the Huygens-Fresnel principle which
postulates that each point on a wave surface is a source of elementary
fields (cf. Sec. 4.1), and the results of the Greens theorem integration
are to be regarded as furnishing the appropriate identification for the
sources as stated in Eqs. (3.111). The requirement that the source
distribution must satisfy the equation of continuity then leads to the
addition of a line distribution of electric and magnetic charge along the
boundary of the aperture surface. Thus Eqs. (lOla) and (1016) and
hence Eqs. (102a) and (102b) derived from them apply to all diffraction
problems in the high-frequency approximation method; the fields E,, H, are
to be interpreted quite generally as the fields over the aperture surface.
6.14. Reduction to a Scalar Diffraction Problem.In many antennas
the field over the aperture is almost completely linearly polarized, only
a small fraction of the energy being in the cross-polarization component
of the field. If the latter is neglected, the calculation of the diffraction
field is simplified; by a further approximation, consistent with the
high-frequency approximations made already, the problem can be reduced
to a scalar diffraction problem.
The analysis will be restricted to a plane aperture; the aperture will
be taken in the zy-plane as in Fig. 510, and the electric field will be taken
to be polarized in the z-direction.
It was pointed out in Sec. 4.1 that the integrals of Eqs. (99a) and
(99b) can be transformed into

provided that 2 is a closed surface over the whole of which E and H

are continuous. Equations (1 12a) and (112b) are each a set of three
equations for the three cartesian components of the field vectors; the
normal derivative d/dn is applied component by component. If z is an
open surface, as in the case of the diffraction problem, a similar trans-
formation can be effected; additional terms appear that vanish in the
former case of integration over a closed surface. The reader can verify
that for the aperture the integrals transform as follows:

& / ~ [jw(n XW + (n xIV x V$ + (n. E) v+]cLS

SEC.5.14] REDUCTION TO A SCALAR DIFFRACTION PROBLEM 165

a similar expression holds for Eq, (99 b). The line integral around the
boundary is different from that of the line distribution of charge but
arises from the transformation of the surface integrals into one another.
The diffraction field is the sum of the contribution of the surface integral
and the line distribution of charges on the boundary. The complete
expression is then

+ :$,. 4(E x ~) ds

- ~+$rA V*(T . H) ds. (1I4)

If the field over the aperture is linearly polarized \vith, say,

EV=E. =O,
the surface integral contributes only to E.(P). As is seen in Fig. 5.11,
this leads to a component of the field in the direction of propagation.
x E%(P)
\
\ ,.-

$%
/f t
e
z
t

@
v
FIG, 5.11.ILcduction to a scalar diffraction problem.

For a given angle d, the order of magnitude of this longitudinal component

is at most EZ(P) sin 6. The vector E x ~ is normal to &he aperture.
It therefore gives rise only to a component E.(P); the contribution of the
latter to the transverse field is again proportional to sin 0. iYow the
high-frequency approximation method is based on the assumption that
the diffraction field is contained almost entirely in the region of small
values of 6; therefore in the significant region of the field the longi-
tudinal and transverse components arising from the surface integral and
E x ~ respectively are negligible, and the surface integral may be taken
alone to calculate the transverse field. As regards the longitudinal com-
ponent introduced by E x ~, it will be recalled that the third integral
of Eq. (110) was such as just to cancel the longitudinal component of the
field introduced by the first t\vo terms. Therefore, the last t\vo integrals
of Eqs. (110) virtually cancel each other for small angles O.
The diffraction field is thus given by the scalar integral formula

(115)
166 SCATTERING AND DIFFRACTION [SEC.514

where u stands for the particular component of the field involved. It

will be recognized that this is the Kirchhoff diffraction formula used in
physical optics.1
In the geometrical-optics approximation the field in the region of
theaperture has the form (Sec. 4.5)
u = .4(Z,y,z)eikd~(l,u,a) (116)

whereA(x,y,z) istheamplitude and L(z, g,z) = constant aretheequiphase

surfaces. Then

(117)

If the wavelength is short, k, is large and the second term may be neg-
lected in comparison with the first:

au
m= jk,un . vL. (118)

The field over the aperture is usually given in terms of the amplitudr
A(z,y,z) andthephase distributionW(z,y) = kaL(z,y,O). Ifsis theunit
vector in the direction of ray at a given point on the aperture, we have
byllq. (4.10)
k,VL =ks; (119)
then
au
= jkun . s. (120)
an
The components of s in terms of the phase distribution w(z,y) have been
given previously in Eq. (5.107),
With regard to d~/dn it is observed that

(121)

where rl is a unit vector from the point on the aperture to the field point.
Collecting the terms in Eq. (1 15), we obtain

I See, for example, }1. Born, Optzk, reprint by Edwards Bros., Ann Arbor, Mich.,
1943.
z This covers the general case in \vhichthe wavelength in the region of the aperture
differs from that in free space.
SEC.515] BABIAETS PRINCIPLE FOR ELECTROMAGNETIC FIELD 167

For the far-zone field the customary approximations are made with
regard to e~~r/r. In addition, n . r, = n o R, = cos O is constant over
the aperture, and l/r is negligible compared with j%. The far-zone ficlcl
is, therefore,
jk _.kE @@.R, (n . s + Cos O) ds,
uP=el (123)
4.R ! A

where p is the vector from the origin to the surface element dS. Atten-
tion should be directed t(~the n . s term. Only if the phase distribution
over the aperture represents a small deviation from constant phase can
n . s be set equal to unity with little error. In that case, we have

up = ~ e-jkR (1 + COS6) ~ ~eiW.R,ds, (124)

41rR /
5.15. Babinets Principle for the Electromagnetic Field.Consider-
able progress has been made during the past few years in obtaining rigor-
ous solutions of diffraction problems. Discussion of these would carry
us beyond the scope of this \vork. Attention must be called, however, to
the relation of Babinets principle to the electromagnetic field that results
from the solution of the problem of the diffraction of a wave by a plane
metal screen of infinite conductivityy. 1
It will be well to recall the form of Babinets principle as it applies
to a scalar wave field. 2 Suppose that ~ve have a plane opaque screen in
the xv-plane; Am is the area covered by the screen and .40 is the aperture
area. The complementary screen is defined to be that covering the area
A Oand having aperture area Am. In both cases let there be an initial
field u, arising from sources in the negative z-region of space, and let
UI and U2be the diffraction field produced in the positive z-region by the
respective screens. The optical Babinets principle states that the sum
of the two complementary fields at any point is equal to the initial \vave
amplitude at the point in the absence of any screen:

Ui = UI + U2. (125)

This relates the problem of diffraction around a metal sheet of finite

area to the diffraction of a \vave through an aperture of the same size
a,nd shape in an infinite plane sheet.
The principle for the electromagnetic field is fundamentally different
in that the initial fields are complementary as ~vell as the screens. I,et
E; = F, H, = G be the initiai field arising from sources in the negative
z-region in the case of one of the screens, and let E 1,H, t)c the diffraction
1H. C.. Booker, (13abinctsPrinriplc Mid +,hcTheory of lksonmt Slot~, TIW;
(Great Britain) Ikport No. 29, Decemhcr 1941; E, T. Copson, I%oc. Roy, ,Sor., A,
186, 100 (1946).
ZNf. Born, Op. nut.
168 SCA TTERINQ AND DIFFRACTION ~SEC.
515

field in the positive z-region. Let EL = G, Hi = F be the initial field

in the case of the complementary screen and Ez, Hz its diffraction field.
Then
E,+ H2=F, (126)
H, H,=G. (127)

The incident field for the complementary screen is rotated 90 with

respect to the first field, and the complementary relation exists between

?vm~;
;~j<j n-cl
l/_
(a) (b)
FIC.5.12,llelationbetweena slot anda dipoleradiator.

the electric and magnetic field vector of the respective diffraction fields.
This principle leads to a usefui relation between the radiation field
of a slot and that of a dipole. Let S be a slot in an infinite plane con-
ducting sheet, excited by a generator across its center as shown in Fig.
5.12a. The complementary dipole is a similar thin metal strip (Fig. 512b)
energized by a generator across an infinitesimal gap at its center. The
field vectors over the slot are perpendicular to the corresponding field
vectors in the case of the dipole. It then follows from the Babinets
principle that the radiation field of the slot is the same as that for the
dipole, but with the electric and magnetic field vectors interchanged; full
details of the proof will be found in the paper by Booker.
 CHAPTER 6

APERTURE ILLUMINATION AND ANTENNA PATTERNS

By S. SII.VER

6.1. Primary and Secondary Patterns.The discussion of aperture

systems will be continued in the present chapter with the object of
developing in more dettiil the [elations between the aperture field and the
diffraction field. The results ~vill furnish a basis for the design of the
reflectors and lenses used in directive microwa~e antennas. The design
considerations for such systems fall into two major groups: (1) trans-
formation of the specifications that the radiation pattern of the antenna
as a whole is required to meet into requirements on the aperture-field
distribution, and (2) the design of the primary feed and reflector or lens to
produce the required aperture field. The radiation pattern of the com-
posite antenna will be referred to as the secondary pattern L1 distinction
to the primary pattern of the feed system.
It must be kept in mind that strictly speaking the secondary pattern
is a superposition of the diffraction field from the aperture ani the field
of the primary feed (cf. Sec. 51). Microwave feeds, however, are
designed to have such directivity that the major portion of the;r energy
is directed into illuminating the optical device. The overlapping of the
field of the primary feed and the diffraction field gives rise, therefore, only
to second-order effects; these will be treated in later chapters in the
discussion of specific antennas. The secondary pattern can thus be
resolved into two parts: (1) the diffraction field of the aperture and (2)
the portion of the primary feed field that is not intercepted by the optical
system. Specifications imposed on the secondary pattern (intensity level
relative to peak intensity) in the second region are therefore require-
ments imposed on the primary feed pattern in addition to the require-
ments pertaining to the production of a desired aperture field.
6.2. The Diffraction Field.The discussion \vill be restricted to a
plane aperture and will be based on the scalar field approximation devel-
oped in Sec. 5.14. It is therefore being assumed that the field over the
aperture is uniformly polarized in one direction, which, to fix our ideas, is,
say, the z-direction, the aperture being taken in the zy-plane (Fig. 6.1).
Let the coordinates of a point in the aperture be f, q and those of a
field point P be r, y, z. It will prove convenient to change the notation
somewhat from that used in the preceding chapter. The field over ihe
169
170 APERTURE ILLUMINATION AND ANTENNA PA TTER.VS [SEC. 62

aperture will be designated by F(.$,q); A (~,~) will be the amplitude distribu-

tion, and ~(~,~) the phase distribution, so that
F((,~) = A (.f,q)e-jv(~~). (1)

The method of determining the system of optical rays through the aper-
ture associated with the phase distribution T ( ~,v) was discussed in Sec.
x

Y
FIG. 6.1.On the diffraction field.

5.12; the explicit relations between the components of a unit vector s

along a ray and the phase function are given in Eq. (5.107). If rl is a
unit vector in the direction from the aperture point (L,q) to the field point
P, then according to Eq. (5122) the diffraction field U, is given by

The diffraction field may be divided into three general zones which
are determined mathematically by the nature of the approximations
that may be made in the integral [Eq. (2)]. The three zones are also
differentiated by the structure of the field, but it should be noted that the
boundaries of these regions are not sharply defined.
First there is the new--zone region of points in the immediate neighbor-
hood of the aperture for which no simplifying approximations can be
made in Eq. (2). Although the dimensions of the aperture are large
compared with the wavelength-an assumption that underlies the use of
Eq. (2)there is in general, for a given field point in this region, an
appreciable area of the aperture for the points on which the I/r term in
the brackets of the integral is not negligible compared with k = 27r/A.
The region extends several wavelengths outward from the aperture, tm.cl
it will be readily appreciated that this is not exactly infinitesimal for the
SEC.6.2] THE DIFFRACTIO,V FIELD 171

wavelengths of the orders of magnitude of the microwave region. Also,

for the near-zone region, the variation of i,. rl over the aperture for a
given field point must be taken into account. The integrations are in
general dificult to carry out, andadetailed study of theintegral for this
region is beyond the scope and purpose of our present discussion. In
such cases where the field has been worked out in detail it has been found
that the near-zone field is determined essentially by geometrical propaga-
tion along the aperture-ray system with fluctuations in intensity over the
phase surface due to interference effects; the mean value of the intensity,
however, differs little from that of the geometrically propagated field.
The shadow region boundary is quite sharply defined.
Attention should be called to the fact that the scalar diffraction
integral [Eq. (2)] can at best yield only qualitative results for the near
field zone. In this region the contribl~tions of the line integrals along the
aperture boundary in I?q. (5, 114) will make significant contributions to
the field and must be taken into account if the results are to have a
quantitative value.
From the near zone we pass into the region of the diffraction field
which we shall call the optical-Fresnel field by virtue of its correspond-
ence to the Fresnel region of optical diffraction problems. Several
simplifying approximations are introduced; the orders of magnitudes of
the errors involved must necessarily be evaluated for each case separately,
To start \vith, the term I/r in the brackets of Eq. (2) is considered to be
negligible with respect to k; at a distance of several wavelengths from the
aperture this approximation is reasonable. Second, the variation of
(i. . r,) over the aperture is neglected, and the term is replaced by the
constant iZ. R, = cos d, where RI is a unit vector directed from the origin
to the field point. A third approximation in the same category is to
neglect the variation of the l/r term outside the brackets; it is set equal
to the reciprocal distance l/R from the origin to the field point.
The variation of r over the aperture must be treated more carefully
in the phase term elk. lVe have in general
~ = [(z {)2 + (y 7)2 + 22]J;. (3)
If the field is concentrated in the region around the z-axis, a distance z
from the aperture tvill be reached at which for the points in the significant
region of the field z >> I.r ~1, Iy TI. Equation (3) can then be
expanded as follows:

r
~z+(zt) +(vd+. ..=z+ ,a +..
. . (4)
22 22
Terms higher than the second order are neglected in the Fresnel field
approximation. An alternative form of expansion is obtained by express-
ing the coordinates of the field point in spherical coordinates:
172 APERTURE ILLUMINATION AND AiVTENVA PA T1ERNS [SEC. 62

x= Rsin Ocos@=Ra,
y= Rsin Osin@=R~, (5)
z= Rcos O. /

Introducing these into Eq. (3), we obtain

~R_(at+ Bn)+t+v(a:+D7)
r R+Tb, (6)
2R

neglecting terms of order higher than the second. It is seen that this
assumes a~/R <<1, Pq/F-i <<1. The expansion in the form of Eq. (4)
is suited for discussing the field over planes z = constant, \vhereasEq. (6)
is best for discussing the field over a sphere of radius R about the origin.
Since both expansions actually assume that the field is concentrated in
the neighborhood of the z-axis, there is no significant difference between
the results obtained with one or the other. The diffraction integral for
the optical-Fresnel region thus becomes

(7a)

or

(7b)

Eqs. (7a) and (7b) differ from the expressions for the Fresnel field
generally found in the literature in the presence of the term i. . s ivhich
arises from a nonuniform phase distribution over the aperture. It is to
be noted that a phase distribution which represents ~}-idedeviations from
constant phase has associated with it a highly dispersed system of rays.
Under such conditions the assumption that the energy in the diffraction
field is concentrated around the z-axis is not valid and the approximations
entering into Eqs. (4) and (6) may not be justified. If, however, the
phase distribution represents only small deviations from uniform phase,
the deviation of the rays from a system of parallel rays that are normal
to the aperture is small; the term i, . s may then be treated as constant
and equal to unity over the aperture.
The Fresnel region is characterized by the onset of diffusion of the
field and the wavefront outside the boundaries defined by the extension
of the rays through the aperture. The latter, however, still define the
propagation of the major portion of the field; further details of the Fresnel
region will be developed in later sections.
With increasing distance from the aperture we finally pass into the
Fraunhofer or far-zone region of the field. This is the region \vith ~vhich
the secondary pattern is concerned. The far-zone approximations ha\-c
been discussed a number of times before. In the present connection it
SEC. 6.2] THE I) IFPIM CYIO,Y FIELD 173

will be noted that the Fraunhofer region is differentiated from the

Fresnel region by further approximations that are made in the phase term
~~~r:the Fraunhofer field approximation neglects al~ terms in Eq. (6)
above the first order in the aperture coordinates therehy considering in
Fig. 6.1 that the unit vector rl is
parallel to R,. ~Te have then

Like all other far-zone fields en-

countered previously, the Fraun-
hofer field is a quasi-point source
field. The field distribution is the
_I12- --zhL
l(a) l(b)
same over all spheres about the
origin; in a given direction e, +
the amplitude varies monotonically
J+l ](c)
+ (d)
as I/l? and the intensity as I/Rz.
Again if the phase error orer the
aperturedeviations from constant
phaseare small, the i,. s term may
-+4& -+ 1
be replaced by unity. Equation
(8) then becomes
(9)
I:r~. 6.2,Trwls]tion from lresnel LO
Fraunhofer diffrart,on for a slit; (a)
$.f) depirt the field distribution across planes
]n the Frcsnel region at inrmasing di. tances
from the At, showing progressive cfaffusioll
of the field into the shadow region; (Q) IS the
Iraunhofer pattern. (Reproduced from .1.
C. SMer and .+. H. Frank, I?LlrodtLclio?Lto
Theoretical Ph@cs, JfcGraw-Hill, ,Yew
It will be found that with nearly uni- York, bg courtesy of the authors. )
form phase over the aperture, al-
most all of the energy in the field is contained in a small angular re~ion
about the %axis (corresponding to the geometrical property that the aper-
ture rays are all parallel to the z-axis). The variation of cos o over the im-
portant region of the secondary pattern may then be neglected, and }ve have
as our final approximation

CP = & e-) l(g,q)eik.~.t~o. $+~~n$d~ dq. (9)

/ .4

Equation (9) is frequently used indiscriminately for both small and large
phase errors over the aperture. This \villI)e done in the present chapter
and it should be remembered that for the latter cases the results ha I:Conly
qualitative value.
174 AP.ERYuRE lLL~l~INAYIO.V A.VD AAVTE.V.I.4P.4 TTER.%S [SEC. 63

It was pointed out earlier that the boundaries of the three regions
of the field cannot be sharply defined. It is clear that the passage from
the Fresnel approximations to the Fraunhofer approximations is a gradual
one and is determined to a large extent by the criteria of the acceptable
error in the approximations that are made. In later sections n-e shall
attempt to define an inner boundary for the Fraunhofer region on prac-
tical considerations for special types of apertures. The gradual transi-
tion of the physical characteristics of the field from one region to the next
is illustrated very nicely in Fig. 6.2 taken from Slater and Frank. The
figures pertain to a sIit over which the field is uniform in amplitude and
phase. The near-zone pattern (Fig. 6.2b) is seen to consist essentially
of the column of radiation propagated geometrically from the aperture.
With increasing distance from the aperture the field diffuses into the
shadow region, the system of parallel aperture rays finally passing over
into a cone of rays in the Fraunhofer region.
6.3. Fourier Integral Representation of the Fraunhofer Region.The
final approximate expression that was obtained for the Fraunhofer region
[Eq. (9)] has an interesting interpretation. Let us define

k= = k sin 0 cos ~, (lOa)

k. = k sin Osin rj; (lOb)
Eq. (9) then becomes

Up = +Re-]kg(k=,kv) (1 la)

with

g(k.,k,) = ~ F(/,q)ej@f+k~~) df dq. (llb)

/
Consider the plane z = O. The aperture field can be regarded as the
function u(z,y) over the entire plane:

u(x,y) = F(x,y) inside .4,

(12)
u[z,y) = o outside A. )
The function U(Z,V) is stepwise continuous over the entire plane and cam,
therefore, be expressed as a Fourier expansion in the form of the Fourier
integral:

a(%Y) = *2
111 lu(~)ek(-z)ek(rv)d~d~d~d~. 13)
.. -m.
or
mm

U(x, y) = ;T g(kz,kv)e)@~tk@) dkZ dkv, (13a)

//
..
1zr2tr(ld?/rf
iOn(0 TheOrft/r/1[ Xrw
I/iJ/sic..,31cGra\y-I1ill, York, 1!)33,Chap. 27.
 SEC. 64] GENERAL ~EA T~fW5 OF THE SECO.VDAR 1- PATTERN 175

with

..
(13b)

It will be observed that except for the factor of l/27r Eq. (13b) is identical
with Eq. (1 lb).
Let us now examine l?q. (13a). If ~vedefine a vector k,

(14)

the function e~k satisfies the wave equation and represents a plane wave
of unit amplitude traveling in the direction of the vector k. Over the
plane z = O, the wave produces a distribution

The integrand of Eq. (13a) is thus the distribution over the plane z = O
produced by a plane wave in the direction k \vith an amplitude g(k.,ku),
and the arbitrary distribution u (.~,y) is given by Eq. (13a) as a super-
position of plane waves traveling in all directions. Referring to 13q. (] la)
it is then seen that the amplitude of the field in the Fraunhofer region
in the direction defined by k= and ky [Eqs. (10)] is the amplitude of the
plane wave component in that direction which enters into the synthesis
of the arbitrary distribution over the aperture.
Equations (13a) and (13b) are referred to as the pair of mates of a
Fourier transform. If the function g(k.,kv) is given, that is, the Fraun-
hofer field is prescribed completely both as regards to amplit~lde and
phase, Rq. (13a) serves to determine the field clistriblltion over the plane
z = O that is required to produce the prescribed secondary field pattern.
In practice the use of the transform is limited by the fact that the second-
ary pattern is prescribed only in power; the phase of g(k,,liu) can be assigned
at ~vill,and therefore the aperture distribution is not determined uniquely.
Two different choices of the phase of g(k,,k,,) lead to t$vo different
aperture fields, one of ~vhich it may be physically possible to produce,
whereas the other may not be realizable physically at all.
6.4. General Features of the Secondarv Pattern.-The results of
later sections will be anticipatcxl here ~vith-a general summary of the
relation between the secondary pattern and the apert,ure field, Consid-
ering Eq, (8) or (9) again from the point of \~ie\v(If the sllperposit.it)n of
contributions from each element (If s~lrfa,cc on the aperture, the field
 176 APERTURE ILL UlfI.VA TIO.V AND A.VTE.V.V.4 PA TTAR.YS [SEr, 6.4

at a given point is visualized as the resultant of a system of vector

elements. The magnitude of the vector from the element of surface
at a point ~, q is 11(~,q)[ d~ d~; the angle that it makes with the element
from the origin (which is taken as a reference) is determined by the
intrinsic phase difference between them over the aperture and the phase
difference arising from different path lengths to the field point. The
absolute maximum value that the resultant of the system of vectors can
have is equal to the sum of their magnitudes, obtained when the contribu-
tions are all in phase. If the phase over the aperture is constant, the
absolute maximum is attained in the direction normal to the aperture, for
in that direction the path length is the same from all aperture points
to the field point. Since the path-length phase factor

is a linear function of the coordinates on the aperture, the absolute

maximum cannot be obtained in any direction in the case of arbitrary
phase distributions over the aperture unless the distribution is a linear
function of the aperture coordinates, In general, however, there will
always exist directions in space for which the path phase factor makes the
optimum compensation for the aperture phase differences between the
elements as compared with neighboring directions. The secondary
pattern thus has series of maxima and minima. If the phase distribution
does not deviate too widely from constant phase over the aperture, there
will in general be one maximum that is considerably greater in value
than the others. The portion of the secondary pattern possessing this
maximum and contained within the angular region bounded by the
directions of the adjacent minima is know-n as the main lobe or sometimes
as the main beam. The subsidiary maxima arc referred to as side lobes.
The line through the origin and the peak of the main beam is referred to
as the beam axis.
From the practical point of view the pattern is specified by certain
beam characteristics: the direction of the peak intensity; the gain, half-
power, and tenth-power widths of the main lobe; and the magnitudes and
positions of the side lobes. To define the beam widths consider any
plane containing the axis of the beam; the half-poirer \vidth (3 in that
plane is the angular distance between the two directions about the axis
in which the power radiated per unit solid angle is one-half the peak
value; the tenth-power width (3(Ilm) is defined correspondingly. If the
aperture is symmetrical in shape and the field distribution over the
aperture has certain symmetry elements in common with the aperture,
the main lobe will reflect the symmetry of the field distribution. The
symmetry elements arc generally planes of symmetry; these are referred
to as the principal planes of the pattern.
SMC.6,4] GhNllRAL F.EA T LRAX OF THE SBCONBA R Y PA TTE2L41 177

Gain.Let us consider first the relation between the gain and the
aperture field. The power radiated per unit solid angle in a given direc-
tion is [cf. Eq. (5.77a)]

(16)

The total power Pa radiated by the aperture is equal to the power flow
across the aperture, which is the integral of the normal component of the
Poynting vector. The total power is then

(17)

and the gain function in the direction (o, O) is

The more exact form [Eq. (8)] has been used in Eq. (18) for the field
intensity CP in the Fraunhofer region.
It was seen that if the phase is constant over the aperture, the second-
ary pattern attains the absolute maximum in the direction of the z-axis,
6 = O. The aperture rays are parallel to the z-axis so that s, = 1; the
maximum value of the gain function, or simply the gain, is, therefore,

(19)

A case of especial interest is that of uniform illumination over the aper-

ture. F(i,q) is a constant; from Eq. (19) the gain GO for that case is
found to be

Go . 47rA
~. (20)

( onsider now any other intensity distribution. Making use of the

Sch\~artz inequality,

(21)

where ~ and g are any two functions; by taking j = F(~,q) and g = 1,

\vefind
178 APERTURE ILLUMINATION AND A.VTE~VNA PAT1ERNS [S~c. 6.4

Hence,
4mA
G.,, < ~. (23)

Thus, the uniform field distribution over the aperture gives the highest
gain of all constant-phase distributions over the aperture. The ratio
s = G.M/G,, known as the gain factor, may be regarded as the efficiency
of the aperture in concentrating the available energy into the peak
intensity of the beam.
The proof given above that uniform illumination gives maximum gain
is valid strictly for constant-phase distributions only, since Eq. ( 19)
applies only to such distributions. A proof for the more general case
must be based on Eq. (18); so far, to the authors knowledge, no such
proof has been established. If the phase distribution represents a small
deviation from constant phase, however, and is such that the peak
intensity lies in the direction 6 = O, it is certain that the gain is less than
that of the uniform field. The value of the peak intensity is more
sensitive to the interference effects between the vector elements from
the aperture than is the value of the total Power to the slight deviations
of the aperture rays from the normal to the latter. The effect of such
phase errors is, therefore, a reduction in the aperture efficiency.
The aperture efficiency can be given more pictorial significance by
considering the performance of the antenna system on reception. Let
us suppose for the moment that the primary feed is designed to illuminate
the reflector or the lens but to have no radiation in other directions. In
that case the secondary pattern arises entirely from the aperture, and the
gain of the antenna is equal to the aperture gain G,,. If now a plane
wave is incident on a matched antenna along the beam axis, by Eq.
(2.80), the absorption cross section presented by the antenna to the plane
wave is

(24)

From Eq. (20), it is then seen that if the aperture is uniformly illuminated,
the absorption cross section is equal to the physical cross section presented
by the antenna to the incident wave. In the case of any other type of
constant-phase illumination we have

(25)

The effective area is reduced by the gain factor. The aperture efficiency-
may thus be regarded as meas~lriug the efrecti~-e aperture area presented
by the antenna to the incident Irave.
In the practical case the primary feed radiates in directions other
than that required to illuminate the optical device. The energy not
SEC. 6.4] GENERAL FEATURES OF THE S.5ONDAR Y PATTERN 179

intercepted by the latter is referred to as the spill-over energy. Except

for secondary effects the total power in the secondary pattern can be
considered as consisting additively of the power in the Fraunhofer region
of the aperture diffraction field and the spill-over energy of the feed.
The over-all gain of the antenna must be referred to the total energy
radiated in all directions. If Pm is the peak intensity in the secondary
pattern and P is the total power radiated by the feed, the over-all gain is

~ _ 4.Pm
(26)
P

If the aperture intercepts a fraction a of the total polver from the feed,
the power radiated by the aperture is Pm = CYP, whence

G=a~= aG~. (27]

a
Taking again the uniformly illuminated aperture with an idealized feed
as the reference, the over-all efficiency of the system is

(28)

The efficiency of the antenna is thus seen to be a product of two factors:

(1) the fraction of the total power intercepted by the optical device and
(2) the eiliciency of the aperture in concentrating the available energy
into the peak of the main lobe.
Beamwidths and Side Lobes.The beamwidth and side-lobe character-
istics are, of course, intimately related to the dependence of gain on
the aperture field distributions. The follo~ving remarks are based on the
results of the investigation of a number of special cases. Taking the
constant-phase distribution first, we have seen that maximum gain is
realized with uniform illumination. If the illumination over the aperture
is modified so that the intensity is peaked in the central area of the
aperture and tapered down in magnitude toward the aperture boundary,
the diminution in gain is accompanied by an increase in beamwidth and a
decrease in side-lobe intensity relative to the peak intensity of the main
lobe. The prominence of the side lobes can be traced to the discontinuity
at the edge of the aperture, considering the field distribution with regard
t,o the plane z = O as a whole.
The effect of phase errors over the aperture, with the types of aperture
fields that are commonly encountered, is in general to reduce the gain
and broaden the main lobe. Side-lobe levels may be either raised or
depressed depending on both the type of phase distortion and the inten-
sity distribution over the aperture, Quite generally the sharpness of the
minima is reduced and their levels are raised. Severe phase errors ovel
180 APERTURE ILL UJf INA TION AND A.VTE.VNA PA lTER\S [SEC. 6.5

the aperture may result in splitting of the main lobe and enhancement
of the side lobes to such an extent that it is no longer possible to identify
a major lobe.
6.5. The Rectangular Aperture.-A number of special problems
associated with rectangular and circular apertures will be investigated
to illustrate the general ideas formulated in the preceding section. The
rectangular aperture will be treated first. Let the dimensions of the
aperture be designated by a and b, the orientation of the aperiure in
the xy-plane being shown in Fig. 6.3. The secondary pattern depends
only on the relative distribution over the aperture, and in the following
discussion it will be assumed that the distribution I( g,q) is normalized
to have a maximum value of unity. For the present purposes the com-
pletely simplified expression for the Fraunhofer region [Eq. (9]] will
be used; in so far as the pattern is concerned we need consider only the
factor

Uniform Amplitude and Phase.F( L,q) = 1 for a uniformly ilhlmi-

nated aperture; the integral of Eq. (29) is easily evaluated, and one
finds that the secondary amplitude pattern is

g(e)($) =

The patterns in the principal planes (the zz- and yz-planes) are of par-
titular interest. For the zz-plane @ = O, Eq. (30) simplifies to

g(e,o)
<,n
= A -- ( ~a .
sm 8

7ra
A ). (31)
sin e
x

For the yz-plane + = 7r/2; the pattern in thk plane is likewise given
by Eq. (31) with a replaced by b. Both patterns are of the same form,
sin u/uj but are scaled in the angle e according to the aperture dimensions
in the respective planes. The secondary power pattern, normalized to
a peak value of unity, is plotted in Fig. 6.3 on a logarithmic scale as a

function of the reduced variable u = ~ (r/h) sin e. The minima in

()
this case are eq~lal to zero and occur at the points u. = m, n = f 1,
+2, . . . The full \vidths of the main lobe measured from null point
SWJ.6.5] THE REC7ANGVLAR APE R7(IRE 181

Loo
0.8
0.6

0.4

0.2

0.1
0.08
0.06

(*f
Iz
004

0.02

0.01
0.008
0.006

0.004

0.002

0.001
o 2 4 6 8 10
u
FIG. 6,3.Secondary pattern of a ulliforlnly illuminated rcx%:uwularaperture.

to null point on either side of the axis are

zz-plane: 2 sinl
()~

A
2A
= ,
a

yz-plane: 2 sinl ~ = ~.
()
The half-power point on the main lobe is very closely at u = 1.39; hence
the half widths in the principal planes are

zz-plane: @ = 2 sin
() 1.39A

~a
1.39A
= 0.88$ (32a)

yz-plane: @ = 2 sin] ~ = 0.88 ;. (32b)

()
182 APERTURE ILLUMINATION AND ANTENNA PATTERNS [SEC. 6.6

These illustrate two fundamental points: (1) In a plane of symmetry the

width of the beam is determined by the aperture dimension in that plane,
and (2) the diffraction pattern is confined to a smaller angular region the
larger the dimensions of the aperture as measured in wavelengths.
The side lobas (peaks) are located at the points u- that satisfy the
relation u = tan u. The first of these comes ai u, = 4.51; the second at
U* = 7.73. The side-lobe intensities relative to the peak intensities are
readily found to be 1/(1 + u;), which, from the values of UI and uZ, is
seen to be very nearly equal to 1/u~. Referring to the amplitude
expression [Eq. (31)], it is seen that g(0) is positive over the entire main
lobe, changing sign in passing through the first zero, returning to a
positive value on passing through the second zero, and so on. The
odd-numbered side lobes are, therefore, out of phase with the main lobe,
and the even-numbered ones are in phase. Such phase reversals are
characteristic of all power patterns in which the minima are equal to zero.
Separable Aperture-field Distributions.A common type of aperture-
field distribution is that which arises with a cylindrical reflector or lens
and a line source (cf. Sec. 59) where the distribution over the aperture is
separable into a product of two functions:

F(t,v) = ~l(.wz(~). (33)

Substituting into Eq. (29), we find that the integral is likewise separable:

If we consider again the principal plane patterns, Ive see that the pattern
in a given plane is determined entirely by the field distribution along the
corresponding aspect of the aperture. The principal plane patterns are

zz-plane: g(0) =

yz-plane: g(~) =
u b/2

b/2
./2
F,(q) dq

F,(t) d(
U
a/2

./2
b/2
~l(f)ei~t.in~ d~;

F,(T)e~k~Slu@dq.
(35a)

(35b)
[/ ./2 U b/2

The effects of tapered illumination and phase errors on the principal

plane patterns can thus be studied as two-dimensional problems, provid-
ing, of course, that the aperture field is separable in the form of Eq. (33)
both in amplitude and in phase.
6.6. Two-dimensional Problems.The remaining analysis of the
secondary pattern of a rectangular aperture ~vill be restricted to a separ-
able distribution m which the field is uniform along, say, the ydirection;
that is, F2(7) = 1. The pattern in the plane z = O is just the sin u/u
pattern,
SEC. 6.6] TWO-DIMENSIONAL PROBLEMS 133

d
sinsin e
g(e) = () 1-
~b ,
sin e
T
and we are left with only the pattern in the plane y = O,
a/2
g(8) = _a,2 Fl(&)ti~*~d& (36)
/
to consider. Multiplicative constants are being ignored in Eq. (36).
It is convenient to introduce new variables

2t ~=~asinfi;
~=, (37)
a
the function F1 (f) goes over into a function j(z), and g(0) becomes a
function of u which to avoid difficulties of notation will be designated
as g(u). Equation (36) then becomes
1
g(u) = ; _, j(2)eju= dz. (38)
\
It is seen at once that if the same relative distribution, for example,
j(z) = [1 (4~/a)] = (1 ~z), is produced over two apertures of
different size, the two apertures will produce the same secondary patterns
when regarded as functions of u. The side-lobe intensities relative to the
peak intensity will be the same in the two cases. However, since
sin o = ku/~a, the angular distributions will differ; the diffraction field
of the larger aperture will be contained in a smaller angular region than
that of the smaller aperture, and in particular the main lobe will have a
smaller beamwidth. The larger aperture will yield higher gain, cor-
responding to the fact that the pattern is confined to a smaller angular
region in space. This can be seen directly from the expression for the
gain [Eq. (19)]. For the present we shall consider only constant-phase
distributions. Equation (19) reduces to

(39a)

for the separable type of distribution. On introducing the variable z,

this becomes

(39b)

showing explicitly that the gain is proportional to the area of the aperture.
184 APERTURE ILLUMINATION AND ANTENNA PATTERNS [SEC. 6.6

The distribution over the aperture can be characterized completely

by its moments w~,
1
pm = _ ~X.f(z) dz (rn=o, 1,2, ...). (40)
/
These are very useful in relating the properties of the secondary uattern
to those of the aperture distribution. Expanding the exponential in
Eq. (38) we obtain

(41)

or

(42)

The power pattern p(u) = Ig(u) I is then

()= (~)[l-ti-i)u2+(a-% +a)u4-l

(43)

It is seen from Eq. (42) that an asymmetrical distribution over the aper-
ture results in a g(u) that is complex so that the equiphase surfaces in
the Fraunhofer region are not spheres centered at the origin. If, how-
ever, the aperture distribution is symmetrical, that is, j(z) is an even
function, its odd moments vanish and g(u) is real:

g(u) = 2
(
#Gu ~_&u2+M~4
2/lo 4 !#O
.,-
). (44)

For the latter case, convenient expressions for the beamwidth can be
obtained by simple approximations. In the neighborhood of the beam
axis, we shall approximate the pattern by neglecting all terms in Eq. (44)
beyond the second:

(u)+( -:H2) (45)

The half-power point in the po}ver pattern corresponds to point Q at

which the amplitude has allen to 1/ti~ of its peak value. Hence

I R. C. Spencer,Fourier Integral Methodsof PatternAnalysis, ILLReport ~o.

762-1,Jan. 21, 1946.
SEC. 6.6] TWO-DIMENSIONAL PROBLEMS 185

and the half-power width.is

@) = 2 sin-l [[(2 - @i) ;];)

or

(46)

The effect of tapering the illumination down toward the edges of the
aperture can be seen directly from this expression. Since the moment
M is the average of the distribution function weighted by the factor d,
peaking the function in the neighborhood of x = O decreases the second
moment more rapidly than po which is the average of the function itself.
The effect of such tapering is to increase the ratio pO/~* and hence to
increase the beamwidth.
A more accurate expression for the beamwidth has been obtained, 1
which can be used to construct the main lobe down to its tenth-power
width. The results are applicable only to the cases of symmetrical
aperture distributions. For the latter the expansion of the power
pattern [Eq. (43)] reduces to

(47)
f() =l-(9u+w)+&l~ 4-

The factor (~Oa/2)2 has been dropped to normalize the pattern to a

peak value of unity. The power drop in the pattern relative to the peak
expressed in napiers is N = in p(u). Considering Eq. (47) to be of
the form 1 z, the expansion ff~rin (1 z) is used to obtain

(48)

Solving for u, \vethen get

u = AN~J(l BN), (49)

where
~=(:
) ; =+(-%0 (49a)

Since hr p(u) = 2,303 iog ~(v), the corresponding expression in terms of

the decibel drop D is
u = A~D)!(I _ BJD), (50)
where
A = 1.518A; B = 2.303B. (50a)
1R. C. Spcncrr, op. cit.
186 APERTURE ILLUMINATION AND ANTENNA PATTERNS @C. 6.7

The full angular width of the main lobe at a given decibel level is, there-
fore,

8. = z sin- ~+ (1 ED); . (51)

[ 1
The apertur~field-secondary-pattern relationships are further illus-
trated by the results in Table 61, in which the major secondary-pattern
characteristics are given for several typical aperture distributions. The
integration of E~. (38) is easy to perform for each of these distributions,
and the details need not be given here. The effect of reducing the
disc~mtinuity at the edge of the aperture is shown by the series of para-
bolic distributions 1 (1 A)z2. It is seen that the gain decreases
rapidly as A gets in the neighborhood of zero; and the beam width increases.
The series Cosm(mz/2) shows the effect of a higher-order taper of illumina-
tion. All members of the series (n = 1, 2, . . .) reduce to zero at the
edge of the aperture, but in addition the nth member has n 1 deriva-
tives equal to zero at the edge of the aperture. The gain decreases and
the beamwidth increases with increasing n; the side lobes appear at
increasingly larger angles and with reduced intensity relative to the main
lobe.
6.7. Phase -emor Effects.-A phase-error distribution may arise over
the aperture of an optical system from various causes such as a displace-
ment of the primary feed from the focus or distortion of the reflector or
lens, or it may be caused by phase error in the field of the primary feed;
that is, the wavefront is not spherical or cylindrical as is presupposed in
the design of the optical system.
It will again be assumed in the following that the aperture field is
separable, the field being uniform in the y-direction and the phase error
existing in the z-direction only. If v (2 f/c4) = V(Z) denotes the phase-
error distribution, the expression for the secondary pattern [Eq. (38)]
becomes
1
g(~) = ; _, f(z)@=-W@l ~~, (52)
/
where ~(z) now denotes the amplitude of the aperture field. As in the
preceding section, ~(z) is assumed to be normalized to unity.
The discussion will be limited to a consideration of special forms of
*(z), specifically to the following:1
Linear error: V(Z) = Px.
Quadratic error: V(Z) = @z.
Cubic error: V(Z) = QX.
1The results presented here are taken largely from R. c. Spencer and P. M.
Austin, Tables aud Methods of Calculation for Line Sources, RL Report No. 762-2,
Mar. 30, 1946.
TABLE6.1.--+ECONDAEY PATTERN
CHAR.I
CTERISTICS
PRODUCED
BYVARIOUS
TYPESOF
APERTUREDISTRIBUTIONS.
1
j(z) = 1 Izl >1
g(u) = y
=0 121>1
m
-1 +1
Full width Intensity of first
Angularposition
Gainfactor S. at half power side lobe; db below
0 of first zero
(F),radians peak intensity
A
1 o.88~ 13.2
a
j(z) = 1 ,71 A)zz, IzI <1
g(u) = a
p? [ %+1
-A A(S%)l
(2 + A)2
(A) = 911 :(1 A) + i(l A)*]
A=l. o A
1 0.88~ 13.2
a i
0.8 0.994 0.92~ l,06~ 15,8
a a
0.5 0.970 o.97~ 1.143 17.1
a a
0,0 0,833 1.15; 1,433 20.6
a
f(z) = Cos; 1%1<1
n!cosu
g(u) = :n_l n, odd;

: (2k + 1) 4*]
k=O [
n! sin u
g(u) = a n, even
j u
~~1 (2k)~ 4$2
+& [ 1
4 2.4. 6.. (n-l)Z
S=>[l .3.5,., 1
2.4. 6n. . . 2n
n, odd
( 1.3. 5.. .2ll )
1.3. 5. (n-l)
= [2,4.6... 1
(n+2)(n +;)... 2n
[ (n+l)(n+3). ..2ll 1 even
A
rt=o 1 o.88~ 13,2
a i
1 0.810 1,2: 1.5; 23
2 0.667 1,45; 2? 32
a
3 0.575 1.661 2.5: 40
a
0 515 1.93A 3; 48

j(z) :1 121,121,<1
sin ~
g(u) = 4a +
4A
() 2
0.75 1.28: 2? 26.4
a
188 APERTURE ILLUMINATION AND ANTENNA PATTERNS @~c. 6.7

Linear Error. -Inserting the appropriate expression for V(x) into

Eq. (52) we obtain
1
9(U) = ~ _, ~(~)ejt-~jz dz. (53)
/
It is directly evident that thk is of the same form as Eq. (38) with u
replaced by (u ~). The pattern is, therefore, the same as that of the
constant-phase distribution but displaced by an amount /3. Ihe peak
intensity comes at u = (3, that is, in the direction

~. = sin_, p.
(54)
ar
The pattern is thus the secondary pattern of a constant-phase aperture
field rotated through the angle 00. The physical basis for this is very
simple. On expressing the phase distribution in terms of the original
aperture variable ~ = az/2 and making use of Eq. (5107), it will be
found that the aperture rays form a system of parallel rays traveling in
the direction 00 given by Eq. (54). The aperture field can be considered
to have arisen from a plane wave incident on the aperture in the direction
60.
If the aperture is projected onto a plane normal to the aperture rays,
a new aperture is obtained over which the field distribution has constant
phase. The projected aperture dimension a is

a = a cos O.,

and therefore the gain G~ in the case of linear-phase error will be related
to the gain GM of the constant-phase distribution by

G; = GM cos 00. (55)

This can be verified by a direct calculation on the basis of Eq. (18).

Quadratic Error.The secondary amplitude pattern for this case is
1
9(U) =; _,
j(x)ei(w+) dx. (56)
/
The evaluation of such integrals is generally laborious. For small
phase errors, however, a convenient approximate method can be used.
Expanding the exponential factor e@, we obtain

(57)

The integrals of Eq. (57) can be expressed as derivatives of the pattern,

go(u) obtained in the absence of phase error (p = O), for
SEC. 67] PHASE-ERROR EFFECTS 189

(58)

Hence

(59)

Retaining only the first two terms, we have

g(u) == ; [go(u) + jP91(u)l. (60)

If the amplitude distribution F(z) is symmetrical, g,(u) is real; the power

pattern p(u) is then
~z
p(u) = ~ {[90(U)12+ PII (u)]}. (61)

The effect of quadratic phase errors is illustrated in Fig. 64 for two

types of illumination: uniform amplitude, f(z) = 1, and tapered illumina-
tion, f(z) = cosz (mE/2). In both cases Q = 7r/2, representing a path
length deviation of A/4 from constant phase at the edges of the aperture.
The value of D is ratner large for the use of Eq. (61) to be valid, but the
qualitative features are not seriously affected by the errors involved.
It is seen that the peak intensity still appears in the direction O = O.
Since the phase error is symmetrical with respect to the center of the
aperture, the secondary pattern will always be symmetrical about the
0 = O axis. However, it will be found that when B gets sufficiently large
the main lobe becomes bifurcated, with maxima appearing on either side
of the o = O axis. The general effect of the phase error is to raise both
the side-lobe level and the level of the minima. In the case of the tapered
illumination these effects are so large that the first side lobe is almost
completely absorbed into an extremely broadened main lobe. The effect
of phase error on gain is exemplified by Fig. 6.5 which shows the gain
relative to the constant-phase distribution for a uniformly illuminated
aperture. The phase error is expressed in terms of path length deviation
from constant phase. The loss in gain that can be tolerated in practice
depends, of course, on the operational requirements on the antenna and
the associated system.
Cubic Phase Errors.The cubic phase errors can be treated by the
same approximation technique as was employed in Eq. (60). The
corresponding expression for the ampl; tude pattern is

9(U) = ; [90(U) + d:(u)]. (62)

If)o APERTURE IJ.L1lM INA TION AND ANTENNA PA TT17R.VS [SEC.6.7

In the caw of a symmetrical distribution f(x) over ~he aperture go(u) is

real and the power pattern is

P(u) = : [90(u) + @g;(u)]. (63)

If j(z) is an even function, go(u) is likewise even and hence g;(u) is odd.
In the neighborhood of u = O g;(u) is positive for u >0 a~d negative
for u <0. It is then directly evident from the form of p(u) that the
peak will occur at some value u >0. The effect of the phase error is to

3-

5-

\
\\
10
\
\
\
15 - \
I
\
I i
20
\
I

25 -
1/ \
1/ \
I I \
300
u 31
(b) 2z

FIG. 6.4.Effect of quadratic phase error; maximum phase er,or of 7r/2 at the edge of tl]e
aperture: (a) constant amplitude; (b) tapered illumination, f(z) = COS2(7rz/2).

tilt the beam as in the case of a linear error. In addition, however, the
main lobe becomes asymmetrical and the side lobes increase on the side
of the main lobe nearer 0 = O and decrease on the other side of the main
lobe. The shift in the main lobe is also accompanied by a loss in gain.
Aperture Blocking.The problem of an obstacle in the aperture is of
interest in connection with the use of reflectors, for the primary feed is
located in the path of the reflected rays, thus blocking out a portion of the
aperture. The obstacle may be considered as a particular type of phase
error. Assuming that over the exposed area the presence of the obstacle
does not alter the distribution f(z) which would exist in its absence, the
obstacle can be regarded as producing a field 180 out of phase with f(z)
over the area that it covers.
SEC. 6.7] PHASE-ERROR EFFECTS 191

Let us consider the particular case illustrated in Fig. 66a of an

obstacle located in the center of the aperture. The width w of the
obstacle will be taken to be small compared with the total aperture width
a. To express the pattern in ~w
\ o
terms of thevariablesof Eq. (37),
\
we define the normalized width \
90 \
26 = wfa. Equation (38) for the \
pattern then becomes \
80 \
6 \ - 1
\
9(U) = ~ _l f(x)edz
[/ ,~ 70 g
1 \
\ *;
+ ,f(x)ejzdx , (64) ~ \
/ +6 1 60 \
which can be rewritten as
\ ~
1 50 \
\
9(U) = ~ _lj(z)e~dz \
[/
+6 40 J4
j(z)euzdz (65)
/ -6 1 0 0.1 0.2 0.3 0.4 0.5
Fhaseerror
inwavelewth
Ym
It is seen explicitly that the ob-
F1~. 6.5.Loss of gain as a function cd
stacle can be regarded as an out- quadratic phase error over a uniformly
of-phase field superimposed on the illuminated aperture.
original distribution. Over the region of the obstacle maybe con-
sidered constant and equal to unity; hence
1
g(u) =; (66)
[/ f(~)e~~zw$d)1
Since the width of the obstac~is small compared with the aperture width,
the pattern produced by the former will be very broad compared with
that of the aperture. For qualitative results the obstacle pattern may be
regarded as constant over the region of the main lobe and near in side
lobes of the aperture pattern. The effect of the obstacle is then simply
that illustrated in Fig. 66b of subtracting a constant 26 from the original
amplitude pattern. If the peak amplitude of the original pattern is
1
f(x) dx
a. =
/ 1
and the amplitude of the first side lobe is paO, the intensity of the first
side lobe relative to the peak in the modified pattern is
,_~
_ao 28 aO
(67)
Pfio+w=
P+:.

1R. C, Spencer,(FourierIntegral}Iethods of PatternAnalysis, RL Report No.

761-1, Jan. 21, 1946.
192 APERTURE ILL IJMIiVA TION AND ANTE,VNA PA lTERNS [S... 6.8

The effect of the obstacle is to increase the magnitude of the first side
lobe.
6.8. The Circular Aperture.-The fundamental considerations and
results developed for the rectangular aperture pertaining to the relation
between the aperture field and the secondary pattern apply in general
to the circular aperture, but the quantitative details differ because of the
difference in aperture geometry. In treating circular aperture problems

/
f y)

-1 -26- 1

(a) (b)
FIG. 66,-The effect of aperture blocking: (a) modified aperture distribution; (b) second-
ary pattern.

it is usually convenient to use polar coordinates P, O (1%. 6. i), which are

related to ~, q by
.g = p Cos 4, q = p sin $. (68)

Denoting the aperture field distribution by F(pj +), the expression for
the secondary pattern [Eq. (9)] becomes
2T a
g(e, r#l)= F(P,@)ejP~ino~ <$-@Pdp d~, ((69)
IIo 0

where a is the radius of the aperture. Introducing the variables

Zlm
~=f; U=~Sln O= T~sin O, (70)
a

the function F(P, o) goes over into a function ~(r, o), and g(d, O) goes
over into a new function which ~veshall denote simply as g(u, O). It ~vill
be assumed, as before, that -f(r, @) is normalized to unity. The pattern
is then
27 1
g(u, ~) = a2 f(r, +~)eiur m (++Jr & d,$t. (71)
/! o 0
SEC.68] THE CIRCULAR APERTURE 193

It is observed that as in the case of the rectangular aperture, all apertures

having the same relative distributions produce the same secondary
patterns regarded as functions of w The angular distribution in @ is
the same for all, reflecting the symmetry of the distribution over the
aperture; as seen from Eq. (70), the distribution in 0 again scales by the

100
0.8
\ x
0.6

, -&
0.4 R
P o
z

0.2 4
Y

0.1
0,08
0.06

0.04
I \
[Al(w)]
0.02

0.01
0.008 I \

0.006 \

0004
(

0.002 / \

0.001
2 4 6 8 10
u
FIG. &7.-& condary patternfrom & unirormly illuminated cirrularaperture.

factor k/D; the larger the diameter the smaller is the angular spl-ead
of the pattern about the (0 = O)-axis.
Unijorm Phase and Amplitude. Setting f(r,+) = 1 in Eq. (71) and
carrying out the integration over ~, we obtain
1
g(u) = 2ra2 r~~(ur) dr, (72)
/ o
194 APERTURE ILLUMINATION AND ANTEN.VA PATTERNS [SEC. 643

where .JO(UT) is the Bessel function of order zero. 1 The integration over
r leads to

~(u) = ~=a, A().

(73)
u

The power pattern p(u) normalized to unity is shown in Fig. 6.7, plotted
on a logarithmic scale, as a function of u. The half width of the main
lobe is

@ = 2 sin-
()
0.51 ~ = 1.02 ), (74)

and the first side lobe is 17.5 db down from the peak. These are to be
compared with the half width of 0.88A/a and the 14-db side lobe of the
secondary pat tern of a rectangular aperture.
Tapered Illumination .The effect of tapering the illumination down
toward the edge is the same as with a rectangular aperture: reduction
in gain, increase in beamwidth, and reduction in side lobes. The effects
can be illustrated by considering the series of aperture field distributions
(1- T)~, p=l,2, . ... The secondary patterns are given by
1
gp(u) = 27ra2 (1 r)~o(ur)r dr (75)
/ o
or
Z1p!Jp+,(u) _ ~a if +,(u)
g,(u) = rrag ~P+l (75a)
p+l p

The functions Ap are available in tabular form. 3 The major character-

istics of the patterns are summarized in Table 6.2 and will not be dis-
cussed further.
The circular symmetry of the secondary pattern is associated of
course with the corresponding symmetry of the field distribution. It is
of interest to consider a distribution of the form

,f(r, r$) = 1 r Cos @l (76)

which is tapered in the plane O = O and uniform in the plane O = 7r/2.

Substituting 13q. (76) into Eq. (71), \re obtain

I G. N. Watson, Theory of Bessel Functions, 2d cd., Macmillan, New York, 1945.

ZR, C. Spencer, Paraboloid Diffraction Patterns from the Standpoint of Physical
Optics, RL Report T-7, Oct. 21, 1942.
3E. Jahnke and F. Erode, Tables OJ Functions, reprint b.y Dover Publications,
New York, 1943.
SEC.6.8] THE CIRCULAR APERTURE 195

TABLE62.-SEC0NDAR%PATTERN CHARACTERISTICS
PRODUCED
BY A DISTRIBUTION
(1 P)POVER A CIRCULAR
APERTURE

~, ~, half-power 0, position
First side lobe,

I
P gain factor db below peak
width of firstzero
intensity

I sin_l 1.22X
o 1.00 17.6
D
~in_, 1.63A
1 0.75 24.6
D
Sin_l 2. 03A
2 0,56 30.6
D
Sin_, 2. 42x
3 0,44 . .
D
sin_, 2. 79X
4 0,36 ....
D

This can reevaluated by means of theexpansionl

(78)

of particular interest arethepatterns intheplanes ofsyrnmetry, @ = O

and d = rr/2. We find that these are

d)=o,

_ 3rra
~ Az(u). (79)

(u)= 2maw
-W[- +J(uldl
(80)

Thetwopatterns areshown in Fig. 68. It is seen that the beamwidth

is greater in the plane @ = O than in the plane @ = m/2, corresponding
to the fact that in the first principal plane the illumination over the
aperture is tapered whereas in the seocmd the aperture illumination is
uniform.

lWatsOn, op. cit., Sec. 2.22.

196 APERTURE ILLUMINATION A,VL) A,VI.V.VAVA PA TTER.vS [SEC. 6.9

6.9. The Field on the Axis in the Fresnel Region.An important

consideration in making measurements of secondary patterns is the
minimum distance from the aperture at which the field may be regarded
as being in the Fraunhofer region. To aid in arriving at a criterion it

1.0 I
8
I

4 \
I
t, -4=$ 1
1
2
i 1
I ,
0!10
8 I \
,1
I
6 - 1
\
! \\
I ,. .,
\
~ 1 \
Ii

: ~ ! ~, J ;,
0.010
8 ,, /..
1, , I
6
~~ I
\ j ,,
II: I
4 J
\
,! \
,, !
:! \
\ J
2 II 1
;; 1
(
I 1
O.wlo I I
! Ii
\
8
1 !,
6
I ;1
4 1 1 \
;;
I
lr \
2

II 1

2 4 6 8 10
U=&sino

FICJ,68.-Principal plane patterns for the aperture distribution f(r, +) = 1 rz Cosz+.

will be well to discuss briefly the field on the axis in the Fresnel region
and the transition to the Fraunhofer region. The aperture field will be
taken to be uniform in amplitude and phase.
The method of Fresnel zones used extensively in optics affords a
simple physical basis for understanding the effects that are observed

I
S~C. 6,9] 7HE FIELD O.V THE AX1,T I,V THE FRI?SNEI, REGIO,V 197

in the Fresnel region. In Fig. 69 let the point P at distance R on the

axis be the field point under consideration. Taking the point P as a
center, weshall describe afamily ofspheresofradii R +A/2, R + 2(x/2)
. . . . Their intersections with the aperture divide the latter into annu-
lar regions; these are known as the Fresnel zones. The zones will be
numbered as shown in the figure. Taking any two adjacent zones n and
n + 1, it is seen from the method of construction that for every contribu-
tion to the field at P arising from an element of surface in the first zone

FIG. 6.9.Division of the aperture into Fresnel zones.

there is a contribution from an element in the second zone 180 out of

phase therewith, and the integrated contributions of the two zones are,
therefore, very nearly 180 out of phase with one another. Denoting
the magnitude of the contribution of the nth zone by Sn, the effect of the
entire aperture is
S=sl s2+s3 s,+----- (81)

The contributions l% decrease slowly with increasing n; the resultant

effect of pairs of adjacent zones is therefore virtually equal to zero. A
careful analysisl shows that if the aperture contains a full number of
zones N, the resultant is very closely equal to
s = ;(s, f s.) (82)
depending upon whether N is odd or even. As R increases, N decreases
and S fluctuates between the values
S = ~(~1 S,) = O for N even (83a)
and
S = ~(Sl + A,-) = S, for N odd. (83b)
The amplitude of the field along the axis, therefore, passes through
maxima and minima, the maxima coming at the points that subtend an
odd number of Fresrml zones, the minima at the points subtending an

] See, for example, M. Born, Opfik, p. 145,reprintby EdwardsBros., Ann Arbor,

Mirh., 1943.
198 APERTIJRE JJ,J,~rM INA TJON AND ANTENNA PATTERNS @N. 6.9

even number. There will be no further fluctuations beyond the point

on the axis for which the entire aperture consists of a single Fresnel zone;
this distance is
D2 _ ~2 D2
R,=x (84)
= G
D being the diameter of the aperture. However, the distance Rj cannot
be taken as the beginning point of the Fraunhofer region, for at that point
the contribution from the edge of the aperture is still 180 out of phase
with that from the center; whereas the calculation of the Fraunhofer
I I I 1
I

(a) (b) (d)

FIG. 6.10.Thetrmsition regionbetweenthe Fresneland the Fraunhofer regions: (a)
R= R<= D2/4k; (a) R = DZ12~;( c) R = D2/i; (d) R = m

region, on axis, assumes that the path differences between points on the
aperture to the field point are negligible. Considering the aperture to be
subdivided into small annular zones and resolving the resultant effect
of the aperture into the superposition of the vector elements of these
zones, one finds that the vector diagrams take the forms shown in Fig. 6.10
for distances greater than R,. The slope angle of the vector diagram
at the terminal point is equal to the phase difference between the edge
of the aperture and the center, corresponding to the difference in path
length to the field point. At a distance greater than D2/2A (Fig. 6. 10b)
there is no longer any cancellation between horizontal components of the
vector elements; at a distance D2/k the resultant is a good approximation
to the value for R = m.
To make a more quantitative evaluation we must consider the actual
values of the field intensity and the gain. For this purpose we will
start from the Fresnel approximation [Eq. (7a)], which in the present case
takes the form

(85)

It will be recognized that this is equivalent to the expression for the

on-axis field intensity in the Fraunhof er region of an aperture having a
quadratic phase error p2/2R. Equation (78) is easily integrated for
uniform illumination giving

~TP= 2j sin
()
~ ~jkR . (86)

IThe vector diagrams depict the variation of only the form factorsof the field
that is, the integralsof Eqs. (7)with increasingdistance R.
S.C. 69] THE FIELD ON THE AXIS IN THE FRESNEL REGION 199

This corresponds to radiated power per unit solid angle

m= xY4R2[sin(a12
or

(87)
~=i(i)+(%} =~
The total power radiated by the aperture is simply ~(c/p) ~~A,whence the
gain is

(88)

The factor (sin Z/Z) Zexpresses the ratio of the gain measured at a distance
R to the gain Go of the true Fraunhofer field at infinity. The accepted
values of the minimum distance at which pattern measurements may
properly be made vary between R = D/A and R = 2DZ/A. The
values of the gain ratio of Eq. (89) for the two cases are

There is a little difference between them; for the cases most commonly
encountered the 2D2/x criterion is to be favored. Other considerations
which are discussed in Chap. 15 also point in that direction.
 CHAPTER 7

MICROWAVE TRANSMISSION LINES

BY S. SILVER

We have dwelt at considerable length on general theoretical consider-

ations underlying the design and operation of microwave antennas as a
whole. We now enter upon a program of studying the components of an
antenna, starting with an investigation of microwave transmission lines.
Usually about a foot, or perhaps two, of the line immediately preced-
ing the radiating system is at the disposal of the engineer for the insertion
of matching devices to compensate for the impedance mismatch of his
antenna; this section will be referred to as the feed line. The following
discussion of feed lines will be confined to elementary transmission-line
theory and problems; for more extensive treatments, in particular for the
anal ysis of matching dev~{ es, the reader is referred to the sources indi-
cated below. ]
7.1. Microwave and Long-wave Transmission Lines.-A brief com-
parison of long-wave and microwave lines was made in Sec. 1.4. It was
pointed out that the use of unshielded parallel \virelines becomes imprac-
tical at microwave frequencies largely because the power-carrying
capacity is so sharply limited by the small interline spacing required
if the line is not to radiate. The relation between the interline spacing
and radiation follo~s from the ideas developed in Chap. 3. We may
consider the alternating current in a wire as a line distribution of oscillat-
ing dipoles; corresponding points on a pair of wires carrying equal and
opposite currents are occupied by similar dipoles in opposite phase. If
the spacing between the dipoles is small compared with a wavelength,
their radiation fields will be out of phase at all points in space and annul
each other. On the other hand, if the spacing is comparable to the wave-
length, the double-dipole system can radiate, there being directions in
space for which path-length differences will compensate for the intrinsic
phase difference of the members of the pair. In addition, it should be
noted that large interline spacings can be used at long wavelengths, since
the radiation-field intensity of a dipole varies inversely as the square
of the wavelength [cf. Eqs. (3.148)].
1 J. C. Slater, .Wicrowaue Transmission, McGraw-Hill, New York, 1941,Chaps.3, 4;
R. L. Lament, Waoe Guides, Methuen, London, 1942: Ilontgomcry, Purcell, and
Dickie, The Principles of .Ificrowave Circuits, Vol. 8; and N. Marcuvitz, The Waveguide
Handbook, Vol. 10, of this series.
200
SEC. 72] PROPAGATIO>Y IV WAVE
GLlDES 201

Another hindrance in the-use of unshielded lines is their susceptibility

to interference, which would cause serious installation difficulties. If
the line is not to radiate, its elements must be symmetrically disposed
with respect to near-by conductors, in order that equal and opposite
currents be maintained at all paired uoints on the line. Perturbation
of the line balance would also give rise to impedance difficulties.
It accordingly becomes clear that shielded lines are required for
tra.lsmission at microwave frequencies. Two types are in use: (1) two-
conductor lines consisting of one conductor surrounded by a second,
separated by dielectric, and (2) hollow metal tubes. These lines are
to be considered as waveguides for electromagnetic waves in the enclosed
dielectric rather than as transmission lines carrying current and voltage
waves. In fact, in the hollow waveguideit is not possible to establish
definitions of line current and line ~oltage that ar~ comparable to the
quantities defined in a parallel wire line; under special conditions current
and voltage can be defined for the two-conductor line. However, while
a new approach must be taken in the fundamental analysis, it is found
that under suitable conditions transmission-line analogues of voltage and
current can be defined and use can be made of the line theory summarized
in Chap. 2.
7.2. Propagation in Waveguides of Uniform Cross Section.We shall
confine our discussion to lines of arbitrary but uniform cross section.
that is, waveguides with cylindrical
x
walls. The guide walls will be taken
to have infinite conductivity; the di-
electric in the interior \villbe assumed
to be homogeneous, ~rith dielectric ~n
z
constant c, permeability p, and zero
conductivity y; the dielectric will also
be assumed to be free of charge. d
We shall consider electromagnetic YH
fields in these waveguides which have ~IG.7.1.On waveguide propagation.
a harmonic time dependence; they
will satisfy the homogeneous field equations obtained from Eqs. (332) by
setting the source functions and the conductivity equal to zero.
We shall take the z-axis of the coordinate system to be parallel to the
generator of the cylindrical walls of the waveguide. Since the guide is
homogeneous in structure along the z-direction, a wave of a single fre-
quency will depend on z only through a phase factor and possibly a
damping factor corresponding to progressive attenuation of the wave.
That is, the z-dependence of all field components is of the form e~~z,where
~ is possibly complex:
202 MICRO WAVE TRANSMISS1ON LINES [SEC. 7.2

With the convention that a and @ are both to be positive quantities, the
upper sign in the exponential corresponds to propagation in the positive
z-direction, the lower sign to propagation in the negative z-direction.
Writing out the field equations in component form and taking into
account the postulated form of the z-dependence, we obtain for a wave
traveling in the positive z-direction

(2a)

(2b)

(2C)

and

jucE~ = ~ 1- YHV, (3a)

jwtEW = ~ -YHZ, (3b)

jueE, = ~ ~- (3C)

For most purposes an alternative set of equations is more convenient

RI use. Taking Eqs. (2u), (2b), (3a), and (3b) we find, on substitution.
:md rearrangement,
dH. dE,
K2E. = jv.I y J (4a)
ay az
dH, dEv
U2EU= jiw Y ) (4b)
ax ay
and
;Ez dHz
K2H. = jco y ~ (5a)
dy ax
dE. dHe
K2HU = (5b)
]e z y x

On substitution of these into Eqs. (2c) and (3c), u-e obtain

aEz
(6a)
~+~+K2EZ =0,

d2H.
+
&2
a% + K2H. = O, (6b)

with
KZ= &~c + # = kt + 72. (7)

The structure of this second set of equations shows that there are
two independent field components E., H. from which the others can be
SEC.7.2] PROPAGATION IN WA VEGUIDES 203

derived by Eqs. (4) and (5). We can consequently classify waves in

uniform guides into three fundamental types:
1. TEM-waves (transverse electromagnetic waves) with E. = H, = O.
2. ZE-waves (transverse electric waves) with E. = O, H. # O.
3. TM-waves (transverse magnetic waves) with H, = O, E= # O.
TELf-waves.-These are also known as the principal waves. The
electric and magnetic field vectors lie in a plane transverse to the direc-
tion of propagation as they do in a plane wave. From Eqs. (4) and (5)
it is evident that the principal wave will vanish identically unless ~z = O.
With this condition on K, we find from Eq. (7)

? = (a + m) = jk = j+f)~. (8)

If such a wave can exist, it will propagate without attenuation and

with a phase constant 13= u(w) ~ = 2r/A which is the same as that
for a plane wave in an unbounded medium.
To obtain information about the field vectors we must return to
Eqs. (2) and (3). It is directly evident from Eqs. (2a) and (2b) that

H=
()
~~(i.xE). (9)

Thus E and H are related as in a plane wave; they are mutually perpen-
dicular and transverse to the direction of propagation. Equation (2c)
becomes
aE.
. dE. . o
ax ay

whereas substitution of Eqs. (2a) and (2b) into Eq. (3c) yields

$+$=O.

The first of these states that in the dependence on z and y the field is
derivable from a potential function U(z,y); that is, we can write
E = e- V71(z,y). (lo)

It then follows from the second of the above equations that U(x,g) must
be a solution of the twe-dimensional Laplace equation:
dzu
~+y=o. (11)

The electric vector of Eq. (10) is everywhere normal to the equipoten-

tial surfaces U = constant. Since the conductivity of the guide walls
is infinite, E must be normal to the wails by the boundary conditions
formulated in Sec. 3.3; consequently, the walls of the guide must cor-
respond to equipotential lines of U. This, however, raises an important
204 MICROWAVE TRANSMISSION LIATES [SEC. 72

distinction between single-conductor and two-conductor lines. In the

case of a single-conductor line we are concerned with a solution of
Laplaces equation in a simply connected regiona solution that assumes
a constant value over the boundary. The only such solution is that
for which U is a constant over the entire region; hence the gradient and
the field vectors are zero. There is no TI%iki-wave possible in a hollow
waveguide. In the two-conductor line we seek solutions of Laplaces
equation in a multiply connected region. The required solution assumes
one constant value over one boundary and another constant value over
the second boundary, as in the electrostatic problem of two conductors
at different potentials. Such solutions exist; consequently a TEiW-wave
is possable in any twe-conductor line. Furthermore since y is a pure
imaginary for all frequencies, the lossless two-c ondudor line supports jree
propqation of this wave type at all frequerwies.
TE-waves.These are known also as H-waves. The electric field is
wholly transverse to the direction of propagation, while the magnetic
field has a longitudinal component H, in the direction of propagation.
It is clear from Eqs. (4) and (5) that all the other components can be
derived from Hz. If we write

H, = ~(x,y)e~z, (12)
~ must satisfy Eq. (6b):
V+ + K+ = O. (7.6b)

We must find solutions of Eq. (6fI) that lead to field components satisfy-
ing appropriate boundary conditions at the guide walls. By Eqs. (3,24)
and (3.28), these conditions are

(n x E) = (n,EV - null,)i, = 0, (I)

(n -H) = n.Hl + nUHy = O, (II)

where n is a unit normal to the boundary, directed into the interior of the
guide, From Eqs. (5a) and (5b) it follows that (II) is equivalent to
requiring

(13)

over the boundary. On inserting the values of E., E. from Eqs. (4a)
and (4b) into (I), one finds that condition (I) likewise reduces to Eq. (13).
Thus, the boundary condition (13) is the only one that need be imposed
on the solution.
Solutions to Eq. (6b) which satisfy Eq. (13) are possible only for
definite values of K. These are known as the characteristic values; we
shall designate them by Kmn. To each characteristic value there corre-
sponds a set of wave types which are spoken of as modes of propaga l;(~n;
in most cases of interest there is only one mode for each value of K. Any
SEC. 7.2] PROPAGATION IN WA VEGUIDES 205

one mode is completely specified by giving the field configuration over a

cross section of the line. The propagation constant -y~. for a given mode
is
Tmn= (K;n k)). (14)

It is immediately evident that if .:. < ~, then ~~n is a pure imaginary

and the wave is propagated without attenuation. Conversely, if
K~n > ~2, then -ynmis real and the wave is attenuated. A wave of fre-
quency v = 2rr/w will, therefore, be freely propagated only in those
modes for which O(W)~ = 27r/k > Kmn. The phase constant for a given
mode in which free propagation takes place is

B*7I = ~2:
9..
=(k K:.))i. (15)

If we define the cutoff wavelength X$i by the equation

(16)

then the wavelength in the guide is

(17)

m
=[-(2)21
When the wavelength in unbounded dielectric exceeds the cutoff wave-
length, the wave cannot propagate in that particular mode. A hollo\v
waveguide thus behaves like a high-pass filter, for there is a definite
upper limit to the cutoff wavelength, corresponding to the smallest
characteristic value Kmm. In terms of the free-space wavelength h, and
the specific inductive capacity k, = c/tO, the guide wavelength is given by

=[1-KJT;
(17a)

the permeability P of the medium is assumed to be negligibly different
from that of free space, KO.
The wave type, or mode, corresponding to a characteristic value
K~nis designated as TE_ It follovw from Eqs. (5) and (12) that the
transverse magnetic field is given by

The complete magnetic field is, therefore,

 MICRO WAVE TRANSMISSION LINES [SEX. 72

H=ew ~ ~z V* + *iz . (18)

( )

The electric field is given by E = (1/.ju~) V x H or

E,=~(Hxiz). (19)

!fM-waues.-The magnetic field is wholly transverse to the direction

of propagation, whereas the electric field has a component E. in the
direction of propagation. These waves are known also as E-waves. If
we write
E. = @(z,y)e-~z, (20)

@(w) must satisfy Eq. (6a):

Vzl$ + K2+ = o, (7.6a)

which is the same equation as that for + (z, y) in the case of Tllwaves.
The essential difference between the problems arises from the boundary
conditions. The boundary condition (I) is a statement that at the walls
the tangential electric field must be zero. We thus require

n.EV nvE. = O, (21a)

i. = o; i.e., @(xJy) = o, (21b)

over the walls. Substituting from Eqs. (4a) and (4b), we find that
condition (21a) is equivalent to

(n x VI$) = O. (21C)

If condition (21b) is satisfied, the boundary corresponds to a curve of

constant @; hence V@ is normal to the boundary, and condition (21c) is
automatically satisfied. Further, from Eqs. (5a) and (5b) itfollows that
for ZM-waves Eq. (21a) is equivalent to the boundary condition (II)
on the magnetic field, stated previously. Again, therefore, we have a
single boundary condition, namely, Eq. (21 b), to impose on the solutions
of Eq. (6a).
As in the case ,of ZE-waves, it is found that solutions +(z,Y) of Eq. (6a)
which satisfy the boundary condition exist only for certain character-
istic values Km.; these are, of course, different from the TE-values. To
each characteristic value there corresponds at least one wave type or
TT1-mode. The general remarks concerning the propagation constant
?~. and the conditions for free propagation are equally applicable to the
TM-mode; the guide wavelength is given again by Eqs. (17) and (17a).
It follows from Eqs. (4a) and (Q) that the complete electric field, for a
single mode, is given by
SEC.7.3] ORTHOGONALITY RELATIONS AND POWER FLOW 207

(22)

the magnetic field obtained therefrom by H = (1/.jo.w)(V x E) is

H, = ~ (E x i,), (23)

7.3. Orthogonality Relations and Power Flow. Examination of Eqs.

(18), (19), (22), and (23) shows that in a freely propagated mode, for
which ~ = jp is a pure imaginary, the transverse electric and magnetic
fields are in phase with each other and are in time quadrature with the
longitudinal field component associated with the given mode. The func-
tions ~ and @ are arbitrary to within a multiplicative constant; by a
simple readjust ment of constants which does not affect the relative
magnitudes and phases of the field components, the latter can be put in
the following form:

TE-waves:
fIz = jHa,e-iP.~; Ha, = :A ~a, (24a)
Et = E=~ei&.; Ea, = V~a x i,, (24b)
H, = Ha,e-~&~; Ha, = ~~ V@a. (24c)

TM-waves:

E, = jE=.e@; E., = ~ ~a, (25a)

.
Et = EateN..; E., = VI#Ja, (25b)

H, = Ha~e~~~Z; H., = ~ i, x V.#I=. (25.)

where the functions Haz, Ea., Es,, and Ha, are all real. The subscript a
represents the pair of mode indices m, n. Equations (24) and (25) are,
of course, still to be multiplied by arbitrary constants determining the
amplitudes of the waves. From these expressions it is seen that the
Poynting vector S = ~ Re (E x H*) arises entirely from the transverse
field components; the power flow is, therefore, entirely along the axis
of the waveguide, no power flowing into the walls of the guide.
The same expressions [Eqs. (24) and (25)] with B. replaced by j~~, -r.
being real, serve also for the modes that are beyond cutoff for the given
operating wavelength. It is seen that in these modes the transverse
electric and magnetic fields are in time quadrature; consequently, there
is no energy flow along the axis of the guide. In fact, the Poynting
vector S = ~ Re (E x H*) vanishes completely; the energy associated
with these modes is stored in the waveguide in the neighborhood of the
point of their excitation.
208 MICRO WAVE TRANSMISSION LIAES [SEC. 73

The modes possess important orthogonality properties. 1 The total

power transported through any cross section of a guide that is supporting
free propagation of several modes is the sum of the powers transported
by the separate modes; there is no energy coupling between modes.
For example, let us consider the power transport Pab of the mixed Poyn-
ting vector % = ~(E~~ x Hbo of a pair of modes TEa and TEr,; we have

par, = ~ (E., x Hb,) . i, d~

\
.,0,8
.ection

= & [(V*. x i,) x v&] . i, dS = ~P V~a . v#b dS. (26)

/ /
The last integral transforms as follows:

VIJ. . V+b dS = V . (~. Vir,) dS $aV$b dfi. (27)

[ \ /
By means of Greens theorem, the first integral on the right-hand side
transforms into a line integral over the boundary:

(28)

the positive normal to the boundary being taken as shown in Fig. 7].
Since the function +, satisfies the boundary condition [Eq. (13)] for the
TE-modes, the integral (28) is equal to zero. Making use of 13q. (6b]
we have then

V+a . V+b dS = K; ti.tib ds. (29)

/ /
Interchanging the role of ~a and lb in Eq. (28), we arrive in a similar
manner to

V+. . V+b dS = K; $.4, dS. (29a)

/ /
It is evident that if a # b, Eqs. (29) and (29a) can both be satisfied
only if

We have thus found that

Pa, = o, a#b

.& IV4.12 dS, a=b. (31)

/
I H. A. Bethe, Formal Theory of Waveguidesof Arbitrary Cross Section, RL
Report No. 43-26, March 1943.
SEC.7.4] 17RANLSMlSSION-Ll iVE CONSIDERATIONS 209

It is readily seen that the proof applies without change to the case
where one or both of the modes are beyond cutoff. Similar techniques
I
lead to the result that there is no energy coupling between pairs of
TM-modes or between a TE- and TM-mode. The power relation is
only one of a number of orthogonality properties. The others are given
without proof: if a # b,

7.4. Transmission-line Considerations in Waveguides. We have

concerned ourselves in the foregoing with a wave propagated in the
positive z-direction; this is the physical situation which would exist in a
waveguide extending to z = + m with a generator at some remote point
along the negative z-axis. It was found for every wave type that in a
single mode there is a simple linear relationship between the transverse
components of the electric and magnetic fields:

>4
ZEM-mode: H = ~~ (i= x E), z(o) = f ,
(33a)
()e

TE-mode: E = Zfi~ (H x i,), z:; = tip, (33b)

Ymn

(33C)

These are analogous to the current-voltage relationships in a single wave

on an infinite two-wire transmission line. The quantity Z~~ is known
as the transverse wave impedance.
The general field for a single mode in a waveguide that does not
extend to infinity consists of two waves, one propagating in the positive
z-direction, the other in the negative z-direction. The field expressions
for the latter are fundamentally the same as those given by the sets of
Eqs. (24) and (25), but with e@azreplacing e~~,and with the magnetic
field components reversed in sign to give the proper direction to the
Poynting vector of the wave. Considerj for example, the TE.-mode.
Let A. and Ba be the amplitudes of the electric field in the waves propagat-
ing in the positive and negative z-directions, respectively; from Eqs.
(24b) and (24c) we have then that the transverse fields are

(34b)
210 .![IcRow4 VE TI{,I.VST11*S,51
O! LIVES [SEC. 74

on considering the scalar factors that express the dependence of the

fields on position along the \va~-eguide axis, it is seen that the mode
can be set into equivalence with a two-wire transmission line of character-
istic impedance Z:); the electric and magnetic fields are the analogues
of the voltage and current, respectively. It must be noted that the
characteristic impedance of the equivalent line differs from mode to
mode; and consequently, a waveguide supporting free propagation
of a number of modes cannot be set into correspondence \vith any one
two-wire line.
The definition of the equivalent tl~o-wire line for a. given mode is
arbitrary to a considerable extrnt. Given a function *.(z,y), ~ve may
define a pair of vector functions

g.(.rjv) = c1 V+a x i,, (35a)

ha(x, y) = C, vta, (35b)

where the constants c1 and cz arc required to be such that

The constants c, and CZare arbitrary. In terms of the nc\v vector func-
tions, Eqs. (34a) and (34b) can then be written

The quantities l. and 1. will be named the ~-oltage and current param-
eters of the mode, respectively. The voltage parameter is the sum
of tl\-o voltage waves tra~-cling in opposite directions, of amplitudes
v:+] and 1:-), respectively.
Equations (36a) and (361~) serve to emphasize the arbitrary feature
of the two-wire line equivalent of a l~aveguide mode. The ratio c1 CJ
can be chosen at will; given any ratio, the characteristic impedance of
the equivalent line is

Zo = Z$ :; (37)

the voltage and cimrent parameters represent directly the voltage and
current on the equivalent line. The voltage and clwrent parameters
possess one property that is unique, indrpcndent of the arbitrary choice
of the constants c1 and c2, prorided Elq. (35c) is satisfied. The net po~,ver
passing through the cross section of the guide in t!~e positive z-direction
is

(38)
SEC. 74] TRAASIJISSIO>J-l,l .VE CO.V.$IDERA TION,7 211

Thus, any choice of definition of the volta~e and current parameters

leads to a two-wire line representation in which the polver flow computed
on the basis of the equivalent voltage and c~lrrent is equal to the power
transport along the ~vaveguide of the given mode.
One possible choice of the definition of the equi~-alcnt transmission
line is to take c1 = c! = 1. The fllnct,ion ~(.r,y) is itself arbitrary to
within a multiplicative constant; it can, therefore, be chosen so that it
satisfies the normalization condition

The characteristic impedance of the line in this case is equal to the trans-
verse \vave impedance. This definition has one shortcoming: It is
possible to change the dimensions of the waveguide, other than by a
scale factor, ~vithout changing the characteristic impedance of the
equivalent line. Consider, for example, a pair of two-conductor lines,
having different cross-sectional dimensions and configurations, joined
together to form an infinite line. The ~va~-eimpedance of the !IElf-mode
is independent of the cross-sectional dimensions, and on that basis alone
the hybrid line is equivalent to an infinite homogeneous two-wire line.
The treatment of the junction effect can be simplified considerably by a
different choice of the definition of the characteristic impedance of the
line, obtained by multiplying the \vave impedance by a factor c2/c, that
is a function of the cross-section geometry. In Sec. 7.6 it will be shotrn
that there is a natural physical definition of the voltage and current
parameters for a TE-U-mode which leads to a characteristic impedance
having t!le desired properties. Similar considerations apply to the other
modes; it is possible to choose the ratio cZ/cl in 13q. (37) to be a function
of the cross-sectional dimensions of the \vavc~uide in such a \vay as to
simplify the analysis of problems involring jlmctions bet\veen wave-
guides of different cross section.
The transmission-line analogy develops more fully if \veconsider the
\vaveguide to undergo sudden changes in structure. Such changes may
be produced by obstacles inserted at some point in the guide, a sharp
transition in the properties of the dielectric medium, or a sudden transi-
tion TO a waveguide of different cross section-to mention but a few.
It-e shall consider in detail the simplest of these casesa sharp transition
in the dielectric in a guide of uniform cross section. For convenience
the boundary between the two media will be taken to be in the plane
z = (), as shown in Fig. 7,2. Let the constants of the medium to the
left of z = O be c,, p, and those of the medium to the right of z = O be

I For further clet,ails sw J. C. Slater, .Uirrowave Transmission, McC, r:tw-Hill,

New York, 1942, ~hap. 4.
212 MICROWAVE TRANSMISSION LINES [SEC, 7.4

e2, P2. As a typical case consider a TE-wave of a single mode to be

incident on the boundary from the left. To the right of z = O we will
have a transmitted wave and to the
left a reflected wave, in addition
to the incident wave. No other
waves will occur, since the con-
m tours of the cross section are uni-
Z=o
form and there is no necessary
FIG.7.2.Discontinuity in waveguide struc-
ture. distortion of the field configuration
at the boundary; these three waves
will suffice to satisfy the boundary conditions on the fields at the discon-
tinuityy.
The field vectors of each of the three waves are derived from a scalar
function #(.z,y) according to Eqs. (24a) to (24c). Furthermore, the
scalar functions for the three waves all satisfy the same differential
equation [Eq. (6b)], and the same boundary conditions at the walls of the
waveguide; hence, all three fields derive from the same scalar function.
The ratio c2/cl in Eq. (37) is of no immediate consequence in this case,
because the cross section is uniform and may be chosen equal to unity;
the function ~(z,y) may also be required to satisfy the normalization
condition [Eq. (39)]. The only significant differences between the waves
are the amplitudes and the transverse wave impedance. The field in
region 1 is thenl
(40a)

(40b)

and in region 2,
E,, = [V~+)e-@,]g(x,y), (41a)

H,, = ~Oj [V~+)e-@,Z]h(x,y). (41b)

2

According to the boundary conditions (Sec. 33) the transverse

electric and magnetic fields must be continuous across the plane z = O;
we have then
~~+) + v;-) = ~$+), (42a)

*
1
[w+) - TT] = &
2
Vj+. (42b)

As in the case of a two-wire line, these equations express the continuity

of voltage and current at the junction of tlvo lines of different character-
istic impedance. We can also define an electric-jield reelection coqflcient
r(z),
I The mode subscripta will be droppedto simplifythe notation
SEC.7.4] TRANSMISSION-LINE CONSIDERATIONS 213

(43)

which corresponds to the voltage reflection coefficient of Eq. (2.27).

From Eqs. (42a) and (42b) the value of the reflection coefficient r(0) at
z = O is found to be
z~o) _ Z\O).

r@) = Zjo) + z~o] (44)

It is evident that this is equivalent to the reflection coefficient of a line

of characteristic impedance ZjO1terminated in an impedance Z$O. With
respect to the terminal impedance, it will be noted that the line to the
right, extending to infinity, is equivalent to a line terminated in its own
characteristic impedance and hence presents an input impedance Z~O)at
the plane z = O.
The reflection coefficient r(z) is to be regarded as the fundamental
transmission-line quantity for a waveguide. Evidently it is free from
the arbitrary factors entering into the definition of the voltage and current
parameters and the characteristic impedance of the equivalent trans-
mission line. It is apparent from Eq, (43) that it transforms along the
line just like a voltage reflection coefficient. Also, on computing the
Poynting vectors of the incident and reflected waves, it will be seen that
the electric-field reflection coefficient bears the same relation to the inci-
dent and reflected power as the voltage reflection coefficient (Sec. 27).
At any point along the line we can regard the section to the right as
presenting an input impedance, normalized to the characteristic imped-
ance of the mode,
I + m).
((z) = (45)
1 r(z)

The normalized impedance is also independent of the choice of the

definition of the equivalent transmission line. Making use of the
transformation property of I(z) expressed by Eq. (43) it is found that
the normalized impedance transforms along the waveguide according to

(46)

just as it does on a two-wire line. The normalized admittance can also

be defined in the same manner as was done in Sec. 26,

and it is evident that it also transforms along the waveguide according

to Eq. (46). Thus, the entire discussion in Chap. 2 on impedance mis-
214 MfCROWA VE lRA.\rS.llISSIO.~ 1>1.VES [SEC.7.5

match, standing-wave ratios, and line transformations can be carried

over to the fields on any one mode in a waveguide.
7.5. Network Equivalents of Junctions and Obstacles.The develop-
ment of the problem considered above proceeds in a similar manner for
ZE.lf- and ZM-modes and leads to equivalent two-wire line analogies
for any single mode. The discontinuity that we have considered in that
problem is equivalent to a junction between a pair of two-wire lines of
different characteristic impedances, such that the capacitative and
inductive effects due to the junction are negligible. At such a junction
both the current and voltage are continuous, corresponding to the
continuity in the transverse magnetic and electric fields, respectively, in
the waveguide problem. As a second step in developing the transmis-
sion-line analysis we shall consider junction effects and the problem of
obstacles inserted into a waveguide. The general theory of these prob-
lems is treated extensively in other volumes of this series. 1 We shall
restrict ourselves here to several qualitative remarks.
As a specific problem let us consider a junction between two wave-
guides of the same cross-sectional shape but different dimensions, joined
in the plane z = O (see Fig. 7.3).
The dimensions of both guides are
assumed to be such that they can
support free propagation at the given
Z=o frequency in one mode only; we shall
Fro. 7,3.-Junction effects in wave- refer to the latter as the dominant-
guides. mode wave. We shall assume the
dominant-mode wave, set up by a
generator at a remote point on the negative z-axis, to be inc!ident on the
junction. Since there is a change in cross section at the junction, ~ve
should certainly expect to find a reflected ~vave of the dominant mode on
the left and a transmitted wave of that mode in the waveguide on the
right. The fields must join in tne plane z = O so as to satisfy tne appro-
priate boundary conditions. Over the opening in the junction the trans-
verse fields must be continuous; over the metal surface of the junction
the transverse electric field and the normal component of the magnetic
field must vanish. The latter conditions cannot be satisfied by the three
dominant-mode waves alone; higher modes must be excited in both
waveguides at the junction.
The generation of the higher modes arises from the necessary distor-
tion of the electric and magnetic fields due to the edge of the junction
and its metal surface. The electric-field lines must be normal to the
lattera condition that cannot be met by the dominant mode alone in
the waveguide to the right. However, according to our assumptions as

1Principles of ill icrouaue Circuits, Vol. 8, and The Haueguide lIandbook, VO1,10.
SEC. 7.5] .YETWORI< EQCI i.4LE.YTS OF J1 .YCTIO.YS 215

tothewaveguided irnensions, tbehigherm odesc annot propagate; except

within a short distance of the junction (of the order of a wavelength)
the fields consist essentially of dominant-mode waves. The higher modes
represent electric and magnetic energy stored at the junction. It is
possible to represent these energies as energies stored in a reactive network
equivalent for the junction. 1 In the general case the network takes the
form of a T- or II-section (see Sec. 2.2). The effect of the junction on
the dominant-mode wave thus arises from two factors: (1) a discontinuity
in characteristic impedance and (2) a reactive four-terminal network
inserted between the lines. The precise lalues of the elements of the
latter network again depend on the definition of the characteristic imped-
ance of the equivalent line for the dominant mode. A number of junc-
tion networks are given in the Waveguide Handbock, Yol. 10 of this
series; in each case, the definition of the characteristic impedance (or its
reciprocal, the characteristic admittance) is given. The elements of the
network can, of course, be expressed as normalized values with respect
to the characteristic impedance of either guide.
In the waveguide to the left, at a short distance from the junction,
we have cmly the incident and reflected dominant-mode waves. Here
we can apply transmission-line concepts to the dominant mode and
define the corresponding electric-field reflection coefficient. This reflec-
tion coefficient can be related to an effective impedance terminating the
line at the junction. This impedance, in turn, may be expressed as due
to a junction network across the output terminals of which there has been
connected the characteristic impedance of the guide to the right. These
procedures lead to consistent definitions of the junction impedance.
The junction network necessary to represent the stored energies, when
inserted between the transmission-line representations of the two wave-
guides, gives rise to a reflection coefficient in the region on the left cor-
responding to the electric-field reflection coefficient obtained on the basis
of field-theory analysis.
The theory of obstacles develops along similar lines. It is found, as
in the case of junctions, that an obstacle has the same effects on impedance
and energy stored as a four-terminal net\vork inserted between a pair of
transmission lines \vhose characteristic impedances are the wave imped-
ances of the dominant mode. It must be emphasized that each mode
\vhich can propagate has its onm transmission-line analogue and that
simple transmission-line theory applies to a \vaveguide only when it can
support hut one mode. Transmission theory alone can give no informa-
tion as to the net \vork equivalents of junctions and obstacles; these
must l)c obt ainml I)y field-theory analysis. The equivalent network also
depends t,t] tl~e particular dominant mode being considered. once the
equil-alent netl~ork }ms been estal)lishwl, it can be expressed as a T-sec-
1Se(,~~,;~~c;,,le,y
of l~,,mm[w Circ(/i/,s,VO1. 8 of this series.
216 MICROWAVE TRANSMISSION LINES [SEC, 76

tion, and the impedance transformation properties of such networks can

be used in the conventional manner.
7.6. TEM-mode Transmission Lmes.We have pointed out earlier
that in general it is not possible to set up unique definitions of voltage
and current in waveguides, and we have therefore set up transmission-
Iine analogues in terms of wave impedances and field-reflection coefficients.
In the case of !!EM-modes, however, it is possible to set up transmission-
line quantities that are directly related to the two-wire line quantities
discussed in Chap. 2.
It was found in Sec. 7.2 that the electric field over any cross section
is derivable from a potential. Hencej over a cross section, the line integral
of the electric field from the inner conductor Cl to the outer conductor
C, is independent of the path and indeed is equal to the difference between
the values of the potential over the conductors. This defines the voltage:
c,
ve~z = c dr =
/ c,
E . e~z
/ c1
VU. dr = (Ui ?ll)e-~. (47)

There exists also a relation between surface integrals over any closed
region in a cross section:

+
(Vx H). i.dS=ju~
!$E.i. dS=O. (47(z)

Jt follows that the line integral of H over a closed curve surrounding C,

is independent of the choice of the curve. In particular, let us take a
path along the boundary of Cl. His tangential to Cl and by the bound-
ary condition (Sec. 3.3) is equal in magnitude to the surface current
density K. Hence the line integral of H gives the total current carried
by C,:
~, H , ds = le-~z. (48)
$
The line integral of H along the boundary of C2 gives the total current
carried by the latter; by virtue of the equality of the line integrals the
two currents are equal. On carrying through the details of the vector
calculation, it will be found that the current on C2 is opposite in direction
to that on Cl. There is thus a direct two-wire line analogue with voltage
V and current 1. Corresponding to these we define a characteristic
impedance,

Zo =;=
()
:
6 U2

#c,
ul

IVUI d.
(49)

This is, of course, different from the wave impedance for the mode.
The relationships between the ZOdefined in Eq. (49) and the t\vo-wire-
line impedance become more evident on calculating the equivalent series
SEC. 7.7] COAXIAL LINES: TEM-MODE 217

inductance and shunt capacitance per unit length of the two-conductor

system. The magnetic energy for unit volume is ~~ IHI, and therefore
the magnetic energy per unit length of line is

w.=; lVL1dS. (50)

// cresssection
If L is the equivalent inductance per unit length, then

Wm = ;L12 ;
hence
L = f~~ lv~l~~
12
(50a)

Similarly the electric energy per unit volume is ~elll 2, and the electric
energy per unit length is

we=;
H cromsection Ivulds (51)

The equivalent capacity C per unit length is then

w, = +Cv
or
~ = 6// Ivclzds
~, (51a)

According to Eq. (220) the characteristic impedance of a lossless two-

wire line is
E
z, = ~.
d

Combining this with Eqs. (50a) and (51a), we obtain the quantity defined
in Eq. (49),
For most practical purposes a two-conductor guide supporting the
7EM-mode as its dominant \vave can be treated from the voltage-current
point of vie\v. Applications of this fact will be made in Sees. 7.9 and
7 10 in discussing impedance transformations and matching devices for
coaxial lines.
7.7. Coaxial Lines: TEM-mode.The only type of two-conductor
guide of major importance is the coaxial line formed by a pair of concentric
circular cylinders. Let a be the radius of the inner conductor, b the radius
of the outer conductor. Cylindrical coordinates r, O,z are suited fGr the
discussion of this system, T and 0 being polar coordinates in a cross section
of the line. We shall first consider the ZEM-mode. The solution to the
potential problem is well kno\vn from electrostatics:
218 MICROWAVE TRANSMISSION LINES [SEC. 7.7

l
U(x,y) = ln~, (52)
in ~ ()
() a
where V is the voltage across the line. The electric-field intensity is,

Line of zero field intensity

(a) (b)
FIG. 7.4.Coaxial-linemodes: (a) TEJ-f-mode (no cutoff wavelength); (b) TEII mode
[x,,~c)= (5 + b)7r], electricfield; magneticfield.

therefore,

(53)

and the magnetic-field intensity

(53a)

where i, and ia are unit vectors in the directions of increasing r and 0.

Here -y has been replaced by j~, and the double sign indicates a wave
traveling in either the positive or negative z-direction. The field con-
figuration and the current distributions on the conductors are completely
symmetrical about the z-axis; the former is shown in Fig. 74.
The current is given simply by 27rrH(r):
e~~v.
I=2T-
()
P
In b
() a
It follows then directly that the characteristic impedance, in the sense
of the previous section, is

O=ww)
For most dielectrics of interest p differs negligibly from the free-space
(54)

value ,uO. On introduction of the specific inductive capacity k, = ~/co,

the characteristic impedance becomes
SE{,. 78] (70.4 XI.4L T.I.VES: T.If - .4 Vl) TE-MODES 219

0=%1:
L () (54a)

The series inductance and shunt capacitance per unit length of the line,
computed from Eqs. (50a) and (51 a), are fo~md to be

L=~ln
()
~, (55a)

c=~. (55b)
in ~
() a
7.8. Coaxial Lines: TM- and TE-modes.In the study of the lE-
and ZM-modes we are concerned with the solutions of equations of the
form
JZF
~2+$+K2~ =0, (76a)

where F will stand for either of the functions 4(x,Y) or I$(x,Y) of ~qs. (1z)
and (20) respectively. On introduction of the polar coordinates r-, 0,
the differential equation becomes

(56)

The equation is separable in the variables r and 0; in particular we shall

writ e

F = R(r) ;~~;;
{
then R(r) satisfies the equation

(57)

This is the differential equation for the cylinder functions of order m, in

the variable Kr. The pair of linearly independent solutions suited to
the finite region with which \veare concerned here consists of the Bessel
function ~fi(~r) and the Xeumann function J~~(K~). The latter ~ec~mes
infinite at r = O; however, since the origin is excluded by the inner con-
ductor, the Xeumann function is admissible as a solution. The general
solutions of Ilq. (56) are therefore

(zy)
@(xjy) 1
= [.1,1.,(.,) + ~.~~(.r)](c cos mfi + D sin w@.

The field must be single-valuml in 0; as a consequence m can have only

Integral values. l~(;r any given value of m it is possible to eliminate one
220 MICROWAVE TRANSMISSION LINES [SEC.78

of the trigonometric functions by proper orientation of the z ,y-axes.

Without loss of generality we can set D = O, taking as the solutions

(58)

a. TM-modes.We must consider the TM- and TE-modes sepa-

rately. In the case of the TM-modes we are concerned with the function
t$(z,y) and the boundary condition of Eq. (21b); we require @ = 0 for

aHvalues of Oat r = a and r = b. This gives two homogeneous equations,

~~_(KU) + ~~~(KU) = 0,
(59)
fi~~(.b) + BIV*(Kb) = O, }

for the determination of the ratio B/A. Solutions other than A = B = O

exist only if the determinant of the coefficients vanishes:

~m(.a) ~m(KU)
= ~m(KU)~m(Kb) ~m(Kb)~na(KU) = 0. (60)
~~(.b) ~~(.b)

This in turn is satisfied only for a discrete set of values of K; the latter are
the characteristic values which, arranged in order of increasing magnitude,
we shall designate by Kmn. If we write u = Ka, a = b/a, the equation
appears in the standard form
Jm(u)Nm(au) Nm(u)Jm(au) = o. (60a)

Roots of this equation are given in Jahnke and Erode. For a given value
of a the smallest value of W- occurs for m = O; this gives the longest
cutoff wavelength for these modes. Examination of the roots shows that
forl~a~7,
3a
~a > UOI = (LKol>
ba
Therefore the cutoff wavelength X$] for the mode is given approximately
by
~;~ =2( b a). (61)

We recall that propagation in a given mode can take place only if the
wavelength in unbounded dielectric is shorter than X(c). In all practical
cases the spacing between the conductors is much smaller than the
wavelength, and there is no need to be concerned about the simultaneous
excitation of TiW- and !!EM-modes.
b. TE-waoes.Here we are concerned with the function +(z,v) and
the boundary condition of Eq. (13); for the case at hand the latte~
becomes dl/~r = O for r = a and r = b. This leads to the conditions
1E. Jahnke and F. F;mde, Tables of Functions, Fig. 204, Dover Pul>licati<;rs
Reprint, New York, 1943.
SEC.7.9] CASCADE TRANSFORMERS: TEM-MODE 221

~~~(KU) + ~~~(KLI) = 0,
(62)
~~:(Kb) + ~~:(Kb) = O, }

on the constants A and B. Aont rivial solutions for the latter exist
again only for the characteristic values .~n that satisfy

&(KU)~~(.b) ~~(.b)~~(.~) = (). (63)

For m = o we have the relation J:(z) = .Jt(z), and similarly for the
Neumann function; the characteristic values of the !iEO.-modes are there-
fore given by the roots of

J,(u) .vl(au) J,(aU)N,(u) = o, (63a)

where u, a have the same meanings as previously.

From what has been said about the roots of Eq. (60a), it is evident
that the cutoff wavelength of the ZEO-modes is shorter than that of
the ZMOl-mode and that the former are of no consequence as propa-
gating modes in a practical case. The roots of Eq. (63) for m >0 have
been discussed by Truell. For our immediate purposes we need con-
cern ourselves only wit h the lowest mode of the series, the TE1 l-mode.
The field configuration for this mode is illustrated in Fig. 7.4. For this
case lt is found that the characteristic value is given very closely by

2
(64)
l=a+b;

thus the cutoff wavelength is

h~~ = 7r(a + b). (64a)

This is the mean circumference of the inner and outer conductors. TO

p~event propagation of the TEl,-mode the mean circumference must be smaller
than the ope~ating wavelength. This imposes limitations on the dimen-
sions of the line and in particular on the spacing between the conductors;
the latter in turn limits the power-carrying capacity of the line.
7.9. Cascade Transformers: TEM-mode.The termination of the
line in a radiating system in general gives rise to a reflected TEM-wave
and to excitation of TAT- and TE-modes. We shall assume that the
line dimensions are such that the latter modes cannot propagate and
confine our attention to the region of the line where only the incident
and reflected TE.W-waves exist. The reflected ~vave represents an
impedance mismatch, and it is necessary to consider a correction for it.
Perhaps the most useful device is a cascade transformer, a section of
coaxial line of characteristic impedance different from that of the main
line. Two such transformers are illustrated in Fig. 7.5: (a) the sleeve
I R. Truell, .Jour..4pplied Phys,, 14, 350 (1943).
222 MICROWAVE TRA.VSMISSIOA7 LIMES [SEC. 7.9

type with characteristic impedance smaller than the line impedance and
(b) the undercut type with characteristic impedance larger than that of
the line. As has been pointed out before, the junctions give rise to other
modes; however, if the change in radius is small, the junction effect is
small. Data on the latter will be given below.
A B
.1-

%GeratOr Lo=
(a)

A B
r 1-

(6)
FIG. 7.5.Cascade impedance transformers: (a) sleeve section; (b) undercut section,

The dimensions desired in a transformer can be determined as fol-

lows: Except for junction effects, the voltage and current and the input
impedance looking toward the right ha~-e the same values at adjacent.
points on either side of the junction. Let Z: be the characteristic imped-
ance of the transformed, 20 that of the line, and Zl the input impedance
at 1?. Then, from Eq. (2.32), Sec. 2.6, the input impedance at A is

(65a)

Impedance matching requires Z(A) = Z,; that is,

(65b)

Separation of real and imaginary parts gives two equations from which,
for a given value of A, one can obtain Z: and the length of the transformer
that matches Z into ZO; the dimensions of the transformers are obtained
from Z~ by means of Eq. (54a).
There are points along the line at which Z is real. These points are
h/4 apart, and the impedance is alternately rZ, and ZO/r, ~vhere r is the
voltage standing-wave ratio. If either of tlwse is taken as the junction
point 1?, it is found from Eq. (6.5b) that 1 = A/4. The characteristic
impedance of the quarter-wave section is found to be related to ZO as
follows :
SEC. 7.10] PARALLEL STUBS A.VD SERIES REACTANCE 223

Z: >20 if z = r-z,.
The first of these corresponds to a sleeve section; the second to an under-
cut section. In so far as matching is concerned, either can be used.
The sleeve section has the advantage of simplicity of insertion, since
it is necessary only to slip a piece of tubing over the inner conductor
and to solder the seam to ensure good contact; it also has the advantage
of strengthening the line mechanically. An undercut section requires
machining and weakens the line. On the other hand, the sleeve section
reduces the clearance between the conductors and consequently the power
capacity. In both cases the edges of the junction increase the break-
down tendency; this difficulty can be minimized by rounding the edges
of the junction without impairing the matching relations.
It must be emphasized that a single transformer matches properly
at only one wavelength. In general the load impedance is a function of
frequency. Matching over a frequency band, such that the standing-
wave ratio remains less than a prescribed value, can often be achieved
by a series of transformer sections of different lengths and characteristic
impedances. It is difficult to carry the analysis through analytically for
an arbitrary load Z(k). A method of rather limited applicability employ-
ing a tandem of quarter-~vave sections has been developed by Fubini,
Sutro, and Lewis.
While the matching condition of Eq. (65b) always leads to a solution
of the mathematical problem, it is not necessarily true that the trans-
former will be satisfactory. If a large change in radius is required at the
junction, the junction effect becomes significant, and we must add to the
equivalent transmission-line reactive net\vorks at A and B corresponding
to the junction effects. It is found that the network consists of a
capacity across the transmission line at the junction points. The junc-
tion effect can be studied experimentally by means of a half-wave section.
From Eq. (65a) it is seen from transmisson-line considerations alone
that if 1 = A/2, then Z(A) = Z regardless of the value of Z:; this means
that the standing-wave ratio should be the same on either side of the
transformer. Figure 7.6 shows experimental results obtained with a half-
wavelength sleeve section on a 50-ohm coaxial line with inner diameter
0.375 in. It is seen that the de~-iation from simple transmission-line
behavior increases rapidly with increasing diameter of the sleeve section.
7.10. Parallel Stubs and Series Reactances.-Another useful device in
coaxial-line design is the parallel stub consisting of a section of coaxial
I IIFreq,lency characteristics of Wide-band Ilatching Sections, Radio Research
Laboratory (Harvard University) Report So. 23, April 1943.
2 Waueguide Handbook, tol. 10 of this series.
224 MICROWAVE TRANSMISSION LINES [SEC. 710 I

line at right angles to the main line. The arrangement is shown sche-
matically in Fig. 7.7a. The stub is terminated by a metal cap to prevent
radiation. Electrically the stub is a shorted section of transmission line.
If 2{ is the characteristic impedance of the stub and 1 its length, then
its input impedance, obtained from Eq. (65a) by setting Z = O, is

1.9 -

1.8 0

)
1.7

0
1.6

$ 1.5
/
g
~ 1.4 /
o

1.3
/
o
1,2
0
0
1.1 /
/
1.0 --
0 0.10 0.20 0.30
Transformer dia. D_O.375,,
FIG. 7.6.J um%ion effects with cascade transformers; mismatch of a A/2 transformer as a
function of diameter in a coaxial line of dimensions OD = 0.S11, ID = 0.375 in.

