PULSE-WIDTH AND PULSE-TIME MDULATORS by Seth H. Washburn B. S. Mass. Inst. of Technology 1944 SUBMITTED IN PARTIAL FULFILLMENT OF TIE REQUIREENTS FOR THE DEGREE OF MASTER OF SCIENCE - from the VASSACHUSETTS INSTITUTE OF TECHNOLOGY 1947 Signature redacted Signature of Author Department of Electrical Engineering, May 22, 1947 Signature of Professor in Charge of Research Signature redacted Signature of Chairman of Department Committee on Graduate Students Signature redacted -ii(.I Q TABLE OF CONTENTS Page INTRODUCTION........................1 Chapter I. A. A Brief Discussion of Pulse-Width and Pulse-Time 1fodulation..........,...1 1. Pulse-Vidth Modulation....,........,.l 2. Pulse-Time "&Iodulation.................8 3. Demodulation ........ B. .... . .. .. 9 The Historical Background.............10 1. The Heising Modulator................10 2. Finch's Modulation System............12 3. R. D. Kell's Modulators...........'...14 4. Cathode-Ray-Tube Modulators..........16 5. Modulators for Multiplex Systems.....18 Nature of Problem under Concideration.20 C. Chapter II. A. THE MIXER-TYPE MODULATORS............24 The Kell-Kretzmer Circuit.............24 1. Kell-Kretzmer Circuit Design.........25 2. Experimental Procedure...............26 3. Experimental Results.... B. .29 An OriginalModulator.................33 1. The CircuitDesign...................33 2. Experimental Results................36 C. Stability of Mixer-Type Modulators....38 28"783. -iii- Page Chapter III. THE NULTIVIBRATOR 1LDULATORS........43 The First Multivibrator Modulator.....48 A. 1. Circuit Design....................48 2, Experimental Results................0.50 A Second Multivibrator Modulator......56 B. 1. Design of the Circuit................56 2. Experimental Results................57 C. Stability of the Multivibrator Modulators. Chapter IV. . .. .. .. .. .. .. . .. . .. . . ..... 61 CONCLUSION AND INTERPRETATION OF tRESULTS . . . . . . . .. . . . . .. .. . . ... * . . * *.. . 65 BIBLIOGRAPHY.............................72 -iv- ACKNRIfEDGENENT The author wishes to express his gratitude to the supervisor of this thesis, Professor L. B. Arguimbau, for his patient assistance. Professor Arguimbau, as well as serving as the source of essential advice and guidance, provided a certain freshness of approach which was, at many times, most sorely needed. Many thanks are also due Professor H. J. Zimmermann and Mr. E. R. Kretzmer for the time they spent in helping the author wrestle with the problem of multivibrator modulation. -l- CHAPTER I INTRODUCTION A. A Brief Discussion of Pulse-Width and Pulse-Time Modulation. Modulption may be defined as "the process of producing a wave, some characteristic of which varies as a function of the instantaneous amplitude of a second wave, the modulating signal." Thus, pulse modulation refers to the modulation of a chain of rectangular pulses by a signal voltage. Characteristics of a pulse chain which may be varied are the widths of the pulses, the positions, or times, of the pulses with respect to other pulses in the chain, the amplitudes of the pulses, and, finally, the pulse-i-epetition frequency. In this discussion only the first two methods are of interest. 1. Pulse-Width Modulation. In pulse-width modulation, the widths or durations of the pulses are varied as a function of the instantaneous amplitude of the signal wave. For example, consider a train of pulses of voltage amplitude E, duration d, and pulse repetition rate p, as shown in Figure la. Consider also a signal voltage of complex waveform as in Figure 1b. It is possible by electronic .. 2-. means to combine the two waveforms in the process of modulation so that the resultant waveform will appear as in Figure 1c. (a) Unmodulated Pulse Train (b) Signal Voltage (c) Width-Modulated Train Figure 1. Pulse-Width Modulation. The widths of the pulses in the modulated chain depend on the amplitude of the signal at certain instants, either fixed in time or at times dependent on the signal itself. If the pulse widths are a function of the signal amplitude at instants corresponding to the midpointl of each pulse in the unmodulated chain, as shown in Figure 2, the modulation would be termed "symmetrical" since the pulses vary in width symmetrically about their midpoints. In effect, the signal waveform is "sampled' at regular time intervals, t, in Figure 2, the widths of the pulses in the chain depending upon the amplitudes of these samples. Signal i Width-Modulated Pulse Train Figure 2. Symmetrical Width-Modulation. Apposed to symmetrical width modulation is the asymmetrical, or disymmetrical, type. Here, the leading and lagging edges of the pulses vary in position by unequal amounts. There are two general modes of asymmetrical modulation, In "Mode 1," shown in Figure 3, the pulse edges shift by an amount dependent on the signal voltage at fixed instants; namely, at the instants of the leading and lagging edges of the unmodulated pulses. Thus, the leading edge of each pulse in the modulated chain will be shifted as a function of the signal at one fixed instant, and the position -4- of the lagging edge is a function of the signal at another fixed instant. Sigal Width-Modulated Pulse Train Figure 3. Asymmetrical Width Modulation, Mode 1. Dotted lines indicate positions of unmodulated pulses. In the second mode of operation, illustrated in Figure 4, the shifts of the pulse edges are again asymmetrical, but the magnitudes of the shifts depend not upon the amplitudes of the signal voltage at fixed times but on the signal at the actual instants at which the leading and lagging edges of the pulses in the modulated chain occur. Thus, in "Mode 2" operation, the pulse widths are a function of the signal amplitude at instants which are themselves a function of the signal amplitude. E. R. Kretzmer, of the Electronics Research Laboratories at Massachusetts Institute of Technology, has shown that a Foutier series may be derived for Signal IF, Width-Modulated Pulse Train Figure 4. Asymmetrical Width Modulation, Mode 2. "Mode 2" pulse-width modulation. For a width-modulated pulse train with voltage amplitude of unity, an unmodulated pulse width of d' ( sec.), pulse repetition rate p (rad./sec.), and signal frequency q (rad./sec.), the f ollowing series obtains: pd'(l+ks int )+ 21JJ 41W + Iona, ,- +I 0.01,-- #j.g, n (n f) sin(nRpd')) cos (npt ) e, 2 2 (npd') sin (npd') (cos (np~mg) t+c os (np-mq) t) 2 (ndi) 2 2 cos (ap~d') (sin (np*-mq) t-sin (np-mq) t 2 v,.here k modulat ion index (dma.- &'in, ( tax.+ d'mi n.. and m and n are integers. This is-.obtaindd by -subs titut ing- in'to the"-filzilikr- series for a rectangular pulse chain, ZZ1 s in(]Ud)l cos (npt) 2 v.=, n * e rp 2-W the equation indicating the dependence of d upon the signal: dz (d'+kd'(sin qt), where d' is the unmodulated pulse width, and simplifying the result. Kretzmer has adequately justified this substitution in his paper. From the above expression for the modulated chain it can be seen that this chain contains a directcurrent component of amplitude (pd')/(2r), a signal- frequency component of amplitude (pdA)/(2w), and an infinite number of components with frequencies (np~mq) and amplitudes dependent on the Bessel Function of (npd'k)/2. It is interesting to note that, in the above expression, n cannot equal zero. This indicates that components of frequencies mq are not possible. In other words, harmonics of the signal frequency, theoretically, are riot geerafted in Atbe .- odulation process. Obviously, the higher the pulse repetition rate in comparison with the signal frequency, the smaller will be the amplitudes of the undesired (nprmq) components. It has been found that a ratio of p/q equal to 2.5 or 3.0 is high enough for practical A purposes. The series which expresses the width-modulated 3 chain obtained by "Mode 1" modulation is as follows: .. so mnp) c os((np~mq)(t-') k m~Jnd e a + (1) J m) d + y(-177 Jj'/2) (n+s1 cos ((np-mq) (t- d')) J(d'k(np+sqj) sin ((np*sq) 2 sin(npd'/2) cos(pn(t-d'/2)) nTr where sviatb, and a and b are integers. All other letters are as defined above. This expression for the is derived in a manner similar to that used "lMde 2" series, except that the process is somewhat 3 longer and more involved. The expression indicates the presence of (np*mq) components as well as signal frequency and direct-current components. It can be seen that, in addition, since n can equal zero, harmonics of the signal frequency are generated in the modulation to process. By setting n equal to zero and m equal the one, two, and three, it is easily found that magnitude of the signal frequency component is / J, (kdq). (ej= 2e=i The second harmonic is given by le.1 J, (2dkq) 2 4_XL The expression for the third harmonic is 2 7ri Js(3dkQ) 3q/p* -8. Other methods of modulation which are based on these two modes are, of course, possible. One example might produce a modulated pulse train in which the leeding edges of the pulses were fixed in position and the lagging edges varied in position in accordance with either "Mode 1" or "Mode 2." Another possibility might be a train in which the leading edges varied as in "Mode 1" and the lagging edges, as in "Mode 2. 