Antenna Theory and Design
Antenna Theory and Design
Antenna Theory and Design
VOLOG1
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
\ ,.
MICROWAVE ANTENNA
THEORY AND DESIGN
Ediied by
SAMUEL SILVER
ASSOCIATE PROFESSOR OF ELECTRICAL ENGINEERING
UNNEB.SITY OF CALIFORNIA, i3EP.KELEY
FIRST EDITION
*ienCe
m
,,
)/
/y,,
,, . \
EDITORIAL STAFF
SAMUEL SILVER
HUBERT M. JAMES
L. J. I;YGES J. R. RISSER
T. J. KEARY S. SILVER
H. KRUTTER O. A. TYSON
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
:lprd, 1947.
.
Contents
PRE1744CE
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
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
ARRAYS . . . . . . . . . . . . . . . . . . . . .. 312
I)ll:l,l:C,rRICI
JEXSF;S . . . . . . . . . . . . . . . . . . . . . . . 389
hfETAL-I,L~TELE
XSES . . . . . . . . . . . . . . . . . . . . . ...402
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
]]Y s. SII,V1;ll
(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
(2)
47r
SEC. 1.2] ANTENNA PATTERNS 3
r = G(f3,@) (3)
(4)
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$1.1<.11.V [SW. 12
of incidence of the lvave, This function, like the gain function, is repre-
sented conveniently as the surface
? = .4, (0,0). (6 J
The receiving pattern of an antenna is drfincd, :malogolls]y t(
the transmitting pattern, as the above surface normalized to unit maxi-
mum radius:
(7)
(11)
P$i A
R .,., = ~; (G,G,) ~ (12:
P
(-) ,m
SEC. 12] ANTENNA PATTERNS 5
(13)
(14)
(15)
(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
(b)
(c) (d)
FIG.I.I.Typicalgain-functionsurfacesformicrowaveantennas. (a) Toroidal(omni-
directional)pattern;(b) pencil-beampattern;(c) flat-topflaredbeam; (d) asymmetrically
flaredbeam.
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.
(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.)
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.
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
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. }
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
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)
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
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;.
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)
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
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
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 =
(15b)
with the square root taken to be such that both a and p are positive
quantities, ~vefind the solution of Eq. (16) to be
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)
(21)
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)
(1+%)(7+(1
-%)2o =
(24)
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)
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)
(28)
(z
~)
=1 . ~ ~(z) + Y,, tanh (71)
LY, + Y@) tanh (Tl) 1 (33)
l, = ;;,. (34)
(37b)
SEC.2.7! 7HE L(),?,5I,I<SX LINE 27
~(z) +~tanf?l
r(z1) ==+jrtanpl, (38)
q(z) +j tan @
7(2 1) = ~tan~i (39)
(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)
~11(z)] = ~VV* = ; [~.4,~ + \.4,1 + 21Au42) cOS (262 0,+ 0,)]. (43)
along the line. Ihe maxima and minima occur at those points for \vhich
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)
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
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)
(49)
[
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.
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
\ I ,---T I /\ I {
J2,6
Realcomponent
IJIG.2.9.Thebipolari,],pcdar,ccchart,
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
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
(581))
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)
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
.
where
2
~ (6%)
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
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
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
(66)
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
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
(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
(74)
p _ IPC12.
(75)
8RA
(76)
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)
(81)
(~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)
and
2.=20
()
~r!
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)
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
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
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,
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.
,
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.:
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 = ~,
.1 = I b=b@,,a,,. I
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
It then follows from I?qs. (98a) that the transfer impedance coefficients
obey the reciprocity relation
Zlz = Z21.
CHAPTER 3
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)
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
(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)
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)
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
(9)
VSOK+~t=O, (lo)
(11)
F = Eq, (12)
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.
,
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)
$c E.ds= -~
al / ,y
B .11 ds, (17fl)
$ c ds=NJ+ands (17[,)
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
vxH=J+~, (18b)
v. B=p~, (18C)
v. D=P, (18d)
with two equations of continuity
VOJ+~=O, (18e)
V. Jm+~=O. (18f)
~=,D (19a)
E
~=, B (19b)
H
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,
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:
aq,n
VS . K,,, + ~1 = O, (31)
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
(EXH+EJ=-W+W (34)
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
=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)
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)
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
(49)
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))
~= 1
(53)
Vz
n=:=~k=. (55a)
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:
(58)
.
(59b)
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
(70)
(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
.
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)
(80\
as was to be expected.
78 l<A1)lA I10A FROM CURRENT I)ISTRIBU1710NS [SEC, 37
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,
/ \
/ \
/
z
FIG.3.3,Sphericalrmrdinates.
