U.D.C. 621.

375/6

Circuits with Time-Varying Parameters

(Modulators, Frequency-changers and Parametric Amplifiers)
By Professor D. G. TUCKER, D.Sc.
{Member) f

This paper is sponsored by the Institution's Education Committee.

Summary: It is shown that the theory of circuits with time-varying parameters may be presented in a manner sufSciently simple for undergraduate instruction, and that the basic conceptions of modulators with complex impedances and of parametric amplifiers may thus be made clear. Moreover, the treatment is general in nature, and so brings out the similarities and differences between circuits with time-varying resistance and those with time-varying inductance and capacitance. Non-linear effects are also touched upon.

1. Introduction

From Fig. l(c) (i) and Fig. \{b) (ii) respectively,

There is no doubt that, as with so many branches of E cos coqt = i(t).dt (5) engineering nowadays, the subject of circuits with time-varying parameters is often treated in such an advanced mathematical manner (see, e.g., the series dt (6) of papers in the I.R.E. Transactions on Circuit Theory, \ ) v March 1955) that it seems to be quite beyond the These equations are true for every instant of time. scope of undergraduate work. It is possible, however, The practical applications of these circuits are to present the basic concepts very simply, and this simple treatment is adequate for the solution of many mainly to rectifier modulators for time-varying important practical circuits in the steady state. resistance, and to parametric amplifiers for timeExamples of such circuits are amplitude modulators varying inductance and capacitance. and parametric amplifiers. i(t) There is nothing in Kirchhoff's Laws to prevent their I COS (ji application to time-varying circuits, and we therefore start with a statement of these laws for single loop v(t) r(t) and single node-pair circuits containing, in addition Z? v(t) krit) Ecos^tQj to ordinary impedances, elements of resistance, inductance or capacitance which are caused to vary (i) (ii) with time.
n

(a) Circuits with time-varying resistance r(t)

2. Kirchhoff's Laws for Time-varying Circuits

The basic circuits are shown in Fig. 1. The single node-pair circuits can be re-arranged, as will soon become evident, as exact duals of the single loop circuits. Thus the equations may be written in dual pairs, where Y = 1/Z, g{t) = 1/KOThus from Fig. \(a), (i) and (ii) respectively,
E cos coqt = r(t). i(

vlU E cos u>nt (i)

5>L(t)

(ii)

(b) Circuits with time-varying inductance L(t)

/ cos coqt=Y. v(t) + g(t). v(t) From Fig. l(b) (i) and Fig. l(c) (ii) respectively,
E cos coqt =

(1) (2)
I COS CJnt A \C(t) \'i(t)

^ [L(0. i(

(3) (4)

Z 5

v(t)

(t) Sic(t)

/ cos coqt = Y. v(t) + ^ [C(0. v(t)-]

(i) (ii) (c) Circuits with time-varying capacitance C(t)

t Electrical Engineering Department, University of Birmingham.

Journal Brit.I.R.E., March 1963

Fig. 1. Basic circuits. 263

D. G. TUCKER

Now, in practice, the time-varying element will vary in a periodic manner, and we call the fundamental angular frequency of this variation cop. Although it is not necessary, it is convenient in order to simplify the mathematical working (and also reasonably representative of practice) to restrict the Fourier series representing the time variation to cosine terms. Thus K0= Z Z rncosnot)pt
n= o0

(7) (8) (9)

0(0= Z gncosncopt
n= 0 oo

it seems preferable to use the cosine notation. Moreover, the complex symbolism using a single exponential for each frequency relies on the superposition theorem for its justification, and clearly applies only to linear circuits. It is surely very wrong, when it is avoidable, to teach students to think in a system which is restricted to linear circuits when so many real problems involve non-linearity. Often, too, the working of a problem is greatly complicated by the use of exponentials. One other point requires noting before proceeding to the expansion of eqns. (l)-(6). The term Z.i(t) is handled by using at each frequency (coq+mcop) the value Zm which Z has at that frequency, thus Z.i(0= Z z mimcos[(c,
oo

L(0= Z Z Lncosncopt
n=0

(15)

L(f)

nt,ouJ
oo

cosnfi)

pf

(10) (11) (12)

and correspondingly for Y.v(t).

