Lectures On Stability: Analog Circuit Design

EECS140 ANALOG CIRCUIT DESIGN LECTURES ON STABILITY
University of California
Berkeley
College of Engineering
Department of Electrical Engineering
and Computer Science
Robert W. Brodersen
EECS140
Analog Circuit Design
Lectures
on
STABILITY
Effect of Feedback on Frequency Response SB-1
a(ω)
+ νO ( ω )
νIN ( ω ) Σ
-
macro block of a ( ω ) aO
R
C
Effect of Feedback on Frequency Response (Cont.) SB-2
jω RHP
exp ( p 1 ⋅ t )
X
p1 σ
Impulse
jω LHP
exp ( p 2 ⋅ t )
X
p2 σ
Impulse
Let a ( ω ) be a single pole response,
aO
a O ⇔ a ( ω ) = ------------------
a ( s ) = -------------
s ω
1– ---- 1 + j -
------
p1 ω p1
p 1 = – ω p1
ν OUT ( s ) a(s) T (s ) 
---------------- = A ( s ) = -------------------------- = 1---  --------------------
ν in ( s ) 1 + a(s) ⋅ f f  1 + T ( s )
aO
-------------
s
1 – ----  
p a O  -------------------------------------------
1 
A ( s ) = -----------------------------
1
= --------------------
aO ⋅ f 1 + aO ⋅ f  s 
1 + -------------  1 – ---------------------------------- 
s p1 ⋅ ( 1 + aO ⋅ f )
1 – ----
p1
Let T O = a O ⋅ f
 
A ( s ) = -------------- ⋅  --------------------------------------
aO 1
1 + TO s
 1 – ----------------------------
p1 ⋅ ( 1 + T O )
Pole is at p 1 ⋅ ( 1 + T O ) ⇒ – ω p1 ⋅ ( 1 + T O )
∞ ← To To = 0
X X
– ω p1 ⋅ ( 1 + T o ) – ω p1
Root Locus - motion of poles as loop Gain is increased .

open loop a ( ω )
20 log ( a o ) 1
a ∝ ----
ω
– 20 log ( 1 + T o )
closed loop A ( ω )
20 log  --------------
ao 
 1 + T o
ω p1 ω p1 ⋅ ( 1 + T o )
5HZ 50kHZ
T o = 104
 --------------
1 
Gain reduction by negative feedback reduces Gain by  1 + T o
and increases bandwidth by ( 1 + T ) o

Why not let T O → ∞ ? Problems if we have more than one pole.
( a o ) dB
aO
------------------------------------------------------------------
 1 – ---- s 
⋅ 1 – ---- ⋅  1 – ---- 
s s
 p 1  p 2  p 3
a(ω)
0.1ω p1 0.1ω p2 0.1ω p3

ω p1 ω p2 ω p3
10ω p1 10ω p2 10ω p3
0
– 90
– 180 0
corresponds to a sign inversion
0
– 270
At the frequency ( 10ω p2 ) the pahse shift is 180 0 or negative feedback

at DC is now positive feedback.

Lets look at the motion of a single pole with positive feedback :
a(ω)
+ νO ( ω )
νIN ( ω ) Σ
-
aO p 1 = – ω p1
a ( ω ) = – -------------
s
1 – ----
p1
a O ⋅ -------------------------------------
s
A ( s ) = --------------
1 + T O p1 ⋅ ( 1 – aO ⋅ f )
Since,
To = – ao ⋅ f
ao ⋅ f = 1
Stable Unstable
ω p1
ao ⋅ f < 1 ao ⋅ f > 1
pole is at – ω p1 ⋅ ( 1 – a o ⋅ f )
If T < – 1 or (1+T) < 0 the circuit is unstable

The condition for stability of a multipole response is the Nyquist Criteria.
a(s) a (s )
A ( s ) = -------------------------- = --------------------
1 + a(s) ⋅ f 1 + T( s)
Simple Version :
If Tjω > 1 at the frequency where the phase of T ( jω ) = – 180 0 ,
then the circuit is unstable.
T (j ω ) = T (s ) s = jω
{ T ( jω ) }
θ T( jω) = arc tan Im
----------------------------
Re {T ( jω ) }
Complex Nyquist Criteria :
Plot T ( jω ) on complex plane. As ω increases count number of times

