ENGINEERING RESEARCH INSTITUTE THE UNIVERSITY OF MICHIGAN ANN ARBOR AN EVALUATION OF A SPECIAL BROADBAND RECEIVER INPUT CIRCUIT Technical Report No. 67 Department of Electrical Engineering Electronic Defense Group 3. F. ngon Approved by: % By: I D. bHafiburg A. B. Macnee Project 2262 TASK ORDER NO. EDG-1 CONTRACT NO. DA-36-039 sc-63203 SIGNAL CORPS, DEPARTMENT OF THE ARMY DEPARTMENT OF ARMY PROJECT NO. 3-99-04-042 SIGNAL CORPS PROJECT NO. 194B November 1956

TABLE OF CONTENTS Page LIST OF ILLUSTRATIONS iii ABSTRACT iv 1. INTRODUCTION 1 2. SUMMARY OF CRYSTAL THEORY 3 3. SOME ELEMENTARY NOISE CONSIDERATIONS 6 4. DESCRIPTION OF EXPERIMENTAL TECHNIQUES UTILIZED IN THE INVESTIGATION 11 5- SYNTHESIZED SYSTEM MEASUREMENTS 14 5.1 Basic Setup 14 5.2 Basic Procedure 16 5.3 Results of Synthesized System Measurements 20 6. BRETT CIRCUIT MEASUREMENTS 23 6.1 Setup 23 6.2 Procedure 25 6.3 Results of Brett Circuit Measurements 27 7. CONCLUSIONS 30 APPENDIX 31 REFERENCES 34 DISTRIBUTION LIST 35 ii

LIST OF ILLUSTRATIONS Page Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 Figure 13 Figure 14 Figure 15 Figure 16 Figure 17 Figure 18 Figure 19 Brett Receiver Input Circuit Assumed Equivalent Circuit of Fig. 1 r vs. Forward Bias Equivalent Tube Noise Resistance Illustration of Output Spectra Experimental Circuit for Impedance Determination Twice Amplitude Modulated Signal Spectrum Synthesized Demodulation Crystal Input Block Diagram of Experimental Setup Used for Synthesized System Investigation Setup for Measuring Standing-Waves at 200 and 400 me Setup for Measuring Standing-Waves at 1830 and 1950 me Uncorrected Sensitivities Obtained Using Synthesized Brett System Corrected Sensitivities Obtained with Synthesized Brett System Variation of Crystal Video Sensitivity with Input Modulation Frequency Experimental Brett Circuit Sensitivities Obtained with Actual Brett Input Circuit for Various Chopper Voltages Using HP 608 as Ext. Generator Sensitities Obtained with Actual Brett Input Circuit for Various Chopper Voltages Using Specially Built Crystal-Controlled Oscillator as Ext. Generator Sensitivities Obtained with Actual Brett Input Circuit Circuit Diagram of 2 me Crystal-Controlled Oscillator 1 2 5 8 9 10 12 12 15 17 18 21 22 24 25 28 28 28 29 iii

ABSTRACT A theoretical discussion of a somewhat novel broadband receiver input circuit is presented. The circuit utilizes two crystals, one in a chopper which acts on the incoming signal, and a second in a subsequent envelope detector. The amplifier which follows can be band.pass rather than video, as in the conventional crystal video circuit. Various problems related to sensitivity are discussed. Considerable experimental data which corroborates the discussion are also included. iv

ELFCT IC DEFUNSE GROUP T-OBIzCAL EPOR Noo 67 EEATA i Should be Uk Order EDG-4 10 Comma after "amplifier". 18 Fig. 7 f should be f 13 PFranethesesa mis:ig at esd of line 9 13 Should be e' in bottom tine iawtead of e ' out 22 Should be sinus sig n vf t of 520 6db for 11B 3(6). 27 Footnate 1 - Should be Fig. 19 (Not Fig. 21). 28 FFig. 18, line 5 - omit ai 1onal line after -2.9. 32 6biLast eqation sbate d bave 4 In denmaittor instead of 2,

I I ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN AN EVALUATION OF A SPECIAL BROADBAND RECEIVER INPUT CIRCUIT I. INTRODUCTION The receiver input circuit shown in Fig. 1 was recently developed by Mr. Herbert Brett at Evans Signal Laboratory. A number of other forms of this circuit have been investigated at Evans, but it is believed that the principle of operation is the same in each case. Phenomenologically, it has been observed that if a modulated RF signal and a sinusoidal external generator signal are impressed on the circuit as shown, the modulation can be transferred to a carrier at the external generator frequency independently of the RF carrier frequency. It is also possible to transfer the modulation to carrier frequencies which are harmonics of the external generator fundamental frequency. L I! XTAL A2 PLUNGER XTAL I I i = XTEI R NAL IDEAL EXTERNAL I (GENERATOR GENERATOR CURRENT PATH- G ENERATOR FIG I BRETT RECEIVER INPUT CIRCUIT * The analysis and measurements described in this report were carried out at the request of Evans Signal Laboratory.

I I - ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN In the Appendix it is demonstrated that the operation of the circuit of Fig. 1 can be interpreted as in Fig. 2. An input signal of carrier frequency T~o1~~~~sc ~CENTER FREQUENCY Acoswnt+ + fn X l_____ | ______ FIXEDR-F INPUT R' NMOPUT ^FILTER DEMOD. [FREQ circuit of Fig. 1 by crystal 1. The waveform of the modulation resulting at crystal 1 may be supposed to have a Fourier Series representation containing a component (at frequency fn) of interest. After filtering out undesired components such as at the oscillator fundamental frequency and harmonics, demodulation (at crystal 2 in Fig. 1) produces a signal of carrier frequency fn with modulation f[a(t); i.e., the modulation at the output of the circuit is a function f of the input modulation. In contrast to a crystal video circuit where subsequent amplification is video, the subsequent amplification of the above circuit is centered around a frequency fn. The circuit is similar to a crystal video circuit, however, in that envelope detection occurs prior to amplification. In a superheterodyne receiver, on the other hand, higher sensitivities are achieved by avoiding such a low level of detection. The comparison is expressed mathematically in the Appendix. The application of a circuit such as Fig. 1 in a broadband (microwave) receiver has been suggested. A loss of somewhat more than 3 db is expected 2

- ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN 1 in a crystal modulator. This advantage, relative to a conventional crystal video receiver, might be substantially offset through the use of a phasesensitive second detector, provided the circuit between the modulator and demodulator is non-dispersive over the RF band.2 2. SUMMARY OF CRYSTAL THEORY A principle remaining factor necessary for a sensitivity comparison between the circuit of Fig. 1 and the crystal video circuit is detector efficiency. In References 2 and 3 material is presented from which the following conclusions are drawn: The output signal to noise voltage ratio of a crystal detector is given by E: P M (1) N TB where M is called the figure of merit and is given by M = ' R (2) /Rt + Ra where, P = RF power into crystal, R = video (or output) crystal resistance, P = current sensitivity = current out (amps) power in (watts) t = effective crystal noise temperature ratio, and Ra = equivalent noise resistance of the subsequent amplifier. Note that the figure of merit depends on the noise characteristics of the amplifier following the detector. 1. It is assumed that for a fixed character of input signal envelope the detectability is measured by the average signal power absorbed by crystal 2 of Fig. 1. 2. An advantage of 3 db accrues for a phase sensitive detector (Ref. 1) compared to a linear detector at unity output signal to noise ratio, if the noise is noncoherent. Larger advantages can be realized at lower signal to noise ratios. 3

- ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN The output noise from a crystal consists of Johnson noise plus additional components in the presence of excitation. The amplitude of the excitation induced noise varies inversely with frequency over a range of frequencies from at least 50 cps to 60 me. The parameter t is thus a function of output frequency; i.e., the average value of t over a narrow band around a frequency f is referred to as the value of t at the frequency f. The parameter t varies widely with the degree of excitation. Values as high as 100,000 have been observed with reverse bias in the neighborhood of 50 cps. In the presence of a typical size local oscillator signal, large values of t are observed at audio frequencies. The value of t is typically less than 1.5 in the frequency range above perhaps 30 me with the usual local oscillator drive, however, which accounts in part for the choice of conventional intermediate frequencies. In the absence of excitation, the laws of thermodynamics require that the crystal noise temperature (ratio) t be unity. Now, if only a signal to be detected is applied to a crystal, the output noise power available in a limited frequency band is well below the input power. It has, in fact, been established experimentally that if only a signal near the limit of sensitivity is applied, the effective value of t is unity; i.e., the excitation induced noise output does not in this case limit sensitivity.1 Applying this result in Eqs 1 and 2, it is found that at unity output signal to noise ratio and with only a signal to be detected applied, p ~/V4kTB V/kTB R+R (3) P =- _ _ RE (3) M 5 R 1. See Ref. 2, pg. 346 and Ref. 3, p. 8. The assumption of unity noise ratio may be invalid below some frequency due to the 1/f character of the excess noise. This point is discussed in Section 3 along with flicker noise of the amplifiers, which has a similar character. -!

ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN Experimental studies indicate that for a video frequency output band the optimum M occurs at essentially zero bias. This is expected since t has been noted to increase extremely rapidly at low frequencies in the presence of excitation. It is interesting to consider whether a significant increase in M might be obtained by utilizing a crystal bias at a detector output center frequency above perhaps 10 me. In this frequency range t ~ 1 even for biases of a substantial fraction of a volt. Further, the ideal crystal characteristic, / qV \ i = Ioe -1), (4) indicates a monotonically increasing curvature (and thus 3) with increasing forward bias. An actual crystal characteristic becomes linear for large forward signals, however, so that an experimental curve of | vs forward bias may be expected to be in the form suggested by Fig. 3. If a transformer of turns ratio FORWARD BIAS FIG. 3 l:n at the crystal output and t k 1 are assumed, then M PnR (5) If, favorably, one makes n2R >> Ra ' then 5

ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN Elementary theory indicates that M should increase with forward bias under these conditions. Specifically: / qVb qAV l) ) (kT FikT Ai = Io e e - I - I - = I e e - 1(6) From Eq 6, for small AV, qVb AV kT kT then, MR -K/ e (9) For actual crystals, both 3 and R decrease rather rapidly beyond a certain small bias. It therefore seemed worthwhile to determine experimentally whether an increase in sensitivity occurs with moderate bias as the detector output center frequency is raised above perhaps 10 me. 3. SOME ELEMENTARY NOISE CONSIDERATIONS Before describing the experimental evaluation of the Brett circuit (Fig. 1), it is worthwhile to discuss briefly some noise problems which are of importance in assessing the advantages of a receiver employing this circuit 6Vb kT qVb compared to a video oheterodyne system. These problems have to do with flicker noise, the bof crystal noise spectrum already mentioned, and the design small bias. It therefore seemed worthwhile to determine experimentally whether of the input circuit of the amplifier followingwith moderate bias as the detector rystal. centFler frequency is raise in tubes is thoht to occur as a result of a "shimmering" 3. SOME ELEMENTARY NOISE CONSIDERATIONS (Fig. 1), it is worthwhile to discuss briefly some noise problems which are of compared to a video or superheterodyne system. These problems have to do with flicker noise, the 1/f crystal noise spectrum already mentioned, and the design f the emnoission from the cathought to occurface Flicker noise results in an equivalent of the emission from the cathode surface. Flicker noise results in an equivalent 1. Extensive measurements of flicker noise characteristics have been made at the University of Minnesota under Professor August Van der Ziel. J [ 6

ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN tube noise resistance plot with frequency similar in shape to that of Fig. 4.l The ordinate is proportional to available tube output noise power. The noise depicted consists in part of shot noise which is essentially independent of frequency and is the principal noise contribution at the high frequency end of the graph. The other major noise contributions is flicker noise which has nearly a 1/f power spectrum. In a broadband microwave monitoring application, input modulation bandwidths of the order of 1 me would be likely. It is apparent that even large excess (flicker) noise components over a band of a few kilocycles per second may only moderately increase the rms value of the output noise of a receiver with a 1 me bandwidth. Further, this audio frequency noise would be relatively ineffective in masking, for example, 1 it sec pulses. The sensitivity for a given signal should depend rather critically on the signal to noise ratio existing in a limited frequency region containing the first few harmonics of the input modulation. As the center frequency of this limited frequency region is lowered, one would expect that the sensitivity would remain nearly constant until flicker noise became substantial in the region. The sensitivity should then decrease at roughly 3 db per octave, with further decreases in center frequency. The above facts suggested the use of a narrow band filter, centered around the input modulation fundamental frequency component, for measuring sensitivity in the experimental circuit. A substantial advantage of this technique is that relative sensitivities can be determined from meter readings, by observing the RF signal level at which a high prescribed narrow band signalto-noise ratio is realized. For the tests described in Section 5 a three to 1. The 6CB6 is one of the best tubes commercially available for applications where low flicker noise is desired. Other good tubes in this respect are the WE416A, Phillips EF40, and the 6AG5. It is in general advisable to select tubes. 7

9 p01 (la!Z Jap UDoA JaV) 33NV.LSIS38 3SION 3e8n1 LN3lVAInO3 " 't1 SdO' AON3nO38.I ~01 01 ow-~~~.... - 01 1033NN03 3001W1 9839 "" ~~" S ----..... -— _ )s -— ''-,, "% C,) 0O.... ___ - - - _ _ _ -^x S^Io \X=01L H t ==t I I I I I I 1 1 = 1= ==== === t0 o01 2262 A-GI-277 M.B. 11-1-56

ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN one "signal-plus-noise" to "noise" criterion for an 8 cps band centered at about 1.1 kc was used. The input signal modulation was a 1.1 kc sinusoid. The 3:1 S+N/N ratio is estimated to correspond to a somewhat less than unity signal-to-noise ratio for a video system with a bandwidth from 10 cps to 250 kc. The objective of the experimental study was to make a comparison between the sensitivities of a system employing the Brett circuit and a crystal video system, rather than to measure the absolute sensitivities achievable. The meter criterion permitted this comparison to be made easily. The expected noise variation with frequency at the output of the second detector is in the form of Fig. 5. A pulse envelope spectrum and a narrow band filter characteristic are also illustrated. The noise amplitude \ NARROW BAND FILTER CHARACTERISTIC N SPERUM NOISE SPECTRUM FREQUENCY FIG. 5 ILLUSTRATION OF OUTPUT SPECTRA is a function of frequency, as previously described, so that the "sensitivity" using the above meter criterion can be expected to vary at low input modulation frequencies. The sensitivity decreases with decreasing frequency when flicker or 1/f noise becomes appreciable compared to tube shot noise or Johnson noise. At this frequency the spot noise figure starts to disintegrate. The output impedance of the low level detector crystal depends on the crystal type and sample and crystal bias. A source impedance exists, for a 9

I - ENGINEERING RESEARCH INSTITUTE ~ UNIVERSITY OF MICHIGAN given fixed frequency amplifier which produces the minimum obtainable noise figure. An analysis of this problem is presented in Ref. 4. It is shown that the optimum noise factor, Fopt = 1 +2 Req (10) Rs opt where Re is the equivalent noise resistance for the input tube and its load. Rsopt is the optimum source impedance given by R - / Rgeq (11) Rspt Tg/T where Rg = input resistance of the tube and input circuit Tg T = input resistance temperature to ambient temperature ratio For other source impedances F = 1 + T Rs e (1 R) (12) T R Rg Rs For the cascode input circuits utilized, Req is known to sufficient accuracy. R and Rs are determined using the circuit of Fig. 6. The frequency of the g HIGH IMPEDANCE GENERATOR AMPLIFIER FIG. 6 high impedance source is varied with Rs removed and, if the amplifier is sufficiently broadband, Rg is determined from the 3 db bandwidth of the response. A similar measurement with the source connected determines Rs. The noise factor of the fixedfrequency amplifier was estimated from the data obtained using this J I - 10

I - ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN procedure. The sensitivity data were corrected to.5 db noise figure at 2 me and 2. db at 60 mc. It is noted that the bandwidth varies with the source impedance in the above circuit. The amplifier is followed in the experimental circuit by a linear detector. Further, the zero signal noise output of the linear detector is sampled at a frequency much lower than the bandwidth of the amplifier. This noise is relatively independent of the bandwidth of a fixed gain amplifier under these conditions (see Ref. 5 ). It is concluded that the noise factor correction previously described is sufficient for the relatively high signalplus-noise to noise ratio utilized at the linear detector output. 4. DESCRIPTION OF EXPERIMENTAL TECHNIQUES UTILIZED IN THE INVESTIGATION Two experimental methods were employed in obtaining the data in this report. Some sensitivity measurements were made directly on the circuit of Fig. 1. The second method of measurement stems from the explanation of the circuit operation which was illustrated in Fig. 2. It is easily shown than the spectrum of a filtered, twice sinusoidally amplitude modulated signal (as assumed at x-x in Fig. 2 in the Introduction), has the form of Fig. 7. If either or both modulations are not sinsuoidal, additional sidebands are present. Then the demodulation process can be regarded as mixing of the carrier and external generator frequency sidebands. The significant difference between this circuit and the conventional superheterodyne circuit is that the two signals in Fig. 7 are of comparable size, while in the superheterodyne mixer one of the signals (the local oscillator signal) is much larger. From this viewpoint, one sees 1. An alternative was to actually optimize the input match for each run. 11

ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN -If Vl fo= SIGNAL CARRIER FREQUENCY f S SIGNAL MODULATION FREQUENCY fn= EXTERNAL GENERATOR FREQUENCY FIG. 7. TWICE AMPLITUDE MODULATED SIGNAL SPECTRUM that the desired modulator output centered at fn will be obtained with the input of Fig. 8 where one of the groups of sideband spectra has been omitted. Since the signals are small, the reduction of sensitivity caused by this omission can be calculated assuming a square-law demodulator characteristic. -- f n o — I I I 0 0 m~~~ FIG. 8. SYNTHESIZED DEMODULATION CRYSTAL INPUT Sensitivity comparisons can thus be made using a carrier and a sideband signal whose frequency separation is the desired external generator frequency, both signals being modulated by some prescribed (input) modulation signal. The advantage of synthesizing the modulator output rather than using the actual circuit is that tests at various RF center frequencies and external generator frequencies can be made without having to construct a series of modulators

- ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN and filters. The advantage of omitting one sideband is that the problem of phasing three oscillators representing the carrier and both sidebands is avoided. Let the twice modulated signal at x-x in Fig. 2 be e = E(1 + mm cos cmt)(l + mn cos nt) cos coot E = E(l + mm cos Cant) cos wot + E2 nn (1 + mm cos oat) cos (X0 - un)t (13) +. mn (l + mm cos wDt) cos (co + on)t which corresponds to the spectrum of Fig. 7. When passed through a square law device, and when mm = mn = 1, the output is (14) io = PE2(1 + 2 cos Io t + 1/2 + 1/2 cos 2uat)(l + 2 cos ant + 1/2 + 1/2 cos 2cint x (1/2 + 1/2 cos 2oot) After amplifying at a center frequency On and passing through a linear detector, the output with frequency uc is eout = 2P AE2 cos aDnt (15) where A is a gain factor. If, however, one sideband of the RF signal is omitted or filtered, the signal applied to the square law device is in the form e' = E(1 + m cos cmt) [cos Cot + 1/2 mn cos (W - Cwh)t] (16) For nmn = mn = 1, the output from the square law device is i'= PE2 (1 + 2 cos aot + 1/2 + 1/2 cos 2 wat)x 0 0 (17) [l/2 + 1/2 cos 2wot + 1/2 cos %ht + 1/2 cos (2Xo - h)t + 1/8 + 1/8 cos 2(o-<n)t] After amplifying at a center frequency cDn and passing through a linear detector, the component with frequency (am is e' = PAE2 cos ant (18) 13

ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN Hence the omission of a sideband reduces sensitivity of the demodulator portion of the circuit by 6 db. The experimental circuit based on this analysis will be referred to as the synthesized Brett system. 5. SYNTHESIZED SYSTEM MEASUREMENTS In making sensitivity measurements with the synthesized Brett system, the parameters varied were RF frequency, chopper1 frequency, demodulator crystal (type and sample), and crystal bias. The sensitivity data thus obtained were compared with crystal video data obtained at the same RF frequencies after converting the experimental setup to a video system. 5.1 Basic Setup A block diagram of the basic experimental setup for the synthesized system investigation is shown in Fig. 9. As pointed out earlier, the input to the demodulator crystal (crystal 2 in Fig. 1) of the Brett circuit can be effectively synthesized with two commonly-modulated signals whose difference frequency is the desired chopper frequency. Generators A and B supply the two RF signals and the audio oscillator (Hewlett-Packard model 200AB) supplies the common modulation. The 10 db pads in each generator line provide isolation of the generators while the 10 db pad in the common line makes the impedance more nearly 50 ohms. The two tuning stubs provide a means of matching the crystal to the 50 ohm source. The signal out of the demodulator crystal is amplified by a fixed-frequency amplifier. The particular amplifiers used have tuned circuit inputs with provision for applying crystal bias. 1. The external generator signal will be referred to as the chopper signal in the remainder of the report. 2. The RF matching problem is discussed in Section 5.2 J I 14

V) 0 cN U) ax 0 CM ed 10 DB PADS TUNING STUBS DUMONT 304 A OSCILLOSCOPE XTAL MIXER X SLOTTED LINE AND SLIDING PROBE \J1H / B SYMETRICAL PATHS FIGURE 9 BLOCK DIAGRAM OF EXPERIMENTAL SETUP USED FOR SYNTHESIZED SYSTEM INVESTIGATION

- ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN After sufficient fixed-frequency amplification, a crystal diode detector is used to give linear detection. Subsequent video amplification is provided by the vacuum tube voltmeters (Hewlett-Packard 400-D and General Radio 1231-B). The use of the 8 cps bandwidth filter and the meter criterion was discussed in Section 3. The 400-B VTVM placed after the filter measures the signalplus-noise or noise voltages, as the case may be, in the narrow band. It was necessary to incorporate 5000 j~if in the meter to perform the averaging of the 8 cps band of noise, and a time constant of several seconds resulted. The oscilloscope (Dumont 304-A) provides visual monitoring. To convert the above setup to a crystal video system, one of the RF generators is turned off, and the fixed-frequency amplifier and the linear detector are removed. 5.2 Basic Procedure The first step in obtaining a particular sensitivity measurement is to tune generators A and B. One of the generators is tuned to the RF frequency desired and the other to a frequency that causes the difference frequency to fall in the passband of the fixed-frequency amplifier. The second step is to adjust the audio oscillator to the center frequency of the 8 cps bandwidth filter. The third step is to adjust the two tuning stubs to give a maximum signal output as indicated by the HP 400-B VTVM. The fourth step is to adjust the relative attenuator settings of generators A and B until the signal-plusnoise output in the 8 cps video band is three times the noise alone in the same band; i.e., until the meter criterion described is satisfied. Since it is unlikely that an impedance match can be obtained at two different frequencies by means of double-stub matching, the final step is to measure the standing waves associated with each generators mismatch. Two methods of measurement were employed. depending upon the RF fre-qpiency..

i ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN For RF frequencies of 200 and 400 me, the measurement of standing waves was accomplished with a HP 803-A VHF bridge. A block diagram of the, setup is shown in Fig. 10. Generators A and B and the 10 db pads are HP 417 A VHF D ET. HP 608 HP 803A lI l RSLOTTED LINE STUBS ---— XTAL FIG. 10. SETUP FOR MEASURING STANDING-WAVES AT 200 &400 MC removed from the setup of Fig. 9 and the VHF bridge with its associated equipment is inserted in their place. The uncorrected impedance at the one generator's frequency is then determined. This impedance is used to obtain the standing-wave ratio in the line of that generator. From the standingwave ratio, the associated power loss due to mismatch is determined. The measurement is then repeated at the frequency of the other generator, and its power loss due to mismatch is determined. For the RF frequencies of 1830 and 1950 me, the measurement of standingwaves was accomplished with a slotted line and sliding probe. Since the RF signal level in the line was quite low, and since a small amount of coupling between the probe and the line was desired, superheterodyne receiver sensitivity was required to detect the signal picked up by the probe as it was moved along the line. A block diagram of the setup is shown in Fig. 11. J I 17

ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN I - 1 HP 400 B SUPERHETERODYNE RECEIVER - H l |VTVM I IL uGR874 I UHF TUNING MIXER SIG. GEN. I STUBS HP614A 10 DB PADS UHF' SIG. GEN. OSCT GR 874 LB SLOTTED LINE TUNING FIG. II. SETUP FOR MEASURING STANDING-WAVES AT 1830 AND 1950 MC The superheterodyne receiver consists of the two tuning stubs for matching, a UHF signal generator for heterodyning, a mixer, and a HP 400-B VTVM for amplifying and indicating the mixer output signal. One of the RF generators is turned off. The standing-waves measured are then due to the other generator's mismatch. After the local oscillator frequency and the two stubs in the superheterodyne receiver circuit are adjusted, the probe is moved along the slotted line and the maximum and minimum readings on the HP 400-B are recorded. The voltage-standing-wave ratio is then calculated as follows: VSWR = (S+N)2ax - (N (19) VSWR = (19) 2 2 (s+N) min (No) where (S+N)max and (S+N)min refer to the maximum and minimum HP 400-B readings obtained with the generator on, and (No) refers to the no-signal noise reading. From the voltage standing-wave ratio, the mismatch loss for one generator is - I

- ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN determined. The entire procedure is then repeated to obtain the mismatch loss for the other generator. It was desirable to obtain a number for each run, determined by the RF signal generators' attenuation readings and various corrections that are summarized in Section 5.3 indicative of the sensitivity achieved in the run. This number is referred to in this report as the relative sensitivity. The attenuator reading for the larger of the two RF signals representing the carrier and one sideband of the twice-modulated signal was arbitrarily chosen as the uncorrected reference level. The level of the two signals differed by 6 db, which is the difference between the carrier and sideband of a 100% sinusoidally modulated signal. It was observed that if one attenuator were set higher by a certain amount and the other set lower by the same amount for a particular setup, the output remained unchanged. Therefore, after corrections were applied to the attenuator readings for a particular run, the corrected readings were referred to this 6 db separation. The above steps were carried out for zero crystal bias on each crystal and the raw data were recorded. The crystal bias was then increased to the point where the signal-plus-noise indicated by the HP 400-B VTVM was maximum, and the bias voltage and sensitivity at this point were recorded. Increasing the crystal bias produced two noticeable effects. The first effect was that the no-signal noise reading decreased steadily and finally reached a constant value. The second effect was that the signal-plus-noise reading initially increased and then dropped off rapidly. It would be expected, then, that the optimum bias would be greater than the bias associated with the maximum HP 400-B VTVM signal-plus-noise reading since the no-signal noise is 1. This statement was true as long as the two signals were within perhaps 20 db of each other. 19

ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN still decreasing at the maximum point. Actually, however, the signal-plus-noise drop-off is so rapid that there is essentially no difference between the true optimum bias and the bias associated with the maximum reading. The bias voltage recorded has thus been referred to as the optimum bias. The RF frequencies investigated were 200, 400, 1830, and 1950 mc. At 200 and 400 me, HP 608 VHF signal generators were used for generators A and B. At 1830 and 1950 me, HP 614-A and HP 616-A UHF signal generators were used. Chopper frequencies investigated were 60 me and 2 mc. Both the 60 mc and the 2 me fixed frequency amplifiers have a cascode input circuit. The noise figure of the 60 me amplifier measured with a 4,000 ohm generator was about 3 db. The noise figure of the 2 me fixed-frequency amplifier measured with a 12,000 ohm generator was also about 3 db. The corresponding crystal video sensitivities are obtained as above, except there is only one RF generator to tune and no difference frequency adjustments to be made. Furthermore, the measurements are made only at zero crystal bias since it was determined experimentally that this was essentially optimum.. 5.3 Results of Synthesized System Measurements The raw data for the synthesized Brett system and the corresponding crystal video system are presented in Fig. 12. The sensitivities in Fig. 13 are relative sensitivities; i.e., the various corrections previously discussed have been applied. These corrections are associated with RF mismatch loss (discussed in Section 5.2), omission of the one side band in the synthesized signal (discussed in Section 4 ), fixed-frequency amplifier input circuit noise figure variation with crystal type and bias (discussed in Section 3 ), different peak voltages between the synthesized signal and the crystal _____________________________ 20 -

