PROCEEDINGS OF THE I.R.E.

Noise-Suppression Characteristics of

Pulse-Time Modulation*SIDNEY MOSKOWITZt, MEMBER, I.R.E., AND DONALD D. GRIEGt, SENIOR MEMBER, I.R.E.

Summary-An experimental investigation of the noise-suppres-sion characteristics of pulse-time modulation is outlined. Impulsenoise and thermal-agitation or fluctuation noise are treated. Theeffects of these types of noise and the improvements obtainedthrough the use of limiters, differentiators, and multivibrators arepresented graphically.

INTRODUCTION

C, OMMUNICATION SYSTEMS utilizing pulse-time modulation and the general properties ofthis type of modulation have been described in

the technical literature.1" Briefly, in this method,instantaneous samples of the modulating wave vary thetime of occurrence of a pulse subcarrier. Thus, a par-ticular value or sample of the modulating signal is rep-resented by the displacement of the pulse in time withrespect to a synchronizing pulse or time reference, andthe frequency of the modulating signal is given by therate of change of pulse displacement.One of the important characteristics of pulse-time

modulation is its noise-reducing properties. The noisecan be of two distinct types: impulse noise, and thermal-agitation or fluctuation noise. The first may be shortimpulses caused by electrical disturbances or they mayoriginate in neighboring systems. Thermal-agitationand other noises that have a similar spectral distributionsuch as "shot" noise are usually contributed by thefirst few stages of the receiving equipment and, to alesser degree, by the transmitter if "jitter" exists.6'7 Ingeneral, thermal-noise effects are most important be-cause they determine the lower limit of sensitivity and,hence, such transmission parameters as bandwidth andpower.

* Decimal classification: R148.6. Original manuscript received bythe Institute, April 30, 1947; revised manuscript received, September15, 1947. Presented, 1947 I.R.E. National Convention, March 6,1947, New York, N. Y.

t Federal Telecommunication Laboratories, Inc., New York,N. Y.

I E. M. Deloraine and E. Labin, "Pulse-time modulation," Elec.Commun., vol. 22, no. 2, pp. 91-98; 1944.

2 D. D. Grieg and A. M. Levine, "Pulse-time-modulated multi-plex radio relay system-Terminal equipment," Elec. Commun., vol.23, pp. 159-178; June, 1946.

3 F. F. Roberts and J. C. Simmonds, "Multichannel communica-tion system," Wireless Eng., vol. 22, p. 538; November, 1945.

4 R. E. Lacey, "Two multichannel microwave relay equipmentsfor the United States Army communication network," PROC. I.R.E.,vol. 35, pp. 65-70; January, 1947.

6 B. Trevor, 0. E. Dow, and W. D. Houghton, "Pulse-time divi-sion radio relay," RCA Rev., vol. 7, pp. 561-575; December, 1946.

6 H. T. Friis, "Noise figures of radio receivers," PROC. I.R.E.,vol. 32, pp. 419-422; July, 1944.

7 C. W. Hansell, "Radio-relay-systems development by the RadioCorporation of America," PROC. I.R.E., vol. 33, pp. 156-168; March,1945.

In a pulse-time system in which the transmitted in-telligence is derived from the timing of a pulse edge,noise may displace the pulse edge from the value cor-responding to the modulating signal. Noise impulsesalso may modulate other characteristics of the signalpulses such as amplitude, width, and slope of the pulseedges, but are ultimately translated into pulse-time dis-placement. The optimum signal-to-noise ratio is realizedwhen all effects of noise, other than time displacement,are eliminated by suppression devices in the receiver.The method by which noise distorts the signal pulses

and causes a distortion of the pulse edge timing is shownin Fig. 1. It should be noted that only d.c. or "video"pulses are treated; they are considered independently ofthe method of transmission. Where amplitude modula-tion of an r.f. carrier is utilized, noise-reduction prop-erties are the same as at video frequencies. Wherefrequency modulation of the carrier is by means of time-modulated pulses (frequency-shift keying), the video-frequency relationships with respect to noise are like-wise similar, but reduced by the ratio of the improve-ment factor attributable to f.m. transmission.

Fig. 1-S and N are signal and noise impulses. Output noise =a'b'.(S/N) output =D/a'b', where D = modulation displacement.ab =bd = G. a'b'/ab ==a'b'/G = (N/S) input. (S/N) output = (S/N)input DIG.