Ze = jZ~ tan fll. It is thus a reactive element. Consideration of the

current division at A shows that, neglecting junction networks, the stub
is to be regarded as a reactance shunted across the main line. If 1 = i/4,
then Z = cc, and the stub introduces no change in impedance at A;
such a quarter-wave stub is useful as a mechanical support for the inner
conductor. We shall not consider here the refinements required to
eliminate the frequency sensitivity.
The stub can also serve as a matching device. In this connection it
is more convenient to speak in terms of admittances. Let Y, = I/Z, be
the characteristic admittance of the main line, Y the admittance seen
to the right of A, and Y, = jY~ cot P1the admittance of the stub. It is
possible to locate the point A so that the admittance Y is Y = Y, + jB.
SEC. 7.10] PARALLEL STUBS AND SERIES REACTANCE 225

Insertion of the stub gives an admittance at the left of A equal to

Y+ Y,= Yo+j(13- Yjcot (51). For matching we require simply
that
Y~ cot @ = B. (66)

The structure illustrated in Fig. 7.7b is less widely used but is worth
consideration. The region bet ween A B and the outer conductor C2 acts

r 1 z;

zrJ
. ___ _VA

(a)
.Z5 Z.
Z. z; [
c,
A lB

(b)
FIG.7.7.(a) Parallelstubreactance;(b) seriesreactanceelement.
as a cascade transformer. If Z is the impedance at B, the impedance
just to the right of A in the transformer space is

where 2$ is the characteristic impedance of the transform space. The

region AB within the inner conductor acts as a shorted section of line
which presents an impedance at A equal to Zc,A = jZ~ tan ~1, where Z;
is the characteristic impedance of the inner region. At A we have the
voltage relation VCZC,= 1.,. + VAC,; the impedance just to the left of A
is given by
Z(A) = ZC,A + .ZAC,

(67)

The structure thus introduces a series impedance at A. It is of interest

to note that the length of the inner region can be made shorter than the
length of the outer region. If the latter is made equal to an integral
number of half wavelengths, the effect of the transformer region is
eliminated and at A we have simply an impedance ZC,~ in series with Z.
226 MICROWAVE 1RANSMISWION LINES [SEC.7.11

It is found in practice that stubs and series reactance transformers

with dimensions calculated on the basis of the transmission-line formulas
given above do not quite meet the simple theoretical expectations. This
is due to the junction effects neglected in the transmission-line arguments.
The errors, however, are generally small and can be eliminated by small
adjustments of the lengths of the structures. In the case of stubs, the
shorting cap can be replaced by a sliding
z
plunger in the experimental model to allow
easy adjustment of the length. The use of
series reactance transformers limits the
/ power capacity of the line; the standing
/
waves in the inner region produce intense
I,// fields at the open end and increase the
k__- tendency toward electrical breakdown and
)
/ sparking. An alternative form, which
// mounts the transformer on the outer con-
/ ductor where the electric field is weaker, is
y? / more satisfactory with respect to breakdown
/ characteristic but is less desirable from
~Lz assembly considerations.
I---- U----4 7s11. Rectangular Waveguides: TE- and
FIG. 7.8.Rectangukm wave-
guide. TM-modes.The hollow guide of rectangu-
lar cross section is the most widely used
line in microwave antennas. We shall take the z,,-v-axes to be oriented
as shown in Fig. 7.8; a is the broad dimension of the guide; b the narrow
dimension. The Helmholtz equation

(76a)

is in this case separable in the form

F = x(z) Y(y).

Substitution into Eq. (6a) leads to the two equations

(68)

with
K: + K; = K2. (68a)

The solutions have the same form for both members of Eq. (68); for
example,
x (x) ==A cos (K#) + B sin (K&).

The general solution of Eq. (6a) is

SEC. 7.11] RECTA XGC7LA R WA 1EGVIDES: TE- A.VD TM-MODES 227

$(z~)
l$(x,y) 1 = [A .o. (.=X) + B sin (K4)][C cos (by) + D sin (wY)]. (69)

a. ZE-waues.The solution ~(x,y) must satisfy the boundary con-

dition d~/&z = O over the walls. For the walls z = O and z = a,
al/an = d+/&r; we thus require that for all values of y

%1
a+

a+
.=O
= Kz~](?J) = O, (70a)

.=[.4 sin (K@) B cos (Ksa)]Y(y) = O. (70b)

ax ,=.
-1 =
This requires that ,?3 = O and that K. have the characteristic values

mr
~z = , m =0,1,2,.. (71)
a

Over the walls y = O and y = b, d~jdn = d~idy. This boundary con-

dition requires that D = O and that KY ha~-e the characteristic values

nr
Ku = , n = 0,1, 2,.... (72)
b

The characteristic values .~ti for the !fE,m-\raveare therefore

(92+(92
= (73)

By use of Eqs. (18) and (19), the complete set of field components for
the \\ravein the positive z-direction is found to be

H. = cos 7X cm ~ e~-z, E. = O,

Hz = &E. = ~~~
in(?;)cos(%e-z

u zEz= ?~coe)sint~)e-m
The significance of the integers m, n, is directly apparent: They represent
the number of sinusoids in the intensity of the field components E and
E=, respectil,ely, over the cross section of the gllide.
The cutoff wa\elengthj the gllide ~vavelength, and the trans~erse
\va\~e
impedance for a TE~fi-mode are respectively

(75)
228 MZCROWA VE TRANSfi ISSION LINES [SEC. 7.11

(76)

o
={1- [(4+(311

u-[(4+(2)1}
z:: = (77)

The ZEl~-mode (m = 1, n = O) has the longest cutoff wavelength. It

is by far the most important mode for antenna work. The electric field
has but one component, E., which is uniform in the y-direction and varies
sinusoidally along the x-direction with symmetry about the central sec-
tion of the guide. The field configurations for this and several other
TE-modes are shown in Fig. 7.9. It will be seen from Eq. (75) that to

(b)

m
k---a -

4 +-$--
[
*b

lt-t-l
i

(d) -
Fm. 7,9.Tll-modesin rectangularwaveguides: (a) ZE,o-mode [AIO) = 2a]; (b) TEII-
mode [XII() = 2ab/d-; (c) !ZEzo-mode [Az.a(=)
= a]; (d) TEo,-mode (kOI[) = 2b).
electric field; magnetic field.

ensure propagation of the TE10-mode alone the dimensions of the guide

must be such that
a<h <2a; 2b < A.
b. TM-waves.The solution @(zjy) must satisfy the boundary con-
dition @ = O over the walls. It is evident from Eq. (69) that we must
set A = ~ = O to satisfy the condition over the surface x = O and y = O.
Over the walls z = a, y = b the conditions are satisfied only for the
characteristic values

~z = m~
, m = 0,1, 2,.... (7.71)
a
~=l!, n=o, 1,2, . . . . (7.72)
SEC. 7,11] IMPEDANCE TRANSFORMERS 229

Thus the characteristic value .nn for the TM~m-mode line, like that of a
TE~m-wave, is given by

nH+(Y (773)

The cutoff and guide wavelengths are given by Eqs. (75) and (76); the
characteristic wave impedance, however, differs from that of the ZE-wave.
It is

(78)
(9+(9211

(a) (b)
FIQ.7.10.TM-modes in rectangular wave guides: (a) TM, ~-mode [AI,() = 2ab/<~;
(b) ZM,,-mode [AZ,(C) = 2ab/Vai + b!]. electric field; magnetic field.

The complete set of field components obtained by means of Eqs. (22)

and (23) is

zSin(%sin(we-
z 01
Ez=~Hv=m~
m. Cos(%)sint+)e-z, 7)

&=~Hz=_nYmn
msin(?)cos(?)e-mz I

There is no mode for which either m or n is zero; the lowest is the TMll-
mode.. It follows accordingly that a guide designed to cut off the TE-
modes other than the TEI o will likewise not support free propagation
of any of the TM-modes. The field configurations for several of the
latter are shown in Fig. 7.10.
7.12. Impedance Transformers for Rectangular Guides.Equivalent
networks have been established for a number of types of obstacles in
waveguides; these can serve to match out the reflected dominant mode
wave set up by the line termination. We shall present here the pertinent
data on elements designed for the TEIO-mode in rectangular guide and
shall indicate their applicability. The simplest, from the point of view
of the equivalent networks, are the mndows: metal diaphragms inserted
in the cross section of the guide. Typical forms are illustrated in Fig.
711. In the idealized case of infinite conductivity these elements behave
like capacities or inductances shunted across the two-wire t,ransmisaion-
230 MICROWAVE TRANSMISSION I.I.VES [SEC. 712

line representation of the TElo-mode. Accordingly, in using these ele-

ments it is convenient to treat the line in terms of admittance rather than
impedance. Let Y j~) = l/Zj~ be the characteristic wave admittance of
the TEIO-mode line. 1 l$-ith any arbitrary termination, there exist points
along the line, at quarter-wavelength intervals, at which the input admit-
tance looking toward the load is alternately (1) Y = Y, jB and (2)
Y = YO -t- jB. At the points 1 the load susceptance is inductive and a
parallel capacity is required for matching. Points (2), where the load
susceptance is ca,pacitative, require a parallel inductance. For the

(a) --16 i--

(b)

El I-6 +! 1--. --1

(c) (d)
F.6. 7.13.Windows for rectangular guides; (a) symmetrical capacitative; (b) s.vrnrnetrical
inductive; (c) asymmetrical inductive; (d) resonant.

former case the capacitative window (Fig. 7.1 la), is suited, while for
points 2 the inductive windows (Fig. 7.1 lb and c) are appropriate. For-
mulas and graphs for the susceptance of these and other windows,
referred to the characteristic wave admittance of the T~10-mode, are
available in the literature. z
In practice the inductive windows are to be preferred, because the
capacitative window-, in presenting an edge across the electric-field lines,
is more susceptible to electrical breakdown. Asymmetrical windows
have experimental and design advantages in that only one side of the
guide need be milled for an insertion. 3 This reduces the amount of
machining required in making test runs on impedance and eliminates the

1We shall drop the mode notation hereafterand write simply ZOand YOfor the
characteristicimpedanceand admittancerespectively,
2Microwave TramwnissionDesign Data, Sperry Gyroscope Company, 1944;
( Waveguide Handbook, RL Group Report No. 4?, Feb. 7, 1944; }Yaveguide
Handbook Supplement, RL Group Report SO. 41, ,Jan.23, 1!145;11aveguide Hand-
book, Vol. 10 of this series.
3 W. Sirhak, One-sided Inductive Irises and Quarter-wive (apacitative Trans-

formersin Waveguide, RI, Report No. 426, Nov. 17, 1943

SEC. 712] 1.lfPEDANC13 TRANSFORMERS 231

problem of alignment of two halves of a symmetrical window. On the

other hand, symmetrical windows lend themselves to use as pressuriza-
tion devices; the two metal borders can serve as supports for a thin
dielectric sheet. Such a sheet introduces an additional capacity in
parallel with the windo~vj the magnitude of which depends on the thick-
ness and dielectric constant. No systematic design information seems
to be available on this point at present, and the design of the pressurized
window must be developed experimentally.
The circuit equivalents of the windows immediately suggest the pos-
sibility of combining a capacitative and inductive window to make the
net susceptance zero, that is, to produce a resonant de~-ice that introduces
no reflection in the guide. Such a resonant window is illustrated in Fig.

1--1--1

Generator Load
(a)
(b)
FIG. 7.12.Step transformer in rectangular wa~wguide: (a) transverse cross section; (b)
longitudinal cross sect, on.

7.1 ld. To a first approximation, the dimensions cL, ~~ can be so choeen

that the capacitative and inductive susceptances are equal in magnitude.
The resonant window transmits all the incident power and, therefore,
cannot be used as a matching device. It is useful as a pressurizing ele-
ment to seal the waveguide; either the window frame serves as a support
for a thin dielectric sheet, or the open area of the window is filled with a
dielectric block. The dimensions of the window must be adjusted to
compensate for the dielectric; this again must be determined empirically.
It is obvious that true resonance behavior can be achieved at only one
wavelength with a given window.
The use of windows at wavelengths shorter than 3 cm is rather limited.
Several difficulties arise due to the decrease in the dimensions of the
waveguide with decreasing wavelength. The most striking of these are
(1) the increased liability to electrical breakdown in the neighborhood
of a window, (2) errors in determining the position of the element, and
(3) the machining and insertion of small parts. For wavelengths shorter
than 3 cm the step transformer, illustrated in Fig. 7.12, is recommended.
YiLis is analogous to the cascade section discussed for coaxial lines. The
characteristics of the step transformer can be expressed in terms of the
input admittance Presented at the generator sicle when the guide is
terminated beyond the section in a matched load:1

I Sichak, op. cit., p. 2.

232 MICROWAVE TRANSMISSION L1.VBS [SEC. 7.12

The parameters 1, b, d are defined in the figure; A, is the wavelength

in the waveguide; and F is a function of d/b alone, a few values of which
are given in Table 7.1. When 1 = Xg/4, Eq. (81) reduces to

TABLE 7. 1.F-FUNCTION FOE STEP TRANSFORMER

1$$% F-function
o 0
10 0.020
20 0.063
30 0.130
40 0.235
50 0.395
60 0.598
70 0.820

To design a transformer from either Eq. (80) or (80a) it is necessary to

construct a graph, based on Table 71, from which the required values
of F may be obtained. Over the range of useful values of b/d the section
can be regarded as a quarter-wave transformer with a phase correction
due to the capacitative effects at the junction. The phase correction
makes itself felt in that the load end of the transformer is not placed at
the point of a voltage minimum (the point of maximum load admit-
tance) but is displaced slightly from that point toward the generator.
To a first approximation the matching condition is that the conductance
in Eq. (80a) be equal to the maximum normalized load admittance,
b,
~
g, = . (81)
4b2F2
l+T
9
The latter directly equals the voltage standing-wave ratio due to the
load. Accordingly it is suggested that the designer prepare for himself
a set of charts of r or g, against d/b over the range of k, with which he
will be chiefly concerned. For a given case the transformer with dimen-
sions determined in the indicated manner can be prepared to slide in th~
SEC. 7.13] CIRCULAR WA VEGUIDE: TM- AND TE-MODES 233

guide, the bottom of the transformer being tinned before insertion. The
transformer is moved along the guide until the best matching position
is located and then soldered into place by heating the outside of the guide.
7.13. Circular Waveguide: TM- and TE-Modes.Let us consider
next a hollow guide of circular cross section of radius a. As in the case
of the coaxial line we are here concerned with solutions of the scalar
Helmholtz equation in a circular region. The general solutions are the
same as for the coaxial line:

~(z~~)
= [AJm(Kr) + ILVm(.r)] cos mO. (758)
@(z,y)
Here again r, o are polar coordinates over the cross section, and m is an
integer. In the present case, since there is no inner conductor, there are
no sources in the interior and the fields must be finite at all points. The
Neumann function, however, becomes infinite at r = O; accordingly it
must be removed from the solution: B must be equal to zero. The funda-
mental solutions are, therefore,

4(%Y) = ~Jm(.r) cos mf3.

(82)
41(x,y)

a. TM-modes.By the boundary condition of Eq. (21b) we require

@ = O at r = a for all values of d. This leads to characteristic values
K~~which satisfy the relationl

Jm(~mna) = O.

The complete set of field components for the TMn-mode, obtained from
Eqs. (22) and (23), are

E, = K;n COSmd~~(.nn~)e-y ; H, = O,

E, = 7= Ho = ~.,~.m. cos me~~(.~.r)~-y,

(83)
~nt(Knw,~) e_7,,,nz
E,g= ~H,=mymnsinmO
r .1

The field configurations for several of these modes, together with the
cutoff wavelengths, are shown in Fig. 713.
b. TE-modes.The function t(x,y) is subject to the boundary con-
dition of Eq. (13): d~/dr]~~ = O for all 0. The characteristic values .nn
satisfy the relationz

Ym(Kmfla)= o. (84)

1For lowerroots z~. = .~.a of this equationseeE. Jahnkeand F. Erode, Tables of

Functions, Dover PublicationsReprint, New York, 1943,p. 168.
2For the lowerroots .A = K,,AIof this equation see ibid.
234 MICROWAVE TRANSMISS1O.V LIKES [SEC. 7.13

The field components for the general TE~.-mode are found to be

H. = K~. cos mdJ~(Kfi.r)e-~z; E. = O,

H, = ~~ Eo = Kn.ymN cos mO~~(K~sr)eYZ,

(85)
Jm(Km.?)~_,mnz
HO = ~~ E, = mym. sin mo
r .1

These modes are illustrated in Fig. 7.14. On examination of the roots

of the Bessel functions and their derivatives it will be seen that the lowest
mode, that is, the mode with the longest cutoff wavelength, is the TEI,.

(a) (b)

(c)
FIO. 7.13.7M-mode in circular waveguide: (a) ZWo,-mode [M,() = 1.31dl; (~)
!lMo,-rnode [km{) = 1.07d]; (c) TM,~-mode [xI,(I = 0.82d]. electric field;
magnetic field.

mode. This is the mode generally utilized in antenna systems. It is the

circular guide analogue of the TE10-mode in rectangular waveguide.
The use of circular waveguide is limited by several factors, of which
perhaps the most significant is instability in orientation of the field con-
figurations. Since the guide has rotational symmetry, the field configura-
tion can be rotated about the z-axis without violating boundary conditions;
there is no preferred direction 0 = O. Small irregularities in the wall
of the guide or matching windows can cause such rtitatirm of the fields
giving rise to subsequent difficulties in designing the radiating system.
In rectangular guide, on the other hand, the orientation of the field
configuration is uniquely determined by the orientation of the cross sec-
tion. Another difficulty in round guide is mode control over an appre-
ciable frequency band. The radius is the only parameter availahlc to
determine the cutoff ivavelength; in rectangular guide, both the dimel,-
sions a and b enter into the characteristic values of the higher modes.
Sk:(, 7.15] l.i liiil.1.IiI,-II,A YE }} :1 }EC[ IIIE 235

Other comparative factors will be pointed out in the discussion of wave-

guide and horn feeds.
7.14. Windows for Use in Circular Guides.As with rectangular
guides, metal diaphragms can be inserted into circular guides to serve as
matching devices for the Z1111-mode. The circuit equivalents of these
windows are again reactive elements shunted across the two-wire line
representation of the dominant mode. Capacitative windows cut across
the E-lines, \vhile inductive windows cut across the transverse magnetic
field inthe cross section. Theadmittance characteristics ofsuchwindows

@@

(a) (b)

(c)
1:1,;. 7.14. 7E-,,Iodr* i,, r,rrular waveguicles: (a) Tfl,,-mode [ho,(c) =0.!32d]; (b)
Il?,l-!llode [A,, == 1.71(1]; (c) TEI,-111OCIC
[k) = 1.OM]. electric field;
n~aguctic field.

may be found in the literature on the subject. There is also available

aresmmnt window which can beusedasafrarne to support a thin dielec-
tric sheet to seal the waveguide.
7.16. Parallel-plate Waveguide..$mother type of waveguide that is
used in micro~rave antennas is that formed by a pair of parallel plates.
The modes can be derived, asin thepre~-ious sections, by a direct solu-
tion of the field equations, in the present case for a region bounded by a
pair of parallel perfectly conducting surfaces of infinite extent. It will
be instructive, ho~veverj to treat the parallel-plate system as a limiting
case of the coaxial line and the rectangular waveguide.
The parallel-plate \vaveguide can be derived from the coaxial line

I ,Vicro?L,ave Tru7Lst/tission L)esfgn Dflta, SperryGyroscopeCompany, 1944; Wave-

g~iideHandbook,) 1{1.Group Report 3-0.43, Feb. 7, 1945; JVaveguide Handbook
S,, pplemer,t,] RI, Grollp lleport Xo. 41, Jan. 23, 1945; 11awguideHandbook, Vol. 10
~t (his series.
236 MICROWAVE TRANSMISSION LINES [SEC.7.15

by allowing the radii a and b of the inner and outer conductors to become
infinite in such a way that the spacing b a between the conductors
remains constant:
ba=s. (86)
It will be recalled that the ZEM-mode is independent of the radii of the
conductors and is supported by the line for all frequencies with a wave-
length equal to that in free space. We thus arrive directly at the result
that the parallel-plate guide supports free propagation of a TEM-mode
at all frequencies. The electric vector is perpendicular to the plates,

~ ,.=S,Z

o
Fm. 7.15.Theparallel-platewaveguideas a limitingcaseof a coaxialline.
and the magnetic vector is parallel to the plates; neither field vector has
a component in the direction of propagation. Taking Eq. (53) for the
electric vector of the !fEilf-mode and writing r = a + y, b = a + s (cj.
Fig. 7.15), we find that the magnitude of the electric field is
1 v
(s7)
+!J):_
g,+
a

Letting a become infinite we obtain

v
Iim IEI . (88)
a+ cc s
The magnitude of the electric-field vector is independent of position
between the plates; the same result is obtained for the magnetic field.
It will be recognized that Eq. (88) is the same expression as for the static
electric field between a pair of plates at a difference of potential V.
Considering next the TE- and TM-modes of the coaxial line we note
~hat as the radii become infinite, the periodicity condition disappears;
that is, we need concern ourselves only with the modes of order m = O
[Eq, (5S)].
TM-modes.-The longitudinal component of the electric field [the
function o(z,y) in Eq. (58)] is
E. = A.J,(Kr) + BNo(Kr). (89)
Making use of the asymptotic forms of the Bessel functional for large
(.I-), we get
1G. N. Watson,Bessel Fundiom, 2d cd., Macmillan, New York, 1945,Chap. 7.
SEC. 715] PARALLEL-PLATE WA VEGUIDE 237

+JIAcOs(K+Ka-:)+
Bsin(K+a-$1
We have here introduced again r = a + y. In the limit a = m, the
solution takes the form

E, A COS(K~ +,). (90)

Applying the boundary conditions EZ = O at y = O and y = s, ~ve find
that 7 = (7r/2) f 2rmrand that the characteristic values of the modes are

% = ?~. n=l,,2, . . . . (91)

s

Equation (9o) can thus be re\vritten as

(92)

The cutoff wavelength for the TMm-mode is [Eq. (16)]

(93)

and the guide wavelength for the freely propagated mode [Eq. (17)] is

[1-/$1
(94)

g=
The transverse components of the field are obtained from E, by means
of the set of Eqs. (4) and (5):

(95a)

(95b)

the constant ~n is defined by Eq. (14).

TE-modes.The derivation of the TE-modes proceeds in a similar
manner. Equation (90) in this case represents the longitudinal com-
ponent of the magnetic field; that is,

H, = ~ 0S (KY + T). (96)

The boundary conditions c?H./dy = O at y = O and y = s lead to the

result that 7 = + 2mrr and

n = El, n=l,2, .,, (97}

s
238 If Iclto JVA Vfl TRAiVS.flf i SSIOiV LI.i-IiS [Sm. 716

The cutoff wavelength for the TE.-mode is given by Eq. (93), and the
guide wavelength by Eq. (94). The complete set of field components is

H.= Acos
() ~, (97a)

E= = ~~!! Al sin Zy , E, = O, (97tl)

K. () s

Hu=~A sin ~ ; Hz = O. (!17c)

Km ()
It will be recognized that the field distributions and guide \~ave-
lengths correspond to TM- and TE-modes of the rectangular guide.
The T.kfn-rnodes of the parallel-plate system are the analogues of the
TM1, n-modes, and the T.En-modes are the analogues of the TEO,n-modes.
As the broadside dimension a of the rectangular guide becomes infinite,
the modes of the latter pass into parallel plate modes.
7.16. Design Notes.Several remarks on design practice may prove
of interest to the reader. These are particularly concerned with coaxial
lines and circular waveguides. Unless an antenna is being developed as
a single experimental model or for production in very limited numbers,
some attention should be given to the production problem or the avail-
ability of parts. With respect to the coaxial lines and circular guides,
dimensions should be chosen as near as possible to those of commercially
standardized tubing. The primary considerations in the choice of
dimensions are, of course, the characteristic impedance of the line and
the control of higher modes; these, however, allow some latitude in
design.
Special care should be taken in the inspection of tubing. Erratic
results in standing-wave measurements on lines have frequently been
traced to irregularities in the cross section of the line. Ridges and \vaves
are found in the tube wall if the die through which the tubing was
extruded is worn or if the driving unit is faulty. Such ridges and waves
can be detected only by cutting the tube in half. It is recommcmded that
a sample length of tubing from each ne\v lot be cut down the middle
for inspection before using the material. It is often useful to force a steel
ball of proper diameter through the tubing under pressure, thus sizing
and polisbg the inside surface.
 CHAPTER 8

MICROWAVE DIPOLE ANTENNAS AND FEEDS

BY S. SILVER

The early trends in microwave antenna design grew out of the prac-
tice of using dipole systems at longer \\-avelengths. Nevertheless, little
systematic information has been obtained about microwave dipole sys-
tems. This is partly due to the greater difficulty in applying theory to
practically useful microwave dipoles and partly to the urgent military
needs which prevented systematic research during the early development
in this field. More recently, attention has been concentrated on wave-
guide and horn radiators, which are more amenable to quantitative
analysis. Consequently, the design of microwave dipole antennas is
s$ill in the empirical stage; quantitative data are available only with
reference to particular systems.
8.1. Characteristics of Antenna Feeds.The dipole systems that we
shall consider in this chapter are, ~\-itha few exceptions, designed to
serve as primary feeds to illuminate reflectors; it will be assumed through-
out, unless the contrary is noted, that this is the end in view. The
general design requirements and specifications imposed on primary feeds
are the following:
Radiation Patter-n.-It is evident that a primary feed radiation pat-
tern must be directive, with the major fraction of the energy radiated
toward the reflector. We have studied in Chap. 6 the relation between
the radiation pattern of the antenna as a whole and the intensity and
phase distribution over the aperture. The relation between the latter
and the primary pattern ~rill be developed in later chapters on the design
problems of special types of antennas. It may be noted here, however,
that the design of a reflector-or ~.lensis generally based on the assump-
tion that the feed is a point source. De\-iations of the feed from a point-
source radiator result in phase errors over the aperture of the antenna.
Particular attention must be paid to the phase. It was shown in
Chap. 3 that many idealized radiating systems are effectively point
sources in the sense that the equiphase surfaces constitute a family of
concentric spheres. This situation is realized only approximately in the
case of an actual feed. The pattern of the latter is usually specified in
terms of the principal E- and H-plane patterns (Sec. 3 18). In each of
these planes it should be possible to find an equi~alent center oj feed,
239
240 MICRO WAVE DIPOLE ANTENNAS AND FEEDS [SEC. 82

with respect to which the equiphase lines are circular (to within a pre-
scribed limit of error) over the region to be covered by the reflector.
To minimize the problems of reflector design it is essential that the centers
of feed for the principal planes be coincident. In general, it is desirable
that on a sphere about the center of feed the phase shall be constant to
within t r/8, corresponding to path differences of f A/16; for some
purposes, path differences of + A/8 can be tolerated. The cone within
which the feed is a point source in the sense of these criteria will be referred
to as the point-source cone.
Impedance. Impedance match is required over as broad a frequency
band as possible. An antenna is generally considered to be usable
throughout the frequency band in which the voltage standing-wave ratio
is less than 1.4. Since interactions with the reflector tend to increase
the total mismatch, it is desirable to keep the feed mismatch below the
figure given above.
Power-carrying Capacity .This is limited by electrical breakdown
which may occur within the feed line and around the feed components
under the peak voltage of a transmitted signal. The effect of matching
devices on breakdown characteristics was noted in Sec. 7.9. The break-
down problem is particularly significant in antennas intended for air-
craft, because the breakdown potential decreases with increasing altitude,
due to the decrease in atmospheric pressure and the increase of free ion
content. Feeds for high-altitude airborne systems must therefore be so
designed that air can be held in the r-f line under pressure. The average
requirement is 10 to 15-lb gauge pressure relative to sea level atmospheric
pressure.
Weather Protection. Antennas must be protected from the weather
to prevent corrosion and consequent power dissipation in the antenna
structure. Weatherization is an important consideration in shipborne
antennas, which are exposed to sea-water sprays.
Mechanical Strength, Light Weight .Antennas installed in aircraft
and ships are subject to high stresses due to rapid changes in the motion
of the airplane or oscillations of masts of the ship in a high wind. In
aircraft systems, mechanical strength must be attained with economy of
weight.
Reasonable Tolerances. Tolerances should not be so close that pro-
duction methods cannot be used effectively.
8.2. Coaxial Line Terminations: The Skirt Dipole.-The theoretical
prototype of the dipole radiators is the half-wave dipole fed at the center
from a balanced two-wire transmission line. The significant features of
this system are the following:

1. The two wings of the dipole carry equal currents.

SEC. &2] COAXIAL LINE IERA{INA TIONS: THE ISKIRT DIPOLE 241

2. The current distribution is determined by the dipole structure,

interaction between the dipole and the transmission line being
negligible,
3. The dipole termination does not upset the balanced condition of the
line.

While it is true that a coaxial line propagating the 7EM-mode is equiva-

lent to a balanced two-wire line, it is virtually impossible to make a

T1,

1
4
12

Die-
lectric
bead
support

(a) (b) (c)

FIG. S. I. Skirt dipole: (a) simple form; (b) tapered gap to improve the impedance
characteristics; (c) decoupling choke Ca to prevent current leakage along the outer wall.

microwave dipole termination that behaves like the theoretical prototype.

The skirt dipole illustrated in Fig. 8.1 is an example of a coaxial-line
termination that is used extensive] y at longer wavelengths and to a lesser
extent in the microwave region. The two wings of the dipole consist of
the unshielded section of the inner conductor and the folded-back section
of the outer conductor (A in Fig. 8.1); we shall refer to the latter as the
skirt. If the lengths 11, 12, of the respective elements are each about
A/4, the system approximates a center-driven half-wave dipole. This
termination maintains the radial symmetry of the line; the current dis-
tribution over the wings is radially symmetric, and the radiation pattern
has the axial symmetry of the idealized system.
It is to be expected, however, that the meridional pattern will differ
from that of the line radiator. One reason for this is the fact that the
current distribution is spread over a finite area instead of being confined
to a line. The currents at different points on a circumference of the
skirt are consequently at different distances from a field point and give
242 JIIChOWAVE I) IPOI,E A.VTE.Y,VAS A.VI) IEEDS [SEC.83

contributions to the field that have correspondingly different phases.

These phase differences are more significant \vith microwavesthan \vith
long waves, since they depend on the ratio of the skirt diameter to the
wavelength. Thepatterni salso affectedly the finite dimension of the
gap at the driving point; this causes the current distribution along
the length of the dipole to deviate from the sinusoidal distribution of the
line dipole with an infinitesimal gap.
A further major factor is the coupling between the field of the dipole
and the outside wall of the line, which produces a current distribution
down the line beyond the skirt. This current distribution also radiates;
the total pattern arises from superposition of this field and the dipole
field. Thepattern rapidly becomes less satisfactory asthec~lrrent on the
line increases; so the line current must therefore be kept as small as
possible. It can be controlled in part by changing the cavity Cl formed
by the skirt and the outside wall of the line. This region constitutes a
shorted section of line and as such presents at the open end of the skirt
a reactive impedance in series \vith the dipole and the outer wall of the
line; by making the depth k/4, the reactive impedance can be made
infinite. In practice it is found that best results are obtained with a skirt
of length somewhat less than A/4. Proper operation is obtained only
at the design frequency, since the impedance of the choke Cl varies rapidly
with frequency. Improved over-all impedance characteristics have been
obtained by shortening the skirt and compensating for the reduced
physical length of Cl by filling it with dielectric to bring the electrical
length up to A/4. It has also been found that, more efficient decoupling
between the dipole and the outer line can be effected by means of a second
choke Cz mounted as shown in Fig. 8. lc. The electrical depth of C2
should again be a quarter wavelength, so that the choke presents an
infinite impedance at the open end. Experimentally it is found that the
decoupling is most complete when the separation of C, and C, is 0. 15A.
The structure of the gap G plays a significant part in determining the
over-all impedance characteristics of the antenna. Because an abrupt
discontinuity in structure gives rise to a reflected wave in the line, it is
natural to replace the region G in Fig. 8. la by the tapered structure shown
in Fig, 8,1 b. The increased diameter of the dipole stub also contributes
to maintaining uniform impedance over a larger frequency band (cj.
Sec. 8.5). Further methods of controlling the impedance characteristics,
such as decreasing the length 11and loading the stub with a sphere (capaci-
tative loading), will occur to the reader; w-e shall not dwell upon them
here.
8.3. Asymmetrical Dipole Termination.-The asymmetric dipole
terminations sho\vn in Fig. 8.2 are designed to give a radiation pattern
with peak intensity along the axis of the feed line. The dipole in Fig.
82a is center. fed from a two-wire line. The asymmetry of the tcrmina-
tion unbalances the current distributions on the inner and outer conduc-
tors of the line, with the result that the two ~vings of the dipole are not
equally excited. Also, strong coupling exists between the dipole system
and the outer wall of the line, giving rise to radiating currents on the
latter, just as in the case of the skirt dipole. The choke C remedies the
situation to some extent; with a depth 1. of about A/4 the choke presents
at its open end an infinite impedance, in series bet~reen the outer wall of

c
. t

I
-
. L 1=;

~
1,

+
I
(a)

L
c
- -- +--s
7
- - / _ -
__ _
------
!
b!
ls +
D
(b)
FIG. S2.-Asymmetric dipole ter,ninatinns: (a) opewend termination; (b) stub-support
terlmnati on.

the line to the left of the choke and the region of the line to the right.
This serves to confine most of the outer-wall currents to the region between
the dipole and the choke.
The open-ended termination has poor structural properties. In order
to maintain alignment, of the dipole \rings it is necessary to fill the termi-
nal region of the line \vith a dielectric plug. The latter gives rise to
further problems of impedance mismatch and to poor contact between
the dielectric and the conductors, \\-hichmay lead to electrical breakdovm;
the seals generally deteriorate under exposure to moisture and thermal
and mechanical stresses. In addition, radiation from the open-ended
coaxial line distorts the dipole pattern. These defects are absent in the
stub-support termination she\\-nin Fig. 8.2b. The coaxial line is con-
tinued for a distance 1.= X/4 Imyond t}~e dipole system and terminated
there in a metal plate. The latter rwgion, known as the terminating
244 MICROWAVE DIPOLE ANTENNAS AND 11EEDS [SEC. 83

stub, is again a shorted section of line, presenting a reactive impedance

at the input end. Consideration of the current division at the driving
point of the dipole shows that the stub is equivalent to an impedance
shunted across the gap between the dipole wings. With 1. = A/4, this
impedance is practically infinite; electrically the system is equivalent
to an open-ended termination.
The disparity in the currents on the two wings of a stub termination
is even greater than that in the open-ended termination. The dead
wing (or stub) D is excited only by leakage currents which make their

4
0 3
s I(A)
0
~
-f(A) ~_ _

I(B)

% A = 10.>0 cm
u

~~
5 10 15 20 25 30
Oistance from center of the dipole, cm
FIG. S.3.Leakage currents along the line; stub-supported dipole-disk feed without choke.

way through the opening in the outer wall and by coupling with the field
of the live stub L. As in the case of the open-ended termination, coupling I
exists between the dipole system and the center wall of the line. A
measure of the relative excitation of the dipole stubs is afforded by the
intensity of the outer-wall line currents along lines in a plane containing
the dipole axis. Figure 8.3 shows results of line-current studies made on a
dipole system carrying a reflecting plate on the terminal stub. The
standing-wave structure in the current is due to some obstruction on the
outside surface at the input end of the coaxial line.
Control of the outer wall currents is achieved by means of the choke

I
Sxc. 8.4] SYMMETRICALLY ENERGIZED DIPOLES 245

C (Fig. 8.2 b), as in the systems discussed previously. To present an

infinite impedance at its open end the choke should, nominally, have a
depth of A/4. However, because of junction effects at the open end and
coupling with the dipole system,
the optimum value is somewhat
less than A/4. Figure 8.4 shows
the line-current strength at a fixed
point on the outer wall as a func-
tion of choke depth for the system
studied in Fig. 8.3; the optimum
depth is 0.23A. Although this
value is strictly significant only
for the system illustrated, it has
been found to give good results in
other dipole systems employing
chokes; it is a suitable value for the
depth of the choke Cl of the skirt
dipole considered earlier.
It has been noted that the
effect of the choke is to confine the 0.1 0.2 0.3
outerwall current to the region be- Choke depth .f wavelengths
tween the choke and the end of the lSIG. 8.4.Leakage current as a function of
choke depth.
line. This current distribution
serves as a linear radiator along the axis of the feed line. From the
general considerations of Sec. 3.15 it will be evident that this radiates no
energy in the direction of the line axis; it will, in general, give rise to a
pattern with peak intensity on a cone having its axis coincident with the
line axis. The phase of the line current with respect to the dipole cpr-
rent is determined by the position of the choke with respect to the dipole
system. In combination with a paraboloidal mirror, in which the feed
line lies on the axis of the mirror, the interaction between the dipole and
line-current system produces a phenomenon known as squint, in which
the over-all antenna beam is pointed, not along the axis of symmetry of
the system, but in a direction making a small angle with that axis. Use
is made of this phenomenon for scanning.
In closing the discussion of the asymmetric terminations, it should be
noted that the input impedances of both the choke and the terminating
stub vary rapidly \vith frequency. As a result, these structures are
strong contributing factors in the frequency sensitivity of the impedance
of these antennas. In addition, the cut-a\vay region of the line introduces
distributed capacities and inductances. lhcse factors restrict the
usability of the antenna to a narrow frequency band.
8.4. Symmetrically Energized Dipoles: Slot-fed Systems.The
shortcoming of unequal excitation of the dipole stubs, which charac-
246 MICROWAVE DIPOLE AAT7Ii,\,V.lS :l\Il F>;);l)S [SE(. 8.4

terizes the terminations discussed above, is eliminated in the slot-fed

systems shown in Fig. 8.5. Both wings of the dipole are mounted on the
outer conductor, in which a pair of slots S is milled in a plane normal to
the dipole axis. The inner conductor is short-circuited to the outer
conductor on one side by the post P, which usually is in the line of the
dipole axis but may be inserted at any point along the line in the slotted
region. Both open-ended and stub-terminated systems are used, ana-

--- 6!!?
s-s
P

Y
(a)

n
[

..
-i

+
(b)
FIG. 85.-Slot-fed dipole terminations on coaxial line: (a) opeme,,ded termination; (b)
stub termination.

logous to the systems discussed in the preceding section. The open-

ended type is used as a radiating element in linear arrays (cf. Sec. 9.8).
The operation of the dipole can be interpreted from various points of
view. Perhaps the simplest picture is that the radiating system is
energized by a voltage impressed across the slot. The origin of the
voltage becomes evident on consideration of mode relationships in the
slotted region. In the absence of the short-circuiting post P we would
have the ZEM-mode and possibly higher modes generated in the open-
ended termination or, in the case of a wide slot, generated by the slot
itself. All these modes, however, would be symmetric with respect to
the plane containing the axes of the slots and give rise to no impressed
field across the slot; under these conditions the dipole is not excited.
With the insertion of the post, modes are generated that are symmetric
SEC. 8.4] SYMMETRICALLY ENERGIZED DIPOLES 247

with respect to the plane determined by the axis of the post and the axis
of theinner conductor. These modes, whensuperposedo nthep receding
set, must give rise to a field such that the tangential electric field is
zero over the surface of the post. In the case of a narrow slot we can
ignore (for the qualitative picture) themodes generated bythe slot itself;
the prime effect of these modes istorelax thecutoff conditions and allow
propagation within the slotted region of some of the modes generated
by the post. The most significant of the latter is the ZE,l-mode; Fig.
86 shows how superposition of the Z11,,-mode on the 7EM-mode leads

TEM TE, , E =Ern,,

W,>. S6.-Superl,c,siti(>,, O{ the TE.11- a,)d TE, I-mncies in the slotted region of the slot-fed
dlpol c.,

to a field contig~lration that satisfies the requisite boundary condition

on the electric field in the case of a thin post. The resultant configura-
tion gives a field th:~t is zero along the post and increases with angle to a
maximum value directlyf opposite to the post. It is readily seen that
this impresses a voltage across the slot, with resulting excitation of the
dipole structure.
The slotted dipole can also be analyzed from the transmission-line
point of view. The slotted region is conceived as a three-\vire trans-
mission line; this is the appropriate represcntat ion of a waveguide sup-
porting simultaneous propagation of two modes, just as the t\vo-wire
line reprments single-mode propagation. It \vill carry us too far afield
to discuss tl~e general theory of three-\vire lines. z The equivalent cir-
cuit relJr(:sellt:LtiL)Ilsfor the open-ended and stub-terminated systems are
sholvn in l~ig. 8.ia :LIK1 b rcspwtivciy, for t,hc case in wh~ch the post
lies along the dipole axis. Here Z,, is the impe(lance across the pair of
lines ronncci cd I)y the I)ON; Z,; the impedance at, the gap opposite to the
Post; ~A is tile irlpllt inlpe(kn(() of ~~dipole ha~in~ the same \~in~strut-
trlrc as in (1w given systcrn, I)Ilt cent rl-fed from a balanred two-\vire line;
(, is the length of the slot ; :ln(l 1, tl~c length of the terrninating stub.
it the end of the slot tlL( OIIter lines are short-circuited, the three-wire
line passing into the tIV(j-ifire line
I 11. I{ll)kt,Sl~ltt[,{l l)ill{llt IIIIIIr(l:LIIrCThm)ry, 1{1. Report No. 772, Nov. 21,
I!)45.
z SW S. 0. Rice, St(,:l{ly St:ltt, ,S(jl(ti[[)lis {~f Tr:msmissioll I,ine I; qtlations,
/if// s!/sh-m 1ech Jo(()., 20, I:1I ( I!1-11 1.
248 MICROWAVE DIPOLE ANTENNAS AND FEEDS [SEC. 8.5

In the case of the open-ended termination with a riarrow slot it is

possible to reduce the system to a two-wire line with appropriate loading,
as shown in Fig. 8.7c. The impedance ZP has been taken to be zero;
Z, is the characteristic impedance of the three-wire line under the con-
dition that no current is flowing in the central line. It will be seen that
in this particular case the slot con-
1----- ls -----1 tributes only a susceptance, like a
short-circuited section of two-wire

4BD
line. This circuit representation
indicates that the length of the slot
can be so chosen m to match out
the other reactive impedance ele-
(a) ments involved in the termination.
The slot not only equalizes the
excitation of the wings but also
p- !s,
serves as a choke element to de-
couple the dipole system from the
J outer wall of the line. The result-
ing system is completely free from
the squint phenomenon associated
(b) with the asymmetric termination.
In the case of open-ended termina-
tions it is possible to design units
with high power capacity; these
j+ tan% have found application in linear-
array antennas. The stub-termin-
ated units, on the other hand, are
more limited in their power capacity
ZE@ than the corresponding asymmetri-
(c) cal terminations and have been
FIG, S7,-Three-wire line representa- used in place of the latter only where
tion of the slot-fed dipole: (a) opeu-euded it is imperative to have a squint-free
termination; (b) stub-terminated line; (c)
reduced equivalent loading for Case a. system and relatively lower po\ver
levels are acceptable.
8.6. Shape and Size of the Dipole.The impedance problem has been
a troublesome one with dipole feeds, largely because of the frequency-
sensitive elementssuch as the choke, terminating stub, and slot-
needed in making various types of terminations. A certain measure of
adjustment is available in the size and shape of the dipole. The depend-
ence of the impedance of a center-fed dipole on its size and shape has
been the subject of considerable theoretical work. All of the work
1S. A. Schelkunoff, ,?itectromagrLeticWaves, tan Nostrand, .New York, 1943,
Chap. 11; L. J. Chu and J. A, Stratton, Jour. Appli.d Phys., 12,241 (1941); R. W. P.
King and D. D. King, low. Applied Phys., 16,445 (1945).
SEC. 85] SHAPE AND SIZE OF THE DIPOLE 249

applies to an idealized system in which the dipole is driven from a balanced

system across an infinitesimal gap; it is assumed that the coupling
between the dipole and line plays no part in determining the current
distribution of the wings. As we have noted above, this condition is
never realized in microwave systems where the dimensions of the feed-
line cross section are comparable to those of the dipole structure.
The theoretical results, however, are helpful in a qualitative way.
The various theories differ in quantitative details concerning the values
of the impedance, but all show the same general qualitative features.

g 600 300
5~. 500 200
f
: 400 g lofJ
g
; 300 o
~
~ 200
,*Q : 100
: 100 $ 200
so 300
0.2 0.3 0,4 0.5 0,6 0.70.8 0.20.30.4 0,50.60,7 0.8
-VA %
(a) (b)
FIG. S41-Inrwt impedanceof spheroidaldipoleswithmajoraxisL and minoraxisD:
(a) realcomponentor radiationresistmce:(h) imaginarycomponentor reactance. (From
L. J. Chu and J. A. Stratton, J. A PPL physic-s, by courtesy of tb au.thuraand the American
Instiiute of Phwti.)

The curves shown in Fig. 88 are taken from the work of Chu and
Stratton. They apply to spheroidal dipoles, the major axis of which is
designated by L and the minor axis by D.
The curves show the dependence of the real and imaginary compo-
nents of the impedance on wavelength for various values of the ratio
I./D. It is observed that in the neighborhood of the resonant point,
which corresponds closely to a length equal to X/2, the resistive compo-
nent is virtually independent of the value of L/D and is equal to about
70 ohms. The dependence of the resistance on wavelength does not
become marked until the length is considerably larger than the resonant
value. The reactive component, however, is seen to be a decided func-
tion of the frequency. The larger the ratio L/D, that is, the thinner the
dipole, the more rapidly does the reactance vary and the sharper is the
resonant point. Thus, a thin dipole is more frequency-sensitive than a
fat dipole. The dipole dimensions can be chosen such that its reactive
component balances the reactance which is associated with the termina-
tion; this in general ~vill lead to better over-all impedance characteristics
for the antenna than the choice of a dipole that alone has a flat reactance
characteristic. The impedance characteristics of the dipole can also be
250 MICRO WA VB DIPOLE A.V7hiV.VA S .I.VD FEJ91M [SEC. 86

controlled by such processes as top loading ~vith a sphere or other struc-

ture in the same manner as is done at longer wavelengths. Here again
the procedure is entirely empirical, and we shall not dwell upon it any
further.
8.6. Waveguide-line-fed Dipoles.It is much simpler to feed a dipole
irom a waveguide line than from a coaxial line. The technique of termi-
nation is shown in Fig. 8.9. The dipole is mounted on a web that fits
into the mouth of the guide, parallel
to the broad face of the guide and
transverse to the electric vector in
the dominant TE,,-mode. The E-
FIG. S9,-Dipole terminationon wave-
euide. vector is thus parallel to the dipole,
which is driven by the radiation
incident on it from the mouth of the guide. It is obvious that if the
web is inserted ~ymmetrically, the two ~vings of the dipole are excited
equally. The taper shown in the diagram serves as an impedance-
matching device; it also improves the radiation pattern in that it decouples
the outer wall of the line from the dipole. The impedance of the system
is also determined by the depth of insertion of the web and the position
of the dipole with respect to the mouth.
8.7. Directive Dipole Feeds.The design of directive feeds is based
on the principle of interference bct~veen dipoles properly spaced and
phased (Sec. 3.17) and on the principle of images (Sec. 5.3). Early
designs utilized the skirt dipole with a reflecting plate and the open-ended
asymmetric termination followed by a second dipole or a reflecting plate.
These designs have very poor structural characteristics; they will not be
discussed here. Stub-terminated coaxial systems and waveguide systems
lend themselves admirably to the construction of directive feeds, the
stub or web providing mechanical support for the system of dipoles
involved or for the reflecting plate. These directive systems are designed
to radiate maximum power back along the feed ~ine; the reflector that is
to be illuminated by the feed is then also mounted on the feed line.
This rear-feed type of installation (examples of which are to be seen in
Sec. 12.1 1) minimizes the length of line and the series of bends and joints
required (factors of considerable importance for generator stability) and
forms a compact and rugged system.
The directive system employing a reflecting plate, which may be
termed a dipole-plate or dipole-disk feed, is based on the principle of
images. In accordance with the general theory, to produce peak intensity
along the feed line the reflecting plate is mounted a distance A/4 behind
ihe dipole. The principle of images assumes, of course, a reflecting
plate of infinite extent. In the case of the feed system the plate must be
kept as small as possible. otherwise the feed will present too extended
an obstruction in the path of the energy reflected from the large mirror;
SEC. 8.8] l) IPOLE-I)ISA- FEEl~S 251

the effects of such aperture blocking on the over-all alj bmna patterri
are discussed in Sec. 6.7. It is thus necessary to sacrifice a certain
measure of directivity, with the result that the primary feed has a back
lobe, that is, radiation behind the reflector plate; this, too, has a sig-
nificant effect on the over-all antenna pattern (cf. Sec. 125).
The coaxial-line-fed rnultidipole systems are usually designed so that
orily one dipole is excited directly from the line. The other members
(dummy or parasitic dipoles), arranged in a linear array, are fed by
coupling with the directly excited element. Micro\vave feeds have
usually included a single du rnmy element to complete a double-dipole
system such as that discussed in Sec. 3.18. In that section the case of
h/4 spacing and relative phase i = 7r/2 was considered in detail. How-
ever, by reference to 13q. (3174) of Sec. 318, it may be seen that an~
pair of values of spacing a and phase }vhich satisfy the relation

ira -!!
rr7=o,1,2, . . .
x 2= m

will give peak intensity along the direction normal to both dipole axes,
that is, along the feed line in the practical case. These other systems,
however, unlike the (k/4, m/2) system, in general also give rise to a back
lobe in the direction 180 away from the peak. In practice, the phase of
the dummy relative to the driven element is controlled I)y the relative
dimensions of the dipoles as well as by their spacing; from Sec. 85 \ve
see that it is possible to make one dipole capacitative or inductive relative
to the other, by proper choice of dimensions.
Directive feeds will be further discussed with reference to particular
systems. In the following sections design data are presented on a num-
ber of feeds that have been developed in the Radiation I,aboratory and
used extensively. It is not to be assumed that the results given here
represetrt the ultimate that can be achieved with these systems.
8.8. Dipole-disk Feeds.Two dipole-disk systems have been devel-
oped, employing respectively the stub-terminated asymmetric dipole
and the stub-terminated slot-fed dipole.
a. Asymmetric Dipole Termination .Three such feeds have been
designed to illuminate paraboloidal mirrors, of focal length 10.6 in. and
30-in. aperture, at ~ravelengths of 9.1, 10.0, and 10.7 cm respectively.
Details of the feed assembly are given in Fig. 8.10. The line has a
characteristic impedance of 46 ohms; its dimensions are outer conductor,
OD = 0.875 in. \vith ~vallthickness of 0.032 to 0.035 in.; inner conductor,
OD = 0.375 in. Reasonable directivity was obtained with a reflector
plate with diameter about 0.8X. Thr principal E- and H-plane feed
patterns are showm in Fig, 8.11. The peak intensity of the pattern is

I S. Brwn and R, Hiatt, IL 1, Ib,lmrt So. .54-23, .Junc 21, 1943,

252 MICRO WAVE DIPOLE ANTE.V.VAS AND FEEDS [SEC. 88

Effective center
Transparent sphere of feed
r

\
\seam

FIG.8. 10.Dipole-disk feed assembly.

.200 0 20 00 3400

400

320
6(Y

300
800
280

260
1000

120 140160180 120 140160180200220 240

(a) (b)
FIG. 811. -Prin!ary pattern of dipole-disk feed: (a) H-plane: (b) -E-plane.
dipole-disk feed of Fig. 8. 10; --- tbcoret,ical pattern of a dj~ole at a dlst,ance X/4 from
an infinite plane,
SEC, 8.9] DOUBLE-DIPOLE FEEDS 253

directed along the feed line, and the data in this region are, therefore,
somewhat uncertain. The dotted portion of the curves have been
obtained by extrapolation. It is observed that the E-plane pattern is
not symmetrical. This is due to the fact that one wing of the dipole is
excited more strongly than the other in the asymmetric dipole termina-
tion, as was pointed out in Sec. 83. The peak appears on that side of
the axis which corresponds to the dipole wing carrying the major portion
of the current. The H-plane pattern, on the other hand, was found to
be accurately symmetrical corresponding to the symmetry of the dipole
structure in the plane. For comparison, there are plotted the theoretical
patterns for the ideal system of a dipole placed A/4 in front of an infinite
reflecting plane. It is seen that the feed pattern is considerably more
directive; the gain of the feed is found to be equal to 7.
The E- and H-plane centers of feed are coincident, lying between the
dipole and the disk, somewhat nearer to the latter. The point-source
cone is more than adequate to cover a mirror with dimensions given
above. The unpressurized feed has a peak power capacity of 35o ~ 35
kw. With suitable matching transformers it has been possible to realize
an impedance characteristic for the composite system (feed and para-
boloid) such that the standing-wave ratio r did not exceed 1.23 over a
band of i 3 per cent about the matching frequency.
b. Slot-jed Terminataon.-A unit designed to operate at a wavelength
of 9.1 cm with a paraboloidal mirror of 3.6-in, focal length and 12-in.
aperture is illustrated in Fig. 812.1 The line has a characteristic imped-
ance of 45 ohms, with an inner conductor of T5K in. diameter. The smaller
line was used here to reduce weight, the power requirements on the feed
having been smaller than in the preceding case. It will be noted that
the disk diameter here is about 0.5k. The system has a single center of
feed for both principal planes and is completely free from squint. The
composite antenna made up of the feed and the mirror indicated above
has an impedance band of + 1.25 per cent about the design frequency
over which r < 1.23.
8.9. Double-dipole Feeds. a. Coaxial-line-jed Sysl em ,Such a feed 2
is illustrated in Fig. 8.13; it is a light\veight unit employing a +-in. line
like that discussed in Sec. 88b. The spacing between the dipoles is very
nearly h/8; correspondingly, the parasite element is longer than the driven
element in order to produce the proper phase relationships. This SYS-
tem, like those discussed above, has a unique center of feed. An antenna
consisting of this feed and a paraboloidal reflector of 3.6-in. focal length
and 12-in. aperture ha ~a standing-~vave ratio r s 1.23 in a band of + 1
per cent about the design frequency.

1 \V.~. Xowak, RL Report No, 54-26, Juiy 5, 1943,

2 [bid.
 -

~OD x 0.25 wall

FIG.8.12.Dipole-disk feed.

Center of feed

*
0

+o.l153-
A
0.250 A
FT,:.X.13. l)nlll )le-,lil,nle fwd vn maxial line,
SEC. 8 !Jj Do(BLE-DIPOLE FEEDS 255

b. JVavegui(ie Systems.-.kdo:iLie-dipole feed built upona~,aveguide

termination for use at a wavelength of 3.2 cm is shown in Fig. 814.
The two dipoles are mounted on the web so that their axes lie on the
plane of symmetry of the guide. The spacing between the dipoles is

Materml 1 x:x 0.05 wall Cut out V to dimensions as

rectangular tubing gwen, bend, and hard solder,
g
Bevel end to make it sauare, g
0 :UJ
Q ~
2 Mill after
0.125 soldering
.
1 ____ _ @O

d . . ~
:; ~:;:.2
% Hard solder
0.250
These angles
T ust be equal

~~ 0.275

0.506
-1-q
i o J221
0.256-+
4}500

~ ;

_________ +P+L7
+ u U& ~~ 0!652

FIG. S 14.Waveguide double-dipole feed: k = 3.2 cm.

about h/2.,5; again the coupling (and hence the relative phase) of the
elements is adjusted by the suitable choice of their relative dimensions.
The radiation pattern has an appreciable back lobe which is }n some
measure due to the guide itself; this is reduced by tapering the terminal
region as shown in the figure. The E- and 11- plane centers of feed are
not concident; however, their separation is negligible for most purposes,
and the equivalent center of feed can be taken to be located just behind
the first dipole.
We have previously pointed out the dependence of the impedance on
the taper, depth of insertion of the web, and the dipole factors. To
obtain reproducible results, special care must be taken to remove excess
1 }$. Sichak, DrIl]hle Dipole Rectangular \\ave (luicic Antennas, RL Report
No. 54-?5. ,June 26. 19+3
256 MI(7RO WA VE DIPOLE ANTENNAS AND FEEDS [SEC 8.10

solder at the base of the dipole and at the seams between web and wave-
guide. Antennas made up of this feed and paraboloidal mirrors of
18-in. aperture and focal length either 4.5 or 5.67 in. have a bandwidth
of + 1.5 per cent over which r ~ 1.23 if the antenna is matched by an
inductive window at k = 3.2 cm. The unpressurized antenna has a peak
power capacity of 375 kw, corresponding to 50 kw at 50,000-ft altitude.
8.10. Multidipole Systems.The web termination on a waveguide
provides atconvenient base on which to build multidipole systems in the
form of two-dimensional arrays. Two such arrays have been designed

Fm. 8.15.Four-dipole feed.

for the 3-cm band, one a triangular array of three dipoles, the other a
rectangular array of four dipoles. Only the latter has been used in final
antenna design. The four-dipole array shown schematically in Fig.
8.15 can be regarded as a pair of the double dipole units discussed in Sec.
8.9b, separated by a distance of approximately A/2. Each double-
dipole unit can be replaced by its equivalent point source, reducing the
system to two directive sources in phase, spaced ~/2 apart. It is evident
that no appreciable change is to be expected in the E-plane pattern.
The H-plane pattern, however, must be multiplied by the directivity
factor of two isotropic sources in phase and with k/2 separation. This
factor is readily found to be [COS(7r/2 cos ~)] where @ is the angle with
respect to the axis in the H-plane. Hence if P,(*) is the H-plane pattern
of the double-dipole system, the pattern P1(I#I) of the four-dipole system
is given closely by

p,J@[cos@cos@
4(4)
 CHAPTER 9

LINEAR-ARMY ANTENNAS AND FEEDS

BY J. E. EATON, L. J. EYGES, AND G. G. MACFARLANE

9.1. General Considerations. -The technique of producing directive

beams by means of arrays of radiators that are suitably spaced and driven
with appropriate relative amplitudes and phases has been used widely
at the longer wavelengths. These arrays have generally been in the
form of two-dimensional lattices with the possible addition of a reflecting
surface to confine the radiation to a single hemisphere in space. In the
microwave region, attention has been confined almost exclusively to the
one-dimensional, that is, linear, arrays. The wavelength advantage
becomes evident at once, for with economy in physical size it is still pos-
sible to have an array that is long measured in wavelengths and hence
highly directive.
The arrays that have been designed to date can be grouped into two
general clfisses: (1) end-fire arrays producing a beam directed along the
axis of the array and (2) broadside arrays producing beams the peak
intensity of which is in a direction normal to or nearly normal to the
axis. End-fire arrays have proved to be particularly useful where it is
necessary to mount an antenna close to an object; for example, such arrays
have been mounted along a gun barrel in airplanes to furnish gunfire
range information and to serve as gunfire directors. Axially symmetrical
broadside arrays which produce beams symmetrical about the axis have
been designed for use as beacons; installed both in ground or ship and
on aircraft they provide a communication system bet\veen ground (or
ship) and aircraft. The patterns of these arrays are axially symmetrical
like the dipole patterns but have increased directivity in the meridional
plane to give increased range. Other types of broadside arrays have been
developed whose beams have a fair measure of directivity also in the
plane perpendicular to the array axis. In a few cases, arrays of this type
have been used as the terminal antenna system; more frequently these
arrays have been used as line sources for illuminating cylindrical reflectors,
in which case the reflector is placed sufficiently close to the array so as
to be in its cylindrical ~vave zone.
While there is no fundamental difference in principle between long-
wave and microwave arrays, the microwave arrays present problems of
257
258 LI.VEAR-ARRA Y ANTENNAS AND FEEDS [SEC. 92

their own which are due to the wavelength region involved. In long-
wave arrays it is possible to isolate to a large degree the feeding of one
element of the array from another. Microwave arrays must be built
on coaxial line or waveguides with the result that the feeding of the
element becomes a mutual interaction problem. This type of feeding
also requires special designs in the radiating elements of which there is
quite a variety. at microwave frequencies. The physical size of the
radiating elements is generally small, and tolerance problems are asso-
ciated with microwave arrays that are generally uncommon at longer
wavelengths.
The problems and techniques of linear-array design have been divided
in this chapter into three general parts. The first concerns itself with
general pattern theory, that is, the relation between the far-zone pattern
of an array and the amplitude and phase dist ribut ion among the elements
and their spacing; in this section no attention is paid to the problem of
realizing a given amplitude and phase distribution. The second part is
a survey of the radiating elements that have been developed for micro-
wave arrays. The final division treats the problems associated with
combination of the elements into linear arrays and the techniques avail-
able to produce the desired amplitude and phase distributions.

PATTERN THEORY

9.2. General Array Formula.A linear array is a specialization of the

general space array discussed in Sec. 3.19. The space factor of the system
can be obtained immediately from

/ Eqs. (3.179) and (3.180) by impos-

ing on those equations the simpli-
/ fications gained in working with
a one-dimensional rather than a
(, three-dimensional complex. It
/ .., may be instructive however, to
/# . derive the space factor directly
,/ \
. from the superposition of fields;
/]
\ve shtill he concerned only with
the far-zone field of the array.
1,1<;.9.1 . Difference in distal)ve frclnl the Suppose that there are n ele-
tth elenlerlt and fro,ll the pole to a dmtallt ments in the array under consider-
point ill the direction O, c$.
ation, and let the rrfm-cnce line of
the array be taken as the polar axis. The ordering of the elements P,, P,j
. . . . P.1 is shown in Fig. 9.1 with the element PO taken at thc origin;
the distance between t\vo adj scent elements is s. I,et, us consider the field
at a point (R, 0, O) in the far zone. :~ccording to llqs. (3.168a) and
(3. 168/)) the field due to the ith element at a distance r, from t}le element is
 259

If the customary far-zone field approximations are made, r, can be set

equal to R in the denominator, while in the phase term we have
ri=Riscos O (1)

as shown in Fig, 91. The component fields are then

(2a)

(2b)

where +, is the phase difference bet}reen the ith element and the origin
due to the difference in path length to the field point:
27ris cos O
+,= ~ (3)

The elements of the array are identical in structure and carry similar
current distributions. They differ only in the amplitude and phase.
We can, therefore, write
F,,(6,0) = a,F, (O,@), (4a)
F2,(0,0) = alF2(0, f$). (4b)

The complex coefficients at express the amplitude and phase of the ith
element with respect, say, to the zeroth element; they will be called the
feeding coefficients.
By the superposition principle, the field of the array is
nl

The last two factors in each instance represent the corresponding space
factor of the array. The po\rer pattern is proportional to the sum
of the squares of the absolute value of the two space factors; that is,
.1 ,,1
2
~ici(:=i.ms8]/A
I(e,@) = F,(e,@) (Mut- + F,(@, o)
z
i=o
z
,=0
 260 LINEAR-ARRAY ANTENNAfl AND FEEDS [SEC. 92

Since the absolute value of a product is the product of the absolute values,
nl
2

or
P(e, fj) = [lF1(e, @)l + l~2(@,@)121
12 ~=o
~ie!(2A
1,
. . . .9)/

The first factor is the power pattern of an individual element of the

array. The second factor depends on the number of elements in the
array, their amplitudes and phases, and their spacing. It is formally
independent of the type of element used, although in practice the value
of the feeding coefficients at is intimately connected with the character-
istics of the elements of the array. We ,shall call this function the
array fact or and denote it by v (t?).

(7)

This factor is the power pattern of a similar array of isotropic radiators,

for which P~(O,@) = 1. Moreover, it is independent of @ as was to be
expect ed.
If the feeding coefficient a~is written as

the array factor is seen to be the square of the magnitude of the resultant
of n vectors; the magnitude of the ith element vector is [ail and the angle
bet ween it and the zerot h-element vector is X, + vi. The angles between
the vectors vary with the angular position o of the field point, with cor-
responding variation in the resultant vector. In general as 8 covers the
entire range from 0 = O to 0 = T, the magnitude of the resultant passes
through maximum and minimum values. The absolute maximum value
that could be attained by the resultant is the sum of the vectors when they
are colinear and in the same direction. With arbitrary xl, however,
there may be no angle 0 for which this condition is realized and the
i
maxima are less than the absolute maximum. Similarly, there may be
no value of 8 for which the minimum value of the resultant takes on the
absolute minimum value of zero. However, with special relations
between the x, it is possible to have directions 6 for which the path-length
phases ~i compensate for the intrinsic phase differences x; between the
elements to bring all the component field vectors in phase; in this case,
the absolute maximum resultant is attained.

I
SEC. 9.3] THE ASSOCIATED POLYNOMIAL, 261

A particularly simple and useful case is that in which the coefficients

ai are all real. Thk implies that the angle between any two vectors
associated with adjacent elements is (27rscos t?)/A.
If the coefficients u; are equal, it is readily apparent that the resultant
vector is O whenever the vectors constituting the sum permute among
themselves under a rotation of less than 27r. For then the resultant
vector both rotates and remains unchanged and hence is O. This occurs
whenever (2ms cos 0)/h is any integral multiple of 2r/n less than n.
When (27rscos 0)/X = 2T, the vectors obviously reinforce one another
and an absolute maximum results.
Whenever *(o) = O, then E, = E+ = P(r9, @) = O [Eqs. (5a), (5b),
~d (6)] for all values of O. The surface in spherical coordinates for
which o is constant is a right circular cone. The cones on which v(6) = O
are commonly called cones of silence.
9.3. The Associated PoIynoxniaL-The vector representation of the
array factor provides a method of rapidly analyzing the simple arrays
frequently encountered in practice. Vector language is not, however,
well suited to a more general study of arrays. An alternate method has
been developed that associates a polynomial with any linear array.
The array factor may be completely analyzed in terms of properties of
this polynomial.
Let z be the complex number z = z + jy. The polynomial associ-
ated with the linear array of elements having feeding coefficients a; is

f(~) = aO+ w + . . + a._lzn-l.

The value of the polynomial for the complex number

is the sum entering into Eqs. (6) and (7); the array factor is thus the
normz of the associated polynomial for z = ~,
v(e) = Ij(f) I2. (8)
The complex number ~ is a vector from the origin in the complex plane,
of magnitude unity, making an angle v = (27rscos 0)/k with the real
axis. As 8 varies z = ~ describes a circle of unit radius about the origin.
In the future we shall not distinguish between z and ~; it is to be under-
stood that z lies on the unit circle whenever V(O) is to be computed from
the associated polynomial. When 0 = O, * = 27rs/A. As 0 moves
toward r, z moves along the unit circle clockwise toward the point where
its angle ~ = 27rs/k. In that interval z may traverse but a portion of
1 S.A, Schelkunoff, A IIIathematical Theory of Linear Arrays, BeZl System Tech.
Jour.,22, 80 (1943).
z The norm of a complex number as used here is the square of its abwlute value.
It may have a more general meaning.
262 I>[~VEA R-ARRA 1- A ,V7E,V,VA,S A ,\-l) FEEI),9 [SEC<,93

the unit circle or may complete several circuits of it depending on the

value of s, Its path will be referred to as the range of z. In Fig. 9.2
the range of z is shown for three values of .s. Since the angular distance
traversed by z is 47rs/x, the range of z is exactly one circuit of the unit
circle when s = A/2, is less than one circuit when s < A/2, and is more
than one circuit when s > A/2.

s +

FIG. 9.2.The portion of the unit circle in the complex Plane that is the range of z.
The real axis is horizontal, The figures on the perimeter show tile rnrrespouding values of
O for certain values of z.

Any polynomial can be expressed as a product of linear factors. In

particular the associated polynomial may be written in the form

j(z) = an-l(z z,) (2 22) . . (z z.-J. (~)

Since the feeding coefficients give only the relative phases and amplitudes
of the elements of the array, a~.-l can be taken to be any con~enient non-
zero number. The complex numbers z, (known as the zeros of the
polynomial) are unaffected. Their values depend only w the set of
ratios a,/am_l. The factorization of j(z) in Eq. (9) lends itself to a simple
geometric interpretation of the array factor. Since the norm of a product
is the product of the norms, Eq. (9) may be written

v(o) = ]2 2,{212 2,12 . . . Iz z,_1[2

for z on the unit circle. The zeros of ~(z) are \rell-defined points in the
complex plane but do not necessarily lie on the unit circle. For any value
of z, Iz z,Iz is the square of the distance between the point z and the
point z,. The array factor is then the square of the product of the dis-
tances of n 1 fixed points to a variable point moving on the unit circle.
It is immediately obvious that v(6) = O if and only if some Z, lies on the
unit circle within the prescribed range of z. Shown in Fig. 9.3 is the range
of z when s = k/4. The zeros of j(z) are shonm for the case n = 9 and
aO= al=.. . = a.l = 1. The array factor then vanishes for four
values of o and attains a maximum vaiue at three points, each lying
between an adjacent pair of nulls. The predominating influence on
the value of T(6) is the distance from the corresponding value of z to the
SEC. 9.3] THE ASSOCIA TED POLYNOMIAL 263

nearest zero of j(z). In this connection it should be noted that the

zeros of j(z) lying outside the range of z have for the most part but small
effect on the relative value of ~(d). For the case illustrated, the point
z = 1 (corresponding to o = m/2)
o
is farthest from a zero off(z), and
one would expect, as is the case,
that v(0) has an absolute maxi-
mum at d = 7r/2. This is directly
evident from the vector view-
point of the previous section. The
elements themselves are al1 in
phase, and in the direction e = r/2
the distance is the same from all
the elements to the field point.
The contributions from the ele-
ments in that direction are, there-
u
fore, in phase, and the vectors are FIG. 9.3.The location of the zeros of
all in the same direction. ~(z)=ltz+z+.. +Zn. The range
of z for s = A/4 is also shown together with a
It is frequently advantageous
few of the corresponding values of 0,
to separate from the relative phase
xi of each element as expressed in the coefficients a, a constant-phase
delay ~, between each pair of adjacent elements. This is equivalent to
writing
~i = ~iel(%)

where now the angle of d, is the deviation from i~~ of the difference in
phase between the ith element and the element with index O. Let
~ = ~>+o~.
J (lo)

then j(z) transforms into the polynomial

%1
j(z) = ~ titz. (11)
,=0

When z lies on the unit circle, i?given by Eq. (10) does likewise, and we
have j(z) = ](2). Since Eq. (10) is equivalent to a rotation of the com-
plex plane through an angle +0 in the clockwise direction, the array factor
may be computed from the zeros of j(~) in the same manner as before
save that the range of z is the original range of z rotated clockwise through
the angle ~o (Fig. 9.4). Symbolically

v(e) = IJ(Z)],
where
~ = ~al(z~$
cm~)/Aw.
264 LINEAR-ARRAY ANTENNAS AND FEEDS [SEC. 9.4

The association of a polynomial with any linear array of prescribed

spacing provides a simple and elegant method for compounding array
factors. Suppose v,(o) and *Z(0) are the respective array factors of two
arrays with the same spacing.
If ~,(z) and $2(z) are the asso-
ciated polynomials of the arrays,
then the array whose associated
polynomial is
f(z) = j,(z)j,(z)
z
will have as its array factor

w(o) = V,(e) v,(e), (12)

for, as has already been observed,

0.67?r the norm of a product is the prod-
uct of the norms. Explicit
values of the feeding coefficients
O.!xm
FIG. 9,4.The range of z due to the con- of an array whose array factor is
stant Phase delay #o = m/2 for a = X/4. given by Eq. (12) can thus be ob-
The dotted curve indicates the range of z
associated with that of 2. The figures on the tained by simply multiplying to-
perimeter indicate the corresponding values gether the polynomials j,(z) and
of .9for certainvaluesof z and i.
f,(z).
9.4. Uniform Arrays.-A linear array that is made up of elements
having equal amplitudes and a constant-phase difference between adja-
cent ones is of considerable importance. Such an array is called a uni-
form array. Its feeding coefficients are

Although the associated polynomial of a uniform array will in general

have complex coefficients, the related polynomial .f(~) may, as was shown
in the previous section, be used with equal effactiveness. Then

The array factor is then

with

ti=ycoso-+o. (13)
SEc.9.4] UNIFORM ARRAYS 265

However, Ie*1= 1 and lei~ e-i$l = 4 sinz T$. Thus the array factor
of a uniform array is given by

(14)

together with Eq. (13). The number nz has been inserted in the denomi-
nator as a normalizing factor.
When ~ = 2kr and k = O, +1, t 2, . . . , v(8) is indeterminate.
It can be readily shown, however, that it approaches the value unity at
those points. The corresponding values of e are given by

Cos e = ;8 (2k7r + *O) (k=o, *l, *2,...). (15)

For every real value of 0 satisfying Eq. (15) V(O) has an absolute maxi-
mum. In these directions the differences in phase between the vector
contributions of successive elements that are due to differences in path
to the field point just compensate for the intrinsic phase difference between
the elements. The contributions are then all in phase, and we have,
therefore, an absolute maximum equal to the sum of the lengths of then
vectors. For values of s < ~/2, however, there will be values of *O for
which Eq. (15) has no real solution.
Since V(O) is never negative, its absolute minima will occur when
V(O) = O, that is, for any value of 0 satisfying

cOse=M+++r)
=0+1+2) 1)
other than those satisfying Eq. (15); for at those points the numerator
in Eq. (14) vanishes while the denominator does not. The points .9 = O
and o = m may also be minimum points. Certainly W(d) is an extremum
at each of these values because it has the period 27r and is symmetrical
with respect to the line o = O.
No other minima of V(O) exist. 1 The maxima, other than those given
by Eq. (15), will occur close to the point where the numerator in Eq. (14)
reaches its maximum value of unity; for the numerator is changing much

1Differentiate v (0) with respect to 0.

The points at which sin n~/2, and sin 8 vanish have already been examined. The
only other critical values can arise from the factor

*sin(n+)-n+in(+)
266 LINEAR-ARRAY AN1ENNALS AND FEEDS [SEC.95

more rapidly than the denominator. An excellent approximation then

for the remaining maximum points of ~(0) is

(2k+l)=+*o
cose=~ (I C =+1, *2,-..)
2TS [ n 1
in which values of k divisible by n are excluded.
Figure 9.5 shows w as a function of + [Eq. (14)] for n = 12. For thk
functional dependence v has the period 21rand is symmetrical with respect
o

0 r Zlr
J

FIG.5.The unction
(nZ:*)
(n, Sin, ++)
for n = 12.

to the line # = O. An idea of the shape of the array factor for various
values of s/h and $0 may be obtained from the graph. Because
2TS Cos o
+ . h +0) (9.13)

This function, however, is monotone in any region between adjacent solutions of

Eq. (16) because its derivative is
rl T-l.n
sm ~ # sin ~ ~.
2
Thus it can vanish no more than once in the region whew end points are successive
roots of Eq. (16).
SEC. 9.5] BROADSIDE BEAMS 267

the portion of V that represents the array factor lies in the region for
which

Thus the values of ~ that determine the array factor extend over an inter-
val of length 4rs/A; this may be less than the period of V or several of its
periods.. In the language of Sec. 9.3 the range of z (whose angle is +)
may be less than one circuit of the unit circle or several circuits.
9.6. Broadside Beams.A linear array whose form factor has its
absolute maxima only in directions normal to the axis of the array is
known as a broadside array. The array factor of such an array should
then have a single absolute maximum in the direction 0 = 7/2. The
power pattern of the radiating element employed will, of course, deter-
mine whether, among other possibilities, the array will have a single
direction of maximum intensity in the plane 0 = 7r/2 or its intensity will
be maximum in every direction in that plane. Both of these types of
arrays have widespread application; the latter is sometimes called an
omnidirectional antenna. 1 In the microwave region the principal use
of these antennas is as beacons, and in the following sections all such
antennas will be referred to as beacons.
We have seen that the array factor for a uniform array has absolute
maxima for the values 0 satisfying the relation [Eq. (15)]

This will have the solution 0 = m/2 if *O = O, and it will be the only such
solution if s < k. Arrays in which the elements all have the same ampli-
tude and phase (4o = 0) are commonly referred to as uniformly illumi-
nated arrays. For the moment let attention be restricted to the case
s = A/2. It will be shown later that this restriction is desirable. The
array factor is then, from Eq. (14),

w(e) = (
sin~ ~T cos .9
2 ). (17)
n2 sin2 ~ Cos e
() 2
Equation (17) is plotted in Fig. 9.6 on a decibel scale for n = 6 and n = 12.
It will be observed that the side lobes (secondary maxima) on either
side of the main beam decrease. Moreover, on the decibel scale used in
Fig. 9.6, a straight line joining any two peaks on the same side of the main
beam lies entirely above any intervening peak. That this is always true
I It should be remembered, however, should this usage be encountered, that
omnidirectional means all directions in a plane.
268 LINEAR-ARRAY ANTENNAS AND FEEDS [SEC. 95

may be verified by noting that the peaks of the side lobes lie approxi-
mately on the curve
1
*(l =
r
n2 sin2 cos e
()2
The second derivative of in ~, with respect to e is

~[icsc26os)sin2+ co6c0so)c0sel

5
db
10
1

15

20

25

30

35

40
010 30 50 70 90 110 130 150 :76
e

FIQ. [G )1 for n = 6 (full curve) and n = 12 (dotted

+raph f )1
[nsh(ices
curve).

This is positive in the interval on either side of the main beam; hence the
peaks of the lobes on each side of the main beam lie on curves that are
concave upward.
Direct computation shows that the height of the first side lobe, that
is, the one nearest the main beam, varies from 0.056 for n = 6 to 0.047
for n = 12 and 0.045 for all sufficiently large n. The height of the last
side lobe is 1/n2 for odd n and approximateely that for even n.
If the sine appearing in the denominator of Eq. (17) is replaced by its
SEC. 95] BROADSIDE BEAMS 269

argument, an approximation for the half-power width of the array factor

may be obtained namely,
101.8.
@=
n

An indication of the magnitude of the error is contained in Table 91 in

which both the actual half-power widths and those computed by the
approximate formula are given.

TABLE9.1 .MAGNITUDEOFERRORRESULTINGFROMTHEUSE OFTHEAFPROXI~ATION

FOR THE HALF-POWERWIDTH

n 2 3 4 5 6 12 50

101,8
50.9 33.9 25.4 20.4 16.97 8.48 2,036
; 60.0 36.3 26.3 20.8 17.19 8,.50 2.039

The relatively large height of the first side lobe is characteristic of a

uniform array and at times may be annoying. Broadside arrays may
readily be formulated, at least in theory, that have side lobes as small as
desired. For example, consider an array whose associated polynomial is

j(z) =(l+z+z + +z~-)

or
f(z) = 1 + 22 + . 0 . + red- + . . . + 2z~- + 2-2.

The elements are all in phase, but their amplitudes decrease uniformly
from the central element. This is a special case of what is commonly
called a gabled illumination. Its array factor is the square of the array
factor of a uniform array; hence its first side lobe will have a height of
but ~ per cent of the height of the main beam instead of the 5 per cent
height of the uniform array. All of the other side lobes will be reduced
in a similar fashion, but the main beam will be somewhat broader than
the main beam of a uniform array with the same number of elements.
The half-power widths of the gabled and uniform arrays are approxi-
mately 1460/n and 1020/n, respectively, where n is the number of ele-
ments. Successively higher powers of the polynomial may be computed;
the reduction inside lobes is accompanied by a rapid growth in beam width.
A general discussion of the problem of constructing high-gain broad-
side arrays with side lobes below a prescribed value will be given in
Sec. 9.7. Attention may be called here to an array that eliminates side
lobes completely. The feeding coefficients are equal to the binomial
coefficients
~!
c,,, = (18)
k~r k)!
270 LINEAR-ARRAY ANTENNAS AND FEEDS [SEC. 9.5

The array is derived from the two-element half-wavelength-spaced uni-

form array. The latter has an array factor

v(e) = COS2 ; Cos e ;

()
o

db lo

20

30

40
10 30 50 70 90 110 130 150 170
e
l;l[s. 9~.--Array factorz for three 1 l-.le,TIent A/2-spacecf arrays: (a) & uniform ~rray

this has no side lobes and has nulls at 0 = O and O = r. Its associated
polynomial is
f(z) = 1 + z.

From Eq. (12) it follmvs then that the array JV1](JSC

polynomial is

j(z) = (1 + 2) = c.,, + Cr,,z + C,, *2 + + c,,#

has an array factor

z = COS2 ~ Cos
2 () .9

An irispection of the three space factors given in Fig. 9.7 shows that, of I
the three, the uniform broadside array concentrates the greatest per-
centage of the radiated energy in the direction normal to the array. It I
can be readily shown that of all arrays in which the elements have the
h
same phase and the spacing is ~ the uniformly illuminated array has I
the maximum gain. Let
I

I
SEC. 95] BEOADSJDE BEA. IIS 271

j(z)
7L1

z
,=0
a,

be the polynomial associated with an arbitrary array of n elements,

normalized so that j(1) = 1.
Since the elements are spaced a half \,-avelengt haPart,. = e~~oos.
The gain G in the direction .9 = 7r/2 (which corresponds to z = 1) is
given by

=~2r~-lf(z~sin,d@d@ =ecme
or

G=;, (19)
where

(20)

Tomaximize G, I must be minimized, In Sec. 9.7itis shown that there

is no loss in generality by assuming that f(z) has real coefficients. If we
let # = cos d, then Eq. (20) becomes, after expansion,

where the numbers C, are combinations of the polynomial coefficients G,

Hence
,11
2~a~
1= ~:;o , (21)
2
a,
(y,=0
k )
and
nl nl ,?1

131 a(~a,)-za,za:
,=0 ,=0 ,=0
=4 ,1I (/%=0,1, . . ..? 2-1).
dak
Ya,
(u,=0 )

\Ve then ha~-e for a minimum the system of equations

272 LINEAR-ARRAY ANTENNAS AND FEEDS [SEC. 9.5
nl nl

(IJc=o,l, . . ..1)l). (22)

akxa-2a=0 ,=0
~=o

The difference between any two of Eqs. (22) is

*1

(a, G) a, = O.
z
,=0

Thus a~ = ai for all i and k. That these conditions actually yield a

minimum may be shown by examining the second derivatives. Equa-
tions (19) and (21) show that the gain of such a uniform array is n,
the number of elements.
The situation becomes more complicated if the restriction to half-
wavelength spacing is removed. We shall attempt an answer only for
broadside arrays having a total length (the distance between the first and
last elements) of k/2. In that event, the half-wavelength-spaced unif orm
array has the largest gain of all uniform arrays of the prescribed length.
However, the uniform array does nol yield maximum gain of all arrays oj a
given length.
The array factor of an n-element uniform broadside array ~/2 long is,
from Eq. (14),

*=
sinz
n
n12( ~ Cos 6

1 TCOS6
)

n2 sinz
(n1~ )

When n = 2, we have

V@=cos
();Cose.

The two-element array will have maximum gain if

2 ()
T
Cos -- Cos e s
ln ( ~ )
n
n12
Cos

1 : Cos-o
e

n sin
( n12 )

in the interval O ~ 8 < T/2. This inequality, however, is equivalent to

sin
( E
n12
~ Cose
)( < 1
n2~
-
n12
cos e
. )

n+icose - :+%
SEC. 95] BROADSIDE BEAMS 273

Because (sin z)/xisa decreasing function inthefirst twoquadrants, the

inequality is established and the two-element array has the largest gain
of all unijorm arrays whose total length is A/2.
To show that the uniform broadside array has not, however, the
largest gain of all broadside arrays of the same length, we shall consider
as a specific example the maximum gain of three-element quarter-wave-
length-spaced arrays. Let

be the polynomial associated with any such array. As before, a, b, and

c are assumed to be real. Then from Eq, (20)

i (a + b + c) + 2(ab + bc) cos ~ + 2ac cos 2+ ~i

I=!
To/ (a+ b+ c)
or

: (aj + b + c) + 2(ab + bc)

~=~2
7r (a+ b+ c)

The equations for minimizing 1 are

(a+ b+c)2(ra +2b)2(a+b+c)lV= O,

(a+ b+c)2(rb +2a+2c)2(a+b+c)N =0,
(a+ b+c)2(~c +2b)2(a+b+c)iV= O,
where

N = ~ (a + b + c) + 2(ab + be).

Thus
a=c,
T-l
b= a = 0.7519a.
T2

Hence G = 2.4 as contrasted to G = 2.0 for the two-element half-wave-

length-spaced uniform array. In Fig, 9.8 are drawn the array factors
of the uniform array and the array whose gain was just computed. Also
shown is the array factor for the uniform continuous array, that is, one
in which n has been allowed to increase without limit, subject only to the
restriction (n 1)s = k/2.
In practice it is frequently desirable to avoid half-wavelength spacing
because of the resonance that may occur at that spacing. As far as gain
is concerned this is quite feasible; for the gain of a uniform array suffi-
ciently long is nearly independent of the number of elements, provided
only that the spacing does not greatly exceed a half wavelength. How-
ever, it is only at resonance that the requirement that the radiating ele-
274 LIAEAR-ARRA Y ANTENNAS A,VJ9 FEEDS [SEC.96

ments be in phase can be readily met.1 N-onetheless, if the spacing does

not differ by much from a half \\-avelength, the progressive phase delay
thereby introduced [xOin Eq. (13)] causes but a small deflection of the main
beam from the normal to the array; the exact amount is given by
~= sin-l k!?. (23)
2TS

o
(c) 11

e
FIG. 9S.-Array fart,,rs of four broadside arra>-s whose lengths are i/2: (a) the two-
element A/2-spaced uniform array cosz ? cos O ; (b) the four-element k/6-spaced uni-
()
sins ~ cos O
n(%) 2
form array ; (c) the mntinuous uniform array ~; and (d) the
16 sinz ; Cos e ; Cos e
() ()
+,hree-dement h/4-spaced array with maximum gain.

9.6. End-fire Beams.The feeding coefficients of a linear array may

be chosen so that the array factor has an absolute maximum along the
axis of the array. If an element of the array produces a pattern having
an absolute maximum in the same direction and if the product of tl-~e
array factor and the elcmcmt pattern has no other absolute maximum,
the array is called an end-fire array.
If the elements of the array are not directive, the radiation pattern
of the array is determined entirely by the arr~y factor T(0). The pattern
is the surface in spherical coordinates given by

r ===v(f3)

and is therefore a surface of revolution symmetric with respect to the

axis of the array. It is only b~. considering the three-dimensional picture
that the great difference bct\veen end-fire and broadside arrays becomes
apparent. The major lobe of an end-fire array is a pencil beam; thus a

LSee Sec. 9,17.

SEC,96] EAD-FIRE BEAMS 275

on~dimensional configuration of sources produces radiation directive in

two planes and does so without relying on any directivity of the individual
sources. A broadside array on the other hand is directive in only one
plane; it is omnidirectional in the plane perpendicular to the axis of the
array.
Pencil beams whose half-power widths are in the region from about
15 to 35 can be produced quite readily by end-fire arrays that have
lengths ranging from 3 to 18 wavelengths. The length of the array,
however, varies inversely with the square of the beamwidth; narrow
beams would require very long arrays. By properly choosing the feeding
coefficients, end-fire arrays can be designed whose gains are almost double
those of broadside arrays with the same length. This increase in gain
has the greatest practical significance for arrays about 5 wavelengths
long.
In order to eliminate from the array factor large lobes in any direction
except 0 = 0, it is necessary to restrict the spacing of the elements. The
necessary relations can be obtained
o.33fr
from a study of the associated poly-
nomial and an examination of the
range of z on the unit circle in the
complex plane. Since the angle ~ 02-s,,,
\\
of z is given by

~ assumes all values from 27rs/h

\ 0.5(q I o.5olr

to 2ms/A as 0 ranges from O to m.

Because the array factor has a
period 27r as a function of ~, the
~..
-..= 0.671r
spacing s/x must be such that FIG. 9.9.The range of $, the angle of z,
over the range of 0 the total varia- as .9 varies from O to n. The outer ring of
figures shows corres~ondine values of o at
tion of ~ is less than 27r; that is, various points on the unit cirrle in the Z-plarle
the range of z in the unit circle for s = ),/2. The inner set of figures S11OWW
values of 9 for s = ),/4.
is less than one revolution. In
that case a principal maximum, which in the case of an end-fire occurs
at ~ = 27rs/A, will not be repeated. In Fig. 9.9 the mapping of o on
x is shown schematically. On the exterior of the unit circle in the
z-plane are shown values of O, corresponding to the indicated points on
the circle for the spacing s = A/2. The values shown on the interior
are for the spacing s = A/4. The array factor of a half-wavelength-
spaced end-fire array will duplicate its value for 0 = O again at 9 = m.
To suppress such an undesirable back lobe it is necessary to separate the
end points of the range of z. The quarter-wavelength-spaced array wili
be examined as a typical array that satisfies this condition. Figure 99
276 LINEAR-ARRA S ANTENNAS AND FEEDS [SEC. 9.6

shows that the range of ~ for the quarter-wavelength spacing of elements

is from 7r/2 to 7r/2. From the discussion in the previous sections it is
seen that if such an array has equal feeding coefficients, it will have its
principal maximum at o = 7r/2, that is, at ~ = O.
We have seen, however (Sec. 93, Fig. 9.4), that if a given array is
altered by havir,g superimposed on it a constant-phase difference from
element to element, the effect on the array factor is to rotate the range
of z through an angle equal to that phase difference. If, in particular,
a quarter-wavelength-spaced uniform array is adjusted to introduce

a phase difference of ~ between each pair of adjacent elements, the

principal maximum that occurs when

$ = O corresponds to 0 = O, and an end-
fire antenna results (Fig. 9.10). The
array factor of such an array is obtained
immediately from Eq. (14) and is

sin2 ~ (COS0 1)
v(o) =
nz sin2 ~ (cos 0 1)

The factor n has been inserted in the

0.507 denominator so that v(O) = 1.
FIG. 910.-The effect of intro-
ducing a constant-phase difference The gain is more easily computed
T/2 on the elements of a quarter- from the polynomial associated with the
wavelength-spaced uniform array.
The inner semicircle shows the array
original range of z and the correspond-
ingvaluesof 8. Theouter sen,icircle
is the range of z due to the phase dif -
ference; the corresponding values of
o are indicated.

Equation (20) becomes, after substituting and making the change of

variable ~ = T(cos 0 - 1)/2,
nl

=:/:. [n+zI(n-k)cos+l+
k=l

Thus from Eq. (19) G = n, the same gain as the longer half-\vavelengtk-
spaced uniform broadside array.
A uniform array with constant-phase difference between adjacent
elements is one readily realized in practice. It is then well to inquire
if the choice 7r/2 for the phase difference is optimum for an end-fire
array. If this difference is slightly less than 7r/2, then the range of z
is displaced slightly more than 7r/2. The direction 6 = O no longer repre-
sents the principal maximum of
SEC. 9.6] END-FIRE BEAMS 277

~i*~ ?+
2
(24)
+
sin=
2

but is displaced slightly from it (Fig. 9.11). However, the principal

maximum of Expression (24) is not contained in the range of z. The
end point of that range corresponds to 0 = O which then is a maximum
of Expression (24) considered as a function of 0. The net effect is two-

2.
FIQ. 9.1 I.The portion (cross-hatched) of the fiinction included in the range of
25 sinz ~
z due to a displacement slightly more than u/2.

fold: Because the value of Expression (24) at 0 = O has been reduced,

the relative heights of the side lobes are increased and the gain tends to
be reduced. On the other hand, the width of the main beam has been
diminished, which has an opposite influence on the gain.
An estimate of the displacement yielding maximum gain maybe made
by approximate methods valid for large n. Suppose the phase difference
between any two elements is (7r/2) ~,; that is,

+=; (COSOl) *0.

Then the array factor is

~in* ~ sin, n7r

~(cosel)y
v(e) = J [ 1.
n#O
si~2 _ sin2 T .$
2 4 (Cos e 1)
[ 1

The first factor has been inserted so that *(O) = i. An approximation

of i;he gain may be obtained. If y = (n/2)#, Eq. (20) becomes
278 LINEAR-ARRAY AATENNAS A,YD FEEDS [SEC.96

If n is large, the distant side lobes have little effect on the gain. We
may replace sinz y/r~ by its argument, thus reducing the height of the
distant side lobes. The range of integration may then be extended to
cc and sinz YO/2 may be replaced by its argument with but a negligible
effect on the value of 1. Then Eq. (25) becomes

By graphical methods~ it has been shown that 1 is a minimum when

$0 = 2.9~/n. Then the direction 6 = O corresponds to the point where
the function given in Expression (24) is 46 per cent of its maximum value.
The main beam is about half as broad as that of a uniform array with a
phase shift of 7r/2. On the other hand the heights of the side lobes
have been more than doubled. The gain of such an array is

G = 1.82n.

Its half-power width is approximated by

~ _ 125
fin

For a more general spacing s between adjacent elements, maximum

gain occurs when
*O = 242.

Here ~, still refers to the additional displacement of the range of z beyond

rr/2. The phase difference bet\reen adjacent elements is 2ms/A *O.
The gain for the general case is

However, s is not completely arbitrary. We still must conform to our

assumption that the distant side lobes have small effect on the gain. It
has been suggested that the approximations are valid for s < A/3.

I W, W. Hansen and J. R. Woodyard, A New Principle in Directional .4ntenn~,

Design, Proc. ZRI?, 26, 333 (1938).
2 Hansen and Woodyard, op, cit.
SEC. 97] BEAM SYNTHESIS 279

An entirely different technique for increasing the gain of an end-fire

array has been given. 1 .4gain we start with a quarter-wavelength-spaced
uniform array with a phase delay 7r,/2 between adjacent elements.
The polynomial associated with such an array is
nl n1
j(z) = (jz)k = (_j).-1 (z jti) (26)
rI
z
k=o k=l

where u is the nth root of unity with the smallest positive angle. The
array factor is then
n1

v(e) = ~ \z ju1.
k=l

The numbers .ju~ lie on the unit circle, and, it will be recalled, the array
factor is formed by compubing the square of the product of the distances
from these numbers to the variable point z.
It is apparent (Fig. 9.12) that the zeros of
j(z) lying outside the range of z add little
to the directivity of the array. Thus, an
array whose polynomial is
nl
f(z) = ~ (z ju),
k.[~]

retaining as it does only those zeros which

lie on the range of z, will have the same FIC. 9.12.The location O,
nulls as formerly and at the same location. the zerosof Eq. (26)withrelatlon
Its Win will have been reduced but ~itt]e, 0 he nge Ofz for n = S.
while its number of elements has been almost halved.
9.7. Beam Synthesis.-The preceding sections have dealt principally
with the problem of analyzing the properties of the array factors of given
linear arrays. The inverse problem, that of finding an array which will
yield an array factor having prescribed characteristics, is far more diffi-
cult, The prt=sentsection \villtreat some aspects of the synthesis problem.
The nature of the synthesis problem depends on the manner in which
the desired pattern is specified. The latter may be prescribed as a
complete function of d over the physical range O s o s T. In general a
solution is sought that gives an acceptable approximation to the desired
pattern. There can be no unique solution to such a problem, since the
pattern is prescribed only as regards intensity distribution of the radia-
1S. A, Schelkunoff, A hlathematical Theory of Linear Arrays, Bell Systenl
Tech. Jour., 22, 80 (1943).
280 LINEAR-ARRAY ANTENNAS AND FEEDS [SEC. 9.7

tion field; the phase distribution is arbitrary, and each choice of such a
distribution will lead to a different array. Only a partial solution to the
problem will be given. It will be based on the general characterization
of the array factor of an n-element array and the formulation of the prop
erties that a pattern must possess in order to be the array factor of a
linear array. All n-element arrays will be found that have a given array
factor. The problem of finding the best approximation to a given pattern
by a realizable array factor is beyond the scope of the present discussion.
The desired pattern may be specified with regard to general proper-
ties rather than as a complete function of 9. Examples of such synthesis
problems are the design of a broadside array having minimum beamwidth
for a given side-lobe level and the design of one having a minimum side-
lobe level for a given beamwidth. These problems have exact solutions
when the spacing of the elements of the array is A/2 or greater; they will
be discussed later in this section.
We shall consider first the characterization of the arrav factor of an
n-element array. The array factor can be obtained from Eq. (8) by
replacing ~ by ei(z- )X and expanding. If the real numbers A k and Bk
are defined by
nlk

Eq. (8) becomes

/ik +j& =
z
,=0
ara~+k, (27)

nl

*@)=
Ao+2z[Ac+:cOs) k=l
+Bksin(k?c@)l 28)

Thus the array factor of any n-element linear array with spacing s is a
trigonometric sum of order n 1 in the angle $ = (2%s cos 0)/k. The
trigonometric sum is nonnegative for all real values of *. Conversely,
every nonnegative trigonometric sum can be realized as the array factor
of a linear array. It follows then that the necessary and sufficient con-
dition that there exists a linear array having the prescribed pattern as its
array factor is that the prescribed pattern can be expressed as a nonnega-
tive trigonometric sum of a finite number of terms. Expressing the
prescribed pattern as such a sum determines the coefficients Ak and B~,
and in principle the feeding coefficients a, of the array can be determined
from Eq. (27).
To find an n-element array that will approximate the prescribed
pattern, the latter may be approximated Lby the terms of order less than
I The method of approximation selected will depend on how the prescribed pattern
is specified and what deviation from it is acceptable. For a general discussion of this
problem see C. de la Vallee Poussin, Lecona 8UTt A pprommation da Fwwtiona d unz
Variable reLle, Paris, 1919.
SEC. 9.7] BEAM SYNTHESIS 281

n in its Fourier series expansion in the angle $ = (%s cos 6)/x. When
s > x/2, the periodicity of-the Fourier series may present difficulty. If
these terms form a trigonometric sum that is nonnegative for all ~, the
coefficients A ~ and Bk may be used to determine the feeding coefficients
of the n-element array.
A direct solution of Eq, (27) is, however, difficult, Instead, define
an auxiliary polynomial F(z) by
nl nl
F(z) = (A, jBk)z-+ + AOZ-l + (Ak + jB~)Zn-l-. (29)
z z
k=l k=l

Then Eq. (28) becomes

We) = l~(e(r-)h)l. (30)

If Zkis a zero of F(z), so also is its conjugate reciprocal l/z$. The assump-
tion that the trigonometric sum is nonnegative thus implies that the zeros
lying on the unit circle, which are their own conjugate reciprocals, occur
with even multiplicities. Hence the zeros of F(z) may be grouped in
pairs; and aside from a constant multiplier,
n1

F(z) = ~ [(Z zk) 2>. : (31)

( )1
k=l

One zero in each pair may be selected as a zero of a new polynomial

n1
f(z) = ~ (z z,). (32)
k-l

For values of z on the unit circle

W)l = l~(z)l = v(e),

where again a constant multiplier has been dropped. Eauation (30)

.,
then implies that W(O) is the a~ray factor of the a~~ay whose associated
polynomial is given by Eq. (32). The separation of the zeros of l(z)
into two sets can in general be done in many ways. Each such partition
will usutilly lead to two different arrays; all arrays will have the same
array factor. When all the zeros lie on the unit circle (as, for example,
in the uniform array), only one method of division is possible and the
two sets obtained are the same. It should not be assumed that finding
an array having a given array factor is an easy computational problem,
even when n is as small as 5. It is necessary to find the zeros of a poly-
nomi al of degree 2n 2 and then perform the multiplications indicated
282 LINEAR-ARRAY ANTENNAS AND FEEDS [SEC. 97

in Eq. (32) to find the feeding coefficients. Simpler but less general
methods have been devised. 1
It is now possible to verify the assumption made in Sec. 9.5 that as
far as the gain of broadside arrays is concerned, attention maybe restricted
to arrays whose elements are either in phase or out of phase by 180,
that is, to arrays whose feeding coefficients are real. The array factor
of an arbitrary array is given by Eq. (28). If the sine terms are dropped,
the resulting V(O) is still the array factor of some array. Moreover the
gain in the direction 0 = rr/2 is unchanged, since both the field intensity
in the direction 8 = 7r/2 and the integral in Eq. (20) are unchanged by ~
eliminating the sine terms. The corresponding polynomial F(z) in Eq.
(29) will have real coefficients, and its nonreal zeros occur in conjugate ~
pairs. Hence in forming the associated polynomial f(z), the pairing of
conjugate zeros may be maintained and j(z) will have real coefficients.
Let us consider next the problem of minimizing the side-lobe level of
broadside arrays with a fixed beamwidth or maximizing the beamwidth
for a given side-lobe level. The problem has received an exact solution?
when the spacing between elements is at least A/2 and sufficiently less
than A to eliminate any large end-fire lobe. For the present purposes
a convenient definition of beamwidth is the angular difference between the .
position of the two nulls enclosing the main beam. Only those arrays
will be considered whose main-beam nulls are symmetrically located with
..
respect to the direction 0 = 7r/2.
The array factor having either the minimum beamwidth or the lowest
side-lobe level may be expressed in terms of the Tchebyscheff polynomial
Z,.(Z) = cos (2n cos- z). This polynomial falls between 1 and + 1
in the interval 1 S z s 1, assumes the value + 1 at the end points of
the interval, increases steadily outside the interval, and is symmetric
with respect to the line z = O. The actual array factor is given by

rs Cos o
v(e) = ; [1 + T,. (az)l, z = Cos (33)
() A

Figure 9.13 is a graph of ~[1 + T,n(ax)] for n = 4 and a = 1. In Eq.

(33) the direction 8 = r/2 corresponds to z = 1. If

~[1 + Zzm(a)] = r, (34)

the relative height of each side lobe is l/r. The array factor with this

1S. A. Schelkunoff, A Mathematical Theory of Linear Array s, Bell System Tech.

Jour., 22, 80 (1943); and Irving Wolff, Determination of the Radiating System
Which Will Produce a Specified Directional Char~cteristic, Proc. IRE, 26, 630 (1937).
C. L. Dolph, A fhrrent Distribution for llroadside Arrays JYhich Optimizes the
Relationship between Beam Width and Side I,obe I,evel, Proc. IRE, 34, 3.35 (1946).
The results of Dolph have been generalized by Henry J. Riblet, Proc. IRE, S6, 489
(1947).
 SEC. 9.7] BEAM SYNTHESIS 283

side-lobe level and having the smallest beamwidth is given by Eq. (33)
with a a solution of Eq. (34). The null nearest z = 1 occurs for

1
(35)
= i Cos &

This, together with z = cos [(7s cos O)/A], gives the beamwidth. If the
side-lobe level is to be minimized for a prescribed beam width, Eq. (35)
is used to determine a and Eq. (34) to find the height of the side lobe.

-1.1-1.0-0.9-0.8-0.7 -0.6 -0.5-0.4-0.3-0.2 -0.1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1
FIG. 9.13.The function ~[1 + ZS(Z)].

The substitution x = +(zM + z-~$) transforms ~[1 + Tjn(az)] into

2W(Z), where F(z) is a polynomial of the form of Eq. (29) with Bk = O
for all k. The symmetry of Z,.(az) ensures that the fractional powers
of z in F(z) are missing. The same substitution transforms

=cs[(=sTe)l

into z = e~fz-m Ja. Hence this substitution transforms Eq. (33) into
Eq. (30), and thus Eq. (33) represents an array factor of some linear
array. The feeding coefficients are obtained most easily from the zeros
of ~[1 + TZ~(az)], for these transform into the zeros off(z), the associated
polynomial of the array.
The optimum properties of the Tchebyscheff array are readily estab-
lished. An argument similar to one used earlier in this section is sufficient
to show that attention may be restricted to arrays whose associated
polynomials have real coefficients. The array factor can be represented
> in the form of Eq. (30). The polynomial F(z) defined in Eq. (29) has
only real coefficients. Hence the substitution x = ~(z$fi+ .z-~~) trans-
forms Znl(z) into a polynomial G(x) symmetric with respect to the line
z = O. Equation (30) is transformed into
284 LINEAR-ARRAY ANTENNAS AND FEEDS [SEC. 98

w(e) =G(z), x = Cos

()A
7rsCos e

Suppose G(z) isnormalized sothat G(l) =rwithrasin Eq. (34). Itis

impossible for G(z) to have a zero for z > xl [Eq. (35)] and at the same
time lie between O and 1 for O s z S zl. Any such polynomial would
then have (n + 1) points in common with ~[1 + TZ.(z)], double points
being counted as such. The symmetry then shows that the two poly-
nomials, each of degree 2n, have 2n + 2 points of intersection and so
must coincide. If s a k/2, there are real values of o corresponding to
any x in the interval O S x ~ 1. Hence if the side-lobe level of the
array is I/r, the requirement that O s G(z) s 1 for O S z S xl must be
met, and the only array possible is the Tchebyscheff array.

RADIATING ELEMENTS
9.8. Dipole Radiators.-The various forms of coaxial line-fed dipoles
discussed in Chap. 8 can be adapted for use as a linear-array element to
be mounted on either coaxial line or waveguide. Design and perform-
ance are discussed here in terms of a rectangular guide; however, the
fundamental ideas apply to all types of lines. The general properties ,
desired of a dipole element are (1) a balanced excitation of the wings to
give a symmetrical pattern, (2) a resistive load presented by the dipole
because a reactive component means large reflections in the line, (3) an
easily adjustable resistance with minimum frequency dependence, and
(4) high power capacity.
The requirement for balanced excitation of the wings favors the use
of the slot-fed dipole (cj. Sec. 8.4). The open-end termination has been
used almost exclusively; the stub-terminated units are more frequency
sensitive and are also limited in power capacity by the standing waves
in the stub section. The general arrangement of a slot.fed dipole adapted
to a rectangular guide is illustrated in Fig. 9.14. The inner conductor 1
of the coaxial line serves as a coupling probe to the waveguide; it is evi-
dent that the probe should be parallel to the electric field in the guide for
efficient coupling.
The important parameters of the dipole are slot depth, wing length,
and outer-conductor diameter. The properties of the element are com- 1
plicated functions of these parameters, and little is available in the form
of systematic data. Breakdown tends to occur between the conductors
of the coaxial section. The breakdown potential can be increased by
increasing the slot width and the outer-conductor diameter; the extent
to which this can be pursued is limited, however, by the unbalancing of
I
the wing excitation. The unbalancing is due to higher modes becoming
prominent and producing an asymmetrical field across the line; the simple
mclde picture drawn in Fig. 8% is applicable only for slot widths and
 SEC. 9.81 DIPOLE RADIATORS 285

coaxial-line dimensions that suppress the higher modes. The element

illustratedin Fig. 9.14 designed for use in the 10.7-cm region has a high
power capacity. With the values of the parameters indicated in the
figure the balanced condition is maintained, as evidenced by a symmetrical
radiation pattern; furthermore, studies of the phase fronts indicat,e that
the unit has a center of feed located in the inner conductor. The wave-
guide serves as a reflector so that the unit mounted in guide forms ~!ssen-
tially a dipole-plate system.
I

P:--AA
13 ~

FIG. 9.14.CIOSSsection of a dipole on rectangular waveguide.

