Stability

8.7 Determining the Loop Gain

-xtbA
-xtb
xs = 0

xtb xt

xo   xt bA
xo
bA  
xt
It should also be obvious that this applies regardless of where the loop
is broken.
 Vr
bA  
Vt

 1 1 
bA   1   
 Toc Tsc 
Figure 8.26 A conceptual feedback loop is broken at XXand a test voltage Vt is applied. The impedance Zt is equal to that previously seen looking to
the left of XX. The loop gain Ab = –Vr/Vt, where Vr is the returned voltage. As an alternative, Ab can be determined by finding the open-circuit transfer
function Toc, as in (c), and the short-circuit transfer function Tsc, as in (d), and combining them as indicated.
 RL //R2  R1 // Rid  R  R1 // Rid  R  Rid
Vr   V1
RL //R2  R1 // Rid  R  ro R1 // Rid  R   R2 Rid  R
RL //R2  R1 // Rid  R 
L  bA  Vr Vt  Vr V1  V1
RL //R2  R1 // Rid  R  ro

 R1 // Rid  R  Rid
R1 // Rid  R   R2 Rid  R
8.8 The Stability Problem
Closed-loop transfer function

As 
A f s  
1  As b s 

For physical frequencies s = jw

A jw 
A f  jw  
1  A jw b  jw 
L jw   A jw b  jw 
 A jw b  jw  e j w 

The loop gain varies with frequency that determines the stability or
instability of the feedback amplifier.
 A jw 
A f  jw  
1  A jw b  jw 
L jw   A jw b  jw 
 A jw b  jw  e j w 

If at w  w180
L jw180   1
 A f  jw180   A jw180 
The feedback amplifier will still be stable.
If at w  w180
L jw180   1
 A f  jw180   

The feedback amplifier will be an oscillator.

 signal X i sin w180 t 

xs  0

X f  A jw180 b  jw180  X i   X i
The amplifier is said to oscillate at the frequency w180.
What happens if at w180 the magnitude of the loop gain is greater
than unity?
The Nyquist Plot

The Nyquist plot is a formalized approach for testing for stability.

The radial distance is ∣ Ab∣ and the angle is the phase angle .
If the Nyquist plot encircles the point (-1,0) then the amplifier will be
unstable.
8.9 Effect of Feedback on the Amplifier Poles

The amplifier frequency response and stability are determined directly

by its poles.

Stability and Pole Location

Consider an amplifier with a pole pair at s   0  jwn

If the amplifier is subjected to a disturbance, its transient response
will contain terms of the form.
v t   e 0t e  jwnt  e  jwnt   2e 0t coswn t 
v t   e 0t e  jwnt  e  jwnt   2e 0t coswn t 
For an amplifier or any other system to be stable, its poles should lie
in the left half of the s plane.
A pair of complex conjugate poles on the jw axis gives rise to
sustained sinusoidal oscillations.
Poles in the right half of the s plane give rise to growing oscillations.
Poles of the Feedback Amplifier

A feedback amplifier

As 
A f s  
1  As b s 

The poles are obtained by solving the equation

1  As b s   0

which is called the characteristic equation of the feedback loop.

Amplifier with Single-Pole Response
Open-loop transfer function

A0
As  
1  s wP

Closed-loop transfer function

A0 1  A0 b 
A f s  
1  s wP 1  A0 b 

Thus the feedback moves the pole along the negative real axis to a
a frequency wPf,

wPf  wP 1  A0 b 
For frequencie s w  wP 1  A0 b  :

A0 1  A0 b  Aw
A f s    0 P
1  s wP 1  A0 b  s
A0 A0wP
As   
1  s wP s
 Af s   As 
Since the pole of the closed-loop amplifier never enters the right half
of the s plane, the single-pole amplifier is stable for any value of b.
Thus this amplifier is said to be unconditionally stable.
Amplifier with Two-Pole Response
Open-loop transfer function
A0
As  
1  s wP1 1  s wP 2 
Closed - loop poles are obtained from 1  As b  0, which leads to
s 2  s wP1  wP 2   1  A0 b wP1wP 2  0
Thus the closed-loop poles are given by
1
s wP1  wP 2   1 wP1  wP 2 2  41  A0 b wP1wP 2
2 2
 1 1
s   wP1  wP 2   wP1  wP 2 2  41  A0 b wP1wP 2
2 2

s  wP1 ,  wP 2 for A0 b  0

This feedback amplifier also is

unconditionally stable.

Root-locus diagram
 The characteristic equation of a second-order network can be written
in the standard form
w0
s s
2
 w02  0
Q

where w0 is called the pole frequency and Q is called pole Q factor.

w0 w02
s  j w02 
2Q 4Q 2

2
 w0   
2
w 2
  
  0w 2
 0
2
  w02
 2Q   4Q 

The poles are complex if Q is greater than 0.5.

 By comparing Eqs. (8.35) and (8.37) we obtain the Q factor for the
Poles of the feedback amplifier as
1  A0 b wP1wP 2
Q
w P1  w P 2
The response shows no peaking for Q ≤ 0.707.
The boundary case corresponding to Q = 0.707 results in the maximally
flat response.
Example
Consider the positive-feedback circuit. Find the loop transmission L(s)
and the characteristic equation. Sketch a root-locus diagram for varying
K, and find the value of K that results in a maximally flat response, and
the value of K that makes the circuit oscillate. Assume that the amplifier
has infinite impedance and zero output impedance.
solution
Loop transmission L(s)
Vr
Ls      KT s 
Vt
Vr s 1 CR 
T s    2
V1 s  s 3 CR   1 CR 2
s K CR 
Thus, Ls   2
s  s 3 CR   1 CR 
2

The characteristic equation is

1  L s   0
2
3  1  K
s s
2
  s 0
CR  CR  CR
3 K  1 
2

s s
2
  0
CR  CR 
Comparing these two equations
w0
s s
2
 w02  0
Q
3 K  1 
2

s s
2
  0
CR  CR 

1
w0 
CR
1
Q
3 K
 1
Q
3 K
1
K 0Q 
3
Poles are located on the negative real
axis.