Crystal 60 mc Difference Frequency 2 mc Difference Frequency Crystal Freq. Video Sensitivity (dbm) Sensitivity (dbm) Type (mc.) Sensitiv- _ ity (dbm) bias Opt. bias O bias Opt. bias 1N56 ( 200 -49.5 1N56 400 -48.0 1N82A2) 200 -55.0 -36.5 Opt bias is 0 -50.2 -51.5 (.04 v.) 1N82A 400 -58.0 -39.5 Opt bias is 0 -48.0 -50.0 (.035 v.) 1N82P3)1 200 -50.5 1N82A 400 -54.5 (4) 1N69 200 -52.0 -35.0 Opt bias is 0 -44.5 Opt bias is 0 1N69 400 -49.0 -35.0 Opt bias is 0 -40.0 Opt bias is 0 (6) lN21B 200 -46.0 -29.5 -39.5 (.14 v.) -42.5 -43.0 (.12 v.) 1N21B 400 -49.5 -33.0 -42.2 (.134v.) -42.0 -47.0 (.11 v.) 1N21B 1830 -33.5 -42.5 (.13 v.) 1N21B 1950 -33.0 -42.5 (.15 v.) 1N21B(" 1830 -56.4 -38.5 -44.5 (.08 v.) -45.5 Opt bias is 0 1N21B 1950 -59.1 -38.7 -43.0 (.065 v.) -47.5 -48.0 (.02 v.) lN212) 1830 -54.0 -34.5 -44.0 (.10 v.) 1N21B 1950 -55.5 -33.7 -41.0 (.17 v.) 1N21 1830 -54.0 -34.0 -41.0 (.7 v.) 1N21A 1950 -55-5 -33.7 -41.0 (.17 v.) 1N2314) 1830 -52.3 -35.5 -44.0 (.09 v.) -42.5 -46.5 (.05 v.) 1N23 1950 -54.2 -35.5 -42.0 (.09 v.) -45.0 Opt bias is 0 ('5) 1N23 1830 -36.0 -43.5 (.10 v.) 1N23 1950 -36.5 -43.0 (.09 v.) UNCORRECTED SENSITIVITIES OBTAINED WITH SYNTHESIZED BRETT SYSTEM Figure 12 Superscripts on crystal types designate particular samples. NOTE: 21

CCry l Fq rystal 60 mc Difference Frequency 2 me Difference Frequency Type |(mc Sensitiv- Sensitivity (dbm) Sensitivity (dbm) Type (mcc) Sensitiv|ity (dbm) 0 bias Opt. bias 0 bias Opt. bias 1N82A( 200 -61.6 -49.3 Opt. bias is 0 -53.1 -53.4 (.04v.) 1N82A 400 -63.6 -49.7 Opt. bias is 0 -52.6 -52.7 (.04 v.) lN69 4) 200 -55-0 -48.2 Opt. bias is 0 -48.2 Opt. bias is 0 1N69 400 -50.5 -45.0 Opt. bias is 0 -43.1 Opt. bias is 0 N21B(6) 200 52.0 -44.2 -46.4 (.14 v.) -45.5 -49.8 (.12 v.) 1N21B 400 -51.8 -43.1 -44.6 (.13 v.) -44.9 -48.7 (.11 v.) 1N21B01 1830 -56.4 -52.0 Opt. bias is 0 -49.8 Opt. bias is 0 1N21B 1950 -59.1 -52.7 Opt. bias is 0 -52.6 Opt. bias is 0 1N23 1830 -52.3 -49.2 -49.8 (.09 v.) -47.0 -49.8 (.05 v.) 1N23 1950 -54.2 -48.2 -49.9 (.09 v.) -49.6 Opt. bias is 0 Figure 13 CORRECTED SENSITIVITIES OBTAINED WITH SYNTHESIZED BRETT SYSTEM Superscripts on crystal types designate particular samples. NOTE: 22

ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN video input signal', and different signal generator calibrations. The above mentioned sensitivities were obtained using an input modulation frequency of about 1.1 kc. For one particular setup3 the input modulation frequency was varied from about 275 to 4600 cps. There was essentially no change in the synthesized Brett system sensitivities as expected, while the crystal video sensitivities varied as shown in Fig. 14. 6. BRETT CIRCUIT MEASUREMENTS In making sensitivity measurements with the actual Brett circuit (sketched in Fig. 1), the parameters varied were RF frequency, chopper frequency, and chopper voltage. The sensitivity data thus obtained was compared with crystal video data obtained at the same RF frequencies after converting the experimental setup to a crystal video setup. 6.1 Setup A block diagram of the setup for obtaining sensitivity data with the Brett circuit is shown in Fig. 15. The HP 620-A BF signal generator supplies 1. The synthesized double modulated signal can be written el = El(l + mm cos omt)(l + mn cos cnt) cos Wot. For mr and mn = 1, the peak value of this signal is 4E1. The crystal video input signal can be written e2 = E2(1 + nm cos Wmt) cos Wot. For mm = 1, the peak value of this signal is 2E2. Thus, to correct to the same peak input voltage, the synthesized signal for which the carrier level is taken as reference must be penalized 6 db. 2. One generator was calibrated in terms of the other by first using one and then the other in a crystal video setup. 3. 1N82A crystal No. 2, RF frequency of 400 me, and chopper frequency of 2 mc. 23

CIRCUIT OF FIG.4 N82A CRYSTAL RF FREQUENCY 400 MC 50 100 500 1000 FREQUENCY CPS FIG.14. VARIATION OF CRYSTAL VIDEO SENSITIVITY WITH INPUT MODULATION FREQUENCY:_::: - z"-// * --- —-j*~~ — ^ --- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~"-:=- z =-::::::^===:::::===:~~~~~~~~~~ iP. --- —l --------— ^ --- —-------------------------- - ------ ------------------------- -~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~..; -- - ^ ----------------- - - _ _ -------------- / ICI FFG4 N2 RSA ' ___RFFEUNY40M m, 0 4 0 cM lO. 2 5000

-- ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN the RF signal to be detected, and the HP 200-AB audio oscillator supplies the input modulation. The chopper signal is supplied by the external generator. The output from the Brett circuit is fed to the circuit shown to the right of line "xx" in Fig. 9. i I:FIXED- FREC AMPLIFIER, I ETC. PLUNGER w- - GUIDE.,XTALI ~OSC SIG. GEN. ADAPTER r -C TO r EXTERNAL GENERATOR FIG. 15. EXPERIMENTAL BRETT CIRCUIT To convert the above setup to a crystal video system the chopper crystal (crystal 1) is replaced by a solid piece of brass (which provides a short circuit to the case of the Brett structure), and the external generator, the fixed-frequency amplifier, and the linear detector are removed. 6.2 Procedure The first step in obtaining a sensitivity measurement with the Brett circuit is to tune the HP 620-A SHF signal generator to the RF frequency desired. The second step is to tune the external generator to the chopper frequency desired and adjust its output level to a prescribed voltage. The third step is to adjust the HP 200-AB audio oscillator to the center frequency of the 8 cps bandwidth filter. The fourth step is to adjust the tuning stubs and waveguide 25

ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN plunger to give the maximum signal output as indicated by the HP 400-B VTVM. The final step is to adjust the attenuator on the HP 620-A generator until the meter criterion previously described has been satisfied. Chopper frequencies investigated were 60 and 2 me. The fixedfrequency amplifiers used were those previously described. A HP 608-A VHF signal generator was used for the external generator in the 60 me case, while a GR 1001-A signal generator was used in the 2 me case. There was a danger that a significant component of an observed output signal (HP 400-B reading) was due to direct feed-through of energy from the external generator. Such an output, which must occur in the audio band, would result because of incidental modulation of the external generator signal. To determine whether such an extraneous signal were present, the RF generator was turned off. The external generator signal level was then varied, and the variation in output observed. An external generator signal level was chosen at which a very small increase in output occurred. It was found that an appreciable extraneous output did not occur with an external generator signal level of 100 mv. This level was used in obtaining the data for making a comparison with the video system results. To determine whether the increase in the no-signal output was actually feed-through, and not due to an increase in the effective crystal noise temperature ratio (t), an oscilloscope was attached to the output of the fixedfrequency amplifier. The external generator drive was then varied, and the resulting variation in output was observed on the oscilloscope. By this procedure, it was determined that the increase in output was primarily due to feed-through of the external generator signal. 26

ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN Relative sensitivities for external generator voltages greater than 100 my were obtained using the HP 608-A generator. These sensitivities, which are for a particular crystal and RF frequency, are presented in Fig. 16. Since the sensitivity increases for larger external generator signals than that which just produced a very small increase in the no-signal audio band output, a low noise external generator could be expected to give results superior to those of Fig. 16. It was felt that a crystal controlled oscillator would have the best output spectrum that could be readily obtained, and hence greater external generator drives might be employed before the no-signal audio band output increased appreciably. A 2 me crystal-controlled oscillator was built, and the results obtained using it are presented in Fig.17.1 It was found in this case that an appreciable extraneous output did not occur with an external generator signal level of 250 mv. The corresponding crystal video sensitivities are obtained as outlined in the steps above except there is no chopper generator to adjust. 6.3 Results of Brett Circuit Measurements Relative sensitivity for the actual Brett system and the corresponding 2 crystal video system are presented in Fig. 18. The Brett sensitivities were obtained with chopper voltages of 100 my as previously explained. The variations of sensitivity with external generator voltages are presented in Figs. 16 and 17 as already described. As indicated by these results, the sensitivities of Fig. 18 could be increased by about 4 db before being limited by feed-through by employing a crystal controlled oscillator. 1. The oscillator circuit is presented in Fig. 21. 2. The external generator for the data of Fig. 18 was the HP 608A. 27

Chopper Voltage (mi 1 1 Tivonl-. 0 bias Sensitivity (dbm) Opt. bias Sensitivity tmhT') \ J L V % -, %A. \ k.I.L / 100 -65.0 -67.9 (.08 v.) 150 -68.0 -70.9 (.08 v.) 200 -68.5 -71.4 (.08 v.) 250 -69.0 -71.4 (.08 v.) 400 -69.0 -71.4 (.08 v.) Figure 16 Sensitivities Obtained With Actual Brett Input Circuit for Various Chopper Voltages Using HP 608 as Ext. Generator (1N369A crystals used as chopper and detector, 60 me chopper frequency 8.5 Kmc RF frequency.) Chopper Voltage 0 bias Sensitivity Opt. bias Sensitivity (millivolts) (dbm) (dbm) 100 -65.9 -68.6 (.05 v.) 250 -68.9 -71.6 (.05 v.) 500 -73.4 -75.6 (.05 v.) 1000 -73.4 -75.6 (.05 v.) 1400 -73.4 -75.6 (.05 v.) Figure 17 Sensitivities Obtained With Actual Brett Input Circuit for Various Chopper Voltages.Using Specially Built CrystalControlled Oscillator as Ext. Generator (1N369A crystals used as chopper and detector, 2 me chopper frequency 8.5 Kmc RF frequency.) Freq. Crystal 60 mc Chopper Frequency 2 me Chopper Frequency Video Sensitivity (dbm) (kmc) S i Sensitiv-ity dbSensitivity (dbm) ity (dbm) 0 bias Opt. bs 0 bias Opt. bias 8.5 -75.0 -65.0 -67.9 (.08 v.) -66.4 -67.1 (.08 v.) 9.25 -74.5 -64.5 -66.4 (.08 v.) -64.4 -66.1 (.08 v.) 10.0 -73.0 -61.0 -62.9/(.08 v.) -61.9 -63.1 (.08 v.) 10.4 -71.5 -54.0 -57.9 (.08 v.) -56.9 -58.6 (.08 v.) Figure 18 Sensitivities Obtained With Actual Brett Input Circuit (1N369A crystals used as chopper and detector, chopper voltage 100 mv.) 28

6AG7.002,d.IM 12AT7 2 40 In I 10 O ro In U (M N FIG. 19. CIRCUIT DIAGRAM OF 2 MC CRYSTAL- CONTROLLED OSCI L LATOR

ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN - 7. CONCLUSIONS The results of the experimental investigation and the accompanying analysis indicate that a crystal video system has 0 to 6 db greater sensitivity than the Brett system for pulse repetition frequencies above 1000 pps. The Brett system appears to offer superior sensitivity for signals with a large duty factor (say 50%) at pulse rates less than approximately 100 pps. The Brett system incorporates a conventional IF strip in place of the video amplifier of a crystal video system. Particular advantages which accrue relate to size and microphonics. In a mobile or portable broadband detection system, these advantages may be substantial. L J 30

ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN APPENDIX In this section, some general remarks concerning the principle of operation of the ESL circuits is presented. It seems desirable to preface this discussion with a brief mathematical statement of detection, mixing, and the apparent manner of operation of the Brett circuit. Detection, as in a crystal video circuit, is achieved by impressing -7 a signal I = a(t) cos Cot on a non-linear circuit. It is assumed that the transfer characteristic of the circuit can be expressed in the form (A.1) 2 iout = ain + in +.... (A.2) For the case of low level detection the important output is due to the square I law termt) a 2(t iout = C a2(t) cos2 CDot = (1 + cos 2 Cot) (A.3) 2 Ba2(t) from which the useful output 2a is selected by filtering. The output signal current is thus proportional to input power. In the superheterodyne circuit, a local oscillator signal of frequency f2 is utilized. For a diode mixer, then, a signal I = a(t) cos Cot + A cos m2t (A.4) is impressed on a circuit with a transfer characteristic such as in Eq A.2. The first few terms resulting from the substitution of Eq A.4 in Eq A.2, after utilizing some well known trigonometry identities, are: linear terms square law terms iout = ca(t) cos cwot + cA cos o2t + a(t) (1 + cos 2Uot) + (1+ cos ao2t) 2 square law terms + PAa(t) cos (0o-<2)t + Aa(t) cos (wo + ac2)t +.... (A.5) 31

ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN Ordinarily one of the latter two terms indicated would be filtered out as the useful output. Here the output signal current is proportional to the input signal voltage. Consider the operation of the circuit of Fig. 2. After modulation and filtering, a signal a(t) r ) ~ = 1 + m g(t)j cos Cot (A.6) 2 is impressed on the demodulator, which has a transfer characteristic such as in Eq A.2. The function g(t) depends on the modulation wave form chosen1. Substituting again in Eq A.2, the initial terms are (A.7) i aa(t) + mg(t] Cos C+t + (t + 2 (t) + 2 mg(t + m(l+cos 2mot) Let us suppose that.the "periodic in time" function g(t) can be expressed as a Fourier Series, and that a particular component of interest has amplitude gn and frequency fn. Writing only the outputs from the first two terms above which are due to the modulation component of frequency fn, iout = aa t1 + m gn cos (Wnt + 0n)]cos wot 2 + t)[1 + 2 mgn cos (nt + 0n) 2 2 + mgn {1 + cos 2(Tht + 0J](l + cos 2wo)t) +. (A.8) Consider the term n~a2 (t) 2 gncos (nt + n) This output current component, which is centered around the frequency o., contains the input modulation information, and is zero when the modulator is turned off. It is proportional to the curvature of the crystal characteristic 1. The reduction in peak amplitude due to modulator losses is not considered here. The function g(t) is assumed to vary periodically between +1 and -1. The degree of modulation achieved at the modulator is indicated by the modulation index m. 32

- ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN and to input power, just as for straight detection. The other signal components displayed above are video or RF except for a term centered around an which is also potentially useful but is smaller than the ao terms. If higher order terms of the power series of the transfer characteristic are considered, outputs centered at higher harmonics of a% are among those present, but generally can be expected to be smaller than the output at %n. It is perhaps worth pointing out that for sinusoidal modulation, g = cos (ot + 9). The Fourier Series is in this case very simple, containing just one component of unit amplitude and frequency.f. The largest g which can occur is readily shown to be realized for the fundamental component of a square wave, when gl = 4//. In a practical circuit a loss accrues from the fact that m is less than unity. It is interesting to consider whether the analysis of Fig. 2 just presented is a valid description of the operation of the circuit of Fig. 1. Referring to Fig. 1, the leads designated as filters are intended to be essentially short circuits to the external generator signal, and to have nominal RF impedances. Then the external generator voltage causes a current to flow which is largely confined to the path indicated in Fig. 1. This current can reasonably be expected to effect a variation of the RF impedance of crystal 1, and thus a variation of RF absorption of crystal 1. In fact, this clearly would be the only important effect of the external generator signal if the external generator current were confined to the path mentioned above. Now, one might expect that as a result of the varying absorption of crystal 1, the RF signal absorbed by the circuit of crystal 2 would also vary (with, of course, the same periodicity). This is tantamount to modulation of the RF signal by the circuit of crystal 1. 33

REFERENCES 1. R. A. Smith, "The Relative Advantages of Coherent and Incoherent Detectors: A Study of Their Noise Spectra Under Various Conditions," Proc. IEE, 89, Part 3, Sept. 1951. 2. Torrey and Whitmer, "Crystal Rectifiers", MIT Rad. Lab. Series, Vol. 15, McGraw-Hill, 1948. 3. Van Voorhis, et. al., "Crystal Detectors and the Crystal-Video Receiver," Report 638, MIT Radiation Laboratory. 4. H. Wallman, C. P. Gadsten, A. B. Macnee, "A Low Noise Amplifier," Proc.IRE, June 1948. 5. J. L. Lawson, G. E. Uhlenbeck, "Threshold Signals," Rad, Lab. Series, Vol. 24, McGraw-Hill, 1950. 34

DSTBUIOISTRIBUTION LIST 1 Copy Document Room Stanford Electronic Laboratories Stanford University Stanford, California 1 Copy Commanding General Aimy Electronic Proving Ground Fort Huachuca, Arizona Attn: Director, Electronic Warfare Department 1 Copy Chief, Research and Development Division Office of the Chief Signal Officer Department of the Army Washington 25, D. C. Attn: SIGEB 1 Copy Chief, Plans and Operations Division Office of the Chief Signal Officer Washington 25, D. C. Attn: SIGEW 1 Copy Commanding Officer Signal Corps Electronics Research Unit 9560 TSU Mountain View, California 40 Copies Transportation Officer, SCEL Evans Signal Laboratory Building No. 42, Belmar, New Jersey FOR- SCEL Accountable Officer Inspect at Destination File No. 22824-PH-54-91 (1701) 1 Copy J. A. Boyd Engineering Research Institute University of Michigan Ann Arbor, Michigan 1 Copy Document Room Willow Run Laboratories University of Michigan Willow Run, Michigan 12 Copies Electronic Defense Group Project File University of Michigan Ann Arbor, Michigan 1 Copy Engineering Research Institute Project File University of Michigan Ann Arbor, Michigan 35

UNIVERSITY OF MICHIGAN 3 9015 02523 02051111 3 9015 02523 0205