Design of Fully Differential Operational Amplifier With High Gain, Large Bandwidth and Large Dynamic Range

Download as pdf or txt
Download as pdf or txt
You are on page 1of 93
At a glance
Powered by AI
The thesis discusses the design of a fully differential operational amplifier with high gain, large bandwidth and large dynamic range. It covers topics like topologies, common mode feedback, frequency compensation techniques and the complete design flow.

The thesis discusses the design of a two-stage fully differential operational amplifier. It covers the selection of topology, design specifications, first and second stage design, compensation circuitry, common mode feedback and layout of the op-amp.

The thesis discusses the design methodology of a two-stage op-amp including the design of first and second stages, compensation techniques and common mode feedback circuitry.

Design of Fully Differential Operational Amplifier with High

Gain, Large Bandwidth and Large Dynamic Range


A Thesis Submitted in partial fulfillment of the
requirements for the award of degree of
Master of Technology
in
VLSI Design & CAD
Submitted by
Manish Kumar
Roll. No: 60761010
Under Guidance of
Ms. Alpana Agarwal
Assistant Professor, ECED
Thapar University, Patiala
Department of Electronics and Communication Engineering
THAPAR UNIVERSITY
PATIALA-147004
July- 2009
i
ii
ACKNOWLEDGEMENT
With a deep sense of gratitude, I wish to express my sincere thanks to my supervisor,
Ms. Alpana Agarwal, for her immense help in planning and executing the work in time. The
confidence and dynamism with which she guided the work requires no elaboration. Her
company and assurance at the time of crisis would be remembered lifelong. Her valuable
suggestions as final words during the course of work are greatly acknowledged. What I know
today about the process of research, I learned from Ms Alpana Agarwal.
My sincere thanks are due to Prof. A. K. CHATTERJEE, head of the department,
for providing me constant encouragement. I specially thank Mr. Mohd Iliyas SMDP VLSI
Project Faculty, ECED for the help extended to me when I approached him and the
valuable discussion that I had with him during the course of thesis.
Special thanks are due to Mr. Sanjay Kumar and Mr. B. K. Hemant for extending
timely help in carrying out my important pieces work. The cooperation I received from other
faculty members of this department is gratefully acknowledged. I will be failing in my duty if
I do not mention the laboratory staff and administrative staff of this department for their
timely help.
I acknowledge the Hardware & Software support provided by Department of
Information Technology (Govt. of India) through project Special Manpower Development
Program for VLSI Design & Related Software (Phase - II).
I also want to thank my friends and parents, who taught me the value of hard work by
their own example. I would like to share this moment of happiness with my father, mother,
brother and sister. They rendered me enormous support during the whole tenure of my thesis
work.
Finally, I would like to thank all whose direct and indirect support helped me
completing my thesis in time.
Date: MANISH KUMAR
(60761010)
iii
TABLE OF CONTENTS
Certificate i
Acknowledgement ii
Table of Contents iii
List of Figures v
List of Tables viii
List of Abbreviations ix
Abstract xi
1. INTRODUCTION 1
1.1 Background 1
1.2 Need of Fully Differential Amplifier 2
1.3 Applications of Fully Differential Amplifier 3
1.4 Motivation 4
1.5 Organization of Thesis 4
2. LITEATURE SURVEY 5
2.1 Fundamentals of Ideal Fully Differential Amplifier 7
2.2 Need of Differential Output Op-Amp 8
2.3 Design Procedure Of Two Stage Op-Amp 10
2.4 Topologies for Fully Differential Op-Amp 14
2.4.1 Folded Cascode Fully Differential Amplifier 14
2.4.2 Telescopic Op-Amp 16
2.5 Common Mode Feedback 20
2.5.1 Primary Reason for Common Mode Feedback 21
2.6 Frequency compensation techniques 24
2.6.1 Parallel Compensation 25
2.6.2 Pole Spilliting Single Miller Compensation (SMC) 25
2.6.3 Single Miller Feedforward Compensation (SMFFC) 26
iv
2.6.4 Negative Miller Capacitances Compensation (SMCC) 27
3. DESIGN OF TWO STAGE FULLY DIFFERENTIAL 28
3.1 Topology Selection 28
3.2 Design Specifications 31
3.3 Design of Two Stage Fully Differential OTA 31
3.3.1 Design of First Stage Telescopic Amplifier 32
3.4 Compensation Of Two Stage Op-Amp 36
3.5 Common Mode Feedback Circuit 39
3.6 Output Stage Design 40
3.7 Schematic of Two Stage Fully Differential Amplifier 41
3.8 Layout of Op-Amp 43
3.8.1 Issues in Analog Layout 43
3.9 Complete Layout of Op-Amp 47

4. SIMULATION RESULTS AND LAYOUT 51
4.1 Schematic Simulation 51
4.1.1 AC Response 54
4.1.2 Common Mode Rejection Ratio (CMRR) 54
4.1.3 Power Supply Rejection Ratio (PSRR) 56
4.1.4 Input Common Mode Range (ICMR) 57
4.1.5 Transient Response 59
4.1.6 Transient Step Response 60
4.1.7 Settling Time 62
4.1.8 Output Dynamic Range of Op-Amp 63
4.2 Post Layout Simulations 67
4.3 Process Corner Simulations 68
4.4 Monte Carlo Simulations 75
5. CONCLUSION AND FUTURE PROSPECTS 78
5.1 Conclusion 78
5.2 Future Scope 78
REFERENCES 79
v
LIST OF FIGURES
2.1 Settling time 6
2.2 Input/ Output of fully differential op-amp 7
2.3 Input/ Output of balanced fully differential op-amp 8
2.4 Difference signal and differential output 9
2.5 Symbol of fully differential op-amp 9
2.6 Illustration of output common mode 9
2.7 Block diagram of basic op-amp 10
2.8 CMOS differential input stage amplifier 11
2.9 Schematic of two stage OTA 12
2.10 Small signal equivalent of two stage OTA 12
2.11 Fully differential folded cascode amplifier 15
2.12 Fully differential telescopic single stage op-amp 17
2.13 Voltage swing of cascode amplifier 18
2.14 Current mismatch problem 21
2.15 Fully differential amplifier with CMFB 22
2.16 Conceptual block diagram of CMFB loop 23
2.17 Phase margin plot 24
2.18 Single miller compensation(SMC) 25
2.19 Single miller capacitor nulling resistor compensation(SMCNRC) 26
2.20 Single miller feed forward compensation (SMFFC) 27
3.1 Telescopic amplifier 30
3.2 Half circuit of telescopic cascode 32
3.3 Two Stage telescopic Op-Amp 36
3.4 Miller compensation of two stage op-amp 37
3.5 Bode plots of loop gain of two stage op-amp 37
3.6 Common mode feedback schematic 39
3.7 Schematic of class A output stage 40
3.8 Schematic of two stage of fully differential op-amp 41
vi
3.9 Interdigited MOSFETs 44
3.10 Basic structure of common centroid layout 45
3.11 Stack layout design of MN5 48
3.12 Complete layout of fully differential op-amp 48
3.13 Complete LVS report 49
3.14 LVS netlist report 50
4.1 Test setup for AC response of op-amp 51
4.2 (a) Frequency response plot with C
L
= 5pf 52
(b) Frequency response plot with C
L
= 1pf 52
(c) Frequency response plot with C
L
= 3pf 53
4.3 Frequency response with temperature variations 53
4.4 Test setup for common mode response 54
4.5 CMRR at 27with C
L
=5pf 54
4.6 CMRR with temperature variation 55
4.7 CMRR with load capacitance variation 55
4.8 Test setup for PSRR 56
4.9 PSRR of op-amp 56
4.10 PSRR with temperature variation 57
4.11 Test setup for ICMR 57
4.12 ICMR of op-amp 58
4.13 ICMR with temperature variation 58
4.14 Schematic for sinusoidal transient response 59
4.15 Sinusoidal transient differential outputs 59
4.16 Sinusoidal transient differential outputs with temperature variation 60
4.17 Schematic for simulation and measurement of slew rate 60
4.18 Transient pulse response of op-amp for slew rate measurement 61
4.19 Effect of temperature on slew rate 61
4.20 Settle time at various tolerance level of differential output 62
4.21 Output noise voltages 64
4.22 Output swing at differential output at various 64
vii
4.23 Post layout gain phase plot 67
4.24 Post layout gain phase plot with temperature variation 67
4.25 CMRR of post layout 68
4.26 Settling time for SS corner 69
4.27 Slew rate for SS corner 69
4.28 Gain phase plot for SS corner 70
4.29 Settling time for SF corner 70
4.30 Slew rate for SF corner 71
4.31 Gain phase plot for SF corner 71
4.32 Settling time for FS corner 72
4.33 Slew rate for FS corner 72
4.34 Gain plot for FS corner 73
4.35 Slew rate for FF corner 73
4.36 Settling time for FF corner 74
4.37 Gain Phase plot for FF corner 74
4.38 Monte Carlo simulation results 75
viii
LIST OF TABLES
2.1 Performance of four different topologies 14
3.1 Target specifications of design 31
3.2 Aspect ratio of input stage transistor 35
3.3 Aspect ratio of transistors and their functions in op-amp 42
3.4 Capacitance values 42
3.5 Biasing voltages 42
3.6 Primary trades off of our topology 44
4.1 AC result due to load capacitance variation 53
4.2 CMRR with load capacitance variation 56
4.3 Show the effect of temperature on slew rate 62
4.4 Settling time at differential output 63
4.5 Dynamic Range of op-amp at various frequencies 66
4.6 Simulation Results of fully differential Op-Amp 66
4.7 Post layout simulation results 68
4.8 Results of process corner simulations 75
4.9 Results of Monte Carlo simulation 77
ix
LIST OF ABBREVIATIONS AND SYMBOLS
Symbol Quantity Units
Charge carrier mobility cm
2
/V
S
A
o
DC open-loop gain dB
Av Closed loop voltage gain dB
B Bandwidth Hz
C
gs
Gate-source capacitance f
CMRR Common-Mode Rejection Ratio dB
C
L
Load capacitor f
C
OX
Normalized oxide capacitance f/m
2
DR Dynamic Range dB
DM Differential mode signal
F Frequency Hz
GBW Unity gain bandwidth Hz
g
m
Trans-conductance
-1
g
m,n
Trans-conductance of n-transistor
-1
g
m,p
Trans-conductance of p-transistor
-1
g
m,T
Total trans-conductance
-1
ICMR Input Common Mode Range dB
I
d
Drains current A
K Boltzmanns constant J/K
K
p
PMOS process trans-conductance parameter A/V
2
K
n
NMOS process trans-conductance parameter A/V
2
L Channel length m
W Channel width m
LVS Layout Vs Schematic
PSRR Power Supply Rejection Ratio dB
P
peak-signal
Peak to peak signal power V
2
/Hz
x
P
noise
Output noise power V
2
SNR Signal-to-Noise Ratio dB
SR Slew rate V/s
UGB Unity gain bandwidth Hz
V
CM
Common-mode input voltage V
V
DD
Positive supply V
V
o,swing
Output voltage swing V
V
DS
Drain-source voltage V
V
d,sat
Saturation voltage V
V
GS
Gate-source voltage V
V
on
Output noise voltage V
GND Ground
V
th
Thermal voltage V
V
tn
Threshold voltage V
V
tn0
Threshold voltage at V
sb
=0V V
S-S Slow-Slow
S-F Slow-Fast
F-S Fast-Slow
F-F Fast-Fast
Z Impedance
xi
ABSTRACT
This thesis work presents the full custom design of a two-stage fully differential CMOS
amplifier with outstanding characteristics of high unity-gain bandwidth and large dynamic
range at output. The circuit has been designed. The simulation results in TSMC 0.35m
CMOS process from a 3.3V voltage-supply demonstrate that the designed operational
amplifier has satisfied state-of-the-art design specifications with a dedicated optimization
technique. This fully differential op-amp can be used in a pipeline ADC stage. The amplifier
consists of four parts, a telescopic differential amplifier that provides most of the open loop
gain, second stage common source amplifier that provide the output swing, a common mode
feedback circuit to maintain the constant output voltage and a compensation circuitry for
achieving desired phase margin with high unity gain bandwidth.
The amplifier exhibit large output swing of 3.1 V peak-to-peak, a fast settling time 2.8 ns to
the accuracy of 1%. While the open loop amplifier have the considerable gain as high as
86.02 dB, the differential AC loop unity gain bandwidth reaches 270 MHz and the dynamic
range at the output gets 85.15dB with phase margin of 50. The present work also
demonstrate how to efficiently design and further optimize the circuit topology and
parameters in order to meet some strict specifications given in advance for a specific
engineering project.
1
CHAPTER 1
INTRODUCTION
This chapter discusses the background and motivation behind the fully differential
operational amplifier and ideal requirements of fully differential operational amplifier for
input stage. It also discusses various applications of fully differential operational
amplifier.
1.1 Background
Constraints imposed by advanced IC process technologies, modern electronic system
requirements, and the economics of circuit integration have created new challenges in
analog circuit design. With the advancement of CMOS process technologies and the
increasing popularity of battery-powered mobile electronic systems comes the demand for
lower-voltage analog circuit designs. In addition, the drive to reduce system costs is
forcing the integration of both analog and digital circuitry onto a single die. Both of these
changes have a detrimental impact on analog circuit performance. With a reduction in
power supply voltage comes a decrease in both the peak SNR and the dynamic range of
an analog circuit. Integrating analog circuitry and noisy digital circuitry on the same die
further degrades analog performance due to noise injection through a common power
supply and/or power distribution network, the die substrate, and/or capacitive coupling
between conductors.
Many analog design techniques and methodologies have been devised to enable high
performance analog signal processing in todays environment. Fully differential analog
signal processing is one technique that has become widespread because it reduces the
problems associated with both reduced signal swings and noise coupling. Using a
differential design technique effectively doubles the maximum signal swing in the circuit.
Also, all external noise sources that influence both signal paths of a balanced differential
system in the same way, to a first order approximation, will be rejected. This is due to the
fact that, in a differential system, the signal of interest is the difference between the
signals in the two signal paths. Thus any noise common to both signal paths will be
subtracted away. For the same reason, the total harmonic distortion of the circuit due to
non-linear elements can be reduced. Each distortion component at a frequency that is an
2
even harmonics of the fundamental signal frequency will be subtracted away from the
differential signal because it is a common in both signal paths [1].
Operational amplifiers are the backbone for many analog circuit designs. It is a
fundamental building block for many circuit designs that utilize its high gain, high input
impedance, low output impedance, high bandwidth and fast settling time. Operational
amplifier (Op-Amp) is one of the basic and important circuits which have a wide
application in several analog circuits such as switched-capacitor filters, algorithmic,
pipelined and sigma-delta A/D converters, sample-and hold amplifiers etc. The speed and
accuracy of these circuits depend on the bandwidth and DC gain of the Op-amp. Larger
the bandwidth and gain, higher the speed and accuracy of the amplifier[2]. Operational
amplifiers are a critical element in analog sampled-data circuits, such as SC filters,
modulators. Higher and higher clock frequency requirement for these circuits translates
directly to higher frequency requirement for the Op-amp. A high gain bandwidth (GBW)
is essential for accurate dynamic charge transfer in an switch-capacitor (SC) circuit in a
short sampling period. Applications of the high speed opamp range from video amplifiers
to sampling circuits. Many fiber optic applications also require analog drivers and
receivers operating in the megahertz range wide-band op amps are necessary.
In recent years, CMOS analog-digital converters (ADC) are expected to achieve a high
gain and unity gain frequency, and a fast settling time. However, the problem is that high
speed and high open-loop gain are two contradictory demands [3].
An integrated, fully-differential amplifier is very similar in architecture to a standard,
voltage feedback operational amplifier. Fully differential amplifiers have differential
outputs, while a standard operational amplifiers output is single-ended. There is typically
one feedback path from the output to the negative input in a standard operational
amplifier. A fully-differential amplifier has multiple feedback paths.
1.2 Need of Fully Differential Amplifier
1 Increase noise immunity, invariably, when signals are routed from one place to
another, noise is coupled into the wiring. In a differential system, keeping the
transport wires as close as possible to one another makes the noise coupled
into the conductors appear as a common-mode voltage. Noise that is common
to the power supplies also appears as a common-mode voltage. Since the
differential amplifier rejects common-mode voltages, the system is more
immune to external noise.
3
2. Increased output voltage swing, due to the change in phase between the
differential outputs, the output voltage swing increases by a factor of 2 over a
single-ended output with the same voltage swing. This makes them ideal for
low voltage applications.
3. Reduced even order harmonics, expanding the transfer functions of circuits
into a power series is a typical way to quantify the distortion products.
Taking generic expansion of outputs and assuming matched amplifiers, we get
..... ..........
3
3
2
2 1
+ + + =
+ in in in out
V K V K V K V and (1.1)
( ) ( ) .... ) (
3
3
2
2 1
+ + + =
in in in out
V K V K V K V (1.2)
taking the differential output
....... ) ( 2 2
3
3 1
+ + =
in in od
V K V K V where K
1,
K
2
... are constants. (1.3)
The quadratic terms gives rise to second-order harmonic distortion, the cubic
terms gives rise to third-order harmonic distortion, and so on. In a fully-
differential amplifier, the odd-order terms retain their polarity, while the even-
order terms are always positive. When the differential is taken, the even order
terms cancel.
4. Fully differential amplifier had large output dynamic range, due to its noise
immune property.
5. The differential pair provides a built-in level shift that allows for all-NMOS
devices in the signal path. This would allow for a rough 2X increase in speed
for the same power, or a decrease in power for the same speed.
6. Fully differential telescopic op-amp consume much less power than their
counter folded cascode fully differential op-amp.
1.3 Applications of Fully Differential Operational Amplifier
- Digitally programmable voltage gain amplifier.
- Fully differential amplifier can be used as pipeline ADC stage.
- Telescopic fully differential op-amp is employed as main stage of CMOS ADC.
- Switched- capacitor filter.
- RF modulator and audio systems.
- Fully differential op-amp use in Delta Sigma modulator.
4
1.4 MOTIVATION
The design of high-accuracy analogue circuits is becoming a difficult task with the
scaling down of supply voltages and transistor channel lengths of the current
mixed-signal integrated circuits. Most of this kind of circuits requires the use of
the highest performance active cell: the operational amplifier (op amp). Designers
are continuously working toward trades off solutions between gain, input/output
swings, speed, power dissipation and noise. Basically, the principal topologies of
op amps are based on the telescopic cascode, folded cascode, two-stage or gain
boosting schemes. Op-amp with active cascode circuits achieve to increase the
open-loop gain without adding cascade stage and cascode devices. In this way,
high speed circuits with low headroom can be obtained.
1.5 Organisation of Thesis
Introduction to fully differential operational amplifier its benefits over single ended
opamp and applications of fully differential opamps.
Chapter two starts with basic parameters of opamp. Discussing the various property
opamp and design procedure of two stage opamp. Comparative study of various
topologies of fully differential amplifier has been given. The idea of common mode
feedback technique has been present. How to use various compensation techniques for
high performance and high stable op-amp are also discuss.
Chapter three starts with the choice of topology for input stage of the fully differential
op-amp. Design of fully differential telescopic op-amp for targets specifications with
using common mode feedback and compensation techniques.
Chapter four presents the simulations results for schematic of fully differential amplifier.
It also discusses the results of post layout, process corner and Monte Carlo simulations.
Chapter five concludes the present work and some scope of further improvements, have
been suggested.

