THE EFFECTS OF AN UNWANTED SIGNAL MIXED WITH THE CARRIER SUPPLY OF RING AND COWAN MODULATORS

By D. G. TUCKER, Ph.D., Associate Member.

SUMMARY It is shown how the performance of a ring or Cowan modulator, when the carrier supply has an unwanted signal mixed with it, can be investigated by a frequency-analysis of its modulating function. Results obtained for particular cases lead to certain general conclusions, among which are the following: {a) The primary modulating effect of an unwanted signal mixed with the carrier is largely independent of the resistance of the circuit supplying the carrier, even though this affects the shape of the modulating function very considerably. The amplitude of the unwanted component in the modulation is, relative to the wanted component, only about one-half of that in the carrier input. (b) Ring and Cowan modulators give a similar performance in this respect. (c) When the frequency of the input signal is equal to the carrier frequency, the output of the difference frequency between the carrier and unwanted signal is zero for an ideal modulator, and very small for one with a carrier supply of high resistance; as the resistance is reduced, the output of difference frequency increases, until with zero resistance it has a ratio relative to a main sideband of about onequarter of the amplitude ratio of the unwanted signal relative to the carrier in the carrier-supply circuit.

impracticable, but a practical method is to plot the distorted modulating function as a graph, and to analyse it into its frequency components by means of suitable apparatus. Specific results can be obtained in this way for any particular modulator circuit. Results and conclusions which have general significance can be derived, however, from the treatment of one typical series of each of ring and Cowan types. The modulator circuits which are considered here use rectifiers with the resistance laws3 R = 100 + 1046-4^ for the d.c. resistance (which concerns the carrier ciicuit), r = 60 + 5 OOOe-6^ and for the a.c. resistance (which concerns the signal circuit), where V is the voltage across the rectifier due to the carrier (and to any bias that may be applied, although none is considered in this paper). These resistance laws correspond fairly closely to those associated with copper-oxide rectifiers, say type G2. The circuit impedance considered is 600 ohms for the ring modulators, and 6 000 ohms for the Cowan modulator. The one variable circuit factor considered is the resistance of the circuit supplying the carrier; this has the most marked effect on the modulating function, and is the design factor which is usually most independent of other circuit requirements. In all cases, as this resistance is increased, the e.m.f. of the supply is also increased to maintain a peak value of V of about 1 volt. To make the graphical part of the process reasonably compact, it is desirable to choose an interfering signal of frequency so related to the carrier that the repetition of the pattern of the distorted modulating function occurs every few cycles of the carrier. Some generality is lost by this, as confusion of coincident frequency components may occur; but a check of probable relative magnitudes can be made from the analysis of the ideal case, and the smaller of two coincident components will have to be ignored. In the present work, an interfering signal of frequency 2 5 times that of the carrier, and of -J-th the amplitude, was chosen. The pattern thus repeats every two cycles.

(1) INTRODUCTION The problem of determining the effect of unwanted signals mixed with the carrier supply of a ring or Cowan modulator is no new one; it has arisen chiefly in connection with synchronized carrier systems.1-2 The method of attack has been to assume an ideal modulator, in which the rectifiers are switched instantaneously from infinite to zero resistance and vice versa by the action of the carrier. In these circumstances the amplitude of the carrier has no effect on the operation of the modulator and the only effect of an unwanted signal is to alter the time of switching, i.e. to produce a phase modulation on the output signal. The analysis of this case is, therefore, quite straightforward, and a summary of it is given in the Appendix. In practice modulators are not ideal, and their practical performance can best be studied by means of the "modulating function," which is the factor by which the modulation process multiplies the input signal; it is, of course, a time-function. This has been discussed in previous papers, 3 ' 4 and it has been shown (2) ANALYSIS OF DISTORTED MODULATING FUNCTION that, although for an ideal modulator the modulating function Fig. 1 shows the modulating function, <0(/), and the disis a square wave, in practice it is never quite square. Its shape torted modulating function, <f>i{t), of an ideal ring modulator depends primarily on three factors: and of real ring modulators in which the carrier supply has (a) The resistance-voltage characteristic of the rectifiers. resistances of 1 000, 50 and 0 ohms, respectively. Although the {b) The impedance of the circuit in which the modulator is used. pattern of ^>x{t) repeats every two cycles, the second cycle is an odd image of the first, and therefore only the first is drawn. (f) The resistance of the circuit providing the carrier. The effect of an interfering signal in the carrier supply is to The distorted functions were all analysed by an automatic change the shape of the modulating function. A frequency- frequency-analyser designed and installed in the Admiralty analysis of the distorted function gives all the frequency com- Research Laboratories.* The resulting spectra are shown in ponents which multiply (or modulate with) the input signal; the Fig. 2. The accuracy of the amplitudes is probably much new frequencies introduced into the output, and the change of better than 1 0 % within one spectrum, but the error between amplitude of existing ones, can therefore be determined. one spectrum and another is possibly greater, but not exceeding To deal with the matter by strict mathematical analysis appears 20%. By comparison with the mathematical analysis of the Radio Section paper. Written contributions on papers published without being read ideal case, given in the Appendix, the various frequency comat meetings are invited for consideration with a view to publication. ponents were identified in terms of u> , the carrier frequency, and Dr. Tucker is at the Post Office Research Station.
[173]