– 1 is circled - even number means unstable (I Think).
ω p3 3 - Poles
ω<0
ω→∞ ω = 0
ω>0
ω p2 ω p1

a(ω)
+ νO ( ω )
νIN ( ω ) Σ
-
T ( jω ) 180 < 1
2
-
Worst Case Stability Condition
+
a O dB
f = 1
f = 0.1 T (ω ) = a ( ω )
T ( jω ) dB
20dB Frequency where θ = – 180 0
T( ω ) = 1
θ = 180 0 0.1ω p3
0dB
0.1ω p1 ω p1 ω p2 ω p3
10ω p1 0.1ω p2 10ω p2 10ω p3
0
– 90
– 180 0
– 270 0
PHASE MARGIN : Difference between the actual phase shift and – 180 0
when T( ω ) = 1
i.e. θ m ≡ Phase Margin = θ [ T ( ω ) ] – ( – 180 0 )
if θ m > 0 then the amplifier is stable - typically 45 0 – 60 0
1
A = -- more gain more stable
f
ro
R OUT = ------------ higher R OUT with more gain
1+T
a(ω) Always Stable

νIN ( ω ) Σ νO( ω )
a(ω)
+ νO ( ω )
νIN ( ω ) Σ
-
Worst Case for Stability

a(ω)
+ νO ( ω )
νIN ( ω ) Σ
-
As θm approaches 0 the amplifier is becoming unstable.
θ m = 30 0
A ( ω ) dB
closed loop gain
θ m = 45 0
( A O )dB
θ m = 60 0
θ m = 90 0
ω p1 ⋅ ( 1 + T ) ω
a(ω)
A ( ω ) = ---------------------------
1 + a(ω) ⋅ f
N( s)
a ( s ) = -----------
D(s)
N( s)
-----------
D (s ) N( s) zeros of a(s)
A ( s ) = --------------------------- = -----------------------------------
N( s) D ( s ) + N ( s) ⋅ f
1 + ----------- ⋅ f
D(s) poles of a(s)
if the feedback factor is frequency dependent, then,

Nf(s )
f ( s ) = ------------
Df ( s )
N ( s )D ( s )
A ( s ) = -------------------------------------------------------
D ( s )D f ( s ) + N ( s )N f ( s )
Compensation SB-17
Compensation is the method in which an amplifier is modified

so that it is stable.
One way is to decrease f (less feedback).
If ω180 is the frequency where,
θ ( a ( ω 180 ) ) = – 180 0
0
then if,
1
f < -----------------
a ( ω 180 ) 2
then,
T ( ω 180 ) = f ⋅ a ( ω 180 )
0 0
and stability is ensured.

Narrowbanding for Compensation SB-18

This entails the addition of a dominant pole
ω p1 ω p2 ω p3
a O dB
T ( jω ) dB
ω p1
----------- = ωC
aO ⋅ f
θ m = 45 0
0dB
Compensation pole ω 180 0
– 90 0
– 180 0
– 135 0 or θ m = 45 0
– 270 0
Narrowbanding for Compensation (Cont.) SB-19
For,
θ m = 45°
add a compensation pole, ω C at the fequency,
ω P1
--------- = ωC
ao f
ω P = 1MHz
a o f = 10 4
ω C = 100Hz
– 90 ° of phase shift from the new compensation pole.

– 45 ° from the second pole.
Pole Splitting SB-20

It is better to use an existing pole rather than add another.
RL
C GD
ν +in +
Diff. Output ν OUT
Stage
Pair CD
ν --
in - C GB C GS
a1 gm RL a2
------------------ ----------------------------------------------------- ------------------
1 + j ---ω ----  1 + j ---ω ----  ⋅  1 + j ---ω ----  1 + j ---ω ----
ω p1  ω p2  ω p3 ω p4
Pole Splitting (Cont.) SB-21

Lets say ωp1 and ωp4 are given with,
ω p4 » ω p1
C GD « C GS , C D
then,
1 1
ω p2 = ------------------ ω p3 = -----------
R DIFF C GS R LCD
gm
ω Z = --------
CG D
X X X X X X O
ω p4 ω p1 ω p3 ω p2 gm
ω p2 with C C --------
after CC C GD
SB-22
A ( ω ) dB
closed loop gain
< 3dB
( A O )dB
tRISE + t FALL
t SETTLE = ----------------------
2
tFALL
t RISE

If we add a compensation capacitor, C in parallel with C : C GD
1
ω p2 = ------------------------------------------------------
R DIFF ⋅ ( 1 + g m ⋅ R L ) ⋅ CC
gm
ω p3 = --------------------
C GS + C D
Lets put numbers in :

rad
R DIFF = 10MegΩ ω p1 = 10 ⋅ 10 6 ---------
sec
R L = 5MegΩ rad
ω p4 = 100 ⋅ 106 ---------
C GS = 0.1pF sec
a 1 = 10 3
C D = 0.1pF
g m = 10 –3 Mhos a2 = 1