2. Pulse-Time Modulation. In pulse-time modulation, the positions, or times, of the pulses in the modulated chain vary as a function of the signal voltage. There is a clear correspondence between pulse-width and pulse-time modulation since any time-modulated chain may be thought of as the difference between two width-modulated trains. Figure 5 illustrates a time-modulated chain composed of pairs of pulses, the first pulse in each pair fixed in time, and the second vaiying as a function of the signal. The figure also shows the corresponding widthmodulated chains. It can be seen that the discussion of "Mode 1" and "Mode 2" modulation will also apply to pulsetime modulation. Expressions for time-varying pulse chains may be written directly from the known series for the corresponding width-modulated pulse train. -9- Time -odulated Train Corresponding Width-Modulated Trains Figure 5. Pulse-Time Modulation. Because of the simple relationship between timemodulation and width-modulation it is obvious that corresponding chains will have the same frequency components (though the amplitudes of corresponding components will be different). 3. Demodulation. Demodulation is the process of recovering the signal waveform from the modulated pulse train. The standard method of demodulation is by the use of a low-pass filter adjusted to pass only signal frequencies. Of course, those (nptaq) components with frequencies less than that of the highest signal frequency to be passed will appear at the output of the filter, as well as the signal frequencies. Also, some harmonics -10- of the signal frequencies, either caused by nonlinearity of modulation or inherent in the mode of modulation, will be passed by the demodulating filter. In order to demodulate a chain of time-varying pulses by means of a low-pass filter, the chain is converted to the corresponding width-modulated pulse train. Thus, to determine the components appearing in the output of any pulse-time or pulse-width system, it is sufficient to investigate only the corresponding w-idth-varying pulse chain. B. The Historical Background. 1. The Heising Modulator. R. A. Heising of the American Telephone and Telegraph Company applied for a patent on the fundamental pulse modulator in 1924. In his system, an audio signal voltage was added to a 20,000 cps, sinusoidal waveform, and the summation was passed through two overdriven amplifier stages, as shown in Figure 6. The overdriven amplifiers were biased in such a manner as to clip the tops and bottoms from the input summation waveform. Thus, the output consisted of a series of nearly-rectangular widthvarying pulses. The signal waveform, the summation wave, and the output of the overdriven amplifier - -11 Figure 6. Heising's Pulse Modulator. stages are pictured in Figure 7. Examination of. this modulation method indicates that there was some lack of linearity in the system. The base widths of the pulses at the output of the overdriven amplifiers will depend upon the distances, or times, between points of intersection of the summation waveform and the line of cut-off voltage for the first amplifier tube. If the modulation were to be exactly linear, these distances would need to vary linearly with the signal amplitude. It is obvious that, in the Heising modulator, the points of intersection between the summation waveform and the cutoff voltage line do not vary linearly with the amplitude of the signal, but rather as a trigonometric function of the signal amplitude. As the ratio of the amplitude of the sign wave to that of the signal is increased, the non-linearity decreases, since the points of intersection fall upon the more nearly straight portions of the sine wave. Thus, for a very small -12- Sumnxat ion Wavef orm Output of Modulator Figure 7. Waveforms in Heising's Modulator. percentage modulation, the harmonic distortion would be low. It should be noted that Heising's modulation was essentially of the "Mode 2" type since the widths of the pulses in the modulated chain depended on the signal amplitude not at fixed intervals but at instants determined by the signal amplitude itself. 2, Finch's Modulation System. In 1929, J. L. Finch of the Radio Corporation of America substituted a supersonic saw-tooth voltage wave f or the sinusoid of Heising. The summation of the saw-tooth and signal voltages was fed to the grid of -13- a triode amplifier, as shown in Figure 8, driving the tube below cut-off for a portion of each negative swing of the summation wave. The tops of the summation waveform were clipped by a glow-discharge tube in the AM6 8* Figure 8. Finch's Pulse Modulator. plate circuit of this overdriven triode stage. The resulting waveform was further shaped by two more overdriven stages to give a final output of widthvarying pulses. Since a saw-tooth rather than a sinusoid was mixed with the signal, the modulation was, of the theoretically, perfectly linear. The widths at pulses var.ied directly as the signal amplitude pulses the instants of the leading edges of the in the modulated chain as in "Mode 2" modulation, as a The lagging edges were fixed in position can be result of the shape of the saw-tooth. It pulses at seen that the widths of the rectangular could be the output, with no modulating signal, .14- varied by adjusting the bias voltages on the first overdriven triode and glow-discharge tube. If these biases were of such values as to give output pulses of a duration equal to one-half the pulse repetition frequency, a modulation index of unity (or 100 percent modulation) would be obtained when the peak-topeak signal voltage at the secondary of the input transformer was equal to the saw-tooth voltage swing. 3. R. R, Kell's Modulators. R. D. Kell, in 1934, applied for a patent on a pulse modulator somewhat similar to that of Finch! However, instead of mixing a sawtooth wave with the modulating signal voltage across a common resistor, Kell applied a supersonic triangular wave to the grid of a pettode, and coupled the signal to the cathode, as diagramed in Figure 9. Thus, the mixing process was accomplished by the tube. The pebtode also served as a clipper since the grid was biased at the cut-off point, with no input. An overdriven triode served to remove the peaks of the output of the mixerclipper, and further pulse-shaping was carried out by two more overdriven stages. The similarity between Kell's system and that of J. L. Finch is evident, and, except for the -15- JUL Figure 9. R. D. Kell's Modulator. substitution of a triangular waveform for the sawtooth, the observed waveforms at various points in this system are similar to those obtained at corresponding points in Finch's modulator. Since a triangular wave -was employed, both leading and lagging edges of the pulses in the modulated chain varied in position. Again, the modulator was of the "Mode 2" type. Kell suggested a modification to his basic modulator in which a reristop-capacitor differentiating network and negative-pulse inverter were placed after the clippers in the basic circuit. The inverter had for an output a series of short pulses marking the leading and lagging edge of each input width-modulated pulse, as shown in Figure 10. Thus, he obtained a chain of time-varying pulses. At the receiver, the pulse-time modulated chiin was converted to the corresponding chain of width-varying pulses by means of -16- Figure 10. Kell s Conversion from Width- Varying to Time-Varying Pulses, a free-running multivibrator. This multivibrator was synchronized by the time-varying pulses in such a manner as to produce the desired width-modulated train at the output. 