174

TUCKER: THE EFFECTS OF AN UNWANTED SIGNAL MIXED WITH 10

10

! I -10 60 120
i i i

(a) -10 240 300 360 ^ 60 120 Wet 240 300 360

180

10

(c)

f
-10 0

at

^ *l(t)

-10 0"

60

120

180

240

300

360

60

120

180 Wet

240

300

360

Fig. 1.Modulating function with and without interference for various ring modulators. (b) Carrier-generator resistance 1 000 ohms. (a) Ideal. id)' Carrier-generator resistance zero. (c) Carrier-generator resistance 50 ohms. (a) The second cycle of coc is an odd image of the first cycle. The whole pattern repeats every 720.

H 10 ,08

1-2 10 0-8

1
01 02 0

I
G4 0-2

_L 1 1 1
1 2 3

1 1 i
8

I 9

4 5 6 7 Frequency/earner frequency

J_ 1 1 1 1

Frequency/carrier frequency 1-2 10 08 (.d)

u
10 08

04 02 0

0-4 0-2

J_

3 4 5 6 Frequency/earner frequency

|
3 4 5 r 3 Frequency/carrier frequency

Fig. 2.Frequency analysis of distorted modulating function for various ring modulators. (a) Ideal. (c) Carrier-generator resistance SO ohms. (b) Carrier-generator resistance 1 000 ohms. (d) Carrier-generator resistance zero.

THE CARRIER SUPPLY OF RING AND COWAN MODULATORS 110

175

the full two cycles. The analysis of this gives the spectrum of Fig. 4, and the amplitudes are also entered in the Table. Bearing in mind that the efficiency of a Cowan modulator is only fo(t)\J about one-half that of a ring modulator, so that for comparison purposes the figures in the Table should be doubled, it can be seen that the relative magnitude of the primary component (to) of the interfering signal is the same as in the ring modulator. 90 IKF 270 360 450 540 630 720 The appearance of even-order components is due to the unsymmetrical shape of the modulating function of a Cowan modulator Fig. 3.Modulating function with and without interference for Cowan in practice. modulator with carrier-generator resistance zero. The presence of the interfering tone in the carrier supply causes a small reduction in efficiency of the main modulation, but the 06 accuracy of the analysis of the modulating function is not good enough to detect this. For an interfering signal of | t h the 0-5 amplitude of the carrier, which is a much larger amplitude than any normally encountered, this reduction in efficiency is no more 04 than about 1 %, i.e. 0-1 db, and the effect may safely be neglected o> in practice. It can be seen from the Appendix that, for an ideal 03 modulator, the change in efficiency due to an interfering signal of amplitude x times the carrier amplitude is given by the change from 1 to J0(x); and for x = 0-2, J0(x) is 0-990. Careful a. laboratory tests have shown that the same order of change is obtained in practice. *>2 01 1 (3) GENERAL CONCLUSIONS TO BE DRAWN FROM THE CASES CONSIDERED 0 1 2 3 4 5 6 7 8 9 Fr^fjuency/carrier frequency Although only certain specific examples have been considered, Fig. 4.Frequency analysis of distorted modulating function of Cowan it is nevertheless possible to draw some general conclusions regarding the effect of an unwanted signal mixed with the carrier. modulator with carrier-generator resistance zero.

Urn

V. J

Table 1
FREQUENCY ANALYSIS OF MODULATING Ideal modulator Type of product Theoretical (Oc 3ft)c By analysis of modulating function Re* - 1 000 FUNCTION