In the arrangement shown in the figure, the dipole behaves like a

load shunted across the line. This is proved experimentally by measur-
ing the input admittance of the dipole when it is followed by a variable
reactance, which is provided by a movable plunger in the end of the guide;
it is found that the conductance of the system is independent of the ter-
minating reactance. The admittance of the dipole is a function of probe
depth. With no probe the element presents an inductive susceptance
component; the probe, like a tuning screw, is a capacitative susceptance
? (except for extreme depths of insertion); accordingly it is possible to
find a probe depth at which the susceptance of the element as a whole
1J. Whelpton, Admittance characteristics of Some S-band Waveguide Fed
Dipoles, RL Report No. 1082, January, 1946.
 286 LINEAR-.4RRAY ANTENNAS AND FEEDS [SEC,99

vanishes. These relations areillustrated in Fig. 9.15, aplotof the dipole !-

admittance asa function of probe depth. The depth towhich the probe
may be inserted is limited by breakdown, which can occur between the
end of the probe and the bottom of the guide, This difficulty can be
ohvi~ted in some measure by terminating the probe in a small sphere.
1.0

YY
..? 01 6[ 61 01 01 .+ ,-i
go
w
.. 0.0 Conductance
., &n~onenJ($ F

~.*-
F,G. 915.~Dipole admittance as a function of probe depth in inches (X = 10,7 cm),
r
For a given depth of insertion, the sphere causes a slight increase in the
capacitative effect of the probe.
The impedances of these dipoles as single elements are practically
independent of the orientation with respect to the axis of the line. In
an assemblage of elements there are mutual interactions which are
decided functions of orientation.
For assemblages of elements the question of reproducibility of an
element in production is of considerable importance: it has been found
that characteristics can be reproduced quite accurately by centrifugal
or die-casting production methods.
9.9. Slots in Waveguide Walls.It was noted in Chap. 7 that the
electromagnetic field in the interior of a waveguide has associated with
it a distribution of current over the boundary surfaces of the guide.
:
This current sheet may be regarded properly as that required to prevent
 #
SEC. 910] TIIEORk OF .~I,Ol RAD1A7OR,S 287

penetration of the field into the region exterior to the boundaries; it is

indeed true that the metallic structure can be removed, providing the
current sheet is maintained, \vithout leakagcuf cmmgyac rossthebound-
aries. If a narrow slot is cut in the wall of a ~vavcguide such that the
long dimension of the slot runs along a current line or along the region
of the wall \vherc the current is zero, it produces only a minor perturba-
tion of the current distribution and correspondingly very little coupling
of the internal field to space. Examples of such slots are elements cut
in a coaxial line with the long dimension parallel to the axis of the line
or elements of the type c and e cut in a rectangular guide as illustrated in
Fig. 916; the slot c lying along the central line of the guide is in a region
of ?ero current density. ATonradiating slots offer a means of entry into
the guide for studying the internal field and are used for this purpose in
impedance measurements (qf. Chap. 15).
On the other hand, a slot cut in a guide wall in a direction transverse
to the current lines produces a significant perturbation of the current
sheet, with the result that the in-
ternal field is coupled to space.
A slot of this type constitutes a
radiating element. Ihe degree of
coupling depends on the current
. density intercepted by the slot and
the component of the length of the
slot transverse to tbe current lines.
Thus the coupling at a given posi-
tion on the guide can be adjusted Fr~. 9. lfi.-Slots in the wall of rectangular
by the orientation of the slot as is waveguidc.
indicated for the elements d and j
in Fig. 9.16, or the coupling can be adjusted by position like the radiating
slot b and non-radiating slot c in the figure. The type of circuit element
that the radiating slot presents to the transmission-line representation of
the wave-guide is again a function of position and orientation. Under
certain conditions the slot is in effect a shunt element; in others a series
clement; under very general conditions the slot can be represented ade-
quately only by a T- or II-section inserted in the line. The general circuit
relations and the fundamental properties of slots will be developed in the
follo\ving section.
9.10. Theory of Slot Radiators.Let us consider a cylindrical wave-
guide of arbitrary cross section with its axis the z-axis. It was found in
Chap. 7 that, the normal modes of such a guide fall into two classes:
* 7E-m,Jdw having an 11,- hut no E,-component and TM-modes having
an E.- bllt no If Z-componentj. Each mode is characterized by its char-
acterist ic admit tance Yfi~ and propa~ation constant 13~~; the latter is
real for a freely propagated mode but, is to be taken equal to ~~~~for a
288 LINEAR-ARRAY ANTENNAS AND FEEDS [SEC.910

mode beyond cutoff. From the general discussion in Sec. 73 it is seen

that the field components of a TE-mode of order a = mn can be written

H, = jH.. exp (: j~az),

E, = E., exp ( +jb~z),
H, = tHa, exp ( +jpaz), 1
where El and H~ represent the transverse electric and magnetic field vec-
(36)

tors and the upper or lower signs are taken according as the wave is
going in the positive or the negative z-direction. The general form of the
ZM-mode field components is the same as in Eq. (36) with H, replaced
by
E. = jE. exp ( T j~az). (37)

If D. is real, the functions E.,, H.., Es,, and Ho, are all real and depend
only on a, x, and y. We have also seen (cj. Sec. 7.3) that the component
vector functions E., and H~i have the orthogonality property

(Eat x H,,) . i, dS = O, a#b,

/ (38)

s., a=b, )
where & is twice the Poynting energy flux for a freely propagated mode
and i, is a unit vector in the direction Oz. The normal modes of the guide
form a complete set in terms of which an arbitrary field distribution over
the wall of the guide can be expressed in the form of a Fourier expansion.
Now consider a slot from ZI to .2, in the wall of the infinite guide.
We assume that the guide is to be excited by a known field distribution
along the slot. Then the field in the guide, which is denoted by subscript
1, will consist of outgoing waves on either side of the slot; that is, it will
contain only waves going to the right for z > ZZand only waves going to
the left for z < zI,

E,, = AOE=,exp ( ~paz), z > 22,

z
a
E,, = BaEa, exp (jp.z), z < 21,
z
a
(39)
H,, = AaHa, exp ( jf?az), z > 22,
z
a

H,, = B.H., exp (JLz), z < 21.

z
a

The amplitudes of waves going to the right and left are not necessarily
equal and are denoted by Aa and Ba respectively; they must be such
SW!. 9.10] 7711EOR 1 OtI S1.01 RADIA 7ORS 289

that on superposing the two sets of waves a field is produced which

matches the field over the slot according to the general boundary condi-
tions formulated in (lap. 3.
In order to evaluate the amplitudes A. and Ba an auxiliary relation
must first be derived. Consider two fields E,, H, and EZ, H, of the same
frequency and both satisfying the homogeneous field equations. By
virtue of these equations we find

V . (E, X HJ = V . (EZ xH, ) = jti(cE, . Ez + ~H, .HJZ

1Ience
V . [(E, xH,) (E, xH, )] = f).

If 1 is any closed rcgi{m bolludml I),ya sllrfare ,S, it fullmvs by the diver-
gence theorcm that,

(E, xH, E, xH, ) .rrd,l = O, (40)

\ s
where n is the unit vector normal to dfl and directed outward from V.
lirst, we shall evaluate B.. I.et the field E,, HI be the field set up in
the guide by the slot as formulated in 13(1s.(39). For the field EZ, H, let
us take a normal mode, free propfigation of which is supported by the
guide, traveling toward the right, and lot a be the index of this mode.
Furthermore, take as the region V the section of the guide containing the
slot, bounded on the left by the plane z = Z3 < z, and on the right by the
plane z = z, > z,. The surface LSto \vhich I;q. (40) is to be applied
consists then of thrsc t\vo r]hmcs and the \vall of the ., guide. Over the
plane z = 24 the fields 1 and 2 consist of s,vstjems of waves traveling in
the same direction. When the indicatrd substitutions are made and the
orthogonality property of Eq. 38 is used, the intcgmd vanishes. fln
the plane z = z~ the fields 1 and 2 arc composed of waves traveling in
opposit c directions. M~king usc of the ortho~ona]ity rrkt,ion again
and nutiug that for this s~lrfacc n = i., the integral over this surface
is 21;,.,S.. ( t)nsi(lering the integral over the wall, the sccxmd term in
the intcgrand is zero cvcr,v\vhcrc, for it can bc \rritten as H, c (n x E*),
and n x E~ is zero over the wall, since EZ is a normal mode. Similarly
the field E, must satisfy tl)e c{mdition n x E, = O over the metal wall
Imlln(lary. The only nonvanislliug c(mt ribllt,ion from the ]vall area arises
from the first term of the intcgrand over the region of the slot. One
thus finds

2B.S. = (E, xHJ .ndS, (41)

1 ,I!)t,
\rhcrc n is a llnit vector normal 1t) the \vall and dircctrd into the interior
of t,llc gllide. If z is a llnit vector pcrpendiclllar to the axis of the guide
and tangent to the sllrfacc of the gllidc, then
290 LINEAR-A RRAI ANTENNAS AND FEEDS [SEC. 910

Substituting this last relation into Eq, (41), we obtain finally

Second, we shall evaluate A.. The field El, H, is again taken to be that
set up in the guide by the slot, and the field Ej, HZ is taken to be the
normal mode of index a traveling to the left. In this case the plane
z = 23 does not contribute to the integral in Eq. (40), and the plane
z = 2A contributes 2A,S~; over the wall of the guide the only non-
vanishing contribution arises again from the first term of the integrand
over the area of the slot. It is thus found that

2AaSa = (E, x H,) . n CM

\ *lot
. (jE,,H.. + EI,HC,) exp (j3az) dS. (44)
/ slot
The interpretation of Eqs. (43) and (44) for the amplitudes becomes
clearer if the magnetic field components H., and Ha, are replaced by
surface current densities K.. and Kc, respectively. These are the com-
ponents of the surface current, in the direction of the axis of the guide
and in the direction transverse to it, that exists over the area of the slot f
in the nonslotted guide supporting the ath mode. In terms of these
currents the amplitudes become

(46)

It is evident from these equations that in general the slot does not radiate
equally in both directions within the guide. The formulas also sho~v that
the slot will couple the ath mode to space only if it cuts across current
lines corresponding to that mode, There are various special conditions
under which a small slot is symmetrical with respect to the ath mode.
If all the dimensions of a slot are small compared \vith the ~vavelength,
the variation of a phase factor exp ( +j~az) across the slot can be neglected;
without loss of generality the slot can be located at z = O, in which case
the phase factors are replaced by unity. lVe then observe that

1. A. = B. if El, or Ke, is zero, Reference to Eqs, (39) shows that

?
as far as the ath-mode contribution is concerned, El~ is continuous
at the plane z = O while the magnetic field is discontinuous; in
fact, H~ is in opposite phase to H;. With respect to the ath
mode the slot acts like a shl(nt element in a transmission line.
SEC. 9.1 1] SLOTS IN RECTA .~GULAR WA VEGClDE; TE,o-MODE 291

2. A. = B. if E,, or K., is zero. In this case El, is discontinuous

and H It is continuous at the plane z = O as far as the ath mode is
concerned; the slot beha\es like a series element in the ath-mode
transmission line.
The slots of more general interest are narrow ones having a length of
about x/2 and a ~~idth small compared ~vith the length. The electric-
field distribution in such a slot is nearly sinusoidal along the length and
independent of the feeding system; the direction of the field is transverse
to the long dimension. There are also special conditions under which
such slots reduce to series or shunt elements:
1. Axis of the slot perpendicular to the guide axis. In this case the
phase factor exp ( tj~~z) can again be replaced by unity. Further-
more El, = O; hence if Ka. # 0, Aa = ~a and the slot behaves
like a series element in the ath-mode transmission line.
2. Axis of the slot parallel to the guide axis. In this case El, = O and
the second members of the integrands of Eqs. (45) and (46)
vanish. The variation of the phase factors exp ( TjII?az) cannot
be neglected; however, K.. is constant, and E,, is an even function
along the slot; therefore only the real parts of the phase factors
contribute to nonvanishing integrals, and one has A. = Ba. The
slot oriented in this manner behaves like a shunt element.
Except when special conditions of symmetry are imposed on the field
and on the currents in the slot, for orientations more general than (1)
and (2) above, B. # t Aa, and the slot behaves like a more complicated
combination of shunt and series element. In this case the slot is repre-
sented by a 2- or II-section equivalent in the ath-mode transmission line.
9.11. Slots in Rectangular Waveguide; ZE,,-mode.-The theory of
slots in rectangular guide that supports only the TE10-mode will be devel-
oped in detail. The discussion will be based on the following assumptions:
1. The slot is narro~v; i.e., 2 log,, (length/width) >>1.
2. The slot is cut so that it is to be near the first resonance (length
of the slot = A/2).
3, The field in the slot is transverse to the long dimension and varies
sinusoidally along the slot, independent of the exciting system.
4, The guide walls are perfectly conducting and infinitely thin.
5. The field in the region behind the face containing the slot is
negligible with respect to the field outside the guide; this is tanta-
mount to extending the face containing the slot into an infinite
perfectly conducting plane.
The third assumption concerning the field distribution is closely in
accord with experimental conditions. The fifth assumption is probably
the most radical in its departure from the actual conditions.
292 LINEAR-ARRAY ANTENNAS AND FEEDS [SEC.911

Fh_st the equivalent circuits arc given for the common types of slot,
and then the method is given for calculating the values of the elements
by means of Eqs. (45) and (46) and the electromagnetic formula-
tion of 13abinets principle (Sec.
5.15), provided the reactive field
of the slot is zero. 1 The rectan-
gular guide has the dimensions
shown in Fig. 9.17. The shunt
conductance of a slot normalized
to the characteristic admittance
(b)
of the TE10-mode line is g, and the
series resistance normalized with
respect to the line characteristic
r+jz
impedance is r. We have then
0 0 (1) for a longitudinal slot in the
broad face (shunt element b in
(c) Fig. 9.17)

r+jz

where
g=glsin2
()
TXI
~, (47a)

0 0

(d)

0 0
g, = 2.09 ~ : COS2
() A
2X
(2) for a transverse slot in the
; (47b) i

broad face (series element c in

Fig, 9.17)
FIG. 9. 17.Parameters and equivalent
circmts of slots in rectangular waveguide
(reference point for circuit elements is the
center of the dot). (a) waveguide dimen-
sions; (b) longitudinal slot in broad face, where
() T= TOCOS2 ,
7rx,
a
(48a)

shunt element; (c) transverse slot in broad

face, series element; (d) centered inclined slot
in broad face, series element; (e) inclined slot
in narrow face, shunt element.
0=0523(
333c0s2(a (48b)
(3) for a centered inclined slot in the broad face (series element d in Fig.
9.17)

where

(49b) t

I The resultsto be quoted are due to A. F. Stevenson, Series of Slots in Rectan-

gular lVaveguides, Parts I and II, Special ~ommittee on Applied Mathcc]mtics,
National Research council of Canada, Radio Reports 12 and 13, 1944.
SEC. %11] SLOTS lN RECTANGULAR WA VEGUIDE; TE,o-MODE 293

$= ~ cos 0 F ~a sin 9; (49C)

T,)
and (4) for an inclined slot in the narrow face (shunt element e in Fig.
917)

inecos(%sino)
. (~~)

()1 }
0
sin@
1
As an illustration of the method of deriving the above relations we
shall conclude this section with a summary of the procedure for the
longitudinal slot in the broad face of the guide, Case (1) above. Choose
dimensions as indicated in Fig. 9.17b. Suppose a TE,,-wave of ampli-
tude unity to be incident on the slot from the left; this field induces a
field across the slot so that the slot radiates waves in both directions in
the guide and into space outside the guide. The amplitudes BIOand A ~0
(the mode index a is here replaced by 10) of the waves radiated in the
interior are given by Eqs. (45) and (46) in terms of the field in the slot;
the field, according to the third of our initial assumptions, is

E,, = E, COS (kZ),

(51a)
El. = O, }
where EOis the field at the center of the slot. We have also for the other
quantities entering into Eqs. (45) and (46)

(511))

where Y$j) is the characteristic ~vavc admittance of the TE, O-mode.1 On

inserting these quantities into the expressions for the amplitudes it is
seen at once that A 10 = BIO; that is, the slot is a sh~lnt element, in agree-
ment, with the previous conclusions relative to slots parallrl to the guide
axis. The amplitudes are given explicitly by

lO=BIO=-Eo:(:Ysint:)(
osf+)J52)
1 TIlr rollst:lllts of (K,), and ,S, flcorrespond to the nlmic being so norlllalimd thtt

the electric Iicl(l :wross t,llc guide is given by *sill ~q:.

294 LINEAR-ARRAY A.kT7ENNAS AND FEED,q [SEC. 911

where w is the width of the slot. It is useful to express the slot excita-
tion in terms of a voltage transformation ratio. The ( voltage
across the slot is defined to be the line integral of the field across the slot
at its center, i.e.,
VO = wE,,

while the voltage in the guide corresponding to any one of the dominant-
mode waves is defined as the line integral of the field across the center of
the guide, i.e.,
v, = /).4,0 = blllo.

The voltage transformation ratio is then

(53)

It is recognized further that the amplitude A,, measures directly the

reflection coefficient f7 (at z = O) in the transmission-line equivalent of
the dominant-mode wave. If the slot is resonant, the value of r at z = O
must be real, because the impedance looking to the right is real at that
point at resonance. Then if the slot is a shunt element of normalized
conductance g, the total admittance at z = O is 1 + g; while if the slot
is a series element of resistance r, the input impedance at z = O is 1 + r.
From Eqs. (2.30) and (2.36) g and r may be expressed in terms of r by

(54)

The value of r can be evaluated for a resonant dot by energy-balance

relations. The total energy incident on the slot is equal to the sum of
the reflected, transmitted, and radiated energy. The incident po~~er is
S./2 for an incident ~rave of unit amplitude; the reflected po\ver is
(A ,O)~~fl. The total amplit~lde of the dominant-mode ~vave to the right

of the slot is 1 + B1o; hence, the transmitted power is

In computing the power radiated by the slot use is made of a result

obtained, by means of an electromagnetic Babincts principle,l for the
radiation resistance of a center-driven narro~v slot in an infinite perfectly
conducting plane sheet of zero thickness. In this (ase the input resistance
is
1H. Booker, Babinets Principle and the Theory of Iteeonant Slots, TR13 Report
No, T-1028
 (
i
SEC. 912] hXPERI.WE.V7AL l)A7A O.Y SLOT RAL)IATORS 295

In the infinite sheet problem the slot radiates to both sides of the sheet;
in our case the slot radiates to one side so that the radiation resistance
is assumed to be simply twice the above value. The power radiated
by the slot is then given by

1 Vg
= 73 V~ g \vatts
22Rr PO

Writing the energy balance equation and remembering that A,0 = BIO,
\ve have

s. _.. 9
~ IAIoI 2 + ~[1 + IA I012+ z~e (.AIo)]+ 7Svi~.
2

Finally, since .4,0 = r is real, \ve obtain from the above

vi
l+; =73Q (55)
POS.(A10)2

Making use of Eqs. (51a) and (51b) and substituting this last result into
Eq. (54), the conductance of the resonant shunt slot is

(56)

We already have the voltage transformation ratio in Eq. (53); substituting

this into Eq. (56) gives the final expression for the normalized shunt
conductance,

(57)

9.12. Experimental Data on Slot Radiators. Confirmation of the

theory developed in the last section has been obtained by experiment
for the longitudinal slot in the broad face of the guide (Case b, Fig. 9.17)
and for the inclined slot in the narrow face (Case e, Fig. 9.17).1 The
resistance of a longitudinal slot as a function of its position with respect
to the center of the guide is shown in Fig. 9.18; the points are in good
agreement with the formula

I A. L. ~ullen, The characteristics of Some Slot Radiators in Rectangular

Waveguides, Royal Aircraft %tablishmcnt, Great Britainj Tech. Note No. Rad. 200;
Dodds and Watson, Frequency characteristics of Slots, McGill University, PRA-
108; Dodds, GuptiO, and Watson, Further Data on Resonant Slots, McGil[ Univer-
sity, PRA-109; E. JV, C,upti]l and W. H. ~~atson, Longitudinally Polarized Arrays
of Slots, hlcGill University, PRA-104.
 .4

296 LINEAR-AURA Y AN TENNAS AND FEEDS [SEC. 9.12

g = gl Sinj ~ , (9.47a)
() a
but the numerical constant g, is 1.73 whereas the theoretical value given
by Eq. (47b) is 1.63. The discrepancy is probably due to the assumptions
underlying the theory. The frequency characteristics of longitudinal

100
80 ?

60

40 \

20

10 -
~8
%6 \,

I
..-
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Slot displacement from center of guide to center of slot, in.
FIG. 9. 18.Resistance offered by a longitudinal slot as a function of its displacement
from the center. The slot dimensions are & by 2 in., the waveguide is 11 by 3 in., X = 10.7
cm. The data fit the relation G = Zo/R = 1,73 sinz [(mz/Iz) 1], (From J. W. Dodd-s, E W.
Guptdl, and W. H. W&son bv permi.wion of the Nalional Research Council of Canada.)

slots as a function of slot width are presented in Fig. 9.19, which shows
that the wider the slot the flatter the frequency response. The maxi-
mum of conductance does not coincide with the vanishing of susceptance.
For practical convenience dumbbell-shaped slots such as the one illus-
trated in Fig. 9.20 have been used in arrays in place of rectangular slots.
The perimeter of a resonant slot is generally equal to a wavelength. The
length of a resonant dumbbell slot is therefore less than that of rectangular
ones; they can be used with less sacrifice of mechanical strength, since
less guide is cut away. The dumbbell slot is also simpler to machine
SEC. 912] EXPERIMENTAJ, DATA ON SLOT RADIATORS 297

0.6 -

0.5 II

~.- .
0.4 <>
?
* \
\
0.3 -
1~ !
Slot width, In.
\ \ i
0.2 0 ------ $fS
\ :\ %
-- y4
)1
A 1/2
\\
\o
+

I
+++13 I

I I
1%=4 I I
2.8 2.9 3.0 3.1 3.2 3.3
Frequency 109CPS

0.9 -

0.8
&
,
0.7 ,
d \
/ /
00.6 >i
$ u /
,
: 0.5
~ ,
% /
~ 0.4 , \
/ \ \
5 \
/ \ b ~~
0.3 v .
/ \
. \ -a
0.2
d U. %Y *
- . Q
0.1
I
n
.
2.7 2.8 2.9 3.0 3.1 3.2 3.3
Frequency x 109 cps
F1~ 9.19.Admittance of longitudinal slot as a function of frequency (center of slot
is l,98cm from the center of the waveguide). (From the work oj.T, W. Doddaarui W. H.
Wa180nbg permis.sion of the National Re.earth Count?Jof Canada.)
298 LINEAR-ARRAY ANTENNAS AND FEEDS [SEC.9.12

because the dumbbell areas are drilled rather than cut by a milling
machinp. Another technique for shortening the resonant length is to
place a thin sheet of dielectric over the slot; a sheet of polystyrene of
0.007-in. thickness reduces the resonant length by 1.13 percent at 10.7
cm. The dielectric sheet also serves asa pressurizing device.
If the conductance and frequency characteristics of each element of a
slotted linear array are known, it is possible to place a given number of
longitudinal slots k,/2 apart so that they are effectively in parallel and
to short-circuit the far end of the guide A,/4 from
the last slot so that the admittance in parallel
with the last slot is zero. Then if there aren
m elements, the relative conductance of each slot
FIw9.20.-Dumbbell-shaped must (bysuitably choosingx J bemade to equal
slot. to l/n in order to provide a good match.
Because the slots are placed in the same way as a set of dipoles, end to
end, the mutual impedance of the slots is negligible.
The conductance of a longitudinal slot cut in the broad face of the
guide can be readily determined by measuring the input impedance of
n slots in parallel because the mutual impedance between slots is negligi-
ble. This is not so when the slots are cut in the narrow face. The effec-
tive conductance of this slot may be found by measuring the additional
conductance produced when one slot is added to an array. In practice
a number of slots, for example 10, are cut and the input admittance
determined. The input admittance is then again determined when addi-
tional slots are cut in sets of, say, 3. Eventually the total susceptance
becomes constant and the conductance linearly proportional to n (if the
susceptance is also proportional to n, the slot depth is adjusted for
resonance). The incremental and ordinary conductance are plotted
in Fig. 9.21 as functions of the angle 0. Both obey very well the law

g = go sinz 0

over the measured range. This is in good agreement with Eq. (5o) for
small angles e.
Slots cut in the narrow face have the very useful feature that the
variation of susceptance with frequency is very small compared with
that for slots in other positions in the guide. The variation of admittance
with slot depth is also small as is shown by Fig. 9.22. Thus a change of
~ 1 mm in depth from the resonant point produces a change of only 4 per
cent in conductance and only a small change in susceptance. Because
bhe depth of cut can always be accurately controlled in a milling
operation, this represents a tolerance which can easily be attained.
Since the angle of the slot to the guide axis can also be accurately !:eld,
the system represents a satisfactory array from the constructional point
 SEC. !)13] PROBE-FED SLOTS 299

of view. A possible objection is that there is an appreciable degree of

unwanted polarization in these beams. The field over the slot has a
longitudinal component proportional to cos 0; the transverse component
of the field does not reverse direction with reversal of the direction of
inclination of the slot and gives rise to an unwanted side lobe at about
40 to the main beam. For tilt angles up to 15, however, the unwanted
polarization is less than 1 per cent of the radiated po~ver.1

0.1
0.08

0.06

0.04

Cl
3
~ 0.02
%
~
8
0.01
0.008
0.006

0004

4
0.002 I
0 5 10 15 20 25 30 35
Inclination of slot ( O), deg
FIQ. 9.21.Incremental and ordinary conductance as a fu,,cti<~]l of slot inclination.
(From the work oj E. W. Guptill and W. H, Watson b~ permission of lhe National Research
Coun~l of Canada.)
I
9.13. Probe-fed Slots.It was pointed out in Sec. 9.9 that there are
various positions in a guide and various orientations of the slot axis for
which no radiation takes place. It is possible, however, to make any
slot of this type radiate by inserting a suitable probe into the guide adja-
cent to the slot. 2 The probe introduces the necessary asymmetry in the
field and current distributions for excitation of a field across the slot.
The probe-fed unit has many advantages. In particular the direction
of the field across the slot depends on the side in which the probe is

1Dorlds, Guptill, and Watson, op. cit.

2 R. 1. Clapp, Probe-fed Slots as Radiating Elements in Linear Arrays, RI,
Report No. 455, Jan. 25, 1944.
 T

&

300 LINEAR-ARRAY ANTENNAS AND FEEDS [SEC. 9.13

inserted; the phase of a given slot can be shifted 180 by switching the
probe position. An example of this phase reversal is afforded by the
array of slots on rectangular guide illustrated in Fig. 9.23; here the phase

I I I I I I I
0.05

0.04
v
al
:
~ 0.03

:
v

0.02

Depth of cut 11cm

0.01

0.004

; 0.002

g
n
8
2 o
u-l

-0.CQ2
I WI

FIG. 9.22.Admittance of a 15 inclined slot on narrow edge of rectangular wavegulde.

The waveguidedimensionsare 1# by 2~ in., X = 10,7cm,, and the width of the slot is }
in. (From the work of J. W. Dodda, E. W. GuPM1, and W. H. Watsonb~ permission of
the Naiionol Rmearch Council of CLI-wxia.)

reversal of the probe is used to compensate for the 180 phase clifference
corresponding to the k~/2 spacing of the slots; the result is an array of
equiphased slot radiators.
Another advantage of the probe-fed unit is that the amount of energy
radiated by the slot is controlled by the probe insertion. For the case
illustrated in Fig. 9.23 where the probe is parallel to the field, the coupling
SEC. 9.14] WA VEGUIDE RADIATORS 301

is adjusted by the probe depth. To excite a slot in the narrow side of a

rectangular guide a bent probe is used, as shown in Fig. 9.24; here the
coupling can be varied by the angle between the hook of the probe and
the electric field. In some cases the screw head of the probe introduces
undesirable impedance characteristics; the head of the screw can be

FIG. 9.23.Probe-fed slots on rectangular waveguide. The arrcws show lines of current
flow.

ground off after the desired coupling has been obtained, or the unit can
be balanced externally by a dummy screw head. Many variants of the
probe can be developed for various types of guides and modes; the reader
is referred to Clapps report for details.
9.14. Waveguide Radiators.The impedance of a radiating element
has been seen to consist in general of a resistive and a reactive component.
The reactive component is gener-
ally undesirable, since it enhances
the frequency sensitivity. The [Col
reactance vanishes under special ~Q
conditions, but these are not -
w %~g
always optimum operating con- FIG, 9,24.Prone-fed transverse slots 0]1 the
ditions; for example, in the case narrow face of rectangular waveguide.
of the dipole element discussed in
Sec. 98 resonance occurs at a probe depth that is generally too small
to meet power-extraction requirements. Slots and dipoles suffer another
severe disadvantage at short wave~engths as in the l-cm region where
they become so small that they have an insufficient power-handling capac-
ity and the tolerances on the dimensions become impractically restrictive.
The waveguide radiators illustrated in Fig. 9.25 are less subject to
the above limitations. 1 The element consists of a waveguide coupled to
the main guide bv a T-junction. As shown in the figure, two arrange-
] W, Sichak and E. M, Purcell, (krsecz .4ntennas with a Line Source and Shaped
Cylindrical Reflector, RL Repart So. 624, Nov. 3, 1944, pp. 7-13.
302 LINEAR-ARRAY ANTBNiY-AS AVD FEEDS [SEC,9.1-I

ments are possible corresponding to longitudinal and trans~erse polariza-

tions. The longitudinally polarized element, just like a slot with axis
transverse to the guide axis, presents an impedance in series with the
main line; thetransversely polarized element, like a slot with axis paral]el
tothe guide axis, is equivalent to a shunt element across tbe transmission
line, inserted in the plane of symmetry of the radiator that is perpendicu-
larto the guide axis. It has been found experimentally (cf. Sec. 10.11)
that the open end of a waveguide can be represented by a load admittance
consisting of the radiation resistance in parallel \vith a capacitive react-

(a) (b)
~,~. $~~.-l~avegu,de radiators: (a) longitudinally polarized; (b) transversely polarized.

ante. lhe net~vork equivalent of the T-junction consists in a similar

manner of a capacitative reactance in parallel ~rith the input impedance
of the branched guide. Both the input and termination capacities are
junction effects and may be expected to be of the same order of magnitude.
If the length of the branch guide is k~/4 where k: is the g~lide~~a~elength
in the branch, the terminal rapacitative reactance is transformed into an
inductive component at the input end; and since the inductive component
is in parallel with the T-junction capacitance, a near-resonant condition
should result. In actual practice, ho~~e~-er,the length of the section is
different from A; 4. The correct length has been found to be given
closely by the result of an analysis of a branched waveguide which takes
the junction effects into account, namely,

=wl+ln*) (58)

The dimensions b and b are defined in Fig. 9.25. IYith the above length
the element has been found to be very closely a pure resistance.
The coupling of the element to the line, i.e., its resistance or conduc-
tance, is a function only of the relati~-e dimensions of the branch guide and
the mam gmde. It is the particular advantage of the waveguide element
T

SEC. 915] AXIAI,LI S1Jl>!fE7RICAL RADIATOR,Y 303

that the coupling factor can be adjusted independently of the resonance

condition. On the assumption that the impedance presented to the
main guide when the branched guide radiates into free space is not very
different from that when the branched guide couples to a second guide
parallel to the main guide, the resistance of the longitudinally polarized
element has been calculated to be

()
lb2
=27

The quadratic dependence on b/b is in accord with the experimental

I
results; these results indicate, however, that thenumerical factor is not ~.
I No systematic study of a single element has been made as yet. The
coupling factor for the transversely polarized element has meaning only
in terms of a complete array because with these elements mutual inter-
actions become very significant. The results will be given later in the
discussion of nonresonant arrays which make use of these elements.
The length of the element given in 13q. (58) can be increased by any
integral multiple of A~/2 \vithout affecting either the resonance or the
coupling factor. This is advantageous in that it provides a method for
1 shifting the phase of the radiator 180. For the same power extraction
the b dimension of the transversely polarized element must be larger
than that of the longitudinally polarized element because the former cuts
across transverse currents that are smaller than the longitudinal currents
on the broad face of the guide. Consequently, the tolerances arc less
restrictive for the transversely polarized element, and it therefore is
preferred if all other considerations are equal. At short ~vavelengthsj
e.g., at 1 cm, the length of the radiator is so small that the wall of the
main guide can be constructed of that thickness, and the radiating ele-
ment then takes the form of a slot in this wall. This produces a sturdy
mechanical system.
9.15. Axially Symmetrical Radiators.For general communication
~ purposes it is desired to have a stationary antenna with an axially sym-
metrical pattern covering a large region of space. The simplest antenna
of this type is a half-~va~e dipole. The gain of the dipole, however, is
too low to meet the usual requirements on range, and it is therefore
necessary to design an antenna having the axial symmetry of the dipole
but with a more directive meridional pattern. The Ia,ttcr can be achieved
by means of a linear array of axially symmetrical r~diating elements,
an example of which is illustrated schematically in Fig. 9.26. The ele-
! ments to be discussed fall into t~vo gro~lps distinguished by the polariza-
tion of the field: (1) transversely polarized radiators producing a field in
which the electric \ector lies in planes normal to the axis of the array,
..
(2) longitlldinally polarized radiators prodllring a field in ~vhirh the elec-
 r

f
1,
304 LINEAR-ARRAY AX TENNAS AND FEEDS [sm. !315 ,

tric vectors lie in meridional planes having the axis of the array as a
common line of intersection; the transverse element is analogous to a
magnetic dipole, and the longitudinal element to an electric dipole.
The transverse element in its ideal form should consist of a circular ring
of uniform current, while the idealized
z longitudinal radiator should consist

I of a short circular cylindrical current

sheet of uniform density running
parallel to the axis of the cylinder.
In practice these elements can best
be approximated by an array of ele-
ments located at points disposed
symmetrically about the array axis in
a plane normal to it. Thus the an-
tenna as a whole is, in fact, a three-
dimensional array; however, design
problems for the azimuthal and merid-
ional patterns are completely separ-
V able. The meridional pattern is a
FIG. 926.-Array of axially symmetrical
radiators. straightforward linear-array problem.
The azimuth pattern reflects the
symmetry of the arrangement of the radiators about the array axis and
consequently deviates from a uniform pattern, showing maxima and
minima. The ratio of maximum power to minimum is referred to as the
azimuth ratio; it is generally required that this ratio be less than 2.

~f$%zzfi
I~IG. 9.27,Dipole fed I)y three-wire line,

The Tridipole lransverse ~lement.lirst the elements designed for

transverse polarization ~vill be considered. A simple approximation to
the circular current ring is obtained by arranging three half-]vave dipoles
on the circumference of a circle. The basic unit illustrated in Fig. 927
is a three-wire-line-fed dipole analogous to the slotted dipole discussed
earlier. The central line serves as a probe to couple the dipole to the
interior of a v-aveguide. The axially symmet,riczl tridipole array sho\\n
in Fig. 9.28 is designed for use w-ith a coaxial line. The clement is made
so as to slide over the outer conductor and is soldered to the latter at the
appropriate location. In order to maintain the azimuth pattmn sym-
metry it is essential thiit, the three probes I)c inserted t,o eql)al depths.
SEC. 915] AXIALLY SYMMETRICAL RADIATORS 305

The line coupling can be achieved either by inserting the probes so as to

make contact with the inner conductor of the line or bv. capacitive .
coupling in which the probes do not make contact with the inner conduc-
tor. In the former case the probes are soldered to the inner line; a more
reliable procedure is to have threaded holes in the inner conductor into
which the probes can be screwed and then soldered to ensure good contact.
For capacitative coupling, probe settings can
be made by slipping a shim of suitable thick-
ness over the inner conductor; the shim is
subsequently removed.
Satisfactory results have been obtained
with tridipole elements over the 10-cm bandl
and at various longer wavelengths. The im-
pedance characteristics of a single unit can be
adjusted in the course of design by the choice
of the dimensions of the dipole wings; the
impedance characteristics of an array of units
are adjustable by means of the probe depth.
Figure 9.29 shows the frequency sensitivity of
Diameter
the pattern of a tridipole unit designed for the
0.56h
10-cm band, the unit being fed from a 50-ohm Fm.9.2S.Tridipoleradiator.
line with a ~ in. OD. The pattern exhibits a
high degree of stability. The same element with its probes 0.5 mm from
the inner conductor handles 10-kw ~eak . .Dower without breakdown. It
was found that at 10 cm the dimensions of the unit are not critical and
the elements can be produced in quantity by die-casting techniques with
good reproduction of performance. A 3-cm version, however, requires
manufacturing tolerances too close for practical use.
Axially Symmetrical Slot Array.Another type of unit for transverse
polarization is provided by an array of slots along the circumference in the
wall of a circular guide or coaxial line, the long dimension of the slots
being parallel to the axis of the guide. A number of factors enter into the
design of the unit.
The most important is that the line must carry a radially symmetrical
mode so that the slots are excited equally. This condition is fulfilled by
a coaxial line supporting only the ZElll-mode and by a circular wave-
guide propagating the Zllfol-mode as indicated in Fig. 9.30. However,
for both cases a slot cut parallel to the guide axis does not radiate. It
is therefore necessary to excite the slots by means of probes as shown in
the figure, and to ensure symmetrical excitation the probe depths must
be uniform.

1 H. Riblet, Horizontally Polarized Nondirectional Antennas, RL Report No.

517, Feb. 14, 1944.
306 I,INEAR-ARRA Y AN 1BNNAS AN.!) FhhI)S [SEC. 915

220 210 200 190 Moo 1700 1600 150 140

230 130

2400 1200
I
250 110 i
I
260 1000

270 900 i

2800 800

290 700

300 60

i
310 500

i
320 330 340 350 00 100 200
30 400
FIG, 929.-Patterns of a tridipole unit in the plane of the unit.

. Current flow induced by probes,

Transverse electric-field config.
uration of unperturbed mode,
9.30. .\xially symmetrical radiating unit formed by a circular array of dots.
14IG.
 SEC. 9.15] AXIA I.LI .71.if,\1E7RI(:,lI, RADIATOILT 307

A second factor is the minimum number of slots required to produce

: a pattern having a satisfactory azimuth ratio. This is found to depend
on the size of the line; the larger the guide diameter the greater the num-
ber of elements. Fora, l-in. ODcoaxial line operating inthe3-cm band

2100 200 190 180 1700 160 150

220 140

230 1300

240 1200

250 1100

260 100Q

270 900

280 800

290 700

300 600

3103 500

320 40

330 340 3509 00 100 zoo 300

F1~. 9.31. P~tterns of circular arraj-s of four and six slots on coaxial line of 1 i. OD.

the minimum number is six; Fig. 9.31 shows the patterns obtained from
four and six elements; the former reflects strongly the fourfold symmetry
of the array. Figure 9.32 illustrates the pattern resulting from a seven-
element array on 1~-in. circular guide, again for the 3-cm band.
(
The minimum number of elements is also related to a problem of
30s LIA EAR. ARRAY ANTENNAS AND FEEDS [SEC. 915

mode control. The conventional coaxial line in the 3-cm band, for which
all modes other than the lEM-mode are beyond cutoff, is too limited in
its breakdown properties and mechanical strength, the latter being an
important factor in long arrays. I,ine of a larger size is therefore used
which can support other modes. Considerable care must be taken at
the input end of the line to ensure
radially symmetrical e x c i t a t i on.
The probe inserts for exciting the
slots likewise excite higher modes. -i-
No mode will be excited, however, if
its planes of symmetry do not con-
tain the symmetries of the geomet-
rical configuration. There is thus
a minimum number of probes for
which the higher modes excited ~vill ===1-
attenuate.
T
180

J
+d
1

1 _

_
__
00
FIG, 9.32.Pattern of circular array of FIG. 9.33. TELo- to TMol-mode con.
seven slots on circular waveguidc with 1$ in. verter. >
OD.

A similar mode-control problem exists in the circular guide, for a ,

circular guide that can support propagation of the l,l~ol-mode neces-
sarily supports the ZE, l-mode. It is therefore necessary to feed the guide
in such a manner that the ZEll-mode is not excited, and again there is a
minimum number of slots required. The proper feeding of the circular
guide is achieved by transition from the ZE,O-mocfe in rectangular guide
through a TE,O- to ZMOl-mode converter, which is illustrated in Fig.
933. Briefly the principle of its operation is as follows: The distance 1
is equal to xl,/4 or 3A11/4 where xII is the guide wavelength for the TEll- 1
mode; this puts a large series reactance for this mode at P between the
rectangular and circular guide so that, the mode is not fed into the latter
guide. The diameter d is chosen to be AO1/2 where AOIis the TMO,-
 SEC. 9.15] AXIALLY SYMMETRICAL RADIATORS 309

mode guide wavelength; for 1 = 3A11/4, this gives a good match for the
TMO,-mode.
Longitudinally Polarized Elements,Satisf actory elements of this type
in the microwave region have thus far been developed only for the 10-cm
band. A longitudinal element analogous
to the tridipole unit can be produced by
a circular array of dipoles with axes
A
parallel to the guide axis. It is found,
however, that a longitudinally polarized
tridipole array gives rise to a pattern
having a decided threefold symmetry
while a larger number of dipoles results
in a unit whose design is very critical. A
cylindrical element with three-point ex-
citation provides a simple solution; the
element is shown in Fig. 9.34. It can
be thought of as being derived from a
system of three longitudinal dipoles of
the type illustrated in Fig. 9.27 in which
the wings have been extended laterally
and joined into a cylinder. The currents FIG. 9.34.Longitudinally polarized
axiaUy symmetrical radiating unit.
tend to spread out uniformly over the sur-
face giving a uniform azimuth pattern. The unit is made in t~vo parts,
one consisting of a die-cast spider carrying the two outer lines of the
three-wire line-feeding system and the other the pair of cylinders that
correspond to the dipole wings. As ~vith the transverse unit the svstem
is fed by probes which couple the cylinders to
the line; the general remarks made previously
ielectric concerning the insertion and alignment of the
plug
probes likewise apply here.
Attention should be called to a longitudi-
nally polarized slot radiator which can be design-
ed ~vith a coaxial line. The element illustrated
in Fig, 935 consists of a slot running completely
around the wall, Mechanical support is pro-
vided by filling the line with dielectric. The
FIG 9.35. Long1tudi-
element obviously gives a uniform pattern but
nallv polarlzed slot radiator suffers from a number of disadvantages. It is
for coaxial line,
very frequency-sensitive; mechanical properties,
particularly of long arrays, are poor; satisfactory contact between the
) dielectric and metal is difficult to maintain particularly under mechanical
and thermal stresses with the result that the system becomes susceptible to
electrical breakdown. Development of arrays with these units was
finally given up because of these limitations and difficulties,
310 LINEAR-ARRAY ANTENNAS AND FEEDL~ [%c. 9.16

9.16. Streamlined Radiators. Arrays of axially symmetrical radia-

tors have been developed for airbornr, ground, and ship installations.
With the development of high-speed planes, however, aerodynamic con-
siderations have become in-
creasingly significant in antenna
installations. Arrays of elements
of the types already discussed
produce sufficient aerodynamic
drag to present a serious installa-
tion problem. It has therefore
been necessary to make some com-
promise between pattern and aero-
dynamic requirements and to
design elements whose geometry
has a less deleterious effect on the
aircraft. For this purpose various
types of streamlined elements have
been developed which, though
lacking the uniform coverage of
the axially symmetrical units, still
produce patterns with not too
large an azimuth ratio.
150 180 210 Two types have been devel-
l;IG. 9:M-Pattern prwfui-ciil)> a pair of oped, one for transverse, the other
a]lt,phahed dot, s
for longitudinal polarization. I,et
us consider the transverse radiator first. It has bcrn found that two slots
cut opposite each other on a coaxial line and excited 180 out of phase
produce a pattern with an azimuth ratio not exceeding 5 or 6; this is sho~vn

&
o

o
o X=+%lsk.h.
tim/[;yjLy/,,,,,,
#))QJu12>>,,,>,>,
,,,/),,,,, \
00
~lG.937,-.lrray of thrw pairs of slots 011 strcarl]litlwl elliptlral waveguide.

in Fig. 936. The mlrrcnts tcmd to run cornplctrly around the cylinder,
giving a continuous, if not completely symmetrical, current distribution.
Starting from this obscrvat,ion, one can prm,cwl in several directions to the
design of streamlined elements. First, the {)~iter cund~lctor instead of
 SMC. 916] STREAMLINED RADIA TORA 311

Ftc. 9.38.T ransversel y polarized streamlined radiator.

being cylindrical can be made elliptical or streamlined, or the inner con-

ductor can be omitted entirely and a streamlined section of sufficient
size used instead as a waveguide. Figure 937 for example, shows an
array of three slots on streamlined
300
elliptical guide for the 3-cm band. 0 330

Transition to the elliptical guide

from rectangular guide is effected
500 310
by a tapered section. It has been
found that if the ratio of major to
the minor axis of the ellipse is at 700 290
least 4 and if the minor axis is
900
approximately A/4, the azimuth 270
ratio is in the neighborhood of 2;
this figure has been obtained with 11OQ
250
the three-unit array referred to
above. 1300
230
A method of introducing the r-f
that provides a good impedance
match is shown in Fig. 938 for a
10-cm band system. Hcrethe two 150 180 2100
FIG. 939. -l<adiation pattern of a trans-
slots are cut at the point of maxi- versely polarized streamlined radiator,
mum width of the guide. .4 slotted
1
dipole on the end of small coaxial line is used to excite the slots. The
wings of the dipole are cut to fit, and each wing acts as an exciting antenna
for one slot. The ~SWR obtained \vith a single element is less than 1.2
312 LINEAR-ARRAY ANTENNAS AND FEEDS [SEC. 916

over a 16 percent band. The pattern shown in Fig. 9.39 is likewise satisfac-
tory. The only longitudinally polarized unit that has been built is for
the 10-cm region. It consists essentially of two vertical dipoles about a
quarter wavelength apart supported on opposite sides of coaxial line.
Two such dipoles in free space would have an oval pattern which, how-
ever, does not have too large an azimuth ratio. This, of course, is modi-
fied by the coaxial line; but if the line is small enougn (ordinarily ~ in.

300 & 3300

500 310

703 290

90 270

1100 ~~o

1300 23W

1500 1800 210

FIQ. 9.40. Longitudinally polarized FIG. 9.41.Radiation pattern of a lon-
streamlined radiator (H-element). gitudmally polarized streamlined radiator
(H-element).

OD), the effect is small and does not seriously impair the pattern. The
unit, generally referred to as an H-element, is shown in Fig. 9.40. The
two dipole ~~ings are supported by elliptical straps, and the whole unit is
placed over the coaxial line and excited by the probes projecting into it,
The elliptical straps serve also as a \vave trap, to prevent currents running
along the coaxial line. The pattern produced by the unit is sho~~n in
Fig. 9.41. An array of such elements is ordinarily enclosed in a close
fitting elliptical housing.

ARRAYS
It is shown in the sections on general pattern theory that the pattern
of a linear array is determined essentially by three factors: (1) the rela-
(
tive amplitude and phase of the current distributions on the elements of
ihe array, (2) the spacing of elements along the axis, and (3) the form
factor of the pattern of a single e!emrnt. 1n practice these factors are
 SEC. 917] LOADED-LINE ANALYSIS 313

not independent variables; the amplitude and phase of the elements are
1 determined in part by interactions between the elements, which in turn
are functions of their spacing. At longer wavelengths feeding techniques
are available whereby the amplitude and phase, except for external field
coupling between the radiators, are independent of spacing. In micro-
wave antennas the elements must be fed in cascade from a transmission
line; the phase of the radiator thus depends on the phase velocity in the
line and the position of the element along the line; phase and spacing are
thereby most intimately related. The relation becomes complicated
. further because the feeding arrangement results in a loaded transmission
line with propagation constant and characteristic impedance different
from those of the unloaded line. Finally mutual interactions between
the elements because of their external fields must be considered. The
result of these interrelations is that the transition from the properties
of a single element to a composite array is not a calculable design pro-
cedure but must be determined to a large extent on an empirical basis.
9.17. Loaded-line Analysis.-The relation between the parameters of
loaded and u] loaded lines will be investigated first. Consider a line,
whose unloaded parameters are the characteristic impedance ZO and the
complex propagation constant ~ = a + j~, loaded at regular intervals t
with identical radiating elements. Taking a fixed reference point in a
radiator, the radiator in general can be regarded as a bilateral passive
four-terminal network inserted at the reference point between two seg-
ments of line. It was shown in Sec. 2.2 that such a network can be
replaced by a T- or II-section equivalent; in the no~ation of Sec. 22 the
three impedance elements of the T-section will be designated by 21, 22,
and 23 and the elements of the ~-section by 2A, ZB, and Zc. The rela-
tion between the T- and H-section elements is given in Eq. (2.10).
The radiating elements that have been discussed in the earlier sections
all have at least one plane of symmetry; if the reference point is taken in
this plane, the T- or II-section equivalent of the radiator is symmetrical;
i.e., 2, = Zt and Z~ = ZC. It was shown further in Sec. 2.9 that a sec-
tion of homogeneous transmission line of length 1 has a symmetrical
T- and II-section equivalent; from Eqs. (28), (256a), and (2.56b) the
elements of the T-equivalent are found to be

(59)

By means of Eqs. (27), (28), and (2 10) the elements of the equivalent
II-sections are obtained from these. The II-elements are
*

(60)
 ,4

314 LINEAR-ARRAY ANTENNAS AND FEEDS [SEC. 917

On replacing both the radiators and the line segments by their equivalent
T-sectior.s, the loaded line is reduced to a cascade of networks as shown
in Fig. 9.42a; the points A and A are the reference points in the radiating
elements. By splitting the shunt element 23 into a pair of impedances
2Z~ in parallel, the line is further reduced to a chain of symmetrical net-
works, a single unit of which is shown in Fig. 9.42b.

(c)
FIG. 942.-Network system equivalent to a loaded trammission line: (a) T-section
replacements of radiators and hne segments; (b) reduction to symmetrical networks; (c)
II-section equivalent of the network in (b).

The characteristic impedance Z: and propagation constant -y of the

loaded line are obtained by reduction of the network in Fig. 9.42b to its
equivalent T- or II-section and subsequently determining the parameters
of a homogeneous line having a length 1 for which the above T-section
(or II-section) constitutes an equivalent representation. In the present
case the simplest procedure is to reduce the network to a II-section by
replacing the T-network of elements 21 + ~ 1and ,~s by its H-equivalent.
The completely reduced network is shown in Fig. 942c. If .~~ and Rj
are the elements of the reduced network, the loaded-line parameters are
given by
 SEC. 9.17] LOAI)E1)-1,1.VE A .V.4 L Y,YI,T 315

Z: coth <L = ~~ ; Z: sinh (~1) = ~~. (61 )

() 2

lf the values of ~~ and ~~ given in Fig. 942c together with the values
of al and as given in 13q. (59) are inserted in Eq. (61), the half-argument
identities for hyperbolic functions may be used to obtain Campbells
formulas:

cosh (-/1) =
() 1 + ~;
.
cosh (+)
b

.
+(%3+~.+~-~$,
) Sinh (IJ); (62)

Zj sinh (-yl) =
[
Z, + 2, tanh

20
()1
71
~

2

sinh (71)

+2
[
Z,+ Zotanh
()1
~ . (63)

For the present purposes the attenuation in unloaded waveguides due to

conduction losses in the walls may be neglected; under these conditions
~ = jp and the propagation constant of the loaded line is given by

(+%)
Cosh() cOs1+(2+2+2%)
in
64)
It is seen at once that the loaded line has a complex propagation con-
stant ~ = a + j~ in which both the attenuation and phase constants
are functions of the loading and the spacing of the elements. Equation
(64) shows, however, that if the spacing is equal to half the wavelength in
the unloaded line, the relation reduces to

cosh (~1) =
() 1 + -~a ;
J
l=;. (65)

If the radiating element is a pure shunt element so that Z, = O, it is

found directly from Eq. (65) that

+=j$=y.
9
Similarly if the element is a pure series element, in which case ZS = cc, it
is found that y = -y. lh USthere is no attenuation in a line loaded with

pure series or pure shunt elements at half-wavelength intervals. The same

is true of a line loaded at wavelength intervals, For arbitrary spacings
0 the propagation constant of the shunt-loaded line is given by

cosh (~1) = cosh (yl) + ~ sinh (71) (66)

316 LINEAR-ARRAY ANTENNAS AND FEEDS [SEC, 9.18

and for the series-loaded line

cosh (~1) = cosh (Y1) + ~ sinh (+), (67)

where 24 is the impedance of a single element.

The pure series- and shunt-loaded lines with half-wavelength spacing
have the additional property of producing a uniformly illuminated array
when the line is suitably terminated in a short circuit. For the series-
Ioaded line the short circuit is made an integral number of half wave-
lengths beyond the final element; by virtue of the half-wavelength spacing
the array is equivalent to a simple series circuit of equal impedances; all
the elements therefore dissipate equal amounts of power. The shunt-
loaded line is terminated A./4 + nAo/2 beyond the final element, n being
an integer; this array is equivalent to a system of equal impedances all
in parallel, and again all the elements dissipate equal amounts of power.
The loaded-line analysis takes no account of coupling between the
elements by means of the external fields. Campbells formulas do show,
however, the interrelation between the amplitude and phase of the ele-
ments and the spacing and also the relation between the amplitude of
the element and its phase, for the phase velocity is a function of the
coupling between the radiator and the guide.
9.18. End-fire Array .-The only important examples of end-fire
arrays for the microwave region were two very similar antennas for opera-
tion at wavelengths of 10.7 and 11.7 cm. 1 They consist of 18 individual
radiators a quarter wavelength apart and fed from a coaxial line. The
antenna shown in Fig. 943 is a 14-element ex~erimental model. An
antenna of this type must be terminated in a dummy load to absorb
the unradiated power in the line thereby eliminating a reflected wave;
otherwise the reflected wave would give rise to an end-fire pattern in its
direction of propagation, that is, in a direction 180 away from the prin-
cipal beam.
The elements are built up from the fundamental dipole shown in
Fig. 9.27. Each consists of two such dipoles having their wings bent into
arcs of circles and joined to form a unit. Like the axially symmetrical
tndipole units these elements are simply slipped over the outer conductor
of a coaxial line, and they also can be represented by a shunt impedance.
Two conditions must be satisfied if an end-fire array of this type is to

1H. J. Riblet and B. L. Birchard, End-6reArray Antenna, RL Report No. 577,

July 11, 1944. Dielectric-rodantennasmay be designedto have end-firepatterns
with gain, beamwidth, and side-lobe propertiesas good as those of linear arrays.
C!. C. E, Mueller,The DielectricAntennaor Polyrod, BTL Report No, 251, Jan.26,
1942; J. E. Eaton, Dielectric Rod End-fire Antennas Close to Metal Surfaces, RL
Report No. 969, Jan. 23, 1946; R. E. Dillon and L. J. Eyges, Compact Horns Inter-
mediate between Polyrods and Reflectors, RL Report No. 961, Jan. 31, 1946.
 SEC.9.18] END-FIRE ARRAY 317

have maximum gain. (1) There is an optimum value for the wavelength
~, of the coaxial line. It was shown in Sec. 9.9 that maximum gain for
quarter-wavelength-spaced end-fire arrays with a constant phase delay
occurred when the phase delay between adjacent elements was

2.94.
;+
n

The total phase delay between the first and last elements is then approxi-
mately

*= T;+l. (68)
()

If L = nk/4 is taken as the length of the array, Eq. (68) becomes ()

or
:. 1
L
=:+; b 9j ,>
G
;
(2) The attenuation has a definite optimum value; it must be neithe~ so
large that most of the power is radiated from the first few elements Qr .;
so small that an excessive amount of power is lost in the dummy 1& d. ~
This optimum attenuation is ordinarily assumed to be that which all~wa .,
from 5 to 10 per cent of the total power to be absorbed in the du~rny ~
load. \. A
The desired attenuation and phase shift can be obtained in prinw --
in a very simple way. From Sec. 917 we have seen that periodic loading
of a transmission line changes the propagation constant of the line.
Hence it should be possible to choose the impedance of individual radi-
ators so that they cause just the right change in attenuation and phase
velocity. In fact, if the impedances of the elements are known as a func-
t
tion of several parameters, the propagation constant can be calculated
from Eq, (62) as a function of the parameters and the best value chosen.
In the design of the particular arrays described above, the impedances
 .{

of the individual elements ~verenot known in enough detail to alloli- this.

Hence a different approach was used. The gain Tvas measured as a func-
tion of probe depth for various lengths of the dipole ~vings. This gave
the t~vo parameters necessary to adjust, for the correct phase ve]ocii y and
attenuation. The gain of the 11 .7-rm array finally obtained in this
manner \~asaround 15.4 dbj slightly greater than thr theoretical value of
15,2 db for such an array; the theoretical value, ho~vcvm-, is based on
isotropic radiators. The gain of the 10.7-cm array, which used

0 S@ 70 60

lIU. 944.-E-plane
D
1

p:,ttern
I 1
40 30 20 10 0
Angle
B

10 20 30 40 50 60 70 80 90

of an 18-4ement end-fire array, A = 11.7. A is the pat-

tern in tile clirectlun of the r]!ain lube, :LIIdB is the pattern i)] the direction of the back
lube.

11.7-cm elements, was 14,8 db. The E-plane pattern of the 11,7-cm
antenna is shown in Fig. 9-N; the H-plane pattern differs from it only in
minor details.
.

BROADSIDE ARRAYS
9.19. Suppression of Extraneous Major Lobes.The majority of the
applications of microwave arrays have called for a beam having the
principal maximum in a direction normal to or nearly normal to the axis
of the array. &rays of this type will be referred to as broadside arrays
\vith the arbitrary limit on the classification that the principal maximum
lies within 25 of the normal to the array. In general there must be no
principal maximum other than that of the broadside lobe, that is, all
other maxima must be in the form of side lobes at considerably lower
levels. This requirement gives rise to a spacing and phase problem
 SEC. 919] SUPPRESSION OF EXTRA,VEO US MAJOR LOBES 319

common to all arrays of this type. It was seen in Sec. 9.5 that the ele-
ments of a uniform array must all be in phase for an accurately normal
main lobe while to produce an off-normal lobe [cj. Eq. (23)] there must be
a small progressive phase delay. If there are to be no other major lobes,
the spacing between isotropic radiators must be somewhat less than A,
the free-space wavelength. The exact amount depends on n and the
acceptable side-lobe level; no portion of an accurately normal main beam
will be repeated in the direction t? = O if s = (1 I/n)X. To produce
uniform phase, the radiators must be spaced at intervals of &j the guide
wavelength. However, for all the air-filled microwave lines discussed
in Chap. 7, it was found that & ~ A with the result that the spacing
exceeds the limit stated above.
There are various techniques for circumventing the difficulty. The
less-than-wavelength spacing limit applies strictlv to an arra.v of isotropic
radiators. However, in Sec. 9.2 it
was shown that the pattern of an array II
is a product of an array factor corre-
sponding to the pattern of an array
of isotropic radiators and the pattern
of an individual radiator. If the
latter pattern is made sufficiently dire-
. tive with a maximum in the direc-
tion normal to the array, a principal FIG. 9.45,Array of transverse slots
with horns to eliminate end-fire lobes.
maximum will occur only in the
region where the array factor and the radiator pattern simultaneously
have appreciable values. In this case the spacing can exceed A without
the appearance of extraneous major lobes. Illustrative of such a direc-
tive device is a horn fed by a slot; an array of this type is shown sche-
matically in Fig. 9.45.
A procedure that suggests itself immediately is to shorten the guide
wavelength to a value below the allowed spacing limit. The methods that
have been used to do this are described here because they hare been gener-
ally unsatisfactory. The simplest technique is to fill the guide with
dielectric and thus reduce the guide wavelength. However, the use of
dielectrics gives rise to a number of problems: the loss, particularly in long
arrays, results in diminution of the gain; it is difficult to maintain proper
contact between the guide walls and the dielectric, with the result that
electrical break do\vn tends to occur and with it reduction in the power-
handling capacity of the array; and also of no small significance is the
increase in the vieight of the antenna. Another method that has been
tried is that of using a corrugated line, With coaxial line the inner
conductor is corrugated as shoum in Fig. 94(;a, \\-bilewith rectangular
guide one of the broad faces is replaced bv a corrugated wall as shown in
Fig. 946b. The systems can be thought of as a transmission line loaded

.

320 LINEAR-ARRAY ANTENNAS AND FEEDS [SBC. 919

periodically with reactance. The wavelength in the loaded linel has been
found to be given approximately for coaxial line by

f -+--=FP -
L-
(a)

El a
a
x

y+-1-J
dy
L
.-- --- _________

(b)
~IG. 946.-Corrugated lines for shortening Xti:(a) coaxial line; (b) waveguide.

and for rectangular guide by

where ~, is a solution of

These lines have proved impractical for the same general reasons as the
dielectric-filled line: There is a significant increase in weight and great
reduction in poww-handling capacity, and in addition the corrugated
sections are difficult to manufacture. Some of the difficulties, however,
are due to the high percentage reduction in wavelength that is being
I H. Goldstein, The Theory of Corrugated Transmission Lines and Waveguides,
RL Report No. 494, Apr. 3, 1944.
 J.i
SEC. 920] RESONANT ARRAYS 321

etfected. The corrugated line has been used with more success in other
antenna designs where only a small wavelength reduction was attempted.
The most successful technique that has been developed is in the design
of radiators whose phase can be shifted lSOO by simple structural
changes; the elements can then be spaced at intervals of XO/2 and brought
into phase by the structural phase reversal. Since the guide wavelength
is generally in the range A S Aa S 1.5A, this spacing is acceptable. The
procedure is also satisfactory from the point of view of the loaded-line
analysis. If the elements are pure series or shunt elements, the propaga-
tion constant is unaffected by the loading and a uniformly illuminated
array results. The phase reversal does not alter the impedance presented
by the radiator to the line. A brief summary of the phase-reversal tech-
niques for the various types of elements discussed previously is given
below:

1. Slotted dipole, Fig. 9.47a, the dipole is rotated through 180 about
the coupling probe.
2. Tridipole radiator, same as for the slotted dipole.
3. Shunt slots in brood face of rectangular guide, Fig. 9.47b, the slots
are placed on alternate sides of the axis of the guide.
4. Shunt inclined slots on the narrow face, Fig. 9.47c, the inclination of
alternate slots is reversed.
5. Probe-fed slots, coupling probe is placed on opposite sides in alter-
nate slots, or the orientation of the probe in the guide is reversed
(see Figs. 9.23 and 9.24).
6. Longitudinally polarized waveguide radiator, the length of alternate
slots differs by x~/2.
7. Transversely polarized waveguide radiator, elements are staggered
with respect to the guide axis just like theshunt slots in Fig.-~.47b.

9.20. Resonant Array s.Broadside arrays can be divided into two

general classes: resonant and nonresonant arrays. The resonant type
yields an accurately normal beam and is well matched at the design fre-
quency; the impedance match, however, deteriorates rapidly with depar-
ture from the design frequency, and the array can be used only over a
very narrow frequency band. An array of this type consists of a number
of single series or shunt elements, spaced a guide half-wavelength apart
on waveguide or coaxial line, with successive elements mechanically
reversed in their feeding to give the phase reversal discussed in the
preceding section. The resonant array is uniformly illuminated, since,
4 as Eqs. (66) and (67) show, there is no attenuation in a line loaded with
half-wavelength-spaced single series or shunt elements. The fact that
uniform illumination is produced has been verified experimentally by
measurements of the radiation directly in front of the array with a small
 .(

322 LINEAIL-ARRA 1 ANTENNAS AND FEELAS [SEC. 920

exploring horn. Furthermore, the secondary patterns of these arrays

are in agreement with the patterns of an array of uniformly excited ele-
ments. The uniform illumination is an advantageous feature where the

t+ ---1

d Itt+++lhl+l
(a)

!
I
FIG. Q.47.Phase-reversal technique: (a) phase reversal of dipoles; (b) phase reversal
of longitudinal slots in the broad face of rectangular waveguide; (c) phase reversal of inclined
slots in the narrow face of rectangular waveguide.

prime requirement is high gain; on the other hand the array is unsatis-
factory when side lobes are the major consideration, since the first side
lobe is over 4 per cent of the peak intensity.
The impedance match of the array is obtained by choosing the imped-
;
ances of the elements properly and by adjusting a short-circuiting plunger
 i

:il~c. 920] RESONA.VI ARRAYS 323

it the end of the array. The short-circuit termination is a characteristic

feature of broadside arrays; the reflected wave causes no difficulties such
as would arise in end-fire arrays; for since the elements are half wave-
lengths apart, the radiation pattern due to the reflected wave is again a
normal beam. The well-matched condition on the design frequency and !
the narrow bandwidth property of the array will be discussed for n series
elements; the argument, phrased in terms of admittances, is similar for i
shunt elements. We assume that the impedance of each element has
been adjusted to Zo/n, where Z, is the characteristic impedance of the
# line. The line is terminated in a short circuit at a distance x,/2 from the
last element. Since the spacing is ~g/2, the entire array is equivalent
to n elements in series. The input impedance is therefore n(ZO/n) = 2,;
that is, the array is matched.
When the exciting frequency is not the design frequency, the elements
are no longer exactly a half wavelength apart. Then the impedances
do not add up to 2,, and the array is not matched. The mismatch for
frequencies off resonance cannot be calculated unless the frequency varia-
tion of the impedances of the elements is known. Their variation can often
be neglected over the bands in \vhich one is interested. A simple
graphical analysis can then be carried out on an impedance chart. Let
us take for example a 10-element 10/2-spaced array of series elements,
each of resistance 0.12., and plot on an impedance chart, starting from
the terminal short circuit, the input impedance seen looking to the right
from a point just to the left of each successive element. At the design
frequency these points fall along the R/Z,-axis as indicated on the line S
in Fig. 948.
Suppose, for example, that the wavelength decreases by I per cent.
The spacing between elements is now greater than As/2. The short
circuit now presents a small positi~-e reactance in series with the tenth
element. As one proceeds from element 10 to 9, the path traversed is
greater than XQ/2 so that the reacti\e component increases more than for
the resonant wavelength. With each transformation to the next element
there is an increase in reactive component due to the excess of the path
over ~g/2, with the result that the input impedance to the array as a whole
has an appreciable reactive element. The transformation is shown as
line S in Fig. 9.48. The frequency sensiti~ity is evidently greater the
longer the line. Common practice has been to limit the length of the
array to 15 wavelengths, because longer arrays have been found to be
too frequency-sensitive.
There are additional frequency-sensitive characteristics that should
be noted. (1) Because the spacing is no longer equal to k~/2 for fre-
quencies off resonance, attenuation sets in [c~. Eq. (67)] and the array is
not uniformly illuminated. (2) The beam is no longer accurately normal
to the array. These effects are gerwra!ly less important than the imped-
324 LINEAR-ARRAY ANTENNAS AND FEEDS [SEC, 920

ante sensitivity because they are relatively insignificant for the narrow
band over which the impedance match is acceptable.
The impedance characteristics of a resonant array can be improved
by a process of overloading the line, i.e., using elements with imped-
ances greater than ZO/n. Of course, the array is then not matched, and

1.0

:
L 10

-0.5
+0
t-

-1.01 / 1X1
0 0.5 1.0 1.5 2.0 2.5
FIG. 9.48.Input impedance of IO-element resonant array.

a matching transformer must be used. However, the combination of

overloaded elements and transformer will generally have a broader band
than the array matched by itself. The theory is best shown by example.
Let us take again a 10-element k,/2-spaced array with series elements.
Suppose now that the resistances of the elements are 0.220 and, for
definiteness, that the array is matched by a tuning screw. As before,
the line 10, 9, 8, . . . in Fig. 9.49 represents the input impedances to
successive elements for the frequency at which the spacing is As/2, and
1 r~presents the input impedance to the array as a whole.
The array is matched at the design frequency by traveling clockwise
on a constant VSWR circle to point P on the unity R/ZO line and by
inserting a tuning screw there to transform to the point Q where Z = 2..
If the wavelength is again assumed to decrease by 1 per cent, the imped-
ance to the array is given by 1. This is quite close to point 1. If this
impedance is transformed to the scre~v, it falls on the point P and
 i
SEC. 9.20] RESONANT ARRAYS 325

the input impedance to the array is then Q, which is quite close to 2..
Thus the array is still fairly well matched. In practice this method is
quite successful, sometimes to the extent of doublipg or tripling the
bandwidth.
1.0
90
.-

, 0.5 ~.
60
lx
/ I

mti --

FIG. 9.49.Input impedance of overloaded resonant array.

As an example of the performance of resonant arrays one antenna of

this type will be discussed in detail. 1 Figure 9.50 shows an axially sym-
metrical array for transverse polarization designed for the 3-cm band,

FIG. 9.50.An axially symmetrical array for transverse polarization.

The elements consist of the axially symmetrical units of slot radiators

shown in Fig. 9.30. The line is a circular waveguide having an outer
diameter of l+ in. and supporting the ZJf,l-mode. It is fed by the con-
verter shown in Fig. 933. The distance from slot to slot along the axis
4 of the guide is &/2. Phase reversal of the slots is achieved by putting

1H. .J. Rihlet, Horizontally Polarized Non-directional Antennas, RL Report

>~o.489, Apr. 22, 1944.
 326 LZATEA R-II RRA Y AN TE.V.$A,9 A,VD FEEDS [SEC. 920
20

. , -.

--
=15 .
,7
:

%2-- .

10
3180 3187 3197 3207 3214
Waelm@ cm,

~lG. 9.51.Frequency sensitivity of axially symmetrical transversely polarized array.

:,

,,

Elevation,
dq
l,1<;, 9,52. >leri(li<>]]:Ll pattenls of an ~xially sy!r],]letri<al trar, sverw]y pu]:trized array
 SEC. 921] BEA(!ON A,VTB,V.VA SYSTEM, S 327

the exciting screws on one side of a given slot and on the opposi ~e side
of the next slot. Between each bay of slots there is shown another set
of screws. It was found that an array without these screws \vas exces-
sively frequency-sensitive, The screws partially cancel the reflected

waves from each bay of slots and hence increase the band\vidth of the
array. The input VSTVR to this
array is shown in Figs. 951 and

,
952 shows the meridional pattern.
The beamwidth is about 4 .5; the
theoretical width calculated from
Eq. (14) with s = .870A is 4,9.
Recei
l=!
-1l.

The first side lobes are about 4* pcr ante

cent, a value expected for uniform
illumination. The asymmetry in the
pattern is due to spurious reflections
from objects surrounding the pat-
tern-measuring equipment.
9.21. Beacon Antenna Systems.
In beacon systems the responder Transmi
(receiver) and transponder (trans- anten
mitter) are ordinarily on t~\-odiffer-
ent frequencies. This necessitates
two different antennas, one for trans-
mitting and one for receiving,
These t\vo antennas must bc so
arranged that there is no cross-
talk between them; i.e., ~-ery little
energy from the transmitter is pick-
ed up directly by the receiver.
Actually a little is always picked
~lG. 953. -&aeon antenna with an exter-
up, but in satisfactory antennas it nal feed hne,
is at least 40 db down.
The ordinary way of arranging a transmitter and receiver is to place
one directly above and on the same axis as the other. The major prob-
lem is then to feed the upper antenna. This has been solved in tlvo
different ~vays. First, an external feed line can be used. Such an arrange-
ment is shown in Fig. 953. The transmitter and receiver of this beacon
antenna are resonant arrtiys of slot-type axially symmetrical radiators.
The external feed naturally has an effect on t}lc azimuth pattern of the
bottom antennti. This effect is relatively small and not intolerable.
4 It usually takes the form of superimposing a series of sharp maxima and
minima On the or(lillar~ azi]ll(ltll Piltte]n.
For some IIsrs, particularly for airborne beacons, an external feed is
so }Julky an(l clllrnsy that an alternati!c design is used. It is applicable
328 LINEAR-ARRAY ANTENNAS AND FEEDS [SEC. 922

only when the antennas are built on coaxial line. In this design the
inner conductor for the bottom antenna is made hollow and another
conductor runs inside it, forming a coaxial feed line for the upper antenna.
This inner inner conductor is then tapered to normal size as it enters
the upper antenna.
Such double antenna systems have been built at both 3 and 10 cm,
and almost all the coaxially-fed axially symmetrical radiators previously
dk,cussed have been used. Figure 9.54 shows such an antenna for 3

FIC. 954.-Double antenna system, X = 3 cm.

cm. The elements are axially symmetrical radiators consisting of slots

on a coaxial line having a 1 in. OD. The rectangular waveguides, A and
B feed the coaxial lines for the top and bottom antennas respectively.
C is the hollow tube that serves as inner conductor for the bottom antenna
and as the outer conductor for the coaxial line feeding the upper antenna.
D is the inner conductor of the latter coaxial line. E is a tapered section
of coaxial line.
9.22. Nonresonant Array s.The nonresonant broadside array may
consist of a number of elements spaced a little more or a little less than
AO/2 apart. Consequently the beam is not normal to the arrhy but at an
angle given by Eq. (23). This may be a disadvantage in some applica-
tions. The advantage of this type of array is that its impedance match
is generally good. Because the elements are not k,/2 apart, reflections
from later elements tend to cancel reflections from earlier ones so that the
array remains matched over a much wider band than the resonant array
of the same length.
Although the nonresonant array eliminates the matching problem
inherent in the resonant array, it presents an illumination problem that
the resonant array does not have. The elements of the nonresonant
array are not equally excited as in the resonant array; less po~ver reaches
the later elements; and if the elements are all alike, an exponential illu-
mination results. Such illumination is undesirable because it reduces
L. J. Eyges, Omnidirectional Antennas for 13UPX, RL Report No. 996, Jan. 17,
1946.
 SEC. $22] NONRESONANT ARRAYS 329

the gain. There area number of ways tocontrol theillumination andin

particular to make the last elements radiate as much as the first. One
method is to vary the elements themselves so that the first elements take
out small fractions of the power incident on them while the later ele-
ments couple out larger and larger fractions. Thus if the elements are
dipoles, successive dipoles can have deeper probes; if they are inclined
slots on the narrow side of the guide, the inclination can be increased
with distance along the array. Another useful and advantageous
method permits the elements to be all identical. This consists in taper-
,

h
b

Fm. 9.55.Transversely polarized array of waveguide radiators.

ing the guide in its narrow dimension so that it is smaller toward the end
of the array. If there were no radiating elements, this would mean that
the energy density would become larger toward the end of the array
because a given amount of energy would be flowing through a smaller and
smaller area. When there are radiating elements, the taper can be made
to compensate for the loss of energy, thus maintaining a constant energy
density in the guide.
In such an array there must be no appreciable wave reflected from
the end. If the original wave radiates a beam at an angle 19to the normal,
the reflected wave will radiate an undesirable lobe at an angle 0. To
avoid this lobe the array is usually terminated in a matched load. This
may be a dissipative load, and ordinarily arrays are designed so that about
5 per cent of the total power gets beyond the last element and is dissi-
pated as heat. To avoid this waste of power, a matched load can be
made of one of the radiators backed by a short circuit and matched with
an iris. With this on the end of the array there is no reflected wave and
all the energy is radiated.
r
Shown in Fig. 9.55 is a section of a nonresonant array, built for the
l-cm region. The elements are the transversely polarized waveguide
radiators shown in Fig. 9.25b. The wall of the guide in which they are
 ;<

330 LINEAR-A RRA 1 ANTENNAS AND FEEDS [SEC. 922

cut is a quarter wavelength thick. The elements are spaced nearly ~

apart; and for phase reversal, alternate elements are staggered with
respect to the center line. The whole array feeds into parallel plates that
flare out to the proper size for beam shaping in the other plane. The
guide is tapered for uniform illumination. The coupling formula for the
fraction of incident power abstracted by each element is

I-sl I .1

II
.. ___ __

T
a

___
il
___ __ J
l--a<
FIG, 956.-Geometrical parameters in Eq. (70)

where P/PO is the average fraction of power abstracted per radiator and
Xo, X;, and ~0 are respectively the guide wavelengths in the main guide,
the branching guide, and the parallel plates. The geometrical parameters
are defined in Fig. 9.56. The physical length 1 of the branching guides
must be chosen so that its effective electrical length is A~/4. An approxi-
mate formula for 1 is

=$-:(+$)
Equation (70) has not been checked directly, but arrays based on it
have been built, and their performance \vas almost that expected.
Another type of nonresonant array has been designed that has a
normal or closely normal beam, like the resonant array, but is much more
broadband in impedance. Like the resonant array it has its element
spaced at half-wavelength intervals. In order that the array be matched,
1 For a derivation of this formula see V. Sirhak and l?. 11. Purcell, 1(70scc?
Antennas with a Line Source and Shaped ~ylindrical Reflector, RL Report No. 624,
Nov. 3, 1944, pp. 7-13.
 SEC. 9.23] BIWADBA,VD SYSTEMS 1!ITIJ .VOR.UAL 13EA.IIS 331

each element is matched to the guide; i.e., it is nonreflecting \vhen termi-

nated in the characteristic impedance of the guide.
There are a number of different ways to realize such a matched ele-
ment. One obvious method is to match each element individually by a
tuning screw or iris in front of it. This is always theoretically possible,
but in practice it may be difficult; because the transmission-line equations
are not valid close to the radiating elements, it is not al\vays easy to find the
proper size and position of the iris or screw to match the elements. This
difficulty is avoided by the use of inclined displaced slots. It is possible
, to choose the length, displacement, and inclination of these slots so that
they present an input conductance of unity, shunted by a susceptance.

!3
4.37+
FIG, 957.-Section of nonresonantarraywithspacing(X,/2)(k = 1.25CIT,).

A tuning screw placed at the center of the slot will match it. Another
type of element that is matched without tuning screws or irises has been
built for the l-cm band. It combines features of the waveguide radiators
and inclined displaced slots in that it consists of asymmetrical inclined
slots cut through the quarter-wavelength thickness of the broad wall of
the waveguide. Successive slots are set on opposite sides of the center
of the guide, and succeeding slots run together. The exact dimensions
of these slots had to be determined experimentally. Figure 9.57 shows
a sketch of these slots. 1
It is obvious that an element which has an input impedance of 20
when terminated in 20 cannot be either a simple series or a simple shunt
element; it must be represented by some T- or n-network. Thus the
waveguide is equivalent to a line loaded with T- or II-networks, and (see
Sec. 9. 17) there is attenuation in such a line, For uniform illumination
some device must be used to enable the later elements to abstract as
much pou-er as the first. This can be done by increasing the coupling
of later elements or by tapering the guide,
9.23. Broadband Systems with Normal Beams.The various arrays
we have discussed thus far have one feature in common: The direction of
the beam is a function of frequency. Whether the beam is normal for
the design frequency as in the resonant and second type of nonresonant
arrays or is not normal as in the first type of nonresonant array, the beam

1 J. Steinbcrgrr and F.. Il. ~hisbolrn, Linear Array, RL RePort No, 77I, Ja,l 31
j 1946.
332 LINEAR-A RRA Y ANTENNAS AND FEEDS [SEC. 9.23

angle shifts when the frequency changes. This feature is disadvantageous

for many applications. This section treats systems of arrays that have
the two properties of constant beam angle and broadband impedance
match.
The one feature common to all the array systems discussed in this
section that causes the beam to remain normal over a band is that they
are excited in the center. Such a system can be considered as two end-
fed component arrays. These two arrays are arranged so that at the
design frequency their component patterns add up exactly to give a result-
ant normal pattern. Wheh the frequency changes, the beams from the
individual arrays move in opposite directions; the resultant beam is
still normal to the array. Of course, the resultant beam broadens some-
what and, if the frequency changes excessively, begins to split, but it
remains normal to the array system. Three different arrays of this type
have been built; the differences among them lie in the methods of obtain-
ing a broadband impedance match.
One consists of two nonresonant arrays each with a beam at an angle
8 to the array. 1 The two arrays are arranged in a V of angle 180 2tl,
and the power is applied at the vertex of the V. Generally a parallel-
plate waveguide is placed in front of the array to give a satisfactory pat-
tern in the other plane. The impedance properties of such an array are
very similar to those for a single nonresonant array, and it remains
matched over a broad frequency band. The main disadvantage of this
array is that it is not linear. The V-shape and the flaps on the parallel.
plate section make it clumsy and heavy.
The disadvantages of size and weight are eliminated in the second
example of broadband array with a normal beam. This array consists
also of two component arrays excited in the center, but these are of the
second type of nonresonant array. 2 Since the beam of each component
array is normal to [t, the two arrays can be placed in a straight line.
Thus, the major disadvantage of the clumsiness of the V-shape is over-
come, but there is a neti disadvantage in that the component arrays are
more complicated.
The third example of array, like the other two, consists of two com-
ponents excited in the center. In this arrav, each component is a
resonant array. 3 The broadband impedance match is obtained by dis-
placing one array with respect to the other until the reflections from th~
components cancel each other. Such a system is illustrated in Fig. 9.58
and is made of two arrays, I and II, with identical spacing and phasing;

1J, R. Risser et al., Linear Array for Use in the AN/APS-23 Antenna, RL Report
No. 973, Mar. 19, 1946, pp. 17.
2J. Steinbergerand E. B. ~hisholm, Linear Array, RL Report No. 771, Jan. 31,
1946.
3 Risser et al., op. cit., pp. 713.
.

Sm. 9.23] BROADBAND SYSTEMS WITH NORMAL BEAMA 333

but array I is placed a distance A ahead of the other, and the distance z
to the first element of array I is different from d, the corresponding
distance for array II. There are two requirements for a satisfactory
pattern and a broadband impedance match. First, for complete can-
cellation of the reflected waves at any point P in the main guide, the length
from P to the first element of array I must be A,/4 longer than the cor-
responding distance for array II; then the reflected waves from the two
,t
I
<
A 1 . _B

t 1
II
P

FIG, 9.58. - Scl)elrlatic of brozdlnmd normal-firing resonant array system.

arrays \vill be just a half ~vavclength out of phase and \vill cancel. The
condition for this hO/4 path difference is

~d+A=& (71)
4

For a satisfactory radiation pattern there is another condition. The line

Al? must be a line of constant phase. For generality suppose that array
II feeds into some sort of parallel-plate system in which the wavelength
X is not necessarily the free-space wavelength. lhe condition for equi-
.
phase along All is then

(72)

A simultaneous solution of Eqs. (71) and (72) is A = A/4 and

4
 CHAPTER 10

WAVEGUIDE AND HORN FEEDS

BY J. R. RISSER

10c1. Radiation from Waveguide of Arbitrary Cross Section.The

problem of radiation from the open end of a waveguide could be dis-
cussed in principle from several points of view. Rigorously, the radiation
can be considered to arise from the current distribution on the inside
walls of the guide, which is just the current distribution associated with
the fields propagated in the interior of the guide, together with the cur-
rents flowing from the open end out upon the exterior guide surface.
Were it not for difficulties in the analysis, this current distribution and the
radiation field at an external point co~dd be calculated. This has, how-
ever, not yet been accomplished. On the other hand, the approximate
methods of diffraction theory developed in Sees. 5.11 and 5.12 have been
applied to the problem with some degree of success. I The guide opening
is presumed to act like a hole or aperture in an infinite screen, the trans-
verse fields in the aperture being assumed to be identical with those in a
parallel cross section inside the guide. The vector Huygens principle is
applied to obtain the radiation field from the aperture field distribution
as discussed in Sees. 5.11 and 5.12.
In all important practical cases the guide allows propagation of only
one mode, called the dominant mode. Over a cross section inside the
guide sufficiently far from the aperture, any component of the field is
the vector sum of the components associated with incident and reflected
waves of the dominant mode. In the aperture, however, additional
higher-mode fields exist locally, excited by the discontinuity in the guide.
It is not possible to determine the details of the higher-mode field distri-
bution empirically; they can be obtained only from a rigorous solution of
the boundary problem. The contribution of the higher-mode fields are
neglected in the approximate diffraction theory used in this chapter.
This is one source of inaccuracy in the method.
The effects of the reflected dominant mode \vave can, ho~vever, be
taken into account. They are expressible in terms of a reflection coeffi-
cient r which can be determined empirically by standing-wave measure-
ments in the guide. The reflection coefficient r is the ratio (Ei)./(Ei)i
of the transverse components of the reflected and incident ~lectric field
1L, J. ~hu, Calculat,lon of the Radiation Properties of Hollow Pipes and Horns,
Jour. Applied Phys., 11, 603-610 (1940).
334
SEC. 101] JVA JIWI IDE OF A I(BI TRA RI CROSS ,5ECTION 335

vectors; it varies in phase but not in magnitude along the guide. When
extrapolated to the plane of the aperture, r can be interpreted in terms of
an equivalent circuit admittance q for the aperture by the relation
~=lq
(1)
l+q;
the admittance q is normalized to the characteristic wave admittance of
the guide. The characteristics of q and r \vill be discussed later (Sec.
10.1 l). It is assumed here that r is a knonm quantity. The total trans-
verse electric field Et of the dominant mode in the aperture is then given
by
E, = (1 + r)(E,),, (2)

where (Et), is the transverse electric field of the incidmt dominant-mode

wave. The transverse magnetic field H, can be obtained from Et using
Eqs. (7.33b) and (i33c) :

(H,), = t[i, x (E,),],

(3)
(H,). = t[i, x (E,),], )
where
t=:B for TE-modes
UP

(de
for T~lI-modes.
P,.n
Therefore H, can be \vritt,en

H, = t(l r)[i. x (E,),]

. t ~ [i, x E,]. (4)

()
The relation between the electric and magnetic fields over the aperture
is thus of the form of Eq. (5,104) with the constant a = t(l I)/(] + r),
It should be kept in mind that the value of r is not altered by the inser-
tion of a matching transformer in the guide because the refiected wave
still exists in the region bet~vmm the transformer and the aperture.
To calculate the radiation field at a point Z outside the pipe, we sur-
round P by a closed surface containing the apertorc. This surface con-
sists of the aperture, the exterior surf am of the guide and the sphere at,
infinity. The vector Huygens principle is applied to this surface, .+s
in other diffraction problems the sphere at, infinity contrib~ltes nothing.
Over the exterior surface of the guide the electric field is necessarily normal
to the surface] rind thervfore E, is zero. flcre is, holvever, a tangrntia]
component of the magnetic field assoriate(l ivith cllrrents originating at
the aperture. As in the case of the higher modes in the aperture, inability
to solve the bollnrlary problem at the end of the }~tiveg~lide means that
336 WA VEGUIDE AND HORN FEEDS [SEC. 102

these currents are unknown, and H, is assumed to be zero on the guide

surface. This is a second source of error in the method.
The effect of neglecting the higher-mode fields in the aperture and the
tangential component of the magnetic field on the outside surface of the
guide depends on the dimensions of the aperture as measured in wave-
lengths. It is reasonable to assume that both factors contribute a smaller
fraction of the total radiation field as the aperture dimensions increase.
It is, in fact, the case that the calculated radiation field is in increasingly
better agreement with experiment as the aperture dimensions increase,
so that the limitations of the theory are apparent principally for small
apertures. Unfortunately the dimensions of waveguide actually used are
fractions of a wavelength. A more rigorous treatment of the problem
would be desirable.
By neglecting the higher modes and the current distribution over the
exterior surface of the waveguide, the problem is reduced to a simple
aperture problem. The radiation field is calculated by means of Eqs.
(5110) and (5110a). The trans-
x verse electric field E, appearing in
the latter is replaced in the pres-
ent case by the resultant electric
field Et of the dominant mode over
-2 the aperture. The latter, in turn,
is expressed in terms of the inci-
dent electric field by means of Eq.
/
Y i (2).
FIG. 10.1.Coordinate system used in dis- The coordinate system is
cussing radiation Irom open waveguide. shown in Fig. 10.1. Rectangular
coordinates (z,y) are used in the aperture, taken to be the plane z = O,
and spherical coordinates R, @and @ are used to locate the point P. From
Eqs. (5.llla) and (5.lllb) the components of E, become

(5)

where N is the vector

JA

(1 + r) / (E,),e~kfZ~L- ~+~~im
o~~~jdS. (6)
A
10.2. Radiation from Circular Waveguide.The radiation vector
N of Eq. (6) can be computed for waveguide of circular cross section using
the exrwession for the transverse field vector of the dominant mode given
 ~
i,
SEC, 102] liAl)lA TIO.V PIM.lf CIliCLrLAR WA VEG(iIDh 337

in Sec. 7.13. In computing N it is convenient to express (E,),, the

incident wave field in the aperture, in rectangular components.
Case 1. !IE-uxwes.-In this case the rectangular components of (E~)i
are 1

E z = ~+ [,lm_,(.mnP) sin (m 1)* + Jm+,(.n.p) sin (m+ 1)*],

(7)
E, = J* [Jm_,(Km,@)cm (m 1)* Jm+,(Kmnp)cm (m + 1)4].
I
Writing z = p cos ~, y = p sin ~, the expressions to be evaluated become

These are evaluated ~vith the help of the Bessel-Fourier series

e- -$) = JO(~P) + 2jJ.(Ap) Cos n(r$ *) (9)

z
n =,1
and the Lommel integral formula

(lo)
Using these together \vith the recurrence relations and recalling that
~~(.m.a) = O, the field components are obtained as follows:

(11)
~+ . jn+l kaw
_
2R [
h,.
~+cOsor
( 6:Cos o
)1
~.t(K,,,.a)~~(ka sin 0) ~os ,,l@e_,,n
~_ksin02
() %m 1
1The following recurrence relations arc neeclrd for this section:

.lm(z) = ; J,,,(z) J,n+, (z) = ; [J,?-,(z) .Tm+,

(z)]

; Jm(z) + J.,-,(z),

;Jm(z) = ; [J+,(z) + J-,(z)].

338 WA VEGUIDE A.VD HOILV YEEJ9S [sm. 102

Case 2. T~ll-waues.Follo\\ing the same procedure as above, the

integrals to be evaluated are found to be the same. Specifically the
integrals in the two cases are related as follows:

I
(N=)..,, = ~ (N,),.;
(12)
(Nu) . .. = & (N,),..

It w-ill be recalled that the characteristic values of Km. for TM-waves are
obtained from the roots of ~fi(KnJZ) = O. On evaluating the field com-
ponents, it is found that due to this condition, E+ is zero and there is but
one component:

The TE,,-mode which has the lowest cutoff frequency is the one most
commonly used in circular-guide antenna feeds. The remainder of the
discussion will be confined to this mode. on setting m = 1 into Eq. (7)
it can be seen that the electric field over the aperture is symmetrical with
respect to the yz-plane, which is thus the E-plane of the system. Figure
10.2 taken from Chus paper shows the calculated E- and H-plane pat-
terns as a function of aperture. The effect of the reflected wave in the
pipe on the aperture distribution has been neglected (i.e., r has been set
equal to zero) in computing these patterns. Figure 10.3 shows a com-
parison between an observed pattern and the corresponding theoretical
pattern. The agreement is quite good considering the factors neglected
in the theory.
There are various measures of the sharpness of the beam. One cri-
terion that has been used in the literature is the angle from zero to zero
inckiding the main beam. In the E-plane (O = r/2), Ee is zero when
ka sin 0 = 3.83. The beam angles in the A- and H-planes are then

8E = 2 sinl
()3.83A

27ra
5.33A
(14)
8. = z sinl
( 2~a )1
The beam is thus sharper in the E-plane than in the H-plane. Equation
(14) is, of course, meaningless for the E-plane when 2a/X <1.22 and for
1~bi~is a good apprcjxinmtion for circular guide. For standard Radiation Labora-
tory waveguide (2a = 0.75k) r is found to be small.
 \
l
,,
/
(
340 WA VEGUIDE AND HORN FEEDS [SEC. 102

the H-plane when 2a/A < 1.7. From apractical standpoint, more useful
measures of the beam sharpness are the full angular widths between half-
power points and tenth-power points. For values of ~/a < 1, the half-

3cP 2LY 10 @lo 300 2@ NY 01,0

[ Theoretical\ ~~

~oo
0.9

0.8

0.7
,Ok-mo 0.8

0.7

~oo 0.6 500 0.6

0.5 0.5

&JO 0.4 600 0.4

0.3 0.3
700 700
0.2 0.2

800 800
0.1

90 S@

1w l~o

1100 12OO 131Y 14031601S00 110 1200 13tY140 160180

(d (b)
FIG. 10.3.Theoretical and observed radiation patterns from waveguide of circular cross
section; h = 3.2 cm. (n) E-plane; (b) H-plane.

power and tenth-power widths in the principal planes are given in degrees
by

@.= 14.7:; @H = 18.6:;

(15)
~, & = 25.0 ); @. ~ = 32,3:.
() () 1

Another characteristic of interest is the gain relative to an isotropic

source. It is given by

G = 47r +)
t

where PI is the total power radiated and P(O,O) is the maximum power
radiated per unit solid angle, which is in the direction o = @ = O. This
power is
5k2a2@2U21+{+r l~ 2 J~(Kl,a). (16)
P(o,o) = & :
() ( )1
I

I
SEC. 103] RADIATION FROM RECTANGULAR GUIDE 341

To find the total power radiated, the Poynting vector ~ Re (E, x H~ ) is

integrated over the aperture. This is evaluated as follows:

inserting the values of E, and EV from Eq. (7), we have for the TE1l-mode,

The last integral is evaluated by means of a Lommel formula resulting in

Making use of the recurrence relations and the boundary condition

~~(Klla) = O, we obtain finally

(17)

The gain is, therefore,

2

G=
lc%l+:+r

4.775P(1
( l~

lr12)
)1

(18)

~vhere the value of Klla = 1.841 has been inserted. For the region far
enough away from cutoff, r = O, P/k = 1, the gain is approximately

area of aperture
G = 10.5 (19)
( A= )

10.3. Radiation from Rectangular Guide.The tangential field com-

ponents of the dominant mode in the aperture of rectangular guide are
obtained from Eq. (7.74) or (7.79) by placing z equal to zero. Then,
in the same manner as for circular guide, the radiation vector N is calcu-
lated from Eq. (6).

1 G. N. Watson, Bessel Functions, 2d cd., lIacn~illan, Xew l-ork, 1945, p. 135,

Eq, (11).
z L, J, Chu, Jot{r. .4pp11ed Phys., 11, 603610 (1940).
 7

342 WA VEGUIDE AND HORN FEEDS [sMC. 103

Case 1. TE-modes.The components N. and N. are

N==j
To::.: r)rcOs@)ekmdd
b
nz~zti~(l + r)k sin 0 cos @
sin ~y eik~inoi ~ dy = #~b2
/ o b
1 ~i(ka,k, e.~ d+nm) 1 eI(kb sir.8.in~+nr)
x ~2r2 ]
~z sin2 O cos @ ~ k~sin20sin2 ~ ~
[ 1[ 1

NV = j
=:!:
rilsine)ekzBi8m~
[coS(?) ~~ku.in
Oain+ dy =

1 ei(ka.i. 9 m +-!-m.)
mmz~~(l + r)k sin O sin 4
kkna
1 e] (M .inOim++nm)
m2T2 n27r2
k2 sinz OCOS2+ ~ k sinz 0 sin2 @ ~
[ 1[ 1

The electric-field components of the radiation field are then

~i (rab)2 sin 8
Es= : 2~,Rk~n l+~cOsd+r IP~coso
() [ ( )1

[Rsin)-(~cosYlq~.
:()5
.?+= (rab)2 sin 0 sin o cos @
z~3R

[ cos+%+r(cose-%)lvfi(e~) I (20)
Vm.(o, dl)

1[
sin ~sin Ocosf$+~ sin

( )( )

I
[ ( +incos)-w esf:::i~(z) II
ej[k~-~sin b(acosd+bsin+ )-(rn+n+l)~] /

Case 2. TM-modes.The components N= and N. are related to those

of the TE-modes by

(N.),., = % (N.) T,,

1
(N.) TM = *
mbpu (N,).,.

As in the case of the TM-modes in a circular guide, the radiation field is

SEC, 103] R.4DI.4 TIO.V FROM RECTA AGULAR GUIDE 343

found to have only one component:

+r
( l~coso
n. )1
~lnn(e,o), (21)

while Ed = O.
The Ill-mode, m = 1, n = O, is of s~ecial interest, In this case the
radiation field reduces to

Jj kR~sin E(ac<,s @+ bsir, @)

1
(22)

(
7ra
Cos Sln
A
6
)
Cos @

(- ) ()
T(Z
slnocoso ~-
A

~-[+=- bwn +)
1

\vhere K,Ohas been replacecf by r/a. The phase factor

kII?
()
f sin /3 (a cos @ + b sin @)

can be simplifimf, It ~vill he rccallcd that in deriving the field expressions

the origin \ras taken at ~ corner of the guide. It, is easily found that if
the origin is shifted to the center of the aperture, the phase factor trans-
forms into l:R, 1/ nwv being measured from the new origin. In the case
of large alwrturcs r = O, so that the space factor is, therefore, real and
the guide is a dircctivc point-source frrd, the center of feed being the
center of the a!wrturr. In small aprrturm \\-herer is complex, there is no
exact center of feed; the yllide is only approximately a point source from
the point of vimv of th~: cqlliphasc surfaces of the radiation pattern.
The clwtric field ov(r the alwrturc is polarized in the Y-direction
so t}lat thr yz-l)l:ln(, is thr E-}ll:lnr of the system Ivhile the .rz-plane is
the If-plan(. ] ]1(, [}attrlrls ill t]l(+(> t\\-() plin(ip:l] planCs are
 T

344 WA VEGUIDE AND HORN FEEDS [SEC. 102

a. E-plane, @ = 7r/2

E~ =
2(:YR[1 +?

+r
(
l~~coso
k )1
( )
In
mh
sin 0
xb
A e-ikR. (23a)
sin 0
T
b. H-plane, @ = O

Ed =

+r(cs-?)~::~
e-
23b)
Y 4

It is observed that the predominant factors in the patterns

sin~sinO)l~sinO and
cOstfsinO)/[t+si OY-:l
are determined by the dimensions of the apertures in the respective planes.
It will be further observed that the E-plane pattern is essentially the pat-
tern due to uniformly illuminated slit of width b. The pattern in the
ILplane is essentially that due to a slit of width a over which the illumina-
tion is distributed sinusoidally as it is across the guide in the z-direction.
This is illustrative of a fairly general characteristic that the patterns in
the two principal planes are independent and are determined by the
aperture dimension and the distribution of illumination across the aper-
ture in the respective planes. The angular distances between the first
zeros on either side of the peak are given by

(24)

Figure 10.4 is a plot of E-plane and H-plane patterns of 3.2-cm wave-

guide calculated from Eqs. (23a) and (23 b), together with experimentally
observed values. Since the guide dimensions are appreciably smaller
than a wavelength (a/h = 0.71; b/A = 0.32), agreement would not be
expected to be particularly good in view of the approximations in the
theory. Better agreement would be expected with larger aperture
dimensions, although from a practical standpoint limitations on size of
aperture are imposed by the necessity of suppressing higher modes.
The predictions of Eqs. (23a) and (23b) for large apertures are, however,
[SEC. 103 RADIATION FROM RECTA NG[iLAR GUIDE 345

1.0
7
0.9

0.8
\
0.7 i
/
0.6
\
P(e) 0.5 h
c\
\
0.4

0.3

0.2 0
\.
\ )%7 LQ !:__
0.1

o
o 200
40 600 800
100 120 140 160 1800 20W

(:)
1.0

0.9
\\
0.8

0.7

0.6 n

P(e) 0.5

0.4
(>
0.3 ,
\
0.2 \
o
With reflection =
() coefficient
0.1 \: . -o
g-3- V
..
0 ~~- Ja- ~
00 200
40 60 80 1000 1200 140 1600 180
o
(b)
Fxc. 10.4. Theoretica1 and observed radiation patterns from rectangular waxw-
guide; X = 3.2 cm; ~= 0.71; ~ = 0.32; calculated patterns; o observed values; (a)
E-plane; (b) H-plane.
346 WA VEGUIDE AND HORN FEEDL3 [SEC. 103

of interest as a convenient means of predicting the radiation patterns of

sectoral horns, with small flare angles. Figure 10.5 is a graph of the
3- and 10-db-widths of the E- and H-plane patterns as a function of the
respective aperture dimensions. In computing the latter, 1?was taken
equal to zero.

1!

0 0.5 1.0 1.5 2.0

Wavelength/aperture
FIG. 10.5.Relation between the aperture dimension and the 3-db and 10-db widths of the r
radiation pattern of rectangular wave guide; E-plane; H-plane.

Finally the gain relative to an isotropic point solirce can be-calculated.

The power radiated per unit solid angle in the peak direction, 0 = + = O,
is

(070) (:)t+)l+%+r(
The total power radiated is obtained as in the case of circular guide by
-N
integrating the Poynting vector ~ Re (EL x H:) over the aperture. This F
integration is easy to carry through in the present case. We obtain

Pt=( 1 lrl*)@p~w310. (25)

w
 SF,C.104] W.4VEG[ IDE A .VTE,Y .VA FEEDS 347

The gain is, therefore,

G=4.~+=a,: iri,)&ll+y+r 1-+1$. (26)

()
For large apertures, (3/k = 1, r = f), the gain is approximately
area of the aperture
G = 10.2 (27)
( x )
10.4. Waveguide Antenna Feeds. Waveguide can be used satis-
factorily as an antenna feed, but only for very restricted applications.
It will be shovm in a later chapter that the po\rer radiated by the feed
should be do\vn approximately 10 db in the direction of the reflector
edge (cf. Chap. 12). This requirement determines the refiector shape
that can be used efficiently with ~~avcguide feeds. ?ior rectangular \~ave-
guide with b/A = 0.32 and a/k = 0.71, the reflector aperture should
subtend an angle at the feed of approximately 180 in the electric plane
and 120 in the magnetic plane (see Fig. 10.4). For circular \vaveguidc
with 2a/k = 0.75, these angles should be approximately 150 for the
electric plane and 140 for the magnetic plane, 11bile these figures are
necessarily approximate because the 10-db specification has some arbi-
trariness, reflectors of markedly different shape cannot be used ~vithout
sacrifice in gain or side lobes. In general, the reflector dimensions are
t determined by the application, and the feed aperture dimensions must be
selected accordingly. Flaring the terminal region of the guide to form
a simple rectangular or conical horn and placing beam-shaping obstacles
in the aperture of the feed constitute the usual solutions to the prot)lem.
Circular waveguide has found a more restricted application than rec-
tangular guide as an antenna feed; in fact its use has been confined to
conically scanning antennas. Since long lengths of circular guide are
found unsatisfactory, a circular-guide feed is generally excited from rec-
tangular guide through an intermediate tapered section. The feature
of circular guide that makes it suitable for conically scanning antennas is
that the terminal section can bc rotated without distortion of the mode
of propagation or rotation of the polarization of the radiated beam. If
a circular-guide feed is placed a small distance from the axis of a para-
boloidal reflector and is rotated about this axis, the peak of the beam from
the paraboloid will describe a cone \rhose axis coincides with the para-
boloid axis. The direction of polarization remains fixed in the course of
the rotation. The greatest care must be taken not to deform the guide
in bending, because deformations act as transformers converting plane to
elliptical polarization.
#
A number of structures have been developed to enable waveguide to
be used in rear feed systems. A ( rear feed is one that enters the para-
boloid at or near the vertex from behind and provides a means of deflect-
 ,

348 WA VEGIJIDE A Nl) HORN FEEDS [SEC. 105

ing the direction of propagation of the radiation so as to illuminate the

paraboloid. l~ith circular guide this is done by a reflecting disk (Fig.
10%), However, a study of the
Rmgsourceof rad(at,on
equiphase surfaces shows that
O
such a feed does not have a point
1 Fn
center of feed but behaves rather

,1 i-~
:-

I
like a ring source. It is, there-
fore, not suited foruse with are-
flector having a point focus.
/ In the case of rectangular
guide, a rear feed system can be
\
R / obtained by making a U-shaped
\ /
bend in the guide; this infeasible
at short wavelengths where bend-
ing the guide is mechanically feas-
ible and the added weight and feed
FIG. 106.- Circular waveguideand re- shadow are not important factors.
flectingdu+kstlowi]lgpositionof rin~source
of radiation]). When a more compact rear feed on
rectangular guide was needed,
modifications such as the t\vo- and four-dipole feeds (Sees. 8.10 and 8.11) ,4

(n- the double-slot feed were been used. The latter feed will be described
in the next section. f
A
10.5. The Double-slot Feed.