3. Expansion for Time-varying Resistance

C{t) =

YJ
n=0

Cn cos n(op t I c o s na>r

C(t)

Note that, except for square-wave variations of r(f), the gn are not explicitly related to the rn; similarly for (1/L) and L and for (1/C) and C. Although the term "elastance" with symbol S is often used for (1/C), there seems to be no corresponding term and symbol for (1/L) in common use; we shall therefore continue to use the expressions (1/C) and (1/L). Evidently the current i(t) and the voltage v(t) must, in general, contain components of all possible modulation frequencies na>p coq. They may therefore be written

n = 0 \C>

We shall deal with eqn. (1). Obviously eqn. (2) is handled in an exactly similar way. The product r(t).i(t) gives rise to a new series of terms, but without introducing any frequencies not already in i(t). Thus
rn cos ncop t. im cos [((Qq + m(op)t + 0m] = n im cos {[coq + (m + ^cop] t + 9m} +

+JErnimcos{[coq + (m-n)(Op]t+9m} (16) It is important to note that the phase angle still carries the same subscript as the current magnitude, i.e. although the frequency has been changed, the phase angle and magnitude still correspond. This means that it is quite unnecessary to retain the phase angles if the currents im are regarded as vector quantities. Since the eqn. (1) is true at all instants of time, it (13) i(0= Z must hold for each individual frequency taken m = oo separately. In the statement of equilibrium for each (14) frequency, the cos (coq+mcop)t will appear in every term and may be cancelled out. The expansion of In these expressions the frequency has been written eqn. (1) thus appears: coq+mcop, although in practice cop often exceeds coq, so that for m negative this frequency is then negative. at frequency coq: It is important, however, to express the frequency in 2 this way, as it avoids the difficulty of reversal of (17) phase angle which arises if the frequency is taken as at frequency (oq + a>p: mcop coq (see Ref. 1). If, in a practical problem, Z is i+2+... specified at frequencies mcopcoq, then in the equations 0 = . . . + i r 2 i _ 1 + i r 1 (18) used here, the conjugate of the specified value must be at frequency co co : q p used to give the appropriate value for frequencies coq-mcop. 0 = ... i + 2 + ... (19) It is possible, and some would say preferable, to use complex symbolism, i.e. exp(jcot), in place of the and at frequency a>q+mcop(m # 0): cosine notation.2 The present authorfindsit hard to see any real advantage in this, however, and as students 0 = ...+ir |M ,io+ir| m _ 1 ,i +1 + ... (20) probably think more easily in terms of cosine waves
264
Journal Brit.l.R.E.

CIRCUITS WITH TIME-VARYING PARAMETERS We thus have an infinite number of equations, each with an infinite number of terms. Explicit solution is not generally possible. But as soon as we can make some restrictions, solution becomes possible. A very simple example of such restrictions is to specify that Z is infinite except at the two frequencies coq and (coq + cop). This might, in practice, be a rectifier modulator of the "series" type with filters for source and load terminations, although it happens to be an inefficient arrangement and not recommended for general use. It is chosen here only for its analytical simplicity. Then all currents are zero except i0 and i+1. Equations (17) and (19) then become E = (Z0 + r0)i0 + $r1i + 1 (21) (22) + r o )i + 1 giving the solution
(23)

It is clear that knowledge of the absence of evenorder products is a great help in this problem, as it is indeed in many others.
4. Expansion for Time-varying Inductance or Capacitance

4.1. Solution of Equations (3) and (4) Here we shall deal with eqn. (3). It is clear that the same working applies also to eqn. (4). Now
(28)

From eqn. (9), dL(Q _ dt and from eqn. (11), di{t) ,

dt
m = oo

nLnsinncopt

(29)

It should be noted that although the two eqns. (17) and (19) are sufficient for solving for i+1 (and i0), the other equations do not disappear, as some finite terms (including Zmim) remain in each. An example of a somewhat different kind is to specify that Z is a constant pure resistance (i?) at all relevant frequencies, and that r(t) is a square-wave. Then it can be shown by various methods1'3 that no even-order products exist other than i0i.e. im = 0 when m is even and not zero. Equation (17) then becomes
)o+i Z
m = l , 3, 5,-

