UNCLASSIFIED

AD 273 842

ARMED SERVICES TECHNICAL INFORMATION AGENCY

ARLINGON HALL STATION
ARLINGION 12, VIRGINIA

UNCLASSIFIED
NOTICI: When goverment or other drawings, speci-
fications or other data are used for any purpose
other than in connection with a definitely related
government procurment operation, the U. S.
Goverment thereby incurs no responsibility, nor any
obligation whatsoever; and the fact that the Govern-
ment may have fozrilated, furnished, or in any way
supplied the said dravings, specifications, or other
data is not to be regarded by implication or other-
wise as in any manner licensing the holder or any
other person or corporation, or conveying any rights
or permission to manufacture, use or sell any
patented invention that my in any way be related
thereto.
 213 842
C MEMORANDUM
tRM-3080-PR
10 MARCH 1962

S THE ANALYTIC SIGNAL REPRESENTATION

- OF MODULATED WAVEFORMS
Edward Bedrosian

-PR
PREPARED FOR:

UNITED STATES AIR FORCE PROJECT RAND uJ

u- L ,

. - S N IAA

SANTA MONICA"* CALIFORNIA

MEMORANDUM
RM-3080-PR
MARCH 1962

THE ANALYTIC SIGNAL REPRESENTATION

OF MODULATED WAVEFORMS
Edward Bedrosian

This research is sponsored by the United States Air Force under Project RAND - Con-
tract No. AF 49(6381.700-monitored by the Directorate of Development Planning,
Deputy Chief of Staff, Research and Technology, Hq USAF. Views or conclusions con-
tained in this Memorandum should not be interpreted as representing the official opinion
or policy of the United States Air Force. Permission to quote from or reproduce portions
of this Memorandum must be obtained from The RAND Corporation.

1700 MAIN ST * SANTA MONICA * CALIFORNIA

 iii

PREFACE

This report is a result of RAID's continuing study in comimmications

and propagation. It is prompted by the need to develop new and efficient

modulation techniques which will make better use of existing equipment

and available bandwidth. This report reveals a new type of frequency

modulation (FM) that retains many of the advantages of existing FM while

reducing the bandwidth needed. A patent disclosure is contemplated. The

report should be of interest to agencies and contractors concerned with

the design and application of FM in communications systems.

 V

SIUARf

This report offers a rather general and mathematically convenient

formulation of analog-modulated signals which makes use of the analytic

signal concept. Xnown types of modulation are readily identified as

special cases. As a result of examining the various cases which the

model embraces, a new type of modulation has been discovered--single

sideband frequency modulation (SSB PH)--which can be derived from a con-

ventional phase-modulated signal bV an additional amplitude modulation,

using the exponential function of the modulating signa's Hilbert trans-

form. The resulting modulated signal will have a one-sided spectrua

about the carrier frequency, will be compatible with existing PH receivers,

and will cause a decrease in signal bandwidth.

 vii

ACKNWGMV

Dr. W. DaVle of On wN Corporatlon provided mmy stimlating

discussions and useful suggestions. His contribution is gptefulW

aeiwvle ed.
 ix

CORTEI

AWKW cT. ............. ...... .. *.*........... .. vii

LIST OF FP 0R ...................................... xi

Section
1. INTODUCTION ......................

II. THE ANALYTIC SIGNAL ................................. 2

III. THE MODUIATED SIGNAL ...................... ............. 3
The Carrier Function ..................................
The Modulation Function ..............................
IV. HARMNIC-CA&RI'R SYSTEM................
Linear (Amplitude) Modulation ......................... 7
Exponential (Phase) Modulation ........................ 9
Exponential (Envelope) Modulation ..................... 12
V9 CONCLUSIO0N ........................ 16

REFERENCES ......................... .. 18
 xi

LIST O' FIGURU

1. Spectral translations due to introduction of analytic

carrier ...................... ........... .... ...... .....

2. Mbnitudes of spectral components for conventional and SSB

forms of FM with sinusoidal modulation ..................... 13
 1

I. INTROCTIC

Recent studies of compatible single sideband (SSB) modulation system

have stimulated interest in the theoretical aspects of simultaneous ampli-

tude and phase modulation. ( ) These studies are notable for at least

two reasons which are not connected with the specific application. First,

they underscore the fact that although amplitude and phase modulation are

distinct techniques, they have an intimate connection. Second, theoretical

studies invariably introduce the Hilbert transform ( 5 . 6 ) sooner or later

(7'8 )
and become, in fact, applications of the "analytic signal."