()
72=;2: ~z; +
la
sin 0 & + ~~ (86)
rz sin 0 --(
do ) r- smz e W
(87)
(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
v. E=~, (332d)
E
V. J+@P=O, (332e)
FIG. 3.4.Notation for Greens theorem. V. Jm+jup. =O, (332f)
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
.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
and
82 RAI)IA7ION PROM CURXi?N17 DISTRIBUTIO.VA [SEC. 38
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:
To the surface integrals cm the right-hand side of Eq. (W) \veapply the
following transformations:
a. ju~+ + m X V+ : v+ dv
Nv )
(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
.
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
e-ik,
1 x
[(nx
j4irTof?-ik0
E)xn+(n.
(wn
E)n]dQ
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
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
.
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
K=(nx H),
Km = (n xE),
(111)
~ = ~(n . E),
q~ = p(n H), 1
&
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))
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
(115)
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)
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
and
(3 (+:)=1 (125)
-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
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
+ O & (129)
()
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$) (13k)
47rR
and
S=~Re(Ex H*)=~
()
lCYJ
~ +Re[Ex(R, xE*)]
or
(136)
where
l(e,@) = IF1(8,4)12 + ]F2(e,@)]. (137)
The power flow is radially outward from the origin, with an intensity of
flow that falls off with the square of R and depends also upon 8 and ~;
with respect to power flow the current distribution is, in effect, a directive
point source at the origin.
In discussing the power flow it is convenient to use, instead of the
Poynting vector, the power P(L9,c$) radiated per unit solid angle in the
direction O, 0. This is given by
(138)
G(e,l#l) =
1
,2. . (04)
F(L9,1#J)
sin OdO do
Goo//
%k(e,f$)
. (139)
2= x
H
00
Y(f?,@) sin @de d+
The maximum value of the gain function is termed the absolute gain.
In design specifications this is generally quoted in decibels above the gain
of an isotropic radiator (which is unity):
Here the As and ~s are real, and YI and -yl are in general not equal.
The vector E, is thus the resu!tant of a pair of time-periodic vectors
E,i8 and E@i@at right angles to each other, with relative amplitude and
W4 1(8,0)
EO = sin (tit kR -yl) = cw sin (d kR TJ, (142a)
4.R
~ = @w42(d, @)
* sin (cd kR ~.J = a+ sin (cd kR T1 6), (142b)
4.R
(143)
handed according as the terminus of the vector E, traces out the ellipse
in the clockwise or counterclockwise sense, respectively.
If the phase difference 6 is an odd multiple of 7r/2 and the amplitudes
are equal, the ellipse becomes a circle; right-handed and left-handed
circular polarization are defined in the same manner as for elliptical
polarization. If the phase 6 is an integral multiple of m, the ellipse degen-
erates into a straight line traced out by a linearly polarized resultant.
As o and @ are varied, both 6(0, O) and EJE$ 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)
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)
@ = 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)
(153)
+ H9Y:-iwmoe(-k) (156a)
96 RADIATION FROM C1RRE.VT D1,YTIUII 1I1O.VS [sEC. 315
(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)
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
E+ = O. (160)
or
(161b)
(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
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
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
Figure 3.12 shows the meridional polar diagram in comparison with that
of the dipole. The gain of the half-wave dipole is
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
/
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$lelkeROdodo
(168a)
SEC. 318] THE DOUBLE-DIPOLE SYSTEM 101
and
. ~+ &+ri~2i(@,@) .
47rr,
(170a)
where
(170b)
and
(171a)
where
%(e,f$) =
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 $, and & in
terms of these and the phase differences arising from the relative positions
of the sources is a procedure useful in many other fields-for example,
the theory of X-ray diffraction.
3.18. The Double-dipole System.The radiation patterns of com-
pound systems are usually more directive than the patterns of the
component systems; destructive interference between the fields of the com-
ponent systems takes place in certain directions, constructive interference
!n others, with the consequence that the total power density changes
more rapidly with angle and reaches more extreme values than does the
power density for any component system.
An important compound system with wide application to microwave
102 RADIATION FROM CURRENT DISTRIBUTIONS [SEC. 3.18
xJ
~lG. 3.14.The double-dipole system.
E+-component in the far field, the total field can have no such component.
The space factors of the dipoles are alike, except for the current phase
term e-i$. 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
/, --
,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,
3(0,0)
Cos
= ()
T
5
sin @
Cos e
Cos : ( lsin
[
Osin@)
1
~ (175)
H. = 76, @~ = 180,
(176)
E).(;V) = 130, @.(fi) = 252. 1
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
We need consider only the space factors $, and % defined in Eqs. (170b)
and (171 b). The space factors Fli and F2i are independent of i except for
5,(8,+) = II(19,c$)A(0,+),
(179)
34 L9,1$)= F,(O,+)A(O,+), 1
where
Here
A(e)+) =
m N,
na=On*=Ona=O
N,
A ~,n,naexp j
3
[Z i=l
n;(kRO . ai ~i) .
1
(180)
The total space factor is thus a product of the space factor for a radiating
unit by a lattice factor. The lattice factor, it will be noted, is itself the
space factor of a ~attice array of isotropic radiators with relative ampli-
tudes An,m~, and relatlve phases t~,~,~,.