5
CHAPTER 2
LITERATURE SURVEY
This chapter examines a number of various input stage topologies for the fully differential
op-amp and explains how the input stage design is shaped by the constraints from
specifications. The related constraints are common-mode rejection ratio, DC open loop
gain, input offset voltage, settling time, slew rate, dynamic range and unity gain
bandwidth are the discussed in this context.
Input Common Mode Range:
One of the primary specifications of the op amp in design is to have the input common
mode range that includes ground for the single supply operation as well as mid supply
voltage for dual-supply operation.
DC Open Loop Gain
The ultimate settling accuracy is limited by the finite opamp DC gain. The exact settling
error is depends not only on the gain but also on the feedback factor in the circuit utilizing
the op-amp. Typically, the DC gain requirement is from 60 dB up to 100 dB. In some
circuits, such as a front-end S/H circuit, insufficient opamp DC gain results only in a gain
error which is usually tolerable. The DC gain, however, has to be constant over the op-
amp output voltage range in order to avoid harmonic distortion [3].
Settling Time
It takes a finite time for a signal to propagate through the internal circuitry of an op amp.
Therefore, it takes a certain period of time for the output to react to a step change in the
input.
The settling time consists of 30% of slewing time and 70% of linear settling time. The
settling time constant is given by [3]:

m
L
g
C
= (2.1)
The number of settling time constants to obtain the required accuracy is given by
Williams et al [3]:

2
s
T
n = (2.2)
6
If the system has no slew rate limiting then the number of time constants can be less than
five, i.e < 5[3].
Figure 2.1: Settling time
Slew Rate
The Slew rate (SR) is the current available to drive the capacitance present at the output
of the amplifier. Slew rate is defined as in [4]

L
bais
C
I
SR = (2.3)
Where, I
bais
is the bias current and C
L
is the output load capacitance. SR can also be
expressed as:
UGB
V
SR
ov
2
= (2.4)
Where, Vov is the overdrive voltage and UGB is the unity gain bandwidth. If the unity
gain bandwidth is kept constant, the slew rate is improved by increasing the overdrive
voltage. SR can be further improved by using larger lengths or decreasing the
transconductance by keeping the current and gain bandwidth constant. It is recommended
that the SR should be five times the sampling frequency of the system [3].
7
Unity Gain Bandwidth
Unity gain bandwidth (UGB) and gain bandwidth product (GBW) are similar and
specifies as the frequency at which differential DC gain of opamp is unity. GBW specifies
the gain-bandwidth product of the op amp in an open loop configuration and the output
loaded:
GBW= A
D
*f, where A
D
is differential DC gain and f is unity gain frequency (2.5)

Dynamic Range
Dynamic range is defined as:
|
|
.
|

\
|
=
noise
peaksignal
P
P
DR log 10 (2.6)
The peak signal power is the power of the maximum differential sinusoidal signal that
does not overload the amplifier. The noise power is the total noise at the amplifier output
integrated from 1Hz to infinity [5].
2.1 Fundamentals of Ideal Fully Differential Op-Amp
The differential op-amp has two input signals, V
i1
and V
i2
, and two output signals, V
O1
and V
O2
as shown in figure 2.2. However, the input and output signals of interest in this
system is the difference between the two input terminals and the two output terminals,
respectively. The difference between these signals is called the differential mode input
and differential-mode output, or V
iDM
and VoDM. If this is a balanced system with
balanced inputs, the input and output signals can be referenced to a common mode, or
average voltage, ViCM and VoCM, respectively as shown in figure 2.3. If the common-
mode voltage is set to analog ground, as is usually the case, then the following relation
holds:
2 1
V V =
Figure 2.2: Input/ Output of fully differential op-amp
8
There are 4 gain parameters of interest. Gain A
DD
relates the differential output
signal, Vo
DM
, and the differential input signal, Vi
DM
.
Vidm Vicm V
i
2
1
1
+ = Vidm Vicm V
i
2
1
2
= (2.7)
Vodm Vocm V
o
2
1
1
+ = Vodm Vocm V
o
2
1
2
= (2.8)
2 1 i i
V V Vidm = | |
2 1
2
1
i i
V V Vicm = (2.9)

2 1 O O
V V Vodm = | |
2 1
2
1
O O
V V Vocm = (2.10)
Figure 2.3 Input/ Output of balanced fully differential op-amp
This is the most important gain parameter for a differential op-amp and ideally it
approaches infinity. This high differential gain parameter is what creates a differential-
mode virtual short between the Vi1 and Vi2 terminals when the op-amp is used in a
negative feedback configuration. Gain A
CD
relates the differential output signal, V
ODM
,
and the common-mode input signal, V
iCM
. Ideally, V
ODM
is not related to V
iCM
, so A
CM
approaches 0. The ratio of A
DD
to A
CM
is called the common-mode rejection ratio, or
CMRR of the op-amp. The higher the CMRR is better, and ideally it approaches infinity.
Gain A
DC
relates V
OCM
and V
iDM
. Ideally, the output common mode signal has no relation
to the input differential signal, so A
DC
approaches 0. Gain A
CC
relates V
OCM
and V
iCM
.
There should not be a relation between the common-mode output and common-mode
input, so ideally A
CC
approaches zero.
So input/output signal relationship for fully differential op-amp is [6]:
(