A gate limiter will remove noise amplitude modula-tion as well as noise occurring between pulses. The fol-lowing discussion assumes an idealized pulse that buildsup to maximum amplitude and decays in a time de-termined by the transmission bandwidth. Under theseconditions, both the noise and signal pulses can be rep-resented approximately by triangular shapes.The time displacement of the pulse edge caused by a

noise impulse is shown in Fig. 1 as a'b'. A narrow gatelimiter is set at the pulse amplitude corresponding tothe peak of the noise. Hence, as shown, the signal-to-noise ratio at the threshold level represented by timemodulation is improved over that obtained at the re-ceiver input by the factor D/G, where D is the modula-tion displacement and G is the build-up time. It is well

446 A pril

Moskowitz and Grieg: Noise-Suppression Characteristics of Pulse- Time Modulation

known that the frequency band necessary to support apulse build-up time G is inversely proportional to G.Therefore, the signal-to-noise improvement ratio is di-rectly proportional to the frequency bandwidth of thereceiver, provided the transmitted bandwidth is equalto or greater than the receiver bandwidth.

It should be pointed out that it is not possible to de-rive in a simple manner the exact constants of propor-tionality. In practice, purely triangular pulses are notcommon, nor is the pulse edge truly linear. Further-more, it is necessary to know the relation between theequivalent noise peak (N) and the r.m.s. noise voltage.

It is interesting to note that the input signal-to-noiseratio in a time-modulation system in which the fre-quency band is optimum for a given pulse width is con-stant with respect to the frequency band, and dependsonly on the average power. Corresponding to the in-crease in noise amplitude with increasing bandwidth,the pulse amplitude will be increased in the same pro-portion, because the narrower pulse for the same aver-age power will represent greater peak power. Thus, for agiven average power, the improvement in signal-to-noiseratio that can be realized with time-modulated pulses isproportional to the frequency band, as is the case withfrequency modulation, but, unlike frequency modula-tion, the improvement ratio continues to increase withincreasing bandwidth.

This analysis of pulse-time modulation is based on ademodulation system in which the pulse edge defines thepulse timing. It is possible for the leading and trailingedges of the pulse to be distorted in opposite directionsby noise pulses. A further gain of approximately 3 db insignal-to-noise ratio may be obtained by utilizing thecenter of the pulse for demodulation. This gain is real-ized, however, at the expense of system complication.For example, a system may be visualized whereby bothpulse edges are demodulated and the outputs added toreinforce the modulating signal, but partially cancelnoise.Many types of noise pulses, which run the gamut of

all shapes and variations in time consistent with thebandwidth of the receiver, might be imagined. As far astheir interfering effects are concerned, only those edgesof noise pulses that actually coincide in time with thesignal pulse edge will cause an a.f. noise output.

I. NOISE TESTS

The types of noise suppressors described in this paperthat have been used in pulse-time-modulation receiversare gate limiters, differentiators, and multivibrators.Tests were conducted wherein a train of time-modu-lated pulses was transmitted to a pulse-time demodu-lator over a wire link in which the noise-suppression cir-cuit under test was inserted.A block diagram of the apparatus is shown in Fig. 2.

The wide-band fluctuation noise contained frequencycomponents from 30 c.p.s. to 1.5 Mc. The noise gener-ated in a resistor was amplified by an 11-Mc. i.f. am.

plifier having a bandpass characteristic of ± 2.5 Mc.The band of noise at the intermediate frequency was

PULSE-TIME FILTER AUU OUTPUTDEMOMATO 10-300 EQUNCYPOWER

CYCLSAPLIIERMETER

Fig. 2-Block diagram of pulse-time-modulation noise-test setup.

transposed to video frequency and the bandwidth lim-ited to 1.5 Mc. by an adjustable output filter.

In carrying out the noise tests, provision was madefor substituting a pulse-interfering source for the fluc-tuation-noise generator. The pulse-interference sourceconsisted of a combination multivibrator, differentiator,and shaper circuit. This device generated pulses of aconstant width and with a repetition rate continuouslyvariable from 250 to 1000 pulses per second. The ampli-tude of this interfering signal was continuously adjusta-ble without destroying the pulse shape.The double-gate limiter, shown in Fig. 3, consisted of

two pentodes having individually adjustable grid-biascontrols to determine the position of the upper andlower levels of limiting.