If the radiating units all have the same amplitude, say equal to unity,
the sums in Eq. (180) can be evaluated. The term on the right becomes
a product of three factors:
106 RADIATION FROM CURRENT DISTRIBUTIONS [SEC. 319
where
N.
A, = @! (kalult,
) (183)
z
71,=0
This geometric series is easily summed. One finds
~i(hr;+l/ 2)(k.,u,$, )
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)
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
and
respectively. z Relations of the same form must, of course, hold for the
(3)
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)
/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
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)
IvLI = ~ = n, (8)
(9)
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
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
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)
s = ;
()2
; pq. (17)
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.
(20)
~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)
(27a)
(27b)
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
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)
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
1 aF
Vu = jkuel + ~ Pl+~#l+
( ~+%)+($) 34)
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,
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)
(40)
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)
P(O,0,P)
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)
=-(&+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)
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)
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)
I
SEC. 4.8] F17RMA T,~ PRINCIPLE 123
(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
by our postulate this must vanish to terms of the first order in al. From
Fig. 4.6 it is clear that
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. ]
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)
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,
FIG. 47.-On the law of the optiral path.
and
c c B
~11) n& . d = nzsa . dl + m& . d + nzsz . df
+ / B / / c
[rj~
B
Vxs=o;
(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
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
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)
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 /
E,= &e
/ ~ [(K . v)v + lcK] ~dS, (7)
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
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
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
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)
(3) their components normal to the zy-plane are equal in magnitude but
opposite in direction,
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)
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
or, alternatively,
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
J-
.l+~
?4
a
<M
(a)
a
-M
M
/ z
(b)
M
-.
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)
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)
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)
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)
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
(41)
and
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 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)
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
cos i
R,R, = ~ Cos ~
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:
(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.
i=[2(9%GsOreos% (52)
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
where
e, = [n o eO(sO)]n [n x eO(sO)]x n. (56a)
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(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
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
flowfrom the area is equal to the rate of decrease of the charge enclosed,
we obtain
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
and
() vx H,:
.
=V~xHi+~VxHij
a a
(65)
or
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)
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)
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)
Et = Ei@ + E.+ = R z=
-k [yYTi~
(@@ (7,,
F.J@@) = [G~(@3)]~+i~ o e; ~ i* I 1
{ }
or
G(@)@) = IF,I + IF,12. (77b)
(78a)
and
(78b)
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 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:
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
of the line. The total power transported across thereference cross see-
tion is
P = alv,l, (90)
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):
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
or
Er= *v[~(;)Gf(@lcOsie
3)
The magnitude of the Poynting vector of the spherical walrelet is
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
x
P
r
-2
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
(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
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
~k nxE,
;~
,kRR1x
E,(P) = ~Re-
/[ .4
!
()
6
R, x(n xH,)
1
e~kQRldS. (103)
E(P) =
~k
~
~,jk.? RI x
H ~
nxE, a
()
~ [R,. (s xE,)n
(s x E,)(n . R,)]
I e~@RdS. (lo5)
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:
(109)
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
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
$%
/f t
e
z
t
@
v
FIG, 5.11.ILcduction to a scalar diffraction problem.
(115)
166 SCATTERING AND DIFFRACTION [SEC.514
(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
Ui = UI + U2. (125)
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
By S. SII.VER
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.
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
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. /
~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
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
Up = +Re-]kg(k=,kv) (1 la)
with
a(%Y) = *2
111 lu(~)ek(-z)ek(rv)d~d~d~d~. 13)
.. -m.
or
mm
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
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)
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)
Go . 47rA
~. (20)
(21)
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
~ _ 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
(28)
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
g(e)($) =
The patterns in the principal planes (the zz- and yz-planes) are of par-
titular interest. For the zz-plane @ = O, Eq. (30) simplifies to
g(e,o)
<,n
= A -- ( ~a .
sm 8
7ra
A ). (31)
sin e
x
For the yz-plane + = 7r/2; the pattern in thk plane is likewise given
by Eq. (31) with a replaced by b. Both patterns are of the same form,
sin u/uj but are scaled in the angle e according to the aperture dimensions
in the respective planes. The secondary power pattern, normalized to
a peak value of unity, is plotted in Fig. 6.3 on a logarithmic scale as a
Loo
0.8
0.6
0.4
0.2
0.1
0.08
0.06
(*f
Iz
004
0.02
0.01
0.008
0.006
0.004
0.002
0.001
o 2 4 6 8 10
u
FIG. 6,3.Secondary pattern of a ulliforlnly illuminated rcx%:uwularaperture.
zz-plane: 2 sinl
()~
A
2A
= ,
a
yz-plane: 2 sinl ~ = ~.