=
(

iCM
IDM
CC DC
CD DD
OCM
ODM
V
V
A A
A A
V
V
(2.11)
2.2 Need of Differential Output Op-Amp
This section discussed about the need fully differential opamp. The basic necessity of
fully differential operational amplifier are followings:
9
- Increased signal swing as shown in figure 2.4.
- Cancellation of common mode signals including clock feed through as shown in
figure 2.6.
- Cancellation of even-order harmonics
- Common mode output voltage stabilisation. If the common mode gain not small, it
may cause the common mode output voltage to be poorly defined.
Figure 2.4 Difference Signal and differential output[7]
Figure 2.5 Symbol of fully differential op-amp
Figure 2.6 Illustration of output common mode [7]
10
2.3 Design Procedure of Two Stage Opamp
This section presents a basic two-stage CMOS opamp design procedure that provides the
circuit designer with a means to strike a balance between two important characteristics in
electronic circuit design, namely noise performance and power consumption.
CMOS opamps are ubiquitous integral parts in various analog and mixed-signal circuits
and systems. The two-stage CMOS opamp is widely used because of its simple structure
and robustness. In designing an opamp, numerous electrical characteristics, e.g., gain-
bandwidth, slew rate, common-mode range, output swing, offset, all have to be taken into
consideration. Furthermore, since opamps are designed to be operated with negative-
feedback connection, frequency compensation is necessary for closed-loop stability.
Unfortunately, in order to achieve the required degree of stability, generally indicated by
phase margin, other performance parameters are usually compromised. As a result,
designing an opamp that meets all specifications needs a good compensation strategy and
design methodology.
Figure 2.7 Block diagram of basic op-amp [1]
A classic op-amp architecture is made up of three stages as shown in figure 2.7, even
though it is often referred to as a two-stage op-amp, ignoring the buffer stage. The first
stage usually consists of a high-gain, differential amplifier. This stage has the most
dominant pole of the system. A common source amplifier usually meets the specifications
of the second stage, having a moderate gain. The third stage is most commonly
implemented as a unity gain source follower with a high frequency and negligible pole
[2].
A typical CMOS differential amplifier stage is given in figure 2.8. Differential amplifiers
are often desired as the first stage in an op-amp due to their differential input to single
ended output conversion and their high gain.
The high gain requirement indicates that either a very high-gain single stage or two
modest gain stages are needed. The main disadvantage of the single stage implementation
is the low output range [7].
One of the biggest benefits of the two-stage approach is that the net open loop gain can be
achieved with two distinct stages, thereby eliminating much of the complexity involved in
11
designing a single gain stage and leaving the distribution of gain in each stage up to the
designers discretion. The first-stage does not have to drive the large capacitive load at the
output of the second stage. The most logical approach would be to have a large gain in the
first stage and a small gain and high swing in the second; the rationale being that low
second stage gain would not greatly amply first stage noise and high swing would give
better dynamic range [8].
Figure 2.8 CMOS differential input stage amplifier [5]
.
There are several benefits of a two-stage topology compared to that of a single stage
amplifier. First and foremost is that the net amplifier gain required can be realized in two
stages, thus de-coupling gain and headroom considerations. Typically, the majority of the
gain is realized in the first stage. Secondly, the second stage or output stage can be
designed to simultaneously source large currents and have a large output swing since its
gain requirements are significantly reduced by the two-stage topology. Also, the noise in
a two-stage amplifier is comparable to that of a single stage amplifier. This is because
very little input referred noise is contributed by the second stage due to the typically large
gain in the first stage. Consequently, the total output noise is primarily due to only the
first stage of the amplifier.
The major drawback of a two-stage design is that of speed. This is due to the additional
non-dominant pole added by cascading the two stages. However, a compensation
capacitor can be used to offset this issue. Thus the two-stage design with compensation
capacitor was chosen over the single stage. However, the two-stage design is more
12
flexible than that of a single stage design. This suggests that the slowness of the two-stage
design can be compensated for without having to trade off significantly in terms of gain,
and dynamic range. Upon comparison of the single stage to two-stage topologies, we
found that the advantages of the two -stage topology outweighed its disadvantages.
Consequently, we opted for a two-stage design over single stage design.
The two-stage operational transconductance amplifier (OTA) in figure 2.9 is a widely
used analog building block. Indeed, it identifies a very simple and robust topology which
provides good values for most of its electrical parameters such as dc gain, output swing,
linearity, CMRR, etc. Although the term OTA was originally conceived for operational
transconductance amplifiers with linear transconductance.
Figure 2.9 Schematic of two stage OTA
The design procedure is based on the following main parameters: noise, phase margin
(M

), gain-bandwidth product (f
GBW
), load capacitance (C
L
), slew rate (SR), input
common mode range (CMR), output swing (OS), and input offset voltage (due to
systematic errors).

Figure 2.10 Small signal equivalent of two stage OTA [7]
13
The equations for determining the various opamp characteristics can be shown as follows:
(a) Gain and Bandwidth
According to the equivalent circuit shown in Figure 2.10 under typical conditions
C gs B m
C C R g /
6 6
>> ,
C L A m
C C R g /
6
>> and R
C
<< R
A
, R
B.
The dc gain of opamp is given by
A
v0
=g
m1
g
m6
R
A
R
B
(2.12)
The op-amps dominant pole frequency and unity-gain bandwidth, also commonly known
as gain-bandwidth, can be found to be
C B A m
P
C R R g
6
1
1
= and
C
m
U
C
g
1
= =
1 1 0 P m V
g A = (2.13)
Where
P1
and
U
are dominant pole and unity gain bandwidth respectively. The transfer
function of two stage OTA is given by [7]:
( )
3
6
2
6 6 6
6
0
1
1
1
s C C C R R R s C C C C C C R R s C R R g
R
g
sC
A
V
V
C L g C B A C L C g L g B A C B A m
C
m
C
V
ID
OUT
+ + + + +
|
|
.
|

\
|

=
(2.14)
(b) Output Swing
By defining
OUT
HR
V as the voltage head room voltage at output, i.e

+ OUT
HR
V =
V
DD
V
out
(max) and
OUT
HR
V
=
V
out
(min) V
SS

(c) Common Mode range
If we define V
CM
as the opamp input common mode range i.e
V
CM
+
= V
DD
- V
CM
(max) and V
CM
-
= V
CM
(min) V
SS
According to Figure 2.8, it can be shown that
V
CM
+
= V
eff3
V
tn
and V
CM
-
= V
eff5
+ V
eff1,2
+ V
tn
(d) Internal Slew Rate
The slew rate associated with C
C
is found to be

C
D
C
I
SR
5
=

(2.15)
The slew rate associated with C
L
is found to be

L
D D
C
I I
SR
5 7

= (2.16)
Combining both (2.15) and (2.16) we obtain
) (
7 L C D
C C SR I + = (2.17)
14
2.4 Topologies for Input Stage Fully Differential Amplifier
This section discusses various topologies for the input differential amplifier for fully
differential amplifier.
There are many different topologies that can be considered for this design. Four of the
most relevant methods are compared in Table 2.1 [2, 9].
Table 2.1 Performance of four different topologies [2]
Topology Gain Output swing Speed Power
Consumption
Telescopic
Cascode
Medium Medium Highest Lowest
Folded Cascode Medium Medium High Medium
Two Stage High Highest Low Medium
Regulated
Cascode
High Medium Medium High
From the above table it is clearly visible that telescopic and folded cascode opamp are
best suited for high speed application although the gain provided by them is medium. As
in this thesis work high speed opamp is being desired so folded cascade and telescopic
opamp are studied in detail.
2.4.1 Folded Cascode Fully Differential [10-13]
A fully-differential folded-cascode op-amp is shown in figure. 2.11. Four current sources
are needed to drive the amplifier. The current sources are implemented using the cascode
technique. A folded topology enables the amplifier to have a high output voltage swing at
the cost of having a common-mode level sensitive to device mismatches. The folded
cascode op amp has a push pull output stage which can sink or source current from the
load. The exact match of the currents in the differential amplifier is not demanded by the
folded cascode op amp since extra current can flow in or out of the current mirrors. While
the bias current of the conventional cascode delivers the current to both the input devices
and the cascode devices since they are stacked together, the bias current I
SS
of the folded
cascode supplies only the input devices. Additional bias currents are required to add
necessary bias current. In general, the folded cascode connection dissipates more power.
The gain of a folded cascode op amp is normally lower than that of a corresponding
conventional cascode op amp due to the lower impedance of the devices in parallel. A
15
folded cascode op amp has a pole at the folding connection which is lower compared to
that node pole of the conventional cascode op-amp.
Figure 2.11 Fully differential folded cascode amplifier
This is due to the larger parasitic capacitance of extra and possibly wider devices in the
folded structure. Sometimes this low folding pole can self-compensate a folded cascode if
the phase margin is good enough.
By applying good approximations, the voltage gain of the Op-Amp is given by:

O m V
R G A = (2.18)
where ( ) ( )
9 1 7 7
||
O O O m O
r r r g R = || ( )
3 5 5 O O m
r r g
There are two dominant poles and a zero are considered in the studying of the frequency
response:
1. First dominant pole is given by the high output resistance(R
O
) occurs at output
node [11]:

O O
P
C R
1
1

= (2.19)

7 5 D D L O
C C C C + + = (2.20)
16
2. The second pole which have frequency much greater than dominant pole
frequency and is given by

CASC CASC
P
C R
w
1
2

= (2.21)

9 7 1 D S D CASC
C C C C + + = (2.22)

9 7 1
7
|| || ||
1
O O O
m
CASC
r r r
g
R
|
.
|

\
|
~ (2.23)
3. Due to the existing a second signal path through C
gd1
, a right half plane zero is
introduced in transfer function is calculated as

1
1
1
gd
m
Z
C
g
= (2.24)
There is pole at folding point i.e the source of M7 & M8 is quite closer to the origin that
associated with the sources of cascode devices.
The maximum output voltage swing of the folded cascode with proper choice of bias
voltages, the lower end of the swing is given by V
OD3
+V
OD5
and the upper end by
( )
9 7 OD OD DD
V V V + . Thus, the peak-to-peak swing on each side is equal to
) (
9 7 5 3 OD OD OD OD DD
V V V V V + + + .
Thus, voltage swing of folded casode amplifier is slightly higher than that of telescopic
configuration. This advantage comes at the cost of higher power dissipation, lower
voltage gain, lower pole frequencies and higher noise [12, 14].
2.4.2. Telescopic OP-AMP
Cascode configurations may be used to increase the voltage gain of CMOS transistor
amplifier stages. This structure has been called a telescopic-cascode opamp because the
cascodes are connected between the power supplies in series with the transistors in the
differential pair, resulting in a structure in which the transistors in each branch are
connected along a straight line. The main potential advantage of telescopic cascode op-
amps is that they can be designed so that the signal variations are entirely handled by the
fastest-polarity transistors in a given process [15, 16]. In the first stage, we were simply
looking for a configuration that allowed for high gain, low noise and minimal current
since output swing is less critical. The folded cascode and the telescopic configurations
were considered since we required at least one cascoded stage for a gain on the order of
( )
2
o m
r g . A high swing configuration still needs to be used to insure that all the devices in
this stage are in saturation. In comparing the two topologies, the folded cascode has more
17
current legs and more devices in the signal path. This leads to larger static current and
more noise contributors [17]. On the other hand, the telescopic configuration has half as
many current legs and fewer devices.
Figure 2.12 Fully Differential Telescopic Single Stage op-amp
A telescopic cascode op-amp, as shown in figure 2.12, typically has higher frequency
capability and consumes less power than other topologies. Its high-frequency response
stems from the fact that its second pole corresponding to the source of the n-channel
cascode devices is determined by the transconductance of n-channel devices as opposed
to p- channel devices, as in the case of a folded cascode. Also the parasitic capacitance at
this node arises from only two transistors instead of three, as in the latter. The single stage
architecture naturally suggests low power consumption.
M1 M2
M2A
M1A
M3
M3A
M4
M4A
M5
Vcm
Vb1
Vb2
Vb3
Vi+ Vi-
Vdd
Vss
18
The disadvantage of a telescopic op-amp is severely limited output swing. It is smaller
than that of the folded cascode because the tail transistor directly cuts into the output
swing from both sides of the output. In the telescopic op-amp shown in figure 2.12, all
transistors are biased in the saturation region. Transistors M1M2, M7M8, and the tail
current source M9 must have at least
DSAT
V to offer good common-mode rejection,
frequency response, and gain [18].
The maximum output voltage depends on the common-mode input. However, this
limitation as well as the limitation on the common-mode input range can be overcome in
switched-capacitor circuits. Such circuits allow the op-amp common-mode input voltage
to be set to a level that is independent of all other common-mode voltages on the same
integrated circuit [19].

