Electronic Circuits
Stanford University
Department of Electrical
Engineering
© 1997 Gregory T. A. Kovacs, All Rights Reserved
TABLE OF CONTENTS
Chapter 1: INTRODUCTION..........................................................................................1
1: OBJECTIVES.........................................................................................................1
2: EE113 AND EE122  A WINNING COMBINATION............................................... 1
3: COURSE INFORMATION......................................................................................2
4: OUTLINE OF THE COURSE................................................................................. 3
Chapter 2: BASIC CONCEPTS......................................................................................5
1: OBJECTIVES.........................................................................................................5
2: SOURCES .............................................................................................................5
3: SIGNALS ............................................................................................................... 6
4: AMPLIFIERS..........................................................................................................8
5: NOISE.................................................................................................................. 11
6: DISTORTION....................................................................................................... 12
7: AMPLIFIER POWER SUPPLIES & EFFICIENCY............................................... 14
8: LARGE AND SMALL SIGNALS ..........................................................................15
9: TRANSFER FUNCTIONS.................................................................................... 16
10: BODE PLOTS.................................................................................................... 20
Chapter 3: BJT REVIEW.............................................................................................. 25
1: OBJECTIVES....................................................................................................... 25
2: TYPES OF BIPOLAR JUNCTION TRANSISTORS............................................. 25
3: DC ANALYSIS (LARGESIGNAL) ....................................................................... 27
4: ANALYSIS OF BIASING ..................................................................................... 29
5: DC LOAD LINES.................................................................................................. 31
6: SMALLSIGNAL BEHAVIOR ............................................................................... 32
7: HYBRIDπ MODEL FOR (AC) SMALL SIGNALS................................................ 33
8: SCALING RESISTANCES BETWEEN BASE AND EMITTER............................ 35
9: AC LOAD LINES..................................................................................................36
Chapter 4: Operational Amplifiers............................................................................. 38
1: OBJECTIVES....................................................................................................... 38
2: OPERATIONAL AMPLIFIERS CONCEPTS........................................................ 39
3: BASIC OPAMP CIRCUITS................................................................................. 40
3.1 VOLTAGE FOLLOWER................................................................................. 40
3.2 INVERTING AMPLIFIER............................................................................... 41
3.3 LOGARITHMIC AMPLIFIER.......................................................................... 42
3.4 EXPONENTIAL AMPLIFIER.......................................................................... 42
3.5 NONINVERTING AMPLIFIER...................................................................... 43
3.6 SUMMING AMPLIFIER................................................................................. 43
3.7 INTEGRATOR............................................................................................... 44
3.8 DIFFERENTIATOR........................................................................................ 45
4: “REAL” VERSUS “IDEAL” OPAMPS.................................................................. 46
5: FREQUENCY RESPONSE OF OPAMPS.......................................................... 47
6: THE GUTS OF REAL OPAMP CHIPS............................................................... 50
7: A PREVIEW OF FEEDBACK .............................................................................. 53
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER
AMPLIFIER.................................................................................................................... 54
1: OBJECTIVES....................................................................................................... 54
2: OVERVIEW OF SINGLESTAGE BJT AMPLIFIERS.......................................... 54
3: THE COMMONEMITTER AMPLIFIER............................................................... 56
4: CLASSIC BIASING SCHEME FOR CE AMPLIFIERS......................................... 59
5: BASE CIRCUIT DESIGN STRATEGY  THE BASICS........................................ 59
6: LOCAL FEEDBACK THROUGH AN EMITTER RESISTOR................................. 60
7: BASE CIRCUIT ANALYSIS STRATEGY............................................................. 61
8: UNBYPASSED EMITTER RESISTANCE FOR AC AND DC FEEDBACK......... 62
9: EMITTER DEGENERATION................................................................................ 66
10: THE BOTTOM LINE ON DESIGN OF CE AMPLIFIERS...................................71
11: THE ART OF CE DESIGN (AN INTRO) ............................................................ 72
12: BIAS CIRCUITRY DESIGN................................................................................ 74
13: REVISITING THE LOAD LINES........................................................................ 75
14: DESIGN EXAMPLE #1: METHOD OF SEDRA & SMITH................................. 79
Chapter 6: BJT FREQUENCY RESPONSE ................................................................ 84
1: OBJECTIVES....................................................................................................... 84
2: JUNCTION CAPACITANCES OF BJT’S ............................................................. 84
3: DIFFUSION CAPACITANCE............................................................................... 87
4: COMPLETE HYBRIDp MODEL.......................................................................... 88
5: GAIN (β) VERSUS FREQUENCY FOR BJT'S .................................................... 89
6: COMMONEMITTER CONFIGURATION AND MILLER CAPACITANCE........... 93
7: USING THE MILLER IDEA UP FRONT............................................................... 95
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE................................. 97
1: OBJECTIVES....................................................................................................... 97
2: GENERALIZED CAPACITORCOUPLED AMPLIFIER FREQUENCY RESPONSE
.............................................................................................................................98
3: DOMINANT POLES............................................................................................. 99
4: SHORTCIRCUIT AND OPENCIRCUIT TIME CONSTANT METHODS FOR ......
APPROXIMATING THE RESPONSE OF AMPLIFIERS......................................... 100
4.1 OPENCIRCUIT TIME CONSTANTS FOR UPPER CUTOFF FREQUENCY
APPROXIMATION.............................................................................................101
4.2 SHORTCIRCUIT TIME CONSTANTS FOR LOWER CUTOFF FREQUENCY
APPROXIMATION.............................................................................................101
5: FREQUENCY RESPONSE OF THE CE AMPLIFIER ......................................102
5.1 LOW FREQUENCY RESPONSE................................................................ 103
5.2 MIDFREQUENCY RESPONSE .................................................................108
5.3 HIGHFREQUENCY RESPONSE...............................................................109
6: EXAMPLE FREQUENCY RESPONSE ANALYSIS........................................... 112
6.1 MIDBAND GAIN CALCULATION...............................................................112
6.2 LOWFREQUENCY RESPONSE................................................................ 113
Chapter 8: THE COMMON BASE AMPLIFIER .........................................................117
1: OBJECTIVES.....................................................................................................117
2: THE COMMON BASE CIRCUIT........................................................................118
3: MIDBAND GAIN CALCULATIONS................................................................... 118
4: LOWFREQUENCY RESPONSE...................................................................... 121
5: HIGHFREQUENCY RESPONSE .....................................................................122
6: CASCODE AMPLIFIERS = CE + CB.................................................................126
Chapter 9: THE COMMON COLLECTOR AMPLIFIER.............................................127
1: OBJECTIVES.....................................................................................................127
2: COMMON COLLECTOR CIRCUIT....................................................................128
3: PRACTICAL CC CIRCUIT................................................................................. 130
4: THE PHASE SPLITTER CIRCUIT.....................................................................132
5: QUICK LOOK AT CC FREQUENCY RESPONSE ............................................133
6: DESIGN OF CC STAGES.................................................................................. 134
Chapter 10: CASCADED AND CASCODE AMPLIFIERS.........................................135
1: OBJECTIVES.....................................................................................................135
2: AMPLIFIERS IN SERIES (CASCADED)............................................................ 136
3: THE CASCODE AMPLIFIER............................................................................. 138
4: PRACTICAL CASCODE AMPLIFIER CIRCUIT................................................. 140
5: CASCODE EXAMPLE .......................................................................................146
6: CASCODE AMPLIFIER DESIGN EXAMPLE #2 ............................................... 148
Chapter 11: DIFFERENTIAL AMPLIFIERS...............................................................153
1: OBJECTIVES.....................................................................................................153
2: DIFFERENTIAL AMPLIFIER BASIC CONCEPTS.............................................154
3: MODES OF OPERATION OF THE DIFFERENTIAL PAIR................................ 155
3.1 COMMONMODE OPERATION..................................................................156
3.2 “LARGE” SIGNAL DIFFERENTIALMODE OPERATION........................... 157
3.3 SMALLSIGNAL DIFFERENTIALMODE OPERATION..............................158
4: DETAILS OF SMALLSIGNAL OPERATION.....................................................159
5: HALFCIRCUIT MODEL OF DIFFERENTIAL PAIR ..........................................166
6: HALFCIRCUIT MODEL AND COMMONMODE OPERATION........................168
7: COMMONMODE REJECTION RATIO............................................................. 170
8: FREQUENCY RESPONSE OF THE DIFFERENTIAL AMPLIFIER................... 171
9: SUMMARY OF DIFFERENTIAL PAIR SMALLSIGNAL OPERATION............. 182
10: OTHER IMPERFECTIONS OF “REALISTIC” DIFFERENTIAL AMPLIFIERS. 183
10.1 INPUT OFFSET VOLTAGE...................................................................... 183
10.2 INPUT BIAS AND OFFSET CURRENTS................................................. 186
Chapter 12: CURRENT SOURCES ...........................................................................188
1: OBJECTIVES.....................................................................................................188
2: THE DIODECONNECTED TRANSISTOR....................................................... 189
3: THE CURRENT MIRROR.................................................................................. 190
4: A SIMPLE (LOUSY) CURRENT SOURCE........................................................ 191
5: THE WIDLAR CURRENT SOURCE..................................................................192
6: CURRENT MIRRORS........................................................................................ 193
7: NONIDEALITIES OF BJT CURRENT SOURCES............................................194
7.1 EFFECTS OF r
o
ON CURRENT SOURCE PERFORMANCE..................... 195
Chapter 13: MULTISTAGE AMPLIFIERS................................................................ 197
1: OBJECTIVE....................................................................................................... 197
2: ANALYSIS OF AN EXAMPLE AMPLIFIER........................................................ 197
Chapter 14: FEEDBACK............................................................................................203
1: OBJECTIVES.....................................................................................................203
2: INTRODUCTION TO FEEDBACK..................................................................... 204
2.1 PROPERTIES OF NEGATIVE FEEDBACK............................................... 204
2.2 THE BASIC FEEDBACK CIRCUIT..............................................................205
3: HOW FEEDBACK AFFECTS BANDWIDTH......................................................207
4: FROM BASIC BLOCK DIAGRAM TO ACTUAL FEEDBACK CIRCUITS.......... 209
4.1 REMINDER: TYPES OF AMPLIFIERS ......................................................209
4.2 BASIC STRUCTURE OF THE CIRCUIT.................................................... 211
4.3 TYPES OF MIXER...................................................................................... 212
4.4 TYPES OF SAMPLER................................................................................212
5: SERIESSHUNT FEEDBACK > VOLTAGE AMP............................................. 215
6: SHUNTSERIES FEEDBACK > CURRENT AMP............................................ 219
7: SERIESSERIES FEEDBACK > TRANSCONDUCTANCE AMPLIFIER.......... 221
8: SHUNTSHUNT FEEDBACK > TRANSRESISTANCE AMPLIFIER................ 223
9: FEEDBACK ANALYSIS IN REALISTIC CIRCUITS........................................... 224
9.1 SUMMARY OF STEPS YOU WILL USE.................................................... 224
10: SERIESSHUNT FEEDBACK.......................................................................... 225
11: RULES FOR SERIESSERIES (TRANSCONDUCTANCE AMPLIFIER) ........235
12: RULES FOR SHUNTSHUNT (TRANSRESISTANCE AMPLIFIER) ...............236
13: RULES FOR SHUNTSERIES (CURRENT AMPLIFIER)................................237
14: RULES FOR SERIESSHUNT (VOLTAGE AMPLIFIER) ................................238
15: STABILITY AND POLE LOCATION................................................................. 239
15.1 SINGLE POLE WITH FEEDBACK ...........................................................240
15.2 TWO POLES WITH FEEDBACK..............................................................241
15.3 THREE OR MORE POLES WITH FEEDBACK........................................ 243
15.4 STABILITY................................................................................................244
16: COMPENSATION............................................................................................ 245
17: "TENT" MODEL FOR VISUALIZING POLES AND ZEROS............................ 246
18: LINEAR OSCILLATORS (VERY BRIEFLY!).................................................... 247
Chapter 1: INTRODUCTION
Chapter 1: INTRODUCTION
Minimalism is in, and there’s nothing
more minimal than nothing...
Barden N. Shimbo, Former EE122 Student
1: OBJECTIVES
• Get some practical knowledge about the design and analysis of basic analog
circuits.
• Learn about operational amplifiers and circuits that use them.
• Find out how bipolar transistors really work in circuits.
• Learn how to design, analyze, and test basic amplifiers.
• Learn about differential pairs, current sources and multistage amplifier design.
• Learn how to design, analyze and test multistage amplifiers.
• Learn about feedback as it applies to amplifiers.
2: EE113 AND EE122  A WINNING COMBINATION
To the extent possible, EE113 and EE122 are parallel courses. In general, the lectures,
problem sets, and overall topics will be timed so that you will have the background
knowledge to cope with the laboratories. The point is that (finally) you get to apply what
you learn in the classroom to handson designing and testing of circuits. A potential
problem arises if you don’t keep up: you can be lost in two classes at once!
PLEASE DON’T GET LEFT BEHIND!!! IF YOU FEEL
THIS HAPPENING, PLEASE LET US KNOW!
THE COURSES ARE ROUGHLY SYNCHRONIZED, BUT
YOU WILL NEED TO READ AHEAD (TEXT AND NOTES)
SOMETIMES!
G. Kovacs ©1997 Page 1
Chapter 1: INTRODUCTION
3: COURSE INFORMATION
Instructor: Prof. Greg Kovacs
Office: CISX 202
Email: kovacs@cis
Phone: 7253637 (email greatly preferred for setting up
meetings)
Lectures: Skilling 193
MonWedFri , 10:00am  10:50 am
Office Hours: Mon 11:00 am  12:00 pm
(Subject to Change) Wed 2:00 pm  3:00 pm
Fri 1:00 pm  2:00 pm
(Or by appointment please.)
Admin. Assistant: Ms. Susan Kahn
CISX 203
Phone: 7230720
email: skahn@leland
Hours: 10:00 am  3:00 pm MTWTF
TA: Nolan Sharp
Email: nolans@leland
Office Hours TBD Review Sessions TBD
Texts/Notes: Sedra and Smith, “Microelectronic Circuits,” HRW
Notes handed out in class + supplements as needed.
References: Horowitz and Hill, “The Art of Electronics,” Cambridge Press
Savant, Roden, and Carpenter, “Electronic Design,” Benjamin
Cummings
Gray and Meyer, “Analysis and Design of Analog Integrated
Circuits,” 2nd Edition, Wiley
Neudeck, “PN Junction Diodes,” and “The Bipolar Junction
Transistor,” AddisonWesley
Muller and Kamins, “Device Electronics for Integrated Circuits,”
2nd Edition, Wiley
G. Kovacs ©1997 Page 2
Chapter 1: INTRODUCTION
Credit: 3 units, letter grade only
Grading: Homework = 10% assigned weekly, due at 5pm on date
marked on assignment in box at CISX 203. Please don’t be
late (penalties may apply).
Midterm Exam ± Quizzes = 35% (OPEN BOOK)
Final Exam = 55% (OPEN BOOK)
4: OUTLINE OF THE COURSE
INTRODUCTION AND REVIEW (handed out, but not covered in class)
BJT REVIEW (handed out, covered briefly in class if needed)
THE OPERATIONAL AMPLIFIER
COMMON EMITTER BJT AMPLIFIERS
GENERALIZED FREQUENCY RESPONSE ESTIMATION
COMMON BASE BJT AMPLIFIERS
COMMON COLLECTOR BJT AMPLIFIERS
CASCADED/CASCODED BJT AMPLIFIERS
DIFFERENTIAL AMPLIFIERS
CURRENT SOURCES
MULTISTAGE AMPLIFIERS
FEEDBACK: THEORY AND PRACTICE
G. Kovacs ©1997 Page 3
Chapter 1: INTRODUCTION
Analog Boot Camp Drill Routine
by G. Kovacs
(The words are first barked out by the professor,
then shouted back by students marching in
formation.)
Analog circuits sure are fine,
Just can’t get ‘em off my mind.
Digital circuits ain’t my kind,
Zeros and ones for simple minds.
I guess NAND gates aren’t all that bad,’
’Cause I need them for circuit CAD.
One, two, three, four,
Gain and bandwidth, we want more.
Five, six, seven, eight,
We don’t want to oscillate.
Widlar, Wilson, Brokaw too,
They’ve got circuits, how ‘bout you?
(repeat)
G. Kovacs ©1997 Page 4
Chapter 2: BASIC CONCEPTS
Chapter 2: BASIC CONCEPTS
I’m a bilingual illiterate...
I can’t read or write in two languages.
Former EE113 Student
1: OBJECTIVES
• To review sources, signals, amplifiers, transfer functions, and Bode plots.
• To discuss noise, distortion and large versus small signals.
READING:
READ S&S Sections 1.1  1.5
READ APPENDIX E in S&S!!!
2: SOURCES
Thévenin & Norton I
SC
=
V
OC
R
o
+

R
o
V
oc
I
sc
R
o
You should review this briefly and know how to convert back and forth. The Norton
form has a current source whose value is found by shortcircuiting the Thévenin form.
The Thévenin's voltage source is found by taking the opencircuit voltage of the Norton
form. The output resistance, R
o
, remains the same in both forms, but its orientation in
the circuit changes.
G. Kovacs ©1997 Page 5
Chapter 2: BASIC CONCEPTS
3: SIGNALS
Fourier Series  a way to represent periodic signals as a sum of sine and cosine
“harmonics”
x t = a
n
cos nω
o
t + b
n
sin nω
o
t
∑
n = 0
∞
One can determine the fourier coefficients (a
n
and b
n
) using the following equations:
a
o
=
1
T
x t dt

T
2
T
2
a
n
=
2
T
x t cos nω
o
t dt

T
2
T
2
b
n
=
2
T
x t sin nω
o
t dt

T
2
T
2
The DC term
For Even Functions
For Odd Functions
It is often useful to look at the function and decide if it’s even or odd.
Example 1: symmetrical square wave (odd > sine terms only)
x t =
4
π
1
n
sin nω
o
t
∑
n=1
∞
n's = 1,3,5... (odd)
b
n
's =
4
π
,
4
3π
,
4
5π
...
G. Kovacs ©1997 Page 6
(since the signal is odd, all a
n
’s are zero)
Chapter 2: BASIC CONCEPTS
Example 2: Full Wave Rectified Sinewave. This is a common signal found in power
supplies after the sinewave input power has been rectified using a diode or diodes and
if an output filter capacitor is not used.
x t =
2
π

4
π
cos nω
o
t
n
2
 1
∑
n=2,4,6,...
∞
RMS Value of an AC signal  the amount of DC power required to provide an equivalent
amount of heating in a resistive load... it is very useful when measuring the energy or
power in signals that might not be a wellknown waveform.
To compute the RMS, take the signal, square it, average it, and take the square root...
V
RMS
= AVG v t
2
=
1
T
v
2
t dt
0
T
Some useful RMS formulas:
Sinewave RMS =
V
peak
2
=
V
peaktopeak
2 2
Symmetrical Squarewave RMS = V
peak
=
V
peaktopeak
2
Triangle Wave RMS =
V
peak
3
=
V
peaktopeak
2 3
The AD536/AD636 AC/RMS Converter chip from Analog Devices directly computes the
RMS value of an analog signal and it, or a close relative, is included in most handheld
G. Kovacs ©1997 Page 7
digital multimeters.
Chapter 2: BASIC CONCEPTS
4: AMPLIFIERS
TYPES OF AMPLIFIERS
Amplifiers always increase (or at least maintain) the signal power. The gain of an
amplifier is expressed as a voltage gain, transconductance gain (voltage input, current
output), transresistance (current input, voltage output) or current gain. Thus, there are
four basic types of amplifiers, depending on what it is that they amplify (voltage or
current) and what it is that you want as their output (voltage or current).
One can model any amplifier as any of the four types, but the intended use of the
amplifier usually makes one choice usually the best. In other words, an amplifier is
usually designed to be a particular type.
Source
Parameter to
be Amplified
Desired
Output
Parameter
Type
of
Amplifier
Gain
Expression
source voltage, v
s
v
o Voltage
source voltage, v
s
i
o Transconductance
source current, i
s
v
o Transresistance
source current, i
s
i
o Current
A
v
=
v
o
v
s
= voltage gain
(dimensionless)
G
m
=
i
o
v
s
= transconductance
in Ω
1
or Siemens
R
m
=
v
o
i
s
= transresistance
in Ω
A
i
=
i
o
i
s
= current gain
(dimensionless)
G. Kovacs ©1997 Page 8
Chapter 2: BASIC CONCEPTS
SCHEMATICS OF THE BASIC AMPLIFIER TYPES
+

+

Ideal Voltage Amplifier
Ideal Transconductance Amplifier
Ideal Transresistance Amplifier
Ideal Current Amplifier
R
A
v
A
G
M
M
i
L
R
v
i
i
i
v
s
i
s
v
i
i
i
v
s
i
s
L
R
L
R
L
R
s
R
s
R
s
R
v
o
v
o
v
i
v
i
i
i
i
i
+

+

+

+

i
o
i
o
R
i
R
i
R
i
R
i
R
o
R
o
R
o
R
o
s
R
NOTE 1: in general use Z (for impedance) rather than R, since most inputs and outputs
are not purely resistive!
NOTE 2: R
S
is shown as a resistor at the input of the amplifier that effectively attenuates
the input signal if the amplifier is not ideal (i.e. if the voltage input amplifiers have input
resistances less than infinity or if the current input amplifiers have input resistances
G. Kovacs ©1997 Page 9
greater than zero).
Chapter 2: BASIC CONCEPTS
Type
of
Amplifier
Gain
Expression
Ideal
Input
Impedance
Ideal
Output
Impedance
Voltage
Z
i
= ∞ Z
o
= 0
Transconductance
Z
i
= ∞ Z
o
= ∞
Transresistance
Z
i
= 0 Z
o
= 0
Current
Z
i
= 0 Z
o
= ∞
A
v
=
v
o
v
s
= voltage gain
(dimensionless)
G
m
=
i
o
v
s
= transconductance
in Ω
1
or Siemens
R
m
=
v
o
i
s
= transresistance
in Ω
A
i
=
i
o
i
s
= current gain
(dimensionless)
BASIC AMPLIFIER SPECIFICATIONS
• GAIN is usually expressed in decibels in terms of the input and output parameters:
VOLTAGE GAIN = A
v
=
v
o
v
i
• Decibels (after Alexander Graham Bell) are a common unit of measure. POWER gain
is expressed as:
dB = 20 log
10
A
v
NOTE! if the input and output signals are already POWER, then use,
dB = 10 log
10
P
out
P
in
dB = 20 log
10
A
v
POWER GAIN = A
p
=
p
o
p
i
=
i
o
v
o
i
i
v
i
When calculating the overall gain of a cascaded amplifier, simply add up the gain (or
loss) of each stage in dB to get the overall gain.
G. Kovacs ©1997 Page 10
Chapter 2: BASIC CONCEPTS
5: NOISE
• NOISE is unwanted signal(s) that end up added to the desired signals.
• Noise can originate outside of an amplifier or come from inside of the amplifier.
• Interference > electromagnetic interference (EMI) is “pickedup” signals from external
noise sources such as household wiring, automobile ignitions, etc.
SOLUTION = shielding!
• Thermal Noise > noise from random motion of electrons in conductors (proportional
to temperature) > if the conductor is a resistor, the higher its value, the more noise
voltage is generated (thermal noise is also a function of the signal bandwidth).
v
n
RMS = 4kTR∆f
where k is Boltzmann’s constant (1.38 X 10
23
J/°K), T is the absolute temperature in °K,
R is the component’s resistance in Ω, and ∆f is the bandwidth of interest in Hz.
SOLUTION = use lowresistance values or cool your circuit
• Shot Noise > noise current that occurs in active semiconductor devices (BJT’s,
FET’s, etc.) due to the arrival and departure of individual carriers in the device...
i
n
RMS = 2eI∆f
where e is the charge on the electron (1.6 X 10
19
C), I is the DC current flowing, and ∆f
is the bandwidth of interest in Hz. This is proportional to the current.
SOLUTION = use lower currents.
• Flicker Noise > lowfrequency noise (1/f dependence of frequency spectrum) generated
by the random recombination of electrons and holes in semiconductors, or by fluctuations
in component values, generally only containing frequencies in the audio spectrum (bad
for stereo gear!).
G. Kovacs ©1997 Page 11
SOLUTION = move to another universe.
Chapter 2: BASIC CONCEPTS
• “White” noise is noise that has a flat frequency spectrum (i.e. contains all frequencies
in equal proportion). In practice, noise is only “white” over a finite bandwidth. The
sound from an FM receiver between channels is moreorless white.
 1 . 0
 0 . 5
0.0
0.6
1.1
500 400 300 200 100 0
x10
6
An actual "white noise" signal with bandwidth limited to 20 kHz.
• White noise can be really useful for determining the frequency response of circuits
using a spectrum analyzer  all frequencies are equally represented in the spectrum of
white noise, so you can input it into a circuit you are testing and look at which frequencies
come out! If you average over a long enough time, you can obtain a frequency
response for the circuit under test.
6: DISTORTION
• DISTORTION of a signal occurs when the amplified version of the signal coming out of
the amplifier is not simply a scaled copy of the input signal, but is differently shaped
(distorted).
• Distortion can be noted as a difference in waveform shape the ideal scaled copy of the
input, a difference in the spectrum of the input and output signals, or sometimes observed
by listening to the output of an amplifier (if it is used for audio).
• Distortion is due to nonlinearities, generally because the semiconductor (or tube)
amplifier are not perfectly linear. In some cases, distortion can come from amplifier
saturation ("clipping," or the amplifier simply reaching one of the voltage or current
G. Kovacs ©1997 Page 12
extremes beyond which it cannot swing).
Chapter 2: BASIC CONCEPTS
• Another common type of distortion in amplifiers that use both PNP and NPN transistors
at their outputs is crossover distortion, which is caused by the slight "gap" in voltage
between one type of transistor turning off and the other turning on.
• You can test for distortion by using a pure (single frequency) sinusoidal input signal
and looking at the output either visually or with a spectrum analyzer. A linear system
will only have the same frequency at its output. Nonlinearities will give rise to harmonics
(signals at frequencies other than the one input to the system) which are measurable
with the spectrum analyzer (and sometimes by your eye on an oscilloscope screen).
• The term total harmonic distortion (THD) represents the percentage of the total
output signal of an amplifier that is at frequencies other than the one put in... in other
words, you drive the amplifier with a pure sinewave at a frequency f
o
and make a ratio of
the power in the harmonics (i.e. sum of signal frequencies other than f
o
, with amplitudes
given by A
i
(f
i
)) to the input signal power.
%THD ·
A
i
f
i
( )
i·1
∞
∑
A
o
f
o
( )
×100%
• In practice, one does not add up harmonic amplitudes to infinite frequency, but through
the range of interest (e.g. up through 20 kHz for audio).
• Other amplifier specifications such as frequency response (bandwidth), gain, etc., will
G. Kovacs ©1997 Page 13
be discussed below.
Crossover Distortion
Chapter 2: BASIC CONCEPTS
7: AMPLIFIER POWER SUPPLIES & EFFICIENCY
• All amplifiers need some type of power supply to supply the extra energy that is
delivered to the load.
• Most analog amplifiers use two power supply voltages or “rails,” as shown below,
V+
V
R
L
v
i
i
i
o
i
I
dc
• Some amplifiers use only a single power supply voltage, but sometimes they internally
"split" that single voltage into two rails by making an artificial "ground" voltage half way
from "real ground" to the supply voltage.
• The efficiency of an amplifier reflects the amount of power delivered to the load as a
fraction of the total power drawn from the power supply, and can be computed using:
η =
P
L
P
dc
× 100% =
Power Delivered to Load
Power Used from Power Supplies
× 100%
G. Kovacs ©1997 Page 14
Chapter 2: BASIC CONCEPTS
8: LARGE AND SMALL SIGNALS
• Most circuits are linear if the input signals are small enough! If the signal
amplitude is increased enough some type of nonlinearity will make itself obvious! All
semiconductor devices (and vacuum tubes!) are very nonlinear, and the only reason we
get nice, clean amplifier outputs is that we are keeping signal swings small enough
through various techniques.
• Examples of large signal effects (as discussed above in "Distortion"):
• Amplifier clipping (saturation) > here you have a case where the amplifier’s output
cannot swing above and below certain maximum and minimum voltages (that makes
sense)... you have probably heard clipping when someone turned up a stereo too loud!
• Amplifier distortion due to transistor nonlinearities > this is simplest to understand by
considering that basically, all transistors are nonlinear devices and we work very hard to
“coax” linearity out of them over certain ranges of signal level... this type of distortion
can be minimized but can never be completely avoided.
• Amplifier exploding (very nonlinear) due to extremely large input signal:
When we talk about transfer functions, AC small
signal equivalent models, Bode plots, etc., we are
always assuming that the circuit is in smallsignal
operation!
G. Kovacs ©1997 Page 15
Chapter 2: BASIC CONCEPTS
9: TRANSFER FUNCTIONS
• You already know what these are if you have ever looked at stereo equipment catalogs
or specification sheets, since they are plotted to illustrate the frequency range over
which the amplifier will operate properly.
• Amplifiers are either DC or AC coupled, meaning that the inputs are sensitive to both
DC and AC signals ("DCcoupled") or only AC signals ("ACcoupled). You should note
that oscilloscope amplifiers always have the option of choosing one or the other coupling
modes.
• You can use capacitors at the inputs and outputs of amplifiers to “block” DC signals,
as long as they are large enough to "look like shorts" in the frequency range of interest...
This type of amplifier is called AC or capacitivelycoupled.
• The frequency response of such an ACcoupled circuit cannot extend to zero Hz.
• Here is a brief complex number review for your reference so you can compute transfer
functions of circuits:
V = Re + jIm = Ae
jθ
Re = Acos θ
Im = Asin θ
V = Re
2
+ Im
2
θ = arctan
Im
Re
Remember that s = σ + j ω
POLES & ZEROS > WHAT DO THEY MEAN?
Transfer Function Notation:
T s =
v
out
s
v
in
s
= A
o
s  z
1
s  z
2
s  z
N
s  p
1
s  p
2
s  p
M
z
1
through z
N
are the “zeros,” or the complex frequencies at which the numerator
becomes zero (the transfer function goes to zero)...
p
1
through p
M
are the “poles,” or the complex frequencies at which the denominator
becomes zero (the transfer function goes to infinity)...
G. Kovacs ©1997 Page 16
Chapter 2: BASIC CONCEPTS
• The “order” of the system = N + M
• A zero alone would make the output amplitude of a circuit increase forever with
increasing frequency. In real circuits, one cannot have a zero without a pole to “cancel”
it out at some frequency.
SIMPLE FILTERS
• The cutoff (or 3dB) frequency is the point at which the response is 3 dB lower than in
the passband ( ≈ 0.707 times the passband amplitude).
HERE IS A GOOD IDEA TO GET A SENSE FOR CIRCUIT BEHAVIOR: Look at the
circuit first before doing any math! The capacitors are all infinite impedance for
DC and their impedance decreases toward zero as the frequency increases.
FIRSTORDER RC LOWPASS FILTER:
V V
in out
R
C
Hs =
1
Cs
R +
1
Cs
=
1
RCs + 1
=
1
1 +
s
ω
o
• The general form is Hs =
K
1 +
s
ω
o
OR
Kω
o
ω
o
+ s
=
A
ω
o
+ s
where K is the gain for low frequencies and A = Kω
o
• The pole is at S = ω
o
• The “cutoff” frequency is at ω
0
= 2πf
c
=
1
RC
G. Kovacs ©1997 Page 17
Chapter 2: BASIC CONCEPTS
Gain and phase responses of the firstorder, RC lowpass filter:
0.707 V
 45°
1 KHz
FIRSTORDER RC HIGHPASS FILTER:
R
C
V
V
in
out
Hs =
R
R +
1
Cs
=
RCs
RCs + 1
=
s
s +
1
RC
=
s
s + ω
o
• The general form is Hs =
Ks
s + ω
o
where K is the gain for high frequencies
• The pole is at S = ω
o
and the zero is at S = 0
G. Kovacs ©1997 Page 18
• The cutoff frequency is the same as for the lowpass filter.
Chapter 2: BASIC CONCEPTS
Gain and phase responses of the firstorder, RC highpass filter:
0.707 V
+ 45°
1 KHz
G. Kovacs ©1997 Page 19
Chapter 2: BASIC CONCEPTS
10: BODE PLOTS
This material is covered in the prerequisites to EE113, but typically, people are a bit
"shaky" on phase plots. The key point is that the effect of poles or zeros starts to
take place a decade before or a decade after the pole or zero frequency when you
are dealing with phase (unlike at the frequency for gain plots)!
If you are already familiar with all of this, you can skip to the next section.
THE BASIC IDEA:
The point is to be able to draw a "quick" sketch of a transfer function of a circuit.
It is assumed that you have an equation for the transfer function.
This is an important technique, despite the availability of computers!
You will need to be able to do Bode plots on exams and in some "real world"
situations.
Remember that you can use computer programs like Matlab™, Mathematica™, Theorist™,
etc. to check your work!
ADD up the individual responses of all of the poles and zeros of the transfer function.
They each affect the frequency response “only” after they take effect at their respective
“break” frequencies....
• a pole makes the amplitude fall with frequency by 20 dB /decade and “has no effect”
before its break frequency
• a zero makes the amplitude rise with frequency by 20 dB/decade and “has no effect”
before its break frequency
• a pole causes the phase to fall from 0° to 90° over two decades of frequency
starting one decade before the break frequency > the phase is 45° at its break
frequency
• a zero causes the phase to rise from 0° to +90° over two decades of frequency
starting one decade before the break frequency > the phase is +45° at its break
frequency
• the effects of poles or zeros at “zero” frequency have already “maxed out” by
the time you start your plot (i.e. phase is 90°for a pole at zero frequency, and +90°
G. Kovacs ©1997 Page 20
for a zero at zero frequency)
Chapter 2: BASIC CONCEPTS
HOW TO MAKE A GAIN PLOT
1) Write the transfer function equation in a form so that you can see the break
frequencies of the poles and zeros.
2) Try to begin the sketch at a frequency where you know the gain (from looking at the
equation).
If it is not obvious, draw a rough draft of the curve and select the frequencies corresponding
to the "flat" parts, plug those frequencies in for "S", estimate the gain and convert to dB.
Remember about poles and/or zeros that may have already “taken effect” at low enough
frequencies that they are “maxed out” before you start your sketch.
3) For each zero, add a +20 dB/decade slope to the slope of the sketch at the break
frequency of that zero.
4) For each pole, add a 20 dB/decade slope to the slope of the sketch at the break
frequency of that pole.
5) Draw a “smooth” curve over the sketch (the curves differ by about 3 dB at each
single break)...
HOW TO MAKE A PHASE PLOT
1) Write the transfer function equation in a form so that you can see the break
frequencies of the poles and zeros.
2) Try to begin the sketch at a frequency where you know the phase (from looking at
the equation).
Remember about poles and/or zeros that may have already “taken effect” at low enough
frequencies that they are “maxed out” before you start your sketch.
One way to make it easier is to start out assuming 0°at "super low" frequencies, then
shift the whole phase sketch:
a) + 90°for any zeros at "zero frequency"
b)  90°for any poles at "zero frequency"
c) +/ 180°if there is a negative sign
Remember that a negative sign on a gain is a 180°phase shift!
G. Kovacs ©1997 Page 21
Chapter 2: BASIC CONCEPTS
3) Each zero contributes a phase slope of +45° per decade starting one decade
below and lasting through one decade above the break frequency.
The phase contribution from that zero is “half way there” (or contributing +45°) at
the break frequency.
The contribution of that zero to phase at frequencies less that one tenth of the break
frequency and greater than ten times the break frequency is zero!
4) Each pole contributes a phase slope of 45° per decade starting one decade
below and lasting through one decade above the break frequency.
The phase contribution from that pole is “half way there” (or contributing 45°) at
the break frequency.
The contribution of that pole to phase at frequencies less that one tenth of the break
frequency and greater than ten times the break frequency is zero!
5) Draw a “smooth” curve over the sketch (the curves differ by about 6°at each single
break).
G. Kovacs ©1997 Page 22
Chapter 2: BASIC CONCEPTS
EXAMPLE 1: Hs =
10
5
s + 10
s + 10
3
120 dB
100 dB
80 dB
60 dB
40 dB
20 dB
0 dB
20 dB
1 10 10
10 10 10 10 10 10
2 3 4 5 6 7 8
10
1
EXAMPLE 2: Typical CapacitorCoupled Amplifier
Hs = 100
s
s + 10
3
1
1 +
s
10
6
=
10
8
s
s + 10
3
s + 10
6
Here we have a zero at zero (‘maxed out” by the time you start your plot), a pole at 10
3
radians/S and a pole at 10
6
radians/S.
In the midband, the first pole will have “cancelled” the zero and the gain should be flat at
100...
G. Kovacs ©1997 Page 23
Chapter 2: BASIC CONCEPTS
Below the first pole frequency, you should only need to consider the zero...
120 dB
100 dB
80 dB
60 dB
40 dB
20 dB
0 dB
20 dB
1 10 10
10 10 10 10 10 10
2 3 4 5 6 7 8
10
1
G. Kovacs ©1997 Page 24
Chapter 3: BJT REVIEW
Chapter 3: BJT REVIEW
Without bipolar transistors, the world
would be a pretty dull place!
Shockley, Bardeen and Brattain, Inventors of the bipolar transistor
(they didn't really say it).
1: OBJECTIVES
BJT operation
BJT DC Analysis and DC Load Lines
BJT smallsignal model > hybrid π
Looking “in” each terminal
AC Load Lines
READ S&S Chapter 4, Sections 4.1  4.9
2: TYPES OF BIPOLAR JUNCTION TRANSISTORS
• There are two polarities: NPN (“not pointin’ in”) and PNP (“pointin’ in”), where the
“pointin’” part refers to the direction of the arrow on the emitter terminal.
• BJT’s can operate as amplifiers (active mode) or switches (saturation = “ON,” cutoff =
“OFF”)
Mode EBJ CBJ
Cutoff Reverse Reverse
Active Forward Reverse
Saturation Forward Forward
G. Kovacs ©1997 Page 25
Chapter 3: BJT REVIEW
G. Kovacs ©1997 Page 26
Chapter 3: BJT REVIEW
3: DC ANALYSIS (LARGESIGNAL)
• The collector current versus V
BE
for a BJT is given by,
I
C
= I
s
e
qV
BE
kT
 e
qV
BC
kT
I
B
=
I
s
β
F
e
qV
BE
kT
 1 +
I
s
β
R
e
qV
BC
kT
 1
where, in the lower equation, the forward current gain, β
F
and the reverse current gain,
β
R
are considered to account for all possible operating modes. Generally, as seen
below, we only use "β" and assume we are talking about β
F
.
• For normal operation (assuming NPN), V
BE
> 0 and V
BC
< 0 so one can neglect the
second exponential terms in both equations.
• I
S
is the “current scale factor” which depends on the size of the transistor (in particular,
the geometrical area of its emitter) and the process by which it was made.
• If it is certain that V
BC
< 0 (basecollector junction is reverse biased) then one can use
simplified forms of the above equations, simply neglecting the β
R
terms and substituting
the generic β for β
F
.
I
C
= I
s
e
qV
BE
kT
= I
s
e
V
BE
V
T
I
S
may be on the order of 10
14
to 10
15
A
I
B
·
I
C
β
where kT/q ≈ 26 mV at room temperature.
• Another essential equation relates the emitter current to the base current:
G. Kovacs ©1997 Page 27
I
E
· I
C
+ I
B
· 1+ β ( )I
B
≈ βI
B
if β is large!
Chapter 3: BJT REVIEW
• The plot above shows the three regions of BJT operation: active (where analog
amplifiers typically operate), saturation (when the transistor is “fully on”) and cutoff
(where the transistor is “fully off”).
• The plot above shows the exponential relationship between I
C
and V
BE
... the point is
that tiny changes in V
BE
cause huge changes in I
C
.
• Note that I
B
vs. V
BE
curve is just a scaled copy of the I
C
vs. V
BE
curve.
G. Kovacs ©1997 Page 28
Chapter 3: BJT REVIEW
4: ANALYSIS OF BIASING
(Be careful... NPN & PNP are different for DC biasing! Here we look at NPN’s)
1) Typically, you start out assuming that the baseemitter junction is forward biased (but
later should verify that!)... V
BE
≈ +0.7V
2) Check that the basecollector junction is reverse biased: V
BC
< 0 (i.e. if so, then the
transistor is not saturated). If V
BC
< 0 and if I
B
is uniquely determined, then I
C
= β I
B
In practice, we generally assume that a transistor is not fully saturated until V
CE
≈ 0.2V...
this means that V
BC
can be as much above zero as 0.5V before “full” saturation (see the
curves above), but we really don’t want to cut it that close for a real amplifier because
the I
C
vs V
CE
curves start to really bend there, leading to distortion .
CC
v
o
v
s
R
B
R
C
V coupling
capacitor
to block
DC...
50 kΩ
+15V
500 Ω
• Assume ß = 100...
• Start out assuming active mode, so V
BE
= 0.7V.
I
B
·
15− 0.7
50kΩ
· 0.286 mA
I
C
· 100I
B
·28.6 mA
V
CE
· 15V− 28.6 mA ( ) 500 Ω ( ) · 0.7 V
• This means that the basecollector junction is at zero volts (on the edge between
reverse and forward biased, but still not “on”)... this means that the input signal would
not have to swing very far to put the transistor into saturation (not good for an amplifier!).
• What if β > 100? For example, if β = 200, V
BC
> 0 !!! We don't want a forward biased
G. Kovacs ©1997 Page 29
BC junction...
THIS IS NOT A GREAT
BIASING SCHEME... IT
IS VERY SENSITIVE TO
VARIATIONS IN ß.
Chapter 3: BJT REVIEW
CC
V = +15 V
V = 15 V
EE
C
V
E
V
R
C
R
E
R
C
= 1KΩ R
E
= 1KΩ β
F
= 100
V
E
=  0.7 V I
E
=
0.7  15V
1K
= 14.3mA
I
C
= 14.3mA
β
β + 1
= 14.3mA
100
101
= 14.16mA
V
C
= 15V  14.16mA 1K = 0.84V
V
CE
= _______
V
BC
= _______
• The answers are: V
CE
= 0.84  (0.7) = 1.54V (not saturated) and V
BC
= 0  0.84 = 0.84
(reverse biased).
• What if ß is 200 now?... not much happens > this is better biasing!
A MORE REALISTIC BIAS CIRCUIT (COVERED IN THE CE AMPLIFIER
SECTION):
CC
v
s
R
B2
R
C
V
B
V
V
EE
R
E
R
B1
I
B
I
B1
I
B2
v
o
G. Kovacs ©1997 Page 30
Chapter 3: BJT REVIEW
5: DC LOAD LINES
• This is a graphical approach to solving two simultaneous equations: the current vs.
voltage characteristic of the resistor in the collector circuit (just Ohm’s Law) and the I
C
vs
V
CE
characteristic of the transistor (for a given base current).
• The intersection of the two current vs. voltage curves is the quiescent point (“Qpoint”),
which is where the amplifier “idles” with no AC input signal.
G. Kovacs ©1997 Page 31
Chapter 3: BJT REVIEW
6: SMALLSIGNAL BEHAVIOR
• The goal when modeling smallsignal behavior is to make a model of a the transistor
that works for signals small enough to “keep things linear” (i.e. not distort too much).
• The basic trick is to linearize the very nonlinear exponential relationship between V
BE
and I
C
by looking at a small enough region of the exponential... The transistor’s behavior
is almost linear if the region is quite small...
I
C
+ i
c
= I′
C
= I
s
e
V
BE
+ v
be
V
T
DC + AC current
= I
s
e
qV
BE
kT
e
v
be
V
T
= I
C
e
v
be
V
T
= I
C
1 +
v
be
V
T
+ ... expand....
I′
C
= I
C
+
I
C
V
T
v
be
= I
C
+ g
m
v
be
linearize...
This assumes v
be
<<
kT
q
= V
T
≈ 25.9 mV
g
m
≡
I
C
V
T
THE TRANSCONDUCTANCE
G. Kovacs ©1997 Page 32
Chapter 3: BJT REVIEW
• Now one can look at i
b
versus v
be
... want to look at it as an equivalent input resistance,
r
π
, that models the input of the transistor for small signals.
• The AC component of the collector current from above is,
i
c
· g
m
v
be
and
i
c
· βi
b
therefore,
βi
b
· g
m
v
be
→ r
π
·
v
be
i
b
·
β
g
m
• The concept of g
m
and r
π
can be usefully combined to obtain the
7: HYBRIDπ MODEL FOR (AC) SMALL SIGNALS
C B
E
g
m
v
be
r
π
v
be
i
c
i
b
i
e
+

• This simple model for the BJT can be plugged into amplifier circuits to figure out gain,
etc. More detailed versions include parasitic capacitances and resistances.
• REMEMBER that for AC analysis, the DC power supply voltages get shorted to
ground, capacitors become shortcircuits, and current sources (other than those
defined as AC signals, as above) become opencircuits!
G. Kovacs ©1997 Page 33
AND
AND
BECOME
SHORTS!
BECOME
OPENCIRCUITS!
Stanford
Chapter 3: BJT REVIEW
NOTE that the smallsignal equivalent circuit shown
above works for NPN or PNP transistors with NO
CHANGE OF POLARITY!!!
It is CRITICAL, however, to consider the differences
between NPN and PNP transistors when analyzing the
DC bias conditions!
It is important to keep in mind the effect that the DC
bias conditions (particularly I
C
) have on the AC model...
basically, I
C
directly controls g
m
and hence the gain
when used in an amplifier circuit!
G. Kovacs ©1997 Page 34
Chapter 3: BJT REVIEW
8: SCALING RESISTANCES BETWEEN BASE AND
EMITTER
• The resistance between base and emitter depends on which way you look!
• The conceptual explanation is: from the base’s viewpoint, a small change in voltage
causes a relatively small change in base current, but from the emitter’s viewpoint, a
small change in voltage (a change in v
be
) causes a much larger change in current due to
the g
m
generator!
•This means the emittertobase resistance, r
e
seems much smaller... looking in through
the emitter, a small change in the voltage relative to the base causes a huge amount of
current to flow (compared to the same change in voltage applied looking into the base)
because you get not only the base current flowing through r
π
, but also the extra current
from the transistor’s amplifying action.
• Looking into the base toward the emitter, one “sees” r
π
.
C B
E
g
m
v
be
r
π
• Looking into the emitter toward the base, one “sees” r
e
. For a change in emitter
voltage, (β + 1) times as much current change will occur than for a comparable voltage
change at the base. Therefore,
r
e
=
r
π
β + 1
also... r
e
≈
r
π
β
=
1
g
m
G. Kovacs ©1997 Page 35
Chapter 3: BJT REVIEW
KEY POINT: You can “transform” a resistance from the emitter or base
side of the baseemitter circuit to the opposite side by multiplying by (β + 1)
if looking in the base, or dividing by (β + 1) if looking in the emitter.
9: AC LOAD LINES
G. Kovacs ©1997 Page 36
Chapter 3: BJT REVIEW
• The idea here is to extend the DC load line concept to allow for “wiggling” the base
current using an applied (AC) signal and looking at the corresponding “wiggling” of V
CE
along the constraining line of the resistor (Ohm’s Law) to see what the output of the
amplifier is doing (i.e. how much it is amplifying).
• The plot below shows how the base current, i
b
, varies with v
be
. Such variations in i
b
lead to variations in collector current (and finally output voltage) by causing the movement
between the various I
C
vs V
CE
curves shown above (movement from one line to the
other on the plot is constrained by the output resistor load line, giving rise to the output
voltage waveforms shown).
v
be
B
i
Base
Bias
Current
(DC)
B
I
B2
i
B1
i
b
i
v
be
V
BE
Quiescent BaseEmitter Voltage
Nearly Linear
Small Segment
of Curve
Quiescent
Point
0
Time
G. Kovacs ©1997 Page 37
Chapter 4: OPERATIONAL AMPLIFIERS
Chapter 4: Operational Amplifiers
Opamps are great, opamps are neat,
between you and me, they just can’t be beat...
Former EE113 BrownNosing Student
Presently employed as freelance poet.
1: OBJECTIVES
• To learn:
what an opamp is...
the basic opamp circuits
the differences between “ideal” and “real” opamps
the frequency response of opamps
a bit about feedback
READ S&S CHAPTER 2
G. Kovacs ©1997 Page 38
Chapter 4: OPERATIONAL AMPLIFIERS
2: OPERATIONAL AMPLIFIERS CONCEPTS
• Opamps are amplifiers that provide the amplified difference between two input
signals as their output...
+


+
V2
V1
A(V2V1)
1) The input impedance is infinite  i.e. no current ever flows into either input of the
opamp.
2) The output impedance is zero  i.e. the opamp can drive any load impedance to
any voltage.
3) The openloop gain (A) is infinite.
4) The bandwidth is infinite.
5) The output voltage is zero when the input voltage difference is zero.
COMMENTS ON FEEDBACK (more later!)
• The gain of the circuit is made less sensitive to the values of individual components.
• Nonlinear distortion can be reduced.
• The effects of noise can be reduced.
• The input and output impedances of the amplifier can be modified.
• The bandwidth of the amplifier can be extended.
G. Kovacs ©1997 Page 39
Chapter 4: OPERATIONAL AMPLIFIERS
3: BASIC OPAMP CIRCUITS
3.1 VOLTAGE FOLLOWER
V

V
OUT
V
+
V
IN
• Note that the opamp can swing its output in either direction to keep the voltage
difference between the input zero.
V

= V
OUT
substitute into the basic opamp equation to get...
V
OUT
= AV
+
 V

⇒ V
OUT
= AV
IN
 V
OUT
⇒ 1 + A V
OUT
= AV
IN
which yields, V
OUT
=
A
1 + A
V
+
⇒ V
OUT
= V
+
as A → ∞
• Here the input impedance is that of the opamp itself (very high).
• What about this configuration? Does it work?
V

V
OUT
V
+
V
IN
• Well, if you do the math, it looks like it should...
V
OUT
· A V
+
− V
−
( ) · A V
OUT
− V
−
( )
V
OUT
·
A
1−A
¸
¸
_
,
V
−
→ V
OUT
· −V
−
as A→∞
• This would work only if the opamp had no poles (response flat out to infinite
frequency)... if you have any poles, they can move to the right halfplane and the whole
G. Kovacs ©1997 Page 40
thing oscillates (positive feedback).
Chapter 4: OPERATIONAL AMPLIFIERS
3.2 INVERTING AMPLIFIER
R2
R1
i
i n
i
fb
V
OUT
V
IN
V

V
+
• By definition, v
out
· A v
+
− v
−
( ) and thus, v
+
− v
−
·
v
out
A
which means that as A
approaches infinity, v
+
−v
−
( ) →0 (this is the virtual ground assumption). Since, for
an ideal opamp, the input impedance is infinite, the input current is zero. This means
that the current in through R
1
must come out via R
2
and none enters the v

terminal.
This means that one can write,
v
in
R
1
+
v
out
R
2
· 0 which gives A
V
·
v
out
v
in
· −
R
2
R
1
• Note that the input impedance is simply R
1
since it is connected to a virtual ground...
KEY POINT: With negative feedback (as is generally
used with opamps except as oscillators or
comparators), the two input terminals are forced
to the same potential by the feedback.
If v
+
is grounded, v

is also forced to ground. This
is a common configuration, and we refer to the v
terminal as a “virtual ground.”
If v
+
is not grounded, v

is still forced to the same
voltage that is applied to v
+
.
This simple rule makes opamp analysis quite
simple in many cases.
G. Kovacs ©1997 Page 41
Chapter 4: OPERATIONAL AMPLIFIERS
3.3 LOGARITHMIC AMPLIFIER
• Replace R
2
of the Inverting Amplifier with a diode....
R1 R1
i
i n
i
fb
V
OUT
V
IN
V

V
+
• Start with the diode transfer function,
I
diode
· I
s
e
v
nv
T
−1
¸
¸
_
,
≈ I
s
e
v
nv
T
thus, the feedback current is i
fb
≈ I
s
e
v
−
−v
out
nv
T
(assume n = 2 for discrete diodes)
• Using the virtual ground assumption,
i
in
·
v
in
R
1
and i
fb
≈ I
s
e
−v
out
nv
T
(remember that for an inverting amp, i
fb
· −
v
out
R
2
)
and their sum must be equal to zero since no current enters the v

terminal.
• We also know that,
v
out
· −nv
T
ln
i
fb
I
S
¸
¸
_
,
· −nv
T
ln i
fb
( ) − ln I
S
( ) [ ]
• Since i
fb
= i
in
, one can substitute i
in
=
v
in
R
1
for i
fb
...
v
out
· −nv
T
ln
v
in
R
1
¸
¸
_
,
− ln I
S
( )
¸
1
]
1
• In practice, this circuit would probably not be used for a log amplifier. Instead, one
would use a groundedbase transistor in place of the diode to eliminate the effect of the
“n” term (dependent upon the current through the diode) and also take advantage of
the fact that the exponential iv relationship extends over a much larger range for
transistors.
G. Kovacs ©1997 Page 42
Chapter 4: OPERATIONAL AMPLIFIERS
3.4 EXPONENTIAL AMPLIFIER
• Replace R
1
of the Inverting Amplifier with a diode.
R1
R2
i
i n
i
fb
V
OUT
V
IN
V

V
+
• Note that i
in
≈ I
s
e
v
in
nv
T
· −
v
out
R
2
and thus, v
out
· −R
2
I
s
e
v
in
nv
T
• With exponential and logarithmic functions, one can multiply and divide if one can add
and subtract (this is easy with opamps, as discussed below)! This is accomplished as
follows:
A× B · e
ln A ( ) +ln B ( ) [ ]
A
B
· e
ln A ( )−ln B ( ) [ ]
• This principle, along with the addition, subtraction, integral and derivative opamp
functions described below, were the basis for analog computers.
3.5 NONINVERTING AMPLIFIER
R1
R2
i
1
i
fb
V
OUT
V
IN
V

V
+
• Since feedback is operating, the opamp tries to hold the difference between its input
terminals at zero (remember the idea behind the virtual ground assumption... this is
similar). Again, if you write the basic opamp equation, and substitute v
in
for v
+
,
v
out
· A v
+
− v
−
( ) · A v
in
− v
−
( ) →
v
out
A
· v
in
− v
−
( )
Thus, as the opamp's gain A becomes very large, the difference between vin and v

must go to zero. This means that v
in
= v

if A goes to infinity. In other words, the voltage
G. Kovacs ©1997 Page 43
at the v

terminal is forced to equal the input voltage.
Chapter 4: OPERATIONAL AMPLIFIERS
• Thus, start with the assumption that, v
−
· v
+
· v
in
, and, considering the fact that no
current will enter the opamp terminals, write,
v
out
− v
in
R
2
−
v
in
R
1
· 0 which can be rearranged to yield,
v
out
v
in
· 1+
R
2
R
1
¸
¸
_
,
• This circuit is very useful when you need to use the full input impedance of the
opamp, such as in instrumentation. (Of course, you can obtain a noninverting amplifier
by connecting two inverting amplifiers in series, but that uses on additional amplifier.)
3.6 SUMMING AMPLIFIER
Rf
R3
R2
R1
R4
Rn
i
fb
V
OUT
V
1
V
V

+
V
2
V
3
V
4
V
n
i
1
• This amplifier's operation can be "summed up" by this equation, indicating that the
input and feedback currents must add up to zero (no current enters the operational
amplifier's input terminals), i
k
+i
fb
· 0
k·1
n
∑
• Summing currents, one can write,
v
out
R
f
· −
v
1
R
1
+
v
2
R
2
+
v
3
R
3
+L+
v
n
R
n
¸
¸
_
,
Which can be rearranged to give,
v
out
· − v
1
R
f
R
1
+ v
2
R
f
R
2
+ v
3
R
f
R
3
+L+ v
n
R
f
R
n
¸
¸
_
,
which
has the form of a series of "n" inverting amplifiers, whose outputs are summed together.
• The summing amplifier is often used as an audio mixer, where the inputs are voltages
from several microphones or other sources, and the input resistors, R
1
••• R
n
are varied
to control the individual channel gains.
G. Kovacs ©1997 Page 44
• This circuit can also be used as an averager if R
1
= R
2
= ••• = R
n
and R
f
= (R
1
/n)
Chapter 4: OPERATIONAL AMPLIFIERS
3.7 INTEGRATOR
R2
R1
C1
V
OUT
V
V

+
V
IN
i
i n
• For DC, it is just like an inverting amplifier,
v
out
v
in
· −
R
2
R
1
• For higher frequencies, the capacitor begins to matter,
v
in
R
1
· −C
1
dv
out
dt
• It is simplest to replace R
2
for the inverting amplifier with the impedance of the parallel
combination of R
2
and C
1
,
v
out
v
in
· −
Z
2
R
1
· −
1
C
1
S
R
2
¸
¸
_
,
R
1
· −
R
2
C
1
S
R
2
+
1
C
1
S
¸
¸
_
,
R
1
· −
R
2
R
2
C
1
S+1
¸
¸
_
,
R
1
· −
R
2
R
1
¸
¸
_
,
R
2
C
1
S+ 1
• The form of this equation is a lowpass filter with a cutoff frequency of f
c
·
1
2πR
2
C
1
(we will refer to it as f
min
) and a DC gain of
v
out
v
in
· −
R
2
R
1
• This is, in fact, an integrator for frequencies above f
c
, and an inverting amplifier for
frequencies below f
c
. (When you see capacitors in a circuit, always consider what
happens at zero and infinite frequency!)
G. Kovacs ©1997 Page 45
Chapter 4: OPERATIONAL AMPLIFIERS
• For the theoretical “ideal” integrator circuit, R
2
would be omitted, and the equation
would be,
v
out
· −
1
R
1
C
1
v
in
dt
∫
• However, small DC offsets at the input (and those of realistic, "nonideal" opamps)
would quickly "charge up" the integrator... remember that the integral of a constant is
the constant times time! We want to limit the integrator's theoretical infinite gain at DC
to something less.
• For the realistic integrator shown above, the integral equation is correct, but only for
frequencies above the point where the effect of C
2
begins to dominate over R
2
to set the
gain of the circuit. The frequency below which the circuit’s behavior becomes more like
a DC amplifier than an integrator is given by the lowpass filter's cutoff frequency.
G. Kovacs ©1997 Page 46
R
2
R
1
Gain
Frequency
"Integrator"
"Amplifier"
f
min
Chapter 4: OPERATIONAL AMPLIFIERS
3.8 DIFFERENTIATOR
R2
R1
C1
V
OUT
V
V

+
V
IN
i
i n
i
f
• In the differentiator circuit shown here, there is also a component, R
1
, that is sometimes
not shown in “textbook” opamp differentiator circuits.
• Its purpose is to limit the highfrequency gain of the differentiator so that it does not get
swamped by high frequency noise (which may have a large derivative despite a small
amplitude).
• For “lowenough” frequencies that the input impedance is dominated by C
1
, i
in
and i
f
can be equated to show that,
v
out
· −R
2
C
1
dv
in
dt
• As for the integrator above, there is a frequency range over which the differentiator will
not work very well.
• Again, one can start with the equation for an inverting amplifier and substitute the
impedance of R
1
and C
1
in series for R
1
,
v
out
v
in
· −
R
2
Z
1
· −
R
2
R
1
+
1
C
1
S
· −
R
2
C
1
S
R
1
C
1
S +1
which yields an equation for a highpass filter in classical form, with a cutoff frequency
of, f
max
·
1
2πR
1
C
1
, above which the filter gives a steady gain of
v
out
v
in
· −
R
2
R
1
and
below which it acts as a differentiator.
• Again, substitute zero and infinity for S in the above equation and be sure you understand
what happens at both limits.
G. Kovacs ©1997 Page 47
Chapter 4: OPERATIONAL AMPLIFIERS
4: “REAL” VERSUS “IDEAL” OPAMPS
+ +
+
Acm ( Vi+ + Vi) / 2
Adm ( Vi+  Vi)
Ro
Ri
Vos
Ios
Ibias
Ibias
Vi +
Vi 
Vo
VCC
 VEE
• There are unwanted currents at the inputs!
• There are offset currents and voltages!
• Signals applied to both inputs (which should not be amplified) are amplified to
some extent!
• The input resistance is not infinity!
• The output resistance is not zero!
• However, for most practical applications, they are pretty close to being ideal, and
G. Kovacs ©1997 Page 48
negative feedback helps make them even closer!
Chapter 4: OPERATIONAL AMPLIFIERS
5: FREQUENCY RESPONSE OF OPAMPS
• As you might have guessed, opamps do not have infinite frequency response
either...
• When used without external feedback, the opamp is in OPENLOOP mode, and the
gain is quite large, but the frequency response is TERRIBLE!.
• As you increase the amount of negative feedback, the bandwidth increases and the
gain decreases.
• The product of GAIN X BANDWIDTH is a constant for a given opamp (guess what?...
it is referred to as the “gainbandwidth product”).
• The opamp has (by design) a FIRSTORDER LOWPASS RESPONSE, as we saw in
the first lectures....
• The openloop gain is given by:
A
op−amp
S ( ) ·
A
o
1+
S
ω
o
where A
o
is the open loop gain at DC and ω
o
is the 3dB or “break” frequency at which
the openloop gain starts to roll off.
KEY POINT: You can model an opamp’s openloop frequency response as a
firstorder RC lowpass filter with,
ω
o
·
1
RC ( )
equivalent
• Another key definition is the UNITYGAIN BANDWIDTH, which is the frequency, ω
t
, at
which the gain reaches ONE.
• By thinking about the gainbandwidth product concept, it follows that,
ω
t
· A
o
ω
o
G. Kovacs ©1997 Page 49
Chapter 4: OPERATIONAL AMPLIFIERS
• If you rederive the gain equation for the inverting amplifier assuming a finite opamp
gain (see page 55, Sedra & Smith), you obtain,
A
v
S ( ) · −
R
2
R
1
1+
1+
R
2
R
1
¸
¸
_
,
A
op−amp
S ( )
• Substituting the opamp’s frequency response in for A
opamp
, you (eventually) obtain,
A
v
S ( ) · −
R
2
R
1
1+
1+
R
2
R
1
¸
¸
_
,
A
o
1+ Sω
o
¸
¸
_
,
· −
R
2
R
1
1+
1
A
o
1+
R
2
R
1
¸
¸
_
,
+
S
ω
t
1+
R
2
R
1
¸
¸
_
,
A
v
S ( ) ≈ −
R
2
R
1
1+
S
ω
t
1+
R
2
R
1
¸
¸
_
,
for A
o
>> 1+
R
2
R
1
¸
¸
_
,
• This is a firstorder lowpass filter.
G. Kovacs ©1997 Page 50
Chapter 4: OPERATIONAL AMPLIFIERS
V V
in out
R
C
Hs =
K
1 +
s
ω
o
where K is the gain for low frequencies
the pole is at S = ω
o
and the “cutoff” frequency is at ω
0
= 2πf
c
=
1
RC
THE DC GAIN IS 
R
2
R
1
AND THE BREAK FREQUENCY IS
ω
t
1 +
R
2
R
1
CONCLUSION: You can model the frequency response of the inverting configuration
also using a firstorder lowpass filter with,
RC
equivalent
=
1 +
R
2
R
1
ω
t
• A similar analysis for the noninverting case can be found in Sedra and Smith (Section
2.7).
G. Kovacs ©1997 Page 51
Chapter 4: OPERATIONAL AMPLIFIERS
G
A
I
N
FREQUENCY
A = 1K
A = 100
A = 10
A = 1
FREQUENCY RESPONSE OF INVERTING
OPAMP CIRCUIT FOR VARIOUS GAINS,
ASSUMING GBPRODUCT = 1 MHZ AND
fb = 10 Hz
• The above plots illustrate that, no matter what the gain of the opamp is set to via
external feedback resistors, the overall frequency response is constrained by the outer
boundary set by the opamp's open looproll off. As the gain is reduced, the frequency
at which that new gain is reduced by 3dB moves up in frequency. This is another way
of looking at the gainbandwidth product rule.
G. Kovacs ©1997 Page 52
Chapter 4: OPERATIONAL AMPLIFIERS
6: THE GUTS OF REAL OPAMP CHIPS
• In real opamps, there is generally an internal (sometimes external) capacitor that sets
the opamp’s dominant pole and forces it to roll off before other, less well controlled
parasitic capacitances would cause that.
• In practical integrated circuit implementations, the capacitances available are very
small (a few pF)... how can such small capacitances give rise to such low cutoff
frequencies (a few hundred Hz)?
• The answer is provided by an understanding of the Miller Effect, which basically
explains how an impedance sitting across an amplifier is effectively converted into a
smaller impedance at the input and roughly an equal impedance at the output of the
amplifier... the amount that the effective input impedance is smaller than the actual one
across the amplifier is a function of the amplifier’s gain.
• This explains how an opamp, with a huge gain and a very small internal capacitance,
can act like it has a huge capacitance at its input (a smaller impedance corresponds to
a larger capacitance).
• As mentioned above, the Miller Effect effectively takes an impedance across an amplifier
and replaces it with a smaller impedance at the input and a roughly equivalent one at
the output.
• This effect works only for amplifiers that share a common terminal between input and
output (e.g. ground).
• The effect does work for negative and positive gains, but for positive gains greater
than one, gives rise to negative impedances at the input (reverse polarity of current that
flows for an applied voltage, but Ohm’s Law still applies). (The output impedance is still
positive.)
• Here, we only consider negative (inverting) amplifiers.
G. Kovacs ©1997 Page 53
Chapter 4: OPERATIONAL AMPLIFIERS
G. Kovacs ©1997 Page 54
i
1
=
v
in
 v
out
Z
f
=
v
in
 Av
in
Z
f
=
v
in
Z
f
1  A
i
2
=
v
out
 v
in
Z
f
=
v
out

v
out
A
Z
f
=
v
out
Z
f
A
1  A
i
1
*
=
v
in
Z
1
i
2
*
=
v
out
Z
2
i
1
*
= i
1
i
2
*
= i
2
Z
1
=
Z
f
1  A
Z
2
= Z
f
A
1  A
and By definition,
Chapter 4: OPERATIONAL AMPLIFIERS
• Intuitively, one can think of “looking” into the input of the amplifier and seeing one side
of an impedance... with an inverting gain, the amplifier is going to “pull” against the
other side of that impedance in response to any input voltage wiggle...
• If you wiggle the input voltage up slightly and try to force current into the impedance
(remember that you can’t force current into the input of the amplifier!), the amplifier will
pull a lot more current away from you than if you were pushing against just the impedance
tied to ground.
• This seems, “looking” from the input terminal, like a smaller impedance than what is
sitting across the amplifier... you get more current flowing for a small voltage wiggle
than if the impedance were just tied to ground.
• Note that taking A through the gain range of 1 to ∞, the corresponding range of Z
2
is
Z
2
/2 and Z
2
. Thus for a capacitor in the Miller configuration, the maximum value of the
effective C
out
is 2C
f
and the minimum is C
out
= C
f
.
• Using Miller’s Theorem, we can look at the opamp “guts” model to figure out why we
get the frequency response observed (after Sedra & Smith, Section 2.8)....
+