+!!!
One form of this type of feed is s,
shown in Fig, 10.7. Essentially =
the \vaveguide splits into two =
c
x-aveguide-like branches which
turn back and have their open
A
ends directed toward the parabo-
loid. One opening is above and
the other is beloir the input wave-
guide, \~hichis tapered to less than
normal height to decrease the
separation of the slots and con-
sequently the directivity of the
feed in the electric plane. As
shown in the figure, the two
branch paths are contained in a SectionAA
compact cylindrical head designed FIG. 10 .7.A double-slot feed.
for ease of manufacture. Each
branch consists of half the cylindrical cavity C and the v-aveguide-like slot *
S. The slots are pressurized by mica windows. The dimensions of the
feed \vere\vorked out empirically to obtain good match and pattern over a
6 per cent band (AA/AO = + 3 per cent) centered at 3.2 cm. over this
 SEC. 106] ELECTROMAGNETIC HOR.VS 349

band the VSWR is less than 1.3. The pattern is somewhat narrowerin ,!

the electric plane than in the magnetic plane, This feed is useful where I
a compact straight rear feed is needed.
10.6. Electromagnetic Horns.It appears at first sight that a radia-
tion pattern of any desired directivity can be obtained from a ~vaveguide
by a suitable choice of its dimensions. However, if the dimensions are
sufficiently large to allow free propagation of more than one mode, the
serious problem of controlling the modes arises. It is difficult to excite
a large-sized waveguide so that only a single mode is generated; if several
, modes are present, their relative phases at the aperture and hence the
resultant field over the latter are a function of the length of the guide.
The required large aperture with a single-mode-field excitation can be
achieved by a gradual transition produced by flaring the terminal section
of the waveguide to form an electromagnet c horn. Of course, a number
of modes are excited in the throat of the horn at the junction between the
latter and the waveguide. However, the throat serves as a filter device,
allowing only a single mode to be propagated freely to the aperture.
Each mode in the horn can be set into correspondence with a mode in
the waveguide into which it passes as the flare angle of the horn is reduced
to zero, The horn will not support free propagation of a particular mode
until roughly the transverse dimensions of the horn exceed those of a
waveguide which would support the given mode. Thus, unless the flare
angle is too large, all but the dominant mode ~vill be attenuated to a
negligible amplitude in the throat region before free propagation in the
horn space is possible.
The discussion in the following sections will be restricted to horns
that are derived from a rectangular wavepuide. Comparatively little is
known about conical horns derived from a circular ~vaveguidej and they
l,ave found comparatively few applications in microwave antennas.
Rectangular horns are treated in considerable detail in the literature. 2
The reader is referred to the original papers for a complete treatment of
the modes in a rectangular horn and the analysis of the filter properties
of the throat. If the horn is to serve as the terminal antenna element,
there exist optimum relations between the horn length and flare angle
for achieving maximum directivity; these relations are given in the sources
referred to previously. Horns are used in microwave antennas primarily
as a feed to illuminate a reflector or lens. In this case the important
design considerations are the impedance characteristics and the efficient

L C,. C;. Sollth~vorth znd .\. P. King, Proc. IRE, 27, 95 (1939); A, P. King, Bell
[,dmratorfes RPcord, 18, 247 (1940).
# 27, 51 (1939) ; W. 1,. Barrow and F. D,
z If-. 1, Ilarrolv a])d 1,. .J, (hll, Froe, Jlfll,
Imwis,Imc. 27, 41 (193!)) T,.J. (})(! and \V.1,. IIarrow, Trans. AIEE, 68, 333
IFtl~,
f 1g~!),, Thf, dcsigt] data mrr slimrllarizc(l IJY F, 1. Tcrlnan, Radio EngineersHand-
l,mk, \[r(; r:i~v-Hill,SCJV}ork, I!).J3,pp 824-837.
350 WA VEGUIDE AVD HOR,V FEED,? [SK,. 107

illumination of the optical system rather than maximum gain from the
horn. Only the material not readily available in the literature will be
presented in the following sections, with attention being centered on the
microwave design problems.
Various types of horns are ill~lstrated in Fig. 108. lhe horns shown
in Fig. 10.8a and b are known as sectoral horns; they are flared in one
plane only. Thi fields in the sec-
toral horns consist of cylindrical
waves the axes of which coincide with
the line of intersection of the planes
containing the flared sides. The
compound horn (Fig. 108c), allo~~s
variation of both aperture dimen-
(a)
sions. An alternative procedure to
that shown in the figure is to flare
both sides of the horn directly from
the junction with the waveguide to
form a quasi-pyramidal. structure.
From the point of view of the im-
pedance characteristics the former
procedure is preferable.
10.7. Modes in E-plane Sectoral
Electric rdane throat .
Horns.The sectoral horns to be
considered first are those in ~vhich
the flare increases the aperture in the
direction of the electric ~ector (Fig.
10.8a). lhey ~~ill be referred to as
E-plane sertoral horns. The sec-
toral character of the space inside the
flare and the cylindrical coordinate
(c)
system (z,r, O) appropriate to this
~lG. 10.8. Horn feed t~-ws : (a)
eiertric plane horn: (h) r,,awl?tlc,plane space can be seen in Fig. 10%.
horn; (c) compoundhorn. The z-axis coincides with the line of .
intersection of the planes containing the flared sides; the planes of con-
stant z are thus parallel to the unflared sides of the horn. The polar
coordinates r and 0 locate points in these planes. The unflared sides of
the horns are in the planes x = k a/2. Propagation in the flare is along
the radius vector, the wavefronts being coaxial cylindrical surfaces of
constant r. The portion of the flare included bet~veen any t\vo of these
surfaces can be thought of as a length of sectoral guide. In particular,
the horn ffare is a section of sect oral guide \vhose length is (r~ rl),
\vhere the surfaces r = r-l and r = r? locate the throat and (mouth
of the horn respectively. Maxwell)s equations for the sectoral guide
space may be written I
SEC. 10.7] MODES I.V E-PLA ,%TEL$ECTORAL HORNS 351

(28a)

(28b)

(28c)

(28d)

(28e)

(28f)

(28g)

(28h)

~,eld~ electric

{ magnetic- ----

(a)

Coordinate system

(b)
FIG. 10 !J -Coorri, nate .>btell~ atId Iowest-mn<k field ronfigl]ration in wetoral guide

d

352 WA VEGUIDE AND HORN FEEDS [SEC. 10,7

where c and ~ are the inductive capacities of the medium filling the sec-
toral guide.
The Dominant-mode Fields.The wavetmide feeding the horn is
assumed to support free propagation of only the TE10-mode which is then
the exciting field impressed on the horn. The lowest sectoral guide
mode, which is the analogue of the TE10-mode in the uniform guide; will
predominate, all the other modes being attenuated in the region of the
throat. This mode is characterized by vanishing of all field components
except E@, H,, and Ht. hIaps of the field lines in the cylindrical wave-
fronts are qualitatively the same as those in the plane wavefronts of the .!
uniform guide. The electric lines are arcs normal to the flared sides of
the guide. The electric field l?~ varies sinusoidally in the z-direction, (
vanishing at the parallel ~valls of the guide.
To derive expressions for the dominant-mode field components, the
simplifications E, = Ez = 119 = O are introduced into Maxwells equations
{28~) to (28h), \vhich then become

jupH. 1~
= ; ~r (?-E,), (29d)

; $ (?-H,) + g = o. (29e)

Equations (29c) and (29d) serve to express H, and Hz in terms of the

derivatives of Iie. Substituting the expressions so obtained in ~q. (29b),
the following equation is obtained for Efl: ,

%+::+%+(-:)
=0
.4s was pointed out previousl.v the elpctric field lit varies sinusoidally
(30)

along the x-direction as in the case of the lh1~-mode of the uniform

guide. We have then

E, = COS y j(r), (31)

()
where f(r) is a function of r only. The expression for E6 satisfies the
boundary condition that Ed = O at x = t a, 2. Substituting Eq. (3 I )
I
into Eq. (3o), we obtain the follo~ving dillerent ial e(luation for f(r) :

(32)
sEC. 107] MODES I.% E-PLA.VE SEC7ORAL HOR,VS 353

where

(33)

Equation (32) is the Bessel equation of order unity in the argument

(L3r). The solutions to the equation take a number of clifferent forms;
any linearly independent pair of solutions may be taken to construct the
general solution. Denoting by 21 (Pr) any solution, we have

E8 = Cos
(); 2,(/3?-), (34a)

and the corresponding components of the magnetic field are

(34b)

(34C)

In obtaining H, use is made of the recurrence relational

$ [pZ.(p)] = ~Z._l(p). (35)

The linearly independent solutions to Eq. (32) which are particularly

suited to the present problem are the Bessel functions of the second kind
the Hankel functions ll~)(pr), H~2(DT). Z, (@r) in Eqs. (34a) and (34b)
is LObe taken as representing either one of the two functions; similarly
Z,(PT) denotes either of the Hankel functions H$) (&), H~2)(L3r-)of ordel,
zero. These solutions represent traveling \vaves as is evident from the
asymptotic forms of the functions of order n for large ~r:

(36)

It is seen that the first of these represents a wave traveling in the negative
r direction, i.e., a wave converging on the cylinder axis r = O, and the
second a wave traveling in the direction of increasing r. The solutions
correspond to e~~zand e@zin the uniform guide. For large Br the phase
fronts are spaced radially in the sectoral guide exactly as they are in
the z-direction in the uniform guide. The amplitude is proportional to
~>~because the energy density associated with a traveling cylinder wave
1G. N. Watson, Bessel Ftmctzons, 2d cd., Macmillan, NPWYork, 1945.
354 WA VEGUIDE AND HORN FEEDS [SEC. 107

is proportional to Tl, that is, to the reciprocal of the area of the wave-
front. It will be noted that the wavelength of propagation k, = 2r/D
is the same as in the uniform guide. The cutoff condition for the mode
is the same as that of the TE10-mode in the uniform guide. In fact, the
cutoff conditions for the TE~o-modes in the horn are all the same as for
corresponding modes in the uniform guide; this explains the attenuation
of the higher modes generated at the throat. For small BT values, the
interpretation of H~l)(13r) and H~2)(or) is not so simple because there are
quadrature terms in the function that represent energy stored in the
electromagnetic fields. These terms become more important as @ !
becomes smaller. For numerical tables of H$) (pr) and H~2)(Dr) for small
(~r) the reader is referred to Watson.
Having selected the Hankel functions as particular solutions of Eq.
(32), the general solution for the dominant-mode field components in the
sectoral guide can be written down as follows:

(37a)

(37b)

.,,
where the constants of integration A and a are in general complex. The
general field of the dominant mode consists of the incident wave gener-
ated at the throat and the wave reflected by the mouth of the horn. The
magnitude of a is less than unity because it represents the ratio of the
amplitude of the field components in the reflected and incident waves.
Higher-mode Fields.In addition to the dominant-mode fields which
have been considered in detail, fields of other modes exist locally in the
sectoral guide. The mouth and throat discontinuities give rise to these
modes because the boundary conditions at these points cannot be set up
in terms of dominant-mode fields alone. At the throat the amplitudes
of the higher-mode fields are small compared ~vith those of the dominant
mode unless the flare angle dois large, and they exist only in the immediate
neighborhood of r = TI because the sectoral guide dimensions are below
cutoff. It will be shown (Sec. 10.11) that E-plane sectoral guide admit-
tances can be calculated from the dominant-mode fields alone without
appreciable error. At the horn mouth the effect of higher modes is prob-
ably not negligible, especially for smal lapertures. The boundary problem
is a difficult one, and no rigorous solution has been obtained. Experi-
mental values of the mouth admit tante contain hi~her-rnode cent ribu-
tions but in an unknown proportion. In calculating the radiation field
1Watson, op. cd., lkble I, .4ppendix.
 from sectoral horns neglect of higher-mode fields in the aperture probably
introduces an appreciable error.
1008. Modes in H-plane Sectoral Horns..k horn will be referred to
as an H-plane sectoral horn ~vhcn flaring increases the aperture in a plane
perpendicular to the electric vector (Fig. 1086). The sectoral character
of the space in the fiare dictates the choice of cylindrical coordinates
as in the E-plane case. In this case, the coordinates r, 0, and y will be
used (Fig. 10.9b) because the axis of the cylindrical waves in the horn is
parallel to the y-axis in the uniform guide. The portion of the flare
r between any two surfaces of constant r can be considered as a length of
H-plane sectoral guide, the flare as a ~vhole being of length (r, r,),
where r~ and rl are chosen as sho~vn in Fig. 10.8, Maxwells equations
for the H-plane sectoral guide are the same as for the 13-plane guide
[Ills. (28a) to (28h)] except that z is replaced by y.
The Dominant-mode Fields.k in the electric-plane case, energy is
propagated in only one mode because of the dimensions of the H-plane
sectoral guide at the horn throat andthe nature of the excitation by ll?l O-
modc in the uniform guide. Tile dominant mode is characterized by
vanishing except E,, 11,, and 116. The cylindrical
of all field components
character of the space requires that the wavefronts of this mode be sur-
faces of constant r-. To satisfy the boundary conditions E, must vanish
at the \ralls 0 = f 00.
+1 The expressions for the dominant-mode field components are derived
from klaxwells equations for the If-plane sectoral guide after introducing
the simplifications E, = Ej = 11. = O. Substituting y for z in Eqs.
(28a) to (28h) and dropping terms invol,ing E,, E,, and Hv, Maxwells
equations for the dominant mode become

(38a)

(381))

(38c)

(38d)

(38,)

Equations (38c) and (38d) serve to express 11, and IIH in terms of the
derivatives of E.. Substituting the expressions so obtuined in Eq. (38b),
4 the follo~ving equation for I]Uis obtained:

(39)
356 WA VEGITIDE AND HOR.%- I~EEIIS [sEc. 108

Since the boundary conditions require that E, vanish on the walls

0 = ~ O., Ez is of the form
EV = cos pO F(r), (40)
where
?r
(41)
P=zo
and F(r) is a function of r only. Substituting in Eq. (39), the follo~ving
equation for F(r) is obtained:
[

=+;% (42)
a(kr)z +[-MF=O,
where

kz=uzpe= ~Zlr
()

Equation (42) is the form of Bessels differential equation whose solu-

tions are cylinder functions of order p. The Hankel functions H>)(kr)
and H~] (kr) are chosen as particular solutions of this equation again
because they represent traveling waves (Sec. 10.7). The general solu-
tion for the electric field E. is therefore

E, = A cos pe[ll$) (kr) + aH$ (kr)]. (43a)

From Eqs. (38c) and (38d)

~, = PA Sinrpe [H&(kr) + aH~(kr)], (43b)

yop

He = &p cos pd [H~ (k.) + aH$ (k.)], (43C)

,
where the primes indicate differentiation with respect to kr.
The solutions for the field components in the H-plane sectoral guide
[Eqs. (43)] are of the same form as those for the E-plane sectoral guide,
but they differ in two noteworthy respects. In the H-plane guide the
order p of the functions depends on flare angle O.. It is high for small flare
angles and is in general not an integer. Thus, for a flare angle of 20,
which is often used in practice, the order is J. In the H-plane guide the
argument of the Hankel functions is kr( = 2m-/A). From the asymp-
totic expressions [Eqs. (36)] it is seen that at large k. the equiphase sur-
faces are separated by a free-space wavelength in contrast to the guide
wavelength of the E-plane horn. This is reasonable because the H-plane
flare increases the separation of the walls that determine the guide wave-
length in the uniform guide. For small 0, and high-order p, the asymp-
totic expressions of the Hankel functions are good approximations only
 at very large lcr, corresponding to the fact that this wall separation
becomes large only at very large kr.
Higher-mode Fields.As in E-plane sectoral guide higher-mode fields
are necessarily present at throat and mouth cliscontinuities. The effect
on impedance and radiation patterns of the higher-mode fields at the
mouth is not negligible, although it is at present impossible to take them
into account in sectoral guide theory.
10.9. Vector Diffraction Theory Applied to Sectoral Horns.The
same considerations discussed in connection with radiation from open
F waveguide (Sees. 10.1 to 10.3) apply to radiation from horns. If the
current distribution on the inside and outside walls of the horn were
known, it would be possible to calculate the radiation field at a point
outside the guide from this distribution. In the absence of this knowledge,
the aperture diffraction method is used as in the case of the waveguides.
In the present case the aperture surface is taken to coincide with the
cylindrical wavefront of the dominant mode at the mouth of the horn.
The aperture field is assumed to be that of the incident wave, the effect
of the reflected wave being neglected. The radiation field is computed by
means of Eq. (5.103). As usual the radiation field is expressed in terms
of spherical coordinates, the origin of ivhich is here taken to coincide \vith
that of the coordinate systems shown in Fig. 10.9. The z-axis of the
.!f latter forms the polar axis of the spherical coordinate system, azimuth
being measured \vith respect to the zz-plane in each case. The results
are written down in the following paragraphs for the cases in which the
medium is the same inside and outside the horn.
Radiation jrom E-plane Sectoral Horns.For the E-plane the radiat-
ion field at an external point P, as derived from Eq. (5.103), can be sho~vn
to be

((i, x R,)Ed +
o>~
~ [i, R,(i,.R,)]II=
1
dO dx. (44)

The quantities R, @, and @ are the spherical coordinates of the point P;

rz, ~, and ~ are the coordinates on the surface of integration, which is
taken to coincide \vith a ~vavefront (r = r~) at the mouth of the horn;
i,, i~, and RO are unit vectors in the directions x, d, and R increasing.
For the plane @ = 90 (electric plane):
358 11.4LEG[II)E A .Vl) HOR.V FEED,7 [SEC. 101[.

For the plane @ = O (magnetic plane):

Expressions for EO and Hz over the aperture are .gi\-cnby Eq. (37), Sec.
108, when r is replaced by r~.
Radiation jrom H-plane Sector-al IIorns.-For the H-plane sectoral ,
horn

For the plane @ = 90 (electric plane):

jkr2eikR ~ e.
Ep = ie ~TR ~ dy _ e,e+(. sin(-)+,,...0.., ~)
//

[ Ycose-(:),cos@ld, 47a)
For the plane @ = O (magnetic plane):

Expressions for E. and H, are given by Eq. (43) when r is replaced by r,,
10.10. Characteristics of Observed Radiation Patterns from Horns
of Rectangular Cross Section. J$Then radiation patterns from srctoral
horns are observed and compared l~ith the patterns obtained from Eqs.
(45) and (47) by numerical integration, in general it is found that they
do not agree in detail. In vie\v of the fact, that the theory neglects the
cllrrent on the outside walls of the scctoral guide and the higher-mode
fields in the aperture, this is not particularly surprising. only a brief
summary of the experimental data will be attempted here.
Figures 10.10 to 10.13 are compilations of patterns from a number of
E- and H-plane sectoral horns of large aperture; the apcrt m-es ~vcre plane
surfaces perpendicular to the axis of the guide. They are classified by
flare angle and radial length meawrrd in wavelengths (r, ~X of Fig.
10.9). It will be obserwd that for a horn of constant flare angle the
main lobe undergoes wide changes in ~vidth and structure as the horn
length increases. This can be correlated qualitatively If-itb the changes
in the field over the mouth of the horn. For a giwn flare angle the aper-
ture area increases directly \vith the horn length; this alone n-ollld tend to
sm. 1010] RA1)IA TIO.V 1.4 TTER.V,V FROJ1 HOR.VS 359

narro!y the beam as the length increases. Ho\}-ever, the effect of in- I
creasing aperture is m-ershaclowed by the phase-error effects. I.et 8
represent the maximum dcpart~u-c of the ~ravefront r-2from the aperture
plane (Fig. 1014). Then 27r~,Afl,~rhere AOis the ~vavelength in the
scctoral guide at the month, is the phase difference between the center of

..
E- 10 + =8.97 E- 10 + =13,7 E.lOO %=23

00 100
(a)
20

E-20 ~=3.96 E-20 ~ =6.65 .E-20f =11.6

(b)
lI,; . lL1.10.-Itadiat~onpatternsof E-planewrtoral horns of vario,mlengthsand tmre
an~lcb:(a) flareangleof 10; (h) flare angle of 20.
360 WA VEGUIDE AND HORN FEEDS [SEC. 10.10 ,

theaperture and the edge. Itcaneasily be seen that

6 (1 cost?,).
= r2 (48)
~ A,
Under certain conditions ~f length and flare angle a phase error is pro-
0
10 20

5 db
30

10 db
40

15 db 5(Y

60
20 db
70
80
90

E-30 :=1.77

0 10 20

5 db

10 db

15 db

?0 db

16O1OY
MY 130120
E-40 ~=1.53 E-40 Q:=2.4 E-40; =3
~-40 ~=1.21
A
(b)
FIG. 10.11.Radiation
patternsof ,?-plane.ectoralhorns:(a) flareangleof 30; (~) flare
angleof 40.
 SEC 1010] RADIA TIO.V PA TTER.VS lRO]Jl HOR.VS 361

duced over the aperture that leads to a minimum in the main lobe in the
forward direction such as may be seen in Fig. 10.lla.
When the aperture or flare angle of a horn is small, ah, is small and
the wavefront at the aperture approximates a plane. Horns are charac-

00 10 20

5 db 30

10db
40

15db 5P

w
20 db
1P
8P
w
H-l@ + =10.0 H- 100 ~ =14.8 H-1OO +24.3
(a)

00 100 200

5 db 3F

10db
40

15db
504
Ho
20 db
70
Ml
90+
H-20 ; = 4.57 H-20v ; =7.22 H-200 ~ = 12,2
f
(b)
FIG. 10.12,Radiation patterns of H-plane sectoral t,orns: (a) flare angle of 10; (6) flare
angle of 20,
 362 WA VEGIJIIJE A-VII HOR.V FEEDS [SEC, 1010

0 10 20

5 db

10 db

15 db

20 db

180 170 160 150

H-30: =4.17 H- 30; = 2.54 H-30 +=1.92
H- 30 := 5.63
(a)
l:TG.1013--f{adiation patternsof H-plane sertoral

terized by uniform amplitude distribution across the aperture in the

E-plane and sinusoidal distribution in the H-plane [Eqs. (37a) and (43a)
and Fig. 10.9]. In the case of approximately uniform phase, therefore,
the gain and main lobe width should be functions of aperture correspond-
ing to uniform illumination in the E-plane and sinusoidal illumination
in the H-plane. At present these functions can be determined only from
experimental data. In Fig. 1015 the observed 10-db widths of a num-
ber of horn patterns are graphed against the reciprocal of the aperture
in wavelengths. For all the horns 5/kn \vas less than ~. Results have
been obtained for both the sectoral horns and compound horns (If the
type illustrated in Fig. 10.8; in the latter case the flare again ~vassuch that
the phase over the aperture \vas substantially uniform. The E-plane
10-db width (tenth-po\ver width) for all horns lies on the same curve,
showing that the E-plane pattern is a function only of the E-plane aper-
ture. The 11-plane patterns, on the other hand, depend on both aperture (
dimensions. Thus, the values obtained from H-plane sectoral horns on
!: standard guide (E-plane aperture approximately A/3) fall on Curve 11;
{.,. \rt]ereas in the casr of the rompollnd horns \vith an E-plane aperture of a
 SEC. 10101 RAL)I.411O,V 1A TThR.VS FROJ1 HOR.brS 363

00 100

Ilorns;(a) flareangleof :30;(~) flareangleof 40

\vavelength or greater, the H-plane 10-db ~vidths fall on (urve I. Points

for intermediate E-plane apertures which fall bet~veen (urves I and II
are not shown. At first glance this is some-
what surprising if one assumes that the pat- h
~t A
terns depend only on amplitude and phase
\ 7\
distribution of dominant-mode fields in the @o\\
aperture. It means, however, that the other < 1A
factors, namely, higher-mode fields in the \
//
aperture and currents on the outside \valls ,1
of the horn, contribute in the case of small E- 631
FIG, 1014.-Orlgin of
plane apertures and are relatively unim-portant phase variation across the
for E-plane apertures greater than a ~vave- awrturc of ~ .C~tOralhorn
length. These factors are apparently not Ad = 2=6X)
dependent on n-plane aperture, at least ~vhcnthis aperture is 0,7k or more,
,, as in horns on standard rectangular guide, because the observed E-plane
~vidths do not depend on H-plane aperture.
The phase variation across the aperture of a horn is small, for a given
aperture dimension A in the plane of the flare, only if the flare angle is
 w

o
t

364 WA VEGUIDE AND HORN FEEDS [SEC. 10.10 !

;
less than a maximum value (or the length r, greater than a minimum
r
value) which depends on aperture and can be obtained from the condi-

I , , 1 ,. I ,
I ] 1/ /

/ A / I I I I I I I

00 ~1 0.5 1.0 1.5 2.0

Wavelength/apartura
FIG. 10.15.ExPerimental 10-db widths of horns having small phase variations over the
aperture ~ < ~ E-plane; H-plane sectoral horns; H-plane
()
of compoun~ horns with E-plane aperture equal to or greater than a wavelength.

tion that 6/x. shall be small. Using the relation for the separation of an
arc and its chord, it is easily shown that

(49)
or
(50)

Using # as the allowable upper limit for 5/As,

(r,)d. = $0

(th)max= sin-
()
~ (51)
 sEC. 1010] RADIATION PA7TERNS FROM HORNS 365

For many applications the aperture is small and Eq. (51) is satisfied by
t convenient values of 190and rz. For large apertures a horn satisfying
condition (51 ) is long and possibly too bulky or heavy for practical appli-
cations. In this case the horn designer is forced to compromise on flare
angle and aperture. If he is to use a horn feed, he must increase the flare
angle and allow for broadening due to phase variation in the aperture
by choosing a larger aperture than that predicted by the curves of Fig.
# 10.15.
In horns of small flare angle (or large rz/x@) the dominant-mode fields
near the aperture are described by the asymptotic forms of the Hankel
functions [Eqs. (36)], which are exponential functions with slowly varying
8 amplitude. Moreover the departure of the wavefronts from plane sur-
faces is small. Consequently the dominant-mode fields in the horn
closely resemble those in uniform guide. The problem of radiation from
h~rns of s.m_allflare angle is therefore approximately the same as that
from uniform waveguide, and the predictions
-- _._.
.._ [Eqs. (23)] of the vector
diffraction_J~y for waveguide can be apphed without serious error to
horns satisfying condition (51 ). It is therefore interesting to compare the
curves of Fig. 10.15 with the corresponding theoretical curves for wave-
, guide in Fig. 10.5. Agreement is good for apertures greater than about
2X/3 in the electric plane and 5A/4 in the magnetic plane, indicating
the probable lower limits at which the factors neglected in the theory are
J
really negligible. It is believed that the predictions of Eq. (23) for w-ave-
guide patterns can be useful when properly applied to horns because gain,
main-lobe ~vidths at various po~ver levels, side-lobe amplitudes, etc., can
be determined for different apertures ~rith relative ease.
Several empirical formulas have been ~vorked out for the 10-db \vidth
as a function of aperture for the average horn feed.
1. For the electric plane:

(52)

2. For the magnetic plane:

(53)

The symbol ~ represents the 10-db width, and 1? and A are the apertures
in the electric and magnetic planes res~ectively. These formulas \vere
obtained from a large number of 10-db !v~rithsmeasured at the Radiation
Laboratory over a period of several years The fiare angle of the auerage
horn is probably abollt 20. Since phase variation is not taken into
account, the formulas ~annot be expected to predict the 10-db ~vidths of
individual hc,ms zccurate]y, bllt they ila~e proved \ery useful as a first
approximation in defiigning horns,
366 WA VEGUIDE A .VD HORN FEEDS [SEC. 1011

10.11. Admittance of Waveguide and Horns. Admittance oj Open

Waveguide.It is observed experimentally that rectangular waveguide of 4
ordinary dimensions when open to space is terminated at the plane of
the opening by a capacitive admittance. This type of admittance is
to be expected in view of the close spacing between the waveguide walls
that are perpendicular to the electric vector (about k/3), It is of inter-
est to note that a rigorous treatment of the radiation from the open end
of a semi-infinite parallel-plate line carrying the ZElf-mode leads to the ;
result that the line is terminated by a capacitive admittance. I This
property of the waveguide will be useful in analyzing the admittance I
characteristics of horns.
Admittance of Sectoral Horns. From the transmission-line point of 1
view a sectoral horn consists of a length of sectoral guide terminated by
a mouth admittance at one end and joined to uniform guide at the other.

The discussion of its admittance characteristics will be based on sectoral

guide transmission-line arguments. The input horn admittance observed
in the uniform guide depends on the aperture admittance terminating the
sectoral guide, the guide length, and the transformation associated with
the junction to uniform guide at the horn throat. 2 In the following dis-
cussion sectoral guide characteristics will be summarized. The sum-
mary will be followed by a discussion of mouth admittances, junction #
effects at the throat, and the influence of both factors on horn admittances.
Particular attention will be given to E-plane sectoral horns, In what .
follows, when the term horn admittance is used, it will be understood
to refer to the admittance measured in the uniform guide and referred to
the plane of the junction between the guide and the horn.
Charozteristics oj E-plane Sectoral Guide.In Sec. 10.7 expressions
were developed for the lowest-mode field components in E-plane sectoral
guides [Eqs. (37)]. .4s in uniform guide, one can define and use a wave
admittance, consisting of the ratio of the transverse magnetic to trans-
verse electric fields. If the admittance is expressed in units that make 4
the characteristic admittance of the ZE,O-mode in the uniform guide
equal to unity, the admittance for the E-plane guide becomes #

y= (%?%)
By inspection of the expressions for the field components [Eqs. (37)] it
can be seen that this ratio is a function of f?r only:

(54)

] The results of the analysis are given in Waveguide Handbook Supplement,

I
RI. Group Report No. 41, ,Jan 23, 1945, Sec. 60c.
2J. R. Risser, Characteristics of Horn Feeds on l{ectangular \Vavcgllidc, IiL
Report No. 656, December 1945. I
SEC. 1011] ADJIITT.4NCE OF JV.I TEG1-IIIE .4.YD H(lR.h-.S 367

The complex constant a is determined in magnitude and phase from the

ratio of incident and reflected waves in the sectoral guide. It can be
expressed in terms of the output admittance Y2 evaluated at the aperture
end of the guide where r = rz:

(55)

Substituting this value of a in I?q. (54) an expression is obtained for

the admittance at a general point r in terms of Y~ and line parameters.

(56)

where the symbol K,, is used to represent combinations of Hankel func-

tions as follows:

W, = H:)(@r, )II:(@r) H:(@r,)H$~(&) (57)

A degree of simplification of Eq. (56) is obtained by expressing the Hankel

functions in terms of amplitude and phase, using the property that for
real values of @r, H~21(~r) is the complex conjugate of H$j)(&). Let

H$)(&) = Fe@,
(58)
Hf) (@r) = Gel@, )

where F, G, $, and @ are real functions of Dr. Numerical values of these

functions are listed in tables of Bessel functions for small values of pr;
for large /3r the asymptotic values can be used. Substituting in Eq.
(56) and using subscripts 2 for the functions evaluated at the aperture
where r = r~,

~ sin (42 *) + j 1~~ sin (@? *)

Y(f?r) = (59)
G.. } .,<;,,
~ sln (~, @) -fi,: sin (@, ~)

The characteristic admittanc,c l~(~r) of E-plant> sectural guide can

be ~vritten do]rrr from Eq. (54) by mdiing a e(ll[:d to zero,

(60)
368 WA VEGUIDE A .VII HOR.V FEEDS [SEC. 1011

For large @i-, Y. approaches unity, since F/G -1 and (O ~) ~ 7/2.

For small @r, Y.(fh-) is complex. It is graphed in Fig. 10.16. The com-
plex character of Y, for small Pr is due to the fact that the fields in the
region of the horn apex store as well as transmit energy.
1.2

1.0

; 0.8

0.2 .; \ I I
Imaginary pati

o I
o 1.0 2.0 3.0 4.0 5.0 6,0 7.0 8.0 9.0 10.0
(J%)
FIG, 10.16. Characteristic wave admittance Y.(L3r) of a sectoral guide.

It is often useful to speak in terms of a reflection coefficient r, in

the sectoral guide and to use its transformation properties along the
guide; it is defined as the ratio of the electric vector in the reflected wave
to that in the incident wave. Then

Since r,e-~ is equal to the complex constant a, r, transforms down the

sectoral guide according to the relation

\vhere r,, @ are evaluated at the point r and r:, O at r. It can then be
shown that

1;
r,=yc. (63)
_ e;z(*+)
Y,
Conversely
Y 1 + r~e,l{i$j
(64)
x= l+r,
 SEC.1012] 1RANSFORMA !lION OF THE E-PLANE HoR.V 369
~ I
I For large ,&-, 2(4 ~) is equal to r, so that Eq. (63) becomes identical
with the corresponding expression for r in uniform guide.
Mouth Admittance of E-plane Sectoral Horns,The admittance of an
open sectoral guide has not been successfully treated theoretically,
Qualitative arguments can be applied from the theory of open parallel-
plate transmission lines. Aslongas theelectric plane aperture is small,
the mouth admittance should be capacitive as in the case of the wave-
guide. As the electric plane aperture is increased, the capacitive term
in the admittance should decrease. In the limit of large electric-tiane
aperture the admittances probably determinedly the separation of the
guide ~valls in the magnetic plane.
Experimentally,., the determination of the mouth admittance of an
E-plane sectoral horn is compar-
?
atively easy. It has been ob-
0.15 L
tained in a number of cases by . 00=150
\
using Eq. (59) to extrapolate down x00=300
\ BJig=0.72B/k
the sectoral guide from the throat
\=
where the admittance is deter- : 0.10
mined from measurements in the k .\.
uniform guide. 1 The capacitive = .
0.05
susceptance term, which decreases
with increasing aperture, is seen to
be present. For large apertures
o
the admittance is independent of 0.5 1.0 1,5 2.0
aperture to a first approximation. B/h
The magnitude of r,,, the reflec- FIG. 10.17.Reflecti0n coefficient in
sectoral guide referred to the aperture B,
tion coefficient in the sectoral where B is the aperture dimension in the
guide referred to the aperture, E-plane.
decreases rapidly with increasing aperture, being small and approximately
independent of aperture for apertures above 3A/4 (Fig. 10.17).
Except in the region near cutoff the mouth admittance and reflection
coefficient r,~ are not sensitive to wavelength changes of the order of
10 per cent. This is due to the fact that at large apertures for which
the aperture-to-wavelength ratio changes rapidly with wavelength, the
admittance is practically independent of 1?, where B is the aperture
dimension in the E-plane.
10.12. Transformation of the E-plane Horn Admittance from the
Throat to the Uniform Guide. The E-plane Throat Transition .The
effect on admittance of the transition from sectoral to uniform guide at
the horn throat depends primarily on pr,, where r, is the inner radius of
the sectoral guide (Fig. 108). It is informative to consider first the case
where there is no reflected wave h the sectoral guide, so that the char-

1Risser,op. cd., p. 19.

370 lVAth(; l[DE .4.V1) HOR.V FIihDS [SEC. 1012
I
act,eristic admittance of sectoral gllide at r = rl is YC(@l). As can be
seen by inspection of Fig, 10.16 for values of @r > 5, Y.(&l) approaches i
unity, that is, it becomes equal to the characteristic admittance of the
wa~eguide. The throat mismatch becomes negligible as@l increases or
as the flare angle decreases; @-1 depends on waveguicle height b, flare
angle 02, and guide !vavelength, as follo~vs:

~b I
@r, = (05)
A,, sin O.

To sho~v the effect of d on the throat admittance, a plot of calculated -

admittances for a series of (lovalues at 10.0 cm for horns built on stmd-
!

G G
component
Conducbnce To component
Conducbnce Yj 1

(a) (b)
l:l~. 10,18.Throat adr]]ittar~re asafllnrtionof flare an~lealld frequer]cy: (a) X = 10cnI,
80 varied; (h) kvaried for flare :mgleb of 1.5 and 30.

ard guide (b = 0.34A; i,~k = 1,39) is sho~vn in Fig. 10.18. rhr throat
mismatch is small for flare angles less than 10 ond incrrascs ~vith increas-
ing 0,, For the 15 and 30 cases the admittances are plot ted in Fig,
10.18b for wavelengths from 9 to 11 cm, From this the mismatch can
be seen to increase in the direction of the long ~vavelength end of the band,
becoming very large ~vhen the wavelength approaches cutoff. In choos-
ing b values for applications involving nonstandard guide, it is necessary
to be careful because small values of b are equivalent to large values of
00 or X., 1
When the sectoral guide is not matched, the admittance Y(pr-l) in
the sectoral guide at the throat is given by Eq. (59) \vith r- = rl. In
either case, when computing the admittance in the uniform guide at the
throat, the admittance in the sectoral guide at r = r, must be multiplied
by a factor that ensures continuity of voltage and current at the junction.
The continuity of current is ensured by continuity in 11,. However,
voltage is proportional to the product of the length of the electric-field
lines and the field strength. Thus, in the uniform guide the voltage is
bEu( = 2r, sin 00E,) and in the sectoral guide 2r160E~. Therefore, the
current-voltage ratio is proportional to H. i(2rld@6) in the sectoral guide
I
and to H,l(2rl sin 19@U)in the uniform guide, so that
SW=.10.12] TR.[.\.SlY)R.lf.4
7IOaVOF THE E-Pli.[.Vh HOR.V 371

Y. = s%Y(&-J, (66)

\~here Y~ is the horn admittance in the uniform guide referred to the

plane of the junction \rith the throat of the horn. For values of 00 up to
300 the ratio (sin 00)/80 differs only slightly from unity, and the correction
is not important. In general, \vhen guides of different geometry are
joined, account must be taken of the contribution of higher-mode fields
to the admittance. However, in sectoral horns ~vith va]urs of 00 up to
3(Y this effect can be neglected.
Admittances of E-plane Sectoral Horns.-A discussion of the relation
bet~leen the parameters of an E-plane horn and its admittance can best
be carried out in terms of reflection coefficients. If rH k the reflection
coefficient in the 7miform yuide referred to the junction \~ithhorn throat,
\rhere according to the usual definition of reflection coefficient

1,,, = .IYH
..
1 YH

it can be shown from Eqs. (54), (60), (61), and (66) that if the reflection
coefficients are small, in particular Ir, r,ll <<1,
r,, = r, + r2, (67)
\vhere

l s+ Yc(&-,)
rl= (68)
1+ s+ Y.(pr-,)

,1 is the reflection coefficient in the sectoral guide at the horn throat,

[1;q. ((}1)], and c is a small angle given by

sin 00Fl(jrl)
Cos (*1 @l)
~ = ~ tan_l en ,G, (&,)
(70)
sm (?OF,
1+ ~m sm (+1 @l)

Equation (67) states that r~ is the sum of two components , and

~; II, is the reflection coefficient in the uniform guide at the throat when
the sectoral guide is matched, i.e., when r,, = O, and the admittance of
the sectoral guide at r = i-, is its characteristic admittance. For any
of the IO-cm horns ]vhose throat admittances (sin O()00) Y,(@r]) are plotted
372 WA VEGIJIDE AND HORN FEEDS [SEC, 1012

in Fig. 10.18, r 1 is the vector drawn from Y/ Yo = 1 on the chart to the

admittance point. From Eqs, (62) and (69)

where
A = c + 2[I#J(8n) o(~r,)l.
For reasonable flare angles (sin 0~)/00 = 1 and F, (Br,)/G, (drJ = 1 so
that r, differs essentially only in phase from r,?, the reflection coefficient
at the horn mouth. The phase angle A consists of the sum of a term

ponen
o

Conductance component (~) Conductance component ($.)

0.=20 O.= 30
FIG. 10.19.Admittances of 15, 20, and 30 electric plane horns for different flare lengths
(T2 T,)/hp

depending on the horn length, 2[q$(@rz) o(dr,)], and C, a small Phase

shift at the throat. For the 10-cm horns of Figs. 101S, ~ varies from 3
when 00 = 5 to 15 when (lo = 30.
The manner in which the admittance YH of a sectoral horn and its I
reflection coefficient ~ depend on the vector sum of 1 and z is illus-
trated in Fig. 10.19, where the admittances at 10 cm of a series of horns
 SEC. 10.12] TRANSFORMATION OF THE E-PLANE HORN 373

with different lengths are plotted on a reflection coefficient chart for

three different values of 00. These admittances were determined experi-
mentally. The admittance of all horns with the same 00 lie on a spiral
whose center is determined by r j and whose periphery is determined by
rl + rj. The decrease in the radius of the spiral with increasing horn
length (and aperture) is due to the decrease in the magnitude of I., ~vith
1.8

1.7
x 150
.--_ Zoo
1.6
o 250
~-3LP

>

1.3

1.2 \

1.1 !

1.0
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1,8 2.0
Flare length, (rz - rl) Ag
11,+. 10.20 .--Standinwaveve ratio vs. flare length for typical elertrir plane horns,

increasing aperture (Fig, 10 17). Nloreover, since the aperture height B

increases more rapidly \vith increasing length for larger flare angles, the
inner portion of the spiral is rcachcd for smaller values, of /3(r2 r-l) for
larger flare angles. The relation bet}vecn B and (TZ rl) is

B = 2(T, r-l) sin 00 + b. (72)

Since the inner radii of the spirals arc independent of aperture for long
horns of large aperture, r,j must have a small constant value independent
of B/X for large values of B.
For a given ~, a series of horn Iengt hs exist for \rhich r, and r, are
180 out of phase and the match is optimum. An empirical formula for
these optimum lengths is

\ From Eqs. (72) and (73) horns can be designed to bc matched at any
374 TVAf-E(I[ll)ti .i .Vl) HOI{.V FEE]),? [SE(. 1013

aperture and ~ravelength, since the value of dois not critical. lhe degree
1
of match attainable is indicatedin Fig. 10.20.
The mouth and throat admittances and conseql~ently r.z and I, are
not sensitive to ~vavelength changes of the order of 10 per cent (Fig.
10.18b). Theprincipal frequency sensitivity of an E-plane sectoral horn
therefore arises from variation of the effective scctoral guide length
2[@(13i-,) o(pr-,)]. This is clearly sho\vn by the admittance-frequency
curves for several typical horns she\\-nin Fig, 10.21. For the long horns

Conductance component

(a) (b) (c)

l:l,;. t021.--. &<{!rlitt:ltl<,psoftIlr?p l,~ clet,tric plaue 11 OTIIS
of different lengths as a fun,,.
tiocl<](x%x,elp,lgtll: (a) rz r, = 1().(icr,), Avarledfrom9. Oto 11.5cm with best nlatrhat
11.4 r,,,; (h) ,, r, = 22. S c,,>, h VO,WC1fro,,, 9.1 to 11.5 c,,, with best match at 10.1 en,;
(c) l, r, = 26.3 crr1,A varied Crotu S.1 to 11.5cn] with bebt nmtc hat10,9 cm.

the frequency variation causes the admittance to traverse more than a

complete loop, corresponding to a change in A [Eq. (71)] of more than
360. As discussed previouslyin this section (see Fig. 10.18) increasing
\va\,elength results in incr~asing mismatch at the mouth and throat, so
that r,, and II, both increase in magnitude as the long-wavelength end
of the band is approached. To obtain a low SWR over a very wide
band, it is advisable to choose the horn length for optimum match at
the long-wavelength end of the band; a comparison of curves b and c
of Fig. 1021 indiratw that improvement is obtained by so doing.
10013. Admittance Characteristics of H-plane Sectoral Horns.The
wave admittance for If-plane scctoral guidcis (k)6)(P/() ~(HO/EU), using
units in ~vhich the characteristic admittance of the T,EIO-mode in uniform
guide is unity, Then

(74)

\\-herethe prime indiratcs the derivative of the flmction \vith respect to

(kr). Noting that Eq. (74) forthe If-plan ediffersfromEq. (54) forthe
E-plane in having 1[~(kr) in plare of H~)(@r), ll~)(kr) in place of
H~)(~r), and (jk ~) in place ofj, eqllations for the 11-plane analogous
to each of E-plane equations [Eqs, (,54) to (64)] can be written do~vn.
There are practical limitations to the llsefulness ef ttle H-plane sectoral

-1_
SEC. 1013] AD JIITT.4. TCE CH.iR.l( TEHI,~7IC,Y OF H-PJ,A,VE 375

guide equations, since the orderp

() 90
00
is not the same for all horns,

as it is in the E-plane case. The I.jrder is in general high, horns for most
applications having small flare
angles.
While mouth admittances of
H-plane sectoral horns are some-
what difficult to calculate because
of the high p, it is not difficult to
predict them approximately from
E-plane horn measurements and
parallel-plate theory. Since the
aperture B in the electric plane
is small compared with the mag-
netic plane aperture and is equal
to the height b of the uniform
guide on which the horn is built,
the mouth admittances of all H- (a)
plane sectoral horns would be ex-
pected to be approximately equal x /.\ A.
to the mouth admit tanrc of open
uniform ~raveguidc and to exhibit
the same degree of frequency sen-
sitivityy.
The throat transition in mag-
netic plane horns of small angle
has a minor effect on the horn ad-
mittance, and the throat reflection
is very small compared \vith the
reflection at the mouth. This has
been shown experimentally by
making standing-wave measure- . .
ment on magnetic plane horns
(b)
over a band of frequcncirs and 1IG. 10.22 .td, t,lttanres of 15 atld
again at a constant frequency 30 Itlamet)r[,I:L!lP Ilorw for differer(tflare
while the horn length Ivas cut lelllqlls.(IL)O = .5;(1))0 = W.
down (decreasing the aperture simultaneous y \\-itht}le length). In both
cases the shift in phase in the standing-~vave pattern indicated that the
large reflection occurred at the mouth.
The admittances measurrd at 1().0 cm of a series of H-plane sectoral
horns are shown in Fig. 1022. The admittance ~ralues correspond to
different lengths (and apertures). The form of the ~lot can be explained
by postulating tlvo cornponrnts of the rrflcrtion r(wfficient: a small fixed
376 WA VEGIJIDE AND HORN FEEDS [SEC. 10.14

component representing the throat reflection and a large component due

to the mouth which has a magnitude independent of aperture and a phase
dependent on sectoral guide length. Experiments indicate that the
frequency sensitivity of the horn admittance arises primarily because the
effective sectoral guide length varies.
10014. Compound Horns.Doubly flared horns must be used to
enlarge both the E- and H-plane dimensions. The admittance charac-
teristics of the sectoral horns provide the basis for the design of a broad-
band compound horn of the type illustrated in Fig. 108c. The principle
is to use the frequency characteristics of an E-plane flare to compensate
those of an H-plane flare. Consider, for example, the problem of pro-
ducing an aperture having an H-plane dimension larger than the E-plane
dimension. The desired aperture is obtained by first flaring in the
H-plane to the required dimension and subsequently flaring in the E-plane.
Since the H-plane flare introduces a negligible mismatch at the throat,
the mismatch arises entirely in the E-plane horn, one component at the
junction with the H-plane sectoral guide and the second component at
the mouth. The E-plane flare in this case does not differ much from that
on uniform guide, and the data obtained for E-plane sectoral horns can
be used to determine the E-plane flare angle and horn length so that the
mouth reflection cancels the junction reflection (reflection at the throat
of the E-plane flare). The shortest possible length is chosen for the
E-plane flare consistent with the matching conditions. This is to
eliminate the long-line effect which would cause the phase of the
mouth reflection transformed to the throat to vary rapidly with respect
to the throat reflection as the frequency changed. In actual practice, the
mouth reflection is chosen to cancel only partially the throat reflection,
because it is necessary to close the mouth of the horn by a pressuriz-
ing device which likewise gives rise to a reflection. The reflection coeffi-
cient of the pressurizing device can be designed to be equal to the residual
mismatch of the mouth and throat and phased properly relative thereto
by positioning the device with respect to the mouth to give an over-all
reflection-free system (see Sec. 10, 17).
The technique is essentially the same in the case where the E-plane
aperture dimension is larger than the H-plane dimension. The E-plane
flare is introduced first followed by the H-plane flare. In this case, the
major sources of the mismatch are widely separatedone at the junction
between the uniform guide and the E-plane sectoral guide and the second
at the mouth. Although the latter reflection can be chosen to counteract
the first at some one frequency, the bandwidth is small because of the
long-line effect. It is preferable to cancel the mouth reflection by means r
of the pressurizing device and to cancel the throat reflection by matching
window in the uniform guide. In this way the sources between any pair
of compensating reflections can be put close together so that rapid phase
variations due to long electrical paths are eliminated.
SEC. 10.15] THE BOX HORN 377

10015. The Box Horn.The box horn is a special horn type devised
to have greater directivity in the H-plane than a flared horn of the same
aperture, It is so constructed as to introduce a third harmonic 180
out of phase with the fundamental mode in the aperture plane. This
alters the amplitude distribution across the aperture from the cosine
type associated with the funda- l--~+
mental mode to one more nearly
uniform. 1
The box horn is not a true horn
;4
in that there is no throat that is used
1
to filter out higher modes. Its es- A
A
sential features are sketched in Fig.
10.23a. The horn consists prima- .4
rily of a piece of waveguide of length
L, frequently referred to as a box, ,1--- 1
a
whose magnetic plane dimension A
is large enough to support TE.,O- (a)
modes with values of n up to 4. It
is open to space at one end and fed
at the other by a waveguide or H-
plane sectoral horn of aperture A
located centrally so as to excite only
the modes having nonzero ampli-
tude at the center, i.e., the TEIO-
and TE,,-modes. The ratio of the
~-- Z
amplitudes of the TESIJ-and TEIO-
modes depends on the ratio A /A.
Since the velocity of propagation
of the two modes is not the same,
the length L of the box determines
I----LA
their relative phase at the aperture.
The horn may be made as directive (b)
as desired in the E-plane by intro- F1~.1O.23.BOX horn:(a)directjunction
type; (b) sectoral guide junction type,
ducing an E-plane flare.
It is easy to show approximately how the ratio of the amplitudes of
the TESO-and TElo-modes depends on the ratio A /A of the dimensions
of the two guides. The fields EV(Z,Z) in the box can be represented
as a superposition of the modes excited at the junction. Neglecting
the effect of the nonpropagating modes, we have

7rx %x ~lRa#.
EU(Z,Z) = al cos e~~,oz+ as cos (75)
() A () A

The junction between the two guides is taken to be in the plane z = O.

LS. J. Mason, Flared Box Horn, RL Report No. 653, July 1945.
378 }V.1 V17G( Il~E A.V1) HORiV FEEDS [SEC. 1015

It is assumed that the field over the common area between the guides is
that of the dominant mode in the smaller guide; that is,

+;
I
Ev($,o) = Cos;, /xl <
(76)
EU(z,o) = o, +<lxI <;.

Onsettingz = Oin Eq. (75) andmaking use of Eq. (76) forthefield, we

1.0 can obtain the coefficients al and a
bytheusual procedure fordetermin-
ing Fourier coefficients. The ratio
(
of the harmonic components is found
. to be
% 0,5 \
e
as = /::2c05(:)cos(F)
_

a /:;2cos($)co5t2)
o (77)
o 0.5 1.0
A~A Figure 10.24 is a plot of aS/al vs.
~x~. 10.24.ltatio of the anl~l,tucirs A/.4. The ratio aS/al decreases
v
of the TEm- and Tk?,o-modes ,U a box wit h increasing A /A, reaching zero
hornVS.A/A.
when A /A is unity. If it were pos-
sible to make A /.4 very small, values of CL3 lal approximately equal to 1
could be obtained. The lower limit of A /.4 is 0.20, corresponding to

I
10
0.9
0.8
0.7
~ 0.6
E 0.5
Q
0.4
0.3
0.2
0.1
0
7Y2 r/3 76 0 ?76 ~/3 77/2
7rXL
FIG.1025.-Aperture illul,)lllatlon (a]nplltude) of a Imx honl for a series of talum of tl,e ratio
113n,.
 SEC,1015] 7Htt B().Y HOtt.V 379

A = 0.5A the cutoff in the input guide, so that the maximum aJal is 0.93.
Actually much smaller values of a,~a, arc desirable. Figure 10.25 is a plot
of amplitude illumination across the aperture for as/al values from Oto 0.5,
calculated on the assumption that the horn length is correctly chosen to

lo

~ ire:: 0 0.2 0.3 0.4

0.1
a3 a,
/
1Ic,.10.26.Rel~tivegain 01 a box
0.5

horn
0 0.1 0.2

FIG. 1027.-Magnitude
0.3
a3 a,
/
0.4
of the first side
vs. asjal. lobe m the power pattern of a box horn vs.
as/a,.

make the two propagating modes 180 out of phase at z = L With respect
to their relative phase at z = O. The curve for a, al equal to 0.3 is a
fairly good approximation to uniform illumination.
, The H-plane radiation patterns in the Fraunhofer region have been
calculated from scalar diffraction
theory (Chap. 6) for the amplitude So
distributions of Fig. 10.25. The
results are summarized in Figs. ~
10.26 to 10.28, n-here Fig. 10.26 ~ VA
sho\vs relative gain, Fig. 10.27 the ~ ~0 \ 1;
magnitude of the first side lobe in g
~ .1:
the power pattern in decibels ~
dotrn from peak, and Fig. 10.28 > ~ .2
the full angular ~vidth of the pat- ~o. \ > ~ \ Z$
tern at tenth power. Gain is seen ~
~ .2+
to be maximum in the neighbor-
0.1 0,2 0.3 0.4 0.5
hood of aS/a, equal to 0.35, where 0
a/a,
the amplitude distribution across
FI(+.102s,-~ull mdth at tenthpowerof a
the aperture approximates uni- ImxhOrllpatternvs. a 3/al.
formity. In this region also the
first side lobes are approximately 13 db do\\-n,the theoretical value for
uniform illumination. Illuminating the edges of the aperture more
strongly increases side lobes and cuts down gain, although it somewhat
increases the directivity.
The length L of the box is obtained from the relation

r = (BI 133)L, (78)

380 WA VEGUIDE AND HORN FEEDS [SEC. 10.16

where

1=:[1 -(4TJ (79)

3=:[1 -w] 1

The ratio A/A can be varied over wider limits than that set by the
uniform guides by means of a sectoral guide transition as shown in Fig.
10.23b. In this case, ho\vever, because of the cylindrical waves in the
sectoral guide, a phase error is introduced into the field over the plane
of the junction with the box horn. This error is not taken into account
in Eq. (77); it is usually neglected in designing the horn. From existing
measurements it is difficult to evaluate the influence of this factor on box
horn performance. It is found true in general that the effect of replacing
a plane wave front by a cylindrical one is barely detectable experimentally
where Ar5/x. is r/4 or less. The same criterion would be expected to be
valid for box horns. In terms of A and 0,, it is written [Eq. (49)]

The use of a box horn is, of course, limited to applications requiring

H-plane tenth-power widths from about 36 to 70. Its principal advan- ?
tage is compactness. The contrast in size between a box horn and flared
horn of the same tenth-power width is greatest in the region of small
apertures. Here, near cutoff for the T~,-mode, L is small because
(01 BJ is approximately equal to dl. A box horn ~vith A = 1.6A,
as/al = 0.5, and @(~) = 67 is 1.3A long, while the flared horn of the
same tenth-power width is twice as long. However, for A from 2.OA to
2.5X, a box horn is only about 20 per cent shorter than the corresponding
flared horn. The box horn is especially useful therefore for applications
requiring tenth-power widths from about 55 to 65, with apertures from
1.6A to 1.7A. A flared horn has an advantage over the box horn in hav-
ing side lobes 5 to 10 db lower, although this is of no concern in many
applications.
10s16. Beam Shaping by Means of Obstacles in Horn and Waveguide
Apertures. There are antenna applications requiring very broad or
very narrow feed patterns for which waveguide and horn feeds are not
strictly suitable. Thus, to obtain primary patterns in the H-plane with
10-db widths n-eater than about 120, the 10-db width obtained from
ordinary waveguide, special beam-shaping techniques are required. At r
the other extreme of very narrow 10-db widths, horns become too bulky
for many applications. To solve these problems beam shaping by means $
] Mason, op. cit., p 18, Fig. 13.
SEC. 10.16] BEAM SHA PING BY MEANS OF OBSTACLES 381

of obstacles in horn and waveguide apertures has been investigated and

techniques developed. 1
a. Beam Broadening: H-plane.
To obtain a broader pattern than I
that of ordinary waveguide, the
device shown in Fig. 10.29 is used. I
A metallic post is inserted across
the waveguide just inside the aper-
ture at the center, and the wave-
guide corners are removed by
symmetrical cuts AA. Figure
10.30 shows the 10-db widths ob-
served with this arrangement on
3.2 cm guide for a series of values ,
of C and O, where C is the dis- I

tance from the waveguide edge at

FIG. 10.29.Device for broadening the
which the diagonal cut begins and pattern in the H-plane.
@ is the angle of cut. Figure
10.31 shows a comparison of a waveguide pattern with a broadened
pattern

300

-T
z

2od

100

~, deg.
Fl{;. 10.30 .l O-dll widths of the pattern obtained with the device of Fig. 10.29 as a function
of C/k and @(k = 3.2 crll, a =0.900 cm, b = 0.400 em, d = 0,063 cm).

The wavefronts, or surfaces of constant phase, from any device yield-

ing a very broad pattern must receive critical examination by the antenna

C. S. Pao, Shaping the Primary Pattern of a Horn Feed, RL Report No. 6L5,
.Tanuary 1945.
382 1~.1JEGUIlih .1 .Vl) JIOH.\- FEEDS [SK(. 10.16

designer. lhe gain obtainable frf;m a properly illuminated secondary

reflector may not be realized if phase irregularities exist over its aperture. .
100 0--- 0
/ - 2
~ 50 \ - 4
/ / \
: ~ (1) (2) - 6n
g /1
~ \ A- :
% / I I I \ I I - Zal
35
. / \ ,- 45
( I \i
g \ / 6ti
\ I \ / .
G \ I \ / 8~
3 \
g 1.0 / L -- A 20 ;
\l \ / 2- ~
.: 0.5 t
Q 1,/ - 4$
1% 6
- 8
0,1 30
1600 -120 -800 400
00 400 800 1200 160
Angle
FI~, 10.31 .Primary~$ys of (1) an open wa~.ecuide and (2) a cut corner waveguide
.
- (+ =%3.4 and C = *)

The tendency to irregularities on the phase fronts of wide-angle devices

arises from the fact that they usually consist of several radiating elements
~~ith noncoincident directive pat-
terns located an appreciable frac-

tion of a wavelength apart. An

analysis of a simple case (Fig.
10.32) }vill illustrate this. Con-
sider two identical radiating ele-
ments, 1 and 2, separated by a
distance 2d which have pattern
maxima in the Fraunhofer region
at angles 01 and O*with the for-
I 1(; 10.32, Phahe distortion ill the fields of
urilnary feed. of low dlrcwtivity.
ward z direction, respectively.
Then letting f, = (1/r)f(O .9,)
and ,fl = ( 1lr)./(o (1?)equal the amplitude functions of the two sources,
the amplitude and phase on a circle of radius R about 0 is given by
E = ,flel~t-~]$+~cl,,m
6) + f2e, (ulLl{kd,iD .9)

= [,f; + .f~ + zjIf2 COS (2kd sin e)])eiti-l ei, (81)

}\here
tan ~ = f-_: f~ tan ~kd sin ~),
(81a)
f, + f,
If j, = j,, the phase is constant on a circle of radius R. Ho~ve~-er, if
f
~ ,~h{~~et
e pt]ae at angle o departs from its value at o = o by an ang](.

f, j,
A = tan- ~, + ~ tan (kd sin @) . (82)
[ 1
 SEC,10.17] PRESSURIZE.VG A .}.0 ,tf.41CHI,VG 383

For large values of kd, A may exceed the allowable deviation from con-
1 stant phase, particularly if .fI is appreciably greater or smaller than jj.
Usually the beam shaping need not be carried to the point where the
phase departure A has to be reckoned with.
b. Narrowing the Primary Pattern: H-plane .The pattern in the
H-plane is narrowed -when several metallic pins are placed in the aperture
of an H-plane horn. For example, the 10-db width of a 3.2-cm horn with
a flare angle 30 and magnetic plane aperture 2X decreases from 78 to
56 when two +in.-diameter pins are placed just inside the aperture
* at a distance 0.44x on either side of the center. The impedance match
is improved rather than impaired by the presence of the pins. A number
of other arrangements also have been found to be effective. 1

llG. 10.33 .FlanR for E-plane beam FIG. 10.34.Strip for narrowi,,g of the
shaping. beam in the E-plane.

Beam Shaping: E-plane .Since ordinary waveguide has an E-plane

aperture of about k/3, there is rarely a need for special techniques for
beam broadening. However, the form of the pattern can~j.np~oyed
.....
a flange a~ roximately ~~~~~;~n=he
by adding.- E-plane as shown in
Fig. 1033. With the flange~~~p=~= rnoi-e;api~ly- at small
fingles from the forward direction and less rapidly at large angles, thereby
yielding a more nearly uniform illumination across the secondary aperture.
Narrowing of the E-plane pattern has been accomplished by inserting
a metallic strip with a considerable H-plane width, as shown in Fig.
10.34. This strip, however, causes a troublesome mismatch, 1 and it
seems questionable from the data if it is more effective than straightening
the phase fronts in the horn by making the horn longer (Sec. 10.10).
9 10.17. Pressurizing and Matching.In most applications waveguide
and horn feeds must be at least weatherproof and preferably capable of
holding pressure. Several techniques have been evo!ved to utilize the

1 Pao, op. cit.

384 WA V.YG(lDE AND HORN FEEDS [SEC. 10.17

pressurizing or weatherizing enclosure as a matching device. One of the

most successful is illustrated in Fig. 10.35. The enclosure, roughly .
cylindrical in form, is placed over the end of the waveguide or horn with
the axis of the cylinder approximately in the aperture plane. LIatching
is accomplished by adjusting the thickness t and position d so that the
reflections fror,~ enclosure and feed are equal in magnitude and 180
out of phase in the direction of the generator, While t and d can be

I~IG. 10.35 .-Press~lrizil]g and matching enclosure,

calculated to a rough approximation for a feed with a given mismatch,

the matching procedure is actually empirical. Cylinders are formed with
~hicknesses and radii ranging in value about the calculated t and d.
Each cylinder is placed over the mouth of the feed, and impedance
measurements are made for a series of d values. From these measure-
ments the final choice of enclosure parameters is made. Figure 1036
is a typical impedance plot obtained during this procedure. It will be
observed that the best impedance match for this example is obtained ]~ith b
an enclosure of thickness 0.030 in. and mounted at a distance d = 1.2 cm
(Curve III), It is desirable, although not essential, that the enclosure
ra,ii ~E be approximate y equal to d, The mounting flange is positional
 SEC.10.17] II{ESS(RIZI. VG A.VI) MA7(HI.t-G 385

behind the aperture plane and a position can be found that improves the
1
feed SWR; it should be in place during the determination of enclosure
parameters and the pattern of the horn.

1(1.36, Impedance of a typical horn as a function of the positioll of the pressurizing

enclosure for differellt wall thicknesses.

The materials used for pressurizing enclosures have been low-loss or

glass thermoplastics: plexiglas or lucite

[ =k)
ndexofefraction =601
for wavelengths above about 6 cm, polystyrene (n = 1,60) and styraloy
(n = 1.60) at 3 cm and above. Laminates can be used, but because of
the high dielectric constants of these materials the enclosure walls are
thin. At 1.25 cm Corning 707 glass has been used for two-dimensional
pressurizing enclosures, but it presents obvious, although possibly not
insuperable, difficulties in three dimensions. It also necessitates building
the feed of metals with low coefficients of thermal expansion, such as
invar and covar. llastic enclosures are sealed by means of gaskets;
glass by platinizing or bonding metal to the edges and soft soldering.
For a given material the wall thickness f and position d of an enclosure
can be estimated from the measured reflection coefficient

rp(= lrFle~$) (83)

386 WA VEGUIDE AND HORN FEEI),7 [SEC. 1017

of the feed referred to the aperture. The assumption is made that the
reflection coefficient rE from the enclosure is the same as that of an
infinite sheet of the same material for plane waves. The latter is given
by

r,=~
~2_l
sin

~ + (n24j, 1) sin,
()
2?rnt
A
2~t
$5e
j{.+h.-, [7*IC0t(%)l}

[ ( )1
(84)

On this assumption the feed will be matched when the thickness I is

such that

712 1
2n ~ + (n
sin

()27rnt

l) k.Slnz 2irnt
(85)

[ 4?12 ( A )1

and the distance d satisfies the relation

d 1 tan_, 2n
cot (tit - @l. - 27M (86)
A=k -{ [ ~2+1 \A )1 }
where m is an integer. If the enclosure were in the Fraunhofer region and
were designed so that its surfaces coincide ~vith equiphase surfaces from -v.
the feed, the assumption involved in using Eqs. (84) to (86) would be
justified. Because the enclosure is actually close to the feed aperture
and cuts across equiphase surfaces, a given thickness t corresponds to
smaller 117~1than is indicated by 13q. (85). For this reason also, the
average path from the aperture to the inner surface of the enclosure is
less than d, so that the experimental optimum d is larger than the value
indicated by Eq. (86). In fact, for feeds with small E-plane apertures
(B < 3A/4), where there is a capacitive mismatch localized at the aperture
(or is approximately 2700), the value of d calculated from Eq. (85) is
1/4; experimentally d is about 0.351 for a number of typical cases.
Where the feed mismatch is small or the wavelength very short, it may
not be practical to use the smallest thickness t, calculated from Eq. (85).
For these cases, as inspection of Eq. (85) will show, a reflection of the
same magnitude can be obtained by using thicknesses [t] + (h/2n)];
the spacing is unaffected.
The use of an external pressurizing enclosure has a number of advan-
tages to recommend it. The possibility of breakdown at high power is
minimized because the pressurized region extends beyond the feed aper-
ture and the dielectric housing is located in a region of low field strengths.
The impedance match is reprodllcible in the sense that it is independent
1( Waveguide Handbook Supplemmt,) RI, Group Report No. 41, .Jan. 23, 1945
 SE( 1017] IRE,VLS1 lTlZI.VG .t .Vl) .11.17(HI. V(7 3s7

of small fluctuations in enclosure dimensions and positioning. This iS

\ in distinct contrast to the properties of pressurizing diaphragms \vith
flanges in the plane of the aperture, ~~herethe geometry of the flange and
even the tightness of the retaining bolts have an dfect on pattern and
match. As a matching device the pressurizing enclosure is especially
effective when the mismatch of the feed arises at the aperture. lhis is
true for open waveguide and properly designed horns whose elrctric plane
aperture is less than about 3A/4. 1hen, since enclosure and apertllrc
are closely spaced (cl -- A/3), the impedarwe match is insensitive tu lrc-

----- 0.1

1:1(,. 1o:17.-

quency changes in which Ax/k is of the order of 10 per cent. ~he impeci-
ante-wavelength curve of a properly designed pressurized horn is she\\-n
in Fig. 10.37 to illustrate the bandwidth of the device.
Another successful pressurization technique Irhich ~as received only
preliminary triall consists in soldering a diaphragm of (orning 70i glass
with a bonded metal rim inside the feed near the apcrtum. The glass
window is somewhat smaller than the inside dimensions of the feed.
,
The thickness of the glass and the dimensi(ms of the opening (an be
chosen to make the effect of the diaphragm resonant, inductive, or capaci-
tive so that any feed can be matched by this techni{lue.
I 3[, I). Fiske, Resonant l~indows for Vficuun~ Seals in ltcctangular J\-ave-
guides, Rev, Sci. {;strurnents, 17, 478 (1946).
 CHAPTER 11

DIELECTRIC AND METAL-PLATE LENSES

BY J. R. RISSER

11.1. Use of Lenses in Microwave Antennas.-The utilization of

optical met hods is an outstanding feature of microwave antenna design.
It is natural, therefore, to consider a much-used optical device, the lens.
Dielectric lenses of conventional optical design are, in general, too cum-
bersome for use in microwave antenna systems, but when they are zoned
so that the dielectric is nowhere more than several wavelengths thick,
their use is a distinct possibility. At wavelengths in the microwave
region, a practical lens can also be constructed using parallel metal plates
spaced a fraction of a wavelength apart, because for radiation with the
electric vector parallel to the plate surfaces, the space between the plates
is characterized by a longer wavelength than the free-space wavelength
and consequently has the properties of a refracting medium with an
index of refraction less than unity,
Lenses and reflectors are interchangeable in microwave antennas,
v.
because both perform the same basic functionmodification of phase.
Thus, for example, lenses can be substituted for reflectors to produce
pencil. beams; cylindrical lenses of suitable contour can be used with line
sources to obtain asymmetrically flared beams; and line sources can be
formed by the use of two-dimensional lenses between parallel plates.
Reflectors have many advantageous features: mechanical simplicity,
lightness, and freedom from chromatic aberration. Lenses, in turn, have
characteristics that render them invaluable for many applications. Thus,
in a lens system, the feed is out of the path of the main beam, a considera-
tion of particular importance in parallel-plate line sources. Lenses are
also particularly suited for insertion into optical systems to perform special
corrective functions; for example, correcting lenses of the Schmidt type
can be used in conjunction with a reflector to obtain a wide field in
antennas for rapid scanning, and metal-plate lenses are used for phase-
front correction in sectoral horns.
Because of the difference in wavelengths, microwave lens techniques
are free of certain restrictions which obtain in optics. Surface tolerances
are large. Dielectric lenses can be made of relatively soft thermoplastics,
such as lucite or polystyrene, instead of glass, and the lens surfaces can
be turned on a lathe or molded. ~onsequently, the surfaces need not,
be spherical but can be cut to contours appropriate to the function of
388
 I
sEC. 11 .2] PRINCIPLES OF DESIGIV 389

the lens. Metal-plate lenses, likewise, can be produced by ordinary

machine-shop methods. Since the fundamental function of a lens is to
modify the phase fronts from a radiating source, the lens surfaces are
designed using the laws of geometrical optics. However, the radiation
pattern of the antenna as a whole-lens and primary sourcemust be
considered from the standpoint of diffraction theory. The far zone field
produced by the lens is obtained from the amplitude and phase distribu-
tion over its aperture by the methods of Chaps. 5 and 6.
It is the aim of this chapter to point out the methods of design, types
f of structure, and general problems involved in the use of lenses. Cor-
recting lenses and other lenses designed for special purposes will not be
considered. Discussion will be confined to those lenses whose function is
to convert the spherical (or cylindrical) phase front from a point (or
line) source at the focus of the lens into a plane phase front across the
aperture. This is the most frequently recurrent problem in microwave
antenna design, because, by diffraction theory, a plane phase front results
in the most directive pattern for an aperture of a given size with a given
amplitude distribute on across it.

DIELECTRIC LENSES
11.2. Principles of Design.The general principles of geometrical
optics were formulated in Chap. 4. Lens design is based on two of these
-4. principles (Sees. 4.8 and 4.9): (1) the principle of equality of optical paths
along rays between pairs of wavefronts and (2) Snells law of refraction.
The procedure of lens design is commonly referred to as ray tracing
because it deals exclusively with the optical paths or rays, that is, the
normals to the equiphase surfaces or wavefronts. In a homogeneous
medium the rays are straight-line segments. In empty space the optical
path length is just the length of the ray segment; in a dielectric medium,
it is the length times the index
~]f refraction n (equal to v%).
The rays are refracted at the
lens surfaces in the way described
by Snells law. The ray-tracing
method consists in determining
the lens surfaces, so that the
11~. 1l.1.O@lcal paths and equiphase
combined optical length (1I + surfaces.
nl~ + 1S) (see Fig. 11.1) along
any one ray between t\vo equiphase surfaces ASIand SZ on opposite sides
of the lens is the same as the length (l; + nlj + 1~)along any other ray
+ between SI and S2. The reciprocity theorem can be invoked to show
that it is immaterial whether the direction of propagation is from S,
to\vard Sz or the reverse; a lens designed to convert a spherical equiphase
d
surface from a point source F, located to the left of the lens, into a plane
 ~
4

390 DIE1,E(71(lC A ,Vl) .111<7.~1,-I J>,I 7E J,ll,Y,?E,~ [SEC.113

equiphase surface to the right, will function equally \vell in bringing to a

focus at F the energy in a plane \vave incident from the right.
11.3. Simple Lenses without Zoning. Simple lenses will be divided
into two categories according to the number of refracting surfaces, If
one lens surface is an equiphase surface of the incident or emergent ~vave
with the result that the rays are normal to the surface and pass through
undeviated, the term <one-surface lens ~villbe applied. A two-surface
lens is one in which refraction occurs at both lens surfaces. The design
of a one-surface lens is a relatively simple problem and \villaccordingly be
treated first.
The first to be considered is shown in Fig. 11.2. The boundary TT
between air on the left and the dielectric of index of refraction n on the
right is to be determined so that the phase front SI from a source at F
is converted into the plane phase front SZ in the dielectric. The dielectric
is terminated on the right in a plane parallel to S2. The problem is
essentially tivo-dimensional, whether the lens is cylindrical and F is a
line sourcein which case the line source and cylinder axis are perpen-
dicular to the plane of the diagramor the source at F is a point source
:md the lens is spherical, 1 that is, has rotational symmetry about the
optical axis FQZ, the line through F perpendicular to Sj, The equation
of the lens surface is obtained from the condition that the optical path
length [(PIP) + n(PPZ)] through an arbitrary point P shall be equal to
v.
the optical path length [(Q,Q) + n(QQ,)] on the axis or, more simply,
that
(FP) = (FQ) + n(QQ4). (1)

In terms of FQ ( = j) and the polar coordinates (r,d) of the point P,

Eq. (1) can be written
r=j+n(rcoso j). (2)

Solving for r, the equation for 7T, the generating curve of the lens
sllrface, is given by

r=
(n [)~. (3) ,
?1 COS @ ]

Since n > 1, this is the equation of a hyperbo]a of eccentricity n with

the origin at one focus. The asymptotes make an angle 9. with the lens
ayis given by
e= = CoSl:. (4)

~hc angle dais 51 for polystyrene,and plexiglas for each of which n = 1,6.
It may be noted that the la\r of refraction is not used in deri~ing Eq. t
1For convcnienre, a I(v]s \rill he dcsi~n:~tctl :]s spl)crical if its s{lrfaccs zrc
geuerated tjy a rotation almtlt the axis and cylindrical if tl)c gencrzting motion is u
tratlslation p:irallcl to the Llrle source II.
 4
i,
SEC.11+] SIMPLE l, ENSES JfITHOIJT ZO,VIIVG S(J]

(3); it is easy to show, however] that the law of refraction is satisfied

at this boundary for the ray construction of Fig. 11.2, and it is, in fact,
known from general considmations (Sec. 4.9) that it must be satisfied,
It is likc\visc easily seen that the right-hand boundary of the dielectric
has no effect on the optical paths because it coincides with an equiphase
surface and, consequently, is normal
T
to the rays PPZ, QQZ, etc.
In evaluating the usefulness of
the lens, its effect on the amplitude
I
dist rihut ion o~-er the aperture must
be ascertained, because this as ~vell
as the phase determines the diffrac-
F
tion pattern. The effect differs in
spherical and cylindrical lenses. For
a spherical lens ~vith a point source
at F that has an axially symmetric
pattern, if P(d) is the po\\er radiated
per unit solid angle by the point 1-
s{)~lr(,e in the O-direction and F(o) l1,;. 112. - ol,p-hurfaw II-,,. with Ill ,,,r-
I,olic rol)tour,
the c{)rrcsponding po\ver per unit
artw in the aperture at, a distancr from thr axis P( = r sin 6), then from
9. geometrical considerations,
P(p) sin /3df9
(5)
P(o) p (/p

Refiertitnl a( i he lrns s~lrfare i~ being neglected. For the hyperbolic

surface generated by the cllrvc of l;q. (3)
P(p) (n ros e 1)
(6)
P(o) = 7(n 1) (n Cos 0)

}vhereas the corresponding amplitude ratio is

(7)

Iror a cylindrical lens ~vith a line source at F, F(o) do is the power radi-
ated per ~lnit length I)y Ihe line s(nlrce bet~vem the angles o and @ + de.
Then if F(!/) d?yis the po~vcr per unit lrngth in the corresponding aperture
interval Iwt\vcen y and y + dy, where y is again the distance r sin d
fmrn the axis,
P(y) _ g
(8)
P(e) fly
t
IJor the hypcrboli( s~lrfare,

J(V) = __()/ (os e 1)

(9)
r(e) (,, l),f()/ (0s 0)
 ~
d
,.
I

392 DIELEC1RIC AND METAL-PLATE LENSES [SEC. 113

and
A(Y) _ (n cm e 1)
(lo)
A(9) J (n l)f(ncos e)
The amplitude ratios for spherical and cylindrical lenses normalized
at tJ= O are plotted in Fig. 11.3. They drop off rapidly with increasing
8, an effect that impairs the usefulness of
the lens. For the spherical lens of poly-
styrene or plexiglas, the aperture ampli-
tude has dropped off 50 per cent relative
to the feed amplitude at an angle of 30
with the axis, with the result that it is
scarcely feasible to use a lens aperture
extending beyond this point because of the
high degree of taper in the illumination.
This results in a serious reduction in the
gain and an increased width of the main
lobe of the antenna pattern (cf. Sec. 6.6).
Another design for a one-surface lens
o 20 40 60 can be obtained by considering the source
0, deg. F (see Fig. 11.4) to be immersed in a di-
klG. 11.3. A(p) /A(O) for a
electric medium of index of refraction n
spherical lens and A(y) /A(fl) for a
cylindrical lens wkh hyperbolic bounded on the right by a dielectric-air
contours; n = 1.6. surface. In this case the equation for the
generating curve TITZ is found to be
,= f(nl)
(11)
n Cos e
where j is the distance along the axis from the focus to Tl!la. This is
the equation of an ellipse of eccen-
tricity l/n with the origin at the
T1
focus farther from Q. An actual
lens would be constructed, as shown
in Fig. 11.4, where the source F is
outside the dielectric and the inci-
dent dielectric surface T; T; is F
. ,_
spherical or cylindrical, as the case
may be, and normal to the rays FP
and FQ, For a given focal length,
the aperture of the lens cannot be
larger than 2b, where b is the semi-
T; I
minor axis of the ellipse:
FIC. 11.4.One-surface lens with elliptical
contour. i
b= $ j. (12)
d
 393

The semiminor axis b subtends an angle 9~ at the feed given by

1
em = COS1 .
n () (13)

The minimum ratio of the focal length to the diameter is

(14)

\vhich is 1.04 for plexiglas or polystyrene. For a spherical lens ~vitb an

I
elliptical contour, .4 (P),.4 (o) is given by

A (P) _ (?3 Cos e)

(15)
A(L9) J (n 1) ~(?z (0s e 1)

and for a cylindrical lens by

A (Y) = (n Cos 0)
(16)
@ej d (n l)j(n cos 8 1)

Plots of these amplitude ratios normalized to unity at e = O are given in

Fig. 11.5. The amplitude in the aperture increases relative to ~he feed
amplitude \vith increasing d. This property of the elliptical contours is
desirable for micro\vave work, because the lens compensates for the
~ directivity of the feed pattern, producing ~..
a more efficient illumination over the
aperture from the standpoint of antenna
gain. On the other hand, the more uni-
form illumination enhances the side-lobe
/
structure of the pattern, and the lens is not
suited for use with an antenna feed of too
2.0 T
low a directivity.
Practical considerations of bulk and A(8)

efficiency place a limit on the useful aper-

4 ture of this lens. For polystyrene or plexi-
glas, the maximum diameter 5b is 0,96f,
and the half angle L subtended at the feed
F is 51. For a 30 half angle, on the 10o 20 40 60
other hand, the lens diameter is 85 per e,deg.
I:lG. 11.5.~(P)/A(d) for a
cent of maximum; and in the case of the sphericallens and A (v) /A (6) for a
spherical lens, the volume of dielectric is c~lindricallens with elliutica] co,, -
tours;n = 1.6.
approximately one-fourth of the volume
with maximum diameter. For a spherical lens cut at 8 = 30, A (P)/A (d)
1
is 1.7, with the result that the feed pattern must be donm 14.5 db from
0 = O in order to make the power at the edge of the aperture 10 db down
: from center.
 394 DIELECTRIC A.VI) MET.41]-PI,A TE I. E.V,$E,V [SEC. 11.3

A two-surface lens is one in which neither of the lens surfaces

coincides with a surface of equal phase and the rays undergo refraction
upon entering and leaving the lens. The use of a second refracting sur-
face is one way of increasing the versatility of the lens. The optics of a
two-surface lens will be outlined briefly, although there will be no detailed
discussion of an example. Figure 116 is a sketch of the geometry. As
before, F is a point or line source to
Iq
TI T2 the left of the first surface whose con-
X,yl tour is denoted by 111~. Since re-
~ fraction occurs at both surfaces, there
Ir
Oe x is no unique form for the equiphase
F It is con-
surfaces in the dielectric.
n venient in this case to discuss the
problem in terms of the angle 6 (Fig.
+ 2; T; [ 11.6) which a ray entering the di-
FIC.116.-Comdinate systemfor a two- electric at the point (r-,@ makes with
surfacelens.
the axis of the lens. The angle 0 is
determined from the form of the contour TI 2{, or vice versa, in accordance
with Snells law, In terms of 0 the differential equation for the contour
T,7~ can be shown to be

1 dr ~ sin (0 0)
(17)
;G= n Cos (0 0) 1

The coordinates z, y of the point where the ray from (r, 6) intersects
the second surface TZTj are determined by the geometrical relationship

yrsin O=tan O,
(18)
xr cost?

and by the condition for the equality of optical paths. Since the equi-
phase surfaces to the right of ZjTj are required to be planes, the condition
on the optical paths is given by

r +n <(g rsin 0)2 + (z rcos 0)2 z = constant. (19)

As is easily seen, Eqs. (17) to (19) are not sufficient to determine uniquel,v
the coordinates of both surfaces. Another condition, essentially equiva-
lent to a condition on W, may be imposed. For example, it may be
required that the lens be free of coma (to render it suitable for use in a
scanning antenna) or that the amplitude ratio A (y)/A (0) of Eq. (5) or
(8) be specified as a function of y or 6. For a general-purpose microwave
antenna it would be desirable that the amplitude ratio be constant or at
least a slo~vly varying function of y or O,~vith the result that the taper in
the angular pattern of the feed is reprod~lced in the aperture. As far as is
known, a practical sol{ltion of this problem has not, yet been obtained.

-% -- - -= _ __
SE(J.11.4] ZO.VED DIELECTRIC I,E.J~,TE,5 395

lVhen the contour 1,1~is determined by arbitrarily setting 8 equal to

8/2, it is found that A(y)/.4(@) is very nearly constant for a cylin-
drical lens of moderate aperture and only ~-ery slowly varying for a
spherical lens. This lens is thick, ho\vever, for reasonable apertures.
In general, \viththe large aperture suse dinmicron-ave applications, con-
siderations of weight and bulk make zoned construction a practical
necessity. Consequently, the design of lenses without zoning is of some-
what academic interest, and the conditions on the lens s~lrfaces are not
complete without provision for the zone steps. This will be treated in
the next section.
Before leaving the subject of unzoned lenses, however, we may take
note of an expression for the thickness of a simple converging lens on
the axis. This expression is useful in estimating lens proportions,
It can easily be seen from the principle of equality of optical paths that
if R and 60are the coordinates of the apex of the lens ~vherethe dielectric
reaches zero thickness, the thickness t on the axis is given by
R(I (0s O.) = (n l)f. (20)
For example, a polystyrene lens ~vith R equal to 20 ~vave]engths and 00
equal to 30 is 4.5 wavelengths (free-space) thick,
11.4. Zoned Dielectric Lenses.A simple lucite cr polystyrene lens
of the type described in the preceding section is man~, ~~p.velengths thick
if its focal length and aperture are large
compared with a wavelength. For a
simple lens the optical path length along
the axis is the same as the length by way
of the edge. This condition is unneces-
sarily restrictive, however, at microwave
frequencies w-herethe wavelength is large
compared with ordinary manufacturing
tolerances. The surfaces of microwave
lenses can be di~-ided into zones \rith the
optical paths differing by integral multi-
ples of a wavelength from one zone to (u) (b)
another. .L lens may accordingly be FIG. 11.7. -Zoning of nonrefrarting
lens surf arcs
designed ~vith its cross section similar to
those shown in Fig. 11.7, Starting ~vith zero thickness at the ccfgc of the
lens the thickness of dielectric may be progressively increased to!~arcf the
lens axis, as required by the phase condition, until the path difference intro-
duced by the presence of the dielectric is equal to a ~!-avelength. .%t this
point the path in the dielectric can be reduced to zero without altering
the ~vavefronts from the lens. This is then the outer boundary of another
zone, through ~vhich the optical path lengths are one ~vavelength less
than those throllgh the outermost zone. This zone Iike\visc increases
396 DIELI!CTRIC AND METAL-PLATE LENSES [SEC. 114

in thickness in the direction of the axis until the point is reached where
reduction of the dielectric thickness to zero results in an optical path
length smaller by another wavelength, and so on. The resulting lens is
similar to a conventional zone plate except that the path difference
between zones is equal to a single wavelength. When the lens has K
zones, the optical paths through the outermost zone are (K 1)A longer
than those passing through the zone on the lens axis.
The maximum thickness of a zoned lens is approximately equal to
h/(n 1), because the maximum path difference (n l)t introduced in
a path of length t by the presence of the dielectric is approximately a
wavelength. In actual practice a small thickness t~ (Figs. 117 and 118)
must be left at the thinnest points for reasons of mechanical strength,
so that the maximum thickness is greater than h/(n 1) by this amount.
A good example of a zoned two-surface lens which has been tested
and used is shown in Fig. 11.8. This is a plexiglas lens, 13.5 in. in diam-
eter, for use at 3.3 cm. The surface toward the feed is chosen somewhat
arbitrarily to be a plane. This choice has much to recommend it, how-
ever, because a. plane surface should have somewhat less back reflection
to the feed than a concave surface and better illumination characteristics
than a convex surface. Except for the inclusion of zones, the lens is
designed in the manner described for two-surface lenses in Sec. 113.
Once a plane for the first surface and the distance from this plane to the
focus (6 in. in this case) are chosen, the lens structure is completely deter- d

mined by the requirement that the equiphase surfaces to the right of the
lens shall be planes: Snells law determines the directions of the rays in
the dielectric of the lens; then the second surface is determined from the
principle of equality of optical paths with the provision that the paths
differ by a wavelength from one zone to another. The following equa-
tions describe the zoned surface:

~= (Kl)A+ (nl)D+j~j2+r1
rz ]- 21)
n l
[ d nz(r + j)
and
d
R=rl+ (22)
[- ~n(fl + r) r 1

The notation is that of Fig. 11.8. The zone number K is unity for the
central portion of the lens. This zone is carried out from the axis to a
point where the thickness of material L is considered a minimum for
mechanical strength. There the step is made to the surface determined
by K = 2, and so on to the edge of the lens.
1A. M. Sbwllptt,(PlexiglasLens Antenna for Micro
waves, BTI, Report MM-43-
170-15,September1943.

.
SEC. 11.4] ZONED DIELECTRIC LENSES 397

In designing a zoned lens care must be taken to avoid excessive shadow

area between zones inthe lens aperture. Shadow invariably occurs when
zone steps are cut in a refracting surface of the lens because the ray

F1o. llfl. Zonedtwo-surface lens.

. .
F
F1~. 114-Shadow introduced by zoning.

adjacent to a step inside the dielectric has a different direction from that
of the adjacent ray outside. This is shown in Fig. 11.9 where the rays
FPP and FQQ, which are together on entering the lens, have undergone
considerable separation at the step. In transmission the aperture illu-
mination is zero between P and Q; in reception the energy incident
between P and Q does not reach F and is therefore lost. Shado\v does
~g~ l) IEI,I?CTRI(: A.VI) .llETil I,-II,ATh LE.VSE.T

not occur where the step is cut in an cquiphase lens surface because the
directions of the rays are nut altered at the surface and hence undergo (
no separation. Theeffect of shado\r regions, of course, shows up in the
secondary pattern of the lens as a decrease in gain and increase in side
lobes. Data are lackingat present on the r-elation between these effects
and the size and position of the shadmv rrgions.
Frequenr~ .~en~iti?il~.sin~e dielectric constants are independent of
frequency in the micro~rave region, an unzoned dielectric lens performs i
its function regardless of frequency. }}ith zoning, ho\\-ever,frefl~lcncy
sensitivity is introduced. Let P, be the length of optical path from the
focus F along the axis to any plane perpmdicular to the axis on the far
side of the lens. If Kis the total number of zones, the optical pathP, )
through the Kth zone is PI + (K l)Ao where AOis the design wave-
length. At wavelength ~, the radiation from the first and ZIth zones
wi!] be in phase in any plane perpendicular to the optical axis, IIo\\ever,
at a near-by wa~elength XO+ Ax, the \va\-cfront from the Kth zone \vill
be displaced a distance 8 along the axis relative to the wal-efront from t}~e
first zone, l~hen the changes in ~va~clength m-e small, 3 in wavelengths
is given by

\\hence
8 d
(K 1) :.
A=
The usual criterion for mirrul~avc work is that the displacement 6 shall
not (W((!ed (). 125A.1 [sing this criteriun and ddining ~he bundivi(lth as
t!rice the maximum allo!vable fractional change in lva~-elengtll exprcsswl
in per cent,
per ccut.
,Ia n d ,v i d t,l _ __~

(24)
K 1

.l lens of ! per cent bandlvid(h can have se~-cn zOII(,Sll-ith AOsteps bel \\ccu
zones. The formula is a})proxinmtr bec:tuse AA is assumed small in the /
(h?rivation. It should bc noted that (K ] ) is :Lct IIa]ly (IIC numlxv of
\\avclcllgt}l strps introdllced })Y zoning; if there ar( stops ot t!l-o or more
Ivavelengths, this must he taken into a(vount, It is belicrwl that 1;{[.
(24) gives a consrrvtitive estimate of ban(ill idt 11,bccallse \\ith talx,rcd
aperture illurninaticm, high(,r values of 8, k might be tolemted for (,c~rtair~
applications.
11.5. Use of Materials with High Refractive Indexes. Recently

1 1!1(, relat 1o11 hrt,ivren thc gain of a penril beam antc]ln:i :illrt the ]Ih:Iw (~rror I
II\-rr tllr :Ilwrt([rr, is (Ii.wl].sd in SeC. 12,5.
I
SEC. 116] I) IEI,E(TI<I(: LOSSES it .Vl) TOI>EI<.I .V(E.? 399

Irnscs, Two distinct typrs exist. Iolyglasl mixtures ~vith titanium

dioxide or titanate fillers ha~e refractive indexrs from 1.7 to 4.9, depend-
ing on composition. Io~rer factors am 0.002 to 0.003 at 101 cps. These
materials have the ad~antage of possessing cocfKcicnts of thermal expan-
sion near those of copper and brass. lery high refractive indexes (about
10) are obtained from titanium dioxiclc and tit anat e ceramics. z The
titanium ceramics have pmrer factors brlom- 103 at 3 X 10g cps.
dioxide
There is no fundamental obstacle preventing the use of these materials
for micro~rave lenses. At present lack of development of manufacturing
techni(lues for heat treating and molding of large samples is the principal
difficulty. The degree of control necessary for successful manufacture
is indicated in the reports on the rnatcrials. IIecause tolerances against
warpage and twisting are large for lens surfaces (Sec. 11.(;), ~ lens could
be made as an assemblage of small sections, in order to decrease the size
of furnaces and molds.
In addition to techniques of manufact~~rej matm-ials ~vith hi~h refrac-
tive indexes present several prtlblems to the lens designer. The tolm-
ances on lens thickness become important (SrC. 11,0), :md reflections
from lens surfaces result in prohibit i~ely high transmission 10SSW(Sec.
11.7) unless slwface-matching sections are adtied.
11.6. Dielectric Losses and Tolerances on Lens Parameters.In
evaluating the usefulness of micro]va~-e lenses, it is ne(essarv to consider
the practical problems arising from properties of lens materials and pos-
sible limitations in methods of manufacture. Attenuation in th(, dielec-
tric must be reasonably small. It must be possible to fabricate the lens
to satisfactory tolrrunces on the contours of both surfaces and on the
thickness. 31any materials ]1-ith suitable refractive indexes arc lossy at
microlval-e frequencies. The attenuation in decibels per (free-space)
wavelength in an unboundrd dielectric medium is gi~en by

where n is the index of refraction and c and are the real an(i imaginary
parts of ~he complex dielectric constant (c j,) characteristic ot l(,ssy
materials, (r~hc index ,)f refraction is the square root of c,, ,rhich is
the inductive capacity k, ordinarily q{lcted. ) llc ratio 6,16 is equal to
the power factor of the material lvhrn it is small u)rnparcd \rith Ilnii y
Since the maximum thickness times (n 1) is almllt a Navelen~th for :
zoned microlvave lens, the upper limit to the attcrl~lation in L zoned lens
1.1. von I{ippt,l, S. 31. l<i]l~sl>llry, :ITIII1,. C, JII,s.s(JII, I,{)!v ll{,rl)]:~l 1I\lm]lsioll
[last its, Sl)Iic 14-5:39, O(tol)(,l 1!)4.5.
~ .[. VOI1 Hippcl, Il. (i. llrr(,k(,nrl[lg~,, .1. 1. dr l{!(~tt(,\lll(,, Jr , J II. llrI)\\t)lo\u,
F (;, (tlesl(.y, (i. ester, 1, 1isz:i, :111(1\\. }{. \\lw])ll; ll, 4ili~ll 1)it,ll,t,(r][, Lollst:illt
(crao]ics, Kl)l{( !4-300, .\llgllll 1!)41.
 400 DIELECTRIC A.>]) METAL-PLATE LENSES [sEC. 11.6

is approximated by

A. = 27.3 : ~ (26)
() nl

which is about 70~/c for polystyrene or lucite and about 35e/c for a
dielectric with nequal to 4.5. Thus, power factors up to 0.003 can be
used without introducing more than a few tenths of a decibel attenuation.
For polystyrene (c/c = 0.0003) the ai,tenuation is negligible.
The tolerance to be placed on any lens parameters proportional to
the maximum allowable irregularity in the wavefronts, or equiphase
surfaces, formed by the lens. This isagain taken to be A/8. Thetoler-
ances on thickness and index of refraction are interrelated because the
compensation in optical path introduced by the presence of dielectric
of thickness t is (n I)t. Setting X/16 as the upper limit on wave-
front irregularities arising from variation in either t or n to allow for
variations in both quantities, we have, approximately,

A
(27)
t S 16(n 1)
and

(28)

Since (n 1)Lis of the order of a ~vavelength for zoned micr~wave lenses,

Eq. (28) becomes
(29)
i%+
The tolerance on thickness becomes important only for materials with
high index of refraction. For polystyrene at a wavelength of 3.2 cm,
Eq. (27) gives a thickness tolerance of ~ in., whereas for a substance
with a refractive index of 4.5, it is 0.020 in. at the same wavelength.
As regards the dielectric constant, a variation greater than 3 per cent is
i
not likely to occur when reasonable care is taken in manufacture, even
in materials of high dielectric constants. This variation is ~vell within
the limits prescribed by Eq. (29). If variation in thickness alone or
dielectric constant alone is considered, the tolerances given by Eqs. (27)
and (28) may be increased by a factor of 2.
Some restriction on the surface contours of a lens arises from the
tolerance on the thickness t. The two surfaces can be deformed simul-
taneously by warping, however, without affecting the thickness appre-
ciably. From Fermats principle, the length of optical path through
any portion of a lens has an extremum value and consequently, small
displacements of any section of the lens result in changes in optical
path that are small compared with the displacements, The tolerance
 ,
SEC. 11.7] REFLECTIONIS FROM DIELECTRIC SURFACES 401

on a surface contour, except when it affects thickness, is therefore at

least as large as the maximum allowable irregularity in the wavefronts.
11.7. Reflections from Dielectric Surfaces. Reflections from dielec-
tric surfaces can cause feed mismatch and power loss. Feed mismatch
I
is most likely to occur when a lens surface coincides with an equiphase
surface because the reflection from the entire surface is then in phase at
the feed. In this case the surface reflection coefficient is given by the
well-known expression for normal f
incidence

0.50 /
R=~. (30)
n+l
Equation (30) yields a value of 0.23
forplexiglas or polystyrene and, of 0.40
course, larger values for higher di- a
electric constants. A reflection of
0.30
this magnitude is too large if picked IRI2
up by the feed, and so the use of an
equiphase surface as a lens surface I
0.20
is to be avoided whenever possible.
For high indexes of refraction
and large angles of incidence, power b
0.10
loss itself becomes important. For /
a refractive index of 4.5 the loss is
40 per cent at normal incidence, so o
o 30 60 90
that surface-matching sections are Angle of incidence
necessary. The reflection coeffi- FIG. Ill O.-Fraction of incident power
cient R depends not only on n but reflected from the surface of an infinite
dielectric slab vs. angle of incidence: (a)
also on the angle of incidence. electric vector per~ndicular to the plane
This is shown in Fig. 11.10, where of incidence; (b) electric vector in the plane
of incidence; n = 1,6.
IRI is plotted for a plane wave inci-
dent on the plane surface of an infinite dielectric slab (n = 1.6) at angles
up to 90, The curve for polarization with the electric vector perpendicu-
lar to the plane of incidence is a plot of the well-known relation

sin (i r)
RI = sin (i+ 7-)
with (31)
I
r = sinl ;sini .
()
In this case IRl increases ]~ith angle of incidence from its value at normal
incidence, slo!rly at small angles and rapidly in the neighborhood of 90.
The power loss reaches a value of 10 per cent at about 40 for n = 1.6.
.4ccount must be taken of this effect in the design of lenses. For polariza-
 402 DIELECTRIC A.VD ME1AL-PLAYE LENSES [SEC. 118

tion with theelectric vector in the plane of incidence,

~=tan(ir)
(32)
tan (2 + r)
In this case IRI decreases with angle of incidence until it reaches zero at
the Brewster angle, tan n; beyond this angle it again increases. Match-
ing devices can be used to cut down surface reflections, as, for example,
a quarter-wavelength-thick surface layer of material whose refractive
index is the geometric mean of the refractive index of the lens dielectric
and that of air. Such a matching section is, of course, an additional
complication in the design and manufacture of the lens.

METAL-PLATE LENSES
1108. Parallel-plate Lenses. A lens structure using spaced conduct-
ingplanes instead of a dielectric has been developed lfor use at microwave
frequencies. A common form consists of parallel strips of sheet metal
held apart by accurate nonreflecting spacers. Where the electric vector
I
I
is parallel to the plate surfaces and the plate spacing a is less than k,

but greater than ~ (cf. Sec. 7.15), the wavelength betlveen the plates is
given by
!

(33)

Since this wavelength is greater than the free-space wavelength, the

parallel-plate space has the properties of a refracting medium with index
of refraction less than unity. When a thickness of this medium is intro-
duced into the paths of the rays, optical path lengths are reduced from
their free-space values instead of increased as ip the case of dielectrics.
Thus a converging parallel-plate lens is thinner on the axis than at the
edge of the lens, and a diverging lens is thicker. This is in contrast
to dielectric-lens structure. A parallel-plate lens can be designed to
have variable thickness, like a dielectric lens, or it can ha\e a uniform
thickness and \ariable plate spacing. The former, a more common
design, ~vill be considered here.
The refractive index of the parallel-plate space is given by

(34)

1 W. E. Kock, Iixpcrimcnts \vith 31ct:~l Il:itc I,enses for 31icro\vavcs, BTJ,

l{,port \fX1-44-160-67,lIarch 1%4, \\ire I,cns Antennas,] 11 XI-44 -I6O-1OO, .{pri]
1!)44; }fctal Plate I,ens l)twgn considerations, \f}f-44-160-1!},5, August 1!)44;
i JIetal Plate Lenses for lficrowaves, >111-45-160-23, iIarch 1945; 31ctal-Lens
Antennas, P70c. IRE, 34, 828836, November 1946.

SW, 11.tq PA RAI.LhI.-IL.47h LIi.VSE,q 403

with~/2 < a < A. Values ofnliebetween Oandabout O.86. Naturally

the smallest practical value of nshouldbe usedin order to minimize the
thickness of the lens. Owing to difficulties arising from the use of plate
separations near cutoff, namely,
frequency sensitivity, mismatch at
the lens surface, and difficulty in
maintaining the tolerances on
plate spacing, 0.5 is generally con-
sidered the minimum practical
value ofn. Oneimportant differ-
ence between dielectric and paral-
lel-plate lenses consists in the
constraints placed on the rays in
the lens by the presence of the par- ~i-l(*)
allel plates. The direction of FIG. 11.11.C~l;ndri.al parallel-plate
propagation in the lens must be lensin whichchangeof angleat the surface
is determined~YSnellslaw.
parallel to the plate surfaces.
Thus Snellslaw does not, in general, describe thechange indirection of
arayat the lens surface. In the case of cylindrical lenses this fact does
not essentially change the lens design. Here two cases maybe distin-
guished. When the electric
Actual direti!on of vector and lens plates are perpen-
dicular to the cylinder axis, Snells
law is valid and the constraints do
not enter (Fig. 11.1 1). When the
electric vector and plate edges lie in
planes parallel to the cylinder axis,
Snells law is replaced by the con-
dition of the constraint (Fig.
11.12). However, the design of a
spherical lens consisting only of
plates parallel to the E-plane of
FIG. 11.12 .CJ-lindrical parallel-plate
the feed becomes a three-dimen-
lens in which change of angle at the lens
surface is determined by the corldition that sional problem instead of the two-
the ray is constrained to pass between the dimensional one discussed for
plates.
dielectrics. The constraint exists
in the magnetic but not in the electric plane, and the lens surface is not
symmetrical with respect to a rotation about the axis. The design can, of
course, be reduced to a two-dimensional problem by use of a cellular con-
struction to introduce identical constraint in the electric plane.
There is one useful example in which the constraint does not enter
explicitly, and most spherical parallel-plate lenses used hitherto have
been of this type. This is the one-surface lens in which the refracting
surface is on the side of the feed with the second surface a plane per-
 1-
;
,4

404 DIELECTRIC A.VD JIETA1.-P1,ATE I>E.V,$E,T [SEC. 118

pendicularto the lens axis. The rays are then parallel to the axis, and
the constraints are automatically satisfied. .4s discussed in Sec. 11.3,
the contour of this lens may be designed solely from the condition of
equality of optical paths. Referring to Sec. 11.3, to the case where the
refracting medium is to the right of the contour T1 (Fig. 11.2), the
coordinates of 22 are given again byEq. (3):
(1 njf
(3) I
=lncose
In the present case, however, with n <1, this is the equation of an ellipse;
the hyperbolic face of the dielectric lens is replaced by an elliptical face
in the metal-plate lens. Since little use is made of unzoned lenses, it is
I
desirable to rewrite this equation to apply to a zoned lens. If the zone on
the lens axis-is taken as the first, the equation
for the surface of the Kth zone is given by
(1 n)fx
(35)
K=lncose
v.ith

In this equation the assumption is made that

there are steps of one wavelength between
adjacent zones; other~rise (K 1)in Eq. (36)
2.0

A(P)
A(O) Without /
zoning /
1.5
/
/
/
/

EElll1.0
o 10 200 30 400
0
1,1~. 11.13. Elllptiral vuu- 11,,. I 114. -Rat,o A(o) /A(@ [,)r zo,,ed
tour with five mues; n = 0.5; contour of Fig. 11.13.
f ==40A.

must be replaced by the total number of wavelengths in the steps between

the axis and the Kth zone. Figure 11.13 shows a five-zone arrangement
of this contour using 0.5 for n and 40k for fl. The ratio of jl to aperture
is 0.86. The unzoned contour is shown for comparison. The ampli-
tude-illumination ratio (Sec. 11.3) which relates the amplitude across
the aperture to the amplitude pattern of the feed is given for a spherical
lens by

(37)
%=~
L%C. ~] 8] PARALLEL-PLATE L.V.VShS

and fora cylindrical lens by

%= K%E5G (3s)

The ratio A(p)/A(0) of Eq. (37) for the spherical case is plotted in Fig,
11.14 for the five-zoned elliptical contour of Fig. 11.13. The ratio
exhibits a slow stepwise increase from the center to the edge of the lens

FIG. 11.15.Rear view of a 1.25-cm parallel-plate lenS. (Courteav of the Bell Telephone
Laborat.rg.)

in a fashion favorable for microwave use. The improvement introduced

by zoning is seen by comparing with the dotted curve for an unzoned
contour. Figure 11.15 is a photograph of the rearview of a spherical
Iens;it is48in. indiameter anddesigned foruse in the l-cm range. It
is a zoned single-surface lens with elliptical contours. This lens is
constructed of thin, equally spaced metal plates parallel to the -E-plane of
the feed. Thetirn einvolve dincuttin gtheplate sforlenses of this type
is a factor, because the plates differ in contour except for corresponding
pairs on opposite sides of the optical axis. Since surface tolerances are
large (Sec. 11.10), the contours can be cut by sawing or filing to ascribe
line or by stacking the plates with temporary wooden spacers and cutting
on a lathe.
1W. F,. Km-k, lIct:il
Ilate Lensm for Lficrow:,vw,) RT1, Report 3131-44-160-
100, April 1944.
 7

#d

406 DIELECTR~C AND METAL-PLATE LENSES [sm. 119

11.9. Other Metal-lens Structures. -Several other methods of lens

construction have been tried with success. 1 One urocedure that leads to
good structural characteristics makes use of polys~yrene foam (e = 1.018
and weight 1 to 2 lb/ft 3, as a dielectric medium between the plates.
The plates are actually sheets of
metal foil bonded to the polysty-
rene. The slabs of polystyrene foam
with the metal foil sidings may be
molded into a block, and the lens
contour can then be cut out on a
lathe.
Another method is to replace the
lens plates by a system of parallel
wires. Lenses ofthistype areinter-
esting principally because they point
to the possibility of using lenses at
long wavelengths where metal plates
are out of question but curtains of
wire suspended from poles are feas-
FIG. 11.16. Polyst yrene-foam lens. ible. The lens structure is based on
(Courtesy of the Bell Telephone Labora- (
tory.)
the fact that slots in the wall of a
waveguide which are parallel to the
electric-field vector do not radiate. The system of parallel wires maybe
thought of as a limiting condition arrived at by cutting slots in the parallel
plates constituting the lens. The primary problem in the design is the

yIG, 11 .17, Wire lenses: (a) parallel-wire lens; (b) wire mesh lens. ((lwmtes~ oj the Bell
Telephone Laboratory.)

determination of the practical ratio of slot width to conductor width.

It has been found that at a wavelength of 3.2 cm wires of diameter O.O49
in. and spaced 0.3 in. center to center have a loss of 0.1 db per inch or
about & db per wavelength. Since the lens is only a few wavelengths
1 w. E. Koclc, kfetal Plat~ Lenses for Microwaves, ?3TL Report MN1.45-160-23,
March 1945.
~ IV. IL Kock, Wire Lens Antennas, J3~L Report MM-44-160-100, April 1944.

