RC Phase-Shift Oscillators

Ramn Vargas Patrn

[email protected]
INICTEL-UNI
Oscillators generate periodic signals in the time domain. They convert DC power into
AC signal power. Signal generation implies production of self-sustained oscillations.
According to the types of waveforms produced oscillators can be classified into one of
four generic types:
-Harmonic oscillators: used for sine-wave generation.
-Sawtooth oscillators: used for the generation of exponential or linear sawtooth waves.
-Relaxation oscillators: used for current or voltage pulse generation with negative
resistance devices.
-Astable multivibrators: used for the generation of rectangular or square waves.
Each type of oscillator resorts to either a positive-feedback or negative-resistance
principle to operate. On the other hand, very complex waveforms can be generated
mixing analog and digital techniques.
The Positive Feedback Approach for a Harmonic Oscillator
As a system using positive feedback, the harmonic oscillator can be represented in
block-diagram form as depicted by Fig.1, where G(s) is the Laplace transform of the
open-loop voltage-gain function of the amplifier stage and H(s) is the transfer function
of the passive feedback network.

Fig.1 An oscillator as a feedback system

The closed-loop gain function of the system is found to be:

V0
(s ) = G (s )
V1
1 G (s )H (s )
Mathematically, it represents the response of the system to an impulse function, i.e., a
function for which V1 (s ) = 1 . In the real-world, impulse functions are approximated by
narrow noise impulses.

-1-

The expression:

1 G (s )H (s ) = 0

(1)

is known as the characteristic equation and contains important information regarding

the systems stability. Generally speaking, the roots of the equation are of the type
s = j . When = 0 the system is marginally stable and s = j . If certain
conditions are met, this situation will describe an oscillator in the steady state delivering
a constant-amplitude sine wave of radian frequency . A real-world oscillator requires
that > 0 for oscillations to start and build up. The power supply start-up impulse or
circuit thermal noise will generally trigger the response. Then, some type of amplitude
stabilization mechanism in the system will gradually reduce and stabilize the
oscillations. An amplitude-stabilized oscillator will usually exhibit good frequency
stability too. One final note here is that once the circuit bursts into oscillation no
external perturbation v1 (t ) is needed for sustained operation.
Equation (1) tells us that after amplitude stabilization:

G ( j ) H ( j ) = 1

(2)

This is Barkhausens criterion for oscillations. The left hand member is a complex
number and consists of a real and an imaginary part. Thus, we may write the equation:

Re( ) + j Im( ) = 1 + j 0
Re( ) = 1
Im( ) = 0
Im() equated to zero will generally give the frequency of oscillation. Re() equated to
unity will yield the conditions to be met for oscillations.
If we open the loop at the input of block G(s) in Fig.1 and inject a probe signal VT(s) at
this point, the output from block H(s) would be:

VL (s ) = VT (s ) G (s )H (s )
We now define the gain function:

AL (s ) =

V L (s )
= G (s )H (s )
VT (s )

This is the loop gain of our system. Barkhausens criterion then states that the
oscillators loop gain must be unity for oscillations to develop in the circuit.

-2-

Some Phase-Shift Oscillators

The most simple RC phase-shift oscillator configuration uses three buffered RC cells
and a voltage amplifier with very high input impedance and very low output impedance.
Fig.2 shows a typical schematic.
Because the RC cells wont load each other, the loop gain may be found to be:

sRC
AL (s ) = G (s )H (s ) = A

sRC + 1

s 3 R 3C 3

= A 3 3 3
2 2 2
s
R
C
+
3
s
R
C
+
3
sRC
+
1

Fig.2 Simple RC Phase-Shift Oscillator

For sine-wave steady-state operation, s = j , and the following may be written:
j 3 R 3 C 3
1
=
2 2 2
2 2 2
A
1 3 R C + jRC 3 R C

(3)

Equating the real part of the denominator to zero will yield the frequency of oscillation:
1 3 2 R 2 C 2 = 0 0 =

1
3RC

This result fed back into expression (3) will render information on amplifier gain A.
Accordingly:
A = 8
The phase shift introduced by each RC cell can be obtained from its transfer function:
F ( j ) =

jRC
jRC + 1

-3-

F ( j 0 ) =

j
3
j

+ 1

Then:
1
= 90 30 = 60
3

= 90 tan 1

The total phase shift introduced by the RC network at = 0 will be three times , or
180.

The BJT RC Phase-Shift Oscillator

Alike the preceding configuration, the BJT RC Phase-Shift Oscillator is a popular
configuration for the generation of low-frequency sine waves, starting at a few Hertz
and up to about 100 kHz. A schematic diagram of a basic implementation can be seen in
Fig.3. If the RC cells were isolated from each other, the phase shift per cell would be
60. However, not being it the case, we need to perform a detailed analysis considering
loading effects.

Fig.3 BJT-based RC Phase-Shift Oscillator

The bipolar transistor is a current amplifier. Therefore, it is convenient to express the
loop gain as a current ratio:
I
AL (s ) = b'
Ib
where I b' is a test signal current injected at the transistors base and I b is the
corresponding base current following loop excitation. Please see Fig.4.