Figure 2.13 Voltage swing of cascode amplifier [20]
The voltage swing of cascode amplifier is being shown in figure 2.13.
This property holds because the only coupling of signals to the op-amp inputs is through
capacitors, which conduct zero dc current even with a nonzero dc voltage drop.
OV DD OmaX
V V V 2 = (2.25)
This equation (2.25) shows that the maximum output voltage of a telescopic op amp that
consists of the first stage with optimum common-mode input biasing is three overdrives
19
less than the positive supply. This result stems from the observation that three transistors
are connected between V
DD
and the output.
OV SS OMIN
V V V 3 + = (2.26)
To determine the minimum required supply voltage difference, we will subtract (2.25)
from (2.26), this gives
OV SS DD MIN O MAX O
V V V V V 5 ) (
) ( 1 ) ( 1
= (2.27)
Assuming that the magnitudes of all the overdrives are all equals. Rearranging this
equation gives
OV MIN O MAX O SS DD
V V V V V 5 ) (
) ( ! ) ( 1
+ = (2.28)
This equation shows that the minimum difference between the supply voltages in a
telescopic cascode op amp must be at least equal to the peak-to-peak output signal swing
plus five overdrive voltages to operate all transistors in the active region. For example,
with a peak-to-peak output swing of 1 V and V
OV
= 100 mV for each transistor, the
minimum difference between the supply voltages is 1.5 V. To avoid this problem,
transistors in practical op amps are usually biased so that the magnitude of the drain-
source voltage of each transistor is more than the corresponding overdrive by a margin of
typically at least one hundred milli volts. The margin allowed for each transistor directly
adds to the minimum required supply voltage difference.
The extra three overdrive terms stem from the two cascode devices and the tail current
source. The overdrives from the cascodes should be viewed as the cost of using cascodes.
A fully differential cascode op amp is shown in Figure 2.12. Compared to its single-ended
counterpart; the main difference is that diode connections are removed from transistors
M
3
and M
3A
. Also, the gates of cascode transistors M
1A
-M
4A
are connected to bias
voltages here. The op-amp outputs are taken from the drains of M
IA
and M
2A
.The
resulting circuit is symmetric with each output loaded by a cascoded current source [2].
An advantage of the topology in figure.2.12 is that the DM signal path consists only of
PMOS transistors. That is, only the PMOS transistors conduct time-varying currents. The
NMOS transistors conduct constant currents. Such a configuration maximizes the op-amp
speed because n-channel transistors have higher mobility and f
t
than their NMOS
counterparts.
20
The telescopic-cascode topology, although has extremely high gain on the order of
( )
3
O m
r g can provide enough gain and minimize power due to the fact that we have
current flowing through only one branch [21].
To minimize differential noise, it is vital to keep a large V
eff
in all current sources in
differential branches. So if we reduce V
eff
in those transistors in hopes of increasing the
dynamic range, we risk increasing its effective noise factors thus nullifying any potential
dynamic range increase [8].
The second stage differential amplifier can provide a large output swing thus relaxing the
output noise allowed. Meanwhile, while most of the gain will be generated in the first
telescopic stage, it will not be driving the large capacitive load and can be optimized for
gain and range. This division of roles allows a large degree of flexibility in its design.
Meanwhile, while most of the gain will be generated in the first telescopic stage, it will
not be driving the large capacitive load and can be optimized for gain and range. This
division of roles allows a large degree of flexibility in its design. However, in any two-
stage amplifier design, a potential compensation capacitance must be placed between the
two stages to adjust its phase margin for settling requirements. The second stage may be
replaced with a common-source amplifier where the increased output swing would be at
the cost of decreased CMRR.
2.5 Common Mode Feedback
For proper operation of a fully differential amplifier, common mode feedback (CMFB) is
required to fix the voltages at high impedance nodes in the circuit to their desired values.
Because we employ a two stage design with two inversions, the CMFB also must be
inverting. This is accomplished by a switched-capacitor circuit and PMOS differential
pair which adjusts the common mode level of the first stage by either injecting current
into or bleeding current from the input legs as needed. The common mode output of the
first stage is set to the point which minimizes the quiescent current in the second stage.
The common mode voltage of the second stage is also dynamically adjusted using the
transistors M1 and M2. These transistors help to correct the inherent imbalance in pulling
between NMOS in PMOS in a class AB stage during switching. In order not to degrade
the overall speed of the amplifier, the unity gain bandwidth of the CMFB circuits must be
greater than that of the main amplifier.
21
2.5.1 Primary Reason for Common Mode Feedback
Since each of input transistors carries a current of I
SS
/2, the CM level depends on how
close I
D3
and I
D4
are to this value. In practice, as expemilfied by figure 2.14, mismatches
in the PMOS and NMOS current mirrors defining I
SS
and I
D3,4
create a finite error
between I
D3,4
and I
SS
. For high gain amplifiers, we wish a p-type current source to balance
an n-type current source. As illustrated in figure 2.14, the difference between I
P
and I
N
must flow through intrinsic output impedance of the amplifier, creating an output voltage
change of ( )( )
N P N P
R R I I || . Since the current error depends on mismatches and
N P
R R || is quite high, the voltage error may be large, thus driving p-type and n-type
current sources into triode region. As a general rule, if the output CM level cannot be
determined by visual inspection and requires calculation based on device properties. Thus
we emphasize that differential feedback cannot define the CM level.

Figure 2.14 Current mismatch problem
The sensing method suffers from an important drawbacks, it limits the differential output
swings. Since the common-mode loop gain is not large enough to keep the common-mode
voltages steady we must add an external circuit to implement a high-gain Common-Mode
Feedback (CMFB) loop. A block diagram of a fully-differential op-amp with CMFB is
shown below in figure 2.15 the gain parameters A
DD
, A
CD
, A
DC
. The gain parameters A
DD
,
A
CD
, A
DC
, and A
CC
are the same as shown in figure 2.15. The new gain parameters A
SD
,
A
SD
, A
DS
, and A
CS
model how the CMFB circuit will effects the op-amps behaviour.
The gain parameters A
SD
and A
SC
model how the CM control signal Vs effects both V
ODM
and V
OCM
. Ideally, the CMFB circuit should keep V
CM
stable without influencing V
ODM
.
Thus, A
SC
should be large and ideally approach infinity and A
SD
should be small and
ideally approach 0. The gain parameters A
DS
and A
CS
relate the CM control voltage Vs to
V
ODM
and V
OCM
, respectively. We want the CM sense circuit to generate a control voltage
22
Vs which is dependant only on the output CM voltage and reject the differential mode
(DM) output voltage. Thus, we want A
DS
to be small and ideally approach 0, and A
CS
to
be large and ideally approach infinity. Given this the following approximations hold:
V
OCM
A
SC
*V
S
and V
S
A
CS
* V
OCM
. From this we can get one of the most important
performance parameters of the CMFB loop, namely the CMFB loop gain which is equal
to A
CML
=(A
SC
*A
CS
).
Figure 2.15 Fully differential amplifier with CMFB
To minimize the offset in V
OCM
, A
CML
should be designed to be as large as possible, and
ideally it approaches infinity. Since in a real world implementation the magnitude of A
SC
and A
CS
will be a function of frequency, we want a large magnitude for A
CML
with a
bandwidth as large as the bandwidth of the differential mode bandwidth of the op-amp. A
CMFB circuit averages both differential output voltages to produce a common mode
voltage V
CM
. Voltage V
CM
is then compared to a desired reference common-mode
voltage, V
CM
, usually equal to the average of the two power supplies, or analog ground.
The difference between V
CM
and V
CMS
is amplified and this error voltage is used to
change the common-mode bias current of the op-amp to force V
CM
and V
CMS
to be equal.
Figure 2.16 shows the connectivity between a current-mirror differential amplifier and its
CMFB circuitry.
Notice that the CMFB control voltage Vs changes the common-mode bias current of the
op-amp by controlling the gate to source voltage of M3 and M5. As Vs increases, the bias
current being sunk into the drain of M3 and M5 also increases which causes an equal
reduction in the voltage of nodes Vo+ and Vo-, thereby decreasing Vcm. As Vs
23
decreases, the bias current sunk into the drain of M3 and M5 decreases thereby increasing
Vcm.

Figure 2.16 Conceptual block diagram of CMFB loop [24]
The I/O signal relationships for fully differential Op-amp with CMFB [24]

(
(
(

=
(

S
iCM
iDM
SC CC
DC
SD CD DD
OCM
ODM
V
V
V
A A A
A A A
V
V
(2.28)
| | | |
(

=
OCM
ODM
CS DS S
V
V
A A V (2.29)
The current in the CMFB circuit does not need to be large as long as the currents through
the top and bottom of the OTA are fairly well balanced. Since the common mode
feedback circuit only adds to the bias current in the bottom of the circuit, it is expected
that the bias currents in top half will be slightly lower.
Most CMFB circuits can be divided into three general categories: Switched-Capacitor
(SC) CMFB circuits, differential difference amplifier (DDA) CMFB circuits or resistor-
averaged CMFB circuits. The major distinction between these three categories is the
technique used to average the differential output voltages to produce the common mode
voltage V
CM
.
24
2.6 Frequency Compensation Techniques
The single stage amplifier typically has good frequency response and could achieve a
phase margin of 90 assuming the gain bandwidth is ten times higher than the single pole.
However, the dc gain of the single amplifier is generally not high enough and is even less
for submicron CMOS transistors. In general, op amps require at least two gain stages. As
a result, op amp circuits have multiple poles. The poles contribute to the negative phase
shift and may cause phase angle -180, before the unity gain frequency. The circuit will
then oscillate due to the negative phase margin. It leads to the necessity of altering the
amplifier circuit to increase the phase margin and stabilize the closed loop circuit. This
process is called compensation.
Figure 2.17 Phase Margin Plot
The most straightforward way is to make the gain drop faster in order for the phase shift
to be less than 180 at the unity gain frequency. This method achieves stability by
reducing the bandwidth of the amplifier. The most popular pole splitting method uses this
procedure. Another compensation method pushes the phase crossover frequency out by
decreasing the total phase shift. In this case, the number of the poles of the op amp needs
to be minimized while still providing enough gain. Pushing the phase crossover frequency
out is the basic idea of approaches like introducing zeros to cancel the poles or using feed
forward paths to improve the phase margin without narrow-banding the bandwidth as
much as the pole splitting method does. The feed forward method modifies the open loop
transfer function or the closed loop transfer function to increase the phase margin [23].
The frequency response of the multistage amplifier is not as good as that of the single
25
stage and this amplifier has a higher probability of oscillation in feedback circuits. One
popular way to predict the closed loop stability is by measuring the phase margin of the
open loop gain response. PM must be greater than 0 for no oscillation to occur. A good
performing amplifier will need a PM of about 45 to 60. Otherwise, the amplifier may
exhibit ringing in the time domain and peaking in the frequency domain.
2.6.1 PARALLEL COMPENSATION
Parallel compensation is a classical way to compensate the op amp. A capacitor is
connected in parallel to the output resistance of a gain stage of the operational amplifier to
modify the pole. It is not commonly used in the integrated circuit due to the large
capacitance value required to compensate the op amp, which costs considerable die area.
2.6.2 POLE SPILITTING SINGLE MILLER COMPENSATION (SMC)
By putting a compensation capacitor between the input and output nodes of the second
inverting stage of the op amp, the dominant pole is created due to Miller feedback. This
method maintains a high midband gain for the op amp since the capacitor does not affect
the dc response of the amplifier.
Figure 2.18 Single Miller Compensation (SMC)
Figure 2.18 shows standard SMC topology. As the transistor gain of the second stage
increases, the dominant pole decreases and the non-dominant pole increases. In this way
the two poles are being split apart and stabilize the feedback amplifiers by greatly
narrowing the bandwidth. This simple pole splitting method also introduces a right half
plane zero which causes negative phase shift, as a result, the stability is made a little
poorer. The zero comes from the direct feed through of the input to the output through the
26
Miller capacitor. To eliminate the RHP zero due to the feed through and increase the
phase margin of the op amp, lead compensation which adds a nulling resistor in series
with the compensation capacitor (SMCNR) to increase the impedance of the feed through
path is reported.
.
Figure 2.19 Single Miller Capacitor Nulling Resistor (SMCNR)
The effect of the nulling resistor to the positions of the poles as well as that of the zero
and pointed out the pole splitting would break down if the resistor becomes too big. When
the resistor gets very large, there is no pole splitting since the compensation capacitor is
actually open circuit. Single Miller capacitor nulling resistor (SMCNR) configuration is
shown in figure 2.19.
2.6.3 SINGLE MILLER FEEDFORWARD COMPENSATION (SMFFC)
Many compensation techniques mentioned above are not suitable for large load
capacitors. The demand for lower power consumption, lower chip integration area,
capability for driving large capacitive loads and stable high gain bandwidth of amplifiers
calls for improved frequency compensation patterns. The topologies using a single Miller
capacitor in three stage amplifiers could greatly reduce the needed sizes of the
compensation capacitors compared to NMC related schemes and result in amplifiers with
smaller chip area. The presented SMFFC and the modified SMC with the additional feed
forward path from the output of the first stage to the output load stage.
The topology of the SMFFC op amp is represented in figure. 2.20. Instead of using pole
zero cancellation, SMC with one forward path adopts the separate pole approach for
compensation in the situation of large capacitive loads. SMFFC employs two forward
27
paths and provide a LHP zero to compensate the first non dominant pole to alleviate the
bandwidth reduction and improve the phase margin.

