1954 PROCEEDINGS OF THE I-R-E 1169

Generalized Equations for RC Phase-Shift Oscillators *

SOL SHERRt, SENIOR MEMBER, IRE
Summary-General solutions are obtained for three- and four- This equation defines the frequency of oscillation and
section phase-shift networks. These solutions are reduced to design the gain required to produce oscillation. The frequency
equations for a number of specific circuit configurations, and the of oscillation, "F" is determined by setting the imagi-
complete solution is given for a four-section phase-shift oscillator,
including effects of plate-load resistance, and input-miller capacity. nary, or odd-order terms equal to zero, which satisfies
the condition that the result be a negative real quantity.
1fl C PHASE-SHIFT OSCILLATORS exhibit sev- The required gain, "A" is found by substituting the
eral useful characteristics, foremost among which values of Z, derived from the frequency solution, into
are cost, weight, and space reduction. The fre- the real part of the transfer characteristic and solving
quency-determining components of these oscillators are for the loss of the network. This loss is what must be
relatively tiny and inexpensive in comparison to the compensated for by external gain in order to satisfy the
coils and capacitors of conventional resonant circuits, requirement that the entire circuit be lossless.
the difference being particularly striking in the low If the substitutions ZA=Zs=Z6, and Z2=Z4=Z6 are
audio-frequency region. This type of oscillator is being made, the general equation reduces to
used with increasing frequency by designers who wish
the advantage of reduced weight and cost while main- e Z23
taining good waveform and stability in the audio range. (2)
E Z23 + 6Z22Z1 + 5Zl2Z2 + Z12
The subject of this class of oscillators has been treated
rather extensively in the literature.'-6 However, the treat- which is the one derived by Gamertsfelder.7 Further
ment has usually been in terms of a single type or re- substitutions of Z = -jx, and Z2 = R, will give the re-
stricted application, with the resultant lack of general- sults
ity. This article proposes to develop completely gen-
eralized formulas for phase-shift oscillators, then reduce F =2rRV (2A)
these formulas to various specific cases, some well known 2 7rRC-/6
and some new. The analysis will be restricted, for prac- A = -29, (2B)
tical reasons, to three- and four-section networks.
which are the ones derived by Ginzton and Hollings-
worth.2 Similarly letting ZA=R and Z2 = - jXc will give
IN E -s 1.
- e out 2r6
2 7,RC
(2C)
Fig. 1 ;-Generalized three-mesh network. which was derived by Ginzton and Hollingsworth2 and
The minimum number of rc sections with which it is A = -29,
possible to achieve oscillation is three. This network is (2D)
shown in Fig. 1. which is the correction made by Blanchard3 of the Ginz-
Solving for I3, we have ton and Hollingsworth result of A = -5.

EZ2Z4
13 =
(Z4 + Z6 + Z6) (Z1 + Z2) (Z2 + Z3 + Z4) Z22(Z4 + ZS + Z)
- -
Z42(Z1 + Z2)
and e=I3Z6.
The transfer characteristic is defined by
e Z2Z4Z6
E (Zl + Z2) (Z2 + Z3 + Z4) (Z4 + Z6 + Z6) -
Z22(Z4 + Z5 + Z) -
Z42(Z1 + Z2) (1)