-4-

Fig.4 Equivalent circuit for loop-gain calculation

The node-potential set of equations in the Laplace domain can be written as:
h fe I b' =

VA
+ (V A V B )sC
RC

VB
+ (V B V A )sC + (V B VC )sC
R
V
0 = C + (VC V B )sC + (VC V D )sC
R
V
0 = D + (V D VC )sC
R

0=

A small simplification leads to:

1

h fe I b' = V A
+ sC V B sC
RC

1
0 = V A sC + V B + 2 sC VC sC
R

0 = V B sC + VC + 2 sC V D sC
R

0 = VC sC + V D + sC
R

... (4.a )
... (4.b )
... (4.c )
... (4.d )

On the other hand, usually, Rb >> hie . Then:

V D = RI b R ' + hie I b
From Eq.(4.d):

VC = V D
+ 1
sRC

-5-

... (4.e )

Substituting into Eq.(4.c):

1
1

0 = V B sC + V D
+ 1 + 2 sC V D sC
sRC R

which simplifies to:

1
3

VB = VD 2 2 2 +
+ 1
sRC
s R C
Substituting for VB and VC in Eq.(4.b) we get:
1
3

0 = V A sC + V D 2 2 2 +
+ 1 + 2 sC V D
+ 1 sC
sRC R
s R C

sRC
yielding:
5
6
1

V A = VD 3 3 3 + 2 2 2 +
+ 1
sRC
s R C
s R C
Substituting for V A and VB in Eq.(4.a):

5
6
1
3
1
1

h fe I b' = VD 3 3 3 + 2 2 2 +
+ 1
+ sC VD 2 2 2 +
+ 1 sC
sRC RC
sRC
s R C
s R C
s R C

After some simplification:

1
5
6
1
1
4
3
h fe I b' = VD 3 3
+ 2 2
+
+
+ 2 3 2 + 2 +
3
2
sRRC C RC s R C
s R RC C
sR C R
s R RC C
Bearing in mind Eq.(4.e), the last expression transforms to:

1
5
6
R
1
4
h fe I b' = I b 3 2
+ 2
+
+
+ 2 2 2 +
+ 3
3
2
sRC C RC s R C
sRC
s RRC C
s R RC C

... (5)

If AL (s ) = 1 , then I b = I b' . For sine wave operation, s = j and Eq.(5) is rewritten as:
h fe = j

1
6
4
5
1
R
j
j
2
2 2 2 +
+3
3
2
RC C
RC RRC C
RC
R RC C
R C
3

... (6 )

The frequency of oscillation is obtained equating the imaginary part of the right-hand
member to zero:
1
6
4
0= 3 2

0 R RC C 3 0 RC C 0 RC

-6-

Knowing that 0 can not be zero we may write:

0=

0 R RC C

6
4

RC R

and solve for 0 :

0 =

1
6
4
RC RC
+
RC R

or in a more suitable form:

1

0 =

RC 6 +

4 RC
R

in radians per second. The condition for oscillation is obtained equating to h fe the real
part of the right-hand member of Eq.(6) while making = 0 :

h fe = 3 +

5
1
R
2
2 2 2
2
RC 0 RRC C
0 R C

6R

4R
R
5
+ 4 6 + C
RC
R
RC

R
R
= 23 29
4 C
RC
R
= 3+

Then:

h fe = 23 + 29

R
R
+4 C
RC
R

... (7)

The minimum value of h fe required for oscillations is h fe min = 44.5 , and occurs when
R
= 0.37 . If h fe is less than the said value the circuit wont oscillate, because AL ( j )
RC
would be less than unity. We can write Eq.(7) in the alternate form:

h fe 23
R
=
+
RC
58

h fe 23
4

29
58

-7-

R
to be a real number. If h fe = 44.5 , then
RC
R
R
. Then
= 0.37 . The design process would be then: Given h fe and 0 , find
RC
RC
compute the RC product and select convenient values for C and R. Design the DC bias
network for Class-A operation and symmetrical signal excursion. Amplitude distortion
at the output may be reduced by introducing negative feedback in the emitter branch
with a series added small resistor. Of course, a little more gain would be needed to
compensate for the reduction in the effective transconductance g m' of the circuit, i.e.:
Again, h fe must be greater than 44.5 for

gm
1 + g m Re
where g m is the transconductance of the transistor and Re is the series added resistor.
g m' =

The JFET RC Phase-Shift Oscillator

Another approach uses a discrete high input-impedance voltage amplifier stage. The
JFET is a high input-impedance semiconductor device and is very well suited for
voltage amplification. Fig.5 shows a schematic diagram of a typical JFET-based RC
Phase-Shift oscillator, yet another popular alternative for low-frequency sine wave
generation. As may be noted, resistor R of the far right RC cell substitutes for the gatesource bias return resistor commonly used in biasing schemes.
A general approach for the passive phase-shift network used in this type of oscillator
can be seen in Fig.6, where Z1 and Z2 are a capacitor C and a resistor R. Either
impedance can be selected to be the capacitor. However, the preferred configuration is
that depicted by Fig.5. Given the JFETs high input impedance, loading effects
occurring in the gate circuit are virtually eliminated. The amplifying device is biased for
Class-A operation and minimum signal distortion at the output. If needed, negative
feedback may be used for wave shape correction, as in the bipolar transistor case.

Fig.5 JFET-based RC Phase-Shift Oscillator

-8-

Fig.6 The three-cell RC network general approach

Design work requires some knowledge of the JFETs small-signal parameters. These
are defined by:
v
rds = ds when v gs = 0
id
gm =

id
when v ds = 0
v gs

v ds
when i d = 0
v gs

where v ds , id , v gs are small-signal variations about a quiescent point (please see Fig.7
below).