4. The Cathode-RZa-Tube Modulators. An entirely different method of modulation was devised by R. E. Shelby in 1937, and later modified by by W. T. Beatty. The heart of this method was a cathoderay tube with the normal electron gun and electrostatic deflection plate assembly but with a specially-shaped anode. The output of the modulator was taken from this anode. Pulses were generated between the anode and ground by the electron beam which was caused to sweep repeatedly at a constant rate over the anode. The section of the anode swept over by the beam was controlled by the modulating signal which was coupled to the -17- deflection plates. Thus, the width or position of the output pulses, depending on the design of the anode, could be made a linear function of the signal amplitude. A typical modulator of this type had for an anode a plate, triangular in shape, in a plane normal to the axis of the tube, as in Figure 11. The electron beam was swept across the triangle by sweep voltages Figure 11. Anode of Typical Beatty 'Modulator. applied to the hoizontal deflection plates. The signal varied the potentials of the vertical plates, thereby moving the sweep in a vertical direction. The widths of the output pulses depended upon the section of the anode swept over and, therefore, on the signal voltage. Since modulation by means of specially- designed tubes was beyond the scope of this thesis, the brief mention above of the cathode-ray-tube modulators should suffice. For further information on these modulators, as well as on the history of pulse modulation in general, the reader is referred to the Electrical Engineering Department seminat paper by the author. -18- f. Modulators for Multiplex Systems. Probably the most practical application of pulse modulation is in the field of multiplex communication. Several pulse systems providing for the transmission of a group of communication-channels using a single transmitter were devised during the war years. The most important of these were the AJ/TRC-5 and AN/TRC-6. These systems are basically very similar. The pulse chain which keys the radio--frequency transmitter is composed of a train of fixed marker pulses, as illustrated in Figure 12, each followed by timemodulated channel pulses, one for each channel to be transmitted. There is a modulator for each channel which varies the position of the proper channel pulse with relation to the common marker pulse. As.I M ARKEA Figure 12. PULSE !. 7/t CHANNEL PVLFiS" AN/TRC-6 Multiplex Pulse Chain. The pulse chain in Figure 12 Jir labelled to indicate the time relationships existing for the AN/TRC-6. One hundred percent modulation of one channel p -.19- causes the corresponding channel pulse to shift by six microseconds to either side of its unmodulated position, a total pulse displacement of 12 microseconds. Therefore, rith a maximum positive displacement of one pulse and a maximum negative displacement of the next adjacent pulse, the time interval between the trailing edge of the former and the leading edge of the latter is about 2 microseconds. This interval is called the safety, or guard, interval, and is necessary to reduce interference between channels. Unfortunately, no detailed information of an unclassified nature has been published concerning the modulation methods employed in these systems. Although a brief account of the modulator used in the AT/TRC-6 is contained in one of the available references, the writer has been informed that this account is inaccurate, and that actual modulator is still under military classification. Assuming that width-modulated pulses are produced by the modulator and that these pulses are, in turn, converted to corresponding time-varying pulses, it can be seen that 100 percent modulation of the time-varying pulse would not necessarily * According to information received from Prof. S. T. Martin, Mass. Inst. of Tech. ] -20- correspond to 100 percent modulation of the widthvarying pulse. For example, consider the second channel pulse in the chain shown in Figure 12. If this pulse is converted from a width-modulated pulse whose leading edge is fixed and occurs at the instant of the marker pulse, the index of modulation of the widthvarying pulse necessary to give 100 percent modulation of the time-varying pulse would be: (dmax.-d'min./ dmax.+d'min. ) x 12/54 . 0.22. On the other hand, a timing system might conceivably be used which would generate the leading edge of the width-modulated pulse at some fixed time after the marker pulse. Thus, no fixed relation can be derived between the modulation index of the width-modulation and the percentage modulation of the time-modulation without detailed knowledge of the system under consideration. C. Nature of Problem Under Consideration. In the published work on pulse modulation there has been little information on the design and operation of pulse modulators from the standpoint of harmonic distortion. Theoretically, as shown by the expression for the modulated pulse chain, any modulation method which produces a train in which the widths of the pulses (or the positions of the -21- pulses in a corresponding time-modulated chain) depend upon the signal amplitude at the instants of the leading and lagging edges will not generate harmonics of the signal frequency. Such a method should be superior to a system in which signal harmonics are inherent, as in a modulator of the "Mode 1" type. However, it might be that, by reason of increased stability or simplicity of design, the second system would be preferable to the first. Also, although the mathematics might indicate no harmonic distortion, the modulator itself might operate non-linearly in such a way as to introduce the undesired harmonics. It was decided to investigate the behavior of several modulators, all employing standard vacuum tubes, in regards to the generation of harmonics of the modulating signal as a function of the modulating frequency and the modulation index. Moreover, some information as to the stability of the modulators in the face of changes in vacuum-tube characteristics would be derived. Unfortunately, time did not permit as complete an invstigation as originally envisioned. possible, however, It was to take sufficient data on the modulators considered so that conclusions could be -22- drawn as to the comparitive merits of each. The investigation was largely experimental since, except in the case of the basic modes of operation as discussed in the first part 6f this chapter, the mathematical analysis was too lengthy to warrant its inclusion in the relatively short time allowed for this thesis research. All the modulators designed had width-modulated pulse chains as outputs. Any one of them, however, could be used as pulse-time modulators by the addition of a differentiating network and negative-pulse inverter at the output. Since, in any pulse system, a timemodulated train is converted to the corresponding width-modulated train before demodulation, it is un- necessary to investigate the time-modulated train in order to draw conclusions as to the quality of modulation. Thus, in general, Pny conclusions as to a given modulator will hold whether it is to, be used in a pulse-width or pulse-time application. Any distortion which appears as a result of converting the width-modulated train to a pulse-time train is not a function of the basic modulation. On the other hand, inwlight of the preceding discussion of multiplex pulse chains, a modulator which generated low distortion for small percentage modulation but high distortion for larger percentages .23- might be used to advantage in a pulse-time system. In such a system a small variation in pulse width would give a 100 percent of the final time-varying pulse. Of course, the above paragraph would not apply for modulators which varied the pulse positions directly, without first generating a width-modulated chain. However, the writer was unable to design or find any account in published literature of such a modulator other than the specially-designed cathode-ray tube. This is not meant to imply, of course, that such a system is impossible. -24- CHAPTER II THE MIXER-TYPE MODULATORS A. The Kell-Kretzmer Circuit. The first modulator to be constructed was very similar to that patented by R. D. Kell. The circuit was taken, 2 in large measure, from work of E. R. Kretzmer. In operation, the modulatbr mixed the modulating signal with a triangular waveform. This summation was amplified and clipped to produce at the output a chain of width-varying rectangular pulses. A repetition frequency of 20,000 cps. was arbitrarily selected for the triangular wave. As stated above, it has been shown that the interference frequencies, i.e., the sum and difference of multiples of pulse repetition frequency and signal frequency, will be small for practical purposes if the ratio of pulse frequency to signal frequency is 2.5 or above. Thus, a pulse frequency of 20,000 cps. will allow a maximum signal frequency of 8,000 cps. This maximum is considerably higher than would be necessary in a system to be used solely for communications purposes. However, if