notes, M.

Rodwell, copyrighted

ECE145A / 218A Notes :

Basic Analysis of Analog Circuits
Mark Rodwell
University of California, Santa Barbara

rodwell@ece.ucsb.edu 805-893-3244, 805-893-3262 fax

Comment

notes, M. Rodwell, copyrighted

This (2009) is a transitional year:

Next year 145abc will be reorganized,
reorganized
145a: fundamentals (devices, analog & RF analysis, models)
145cb: RF systems at IC and system level
This year:
some students have taken 145c:
already
l d have
h
device
d i models
d l
already know analog circuit analysis well
some students have not
must cover device models
must review some circuit analysis methods
These notes: shortened version (2009 only) of device models

Transistor Circuit Design

notes, M. Rodwell, copyrighted

This note set

-reviews the basics
-starts at the level of a first IC design course
-moves very quickly

This will
-establish
establish a common terminology
-accommodate capable students having minimal background in ICs.

notes, M. Rodwell, copyrighted

DC models
d l
DC bi
bias analysis
l i

Large-Signal Model For Bias Analysis

Ib

notes, M. Rodwell, copyrighted

Ic

+
Vbe

Provided that Vce > 0,

I c = I s exp(Vbe / VT ) and I b = I c / , where VT = kT / q
...note that Vbe is specified internal to the emitter resistance Rex
Th I e Rex drop
The
d is
i significan
i ifi t for
f HBTs
HBT operating
ti att currentt densities
d iti
near that required for peak transistor bandwidth.

notes, M. Rodwell, copyrighted

DC Bias Example: Current Mirror

I ref

Ic2

Q2

Q1
I b 2 I b1

Rex 2

We have Vbe
b 1 + I e1 ( Rex1 + Ree1 ) = Vbe
b 2 + I e 2 ( Rex 2 + Ree 2 )
and Vbe1 = Vt ln( I c1 / I s1 ), Vbe 2 = Vt ln( I c 2 / I s 2 )
Assume that >> 1, Ree 2 = 2 Ree1
& assume that AE1 = AE 2 ( AE is the emitter area).
area)
This implies Rex1 = Rex 2 / 2 , and I s1 = 2 I s 2 ,
from which we find I c 2 = I c1 / 2

Ree 2

Ie2

I e1

Rex1
Ree1

notes, M. Rodwell, copyrighted

Simpler DC Model for Bias Analysis

Ib

Ib

Ic
Vbe

Ic

Vbe

It is often sufficient in bias analysis to ignore the

variation of Vbe
b with I c and instead take Vbe
b = Vbe
b ,on = .
Vbe,on depends
d
d upon currentt density
d it and
d technolog
t h l y.
Biased at current densities within ~ 10% of peak bandwidth bias,
Vbe,on

0.9 V Modern Si/SiGe HBTs

= ~ 0.7 or 0.9 V InGaAs/InP HBTs

1.4 V GaAs/GaInP HBTs

notes, M. Rodwell, copyrighted

Simple DC Bias Example

Rc
Ib2

Rb
Q2

Ib11

Q1
Ree

-Vee

If we neglect the I b Rb drops, then Vb1 = Vb 2 = 0 Volts.

Approximate Ve1 = Ve 2 = 0.9 V (SiGe).
I c1 + I c 2 = 2 I c1 = (Vee 0.9V ) / Ree
I c1 = I c 2 = (Vee 0.9V ) / 2 Ree

notes, M. Rodwell, copyrighted

Efficiently Handling Base Currents In Bias Analysis

If I b Rbdrop is singificant,
one can solve simultaneous equations :

Rc
Ib2

Rb

I c1 + I c 2 = 2 I c1 = ( Vee I b1Rb ) / Ree

where I b1 = I c1 / ,

Q2

Q1
Ree

Q i k : find
Quicker
fi d by
b iteration
it ti :
1) solve I c1 = ( Vee ) / 2 Ree
2) solve I b1 I c1 /

-Vee

3) use this value of I b to solve I c1 = ( Vee I b1Rb ) / 2 Ree

Works because any well - designed circuit has DC bias
only weakly dependent upon .