Other studies dealing with various types of signal and noise vave-

forms (9-12) further attest to the growing utility of the analytic signal

representation as a theoretical device. This paper considers the applica-

tion of the analytic signal to Bfthdad's representation of analog modula-

tion(13) in order to obtain a more general formulation which embraces

known types of analog modulation and in which signals having one-sided

spectra are implicit.

*This list of references is by no means complete. The first three

are merely the most recent which have come to the author's attention.
IrannW (Ref. 4) reviewed the previous literature giving numrous refer-
ences and pointed out that the notion of simultaneous amplitude and phase
modulation originated with Tetel baum in 1937. Additional references are
given by Kahn (Ref. 3).
2

I. TH ANALUTIC SNRAL

Basically, the analytic signal is a complex function of a real

variable whose real and iaginary parts are a Hilbert pair. In practice,

the actual vemeform under consideration is identified with either the

real or the iaginary part of the analytic signal; the analytic signal

is then simply substituted for the actual signal for the purpose of

analysis.

Frequently the result is a more convenient and compact notation

which can be used in convolution Integrals or in conjunction with transfer

functions in the sam fashion as the real signal and which, in addition,

has several interesting and useful properties:

1) The Fourier transform of the analytic signal vanishes for

negative frequencies (also, the real end iLignary parts have
identical poer spectra).
2) The aegnitule of the analytic signal Is the envelope of the
actual wveform. Though defined purely mnth~stialW, the
envelope Wa physical significance for narrwo-band signals.

3) The phase of the analytic signal is equal to the phase of the

actual waveform.

It is clear from the foregoing that the analytic signal is simply

a formalised version of the "rotating vector" or "phsor" that is fre.

quently used in circuit analysis and studies of omunication systems.

The theoretical justification of this representation is grati ying because

it brings to light some of the otherwise unsuspected properties listed

above as vell as eliminating the need for its Intuitive introduction.

Details of these properties are given in Rsfs. 5, 7-12, 15, and 16.

The date at which the rotating vector was introduced is difficult to

determine precisely. Ma of the modern concepts associated with analysis
of PH and SSB systems are given bV Wheeler (Bef. 1).
 III. THE MODULATED SIGNAL

While the actual processes employed in generating an analog-modulated

signal may differ considerably for various types of modulation, the re-
sulting signal appears representable as the product of two time functions,
viz.,

S(t) = c(t) m(f(t)) (1)

where c is a function representing the carrier and m is the modulation

functional representing an operation on the modulating signal f. The

form of Eq. (1) has the advantage of mathematically disassociating the

operation of providing a carrier from that of characterizing the actual
modulation technique.

THE CARRIE FUNCTION

The purpose of the carrier function in Eq. (1) is to transfer the

intelligence spectrum to a frequency region which is more suitable for

propagation. The effect on the resulting spectrum is seen from the Fourier

transform of the product(17) in Eq. (1),

= (2)

If the carrier is a narrow-band waveform as might be employed for

noise modulation, then the convolution expressed in Eq. (2) results in a
spreading of the signal band as well as its translation to the vicinity

of the carrier frequency. If the carrier is a pure sinusoid, then a

simple frequency translation is effected and Eq. (2) becomes

4

S(wJ) = M(* W. + M(.U+wo (3

If the carrier is written as an analytic signal whose spectrum,

therefore, exists only for positive frequencies, then the frequency trans-

fer is only toward the positive frequencies as illustrated in Figs. la

and lb for jo > 0. Whether or not s(t) is in turn analytic depends on

the spectral extent of both c(t) and m(t), the test being whether or not
S(..:)
vanishes for w negative.

THE MODULATION FUNCTION

Usable modulation functions must be limited, of course, to those
which yield modulated signals amenable to subsequent demodulation. The
techniques currently available are: coherent (amplitude) detection in
which a replica of the carrier function is used to render the amplitude
of the modulation function, phase or frequency detection in wh'ch the
oscillatory character of a narrow-band modulated signal permits an
approximate measurement of its phase or frequency, and envelope detection
in which a suitable rectifying and filtering network can yield an approxi-
mation to the envelope of a narrow-band modulated signal. Thus, the
modulation function must cause the modulating signal to appear (though not
necessarily linearly) in either the amplitude, phase/frequency, or
envelope of the modulated signal. The linear and exponential functionals

are unique in their ability to produce Just these effects and are, there-

fore, the only ones considered here.