Fig.7 Basic JFET amplifier and small-signal equations

Here, gm is the low-frequency forward transconductance, rds is the drains dynamic
output resistance and is the amplification factor. The latter equals the product g m rds .
Sine wave operation permits us to perform calculations in the frequency domain.
Accordingly, we can write the mesh equations for the network of Fig.6 in phasor form
as:
V1 = I 1 (Z 1 + Z 2 ) I 2 Z 2
0 = I 1 Z 2 + I 2 (Z 1 + 2 Z 2 ) I 3 Z 2
0 = I 2 Z 2 + I 3 (Z 1 + 2 Z 2 )
The output voltage is given by:
V2 = I 3 Z 2

-9-

...(8)

By Cramers rule:
I3 =

The determinant of the coefficient matrix is calculated as:

(Z1 + Z 2 )

Z2

Z2
0

(Z 1 + 2 Z 2 )

Z2
(Z 1 + 2 Z 2 )

Z2

which reduces to:

= (Z 1 + Z 2 )

(Z 1 + 2Z 2 )
Z2

Z2
+ Z [( Z 2 )(Z 1 + 2Z 2 )]
(Z 1 + 2 Z 2 ) 2

After performing the indicated algebraic operations we obtain:

3

= Z 1 + 5Z 1 Z 2 + 6 Z 1 Z 2 + Z 2

... (9)

The determinant 3 is defined by:

(Z1 + Z 2 )

Z2

V1

Z2

(Z 1 + 2 Z 2 )

Z2

3 =

= V1 Z 2

Then:

I3 =

Z2
3

V1

V1

Z 1 + 5Z 1 Z 2 + 6 Z 1 Z 2 + Z 2

Substituting into Eq.(8):

V2 =

Z2
3

Z 1 + 5Z 1 Z 2 + 6 Z 1 Z 2 + Z 2

The RC networks transfer function is then:

3

V2
Z2
= 3
2
V1 Z 1 + 5Z 1 Z 2 + 6Z 1 Z 2 2 + Z 2 3

...(10)

The networks input impedance is given by:

Z IN =

-10-

V1
I1

...(11)

Observing that:
I1 =

...(12)

where:
V1
1 = 0

Z2

(Z 1 + 2 Z 2 )

Z2

Z2

(Z 1 + 2 Z 2 )

= V1 (Z 1 + 2 Z 2 ) Z 2

= V1 Z 1 + 4 Z 1 Z 2 + 3Z 2

...(13)

we get, from equations (9), (11), (12) and (13):

3

V1 Z 1 + 5Z 1 Z 2 + 6Z 1 Z 2 + Z 2
=
2
2
1
Z 1 + 4Z 1 Z 2 + 3Z 2

Z IN =

...(14)

Next, we choose Z1 and Z 2 to be:

1
sC
Z2 = R
Z1 =

Substituting in equations (10) and (14) we arrive to:

V2
s 3 R 3C 3
= 3 3 3
V1 s R C + 6 s 2 R 2 C 2 + 5sRC + 1

...(15)

and:
s 3 R 3C 3 + 6 s 2 R 2 C 2 + 5sRC + 1
3s 3 R 2 C 3 + 4 s 2 RC 2 + sC
s 3 R 3C 3 + 6 s 2 R 2 C 2 + 5sRC + 1
=R
3s 3 R 3 C 3 + 4 s 2 R 2 C 2 + sRC

Z IN =

...(16)

From Eq.(15), and for s = j :

H ( j ) =

V2
j 3 R 3 C 3
( j ) =
V1
1 6 2 R 2 C 2 + jRC 5 2 R 2 C 2

-11-

From Eq.(16), and for s = j :

Z IN = R

(1 6

R 2 C 2 + jRC 5 2 R 2 C 2
4 2 R 2 C 2 + jRC 1 3 2 R 2 C 2

...(17)

With regards to Fig.5, the expression for the voltage gain of the amplifying device is:
AV ( j ) = g m (R0 // Z IN )
where R0 = rds // R D , and Z IN is the RC networks input impedance. Barkhausens
criterion for the oscillators loop gain, Eq.(2), states that:
AV ( j ) H ( j ) = g m (R0 // Z IN ) H ( j ) = 1 + j 0

V2
( j ) is the passive networks transfer function already calculated.
V1
If we assume that the RC network doesnt load the JFETs output, a situation that can be
met with an adequate selection of circuit values, Barkhausens criterion will read as:

where H ( j ) =

AV ( j ) H ( j ) = g m R0 H ( j ) = 1 + j 0
Substituting for H ( j ) in the loop-gain condition yields:
g m R0

j 3 R 3 C 3
= 1 + j0
1 6 2 R 2 C 2 + jRC 5 2 R 2 C 2

which is satisfied when:

1 6 2 R 2 C 2 = 0
This sets the radian frequency of oscillation at:
1
0 =
6 RC
The condition for oscillation is:
0 2 R 2C 2
g m R0
=1
2
5 0 R 2C 2
which shows that the minimum value required of voltage gain is g m R0 = 29 . At the
frequency of oscillation the input impedance of the RC network is given by Eq.(17) as:

j 29
R ohms
4 6 + j3
= (0.83 j 2.70 )R ohms

Z IN =

= 2.82 R exp( j 72.9 ) ohms

-12-

An interesting result is noticed here, and is that Z IN is independent of the frequency of

oscillation, depending only upon R.
In order to avoid loading sensibly the JFETs output, the following should be satisfied
as a rule of thumb:
2.82 R >> R0 R >> 0.355 R0 R > 3.55 R0
We would now like to consider the effect of the RC network loading the JFETs output.
Observing Fig.4 we can write the following circuit analogies:
V gs' Vbe'
V gs Vbe = I b R

for the case in which R ' = 0 . Then:

V gs
g m JFET V

'
gs

Vbe
I R
b '
'
g m BJT Vbe
h fe I b

Barkhausens criterion stablishes that:

V gs

'
gs

Ib
=1
I b'

We may write then:

1
g m JFET

R
h fe

and:
...(18)

h fe g m JFET R

We already arrived, in the BJT case to:

h fe > 23 + 4k +

29
k

with k =

R0
R

Using analogy (18):

g m JFET R > 23 + 4k +

29
k

R0
29
> 23 + 4k +
k
k
2
g m JFET R0 > 23k + 4k + 29

g m JFET

Finally, the minimum voltage gain required from the JFET stage considering loading
effects is:
2

R
R
Av = g m R0 = 4 0 + 23 0 + 29
R
R

-13-

where g m , R0 and R are quantities previously defined during the analysis process of the
JFET oscillator.

The Five-Cell RC Phase-Shift Oscillator

The five-cell RC phase-shift oscillator is an extension of the three-cell case of the
preceding section, and will also require a high input-impedance voltage amplifier for
ease of analysis and design. A medium input-impedance amplifier could also be used,
but Z2 would need to be considered part of the amplifiers input circuitry, requiring
further analysis prior to design work. There exists also some concern with regards to the
loading effect of the RC network on the amplifiers output. This issue will be given
detailed consideration in our study. As will be demonstrated, the use of five RC cells
reduces considerably the gain threshold for oscillation, as compared to the three-cell
circuit. The amplifying device will be considered to be a vacuum triode. Fig.8 shows a
five-cell RC network as used in a phase-shift oscillator.

Fig.8 Five-cell RC network

The mesh equations for the passive network with input V1 may be written in phasor
form as:
V1 = I 1 (Z 1 + Z 2 ) I 2 Z 2

0 = I 1 Z 2 + I 2 (Z 1 + 2 Z 2 ) I 3 Z 2
0 = I 2 Z 2 + I 3 (Z 1 + 2 Z 2 ) I 4 Z 2
0 = I 3 Z 2 + I 4 (Z 1 + 2 Z 2 ) I 5 Z 2
0 = I 4 Z 2 + I 5 (Z 1 + 2 Z 2 )
and V2 = I 5 Z 2 . The system of five equations with five unknowns will be solved using,
again, Cramers rule. The determinant of the coefficient matrix is:

(Z 1 + Z 2 )
=

Z2
0
0
0

Z2
(Z 1 + 2Z 2 )
Z2
0
0

0
Z2
(Z 1 + 2 Z 2 )
Z2
0

-14-

0
0
Z2
(Z 1 + 2 Z 2 )
Z2

0
0
0
Z2
(Z 1 + 2 Z 2 )

which can be reduced to:

(Z1 + 2Z 2 )
= (Z 1 + Z 2 )

Z2
0
0

Z2
(Z 1 + 2 Z 2 )
Z2
0

0
Z2
(Z 1 + 2 Z 2 )
Z2

0
(Z 1 + 2 Z 2 )
0
2
Z2
Z2
Z2
0
(Z 1 + 2 Z 2 )

Z2

(Z 1 + 2 Z 2 )

Z2
(Z 1 + 2 Z 2 )

Z2

After calculating the determinants and performing the remaining algebraic operations
we arrive to:
5
4
3
2
2
3
4
5
= Z 1 + 9 Z 1 Z 2 + 28Z 1 Z 2 + 35Z 1 Z 2 + 15Z 1 Z 2 + Z 2

Mesh current I5 is calculated from I 5 = 5 . The determinant 5 is given by:

(Z 1 + Z 2 )
5 =

Z2
(Z 1 + 2Z 2 )
Z2
0
0

Z2
0
0
0

0
Z2
(Z 1 + 2 Z 2 )
Z2
0

0
V1
0
0
Z2
0
(Z 1 + 2 Z 2 ) 0
Z2
0

which readily simplifies to:

(Z 1 + Z 2 )

Z2

V1

Z2
0
0

(Z 1 + 2 Z 2 )

Z2
(Z1 + 2Z 2 )
Z2

0
0
0

5 = Z 2

(Z 1 + Z 2 )
= Z2

= Z2

Z2
0

(Z 1 + Z 2 )
Z2

= V1 Z 2

Z2
0

Z2
V1
(Z 1 + 2 Z 2 ) 0
Z2
0
V1
0

Then:
5

V2 = I 5 Z 2 =

5
Z
Z 2 = 2 V1

The RC networks transfer function is defined as H ( j ) =

-15-

V2 Z 2
=
, yielding:
V1

Z2

H ( j ) =

Z 1 + 9Z 1 Z 2 + 28Z 1 Z 2 + 35Z 1 Z 2 + 15Z 1 Z 2 + Z 2

Next, we need to compute the RC networks input impedance. It is given by:

Z IN =
with current I 1 obtained from I 1 =

V1
0
1 = 0
0
0

V1
I1

1
.