Ib1

notes, M. Rodwell, copyrighted

small-signal
ll i
l
b
baseband
b d analysis
l i

Hybrid- Bipolar Transistor Model

notes, M. Rodwell, copyrighted

Ccbx
B

Rbb

Ccbi

Rc

Rbe = / g m

f = b + c

Rbe
Cbe,diff
gm f Cje
b diff =g

gm Vbee -jc

Vbe
Rex
E

Accurate model, but too detailed for quick hand analysis

Oversimplified Model for Quick Hand Analysis

Cbe Rbe Ccb g mVbe
B

Vbe

Rbe
B

g mVbe
C

Vbe

notes, M. Rodwell, copyrighted

In most high
high-frequency
frequency circuits
circuits, the node impedance is low and
Rce is therefore negligible.
Neglecting Rbb in high-frequency
high frequency analysis is a poor approximation
but is nevertheless common in introductory treatments.

The "textbook" analyses which follow use this oversimplified

model. These introductoryy treatments will later be refined.

notes, M. Rodwell, copyrighted

Common Emitter Stage: Basics

2) VC = RLeqq I C

RLeq

7) Rin ,T = Vb / I b = (re + Re )
5) I b = I c /
1) I e I c

4) Vbe = I e / g m

3) Ve = RE I e
6) Vb = I e (re + Re ) = I b (re + Re )

RE

8) Vout / Vin = Vout / VB = RLeq /(re + Re ) = RLeq /( Re + 1 / g m )

Gain is - RLeq /( Re + 1 / g m ) ; Transistor Rin is (re + RE )

notes, M. Rodwell, copyrighted

Emitter Follower Stage: Basics

6) Rin ,T = Vb / I b = (re + RLeq )
4) I b = I c /

1) I e I c

3) Vbe = I e / g m

2) Ve = RLeq I e
5) Vb = I e (re + RLeqq ) = I b (re + RLeqq )

RLeq

7) Vout / Vin = Vout / VE = RLeq /(re + RLeq ) = RLeq /( RLeq + 1 / g m )

Gain is RLeq /( RLeq + 1 / g m ) ; Transistor Rin is (re + RE )

notes, M. Rodwell, copyrighted

Common-Base Stage: Basics

2) Vout = RLeq I C

6) Vin = I e (re + Rb / )
7) Rin ,T = Ve / I e = re + RB /
+

5) Vbe = I e / g m
RB

1) I e I c
RLeq

4) Vb = I c Rb /
3) I b = I c /
7) Vout / Vin = RLeq /(re + Rb / )
Gain is RLeq /(re + Rb / ) ; Transistor Rin is re + Rb /

notes, M. Rodwell, copyrighted

Emitter Follower Output Impedance

ECE145C /218C notes, M. Rodwell, copyrighted

Common-Base Stage: Basics

2) Vout = RLeq I C

6) Vin = I e (re + Rb / )

Rout ,emitter
RB

Rout ,amp

7) Rin ,T = Ve / I e = re + RB /
+

5) Vbe = I e / g m
RB

4) Vb = I c Rb /

1) I e I c
RLeq

3) I b = I c /

7) Vout / Vin = RLeq /(re + Rb / )

REE

Gain is RLeq /(re + Rb / ) ; Transistor Rin is re + Rb /

E.F. output impedance is same problem as C.B. input impedance

Rout ,emitter = re + RB / = 1 / g m + RB /

Rout ,amp = Rout ,emitter REE

notes, M. Rodwell, copyrighted

Including Bias Circuit Resistances

Rin , Amp = RB1 RB 2 RinT

Rgen

Vgen

RB1

RLeq = RC RL

RC

Rin ,T

Vin

RB1

Th are (trivially
These
(t i i ll ) added
dd d in
i parallel
ll l with
ith the
th transisto
t
it r
terminal impedances to determine the net circuit impedances.
From which, Vin / Vgen = Rin ,amp /( Rin ,amp + Rgen ) , etc.

RL

notes, M. Rodwell, copyrighted

Baseband Analysis Of Multistage Circuits

For baseband analysis of multi-stage circuits,

simply break into individual stages.
Q1

Q2

Q3

Load impedance of the Nth stage includes the input

impedance of the (N+1)th stage

Q4

Analysis is then trivial...