V
id
V
o
+

R
o1
R
i2
V
i2
sC(1+µ)V
i2
V
i2
+

+

G
m
V
id
 µV
i2
G
m
V
id
R = R  R
o1
i2
C (1 + µ)
+

• From this it can be seen that the feedback capacitor’s value is multiplied by a factor of
1 + the gain of the internal voltage amplifier (µ) and placed in parallel with the combination
G. Kovacs ©1997 Page 55
of the first stage’s output resistance and the second stage’s input resistance.
Chapter 4: OPERATIONAL AMPLIFIERS
• This determines the effective RC time constant of the opamp (see the Appendix
below).
• At the output, an equivalent capacitance is present with a value of C, for large µ... for
an ideal (voltage source) output, that makes no difference, but for a real voltage source
output (i.e. with a series resistance other than zero), it can make quite an important
difference!
7: A PREVIEW OF FEEDBACK
R
R
A
1
2
Feedback Loop
V
in
• This schematic shows the feedback path for the noninverting opamp configuration.
• This is a useful point to define some of the feedback terms that will be used later in the
course.
FEEDBACK RATIO (FACTOR) β ·
R
1
R
1
+ R
2
LOOP GAIN: T · −Aβ
AMOUNT OF FEEDBACK ≡ 1  T = 1 + A β
GAIN ·
A
AMOUNT OF FEEDBACK
·
A
1+ Aβ
G. Kovacs ©1997 Page 56
Gain = A
V
·
A
o
1+ A
o
R
1
R
1
+ R
2
¸
¸
_
,
In this case, it is just
voltage division and
ß is the fraction of the
output voltage fed back.
Chapter 4: OPERATIONAL AMPLIFIERS
APPENDIX TO CHAPTER 4:
A Bit More On Miller
Why isn't it called the Budweiser Effect?
Bobby Twistoffski
Stanford Senior, Electrical Engineering
WHY DO WE CARE?
• The Miller Effect can turn a small, parasitic capacitance sitting across an amplifier into
a real problem!
• The Miller Effect is basically a multiplication of the capacitance's value, as seen at the
input of the circuit by a factor related to the gain of the amplifier.
• In other words, we can get a HUGE effective capacitance at the input of a highgain
amplifier for having only a small (i.e. a few picoFarads) capacitance across it (from input
to output)...
• We can use the Miller Theorem to help analyze circuits where we have an impedance
across an amplifier.
• Again, that assumes that the gain of the amplifier is not changed by having the
capacitor placed across it (if it is changed slightly, we can still use the Miller
G. Kovacs ©1997 Page 57
approximation).
Chapter 4: OPERATIONAL AMPLIFIERS
• Previously, we saw that the Miller Theorem lets us "convert a circuit with an impedance
across an amplifier in to one that is easier to analyze because we simplify it into two
"separate" circuits!
• A KEY assumption is that the gain of the amplifier itself is not changed by having
the impedance placed between its input and output...
Amplifier
with
Gain=A
V
in
V
out
Z
f
1
i
i
2
Amplifier
with
Gain=A
V
in
V
out
Z
1
Z
2
1
i i
2
• We showed that,
Z
1
=
Z
f
1  A
Z
2
= Z
f
A
1  A
and for an "ideal" amplifier, we can let A > ∞ and we get,
Z
1
→ 0 Z
2
→ Z
f
• We can replace an impedance across the amplifier to two impedances to ground
(usually easier to work with).
• If our impedance was actually a capacitance, for a given frequency, that would mean
that C
1
would tend to increase (to decrease the impedance at the input to zero) and C
2
would tend to decrease (to increase the impedance at the output to infinity).
• It's easier to call them C
in
and C
out
to keep things straight, and that is what we'll do...
G. Kovacs ©1997 Page 58
Chapter 4: OPERATIONAL AMPLIFIERS
• Now actually do it with a capacitor to convince yourself...
Amplifier
with
Gain=A
V
in
V
out
Amplifier
with
Gain=A
V
in
V
out
1
i
i
2
1
i
i
2
C
f
C
out
C
in
• The effective impedances are,
Z
Cin
·
1
C
f
S
¸
¸
_
,
1− A
·
1
C
f
1− A ( )S
Z
Cout
·
1
C
f
S
¸
¸
_
,
−A
1− A
¸
¸
_
,
·
1
C
f
1− A
−A
¸
¸
_
,
S
which is the same as saying that there are "scaled copies" of C
f
at the input and output
of the amplifier,
C
in
· C
f
1− A ( ) C
out
· C
f
1− A
−A
¸
¸
_
,
• If the amplifier was really "ideal" and we were not worrying about input and output
resistances of the driving circuit and the load, these capacitances might not be a
problem... (for example, if an "ideal" voltage source with zero output impedance was
driving C
in
, the fact that C
in
was large would not matter)
G. Kovacs ©1997 Page 59
Chapter 4: OPERATIONAL AMPLIFIERS
• Now look at a "realistic" amplifier (input impedance is less than infinity and output
resistance is greater than zero)...
Amplifier
with
Gain=A
V
in
V
out
1
i
i
2
C
out
C
in
R
in
R
out
• Now we have created two RC time constants (e.g. two lowpass filters!), one of which
(input) turns out to usually be dominant in terms of the overall frequency response...
τ
in
· R
in
C
in
· R
in
C
f
1− A ( ) τ
out
· R
out
C
out
· R
out
C
f
1− A
−A
¸
¸
_
,
• From this you can see that we have a relatively small contribution from C
f
at the output
and a relatively large contribution at the input...
• For typical amplifiers, R
in
is large and R
out
is small, so we get a large time constant at
the input and a small one at the output.
• In other words, the input pole will dominate the frequency response since it will
take effect at a lower frequency that the output pole.
• Still more realistically, we have to take into account the output resistance of the circuit
driving our amplifier and the load resistance that it has to drive... now we could
compute more accurate time constants (and thus the frequency response).
G. Kovacs ©1997 Page 60
Amplifier
with
Gain=A
V
in
V
out
1
i
i
2
C
out
C
in
R
in
R
out
R
s
R
load
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
Chapter 5: SINGLESTAGE BJT
AMPLIFIERS AND THE COMMON
EMITTER AMPLIFIER
This amplifier goes to eleven!
Nigel, member of the rock band “Spinal Tap.”
1: OBJECTIVES
• To learn about:
The similarities and differences between the three basic BJT amplifiers.
The basic analysis and design approaches to the commonemitter amplifier.
READ S&S Sections 4.10  4.12
2: OVERVIEW OF SINGLESTAGE BJT AMPLIFIERS
• There are three basic BJT amplifier configurations, named depending upon which of
the three terminals (B, C, or E) is common to the input and output of the amplifier
(commonbase, commoncollector, or commonemitter, respectively).
• In practice, it is the terminal that is connected (either directly or through a capacitor
bypassed resistor > directly for AC signals) to a power or ground voltage that is the
“common” terminal.
• Each of the three possible configurations has unique characteristics that make them
useful in different situations. The table below compares the three basic configurations,
as well as a variant of the common emitter amplifier.
• The common emitter amplifier is the most common version, and is an inverting amplifier
(that is, the output is 180° out of phase with the input). In many cases, a resistor is
placed between the emitter and ground (usually bypassed with a large capacitor). This
G. Kovacs ©1997 Page 61
emitter resistor, R
E
, serves to stabilize the bias point of the transistor using local feedback
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
(this is discussed in detail below). In order to obtain this stabilization, gain is traded off
(reduced). This form of amplifier is, however, still a common emitter type.
• The common collector amplifier is often used as a buffer amplifier, since its noninverting
gain is almost exactly one. The current gain, however, is large (approximately ß+1),
and this circuit is often used to add power to a signal.
• The common base amplifier, like the common collector, is noninverting. It can provide
high voltage gains, like the common emitter amplifier. The main difference between the
common base and the other amplifier types is that the input impedance is very low.
This can be a problem if the signal source is high impedance (in other words if the
signal source cannot easily drive a lowimpedance load), but it can be very useful if the
signal source is low impedance (e.g. radio frequency signals arriving via a coaxial cable
of characteristic impedance 50 Ω). The common base amplifier is generally very fast.
• The relative merits of each amplifier type is discussed below, including frequency
response.
Parameter CE CE with RE CC CB
Inverting? YES YES NO NO
A
v
HIGH HIGH (reduced) LOW HIGH
A
i
HIGH HIGH (reduced) HIGH LOW
R
in
MEDIUM MEDIUM
(increased)
HIGH LOW
R
out
HIGH HIGH
(increased)
LOW HIGH
Typical
Schematic
Vcc
CC
VCC VCC VCCVEE
G. Kovacs ©1997 Page 62
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
3: THE COMMON EMITTER AMPLIFIER
• As mentioned above, the common emitter amplifier is by far the most common single
transistor amplifier configuration. You need to be quite familiar with it, its properties
and, eventually, its variations. It is important to note that it is an inverting amplifier.
USING THE PREVIOUS EXAMPLE (not the best biasing, but simple)
CC
v
o
v
s
R
B
R
C
V
V
CC
= 15V R
B
= 50K
R
C
= 500 Ω β = 100
I
B
=
15  0.7
50K
= 0.286mA
I
C
= 100 I
B
= 28.6mA
v
o
= g
m
v
be
R
C
v
be
= v
s
• One can compute the voltage gain A
v
...
A
v
·
v
o
v
s
·
v
o
v
be
· −g
m
R
C
g
m
=
I
C
V
T
=
28.6 mA
25.9 mV
= 1.1Ω
1
A
v
=  1.1 500 = 550
NOW MAKE IT MORE REALISTIC... ADD AN OUTPUT COUPLING CAPACITOR
G. Kovacs ©1997 Page 63
AND A SOURCE RESISTANCE!
This capacitor
is generally
quite large, and
serves to block
DC voltages
from affecting
the bias point.
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
CC
v
o
v
s
R
B
R
C
R
S
V
• To make an AC smallsignal model, short the caps and V
CC
to ground and you get....
R
s
v
s
v
be
· v
s
R
B
r
π
R
S
+ R
B
 r
π
¸
¸
_
,
v
o
· −g
m
v
be
R
C
A
v
·
v
o
v
s
· −
R
B
 r
π
R
S
+ R
B
 r
π
¸
¸
_
,
g
m
R
C
• Note that with a source resistance, R
S
, there is a voltage divider at the input, reducing
the gain. The larger the source resistance is, the lower the gain will be. If the input
resistance of the amplifier (R
B
r
π
in this case) is made very large, the effects of R
S
are
reduced.
G. Kovacs ©1997 Page 64
m
Note that it is I
that links the DC
and AC models
through g .
c
m
input coupling capacitor
blocks DC but is intended
to pass AC signals (i.e. use
a fairly large capacitor here)
output coupling capacitor
blocks DC but is intended
to pass AC signals (i.e. use
a fairly large capacitor here)
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
TRY THE EXAMPLE AGAIN WITH A 50 Ω SOURCE RESISTANCE (typical for a
laboratory signal generator):
R
B
= 50 KΩ R
C
= 500 Ω I
C
= 28.6 mA β
F
= 100
g
m
=
I
C
V
T
=
28.6 mA
25.9 mV
= 1.1Ω
1
r
π
=
β
g
m
=
100
1.1
= 90.6Ω
A
v
=  1.1 500
50K  90.6Ω
50Ω + 50K  90.6 Ω
≈  1.1 500
90.6
50 + 90.6
= 356 Not Bad!
• REMEMBER this is SMALL signal only! (i.e. v
be
<< v
T
= 25.9 mV) Note that V
C
≈ 0.7
V, and the transistor is on the edge of saturation. Running it at a lower current would
raise V
C
and allow for a larger voltage swing at the output, but would reduce gain.
• Things can get nonlinear for larger signals (i.e. DISTORTION)!
• A potential problem here is that you have an overall gain that is a function of ß!
• You cannot control that (unless you make your own transistors, and then only somewhat)
and it varies with temperature!
• To get around that, you can make sure the other terms in the input voltage divider
“dwarf out” R
S
... in this example, if a smaller I
C
were used, g
m
would fall and r
π
would
increase to help accomplish this (of course, gain would also fall unless R
C
were increased
to compensate).
• It turns out that a potentially bigger problem is that g
m
varies with I
C
which varies with
G. Kovacs ©1997 Page 65
ß!... that can be fixed using a better biasing scheme...
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
4: CLASSIC BIASING SCHEME FOR CE AMPLIFIERS
CC
v
in
R
B2
R
C
V
B
V
V
EE
R
E
R
B1
I
B
I
B1
I
B2
v
o
• Typical V
CC
= +12 or +15 V, typical V
EE
= 0, 12V, or 15 V
5: BASE CIRCUIT DESIGN STRATEGY  THE BASICS
• It is necessary to keep V
B
more than one diode drop below V
C
even at maximum signal
swing to prevent saturation. The method presented here is described in Sedra & Smith
and is commonly used... it is more of a rule of thumb than an absolute, however!
• The basic idea is to set up the bias resistors to provide a nearly constant voltage at the
base of the transistor, independent of its ß.
• As explained below, the emitter resistor, R
E
, serves (through localized negative feedback)
to trade some of the available gain of the circuit for stability in I
C
, which in turn stabilizes
g
m
. If R
E
is bypassed using a capacitor (as shown above, and commonly done), the AC
gain is unaffected by R
E
(since large bypass capacitors "become shorts" in the AC
model). If R
E
is not bypassed, the AC gain of the amplifier is reduced.
• Choose the overall magnitude of R
B1
and R
B2
to ensure that the current through them
is about 10  20 X greater than the base current (i.e. only about 10% of the total current
flowing in the base circuit is shunted to the base as I
B
) to ensure bias stability.
• As a general rule, choose R
B1
:R
B2
ratio to set V
B
around onethird of the way up from
V
EE
(or ground) to V
CC
... this makes sure that V
B
is much larger than any changes in V
BE
expected due to temperature variation (V
BE
decreases by 2 mV/°C).
G. Kovacs ©1997 Page 66
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
6: LOCAL FEEDBACK THROUGH AN EMITTER
RESISTOR
• As mentioned above, R
E
is present in the CE circuit shown above to help stabilize it.
• The previously discussed biasing schemes (without R
E
) are really not very good
because:
1) β varies from transistor to transistor (think about what would happen in production!)
2) g
m
and β vary with I
C
and temperature! Remember g
m
=
I
C
V
T
• GOOD biasing would make the bias current stable (i.e. keep I
C
constant as the
temperature and devicedependent parameters changed), since variations in I
C
cause
variations in gain, etc.
• R
E
allows for stabilization of I
C
by feeding back an error signal if I
C
increases. The
error signal reduces V
BE
enough to reduce I
C
back down to where it should be.
• In general, a bypass capacitor, C
E
, is placed in parallel with the emitter resistance, R
E
,
to "short out" R
E
for AC signals but not disturb R
E
's effect of stabilizing the bias point
(which only matters at DC). Thus, a bypassed R
E
provides DC feedback only.
VCC
As usual for smallsignal analysis,
Vcc shorts to ground, and capacitors
become shortcircuits...
E
R
E C
R
C
s
v
R
v
o
S
B1
B2
R
R
G. Kovacs ©1997 Page 67
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
• With R
E
shunted by a large enough capacitor, this gives exactly the same gain equations
as we saw without R
E
...
v
be
· v
s
R
B
r
π
R
S
+ R
B
 r
π
¸
¸
_
,
v
o
· −g
m
v
be
R
C
A
V
·
v
o
V
s
· −
R
B
 r
π
R
S
+ R
B
 r
π
¸
¸
_
,
g
m
R
C
• The details of designing the whole thing, including the base circuit and R
E
are
presented below after some discussion of analysis and the effects of nonbypassed
R
E
!
7: BASE CIRCUIT ANALYSIS STRATEGY
• This section is concerned with how one analyzes the base circuit to check that it is
designed correctly or to understand the operation of a circuit.
• If I
B
is much smaller than the current running through R
B1
and R
B2
, then V
B
is set by a
simple voltage divider.... then one can assume a V
BE
, compute the currents, and check
the V
BE
assumption (i.e. is the transistor in active mode versus cutoff or saturation).
• Otherwise use a Thévenin equivalent for the input circuit and solve the loop equation
to compute I
B
(see below).
V
BB
R
B
R
B1
R
B2
EE
V
CC
V
CC
V
EE
V
V
B
V
B
ASSUME
V = 0 V
EE
V
BB
= V
CC
R
B2
R
B1
+ R
B2 R
B
= R
B1
R
B2
G. Kovacs ©1997 Page 68
Assuming V = 0
EE
Clearly V = V if the base current
is zero! Then you just have a voltage
divider.... The equivalent resistance R
is useful to understand how the base
voltage drops if the base current goes
up.
B
BB B
Rule of thumb: this is a good approximation when the base
current is <10% of the total basecircuit current...
This is true as long as the bypass capacitor
is large enough and/or the frequency is high
enough!!!!
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
BIAS CIRCUIT ANALYSIS: SEE SEDRA & SMITH EXAMPLE 4.21
• Do the example two ways...
1) Assume I
B
is small relative to the total current through the bias resistors and
use a simple voltage divider to compute V
B
...
2) Do not assume anything....
COMPARE THE RESULTS! > The point is that if you assume a simple voltage
divider, the base current must be small relative to the current through the bias resistors!
8: UNBYPASSED EMITTER RESISTANCE FOR AC
AND DC FEEDBACK
• With R
E
not bypassed, the situation is changed so that there is a resistance from the
emitter of the hybridπ model to ground (i.e. the AC model IS affected by R
E
).
• You can cleverly handle this by “pulling” R
E
into the hybridπ model and redefining g
m
and r
π
as g
m
' and r
π
'... this makes the overall circuit analysis much simpler, since we can
redraw the AC equivalent circuit with the emitter of the hybridπ model grounded,
avoiding the need to write loop equations!
C B
E
g
m
v
be
r
π
C
B
E
g
m
v
be
r
π
R
E
' '
G. Kovacs ©1997 Page 69
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
CC
v
s
R
B2
R
C
V
B
V
R
E
R
B1
v
o
R
B
R
C
E
R
g
m
v
be
r
π
v
s
v
be
+

i
in
R
in
+

v
in
SIMPLIFYING ASSUMPTIONS / OBSERVATIONS:
• Ignore the source resistance R
S
(not shown above) for the analysis.
• Ignore the transistor’s output resistance r
o
which would be in parallel with the g
m
current generator....
• Take note of the fact that the (ideal) signal source is directly driving R
B
, so it does not
enter into our calculations...
• Calculate the gain and input resistance and compare to CE amp without R
E
...
v
in
= v
be
+
1
r
π
+ g
m
v
be
R
E
(the second term is the voltage across R
E
)
• The other way to get this equation is from the base current,
v
in
= v
be
+ 1 + β i
in
R
E
= v
be
+ 1 + β i
B
R
E
v
in
= v
be
+ 1 + β
v
be
r
π
R
E
= v
be
+
1
r
π
+ g
m
v
be
R
E
G. Kovacs ©1997 Page 70
i
e
sum of conductances seen looking into base terminal
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
• Solve for v
be
...
v
be
=
v
in
1 +
1
r
π
+ g
m
R
E
≈
v
in
1 + g
m
R
E
i
in
=
v
be
r
π
=
v
in
r
π
+ 1 + g
m
r
π
R
E
≈
v
in
r
π
1 + g
m
R
E
• Therefore, one can now redefine the input resistance of the transistor “taking into
account” R
E
,
R
in
≡
v
in
i
in
= r
π
+ 1 + β R
E
≈ r
π
1 + g
m
R
E
≡ r
π
'
• Similarly, one can look at the gain,
v
o
=  g
m
v
be
R
C
• Substituting in v
be
≈
v
in
1+g
m
R
E
, we can define a new g
m
“taking into account” R
E
,
v
o
· −R
c
g
m
1+ g
m
R
E
¸
¸
_
,
v
in
• Finally, plugging into the original CE amplifier gain equation,
A
V
·
v
out
v
in
· −g
m
'
R
C
· −
g
m
1+ g
m
R
E
¸
¸
_
,
R
C
≈ −
R
C
R
E
if g
m
R
E
>>1
G. Kovacs ©1997 Page 71
this term is defined as
g
m
'
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
• Thus, one can effectively replace the original hybridπ model with a new one
with an INCREASED r
π
(now called r
π
’) and a DECREASED g
m
(now called g
m
’)....
r
π
'
≡ r
π
1+ g
m
R
E
( ) g
m
'
≡
g
m
1+ g
m
R
E
• If the input divider losses are significant, they should be included in the gain equation
as shown below,
A
V
·
v
out
v
in
· −
R
B
 r
π
'
R
S
+ R
B
 r
π
'
¸
¸
_
,
g
m
'
R
C
· −
R
B
r
π
'
R
S
+R
B
r
π
'
¸
¸
_
,
g
m
1+g
m
R
E
¸
¸
_
,
R
C
≈ −
R
B
r
π
'
R
S
+R
B
r
π
'
¸
¸
_
,
R
C
R
E
if g
m
R
E
>> 1
• The g
m
' equation should look like the feedback equation, because that is exactly what
it shows...
g
m
'
≡
g
m
1+ g
m
R
E
GAIN ·
A
AMOUNT OF FEEDBACK
·
A
1+ Aβ
• What this means is that, treating the transistor as a transconductance amplifier, we are
cutting down its "open loop" transconductance g
m
by a factor (1+g
m
R
E
).
G. Kovacs ©1997 Page 72
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
• The feedback situation for this circuit is illustrated conceptually below:
+

v
in
v
be
R
E
g
m
i
c
Negative
Feedback
• Later in the notes, the tools to handle general feedback cases are presented.
• The same approach works if you have an IMPEDANCE in the emitter circuit instead
of a resistance.
• It usually makes sense to use the simplifying assumption up front, to simplify the AC
equivalent circuit, the gain of which is then obtained using the standard equation for the
common emitter amplifier.
G. Kovacs ©1997 Page 73
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
CC
v
s
R
B2
R
C
V
B
V
R
E
R
B1
v
o
R
B
R
C
E
R
g
m
v
be
r
π
v
s
v
be
+

i
in
R
in
+

v
in
v
o
R
B
R
C
g
m
v
be
r
π
v
s
v
be
+

i
in
R
in
+

v
in
'
'
G. Kovacs ©1997 Page 74
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
9: EMITTER DEGENERATION
• “Emitter degeneration” describes what happens to the DC transfer function of a CE
amplifier that has an unbypassed R
E
.... (note that in the active region, the slope is
R
C
/R
E
if g
m
R
E
>1)
R
E
R
C
+ R
E
V
CC
+ V
BE
V
CC

R
C
R
C
+ R
E
V
CC
• Looking at the circuit, we can see the three cases much more clearly... For active
mode, it looks like this:
G. Kovacs ©1997 Page 75
CC
V
IN
R
B2
R
C
V
B
V
R
E
R
B1
+

V
OUT
CC
vs
R
B2
RC
V
B
V
R
E
R
B1
v
o
0 V (ground)
dc
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
• For cutoff, the circuit looks roughly like this:
CC
V
IN
R
B2
R
C
V
B
V
R
E
R
B1
+

V
OUT
No connection!
Of course, leakage
currents DO flow,
but this is roughly
the case...
• For saturation, the circuit looks like this:
CC
V
IN
R
B2
R
C
V
B
V
R
E
R
B1
+

V
OUT
≈ 0.2 V
+

I
B
• V
CE
becomes V
SAT
(which is roughly 0.2  0.3 V in practice) when the transistor
saturates...
G. Kovacs ©1997 Page 76
• The transistor is now delivering the maximum I
C
it can, so I
C
remains at I
CSAT
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
• Further increases in I
B
simply go directly to the emitter circuit, raising V
E
• Because V
CE
is now fixed at V
SAT
, raising V
E
ends up raising V
C
(V
OUT
)
• This means that as V
B
(V
IN
) goes up, V
E
just tracks it, after losing a forward biased
junction drop V
BE(SAT)
which is between 0.7 and 0.8 Volts.
• That last point explains why the slope of the transfer function becomes one in
saturation... the input voltage V
B
is essentially shorted to V
E
, which is essentially
shorted to V
OUT
!
• This means that V
OUT
just tracks further increases in V
IN
(certainly, if V
IN
is an ideal
voltage source, it can drive V
OUT
just fine!).
• Of course, when saying things are "shorted," it is assumed that we can ignore the
small voltage drops (0.7 V between B and E, and 0.2 V between C and E).
• Note that with 0.2 V between collector and emitter, the BC junction voltage is actually
about 0.5 V, so the BC junction is not fully forward biased. Also, the extra base current
flowing into the transistor does not flow out the collector because it is at a higher
potential than the emitter!
• Now it is worthwhile to modify our previous Spice deck (our design example) for
timedomain analysis and put in a single cycle of a BIG sinewave (7 V!) to look at what
happens when the amplifier hits the limits of its voltage transfer function (cutoff and
saturation)...
EE113 Example Degenerated CommonEmitter Amplifier
*dc components
Vcc VCC 0 12
R1 VCC VBG 39K
Q1 VCG VBG VEG TRANSMODEL
R2 VBG 0 18K
RC VCC VCG 680
RE VEG 0 1.8K
*.MODEL TRANSMODEL NPN (IS=1.3E14)
*ac components
Vss 0 Vin sin(0 7 1000)
RS Vin VX 1
C3 VX VBG 100UF
*C6 VEG 0 100UF
C5 Vout VCG 100UF
RLL Vout 0 50K
.MODEL TRANSMODEL NPN (BF=150 IS=1.3E14
+ TF=.9N CJE=6P CJC=5P)
*Time Domain Response
.TRAN 100nS 1mS
.PROBE
.end
G. Kovacs ©1997 Page 77
E
The asterisk in front of this line
removes the bypass capacitor
from across R .....
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
• Here is the result, showing input and output signals...
NOTE THAT THE SPICE DECK CONTAINS AN OUTPUT COUPLING CAPACITOR,
which explains why the ACcoupled output is centered around 0 V while the
collector voltage si swinging near to + 12 V!
saturation
cutoff (clipping)
G. Kovacs ©1997 Page 78
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
• Below is an expanded view of the output signal (ACcoupled)...
0 s 0 . 2 ms 0 . 4 ms 0 . 6 ms 0 . 8 ms 1 . 0 ms
Ti me
V( Vo u t )
2 . 0 V
1 . 0 V
0 . 0 V
 1 . 0 V
 2 . 0 V
 3 . 0 V
Ex i t Ad d _ t r a c e Re mo v e _ t r a c e X_ a x i s Y_ a x i s Pl o t _ c o n t r o l
Di s p l a y _ c o n t r o l Ma c r o s Ha r d _ c o p y Cu r s o r Zo o m La b e l
CUTOFF (maximum
positive swing)
SATURATION
(slope becomes +1)
• Now you can see why we bother looking at the transfer function... it predicts the kind
of distortions you see when you overdrive an emitterdegenerated (unbypassed R
E
)
commonemitter amplifier.
G. Kovacs ©1997 Page 79
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
10: THE BOTTOM LINE ON DESIGN OF CE AMPLIFIERS
• Typically you design to meet a power dissipation specification (i.e. the maximum I
C
tolerable for a reasonable battery life...) and want maximum output signal swing... let’s
go through the thinking behind it and then a design example using the methods of
Sedra and Smith.
V
CC
V
EE
+v
o
v
o
V
C
GROUND
CC
v
o
v
in
R
B2
R
C
V
B
V
V
EE
R
E
R
B1
I
B
I
B1
I
B2
v
o
V
B
1
3
1
3
1
3
(usually
bypassed)
• Typical V
CC
= +15 V, typical V
EE
= 15 V (other common voltages = +/ 12 V)
• The rule of thumb approach is to design for the voltage between the power supplies to
be split (roughly) into thirds....
V
B
=
1
3
V
CC
 V
EE
V
C
≈
2
3
V
CC
− V
EE
( ) NOTE that this is not a design constraint!
Note that the V
C
value specified is the quiescent (no input signal) value). Do not try to
force it to that value. V
C
may be considerably lower if the current (and gain) is increased.
1) Assuming that you know the desired DC collector current, you compute the resistor
values (for DC) that give you this current.
2) Then choose the bias resistors R
B1
and R
B2
so that the 1/3 rule is obeyed and so that
the base current is in the range I
B
≈ 0.1 I
B1
so that the “voltage divider” assumption is
G. Kovacs ©1997 Page 80
reasonably close (i.e. the base current doesn’t change V
B
much).
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
Rules of thumb for this are:
V
BB
>> V
BE
This ensures that small variations in V
BE
(near 0.7V) will be dwarfed
by V
BB
being so much higher so that I
B
will not change much...
Remember that I
E
·
V
BB
−V
BE
R
E
R
E
>>
R
B
β + 1
This ensures that R
E
is large enough to provide adequate local
feedback to stabilize I
E
(and hence, I
C
)... The term on the right is
simply the Thévenin equivalent of the bias resistors as “seen” from
the emitter (remember that dividing by (β + 1) “transforms”
resistances from the base circuit to the perspective of the emitter
circuit).
11: THE ART OF CE DESIGN (AN INTRO)
• Which ever approach you take to designing CE amplifiers, you will typically be designing
them to meet certain specifications, such as gain, power dissipation (quiescent), specific
collector (or emitter current), etc.
• The approach you take depends on which of these specifications are given and which
are most important.
• For example, if the gain was specified, you would probably start with the voltage
gain equation for the configuration you wish to use to get a feel for the constraint. The
equation for the case where R
E
is bypassed is,
A
v
·
v
o
v
s
· −
R
B
 r
π
R
S
+ R
B
 r
π
¸
¸
_
,
g
m
R
C
and for when R
E
is not bypassed,
G. Kovacs ©1997 Page 81
A
V
·
v
out
v
in
· −
R
B
 r
π
'
R
S
+ R
B
 r
π
'
¸
¸
_
,
g
m
'
R
C
· −
R
B
r
π
'
R
S
+R
B
r
π
'
¸
¸
_
,
g
m
1+g
m
R
E
¸
¸
_
,
R
C
≈ −
R
B
r
π
'
R
S
+R
B
r
π
'
¸
¸
_
,
R
C
R
E
if g
m
R
E
>> 1
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
• You then look to see what determines the gain. In this case, if g
m
R
E
>> 1, it is purely
the ratio of the resistances (neglecting the effects of the input divider)....
A
v
≈ 
R
C
R
E
• For a nondegenerated CE amplifier (R
E
bypassed by a large enough capacitor), the
gain is g
m
R
C
, so you need to be careful to set g
m
to an appropriate value by choosing I
C
,
knowing that,
g
m
=
I
C
V
T
• This is where your design begins in this case, with the collector emitter circuit, followed
by the design of the bias circuitry.
• If you have a power constraint, you typically start by determining the maximum allowable
collector current, and hence the maximum available gain.
Typically, you would derive exact component values using a design
method you choose (see below), substitute the nearest “real”
component values, and simulate the circuit using SPICE before building
a prototype.
• You may need to consider the effects of load resistances on the overall gain. For
capacitivelycoupled amplifiers, these loads do not affect the bias point.
• At this point we won’t worry much about designing to take into account the frequency
response of the BJT amplifier (this is covered below).
REMINDER: You should always remember these
relationships (they are very handy and we use them below!).
I
E
= β + 1 I
B
and I
B
=
I
C
β
thus,
I
E
=
β + 1
β
I
C
(That makes sense because I
E
is larger than I
C
...)
G. Kovacs ©1997 Page 82
If R is bypassed, you can
adjust its value to control I
and hence gain.
E
C
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
12: BIAS CIRCUITRY DESIGN
• Try to choose R
B1
and R
B2
so that I
B
is small compared to the current flowing through
R
B1
and R
B2
...
• To select R
B
for bias stability (resistant to changes in β and temperature, T), a rule of
thumb is to set the bias network (R
B1
and R
B2
) up so that approximately 10% of the total
current through R
B1
and R
B2
goes into the base of the transistor.
• Sedra and Smith suggest a rule of thumb for selecting the bias resistors that is to set
the voltage divider current to 0.1 I
E
.
• So, if you have I
E
, you can do it directly.
• If you don't, you can calculate I
E
from I
B
...
I
BIAS
≈
V
CC
R
B1
+ R
B2
= 0.1 I
E
= 0.1 1 + β I
B
• To set V
B
at roughly 1/3 of the supply voltage, you can write...
1
3
=
R
B2
R
B1
+ R
B2
...this is
1
3
V
CC
 V
EE
=
R
B2
R
B1
+ R
B2
V
CC
 V
EE
• Combining these, you get,
R
B2
=
V
CC
0.3 I
E
and
R
B1
=
V
CC
0.1 I
E
 R
B2
G. Kovacs ©1997 Page 83
• Substitute V
CC
V
EE
for V
CC
in the above equations if dual power supplies are used.
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
13: REVISITING LOAD LINES
• The load line approach lets us solve two simultaneous equations: 1) the I
C
versus V
CE
curve of the BJT (remember that only particular curves are plotted for specific I
B
values, but effectively there are an infinite number of them... it is really a 3D plot) and 2)
the resistor's I
C
versus V
CE
line due to the collector resistor (Ohm's Law).
• "Wiggling" the base current moves you up and down along the constraint of the
resistor load line, generating a corresponding output "wiggle" in v
ce
and consequently
v
out
!
• NOTE that here we assume a simple CE amplifier (no R
E
), so V
CE
= V
C
.
V
CE
C
I
I
B
Resistor load line with slope = 
1
V
CC
V
CC
c
i
v
ce
R
C
R
C
• The slope of the resistor load line controls the gain  the larger R
C
, the shallower the
G. Kovacs ©1997 Page 84
slope of the line and the larger the output signal swing...
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
V
CE
C
I
V
CC
V
CE MAX
V
CC
=
R
C
Increasing R
C
• This helps explain the effect of R
C
on the gain of the simple CE amplifier, where,
A
V
· −g
m
R
C
• Combining two resistive load lines into one (somewhat crowded) plot helps bring it
together.
G. Kovacs ©1997 Page 85
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
V
CE
C
I
I
B
V
CC
V
CC
c
i
v
ce
R
C
Intercept =
• Note that g
m
is effectively set by the quiescentpoint (no signal current, or
Qpoint) current!!!... moving to a lower Qpoint also decreases gain... what is not
illustrated by the simple load line drawings here is that the slopes of the V
CE
versus I
C
curves for the BJT increase with increasing I
C
.
• The Early Voltage, discussed later on with respect to BJT current sources, represents
an effective output resistance, r
o
, between the collector and emitter... this gives rise to
the gradually increasing slope of the I
C
versus V
CE
curves for increasing I
C
(see
pages 207  208 in S & S).
r
o
=
V
A
I
C
G. Kovacs ©1997 Page 86
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
V
I
I
V
Slope = 1/r
o
• Remember that in the above figure, the I
C
versus V
CE
curves represent equal base
current steps between them.
• If I
C
is increased to increase g
m
, the DC collector voltage, V
C
, will come closer to
ground (i.e. V
CE
will decrease) until the transistor is saturated.
V
CE
C
I
I
B
Cutoff Region
Active
Region
Saturation
Region
• This means that, for a given supply voltage, there is a limit to the amount of gain one
can get from a given transistor and power supply voltage... note that the V
CE
intercept
of the resistor load line is determined entirely by the supply voltage.
• One can solve for this maximum gain by setting V
CE
to some minimum voltage, V
CEmin
to allow a reasonable swing (e.g. 1 V) and observing that the gain equation for a
common emitter amplifier is independent of R
C
... (the input divider is ignored here),
G. Kovacs ©1997 Page 87
A
V
· −g
m
R
C
· −
I
C
v
T
R
C
· −
V
cc
− V
CEmin
( )
R
C
1
v
T
¸
¸
_
,
R
C
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
14: DESIGN EXAMPLE #1: METHOD OF SEDRA &
SMITH
Specifications:
DC power dissipation: P
D
< 30 mW
Power supply: 12 VDC
Voltage Gain: 50X
Load: Resistive, 50 kΩ
Assume R
S
= 0 Ω
Must use 2N2222A Transistor (NPN, β = 150 measured)
CC
v
s
R
B2
R
C
V
B
V
R
E
R
B1
R
S
R
L
All capacitors are
large (i.e. 100 µF)
so they are essentially
shortcircuits for AC.
v
o
G. Kovacs ©1997 Page 88
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
We know that the total power dissipation (DC, or quiescent) and the voltage gain
both are determined by the choice of current in the collectoremitter circuit.
SO, START THERE!
1) Calculate I
max
at 30mW/12VDC = 2.5 mA
2) We know that the gain of this amplifier with R
E
bypassed (shorted to ground
for AC signals) is,
A
v
·
v
o
v
be
· −g
m
R
C
and that
g
m
=
I
C
V
T
choosing I
C
= 2.0 mA, we see that g
m
is,
g
m
·
I
C
v
T
·
2 mA
25.9 mV
· 0.0772 Ω
1
now we can calculate R
C
(assuming R
E
is fully bypassed for AC)
R
C
·
−A
V
g
m
·
50
0.0772
· 647.5 Ω
assuming the 1/3, 1/3, 1/3 rule of Sedra & Smith, we also allow V
CE
= 12/3 = 4V and
can choose R
E
by,
R
E
·
V
B
− V
BE
I
E
·
V
B
− V
BE
β+ 1
β
¸
¸
_
,
I
C
·
4 − 0.7
150 +1
150
¸
¸
_
,
0.002
·1,639 Ω
3) Now calculate the required base current,
I
B
=
I
C
β
=
2 mA
150
= 0.0133 mA
G. Kovacs ©1997 Page 89
In this "quick" design, we are ignoring the
effects of the input voltage divider (bias
resistors). This should not be done in general,
since these resistors will typically reduce
overall gain (unless R is zero).
S
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
4) Knowing the base current, and that V
B
should be 1/3 of the way up from
ground to +12 V, and assuming that the current in the voltage divider is 0.1I
E
, we
can calculate the bias resistor circuit,
R
B2
·
V
CC
0.3I
E
·
V
CC
0.3 1+ β ( )I
B
·
12
0.3 1+150 ( ) 13.3 µA
·19.92 kΩ
R
B1
·
V
CC
0.1I
E
− R
B2
·
V
CC
0.11+β ( )I
B
− R
B2
·
12
0.11+150 ( ) 13.3 µA
−19.92 kΩ · 39.83 kΩ
5) Choose “real” component values close to those calculated...
R
C
= 680 Ω R
E
= 1.8 kΩ
R
B1
= 39 kΩ R
B2
= 18 kΩ
6) SIMULATE the circuit using SPICE...
EE113 Example Design #1 CommonEmitter Amplifier
*dc components
Vcc VCC 0 12
R1 VCC VBG 39K
Q1 VCG VBG VEG TRANSMODEL
R2 VBG 0 18K
RC VCC VCG 680
RE VEG 0 1.8K
*.MODEL TRANSMODEL NPN (IS=1.3E14)
*ac components
Vss 0 Vin AC 10mV
RS Vin VX 1
C3 VX VBG 100UF
C6 VEG 0 100UF
C5 Vout VCG 100UF
RLL Vout 0 50K
.MODEL TRANSMODEL NPN (BF=150 IS=1.3E14
+ TF=.9N CJE=6P CJC=5P)
*input sweep for Bode plot
.AC DEC 10 1 100MEG
.PROBE
.end
G. Kovacs ©1997 Page 90
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
1 . 0 h 1 0 0 h 1 0 Kh 1 . 0 Mh 1 0 0 Mh
Fr e q u e n c y
V( Vo u t ) / V( Vi n )
5 0
4 0
3 0
2 0
1 0
0
C1 =
C2 =
d i f =
5 9 . 3 7 2 K, 4 2 . 8 1 6
5 9 . 3 7 1 K, 4 2 . 2 7 1
1 . 0 0 0 0 , 5 4 5 . 3 3 5 m
5 9 . 3 7 1 K, 4 2 . 2 7 1
C1 =
C2 =
d i f =
5 9 . 3 7 2 K, 4 2 . 8 1 6
5 9 . 3 7 1 K, 4 2 . 2 7 1
1 . 0 0 0 0 , 5 4 5 . 3 3 5 m
We see that the midband gain is only 43X !!!!
Noting that,
g
m
·
I
C
v
T
·
β
β+ 1
¸
¸
_
,
I
E
v
T
·
V
B
− V
BE
β
β +1
¸
¸
_
,
R
E
v
T
≈
127.4
R
E
Ω
1
we see that the gain can be directly controlled via R
E
. To do this, we scale the original
computed R
E
(1.639 kΩ) by the ratio of the actual to desired gain (0.84) to obtain 1.41
G. Kovacs ©1997 Page 91
kΩ, and choose a real valued R
E
= 1.5 kΩ and simulate again....
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
1 . 0 h 1 0 0 h 1 0 Kh 1 . 0 Mh 1 0 0 Mh
Fr e q u e n c y
V( Vo u t ) / V( Vi n )
6 0
5 0
4 0
3 0
2 0
1 0
0
C1 =
C2 =
d i f =
7 3 . 7 7 8 K, 5 0 . 8 6 1
7 3 . 7 7 7 K, 5 0 . 2 6 6
1 . 0 0 0 0 , 5 9 5 . 5 2 1 m
7 3 . 7 7 7 K, 5 0 . 2 6 6
C1 =
C2 =
d i f =
7 3 . 7 7 8 K, 5 0 . 8 6 1
7 3 . 7 7 7 K, 5 0 . 2 6 6
1 . 0 0 0 0 , 5 9 5 . 5 2 1 m
A gain of 50.9 is probably close enough!
Checking the power drain as computed by SPICE,
VOLTAGE SOURCE CURRENTS
NAME CURRENT
Vcc 2.177E03
Vss 0.000E+00
TOTAL POWER DISSIPATION 2.61E02 WATTS
G. Kovacs ©1997 Page 92
Looks fine! Build it!
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
15: WHEN DOES THE INPUT DIVIDER MATTER?
• If the input divider is considered upfront in a design where R
S
≠ 0, the power
dissipation constraint in this example still is the best place to start, and we begin by
choosing I
C
= 2 mA, giving g
m
= 0.0772 Ω
1
.
• Then you would refer to the "full" gain equation and note that the input divider values
would need to be determined before choosing R
C
,
−50 · A
v
·
v
o
v
s
· −
R
B
 r
π
R
S
+ R
B
r
π
¸
¸
_
,
g
m
R
C
• Assuming the 1/3, 1/3, 1/3 rule of Sedra & Smith, we also allow V
CE
= 12/3 = 4V and
can choose R
E
by (same as before),
R
E
=
V
B
 V
BE
I
E
=
V
B
 V
BE
β + 1
β
I
C
=
4  0.7
150 + 1
150
2 mA
= 1,639 Ω
• Again, we calculate the required base current,
I
B
=
I
C
β
=
2 mA
150
= 0.0133 mA
• Knowing the base current, and that V
B
should be 1/3 of the way up from ground to +12
V, and assuming that the current in the voltage divider is 0.1I
E
, we can calculate the bias
resistor circuit,
R
B2
=
V
CC
0.3 I
E
=
V
CC
0.3 1 + β I
B
=
12
0.3 1 + 150 0.0133 mA
= 19.92 KΩ
R
B1
=
V
CC
0.1 I
E
 R
B2
=
12
0.1 1 + 150 0.0133 mA
 19.92KΩ = 39.83 KΩ
• Thus far, the calculations have been the same, but the final calculation of R
C
must be
done differently. First, you need to compute r
π
,
r
π
·
β
g
m
·
150
0.0772
·1943 Ω
G. Kovacs ©1997 Page 93
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
• Here there is a choice. You can pick real values for R
B1
and R
B2
and solve the "full"
gain equation for R
C
(knowing R
S
). Or, you can make the calculation using the
"accurate" values and then pick "real" ones later. The first choice is usually better, since
then there is only one resistor that is not accurately represented in the calculations
when the nearest "real" value is chosen (R
C
).
• Choosing "real" values, R
B1
= 39 kΩ and R
B2
= 18 kΩ. Then the gain equation can be
solved for R
C
(assuming R
S
= 50Ω in this example),
R
C
·
−A
v
R
B
 r
π
R
S
+ R
B
r
π
¸
¸
_
,
g
m
R
C
·
−A
v
R
B1
 R
B2
 r
π
R
S
+ R
B1
 R
B2
 r
π
¸
¸
_
,
g
m
·
50
39kΩ 18kΩ1.9kΩ
50Ω +39kΩ18kΩ1.9kΩ
¸
¸
_
,
0.0772Ω
−1
· 667.3Ω
For which you would still choose 680 Ω as the nearest "real" value.
• The only remaining task is to choose a "real" value for R
E
, and again the nearest value
is 1.8 kΩ.
• This gives the same result as we got by ignoring the input divider!!!
R
C
= 680 Ω R
E
= 1.8 kΩ
R
B1
= 39 kΩ R
B2
= 18 kΩ
• We would still have had to adjust R
E
to get the correct gain after simulating the circuit!
HOWEVER (!), if R
S
was greater than 50 Ω (for which the input divider gives a gain
factor of 0.971), the input divider could have been much more significant. For example,
if R
S
were 1 kΩ, the gain gain factor from the input divider would be 0.622, which would
mean that for an overall gain of 50, the CE stage itself would need a gain boost of
(0.622)
1
, or 1.61 times.
• The moral of this story is that the way to tell whether or not the input divider is
significant is to compare the magnitude of R
B1
R
B2
r
π
to R
S
. If R
S
is much smaller, it can
often be ignored. If R
S
is comparable, the attenuation from the input divider can be
quite significant.
• Also note that r
π
is a function of I
C
, so adjusting I
C
does have an effect on the input
divider.
G. Kovacs ©1997 Page 94
r
π
·
β
g
m
·
βv
T
I
C
Chapter 6: BJT FREQUENCY RESPONSE
Chapter 6: BJT FREQUENCY
RESPONSE
... I have wasted a lot of time being fooled by a bad component 
what do I do? I usually WIDLARIZE it, and it makes me feel a
lot better. How do you WIDLARIZE something? You take it over
to the anvil part of the vice, and you beat on it with a hammer,
until it is all crunched down to tiny little pieces, so small that you
don’t even have to sweep it off the floor. It makes you feel better.
Bob Pease, Analog Guru, National Semiconductor
1: OBJECTIVES
• To begin estimating frequency responses of BJT amplifiers by learning about:
The junction capacitances of BJT’s and their effects on frequency response.
The diffusion capacitance of BJT’s and its effect on frequency response.
A frequency dependent and more complete Hybridπ model.
Estimating frequency response of a CE amplifier and seeing how the Miller
Effect applies in that case.
READ S&S Section 7.5
2: JUNCTION CAPACITANCES OF BJT’S
• Both of the PN junctions in a BJT have junction capacitances which end up reducing
the frequency response of the transistor.
• They are often referred to as C
JE
and C
JC
(or C
BE
and C
BC
, for the baseemitter and
G. Kovacs ©1997 Page 95
basecollector capacitances, respectively) as illustrated below:
=
ε
o
ε
R
A
d
Chapter 6: BJT FREQUENCY RESPONSE
G. Kovacs ©1997 Page 96
C =
ε
o
ε
R
A
d
This is analogous to a
parallelplate capacitor.
Chapter 6: BJT FREQUENCY RESPONSE
BIAS DEPENDENCE OF JUNCTION CAPACITANCES
• The junction capacitances are (nonlinearly) determined by the bias voltages as shown
below.
• Actual estimation of the junction capacitances can be tricky. The equation shown
below works fairly well for reversebiased junctions.
C
j
·
C
jo
1−
V
V
j
¸
¸
_
,
m
j
where C
jo
is the unbiased junction capacitance, V
j
is the builtin potential of the junction
(often called Ψ
o
), and m
j
is a constant between 0.3 and 0.5 depending on whether the
junction is graded (0.3) or abrupt (0.5).
• For forwardbiased junctions, the approximately generally used is that the capacitance
is double that of the same junction under no bias.
G. Kovacs ©1997 Page 97
Chapter 6: BJT FREQUENCY RESPONSE
3: DIFFUSION CAPACITANCE
• The diffusion capacitance appears between the BASE and EMITTER and is the result
of the diffusion of minority carriers across the base.
• One can think about it as follows: if there is a certain number of carriers
diffusing through the base, it takes a certain amount of time to adjust that number
if the applied signal changes. Time delays in controlling currents are the hallmark
of capacitances.
• For NPN transistors, electrons are the minority carriers in the base.
I
C
·
Q
B
τ
f
·
charge in base
transit time for that charge to be swept through the base
Q
B
= I
C
τ
f
C
d
≡
dQ
B
dV
BE
= τ
f
dI
C
dV
BE
= g
m
τ
f
= g
m
W
B
2
2 D
B
G. Kovacs ©1997 Page 98
diffusivity of minority
carriers in base
base
width
Chapter 6: BJT FREQUENCY RESPONSE
SUMMARY:
• AT LOW I
C
> C
jc
and C
je
DOMINATE
• AT HIGH I
C
> C
d
DOMINATES
4: COMPLETE HYBRIDπ MODEL
B
r
r
x
π
v
be
+

C
E
r
o
g
m
v
be
BC
C
C C
d JE
Both voltage and currentdependent
capacitances appear here...
This capacitance is voltagedependent only.
• r
x
is the ohmic resistance of the base contact and is a few tenths of ohms normally.
• r
o
is a resistance that models the slight effect of collector voltage on collector current in
the active region of operation (the curves are not exactly flat!).
• r
o
is inversely proportional to the DC bias current and is typically tens of thousands of
Ohms...
r
o
=
V
A
I
C
V
I
I
V
Slope = 1/r
o
• V
A
is the Early voltage, which is discussed again below with respect to current
G. Kovacs ©1997 Page 99
sources.
These should be
horizontal lines for an
ideal current source...
extra current through
r determines slope.
o
Chapter 6: BJT FREQUENCY RESPONSE
• In practice, the two baseemitter capacitances are lumped together as c
π
(or C
BE
in
SPICE).
B C
E
r
r
r
x
o
π
g
m
v
be
v
be
π
BC
C
C
+

• Typically r
x
is in series with R
s
(source resistance of the input signal generator) and r
o
shunts R
C
(the collector resistor, not shown above because it would be a part of the
external circuit).
AVOID NOTATION CONFUSION!!!
C
π
≡ C
d
+ C
JE
V
BE
in SPICE, this is C
BE
C
JC
V
BE
= C
BC
in SPICE
5: GAIN (β) VERSUS FREQUENCY FOR BJT'S
• In order to understand the frequency response characteristics of the BJT, one needs
to study the time constants of the BJT alone.
g
m
v
be π
BC
C
C
i
c
i
b
r
r
x
π
v
be
+

r
o
G. Kovacs ©1997 Page 100
Chapter 6: BJT FREQUENCY RESPONSE
β S ( ) ·
i
c
i
b
·
g
m
v
be
i
b
v
be
·
i
b
1
r
π
+ S C
π
+ C
BC
( )
β S ( ) ·
i
c
i
b
·
g
m
1
r
π
+S C
π
+ C
BC
( )
·
g
m
r
π
1+ Sr
π
C
π
+ C
BC
( )
·
β
o
1+Sr
π
C
π
+ C
BC
( )
LOOKS LIKE A LOWPASS FILTER!
β S =
β
o
1 + S r
π
C
π
+ C
BC
Hs =
K
1 +
s
ω
o
OR
Kω
o
ω
o
+ s
=
A
ω
o
+ s
where ω
0
= 2πf
c
=
1
"RC"
=
1
r
π
C
π
+ C
BC
β
ω
β
o
or β
DC
β · 1
ω
o
ω
t
A graphical representation of the decrease of β as frequency increases. β is 3 dB down
G. Kovacs ©1997 Page 101
at ω
o
and β = 1 at ω
t
. Note that Sedra and Smith refers to ω
o
as ω
b
.
Chapter 6: BJT FREQUENCY RESPONSE
• A key transistor parameter, ω
t
, can now be derived. The unitygain bandwidth, ω
t
(or f
t
in Hertz) is the frequency at which the current gain (β) of a transistor in commonemitter
configuration, with its output shortcircuited, drops to unity.
β ω
t
= 1 =
β
o
1 + S r
π
C
π
+ C
BC
=
β
o
1 + S
β
o
g
m
C
π
+ C
BC
solving,
1 + S
β
o
g
m
C
π
+ C
BC
= β
o
S = ω
t
=
β
o
 1
β
o
g
m
C
π
+ C
BC
≈
1
1
g
m
C
π
+ C
BC
• One can write this to show the contributions from different capacitance mechanisms,
2πf
t
= ω
t
≈
1
1
g
m
C
π
+ C
BC
=
1
τ
f
+
C
JE
g
m
+
C
BC
g
m
• One can increase I
c
to increase bandwidth until τ
f
dominates.
DESIGN NOTE: As you increase I
C
, g
m
goes up, so the last two terms decrease... The
maximum possible f
t
is then obtained, but then falls off if I
C
continues to be increased
because β
o
falls off at high current anyway.
G. Kovacs ©1997 Page 102
This is independent of ß !
o
Chapter 6: BJT FREQUENCY RESPONSE
• By inspection of the above, it can be seen that that,
ω
o
=
1
r
π
C
π
+ C
BC
(the 3dB frequency for β)
TYPICAL VALUES FROM SPICE (EE122 PRELAB)
C
BE
= 2.2 X 10
11
(C
d
+ C
JE
(V
BE
))
C
BC
= 2.3 X 10
12
g
m
= 1.43 X 10
2
2πf
t
= ω
t
≈
1
2.2 X 10
11
1.43 X 10
2
+
2.3 X 10
12
1.43 X 10
2
= 5.88 X 10
8
Radians
f
t
= 9.35 X 10
7
Hertz > 93.5 MHz
THIS IS ALMOST EXACTLY WHAT SPICE PREDICTS (See Prelab)
G. Kovacs ©1997 Page 103
τ Dominates
f
g Controls Bandwidth
m
Chapter 6: BJT FREQUENCY RESPONSE
6: COMMONEMITTER CONFIGURATION AND
MILLER CAPACITANCE
• The purpose of this effort is to examine the frequency response of the common emitter
amplifier and see how the gain of the amplifier makes C
BC
look really large thanks to the
Miller Effect.
g
m
v
be
BC
C
r
r
x
π
v
be
+

π
C
R
C
R
s 1 2
v
o
v
s
• For convenience, use R
S
'
= R
S
+ r
x
• START by looking at the circuit... It looks like any capacitance in the input circuit
shorts out v
be
as the frequency goes up...
• Summing currents at node
1
v
s
 v
be
R
s
'
+ v
o
 v
be
S C
BC
=
1
r
π
+ S C
π
v
be
• Summing currents at node 2
g
m
v
be
= 
v
o
R
C
 v
o
 v
be
S C
BC
• Assuming that v
o
>> v
be
, can write,
v
o
≈ 
g
m
v
be
1
R
C
+ S C
BC
• To further simplify, use the basic CE amplifier gain equation (assuming 1/R
c
>> SC
BC
),
v
o
≈  g
m
R
C
v
be
G. Kovacs ©1997 Page 104
Chapter 6: BJT FREQUENCY RESPONSE
• PLUG IT ALL TOGETHER....
A
v
=
v
o
v
s
≈
 g
m
R
C
r
π
r
π
+ R
s
'
1 + S r
π
 R
s
'
C
π
+ C
BC
1 + g
m
R
C
• Reminder of Miller's Theorem:
Z
1
=
Z
f
1  A
> Miller impedance at the input of the amplifier
Z
2
= Z
f
A
1  A
> Miller impedance at the output of the amplifier
THE CIRCUIT REWRITTEN AS THE MILLER EQUIVALENT
• (Note that the output Miller equivalent of C
BC
has been neglected because it is very
small... also we are ASSUMING that the hybridπ amplifier model's gain is not changed
by C
BC
... this is true for a model, but not always for a real amplifier and especially for
larger capacitances, perhaps external to the transistors!)
r
π
v
be
+
C
π 
R
s
'
C
M
g
m
v
be
R
C
v
o
≈C
BC
v
s
C
M
= C
BC
1 + g
m
R
C
• So, in effect, a small input current makes g
m
R
C
times as much current flow into the
output circuit... This makes C
BC
seem much larger from the point of view of the input
circuit!
• The dominant pole (more on this later) of CE amplifiers is determined by the input
G. Kovacs ©1997 Page 105
resistance and the input capacitances, C
π
and C
M
...
C
MILLER
The numerator is the DC gain.
Chapter 6: BJT FREQUENCY RESPONSE
7: USING THE MILLER IDEA UP FRONT
• Here we are going to apply the Miller Effect idea directly to the complete hybridπ
model to simplify the circuit analysis. Remember that the idea is to convert an impedance
across an amplifier into two separate (grounded) ones at the input and output.
• Here (again) is the "complete" hybridπ model for a bipolar transistor:
B
r
r
x
π
v
be
+

C
E
r
o
g
m
v
be
BC
C
C C
d JE
Both voltage and currentdependent
capacitances appear here...
This capacitance is voltagedependent only.
• Now throw in a collector resistance, R
C
, to make it into a common emitter amplifier (no
R
E
)... also, let's neglect r
x
and r
o
!
g
m
v
be
BC
C
r
π
v
be
+

π
C
R
C
R
s
v
o
v
s
The
Amplifier
The
Impedance
Across the
Amplifier
• A very important point here is that the gain used in the Miller equation must be the
gain seen by the component in question (here, that is C
BC
). Thus, if there is an input
voltage divider due to R
S
, for example, one needs to compute the gain C
BC
deals with,
and the input divider does not affect that!
G. Kovacs ©1997 Page 106
Chapter 6: BJT FREQUENCY RESPONSE
• The internal gain (relative to C
BC
) is g
m
R
c
and one can rewrite the circuit as the Miller
equivalent, by multiplying C
BC
by the (1  gain) and putting it at the input, and multiplying
C
BC
by ≈1 and putting it at the output!
r
π
v
be
+
C
π 
R
s
'
C
M
g
m
v
be
R
C
v
o
≈C
BC
C
M
= C
BC
1 + g
m
R
C
• Note that the maximum possible range of the scaled copy of C
BC
at the output is
between one and two times the value of C
BC
, corresponding to infinite gain and a gain of
one for the amplifier that C
BC
straddles. This is discussed in detail below, and its "exact"
value is given by,
C
out
· C
BC
1+ g
m
R
C
 R
L
r
o
( )
g
m
R
C
 R
L
 r
o
( )
• One can sometimes ignore C
out
at the output because if R
C
is relatively low, the time
constant due to a few picoFarads is not much... however, at high frequencies, it may
start to matter!
• Also, if we are talking about a multistage amplifier, C
out
now appears at the INPUT of
the next stage... then it can be significant!
PLEASE REMEMBER THIS STUFF... WE WILL
USE THIS TECHNIQUE OF TAKING THE
MILLER EQUIVALENT OF C
BC
MANY TIMES IN
THE FUTURE!
G. Kovacs ©1997 Page 107
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
Chapter 7: GENERAL AMPLIFIER
FREQUENCY RESPONSE
If I tell you to neglect the effects of the junction
capacitances, you’d better just do it, little man!
Arnold Schwarzenegger as RoboProf
1: OBJECTIVES
• To learn about:
The generalized frequency response of capacitorcoupled amplifiers.
Shortcircuit and opencircuit time constants for approximating the
response of amplifiers.
How coupling and bypass capacitors affect the (LOW) frequency
response of amplifiers by studying a common emitter example.
READ S&S Sections 7.2, 7.4 (don’t worry that
it’s done with a FET!), and 7.6
REVIEW: How to make Bode Plots (S & S Section 7.1)
THE POINT OF ALL OF THIS: We want to know how to
take a circuit and determine its frequency response.
An overall goal of this is to be able to analyze a circuit
and then know how to DESIGN a circuit with the
frequency response we need!
G. Kovacs ©1997 Page 108
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
2: GENERALIZED CAPACITORCOUPLED AMPLIFIER
FREQUENCY RESPONSE
• The two plots above represent DCcoupled (left) and ACcoupled (right) amplifiers.
The single transistor amplifiers generally need input and output coupling capacitors so
that external DC voltages do not shift the bias point. Thus, these will have ACcoupled
responses. Differential amplifiers (presented below) are DCcoupled.
• For the ACcoupled response, near ω
L
and below, the amplifier’s response acts like a
highpass filter.
• For both cases, near ω
H
and above, the amplifier’s response acts like a lowpass
filter.
• The (hopefully) flat region in the middle is the “midband.” We typically talk about the
“midband gain” of an ACcoupled amplifier.
• It is important to note that one can look at the transfer function of an ACcoupled
amplifier as the product of the highpass, midband, and lowpass transfer functions,
A S = A
M
F
L
S F
H
S
Reminder: The impedance of a capacitor is,
Z
C
·
1
SC
or, for sinusoidal steady  state =
1
2πfC
G. Kovacs ©1997 Page 109
Thus, they are opencircuits at DC and shorts at very high frequencies.
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
3: DOMINANT POLES
• Dominant poles are poles that dominate over the others in the transfer function so that
the low or highfrequency response of an amplifier can be approximated as a firstorder
high or lowpass filter. (REMEMBER that opamps are DELIBERATELY designed to
have a single dominant pole > this is so we will be able to predict their response and
use it!).
DOMINANT LOWFREQUENCY POLE
• If there is a dominant lowfrequency pole (this is the one that acts like a HIGHPASS
filter), the lowfrequency response can be approximated as,
F
L
S ≈
S
S + ω
P1
• So, if the “full” lowfrequency response is,
F
L
S =
S + ω
Z1
S + ω
Z2
•••• S + ω
ZNL
S + ω
P1
S + ω
P2
•••• S + ω
PNL
this means that ω
P1
is at a much HIGHER frequency than the other poles so that it can
DOMINATE...
• Note that the number of poles and zeros must be equal (at that point) if the
function “flattens out” in the midband!
• If there is no dominant pole, you can make the following approximation,
ω
L
≈ ω
Pn
2
∑
n = 1
N
 2 ω
Zn
2
∑
n = 1
N
DOMINANT HIGHFREQUENCY POLE
• If there is a dominant highfrequency pole (this is the one that acts like a LOWPASS
filter), the highfrequency response can be approximated as,
F
H
S ≈
1
1 +
S
ω
P1
G. Kovacs ©1997 Page 110
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
• So, if the “full” highfrequency response is,
F
H
S =
1 +
S
ω
Z1
1 +
S
ω
Z2
•••• 1 +
S
ω
ZNH
1 +
S
ω
P1
1 +
S
ω
P2
•••• 1 +
S
ω
PNH
this means that ω
P1
is at a much LOWER frequency than the other poles so that it can
DOMINATE...
Note that the number of poles and zeros must be equal (at that point) if the
function “flattens out” in the midband!
• If there is no dominant pole, you can make the following approximation,
ω
H
≈
1
1
ω
Pn
2
∑
n = 1
N
 2
1
ω
Zn
2
∑
n = 1
N
4: SHORTCIRCUIT AND OPENCIRCUIT TIME
CONSTANT METHODS FOR APPROXIMATING THE
RESPONSE OF AMPLIFIERS
• This is a very useful method for approximating upper and lower cutoff frequencies,
assuming that there is a dominant pole (it works pretty well if there isn’t too!).
• Depending on which of the two methods you use, for each capacitor in the circuit, you
replace the capacitors in the circuit, except for the one you are looking at, with either
shorts or opens.
• Two key rules apply to both methods:
1) DISABLE all independent sources...
voltage sources > SHORT CIRCUIT
current sources > OPEN CIRCUIT
2) DO NOT remove or “disable” dependent
sources!
G. Kovacs ©1997 Page 111
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
4.1 OPENCIRCUIT TIME CONSTANTS FOR UPPER
CUTOFF FREQUENCY APPROXIMATION
• All you have to do is step through the process of calculating a “local” RC timeconstant
for each capacitor in the circuit.
For each capacitor you set all other capacitances (other
than the one you are looking at) to zero (i.e. they become
OPEN CIRCUITS as if they have not yet had any effect on
the amplifier's rolloff) and determine the resistance, R
io
,
seen by C
i
(the capacitor you are working with).
• Repeat this process for each capacitor.
• The upper cutoff frequency is then approximately given by,
ω
H
≈
1
C
i
R
io ∑
i
This approach can tell you which of the capacitances in a circuit is most significant
in determining the highfrequency response!
4.2 SHORTCIRCUIT TIME CONSTANTS FOR LOWER
CUTOFF FREQUENCY APPROXIMATION
• Again, all you have to do is step through the process of calculating a “local” RC
timeconstant for each capacitor in the circuit.
For each capacitor you set all other capacitances (other
than the one you are looking at) to INFINITY (i.e. they
become SHORT CIRCUITS as if they have already become
low enough impedance to neglect) and determine the
resistance, R
iS
, seen by C
i
(the capacitor you are working
with).
G. Kovacs ©1997 Page 112
• Repeat this process for each capacitor.
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
• The lower cutoff frequency is then approximately given by,
ω
L
≈
1
C
i
R
iS
∑
i
This approach can tell you which of the capacitances in a circuit is most significant
in determining the lowfrequency response!
5: FREQUENCY RESPONSE OF THE CE AMPLIFIER
• This circuit should be familiar from previous lectures and from EE122! The point of
this section is to GRIND through the derivation of its frequency response “the hard way,”
and then compare to the response obtained using the open and shortcircuit time
constant approximations....
VCC
v
s
R
B2
R
C
V
B
R
E
R
B1
E
C
C1
C
C2
C
v
o
S
R
G. Kovacs ©1997 Page 113
since the emitter resistance cuts
down the gain, as it gets shorted
out, the gain GOES UP...
that means C introduces a
ZERO!!!!
E
As the frequency goes up,
the emitter resistance gets
shorted out by this capacitance...
This capacitance is meant
to block DC and pass AC
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
v
o
R
C
g
m
v
be
R
E
C
E
Z
E
Z
S
R
S
R
B
r
π
v
be
+