Z2
(Z 1 + 2 Z 2 )
Z2
0
0

0
Z2
(Z 1 + 2 Z 2 )
Z2
0

0
0
Z2
(Z 1 + 2 Z 2 )
Z2

0
0
0
Z2
(Z 1 + 2 Z 2 )

which can be shown to reduce to:

1 = V1 (Z 1 + 2Z 2 )

(Z 1 + 2 Z 2 )

Z2

Z2
0

(Z 1 + 2 Z 2 )

Z2

Z2
V1 Z 2
(Z 1 + 2 Z 2 )

(Z 1 + 2 Z 2 )
Z2
Z2
(Z 1 + 2 Z 2 )

It further simplifies to:

) (

1 = V1 (Z 1 + 2 Z 2 ) Z 1 6 Z 1 Z 2 10 Z 1 Z 2 4 Z 2 V1 Z 1 Z 2 + 4 Z 1 Z 2 + 3Z 2

giving:

1 = V1 Z 1 + 8Z 1 Z 2 + 21Z 1 Z 2 + 20 Z 1 Z 2 + 5Z 2

Then:
5

V V
Z + 9Z 1 Z 2 + 28Z 1 Z 2 + 35Z 1 Z 2 + 15Z 1 Z 2 + Z 2

Z IN = 1 = 1 =
= 1
4
3
2
2
3
4
I1
1 1
Z 1 + 8Z 1 Z 2 + 21Z 1 Z 2 + 20Z 1 Z 2 + 5Z 2

V1
1
When Z1 is a capacitor C and Z 2 a resistor R, the impedances are Z 1 =
and
j C
Z 2 = R . Hence, substituting for Z1 and Z 2 we obtain:

-16-

j
Z IN =

1
9R
28 R 2
R3
15 R 4
j
j
+
+

35

+ R5
C
5C 5 4C 4
3C 3
2C 2
1
8R
R2
20 R 3
+
j

21

j
+ 5R 4
C
4C 4
3C 3
2C 2

Multiplying the numerator and denominator of the above expression by 5 RC 5 we get:

Z IN

5 R 5 C 5 j15 4 R 4 C 4 35 3 R 3C 3 + j 28 2 R 2 C 2 + 9RC j
=R
5 5 R 5 C 5 j 20 4 R 4 C 4 21 3 R 3 C 3 + j8 2 R 2 C 2 + RC

Rearranging real and imaginary terms:

Z IN

5 R 5 C 5 35 3 R 3C 3 + 9RC + j ( 15 4 R 4 C 4 + 28 2 R 2 C 2 1)
=R
(19)
5 5 R 5 C 5 21 3 R 3C 3 + RC + j ( 20 4 R 4 C 4 + 8 2 R 2 C 2 )

Considering the loading effect of the RC network upon the triodes output yields the
familiar expression for the voltage gain of the vacuum tube:
AV ( j ) = g m (R0 // Z IN )
where R0 = rP // RP . The two paralleled resistances here are the plates dynamic output
resistance and the external plate bias resistor, respectively. Barkhausens criterion for
the oscillators loop gain, Eq.(2), gives:
AV ( j ) H ( j ) = g m (rP // RP // Z IN ) H ( j ) = 1 + j 0

V2
( j ) is the passive networks transfer function. Substituting for
V1
AV ( j ) and H ( j ) their individual expressions we obtain for the loop gain:

where H ( j ) =

AV ( j ) H ( j ) = g m (R0 // Z IN )

Z2
5
4
3
2
2
3
4
5
Z1 + 9 Z1 Z 2 + 28Z1 Z 2 + 35Z1 Z 2 + 15Z1Z 2 + Z 2

RZ 5
g m 0 IN Z 2
R0 + Z IN
= 5
4
3
2
2
3
4
5
Z1 + 9 Z1 Z 2 + 28Z1 Z 2 + 35Z1 Z 2 + 15Z1Z 2 + Z 2

Z IN

RZ 5
gm 0 IN Z 2
R0 + Z IN
4
3
2
2
3
4
Z1 + 8Z1 Z 2 + 21Z1 Z 2 + 20 Z1Z 2 + 5Z 2

-17-

After some simple algebraic work and simplification we arrive to:

g m R0 Z 25
AV ( j ) H ( j ) =
R0 Z 14 + 8Z 13 Z 2 + 21Z 12 Z 22 + 20 Z 1 Z 23 + 5Z 24 + Z 15 + 9 Z 14 Z 2 + 28Z 13 Z 22 + 35Z 12 Z 23 + 15Z 1 Z 24 + Z 25

) (

Substituting for Z1 and Z 2 their respective expressions, we get, after some manipulation:

AV ( j ) H ( j ) =

(R0 + 9 R ) 21R0 R 2 + 35 R 3

4 4
2C 2
C

g m R0 R 5

8 R0 R + 28 R 2
20 R0 R 3 + 15 R 4
1
4
5
+ 5 R0 R + R + j 5 5 +

C
3C 3

) (

) (

= 1 + j0
The denominator of the fraction above must be a real negative quantity for the loop gain
equation to hold. Then:

) (

8 R0 R + 28 R 2
20 R0 R 3 + 15 R 4
1
+

=0
C
5C 5
3C 3

Knowing that C can not be equal to zero, the above equation simplifies to:

) (

8 R0 R + 28 R 2
1
4 4 +
20 R0 R 3 + 15 R 4 = 0
2 2
C
C

which may be rewritten as:

(20R R
0

+ 15 R 4 ) 4 C 4 (8 R0 R + 28 R 2 ) 2 C 2 + 1 = 0

or in the alternate form:

R0

+ 15 R 4 4 C 4 8 0 + 28 R 2 2 C 2 + 1 = 0
20
R

Solving for 2 we get:

2

R0

R
R
+ 28 64 0 + 368 0 + 724
8
K2
R

R
R
2 =
= 2 2
R C
R0

+ 30 R 2 C 2
40
R

-18-

(20)