trivial

Rin3

Rin2
Vin1

Q3

Q1 Vout1

Vout3
Q4

Q2
Rin3

Rin2

Rin4

Vout2

Vin2=V
Vout1

Vout2=Vin3
Rin4

Vout3=Vin4

notes, M. Rodwell, copyrighted

small-signal
ll i
l
b
baseband
b d analysis
l i

High-Frequency Analysis: The General Problem

Ccbx = 6.4 fF
B

Rc = 6500
Ccbi = 3.7 fF

Rbb = 49
Vb'e

R =
350

g mVb 'e

Cdiff = C je =
182 fF 38 fF

Rex = 4.3
E
B

Ccbx = 6.4 fF

Q2

Cdiff = C je =
182 fF 38 fF

Rc = 6500
Ccbi = 3.7 fF

Rbb = 49
Vb'e

Q1

notes, M. Rodwell, copyrighted

R =
350

g mVb 'e
Rex = 4.3

Vgen
E

Analyzing frequency response is difficult: cannot separate stage-by-stage

y accurate,, ggeneral,, tedious.
Method #1: nodal analysis:
Method #2: method of time constants: accurate, limited applicabilty, quick & intuitive

notes, M. Rodwell, copyrighted

N d l Analysis
Nodal
A l i

notes, M. Rodwell, copyrighted

Tutorial: Transfer Function Analysis: Nodal Analysis I

Simple & very familiar example : common - emitter amplifier.
Vout

Vin
Rgen
Vgen
Rb

Vin

C L RL

Ccb

Vout

Rgen
Vgen

Rb

Cbe Rbe

gmVbe

CL RL

notes, M. Rodwell, copyrighted

Tutorial: Transfer Function Analysis: Nodal Analysis II

Reduced circuit :

Vin

Vout

( Ri = Rgen || Rbe || Rb )
Ii=Vgen/Rgen Cbe Ri

gmVbe CL RL

Step 1 : Write Nodal Equations from KCL

Gi + sCbe + sCcb
g sC
m
cb

sCcb

Vin I i
=

GL + sCL + sCcb Vout 0

notes, M. Rodwell, copyrighted

Tutorial: Transfer Function Analysis: Nodal Analysis III

Step 2 : Solve Nodal Equations :
Vout / I in = N ( s ) / D ( s )
Gi + sCbe + sCcb
N ( s) =
g m sCcb

1
= ( g m sCcb )
0

Gi + sCbe + sCcb
D( s) =
g m sCcb

sCcb
GL + sCL + sCcb

D ( s ) = (Gi + sCbe + sCcb )(GL + sCL + sCcb ) (g m sCcb )( sCcb )

Step 3 : Organize in powers of s
D ( s ) = GiGL
+ s(GiCL + GiCcb + GLCbe + GLCcb + g mCcb
+ s 2 (CbeCL + CbeCcb + CcbCL + CcbCcb CcbCcb )

notes, M. Rodwell, copyrighted

Tutorial: Transfer Function Analysis: Nodal Analysis IV

Step 4 : Separate into dimensionless ratio - of - polynomials form,
separating
p
g constants and g
gains from the transfer function...
Vout
Vout
N ( s)
=
=
I in Vgen / Rgen D ( s )
( g m sCcb )
=
GiGL + s(GiCL + GiCcbb + GLCbe
b + GLCcb
b + g mCcb
b

+
s
(
C
C
+
C
C
+
C
C
)
be L
be cb
cb L

( g m sCcb ) Ri RL / Rggen
Vout
=
Vgen
1 + s( RLCL + RLCcb + RiCbe + RiCcb + g m Ri RLCcb

+
s
(
C
C
+
C
C
+
C
C
)
R
R
be L
be cb
cb L
i L

notes, M. Rodwell, copyrighted

Tutorial: Transfer Function Analysis: Nodal Analysis V

Rin , Amp
( Rbe || Rb || Rgen )
Ri
( Rbe || Rb )
note that
=
=
=
Rggen
Rggen
( Rbe || Rb ) + Rggen Rin , Ampp + Rggen
so...
Rin , Amp

Vout
( g m RL )
=
Vgen Rin , Amp + Rgen
b1
(1 sC
Ccb / g m )

1 + s( RLCL + RLCcb + RiCbe + RiCcb + g m Ri RLCcb )

+
s
(
C
C
+
C
C
+
C
C
)
R
R
be L
be cb
cb L
i L

Vout ( s ) Vout

=
Vgen ( s ) Vgen

mid band

1 + b1s
1 + a1s + a2 s 2

a1

a2

notes, M. Rodwell, copyrighted

Tutorial: Transfer Function Analysis: Nodal Analysis VI

Step 5 : Find the roots (poles & zeros) of the polynomial
Vout ( s ) Vout
=
Vgen ( s ) Vgen

mid band

1 + b1s
Vout
=
2
1 + a1s + a2 s
Vgen

mid band

1 + b1s
(1 s / s p1 )(1 s / s p 2 )

what are efficient methods of finding the poles ?