C
C1
v
s
Z
in
Z
S
=
1 + S R
S
C
C1
S C
C1
Z
E
=
R
E
1 + S R
E
C
E
• REMEMBER from the case where R
E
is not bypassed, that one can take R
E
into
account when looking from the base using the rule that lets you multiply its resistance
by (1 + β). HERE IS THAT POINT AGAIN, GENERALIZED FOR IMPEDANCES!
One can "transform" an impedance from the emitter or base side of the base
emitter circuit to the opposite side just as one would for just a resistance. An
impedance in the emitter circuit will appear as being (β+1) times larger when
viewed from the base terminal. Similarly, an impedance in the base circuit will
appear as being (β+1) times smaller when viewed from the emitter terminal.
Using this, we can write,
Z
in
· r
π
+ β +1 ( )Z
E
(here R
in
is for the transistor, NOT the entire amplifier!)
G. Kovacs ©1997 Page 114
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
5.1 LOW FREQUENCY RESPONSE
Z
S
v
s
R
B
Z
in
i
in
i
in
·
v
s
1+
Z
in
R
B
¸
¸
_
,
Z
S
+ Z
in
FROM HERE ON, ASSUME R
B
= INFINITY TO SIMPLIFY THINGS!
i
in
·
v
s
Z
s
+ Z
in
·
v
s
Z
s
+ r
π
+ β+ 1 ( )Z
E
·
v
s
1+ SR
S
C
C1
SC
C1
+ r
π
+ β +1 ( )
R
E
1+ SR
E
C
E
¸
1
]
1
• Grinding through the math and regrouping terms,
i
in
v
s
=
S C
C1
1 + S R
E
C
E
1 + S C
C1
R
S
+ r
π
+ β + 1 R
E
+ R
E
C
E
+ S
2
R
S
C
C1
R
E
C
E
+ r
π
C
C1
R
E
C
E
• A nice, secondorder transfer function for the input conductance....
• Looking at the numerator first, there is a ZERO at ω = 0 (because of C
C1
), and
ZEROs are what we are typically expecting for low frequency analysis.
• There is a ZERO at ω =
1
R
E
C
E
v
o
v
s
= β R
C
i
in
v
o
v
s
=
βR
C
C
C1
R
E
C
E
S S +
1
R
E
C
E
1 + S C
C1
R
S
+ r
π
+ β + 1 R
E
+ R
E
C
E
+ S
2
R
S
C
C1
R
E
C
E
+ r
π
C
C1
R
E
C
E
G. Kovacs ©1997 Page 115
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
v
o
v
s
=
β R
C
R
S
+ r
π
S S +
1
R
E
C
E
S
2
+ S
1
C
E
R
E

R
S
+ r
π
β + 1
+
1
C
C1
R
S
+ r
π
+
1
R
E
C
E
C
C1
R
S
+ r
π
• This is of the form,
F
L
S =
S S + ω
Z1
S + ω
P1
S + ω
P2
=
S S + ω
Z1
S
2
+ S ω
P1
+ ω
P2
+ ω
P1
ω
P2
• This was a lot of work!!!! Compare and see if the short and opencircuit time constant
methods agree. Begin by computing ω
L
using the shortcircuit time constant
approximation.
• For C
E
, (REMEMBER that we assumed R
B
= INFINITY for simplicity) short C
C1
,
R
E
C
E
Z
E
R
S
R
B
r
π
∞
R
g
m
v
be
C
v
be
+

v
o
• Note that R
C
is effectively "invisible" to C
E
, since it cannot effect v
be
(that might not be
true with feedback!) and hence the current flowing in R
E
and C
E
... you can prove it to
yourself by applying a test current... intuitively, R
C
can't be "seen" through the current
source.
• Taking into account (ß+1) scaling,
R
CES
· R
E

R
S
+ r
π
β +1
· R
E
 r
e
+
R
S
β +1
¸
¸
_
,
G. Kovacs ©1997 Page 116
The input voltage
replaced by a short...
ALWAYS KILL
INDEPENDENT
SOURCES!
C replaced by a short
C1
C1S
= R
s
+ r
π
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
• If R
B
≠ ∞ (or is not large enough to assume that), one obtains,
R
CES
· R
E

R
S
R
B
( ) + r
π
β + 1
· R
E
 r
e
+
R
S
 R
B
( )
β + 1
¸
¸
_
,
• For C
C1
short out C
E
,
R
S
R
B
r
π
C
C1
∞
R
C1S
· R
S
+ r
π
• These two RC time constants are the same as those found analytically....
v
o
v
s
=
β R
C
R
S
+ r
π
S S +
1
R
E
C
E
S
2
+ S
1
C
E
R
E

R
S
+ r
π
β + 1
+
1
C
C1
R
S
+ r
π
+
1
R
E
C
E
C
C1
R
S
+ r
π
G. Kovacs ©1997 Page 117
R
C1S
= R
s
+ r
π
R
ES
· R
E

R
S
+ r
π
β +1
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
• If you do not assume R
B
= ∞, include r
x
and r
o
, consider the load resistance R
L
, and
consider an output coupling capacitor C
C2
between the g
m
generator and R
C
 R
L
, you
get the following results (S & S p. 529),
r
r
x
π
v
be
+

g
m
v
be
r
o
R
E
C
E
R
S
R
B
R
C
R
L
v
o
C
C1 C
C2
v
s
Using the method of shortcircuit time constants,
ω
L
≈
1
C
i
R
iS
∑
i
ω
L
≈
1
C
C1
R
C1S
+
1
C
E
R
ES
+
1
C
C2
R
C2
Where,
R
C1S
· R
S
+ R
B
 r
x
+ r
π
( ) [ ]
(INPUT)
(remember that the other caps are shorts!)
R
ES
· R
E

r
x
+ r
π
+ R
B
R
S
( )
β +1
¸
1
]
1
(EMITTER)
(remember about dividing by [β + 1] to reflect the
input circuit impedances to the emitter circuit!)
R
C2S
· R
L
+ R
C
 r
o
( ) (OUTPUT)
G. Kovacs ©1997 Page 118
The zero introduced by C
E
is at ω
ZE
=
1
C
E
R
E
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
5.2 MIDFREQUENCY RESPONSE
• Here assume all coupling and bypass capacitors are shorted (i.e. C
C1
, C
C2
, and C
E
).
You do not really need to worry too much yet (at midband frequencies) about the
parasitic capacitances inside the BJT. They matter at higher frequencies.
v
o
R
C
g
m
v
be
R
S
R
B
r
π
v
be
+

v
s
R
in
For the amplifier!
R
L
R
in
= R
B
 r
π
R
B
= R
B1
 R
B2
v
be
=
R
in
R
s
+ R
in
v
s
v
o
=  g
m
v
be
R
C
 R
L
A
v
=
v
o
v
s
=
R
in
R
s
+ R
in
 g
m
R
C
 R
L
• NOTE that r
o
is not shown above... to take it into account, it can be placed in parallel
with R
C
and R
L
.
• Remember that the overall gain is,
A
V
S ( ) · LFgain S ( ) × MBgain ×HFgain S ( )
G. Kovacs ©1997 Page 119
THIS IS MIDBAND,
SO IGNORE CAPS!
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
• WHAT IF WE ADD IN THE “DETAILS” (i.e. CONSIDER R
L
, r
o
and r
x
)?
g
m
v
be
BC
C
r
r
x
π
v
be
+

π
C
R  R
C
R
s
v
o
v
s
L
in
R
R
B
r
o
R
in
= R
B
 r
x
+ r
π
R
B
= R
B1
 R
B2
v
be
=
R
in
R
s
+ R
in
r
π
r
π
+ r
x
v
s
v
o
=  g
m
v
be
R
C
 R
L
 r
o
A
v
·
v
o
v
s
·
R
in
R
S
+ R
in
¸
¸
_
,
r
π
r
x
+ r
π
¸
¸
_
,
−g
m
( ) R
C
 R
L
 r
o
( )
G. Kovacs ©1997 Page 120
THIS IS MIDBAND, SO IGNORE CAPS!
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
5.3 HIGHFREQUENCY RESPONSE
• From here up in frequency, we will start to worry about the details of what’s inside
the BJT! Now it matters, so use the “full” Hybridπ model...
g
m
v
be
BC
C
r
r
x
π
v
be
+

π
C
R  R
C
R
s
v
o
v
s
L
in
R
R
B
r
o
• Note that in some cases, you may need to compute the required capacitances for the
model using,
2πf
t
=
1
C
π
g
m
+
C
BC
g
m
(remember that C
π
= C
d
+ C
JE
)
• First, it is necessary to calculate the Miller capacitance to obtain a value for the total
input capacitance,
C
in
= C
π
+ C
M
= C
π
+ C
BC
1  A
VQ
NOTE that S & S use C
µ
instead of C
BC
• This is based on A
VQ
which we will define as the voltage gain between the two
transistor terminals straddled by C
BC
.... (IN OTHER WORDS, NOT SIMPLY THE
OVERALL GAIN OF THE AMP!) For this amplifier, you know that you want,
A
VQ
=
v
o
v
be
=  g
m
R
C
 R
L
 r
o
• Thus, we find that the Miller capacitance is given by,
C
M
= C
BC
1 + g
m
R
C
 R
L
 r
o
• The output capacitance, C
out
is given by,
G. Kovacs ©1997 Page 121
C
out
· C
BC
1+ g
m
R
C
 R
L
r
o
( )
g
m
R
C
 R
L
 r
o
( )
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
• This reduces the input circuit to a firstorder lowpass filter (unless C
out
is significant),
so we do not need to use the method of opencircuit time constants.
g
m
v
be
r
r
x
π
v
be
+

in
C
R  R
C
R
s
v
o
v
s
L
in
R
R
B
r
o
C
out
The upper 3 dB frequency of the input circuit is often given by,
ω
H
·
1
R
'
C
in
Where,
R
'
· r
π
 r
x
+ R
B
R
S
( )
and
C
in
· C
π
+C
M
· C
π
+C
BC
1+ g
m
R
C
 R
L
 r
o
( ) [ ]
• NOTE that in some cases, the upper cutoff frequency may be determined by the
output circuit (i.e. C
µ
and r
o
R
C
R
L
). This is very important to check!
• A more general approach would be to use Miller and then opencircuit time constants,
ω
H
≈
1
C
i
R
io
i
∑
¸
1
]
1
·
1
C
in
R
'
+ C
out
R
out
·
1
C
in
r
π
 r
x
+R
B
 R
S
[ ] ( )
+ C
out
r
o
 R
C
 R
L
( )
G. Kovacs ©1997 Page 122
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
6: EXAMPLE FREQUENCY RESPONSE ANALYSIS
VCC
v
s
R
B2
R
C
V
B
R
E
R
B1
E
C
C1
C
C2
C
v
o
S
R
L
R
• Following the example on page 530, Sedra & Smith (Exercises 7.14  7.19)
R
S
= 4 kΩ R
B1
= 8 kΩ R
B2
= 4 kΩ R
E
= 3.3 kΩ R
C
= 6 kΩ R
L
= 4 kΩ
V
CC
= 12 V I
E
≈ 1 mA β
o
= 100 C
π
= 13.9 pF C
µ
= 2 pF r
o
= 100 kΩ
r
x
= 50 Ω
6.1 MIDBAND GAIN CALCULATION
• NEGLECTING R
L
, r
o
and r
x
...
A
V
·
v
o
v
s
·
R
B
r
π
R
S
+ R
B
 r
π
¸
¸
_
,
−g
m
( ) R
C
 R
L
( )
g
m
·
I
C
v
T
≈
1 mA
25.9 mV
· 0.039 Ω
1
r
π
·
β
g
m
·
100
0.039 Ω
1
· 2.56 kΩ
R
B
· R
B1
 R
B2
·8 kΩ4 kΩ· 2.67 kΩ
G. Kovacs ©1997 Page 123
R
in
· R
B
r
π
· 2.67 kΩ2.56 kΩ ·1.31 kΩ
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
A
V
·
v
o
v
s
·
2.67 kΩ 2.56 kΩ
4 kΩ+2.67 kΩ2.56 kΩ
¸
¸
_
,
−0.039 Ω
1
( ) 6 kΩ 4 kΩ ( )
A
V
· 0.247 ( ) −0.039 Ω
1
( ) 2.4 kΩ ( ) · −23.1
• Do the analysis again with "the details,"
R
in
· R
B
 r
x
+ r
π
( ) · 2.67 kΩ 50 Ω+ 2.56 kΩ ( ) · 1.32 kΩ
A
v
·
v
o
v
s
·
R
B
 r
x
+ r
π
( )
R
S
+ R
B
 r
x
+ r
π
( )
¸
¸
_
,
r
π
r
x
+ r
π
¸
¸
_
,
−g
m
( ) R
C
 R
L
 r
o
( )
·
1.32 kΩ
4 kΩ+ 1.32 kΩ
¸
¸
_
,
2.56 kΩ
50 Ω+2.56 kΩ
¸
¸
_
,
−0.037 ( ) 6 kΩ4 kΩ100 kΩ ( )
· 0.248 ( ) 0.981 ( ) −0.039 Ω
1
( )
2.34 kΩ ( ) · −22.2
A
v
= 0.248 0.981  0.039 2.34 KΩ = 22.2 V/V
• There is not much difference between the two gain values. The key
to this type of analysis is to know when you can approximate and
when you cannot.
G. Kovacs ©1997 Page 124
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
6.2 LOWFREQUENCY RESPONSE
r
r
x
π
v
be
+

g
m
v
be
r
o
R
E
C
E
R
S
R
B
R
C
R
L
v
o
C
C1 C
C2
v
s
• Given: C
C1
= C
C2
= 1 µF and C
E
= 10 µF (not infinity, so must use SC time
constants), use,
ω
L
≈
1
C
C1
R
C1S
+
1
C
E
R
ES
+
1
C
C2
R
C2
Where,
R
C1S
· R
S
+ R
B
 r
x
+ r
π
( ) [ ]
· 4 kΩ+ 2.67 kΩ 50 Ω+ 2.56 kΩ ( ) [ ] · 5.32 kΩ
R
ES
· R
E

r
x
+ r
π
+ R
B
R
S
( )
β +1
¸
1
]
1
R
ES
· 3.3 kΩ 
50 Ω+ 2.56 kΩ+ 2.67 kΩ 4 kΩ ( )
100 +1
¸
1
]
1
· 41.2 Ω
R
C2S
· R
L
+ R
C
 r
o
( ) · 4 kΩ + 6 kΩ100 kΩ ( ) · 9.66 kΩ
ω
L
≈
1
C
C1
R
C1S
+
1
C
E
R
ES
+
1
C
C2
R
C2
G. Kovacs ©1997 Page 125
ω
L
≈
1
1 X 10
6
5.32 K
+
1
10 X 10
6
41.2
+
1
1 X 10
6
9.66 K
What components
control the low cut
off frequency?
Can you "design" it?
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
ω
L
≈ 188.0 + 2,427 + 103.5 = 2,719 Radians > f
L
≈ 432.7 Hz
And, the frequency of the zero caused by the bypass capacitor on the emitter resistor
being finite in size is,
ω
ZE
=
1
C
E
R
E
=
1
10 X 10
6
3.3 KΩ
> f
ZE
= 4.8 Hz
(note that it is not R
E
' above, but R
E
)
6.3 HIGHFREQUENCY RESPONSE
• From here up in frequency, we will start to worry about the details of what’s inside the
BJT. Now those junction capacitances matter, so one needs to use the “full” Hybridπ
model.
g
m
v
be
BC
C
r
r
x
π
v
be
+

π
C
R  R
C
R
s
v
o
v
s
L
in
R
R
B
r
o
= C
µ
• Note (for homework) that you may need compute the required capacitances for the
model using,
2πf
t
=
1
C
π
g
m
+
C
BC
g
m
• In the present example, we are given the values,
C
π
= 13.9 pF C
µ
= C
BC
= 2 pF
• First, we calculate the Miller capacitance to obtain a value for the total input capacitance
(this gives the dominant highfrequency pole...)
C
M
= C
BC
1 + g
m
R
C
 R
L
 r
o
(note dependence on load resistance!)
C
M
= 2 X 10
12
1 + 0.039 6 KΩ  4 KΩ  100 KΩ = 184.8 pF
G. Kovacs ©1997 Page 126
• From this, we see that the total input capacitance is,
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
C
in
= C
π
+ C
M
= 198.7 pF
• The upper 3 dB frequency of the input circuit is given by,
ω
H
=
1
R
'
C
in
Where,
R
'
= r
π
 r
x
+ R
B
 R
s
= 2.56 KΩ  50 Ω + 2.67 KΩ  4 KΩ = 1.00 KΩ
• So, ω
H
=
1
1.00 K 199 X 10
12
= 5.025 X 10
6
Radians > f
H
= 800 KHz
• Remember that there is a scaled "copy" of C
BC
connected from the output to ground
due to the Miller Effect.... what would that time constant be?
C
out
· C
BC
1+ g
m
R
C
 R
L
 r
o
( )
g
m
R
C
 R
L
 r
o
( )
· 2 X 10
−12
( )
1 + 0.039 Ω
−1
( )
6 kΩ4 kΩ100 kΩ ( )
0.039 Ω
−1
( )
6 kΩ 4 kΩ100 kΩ ( )
C
out
· 2 X 10
−12
( )
1.01 ( ) · 2.02 X 10
−12
• Note that the value of C
out
, for reasonably small gains like those typical in singletransistor
amplifiers, is nearly the same as C
BC
.
g
m
v
be
out
C
R  R
C
v
o
L
r
o
≈ C = C
µ
BC
R
out
'
= r
o
 R
L
 R
C
= 100KΩ  4KΩ  6KΩ = 2.34 KΩ
ω
out
=
1
R
out
'
C
BC
=
1
2.34 KΩ 2 pF
= 214 X 10
6
Radians = 34 MHz
G. Kovacs ©1997 Page 127
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE
• So, in this case , it is pretty clear that the input time constant dominates at 800 kHz!
• One can always use the "full" approximation:
ω
H
≈
1
C
i
R
io
i
∑
¸
1
]
1
·
1
C
in
R
'
+C
out
R
out
·
1
C
in
r
π
 r
x
+R
B
 R
S
[ ] ( )
+C
out
r
o
 R
C
 R
L
( )
DO NOT ASSUME THAT YOU CAN IGNORE
THE OUTPUT TIME CONSTANT! (i.e. make sure
the time constant caused by the Miller copy of
C
BC
at the output is not a significant one!)
•A "full" analysis for the upper cutoff frequency could also be done using the opencircuit
time constants method, but would not provide additional information in this case due to
the relative simplicity of the circuit.
G. Kovacs ©1997 Page 128
Chapter 8: THE COMMON BASE AMPLIFIER
Chapter 8: THE COMMON BASE
AMPLIFIER
22% of fast food employees in the U.S.
admit to doing, “slow, sloppy work on purpose.”
Harper’s Index
1: OBJECTIVES
• To learn about:
Common base amplifier analysis.
Common base amplifier frequency response.
The basics of a twotransistor amplifier, the cascode amplifier, based on the
CE and CB stages in combination.
READ S&S Section 7.7
G. Kovacs ©1997 Page 129
Chapter 8: THE COMMON BASE AMPLIFIER
2: THE COMMON BASE CIRCUIT
VCC
VCC
R
B2
R
C
V
B
R
E
R
B1
v
s
v
o
R
S
RB2
C C1
C
C2
C
• Note that the base can either be tied directly to ground or (more common) it can be
biased using the “classic” scheme discussed for the commonemitter amplifier
with one difference > one of the bias resistors should be bypassed so that the parallel
combination of R
B1
and R
B2
has "zero" impedance for AC signals!
3: MIDBAND GAIN CALCULATIONS
• First, make a smallsignal equivalent of the circuit on the above right (more general):
G. Kovacs ©1997 Page 130
g
m
v
be
r
o
R  R
C L
R
s
r
x
r
π
v
be
+

R
E
v
s
Drive the
emitter!
E
B
C
Chapter 8: THE COMMON BASE AMPLIFIER
• Transform the base resistances to the emitter circuit by dividing by (β + 1),
R
E
g
m
v
be
r
o
R  R
C L
r
π
β + 1
r
x
β + 1
R
s
v
s
v
be
+

v
1
"B"
v
o
R
in
R
i
then compute v
be
as a function of v
s
, starting with an intermediate voltage, v
1
,
v
1
=
R
E

r
π
+ r
x
β + 1
R
E

r
π
+ r
x
β + 1
+ R
S
v
s
v
1
is a fraction of v
s
given by voltage division
v
be
= 
r
π
β + 1
r
π
β + 1
+
r
x
β + 1
v
1
= 
r
π
r
π
+ r
x
v
1
similarly, v
be
is a fraction of v
1
combining these gives,
v
be
= 
r
π
r
π
+ r
x
R
E

r
π
+ r
x
β + 1
R
E

r
π
+ r
x
β + 1
+ R
S
v
s
• If we assume that r
x
≈ 0 (not an unreasonable assumption, since typical values are
less than 50 Ω, but certainly r
x
is << r
π
),
v
be
= 
R
E

r
π
β + 1
R
E

r
π
β + 1
+ R
S
v
s
= 
R
E
 r
e
R
E
 r
e
+ R
S
v
s
G. Kovacs ©1997 Page 131
Chapter 8: THE COMMON BASE AMPLIFIER
• This further simplifies if R
s
= 0 (i.e. the CB amplifier is driven by a very low output
impedance stage)...
v
be
=  v
s
• Now for the output stage, assuming r
o
is "approximately grounded" or, simply very
large compared to R
C
and R
L
,
v
o
≈ −g
m
v
be
R
C
 R
L
 r
o
( ) ≈ −g
m
v
be
R
C
R
L
( )
which gives,
A
V
≡
v
o
v
s
·
R
E
 r
e
( )
R
E
 r
e
( ) + R
S
g
m
R
C
R
L
( )
and for R
E
>> r
e
, this simplifies to,
A
V
·
r
e
r
e
+ R
S
g
m
R
C
R
L
( )
• If r
o
were considered in parallel with R
C
and R
L
, it would clearly reduce gain.
• Again neglecting r
o
and r
x
, the input resistance is seen to be,
R
i
· R
S
+ R
E
r
e
≈ R
S
+ r
e
• NOTE that we typically do NOT include R
S
into our expression for R
IN
because it is not
a part of the commonbase amplifier itself!.... don't let this confuse you!
• Remember that,
r
e
·
r
π
1+β
≈
r
π
β
·
1
g
m
·
V
T
I
C
• This means you can control r
e
and hence the input resistance via I
C
.
G. Kovacs ©1997 Page 132
NONINVERTING!
Chapter 8: THE COMMON BASE AMPLIFIER
R
IN
· R
E
 r
e
≈ r
e
• To determine the output resistance, we set v
s
= 0 and remove R
L
(since it is not part of
the amplifier itself) and see that,
R
o
· R
C
• One can include r
o
in parallel with R
C
for added accuracy, if necessary... if r
o
is
comparable to R
C
this can be quite significant and reduce gain as well.
4: LOWFREQUENCY RESPONSE
VCC
R
B2
V
B
R
B1
E
R
v
s
R
S
R
C
R
L
r
π
v
be
+

g
m
v
be
RB2
C
C1
C
C2
C
G. Kovacs ©1997 Page 133
• Here the coupling and bypass capacitors matter!
r not shown here
o
Chapter 8: THE COMMON BASE AMPLIFIER
• Using the method of shortcircuit time constants (the one for low frequencies),
ω
L
≈
1
C
RB2
R
RB2
+
1
C
C1
R
C1
+
1
C
C2
R
C2
where,
R
RB2
· R
B2
R
B1
 r
π
+ β +1 ( ) R
E
 R
S
( ) [ ]
R
C1
· R
S
+ R
E
r
e
R
C2
· R
C
+ R
L
• Hopefully it is clear that if you literally ground the base, you don’t need to consider
R
RB2
...
• NOTE: If you take into account r
o
in parallel with the current source, you get a new
time constant for C
C2
,
R
C2
· R
C
 r
o
+ R
E
R
S
 r
e
( ) + R
L
but since r
o
is generally >> r
e
, this doesn't change things much.
5: HIGHFREQUENCY RESPONSE
g
m
v
be
BC
C
R
s
π
C
r
π
v
be
+

r
x
E
R
R
C
 R
L
r
o
v
s
• Assuming r
x
= 0, it is immediately clear the one terminal of C
BC
is grounded... This
means there is no Miller multiplication of its capacitance!
G. Kovacs ©1997 Page 134
• The circuit is more intuitive if it is redrawn knowing this (r
x
no longer shown),
Chapter 8: THE COMMON BASE AMPLIFIER
E
R
R
C
 R
L
g
m
v
be
r
o
r
π
v
be
+

π
C
BC
C
R
s
v
s
• We can see that the input circuit (after transforming r
π
into r
e
in the emitter circuit) looks
like this:
R
s
v
s
r
e
π
C
E
R
• NOTE: We don't scale C
π
by (ß + 1) because the extra current flowing due to the
current source is already taken into account in transforming r
π
into r
e
(if we scaled C
π
as
well, the resulting pole would be "corrected" twice and therefore be wrong!).
• In other words, the extra emitter current that we model in r
e
is "real" as opposed to the
capacitor current, which is "imaginary." Therefore, the extra current is fully modeled by
r
e
without modifying C
π
.
• Still another way of looking at it is to consider that the capacitor current is determined
by the voltage across it, v
be
, which is the same, no matter which way you look (base or
emitter). For r
π
and r
e
, the current is a function of which terminal you look into because
at the emitter, you see the extra current from the source.
• Now we can determine the input pole frequency by setting the signal source to
zero...
ω
P1
·
1
C
π
r
e
 R
E
R
S
( )
G. Kovacs ©1997 Page 135
Chapter 8: THE COMMON BASE AMPLIFIER
• Similarly, the output stage’s pole frequency can be determined (neglecting r
o
),
R
C
 R
L
g
m
v
be
BC
C
ω
P2
·
1
C
BC
R
C
R
L
( )
• Since r
e
is small, ω
P1
will be a very high frequency. Since C
BC
is small, ω
P2
will also be
a very high frequency.
• The high frequency response is (neglecting r
o
),
ω
H
≈
1
C
i
R
io
i
∑
¸
1
]
1
·
1
C
π
r
e
 R
E
 R
S
( ) +C
BC
R
C
 R
L
( )
• For the CE amplifier we had derived (neglecting the output pole at C
BC
(r
o
R
C
R
L
)),
ω
H
≈
1
R
'
C
in
·
1
r
π
 r
x
+ R
B
 R
S
( ) C
π
+ C
BC
1+ g
m
R
C
 R
L
 r
o
( )
[ ] ( )
assuming r
x
is very small and r
o
is very large, this simplifies to,
ω
H
=
1
R
'
C
in
=
1
r
π
 R
B
 R
s
C
π
+ C
BC
+ C
BC
g
m
R
C
 R
L
G. Kovacs ©1997 Page 136
r not shown here
o
Chapter 8: THE COMMON BASE AMPLIFIER
• Take our previous example and compare...
R
S
= 4 KΩ R
B1
= 8 KΩ R
B2
= 4 KΩ R
E
= 3.3 KΩ R
C
= 6 KΩ R
L
= 4 KΩ
V
CC
= 12 V I
E
≈ 1 mA β
o
= 100 C
π
= 13.9 pF C
µ
= 2 pF r
o
= 100 KΩ
r
x
= 50 Ω
g
m
=
I
C
V
T
≈
1 mA
25.9 mV
= 0.039 Ω
1
r
π
=
β
g
m
=
100
0.039 Ω
1
= 2.56 KΩ
r
e
=
1
g
m
= 25.6 Ω
f
HCE
= 800 KHz
f
HCB
= 30.9 MHz
• Two comments from this:
1) The CB amplifier, despite its very low input impedance (≈ r
e
), has very
good highfrequency response.
2) The secondorder transistor model effects (such as r
x
) may become
significant at high frequencies for the CB configuration.
G. Kovacs ©1997 Page 137
Chapter 8: THE COMMON BASE AMPLIFIER
6: CASCODE AMPLIFIERS = CE + CB
• This is a handy type of amplifier that is a CE stage driving a CB stage...
VCC
v
s
E
R
v
o
C2
C R
C
R
B2
V
B2
R
B1
R
B3
B
C
R
S
C1
C
R
L
V
B1
E
C
• The biasing for both stages is set up using R
B1
, R
B2
, and R
B3
as a voltage divider, and
R
E
sets the total quiescent current in the output stage.
• In a nutshell, the CE stage's "R
C
" is the input of the CB stage (≈ r
e
) so the CE stage
has a gain of about one and no real Miller multiplication of C
µ
... the gain comes from
the CB stage.
• It is a fast amplifier.
• The details are presented in a separate section below.
G. Kovacs ©1997 Page 138
Chapter 9: THE COMMON COLLECTOR AMPLIFIER
Chapter 9: THE COMMON
COLLECTOR AMPLIFIER
Although we modern persons tend to take our electric lights, radios,
mixers, etc., for granted, hundreds of years ago, people did not have
any of them, which is just as well because there was no place to plug
them in.
Dave Barry
1: OBJECTIVES
• To learn about:
• The common collector or emitter follower BJT amplifier configuration (the last
one with only one BJT in it!) in terms of “what is it good for...”
• The basic properties of the CC amplifier.
• The frequency response of the CC amplifier
READ S&S Pages 259  265
and S & S Section 7.8
G. Kovacs ©1997 Page 139
Chapter 9: THE COMMON COLLECTOR AMPLIFIER
2: COMMON COLLECTOR CIRCUIT
VCC
v
s
v
o
R
E
r
π
g
m
v
be
i
in
v
be
v
o
v
s
R
E
Bias resistors, coupling caps not shown.
• We know that,
v
s
· i
in
r
π
+ i
in
+ g
m
v
be
( )R
E
· i
in
r
π
+ i
in
+ g
m
r
π
i
in
( )R
E
· i
in
r
π
+ i
in
+βi
in
( )R
E
v
s
· i
in
r
π
+ β+ 1 ( )R
E
[ ]
• The input current is thus,
i
in
·
v
s
r
π
+ β+ 1 ( )R
E
• Note that the extra current supplied by the g
m
generator in addition to i
in
makes R
E
look
like a much larger resistor in series with r
π
.
• The current flowing in the collector circuit is
g
m
v
be
= β
F
i
b
• Therefore, the output voltage is given by,
v
o
· β +1 ( )i
in
R
E
·
β +1 ( )R
E
r
π
+ β+ 1 ( )R
E
v
s
G. Kovacs ©1997 Page 140
Chapter 9: THE COMMON COLLECTOR AMPLIFIER
which gives a voltage gain of,
A
V
·
v
o
v
s
·
β +1 ( )R
E
r
π
+ β +1 ( )R
E
(in a voltage divider form)
• BUT since (β
F
+ 1)R
E
>> r
π
> A
v
≈ 1
• You can also rearrange the gain equation into,
A
V
·
R
E
r
π
β +1 ( )
+ R
E
·
R
E
r
e
+ R
E
which may be a more intuitive way of looking at the "voltage divider."
• The circuit is a “voltage follower” or “emitter follower” (i.e. the emitter follows the
base voltage) and is similar to the unitygain opamp circuit...
• Note that the output voltage “follows” the input voltage but is always one v
be
drop lower!!! This only matters in DCcoupled applications, that are generally not
of importance here....
• It is can be seen from
i
in
·
v
s
r
π
+ β+ 1 ( )R
E
that
R
in
· r
π
+ β +1 ( )R
E
• This is typically in the > 100 KΩ range, which is a reasonably high input impedance.
• Note that in Sedra and Smith (pages 259  265), the series resistance of the signal
source, R
S
, a parallel input resistance from base to ground, R
B
, a load resistance R
L
,
and the output resistance of the transistor, r
o
, are also taken into account in the derivations
as shown in Fig. 4.46.
G. Kovacs ©1997 Page 141
Chapter 9: THE COMMON COLLECTOR AMPLIFIER
3: PRACTICAL CC CIRCUIT
CC
v
s
R
B2
V
B
V
R
E
R
B1
L
R
Assume all capacitors > 0 impedance
for AC signals.
R
S
r
π
g
m
v
be
v
be
v
o
R
E
+

r R
o
L
B
S
R
R
v
s
R
o
i
R
R
ie
• For the “full” circuit, where R
B
= R
B1
 R
B2
, one obtains an input impedance of,
R
i
· R
B
 β +1 ( ) r
e
+ R
E
r
o
 R
L
( )
[ ]
· R
B
 r
π
+ β + 1 ( ) R
E
 r
o
 R
L
( )
[ ]
G. Kovacs ©1997 Page 142
Chapter 9: THE COMMON COLLECTOR AMPLIFIER
• The latter form can be found by inspection...
• The output resistance can be determined by replacing R
L
with a test voltage source v
x
and using,
R
O
≡
v
x
i
x
• The resistance looking into the emitter, R
ie
, can be computed using,
R
ie
≡
v
x
i
e
i
e
=  i
b
 βi
b
=  1 + β i
b
i
b
= 
v
x
r
π
+ R
S
 R
B
• So R
ie
is,
R
ie
≡
v
x
i
e
=
r
π
+ R
S
 R
B
β + 1
which gives R
o
(by placing R
ie
in parallel with R
E
and r
o
),
R
o
≡
v
x
i
x
· R
E
 r
o

r
π
+ R
S
R
B
( )
β +1 ( )
• The voltage gain for the “full” circuit can be shown to be,
A
V
≡
v
o
v
s
·
R
i
R
i
+R
S
¸
¸
_
,
R
E
 r
o
 R
L
r
e
+ R
E
 r
o
R
L
[ ]
¸
¸
_
,
·
R
i
R
i
+ R
S
¸
¸
_
,
β +1 [ ] R
E
 r
o
 R
L
[ ]
r
π
+ β +1 [ ] R
E
r
o
 R
L
[ ]
¸
¸
_
,
• This includes a first term which takes into account the voltage division at the input and
a second term where we have modified the original gain equation to take into account
the fact that r
o
and R
L
are in parallel with R
E
in the smallsignal equivalent circuit.
G. Kovacs ©1997 Page 143
Chapter 9: THE COMMON COLLECTOR AMPLIFIER
• Again, if R
S
is small, A
V
≈ 1.... (typically what we want when we use a CC stage)
• The current gain is quite high, and for R
L
<< (R
E
 r
o
) (i.e. R
L
is like a short circuit),
the current gain is approximately,
A
i
≡
i
o
i
i
·
v
o
R
L
v
s
R
S
+ R
i
( )
≈
R
i
R
L
≈
r
π
+ β +1 ( )R
L
R
L
≈ β +1 ( )
• Note that the r
π
/R
L
term may be significant if it is large relative to (ß+1), but this is
generally not the case.
• Thus, the CC amplifier can be used as a current booster or voltage follower.
4: THE PHASE SPLITTER CIRCUIT
VCC
INVERTED
OUTPUT
NON
INVERTED
OUTPUT
• This circuit gives inverted and noninverted outputs from the same
G. Kovacs ©1997 Page 144
input signal... It’s not really CE or CC.... Think about its operation...
Chapter 9: THE COMMON COLLECTOR AMPLIFIER
5: QUICK LOOK AT CC FREQUENCY RESPONSE
• The details of the derivation are presented in S & S Section 7.8, but an overview is
useful here....
S
R
v
s
r
π
v
be
+

R
B
C
π
C
µ
g
m
v
be
r
o
v
o
R
E
L
R
• Based on a “full” analysis (See S & S), the dominant pole frequency is given by,
ω
p
·
1
R
S
+ r
x
( )  1+ g
m
R
E
R
L
[ ] ( )r
π [ ]
C
µ
+
C
π
1+ g
m
R
E
 R
L
[ ]
¸
¸
_
,
(NOTE THAT r
x
is not shown in the above AC equivalent circuit and, in fact, can
typically be neglected since it is usually on the order of 10Ω... also note that R
B
is considered to be very large here...)
• Practically speaking, the input pole can usually be approximated by,
ω
p
≈
1
R
S
+ r
x
( )C
µ
• An alternative is to approximate the high frequency response using the method of
G. Kovacs ©1997 Page 145
opencircuit time constants....
Chapter 9: THE COMMON COLLECTOR AMPLIFIER
6: DESIGN OF CC STAGES
• The design process is similar to that for the other singleBJT amplifiers!
1) Assuming you need a basic follower, you don’t worry about gain (unless you need a
particular current gain, and then you choose a transistor with the right β).
2) You choose an I
C
value that you want to operate at and calculate g
m
, etc.
3) Bias stages are done the same way as before.
TYPICAL CC CIRCUIT:
CC
v
s
R
B2
V
B
V
R
E
R
B1
L
R
G. Kovacs ©1997 Page 146
Chapter 10: CASCADED AND CASCODE AMPLIFIERS
Chapter 10: CASCADED AND
CASCODE AMPLIFIERS
Gee Toto, I guess we're not in Terman anymore!
Stanford EE at first job interview in Silicon Valley...
1: OBJECTIVES
• To learn about:
What happens when you put a bunch of amplifiers in series.... in other
words, look at interstage loading effects (affecting gain and frequency response).
To study a very useful multitransistor amplifier: the cascode configuration.
This will combine what we know about common emitter and common base
amplifiers.
READ S&S Sections 7.7, 7.8 and 6.1  6.3
G. Kovacs ©1997 Page 147
NOTE: we will use S & S notation for C
BC
= C
µ
here.
Chapter 10: CASCADED AND CASCODE AMPLIFIERS
2: AMPLIFIERS IN SERIES (CASCADED)
Vo
V
in1
V
in2
R
i1
R
i2
R
o1
R
o2
R
L
AV
in1
AV
in2
Amp #1
Amp #2
+

+

• Now it is necessary to consider what happens when amplifiers are put in series.
Looking at the above example, it is clear that the input and output resistances (or
impedances!) come into play by reducing the overall gain.
• If the amplifiers were ideal (R
out
= 0 and R
in
= ∞), and each had a gain of A, the
overall gain would simply be A
2
!
• In the above example, let's work out the gain assuming nothing about the R
in
and R
out
of each stage, looking at them as voltage dividers between each stage and between
the last stage and the load.
• Note that in practice, impedances, Zi, would normally be used, not resistances, but
they serve to illustrate the point here.
v
in2
· Av
in1
R
i2
R
i2
+ R
o1
¸
¸
_
,
< Losses between stages and first amplification
v
o
· Av
in2
R
L
R
L
+ R
o2
¸
¸
_
,
< Losses Due To R
o2
and second amplification
v
o
v
in
· A
2
R
i2
R
i2
+ R
o1
¸
¸
_
,
R
L
R
L
+ R
o2
¸
¸
_
,
< Overall eqn. assuming equal gains (A)!
G. Kovacs ©1997 Page 148
Chapter 10: CASCADED AND CASCODE AMPLIFIERS
• As expected, the equation reduces to the ideal case of A
V
= A
2
for two identical stages
if we let the R
o
's go to 0 and the R
i
's go to infinity.
• The above equations assume that the individual amplifier gains ("A") do not change
with output loading.
• For most opamps since R
in
= MΩ to GΩ range, and R
o
≈ 50  100 Ω, the gains are
pretty close to being A
N
(where N = number of equalgain stages).
• Just to check that, assume a "notsohot" opamp with R
o
= 100Ω and R
in
= 1MΩ, what
is the gain with two stages of gain A in series? (assume R
L
= 1 MΩ too)
v
o
v
in
= A
2
1MΩ
1MΩ+100Ω
1MΩ
1MΩ+100Ω
= A
2
(0.9999) (0.9999) = 0.9998 A
2
• That is pretty close to A
2
!
• In fact, you would have to go to a ONE HUNDRED stages with these specifications
before you even lost 1% of the expected "ideal" gain (i.e. to get 0.99 A
100
)...
• By the time you reached that point, other effects would have caused much more
trouble (for example, the fact that noise from each successive stage is added to the
noise coming into that stage and amplified... on down the line!).
• There are practical reasons why you just can't continue cascading stages "forever..."
If DCcoupled, realworld offsets can be impossible to trim out!
Even if ACcoupled, noise from preceding stages gets amplified by each
downstream amplifier stage, making for nothing but a noise source after a while!
• We normally refer all noise to the input of an amplifier, taking out the effects of the
G. Kovacs ©1997 Page 149
gain stages.
Chapter 10: CASCADED AND CASCODE AMPLIFIERS
3: THE CASCODE AMPLIFIER
• Now... how do we make a fast amplifier (i.e. for oscilloscopes, etc.)?
• The idea: Combine the advantages of common emitter (high R
in
) with common base
(no Miller effect) to get greatly improved performance.
• Remember that the CE amplifier suffers the most from the rolloff caused by the input
pole.
Vo
+

R
C
R
L

g
m
V
be
Cπ V
be
r
π
Cµ
BECOMES ...
V
+

V
be
r
π
Cπ C
M
g
m
V
be
R
C
R
L

≈Cµ
o
where the Miller Capacitance is C
M
= C
µ
1+g
m
R
C
R
L
(Note that C
µ
is just another notation for C
BC
.)
• The cascode amplifier is a CE amplifier driving a CB stage.
• The CE stage, where the Miller Effect could be a problem, is deliberately set up for a
low gain to minimize it!
• The gain is obtained from the CB output stage, not the CE stage.
G. Kovacs ©1997 Page 150
Chapter 10: CASCADED AND CASCODE AMPLIFIERS
R
R
No Miller Here!
Miller Effect Here
Low Input R
for CB Stage
in2
C
R
L

v
in
• Let's look at the Miller capacitance at the CE stage... First figure out what the "R
C
" of
the CE stage is (the input of the CB stage!).
• Remember the input resistance of a commonbase amplifier is,
R
in2
≈ r
e
≈
1
g
m2
• Now we can calculate the CE gain and substitute into the Miller equation (remember
that we multiply the capacitance by [1  A
V
], where A
V
is the voltage gain),
C
M
· C
µ
1+ g
m
R
C
 R
L
( ) ≈ C
µ1
1+ g
m1
1
g
m2
¸
¸
_
,
(Note that r
o
of the transistors is neglected here, and would appear in parallel with R
C
and R
L
if considered.)
• If g
m1
= g
m2
, then,
C
M
≈ C
µ1
1+1 ( ) · 2C
µ1
• Thus the Miller Effect is minimized in this configuration and the CE gain is ≈ 1.
G. Kovacs ©1997 Page 151
Chapter 10: CASCADED AND CASCODE AMPLIFIERS
4: PRACTICAL CASCODE AMPLIFIER CIRCUIT
V
C
B
Q
2
Q
1
R
C
R
3
R
2
R
1
R
E
R
S
C
C2
R
L
v
s C
E
C
C1
v
o
CC
• One can use a "1/4, 1/4, 1/4, 1/4" biasing scheme here, where V
B1
is 1/4 V
cc
, V
B2
is 1/2 V
cc
and so on.
• R
E
sets the current through BOTH Q
1
and Q
2
. In fact, if the ß values are relatively
large, you can assume that the collector currents are equal...
I
E1
·
β +1
β
I
C1
≈ I
C1
I
E1
≈ I
C1
· I
E2
≈ I
C2
• R
1
, R
2
, and R
3
set the bias points, and R
E
can be used to adjust the current
through both transistors.
G. Kovacs ©1997 Page 152
Chapter 10: CASCADED AND CASCODE AMPLIFIERS
The AC Equivalent Circuit (High Frequency)
g
m
V
be2
Cµ2
Cπ2 r
π2
g
m
V
be1
r
o1
Cµ1
Cπ1 r
π1
R
2
R
1
R
s
v
s
r
x1 B1
E1
E2
C2 B2
C1
r
o2
R
C
R
L

v
o
NOTE: grounded here...
CB has no Miller!
NOTE: assuming r
is zero here...
X2
4.1 THE INPUT SECTION
• Assuming r
x1
= 0 (See S&S p. 534)...
• As is usually the case, the input resistances just act as a voltage divider...
v
be1
· v
s
r
π1
R
1
 R
2
R
S
+ r
π1
R
1
 R
2
¸
¸
_
,
R
s
R
1
R
2
r
π1
V
s
4.2 THE COMMON EMITTER STAGE
• The CE amplifier output is looking at r
e2
.
C
M
= C
µ1
1+g
m1
r
e2
= 2 C
µ1
if r
e2
≈
1
g
m2
=
1
g
m1
and r
o1
=∞
G. Kovacs ©1997 Page 153
Chapter 10: CASCADED AND CASCODE AMPLIFIERS
V
be1
r
π1
C
π1
C
M
r
o1
r
e2
g
m
V
be1
+

2C
µ1
• We know that the gain for a nonemitterdegenerated CE stage (note that R
E
is
bypassed by C
E
!) is,
A
V
=  g
m1
R
C
 R
L
and that here, R
C
 R
L
is simply r
e2
... the CE GAIN ≈ 1 in this case.
INPUT TIME CONSTANT FOR THE CE STAGE
R
s
R
1
R
2
r
π1 C
π1
C
M
r
x
C
M
= 2C
µ1
ω
1
·
1
R
S
 R
1
 R
2
+ r
x1
( )  r
π1 [ ]
C
π1
+2C
µ1
( )
≈
1
R
S
 R
1
 R
2
 r
π1
( ) C
π1
+ 2C
µ1
( )
• Note that the latter approximation assumes r
x
≈ 0, but if R
s
is very small, r
x
may be
significant (otherwise, the input pole theoretically goes to infinite frequency).
G. Kovacs ©1997 Page 154
Chapter 10: CASCADED AND CASCODE AMPLIFIERS
4.3 THE COMMON BASE STAGE
• Now look at the CB Stage to get the overall gain...
• Since the base is grounded, the nongrounded half of C
π2
must look at r
π2
"through the
emitter."
g
m
V
be1
r
π2
R
C
R
L

v
Cπ2
g
m
V
be2
C
µ2
r
o2
v
x
o r
e2
E2
B2
C1
Note that we use r
because we are looking
into the emitter of Q2.
e2
2C
µ1
• It is known that since the gain of the CE stage is ≈ 1, V
x
≈ V
s
.
• We know that the gain of CB amplifier is,
A
VCB
·
v
o
v
x
· +g
m2
r
o2
 R
C
R
L
( ) ≈ g
m2
R
C
 R
L
( )
• We can thus write the complete midband gain (neglecting r
x
and r
o
) as:
A
V
≡
v
o
v
s
·
r
π1
 R
1
R
2
R
S
+ r
π1
 R
1
R
2
¸
¸
_
,
−1 ( )g
m2
R
C
R
L
( )
G. Kovacs ©1997 Page 155
Input
Divider
CE stage
CB stage
Chapter 10: CASCADED AND CASCODE AMPLIFIERS
4.4 THE OTHER TIME CONSTANTS
• So far, we only found one time constant (the input) so, let's find the others, in between
stages and at the output...
• In between the stages...
g
m
V
be1
r
π2
R
C
R
L

v
Cπ2
g
m
V
be2
C
µ2
r
o2
v
x
o
r
e2
2C
µ1
Emitter of Q2
ω
2
=
1
C
π2
r
e2
• NOTE: Use r
e
not r
π
because we are looking into Q
2
's emitter!!!
• Note that we could also consider two lessimportant components in between stages,
r
o1
and C
µ1
(C
µ1
appears at the output because of the Miller Effect)...
g
m
V
be1
r
π2
R
C
R
L

v
Cπ2
g
m
V
be2
C
µ2
r
o2
o
r
o1
2C
µ1
v
x
r
e2
C1 E2
• You will note that sometimes these other two components are considered, other times
they are not (particularly in the text book!). Since it is relatively easy to consider them,
you can include them OR show that they don't change things much (i.e. C
µ1
<< C
π2
and
r
e2
<< r
o1
). The "complete" equation would be,
ω
2
·
1
C
π2
+ 2C
µ1
( )
r
e2
 r
o1
( )
G. Kovacs ©1997 Page 156
Chapter 10: CASCADED AND CASCODE AMPLIFIERS
• r
o1
is typically ≈ 100 KΩ and r
e2
is typically a few Ohms, so r
e2
typically determines the
net resistance. You may have to look at C
π2
and C
µ1
to decide...
• This interstage time constant typically corresponds to a very high frequency. Its value
doesn’t depend on R
s
or the Load (R
L
).
• The output stage time constant...
g
m
V
be1
r
π2
R
C
R
L

v
Cπ2
g
m
V
be2
C
µ2
r
o2
o
r
o1
2C
µ1
v
x
• Note that r
o2
is shown with one end grounded to simplify the analysis, but the circuit
actually would be something like (C
µ1
and C
π2
are shown as opencircuits as well as the
current source),
r
π2
R
C
R
L

v
Cπ2
g
m
V
be2
C
µ2
o
r
o1
2C
µ1
v
x
r
o2
r
e
to ground
• Effectively, you have r
o2
connected to a very small resistance (r
e
) to ground, so we can
just look at it as having one end grounded as shown above (i.e. r
o2
+ r
e
≈ r
o2
).
ω
3
=
1
C
µ2
r
o2
R
C
R
L
≈
1
C
µ2
R
C
R
L
G. Kovacs ©1997 Page 157
r
e2
r
e2
Chapter 10: CASCADED AND CASCODE AMPLIFIERS
• Now recall the input pole...
ω
1
=
1
R
S
R
1
R
2
r
π2
C
π1
+2C
µ1
• If R
s
is large, this frequency is lower than ω
2
and ω
3
...
• So, like the CE stages alone, the input pole typically (but not always!) determines
f
H
, but here there is only a 2X Miller multiplication!
G. Kovacs ©1997 Page 158
Chapter 10: CASCADED AND CASCODE AMPLIFIERS
5: QUICK CASCODE EXAMPLE
(S&S EX 7.20, p. 536)
VCC
18K
6K
3.3K
4K
8K 10µF
4K
1µF
10µF
1µF
4K
Q
1
Q
2
R
3
R
2
R
1
R
E
C
E
R
L
R
S
v
s
C
B
C
C1
C
C2
• Given : V
cc
= +15 V, I
E
≈ 1mA, Q
1
= Q
2
(identical devices), ß = 100
• Doing the analysis is not a big problem ...
r
π
=
ß
g
m
= 2.56 KΩ
g
m
≈
I
C
V
T
=
1mA
25.9
= 0.039 Ω
1
A
M
= 
R
π
R
1
R
2
R
S
+ R
π
R
1
R
2
g
m
R
C
R
L
= 23.1
f
1
=
ω
1
2π
=
1
2π
1
R
S
R
1
R
2
r
π2
C
π1
+2C
µ1
= 8.75 MHz
f
2
=
ω
2
2π
=
1
2π
1
C
π2
r
e2
=
1
2π
1
13.9 pF 25.6Ω
= 447 MHz
G. Kovacs ©1997 Page 159
Chapter 10: CASCADED AND CASCODE AMPLIFIERS
f
3
=
ω
3
2π
=
1
2πC
µ2
R
C
R
L
=
1
2π 2 pF 4K  6K
= 33.2 MHz < neglect r
o
• Clearly, f
1
is the lowest!
• Now, what if we let R
s
go to 50Ω (typical for signal generators)...
f
1
goes way up (around 700 MHz) and the output pole (f
3
) dominates in this case!
• The message here is DON'T ASSUME you know which pole is dominant !!!!
f
H
=
1
1
f
1
2
+
1
f
2
2
+
1
f
3
2
= 8.46 MHz
• Compare this to the previous CE example!
A
V
= 23.1
f
H
= 800 KHz
G. Kovacs ©1997 Page 160
Chapter 10: CASCADED AND CASCODE AMPLIFIERS
6: CASCODE AMPLIFIER DESIGN EXAMPLE
• This example is analogous to the CommonEmitter design done previously in the
notes. The purpose is for you to compare the two designs for differences in characteristics.
Specifications:
DC power dissipation: P
D
< 25 mW
Power Supply: 12 VDC
Voltage Gain: 50X
Load: Resistive, 50 KΩ
Assume R
S
= 0 Ω
Must use 2N2222A Transistors (NPN, ß = 150 measured)
VCC
C
B
Q
2
Q
1
R
C
R
3
R
2
R
1
R
E
R
S
C
C2
R
L
V
S C
E
C
C1
V
o
All capacitors are large
(100 µF) so they are
essentially shortcircuits
for AC
Key point: The collectoremitter current running through Q
2
is assumed to be
equal to that through Q
1
. Further, the total power dissipation (DC, or quiescent)
and the voltage gain are both determined by the choice of this current.
DESIGN PROCESS:
1) Pick V
cc
unless specified.
V
cc
= 12 VDC
G. Kovacs ©1997 Page 161
Chapter 10: CASCADED AND CASCODE AMPLIFIERS
2) Calculate I
max
.
I
MAX
=
Max DC Power
Vcc
=
25 mW
12 V
= 2.083 mA
3) Select an I
C
< I
MAX
and solve for g
m2
.
Let I
C
= 1.8 mA
g
m2
=
1.8 mA
0.0259 V
= 0.0695 Ω
1
4) Let R
S
= 0 or be too small to matter (for now) and solve for R
C
:
A =  g
m2
R
C
R
L
r
π
R
1
R
2
R
S
+ r
π
R
1
R
2
R
C
=
g
m2
A

1
R
L
1
R
C
= 730 Ω
5) Use a 1/4, 1/4, 1/4, 1/4 biasing rule to set up bias resistors. Let V
CE
of the
transistors = V
cc
/4 = 3V and solve the following for R
E
.
R
E
=
V
B1
 V
BE
I
E
=
V
CC
4
 0.7
I
C
ß + 1
ß
=
3  0.7
0.001812
= 1.27 kΩ
6) The required base current is:
I
B
=
I
C
ß
=
0.0018
150
= 0.012 mA
7) Solve for the biasing resistors.
I
BIAS
> 0.1I
E
≈ 0.1I
C
R
BNET
· R
1
+ R
2
+ R
3
·
V
CC
0.1I
C
·
12 V
0.1 ( ) 1.8 mA ( )
· 66.7kΩ
V
B1
=
V
CC
4
= V
CC
R
1
R
1
+ R
2
+ R
3
R
1
·
R
BNET
4
·16.7 kΩ
G. Kovacs ©1997 Page 162
Chapter 10: CASCADED AND CASCODE AMPLIFIERS
V
B2
·
V
CC
2
∴ R
2
· R
1
∴ R
3
· 2R
1
· 2R
2
·
R
BNET
2
· 33.3 kΩ
8) Now check that any R
S
that may be present should not have a major impact on
the gain....
 If little or no impact, move on.
 If much impact, tweak I
C
and iterate if necessary.
9) SIMULATE THE CIRCUIT ON SPICE USING THE FOLLOWING VALUES:
R
1
= 15 KΩ R
2
= 15 KΩ R
3
= 30 KΩ
R
C
= 750 Ω R
E
= 1.3 KΩ
EE113 Demo  Cascode Amplifier
*Resistors
R1 VB1 0 15K
R2 VB2 VB1 15K
R3 VCC VB2 30K
RC VCC VC2 750
RE VE1 0 1.3K
RS Vin VX 1
RLL Vout 0 50K
*Capacitors
C1 VX VB1 100UF
C2 VB2 0 100UF
C5 VE1 0 100UF
C6 Vout VC2 100UF
*Sources
Vcc VCC 0 12
Vss 0 Vin AC 10mV
*Transistors
Q1 V01 VB1 VE1 TRANSMODEL
Q2 VC2 VB2 V01 TRANSMODEL
.MODEL TRANSMODEL NPN (BF=150 IS=1.3E14
+ TF=.9N CJE=6P CJC=5P)
*input frequency sweep
.AC DEC 10 1 100MEG
.PROBE
.end
G. Kovacs ©1997 Page 163
Chapter 10: CASCADED AND CASCODE AMPLIFIERS
1 . 0 h 1 0 0 h 1 0 Kh 1 . 0 Mh 1 0 0 Mh
Fr e q u e n c y
V( Vo u t ) / V( Vi n )
5 0
0
C1 =
C2 =
d i f =
1 0 . 0 0 0 K, 4 6 . 2 8 4
9 . 9 9 9 K, 4 5 . 5 9 7
1 . 0 0 0 0 , 6 8 7 . 5 0 6 m
9 . 9 9 9 K, 4 5 . 5 9 7
C1 =
C2 =
d i f =
1 0 . 0 0 0 K, 4 6 . 2 8 4
9 . 9 9 9 K, 4 5 . 5 9 7
1 . 0 0 0 0 , 6 8 7 . 5 0 6 m
10) Adjust R
E
to modify the CollectorEmitter current by multiplying by the ratio of
simulated to desired gain.
The gain was (46.284 / 50 =) 0.9257 of the desired gain of 50. Let
R
E
= (0.9257)(1.3K) = 1.2K
1 . 0 h 1 0 0 h 1 0 Kh 1 . 0 Mh 1 0 0 Mh
Fr e q u e n c y
V( Vo u t ) / V( Vi n )
5 0
0
C1 =
C2 =
d i f =
1 0 . 4 4 4 K, 4 9 . 7 6 8
1 0 . 4 4 3 K, 4 9 . 0 4 6
1 . 0 0 0 0 , 7 2 2 . 3 9 1 m
1 0 . 4 4 3 K, 4 9 . 0 4 6
C1 =
C2 =
d i f =
1 0 . 4 4 4 K, 4 9 . 7 6 8
1 0 . 4 4 3 K, 4 9 . 0 4 6
1 . 0 0 0 0 , 7 2 2 . 3 9 1 m
The gain is now 49.768, and should be acceptable!
From SPICE Output Deck:
TOTAL POWER DISSIPATION 2.40E02 WATTS
G. Kovacs ©1997 Page 164
Everything looks like it's in order!
Chapter 11: DIFFERENTIAL AMPLIFIERS
Chapter 11: DIFFERENTIAL
AMPLIFIERS
Ow! That’s hot!
EE122 student’s first encounter with a soldering iron...
1: OBJECTIVES
• To consider:
The basic idea of the differential pair.
The benefits of the differential pair (no coupling caps > DC response, etc.).
DC operation of the differential pair.
The basic smallsignal equivalent circuit for the differential pair.
Commonmode operation of the differential pair.
The frequency response of the differential pair.
Nonideal properties of the differential pair.
READ S&S Sections 6.1  6.3,
7.10 and 7.11
G. Kovacs ©1997 Page 165
Chapter 11: DIFFERENTIAL AMPLIFIERS
2: DIFFERENTIAL AMPLIFIER BASIC CONCEPTS
+V
cc
v
o1
v
o2
v
B1
v
B2
Q
1
Q
2
R
C1 R
C2
+V
cc
V
EE
I
E
I
E
I
E
2
v
1
v
2
• The differential amplifier circuit (as shown above, it is often referred to as a differential
pair) is the basis for any operational amplifier (you can find one on the input of nearly
any opamp type chip). The two transistors are ideally identical and are generally
assumed to be below.
• There are two inputs, v
1
and v
2
, and one can apply inputs to both (differentially) or
ground one and use the amplifier in a singleended (ground referenced) way.
• This circuit can compute the difference between two input signals.
• While the opamp symbol ("triangle") has only one output, the differential pair shown
above has two (v
o1
and v
o2
). One can take outputs from both (differentially) or use just
one relative to ground (as you would with a typical opamp).
• The differential pair is basically two common emitter amplifiers sharing a common
G. Kovacs ©1997 Page 166
current source that sets the total DC bias current through both transistors at 2I
E
. In
+

Chapter 11: DIFFERENTIAL AMPLIFIERS
seeking to understand its operation, we will begin by looking at the shifting of this DC
current between the two transistors when there is a largesignal imbalance between
their two inputs (i.e. the difference between v
1
and v
2
is large).
• With NO differential signal input (v
1
and v
2
are the same), the current flowing out the
emitter of each transistor is I
E
and therefore, v
o1
= v
o2
in this situation (if R
C1
= R
C2
, which
is usually the case).
• For the smallsignal case, if the collector currents are the same in both transistors (I
E
≈
I
C
), their g
m
's and other parameters will be equal (both transistors are as close to
identical as possible).
• As will be seen when the circuit is studied more closely, its symmetry is what gives rise
to its "special" properties.
3: MODES OF OPERATION OF THE DIFFERENTIAL
PAIR
+

+

SingleEnded Mode Common Mode Differential Mode
+

G. Kovacs ©1997 Page 167
Chapter 11: DIFFERENTIAL AMPLIFIERS
3.1 LARGE SIGNAL OPERATION
+V
cc
v
C1
v C2
v
B1
v
B2 Q
1
Q
2
R
C1
R
C2
+V
cc
V
EE
I
αI
V
CC
 αI R
C
I
0.3 V
ONE INPUT ZERO,
THE OTHER SOME
"LARGE" DC
VOLTAGE...
V
CC
0
0
1 V
ON OFF
+V
cc
v
C1
v C2
v
B1
v
B2 Q
1
Q
2
R
C1 R
C2
+V
cc
V
EE
I
αI
V
CC
 αI R
C
I
 0.7 V
REVERSE THE
POLARITY OF THE
INPUT SIGNAL...
VCC
0
0
 1 V
ON
OFF
G. Kovacs ©1997 Page 168
Note: singleended operation shown.
Chapter 11: DIFFERENTIAL AMPLIFIERS
3.2 SMALLSIGNAL OPERATION
+V
cc
v
B1
v
B2 Q
1
Q
2
R
C1
R
C2
+V
cc
V
EE
I
approximately  0.7 V
SIGNAL IS A
FEW
MILLIVOLTS...
small
voltage
V
CC

αI
2
R
C
 α∆IR
C
V
CC

αI
2
R
C
+ α∆IR
C
I
2
+ ∆I
I
2
 ∆I
v
o
= 2α∆IR
C

+
Note: singleended operation shown.
• For a small input signal (this is like our previous example with one input grounded), the
current distributes itself between the transistors so that one gets an incremental amount
∆I more and the other one gets ∆I less....
• NOTE that the differential output voltage is,
v
o
= 2α∆IR
C
• NOTE the polarity of v
o
...
For a POSITIVE small input voltage, v
C1
DECREASES
For a NEGATIVE small input voltage, v
C1
INCREASES
v
C2
does the opposite....
• NOTE that the current source node is at approximately 0.7V
G. Kovacs ©1997 Page 169
Chapter 11: DIFFERENTIAL AMPLIFIERS
3.3 COMMONMODE OPERATION
+V
cc
v
C1
v C2
v
B1
v
B2 Q
1
Q
2
R
C1
R
C2
+V
cc
V
EE
I
v
CM
αI
2
αI
2
V
CC

αI
2
R
C V
CC

αI
2
R
C
I
2
I
2
v
cm
 0.7 V
SAME SIGNAL
ON BOTH INPUTS...
"COMMON" MODE
• Both inputs are driven by the same signal in commonmode operation.
• Since both transistors are the same (matched), the current will remain symmetrical
between the two transistors and so the output voltages v
C1
and v
C2
will exactly track
each other. Therefore the difference between v
C1
and v
C2
will be ZERO.
• This means that common mode signals are rejected! (as long as R
C1
= R
C2
= R
C
)
• Remember from the opamp section that commonmode rejection is supposed to be
high.
G. Kovacs ©1997 Page 170
Chapter 11: DIFFERENTIAL AMPLIFIERS
4: DETAILS OF SMALLSIGNAL OPERATION
+V
cc
Q
1
Q
2
R
C1
R
C2
+V
cc
V
EE
I

+
+

+

+

αI
2
 g
m
v
d
2
αI
2
+ g
m
v
d
2
g
m
R
C
v
d
V
CC

αI
2
R
C
 g
m
R
C
v
d
2
V
CC

αI
2
R
C
+ g
m
R
C
v
d
2
v
BE1
= V
BE
+
v
d
2
v
BE2
= V
BE

v
d
2
v
E
v
d
<< 2 V
T
g
m
≡
I
C
V
T
=
α I
2 V
T
v
d
• In this section, the goal is to look at small signal operation with differential inputs,
which are seen to be split evenly between the two transitors BE junctions.
• By defining the voltage at the common emitter node as v
E
and taking note of the
polarities shown above, one can write expressions for the emitter currents in terms of a
common voltage...
i
E1
·
I
S
α
e
v
B1
− v
E
( )
v
T
i
E2
·
I
S
α
e
v
B2
−v
E
( )
v
T
• Which can be combined into,
i
E1
i
E2
· e
v
B1
−v
B2
( )
v
T
• Remember that I
C
· I
S
e
v
BE
v
T
G. Kovacs ©1997 Page 171
Chapter 11: DIFFERENTIAL AMPLIFIERS
• These equations can be rearranged into the forms,
i
E1
i
E1
+ i
E2
=
1
1 + e
vB2  vB1
VT
and
i
E2
i
E1
+ i
E2
=
1
1 + e
vB1  vB2
VT
• Combined with the constraint imposed by the current source that they both sum to I,
i
E1
+ i
E2
= I
i
E1
=
I
1 + e
vB2  vB1
VT
and i
E2
=
I
1 + e
vB1  vB2
VT
• Which represent the currents as functions of the input voltage and shows that we have
an extremely sensitive variation of the i
E
’s with the differentialmode input voltages.
LINEAR REGION
I
C1
I
I
C2
I
NORMALIZED DIFFERENTIAL INPUT VOLTAGE
N
O
R
M
A
L
I
Z
E
D
C
O
L
L
E
C
T
O
R
C
U
R
R
E
N
T
v
B1
 v
B2
V
T
WITH NO
DIFFERENTIAL
INPUT SIGNAL
CURRENT SPLITS
50:50
G. Kovacs ©1997 Page 172
Chapter 11: DIFFERENTIAL AMPLIFIERS
+V
cc
Q
1
Q
2
R
C1
R
C2
+V
cc
V
EE
I