()
The half-power point on the main lobe is very closely at u = 1.39; hence
the half widths in the principal planes are
zz-plane: @ = 2 sin
() 1.39A
~a
1.39A
= 0.88$ (32a)
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
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)
(39b)
showing explicitly that the gain is proportional to the area of the aperture.
184 APERTURE ILLUMINATION AND ANTENNA PATTERNS [SEC. 6.6
(41)
or
(42)
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:
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-
(48)
The full angular width of the main lobe at a given decibel level is, there-
fore,
: (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
~. = 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
(57)
(58)
Hence
(59)
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
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.
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
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,$t. (71)
/! o 0
SEC.68] THE CIRCULAR APERTURE 193
100
0.8
\ x
0.6
, -&
0.4 R
P o
z
0.2 4
Y
0.1
0,08
0.06
0.04
I \
[Al(w)]
0.02
0.01
0.008 I \
0.006 \
0004
(
0.002 / \
0.001
2 4 6 8 10
u
FIG. &7.-& condary patternfrom & unirormly illuminated cirrularaperture.
factor k/D; the larger the diameter the smaller is the angular spl-ead
of the pattern about the (0 = O)-axis.
Unijorm Phase and Amplitude. Setting f(r,+) = 1 in Eq. (71) and
carrying out the integration over ~, we obtain
1
g(u) = 2ra2 r~~(ur) dr, (72)
/ o
194 APERTURE ILLUMINATION AND ANTEN.VA PATTERNS [SEC. 643
where .JO(UT) is the Bessel function of order zero. 1 The integration over
r leads to
The power pattern p(u) normalized to unity is shown in Fig. 6.7, plotted
on a logarithmic scale, as a function of u. The half width of the main
lobe is
@ = 2 sin-
()
0.51 ~ = 1.02 ), (74)
and the first side lobe is 17.5 db down from the peak. These are to be
compared with the half width of 0.88A/a and the 14-db side lobe of the
secondary pat tern of a rectangular aperture.
Tapered Illumination .The effect of tapering the illumination down
toward the edge is the same as with a rectangular aperture: reduction
in gain, increase in beamwidth, and reduction in side lobes. The effects
can be illustrated by considering the series of aperture field distributions
(1- T)~, p=l,2, . ... The secondary patterns are given by
1
gp(u) = 27ra2 (1 r)~o(ur)r dr (75)
/ o
or
Z1p!Jp+,(u) _ ~a if +,(u)
g,(u) = rrag ~P+l (75a)
p+l p
TABLE62.-SEC0NDAR%PATTERN CHARACTERISTICS
PRODUCED
BY A DISTRIBUTION
(1 P)POVER A CIRCULAR
APERTURE
~, ~, half-power 0, position
First side lobe,
I
P gain factor db below peak
width of firstzero
intensity
I sin_l 1.22X
o 1.00 17.6
D
~in_, 1.63A
1 0.75 24.6
D
Sin_l 2. 03A
2 0,56 30.6
D
Sin_, 2. 42x
3 0,44 . .
D
sin_, 2. 79X
4 0,36 ....
D
(78)
d)=o,
_ 3rra
~ Az(u). (79)
(u)= 2maw
-W[- +J(uldl
(80)
1.0 I
8
I
4 \
I
t, -4=$ 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
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
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)
~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
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
BY S. SILVER
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
jueE, = ~ ~- (3C)
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
? = (a + m) = jk = j+f)~. (8)
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)
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
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
B*7I = ~2:
9..
=(k K:.))i. (15)
(16)
(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
E,=~(Hxiz). (19)
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
over the walls. Substituting from Eqs. (4a) and (4b), we find that
condition (21a) is equivalent to
(n x VI$) = O. (21C)
(22)
H, = ~ (E x i,), (23)
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:
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
(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
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,
>4
ZEM-mode: H = ~~ (i= x E), z(o) = f ,
(33a)
()e
(33C)
(34b)
210 .![IcRow4 VE TI{,I.VST11*S,51
O! LIVES [SEC. 74
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
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
(40b)
and in region 2,
E,, = [V~+)e-@,]g(x,y), (41a)
*
1
[w+) - TT] = &
2
Vj+. (42b)
(43)
(46)
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
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)
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
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)
w, = +Cv
or
~ = 6// Ivclzds
~, (51a)
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,
(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)
(53a)
O=ww)
For most dielectrics of interest p differs negligibly from the free-space
(54)
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)
F = R(r) ;~~;;
{
then R(r) satisfies the equation
(57)
(zy)
@(xjy) 1
= [.1,1.,(.,) + ~.~~(.r)](c cos mfi + D sin w@.
(58)
~~_(KU) + ~~~(KU) = 0,
(59)
fi~~(.b) + BIV*(Kb) = O, }
~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
2
(64)
l=a+b;
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,
(65a)
(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.
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
(67)
(76a)
F = x(z) Y(y).
(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&).