)im sin l(co + mcop)t + dn

(30)

Thus eqn. (3) becomes

E cos coqt =
m = oo

Z m im cos [(coq + mcop)t + 0 M ] -

nLnsinno)ptx

m = oo

rm(i+m+i_m)

(24)

Z Lncosncoptx
n=0

and eqn. (20) becomes, for m odd, Q = (R + ro)im+$rmio (25) (31) These give immediately a solution for any particular It will be observed that in all resultant terms, 6m is current, e.g. always associated with im as in Section 3, but there are i+i= "^ri (26) now sine as well as cosine terms. This can be taken care of by the use of the " j " operator with the sine r o ) 2 -i m=l, 3, 5, terms, and a set of equations, in which only the vector Now as r(t) is a square-wave, coefficients appear, can then be formed, on the lines of, but more complicated than, eqns. (17)-(20). oo (27) If eqn. (31) is expanded, and the terms at each m = l , 3, 5,- separate frequency are grouped together, it is readily seen that the following system of equations is obtained: so that the solution is explicit.
at frequency coq:

E= ... +ij(oqLli_1+(Z0+icoqL0)i0+ q at frequency a>q + cop: 0= ... +iJ(o,+Op)L 2 i_ 1 +ij(o fl + Op)L1i0+ [Z and at frequency coq+mcop (m # 0 ) : 0 = . . . 4- i JK+ mco p )L M i0 + \]{(oq + mcop)L]m_

oi+2+...

(32)
(33)

+ mcop)Lt im _ x + [Zm +](coq + mo p )L 0 ] im+ (34)

265

March 1963

D. G. TUCKER

As with the time-varying resistance, no explicit solution is possible for the general case. Particular cases in which specified restrictions are imposed are, however, soluble. A most important such case, which forms the basis of the parametric amplifier, is that in which Z, in Fig. 10) (i), is infinite at all frequencies
except coq and (coq + cop)(or (coq-cop) as will be discussed later). Then i0 and i+1 are the only currents which are not zero. Using eqns. (32) and (33) we
1i+1

4.2. Solution of Equations (5) and (6)

(35)

j n o r ( j e r t 0 r e t a m t n e wor king for a timeinductance, we shall consider eqn. (6). clearly eqn. (5) is worked out exactly similarly. It is a l s o v e r y important to work as often as possible i n t e r m s o f admittance rather than impedance, as s t u d e n ts seem to find this difficult, and eqn. (6) g i v e s prac tice in this. Using the expressions given m eqns. (10) and (14), eqn. (6) may be written
Here? varyjng

and 0 = iKco + o>j,)^i i 0 + Cz+1 + J(w + w P ) L o] * (36) whence

I cos coqt =
m = oo

cos[(coq

)t + 0 m ]

+ n=o\L Z (7

(37) at frequency co
q:

equations for each frequency, and regarding the coefficients vm as vectors, we obtain:

E x p a n d i n g t h i s i n t o s e p arate

2 and at frequency a>q + mcop (m 0):

(39)

J/l
L/0\'

(40)

It can be seen at once that this is quite different from eqns. (32)-(34); in particular, different frequencies occur in the coefficients of the various terms of one equation, instead of just the one frequency. Nevertheless, this set of equations is used in just the same way as before. A particular case, which can also be used as the basis of a parametric amplifier, is that in which Y is infinite at all frequencies except a>q and (coq + cop), so that only v0 and v+t are finite, and only two equations are required to solve for v+x. This case gives )
q\LJ0

and (42) 2 j whence

v +1 =

o+

ja),WoJ
Journal Brit.l.R.E.

266

CIRCUITS WITH TIME-VARYING PARAMETERS

TUNED TO U)q AND GIVING Z - o o E L S W H E R E TUNED TO coq+(vp GIVING Z=oo AND

Now eqn. (21) gives us

ELSEWHERE

Vand since E+1Z{ + 1 / + 1 = Z + 1 / + 1 , then, from eqn. (22), E+1-Zi> + 1i+1 = -ir1io-roi+1

(45) so that

-ir.E
and

(46)

(47)
TUNED JOOOq AND TUNED TO cvq+ U)p AND GIVING f=oELSEWHERE GIVING Y=oo ELSEWHERE

(b)

Fig. 2. Time-vafying element in a circuit with (a) only two non-zero currents (b) only two non-zero voltages.