+
+

+

+

g
m
R
C
v
d
v
BE1
= V
BE
+
v
d
2
v
BE2
= V
BE

v
d
2
v
E
v
d
<< 2 V
T
g
m
≡
I
C
V
T
=
α I
2 V
T
i
C1
= I
C
+ g
m
v
d
2
i
C2
= I
C
 g
m
v
d
2
V
CC
 I
C
R
C
 g
m
R
C
v
d
2
V
CC
 I
C
R
C
+ g
m
R
C
v
d
2
v
d
• Redefining the steadystate current in each collector as,
I
C
=
αI
2
• We see that the collector currents can be expressed in a form that looks like a
common emitter amplifier gain equation with v
be
given by v
d
/2,
i
C1
= I
C
+ g
m
v
d
2
i
C2
= I
C
 g
m
v
d
2
• And the output voltages (considering both DC and AC components) are,
v
C1
= V
CC
 I
C
R
C
 g
m
R
C
v
d
2
v
C2
= V
CC
 I
C
R
C
+ g
m
R
C
v
d
2
G. Kovacs ©1997 Page 173
quiescent
signal
quiescent
signal
Chapter 11: DIFFERENTIAL AMPLIFIERS
• The differential gain (taken across the two outputs) is then,
A
d
=
v
C1
 v
C2
v
d
=  g
m
R
C
(written as for a CE amplifier)
(To reverse the sign, simply reverse your connections to the outputs!)
• And the singleended gain (looking at only one of the two outputs is 1/2 of that....
A
d
=
v
C1
v
d
= 
1
2
g
m
R
C
• In order to understand the smallsignal behavior of the differential pair, treat the
circuit as a cascade of two singletransistor amplifiers, a CE and a CB (a CASCODE)
and ask: WHAT IS THE GAIN?
+v
cc
Q
1
Q
2
v
o2
v
cc v
ee
2 I
E
v
s
v
o1
+
+Vcc
v
o1 v
o2
v1
v 2
Q
1
Q
2
R
C1
R
C2
2I E
+Vcc
VEE
v
s
2
+

v
s
2
+

R
C1
R
C2
G. Kovacs ©1997 Page 174
• Start by grounding one of the inputs (to make it simpler to understand)...
v
2
Chapter 11: DIFFERENTIAL AMPLIFIERS
Note that, it looks like two different configurations, depending upon
whether you are looking at v
o1
or v
o2
(think about this for a minute!)...
From the point of view of v
o1
, it is a common emitter amplifier that
is degenerated by the load on its emitter (Q2).
From the point of view of v
o2
, it is a common collector (emitter
follower) amplifier with a common base amplifier as its load.
(Remember: the 2I
E
current source is only there for the DC circuit!)
• MAKE A SMALLSIGNAL EQUIVALENT CIRCUIT....
R
i
=
r
π2
β + 1
= r
e2
≈
1
g
m2
2
g
m1
R
in
r
π1
R
c1
v
o2
v
o1
r
π2
R
c2

+
v
1

+
v
v
2
v
1
g
m2
+

i
2
i
in
NOTE
v
S
i
2
= 1 + β i
in
(Here, let’s use separate notation of the r
π
’s, g
m
’s, etc.... below we will assume they are
the same.)
• First look at the input resistance of the amplifier...
G. Kovacs ©1997 Page 175
Common
Base
Amplifier
B1
C1
E1
B2
C2 E2
Chapter 11: DIFFERENTIAL AMPLIFIERS
• By inspection, looking into the base of Q
1
, (Assuming r
π
’s are the same.)
R
in
= r
π
+ β + 1
r
π
β + 1
= 2r
π
• So, the input resistance is twice what it would be for a common emitter amplifier with
the same transistor.
• Now try to figure out the gain.... Start by computing the voltage at the input of the
common base stage, v
2
...
v
2
=  i
2
r
e
= 
v
s
R
in
β + 1
r
π
β + 1
= 
r
π
2r
π
v
s
= 
v
s
2
• Now we can calculate the output voltage v
o2
v
o2
= g
m
v
2
R
c
= g
m
 v
s
2
R
c
= g
m
R
c
v
s
2
Interesting Result!
G. Kovacs ©1997 Page 176
• From this we can find the gain...
This term is the
input current to
Q1
2
Minus sign because
of the way we defined
v (look at the
schematic)...
The load of
This term is
the current gain
of Q1 from B to E
This term is
just r e
Scaling because you
are looking into the
base of Q1
Divide by (ß + 1) because
you are looking into the
emitter of Q2. This is also
the same as r for Q2.
e
Chapter 11: DIFFERENTIAL AMPLIFIERS
A
v2
=
v
o2
v
s
=
g
m
R
c
2
• This is 1/2 the gain of a comparable common base amplifier....
• Also note that the other output (v
o1
) is an inverting version of (v
o2
) :
v
1
=
v
s
2
v
o1
= g
m
v
s
2
R
c
A
v1
=
v
o1
v
s
= 
g
m
R
c
2
• Note that you can obtain the same result using r
π
' and g
m
' for an unbypassed emitter
resistance of R
E
= r
e
= 1/g
m
.
• So, at this point we know that:
• the input resistance is 2r
π
• the gain is
g
m
R
c
2
(or 1 times that, depending on which output you use) > this is 1/2
the gain of a comparable commonemitter amplifier.
• This circuit can be used as a symmetrical “phasesplitter” to get
two equal amplitude versions of a signal 180° out of phase with
each other
G. Kovacs ©1997 Page 177
Hopefully you can see
why this is the case!
Chapter 11: DIFFERENTIAL AMPLIFIERS
5: HALFCIRCUIT MODEL OF DIFFERENTIAL PAIR
• The halfcircuit model is a way of taking advantage of the symmetry of the differential
pair circuit to simplify analysis in settings where we either drive the inputs differentially
(and symetrically) or we drive them in commonmode. The halfcircuit model does
not work for asymmetrically driven differential pairs (i.e. one input grounded, the other
driven).
• In this example, we drive the circuit differentially (when one output swings up, the
other swings down)... this is not the same as the previous case where we studied the
frequency response because then we grounded one input!
+V
cc
Q
1
Q
2
R
C1
R
C2
+V
cc
V
EE
I

+
+

+

+

g
m
R
C
v
d
v
d
2
v
E v
d
v
d
2
v
C2
v
C1
THIS VOLTAGE
DOES NOT CHANGE!
• The idea there is that the voltages pivot about the common point between the emitters
(the current source)
> v
E
does not change
> can thus treat v
E
as a power supply, so it is a virtual ground for small signals...
G. Kovacs ©1997 Page 178
Chapter 11: DIFFERENTIAL AMPLIFIERS
• Just split the differential pair in half (it looks like two identical commonemitter
circuits)...
+V
cc
Q
1
Q
2
R
C1
R
C2
+V
cc

+
+

+

v
d
2
v
d
2
v
C2
v
C1
d
g
m
R
C
v
2
d
g
m
R
C
v
2
• The halfcircuit idea works for any connections between the two sides. For example, if
there were a resistor of value R connected between the collectors, you could split it into
two resistors of value R/2, each connected with one end to one of the collectors and the
other end to ground.
G. Kovacs ©1997 Page 179
Chapter 11: DIFFERENTIAL AMPLIFIERS
6: HALFCIRCUIT MODEL AND COMMONMODE
OPERATION
v
cm
+

v
cm
+

v
cm
+

v
cm
+

Q
1
Q
2
Q
1
Q
2
v
C1
v
C2
v
C1
v
C2
R
C
R
C
R
C
R
C
+V
cc
+V
cc
+V
cc
+V
cc
V
EE
V
EE
V
EE
R
o
2R
o
2R
o
I
2
I
2 I
(Note that R
o
can be shown connected to ground, depending upon the current source.)
• Use the halfcircuit concept from above to analyze commonmode operation (which
preserves the symmetry that allows the halfcircuit model to be used in the first place).
• Relationship of halfcircuit to commonmode circuit: the two halves are identical so the
potential at the emitters is the same. Circuit theory says we can connect together two
nodes that have the identical voltages without affecting the operation of the circuit >
connecting the emitters in the halfcircuit leads to the physical equivalent of the circuit
on the left.
• This time, assume a current source with a shunt resistance R
o
... (REMEMBER the
output resistance, R
o
, of the “realistic” current sources we looked at before...)
• Each of the independent current sources provides a current of
I
2
and has an equivalent
shunt resistance of 2R
o
...
• Both inputs are driven by the same signal (otherwise it wouldn’t be common mode!)...
• IDEALLY common mode signals would be rejected! What happens here?
G. Kovacs ©1997 Page 180
Chapter 11: DIFFERENTIAL AMPLIFIERS
• With the nonideal current source, the voltage and the current at the V
E
node changes
with the input signal.
v
cm
+

Q
1
C1
v
R
C
+V
cc
V
EE
2R
o
This is the halfcircuit for AC signals....
(This method is easier to understand than Sedra & Smith’s.)
• Note that the current source is not shown because this is an AC circuit, but you would
still redraw it to obtain a smallsignal circuit!.
• This is just an emitter degenerated common emitter amplifier! So, just write down
the equation from memory (yeah, sure...),
v
c1
=  g
m
'
R
C
v
be
= 
g
m
1 + g
m
2R
o
R
C
v
cm
• If g
m
2R
o
>> 1, we can make the approximation,
v
c1
= 
R
C
2R
o
v
cm
and, by symmetry, also know that, v
c2
= 
R
C
2R
o
v
cm
• This gives a (singleended output) gain of,
A
cm
≈ 
R
C
2R
o
G. Kovacs ©1997 Page 181
Chapter 11: DIFFERENTIAL AMPLIFIERS
• For R
C
of 5 kΩ and R
o
of 250 kΩ, this gives,
A
cm
≈ 
R
C
2R
o
= 
1
100
(40dB, taking the absolute value, of course)
• This is attenuation, which is exactly what you want for common mode signals (which
are often noise that you do not want to amplify)!
7: COMMONMODE REJECTION RATIO
• This parameter is used to express how well a differential amplifier can reject common
mode signals compared to how well it can amplify differentialmode signals.
CMRR ≡ 20 log
Differential Gain
CommonMode Gain
= 20 log
A
d
A
cm
dB
(CMRR is almost always expressed in dB...)
• KEY POINT! If the output voltage is taken differentially, the commonmode gain
will be zero since both outputs do exactly the same thing (unless there are mismatches
in transistors and/or resistors)!
• If the output is taken in a singleended fashion, one can compute the CMRR...
• The singleended output, differential gain (from before) is,
A
d
= 
1
2
g
m
R
C
• Therefore, the CMRR can be written as,
CMRR = 20 log

1
2
g
m
R
C

R
C
2R
o
= 20 log g
m
R
o
dB
• It should be clear from the above equation that improving CMRR can be done via
increasing g
m
and/or R
o
.
G. Kovacs ©1997 Page 182
Chapter 11: DIFFERENTIAL AMPLIFIERS
8: FREQUENCY RESPONSE OF THE DIFFERENTIAL
AMPLIFIER
v
2
vo2
v
o1
R
s
R
c
R
c
C
BC
C
BC
C
π
C
π
r
π
r
π
+

g
m
v
2
g
m
v
v
s
1
v
1
+

v
X
+

• For this analysis, add R
s
, neglect r
o
and r
X
and assume that both transistors and their
bias conditions are the same.
Break this into two problems:
Part I going from the input signal v
s
to output V
o1
R
s
v
s
v
x
v
o1
r
π
c
π
C
BC
1
g
m
c
π
R
c
g
m
v
1
v
1
+

NOTE!
this is r
e
G. Kovacs ©1997 Page 183
Chapter 11: DIFFERENTIAL AMPLIFIERS
Part II going from the intermediate signal v
x
to output V
o2
 g
m
v
x
C
BC
R
C
v
o2
• This makes the analysis much easier to follow!
BEFORE LOOKING AT PART I...
• Remember about incorporating an emitter resistance into a “new” small signal
model for the transistor? This was the principle of local feedback that led to decreased
gain for emitter degenerated common emitter amplifiers.
CC
v
s
R
B2
R
C
V
B
V
R
E
R
B1
v
o
R
B
R
C
E
R
g
m
v
be
r
π
v
s
v
be
+

i
in
R
in
+

v
in
G. Kovacs ©1997 Page 184
Chapter 11: DIFFERENTIAL AMPLIFIERS
• Thus, you have effectively replaced the original hybridπ model with a new one
with an INCREASED r
π
(now called r
π
’) and a DECREASED g
m
(now called g
m
’)....
r
π
'
≡ r
π
1 + g
m
R
E g
m
'
≡
g
m
1 + g
m
R
E
• Now we will see that the same approach works if you have an IMPEDANCE in the
emitter circuit instead of a resistance!
• We will see that we can also suitably “modify” C
π
! This is also the result of local
feedback effects.
• Consider the same circuit we used in the common emitter lecture but with C
π
taken
into consideration...
g
m
v
be
C
π r
π
v
be
+

R
E
i
v
tot
v
e
+

+

Z
i
r
π
C
π
Z
π
=
i
RE
Note that
v = v
tot
s
• The approach is to consider the current and voltage at R
E
and the total voltage and
current into the base circuit (as shown above) and then to try to write a new expression
that gives the correct total voltage and current for a “new” parallel RC combination at
the base circuit....
• In other words, we want to modify the base circuit to take into account R
E
so we can
"ground" the emitter and treat it like a commonemitter amplifier after that.
• The impedance in the base circuit is,
z
π
= r
π

1
sC
π
=
r
π
1 + sC
π
r
π
=
1
y
π
• And the current flowing in the emitter resistance R
E
is,
G. Kovacs ©1997 Page 185
i
RE
= y
π
+ g
m
v
be
Chapter 11: DIFFERENTIAL AMPLIFIERS
• And the voltage across R
E
is,
v
e
= R
E
y
π
+ g
m
v
be
• Now write an expression for the total input voltage,
v
tot
= v
be
+ R
E
y
π
+ g
m
v
be
• And the input current into the “new” basecircuit impedance z
π
is,
i =
v
be
z
π
• Now (finally!) we can write an expression for the TOTAL input impedance considering
both the input impedance of the base circuit plus the effects of the emitter circuit...
z
in
=
v
tot
i
=
z
π
v
be
v
be
+ R
E
y
π
+ g
m
v
be
= z
π
+ R
E
+ z
π
R
E
g
m
= R
E
+ z
π
1 + g
m
R
E
• Substitute for z
π
.....
= R
E
+
r
π
1 + g
m
R
E
1 + s C
π
r
π
• At last... PAY DIRT!! It looks like a parallel RC combination in series with R
E
!
z
in
= R
E
+ r
π
1 + g
m
R
E

1
S
C
π
1 + g
m
R
E
G. Kovacs ©1997 Page 186
v
e
Chapter 11: DIFFERENTIAL AMPLIFIERS
• Our “new” equivalent circuit (neglecting R
E
)....
v
+

C
π
1 + g
m
R
E
g
m
1 + g
m
R
E
v
tot i
r
π
1 + g
m
R
E
• NOTE: here we have neglected R
E
because it is 1/g
m2
when the load is a CB stage, as
it is here.... if you put it into the above schematic, it makes things a bit less clear (but
more accurate!). You can look at the "full" derivation in Gray and Meyer, "Analysis and
Design of Analog Integrated Circuits," on pages 491  498.
• Note that for a differential pair, taking R
E
= r
e
into account yields r
π
' = 2r
π
, C
π
' = C
π
/2
and g
m
' = g
m
/2.
NOW FOR PART I...
• It looks just like the above, but with an emitter IMPEDANCE rather than just an
emitter resistance....
R
s
v
s
v
x
v
o1
r
π
c
π
C
BC
1
g
m
c
π
R
c
g
m
v
1
v
1
+

NOTE!
Z
E
• In order to get transfer functions (i.e. Both
v
o1
v
s
and
v
x
v
s
) look at the results we just
obtained.
G. Kovacs ©1997 Page 187
Chapter 11: DIFFERENTIAL AMPLIFIERS
Z
Z
E v
x
r
π c
π
g
m
v
v
i
v

+
+
+


be
be
r
π
1 + g
m
Z
E
v
i
+

r
π
'
C
π
'
C
π
1 + g
m
Z
E
g
m
1 + g
m
Z
E
v
i
= g
m
'
v
i
• For PART I, we know that,
Z
E
=
1
g
m
1 + s
c
π
g
m
and thus our “scaling factor”....
1 + g
m
Z
E
= 2
1 + s
c
π
2 g
m
1 + s
c
π
g
m
(Note that C
π
/g
m
= C
π
r
e
... perhaps more intuitive as an RC time constant.)
• This means there is a zero at w
z1
=
2 g
m
c
π
and a pole at w
p1
=
1
2
w
z1
=
g
m
c
π
G. Kovacs ©1997 Page 188
r
e
Here we see that the circuit reduces
to the familiar common emitter circuit
with the emitter grounded... this
makes the analysis easier and allows
us to use the Miller approximation
(we'll do that later).
Chapter 11: DIFFERENTIAL AMPLIFIERS
• Practically... assuming the pole and zero cancel since they are so close in
frequency.... the “scaling factor” reduces to,
1 + g
m
Z
E
≈ 2
• Thus, we get an answer that looks a bit familiar (remember all those factors of two?).
r
π
'
≈ 2r
π
c
π
'
≈
c
π
2
g
m
'
≈
g
m
2
G. Kovacs ©1997 Page 189
Chapter 11: DIFFERENTIAL AMPLIFIERS
NOW, USING THESE NEW VALUES FOR THE “EQUIVALENT”
TRANSISTOR, SOLVE PART 1....
R
s
v
s
v
x
v
o1
r
π
c
π
C
BC
1
g
m
c
π
R
c
g
m
v
1
v
1
+

NOTE!
R
s
v
s
R
s
v
s
r
π
' C
π
'
C
BC
g
m
v
1
'
v
1
+

'
r
π
'
v
1
+

'
C
π
'
v
o1
R
c
1
2
C
BC
1 + g
m
'
R
C
'
v
o1
R
c
C
BC
g
m
v
1
' '
USE MILLER
APPROXIMATION
SUBSTITUTE
EQUIVALENT
CIRCUIT
G. Kovacs ©1997 Page 190
NOTE THAT: A
v
'
=  g
m
'
R
C
so the Miller capacitance is C
BC
1 + g
m
'
R
C
Chapter 11: DIFFERENTIAL AMPLIFIERS
• The input time constant is given by,
τ
in
= R
S
 r
π
'
c
π
'
+ C
BC
1 + g
m
'
R
C
= R
S
 2r
π
c
π
2
+ C
BC
1 +
g
m
2
R
C
• NOTE that if R
s
= 0, r
x
is the main thing that sets the upper cutoff frequency....
The above equation (approximate!) considering r
x
and r
o
is:
τ
in
= R
S
+ r
x
 2r
π
c
π
2
+ C
BC
1 +
g
m
2
R
C
 r
o
• The output circuit time constant is,
τ
out
= C
BC
R
C
• And we can now write the overall transfer function equation for the
v
o1
v
s
path.....
v
o1
v
s
= T S = 
r
π
'
R
S
+ r
π
'
g
m
'
R
C
1
1 + Sτ
in
1
1 + Sτ
out
• Substituting in for the “primes”....
v
o1
v
s
= T S = 
2r
π
R
S
+ 2r
π
g
m
2
R
C
1
1 + Sτ
in
1
1 + Sτ
out
This is a lowpass response only (no highpass)... this
is a DCcoupled amplifier!
G. Kovacs ©1997 Page 191
DC
AC
Chapter 11: DIFFERENTIAL AMPLIFIERS
NOW DO PART II...
 g
m
v
x
C
BC
R
C
v
o2
• Assuming (see above) that,
v
x
=
v
i
2
(where v
i
is the voltage input to the base of Q
1
, AFTER R
s
...)
• We look at the schematic from PART I again....
R
s
v
s
v
x
v
o1
r
π
c
π
C
BC
1
g
m
c
π
R
c
g
m
v
1
v
1
+

NOTE!
+

v
i
 g
m
v
x
C
BC
R
C
v
o2
• Knowing (from above) that the rolloff of the voltage v
i
is given by,
1
1 + S τ
in
G. Kovacs ©1997 Page 192
Chapter 11: DIFFERENTIAL AMPLIFIERS
• v
i
can be found as a function of v
S
by simple voltage division, knowing that the input
resistance (after R
S
) is 2r
π
v
i
=
2 r
π
R
S
+ 2 r
π
v
s
1
1 + S τ
in
• Again, we use the fact that v
X
is simply 1/2 of v
i
....
• We can also see that the output signal v
o2
is,
v
o2
= g
m
v
x
R
C
1 + S R
C
C
BC
= g
m
v
i
2
R
C
1 + S τ
out
• Now we can find the complete transfer function for v
o2
,
v
o2
v
s
= T S =
2r
π
R
S
+ 2r
π
g
m
2
R
C
1
1 + Sτ
in
1
1 + Sτ
out
= 
v
o1
v
s
• THIS IS THE SAME AS FOR v
o1
, but it is 180° out of phase....
G. Kovacs ©1997 Page 193
input voltage
divider term
rolloff due to
input circuit
Chapter 11: DIFFERENTIAL AMPLIFIERS
9: SUMMARY OF DIFFERENTIAL PAIR SMALL
SIGNAL OPERATION
• R
in
= 2r
π
• Both signal paths have a midband gain of
g
m
2
R
C
with an input voltage divider of
R
in
R
S
+ R
in
• One path is inverting and the other one is noninverting.
• Both paths have the same time constants!
• Small signal gain:
+v
cc
Q
1
Q
2
v
o2
v
cc v
ee
2 I
E
v
s
v
o1
+
v
o1
= 
2 r
π
R
s
+ 2 r
π
g
m
2
R
C
v
o2
=
2 r
π
R
s
+ 2 r
π
g
m
2
R
C
R
S
R
in
v
s
v
s
G. Kovacs ©1997 Page 194
NOTE!
Chapter 11: DIFFERENTIAL AMPLIFIERS
10: OTHER IMPERFECTIONS OF “REALISTIC”
DIFFERENTIAL AMPLIFIERS
• This section deals with imperfections of differential amplifiers beyond just the common
mode issues considered thus far.
THESE ARE SOME KEY TERMS TO KNOW!
10.1 INPUT OFFSET VOLTAGE
• If you ground both inputs of the differential pair, the output (taken differentially) should
be zero, right? Well, in practice, it may not be....
• The input offset voltage is defined as the DC voltage that appears at the output
with the inputs grounded, divided by the gain of the amplifier (to refer the offset
to the input of the amplifier...).
V
OS
≡
Output voltage with inputs grounded
Differential Gain
=
V
O
A
d
• If you apply a voltage V
OS
to the inputs of the amplifier, the output goes to zero
(in theory) !!!!
G. Kovacs ©1997 Page 195
+

V
o
≠ 0
A
v
+

V
os
now = 0
(in theory)
A
v
V
os
A
v
Chapter 11: DIFFERENTIAL AMPLIFIERS
• In other words, with an imperfect amp, you get V
o
out if you ground the inputs... V
o
is
what you would get if you put V
os
into a perfect amp with no offset!
 V
OS
+

Q
1
Q
2
Q
1
Q
2
0 V
R
C
R
C
R
C
R
C
+V
cc
+V
cc
+V
cc
+V
cc
V
EE
I
o
V
+ 
V
EE
I
• Where could this be a problem? If you cascade several DCcoupled amplifiers,
the offset voltages get multiplied by the gains and eventually saturate the amplifier!
• Where does the offset come from?
• Mismatches in R
C
’s, mismatches in the transistors, etc....
• Consider offsets in the R
C
’s... (assume matched transistors)
• If the mismatch is evenly distributed between the two resistors (this is not as weird as
it sounds, since you simply “split the difference” between the resistors and call the
average value R
C
...)
R
C1
= R
C
+
∆R
C
2
R
C2
= R
C

∆R
C
2
• If the transistors are matched, the current will divide equally between them, allowing us
to write,
V
C1
= V
CC

αI
2
R
C
+
∆R
C
2
V
C2
= V
CC

αI
2
R
C

∆R
C
2
G. Kovacs ©1997 Page 196
Chapter 11: DIFFERENTIAL AMPLIFIERS
• Which gives an output voltage of,
V
O
= V
C1
 V
C2
=
αI
2
∆R
C
• To calculate the input offset voltage, we need to recall that A
d
is given by,
A
d
= g
m
R
C
where g
m
=
I
C
V
T
=
αI
2
V
T
(note that this is not
g
m
2
R
C
because the output is taken differentially)
• Therefore, the input offset voltage is a function of the matching of collector resistors,
not their absolute value,
V
OS
=
V
O
A
d
=
αI
2
∆R
C
αI
2
V
T
R
C
= V
T
∆R
C
R
C
• Thus the offset is clearly a function of temperature V
T
·
kT
q
¸
¸
_
,
and the matching of
the resistors,
V
os
T
Increasing ∆R
C
R
C
G. Kovacs ©1997 Page 197
≈ 26 millivolts
Units of % change
Chapter 11: DIFFERENTIAL AMPLIFIERS
• For +/ 1% resistors, this gives a maximum mismatch of 2% and maximum V
OS
of 0.5
mV.
• You can do the same type of analysis for differences in the transistor parameters
(Sedra and Smith consider differences in the emitterbase junction areas that in turn
cause differences in I
S
between the two devices).
• If you work out the same thing for this case, you end up with a similarlooking equation
that gives V
OS
in terms of I
S
mismatch,
V
OS
=
V
O
A
d
= V
T
∆I
S
I
S
• Note that I
S
is also a function of T... I
S
∝ T
α
10.2 INPUT BIAS AND OFFSET CURRENTS
• Input bias currents represent the currents required to turn on the input transistors (i.e.
the BASE CURRENTS!!!).
• The input bias current is given by,
I
B1
= I
B2
=
I
2
β + 1
• If the β’s are mismatched, the input bias currents will also be mismatched!
• The INPUT OFFSET CURRENT is simply defined as the difference between them!
I
OS
= I
B1
 I
B2
• Using similar math as for the offset voltage, and assuming β mismatch given by,
G. Kovacs ©1997 Page 198
β
1
= β +
∆β
2
β
2
= β 
∆β
2
the emitter current in
each of the two transistors
Chapter 11: DIFFERENTIAL AMPLIFIERS
• It can be shown that (see Sedra and Smith),
I
OS
= I
B
∆β
β
where I
B
is the "average" base current I
B
=
I
B1
+ I
B2
2
=
I
2 β + 1
• Here can improve things by reducing I
C
(and hence I
B
).
• Remember that the same terms apply to operational amplifiers, but may be brought
about by more complicated effects (i.e. the same formulas may not always apply, but
the basic principles do!).
+ +
+
Acm ( Vi+ + Vi) / 2
Adm ( Vi+  Vi)
Ro
Ri
Vos
Ios
Ibias
Ibias
Vi +
Vi 
Vo
VCC
 VEE
G. Kovacs ©1997 Page 199
Chapter 12: CURRENT SOURCES
Chapter 12: CURRENT SOURCES
Oh, you’re supposed to pick up the OTHER end!!!
Berkeley EE student’s thirtysecond encounter with a soldering iron...
1: OBJECTIVES
• To consider:
The basics current source circuits.
The limitations of common current sources.
Design and use of current sources in multistage amplifiers.
• In order to build discrete amplifiers and integrated versions, current sources are
necessary. In many cases, a single reference current must be scaled and "copied" (the
term used is "mirrored") so that the bias points of several transistors on a chip are
proportional to each other and the reference.
• The purpose of this chapter is to discuss such circuits and their uses.
G. Kovacs ©1997 Page 200
Chapter 12: CURRENT SOURCES
2: THE DIODECONNECTED TRANSISTOR
• This simple circuit is a buildingblock for current sources.
• A diode can be formed by connecting the base to the collector.
• The transistor now behaves like a diode with an IV characteristic equal to the I
C
 V
BE
curve for the transistor. The resulting VBE can be applied to the baseemitter junction
of a second transistor, which will have almost exactly the same collector current as the
current flowing through the diode, making a "replica."
base current =
i
β + 1
β
β + 1
i = αi
i
+

v
BE
+

+

Same
for PNP...
G. Kovacs ©1997 Page 201
Chapter 12: CURRENT SOURCES
3: THE CURRENT MIRROR
• This circuit makes good use of the fact that the diode formed from a transistor has the
same temperature coefficient as the transistor.
• Use one transistor as a v
BE
reference > feed in a desired reference current and the
“diode” will establish the exact v
BE
on the other transistor to get the same current to flow
in it....
+ V
CC
 V
EE
I
REF
I
E
β + 1
I
E
β + 1
2I
E
β + 1
β
β + 1
I
E
I
o
=
β
β + 1
I
E
I
E
I
E
V
o
• Looking at the circuit above, it is obvious that,
I
o
=
β
β + 1
I
E
and therefore, we need (to get the desired I
0
out...),
I
REF
=
β
β + 1
I
E
+
2I
E
β + 1
=
β + 2
β + 1
I
E
G. Kovacs ©1997 Page 202
• So, for large β, I
REF
= I
o
... For smaller β, need to make I
REF
> I
o
by the correct amount...
Chapter 12: CURRENT SOURCES
4: A SIMPLE CURRENT SOURCE
+ V
CC
 V
EE
I
REF
V
o
+

v
BE
(or ground...)
I
o
Q1 Q2
I
REF
=
V
CC
 V
BE
R
I
REF
=
V
CC
 V
BE
R
= I
C1
+ I
B1
+ I
B2
= I
C1
+
I
C1
β
+
I
C2
β
= I
C1
+
2I
C1
β
I
C1
= I
C2
= I
o
• As for the above case,
As β → ∞ I
o
= I
REF
(losing no current in the bases)
G. Kovacs ©1997 Page 203
BE
V
's are the same....
I
C2
is the same as I
o
Chapter 12: CURRENT SOURCES
5: THE WIDLAR CURRENT SOURCE
+ V
CC

EE
I
REF
+

V
BE1
(or ground...)
I
o
Q1
Q2
V
V
BE2
+

R
E
• Add in an emitter resistance in the output transistor....
• As before, R
E
stabilizes V
BE2
against variations with Q
2
’s β, increases the output
impedance (makes the current source more ideal... see below) and allows scaling of
the current to any value less than I
REF
.
• Looking at the circuit one can see (base currents are neglected for high β ) that,
I
o
·
V
BE1
− V
BE2
R
E
·
V
E
R
E
• The equation is not very useful in terms of V
BE
’s... We know that,
V
BE1
· V
T
ln
I
REF
I
S
¸
¸
_
,
from I
C
· I
S
e
v
BE
V
T
¸
¸
_
,
V
BE2
= V
T
ln
I
o
I
S
• So, plug into the equation for I
o
...
G. Kovacs ©1997 Page 204
I
o
R
E
= V
T
ln
I
REF
I
o
You could get the
reference current
from a voltage
reference and a
resistor, for
example...
+

V
E
Chapter 12: CURRENT SOURCES
• To design a Widlar current source: 1) Choose I
REF
and 2) Compute R
E
knowing the
I
o
needed.
6: CURRENT MIRRORS
• This concept can easily be multiplied... You can use one diodeconnected transistor
to “steer” the current through many others by connecting their bases together...
• Since the collector current as a function of v
BE
is given by,
I
C
= I
S
e
vBE
VT
• We can deliberately adjust I
S
for any given transistor to control its I
C
...
• I
S
is a function of the area of the emitter, so in integrated circuits we can just make a
bigger emitter to get more current....
+ V
CC
I
REF
Q1
 V
EE
(or ground...)
+

v
BE
+

v
BE
+

v
BE
+

v
BE
 V
EE
 V
EE
 V
EE
2 I
REF
I
REF
3 I
REF
This transistor
has twice the
emitter area as
the reference...
This transistor
has three times the
emitter area as
the reference...
G. Kovacs ©1997 Page 205
Chapter 12: CURRENT SOURCES
7: NONIDEALITIES OF BJT CURRENT SOURCES
• Real transistors are not ideal current sources!
• Remember the Early Voltage and r
o
?
• r
o
is a resistance that models the slight effect of collector voltage on collector current in
the active region of operation (the curves are not exactly flat!).
• r
o
is inversely proportional to the DC bias current, so that as you increase I
C
to get
more g
m
and more gain, you also get a lower (worse) resistance between the collector
and emitter (i.e. the slope of the current/voltage curves is steeper).
r
o
=
V
A
I
C
where V
A
is the Early voltage... (see pages 207  208 in S & S)
V
I
I
V
Slope = 1/r
o
• Remember that in the above figure, the I
C
versus V
CE
curves represent equal base
current steps between them.
• r
o
is typically a few thousand Ohms...
B C
E
r
r
r
x
o
π
g
m
v
be
v
be
π
BC
C
C
+

THIS
THING
HERE!
G. Kovacs ©1997 Page 206
Chapter 12: CURRENT SOURCES
7.1 EFFECTS OF r
o
ON CURRENT SOURCE
PERFORMANCE
• An ideal current source has an infinite output resistance, but a “real” one with a BJT
does not....
r
o
I
o
I
o
• We will see later that this means that you can’t simply “neglect” the bias current
sources in your smallsignal equivalent circuits if you want to get a detailed idea of
performance....
• Try to find out what the output resistance of a BJT current source is (look at Widlar)...
+

r
π v
π
+

R
E
g
m
v
π
r
o
C
v
x
i
x
B
E
Use a test
voltage...
• As usual, want to use Ohm’s Law to compute R
o
...
G. Kovacs ©1997 Page 207
Chapter 12: CURRENT SOURCES
• Use R
E
'
= r
π
 R
E
and redraw the circuit for clarity...
+

v
x
r
o
C
v
π

+
R
E
'
= r
π
 R
E
g
m
v
π
i
x
g
m
+
1
R
E
'
v
π
v
π
R
E
'
• The currents are relatively easy to see.... (careful with signs!)
i
x
= g
m
v
π
 g
m
+
1
R
E
'
v
π
= 
1
R
E
'
v
π
• And the voltages can be summed around the loop to get v
x
...
v
x
=  v
π
 g
m
+
1
R
E
'
v
π
r
o
• Now you can go ahead and use Ohm’s Law...
R
o
≡
v
x
i
x
=
1 + g
m
+
1
R
E
'
r
o
1
R
E
'
= R
E
'
+ 1 + g
m
R
E
'
r
o
• Which, for small R
E
’
(where R
E
'
· r
π
 R
E
) gives (more intuitive)...
R
o
≈ 1 + g
m
R
E
'
r
o
G. Kovacs ©1997 Page 208
Chapter 13: MULTISTAGE AMPLIFIERS
Chapter 13: MULTISTAGE
AMPLIFIERS
OW, OW, OW, OW!!!!
Berkeley EE student’s thirtythird encounter with a soldering iron...
1: OBJECTIVE
• To consider:
How to analyze and design multistage amplifiers based on the differential pair.
REVIEW SEDRA & SMITH SECTION 6.10
2: ANALYSIS OF AN EXAMPLE AMPLIFIER
• Begin with a simplified version of the EE122 discrete component opamp....
G. Kovacs ©1997 Page 209
+V
cc
Q
1
Q
2
R
C1
R
C2
+V
cc
V
EE
I
I
c
v
o
R
E
V
EE
+V
cc
R
S
v
s
Q
3
I
2
I
I
2
 I
B
I
B
EQ3
C
C
Chapter 13: MULTISTAGE AMPLIFIERS
• A small signal model for the differential pair with one input grounded has already been
discussed, and is shown below.
• Looking at the output stage, there is emitter degeneration, and the load is also a
function of R
o
of the current source (R
o
not shown above, but it is the output resistance
of the current source driving Q
3
) and the load resistance (R
L
, not shown in the schematic)...
A
voutput
≈ g
m
'
R
L
 R
o
( )
• To evaluate the effect of the base current of Q
3
on the differential pair (tending to
unbalance it), look at the voltages...
I
2
 I
B
R
C1
= I
EQ3
R
E
 V
BE3
• One needs to compensate for the extra current flowing into Q
1
! Do this by adjusting
R
C
’s... THINK ABOUT THIS! To get balanced output voltages for the differential pair if
one side has less current flowing in it (due to the base current of Q
3
), that R
C
can be
made larger.
• Next consider the frequency response of this amplifier by considering it with one input
grounded and using the model discussed in the Differential Amplifiers Chapter.
• For this situation, the common emitter amplifier formed by Q
1
has an emitter resistance
of r
e
(of Q
2
) and drives the output stage formed by Q
3
.
• We already looked at the response of the differential pair with one input grounded and
arrived at a smallsignal equivalent circuit like the one shown below:
G. Kovacs ©1997 Page 210
Chapter 13: MULTISTAGE AMPLIFIERS
v
2
v
o2
v
o1
R
s
R
c
R
c
C
BC1
C
BC2
C
π2
C
π1
r
π1
r
π2
+

g
m2
v
2
g
m1
v
v
s
1
v
1
+

• The signal path to v
o1
was drawn as shown below, but now we will have to consider
the loading effects of Q
3
on this circuit (not shown in the schematic immediately below),
R
s
v
s
r
π1
'
v
1
+

'
C
π1
'
v
o1
R
C
C
BC1
g
m1
v
1
' '
C
BC1
1 + g
m1
'
R
C
where r
π1
'
≈ 2r
π1
C
π1
'
≈
C
π
2
g
m
'
≈
g
m1
2
• The output stage that is driven by v
o1
must then be considered at this stage's load.
• This load includes the input resistance of the output stage in parallel with the R
C
of Q
1
(otherwise the Miller approximation shown above will be incorrect since the gain at the
transistor in question will change).
• The output is a common emitter amplifier with an unbypassed emitter resistance R
E
,
and a collector resistance that is the output resistance of the current source (not shown
in the simplified schematic at the beginning of this chapter).
G. Kovacs ©1997 Page 211
• The input resistance of the output stage becomes the output load for the first stage.
Chapter 13: MULTISTAGE AMPLIFIERS
R
s
v
s
r
π1
'
v
1
+

'
C
π1
'
R
C
C
BC1
gm1 v
1
' '
r
π3
'
v
3
+

'
C
π3
'
R
L
gm3 v
3
' '
C
BC3
+ C
C C
BC3
+ C
C
1 + g
m3
'
R
L
C
BC1
1 + g
m
'
R
C
 r
π3
'
v
o
Where,
r
π1
'
= 2 r
π1
g
m1
'
=
g
m1
2
r
π3
'
= r
π3
1 + g
m3
R
E
g
m3
'
=
g
m3
1 + g
m3
R
E
C
π1
'
=
C
π1
2
C
π3
'
=
C
π3
1 + g
m3
R
E
• For the midband gain (ignore the capacitances)...
A
M
=
r
π1
'
R
S
+ r
π1
'
g
m1
'
R
C
 r
π3
'
g
m3
'
R
L
NOTE that this does not consider the current source output resistance (R
o
) here, and if
we did, it would be in parallel with R
L
in the above equation. In the case above, the gain
G. Kovacs ©1997 Page 212
is a direct function of R
L
.
Input voltage
divider....
Firststage
Gain...
Secondstage
Gain...
Chapter 13: MULTISTAGE AMPLIFIERS
R
s
v
s
r
π1
'
v
1
+

'
C
π1
'
R
C
C
BC1
gm1 v
1
' '
r
π3
'
v
3
+

'
C
π3
'
R
L
gm3 v
3
' '
C
BC3
+ C
C C
BC3
+ C
C
1 + g
m3
'
R
L
C
BC1
1 + g
m
'
R
C
 r
π3
'
v
o
• For the high frequency response, you need to consider the Miller multiplication of the
capacitances in both stages...
• NOTE that the compensation capacitor C
C
gets Miller multiplied too, as intended!!!
1) The input time constant is,
R
S
 r
π1
'
C
π1
'
+ C
M1
where C
M1
= C
BC1
1 + g
m1
'
R
C
 r
π3
'
Since R
S
is typically very small, and thus corresponds to a very high frequency...
2) The interstage time constant is (for extra nerdiness we could include C
BC1
but it is a
relatively small contribution),
r
π3
'
 R
C
C
π3
'
+ C
M3
where C
M3
= C
BC3
+ C
C
1 + g
m3
'
R
L
• This is a relatively long time constant because C
M3
is very large (both C
BC3
and C
C
get
Miller multiplied) and because the resistance here is moderate in magnitude.
3) And the output time constant is,
R
L
C
BC3
+ C
C
• This time constant will be fairly short, but may be longer than the input circuit, depending
G. Kovacs ©1997 Page 213
on R
L
, and external capacitances that may be in parallel with R
L
.
Chapter 13: MULTISTAGE AMPLIFIERS
• Look at the example in Sedra and Smith, page 438... First, look at it in chunks that
are easy to understand (the reader may wish to add some lines and notations here)...
VCC
VEE
• Note the current mirrors to distribute current to the circuit, with the currents referenced
to the resistor feeding the diodeconnected transistor at lower left.
• Note the cascaded differential amplifier for increased gain (inputs are at the left, both
shown grounded).
• Note that one of the transistors in the second differential amplifier has no collector
resistor. It is not needed since no voltage is needed from this collector and it is the
current there that is required for proper operation of the differential amplifier (the output
beyond this point is singleended, and the voltage from the second differential pair is
taken at the other transistor, which does have a collector resistor).
• Note the PNP transistor used as an emitter degenerated common emitter stage just
before the emitter follower output. If PNP transistors were not available, some other
sort of levelshifter circuit (see below) would be necessary since the DC voltage in the
signal path would otherwise be too close to V
cc
.
G. Kovacs ©1997 Page 214
Chapter 13: MULTISTAGE AMPLIFIERS
3: LEVEL SHIFTERS
• In multistage amplifiers on integrated circuits, coupling capacitors between stages are
almost always not used because they cannot be made large enough for reasonable
lowfrequency operation.
• Thus, stages are DCcoupled. This means that voltage offsets like V
BE
drops between
stages can start to add up... stages referred to as level shifters can be used to
compensate where necessary.
• A typical level shifter is simply a degenerated CE amplifier (it can provide gain or
simply act as an inverting buffer if R
E
= R
C
). The choice of R
C
"programs" the quiescent
collector voltage of the level shifter and can thus be used to "center" the output voltage
of an opamp so that it can swing both positive and negative from its quiescent point.
• Also, emitter followers and diodes can be used to drop ≈ 0.7 V per stage.
• By using PNP and NPN transistors, the sign of the voltage drop per shifter can be
changed.
G. Kovacs ©1997 Page 215
Chapter 14: FEEDBACK
Chapter 14: FEEDBACK
.... by building an amplifier whose gain is made deliberately, say 40 decibels
higher than necessary (10,000fold excess on an energy basis) and then feeding
the output back to the input in such a way as to throw away the excess gain, it
has been found possible to effect extraordinary improvement in constancy of
amplification and freedom from nonlinearity.
Harold Black, inventor of negative feedback, 1934
1: OBJECTIVES
• To consider:
The basics of feedback.
The properties of negative feedback.
The basic feedback topologies.
An example of the “ideal” feedback case.
Some realistic circuit examples and how to analyze them.
READ S&S Sections 8.1  8.7
G. Kovacs ©1997 Page 216
Chapter 14: FEEDBACK
2: INTRODUCTION TO FEEDBACK
• There are two types of feedback: regenerative (positive feedback) and degenerative
(negative feedback).
• Unless you want your circuit to oscillate, we usually use NEGATIVE FEEDBACK...
• This idea came about in the late 1920’s when they were able to build amplifiers with
reasonable gains, but with gains that were difficult to control from amplifier to amplifier...
• One day, while riding the Staten Island Ferry, Harold Black invented negative feedback....
2.1 PROPERTIES OF NEGATIVE FEEDBACK
• The gain of the circuit is made less sensitive to the values
of individual components.
• Nonlinear distortion can be reduced.
• The effects of noise can be reduced (but not the noise
itself).
• The input and output impedances of the amplifier can be
modified.
• The bandwidth of an amplifier can be extended.
• All you have to do to “get some feedback” (of the negative kind) is to supply a scaled
replica of the amplifier’s output to the inverting (negative) input (more on this below) and
presto!
• Of course, if you use negative feedback, overall gain of the amplifier is always less
than the maximum achievable by the amplifier without feedback.
G. Kovacs ©1997 Page 217
Chapter 14: FEEDBACK
2.2 THE BASIC FEEDBACK CIRCUIT
• With an input signal x
s
, an output signal x
o
, a feedback signal x
f
, and an amplifier input
signal x
i
, let’s look at the basic feedback circuit illustrated above.
• The amplifier has a gain of A and the feedback network has a gain of β...
• The input to the amplifier is,
x
i
= x
s
 x
f
• The output of the amplifier is,
x
o
= Ax
i
• So we can obtain an expression for the output signal in terms of the input signal and
the feedback gain...
x
o
= A x
s
 x
f
= A x
s
 βx
o
• Rearranging,
x
o
= Ax
s
 Aβx
o
→ x
o
1 + Aβ = Ax
s
G. Kovacs ©1997 Page 218
Chapter 14: FEEDBACK
• From which we obtain the negative feedback equation by solving for the overall gain
x
o
x
s
,
A
fb
≡
x
o
x
s
=
A
1 + Aβ
• This means that the gain is almost entirely determined by the feedback circuit!!!
A
fb
≡
x
o
x
s
=
A
1 + Aβ
≈
1
β
for large A
• For positive feedback, you only need to change the “+” sign in the denominator to a “”
sign....
• It is easy to obtain the equation for the feedback signal, x
f
,
x
f
=
Aβ
1 + Aβ
x
s
• If the amplifier gain and the loop gain are large (i.e. Aβ >> 1), then the feedback signal
x
f
becomes nearly an identical copy of the input signal x
s
....
• This explains why the two terminals of an opamp become nearly identical when
using negative feedback... (Remember the “virtual ground” stuff?)
G. Kovacs ©1997 Page 219
Chapter 14: FEEDBACK
3: HOW FEEDBACK AFFECTS BANDWIDTH
• Assuming an amplifier with a singlepole frequency response (i.e. an “ideal” opamp),
its frequency response is given by,
A s =
A
M
1 +
s
ω
H
• If you use the amplifier with negative feedback, the gain becomes,
A
f
s =
A s
1 + βA s
• Substituting, you get,
A
f
s =
A
M
1 + A
M
β
1 +
s
ω
H
1 + A
M
β
• Thus we have another singlepole response, but with a high cutoff frequency given
by,
ω
Hf
= ω
H
1 + A
M
β
• THIS MEANS THAT THE UPPER CUTOFF FREQUENCY IS INCREASED
BY A FACTOR EQUAL TO THE AMOUNT OF FEEDBACK.
THIS IS THE FUNDAMENTAL GAINBANDWIDTH PRODUCT
TRADEOFF THAT WE STUDIED WITH OPAMPS!!!
• The same is true for the lower cutoff frequency, and the amplifier will have a lower
cutoff frequency by a factor again equal to the amount of feedback,
ω
Lf
=
ω
L
1 + A
M
β
G. Kovacs ©1997 Page 220
Chapter 14: FEEDBACK
THE BANDWIDTH IS EXTENDED BY THE SAME
AMOUNT THE MIDBAND GAIN IS DECREASED.
• A KEY ASSUMPTION HERE IS THAT THE SOURCE, LOAD AND FEEDBACK
NETWORK DO NOT AFFECT THE GAIN OF THE AMPLIFIER!
• IN PRACTICE THIS WILL NOT BE THE CASE, SO WE WILL LEARN HOW TO TAKE
AMPLIFIER LOADING INTO CONSIDERATION.
FUNDAMENTAL ASSUMPTIONS:
1) The feedback network is UNILATERAL. This means that it
transmits signals only one way (from output to input of the
composite circuit). This is usually only approximately true.
2) The amplifier is UNILATERAL. This means that it transmits
signals only one way (from the input to the output of the composite
circuit).
3) The loop gain, β, is INDEPENDENT of the load and source
resistances seen by the composite circuit.
G. Kovacs ©1997 Page 221
Chapter 14: FEEDBACK
4: FROM BASIC BLOCK DIAGRAM TO ACTUAL
FEEDBACK CIRCUITS
4.1 REMINDER: TYPES OF AMPLIFIERS
+

+

Ideal Voltage Amplifier
Ideal Transconductance Amplifier
Ideal Transresistance Amplifier
Ideal Current Amplifier
R
A
v
A
G
M
M
i
L
R
v
i
i
i
v
s
i
s
v
i
i
i
v
s
i
s
L
R
L
R
L
R
s
R
s
R
s
R
s
R
v
o
v
o
v
i
v
i
i
i
i
i
+

+

+

+

i
o
i
o
R
i
R
i
R
i
R
i
R
o
R
o
R
o
R
o
∞ 0
∞ ∞
∞
0
0
0
G. Kovacs ©1997 Page 222
NOTE THAT THE IDEAL INPUT AND OUTPUT RESISTANCES ARE SHOWN.
Chapter 14: FEEDBACK
Type
of
Amplifier
Gain
Expression
Ideal
Input
Impedance
Ideal
Output
Impedance
Voltage
Z
i
= ∞ Z
o
= 0
Transconductance
Z
i
= ∞ Z
o
= ∞
Transresistance
Z
i
= 0 Z
o
= 0
Current
Z
i
= 0 Z
o
= ∞
A
v
=
v
o
v
s
= voltage gain
(dimensionless)
G
m
=
i
o
v
s
= transconductance
in Ω
1
or Siemens
R
m
=
v
o
i
s
= transresistance
in Ω
A
i
=
i
o
i
s
= current gain
(dimensionless)
• The distinction between types of amplifiers is essential because the type of feedback
used with each type is distinct and requires analysis in the “native” units of current or
voltage that is used for input and output.
THIS STUFF IS EASY TO GET CONFUSED ON!!! PLEASE
SPEND A LITTLE TIME TO TRY TO MEMORIZE THE
DIFFERENT AMPLIFIER TYPES AND THE RELEVANT
FEEDBACK TOPOLOGIES...
G. Kovacs ©1997 Page 223
Chapter 14: FEEDBACK
4.2 BASIC STRUCTURE OF THE CIRCUIT
• Here we have assumed that there was an input “comparator” or “mixer” and an output
“sampler” that provided us with a copy of the output signal for use as a feedback signal.
• The form these devices take depends upon whether the amplifier’s input and output
are current or voltage based...
FOOD FOR THOUGHT...
• If x
o
decreases due to negative feedback, won’t βx
o
also decrease, making the output
eventually go to zero????
• NO! This is because subtracting x
f
= βx
o
causes x
i
to decrease slightly so βx
o
decreases
slightly so x
i
is allowed to INCREASE slightly! This acts to stabilize the gain!
G. Kovacs ©1997 Page 224
Mixer or
Comparator
Sampler
Chapter 14: FEEDBACK
4.3 TYPES OF MIXER
+

R
s
v
s
I
s
R
s
Amplifier
Feedback
Network
Amplifier
Feedback
Network
SERIES
MIXER
SHUNT
MIXER
v
f
v
i
i
i
i
f
i
f
+

+ 
4.4 TYPES OF SAMPLER
G. Kovacs ©1997 Page 225
Amplifier
Feedback
Network
Amplifier
Feedback
Network
SHUNT
(VOLTAGE)
SAMPLER
SERIES
(CURRENT)
SAMPLER
v
o
i
o
+

i
o
R
L
R
L
v
o
+

Chapter 14: FEEDBACK
NOTE the effect of feedback on input and output impedance is just a
function of the mixing and sampling type, respectively... e.g. if you put
something in series with the input, its impedance goes up... it goes down
if you put something in shunt (same idea at the output).
• We refer to a given feedback amplifier in terms of the “MIXING  SAMPLING” feedback,
where MIXING and SAMPLING are either SHUNT or SERIES...
• There are four possible types... (for sampling think of how you would measure I or V in
the lab... series for current and shunt for voltage).
INPUTOUTPUT (MIXING  SAMPLING)
SERIESSHUNT series (voltage) mixing, voltagesampling VV
SHUNTSERIES shunt (current) mixing, currentsampling, II
SERIESSERIES series (voltage) mixing, currentsampling VI
SHUNTSHUNT shunt (current) mixing, voltagesampling IV
• Let's consider a familiar example.... the degenerated commonemitter amplifier.
CC
v
s
R
B2
R
C
V
B
V
R
E
R
B1
v
o
i
C
Here we consider i to be the
output of the amplifier (i.e. treat
it as a voltagein, currentout, or
transconductance amplifier)...
C
Feedback network... sample the output
current and produce a feedback voltage
+

v
f
v
i
+

i ≈
E
i
C
G. Kovacs ©1997 Page 226
Chapter 14: FEEDBACK
• Since we sample the output current and generate a voltage feedback signal, this is
a seriesseries feedback topology.
• Considering the output current to be the output signal (e.g. i
o
= i
c
) and the input to be v
s
(for simplicity, assume that R
B1
and R
B2
are very large), the units of the basic amplifier
are,
Overall Gain = A =
i
o
v
s
≡ G
m
in Ω
1
• We know that the feedback voltage is given by Ohm's Law as,
v
f
= i
o
R
E
so the feedback network gain, β = R
E
(NOTE: don't get confused! This ß is NOT the transistor's ß!)
Transistor's current gain =
i
o
v
i
≡ g
m
in Ω
1
• The output current is given by,
i
o
= g
m
v
s
 v
f
= g
m
v
s
 i
o
R
E
• Combining these equations to find the overall gain for the amplifier, G
m
, we end up
with an equation we have seen before!
G
m
=
i
o
v
s
=
g
m
v
s
 i
o
R
E
v
s
= g
m
1 
i
o
v
s
R
E
= g
m
1  G
m
R
E
= g
m
 g
m
G
m
R
E
G
m
1 + g
m
R
E
= g
m
G
m
=
g
m
1 + g
m
R
E
THIS IS g
m
'
THAT WE SAW BEFORE!
G. Kovacs ©1997 Page 227
Chapter 14: FEEDBACK
Note that,
G
m
=
g
m
1 + g
m
R
E
=
A
1 + Aβ
(Again, this ß is not ß for the transistor!)
5: SERIESSHUNT FEEDBACK > VOLTAGE AMP
(SERIES [VOLTAGE] MIXING, VOLTAGESAMPLING)
+

v
s
R
s
Amplifier
Feedback
Network
SERIES
MIXER
v
f
v
i
+

+ 
SHUNT
(VOLTAGE)
SAMPLER
v
o
+

R
L
• An “ideal” example... The noninverting opamp configuration using an ideal opamp
(infinite input impedance, zero output impedance)...
G. Kovacs ©1997 Page 228
Chapter 14: FEEDBACK
+

R
1
R
2
R
s
R
L
V
s
V
o
v
f
+

v
o
+

• Notice how you can redraw the two feedback resistors as a feedback network of the
form we are discussing...
• Note that there are often “implied” ground connections to make the feedback network
a fourterminal device...
• The feedback network gain can be obtained directly by voltage division,
β =
R
1
R
1
+ R
2
• This can be plugged into the feedback gain equation to find the overall gain,
A
fb
≡
x
o
x
s
=
A
1 + A
R
1
R
1
+ R
2
≈
1
R
1
R
1
+ R
2
= 1 +
R
2
R
1
• Continuing with the seriesshunt case, but including the input and output resistance
G. Kovacs ©1997 Page 229
terms (R
i
and R
o
),
Chapter 14: FEEDBACK
+

+

+

+

v
s
SERIES
MIXER
v
f
v
i
+

+ 
SHUNT
(VOLTAGE)
SAMPLER
v
o
+

R
L
v
o
+

β v
o
R
i
R
o
R
if
R
of
v
s
+

v
o
+

A
f
v
s
s
s'
o
o'
s
s'
o
o'
A Circuit
B Circuit
Equivalent Circuit
I
i
I
i
v
i
A
• We can obtain an expression for the equivalent input and output resistance...
R
if
≡
V
s
I
i
=
V
s
V
i
R
i
= R
i
V
s
V
i
= R
i
V
i
+ βAV
i
V
i
= R
i
1 + Aβ
• NOTE that later on, we will generalize this to include IMPEDANCES....
Z
if
S = Z
i
S 1 + A S β S (always true for series mixing)
• The effect of feedback on input resistance or impedance is only a function of the
method of mixing
• The output resistance can also be obtained by the same method we used previously:
• Reduce the input signal (V
s
) to zero and apply a test voltage V
t
at the output....
G. Kovacs ©1997 Page 230
f
V
Chapter 14: FEEDBACK
+

+

SERIES
MIXER
v
f
+

+ 
SHUNT
(VOLTAGE)
SAMPLER
v
t
+

v
o
+

β v
o
R
i
R
o
s
s'
o
o'
A Circuit
B Circuit
I
i
v
i
A
+

I
0
v
i
• Note that you "kill" the input voltage by shorting the input terminals.
• Starting with the definition,
R
of
≡
V
t
I
> I =
V
t
 AV
i
R
o
• Since V
s
= 0,
V
i
=  V
f
=  β V
o
=  β V
t
• Thus,
I =
V
t
 AV
i
R
o
=
V
t
+ Aβ V
t
R
o
• Therefore,
R
of
·
R
o
1+ Aβ
or, more generally, Z
of
·
Z
o
1+ Aβ
G. Kovacs ©1997 Page 231
• Again, this could be written with frequencydependent A and β...
True for all shunt
sampling cases.
Chapter 14: FEEDBACK
6: SHUNTSERIES FEEDBACK > CURRENT AMP
(SHUNT [CURRENT] MIXING, CURRENTSAMPLING)
R
s
I
s
SHUNT
MIXER
i
i
i
f
i
f
SERIES
(CURRENT)
SAMPLER
i
o
i
o
R
L
v
o
+

Amplifier
Feedback
Network
• Here we have series sampling (like connecting an ammeter in series with a circuit in
which you want to know the current flow)...
• Shunt mixing feeds back a βscaled version of the sampled current...
G. Kovacs ©1997 Page 232
• Now look at a practical circuit example, a twostage commonemitter amplifier.
Chapter 14: FEEDBACK
I
s R
s
I
f
Q
1
R
L1
R
F
R
E
R
L2 Q
2
I
o
I
o
α
I
in
• NOTE that the feedback current, I
f
is shown SUBTRACTING from the input current...
This is the notation we will use!
• BE SURE TO “GO AROUND THE LOOP” to verify that the polarity of the feedback
is negative (this means that the loop gain (1 + Aβ) must be positive...)
‘ROUND THE LOOP....
Is R
s
I
f
Q
1
RL1
RF
R
E
R
L2 Q
2
I
o
I
o
α
1) Let I increase...
s
2) Q1's base and collector
currents increase...
3) This voltage
DROPS..
4) I DECREASES...
o
5) DECREASES...
Io
α
6) I INCREASES (in the
direction shown!!!!) [think about V
decreasing... more current can
flow here...]
f
E
CONCLUSION: the feedback is negative since the increase
in I SUBTRACTS from I ......
f s
G. Kovacs ©1997 Page 233
Chapter 14: FEEDBACK
• This sort of “sanity check” tends to be essential....
• Note that here we did not sample the OUTPUT current directly, but rather the nearly
equal emitter current of Q
2
... This is quite reasonable, and we will do this in the next
example too....
7: SERIESSERIES FEEDBACK >
TRANSCONDUCTANCE AMPLIFIER
VOLTAGEIN, CURRENTOUT
(SERIES [VOLTAGE] MIXING, CURRENTSAMPLING)
+

v
s
R
s
Amplifier
Feedback
Network
SERIES
MIXER
v
f
v
i
+

+ 
SERIES
(CURRENT)
SAMPLER
i
o
i
o
R
L
v
o
+

• This type of circuit is used when you need to generate an output current proportional
to a command voltage, such as when you need to drive the deflection coils in a
television...
G. Kovacs ©1997 Page 234
Chapter 14: FEEDBACK
+

I
o
α
I
o
V
s
V
f
R
s
R
E1
R
F
R
E2
R
L3
Q
1
Q
2
Q
3
R
L2
R
L1
+

SERIES
SERIES
Two approximations...
1) Emitter current ≈ output current.
2) The series mixer is through the Q1
EB junction rather than direct....
G. Kovacs ©1997 Page 235
Chapter 14: FEEDBACK
8: SHUNTSHUNT FEEDBACK >
TRANSRESISTANCE AMPLIFIER
CURRENTIN, VOLTAGEOUT
(SHUNT [CURRENT] MIXING, VOLTAGESAMPLING)
R
s
I
s
SHUNT
(CURRENT)
MIXER
i
i
i
f
i
f
Amplifier
Feedback
Network
SHUNT
(VOLTAGE)
SAMPLER
v
o
+

R
L
• If you redraw the inverting opamp configuration with a Norton input source, you can
treat the circuit as a transresistance amplifier....
+

R
S
R
F
R
S
R
F
I
S
=
V
S
R
S
V
S
V
o
V
o

+

+
I
F
SHUNT
SHUNT
• ‘Round the loop... If I
S
increases, V
o
decreases, allowing I
F
to increase. I
F
subtracts
G. Kovacs ©1997 Page 236
from I
S
, so an increased I
F
means negative feedback (as you expected, I hope!)....
Chapter 14: FEEDBACK
9: FEEDBACK ANALYSIS IN REALISTIC CIRCUITS
• In real situations, the feedback network is generally passive, and loads the basic
amplifier in the loop and thus affects A, R
i
and R
o
....
• The source and load resistances will also affect these three parameters...
• The trick is to lump these “bad” effects into the previously mentioned “A” circuits (and
sometimes “β” circuits, but we lump it all into "A" here)....
• These “lumped” circuits can be plugged into the ideal feedback circuit case and used
to determine the properties of the closedloop case....
• For each of the four feedback topologies, there is a set of rules about how to
incorporate the loading effects and how to compute β.
9.1 SUMMARY OF STEPS YOU WILL USE
1) Redraw the amplifier as an "A" circuit which takes into account the fact that
introducing the feedback network actually affects the openloop gain of the amplifier!
Also lump R
S
and R
L
into the "A" circuit. The fact that this loading occurs means that
you can't just use,
A
f
=
A
1 + Aβ
until you compute a new "A" value that takes into account the loading!!!
2) After determining the "A" circuit, find the value for β.
3) Now use the corrected "A" gain (call it A') to compute A
f
...
A
f
=
A'
1 + A' β
4) Now, if you need to, you can compute the input and output impedances (with
feedback) using the appropriate (1 + A'β) scaling (how to do this is explained below).
5) Also, if you need to, you can compute the input and output impedances of the
amplifier (with feedback) looking into its terminals but without R
S
and R
L
(see below).
You simply mathematically remove the effects of R
S
and R
L
. Why not just start without
them? The reason is that you have to take into account their loading effects on the
G. Kovacs ©1997 Page 237
amplifier BEFORE doing the feedback calculations.
Chapter 14: FEEDBACK
10: SERIESSHUNT FEEDBACK
(SERIES [VOLTAGE] MIXING, VOLTAGESAMPLING)
+

+

+

+

v
s
SERIES
MIXER
v
f
v
i
+

+ 
SHUNT
(VOLTAGE)
SAMPLER
v
o
+

R
L
v
o
+

β v
o
R
i
R
o
R
if
R
of
v
s
+

v
o
+

A
f
v
s
s
s'
o
o'
s
s'
o
o'
A Circuit
ß Circuit
Equivalent Circuit
I
i
I
i
v
i
A
NOTE that this
is an "ideal" case
so no R is included!
S
• The above equivalent circuit is the "goal," incorporating all of the
effects of loading and feedback into a final equivalent.
• We will learn a general set of rules (one for each of the four types of
feedback configurations) to "pull" the effects of R
S
, R
L
and ßnetwork
loading into the a new "A" circuit and then apply the basic feedback
equation with the ßnetwork taken to be "ideal."
G. Kovacs ©1997 Page 238
Chapter 14: FEEDBACK
• Determine the loading effects on input and output by “destroying” the feedback from
each end of the feedback network to the other...
SHORT SHUNTS
and
SEVER SERIES
RULES FOR SERIESSHUNT
G. Kovacs ©1997 Page 239
+

+

BASIC
AMPLIFIER
FEEDBACK
NETWORK
FEEDBACK
NETWORK
FEEDBACK
NETWORK
R
S
R
11
V
i
'
R
22
R
L
V
o
'
+

R
11
R
22
1 2
1 2
1 2
+

V
f
'
I
1
V
o
'
A' ≡
V
o
'
V
i
'
β ≡
V
f
'
V
o
'
I
1
= 0
SHORT
the
SHUNT
SEVER
the
SERIES
This model for the amplifier now takes
into account loading from R and R
s L
Here we just want "pure" feedback (no worrying about loading anymore).
Chapter 14: FEEDBACK
LET’S TRY IT!!!! SEDRA & SMITH EXERCISE 8.5
+