$(z~)
l$(x,y) 1 = [A .o. (.=X) + B sin (K4)][C cos (by) + D sin (wY)]. (69)
%1
a+
a+
.=O
= Kz~](?J) = O, (70a)
mr
~z = , m =0,1,2,.. (71)
a
nr
Ku = , n = 0,1, 2,.... (72)
b
(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)
(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.
~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.
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
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-
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.
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
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,
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,
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)
(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.
@@
(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.
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
+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
(92)
(93)
and the guide wavelength for the freely propagated mode [Eq. (17)] is
[1-/$1
(94)
g=
The transverse components of the field are obtained from E, by means
of the set of Eqs. (4) and (5):
(95a)
(95b)
The cutoff wavelength for the TE.-mode is given by Eq. (93), and the
guide wavelength by Eq. (94). The complete set of field components is
H.= Acos
() ~, (97a)
BY S. SILVER
The early trends in microwave antenna design grew out of the prac-
tice of using dipole systems at longer \\-avelengths. Nevertheless, little
systematic information has been obtained about microwave dipole sys-
tems. This is partly due to the greater difficulty in applying theory to
practically useful microwave dipoles and partly to the urgent military
needs which prevented systematic research during the early development
in this field. More recently, attention has been concentrated on wave-
guide and horn radiators, which are more amenable to quantitative
analysis. Consequently, the design of microwave dipole antennas is
s$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:
T1,
1
4
12
Die-
lectric
bead
support
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
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
--- 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.
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
W,>. S6.-Superl,c,siti(>,, O{ the TE.11- a,)d TE, I-mncies in the slotted region of the slot-fed
dlpol c.,
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
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
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
Effective center
Transparent sphere of feed
r
\
\seam
.200 0 20 00 3400
400
320
6(Y
300
800
280
260
1000
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.
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
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
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
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
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
(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)/
(7)
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
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
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.
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
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
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)
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
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
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
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
(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)
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
+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
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 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
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.
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,
=~2r~-lf(z~sin,d@d@ =ecme
or
G=;, (19)
where
(20)
131 a(~a,)-za,za:
,=0 ,=0 ,=0
=4 ,1I (/%=0,1, . . ..? 2-1).
dak
Ya,
(u,=0 )
(a, G) a, = O.
z
,=0
*=
sinz
n
n12( ~ Cos 6
1 TCOS6
)
n2 sinz
(n1~ )
When n = 2, we have
V@=cos
();Cose.
2 ()
T
Cos -- Cos e s
ln ( ~ )
n
n12
Cos
1 : Cos-o
e
n sin
( n12 )
sin
( E
n12
~ Cose
)( < 1
n2~
-
n12
cos e
. )
n+icose - :+%
SEC. 95] BROADSIDE BEAMS 273
N = ~ (a + b + c) + 2(ab + be).
Thus
a=c,
T-l
b= a = 0.7519a.
T2
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.
r ===v(f3)
sin2 ~ (COS0 1)
v(o) =
nz sin2 ~ (cos 0 1)
=:/:. [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
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.
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
G = 1.82n.
~ _ 125
fin
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.[~]
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
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
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
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
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)].
=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
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
P:--AA
13 ~
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
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
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
If 1 is any closed rcgi{m bolludml I),ya sllrfare ,S, it fullmvs by the diver-
gence theorcm that,
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
(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
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
where
(49b) t
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
(511))
lO=BIO=-Eo:(:Ysint:)(
osf+)J52)
1 TIlr rollst:lllts of (K,), and ,S, flcorrespond to the nlmic being so norlllalimd thtt
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.
(53)
(54)
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
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)
(57)
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
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
&
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
0.004
; 0.002
g
n
8
2 o
u-l
-0.CQ2
I WI
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
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
(a) (b)
~,~. $~~.-l~avegu,de radiators: (a) longitudinally polarized; (b) transversely polarized.
=wl+ln*) (58)
The dimensions b and b are defined in Fig. 9.25. IYith the above length
the element has been found to be very closely a pure resistance.
The coupling of the element to the line, i.e., its resistance or conduc-
tance, is a function only of the relati~-e dimensions of the branch guide and
the mam gmde. It is the particular advantage of the waveguide element
T
()
lb2
=27
f
1,
304 LINEAR-ARRAY AX TENNAS AND FEEDS [sm. !315 ,
tric vectors lie in meridional planes having the axis of the array as a
common line of intersection; the transverse element is analogous to a
magnetic dipole, and the longitudinal element to an electric dipole.
The transverse element in its ideal form should consist of a circular ring
of uniform current, while the idealized
z longitudinal radiator should consist
~f$%zzfi
I~IG. 9.27,Dipole fed I)y three-wire line,
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.
220 140
230 1300
240 1200
250 1100
260 100Q
270 900
280 800
290 700
300 600
3103 500
320 40
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.
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
&
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
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.
500 310
703 290
90 270
1100 ~~o
1300 23W
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
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).