Before compensation, and with,
C GD = 0
1 6 rad
ω p2 = ----------------------
–13 = 10 -------- -
10 ⋅ 107
sec
1 6 rad
ω p3 – 13 = 2 ⋅ 10 --------
= ------------------------------ -
5 ⋅ 10 ⋅ 10 6
sec
     
 ⋅ ⋅ 
a ( ω ) =  --------------------------  ⋅  -------------------
10 3
10 –3
5 10 6
1 
⋅  ------------------------------------------------------------- 
 ω    1 + j ------- ω  ⋅  1 + j --------------- ω   ω
 1 + j ---------------    1 + j ------- 
4 ⋅ 10 6 10 6  2 ⋅ 10 6  10 8
Compensate this amplifier for the worst case,

f = 1
with,
θ m = 45 0
-
Worst Case Stability Condition
ν in + νOUT

( 5 ⋅ 10 6 ) 10 6 2 ⋅ 10 6 10 ⋅ 10 6
13.4dB
10 5 θ = – 135 0
θ = – 225 0
Somewhere between 2MHz and 10MHz,

θ = – 180 0
but the loop gain,
1
T»1 T ∼ 10 5
So how to compensate it?
Add Cc so that the gain at the first non-dominant
pole (w p1). since wp3 will move to a higer frequency
and wp2 will move lower
ω 7
10
ω P2 = ----------------6 = ----------------6 = 2 ⋅ rad
P1
---------
5 × 10 5 × 10 sec
Formula for ω P2 & ω P3 with C : C
C C » CGS , C D
1
ω P 2 = ------------------------------------------------------
R DIFF ⋅ ( 1 + g m ⋅ R L ) ⋅ CC
gm
ω P3 = --------------------
C GS + C D
g
ω Z = ------m
CC
C C = 10 pF
1
ω P2 = ------------------------------------------
-
10 × 5 × 10 × CC
7 3
–3
10
ω P3 – 12 = 5 × 10 ⋅ rad ⁄ sec = 5000Mrad ⁄ sec
= -----------------------
- 9
0.2 × 10
–3
10 rad
ω Z = ---------- ⋅
8
= 10 ---------
10 –11 sec
So by adding a 10pF capacitor this circuit is made stable.

SB-29
ν in +
ν OUT
-
Slew Rate Limited

SB-30
small θ m
large θ m
Slew Rate & Compensation Miller Op Amp SB-31
CC
I SS
I SS Output Stage
GAIN Aν = 1
νo
0
ν id
----- M1 M2
2
ν id I SS
– -----
2
I SS
Gm
ν cd
– I SS
Slew Rate & Compensation Miller Op Amp (Cont.) SB-32
ν in +
ν OUT
-
few 100mV
Slew Rate Volts/µsec 10 low

20 – 50 med
100 high
Circuit situation with large ν id

basically
CC VB
Aν
νo
I SS Slewrate
νo ν o = – -----
I SS-
I SS = – CC ⋅ d------- d-------
or = slew rate
dt dt CC
1 I
ν o = ---- ⋅ ∫ I SS ⋅ dt = ----SS- ⋅ t ← linear with time
C C
ω p1 ω p2
aO
T( ω )
1
θ m = 45°
f = 1
a o = a DIFF ⋅ a g
Location of compensation pole :

ω
ω C = -------
P2
ao

CC
+
ν id a DIFF
-
ag νo
R D, DIFF
a DIFF = g m ⋅ R D, DIFF
1 ω ωP 2
ωC = ------------------------------ ------- ------------------
P2
= =
R D, diff ⋅ a g ⋅ C C ao a DIFF ⋅ a g
ω P2 1 ω P2
--------- = ------------------------ = -----------------------
a DIFF R D, DIFF ⋅ C C g m ⋅ R D, DIFF
gm
CC = ------- The size of the compensation depends only on g & ω
ω P2 m P2
2 ⋅ IDS
g m = -------------
V DSAT
⋅ IDS
2-------------
V DSAT =
gm
gm 1 IS S
IDS = ------
2
I-----SS ⋅ 2
νo
d------- ISS- ISS ⋅ ω 2
= ----- = ------ P2 = ------------ ⋅ ω P2
dt CC gm 1 g m1
νo = V
Slewrate = d------- DSAT1 ⋅ ω P2
dt
V DSAT = 0.1V
ω P2 = 10MHz ⋅ 2π SR = 6.3V ⁄ µsec
How to increase slew rate :