As might be suspected, the linear functional corresponds to amplitude

modulation. The exponential functional, however, corresponds not only to

 IM(w)I

0
IC(W)I
WO
IS 0w)

0ow
a) General (bond limited) analytic carrier

IM(w)I

0 WO

wo t
b) Harmonic analytic carrier, ei

Fig. I - Spectral translations due to introduction

of analytic carrier
6

the familiar phase modulatiorn but also to the emergent technique of

simultaneous amplitude and phase modulation which may be better charac-

terized as envelope modulation. The latter will be discussed in greater

detail in a subsequent section.

If the modulation function is written for the analytic as well as

the purely real modulating signals, then both the SSB and double sideband

(DSB) forms of modulated signal appear. The SSB form results, of course,

from the fact that an analytic function m of an analytic function f is

itself analytic and therefore has a spectrum which vanishes for negative

frequencies. Frequency translation by the carrier function then results

in a modulated signal which has only an upper sideband. A one-sided

spectrum having only a lower sideband can be generated by using the cm-

plex conjugate of the analytic function f, since its spectrum vanishes

for positive frequencies.

No generic differentiation need be made between phase and frequency

modulation since they differ only by a linear operation (differentiation
or integration) on the modulating signal. Frequency modulation can be
obtained by phase modulating with the integral of the modulating signal
and conversely using differentiation.
 7

IV. HAR4ONIC-CARRIER SYSTW

The modulated signals resulting from the foregoing modulation

functions are best illustrated by using the harmonic analytic carrier

iwt
e in Eq. (1). For each modulation function m(t) a complex modulated

signal s(t) results. When analytic, both its real and imaginary parts

are valid representations of the actual modulated signal) and its mag-

nitude gives the envelope. The modulating signal is denoted by f(t) and
A
its Hilbert transform by f(t).

LINAR (AWLrfI ) I4UWION

Double Sideband

M(t) - f(t)

s(t) = f(t) cos wot +i f(t) sinwt (w)

Is(t)I = f(t) , W ?_ Wmax

The representation of a conventional amplitude-modulated signal

is izuediately apparent in Eq. (4). It is analytic for .u - w1 1 , where

wmax is the highest angular frequency component in f(t), as noted both

by 0svald(9) and by Wozencraft. (16) Coherent detection can recover f(t)

exactly for all 0 . Theoretically, the signal must be analytic to permit

envelope detection; practically, it must also be narrow-band.

The carrier is suppressed in this representation, but the distinction

is trivial since it is easily inserted by adding a constant to m(t) in
Eqs. (4) and (5).
Single Sideband

A
m(t) - r(t) + i r(t)
S(t) - [r(t) coo wot - fAt sin wot]

+ i [r(t) sin Wot + A~t Cos W~t]

Is(t)t - [r 2 (t) +Ar2(t)]1/2,0 WO > 0

That Eq. (5) represents a SSB azplitiu1e-modulated signal is easilyr

shown 1tr noting that coo at and sin at awe a Hilbert pair, so the Fourier
A
series representations of f(t) aid f(t) wre
CD CD
f~)c.cos(wt + Ynr); fAt sin(wt+T
n-o

Thuo, Eq. ()can be written

s(t) C CcOS[ (wh + wo) t + T.n] + iX cn sin[(w. + wo) t + 9.]

where the simple translation of all frequency components bc an amount

wis apparent. The representation is analytic for all values of wo > 0
and =Uy~ coherent detection can recover f(t) exactlyr.
EXPONh1IIAL (POiSE) 1MVI2JATION

Double Sideband

,n(t) - •eIf(t)

8(t) - Cos [uw0t + f(t)] + i sin [w~t + f(t)] (6)

Is(t)l S! 1, , > imx

The conventional representation of phase modulation is apparent in

Eq. (6); it is designated as DSB to emphasize that the spectrum is two-

sided about wo . Even if the modulating signal f(t) is band-limited, the

exponential operation used in generating m(t) assures that it will have

a spectrum of infinite extent in both positive and negative directions.

Shifting the center frequency to w0 cannot, therefore, produce a spectrum

(11)
containing only positive frequencies so, as noted also br Dagundji,

the representation is not analytic. In a practical sense, of course, it

is known that the spectrum falls off rapidly beyond some frequency. Thus,

the representation can be taken approximately to be analytic if the center

frequency Is high caqopa with the highest significant frequency component

Clxin m(t).