notes, M. Rodwell, copyrighted

Finding
Fi
di Poles
P l
f
from
T
Transfer
f Functions
F
ti

Finding Poles and Zeros

Ratio - of - Polynomial Form :
Vout ( s ) Vout
=
Vgen ( s ) Vgen

at mid- band

2
1
+
b
s
+
b
s
+ ...
m
1
2
*s
1 + a1s + a2 s 2 + ...

Poles and Zeros :

Vout ( s ) Vout
=
Vgen ( s ) Vgen

(1 s / sz1 )(1 s / sz1 )(1 s / sz1 )...

*s
(1 s / s p1 )(1 s / s p1 )(1 s / s p1 ))...
m

at mid- band

notes, M. Rodwell, copyrighted

notes, M. Rodwell, copyrighted

Finding Poles: Complex Poles

Vout
1 + b1s + b2 s 2 + ...
=k
Vgen
1 + a1s + a2 s 2 + a3s 3
If a3 / a2 << a2 then we can ignore the s 3 at moderate frequencie s and
1 + b1s + b2 s 2 + ...
Vout

k
2
Vgen
1
a
s
a
s
+
+
1
2

jd

If the roots of this are complex,

complex then
1 + b1s + b2 s 2 + ...
1 + b1s + b2 s 2 + ...
Vout

= k
= k
2
2
2
Vgen
1 + a1s + a2 s
1 + (2 / n ) s + s / n
Vout
Vgen

1 + b1s + b2 s 2 + ...

= k

s
s
1

1
+

n + jd n jd

d2 = n2 (1 2 )

Finding Poles: Separated Pole Approximation

notes, M. Rodwell, copyrighted

If the roots are widely separated

e.g.
g ( a2 /a1 ) << a1 , then
1 + b1s + b2 s 2 + ...
Vout

= k
2
I in
1 + a1s + a2 s
Vout
1 + b1s + b2 s 2 + ...
k
I in
a2
(1 + a1s )1 + s
a1

a1 is the dominant pole.

j
s = a2 / a1
s = a1

notes, M. Rodwell, copyrighted

IIntroductory
t d t
Ci
Circuit
it Design:
D i
summary

Gain Stages: Elementary Bandwidth Analysis

notes, M. Rodwell, copyrighted

Using the oversimplified device model below, with Cpi denoting the sum of
base-emitter depletion and diffusion capacitances, bandwidth of CE/CB/CC
stages
g can be found.

Cbe Rbe Ccb g mVbe

B

Vbe

notes, M. Rodwell, copyrighted

CE Stage: Elementary Bandwidth Analysis

CL
Rgen
RL

Rc
Rin

Ri is the parallel combination of Rgen, Rin, and Rpi

RLeq is the parallel combination of RL, Rc, and Ro
Note in the dominant pole (a1) the miller-multiplication of the collector base
capacitance

CC Stage: Elementary Bandwidth Analysis

Rgen

notes, M. Rodwell, copyrighted

CL
RL
Rin
REE

Ri is the parallel combination of Rgen, and Rin,

RLeq is the parallel combination of Ree and RL
Note that the frequency response is a mess. Given CL, the transfer function
very often has complex poles
poles, and may show strong gain peaking
peaking, hence
ringing in the pulse response.

CB Stage: Elementary Bandwidth Analysis

Rgen

Rcc

CL
RL

Rin
Rgen Vin

Cin,cb

Ccb Rcc

V
Vgen

CL
RL

Rin

Rin

notes, M. Rodwell, copyrighted

Cin ,cb = C je + g m b

qI / kT
g m ( j ) =
1 + j c

gm*Vin

Here we have a problem. To the extent that the CB stage is modeled by a very
very simple hybrid-pi model (explicitly, with zero Rbb), we find (by very simple
analysis) very high bandwidth
bandwidth, with poles having time constants equal to tau
tau_b,
b
to tau_c, and to the product of the load resistance times (Ccb+CL).
Note that
1) IInput capacitance
i
is
i indeed
i d d as noted.
d Does
D
not include
i l d effect
ff
off tau_c
2) Ignoring Rbb in CB stage analysis, while appealing for simplicity (e.g.
g
classes)) is quite
q
unreasonable,, as CcbRbb often dominates
undergrad
high frequency rolloff. More regarding this later.

notes, M. Rodwell, copyrighted

Method
M
th d off
Ti
Time
Constants
C
t t

notes, M. Rodwell, copyrighted

make revision for 2009----first before MOTC, give by summary

without derivation the standard stage expressions
expressions.
then define MOTC, first and second order
then show a 1-stage Darlington diff amp, and say caps to
ground, caps between inputs and outputs.