'C-

Fig. 3-Double-gate limiter used as protection against interference.

A resistance-capacitance type of differentiator wasused in conjunction with a limiter, so that the leadingedge of the signal pulse could be selected and demodu-lated. The multivibrator was of conventional type, asmay be seen from Fig. 4. Input and output couplingstages isolate the multivibrator from external effects. Inaddition to variable time constants, an input attenuatorwas provided to control the amplitude of the synchroniz-ing signal supplied to the multivibrator. This input at-

1948 447

PROCEEDINGS OF THE I.R.E.

tenuator was constructed so that neither the input pulseshape., output pulse shape, nor the multivibrator timeconstant was affected during manipulation.

8+

Fig. 4-Multivibrator used as protection against interference.

The characteristics of the pulse-time transmissionsystem were as follows:

Pulse repetition rate (p.p.s.)Pulse period (,s.)Modulation displacement (As.)Pulse build-up time (Mus.)Pulse decay time (Ms.)Pulse width at base (Mus.)Audio modulation frequency (c.p.s.)Demodulator audio-frequency pass band (c.p.s.)

12,00083±80.751.52.5400100 to 3000

Oscillographic comparison was made of the inputpulse signals and interfering noise. For this measure-ment the horizontal sweep voltage usually was re-moved from the deflecting plates, and the magnitude ofthe vertical traces compared.Peak values of noise were measured in this manner

for convenience. A comparison measurement by meansof a thermocouple was made to determine the ratio ofthe peak value of noise as measured by the oscilloscopeto the r.m.s. value, and this ratio was found to be 3.5.

IL. FLUCTUATION NOISE

In the first test on fluctuation noise, the output signal-to-noise ratio was compared to that at the demodulatorinput without noise-suppression devices. The resultsare shown graphically in curve A of Fig. 5. The inputsignal-to-noise ratio is given in terms of peak ampli-tude, because of the oscillographic method of measure-ment. Thus, 6 db here corresponds to a noise peak equalto one-half the modulation pulse peak.The output signal-to-noise ratio is shown to be pro-

portional to the input ratio with no improvement. Un-der these conditions, the output signal contains noisefor two main reasons. First, noise is introduced directlyas amplitude modulation in the demodulator. Secondly,some noise is introduced as time modulation because ofthe inherent nonlinearity of the demodulator. It is ob-vious that, in a multichannel system, where only se-lected groups of pulses are applied to the demodulator,some improvement would be obtained because a largeportion of the pulse-repetition period would be blankedout. Thus, only noise appearing within the time allottedto one channel would affect the output signal. The sys-

tem used here may be compared to a multiplexed systemby extrapolating the results obtained. The output signalis proportional to the modulation displacement, so that,for a maximum modulation displacement of ± 40 us., anoutput signal-to-noise ratio 5 times greater would beobtained. In the multichannel pulse system, the samesignal-to-noise conditions for maximum individual chan-nel displacement would be obtained, since the noisepower is less per channel by the ratio of channel time tobase pulse period. (This example holds, of course, onlyfor the same pulse peak power for both systems.)

hiw0

O J

_'z LoO20aX

D-j. Z00 o-_; -

en- mm

0 2 4 6 8 10 12 14 16 18 20 22 24 26

INPUT SIGNAL-TO-NOISE PEAK-AMPLITUDERATIO IN DECIBELS

Fig. 5-Output signal-to-noise power ratio plotted against inputsignal-to-noise peak-amplitude ratio for fluctuation noise and apass band of 1.5 Mc. Curve A is with no protection. Curve B isfor a double-gate limiter. Curve C is for two double-gate limitersseparated by a differentiator. Curve D is for three double-gatelimiters and two differentiators alternately connected.

Curve B of Fig. 5 illustrates the results of using adouble-gate limiter to remove a.m. noise. It can be seenthat no critical threshold occurs, although the outputsignal-to-noise ratio begins to increase more rapidlywhen a 2:1 ratio is obtained at the input. There is a gainof about 12 db over the ratio obtained in the precedingtest. The slight effect of the limiter is accounted for bythe presence of width-modulation noise, which may beremoved by a differentiator and second double-gate lim-iter. The action of such devices is shown by curve C ofFig. 5. A definite threshold level is obtained, above whichnoise suppression is considerable. By further differentia-tion and limiting, the improvement above the thresholdis further increased as illustrated by curve D.The function of successive stages of differentiation

may be accomplished by a multivibrator that is syn-chronized by the signal pulses. The multivibrator fur-nishes a pulse whose leading edge corresponds in timeto the leading edge of the synchronizing pulse, andwhose trailing edge is a function only of the multivibra-tor time constants. Only the leading edge is selected fordemodulation. In addition, the limiting effect is ob-tained by the triggering action of the multivibrator. Theresults obtained with this device are shown in Fig. 6,