Increase V DSAT 1 ⇒ More current, smaller W
-----
L
gm
Increase ω P2 --------------------
1 + g mRS
Slew rate limits max change :

A
A sin ω S ⋅ t
Max rate of change here

–A
d-----------
V SIG
= ω S ⋅ A cos ω S ⋅ t
A = 2V ω S = 10 6 ⋅ 2π dt
SR = 13V ⁄ µsec Max Value ω S ⋅ A
MOS Miller Amp - Right Half Plane Zero SB-38
CC ignore for now

νo
+ gm ν 1 C2
idiff ν1 C1
R o, diff - R OUT, GAIN
RZ CC CC
νo
RL
also
idiff
RL
removes zero entirely
SB-39
MOS Miller Amp - Right Half Plane Zero (Cont.)
1
ω Z = --------------------------------
C C ⋅  ----- – R Z
1
 gm 
gm ⁄ C C
O O
1
-------------------------------
-
C C ⋅  ----- – R t
1
gm ∞

Lectures On Stability: Analog Circuit Design

Uploaded by

Copyright:

Available Formats

Lectures On Stability: Analog Circuit Design

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lectures On Stability: Analog Circuit Design

Uploaded by

Copyright:

Available Formats

EECS140 ANALOG CIRCUIT DESIGN LECTURES ON STABILITY

Effect of Feedback on Frequency Response SB-1

Effect of Feedback on Frequency Response (Cont.) SB-2

Effect of Feedback on Frequency Response (Cont.) SB-3

Let a ( ω ) be a single pole response,

Effect of Feedback on Frequency Response (Cont.) SB-4

Root Locus - motion of poles as loop Gain is increased .

Effect of Feedback on Frequency Response (Cont.) SB-5

Effect of Feedback on Frequency Response (Cont.) SB-6

0.1ω p1 0.1ω p2 0.1ω p3

At the frequency ( 10ω p2 ) the pahse shift is 180 0 or negative feedback

Effect of Feedback on Frequency Response (Cont.) SB-7

Effect of Feedback on Frequency Response (Cont.) SB-8

If T < – 1 or (1+T) < 0 the circuit is unstable

Effect of Feedback on Frequency Response (Cont.) SB-9

The condition for stability of a multipole response is the Nyquist Criteria.

Effect of Feedback on Frequency Response (Cont.) SB-10

Complex Nyquist Criteria :

Plot T ( jω ) on complex plane. As ω increases count number of times

Effect of Feedback on Frequency Response (Cont.) SB-11

Effect of Feedback on Frequency Response (Cont.) SB-12

Effect of Feedback on Frequency Response (Cont.) SB-13

i.e. θ m ≡ Phase Margin = θ [ T ( ω ) ] – ( – 180 0 )

if θ m > 0 then the amplifier is stable - typically 45 0 – 60 0

Effect of Feedback on Frequency Response (Cont.) SB-14

a(ω) Always Stable

Worst Case for Stability

Effect of Feedback on Frequency Response (Cont.) SB-15

As θm approaches 0 the amplifier is becoming unstable.

Effect of Feedback on Frequency Response (Cont.) SB-16

if the feedback factor is frequency dependent, then,

Compensation is the method in which an amplifier is modified

and stability is ensured.

Narrowbanding for Compensation SB-18

Narrowbanding for Compensation (Cont.) SB-19

– 90 ° of phase shift from the new compensation pole.

Pole Splitting SB-20

Pole Splitting (Cont.) SB-21

Pole Splitting (Cont.) SB-23

Lets put numbers in :

Pole Splitting (Cont.) SB-24

Pole Splitting (Cont.) SB-25

Compensate this amplifier for the worst case,

Pole Splitting (Cont.) SB-26

Somewhere between 2MHz and 10MHz,

Pole Splitting (Cont.) SB-27

Formula for ω P2 & ω P3 with C : C

Pole Splitting (Cont.) SB-28

So by adding a 10pF capacitor this circuit is made stable.

Slew Rate Limited

Slew Rate & Compensation Miller Op Amp SB-31

Slew Rate & Compensation Miller Op Amp (Cont.) SB-32

Slew Rate Volts/µsec 10 low

Slew Rate & Compensation Miller Op Amp (Cont.) SB-33

Circuit situation with large ν id

Slew Rate & Compensation Miller Op Amp (Cont.) SB-34

Location of compensation pole :

Slew Rate & Compensation Miller Op Amp (Cont.) SB-35