This equation gives the radian frequency of oscillation. We keep the positive sign for the
square root above because, as can be shown, the negative sign would not permit
satisfaction of the loop-gain condition.
We now need to calculate the minimum gain for oscillations to occur. The loop-gain
equation dictates that:
g m R0 R 5
=1
(R0 + 9 R ) 21R0 R 2 + 35R 3
4
5

+ 5 R0 R + R
4C 4
2C 2

) (

This expression can be rewritten as:

g m R0
=1
R0
R0

+ 9 21 + 35

R
R
+ 5 R 0 + 1

4 4 4
2 2 2
R C
R C
R

and solved for g m R0 , giving:

R0
R

+ 9 21 0 + 35

R
R
R

g m R0 = 4 4 4 + 2 2 2 5 0 + 1
R C
R C
R

(21)

This is the minimum value of the product g m R0 for sustained oscillations in the circuit.
Tables I and II below show how the ratio

R0
influences results for the frequency of
R

K
and the triodes unloaded-case small-signal voltage gain,
2RC
R
g m R0 . The particular case 0 0 corresponds to situations where the RC network
R
wont load the vacuum tubes output.

oscillation fosc =

-19-

TABLE I
1

R0
R

R
8 0 + 28
R

R
40 0 + 30
R

R0 2
2
R0
64 + 368 + 724
R
R

K 2 = 2 R 2C 2

K = RC

12
14
18
1 16
1 20
1 40

32
30
29
28.5
28.4
28.2
28

50
40
35
32.5
32
31
30

30.39
28.63
27.77
27.33
27.25
27.08
26.91

1.248
1.466
1.622
1.718
1.739
1.783
1.830

1.117
1.211
1.273
1.311
1.319
1.335
1.353

TABLE II

R0
R
12
14
18
1 16
1 20
1 40
0

R0
+9
R
9 .5
9.25
9.125
9.0625
9.05
9.025
9

R0
+ 35
R
45.5
40.25
37.63
36.31
36.05
35.53
35

21

R0
+1
R
3 .5
2.25
1.625
1.313
1.25
1.125
1

-20-

2 R 2C 2

4 R 4C 4

g m R0

1.248
1.466
1.622
1.718
1.739
1.783
1.830

1.557
2.149
2.631
2.951
3.024
3.179
3.349

26.85
20.90
18.11
16.75
16.49
15.96
15.44

We shall work out an example with the preceding results in an unusual application of a
vintage 1629 magic-eye tube in a single-stage 1-Hz sine-wave oscillator. This circuit
can be used to check the functionality of the tube while observing the opening and
closing of the eye once per second. The 1629 is a later version of the original 6E5
magic-eye tuning indicator tube in an octal base and a 12.6V/0.15A filament-voltage
regimen. It was manufactured by RCA, Tung-Sol and other companies. Characteristic
curves of the 1629 are similar to those of the 6E5.

Fig.9 Average Characteristics of the 6E5 Magic-Eye tube

-21-

Fig.10 Data for the 1629 Magic-Eye tube

-22-

The oscillator makes use of the triode section of the 1629 magic-eye tube for signal
amplification. The triodes plate potential to ground controls the opening and closing of
the eye, following the 1-Hz oscillation. Design work requires some knowledge on the
tubes triode small-signal parameters. For a graphical computation of these we can use
the following definitions in conjunction with data obtained from the tubes
characteristic curves:
vp
rp =
when v gk = 0
ip

gm =

ip
v gk
vp
v gk

when v p = 0
when i p = 0

where v p , i p , v gk are small-signal variations about a quiescent point (see Fig.11

below).

Fig.11 Basic triode amplifier and small-signal equations

The Tung-Sol 1629/6E5 specifications sheet shows data for DC power supply voltages
of 100, 200 and 250 Volts. We have at our disposal a 150-Volt DC power supply on the
workbench, so we will have to work out average 1629 data for this supply. The triode
sections peak current will have to be limited to some 0.2 mA. This suggests a series
triode plate resistor of 750k ohms. The triode grid voltage for a 0 shadow angle will be
something in between -3.3 Volts and -6.5 Volts (cathode at zero potential). Choosing a
geometric mean for this grid voltage, we get -4.63 Volts. One half of this value, or -2.32
Volts, will be the quiescent triode grid voltage for the best expected grid to cathode
waveform. The small-signal plate output resistance rp will be calculated upon this
value.
The plate will be loaded externally by the 750k-ohm resistor paralleled by the RC
networks input impedance Z IN . We should not expect, then, a large net load impedance
on the tubes triode plate. It would be advisable to maximize the plates signal current
excursion in this sense. For this reason, we select 0.1mA for the value of the quiescent
triode plate current (about one half of the average peak plate current), assuming Class-A
operation. On the graphs, for a plate current of 125uA, the average grid bias is -4.7
Volts with a 250-Volt supply and -1.5 Volts for a 100-Volt supply. The geometric mean