54 D

48 ..000001- c-000042

36

30

24

8

1 2 -00006

0

6

448 April

Moskowitz and Grieg: Noise-Suppression Characteristics of Pulse-Time Modulation

whence it can be seen that superior improvement is ob-tained at and above the threshold level.

w03 48

042 4 PI

036

0

24

O 12

CL- 6

012

0 2 4 6 8 10 12 14 16 18 20 22 24 26

INPUT SIGNAL-TO-NOISE PEAK-AMPLITUDE RATIO IN DECIBELS

Fig. 6-Conditions similar to Fig. 5, but with adouble-gate limiter and multivibrator.

Further tests were made in which a stage of differ-entiation was added to the multivibrator. No significantfurther improvement was noted, showing that the mul-tivibrator entirely removed the width-modulation noise.

It can be concluded from the foregoing tests that themaximum signal-to-noise ratio improvement can be ob-tained from a pulse-time-modulation system either byincluding successive stages of limiting and differentia-tion, or by incorporating these functions in a multivi-brator. In this manner, the reduction of output noise bythe elimination of all noise modulations, except that ofedge timing, is accomplished.To determine the noise improvement of pulse-time

modulation as a function of the bandwidth utilized, atest was made wherein the video-frequency and noisebandwidths were simultaneously varied, and the outputsignal-to-noise ratios at the threshold were measured.The resulting curve, Fig. 7, shows that the threshold

w

CD4c-J0

4

Ci

n

00

z

0

48

0i-j0

Cohi

I.-40

4i

portionality between input and output signal-to-noiseratio may be obtained, since

output peak S/N = input peak S/N (KDFV)

where K is a constant, D is the modulation displace-ment, and F, is the video-frequency bandwidth. Thevalue of K can be determined from Fig. 7 to be equal to7.5.

Since the optimum a.f. bandwidth, equal to 0.5 thepulse-repetition rate, was not used in these tests, theabove equation may be corrected by a factor of 1/V/2.Thus, for the optimum system, we have

output peak S/N = input peak S/N 5.3 DF,.Furthermore, the maximum modulation displacement

is equal to 1/2fp, where f, is the pulse-repetition rate,and sincefp may equal twice the highest modulation fre-quencyf., we may then conclude that

output peak S/N = input peak S/N 1.3-fa

This equation defines the signal-to-noise improve-ment for the optimum pulse-time-modulation system.The result may be applied to a multiplexed system byassuming the highest modulation frequency to be thevalue for one channel multiplied by the number of chan-nels in the system.

III. IMPULSE NOISEUnder some conditions, the suppression of impulse

noise may be of interest. A study, similar to the forego-ing for fluctuation noise, has been made in which theimpulse-noise-suppression characteristics of pulse-timemodulation have been measured. The noise source forthe previous tests was replaced by a generator of pulseswhose repetition rate was variable. Fig. 8 illustrates theresults obtained without any suppression devices. Theoutput interference varies almost directly with inputinterference. However, the lower-repetition-frequency

It

WoAJ

o20

ow

-

20

z osn a

o~~~~~~~~~~~~~

tLa0.ID-

0 0.1 0.2 0.3 0.4 0.5

BANDWIDTH IN MEGACYCLESFig. 7-Signal-to-noise ratio plotted against bandwidth for thermal-

agitation noise and signal-pulse displacement of ± 8 ,us.

signal-to-noise ratio is proportional to the bandwidthutilized. From this curve, the empirical constant of pro-

24

18

12

6

0

.6

-1-I-

i |i

-F,.F2

. _~z

0 0 2 4 6 8 10 12 14 16 18 20 22 24 26

INPUT SIGNAL-TO-NOISE PEAK-AMPLITUDE RATIO IN DECIBELS

Fig. 8-Output signal-to-noise power ratio plotted against inputsignal-to-noise peak-amplitude ratio for 3-,us. noise pulse havingrepetition frequencies of F1=250, F2=500, and F3=1000. Noprotection.

pulses cause less interference than those at higher rates.The pulses being of constant width, doubling the pulsefrequency increases the noise power by 3 db, accountingfor the variation shown on the curves. All pulses react

I e&2fr-.A!n I

PIi-.