4 -
 SEC. 11.10] METAL-PLATE LENS TOLERANCES 407

thick, thkis not prohibitive. The over-all performance of a lens of this

type has been found to compare favorably with an equivalent lens mak-
ing use of solid plates.
Several of these types of lenses are illustrated in Figs. 11.16 and
11.17. The former shows a lens making use of a polystyrene-foam base.
Figure 11.17a shows a parallel wire lens, and Fig. 11.17b shows a wire
mesh lens. The latter takes as its starting point the use of rectangular
waveguides as the lens medium. The waveguide walls are replaced by
wire mesh, again making use of the fact that slots in a waveguide wall,
when suitably oriented, do not radiate.
11.10. Metal-plate Lens Tolerances.-Tolerances on the lens surface
are large with respect to deformation by warping and twisting, as they
are for dielectric lenses. Extreme rigidity in the cellular structure is
consequently not necessary. Tolerances on the lens thickness and plate
spacing are interrelated. Using the same criteria as for the dielectric
lenses (allowing an error either in thickness or refractive index alone
to cause more than half the allowable phase error X/8),

A
(39)
t S 16(1 n)
\ and

(40)

where An depends both on plate spacing and wavelength. Reserving

discussion of changes with wavelength until later (Sec. 11.11), the varia-
tion of n with plate spacing a is given by

An = (1 n) Aa
. (41)
n a

If in addition (1 n)t is assumed to be approximately a wavelength,

as it is for zoned lenses, Eq. (40) becomes

Aa <
(42)
x = 16 (nn+ 1)

When n is equal to 0.5, Aa must be less than 0.024k, which is 0.030 in.
at 3.2 cm. At this wavelength the tolerance on a would place a lower
practical limit of about 0.3 on n, because this would lead to a tolerance
of about 0.014 in. From Eq. (39) a value of 0.5 for n leads to an exceed-
ingly liberal tolerance on t, i.e., At s x/8. This is about twice the
tolerance on a reflector contour for the same over-all phase error of A/8;
I a discrepancy of A/16 in a reflector contour leads to a phase error of
approximately k/8. The tolerances given here are conservative, based on
the assumption that both thickness and plate spacing are in error. If
 *d

408 DIELECTRIC AND ME1AL-PLA TE LENSES [SEC. 11.11

the process of fabrication is such as to hold one or the other to better

!
than the stated values, the alternate tolerance may be increased.
11.11. Bandwidth of Metal-plate Lenses; Achromatic Doublets.
Since the index of refraction of a metal-plate medium depends on the
ratio of plate spacing to wavelength, metal-plate lenses are frequency-
sensitive devices. The relation between small changes in x and corre-
: spending changes in n is obtained by differentiating Eq. (34):
1
1 nz Ak
An= TT. (43)
.,
The effect on the wavefronts from an unzoned lens will be considered
first. Let L, represent the total length of the line segments FP and
PP (Fig. 11.18) passing through the edge of the lens and LI the length
FQQ on the axis. If Pa and PI are
I t F1h- the optical lengths of L, and L,
p! I respectively and t is the difference
I between the thickness of the lens at
F the edge and at the center, the op-
Q IQ tical path difference (P2 Pl) is
given by
l\
l\ P, P,=(L, t+nt) L,.
FIG. il.lS,Effect on a wavefmnt of (44) I
change in wavelength from the design
wavelength. The path difference (P2 P,) is a
function of wavelength because n depends on wavelength. At the design
wavelength AO,(Pj Pl) must be zero in order that PQ may represent a
wavefront. Hence
L, (1 no)t = L,. (45)

At a near-by wavelength AO+ AX, the wavefront at P is displaced a dis-

tance b along the axis with respect to the wavefront at Q, where ti/h
is given by

(23)

The approximation is good only for very small Ax. Substituting from
Eqs. (43) and (45),
6 1 + nO (1 n,)t AA
= (46)
A nO Arj x

If the limits of bandwidth are defined by Ial/k equal in magnitude to

0.125, the band\vidth is given by

25no X.
Bandwidth = per cent. (47)
1 + n, (1 ?,o)t

-=4 .-.- -- .- ..
 SEC. 11.11] BAN I) WIIMH OF METAL-PLATE LENSES 409

In the special case n, equal to 0.5,

Bandwidth = 8.3 per cent. (48)

(1 10nO)t

Since (1 na)tis at least several times ~,, bandwidths are of the order
of a few percent. It is believed that Eq. (48) gives aconservativeesti-
mate of the bandwidth, because, with considerable taper in the aperture
illumination, values of (161/h)_up to O.25 might be tolerated for certain
applications.
In determining the bandwidth of zoned lenses, the same type of
procedure is followed. The frequency sensitivity of zoned lenses is due
not only to the variation in n but also to the steps. As for unzoned
lenses the maximum deviation in the wavefront can be calculated by
comparing the optical path along the lens axis \vith the path by way of
the edge. Using the same notation as before, we have

P, P,=(L2t+?zt)-L1, (44)

with the somewhat different condition at XOintroduced by the zoning

P, P,(Kl)XO=O. (49)

Here K k the number of zones, counting the zone on the axis as the
first. In this case

Ak= (K1)% (l+nO)(lno)t~ (50)

[:r+)lx=xo nO ko

and the bandwidth is approximately given by

1
Bandwidth = 25 per cent. (51)
1 + nO(1 nO)t
(Kl)+= ~.

For zoned lenses (1 no)t is approximately one \vavelength at the

thickest portions. For practical purposes, therefore,

Bandwidth = 25 0 per cent. (52)

1 + Kno
For the special case no = 0.5,
25
Bandwidth = per cent. (53)
2+K

Zoning increases the bandwidth of a lens. For example, the use of

Eqs. (48) and (53) to compare equivalent zoned and unzoned lenses
which introduce compensation of five wavelengths in the longest optical
v
path [(1 nO)t = 5A0 for the unzoned lens; (1 n~)t = A,, K = 5, for
the zoned lens] shows that the zoned lens has slightly more than twice
the bandwidth of the unzoned lens, 3.57 per cent as compared with 1.67
410 DIELECTRIC AND METAL-PLATE LENSES [SHC. 11.12

per cent. For large values of K the zoned lens with m equal to 0.5 has
approximately three times the bandwidth of the equivalent unzoned lens
since a lens with K zones is equivalent to an unzoned lens with (1 nO)t
equal to KhO. This indicates the advisability of zoning on the basis of
bandwidth alone. Moreover it suggests the possibifityl of using a
doublet consisting of zoned and unzoned lenses with opposite frequency
characteristics to obtain increased bandwidth. A zoned converging lens,
for example, is stronger than an unzoned diverging lens of opposite fre-
quency sensitivity, and the combination is a converging lens. It must
be remembered, however, that the focal length of the doublet is much
longer than that of the uncorrected converging lens unless the compensat-
1

n
FKQ. 11,19.Power reflection at normal incidence as a function of n.

ing lens can be made optically thin and given the requisite frequency sen-
sitivity by spacing the plates closely to yield small n.
For any lens the first-order effect of change in frequency is to alter the
effective focal length of the lens. For frequencies close to the design
frequency the deformed wavefronts are so nearly spherical that moving
the feed along the axis effectively removes the deformation. The effec-
Q
tive bandwidth of a lens is consequently increased by a provision in the
antenna system for feed motion.
11.12. Reflections from Surfaces of Parallel-plate Lenses. While the r
general problem of reflection from the surfaces of a parallel-plate kns
has not been solved, some indication of magnitude can be obtained from
a study of the reflection of a plane wave from the edges of an array of

parallel, equally spaced plates when the edges lie in a plane. This
problem has received rigorous theoretical treatment. 2 It seems reason-
able to expect that the values of R derived for this case at various angles
of incidence should be a good approximation to local values of R on a
1 W. E. Kock, I;xperiments with Metal Plate Lenses for ilIicrowaves, BTL r
Report M Xf-160-67, Afarch 1944,
2 J. F. (;arlson and Albert E. Heins, The Reflection of an Electromagnetic Plane
Wave by an Infinite Set of Plates, I, Quart. Applied Math,, 4,313-329, January 1947, :

I
SEC. 11.12] REFLECTIONS FROM PARALLEL-PLATE LENSES 411

lens surface, providing that lens surfaces and wavefronts do not appre-
ciably depart from a plane over distances comparable to a wavelength.
For normal incidence R is given by
,
~=lnejm
(54)
l+n
This expression differs from the expression for normal incidence on o
dielectric surface only by the presence of the phase angle @. Because
n is less than unity, the magnitude of R is larger, however, than that
from a dielectric surface with the same value of 11 n[. The magnitude
f
I

FIQ. 11.20.Planes of incidence at the plane face of an infinite set of parallel plates.

of IR12at normal incidence is plotted in Fig. 11.19. Its value for small n
obviously puts a lower practical limit on n, especially for equiphase lens
surfaces where the reflected wave is in phase at the feed. For n equal
to 0.5, the reflection is already quite large (11 per cent power reflection).
Surface-matching devices are, of course, a possibility.
The average reflection over a lens surface is probably less than the
value derived for normal incidencq because the magnitude of the reflected
wav% probably decreases with the angle of incidence in both planes.
For the magnetic plane (MF in Fig. 11.20) where the change in angle
at the surface is determined by the constraint and not by Snells law,
an expression has been derived for [RI as follows:

(55)

This expression is valid for a restricted range of angles:

412 DIELECTRIC AND METAL-PLATE LENSES [SEC, 1112

with ,.
1<; <2.

The restriction on i arises from the fact that grating lobes become possible

for angles of incidence larger than sin]

~ 1
a ()
Inspection of Fig.

11.20 and a plot of lR12in Fig. 11.21 shows that IRI decreases wlthincreas-
ingi, reaches zero foriequaltocos1 n, and increases beyond that angle.
0.5

e
e=
= =
~;
SM?*-1)
~ //
oE
30 60 90
Angle
FIQ. 11.21.Power reflection at the plane fare of an infinite set of parallel plates as a
function of angle of incidence.

Forthe electric plane (eeFof Fig. 11.20) theexpression forthevaria-

tion of R with angle of incidence has not been derived. In this plane
the change of angle upon refraction is described by Snells law. One t
might expect some similarity to the dielectric case where the magnitude
of the reflected waw? decreases from its value at normal incidence with
increasing i until it reaches zero at the %ewster angle, tan n, and then
increases. It is hoped that a solution for this plane will become available
in the future.

I
r I
 i.:
!:
,,
,.
.,
:
/

CHAPTER 12

PENCIL-BEAM AND SIMPLE FANNED-BEAM ANTENNAS

BY S. SILVER

PENCIL-BEAM ANTENNAS

12.1. Pencil-beam Requirements and Techniques.-The term pencil

6 beam is applied to a highly directive antenna pattern consisting of a
single major lobe contained within a cone of small solid angle and almost
circularly symmetrical about the direction of peak intensity. As used
here, it will apply to beams with half-power width less than 15. These
beams are analogous to searchlight beams, and, as with an optical search-
light, the elevation and azimuth coordinates of a target in space can be
simply correlated with the similar coordinates that define the orientations
of the antenna. In connection with the technique of using radar echoes
,. for obtaining range information, the pencil-beam antenna serves to define
the position of a target completely.
There are several possible techniques for producing pencil beams.
*
The simplest in conception and from the point of view of practical
design is that of placing a point source at the focus of an optical
system, such as a reflector or lens, to produce a beam of parallel rays.
It is evident that to produce a circularly symmetrical beam, the optcial
system should have rotational symmetry with the feed located on the
axis of rotation (optical axis). This presupposes that the primary feed
pattern likewise has rotational symmetry about the same axis; in prac-
tice this is approximated by a feed pattern having a pair of orthogonal
t principal planes (symmetry planes) that intersect along the optical axis,
with nearly equal half-power \vidths in the t\vo planes. In many calcu-
1 Iations this actual feed pattern can be replaced by an equi~ralent cir-
cularly symmetric pattern that is the average of the patterns in the two
principal planes. If the simple geometrical picture-that the beam
produced by the optical system consists of a family of parallel ra.v
,,jere strictly y-slid, the beam would have zero ~vidth as plotted in a
polar diagram, Ho\\ever, this simple picture is markedl.v modified by
diffraction phenomena due to the limited aperture of the optical system.
The aperture is the projected area of the reflector or lens on a plane
* normal to the optical axis, and for a rotationally symmetrical system it
is [ir(,lllar in shape. .1s a result cf diffraction the antenna pattern has
a major lobe of finite width and characteristic side-lobe structure.
413
414 PENCIL-BEAM AND SIMPLE FANNED-BEAM ANTENNAS [SEC. 121

The general theory of apertures and diffraction has been developed

in Chap. 6. It was shown there that of all the phase and amplitude dis- .
tributions over a plane aperture that give rise to a beam with maximum
intensity in the direction normal to the aperture, a uniform amplitude
and phase distribution gives rise to maximum gain; in general, minimum
beamwidth is concomitant with maximum gain.
The relation between antenna gain and range in radar systems has
been noted in Sec. 1.2 [Eq. (1.12)]. The beamwidth is also an important
factor in the precision with which target location can be effected. The
considerations here are partly optical (of exactly the same nature as
those which determine the resolving power of a telescope) and in part
involve system factors such as pulse width; a rather complete discussion
of resolving power of a radar set and its bearing on beamwidth require-
ments is given in Vol. 1 of this series. 2 On the basis of gain and beam-
width considerations a fundamental design requirement jor radar antennas
is that the phase distribution over the aperture be uniform-n terms oj
geometrical optics, that the optical system produce a beam oj parallel rays.
It should be noted that in addition to the gain and beamwidth require-
ments, the greatest possible suppression of all secondary lobes is desir-
able; for if a target is sufficiently close to be detected by the side lobe,
it becomes indistinguishable from a target detected by the main lobe
at the same range. However, as was found in the treatment of general
diffraction theory, requirements of maximum gain and minimum side- T
lobe level are generally incompatible. The necessary compromise
between them in antenna design is made in optical systems by adjusting
the illumination, that is, the amplitude distribution, over the aperture.
The advantages of microwaves become strongly evident in the design
of the pencil-beam antennas. Within reasonable limits on the over-all
size of the antenna, the distance from the reflector (or lens) to the antenna
feed can be made so large that the optical device is in the radiation zone
of the feed. Thus the difficulties associated with the phase quadrature
of the induction field are avoided; that is, it is possible to operate in
that region of the feed pattern where the feed is essentially a point ,
source. Because the dimensions of the reflectors and lenses are fairly
large compared with the wavelength, it is possible to simplify the theo-
retical considerations by suitable approximations. As a result the design
of a pencil-beam antenna becomes to a large extent a calculable procedure.
Inasmuch as lenses have been discussed in detail in Chap. 11, the
treatment of design problems in the present chapter will be confined
almost entirely to reflectors. Many problems are common to both: The
secondary pattern is determined essentially by the field over the aper-
r
ture, and the requirements to be imposed on the latter, which will be
1}1, Born,Optdc,EdwardsBins., Inc., Ann Arbor, \fich., Chap,4.
2Ridenour, Radar ,$ystem Engtneerzng, Vol. 1, RI. rechnical Series, ~hap. 14.
 SEC. 12.2] GEOMETRICAL PARAMETERS 415

arrived at from the discussion of reflectors, can be transferred directly

to lenses. The latter just began to commond serious attention at the
close of the war period, and their study and use are still in the initial
stages.
PARABOLOIDAL REFLECTORS

12.2. Geometrical Parameters.-The nature of a reflector that trans-

forms a spherical wave, arising from a point source, into a plane wave-
front was discussed in Sec. 4.9, where it was found to be a paraboloid
of revolution with the source at the focus. In discussing these systems
it is convenient to use several different coordinate systems simultane-
ously; these are defined in Fig. 12.1. .4 rectangular coordinate system
z, y, z will be used, with the origin at the vertex of the paraboloid and
the z-axis the axis of revolution. In these coordinates the equation of
the paraboloidal surface is
X2 + Y2 = 4f% (1)

where j = OF is the focal length. We shall also use cylindrical coordi-

nates r, ~, z, where r and &are polar coordinates in the planes z = constant,
~ being measured from the zz-plane. In these coordinates the equation
of the surface is
7-2= 4fz. (2)
.
In expressing the relation of the primary feed pattern to the reflector,
there is employed a spherical coordinate system p, +, ~, with the origin
at the focus F and the polar axis directed in the negative z-direction;
the aximuth angle ~ is the same as that defined in the cylindrical system,
and ~ is the polar angle. The equation of the surface referred to these
spherical coordinates was obtained in Sec. 4.9 [Eq. (4.69)]; it is

2f
~=l+cos~
=jsecz
()
~ (3)

Lastly, to discuss the final antenna pattern we use a spherical coordinate

system with polar axis in the ~ositive z-direction and origin again at the

1The material to he presented in the following sections represents a summary of

British and American work done during the >varperiod; the following is a partial bibli-
ography of reports on paraholoidal reflectors: L. J. Chu, ( Theory of Radiation from
Paraboloidal Reflectors, ILL,Report h-o. V-18, Feb. 12, 1941; 13.U. Condon, Theory
of Radiation from Paraboloid Reflectors, l~cstinghouse Report No. 15, Sept. 24, 1941;
G. F. Hull, Jr., Application of Principles of Physical Optics to Design of UHF
Paraboloid Antennas, BTL Report ~f31-43-110-2, Feb. 8, 1943; and F. R. N.
Nabarro, ( Theoretical \Vork on the Paraboloid Mirror, British Report, Ministrv
of Supply, .4. C. 1435, RL)F 103, ~om. 72, NTOV.27, 1941. References to earlier
French and German work are given by F. E. Terman, Radio Engineers Handbook.
MoGraw-Hill, New l-ork, 1943, p. 837.
416 PENCIL-BEAM AND SIMPLE FANNED-BEAM ANTENNAS ~Sm,, 12.2

focus; the coordinates are R, O, +, with O the pola~ angle and I#Jthe
.
azimuth angle, the latter being measured from the zz-plane.
The reflector is cut off by the aperture plane A at z = Zo. The
diameter of the aperture will be designated by D, and iix area by A.
The shape of the reflector is specified by the ratio of focal length to
z 1

A
I
I
I
1!
.f+zo I

Y
/

v A

FIG. 12. 1.Geometrical parameters for the paralmlo idal reflector.

diameter, j/D, or alternatively by the angular aperture T, that is, the

angle subtended at the focus by a radius of the aperture. The relation
between the f/D ratio and the angular aperture is given by

(4b)

One of the most important design problems is the determination of the

shape that gives maximum antenna gain for a given aperture diamct:r
t
and a given primary feed pattern.
The geometrical properties of paraboloids are well known. .4ny rcc-
tion of the surface containing the z-axis is, of collrse, a parabola ~vith
 SEC. 12.3] THE SURFACE-CURRENT DISTRIBUTION 4A7

focus at F. In addition, however, the curve of intersection of the surface

. with any plane parallel to the z-axis (normal to the zy-plane) is also a
parabola of the same focal length f as the paraboloid. As a consequence
of the property, only a single parabolic template is needed in the con-
struction of the reflector to test the accuracy of all parts of the surface.
The normal to the surface at a point p, ~, ~ lies in the plane containing
this point and the z-axis and makes an angle $/2 with the incident ray
from F,
12.3. The Surface-current and Aperture-field Distributions.In the
treatment of the general theory of reflectors developed in Chap. 5 it
is shown that the over-all pattern of the antenna, that is, the secondary
pattern, arises by the superposition upon the radiation field of the antenna
feed of the radiation field of the distribution of current generated on the
surface of the reflector in the presence of the feed. It was shown further
that the reflector field can be determined either from the surface-current
distribution directly or in the form of a diffraction pattern from the field
distribution over the aperture of the mirror. Before proceeding with
the calculation of the surface-current and aperture-field distributions,
some fundamental ideas and assumptions which underlie all of the sub-
sequent discussions should be noted. The feed pattern in the presence
of a reflector, in general, differs from its free-space pattern, because the
reaction of the reflector on the antenna feed modifies its current system.
v
If, however, the focal length of the paraboloid is at least several wave-
lengths in magnitude and the mirror is in the radiation zone of the free-
space pattern of the feed, the interaction between the mirror and the
antenna feed is a second-order effect as far as the primary pattern is
concerned. These conditions are usually realized in microwave antennas;
and subject to their realization, it will be assumed that the feed pattern
in the presence of the reflector is the same as under free-space conditions.
To avoid the complex problem of interference between the fields
of the reflector and the antenna feed in the formation of the main struc-
ture of the antenna pattern [cf Eqs. (5.75) and (5.76)], the directivity
of the feed pattern should be such that the major portion of the energy
lies within the cone defined by the feed and the reflector. Referring
to Fig. 12.1, iftaking an ideal casethe primary pattern is zero for
angles $ > 90, the main structure of the beam is determined by the
reflector currents alone. The directive feeds discussed in Chaps. 8 and
10 approximate this condition rather closely; their back lobes, however,
are not completely negligible and have significant effects not only on
the wide-angle side lobes where the back-lobe field is comparable to the
# weak reflector field but also on the peak intensity, that is, on the antenna
gain. The effect of the back lobe on gain will be in~estigatecl in Sec.
12-5.
The general approximation procedure bawd on geomctriral optics
418 PENCIL-BEAM AND SIMPLE FANNED-BEAM ANTENNAS ~~EC. 12.3

and plane-wave boundary conditions, which is discussed in Chap. 5, will

be used to evaluate the surface-current and aperture-field distributions.
Let the principal E- and H-planes of the primary feed pattern coincide
with the ZZ- and yz-planes, respectively, in Fig. 12.1. If PT is the total
power radiated by the feed and G,(f,v) its gain function, the power P((,~)
radiated per unit solid angle in the direction .f, ~ is

F(.E,I)) = ~ G,(&x).

The electric-field-intensity primary pattern, reduced to the unit sphere

about the center of feed, is then given by

(5)
E()*)]P-l=
12(:)H[:Gf(*11e0(J~
where eOis a unit vector defining the polarization in the primary pattern.1
The field intensity in the incident wave at a point p, & $ on the reflector
is therefore given by

(6)
0= [wwG(Y%-e@e@
The field intensity El in the reflected wave at the same point is

(7) w
1= [2(YwG(Ye$e$
where e ~ defines the polarization in the reflected wave; according to the
plane-wave boundary condition [Eq. (5.25)] the vectors e, and el are
connected by the relation
nx(eO+el)=O, (8)
in which n is the unit vector normal to the reflector at the point of inci-
dence. The vector n will be taken to be directed outward from the I
reflector into free space. Following Eq. (5.57a), the surface-current
density K is given in terms of the incident wave by
t

= [8(;)fi21[G(:*)]4 -k[nx(Qoxe0
g)
where pOis a unit vector in the direction of the incident ray. Expanding
the vector product, we obtain

= [8(:Y:l$[Gf(:~)]-]k[eocos:+(e0n)@o] (ga) t
The current can be expressed in a similar manner in terms of the reflected
] It is being assumed that the radiation field of the primary feed is linearly polarized
at every point but that e~ is a function of L #.
 SEC. 12.3] THE SURFACE-CURRENT DISTRIBUTION 419

field bymakinguseof Eq. (5.57 b), noting that thereflected rayis parallel
to the z-axis; we have then

or
=[8(:) =l}G(:*)]-knx ixel110)
=[8(YwG(Y%e-k[-ec0s: +iJned oa)
To obtain the field over the aperture we note that since the reflected
rays are all parallel, the field intensity remains constant in magnitude

ne
!

H-plane

FIO. 12.2.Typica1 aperture-field distribution; the field is resolved into principal and
cross-polarization compments.

along the reflected ray (cj. Sec. 4.4). The electric-field intensity E(T,f)
at a point (r, () on the aperture is thus given directly by El at the cor-
.
responding point (p,~,~), except for the phase retardation corresponding
to the path from the reflector to the aperture plane. The relation is

E(r, f) = El(P, f,#)ek@-zJ

= [2(:) H[G(Y%-k(tizz-)ele1. 1)
The distance p + zO z is the total optics! path from F to the aperture
plane; it is therefore independent of the point (r, ~), and more specifically
t it is equal to f + z,. Comparing the surface-current distribution as
given by Eq. (lOa) with the aperture-field distribution [Eq. (1 1)], it is
seen that except for constants, the aperture field is the projection of the
 4

420 PENCIL-BEAM AND SIMPLE FANNED-BEAM ANTENNAS [~EC, 124

surface-current distribution into the aperture plane. In this connection

it should be noted that the longitudinal component of the current given
by the term (e, c n)i. in Eq. (10) finds no counterpart in the aperture
distribution because the field over the aperture is wholly transverse to
the z-axis. This longitudinal component of the current has generally
been neglected in paraboloid theory. It contributes nothing to the field
in the forward direction and therefore does not enter into the computa-
tion of the peak intensity, but it does modify the side-lobe structure of
the beam. The general character of the aperture distribution (and the
transverse component of the current distribution) is illustrated in Fig.
12.2. It is seen that the polarization reflects the symmetry of the pri-
mary feed pattern. The component e,z of the aperture polarization
which is parallel to the principal E-plane of the feed is known as the
principal polarization component, and the component el~, which is
transverse thereto, is known as the cross-polarization component. By
virtue of the symmetry conditions the cross-polarization components
at any pair of points that are symmetrical with respect to tbe principal
planes are effectively 180 out of phase with one another.
12.4. The Radiation Field of the Reflector.-The secondary pattern
produced by the reflector may now be calculated from either the current
distribution or the aperture field, using the methods of Sees. 5.9 and 5.12,
respectively. The two calculations do not lead to completely concordant
y
results; the differences between them vanish, however, in the limit of
zero walelength. The discrepancy lies in the fact that the aperture
field which would be calculated as produced by the surface currents is
equal to that calculated on the basis of the reflected rays only under
the limiting condition of zero wavelength. To exhibit the relationships
we shall set up the expressions for the radiation field as obtained from
the current distribution. Letting Rl, ie, i+ be unit vectors associated
with the spherical coordinates R, 0, @ (lFig. 12.1), we have, by Eqs.
(5.74a) to (5.74c), that the radiation field of the reflector is

(12)

where the vector 1, expressed in terms of the incident field on the reflector

X p sin + sec ~ W df (12a)

and, in terms of the reflected field,
 SEC. 12.4] THE RADIATIO.V FIEI>IJ OF THE REFLECTOR 421

Comparing with Eqs. (10) and (lOa) it is seen that in the form of Eq.
(12f)), the vector I is resolved into a transverse component parallel to
the q-plane,

and a longitudinal component

As regards the longitudinal component it is observed that 1, makes no

contribution to the Ed-component of the field because i. is always in a
plane normal to i,. Furthermore, since iZ . i~ = sin O, the longitudinal
component makes no contribution to the field in the direction o = &
the physical basis for this being that a current element is equivalent to
a dipole and does not radiate in the direction along its axis. The con-
tribution of 1, is significant only at wide angles. For the systems with
which \ve are concerned that produce narrow beams, the contribution
of I= is a second-order effect; it vanishes in the limit of zero wavelength.
There is no counterpart of the 1, contribution in the calculation of the
pattern from the aperture-field distribution.
y Considering the transverse component 1,, it will be observed that if
the radiation field is confined to a small angular region about the d = O
axis, the variation of cos 0 in the phase term of Eq. (13a) can be neglected;
we have then P(1 + cos ~ cos 0) = 2,f. Also it will be noted on com-
paring with Eq. (11) that except for a multiplicative constantwhich is
contained in the field expressions [Eq. (12)]the factor [G1(~,~)]~$el/p
is the field in the aperture plane at the point (r, ~) which corresponds to
the point (p,~,~) on the reflector. Equation (13a) is, therefore, given
approximatel.v by the integral

over the aperture plane, on setting up the radiation field on tile l):~sis
of the aperture field by the methods of Sec. .5.12 it ~vill be fount] th:lt (ho
same result is obtained for the pattern as that from the ~lsc of E(I, (14)
in conjlmction ~vith Eqs. (12). Thus, the current-distribution metlu.xl
passes into the aperture-field method as the angular spread of the pa~-
tern decreases, that is, as the ratio of the \vavelength to aperture diam-
eter, x/D, apprwwhcs zero. The significant difference bet~veen the
f results of the two methods is the dependence of the pattern on the ratio
h/D. It ~1-asshoim in Sec. 68 that on the basis of the aperture-fiel(l
calculation, the ang~llar distribution of the secondary pattern is propor-
tiomd to k 1) for a given relative distrib~ltion ovrr the aperture and the
side-lobe intrnsitics arr independent of k I). On the nther hand, it has
422 PENCIL-BEAM AND SIMPLE FA NNIID-BEA Al ANTENNAS ~f+EC. 12.4

been found in a study of special forms of Gl(f,Y) by the current-distribu-

tion method, using the complete expressions Eqs. (12a) and (12b), that
the side-lobe intensity is also a function of A/D, which asymptotically
approaches the value given by the aperture method as A/D approaches
zero.
Principal E- and H-plane Patterns.In an arbitrary direction the
field has both the Ee- and the E$-component. They are generally out of
phase with the result that the field is elliptically polarized (cj. Sec. 3.12).
However, in the principal planesthe planes + = O and o = r/2the
field is linearly polarized in the direction determined essentially by the
principal polarization component of the aperture field. Considering
the E-plane, @ = O, we see that the y-component of It which arises from
the cross-polarization component of the current distribution (or aperture
field) vanishes because contributions from points in the reflector that
are symmetrically located with respect to the zz-plane are 180 out of
phase. The field is produced by the 1,- and I,=-components and,
therefore, has only an Ep-component which lies in the E-plane.
Similarly it is found that in the H-plane the field has only an E$-compo-
nent and is, therefore, everywhere normal to the H-plane and parallel
to the principal component of the aperture field. Again, since the
cross-polarization components of the current at a pair of points on the
reflector that are symmetrically located with respect to the yz-plane are ?.~
180 out of phase, their resultant contribution to the H-plane vanishes.
It was noted earlier that the longitudinal current element contributes
nothing to the E+-component; therefore the H-plane field is produced
entirely by the principal component of the aperture field. Using the
aperture-field approximation [Eq. (14)], we find that the principal plane
patterns are

a, E-plane:

(15a)

b. H-plane:

(16a)

The two patterns have the same value, of course, along the axis (in the
direction 8 = 0).
 SW. 12.5] THE ANTENNA GAIN 423

Cross Polarization.The polarization of the field in a pencil beam is

generally expressed with reference to the m and y-axes rather than
the spherical coordinate directions as we have done above. The use
of the cartesian components has associated with it an error in that the
field is transverse to the radial direction from the origin and not to the
z-axis; but if the beam is narrow, the error is small. The latter mode of
description has the advantage that the E.-component is associated directly
with principal polarization component of the aperture field and the
Eu-component with the crosr-polarization component. The E=- and
Ev-components are designated correspondingly as the principal polariza-
tion and cross-polarization components of the secondary pattern. They
aregiven by relation sequivalent to Eqs. (12) withi~. Iandi~. I replaced
byl,. andl,V, respectively.
Byusing the cartesian components, thesecondary pattern is resolved
into a principal polarization pattern and a cross-polarization pattern.
The E- and H-plane patterns given by Eqs. (15a) and (16a) belong to
the former. It is obvious that the symmetry properties of the aperture
field with respect to the principal planes, which lead to zero cross polari-
zation in those planes, do not hold for other directions in space. The
cross-polarization pattern must, therefore, have maxima in the four
quadrants between the principal planes. A detailed analysisl shows
b.: that the cross-polarization pattern takes the form of four lobes whose
maxima lie in the 45 planes between the principal planes. Any
two lobes related by reflection in a given principal plane are out of
phase by 180. The maxima of one set of lobes occur at angular dLs-
tances from the paraboloid axis equal to the position of the first minimum
of the principal polarization pattern, which is very closely equal to
the half-power width of the main lobe. A second set of cross-polarization
lobes appears at much wider angles; the peaks are quite low, but the
lobes are very broad and therefore represent a not completely negligible
fraction of the total energy.
Cross-polarization studies should be made on all antennas on which
I
the side-lobe specifications are very stringent. Although the principal
polarization lobes may meet the operational requirements, the cross-
polarization lobes may not. Furthermore, since they lie close into the
main beam, they effectively increase its width.
12.6. The Antenna Gain.-The gain is generally the primary con-
sideration in the design of the antenna. The factors affecting the gain
are treated conveniently in three parts: (1) the dependence of the opti-
mum angular aperture v on the feed pattern, D in
for a fixed diameter
f the apcrt ure plane, assuming that Gj(t,l) = O for V > 90; (2) the back-
labe interference effect; and (3) phase-error considerations. In this
LF,.U. Condon, Theoryof Radiationfrom ParaboloidReflector, Westinghouse
Report No. 15, Sept. 24, 1941.
 424 PENCIL-BEAM AND SIMPLE FANNED-BEAM ANTENNAS [SEC. 125

discussion the primary pattern will be taken to be circularly symmetrical,

independent of ~; as was noted in Sec. 12.1 this means that in practice
the feed pattern is replaced by the arithmetic mean of its principal
E- and H-plane patterns. It is immaterial for the calculation of gain
whether the surface-current or aperture-field distribution is taken as
the starting point, because as was pointed out above, the longitudinal
component of the current is ineffective in determining the peak intensity.
Optimum Angular Aperture Relations.The field intensity in the
secondary pattern on the axis at a distance RO from the focus is given by
either Eq. (15a) or (16a) for o = O. For the present purpose, it is more
convenient to express Ilt in the form of Eq. (13a) as an integral over the
surface of the reflector rather than in the form of Eq. (15b); we have

The polarization component el, is in general a function of ~ and v because

of the presence of cross polarization. Ho\vever, in most cases of interest
the cross polarization e,. is a very small fraction of the total field and the
variation of elz over the aperture may then be neglected. Introducing
the equation of the paraboloid [Eq. (3)] and performing the integration 1.$

over f, we get

=%[(wle-yk(+) JIGf(+)an@ 8)
The power per unit solid angle P(O,O) radiated in the forward direction
is given by
~ }i
P(o,oj = ; R; ~ lE(Ro,o,0)12, (19)
()
and the antenna gain is ubtained from it as
G = P(o,o)
(20)
P,

h
because the total power radiated by the antenna as a whole equals that
radiated by the feed. The gain is thus found to be

(21)

The focal length is related to the angular aperture and the aperture dim-
eter D by
f = : cot;. (22

_. *._.. -. _, -. --...
~
 --
SEC. 125[ TIIE ANTENNA GAIN 425

Substituting into the preceding relation, we obtain finally the working

formula

G=t+Yc0t2~l/ovG*ta
23)
The factor (mD/x) 2 is the gain for a uniformly illuminated constant-
phase aperture; the rest is the gain factor or efficiency

(23a)

Thus the efficiency is a function only of the feed pattern and the angular
aperture; that is, for a given feed pattern, the efficiency is the same for
all paraboloids having the same j/D ratio.
It is instructive to consider the class of feed patterns defined b-y

Gf(+) = G$) COSn~, ()~$~;,

=0 p;. (24)

Many feed patterns can be represented by some one member of this class
over a sizable portion of the main lobe. The gain G~) is determined by
the condition that
G,(+) da = 47r,
/
dfl being the element of solid angle; this gives
G$) = 2(n + 1). (24a)

Substituting Eqs. (24) and (24a) into Eq. (23a), we obtain

+
S = 2(n + 1) cot ~ ,* cosn ~ tan ijd$) (25)
[/ () 1
with the follo~ving explicit expressions for the even values of n between
n=2andn =8:
2
S2 = 24
( sin ~ + In cm ~
) 2
*
cot
2

$4 = 40
( sin; + In cos~
)
(1 Cos )3
cot
~;

+ ~sin, V cot2x.
GG=14 21ncos~+ ~
[ 2 1 2
1 Cos * (1 COSZ) 1 . , coty.
s, =18 ~ 21ncos~
[ 3 ~sn q1 2
These results are shown graphically in Fig. 12.3, \vhere S. is plotted as
 426 PENCIL-BEAM AND SIMPLE FANNED-BEAM ANTENNAS [SEC. 12.5

a function of the angular aperture ~. For each primary pattern there

is an optimum aperture for which the maximum gain factor is attained.
The more directive the feed pattern the smaller is the optimum aper-
ture and, since the diameter of the aperture plane is const~.nt, the longer
is the optimum focal length. The general course of the curves and the

1,0

0.9

0.8

0.7

0.6
C%

i
-0.5
c
.=
a
0.4

0.3

0.2

0.1

0
00 100 200 30J 40 w 6(Y 70 800 900
Angular a~ure ~
Fro. 12.3.Dependence of the gain factor on angular aperture and pri rnary feed pattern.

G,(t) = 2(7L+ 1) COB+, O<$<T

2
= o, +>;.

existence of a maximum are readily understood ~vhen it is recognized

that the gain factor arises essentially as a product of two factors: (1)
the fraction of the total power radiated by the antenna feed that is
intercepted by the reflector and is thus made available to its aperture
for the main beam and (2) the efficiency with which the aperture
concentrates the available energy in the forward direction. The first
factor obviously increases with increasing angular aperture. The
second factor, determined by the field distriblltion over the aperturr,

.- ..
 SEC. 12.5] THE ANTENNA GAIN 427

decreases with increasing v; for as Z increases, the illumination over the I

aperture becomes more and more tapered toward the edge relative to
the center. This tapering is accentuated by the superposition of the
space attenuation factor I/p on the already directive feed pattern
[G,(+) ]~fi; as was shown in Chap. 6, such tapering of the illumination
results in a decrease in aperture efficiency. The optimum angular
aperture represents the proper compromise between spillover of the
feed energy and aperture efficiency. For an arbitrary Gf(~), the optimum
angular aperture is obtained as a solution of

(26)

a relation obtained by setting the derivative of Eq. (23a), dS/d~, equal

to zero.
The values of the gain factor at the maxima in Fig. 123 are consid-
erably higher than the values realized in practice. This is because ideal-
ized feed patterns have been assumed in which no feed energy is radiated
beyond ~ = 90. As a result the gain G$) of the idealized pattern is
much greater than the gain GfO of an actual feed whose main lobe can
be represented closely by GJOCOS+ but which in addition radiates
beyond 90. The gain factor S re-
alized with the actual feed is re-
lated to S. by

The value of the optimum angular

aperture is unaffected by this scal-
ing in the primary feed gain. It
will be observed that the value of
the maximum varies but slowly
with the illumination function.
The broader the primary feed
pattern the broader is the maxi- n
mum in the S-curve and the less FIG. 12.4Cutoff point ,n primary feed
pattern for n,aximurn gain M a function of the
critical is the choice of angular
sharpness of the feed pattern.
a~erture. It is convenient to
d~signate the optimum angular aperture in terms of the decibel level
of the primary pattern at the edge of the aperture relative to its maxi-
mum. Thus for a cosine-squared pattern the optimum value of v cor-
responds ta that angle in the primary feed pattern at which the power
d The decibel-cutoff point in the
is 8 db down from the peak intensity.
primary pattern is plotted as a function of the directivity in Fig. 12.4.
The decibel-clltotl point again is not a sensitive function of the directivity.
For most feeds the a~wmge ontimurn figure is from 9 to 10 db.
1
428 PENCIL-BEAM AND SIMPLE FANNED-BEAM AN TENAAS [SEC. 12.5

The optimum angular aperture can also be expressed in terms of the

intensity of illumination at the edge of the reflector relative to that at
the vertex. This is obtained by multiplying the ratio of the primary
pattern intensities GfO/Gf(J) by the ratio of the space attenuation
factors pif. It is found in the case of each of the distributions studied
above that the optimum angular aperture corresponds to an edge illu-
mination 11 db below the vertex illumination.
Back-1obe lnter~erence.The above results may be modified signifi- ;
cantly by the effect of interference between the back lobe of the primary
feed pattern and the reflector field. I.et G. be the gain of the antenna
feed in the direction ~ = 180. The back-lobe field intensity at the field
point (RO,O,O) along the axis is then
)

(28)

The choice of positive or negative sign is made according to whether the

field of the feed in the direction x = T is parallel (in phase) or antiparallel
(180 out of phase) to that in the direction ~ = O. Superposition of the
back-lobe field cm the reflector field [Eq. (18)] yields the total field
intensity

By the procedure followed previously the gain factor is found to be

where
U. = ~ [G,(+)]; tan ~ d~. (30a)
\
In most cases of interest G= is so small that the last term in Eq. (30) is
negligible; under this condition the gain factor becomes

$=(oCot:)[ =%n:si%cw
The term in brackets is the modification of the previous result introduced
3)

by the back-lobe interference. This modification introduces an addi-

tional X/D dependence; the interference effect depends on the ratio of
:
the back-lobe field intensity to the reflector field intensity, and the latter
is proportional to D/x. For a given primary pattern S is no longer a
function of the paraboloid shape alone.
~
I
 4

SEC. 12.5] THE ANTENNA GAIN 429

The back-lobe effect is illustrated graphically in Fig. 12.5. The

curves pertain to an actual feedthe +in. stub-supported dipole-disk
feed of Sec. 8.8and a reflector with a, 30-in. aperture diameter; the
wavelength is 10 cm. The main lobe of this particular feed is fitted
closely by the function
G(+) = 7.0 COS4 +.

It will be noted that the gain is 7, as compared with G~4J= 10 for the
idealized cos~ ~ pattern used previously, The back-lobe gain Gr is
0.142. Curve A is the relation between the gain factor and aperture,

0.7
J

0.6

0.5

~o.4
z
~

~ 0.3

0.2

u.1

n
-00 100 20 300 400 500 600 700 800 900
Angular aperture ~
FIG. 12.5.Effect of back lobe on gain.

neglecting the back lobe, while Curve B includes the interference effect.
The gain falls above or below Curve A according to whether the back
lobe is in phase or out of phase with the reflector field. The two fields
add when the focal dist ante is such that, together with the 180 phase
change at the reflector, the field of the latter is brought into phase with
the back lobe. The points of maximum deviation from Curve A corre-
spond to differences in focal length very nearly equal to x/2. The
optimum aperture is not altered noticeably, but the maximum realizable
gain factor increases by 2,5 per cent. The effect is small for this par-
ticular feed because the back-lobe level is so low relative to the main lobe.
j With ieeds such as the 3-cm-band double-dipole feed discussed in Sec.
(89), having a comparatively high back-lobe level, the back-lobe inter-
~ ference effect is much more significant.
430 PENCIL-BEAM AND SIMPLE FANNED-BEAM ANTENNAS [%C. 12.5

Phase-error Effects.-It was pointed out earlier that a diminution of

gain results from any departure from uniform phase over the aperture
that, however, leaves the peak intensity on the axis of the paraboloid.
The direction of peak intensity remains unchanged if the phase-error
distribution over the aperture is independent of (; this discussion is con-
fined to such distributions. The phase deviation can arise from a
number of scurces: (1) deviation of the reflector from a paraboloidal
shape, (2) defocusing (displacement of the feed center from the focus), or
(3) deviation of the antenna-feed wavefronts from spherical wavefronts.
From the point of view of the aperture it is immaterial which of the three
factors is operative. To tie in with the preceding discussion of the rela-
tion between the reflector and the feed pattern, the phase-error source will
be taken to be the third of the above, that is, the absence of a true center
of feed. Back-lobe interference will be neglected. The final results can
easily be interpreted in terms of equivalent errors arising from surface
distortion or defocusing. Let us then assume that the field-intensity
pattern of the antenna feed has the form

(32)

where d(~) represents the phase error in the feed pattern. A review
of the steps leading to the field intensity E(Ro, O,O) of the secondary
pattern on the axis, given in Eq. (18), will show that the only change
introduced in Eq. (18) is the replacement of [Gj(+)]~~ by [GJ(~)]~~e~~6t$J.
By precisely the same development as before, the gain factor is given by

By way of illustration, the effect of a quadratic phase error has been

computed for the primary pattern of the dipole-disk feed considered
above in connection with back-lobe interference. The phase function
is taken to be

The optimum angular aperture in the absence of phase error is taken as a

base for comparison and a is adjusted to produce a preassigned phase
error at the edge of the aperture for that case. The curves given in Fig.
12.6 are for values of a that result in phase errors of k/24, k/16, k/8, and
~/4 at the edge of an aperture of angle w = 61. The loss in gain is 2 per
cent for an error of A/l 6 at the edge, 6 per cent for A/8, and 20 per cent for
A,4. The effwt of a highly tapered illumination is shown in Fig. 126 by
SEC. 12.5] THE ANTENNA GAIN 431

thevalues of thegain factor forthe aperture of angle~ = 90. For this

value of Vthealcurve represents aphase error of approximately ~/11 at
the edge, while the a, curve represents an error of approximately X/6; the
corresponding losses in gain relative to the a = O curve are 2.7 and 3.6
per cent respectively. Since the gain curves are not very sensitive to
the illumination, the results obtained here for the cos4~ distribution
may be taken as characteristic; a conservative evaluation sets X/8 as
the maximum allowable phase deviation over the aperture.

0.7

0.6

0.5

0.2

0.1

0
0 10 20 30 40 503 600 700 ~o go.
Angular aperture ~
FIG. 12,6. Phase+rror effects on gain.

There is another aspect of the feed pattern that should be noted in

connection with phase-error effects. The discussion above is based on
the assumption that the reflector is illuminated by the main lobe of the
feed. For some purposes it may be desirable to accept the loss in gain
associated with a large angular aperture in order to suppress the side
lobes. The angular aperture, however, must not extend beyond the
first minimum of the feed pattern. Generally, in passing through a
minimum (more exactly a null) inthefeed pattern there is a discontinuity
of 180 in the phase. Inclusion of any portion of the pattern beyond
the minimum thus introduces completely out-of-phase illumination at
the periphery of the aperture, with a very serious reduction in gain.
Results similar to those obtained above are obtained when the phase
error arises from defocusing. .As shown in Fig. 12.7, if the center of feed
is displaced a distance &l from the focus along the axis, the phase-error
432 PENCIL-BEAM AND SIMPLE FANNED-BEAM ANTENNAS [SEC!. 125

function is
6 6,
Cos +. (34)
X=x

The k/8 criterion indicates a focusing tolerance related to the angular

aperture by

(===
130= ~ sec V.
.
=.
+ . In practice, the focusing condi-
L5=60COS+
.
60 F tion is not adhered to rigidly. It
is not practicable to tailor every
reflector to the feed, because fre-
quently it is necessary to inter-
change feed systems. In these
FIQ. 12.7.Defocusing phase errors.
cases the back-lobe interference
effect may be a decided asset; by defocusing to bring the back lobe in
phase with the main beam it may be possible to achieve an increase in
gain that far exceeds the loss due to defocusing phase errors. This is
particularly true with feeds such as the 3-cm-band double-dipole feed
which has a very large back lobe.
Design Procedures.The theoretical analysis may be summed up in
terms of design procedures for realizing a maximum gain factor:

1. The shape factor j/D is to be chosen so that the full angle sub-
tended by the reflector,at the feed is in the range between the 9- and
10-db widths of the primary feed pattern. A more exact value for
a given primary pattern is obtained by solving Eq. (26).
2. The focal length of the paraboloid should be an integral number of
half wavelengths j~ = mk /2 if the back lobe of the primary pat-
tern is 180 out of phase with the main lobe; if the back lobe and
main lobe of the primary pattern are in phase, the focal length
should be f~ = (2m + 1)A/4 where m again is an integer, Lnder
these conditions the back lobe ~vill be in phase with the paraboloid
beam and add to the gain. If it is not possible to satisfy these
requirements exactly, the feed should be placed at the point,
nearest the focus, at which the distance to the vertex satisfies the
half- or quarter-wavelength requirement.
3. Deviations from constant phase of the aperture should be kept
within A/8 and certainly should not exceed A/4. Two factors
contribute to phase error: distortion of the paraboloid surface and
deviation of the primary m-a~,efronts from spherical \vaves. With
reference to the first of these the phase-error criterion can readily
be converted to tolerances that may be allowed in constructing the
reflector. .4s concerns the feed, the phase-error criterion serves
S E<. 127] EXPERIMENTAL RESULTS ON SECONDARY PATTERNS 433

to define the point-source cone (cj. Sec. 8.1). The angular aperture
of the paraboloid should lie within the point-source cone.

12.6. Primary Pattern Designs for Maximizing GainMention

should be made of the technique of shaping the primary feed pattern
so as to produce uniform illumination over the aperture and thereby to
maximize the gain. The required primary pattern is obtained directly
from the expression for the aperture field in Eq. (11), For E(r, &) to be
constant, the primary pattern must be such that within the cone sub-
tended by the reflector at the feed

[GAW)l! = Const (35)

P
or

Gf(&,v) = Gf, S(7Ci~. (35a)

In addition, the feed must radiate no energy outside the angular aperture
T in order to realize the gain of 4mA/A. The value of GJOis obtained
from the condition

~vhence

(36)

It is, of course, impossible to produce a pattern having a sharp cut-

off, but the rec~uired pattern can be approximated quite closely. Tech-
niques of shaping the primary patterns of horn feeds are discussed in
Sec. 10.16. It will be noted that the paitern [Eq. (35a)] has a minimum
in the direction v = O. In order to produce such a minimum consid-
erable phase distortion must be introduced over the mouth of the horn.
Such feeds must be used with caution, for a concomitant effect of the phase
distortion to that of producing the desired intensity distribution may be
that of eliminating the center of feed. This \\-illresult in phase errors in
the field over the aperture of the reflector that may \\-ellcancel the gains
which might have been made by the uniform illumination.
12.7. Experimental Results on Secondary Patterns.The relation
between the secondary pattern and the apertllre-field distribution can
be studied by evaluating the expressions in Sec. 124 for the secondary
pattern for a number of different tylws of gain functions G, ( f,+). The
essential results of such calculations have been summarized in Sec. 6.8.
In this section the relation between the principal-plane patterns and the
aperture will be discussed by refelence to experimental data, The
 1

material will also serve as a presentation of the performance of several

of the more important types of feeds described in earlier chapters.

100 0 100 0
2 8 2
: 6
4 4 4
6 :
2 2 8
8
10 10 10 10
8 8 2
6 2 6
4 4 I
~ 4 G+ Z4
6+
%2 Z2 8%
s:
: 1.0 20 z E 1.0 20 m
~8 ?% S8
6 2$
t*4 & 6 4g
4s 54 1
22 6 6
~2 8 y
0.1 :0 0.1 30
t ; : 2
4 4 4
2 6 6
2
8
0.01 40 0.01 :0
10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90
Angle in electricplane Anglein magnetic plane
(a) (b)
l:IG. 12. S.Principal plane patterns as a function of diameter for a series of paraboloids of
j/D = 0.25; k = 4.00 in.: (a) E-plane; (b) H-plane.

The dependence of the pattern on the diameter for a given relative I

distribution over the aperture is exemplified by the series of patterns, 1

,
10 I I I I I -o
shown in Fig. 12.8, for a set of
8 %
6 I . 2 paraboloidal antennas all of the
Side lobes in
.# 4 the ~agn~tic plane
same shape, j/D = 0.25, and il-
\ 4
~ \ - 6 luminated by the same antenna
.: 2 \,) Flr~{~idelobe - 8 feed. The latter is a coaxial-line-
; 1.0 L 10 fed double-dipole feed of the same
8 \,
%6 \ 12 general type as was discussed in
E4 \
~ 14 ~ Sec. 8.9. The installations are
\\ Second
side lobe - 16 ~ the rear-feed type in which the
.@ 2
g b\ - 18 feed line lies along the axis of the
s 0.1 LargeangleI~ \
_side lobe 4 7 , 20 reflector, passing through its ver-
.S 8 \
.-%6 . 22 tex (cf. Sec. 12.11). The focal
54 =. , *4
length in each case is an integral
$ - 26
2 multiple of a half wavelength so
- 28 that the interaction between the
0.01 30 aperture beam and back lobe of
0 8 16 24 32 40 48
Diameter, in. the feed along the axis is the
Mo, 12.9. Depende,lce of relative height of same for all members of the
side lobes On aperture dialneter. series.
It is observed that ~vith increasing diameter the bearnw-idth decreases
IL. C,. Van Atta, Effect of Paraboloid Size and Shape on Beam Patterns,
RI, Report No. 54-9, .4ug, 5, 1942.
!

S~c. 12.7] IIXPI?RIMENTAI. ILE,~[JI,l.V

{)IV SllCOhrD,lR 1 PA TTERN,q 43,-)

and the side lobes move in toward the axis. The intensities of the side
\ lobes are diameter dependent, contrary to the results of the aperture
theory. However, as was noted in Sec. 12.4, such deviations are to be
expected for large values of x/D. Figure 12.9 shows the variation of
side-lobe intensity with diameter. The intensity approaches an asymp-
totic value as x/D decreases, becoming independent of the diameter as
the latter becomes large compared with the wavelength; the asymptotic
limit agrees with aperture theory predictions. The diameter dependence
of the side lobes may be accounted for only in part by the corrections
! to the aperture theory that are contained in the current-distribution
method for calculating the pattern (cf. Sec. 124). Another significant
1, factor is the overlapping bet~veen the primary feed pattern and the
aperture pattern. The overlapping also has the effect of filling in the
minima. It is seen that in some cases the side lobes have been fused
into the main lobe and appear only as shoulders. The same effect is
produced by phase errors in the aperture field (cj. Sec. 6.7).
The beamwidth also shows an anomalous behavior from the point
of view of aperture theory. According to the latter the product of the
beamwidth and D/A is a constant for a given distribution over the
f aperture. The products for each of the principal plane patterns of
TABLE12.1.BEAMWIDTH
AND GAIN FACTOR AS A FUNCTION OF DIAMETERS.
-! (E)is in radians

8 1 22 1.07 0 66
16 1 44 1.15 0.63
24 1.42 1.25 0.62
32 1.46 1.28 0.59
48 1 47 1 38 0 50

\I
Fig. 12S are listed in Table 12.1; it is seen that the product, for each of
the principal planes, varies with the diameter. The B-plane half ~vidth
appears to be approaching an asymptotic value that is proportional to
A/D. The difference between the E- and H-plane beamwidths can be
correlated with the directivity of the feed. Because of the directivity
of a single dipole in the E-plane, the pattern of the double-dipole system
is like~vise more directive in the E-plane than in the H-plane. Conse-
quently, the aperture field is more tapered in the E-plane than the
H-plane, and the former has a broader secondary pattern.
The variation of the gain factor S with diameter, as shown by Table
12.1, arises from the back-lobe interference effect. Along the axis in
each case, the back lobe of the feed adds to the field produced by the
reflector. Since the latter is proportional to D/A, the addition of the

I
436 PENCIL-BRA M AND ,SIMPI.E FAN VEI)-BEAM ANTENNAS [SEC. 127

back-lobe intensity produces a greater fractional increase in the total

intensity and peak power for smaller diameters than for large diameters
and correspondingly larger gain factors.
It will be of interest to record the data on the performance of the
waveguide double-dipole feed shown in Fig. 8.14 and of the stub-ter-
minated dipole-disk feed shown in Fig. 8.10 because of their extensive
use. The beam characteristics obtained with the double-dipole feed

TABLE 12.2.PERFORMANCE OF THE DOUBLE-DIPOLE FEED IN VARIOU SPARfiOLOIDS.

V is the distance from the vertex of the paraboloid to the front edge of the waveguide 7

Paraboloid Beamwidths I Side lobes, db down
1
v, s
Diam,
in.
Focal
length,
in.
cm i 3%
H, E, I E,

1.2 1.27 27 30 25

J.----L
30
18 4.5 10.8 0.61
1.25 1.2 24 29 25 29
18 5.67 13.9 0.61
1.15 1,15 26 27 26 30
18 6.0 14.2 0.64
1.13 1.20 22 28 23 28
24 8.0 19.5 0.63
1.25 1,16 22 28 25 28
30 10.0 24.0 0.60

at a wavelength of 3.2 cm are summarized in Table 122. HI and H2

are the first and second side lobes in the H-plane; El and Ez designate
the corresponding lobes in the E-plane.
The data for the three 18-in. diameter paraboloids can be compared
for the effect of tapered illumination; the longer the focal length the
less tapered is the aperture illumination with a given primary pattern.
The effects are quite evident in the decrease in the H-plane beamwidth
and the rise in the H-plane side-lobe intensity levels; the HI lobe of the
paraboloid of 5.67-in. focal length is an exception to the general behavior.
The E-plane characteristics are also anomalous. The discrepancies are
caused by the peculiar properties of the feed. As was pointed out in
Sec. 8.9, the centers of feed are different in the E- and H-planes; this
gives rise to small defocusing phase errors. In addition, the back-lobe
intensitv is large, and the position of the feed on the axis is determined
primarily by the optimum interaction between the back lobe and main
lobe rather than by the focal point of the reflector. The last three rows
form a sequence of paraboloids of the same shape; here too it is seen that
the behavior is not in accord ~vith the more systematic characteristics
observed in the set of patterns considered in Fig. 12.8. While the char-
acteristics of the feed leave much to be desired from the standpoint of
9
theoretical analysis of the patterns, the pattern characteristics given in
Table 122 are highly satisfactory for operational purposes.

1IV. Sichak, T)ouhlwdipo]eRectangularWave Guide Antennas, RL Report No.

51-25, June 26, 1943.
 SEC. 12.7] EXPERIMENTAL RESULTS ON SECONDARY PATTERNS 437

The stub-supported dipole-disk feed is the one to which the gain

factor curves in Fig. 12.5 apply. The patterns obtainedl with a reflector
of 30-in. diameter and 10.6-in. focal length, at a wavelength of 10 cm,
are shown in Fig. 1210. The angular aperture of the paraboloid is
100 o
8 \
6 - 2
4 \
4
\
- 6
2 \
\ - 8
10 10
8 - ,I
6 \ -.2
~
;4 \ 4g
n
I - 6$
~2
3 - 8:
\ g
~ 1.0 20&
z \ -2 2 I
I g:
~4 \ .4.~
( .- \ / 1,
.1 -6 g
r \\ a
L \
\ ,/ / - -A 8
0.1 \ ,- . / ;-
I --- \ 30
8
6 \ / . -2
/
4 i /
4
-6
2
-8
0.01
00 100 200 300 400 500 600 700 800 90!0
Angle
IrIG. 12.10.-Principal plane patterns of 30-in. -diameter paraboloid (f/D = 0.354
illuminated by the dipole-disk feed; A = 10 cm; E-plane; H-plane; @E = 10.2;
@H = 9.5.

v = 70.5, larger than the theoretical value of 60 for a maximum gain

factor. Whereas this represents a small loss in gain, the larger angular
aperture results in a more tapered illumination over the aperture plane
and better side-lobe characteristics. The half-power widths (in radians)
I are
@, = 1.403; e. = 1,26;.

1S. Breen and R. Hi;il t, Ar]tcl)na Feeds for &m. Stub-supported (~oaxial Line,
l{L I{eport So. 54-23, .Tunc 21. 1943.
438 PENCIL-BEAM AND SIMPLE FANNED-BhAM ANTENNAS @EC. 12.7

Attention was called in Sec. 8.3 to the fact that with a feed of this
type the axis of the beam does not coincide with the axis of the reflector.
The deviation is not shown in Fig. 12.10 because of its small magnitude;
it is less than half a degree. The squint phenomenon has great opera-
tional value; by rotating the feed about its axis, the antenna beam
is made to describe a cone, thus creating an effective cusp-shaped mini-

A

+

(a)

TB
A

(b)
FIG. 12.11.Production of squint by the asymmetric dipole: (a) current on the feed; (b)
distortion of the phase front.

mum along the axis of the paraboloid. The intensity differentiation in

the cusp is more sensitive than on the peak of the beam, and by this
technique the accuracy of pointing the antenna at a target is increased,
The production of the squint may be understood by reference to
Fig. 12.11. It will be recalled (Sec. 8.3) that the asymmetric dipole
termination gives rise to currents along the outer conductor of the coaxial
line, and the effect of the choke is to confine the line current to the
terminal region as shown in Fig. 12.1 la. The feed can be regarded as
two radiating elements: A the transverse dipole current and B the axial
current. The relative magnitudes and phases are determined by the
position of the choke. The primary pattern of A is the normal type
of pattern shown in Fig. 12.1 lb and gives rise to the field distribution in
the aperture that we have discussed previously (cf. Fig. 12.2). The
pattern of the element B has a null along the axis; it produces a
field distribution over the aperture in which the electric vector along
any diameter undergoes a reversal in direction through the center,
 SEC. 12.8] IMPEDANCE CHARACTERISTICS 439

which is equivalent to a 180 reversal in phase. The space relationship

between theaperture fields of A and Binthe E-plane areshown in Fig.
12.llb. If the current elements A and B are in phase, the fields at two
diametrically opposite points in the E-plane, such as z, and ztin the
figure, are a b and a +b, respectively; there is no distortion of the
phase front, providing b is always less than a. If, however, there is a
phase difference @between the currents, the resultant fields at the same
two points area be-@ anda + be@, respectively ;the resultant phases
are
b sin @
xl; tanl
abcos@
b sin @
xz; tanl
a+ bcos+

The aperture is no longer an equiphase surface; the phase front is tipped

with respect to the aperture as shown in the figure.
For a given position of the choke, the beam deviation varies with
frequency. In the case of the antenna whose patterns are shown in Fig.
12.10, the observed variation is as f ollows: A = 9.7 cm, deviation = 0.3;
A = 10.0 cm, deviation = 0.19; k = 10.3 cm, deviation = 0.38.
Finally, it should be mentioned that the current element B produces
cross polarization in the H-plane. This, however, does not affect the
accuracy of pointing, since the cross polarization is zero along the axis.
12.8. Impedance Characteristics.Another consideration of major II
importance in the design of an antenna is the impedance bandwidth. !
The impedance characteristics are the resultant effects of the impedance
characteristics of the antenna feed in free space and the mismatch pro-
duced by the interaction between the reflector and the antenna feed.
The latter problem was treated quite generally in Sec. 5.10. It was
shown that if the feed in free space is itself matched to the line, the
reflector gives rise to a reflection coefficient

(5.97)

in the transmission line. GJ($, ~) is the gain function of the feed; i is

the angle of incidence at the point (p,+, ~) on the reflector. If the feed
in free space is mismatched, with a reflection coefficient rj, measured
at the same point in the line to which r, is referred, the total mismatch
of the antenna is to a good approximation the sum

! r=rf+ r,; (37)

that is, the reflection coefficients add vectorially on the reflection

, coefficient chart (Sec. 2.8),
440 PENCIL-BEAM AND SIMPLE FANNED-BEAM ANTENNAS [%c. 12.8

If the wavelength is small compared with the focal length and aper-
ture diameter, the asymptotic value of Eq. (597), given by Eq. (5.98),
may be used. For the present case of the feed at the focus of the parab-
oloid, the radii of curvature Rt and Rn at the point of normal incidence,
which is the vertex, are both equal to 2j, and p. is equal to j. We have
then
> ~j(2k/+6)e
(38)
= 47rj

More generally, if the feed is on the axis near the focus, but at a distance
p from the vertex, the reflection coefficient of the reflector is

r,=foA
4?rj
~j(2kp+6).
(38a)

The magnitude of the reflection coefficient

(39)

,
,

[email protected]. 12.Variation of the reflection coefficient with position of the feed along the axis;
- experimental, theoretical curve as obtained from 13q. (38a).

can be determined by measuring the total reflection coefficient r as a

function of position of the feed along the axis. The feed reflection coeffi-
cient rj remains fixed, whereas r, undergoes a cyclic variation by virtue of

..
 SEC.128] IMPEDANCE cHARAcTERIS7ICS 441

the changing distance to the reflector. Over small distances about the
focal point Ir,l is essentially constant. As the feed is moved along the axis,
the total reflection coefficient therefore describes a circle in the reflection
coefficient plane corresponding to the rotation of r, about the terminal
point of the vector r,; this is illustrated in Fig. 12.12. The magnitude of
r, is determined directly from the radius of the circle.1 The measure-
ments can, in fact, be used to obtain the gain GO{of the feed pattern by use
1 of Eq. (39).

t Magnitude of I vs,~
Fead:~;n. stub-supported-
dipole-disk.
0.5
Wavelength: 9.1 cm
Theoretical
R
- Experimental
~ 0.4
18 dia.
z
.I!
*Q .4
f24 p /
~ 0,3 /
Y
.-E
5
=0
a 0.2 7
r -48. ~ ~30

!Aa!f3fk
IJIo. 12.13.Contribution
1 2 3
~ (cm-j
4
x 102
of the paraboloid to the reflection coefficient
5 6 7

as a function of
focal length.

The data presented in Fig. 12.12 \vere obtained with the stub-sup-
ported dipole-disk feed shovm in Fig. 8.10 and a paraboloidal reflector
of 10.6 in focal length and having an aperture diameter of 30 in. The
gain of the feed was evaluated by graphical integration of its primary
pattern, and the theoretical curve of Fig. 12.12 was then obtained from
Eq. (38a), the constant 6 being adjusted to make the theoretical and
experimental values agree at the focal point. Similar studies with the
same feed in a series of reflectors of different focal lengths gave the results
shown in Fig. 12,13, demonstrating the applicability of Eq. (3~).2
It is seen from Eq, (37) that, the process of matching the antenna

1 S. Silver, contribution of the Dish to the Impedance \Iismatch of an Antenna,

RI. Report No. 442. Sept. 17. 1943.
z Ibid.
442 PENCIL-BEAM AND SIMPLE FANNED-BEAM ANTENNAS lSEc. 12.8

reducing r to zer~by means of a transformer in the line can be regarded

as that of transforming the mismatch rf of the feed in free space into
r,. Therefore, if an antenna is matched with the feed at a position
pl, it will also be matched with the feed at positions pl f nX/2, where n
is an integer, for r, has the same value at all these points [cf. Eq. (38a)].
Furthermore, the feed can be placed in an entirely different paraboloid;
and providing the distance from the vertex is pz = pl + nA/2, a good
impedance match will be obtained. A small difference will be observed
from the value obtained with the original reflector because of the different
magnitude of r,, but the phase relations between rf and r, in each case
are the optimum for minimizing the total reflection coefficient. If the
distance pl is chosen to be the closest to the focal length that is equal to
an integral number of half wavelengths, the feed may be placed at the
corresponding half-wave points in other paraboloids with both the proper
conditions for impedance match and the constructive superposition of
the back lobe and main lobe being maintained. 1
The seriousness of the mismatch caused by the reflector lies in its
frequency sensitivity. Since the focal length is large compared with A,
a small change in the latter produces a large change in the phase of r,.
The antenna can easily be matched at one wavelength kOby a conven-
tional type of matching transformer (cj. Chap. 7). However, the char-
acteristics of the transformer do not vary rapidly enough with frequency
to follow the rapid change in the phase of r, and in any case do not
necessarily vary in the proper direction. The total reflection coefficient,
therefore, varies rapidly with frequency. For this reason it is necessary
to eliminate the mismatch caused by the reflector by other methods in
order to realize satisfactory impedance characteristics over a wide fre-
quency band.
There are two obvious solutions to the problem. One is to reduce the
reflection coefficient of the reflector to zero. 2 For this purpose we musi
return to Eq. (5.97), which formulates the reflection coefficient as a
superposition of contributions from the entire reflector surface. The
matching technique that suggests itself immediately is to divide the
reflector into two areas, which give integrated effects of equal mag-
nitude, and then by a small displacement of one of the areas with
respect to the other to make their contributions 180 out of phase.
Since only a small displacement of one area with respect to the other is

LH. Krutter, R. Hiatt, J. Bohncrt, Some hlatching Properties of Antenna

Feeds, RL Report No. 54-13, Nov. 17, 1942.
a N. I?lson and A. B. Pippard, Wide Band llatchir,g of tVavcguide Radiators and
Paraboloids) AD RDIIU (British) Report Xo. 220; W. Kock, llethod for Reducing
Reflection IXfects in Antenna Feeds, BIL Report 3111-42-160-92; S. Silver, Analy-
sis and Correction cf the Impedance Jlismatch Due to a Reflector, RL Report No.
810,
 SEC. 12.9] THE VIIRT17X-PI.A 71( Jf.17(CHJiVG TE(:HN[Q[E 443

involved, the matching process is not very frequency sensitive. The

bandwidth of the antenna is then largely determined by the impedance
characteristics of the feed in free space. A second solution is to render
the feed insensitive to the reflected radiation. This will be accomplished
if the polarization of the latter is rotated through 90 by the reflecting
surface. Such a rotation can be effected by introducing a suitable grating
over the surface of the reflector. The details of the two methods will
be developed in the following two sections.
12.9. The Vertex-plate Matching Tech.nique.-A complete evalua-
tion of Eq. (5.97) involves considerable numerical work. For the present
purposes the computation can be simplified by replacing G(*, ~) by a
circularly symmetrical function G(+) which is the mean value of the gain
functions in the E- and H-planes of the feed pattern. By virtue of the
symmetry we can take as the element of area d.! the circular zone sub-
tending the angle d+ at the focus. It is more convenient to base the
integral upon the projection of dS on the aperture plane:

2m- dr
dS = . (40)
*
Cos
() 2

The gain function G(x) can be expressed as a function of r through the

relation

(41)

Since z = 4/2, we have for the reflection coefficient ccmtributed by the

portion of the reflector of aperture radius r,

(42)

constant terms in the phase being discarded. Changing variables to

=5 (+3)

we
b
get

G(v) _
jzf

h
r(~) = ~ ,e A dv. (44)

o
1+;
)
444 PENCIL-BEAM AND SIMPLE FANNED-BEAM ANTENNAS ISEC. 12.9

If we take the real and imaginary parts of r(u),

~(~) = + G(v)
~ ()
COS d
A
V dv,

~(v) = ~
/ 0
()

1+;

G(v) d
(45)
~sin u dv,
/ o 1+; () A
() 1

and plot Z(v) against R(v), we obtain a reflection coefficient spiral for the
paraboloid as shown in Fig. 12.14a. The vector r. from the origin to any
point on the spiral is the reflection coefficient due to the portion of the
I
I (v)

t~ I

(a)

t (b)

Zc
Jt- c

(c)
FIG. 12. 14.Elimination of the mismatch caused by the reflector: (a) reflection coefficient
spiral; (b) effect of infinitely thin zone plate; (c) final position of zone plate.

paraboloid whose aperture radius r corresponds to that point v on the

spiral. The reflection coefficient due to the entire paraboloid is given by
the vector to the terminal point corresponding to
D,

= p

For any particular case R(v) and Z(U) can be evaluated numerically once
the gain function of the feed has been measured, and the spiral con-
structed accordingly. It has been found that in many cases the function
 SEC, 1242] THE VERTEX-PLATE MATCHING TECHNIQUE 445

can be fitted satisfactorily by an exponential

~r,
GOe P ~GOemV, (47)

GO being the gain of the feed, Assuming this form we can evaluate
!,
R(o), l(o) analytically:

Thelirnit point of the spiral, corresponding too = ~, comesat

=;(a:~~); g) =44%
This isto be contrasted on theone hand with the Cornu spiral in which
R~ = 1~ and on the other hand with the circular aperture diffraction
spiral in which R~ = O and 1~ is the radius of curvature of the initial
portion of the spiral. Thernagnitude of the vector to the limit point is

(50)

,- If the aperture of theparaboloid is large, thedifference between lr,l, the

reflection coefficient due to the entire paraboloid, and Irml is small.
. We observe further that if j>> A, Irml becomesGok4~~, independent of
the illumination function. This is the result obtained previously from
Eq, (38).
It is further of interest to note that the radius of curvature of the
spiral is
G,A
(51)
p = Gf -a
In the limit of a very large aperture and ~ >> X, the center of curvature of
t the spiral in the neighborhood of , = O coincides with the limit point.
Under these conditions \ve obtain the result that the spiral has the
form of the diffraction spiral for a point on the axis of a circular aperture,
:1 independent of the feed illumination function.
 .4

446 Ifi:.V(!II,-I! l</l.1/ .t .Vl) 51. }111>1<1,.1.V.VI(l)-1<1<.I .lf .INIEJV,VA,Y [Ss,, 1210

l_he method of impedance correction is as follo\\-s: The perpendicular

bisector of II,, the reflection coefficient due to the reflector, is erected,
and its intersection point ~~, on the spiral is determined [refer back to
Fig. 1214a). This divides the surface intt~ tlro zones, one \\-it,hinthe
radius r. = j(~,) contributing thr vector 1., the other the region outside
r, contributing Ib, The magnitudes of r. and I~ are equal, No\v sup-
pose that an infinitely thin plate of ra(lius r, is placed against the
surface (Fig. 12.14 c), The path from F to the edge isj + z,; the path to
the center is j z.; the average path length is ~, The plate thus
rotates ra onto the R-axis to coincide in phase \vith the contribution
from the vertex area of thr partiboloid (Fig, 12.14b). It is desired to
bring ra 180 out of phase \vith [~, This is acl;ieved by rotating r.
by moving the plate for\vard a liistancc ( (<mmaking th(~-plate of ~hat
thickness),
Al),
t= (2n+l); -.<T! /L== (), 1,2, , (52)

~. being the angle between II* and the R-axis. It is evident that only
a small portion of the spiral in the neighborhood of L = O and the terminal
vector It arc rcqllirecf to determinr tile pw-ametrrs of the correction
plate, In most cases r, can be replaced for this purpose by ]~, and the r
final position of the plate is adjusted empirically to compensate for the
error. ,
In the limit A/~ = O and large apertures, the parameters of the correc-
tion plate become practically independent of the aperture and primary
feed illumination, providing the latter is not too sharply peaked. We
ha~-e noted above that in the limit indicated the resultant r, differs
negligibly from Im ~vhich (in this case) lies on the I-axis. .Mso from
Eq. (51) it is seen that if the primary feed illumination is not too sharp,
i.e., magnitude of a is not too large, the initial portion of the spiral can
be regarded 11-ithsmall error as a circle of constant ra(lills ir~l und renter ,
on the I-axis. W%en the procedure t)(ltlined in tllc preceding paragraph
is applied to this case, it is foun(l that the diameter of the corre(,tion
plate is

(53)
and its thirkness

We have assumed that the current distribution over the correction .,,I
plate is, except for phase, the samr as that (Jvcr the corresponding area
of th( ptir:ll)t)lfji(l. In general the area of the correction plate is small, and i
ii.hrn a small obstacle is irradiated, there is an appreciable (,~lrrent dis-
[ rll)lltif)n over the shad(nv area of the (Jhstacle as lvel] as on th(, illllminated )
 SEC.1210] R07A7ION OF POl>At{lZATloK 7IWHVIQIJR 447

region. To eliminate the former it is preferable to use a plate of the

thickness specified by Eq. (52), making good electrical contact with the
paraboloid, rather than athin plate set at the specified distance.
The one major objection to the vertex plate is the deleterious effect
onthe secondary pattern. The displacement of thevertex area produces
a phase error in the field over a corresponding area of the aperture, with
a resulting loss in gain, increase in beamwidth and side-lobe intensity.
If the specifications on the side lobes are very stringent, the vertex-plate
technique cannot be used.

I?IG. 12.15,Quarter-wave grating to rotate the polarization of the electric vector and
1 elimm
ate the n!islllatch.

12.10. Rotation of Polarization Technique.-The electric vector of

the wave reflected by the paraboloid can be rotated through 90 with
respect to the incident wave by means of a quarter-wave grating. The
system is illustrated in Fig. 12.15. The grating is made up of parallel
!. plates cut to the contour of the reflector; the plates are oriented to make
an angle of 45 with the E-plane of the feed.
The grating makes use of the property of parallel plates (Sec. 7.15)
*
that they v-ill not support free propagation of a wave having the electric
vector parallel to the plates unless the spacing s between them is greater
\ than X/2. If s < A/2, the wave is attenuated; if s < X/S, the parallel
plates reflect almost completely an incident ~rave with the electric vector
parallel to the plates. With the grating oriented at angle of 45 with
respect to the E-planej the incident electric vector can be resolved into
@ two equal components, one parallel to the plates and one perpendicular
to them. The plate spacing is such that the parallel component is
F. reflected, \vith a change in phase of 180. The perpendicular component,
i on the other hand, propagates between the plates with free-space velocity.
i If the depth of the plates d is h/4, the latter component after reflection
from the paraboloid emerges from the plates in the same direction as it
had on entry. Combination with the reversed parallel component then
results in a resultant vector perpendicular to the E-plane.
 Preliminary experiments conducted at the Rodiat,ion Laboratory to
test the effectiveness of the techniclllc gave promising rrs~llts in so far
as the impedance characteristics ]fere concrrned. The effects on the
secondary pattern ~verc not determined. It is to be expected that the
grating does not function properly at the edges because of the oblique
incidence of the primaly radiation, th~ls intr(xlucing phase distortion.
Further study of the sl[l)ject is needed in order to evaluate the relative
values of the grating and vertex-plate twhni([nes,
12.11. Structural Design Problems.tn antenna must generally
mret certain mcchani(~~l specifications sllch as a minimum ~vright strengtl-.
factt)r, l(JIV~i-ind rcsist~n(c, an(l yisllal trans~)alcncy in addition to ful-
filling the requirements on the secondary pattern.

((1) (/,) (c)

VI<;. 12.16,- I{e:tr-feed ;Lt)dfrollt-feed ltlstdl:htlotls: (a) re:u-fw<l tcrlll]iquc (or a dipole-dis~
feed; (b) rear-feed [erlLIIJque for :1 horII; (c) fro!lt-feed terl, nique for :L horII.

//ww-~ecd a~d Jron (-jecd Sys[cws.-rlle first ftictor to be considere(i

is the type of feed installation. f\vo grnertil inetho(ls-rear-f eed and
f rent-feed inst allatioll-arr illust rtited in l~ig. 121 (i. ~he rear-fce(f
install~tti[)ns (1ig. 12.1 (k] and b) ha~e the advant +yw of compactness
and retluiring a minimum length of trunsmissif)n line. The latter has
important bearing on the impwlance presrnt ed by the syst cm at the
grurrator ternlinals, If the focal length is short, a simple flange con-
nection bet Iveen the transmission line and re~lwtor is sufficient to 9
?Ilpp, )rt, the fecd system. If the f,)(:d Irngth is Iargc, a more extended
rollar sllch as is sho\vn diagr:lnlrn:~ti(,:~lly in 1<ig. 12.1(k is nrcessary to
I)f-cv(mt free phiy of t tle fe(xl. The r(ar-frcd inst;lllatiou of a horn, such
as illust ra(,d in I~ig. I2. I(ib, is fe:lsil)l{~ only at s}ltjrt Iral-elengtlls (3 cm
or less). ]~vcn for tllc Iattrr it is not to be r(,(t)l]~i~~(,]lcl(([l)(WaUSC of the
asymmetry and possible phase distort ion int reduced into thc primary
pattern.
lhe front-feed installation (l~ig. 121 (k-) is r(,(,onll~~ell(l(,(ifor all horn i
feeds. It suffers from onc scriolls (1{fe(t of OIW(rllcting too much of thr
aperture. The intrrfcrenee is re(luced some\\.hat if the ~~tivcgllide is I
placed in the 11-plane. This may make it necessary to put a t~vist in
 SEC. 12,11] STRUCTIJRAL DESIGN PROBLEMS 449

the waveguide in order to orient the horn properly with respect to the
horizontal plane.
Grating and Screen Reelectors,-The weight and wind resistance of
the paraboloid can be reduced considerably by replacing the continuous
reflector surface by a perforated surface or a grating structure. An
example of an antenna using a perforated paraboloid is showm in Fig.
1.5; examples of grating reflectors will be found in Fig. 12.23 and in
several photographs in Chap. 13.
The reflectivity of the perforated surface is insensitive to polarization.
.( The perforations can be regarded as short waveguides designed to +]
far beyond cutoff for the frequency band over which the antenna is to
1
used. For example, if the reflector is a wire screen with square openin s, ~
the edge length a of the openings must be such that t
r i

{.
This is the condition for cutoff in a square waveguide.
The gratings are sensitive to polarization. The space between the
grating elements may be thought of as \vaveguides beyond cutoff fo~
t the electric vector parallel to grating element. The grating element$-

~ :.
I

,,, ffl-xis:
FIG. 12
(a)
17.Grating reflectors:
(b)
(a) broadside
(c) L.:!

strips; (b) round bars; (c) edgewias strips.

may be divided into three groups: (1) broadside strips, (2) bars, and (3)
edgewise strips; these are illustrated in Fig. 12.17. The various types
of gratings have been studied experimentally to determine the relation-
ships bet~veen the grating dimensions and wind resistance and transmis-
sivity. There are t~vo major restrictions that apply to all gratings:
1. The electric vector of the incident ~vave must be in the plane
determined by the incident ray and the axis of the grating element.
i 2. The center to center spacing of the elements must be less than
A/(1 + sin 6), where o is the angle between the incident ray and

4 1W. D, Hayes, Grating and Screens as Nlicrowave Reflectors, RI, Report No.
54-20, Apr. 1, 1943.
450 1E,V(,lL-BA.4M AND SIMILA IA,YNEI)-BEA lvf A,VTENN.4,Y [Sm. 12.12

the normal to the axis of the grating element. Larger spacings

cause the appearance of undesirable higher-order lobes in the
secondary pattern.
The edgewise strips are generally to be preferred. Their transmission
characteristics are summmized in Fig. 12.18, ~vhich gives the relation
between the strip depth and the spacing for fixed values of transmissivity.
The properties vary, of course, with the width of the strips; the reader
is referred to the report by Hayes for more extensive data,. The variation
of the depth of the strips to control the r-f transmission has a negligible
effect on ~vind resistance; both can be made quite lo~v. Jlechanical
0.40 I 1
v-n?

t-----r
0.4 1 35
T=per cent r-f transmission * - ~ i , II
0.35 normal incidence J I 1 [1
II r r

0.30

Bti
&

\_

0.05

0 1 I I I I I I I I I J
o 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0,50
s/h, center to center spacing, wavelengths
FIG. 12 lS. GratinR of edgewise strips: l{el:ltiotl between strip depth and spaci,, g f<,r
constant tranhmisbion.

rigidity can be obtained by proper bracing. The strips also have the
advautfige that the reflector shape can bc (Jbttiined by a cutting oper-
ation; in makinz up a paraboloid all the strips ran be identical punrhings .
of flat sheet metal.

SIMPLE FANNED-BEAM ANTENNAS

12.12. Applications of Fanned Beams and Methods of Production. i
The singnhm advantage of a pencil beam for lorating a target \vith accu-
racy is offset by the difficulty oi intercepting L target in the course of a
t
random. search Lwcallse t}le beam u)~,ers {July a narn)!v cone of space at a
 SE{. 1213] SYMA4k7RICALL,Y CUT P.4&.t BOLol~S 451

given instant. Further difficulties are encountered in the case of

antennas on ships; in the course of the roll and pitch of the ship the beam
swings into the water where it serves no purpose or tip above the horizon
losing its effectiveness in locating surface vessels. To counter these
various difficulties it is necessary to sacrifice the directivity by flaring
the beam in one of its principal planesgenerally the vertical plane.
By retaining the narrow width in azimuth, resolution is maintained in
this aspect and the radar echo technique supplies information on range.
The present chapter concerns itself only with simple fanned beams
which may be thought of as being developed by distorting the almost
circularly symmetrical beam into a symmetrical elliptical beam. The
more complex fanned beams which are designed for highly specialized I
operational functions will be treated in the next chapter. From the
general relations developed in Chap. 6, between the symmetry of the
aperture and aperture field and the symmetry of the beam, two basic
techniques suggest themselves: (1) to use an aperture ~~ith two highly
different dimensions in the principal planes, the beamwidths in the prin-
cipal planes being inversely proportional to the aperture dimensions, and
(2) to taper the illumination differently in the t~vo principal planes.
The second of these may be dismissed as an isolated technique because
*
the bearnwidth is not sufficiently sensitive to the illumination. The
only practical technique, therefore, is the first, of using an aperture with
$. suitable dimensions in the principal planes. The illumination technique
may be used as auxiliary to the other method.
The fanned-beam antennas take the following forms: (1) an ovoid
section of a paraboloidal reffector with a point-source feed at the focus,
(2) a parabolic cylinder with a line source producing a rectangular
aperture, and (3) a parallel-plate antenna consisting of a parabolic cylin-
drical reflector illuminated by a simple feed at the focus and located
between parallel plates that are perpendicular to the generator of the
cylinder; this likewise produces a rectangular aperture. Design tech-
niques will be presented for each of these types of antennas.
. 12.13. Symmetrically Cut Paraboloids.The simplest procedure
is to cut a paraboloid symmetrically by a pair of parallel planes as shovm
in Fig. 12.19a. The long dimension will be denoted by d,, and the nar-
row dimension by d~. The results obtained from a circular aperture ~vith
many types of feeds and paraboloid shapes show that the beam~vidth
is in the range (1.2 + 0.2)h/D. These results have been extrapolated
to the cut paraboloid, and the relations betneen the dimensions of the
I
latter and the principal-plane beam \\-idthsare generally taken to be
i A A
(56)
1=12Z; =2z
f symmetrical primary feed pattern
It is qllilc e\.iflenl 111:{( a rirclllady
 4
I
452 IEiVCIL-BEAM AND SIMII>E FA.%-NED-B17A If A NTENNAS [SF,C.12.13

is unsuited to illuminate the reflector; a large fraction of the energy would

be ~vasted in spillover. The primary feed pattern must be shaped to
the same symmetry as the reflector. Taking the results for optimum
performance of a circular aperture again as a criterion, we may require
that the 10-db width of the feed pattern in a given principal plane be
equal to the angle subtended by the reflector at the feed in the given
principal plane.
Horns ~vith rectangular apertures lend themsel\-es particularly well
to the design of suitable freds, since the beam\vidths in the tlvu principal
_ _ ,/ --\
/ \ Y
/ \ / \
/ \ / \
/ / \
/ \
,r_._ . . ..].g.g

\ / 1,
\ / 4
\ / \ /
\ / \ /
\ /
=. / +____
- _-

planrs can bc controlled virtually independ(,nt]y of one another by choice

of the prin(ipal plane dimensions, The relation bct\vecns the primary
pattern 10-clb beam~vidth and the horn dimcnsiorrs are given by Eqs.
(10.52) and ( 10.53). The primary patt crn beaulwidths that are required
are dutcrrnined by the dimensions 01and d? of the reflector aperture and
the focal lvrrgth. The latter should br chosen as small as possible to keep
the primary pat tern 10-db \\-idthlarge; othrrnvisc tllc design of a pra(,tiral
horn brcomcs vvry dificlllt. Difhcllltics are enco~lntcred if the ratio ,
dl ~dz is too hwgc; sin(,e t}lc dimension of the horn in the all-plane must be
so n~ll(,h sm:Lller than thtit in the CZz-planethat the resulting horn has
li id(,ly differcnt centers of fcwl in the t\~o planes. This will give rise to
serious phase errors :md 10ss in gain.
Thr primary pattern of the horn designed to meet the principal plane
reqllirrments has an clliptictil cross section, Consequently, the equi-
intensity illumination contours on the reflc(tor are also elliptical in shape.
~herc fire several rcastlns for r(l(t ing the ref{ector along s(lch a conto(lr 4
as shown in lig. 12. IW. It is f,]{lnd in ~encrul that the gain factor
increases and the gen-r:ll fc:lt IIlcs of 1he pattern arc improved I]Y a redur-
1
tjion in side lolms in thr prinripal planes. Tllc basis for this lies in thr
fart, that t hc cflc(,ii\e illllnlill:it ion for s:lJ, thr d-prin(ilxd plan{ at J
Sm. 12.14] (!ONTO(JR C1JT7ING OF lcEFI,ECTOR,$ 453

given point on the dz-axis is the integrated intensity across the aperture
parallel to the d,-plane. With the aperture of the type shown in Fig.
12.19b the integrated area tapers along the all-axis, and the effective
illumination is, therefore, more tapered than in the corresponding case
of Fig. 12.19a, hence the improved side-lobe characteristics. The ovoid
shape of Fig. 12. 19b also has advantages of low wind resistance and
smaller moments of area and inertia which are of considerable importance
in connection with the mechanical problems of support and rotation of
the antenna.

100 0
8
6
4
2
4
I
6
8
-- 10
8s zg
g
~ : 4A
6?
32 8~
g 1:0 20:
8 2$
:6
4K
z 4
6 .2
.-%!2 8:
~
; 0.1 30
s
2
: 4
6
2
8
0.01
-180 _160_140-120-100-80-60-40 -20 oO 20 40 60 80 100 120 140 160 18:0
FIG. 12.20.E- and H-planepatterns of the beavertail antenna, shown in Fig. 13.12a.

.kn antenna using the symmetrical ovoid-shaped reflector is shown in

the follmving chapter in Fig. 13.12a. The dimensions of the reflector
are dl = 20 ft, da = 5 ft, and the focal length f = 5 ft. The
feed that was finally adopted for this antenna is a flared box hornz
designed to meet the illumination requirements in the principal planes,
The secondary patterns of the antenna are shown in Fig. 12.20, The
ratio of the half-power widths @E/@H is 0.29, and the ratio of the aperture
dimensions d,/dl = 0.25. The H-plane side-lobe levels are all down
belo~v 17 db with no prominent wide-angles lobes; the E-plane side
lobes are all below 23 db, showing that the illumination is properly dis-
tributed over the reflector.
12.14. Feed Offset and Contour Cutting of Reflectors.-The symmet-
rically cut paraboloids have the dra~vback that the feed must be located
1c. s. Pno, The Beavertail .4ntenna, RL Report No. 1027, Apr. 9, 1946.
z S. J. ~hmn, Flared BOX Horn,; RL Report ho. 653, July 9, 1945; see also
Chap. 10.
in the center of the aperture. In this position it is in the path of the
reflected rays from the most intensely illuminated area, and hence the
mismatch introduced by the reflector is quite significant. Furthermore,
the use of a horn feed introduces a large section of waveguide, which in
large reflectors necessitates additional supporting structures; these
together with the feed block out aperture area, causing a loss in gain and
increase in side lobes (cf. Sec. 6,7).
Both of these defects are eliminated by the offset feeding technique
which is illustrated schematically in Fig. 12.21. The center of feed is
placed at focus of the paraboloid ?s in the previous case, but the horn is
tipped so that the peak of the primary pat-
tern makes some angle ~0 with the para-
Peak ~oloid axis. The major portion of the
illumination
w. A . lower section of the paraboloid is discarded.
. az The dimension d, is again determined from

([
.
the secondary pattern beamwidth by Eq.
+;q \ (56), The offset feeding removes the horn
& and its supporting structure out of the way
,:,,;, 1227 I.offset
of the most intensely illuminated area of
feeditig twll-
lllque.
the aperture with resulting improvement
in gain and in side-lobe characteristics.
The redllction of the mismatch can bc understood in terms of the geomet-
,
rical-o[>tics Dicturc that the radiation returning to the feed comes from the
:Lreaal:ollndthe vertex of the paraboloid. The-magnitude of the mismatch
is given b.v a relation equivalent to Eq. (39):

(57)

Ithere GJ($O)is now the gain of the feed in the direction along the axis.
1ly offsetting the feed the reflection coefficient is reduced by the ratio
GJ(+o)/GJ,, where GJ, is the peak gain.
The design proccdurc is essentially the follo\ving: The dimensions d,
and d.l arc chosen in accordance \vith the bearnwidth relations [13q. (56)].
The focal length and the dimensions of the horn aperture are chosen as
though the reflector is to be cut symmetrically; the angle subtended by
d, at thc focus sho~dd not exceed 160. The horn is constructed, pres-
surized, and matched by the methods discussed in Chap. 10. I,et r, be
the residual mismatch of the feed and r the allowable total mismatch
of the antenna; the allo~vable reflector mismatch is then

(58)

[sing a circular paraboloid of the focal length of the final antenna, the
I
paraboloid reflection coefficient is determined as a function of the feed

d
!,

s~c, 1214] CWN7Or?licUYYING OF RIWLIKTORS 455

offset ~0 by the circle diagram method referred to prewcusly (qf. Fig.

12.12), or the peak gain G,o is determined from the mismatch at the
angle lo = O, and the mismatch at any other angle is computed by means
of Eq. (57) from the knowledge of the primary pattern.
For the chosen value of ~o the primary feed pattern is transformed
into equi-intensity illumination contours on the surface of the paraboloid
by taking into account space attenuation according to the inverse square

1.2

1.0

0.8

0.6

0.4
g
f 0.2

-0.2

-0.4

-0.6
o 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
x
-7
J
FICJ.12.22.Constant intensity contours in paraboloid aperture (horn feed axis tilted 20
relatlve to paraboloid axis).

law, An example of such an equi-intensity contour plot is sho\vn in

Fig. 12.22. The paraboloid is then cut to follow an equi-intensity con-
tour, generally chosen as the 14-db contour.
A number of antennas have been designed according to this procedure
with very successful results.1 Figure 12.23 is a photograph of the
antennaz to which the constant intensity contours (Fig. 12.22) apply.
Thereflector dimensions aredl = 54 in., dz = 24 in., j= 14.5 in. The
horn aperture dimensions are2cmin the E-plane, 6.Ocminthe H-plane
\vith flare angles of 10 and 40 in the respective planes. The offset
angle is20; this waschosen so that r, < 0.04in order that the resultant
mismatch of the feed and paraboloid over the entire band of 8600 to
1 T. J, Kcary and J. I. J30hnert, RL Report h-o. 659, hfar. i, 1345; RI, Iteport ho.
660, Feb. 19, 1945; RL Report No. 779, Aug. 30, 1943; J. I. Bohnert and H, Krutter,
It I, ReportNo,665, Feb. 7, 1945.
T. .J, Keary and J. I. Bohnert, RI, Report h-o. 659, Jlar. 7, 1945.
 456 PLNCIL-BEA M AND SIM1>LE lAIVNEl~-B13Ahf AIVT~JVN/!S [SIX 121!

Fm. 12.23.Fanned-beam antenna using the offset feed technique.

100
8
6
4

2
0.01
-40 -30 -200 -10 0 100 200 3f30 400
Angle
FIQ.12.24.Principal E- and H-t)la:le polar diagrams of the antenna shown in Fig. 12.2:3.

.-
- .
 skc. 12.15] TIIE PARABOLIC CYLINDER AND LINE SO lJltCE 457

I
9700 Mc per sec should represent a reflection coefficient less than 0.091.
The performance of the antenna is demonstrated by the E- and H-plane
patterns shown in Fig. 12.24. The ratio of the beamlvidths @./@.
again is very closely equal to the ratio dz/dl. The low level of the side
lobes attest further to the validity of the design procedure.
The elimination of one of the planes of symmetry by the offset feeding
technique produces one serious effect. The process destroys the sym-
metry of the cross-polarization component of the aperture field leading
to cross polarization in the plane of the large dimension of the aperture.
The cross-polarization pattern has lobes on either side of the main lobe
in the plane of the narrower beamwidth, which may seriously affect the
0

4
121G.12.25 --- Parabolic cylinder and line source.

performance of the system. Cross-polarization studies should be made

in the narrow-width plane for all antennas of this type.
12.15. The Parabolic Cylinder and Line Source.In principle a sim-
ple fanned beam is most easily produced by using a rectangular aperture
with a separable type of aperture field such as was discussed in Sees. 65
and 6,6. The principal plane patterns are then determined completely
by the aperture dimension in the given plane and the field distribution in
that aspect. There is no interaction between the distributions in the
principal planes. A second advantage is the reduction of cross
polarization.
The required aperture configuration and field distribution are readily
obtained by illuminating a parabolic cylinder by a line source located
along its focal line. An antenna of this type is shown in Chap. 1, Fig.
1.6. The general theory of such systems has been developed from the
standpoint of the reflector currents in Sec. 5.9 and from the aperture
field standpoint in Sees. 68 and 69. We shall here simply state the
resu!ts which are particular to the parabolic cylinder. In Fig. 12.25,
the line source is taken along the x-axis which is also the focal line of the
parabolic cylinder. Let 1be the length of the source and v be the angular
aperture of the reflector. The performance of the system depends on the
 ~
4

458 PENCIL-BEAM AND SIMPLE FA NNE1)-BEAM AN TE.VNAS [SEC. 1215

fact that the reflector is in the cylindrical wave cone of the source. It is,
therefore, necessary that
1,
l>> A, pm < ~) (59)

where p- is the maximum radial distance from the source to the reflector.
For wavelengths greater than about 10 cm conditions (59) imply that
the length of the cylinder is greater than the height of the aperture.
It is clear that all rays from the source incident on the reflector in a
plane parallel to the yz-plane are reflected in that plane into a family
of rays parallel to the z-axis. The reflector thus produces a uniform
phase distribution over the aperture. Also, since the reflected rays arc
parallel, the field intensity at a given point on the aperture is the same
in magnitude as that of the reflected field (or incident field) at the corre-
sponding point on the reflector. The intensity distribution, F(z), in the
z-direction over the aperture is, therefore, the same as that of the line
source, and the aperture distribution in the transverse direction is deter-
mined entirely by the two-dimensional gain funct ion G(+) of the cylin-
drical wave zone of the line source (cf. Sec. 5.9). Evaluating the field in
the forward direction by means of Eqs. (5.86) and (588), we find that
for both the longitudinally and transversely polarized systems the gain
is given by

The gain factor S = G.,fx2/4mA is, therefore,

The term involving F(z) gives the effect of the deviation from uniform
illumination along the z-direction. The second term gives the depend-
ence on the angular distribution of the primary pattern. As in the case
of the paraboloid of revolution there is an optimum angular aperture
for a given feed pattern that represents the compromise between spill-
over and tapered illumination over the aperture in the y-direction. The
optimum angular aperture can be found by graphical methods as \vas
done in Sec. 125 for the paraboloid of revolution.
For maximum gain, the distribution F(x) should be equal to unity,
This, however, gives maximum side lobes in the longitudinal pattern
that is, in the planes containing the line sourceas compared ~rith !
tapered distributions. The longitudinal pattern can be studied as a two-
dimensional problem, independently of the transverse pattern. .\ll of ,
f the results of Sec. 6.6, other than the actual values of the gain, ran l)e
r
sEc. 1216] PARALLEL-PLATE SYS7EMS 459

applied here without modification. The gain is affected by the trans-

verse distribution as is shown by Eq, (60).
The essential difficulty with antennas of this type is in producing an
efficient line source. Linear arrays such as are discussed in Chap. 9 are
frequently used. The impedance characteristics are generally poor due
to strong interaction between the reflector and the source.
12.16. Parallel-plate Systems. Cheese and Pillbox Antennas.The
limitations imposed on the antenna design by conditions (59) can be
eliminated by placing the parabolic cylinder between parallel plates as
shown in Fig. 12.26. The feed may then be a waveguide or a horn with
one of its aperture dimensions equal to the distance h between the plates.

(a) (b)
FIG. 12.26.Parallel-plate systems: (b) uillbox al, tenna; (a) cheese antenna.

From the point of view of the rays bet\veen the plates, t!le system is
equivalent to a segment of an extended line source and parabolic cylinder.
The antennas differ from the open system of the preceding section
in that propagation can take place bet~veen the plates in various modes
(cf. Sec. 7.15). The parallel plates support free propagation of a principal
wave-the TE.W-mode-in which the electric vector is normal to the
plates; the velocity of propagation and the ~vavelength is the same as in
free space. TE- and TJ1-modes are also possible, ~vhich are equivalent
to the modes in a rectangular ~vaveguide. We need concern ourselves
only with the lo~vest ZE-mode in which the electric vector is parallel to
the plates and varies in magnitude along the line normal to the plates
according to sin (z/h), where h is the distance between them, The plates
~villsupport propagation in this mode only for free-space wavelengths that
satisfy the condition
x < 2h. (62)
460 PENCIL-BEAM AND SIMPLE FANNED-BEAM ANTENNAS [SW. 12.17

The wavelength of propagation is

(63)

g=
[ -(W

lhe cutoff condition for the next higher mode is

The parallel-plate systems may be classed into two groups: (1) those
with spacing h < k which support free propagation in the 7E.M-mode
and possibly the Z111-mode if h > A/2, (2) those with spacing h > x
which support additional modes. The two groups are labeled pictorially,
the former being called the pillbox antennas, the latter cheese antennas.
The cheese antennas can be designed to meet any length-to-height
ratio desired. If onl.v the 2WM-mode is desired, the feed must be
designed with great ca~e in order to avoid the excitation of other modes.
The difficulty of eliminating other modes is the major objection to the
TEM-cheese antenna. On the other hand, the feed can be designed to
excite various TE- and TM-modes purposefully. Each mode travels with
a characteristic phase velocity, and the superposition of the modes is
used to synthesize various types of phase distributions over the aperture. 1
The limitations imposed on the secondary pattern of a pillbox in the
plane containing the h-dimension, because of the restrictions on the latter,
can be obviated to some extent by flaring the mouth of the pillbox into
a two-dimensional horn. This has a further advantage of reducing the
reflection by the aperture of the wave between the plates. Another
method of controlling the pattern is by means of flaps such as are shown
in the half-beacon antennaz in Fig. 12.27. Half a pillbox is used in this
particular case; it is fed by an H-plane sectoral horn. The h-dimension
is equal to k/3 so that the plates can support only the TElf-mode.
Attention should be called to the curled edge of the flap; the curl follows
an exponential spiral in order to reduce the impedance mismatch arising
from the discontinuity at the edge of the flap. The pattern obtained
in the plane of the h-dimension is shown in Fig. 1228.
12.17. Pillbox Design Problems.There are three major problems to
be considered in the design of the pillbox: (1) the f/d ratio for maximum
gain factor, (2) impedance mismatch, (3) structural problems.

I The cheese antenna received more attention in Britain than in the L-nited States.
Information pertaining to British reports may be obtained from the British Scientific
Commission office in Washington, D.(. or the British central Radio Bureau in ~
I,ondon. Much of the Britishworkis appearingin the ne~vsectionPart IIIa, Radio-
location, of the jo{lrnalof the Instituteof Electrical Fhgilleers.
ZA, IIraunlich, Half Beacon .4ntcnna, RI, Rrport No. 41!), Scpt, (i, 1!)43. f

,,
 ,
4

SEC. 1217] PILLBOX DESIGN PROBLEMS 461

Optimum Shape.-The analysis of the gain factor proceeds along the

same lines as for the parabolic cylinder in Sec. 12.15. The result is
essential y that of Eq. (61) except for multiplicative constants. The

. Ii3==aveguide
FIG. 1227.-Half-beacon antenna.

optimum angular aperture v (cJ. Fig. 12.29) is the value for which the
expression
2
*
cot ~,* [GJ(~)]J~sec ~ d~ (65)
2
()[~ 1
hfis its maximum value. The gain function G,(+) is that of the feed
radiating ,bet~veen parallel plates, not in free space; like that of the cylin-
drical ~vave zone of a line source it is two-dimensional.
) The optimum angular aperture is generally less than 90. The pill-
box is then constructed as shown in Fig. 12.29 with the parallel plates
j 1 T. J. Keary, .4. R. Poole, J. R. Hisser, H. Wolfe, Airborne havigational Radar
Antennas, RL Report No. 808, Mar. 1.5, 1946.
I

462 PENCIL-BEAM AND SIMPLE FANNED-BEAM AN TENAAS [SW. 12.17

3200 3300 3400 3500 0 100 200 300 400

3103 500

3000 600

290 700

280 80

270 900

260 1000

250 1100

2400 120

230 130

220 210 200 1900 1800 1700 160J 150 140

F1~.12.28.Patternof the half-beacon antenna in the h-dimension plane.

F1~.12.29.A pillbox of angular aperture W.

,-+- .
i

SEC. 1217] III,LBOX I) E,VIGN PRO BLEM,? 463

extending a little beyond the focal point. The gain factors realized by
)
pillboxes are considerably higher than that of paraboloidal antennas,
ranging in value up~rard from 0.8.
Impedance lWiwnatch.-The mismat~h produced by the parallel-plate
system arises from the parabolic strip and from reflection at the aperture.
The latter can be reduced, as was noted before, by flaring the mouth of
the pillbox into a two-dimensional horn. The reflection coefficient pro-
; duced by the parabolic strip can be developed along lines similar to
that followed in the case of the paraboloid The essential difference is
that the field between the parallel plates is in the form of a cylindrical
wave rather than a spherical \rave. The reflection coefficient is found
? to be
* ~fkp
rr=: G(*)
+
COS - ds, (66)
7rlJ
/ P 2

where ds is the element of length along the reflecting strip. Let z measure
position along a line parallel to the aperture; on introducing the variable

v=3=2t,an+ (67)
f 2

the refferti{m coefficient becomes

u G(v) _ ,.yfu
,
r.(v) = : - e dv, (68)
To/
1 +g

disregarding all constant-phase terms. If fh >>1, the integral of Eq.

(68) is very closely equal to

(69)

where GO is the peak gain of the feed. This is the two-dimensional

analogue of Eq. (39).
The mismatch can be eliminated by means of a vertex plate as in
the case of the circular paraboloid. The technique of determining the
dimensions of the plate is the same as that described in Sec. 12.9. It
should be noted that the intersection point v. on the spiral gives x. which
is only half the length of the plate. The vertex plate has the same unde-
sirable effects on the secondary pattern as in the paraboloid: reduction in
gain, increase in side-lobe intensity.
Structural Problems. Special attention must be paid to the structure
and assembly of the pillbox. The feed must make good electrical con-
tact with the parallel plates. The contact can be established by solder-
I Details are given by S. Silver, Analysis and Correction of the Impedance Mis-
match Due to a Reflector, RI, Report No, 810, Sept. 25, 1945.
464 PI?NCIL-BLA .}{ ,1ND ~$IMPLIc IANNI$JBEA M AiVTllIVNAS [SEC. 1217

ing; a better technique, however, is to make a choke joint between the

feed and the pillbox. 1 Good electrical contact must also be maintained
between the reflecting strip and the parallel plates. This is a more
important consideration for the Zltlf-rnode, in which the electric-field
vector is normal to the plates, than the TE-mode, in which it is parallel
to the plates, and zero at the surface of the plates. The space between
the reflecting strip and the parallel plate, in the case of poor contact, is
too small to propagate a Z11-mode.
It is recommended that the parabolic strip be cut out of a metal
plate and cut to a sizable thickness so that the plates can be bolted to
the strip without warping the parabolic curve.

l:IG, 1230,- Structural design of a pillbox.

llaintaining a uniform spacing between the plates poses a number

of difficult problems. The spacing problem is not too serious for the
ZE.lf-mode. If the spacing h is well below x/2, there is no significant
mode control problem. In this case, proper reinforcement of the parallel
platesz as sho~~n in Fig. 12.30 together with the spacing support pro-
vided by the feed is sufficient. Additional support is necessary only in
extremely large structures.
The tolerances on the spacing are more restrictive in the case of the
TE-mode. The phase velocity varies with the spacing [Eq. (63)]; a
nonuniform spacing produces phase distortion over the aperture. The
spacing can be maintained by a distribution of metal or dielectric posts.
These scatter the energy, however, producing both a mismatch and dis-
tortion of the field over the aperture. The pins should be kept out of
the high intensity region of the primary pattern of the feed.
1 T. J. IGYWYd cl., .iirbornc XTavigational Itadar .4ntcnnas, RI, Report No.
808, Ilar. 1.5, 1!)46.
2 \J, Sirlmk a]]d 1<:.P(lrrell, (osec .intcnnas \vith a I,ine Source and Shaped
~yliudriral lteflcctor, lU. lteport Xo. 624, Nov. 3, 1944.
 4

CHAPTER 13

SHAPED-BEAM ANTENNAS
BY L. C. VAN ATTA AND T. J. KEARY

The highly directive beams attainable with microwave antennas have

been utilized to achieve large antenna gain, precision direction finding,
and a high degree of resolution of complex targets. The exploration
of a wide angular region with such sharp beams requires an involved
scanning operation in which the scanning time becomes a limiting factor.
This problem is much simplified if the required scanning can be reduced
to only one direction, the coverage of the angular region being completed
by fanning the beam broadly. The characteristics of simple fanned-
beam antennas have been discussed in Chap. 12. For many applica-
tions, however, the characteristic shape of the fanned beam obtained

FIG. 13.1.Beam from ground-based or shiplmrne antenna providing coverage on aircraft.

by simply reducing the corresponding dimension of the aperture is unsat-

isfactory; it may be wasteful of the limited microwave power, or it may
result in a very unequal illumination of targets in different directions.
To overcome these limitations it is necessary to impose on the beam
by special design techniques some shape not characteristic of the normal
diffraction lobe. These beams are referred to as shaped beams, and the
antennas that produce them as shaped-beam antennas.
The purpose of this chapter is to describe several applications for
shaped beams, to discuss requirements imposed on the beam by these
applications, and to present a number of design techniques for producing
shaped-beam antennas.
13.1. Shaped-beam Applications and Requirements. There are a
number of radar applications for microwave systems that impose more
or less severe beam-shaping requirements upon the antenna. The
applications and requirements ~vill be considered here; the means for
realizing the shaped beams \vill be deferred to later sections.
S/(r~ace .-ln(enna jor Air Search.For use in search for aircraft, an
antenna on the ground or on a ship is required to produce a beam sharp
in azimuth but shaped in elevation; the azimuth coverage is obtained
465
by scanning. The elevation shape of the beam must provide coverage
on aircraft up to a certain altitude and angle of elevation and out to the
maximum range of the system. This is to be accomplished without
wasteful use of available pofver. Figure 13.1 indicates the general shape
of the co~erage required in the ~ertical plane. Tbe antenna beam need
not meet the coverage requirement very accurately, since conservation
of po\ver and a relatively constant signal on a plane at a fixed altitude
are the only objectives.
In order to maintain a fixed minimum of illumination on the aircraft
at various points along the uppm contour of the coverage diagram, it is
necessary that the amplitude of the antenna pattern be proportional to

/ /
l:l<;, 132,-Beam froxu antenna of airlmrne radar for surface search.

the (Iistanrc r from the antenna to the aircraft on that contoul. In

other ]vords, the coverage contour of Fig. 13,1 can be taken to be the
arnplitllde pattern of the antenna (c~. Sec. 1.2). Since r = h csc 0, the
amplitude pattern must be proportional to csc O, or the po}}-er pattern
must be proportional to Cscz 0. The proportionality must hold over the
. .
region from a mmlmum angle arc sln h/rmax, to the maximum elevation
angle for which coverage is required.
.4ir+or-neAntenna jor Surjacc Smrch.-.&n airborne antenna is required
to pro(lucc a beam sharp in azimuth but, so shaped in elevation as to
provide ~miform illumination on the ground; azimuthal coverage again
is achieve(i by scanning. Figure 13.2 illustrates the vertical coverage
requirement; this ]vas showm in the previous paragraph to be identical
~vith the vertical amplitude pattern. Both this and the previous pattern
assume isotropic scattering b,y target objects. Deviations from this
assumption for various tar~et objects w-ill be discussrd in the next section.
lVhen an airborne antenna is used primarily for surface search over
s?a against such point targets as ships and buovs, the purpose of beam
shaping is to conser}e power, to maintain a relatively constant signal as
the target is approached, and to avoid overloading the indicator scope
with sea return. h-one of these objectives impose exacting require-
ments on the beam shape. IIowever, for successful surface search o~rr
land it is necessary to illuminate the ~round \cry uniformly in ordrr to
obtain [ solid painting on the indicator scope and a fully intelligible
picture. he results from the opcratous \ie\rp(,int of Satisfactt)ry an[l
umatistat,tor.v rlevalion patterns ar(: des(,rilxvl in SC(I, 14 I aIILi its ;it,(t)m-
pauying figllres.
 Sm. 131] SHAPED-BEAM APPLICA 7TIONS AND RAQ (JIREbfE,VTS 4(;7

5hipborne Antenna jor Surjace Searth.-A shipborne antenna for use

in surface search must scan in azimuth with a sharp azimuth pattern.
To accommodate roll and pitch the beam of an unstabilized antenna must
be broad in elevation. This broadening \vill be more conservative of
power and will provide a mm-e constant illumination of the target if it is
accomplished with a shaped beam (Fig. 13,3) rather than a simple fanned

Shaped-
beam
antenna ------ --= ==. ----- __

1
I Ship

,ntenn
! 1TG. 133. -Sertor shaped beam for surface search by shiphorne antenna.