Figure 2.20 Single Miller Feed Forward Compensation (SMFFC)
The second and third poles of the amplifier would be placed at higher frequencies that
lead to a coarse single pole system for an easier frequency compensation strategy. The
appropriate selection of the moderate gain of the second stage will then decrease the
compensation capacitor size. Unfortunately, this method does not truly resolve the
compressed gain bandwidth issue due to the super high gain of the first stage and the
nature of the pole separation. Gain enhanced feed forward path compensation.
2.6.4 NEGATIVE MILLER CAPACITANCE COMPENSATION (NMCC)
The negative Miller capacitance compensates high speed CMOS op amp, that consists of
an operational transconductance amplifier (OTA) and a buffer. The buffer with a dc gain
of A
V
is used to detach the OTA from the load. The OTA is compensated with a capacitor
Cc connected between the input and output of the buffer. Assuming the op amp drives a
load with a parallel combination of a resistor R
L
and a capacitor C
L
, the effective
capacitance seen at the input of the buffer is ( )
V C in
A C C = 1 and
|
|
.
|

\
|
+ =
V
C L out
A
C C C
1
1 at the output of buffer. Since the gain of the buffer is positive
and smaller than one, the reflected Miller capacitor
|
|
.
|

\
|

V
C
A
C
1
1 at the output will be
negative. The total effective output capacitance is reduced to be smaller than the original
load capacitance due to the negative Miller capacitance. NMCC can be applied to drive a
large capacitive load. The experimental results show that the NMCC design shifts the first
non dominant pole to a higher frequency while keeping the position of the dominant pole
almost the same, could increase both the bandwidth and phase margin.
28
CHAPTER 3
DESIGN OF TWO STAGE FULLY DIFFERENTIAL
OPERATIONAL AMPLIFIER
3.1 Topology Selection
For high speed and high accuracy circuits, op amps with high open loop DC gain for the
settling accuracy of 1% in less than 30ns, high output swing due to large output dynamic
range specification of 85dB, high unity-gain bandwidth are required. This indicates that
the chosen topology must have large output swing as well as minimal noise. Our target is
to design an amplifier with 300MHz unity-gain bandwidth and a DC-gain higher than
80dB, with a 5pF load. This op amp is intended to work in a switched-capacitor integrator
in the first stage of a delta-sigma modulator and CMOS ADC.
The first, and probably the most important, step of the design is to select an amplifier
topology that possesses the ability to meet all the specifications using the least amount of
power. Due to the high close-loop gain and the extremely small allowable settling error
of 0.01%, a quick calculation reveals that the amplifier must be able to deliver an open-
loop gain of 80dB, which is on the order of ( )
3
O m
r g ~ ( )
4
O m
r g . Gain boosting is very
attractive because boosters can multiply up the gain of the main amplifier while
consuming minimal power. An important concern when utilizing gain boosters is the
pole-zero doublet, which must be placed at a high enough frequency to prevent it from
slowing the settling but not so high that it causes significant ringing around the local gain-
booster loop.
Topologies that will certainly satisfy this magnitude of DC gain include a two stage
amplifier with either a telescopic or folded cascode first stage, a gain-boosted amplifier or
a folded triple cascode. To avoid much complexity in the design process a two-stage
topology is chosen.
The 85dB dynamic range specification calls for a careful comparison between single-
stage and two stage topologies. In a two-stage implementation, the second stage can be
designed to provide a larger output swing of 3V, thereby reducing the capacitance driven
by the input transistors and consequently lowering the required current in the first stage.
However, the second stage generally needs to flow at least the same amount of current as
the first stage to allow for slewing during amplifier operation. The total currents required
29
for the two-stage topology would therefore be at least double that of the first stage.
Hence, if a single stage amplifier can achieve output swing higher than 2.3V (3.3 divided
by 2 ), it would dissipate lower power than a two-stage amplifier. In terms of circuit
topology, telescopic is the necessary choice over folded-cascode in order to realize this
benefit in decreased power since a telescopic amplifier consumes only half as much
power and produces less noise at the expense of a slightly tighter output swing. For the a
fore mentioned reasons, telescopic single-stage with gain boosting is chosen as the
topology of our amplifier. Output swing of at least 2V zero-to-peak is set as an additional
design specification, which leaves 1V for 5 stacking transistors. This, combined with a
large tail current stemming from the stringent settling accuracy requirement, requires us
to pursue a design with large transistors. To prevent the large input parasitic capacitance
from lowering the feedback factor, C
f
and C
s
would have to be increased accordingly,
leading to huge on-chip area consumption if we were to layout the amplifier. Fortunately,
area is not a specification or an objective to be optimized in this project. NMOS is
selected to be the input transistors over PMOS because smaller area is needed to realize
the same trans-conductance. Another shortcoming of this design is a poor CMRR because
we are forced not to cascode the tail current source due to limited voltage headroom.
Designing a one stage gain-boosted amplifier to have a high output swing, low noise, and
high gain at the same time is obviously not an easy task. Although two-stage design
seems to be less power efficient (four main current legs), it gives us more degrees of
freedom in the design. Our basic calculations indicates that by combining the high-output
swing and relatively low noise properties, two-stages amplifier tends to consume less
power than the gain-boosted one stage amplifier for reaching the given specifications. So
we select the two stage topology for our design. Noting that using two stages allows one
to separates the gain stage from the stage that determines the output swing, we further
looked into the two-stage design. We liked the fact that the main gain stage would not
have to drive the large capacitive load. Additionally the two-stage topology would give us
more degrees of freedom to optimize our design, at the cost of complexity. We made our
decision after verifying that a two-stage topology would not cause problems meeting the
speed requirements. We decided on the telescopic design over folded cascode for our first
stage because it has fewer current legs, consumes less power, and adds less noise to the
signal path. A large signal swing was not required of the first stage, which also is in
favour of the telescopic design. the telescopic amplifier has only four transistors in the
signal path contributing significant noise power, whereas the folded cascode has six
30
transistors. Because it consumes less power and has less noise, the telescopic amplifier
was chosen for the first stage.
Telescopic cascode op-amp as shown in figure 3.1, typically has higher frequency
capability and consumes less power than other topologies. Its high-frequency response
stems from the fact that its second pole corresponding to the source nodes of the n-
channel cascode devices is determined by the transconductance of n-channel devices as
opposed to p-channel devices, as in the case of a folded cascode. Also the parasitic
capacitance at this node arises from only two transistors instead of three, as in the latter.
The single stage architecture naturally suggests low power consumption.
Figure 3.1 Telescopic differential amplifier
M1 M2
M4
M3
M7
M5
M8
M6
M9
Vcm
Vb1
Vb2
Vb3
Vi+ Vi-
Vdd
Vss
Vo+
Vo-
C
L
C
L
31
3.2. Design Specifications
The target specifications of the design are given in the Table 3.1.
Table 3.1 Target Specifications of Design
Specification Target Value
DC open loop gain 80 dB
Unity gain bandwidth 300 MHz
Phase margin 50
f
-3dB
frequency 5KHz
Output dynamic range 85 dB
CMRR 85 dB
Output voltage swing 3V (peak to peak)
Power dissipation Minimize ( 5mW)
Slew rate 20 V/S
Settling time 30 nS
Load capacitance 5pF
Supply voltage 3.3V
3.3. Design of Two Stage Telescopic Fully Differential OTA
In this section we discuss various steps to follows in the designing of the fully differential
with having target specification as given in Table 3.1.
The first step of any two stage design is the design of input differential stage amplifier,
which is telescopic differential amplifier in our case. The input transistors of the
telescopic cascode stage are NMOS devices to maximize gm/Id.
For the second/output stage, a common source amplifier for each differential output was
utilized as typically done. Differential amplifiers as output stage would be more
complicated due to the design of another common mode feedback circuit. In addition, the
tail current transistor for the differential amp will only consume voltage headroom, thus
lowering the output swing. To ensure stability of the amplifiers, miller compensation is
used despite the cascode compensations better high frequency performance. Miller
compensation however, can be easily designed and will introduce non-negligible noise
depending on how the circuit is compensated.
32
3.3.1 Design of First Stage Telescopic Amplifier
Telescopic differential amplifier as shown in figure 3.1, all transistors must be in
saturation region of operation. Since schematic is of balance in nature so we adopt half
circuit method to find its dc gain.

Figure 3.2 Half circuit of telescopic cascode
Let us now consider the small signal characteristics of a cascode stage, assuming all
transistors operates in saturation. The voltage gain is equal to that of common source
stage because the drain current produced by input device must flow through the cascode
device. An important property of the cascode structure is its high output impedance. The
output impedance (R
OUT
) of the schematic as shown in figure 3.2 is calculated as;
( ) ( ) { } ( ) ( ) { }
4 4 3 3 3 2 1 2 2 2
1 || 1
O O O mb m O O O mb m OUT
r r r g g r r r g g R + + + + + + = (3.1)
We calculated the gain as

OUT m V
R G A ~ and
1 m m
g G ~ (3.2)
Where,
1 m
g is the trans-conductance of input transistor M1.
The transconductance g
m
of MOS transistor is defined as

GS
D
m
V
I
g

= at constant V
DS
(3.3)
( )
th GS OX n m
V V
L
W
C g = (3.3(a))
33
=
D OX n
I
L
W
C 2 (3.3(b))
=
th GS
D
V V
I

2
(3.3(c))
The total gain of the cascode structure is give by;
~
V
A ( ) ( ) { } ( ) ( ) { }
4 4 3 3 3 2 1 2 2 2 1
1 || 1
O O O mb m O O O mb m m
r r r g g r r r g g g + + + + + + (3.4)
(3.5)
Therefore, cascading of transistors increases the differential gain substantially but at the
cost of consuming more voltage headroom.
The power dissipation of two stage operational amplifier is given by;
P
d
=V
DD
(I
TAIL
+ I
SECOND
), where I
TAIL
is input tail current of first stage and I
SECOND
is total
sum of output branch current which is twice of individual output branch current.
Since input differential amplifier is gain stage of our design, the telescopic cascode
amplifier is of gain of the order of ( )
2
O m
r g . Since the total gain of operation amplifier is
80 dB, therefore we divide this much of gain as 55db for first stage and rest 30 db for
second stage which is a common source stage with appropriate phase margin.
To achieve the power dissipation (<5mW) and gain parameter. We distributed the current
as the current in second stage must be at least 4-5 times of the input stage current, so that
there is no overlapping of non-dominant pole with dominant pole for getting good phase
margin and proper high bandwidth. Let input tail transistor draw 100A of current, i.e.
I
SS
=100A=2I
D
.
( )
2
|
2
1
tp SG OX P DP
V V C I = (3.6(a))
( )
2
2
1
tn GS OX n Dn
V V C I = (3.6(b))
The technology parameters used in calculations are:
6
10 5 . 132

=
OX n
C A/V
2

n
=0.02 V
-1
V V
tn
549 . 0 =
6
10 12 . 50

=
OX p
C A/V
2

p
=|0.04| V
-1 V V
tp
68 . 0 =
And
D
O
I
r

1
~
In figure 3.1, the input common mode (CM) level and the bias voltages V
b1
and V
b2
must
be chosen so as to allow maximum output swings. The maximum allowable input CM
( ) ( )
4 3 3 2 1 2 1
||
O O m O O m m V
r r g r r g g A ~
34
level equals
9 1 OD GS
V V + =
9 1 1 OD OD th
V V V + + . The minimum value of V
B1
is given by
9 1 3 OD OD GS
V V + + . Similarly, =
max , 2 B
V ) (
7 5 OD GS DD
V V V + . In practise, some margin
must be included in the value of V
B1
and V
B2
to allow the process variation.
The telescopic cascode design starts with the sizing of the main differential input pair of
transistors M1 and M2 as shown in figure 3.2, using the desired phase margin and gain
bandwidth specifications. As these are the two factors that affect the input pair transistors
they are first designed to meet the required specifications. Care has to be taken not to
make the input pair too big to affect the bandwidth and at the same time making them big
enough to provide enough transconductance and hence the gain. The NMOS Ml and M2
cascode transistors are then sized to act as a buffer between the input pair and the output.
These transistors are sized such as not to load the output nodes with huge parasitic
capacitances so as to affect the bandwidth of the amplifier.
The PMOS cascode load transistors are designed to steer the required amount of current
through both the legs. Here again the PMOS cascode transistors are sized so as not to load
the output with huge parasitic capacitances. The overall gain is in general increased by
approximately the gain of the gain-boost amplifiers; the gain specifications of the gain-
boost amplifiers were thereby known. The unity gain frequency of the gain-boost
amplifiers should be large enough so that they do not significantly affect the frequency
behaviour of the overall amplifier. They will reduce the unity gain frequency of the
overall amplifier since by adding the gain-boost amplifiers to the output side, extra
capacitance and thereby some extra poles are added.
The gain had increased, but both the unity gain frequency as well as the phase margin was
degraded too much. The bias currents of the main amplifier were then increased and some
of the bias voltages were altered. Since the output DC-level of the gain boost amplifiers
can be chosen from a wider range than the bias voltages, this was another benefit of using
gain boosting. In a way they give both increased gain and voltage level shift although at
the cost of added complexity.
Let the gain for first stage is of 55 dB and tail current of 100A
I
D
=I
SS
/2=50A
Hence
D n
O
I
r