V
s
Q1 Q2
Q3
+ V
cc
= 10.7 V
+ V
cc
 V
EE
 V
EE
R
1
= 1 KΩ
R
2
= 9 KΩ
R
C
= 20 KΩ
R
S
= 10 KΩ
1 mA
5 mA
R
L
= 2 KΩ
V
o
R
if
'
R
of
'
SERIES
MIXING
SHUNT
SAMPLING
You already know that the
output voltage of the differential
pair is the difference between the
inputs!!! As connected, Q2 is the
inverting input!
• First look at the overall design...
• This is a differential amplifier feeding an emitter follower stage at the output.
• Q
1
’s collector is tied directly to V
CC
(no collector resistor) to eliminate its Miller capacitance
multiplication.
• Look at the circuit and convince yourself that it is really SERIESSHUNT feedback...
(You were supposed to read section 7.11 before... If you haven’t, please do so...)
• DC Voltages.... (β = 100 for all three transistors)
I
E1
= I
E2
= 0.5 mA
V
C2
≈ 10.7  0.5 mA X 20 KΩ = + 0.7 V > V
o
= 0.7  V
BE3
≈ 0 V
G. Kovacs ©1997 Page 240
Chapter 14: FEEDBACK
+

v
s
Q1 Q2
+ V
cc
R
C
+ V
cc
V
o
Later on we can put a
resistance here, and take
it into account...
Later on we can put a
resistance here, and take
it into account...
I
• For the modified (no collector resistor on Q1) differential pair (shown above in simplified
form), the voltage gain (to the collector of Q
2
) can be quickly determined...
• The circuit is basically a common collector stage driving a common base stage!
• For the common collector stage, we know that the output voltage (at the emitter) is
given by,
v
e
= v
b
R
e
R
e
+ r
e
Re
where R
e
is the total resistance in the emitter circuit (effectively the load of the CC
amplifier!) and r
e
is the emitter resistance of the transistor in the CC amplifier...
• In this case, R
e
= r
e2
, the emitter resistance of Q
2
... Therefore, the voltage feeding into
the common base stage is just split in half (surprise, surprise!!!),
v
e1
= v
s
r
e2
r
e2
+ r
e1
=
v
s
2
since r
e2
= r
e1
G. Kovacs ©1997 Page 241
Chapter 14: FEEDBACK
• The current into the emitter of Q
2
is,
i
e2
=
v
e1
r
e
=
v
s
2
1
r
e
=
v
s
2r
e
• Therefore, the collector current of Q
2
is,
i
c2
= α
v
s
2r
e
≈
v
s
2r
e
• Finally giving a voltage gain of,
A =
v
C2
v
s
=
i
c2
R
C
v
s
=
R
C
2r
e
• If we load the input stage with an output stage (Q
3
) we need to include that load
resistance into a “new” R
C
...
• We also know that the input resistance of the CC amplifier is large,
R
iCC
= r
π
+ β + 1 R
e
• But in this case it is only about 10 kΩ, so later on we will take into account the effects
of R
S
... (you may more frequently see a smaller R
S
anyway...)
NOW APPLY THE SERIESSHUNT RULES TO FIND THE
“A” CIRCUIT:
+

BASIC
AMPLIFIER
R
S
R
11
V
i
'
R22
R
L
Vo
'
+

• Again, this is the amplifier with all loading effects "lumped" into it.
1) Input loading > short the shunt at the output and determine the resistance
seen looking into the INPUT side of the feedback network...
G. Kovacs ©1997 Page 242
FEEDBACK
NETWORK
R
11
1 2
Chapter 14: FEEDBACK
R1 = 1 KΩ
R
2
= 9 KΩ
R
11
R
11
= 1KΩ  9KΩ
2) Output loading > open circuit (sever) the series connection at the input and
determine the resistance seen looking into the OUTPUT side of the feedback network....
FEEDBACK
NETWORK
R
22
1 2
R1 = 1 KΩ
R
2
= 9 KΩ
R
22
R
22
= 1KΩ + 9KΩ
G. Kovacs ©1997 Page 243
Chapter 14: FEEDBACK
3) Draw the “A” circuit and determine its gain....
 V
EE
Q1 Q2
+ V
cc
= 10.7 V
+ V
cc
 V
EE
R
C
= 20 KΩ
1 mA
R
S
= 10 KΩ
9 KΩ
1 KΩ
1 KΩ
9 KΩ
Q3
5 mA
R
L
= 2 KΩ
V
i
'
+

V
o
'
• Modifying our previously determined gain equations to take into account the base
resistance of Q
2
and the loading of Q
3
,
A' =
V
o
'
V
i
'
=
R
C
 loading from Q
3
circuit
r
e1
+ r
e2
+ other loading in the CC emitter circuit
Gain of the Q
3
stage
• Plugging it all together and using (β + 1) scaling where necessary,
A' =
V
o
'
V
i
'
=
R
C
 r
π3
+ β
3
+ 1 1KΩ + 9KΩ  R
L
r
e1
+ r
e2
+
R
S
β
1
+ 1
+
1KΩ  9KΩ
β
2
+ 1
1KΩ + 9KΩ  R
L
1KΩ + 9KΩ  R
L
+ r
e3
• Using the values for r
π
and r
e
of,
r
π
=
β
g
m
=
β V
T
I
C
> r
π3
≈ 520 Ω
G. Kovacs ©1997 Page 244
this stuff is the load on the CC
amplifier formed by Q1....
Chapter 14: FEEDBACK
r
e1
= r
e2
=
V
T
I
C
≈ 52 Ω
r
e3
=
V
T
I
C3
≈ 5.2 Ω
• The result is,
A' =
V
o
'
V
i
'
= 84.4 0.997 = 84.1
4) Now find the “β” circuit.... (simple voltage divider)
R
1
= 1 KΩ
R
2
= 9 KΩ
V
+

V
+

f o
' '
β ≡
V
f
'
V
o
'
=
R
1
R
1
+ R
2
=
1 KΩ
1 KΩ + 9 KΩ
= 0.1
4) Compute the closedloop gain...
A
f
=
V
o
'
V
s
'
=
A'
1 + A' β
=
84.1
1 + 84.1 0.1
= 8.94
• Compare this to the "ideal" noninverting opamp configuration... the gain would be,
G. Kovacs ©1997 Page 245
A
V
· 1+
R
f
R
g
¸
¸
_
,
· 1+
9kΩ
1kΩ
¸
¸
_
,
·10
Chapter 14: FEEDBACK
5) To get the input and output resistances, you need to compute
them for the “A” circuit and then apply the feedback equations that
you know for the seriesshunt case... KEY: you FIRST have to take
into account loading and go through the feedback stuff to compute
the input and output resistances!
 V
EE
Q1 Q2
+ V
cc
= 10.7 V
+ V
cc
 V
EE
R
C
= 20 KΩ
1 mA
R
S = 10 KΩ
9 KΩ
1 KΩ
1 KΩ
9 KΩ
Q3
5 mA
R
L
= 2 KΩ
V
i
'
+

V
o
'
R
if
'
R
of
'
R
i
R
o
R
i
= 10 KΩ + r
π1
+ β
1
+ 1 r
e2
+ β
1
+ 1
1 KΩ  9 KΩ
β
2
+ 1
• Which gives (after computing that r
π1
≈ 5.2 kΩ) ,
R
i
= 10 KΩ + 5.2 KΩ + 101 52 + 101
1 KΩ  9 KΩ
101
= 21.4 KΩ
• Applying the appropriate feedback equation,
R
if
= R
i
1 + A' β = 21.4 KΩ 9.41 = 201 KΩ
G. Kovacs ©1997 Page 246
looking at the 1KΩ  9KΩ resistance
through the base of Q1 and through
the emitter of Q2
looking at Q2's emitter
resistance through the
base of Q1
Chapter 14: FEEDBACK
• The actual input resistance of the amplifier (after R
S
) can be determined by subtracting
R
S
...
R
if
'
= R
if
 R
S
= 201 KΩ  10 KΩ = 191 KΩ
• Similarly, we start with the output resistance of the “A” circuit,
R
o
= R
L
 1 KΩ + 9 KΩ  r
e3
+
R
C
β
3
+ 1
which gives,
R
o
= 2 KΩ  1 KΩ + 9 KΩ  5.2 Ω +
20 KΩ
101
= 181 Ω
• Applying the appropriate feedback equation,
R
of
=
R
o
1 + Aβ
=
181 Ω
9.41
= 19.24 Ω
• Finally, to get the output resistance of the amplifier ALONE (R
of
’), we need to remove
the effects of R
L
in parallel with it...
1
R
of
'
+
1
R
L
=
1
R
of
> R
of
'
=
1
1
R
of

1
R
L
• This gives,
R
of
'
=
1
1
19.24 Ω

1
2 KΩ
= 19.42 Ω
G. Kovacs ©1997 Page 247
looking at Rc through
the emitter of Q3...
Chapter 14: FEEDBACK
11: RULES FOR SERIESSERIES
(TRANSCONDUCTANCE AMPLIFIER)
+

BASIC
AMPLIFIER
FEEDBACK
NETWORK
FEEDBACK
NETWORK
R
S
R
11
V
i
'
R
11
R
22
+

V
f
'
1 2
1 2
1 2
SEVER
the
SERIES
SEVER
the
SERIES
I
1
= 0
I
o
'
I
o
'
R
22
R
L
A ≡
I
o
'
V
i
'
FEEDBACK
NETWORK
β ≡
V
f
'
I
o
'
I
1
= 0
G. Kovacs ©1997 Page 248
Chapter 14: FEEDBACK
12: RULES FOR SHUNTSHUNT (TRANSRESISTANCE
AMPLIFIER)
+

BASIC
AMPLIFIER
FEEDBACK
NETWORK
FEEDBACK
NETWORK
FEEDBACK
NETWORK
R
22
R
L
V
o
'
+

R
11
R
22
1 2
1 2
1 2
V
o
'
SHORT
the
SHUNT
SHORT
the
SHUNT
I
i
'
R
11
R
S
V
1
= 0
+

I
f
'
A ≡
V
o
'
I
i
'
β ≡
I
f
'
V
o
'
V
1
= 0
G. Kovacs ©1997 Page 249
Chapter 14: FEEDBACK
13: RULES FOR SHUNTSERIES (CURRENT
AMPLIFIER)
BASIC
AMPLIFIER
FEEDBACK
NETWORK
R
11
R
22
1 2
1 2
1 2
SHORT
the
SHUNT
I
i
'
R
11
R
S
V
1
= 0
+

I
f
'
FEEDBACK
NETWORK
FEEDBACK
NETWORK
SEVER
the
SERIES
I
o
'
I
o
'
R
22
R
L
β ≡
I
f
'
I
o
'
A ≡
I
o
'
I
i
'
V
1
= 0
G. Kovacs ©1997 Page 250
Chapter 14: FEEDBACK
14: RULES FOR SERIESSHUNT (VOLTAGE
AMPLIFIER)
(repeated for convenience)
+