Decimal classification: R355.914.31. Original manuscript re-

A further substitution of Z3 = KZ1, Z5 = K2Z1, Z
*

ceived by the IRE, September 17, 1953; revised manuscript received,

February 10, 1954. = KZ2, Z6 = K2Z2 results in the general formula
t General Precision Laboratory Inc., Pleasantville, N. Y.
I Britton Chance et al., "Waveforms," McGraw-Hill Book Co.,
New York, N. Y.; 1949.
e 1
2 E. L. Ginzton and L. M. Hollingsworth, 'Phase shift oscilla- (3)
E Z13 (3K+2) Z12 (3K2+2K+1) Z1
tors," PROC. I.R.E., vol. 29, pp. 43-49; February, 1941.
3A. Blanchard, "Correction to phase shift oscillators by Ginzton Z23
K-+
K
Z
Z22 KI
K2
+
Z2
and Hollingsworth," PROC. I.R.E., vol. 32, p. 641; October, 1944.
4R. W. Johnson, "Extending the frequency range of the phase
shift oscillator," PROC. I.R.E., vol. 33, pp. 597-603; September, 1945. This network is shown in Fig. 2. The specific results
p. G. Sulzer, "The tapered phase shift oscillator," PROC. I.R.E, derived by Johnson,4 namely
vol. 36, pp. 1302-1305. October, 1948.
6 W. C. Vaughan, "Phase shift oscillator," Wireless Eng., vol. 26,
pp. 391-399; December, 1949. 7 B. Chance, op. cit., p. 111.
1170 PROCEEDINGS OF THE I-R-E July
1 EZ-Z4Z_,6
(3A) i4=
a(Z2+Z3+Z4) (Zl+Z2) -Z22a - (ZI+Z2) (Z6+Z7+Z8)
27rRC/3+ - +- + (2')G ±1)
where
and a = (Z6 + Z7 + Z8) (Z4 + Z5 + Z6) -Z62,
R1 -11 4 and the transfer characteristic is defined by
A = -8--- + _- + 8
e 1 _ =~~~~~~~
(6)
E (bcde') Z6(de') Z4(be') Z2(bc) Z2Z6
-( 1) (+ 2) (3B)
Z2Z4Z6Z8 Z2Z4Z8 Z2Z6Z8 Z4Z6Z8
+
Z4Z8
can also be found from (1) by making the appropriate where
substitutions
b = (Z6 + Z7 + Z8)
Z, = (R1 - jxc), Z3 = Z6 = - jX,

C = (Z4 + Z6 + Z6)
Z2 =
Z4 = ZB R.
d = (Z2 + Z3 + Z4)
and
el = (Z1 + Z2).

Fig. 2-Tapered three-mesh network.

Any other desired combination of impedances may

be used and the specific formulas derived from the gen-
eralized one given above. For example let Z2=Z4=Z6, Fig. 4-Generalized four-mesh network.
and Z3waZ5. The network is shown in Fig. 3. The equa-
tion becomes
This equation defines the frequency of oscillation
e 1 -____________ (4) "F" and the gain "A" necessary for oscillation in the
E Z1 ZlZ3 ZlZ32 Z3 Z32 same manner as (1) does for the three-section network.
3- +4 - + + 1 + 3-+ -z If
Z2 z22 Z23 Z2 Z22
Z1 = Z3 = Z6 = Z7,
and
Z2 = Z4= Z6 = Zg,
then
Fig. 3-Shunt R three-mesh network. e 1
(7)
E (f2g) 2Z2(fg) Z2f2 +-
- 1
Equating the odd-order terms to zero, substituting Z24 Z23 Z2 3
Z2 = Z4-=Z6 = R, where
Zi = - jXci, f = (Z1 + 2Z2)
and g= (Z1 +Z2).
Z3 = Z5 = - jXC2, Let Ko = Z,/Z2 and simplify, then
and simplifying, gives e 1
(8)
1 E K4 + 7Ko3 + 15Ko2 + 1OKo + 1
F=2i7rRV\3 (C22 ± C1C2) (5)
and equating odd-order terms to zero gives
which is equivalent to the result given by Ginzton and
Hollingsworth.2 Formulas for any combinations of im- V10
Ko = /-
pedances may be derived by proper substitution into the
general formula.
Let
The general four-mesh network is shown in Fig. 4.
Solving for i4, we have =
jxc,
1954 Sherr: Generalized Equations for RC Phase-Shift Oscillators 1171

and 2(Zl' + 2Z2)(Z1' + Z2 + R')

Z2 = R, z22
then
(Z1' + 2Z2)2
F= (9) Z22 +1.
10
2irCR
V7 Let Z2 = R and
which is the formula given by Ginzton and Hollings- Z1,
-- = K1,
worth.2 Substituting for Ko in the real terms, we have Z2
the result,
100 150 then
A = -- -
+ --
18.4 1 = (9A) -

E RI
49 7 -= K14 + 17+-K13+ t15+ 6 -K12
as the minimum gain required for oscillation. This cor- eR R,
responds to the correction of the Ginzton and Hollings- R'\ R'
worth2 result given by Blanchard.3 In addition, if the + 10 + 10R )Kl +4-+ 1.
substitutions ZA = Z3 = -jXc, Zf = Z7 = -jXc,l and Z2
=Z4=Z6=Z8=R are made, then the final correction Equating j terms to zero and substituting Z1'=-jX,
by Blanchard,3 which is K, = -jX,/R gives
C) /C 2 \ C
9 -+ 114 -+ 352~~+ 342 ± 4 2
(9B)
C
(4 + 3 -

can be derived. 1
If we wish to find the effect of the vacuum tube in the * (10)
complete circuit shown in Fig. 5(a) we may use the R 1OR(1+-)
equivalent circuit of Fig. 5(b) and make the following
substitutions: V7R+R'
Z= R' + Z1', If we wish to assess the effect of input and Miller
Z2 =
Z4 =
Z6 =
Z8, capacity we let Z1=Z3 = Zb = Z7, and Z2 = Z4 = Z6.
and The equation becomes
Z3 -Z6 = Z7 = Z1'. E Z14 6Z13 1OZ12 4Z1 Z13
=___+ +
+ 4_