(48) Therefore for a conjugate match at 2,2', R+1 = r0and *ri(Ro + r)2 Xr2Y~ irjXo (49)

5. Further Consideration of Circuits Restricted to Two Non-zero Currents or Voltages The examples of specific solutions which we have taken above have restricted the circuit to have either (a) non-zero currents at only two frequencies, or (b) non-zero voltages at only two frequencies. It is thus convenient to consider the circuit in its practical form as having separate parts for the input and output signals respectively, as shown in Fig. 2(a) and (b). It is clear that from the point of view of the previous analysis, since the two parts function at different frequencies, they may be superposed to give the single loops or single node-pairs of Fig. 1. Then Z o and Z + 1 are merely the values of Z at frequencies coq and (coq+a>p); and similarly for Y. The tuned circuits shown are symbolic only, to indicate a practical way of obtaining an approximation to the required conditions. We shall now show how the matching conditions and conversion loss or gain may be determined, and draw some general conclusions regarding these circuits. 5.1. Time-varying Resistance Let us assume first of all that Z o = R0+jX0 is given, and that we require to find the value of Z + 1 = R+1+jX+1 needed to give a conjugate match (i.e. maximum power transfer) at the terminals 2,2' in Fig. 2(a). The circuit to the left of 2,2' may be represented as a generator of e.m.f. E+1 and internal impedance Ztt+1 at the frequency coq + cop. Obviously then, if we find Z i + 1 , we have to make Z+x Z * + 1 .
March 1963

(50)

When this match is obtained, the power in the output load

=

F-+

o[(Ro

+ r 0 ) 2 + X$] - jr2(R0 + r 0 )}

(51) and for the given Z o and r(t), the optimum conversion power-loss ratio _ power available from signal source at a>q power in load at coq + cop ^~ (52)

If the load Z+ 1 were given, and we had to find the value of Z o required for maximum power transfer, then we would calculate the input impedance at frequency coq looking from the left into terminals 1,1' as E Zoio

(53)
267

D. G. TUCKER

We would then choose Z o to be the conjugate of Zi0. Equations (47) and (53) make it clear that the circuit is symmetrical, so that if Z o and Z + 1 are both adjustable, the conversion loss will clearly be a minimum when Xo = X+1 = 0 and The minimum conversion power-loss ratio is then

Working with the circuit of Fig. 2(a), and using time-varying inductance, we have, from eqns. (35) and (36), the following values for the effective generator at frequency (coq + cop) seen to the left of terminals 2,2': Z0+jo) Lo (55)