I_

4491948

PROCEEDINGS OF THE I.R.E.

C 24 W_

en_j IsB00-0§ | 1 <

o 12

I - 6

sz _ =F

o 0 2 4 6 8 10 12 14 16 18 20 22 24 26

Fig. 9-Conditions similar to Fig. 8, but with double-gate limiter.

on the demodulator, and their fundamental and somehigher harmonics are within the pass band of the a.f.system. As a result, there is very little, if any, improve-ment in this unprotected system.

54 - = - - - -

48 - - - -.

42--

36 - - .- - -

24 - - - - -

18 _ __ _

O--=-.-=-_ -_6--

-6 - - - - - -

0 2 4 6 8 10 12 14 16 18 20 22 24 26

INPUT SIGNAL-TO-NOISE PEAK-AMPLITUDE RATIO IN DECIBELS

Fig. 10-Conditions similar to Fig. 8, but with two double-gatelimiters separated by a differentiator. Noise pulses are at a rateof 500 p.p.s.

By incorporating a double-gate limiter, a 6-db im-provement is obtained as shown in Fig. 9. When thesignal-to-interfering-pulse ratio is above the 2: 1 thresh-old, only those pulses that occur at the same time as thesignal pulses can cause interference. This effect takesplace at the beat frequency of the two sets of pulses,causing a distortion of the signal in the same manner asthe fluctuation noise, but with a more limited frequencyspectrum. Below the threshold, the signal and interfer-ing pulses are limited to the same amplitude, so that theoutput signal-to-noise ratio is constant.By adding a differentiating circuit and a second stage

of limiting, a sharply defined threshold is obtained. Theresults of a test using this suppression device are shownin Fig. 10. The steep threshold indicates that the sup-pression is more complete than that obtained with fluc-tuation noise.

IV. CONCLUSION

The described tests and results have illustrated thesignal-noise capabilities of a pulse-time-modulationsystem. In addition, the effectiveness of limiters, dif-ferentiators, and multivibrators in realizing optimumnoise improvement for both thermal and agitation noiseand impulse interference has been demonstrated. In ad-dition to normal pulse displacement, noise may resultfrom variations in pulse amplitude, width, and edgeslope.The various devices tested have proved effective

in reducing such noise. The uses of these devices are,therefore, indicated to take maximum advantage of thebandwidths utilized for the system. A communicationsystem operating in this manner may then be designedfor minimum transmitter power necessary to produce aconservative output signal-to-noise ratio.

Magnetoionic Multiple Refraction at High Latitudes*S. L. SEATONt, SENIOR MEMBER, I.R.E.

Summary-Experimental ionospheric soundings examined byScott and Davies are cited, with a short discussion of the interpreta-tion these authors offer for multiple refraction at high latitudes. Thetheory of magnetoionic multiple refraction of Appleton and Builderis discussed with especial regard to effects to be expected in highgeomagnetic latitudes. Experimental evidence is offered to show thatthe "Z" component of Scott and Davies is probably the longitudinalordinary ray predicted by Appleton and Builder and by Taylor whencollisional friction is appreciable. On the basis of certain assumptions,the collisional frequency near Fairbanks, Alaska, is calculated asabout 4(10)4 at 300 km. height.

* Decimal classification R113.613. Original manuscript receivedby the Institute, July 21, 1947.

t The Geophysical Observatory, University of Alaska, College,Alaska.

ECENTLY Scott and Davies' have discussed thefine structure of the ionosphere in high northernlatitudes and have shown that three wave com-

ponents, corresponding to ordinary ray, extraordinaryray, and a third which they designate the Z com-

ponent, are frequently returned at vertical incidencefrom the ionosphere. Determinations were made bymeans of ionospheric soundings, and the three com-

ponents were interpreted as resulting from magnetoionic

I Joint Meeting, The International Scientific Radio Union, andAmerican Section, The Institute of Radio Engineers, paper No. 25,Washington, D. C., May 6, 1947.

450 April

(A-Jw

Uto

0Z

0Ct:

30

z0

a.J

~2

0

z

(D

5