-23-

is -2.65 Volts, close enough to our previously selected grid bias. In order to minimize
estimation errors from the curves, we shall select a grid bias voltage of -2 Volts on the
graphics. During implementation, this will require the use of an AC-decoupled cathode
bias resistor.
We are now able to graphically (and grossly) estimate the small-signal parameters. For a
100-Volt DC supply and grid bias voltage of -2 Volts we find on the respective curve a
plate current of 106.25 uA, which will be labeled as I1. The external plate resistor is
0.5M ohms. The plate-to-cathode voltage drop is then VP1 = 100-(106.25)(0.5) = 46.87
Volts. On the other hand, for a 250-Volt DC supply and grid bias voltage of -2 Volts the
plate current is I2 = 193.75 uA. The external plate resistor is 1M ohm. Accordingly, the
plate-to-cathode voltage drop is VP2 = 250-(193.75)(1) = 56.25 Volts. The small-signal
plate output resistance rp will be estimated from the expression:

rp =

V P 2 V P1
I 2 I1

with grid bias voltage, VGK , constant at -2 Volts. We get:

rp =

(56.25 46.87 )Volts = 9.38 V = 107.2k ohms

(193.75 106.25) A 87.5 A

We shall now proceed to estimate a value for the amplification factor . The calculation
requires that the plate current stays constant at a certain fixed value, for instance, 125
A (a value close to the selected 0.1-mA quiescent point current).
From the characteristic curves, for a 100-Volt DC supply, the plate current is 125 A
when the grid bias voltage (cathode at zero potential) is VGK1 = -1.5 Volts. The plate-tocathode voltage drop is VP1 = 100-(125)(0.5) = 37.5 Volts. For a 250-Volt DC supply,
the plate current is 125 A when the grid bias voltage is VGK2 = -4.7 Volts. The plate-tocathode voltage drop is VP2 = 250-(125)(1) = 125 Volts. The amplification factor is
estimated from the expression:

VP 2 VP1
VGK 1 VGK 2

with constant plate current equal to 125 A. We get:

125 37.5
= 27.34
1 .5 + 4 .7

-24-

In terms of small-signal parameters, the amplification factor is = rp g m . We can

readily estimate the triodes transconductance as g m =

rp

= 0.255 mA / V .

R0 as defined before is given by:

R0 = r p // R P = 107.2k ohms // 750k ohms = 93.79k ohms
The expression for the radian frequency of oscillation, Eq.(20), yields OSC RC = 1.353
when

R0
0 . In this case, the RC network input impedance is found to be:
R
(69.953 + j 0.005) ohms
Z IN = R
( 27.993 j 52.34)
= (0.5555 j1.0388)R ohms
= 1.178 R exp( j 61.865 ) ohms

after Eq.(19) and values from Tables I and II. Thus, for a given vacuum triode and bias
resistors, Z IN is independent of the frequency of oscillation, depending only upon the
value selected for R.
As a rule of thumb for design, the RC network will not load the triodes plate output if
the following is accomplished:

Z IN >> R0 1.178R >> R0 R >> 0.8489 R0 R 8.5R0

At this point, we may start deciding on part values for the oscillator circuit. First, we
R
find the ratio 0 , knowing that recommended values for R lie between 1M ohm and 3M
R
ohms for the selected vacuum tube (triode section). A 1M-ohm value is chosen because
R
it is a handy one. Then, 0 = 0.09379 and, from formulas previously developed,
R
OSC RC = 1.292 . For a 1-Hz oscillation we would need, accordingly, a 0.205-F
capacitor. A standard 0.22-F part could be used if accuracy is not an issue.
To evaluate the loading effect of the passive network we just recall the Barkhausen
criterion stating that:
g m (R0 // Z IN ) H ( j ) = 1 + j 0

-25-

The RC network input impedance may be found to be:

60.177 + j 3.974
M ohms
25.9775 j 42.296
= (0.5663 j1.0749 ) M ohms

Z IN =

= 1.215 exp( j 62.22 ) M ohms

The net plate load impedance is:
R0 // Z IN = 93.79k ohms // (0.5663 j1.0749 ) M ohms
=

(0.09379) (0.5663 j1.0749) M ohms

(0.09379 + 0.5663 j1.0749)

0.0531 j 0.1008
M ohms
(0.6601 j1.0749)
= (90.094 j 5.943) k ohms

= 90.29 exp( j 3.77 ) k ohms

The overall voltage gain of the triode amplifier is:
AV = g m (R0 // Z IN )
= 0.255 mA / V [90.29 exp( j 3.77 ) ]k ohms
= [0.255 exp( j180 )]mA / V . [90.29 exp( j 3.77 )] k ohms

(22)

= 23.02 exp( j176.23 )

We see that the input impedance of the RC network introduces a 3.77 phase lag in the
overall voltage gain, compared to the case where the plate load is purely resistive.
According to Eq.(21), the triodes unloaded-case small-signal voltage gain, g m R0 ,
requires to be greater than 17.42 for oscillations to start in the circuit. From
experimental data, we are actually having:
g m R0 = (0.255 mA / V ) (93.79 k ohms ) = 23.92
enough gain for sustained oscillations to occur.

The successful 1-Hz oscillator

With theoretical and experimental data clearly understood, a working oscillator was
built accordingly on a vintage KEL CORP. Model 5710-100-044 experimental printed
circuit board (Fig.12). The 1629 magic-eye tube was plugged into a glass-ceramic 8-pin
octal socket located on top of the circuit board. The RC network consisted of five
0.22uF/400V capacitors and five 1M-ohm 1/4-watt resistors. A 750k-ohm 1/4-watt
resistor was used for plate bias from a 150-Volt DC power supply. A 5k-ohm
potentiometer wired as a rheostat was connected in series with the cathode of the
-26-

vacuum tube for adjustment of the optimal grid bias voltage. A 220V/12V 60-Hz 25watt AC mains transformer heated the 12.6V/0.15A filament of the tube. The oscillation
frequency, when checked with an electronic wrist watch, was found to be very close to
the design value of 1Hz. As expected, the eye opened and fully closed once per second.
The schematic diagram of the oscillator is shown in Fig.14.