(54) (56) It happens that the value of optimum circuit resistance thus obtained is usually inconveniently For a given Z o , therefore, the value of high; if r(t) is a square-wave variation between Z+i = R+l+}X+l values rf (e.g. forward resistance of a rectifier) and rb (e.g. backward resistance of a rectifier), then to give a conjugate match at 2,2' (and hence maximum Ro = R+x ~ 0-39 rb if it is assumed that rb^>rf. The power transfer) is Z* so that +1 conversion loss may then be as low as 8 -9 dB, but if the circuit resistance has to be reduced to some small (57) +1 fraction of rb for practical reasons, the loss becomes high. If the time-varying resistance were, in fact, the opening and closing of a perfect switch, then it could and be shown that the loss becomes infinite. The method can, of course, be applied (but with more complexity) to other frequency-changer arrange(58) ments where more than two currents or voltages exist, and where lower losses may be obtained. For With this match, the power in the load a tabulation of such arrangements and their mini= \E+l\2JSR+1 mum losses, see reference 1. E2 The minimum conversion power-loss ratio it is (59) possible to have in any circuit with time-varying resistance is unityi.e. no loss at alland this occurs so that the power gain in a single-loop circuit when the following conditions _ coq+cop are met: (60) (O (a) r(t) is a square-wave function switching between Evidently, therefore, if cop^>coq, this gain is very values of zero and infinity, large. This is the basis of the parametric amplifier (b) Z = Ro at frequency coq, R+l at frequency (coq + a>p), zero at all other odd-order product of the so-called upper-sideband "up-converter" type. frequencies, and infinity at all other even-order It is an amplifier so long as the wanted output signal is of higher frequency than the input signal. The power product frequencies, gain is derived, of course, from the work done in 2 (C) Ro = (4/7T )i?+1. varying Lcommonly called the "pumping" action. There is, of course, a corresponding dual circuit. The gain given by eqn. (60) is obtained on making This zero-loss condition is hardly a practical one, as Z + 1 a matched value for any given Z o . It is interesting the impedance requirements are almost impracticable; that no optimization of Z o and Z + 1 in relation to L(t) nor is it easily deducible from the general theory is required. In other words, if Z o is given, then a given here. But it is of basic theoretical importance choice of Z + x to give a conjugate match according to in showing that it is, in principle, possible to convert eqns. (57) and (58) will automatically give the maxiall available signal power to another frequency; and mum possible power gain of (coq + cop)/(oq. This is it can be readily realized in practice with a ring quite different from the behaviour of the circuit with modulatori.e. with a 3-loop circuit. The matter is time-varying resistance discussed earlier, since in fully discussed by Belevitch.4 that circuit the minimum conversion loss was obtained only when Z o and Z+1 were both chosen to have a 5.2. Time-varying Inductance or Capacitance particular, unique pair of values in relation to the 5.2.1. Use as a frequency-converter Fourier coefficients of r(t); mere conjugate matching with an arbitrary choice of Z o or Z + 1 gave, in general, The same method is used as that discussed in the greater loss. previous section.
n

[ro + V ^ F ? ] 2 / ^

268

Journal Brit.I.R.E.

CIRCUITS WITH TIME-VARYING PARAMETERS For two extreme cases, interesting results are obtained: (a) Xo = -coqL0 and X+1 = -(coq+cop)L0 (This is the condition usually assumed in the literature.) Then R0R+1 = icoq((oq + (op)L2 (61) is the condition for maximum gain, (b) Xo = 0. Then for maximum gain, (62) and Therefore and

(66)

mi-+km
(67) This match gives a power gain of (coq + cop)/coq as for the previous circuit. It will be found, indeed, that all the four circuits of Fig. l(b) and (c), give the same performance, and expressions for matching, when proper allowance is made for the dual relationships. This means that an up-converter (and indeed, the negative-resistance parametric amplifier discussed below) can be made with time-varying inductance or capacitance, and with the circuit impedance made infinite or zero at the unwanted frequencies. It is extremely interesting that the maximum power gain, as given by eqn. (60), is independent of the amount of inductance variation, as represented by L or (1/L)i. One would at first have supposed that as the variation was reduced, so also would the power delivered into the output circuit be reduced. But, taking the circuit of Fig. 2(a), the optimum resistance termination (R+1) is seen from eqn. (57) to be a function of Lu so that as Lt is reduced, so also is R+1. Thus, as we approach the limit, variations in the inductance (although small) are opposed (according to Lenz's law) by increased forces due to the almost short-circuited condition; this makes it possible for the power absorbed in producing the variations to remain constant in spite of the reduced variation. There is, of course, a discontinuity at the limit where Lt actually becomes zero, since no gain can then be produced.

i
coqL0

coq X.

2 2 + co qL \

(63) (64)

con

Taking case (a), as it represents a usual practical arrangement, we can examine the effect of changing the signal frequency by a small amount. Since in this case X0 + coqL0 ~ 0 for small changes in coq, we can write eqn. (58) as

L\(X

(65) Since Xo is due to a capacitance (or, at any rate, its circuit is dominantly a capacitance), a small increase in coq will cause the expression between the large brackets to diminish in magnitude, so that the value of X+ x required for conjugate matching becomes less negative. But since X+x is also necessarily due to a capacitance, this is just the way the value of X+l would tend to alter due to the change of frequency. Thus an approximation to conjugate matching is obtained over a relatively wide frequency band. This means, therefore, that the up-converter is inherently a wide-band device.