Then e Z23Z8 Z22Z8 Z2Z8 Z8 Z23

E (Z1' + 2Z2)3(Z1' + Z2 + R') 5Z12 6Z1
e Z24 22 Z2
Substituting
-jRXc1
Z8-~~
(a) R - jXc1

where Ci=input and Miller capacity,

z1 =
I-i
and
Z2e=Ro
(b)
and equating j terms to zero we get
1
(11)
2rCR (C1 + C)10
C1 + 7C
Fig. 5-(a) Vacuum tube with generalized four-mesh network. (b)
Vacuum tube equivalent circuit with generalized four-mesh net- If (R1CI/RC) <<1, this may be combined with the
work. previous result to give
1172 PROCEEDINGS OF THE I-R-E July
1 (4K3 + 3K2 + 2K + 1) Z1
F= (12) + - + 1.
- Cl R') K3 Z

2irCR1/ ( +- Equating odd terms to zero gives

1 7 (1+ Cl + R' /4K3 + 3K2 + 2K + 1
Z,
which is the complete solution for this circuit.8 If Z2
=
1 K2(4K + 3)
(13)

(R'C1/RC) 0.01, the error in calculation will be less and

than 1 per cent.
Z14 (6K2+ 6K + 3) Z12
A =-+ -- Z22
- + 1.
Z24 K2

Substituting for Zl/Z2 we have

64K'+ 192K5+260K4+214K3+ 1 19K2+44K+ 8
A (14)
16K6+24K5+9K4
which corresponds to Sulzer's5 results, with greater ac-
curacy in the K2 term. Substituting K= 1 gives A = 18.4
as in the simpler case to which it is equivalent, while
substituting K=2 gives A=8.65, which indicates the
advantage to be gained from using a tapered network.
Other equations for any combination of resistors and
capacitors can be derived from these general equations.
Fig. 6-Test circuit using four-mesh network. Thus, it is possible by direct substitution to solve net-
works employing these elements.
This formula was used to calculate the frequency for
the circuit shown in Fig. 6. The values of R' and C1 were ACKNOWLEDGMENT
assumed to be approximately 11,000 ohms and 47 The writer thanks H. Strell for checking the mathe-
MMF respectively. Then matics, R. Bernstein for doing experiments on final cir-
1 cuit, and some calculations. Also E. B. Hales and I. A.
Greenwood, Jr. helped offer valuable suggestions.
10 +
6.28X470X10-'2X110X103 1 1 APPENDIX I
70 70 Equating j terms to zero gives
2,300 cycles per second. 1 i C, R'\
0= 7+ +
The measured frequency differed from the calculated (c,CR) 3\ C R,
frequency by less than 3 per cent, with the difference 1 / lo0C1 10R' 5R'C1\
C +~
_10( + + l~
attributable to variations in R' and Cl from the assumed OCR R RC)
values. + XC1R'
A final example is the tapered four-section network
discussed by Sulzer.6 The substitutions here are Z3
=KZ1, Z4=KZ2, Z5=K2Z1, Z6=K2Z2, Z7=K3Z1 and a±+ ,/a2 - 4(7+ ±+)(RC1)
Z8= KZ2. Therefore .. (coCR)2=
R'CR
E (Lmnp) K2Z2(np) KZ2(Lp) Z2(Lm) 1 2
RC
e K6Z24 K4Z28 K5Z23 K6Z23 K2 where
Where / C1
R-
R'
RCl
R'C1\
a= 10 1+ C + +
L = (K2Z2 + K3Z1 + K3Z2) 2RC/

m = (KZ2 + K2Z1 + K2Z2) Then

n = (Z2 + KZ1 + KZ2)
p = (Z1 + Z2).
/a+ [a2-4 (7+C1+R' X

F=/ C~~~C R
/kRC
This reduces to 27rCR 2 R'Cl
E Z14 (4K + 3) Z13 (6K2 + 6K + 3) Z12 RC
+*

e Z24 K Z23 K2 Z22 The plus solution corresponds to the zero-phase shift
See Appendix I for the exact solution. case and does not satisfy requirements for oscillation.