+

BASIC
AMPLIFIER
FEEDBACK
NETWORK
FEEDBACK
NETWORK
FEEDBACK
NETWORK
R
S
R
11
V
i
'
R
22
R
L
V
o
'
+

R
11
R
22
1 2
1 2
1 2
+

V
f
'
I
1
V
o
'
A' ≡
V
o
'
V
i
'
β ≡
V
f
'
V
o
'
I
1
= 0
SHORT
the
SHUNT
SEVER
the
SERIES
This model for the amplifier now takes
into account loading from R and R
s L
Here we just want "pure" feedback (no worrying about loading anymore).
G. Kovacs ©1997 Page 251
Chapter 14: FEEDBACK
15: STABILITY AND POLE LOCATION
σ = 0.05 Damped oscillations.
σ = 0 An oscillator.
σ = +0.05 Positive feedback out of control!
G. Kovacs ©1997 Page 252
Chapter 14: FEEDBACK
15.1 SINGLE POLE WITH FEEDBACK
• Remember the response of a singlepole feedback amplifier that we previously looked
at can be derived from the response of the amplifier without feedback, given by,
A s =
A
M
1 +
s
ω
H
(Sedra & Smith use A
M
in section 8.2 and Ao in section 8.9 > A
M
= A
o
for a DCcoupled amplifier.)
• If you use the amplifier with negative feedback, the gain becomes,
A
f
s =
A s
1 + βA s
• Substituting, you get,
A
f
s ( ) ·
A
M
1+ A
M
β ( )
1+
s
ω
H
1+ A
M
β ( )
·
DC gain
1+
s
Cut Off frequency
• Thus we have another singlepole response, but with a high cutoff frequency given
by,
ω
Hf
= ω
H
1 + A
M
β
• This means that the pole is sliding along the real axis from its original distance
G. Kovacs ©1997 Page 253
(frequency) ω
H
to its new one, ω
Hf
....
Chapter 14: FEEDBACK
ωHf = ωH 1 + AMβ
ωH
Pole moves this way
as you increase β
NOTE that the new pole is given by 0 = 1 + A S β S .
• "Single pole" opamps are always stable, but real opamps always have more
than one pole!
• Meanwhile, the frequency response is changing in accord with the gainbandwidth
product....
β = 1
β = 0.1
β = 0.01
β = 0.001
β = 0.0001
β = 0.00001
β = 0
G. Kovacs ©1997 Page 254
Chapter 14: FEEDBACK
15.2 TWO POLES WITH FEEDBACK
• For a twopole basic amplifier, its frequency response can be written as,
A S =
A
M
1 +
S
ω
p1
1 +
S
ω
p2
• The closedloop poles are obtained from 0 = 1 + A S β S , which can be written,
S
2
+ S ω
p1
+ ω
p2
+ 1 + A
M
β ω
p1
ω
p2
= 0
• With the closedloop poles given by,
S = 
1
2
ω
p1
+ ω
p2
t
1
2
ω
p1
+ ω
p2
2
 4 1 + A
M
β ω
p1
ω
p2
• As ß is increased, the two poles move together on the Splane until they overlap, and
then they separate along a line,
σ =
ω
p1
+ ω
p2
2
G. Kovacs ©1997 Page 255
Chapter 14: FEEDBACK
 ω
p1
 ω
p2
σ =
ω
p1
+ ω
p2
2
• For a secondorder system such as this, the poles never enter the right halfplane (the
maximum phase shift is 180°, but it is not reached until ω = ∞ ).
• So, such an amplifier is UNCONDITIONALLY STABLE!
G. Kovacs ©1997 Page 256
Chapter 14: FEEDBACK
15.3 THREE OR MORE POLES WITH FEEDBACK
• For a basic amplifier with three poles, there is always some value of Aβ that makes a
pair of poles enter the RHP! This makes sense for any number of poles greater than
two since there is a phase shift of at least 270 °, so that 180°is reached at a finite
frequency....
• Two of the poles become coincident, complex and conjugate.
• There is a value of β at which these two poles enter the right halfplane, causing the
amplifier to become unstable.
• Because there is a value of β for which the amplifier becomes unstable, one can look
G. Kovacs ©1997 Page 257
at it as a maximum value of β or a minimum gain that guarantees stability.
Chapter 14: FEEDBACK
15.4 STABILITY
• Loop gain = Aβ is a function of frequency...
• Look at the basic feedback equation,
A
f
s =
A s
1 + βA s
• If  Aβ  > 1 (i.e. there is still some GAIN left) when – Aβ =  180°the feedback will
become POSITIVE!!!
• This leads to instability (i.e. oscillation).
• Gain Margin is defined as the difference between the value of  Aβ  at the 180°phase
frequency (ω
180
) and unity....
Gain Margin ≡ gain at ω
180
 gain at ω
unity gain
= gain at ω
180
 1
• Phase Margin is defined as the difference between the phase angle at the unitygain
frequency and at the frequency where the phase reaches 180°...
Phase Margin ≡ phase at ω
180
+ phase at ω
unity gain
= 180 + phase at ω
unity gain
G. Kovacs ©1997 Page 258
a negative number usually!
Chapter 14: FEEDBACK
ω (log scale)
ω (log scale)
0
0
0°
ω
180
ω
unity gain
ω
180
ω
unity gain
+90°
90°
180°
Gain Margin
Phase Margin
360°
270°
(Figure adapted from A. S. Sedra and K. C. Smith, Microelectronic Circuits, HRW Saunders, 1991)
16: COMPENSATION
• Real opamps have an openloop gain rolloff with frequency that is approximately
firstorder (20 dB/decade) over much of their useful bandwidth.
• The major internal capacitance that causes this rolloff is often referred to as the
“dominant pole” of the amplifier.
• At higher and higher frequencies, other capacitance effects come into play as additional
poles (sometimes there are three or more).
• This means that the openloop phase response of the amplifier will eventually reach
G. Kovacs ©1997 Page 259
90°times the number of poles....
Chapter 14: FEEDBACK
• If the phase is less than 180°when the gain of the amplifier reaches a gain of unity (0
dB), everything remains stable.
• However, if the phase crosses 180° before the gain falls to unity, oscillations
will probably occur (since an inverted replica of the amplifier’s output is fed back
into it)! If an opamp circuit is unstable, almost any noise present in the circuit
will have enough of a highfrequency component to cause the circuit to oscillate.
• If the phase crosses 180° at a frequency where the gain of the
amplifier is less than unity, the amplifier is “unconditionally stable!”
• The most common way to guarantee stability is to compensate the amplifier with some
additional components that shape its frequency response so that its gain is less than
unity by the time the phase hits 180°. This does, however, compromise the highfrequency
response of the amplifier!
• Most opamps available are “internally compensated,” which means that the
component(s) required to guarantee stability are included on the chip itself. Some of
the older opamps, and those designed for highspeed operation, are “externally
compensated,” which means that you, the designer, must choose external components
to assure that the amplifier will not become unstable.
G. Kovacs ©1997 Page 260
Chapter 14: FEEDBACK
17: "TENT" MODEL FOR VISUALIZING POLES AND
ZEROS
G. Kovacs ©1997 Page 261
Chapter 14: FEEDBACK
18: LINEAR OSCILLATORS (VERY BRIEFLY!)
• A linear oscillator ideally produces a pure sinusoidal output at a single frequency
(hopefully).
• To achieve linear oscillation, a linear amplifier must oscillate without external stimuli
(other than a startup transient to get it going, perhaps).
• In order to understand this type of oscillator, a minor excursion into theory will be
required (it’s worth it, since a little bit of intuitive understanding goes a long way!).
• What is required to make a linear oscillator (that works, that is!) is the arrangement
shown below (this is just POSITIVE feedback)...
• A linear oscillator of gain A provides an output voltage v
out
that is fed through a
feedback loop with gain β, and summed back into the input of the amplifier. The overall
gain of the circuit with feedback, A
f
(S), is given by,
A
f
S =
A S
1  A S β S
• Without going into all of the details, but by examining the denominator of the above
equation, it is easy to see that the overall gain can be made infinity by setting the
“roundtrip” gain around the entire feedback loop so that,
A S β S = 1
• This condition, known to those who care as the Barkhausen Criterion, appears to
make the circuit blow up!
G. Kovacs ©1997 Page 262
Chapter 14: FEEDBACK
• Actually, it is a necessary condition for this type of oscillator (linear) to work.
• Intuitively, however the fact that the overall gain is infinity means that the output of the
circuit is some signal (to be determined!), even with NO input at all!
• If one can arrange it so that the Barkhausen Criterion is met at only a single frequency,
it is possible to obtain a very pure sinewave output (if it is met at multiple frequencies,
you might get an interesting mix of frequencies).
• The classic Wein Bridge Oscillator achieves this through a combination of negative
and positive feedback...
C
C
R
R
R1
R2
V
out
• The oscillation frequency is given by,
f
osc
=
1
2πRC
• The details of the circuit (i.e. actually getting it to start oscillating, keeping it stable,
G. Kovacs ©1997 Page 263
etc.) are covered in EE122 and many good EE reference books...
TABLE OF CONTENTS
Chapter 1: INTRODUCTION ..........................................................................................1 1: OBJECTIVES.........................................................................................................1 2: EE113 AND EE122  A WINNING COMBINATION...............................................1 3: COURSE INFORMATION......................................................................................2 4: OUTLINE OF THE COURSE.................................................................................3 Chapter 2: BASIC CONCEPTS......................................................................................5 1: OBJECTIVES.........................................................................................................5 2: SOURCES .............................................................................................................5 3: SIGNALS ...............................................................................................................6 4: AMPLIFIERS..........................................................................................................8 5: NOISE.................................................................................................................. 11 6: DISTORTION....................................................................................................... 12 7: AMPLIFIER POWER SUPPLIES & EFFICIENCY ............................................... 14 8: LARGE AND SMALL SIGNALS ..........................................................................15 9: TRANSFER FUNCTIONS.................................................................................... 16 10: BODE PLOTS .................................................................................................... 20 Chapter 3: BJT REVIEW..............................................................................................25 1: OBJECTIVES.......................................................................................................25 2: TYPES OF BIPOLAR JUNCTION TRANSISTORS............................................. 25 3: DC ANALYSIS (LARGESIGNAL) .......................................................................27 4: ANALYSIS OF BIASING ..................................................................................... 29 5: DC LOAD LINES..................................................................................................31 6: SMALLSIGNAL BEHAVIOR ...............................................................................32 7: HYBRIDπ MODEL FOR (AC) SMALL SIGNALS ................................................ 33 8: SCALING RESISTANCES BETWEEN BASE AND EMITTER ............................ 35 9: AC LOAD LINES ..................................................................................................36 Chapter 4: Operational Amplifiers ............................................................................. 38 1: OBJECTIVES.......................................................................................................38 2: OPERATIONAL AMPLIFIERS CONCEPTS ........................................................ 39 3: BASIC OPAMP CIRCUITS ................................................................................. 40 3.1 VOLTAGE FOLLOWER................................................................................. 40 3.2 INVERTING AMPLIFIER ............................................................................... 41 3.3 LOGARITHMIC AMPLIFIER..........................................................................42 3.4 EXPONENTIAL AMPLIFIER.......................................................................... 42 3.5 NONINVERTING AMPLIFIER...................................................................... 43 3.6 SUMMING AMPLIFIER ................................................................................. 43 3.7 INTEGRATOR ............................................................................................... 44 3.8 DIFFERENTIATOR........................................................................................ 45 4: “REAL” VERSUS “IDEAL” OPAMPS .................................................................. 46 5: FREQUENCY RESPONSE OF OPAMPS.......................................................... 47 6: THE GUTS OF REAL OPAMP CHIPS ............................................................... 50 7: A PREVIEW OF FEEDBACK .............................................................................. 53
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER .................................................................................................................... 54 1: OBJECTIVES.......................................................................................................54 2: OVERVIEW OF SINGLESTAGE BJT AMPLIFIERS .......................................... 54 3: THE COMMONEMITTER AMPLIFIER ............................................................... 56 4: CLASSIC BIASING SCHEME FOR CE AMPLIFIERS.........................................59 5: BASE CIRCUIT DESIGN STRATEGY  THE BASICS ........................................ 59 6: LOCAL FEEDBACK THROUGH AN EMITTER RESISTOR................................. 60 7: BASE CIRCUIT ANALYSIS STRATEGY............................................................. 61 8: UNBYPASSED EMITTER RESISTANCE FOR AC AND DC FEEDBACK......... 62 9: EMITTER DEGENERATION................................................................................ 66 10: THE BOTTOM LINE ON DESIGN OF CE AMPLIFIERS ...................................71 11: THE ART OF CE DESIGN (AN INTRO) ............................................................ 72 12: BIAS CIRCUITRY DESIGN................................................................................ 74 13: REVISITING THE LOAD LINES ........................................................................ 75 14: DESIGN EXAMPLE #1: METHOD OF SEDRA & SMITH................................. 79 Chapter 6: BJT FREQUENCY RESPONSE ................................................................84 1: OBJECTIVES.......................................................................................................84 2: JUNCTION CAPACITANCES OF BJT’S .............................................................84 3: DIFFUSION CAPACITANCE ............................................................................... 87 4: COMPLETE HYBRIDp MODEL..........................................................................88 5: GAIN (β) VERSUS FREQUENCY FOR BJT'S .................................................... 89 6: COMMONEMITTER CONFIGURATION AND MILLER CAPACITANCE ...........93 7: USING THE MILLER IDEA UP FRONT............................................................... 95 Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE ................................. 97 1: OBJECTIVES.......................................................................................................97 2: GENERALIZED CAPACITORCOUPLED AMPLIFIER FREQUENCY RESPONSE .............................................................................................................................98 3: DOMINANT POLES............................................................................................. 99 4: SHORTCIRCUIT AND OPENCIRCUIT TIME CONSTANT METHODS FOR ...... APPROXIMATING THE RESPONSE OF AMPLIFIERS......................................... 100 4.1 OPENCIRCUIT TIME CONSTANTS FOR UPPER CUTOFF FREQUENCY APPROXIMATION.............................................................................................101 4.2 SHORTCIRCUIT TIME CONSTANTS FOR LOWER CUTOFF FREQUENCY APPROXIMATION.............................................................................................101 5: FREQUENCY RESPONSE OF THE CE AMPLIFIER ......................................102 5.1 LOW FREQUENCY RESPONSE................................................................ 103 5.2 MIDFREQUENCY RESPONSE .................................................................108 5.3 HIGHFREQUENCY RESPONSE ...............................................................109 6: EXAMPLE FREQUENCY RESPONSE ANALYSIS........................................... 112 6.1 MIDBAND GAIN CALCULATION ...............................................................112 6.2 LOWFREQUENCY RESPONSE................................................................ 113 Chapter 8: THE COMMON BASE AMPLIFIER .........................................................117 1: OBJECTIVES.....................................................................................................117 2: THE COMMON BASE CIRCUIT ........................................................................118 3: MIDBAND GAIN CALCULATIONS................................................................... 118 4: LOWFREQUENCY RESPONSE ......................................................................121
5: HIGHFREQUENCY RESPONSE .....................................................................122 6: CASCODE AMPLIFIERS = CE + CB.................................................................126 Chapter 9: THE COMMON COLLECTOR AMPLIFIER .............................................127 1: OBJECTIVES.....................................................................................................127 2: COMMON COLLECTOR CIRCUIT....................................................................128 3: PRACTICAL CC CIRCUIT .................................................................................130 4: THE PHASE SPLITTER CIRCUIT .....................................................................132 5: QUICK LOOK AT CC FREQUENCY RESPONSE ............................................133 6: DESIGN OF CC STAGES..................................................................................134 Chapter 10: CASCADED AND CASCODE AMPLIFIERS.........................................135 1: OBJECTIVES.....................................................................................................135 2: AMPLIFIERS IN SERIES (CASCADED)............................................................136 3: THE CASCODE AMPLIFIER .............................................................................138 4: PRACTICAL CASCODE AMPLIFIER CIRCUIT................................................. 140 5: CASCODE EXAMPLE .......................................................................................146 6: CASCODE AMPLIFIER DESIGN EXAMPLE #2 ...............................................148 Chapter 11: DIFFERENTIAL AMPLIFIERS ...............................................................153 1: OBJECTIVES.....................................................................................................153 2: DIFFERENTIAL AMPLIFIER BASIC CONCEPTS.............................................154 3: MODES OF OPERATION OF THE DIFFERENTIAL PAIR................................ 155 3.1 COMMONMODE OPERATION..................................................................156 3.2 “LARGE” SIGNAL DIFFERENTIALMODE OPERATION ........................... 157 3.3 SMALLSIGNAL DIFFERENTIALMODE OPERATION..............................158 4: DETAILS OF SMALLSIGNAL OPERATION.....................................................159 5: HALFCIRCUIT MODEL OF DIFFERENTIAL PAIR ..........................................166 6: HALFCIRCUIT MODEL AND COMMONMODE OPERATION........................168 7: COMMONMODE REJECTION RATIO............................................................. 170 8: FREQUENCY RESPONSE OF THE DIFFERENTIAL AMPLIFIER................... 171 9: SUMMARY OF DIFFERENTIAL PAIR SMALLSIGNAL OPERATION ............. 182 10: OTHER IMPERFECTIONS OF “REALISTIC” DIFFERENTIAL AMPLIFIERS. 183 10.1 INPUT OFFSET VOLTAGE...................................................................... 183 10.2 INPUT BIAS AND OFFSET CURRENTS................................................. 186 Chapter 12: CURRENT SOURCES ...........................................................................188 1: OBJECTIVES.....................................................................................................188 2: THE DIODECONNECTED TRANSISTOR ....................................................... 189 3: THE CURRENT MIRROR.................................................................................. 190 4: A SIMPLE (LOUSY) CURRENT SOURCE........................................................ 191 5: THE WIDLAR CURRENT SOURCE ..................................................................192 6: CURRENT MIRRORS........................................................................................193 7: NONIDEALITIES OF BJT CURRENT SOURCES............................................194 7.1 EFFECTS OF ro ON CURRENT SOURCE PERFORMANCE..................... 195 Chapter 13: MULTISTAGE AMPLIFIERS ................................................................ 197 1: OBJECTIVE .......................................................................................................197 2: ANALYSIS OF AN EXAMPLE AMPLIFIER........................................................197
Chapter 14: FEEDBACK............................................................................................203 1: OBJECTIVES.....................................................................................................203 2: INTRODUCTION TO FEEDBACK .....................................................................204 2.1 PROPERTIES OF NEGATIVE FEEDBACK............................................... 204 2.2 THE BASIC FEEDBACK CIRCUIT..............................................................205 3: HOW FEEDBACK AFFECTS BANDWIDTH......................................................207 4: FROM BASIC BLOCK DIAGRAM TO ACTUAL FEEDBACK CIRCUITS .......... 209 4.1 REMINDER: TYPES OF AMPLIFIERS ......................................................209 4.2 BASIC STRUCTURE OF THE CIRCUIT.................................................... 211 4.3 TYPES OF MIXER......................................................................................212 4.4 TYPES OF SAMPLER................................................................................212 5: SERIESSHUNT FEEDBACK > VOLTAGE AMP.............................................215 6: SHUNTSERIES FEEDBACK > CURRENT AMP ............................................ 219 7: SERIESSERIES FEEDBACK > TRANSCONDUCTANCE AMPLIFIER.......... 221 8: SHUNTSHUNT FEEDBACK > TRANSRESISTANCE AMPLIFIER ................ 223 9: FEEDBACK ANALYSIS IN REALISTIC CIRCUITS........................................... 224 9.1 SUMMARY OF STEPS YOU WILL USE.................................................... 224 10: SERIESSHUNT FEEDBACK.......................................................................... 225 11: RULES FOR SERIESSERIES (TRANSCONDUCTANCE AMPLIFIER) ........235 12: RULES FOR SHUNTSHUNT (TRANSRESISTANCE AMPLIFIER) ...............236 13: RULES FOR SHUNTSERIES (CURRENT AMPLIFIER)................................237 14: RULES FOR SERIESSHUNT (VOLTAGE AMPLIFIER) ................................238 15: STABILITY AND POLE LOCATION.................................................................239 15.1 SINGLE POLE WITH FEEDBACK ...........................................................240 15.2 TWO POLES WITH FEEDBACK..............................................................241 15.3 THREE OR MORE POLES WITH FEEDBACK........................................ 243 15.4 STABILITY................................................................................................244 16: COMPENSATION............................................................................................ 245 17: "TENT" MODEL FOR VISUALIZING POLES AND ZEROS ............................ 246 18: LINEAR OSCILLATORS (VERY BRIEFLY!).................................................... 247
2: EE113 AND EE122 . Shimbo. • Learn about differential pairs. • Find out how bipolar transistors really work in circuits. • Learn how to design.. Kovacs ©1997 Page 1 . • Learn how to design. • Learn about feedback as it applies to amplifiers. A potential problem arises if you don’t keep up: you can be lost in two classes at once! PLEASE DON’T GET LEFT BEHIND!!! IF YOU FEEL THIS HAPPENING.. Barden N. and there’s nothing more minimal than nothing. analyze and test multistage amplifiers.Chapter 1: INTRODUCTION Chapter 1: INTRODUCTION Minimalism is in. In general. and overall topics will be timed so that you will have the background knowledge to cope with the laboratories. • Learn about operational amplifiers and circuits that use them.A WINNING COMBINATION To the extent possible. The point is that (finally) you get to apply what you learn in the classroom to handson designing and testing of circuits. problem sets. and test basic amplifiers. current sources and multistage amplifier design. BUT YOU WILL NEED TO READ AHEAD (TEXT AND NOTES) SOMETIMES! G. Former EE122 Student 1: OBJECTIVES • Get some practical knowledge about the design and analysis of basic analog circuits. EE113 and EE122 are parallel courses. the lectures. PLEASE LET US KNOW! THE COURSES ARE ROUGHLY SYNCHRONIZED. analyze.
Wiley Lectures: Office Hours: (Subject to Change) Admin.” 2nd Edition. References: Horowitz and Hill. “Electronic Design.10:50 am Mon 11:00 am . “The Art of Electronics. Roden.3:00 pm Fri 1:00 pm . “Device Electronics for Integrated Circuits.Chapter 1: INTRODUCTION 3: COURSE INFORMATION Instructor: Prof.” AddisonWesley Muller and Kamins.” and “The Bipolar Junction Transistor. “Analysis and Design of Analog Integrated Circuits. “PN Junction Diodes.” Cambridge Press Savant. Greg Kovacs Office: CISX 202 Email: kovacs@cis Phone: 7253637 (email greatly preferred for setting up meetings) Skilling 193 MonWedFri .” HRW Notes handed out in class + supplements as needed. “Microelectronic Circuits. Kovacs ©1997 Page 2 . Assistant: TA: Texts/Notes: G.” 2nd Edition.2:00 pm (Or by appointment please. and Carpenter.12:00 pm Wed 2:00 pm .” Benjamin Cummings Gray and Meyer.) Ms.3:00 pm MTWTF Nolan Sharp Email: nolans@leland Office Hours TBD Review Sessions TBD Sedra and Smith. Susan Kahn CISX 203 Phone: 7230720 email: skahn@leland Hours: 10:00 am . Wiley Neudeck. 10:00am .
but not covered in class) BJT REVIEW (handed out.Chapter 1: INTRODUCTION Credit: Grading: 3 units. Midterm Exam ± Quizzes = 35% (OPEN BOOK) Final Exam = 55% (OPEN BOOK) 4: OUTLINE OF THE COURSE INTRODUCTION AND REVIEW (handed out. letter grade only Homework = 10% assigned weekly. Please don’t be late (penalties may apply). Kovacs ©1997 Page 3 . covered briefly in class if needed) THE OPERATIONAL AMPLIFIER COMMON EMITTER BJT AMPLIFIERS GENERALIZED FREQUENCY RESPONSE ESTIMATION COMMON BASE BJT AMPLIFIERS COMMON COLLECTOR BJT AMPLIFIERS CASCADED/CASCODED BJT AMPLIFIERS DIFFERENTIAL AMPLIFIERS CURRENT SOURCES MULTISTAGE AMPLIFIERS FEEDBACK: THEORY AND PRACTICE G. due at 5pm on date marked on assignment in box at CISX 203.
Zeros and ones for simple minds. we want more. Kovacs ©1997 Page 4 . three. two. Gain and bandwidth. We don’t want to oscillate. Kovacs (The words are first barked out by the professor. four. Wilson. They’ve got circuits. I guess NAND gates aren’t all that bad. Just can’t get ‘em off my mind.Chapter 1: INTRODUCTION Analog Boot Camp Drill Routine by G.) Analog circuits sure are fine. One. Brokaw too. six. Widlar. Digital circuits ain’t my kind.’ ’Cause I need them for circuit CAD. seven. then shouted back by students marching in formation. Five. how ‘bout you? (repeat) G. eight.
remains the same in both forms. G.5 READ APPENDIX E in S&S!!! 2: SOURCES Thévenin & Norton ISC = VOC Ro Ro Voc +  I sc Ro You should review this briefly and know how to convert back and forth. but its orientation in the circuit changes. The output resistance. distortion and large versus small signals. Former EE113 Student 1: OBJECTIVES • To review sources. The Norton form has a current source whose value is found by shortcircuiting the Thévenin form. I can’t read or write in two languages..Chapter 2: BASIC CONCEPTS Chapter 2: BASIC CONCEPTS I’m a bilingual illiterate.1 .. READING: READ S&S Sections 1. amplifiers. Ro. transfer functions. signals. The Thévenin's voltage source is found by taking the opencircuit voltage of the Norton form. Kovacs ©1997 Page 5 . • To discuss noise. and Bode plots.1.
.5. 4 .3. π 3π 5π (since the signal is odd. all an’s are zero) G.a way to represent periodic signals as a sum of sine and cosine “harmonics” xt = n=0 ∑ ∞ an cos nω ot + b n sin nω ot One can determine the fourier coefficients (an and bn) using the following equations: T 2 ao = 1 T T 2 x t dt T 2 The DC term an = 2 T x t cos nωo t dt T 2 T 2 For Even Functions bn = 2 T x t sin nωo t dt T 2 For Odd Functions It is often useful to look at the function and decide if it’s even or odd. (odd) bn's = 4 . Kovacs ©1997 Page 6 ∞ . 4 ..Chapter 2: BASIC CONCEPTS 3: SIGNALS Fourier Series .. Example 1: symmetrical square wave (odd > sine terms only) x t = 4 ∑ 1 sin nω ot π n=1 n n's = 1..
n2 ... This is a common signal found in power supplies after the sinewave input power has been rectified using a diode or diodes and if an output filter capacitor is not used.Chapter 2: BASIC CONCEPTS Example 2: Full Wave Rectified Sinewave. is included in most handheld digital multimeters. G..the amount of DC power required to provide an equivalent amount of heating in a resistive load.. To compute the RMS. it is very useful when measuring the energy or power in signals that might not be a wellknown waveform. or a close relative..1 RMS Value of an AC signal . take the signal. T ∞ VRMS = AVG v t 2 = Some useful RMS formulas: 1 T 0 v2 t dt Sinewave RMS = Vpeak 2 = Vpeaktopeak 2 2 Symmetrical Squarewave RMS = Vpeak = Triangle Wave RMS = Vpeak 3 Vpeaktopeak 2 = Vpeaktopeak 2 3 The AD536/AD636 AC/RMS Converter chip from Analog Devices directly computes the RMS value of an analog signal and it. cos nω ot x t = 2.4. average it.6. Kovacs ©1997 Page 7 . square it.. and take the square root.4 ∑ π π n=2..
Thus. is vo Transresistance Rm = vo = transresistance is in Ω source current. One can model any amplifier as any of the four types. Source Parameter to be Amplified source voltage . is io Current Ai = io = current gain is (dimensionless) G. an amplifier is usually designed to be a particular type. In other words.Chapter 2: BASIC CONCEPTS 4: AMPLIFIERS TYPES OF AMPLIFIERS Amplifiers always increase (or at least maintain) the signal power. Kovacs ©1997 Page 8 . vs Desired Output Parameter vo Type of Amplifier Voltage Gain Expression Av = vo = voltage gain vs (dimensionless) source voltage . vs io Transconductance i Gm = vo = transconductance s in Ω1 or Siemens source current. there are four basic types of amplifiers. but the intended use of the amplifier usually makes one choice usually the best. The gain of an amplifier is expressed as a voltage gain. depending on what it is that they amplify (voltage or current) and what it is that you want as their output (voltage or current). voltage output) or current gain. transresistance (current input. current output). transconductance gain (voltage input.
since most inputs and outputs are not purely resistive! NOTE 2: RS is shown as a resistor at the input of the amplifier that effectively attenuates the input signal if the amplifier is not ideal (i. G.Chapter 2: BASIC CONCEPTS SCHEMATICS OF THE BASIC AMPLIFIER TYPES Rs vs Ri Ideal Voltage Amplifier + + vi  A v vi +  Ro R L vo  Rs vs Ri Ideal Transconductance Amplifier + io Ro vi  G M vi RL Ideal Transresistance Amplifier is Rs Ri RMi i ii + +  Ro R L vo  Ideal Current Amplifier Rs is Ri ii Ai i i io Ro RL NOTE 1: in general use Z (for impedance) rather than R. if the voltage input amplifiers have input resistances less than infinity or if the current input amplifiers have input resistances greater than zero). Kovacs ©1997 Page 9 .e.
Chapter 2: BASIC CONCEPTS Type of Amplifier Voltage Gain Expression Av = vo = voltage gain vs (dimensionless) Ideal Input Impedance Zi = ∞ Ideal Output Impedance Zo= 0 Transconductance i Gm = vo = transconductance s in Ω1 or Siemens Zi = ∞ Zo = ∞ Transresistance Rm = vo = transresistance is in Ω Zi = 0 Zo = 0 Current Ai = io = current gain is (dimensionless) Zi = 0 Zo = ∞ BASIC AMPLIFIER SPECIFICATIONS • GAIN is usually expressed in decibels in terms of the input and output parameters: VOLTAGE GAIN = Av = vo v i • Decibels (after Alexander Graham Bell) are a common unit of measure. dB = 10 log10 Pout dB = 20 log10 Av Pin p POWER GAIN = Ap = po = iovo i iivi When calculating the overall gain of a cascaded amplifier. simply add up the gain (or loss) of each stage in dB to get the overall gain. Kovacs ©1997 Page 10 . then use. G. POWER gain is expressed as: dB = 20 log10 Av NOTE! if the input and output signals are already POWER.
. T is the absolute temperature in °K. or by fluctuations in component values. FET’s.38 X 10 23 J/°K).Chapter 2: BASIC CONCEPTS 5: NOISE • NOISE is unwanted signal(s) that end up added to the desired signals. automobile ignitions. SOLUTION = use lower currents. SOLUTION = use lowresistance values or cool your circuit • Shot Noise > noise current that occurs in active semiconductor devices (BJT’s. the more noise voltage is generated (thermal noise is also a function of the signal bandwidth). in RMS = 2eI∆f where e is the charge on the electron (1. R is the component’s resistance in Ω. • Noise can originate outside of an amplifier or come from inside of the amplifier. This is proportional to the current.) due to the arrival and departure of individual carriers in the device. • Flicker Noise > lowfrequency noise (1/f dependence of frequency spectrum) generated by the random recombination of electrons and holes in semiconductors. I is the DC current flowing. SOLUTION = shielding! • Thermal Noise > noise from random motion of electrons in conductors (proportional to temperature) > if the conductor is a resistor. and ∆f is the bandwidth of interest in Hz. Kovacs ©1997 Page 11 . etc. etc. the higher its value. vn RMS = 4kTR∆f where k is Boltzmann’s constant (1. G.6 X 10 19 C). SOLUTION = move to another universe.. generally only containing frequencies in the audio spectrum (bad for stereo gear!). and ∆f is the bandwidth of interest in Hz. • Interference > electromagnetic interference (EMI) is “pickedup” signals from external noise sources such as household wiring.
In practice.0 0.e. Kovacs ©1997 Page 12 .0 0 100 200 x10 6 300 400 500 An actual "white noise" signal with bandwidth limited to 20 kHz. a difference in the spectrum of the input and output signals. noise is only “white” over a finite bandwidth. generally because the semiconductor (or tube) amplifier are not perfectly linear. 1.5 1.Chapter 2: BASIC CONCEPTS • “White” noise is noise that has a flat frequency spectrum (i. • White noise can be really useful for determining the frequency response of circuits using a spectrum analyzer .all frequencies are equally represented in the spectrum of white noise." or the amplifier simply reaching one of the voltage or current extremes beyond which it cannot swing). The sound from an FM receiver between channels is moreorless white. • Distortion can be noted as a difference in waveform shape the ideal scaled copy of the input.1 0. or sometimes observed by listening to the output of an amplifier (if it is used for audio). contains all frequencies in equal proportion). distortion can come from amplifier saturation ("clipping.6 0. In some cases. 6: DISTORTION • DISTORTION of a signal occurs when the amplified version of the signal coming out of the amplifier is not simply a scaled copy of the input signal. • Distortion is due to nonlinearities. G. you can obtain a frequency response for the circuit under test. but is differently shaped (distorted). so you can input it into a circuit you are testing and look at which frequencies come out! If you average over a long enough time.
in other words. Nonlinearities will give rise to harmonics (signals at frequencies other than the one input to the system) which are measurable with the spectrum analyzer (and sometimes by your eye on an oscilloscope screen). with amplitudes given by Ai (fi)) to the input signal power. etc. • The term total harmonic distortion (THD) represents the percentage of the total output signal of an amplifier that is at frequencies other than the one put in. G. gain. • Other amplifier specifications such as frequency response (bandwidth). %THD = ∑ A i (f i ) i=1 ∞ A o ( fo ) × 100% • In practice. A linear system will only have the same frequency at its output.g. one does not add up harmonic amplitudes to infinite frequency.Chapter 2: BASIC CONCEPTS • Another common type of distortion in amplifiers that use both PNP and NPN transistors at their outputs is crossover distortion. sum of signal frequencies other than fo. but through the range of interest (e. Kovacs ©1997 Page 13 . which is caused by the slight "gap" in voltage between one type of transistor turning off and the other turning on. up through 20 kHz for audio).. will be discussed below. you drive the amplifier with a pure sinewave at a frequency f o and make a ratio of the power in the harmonics (i. Crossover Distortion • You can test for distortion by using a pure (single frequency) sinusoidal input signal and looking at the output either visually or with a spectrum analyzer...e.
• The efficiency of an amplifier reflects the amount of power delivered to the load as a fraction of the total power drawn from the power supply. • Most analog amplifiers use two power supply voltages or “rails.Chapter 2: BASIC CONCEPTS 7: AMPLIFIER POWER SUPPLIES & EFFICIENCY • All amplifiers need some type of power supply to supply the extra energy that is delivered to the load. but sometimes they internally "split" that single voltage into two rails by making an artificial "ground" voltage half way from "real ground" to the supply voltage. V+ I dc ii vi io RL V • Some amplifiers use only a single power supply voltage. and can be computed using: Power Delivered to Load η = PL × 100% = × 100% Pdc Power Used from Power Supplies G. Kovacs ©1997 Page 14 .” as shown below.
. Bode plots. you have probably heard clipping when someone turned up a stereo too loud! • Amplifier distortion due to transistor nonlinearities > this is simplest to understand by considering that basically. AC smallsignal equivalent models. • Examples of large signal effects (as discussed above in "Distortion"): • Amplifier clipping (saturation) > here you have a case where the amplifier’s output cannot swing above and below certain maximum and minimum voltages (that makes sense). • Amplifier exploding (very nonlinear) due to extremely large input signal: When we talk about transfer functions. clean amplifier outputs is that we are keeping signal swings small enough through various techniques. Kovacs ©1997 Page 15 . we are always assuming that the circuit is in smallsignal operation! G. and the only reason we get nice... all transistors are nonlinear devices and we work very hard to “coax” linearity out of them over certain ranges of signal level. this type of distortion can be minimized but can never be completely avoided.. etc.Chapter 2: BASIC CONCEPTS 8: LARGE AND SMALL SIGNALS • Most circuits are linear if the input signals are small enough! If the signal amplitude is increased enough some type of nonlinearity will make itself obvious! All semiconductor devices (and vacuum tubes!) are very nonlinear..
G.” or the complex frequencies at which the numerator becomes zero (the transfer function goes to zero). • You can use capacitors at the inputs and outputs of amplifiers to “block” DC signals. This type of amplifier is called AC.pM z1 through zN are the “zeros.. as long as they are large enough to "look like shorts" in the frequency range of interest.. • The frequency response of such an ACcoupled circuit cannot extend to zero Hz. since they are plotted to illustrate the frequency range over which the amplifier will operate properly.z2 = Ao vin s s . You should note that oscilloscope amplifiers always have the option of choosing one or the other coupling modes. • Amplifiers are either DC or AC coupled.” or the complex frequencies at which the denominator becomes zero (the transfer function goes to infinity). p1 through pM are the “poles..or capacitivelycoupled..zN s ...p1 s .p2 s . meaning that the inputs are sensitive to both DC and AC signals ("DCcoupled") or only AC signals ("ACcoupled).z1 s .Chapter 2: BASIC CONCEPTS 9: TRANSFER FUNCTIONS • You already know what these are if you have ever looked at stereo equipment catalogs or specification sheets. • Here is a brief complex number review for your reference so you can compute transfer functions of circuits: V = Re + jIm = A ejθ Re = Acos θ Im = Asin θ V = Re2 + Im2 θ = arctan Im Re Remember that s = σ + j ω POLES & ZEROS > WHAT DO THEY MEAN? Transfer Function Notation: Ts = vout s s . Kovacs ©1997 Page 16 .
SIMPLE FILTERS • The cutoff (or 3dB) frequency is the point at which the response is 3 dB lower than in the passband ( ≈ 0. Kovacs ©1997 Page 17 . In real circuits. FIRSTORDER RC LOWPASS FILTER: R V in C V out 1 Cs = 1 1 Hs = = 1+ s R + 1 RCs + 1 ωo Cs • The general form is Hs = 1+ s ωo K OR ωo + s Kω o = ωo + s A where K is the gain for low frequencies and A = Kω o • The pole is at S = ωo • The “cutoff” frequency is at ω0 = 2πfc = 1 RC G. one cannot have a zero without a pole to “cancel” it out at some frequency. HERE IS A GOOD IDEA TO GET A SENSE FOR CIRCUIT BEHAVIOR: Look at the circuit first before doing any math! The capacitors are all infinite impedance for DC and their impedance decreases toward zero as the frequency increases.Chapter 2: BASIC CONCEPTS • The “order” of the system = N + M • A zero alone would make the output amplitude of a circuit increase forever with increasing frequency.707 times the passband amplitude).
G.45° 1 KHz FIRSTORDER RC HIGHPASS FILTER: C V in V R out Hs = R = RCs = s = s s + ωo R + 1 RCs + 1 s + 1 Cs RC Hs = Ks s + ωo where K is the gain for high frequencies • The general form is • The pole is at S = ωo and the zero is at S = 0 • The cutoff frequency is the same as for the lowpass filter. Kovacs ©1997 Page 18 .707 V .Chapter 2: BASIC CONCEPTS Gain and phase responses of the firstorder. RC lowpass filter: 0.
Kovacs ©1997 Page 19 . RC highpass filter: 0.Chapter 2: BASIC CONCEPTS Gain and phase responses of the firstorder.707 V + 45° 1 KHz G.
you can skip to the next section. This is an important technique. • a pole makes the amplitude fall with frequency by 20 dB /decade and “has no effect” before its break frequency • a zero makes the amplitude rise with frequency by 20 dB/decade and “has no effect” before its break frequency • a pole causes the phase to fall from 0° to 90° over two decades of frequency starting one decade before the break frequency > the phase is 45° at its break frequency • a zero causes the phase to rise from 0° to +90° over two decades of frequency starting one decade before the break frequency > the phase is +45° at its break frequency • the effects of poles or zeros at “zero” frequency have already “maxed out” by the time you start your plot (i. despite the availability of computers! You will need to be able to do Bode plots on exams and in some "real world" situations.Chapter 2: BASIC CONCEPTS 10: BODE PLOTS This material is covered in the prerequisites to EE113. and +90° for a zero at zero frequency) G. Kovacs ©1997 Page 20 . but typically.e. to check your work! ADD up the individual responses of all of the poles and zeros of the transfer function.. It is assumed that you have an equation for the transfer function.. They each affect the frequency response “only” after they take effect at their respective “break” frequencies. people are a bit "shaky" on phase plots. The key point is that the effect of poles or zeros starts to take place a decade before or a decade after the pole or zero frequency when you are dealing with phase (unlike at the frequency for gain plots)! If you are already familiar with all of this. Remember that you can use computer programs like Matlab™. THE BASIC IDEA: The point is to be able to draw a "quick" sketch of a transfer function of a circuit. Theorist™. phase is 90° for a pole at zero frequency. etc.. Mathematica™.
then shift the whole phase sketch: a) + 90° for any zeros at "zero frequency" b) .180° if there is a negative sign Remember that a negative sign on a gain is a 180° phase shift! G. add a +20 dB/decade slope to the slope of the sketch at the break frequency of that zero. estimate the gain and convert to dB.Chapter 2: BASIC CONCEPTS HOW TO MAKE A GAIN PLOT 1) Write the transfer function equation in a form so that you can see the break frequencies of the poles and zeros.. add a 20 dB/decade slope to the slope of the sketch at the break frequency of that pole. plug those frequencies in for "S". 2) Try to begin the sketch at a frequency where you know the gain (from looking at the equation). 2) Try to begin the sketch at a frequency where you know the phase (from looking at the equation).90° for any poles at "zero frequency" c) +/. Kovacs ©1997 Page 21 . 5) Draw a “smooth” curve over the sketch (the curves differ by about 3 dB at each single break).. Remember about poles and/or zeros that may have already “taken effect” at low enough frequencies that they are “maxed out” before you start your sketch. 3) For each zero. Remember about poles and/or zeros that may have already “taken effect” at low enough frequencies that they are “maxed out” before you start your sketch. One way to make it easier is to start out assuming 0° at "super low" frequencies. draw a rough draft of the curve and select the frequencies corresponding to the "flat" parts. 4) For each pole. HOW TO MAKE A PHASE PLOT 1) Write the transfer function equation in a form so that you can see the break frequencies of the poles and zeros. If it is not obvious.
The phase contribution from that pole is “half way there” (or contributing 45°) at the break frequency. The contribution of that pole to phase at frequencies less that one tenth of the break frequency and greater than ten times the break frequency is zero! 5) Draw a “smooth” curve over the sketch (the curves differ by about 6° at each single break).Chapter 2: BASIC CONCEPTS 3) Each zero contributes a phase slope of +45° per decade starting one decade below and lasting through one decade above the break frequency. Kovacs ©1997 Page 22 . The contribution of that zero to phase at frequencies less that one tenth of the break frequency and greater than ten times the break frequency is zero! 4) Each pole contributes a phase slope of 45° per decade starting one decade below and lasting through one decade above the break frequency. The phase contribution from that zero is “half way there” (or contributing +45°) at the break frequency. G.
a pole at 103 radians/S and a pole at 106 radians/S.Chapter 2: BASIC CONCEPTS EXAMPLE 1: 120 dB 100 dB 80 dB 60 dB 40 dB 20 dB 0 dB 20 dB 1 10 1 Hs = 10 s + 10 3 s + 10 5 10 10 2 10 3 10 4 10 5 10 6 10 7 10 8 EXAMPLE 2: Typical CapacitorCoupled Amplifier s 1 10 8s Hs = 100 = s + 10 3 1 + s6 s + 10 3 s + 10 6 10 Here we have a zero at zero (‘maxed out” by the time you start your plot). In the midband. Kovacs ©1997 Page 23 . the first pole will have “cancelled” the zero and the gain should be flat at 100... G.
Kovacs ©1997 Page 24 . you should only need to consider the zero. 120 dB 100 dB 80 dB 60 dB 40 dB 20 dB 0 dB 20 dB 1 10 1 10 10 2 3 4 5 6 7 8 10 10 10 10 10 10 G.Chapter 2: BASIC CONCEPTS Below the first pole frequency...
Inventors of the bipolar transistor (they didn't really say it). where the “pointin’” part refers to the direction of the arrow on the emitter terminal. 1: OBJECTIVES BJT operation BJT DC Analysis and DC Load Lines BJT smallsignal model > hybrid π Looking “in” each terminal AC Load Lines READ S&S Chapter 4.9 2: TYPES OF BIPOLAR JUNCTION TRANSISTORS • There are two polarities: NPN (“not pointin’ in”) and PNP (“pointin’ in”). Kovacs ©1997 Page 25 . the world would be a pretty dull place! Shockley.” cutoff = “OFF”) Mode Cutoff Active Saturation EBJ Reverse Forward Forward CBJ Reverse Reverse Forward G.1 .4. Bardeen and Brattain. • BJT’s can operate as amplifiers (active mode) or switches (saturation = “ON. Sections 4.Chapter 3: BJT REVIEW Chapter 3: BJT REVIEW Without bipolar transistors.
Kovacs ©1997 Page 26 .Chapter 3: BJT REVIEW G.
Chapter 3: BJT REVIEW 3: DC ANALYSIS (LARGESIGNAL) • The collector current versus VBE for a BJT is given by. β F and the reverse current gain. the forward current gain.1 βF βR where. • If it is certain that VBC < 0 (basecollector junction is reverse biased) then one can use simplified forms of the above equations. as seen below. Kovacs ©1997 Page 27 . we only use "β" and assume we are talking about βF.e kT qVBE qVBC IB = Is e kT . VBE > 0 and VBC < 0 so one can neglect the second exponential terms in both equations. in the lower equation. qVBE VBE IC = I s e kT IB = IC β = Is e VT IS may be on the order of 1014 to 1015A where kT/q ≈ 26 mV at room temperature. • IS is the “current scale factor” which depends on the size of the transistor (in particular. β R are considered to account for all possible operating modes. simply neglecting the βR terms and substituting the generic β for β F. qVBE qVBC IC = I s e kT . • For normal operation (assuming NPN). • Another essential equation relates the emitter current to the base current: I E = I C + I B = (1 + β) IB ≈ βI B if β is large! G. Generally.1 + Is e kT . the geometrical area of its emitter) and the process by which it was made.
G. the point is that tiny changes in VBE cause huge changes in I C. saturation (when the transistor is “fully on”) and cutoff (where the transistor is “fully off”). Kovacs ©1997 Page 28 . • The plot above shows the exponential relationship between IC and VBE. VBE curve...Chapter 3: BJT REVIEW • The plot above shows the three regions of BJT operation: active (where analog amplifiers typically operate). • Note that IB vs. VBE curve is just a scaled copy of the IC vs.
.. if so. NPN & PNP are different for DC biasing! Here we look at NPN’s) 1) Typically. VBE ≈ +0.2V.. coupling capacitor to block DC.6 mA )( 500 Ω ) = 0.. G. this means that the input signal would not have to swing very far to put the transistor into saturation (not good for an amplifier!).6 mA VCE = 15V − (28.. • Start out assuming active mode...7 V • This means that the basecollector junction is at zero volts (on the edge between reverse and forward biased... • Assume ß = 100.e. VBC > 0 !!! We don't want a forward biased BC junction.286 mA 50kΩ I C = 100I B = 28. If VBC < 0 and if IB is uniquely determined. we generally assume that a transistor is not fully saturated until VCE ≈ 0. Kovacs ©1997 Page 29 .. but we really don’t want to cut it that close for a real amplifier because the IC vs VCE curves start to really bend there... if β = 200. leading to distortion .. this means that VBC can be as much above zero as 0..7V. • What if β > 100? For example.. VCC +15V RB 50 kΩ 500 Ω RC vo vs THIS IS NOT A GREAT BIASING SCHEME. IB = 15 − 0.7 = 0.7V 2) Check that the basecollector junction is reverse biased: VBC < 0 (i.5V before “full” saturation (see the curves above). you start out assuming that the baseemitter junction is forward biased (but later should verify that!). IT IS VERY SENSITIVE TO VARIATIONS IN ß. but still not “on”). then IC = β IB In practice.Chapter 3: BJT REVIEW 4: ANALYSIS OF BIASING (Be careful. so VBE = 0.. then the transistor is not saturated).
84 .15V VE = . Kovacs ©1997 Page 30 .7 V IE = = 14.54V (not saturated) and V BC = 0 .16mA IC = 14.(0.7) = 1.3mA 100 = 14.Chapter 3: BJT REVIEW VCC = +15 V RC V C V RE E V EE = 15 V RE = 1KΩ βF = 100 0. • What if ß is 200 now?.84 (reverse biased).0..3mA β+1 101 VC = 15V . not much happens > this is better biasing! A MORE REALISTIC BIAS CIRCUIT (COVERED IN THE CE AMPLIFIER SECTION): VCC RB1 VB vs I B1 IB RC vo R B2 I B2 RE VEE G..7 .0.16mA 1K = 0.3mA 1K β = 14.84V VCE = _______ VBC = _______ RC = 1KΩ • The answers are: VCE = 0.14.84 = 0.
voltage characteristic of the resistor in the collector circuit (just Ohm’s Law) and the IC vs VCE characteristic of the transistor (for a given base current). G. Kovacs ©1997 Page 31 . • The intersection of the two current vs.Chapter 3: BJT REVIEW 5: DC LOAD LINES • This is a graphical approach to solving two simultaneous equations: the current vs. which is where the amplifier “idles” with no AC input signal. voltage curves is the quiescent point (“Qpoint”).
.. not distort too much). Kovacs ©1997 Page 32 .. VBE + vbe IC + ic = I′ C = Is e VT DC + AC current = Is e kT = I C e VT vbe qVBE e VT vbe = I C 1 + vbe + .. The transistor’s behavior is almost linear if the region is quite small...Chapter 3: BJT REVIEW 6: SMALLSIGNAL BEHAVIOR • The goal when modeling smallsignal behavior is to make a model of a the transistor that works for signals small enough to “keep things linear” (i..e. linearize... I′ C = IC + IC vbe = I C + gm vbe VT This assumes vbe << kT = VT ≈ 25..9 mV q gm ≡ IC VT THE TRANSCONDUCTANCE G. • The basic trick is to linearize the very nonlinear exponential relationship between VBE and IC by looking at a small enough region of the exponential.. VT expand.
Kovacs ©1997 Page 33 . as above) become opencircuits! AND BECOME SHORTS! Stanford AND BECOME OPENCIRCUITS! G.. that models the input of the transistor for small signals. the DC power supply voltages get shorted to ground.. βi b = g m v be → rπ = v be β = ib gm • The concept of gm and rπ can be usefully combined to obtain the 7: HYBRIDπ MODEL FOR (AC) SMALL SIGNALS B ib + rπ vbe ie E ic C gmvbe • This simple model for the BJT can be plugged into amplifier circuits to figure out gain. • REMEMBER that for AC analysis.Chapter 3: BJT REVIEW • Now one can look at i b versus v be. rπ . etc. and current sources (other than those defined as AC signals. capacitors become shortcircuits. want to look at it as an equivalent input resistance. More detailed versions include parasitic capacitances and resistances. i c = g m v be and i c = βi b therefore. • The AC component of the collector current from above is.
. however. IC directly controls gm and hence the gain when used in an amplifier circuit! G. to consider the differences between NPN and PNP transistors when analyzing the DC bias conditions! It is important to keep in mind the effect that the DC bias conditions (particularly IC) have on the AC model. Kovacs ©1997 Page 34 . basically.Chapter 3: BJT REVIEW NOTE that the smallsignal equivalent circuit shown above works for NPN or PNP transistors with NO CHANGE OF POLARITY!!! It is CRITICAL..
B C rπ gmvbe E • Looking into the emitter toward the base. one “sees” rπ. re seems much smaller. For a change in emitter voltage. re = rπ β+1 also.. looking in through the emitter.. but also the extra current from the transistor’s amplifying action. Therefore.. but from the emitter’s viewpoint. re ≈ r π = 1 gm β G. a small change in voltage (a change in vbe) causes a much larger change in current due to the gm generator! •This means the emittertobase resistance. Kovacs ©1997 Page 35 . one “sees” re. • Looking into the base toward the emitter. a small change in voltage causes a relatively small change in base current..Chapter 3: BJT REVIEW 8: SCALING RESISTANCES BETWEEN BASE AND EMITTER • The resistance between base and emitter depends on which way you look! • The conceptual explanation is: from the base’s viewpoint. (β + 1) times as much current change will occur than for a comparable voltage change at the base. a small change in the voltage relative to the base causes a huge amount of current to flow (compared to the same change in voltage applied looking into the base) because you get not only the base current flowing through rπ.
or dividing by (β + 1) if looking in the emitter. 9: AC LOAD LINES G.Chapter 3: BJT REVIEW KEY POINT: You can “transform” a resistance from the emitter or base side of the baseemitter circuit to the opposite side by multiplying by (β + 1) if looking in the base. Kovacs ©1997 Page 36 .
iB Nearly Linear Small Segment of Curve i B2 Base Bias Current B (DC) ib Time I Quiescent Point i B1 0 vbe vbe V BE Quiescent BaseEmitter Voltage G. Kovacs ©1997 Page 37 . varies with vbe. how much it is amplifying). Such variations in ib lead to variations in collector current (and finally output voltage) by causing the movement between the various IC vs VCE curves shown above (movement from one line to the other on the plot is constrained by the output resistor load line. ib.Chapter 3: BJT REVIEW • The idea here is to extend the DC load line concept to allow for “wiggling” the base current using an applied (AC) signal and looking at the corresponding “wiggling” of VCE along the constraining line of the resistor (Ohm’s Law) to see what the output of the amplifier is doing (i. • The plot below shows how the base current. giving rise to the output voltage waveforms shown).e.
the basic opamp circuits the differences between “ideal” and “real” opamps the frequency response of opamps a bit about feedback READ S&S CHAPTER 2 G.. Former EE113 BrownNosing Student Presently employed as freelance poet.. Kovacs ©1997 Page 38 .. 1: OBJECTIVES • To learn: what an opamp is. opamps are neat.. they just can’t be beat.Chapter 4: OPERATIONAL AMPLIFIERS Chapter 4: Operational Amplifiers Opamps are great. between you and me.
G. • Nonlinear distortion can be reduced. V1 + A(V2V1) +  V2 1) The input impedance is infinite . • The input and output impedances of the amplifier can be modified. • The bandwidth of the amplifier can be extended. Kovacs ©1997 Page 39 . 5) The output voltage is zero when the input voltage difference is zero. • The effects of noise can be reduced. 3) The openloop gain (A) is infinite.Chapter 4: OPERATIONAL AMPLIFIERS 2: OPERATIONAL AMPLIFIERS CONCEPTS • Opamps are amplifiers that provide the amplified difference between two input signals as their output.i. COMMENTS ON FEEDBACK (more later!) • The gain of the circuit is made less sensitive to the values of individual components.e.. the opamp can drive any load impedance to any voltage.e.. 2) The output impedance is zero . no current ever flows into either input of the opamp. 4) The bandwidth is infinite.i.
Vwhich yields.. • What about this configuration? Does it work? VIN VV+ VOUT • Well..= VOUT substitute into the basic opamp equation to get.Chapter 4: OPERATIONAL AMPLIFIERS 3: BASIC OPAMP CIRCUITS 3. it looks like it should.1 VOLTAGE FOLLOWER VVIN V+ VOUT • Note that the opamp can swing its output in either direction to keep the voltage difference between the input zero. if you do the math.... VOUT = A V+ . Kovacs ©1997 Page 40 . ⇒ VOUT = VOUT = A VIN . they can move to the right halfplane and the whole thing oscillates (positive feedback).VOUT A V+ 1+A ⇒ ⇒ 1 + A VOUT = AVIN as A→ ∞ VOUT = V+ • Here the input impedance is that of the opamp itself (very high). VOUT = A( V+ − V− ) = A( VOUT − V− ) A V → VOUT = −V− as A → ∞ VOUT = 1−A − • This would work only if the opamp had no poles (response flat out to infinite frequency). if you have any poles. G. V..
This means that the current in through R1 must come out via R2 and none enters the v.. This means that one can write. KEY POINT: With negative feedback (as is generally used with opamps except as oscillators or comparators). This is a common configuration. for an ideal opamp. v out = A ( v + − v − ) and thus.2 INVERTING AMPLIFIER R2 VIN R1 VV+ i fb VOUT i in • By definition. If v + is grounded. This simple rule makes opamp analysis quite simple in many cases. the input current is zero. Since.is also forced to ground.. v. v. v + − v − = approaches infinity. the input impedance is infinite. the two input terminals are forced to the same potential by the feedback. Kovacs ©1997 Page 41 . and we refer to the vterminal as a “virtual ground. G.is still forced to the same voltage that is applied to v+.terminal. v out which means that as A A v in v out + =0 R1 R2 which gives A V = vout R =− 2 v in R1 • Note that the input impedance is simply R1 since it is connected to a virtual ground.” If v+ is not grounded. ( v + − v − ) → 0 (this is the virtual ground assumption).Chapter 4: OPERATIONAL AMPLIFIERS 3.
Chapter 4: OPERATIONAL AMPLIFIERS 3.. this circuit would probably not be used for a log amplifier..3 LOGARITHMIC AMPLIFIER • Replace R2 of the Inverting Amplifier with a diode. Instead.terminal. • We also know that. i fb = − v out ) R2 and their sum must be equal to zero since no current enters the v. v v− − v out v nvT nv T nv T = Is e − 1 ≈ Ise thus. one would use a groundedbase transistor in place of the diode to eliminate the effect of the “n” term (dependent upon the current through the diode) and also take advantage of the fact that the exponential iv relationship extends over a much larger range for transistors. i v out = −nv T ln fb = −nv T [ln(i fb ) − ln( I S ) ] IS • Since i fb = i in.. one can substitute iin = vin for ifb. the feedback current is i fb ≈ Is e I diode (assume n = 2 for discrete diodes) • Using the virtual ground assumption. R1 v v out = −nv T ln in − ln (I S ) R1 • In practice. G.. VIN R1 VV+ i fb VOUT i in • Start with the diode transfer function. v i in = in R1 and i fb ≈ Is e −vout nvT (remember that for an inverting amp. Kovacs ©1997 Page 42 ..
if A goes to infinity. In other words. as discussed below)! This is accomplished as follows: A × B = e[ ln( A) + ln (B )] A ln A − ln B = e [ ( ) ( )] B • This principle. This means that v in = v. along with the addition.. the opamp tries to hold the difference between its input terminals at zero (remember the idea behind the virtual ground assumption.terminal is forced to equal the input voltage. if you write the basic opamp equation. v out = A ( v + − v − ) = A ( v in − v − ) → v out = (v in − v − ) A Thus. were the basis for analog computers.4 EXPONENTIAL AMPLIFIER • Replace R1 of the Inverting Amplifier with a diode. this is similar).. the difference between vin and vmust go to zero. Kovacs ©1997 Page 43 . v out = −R 2 I s e nv T R2 vin • With exponential and logarithmic functions.Chapter 4: OPERATIONAL AMPLIFIERS 3. subtraction. Again. and substitute vin for v+.5 NONINVERTING AMPLIFIER R2 R1 VV+ VIN i fb VOUT i1 • Since feedback is operating. 3. as the opamp's gain A becomes very large. the voltage at the v. one can multiply and divide if one can add and subtract (this is easy with opamps. R2 VIN R1 VV+ i fb VOUT i in v in nv T • Note that i in ≈ I se v = − out and thus. integral and derivative opamp functions described below. G.
6 SUMMING AMPLIFIER V1 V2 V3 V4 Vn R1 Rf i1 R2 R3 R4 VV+ i fb VOUT Rn • This amplifier's operation can be "summed up" by this equation. ∑i k=1 n k + i fb = 0 v v out v v v = − 1 + 2 + 3 +L + n Rf Rn R 1 R 2 R3 • Summing currents. • The summing amplifier is often used as an audio mixer. but that uses on additional amplifier. such as in instrumentation. v − = v + = v in .Chapter 4: OPERATIONAL AMPLIFIERS • Thus. (Of course. Kovacs ©1997 Page 44 . v out R 2 = 1 + vin R1 • This circuit is very useful when you need to use the full input impedance of the opamp.) 3. write. considering the fact that no current will enter the opamp terminals. and. Which can be rearranged to give. you can obtain a noninverting amplifier by connecting two inverting amplifiers in series. indicating that the input and feedback currents must add up to zero (no current enters the operational amplifier's input terminals). and the input resistors. one can write. whose outputs are summed together. • This circuit can also be used as an averager if R1 = R 2 = ••• = Rn and R f = (R 1/n) G. v out = − v1 Rf R R R + v2 f + v 3 f + L + v n f which R1 R2 R3 Rn has the form of a series of "n" inverting amplifiers. v out − v in v in − =0 R2 R1 which can be rearranged to yield. start with the assumption that. where the inputs are voltages from several microphones or other sources. R1 ••• Rn are varied to control the individual channel gains.
v in dv = −C1 out R1 dt • It is simplest to replace R 2 for the inverting amplifier with the impedance of the parallel combination of R2 and C1. the capacitor begins to matter.7 INTEGRATOR R2 C1 VIN R1 VV+ VOUT i in • For DC. in fact. (When you see capacitors in a circuit. an integrator for frequencies above f c. it is just like an inverting amplifier. v out R =− 2 vin R1 • For higher frequencies. and an inverting amplifier for frequencies below f c. Kovacs ©1997 Page 45 .Chapter 4: OPERATIONAL AMPLIFIERS 3. v out vin R2 C1 S 1 R2 1 R2 R2 + R 2 C1 S Z C S R C S + 1 R1 =− 2 =− 1 =− =− 2 1 =− R1 R1 R1 R1 R2 C 1S + 1 1 2πR 2C 1 • The form of this equation is a lowpass filter with a cutoff frequency of fc = (we will refer to it as fmin) and a DC gain of v out R =− 2 vin R1 • This is. always consider what happens at zero and infinite frequency!) G.
but only for frequencies above the point where the effect of C2 begins to dominate over R 2 to set the gain of the circuit.Chapter 4: OPERATIONAL AMPLIFIERS • For the theoretical “ideal” integrator circuit.. "nonideal" opamps) would quickly "charge up" the integrator. and the equation would be.. The frequency below which the circuit’s behavior becomes more like a DC amplifier than an integrator is given by the lowpass filter's cutoff frequency. R2 would be omitted. v out = − 1 v in dt R 1C 1 ∫ • However. remember that the integral of a constant is the constant times time! We want to limit the integrator's theoretical infinite gain at DC to something less. small DC offsets at the input (and those of realistic. • For the realistic integrator shown above. Kovacs ©1997 Page 46 . the integral equation is correct. Gain R2 R1 "Integrator" "Amplifier" fmin Frequency G.
G. one can start with the equation for an inverting amplifier and substitute the impedance of R1 and C1 in series for R1. that is sometimes not shown in “textbook” opamp differentiator circuits. v out = −R 2 C1 dv in dt • As for the integrator above. v out R R2 R 2 C1 S =− 2 = − =− 1 vin Z1 R 1C 1S + 1 R1 + C 1S which yields an equation for a highpass filter in classical form. Kovacs ©1997 Page 47 . iin and if can be equated to show that. • Again.Chapter 4: OPERATIONAL AMPLIFIERS 3. with a cutoff frequency of. there is a frequency range over which the differentiator will not work very well. R 1.8 DIFFERENTIATOR R2 R1 C1 if VV+ VOUT VIN i in • In the differentiator circuit shown here. • Its purpose is to limit the highfrequency gain of the differentiator so that it does not get swamped by high frequency noise (which may have a large derivative despite a small amplitude). • Again. fmax = 1 v R . above which the filter gives a steady gain of out = − 2 and 2πR 1C1 vin R1 below which it acts as a differentiator. substitute zero and infinity for S in the above equation and be sure you understand what happens at both limits. • For “lowenough” frequencies that the input impedance is dominated by C 1. there is also a component.
Kovacs ©1997 Page 48 .Vi) Vo ViIbias . for most practical applications.VEE • There are unwanted currents at the inputs! • There are offset currents and voltages! • Signals applied to both inputs (which should not be amplified) are amplified to some extent! • The input resistance is not infinity! • The output resistance is not zero! • However. and negative feedback helps make them even closer! G.Chapter 4: OPERATIONAL AMPLIFIERS 4: “REAL” VERSUS “IDEAL” OPAMPS VCC Ro Vi+ Ibias Ios + Vos + Acm (Vi+ + Vi)/2 Ri + Adm (Vi+ . they are pretty close to being ideal.
KEY POINT: You can model an opamp’s openloop frequency response as a firstorder RC lowpass filter with....Chapter 4: OPERATIONAL AMPLIFIERS 5: FREQUENCY RESPONSE OF OPAMPS • As you might have guessed. Kovacs ©1997 Page 49 . at which the gain reaches ONE. • As you increase the amount of negative feedback. but the frequency response is TERRIBLE!. • The product of GAIN X BANDWIDTH is a constant for a given opamp (guess what?. the bandwidth increases and the gain decreases.. • The openloop gain is given by: A op − amp (S ) = Ao S 1+ ωo where Ao is the open loop gain at DC and ωo is the 3dB or “break” frequency at which the openloop gain starts to roll off.. ωt . • The opamp has (by design) a FIRSTORDER LOWPASS RESPONSE. it is referred to as the “gainbandwidth product”). ωt = Aoωo G. ωo = (RC )equivalent 1 • Another key definition is the UNITYGAIN BANDWIDTH. the opamp is in OPENLOOP mode. opamps do not either. as we saw in the first lectures. it follows that.. and the gain is quite large.. have infinite frequency response • When used without external feedback. which is the frequency. • By thinking about the gainbandwidth product concept.
Chapter 4: OPERATIONAL AMPLIFIERS • If you rederive the gain equation for the inverting amplifier assuming a finite opamp gain (see page 55. Kovacs ©1997 Page 50 . A v (S ) = − R2 1 + R1 1+ Ao 1+ Sωo R2 R1 R2 R1 =− 1 R2 S 1+ 1 + + Ao R1 ω t R2 1 + R1 R2 R1 A v (S ) ≈ − S R2 1+ 1 + ωt R1 for R A o >> 1 + 2 R1 • This is a firstorder lowpass filter. A v (S ) = − R2 1 + R1 1+ A op −amp (S ) R2 R1 • Substituting the opamp’s frequency response in for Aopamp. you obtain. G. Sedra & Smith). you (eventually) obtain.
R2 AND THE BREAK FREQUENCY IS R1 ωt 1 + R2 R1 CONCLUSION: You can model the frequency response of the inverting configuration also using a firstorder lowpass filter with. Kovacs ©1997 Page 51 . 1 + R2 R1 RCequivalent= ωt • A similar analysis for the noninverting case can be found in Sedra and Smith (Section 2.Chapter 4: OPERATIONAL AMPLIFIERS R V in C V out Hs = 1+ s ωo K where K is the gain for low frequencies the pole is at S = ωo and the “cutoff” frequency is at ω0 = 2πfc = 1 RC THE DC GAIN IS . G.7).
As the gain is reduced. no matter what the gain of the opamp is set to via external feedback resistors. Kovacs ©1997 Page 52 . This is another way of looking at the gainbandwidth product rule. the overall frequency response is constrained by the outer boundary set by the opamp's open looproll off.Chapter 4: OPERATIONAL AMPLIFIERS A = 1K FREQUENCY RESPONSE OF INVERTING OPAMP CIRCUIT FOR VARIOUS GAINS. the frequency at which that new gain is reduced by 3dB moves up in frequency. ASSUMING GBPRODUCT = 1 MHZ AND fb = 10 Hz A = 100 GAIN A = 10 A=1 FREQUENCY • The above plots illustrate that. G.
with a huge gain and a very small internal capacitance. there is generally an internal (sometimes external) capacitor that sets the opamp’s dominant pole and forces it to roll off before other.) • Here. • In practical integrated circuit implementations.g. • As mentioned above.Chapter 4: OPERATIONAL AMPLIFIERS 6: THE GUTS OF REAL OPAMP CHIPS • In real opamps. the Miller Effect effectively takes an impedance across an amplifier and replaces it with a smaller impedance at the input and a roughly equivalent one at the output.. the capacitances available are very small (a few pF). ground). but Ohm’s Law still applies). which basically explains how an impedance sitting across an amplifier is effectively converted into a smaller impedance at the input and roughly an equal impedance at the output of the amplifier. less well controlled parasitic capacitances would cause that.. • This explains how an opamp. Kovacs ©1997 Page 53 . gives rise to negative impedances at the input (reverse polarity of current that flows for an applied voltage. G. can act like it has a huge capacitance at its input (a smaller impedance corresponds to a larger capacitance). how can such small capacitances give rise to such low cutoff frequencies (a few hundred Hz)? • The answer is provided by an understanding of the Miller Effect. we only consider negative (inverting) amplifiers... (The output impedance is still positive. • The effect does work for negative and positive gains. the amount that the effective input impedance is smaller than the actual one across the amplifier is a function of the amplifier’s gain. but for positive gains greater than one. • This effect works only for amplifiers that share a common terminal between input and output (e.
Avin = vin Zf Zf Zf 1A vout .vout A = vout i2 = vout .vout = vin .Chapter 4: OPERATIONAL AMPLIFIERS i1 = vin . i* = i1 and i* = i2 1 2 Z1 = Zf 1A Z2 = Zf A 1A G.vin = Zf Zf Zf A 1A i* = vin 1 Z1 i* = vout 2 Z2 By definition. Kovacs ©1997 Page 54 .
Section 2...... • This seems.µ V i2  o C (1 + µ) + V i2 R = R o1  R i2 .8). you get more current flowing for a small voltage wiggle than if the impedance were just tied to ground. “looking” from the input terminal. the amplifier will pull a lot more current away from you than if you were pushing against just the impedance tied to ground... the maximum value of the effective Cout is 2Cf and the minimum is Cout = Cf. the amplifier is going to “pull” against the other side of that impedance in response to any input voltage wiggle.. with an inverting gain. sC(1+µ)V i2 G m V id • From this it can be seen that the feedback capacitor’s value is multiplied by a factor of 1 + the gain of the internal voltage amplifier (µ) and placed in parallel with the combination of the first stage’s output resistance and the second stage’s input resistance.Chapter 4: OPERATIONAL AMPLIFIERS • Intuitively. • Note that taking A through the gain range of 1 to ∞. we can look at the opamp “guts” model to figure out why we get the frequency response observed (after Sedra & Smith. • If you wiggle the input voltage up slightly and try to force current into the impedance (remember that you can’t force current into the input of the amplifier!). • Using Miller’s Theorem. G. the corresponding range of Z2 is Z2/2 and Z 2. one can think of “looking” into the input of the amplifier and seeing one side of an impedance. like a smaller impedance than what is sitting across the amplifier.. Kovacs ©1997 Page 55 + G m V id V id  + Ro1 R i2 + V i2 +  V  . Thus for a capacitor in the Miller configuration.
FEEDBACK RATIO (FACTOR) LOOP GAIN: T = −Aβ β= R1 R1 + R 2 In this case.. that makes no difference.e. • At the output. for large µ.Chapter 4: OPERATIONAL AMPLIFIERS • This determines the effective RC time constant of the opamp (see the Appendix below). • This is a useful point to define some of the feedback terms that will be used later in the course. it is just voltage division and ß is the fraction of the output voltage fed back. an equivalent capacitance is present with a value of C. for an ideal (voltage source) output. AMOUNT OF FEEDBACK ≡1T=1+A β GAIN = A A = AMOUNT OF FEEDBACK 1 + Aβ Gain = A V = Ao R1 1 + Ao R1 + R 2 Page 56 G. it can make quite an important difference! 7: A PREVIEW OF FEEDBACK R 2 Feedback Loop R 1 Vin A • This schematic shows the feedback path for the noninverting opamp configuration.. but for a real voltage source output (i. with a series resistance other than zero). Kovacs ©1997 .
Chapter 4: OPERATIONAL AMPLIFIERS
APPENDIX TO CHAPTER 4: A Bit More On Miller
Why isn't it called the Budweiser Effect? Bobby Twistoffski Stanford Senior, Electrical Engineering
WHY DO WE CARE?
• The Miller Effect can turn a small, parasitic capacitance sitting across an amplifier into a real problem! • The Miller Effect is basically a multiplication of the capacitance's value, as seen at the input of the circuit by a factor related to the gain of the amplifier. • In other words, we can get a HUGE effective capacitance at the input of a highgain amplifier for having only a small (i.e. a few picoFarads) capacitance across it (from input to output)... • We can use the Miller Theorem to help analyze circuits where we have an impedance across an amplifier. • Again, that assumes that the gain of the amplifier is not changed by having the capacitor placed across it (if it is changed slightly, we can still use the Miller approximation).
G. Kovacs ©1997
Page 57
Chapter 4: OPERATIONAL AMPLIFIERS
• Previously, we saw that the Miller Theorem lets us "convert a circuit with an impedance across an amplifier in to one that is easier to analyze because we simplify it into two "separate" circuits! • A KEY assumption is that the gain of the amplifier itself is not changed by having the impedance placed between its input and output...
i
1 Z f
i2
Vin
Amplifier with Gain=A
V
out
Vin
Amplifier with Gain=A Z Z2
V
out
i1
1
i2
• We showed that,
Z1 = Zf 1A
Z 2 = Z f A 1A
and for an "ideal" amplifier, we can let A > ∞ and we get,
Z1 → 0
Z 2 → Zf
• We can replace an impedance across the amplifier to two impedances to ground (usually easier to work with).
• If our impedance was actually a capacitance, for a given frequency, that would mean that C 1 would tend to increase (to decrease the impedance at the input to zero) and C2 would tend to decrease (to increase the impedance at the output to infinity).
• It's easier to call them Cin and Cout to keep things straight, and that is what we'll do...
G. Kovacs ©1997
Page 58
Chapter 4: OPERATIONAL AMPLIFIERS
• Now actually do it with a capacitor to convince yourself...
i1
Cf
i2
V
in
Amplifier with Gain=A
V
out
V
in i 1
Amplifier with Gain=A
V
out
Cin
C out
i2
• The effective impedances are,
Z Cin
1 Cf S 1 = = 1 − A C f (1 − A )S
1 −A 1 = Z Cout = C f S 1 − A C 1 − A S f −A
which is the same as saying that there are "scaled copies" of Cf at the input and output of the amplifier,
C in = C f (1 − A)
1− A C out = C f −A
• If the amplifier was really "ideal" and we were not worrying about input and output resistances of the driving circuit and the load, these capacitances might not be a problem... (for example, if an "ideal" voltage source with zero output impedance was driving Cin , the fact that Cin was large would not matter)
G. Kovacs ©1997
Page 59
Chapter 4: OPERATIONAL AMPLIFIERS
• Now look at a "realistic" amplifier (input impedance is less than infinity and output resistance is greater than zero)...
Vin i 1
Amplifier with Gain=A
R out
V out
Cin
Rin
C
i
2
out
• Now we have created two RC time constants (e.g. two lowpass filters!), one of which (input) turns out to usually be dominant in terms of the overall frequency response...
τ in = R in Cin = R in C f (1 − A )
1 − A τ out = Rout Cout = R out C f −A
• From this you can see that we have a relatively small contribution from Cf at the output and a relatively large contribution at the input... • For typical amplifiers, R in is large and R out is small, so we get a large time constant at the input and a small one at the output. • In other words, the input pole will dominate the frequency response since it will take effect at a lower frequency that the output pole. • Still more realistically, we have to take into account the output resistance of the circuit driving our amplifier and the load resistance that it has to drive... now we could compute more accurate time constants (and thus the frequency response).
Rs
Vin i1
Amplifier with Gain=A
R out
V out
Cin
Rin
Cout
i2
Rload
G. Kovacs ©1997
Page 60
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
This amplifier goes to eleven! Nigel, member of the rock band “Spinal Tap.”
1: OBJECTIVES
• To learn about: The similarities and differences between the three basic BJT amplifiers. The basic analysis and design approaches to the commonemitter amplifier.
READ S&S Sections 4.10  4.12
2: OVERVIEW OF SINGLESTAGE BJT AMPLIFIERS
• There are three basic BJT amplifier configurations, named depending upon which of the three terminals (B, C, or E) is common to the input and output of the amplifier (commonbase, commoncollector, or commonemitter, respectively). • In practice, it is the terminal that is connected (either directly or through a capacitor bypassed resistor > directly for AC signals) to a power or ground voltage that is the “common” terminal. • Each of the three possible configurations has unique characteristics that make them useful in different situations. The table below compares the three basic configurations, as well as a variant of the common emitter amplifier. • The common emitter amplifier is the most common version, and is an inverting amplifier (that is, the output is 180° out of phase with the input). In many cases, a resistor is placed between the emitter and ground (usually bypassed with a large capacitor). This emitter resistor, RE, serves to stabilize the bias point of the transistor using local feedback
G. Kovacs ©1997 Page 61
radio frequency signals arriving via a coaxial cable of characteristic impedance 50 Ω). including frequency response. gain is traded off (reduced). still a common emitter type. is noninverting. • The relative merits of each amplifier type is discussed below. Kovacs ©1997 Page 62 . like the common collector. This can be a problem if the signal source is high impedance (in other words if the signal source cannot easily drive a lowimpedance load). is large (approximately ß+1). The common base amplifier is generally very fast.Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER (this is discussed in detail below). The main difference between the common base and the other amplifier types is that the input impedance is very low. since its noninverting gain is almost exactly one. but it can be very useful if the signal source is low impedance (e. It can provide high voltage gains. The current gain. however. This form of amplifier is. • The common collector amplifier is often used as a buffer amplifier. and this circuit is often used to add power to a signal. • The common base amplifier. Parameter Inverting? Av Ai Rin Rout Typical Schematic CE YES HIGH HIGH MEDIUM HIGH Vcc CC CE with RE YES HIGH (reduced) HIGH (reduced) MEDIUM (increased) HIGH (increased) VCC CC NO LOW HIGH HIGH LOW VCC CB NO HIGH LOW LOW HIGH VCC VEE G. however. like the common emitter amplifier.g. In order to obtain this stabilization.
1.6 mA = 1. its properties and.0.. You need to be quite familiar with it. but simple) V CC RB RC vo vs This capacitor is generally quite large. ADD AN OUTPUT COUPLING CAPACITOR AND A SOURCE RESISTANCE! G. VCC = 15V RB = 50K R C = 500 Ω β = 100 I B = 15 .1Ω VT 25.1 500 = 550 NOW MAKE IT MORE REALISTIC.9 mV Av = .. its variations. Kovacs ©1997 Page 63 .286mA 50K I C = 100 I B = 28.. USING THE PREVIOUS EXAMPLE (not the best biasing. eventually.7 = 0. Av = vo v o = = − gm RC v s v be 1 gm = IC = 28.6mA vo = g m vbeRC vbe = v s • One can compute the voltage gain Av.. the common emitter amplifier is by far the most common singletransistor amplifier configuration.Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER 3: THE COMMON EMITTER AMPLIFIER • As mentioned above. and serves to block DC voltages from affecting the bias point. It is important to note that it is an inverting amplifier.
Av = RB  rπ vo = − g m RC vs R S + R B  rπ • Note that with a source resistance. the effects of R S are reduced. there is a voltage divider at the input. use a fairly large capacitor here) RB RS RC vo vs output coupling capacitor blocks DC but is intended to pass AC signals (i..e. the lower the gain will be. reducing the gain. If the input resistance of the amplifier (RBrπ in this case) is made very large.. short the caps and VCC to ground and you get. R S.. vs Rs R B r π v be = v s RS + RB  rπ v o = −g m v be RC Note that it is I c that links the DC and AC models through gm. use a fairly large capacitor here) • To make an AC smallsignal model. G. The larger the source resistance is. Kovacs ©1997 Page 64 .e.Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER VCC input coupling capacitor blocks DC but is intended to pass AC signals (i.
vbe << v T = 25. DISTORTION)! • A potential problem here is that you have an overall gain that is a function of ß! • You cannot control that (unless you make your own transistors.e..6 Ω 50Ω + 50K  90. in this example.e.6 mA = 1.1.9 mV Av = . but would reduce gain..6 Ω gm 1. that can be fixed using a better biasing scheme.7 V. and then only somewhat) and it varies with temperature! • To get around that.6 Ω 90.6 50 + 90.Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER TRY THE EXAMPLE AGAIN WITH A 50 Ω SOURCE RESISTANCE (typical for a laboratory signal generator): RB = 50 KΩ RC = 500 Ω IC = 28.1 500 = 356 r π = β = 100 = 90. • It turns out that a potentially bigger problem is that gm varies with IC which varies with ß!. gm would fall and rπ would increase to help accomplish this (of course.. if a smaller I C were used. you can make sure the other terms in the input voltage divider “dwarf out” RS.9 mV) Note that V C ≈ 0.6 mA β F = 100 gm = IC = 28. gain would also fall unless R C were increased to compensate).1.1 50K  90. and the transistor is on the edge of saturation.6 Not Bad! • REMEMBER this is SMALL signal only! (i. • Things can get nonlinear for larger signals (i.1 500 ≈ .. Kovacs ©1997 Page 65 . Running it at a lower current would raise VC and allow for a larger voltage swing at the output.1 Ω1 VT 25... G.
only about 10% of the total current flowing in the base circuit is shunted to the base as IB) to ensure bias stability.e. it is more of a rule of thumb than an absolute. however! • The basic idea is to set up the bias resistors to provide a nearly constant voltage at the base of the transistor. this makes sure that V B is much larger than any changes in V BE expected due to temperature variation (VBE decreases by 2 mV/°C).THE BASICS • It is necessary to keep V B more than one diode drop below VC even at maximum signal swing to prevent saturation. the emitter resistor..Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER 4: CLASSIC BIASING SCHEME FOR CE AMPLIFIERS VCC RB1 VB vin I B1 IB RC vo R B2 I B2 RE VEE • Typical VCC = +12 or +15 V. • As a general rule. the AC gain of the amplifier is reduced. or 15 V 5: BASE CIRCUIT DESIGN STRATEGY . typical VEE = 0. serves (through localized negative feedback) to trade some of the available gain of the circuit for stability in IC. If RE is not bypassed.. • Choose the overall magnitude of R B1 and RB2 to ensure that the current through them is about 10 . Kovacs ©1997 Page 66 . the AC gain is unaffected by RE (since large bypass capacitors "become shorts" in the AC model). choose RB1 :RB2 ratio to set V B around onethird of the way up from VEE (or ground) to V CC. 12V. which in turn stabilizes gm. G.20 X greater than the base current (i. • As explained below. The method presented here is described in Sedra & Smith and is commonly used.. If RE is bypassed using a capacitor (as shown above. independent of its ß. R E.. and commonly done).
e. and capacitors become shortcircuits.and devicedependent parameters changed). a bypassed RE provides DC feedback only. • In general.. RE is present in the CE circuit shown above to help stabilize it. Kovacs ©1997 Page 67 . • RE allows for stabilization of IC by feeding back an error signal if IC increases. RS vs R B2 RE CE G. to "short out" RE for AC signals but not disturb RE's effect of stabilizing the bias point (which only matters at DC). is placed in parallel with the emitter resistance. Thus. • The previously discussed biasing schemes (without RE) are really not very good because: 1) β varies from transistor to transistor (think about what would happen in production!) 2) gm and β vary with IC and temperature! Remember gm = IC VT • GOOD biasing would make the bias current stable (i. CE.Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER 6: LOCAL RESISTOR FEEDBACK THROUGH AN EMITTER • As mentioned above. etc. keep I C constant as the temperature. since variations in I C cause variations in gain. Vcc shorts to ground. VCC R B1 RC vo As usual for smallsignal analysis. The error signal reduces VBE enough to reduce IC back down to where it should be. a bypass capacitor. RE..
R B r π v be = v s RS + RB  rπ v o = −g m v be RC AV = vo Vs This is true as long as the bypass capacitor is large enough and/or the frequency is high enough!!!! RB  rπ = − g m RC R S + R B  rπ • The details of designing the whole thing.. • Otherwise use a Thévenin equivalent for the input circuit and solve the loop equation to compute IB (see below)... including the base circuit and RE are presented below after some discussion of analysis and the effects of nonbypassed RE! 7: BASE CIRCUIT ANALYSIS STRATEGY • This section is concerned with how one analyzes the base circuit to check that it is designed correctly or to understand the operation of a circuit.e.. is the transistor in active mode versus cutoff or saturation). then VB is set by a simple voltage divider.Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER • With RE shunted by a large enough capacitor.... • If I Bis much smaller than the current running through RB1 and RB2 . VCC VCC RB1 VB V BB RB VB Clearly V B = VBB if the base current is zero! Then you just have a voltage divider. The equivalent resistance R B is useful to understand how the base voltage drops if the base current goes up.. compute the currents. then one can assume a VBE.. Kovacs ©1997 Page 68 . RB2 ASSUME VEE = 0 V V EE V VBB = V CC EE RB2 RB1 + RB2 RB = R B1RB2 Assuming VEE = 0 G. and check the VBE assumption (i. this gives exactly the same gain equations as we saw without RE.. Rule of thumb: this is a good approximation when the base current is <10% of the total basecircuit current.
. the AC model IS affected by RE). this makes the overall circuit analysis much simpler. 2) Do not assume anything.e. Kovacs ©1997 Page 69 . • You can cleverly handle this by “pulling” RE into the hybridπ model and redefining gm and rπ as gm' and rπ'. since we can redraw the AC equivalent circuit with the emitter of the hybridπ model grounded.. the base current must be small relative to the current through the bias resistors! 8: UNBYPASSED EMITTER RESISTANCE FOR AC AND DC FEEDBACK • With RE not bypassed... 1) Assume IB is small relative to the total current through the bias resistors and use a simple voltage divider to compute VB.. COMPARE THE RESULTS! > The point is that if you assume a simple voltage divider.21 • Do the example two ways. the situation is changed so that there is a resistance from the emitter of the hybridπ model to ground (i. avoiding the need to write loop equations! B C B C rπ E RE gmvbe r' π g' vbe m E G.Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER BIAS CIRCUIT ANALYSIS: SEE SEDRA & SMITH EXAMPLE 4.....
Kovacs ©1997 Page 70 ... vin = vbe + 1 + β i inRE = vbe + 1 + β i BRE vin = vbe + 1 + β vbe RE = vbe + 1 + gm v beRE rπ rπ ie G.. so it does not enter into our calculations... • Take note of the fact that the (ideal) signal source is directly driving RB.. • Calculate the gain and input resistance and compare to CE amp without RE.. • Ignore the transistor’s output resistance r o which would be in parallel with the gm current generator. vin = vbe + 1 + gm v be RE rπ (the second term is the voltage across RE) sum of conductances seen looking into base terminal • The other way to get this equation is from the base current.Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER VCC RB1 VB vs RC Rin i in + vs vin R B2 RE  RB + vo gmvbe RC RE rπ  vbe SIMPLIFYING ASSUMPTIONS / OBSERVATIONS: • Ignore the source resistance RS (not shown above) for the analysis.
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER • Solve for vbe . one can look at the gain. vbe = 1 + vin 1 + g m RE rπ ≈ vin 1 + gm RE vin vin iin = vbe = ≈ rπ r π + 1 + gmr π RE r π 1 + gm RE • Therefore. we can define a new gm “taking into account” RE. one can now redefine the input resistance of the transistor “taking into account” RE.gm vbe RC • Substituting in v be ≈ v in . Kovacs ©1997 Page 71 .. 1 + gm RE gm v o = −R c v in 1 + gmR E this term is defined as g'm • Finally.. vo = . AV = gm vout RC = −g'm R C = − RC ≈ − v in RE 1 + g m R E if g m R E >> 1 G. plugging into the original CE amplifier gain equation. Rin ≡ vin = r π + 1 + β RE ≈ r π 1 + gmRE ≡ r 'π iin • Similarly.
.. they should be included in the gain equation as shown below. ' ' RB  rπ ' R B r π g m vout = − g m R C = − R ' ' v in R S + R B  rπ R S + R B r π 1 + g m R E C AV = ' R B r π RC ≈ − ' R S + R B r π R E if g m R E >> 1 • The g m' equation should look like the feedback equation. because that is exactly what it shows. one can effectively replace the original hybridπ model with a new one with an INCREASED rπ (now called rπ’) and a DECREASED gm (now called g m’). G.Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER • Thus. Kovacs ©1997 Page 72 . treating the transistor as a transconductance amplifier. we are cutting down its "open loop" transconductance gm by a factor (1+gmRE). g'm ≡ gm 1 + gmR E GAIN = A A = AMOUNT OF FEEDBACK 1 + Aβ • What this means is that.. ' rπ ≡ rπ (1 + g m RE ) g'm ≡ gm 1 + gmR E • If the input divider losses are significant...
to simplify the AC equivalent circuit. the gain of which is then obtained using the standard equation for the common emitter amplifier. • The same approach works if you have an IMPEDANCE in the emitter circuit instead of a resistance. • It usually makes sense to use the simplifying assumption up front. G. Kovacs ©1997 Page 73 .Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER • The feedback situation for this circuit is illustrated conceptually below: vin + Negative Feedback vbe RE gm ic • Later in the notes. the tools to handle general feedback cases are presented.