Fig.12 Top view of the 1-Hz sine-wave oscillator construction based on the 1629
Magic-Eye tube

Fig.13 View showing adjustment of grid voltage for maximum opening/closing deltaangle of the eye

-27-

Fig.14 Schematic diagram of the sine-wave 1-Hz oscillator

Measured data for the 1629 Magic-Eye tube used in the experimental oscillator
A. Static bias operation measured with a B&K Test Bench 388A DMM
-The plate current of the triode amplifier section reads 0.16mA when the cathode bias
resistor RK is set to zero ohms (cathode at ground potential, i.e., VK = 0 Volts). Under
these conditions, the triode will also exhibit a contact potential of -0.64 Volts, i.e., VGK
= -0.64 Volts.
-If VK is increased to 1.2 Volts, the contact potential vanishes (VG shows zero Volts).
-In order to inhibit oscillations in the circuit for static bias operation, one of the Rlabeled resistors is shunted to ground with a jumper (do not select the 1M-ohm grid
resistor). Under these conditions, an approximately 1 shadow angle is obtained with a
cathode potential of VK = 2.83 Volts. An 80 shadow angle is obtained with VK = 0
Volts.

-28-

-At the operating point, the voltage drop across the 750k-ohm plate resistor is 103.7
Volts, yielding a plate current of 0.138 mA. Please compare with Eq.(23)
B. Dynamic operation
-Removing the jumper restores free oscillations and the eye opens and closes once per
second. A totally closed eye followed by the widest opening occurs when VK is adjusted
to a quiescent value of 1.85 Volts. At this point, RK measures 2.92k ohms, giving 0.633
mA for the total cathode current at the operating point. Although the potentiometer is a
5k-ohm part, out of the circuit shows a total resistance of 4.44k ohms, yielding 0.637
mA for the total cathode current at a 1 shadow angle.
-Our DMM has a peak-hold mode capability. Very careful measurements of the slowly
varying DC voltages were taken in order to determine the peak signal values existing in
the circuit. During oscillation, the voltage drop across the 750k-ohm plate resistor varies
from 79.70 Volts to 128.20 Volts. Assuming Class-A operation, the mean value is:

V Rp =

128.20 + 79.70
= 103.95Volts
2

(23)

The plate signal voltage is:

128.20 79.70
= 24.2 5Volts
2
after observing that the cathodes potential is approximately constant during oscillation,
due to the almost constant cathode current occurring during the opening and closing of
the eye. The cathode doesnt need a decoupling capacitor in this application then.
vp =

The grid potential varies between -1.13 Volts and +0.76 Volts. The mean value is:
0.76 1.13
= 0.185Volts
2
The negative value is presumably due to contact potential action in the tube. If contact

Vg =

potential were zero, then V g would be also zero for Class-A operation. The grid signal
voltage is:
0.76 + 1.13
vg =
= 0.945Volts
2
The grid-to-cathode voltage is defined as V gk = V g Vk , where Vg and Vk are the DC
plus AC voltages present at the grid and cathode electrodes, respectively. Hence, Vgk
varies between (-1.13-1.85) Volts = -2.98 Volts and (0.76-1.85) Volts = -1.09 Volts.
The most negative value, i.e., -2.98 Volts, is very close to the static value of -2.83 Volts
needed for a 1 shadow angle. For a 0 shadow angle a slightly more negative voltage is
needed, hence, the -2.98-Volt figure is reasonably correct. On the other hand, the mean
value for Vgk is:

V gk =

1.09 2.98
= 2.035Volts
2
-29-

The grid-to-cathode signal voltage is:

v gk =

1.09 + 2.98
= 0.945Volts = v g
2

The magnitude of the overall voltage gain would be then:

Av =

vp
v gk

24.25
= 25.66
0.945

reasonably close to the magnitude of the complex value given by Eq.(22).

We shall now refer to Fig.15. It shows the small-signal model for the triode sections
output.

Fig.15 Triode section output equivalent

The following may be written:

1
1
1
g m v gk = v p +
+
r
p R P Z IN

An equivalent form of the above equation is:

vp
1
1
g m v gk +
= i p = v p
+
rp
RP Z IN

The right hand member sets the contribution of external components RP and ZIN.
Hereinafter, vgk, vp, ip and ZIN will be treated as phasors. Accordingly, if a 0 phase
reference is assigned to vp, then:

v p = 24.25 exp( j 0 )Volts

-30-

The external plate load consists of R P = 750 k ohms and Z IN = (0.5663 j1.0749 ) M ohms

connected in parallel:

RP // Z IN =

(0.75) (0.5663 j1.0749) M ohms

(0.75) + (0.5663 j1.0749)

0.4247 j 0.8062
M ohms
1.3163 j1.0749
0.9112 exp( j 62.22 )
=
M ohms
1.6994 exp( j 39.23 )

= 0.5362 exp( j 22.99 ) M ohms

The plate signal current is given by:
vp
ip =
RP // Z IN

24.25Volts
0.5362 exp( j 22.99 ) M ohms

= 45.22 exp( j 22.99 )uA

= 45.22 exp( j 202.99 ) uA
and leads the plate signal voltage by 202.99, or equivalently, lags the signal voltage by
157.01.
Ramn Vargas Patrn
[email protected]
Lima-Peru, South America
October 17th 2013

-31-