a degree of fidelity were desired, as in the transmission of music, a maximum signal frequency of 8,000 cps. is none too high. -25- 1. Kell-Kretzmer Circuit Design. The circuit diagram of the Kell-Kretzmer modulator is shown in Figure 13. A square-wave with a pulse repetition rate of 20,000 cps. was passed through a resistor-capacitor integrator with a timeconstant of 100 microseconds. Thus, the voltage waveform across the condenser C, was triangular in shape. The familiar equation for the instantaneous potential across the capacitance for a step-voltage )E, where E is the paplitude of the input is e= (1- e step input, T is the time-constant of the integrator, and t is the time between the step and the instant of measurement of e. The amplitude of the triangular wave, then, was e = (l- e OlK 6SN7 )E 0.22E. 1N 0 -EOeurr .C 300v AfVl -** 5sNAL '/SOV Figure 13. i' I~ a - -~~~~2 The Kell-Kretzmer Modulator. This triangular waveform was amplified by the triode V, . This tube was biased by the action of C,,and R,. C tended to charge to the peak of the triangular -26- wave so that grid current was drawn only at the instants of the positive peaks of this wave, A cathodefollower VA coupled the triode V, to the mixer. As in the Kell modulator, the triangular wave was impressed on the grid of the modulator V,, and the signal was applied through a coupling transformer to the cathode. This tube, without input voltages, was biased to the cut-off point by a variable-bias source. Thus, the mixer also'clipped the negative swings of the summation of signal and triangular wave. The amplitude of the triangular wave was adjusted so that no grid current would be drawn for 100 percent modulation. Overdriven amplifiers V. and V,, similar to Kell's, completed the clipping prooess. In order to prevent a variation of bias at low frequencies, fixed biasing was employed. Especial care was taken to reduce power-frequency iterfetence. Signal-carrying leads were carefully shielded, and filament leads were well segregated. The filaments were held at a d-c voltage of +20 volts with respect to the cathodes to eliminate emission between heaters and cathodes. 2. Experimental Procedure. A block diagram of the test set-up is shown in Figure 14. A General Radio Square-Wave Generator -27- 154U~AM-WANg staivAl. GEN6EATOR Low-PASS /a3AN- PASS CArmoos-AAY WAvE OsCILLosc.,a ANAY.zaeR Figure 14. Block Diagram of Experimental Set-up. Mpdel 769 supplied & square wave to the integrating circuit in the modulator. The modulating signal was obtained from a Hewlett-Packard Audio Oscillator, Model 200b. Since second and third harmonics of the order of 0.3 and 0.1 percent of the fundamental,.-respectively, appeared at the output of the audio oscillator, a band-pass or low-pass filter was inserted between the oscillator and the modulator. Thus, the signal which was applied to the modulator had a harmonic distortion of less than 0.02 percent of the fundamental component. The pulse-chain output of the modulator was fed to a load consisting of either a low-pass filter if the signal was to be recovered, or aeresistinde box. A Dumont Cathode-Ray Oscilloscope, Model 208, .28- was used to view the waveforms at various points throughout the circuit. The sweep of the oscilloscope was synchronized with a voltage either from the squarewave generator or from the signal source, depending upon the waveform of interest. To measure the amplitudes of the harmonic components in the modulated pulse chain, a General Radio Wave Analyzer, Model 736-A, was employed. The fact that harmonics were present in the output of the audio oscillator proved fortunate. The wave analyzer was connected to the oscillator, and was tuned to the approximate frequency of the desired harmonic. The exact setting of the analyzer could be obtained by tuning for maximum swing caused by the harmonic. Thus, even though small errors in calibration existed either in the audio oscillator or in the analyzer, it was still possible to adjust the desired frequency quickly and easily. After the correct dial settings for the second, third, and fourth harmonics had been obtained, the analyzer was connected to the output of the modulator for actual measurement. By carrying out this system of comparative calibration at frequent intervals, one could be sure that measurements were being made of the harmonic components and not of extraneous -29- interference voltages. Since the harmonic distortion of modulation as a function of frequency was to be investigated, various frequencies of the modulating signal were selected. In order to permit valid measurements it was necessary to choose these frequencies carefully so that no harmonic would have the same frequency as any of the larger (np mq) components. As the pulse repetition frequency was 20,000 cps., such.signal'frequenties.las 100, 200, 400, 500, 800, 1000, 1200, 1500, 2000, 4000, 5000, and 8000 cps. were avoided. 3. Experimental Results. Amplitudes of the fundamental and all measurable harmonics of the signal frequency as a function of the modulation index, (cmax.- dmin.)/(d'iax.+d'min.), were measured at the output of the mixer tube and also at the output of the overall modulator at three signal frequencies, 355, 1170, and 3550 cps. Waveforms, shown in Figure 15, were observed at various points in the circuit to check, at intervals, on the operation of the modulator. Care was taken to keep the mixer-grid bias at the cut-off point, thus assuring an output wave in which the unmodulated pulse widths were equal to one-half the pulse repetition period. No Modulation 50 % Modulation Output of Mixer Tube Output of Clippers Figure 15. Waveforms of Kell-Kretzmer Modulator. The results of the measurements are represented on Graph I. It was found that the harmonic content for any value of modulation was independent of the signal frequency. Therefore, only those measurements taken for a signal frequency of 3550 cps. are shown. Obviously, the performance of this modulator is not of very high quality. If the modulator were to be used in a system where the total allowable harmonic distortion was to be below three percent, the modulation index could never be permitted to exceed about 0.5, or 50 percent modulation. It is of interest that, although the second-harmonic distortion at the output of the mixer reaches a maximum of 10.2 percent of the fundamental, the second-harmonic component at the output of the entire system has a maximum of only about 311 percent. The action of the (IL 0 4 0i.4 ri F-- KI 0.- -7vANMWVO~~nas Q0 4-4vq.N::wC -32- clipper-amplifiers is, apparently, to decrease the second-harmonic distortion and increase the distortion caused by the third harmonic of the signal frequency. It is impossible to judge from the data in GraphiI .the degree of distortion which can be credited to the clipper-amplifiers. Apparently: though, the major source of the distortion lies in the process of mixing the modulating signal with the triangular wav. This distortion might be a result of non-lineprity of the sides of the triangular w.ave or non-linearity in the operation of the mixer tube. It will be remembered that the mixing triode was biased at cut-off, with no input signal, and was driven by the modulating signal and triangular wave to a point just below that at which grid current would be drawn. Because of this, the operating point traveled over a considerable portion of the plate voltage-plate current characteristics of the tube, from cut-off to e,= 0. For this reason, it was believed that the major part of the distortion was caused by the non-linearity of the mixer tube. It was decided to design a modulator in which the mixer tube would be driven over only a small and linear portion of its characteristics. Also, a more effective clipping system would be devised to replace the tetrode clippers -33- of the Kell-Kretzmer circuit. B. An Original Modulator. 