Kovacs ©1997 Page 74 .Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER VCC RB1 VB vs RC Rin i in R B2 RE + vs RB vin  + vo gmvbe RC RE rπ  vbe Rin i in + vs vin  RB + ' rπ  vo ' gmvbe vbe RC G.
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
9: EMITTER DEGENERATION
• “Emitter degeneration” describes what happens to the DC transfer function of a CE amplifier that has an unbypassed RE.... (note that in the active region, the slope is RC/RE if g mRE>1)
VCC
R
B1
RC vo
VB vs
dc
R B2 RE
0 V (ground)
VCC 
RC V CC RC + RE
RE V + V CC BE RC + RE
• Looking at the circuit, we can see the three cases much more clearly... For mode, it looks like this: VCC
active
RB1 VIN
+ 
RC VOUT
VB
R B2
RE
G. Kovacs ©1997
Page 75
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
• For cutoff, the circuit looks roughly like this: VCC
RB1 VIN
+ 
RC VOUT
No connection!
VB
R B2
RE
Of course, leakage currents DO flow, but this is roughly the case...
• For saturation, the circuit looks like this: VCC
RB1 VIN
+ 
RC VOUT
VB IB R B2 RE
+ 
≈ 0.2 V
• VCE becomes VSAT (which is roughly 0.2  0.3 V in practice) when the transistor saturates... • The transistor is now delivering the maximum IC it can, so IC remains at ICSAT
G. Kovacs ©1997 Page 76
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
• Further increases in IB simply go directly to the emitter circuit, raising VE • Because V CE is now fixed at VSAT , raising VE ends up raising VC (VOUT) • This means that as VB (VIN) goes up, VE just tracks it, after losing a forward biased junction drop VBE(SAT) which is between 0.7 and 0.8 Volts. • That last point explains why the slope of the transfer function becomes one in saturation... the input voltage VB is essentially shorted to V E, which is essentially shorted to VOUT! • This means that VOUT just tracks further increases in VIN (certainly, if VIN is an ideal voltage source, it can drive VOUT just fine!). • Of course, when saying things are "shorted," it is assumed that we can ignore the small voltage drops (0.7 V between B and E, and 0.2 V between C and E). • Note that with 0.2 V between collector and emitter, the BC junction voltage is actually about 0.5 V, so the BC junction is not fully forward biased. Also, the extra base current flowing into the transistor does not flow out the collector because it is at a higher potential than the emitter! • Now it is worthwhile to modify our previous Spice deck (our design example) for timedomain analysis and put in a single cycle of a BIG sinewave (7 V!) to look at what happens when the amplifier hits the limits of its voltage transfer function (cutoff and saturation)... EE113 Example Degenerated CommonEmitter Amplifier *dc components Vcc VCC 0 12 R1 VCC VBG 39K Q1 VCG VBG VEG TRANSMODEL R2 VBG 0 18K RC VCC VCG 680 RE VEG 0 1.8K *.MODEL TRANSMODEL NPN (IS=1.3E14) *ac components Vss 0 Vin sin(0 7 1000) RS Vin VX 1 The asterisk in front of this line C3 VX VBG 100UF removes the bypass capacitor *C6 VEG 0 100UF from across R E ..... C5 Vout VCG 100UF RLL Vout 0 50K .MODEL TRANSMODEL NPN (BF=150 IS=1.3E14 + TF=.9N CJE=6P CJC=5P) *Time Domain Response .TRAN 100nS 1mS .PROBE .end
G. Kovacs ©1997 Page 77
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
• Here is the result, showing input and output signals... NOTE THAT THE SPICE DECK CONTAINS AN OUTPUT COUPLING CAPACITOR, which explains why the ACcoupled output is centered around 0 V while the collector voltage si swinging near to + 12 V!
cutoff (clipping)
saturation
G. Kovacs ©1997
Page 78
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER
• Below is an expanded view of the output signal (ACcoupled)...
2.0V
CUTOFF (maximum positive swing)
1.0V
0.0V
1.0V
2.0V
SATURATION (slope becomes +1)
3.0V 0s V(Vout) Exit Add_trace Display_control 0.2ms 0.4ms Time 0.6ms 0.8ms 1.0ms
Remove_trace X_axis Y_axis Plot_control Macros Hard_copy Cursor Zoom Label
• Now you can see why we bother looking at the transfer function... it predicts the kind of distortions you see when you overdrive an emitterdegenerated (unbypassed RE) commonemitter amplifier.
G. Kovacs ©1997
Page 79
. VC may be considerably lower if the current (and gain) is increased.V EE 3 VC ≈ 2 (V − VEE ) NOTE that this is not a design constraint! 3 CC Note that the VC value specified is the quiescent (no input signal) value). typical VEE = 15 V (other common voltages = +/. VB = 1 VCC . let’s go through the thinking behind it and then a design example using the methods of Sedra and Smith.) and want maximum output signal swing...12 V) • The rule of thumb approach is to design for the voltage between the power supplies to be split (roughly) into thirds.. 2) Then choose the bias resistors RB1 and R B2 so that the 1/3 rule is obeyed and so that the base current is in the range IB ≈ 0.e.e.Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER 10: THE BOTTOM LINE ON DESIGN OF CE AMPLIFIERS • Typically you design to meet a power dissipation specification (i. the maximum I C tolerable for a reasonable battery life. VCC VCC +vo VC vo GROUND RB1 VB vin I B1 IB RC vo vo 1 3 1 3 R B2 I B2 RE (usually bypassed) VB 1 3 VEE VEE • Typical VCC = +15 V. 1) Assuming that you know the desired DC collector current.1 IB1 so that the “voltage divider” assumption is reasonably close (i. you compute the resistor values (for DC) that give you this current. Do not try to force it to that value.. the base current doesn’t change VB much).. Kovacs ©1997 Page 80 . G..
Av = RB  rπ vo = − g m RC vs R S + R B  rπ and for when RE is not bypassed.7V) will be dwarfed by VBB being so much higher so that IB will not change much. • The approach you take depends on which of these specifications are given and which are most important. 11: THE ART OF CE DESIGN (AN INTRO) • Which ever approach you take to designing CE amplifiers. The term on the right is simply the Thévenin equivalent of the bias resistors as “seen” from the emitter (remember that dividing by (β + 1) “transforms” resistances from the base circuit to the perspective of the emitter circuit). etc. specific collector (or emitter current)..Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER Rules of thumb for this are: VBB >> V BE This ensures that small variations in VBE (near 0. The equation for the case where RE is bypassed is. if the gain was specified. Remember that I E = VBB − VBE RE RE >> RB β+1 This ensures that RE is large enough to provide adequate local feedback to stabilize IE (and hence.. • For example. ' ' RB  rπ ' R B r π g m vout AV = = − g R = − RC ' m C ' v in R S + R B  rπ R S + R B r π 1 + g m R E ' R B r π RC ≈ − ' R S + R B r π R E if g m R E >> 1 Page 81 G. IC).. you will typically be designing them to meet certain specifications. power dissipation (quiescent). Kovacs ©1997 .. you would probably start with the voltage gain equation for the configuration you wish to use to get a feel for the constraint. such as gain.
. so you need to be careful to set g m to an appropriate value by choosing IC. In this case. • At this point we won’t worry much about designing to take into account the frequency response of the BJT amplifier (this is covered below). followed by the design of the bias circuitry. substitute the nearest “real” component values. you would derive exact component values using a design method you choose (see below).RC RE • For a nondegenerated CE amplifier (RE bypassed by a large enough capacitor). these loads do not affect the bias point. you typically start by determining the maximum allowable collector current. Av ≈ . Typically.. For capacitivelycoupled amplifiers. and simulate the circuit using SPICE before building a prototype. you can adjust its value to control I C gm = IC and hence gain. • If you have a power constraint. IE = β + 1 I C β (That makes sense because IE is larger than IC. Kovacs ©1997 Page 82 . REMINDER: You should always remember these relationships (they are very handy and we use them below!). • You may need to consider the effects of load resistances on the overall gain.. and hence the maximum available gain.. the gain is gmRC. IE = β + 1 I B and IB = IC β thus..Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER • You then look to see what determines the gain. knowing that.) G. if gmRE >> 1. VT • This is where your design begins in this case. with the collector emitter circuit. If R E is bypassed. it is purely the ratio of the resistances (neglecting the effects of the input divider).
. you get..1 IE. RB2 1= 3 RB1 + RB2 • Combining these.. G. • If you don't..this is RB2 1 VCC . • To select RB for bias stability (resistant to changes in β and temperature.Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER 12: BIAS CIRCUITRY DESIGN • Try to choose RB1 and RB2 so that IB is small compared to the current flowing through RB1 and R B2.1 I E • Substitute VCCVEE for VCC in the above equations if dual power supplies are used. you can do it directly.VEE = VCC . if you have I E.1 1 + β I B RB1 + RB2 • To set VB at roughly 1/3 of the supply voltage.1 IE = 0. • Sedra and Smith suggest a rule of thumb for selecting the bias resistors that is to set the voltage divider current to 0. Kovacs ©1997 Page 83 . . a rule of thumb is to set the bias network (RB1 and R B2) up so that approximately 10% of the total current through RB1 and RB2 goes into the base of the transistor. IBIAS ≈ VCC = 0....VEE 3 RB1 + RB2 RB2 = VCC 0. • So. you can calculate IE from IB. you can write.3 I E and RB1 = VCC .RB2 0.. T).
.. G.the larger R C. so VCE = VC. Resistor load line with slope =  1 RC IB IC V CC RC ic V CE V CC vce • The slope of the resistor load line controls the gain . • "Wiggling" the base current moves you up and down along the constraint of the resistor load line. but effectively there are an infinite number of them..Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER 13: REVISITING LOAD LINES • The load line approach lets us solve two simultaneous equations: 1) the IC versus VCE curve of the BJT (remember that only particular curves are plotted for specific IB values.. it is really a 3D plot) and 2) the resistor's I C versus VCE line due to the collector resistor (Ohm's Law). the shallower the slope of the line and the larger the output signal swing. Kovacs ©1997 Page 84 . generating a corresponding output "wiggle" in vce and consequently vout! • NOTE that here we assume a simple CE amplifier (no RE).
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER IC Increasing RC V CC RC V CE VCE MAX = V CC • This helps explain the effect of R C on the gain of the simple CE amplifier. G. Kovacs ©1997 Page 85 . A V = −g m R C • Combining two resistive load lines into one (somewhat crowded) plot helps bring it together. where.
this gives rise to the gradually increasing slope of the IC versus VCE curves for increasing I C (see pages 207 ... Kovacs ©1997 Page 86 . • The Early Voltage...208 in S & S). or Qpoint) current!!!. r o = VA IC G. what is not illustrated by the simple load line drawings here is that the slopes of the VCE versus IC curves for the BJT increase with increasing IC. ro.. between the collector and emitter.. represents an effective output resistance. moving to a lower Qpoint also decreases gain.Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER IC IB Intercept = V CC RC ic V CE V CC vce • Note that gm is effectively set by the quiescentpoint (no signal current. discussed later on with respect to BJT current sources.
VCEmin to allow a reasonable swing (e. for a given supply voltage. note that the V CEintercept of the resistor load line is determined entirely by the supply voltage. IC Active Region Saturation Region IB Cutoff Region V CE • This means that. VC..Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER I Slope = 1/ro I V V • Remember that in the above figure.e.g.. 1 V) and observing that the gain equation for a common emitter amplifier is independent of RC. • One can solve for this maximum gain by setting VCE to some minimum voltage.. (the input divider is ignored here). the DC collector voltage. • If I C is increased to increase gm.. there is a limit to the amount of gain one can get from a given transistor and power supply voltage. VCE will decrease) until the transistor is saturated. A V = −g m R C = − IC (V − VCEmin ) 1 R R C = − cc C vT RC vT Page 87 G. the IC versus VCE curves represent equal base current steps between them. will come closer to ground (i. Kovacs ©1997 .
Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER 14: DESIGN EXAMPLE #1: METHOD OF SEDRA & SMITH Specifications: DC power dissipation: PD < 30 mW Power supply: 12 VDC Voltage Gain: 50X Load: Resistive. β = 150 measured) VCC All capacitors are large (i. 100 µF) so they are essentially shortcircuits for AC.e. 50 kΩ Assume RS = 0 Ω Must use 2N2222A Transistor (NPN. Kovacs ©1997 Page 88 . RB1 RS vs R B2 RC vo VB RE RL G.
or quiescent) and the voltage gain both are determined by the choice of current in the collectoremitter circuit. gm = IC 2 mA = = 0.0772 Ω 1 v T 25.0133 mA β 150 G.0 mA. 1/3.639 Ω IE β+ 1 150 + 1 0.7 = = = 1. SO. Av = vo = −g m RC v be and that gm = IC VT In this "quick" design. we see that gm is. IB = IC = 2 mA = 0. S choosing IC = 2. START THERE! 1) Calculate Imax at 30mW/12VDC = 2. 1/3 rule of Sedra & Smith.Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER We know that the total power dissipation (DC.0772 assuming the 1/3. Kovacs ©1997 Page 89 .5 Ω gm 0. This should not be done in general. RE = VB − VBE VB − VBE 4 − 0.002 IC 150 β 3) Now calculate the required base current. we also allow VCE = 12/3 = 4V and can choose RE by. since these resistors will typically reduce overall gain (unless R is zero).9 mV now we can calculate RC (assuming RE is fully bypassed for AC) RC = −A V 50 = = 647. we are ignoring the effects of the input voltage divider (bias resistors).5 mA 2) We know that the gain of this amplifier with RE bypassed (shorted to ground for AC signals) is.
3(1+ β )I B 0.AC DEC 10 1 100MEG ... we can calculate the bias resistor circuit.8 kΩ RB2 = 18 kΩ G.92 kΩ 0.MODEL TRANSMODEL NPN (BF=150 IS=1. and that VB should be 1/3 of the way up from ground to +12 V.3 µA R B1 = 5) Choose “real” component values close to those calculated.3E14) *ac components Vss 0 Vin AC 10mV RS Vin VX 1 C3 VX VBG 100UF C6 VEG 0 100UF C5 Vout VCG 100UF RLL Vout 0 50K . Kovacs ©1997 Page 90 .83 kΩ 0.3 µA VCC VCC 12 − R B2 = − R B2 = −19. and assuming that the current in the voltage divider is 0.9N CJE=6P CJC=5P) *input sweep for Bode plot .92 kΩ = 39.8K *.1(1 + 150) 13.Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER 4) Knowing the base current.1(1 +β )I B 0.MODEL TRANSMODEL NPN (IS=1.3E14 + TF=.1IE.3I E 0.1I E 0. R B2 = VCC VCC 12 = = = 19. EE113 Example Design #1 CommonEmitter Amplifier *dc components Vcc VCC 0 12 R1 VCC VBG 39K Q1 VCG VBG VEG TRANSMODEL R2 VBG 0 18K RC VCC VCG 680 RE VEG 0 1..PROBE ..3(1+ 150) 13.end RE = 1. RC = 680 Ω RB1 = 39 kΩ 6) SIMULATE the circuit using SPICE.
84) to obtain 1.Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER 50 40 30 20 10 0 1.816 545. 1. To do this..4 1 = = ≈ Ω v T β+ 1 v T β RE REvT β +1 we see that the gain can be directly controlled via RE. and choose a real valued RE = 1. we scale the original computed RE (1. 100Mh 42.0Mh 59.0000.41 kΩ. 59.0h 100h V(Vout)/V(Vin) 10Kh Frequency C1 = C2 = dif= 1.639 kΩ) by the ratio of the actual to desired gain (0.. gm = IC β IE VB − VBE 127.5 kΩ and simulate again.372K.335m 42.271 We see that the midband gain is only 43X !!!! Noting that.. Kovacs ©1997 Page 91 .371K. G.
100Mh 50.Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER 60 50 40 30 20 10 0 1. Kovacs ©1997 Page 92 .521m 50.9 is probably close enough! Checking the power drain as computed by SPICE.177E03 Vss 0.000E+00 TOTAL POWER DISSIPATION 2.266 A gain of 50. VOLTAGE SOURCE CURRENTS NAME CURRENT Vcc 2.778K.0000.861 595. 73.61E02 WATTS Looks fine! Build it! G. 1.0Mh 73.777K.0h 100h V(Vout)/V(Vin) 10Kh Frequency C1 = C2 = dif= 1.
we also allow VCE = 12/3 = 4V and can choose RE by (same as before).92KΩ = 39.7 RE = VB .3 1 + β I B 0. the power dissipation constraint in this example still is the best place to start.0772 G.0133 mA • Thus far.92 KΩ 0. First. 4 .0. but the final calculation of RC must be done differently.3 1 + 150 0. VCC 12 RB2 = VCC = = = 19. and we begin by choosing IC= 2 mA. 1/3 rule of Sedra & Smith. you need to compute rπ.0133 mA β 150 • Knowing the base current.19.0133 mA 12 RB1 = VCC . giving gm = 0. 1/3. we calculate the required base current.83 KΩ 0.1 1 + 150 0. we can calculate the bias resistor circuit. and that VB should be 1/3 of the way up from ground to +12 V.Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER 15: WHEN DOES THE INPUT DIVIDER MATTER? • If the input divider is considered upfront in a design where RS ≠ 0. • Then you would refer to the "full" gain equation and note that the input divider values would need to be determined before choosing RC. the calculations have been the same. and assuming that the current in the voltage divider is 0. IB = IC = 2 mA = 0.1 I E 0.3 I E 0.V BE = VB . rπ = β 150 = =1943 Ω g m 0. −50 = A v = R B  rπ vo = − gmRC vs R S + R B r π • Assuming the 1/3.639 Ω IE β+1 I 150 + 1 2 mA C 150 β • Again. Kovacs ©1997 Page 93 .RB2 = .V BE = = 1.1IE.0772 Ω1.
• Choosing "real" values. • This gives the same result as we got by ignoring the input divider!!! RC = 680 Ω RB1 = 39 kΩ RE = 1.622. since then there is only one resistor that is not accurately represented in the calculations when the nearest "real" value is chosen (RC). If RS is comparable. For example. the CE stage itself would need a gain boost of (0. • Also note that rπ is a function of IC.622)1.Chapter 5: SINGLESTAGE BJT AMPLIFIERS AND THE COMMON EMITTER AMPLIFIER • Here there is a choice. so adjusting I C does have an effect on the input divider. You can pick real values for RB1 and RB2 and solve the "full" gain equation for RC (knowing R S). the input divider could have been much more significant. or 1. • The moral of this story is that the way to tell whether or not the input divider is significant is to compare the magnitude of RB1 RB2r π to R S.8 kΩ. If RS is much smaller. the gain gain factor from the input divider would be 0. RB1 = 39 kΩ and RB2 = 18 kΩ.9kΩ 0.61 times. The first choice is usually better. RC = −A v RB  rπ gm R S + R B r π RC = −A v 50 = = 667.0772Ω R S + R B1  RB2  rπ For which you would still choose 680 Ω as the nearest "real" value. • The only remaining task is to choose a "real" value for RE. the attenuation from the input divider can be quite significant. if RS was greater than 50 Ω (for which the input divider gives a gain factor of 0. which would mean that for an overall gain of 50. you can make the calculation using the "accurate" values and then pick "real" ones later. Then the gain equation can be solved for RC (assuming RS = 50Ω in this example).971). and again the nearest value is 1.3Ω 39kΩ 18kΩ 1. Kovacs ©1997 β βv T = gm IC Page 94 .8 kΩ RB2 = 18 kΩ • We would still have had to adjust RE to get the correct gain after simulating the circuit! HOWEVER (!). rπ = G. it can often be ignored.9kΩ RB1  R B2  rπ −1 g m 50Ω + 39kΩ 18kΩ 1. if RS were 1 kΩ. Or.
Estimating frequency response of a CE amplifier and seeing how the Miller Effect applies in that case. How do you WIDLARIZE something? You take it over to the anvil part of the vice. National Semiconductor 1: OBJECTIVES • To begin estimating frequency responses of BJT amplifiers by learning about: The junction capacitances of BJT’s and their effects on frequency response. so small that you don’t even have to sweep it off the floor.. and you beat on it with a hammer. Analog Guru. and it makes me feel a lot better. respectively) as illustrated below: G. The diffusion capacitance of BJT’s and its effect on frequency response.Chapter 6: BJT FREQUENCY RESPONSE Chapter 6: BJT FREQUENCY RESPONSE . I have wasted a lot of time being fooled by a bad component what do I do? I usually WIDLARIZE it. It makes you feel better. for the baseemitter and basecollector capacitances.5 2: JUNCTION CAPACITANCES OF BJT’S • Both of the PN junctions in a BJT have junction capacitances which end up reducing the frequency response of the transistor. A frequency dependent and more complete Hybridπ model. Kovacs ©1997 Page 95 . Bob Pease. READ S&S Section 7. until it is all crunched down to tiny little pieces. • They are often referred to as CJE and C JC (or CBE and C BC..
Chapter 6: BJT FREQUENCY RESPONSE C = εoεRA d This is analogous to a parallelplate capacitor. Kovacs ©1997 Page 96 . G.
Chapter 6: BJT FREQUENCY RESPONSE BIAS DEPENDENCE OF JUNCTION CAPACITANCES • The junction capacitances are (nonlinearly) determined by the bias voltages as shown below.3 and 0. The equation shown below works fairly well for reversebiased junctions. G.3) or abrupt (0.5 depending on whether the junction is graded (0. Kovacs ©1997 Page 97 . Cj = C jo V 1− Vj mj where Cjo is the unbiased junction capacitance.5). Vj is the builtin potential of the junction (often called Ψo). and mj is a constant between 0. the approximately generally used is that the capacitance is double that of the same junction under no bias. • For forwardbiased junctions. • Actual estimation of the junction capacitances can be tricky.
• For NPN transistors. • One can think about it as follows: if there is a certain number of carriers diffusing through the base. electrons are the minority carriers in the base. IC = QB charge in base = τf transit time for that charge to be swept through the base QB = I C τf base width Cd ≡ dQB dVBE = 2 τf dI C = g m τf = g m WB dVBE 2 DB diffusivity of minority carriers in base G. it takes a certain amount of time to adjust that number if the applied signal changes. Time delays in controlling currents are the hallmark of capacitances. Kovacs ©1997 Page 98 .Chapter 6: BJT FREQUENCY RESPONSE 3: DIFFUSION CAPACITANCE • The diffusion capacitance appears between the BASE and EMITTER and is the result of the diffusion of minority carriers across the base.
Chapter 6: BJT FREQUENCY RESPONSE SUMMARY: • AT LOW I C > Cjc and Cje DOMINATE • AT HIGH I C > Cd DOMINATES 4: COMPLETE HYBRIDπ MODEL B rx + v be rπ gmv Cd CJE be This capacitance is voltagedependent only.. • r o is a resistance that models the slight effect of collector voltage on collector current in the active region of operation (the curves are not exactly flat!). E • r x is the ohmic resistance of the base contact and is a few tenths of ohms normally. G. extra current through ro determines slope. • r o is inversely proportional to the DC bias current and is typically tens of thousands of Ohms.. which is discussed again below with respect to current sources.. C BC ro C Both voltage. These should be r o = VA IC I horizontal lines for an ideal current source... Kovacs ©1997 Page 99 .and currentdependent capacitances appear here.. Slope = 1/ro I V V • V A is the Early voltage.
B rx + v be rπ Cπ C BC gmv ro C be E • Typically r x is in series with Rs (source resistance of the input signal generator) and ro shunts RC (the collector resistor. Kovacs ©1997 Page 100 . AVOID NOTATION CONFUSION!!! Cπ ≡ Cd + C JE VBE in SPICE. not shown above because it would be a part of the external circuit).Chapter 6: BJT FREQUENCY RESPONSE • In practice. one needs to study the time constants of the BJT alone. this is in SPICE CBE CJC VBE = CBC 5: GAIN (β) VERSUS FREQUENCY FOR BJT'S • In order to understand the frequency response characteristics of the BJT. the two baseemitter capacitances are lumped together as cπ (or CBE in SPICE). rx + v be  ib C BC Cπ gmv rπ be ro i c G.
Note that Sedra and Smith refers to ωo as ωb.Chapter 6: BJT FREQUENCY RESPONSE β(S) = i c g m v be = ib ib ib v be = 1 + S( Cπ + CBC ) rπ β(S) = g m rπ ic gm βo = 1 = = ib + S( Cπ + CBC ) 1+ Srπ ( Cπ + CBC ) 1 + Srπ (C π + C BC ) rπ LOOKS LIKE A LOWPASS FILTER! β S= βo 1 + S r π Cπ + C BC Hs = 1+ s ωo K OR ωo + s Kω o = ωo + s A 1 where ω0 = 2πfc = 1 = "RC" rπ C π + CBC β β or β o DC β=1 ω o ω t ω A graphical representation of the decrease of β as frequency increases. G. β is 3 dB down at ω o and β = 1 at ω t. Kovacs ©1997 Page 101 .
Kovacs ©1997 Page 102 . ωt (or ft in Hertz) is the frequency at which the current gain (β) of a transistor in commonemitter configuration.Chapter 6: BJT FREQUENCY RESPONSE • A key transistor parameter. 2πf t = ω t ≈ 1 1 = 1 C + C BC τ + CJE + CBC f gm π gm gm • One can increase Ic to increase bandwidth until τ f dominates. The maximum possible ft is then obtained. This is independent of ßo ! S = ωt = βo C + C BC gm π βo . with its output shortcircuited. but then falls off if IC continues to be increased because βo falls off at high current anyway. β ωt = 1 = βo βo = 1 + S r π Cπ + C BC 1 + S βo C + C BC gm π 1 + S βo Cπ + C BC = βo gm solving. drops to unity... DESIGN NOTE: As you increase IC .1 ≈ 1 1 Cπ + C BC gm • One can write this to show the contributions from different capacitance mechanisms. The unitygain bandwidth. ωt . gm goes up. can now be derived. G. so the last two terms decrease.
43 X 102 Hertz > = 5.5 MHz THIS IS ALMOST EXACTLY WHAT SPICE PREDICTS (See Prelab) G.3 X 10 12 gm = 1.43 X 10 2 1.43 X 102 2πf t = ω t ≈ 1 2.35 X 107 93. Kovacs ©1997 Page 103 .Chapter 6: BJT FREQUENCY RESPONSE τ f Dominates g mControls Bandwidth • By inspection of the above.88 X 108 Radians ft = 9. ωo = 1 r π Cπ + CBC (the 3dB frequency for β) TYPICAL VALUES FROM SPICE (EE122 PRELAB) CBE = 2.3 X 10 12 1.2 X 10 11 + 2.2 X 1011 (Cd + CJE(VBE)) CBC = 2. it can be seen that that.
• Summing currents at node 1 vs .g mRCvbe G..v be S CBC RC • Assuming that vo >> vbe. vo ≈ .v S C = 1 + S C v o be BC π be rπ R's • Summing currents at node 2 gmvbe = . Rs v rx + s 1 v be  C BC gmv 2 vo RC rπ Cπ be • For convenience.vo .Chapter 6: BJT FREQUENCY RESPONSE 6: COMMONEMITTER CONFIGURATION AND MILLER CAPACITANCE • The purpose of this effort is to examine the frequency response of the common emitter amplifier and see how the gain of the amplifier makes CBC look really large thanks to the Miller Effect.. can write. It looks like any capacitance in the input circuit shorts out vbe as the frequency goes up.v be + v .vo . vo ≈  gm vbe 1 + S CBC RC • To further simplify. Kovacs ©1997 Page 104 .. use the basic CE amplifier gain equation (assuming 1/Rc >> SC BC). use R'S = R S + r x • START by looking at the circuit..
in effect. but not always for a real amplifier and especially for larger capacitances.g m RC C MILLER • Reminder of Miller's Theorem: Z1 = Zf > Miller impedance at the input of the amplifier 1A Z2 = Zf A 1A > Miller impedance at the output of the amplifier THE CIRCUIT REWRITTEN AS THE MILLER EQUIVALENT • (Note that the output Miller equivalent of CBC has been neglected because it is very small.. Kovacs ©1997 Page 105 .... perhaps external to the transistors!) R 's vo + ≈C BC vs v be gmv be RC rπ Cπ CM  CM = C BC 1 + g mRC • So.... Cπ and CM.Chapter 6: BJT FREQUENCY RESPONSE • PLUG IT ALL TOGETHER. a small input current makes gmRC times as much current flow into the output circuit. this is true for a model. The numerator is the DC gain.. G. This makes CBC seem much larger from the point of view of the input circuit! • The dominant pole (more on this later) of CE amplifiers is determined by the input resistance and the input capacitances.. rπ ' r π + Rs Av = vo ≈ vs 1 + S r  R' C + C 1 + g R π s π BC m C ... also we are ASSUMING that the hybridπ amplifier model's gain is not changed by CBC.
. RC.. • Here (again) is the "complete" hybridπ model for a bipolar transistor: B rx + v be rπ This capacitance is voltagedependent only. Remember that the idea is to convert an impedance across an amplifier into two separate (grounded) ones at the input and output.Chapter 6: BJT FREQUENCY RESPONSE 7: USING THE MILLER IDEA UP FRONT • Here we are going to apply the Miller Effect idea directly to the complete hybridπ model to simplify the circuit analysis.. and the input divider does not affect that! G. also. E • Now throw in a collector resistance. Thus. Kovacs ©1997 Page 106 . for example.. one needs to compute the gain CBC deals with.and currentdependent capacitances appear here. that is CBC). let's neglect r x and ro! The Impedance Across the Amplifier Rs v + v be rπ C BC gmv vo RC s Cπ be The Amplifier • A very important point here is that the gain used in the Miller equation must be the gain seen by the component in question (here. to make it into a common emitter amplifier (no RE). C BC gmv Cd CJE ro C be Both voltage. if there is an input voltage divider due to RS.
however. This is discussed in detail below.. corresponding to infinite gain and a gain of one for the amplifier that CBC straddles. Cout now appears at the INPUT of the next stage. C out = C BC 1+ g m (R C  R L ro ) g m ( RC  RL  ro ) • One can sometimes ignore Cout at the output because if R C is relatively low... it may start to matter! • Also. WE WILL USE THIS TECHNIQUE OF TAKING THE MILLER EQUIVALENT OF C BC MANY TIMES IN THE FUTURE! G. Kovacs ©1997 Page 107 . at high frequencies.. and multiplying CBC by ≈1 and putting it at the output! R 's + v be gmv Cπ CM vo rπ be RC ≈C BC CM = C BC 1 + g mRC • Note that the maximum possible range of the scaled copy of CBC at the output is between one and two times the value of CBC. if we are talking about a multistage amplifier..Chapter 6: BJT FREQUENCY RESPONSE • The internal gain (relative to CBC) is g mRc and one can rewrite the circuit as the Miller equivalent. the time constant due to a few picoFarads is not much.gain) and putting it at the input. by multiplying C BC by the (1 . and its "exact" value is given by. then it can be significant! PLEASE REMEMBER THIS STUFF..
Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE If I tell you to neglect the effects of the junction capacitances.1) THE POINT OF ALL OF THIS: We want to know how to take a circuit and determine its frequency response. How coupling and bypass capacitors affect the (LOW) frequency response of amplifiers by studying a common emitter example.2.6 REVIEW: How to make Bode Plots (S & S Section 7. Kovacs ©1997 Page 108 . An overall goal of this is to be able to analyze a circuit and then know how to DESIGN a circuit with the frequency response we need! G. 7.4 (don’t worry that it’s done with a FET!). Shortcircuit and opencircuit time constants for approximating the response of amplifiers. little man! Arnold Schwarzenegger as RoboProf 1: OBJECTIVES • To learn about: The generalized frequency response of capacitorcoupled amplifiers. READ S&S Sections 7. and 7. you’d better just do it.
• For both cases. • It is important to note that one can look at the transfer function of an ACcoupled amplifier as the product of the highpass.state = SC 2πfC Thus. midband.” We typically talk about the “midband gain” of an ACcoupled amplifier. and lowpass transfer functions. these will have ACcoupled responses. • For the ACcoupled response. the amplifier’s response acts like a highpass filter. G. for sinusoidal steady . the amplifier’s response acts like a lowpass filter. The single transistor amplifiers generally need input and output coupling capacitors so that external DC voltages do not shift the bias point. A S = A M FL S F H S Reminder: The impedance of a capacitor is.Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE 2: GENERALIZED CAPACITORCOUPLED AMPLIFIER FREQUENCY RESPONSE • The two plots above represent DCcoupled (left) and ACcoupled (right) amplifiers. ZC = 1 1 or. near ωL and below. • The (hopefully) flat region in the middle is the “midband. Differential amplifiers (presented below) are DCcoupled. Thus. Kovacs ©1997 Page 109 . near ωH and above. they are opencircuits at DC and shorts at very high frequencies.
ωL ≈ n=1 ∑ N ω2 Pn 2 n=1 ∑ N 2 ω Zn DOMINANT HIGHFREQUENCY POLE • If there is a dominant highfrequency pole (this is the one that acts like a LOWPASS filter). FH S ≈ 1 1+ S ω P1 G. • Note that the number of poles and zeros must be equal (at that point) if the function “flattens out” in the midband! • If there is no dominant pole.Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE 3: DOMINANT POLES • Dominant poles are poles that dominate over the others in the transfer function so that the low.or highfrequency response of an amplifier can be approximated as a firstorder high. the lowfrequency response can be approximated as. (REMEMBER that opamps are DELIBERATELY designed to have a single dominant pole > this is so we will be able to predict their response and use it!). you can make the following approximation. S S + ω P1 FL S = S + ω Z1 S + ω Z2 •••• S + ω ZNL S + ω P1 S + ω P2 •••• S + ω PNL this means that ωP1 is at a much HIGHER frequency than the other poles so that it can DOMINATE. Kovacs ©1997 Page 110 . FL S ≈ • So. the highfrequency response can be approximated as. DOMINANT LOWFREQUENCY POLE • If there is a dominant lowfrequency pole (this is the one that acts like a HIGHPASS filter).. if the “full” lowfrequency response is..or lowpass filter.
assuming that there is a dominant pole (it works pretty well if there isn’t too!). • Depending on which of the two methods you use.. for each capacitor in the circuit. Kovacs ©1997 Page 111 . if the “full” highfrequency response is. voltage sources > SHORT CIRCUIT current sources > OPEN CIRCUIT 2) DO NOT remove or “disable” dependent sources! G. with either shorts or opens. ωH ≈ 1 n=1 ∑ N 1 2∑ 1 2 2 ω Pn n = 1 ω Zn N 4: SHORTCIRCUIT AND OPENCIRCUIT TIME CONSTANT METHODS FOR APPROXIMATING THE RESPONSE OF AMPLIFIERS • This is a very useful method for approximating upper and lower cutoff frequencies.Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE • So.. 1 + S 1 + S •••• 1 + S ω Z1 ω Z2 ω ZNH FH S = 1 + S 1 + S •••• 1 + S ω P1 ω P2 ω PNH this means that ωP1 is at a much LOWER frequency than the other poles so that it can DOMINATE. • Two key rules apply to both methods: 1) DISABLE all independent sources.. except for the one you are looking at. you replace the capacitors in the circuit. Note that the number of poles and zeros must be equal (at that point) if the function “flattens out” in the midband! • If there is no dominant pole.. you can make the following approximation.
Kovacs ©1997 Page 112 . Rio. For each capacitor you set all other capacitances (other than the one you are looking at) to zero (i. they become OPEN CIRCUITS as if they have not yet had any effect on the amplifier's rolloff) and determine the resistance. RiS. seen by Ci (the capacitor you are working with). • Repeat this process for each capacitor. ωH ≈ ∑ i 1 Ci Rio This approach can tell you which of the capacitances in a circuit is most significant in determining the highfrequency response! 4.e. For each capacitor you set all other capacitances (other than the one you are looking at) to INFINITY (i.e. G. • Repeat this process for each capacitor. seen by Ci (the capacitor you are working with).2 SHORTCIRCUIT TIME CONSTANTS FOR LOWER CUTOFF FREQUENCY APPROXIMATION • Again. they become SHORT CIRCUITS as if they have already become low enough impedance to neglect) and determine the resistance. • The upper cutoff frequency is then approximately given by.Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE 4.1 OPENCIRCUIT TIME CONSTANTS FOR UPPER CUTOFF FREQUENCY APPROXIMATION • All you have to do is step through the process of calculating a “local” RC timeconstant for each capacitor in the circuit. all you have to do is step through the process of calculating a “local” RC timeconstant for each capacitor in the circuit.
VCC This capacitance is meant to block DC and pass AC RB1 C C1 RC C C2 vo RS vs VB As the frequency goes up. Kovacs ©1997 Page 113 .” and then compare to the response obtained using the open.. the emitter resistance gets shorted out by this capacitance.... as it gets shorted out.. ωL ≈ ∑ Ci 1 iS R i This approach can tell you which of the capacitances in a circuit is most significant in determining the lowfrequency response! 5: FREQUENCY RESPONSE OF THE CE AMPLIFIER • This circuit should be familiar from previous lectures and from EE122! The point of this section is to GRIND through the derivation of its frequency response “the hard way. that means C introduces a E ZERO!!!! G.. the gain GOES UP.and shortcircuit time constant approximations..Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE • The lower cutoff frequency is then approximately given by. R B2 RE CE since the emitter resistance cuts down the gain.
HERE IS THAT POINT AGAIN. Z in = rπ + (β + 1) ZE (here Rin is for the transistor.Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE ZS RS CC1 RB ZE Z in + gmvbe RC CE vo vs rπ  vbe RE ZS = 1 + S R SCC1 S CC1 ZE = RE 1 + S RECE • REMEMBER from the case where RE is not bypassed. NOT the entire amplifier!) G. GENERALIZED FOR IMPEDANCES! One can "transform" an impedance from the emitter or base side of the base emitter circuit to the opposite side just as one would for just a resistance. an impedance in the base circuit will appear as being (β+1) times smaller when viewed from the emitter terminal. Using this. Similarly. An impedance in the emitter circuit will appear as being (β+1) times larger when viewed from the base terminal. Kovacs ©1997 Page 114 . we can write. that one can take RE into account when looking from the base using the rule that lets you multiply its resistance by (1 + β).
there is a ZERO at ω = 0 (because of CC1).Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE 5.. and ZEROs are what we are typically expecting for low frequency analysis. secondorder transfer function for the input conductance.1 LOW FREQUENCY RESPONSE i in vs ZS RB Z in i in = vs Z 1 + in ZS + Z in RB FROM HERE ON. • Looking at the numerator first.. Kovacs ©1997 Page 115 βRCCC1 RECE S S + . S CC1 1 + S R ECE iin = vs 1 + S C R + r + β + 1 R + R C + S2 R C R C + r C R C C1 S π E E E S C1 E E π C1 E E • A nice. • There is a ZERO at ω = 1 RECE vo = β R i C in vs 1 vo = RECE vs 1 + S C R + r + β + 1 R + R C + S2 R C R C + r C R C C1 S π E E E S C1 E E π C1 E E G. ASSUME R B = INFINITY TO SIMPLIFY THINGS! i in = vs vs vs = = Zs + Z in Zs + rπ + ( β+ 1)Z E 1 + SR SC C1 RE + r π + (β + 1) SC C1 1 + SR EC E • Grinding through the math and regrouping terms..
you can prove it to yourself by applying a test current.. RC can't be "seen" through the current source. R CES = R E  R S + rπ R = R E  re + S β +1 β + 1 G...Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE vo = β RC vs RS + r π S2 + S S S+ 1 CE RE  R S + r π β+1 + CC1 1 RECE 1 1 + RS + r π RECECC1 RS + r π • This is of the form. Kovacs ©1997 Page 116 . intuitively. ALWAYS KILL INDEPENDENT SOURCES! CC1 replaced by a short + gmvbe RC CE vo RB ∞ rπ ZE RE vbe • Note that R C is effectively "invisible" to C E. since it cannot effect v be (that might not be true with feedback!) and hence the current flowing in R E and CE. Begin by computing ωL using the shortcircuit time constant approximation.and opencircuit time constant methods agree. • Taking into account (ß+1) scaling.. (REMEMBER that we assumed RB = INFINITY for simplicity) short C C1.. • For C E.. FL S = S S + ω Z1 S S + ω Z1 = S + ω P1 S + ω P2 S2 + S ω P1 + ω P2 + ω P1ω P2 • This was a lot of work!!!! Compare and see if the short. RS The input voltage replaced by a short.
. (R S R B ) + rπ β+1 ( RS  R B ) = R E  re + β+1 RS CC1 RB ∞ rπ R C1S = R S + r π • These two RC time constants are the same as those found analytically. R CES = R E  • For C C1 short out CE. vo = β RC vs RS + r π S2 + S S S+ 1 CE RE  R S + r π β+1 + CC1 1 RECE 1 1 + RS + r π RECECC1 RS + r π R ES = R E  R S + rπ β +1 RC1S = Rs + r π G. Kovacs ©1997 Page 117 .Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE • If RB ≠ ∞ (or is not large enough to assume that).. one obtains..
and consider an output coupling capacitor CC2 between the gm generator and RC  RL. RS vs CC1 rx + v be gmv rπ CC2 v o be RB ro RC RE CE RL Using the method of shortcircuit time constants. R C1S = R S + [R B  ( rx + rπ )] (INPUT) (remember that the other caps are shorts!) rx + r π + (R B R S ) R ES = R E  β +1 (EMITTER) (remember about dividing by [β + 1] to reflect the input circuit impedances to the emitter circuit!) R C2S = R L + ( R C  ro ) (OUTPUT) The zero introduced by CE is at ω ZE = G.Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE • If you do not assume RB = ∞. Kovacs ©1997 1 CERE Page 118 . 529). consider the load resistance RL. you get the following results (S & S p. include rx and ro. ωL ≈ ∑ Ci 1 iS R i ωL ≈ 1 1 1 + + C C1R C1S C E R ES C C2 R C2 Where.
You do not really need to worry too much yet (at midband frequencies) about the parasitic capacitances inside the BJT. R in For the amplifier! RS + gmvbe RC RL vo vs RB vbe rπ  THIS IS MIDBAND.2 MIDFREQUENCY RESPONSE • Here assume all coupling and bypass capacitors are shorted (i.gm v be RC  R L Rin Av = vo = .. A V (S ) = LFgain(S ) × MBgain × HFgain (S) G.e.g m RC  RL vs Rs + Rin • NOTE that r o is not shown above. CC2. CC1. it can be placed in parallel with RC and R L. Kovacs ©1997 Page 119 .. to take it into account. • Remember that the overall gain is.Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE 5. SO IGNORE CAPS! Rin = R B  r π RB = R B1  R B2 vbe = Rin vs Rs + Rin vo = . They matter at higher frequencies. and CE).
gm v be RC  R L  r o Av = vo R in rπ = ( −g m )( R C  R L  ro ) v s R S + R in rx + rπ G. ro and rx)? THIS IS MIDBAND. CONSIDER RL .e.Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE • WHAT IF WE ADD IN THE “DETAILS” (i. Kovacs ©1997 Page 120 . SO IGNORE CAPS! C BC vo + v be gmv be R C  RL rπ Cπ ro R s R in rx v s RB Rin = R B  r x + r π RB = R B1  R B2 vbe = Rin rπ v rπ + r x s Rs + Rin vo = .
C out is given by. C out = C BC 1+ g m (R C  R L ro ) g m ( RC  RL  ro ) Page 121 G. Cin = C π + C M = C π + C BC 1 . AVQ = vo = . R s R in rx v s C BC + v be rπ Cπ gmv be vo R C  R L RB ro • Note that in some cases. so use the “full” Hybridπ model.AVQ NOTE that S & S use Cµ instead of CBC • This is based on AVQ which we will define as the voltage gain between the two transistor terminals straddled by CBC.g m RC  R L  r o vbe • Thus. we find that the Miller capacitance is given by. 2πf t = 1 Cπ + CBC gm gm (remember that Cπ = C d + C JE) • First.. you know that you want.Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE 5. (IN OTHER WORDS. CM = C BC 1 + gm RC  RL  r o • The output capacitance. it is necessary to calculate the Miller capacitance to obtain a value for the total input capacitance.3 HIGHFREQUENCY RESPONSE • From here up in frequency.. we will start to worry about the details of what’s inside the BJT! Now it matters.. Kovacs ©1997 . NOT SIMPLY THE OVERALL GAIN OF THE AMP!) For this amplifier... you may need to compute the required capacitances for the model using.
Cµ and roRCRL). This is very important to check! • A more general approach would be to use Miller and then opencircuit time constants. R ' = rπ  (rx + R B R S ) and C in = C π + C M = C π + C BC [1 + g m ( R C  R L  ro )] • NOTE that in some cases.e. so we do not need to use the method of opencircuit time constants. ωH = 1 R ' C in Where.Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE • This reduces the input circuit to a firstorder lowpass filter (unless Cout is significant). R s R in v rx s + v be rπ C in gmv be vo Cout R C  R ro L RB The upper 3 dB frequency of the input circuit is often given by. Kovacs ©1997 Page 122 . the upper cutoff frequency may be determined by the output circuit (i. ωH ≈ 1 ∑C R i i io = 1 1 = C in R + C out R out C in ( r π [ rx + R B  R S ]) + Cout ( ro  R C  R L ) ' G.
19) RS = 4 kΩ VCC = 12 V rx = 50 Ω RB1 = 8 kΩ IE ≈ 1 mA RB2 = 4 kΩ β o = 100 RE = 3. Kovacs ©1997 Page 123 .14 .31 kΩ G.039 Ω 1 R B = R B1  R B2 = 8 kΩ 4 kΩ = 2.9 mV gm = rπ = β 100 = = 2.56 kΩ = 1.67 kΩ R in = R B r π = 2. Sedra & Smith (Exercises 7.67 kΩ 2..1 MIDBAND GAIN CALCULATION • NEGLECTING RL.039 Ω 1 v T 25.3 kΩ RC = 6 kΩ Cπ = 13..Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE 6: EXAMPLE FREQUENCY RESPONSE ANALYSIS VCC RB1 C C1 RC C C2 vo RS vs VB R B2 RE CE RL • Following the example on page 530.7.56 kΩ g m 0. AV = vo R B r π = ( −g m ) ( RC  RL ) v s RS + R B  rπ IC 1 mA ≈ = 0.9 pF Cµ = 2 pF RL = 4 kΩ ro = 100 kΩ 6. r o and rx .
248 0.56 kΩ (−0. Kovacs ©1997 Page 124 .67 kΩ 2.2 V/V • There is not much difference between the two gain values.2 Av = 0.32 kΩ vo R B  ( rx + rπ ) rπ = ( −g m ) ( RC  RL  ro ) v s RS + R B  ( rx + rπ ) rx + r π Av = 1.1 • Do the analysis again with "the details.4 kΩ) = −23.56 kΩ (−0.039 Ω1 )(2.248 )(0.32 kΩ 2.34 kΩ ) = −22.039 Ω1 )( 2.981 . G." R in = R B  (rx + r π ) = 2.56 kΩ = (0.039 2.34 KΩ = 22.56 kΩ ) = 1.67 kΩ  (50 Ω + 2.037)(6 kΩ 4 kΩ 100 kΩ ) = 4 kΩ+ 1.0.Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE AV = vo 2.247)(−0. The key to this type of analysis is to know when you can approximate and when you cannot.56 kΩ A V = ( 0.67 kΩ 2.981) (−0.32 kΩ 50 Ω + 2.039 Ω1 )( 6 kΩ  4 kΩ ) = v s 4 kΩ + 2.
56 kΩ) ] = 5.32 kΩ rx + r π + (R B R S ) R ES = R E  β +1 50 Ω + 2.66 kΩ What components control the low cut off frequency? Can you "design" it? ωL ≈ ωL ≈ 1 1 1 + + C C1R C1S C E R ES C C2 R C2 1 1 X 10 6 + 5. use. so must use SC time ωL ≈ Where.56 kΩ + ( 2.67 kΩ  (50 Ω + 2.3 kΩ  = 41.2 LOWFREQUENCY RESPONSE RS vs CC1 rx + v be gmv rπ CC2 v o be RB ro RC RE CE RL • Given: CC1 = CC2 = 1 µF constants).32 K 1 1 + 6 6 10 X 10 41.2 1 X 10 9.66 K Page 125 G.Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE 6. Kovacs ©1997 .2 Ω 100 + 1 R C2S = R L + ( R C  ro ) = 4 kΩ + (6 kΩ 100 kΩ) = 9. 1 1 1 + + C C1R C1S C E R ES C C2 R C2 R C1S = R S + [R B  ( rx + rπ )] = 4 kΩ + [2.67 kΩ  4 kΩ ) R ES = 3. and CE = 10 µF (not infinity.
.3 HIGHFREQUENCY RESPONSE • From here up in frequency.Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE ω L ≈ 188. so one needs to use the “full” Hybridπ model. we will start to worry about the details of what’s inside the BJT. Cπ = 13. Now those junction capacitances matter.039 6 KΩ  4 KΩ  100 KΩ = 184.8 pF • From this.9 pF Cµ = CBC = 2 pF • First. we see that the total input capacitance is. G. ω ZE = 1 = 1 > f ZE = 4. Kovacs ©1997 Page 126 .719 Radians > f L ≈ 432. 2πf t = 1 Cπ + CBC gm gm • In the present example.5 = 2. R s R in rx v s C BC = C µ + v be gmv vo R C  R L RB rπ Cπ be ro • Note (for homework) that you may need compute the required capacitances for the model using.0 + 2.8 Hz CERE 10 X 10 6 3.427 + 103. but RE) 6. we calculate the Miller capacitance to obtain a value for the total input capacitance (this gives the dominant highfrequency pole.) CM = C BC 1 + gm RC  RL  r o (note dependence on load resistance!) CM = 2 X 1012 1 + 0. the frequency of the zero caused by the bypass capacitor on the emitter resistor being finite in size is. we are given the values..7 Hz And.3 KΩ (note that it is not RE' above.
039 Ω −1 )(6 kΩ  4 kΩ 100 kΩ) C out = (2 X 10 −12 )(1.025 X 106 Radians 1.00 KΩ • So.00 K 199 X 10 12 > f H = 800 KHz • Remember that there is a scaled "copy" of CBC connected from the output to ground due to the Miller Effect.34 KΩ 2 pF G.56 KΩ  50 Ω + 2.67 KΩ  4 KΩ = 1.. Kovacs ©1997 Page 127 ..01) = 2.7 pF • The upper 3 dB frequency of the input circuit is given by. for reasonably small gains like those typical in singletransistor amplifiers. what would that time constant be? C out = C BC 1 + g m (R C  R L  ro ) g m ( R C  R L  r o ) = (2 X 10 −12 ) 1 + (0.039 Ω −1 ) (6 kΩ 4 kΩ 100 kΩ) (0. vo gmv be ro R C  R L C out ≈ C BC = C µ R'out = r o  RL  RC = 100KΩ  4KΩ  6KΩ = 2.Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE Cin = C π + C M = 198.02 X 10 −12 • Note that the value of Cout. is nearly the same as CBC. 1 R Cin ' R' = r π  r x + RB  R s = 2.. ωH = Where.34 KΩ ω out = 1 R'outCBC = 1 = 214 X 106 Radians = 34 MHz 2. ωH = 1 = 5.
in this case . G. Kovacs ©1997 Page 128 .Chapter 7: GENERAL AMPLIFIER FREQUENCY RESPONSE • So. make sure the time constant caused by the Miller copy of CBC at the output is not a significant one!) •A "full" analysis for the upper cutoff frequency could also be done using the opencircuit time constants method. but would not provide additional information in this case due to the relative simplicity of the circuit.e. it is pretty clear that the input time constant dominates at 800 kHz! • One can always use the "full" approximation: ωH ≈ 1 ∑ i C i R io = 1 1 = C inR ' + C out R out C in (r π  [ rx + R B  R S ]) + C out (r o  R C  R L ) DO NOT ASSUME THAT YOU CAN IGNORE THE OUTPUT TIME CONSTANT! (i.
admit to doing. Common base amplifier frequency response. based CE and CB stages in combination. “slow. sloppy work on purpose. Kovacs ©1997 Page 129 .S.Chapter 8: THE COMMON BASE AMPLIFIER Chapter 8: THE COMMON BASE AMPLIFIER 22% of fast food employees in the U. The basics of a twotransistor amplifier.7 G. on the READ S&S Section 7. the cascode amplifier.” Harper’s Index 1: OBJECTIVES • To learn about: Common base amplifier analysis.
make a smallsignal equivalent of the circuit on the above right (more general): B rx + v be gmv rπ be C v Rs s ro R C  R L E Drive the emitter! RE G.Chapter 8: THE COMMON BASE AMPLIFIER 2: THE COMMON BASE CIRCUIT VCC VCC RB1 VB RS C RB2 RC vo C C2 R B2 vs C C1 RE • Note that the base can either be tied directly to ground or (more common) it can be biased using the “classic” scheme discussed for the commonemitter amplifier with one difference > one of the bias resistors should be bypassed so that the parallel combination of RB1 and RB2 has "zero" impedance for AC signals! 3: MIDBAND GAIN CALCULATIONS • First. Kovacs ©1997 Page 130 .
vbe is a fraction of v1 combining these gives. Kovacs ©1997 Page 131 . since typical values are less than 50 Ω. rπ + r x β+1 v1 = vs rπ + r x RE  + RS β+1 RE  rπ β+1 rπ + rx β+1 β+1 v1 is a fraction of v s given by voltage division vbe =  v1 =  rπ v rπ + r x 1 similarly. but certainly rx is << rπ). v 1. starting with an intermediate voltage.Chapter 8: THE COMMON BASE AMPLIFIER • Transform the base resistances to the emitter circuit by dividing by (β + 1). RE  vbe = RE  rπ β+1 rπ + RS β+1 vs =  RE  r e vs RE  r e + R S G. vbe =  rπ rπ + r x rπ + r x β+1 vs rπ + r x RE  + RS β+1 RE  • If we assume that rx ≈ 0 (not an unreasonable assumption. R in gmv Rs s Ri v vo be v1 v be + ro R C  R L rπ β+1 rx β+1 RE "B" then compute vbe as a function of vs.
vs • Now for the output stage. this simplifies to. assuming ro is "approximately grounded" or. Kovacs ©1997 Page 132 . v o ≈ −g m v be ( RC  RL  ro ) ≈ −g m v be ( R C R L ) which gives.Chapter 8: THE COMMON BASE AMPLIFIER • This further simplifies if Rs = 0 (i. G. the input resistance is seen to be. • Again neglecting ro and rx . simply very large compared to RC and RL.... R i = R S + R E r e ≈ R S + re • NOTE that we typically do NOT include R S into our expression for RIN because it is not a part of the commonbase amplifier itself!. it would clearly reduce gain. re = rπ r 1 V ≈ π = = T 1 +β β g m I C • This means you can control re and hence the input resistance via IC. NONINVERTING! re g ( R R L ) re + R S m C AV = • If r o were considered in parallel with RC and RL.. AV ≡ vo (R E  re ) g R R = ( L) v s (R E  re ) + RS m C and for RE >> re. don't let this confuse you! • Remember that.e.. the CB amplifier is driven by a very low output impedance stage). vbe = .
if necessary.. if ro is comparable to RC this can be quite significant and reduce gain as well. Ro = RC • One can include ro in parallel with RC for added accuracy. we set vs = 0 and remove RL (since it is not part of the amplifier itself) and see that..Chapter 8: THE COMMON BASE AMPLIFIER R IN = R E  re ≈ re • To determine the output resistance. 4: LOWFREQUENCY RESPONSE VCC RB1 VB + v be gmv rπ be RC C C2 RS C RB2 RL r not shown here C o C1 R B2 vs RE • Here the coupling and bypass capacitors matter! G. Kovacs ©1997 Page 133 .
• NOTE: If you take into account ro in parallel with the current source. it is immediately clear the one terminal of CBC is grounded..Chapter 8: THE COMMON BASE AMPLIFIER • Using the method of shortcircuit time constants (the one for low frequencies). this doesn't change things much. ωL ≈ 1 1 1 + + C RB2 R RB2 C C1R C1 C C2 R C2 where. 5: HIGHFREQUENCY RESPONSE C BC rx Rs vs + v be gmv Cπ rπ be ro RC  R L RE • Assuming rx = 0. Kovacs ©1997 Page 134 . R C2 = R C  ( ro + R E R S  r e ) + R L but since ro is generally >> re.. you get a new time constant for CC2.. This means there is no Miller multiplication of its capacitance! • The circuit is more intuitive if it is redrawn knowing this (r x no longer shown). you don’t need to consider RRB2. G.. R RB2 = R B2 R B1  [ rπ + (β + 1)( R E  R S )] R C1 = R S + R E r e R C2 = R C + R L • Hopefully it is clear that if you literally ground the base.
the extra emitter current that we model in re is "real" as opposed to the capacitor current. • Now we can determine the input pole frequency by setting the signal source to zero. no matter which way you look (base or emitter). which is the same. ω P1 = 1 C π (re  R E R S ) G. • In other words. the extra current is fully modeled by re without modifying Cπ.. vbe. the current is a function of which terminal you look into because at the emitter. • Still another way of looking at it is to consider that the capacitor current is determined by the voltage across it. Kovacs ©1997 Page 135 . which is "imaginary. For rπ and re. the resulting pole would be "corrected" twice and therefore be wrong!)..Chapter 8: THE COMMON BASE AMPLIFIER + v be Rs vs  rπ Cπ gmv be RE C BC ro RC  R L • We can see that the input circuit (after transforming rπ into re in the emitter circuit) looks like this: Rs vs re Cπ RE • NOTE: We don't scale Cπ by (ß + 1) because the extra current flowing due to the current source is already taken into account in transforming rπ into re (if we scaled Cπ as well. you see the extra current from the source." Therefore.
Chapter 8: THE COMMON BASE AMPLIFIER • Similarly. ωH ≈ 1 ∑ C i R io i = 1 C π ( re  RE  R S ) + C BC ( R C  R L ) • For the CE amplifier we had derived (neglecting the output pole at CBC(roRCRL)). ωP1 will be a very high frequency. • The high frequency response is (neglecting ro). the output stage’s pole frequency can be determined (neglecting ro). ωP2 will also be a very high frequency. ωH = 1 = R Cin r π  R B  Rs ' 1 Cπ + C BC + C BC gm RC  R L G. Since CBC is small. ωH ≈ 1 1 = R C in rπ  (rx + R B  R S ) Cπ + CBC [1 + g m ( RC  RL  ro ) ] ' ( ) assuming rx is very small and ro is very large. this simplifies to. Kovacs ©1997 Page 136 . gmv be ro not shown here C BC RC  R L ω P2 = 1 C BC ( R C R L ) • Since re is small.
039 Ω1 VT 25. despite its very low input impedance (≈ r e).9 MHz • Two comments from this: 1) The CB amplifier. Kovacs ©1997 Page 137 .. has very good highfrequency response.6 Ω gm 100 rπ = β = = 2. 2) The secondorder transistor model effects (such as r x) may become significant at high frequencies for the CB configuration.9 mV r e = 1 = 25.9 pF Cµ = 2 pF RL = 4 KΩ ro = 100 KΩ gm = IC ≈ 1 mA = 0.Chapter 8: THE COMMON BASE AMPLIFIER • Take our previous example and compare. RS = 4 KΩ VCC = 12 V rx = 50 Ω RB1 = 8 KΩ IE ≈ 1 mA RB2 = 4 KΩ β o = 100 RE = 3.. G.56 KΩ gm 0.039 Ω1 fHCE = 800 KHz fHCB = 30.3 KΩ RC = 6 KΩ Cπ = 13.
Chapter 8: THE COMMON BASE AMPLIFIER
6: CASCODE AMPLIFIERS = CE + CB
• This is a handy type of amplifier that is a CE stage driving a CB stage...
VCC
RB1
RC
C
C2
vo VB2
CB
R B2
C
C1
VB1
vs
RS
R B3
RE
C
RL
E
• The biasing for both stages is set up using RB1, R B2 , and RB3 as a voltage divider, and RE sets the total quiescent current in the output stage. • In a nutshell, the CE stage's "RC" is the input of the CB stage (≈ re) so the CE stage has a gain of about one and no real Miller multiplication of Cµ... the gain comes from the CB stage. • It is a fast amplifier. • The details are presented in a separate section below.
G. Kovacs ©1997
Page 138
Chapter 9: THE COMMON COLLECTOR AMPLIFIER
Chapter 9: THE COMMON COLLECTOR AMPLIFIER
Although we modern persons tend to take our electric lights, radios, mixers, etc., for granted, hundreds of years ago, people did not have any of them, which is just as well because there was no place to plug them in. Dave Barry
1: OBJECTIVES
• To learn about: • The common collector or emitter follower BJT amplifier configuration (the last one with only one BJT in it!) in terms of “what is it good for...” • The basic properties of the CC amplifier. • The frequency response of the CC amplifier
READ S&S Pages 259  265 and S & S Section 7.8
G. Kovacs ©1997
Page 139
Chapter 9: THE COMMON COLLECTOR AMPLIFIER
2: COMMON COLLECTOR CIRCUIT
VCC
Bias resistors, coupling caps not shown. i in
vs R
vo
E
vs
v be r π
g mv
be
vo R
E
• We know that,
v s = i in rπ + (i in + g m v be )R E = i in rπ + (i in + g m rπ i in )R E = i in rπ + (i in +βi in )R E v s = i in [ rπ + ( β+ 1)R E ]
• The input current is thus,
i in =
vs rπ + ( β+ 1)R E
• Note that the extra current supplied by the g m generator in addition to i in makes RE look like a much larger resistor in series with rπ. • The current flowing in the collector circuit is gmvbe = β Fib • Therefore, the output voltage is given by,
v o = (β +1)i in R E =
G. Kovacs ©1997
(β +1)R E v rπ + (β+ 1)R E s
Page 140
Chapter 9: THE COMMON COLLECTOR AMPLIFIER
which gives a voltage gain of,
AV =
vo (β + 1) RE = v s rπ + (β +1) R E
> Av ≈ 1
(in a voltage divider form)
• BUT since (β F + 1)RE >> rπ
• You can also rearrange the gain equation into,
AV =
RE rπ +R (β +1) E
=
RE re + R E
which may be a more intuitive way of looking at the "voltage divider."
• The circuit is a “voltage follower” or “emitter follower” (i.e. the emitter follows the base voltage) and is similar to the unitygain opamp circuit... • Note that the output voltage “follows” the input voltage but is always one vbe drop lower!!! This only matters in DCcoupled applications, that are generally not of importance here....
• It is can be seen from
i in =
vs rπ + ( β+ 1)R E
that
R in = rπ + (β +1 )R E
• This is typically in the > 100 KΩ range, which is a reasonably high input impedance. • Note that in Sedra and Smith (pages 259  265), the series resistance of the signal source, RS, a parallel input resistance from base to ground, RB, a load resistance R L, and the output resistance of the transistor, ro, are also taken into account in the derivations as shown in Fig. 4.46.
G. Kovacs ©1997
Page 141
Chapter 9: THE COMMON COLLECTOR AMPLIFIER
3: PRACTICAL CC CIRCUIT
VCC
RS vs
RB1 VB
Assume all capacitors > 0 impedance for AC signals.
R B2
RE
RL
Ri vs RS R
B
+ v be r π R ie Ro g mv
be
ro
vo RL
R
E
• For the “full” circuit, where R B = R B1  R B2 , one obtains an input impedance of,
R i = R B  ( β + 1)[ re + ( R E r o  R L )] = R B  [r π + (β + 1)(R E  ro  R L )]
G. Kovacs ©1997
Page 142
RO ≡ vx ix • The resistance looking into the emitter. R ie.Chapter 9: THE COMMON COLLECTOR AMPLIFIER • The latter form can be found by inspection.ib . Kovacs ©1997 Page 143 .βib = ..1 + β i b ib = • So Rie is.. Rie ≡ vx ie ie = . G. • The output resistance can be determined by replacing RL with a test voltage source vx and using. Ro ≡ rπ + ( R S R B ) vx = R E  ro  ix (β + 1) • The voltage gain for the “full” circuit can be shown to be. vx r π + RS  R B Rie ≡ vx = ie r π + RS  RB β+1 which gives Ro (by placing Rie in parallel with RE and ro). AV ≡ vo R i RE  ro  R L R i [β +1][R E  ro  R L ] = = v s R i + R S re + [R E  ro R L ] R i + RS rπ + [β + 1][R E r o  RL ] • This includes a first term which takes into account the voltage division at the input and a second term where we have modified the original gain equation to take into account the fact that ro and RL are in parallel with RE in the smallsignal equivalent circuit. can be computed using.
..e. the CC amplifier can be used as a current booster or voltage follower.. but this is generally not the case. RL is like a short circuit). • Thus.Chapter 9: THE COMMON COLLECTOR AMPLIFIER • Again... Think about its operation. (typically what we want when we use a CC stage) • The current gain is quite high... It’s not really CE or CC. vo r + (β + 1)R L i RL R Ai ≡ o = ≈ i ≈ π ≈ (β + 1) vs ii RL RL (R S + R i ) • Note that the rπ/RL term may be significant if it is large relative to (ß+1). and for RL << (RE  ro) (i. Kovacs ©1997 Page 144 .. 4: THE PHASE SPLITTER CIRCUIT VCC INVERTED OUTPUT NONINVERTED OUTPUT • This circuit gives inverted and noninverted outputs from the same input signal. G.. if RS is small. the current gain is approximately. AV ≈ 1..
G. the dominant pole frequency is given by. ωp ≈ 1 (R S + rx )Cµ • An alternative is to approximate the high frequency response using the method of opencircuit time constants... ωp = 1 [( R S Cπ + rx )  (1 + g m [R E R L ]) rπ C µ + 1 + g m [R E  R L ] ] (NOTE THAT rx is not shown in the above AC equivalent circuit and. Cµ + vs R v be r π B RS Cπ g mv be  ro vo R E RL • Based on a “full” analysis (See S & S).... can typically be neglected since it is usually on the order of 10Ω. the input pole can usually be approximated by... also note that R B is considered to be very large here.) • Practically speaking. Kovacs ©1997 Page 145 ..Chapter 9: THE COMMON COLLECTOR AMPLIFIER 5: QUICK LOOK AT CC FREQUENCY RESPONSE • The details of the derivation are presented in S & S Section 7. in fact. but an overview is useful here..8..
Kovacs ©1997 Page 146 .Chapter 9: THE COMMON COLLECTOR AMPLIFIER 6: DESIGN OF CC STAGES • The design process is similar to that for the other singleBJT amplifiers! 1) Assuming you need a basic follower. and then you choose a transistor with the right β). 2) You choose an IC value that you want to operate at and calculate g m. you don’t worry about gain (unless you need a particular current gain. etc. 3) Bias stages are done the same way as before. TYPICAL CC CIRCUIT: VCC RB1 VB vs R B2 RE RL G.
Chapter 10: CASCADED AND CASCODE AMPLIFIERS Chapter 10: CASCADED AND CASCODE AMPLIFIERS Gee Toto.3 NOTE: we will use S & S notation for CBC = Cµ here.7.. 7. G. This will combine what we know about common emitter and common base amplifiers. READ S&S Sections 7.8 and 6... I guess we're not in Terman anymore! Stanford EE at first job interview in Silicon Valley. 1: OBJECTIVES • To learn about: What happens when you put a bunch of amplifiers in series.1 . To study a very useful multitransistor amplifier: the cascode configuration. in other words. look at interstage loading effects (affecting gain and frequency response)...6. Kovacs ©1997 Page 147 .
it is clear that the input and output resistances (or impedances!) come into play by reducing the overall gain. and each had a gain of A. let's work out the gain assuming nothing about the Rin and Rout of each stage. assuming equal gains (A)! G. Looking at the above example. would normally be used. Zi. looking at them as voltage dividers between each stage and between the last stage and the load. but they serve to illustrate the point here. • Note that in practice. • If the amplifiers were ideal (Rout = 0 and Rin = ∞). Kovacs ©1997 Page 148 .Losses Due To Ro2 and second amplification R i2 RL vo = A2 v in R i2 + R o1 R L + R o2 <.Overall eqn.Chapter 10: CASCADED AND CASCODE AMPLIFIERS 2: AMPLIFIERS IN SERIES (CASCADED) Amp #1 Amp #2 Ro2 Vo R i2 Ro1 V in1 R i1 V in2 +  +  AV in1 AV in2 RL • Now it is necessary to consider what happens when amplifiers are put in series. impedances. R i2 v in2 = Av in1 R i2 + R o1 RL v o = Av in2 R L + R o2 <. the overall gain would simply be A2! • In the above example. not resistances.Losses between stages and first amplification <.
and Ro ≈ 50 . realworld offsets can be impossible to trim out! Even if ACcoupled. taking out the effects of the gain stages. • The above equations assume that the individual amplifier gains ("A") do not change with output loading. the fact that noise from each successive stage is added to the noise coming into that stage and amplified. you would have to go to a ONE HUNDRED stages with these specifications before you even lost 1% of the expected "ideal" gain (i.9999) = 0." If DCcoupled.9999) (0.e... Kovacs ©1997 Page 149 . noise from preceding stages gets amplified by each downstream amplifier stage. the equation reduces to the ideal case of AV = A2 for two identical stages if we let the Ro's go to 0 and the Ri 's go to infinity.. G.99 A100). assume a "notsohot" opamp with Ro = 100Ω and R in = 1MΩ. • There are practical reasons why you just can't continue cascading stages "forever. the gains are pretty close to being AN (where N = number of equalgain stages)..100 Ω. other effects would have caused much more trouble (for example. what is the gain with two stages of gain A in series? (assume RL = 1 MΩ too) 1MΩ 1MΩ vo = A 2 = A 2 (0. • By the time you reached that point.Chapter 10: CASCADED AND CASCODE AMPLIFIERS • As expected.. making for nothing but a noise source after a while! • We normally refer all noise to the input of an amplifier. • Just to check that.. • For most opamps since Rin = MΩ to GΩ range.9998 A 2 vin 1MΩ+100Ω 1MΩ+100Ω • That is pretty close to A2! • In fact. on down the line!). to get 0.
. Kovacs ©1997 Page 150 .Chapter 10: CASCADED AND CASCODE AMPLIFIERS 3: THE CASCODE AMPLIFIER • Now. • The CE stage..) • The cascode amplifier is a CE amplifier driving a CB stage. is deliberately set up for a low gain to minimize it! • The gain is obtained from the CB output stage. Cµ + Vbe rπ Cπ Vo g mV be RC RL BECOMES .e. for oscilloscopes. where the Miller Effect could be a problem. G. how do we make a fast amplifier (i.. etc. • Remember that the CE amplifier suffers the most from the rolloff caused by the input pole. not the CE stage.. + Vbe rπ Cπ CM g mV be ≈C µ Vo RC RL where the Miller Capacitance is CM = C µ 1+gmRCRL (Note that Cµ is just another notation for CBC.)? • The idea: Combine the advantages of common emitter (high Rin) with common base (no Miller effect) to get greatly improved performance.
G. C M ≈ C µ1 (1 +1) = 2C µ1 • Thus the Miller Effect is minimized in this configuration and the CE gain is ≈ 1.) • If gm1 = gm2.Chapter 10: CASCADED AND CASCODE AMPLIFIERS Miller Effect Here Rin2 No Miller Here! RC RL vin Low Input R for CB Stage • Let's look at the Miller capacitance at the CE stage.AV]. then... • Remember the input resistance of a commonbase amplifier is. 1 C M = C µ (1 + g m RC  R L ) ≈ C µ1 1+ g m1 g m2 (Note that r o of the transistors is neglected here. Rin2 ≈ r e ≈ 1 gm2 • Now we can calculate the CE gain and substitute into the Miller equation (remember that we multiply the capacitance by [1 . First figure out what the "RC" of the CE stage is (the input of the CB stage!). where AV is the voltage gain). Kovacs ©1997 Page 151 . and would appear in parallel with R C and RL if considered.
where V B1 is 1/4 V cc. I E1 = β +1 I ≈I β C1 C1 I E1 ≈ IC1 = I E2 ≈ I C2 • R1. if the ß values are relatively large. 1/4. R2..Chapter 10: CASCADED AND CASCODE AMPLIFIERS 4: PRACTICAL CASCODE AMPLIFIER CIRCUIT VCC R3 RC CC2 vo Q2 RL CB RS R2 Q1 CC1 vs R1 RE CE • One can use a "1/4. and R3 set the bias points. G. and RE can be used to adjust the current through both transistors. Kovacs ©1997 Page 152 . • R E sets the current through BOTH Q 1 and Q2. In fact. 1/4. 1/4" biasing scheme here. VB2 is 1/2 Vcc and so on. you can assume that the collector currents are equal..
1 THE INPUT SECTION • Assuming r x1 = 0 (See S&S p.. • As is usually the case. CB has no Miller! B2 C µ1 r π2 C1 ro1 E1 g mVbe1 C π2 Cµ2 C2 ro2 vo B1 R1 g mVbe2 E2 C π1 RC RL 4. CM = C µ1 1+gm1r e2 = 2 Cµ1 if r e2 ≈ 1 = 1 and r o1=∞ gm2 gm1 G.2 THE COMMON EMITTER STAGE • The CE amplifier output is looking at r e2.Chapter 10: CASCADED AND CASCODE AMPLIFIERS The AC Equivalent Circuit (High Frequency) NOTE: assuming rX2 is zero here. rπ1 R 1  R 2 v be1 = v s R S + rπ1 R 1  R 2 Rs Vs R1 R2 r π1 4... Kovacs ©1997 Page 153 .. the input resistances just act as a voltage divider... Rs vs r x1 R2 r π1 NOTE: grounded here. 534)...
the CE GAIN ≈ 1 in this case.. Kovacs ©1997 Page 154 . RC  R L is simply re2. AV = . INPUT TIME CONSTANT FOR THE CE STAGE Rs R1 R rx 2 rπ1 Cπ1 CM CM = 2Cµ1 ω1 = [(R S  R1  R 2 + r x1)  rπ1](C π1 + 2C µ1) 1 ≈ 1 (R S  R1  R 2  rπ1 ) C π1 + 2Cµ 1 ( ) • Note that the latter approximation assumes r x ≈ 0. G.gm1 RC  RL and that here..Chapter 10: CASCADED AND CASCODE AMPLIFIERS re2 + Vbe1  r π1 Cπ1 CM ro1 2C µ1 g m Vbe1 • We know that the gain for a nonemitterdegenerated CE stage (note that RE is bypassed by CE!) is. but if R s is very small. the input pole theoretically goes to infinite frequency). rx may be significant (otherwise.
vo C µ2 ro2 RC RL • It is known that since the gain of the CE stage is ≈ 1...Chapter 10: CASCADED AND CASCODE AMPLIFIERS 4.3 THE COMMON BASE STAGE • Now look at the CB Stage to get the overall gain. the nongrounded half of Cπ2 must look at rπ2 "through the emitter. • Since the base is grounded. • We know that the gain of CB amplifier is. A VCB = vo = +g m2 (ro2  R C R L ) ≈ g m2 ( RC  RL ) vx • We can thus write the complete midband gain (neglecting rx and ro) as: AV ≡ vo r π1  R1 R 2 = ( −1)g m2 ( R C R L ) v s R S + r π1  R1 R 2 Input Divider CE stage CB stage G. Kovacs ©1997 Page 155 ." C1 2C µ1 g mVbe1 r π2 B2 re2 C π2 E2 vx g mVbe2 Note that we use re2 because we are looking into the emitter of Q2. Vx ≈ V s.
C1 r o1 vx 2C µ1 r π2 r e2 C µ2 ro2 C π2 E2 g mVbe2 vo RC R L g mVbe1 • You will note that sometimes these other two components are considered. we only found one time constant (the input) so. vx 2C µ1 g Vbe1 m r π2 Emitter of Q2 C π2 g mVbe2 vo C µ2 ro2 RC R L r e2 ω2 = 1 Cπ2 r e2 • NOTE: Use r e not rπ because we are looking into Q2's emitter!!! • Note that we could also consider two lessimportant components in between stages. Cµ1 << Cπ2 and re2 << ro1)... Since it is relatively easy to consider them.4 THE OTHER TIME CONSTANTS • So far..Chapter 10: CASCADED AND CASCODE AMPLIFIERS 4... you can include them OR show that they don't change things much (i. other times they are not (particularly in the text book!). Kovacs ©1997 Page 156 . • In between the stages. in between stages and at the output. let's find the others.e.. ω2 = (C π2 + 2Cµ1 ( re2  ro1 ) 1 ) G. The "complete" equation would be. ro1 and Cµ1 (Cµ1 appears at the output because of the Miller Effect).
. vx r o1 r π2 C π2 g mVbe2 v re2 C µ2 ro2 o g mVbe1 2C µ1 RC R L • Note that ro2 is shown with one end grounded to simplify the analysis. ω3 = Cµ2 1 1 ≈ r o2RCRL Cµ2 RCRL G. but the circuit actually would be something like (Cµ1 and C π2 are shown as opencircuits as well as the current source).e. Kovacs ©1997 Page 157 . you have r o2 connected to a very small resistance (r e) to ground. vx r o1 2C µ1 re2 r π2 re to ground C π2 g mVbe2 ro2 vo C µ2 RC R L • Effectively. • The output stage time constant... so we can just look at it as having one end grounded as shown above (i. Its value doesn’t depend on Rs or the Load (RL). so re2 typically determines the net resistance. ro2 + re ≈ ro2). You may have to look at Cπ2 and Cµ1 to decide. • This interstage time constant typically corresponds to a very high frequency..Chapter 10: CASCADED AND CASCODE AMPLIFIERS • r o1 is typically ≈ 100 KΩ and r e2 is typically a few Ohms.
. but here there is only a 2X Miller multiplication! G. the input pole typically (but not always!) determines fH .. this frequency is lower than ω2 and ω3 .. • So. like the CE stages alone. ω1 = 1 RS R1R2r π2 Cπ1 +2Cµ1 • If Rs is large.. Kovacs ©1997 Page 158 .Chapter 10: CASCADED AND CASCODE AMPLIFIERS • Now recall the input pole.
9 AM = Rπ R1R2 g m RCRL = 23.Chapter 10: CASCADED AND CASCODE AMPLIFIERS 5: QUICK CASCODE EXAMPLE (S&S EX 7. r π = ß = 2. Kovacs ©1997 Page 159 . Q1 = Q2 (identical devices).039 Ω1 VT 25. I E ≈ 1mA.6Ω G.3K 4K R 2 R3 1µF C C2 Q2 RL 4K Q1 CE 10µF RE • Given : Vcc = +15 V..1 RS + R π R1R2 1 f1 = ω 1 = 1 = 8.75 MHz 2π 2π RS R1R2r π2 Cπ1 +2Cµ1 1 = 1 1 f2 = ω 2 = 1 = 447 MHz 2π 2π Cπ2 r e2 2π 13.20.. p. 536) VCC 6K 18K CB 10µF 4K RS vs CC1 1µF 8K R1 3.56 KΩ gm gm ≈ IC = 1mA = 0.9 pF 25. ß = 100 • Doing the analysis is not a big problem .
1 fH = 800 KHz G. what if we let R s go to 50Ω (typical for signal generators). f1 goes way up (around 700 MHz) and the output pole (f 3) dominates in this case! • The message here is DON'T ASSUME you know which pole is dominant !!!! fH = 1 = 8.2 MHz <. f1 is the lowest! • Now.neglect ro 2π 2πCµ2 RCRL 2π 2 pF 4K  6K • Clearly.Chapter 10: CASCADED AND CASCODE AMPLIFIERS 1 1 f3 = ω 3 = = = 33.46 MHz 1+ 1+ 1 f2 f2 f2 1 2 3 • Compare this to the previous CE example! AV = 23... Kovacs ©1997 Page 160 .
Chapter 10: CASCADED AND CASCODE AMPLIFIERS 6: CASCODE AMPLIFIER DESIGN EXAMPLE • This example is analogous to the CommonEmitter design done previously in the notes. 50 KΩ Assume RS = 0 Ω Must use 2N2222A Transistors (NPN. The purpose is for you to compare the two designs for differences in characteristics. or quiescent) and the voltage gain are both determined by the choice of this current. Further. Specifications: DC power dissipation: PD < 25 mW Power Supply: 12 VDC Voltage Gain: 50X Load: Resistive. Kovacs ©1997 Page 161 . ß = 150 measured) VCC R3 RC CC2 V o Q2 RL All capacitors are large (100 µF) so they are essentially shortcircuits for AC CB RS R2 Q1 CC1 V S R1 RE CE Key point: The collectoremitter current running through Q2 is assumed to be equal to that through Q1. Vcc = 12 VDC G. the total power dissipation (DC. DESIGN PROCESS: 1) Pick Vcc unless specified.
1/4 biasing rule to set up bias resistors.8 mA ) R1 VB1 = VCC = VCC 4 R1 + R2 + R3 R1 = R BNET = 16.7kΩ 0. 1/4.VBE = 4 = 3 .7 RE = VB1 .8 mA = 0. VCC . I BIAS > 0.001812 IC ß 6) The required base current is: IB = IC = 0.7 = 1.1I C R BNET = R1 + R 2 + R 3 = VCC 12 V = = 66.7 kΩ 4 Page 162 G.1I E ≈ 0.083 mA Vcc 12 V 3) Select an IC < IMAX and solve for gm2.gm2 RCRL RC = 730 Ω r π R1R2 RS + r π R1R2 1 RC = gm2 . Let IC = 1. Kovacs ©1997 . 1/4.1I C ( 0.27 kΩ IE ß+1 0.1 A RL 5) Use a 1/4.0259 V 4) Let RS = 0 or be too small to matter (for now) and solve for RC: A = .0695 Ω1 0.0.1)(1.0. Let VCE of the transistors = Vcc/4 = 3V and solve the following for RE.012 mA ß 150 7) Solve for the biasing resistors.Chapter 10: CASCADED AND CASCODE AMPLIFIERS 2) Calculate Imax.0018 = 0. IMAX = Max DC Power = 25 mW = 2.8 mA gm2 = 1.
AC DEC 10 1 100MEG . . .3K RS Vin VX 1 RLL Vout 0 50K *Capacitors C1 VX VB1 100UF C2 VB2 0 100UF C5 VE1 0 100UF C6 Vout VC2 100UF *Sources Vcc VCC 0 12 Vss 0 Vin AC 10mV *Transistors Q1 V01 VB1 VE1 TRANSMODEL Q2 VC2 VB2 V01 TRANSMODEL .If little or no impact..3 KΩ R3= 30 KΩ EE113 Demo .3 kΩ 2 ∴ R 3 = 2R 1 = 2R 2 = 8) Now check that any RS that may be present should not have a major impact on the gain..end G. Kovacs ©1997 Page 163 . move on.9N CJE=6P CJC=5P) *input frequency sweep . 9) SIMULATE THE CIRCUIT ON SPICE USING THE FOLLOWING VALUES: R1 = 15 KΩ RC = 750 Ω R2 = 15 KΩ RE = 1.. tweak IC and iterate if necessary.Cascode Amplifier *Resistors R1 VB1 0 15K R2 VB2 VB1 15K R3 VCC VB2 30K RC VCC VC2 750 RE VE1 0 1.MODEL TRANSMODEL NPN (BF=150 IS=1.Chapter 10: CASCADED AND CASCODE AMPLIFIERS VB2 = VCC 2 ∴ R 2 = R1 R BNET = 33.PROBE .If much impact.3E14 + TF=.
768 722.Chapter 10: CASCADED AND CASCODE AMPLIFIERS 50 0 1.444K. Let RE = (0.443K.597 10) Adjust RE to modify the CollectorEmitter current by multiplying by the ratio of simulated to desired gain. 10. 9.0h 100h V(Vout) / V(Vin) 10Kh Frequency C1 = C2 = dif= 1. 100Mh 46.000K.9257)(1.0000.2K 50 0 1.0h 100h V(Vout) / V(Vin) 10Kh Frequency C1 = C2 = dif= 1.0000. and should be acceptable! From SPICE Output Deck: TOTAL POWER DISSIPATION Everything looks like it's in order! G.506m 45.284 / 50 =) 0.284 687.0Mh 10. 1.40E02 WATTS . The gain was (46. 1.046 The gain is now 49.3K) = 1. Kovacs ©1997 Page 164 2.768.9257 of the desired gain of 50.999K.0Mh 10. 100Mh 49.391m 49.
Nonideal properties of the differential pair. DC operation of the differential pair.6. The benefits of the differential pair (no coupling caps > DC response. Commonmode operation of the differential pair.Chapter 11: DIFFERENTIAL AMPLIFIERS Chapter 11: DIFFERENTIAL AMPLIFIERS Ow! That’s hot! EE122 student’s first encounter with a soldering iron.. etc.10 and 7.3. 7.1 . Kovacs ©1997 Page 165 .). The frequency response of the differential pair..11 G. READ S&S Sections 6. The basic smallsignal equivalent circuit for the differential pair. 1: OBJECTIVES • To consider: The basic idea of the differential pair.
and one can apply inputs to both (differentially) or ground one and use the amplifier in a singleended (ground referenced) way. it is often referred to as a differential pair) is the basis for any operational amplifier (you can find one on the input of nearly any opamp type chip). Kovacs ©1997 Page 166 . One can take outputs from both (differentially) or use just one relative to ground (as you would with a typical opamp). • This circuit can compute the difference between two input signals. • The differential pair is basically two common emitter amplifiers sharing a common current source that sets the total DC bias current through both transistors at 2IE. In G.Chapter 11: DIFFERENTIAL AMPLIFIERS 2: DIFFERENTIAL AMPLIFIER BASIC CONCEPTS +Vcc +Vcc R C1 R C2 v o1 v o2 Q IE 2 v B1 v1 Q1 IE v B2 v 2 2IE V EE • The differential amplifier circuit (as shown above. the differential pair shown above has two (vo1 and vo2). • While the opamp symbol ("triangle") has only one output. + • There are two inputs. The two transistors are ideally identical and are generally assumed to be below. v1 and v 2 .
if the collector currents are the same in both transistors (IE ≈ IC). the current flowing out the emitter of each transistor is IE and therefore. • As will be seen when the circuit is studied more closely. • With NO differential signal input (v 1 and v2 are the same).e. v o1 = vo2 in this situation (if RC1 = R C2. we will begin by looking at the shifting of this DC current between the two transistors when there is a largesignal imbalance between their two inputs (i. Kovacs ©1997 Page 167 . the difference between v1 and v2 is large).Chapter 11: DIFFERENTIAL AMPLIFIERS seeking to understand its operation. • For the smallsignal case. 3: MODES OF OPERATION OF THE DIFFERENTIAL PAIR SingleEnded Mode Common Mode Differential Mode + + +    G. their g m's and other parameters will be equal (both transistors are as close to identical as possible). which is usually the case). its symmetry is what gives rise to its "special" properties.
7 V V EE Note: singleended operation shown. Kovacs ©1997 Page 168 . G.1 LARGE SIGNAL OPERATION +Vcc +Vcc ONE INPUT ZERO.3 V V EE +Vcc 0 R v C1 C1 +Vcc αI R C2 REVERSE THE POLARITY OF THE INPUT SIGNAL.αI R C 1V v B1 2 OFF I ON I .. THE OTHER SOME "LARGE" DC VOLTAGE. αI R C1 0 R C2 VCC .Chapter 11: DIFFERENTIAL AMPLIFIERS 3.0...αI RC 1V v C1 v B1 Q1 Q 0 2 vC2 VCC v B2 ON I OFF I 0.. vC2 Q1 0 Q v B2 VCC VCC .
α∆IRC 2 R C1 R C2 Q1 vo = 2α∆IRC + Q 2 VCC .. the current distributes itself between the transistors so that one gets an incremental amount ∆I more and the other one gets ∆I less.∆I 2 I approximately .αI RC .7V G... Kovacs ©1997 Page 169 ... vC1 INCREASES vC2 does the opposite. • For a small input signal (this is like our previous example with one input grounded). vC1 DECREASES For a NEGATIVE small input voltage.2 SMALLSIGNAL OPERATION +Vcc +Vcc SIGNAL IS A FEW MILLIVOLTS.αI RC + α∆IRC 2 v B2 small voltage v B1 I + ∆I 2 I . VCC . • NOTE that the differential output voltage is.. • NOTE that the current source node is at approximately 0... For a POSITIVE small input voltage. vo = 2α∆IRC • NOTE the polarity of vo..0.7 V V EE Note: singleended operation shown..Chapter 11: DIFFERENTIAL AMPLIFIERS 3.
• This means that common mode signals are rejected! (as long as RC1 = R C2 = R C) • Remember from the opamp section that commonmode rejection is supposed to be high. Kovacs ©1997 Page 170 . • Since both transistors are the same (matched).7 V V EE • Both inputs are driven by the same signal in commonmode operation.αI RC 2 v C1 v B1 2 VCC .. G. Therefore the difference between vC1 and vC2 will be ZERO.0..Chapter 11: DIFFERENTIAL AMPLIFIERS 3.αI RC 2 v CM I 2 I 2 I vcm . "COMMON" MODE R C2 vC2 Q1 Q v B2 R αI 2 C1 αI 2 VCC . the current will remain symmetrical between the two transistors and so the output voltages vC1 and vC2 will exactly track each other.3 COMMONMODE OPERATION +Vcc +Vcc SAME SIGNAL ON BOTH INPUTS.
vd 2 V EE • In this section.αI RC . Kovacs ©1997 Page 171 .gmvd 2 2 R C2 Q1 gmRC vd + Q 2 VCC . i E1 =e i E2 • Remember that I C = I Se vBE vT ( vB1 −vB2 ) vT G. • By defining the voltage at the common emitter node as vE and taking note of the polarities shown above.αI RC + gmRC vd 2 2 + vd vBE1 = VBE + vd 2 + + vE vd<< 2 VT gm ≡ IC = α I VT 2 VT I vBE2 = VBE .. i E1 I = Se α ( vB1 − vE ) vT i E2 I = Se α (v B2 −v E ) vT • Which can be combined into.gmRC vd 2 2 R C1 αI . the goal is to look at small signal operation with differential inputs. one can write expressions for the emitter currents in terms of a common voltage. which are seen to be split evenly between the two transitors BE junctions.Chapter 11: DIFFERENTIAL AMPLIFIERS 4: DETAILS OF SMALLSIGNAL OPERATION +Vcc +Vcc αI + gmvd 2 2 VCC ..
Chapter 11: DIFFERENTIAL AMPLIFIERS • These equations can be rearranged into the forms.v B2 e VT • Which represent the currents as functions of the input voltage and shows that we have an extremely sensitive variation of the iE’s with the differentialmode input voltages. Kovacs ©1997 . iE1 = iE1 + i E2 1 1+e vB2 . iE1 + i E2 = I iE1 = 1+ I vB2 . LINEAR REGION IC1 I NORMALIZED COLLECTOR CURRENT IC2 I WITH NO DIFFERENTIAL INPUT SIGNAL CURRENT SPLITS 50:50 NORMALIZED DIFFERENTIAL INPUT VOLTAGE vB1 .vB2 VT Page 172 G.v B2 VT • Combined with the constraint imposed by the current source that they both sum to I.v B1 VT and iE2 = iE1 + i E2 1 1+e vB1 .v B1 e VT and iE2 = 1+ I vB1 .
gm vd 2 C1 VCC . iC1 = I C + g m vd 2 iC2 = I C . IC = αI 2 • We see that the collector currents can be expressed in a form that looks like a common emitter amplifier gain equation with vbe given by vd /2.Chapter 11: DIFFERENTIAL AMPLIFIERS +Vcc +Vcc iC1 = IC + gm vd 2 R iC2 = IC .I C RC + g mRC vd 2 quiescent signal G.g m vd 2 • And the output voltages (considering both DC and AC components) are.gmRC vd 2 R C2 Q1 gmRC vd + Q 2 VCC .g m RC vd 2 quiescent signal vC2 = VCC .ICRC .ICRC + gmRC vd 2 + vd vBE1 = VBE + vd 2 + + vE vd<< 2 VT gm ≡ IC = α I VT 2 VT I vBE2 = VBE .vd 2 V EE • Redefining the steadystate current in each collector as.I CRC . Kovacs ©1997 Page 173 . vC1 = VCC .
Ad = vC1 = .v C2 = . G... simply reverse your connections to the outputs!) • And the singleended gain (looking at only one of the two outputs is 1/2 of that. a CE and a CB (a CASCODE) and ask: WHAT IS THE GAIN? +vcc +V cc +V cc R C1 R C2 vo1 vo2 v Q2 2 R C1 v1 Q1 vo1 vs + 2I E vs 2  Q1 V EE Q2 + vo2 R C2 v 2 2IE vs 2  vee + vcc • Start by grounding one of the inputs (to make it simpler to understand).Chapter 11: DIFFERENTIAL AMPLIFIERS • The differential gain (taken across the two outputs) is then. Ad = vC1 .. treat the circuit as a cascade of two singletransistor amplifiers.1 gm RC vd 2 • In order to understand the smallsignal behavior of the differential pair.gmRC vd (written as for a CE amplifier) (To reverse the sign.. Kovacs ©1997 Page 174 ..
below we will assume they are the same.. it is a common emitter amplifier that is degenerated by the load on its emitter (Q2)... it looks like two different configurations.. R in vS i in B1 R C1 c1 + v1 v m1 o1 rπ1 g E1 i2 v1 Ri = r π2 = r ≈ 1 e2 gm2 β+1 E2 C2 R c2 v o2 NOTE i2 = 1 + β iin Common Base Amplifier v2 r g m2 v2 π2 + B2 + (Here. G. it is a common collector (emitter follower) amplifier with a common base amplifier as its load.. gm’s. depending upon whether you are looking at vo1 or vo2 (think about this for a minute!). From the point of view of vo2.. etc. (Remember: the 2IE current source is only there for the DC circuit!) • MAKE A SMALLSIGNAL EQUIVALENT CIRCUIT.. From the point of view of v o1. let’s use separate notation of the rπ’s.Chapter 11: DIFFERENTIAL AMPLIFIERS Note that...) • First look at the input resistance of the amplifier. Kovacs ©1997 Page 175 ..
. G.r π vs = . v2.. The load of This term is the current gain of Q1 from B to E v2 = . • Now try to figure out the gain. (Assuming rπ’s are the same.. the input resistance is twice what it would be for a common emitter amplifier with the same transistor. • So..v s Rc = gmRc vs 2 2 Interesting Result! • From this we can find the gain..vs Rin β+1 2r π 2 Minus sign because of the way we defined v 2 (look at the schematic). This is also the same as r e for Q2. looking into the base of Q1... Start by computing the voltage at the input of the common base stage.. This term is the input current to Q1 This term is just r e • Now we can calculate the output voltage vo2 vo2 = gm v 2 Rc = gm .) Rin = r π + β + 1 r π = 2r π β+1 Scaling because you are looking into the base of Q1 Divide by (ß + 1) because you are looking into the emitter of Q2. Kovacs ©1997 Page 176 .Chapter 11: DIFFERENTIAL AMPLIFIERS • By inspection.vs β + 1 r π = .i 2r e = ..
• Also note that the other output (vo1) is an inverting version of (vo2) : v1 = vs 2 vo1= g m vs Rc 2 Hopefully you can see why this is the case! Av1 = vo1 = . • This circuit can be used as a symmetrical “phasesplitter” to get two equal amplitude versions of a signal 180° out of phase with each other G. depending on which output you use) > this is 1/2 the gain of a comparable commonemitter amplifier.. • So..Chapter 11: DIFFERENTIAL AMPLIFIERS Av2 = vo2 = gmRc vs 2 • This is 1/2 the gain of a comparable common base amplifier. Kovacs ©1997 Page 177 . at this point we know that: • the input resistance is 2r π • the gain is gm Rc 2 (or 1 times that..gm Rc vs 2 • Note that you can obtain the same result using rπ' and g m' for an unbypassed emitter resistance of RE = re = 1/g m.
. we drive the circuit differentially (when one output swings up.. The halfcircuit model does not work for asymmetrically driven differential pairs (i. the other swings down).. • In this example. one input grounded. so it is a virtual ground for small signals.. Kovacs ©1997 Page 178 .e. G.Chapter 11: DIFFERENTIAL AMPLIFIERS 5: HALFCIRCUIT MODEL OF DIFFERENTIAL PAIR • The halfcircuit model is a way of taking advantage of the symmetry of the differential pair circuit to simplify analysis in settings where we either drive the inputs differentially (and symetrically) or we drive them in commonmode. this is not the same as the previous case where we studied the frequency response because then we grounded one input! +Vcc +Vcc R C1 R C2 vC1 Q1 gmRC vd + Q 2 vC2 + vd + vd 2  I vE + vd 2  V EE THIS VOLTAGE DOES NOT CHANGE! • The idea there is that the voltages pivot about the common point between the emitters (the current source) > vE does not change > can thus treat vE as a power supply. the other driven).
if there were a resistor of value R connected between the collectors.Chapter 11: DIFFERENTIAL AMPLIFIERS • Just split the differential pair in half (it looks like two identical commonemitter circuits). For example.gmRC vd 2 Q1 Q 2 vC2 + vd 2  + gmRC vd 2 + vd 2  • The halfcircuit idea works for any connections between the two sides . each connected with one end to one of the collectors and the other end to ground. +Vcc +Vcc R C1 R C2 vC1 .. G. Kovacs ©1997 Page 179 . you could split it into two resistors of value R/2..
. assume a current source with a shunt resistance Ro...Chapter 11: DIFFERENTIAL AMPLIFIERS 6: HALFCIRCUIT MODEL AND COMMONMODE OPERATION +Vcc +Vcc +Vcc +Vcc R C R v C1 vC2 R C C R v C1 vC2 C + vcm  Q1 Q2 + vcm  + vcm I 2 Q1 Q2 2R o 2R o + vcm I 2 I Ro V EE VEE V EE (Note that Ro can be shown connected to ground. depending upon the current source..) • Each of the independent current sources provides a current of I and has an equivalent shunt resistance of 2Ro. Ro.) • Use the halfcircuit concept from above to analyze commonmode operation (which preserves the symmetry that allows the halfcircuit model to be used in the first place). Circuit theory says we can connect together two nodes that have the identical voltages without affecting the operation of the circuit > connecting the emitters in the halfcircuit leads to the physical equivalent of the circuit on the left. of the “realistic” current sources we looked at before. • This time.. • Relationship of halfcircuit to commonmode circuit: the two halves are identical so the potential at the emitters is the same.. Kovacs ©1997 Page 180 .. (REMEMBER the output resistance.. 2 • Both inputs are driven by the same signal (otherwise it wouldn’t be common mode!). • IDEALLY common mode signals would be rejected! What happens here? G.
also know that. vc1 = .. +Vcc R C v C1 + vcm  Q1 2R o This is the halfcircuit for AC signals....RC vcm 2Ro • This gives a (singleended output) gain of.RC 2Ro G. by symmetry. just write down the equation from memory (yeah. • This is just an emitter degenerated common emitter amplifier! So. vc2 = .RC vcm 2Ro and. we can make the approximation.. sure.) • Note that the current source is not shown because this is an AC circuit. the voltage and the current at the VE node changes with the input signal.Chapter 11: DIFFERENTIAL AMPLIFIERS • With the nonideal current source. vc1 = .g'mRCvbe =  gm RC vcm 1 + g m2Ro • If gm2Ro >> 1. VEE (This method is easier to understand than Sedra & Smith’s. Kovacs ©1997 Page 181 . but you would still redraw it to obtain a smallsignal circuit!.). Acm ≈ .
differential gain (from before) is. CMRR ≡ 20 log Differential Gain = 20 log Ad dB CommonMode Gain Acm (CMRR is almost always expressed in dB. . the CMRR can be written as. of course) • This is attenuation. • The singleended output.) • KEY POINT! If the output voltage is taken differentially.. Kovacs ©1997 Page 182 . which is exactly what you want for common mode signals (which are often noise that you do not want to amplify)! 7: COMMONMODE REJECTION RATIO • This parameter is used to express how well a differential amplifier can reject commonmode signals compared to how well it can amplify differentialmode signals.1 gmRC CMRR = 20 log 2 = 20 log g mRo dB RC 2Ro • It should be clear from the above equation that improving CMRR can be done via increasing gm and/or Ro..Chapter 11: DIFFERENTIAL AMPLIFIERS • For R C of 5 kΩ and Ro of 250 kΩ.1 gmRC 2 • Therefore.. G. this gives. Acm ≈ ..1 2Ro 100 (40dB. one can compute the CMRR.RC = . Ad = . the commonmode gain will be zero since both outputs do exactly the same thing (unless there are mismatches in transistors and/or resistors)! • If the output is taken in a singleended fashion. taking the absolute value.
add Rs . Break this into two problems: Part I going from the input signal vs to output Vo1 Rs + vs v 1 CBC vo1 gm v Rc  rπ cπ 1 vx 1 gm NOTE! cπ this is r e G. neglect ro and rX and assume that both transistors and their bias conditions are the same.Chapter 11: DIFFERENTIAL AMPLIFIERS 8: FREQUENCY RESPONSE OF THE DIFFERENTIAL AMPLIFIER Rs + v vs 1 CBC vo1 Cπ rπ gm v Rc 1  gm v2 + v  v 2 vo2 X + rπ Cπ Rc CBC • For this analysis. Kovacs ©1997 Page 183 .
VCC RB1 VB vs RC Rin i in + vs vin R B2 RE  RB + vo gmvbe RC RE rπ  vbe G.Chapter 11: DIFFERENTIAL AMPLIFIERS Part II going from the intermediate signal vx to output Vo2 . • Remember about incorporating an emitter resistance into a “new” small signal model for the transistor? This was the principle of local feedback that led to decreased gain for emitter degenerated common emitter amplifiers.. Kovacs ©1997 Page 184 ..gm v x CBC RC vo2 • This makes the analysis much easier to follow! BEFORE LOOKING AT PART I.
... iRE = yπ + g m v be G. Z i + v tot i + v be  rπ Cπ + ve RE gm v be Note that vs = v tot  i RE Zπ = rπ Cπ • The approach is to consider the current and voltage at RE and the total voltage and current into the base circuit (as shown above) and then to try to write a new expression that gives the correct total voltage and current for a “new” parallel RC combination at the base circuit. you have effectively replaced the original hybridπ model with a new one with an INCREASED rπ (now called rπ’) and a DECREASED gm (now called g m’). • In other words.. Kovacs ©1997 Page 185 .... • Consider the same circuit we used in the common emitter lecture but with Cπ taken into consideration. ' r π ≡ r π 1 + gmRE g'm ≡ gm 1 + gm RE • Now we will see that the same approach works if you have an IMPEDANCE in the emitter circuit instead of a resistance! • We will see that we can also suitably “modify” Cπ ! This is also the result of local feedback effects. • The impedance in the base circuit is.Chapter 11: DIFFERENTIAL AMPLIFIERS • Thus.. rπ zπ = r π  1 = = 1 sCπ 1 + sCπ r π yπ • And the current flowing in the emitter resistance RE is. we want to modify the base circuit to take into account RE so we can "ground" the emitter and treat it like a commonemitter amplifier after that.
PAY DIRT!! It looks like a parallel RC combination in series with RE! zin = RE + r π 1 + gm RE  1 Cπ S 1 + g m RE G.. i = vbe zπ • Now (finally!) we can write an expression for the TOTAL input impedance considering both the input impedance of the base circuit plus the effects of the emitter circuit. zin = vtot = zπ vbe + R E yπ + g m v be = z π + RE + z π REgm vbe i = RE + z π 1 + g m R E • Substitute for z π. = RE + r π 1 + g m RE 1 + s Cπ r π • At last. ve = RE yπ + g m vbe • Now write an expression for the total input voltage.....Chapter 11: DIFFERENTIAL AMPLIFIERS • And the voltage across R E is.. vtot = v be + R E yπ + g m v be ve • And the input current into the “new” basecircuit impedance zπ is. Kovacs ©1997 Page 186 ...
Chapter 11: DIFFERENTIAL AMPLIFIERS
• Our “new” equivalent circuit (neglecting R E)....
+
vtot
r π 1 + gmRE