As K increases the poles are brought

closer together and eventually coincide
(Q = 0.5, K=1)
Further increasing K results in the poles
becoming complex and conjugate.
The root locus is a circle because the radial
distance w0 remains constant independent of the value of K.
The maximally flat response is obtained when Q = 0.707, which
results when K = 1.586.
In this case, the poles are at 45◦ angles.
The poles cross the jw axis into the right
half of the s plane at the value of
K that results in Q = ∞, that is K =3.
Thus for K ≥ 3 this circuit becomes
unstable.
Amplifiers with Three or More Poles
 8.10 Stability Study Using Bode Plots
The difference between the value of
Ab at w180 and unity, calls the
gain margin.
The difference between the phase
angle at this frequency and 180o
is termed the phase margin.

Fig. 8.36 Bode plot for the loop gain Ab illustrating the definitions of the gain and phase margins.
 Effect of Phase Margin on Closed-Loop Response

Consider a feedback amplifier with a large low-frequency loop gain,

A0b >> 1. It follows that the closed-loop gain at low frequencies is
approximately 1/b.
Denoting the frequency at which the magnitude of loop gain is unity
by w1 we have

A jw1 b  1 e  j
where
  180  phase margin
At w1 the closed - loop gain is
A jw1  1 b e  j
A f  jw1   
1  A jw1 b 1  e  j
 Thus the magnitude of the gain at w1 is
1b
A f  jw1  
1  e  j
For a phase of 45 ,   135 ; and we obtain
1
A f  jw1   1.3
b

That is, the gain peaks by a factor of 1.3 above the low-frequency
value of 1/b.
 An Alternative Approach

An alternative approach is to construct a Bode plot for the open-loop

gain A(jw) only. Assuming for the time being that b is independent
of frequency, we can plot 20log(1/b) as a horizontal straight line on
the same plane used for 20logA. The difference between the two
curves will be

1
20 log A jw   20 log  20 log Ab
b
which is the loop gain expressed in dB. We may therefore study
stability by examining the difference between the two plots.
 1
20 log A jw   20 log  20 log Ab
b
 8.11 Frequency Compensation

Theory
1. Introducing a new pole in the function A(s) at a sufficiently low
frequency, fD, such that the modified open-loop gain, A’(s),
intersects the 20log(1/b) curve with a slope difference of 20
dB/decade.
2. Shifting the pole from f = fP1 to f = f ’D.
Fig. 8.38 Frequency compensation for b = 10-2. The response labeled A’ is obtained by introducing an additional pole at fD. The A”
response is obtained by moving the original low-frequency pole to f’D.
Implementation
Resistance Rx and capacitance Cx represent the total resistance and
capacitance between the two nodes B and B’. It follows that the pole
fP1 is given by
1
f P1 
2C x Rx

Let us now connect the compensating capacitor CC between nodes

B and B’. The pole can be at any desired lower frequency f ’D:

1
f 
'

2 C x  CC Rx
D

We thus conclude that one can select an appropriate value for CC so

as to shift the pole frequency from fP1 to the value f ’D determined by
point Z’.
A disadvantage of this implementation method is that the required
value of CC is usually quite large.
 Miller Compensation and Pole Splitting

R1 and C1 represents the total resistance and total capacitance between

node B and ground.
R2 and C2 represent the total resistance and total capacitance between
node C and ground. Furthermore, it is assumed that C1 and C2 include
the Miller components due to capacitance C, and C2 includes the
input capacitance of the preceding amplifier stage.
Ii represents the output signal current of the preceding stage.
In the absence of the compensating capacitor Cf.

1 1
f P1  f P2 
2C1 R1 2C2 R2
 With Cf present
Vo sC f  g m R1R2

I i 1  sC1 R1  C2 R2  C f  g m R1 R2  R1  R2   s 2 C1C2  C f C1  C2 R1 R2

The zero is usually at a much higher frequency than the dominant pole,
and we shall neglect its effect. The denominator polynomial D(s) can
be written in the form
 s  s   1 1  S2
Ds   1  ' 1  '   1  S  '  '   ' '
 w P1  w P 2   w P1 w P 2  w P1w P 2
Where w’P1 and w’P2 are the new frequencies of the two poles.
Normally one of the poles will be dominant; w’P1 << w’P2. Thus

s s2
Ds   1  
w '
P1 w P' 1w P' 2
Equating the coefficients of s in the denominator Eq. (8.55) and
Eq. (8.57) results in
1
w P1 
'

C1 R1  C2 R2  C f  g m R1 R2  R1  R2 
1

g m R1C f R2
g mC f
w '

C1C2  C f C1  C2 
P2
Original poles
1 1
f P1  f P2 
2C1 R1 2C2 R2
Compensated poles
1
w '
P1 
g m R1C f R2
g mC f
w '

C1C2  C f C1  C2 
P2

We see that as Cf increased, w’P1 is reduced and w’P2 is increased.

This is referred to as pole splitting.
Note that the increase in w’P2 is highly beneficial; it allows us to move
point Z further to the right, thus resulting in higher compensated open-
loop gain.
Cf is multiplied by the Miller-effect factor gmR2, thus resulting in a
much larger capacitance, gmR2Cf.