Method
M
th d off
Ti
Time
Constants
C
t t

Give expression for caps to ground

Give expression for caps between in and out of general block
then use this for CD stage Cgs only
then use this for CC stage Cbe only
then do for CE stage Ccb only
then work the full Darlington diff amp
then show how CE (with degen) CB CC are same problem
then re-show stage relationship

notes, M. Rodwell, copyrighted

Finding Bandwidth: Method of Time Constants

Ccbx = 6.4 fF
B

Ccbi = 3.7 fF

Rbb = 49
Vb'b'e

Q1

Q2

Rc = 6500

R =
350

g mVb 'e

C diff = C je =
182 fF 38 fF

Rex = 4.3

Vgen
E
B

Ccbx = 6.4 fF

Ccbi = 3.7 fF

Rbb = 49
Vb'e
C diff = C je =
182 fF 38 fF

Rc = 6500

R =
350

g mVb 'e
Rex = 4.3
E

take a general RC network (no inductors or delays

tau), and separate into 2 parts, network without
capacitors, and the capacitors:

notes, M. Rodwell, copyrighted

MOTC: Separation into Capacitors & Resistive N-port

The internal capacitor-free network is now frequencyindependent The MOTC method (not proven here) relies on
independent.
results from n-port network theory

MOTC: Open-Circuit Resistances

notes, M. Rodwell, copyrighted

R110 is
i the
h small
ll signal
i l resistance
i
measuredd at port one with
ih
all other ports open - circuited. This is determined by
applying a test voltage (or current) at the port and computing
from this the resulting current (or voltage)

MOTC: the Dominant Time Constant

The MOTC first - order time constants directly give us the

dominant time constant a1 of the circuit. If (and only if) the
secondary time constant a2 is negligible, the 3 - dB
bandwidth is 1 / 2a1. We must use the second - order
(short - circuit) time constants to determine a2 .

notes, M. Rodwell, copyrighted

MOTC: Are We Saving Any Work ?

notes, M. Rodwell, copyrighted

Are we saving work relative to brute-force nodal analysis:

MOTC would be of only moderate value if we had to calculate
all the Ri's each time. Fortunately, most terms involve quantities
already found in midband stage analysis: input and output
impedances, load impedances, etc.

MOTC: Short-Circuit Resistances

R112 is the small signal resistance measured at port one with

all other ports open - circuited,
except for port 2, which is shorted

notes, M. Rodwell, copyrighted

MOTC: The Second-Order Time-Constant

notes, M. Rodwell, copyrighted

MOTC: Working these Efficiently

notes, M. Rodwell, copyrighted

Because Rxxy R yy0 = Rxx0 R yyx , we always have 2 choices in

finding each term in the MOTC. The trick is to work the
problem so that as much as possible :
1)) terms are related to input,
p ouput,
p load impedances
p
2) terms are ones found earlier, in a1 analysis.
There are 2 " funny
funny"cases which arise so often that I will
give them on the next 2 pages (note these are intimately
related
l d to the
h well
ll - known
k
Miller
ill effect))

MOTC and the Miller Effect

notes, M. Rodwell, copyrighted

R = Ri (1 + Av ) + Rout
y
xx

Av

Ri

Rout
C

Ri

Av

Rout

a1 = = [Ri (1 + Av ) + Rout ] C

notes, M. Rodwell, copyrighted

MOTC: port impedances between collector and base

If we decide explicitly that Rx is to denote the parallel

combination of any external circuit resistances and Rbe ,
and that RLeq similarly denotes the combined effect of
external resistors and Rce , then
R yy0 = Rx (1 + g m RLeq ) + RLeq

MOTC: Port Impedances Between Emitter & Base

notes, M. Rodwell, copyrighted

MOTC: Multistage Example

work on the board...

notes, M. Rodwell, copyrighted

notes, M. Rodwell, copyrighted

E d
End