Horizontal
- __

(CL)

I
Ctennawx
I:lG. 1:14.-Beavertail
(b)
beam for t,eight-finding a,,te,, na: (a) ele~ation pattern; (ii) shaped
azilnuth patter,l.

beam. The ideal beam shape for this purpose n-ould be given by 1 = 1~,.
for angles in the region + 191to 61 and 1 = O for angles outside that
region. This sector shape can be approximated more closely as the
vertical aperture of the antenna is increased, but a close approximation is
not justified.
Surface Antenna for Height Finding.A ground or ship antenna
designed for height finding must have a sharp elevation beam for obtain-
ing precise elevation information and a rapid elevation scan. Provision
must also be made for scanning the antenna slo~vly in azimuth or for
turning the antenna to an assig~ed azimuth. Tbe beam must be rela-
tively broad in azimuth in order that the target ~vill be held in the beam
468 SHAPED-BEAM ANTENNAS [Sm. 132

long enough to obtain height information. If the beam is assumed to

bestationary inazirnuth, anairplane flying across the beam will beillu-
minated for a period proportional to its distance away. To increase
the time of illumination on near-by crossing targets, a low-intensity
broadening of the azimuth beam is required. If a fixed minimum of
illumination is to be achieved at a given linear distance on both sides
of the center line of the azimuth beam, the amplitude pattern must have
the so-called double csc o or beavertail shape illustrated in Fig. 13.4,
13.2. Effect of a Directional Target Response.In the previous sec-
tion it was assumed that the target response is isotropic. The effect of a
directional target response is to alter the beam shape required of the
antenna from that predicted by simple (inverse-square considerations.
The power received by a radar system from a target in a given direction
is proportional to the radar cross section of the target; t,he radar
cross sectionl may be defined as the interception cross section multiplied
by the scattering gain for that direction. Since the radar cross section
of some targets varies widely with direction, this effect must be taken into
consideration in establishing the required beam shape for the antenna.
The power transmitted in a communication from a shaped-beam
system
antenna to a receiving antenna in a given direction is proportional to the
product of the gains of the two antennas along the line joining them.
A directional receiving antenna will modify, therefore, the beam shape
required of the transmitting antenna.
Let us consider in greater detail the case of a radar antenna located
a perpendicular distance h from a plane (Fig. 13.2) and required to
obtain equal signals from identical targets located arbitrarily in that
plane, Let us assume that the antenna is to scan with a sharp beam in
azimuth, as in the case of the first two shaped-beam applications described
in the previous section, Then the specifications for the vertical polar
diagram may be derived if quantities are defined as follows:
P = power emitted by the antenna
G(o) = power gain of the antenna at depression angle 6
r = slant range to the target
U(O) = interception cross section of the target for a plane wave from
the direction of the antennaz

1Also known as the back-scattering coefficient,

Zfhe definition of the scattering cross section is being set up here in more detailed
form than was done in Sec. 12. It is based on the physical picture that the target
presents an interception area such that it rcrnoves from the plane wave all the energy
incident thereon and redistributes it in space in a scattering pattern. Both the
interception area and the scattering pattern vary \vith the aspect presented by the
target to the incident wave. The hack-scattering cross section is the product of the
interception area and the gain of the scattering pattern in the direction of the trans-
mitter. It is the cross section of the equivalent sphere that would produce the same
 SEC. 13,2] EFFECT OF A DIRECTIONAL TARGET RESPONSE 469

7(9) = power gain of the scattering object in the direction of the

antenna
p= power received by the antenna from the target

1
, Nu, 13.5. Effective ground target area and its interception cross sertion for airborne
pulsed radar.

The fraction of the transmitted po~ver received by the antenna is given

by
p _ G(d) u(61)T(0) G(0)k2
P 47r2 ~T 41#
= [G(o)l~2
0(0)7(0). (1)
(47r) %-

+ To impose the condition that, equal signals be received from identical

targets in the plane, let us note that r = h csc 0 and write p/P = Cz,
a constant. Then
CSC2o
G(o) = C(47r)$fi~ (2)
k [a(e) ~(e)p

If [u(0)~(o)] is independent of angle, we obtain the earlier result that

G(6) for the shaped-beam antenna is proportional to CSC20. This should
be recognized as only a crude approximation in the majority of cases of
actual interest.
In the particular case of reflections from ground targets, serious con-
sideration has been given to the angular dependence of the quantit:~
[u(8)~(tI)]. Theeffective area of thetarget ontheground depends upon
beamwidth (6))4), range (r-), pulse length (7), and depression angle (o).
By reference to Fig. 13.5, it is evident that the effective target area A.,,
on the ground as determined by the pulse length is related to depression
angle by the proportionality

Aeff E r-see 8 cc csc Osec O

P
return signal at the transmitter as does the target; thus, the product u(@)T(@) used
here isequal tothescattering cross sectionuof Sec. 12,
lR. 1;, ~lapp, ATheoreticala ndllxperimentalS tudyof Radar Ground Return,
RL Report X-o. 1024, Apr. 10, 1946.
470 SHAPED-BEAM AN1ENNA,Y [SEC. 13.2

and that the projection of this area in the direction of incident radiation is
u(O) = A.if sin O m sec 0. (3)

The angular distribution 7(0) of the radiation scattered by the area u(0)
will depend upon the nature of the target or terrain. .4 mathematical
expression derived for -y(fI) will depend upon the simplified target model
assumed. Best agreement with experience is obtained by assuming a
flat plane made up of closely spaced components which scatter isotrop-
ically. The radiation will thel. be equal in all directions for any given
condition of illumination; i.e.,
-y(f)) = 1. (4)

Combining Eqs. (3) and (4) gives the angular dependence

[u(L9)y(@)]
c sec /?. (5)

Introducing this dependence into Eq. 2 gives the proportion

G(19) a CSC2 e se (6)

for the assumed ground target model.

An antenna with a vertical pattern shaped according to Eq. (6) would
produce, within the limits of the assumptions, a range trace of uniform
brightness on an indicator scope for any given azimuth setting of the
antenna. A succession of range traces from an antenna scanning in
azimuth would still be displayed Ivith uniform brightness on a B-scope
presentation \vhich makes a rectangular plot of range vs. azimuth On
a plan position indicator (PP1), however, the range traces are presented
radially and the azimuth angle, as polar angle. The spacing bet~veen
range traces therefore varies in direct proportion to the range with the
result that the scope is brightened to~vard the center. This effect can
be compensated if the vertical pattern of the antenna is used to modify
the received pm~er by a factor of I/r; i.e., the gain function G(6) of the
antenna should be modified by a factor r = csc O. For the case of PPI
presentation then, Eq, (6) becomes

The several ideal cur~es for G(d) discussed above are presented
in Fig. 13.6:

Curve A. CSC2o dependence for a uniform range trace with isolated

isotropic targets.
Curve B. csc~ OVcos d de.xndence for a uniform range trace with
closely packed isotropic targets,
(urve C. CSC28 #cot O dependence for uniform PPI presentation
with closely packed isotropic targets.
~

SEC. 133] SURVEY OF BEAM-SHAPING TECHNIQUES 471

Curve D. CSC2 e . cm 6 dependence which approximates the experi-

mental optimum pattern shape obtained from considerable flight
experience with a number of antenna designs at wavelengths
between 10 and 1.0 cm.
These curves are all plotted for a minimum depression angle of 10; this
corresponds to the case of an airborne radar system with a maximum
100
100 0
80
60 2

40 4

6
20
8

10 10
78
z 2e
~6 ~
n
~-4 4$
% :
? 6~
92 z
ii 8$
$
1.0 20
0.8
0,6 2

0.4 4

6
0.2
8
01 30
03 ,00 200
30 40 50 60 70 800 900
FIG. 13.6.Ideal curves for dependence of vertical pattern on depression angle.

range about six times the altitude of the aircraft. Several curves of
CSC2 0. cos o for different values of minimum depression angle are pre-
sented in Fig. 13.7.
13.3. Survey of Beam-shaping Techniques.In the preceding sec-
tions we considered various applications for shaped-beam antennas and
the requirements that they impose on the beam shape. In this section
we ~vill discuss the physical principles involved in various beam-shaping
techniques and survey a number been used
of antenna designs that have
? for producing beams of various shapes.
Physical Principles.In Chap. 12, the characteristics of pencil beams
and simple fanned beams were described. Such beams were showm to
have a common shape, characteristic of the main lobe in the diffraction
 n

472 SHA1ED-13EA3i A,VTENNAS [SEC. 13.3

pattern of a constant-phase aperture. This is true independent of the

shape or size of the aperturefor apertures larger than about 2iand
independent of the intensity of illumination across the aperture. The
etfect of these variables is to change only the scale factor for the angular
coordinate of the pattern. The only means available for altering this
characteristic shape is to vary the phase of the illumination across the
aperture.

2j050 10 15 200
100 0
80
60 2

40 4

6
20
8

10 10
~8
: 2e~
?46
~-4 4?
> :
? b=
2* ~
5 82
~
1.0 20
0.8
0.6 2

0.4 4

6
0.2
8

0.1 30
00 loo 200 30. 403 50 600 700 80 900
11~, 1:<7.-Af a,,,ilyofct,rvcs, csc~ r3cos#, for (l~ffcrellt values of Illillilllulll dcprcssiol~ :lt@,,.

The elementary principles of beam shaping can be understood in

terms of gromctriral optics, From Huvgrns principle of propagation
normal to the phase front, it is evident that a cur~ed phase front ~vill
produce a more clispcrsed beam than a plane phase front. For a fixed
aperture size, beam shaping ran be accomplished only at the expense of
antenna gain, since the curvecl portion of the phase front subtracts from
the total apert~lrc a\railal>l(]ft)rcontrib(lti oninthedirecti(~n of maximum
gain. Ther:~cliati~)n intensity f\)ragiven direction inashapcd beam ~~-ill
depend upon the radil~s of curvature of the plmse front normal to the
given direction and upon the intensity of illuminatiurr in that rrgion; the
exact relations involved \vill lx: dvriwxl in Sfw. 12,G. hy shaped-beam

,,
 SEC. 13.3] S UR Vli Y OF BEA M-SHAPING lECHNIQUES 473

antenna can be considered to be a device for obtaining the proper phase

and intensity of illumination across an aperture to realize a specified
beam shape. 1
In the language of ray optics, a constant phase front across an aper-
ture produces a collimated beam of rays from that aperture. This col-
limated beam is obtained usually by focusing the diverging rays from
an antenna feed either with a parabolic reflector or with a lens. The
process of forming a shaped beam can then be visualized as a defocusing

4
llG. 13.S.Formation of a shaped beam by means of a feed array in a paraboloid reflector.

process; the rays emerging from the aperture will not all be collimated
but will be distributed through a range of angles with a variable density
dependent IIpon the pattern required. Defocusing in one plane can be
accomplished either by extending the point source into a line source in
that plane or by modifying the reflector or lens in that plane.
The extension of a point source into a line source can be accomplished
by disposing an array of dipoles or horn feeds in a line in or near the focal
I
plane and by exciting them in the proper intensity and phase. The for-
mati(m of a shaped beam by such a feed array in a paraboloid reflector
4
is illllst,rated in Fig, 13.8. Each of the elements in the array can be
visu:dizcd as a point source that forms its own sharp beam in the parab-
oloid. The intensity of this beam will depend upon the intensity of
excitation of its feed; the angular displacement of the beam from the
axis ~vill be proportional to the angular displacement of the feed point
about the vertex on the opposite side of the axis. The overlapping
beams formed by an array of point sources will synthesize by amplitude
addition into a shaped beam, as illustrated in Fig. 13.8. The resulting
beam can be quite smooth if the component beams are properly spaced
and phased.
1R. C, Spencer, Synthesis of Microwave Diffraction Patterns with Application to
Ckcz 8 Patterns, RL Report No. 54-24, June 23, 1943.
474 SHAPED-BEAM AN ThNNAS [SEC. 133

It has been assumed above that a feed moved off axis from thq focal
point will form in a paraboloid a sharp beam on the other side of the
axis. This is true for small displacements from the axis; for large dis-
placements of a point source, its individual beam is broadened in the
plane of displacement, which is not serious, since the beam is being broad-
ened intentionally in that plane,
but it is also broadened in the per-
pendicular plane, which is serious,
since it reduces resolution and gain
(Sec. 13.4). The extended feed i
method of beam shaping is there-
fore not recommended as a means t
for forming wide-angle patterns.
It is recommended for forming
4 I
shaped beams confined to small
4
(a) angles, since it accomplishes the
beam shaping by increasing the size
of the small feed rather than by in-
creasingfor equal gainsthe size
+ of the relatively large reflector or
* lens.
.i sharp beam formed by a point
. //// source and paraboloid reflector or
by a line source and parabolic cyl-
inder can be dispersed in a con-
trolled way by modifying the shape
of the reflector. The process can
be thought of as one of dispersing
collimated rays into new directions
dictated by the shaped-beam pat-
/ tern, or it can be visualized as one
(b)
FIG, 13.9. lteflector modifications for of controlling the phase and intens-
producing an asyrnmetriral flared beam: ity of illumination across the aper-
(a) by shaping the reflector on the op-
posite side from the flare; (b) !.Y shaping ture. In the latter case, the next
the reflector on the adjacent side to the step to the far-field pattern can be
flare.
made by use of Huygens principle
in some cases and by a ~ouricr transform process in other cases (cf. Chap.
6). Figure 13.9 shows two reflector modifications for obtaining an asym-
metrical flared beam. It has been shown that-one aperture illumination
which gives an asymmetrical beam consists of a sharply peaked ampli-
..
tude distribution v-ith a sudden 180 phase reversal in the region of maxi-
mum amplitude. One means of realizing this is to use a point source feed
\vith two paraboloid reelectors of focal lengths jl and .fz = jl + X/+ for
the top and bottom halves of the aperture respectively. Other methods
 of beam shaping will be described in connection with specific beam-shap-
ing problems.
Symmetrical Shaped Beams.i% sector shaped beam with sharp sides
and a square end, approximating the requirement illustrated in Fig. 13.3,

(a)
I

Reflector.

Two-horn Overlapping
.?, beams
feed
(h) (c)
13. lo.-Two-elel,, ent :irr:iy and paral><,loid for l,mduring sector shapwi IW:L,II (:1)
antenna; (b) o\wrlapl~irlg bea IIu,; (.) aperture ,Hu,,],r)ati<,ll.
476 SHAPED-BEA .11 ANTENNAS [SEC. 133

can be produced by means of either an extended feed or a shaped reflector.

If an array of radiating elements in the focal plane are equally excited,
so spaced that their individual patterns cross over at the half-amplitude

/
point and so phased that the patterns add in amplitude, the result will
be a beam with sharp sidesdetermined by the size of the aperture
and a square end. Figure 13.10a shows two elements combined in this
way. Thedesign procedure can reinterpreted asabove andillllstrated
in Fig. 1310b, or the following. The two feeds excited equally and in
phase will form a symmetrical interference pattern. If each feed is so
dimensioned that its individual pattern properly illuminates the reflector,

&
Path lengths
~ P2=P, + $

A
T
Feed
Focal lengths
,.(
\
1, ~=~++ F

\
k
Composite
ref Iector
(a) (b)
FIG. 13. il.Cut paraboloid method for obtaining sector shaped beam: (a) antenna: (b)
aperture illumination.

and if the two feeds are correctly spaced, their interference pattern will
result in the (sin u)/u aperture illumination shown in Fig. 13, 10c. The
Fourier transform of this illumination curve will be approximately the
sector shaped pattern required (Chap. 6). The antenna shown in Fig.
13.10 produces its sector shaped beam in the horizontal plane. The
reflector is cut with a slight asymmetry to bring the null in the illumina-
tion pattern opposite the feed for improved impedance performance.
The sector shaped beam can also be obtained with a point-source and
modified paraboloid reflector. Let the aperture be divided into three
parts along the lines corresponding to the nulls of Fig. 13. 10c. Let the
two outer portions of the aperture be illuminated with segments of a
paraboloid having a focal length 1/4 longer than that of the central parab-
oloid. This situation is illustrated in Fig. 131 la. Then the phase of
the illumination over the outer portions of the aperture will be delayed
by A/2. If the antenna feed provides a normal primary pattern, the
aperture illumination shown in Fig. 13.1 lb will then be obtained. This
 477

will be a crude approximati[)n t o the (sin II) 1( aperture il]um.ination

required for a sector shaped beum.
The beavertail be:~nlill~lstlaterl in Fig 134 is not obtained conven-
iently by means of an extcmdrd feed; the anglr of co~-cr-age in actual
applications is too hwge, and the t:~i)~rin t~,:o directions from the center
complicates the feed, It is obtained very easily, horevcrj ~rith a reffector
modification. The simplest arrangement is a narro~r ~-ertiral strip down
the middle of the reflector set out from the surface of the main reflector
by a fraction of a ~ravrlength. The factors affecting the ]ridth and
offset of the strip can be appreciated by reference to Fig. 13.12 which
illustrates the design and the mechanism of beam shaping for an actual 1
case. The width of the strip affects the rotal pol~cr that it intercepts
and the directirity of its pattern. Its offset from the main ref!ec~or
establishes the phase relationship for amplitude additi(,n \vhich is impor-
tant in the region where the two am.plitudes arc of the same order of
magnitude. In the actual case at A = 10 cm, a refiector with a 20- by
5-ft aperture and a 5-ft focal length was fitted with a strip r~lnning the
long way of the reflector. The optimum width of the strip pr)ved to
be 8 in., and the offset, + in. It is evident that the presence of ~ strip
of this width ~vill introduce interfmmce side lobes in the pattern of the
remainder of the reflector, ~vhich ~vill impair the quality of the final
beavertail pattern. .4n improved pattern would be obtained if the
flaring of the beam were accomplished by modification at the t}vo edges
of the reflector rather than at the center, in which case the intcrfercmce
lobes would be less prominent.
.4 symmetrical Shaped Beams. Extended feed and modified reflector
designs ha~e both been used surccssfully to ol]ttiin asy-mmetrical shaped
beams. Extended feed designs have been used in genend for ground and
ship antennas for ~1-hich the recluired elevation coveruge was limited
usually to small anglrs and for ]rhich the reflectors 1}.ere too large for
convenient modification. Extended freds arc readily adopted ako to
the use of multiple transmitters IIhen the prescrihcd coverage requires
high power. Modified refiector designs ha~-c been used almost excl~l-
sively in airborne antenntis which are required to provide ~~ide-angle ele-
.
vation coverage and to possess smooth and stable pattern characteristics.
Both dipoles and horns h~ve bern used in linear-array feeds for
shaped-beam antennas; in some cases used in the same
both have been
array. The choice brtl~een a dipole and a horn as the rtidititing element
in a given situation dcpcnds upon the po~rer to bc handled, the required
impedan[,e characteristics, and cf)n~-eniencc in c[)nstrll(tion, The design
and performance? of thrrc extended feed untrnuas ]\-ill be described in
.? (jrdcr of increflsing cornplrxity. .i cut paruh(jloi(i reflector ]vith a tlvo-

I [;, s. 1:10, 1l)t, l+t:ivfvt:lil .\nt{t]t):L, 1{1, licport So. 1027, :Ipr, 0, 1!)46.
 SHAPE II-BEA ,il AN TEKNA,V [sm. 133

c.

(a)
FIG. 13.12.Cent r:+]stri]] in reflrrtor fc>r]Jr,,du<i!,g lwa~. crtail IXX*II1:(<L) antenna with-

dipole array fwd I for olxratifm a+ k = 10 cm is illustrutcd in Fig. 13.13.

The aperture dimensions are S by -1 ft, and thr focal length, 27.5 in.
The elevation pattern is sho\vn in l~ig. 131(;a,. his pattern and the
impedance match (ISIVR < 1.2) \verc maintained satisfactorily over a
2 per cent bandlvidth. .-l cut par~boloid reflrctor with the long dimen-
sion horizontal and with a four-horn array feeclz for X = 10 cm is shown
in Fig. 131}. The aperture dimensions are 5 by 14 ft, and the focal
length, 60 in. The feed dmign will be discussed in the next section.
The elevation pattern is shown in Fig. 1316/). In impedance match, the
feed shoved a }-SJTR < 1.12 for a (i per cent band \vhen tested in free
space and a TSJVR < 1.25 for a 3 per cent band ~1-hentested in the

1 C. F. Portcrficl{l and I,. .J CIIU) .i Si]llplific(l S(,;trrh .tntcnna, RI, Report No. 4
486, Jln. 1, 1!)4.5.
2 },. .J J1-[,st, .1 l;{Jllr-IIOrIl F~d t{) ~iiv(, C*(2 @ ll)t[,nl);~ Pnttern s, 1/1, l{cport
S0. 8!)6, llar, 15, 1!)46.
4

r
SEC. 133] SUR VE I (IF HEA itf-SHAPIArG 7ECH,VIQ UES 479

T
I

Reflector

J.
~~11
8strip

,l-F
+ yl
4

(b)

Contribution
from strip
/

~
Contribution from contours
paraboloid
(c)
out strip; (h) central her, zont:d w{, tion .I>owing strip; (c) nlerhan, smof beam shaping.

reflector. ii lfirge and com~jli[atc(l :~utenna] for k = 10 cm is sho\vn in

Fig, 131.5. The rcfiector is a cut paraboloid 10 by 25 ft \\-ith a 78-in.
focal length. Of the elm-en in the feed army, one clement is
rlcmcnts
fed from a first transmitter, tlvo elements from a second transmitter,
and eight elements from a third transmitter. Horn radiators are used
with the first tw-o transmitters Ix,(,ause of the concentration of po~ver,
and dipoles with the third transmitter. Thr thrcwlobe pattern is sho\vn
in Fig. 13.16c. This elcyation corcrayc rcprcsrnts the limit practicable
with this antenna, since the azimuth bcam~vidt]l of a point fm>d 30 in.
off axis (or 21 referred to the vrrtcx) is increased by abo~lt 70 per
cent. The impcdancc match presented to all three transmitters ~\as
VSWR < 1.12 o~-er at least a 4 pm cent band,
A variety of reflector shapes hare bccm used for beam shaping in
airborne antenna drsigns. Of these the most conuntln and sllcccssful
was the so-callccl antenna. I The shape of this mficctor is
burr-clrf,jlccior
obtained by replacing the top half of a paraboloid reflector \vith a figure
of revolution produced by rotating the generating parabola of the parab-

(a)
 SEC. 133] SURVEY OF BIIA.W-SHAP1.VG TIICH.VIQUES 481

further in Sec. 13.8. Its elevation pattern, shown in Fig. 13.20a, is

~ best suited to high-altitude use. Elevation patterns suitable for low
altitude can be obtained with several reflector shapes. The shovel
r-ejector-l belongs to the family defined in Fig. 13.9b. It is obtained by

(a)
8

(b)
~r~. 13.14.Cut paraboloid reflector with four-horn array feed: (a) antenna; (b) central
vertical section.

] J. H. Gardner, Low Altitude Navigation Antennas, RI, Report No. 615,

Oct. 3, 1944.
482 SHAPED-BEAM A.YTE.V.VAS [SEC, 133

60I

(c)
FIG. 13.15.Cut paraboloid reflector with 11-element array feed: (a) antenna; (~) feed;
(c) central vertical section.
 SEC. 133] S1.RiEl O]: IIE,I .JI-Sll,llI.Y-G lEC)[.~-IQCE.T 483

replacing the lower third or so of a paraboloid reflector with a parabolic

cylinder. The shovel-reflector antenna is illustrated in Fig. 13.18; its
elevation pattern is gi~en in Fig. 13,20b. The general shape and smooth-
ness of the pattern depend upon the point of attachment of the she\-el,
10. / x
8 /
2 / \

2 / \
1.: /
I ,

: \>

4 2
0.1 \
8 I
6
-6. -40 _2. f). 2. 4 6 8 10 12 14 16 18 20 22 24 26 28 300
(a)
10. ----- Y
/ \
: /
4 -

2 / r---
\
Lo J \
8
lb,
\ t \
$

2 I
v
0. \,
$ )
1
+0 -40 -20 00 z 4 6 8 11)9 12 14 ]6 18 20 22 24 26 28 300

(b)
10 A
8 /
6 L 1/ \
4
2 / \
\ \
1.0 / / \
8 I / l\
/ I / \ ._ ~ \
/
2 / J \
/ L - \
2 \
0. \
~: \
-6 4 2 O 2 4 6 8 10 12 14 16 18 .200220 24o 260 28o 300
Elevation angle
4
(c)
l1~. 13. 16.Elevation patterns ol]tained with cut paraboloid reflectors and linear-
array feeds: (u) two-element array of l~]g, 13. 13; (~) four-element array of Ihg. 13.14; (c)
Clcven+ler,lel,t, tl~rcc-trallbrIlitter array of Fig. 13,15.
484 SH.4PED-BEAM A.l;TE.\-.l-AS [sEC. 133

its displacement normal to the surface of attachment, and its tilt ~vith
respect to the tangential direction. .4 low-altitude beam can be obtained ,
also by inserting a narrow horizontal strip just above the center of a

(a)
Orw!al parabola

Sam! surface.

Focal po,nt

Fe
(b)
UT{}. 13.17 .Barm-reflectur ar,t,en,,a for higll.:dtitude LC:LJIIsha~~ing: (a) photograph;
(b) drawing.

paraboloid reflector \rith the proper \\-idth, offset, and tilt. 1 A sb-ip-
r-ejtectorantenna is illustrated in Fig. 13.19, and its elevation pattern is
given in Fig. 1320c. The basic limitation with this strip-reflector design,
as pointed out earlier in connection with the beavertail-shaped beam,
is that the strip divides the aperture into two parts between which inter-

1 C. G. (}utler, ISotes on the Design of Asymmetrical (Closecant) Antennas,

BTL Report illhf-43-160-192, Nov. 12, 1943; J. H. Gardner, Low Altitude Aaviga-
tion Antennas, RL Report N,,. 615, Oct. 3, 1944.
r
Snc. 133] SlrRT7EJ OF BEAL-SHAPING TECHNIQUES 485

ference occurs. Even with the relatively narrow strips required for low-
altitude beams, the interference lobe that appears ontheopposite side of
the peak from the flared portion is only 10 db down from the peak gain.
In many aircraft installations this upward-directed interference lobe would

-.
---
_--
\kb
\
Paraboloid
reflector

Shovel
attachment

\
\
\

(a) (b)
klG, 1~18.--S!l<lV[>, Cfleit<l,1, :L!ltt,!lll:l f<,r lo!v-:dt, fu<lr I>c:l,!l: (11) :Lllte!lrl:l; (b) wtltml

/
/
strip
attachment
p

, Paraboloid
reflector
\

\ .

(a) (b)
l:,<;, 1:1.19.Strill-rcflc, t{]r antrunz for Ioxv-:dtitu<le \)calll: (u) :Lllt CllllX; (l,) (e!(tr:(l
vcrtl{,.al .cct,t,]l of Ic,flcrtflr.

be reflected from the un~lrr side of tlm fusrlagc and \\-ingsand produce
tin interference ripple in the pattern on the grmlnd. The best sirnplr
design for a low-altit~ldc Shape{l-hr:lm antenna has been one that lises
a mlt-down barrel reflector. This ant rnnal romists of a Xl-in. diamct er
by 10,6-in, focal length parahuloi[l \\-it
h a barrrl insert cut cio~vn in
I ,J, H. Gardnrr, I,OTVAltitlldc (scZ .9 Alltcnl]a, 1{1, llcport h-o. 1073, Fch. 21,
1946.
 486 ASHAPIID-BEA <if ANTENNAS [SEC. 133

the vertical to 19 in. (7+ in. above the axis, 1l+ in. below) in order to
eliminate that part of the barrel which contributes to the steep angle
portion of the pattern. This reflector is then fed with a horn so directed
0-

8 \

12
/ q
16

20 h .

24 A

28
\

II
32

36 ~ U

40 1 L 1 1 1
-30 -20 -10 o 10 200 30 W
(b)
o o

4
5
\ \
8

% 10
g 12
1P L 7.
5 s
~ 16 / : 15
1 ~
? .*
:20 - \
2
g 20
~ .-c
24 -
$
.=
~
% : 25
E 28

32 - -.. -_ --- . 30

36 - -- --- - - -
35

40
40 -30 -20 -10 0 10L 20 30 40 -30-20-10 o 100 20 30 400 500
l(l) (d
~IG. 13.20.Elevation patterns obtained with airlmrne shaped-reflector antennas: (a)
barrel-reflector antenna of Fig. 13.17; (h) shovel-reflector antenna of Fig. 13. 18; (c) strip-
reflector antenna of Fig. 13. 19; (d) cut-down barrel-reflector antenna,

as to obtain proper illumination. The antenna can be visualized by

reference totheuncut antennasho\vn in Fig. 13.17. Its vertical pattern
is shown in Fig. 13.20d.
 A number of shaped-beam antennas obtained by reflector modifica-
tions ~vrre described in the previous paragraph. These ~vere truly
reflector modifications in the sense that the design started \rith an existing
paraboloid reflector andprocecded on the basis of obtaining the required
beam shape with a minimum of remodeling. Time-consuming cut-and-
try processes ~vere involved, and the ultimate result ~vas frequently less
than satisfactory. A satisfactory design is obtainable more rapidly
and reliably if the reflector is visualized as a device for transforming the
primary pattern of the feed into the required secondary pattern. In this
sense the reflector shape is dictated by specifying the t~vo patterns; the
problem is one for computation rather than cut-and-try. The computa-
tion of shaped cylindrical reflectors is a straightforward problem and
has received experimental confirmation in a number of antenna designs.
These results ~vill be described in detail in SCc. 13.6. The line source
used with a cylindrical reflector could be used equally Ivell ~vith a shaped
cylindrical lensl to obtain an antenna ~vith certain advantages over the
reflector antenna. The calculation of the complete shape of a double
ctmutwe reflector for beam shaping is more invol~ed and subject to
further study, but experimental confirmation of present design procedures
has been obtained in a limited number of cases.
13.4. Design of Extended Feeds.In the prexious section, several
shapm-beam antennas utilizing extcndrd feeds \~cre described as to
over-m]l design and performance, In actual fact the principal r-f design
problem is c{mumtratml in the rxtrndmi feed itself. It has been pointed
o~lt that hotll dipoles and horns have been IIsed as the radiating elements
in linear-tirray extrndcd feeds and that, th~ choice bet\vecn them depends
llpon lJ(~\vcr-tl~~n(lling, impedance, and mechanical considerations. These
(Illestions and others involved in the design of these arrays will be con-
sidered in t}lis section. lVe \~ill consider first, holvever, some optical
focusing problems common to all extended feeds used in paraboloids
(cf. Scc. 67).
Optical Focusinq Propcrtirs oj Paraboloid Rcjlcctors.-f,et us begin with
a single radiating elrmcnt at the focal point of a circular (uncut) parab-
oloid reflector. The resulting diffraction pattern ~vill depend upon the
directivity of the feed and upon the shape (focal length to diameter ratio,
I,/D) of the paraboloid. As the feed is moved off axis by a rotation
about the \-ertex, the beam \vill move off axis on the side opposite the
feed and in direct proportion to the feed displacement. This propor-
tionality factor wollld be unity for a flat plate according to Snells la~v;
it is slightly lCSSthan unity for paraboloids in the usrful rmgc of shapes,
.1. s. Dllllklr, ~~ct:{l Il:it{,IAIISfor (sc2 At)t(,tl]l:i, 1{1, l{vport so. 1070, F(l).
15, 1!)46.
~ s S,lv,. r :,,, (I (:. ,<. 1:10, l:~r:il~{,l{,i,l .\)]t(,tl]l:l Cl~:ir:[,tf,risti{s :1s :i Ftlnctiou (,(
1~(,(111llt , 1{1, l{eport so. 47!), 1(,1). Ifi, 1!)44.
488 SHAPED-BEAM ANTENNAS [SEC. 134

as shown in Fig. 13.21. As the beam moves off axis, it deteriorates, at

first slowly and then more rapidly as the angle increases. The gain
decreases; the beamwidths increase; and a series of side lobes (the so-called
coma lobes) appear on the axis side
1.0
of the displaced beam. These
; effects can be described uniquely
&
0.9 in terms of the angular displace-
~
.=
.
.- ment expressed in bearnwidths,
z
u
~ 0.8
but only if the reflector shape and
g feed directivity are held constant.
The variation of gain with feed tilt
07
0 0.2 04 0.6 0.8 1.0 for reflectors of different shapes
% with a relatively directive feed is
lIQ. 1321.-Beam deviation propor- plotted in Fig. 13.22. In addition
tionality factor as a function of reflector
shaps. to these pattern changes, an an-
tenna initially matched in imped-
ance \vith the feed on axis will lmdergo a series of impedance changes with
feed tilt. These effects are illustrated by the measured data plotted in
Fig. 13.23, ~vhich are susceptible of quite accurate theoretical verification.
The curves have been rotated into separate quadrants of the Smith Chart
ior clarity in presentation,
100
90

80

-z 70
8
ii 60
n
.~ 50

,$ 40
%
6
K 30

20

10

0
12345678 9 10 11
Feed tilt in beamwidths
1,IG. 1322. -lleper1cier~ce of antenna gain On feed tilt and parxi>nloid shape. FID IS the
focal Ie,lgth to dial,lctcr ratio.

l~hen the paraboloid is cut to reduce the apertllre in (me dimension

or the other, its ofi-axis modified considerably,
focusing properties are
These effects can be undmst ood qllalit atively l~y refcrcu(.c t,) Fig. 1324.
The circle represents the aperture of a paraboloid rrflector, 1~hen the
feed is mo\,ed off axis in the vertiral plane to some position in the rec-
SEC, 134] DESIGN OF RX TENDED FEEDS 489

tangle shown at the center of the circle, rays reflected from various por-
tions of the paraboloid Ivill deviate from parallelism with the ray reflected
from the vertex. Ihe equal deviation contours plotted in the figure
are purely schematic, since actual contours ~vould depend upon reflector
shape and feed displacement. However, they serve to illustrate the
fact that the outer portions of the reflector lying between the principal
planes are responsible for the most serious deviations. From Fig. 13.24a

l~rc. 13.23 .1mpedanre changes at A = 10 rrn in 30-in, paral>ololil refktors of different

shapes (F/Z)) as the feed is t]lted off axis.

it is evident that cutting do~vn the aperture along lines BB \villeliminate

regions of high deviation and increase the allo~vable angle of feed dis-
placement. It is misleading now to express this angle in beam\vidths,
since there are two different beam\vidths in the two planes.
In actual practice the reflector of a carefully designed antenna is not
cut along the straight lines shown in Fig. 13.24a, but rather along one
of the equal illumination curves of the feed (cf. Chap. 12), e.g., the equal
illumination curve 14 db down from the point of maximum illumination.
Three such curves are shown in Fig. 1324b superimposed on the equal
deviation contours: Curve A is appropriate to a feed on axis and pointed
at the vertex of the paraboloid; Curve Z?, to a feed on axis bllt pointed
into the top half of the reflector; ancl ~u rve C, to a feed below the axis
490 SHAPED-BEAM AATTEKNA,T [SE(. 134

but pointed at the vertex. Whereas areflector shape that followsChrve

B is satisfactoryfor a point-source feed on axis, it would introduce serious
defocusing when used with an extended feed because of the high devia-
tion contours that it includes. Forthis reason symmetrically cut reflec-
tors are always used with extended feeds. In the case of a feed array
extending downward from the axis, the top element would favor the
reflector shape given by Curve A while the bottom element would favor
the shape given by Curve C. The actual reflector shape \vill therefore
be a compromise between these.

m
E
B 2-
1 l!-11~1
1 ;1114
B 2-

W (a) (b)
IIG.13,24 .Apertureof paral>oloid reflector shon,ing contours ofequal deviation of rays
frorll direction of central ray as feed is moved off focus in vertiral plane: (a) Lines BB
represent ~trnigt,t mts to narrow the aperture in one plane; (b) Curves A, B, and C repre-
w[lt equal illr ination contours for different feed pohlt]ons and orientations,

In the preceding discussion the feed displacement has been described

as a simple rotation about the vertex. It is not necessarily true that this
places the feed at the distance from the vertex corresponding to maximum
gain. l~easurements macfe under a variety of conditions have sho~vn
that the feed must then be moved away from the vertex to or slightly
beyond the vertical plane through the focal point. This optimum ciis-
tance for a given displacement angle \vill depend upon how the reflector
has been cut.
To obtain optimum performance of a cut paraboloid with an extended
feed, it is necessary to carry out an experimental design procedure. This
procedure \\-ith its results will bc dcscribrd for the 10 by 25-ft reflector
sho\vn in Fig, 13.15 at X = 10 cm,l A dipole feed ~vas mo~,ed transversely
over a range of 30 in. off axis (21 referred to the vrrtcx). At each
displacement the feed was then moved p~rallcl ~rith the axis to find the
point of maximum gain. 1hc optimum feed point proved to bc 5 in.
farther out along the axis for t}lc maximum feral displacement than for
the feed on axis. For these optimum feed points, the variation \\-ith
1C. C,, Strrgi(,p,lli,,,s, 1{1, I?rp,,rt X-,,. !)51, 101]. 12, 1946,
displacement of gain and azimuth beamwidth (perpendicular to the dis-
placement plane) were measured ~vith the results shown in Fig. 13.25.
The reduction in gain is not so serious as it appears, since a major part
of the reduction is due to beam spreading in the vertical plane in which
a flared beam is required. The increase in azimuth beam~vidth is
serious and sets the limit for this reflector on the angular range over which
it is practicable to flare the beam.

1=1=1=11.60
Gain
~ 6,000 \ /
~
m
4,000

rT_T--l&
~~ ~o0.60
20
Oipole paition, in. below focus
FIG. 13.25.Variation of gain and azin, uth I}ea],, width with dipole feed displacement for tlw
reflector of Fig. 13.15,

Dipole-array Extended Feeds. Several problems arise in the design

of a dipole array to be placed near the focal plane of a paraboloid for
the purpose of obtaining a flared beam. First the array should be
located w-ith reference to the focal plane so that the individual overlapping
lobes comprising the flared beam have maximum gain. The dipoles
along the Ivaveguide must be so spaced that the individual lobes are all
in phase ~vith each other. The inpllt impedance of the dipole array
must bc such as to terminate the transmission line properly. Finally,
the available power must be divided among the several dipoles in such
proportions as to obtain the desired beam shape.
The array is oriented with respect to the paraboloid axis and focal
plane on the basis of gain and beamwidth information of the type pre-
sented earlier in this section. The individual lobes n-ill combine in phase
if the radiations from successive dipoles arrive at the vertex in phase.
With the waveguide oriented for maximum gain of the individual lobes,
the relative phases are cent rolled by the spacing dipoles along
bctm-een
the v-aveguide. In terms of the qllantities shown in Fig, 13.26, the
492 SHAPED-B
E.4. V A .VTE.VAAS [SEC. 13.4

spacing for reversed dipoles is given by the relation

The dipole array is usually fed from a traveling wave which is realized
by terminating the array in a dummy load or preferably by using the
last dipole in the array as a load. If the latter technique is used, the
last dipole in the array must have its impedance such that it absorbs
all the power transmitted down the v-avcg~lide to it. This condition
is attained by proper adjustments of the depth of the probe feeding the

l~lu. 13.2ti.-Arrmy of chuoles on wavcguide in paralx,lnid reflector, slmxvin~ dipole spaci)lg

and orient at, un of array with rmpcct to vertex.

last dipole and of the distance from the probe to the s}~orting plug at, t,hc
end of the \vaveguide. The probes of the other dipoles are next inserted
into the waveguide in succession starting from the last dipole in the array
and proceeding to the first dipole. The depths to which the sucmwsive
dipole probes are inserted are determined by the po~ver division among the
dipoles necessary to produce the required flared beam. Oncc the desired
antenna pattern is obtained, the final impedance match of the array may
be accomplished by inserting an inductive iris of appropriate dimensions
at the proper location in the ~vaveguide.
The principal advantage of the dipole array is its simplicity in design
and construction. The disadvantage of this array is the interdependence
of spacing and phasing of the individual elements. Each clement, of the
array independently should provide proper illumination of the reflector.
The dipole elements suffer from the disadvantage that their radiation
SEC. 134] DESIGN OF EX 7E.VI)IiI) FhhI),S 493

patterns are dependent upon the polarization required and are relatively
inflexible.
Horn-array Extended Feeds.The problems in the design of a horn
array are similar to those arising in the design of a dipole array. Opti-
mum gain locations for the horns near the focal plane of the paraboloid
are determined in the same manner as for the dipole elements. The
radiation from the several horns of the array is made to arrive at the

Power dividing
irises

E vector E vector
(a) (b)
Klti, 1327. -Multlplc llorll feeds for illunli,l:ltillg a cut p:tr:%ll<,loid:(u) multiple Y-junctiotl;
(b) succe~Mvc T-ju,,rlio,,s.

paraboloid vertex in phase by a proper choice of the lengths of ~vave-

guide extending to the individual horns.
Tarious methods of dividing the po~ver in multiple horn feeds exist;
two methods are illustrated in Fig. 13.27. In the multiple Y-junction,
the po\ver is proportioned among the se~eral horns by septums extending
into the main waveguide. The impedance match of this multiple feed
may be accomplished by a single iris in the main waveguide. In the
array employing successive T-junctions (Fig. 13.27 b), the division of
power between the upper horn and the middle horn is determined by an
iris in the section of ~faveguide feeding the middle horn. The impedance
match of the combination of upper and middle horns to the waveguide
is accomplished by an iris just below the junctiun. The division of
pmrer bet \\-eenthe combination of top and middle horns and the lo~ver
horn is determined by the iris in the section of waveguide feeding the
494 SHAPE I)-BEA M A NT fi:,VKA,Y [Sl!x. 135

lower horn. The over-all impedance match of the multiple feed is accom-
plished by the iris in the main waveguide. A four-horn array of the
latter type is used in the antenna shovm in I?ig, 13.14.
The advantage of the horn array lies in the fact that the phasing
between successive elements and the radiation pattern of each element
are completely at the disposal of the designer and are susceptible of
calculation, The disadvantages of this type of array are the extreme
complexity of the design and the bulk and weight of the resulting feed.

IJIG. 13.28.-Shaped c> IInchral reflertor llluniillated I)y a pillkx line source

13.5. Cylindrical Reflector Antennas.The techniqlle of obtaining an

asymmetrical beam by a shaped cylindrical reflector and line source has
been used extensively in airborne navigational radar antenna design
where sharp azimuth beams and v-ide-angle vertical coverages arc
required. Figure 13.28 illustrates a shaped cylindrical reflector illu-
minated by a pillbox which serves as a line source parallel to the gener-
ating line of the reflector.
The general theory of cylindrical reflectors and line sources is treated
in Sec. 5.9. It was shonm there that the pattern in the plane perpen-
dicular to the generator of the cylinderthe vertical plane in the present
discussionis determined by the energy distribution of the source in
that plane and the cross-section contour of the reflector, and that the
pattern in the trans~erse planes is determinccf by the energy distribution
along the axis of the source. The cross-section contour of the reflector
is so shaped and oriented with respect to the line source that its lo~ver
section concentrates the rays from the source into approximately parallel
directions, thereby concentrating the energy into the peak of the beam.
The upper part of the reflector is bent forlvard ~~ith increasing curvature
to disperse its rays into a broad flare. The net result is an asymmetrical
Slcc. 13.5] CYLINDRICAL REFLECTOR ANTENNAS 495

beam, the exact vertical Mfraction pattern of which depends upon the
shape of the cylinder and the distribution of illumination from the line
source on the reflector surface.
General Requirements.The specifications on the vertical polar dia-
gram are determined by the operational requirements. For example,
if a navigational radar is to be designed for an aircraft flying at an alti-
tude of about 1 mile and is to be capable of covering a radius of about
20 miles, then the peak of the antenna beam should be depressed about
3 from the horizontal and the beam should be asymmetrical so as to
produce approximately constant illumination of the terrain. Whereas
the beam specifications depend on operational requirements, the over-
all size of the reflector usually depends on the available installation space.
If the antenna is part of an airborne radar, the allowable over-all height
of the reflector is limited; the antenna is installed with the sharply curved
upper portion of the reflector retracted into the fuselage, so that the
protuberance below the fuselage for housing the antenna need not be
large (cj. Sec. 14.3).
In order that the power reflected from the cylinder back into the line
source be negligible and that the line source have no destructive effect
on the vertical polar diagram of the reflector, it is necessary that the
reflector be so shaped that the complete system of reflected rays clears
the line source. To achieve this condition the radiation from the latter
is directed down into the reflector and the major part of the reflector is
belo~v the line source. It is also important that the radiation reflected
optically from the top of the reflector should not strike the bottom of the
reflector. A profile view of a typical reflector and line source is shown in
Fig. 13.29. The orientation of the top of the reflector is such that the
radiation reflected there passes clear of the line source and the bottom
of the reflector.
Line Sow-ces.-With the reflector height limited, it is necessary so to
design the reflector contour and feed aperture that the angle subtended
by the reflector at the feed includes most of its radiation; the illumination
should taper to a low value at the top and bottom of the reflector. This
precaution is necessary if the amplitude of radiation from the line source
going past the edge of the reflector is to be maintained low and if the
side lobes in the vertical diffraction pattern of the reflected radiation are
to be kept down. In addition, the feed must be designed and oriented
in the reflector with a view to minimizing the amount of feed back-lobe
radiation in the angular region of the flared beam. The azimuth diffrac-
tion pattern of the antenna is determined by the design of the line source.
Since the beamwidth should be as narrow as possible in azimuth to secure
good resolution on objects, and since relatively high side lobes in the
azimuth pattern are allowable, the line source is required to have as
sharp an azimuth beam as can be obtained \vith the length of line sourre
49G SHAPIIII-BIIAJ1 A.VTI<,Y.VAS [Six. 135

used. This means that the intensity of the source should be uniform
along its axis.
The various designs of line-source feeds used in this connection can
be classified according to two general types: (1) arrays of radiating e]e-
ments arranged in a line on a ~vaveguide and radiating approximately
broadside and (2) parallel-plate linear focusing systems that have rec-

.A
Line source
Incidentray

I
I

Reflected ray
passing clear
of line source
and boftom
of reflector

FIG. 13.29.Vertical section through SIULpCdC)IIlldrical reflector and line source.

tangular apertures with a large length-to-breadth ratio. In the terms

used in the treatment of linear arrays in Chap. 9, either resonant <Jr
nonresonant arrays may be used;1 design procedures and performance
characteristics are treated in detail in Chap. 9. The aclvantagm of
linear arrays for this application are their compactness and light \vcight.
The disadvantages of the nonresonant array are (1) that the beam scans
through a small conical angle as t}le result of frequency fluctuations and

I J. R. Risser, A. il. Stwnl:md, ,J.Stcinhwger,1,. J. 13ygrs,RI, Report ho. 973,

hlar. 19, 1946.
(2) that the beam must be kept off normal to avoid cumulative impedance
mismatch; this results in a conical beam that illuminates a hyperbolic
trace on the ground. The resonant array suffers from the disadvantages
of frequency sensitivity in impedance match and pattern.
A variety of parallel-plate focusing systems have been examined for
their usefulness in this connection. One type consists of a rapidly flaring
sectoral horn with either a dielectric or metallic lensz in the horn so
designed that radiation across the horn aperture is all in phase.
Another type of parallel-plate line source is the pillbox antenna dis-
cussed in Sec. 12.16. The pillbox line source is simple to design and
relatively simple to construct but suffers from certain basic disadvantages,
The center of the aperture is necessarily obstructed by the feed at the
focus of the parabola; this results not only in side lobes in the azimuth
pattern but also in an impedance frequency sensitivity as the result
of radiation reflected back down the feed line. Efforts to correct one
or both of these effects, have taken various forms: the design of a matching
plate to be put at the vertex of the parabola, 3 the design of new pillbox
feeds,4 and the design of double pillboxes to obtain canceling reflections. s
13.6. Reflector Design on the Basis of Ray Theory.A successful
procedure has been developed for designing the shape of the reflector
that is required to produce a specified vertical-plane pattern. The latter
is usually specified in idealized form: the po~ver distribution is to bc a
prescribed function F(o) bet\veen the depression (or elevation) angles 8,
and 192,and zero for all other angles. It must be recognized, ho~vever,
that it is impossible to realize a discontinuous power distribution of this
type accurately with a reflector of finite extent. Diffraction phenomena
are unavoidable, and the best that one can hope to achieve is to approach
the idealized pattern within acceptable limits of deviation from the pre-
scribed P(tJ) in the range (01,tJZ)and with an acceptable low level of inten-
sity outside the given range.
A good first approximation to the cross-section curve can be arrived
at on the basis of geometrical optics. G If it then proves to be necessary,

1 C. C. Cutler, I,inc %urcrs of llirro~v~vc I:ncrgy for Feeding (cylindrical Reflec-

tors, BTL Report hill-45-160-3, ,Jan. 5, 1!)4,5.
Z M. A. Taggart, Horn with Nfctal Lens, lU, Report No. 863, iYoI,. 13, 1!)45.
3S. Silver, .knalysis and (;orrcction of the Impcdancc llismatch Due to a Reflec-
tor, RL Report No. 810, Scpt, 2.5, 1945.
i Nf. .4, Taggart, A NW Pillhox FrcrJ, IU, lkport No. 862, Nov. 7, 1945; L. J.
Eyges, Lens Fwd for Pillhoxcs, RI, l{cport Xo. 86!), ,Jan. 23, 1946.
W. O. Smith , A Bro:~d-hand 7L.tf Pillhox, RI, ltcport No. 901, Jan. 11, 1946.
GR. C. Spencer, Synthmis of illicro~v:lve Diffr:~(tion l:ittcms ~vith App]ic:ltion
to CSC2(JPattmms, RI, llcport No. 54-24, .JIII]C23, 1943, dcwrihrs J,, ,J. (hlls method
for c:dculating the reflector sh:ipr. SW :L1.o(. (1, (Xutlcr, RTL Iteports 3111-44-160-
37, Feb. 14, 1944; hIM-45-160-4, J:uI. 5, 1!)45.
the curve is modified by means of the more exact analysis given in the
following section. We shall consider here the geometrical-optics tech-
nique. The cross-section configuration of the reflector, the line source,
and the rays is shown in Fig. 13.30. The z-axis is taken in the horizontal
direction; F is the trace of the line source. Positive angles are measured
in the clockwise sense as showm. The reflector subtends a total angle
*Z ~, at the source F. Let p be the radius vector from F to an arbi-

1
I
-z
1

I
I
I
I
1
\

a
Direction of
peak intensity
of the I!ne source
do

Flu. 13.3 U.GeonLctry of ray reflection at surface of reflector.

rrar,vpoint, on the curve, From the law of reflection it follows that the
angle bet!veen the normal to the curve and the incident or reflected ray
is (O ~) /2; the differential equation of the curve is then readily found
to be

(8)

The functional relation between 0 and $ must be determined to pro-

duce the proper dispersion of the primary feed energy into the secondary
pattern distribution. To this end consider the \vedge of incident rays
bet l~een $ and x + d+; on reflection this becomes a ~vedge of rays bet!veen
0 and 8 + df?. IIy the energy balance principle of geometrical optics the
power in the incident wedge is equal to the power in the reflected v-edge.
Let Z(+) be the distribution in the primary pattern about the line source
I. The incident power is then proportional to 1($) d+. Similarly the
reflected power is proportional to P(19) de. IIre have then

The constant K is determined from the rendition that the total primary
feed energy intercepted by the reflector must appear in the secondary
pattern in the required range (@I@z), \~-hence

* 1(+) d+
K=+;, / . (If))
P(o) do
/ 6,
,Similarly the total primary feed energy in an arbitrary range (~1,~) must
appear in a corresponding range (61,6) of the secondary pattern. This
leads to the integral relation

(11)

which serves to determine Oas a function of ~ for prescribed distributions

I(i) and P(6). For example, if F(O) is CSC2EI,17(1.(11) gives

cot 02 rot 0, * I(+)

cot e = cot e, + (w. (12)
1(+) d+ / $,
/ +1

The primary pattern function I(i) is generall.v knolvn only in numerical

form from experimental data. Ihe integrations over ~ must tIIerefore
be carried out numm-ically or graphically.
The functional relation 0($) determined from Eq. (I 1) is then sul)-
stit~lted into Eq. (8) to obtain the equation of the c~lrve. The integral
of hkl.
. (8) is

%=b+wd (13)

This integration must also he performed graphically or numerically.

It is seen that this leads only to the shape and not the absolute scale of
the reflector. The distance po from F to the reflector along the axis is
determined so that the rsquircd height of the reflector conforms to the
total angle ~1 +2 subtended at F.
There are a number of arbitrary ~-ariables in the procedure, the choice
of ~rhirh can be determined only by experience. It \rill be observed
thtii the line solwm is oriented so that its peak intensity is in a direction
+., 1]lis takes co~nizance of i 11(>fact that the lo\\t=rportion of the
refiector is rcqllirrd to prlxlllce th( high-intensity rrgion of the secondary
.500 ,YH,lP1(D-BE.4JIf A.VTfi3NA-.iS [s1,,. 137

pattern. The choice is not critical, and ~o may be as low as 0 and as

much as 25. The angle ~j ~, subtended by the reflector should be
sufficiently large to make efficient use of the primary feed energy. As a
general guide ~Z ~, should correspond closely to the 10-db ~\-icfthof
the primary pattmn l(~).
It is interesting to observe that the lower section of the reflector in
general turns out to be very nearly a parabola with focus at F and axis
FV parallel to the reflected ray in the direction 0,, The physical basis
for this property is readily c~-ident, for this portion of the reftector must
converge the divergent rays from the feed into a nw-ro\vbeam to produce
the required peak intensity. The total flared pattern may be regarded
as a superposition of a narro\v beam produced by the parabolic segment
and a broad beam produced by the dispersive section of the reflector,
13.7. Radiation Pattern Analysis.The secondary pattern that \vill
actually be obtained \rith the reflector determined on the basis of geo-
metrical optics can be calculated to a high degree of accuracy by the
methods of Sec. 59. Such a udculation ser}-es a scvertil-fold purpose.
The extent of the deviations from the idealized pattern can be determined
prior to construction and test of an cxperirnrntal model. The diffrac-
tion effects arc quite sensitive to the scale factor POin Eq. (13). If there
is some latitude allolved in choice of the reflrctor dimensions, the radiation
pattern can bc calculated for sc~rcra.lvalues of P{, for a given shape to
determine the best dimensions within the allo~vect range. For a given
choice of dimensions, the effect of small alterations in the cross-section
centour from the geometrical-optics curve can be studied to arrive at a
curve that yields an acceptable pattern. These results also serve as a
basis for setting the tolerances that are to be required on the reflector
sljape in the production of the antenna.
The re~der is retcr-rwltt) Srcs. 5.7 to 59 for the theoretical details.
Tlw calculation is biLscd on the uswlmptiuns that the reflector is in the
(,yiindrical \\avc zone of the so~lr[e, that the minim\ml value of p is
large c(unparcd ]vitb the ~\uvelength, and that, the m~ximurn va]ue of

p is less than ~, where 1is the length of the smlrce. The primary pattern
of thr stnlrce is specified l),v the distribution fllnction J(r) along the length
of the source and the anglllar (Distribution G(i) armmd the source. Thus,
if P is the total pmvrr rmliatcd, the power radiated in a segment of angular
width d+ and len@ h d.r is

(u = 2:Llf(.r)G(i) (fr d+. (14)

The properties of the distrib~ltion functions are givm in more detail in

1:(1s.(5%1) :Lnd (582).
The pr(mdllrr is lt) ralclll:ltc the cllrrrnt disl ril)lltion induced on the
reflcctur b,v thr primary field and then to calclllatc the radiation pattern
SEC. 137] RADIATION PATTERN ANALYSIS 501

of the surface-current distribution. Two cases are to be distinguished:

(1) longitudinal polarizationhorizontal polarization in the present dis-
cussionin which the electric vector of the primary field is parallel to the
line source and (2) transvers~verticalpolarization in which the electric
vector is parallel to the planes perpendicular to the line source. The
coordinate system is described in detail in Fig. 5.9 and will not be repro-
duced here. Only one change need be noted: In the present section the
positive d,rection of the angle 8 is reversed with respect to that chosen
in Sec. 5.9. The complete expressions for the patterns are given for
the t\vo types of polarization by Eqs. (5.86) and (587), respectively.
k<~rthe vertical plane pattern, that is, the plane @ = O in Fig. 59, these
reduce to
1, Horizontal polarization:
*L
E(E)) = jLl
/ *1 pG(@)cOsi[l+ +fi)2ee-e)d+-e)d+ 5)
2. lertical polarization:

(17)

The angle i is the angle bet\vmm p and the normal n to the surface; the
angle (n,R~) is between the normal and the unit vertor R, in the given
direction of observation in the secondary pattern. l~e note also that
from the preceding section

(18)

The reflector curve being given, p, cm i, and cos (n, R,) are kno\vn func-
tions of position on the reflector. The integrals are then evaluated
numerically for successive values of ~, The po~ver pattern is then
obtained by the usual methods which are discussed in Chap, 5:

P(8) = ~
() : i RIEI, (19a)

the gain in a given direction is

2111(1

G(e) = -). (19b)

1hc
ralr~llaticms arc laborious but straightforward. The reliability
502 SHAPED-BEAM ANTENNAS [SEC. 138

of the method may be judged by the comparison between the calculated

and measured patterns 13.31.1 The positions of the
shown in Fig.
maxima and minima are predicted accurately, and the maximum devia-
tion from the measured values is 2 db. The performance of the reflector
could have been improved by the procedure, outlined at the beginning
-o

4 !
; ---Calculated
\ Observed
r \
8 \
1
, \
I I
* 12
0 I
I
~ 1
( Q ,
3 16 1
1 / ,
1 \J \ 4
\
I ,, ,-.
I .,
\
1
\
I
\
/

. ~
\
: \

\
.J
32

\
36

-40 1 1 I
ZSI)O -200100 00 1~0 ZOO 300 400 ~oO 600 7o0 800 900
Angle
FrG. 13.31.-Calculated arldobs&rved vertical patterns ofcylindrical reflector antenna with
horizontal polarization,

of the section, of modifying the reflector curve until the amplitude fluc-
tuations in the calculated pattern are reduced to an acceptable value.
13.8. Double-curvature Reflector Antennas.-The line-source feeds
required to illuminate the cylindrical reflectors discussed in the preceding
section have a number of disadvantages as compared ~~ith point-source
feeds. They are in general bulkier, heavier, less satisfactory as to imped-
ance properties, and more complicated to design, build, and pressurize.
When a reflector is used with a point-source feed to form a shaped beam,
it is required to provide a pattern of the specified shape in one plane and
to focus in transverse planes. A number of cut-and-try improvisations
for accomplishing this end were described in Sec. 13.3 under Asymmetrical
Shaped Beams.
1Taken from ~r. J. Keary, Calculation of Vertical Polar Diagrams and Poirer
Gains of Antennas for .4irborne Navigational lkdars, ItL Report No. 75o, Sept. 10,
1945.
SEC. 138] DO UBLE-CURVA Ti3RE REFLECTOR AN7ENNA,S 503

BarFel RejZector Antenna.The barrel reflector antenna shown in Fig.

13.17 was the most successful of these improvisations but is subject to
some serious limitations. A consideration of these limitations will pro-
vide helpful guide lines in the design of double-curvature reflectors. In
the first place, the shape of the central vertical section of the reflector
was determined, not as the proper transformation from primary to
secondary pattern, but by the simplest geometry which provides a focus
in transverse planes. The bottom
half of the reflector focuses half
of the radiation from the feed
into a pencil beam. The barrel-
shaped top half of the reflector
focuses the primary pattern in
transverse planes but reflects it
unchanged in the vertical plane.
The superposition of the two
portions of the secondary pattern, 1+~, 1332.-Superposition of barrel and
paraboloidalamplitudepatterrls.
showm schematically in Fig. 13.32,
inevitably results i-n a depression in the pattern near the peak of the
beam, as can be seen in the pattern of Fig. 13.20a. Second, since the
feed is at the center of curvature of the barrel section in the plane of
symmetry, all the reflected rays in that plane pass through the feed.
This causes an excessive amount of power to be returned to the transmission
line resulting in a frequency-sensitive impedance mismatch. A third
difficulty is that the right and left upper portions of the barrel reflector
are set at such angles as to give rise to strong cross-polarized components
in the wide-angle portion of the flared beam. These cancel in the median
plane but add to the sides of the normal polarization lobe to give a con-
siderably broader effective transverse pattern. These several difficulties
with the barrel reflector antenna can be avoided by proper designl of the
generalized surfaces to be discussed below.
General Considerations. The primary feed pattern and the reflector
are to have as a common plane of symmetry the plane in which the beam
is flared-plane XFZ of Fig, 13.33. Two considerations enter the design:
The central vertical section of the reflector is to be adjusted on the basis
of the central vertical pattern of the feed to give the specified secondary
pattern; the remainder of the surface is to be so shaped as to obtain
pencil-beam characteristics in the transverse planes. The second con-
sideration requires that all rays from the point source after reflection
1R. ~. Spencer, Synthesis of lficrowave Diffraction Patterns with Applications
to Ckcz @Patterns, RL Report No. 54-24, June 23, 1943; J. F. Hill, G. G. Lfacfarlane,
W. Walkinshaw, TRI? Report No. 1878, Xlay 17, 1945; S. Silver, Double ~urvaturc
Surfaces for Beam Shaping with Point Source Feeds, RL Report No. 691, June 15,
1945.
504 SHAPED-BEAM AN TE.VN.4S [SEC, 138

from the surface emerge parallel to the central plane; the reflected wave-
front is thus a cylinder whose generator is normal to the central plane.
Transverse Sections of the Reflector.-The transverse sections of the
surface are determined by the requirement that the reflector is to convert
a spherical wave into a cylindrical
wave. The condition is easier to
formulate from the point of vie\\
of reception. Referring to Fig.
13.34, consider the sheet of rays,
all parallel to the central plane,
incident on the reflector in the
plane OA NP. The latter is per-
pendicular to the central plane.
We require all these rays to be
( brought to a focus at F. If then
l<IU. 13.33 .Dout)1e curvature surface and we take any line in the plane
spherird cOordlllate hyhtcm.
OA NP perpendicular to the rays,
the optical path from that line to F is the same for all rays in the sheet.
Let p be the radius vector from F to the central section curve, @ its angle
x

Central
curve p I
Parabolic
section s
x

\/
*A ~\
/\\/

FIG. 13.34.Surfaces of paralmlir section.

of elevation, and u the angle bet]!-een the incirkmt and reflectd ray in
the cent ral section. Through F dra,\- O.r normal tf) the plane 0.1 .IP.
In the plane 0.4 NP set up the orthogonal axes 0?/, OZ. \~ith Oy normal
to the central plane. Writing the condition of the optical path (cf. Sec.
4.9) lye have
AN+.VF=OP+p
or
(P2 sin u + U + z)> z == P(I + IYB 0). (20)
SEC. 13.8] DOUBLE-CURVATURE REFLECTOR ANTENNAS 505

This gives
yz = 4pcos2; (z+p Cos u). (21)

The section of the surface in the plane OANP is, therefore, a parabola
with vertex at P and focal length

f(o) = P(o) COS2y (22)

Given the central section curve, p(.$), j(~) and the associated reflec-
tion plane are determined. It is readily seen that the whole family of
reflection planes corresponds to a cylindrical wavefront. The barrel
section reflector isobtained bysettingp = constant, U(O) = O.
Central Section Curwe.-The cent,ralsectior curve isto redetermined
so that the system will radiate a desired Fraunhofer pattern. There is
no simple decisive procedure forrelatin~ the central curve to this ~at,tern.
L

which takes diffraction effects into account properly. The method dis-
cussed here is based on geometri-
X
cal concepts of energy balance Circle
between the primary feed and
Fraunhofer patterns. The latter
is a three dimensional ~attern for
_~-~
about F

2 - 2%
,-:P

.
I
whose specification we shall use
spherical coordinates such as are ~$$~lLz
shown in Fig. 13.33. The angle /
of elevation with respect to the
\ dO
XY-plane will be denoted by o
FI~, 13.35,GeornetrYof rentralcurve.
(@ being used for the primary feed
pattern). ~ is the azimuth angle in the planes 0 = constant. The pat-
tern is specified in the form P(o, v), the power radiated per unit solid angle
in direction (d,~). The central plane pattern P(8,0), hereafter designated
as P(o), is assumed to be specified.
Referring to Fig. 1335 it is seen that the differential equation of the
central curve is given by

a being the angle between the incident and reflected ray. Or since

1 dp +0
= tan-2- (23)
p d~ ()
The positive directions of o and @ are shown in the figure. The relation
between 0 and O, which is necessary for integrating the above equation,
is obtained from energy balance considerations.
This method is based strictly on geometrical optics. The assumptions
506 SHAPED-BEAM AN1ENNAS [SEC. 138

involved are that the beam is narrow in the transverse (~) directions
and that the transverse aspect of the beam is virtually independent of O.
The effect of diffraction is assumed to be the same in all such planes.
Accordingly we assume that we have a cylindrical wavefront in the far
field corresponding to geometrical optics and that the energy flow through
the cylindrical wavefront between the planes .9and O + dO per unit length
along the generating element is of the form

~((1) do F(y),

where F(y) represents the distribution in the transverse aspect of the

beam.
Let us take a small cone of rays from the source F in the central section
defined by the planes @ and @ + do and azimuth extent chj. The energy
in this cone is
I(o) d~ d+,

where 1(o) is the power radiated per unit solid angle in direction (0,0). ]
On reflection this energy appears in a wedge defined by the planes @
and 6 + de, since the reflected rays are parallel. 2 The width of the wedge
is p d+, so that in terms of P(d) the energy contained therein is

P(e) dep d$.

Equating this to the incident energy gives

pp(d) do = Z(O) dd (24)

or
l(o) g = @ (Ma)
P
Because of the factor p occurring in Eq. (24), the latter cannot be inte-
grated as in the case of the cylindrical reflector to give the relation between
0 and O. Taking logarithmic derivatives \vith respect to @ of Eq, (24a)
and substituting from Eq. (23), we get instead the differential equation

%+[tan(i) -%+1%+%($=0 25)

which is to be integrated numerically. Here

I(o) = dI/d@, pf(e) = g.

The general arrangement of the central curve and feed is shown in

1The solid angle in tl)e spi!cri{,al coordinates IM~d I rr~, i.s ros @ do rf@. For the
central section cos @ = 1.
ZThe mrrespoildence is not strictly trl[c. si]lr(, ravs in all illritlcnt sheet do not lie
in a plane on reflection. \Veare assuming that the error is negligible in the neighbor-
hood of the central section.
SEC. 13,8] IXIUBLE-((:R VA 71 1{1< 1(1<1<l,fi;C7{)li AN7EIV,VA,? 507

Fig. 13.36. The angular limits o,, OZ of the reflector are arbitrary.
A satisfactory choice is to take these to correspond to the 10-db points
in the primary pattern. The anglllar aperture OZ +, must be taken
fairly large, or the s~lrface will be found to curl in too rapidly in
the transverse aspects. By setting up the reflected rays as shown in the
figure the feed is kept out of their paths and the mismatch due to the
reflector is kept at a low level. Since the region @l ~ @ ~ O contributes
to the high energy region of the Fraunhofer pattern, the primary feed

Peak illumination
of the primary
feed pattern

i
FIG, 1336.-General arrangement of central curve apd feed

pattern is tipped to illuminate that area more strongly. The angle of

tip is arbitrary. For cylindrical reflectors with a line source this is
usually taken in the neighborhood of 15. This should be satisfactory
here. With the choice of these various factors decided upon, the integra-
tion problem is defined.
Integration fo~ the CSC20 Pattern-We shall discuss the integration
of the central curve equations with particular reference to the CSC26 pat-
tern used in airborne navigational antennas. P(o) is required to be a
CSC2@ distribution between the angular limits 01 and e~. That is,

P(fJ) = K2 CSC2e, 01<8 <02, e,>o

=0

outside this range. Using Eq. (25), Ire first find 8(o) so that

0=0, when 4 = +1,

0=6, when 4 = +2.
F~)r the CSC2d pattern we get

%+[an(~)-%l:
-otf%)=
2)
-SW nllmerical integration of this equation subject to thr end-point con-
508 SHAPED-BRA 1[ AA TIIN.VA,9 [s,:(. 130

ditions at O, and d, requires a guess as to the value of the slope dO/d@

at the starting point. The integration is carried through for the trial
value of initial slope, and the resulting end value OZis determined. on
the basis of this result a second assumption is made for the initial slope,
and the integration is repeated. The second or, at most, a third gIIess
is sufficient, especially since the end value OZis not critical. The initial
slope may be expected to be small, since this region of the reflector con-
tributes to the peak of the beam and the reflected rays are very nearly
parallel. Throughout the course of the c~lrve dfl ~d@ m~~st be positive.
13,9. Variable Beam Shape.It is frequently specified for an airborne
navigational radar antenna that the shaped elevation pattern be con-
trollable, either continuously or in a limited number of steps. This
requirement is imposed because of the need to operate the aircraft at a
range of altitudes, whereas the Cscz o pattern is designed for a single alti-
tude. Several means have been employed to exercise control over the
elevation beam shape, In one cylindrical reflector antenna the dirrc-
tivity of the feed was altered by means of a flap to direct more of the
radiation onto the sharply curved portion of the reflectcr and thus p(lt,
more power into the wide-angle part of the elevation pattern for high-
altitude operation. In the case of the cylindrical reflector-pillbox
antenna, measurements have been made on the change in shfipc of the
secondary pattern of the antenna as the pillbox is rotated about the
long axis of its aperture. Similar measurements have been made \vit h
encouraging results by rotating a directive horn feed in the barrel reflec-
tor. 2 In all such cases, after the altered pattern shape has b~en obtained
by a change in feed, it is then necessary to rotate the entire antenna to
obtain the correct minimum depressionRotation of the reflector
angle.
about a horizontal axis near the vertex changes the ill.urnination and
depression angle together in such a way as to give satisfactory patterns
in an actual case3 for depression angles from 3 to 13.
It is sometimes necessary to compensate for roll and pitch of the air-
craft by line-of-sight stabilization which maintains a fixed rlevation angle
of the beam with respect to the horizon. For simplified mcchanicfil
control and to eliminate mechanical interferences, tlwre are ad\wntages
in obtaining stabilization in the beam by rotating the reffcctor alone in
the vertical plane. In this case, it is desirable to change the position
of the beam without changing its shape. This is difficult because dis-
placement of the feed cw~ses changes in the illumination over the reflector.
Both of these problems }vould be much simplified if the reflector were

1 C. C. Cutler, BTT, Report lf3~-4.5-160-4.

2 J. H. Gardner, 1I,O\v Altitude Csrz tl Antenna, RI, Report Xo, 1073, Fch, 21,
1946.
3 T. J. Keary, .4. R. Poolr, J. It. Risscr, H. Wolfej 1.iirlltjr]lr N-:tv]g:ltio!)al lt:itl:ir
.intennas, RI, Report Xo. 808, ~Iar. 15, 1946.
not asked to perform the dual function of fomlsing and shaping the beam.
In one antenna design a point source and cut I]araholfji(l procl(lcmf a
collimated beam which ]~as then shaped in the vertical pl:me by a ser(md
reflector. In this case an excellent elevati(m pattern \\asol)t:linml. Fly
control of the second reflector alone it \vas possible to chang~, the shape
and direction of the beam for deprmsion angles ranging from 10 to 9
and to accomplish line-of-sight stabilization over a range of t 1.5. The
serious disadvantage of the double-reflector antenna is 1he spare that
it requires.
I E. B. Chisholm and B. R. Vogel, ( Ilouhlr Reflrctof .fj]tenn:~, 1{1, Report
No. 775, July 16, 1945.
 CH.\PTER 14

ANTENNA INSTALLATION PROBLEMS

BY L. C. 1.m .iTT.L AND R. 31. REDHEFFER

GENERAL SURVEY OF INSTALLATION PROBLEMS

The customary procedure in microwa~-e radar antenna development

has been to design the antenna and to carry out the early experimental
tests on the assumption of free-space conditions surrounding the antenna.
Whereas this represents a good approximation in general, it is necessary
eventually to consider the effect upon the antenna performance of the
supporting structure on which it must be mounted. .Mso, it is generally
necessary to place the antenna in a dielectric housingthe radom~
which likewise affects its performance, It is sometimes possible so to
choose the antenna location on the structure and to design the radome
that the original performance of the antenna is unimpaired, The final
result can be predicted with greater certainty, hofvever, if the electrical
design of the antenna is considered from the beginning in conjunction
with that of the raclome and ~1-itha ~-iel!- to the structure and location
that the antenna-radome system must occupy,
The purpose of this chtipter is to present the problems imposed by
installation requirements and the practices that have been adopted for
dealing ~vith them. These con.. iderations are intended to serve merely
as background for one enga~ed in antenna design.
14.1. Ground Antennas.The meclumiral drsi~n of an antenna for
ground use n~~lst effect a compromise bet\vecn the factors of electrical
rehability, mechanical rugge(lnes, an(l portal) ility. .In antenna in a
permanent site has moderate l! eight and portability requirements, is
not required to sur~irc repciatcd ro(lgh handling, but sho(lld l)e ideally
sited and should be designed to opemte rrlial)ly o~er an extended period
in the face of l(Jcal conditions of !!eathf r and l\ind.
The antenna of a high-l]rrfornli~n(( radar set \\-hichmust be moved
quickly from time to time o~er consi(i(rul)le di+tanc~ to a ne~v site is
still not seriouslv rcwtrirtrd in ]reight bllt must disassemble into rela-
tively light ~~eight components of limited (limexwions for transportation
by air or truck. Its com~tr(lction rnw+t permit precision reassembly after
repeated handling, (omp(ments m(l>t lw >Ilweptiblr to repair or replace-
ment under field conditions.
The ontenna of a +)nle!l-h:ii lixhter ra(l:~r set may be trllck-rnounted
\\it}) I be rest of the set. Tl]i. (,lin~in:~t(+ tl)( nrc(+~i~y for disassembly
,510
SEC, 14.2] SHIP AN TLNIVAS 511

and facilitates ground transport when relocation of the set is required.

At the same time it rules out the desirable features of air transportability.
Very lightweight sets have been designed for transport by mule or pack
for use in inaccessible regions. So much is sacrificed in reliability and per-
formance, however, by this severe portability requirement that such
sets have not been demonstrated to be practicable.
Some ground antennas, particularly truck-mounted units, are pro-
vided with radomes. More often the antenna is exposed. It proves to
be more practicable in these cases to design the antenna for operation in
the open than to provide a satisfactory housing. The antenna feed
must then be weatherized by means of a dielectric cup or plate. The
reflector is generally of mesh, perforated, or grating construction to
reduce windage effects. Mechanical and electrical components must
be enclosed in a metal housing.
14.2. Ship Antennas.-The distinctive features of a ship as an antenna
location are the small number of suitable sites and the large amount of
other gear that invariably interferes with the placement and performance
of an antenna. The top of a mast is obviously the ideal location either
for an antenna with an omnidirectional pattern or for one with a com-
plete azimuth scan. Any other location }vill involve obstruction of the
beam by the mast or by other parts of the superstructure. Such obstruc-
tion will result in blind regions, false signals, or transmitter pulling.
In addition to competing for favorable sites, antennas obstruct each
others view and jam each other \\-hen fre<luency relations permit.
Furthermore, mechanical and electrical considerations are frequently
at cross purposes. A consideration of each antennas performance
argues for placing it at the top of the mast, ]vhereas a consideration of
ship stability would place it beloiv deck. An antenna mounted alongside
a mast on a bracket should be far out from the mast for electrical reasons
but close in for mechanical reasons The electrical performance is seldom
improved by the modifications introduced to provide resistance to shock
and vibration.
A structure of limited extent equipped with a large number of antennas
is referred to as an antenna s,vstem. The antenna system problem then
is to obtain satisfactory performance from the severid antennas by relo-
cating them, by combining their functions or othernrise reducing their
number, by redesigning them, or by redesigning the supporting structure.
The ship antenna s,vstem problem is still far from solution, and each ne~v
ship type and mnr eq([ipmcut serve to increase the over-all problem.
Extensi\e mcwt[ remen ts, illcl ((ding mo(lcl ln(>:~slllcr~](rlts,:tre required t o
assess the ~)(:rfotrll~l]l(,(> of cxistillg tint [)rlnas in prcs(s]ll ]tx,atitms and to
predict their ]J(tlfJrI]l~Lrl(( in ot Iwr ltNat io]ls. Slli[) sIIp(,Ixt rll[,tlllc+ and
antennas m(lst lx more closely integrated in tllc dcsi~n stages. onc
approach to the problem that is commanding considerable attentiou is
512 A.V Y1<.V.Y.i IN, Y1ALLA lION PRI)BL14M, S [Sin. 143

the design of series or parallel coupling circuits to permit the multiple use
of antennas of certain types. Rejection filters can be designed to
eliminate jamming between antennas and to reduce background noise.
Special arrangements have been employed for improving the per-
formance of individual equipments in tile presence of screening structures.
Radars with scmning antennas maybe duplicated fore and aft so that
their regions of effectiveness supplement one another. An antenna with
an omnidirectional pattern may be rep!aced by two antennas on opposite
sides of the superstructure with 180 azimuth patterns.
Another coverage problem is introduced by interference nulls in the
elevation pattern when sufficient radiation from the antenna is reflected
from the deck or sea. One solution to this problem which has been
employed in the case of nonscanning antennas is to replace the antenna
by several properly distributed in height. Such a set of nonscanning
antennas may be connected to different receivers or may be connected
to a single receiver with a diversity hookup which leaves the anter,na
receiving the strongest signal actually connected to the receiver.
In designing an antenna for shipboard use, the effects of dampness,
salt spray, condensation, temperature extremes, high wind velocities,
and icing conditions must be considered. Some small nondirectional
antennas and scanning antennas are enclosed in radomes. The larger
scanning antennas are ~veatherized at the feed and are provided ~~ith
open\rork reflectors of perforated plate or grating construction to reduce
windage effects.
14-3. Aircraft Antennas.An aircraft ~vith its marij antennas for
communication, navigation, instrument landing, radar, identification,
and radar countermeasures pro~-ides an antenna system problem of great
complexity. To the problems of siting, avoiding interference between
antennas, and obtaining a proper pattern with the antenna on the struc-
ture is added the problem of meeting serious aerodynamic requirements.
In the faster aircraft it is desirable to have the antenna totally contained
within the airframe. IVhen this is not possible, the extension should
present minimum frontal area and should be streamlined with a housing
that must have a greater elongation in the direction of motion for a
higher design speed of the aircraft. Any changes in airframe imposed
by antenna requirements must be incorporated in the very early stages
of the aircraft design. To reduce drag, to protect the antenna from
wind forces and weather, and in some cases to provide for pressurization,
every scanning antenna must, be pro~-ided v-ith a racfome. This is true
~rhether the antenna is totally included ~vithin the airframe or is exposed
in a streamlined ho~wing, often referred to as a blister or nacelle. The
electrical and mech:mical dmign req~lirements imposed on such radomes
have become increasingly severe because of the trend to\vard shorter
~~a~,elengths, larger antennas, and more complete streamliuing; the satis-
factory solution of the radome design problem hm rm~uired increasingly
more sophisticated engineering.
Improper siting of the antenna on the aircraft, or unsatisfactory design
of the radome may result in a variety of performance defects in the
system. The radiation pattern of the antenna may br sm-iousl,v altered
by near-by conducting edges or surfaces. Excessive absorption or reflec-
tion by the dielectric ~vall of the radome may introduce a number of
undesirable pattern and impedance effects. Such defects have been
tolerated to a certain extent in the recent past as unavoidable in the face
of the rapid development in aircraft design and in the variety of antenna
installations. However, with a better understanding of the design prob-
lems involved and with the possibility of accommodating antennas more
satisfactorily in new aircraft designs, performance defects in aircraft radio
systems can be drastically reduced. It will be necessar.v, however, to
accomplish this improvement in electrical performance with antenna
designs that are at the same time more satisfactory aerodynamically.
14.4. Scanning Antennas on Aircraft.A scanning antenna employed
in air-to-air search is rec~uired to have a narrow pencil beam and to scan
a for~vard angular region only. Such an antenna can be located in the
nose of a multiengined plane or in a wing nacelle in a single-engine plane.
The wing nacelle can be located in the leading edge or at the tip of the
wing without introducing serious drag. The performance of a narro~v
pencil-beam antenna is not appreciably affected by metal parts of the
aircraft. It need not be affected seriousl~ by the radome ~vall exmpt in
cases of poor radome design. The effect of the latter ~vill be discussed
later and is illustrated in Fig. 148. Difficulty is also encountered \~ith
streamlined radomes designed to meet the aerodynamic requirements of
very high math numbers, since this necessitates near-grazing angles of
incidence of radiation upon the walls,
An antenna scanning in azimuth for air-to-ground search is required
co have a beam that is sharp in azimuth and achieves ~vith high accllracy
a prescribed shape in elevation such as was described in Chap. 13. If an
intelligible picture of the ground is to be presented on the cathode-ray
screen of the radar set, the elevation pattern must follow the CSC2o shape
river a wide range of angles with an accuracy of 1 db for closely spaced
variations. If a maximum range of 50 miles is to be covered from an
altitude of 5 miles, the steep-angle portion of the pattern is at lemt 20
db down frGm the near-horizontal portion. Surfaces or ed~es near the
antenna can reflect, or diffract a small amount of po~~er from nc:ar-
horizontal portions of the beam into directions corresponding to steep
angle port ions of the beam. If such an un~vantcd cent ribut ion at steep
angles is prwent even in power intensities 40 db doll-n from the pcali or 20
cfb down from steep-angle portions of the beam, the resulting iuterferencr
eflect will prodllcr a l-dh ripple in 1his portion of the clc]afion patt~rn
514 ANTENNA INISTALLA TION PROBLEMS [SEC. 14.4

The most serious installation problem is encountered when an antenna

that is employed in air-to-ground search is required to scan through 360
in azimuth. The ideal location for this antenna is at the lowest point
in the fuselage. Since obstructions on this portion of the airframe intro-
duce serious drag, the antenna and radome must be so designed as to
minimize the transverse area exposed and to realize streamlined flow.
The problem is aggravated by the fact that these antennas are generally
large. Certain design features may be incorporated to meet this situa-
tion. The antenna may be so shaped (Sec. 13.4) that the upper por-
tions of the reflector contribute to the steep-angle pattern and only the
lower portion of the reflector protrudes from the fuselage. The well
into which the antenna is recessed may be made large so that metal
edges are further removed from the antenna and are more completely
cleared by the slightly depressed beam. However, we still have a situa-
tion in which an antenna that is required to produce very accurately a
prescribed pattern is closely surrounded by a metal surface into which it
is partially recessed. Furthermore the beam is required to pass through
a radome w-all at near-grazing incidence. This combination of conditions
makes these antennas particularly subject to the performance defects
mentioned in the previous section. Experience with scanning antennas
will serve therefore to illustrate some of the major defects encountered
in airborne microwave radar performance due to faulty design or siting
of the antenna-radome component.
Performance defects from the operators viewpoint are best illustrated
by actual photographs of the cathode-ray tube that supplies him with
visual information. A PPI photograph substantially free of defects is
first presented for the sake of comparison. PPI stands for plan-position-
indicator in which slant range is displayed radially and azimuth angle is
represented by polar angle. Figure 14.1 is a PPI photograph of fairly
flat wooded terrain taken with 3.2-cm radiation. The antenna employed
a 29-in. paraboloid reflector with a barrel-shaped insert in the upper half
to obtain a shaped beam. The uniform illumination of the ground and
especially the absence of lobes in the elevation pattern allow such details
as the small lakes and the river with its islands and bridges to be clearly
recognized. The black disk in the center is a measure of the distance
of the aircraft above the ground; the bright circle surrounding the disk
is caused by perpendicular reflection from the earth and is called the
altitude circle. This photograph is to be compared with some less satis-
factory ones which follow.
There are certain defects in performance that can be traced to the
presence of conducting edges or surfaces near the antenna and can be
interpreted in terms of such physical phenomena as reflection, refraction,
and interference. Microwave radiation shows a sufficiently optical type
of behavior that antenna sites which would involve total obstruction
SEC. 14.4] SC A.VNI.YG .4.JTEN,VAS O.V AIRCRAFT .515

of the beam by portions of the aircraft unimportant directions arc recog-

nized and generally avoided. Two examples of excusable obstruction
are shown, however, toillustrate the effect. Figure 14.2is aPPI photo-
graph taken with 1.25-cm radiation in a shaped beam produced by a
42-in. shaped cylindrical reflector. In addition to hills, a river, and three

FIG. 14.1 .PPI photograph of wooded terrain, This photograph is essentially free of
defects due to antenna pattern.

airfield runways, the photograph shows two black sectors extending in

to steep angles and caused by the lowered landing wheels. In this
PPI the altitude circle had been subtracted out. In Fig. 14.3 the
shaped beam of 3.2-cm radiation from a 30-in. paraboloidal reflector
with a barrel insert has been obstructed by a second aircraft below the
antenna. The effect of this obstruction is to block off a portion of the
radiation and to create a radar ishado\v against the illuminated
background.
Thet~vo elevation patterns sho\~ninlig. 144illustratea neffectt hat
can result from mounting the antenna ,,n an aircraft. This antenna,
516 A.VTENii-A I,YSTALI.A TIOiV PIiO13LE.lfA [SEC, 144

designed to give a, shaped beam of 3.2-cm radiation, employs a pillbox

line-source feed and a W-in. shaped cylindrical reflector. The reffector
is 12-in. high but extends only 4 in, below the center line of the fuselage
forthepattern shown in Fig. 1+.4a. The betim is directed forward along
the line of flight \vith the peak tilted dovm 6 below horizontal; as the
tilt angle is decreased from 6, the pattern becomes rapidly ~vorse. The
intcrterence effect which is evident in this pattern results from a combina-
tion of the direct radiation from the antenna ~vith a small amount of
radiation scattered by the straight rim of the rectangular hole into which

FIG. 142,-PPI photograph showing blank sectors produced by landing wheels.

the antenna is recessed, This scattm-cd radiation is spread broadly in

elevation but is essentially confined in azimuth to the sector of the
original beam \vhen the antenna is pointed for\vard, Reflection from
the underside of the fuselage can bc shov n by geometrical arguments not
to be a contributing factor. Figure 14.5 sho~vs the appearance of this
interference effect in the centrul portion of a PPI photograph. This
photograph \vast akm from an altitude of 25,700 f t over Lake Okecchobee,
Florida, with a 50-mile maximum range setting.
Whereas flat surface reflections did not contribute to the effects
described abovr, there wer~ a number of cases obser~-ed in which inter-
ference effects \\-cred~le to reflections from. the undersurfaces of fuselage
sm. 14.4] ,YCA .V.YI.VG A STll)V.h-A,S O,Y .1 lRCRAFT 517

or wings. Some PPI photographs sho~v small regions of interference

fringes at right angles to the direction of flight. These fringes occur with
very close spacing in angle and can be shown to result from interference
bet\rcen the direct beam and radiation reflected from the undersurfaces
of the \rings. The fringe pattern observable in Fig. 14.6, on the other
hand, is quite extensive fore and aft with a ~vidcr angular spacing between
fringes which also varies with ozimuth angle. Calculations confirm the
assumption that this interference pattern is due to reflections from the
cllrved undersurface of the ftlsclagc.

i1(,. 14.3-PPI photograph showing l>lank area duc to obstruction by motker airrr:, ft

~l]t,rc :~re 01her d(fects in the pcrformanm of a radar set \vhich can
IN, traced tc) elcrt rical effcrts ulxm the beam callsed by the dielect rie
Jvalls of the rad(mle hmlslng the antcuna. These defects may be list d
m c(mlplete blanking out of all sign:lis in certain azimuth sectors, retiur-
tion in ran~c, obscuring of the srrren, false signals, and displacement of
tllr tarxet. Sryere bl:ulliin~ in cert tiin srctors, as shown in Fig. 117, is
(Ilie t [) rc(lcctions at the radome Jvall. These reflect ions direct radiai i(~n
back do~~n the r-f line to the transmitter which is thereby pulle(l in
fre(luency off the pass band of the receiver. A smaller reduction in range
which is at the same time more uniform in azimuth results from excmsi~.c
absorption of r:diati(m in thr w:ill of tllc radome.
 !
,\
,

/ 1
:,+
=. ,< -.. . .,,)
=.
. ../ . .. . . =.! , =..
-+, -.,
[+

E@ 55:

20 50 60 70 90 100
5,4,37 0
,
. 1
. %b

g.
..
-y:

30...- L.,\ -43 ..L Uk 1 Ll _- f_ I / \ / 1

,.
93 em m, 6Y 55, 50~ 45Q 40~ 35. 25. 20~ lW
(b) 30
I:lG. 14.4. Elev:,tion pattern. obtal,,cd ,,-it!, z sl,:,ped.l,ea,,l X,,tcn,,a: (a) ,noullte~ on the

underside of a fuselage; (h) !I,ounted in free spare.

FIG. 14,5.PPI photograph showing interference fringe due to edge reflection.

SEC,14.4

IhG. 14.6.PPI photograph showing interference due to reflections from fuselage surface.

llti. 14.7,PPI ljhotograph showing sector blanking clue to radome wall reflections.
520 .hVTE.V.V.AI.VS!IALLATIO,V 1IWBI]EMS [SEC. 144

Other effects than transmitter pulling can result from radome wall
reflections. Figure 14.8 presents four photographs of a B-scope (range
vs. azimuth angle) taken with 10-cm radiation and a 29-in. paraboloid
antenna located in the streamlined dielectric nose of a two-engine air-

ground which then returned a signal by the same path. Two methods
of reducing this ground signal were proposed: the use of vertical polariza-
SKC. 14,5] IIE.IC().V A.VTI{.Y.V.I,Y ().V A IRCR.APT 521

tion and the use of a radome Ivith a thinner wall. Figure 14.8a shows
the partial obscuration of the tar~et signal by the altitude signal even
v:ith optimum gain setting, when the thick radome and horizontal polari-
zation arc used. The reduction in altitude signal accomplished by
changing to vertical polarization is sholrn in Fig. 14.8b. The improve-
ment resulting from the use of a r-dome with the wall thickness reduced
by one-third is shown in Fig, 1-18c. The improvement realized when
both vertical polarization and the thin radome arc employed is evident
in Fig. l-1.8d, since it has been possible to bring the target signal out
quite strongly ~vithout bringing out the altitude signal. Effects of the
radome on system performance will be discussed more fully later in the
chapter.
14.5. Beacon Antennas on Aircraft.The majority of long-wa~e
antennas on aircraft are required to have omnidirectional patterns.
Because of the strong and unavoidable influence of the aircraft on the
antenna pattern at these wavelengths, the omnidirectional requirement
is usually not N-c1lsatisfied. In the case of microwave beacon antennas,
~vhich are also required to have omnidirectional patterns, the inffucncc
of the aircraft on the pattern is still strong but is more predictable and
also more nearly a~widable. There are a limited number of sites on an
aircraft suitable for a micro~va~~ebeacon antenna; the top of the vertical
stabilizer, the highest point on the upper side of the fuselage, and the
lowest point on the underside of the fuselage are three favored locations.
In selecting such a site the influence of the supporting structure on the
pattern must be considered, not only ~vith the aircraft in level flight but
also under conditions of roll and pitch. The ideal elevation pattern
would have uniform intensity for 10 about the horizontal plane to allow
for roll and pitch of the aircraft and ~vould have an approximately
cosecant-sqrrared dccrcase in intensity on the lower side of the beam to
provide uniform illumination of the ground in to steep angles. The
lm\-est point on the underside of the fuselage is the only location from
\vhich an unobstructed vie~\-of the ground at steep angles can be obtained.
Even ~vith this location the vic~v in near-horizontal directions may be
obstructed by roll of the aircraft if the bottom of the fuselage is flat.
Let us consider the ele~ation patterns that are obtained ~vith a beacon
antenna mounted on the underside of a fuselage. The various factors
that must be considered are the polarization of the radiation, the vertical
dircctivity of the antenna, the distance between the radiating elements
and the fuselage, and the extent and curvature of the fuselage. .4 radi-
ating element located below the fuselage wi]l send some radiation directly
to~vard the ground and some indirectly to~~ard the ground by reflection
in the fuselage. Tbc over-all effect can be simply described by postulat-
ing an image of the radiating elcmrnt in the f~welage. JIrcausc of the
natllre of the reflection process, t,f)eirnagc of :Lvrrticall.v polarize{] element,
522 ANTENNA INSTALLATION PROBLEMS [SEC. 14.5

will be in phase, while the image of a horizontally pola?ized element will

be 180 out of phase with the radiating element. If for the moment
we consider the underside of the fuselage to be an infinite horizontal
plane, it is evident that the over-all elevation pattern from a radiating
element and its image will have a maximum in the horizontal plane in the
case of vertical polarization and a minimum in the horizontal plane in
the case of horizontal polarization.
Let us now assume a requirement of continuous elevation coverage f o r
an angular range from 2 to 30 below horizontal. For vertical polariza-
tion the coverage requirement in the horizontal plane is automatically
met except for the effects of roll and pitch. The steep-angle requirement
can be met by limiting the combined length of the array and its image
in accordance with the relation d = X/sin o where 0 is chosen somewhat
greater than 30. For horizontal polarization it is more difficult to
obtain coverage in directions near the horizontal. Under the assumption
of an infinite plane fuselage it would be necessary to make the linear
array long and to accept an interference pattern in the region to be
covered. Since the fuselage is actually curved and of limited extent, it
is possible to obtain the required coverage by the use of a relatively
directive array either located near the fuselage or set off at some dis-
tance from it, depending upon the shape of the fuselage. It is evident
that this amounts to solving the electrical problem by creating an aero-
dynamic problem.
Microwave beacon antennas have regularly been enclosed in radomes.
Many of the smaller arrays originally were provided with cylindrical
radomes, but later it was recognized that no antenna protrusion is small
enough to justify the omission of streamlining. In the case of vertical
polarization the streamlined radome can have serious effects on the azi-
muth pattern. This situation favors large reflections, since it involves
wide angles of incidence for radiation polarized perpendicular to the plane
of incidence. A solid radome of low-density material affects the azimuth
pattern also because of the focusing effect due to path-length variation
in the material with azimuth direction. Vertical wires properly spaced
in the dielectric offer a possible means to maintain unity dielectric
constant.

RADOME DESIGN PROBLEMS AND PROCEDURES

The relation of the dielectric housing, or radome, to the general
installation problem has been considered in the previous sections. The
principal purpose of this part of the chapter is to analyze the problems
and to describe the procedures associated with radome design. 1,2 It
1Thk subject has been treatedmore fully in Radar Scanners and Radornes, Vol. 26,
Radiation Laboratory Series.
2Thk material is a severe condensationof the subject matter in a seriesof RL
will be necessary to consider several ~vall designs and t~vo general radome
types in terms of the mechanical and aerodynamic, as well as the r-f
requirements that are involved. These considerations will be assisted
by a preliminary examination of the nature and magnitude of the effects
that radomes can have on system performanm, An appreciation of these
effects establishes the need for further investigation and indicates the
difficulties that must be avoided in order to achie~e good radome design.
14.6. Relation of the Radome to System Performance.-In Sec. 14.4
radar system performance defects caused by faulty radome design were
presented from the operators vic\vpoint. In the present section various
reactions of the radume on the radar system will be presented from the
radome designers vie\rptjint.

reports on rmdonlc s(lhjc,(,ts. rh(%c report. arc listed here to avoid repeated multiple
references:
Y. Dowker, l)ielct,tric f70r,st;mt a,,d 1,0ss ~:mgcnt ~omplltation,) No. 483-1!},
Aug. 7, 1!)45.
, Trans,t,ission of I,OSSJ-S:lt,,l}!-ic}lcs, Xo. 483-22, ,Jcn. 23, 1{)46.
Elizabeth Jlvcrh:~rt, Itcccnt I)iclcctric ( onstant and Loss Ymgcnt Illeasuremcnts.
NO. 483-5, ,J\dy 14, l!)M.
, Sand\vich l~alls at V:lri:lhlr .-in~lcs nf IncirJcncc, h-o. 483-8, Dec. 19, 1944.
Radome W-all Iteflcctions at Yarm}]lc Angles of Incidence,f h-o. 483-20,
Jan. 4. 1946.
ill. Hegarty, 1. l)owker, 1{. 31. l{cdhcffcr, 1;. D. Ilinklcr, (urrtnt Irogrcss on IL-f
Research, X-o. 483-17, I[:ly 10, 1!)4,5.
E. B. llclIilhm, Outllnc of l{ado]!lc Dcvclopn]cnt Prohlcms, N-o. 483-1, Der. 2,
1943.
H. A. Perry, Ilc(trical ~rst 31et hods for I{c(lnmcs, A-o. 483-26, ,Jan, 11, 1946.
R. M. Redhcffci-, .ln Wtlinc of the llcctrical Properties of ]{adomcs, No. 483-2,
Dec. 20, 1!)43.
, Transmission and Reflection of Single Pktnc Sbccts, X-o. 483-4, ,July 12,
1944.
, Radomcs ~]ld Svstcm Pc,rformancr, ] No. 483-6, Xov. 17, 1944.
, Trallsl]lis.+ion al]d l{ciie(,tio]l of I%rallcl lla]]c Sheets, A-o. 4S3-12, ,Jan. 2(;,
1945.
, Ilcctric~l Properties of Dollblc l~all and Sandwich Radomcs, ] No. 483-11,
Feb. 1, 1!)4,5.
, I-;lliptir;ll lol:lriz:ltion Irodllt,r(l by Strc:~llllincd Radomes, No. 483-13,
Feb. 12, 1945.
and 1;. D. lf-inklrr, flc 31eastlrc)ncnt of Dielectric (70nstwrts in the Onw
centimeter R:u)d, So. 483-15, lbiy 11, 1945.
, 1The Intcr:~ction of lJicro\v:ive .$ntcnnt~s \vitll Dielectric Sheets, A-o.
483-18, 31:ir. 1, 1946.
E. R. Steele, So,T]r I.;lmtriral .fspects of ]ficrowavc %rdwich Radome Design,
NO. 483-16, lIfly 9, 1!)45.
T. J. SIIrn, I,liz:ihctl~ I;vcril:lrtj Diclcrtric ( onst:mts nlld Loss Tangents of Radome
IIatcriak, X(). 483-2.5, Jan, 11, 1!M6.
J. S. lVhitc, ICC F(,r,lu(tiorl t,), Sl,iphor,,r l{:,[I(,ITIM, Xo. 48;3-3, Fch. 15, 1944.
524 ANTENNA INSTALLATION PROBLEMS [SEC. 14.6

Range.A radome does not transmit all the r-f energy incident upon
it but reflects and absorbs certain fractions of that energy. If sufficient
reflected r-f energy finds its way back down the feed line to the trans-
mitter, frequency pulling of the transmitter can result in total blanking
of the receiver or in the very severe range reduction illustrated in Fig.
14.7. Independent of this effect, an appreciable reduction in range can
result from attenuation of the transmitted signal, especially since attenua-
tion in the outgoing signal is repeated on the return signal. Radar range
is proportional to the fourth root of the transmitter power [cf. Eq. (1.17)],
so that range reduction is related to radome transmission by Eq. (1):

one_,ya.y transmission 10sS ( 70).

Range reduction ( Yo) = (1)
2

To pick large but not unusual values, if one-way attenuation due to

raciome wall reflection is 12 per cent and due to absorption is 8 per cent,
then the range will be reduced by 10 per cent.
Pattern. Reflection and absorption at the radome wall can reduce
the absolute value of the radiation pattern by the processes described
above. Due to a variation in effectiveness of these processes with dh-ec-
tion and to related mechanisms, the shape of the pattern as well as its
magnitude can be altered. Several mechanisms that have been guilty
of pattern distortion will be described because of the seriousness of this
effect in certain cases. Pattern distortion is especially objectionable in
the case of a pencil beam that is being used in a conical scan for precision
direction finding and in the case of a shaped beam that is being used for
uniform illumination of the ground.
Pattern distortions have been traced to a variety of causes. In some
cases the effects were barely appreciable, but in most cases they were
serious enough to require study and elimination. Some of the causative
mechanisms are listed below with brief descriptions.

1. In a streamlined radome the reduction in transmission due to

absorption and reflection varies considerably with azimuth angle
and can therefore produce a minor change in the shape of the
main lobe.
2. An antenna transmitting through a spherical portion of a radome
can experience trouble owing to focusing of reflections from the
radome wall. The focal point of these reflections may lie near
the focal point of the antenna reflector and act as a secondary
source. The beam produced by the focused rays from this
secondary source can combine in various ways with the original
beam to produce distorted patterns. This process is illustrated
in Fig. 14.9a.
SEC. 14.6] RELATION OF THE RADOME TO SYSTEM PERFORMANCE 525

--

Center of spherical
surface

Antenna
Oirection of lobe
i\\\\\\\\\\\\\\ from secondary source

(a)

Streamlined nose radome

/

--
--
~ ~ Original beam

Antenna3:
\
\
\
L Reflection lobe

(b)

Center line of fuselage

Antenna

r
-..
-.

%tlon of flight
~
Twice reflected
lobe
Streamlmed radome
Original beam

(c)
FIG, 14.9. Yfechaniams by which radome wall reflections can distort patterns: (a)
secondary source produced by focused reflections; (b) wide-angle lobe produced by reflec-
tion from the upper port,on of a radome; (c) wide-angle lobe produced by a double reflec.
tion involving the undersurface of the fuselage.
526 .4.VTEAN.4 fNSTALLA7lUN PROBLEMS [SEC. 146

3. With the geometry illustrated in Fig. 14.9b, excessive reflections

from the upper portion of the radome can produce a broad down-
ward lobe at a wide angle from the original beam. This lobe can
produce a strong ground return capable of obscuring aircraft
targets at the same range, as illustratedin Fig. 148. In case the
antenna beam is shaped for surface search, a spurious clo\\-nward
lobe \vill produce a very objectionable interference ripple in the
elevation pattern \rhen combined \vith the original beam.

Streamlined radome
7

Phase front reside radome

l:IG, 14.10 .Distort1or1 of the pl~ase fro]lt of tlm rad~ation fm, ~, an antenna by v;uial~lc
phase del~~-s in sharply rurved p<)rtions of the rwionlc.

4. .-l streamlined radome mounted on the undersurface of an aircraft

can produce a dowm]vard lobe by the process illustrated in Fig.
14.9c. lhis lobe ~vill have the objectionable features described
in the preceding para~raph.
5, l>hasedelays inthetransmittr( iradiati[jn ca~lsccl bythedielet,tric (Jf
the radomc Ivan vary appreciably over sharply curvccf portions
of the wall. Jvhen the antenna apertllrc spans such a portion
of the radome, the constant phase fronts in the near zone of the
antenna arc distorted upon transmission through the ~vall. The
result upon the pattrrn is a distortion of the main lobe. This
process is illustrated in Fig. 14.10.