1. The Circuit Design. A circuit diagram of the resultant original modulator is shown in Figure 16. The triangular-wave generator was identical to that used in the KellKretzmer modulator except for an increased time-constant of the integrating circuit. This increase insured greater linearity of the sides of the triangular waveform. This wave and the signal were mixed in the triode amplifier V, , at a very low amplitude level. The low-level mixing, together with the feedback caused by the un-bypassed cathode resistor, promised good linearity in the mixing process. The grid-toplate amplification of this stage was approximately nine. ISOIC 465mr- Figure 16. Sty TSIC 6* AMC Circuit Diagram of Original Modulator. Further amplification was secured by the use of a second triode V2, also with degenerative feed- -34- back. The feedback was made greater in this second stage than in the first since the input signal was of greater amplitude. As a result, the gain of this stage was only 2.5. Fixed grid bias was employed for both . V, and V2 Double-diodes provided the required clipping at the output of V,. In the absence of a modulating signal, the input to V3 was a triangular wave with a peak-to-peak amplitude of about 55 volts, The conducting plate-cathode resistance of each diode section was approximately 1500 ohms. Thus, the waveform at the output of V3 appeared as in Figure 17. It will be noted that Q.3 volt peaks remain on the somewhat-rectangular pulses at the output of V3 . These are caused by the fact that the conducting resistance of the diodes is not negligible. It is true that the magnitudes of these peaks could have been reduced either by decreasing the amplitude of the input triangular wave or by increasing the input series resistor R,. However, if the amplitude of the input wave wexe decreased, the slopes of the leading and lagging edges of the output pulses would be less. An increase in the value of R1 would also decrease these slopes. The wiring and tube capac- M -'35- Output of First Double-Diode. Output of Second Double-Diode. Figure 17. Waveforms in Clipping Circuit with No Modulation. itances across the input of the diode must be charged through R,. Thus, an increase in R, will increase the charging time-constant, effecting a decrease in the slope of the wave across the diodes. To remove the peaks from the waveform at the output of the first double-diode V3 , a triode amplifier and a second double-diode were employed. The output of this second clipper V., shown also in Figure 17, was the final rectangular pulse chain. The variablebiasing circuit at the input of Vs made it possible to adjust the unmodulated widths of the output pulses. Plainly, this modulator is basically very similar to the Kell-Kretzmer system. The major differences lie in the low-level mixer and the doublediode clipping circuit. .36- The modulator was tested in exactly the same manner as that used for the Kell-Kretzmer circuit. Harmonic components were measured at various points in the circuit as a function of the modulation index. It was found that the amplitudes of these components were again independant of signal frequency so that only the results obtained f or a single signal frequency need be recorded. 2. Experimental Results. The resulting data are plotted in Graph II. It is obvious that this modulator shows a considerable improvement over the Kell-Kretzmer system in regards to harmonic generation. The percentage of the second- harmonic component for the overall circuit reaches a maximum of 0.34 percent at a modulation index of 0.9. The percentage of third-harmonic distortion remains below 0,02 percent for all values of modulation index. Apparently, each stage, as well as the signal transformer, adds a small amount of harmonic distortion. In hopes of eliminating some of this distortion, a direct coupling to the cathode of the mixer was substituted for the transformer. An inpe dance -matching cathodd-follower was placed be tween the signal filter and the mixer. No noticeable ~~~~1~~ 7 4 :t) GRAPH II Harmonic Distortion in the Original Mixer,'Type Modulator JO Ue surable harmonic components are plotted as a function of modulation index. Signal frequency in all cases was 550 cps., and the pulse frequeney, 20,0YO eps. X P ~~14 Seth H. Washburn April 5, 1947 4R d a 0 I Um 'IN 1d 'C 4 ~,t9 . . . - 113 I t!L ~Io wvanWirdr a iii' improvement resulted as the cathode-follower generated about the same values of second-harmonic component as had the original transformer. The only third-harmonic component to be plotted in Graph II was that at the modulator output. This component was so small that it was difficult to measure with accuracy. The fourth .signal harmonic was too weak to record. Unlike the -clipping circuit in the Kell-Kretzmer modulator, the double-diode circuit had no effect on the harmonic content of the wave passed through it. Thus, harmonic percentages measured at the output of the modulator were essentially equal to those at the input of the clippers. C. Stability of Mixer-Type Modulators. As has been seen, the operation of the mixer- type modulators is basically as follows. A triangular wave is passed through a clipping circuit. The modulating signal effectively varies the points on the triangular wave where clipping occurs. Because of this, any change in amplification factor /A or dynamic plate resistance ." , occuring in any tube passing the triangular wave, which varies these clipping points will affect the widths of the pulses in the output chain. .39. AM AAJA Figure 18. Degenerative-Feedback Amplifier. r Figure 19. Output of Amplifier, showing Effect of Amplitude Change on Pulse Width. Consider a single triode with cathode feedback, as shown in Figure 18. The output of this tube is a triangular waveform of peak-to-peak voltage amplitude 2E and period 2T, as in Figure 19. The dotted line in the figure indicates the clipping point, the point at which the clippers in the output of the modulator operate. Thus, the widths of the unmodulated pulses at the output will be d. Let d/T equal a. Referring again to Figure 19, it seen that, by simple geometry, (d/T) = can be' (aE/E). Suppose that some change in tube characteristics -40- causes a change AE in the output voltage amplitude. This, in turn,- will change the output-pulse duration by 4d. Again by geometry, d+d AE + aE AE+E T T Then, Ad AE aE T AE+E a and, Ad: AE(1- a) T +dd T .pa T AE-aE 4E+E 4_d= 4E (1 -al d 4E-+E T d AE (1 E+6E a ) The per-unit change in d will then be: If the change in E, CE, is very small compared with E, d (1) ) .d a E This equation gives the per-unit change in pulse duration with a per-unit change in the amplitude of the output triangular wave. The next step is to discover the effect of changes of tube constants on this output amplitude. For the amplifier in Figure 18, tA= RL+r,+ +(,"l)RoK where A is the grid-to-plate gain. p is the amplification factor. re is the dynamic plate resistance. -41- We are concerned with changes in -AI with changes in rp and A. From elementary calculus, d JAI x (AM/-r,)dr,-+ (JiAI/J ) dA. Thus, the chrnge in JAI will equal the change in JAI due to a variation in rp with A constant plus that due to a variation in ,& with rp held constant, if all changes are very small. a variation in rp. Consider first -.A R &. .+ rM+ (.+1) R4 alAl j r, The per-unit change in JAI will be: IAI JAI B *,r,+ (C+l)P IAI RrO+(/^+l)R_ 6 rp R+r,+ (+1R .Ar F,= (Jsj \ jre- Considering a change in.p, 1 -R&Ric _AIA R. (+1)R* R,.+ r,+R JAI JAIL . Then, A (1-IA1 Now suppose that the input triangular .wave has a peak-to-peek amplitude of 2e. Then E -. AI e , and (aEA)a (A1A$A). 4E=ALAI e . Thus, Substituting equations.,( 2 ) and (3) into (1); for a variation in r,: j.d, d -. I' .A r " R. ) rp-4) rp (2) 42-. and for a variation in A: d d ~)a/ R&/ a (1- - /At (5) The factor a will remain constant as long as the triangular wave is unmodulated. If the unmodulated output-pulse width is one-half the period of the triangular wave, a will equal unity and d will be unaffected by changes in r, andp. However, when the signal is introduced, the value of a will change for each succeeding cycle of the triangular wave. The stability expressions show that, as a decreases, the greater will be the effect of a change in ^ or in r,.on the output pulse. Thus, it appears that a greater degree of stability exists for a condition of small percentage modulation and an unmodulated pulse width of one-half the pulse repetition period. Considering the mixer tube V in the original modulator circuit, shown in Figure 16, where R,=100k, Rk= 5ktA. 2O, and r, 10k, the per-unit stability values are f ound to be: d and . - -. a .