1
1
~ =1M
2 O
r and

D p
O
I
r

1
3
~ =0.5M
4 O
r
35
Keeping V
incm
=1.65V for getting maximum output swing. For biasing all the stack
transistors in saturation, we satisfy the condition of saturation of MOSFET
( )
DS t GS
V V V < for each transistor of cascode stage.
V
incm
=1.65V and V
tail
= 0.6V for current source tail transistor and also assuming the output
DC level at 1.65V for better response. By applying all conditions we gets biasing voltage
condition as follows as shown in half circuit figure 3.2.
V V
b
65 . 2
1
s ; V V
b
35 . 2
2
s and V V
b
15 . 2
3
> (3.7)
First before starting actual design it is important to fix length of the device. Length of the
device is normally kept three or four times of the minimum feature size of the selected
technology. Thus length selected for the design is L = 1.4m (four times of feature size).
After employing above all conditions equation 3.1- 3.7 and using some iteration on tool,
we design first stage of telescopic differential amplifier with following aspect ratio as for
figure 3.1.
Table 3.2 Aspect ratio of input stage transistors
Transistor M1,M2 M3,M4 M5,M6 M7,M8 M9
Aspect ratio 18.9/1.4 14/1.4 56/1.4 112/1.4 694.75/1.4

Now consider the fully differential telescopic cascode, in addition various useful
properties of differential operation, this topology avoids the mirror pole, thereby
exhibiting stable behaviour for a greater bandwidth. In fact, we identify one dominant
pole at each output node and one nondominant pole arising from output node. This
suggests the fully differential telescopic cascode circuits are quite stable.
The capacitance at node N in figure 3.2.

7 7 5 5 DB GD SB GS N
C C C C C + + + = (3.8)
The C
N
shunts the output resistance of M7 at high frequencies, thereby dropping the
output impedance of the cascode. The Z
out
of the single stage fully differential op-amp is;
( )
5 5 5
1
O N O m out
r Z r g Z + + = (3.9)
where body effect is neglected and ( )
1
7
||

= s C r Z
N O N
(3.10)
We have,
( )
1
1
7
7
5 5
+
+ ~
s C r
r
r g Z
N O
O
O m out
(3.11)
36
Now, we take the output load capacitance into account;

( )
s C s C r
r
r g
s C s C r
r
r g
s C
Z
L N O
O
O m
L N O
O
O m
L
out
1
1
) 1 (
1
1
1
1
||
7
7
5 5
7
7
5 5
+
+
+
-
+
+
= (3.12)
=
( )
( ) | | 1 1
1
7 7 5 5
7 5 5
+ + +
+
s C r C r r g
r r g
N O L O O m
O O m
(3.13)
Thus, the parallel combination of Z
out
and load capacitance (C
L
) still contains a single pole
corresponding to a time constant
N O L O O m
C r C r r g
7 7 5 5
) 1 ( + + .
3.4 Compensation of Two Stage Op-Amp
The two stage topologies may prove inevitable if the output swing voltage swing must be
maximized. Thus, the stability and compensation of such op-amps is of interest. The
circuit shown in figure 3.3, we identify three poles: pole at X (or Y), another at E (or F)
and third at A (or B). From our foregoing discussion, we know that the pole at X lies at
relatively high frequencies.
Figure 3.3 Two Stage telescopic Op-Amp
37
The small signal resistance seen at E in figure 3.4 is quite high, even the capacitance of
M3, M5 and M9 create a pole relatively close to the origin. At node A, the small signal
resistance is lower but the value of C
L
may be quite high. Consequently, we say that the
circuit exhibit two dominant poles. One of the dominant poles must move toward the
origin so as to place the gain crossover well below the phase crossover. However, the
unity gain bandwidth after compensation cannot exceed the frequency of the second pole
of the open loop system. Thus, if the magnitude of
P,E
is to reduced, the available
bandwidth is limited to approximately
P,A,
, a low value. Furthermore, the very small
magnitude of the required dominant pole translates to a very large compensation
capacitor.

Figure 3.4 Miller compensation of two stage op-amp
The circuit as shown in figure 3.3 with Miller capacitances gives two poles as follows:

( )( ) | | ( )
L GD C L E GD C L m S
P
C C C R C C C R g R + + + + + +
~
9 9 9
1
1
1
(3.14)
( )( ) | |
( ) ( ) | |
L E L GD C E GD C L S
L GD C L E GD C L m S
P
C C C C C C C C R R
C C C R C C C R g R
+ + + +
+ + + + + +
~
9 9
9 9 9
2
) ( 1
(3.15)
where, R
S
denotes output resistance of the first stage and R
L
=
12 10
||
O O
r r .

Figure 3.5 Bode plots of loop gain of two stage op-amp
38
These expressions are based on the assumption
2 1 P P
<< .Before compensation,
1 P
and
2 P
are of the same order of magnitude. For large C
L
the magnitude of output pole
approximated as
( )
L L
P
C R
1
2
~ and after compensation,
( )
L E
m
P
C C
g
+
~
9
2
.
In cascode topologies, the zeros are quite far from the origin, in two stage op-amps
incorporating Miller compensation, a nearby zero appears in the circuit. The right half
plane zero at =
Z

( )
9
9
GD C
m
C C
g
+
. This is because C
C
+ C
GD9
forms a parasitic signal
path from input to output. The presence of right half zero degrades the stability
considerably. The right half plane zero in two-stage CMOS op amps, given by
( )
GD m
C C g
C
+ / , is a serious issue because g
m
is relatively small and C
C
is chosen large
enough to position the dominant pole properly.
For moving or eliminating zero, there are many method have been develop, one approach
is to place a resistor in series with compensation capacitor, thereby modifying the zero
frequency. The zero frequency is given by:

( )
Z m C
Z
R g C
~
1
9
1
(3.16)
In practise we may even move the zero well into the left half plane so as to cancel the first
nondominant pole. This occurs if
( )
E L
m
Z m C
C C
g
R g C +

9
1
9
1
, (3.17)
that is,

C m
E C L
Z
C g
C C C
R
9
+ +
= (3.18)

C m
C L
C g
C C
9
+
(3.19)
because, C
E
is typically much less than C
L
+C
C
.
The above followed compensation technique is known as nulling resistor compensation
technique. The zero position is pushed away with a resistance in series with Cc.

|
|
.
|

\
|
+
|
|
.
|

\
|
+
|
.
|

\
|
+
~
2 1
2
1 1
1
1
P
s
P
s
C
g
R s
A
V
V
C
m
Z
O
in
O (3.20)
39
3.5 COMMON MODE FEEDBACK CIRCUIT
The high differential gain of a fully differential amplifier stabilizes the differential-mode
signals within the amplifier, but the common-mode signals can float. Extra circuitry,
called a Common-Mode Feedback (CMFB) circuit, is required to increase the common-
mode loop gain of the amplifier so that the common-mode signals are stabilized. This
CMFB circuit implements a negative feedback loop that must be compensated properly to
minimize the loop settling time and maintain stability. The design of the CMFB circuit is
more challenging than the actual op-amp design due to the difficulty of properly
compensating it. The differential amplifier stabilizes the common mode signal by
common mode feedback.

Figure 3.6 Common mode feedback schematic
The voltage VB is chosen such that M1 and M2 are in linear region. The (W/L)
1
and
(W/L)
2
, M1 and M2 are like parallel of two voltage dependent resistances.
( ) }
2
1
{
2
1
1 DS DS t o OX n
V V V V
L
W
C I
|
.
|

\
|
=
+
(3.21)
(3.22)
( )
2
3
2 1
2
1
t DS B OX n OUT
V V V
L
W
C I I I
|
.
|

\
|
= + = (3.23)
With common mode signal: if positive, I
OUT
increases.
With common mode signal: if negative, I
OUT
decreases.
( ) }
2
1
{
2
2
2 DS DS t o OX n
V V V V
L
W
C I
|
.
|

\
|
=

40
Common mode feedback circuits were used to regulate the common mode outputs for the
first and second stages. The dual differential pair sense amp configuration was used, with
a diode connected NMOS tail supplying the common mode control voltage to the bottom
NMOS transistors on the telescopic cascode amplifier.
3.6 OUTPUT STAGE DESIGN
Analog circuits are often required to drive large loads with wide signal swing. This
creates a need for output stages with wide signal swing capability and sufficient
transconductance relative to the load capacitance to meet bandwidth requirements. Output
stages, especially at low supply voltages, are limited to a single n-channel and p-channel
transistor as shown in figure 3.8 to maximise the signal swing.