Manley and Rowe5 have shown that some general power relations exist in circuits with time-varying capacitance, and one of their results is that in this particular circuit the ratio of power absorbed in the It can be shown that if the inductance is associated load at frequency (coq + cop) is always (coq + cop)/coq with a resistance representing its lossesas in practice times the power absorbed from the signal source at it must bethe gain reduces continuously to zero as frequency coq, irrespective of matching. The power Lx is reduced. gain, however, relates the load power to the available signal power, not to the signal power actually absorbed, 5.2.2. Negative-resistance parametric amplifier and so to obtain the gain given by eqn. (60), matching It is evident that if the output frequency were is required. coq cop instead of coq + cop, there would be less gain, If the same process (but using dual parameters) is and there could indeed be a power loss instead of a applied to the circuit of Fig. 2(b) with time-varying gain. But this case can be exploited, nevertheless, to inductance, working with eqns. (41) and (42), and give another kind of parametric amplifier. Suppose putting Yo = G0+jB0, then we find that for a con- then that the system has Z infinite except at frequenjugate match at 2,2', we require Y+t = G+1+jB+l cies coq and a>q cop, according to Fig. 2(a), with timewhere varying inductance, and that we calculate the input
March 1963

269

D. G. TUCKER

impedance (Zi0) seen from the signal source terminals 1,1'. Equations (35) and (36), adapted for frequency coq-cop, give
_
10

E-i0Z0 _

"

single-node-pair modulator.1 Conditions for modulators to have an input impedance which is not timevarying can be specified10 on the analogy of Zobel's constant-resistance networks. It is always important to remember, however, when making circuit transformations, that products such as Z.r(t), which frequently arise, are not straight products; since Z can be regarded as a function of (d/dt), it is an operator, operating on r(t). This frequently makes it difficult to make any use of the transformation.

(68) 7. Non-linear Effects in Circuits with Time-varying Parameters It is thus immediately clear that if a>p ;> a>q, then the The treatment so far given of time-varying parareal part of this input impedance is a negative resismeters has merely assumed that the variation r(f), tance : L(t) or C(t) has been produced by some unspecified icoq(cop-coq)L21R_1 ......(69) external agency. It is necessary, therefore, to give students some idea of how the variation is produced in practice, this being generally by the use of largeIf then a load circuit operating at frequency coq is amplitude local signals which produce a time-varying connected across terminals 1,1', its resistive compo- bias in a non-linear element of r, L or C. The question nent may be arranged to be very nearly cancelled out will then inevitably be asked as to whether the inforby the negative resistance Ri0. A large output may mation-signal voltages or currents, which also appear then be maintained by applying only a small power across or in the non-linear element, have any effect on from the signal source, and so a power amplifier has the time-variation of the element, and whether they are been obtained. The power gain can clearly be made as subject to non-linear distortion due to the non-linearity high as desired; if it is made infinite by making Rl0 of the element. The answer must be given that both completely cancel out the load resistance, then clearly effects do occur, and give rise to non-linear distortion. a self-oscillator is obtained. A general treatment of this matter is prohibitively It is evident that the negative resistance amplifier, complicated (even if possible) for undergraduate by removing most of the resistance component of any courses. But with some special simplifying assumptuned circuit in the signal path, is inherently a narrow- tions, an insight into the nature of the effects may be bandwidth device; in this respect it contrasts markedly given in terms which are acceptable to undergraduates with the up-converter, which is inherently a wide- who ha\e had a course in communication systems, band device, as previously explained. and some previous contact with non-linearity.12 One Exactly corresponding results are obtained with the group of assumptions is as follows: circuit of Fig. 2(b), and with a time-varying capacitance. For high-frequency use, the capacitance, (a) the circuits are purely-resistive with time-varying resistance amplifier is the more suitable in practice.6'7 (b) signal voltage or current small compared with bias An account of the history of parametric amplifiers, 8 wave with a very large bibliography, is given by Mumford. (c) nominally square-wave variation of r(t) produced either 6. More Complicated Circuit Configurations (i) by cosine bias wave applied to bilinear rectifiers In the case of the time-varying resistance, which (i.e. rectifiers with a constant forward resistance represents the important practical circuits known as and a constant back resistance, switching at rectifier modulators, circuits comprising more than zero applied voltage). one loop or one node-pair are common9'10e.g. the or (ii) by square-wave bias applied to any kind of nonring (or lattice) modulator. However, with due care, linear resistance. all the usual circuit theorems and transformations can be applied to them,11 and in some circumstances very Assumption (c) (i) permits the development of simple equivalent circuits can be found. For instance, the idea of a square-wave time-variation, phasewith the assumption of a local carrier (or switching) modulated by the difference-frequency between signal oscillation having "half-cycle" symmetryi.e. con- and bias wave, so that either r(t), or the corresponding taining no even-order harmonicsthe ring modulator time-varying transfer function (j)(t), may be written can be shown to be equivalent to a single-loop or approximately as
270
Journal Brit.I.R.E.