.4 pparent Pattern .The discussion of the previous paragraphs referred

to the radiation pattern \vhich can be defined os the variation of radiation
intensity or reeei\-ing sensitivity ~t-itb angle for a fixed position of the
antenna. In distinction to the radiation pattern, the appm-ent pattern
from the radar obscr~rers vielvpoint (an I)e dctined in terms of the ret(lrn
signal from an cffcctivr point -sollrcc tar~rt as prrwnied on his ra(iar
s(,ope Jf-hilt the ant rnna is sranning. In partic~llar, the apparent pattern
can be defined as the extent along an angular coordinate in the rzdar
scope of the return from a point target. Ilecause of thelimited range in
spot intensity on the scope, the description of an apparent pattern con-
sists simply of a statement of the angular region on the scope coordinate
that is occupied by the target signal.
The apparent pattern depends upon the beamwidth of the radiation
pattern, but it depends on other factors as well. An important modifica-
tion of apparent pattern is introduced by a transmitter frequency pulling
which varies ~vith the scan angle of the antenna. This variation in
pulling arises from a variation in the phase or magnitude of the reflection
back in the r-f line toward the transmitter.
The modification in apparent pattern most pertinent to the subject
matter of this section is that \vhich arises as the result of an angular
variation in the reflection from the radome wall. This pulling effect can
be serious even in the case of a radome that has not appreciably attenu-
ated or distorted the transmitted beam. An antenna mounted slightly
off axis in a cylindrical radome and scanning in azimuth will be subject
to a reflection of constant magnitude but slowly changing phase. A
streamlined radome can introduce large variations of magnitude as ~vell
as more abrupt changes in phase of the reflected \vave. The most
abrupt changes in both magnitude and phase are encountered in the
case of a relatively thin walled radome with internal reinforcing ribs.
Serious operational effects can be traced to a variable frequency pulling
by radome reflections. The resulting change in the apparent pattern,
even though not accompanied by a noticeable reduction in range, can
affect the apparent direction of the target in sufficient amount to ruin
the performance of a radar employed in precision direction finding. This
effect is especially serious, since it is not revealed by careful measurements
of the radiation pattern.
Automatic Frequency Control.No discussion of radomes and system
performance is complete without mention of automatic frequency con-
trol (AFC), since its use profoundly modifies radome design requirements.
AFC causes the local oscillator frequency to follow the transmitter fre-
quency in such a \\-ayas to maintain bet\veen them a constant difference,
the i-f frequency. AFC is strong enough in the case of a relatively
stable magnetron to follow larger frequency changes than those ordinarily
produced by radomes. Ho~vever, the AFC circuit involves a time con-
stant that may be considerably longer thfin the interval required for
the change in the radome reflection. .%n extreme example of a rapid
change in radome reflection is that ~~hich occurs as the antenna s~veeps
by an internal supporting rib in the radornc wall. The AFC \vill not,
follo\v ra;)idly enough to correct for the c}langing reflection and, in addi-
tion, \vill not return to the original fre(lllrncy setting until long after
tI)e radome ref!r(tion lMS rrtllrned to its ori~in:il vall]e. In this case
528 AiVTl?.VAA INSTALLATION I1{O13LI<II{,Y [SEC. 147

the frequency shift has been extended in time rather than reduced by
AFC. This is not to be interpreted as an argument against the use of
AFC but rather against the use of localized supporting ribs in radome
design. In general, AFC changes the radome design problem from one
of reducing reflection to one of reducing rate of change of reflection.
Claswjication oj Radomes.The system performance effects described
in this section are not all present simultaneously for a given radome,
but some are emphasized and some minimized according to the type of
radome considered. For this reason it is convenient to classify radomes
under several headings and to follow a different design procedure in each
case. A natural classification on the basis of use provides three general
groups: pressurizing seals for antenna feeds, beacon housings, and hous-
ingsfor scanning antennas. The first group, pressurizing seals, will not
be discussed here. For the other two it is convenient to make a further
subdivision into cylindrical radomes and streamlined radomes, a dis-
tinction that is of considerable importance in practical work. Not
only are the equations for radomc wall design different in the two
cases, but the underlying objectives are altered. The objective in the
case of cylindrical radomes (normal incidence) is to reduce the reflection,
but in the case of streamlined radomes it is primarily to improve trans-
mission. After presenting the quantitative considerations of the next
section, it will be possible to state this distinction more precisely.
14.7. Radome Wall Design.An ideal raclome \vall would completely
transmit an incident electromagnetic wave \vith neither reflection nor
absorption, and for such a radome wall the deleterious effects associated
~vith one or the other of these causes would be eliminated. Although this
ideal situation cannot be attained in practice, the transmission can be
maximized or the reflection minimized, depending upon the type of
radome under consideration.
The calculation of transmission and reflection by the radome wall is
most conveniently carried out in terms of transmission and reflection
coefficients. }Vhereas s(lc1l coefficients cannot be defined suitably for a
curved surface, a curved surface can be replaced by a plane sheet for
approximate analysis if the radius of curvature is IarKe. The complex
field distribution of the antenna that provides the incident wave is also
rcplaccd by a plane wave of uniform amplitude. The investigation is
thus based on the coefficients that exist and are easily defined for the
simpler situation of plane sheets and uniform plane \vaves.
Ph,yszca.l Pr-inciplcs.-Rcflect ion of an elect romagnctic wa~e occllrs
only at a discontin~lity, that is, at the transition from one mrdillm to
another. F:ver,v radome lvall l\-ithollt cxcrption may br regarded as L
set of pairs of s(lch disco ntinllities. The over-all reflection ~vill reslllt
from superposition of the in(livi(]{lal rcf~cctions; its nu~gnit(l(lr \\ill be
dctrrmille(l 1).y fheir nlagt)ll Il(lw :Ln(l r(~l:lti~,(,])II:LSPS. I{cflecti,)ns c:m
SEC:. 147] RA DOME WALL DhsIGN 529

be reduced or transmission increased either by reducing each of the

individual reflections or by adjusting the spacing between the discon-
tinuitiesto obtain partial or complete phase cancellation.
In Fig. 14.11 a generalized radome wall is represented as a pair of
discontinuities, or sheets. Indi~idual transmission and reflection coeffi-
cients for sheets 1 and 2 and over-all transmission and reflection coeffi-
cients for the wall are defined in the figure. In terms of these coefficients,
the advantage to be gained from minimizing the individual reflections
and adjusting their phases for cancellation can be expressed by the
relations
Shaet 1 Sheet2
[Rim.. = 17-,1+ Ir,l, (2)
and
11?[,,.= Irl Irzl; (3) 1 1
these neglect higher-order interaction
terms and absorption. rl tl rz t2
A precise phase cancellation of the
component reflections requires an ac- s
curate spacing between their sources.
*+
Since deviations from this optimum
spacing must be allowed in manu-
facture, the effect of such deviations
on transmission and reflection must
be investigated. The investigation
can be carried out by dividing the ~vall
configuration into two groups and by
expressing the over-all transmission Radome wall
and reflection in terms of the indi- l<IG. 1411.-Generalized ra do ,)) e
wall showlllg chvisio!l into two groups or
vidual transmission and reflection co- sheet. whlrh are treated as separate
efficients of the two groups. In the sources of reflection, The figure serves
to define symbols used in the text for
case of a single, uniform sheet, the t\vo reflection and transmission coefficients.
groups would be the two air-dielectric
interfaces of the sheet. In the case of a double-wall configuration each
group would represent the total effect of one of the ~valls. The method
can be extended to more complicated configurations. In any case the
nomenclature of Fig. 14.11 applies.
As the geometrical spacing s betlveen the sheets of Fig. 14.11 is
~aried, the over-all transmission coefficient T will reach the maximum
,,ralue
ltd,~
I Tl:.x =
(1 1,1?-21)2
for the optimum spacing s, given by

.s0=$:(41 +4?); )/ = 0,1 ,2,..., (5)

!530 ANTENNA INSTALLATION llWBLEMLS [SEC.147

where x is the wavelength in the region between the sheets and +] and
d, are the ~hases of r-, and r-jrespectively. If 6 = sO s is the deviation
from optimum spacing, the ovei-all transmission will be

~[, = tltzlz
(6)
1 21r,r21cm:: + \T1T212

The cm-responding over-all reflection, for the case of zero absorption, is

IT-II 21r,r-,1cm ~b + ]r,l

A
IR12= (7)

1 21r1rzl cos ~~ + lr,r2i2

From these equations nearly all the results needed for radome wall design
may be obtained directly.
1.0 0.9
9 :0.7
8
7 0.4
6 0.3
5
0.2
0.14 z

0,10 g
.-
0.07 ~
0.050 5
0,040x
0.030:
.
34 0,020:
0 &
3 0.014

2 0.010

0,01
0.(
Spacing error, wavelengths (a/h)
FI~, 14.12.Over-a1l reflection roe firient (R) for a symmetrical roufiwlratinll Of zero
absorption plotted as a fu,lction of the err<)r irl s~>arill~ (6) aIId the ,,lrl,~-iduxl reflert)on
coefficient (r).

The symmetrical radosne \vall consisting of identical groul)s of approxi-

mately zero absorption spaced in accordance \\-it h E(I. (5) \vill bc mcolln-
tered almost invariably in practice. For thr symmetrical config~lrati{ln,
t, = tz = t, rl = Tz = r, 4, = OZ = 1#1and Eqs. (4), (5), ((i), and [7)
SE(,. 14.7] R.4 D(),lf E JIA 1,1. DE,71GN 53 I

assume simplified forms ~rhich will not be repeated here. The expression
for the over-all reflection can be simplified fllrther to

(8)

for small 6lk, or alternatively to

(9)

when [rlis small enough to be neglected. The exact value of Ili?lfrom

Eq. (7) is plotted in Fig, ]4.12 for the symmetrical configuration.
Single WalZ.-The simplest radome wall design is a single, uniform
sheet of dielectric material. In this case the two sources of reflection
treated above and illustrated in Fig. 14.11 are the two air-dielectric
interfaces. If the specific inductive capacity k. of the material is denoted
by p, the over-all reflection is given by Eq. (7) with

(10)

Equation (10) is plotted in Fig. 1413. The optimum sparing is

n~ 71A0
n=o, 1,2, ..., (11)
0=~=2vj

since 01 = f#J2= O; k. is the frec-

space wavelength.
The two design procedures
mentioned abovereduction of r
or adjustment of phasepermit of
simple interpretation for single-
walled radomes. The first
method indicates the use of a lo\r-
dielectric-constant material to re-
duce r [Eq. (10)]. The second
method requires the use of a thin
sheet or in general one that satis-
fies Eq. (11). If a thin sheet is
used, the tolerance 8 in the equa- Dlelectrlcconstant(~)
tions above stands for the thick- FJG. 14 1:3,The reflection coefficient (r)
of an :tir-d~electrlr illterf arc as a fllrlrtion ot
ness itself, The single-wall con-
the spevIfi, i,,<]urt,ve capti.ity (~)
st ru c t ion becomes increasingly
attractive at longer wavelengths where a small ~alue of 6~Acan be realized
vith a wall thickness sufficient for mechanical strength,
.532 A 7VTRIVNA INSTA I,J,A TI().V llTOBI,l<:Jf,S [s,:(,. 147

The assumption of zero absorption loss, even in the case of actual

radome materials, does not invalidate the results for reflection and opti-
mum spacing, but the results for transmission must be modified appro-
priately. The general equation for a single sheet is quite complicated
except when the thickness s~tisfies Eq. (11). For a half-\\-avelcmgth
thickness of relatively lo\v-loss dielectric

T=]-xv%); (12)

L is the loss tangent ~~ for the matcri:il ~vhcrc c and e am the red
and imaginary components of the complex dielectric constant (see ,Sec.

10
0.9

08
~-
~07 1>
0.0018
:
;06 0018

s 0036
:
:05 0072
:
% 0108
o
04
0,144

018
03
00 01 02 03 0.4 05 06 07
Thickness, wavelengths (d/h )

11(;.14.14. LVLI1lM of 10ss

3.2). The rxact values of the prover transmission 11] as a function of

thickness forl-arious values of L is givenin Fi~. 11.14 for the partic~dar
specific indllctive capfieity S = 4.
The sin~l(,-~t-all design is usually not practicable at microwave fr(~-
quencies. lhe half-lrav~lengt,h IIall is freqllently toollc:~vyfc)rairt){)rne
installtitions, ~thercas the ]fall thin ~noug$ to have good electrictil prop-
erties is not satisfactory mechanically.
Sardwich Wdl.-.i major-improvement both mechaniral]y and elec-
trically is realized by substit~ltin% s:ind!t-i(11-tl->illfor single-l~all design.
Thesandnich-\vall dcsigncf)mistsof ashcct of lo\v-den.sity core materia]
fac~don bothsidcswiththinj high-density skins. f+l:ln(]l~-i(,ll const,ru~t,i~~n
has seen considerable use in purely mechanical imtallations \rherc the
strengtl~-to-~~eight ratio m[wt be hixh. Fmrn the electrical point of
view, the skins sho\r lo\\-reflection beca~we they am thin, :md the core,
bccauseit has a!oI\ dielectric constant.. l~llrtll(,rmore the core provides
a means of accurately holding the skin separation to a value favorfil)le
formmrelirrg reflections.
SW. 14.7] RADOME WALL DESIGN 533
~
A double-wall construction which is obtained by omitting the core
material from between the skins loses most of the advantages of the
sandwich construction. It has been used, however, since it provides
means for deicing by circulating hot air between the skins of the radome
wall. The optimum spacing of the skins for reducing reflection when
separated by air is given in terms of their thickness d and specific induc-
ti~e capacity a by the approximate re-
Skin Skin
lation

s, g (Zn + 1) > d >/Q. (13)

\\ :<< \\j

-1
\\p \\ \a Dielectric
The over-all reflection is nearly propor- \\\ ,1 ,i,constants
.\\ \\\
tional to the spacing error and to the

\\\\
.\ \\\\\
reflection from a single sheet when these Geometrical
<\s~~ \;, ~ dimen~ion~

---n-
are small and is given by Eqs. (7), (8), \\\;
and (9). The power transmission is *S D Electrical
given by Eq. (6) in the general case and \\\\ >\~/ dimensions

by 1 R2 when the loss is zero.

The general sandwich wall is shown
in Fig. 14.15 which serves to define some
of the symbols to be used in the discus- Sandwich wall
sion. The electrical thicknesses are TheFIG, 14.15.Sandwich-wall design.
figure serves to define s~mbols
D = d ~a/~OandS = s @/Ao. The used in the text,
individual reflection coefficient is now
that for the three-medium transition from the air through the skin to the
core. It is given by

]r]z = A k ~j (14)
A + 4a <P
where
A = (a+ f?)(a + 1) (d @)(a 1) COS(4mD). (14a)

The value of r obtained from Eq. (14) can be used in Fig. 14.12 together
with the appropriate value of the spacing tolerance to determine the
over-all reflection coefficient. The optimum electrical spacing is similarly
found to be

2 =P (a 1) sin (~~D)
SO = ~ & tanl
(6 ~)(~ + 1) + (a + P)(a 1) Cos (47rD) (15)
Under certain conditions this equation can be approximated by

(16)

The range of validity of Eq. (16) is indicated in Fig. 14.16 which is a

534 A.VTENFA I~YSYALLA YIO,V PROBLEMS [SEC. 14.7

typical family of curves calculated from Eq, (15) for a fixed value of a
and a range of values of D.
In certain methods of sandwich-wall fabrication the skins are glued
rather than molded to the core. The layer of glue introduced by this
process effectively increases the skin thickness and introduces appreciable
error into the calculations. This concept of an effective skin thickness
which includes the effect of the glue is an approximation that ceases to
be valid when the angle of incidence is variable or when both tolerance
0.50

-.
~
g
G 0.40
g
$
3
%
E
%
g 0.30
g
o

0.20
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09
SkirI
thickness, wavelengths (D)
FIG. 14.16. -optimum core thickness as a function of skin thickness for ~=4andfj
ranging from 1.0 to 2.0.

and spacing are involved. It has pro~-ed very useful and is entirely justi-
fied when one is concerned with optimum spacing alone at a fixed angle
of incidence.
Arbitrary Incidence.The results derived hitherto apply to normal
incidence only. They may be used for arbitrary incidence at either
polarization, however, if the quantities D, S, l~o, a, and ,6 are replaced
by appropriate quantities as given in Table 14.1. Similar equivalence
relations for arbitrary 10SSwill not be given here. Limiting values for
TABLE 141.-EQUIVALENCE R~LATIONS FOR CONVERTING FROM NORMAL TO ARBI-
TR.+RY INcIIIF:xc~ VALt-~S
P = sin%: Oo = angle of incidence measared from normal

Normal incidence quantities D s so a B

Arbitrary incidence valaes,

perpendicular polarization. :x/m_p ? >= y~/j_p < op

;A lp
Arbitrary incidence YZIUCS, I
parallel polarization. ~~ ~/~~~ ~~ \ ~ N -az(l p) (P(l p)
V?-Pj~
ix PP
I
grazing. incidence, simplified forms at llre~~sters angle, and approximate
relations for thin skins, though useful for computation, illustrate no new
principles and will not be included in the present discussion.
Considerations up to this point have been based upon the assumption
of a single angle of incidence~rhethcr normal or otherwise-and a single
plane of polarization for the incident \vavc. It \vill be shown in Sec.
149 that a given portion of the wall of a streamlined radome will be
required to transmit for a range of angles of incidence and for a range of
polarizations. If all portions of the radome wall are considered, the
ranges involved will be correspondingly increased. In the interest of
simplicity in fabrication it is very desirable to usc the same \vallstructure
throughout the radome. The rmldting problem, therefore, is to find a
>vall design that will be satisfactory for a ranqc of angles of incidence and
for both polarizations,
To this end the equations for arbitrary incidence are plotted as shown
in Fig, 14.17 \vhere contours of constant reflection for a sandwich wall
are presented as a function of core thickness and angle of incidence.
Similar charts can be used for single-\vall construction. With such
diagr~ms the optimum thickness is readily detcrrnincd for a specified
range of 6and for a specified tolerancc in thickness; the thickness referred
to here is the total \vall thickness of the single-wall design or the core
thickness of the sand\\-ichdesign,
The represent ation for the single wall requires only three variables;
the angle of incidence, the dielectric constant, and the thickness in ~vave-
lengths. A single series of chw-ts for various dielectric constants there-
fore provides complete information for single-ivall design ~vork. In the
case of the sandtvich ~rall, five ~-ariahlcs arc required: skin diclect,ric
constant and thickness in wavelengths, core dielectric constant and
thickness in ~vavelengths, and angle of incidence, .1 very large number
of charts are required; in practice percentage reflection contours are
plotted against core thickness and angle of incidence, each chart, being
for specific Talues of skin thickness, skin (, and core c.
J[echanical Reql{ircmr~l fs-llcsidrs satisfying the electrical require-
ments t}lat form the main subject of the present chapter, a radome must
satisfy certain mechanical rmluircrrmnts as \rell. An airborne radome
must \rithstand the distril)~lted load prodllce(l by windage; it must not,
ddorrn }vhen a concentrated load is applied as is necessary in installation
procedures; and it must often stand the impact of rocks and water in
lundings. To these mechanical rmluircmcnts must be added certain
requirements in physical propert it, s; e.g., the radorne should not be
soluble in ~asoline or in any other solvent likely to he brought in contact
\vith it; it shoilld \\ithstan(l high trmpcrat[lrcs ]t-ithout softening, lo~y
temperat ilrvs \\ltho~lt becoming I)rittle; it must not absorb moistllre to
an,v appreciable extent ; and in certain (,ases ii m~lst be provided \\ith
 2.0

1.8

1.6

1.4

.E 1.2
:
~ 1.0
5
2
g 0.6

0,6

0.4

0.2

0
00 15 300 450 600 750 900
Angleof incidence Angle of incidence
(a)
Fr<;. 14.17 .- Typird rh:wts showing contours of constant, refktion
(b)
as a iunctio,, of row tILirknc.ssand anEIc of incidet, <,efor the f<>llo,vir,gcon.
clit,on.: A,, = 3.2, n,; cure //eO = 1.4; skin e{/eu = 3,7; skin t h,rk,,e~s = 0.050 ill. (u) Folmiz:,tion perpendicular to the plm,e of incidence; (b)
parallel wjlariza~ ion.
SEC. 14.8] .VOILVAI, I.VCIDE.VCE RA IJOME.Y 537

means for removing ice deposits as they are formed. Some of these
requirements are clearly of the utmost import ante; part imllarly i n air-
borne systems, failure of the radome can have very serious c(msequences
and should be avoided m-en at the expense of electrical performance.
It is this simultaneous presence of mechanical and electrical reqllire-
ments, which must, be satisfied ~vithout excmsi~e complication in
manufacture, that constitutes the radome problcm. These mechanical
requirements are met by m~thods standard in the aircraft industry; the
significant change introduced by electrical requirements is the restriction
to suitable materials and to suitable relations, as determined above, for the
linear dimensions, lYithout gi~ing a detailed discussion of fabrication
{Jrmechanical design techniques, it is therefore sufficient here to observe
their grrat importance.
14.8. Normal Incidence Radomes.There ha~e been earlier allusions
in %cs. 14.4, 145, and 146 to the fact that the most serious problem with
normal incidence (cylindrical) radomes is the resulting impedance
mismatch at the transmitter, \vhile the most serious problem with stream-
lined radomes is the resulting attenuation and distortion in the trans-
mitted pattern. These problems and procedures for solving them will
be considered further in this and the following section. First considera-
tion \vill be given to the pattern and impedance effects encountered in
nf)rnlal incidence radomm.
Pat/m-r~ ~~~~rc(s.-~hc effect of a cylindrical radome on antenna gain
(an be minimized in a strai~htforlvard manner by use of the normal
incidence relations of Sec. 14.7. In case the radorne \vallis uniform, this
prnce(lllre also minimizes the effect on pattern, since it leads to small
.,
reflection as \vrll as to large transmission. There is another source for
pattern distortion, honm-crj even in a radonw having complrtc trans-
mission. .4 radorne \vall having a structural rib, overlap, or but [strap
could he designed for complete transmission throll~h both the thin part
and the thick part of the wall. ll~hercas this wolLld maintain the ampli-
tude of the transmitted ~vave unchanged, it \\-onldnot compensate for
the change in phasr. A double thickness se(tion of \vall ~~ould introd~lce
t i~ice the phase delay and so distort the transmitted phase front. This
phasp distortion can become so serious that the presence of a thick clielec-
tric rib may be more harmful than a metal rib of the same dimensions.
4 similar effect introduced by sharply cllr~-ed surfaces is ill~lstratcd in
Fig. 14.10. Tf discontinllities are avoided and tbc reflection is minimize(l,
[hen a cylindrical radome will not ha~-e an appreciable effect upon the
antenna pattern
Impcdanrc E,~~mts-The problem of antenna mismatch and trans-
mitter pulling is milch more seriolls for normal incidence radomcs. For
quantitative ronsidrration of the rase of a slightly tilted antrnna in a
cylindrical raclome, the amplit~lcle of thr reflection back into thr line may
538 A.VTE.VNA I,VSTALLA TIO.V PRoBLEMS [SEC. 14.8

be written
pL = 7R ~~P ~-(,I.,)
(+/@), (17)
19CL

where the symbols have the following definitions:

over-all reflection coefficient for the radome \vall flattened into a
plane sheet
free-space ~vavelength
radius of the cylindrical rarfome
radius of the antcnnti aperture
angle between axis of the pencil beam and a line normal to the
generator of the cylinder
half-power width of the antenna beam
12quation (17) is valid only for small values of @j but this is obviously
the condition for maximum reflect ion into the fine and therefore the
condition of greatest interest.
If the antenna has a sharp vertical beam fixed at zero elevation, p,,
can be greatly reduced by the use of a truncated cone instead of a cylinder
for the radome shape. In general it must be assumed that the sranning
range ~vill cause the maximum reflection to be attained; furthermore it
must be assumed that the phase of the reflection will vary by at least a
half wavelength. Under these conditions the frequency pulling of the
transmitter will be
Af = 5P~(P.j.), (18)

~~here the symbols have the following definitions:

A! = frequency pulling in megacycles
of the transmitter
P~ value from Eq. (17) ~vith the exponential factor set equal to unity
=

p.~. = the pulling figure of the transmitter.

A modified in~-estimation is required if the radar system is equipped
\vith .LFd. The problem is then to estimate the rate of change of fre-
quency due to transmitter pu]]ing. This involves the use of the Ilicke
diagram for the transmitter, results from Eq. (17), similar results for the
phase of the reflection, and a time factor introduced by the scan mt c. ~
These calculations become important in estimating the apparent shift in
the direction of a pencil beam or in the crossover point of a lobe s\vitching
beam.
Design Considerations.The considerations bearing on the design of
c,vlindrical radomes will be summarized in several quasi-chronological
steps. For this purpose it is assumed that the designer is equipped with
1llh~pll]]in~fig,Irc of ,2 magnc?ron h;Ls hem defmcd as the! maximum freq[lcncy
shift in ]llcg:l{,y{, 1(,s from the initial fr: -lucncj-, which can lx, induccrl by % (T-SIVR)
1.5 of :ir},itr:lry PII:IW.
z I{(ff{(rri$jj,!tf
m k,rlflin<>e{ing, Vol. 1, and l?adur ~ca?Lners and iiadomes, Vo]. 26,
I{adiation L:iboratory Series.
sEC. 14.8] AORMA L I:VCII)ENC17 It. t I)(7:lf It:,$ 539

considerable information regarding the dielectric constants and [0sS tan-

gents of materials known to be available commercially and to be suitable
for fabrication purposes. It is assumed also that he is equipped with
radome wall design charts sho\ving the proper dimensional relationshil>s
for different wall designs, dielectric constants, losses, and incident angles
and polarizations.

1. A background of pertinent information must be obtained from

several sources: tactical application of the systcm; r-f requirements
of the system including allowable maximum mismatch and rate
of change of mismatch; information regarding the antenna, such
as size, angle and rate of scan, impedance match, far-field pattern,
near-field ray diagram; mechanical requirements of the actual
installation including method of mounting, windage] shock loading,
chemical exposure, \veather conditions.
2. The most favorable wall design must be selected in the Iight of
available materials and on the basis of electrical and mechanical
requirements from among the several possibilities: thin \vall of
arbitrary dielectric constant, lowdielectri(~-constzmt material (f
arbitrary thickness, half-ivave thick ]vall, double v-all, sandwich
construction.
3. The specific materials to be used in the selcctcd wall design must
be determined. This invol}-w using the dielectric constants and
loss tangents of the several alternative matrrials and the design
charts for the radomc \vall to c~iculatc thr transmission and reflec-
tion that would result from the rhoice of each material .~ com-
parison can then be made bet\~een these r-f performance figures
and the r-f requirements ancl betlveen similar mechanical per-
formance figures and mechanical requirements to a,rrivc at, the
final choice of materials.
4. A number of flat panels must be fabricated using the materials
and dimensions selected. These dimensions and the mrthod of
fabrication may be varied some\vhat to obtain optimum perform-
ance. The panels are tested for all properties dmmcd rclevaut
in the radome under design, e.g., structural strength, moisture
absorption, temperature resistance, transmission, and reflection.
5. A complete radome must be fabricated on the basis of results
obtained in the panel tests. Final tests are made of those char-
acteristics requiring the entire radome, e.g., strains due to con-
tinuous loading and effects on the antenna pattern.
In the sequence of operations outlined abo~-c the less expensive and
time-consuming operations are rarrird out first, ~rhereas the major uncfer-
taking of building the complete radome is delayed until all preliminary
problems have been resolved.
540 A .V lE~VNA IA-STALLA IIO,V PROBLEMS [SEC, 149

14.9. Streamlined Radomes.The cylindrical radomes discussed in

the previous section are suitable for ground and ship installations and
have been used in the past on slow aircraft. For modern airborne instal-
lations, however, it is necessary for the radome to be severely streamlined,
preferably to the extent of being completely incorporated into the original
airframe. These streamlined radomes present problems quite different
from those discussed above; antenna gain and pattern require detailed
investigation, \vhereas the problem of transmitter pulling assumes
secondary importance. Both favorable and unfavorable aspects stem
from the wide angles of incidence usually encountered. A nonscanning
antenna, such as a linear-array beacon antenna, is free of transmitter
pulling because of the fixed relationship of antenna to radome.

Uniform wall, Low-density filler ~ . Low-density filler

@&&ion,
High. density skins x High.density skins>>

(a) (b) (c)

lIti. 14.18. Beacon radon~e designs: (u) uuiform wall; (1,) solid wall; (c) perforated solid
culu,tructmt,.

Iattwn Efiecfs.-The pattern effects caused by streamlined radomw

for beacon antennas can best be discussed in terms of the several beacon
radome designs illustrated in Fig. 1418, In transmission through
sharply curved portions of the uniform radome ~vall shonm in Fig. 14.18a,
the cylindrical phase front is appreciably distorted by variable phase
delays. This mechanism is similar to that illustrated for a plane phase
front in Fig. 14.10. The pattern is further distorted in the case of vertical
polarization by large reflections from the tapered portion of the radome
wall. The use of high-density material in the streamlined shape can
be confined to a single thin skin if a low-density filler is used. This
design (Fig. 14.18b) reduces the reflections below those resulting from
the uniform wall but increases the variations in phase delay. A system
of perforations (Fig. 14. 18c) determined partly by experiment reduces
these variations in phase delay to such a degree that the azimuth pattern
is not seriously affected by the radome. Amplitude reduction in the
transmitted wave is present in sufficient amount to cause appreciable
SEC. 144)] STREA ML I,VED RA1x2MES 541

distortion in the pattern but is not nearly so serious as the other effects
just discussed.
The pattern effects to be expected from a proposed streamlined
radome design for a scanning antenna can be predicted qualitatively by
drawing elevation and plan views of the antenna with its ray diagram
and located in the radome. This type of dra\ring is illustrated schemat-
ically in Fig. 14.19 of a streamlined belly radome and an antenna employ-

Antenna

Axis of
rotation

of flight
(a)
Antenna

shaped beam

(b)

(c)

ing a shaped cylindrical reflector. The rtiys from the antenna are incident
upon the radomc Ivan under a \vide ~ariety of conditions, as can he swm
by a study of the figlu-e. The angle of incidence ~-aries from 0 for
steeply deflected ra~% to almost 90 for near-h(]rizontal rays I(mking
hack!vard. One polarization ~fith respect to the plane of incidrnce ~fill
prevail over the bottom of the radome, the opposite over the sides, with
intermediate polarizations in intermediate directions. A specific por-
tion of the radome may be required to trunsmit rays haying a wide ran~e
of angles of incidrmcc at each (Jf tlvo opposite po]arizati(ms; this can be
seen I)y studying parts (b) and (c) of thr fi~ure \\illl the :antrmna sta-
tionary ah s}l(jf[-n or kjy examining a porti[m of the sidr wall of tllr ra(lomc
in Imrt (u.] of the figllre \vith the antenna scanning. The pr(wedllres to
542 >!iv7Ki\lv.4 IA-S 7.4[,[,.471().V IltollIl?.Jr,? [SEC. 149

be followed in radome wall design to meet situati:ms of this type were

discussed in Sec. 14.7 under Arbitrary Incidence,
Design considerations.It is possible now to outline in several steps
the considerations involved in the proredllrc of dmigning a streamlined
radome for a scanning antenna. In the early stages this procedure is
similar to that described for normal incidence radomps in Sec. 14.8.
The following description is based on the antenna-radorne system shown
in Fig. 14.19:

1. The major dimensions of the radome must be det erminrd. These

depend upon the size of the antenna, the size of the hole in the
fuselage, the amount of retraction of the antenna. ~hese last
t~vo items are interrelated, since a larger fuselage hole will permit
more retraction in the case of a slightly depressed beam. Amount
of roll and pitch of the airplane and presence or absence of antenna
stabilization are ako invol~-ed.
2. The shape of the radome must be established. The first approxi.
mation is based strictly upon aerodynamic considerations. This
shape is then used together with the antenna shape and ray diagram
to construct an accurately dimensioned figure of the type indicated
schematically in Fig. 14.19. The radomc shape is modified to
keep angles of incidence belo\v some maximum figure, say 700,
but subject to approval from aerodynamic considerations.
3. The radome wall design must be calculated. Further study of the
antenna-radome diagram is necessary to fix the ranges of incidence
angles at both polarizations. The wall must be designed to meet
the system mismatch requirements by keeping dmin reflections
at small angles of incidence and to meet pattern and gain require-
ments by keeping do\rn reflection and absorption at large anglrs
of incidence.

Furt her steps, as in the case of cylindrical radomcs, are construction

and testing of flat panels of the radome \rall and construction and testing
of the complete radome. l~cchanical as ~vell as electrical tests are made
on the panels, while tests on the complete radome include aerodynamic
performance and effect on antenna patterns.
I
 CHA1TER 15
ANTENNA MEASUREMENTS-TECHNIQUES
BY H. KRUTTER

15.1. Introduction.lhe principles and techniques of antenna design

were developed in the preceding chapters without consideration of the
methods for obtaining design data and for testing the performance of the
completed antenna. This and the fullo\\ing chapter will be devoted to
a discussion of measurement techniques and a survey of the equipment
required for such measurements.
The antenna characteristics to be measured fall into four groups:
impedance, primary feed patterns, secondary patterns, and gain. The
impedance measurement techniques differ little in detail from those for
other r-f components of microwave systems; the problem is complicated
to a small degree by the fact that the antenna is a radiating load. The
import ante of the primary feed pattern increases \vith the progress that
is made in reducing antenna design from an empirical to a calculable
procedure. The study of pencil beams and fanned beams has shomm the
need for a detailed knowledge of both the phase and intensity distribu-
tion in the primary feed pattern. The over-all characteristics of the
antenna are particularly sensitive to the phase characteristics of the feed.
It is ex-ident from (hap, 13 that the design of shaped-bearn antennas
would be decidedly limited in scope \~-ithout a complete kno~vledge of
this primary feed pattern.
The pattern of the antenna as a whole, referred to as the secondary
pattern, is taken partly in the course of the design and de~-elopmental
research and finally, of course, as a test of the antrnna performance.
A particlllar advantage of the mirro~~-ave region is that, on the one hand,
the secondary pattern ran be determined so as to be closely identical
with that of the mtpnna in frrw spare and, on thp other, by usc of full-
scale models, the distortion of the free-space pattern due to the installa-
tion and hollsin~ can be studied. s(,roncfar~ patterns of pencil-beam and
f anneal-beam ant cnnas are generally confined to the principal l;- and
H-planes; ho\\-e\-er,the importance of complete space patterns is being
recognized part it~llarly in regard to thc effect of cross po]arizat ion in
red~lcing the resolving poivm of the antenna beam. (ompletc space
patterns are, of CO(Irsc, al\\-:LyS
nercss:uy in the design of shaped-beam
antennas; here ag~~inthe polarization of the field m~lst be dctermine(i
complf>tf,ly in order to arrive at a correct evalllation of the antenna
performance.
543
 The range of a system for a given amount of available power is limited
by the gain of the antenna (Sec. 12). From the point of view oft echnique,
the direct measurement of gain is pm-haps more exacting than that of
other antenna measurements. A necessary complement to the instru-
mentation of an antenna laboratory is a set of primary and secondary
gain standards.
IMPEDANCE MEASUREMENTS
15.2. Transmission-line Relations.The subject of two-wire lines
and the relation of waveguidcs thereto were treated in considerable
detail in Chaps. 2 and 7, respectively. For the sake of continuity of the
discussion in the present chapter the principal transmission-line relations
\~illbe revie\ved hrre briefly.
The voltage and current at apositionz along the line are

v(z) = Ale -r. + ..12p= (218)

1
i(z) = ~. (A1e~ A,(r), (219)

where y is the propagation constant, in general a complex number

y=a+.i?. (217)

The constants a and 6 are respertivcly the atten{latifm and phase con-
stants; the latter is related to the \vavelength on the line by

The amplitudes of the component ~vaves A 1and A ? are evaluated in terms

of the conditions at the two ends of the line. Thr origin z = o \\-ill[w
taken at the termination, and the input end of the line ]~ill br located
at z = L. If Z,, is the tcrmirmting (load) impedance, ~ve have by
13q, (223)
.1 2 Z,, zo= r,.
(1)
.1, = z,. + z;
~,, is the load reffertion rxwffirient tlcline(l by Ph. (2.27) on set tin~ z = (),
At the input, end of the ]ine z == /,, we h:LYe

v. = V(L) + zc;i(L), (2)

where V~ is the gynerator emf and Z,; is its int~,rnal impedance. .1 refle[-
{ion ctwffi(,ient ~,, m:ly l)e (Irfine(l (t)lrfsl)ol~(lil~~ to thv mismat (h Ix,tl! ren
the generator impedance Z,, and the c}lar:lcterist ic irnpukmce of the linf,:

Z(, z!.
r(, = (3,)
z,, + z,
SEC, 15.3] S1.4 ,V1)IATG- WA VE RATIOS 545

obtain

(4;

and

2= (a%),-r~:=. (5)

The voltage and current at any point in the line are therefore expressed
in terms of the sollrrrj the transmission line, and the load by

() (=)
z,+ z. 1 X2z -(+) (6)

(z) (z%)l~~:::~Le-(z+) 7)
The line has impedance transformation properties that are described
by Eq. (232), from ~~hich we may obtain the impedance at a point
z = z relative to the load impedance at z = O; i.e.,

V(1) .,. + ., tanh (71) .

= Z(1) = z, (8)
i(1) [ 20+ Z. tanh (7L)1

Similarly the reflection coefficient transforms along the line, and \vehave
from !@. (231 )
r(~) = rL@71. (9)

It is evident that .,, or r, can be determined by these relations from

measured values of .( 1) or r( 1) at any point along the line a distance
1 from the load.
For a Iossless line Eqs. (8) and (9) simplify to

(lo)
and
(11)

The relations for admittance (Y == 1 Z) are given by Eqs. (8) and (10)
with Z everywhere replaced I)y Y. Also ~. in terms of admittance is
given by
rL=Yo YL
(12)
}0 + Y.

15.3. Standing-wave Ratios. Instruments for voltage measurement

used in impedance determinations indicate some function (not neces-
sarily linear) of the time z~-erage of the square of the real voltage; this
is given in terms of the complex voltage by ~ VV*, ~\-hereV* is the complex
conjugate of V. Carrying out the indicated operation on Eq, (6), we
546 ANTENNA MEASUILEMENTS-T ECHNIQUIIS [SEC. 153

obtain
~VV* = ~[e 2[rLlc0S (292+ ~) + \IL\2e2=], (13)

where ~ is the phase angle of I~(r~ = lr~lei~).

The attenuation constant in microwave transmission lines is small
and may usually be neglected in impedance measurements of antennas.
The simplified expressions that follow are strictly true only for a lossless
line. Assuming that a = O, Eq. (13) simplifies to

;VV*=I 1 1 + zlr~l COS (2@Z + @ + lrL12]. (14)

T[

~ VV* is a periodic function of z with maxima occurring at 2pz + 6 equal

to even multiples of mand minima.at 2PZ + J equal to odd multiples of m.
Voltage maxima occur, therefore, every half wavelength, and minima
occur every half wavelength, with adjacent maxima and minima sepa-
rated by a quarter wavelength.
The ratio of maximum voltage squared to minimum voltage squared,
obtained from Eq. (14), is called the power standing-wave ratio and is

1 + lrL] 20
(15)
= [11 lrLl
The square root of r is referred to as the voltage standing-wave ratio and
is given by
1 + ]r.1.
(16)
=llrLl
Accordingly, the magnitude of the reflection coefficient of the load is
given by
Ir.1 = ~ (2.44b)

Following the argument of Sec. 2.7 it is observed that the reflection

coefficient is equal to Ir~] at a voltage maximum and to lr~lat a voltage
minimum. Correspondingly at a voltage maximum the impedance is
real and equal to rZo, and at a voltage minimum the impedance is real
and equal to Zo/r. If 1 is the distance from a voltage minimum to the
load terminals, the load impedance can be expressed in terms of r and
1 by replacing 2( 1) in Eq. (10) by Z,/r, thus obtaining
1 jr tan ~1
z. = z,
rjtan~l
or, separating Z~ into a resistive component R~ and reactive component
.X-L,
Zt, = R,L + .iiYL
r see=@ (r 1) tan /3/
= 20 p + tan2 @ ~~o Tz -~ tin! @l (17)
SEC. 154] ,LIEli.YLliEIJIE,Yl OF }0L1.4GEL YT.I, VDJ.YG-JVAIE KI17lU 547

The load impedance can thus be calculated by means of Eq. (17)

from the measured values of the standing-wave ratio and the distance
from a voltage minimum point to the load. The calculation may be
performed graphically on the reflection coefficient or bipolar charts as
was discussed in Sec. 2.8.
15.4. Measurement of Voltage Standing-wave Ratio.The most
common method of determining lS}YR is by means of the apparatus
Indicalor

7a&a~

! Tuner Attenuator
I & /: Space

/ Slotted section
P7

FIG.15.1.Blnchdiagrarlt of iIllI>cdallce-r,, casure,,lent apparatus.

shown schematically in Fig. t 5. t. The first unit is an r-f po\ver source

which at rnicrolvaves is usually a wlocity-rnodulated tube capable of
being tuned over the wavelength band on which measurements are to
be made. Since these tubes are sensitive to the imwdance mismatch.
an r-f tuner is generally connected close to the source and tuned for
maximum stable output. The tuner is then followed by a variable
attenuator which is prcferablv matched. The attenuator serves to
control the r-f power level and to reduce the pulling effect of a variable
load. .4 wavemeter should be in the set but should be detuncd dur-
ing standing-~vave measurements. The attenuator is follo~ved by a
slotted section of line; the slot is narro\v and cut in such a wa,v that it
does not interrupt the current lines in the wa~-eguidc wall appreciably.
The latter is necessary in order that the coupling between the line and
space be negligible; then the rharactcristic impedance of the slotted sw-
tion does not differ si~uifictintly from that of the uncut line. The field
inside the guide is explored by means of a probe mounted on the slotted
section. The microwave instruments generally emplo,v an electric-field
probea ~vire or needle entering into the guide parallel to the electric
\-ector of the field. Such a probe measures the voltage standing wa\c.
Nlounted in the probe is a detrrtor that suuplies
.. direct current or audio
frequency to an indicator. The response of tbc dctcctm-indicator com-
bination is a measure of the field intensity at the probe. The slotted
548 ANTENNA MEASUREMENTS-TECHNIQUES [SEC. 154

section is placed as close to the antenna as possible in order to reduce

errorsin measurement of electrical length.
The actual measurements consist of moving theprobe along the line
and determining the maximum response and minimum response which
should be separated by a distance of ko/4. From the calibrated response
the VSWRis obtained, The distance of a voltage minimum (the reason
for choosing a voltage minimum rather than a voltage maximum is
given later) from the load is noted. These two quantities together with
knowledge of the wavelength in the transmission line suffice to determine
the impedance of the load.
Precautions in Standing-wale Measurements.The procedure described
above is exactly the same as that used in measuring any r-f component,
except for one very important difference. The antenna is a radiating
load and therefore precautions must be taken so that reflections from
near-by walls or objects do not affect the measurements of VSWR or
phase (position of minimum or maximum voltage). To avoid or diminish
the effect of reflections the most intense portions of the radiation should
be directed toward an open space with as much open space in all other
directions as possilie. Measurements maybe taken inside the laboratory
by directing the main beam at an angle of approximately 45 toward a
wall, preferably of low reflection. To ensure that the space in which
measurements are being made is satisfactory, the antenna should be
moved to several positions (varying the distance to the wall and changing
the angle of incidence plus or minus a few degrees from 45) and the
effect on phase as well as VSWR observed. If no changes occur, the site is
satisfactory. Particular caution must be exercised when impedance
measurements on low gain, nondirective antennas such as beacons are to
be made. If such antennas are to be mounted on a metallic sheet in actual
use, impedance measurements should be made in such a way as to simulate
as closely as possible the actual final conditions.
Several other precautions must be taken to ensure the accuracy of
VSWR and phase measurements. The probe should be loosely coupled
to the line in order to avoid alteration in the standing wave pattern which
will occur if the probe has an appreciable reflection coefficient. The
reflection coefficient of a probe is a function of the tuning of the probe
as well as the probe insertion. The effects of this will be discussed in
more detail in Sec. 15.7. For a matched generator this reflection results
in apparent VSWR less than the true VSWR and asymmetry in the
standing-wave pattern with the maxima and minima not separated by a
quarter wavelength. The position of the minimum is not affected appre-
ciably. If r, is the reflection coefficient of a tuned probe when the
line is terminated in its characteristic impedance, the measured VSWR
corresponding to a load of reflection coefficient rL is less than the true
value by the factor (1 11~1Pl)/(1 + \rLrPl). Thus, for example, a
probe presenting a reflection coefficient r, = 0.05 v-ill measure a stand-
ing-wave ratio rz = 4 as I-2= 3.75. If the internal impedance of the
generator is different from the characteristic impedance of the line
(mismatched generator), these effects will be even more aggravated.
The probe mount should be tunable so as to obtain maximum output
for various frequermies and for a given probe insertion. Timing should
be smooth and not subject to erratic contact; the probe is generally a
fairly high Q device, and the response will be affected easily by inst abilit y.
Furthm-morej for a given probe insertion the reflection coefficient of the
probe \vill be real when the probe is tuned; under the latter condition no
asymmetry will result in the standing-~rave pattern although the meas-
ured TSWR \vill still be lms than the true 173WR.
The response of the probe and indicator must be calibrated. It is
not safe to assume that the response is proportional to the square of the
voltage, A crystal detecto~] is accurately square law only for ver,y
low r-f power levels. The Ia\vof the crystal varies from crystal to crystal;
for a given crystal it is a function of the power level and the load of the
indicating system. A current-biased bolorneter element, such as a T,ittel-
fuse, is very accurately square law over a very large range of power
levels except in the neighborhood of burnout. A simple and convenient
method for calibrating the detector \rill be given in Sec. 15$.
Irregularities in the standing-~rave pattern are frequently due to
structural defects in the slotted section such as erratic contact between
the probe carriage and the line. A common failing of coaxial-line sec-
tions is that the inner conductor is not acmrrately concentric with the
outer. The usual effect is a probe depth varying almost linearly along
the section resulting in a standing-wave pattern that, appears to be super-
imposed on a monotonic voltage; this effect is f-mown as slope. The
effect of slope can be compensated for by calculating the l-SIVR from the
ratio of the geometric mean of tlvo maximum values of response separated
by x,/2 to the minimum vahle between them or by taking a maximum
rrsponsc di vialed by the geometric mean of the t~vo minimum responses
on both sides of the maximum. Both procedures should give the sarr:e
result. Actually an arithmetic avm-a~r is satisfacto~y. The avw-apy of
t~~umaxima divided by the a\~cragcof two minima should not be used in
corrcctin~ fur slope Jlaximurn :mci minimum values shollld be taken
near the c~ntm of thc slot ted section s () as to avoid the edge effect at the
ends f)f the slotted section.
The impmlancc of the slotted section should be the sam~ as that of
111Pfccd lint of 1II(an! (nna bciug invcs( igal [xl and good elcc( ri(,al cent act
bctlvecn tI]c t\\o lines sho~ll(i occllr. l;or most accurate resllits the Iinfs
should LLISO lx; geometrically tllc same. For example, if tl~o 50-ohm
!350 ANTENNA ,Jffi:ilS URE.tfEiV7,~-TI< (! II,VIQIJR,< [SW.155
coaxial lines with an appreciable difference in the radii of their respective
conductors are connected together, the two lines v-ill not be matchwl
because of the capacitance introduced by the junction.
15.5. Determination of Electrical Length.In determining the
proper position for insertion of in~l)f,(iance-mat(hing devices, accurate
knowledge of the position of a voltage minimum point clme to the antenna
terminals is required. For good impedance matching, this information
frequently must be knovm to a higher degree of accuracy than the value
of VSWR. To reduce the error in determining the position of the voltage
minimum the standing-lvave section is always put as close to the antmma
as is physically possible.
loltage minima in the slotted section are first dr+crmined. llrter-
mination of the position of a minimum by adjustment of the probe posi-
tion so that the response is a minimum is inaccurate, since with normal
VSWR the minima are broad and the position is h~rd to determine
exactly. A more accurate method is to determine the position of the
probe for equal response on both sidw of the minimum; the a>-mage of
these positions will then be the location of the minimum. For greater
accuracy the average position for several different resJ>onses mav be
taken. Having located such a minimum position, it miiy be transf~rred
up the line an integral number (Jf h:Llf \~avclengths to a point, nrar the
antenna terminals. In actual practice, except for certain simple cases,
this is easier said than done.
Perhaps the best method for transferring thr posit ion of the minimum
is to short-circuit the transmission line at a point close to the :mtenna
terminals and to use this as a reference point. This mcthtxl, of cmlrsej
assumes that such a short circuit can be made. This \\-illlwlally be tnle
for experimental antennas, hut not usually vith prod(lction antennas.
(onsider as an illustration Fig. 152 \vhich sho\rs the experimental srt~lp
t,~r determining the voltage miniml~m near the tcrmimds of ~ coaxial-
line-fecf antenna. Let 11 be the position of a voltage minimllm in the
~!otted section in Fig. 15,2a. The feed is noit- removed at the fitting,
and a shorting plate is placed at the end of the cotixial line as in Fig. 152b.
A new voltage minimum (zero) is located at 1, on the load side of position
1,. Transferred up the line, this ne~v minimum is at the short circuit
or an integral number of half \vavelengths from the short circuit and
therefore provides a convenient reference point to Ivhich the load voltage
minimum may be referred, l[casl~ring a distan(e 111 1,1 to~vard the
generator ftom the shorting pltitc determines the loa(l voltage minimum
position relative to the short ciucuit.
For air-filled coa,xid lines, the m[a.sllred physical lcngt}ls :md elec-
trical lengths ~~illshow good a~reement, becallse the ~ra,-elength is inde-
pendent of variat iflns in iine inlpedanrc. 1Ioircver, in ~}avrgllidcs the
guide \va\,elength de,,:,llds on the dimensions of the ~llid~>;}lenee the ]lse
SEC. 15.5] DEIERMINA 11ON OF EI,ECTRICAL I,ENGTH 551

of an incorrect guide wavelength will lead Lo an error in the location of a

transferred voltage minimum. For example, suppose that ho were
actually 4.52 cm in the transmission line and a guide wavelength as
determined in the slotted section \rere 4..50 cm. If one wish~d to find
the position of the voltage minimum approximately 90cm (about20AQ)
from the minimum in the slotted section, then the error resulting from
theuseof thevalue 4.50cminstead of4.52cm \\'ol1ldbe2O XO.02 = 0.4
cm. Theshorting method disc(lssed intheprevious section \vould almost

(a)

I a
Padded
,
generator
Shorting
plate
\

(b)
/
FIG. 15.2.Short-rircuitwI line technique for drterminingel ertriral length

completely rancel this error. It shmlld br emphasized that this proce-

dure assumes that the connectors, bends, or small variations in impcdanm
in the transmission line are rrflectionlrss and therefore ca~lse no phas~
shift.
When the shorting method cannot be llswl, the problem becomes rnorp
difficult, and \-ariolls tricks may be used i~ith more or less accurac.v to
give the desired information. one method frequently used in ~vareguidc
matching problems is to have made a set of experimental inductive
irises of various openings, similar to that in Fig. 15.3, so that the outside
dimensions of the frtimr of the experirmmtal iris are such as to ensure a
snug fit inside: the Ii-aveguide with just sufficient clearanre to permit,
sliding. To a given prrcenta~e (Jpcn are:~ of the indllrtire iris there cor-
responds a definite mism:itch lvhich \\ill I)e eliminated by locating thr
iris in the proper position. Knofving the JSIVR of the load from meas-
urement, an iris whose open area is most nearly that required to cancel
the mismatch is chosen from the set and slid along inside the ~vaveguide
to a point in that portion of the lva~-cguide \vhere it is desired that the
matching transformer should be. The position of
heh~z+ the experimental iris is varied until a matched (or
nearly matched) condition is obtained, The posi-
tion of the iris so determined for a minimum JSWR
then determines the proper point for actual inser-
tion of an inductive iris. This method has been
---- found to give the proper position to ~vitbin a milli-
# . .
meter in 1 by ~-in. ~vareguide. The size of the
l~IG. 15.3.Trial iris,
actual iris to be insertrd is obtained from the
\S}VR of the load and kno\vlecfge of the inductive smx-eptance intro-
duced by an iris as a function of the iris opening.
.4nothcr method that may be used in ~vaveguides is to cut a small
hole in the center of the broad side of the \vaveguide near the antenna
terminals and to introdum a capaciti~-e screlr, The JSWR and position
of the minima are obtained for se~-eral screw depths. A ~-oltage mini-
mum position is assumed to bc at the screlv position, Referring to this
point, the admittance of each value of ISIVR and phase are plotted on
the admittance cfiagram. z It ~~illthen be observed that the points do not
fall along a vertical straight line ,orrmponding to capacitance being added
\vith increasing screlv depth. The points \vill all be on a straight
line J\-fmnthey have all been rotated the same phase angle on their
rrspectire r circles. In tbc examplr sho\vn in Fig, 154, rotating each
point 30 on its corresponding VSIVIt rirclc the points then fall on a
~ertical Straight line with added scrc\v inscrt,ion increasing the capacitance
in parallel. This then means for this example that the true \-oltagc
minimum is located a distance / = 301\720 toward the generator from the
position of the center of the scre\r. In practice, this method is subject
to error because there is, in addition to the capacit anre of the scre~v, a
small shunt conductance ~vhich increasrs with increasing dianlrter of
the scre~v. For small w~re~rdiameters and small scrc\v dc~]ths this
method gives an approximate positi(m for the minimum,
15.6. Calibration of Detection System..4s has been stated pre-
viously, the lalv of rrsponse of the detection systcrn must be kno~vn for
accurate mcasurem(,nt of ISIVR. I?an accurate attenuator is availah]c
and the polvcr level is knonm, the response as a fllnction of po]rer input,
can easily be drl ermined. Or a calibration is aI-ailable if thr resDonse
system can be Compared with a kno~~n response system. Both of these
methods hale been llsed. Ho\\-ever, thr metboci dist~lsse{] I)c101v is a
simple one, utilizes the apparatus alrra(ll, a~-ailablc for lSIWl meas~lrc-
Lwawguid? Handbook. (ITO1,10 of this series. )
z See Sec. 2.8.
SEC. 15.6] CALIBRATION OF DETECTION SYSTEM .553

ments, and can be carried out easily from time to time if a variation in the
law of the detection system is suspected (as may happen when a crystal
is being used).

1,0

0.5

____
~ w I
I -t
I
-- ,-
-200
_
WA

-0.5

-1.0
0 0.5 1.0 1.5 2.0 2,5
FIG. 15.4.Admittance diagram for capacitive screw. Increasing srrew depths correspond
to increasing capacitance in parallel.

Indicator

R-f source Wavemeter Probe

-@

t 4
Matched
Tuner variable , 1
attenuator
I Short
/ Slotted circuit
section
FIG, 15.5.Block diagram for calibration of detector.

Figure 155 shows a schematic setup for this experiment showing a

shorted standing-wave section being fed by a generator, the power level
of which is adjusted by a variable matched attenuator. Since with a
short circuit, the voltage is a sinusoidal function of position [Eq. (14)
with 11~1= 1, 6 = r] with period Au/2, the response 1 for the detection
554 AN TE,VNA MEASUREhfENTSTECHNIQ UES [SE(:. 1.5.6

system as read cm a meter will be given by

I = 10 COS
()
2iTl
~
v
I (18)

where the exponent n is the so-called law of the system and 1 is the
maximum response at 1 = O (note that 1 = O here refers to the position
of maximum response). The law of the detector may be a function of
power level, a not uncommon occurrence in the case of crystals. Equa-
tion (18) can be written
I
10$410
~ .
n= (18a)
loglo Cos 2:
9
It is advisable to plot the experimental values of Ioglo ~/~0 VS. log,O
cos 2irl/Aa in order to smooth out experimental irregularities. The slope
of the curve at any indicated level 1 is equal to n.
If n = 2, the system is a square-law system and the meter readings
are therefore proportional to the square of the voltage. The ratio of
maximum response to minimum response for a load then gives directly the
rz of the load. If n # 2, rz is given by

(maximum response)
(minimum response)~

if the law of the system is nl at the maximum reading and n~ at the mini-
mum reading. For low (VSWR) 2 the ratio of maximum to minimum
response may be used uncorrected if n does not differ greatly from 2.
For example, suppose that the meter reading were 15 and 10; then
(VSWR)2 would be written as 1.5; but if the Ia\v of the system at both
levels were 2.2, the true VSWR would be (1.5)222 = 1,45.
It is much simpler, when n does not vary greatly for different power
levels, to determine n experimentally over 3-db po\ver intervals. If
21 = d corresponds to the separation of half response, that is, when
1 = ~1~, then Eq. (18a) may be approximated by

~= 0.2206
(19)
d
0.1397
A.

The error in this approximation is less than 0.02 for the range

Figure 15.6 is a plot of n as given by Eq. ( 18a) as a function of the full

width d/& at half response. It should be noted that d A,, must be deter-
mined accuratei,v Iwca(lse an error of 0.01 in d A,, results in an error in j{
SEC. 15.6] CALIBRA llOAT W DETECTIOV SYSTEM 555

of 0.2. For small wavelengths a dial indicator calibrated in thousandths

of an inch is mounted on the standing-wave smtion for accurate measure-
ments of d,

3.2

3.0 \

2.8

2.6

2.4

2.2

2.0

\
1.8

1,6 -
\ \

1.4
0.20 0,22 0,24 0.26 0.28 0.30

%g-
~IG. 15.6.--~etector law IW. full wvdth dI h. at half response.

A possible procedure to be used is as follolvs: (1) Use the same

.nicroammeter or audio amplifier that \vill be used \vith the crystal or
bolometer in practice. ~his is necessar,v, since the apparent law of the
detector depends on the impedance of the metering arrangement as \vell
as on the r-f po\ver level. (2) .idjust line
the po~ver level by varying the
power level or by varying the probe insertion. The probe insertion sh(ndc!
be so small that the standing-~vave pattern is not disturtwd as it is moved
along. l%is may bc tested by monitoring \vith a second probe held at a
556 ANTENNA M.7ALYLrEEMEA TS-7XECH)T1Q C:E,Y [sm. 15.7

fixed position in a slot tcd section bet~veen the generator and the section
carrying the detector being calibrated. If the fixed probe reading is
independent of the position of the moving probe, the insrrtion depth of
the latter is acceptable. (3) lIeasurc the separation of thr points 3 db
down from the maximum for several maxima positions along the standing-
wave detector. Average these values of d. (4) he distance bet!reen
alternate minima will be the wavelength X. in the transmission line.
In the case of coaxial lines the wavelength is best ol]tained by m(ans of a
precision waverneter. (5) The average ~-alue of n for these conditions is
then determined from Fig. 156. (6) Determine n at other po~ver levels
by a similar procedure.
15.7. Probe Reflections.-The whole of the previous discussion in
regard to making standing-wave measurements by means of a probe in
a standing-wave detector is based on the assumption that the probe itself
does not affect the field being measured. If the probe has a reflection
coefficient different from zero, considerable error may result in the deter-
mination of the law of a detector or in the measurement of standing-~vave
ratios, es~ecially if the mismatch being measured is great. 1 The most
noticeable effect of probe reflections is a distortion of the observed stand-
ing-wave pattern.
For the most part, results that occur in standing-wave measurements
may be accounted for by assuming that the probe is a shunt admittance.
We may obtain the voltage V at any point of the line by Eq. (6) upon
replacing r~ by the combined reflection coefficient of the load and of the
probe. For simplicity, assume that, the generator is matched so that
r~ = O. The voltage will be proportional to (1 + r,,,), \vhere r,,, is the
total reflection coefficient at the probe. The lutter can be sho}rn to be

r; + r, + 211;,111,
re,f= -- - (20)
1 IjIp

in \vhich I~ is the reflection coefficient of th< prolw, gi\-en in terms of the

probe admittance Y, by
1,
r, =
1-,, + 2Y

and 11~is the reflection coefficient of the 10wJ refcrmd to the probe
position
r: = r,d~z.
The voltage variation, therefore, in terms of r, and r, is given as a
function of z by
V = const (21)
1 l-+r:+I - ~z
1 Y. Dowker and R. l\l. I{wlhrffrr, An Investigation of IGF Pmhm, IU, Report
No. 483-14; If-. .Utar, F B. \[:trsl~:dl, I.. P. Hllntcr, Prohc llrrors ill Standing-\vavc
Detectors, IrIJc, IR1:, 34, 1, 33-44,
SEC. 158] PRI.MAI<Y PA TTER.V APPARAT(JS 557

lhe effect of a finite reflection coefficient as given by Eq. (21) in

measuring VSl}TR and phase shows the following results for a matched
generator (for a mismatched generator the state of affairs becomes even
more complicated). For a tuned probe the measured VSWR will always
be less than the true VSWR; however, the minima and maxima will occur
in their proper positions. For an untuned probe the measured VSWR
will also be less than the true VSIVR; the standing-wave pattern, how-
ever, becomes asymmetric ~vith maxima and minima displaced from their
correct positions and maxima and minima no longer separated by k,/4,
despite the fact that adjacent minima are still separated by A/2 as are
also adj scent maxima. The position of the minimum is less affected
than the position of the maximum; the higher the VSWR the less the
minima are affected; and it is for this reason that minimum position has
been recommended i,lstead of maximum position for phase determina-
tions. For example, for a load mismatch 11,1 = 0.1 (r = 1.49), with
17, = O. leTi4 the minimum position is displaced only 0.0063Ar from its
true position: and for r. = 1, the minimum is not displaced at all.
In the calibration method of Sec. 15.6 a probe reflection as small as
r, = .005 will lead to an apparent value of n = 1.97 where the correct
value is n = 2.

PRIMARY FEED PATTERN MEASUREMENTS

15.8. Primary Pattern Apparatus for Point-source Feeds.The
dimensions of microwave radiating systems that are employed as point-
source feeds for reflectors and lenses are such that the physical distance
from the feed at which the radiation zone sets in is within the dimensions
of normal laboratory space. The distance is, of course, large compared
with the feed dimensions in conformity with the requirements of the
far-zone fields (Sec. 3.11). A complete primary pattern consists of the
spatial distribution of radiated energy about the feed, the surfaces of
constant phase, and the orientation of the electric vector (polarization)
at all points on a sphere centered at the feed. The techniques for deter-
mining each of these components of the pattern in the radiation zone will
be considered in the following sections.
The technique of measurement is, in general, the choice of the experi-
menter. It has been found convenient, however, to take transmitting
patterns of primary feeds; that is, the particular feed in question is made
to transmit microwave energy of the required frequency, and a small
horn or pickup is then used to measure the intensity O; the radiated
energy, phase, and polarization. Later, in discussing the measurement of
secondary patternsthe pattern of the composite antennait will be
convenient to consider receiving patterns.
Transmitting patterns may be obtained in principle by receiving the
radiated energy from the transmitting antenna in a polarized pickup
antenna at all points on a large sphere centered about the transmitting
558 ANTENNA MEA8UREMEAVTS-TECHNIQ UE.V [SEC. 15$

antenna. A suitable form of apparatus to accomplish this for an essen-

tially point-source primary feed is sketched in Fig. 157. The r-f com-
ponents illustrated are typical 9400-Mc equipment. The general layout

\
Waveguide - coaxial

Probe

Amplifier

FIG. 15.7,Apparatus for primary pattern measurement on point-source

md ideas, hon-ever, may be used throughout the microwave region.

As an r-f source, a square wave modulated klystron has been utilized
10 give a steady output. The output from the generator is connected
by means of coaxial cable to the waveguide that is terminated by the
feed under studv, which is shown here as a horn. .4 cavity waverneter
is included in the r-f line, The line and feed are clamped in an adjust-
able mount so that the apparatus may be properly aligned. These
adjustments include horizontal and vertical displacements as well as
 Sm. 15.8] PRIMARY PA7TI$RN APPARA7US ,r~ 9
;j

means of rotating the feed about its axis (feed axis being here defined as
the direction of peak radiation). The feed is mounted so that its approxi-
mate center of phase is located on the aximuth axis of rotation of the
pickup antenna. The pickup antenna is mounted on a turntable illus-
trated here as a 30-in. gun mount, the upper ring of which is free to rotate
on ball bearings between it and the lower ring. The pickup may be
rotated about a horizontal axis, and its distance from the axis of the gun
mount is adjustable.
If only intensity and polarization measurements are of interest, the
detecting element may be placed behind the pickup and the output
delivered directly to the indicating system. Illustrated in Fig. 15.7,
however, is a system suitable for the measurement of phase as well as
of intensity and polarization. In order to eliminate cable flexing which
may produce phase errors, the energy is delivered to the detecting system
via a cable, fastened to the pickup MOUIlt, to a rotating joint mounted
\vith its axis coinciding lvith the turntable axis, and from the joint to
the detecting system by a fixed cable or a rigid wavcguide, The detect-
ing system consists of a tuner, a slotted section on which is mounted a
movable tunable probe through which r-f energy from the source may be
introduced for phase determinations, I and finall,v the detecting element,
crystal or Littelfuse. For intensity measurements the cable from the
r-f source to the probe is disconnected.
I A detecting element, ~vhich has been found to be very satisfactory,
is a current-biased Littelfuse or Ilrollaston ~vire, This is in series ~vit,h the
transformer input of a lo~v-noise, narrow bandpass, 10,000-gain, Iineiw
amplifier. The change in the resistance of the element is proportional
to the modulated r-f power input over a very \vide range of po\ver level,
and the resultant readings on a model 300 Ilallantine voltmeter or
other suitable electronic \-oltmeter are very accurately proportional to the
square of the field intensity. A crystal may be used as a detector \vit}l
the above apparatus; but although the sensitivity is of the order of 10 db
greater than the I,ittelfuse, the la~v of the crystal varies as a function of
the r-f power level. Consequently, it is necessary to kno~v the calibration
or law of the crystal as a function of the power level before correct field
intensities may be obtained. This objection to rrystals holds only for
intensity measurements and not for phase measurements. In addition,
the la~v of the crystal has been known to change \vith time and handling.
Crystals are prefmable for both intensity tind phase measurements if a
c-w source and superheterodyne detection system are used instead of the
modulated source and :~mplifi(~r-~roltm(,tersystem, sinrc the r-f po\ver
level required is greatly reduced and the crystal is square la}r for lo~r
po\ver levels.
:1s far as possible, all r-f components shollld he Ivell matched to ol)t:lin
.41)tllt(,rll:itivr 111(.ttlodfor I)]()\,i(lit)g :1 1)11
:{.s(, r(.t(.rvll(,(, is (Iiw[l,w,.(1 III %,,., 1,5 12.
560 AN TENNA IIf EA S L~REM E.V 1STECHA IQ (JE.S [SEC. 15.8

maximum-power transfer and to prevent possible interactions, Gen-

erators should usually be loosely coupled to avoid a change in power
output and frequency due to a variable load, The srune tcchniqups
and cautions required in other r-f measurements are genendly true for
antenna measurements. Ho~~evm-, in some CLWS,mismatch in the var-
ious elements does not resldt in incorrect, measurements. lllerever good
matching is required, this \vill be emphasized in the text,
There are several requirements th~t the pickup antenna should
satisfy. (l) Itshould bep(Jlarized andarrallgeti sothatr eflecti(,1lsfrom
its mount will be negligible. (2) The pickup should be mounted su that
its feed axis is perpendicular to and intersects the turntable azimuth
axis. (3) '~hepickup should ha~es(~nledirtctivityin orciert(>minimize
the effect of possible reflections. (4) The picki~p should be capable of
being rotated about its axis in order todeterrnin ethepolarizatio no fthe
antenna being studied as ndl as to metisure intensity for various orienta-
tions of the transmitting antenna. It is a!so desirable to be able to vary
the distance of the pickup from the turntable axis so that the field
may be measured at different distances from the fed when necessary.
A simple and suitable pickup antenna consists of a rectan~lllar ~raveguide
horn of mouth dimensions such that the apertllrein the clertric planeis
equal to the width of the ~vaveguide in the magrwtic plan?. For rxanl-
ple, in the 3-cm region a fairly well matched pljhwized pickup horn is
obtained by flaring 1- by ~-in. rectangular waveguide to an aperture
1 by 1 in.
The primary pattern apparatus should be Iocatcd in as open a space
as possible to eliminate undwir~hle reflections from s~[rrwunding objects.
It is very difficult toestablish acritcrion for this bec:uwe it is afllnction
of the desired accuracy, the size and directivity of the frrd as Nell as
of the pickup, and the scattering cross sections of the vario~ls objects in
the neighborhood of the apparatus. In many cases it is difficult to
ascertain the callse of unexpected peculiarities in the pattern, If there
are objects near-by suspected of reflecting, the dfects ui thrir remo~,al
naturall,v should be observrd. If a symmet,-ical]y cc)nstr[lcted f?ed is
being studied, the presence of a side lobe on one side and not on the other
at the same angle dcnotw either an error in the :dignrnent of apparatus
or the presence of a reflecting object on one side. A useful precaution
against the effect of reflections, in particular those of neav-by \rallsj
is to cover the scattering objects II-ill] nlicroivave alworbing material.
one must be certain that the material is actually an ubsorber in the
Iravekmgth region bring (onsidcrc(]. part i(,{llarly if a Iossy material
backed (I1>l)Y a metal sheet is ~MJtlits al] absorlwr. 1 [u this (ase, the
angle of inci(lence as uell as the iravelength mllst be consi(lcrc(l,

] .4n Mlv,tiv{, :It)so,-l]i]lz ,s<,r{x,lIO:LJI bc Construct, u~ I)y I):,(kil]g :1 S],11) of wool] $$

 In gf,ncra], the ~PPLi~2L~US itself should be as free as possible of reflecting
surfaces or objects. ltitterns should be taken preferably in azimuth
because reflections from a horizontal surface may be made symmetrical
and with the msual dimctivity of feed and pickup they ~~-illbe negligible.
The primary radiator and pickup should be located as far as possible
from horizontal surfaces such as the table and ceiling in order to minimize
the effect of these reflections. In most cases, if the distance between
the table and the feed is of the order of 10 ~vavelength.s or greater, very
little trouble \vill be encount~red.
A number of other more elaborate primary pattern setups have been
devised, primarily v-ith a ~-imv to greater accuracy in phase measure-
ments. onc possibility is to use a vertical transmission path, thus
eliminating reflections from either the floor or the ceiling which may
conceivably cmse errors in phase measurements made on the apparatus
described above, Holvevcr, a vertical-path device is considerably more
complex rnechtinically and is warranted only where definitive measure-
ments ar~ req~limd.
15.9. Intensity Measurements.The intensity patterns in the princi-
pal electric and magnetic planes (hereafter to be denoted by E- and
H-planes respectively) are us~lally sufficient for simple antenna designs;
sometimes, however, a knowledge of a greater portion of the space pat-
tern is required in more careful antenna design. The method for deter-
mining the field intensity as a function of direction will be discussed first
for the principal 1?- and 11-planes and then for other planes or cuts.
The r-f components need not be \vell matched when only relative
intensity measurements are re(luired. 31ismatch in the various com-
ponents will alfect only ~he power le~-el at the detector and will not
tl-,erefore affect the measurement unless the mismatch is variable. The
separation of feed and pickup should be of the order of 2d2~Xor grrater,
where d is the maximum aperture of the feed. For small d of the order
of a \va\elength, this reqllii-ement may be open to question, and one
should in this case ha~-c a few-to-pickup separation of at least several
wa~,elengths. These criteria ensure that the frwf is being examined in
its Iraunhofer re~ion. The distancw suggested ~~-illminimize interacti(m
bettveen the transmitter and recri\-er. If the feed is to be used in conl.
bination with a reflector such as a paraboloid, it is desirable that the dis-
562 AN TliIV.V.4 .!ll$A,Y LrRli,i{lliV TS-TEL<H.VIQ UhS [SEC.
159

tance from the pickup to the feed should be roughly equal to the distance
from the feed to the contemplated reflector.
A check on whether the Fraunhofer pattern is being measured may
be obtained by mming the pickup a\~ay from the feed at a fixed angle,
say the peak direction, and determining whether the po~ver response, as
measured on the electronic voltmeter, varies in~m-sely \vith the square
of the distance. Figure 158 illustrates the dependence of the measured

FT~. 16.8.Tenth-power width @ ~~oas a function of the path distance R between test h{,r,i
and pickup horn.

10-db width of the intensity pattern upon the separation bet\\-eena

feed and the pickup. It will be seen that the beam~vidth becomes prac-
tically constant for path distances greater than 1.5d/A. This is in agree-
ment with the choice of 2d2/A as a minimum separation (cf. also Sec, 691.
To measure the H-plane intensity distribution, the primary feed
under study is mounted so that the estimated center of feedl is locatevl
directly over the center of rotation of the turntable and the feed oriented
so that the E-vector is vertical. The pickup horn with E-vector vertical
is aligned with the feed so that the axes of the feed and of the pickup, if
extended, become a common horizontal line. The If-plane intcnsiiy
pattern is then determined by moving the pirliup feed in azimuth :1( :1
fixed distance and taking readings from thr \-oltnlctcr as :1 fun(ti{)n tjt
sm. 159] 1.Y-7[<A-,V[ 7)- .l[[<. ! V( li I<.lrl<.v r#Y 563

angle. With a square-law detection system the relative powm per unit
solid angle is thereby determined as a fllnrtion of angle for the principal
H-plane of the radiation field. l;Y rotating the pickup and tccd 90
about their owm axes and c:irrying out the procwill re dwrrihwl for the
H-plane, the principal E-plane pnttern is determined,
Various methods of presmting this cfata have been used, rach
with its own merit. All too fre-
Y
quently in the literature and re-
ports there is a failure to label f
properly the intensity axis as field
intensity or square of field inten-
sity (power pattern). A semi-
logarithmic plot of the po\rer as a
function of angle is commonly
used and has the advantage of
emphasizing the lobe structure,
The relative field intpnsity (or
simply the relative fiel~l intrnsity
squared) as a function of an,qlc on
a polar plot invaluable f<]rol Jtain-
1
ing a good picture of Ibe fiel(l di~- ) z
Turntable ax(s
tribution; in particular for field
1:1<,. 15.9.C<]<,rclit]:Lte sysreln for primary
intensity, this plot gi~-es thr rel:L- ~):ittl?rl, h.
tive range as a function of :mg]e at
which a given rwri~-er wi]l recei~-e a constant signal tronl the fixed trans-
mitting antcmna (qf. Sec. 12).
We now have the principal Z- and H-plane intensity patterns;
let us consider the problem of o})tainirlg the space ciistribution of intc-nsity
and the polarization, }Ye definr :Lcf)[jr(linat~ system a and ~, sh,)~~n in
Fig. 15.9. Thez-axis istheaxis t]ftl]t, )jrin~a],yf(,(,ci;@r[J]~I[>seI]t sr(jt:Lti(Jn
of the pickup aho~{t the tixis of tbr tllrniable; and a reprwents the orien-
tation of the L-ve(tor at points on the feed axis relalive to the plane
defined by the y- and z-tixrs, Tllc, al)p:~ltltllsisfirstlinfdllp, asdmcribe(i
before so as to obtain the principal H-plane pattern. To check the
pohu-izution in this [Jlane, at rach azimuth position of the pickup, the
pickup is rotated abollt its axis until the maximum po\ver is received.
The angle of the pi(,kup then ~l(t(rmin~,stlle:~ngle of polarization of the
radiation at that point. .ifter the polarization and intensity have been
determined for the principal 11-pkme pattern, the feed and pickup are
ret{lrned to their initi:d cc)ndition of a = O und ~ = (), Ihe feed is then
rotated through an anglr a (s:~~-100) abo~lt its mis, and tile pick~lp is
rotated through a simil:Lr an,gle so tll:lt their Ix)lariz:ltioru= are parallel.
The intensity reading s1l{NII(Ibe exartly the s:hm( :LSthe initial reading.
~-;)tc that this mrthtd hastbf> (](sit:l])l(.f(>:Lt\lr(,th:Lt:Nf YLrh cut ist,akrn
 the power level is checked at + = 0, that is, for peak int~nsity. The
pickup is again moved in azimuth; ancl at each an~k> ~, the maximum
intensity and polarization are 01]1ained as alxn-c. This procedure is
repeated for sllccmsi~e values of a. If the field-inf (nsity-sq{lared data
are plotted as a funrtion of a and + (m s(litable coordinate paper, contollrs
of constant intcmsity can be dra~vn.
Direct polarization on a paraboloid is that component of the radiateci
primary field which, after refiertion from tfw paraboloid, is parallel to
the dominant, polarization of the antrnna pattern, Direct- and cross-
polarization components exist for each point on the parabolc}id, If one
is interested only in the direct-polarization component of this feed for
use with a paraboloid, the procedure discussed abo~-e is considerably
simplified. For example, the If-plane patter-n is takrn \vithout rotating
the pickup about its axis for polariztitiorr dderminatiou. I,ike~vise, if the
feed is rotated by an angle a for the a cut, then thr pickup is rotated
the same angle a and kept there \vhilc the intensity as a function of ~
is observed by moving t,hc pickup in azimuth. This procedure is correct
as may be observed by considering the properties of a paraboloid of
revolution. If for any reason the cross-polarization p:ittem of the feed
is desired, it may be obt ainrd cxtict ly as above ~fith the exception that
when the feed is rotated a degrres, the pickup is rotated a + 90.
If \venow desire to determine the contours of constant intensity on a,
paraboloidal reflector, in order to obtain optimum illumination of the
reflector area, I we may, by a sllitable transformation of coordinates and I
remembering to account for space attenu:itionj project the contours of
constant. int rnsity alwad,v dm \rn on the sphere onto the paraboloid, 2
I
1.5.10.Phase Determinations. -In determining the phase front of a
primary radiator it is usually ctmsid~red s~lfficient to mamine the phase
in only the principal E- and 1{-pkmes, , since, if the cur~,es of constant I
phase in these t~vuplanes are circles \vitha common center, it is reawmab]c
to expect that the phase front is sphprical. Ho\\-ever,if other planes are I
of intcrrst, they may be rxamined in a manner similar to that employed
for intensity measurrmcnts, Thr met hods for determining phase involve I
the comparison of energy from thc pickl[p ll[)rn ~rith enmgy trom thr r-f
sourre. The appar~tus for (}~e n~cthfxl described Ilerc is illustrated in I
Fig. 157. me poff-er from the r-t so{lrcr is divi[]ed betl\-een thr pri-
mary radiator and the tllnal)lr +Ii(linx probe inwrtc(i io the ltaveg~lide. I
Encr~y from the pi{kllp anfi frc,m t]lc so(lrce Irill then a({(i in or out of
phase at the cfrtc(,t(]r (Irpen(ling on tllr plolx. l~osil i~)ll.
I
If the Ph:we []istril)~lti(]n in tll( /f-II]an~ is }xirlg in~wtigal cd, then
1Sl,[,sm. 1214,
2.J. 1. Iloll]l{,rt :111(Il..J. I{(,:irt, J{J. l{(l)t)rt \-{,. (1.1<1,\l:\\I,l) 1!14.5: .J. [. Jl[jllt){,rt, I
1{1, Ii!]x,rt X,,. Mi.i l,l,r,l:ir~ 1{,)-lfi: S. ,J, Jl!xt,,], t Il{}rll 1{,{1. Il,r I:ll:,lx,li(. I{rllt,f.
tors, 1{1, I{{]){)rt X(). Ii!W, ,J:,t,[u, ry l!).l(i, ]]]). 21 22,
I
I
SEC. 1510] PHASE DE1ERMINATIONIS 565

the apparatus is lined up as before for H-plane intensity measurements

with the axes of the pickup horn and primary radiator coinciding.
Keeping the pickup fixed, the position of the probe is varied until a
minimum signal is observed on the electronic voltmeter. For this situa-
tion, the monitored powerand thepickuppov-erarc mlt of phase. This
\villbe the reference value, and the position of the probe is rccordcd. It
will be noted that a similar minimum will occur if the nrobe is moved a
distance h, (A if the detection components
are coaxial), since this represents a 3600
change in phase. For this setup, motion of
~he probe toward the bolometer is equivalent,
so far as phase is concerned, to keeping thr
probe fixed and moving the pickup toward
the primary radiator along a radius through
the center of feed. Therefore, moving the
pickup a distance d toward the radiator is
equivalent to moving the probe a distance
p = d(X, /X) toward the bolometer. The
reference value for the probe position hav-
ing been determined, additional points on
I~lG. 15 l(J.Ccnter of feed
the phase front are determined in he follow-
cnrrertlnu. Ihrkup Inoves o!)
ing manner. The pickup is moved in azi- rirrle of ra<lius H. True wave-
muth through an angle ~; again the position front is cirrlc of radius r.
l{cferenre fm,nt @ = 0.
of the minimum is noted; and the difference
p from the reference minimllm position is obtained. This is carried out,
through the required range in angle +.
If the position of minimum deflection is indrpendcnt of +, then the
phase front in that plane is a cirrle whose center is rxact]y the point on
the feed directly over the turntable axis of rotation, This determines
the center of feed in that plane. If the minima positions are different, the
estimated center of feed, ~vhose location is knmvn ttj bc on the z-axis as
showm in Fig. 15.10, may be corrected by means of the cxluation

i-R=[i-
1%. !
(22)
l(.os+-:

where r R is the correction to be ad(le(l t f) t hr assumed center of fec(i

located in this plane and on the axis of the tllrntablc. IAIUCS of d at
corresponding angles + are obtained from d = t p(x A,) \vith due regard
to the choice of sign, the plus sign corresponding to motion of probe away
from bolometer. Sine? R is Iargc :IIId ~ R is small, the correction ternl
ma,y be obtained ~vith a high degree of occ~lracy. The relation :ibo~-e may
be obtained by considcrin< ~i~. 1510, It will Iw noted that d f)r p(~ X,)
is measured along the radius r. IIy the l:LI\-of cosinrs,

(r d) = R + (r R) + 2R(r ~?) COS+,

or solving for r,

[,_co,+_(~v]
T=R

[ 1
Icos+j

when R is subtracted from Iwth sides, the dcsirml relation is obtained.

Eqwimmtal lrrca~~tior~s.
-Severa1 pract ical prvcaut ifms should be
taken in order to ensure proper phase measurements. The po\ver from
the probe and from the pickup into the bolometer should })e of approxi-
mately the same level in order that \\-ell-cfefinedminima should occ(i r.
It is undesirable to change the power level from the generator or the
probe tuning during the phase measurements, since not only is it a
nuisance but there is also the possibility of a change in phase. A recom-
mended procedure is to adjust the tunable probe at first so that the
power which it delivers to the bolomcter is 6 db belo~v the power delivered
by the pickup in peak position. This may be accomplished by first dis-
connecting the probe and then r-otfiting the pickup in azimuth until the
po~ver deli~-ered to the bolometer is b db belo\v pezk value. lhe prol)c
connection is then made; the probe moved to a minim-urn position; the
probe tuning adjusted to d(?cpcn the minimum; and finally, the minirnllm
may be sharpened by adjusting the probe depth. Slightl~- mow t~lning
may be necessary after the probe depth is changed. he pro}w depth
should not be so great as to have a I:lrge rctlection cocfficirnt. After t})(.
above adjustments have been rna[le the prolw tunin~ is not varird (Iliring
the phase measllrcments. The least chanw bctwccn maxima AIN(
minima over :L 12-db range of pickup power frill lhtin bc tb:~t in \\hic]lt}W
powers from the pickup an(l the probe differ by a f:wtor of 4 ((j (lb), and
therefore the amplit(l(lcs IT-illdiftcr l)Y :L factor of 2. llr rwllt]ng r:~tio
of the maxima amplitlldr to the minima I\-illIw (2 + 1), (2 I ) = 3, :In(l
the ratiu of maximum to miniml]m Volt:lgc as ol)s(rvd on the 13all:mtinc
voltmeter ~~ill be 9, which is su~lcicr)tly large for accurately determining
minima.
lVith the system dcsrribed :~ho~-cit i> ntwss:~ry that a goo~l mutch bc
ensllred looking from tlw bolometcr to the pi(kill], sinrc lhr l)()!~(ir
radiated by t,hc prob(~ into the linr di\i[lcs. p:~lt goinx to Ihc lM)IOIU(,(CY,
part going tol~ard 1he pickllp horn. ro (J}ltilin a g{lf)(l m:ll(ll :~ tlln~,r is
placed bet\\-cenIlle prwlm secti(m :111(1the t-f line (tlllnrctcd 10 Ihc I)ifkllll
horn, as in Fig. 157. If this SC(Iion is rnisnmi (IIc(l, i lie lxj\\cr (Iclilrr[d
sm. 15.10] PhASE 1jE7ER.~fl,Y.4 TI0.T,Y 567

bytheprobe tothebolomcter is a function of itsposition along the line,

~vithamaximllm tc~miniml.tm ratio proportionaltothe (VSWR)2 (looking
from bolometer to pirk~lp), and a period of 1/2. .ksaresult false minima
may occur with roughiy k(,/2 sparing instracf of the expected Ag spacing.
A good check on the phase-mraslwing equipment, is to make sure that
no maxima ormiilima ocrurothe rthanthos eexartly~(, apart.
The effect of a mismatched pickup ma,v be illustrated by the following
example. ~Suppose that the voltage amplitude at the bolomet~r from a
matrhecl piukup is unity and that the amplitude from the probe at the

bol{)metrr is 2, its phasr bring dcp(,ndc,nt on its position; then the res~lltant
fiel(l ~t the bolomet~r is
E = 1 + 2@ ~++, (23)

\vherez is the distance of the pro})e from the pickup and @ is a phase angle
that mprewnts the relutive phasr lmtirern the voltfige at the probe and
at the pickup. The po\~er into tl~~ bolometcr is proportional to EE*
\vhicb is
Ei* = 5 + 4 cm (2@z + +). (24)

Illustrated in Fig. 1,5.11 is a plot of EE* as a function of z for the case

@ = m. H(;Iw 1~1 has a minimum ~a]~le of 1 at z = () and a maximum
va]ue of 9 at z = A,,/2 and 01hfir minima at i,,, 2k0 . . . and other maxima
at 3X(,12, 5k0,/2 . This cxaml)le sllo\rs the proper periodicity that is
obtained lvith a IVell-malrll((1 pi(kl]lj.
If, ho\Jever, thr pirkul) is mismatrhwl (for example YSWR = 3), the
result ant field at thc 1)[)lt)m(>lm depc]lds on the probe position and is
given by
E = ] + 2r,(6z+$,+ f,t-~z+~+.), (251
a depends on the phase of the reflection coefficient of the pickup and
will be chosen equal to zero inthis example without affecting the qualita-
tive nature of the illustration. The bolomcter power,foro = r, ~villbc

II; =26cospz+8cos& (26)

The dashed curve in Fig. 15.11 shows the ~ariation of IE12VS. z. It will
be noted that instead of a minimum at z = 0, there is actually a maxi-
mum, and likewise minima do not occur at kg spacing hilt at spacings of
0,378X, and 0.622 k,, and are poorly detined. By variation of the ampli-
tudes and phases other undesirable configurateions may be obtained.

(a)

~ 2;/(2;+ ZJ>

3%402
a

l;l~. 1512.-EquiIwlcl, t {irr(lit of r-f probe.

That the variation of polver as determined by the probe position is a

function of the mismatch of the pickup may be velified by the following
analysis. If a generator of voltage V~ and internal impedance Z. is
connected to a load ZL by means of a transmission line whose charac-
teristic impcdancc is ZO aml length is L, and if Z,, and Z. are assumed to
be real (this assumption \\ill not affect the general result), the po~ver
delivered to the 1o:,(I i ; ~(()){*(()) 12 as obtainwl tr[,m Fkls. ((;) :md (7);
setting ~ = ~~,wc I1:JI.c*

2(,T:
1 r;.
Power to the load = - (27]
2(Z + z(,) 4iI.
I 2rI,r(, cos ~ + r~r;
( 0 )
Note that the po\ver delivered is a funrtion of the line length. The inter-
action between the generator and the load is contained in the expression
(1 21,,1Ccos 4~L/xu + r~r;,) \\-hich~aries bet\veen (1 + r,.r,,) and
(1 r,.r~)z,depending upon the line length.
The properties of the prolw in the phase-mrasuring circuit. as repre-
sented in the equivalent circuit of Fig, 15. 12(1, may he (Ietrrminerl from
Eq. (27). The impedance of the pickup arm is reprcsrntcd by Z*, 1hr
SEC.15.10] PHA, YE l) IWli[tM INA 7IOAS 569

probe is to be considered as a source with very high impedance; motion of

the probe will cause 12and 11 to vary, but their sum will be fixed. J3Y
means of Th6venins them-em (See. 24), the above circuit can bc replaced
by the circuit in Fig. 1(512b in which Z~ is the impedance Z, transformed
to the terminals au. The replacement of J. in ~q. (27) hy vc~.zi~(Z; + ZG)
and of ZQ by Z~Z~/(Z~ + ZG) gi~es as the power into the load

P = +~~ (1 r?.) (3~_r2 cos 2P12+

r;) (28)
. (1 2r,r,, COS2P1. +ri r))

for Zc >> Z. or Z,. The only factor that contains the position of the
probe is (1 + 2~z ros 2(312 + ri) ~vhich has a maximum value of
(1 + rJ and a minimum value of (1 r,). Thus the rai io of maxi-
mum to minimum po\ver received by the bolometer by varying the probe
position is (I + r2)z/(1 rJ2 or the (Jswlt)? of the pickup arm.
To remove this variable effect of position a tuner may be used effec-
tively to make Iz = O. Another possibility is to use a well-matched
attenuator pad. A 3-db matched pad or attenuator (looking both ways)
reduces the reflection coefficient by one-half; a 6-db pad, by one-quartm.
,Suppose, for example, that the pickup (lSWR)~ is {1 (rz = ~) and that a
6-db pad is put between the pirkup and the probe section; then the square
of the resulting mismatch will be

(29)
2=[%-+-i] = 65
thus appreciably reducing the effect of the pickup mismatch. The latter
method can be used only if the loss in po\\-erlevel can be tolerated.
Frequency Sm.srliuity.-In order to reduce frequency sensitivity in
the primary pattern phase measurements, the path lengths from the
r-f source to the mixer must, be chosen so as to minimize the change in
phase of the two paths as a function of the \vavelength. Consider the
path from tuner A (Fig. 157) to the mixer ria the feed and pickup, and
let there be a length La for ~vhirh the ~vavelengtb is the free space wa~c-
length 10 and a lcn.gth L. for ~vhich the ~vavelength is A,. If the path
length Lo includes a section of dielectric-filled coaxial line of length 1.,
then L. = la + l.k,~, \~here la is the path length in f ree space or air-filled
coaxial line and k. is the specific inductil-e capacity of the dielectric
section. The corresponding distances from point A to bolmmcter via the
probe will be iSo and S,. The difference in phase lx%~ieen the t~vo paths is
this:

(30)

(30a)
570 ANTE.VNA ME.4sURJ;.tf E.V T,s-TIKH.VI(J 7_:lM [sm. 1511

For waveguide ~, = AO/Vl (A O/2a) so that dk,/dAO = (~g/Ao) 3, ;vhence,

on clifferentiat ing ~ with respect to AO,we have

d4 __~
La s= L$,, A,
=27r (31)
dAo [ Ao (x,)
-----( x, )1

For a minimum phase variation d@/dA, ,= 0, and [W see that the difference
in path in waveguide for the t\vo paths should he related to the diffcrenm
in path lengths in coaxial line or free space by

s, L, = : ([,. s.). (32)

u

It will be noted that this k not the same rcs~dt \rhich one ~vould obtain
by simply making the electrical path lengths equal, which gives the
result S~ L~ = (A,,lko) (L. S.). Oi course, it only transmission lirms
with wavelength k. are used, equal path lengths in the t~vo directions are
necessary to minimize the frequency sensitivity.
i% an example of the decrease in frequency sensitivity, suppose
that Xo/?IO = 1.4 and that the frequency of the source changes by 0,2 per
cent. Calculating from Eq. (3oa) the change in phase that occm-s for
this change in frequency fvhen the proper distance given by Eq. (32) is
used, one finds it to be ~luu of the phase change obtained by ~wingm~llal
path lengths. If L. S= = 50A0, the change in phase is 0.034 wing
proper path lengths, whereas ~vith equal electrical path lengths a phase
error of the 34.6 results. The former is negligible in phase measure-
ments. In actual practice the rcquircrncnts of Eq. (32) need be met only
approximately if the line Iengt hs in terms of vvivelengt.h arc not ]argc.
15.11. Line-source Primary Pattern.The measurement of the
intensity and phase ciistributiou along the length of a line source such
as a linear array or pillbox is of interest. The measurement of the field
close to the line source, which usually has a small effective ~ertirai
aperture and wide horizontal aperture, affords a ~alual)le check on the
design of the line source. The tcchni(~~m in\olvccf in this measurement
are exactly the same as those {iiscllssed a]mve in connection ~vith point
sources, with the exception th:Lt motion of the pickup must be parallel
to the line source as the intensity find phase measurements are taken.
This does not represent any cliffi(lllt y if intrmsity distribution alune is
desired. For phase determinantif)n+. hf)i~-c.,cr, the appwutus lxx,f)mes
more compliratccf, since sevcrfil lot Ntiug j(~int> luust be ut ilizcd in (Jr(h,r
to eliminate the tfe~ing tal)les i)-llich miLy calbe .3,pparcnt tha])ges ill
phase. Fi~ure 1513 is a skct (I1 of a sllitaljlc :Lpl)a]atllsj collt :Li])i])~
three rotating joints in or(lrr to nMlw Iwwitde thc rrtluirc(l liucar u:t)tion.
These three are necessary if tlw line source Iwing studied is horizontally
SEC. 1511] I. I.l-Ii-oY(il -l{(l~ llil.ll. i 1{~ 1.1 77l.l<.\- 57 I

polarized. For vert ica! polarization the rotary joint attached to the
optical bench can be eliminated by using a coaxial pickup mounted in a
bearing. Care mllst lW taken that the rotary joints u,rp properly designed
so as not to shoJv a changr in phaw !~-ilh rot at ion. They should be
fastened in such a lvay tll:~t the!- If ill not go out L~r alignment. The

nput from I
r-f source

1
-a&k&-. J] sldfv
> /7iMovatle
T tmolmt,,,i
t 7/ er t~ble

~------
I
probe

To ampllfler

..
?&
rotary joints are all mounted in a plane with the first joint fastened to
the traveling arm of the optical bench, the second movable about a
circle whose center is the third fixed rotary joint. The r-f output from
the thi~d rotary joint is she\\-nhere being fed into the phase detection
system. The mixing system illustrated here utilizes the m-called Llagic
T that will be discussed belo\~-. The power coupled from the generator
output for a phase reference is fed into a tunable sliding probe \vhich
excites the \vaveguide, one end of which is terminated in a matched load
and the other end connected by cable to the 31agic T. A dial indicator
(calibrated in O.001 in.) isshownr nounted ontheprobe section so that
the motion of the probe may be accurately measured. For intensity
measurements the cable to the probe is disconnected. Usually the dis-
tance between line source and pickup is only of the order of 2 to 4X in
order to stay within the cylindrical ~vave zone of the source. The
pickup should be as small as feasible in order to pre~-ent any interactions
of the pickup in the field of the line source. The sketch also indicates
that the movable parts of the apparatus are removed from the intense
~ortion of the radiated field so as to reduce extraneous reflections. The
operator should like\vise be out of the strong field.
15.12. Magic T.The Magic T1 may be advantageously used in
phase measurements to reduce interaction effects discussed in connection
\vith the apparatus showm in Fig. 15.7. The waveguide form of this
device is shown in Fig. 1514. If po\\-eris fed into branch P (parallel
arm) and A and B are terminated by matched loads, then the power
divides equally between branches A and B, since a symmetric condition
exists and there is no component of field
s
available to excite arm h. If power is
a fed into arm S (series arm), the po\\er
again divides equally bet\veen A and b
- i-
A )_____ ., - B but the fields in each arm are 18(3 O[lt,
/ of phase and no po~ver is delivered to
+--
/ branch P. In order to prevent reflected
/
/ poiver in arms P and S respecti}-el~-, it
* P is necessary that a match exist, looking
~10. 15.14.The .Magic T.
into arms P and S respectively when arms
A and B are terminated bv matched loads. In this \vav interaction
between the two sources deli~ering po\ver from P and S to ~rms A and B
is made negligible.
In the Magic T as used in the primary pattern apparatus, arm A is
terminated by a matched load and arm B is terminated by a rnatchwi
bolometer. Po\ver from the pickup is fed to P as she\\-nin Fig. 15.13,
and power from the source is fed to arm S (this order can, of course, be

1 SCC \rO1.11 of this series.

SEC. 15.13] BEAPON AZ IMIJTH IATTERNS 573

reversed). Therelative phase of the fields from thetwoe ffectivesourcm

is varied by means of the sliding probe. 310tionof the probe a distance
p = d(A./A) toward the bolometer is exactly equivalent to motion of
the pickup horn a distance d perpendicular to its path toward the
source.
15.13. Beacon Azimuth Patterns.A microwave beacon antenna
may be examined for uniformity of azimuth pattern by utilizing the
apparatus described for point sources, with the simplification mentioned
where intensity only is measured. 1 The accuracy required for azimuth
patterns is not very high because one is usually interested in uniformity
to within 1 or 2 db. Because of the essentially uniform azimuth patterns,
reflections from many sources may affect the results; therefore particu-
lar care must be taken that reflections from surrounding walls and objects
are negligible and the apparatus should be located in as open a space as
possible, since a 360 pattern is desired. The pickup should be as small
as possible if measurements close to the beacon are being made. How-
ever, if sufficient r-f power is available so that a larger distance between
pickup and beacon can be used, then the pickup may be larger and more
directive. Since in the measurement of azimuth patterns only a few deci-
bels variation are observed, greater sensitivity in the apparatus may be
obtained by the use of crystal instead of a bolometer as a detecting
element. Over a range of several decibels the law of the crystal will
not appreciably affect the relative readings. For example, assuming a
square-law crystal, the ratio of peak power to minimum power might
be 100 to 50 or 2; if the exponent in the law of the crystal were 2.3;
the corresponding readings would have a ratio of 2.2 which is within
~ db. For a little more accuracy, the ratio may be corrected to a suffi-
cient degree if the approximate law of the crystal is known.
There may be some question about the distance required for measure-
ment; obviously a distance between pickup and beacon of 2L2/k, where
L is the vertical length of the beacon, would certainly be safe. However,
because the directivity in the azimuth plane is almost nonexistent, this
distance could easily be reduced by a factor of 2 or 4 before an appre-
ciable change in azimuth pattern is observed.
Two schemes have been used for beacon measurements. In one of
these the transmitting beacon is kept fixed and the pickup is rotated in
azimuth about the vertical axis passing through the beacon. The ,other
procedure is to rotate the beacon about its vertical axis, feeding power
into the beacon by means of a rotary joint, while the pickup is kept
fixed. In lining up the apparatus, the pickup is adjusted vertically so
that a cut through the peak of the beam is taken. Usually the axis of
the pickup will be in a horizontal plane bisecting the beacon.

1 Sec. 15.8.
 SECONDARY PATTERN MEASUREMENTS

15.14. Siting Considerations.In general, the term secnndarY

pattern is associated with dirertive antennas that are large, measured
in \va\-~lengths, and with a large distance hetlveen the transmitter and
receiver. A directive antenna will usually consist of a primarl- radiator
together \\-itha reflector or lens or combination of reflectors or lenses.
The secondary pattern is the Fraunhofer pattern of the antenna in
question. '~hetechniques involuted innlcasurementa rcfairlysimpleancl
closely related to those in the discussion of primary feed patterns. The
simplifying consideration here lies in the fact that only the relative field
intensity or po\ver per unit solid angle, and not the phase, is of interest.
The discussion will center about the method of taking receiving pat-
terns. In brief, a distant transmitter sends an essentially plane-polar-
ized \vave toward the receiving antenrm. The power received by the
receiving antenna as a function of its orientation with respect to the linti
of sight bet\veen the transmitting and receiving antennas is recorded
either manually or by a recording device; the data thus obtained provide
a pattern of relative field intensity or relative po!ver per unit solid angle
for the antennaunder study.
The distance between the transmitting and receiving antennas is dic-
tated by the size of the antenna being in\-estigated, Tbe site shoul(i
be cbosenlvith the largest antenna
411.Rk
/ to be in~estigated in mind. It is
fB+___ < required by theory that a plane
\ ~, ~ wave be incident on the receiving
D L4. .f. ___h___h~__ ~-$ antenna; actually, this requirement
// /
/ is met within a certain tolerance.
f / /--- Transmitter
A paraboloidal antenna as a trans-
l~r<;.15.1,5.OII tl,e pattl length for second- mitter w-ill appear to be a point
my natcerns,
source when viewed from a large
distance. If the distance is sufficiently large, then-avefront over asm:,.11
portiun of the main beam ~rill de~iatefrom aplane~vave byonly a small
amount.
Referring to Fig. 15.15, let D be the aperture dimension over \vhich
a plane wa\~eis desired for pattern measurements and R be the distance
between the transmitter, of aperture d, and the antenna to be in\-esti-
gated; then tbe difference in path length bet\veen the outer edge of D
and the center is given by
AR= R R.
Sllmming the squares of the sides of the right triangle 0.1 B ~~e have

r)
R+
()~
2
= (R + JR).
SE(!. 1514] ,s171.V(; (() !YLVI1)EI{A Tloiji5 575

If \ve require that AR shall be a small fraction of a \~avelength, (AR)

is negligible, and there results
~=p_
(33)
8AR

For a path difference M? = A/16 IVCilave

(34)

which is a safe distance to USC. ~he effect (If such small deviation fr{m~
a plane wave only slightly affects any gain detm minationsl and causes
very slight changes in the pattern obtained as compared }~ith that which
would be obtained if the \vave w-ere truly plane. 1his distance ~vill also
minimize any interaction betl!wen transmitter and rcrei~rer. ~ .krt,ually
if space or po\ver limitations do not allo\~ such a g~eiit distaure, then a
distance of D A may be tolerated. This \vill lead, in general, to an
apparent decrease in measured gain, an apparent increase in the minima
of the side-lobe structure, practically no effect on the maxima of the side
lobes, and greater possibility of transmitter and receiver interaction.
Calculations may be carried out for certain ideal cases illustrating the
semiquantitative nature of the above remarks.
In the foregoing discussion the distance R is determined by a con-
sideration of the phase dm-iat ion of the inriderlt \ra\re. Anothw fartor
to be considered is the uniformity of the power distribution over the
aperture D. If we rcqllire that the po~ver at the edge of the apertnrc
shall be a certain fraction of thr polver at the ccntcr, another criterion
for R results. In the vicinity of the peak of the brain of the transmitter
of aperture d, the po~vcr in dirert ion 6may he approximated by

=+ -2(01 (35)

!vhere @l = 1.2k/d is the fllll Ividth of thr t ransrnitting lwarn at. half
power. Then if P = O.!) P,,an(l 0 ~ 1) 2R radians at the edge of the
aperture, there results by substitution

R=~. (36)

Accordingly, if the transmitter alwrture is equal to the receiving aper-

ture, the criteri:l of dkt2LII(C for I)ropcr phase Mld for intensity over the
receiving aperrllrr are tllc s:Lnle. 1hr tramsmit ting antenna is usllall.v
smaller than t Ile receiving
alx, rl IIre; ami thcref(jre umler the previo]ls
criterion [l;q. (34)], tlIc l)()!rer at tlIe LIIlgr(Jttllc rm,eiving :lpcrture is
the same within :L fr\\-per cent as tl)at at 1}ic center of the aperture.
1S(YSW 6!1
2See Ser. 1522
 Other factors determining the separation of transmitter and receiver
are the power available in the r-f source and the sensitivity of the receiv-
ing system. Given two antennas separated by a distance R, the power
~ereived, P,, by the receiving antenna w-hen power P~ is transmitted Ivith
gain G, is given by

(37)

~vith the absorption cross section (A,) given by GEx~/47r. Here one is
really concerned \vith the smallest antenna that can be investigated,
since this w-ill be the limitingfactor on P,.
To illustrate the application of the criteria consider the problem of
choosing a site for measurement of antennas at wavelengths varying from
3 to 10 cm. Assume (1) that r-f (magnetron) sources of 50 watts a~crage
po\ver are available, (2) that antennas to be studied vary from 1 to
10 ft, and (3) that 1 mw of average power received in the bolometer
detecting system corresponds to 100 volts on the electronic voltmeter.
The limiting conditions are most stringent at the shortest wavelength;
and if satisfactory for this wavelength, they will be more than suitable
for the larger wavelengths as far as distances and power are concerned.
Also, higher power sources are usually available at the longer wavelengths.
For the above assumptions the following conditions result:

1. Specifying that the phase variation over the aperture D, which is

taken equal to 10 ft, shall not exceed k/16 for a wavelength of
3 cm, the distance R required is found to be 2130 ft by l;q. (34).
2. The maximum diameter d of the transmitting antenna can be as
large as 10 ft and still satisfy the 90 per cent power requirement
at the edge of D.
3. To read 100 volts on the voltmeter for the smallest antenna being
studied, namely, D = 1 ft, a conservative estimate for the gains of
the receiving and transmitting antennas is given by GR = ~(mD/A) ~
and Gr ==~(ml/N2. Setting R = 2000 ft in accordance \rith
condition 1, NWhave by Itq. (37) that the minimum aperture of the
transmitter is d = 2.5 ft. Thus all the conditions of the problem
are satisfied.
At a wa~elength of 10 cm the transmitter aperture required
would be approximately 7 ft in order to receive power of 1 m\vat
the detector of u 1-ft antenna at the recei~erj assuming the trans-
mitter po\rer is 50 w~tts as before.