Figure 3.7 Schematic of class A output stage
This allows the output range from V
dsat
to V
SUPPLY
- V
dsat
. Class A output stages as shown
in figure 3.8, are simple to design and have low distortion.
The gain of output stage is given by
( )
2 1 1 2
||
O O m V
r r g A ~ (3.24)
The current in the output stage is atleast three times to that of input differential stage for
proper stability and bandwidth.
The output stage transconductance is largely determined by transistor sizing since biasing
variations are limited by headroom at low supply voltages.
The two-stage design was supposed to be good when its first stage gain is much larger
than that of second stage. That was noise could be approximated by the first stage alone
since the gain of the second stage should be negligible. The major setback with that trade-
off is that by decreasing the gain ratio of the first stage to second stage is that the noise
41
from the first stage will be amplified significantly. As a result, due to the large
amplification of the telescopic noise, the measured dynamic range is now reduced. This
assumption is further proved by measuring the output noise of the first stage as compared
to the output noise of the second stage.
3.7 Schematic of Two Stage Fully Differential Op-Amp
The complete schematic of the fully differential telescopic operational amplifier which
meets the target specifications is shown in figure 3.9.
Figure 3.8 Schematic of two stage fully differential op-amp
The aspect ratio of each transistor in circuit design with their function in circuit is given
in Table 3.3. The values of capacitances and bias voltages of circuit are given in Table 3.4
and Table 3.5 respectively.
42
Table 3.3 Aspect ratio of transistors and their functions in op-amp
Transistors Aspect ratio Function Region of operation
MN1& MN2 18.9m/1.4m Input drivers Saturation
MN3 & MN4 14m/1.4m Cascode load Saturation
MN5 694.75/1.4m Bias current source Saturation
MN6 & MN7 7.7m/1.4m Common mode
feedback
Linear
MN8 & MN9 5.95m/1.4m Second stage driver Saturation
MP1 & MP2 56m/1.4m Cascode load Saturation
MP3 & MP4 112m/1.4m Cascode load Saturation
MP5 & MP6 287m/1.4m Second stage load Saturation
MP7 & MP8 210 m/13.3 m Nulling resistor Linear
Table 3.4 Capacitance value
Capacitors Capacitors value function
C1 & C2 5 pf Load capacitances
C3 & C4 5 pf Miller compensation
capacitances
C5 & C6 0.05 pf Feed forward capacitances
Table 3.5 Biasing voltages
Voltage Port Voltage value(Volt)
V
incm
1.65
Vb0 0.7
Vb1 2.25
Vb2 2.05
Vb3 , Vb4 & Vb5 2.45
43
The primary trade off for this circuit is given in Table 3.6.
Table 3.6 Primary design trades off for our topology.
Parameter to improve Method to improve Potential tradeoffs/draw
backs
DC gain Increase L, maintain W/L Feedback factor degraded
Increase g
m
, same L same
I
ds
Reduce V
dsat
Reduce I
ds
at same g
m
Slew rate degrades
Dynamic range Increase C
C
Worse slewing, much more
current, lower UGB
Increase C
L
Reduce phase marin
Increase g
m1
Reduce V
dsat
Decrease g
m3
Increase first stage swing
Unity gain bandwidth Reduced C
C
More noise, lower dynamic
range
Increase C
C
Reduced UGB
3.8 Layout of Op-Amp
Analog layout demand many more layout precautions so as to minimize effects such as
crosstalk, mismatches, noise etc. The main problems of layout for analog ICs are device
matching and unwanted parasitic reduction. Sometimes chip area may also be a concern.
For matching, the techniques are common-centroid and interdigital, which can be used
separately or in combination, as well as dummy devices and device unitization. If there
are no parasitic, then there is no noise coupling. The parasitics reduction techniques
include, shield, guard ring, long distance between noisy digital circuitry etc.
3.8.1 Issues in analog layout
1. Matching of devices
Device mismatch is too often treated as part of the black art of analog design. Random
device mismatch plays an important role in the design of accurate analog circuits. The
device mismatch is due to number factors like local process variation, global lithographic
44
variations, local lithographic variations and process gradients. These factors affect all
devices transistors, resistor, capacitors, and therefore similar techniques can be used to
match all elements.
Matching improves with increasing device area. And accuracy requirements impose a
minimal device area. Use of transistor fingering for large and critical transistors is always
beneficial. In fingering the transistor is fingered into multiple transistors that are
connected in parallel. The folded transistors reduce the source/ Drain junction area and
the gate resistance. The gate resistance can be reduced by decomposing the transistor into
more parallel fingers. There is an advantage with an even number of fingers; the active
capacitance is less, because the drain region is surrounded with gate poly instead of field.
The large transistors are breaks in number of small transistors in parallel. This also
decreases the physical size of the device, and thus can provide a more compact layout.
Minimum size will usually provide a minimum in terms of parasitic capacitance, but still,
this may not provide the optimal device for a given application. For example, an
application that is more sensitive to gate resistance may warrant a non-optimal device
size, in terms of parasitic capacitance. Lower gate resistance is obtained by smaller finger
width. This increases the overall number of fingers and therefore increases the junction
capacitances C
sb
and C
db
.
Figure 3.9 Interdigitated MOSFETs [27]
45
Since MOSFETs allow device folding, many devices with the same finger width can be,
interdigitated with one another. The only requirement for interdigitating devices is that
these devices share source terminals, drain terminals, or source and drain terminals. These
compact, interdigitated devices provide better matching for both process and thermal
gradients. When matched signals or currents are needed, precision device matching is
required. For example, current mirrors and differential pairs. In many cases, these
structures can provide symmetry along two separate axes, completely alleviating the
affects of linear gradients as shown in figure 3.9.
An interdigitation pattern that will minimize source/drain parasitics by placing two same
devices gates back-to-back.
It should also be pointed out that, although the compound device is repetitive, the end
devices do not match the inner devices and therefore device mismatch can exist due to
processing gradients. In order to alleviate this problem, dummy devices should be added
to the structure.
.
Figure 3.10 Basic structure of common centroid layout[27]
46
While on the subject of matching, it should also be mentioned that metal routing over
transistors might affect device matching. Such routing can cause threshold voltage
mismatch as well as transconductance mismatch. Therefore, it is recommended that
routing over transistors be avoided. If such routing cannot be avoided, the routing over
symmetric devices should be made symmetric as well, and if possible, higher metal
layers should be used.
In order to consistently provide symmetry along two axes, a layout technique called a,
common-centroid can be used. Figure 3.10 shows how such a layout provides two axes
of symmetry. Typically, this form of layout is used for precise matching of only two
devices. The method can be extended to more devices; however, the routing for such a
compound device becomes very cumbersome some golden rules for device matching are
- Use of the same W and L and vary M (takes out AW and AL effects, M is
number of fingers)
- Use Ms that are even.
- Use Common-Centroid, or nearly Common-Centroid, layout.
- Use dummy transistors at the ends of the row (takes out poly etch loading and
mask misalignment effects)
- Use plenty of substrate and well taps.
- Route currents a long way, not voltages - IR drops can cause big mismatches.
2. Noise: Noise is important in all analog circuits because it limits dynamic range. In
general there are two types of noise, random noise and environmental noise. Random
noise refers to noise generated by resistors and active devices in an integrated circuit;
environmental noise refers to unwanted signals that are generated by humans. Two
common examples of environmental noise are switching of digital circuits and 60 Hz
'hum'. In general, random noise is dealt with at the circuit design level. However they are
some layout techniques which can help to reduce random noise. Multi-gate finger layout
reduces the gate resistance of the poly-silicon and the neutral body region, which are both
random noise sources. Generous use of Substrate plugs will help to reduce the resistance
of the neutral body region, and thus will minimize the noise contributed by this resistance.
Environmental noise is also dealt with at the circuit level. One common design
technique used to minimize the effects of environmental noise is to employ a 'fully-
differential' circuit design, since environmental noise generally appears as a common-
mode signal. However Substrate plugs is also very useful for reducing substrate noise,
47
which is a particularly troublesome form of environmental noise encountered in highly
integrated mixed-signal systems and Systems-On-a-Chip (SOC). Substrate noise occurs
when large amount digital circuits are present on a chip. The switching of a large number
of circuits discharges large dynamic currents to the substrate, which cause the substrate
voltage to 'bounce'. The modulation of the substrate voltage can then couple into analog
circuits via the body effect or parasitic capacitances. Substrate plugs minimizes substrate
noise because it provides a low impedance path to ground for the noise current.
Issues that are important in digital circuits are still important in analog layout. Foremost
among these is parasitic aware layout. It is important to minimize series resistance in
digital circuits because it slows switching speed. Series resistance also slows analog
circuits, plus it introduces unwanted noise. Parasitic capacitance is avoided in digital
circuits because it slows switching speed and/or increases dynamic power dissipation.
Stray capacitance has the same effect in analog circuits (bias current must be increased to
maintain bandwidth and/or slew rate when extra load capacitance is present) plus it can
lead to instability in high gain feedback systems.
3.9 Complete layout of Op-Amp
From figure 3.9 and table 3.2, the transistors are divided into fingers as given in table 3.7.
Table 3.7 Transistors with aspect ratio and number of finger
Transistor Aspect ratio
(in schematic)
Number of finger Aspect ratio
(of each finger)
M1 & M2 18.9m/1.4m 6 3.2m/1.4m
M3 & M4 14m/1.4m 4 3.5m/1.4m
MN6 & MN7 7.7m/1.4m 2 3.9m/1.4m
MN8 & MN9 5.95m/1.4m 2 3m/1.4m
MP1 & MP2 56m/1.4m 8 7m/1.4m
MP3 & MP4 112m/1.4m 8 14m/1.4m
MP5 & MP6 287m/1.4m 10 28.7m/1.4 m
MP7 & MP8 210m/13.3m 6 35m/13.3 m
Since MN5 have aspect ratio of 694.75m /1.4m is very big transistor, the number of
calculated finger is 70, which is too large to make on single level and also consume large
layout area and make routing complex. Thats why we use here stacking of transistor i.e
one above the other. For number of fingers=70, we use 5 stacks of 10 fingers and each
48
finger of aspect ratio of 14m/1.4m. As, 694.75m/1.4m 5stack * 10 finger *
14m/1.4 m. Stack layout is shown in figure 3.11.
Figure 3.11 Stack layout design of MN5
The complete layout of the op-amp design is shown in figure 3.12
Figure 3.12 Complete layout of fully differential op-amp
49
LVS Report
This design is DRC clean, and layout of the circuit is matched with schematic, both DRC
and LVS are verified by the Calibre.
Figure 3.13 shows there is LVS matching.

Figure 3.13 Complete LVS Report
50
Figure 3.14 LVS netlist report
51
CHAPTER 4
SIMULATION RESULTS
In this chapter the schematic of the circuit has been tested for various parameters of
operational amplifier. The amplifier is powered by supply voltage of 3.3V. The fully
differential CMOS Op-Amp has been designed and simulated on tsmc 0.35m
technology.
This chapter dived in four sections:
- Schematic simulations
- Post layout simulations
- Process corner simulations
- Monte Carlo simulations
4.1 Schematic Simulations
4.1.1 AC Response
Figure. 4.1: Test setup for AC response of the op-amp
In figure 4.2, a Bode and phase plot for 3.3V, 27C and C
L
= 5pf is shown. As can be
seen, the open loop gain is 86.0217dB, and a phase margin is 50.018. The unity gain
bandwidth is 270.455 MHz and f
-3db
bandwidth is 6.105 KHz.
52
Figure 4.2(a) Frequency response plot with C
L
=5pf
Figure 4.2(b) Frequency response plot with C
L
=1pf
53
Figure 4.2(c) Frequency response plot with C
L
=3pf
Table 4.1 shows the effect of load variation on AC response.
Table 4.1 AC result due to load capacitance variation
Capacitor (pF) DC gain(dB) UGB(MHz) f
-3dB
(KHz) Phase margin
1 86.02 284.013 6.09 48.09
3 86.02 277.21 6.098 49.07
5 86.02 270.455 6.105 50.01
Figure 4.3 Frequency responses with temperature variation
54
The temperature has been varied from -20deg to +200deg centegrade. Increasing
temperatures tend to reduce the gain of the circuit. The increasing temperature also
reduces the UGB of the opamp.
4.1.2 Common Mode Rejection Ratio
In order to know the CMRR of the opamp we need to know the differential gain as well
as the common mode gain of the opamp and the CMRR is obtained by A
d
(dB)-A
c
(dB).
The resulting CMRR response is shown in the figure 4.5.
Figure 4.4 Test setup for Common Mode response
Figure 4.5 CMRR at 27C with C
L
=5pf
55
Figure 4.6 CMRR with temperature variation
Figure 4.7 CMRR with load capacitance variation
56
Table 4.2 CMRR with load capacitance variation
Capacitor(pf) CMRR(dB)
1 90.793
3 90.793
5 90.793
The CMRR with load variation is given in Table 4.2. It concludes that CMRR is
independent of load variations.
4.1.3 Power Supply Rejection Ratio
PSRR was measured by placing a 1V AC signal on the power supply. PSRR is equal to
the ratio of the AC signal at the output node to the AC signal on V
DD
.
Figure 4.8 Test setup for PSRR
Figure 4.9 PSRR of op-amp
57
Figure 4.10 PSRR with temperature variation
The PSRR of the fully differential op-amp is 132.725 dB. The effect of temperature on
PSRR is given in figure 4.10. It shows that there is a very little variation in PSRR due to
variation in temperature. As temperature increases PSRR decrease.
4.1.4. Input Common Mode Range (ICMR)
This test is per formed to test the offset voltage and the input common mode range of the
opamp that is the range of opamp for which there is a linear relationship between input
and the output.

Figure 4.11 Test setup for ICMR
58
Figure 4.12 ICMR of op-amp
Figure 4.13 ICMR with temperature variation
From ICMR we also find the offset voltage i.e the output voltage of op-amp when input is
zero. In our case offset is 0.17V.
The output voltage varies linearly with input voltage for a voltage range of 0.94V to 2.2V
so this is called the ICMR range of the op amp.
ICMR with temperature variation is shown in figure 4.13. With rise of temperature ICMR
reduces.
59
4.1.5 Transient Response
For observing the transient sinusoidal response of the op-amp the test setup of figure 4.14
is used. In this setup sinusoidal signals are applied to the two inputs and the effective
value of input signal is the difference of the voltage at the two terminals. DC potential is
also applied along with the sinusoidal signal in order to provide the bias voltage to the
input transistors. The output is taken differentially from the two complementary outputs.

Figure 4.14 Schematic for the sinusoidal transient response
Figure 4.15 Sinusoidal transient differential outputs
60
From figure 4.15, it is measure that the complete voltage swing (peak-to-peak) is
3.1898V.
Figure 4.16 Sinusoidal transient differential outputs with temperature variations
4.1.6 Transient Step Response
In Figure 4.17, a step from ground to V
DD
is applied at the input with unity feedback
configuration.
Figure 4.17 Schematic for the simulation and measurement of the slew rate
The slew rate of op-amp is 40.585V/S for rising edge of pulse and 19.406V/ S for
falling edge of the pulse.
61
Figure 4.18 Transient pulse response of op-amp for slew rate measurement
Figure 4.19 Effect of temperature on slew rate
62
Table 4.3 Show the effect of temperature on slew rate
Temperature (C) Slew rate(V/S) (+ve)
-20 69.683
10 53.348
27 40.585
40 45.496
100 26.240
4.1.7 Settling Time
Settling time is length of the time in which the response of the step input get settle within
certain tolerance of its final value.
Figure 4.20 Settle time at various tolerances level of differential output
63
Table 4.4 Settling time of differential outputs
Tolerance % Vout1 Vout2
1% 2.185 Ns 2.836 Ns
2% 2.0437 Ns 2.7139 Ns
5% 1.7565 Ns 2.3474 Ns
Table 4.3 shows the settling time of differential outputs at various tolerance levels.
4.1.8 Output Dynamic Range of Op-Amp
The dynamic range is a measure of the signal to noise ratio. It requires high swing and
low noise.
First we determined the maximum total output noise we could tolerate for a given output
swing. This translates directly into determining the size of the compensation capacitor and
feedback capacitor to give the desired dynamic range. We initially assumed that we
would have a single-ended output swing of 1.65V (VDD -2V
DSAT
1.65V), which
corresponds to a differential swing of 3.3V peak-to-peak.
The maximum voltage swing essentially voltage rail minus the minimum voltage drop
across series FETs.
The output dynamic range is specified as
noise
signal Peak
P
P
DR