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+~.+~-~$,
) Sinh (IJ); (62)
Zj sinh (-yl) =
[
Z, + 2, tanh
20
()1
71
~
2
sinh (71)
+2
[
Z,+ Zotanh
()1
~ . (63)
(+%)
Cosh() cOs1+(2+2+2%)
in
64)
It is seen at once that the loaded line has a complex propagation con-
stant ~ = a + j~ in which both the attenuation and phase constants
are functions of the loading and the spacing of the elements. Equation
(64) shows, however, that if the spacing is equal to half the wavelength in
the unloaded line, the relation reduces to
cosh (~1) =
() 1 + -~a ;
J
l=;. (65)
+=j$=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
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)
()
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
.{
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
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
.
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.
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.
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
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.
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 --
. , -.
--
=15 .
,7
:
%2-- .
10
3180 3187 3197 3207 3214
Waelm@ cm,
:,
,,
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.
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
h
b
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
;<
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
!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
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.
.
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
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
(72)
4
CHAPTER 10
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)
(de
for T~lI-modes.
P,.n
Therefore H, can be \vritt,en
(5)
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
(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:
; Jm(z) + J.,-,(z),
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-
~oo
0.9
0.8
0.7
,Ok-mo 0.8
0.7
0.5 0.5
0.3 0.3
700 700
0.2 0.2
800 800
0.1
90 S@
1w l~o
(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
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
inserting the values of E, and EV from Eq. (7), we have for the TE1l-mode,
(17)
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= )
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
~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)~] /
+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
(
7ra
Cos Sln
A
6
)
Cos @
(- ) ()
T(Z
slnocoso ~-
A
~-[+=- bwn +)
1
kII?
()
f sin /3 (a cos @ + b sin @)
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
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)
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
1!
(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
,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
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)
%+::+%+(-:)
=0
.4s was pointed out previousl.v the elpctric field lit varies sinusoidally
(30)
(32)
sEC. 107] MODES I.% E-PLA.VE SEC7ORAL HOR,VS 353
where
(33)
E8 = Cos
(); 2,(/3?-), (34a)
(34b)
(34C)
(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
=+;% (42)
a(kr)z +[-MF=O,
where
kz=uzpe= ~Zlr
()
((i, x R,)Ed +
o>~
~ [i, R,(i,.R,)]II=
1
dO dx. (44)
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
[ 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
(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 ,
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
00 100
o
t
I , , 1 ,. I ,
I ] 1/ /
/ A / I I I I I I I
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)
(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)
(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
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)
(55)
(56)
H$)(&) = Fe@,
(58)
Hf) (@r) = Gel@, )
(60)
368 WA VEGUIDE A .VII HOR.V FEEDS [SEC. 1011
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.
\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-
~b I
@r, = (05)
A,, sin O.
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)
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-,)
sin 00Fl(jrl)
Cos (*1 @l)
~ = ~ tan_l en ,G, (&,)
(70)
sm (?OF,
1+ ~m sm (+1 @l)
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
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,
\ 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
(74)
-1_
SEC. 1013] AD JIITT.4. TCE CH.iR.l( TEHI,~7IC,Y OF H-PJ,A,VE 375
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
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
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 <;.
a /:;2cos($)co5t2)
o (77)
o 0.5 1.0
A~A Figure 10.24 is a plot of aS/al vs.
~x~. 10.24.ltatio of the anl~l,tucirs A/.4. The ratio aS/al decreases
v
of the TEm- and Tk?,o-modes ,U a box wit h increasing A /A, reaching zero
hornVS.A/A.
when A /A is unity. If it were pos-
sible to make A /.4 very small, values of CL3 lal approximately equal to 1
could be obtained. The lower limit of A /.4 is 0.20, corresponding to
I
10
0.9
0.8
0.7
~ 0.6
E 0.5
Q
0.4
0.3
0.2
0.1
0
7Y2 r/3 76 0 ?76 ~/3 77/2
7rXL
FIG.1025.-Aperture illul,)lllatlon (a]nplltude) of a Imx honl for a series of talum of tl,e ratio
113n,.
SEC,1015] 7Htt B().Y HOtt.V 379
A = 0.5A the cutoff in the input guide, so that the maximum aJal is 0.93.
Actually much smaller values of a,~a, arc desirable. Figure 10.25 is a plot
of amplitude illumination across the aperture for as/al values from Oto 0.5,
calculated on the assumption that the horn length is correctly chosen to
lo
horn
0 0.1 0.2
FIG. 1027.-Magnitude
0.3
a3 a,
/
0.4
of the first side
vs. asjal. lobe m the power pattern of a box horn vs.
as/a,.
make the two propagating modes 180 out of phase at z = L With respect
to their relative phase at z = O. The curve for a, al equal to 0.3 is a
fairly good approximation to uniform illumination.