~ (1 - )rp a) +0.535) - an -6A .43- CHAPTER III THE MULTIVIBRATOR MODULATORS Although the second mixer-type modulator tested was most encouraging in its very small harmonic generation, it was thought that some other circuit, of simpler design, might also give good results. The cathode-coupled multivibrator suggested itself as a basic circuit from which to work, It was hoped that- a modulator with fewer tubes and of the same order of stability and distortion as the mixer-type of modulator might result. As shall be seen, this hope was not realized, at least in regards to stability and harmonic distortion. The cathode-coupled multivibrator, as illustrated in Figure 20, operates in a manner very similar to that of the more common plate-coupled multivibrator. Consider the circuit at rest. Since the grid of VA is returned to Ebb through Rga, Va draws a relatively high value of plate current. For this reason, the- potential drop across Rk is large. The grid-cathode potential for Va is about zero volts since RgA has a high resistance, of the order of one megohm. If suitable values for R, and RA have been chosen, the grid-to-ground potential of V, is low -44- V RZ Figure 20. R RA R, R~ 2 Ric Cathode-Coupled Multivibrator. compared with the cathode-to-ground voltage el,, V, and is cut-off. These steady-state conditions obtain until the circuit is triggered. A positive pulse of amplitude sufficient to cut-on tube V, is applied to the grid of V, . As this pulse starts the conduction of V, , the plate of VA falls in potential, and, since the voltage across the condenser CA cannot change instantaneously, the grid of V 2 also drops. The decrease in grid-toground voltage for V. causes the cathode voltage to decrease, which, in turn, increases the drop in vol- tage on the plate of V,. This process, an almost instantaneous one, continues until the grid-cathode voltage on V, has fallen considerably below cut-off. The plate and cathode voltages on V, will depend upon the plate current drawn by V,. CA discharges through Rga and V., and, as it -45- discharges, the grid-to-ground potential increases. When the grid of V, reaches cut-off, V, will start to draw current, the cathode will rise in potential, and the plate of V, will also rise, thus cutting off V,. The circuit returns to the steady-state donditions, CA quickly recharging through R. 5. and the grid of VA. The multivibrator remains in this steady-state condition.until the next triggering pulse. The duration of the multivibrator output pulse, or gate, t depends upon the discharge time-constant of CA in series with Rga, the plate resistance of V,, and B.c* If Rga is large compared with the other resistances, this time-constant is approximately equal to C,,XRgz. The duration of the output gate also depends on the initial value of the exponential in the en, waveform, shown in Figure 21, and also upon the value of the same exponential when it reaches the cut- off point for V., If both these values are measured in volts deviation from the steady-state value of Ebb, the gate time can be expressed as t T mi( Initial Value LTJlnL Final Value where Toes is the time-constant C,RgA. From the waveforms in Figure 21, it can be seen that the initial value of the exponential is ..46- Mtv"Aih Fwr.h. Ebb1 eka ecna, Plate-Ground VoltageV, Grid-Ground Voltage, V oOv Cathode-Ground Voltage Plate-Ground Voltage, V& Waveforms for Cathode-Coupled Multivibrator. (Trigger at t, ) Figure 21. (Ebb+ A ebn - ekz) , and the final Value equals (Ebb+ jEcoi-ek, ). EcoI is the grid-cathode cut-off voltage for VA at the end of the gate. ST t aTagIn Therefore, Ebb+ebn Ebb-JE c osj -eks - ek, The initial value of ecn, is a linear function of the plate current of V, during the gate, since Aebn varies linearly with the plate current of V, . Also, since IEcoI and ek, vary linearly with the plate current of V, , the final value of the exponential will also be a direct function of this current. The plate current taries directly with Ece.,, over the linear portion of the tube's characteristics, and, thus, the initial and final values vary linearly with E... If the average current through Q during the gate pulse remains constant for various pulse widths, the pulse width will vary as a direct function of E.,. Since the current through C equals approximately (Ebb -ecn,)Rga, this requirement merely states that the average.e.nmust be a constant. Thus, (Ebb-t Aebn -eka) + (Ebb+ IEco.1 -ek,) K For a change in E,,, Ebb and eka are constants. Since Aebn equals Ebb- ebn,, the above can be simK,. Therefore, plified to (-ebn, +)EcoA-ek,) L0(-ebn,+EcoI-ek, ) = (2) It is obvious that Aebn,/E,, and Aek,/E, are merely the grid-to-plate and grid-to-cathode gains, respectively; and that Eco, is a linear function of ek,, Then, Aebn= -APAEC. A ek Aac4E.. -bAtEa., AEcoz Substituting these identities into equation (2), Ap -bAk-Ak 0 or A,- (1 *b)Af (3) For a triode with unbiased cathode, R .. r, + R,,+ " Ap A,= R, . Ra.n ,"Rj Ak AR = + r, -*R,, 5+1 Substituting into (3), R1. (1 + b)R Iwhere b =IE(.aLOEb (4) S Ath /tek Since the method used to derive this relation -48. is not rigorous, the equation (4) can be regarded as only approximate. However, it can be seen that it is possible to cause the gate width to vary directly with the control voltage Eg,,. A. The First Multivibrator Modulator, The brief analysis above suggests that it might be possible to couple the modulating signal to the grid of V, in the multivibrator, and make the output-gate width vary as a function of the signal. A circuit, shown in Figure 22, incorporating the cathode-coupled multivibrator was designed and constructed, 1. Circuit Design. Examination of the plate current-plate voltage characteristics of the 6SN7 double-triode show that Eco/ Eb equals 0.06. Substituting this value into equation (4) above, it is found that R, should be 1.06 times the value of Rx. By cut-and-try methods of assuming values of Reand RK and computing the pulse widths as a function of E4.,, it was found that for Ri=7.5k and Rs,322.5k, the pulse width varied 3.2 microseconds for a change in 1 volt in E,... This assumes a CaR3L time-constant of 500 microseconds, a high value calculated to make the discharge exponential of e..a essentially linear. -. 49- Voo toou 5#7 a First Multivibrator Modulator, For higher values of B.xand lower values of Rj , Figure 22. sap, 7.5 22. this sensitivity was increased. Since the increased sensitivity did not seem desirable from the standpoint of stability, the values of Rm= 7,5k and R, L422.5k were employed. The cut-and-try methods also showed that if a variation of 0 to 50 microseconds in pulse width were desired, Econ should be variable from 25 to 50 volts. The 25k resistor in the grid circuit of V, served to decouple the signal and the trigger inputs. The method of testing this modulator was essentially the same as that used for previous circuits. The multivibrator was triggered at a rate of 10,000cps. This trigger was obtained by differentiating the output of the square-wave generator and then clipping the resultant negative pulses with a diode limiter. The harmonic content of the width-modulated pulse chain at the output of the multivibrator was meas- -50- ured by the General Radio Wave Analyzer. Again, signal frequencies of 355, 1170, and 3550 cps. were used to investigate the effect of signal frequency on modulation linearity. Since the output of the multivibrator consisted of pulses the leading edges of which were fixed in time, it was convenient to measure the percentage modulation in terms of the number of microseconds the lagging edges varied from their unmodulated positions. Thus, the harmonic content was recorded as a function of the amount of this deviation. These measurements were made at three values of unmodulated-pulse width: 10, 17, and 20 microseconds. 2. Experimental Results. The results of the measurements taken with the multivibrator modulator are illustrated in Graphs III, IV, and V. Certain features are apparent at once. First, the harmonic distortion increases as the signal frequency increases. Then, the harmonic distortion is greater for larger values of the unmodulated pulse widths. The amount of distortion also appears to be greater than would be expected merely from the slight non-linearity of the exponential discharge curve of egm. gild. I ! ! -. I. ~i bfbi~i ~~ r f I 7 - 1___ 71--*i I- A L-l Jo~th H1 T r -1 7- L 1L7 O Hi-,j -i- -T - J- fl74 A.71 thHW~j~ ' L~i~ T,~ I J10-l so th 1-- -. --- I. 1 -J T T*J J~ 7-1 -I I --[,L -- T L T -11 - -I I-: 7 . 4 - -I I1J FER1 ~~1 I- i r~l ar F . I- FIL - I J t - I~~ I . sP41 i0 fAU 7_ --- Ott4 _ _2 rA .L. i a -I *icy F____ T_ _3-1 t -- Tt . 