= log 10 (4.1)
Thus,
2
2
,
on
swing o
noise
signal Peak
V
V
p
P
=

(4.2)
where
signal Peak
P

is signal power (peak-peak) and
noise
P is output noise power.
The noise output voltages of op-amp at different frequencies are given in figure 4.21 and
the output swing at corresponding frequencies are given in figure 4.22.
64
Figure 4.21 Output noise voltages
65
Figure 4.22 Output swing at differential output at various frequencies
The dynamic range with frequency variation is shown in Table 4.4. It concluded that as
the frequency increases output dynamic range also increase for fully differential
operational amplifier
66
Table 4.5 Dynamic Range of op-amp at various frequencies
Frequency Output noise
voltage
Output swing Output dynamic
range
1Hz 237.4157V 3.43V 83.19 dB
1KHz 234.23027V 3.4098V 83.26 dB
10KHz 119.0101V 2.1526V 85.15 db
.
At the end of this section, all simulated results of opamp characterization are tabulated in
Table 4.5 with target specification and specification of the opamp designed by
conventional way.
Table 4.6 Simulation Results of fully differential Op-Amp
Target specification Simulation results
DC gain (dB) 80 86.02
Unity gain bandwidth(MHz) 300 270.45
Phase margin 50 50.01
f
-3dB
frequency(KHz) 5 6.105
Output dynamic range (dB) 85 85.15
CMRR (dB) 85 90.795
PSRR (dB) - 131.725
ICMR (V) - 0.9 to 2.2
Output voltage swing(p-p)(V) 3V 3.1898
Settling time (nS) 30 2.836
Slew rate (V/S) 20 40.585
Power dissipation Minimize ( 5mW) 1.4446 m W
Load capacitance (pf) 5 5
Supply voltage (V) 3.3 3.3
67
4.2 Post Layout Simulations
After completing the layout of the opamp, it is matched with schematic using LVS
simulation. With the successful run of LVS, parasitic and layout netlist extraction
simulation have been done using PEX. Post layout simulations have been done on
extracted netlist. The simulation results of post layout simulation are given below.
AC Response
Figure 4.23 Post layout gain-phase plot
Figure 4.24 Post layout gain-phase plot with temperature variation
68
CMRR
Figure 4.25 CMRR of post layout
Results of post layout simulation are in table 4.7.
Table 4.7 Post layout simulation results
Specifications Schematic results Post layout results
DC gain(dB) 86.02 76.309
f
-3dB
frequency (KHz) 6.105 12.73
CMRR (dB) 90.795 79.655
4.3 Process Corner Simulation
In this section, Process corner simulation has been done on extracted netlist from layout.
The netlist extracted from layout also includes two more netlist, one is distributed RC
network netlist, which contains distributed RC network formed by considering resistances
and parasitic capacitances of different devices and routing in layout. Another netlist
included is the netlist of coupling capacitors in adjacent layers. These parasitic effects can
make a big problem if any process parameter varies during fabrication process.
69
Thus it is necessary to check the circuit performance at every expected corner of the
process variation. The simulation done considering all probabilities of process parameter
variation is called as process corner simulation.
Here in this simulation process parameter like oxide thickness, mobility and electrical
parameter threshold voltage are considered with variations of 10% in each
SS (Slow-Slow)
1) Transient analysis of process corner SS (slow-slow) result shown below;
Figure 4.26 Settling time for SS corner
Figure 4.27 Slew rate for SS corner
70
2) AC Analysis for SS
Figure 4.28 Gain- phase plots for SS
SF (Slow-Fast)
1) Transient Analysis
Figure 4.29 Settling time for SF corner
71
Figure 4.30 Slew rate for SF Corner
2) Ac Analysis for SF
Figure 4.31 Gain-phase plots for SF corner
72
FS (Fast-Slow)
1) Transient Analysis
Figure 4.32 Settling time plots for FS corner
Figure 4.33 Slew rate for FS corner
73
2) Ac Analysis for FS
Figure 4.34 Gain plot for FS corner
FF (Fast-Fast)
1) Transient Analysis
Figure 4.35 Slew rate for FF corner
74

Figure 4.36 Settling time for FF corner
2) AC Analysis

Figure 4.37 Gain Phase plots for FF corner
75
Results of process corner simulations are given in table 4.8.
Table 4.8 Results of process corner simulations
Specifications SS SF FS FF
DC gain (dB) 58.104 49.5 59.55 79.18
Settling time(1% tolerance)(nS) 382.28 58 14.009 424.55
Slew rate (V/S) 20.667 35.193 15.964 50.483
4.4 Monte Carlo Simulations
The Monte Carlo technique provides the greatest flexibility for studying the results of
process variations. All process parameters can be varied either simultaneously or
individually. It is the only technique in which a distribution is obtained, not just a trend or
a small set of individual points. The resultant distribution is more useful than worst-case
analysis since the probability of an occurrence as well as its value is available. This
allows process and device engineers to make realistic tradeoffs in setting process
tolerances.
The Monte Carlo technique is a method of solving problems by simulating original data
with a random number generator. There are two basic requirements for doing a Monte
Carlo analysis. A reasonable model for the parameters of interest must exist and a
mechanism to simulate the model is needed.
The Monte Carlo simulation results are shown in figure 4.38, for 10 run per individual
simulation.
Figure 4.38 (a) Gain phase plots with temperature variation
76
Figure 4.38 (b) Gain phase plots with VDD variation
Figure 4.38 (c) Output swing
77
(d) Settling time
Figure 4.8 Monte Carlo simulation results
Table 4.9 Results of Monte Carlo simulation
Specification Schematic result Monte Carlo result
DC gain (dB) 86.02 85.315
f
-3dB
frequency(KHz) 6.105 5.88
Unity gain bandwidth(MHz) 270.45 288
Settling time(nS) 2.836 5.5149
Output swing (p-p)(V) 3.1898 3.0619
Power dissipation 1.4446mW 2.654mW
Table 4.9 show the Monte Carlo results of the circuit design. It shows that the results of
Monte carlo simulation have quite variation from that of schematic simulation results.
78
CHAPTER 5
CONCLUSION AND FUTURE PROSPECTS
5.1 Conclusion
The fully differential OTA is designed in this thesis work is based on telescopic cascode
topology is chosen over its folded cascode counter parts due its less consumptions of
power, which is in this design is 1.44mW much less than typical fully differential
amplifier. The OTA provide DC gain of 86.02 dB, unity gain of 270 MHz with f
-3dB
of 6.1
KHz. The low value of settling time of 2.836Ns shows the speed of OTA is quite good at
nominal process. The high unity gain bandwidth is achieved by using nulling resistor
capacitor compensation technique with feed forward capacitors at both ends.
There is some variation in performance of op-amp is found at two process corners (SF &
FS) of design.
Monte carlo simulation gives result much nearer to that obtained in schematic simulation
in nominal conditions.
5.2 Future Scope
Applications such as ADC requires high gain as well as large bandwidth and noise
rejection capability is found in this design. In addition of all mention parameters, there is
scope to improve phase margins for good transient operation. Also there is improvement
is require in power dissipation which can we further minimize (<1mW) by using power
reduction techniques.
This topology may use subthereshold current for low power & low voltage operation with
gain boosting techniques.
79
REFRENCES
1. Phillip E. Allen, Douglas R. Holberg, CMOS Analog Circuit Design, Textbook,
Oxford University Press, Second Edition 2007.
2. Behzad Razavi, Design of Analog CMOS Integrated Circuits, Tata McGraw-
Hill Edition 2002.
3. Shailesh B.Nerurkar and Khalid H. Abed, CMOS Fully Differential Operational
Transconductance Amplifier Delta-Sigma Modulators, IEEE Journals of Solid
State Circuits, vol.1, pp 1884-1888, 2008.
4. J.Yuan, N.Farhat, J.V.Spiegel, GBODPCAD: A Synthesis Tool For High-
Performance Gain Boosted Opamp Design, IEEE Transactions on Circuits and
Systems, vol. 52, No.8, 2005.
5. Jirayuth Mahattanakul, Design Procedure for Two Stage CMOS Opamp with
Flexible Nosie Power Balancing Scheme. IEEE Transcations on Circuits and
Systems-I, vol. 52, No. 8, 2005.
6. Allen Y Chen and Pavel Monat, A Fully Differential Two Stage CMOS
Operational Amplifier, Term Paper, Spring 2003.
7. Gray, Hurst, Lewis & Mayer, Analysis And Design Analog Integrated Circuits,
John Wiley & Sons, Fourth edition.
8. A. Chang and D. Fang, A Low Power, Two Stage Telescopic Amplifier with 80
dB Dynamic Range And 10 ns Settling Time, Design project, Spring 2007.
9. P. M VanPeteghem and J.F. Duque-Carrillo, A General Description of Common-
Mode Feedback In Fully Differential Amplifiers, 1990 IEEE International
Symposium on Circuits and Systems, Vol. 4, pp.3209-3212, 1990.
10. J. Mahattanakul, Design Procedure for Two-Stage CMOS Operational
Amplifiers Emploting Current Buffer, IEEE International Transactions on
Circuits and Systems-II, vol. 52, No. 11, 2005.
11. D. Miyazaki, A 10 b 30-MS/s Low Power Pipelined CMOS A/D Converter
Using Pseudo differential Architecture, IEEE Journals of Solid State Circuits,
vol.38, No.2, 2003.
12. T. Lehmann and M. Cassia, 1-V Power Supply CMOS Cascode Amplifier,
IEEE Journals of Solid State Circuits, vol. 36, No. 7, 2001.
13. Wen-Whe Sue, Zhi-Ming Lin, and Chou-Hai Huang, A High DC-Gain Folded-
Cascode CMOS Operational Amplifier, IEEE Journals of Solid State Circuits,
vol 78, No. 3, pp 4391-4394, 1998.
80
14. Sung-Hyun Yang, Kyu-Ho Kim, Yong-Hwan Kim, Younggap You, and Kyoung-
Rok Cho, A Novel CMOS Operational Transconductance Amplifier Based on a
Mobility Compensation Technique, IEEE International Transactions on Circuits
and Systems-II, vol. 52, No. 1, 2005.
15. Richard E. Valle and Ezz I. El-Masry, A Very High-Frequency CMOS
Complementary Folded Cascode Amplifier, IEEE Journals of Solid State
Circuits, vol. 29, No. 2, 1994.
16. Katsufumi Nakamura and L. Richard Carley, An Enhanced Fully Differential
Folded-Cascode OP Amp, IEEE Journals of Solid State Circuits, vol. 27, No. 4,
pp 563-568, 1992.
17. Johns, D.A.; Martin K., Analog Integrated Circuit Design, Textbook, John
Wiley & Sons, First Edition, 1997.
18. A. R. Mortazavi, M. R. Hassanzadeh, J. Talebzadeh, and O. Shoaei, Design
Procedure for a High C-Gain and High- Bandwidth Amplifier, Master of Science
Thesis, University of Tehran, Iran, 2001.
19. A. J. Gano and J. E. Franca, Fully Differential Variable Gain Instrumention
Amplifier Based on A Fully Differential DDA Topology, IEEE Journals of Solid
State Circuits, vol. 78, No. 3, pp 5682- 5689, 1998.
20. M. Walatri, Circuit Techniques for Low-Voltage and High-Speed A/D
Converters, Dissertation for the degree of Doctor of Science in Technology,
Helsinki University of Technology, 2002.
21. Allen Y Chen & Pavel Monat, A Fully Differential Two-Stage CMOS
Operational Amplifier, Term project, University of California, Berkely 2003.
22. T. S. Lee, Chi Chang Lu, Shen Hau Yu and Jiang-Tiang Zhan, A Very High
Speed Low Power Low Voltage Fully Differential CMOS Sample and Hold
Circuit With Low Hold Pedestal, IEEE Journals of Solid State Circuits, vol. 78,
No. 3, pp 8834-8838, 2005.
23. Dony Mota, Fundamentals of Fully Differential Op-Amp and CMFB Circuit
Design, Term Paper, University of Toronto, 2001.
24. Lisha Li, High Gain Low Power Operational Amplifier Design and
Compensation techniques, Term Paper Brigham Young University, 2007.
.
81
25. Bhupendra K. Ahuja , An Improved Frequncy Compensation Technique for
CMOS Operational Amplifiers, IEEE Journal of Solid State Circuits, vol. SC-18,
No.6, December 1983.
26. Allan Hastings Art of Analog Layout,Second Edition, Pearson Education Asia,
2008.
27. Patrick T. McElwee, An Automated Analog Layout Generation Flow, Term
Paper, University of California, Berkely, 2005.
28. F. Maloberti, Analog Design for CMOS VLSI System Integrated Circuits, First
edition, Springer 2007.

You might also like