Sinle Sideband

m(t) = •i(ft) + i A(t)]

A A
8(t) - e'f(t)co[wt + f(t)] + i e'f(t)sin[wot + f(t)l (7)
A
1S(t)1 - e'f(t) 0 %> 0
10

The modulated signal given by Eq. (7) appears to be novel. Since

the modulation function is now analytic, it contains no negative fre-

quencies (thoug the spectrun still extends infinitely in the positive

direction). Thus, s(t) is analytic for ali values of w > 0.

Py analogy with Eq. (5), the modulated signal given by Eq. (7)
represents a SSB version of phase modulation since it contains no fre-

quency components below *o"The implication is that, though the in-

stantaneous frequency

t =
(t)

has the same excursions for both Eqs. (6) and (7), the multiplicative
"
factor e 4(t) causes the lower sideband to disappear in Eq. (7). To

demonstrate that this is indeed the case and, also, to gain insight an

to the effect on the upper sideband, consider frequency modulation ty

the single sinusoid

i(t) = rcos X t

where f is the peak frequency deviation. The phase function and its
Hilbert transform become
A

f(t) = sin w t f(t) coo w t

where ,/t is the deviation ratio or modulation index.

The modulation function in Eq. (7) becomes

e + AAif(t)
ff) 0l] 1 0lk kwt
MMt)
k-o
 11

yielding the ana2!tic modulated signal

•E
m(t) e*iWot =
0

\l1 W*i(wo
(LI)k + kwu)t
kno

The actual modulated signal may be taken as the real part or

= U ( cou(w + k(8

k-o

from vhich the one-sided nature of the spectrum is clear. The mean-
square value of the omplitude factor is 10(20) so the ratio of peak-to-

average power over the modulation cycle becomes

A conventional frequency-modulated (FI) sig l using the same modulating

signal has the familiar expansion

kum
cos[wt + ft(t)] ' 'Jk~g) coa(wv0 + kw)t (9)

It would be difficult to find a more striking example of the

mathematical sinplicity afforded tV the use of the analytic signal than
that deamonstrated IV the develolpmnt leading to the Fourier series ex-
pansion given bV Eq. (8). The conventional approach is to ualtiply the
A G
Fourier series expansion • "f (t) I-cos
I nwt lCr that for

coGEw t + f(t)] given 1tr Eq. (9) and then reduce the double summation;
needless to sor, the procedure is quite tedious.
12

The line spectra of Sqs. (8) and (9) axe plotted in Fig. 2 for

several deviation ratios. The component magnitudes have been adjusted so

that the spectra represent signals of equal average power. Generally

speaking, the cancellation of the lover sideband is accompanied by an

extension of the upper sideband. Nevertheless, the "vidth" of the one-

sided spectra appear reduced W roughly one-third.

The one-sided signal can be generated from a conventional phase-

modulated signal by forming the Hilbert transform (br means, say, of a

x/2 phase shifter) and amplitude modulating with the exponential; limiting

in the receiver will then restore the lower sldebend and permit the use

of a conventional discriminator. Howver, the new threshold behavior

is not clear.

EXPONEWIAL (ENVELOP) W)IIWITI

Double Sideband

,
mM (t) g(t) - a log f(t), f(t) > 0

s(t) = fo(t) cos Wot + i fo(t) sin W0 t (10)

I8(t) I &- f?(t), Wo >>CM

The intermediate function g(t) is introduced in order to produce the

desired modulated effects in Eq. (11). The difference between Eq. (4) and

Eq. (10) is obviously trivial and the latter has been included mainly for

coupleteness; they are, of course, identical for a - 1.

 13
Conventional FM

1I U -a1 Wo WO*f
i
SSB: Pveak 5.1db

Pave
%a-A. %o o+ .

a) Modulation index =

Conventional FM ,11ilii ,
Ppeok

SSB" t =78 8,.

db 1
C%-.a W o+a

b) Modulation index a =3

Conventional FM

ftao- wo Wao+.