7. I -F F I. I. 1- I I .1 1~* .. ---- II. r 7 TTF77 I .1 ~I. i*~} -i ~7T177~ LJLtL' I hL711-1I Aj: 7ff -- ii ~~V-- {.Ui K V~I7 - ------ 4- j .11 I 77 I- T - ~t -; ~ K H ~-4-~J I __ .P J2.4 -~ -- 44~4W ___Y~Fh~+'~4~w I. I. I -I I lUll - 77 J K _Lj if ~ _77~~ F 4 -54- A closer examination of the modulation process may indicate an explanation of these results. It has been shown on preceding pages that the pulse width varies almost linearly with the control voltage Eco. However, this dependence assumed that Ecoswas a d-c, or very slowly varying, voltage. If the control voltage varies rapidly, the value of Econat the end of the pulse will not be the same as that at the beginning of the pulse. The pulse duration depends upon the initial value of the exponential curve of e",.and its value at the end of the pulse. These values are, respectively, (Ebb-+ ,ebn - ek,) and (Ebb-tlEco-ek,). Aebn depends on the control voltage at the leading ledge of the pulse, and JEco4. and ek, are functions of E...at the end of the pulse. Thus, the pulse width depends upon the signal voltage at the instant of the pulse leading edge, a fixed instant, and also upon the signal voltage at the instant of the pulse lagging edge, a varying instant. It was seen in Chapter I that if the pulse width varied as a, function of the signal at the instants of the varying edges themselves, "Mode 2" modulation, there was no harmonic distortion inherent .55- in the modulation system. On the other hand, if the pulse duration depended on the signal at fixed instants, as in "Mode 1" operation, harmonics of the signal frequency would appear in the modulated pulse train. The multivibrator modulator, therefore, appears to incorporate both modes of operation. Thus, it should reasonably follow thpt harmonic distortion is inherent in the system. On page 7, it has been seen that, for "Mode 1" this distortion increases as the signal frequency increases, as the unmodulated pulse width increases, and also as the modulation index, or pulse-edge deviation increases. Graphs III, IV, and V show that the harmonic distortion of the multivibrator modulator behaves similarly. It is, therefore, believed that the major portion of the observed distortion is inherent in the system. Unfortunately, a mathematical analysis of the system to check this belief could not be made in the time available. In light of the large distortion pro- duced by the modulator, it is doubtful whether such an analysis would be, practically, of sufficient value to justify the large amount of time necessary to carry out the analysis. -56- B. .. A Second Multivibrator Modulator, Design of the Circuit, A modulator employing a slightly different method of signal injection was designed, in which the signal varied the potential of the control grid'of V., as in Figure 23. However, this circuit proved to be subject to the same inherent non-linearity of modulation as that found in the first modulator. Again the pulse duration depended upon the signal voltage at the beginning of the pulse, since e ra was a function of the new E.. Also, the duration was a function of the signal at the pulse end, since the signal varied the instant at which the grid of Va reached cut-off at the conclusion of the exponential. 300v Sk 22.5 .Ij1 65s7r 2 V, Itu szp ; Figure 23. aS fE7 20 k 7.r VC a. Tiet Second Multivibrator Modulator. Since the signal varied the initial and final -57- values of the discharge exponential by like amounts (assuming a signal of low amplitude and frequency) for any circuit-component values, no analysis similar to that on page 4 7 was necessary. By cut-and-try experimental methods it was found that the 2 megohm resistor in the signal input circuit was sufficient to decouple the signal source from the modulator. 2, Experimental Results. The harmonic components present in the output wave, shown in Graphs VI, VII, and VIII, were obtained in exactly the same manner as that used for the first multivibrator modulator. Again, it is seen that the distortion increases as a function of the signal frequency, the amount of modulation, and the unmodu-. lated pulse width. This was to be expected as a result of the inherent distortion present in a modulator, the output pulses of which have a width partially dependent upon the signal at fixed instants. The distortion for a given pulse width, frequency, and lagging-edge displacement is slightly higher for the second harmonic than it was for the first. This may be due to an increased dependence, in the second modulator, on the signal amplitude at the fixed instants of the output pulse leading edges. 013 X I *M W4.4W ow ~ j I0- 4 -S K- *1gn.~ re~ OW 1--I --- *Apr'i1 -7- 4 -It II r I I. 4-- 7 .- L i. F, {i TiI I -I Tii1 j _ :_-$ 7U'j- '' .r T H - -7 ~1 ! 'L L - 1~ 7- 777 LL: I 77 I--. rt S1~n r i Apr-Lt 17,i 14Ltd K 7 !-. 71 L LL -I.-.; __ -7- AL- A ....... I___4_ -i -1 K" Ti 77=- - Fry I TT-- !14.. I. It' KIN 7 ~*-v--r-. t- Ia -~ -1 7 ' -~ VLEV-~ 'On! *i SJlb grwe Frqtey i 000~s 2 %, ne -I A I I. I', - J w i 'IL 1 4.. 1ILI F-T 1 Ii. - .- - .1 T7 V f77Y77.-IF% 17 777]TTC.. 7:4 - 2 7 flI 17 I.- Ki 77It7f T tili ~rLLA~rr-- ____ -61- In this second moldulator, the value of ex varies in a one-to-one ratio with the signal voltage. Also, there is some signal leak through the 500 micro-microfarad discharge capacitor causing a very small variation of debn. The initial value of the exponential rise of een depends upon both of these factors. In the first modulator, debn varied as the signal but not to as great an extent as ek, in the second. For this reason, it appears that the second modulator might be subject to greater inherent distortion than was the first. It is believed that if the mathematical analysis of a system in which the pulse width depended on the signal at both fixed and varying instants were carried out, the above conclusion would be found justified. As stated above, however, such an analysis might possibly have little practical value. C. Stab of the Multivibrator Modulators. A study of the stability of the stability of the multivibrator modulators resolves itself into a consideration of how changes in tube constants will affect the duration of the output gate. Thus, it is only necessary to analyze the basic cathode-coupled multivibrator in regards to changes in the amplification factor. and the plate resistance r, for -62- each of thetto triodes in the circuit. The basic equation for the duration of the output pulse, as has been seen, is: t T, Ebb-4ebn-ea ln (1) (Ebb-# Mcftl - eu, From the waveforms on page46, it is obvious that Aebn a Ebb - Ebn, . Now Ebn, , ex, , and e ft are frac- tions of Ebb as follows: Ebn, = Ebb ek, = Ebb eka = Ebb (R6,,+Rk) , where tubes V, r, Rk , f and Y 4- Rk Rk are the static plate resistances of and VA, during and before the pulse, res- pectively. Also, let ,&mEbb/E,,. 1. Substituting these relationships intb the basic equation, t -Tasn (1 n1Fp*Rk) - Rk/(Ra..+FrA+Rk) T Ru/(R+ +eR/&.+ - Rk/(Rb,-. Fp, -+Rk) (2) Now let (R,+ Y,,. Rk) R, and (R45 #f,-tFRk) = Rz. Substituting R, and Ra into equation 2, t : T 1n * 1+- R,/R, -_Rk/R.) 1- +4. - Rk/R, T. l R, RA+R&R&,-ORkR, =~~ TRlnI Ra Evidently, ( 1 -+ 1 /^) - RR' t depends upon the tube constants -63- ,, 7,, , and^. The effect of a change in each on the pulse duration t can be found by taking partial derivatives with respect to each in turn. it Ty, = To' Ta (l+ /.&&)Ra R, Rj, * R, Ra., -- RkO, = osR,+ (Rs RL.) /(R,7-Rk) e)) Rk(4 it ( 3) TDi- RA- .: J, R, /*& T,,L (1+1/.,)R,- Rk = T Fl + 1/^r) R, R, - RkRx T, 1 (1-tA.)R,-Rk -R, Rs, R 0( R- Rg R, - RkR, (1 +1/,6.)RRC-RkRL) ) t R, Ra /& (1+1/,)RR-R (R) T RC -(R.R ) /(R,+ R&,) From the appearance of equations (4) and (5), it is obvious that it is impossible to choose circuit constants such that either bt/laT., or Jt/fp, will be zero. However, the effectof changes in Fp, and F,, may be minimized by making R,,, R&, and Rk as small as possible, in comparison with f,, and i,,. 'Since rp, and f,, will themselves depend on Ra,, Rx, and Rk, only a succession of trial values substituted into the equations will indicate the quantitative effects of various values of the resistances. The plate resistances F,, and fp. are very susceptible to changes in Ebb. A change in plate-supply voltage will produce a variation in ,, and fp., the -64- magnitude of which will depend on the operating points of V, and V . However, a glance at the plate char- acteristic curves will indicate that a one percent variation in Ebb may cause an equally large percentage variation in the static plate resistances. For an idea of the magnitudes of Jt/J'1 ., Jt/dfa, and 4t/d^, the equations have been applied to the first multivibrator modulator. If the circuit is adjusted for an unmodulated pulse width of 12 microseconds, it may be determined from plate characteristics of the 6SN7 that f,, will be 43.3k and fp,, 10.0k. Substituting these values into equations (3), (4), and (5), together with the values of the circuit constants, it is found that t/JceaT,,(0.00385), Jt/df,.