The next mo,t importau~ condition for choming a site, \vhen the
power and distance requirements are s~tisfied, is the tibseuce of reflecting
objects, particularly buildings, If 3000 p~ttmms arc desired, then clear
s~wrounclings for 360 must be available. IJOICdistances as great as those
SK(. 15.14] ,TITI.T(; (():V,SID1<I!. iTI().V,Y 577

required for studying large microwave antennas, one us~lally must

choose a site between tllo high points, such as buildings or hills. The
interference bet;veen the direct beam from the trarwmittrr and the
reflected beam from the grfnmd may result in a poor field di~tributinu
over the aperture that, is being

m
studied. To eliminate this inter-
ference, the transmitting antenna
should be as directive as possible
so that ifs first minimum \villbe
in such a direction that even a Ground

specular reflection will not affect

the fie!d. This is illustrated in
Fig. 15.16. Since the first minimum in the transmitting beam \villm,(llr
at approximately k/d radifins from the peak, then 2}1/1? ~ k/d or

(38a)

the larger the transmitting dish the lo\\-mthe height required, and tl,e
longer the Tvavelength t}le greater the height, for a given disttinre R.
If R = 2D2/k, then

(381))

Tn the problem discussed beforv irith ?) = 10 ft and d = 2,5 ft, Eq,

(381,) \vould mean a rmluired height of 100/2.5 = 40 ft. Ho\vever, since
the criterion for the 10-cm wavelength is not fulfilled, Imd reflections
may occur at the longer ~vavelengtbs. This is one of the difficulties
involved in hal,ing one site for a large spread in wavelengths. A pos-
sible method to minimize the effect of the reflected beam is to place
absorption screens or diffrartin~ edges half\\-ay bet\\-ccd the two sitr%.
[t might appear that a reflr(tiou of 1 per cent in polver may be negligible;
ho\\-ever, one mllst remember that if \re ha~-c tivo \vtivcs of respective
powers 100 and 1, and if they interfere ronst rurt ivrly m dest rurtivdy,
the resulting variation in po!!-er received is not 101 to !)!) but rather
(10 + 1)2 to (10 1) or 121 to 81, since it is the amplitudes that add, n(,~
the powers.
The various conditions hare bcxm stated, and a sllitable compromise
must be made lwt\\-een the variuus factors involvc(ll s(lch :LS heights
and separatism of sites, sprrud uf \\-a\-elengthslleing consif~me(i, pofyer
available, sensitivity of the (lctcction system, and the accuracy desired
in the radiation pattern. In the example discuswl, it might Iw nwws-
sary to resol~-r the conflict txtiwx.n the various ron(litions by the use of
several sites.
After ihr site hm lwen rh(wm, it sht)lll(l lm chet,~c{l for l}niformity
578 ANTENNA MEA SUREMEN T.~TECHNIQ lJES [SEC. 15.15

of field. The procedure to be followed is to direct the transmitter beam

so that the center of its peak is on the center of the receiving site. A
pickup antenna, either a paraboloid or a horn, is then moved over the
aperture of interest, and the field intensity is examined for uniformity.
The field should also be examined in depth to be sure that the intensity
does not fluctuate seriously for motions toward or away from the trans-
mitter corresponding to the depth of the antenna system. If the distance
requirement is satisfied and the field intensity is uniform over the aper-
ture, it will not be necessary to check the phase.
15.15. Pattern Measurernents.-lle mount on which the receiving
antenna is to be placed should have at least t\vo rotation axes: an azimuth
axis and an elevation axis, so that complete space patterns may be
obtained without too much difictilty. Although other axes may be more
convenient for some purposes, the two stated are certainly sufficient.
Whenever possible, patterns should be taken in azimuth, since reflections
from the ground can usually be minimized. For mechanical reasons the
azimuth patterns are likewise desirable, as it is easier to make an accurate
mount for azimuth rotations than for elevation.
The transmitting antenna should be on a mount which permits motion
through an angle sufficient to direct the peak of the beam at the receiving
antenna, and there should be provision for locking the antenna in place.
For convenience, the mount should also ha~-e the property that turning
the antenna 90 for changing the polarization does not require shutting
off the transmitter source and does not change the direction of the peak
of the beam. If a paraboloid antenna is used as a transmitter, the unde-
sirable cross-polarization component may be reduced by fastening to the
aperture of the paraboloid a grating structure \vith spacing approxi-
mately 3k/8 and depth approximately 1/4, with polarization of the
antenna perpendicular to the grating slats. Such a waveguide-beyond-
cutoff grating will decrease the cross polarization of the transmitting
antenna about 10 db belo~v its normal value. 31agnetron sources have
proved satisfactory as a fairly constant, high-level r-f source for the
transmitting antenna. Modulated high-po\ver klystrons may be used
if the potrer requirements are satisfied, or an unmodulated c-w transmit-
ter, if a superheterodyne detection system is used.
The antenna under study is mounted, for example, with its dominant
polarization vertical so that the H-plane pattern may be studied in azi-
muth, The transmitter must ha~-e vertical polarization. The antenna
is adjusted in azimuth and elevation to receive maximum potver, The
transmitter is then adjusted to be sure that its peak is directly pointed
toward the receiving antenna. The antenna is then repeaked for maxi-
mum power, and the mechanism controlling elevation is Iockcd, These
last two steps should be necessary only if the transmitter has been
replaced or moved since the original siting measurements. A n:mi-
toring receiving antenna roughly peaked on the transmitter should bc
available to cherk the transmitting po]ver le~-el at any time. The vertical
axis of the antcnua mount sli(Ju!d be perpendicular to the line of sight
betireen the transmitter Jnd test antenna to ensure that the proper azi-
muth cut is takru, This condition is particularly important, m-hen
narroIv-beam antennas arc being investigated. The bolometer dctect,ion
system nrccl not be m~tche(l to the transmission line, since the mismatch
of the bolometer does not affect relative response. The antenna is then
rotated in azimuth, and the po}ver received as a function of angle is
recorded either manually or by means of a recorder. The value of a
recorder lies mainly in its speed when the effects on the pattern cal]sed by
changing \-ariables are being studied and in thr continuity of data as
}vell as in the permanent Yalue as a record. For single patterns its valllc
is questionable as far as tirw saved is concerned, since most of the time
required for antenna measurements is used in setting up the antenna and
preparing the electrical equipment for measurements.
To obtain the E-plane pattern, the antenna is returned to its peak
direction and locked in azimuth. The antenna is then rotated in eleva-
tion, and its E-plane pattern taken. I)ata taken pointing into the ground
may be questionable; to get the remaining 180 of the pattern it \vill be
necessary to reverse the mounting of the antenna. A better procedure
is to rotate the antenna {10 so that its polarization is horizontal (the
transmitter must also be rotated 90) and its E-plane pattern taken in
azimuth.
The simplest procedure for obtaining complete space cm-erage pat-
terns ~vith a tlro-a,xis rntmnt is to take the normal E- or II-plane pattern
in azimuth and then to rotate the antenna O in elevation and take an
azimuth pattern, thus obtaining the d cut. This is repeated for all
angles o of interest. Space patterns are usually taken or,ly for fanned
beams of the type used for navigation purposes. (ert ain precautions
must be observed, ho\ve\-er,in the choice of axes in the event that pattern
cuts are required for a shap~d-bearn antenna such as described in Chap.
13, The ant enna should be mounted so that the plane cent aining the
flare of the beam is vertical. The angular \\-idthsmeasured in the trans-
verse cut patterns \villthen be true. If the antenna is mounted \~ith the
fanned beam in the horizontal plane, the angular widths, now measured
in the vertical plane, are too large by a factor of sec 0, where o is the
cut angle. These conditions are imposed by the mechanical aspects
of the mount design.
The cross-polarization pattern for any desired cut is obtained by
simply rotating the transmitting antenna 90 and taking the pattern as
usual. ~he grating in front of the transmitting paral~t)l(~i(iensures that
the cross polarization of the receiving antenna is measurml :lnd not that
of the transmitter, Together with this prcxaut i(ul Iirs the addit ion:il
580 AIVTENNA MEASLJREii{EiV7,SlECH.VIQUES [SEC. 15.17

large factor of safety that if thetransmitter isasymmetric paraboloidal

antenna, the cross polarization along the peak of the beam is negligible.
The maximum of the cross polarization occurs roughly at an angle of
X/d radians from the peak and is usually at least 16 db down from peak
power.
15.16. Gain Measurements.Thus far the procedures for determin-
ing relative field intensity or relative power per unit solid angle in all
directions have been discussed. Ho\vever, for the calculation of the
transmission or reception of radiated energy it is necessary to place the
radiation pattern on an absolute basis. To do this a standard uniform
radiator is assumed, and the directive gain of an antenna is then defined
as the ratio of the peak radiated power per unit solid angle to the radi-
ated power per unit solid angle from an isotropic radiator, assuming the
same total radiated power in each case. Knowledge of the gain and the
radiation pattern therefore fixes the radiation in any direction.
For practical applications, one would like to have a quantity that
expresses the power per unit solid angle in the direction of maximum
radiated power in terms of the power delivered to the antenna terminals.
Or conversely, if the antenna is used as a receiving antenna, one would
like to know the maximum power delivered to a load matched to the
antenna transmission line of assumed zero loss when the power per unit
solid angle incident on the antenna is kno\vn. This effective gain, as
defined above, will differ from the definition of directive gain that was
used in previous chapters only in so far as it takes into account heating
10SS?Sin the antenna and the loss of power due to reflection as a result of
having a mismatched antenna. It is assumed that the same losses will
result whether the antenna is used as a transmitter or a receiver, and
therefore the receiving and transmitting gains of an antenna are identical.
Typical procedures for determining directive gain and effective gain
will be discussed, and procedures for determining effective gain standards
~vill be outlined.
15.17. Directive Gain.The directive gain of a transmitting antenna
referred to an ideal isotropic radiator is given by

peak po\ver radiated/unit solid angle

Directive gain =
total power radiated/4n

or that of a receiving antenna by

peak power received

Directive gain =
average power received
This definition does not take into account any heating losses or reflection
losses.
Experimentally the directive gain is obtained directly from the radia-
tion pattern (either receiving or transmitting). If the relative po\\cr
SE(,. 1.51s] G.4 [,+ COMPA RI,VO.V 581

per unit solid angle P(o, 1#1)as a function of orientation o and @ has been
dete~mined, then
4TP(0,0)
Directive gain = 7 (39)
T P(@, @) sin 0 de db
// 00

\vhere o = O and @ = O is the direction of peak radiation. For accuracy

a complete space pattern is required.
For an antenna with essentially a pencil-beam pattern, the assump-
tion is frequently made that P(L9,@) may be replaced by the average of
the radiation pattern in the principal E- ancf If-planes ~vith no depend-
ence on O. If P,, (6] and P~(0) are the patterns in the principal E- and
H-planes respecti~rly, then
4P(0,0)
Directive gain = (40)
~ P,.(o) sin 0 dtl + ~oTP~(0) sin 0 de

For integration purposes (planimeter or Simpsons rule), it is convenient

to lrritle 1 = 1 cos o tind plot P~ and 1,, as a f~mction of T, leading to
the equation
4P(0,0)
Directive gain = (41)
o [PE(X) + PH(r)l (ix
/
This approximate procedure has been found to be fairly accurate (within
10 per cent) for pencil-beam antennas in \rhich not too large a portion
of the radiated energy is contained within the side-lobe structure).
The experimental determination of directive gain is very tedious
and is subject to many possible errors such as incompleteness of radiation
pattern measurements, spurious lobes due to improper setting, inaccuracy
of angle determinations, improper e~-aluation of noise, and errors in
graphical integration. It also suffers from the fact that the time rc(luired
for such a measurement is long and thus rapid gain det erminat ions can-
not be made.
15.18. Gain Comparison .The best method for determining effective
gain, \~hich is the quantity of most int crest, is h!; comparison of the
antenna under investigation wit, h that of a gain standard, either on
reception or on transmission. The procedure for determining such a
gain standard is discussed later.
The experimental setup for gain comparisons of a receiving antenna
is the same as that utilized for receiving pattern measurements. I;ni-
formity of field arross the pattern mount, is esscutial for accurate gain
cornparlsons. The antenna is first peuked in azimuth, and elevation for
rnaximllrn received po\ver and the received potver P,, is noted. The
antenn:~ is thrn replaced by the standard ant mma jfhich is also peaked
in azimuth and elevation for nmximum received po\ver P. using the same
detection apparatus. The effective gain of the antenna is then

G = ~ X gain of standard.
8
If the field is uniform, the gain standard may be clamped to the mount
near the antenna under study. The reciprocal procedure is utilized for
measuring gain on transmission.
Several precautions must be taken to ensure accurate gain com-
parisons, (1) It is essential that the field distribution be uniform; other-
wise the gain comparison \\-illdepend on the relative positions of the t\vo
antennas. (2) A monitoring antenna at the receiving station peaked
on the transmitter should al\vays be available to check the constancy of
the transmitter output. If the output has varied during the measure-
ments, the ratio P./P. must be corrected for the change in power level.
(3) The same detection system should be used for both the antenna and
antenna gain standard. (4) The detection system should be matched
to the transmission line. This may be accomplished either by means of
a tuner or by a suitable matching transformer. In ~ny case, the match-
ing device should be considered M part of the detection system and not
changed during the determination of P,, and P.. (5) The readings P,,
and P, must be corrected if the response of the detection system is not
square lLW. (6) The matched gain stand:ard sho(l]d be (1irective and pref-
erably have a gain comparable (\~ithin 10 db) to that of the antenna
under study.
Assuming that the gain standard is matched to the transmission line
and that the detector is like\vise matched to the line, the gain determined
in the above proredure measures the efficiency of the test antenna com-
pared \vith the gain standard. If the mismatch of the test antenna is
known, the measured gain under matched conditions may be corrected
by multiplying by 1/(1 Irl), where I,, is the reflection coefficient of
the antenna. I In principle the effrctive gain of the matched antenna can
be obtained by inserting a tuner in the transmission line and adjusting
the tuner for maximum received pu\rer, This serves to rmphasizr the
fact that them is actually no difference between (1) matching both the
detector and the antenna to the line and (2) matching the detector to
the antenna, but owing to unavoidable losses in most tuning dcrices it is
more satisfactory to match the detector to the transmission line by other
methods and then make corrections for the mismatch of the antenna,
15.19. Primary Gain Standard Determination .Civcm two identical
matched antennas separated by a large distance R ]~ith po~ver P~ being
delivered to the transmitting antenna, hotr much po\ver \vill be received
in the terminating load at the receiving antenna? It n-ill be assumed that,
free-space propagation exists, that the transmission line lwtlveen the
rereiving antenna terminals and the load is Iossless, and that the load
, S(.e %.,.. ,j,1,5.
SEC. 1519] PRIMA R} GAIN S7.4)V1)ARII l) E7EI<.t/INA YI(),V ,-)8:3

is matched to the transmission line. The two antennas are peaked so

that maximum power is received in the receiving antenna.
The maximum po\ver transmitted per unit solid angle is given by
Pm/4T, where G is the effective gain of the antenna. The solid angle
subtended at the transmitter by the effective absorption area A, of the
receiving antenna is given by A,/R2. The po\verP, received by the latter
is therefore
P,G A , PTG2A2
P,=j%F. (42)
(41r)2R2,

where A, has been replaced by GkJ/4m.

It will be noted that losses in the antenna have been combined into
the factor G and that the receiving gain and transmission gain have been
assumed equal. The factor G determined by means of this equation
is what is meant by effective gain. If such an experiment is performed
and a number G obtained, all other antennas at the same wavelength may
be compared \\-it h this antenna and their effective gains may then be
obtained. The effective gain defined in this manner may be expressed
on transmission by

~P eak power radiated ;unit solid angle

po~ver delivered to antenna/4ir

The experimental determination of

(43)

requires the determination of R, A, FE and P7. Wavelength can be very

accurately measured by means of a \vaverneter. R, the separation
between the t \vo antennas, can be determined Ijy measurement \vith good
accuracy. The exact points between ~vhich R shollld be measurd
is a
little doubtful; howe}er, \vit,h R large (R > 2rlz,;X, corresponding to a
phase variation of Irss than A/16 over the aperture of \vidth d) the use
of the aperture to aperture distance is sufficiently exact. By utilizing a
method that involves measuring the ratio I~11~ instead of IE and P7
separately, G can he clctmmined quite accurately, With suitable pre-
cautions in experimental techniqoc G can ewily be measured to better
than 5 per cent.
The procedure for det erminin~ G is to first r.latch t \vo practically
identical antennas and match the calibrated detection systcr.~ to the
transmission line. The antennas should he separated by a distance R
greater than 2d~A and in a clear space so that reflections from the groun(i
or near-by ol)jects are negligil)le in comparison !vith (he direct beam
bet}veen the tlvo antennas, 1he setup is sho\\l\schematically in Fig.
15.17. The electrical apparatus used is the same as that used in pattern
684 A NT E,VNA ME AS UREMENTS TEC:H:VIQUhl,Y [SEC, 1519

n~~asurements. The transmitting and receiving antennas are peaked for

maximum received power. The procedure for this is first to line up the
antennas roughly and, with the transmitting antenna fixed, to adjust
the receiving antenna for maximum received power. The receiver is
then fixed and the transmitter adjusted so as to make the received power
a maximum. The transmitting antenna is again fixed, and the receiving
antenna repeaked. This should be sufficient for accurately lining up
the antennas. The r-f transmitter should be loosely coupled to the trans-
mission line by means of a matched attenuator pad so that removing the

Transmitter Bolometer
Attenuator

& A-7 h /

/\ ~~R______v
~ Coupling
Ha Am
Coupling ~
/ volt.
Monitoring tuobe to E

~lG. 15.17.On the method for t Ile dctem,ination of the gain of identiral antennas.

antenna does not affect the power output of the r-f transmitter (this can
be verified by using a monitoring probe in the line betlvecn transmitter
and attenuator). Assuming the use of a bolometer-linear-amplifier-
detection system the electronic voltmeter reading is recorded at the
receiving end and is proportional to P~. The bolometer system is then
disconnected from the receiving antenna, and the transmitting antenna
is disconnected. The bolometer detecting system is then brought over
and attached to the transmitter, and a reading on the electronic voltmeter
proportional to P, is obtained. The ratio of these readings is then
PR/P,. The usual experimental procedure of repeating the experiment
several times and with several distances R should be follo~ved in order to
obtain a good degree of accuracy. There is no technical reason for
net being able to perform this experiment to an accuracy within 5 per
cent.
Actually in this experiment G = \/GIGz is obtained ~vhere G, and
72, all bough supposedly identical, may differ by a few per cent due to
inability to make t~vo exactly duplicate antennas. In order to differ-
entiate between the t\vo, a comparison experiment, \vhich has been
described for determining the gain of an unknown antenna, is made.
Antenna G, is placed on the receiving mount for secondary pattern
antenna measurements and set up in the usual fashion, and the po\ver
received is noted on the voltmeter as P,, It is then replaced by antenna
(7Z, and the ~,oltmcter reading P~ is noted. The gain of antenna G, will
be G, = P,G,/Pz so that

This experiment then produces two absolute gain standards at the oper-
ating wavelength.
Even if these two gain standards were lossy, the result of the experi-
ment would be the determination of the desired value of G for practical
application. With no heating 10SSPSthe value G obtained in this fashion
\vould be exactly that obtained on the basis of the definition of the
directive gain.
15.20. Reflection Method for Gain Determination.A modification J
of the preceding method for determining absolute gain utilizes a single

r; -l

-T
;. ,
PT ,\ Pr
T ---
I &J--
h ~ r7--
/
P~
d 1 ://)
A- B

Mirror
1
FIG. 15.lS,On the reflection method for gain determination.

matched antenna and a plane metallic reflecting surface as shown in

Fig. 1518. The second antenna used in the method discussed in the
preceding section is replaced here by the image in the reflector. Energy
incident on the latter is reflected and absorbed by antenna A g,Ting
rise to a reflected wave in the transmission line. The ratio P,/P, is
then found by measuring the stand ing-\rave voltage ratio in the line:

P, T12 G2~2
(44)
c= ()r + I = (4?r)W

where S is the distance from the antenna to its image in the reflector.
Solving for ~, we obtain
~_4mSr-l
A ()
r+]
The practicality of the method depends on the distance S/2 that is
(45)

required from the antenna to the rnirrcr and the required dimensions of
the latter. The use of the Image antenna is based on the ideal situation
of an infinite reflector. The trite-ion for the distance S is the same as
previously discussed, S > M2/x, although expwimenta,lly distances less
than 2d2/~ have been tried without appreciable error. The mirror must
1Il. M. Purcell, A Ilfethod for Mesauring the Absolute Gain of Microwave
Antennas,) RI. Report No. 41-9,
586 A .VTE.V.VA ,lfE.i,7(RE.tf ENTL7-7ECH.VIQ1 E,S [SIX. 1!5.21

be large enough to intercept most of the main beam ~vhose \viclth is (f

the order 2k/d radians. For a square mirror of edge length h, lye ha\-e
then

For S = 2d/k, the dimension h required is equal to or greater than 2d.

The mirror must be flat to a small
fraction of a ]vavelength, at least
1/16.
The experimental procedure con-
sists of setting up the matched an-
tenna on a mount, peaking the
antenna in azimuth and elevation so
that the reflected power, as denoted
(a) by the maximum standing-wave
ratio, is a maximum. The generator
should be well padded so that it ap-
pears as a matched load to the return-
ing energy. The next step is to
measure the WilWR at a distance
S/2 from the mirror at the position
chosen, and again at a position just
A/4 nearer or farther from the mir-
ror. The x/4 displacement reverses
the phase of the returning signal,
];-ith negligible effect on its intensity;
and by taking the arithmetic mean
of the gains computed for the two
positions, most of the error caused by
any small residual mismatch in the
antenna and line is eliminated. This
procedlme also compensates for mul-
tiple reflections from the metallic
surface. The experiment should be
a
repeated at several distances. The
(c)
1:1(J,15.19. Elect r0r,lzg,,eti< honlh: difficulty in the method lies in deter-
(a) E-plm,e bevtoral horu; (IJIIf-pla,, e wc- mining VSWR accurately, since the
toral lLonL; (c) IJJral,, iclal Imnl.
order of magnitude of VSWR ~rill
lx, small. lhe mrtlod is not so accurate as the two-antenna method,
15.21. Secondary Gain Standards.The theoretical gain has been
calculated) for a pyramidal horn \vith an aperture a in the If-plane, an

1S. .\, Sc}lelkllt)ofl, /~~lCiCOltlU(~7L?liC }1fIws, l:w A-f)st

r:m(l, A-eiv Y~)rk. I!)43,
(hap. 9.
SEC. 15.22] 1A TERACTIO.V BE TWEE.V .4/VTANNAS 587

aperture b in the E-plane, and corresponding slant heights 1. and lb as

shown in Fig. 15.19. A few pyramidal horns have been compared with
gain standards at 1, 3, and 10 cm, and the calculated values have agreed
with the comparison values to within 5 per cent. Accordingly, horns
may be used as secondary gain standards having a high degree of accu-
racy. This is particularly valuable when exploring a new \vavelength
region because the horns are easy to make and are fairly well matched,
The pyramidal horn may be thought of as a superposition of an
E-plane sectoral horn and an H-plane sectoral horn. The gain is
expressed in terms of the gains of the component horns, Defining the
Fresnel integrals,

we have the following expressions for gain:

E-plane sectoral horn:

Wcz(+a+s(id] (47)

H-plane sectoral horn:

. = ~ {[c(u) - C(z!)] + [s(u) s(r))]), (48)

Pyramidal horn:

(50)
GGG%G)
(urves of (k/a)G~ as a function of b/X are plotted in Fig, 1520. [or-
responrling curves for (x/b)G~ as a function of a/A are plotted in Fig.
15.21. These curves obviate the necessity of evaluating the Fresnel
integrals (46) for most horn sizes, since the ranges of a/A, lWIX,
of Eq.
f)/A, lb/x are within the limits that have been found convenient for horn
design.
15.22. Interaction between Antennas.-The interaction between
antennas has been mentioned in the discussion of pattern and gain meas-
urements \vithout reference to the orders of magnitude involved. We
shall now discuss the interaction between antennas such as may occur
in the determination of gain. Consider, as shown in Fig. 15.22, two
mat ched systems; one a transmitter, the antenna and generator of which
are both mat,-hed to the transmission line; the other the receiver, in
588 ANTENNA MEA SUREMENTS-TECHNIQ UE,S [%c. 15.22

which the antenna and load are also both matched to their line. The
voltage across the load may be considered as the superposition of com-
ponent voltages generated by a series of waves arising by multiple scat-
tering between the antennas. Also, as a result of the interaction between

29

d
6.2

5.8
27
5.4

4 5.0

25
1 4,6

I
$
<Id

23

#
21

19
20 2.2 2.4 2.6 2

FIG. 15.20.Gain of E-plane sectoral horns as a function of bjh.

the antennas, a reflected wave will be observed in the transmission line

of the transmitter system that may likewise be analyzed in terms of
multiple scattering.
The scattering process of an antenna may be described, as in the
case of absorption, in terms of an interception area, or scattering cross
SEC. 15.22] INTERACTION BETWEEN ANTENA.4S 589

section, presented to a plane wave.1 The scattered field set up by an

antenna is directive and can be specified by a gain function analogous to
the gain function of its trunsm.ission field. Let S be the magnitude of the
Poynting vector in an incident plane wave, .4s the scattering c ross section

35

34

$
I 33

4-32

31

30

29

28

27

26

25

24

23
3.2 3.6 4.0 4,4 4.8 5.2

7A
l:cc; . 1.5 21. Gaill r)f 11-pl:tnP w<tord horns as 4 funrtion of a/A.

of the antcnnu; t}~ell the amplitl[(le of the scattered field at a distance If

in a given direction may be \vrittrn
590 .4,VTENIJ-.4MEA
SUREMENTS-TECHNIQ UES [SEC. 1522

where G, is the scattering gain function in the given direction. The

scattering cross section is a function of the aspect presented by the
antenna to the incident wave; for a given direction in space G, is like~\rise
a function of the aspect of the antenna. It should be noted that the

L Ao,Go %J,90

q d CR=3P T

FIG. 15,22.On the interaction of antennas,

scattered field pattern differs in general from the transmission field pattern.
Consider now the problem of the two antennas. Let A obe the absorp-
tion cross section and Go the transmission gain of the transmitting system
in the direction of the line of sight bet\\-eenthe two antennas; let A,
and G. be respectively its scattering cross section and scattering gain
for the same direction. The corresponding quantities for the receiver
are aO, gO, a,, g., respectively. We shall compute the voltage in the
transmission line of the receiving system in detail. Lrt P, be the total
power radiated by the transmitter in the absence of interacting systems.
The transmitter radiates a primary wave to the receiver \vitll po\rcr per
unit solid angle in the direction of the latter given by G@f/4m. The
receiver would extract from this \\-avealone the power

(52]

(52a)

at a fixed reference point in the line; a and 6 are constants of the receiv-
ing system, the precise values of which are not needed here. The scat-
tering cross section of the recei~-el intercepts the power G@,a./4mR of
the incident wave and sets up a scattered \vave carrying power pcr unit
solid angle
P, = ;;;:: ~

in the direction of the transmitter. The latter is rescattered by the

transmitter; the scattering cross section of the transmitter interce~ts the
po~ver P,A ,,/R~ and reradiates in the direction of the receiver the polver
per unit solid angle
P, 4, G,
p, = -R-y. ~.

From this secondary \rave alone the receiving antenna ~vollld abstract
power
SEC. 15.22] INTERACTION BETWEEiY- AN IENNAS 591

corresponding to a voltage
V2 = v1&,[(klm)+61, (53)
where
(54)

The factor e~c4rRjxJ is introduced to express the phase delay introduced

by the path 2R traversed by the scattered \rave from the receiver to
the transmitter and back. The secondary ~va~e is rescattered by the
recei\er; and follo\ving the process through as before, it is seen that the
voltage excited in the rccei~er line as a result of the second scattering
stage is
T, = v,p(?-J[(~~~+Jl. (55)

The total voltage, as a result of succemi~-e multiple-scattering processes,

(56)

The net powerabsorbed by the recei\erisP, = m21t~,or

the absorption cross section of the rec~i~-ing antenna has Ixwu replaced by
(zIJ= goA?;41r.
The reflected line ~vave \-oltage of the transmitter can be computed
in the same \vay. The magnitude of the reflection coefficient in the line
is then found to be

It is seen ttlat the prover al)sorhed by the recei~-er an(d the standing-\vave
ratio obserx,ed in the transmitter are periodic funrt ions of R ~vith a period
of A/.
Very little information is available on the subjec! of the scattering
cross section and g,ii> fi.mctions. To obtain an ord~r of magnitude of
the intt}raction effect ~~e shall make the crcl hoc awlmption that the
scattering cross section and gain are related in the sanle \wlyas the absorp-
tion cross section and transn]ission gain:

(59)
592 A,\-lENflA MEASUREMENTS-lhtH.l-IQ lES [SK. 1522

If the two antennas are identical, the introduction of Eq. (59) into Eqs.
(57) and (58) expresses the latter in terms of only two parameters A O
and A,. The values of A Oand A. both can then be found from studies
of P,/P, or Ir I as a function of R. In an experiment with a paraboloidal
antenna by the mirror method it was found that A, = j..l 0. Taking this
as a general estimate, the power received in the two-antenna experiment
with identical antennas is seen to vary between limits

(60)

for a displacement of x/4 in distance, If we wish to reduce this variation

in power to less than ~ db, \vearrive at a distance of R = 2d2/k [assuming
ao = 0.6(~d2/4) which is approximately correct for paraboloidal anten-
nas]. For a distance d2/k there is almost a l-db variation. This is,
therefore, another reason for the choice of R = 2d/A rather than d2/k
in pattern and gain comparison measurements, and perhaps R = 3d2/A
is required for accurate gain standard measurements. The magnitude
of the power reflected back into the transmitting antenna will be appre-
ciable and results in a mismatch with respect to the generator of
magnitude

(61)

In the case of two paraboloidal antennas, assuming again that a, = 0.6 ~-

~vefind r = 1.25 for R = dzji and r = 1.0-1 for R = Zdz, ~.
() ,

.4 similar argument for the mirror method leads to the relation that
the power received is given by

g W=U!$==+
It will be noted that in this case ~PR P. is more sensitive to distance
(62)

variations, since the correction term is proportional to the reciprocal

of the distance rather than the square of the distance as in the two-antenna
system. At a given distance S/2, taking the maximum and minimum
VSWR separated by A/4 enables one to determine G.,A. as well as GO;
experimentally of course, several distances S/2 are chosen for accuracy.
 CHAPTER 16

ANTENNA MEASUREMENTSEQUIPMENT
BY O. .4. TYsON

16,1. Survey of Equipment Requirements..Measurements on micro-

wave antennas differ in character from those carried out on most other
radar components. A high order of amplitude stability is required of
measuring equipment for the study of antennas and associated compo-
nents, whereas high accuracy in timing and frequency control are the
main requisites in measurements on other radar components. These
requirements make the design of special e{luiprnent for antenna measure-
ments most desirable.
The preceding chapter has mentioned briefly the chief items of equip-
ment used in making antenna measurements. It has been pointed out
that either a klystron or a magnetron is the most satisfactory source of
power in the micro\vave region (2000 31c/sec and higher). The reflex
klystron has a definite advantage when the pover rcquiremeut is 250
mw or less. The mult iczvit> klystron] if available at the clesired fre-
quency, is useful \vllen the po\~er re(loiremcnt is 10 \\atts or less; if the
power requirement is grefiter than 10 ivatts, magnetrons must be used.
hlention }Ms also been made of the demodulators or detectors com-
monly used, that, is, bolorneters and crystals. Some work has heen
clone \vith diodes, but as yet they are not generally accepted for use above
1000 llc/sec. Bolometers are especially useful because of their uniform
square-law behavior. Crystals arc approximately 15 db more sensitive
than a bolometer but are not uniform in their behavior; they must he
individually calibrated over the entire power range for }vhich they are
to be used. When a bolometer is used together \rith an amplitude-
modulated power source, it is best to employ a tuned audio amplifier
and stable vacuum-tube volt meter for a sensit i~e inclir~ting syst cm.
If a crystal or a diode is used, the amplifier-voltmeter cornhination may
be used \vith an amplitude-modulated source; or with a c-w source, the
detector may be connected to J mirro~mmeter or gal~wnorneter m an
indicating device. Another c-fv method is to use the crystfil m diode
as a mixer, to amplify its output at some suitable i-f frequency, and t o
use as an indicator the low-current meter in the second dctect(]r circuit.
A very important instrllmcnt in antenna Irork is tl~e autt)matic
recorder, which can be used to record my r-f amplitude as a function of
angle, position, or time.
593
594 ANTENNA ,lfE.4S UREM EATS-EQ LIPME,V T ISEC, 16.2

16.2. Sources of R-f Power.The discussion is here confined to a

relative evaluation of various types of sourcm for antenna measurements.
For details on the operation of these tubes, the reader is referred to
Microwaae .Vagnetrons, Vol. 6, and Klystrons and .Wicrowaue Triodes,
Vol. 7, of this series.
Rejlex Klysb-ons.In measuring impedances, primary patterns of
antennas, attenuations, etc., tvhere the power required is between 25
and 250 mw, reflex-klystron sources may be conveniently used, This
type of oscillator has several advantages, among which are

1. Wide tuning range with a single adjustment,

2. Electronic tuning for fine adjustments \rhen precise frequency is
sought.
3. Relatively small power-supply-and-modulator combination.
4. Ease of air cooling, accomplished ~vith loiv-~-elocity quiet-operat-
ing blol}-ers,

The circuit of a typical signal generator is shown schematically in

Fig. 16.1. It consists generally of an electronically regulated anode

Anode .+ -
d-c Square wave
v 0.01
100k
@ ~ +-
8eraZl
1
$&= Amp, Freq,
R-f =
-
_-

Single - -
-.
cavity ,
Reflector - oscdlator
d.c
v
+
.(3 -Q
FIC. 16.1.Diagram of a typical signal generator.

power supply, readily adjustable o~-er a 2-to-1 ~-oltage range, and a

regulated reflector supply consisting of a string of lR tubes capable of
producing the required maximum reflector voltage and a means (usually
a potentiometer) for varying this reflector voltage continuously from the
maximum to a very small minimum. The potentiometer in this circuit
allows satisfactory adjustment because the reflector is always negative
with respect to the cathode and draws no current; hence, no appreciable
change of resistance as a function of applied voltage is encountered in
the load element. hlodulation is most suitably introduced in the
reflector circuit. by a square-lvave voltage superimposed on the steady-
state reflector voltage, which keys (i. e., sw-itches) the oscillator off and
on at the modulation rate. A limiting tube, which consists of a shunting
diode, is used in this arrangement to cut off the positive half cycle of the
SEC. 16.2] SOURCES OF R-F POWER 595

modulation wave, thus preventing the reflector from being driven posi-
tive. Any high-vacuum diode with a low voltage drop, capable of
supporting a peak voltage of several hundred volts, can serve as a limit-
ing tube. The 6X5 is frequently used for this purpose.
This modulation and biasing procedure avoids the frequency modula-
tion that would exist if a sinusoidal modulation ~va~-e\~ere being used.
The relation between the frequency

RLdLfJ
modulation that results and the type
of modulation ~rave used is best dis-
cussed with reference to Fig. 16,2.
Figure 16.2a show-s the r-f power out-
put of a klystron as a function of re- 60 100 200 300 400
flector voltage (the numerical values Reflector voltage
indicated are merely illustrative of (a)
the order of magnitude). It is seen
that po~ver is obtained only over dis-
crete voltage ranges corresponding to
the power modes of the tube. :1
a
N-7%+7K
Reflector voltage
The frcc~nrncy of oscillation of the (b)
tube as a function of reflector volt- F1c;, 162.-C)peratirlg rharacteristi(s
age, ~~ithin any po\\-ermode, is of the of a reflex kl\->tro!L: (a) var]atio,l in out-
put r-f poxvcr with vultage; (h) chalwe
general form illustrated in Fig. 16.2b. in outuut frequerlc> with re. pert to the
freciucurs u,,, of tke maximum Power of
Since the frequency is a function of the rl]ode.
reflector voltage, a sinusoidal modu-
lation ~oltage ]vonld result in a ~vide frequency ~-ariation in the output.
Further, since the r-f power is not a linear function of reflector voltage,
a sinusoidal modulation ~vill not produce a sinusoidal po~ver envelope
but onc that is considcrahly distorted. This is avoided with on-off
squa,re-lvave modulation.
It should be noted thtit the polrer output can be s~vitched off and on
with a square wave of amplitude considerably less than the maximum
reflector volt age. This means that only a relati~cly small modulation
amplitude need l)e superimposed on the steady-state reflector voltage to
obtain complete modolat ion. For instance, as illustrated in Fig. 16.3, the
steady-state volta~c To may be some\vhat less than that required to
acti~-ate any one of the possible modes, and the kryrd voltage V added in
series to this may be justequal to the interval bet\vecn the nonoscillating
condition and the point of maximum polver. The steady-state voltage 10
is generally so chosen that the amplitude of the keying voltage does not
exceed 100 volts, axis to peak, for cornplctc modulation. The keying is
generally sust ainrd :~t a regular rate of some 100 to 2000 cps.
To illustr:ltc the opcr:~ting bcbavior of the signal generator let us
take the Sperry 419B klystrun. The anode voltage \vould then be
adjusted for 1000 volts; the square-]yavc generator \\ould be set for full
596 AN TEN,VA fifE.4SUREME,V7.Y-EQ UII.\fE77 [SEC. 162

output of 100 volts, axis to peak, if modulation is desired. The reflector

voltage can now be increased from the minimum of 25 volts until a mode
of oscillation is found. .4t the instant the tube goes into oscillation the
anode current suddenly begins to increase from its normal steady-state
value of about 40 to 4.5 ma to a val~le of perhaps 43 to 48 ma at maximum
power output. It will be noted that for any tuning adjustment of the
tube cavity and for a fixed anode voltage, there will be from t~vo to
three modes of oscillation ~vith the various possible reflector voltages in
the range of 25 to 450 volts.

-n__La_
) 80 100
Reflector voltage
120

FIG. 16.3. OII the klystron-operating point.

If, in coupling the klystron to the led, a rather large mismatch is

unavoidable, then it is desirable to use a tuner betueen t}le klystron
and load in order to obtain both maximum polrer and stability.
L)ouble-cauity Klyshorr.s.-The measurement of antenna secondary
patterns, large attenuations, etc., usually requires somcll-hht higher
po~ver (approximately 10 watts) than is obtairmble from the reflex kly-
stron. Therefore, a different source is necessary: for example, a double-
cavity klystron, such as the Sperry 4 IO-R klystron, for the range of
approximately 7 cm up~vard. .L magnetron also may be used in this
range.
A po~ver-supply-and-modu] ator combination similar to that sho~~n
for the reflex tube (Fig. 16.1 ) may be usrd for the double-cavity klystron,
with the elimination of the reflector sllpp]y and a modification of the
method of modulation. .i typical circuit for use with the 41O-R klystron
is shown schematically in Fig. 16.4,
The outpllt poiver is a function of anode potential; if the output po\ver
is plotted as a function of applied voltage, it is fo~urd that there is a set,
of po~ver modes similar to those shown in Fig. 16.2 for the reflex klystron.
The double-ca~ity klystron differs from the rcflrx klystron in that there
SE(, 162] ,?()( R(I<S 01 1{.1~ I>() lvI;R ~!)~

is very little change in the frc(lllrncy of oscillation \vith the variation

of the anode voltagr.
Since POI!rr (kws exist, in these discrete modes, it is again necessary
ordy to add a s(luare lwve to a ronst ant, d-r voltage trhen modulation is
desired. This is acromplishrd I)y srt ting the constant d-c voltage to a
l-aiuc rorrcspomling to a position just out of the (kwired mode on the
low si(lc. The s(l~lor(>-11 aYc amplitude is then rhoscn surh that, tile sum
of the constant cl-r and thr s(luarw\vave Toltagcs causes the anode poten-
tid to rise to thc opt imum ~aluc for thc power mode during the peak
position of the ~vave. In a practical case the constant d-c potential

Modulator

L
y!
q
,
Hir-1 )
~ .
Feedback loop

AmD. Frea. l 1%$=-:.r

d K
~{)~ .1 IV YIIIYIVA IlfI<Ai7[i[<I<L111{A TS-EQIJIIM l<,V 7 [SE(;. 1(;.2

nearly resonant at the same frequency. Since it is often required that

the tube be used at a frequency not obtainable with a narrow-range
micrometer tuner, a few words pertaining to the adjustment of these
cavities are in order. This adjustment can best be made by using each
individual cavity as an absorption device; a reduction in power to a
detector will then be noticed whenthecavityisad justedto exact reso-

-J

FIG. 165. -Photographof the Sperry 410-it klystron.

410 R-fiystron
I:lG. 16 .G.Schematic diagram of a circuit for adjusting cavity to resonance.

nance. Figure 16.6 shows asuggested r-f circuit. Themethodis to set

the generator to the exact desired wavelength as read on the wa.vemeter
and indicated by the microammeter. The klystron cavity is connected
to the circuit as shown, and the three adjustment screws are manipulated
until resonance at the generated frequency is obtained (if this frequency
is in the range of the klystron). It is well to note that tightening the
tuning screws (clockwise motion) will cause the frequency to increase;
the screws should not be tightened to the point of causing excessive bulg-
ingof the diaphragm. Also, inthetuning process care should be taken
to keep the tuner frames substantially parallel. When this adjustment
has been made on both cavities, they will be resonant at the same fre-
quency and in a condition to be installed in the signal generator. It is
SEC. 16.2] SOURCES OF R-F POWER 599

possible that final adjustments will be needed after installation; one of

the top tuner screws can then be moved back and forth slightly until
satisfactory oscillation is obtained.
Magnetrons.-The investigation of the secondary pattern of large
antennas requires that a long transmission path be used in order to ensure
a uniform plane wave across the antenna aperture. In view of the rela-
tive insensitivity of accurate measuring devices, this generally means
that an average power in excess of 10 watts is necessary. The magnetron
is the source of power usually employed for these power levels; at wave-
lengths less than 7 cm the magnetron is the only oscillator capable of
generating power high enough for the majority of secondary-pattern
measurements. This type of oscillator is available for most of the ranges
throughout the explored microwave regions. They are constructed in
three principal types:
1. Spot frequency, pulsed operation.
2. Tunable, pulsed operation.
3. C-w, spot frequency, or tunable.
For antenna work it is generally desirable to use pulsed magnetrons
because of the fairly high efficiency and small dimensions of both the
tube and the driving modulators required. There are several ways in
which a magnetron can be pulse modulated, but for antenna work the
so-called (soft -tube line-t ype modulator is preferred, except for powers
in excess of 200-kw pulse peaks. This method of modulation is shown
schematically in Fig. 167.
The modulator provides very short pulses of very large current to
the magnetron, which is to have a duty ratio (V dt) of about 0.001 or less.
(Here titis the duration of the pulse, and v is the repetition frequency, the
number of pulses per unit of time.) The magnetrons used in this fashion
are, of course, designed for radar use, where the short pulses of high
power are needed for measurement of echo time. However, they serve
very well for antenna work when a power-integrating demodulator, such
as a bolometer, is used for the detecting element.
The circuit functions in the following manner. A variable d-c voltage
source charges a pulse-forming network (synthetic transmission line)
through a 60-henry choke during the time that the thyratron is not con-
ducting. At regular intervals a trigger circuit drives the thyratron grid
sufficiently positive to render the thyratron conducting whereupon the
pulse-forming network discharges through the primary of the pulse
transformer. This causes a voltage four to five times greater to appear
across the magnetron, connect ed to the secondary of this transformer.
If the original d-c voltage is adjusted properly, the amplitude of the pulse
to the magnetron will be correct for operation. In this system the pulse
width and shape are controlled by the pulse-forming net~vork. The
number of pulses per unit of time is determined by the repetition rate
of Lw trigger.
In a typical case the d-c voltage may be adjusted to provide a 12-kv
pulsearrms the rr~figl~etr{~n ff)ra period of 1 psec, a thousand times earh.
second. (The applied voltage is approximately the product of the d-c
supply ~'oltage ancltlle lJ~~lse-transformerstepupratio.) Themagnetren
- ___________
60 h 1
D-c +
supply
L ___________________

la]
Trigger =
p
Freq,

Fil. trans.
115V

311K&
60 Cps

I
0-10 amp
2.5mh

I
0-50 ma -
Fr~. lG7,-Sr}wnl:Ltir dia~ram of a mmznetron signal generator.

will then pass an average current of about 10 to 12 ma. The approxi-

mate input potrer to the mn~netron, cxcl{usive of filament polver, is
1,2 X 104 peak volts X 101 amp or 120 \vatts. The peak input power
is given by
Average po{ver input = _P_ = 120
__ \v:Ltt
s
= 120 kw,
10:1--
Duty ratio vM
and the peak current is
1.2 x 105 \\atts
= 10 amp.
1.2 x -10 volts
The filament voltage is measured l)y the ammeter in the filament
circuit, \rhich has previously been calibrated for a measured potential
difference at the magnetron filament of almut 6 volts. This method is
chosen because of the voltage drop across the pulse transformer, \~hich is
dependent on the current through it; a Voltage mc~surwi at the filament
t ransiornwr vill not bc thr true voltage at the filament. Tlm magmt ron
fdament voltage shotdd be reduced to about, 3 volts for a fi-vo]t tnbp
SE(,. 163] I)l<T1<(CTOI<,V (Xl 1

after oscillation starts in order to maintain the cathode at a safe operating

temperature. This is necessary because of the large electron bombard-
ment of the cathode under operating conditions
The load to \rhich po\ver is bein~ supplied I)y the transmission line
from the magnetron mustbe very Jwll snatchwl to the line if it is not to
pull the nlagnetronf l-efl~lency excessively. If the misma,tch issufhcient
to cause a po}ver standing-~vave ratio of se~rerul to 1 ~vith J practical
length of feed line, the tube ~rill probably not owillate at all; at least it
will have poor stal)ility. Poor stjal]i]ity must, particldady lW avoided in
antenna \vork, lvhere a high order of amplitude stal}ility, of I or 2 per
cent, is absolutely nrcmsary. .ky of the undpsiral)lc effects j~lsl mm-
tioned can usually be avoi(lrd I)y coupling the magnetr(m to t}w trans-
mittingantennabv:uisht)rt :ltr:~llsnlission lill[~abi* nl(,(ll:llli(:111~fctisi})le
2 ft or Irssaml maintaining lti- ,
the mismatch belolr 2 to 1 in
power. R,
16.3. Detectors. l3olometers
h. I
and crystals are most trc(l~lcnt]y w --~
o% <
lined as detectorsin micro\vavc an-
tenna measurements. The bo- Y<
Biaspower
Iometer has the decided adyantage
(a)
that its resistance-poiver curve is
Iinear overa \vide range of po\ver, ~
whereas crystals, alth[)u~h murh 3
more sensitive than I)t)lomcters,
have a nonlinear rcsporwe.
The crystal detector is a very
b k
small barrier layer rectifier corn- ~ ~ ~
posed of a chip of silicon in ron- (6) (c) <
tact \\-ith a fine tungsten ~rire. 1:1G. 168, ~;ll:lrartrri.t irbnf:l ix,l,ll!lrtc.r
:IllfliL .iliro,l rr) st:!l (a) rc. i.t:~,,rr v. l,fj>vrr
When this element is properly in- :,i, plied t<, Ix]l,>t,,prcc; (h) l:LIV of lx.l,;{~,<>r
troduced intoan r-f circuit, a\,ery n v.. ]I<]bvcr :Il)i,ve Iu:w I,<, ir]t; (r] I; Lw of
Iml,avior I<v.. ],IIuer {C, r t> 111<:11rrv. t:tl,
sensitive and efhcient rectification
even at very high fre(lllencies results, In sl)itc of tllcirscllsitivity,l ]{)\r-
ever, crystals find compartitive]y limited application in antcnlia lllc>:ls~lre-
mcnts lxw:~lwc of thrir nonlinear charocter.

and ilIIIJliC!d pf)\\er ill :L I)(llt)nle(er, an(l it, nlay he 11(1{((1 that this rc]at ion
islincar ill the Ix)\\et r:[n~(, /), tf)fJ2. .1 ).(,l:li itjtl(,xis( sl)(, t\\c{ll tllepo\vcr
/ :ll)])lietl 10 (Ile r]w:wlili]lg (Ievi(e ati(l (lie (Illtpllt voltage 1 (lcveloped
:L(lOSS th(> ]():L(l:

1 = Ii-l,
602 AiV1EiVIVA ,tfI$A,Yl~RE,\f lt.v7.7EQUIP.41 EK 1 [Sm. 163

where K is a proportionality factor and n the law of behavior. In the

range PI to P2, n = 2 for the boiometer. In crystals for this same range

m
of applied signal power, holvever, n varies considerably. This compari-
son for a typical case is shown
graphically in (b) and (c) of Fig.
1-
Bolometer 168. As a result of this nonlinear
behavior, the crystal, }rhile about
15 db more sensitive than a bolom-
eter, cannot be used for measure-
LIII!WLJ ments v-ithout being previously

. .
calibrated and the measured data

-.
(a)

-
corrected accordingly.
R-f
The bolometer element con-

Irfnl
1 sists of a short platinum wire from

. . Y?
30 to 70 rnicroinches in diameter.
lhis clement has an extremely low
thermal capacity and because of
its very small diameter possesses a
\ery favorable surface-to-volume
L Bolometer ratio.
(b) It has been sho~vn that the re-
FIG. lG!I-CONIIILOIL lmlometcr circuits
sistance of a platinum \virebolom-
eter is linear with po\ver above a certain minimum. Therefore, when a
bolometer element is used in a circuit, it is necessary to provide a bias in
order to work on the linear portion of the curve. Figure 16.9 sho\vs the
two most common circuit arrangements usrd with bolometem.
In circuit a the proper bias is obtained by choosing the current in the
bolometer branch of the bridge Jrhich yields an initial po\ver dissipation
in the bolometer equal at least to PI of the curve a of Fig. 168. Then
if the pov.er dissipation of the holometcr is incrcascd by coupling it to
an r-f field, the consequent linear rise in resistance \rill cause the deflec-
tion of the linear null meter to be directly proportional to the al)sorbed
r-f power.
Similarly in circuit b the bolometer current is controlled by adjusting
the rhcost at to a point l~here the uncoupled po~vrr dissipation is equal
to or greater than P,. When modulated r-f is coupled t o the lmlometer,
a periodic resist ante change Ivill ocrur, can~in~ a I-arying (wrrent that
is proportional to po\ver to flolv in the transformer primary. There is
then induced in the secon[lo.ry a ~-ult~ge tlmt is like\vise prwpurtional to
po\ver.
The Iwst ~rorkal]lr rangr of nlocl~datiou fre([~lencie~ lit>+}wt\veen 100
and 20~() CpS. 111 ]Jig. 1 (j 10 iS gI:l])h(>(lI)olom(t(r S(USitiVit~ VS. fre-
quency for a wire 70 nlicroi~lchcs in dianlftcr 11 is clear from this curve
that greatest sensitivity appetirs bclo~r 100 cps but on the other hand
SEC, 163] J)ETECTORS 603

difficulties in construction of a suitable amplifier eliminate the use of

lower frequencies for this purpose.
The particular bolometer used depends on the radio frequency
involved. For frequencies of less than about 4000 MC the standard
8 AG meter fuse of i+a or ~~ amp made by Littelfuse Company of
Chicago is an excellent performer. For frequencies above 4000 MC it
becomes necessary to design the element for the particular application.

$:\, o
10 100
Frequency, cps
1000

FIO. 16,10.Bolometer sensitivity vs. modulation frea.uencs for Platinum wire (70 micro-
10,000

inches in diameter),

Bolo
). . . . . . . . . . . . . . . . . . . . d
L t-i
Resonant wind

8010meter

meter
ric

(a) (b) (c)

FIG. 16,11.Methods of coupling bolometer to r-f field: (a) coaxial termination; (b)
bolometer wire stretched directly across small dimension of waveguidc; (c) Imlometer in
housing across small dimension of waveguide.

Figure 1611 shows several methods of coupling the Lolometer element to

the r-f field. These methods of coupling along with the geometry and
loss characteristics of the protective enclosure greatly influence the
bolometer element design for use with micro\vaves.
With frequencies below 4000 MC coaxial bolometer terminations are
most frequently used. It is generally quite satisfactory at such fre-
quencies to use either a ~~m- or ,&amp Littelfuse as the r-f integrating
element. lhe choice is dictated by the operating resistance of these
units, since their reactive components are quite similar. The T&arnp
fuse has an operating resistance of about 200 ohms at 10 ma, while the
A-amp fuse operates at about 400 ohms at 5 ma.
Matching a coaxial termination to the line maybe achieved by varying
the length of the stub, which is nominally A/4, and/or the distance
1A detaileddescriptionof tile constructionof Wallastonwirebdometers may be
found in Vol. 11, Chap. 3, of this series.
604 A~VYE.V.V.i .tfl<. 1SlIREMI?)Y 7SIX! lT1lME.VT [SW, 16-I

between the center conductor and the point at \vhich contact is made with
the fuse. By adjusting these t\ro parameters properly a reasonable
match (VSWR ~ 2) may be achiered on i-in. (4(i-ohm) line for about
a 10 per cent band\~idth. If a better match is desired (YSIVR of 1.2),
it may be effected by introducing a coaxial transformer of correct dimen-
sion, but the band~virlth for \vhich this improved match will hold is
reduced to about 1 per cent.
For frequencies above 4000 31c waveguicle bolorneter terminations
are most frequently employed. These are sho\vn in Fig. 16.11 b and c.
When used directly in ~raveguide the bolometer rlement must have
axial electrodes that are parallel to the E-vector so as to achieve a rea-
sonably large coupling. The material of the protccti~-e envrlope, if usecl,
must have a very lo\v loss and in general should be no larger in di~meter
than absolutely necessary for mechanical support. It has linen found
that this envelope need not excewl T3(7in. in cliwnctrr and c:m be suc-
cessfully made of polystyrene ~rith ~lg-in. v LA.
The match in guide is influenced by the choice of thc effect i~-eresistancc
of the wire and the distance d bet Ivmn the \rirc and the sht)rt circuit in
Fig. 16.1 lb and c. With proper choice of Variables very close matching
(VSWR = 1.1) can be achieved for a band\ vidthof 1 per cent or less. HoN-
ever by usin~ a ~vindo!}- (Fig. 16.1 lc) that is resonimt at one frequency,
the impedance match may be held for a bamllri(lth of, say, 8 per cent;
the reactance of the \\-indowvaries \\-itllfrxxluency in the oppositr sense
to the variation of the reactance of the bolometer element \vith the short-
circuited ~vaveguide termination.
16.4. Amplifiers.-The power availa})le at, the detecting clement
is very small (5 mv- or less) in the majority of the methods used in
antenna measurements. For this reason the availal)le voltagp at the
detector output terminals \~illIikei[-ise bc smell. For instanre, the volt-
age at the terminals of a sensitive bolomrtcr ranges from 1[)7 to 102
volt rms over the linear part of the dct rc( ion curve. Crystals also have
an upper useful terminal voltage of almut the surne order of magnit udc,
but the lo\~er limit extends do}rn to arouml 10 volt for the au~lio
region. It is etident, therefore, ttmt some amplification is ncwlrd \\-ith
these micro~vave detectors in order that a prartirtil indicating meter lx,
used for measurements.
The re(luired performance characteristics of surh an amplifier are:
1. Linearity over a range of at lf~ast 1()() dt).
2. An inherent noise level tit least 6 db Iwlo!l the rninimllm input,
signal.
3. Good stability.
4. Freedom from rrsponsc to olllsi(lc fiel(ls.
1S.(> sec. 7.9.
SEC. 16.4] A,J{FLIFIERS 605

Figure 16.12 shows the schematic diagram of a tuned audio amplifier

meeting the above requirements which \vas designed to operate a Bal-
lantine illodel 300 voltmeter as the indicating device. This unit is
linear from O.001 to 100voltsrmsoutputj orinother wm-dsover a range
of 100 dbj ~vhich corresponds to the range of ihc Ballantine meter.

Voltage gain 10,CXJO. Rlw Rlw

Output noise level wth zero

papprOx0O05vm~Rappr0x*00k b l., mPe~::e500k

Freed 12C02 * , 6J5
1620 240 k 0,004
or UTC A-11 - ,0OH
. _ 1W Mica =
--- ---
---
1 lM ~ Thor 1C62
0.004 lW
~ IRC Mica I 1C63 or
~ UTC A-30
- 200k 2M lo 3k 1 0.5 lM 10 6A
F% Iw i m,~-+y Ww Paper ; lW &50V1W
~ ~~ ~
J__


=

Faraday cage
5fl
?5w

(&=J==J
Input impedance 200Q
I

5W4 Thor - 13C27 UTC A-30

Thor - 13R11
Ih 200 h
115V x
Ocp x
m
11/2 a

F x

All fixed resistors 1 watt

X
c x unless otherwise noted.

&
1:1<,. !(,12. 1(l,,c,l :,u<li<):iI,L!IIIficr

1}1(: ~oltagf, ~:~in is 1(),()()(), !Illich Iwrrnits (Jperalifm of tIle in!)llt from
10 to lo? \-olt.
111( t Ilr)f,(l :Ilulio :Lnl[)liflf,r is l)awl 111)1)11
t 11(IIW of :L t II ill-l RC-l)ridge
:LX u i(wll)a(k (Jl(r[l(,rlt to rwlll(,c tlI(, l):m(l!ii(~tl), 1llis r(>(lu(wl b:ln(i-
\\-i(llt! tl:~> :Ln al)pr{v,i:ll)lr clfe(( on tll(, ~igri:~l-(f)-lloiw ratio; in fact, thf
606 A,VTE,V,VA MEASUREMENTS-EQ UIPMEN1 [SEC. 164

This is equivalent to an input voltage of 5 X 10-s, or 6 db below the

minimum signal voltage. Typical performance curves are presented in
Fig. 16.13. In (a) is plotted voltage gain against logarithm of the voltage
output of the amplifier. This curve is clearly linear for practical pur-
poses, over the range from 0.001 to 100 volts. The graph in (b) is a
plot, of relative voltage amplitude vs. frequency for the amplifier, showing
the discrimination of the filter at 1000 cps.
Considerable care mus~ be taken in selecting the components for the
amplifier, also in the layout of the ports and wiring. For instance, it

200 400 600 800 100U12C01401 IWO WI 200il

Frequency, CPS

(b)
FIG. 1613.-Per f<>rrxlallre rurxws for tunrd :Lll(lic, :Il,lr)lifi.r: (~) output-g:>in curve; (5)
curve Sllowitlg <ii.r,rilllltl:Lti{)llOf fi]tcr :Lt IWO CPS.

is imperative to assure excellent rnagnel ic shielding for the input trans-

former and the output chokr; other\vise unflmimllle currents can lm
induced in the ~vindings by surrounding tield>. For a similar reason it is
necessary that ground wires be short and connected to a common point.
Tulws must have very 10JVmicrophonics rcsponsej hence the use of the
1620 tube in the first stage. Also, the first tulle should be shock-mounted.
The po~ver transformer is best packaged separately; it is not included in
the amplifier cabinet but houwf in a separate container and connected to
the amplifier by a flexihlc cal)lc about 3 ft, long. This permits the trans-
former to be placed 2 or 3 ft alvay tvhen in actual operation, a proccdurc
that is not inconvenient, since the required transformer is quite small
and light in \rcight.
The filters are construrtcd as small fixed-tuned plug-in units ~rhicll
are arranged in the amplifier for easy replticement. The frequency range
is normally about 100 to .5000 cps.
There are mtiny ~-aria tions possible \\ith this type of amplifier, most
of which are j~lst a(laptations to special requirements. Ho\vevcr, one
modification is very (lmiral)lc for iml)edance work, lt consists essentially
of the system already shown with the cxccption that the Y2LCUUIn-
tube voltmeter is built as part of tl)e amplifier J\it h an iudi.cat infg meter
placed in the front pa,ncl. ~his nlc,ter is gmlcr:Llly calibrated in volta~e
stand ing-\\avc ratio l\i(ll full-sca]c dvflcction :1s unity. Tllc (:Lli-
SEC. 16.4] A MILIFIEILY 607
bration is based on a square-law detector regardless of the type of detector
used.
Another type receiver, based on a heterodync principle, is used to
some extent in antenna work, especially in the measurement of secondary
patterns either where space attenuation is large or where for some reason
the transmitted power is low. This method is capable of considerably
greater sensitivity than is realizable with the audio method and also
does not require the signal source to be modulated. The order of maxi-
mum practical po~ver sensitivity with the heterodyne system, when using

R-f

Attenuator

o
I

Crystal
mixer

l-f / -
/ attenuator

_ Synchronizing pulse
l-f Video to trigger transmitter
Second
amplifier
detector
30-60 mc r--+ Linear sweep to
frequency modulate
klystron

E!ZIE!EI
FIG. 16.14,l\lock di:~gra,?l (If lleterodsne receiver.

a crystal mixer, is about 10lZ vmtt, ~vhereas the audio amplifim and
crystal combination has a maxim~lm po~ver srmsitivity of xbout 109
watt.1 Figure 1614 is a block diagram of a typical heterodyne receiver
arranged for measurement work.
lhis setup makes use of either an r-f or an i-f attenlmtor or both
and depends largely upon the attenuator as the measuring element.
The attenuator, which is pre~-ioudy calil)rfitetl, is a{ljustwl to have a
minimum insertion in the circuit \\-hrnthe rercivrd sign:d is a minimum,
and the indication on either of the meters or the oscillos(wpr is l~oted.
Then for any signal of greater amplitll{lc the attenll:li or is a(ljusted to
an insertion greater than the origi~l:d setting llntil tllv o(lt l)llt indicator
1Sce T[Ic7/tt7fI7ul)e .1 mp//ji(,r.s, \ol Lq (,I tllih wri~w
 returns to the ~wlue observed for minimum signal. The change in sigmd
amplitude then, of course, is the differcnce indirated by the calihratcd
attenl]atcn-. Artwdly either the c-u- meter or the \-id~o mrter may be
calibrated for a ran~e of, say, 5 or 10 db. Thus, signal diffcrencm can
be measured by the indicating meter and the attenu~tor in com})ination,
This provides a simple means Of measlwing the smallrr variations that,
may be ~vithin the limits of the ndicatin.g meter \vithout readjusting
the attenuator.
The use of an i-f attenuator as the measuring element is to be pre-
ferred if the mixer ctin be she\\n to be linear o~er the desired range to
be colcred. Ysu:Llly, for crystal mixers, this is true to about :30 dl)
above the minimum detect al)le signal tind to about 20 db further \vith
rm-efully selected crystals, This metho(l dots not require a different
attenuator for widely different fre[]uenci(,s c~fmccivrd signal but, on the
ront,rarv., functions equally JVCIIat :LOY rrgion for \rhich a suitable mixer
can he introduced. For very large polfc~ cl Mngcs in the rcceiwxf signtil
(60 db or more), it may bc nercssary to use an r-f attenuator solely or
in coml] ination ~vith an i-f attenuator to maintain the (lesired accuracy.
The hctrrodyne rircllit contains three output in[licatins devices.
( hoim \~illdepcmd upon surh conditions as rharartrr of signal and flexi-
l)ility of indication: (1) If the signal is apprrrial)ly nlodlllatrcl. either
the oscilloscope or meter nx~y lx usrd followin: thc ~i(lro :Im])lifiel.
The srope has the ad~-ant+y of intcrral-t iming, anti Illrrcfore tllc (lrsirr[l
signal and any spurious signal ran (wually bc distinguifihml \vhcn using
synchronized short-time pulse mo<lulatiorr on the transmitter. lhr
undmirml romponcnts arp usually rrflwl ions frt)nl Sllrroun(ling ol)j(,rts
ivhirh pr{xlllcc an crho of different time (Ielay from thr drsirrd signal.
l-lo\vevm, the osrillusropc is (Iiffi(ldt to calil)r:Ltc acrlwatc]y over C>TCII a
fe!y drrilwls of po!,-rr range. Thus, th[ :ltt(Ylll;Ltol 11)11s1 1)(, con(in{l:llly

rcadj IIstr{l to maintain a r~mstant (ilfirrt ion on 1hr sropr \\lNmnwas(lring

the polrer (h:lU~W 1lw mctrr ralmol dist ing~lish lwt j!(,rn the {Irxircxl
sigu;d an(l an intcl.frling sigllai Ililt is cal~~ll)le, ;1S !):1S ] )(>(>11 ])l(>Yioll>l~
pointed out, of Iwing calil)mt d I~llit( :u({lrat rly over a rangr of llossil )Iy
10 db. A synrhrollizr(l s(](lrrc is not USC(1}vith thr nwtrr. (2) If th(,
transmit ted signal is a conl inll[ms \va~r. nu altrrnat illg rl)mponrnt \\-illlx,
av:lilal )Ic at the scrond dcl rrt or. Thrrrfolcj a cl-c alul)lifier :ul(l meter
roml)ination or :L mctrr :dolm, if 1hr (k>lr[i or sigual Ire-cl is high enough,
i:; used. Tll(, owill(wrl)pe is of lit t1( :l(lf:~llt :~~ein this :[ll:lllg(,l~lcrlt, sin(,[~,
\Vitll a (I-W SOIIIW) intc~val-1 iulirlg (ullIlot 1)[, usr(l Jn(l th(,rrft)re the
(l[~sirctl signal (,:ulnt)l lx} (iislill~(lisl)r([ II.0111llIOW (,;LLIs(Ttl ])v r(,fl(,[.~iOlls.
1he i-f anlplifier for :~lly of tllr iIIrIIl if)llftl IIIC(}I(XI+ s}lt)(ll(l }1(> \\-i,lC

rn~mgh ({) allo]y for tllo ins( :~l)ility (ji tIlr st)ll](,r g(,l)(,r:ltor. ~W:L]l~
:Irl :Iml)l]ll(,r 11-1111:t :5- (() 10-.11(. l~:llltljli{llll IS 11x,(1Ifjl oli(,l.~)ii.:1~(~ ir(jrlc.
.1 ,sli~lll nlt){lilir:~li{lll of 11)1Il(,l{,rfdyl)r ~irtllil r(wlilis ill :1 syslcm
SEC. 16..5] RECORDERS 609

that is sometimes useful in c-w measurements, Here the i-f amplifier

is designed to have a very narrofv band~ridth, and the oscilloscope saw-
tooth s\\-eepis used to freqwmcy modulate the local oscillator. Thus,
there \villbe an output pulse from the second detector at the instant the
local oscillator passes through a frequency that is equal to the inter-
mediate frequency above or belo\r the frequency of the rcreived signal.
This pulse isthcn amplified by the ~-ideo circuit and applied as a deflec-
tion voltage to the oscilloscope on which the signal appears as a sharp
pip or vertical line. This pip can bc controlled and measured by the
attenuator as previously described. The advantage of this system for
intensity measurements lies chiefly in the fact that the i-f amplifier has
a quite narro\\- band]ridth and consequently a higher signal-to-noise
ratio. In addition the video amplifier is used ~vhich is less complicated
than a stabled-c amplifier. It is apparentthat any number of separate
pips can appear on the oscilloscope screen if a like number of slightly
different signal frequencies are being produced by the signal source.
Thus, a very accurate means is provided for measuring the frequency
distribution and relative amplitude of the energy gi~renoff from the source.
Units are built using this circuit and are kno\vn as spectrum analyzers
and, as such, are frequently used for amplitude measurements \vith
single-frequency sourcrs.
16.5. Recorders. The use of an automatic antenna pattern recorder
has been pointed out previo:dy in the general discussion of secondary
patterns. The follo\~ing is a discussion of thevarious rnethodsforauto-
matically plotting receiving polver patterns.
lhe simpiest recorder consists merel~ of a recording current meter.
This meter is attached to the output of a linear peaked audio amplifier,
the input of ~vhich is fed from the antenna under test through a square-
law dct ect or. If the angular coordinate is synchronized with the angular
displacement of the antenna, it is possible to obtain a very satisfactory
linear plot of po\u3r. Hoil(vf l) for any antenna of appreciable gain,
much of the desired information in~olving the side lobes will be lost,
bccausc such a linear rcrordcr \\-illnot visibly resolve po~~ers that arc
from 20 to 40 db belolv the main peak po~ver. Also, must recording
meters of this type have the styl~ls attached to a central pi~ot :il~out
which it rotates as a function ()( varying c~mcnt. This does not yield
a plot in rcct:mgular coor(lintt,cs, frcq(lent,ly gi~,ing rise to conf(lsion
Iegar(]irlg thC f3i(tUIC PO~tliL~(& It is evident, then, that in order
to record side 10Ixw an(! iw rcms(ms of clarity other methods nlllst be
contli~r(l.
l;\p(,ri(n((, h:w +IIOJII1 th:Lt t}]e rnmst desirallle s(,ale to IISC Ilith an
a(ltomatic systcm is :L (Ictibcl sc:llr (xt(,n(iillg tJ\er Q range of almllt
40 all). This ma} lx>i~((oll~~jlisll((1in t\\-odititin(t lvays. (1) The linear
recording currcn(. meter may IW drivel] by :~logarithmic amplifier \rhich
610 A.VTB,ViVA h{ EASUl<EMENISE(J UII3fl$.V7 [SEC. 165

is fed by the square-law detector, thus obtaining the desired result by

the use of the nonlinear amplifier. (2) .4 system can be built around a
nonlinear potentiometer or attenuator with servo follow-up for stylus

Signal
<m; =
ampltfier

Potentiometer
amplifier Nonlinear potentiometer

Sewo
amplifier

Drum
Selsyn
motor
.

1,lG. 16.15, -Au[lio a[ltenmz pattern recorder.

displacement. The latter metl]()(l is ,gener:dly preferred, since it does

not require u nonlinear amplifier; a(I(, IIra(,y of s~ich an amplifier is usual ].;

difficult to muintain, SYstCmS bflst:d 011 this second nletlloc~ ,,-ill be

described.
f%!. 16.5] I?ECORDER,$ 611

Figure 16.15 shows a block diagram of a typiral recorder designed to

operate on the audio voltage developed by a bolometer or crvstal used
as the antenna feed termination. It shouhi ...__
be understood, of co~lrse, that the trans-
mitter is modulafed at the pass frequency
of the audio system of the rcmmlm.
The system operates as follows: The
angular displacement of the mount to
\vhich the antenna is affixed is trans-
mitted to the recording drum by a Selsyn
generator-and-motor combination \vhich
synchronizes the angular rotation of the
drum with that of the antenna. 1 The input
to the signal amplifier varies \vith the
orientation of the antrnna; since this ampli-
fier is linear, the voltage appearing across
the calibrated potentiometer is directly
proportional to the power picked up by the
antenna. The tap-of voltage from the
potentiometer is fed into the potentiometer
amplifier, the output, of \\-hichis compared
with a constant voltage. The difference
between the latter two voltages is applied
to the servo amplifier; the servomechanism
then drives the potentiometer tap-off to
a point such as to reduce the diffcrcnce
voltage to zero. Since the stylus is attachc(l
to the same mechanism that dri~es the
potentiometer, it is displaced in a like
manner; consecluently, the stylus displace-
ment is proportional to the amplitude of
the signal.
An instrument based esmntially on the system just described has imm
built and used. z Photographs of the rontrol ond recording console and
the electronic, cabinet of this instrument, are shown in Fig. 1616. With
this instrument it is postiihle to obtain side-lobe information in fine

1C,omlac(wrncy In:ly hc :i(hi(,~(,(l in tlli+ m:lnncr, since the %lsyn gencmtor m the
lnount is +ywrml to t Ilc ll](lt ion t llroII~lI :1 pl(,{i+i[)n ~(wr tr:iill Ivitll :L stcpllp r:ttio of
ahont 1 (0 3C)w]~l tllr St,l+yII l]lt)t{~l drivill~ tll[, (Irlllll is gl,:lr[,cl doirn n likr :Inlount ;
171is mclns th:~t the [,l,(tri,:ll i,,;l((,t:(ti(s 0[ tl](, S(,lsyn sy+tt,lr] :Lrc divided by 36,
:LIICII)y tllc rlloi(,c of gmxl gt,:lrillg, IIl(,rll:ltli,,:l 1 (,rrt,rs ])):Iv I)(, 1)(,1(1(Il,itc sln:tll.
z (). A. ~ysw, .\tltl,lll):L lll:l+llrill~ l;t{l)il]l~l[tlt, 1{1, l{(,p{,rt .\-~).(X)1-4, .J:ul(mrv
1945.
612 A >MTE:V,Y.! ,~f~:( .V[~l<bl.lft<.V7tV -[<()( [1.I1I<.V T Em. 1(15

detail; a typical plot of pow-w vziriation vs. dispkmcment angle is sho~~n

in Fig. 16.17,
Alternative methods of supplyin~ the information to the servomecha-
nism make use of the r-f r:lrrim NU!-Cr:lther th:m thc awlio-modulat ed
output, of a square-law detector. (Me method use< :~ calibrated i-f
attenuator in a hetmodvne cirruii 1 in plaec of the a-f attenuator. This
setup is shown in block diagram in Fig. 16,18. Angular coordinates

Angle
FIG. r6,17..lntenna pattern sllnwing det:iil ol]tain:tl!lc wittl :Lu<liclrm,riler.

are transmitted in a manner similar to tfl~t shmvn in l~ig. 1(j] 5. his

consists of a Selsyn generator geared tu the fintenna molmt :Ln(l a ,Sely,YI
mot or geared t o the recording drum. If the antenna un(ler twt is illu-
minated I)y either a mo(lulatwl or a C-)Vvave, :Lni-f V()]t :L~fi is (kT.CIOP(Ad
at the output of a sm:LH mixrr and local os(i}lator \\-hi(h is usc(l to ter-
minate the antenna feed. MI.= i-f volt ag(~ is 1)2LssNI i hroll~ll a calil)ratcd
logwith mic attcnuut or and amplifie(l; it is thcn rcctif icd and IIsed to
cent rol a servomotor just. M is done in the a~l<lios~-stem. The Styluh,
of course, is again geared to tk wrlomotor, Ifhich drives the r:~lil)rakd
attenuator wd traces a signal amplitu(k plot on the syllrllrtmizcd gwpll
drum.
The chief adranta%c of the hetr[,](l~-ne system is t}ult it is Icry m~l(h
more sensitive to we~~ k si~n:d input<1hau Iwl,,nlc,l,,rs IJrcry~fals ,Iwcl M
) lt. J. ,S)WU)II
(I,>. IIi,r,jlr;,v[ .i]]t[],,);, I:ltf[,!ll l{r(x)!lllr% 1+11. J{(y)rt \[ lI-4-i-
170-,j,j, 1-O1. 1,j, 1W+.
Sk;(. lri5] RECORDERS 613

rectifiers, as ~vas indicated in Sec. 16.4. As a result it is possible tc

reduce transmitter po\ver by possibly 10 db (Jr to extend the range of
po~vm coverage from 40 to approximately 60 db under ideal conditions.
The limitations to this method arise in the mixer, which is often not

Selsyn generator
Antenna +

I M,xer and local oscillator

Detector &
comparator

Selsyn
motor

A
Servo Servo
motor ampllfier

II(; . 16.18,-- .Iut?nna pattern rernrder utilizir]g ~n r-f attenuator in a hctermi~l!e circuit

completely linear over the higher portions of the desired po!ver range.
.41s0, diffimllty sometimes arises from the large hulk of the n~ixcr-oscil-
later termination \\flichm~lst be attached to arelativelysrna]] antenna.
Bolometers, ontheotherh and, are more adaptable because of their size.
 Index

A Antennas, strip reflector, 484

surface, for air search, 465
Absorbing material, 560 transmitting, equivalent circuit of,
Achromatic doublets, 410 37-40
Admittance, characteristic, 26 Aperture, angular, 416
normalized, 26, 213 circular, 192195
terminal, of waveguide, 366 far-zone region of, 172
Admittance characteristics, ~f E-plane gain flmction for, 162
sectoral horns, 369374 near-zone regiou of, 170
of H-plane sectoral horns, 37*376 rectangular, uniformly illuminated.
AFC, 527,538 180-182
Air-to-surface search, 466 Apertur eblockIng,effcct sof, 190-192
Aircraft antennas, 512 .4perture efficiency, 178
Amcraft mstaliations, iutcrfcrcrrcc effects Aperture-field distributions, moments of,
in, 515-51tl 184
41tar, W., 556 wparablc, 182
Altitude circlc, 514 .kpcrture-ticld method for reflectors, 158-
.lmp!ificm, 60460<9 160
audio, 60,5 .lperturc gsiu, maximum, coudition for,
Antenna feed, definition of, 12 177, 178
Antenna mismatch, dependwrcc of ah- Aprrture rel:itions, optimum angular, 424
sorption cross section on, .553 .%pcrture trmtmeut, of radiation from
Antenna pattern recorder, autonlatic, horus, 357
609613 of radiotiou from wavrguides, 334-336
.4ntenna system, 511 .4rmy designs, broadside, 318-333
Antennas, heavertail, 4.53, 477 clld-fire, 31&318
cheese, 45!) uonrcsonant, 328333
conically scannirlg, 347 .Irr:iy ckmrnt, Iongitudiualk polarized,
ground, 510 302, 30!), 312
half-beacon, 460 ::lot-fed dipclc as, 28z-286
interaction bct}vwn, 5875!)2 dots M, 2!)1301
pencil-beam (see Ieucd-beam :tj,tc,,u:{sj streamlined, longitudinally polarizrd,
pill-box, 459464 312
impedance correction fur, 463 translcrselyp olarized, 310, 311
receiving, equivalent circuit of, 4C45 tridipulcradiator as, 304, 305
scanning, 513521 wz~rcguicle radiatOrs as, 301303, 329
shaped-beam, line so~lrccs for, 4954!)7 .Irrayfzictor, 260
secondary patwru uf, m(msurement Arraysj hiuomial, 269
of, 579 broadside (see Broadside arrays)
ship, installation prol]lt,nls{]f, 511 en[i-tire, gaiu of, 277278
shipborue, for stltf:irc SC:{II,II, 467 horn, 319
skirt-dipole, 2-K-2+2 Iinrar, pattrrn synthesis by, 279-284
615
 Arrays, polynomial assori:itcd ,vith, 26- Boundary conditions, grncr:d formulation
264 of, 6668
Tchc+yschrff, 282-284 at infinity, 8486
unifornl, 264267 for srztt~,rillg prol)lcms, 13CK132
uniformly llllinlin:~t{,(l, 267269 Box horns, 377-380
Austinl P. \[., 186 l)r:{llrdich, ..1,, 460
Azin)llth r:Ltio, 304 13reckcnridge, IL. C,., 399
Brccn, S., 251, 437
13 ]Ircwstrr anglr, 402
Brillo!lin, L,, 42
Bahincts prinriplc, 167 Bro:tcll>:irlcl-norm:tl-firi,,g arrays, 331-333
Itack.lohc int(,rf,r{n{,., 42g, 435 llrcm(lsi(lc arrays, binomial illumination
Back-srt~ttrrin# c(wffici(,,,[,, 468 of, 269
Baker. 1;. 1~., 1(M gal)le(l illuminzt,inn of, 269
13arld\vi(ltll, of dirl((tri( lrIIs, 3!)8 g:iin of, 271-274
of In(,t:ll-plltr 1(?11s,408410 m:t~im(lrn g:{in coll(litions for, 270, 271
Barker, (, B., 139 rrsona,,t, 321-327
Ii;wro\v, ly. 1.., 349 llrownlo~v, J. hf., 399
Beacon antrmu~ systems, 327
lkacon :lntrnn as, 011 :tir(mft, 521 c
Bczcon clrrn(,nts (see Gladiators, axially
symmctricd) ( :,mpt,cll]s loa(lml-line forrnldas, 313316
BP:LCOI,pattern mc:ls,,r(,mcnts, 573 Carlson, J. F., 410
Bcaron radome (Iesigrl,s, ,540 ( inter of fcrd, 239, 343, 562
BCaIn) lw:~vcrtnil, 477, 478 d(t(,rmin:~tioll (jf, 564-570
brcmdsidc, 267-27+ Chwdey, F. G., 39!)
cn(i-iirc, 274278 (hisholm, E, R, 331, 332, 509
fanne(l, 7 Chok(, 242, 243, 245
appIic:~tions of, 450 Chu, I,. J., 80, 146, 248, 334, 341, 349,
iowaltitllclr, 484 357, 41.5, 478, 497
omrlichrectiomd, 6 Clapp, 1{. K., 20!), 46!)
pencil, 7 Coaxial li]l(,s, 217226
sector sII;lP(I1, 47,5 cwca(lc transformers in, 221-?23
slI:\pI,(l (we SII:Lp(Yl IM:LIII) pur:~llrl stubs for, 223-22,5
toro;[l:ll, 6 writs rcact:mcc transformer for, 225
11.,:1111(Itv,:,ti(,rl f: Lctt]r, .488 Tit;.11-,,)o(le of, 217-219
11(,:,,11Sh,lp[,, Iwrl:llll(, 508-509 ( 01,1:1 h)hcS, +88
II,,:,,,, sll;ll]i,lg, I,J- ,, Iw(:Lc1(s i,, lI(,rTI :111(1 ( ?,,,(10,,, h:. [-., 415, 423
\\:Lv(,-KIII(lr :1])(M lll[>S, :380 ,38,3 ( (,II(III(I;IIL{,(,,il~trt.]~~,j,t:d, 2!)8
Jh,;llll\rl(lll)s Ys. :I[wrttlrr ill\l!llill:Ltl,,],, (O1)M,ll. 1:. T., 108, 167
17!), 183187, 1!)5 ( U([l]liriK, t]rtwm,n tr:msmittcr and rc-
B(,(hr, II. ,[,, 208 cci ,(r, 45-48
IIiro]]i(al I1(,uI, !) (+ovcr:(gc Imttern, onc-wwy, 4
llIIN,l;Lr tl{art>, 3336 t \W-\\-:lj,5
IIir(llar[l, 11. 1,,, 31(5 (<ross pol:lriz:ltion) of bilrr(,l-ucflc,l, t(lr
I{list,. r, 512 :Illtrllll:lj 503
Jlol)n(,rt, ,J., 4!2, 1.35, 564 Of f:~llllr(l 1)(,:1111 :Lntrl)]lm, 4.57
13010111( tin, ,555, 601, (X)4 lll(:ls(lr (ltl(,llt of, 37!1 580
lh)l(,lll(. t(,r (,1( Ill(,llt, ,$1!) [If [x.jl(,ii-l, (,:ltll :,nt(,llll:\s, .419, 423
Ihmk(,r, 11. (~., 107, 2!)4 ( rww s,vti(,]l. :(llsorpi ion, 3, 42
Born, 11., 125, 166. 167, 197, 414 :lv(,r:1~(, for ]Il:[t(h,.(1 SJ+VII1, 50, 51
Cross section, deprndcnc(, of, on autenna I)ouhle.dipole system, theory of, 101 - ] 04
mismatch, .5153 on waveguide! 2,552,56
interception, 46S Doublet hYIS, 410
radar, 468 Do\vker, Y., 523, 556
receiving, 3 l)unbar, A. S., 480, 487
scattering (see Scattering cross srrtion)
Crystal, 555, 559, 601, 604 1;
calibration of la\v of, 552556
Cscz (1pattern, 466, 470, 507 E-plane, principal, 103
Ckdlcn, A, 1,., 295 E-pkitlf~ pattcrm, principal, of pcnfil-
Current distribution, (liscolltinllo!ls, 146 hearn antennas, 422-423, -433-437
149 E-plane sectord horr,s, a(lrr,ittanm. rl,zw-
far-zone fields of, 87-91 actcrist,ics of, 36!)374
over reflector, 144--149 modes in, 350354
Cut paraboloid, fred-tilt effcrts in, 488, mouth admittance of, 369
489 radiation from, 357, 3.58-365
offset feeding of, 454 terltll-po\vcr \ridths of, 364, 365
for shaped hca, rs,477483, 487-491 throat transition, 3b!)371
for simple fannrd brains, 451-457 trz,)srnissiomline equations for, 366-
Cutler, (7. C,, 484, 4!)7, .508 36!)
Cutoff wa~rrlrngth, 205 Eaton, J. E., 316
Cylinder, paraholirl 457-459 Edge diffmrtion, 516
Edw, reflection, 518
1) I?is(llh:wt, I,. ],, 142
l>lwt,ric:d length, 550552
IX: Brcttcvillr, .f. P., .Jr., 399 Ilson, X,, 442
Dehyc, I., 114 Emdr, 1., 1!M, 220, 233
Detection, sllpcrll~,t(,rrj{lynr, ,5.50, 578, Error, c(~l)ic pllasr, 189
607-60!) lin{mr, 188
I)rtwtion system, calibr:ttion of, ,5!52-,556 quadratic, 188
I)etw,tor, lmlomrtcr, 55.5, 601-004 Evcrbart, 14;., .523
(,ryst:d, 5+!), 554, 555, 601, 604 ]vrritt, IV. 1,., 17, 1+21
I)ctwtor rcspol,sc, 601-603 ];y~(IS, 1,, ,1., 316, 328, 496
l)itki(,, IL. 11., 200
I)illon, R. 1;., 316 1
1)ip<)le, cl[,(tril,, !)295
half-\ vav(, !)8 Far-zone fields, of c,lrrcnt (list riln,t io,,s,
gain of, W 87-!)1
magllt,tic, !)5 of Ii]lwcurrl>rlt {Ii+tributiolls, 96-!)8
sh)t-f(,(l, 245248 Far-zonr rc~ion of :Llx,rt,,r(w, 172
:1s arr:iy (lcmrnt, 284286 l+cml, cc]lt(r of (.s{>,( [,lltcr of f(,~,[lj
sphrroifl:~l, 24!) dipol(,-tlisk, 251- 2.;3
i:np(, (lanr(hof, 24!) clolllllf.-[lip{jl(,, 2<5;32.56,434
I)ipolc-(lisk f(wls, 251-2.53 d(,lll)k-s]ot, 348-34!)
I)ipolc frc(ls, (Iirc(tivt,, 250 Cxt(m(l(,(l, 473, 474, 177, 4874$)4
I)ivrrgcl)cc f:~ctor, I43 dipl}lt,-:LrriLJ-, 4!)14!13
I)odds, ,J. \V., 2!)5, 2!)!) horrl-arr;Ly, 4!13- 4!) I
I)OIIlh, (, I, , 282 follr-[lip,~lr, 256
I)r,llbltwlipolr xystrm, o,, r,,:* xi:,l Iinc, poillt-so(lrcc (.wr Iflillt-sourrc feed)
253-254 ])ri, r,:try, 12
r{J:L\i:il-liri{,.f,,,l, 253 Feed requirements, primary, 239-240
618 MICROWAVE A ,VT.V,V,VA THEORY A .Vl) DESIGN

Feed systems, front, 448 Gain comparison, 581

rear, 347, 434, 448 Gain determination, reflection method
Feed tilt, impedance changes with, 488 for, 585586
Feed-tilt effects, in cut paraboloids, 488, Gain factor, 178, 425
489 vs. aperture illumination, 187, 195
in paraboloidal reflectors, 487, 488 Gain function, 2, 90, 94
Feeding coefficients, 259 for aperture, 162
Fermats prinriple, 122-125 scattering, 590
Field equations for cylindrical wavegllide, Gain measurements, 580-586
201-203 Gain standard determination, 582-585
Fields, dominant-mode, of sectoral horns, Gain standards, secondary, 586
352, 355 Gardner, J. H,, 481, 484, 485, 508
superposition of, 99-101 Goldstein, H., 320
time-periodic, Maxwells eq(mtions for, Grating, quarter-wave, 447
68 Grating reflectors, 449-450
Fkke, M. D., 387 Greens theorem, scalar, 108
Fourier intc.gral representation of l:ra~Ln- vector, 80
hofer region, 174-175 Cround antennas, 510
Frank, N. 11., 173, 174 Ground target arm, effective, 469
Frank, P., 110 Guilleminj E. A., 17, 18
Fraur,hofer region, 16&162, 172 Cuptill, E. If., 295, 299
criteria for, 198, 199, 561, 57.!
Fourier integral representation of, H
174-175
Frequency control, automatic (see AF() H-plane, principal, 103
Frequency pulling, 538 H-plane patterns, principal, of penril-
Fresnel reginr., of circular apertllrc, 106- hmm antennas, 422-423, 433437
199 H-plane sectoral horns, admittance char-
general characteristics of, 171172 acteristics of, 374376
Fresnel zones, 196 modw in, 355357
Front-feed syste,ns, 448 mouth admittances of, 375
radiation from, 35 S365
G tenth-power widths of, 365
throat transition, 375
Gain, 3, 90, 177 Half-power width, 94, 104
absolute, 90 Hansen, W. 1$-., 278
of broadside arrays, 271-274 Hayes, }V. D,, 419
maximum, condition for, 271 Hegarty, Nf., .523
of circular wavegLlides, 340-341 Height finding, 467
dependence of, on tipert(lre illumina- Heins, A. E., 410
tion, 177-178 Hiatt, R., 251, 437, 442
directive, 580-581 Hill, J. F., 503
effective, 583 fIOrn arrays, 319
of end-fire arrays, 277278 as extended feeds, 481, 493-494
of hdf-wave dipole, 99 Horns, biconical, 9
of parabolic cylindrical antennas, 458 box, 377-380
of pencil-beam antennas, 423432 compound, 350, 376
of pyramidal horns, 587 pyramidal, 587
of rectangular wavegui{les, 346 sectoral (see E-plane sertOr:Ll bor]ls;

of scattering pattern, 468 H-plane sectoral }Iorf]s)

of sectoral horns, 587 11~111,C. F., Jr., 415
 INDEX 619

Htlntrr, L. P., 556 Klystron, reflex, 594-596

Huygens-Fresnel principle, 108 Sperry 41o-R, 596
Sperry 41913, 595
I Kock, W. E., 402, 405, 406, 410, 442
Krutter, H,, 442, 455
Illumination, aperture, beamwidths vs.,
179, 183187, 1!35 L
dependence of gain on, 177-178
gain factor vs., 187, 195 Lament, R, I,., 200
side lobes vs., 179, 187, 195 Lattice factor, 105
binomial, of broadsirlc arrays, 269 Lens, dirlwtric, 389-402
gabled, of broadside arrays, 269 attenuation in, 399400
Images, of dipole radiators, 135137 bandtridth of, 398
principle of, 132-137 frequency sensitivity of, 398
for dipole-sources, 135-137 one surface, with elliptical contour,
Impedance, characteristic, 23, 216 302_3~3
normalized, 25, 213
with hyperbole contour, 390392
of spheroidal dipoles, 249
reflections from surfaces of, 401
transverse wave, 209
tolerances on, 400
Impedance changes, with feed tilt, 488
two surface, 390, 394
Impedance correction, for pencil-beam
zoned, 395398
antennas, 443448
doublet, 410
for pillbox antennas, 463
metal-plate, 4024 12
Impedance effects, in normal-incidrmse
achromatic doublets, 410
radome, 537
bandwidth of, 408-410
Impedance-measurement apparatus, 547
one surface, with elliptical contour,
Impedance measurements, probe errors
403-405
in, 548, 556
of parallel plates, 402405
Indicator, response of, 549
of parallel wires, 406
Intensity measurements, primary pat-
polystyrene-foam bon[led, 406
terns, 561564
reflection from surfaces of, 410-412
Interference effects in aircraft installa-
tolerances on, 407408
tions, 515519
spherical, 390
Iris, trial, 552
l,cwis, F. D., 349
Isotropic spherical waves, 78
l,ine,, corrugated, 319
10SS1CSS,26-29
J
I,ine-current distributions, far-zone Iirlds
f, 9&98
Jahnke, E., 194, 220, 233
Junction effects, in waveguidm, 214, 215 I,inr-source primary pattrrn, 570572
Line-source feed, and reflectors, 151-154
K I,ine sources for shaped-beam antennas,
495-497

Keary, T. J., 455, 461, 464, 502, 508, Littelfuse, 549, 559
564 Littelfuse Company, 603
Kelvin, hrd, 121 Loaded-line constants, 313-316
King, .4. P., 349 Loaded-line formulas, Campbells, 313-
King, D. D., 248 316
King, R. W. P., 21, 248 Iossless line, 2&29
Kingsbury, S. hi., 399 Low-altitude cscj@antcnnas, 484-486
Klystron, double-cavity, 596599 Luneberg, ILK,, 126
620 .VICIK)WT.LVEAN7E.V.VA THEolil .I.V1) 1) E5JG.%

M Pamm(,tr:rs, ronstitutivc, 65
current, 210
Mmfarlane, G. G.,503 Voltage, 210
l.lchfillan, E, B.,523 Pattern, :,l,p:ire,Lt, 526
Llagic T, 572 primary (see Irinl:u-y p:lttrrll)
NIagnetrous, 599601 second:wy (see Sccond~Lry pnttrr!l )
JIalus, theorem of, 126 Pattern [Iistortiom by r:ldomm, 524-,526
Narcuvitz, N., 200 Wttrrn synthesis by linear a,rray, 279
kfarshall, F. f3., 556 284
hIason, S, J., 377, 380, 453, 564 lcI1cil }]cam, rcquiremrnts for, 413, 41-1
Matching of waveguide and horn f(,cclsl Pcllcil-h(,:,m ;ultrun: is, ,Ising circilhir p;i-
383387 raholui(lal rcflwtor, aperture fic,l{l
hlaximum-powcrt heorem, 20, 21 of, 419
Ilfaxwells equations, differential form, 64 dcsig]l procrdurcs for, 432
integral form, 64 cross polarization of, 41!), 423
fortime-periodic fields, 68 gain of, 423-432
Mimno, H. R., 21 impcdmcc mrwction for, 443448
Mises, V., 110 paraholoi[l~l, 41 5450
Mismatch, introduced by reflcrtorj 155- using partllmloiclal rcfkxtors, alltt,llrla
158, 439443, 454 gaiu of, 423-433
kfoments, of aperture-field distributmlls, Perry, If. .L, .523
184 lh:,sr constant, 23, 205
&fOntgOmery, C. G., 200 Ihasc c]etorlllirl:ll
iorls on poi[]t-sol]l(,c

Mueller, C. E., 316 feeds, 564570

Phase-error effect.s, on pcn(, il l)~:t]]~ g:lil],
~
430
on secoudary p:ttterlls, 1861!)2
Kabarro, F. R. N.j 415 Phase errors, (Icfo(llsing, 432
Nacelle, 512, 513 Phase mcasurcmcntj linwlcn~tli rffmt ill,
Near-zone region of apertures, 170 56{)
N-etwork, four-terminal, 1719 Ih:w,-reversal tc,huiqucs, i,, :Lrr;,ys, 321
twO-termirl&p:Lir, 17 Phillips, H. B., 62
A-etwork equivalent, four-terminal, of lillho~ dcsi~n prol,lrm., 460-404
section of transmission line, 36 lipp;~rd, A. 11., 442
Network parameters, four-terminal, 17 Point-so,lrre, cone, 240
Nowak, W. B,, 253 Point-soiiwc frcds, prim:iry lJ:It(I,rII ft,r,
557-570
0 and rcfic.tors, 14!)- 151
Puktr dit[grwn, 3, !10
Obsruratinn, of target signal, 520 Polwi7,:~tiou, cirrll]:tr, !)2
offset feeding technique. 453-457 cross (srr ( row pol:triz: it ion)
OpticdFrrwnei field, 171 clliptic:d, !)1
Optical path Irngth, 122 Iolyuolui:il, assolint~,(l \rittl /Lrr:ty, 2til-
Orthogonality of waveguide mode., 207- 264
209 1[)01(, .1. IL, 461, 508
Oster, G,, 399 Porterfif] [l, (. 1:., 478
Pol}-cr mwlm, 5!)5
P Poylltillgs illcor[,lrl, 6!)-71
loylttin~ vct,tor, 6!)
n-section, 19 t ime:lveragv, 70
lb, (. s., 381, 383, 453, 477, 487 WI, 514
 I.\[)EaY 621

IJrcssurizing, 383-387 l((v,iprt,rity rcl:,tior, I](,t!rcc,l trm,sft,r co-

lrcssurizing d(>vicc, 376 cfhci(jlts of )]ctirf~rk, 18
Irinl:lry f(d, 12 I{r[lllell(, r, 1{. 1[ , 523, 556
lr,ru:lry p:, t tern, 12 l{cfltvti{]ll, from cl[rvcd surfaces, 138143
linr-so~lrce, 570572 froll) rtlrvl,l llI1(imsllrf;L(P 0[ fuw+lglj,
for point-source frc(ls, 557,570 ;17
]rim:Lry-p:tttcrn app:Lr:LtIIs, sitill~ of, 560 rtl~f,, ,318
Protw, re.ponsc of, 549 ll\w of, 124, 134
Imhe vrr(jrs, in inlpul:mrr III1,:I>IIrlX- l{(,flt,tif,l] rf,(flific]lt, mlrrent, 32
mrnts, 55fi (,l(,ttri,-f,~,l~l, 212
lr[)ptigilti~)ll ronst:illt, 20,5 of l);~rilt]c,l,,ifl:ll r(,flrl,ttjrs, 43!144;+
lurc(ll, 1;, \t., 200, 301, 330, 4(N, 585 of r:l(lolll(> W;l]]s, 52!)- 537
W)lti[g(,, 25
l{tfl,rti,,,, r[,{, ffiri{,nt cl,:irt, 2!)-33
(,$w ct/,so Smith ch;,rts )
lhular cross swtion, 468 l{(,fl([,ti,)]] lllt,tll~xl for g:~ill {i(t(~rlllirl:l-
ltwlar sllIt{l,)\v, 5 I5 til~ll, 585
It:uli:,tif)n, from E-l)lltlc scrt(,r:,l Il(,rtw, l{(fl(,{t(,r :Lnt(mn:ls, (Ion}) 1(, -(llrvtittll(,,

357 502-50s
from H-pl:mc scrtor:d her,,., 3.jX I{(fl((tor Illislll:ltch, 155-158
from \vavcguide, :lpf,rt(lw t r(:lt lll(l]t ,~f, l{, fi((t(,rx, :11)(,utllr(.fi(,l(l ]11(,tllocl f{lr,
334 1.58-160
ltadiation conditions, 85 1):1,,(.1, 480, .503
Ii:t[liiltiotl pattrrn, !)() c(~lltotlr ((lttitlg of, 453457
l{:uli:~t ion r(mist:ulcc, 39, !15 Cllrrf,tlt (Iistrllll[ti{,]l over, 14414!)
I{atli:ltors, :iki:dly symlll,t ri<:ll, 303-30!) cllt l):\r:limloi<l;(l, 4.51-4571 477
(Iipole, !12!)6 syt]ll,]f,tric;ll tyfm, 4514,53
images o f, 135-137 Cylill[lric:ll. for shaped h(~ilffls, 494 4!)7
strc:LmIinc(l, 310313 gr:ltirlg, 44!)4,50
wavc~lli[lr, 301303, 32!) lil,(-s{,llr((, ft,,,<l :,,)(1, l,jl - 155
Rmlio It[wmrch I,alwratory, 223 Illi,tllateh irlt UX1,,C<,C1l)y, 155-158,
lki~lomr, normzl illci(lonvr. X37 53!) 43!)-443, 4,54
design collsi(l{,rilti(,lls for, .538 IIlo(lific[l, 474, 477, 47$)4!),j, 503
impmlatlrc cffwts 011, X37 l~:lf:~l)[)l[]icltLl, f~,(,{l-tilt (,li(,lts iii, 487,
p;lttlr,l fffccts of, 537 W?
patttrn (Iiht,)rtions hy, 524--.526 St[ll[,t(lr:ll flmi~n l)lt,l)l,,t])s, 4484,30
str(,amlijlctl, 524, 535, ,540-.542 pt,i]]t-w,~lr[x, ftvxl i~ll[l, 14!1 1.51
{l~wigll c(~]l<i[lt,r:itifJIls for, ,342 S(,r(,,,,,, 44!)
pxttcrn (ff(,(ts [if, 540 silo\ (,1-, 4X1 , 483
ll:ldollll. \v:dls, rcficctif)rl rl, rfliricllth of, l{,ir:i,ti,,,,, l:LI,S Of, 125
52!1,537 1{(.1)011([(!, 327
trimslnission coefficients of, 52!)- 537 liiljl[,t, 11., 13!), 247, 282, 30.5, 316, 323
Itangc rwill[tio]l, 524 T{ic(, S, ()., 247
nay cl(rv:lt(lrc, 111, 112 I{i,l[l][,((r, 1,., 414
I{ay velocity, 110 l{lllg .I)llr, (, 3.18
l{:tylri~ll, T,ortl, 1!) I{iwtr, J l{., 3;32, t3(if}, 3(i!l, .l(il, 4!)(), ,508
It:lylcigh reciprocity t,hcorcm, 1!)
l{~,:lr-fwd systcnls, 448 s
ltw(,ivil)g p:ittx,rn, 4
ItL,cilJr(wity, Iwt]recn tranfilllittl]lg :[tl(l S{;il;lr, (~rc(,l]s tl,,~~r,,]]), IOX
rlwriving patterns, .+8-,>() ,St:ir,,,i,,g :(,,t~,,,ll:ls, 51;3-521
622 MICRO WAVE ANTENNA THEORY AAD DhSIGN

Scatteringr multiple, 588 Smith charts, 29

Scattering cross section, 5, 42, 468 Snells laws, 12%125
equivalent sphere. 5 Sommerfcld, .\., 59, 1]4
Scattering gain function, 590 Southworth, G. C,, 349
Scattering pattern, gain of, 468 Space arrays, re,qllar, 104-106
Schelknrroff, S. A., 248, 261, 279, 282, Space fmtors, 89
586 Spatial pattern measur?mentsj 563
Schwartz inequality, 177 Spmccr, R, C., 139, 184, 186, 191, 194,
Screen, absorbing, 560 473, 497, 503
Screen reflectors, 449 Sperry 4 10-1{ klystron, 596
Secondary pattern, 12, 169 Sperry 4191f klystrun, 595
general features of, 1751?.0 Spherical waves, isotropic, 78
phase-error effects on, 186-192 Squint, 245, 438
Sector blanking, ,517 Stabilization, 508
Series reactance, on coaxial lines, 223- Standing-wave measurements, 548-550
226 Stan ing-wave ratio, 28, 545
Shaped beam, applications for, 465468 power, 28
cut paraboloid for, 477483, 4874tJ 1 voltage, 28
cylindrical reflectors for, 494-497 Standing-wave voltage ratio, measure-
requirements for, 465468 ment of, 547549
Ship antennas, 511 Stationary phase, principle of, 119122,
Shipborne antenna for surface search, 157
467 Steele, E. R., 523
Sichak, W., 230, 231, 255, 301, 330, 436, Steenland, .4, 11,, 496
464 Steinbergerj J., 331, 332, 496
Side lobes, 176 Stergiopoulos, ~. G., 479, 490
vs. illumination, 179, 187, 195 Stevenson, .4. F,, 292
Signal generator, 594 Stratton, J. A., 57, 61, 67, 78, 80, 93
Silver, S., 155, 157, 441, 442, 463, 487, 132, 146, 15{), 248
497, 503 Stubs, pardlelj 223226
Siting, of impedance-measurement app:t- Suen, T. J., .523
ratus, 548 Supcrpositinn principle, 66
of primary-pattern apparatus, 560 Surface-to-air search, 465
of secondmy-pattern apparatus, 574 .Surface antenna for air search, 465
578 Symonds, R. ,J., 612
Skellett, A, M., 396 Syng[,, J. L,, 110, 122
Skin thickness, effective, 534
Skirt-dipole antenna, 24W242 T
Slater, J, C., 51, 173, 174, 200, 2] 1
Slope, 549 T-section, 18
Slot array, axially symmetrical, 3115 Taggart, lf. .\., 497
Slot radiators, impedance in wavegui~]e, Tar@ response, directional, 468-471
287-299 Target signal, obscuration of, 520
Slots, dumbbell-shaped, 296 Tchebyscheff arrays, 282-2S4
longitudinal, 291-297 IW-mode, of circular waveguide, 233
nonradiating, 287 of coaxial line, 220
probe-fed, 299-301 of cylindrical wavqzuide, 204206
resonant, 291-299 of ~-plane sectoral guide, 352353
transverse, 291, 292 ~f H-plane sectoral guide, 35>356
Smith, P. H,, 29 of parallel-plate waveguide, 237
Smith, W. O., 497 of rectangular waveguide, 22&221J
 I.YI)E.Y

711,0- to T.lit,-lnCNlr r[lllvrrtrr, 308 JlJltagr rrffrrt iou r{wficicnt, 2,5

?1111-tn[~(!f, of co:lxial li,lm, 217-21!) iron IIippel, A., 399
c~lrrcnt of, 216
of cylinllril:ll w:t~cglli{lr, 203, 204 }$
of par:illr-l~l~ltr w:~vcgtll{lm, 235236
voltage of, 216 Walkinsha\v, IV., 503
Tenth-po\vrr N i,ltll, !14, 104 \f-all, stirld\vich, 532
Terman, F, F;,, 34!), 41.5 single, 531
Thc%enins theorem, 20 \Yatson, G. >-., 76, 194, 195, 236, 341,
Tisza, I,., 3!)9 353, 354, 367
T.V-mode, of rir((ll:tr ~v:lvrRtlidr, 233 lVatson, J\. 11., 2!)5, 299
of coaxial line, 220 lfa~r rqllations, 71-73
of cylindrical wal,rguidt-, 206 l~avr v(,locity, 110
of parallrl-plate wzvcg~[idc, 236 \~:Lvcgui(le, circuhLr, TJ;-mrNlr of, 233
of rectangular w:~vegllidc,, 226- 22!) T$[-mnde of, 233
Transfer admittance cocffirimrts, 1S cyli!l(lrical, field cqll:itions for, 201203
Transfer corllcients of nct\vork, rrri- Th.lf-rnn(l(> of, 203, 204
pmcity relation between, 18 jllnrtiou rffwt in, 214, 215
Transfer impedance rocfflcirnts) 1S radiatiorl from, apcrturr treatment of,
Transformation charts, 2!)36 334
Transfnrmcrs, Cascadoj in coaxi~d li]lrs, rcctanglll at, 7FLm ode of, 22622{)
221-223 T.tf-mn(ir of, 226-22!)
Transmission cocfficicnts, of r:i(lomc
terminal adn]ittanc(, of, 366
walls, 52!)537 }Vavrglli{lc modes, mthognr,:dity of, 2n7-
209
Transmission line, homo~cncotls, 23-26,
544-546 tlvo-~virc line rrI]r(,sc,llt:iti(~Il of, 209-
thrm-~vire, 247 216
two-wire, 2137 lf-aveguide radiation patterns, circular,
336-341
Transmission-line equations, for lt-plune
rectangular, 341347
sectoml horns, 36&36!)
\Vm~,qqlidr radiators, as array elements,
Trammittcr pulling, 517, 537
301-:303, 32!)
Transmitter-receiver system, cquivahmt
JVaves, cylindrical, 75
nrt!vork of, 5360
plan(,, 73
1ransmltting pattern, 3, 557
principal, 203
Transpondcrj 327
Wmson, 1,. G., 399
Tridipnlc transverse clcmc,,t, 304
Wcst, \V. ,J., 478
Trurll, R, 221
}TmtI,hal, TV. B,, 399
Tyson, O. .\., 612
Whelpton, ,J,, 285
White, J. S., .523
v Wind rrsistznce, 450, 453
Winrlo,v, capacitive, 230
Vzllee Potlssin, (;, de la, 280 inductive, 230
Values, characteristic, 204 resonant, 231
Van .ktta, L. ~., 43-1 Wing, .~, H., 21
Variable beam shape, 508-509 Winkler, F;. D., 523
Iector, Greens theorem, 80 Wintncr, A., 122
Vertex plate, 443, 463 Wolfe, H., 461, 500
Vogel, B. It., 509 Wolff, I., 282
Voltage attenuation constant, 23 Wollaston wire, 559, 603
Voltage parameters, 210 }Voodyard, J. R,, 278