, The H-plane radiation patterns in the Fraunhofer region have been
calculated from scalar diffraction
theory (Chap. 6) for the amplitude So
distributions of Fig. 10.25. The
results are summarized in Figs. ~
10.26 to 10.28, n-here Fig. 10.26 ~ VA
sho\vs relative gain, Fig. 10.27 the ~ ~0 \ 1;
magnitude of the first side lobe in g
~ .1:
the power pattern in decibels ~
dotrn from peak, and Fig. 10.28 > ~ .2
the full angular ~vidth of the pat- ~o. \ > ~ \ Z$
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
where
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)]
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).
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
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.
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.
[ =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
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)
712 1
2n ~ + (n
sin
()27rnt
l) k.Slnz 2irnt
(85)
[ 4?12 ( A )1
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
----- 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 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
Solving for r, the equation for 7T, the generating curve of the lens
sllrface, is given by
r=
(n [)~. (3) ,
?1 COS @ ]
~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]
(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,
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
1
em = COS1 .
n () (13)
(14)
A (Y) = (n Cos 0)
(16)
@ej d (n l)j(n cos 8 1)
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
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
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
. .
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
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)
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
~=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)
(34)
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
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.
(37)
%=~
L%C. ~] 8] PARALLEL-PLATE L.V.VShS
%= 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.)
#d
yIG, 11 .17, Wire lenses: (a) parallel-wire lens; (b) wire mesh lens. ((lwmtes~ oj the Bell
Telephone Laboratory.)
4 -
SEC. 11.10] METAL-PLATE LENS TOLERANCES 407
A
(39)
t S 16(1 n)
\ and
(40)
An = (1 n) Aa
. (41)
n a
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
(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
25no X.
Bandwidth = per cent. (47)
1 + n, (1 ?,o)t
-=4 .-.- -- .- ..
SEC. 11.11] BAN I) WIIMH OF METAL-PLATE LENSES 409
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)
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
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)
with ,.
1<; <2.
The restriction on i arises from the fact that grating lobes become possible
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.
I
r I
i.:
!:
,,
,.
.,
:
/
CHAPTER 12
BY S. SILVER
PENCIL-BEAM ANTENNAS
2f
~=l+cos~
=jsecz
()
~ (3)
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
(4b)
F(.E,I)) = ~ G,(&x).
(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
= [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
(12)
where the vector 1, expressed in terms of the incident field on the reflector
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,
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
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
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
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
=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)
+
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
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.
.- ..
SEC. 12.5] THE ANTENNA GAIN 427
(26)
(28)
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
$=(oCot:)[ =%n:si%cw
The term in brackets is the modification of the previous result introduced
3)
It will be noted that the gain is 7, as compared with G~4J= 10 for the
idealized cos~ ~ pattern used previously, The back-lobe gain Gr is
0.142. Curve A is the relation between the gain factor and aperture,
0.7
J
0.6
0.5
~o.4
z
~
~ 0.3
0.2
u.1
n
-00 100 20 300 400 500 600 700 800 900
Angular aperture ~
FIG. 12.5.Effect of back lobe on gain.
neglecting the back lobe, while Curve B includes the interference effect.
The gain falls above or below Curve A according to whether the back
lobe is in phase or out of phase with the reflector field. The two fields
add when the focal dist ante is such that, together with the 180 phase
change at the reflector, the field of the latter is brought into phase with
the back lobe. The points of maximum deviation from Curve A corre-
spond to differences in focal length very nearly equal to x/2. The
optimum aperture is not altered noticeably, but the maximum realizable
gain factor increases by 2,5 per cent. The effect is small for this par-
ticular feed because the back-lobe level is so low relative to the main lobe.
j With ieeds such as the 3-cm-band double-dipole feed discussed in Sec.
(89), having a comparatively high back-lobe level, the back-lobe inter-
~ ference effect is much more significant.
430 PENCIL-BEAM AND SIMPLE FANNED-BEAM ANTENNAS [%C. 12.5
(32)
where d(~) represents the phase error in the feed pattern. A review
of the steps leading to the field intensity E(Ro, O,O) of the secondary
pattern on the axis, given in Eq. (18), will show that the only change
introduced in Eq. (18) is the replacement of [Gj(+)]~~ by [GJ(~)]~~e~~6t$J.
By precisely the same development as before, the gain factor is given by
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.
function is
6 6,
Cos +. (34)
X=x
(===
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.
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)
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.
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
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
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.
(5.97)
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)
(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).
..
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
2m- dr
dS = . (40)
*
Cos
() 2
(41)
(42)
=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
~(~) = + 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.
= 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
GO being the gain of the feed, Assuming this form we can evaluate
!,
R(o), l(o) analytically:
=;(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)
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
~. 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
I?IG. 12.15,Quarter-wave grating to rotate the polarization of the electric vector and
1 elimm
ate the n!islllatch.
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.:!
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
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.
\ / 1,
\ / 4
\ / \ /
\ / \ /
\ /
=. / +____
- _-
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.
([
.
