Lecture 20:

Stability and Frequency Compensation

for CMOS Amplifiers

Gu-Yeon Wei
Division of Engineering and Applied Sciences
Harvard University
[email protected]

Wei ES154 - Lecture 20 1

 Overview

• Reading:
– S&S Chapter 8.8~8.11
• Background
– When we apply feedback, it is important to also consider
stability. Sometimes, we need to frequency compensate
amplifiers to ensure stability.

Wei ES154 - Lecture 20 2

 Feedback Stability

• Consider the feedback circuit where β is constant and the closed-loop transfer
function is:

X(s) Σ A(s) Y(s)

Y
(s ) = A(s )
X 1 + β A(s )
β

– As βA(s=jω) Æ -1, the gain goes to infinity and the amplifier can oscillate
– In other words, oscillation can occur when

β A ( jω 1 ) = 1
∠ β A ( j ω 1 ) = − 180 °

Wei ES154 - Lecture 20 3

 Stability Analysis with Bode Plots

• Bode plot for loop gain Aβ

– Gain margin is an indication of
excess gain before instability
– Phase margin is an indication of
excess phase before -180° phase
shift at unity gain

Wei ES154 - Lecture 20 4

 Stability Analysis using Bode Plot of |A(s)|

• Can use the Bode plot of |A(s)| to

determine β for stable loop gain
– Larger values of β leads to instability

Wei ES154 - Lecture 20 5

 Nyquist Plot

• Polar plot of loop gain where radial distance is |Aβ| and angle is the phase angle
– Unstable if Nyquist plot encircles (-1,0)

Wei ES154 - Lecture 20 6

 Plotting Poles in S-Plane

• For an amplifier to be stable, poles must lie in the left half of the s-plane
– A pair of complex poles on the jω-axis Æ sustained oscillations
– Poles that are right of jω-axis Æ growing oscillations
– Poles that are left of jω-axis Æ stable
• Consider pole pair at σp±jωp
σ t
(
jω t
)
– Results in transient response v(t ) = e p e p + e p = 2e p cos ω p t
− jω t
( )
σ t

Wei ES154 - Lecture 20 7

 Effect of FB on Poles (Amp with Single Pole)

• For a closed loop transfer function, the poles are found by solving for
1 + Aβ = 0
Also known as the characteristic equation of the feedback loop
– Assume only poles in the system (zeros at infinite frequency)
– Assume β is independent of frequency (to simplify analysis)
• Consider an amp with open-loop transfer function A(s) characterized by a single
pole… A(s) = A0/(1+s/ωp)
A0 (1 + A0 β )
A f (s ) =
1 + s ω p (1 + A0 β )
– Feedback moves the pole to
ω pf = ω p (1 + A0 β )

Wei ES154 - Lecture 20 8

 • See pole movement in S-plane and Bode plot

Wei ES154 - Lecture 20 9

 Amplifier with Two-Pole Response

• Now consider an amplifier with open-loop transfer function with two real poles
A0
A(s ) =
(1 + s ω p1 )(1 + s ω p 2 )
• Closed loop poles come from 1+Aβ=0

s 2 + s (ω p1 + ω p 2 ) + (1 + A0 β )ω p1ω p 2 = 0

s = − 12 (ω p1 + ω p 2 ) ± 12 (ω + ω p 2 ) − 4(1 + A0 β )ω p1ω p 2
2
p1

– From the equation, as β increases, the poles move closer together and then
when they are equal, they become complex
– Complex poles cause peaking in the frequency response depending on Q
(see Figure 8.33)
ω0
+ ω0 = 0
2
s2 + s
Q

Wei ES154 - Lecture 20 10

 Frequency Compensation

• Typical op amps have many poles. In a folded-cascode topology, both the

folding node and the output load contribute poles. Due to having multiple poles,
op amps usually have to be compensated – their open-loop transfer function
must be modified in order for the closed-loop circuit to be stable.
– Need compensation b/c Aβ not <1 well before –180 phase shift
• Stabilize by
– Minimizing overall phase shift (zero compensation)
– Dropping gain or pushing dominant pole towards origin

• Watch out for right-half plane zeros

– Magnitude response increases +20dB/dec but…
– Causes negative phase shift

Wei ES154 - Lecture 20 11

 Compensating Two-Stage Op Amp

• Consider the following two-stage op amp:

vb3

vb2

E F

vb1
Vout1 A X Y B Vout2

CL vin CL

– Pole at node X (and Y) at high frequency

– Poles at nodes A (and B) and E (and F) can be close to the origin b/c output
resistance at E is high, CL can be large (even though Rout of 2nd stage is
small) Æ two dominant poles
– Will the amplifier be stable in a unity-gain feedback configuration?

Wei ES154 - Lecture 20 12

 • The resulting Bode plots…

• Assume β=1 and notice that at unity gain, phase shift is <-180°
• Move one of the dominant poles toward the origin to place gain crossover well
below phase crossover frequency

Wei ES154 - Lecture 20 13

 Miller Compensation

• Create a large capacitance at node E (and F) by relying on the moderate gain of

the 2nd stage and the high Rout of the first stage
CC

E A
Av1 Av2

Rout

– Create a large capacitance at node E, (1+Av2)CC

– Pole associated with node E now becomes
1
Rout1 [C E + (1 + Av 2 )CC ]

– Miller multiplication enables us to use a smaller capacitor size versus just

adding a capacitor to Vdd or Gnd (small-signal gnd)
– This also has the effect of moving the output pole (at A and B) away – also
known as pole splitting

Wei ES154 - Lecture 20 14

 • Using the simplified circuit of the two-stage amp Av1 Av2
– From Lecture 17 (HF response of CS amp)
we get (w/ CC=Cgs, no explicit CC, and RS CE
CE=Cgs)
1 CC
ω P1 =
Rs C gs + Rs C gd (1 + g m RL ') + Rs C gd RL ' Rs Vout
RL CL
C gs + C gd (1 + g m RL ') + C gd RL ' Rs gm
ωP 2 = ≅
C gs C gd RL ' C gs
Simplified circuit model of two-
stage amp w/ Miller Compensation
– With CC added… (replace Cgd w/ CC+Cgd)
1
ω P1 =
Rs C gs + Rs (C gd + CC )(1 + g m RL ') + Rs (C gd + CC ) RL ' Rs
C gs + (C gd + CC )(1 + g m RL ') + (CC + C gd ) RL ' Rs
ωP 2 =
C gs (C gd + CC )RL '
– So moves ωp1 to lower frequency and ωp2 to
higher frequency

Wei ES154 - Lecture 20 15

 +Miller Compensation with Zero

• One of the drawbacks of Miller compensation comes from the zero that it
introduces in the right-half plane (right of jω axis)…
– While it increases gain, it causes additional phase shift
• To combat this effect, add in a series resistor Rz
Av1 Av2

RS CE

Vout
Rz CC RL CL

– Then we get a zero frequency at

1
ωz ≈
CC (1 g m − Rz )
– If Rz > 1/gm then ωZ < 0 and get a left-half plane zero

Wei ES154 - Lecture 20 16