= T,,,(-0.OO476), and ut/dip~a the final percentage T,,,(0.005). From these, stability relations may be ob- tained: dt/t = 2.54e/,A/. , dt/t = -8.1d9,/!,, and dt/t 1.9 afp/r, Since Afp,/f,, and AiF,/rp. are of opposite signs, it might be possible to choose circuit constants in such a manner that variations of p. and fr., due to variations in Ebb, might be cancelled against each other. However, changes in either f, or F& not dependent upon Ebb would still be large. -65-. CHAPTER IV CONCLUSION AN INTERPRETATION OF RESULTS In the evaluation of the pulse modulators designed and tested, certain obvious conclusions may be drawn. First, it has been seen that a carefullydesigned mixer-type modulator introduces less harmonic distortion in the modulation process than does a modulator of the multivibrator type. This appears to be true, in general, for any index of modulation and signal-to-pulse frequency ratio. The mixer-type modulator could be so designed that the maximum total harmonic distortion present in the output pulse chain was less than 0.3 percent of the fundamental component. To obtain this low value of distortion, it was necessary to mix the triangular wave with the signal at a very low level and to amplify the summation with good linearity before clipping. Secondly, the mixer-type modulator is found to be inherently more stable than the mnltivibrator modulator for changes in vacuum-tube characteristics. This is apparent from the dependency of the multivibrator output on the static plate resistance of the tubes in the circuit. A change in the emission of W -66- of a vacuum-tube will cause a pronounced variation in the static plate resistance of that tube. In comparing the stability of the two types of modulators, it can be noted that any change in the operating point of a tube in the multivibrator circuit will affect the width of the output pulses. The stability of the mixer modulator depends primarily upon the stability of the dynamic plate resistance of circuit tubes. Thus, if the operating region is small, this region may be shifted an appreciable amount without much effect on the dynamic plate resistance, and, hence, on the output pulse, of the tube. The per-unit relationships derived for both types are also indicative of the comparitive stability of the two modulators. In favor of the multivibrator modulator, it is true that a circuit of the multivibrator type affords a considerable saving in the number of circuit components and tubes to be used. It is, apparently, the general consensus of opinion that the most practical application for pulse modulation lies in the field of multiplex transmission. A modulator, to be of use in a multiplex system, as described in Chapter I, must provide a pulse varying in time with respect to a fixed pulse. The multi- II -67- vibrator modulator would seem to be eminently qualified for application in this field, except for its high inherent distortion generation. However, the multivibrator modulator could be employed if size and weight requirements made the mixer-type modulator undesirable. Suppose that the maximum allowable harmonic distortion were to be limited to three percent. This distortion, in the multivibrator, depends upon three factors: the unmodulated pulse width, the amount of modulation, and the pulse-to-signal frequency ratio, As stated in Chapter I, the ratio of pulse repetition frequency to maximum signal frequency should be greater than 2.5 to 1. However, the use of much higher ratios increases the difficulty of amplifying the modulated pulse train at the 'receiver. Then, a ratio of about 3 to 1 is normally employed. Also, it would be desirable to allow as much pulse shift due to modulation as possible. Referring to Graph III, with these requirements in mind, it can be seen that, for an unmodulated pulse width of d, the maximum modulation or pulse displacement possible would be about 7 microseconds for da 20Asec., 9 microseconds for dx 17^sec., and 14 micro- -68- seconds for d: 10sec, The allowable pulse displacement is limited by the time between unmodulated channel pulses. in the final multiplex pulse chain. If this time were 14 microseconds, as in the AN/TRC-6, the maximum pulse shift due to modulation would be about t6 microseconds, or a total of 12 microseconds pulse displacement or deviation. Thus, the width of the unmodulated pulse at the output of each multivibrator modulator must be less than 10 microseconds. Of course, the mixer-type modulator may easily be converted for use with a multiplex system. Instead of employing a triangular wave, a wave consisting of saw-teeth, each with a vertical leading edge, as in Figure 24, could be mixed with the signal. The output of the modulator would then be a chain of widthvarying pulses, each having a fixed leading edge. A differentiating circuit, similar to that to be used with the multivibrator modulator, could supply the desired position-varying pulse. Here, the problem of distortion would not exist; the only consideration determining the width of the unmodulated pulse at the output of the mixer modulator would be that of stability. Full modulation of the multiplex channel pulse will correspond to some fraction of 100 percent mod- -69- Figure 24. Saw-Tooth Wave for Use with Multiplex Mixer-Type Modulator, ulation of the width-varying pulse at the output of the basic modulator. It is apparent that the smaller this fraction is, the larger will be the effect of a small change in the width of the width-modulated pulse on the percentage modulation of the channel pulse. Thus, for a small fraction, the final result of instability in the modulator will be great. It should be advantageaous then to make the modulation index for the width-modulator as great as possible. If the width of the unmodulated pulse was equal to one-half the total displacement of the position-varying channel pulse, 100 percent modulation of the channel pulse would correspond to a modulation index of unity for the width-modulator. However, a modulation index of unity cannot be obtained in actual practice. In Chapter I, it was shown that for good stability, the duration of the unmodulated pulses in the output of the mixer modulator should be one-half the duration of each mixer saw-tooth. It was also found -70- that a small percentage of modulation would be less affected by changes in tube characteristics than a large percentage. This last is in direct opposition to the conclusion reached in the last paragraph, in which it seemed that a large index of modulation was desirable. In the design of a practical system, it would be necessary to reconcile the two conclusions, taking a mean value for the percentage of modulation for the mixer-type modulator. The equations expressing the stability of the multivibrator modulators indicate that trouble would be experienced if the vacuum-tube characteristics or the plate-supply voltage had a tendency to drift. It would be necessary to make frequent checks and adjustments of.the modulators to compensate for these variations. This lack of good stability in the multivibrator modulator, as well as its inherent harmonic distortion leads to the final 'conclusion that, unless it is of primary importance to employ a modulator having small size and weight, the multivibrator does not lend itself to pulse-modulation applications. It is also concluded that, because of low distortion, good stability, and relative simplicity of design, the mixer-type modulator will give satisfactory results, whether it is to be used with a pulse-width or pulse- time modulation system. -72- BIBLIOGRAPHY 1. Standards on Transmitters and Receivers (New York: The Institute of Radio Engineers, 1938), 3. 2. E. R. Kretzmer, Fidelity in Pulse-Time Modulation, Thesis for M. S. Degree, M. I. T., (1946). 3. Fredendall, Schlesinger, and Schroeder, "Transmission of Television Sound on the Picture Carrier," Proc. 4. .RE., 34(1946), no 2, 49, R. A. Heising, "Transmission System," U.S. Patent 1,655,543, (1928). 5, J. L. Finch, "Signalling System," U.S. Patent 1,887,237, 6. (1932). R. D. Kell, "Signalling System," U.S. Patent 2,061,734, (1936). 7. R. E. Shelby, "Improvements in or relating to Cathode-ray Tubes," British Patent 493,010, (1938). 8. W. A. Beatty, "Pulse Modulation System," U.S. Patent 2,256,336, (1940), (Duplicate of British Patent 523,575). 9. W. A. Beatty, "Pulse Generating and Pulse Modulating System," U.S. Patent 2,265,337, 10. S. H. Washburn, The H (1940). of Pulse Modulation, E.E. Seminar, 11. I. T., (1947). -73- 11. T. Gootee, "Radio Relay Communications," Radio News, 35(1946), no 5, 16. 12. "Pulse Position Lrodulation Technic," Electronic Industries, 4(1945), no 12, 82. 13. D. L. Shapiro, Analysis of Diode Limiter, Thesis for B. S. Degree, V. I. T., (1941). 14. V. I. T. Radar School Staff, Principles of Radar (2nd ed,; New York: McGraw-Hill Book Company, Inc., 1946), 2-53 - 2-58,