SSB: = 8.9 db I
W- A( CWo wen

C) Modulation index 5

Fig.2 -Magnitudes of spectral components for conventional

and SSB forms of FM with sinusoidal modulation
14i

Single Sidebwa

MWt *E-t
e g~
9 g(t) -a log r(t), f(t) > 0

8(t) - fo(t) Co1[ 0t + a log f(t)]

+ ± a(t) sin [1wot + a logfit

f'(tlog
0t)] r

The SSB version given byr Eq. (11) is related to Eq. (7), from which
it can be derived byr considering the logarithm of the modulating signal.
lbs reason these two classes of exponential mdulation have been mae
distinct here is that Ieqs. (6) A (7) contain the modulating sigma in
the pAse of the modulated sigma, while Eqs. (10) and (U1) contain the
modulating signal in the envelope; hence, the designations.
The exponent a Is important because of the relationship which the
choice of its value as either 1 or 1/2 bears on the question of compatible
S53 modulation.* Powers (2) considers the case a a 1/2 so that the modula-
ting sigma is contained in the square of the envelope, thus requiring
a squere-law envelope detector for distortionless reproduction. Since
It is this case which produces a modulated signal having the same spectral
width as a conventional SSB sigma, he concludes that true compatibilityr
(i.e.,A a system usable with a linear envelope detector) is impossible.-
IrancV(4)reaches the sae conclusion and, incidentally, labels the
case optimal Amplitude and Phase 141dulation (OAPI4) for Square-Law Detec-
tion. He also considers a = 1 or OMP for Linear Detection and points
 15

out that this modulated signal occupies a bandwidth Just twice the maximum

modulating frequency and that it is compatible with a linear envelope

detector.

Naturally, there would be no advantage to a compatible SSB system

which has the same spectral width as a conventional DSB system, but

Lyannoy adds that the bandwidth can be halved in practice bV suitable

filtering because of the characteristics of speech. This derives from

the fact that "a real signal has components whose modulation depth approaches

100 per cent only in the range up to 1500 or 2000 c/o, the modulation depth

decreasing rapidly thereafter." Bence, removal of the upper half of the

signal spectrum permits the large-modulation-index, low-frequency components

to be received without distortion; while the small-modulation-index, high-

frequency components are received with a small but tolerable distortion.

Kahn (3 ) makes virtually identical observations with regard to his

Compatible SSB (CSSB) system, but it is outside the scope of this study to

make a comparative analysis. It is apparent, however, that the two system

have a close fundamental relationship.

 V. CONCLUSION

It has been shown that analog-modulated signals are representable in

general as the product of a carrier function and a modulation function.

When an analytic signal is chosen for the carrier, its effect spectrally

becomes one of translating the signal spectrum to the vicinity of the

carrier frequency. The process is one of pure translation when the

carrier is harmonic, exp(L.0 t), or one of translation plus spreading -when

the carrier has a finite spectral extent (i.e., when noiselike).

The useful form of the modulation function are the linear and ex-

ponential functionals from which three classes of modulated signals can

be derived, viz., amplitude, pae, and envelope modulation. Within

these classes, the use of the purely real or analvtic signal forms of the

modulating signal leads naturs.y to the DSB and SSB (i.e., two-sided

and one-sided) spectral forms, respectively, of the modulated signal.

The existing familiar forms of modulation are readily identified

within the above structure. In addition, a new form of modulation which

8 be called SSB F is revealed. It can be derived from a conventional

phase-modulated signal by an additional amplitude modulation using the

exponential function of the modulating signal's Hilbert transform. The

resulting modulated signal will have a one-sided spectrum about the

carrier frequency and be compatible with existing FM receivers. The

advantage is a decrease in signal bandwidth; the disadvantage, the loss

of a constant transmitter output level.

From a practical viewpoint, this form of FM may find little accept-

ance in wide-band systems, not only because of the unfavorable peak-to-

 17

average power ratio required of the transmitter, but also because of the

lack of any particular pressure to conserve the radio spectrum at present

in services, such as broadcast FM or specialized military and civilian

communication systems.

on the other hand, some of the bands assigned to public use are quite

crowded and users are therefore disposed toward modulation techniques

which make more efficient use of the available spectrum. A compatible

narrow-band SSB FM system of the type described here may prove useful in

such applications by narrowing the signal bandwidth for a given modulation

index without sacrificing the "improvement" or the immunity to impulse

noise associated with FM. The peak-to-average power ratio of a narrow-

band SSB FM signal may prove an acceptable burden in that case.