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
!,
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).
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.
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 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
(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
(63)
g=
[ -(W
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
. 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
3103 500
3000 600
290 700
280 80
270 900
260 1000
250 1100
2400 120
230 130
,-+- .
i
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
u G(v) _ ,.yfu
,
r.(v) = : - e dv, (68)
To/
1 +g
(69)
CHAPTER 13
SHAPED-BEAM ANTENNAS
BY L. C. VAN ATTA AND T. J. KEARY
/ /
l:l<;, 132,-Beam froxu antenna of airlmrne radar for surface search.
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
1
, Nu, 13.5. Effective ground target area and its interception cross sertion for airborne
pulsed radar.
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)
The several ideal cur~es for G(d) discussed above are presented
in Fig. 13.6:
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
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@,,.
,,
SEC. 13.3] S UR Vli Y OF BEA M-SHAPING lECHNIQUES 473
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
/
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
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-
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.
(a)
SEC. 133] SURVEY OF BIIA.W-SHAP1.VG TIICH.VIQUES 481
(a)
8
(b)
~r~. 13.14.Cut paraboloid reflector with four-horn array feed: (a) antenna; (b) central
vertical section.
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
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-
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,
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.
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
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,
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,
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
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.
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
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
1
I
-z
1
I
I
I
I
1
\
a
Direction of
peak intensity
of the I!ne source
do
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 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)
%=b+wd (13)
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
(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)
1hc
ralr~llaticms arc laborious but straightforward. The reliability
502 SHAPED-BEAM ANTENNAS [SEC. 138
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
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 ~\
/\\/
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
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 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
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
outside this range. Using Eq. (25), Ire first find 8(o) so that
%+[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
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
FIG. 14.1 .PPI photograph of wooded terrain, This photograph is essentially free of
defects due to antenna pattern.
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:
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
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):
--
Center of spherical
surface
Antenna
Oirection of lobe
i\\\\\\\\\\\\\\ from secondary source
(a)
--
--
~ ~ Original beam
Antenna3:
\
\
\
L Reflection lobe
(b)
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
Streamlined radome
7
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
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
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).
assume simplified forms ~rhich will not be repeated here. The expression
for the over-all reflection can be simplified fllrther to
(8)
(9)
(10)
n~ 71A0
n=o, 1,2, ..., (11)
0=~=2vj
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 )
-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
]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)
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
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
@&&ion,
High. density skins x High.density skins>>
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
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
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.,
Similarly the reflection coefficient transforms along the line, and \vehave
from !@. (231 )
r(~) = rL@71. (9)
(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.
obtain
~VV* = ~[e 2[rLlc0S (292+ ~) + \IL\2e2=], (13)
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)
7a&a~
! Tuner Attenuator
I & /: Space
/ Slotted section
P7
(a)
I a
Padded
,
generator
Shorting
plate
\
(b)
/
FIG. 15.2.Short-rircuitwI line technique for drterminingel ertriral length
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
t 4
Matched
Tuner variable , 1
attenuator
I Short
/ Slotted circuit
section
FIG, 15.5.Block diagram for calibration of detector.
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
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.
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
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
\
Waveguide - coaxial
Probe
Amplifier
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
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.
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
i-R=[i-
1%. !
(22)
l(.os+-:
or solving for r,
[,_co,+_(~v]
T=R
[ 1
Icos+j
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)
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
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
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
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
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 Sec. 15.8.
SECONDARY PATTERN MEASUREMENTS
r)
R+
()~
2
= (R + JR).
SE(!. 1514] ,s171.V(; (() !YLVI1)EI{A Tloiji5 575
(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)
(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:
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
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
(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))
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
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
(43)
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.
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
Wcz(+a+s(id] (47)
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
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.
L Ao,Go %J,90
q d CR=3P T
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, = ;;;:: ~
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)
(56)
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)
(61)
.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)
ANTENNA MEASUREMENTSEQUIPMENT
BY O. .4. TYsON
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.
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
-n__La_
) 80 100
Reflector voltage
120
Modulator
L
y!
q
,
Hir-1 )
~ .
Feedback loop
d K
~{)~ .1 IV YIIIYIVA IlfI<Ai7[i[<I<L111{A TS-EQIJIIM l<,V 7 [SE(;. 1(;.2
-J
410 R-fiystron
I:lG. 16 .G.Schematic diagram of a circuit for adjusting cavity to resonance.
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.
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
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
$:\, 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
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
Faraday cage
5fl
?5w
(&=J==J
Input impedance 200Q
I
F x
&
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
(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.
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
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
Signal
<m; =
ampltfier
Potentiometer
amplifier Nonlinear potentiometer
Sewo
amplifier
Drum
Selsyn
motor
.
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
Angle
FIG. r6,17..lntenna pattern sllnwing det:iil ol]tain:tl!lc wittl :Lu<liclrm,riler.
Selsyn generator
Antenna +
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
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
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