Cπ 1 + gm RE
gm v 1 + gm RE i
• NOTE: here we have neglected R E because it is 1/gm2 when the load is a CB stage, as it is here.... if you put it into the above schematic, it makes things a bit less clear (but more accurate!). You can look at the "full" derivation in Gray and Meyer, "Analysis and Design of Analog Integrated Circuits," on pages 491  498.
• Note that for a differential pair, taking RE = re into account yields r π' = 2rπ, Cπ' = Cπ/2 and gm' = gm/2.
NOW FOR PART I...
• It looks just like the above, but with an emitter IMPEDANCE rather than just an emitter resistance.... Rs + vs v
1
CBC vo1 gm v Rc

rπ
cπ
1
vx 1 gm
NOTE!
cπ
ZE
v • In order to get transfer functions (i.e. Both vo1 and v x ) look at the results we just vs s obtained.
G. Kovacs ©1997
Page 187
Chapter 11: DIFFERENTIAL AMPLIFIERS
+ rπ vi 
+ vbe + vx cπ gm v be
Here we see that the circuit reduces to the familiar common emitter circuit with the emitter grounded... this makes the analysis easier and allows us to use the Miller approximation (we'll do that later).
Z
ZE
Cπ 1 + gmZ E
+ vi rπ' Cπ'
gm vi = g'mvi 1 + gmZ E
r π 1 + gmZ E
• For PART I, we know that,
1 re ZE = gm 1 + s cπ gm
and thus our “scaling factor”....
1 + s cπ 2 gm 1 + gm ZE = 2 1 + s cπ gm
(Note that Cπ/gm = C πre... perhaps more intuitive as an RC time constant.)
• This means there is a zero at
wz1 = 2cg m and a pole at wp1 = 1 wz1 = gm cπ π 2
G. Kovacs ©1997
Page 188
Chapter 11: DIFFERENTIAL AMPLIFIERS
• Practically... assuming the pole and zero cancel since they are so close in frequency.... the “scaling factor” reduces to,
1 + gm ZE ≈ 2
• Thus, we get an answer that looks a bit familiar (remember all those factors of two?).
r'π ≈ 2rπ
c'π ≈ cπ 2
g'm ≈ gm 2
G. Kovacs ©1997
Page 189
Chapter 11: DIFFERENTIAL AMPLIFIERS
NOW, USING THESE NEW VALUES FOR THE “EQUIVALENT” TRANSISTOR, SOLVE PART 1....
Rs + vs v
1
CBC vo1 gm v Rc

rπ
cπ
1
vx 1 gm
NOTE!
cπ
1
C BC rπ' Cπ'
SUBSTITUTE EQUIVALENT CIRCUIT
Rs + vs v' 1
vo1
' gm v ' 1
Rc
Rs vs + v' 1 ' gm v '
2
rπ'
1
USE MILLER APPROXIMATION vo1
Cπ'
Rc
C BC
CBC 1 + g'm RC
NOTE THAT:
A'v =  g'm RC so the Miller capacitance is CBC 1 + g'm RC
Page 190
G. Kovacs ©1997
Chapter 11: DIFFERENTIAL AMPLIFIERS
• The input time constant is given by,
τin = RS  r'π c'π + CBC 1 + g'm RC
= RS  2r π cπ + CBC 1 + gm RC 2 2
• NOTE that if Rs = 0, rx is the main thing that sets the upper cutoff frequency.... The above equation (approximate!) considering rx and ro is:
τin = RS + r x  2r π cπ + CBC 1 + gm RC  r o 2 2
• The output circuit time constant is,
τout = CBCRC
• And we can now write the overall transfer function equation for the vo1 path.....
vs
' vo1 = T S = rπ 1 1 g'm RC vs ' 1 + S τin 1 + S τout RS + r π
DC • Substituting in for the “primes”....
AC
gm R vo1 = T S = 2r π 1 1 C vs 1 + S τin 1 + S τout RS + 2r π 2
This is a lowpass response only (no highpass)... this is a DCcoupled amplifier!
G. Kovacs ©1997
Page 191
..... Rs + vs + v 1 CBC vo1 gm v Rc rπ v cπ 1 vx i NOTE! 1 gm  ..Chapter 11: DIFFERENTIAL AMPLIFIERS NOW DO PART II. 1 1 + S τin G. Kovacs ©1997 Page 192 .gm v x CBC RC vo2 • Assuming (see above) that. vx = vi (where vi is the voltage input to the base of Q1.) 2 • We look at the schematic from PART I again. AFTER Rs..gm v x cπ CBC vo2 RC • Knowing (from above) that the rolloff of the voltage vi is given by. .
• We can also see that the output signal vo2 is.. G..Chapter 11: DIFFERENTIAL AMPLIFIERS • v i can be found as a function of vS by simple voltage division. knowing that the input resistance (after RS) is 2rπ vi = 2 rπ 1 vs RS + 2 r π 1 + S τin input voltage divider term rolloff due to input circuit • Again... vo2 = g mvx RC RC = g mvi 1 + S R CCBC 2 1 + S τout • Now we can find the complete transfer function for vo2.vo1 C vs vs 1 + S τin 1 + S τout RS + 2r π 2 • THIS IS THE SAME AS FOR vo1 .. but it is 180° out of phase. Kovacs ©1997 Page 193 . gm R vo2 = T S = 2r π 1 1 = . we use the fact that vX is simply 1/2 of vi..
Chapter 11: DIFFERENTIAL AMPLIFIERS 9: SUMMARY OF DIFFERENTIAL PAIR SMALLSIGNAL OPERATION • Rin = 2r π • Both signal paths have a midband gain of Rin RS + R in gm R with an input voltage divider of C 2 • One path is inverting and the other one is noninverting. Kovacs ©1997 Page 194 . • Both paths have the same time constants! • Small signal gain: +vcc NOTE! R vs vo1 S vo1 vs = 2 rπ Rs + 2 r π gm R C 2 Q1 Rin 2IE Q2 vo2 vo2 vs vee + vcc = 2 rπ Rs + 2 r π gm R C 2 G.
. the output (taken differentially) should be zero.. • The input offset voltage is defined as the DC voltage that appears at the output with the inputs grounded.. Kovacs ©1997 +A  v V now = 0 os (in theory) Page 195 .1 INPUT OFFSET VOLTAGE • If you ground both inputs of the differential pair..).. THESE ARE SOME KEY TERMS TO KNOW! 10. it may not be. in practice. VOS ≡ Output voltage with inputs grounded = VO Differential Gain Ad • If you apply a voltage VOS to the inputs of the amplifier.Chapter 11: DIFFERENTIAL AMPLIFIERS 10: OTHER IMPERFECTIONS OF “REALISTIC” DIFFERENTIAL AMPLIFIERS • This section deals with imperfections of differential amplifiers beyond just the commonmode issues considered thus far. right? Well. divided by the gain of the amplifier (to refer the offset to the input of the amplifier. the output goes to zero (in theory) !!!! +A  v V ≠0 o V os A v G.
Vo + Q1 Q2 0V + .. with an imperfect amp.αI RC + ∆RC 2 2 VC2 = V CC .∆RC 2 • If the transistors are matched. mismatches in the transistors. (assume matched transistors) • If the mismatch is evenly distributed between the two resistors (this is not as weird as it sounds....∆RC 2 2 G...Chapter 11: DIFFERENTIAL AMPLIFIERS • In other words.. allowing us to write. VC1 = V CC .. • Consider offsets in the RC’s. V o is what you would get if you put Vos into a perfect amp with no offset! +Vcc +Vcc +Vcc +Vcc R C R R C C R C . etc.) RC1 = R C + ∆RC 2 RC2 = RC .VOS  Q1 Q2 I I V EE V EE • Where could this be a problem? If you cascade several DCcoupled amplifiers.αI RC . Kovacs ©1997 Page 196 . the current will divide equally between them.. you get Vo out if you ground the inputs. since you simply “split the difference” between the resistors and call the average value RC. the offset voltages get multiplied by the gains and eventually saturate the amplifier! • Where does the offset come from? • Mismatches in RC’s.
Kovacs ©1997 Page 197 . kT and the matching of q Increasing ∆R V os R C C T G. VO = V C1 . we need to recall that Ad is given by. the input offset voltage is a function of the matching of collector resistors.VC2 = αI ∆RC 2 • To calculate the input offset voltage. Ad = g mRC where αI IC = 2 gm = VT VT (note that this is not gm R because the output is taken differentially) 2 C • Therefore. not their absolute value.Chapter 11: DIFFERENTIAL AMPLIFIERS • Which gives an output voltage of. VOS αI ∆RC = VO = 2 = V T ∆RC Ad αI RC 2 RC VT Units of % change ≈ 26 millivolts • Thus the offset is clearly a function of temperature VT = the resistors.
• If you work out the same thing for this case. the emitter current in each of the two transistors I IB1 = I B2 = 2 β+1 • If the β’s are mismatched.e.1% resistors. VOS = VO = V T ∆I S Ad IS • Note that IS is also a function of T. the input bias currents will also be mismatched! • The INPUT OFFSET CURRENT is simply defined as the difference between them! IOS = I B1 .2 INPUT BIAS AND OFFSET CURRENTS • Input bias currents represent the currents required to turn on the input transistors (i. I S ∝ T α 10.I B2 • Using similar math as for the offset voltage. Kovacs ©1997 β2 = β .5 mV.∆β 2 Page 198 . and assuming β mismatch given by.Chapter 11: DIFFERENTIAL AMPLIFIERS • For +/. • The input bias current is given by... β1 = β + ∆β 2 G. • You can do the same type of analysis for differences in the transistor parameters (Sedra and Smith consider differences in the emitterbase junction areas that in turn cause differences in IS between the two devices). you end up with a similarlooking equation that gives VOS in terms of IS mismatch. the BASE CURRENTS!!!). this gives a maximum mismatch of 2% and maximum VOS of 0.
Kovacs ©1997 Page 199 . but the basic principles do!).Chapter 11: DIFFERENTIAL AMPLIFIERS • It can be shown that (see Sedra and Smith). IOS = I B ∆β β I where I B is the "average" base current IB = IB1 + I B2 = 2 2 β+1 • Here can improve things by reducing I C (and hence IB).VEE G.Vi) Vo ViIbias . but may be brought about by more complicated effects (i. VCC Ro Vi+ Ibias Ios + Vos + Acm (Vi+ + Vi)/2 Ri + Adm (Vi+ .e. the same formulas may not always apply. • Remember that the same terms apply to operational amplifiers.
a single reference current must be scaled and "copied" (the term used is "mirrored") so that the bias points of several transistors on a chip are proportional to each other and the reference. • In order to build discrete amplifiers and integrated versions. 1: OBJECTIVES • To consider: The basics current source circuits. In many cases. • The purpose of this chapter is to discuss such circuits and their uses. Design and use of current sources in multistage amplifiers. Kovacs ©1997 Page 200 . G. The limitations of common current sources... current sources are necessary.Chapter 12: CURRENT SOURCES Chapter 12: CURRENT SOURCES Oh. you’re supposed to pick up the OTHER end!!! Berkeley EE student’s thirtysecond encounter with a soldering iron.
" + β i = αi β+1 base current = i β+1 vBE +  + Same for PNP. • A diode can be formed by connecting the base to the collector.Chapter 12: CURRENT SOURCES 2: THE DIODECONNECTED TRANSISTOR • This simple circuit is a buildingblock for current sources. which will have almost exactly the same collector current as the current flowing through the diode.V BE curve for the transistor.. Kovacs ©1997 Page 201 .. making a "replica. The resulting VBE can be applied to the baseemitter junction of a second transistor. • The transistor now behaves like a diode with an IV characteristic equal to the IC .  i  G.
Kovacs ©1997 Page 202 ... For smaller β.VEE • Looking at the circuit above. IREF = Io... it is obvious that. Io = β I E β+1 and therefore. we need (to get the desired I0 out... need to make IREF > Io by the correct amount.. for large β.)...Chapter 12: CURRENT SOURCES 3: THE CURRENT MIRROR • This circuit makes good use of the fact that the diode formed from a transistor has the same temperature coefficient as the transistor. + V CC IREF 2IE β+1 β I E β+1 IE IE β+1 IE β+1 Io = Vo β I E β+1 IE . • Use one transistor as a vBE reference > feed in a desired reference current and the “diode” will establish the exact vBE on the other transistor to get the same current to flow in it. G. IREF = β I + 2IE = β + 2 I E E β+1 β+1 β+1 • So.
. Kovacs ©1997 Page 203 .VEE (or ground.Chapter 12: CURRENT SOURCES 4: A SIMPLE CURRENT SOURCE + V CC IREF Io Vo Q1 Q2 vBE  + IREF = VCC .. As β → ∞ I o = I REF (losing no current in the bases) G...) V 's are the same.V BE = I C1 + I B1 + I B2 = I C1 + IC1 + IC2 = I C1 + 2IC1 R β β β IC1 = I C2 = I o • As for the above case.V BE R . BE I C2 is the same as I o IREF = VCC ..
We know that... plug into the equation for Io. + E • As before. • Looking at the circuit one can see (base currents are neglected for high β ) that.. Kovacs ©1997 Page 204 .. VBE1 I = VT ln REF from IS VBE2 = V T ln Io IS IC = I Se v BE VT • So..V EE (or ground.) • Add in an emitter resistance in the output transistor. increases the output impedance (makes the current source more ideal... IoRE = V T ln IREF Io G. Io = VBE1 − VBE2 VE = RE RE • The equation is not very useful in terms of VBE’s... RE stabilizes VBE2 against variations with Q2’s β.Chapter 12: CURRENT SOURCES 5: THE WIDLAR CURRENT SOURCE + V CC You could get the reference current from a voltage reference and a resistor.. for example... see below) and allows scaling of the current to any value less than IREF. IREF Io Q1 + VBE1 +V Q2 BE2 RE V ..
.VEE (or ground. so in integrated circuits we can just make a bigger emitter to get more current.VEE ...) . • IS is a function of the area of the emitter. IC = I S vBE e VT • We can deliberately adjust IS for any given transistor to control its IC.. Kovacs ©1997 Page 205 . and 2) Compute RE knowing the 6: CURRENT MIRRORS • This concept can easily be multiplied. + V CC I REF 2 I REF 3 I REF IREF Q1 vBE  + vBE  + vBE  + vBE  + ... G.VEE This transistor has twice the emitter area as the reference.....Chapter 12: CURRENT SOURCES • To design a Widlar current source: 1) Choose I REF Io needed.VEE This transistor has three times the emitter area as the reference. • Since the collector current as a function of vBE is given by..... You can use one diodeconnected transistor to “steer” the current through many others by connecting their bases together. ..
r o = VA IC where VA is the Early voltage. you also get a lower (worse) resistance between the collector and emitter (i.. (see pages 207 .e.Chapter 12: CURRENT SOURCES 7: NONIDEALITIES OF BJT CURRENT SOURCES • Real transistors are not ideal current sources! • Remember the Early Voltage and ro? • r o is a resistance that models the slight effect of collector voltage on collector current in the active region of operation (the curves are not exactly flat!).208 in S & S) I Slope = 1/ro I V V • Remember that in the above figure.. • r o is inversely proportional to the DC bias current.. the slope of the current/voltage curves is steeper). Kovacs ©1997 . so that as you increase IC to get more gm and more gain. • r o is typically a few thousand Ohms. the IC versus VCE curves represent equal base current steps between them.. rx C BC B + v be gmv Cπ be rπ  C ro E THIS THING HERE! Page 206 G.
Chapter 12: CURRENT SOURCES 7... • Try to find out what the output resistance of a BJT current source is (look at Widlar). • As usual....1 EFFECTS OF ro ON CURRENT SOURCE PERFORMANCE • An ideal current source has an infinite output resistance. G.. want to use Ohm’s Law to compute Ro.. ix + B C gmvπ E rπ vπ  ro vx +  RE Use a test voltage... but a “real” one with a BJT does not... Io Io ro • We will see later that this means that you can’t simply “neglect” the bias current sources in your smallsignal equivalent circuits if you want to get a detailed idea of performance. Kovacs ©1997 Page 207 ..
ix gmvπ ro C gm + 1 vπ R'E +  vx vπ R'E R'E = r π  RE vπ + • The currents are relatively easy to see...Chapter 12: CURRENT SOURCES ' • Use RE = r π  R E and redraw the circuit for clarity. for small RE’ (where R E = rπ  R E ) gives (more intuitive). Kovacs ©1997 Page 208 .vπ ...gm + 1 v π = . vx = ..gm + 1 vπ r o ' RE • Now you can go ahead and use Ohm’s Law. 1 + gm + 1 r o ' RE ' Ro ≡ vx = = R'E + 1 + g mRE r o ix 1 ' RE • Which... (careful with signs!) ix = gmvπ .. ' ' Ro ≈ 1 + g mRE r o G...1 v π R'E R'E • And the voltages can be summed around the loop to get vx..
.10 2: ANALYSIS OF AN EXAMPLE AMPLIFIER • Begin with a simplified version of the EE122 discrete component opamp. +Vcc +Vcc +Vcc RE R I C1 2 . OW.. Kovacs ©1997 Page 209 .. OW!!!! Berkeley EE student’s thirtythird encounter with a soldering iron.Chapter 13: MULTISTAGE AMPLIFIERS Chapter 13: MULTISTAGE AMPLIFIERS OW.. 1: OBJECTIVE • To consider: How to analyze and design multistage amplifiers based on the differential pair.IB R C2 IB Q I EQ3 3 R S Q1 I 2 Q 2 C C vo vs I Ic V EE V EE G.. REVIEW SEDRA & SMITH SECTION 6. OW.
• Looking at the output stage.. the common emitter amplifier formed by Q1 has an emitter resistance of re (of Q2) and drives the output stage formed by Q 3. Kovacs ©1997 Page 210 . A voutput ≈ g 'm (R L  R o ) • To evaluate the effect of the base current of Q3 on the differential pair (tending to unbalance it).. • We already looked at the response of the differential pair with one input grounded and arrived at a smallsignal equivalent circuit like the one shown below: G. and is shown below. • For this situation. and the load is also a function of R o of the current source (R o not shown above. but it is the output resistance of the current source driving Q3) and the load resistance (RL...VBE3 2 • One needs to compensate for the extra current flowing into Q1! Do this by adjusting RC’s. • Next consider the frequency response of this amplifier by considering it with one input grounded and using the model discussed in the Differential Amplifiers Chapter. there is emitter degeneration. not shown in the schematic).. look at the voltages.I B RC1 = I EQ3RE . that RC can be made larger. I ..Chapter 13: MULTISTAGE AMPLIFIERS • A small signal model for the differential pair with one input grounded has already been discussed. THINK ABOUT THIS! To get balanced output voltages for the differential pair if one side has less current flowing in it (due to the base current of Q3).
Rs vs + v' 1 ' gm1 v ' vo1 ' rπ1 1 ' Cπ1 RC C BC1 CBC1 1 + g'm1RC where ' rπ1 ≈ 2rπ1 C'π1 ≈ Cπ 2 g'm ≈ g m1 2 • The output stage that is driven by vo1 must then be considered at this stage's load. G. • The output is a common emitter amplifier with an unbypassed emitter resistance RE. • The input resistance of the output stage becomes the output load for the first stage. and a collector resistance that is the output resistance of the current source (not shown in the simplified schematic at the beginning of this chapter).Chapter 13: MULTISTAGE AMPLIFIERS Rs + v vs 1 CBC1 vo1 Cπ1 rπ1 gm1 v Rc 1  gm2 v2 v 2 vo2 + rπ2 Cπ2 Rc CBC2 • The signal path to vo1 was drawn as shown below. Kovacs ©1997 Page 211 . • This load includes the input resistance of the output stage in parallel with the RC of Q1 (otherwise the Miller approximation shown above will be incorrect since the gain at the transistor in question will change). but now we will have to consider the loading effects of Q3 on this circuit (not shown in the schematic immediately below).
. and if we did. In the case above. NOTE that this does not consider the current source output resistance (Ro) here. Secondstage Gain....Chapter 13: MULTISTAGE AMPLIFIERS Rs vs + v' 1 ' gm1 v ' + ' rπ1 1 v ' gm3 v ' o C BC1 RC v'  3 ' Cπ1 ' rπ3 3 ' Cπ3 RL CBC1 1 + g'm RC  r 'π3 CBC3 + CC 1 + g'm3RL CBC3 + CC Where. the gain is a direct function of RL. it would be in parallel with R L in the above equation... AM = r 'π1 g'm1 RC  r 'π3 g'm3 RL ' RS + r π1 Input voltage divider.. Firststage Gain. Kovacs ©1997 Page 212 . ' r π1 = 2 r π1 g'm1 = g m1 2 ' r π3 = r π3 1 + g m3 RE g'm3 = gm3 1 + g m3 RE C'π1 = Cπ1 2 ' Cπ3 = Cπ3 1 + g m3 RE • For the midband gain (ignore the capacitances).. G..
. but may be longer than the input circuit. RL CBC3 + C C • This time constant will be fairly short.Chapter 13: MULTISTAGE AMPLIFIERS Rs vs + v' 1 ' gm1 v ' + ' rπ1 1 v ' gm3 v ' o C BC1 RC v'  3 ' Cπ1 ' rπ3 3 ' Cπ3 RL CBC1 1 + g'm RC  r 'π3 CBC3 + CC 1 + g'm3RL CBC3 + CC • For the high frequency response. G.. as intended!!! 1) The input time constant is. • NOTE that the compensation capacitor CC gets Miller multiplied too. and external capacitances that may be in parallel with R L. RS  r 'π1 C'π1 + CM1 where CM1 = CBC1 1 + g'm1 RC  r 'π3 Since RS is typically very small. depending on RL . Kovacs ©1997 Page 213 . you need to consider the Miller multiplication of the capacitances in both stages. 3) And the output time constant is.. ' r π3  R C C'π3 + C M3 where CM3 = CBC3 + CC 1 + g 'm3RL • This is a relatively long time constant because CM3 is very large (both C BC3 and CC get Miller multiplied) and because the resistance here is moderate in magnitude. and thus corresponds to a very high frequency. 2) The interstage time constant is (for extra nerdiness we could include CBC1 but it is a relatively small contribution)..
Kovacs ©1997 Page 214 .... If PNP transistors were not available. • Note the cascaded differential amplifier for increased gain (inputs are at the left. some other sort of levelshifter circuit (see below) would be necessary since the DC voltage in the signal path would otherwise be too close to Vcc. both shown grounded). with the currents referenced to the resistor feeding the diodeconnected transistor at lower left.Chapter 13: MULTISTAGE AMPLIFIERS • Look at the example in Sedra and Smith. It is not needed since no voltage is needed from this collector and it is the current there that is required for proper operation of the differential amplifier (the output beyond this point is singleended. • Note the PNP transistor used as an emitter degenerated common emitter stage just before the emitter follower output. First.. • Note that one of the transistors in the second differential amplifier has no collector resistor. VCC VEE • Note the current mirrors to distribute current to the circuit. page 438. and the voltage from the second differential pair is taken at the other transistor. look at it in chunks that are easy to understand (the reader may wish to add some lines and notations here). G. which does have a collector resistor).
coupling capacitors between stages are almost always not used because they cannot be made large enough for reasonable lowfrequency operation.. • Thus. Kovacs ©1997 Page 215 . • Also. This means that voltage offsets like V BE drops between stages can start to add up. • A typical level shifter is simply a degenerated CE amplifier (it can provide gain or simply act as an inverting buffer if RE = R C). stages referred to as level shifters can be used to compensate where necessary.. emitter followers and diodes can be used to drop ≈ 0. the sign of the voltage drop per shifter can be changed.Chapter 13: MULTISTAGE AMPLIFIERS 3: LEVEL SHIFTERS • In multistage amplifiers on integrated circuits. The choice of RC "programs" the quiescent collector voltage of the level shifter and can thus be used to "center" the output voltage of an opamp so that it can swing both positive and negative from its quiescent point. stages are DCcoupled. • By using PNP and NPN transistors.7 V per stage. G.
1 . it has been found possible to effect extraordinary improvement in constancy of amplification and freedom from nonlinearity.7 G. Some realistic circuit examples and how to analyze them. 1934 1: OBJECTIVES • To consider: The basics of feedback. inventor of negative feedback.Chapter 14: FEEDBACK Chapter 14: FEEDBACK . say 40 decibels higher than necessary (10. Harold Black. by building an amplifier whose gain is made deliberately.8. The basic feedback topologies.. The properties of negative feedback... READ S&S Sections 8. Kovacs ©1997 Page 216 . An example of the “ideal” feedback case.000fold excess on an energy basis) and then feeding the output back to the input in such a way as to throw away the excess gain.
. • The effects of noise can be reduced (but not the noise itself). but with gains that were difficult to control from amplifier to amplifier. • The input and output impedances of the amplifier can be modified. G. if you use negative feedback. • All you have to do to “get some feedback” (of the negative kind) is to supply a scaled replica of the amplifier’s output to the inverting (negative) input (more on this below) and presto! • Of course.. overall gain of the amplifier is always less than the maximum achievable by the amplifier without feedback.1 PROPERTIES OF NEGATIVE FEEDBACK • The gain of the circuit is made less sensitive to the values of individual components. • One day. • Nonlinear distortion can be reduced.. we usually use NEGATIVE FEEDBACK.Chapter 14: FEEDBACK 2: INTRODUCTION TO FEEDBACK • There are two types of feedback: regenerative (positive feedback) and degenerative (negative feedback). Harold Black invented negative feedback. • The bandwidth of an amplifier can be extended.. Kovacs ©1997 Page 217 . 2. while riding the Staten Island Ferry... • Unless you want your circuit to oscillate.. • This idea came about in the late 1920’s when they were able to build amplifiers with reasonable gains.
Chapter 14: FEEDBACK 2... xo = Ax i • So we can obtain an expression for the output signal in terms of the input signal and the feedback gain.x f • The output of the amplifier is. let’s look at the basic feedback circuit illustrated above. xo = Ax s .2 THE BASIC FEEDBACK CIRCUIT • With an input signal xs . • The amplifier has a gain of A and the feedback network has a gain of β. xo = A xs . • The input to the amplifier is. a feedback signal xf. xi = x s .x f = A xs . an output signal xo.Aβxo → xo 1 + Aβ = Ax s G. Kovacs ©1997 Page 218 .. and an amplifier input signal xi.βxo • Rearranging..
xf = Aβ x s 1 + Aβ • If the amplifier gain and the loop gain are large (i..... • This explains why the two terminals of an opamp become nearly identical when using negative feedback. Kovacs ©1997 Page 219 . xs Afb ≡ xo = A xs 1 + Aβ • This means that the gain is almost entirely determined by the feedback circuit!!! Afb ≡ xo = A ≈ 1 xs 1 + Aβ β for large A • For positive feedback. you only need to change the “+” sign in the denominator to a “” sign.Chapter 14: FEEDBACK • From which we obtain the negative feedback equation by solving for the overall gain xo .. (Remember the “virtual ground” stuff?) G. xf. • It is easy to obtain the equation for the feedback signal. Aβ >> 1)..e. then the feedback signal xf becomes nearly an identical copy of the input signal xs ...
Kovacs ©1997 Page 220 . AM 1 + A Mβ Af s = s 1+ ω H 1 + A Mβ • Thus we have another singlepole response. ω Lf = ωL 1 + AMβ G.Chapter 14: FEEDBACK 3: HOW FEEDBACK AFFECTS BANDWIDTH • Assuming an amplifier with a singlepole frequency response (i. and the amplifier will have a lower cutoff frequency by a factor again equal to the amount of feedback. you get. its frequency response is given by. the gain becomes. Af s = A s 1 + βA s • Substituting. ω Hf = ω H 1 + A Mβ • THIS MEANS THAT THE UPPER CUTOFF FREQUENCY IS INCREASED BY A FACTOR EQUAL TO THE AMOUNT OF FEEDBACK. but with a high cutoff frequency given by. an “ideal” opamp).e. THIS IS THE FUNDAMENTAL GAINBANDWIDTH PRODUCT TRADEOFF THAT WE STUDIED WITH OPAMPS!!! • The same is true for the lower cutoff frequency. A s = AM 1+ s ωH • If you use the amplifier with negative feedback.
G. Kovacs ©1997 Page 221 . FUNDAMENTAL ASSUMPTIONS: 1) The feedback network is UNILATERAL. is INDEPENDENT of the load and source resistances seen by the composite circuit.Chapter 14: FEEDBACK THE BANDWIDTH IS EXTENDED BY THE SAME AMOUNT THE MIDBAND GAIN IS DECREASED. LOAD AND FEEDBACK NETWORK DO NOT AFFECT THE GAIN OF THE AMPLIFIER! • IN PRACTICE THIS WILL NOT BE THE CASE. This means that it transmits signals only one way (from the input to the output of the composite circuit). • A KEY ASSUMPTION HERE IS THAT THE SOURCE. This is usually only approximately true. This means that it transmits signals only one way (from output to input of the composite circuit). β. 2) The amplifier is UNILATERAL. 3) The loop gain. SO WE WILL LEARN HOW TO TAKE AMPLIFIER LOADING INTO CONSIDERATION.
1 REMINDER: TYPES OF AMPLIFIERS Rs vs Ri Ideal Voltage Amplifier + + ∞ vi  A v vi +  Ro 0 R L vo  Rs vs Ri Ideal Transconductance Amplifier + io Ro ∞ vi  G M vi ∞ RL Ideal Transresistance Amplifier is Rs Ri RMi i ii + +  0 Ro 0 R L vo  Ideal Current Amplifier is Rs Ri ii Ai i i io 0 Ro ∞ RL NOTE THAT THE IDEAL INPUT AND OUTPUT RESISTANCES ARE SHOWN. Kovacs ©1997 Page 222 . G.Chapter 14: FEEDBACK 4: FROM BASIC BLOCK DIAGRAM TO ACTUAL FEEDBACK CIRCUITS 4.
Chapter 14: FEEDBACK Type of Amplifier Voltage Gain Expression Av = vo = voltage gain vs (dimensionless) Ideal Input Impedance Zi = ∞ Ideal Output Impedance Zo= 0 Transconductance i Gm = vo = transconductance s in Ω1 or Siemens Zi = ∞ Zo = ∞ Transresistance Rm = vo = transresistance is in Ω Zi = 0 Zo = 0 Current Ai = io = current gain is (dimensionless) Zi = 0 Zo = ∞ • The distinction between types of amplifiers is essential because the type of feedback used with each type is distinct and requires analysis in the “native” units of current or voltage that is used for input and output. THIS STUFF IS EASY TO GET CONFUSED ON!!! PLEASE SPEND A LITTLE TIME TO TRY TO MEMORIZE THE DIFFERENT AMPLIFIER TYPES AND THE RELEVANT FEEDBACK TOPOLOGIES. Kovacs ©1997 Page 223 .. G..
.2 BASIC STRUCTURE OF THE CIRCUIT Mixer or Comparator Sampler • Here we have assumed that there was an input “comparator” or “mixer” and an output “sampler” that provided us with a copy of the output signal for use as a feedback signal.. FOOD FOR THOUGHT. Kovacs ©1997 Page 224 .Chapter 14: FEEDBACK 4... won’t βxo also decrease. making the output eventually go to zero???? • NO! This is because subtracting x f = βxo causes xi to decrease slightly so βxo decreases slightly so xi is allowed to INCREASE slightly! This acts to stabilize the gain! G. • If x o decreases due to negative feedback. • The form these devices take depends upon whether the amplifier’s input and output are current or voltage based.
Chapter 14: FEEDBACK 4.3 TYPES OF MIXER vs +  Rs + ii vi Amplifier Is Rs Amplifier vf+ if SERIES MIXER SHUNT MIXER if Feedback Network Feedback Network 4.4 TYPES OF SAMPLER io + Amplifier vo  + R L Amplifier vo R L  SHUNT (VOLTAGE) SAMPLER io SERIES (CURRENT) SAMPLER Feedback Network Feedback Network G. Kovacs ©1997 Page 225 .
.. respectively..SAMPLING) SERIESSHUNT SHUNTSERIES SERIESSERIES SHUNTSHUNT series (voltage) mixing. it goes down if you put something in shunt (same idea at the output). if you put something in series with the input. the degenerated commonemitter amplifier. series (voltage) mixing. or transconductance amplifier).. currentout.. series for current and shunt for voltage). voltagesampling shunt (current) mixing. sample the output current and produce a feedback voltage G. where MIXING and SAMPLING are either SHUNT or SERIES.. voltagesampling VV II VI IV • Let's consider a familiar example.g. Kovacs ©1997 Page 226 . • There are four possible types.SAMPLING” feedback. treat it as a voltagein..e. currentsampling... R B2 Feedback network. VCC RB1 VB vs + vi + vf  RC vo iC i E≈ i C RE Here we consider iC to be the output of the amplifier (i.... its impedance goes up. • We refer to a given feedback amplifier in terms of the “MIXING ...Chapter 14: FEEDBACK NOTE the effect of feedback on input and output impedance is just a function of the mixing and sampling type. (for sampling think of how you would measure I or V in the lab... e. INPUTOUTPUT (MIXING . currentsampling shunt (current) mixing..
g. assume that RB1 and RB2 are very large).gmGmRE vs vs vs Gm 1 + gmRE = gm Gm = gm 1 + gmRE THIS IS g'm THAT WE SAW BEFORE! G.GmRE = gm . • Considering the output current to be the output signal (e. vf = ioRE so the feedback network gain. β = RE (NOTE: don't get confused! This ß is NOT the transistor's ß!) Transistor's current gain = io ≡ g m in Ω1 vi • The output current is given by.vf = gm vs . we end up with an equation we have seen before! g v i R Gm = io = m s o E = gm 1 . io = gm vs . Kovacs ©1997 Page 227 . this is a seriesseries feedback topology.io RE = gm 1 .Chapter 14: FEEDBACK • Since we sample the output current and generate a voltage feedback signal.i oRE • Combining these equations to find the overall gain for the amplifier. the units of the basic amplifier are. io = i c) and the input to be vs (for simplicity. Overall Gain = A = io ≡ Gm in Ω1 vs • We know that the feedback voltage is given by Ohm's Law as. G m.
. The noninverting opamp configuration using an ideal opamp (infinite input impedance..Chapter 14: FEEDBACK Note that. this ß is not ß for the transistor!) 5: SERIESSHUNT FEEDBACK > VOLTAGE AMP (SERIES [VOLTAGE] MIXING.. Gm = gm = A 1 + gmRE 1 + Aβ (Again. G.. zero output impedance). Kovacs ©1997 Page 228 . VOLTAGESAMPLING) vs +  Rs + + vf +  vi Amplifier vo  R L SERIES MIXER SHUNT (VOLTAGE) SAMPLER Feedback Network • An “ideal” example.
• Note that there are often “implied” ground connections to make the feedback network a fourterminal device. but including the input and output resistance terms (Ri and Ro).. β= R1 R1 + R2 • This can be plugged into the feedback gain equation to find the overall gain.. Kovacs ©1997 Page 229 . • The feedback network gain can be obtained directly by voltage division.Chapter 14: FEEDBACK Rs Vs +  Vo R2 + + vf  R1 vo  RL • Notice how you can redraw the two feedback resistors as a feedback network of the form we are discussing. G.. Afb ≡ xo = xs A 1+A R1 R1 + R 2 ≈ 1 = 1 + R2 R1 R1 R1 + R 2 • Continuing with the seriesshunt case..
Vf Rif ≡ Vs = Vs = R i Vs = Ri Vi + βAVi = R i 1 + Aβ Ii Vi Vi Vi Ri • NOTE that later on.... Kovacs ©1997 Page 230 . Zif S = Z i S 1 + A S β S (always true for series mixing) • The effect of feedback on input resistance or impedance is only a function of the method of mixing • The output resistance can also be obtained by the same method we used previously: • Reduce the input signal (Vs ) to zero and apply a test voltage Vt at the output.Chapter 14: FEEDBACK vs +  s I i + Ro o + s'  vf +  vi Ri +  Av i o' vo  R L A Circuit SERIES MIXER B Circuit + + SHUNT (VOLTAGE) SAMPLER βv o vo  Equivalent Circuit s + I i +  Rof R if o + vs  s' A f vs o' vo  • We can obtain an expression for the equivalent input and output resistance...... we will generalize this to include IMPEDANCES. G.
this could be written with frequencydependent A and β. > I = Vt .. more generally. G.AV i = Vt + Aβ Vt Ro Ro • Therefore. Rof ≡ Vt I • Since Vs = 0.β Vt • Thus. Kovacs ©1997 Page 231 . R of = Ro 1 + Aβ or. Z of = Zo 1 + Aβ True for all shunt sampling cases.Vf = .Chapter 14: FEEDBACK s I i + Ro o I + s'  vi vf+  Ri +  Av i o' vt  +  A Circuit SERIES MIXER B Circuit + + 0 SHUNT (VOLTAGE) SAMPLER βv o vo  • Note that you "kill" the input voltage by shorting the input terminals. I = Vt . • Starting with the definition.AV i Ro Vi = . • Again.β Vo = ..
CURRENTSAMPLING) ii Is Rs io Amplifier if if io + vo R L  SHUNT MIXER SERIES (CURRENT) SAMPLER Feedback Network • Here we have series sampling (like connecting an ammeter in series with a circuit in which you want to know the current flow)... • Shunt mixing feeds back a βscaled version of the sampled current. • Now look at a practical circuit example. Kovacs ©1997 Page 232 .. G. a twostage commonemitter amplifier..Chapter 14: FEEDBACK 6: SHUNTSERIES FEEDBACK > CURRENT AMP (SHUNT [CURRENT] MIXING.
Q2 RL1 Q1 Io α RF RE RL2 Is Rs If 5) Io DECREASES. 2) Q1's base and collector currents increase. I G... 3) This voltage DROPS....... α 6) I f INCREASES (in the direction shown!!!!) [think about V E decreasing..Chapter 14: FEEDBACK Io Q2 I in RL2 RL1 Q1 Io α RF RE Is Rs If • NOTE that the feedback current.. Io 4) I o DECREASES...... 1) Let Is increase.... This is the notation we will use! • BE SURE TO “GO AROUND THE LOOP” to verify that the polarity of the feedback is negative (this means that the loop gain (1 + Aβ) must be positive. more current can flow here.) ‘ROUND THE LOOP... Kovacs ©1997 Page 233 . If is shown SUBTRACTING from the input current...] CONCLUSION: the feedback is negative since the increase in If SUBTRACTS from s ....
... 7: SERIESSERIES FEEDBACK > TRANSCONDUCTANCE AMPLIFIER VOLTAGEIN. CURRENTOUT (SERIES [VOLTAGE] MIXING. but rather the nearly equal emitter current of Q2.Chapter 14: FEEDBACK • This sort of “sanity check” tends to be essential... G. This is quite reasonable. such as when you need to drive the deflection coils in a television... and we will do this in the next example too. Kovacs ©1997 Page 234 .. CURRENTSAMPLING) Rs + io + vs +   vf+ vi  Amplifier io vo R L  SERIES MIXER SERIES (CURRENT) SAMPLER Feedback Network • This type of circuit is used when you need to generate an output current proportional to a command voltage... • Note that here we did not sample the OUTPUT current directly.
Kovacs ©1997 Page 235 . 2) The series mixer is through the Q1 EB junction rather than direct... 1) Emitter current ≈ output current...Chapter 14: FEEDBACK Two approximations. Io Q3 RL2 Rs +  RL3 Q2 Q1 RL1 Io α Vs + SERIES RF RE1 RE2 SERIES Vf  G..
Kovacs ©1997 Page 236 . VOLTAGESAMPLING) ii + Is Rs Amplifier if if vo  R L SHUNT (CURRENT) MIXER SHUNT (VOLTAGE) SAMPLER Feedback Network • If you redraw the inverting opamp configuration with a Norton input source..Chapter 14: FEEDBACK 8: SHUNTSHUNT FEEDBACK > TRANSRESISTANCE AMPLIFIER CURRENTIN... you can treat the circuit as a transresistance amplifier.. VOLTAGEOUT (SHUNT [CURRENT] MIXING. G..... allowing I F to increase. IF subtracts from IS. If I S increases. RF SHUNT IF RF RS VS +  IS = VS RS + SHUNT + Vo  RS Vo  • ‘Round the loop. I hope!). so an increased IF means negative feedback (as you expected. V o decreases.
• The source and load resistances will also affect these three parameters. the feedback network is generally passive. 3) Now use the corrected "A" gain (call it A') to compute A f .. 5) Also... 9. G.. you can compute the input and output impedances (with feedback) using the appropriate (1 + A'β) scaling (how to do this is explained below).. Why not just start without them? The reason is that you have to take into account their loading effects on the amplifier BEFORE doing the feedback calculations.Chapter 14: FEEDBACK 9: FEEDBACK ANALYSIS IN REALISTIC CIRCUITS • In real situations. if you need to. • These “lumped” circuits can be plugged into the ideal feedback circuit case and used to determine the properties of the closedloop case. and loads the basic amplifier in the loop and thus affects A... if you need to. Af = A 1 + Aβ until you compute a new "A" value that takes into account the loading!!! 2) After determining the "A" circuit. You simply mathematically remove the effects of RS and RL. • The trick is to lump these “bad” effects into the previously mentioned “A” circuits (and sometimes “β” circuits..1 SUMMARY OF STEPS YOU WILL USE 1) Redraw the amplifier as an "A" circuit which takes into account the fact that introducing the feedback network actually affects the openloop gain of the amplifier! Also lump RS and R L into the "A" circuit. there is a set of rules about how to incorporate the loading effects and how to compute β. find the value for β.. Ri and R o.. but we lump it all into "A" here).. you can compute the input and output impedances of the amplifier (with feedback) looking into its terminals but without RS and RL (see below).. The fact that this loading occurs means that you can't just use.. • For each of the four feedback topologies. Af = A' 1 + A' β 4) Now. Kovacs ©1997 Page 237 .
" incorporating all of the effects of loading and feedback into a final equivalent." G.Chapter 14: FEEDBACK 10: SERIESSHUNT FEEDBACK (SERIES [VOLTAGE] MIXING. Kovacs ©1997 Page 238 . VOLTAGESAMPLING) vs +  s I i + Ro o + s'  vf +  vi Ri +  Av i o' vo  R L A Circuit SERIES MIXER ß Circuit + + SHUNT (VOLTAGE) SAMPLER βv NOTE that this is an "ideal" case so no R S is included! o vo  Equivalent Circuit s + I i +  Rof R if o + vs  s' A f vs o' vo  • The above equivalent circuit is the "goal. • We will learn a general set of rules (one for each of the four types of feedback configurations) to "pull" the effects of R S. R L and ßnetwork loading into the a new "A" circuit and then apply the basic feedback equation with the ßnetwork taken to be "ideal.
Chapter 14: FEEDBACK • Determine the loading effects on input and output by “destroying” the feedback from each end of the feedback network to the other.. Kovacs ©1997 1 FEEDBACK NETWORK 2 +  V'o Page 239 . SHORT SHUNTS and SEVER SERIES RULES FOR SERIESSHUNT V'i +  RS R11 BASIC AMPLIFIER R22 + V'o RL This model for the amplifier now takes into account loading from R and R s L ' A' ≡ Vo V'i R11 1 FEEDBACK NETWORK 2 SHORT the SHUNT SEVER the SERIES 1 FEEDBACK NETWORK 2 R22 ' β ≡ Vf V'o I1 = 0 Here we just want "pure" feedback (no worrying about loading anymore).. I1 + V'f G.
• Q1’s collector is tied directly to VCC (no collector resistor) to eliminate its Miller capacitance multiplication. If you haven’t.... (You were supposed to read section 7.7 V + V cc RC = 20 KΩ Q3 RS = 10 KΩ Vs Q1 +  R2 = 9 KΩ Q2 RL = 2 KΩ Vo R1 = 1 KΩ 1 mA 5 mA R'if You already know that the output voltage of the differential pair is the difference between the inputs!!! As connected.7 ..11 before.V EE SERIES MIXING .. • Look at the circuit and convince yourself that it is really SERIESSHUNT feedback.5 + V cc = 10..VBE3 ≈ 0 V G. please do so. Kovacs ©1997 Page 240 ..7 .V EE SHUNT SAMPLING • First look at the overall design.Chapter 14: FEEDBACK LET’S TRY IT!!!! SEDRA & SMITH EXERCISE 8...) • DC Voltages.7 V > Vo = 0. Q2 is the inverting input! R'of .5 mA X 20 KΩ = + 0.0. • This is a differential amplifier feeding an emitter follower stage at the output. (β = 100 for all three transistors) IE1 = I E2 = 0..5 mA VC2 ≈ 10..
.. Re ve = v b Re Re + r e where Re is the total resistance in the emitter circuit (effectively the load of the CC amplifier!) and re is the emitter resistance of the transistor in the CC amplifier. and take it into account... Kovacs ©1997 Page 241 . Therefore.. the voltage gain (to the collector of Q2) can be quickly determined. R e = re2. and take it into account.. RC Vo vs +  Q1 Q2 Later on we can put a resistance here. • The circuit is basically a common collector stage driving a common base stage! • For the common collector stage..Chapter 14: FEEDBACK + V cc + V cc Later on we can put a resistance here. the emitter resistance of Q 2.. • In this case. the voltage feeding into the common base stage is just split in half (surprise. surprise!!!). I • For the modified (no collector resistor on Q1) differential pair (shown above in simplified form). we know that the output voltage (at the emitter) is given by. ve1 = vs r e2 = vs r e2 + r e1 2 since r e2 = r e1 G...
(you may more frequently see a smaller RS anyway... • We also know that the input resistance of the CC amplifier is large. so later on we will take into account the effects of RS..) NOW APPLY THE SERIESSHUNT RULES TO FIND THE “A” CIRCUIT: V'i +  RS R11 BASIC AMPLIFIER R22 + ' Vo RL • Again. 1) Input loading > short the shunt at the output and determine the resistance seen looking into the INPUT side of the feedback network. this is the amplifier with all loading effects "lumped" into it. RiCC = r π + β + 1 Re • But in this case it is only about 10 kΩ.. ie2 = ve1 = vs 1 = vs re 2 r e 2r e • Therefore.Chapter 14: FEEDBACK • The current into the emitter of Q 2 is.... A = vC2 = ic2 RC = RC vs vs 2r e • If we load the input stage with an output stage (Q3) we need to include that load resistance into a “new” RC. R11 1 FEEDBACK NETWORK 2 G. Kovacs ©1997 Page 242 .. ic2 = α vs ≈ vs 2r e 2r e • Finally giving a voltage gain of. the collector current of Q2 is.
Chapter 14: FEEDBACK R2 = 9 KΩ R11 R1 = 1 KΩ R11 = 1KΩ  9KΩ 2) Output loading > open circuit (sever) the series connection at the input and determine the resistance seen looking into the OUTPUT side of the feedback network... 1 FEEDBACK NETWORK 2 R22 R2 = 9 KΩ R22 R1 = 1 KΩ R22 = 1KΩ + 9KΩ G. Kovacs ©1997 Page 243 ..
+ V cc = 10..7 V + V cc RC = 20 KΩ Q3 RS = 10 KΩ 9 KΩ Q1 + Q2 V'o RL = 2 KΩ V'i 1 mA 1 KΩ 1 KΩ 5 mA 9 KΩ .. rπ = β = β VT gm IC G.. ' R  r π3 + β3 + 1 1KΩ + 9KΩ  RL A' = Vo = C V'i r e1 + r e2 + RS + 1KΩ  9KΩ β1 + 1 β2 + 1 1KΩ + 9KΩ  RL 1KΩ + 9KΩ  RL + r e3 • Using the values for rπ and re of. ' RC  loading from Q3 circuit A' = Vo = V'i r e1 + r e2 + other loading in the CC emitter circuit this stuff is the load on the CC amplifier formed by Q1.V EE • Modifying our previously determined gain equations to take into account the base resistance of Q2 and the loading of Q3.. Gain of the Q3 stage • Plugging it all together and using (β + 1) scaling where necessary.V EE .Chapter 14: FEEDBACK 3) Draw the “A” circuit and determine its gain.. Kovacs ©1997 > r π3 ≈ 520 Ω Page 244 ..
1 ' R1 + R2 1 KΩ + 9 KΩ Vo 4) Compute the closedloop gain.1 Af = Vo = A' = = 8.1 0....Chapter 14: FEEDBACK r e1 = r e2 = VT ≈ 52 Ω IC r e3 = VT ≈ 5. the gain would be. ' A' = Vo = 84.1 V'i 4) Now find the “β” circuit.4 0. ' 84. (simple voltage divider) R2 = 9 KΩ + + Vf '  R1 = 1 KΩ Vo'  ' 1 KΩ β ≡ Vf = R1 = = 0.. R 9kΩ = 10 A V = 1 + f = 1 + Rg 1kΩ G...2 Ω IC3 • The result is.997 = 84.94 ' 1 + A' β 1 + 84. Kovacs ©1997 Page 245 .1 Vs • Compare this to the "ideal" noninverting opamp configuration..
41 = 201 KΩ G. KEY: you FIRST have to take into account loading and go through the feedback stuff to compute the input and output resistances! + V cc = 10.2 KΩ + 101 52 + 101 1 KΩ  9 KΩ = 21.2 kΩ) ..V EE 5 mA 9 KΩ Ro R'of .Chapter 14: FEEDBACK 5) To get the input and output resistances. Rif = Ri 1 + A' β = 21.V EE Ri = 10 KΩ + r π1 + β1 + 1 r e2 + β1 + 1 1 KΩ  9 KΩ β2 + 1 looking at Q2's emitter resistance through the base of Q1 looking at the 1KΩ  9KΩ resistance through the base of Q1 and through the emitter of Q2 • Which gives (after computing that rπ1 ≈ 5.7 V + V cc RC = 20 KΩ Q3 RS = 10 KΩ 9 KΩ Q1 + Q2 V'o RL = 2 KΩ V'i 1 mA 1 KΩ 1 KΩ Ri R'if . Kovacs ©1997 Page 246 . Ri = 10 KΩ + 5.. you need to compute them for the “A” circuit and then apply the feedback equations that you know for the seriesshunt case.4 KΩ 9.4 KΩ 101 • Applying the appropriate feedback equation.
. Rof = Ro = 181 Ω = 19. we need to remove the effects of RL in parallel with it. 1 + 1 = 1 R'of RL Rof • This gives.1 19.. ' Rif = R if . which gives. Ro = 2 KΩ  1 KΩ + 9 KΩ  5.1 Rof RL R'of = 1 = 19.Chapter 14: FEEDBACK • The actual input resistance of the amplifier (after RS) can be determined by subtracting RS..24 Ω 2 KΩ G..2 Ω + 20 KΩ = 181 Ω 101 • Applying the appropriate feedback equation. Kovacs ©1997 Page 247 .24 Ω 1 + Aβ 9.. Ro = RL  1 KΩ + 9 KΩ  r e3 + RC β3 + 1 looking at Rc through the emitter of Q3. we start with the output resistance of the “A” circuit.10 KΩ = 191 KΩ • Similarly. > R'of = 1 1 .RS = 201 KΩ ..41 • Finally. to get the output resistance of the amplifier ALONE (Rof’).42 Ω 1 .
Chapter 14: FEEDBACK 11: RULES FOR SERIESSERIES (TRANSCONDUCTANCE AMPLIFIER) I'o BASIC AMPLIFIER RL ' A ≡ Io V'i V'i RS +  R11 R22 R11 1 FEEDBACK NETWORK 2 SEVER the SERIES SEVER the SERIES 1 FEEDBACK NETWORK 2 R22 I1 = 0 + V'f 1 FEEDBACK NETWORK 2 ' β ≡ Vf I'o I1 = 0 I'o G. Kovacs ©1997 Page 248 .
Kovacs ©1997 Page 249 .Chapter 14: FEEDBACK 12: RULES FOR SHUNTSHUNT (TRANSRESISTANCE AMPLIFIER) RS I'i R11 BASIC AMPLIFIER RL ' A ≡ Vo I'i R22 + V'o  R11 1 FEEDBACK NETWORK 2 SHORT the SHUNT 1 SHORT the SHUNT FEEDBACK NETWORK 2 R22 ' β ≡ If V'o V1 = 0 I'f + V1 = 0 1  FEEDBACK NETWORK 2 +  V'o G.
Kovacs ©1997 Page 250 .Chapter 14: FEEDBACK 13: RULES FOR SHUNTSERIES (CURRENT AMPLIFIER) RS I'i R11 ' A ≡ Io I'i I'o BASIC AMPLIFIER RL R22 R11 1 FEEDBACK NETWORK 2 SEVER the SERIES 1 SHORT the SHUNT FEEDBACK NETWORK 2 R22 ' β ≡ If I'o V1 = 0 I'f + V1 = 0 1  FEEDBACK NETWORK 2 I'o G.
Chapter 14: FEEDBACK 14: RULES FOR SERIESSHUNT (VOLTAGE AMPLIFIER) (repeated for convenience) V'i +  RS R11 BASIC AMPLIFIER R22 + V'o RL This model for the amplifier now takes into account loading from R and R s L ' A' ≡ Vo V'i R11 1 FEEDBACK NETWORK 2 SHORT the SHUNT SEVER the SERIES 1 FEEDBACK NETWORK 2 R22 ' β ≡ Vf V'o I1 = 0 Here we just want "pure" feedback (no worrying about loading anymore). Kovacs ©1997 Page 251 . I1 + V'f 1 FEEDBACK NETWORK 2 +  V'o G.
Kovacs ©1997 Page 252 .05 Positive feedback out of control! G. σ = +0.Chapter 14: FEEDBACK 15: STABILITY AND POLE LOCATION σ = 0.05 Damped oscillations. σ=0 An oscillator.
9 > AM = Ao for a DCcoupled amplifier.Chapter 14: FEEDBACK 15. you get. A s 1 + βA s AM (1 + A Mβ) = DC gain A f (s ) = s s 1+ 1+ ω H (1 + A Mβ ) Cut Off frequency • Thus we have another singlepole response.1 SINGLE POLE WITH FEEDBACK • Remember the response of a singlepole feedback amplifier that we previously looked at can be derived from the response of the amplifier without feedback.. A s = AM 1+ s ωH (Sedra & Smith use AM in section 8. Af s = • Substituting. but with a high cutoff frequency given by. ω Hf = ω H 1 + A Mβ • This means that the pole is sliding along the real axis from its original distance (frequency) ωH to its new one.2 and Ao in section 8... ω Hf . Kovacs ©1997 Page 253 .) • If you use the amplifier with negative feedback. the gain becomes. given by. G.
.01 β = 0.001 β = 0. the frequency response is changing in accord with the gainbandwidth product.Chapter 14: FEEDBACK Pole moves this way as you increase β ωH ω Hf = ω H 1 + AMβ NOTE that the new pole is given by 0 = 1 + A S β S ...00001 β = 0. β=0 β = 0.0001 β = 0. Kovacs ©1997 Page 254 . but real opamps always have more than one pole! • Meanwhile. • "Single pole" opamps are always stable.1 β=1 G.
1 ω p1 + ω p2 ± 1 2 2 ω p1 + ω p2 2 . S = . the two poles move together on the Splane until they overlap.2 TWO POLES WITH FEEDBACK • For a twopole basic amplifier.4 1 + A Mβ ω p1 ω p2 • As ß is increased. its frequency response can be written as. and then they separate along a line. Kovacs ©1997 Page 255 .Chapter 14: FEEDBACK 15. which can be written. A S = AM 1+ S 1+ S ω p1 ω p2 • The closedloop poles are obtained from 0 = 1 + A S β S . S2 + S ω p1 + ω p2 + 1 + A Mβ ω p1 ω p2 = 0 • With the closedloop poles given by. σ= ω p1 + ω p2 2 G.
but it is not reached until ω = ∞ ). Kovacs ©1997 Page 256 .ω p1 σ= ω p1 + ω p2 2 • For a secondorder system such as this. • So.Chapter 14: FEEDBACK . the poles never enter the right halfplane (the maximum phase shift is 180°. such an amplifier is UNCONDITIONALLY STABLE! G.ω p2 .
. there is always some value of Aβ that makes a pair of poles enter the RHP! This makes sense for any number of poles greater than two since there is a phase shift of at least 270 °. • There is a value of β at which these two poles enter the right halfplane. one can look at it as a maximum value of β or a minimum gain that guarantees stability. so that 180° is reached at a finite frequency. • Two of the poles become coincident.Chapter 14: FEEDBACK 15.. Kovacs ©1997 Page 257 . complex and conjugate.. • Because there is a value of β for which the amplifier becomes unstable. causing the amplifier to become unstable. G.3 THREE OR MORE POLES WITH FEEDBACK • For a basic amplifier with three poles.
.gain at ω unity gain = gain at ω 180 . • Gain Margin is defined as the difference between the value of  Aβ  at the 180° phase frequency (ω180) and unity..1 • Phase Margin is defined as the difference between the phase angle at the unitygain frequency and at the frequency where the phase reaches 180°. Af s = A s 1 + βA s • If  Aβ  > 1 (i.. • Look at the basic feedback equation.180° the feedback will become POSITIVE!!! • This leads to instability (i.Chapter 14: FEEDBACK 15. there is still some GAIN left) when – Aβ = . Phase Margin ≡ phase at ω 180 + phase at ω unity gain = 180 + phase at ω unity gain a negative number usually! G.. Gain Margin ≡ gain at ω 180 ..4 STABILITY • Loop gain = Aβ is a function of frequency. Kovacs ©1997 Page 258 .e.e... oscillation).
Microelectronic Circuits. other capacitance effects come into play as additional poles (sometimes there are three or more). • The major internal capacitance that causes this rolloff is often referred to as the “dominant pole” of the amplifier. Smith. G.. • This means that the openloop phase response of the amplifier will eventually reach 90° times the number of poles.Chapter 14: FEEDBACK Gain Margin 0 ω180 ω unity gain ω (log scale) 0 +90° 0° 90° 180° 270° 360° Phase Margin ω unity gain ω180 ω (log scale) (Figure adapted from A.. S. C. Sedra and K.. Kovacs ©1997 Page 259 . HRW Saunders. 1991) 16: COMPENSATION • Real opamps have an openloop gain rolloff with frequency that is approximately firstorder (20 dB/decade) over much of their useful bandwidth. • At higher and higher frequencies.
• However. if the phase crosses 180° before the gain falls to unity. G. This does. and those designed for highspeed operation.” which means that the component(s) required to guarantee stability are included on the chip itself. compromise the highfrequency response of the amplifier! • Most opamps available are “internally compensated. • If the phase crosses 180° at a frequency where the gain of the amplifier is less than unity.” which means that you. everything remains stable. Some of the older opamps. must choose external components to assure that the amplifier will not become unstable. almost any noise present in the circuit will have enough of a highfrequency component to cause the circuit to oscillate. however.Chapter 14: FEEDBACK • If the phase is less than 180° when the gain of the amplifier reaches a gain of unity (0 dB). oscillations will probably occur (since an inverted replica of the amplifier’s output is fed back into it)! If an opamp circuit is unstable. are “externally compensated. the amplifier is “unconditionally stable!” • The most common way to guarantee stability is to compensate the amplifier with some additional components that shape its frequency response so that its gain is less than unity by the time the phase hits 180°. the designer. Kovacs ©1997 Page 260 .
Chapter 14: FEEDBACK 17: "TENT" MODEL FOR VISUALIZING POLES AND ZEROS G. Kovacs ©1997 Page 261 .
• In order to understand this type of oscillator. but by examining the denominator of the above equation. The overall gain of the circuit with feedback.Chapter 14: FEEDBACK 18: LINEAR OSCILLATORS (VERY BRIEFLY!) • A linear oscillator ideally produces a pure sinusoidal output at a single frequency (hopefully). • To achieve linear oscillation.. Af(S). known to those who care as the Barkhausen Criterion. a linear amplifier must oscillate without external stimuli (other than a startup transient to get it going. is given by. since a little bit of intuitive understanding goes a long way!). Af S = AS 1AS βS • Without going into all of the details. perhaps). Kovacs ©1997 Page 262 . A S β S = 1 • This condition. a minor excursion into theory will be required (it’s worth it. and summed back into the input of the amplifier.. • What is required to make a linear oscillator (that works. appears to make the circuit blow up! G. it is easy to see that the overall gain can be made infinity by setting the “roundtrip” gain around the entire feedback loop so that. • A linear oscillator of gain A provides an output voltage v out that is fed through a feedback loop with gain β. that is!) is the arrangement shown below (this is just POSITIVE feedback).
however the fact that the overall gain is infinity means that the output of the circuit is some signal (to be determined!).. you might get an interesting mix of frequencies).) are covered in EE122 and many good EE reference books. • Intuitively. Kovacs ©1997 Page 263 .Chapter 14: FEEDBACK • Actually. it is a necessary condition for this type of oscillator (linear) to work. fosc = 1 2πRC • The details of the circuit (i. etc. actually getting it to start oscillating.. it is possible to obtain a very pure sinewave output (if it is met at multiple frequencies.. • The classic Wein Bridge Oscillator achieves this through a combination of negative and positive feedback. keeping it stable. G.. even with NO input at all! • If one can arrange it so that the Barkhausen Criterion is met at only a single frequency.e. Vout C R1 R R2 R C • The oscillation frequency is given by.