Observation of the comparable (i.e., SSB) form of envelope modulation

comends Powers' suggestion of the application of the non-compatible,

square-law type of envelope modulation (n - 1/2 in Bq. (9)) as a meant of

SSB data transmission. (2) Conventional SSB is adequate for speech or

music because of the well-known tolerance of the human ear for phase or

even frequency errors in reproduction. However., data vaveforms mst be

reproduced without such errors, so exact carrier re-insertion is required

(a procedure which is frequently difficult). The fact that the modulating

signal is contained in the envelope of Eq. (9) indicates that square-law

envelope detection should result in distortionless reception, thus obvia-

ting carrier re-insertion without sacrificing spectral econorv.

18

REFERENCES

1. Chakrabarti, N. B., "Combined AM and PM for a One-Sided Spectrum,"

Proc. I.R.E., Vol. 47, No. 9, p. 1663, Sept. 1959. Comments:
A. . Taylr, Proc. I.R.E., Vol. 48, No. 5, P. 953, may 1960;
R. M. Goldsen, M. R. Schroeder, and N. B. Chakrabarti, Proc. I.R.E.,
Vol. 48, No. 6, pp. 1094-1095, June 1960.
2. Powers, K. H., "The Compatibility Problem in Single-Sideband Trans-
mission: Proc. I.R.E., Vol. 48, No. 8,p. 1431-1435, Aug. 1960.
Comment: L. R. Kahn, same issue, p. ;5Cg.*

3. Kahn, L. R., "Compatible Single Sideband," Proc. I.R.E., Vol. 49,

No. 10, pp. 1503-1527, Oct. 1961.

4. lyannoV, V. P., "Use of Simultaneous Amplitude and Phase Modulation

in Radio Broadcasting," Izvesti a Yxsshikh Ucheb kh Zaveden y,
.SSO,
Seriya Radiotekhnika (News of Higer EHucational Institutions,
Ministry of Higher and Secondary Specialized Education, Radio
Engineering Series), Vol. 4, No. 1, pp. 84-93, 1961, Kiev. Available
as U.S. Joint Publications Research Service Translation, JPRS; 10920,
Nov. 10, 1961.

5. Titchmarsh, E. C., Introduction to the Theory of Fourier Interalso

Ch. V., Oxford Press, 1937.
6. Erdelyi, A., W. 4agnus, and F. G. Tricomi, Tables of Integral rane-
forms, Vol. II, Ch. XV, McGraw-Hill Book Company, Inc., 191.

7. Gabor, D., "Theory of Communication," Jour. I.E.E., Vol. 93, Part III,
pp. 4.29-4c.7, rov. 1946.

5. Ville, J., "Theorie e- Application de la Notion de Signal Analytique,"

Cables et Trans., Vol. 2, pp. 61-74, Jan. 194$. Available as RAND
Translation T-92, Aug. 1, 1958, by I. 3lin.

9. Oswald, J. R. V., "The Theory of Analytic Band-Limited Signals Applied

to Carrier 3ystezs," Trans. I.R.E., Vol. CT-3, No. 4, pp. 244-251,
Dec. 1956.
10. Arens, R., "Complex Processes for Envelopes of Normal Noise,"
Trans. I.R.E., Vol. IT-3, No. 3, pp. 204-207, Sept. 1957.

11. Dugndji, J., "Envelopes and Pre-Eavelopes of Real Waveforms,"

Trans. I.R.E., Vol. IT-4, No. 4, pp. 53-57, March 1958.

12. Kelly, E. J., I. S. Reed, and W. L. Root, "The Detection of Radar

Echoes in Noise," Jour. S.I.A.M., Vol. 6, No. 2, pp. 309-341,
June 1960; and Vol. F, No. 3, Pp. 481-5O7, Sept. 1960.
 19

13. awgdady, E. J., "Analog Mbdulation Systems," Lectures on Communication

system Theory, Ch. 19, McGraw-Hill Book Company, Inc., New York, 19B1.

14. Wheeler, H. A., "The Solution of Unsymmetrical-Sideband Problems with

the Aid of the Zero-Frequency Carrier," Proc. I.R.E., Vol. 29, No. 8,
pp. 446-458, Aug. 1941.
15. Woodward, P. M., Probability and Information Theory, with Application
to Radar, McGraw-Hill Book Company, Inc., New York, 1953.

16. Wozencraft, J. M., "Sequential Reception of Time-Variant Dispersive

Transmissions, Ch. 12, loc. cit., Ref. 13.

17. Guillemin, E. A., The Mathematics of Circuit Analysis, Ch. VII, Art. 23,
Wiley and Sons, New York, 1949.