Ring modulator Re = 100 .Re = 50

Cowan modulator

*^o 101 0-11 0-6 008

5o)C
llOc
CO

1-26 0-39 0-20

on
012
012

1-27 0-43 019 009 011

1-29 0-43 0-21 010 013

1-25 0-31

on
012

118 0-24 007

on
010 009 006 002

0-11
006 004

2toc co 2coc + co

4coc + ft)
4coc 6(Oc Scoc IOcoc co co ~~ co ~ co

on
010 012

009 007 011 008 007

on

on
0-12 007 012

on on
008 005 003

003 002

on
010 008

on

on
008 006

4coc

007 002

* Rc =r resistance of the circuit supplying the carrier. w, the interfering frequency, and the generalized results are shown in Table 1. It will be seen that the agreement of amplitudes between the two analyses of the ideal case gives considerable confidence in the practical method, and thus in the figures given for the real modulators. It will be observed that the primary interfering component (co) is almost independent of the initial shape of the modulating function, although the higher-order components become smaller as the shape becomes more rounded. Fig. 3 shows the functions </>0(/) and tf>x{t) for a Cowan modulator with a carrier supply of zero resistance. This has to be drawn for The effect of the rectifier characteristic is, in practice, not very large, as most rectifiers have the same kind of characteristic. The effect of circuit impedance is not likely to be very large over the range of variation usually met in practical equipment. The actual frequency of the interfering signal can have no effect, apart from possible coincidences of output components, since there is no part of the circuit which is concerned with the frequency, assuming non-reactive rectifiers. The amplitude of the interfering signal relative to the carrier is, of course, a factor affecting the results, but it is one for which allowance may easily

176 TUCKER: EFFECTS OF UNWANTED SIGNAL MIXED WITH CARRIER SUPPLY OF RING AND COWAN MODULATORS be made; it may be assumed that, by analogy with most nonlinear circuits, the proportion of primary unwanted component (co) (and of any other component involving co) in the modulating function is linearly related to the input amplitude of unwanted signal (co), the proportion of components involving 2co is related by a square law, and so on. This has been confirmed experimentally for the primary component. Thus the following conclusions have a certain general validity: (a) The primary modulating effect of an unwanted signal mixed with the carrier is, to a first approximation, independent of the resistance of the circuit supplying the carrier, even though this affects the shape of the modulating function very considerably. The amplitude of the unwanted component in the modulation is, relative to the wanted component, only about one-half of that in the carrier input. (b) Higher-order products are all much smaller when the resistance of the circuit supplying the carrier is decreased. (c) Ring and Cowan modulators appear to give the same results apart from the introduction of even-order components in the Cowan modulating function. (d) In the special casean important one in certain types of equipmentwhere the input frequency to the modulator is equal to the carrier frequency, the output signal of frequency w coc is derived from the main component co of the modulating function and also from the component co 2coc. These two components are of opposite sign, so that in the ideal modulator they exactly cancel, leaving no output of the difference frequency (co coc). As the resistance of the carrier supply is decreased, however, the amplitude of the co component remains constant, while that of the co - 2cot. component decreases, with the result that the output of the difference frequency increases from zero for the ideal modulator to a maximum for the real modulator with a carrier supply of zero resistance. This maximum value, relative to a main sideband, is about one-quarter of the amplitude of the unwanted signal relative to the carrier in the carrier supply circuit. (4) ACKNOWLEDGMENTS The author would like to thank Dr. G. E. R. Deacon, F.R.S., and Mr. M. J. Tucker, of the Admiralty Research Laboratories, for their courtesy and assistance in providing facilities for the graphical wave-analysis, and Mrs. M. West for carrying out all the calculation and drawing work associated with the investigation. (1)
TUCKER,

(2) Post Office Research Report No. 12492: "The Effect of Interfering Frequencies in the Carrier Supplies to Frequency-changers." (3) TUCKER, D. G.: "Rectifier Resistance Laws," Wireless Engineer, 1948, 25, p. 117. (4) TUCKER, D. G.: "Some Aspects of the Design of Balanced Rectifier Modulators for Precision Applications." See p. 161.
(5) BARBER, N. F., URSELL, F., DARBYSHIRE, J., and TUCKER,

M. J.: "A Frequency Analyzer used in the Study of Ocean Waves," Nature, 1946, 158, p. 329. (6) BELL, D. A.: "F.M. Communication Systems," Wireless Engineer, 1943, 20, p. 233. (6) APPENDIX Mathematical Analysis of Ideal Modulator with an Unwanted Signal in the Carrier Supply Let coc = Angular frequency of the carrier. co Angular frequency of the unwanted signal. x Amplitude of unwanted signal relative to carrier amplitude (assumed small). In the absence of co, the modulating function is given by
CO

<f>o(t) ~ - > sin mcoJ . . (m odd) In the presence of cot the modulating function is therefore
<l>\(t) IT/

4^11

4vil
im

> sin m\co rt + x sin (co r co)/l . . . (m odd) L

m \

4 = -{J(t(x) sin mct + Jy(x) [sin (2coc cS)t sin cot] + J2(x) [sin (3coc - 2co)/ + sin (2co coc)/] + sin 3cocr +
1/0(5JC)

(4coc - co)t sin (2coc + co)/] +

sin 5cocf + . . .

+ etc.} according to the well-known series6 for phase or frequency modulation. The Bessel functions are readily written down as numbers from suitable tables, and it is in this way that the figures in the second column of Table 1 have been computed.

(5) REFERENCES D. G.: "Carrier Frequency Synchronization," Post Office Electrical Engineers Journal, 1940, 33, p. 75.