CIRCUITS WITH TIME-VARYING PARAMETERS 8. References rcos[(2n-l){co t-x^inicop-co^t 1. D. P. Howson and D. G. Tucker, "Rectifier modulators
-x2sin(cop-CQq2)t...}]

fco + ^i Z -^
n

=i

in i

(70) where coql, coq2, etc. are various frequency-components of the information-signal of amplitudes xx, x2, etc. relative to the bias wave. The output signal consists basically of the product of this function and the input signal elcoscoqlt + e2coscoq2t + (71)

2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

and it is clear that an infinite range of harmonics and intermodulation products is produced, with amplitudes dependent on both the relative signal amplitudes (x) and the order of modulation involved {In 1). A full account of this method is available elsewhere.13 Assumption (c) (ii) enables the lattice time-varying network (e.g. the ring modulator) to be regarded13 as a non-linear but constant lattice followed by a reversing switch operating at frequency cop/2n. The non-linear lattice can have a transfer function represented by a power-series, and the non-linear distortion products produced by this are easily calculated. They are then all multiplied by the switching function cos (2n - I K t to give the output spectrum. It is clear these two processes (i) and (ii) give different kinds of result, and give a good indication of the complexity of non-linear analysis without introducing any processes beyond undergraduate level. (72)

12. 13.

with frequency-selective terminations", Proc. Instn Elect. Engrs, 107B, p. 269, May 1960. A. P. Bolle, "Application of complex symbolism to linear variable networks", Trans. Inst. Radio Engrs {Circuit Theory), CT-2, No. 1, p. 32, March 1955. D. G. Tucker, "Elimination of even-order modulation in rectifier modulators", / . Brit.I.R.E., 21, p. 161, 1961. V. Belevitch, "Theorie des Circuits Non-lineaires en Regime Alternatif", (Libfaifie Univefsitaire, Louvain, 1959). J. M. Manley and H. E. Rowe, "Some general properties of non-linear elements", Proc. Inst. Radio Engrs, 44, p. 904, 1956. G. D. Sims and I. M. Stephenson, "Parametric amplifiers", Discovery, December 1960, p. 528. L. A. Blackwell and K. L. Kotzebue, "Semiconductor-diode Parametric Amplifiers", (Prentice-Hall, Englewood Cliffs, N.J., 1961). W. W. Mumford, "Some notes on the history of parametric transducers", Proc. Inst. Radio Engrs, 48, p. 848, 1960. D. G. Tucker, "Modulators and Frequency-changers", (Macdonald, London, 1953). D. G. Tucker, "Constant-resistance modulators", / . Brit. I.R.E., 21, p. 161, 1961. D. P. Howson, "Some Applications of Network Theorems to Linear Circuits with Time-varying Resistance", Electrical Engineering Dept., University of Birmingham, Memorandum No. 40, 1959. D. G. Tucker, "Non-linear circuits: a course for undergraduates", Bull. Elect. Engng Educ, 26, p. 62, June 1961. D. G. Tucker, "Intermodulation distortion in rectifier modulators", Wireless Engineer, 31, p. 145, 1954.

Manuscript first received by the Institution on 2nd March 1962 and in final form on 14th December 1962 {Paper No. 800/Ed9).
The British Institution of Radio Engineers, 1963

March 1963

271