Cadence Op-Amp Schematic Design Tutorial For TSMC CMOSP35

Cadence Op-Amp Schematic Design Tutorial for
TSMC CMOSP35
Till Kuendiger, Joseph Schrey, Iman Taha, Yi Lin,
Tao Dai, Li Liang, Song-Tao Huang, Yue Huang
December 7, 2001
Contents
Preface iv
1 Introduction 1
1.1 Review of CMOS FETs . . . . . . . . . . . . . . . . . . . . . 1
1.2 Creating a New Library in Cadence . . . . . . . . . . . . . . . 2
1.3 Schematic Capture . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3.1 Virtuoso Schematic Editor . . . . . . . . . . . . . . . . 4
1.3.2 Virtuoso Symbol Editor . . . . . . . . . . . . . . . . . 5
1.3.3 Arma Analog Circuit Design Environment . . . . . . 6
1.3.4 The Waveform Window . . . . . . . . . . . . . . . . . 7
1.3.5 The Cadence Calculator . . . . . . . . . . . . . . . . . 8
1.4 Generating the Characteristic MOSFET Curves . . . . . . . . 9
1.4.1 N-channel Enhancement-Type MOSFET . . . . . . . . 9
1.4.2 P-channel Enhancement-Type MOSFET . . . . . . . . 14
2 An Introduction to Op-Amps 17
2.1 Parameters of an Op-Amp . . . . . . . . . . . . . . . . . . . . 17
2.1.1 Oset Voltage . . . . . . . . . . . . . . . . . . . . . . . 17
2.1.2 Input Current . . . . . . . . . . . . . . . . . . . . . . . 18
2.1.3 Input Common Mode Voltage Range . . . . . . . . . . 19
2.1.4 Maximum Output Voltage Swing . . . . . . . . . . . . 19
2.1.5 Output Impedance . . . . . . . . . . . . . . . . . . . . 20
2.1.6 Common-Mode Rejection Ratio . . . . . . . . . . . . . 20
2.1.7 Supply Voltage Rejection Ratio . . . . . . . . . . . . . 21
2.1.8 Slew Rate . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.1.9 Unity Gain Bandwidth and Phase Margin . . . . . . . 22
2.1.10 Settling Time . . . . . . . . . . . . . . . . . . . . . . . 24
2.2 Methodology of Choosing Op-Amp Parameters . . . . . . . . 24
2.3 How to Adjust the Parameters . . . . . . . . . . . . . . . . . 25
i
CONTENTS ii
2.3.1 Specication . . . . . . . . . . . . . . . . . . . . . . . 25
2.3.2 Procedure of Optimization . . . . . . . . . . . . . . . 25
2.3.3 Optimize the Parameters of the Op-Amp . . . . . . . 27
2.3.4 How to get the Quiescent point in a complex circuit . 33
2.4 Target Op-Amp Specications . . . . . . . . . . . . . . . . . . 34
3 Current Mirrors and Biasing Networks 35
3.1 Ideal Characteristics of a Current Mirror . . . . . . . . . . . . 36
3.2 Basic Current Mirror Derivation . . . . . . . . . . . . . . . . 36
3.3 Benchmark Test Circuit . . . . . . . . . . . . . . . . . . . . . 38
3.4 Examined Current Mirrors . . . . . . . . . . . . . . . . . . . . 39
3.4.1 Basic Current Mirror . . . . . . . . . . . . . . . . . . . 39
3.4.2 Cascade/Cascode Current Mirror . . . . . . . . . . . . 41
3.4.3 Wilson Current Mirror . . . . . . . . . . . . . . . . . . 43
3.4.4 Modied Wilson Current Mirror . . . . . . . . . . . . 44
3.4.5 Reduced Cascade Current Mirror . . . . . . . . . . . . 45
3.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4 Dierential Input Stage 48
4.1 The Unbuered Op-Amp . . . . . . . . . . . . . . . . . . . . 48
4.2 Small Signal Equivalent Circuits . . . . . . . . . . . . . . . . 49
4.3 The Frequency Response . . . . . . . . . . . . . . . . . . . . . 53
4.4 Phase Margin . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.5 Compensation . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.6 Adding R
z
in series with C
c
. . . . . . . . . . . . . . . . . . . 60
4.7 Gain Bandwidth Product . . . . . . . . . . . . . . . . . . . . 61
4.8 Large Signal Consideration . . . . . . . . . . . . . . . . . . . 62
4.9 Slew Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.10 The Common-Mode Range . . . . . . . . . . . . . . . . . . . 64
4.11 Important Relationships for The Design . . . . . . . . . . . . 66
4.12 Tradeos for Increasing the Gain of the Two Stage Op-Amp. 66
4.13 Design Methodology for the Two Stage Op-Amp . . . . . . . 67
4.14 Design Example . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.15 Limitations of the Two Stage Op-Amp . . . . . . . . . . . . . 72
4.16 The Cascode Op-Amp . . . . . . . . . . . . . . . . . . . . . . 72
5 Inverting Ampliers 76
5.1 Inverter with Active Resistor Load . . . . . . . . . . . . . . . 77
5.2 Inverter with Current Source/Sink Load . . . . . . . . . . . . 83
5.3 Push-Pull Inverter . . . . . . . . . . . . . . . . . . . . . . . . 87
CONTENTS iii
5.4 Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.5 Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
6 Control Network and Output Stage 95
6.1 Classication of Output Stage . . . . . . . . . . . . . . . . . . 96
6.2 Class-A Output Stage . . . . . . . . . . . . . . . . . . . . . . 96
6.2.1 Simple output amplier using a Class-A, current-source
inverter . . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.2.2 Common-Drain (Source-Follower) Output Amplier . 98
6.2.3 Power Analysis . . . . . . . . . . . . . . . . . . . . . . 98
6.3 Class-B Output Stage . . . . . . . . . . . . . . . . . . . . . . 99
6.3.1 Push-Pull, Inverting CMOS amplier . . . . . . . . . 99
6.3.2 Power Analysis . . . . . . . . . . . . . . . . . . . . . . 101
6.4 Class-AB Output Stage . . . . . . . . . . . . . . . . . . . . . 101
6.5 Short Circuit Protection . . . . . . . . . . . . . . . . . . . . . 102
6.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
6.7 Design Considerations . . . . . . . . . . . . . . . . . . . . . . 106
6.7.1 Negative Feedback . . . . . . . . . . . . . . . . . . . . 106
6.7.2 Frequency Compensation . . . . . . . . . . . . . . . . 106
7 Integrating the Sub-Circuits 108
7.1 Overall Performance . . . . . . . . . . . . . . . . . . . . . . . 109
7.2 The Measurement of Some Main Parameters . . . . . . . . . 110
7.2.1 Input Oset Voltage . . . . . . . . . . . . . . . . . . . 110
7.2.2 Common-Mode Rejection Ratio (CMRR) . . . . . . . 112
7.2.3 Output Resistance - R
o
. . . . . . . . . . . . . . . . . 114
7.3 Delivering Power to the Load/Instantaneous Power . . . . . . 117
7.4 Improving the Output Buer . . . . . . . . . . . . . . . . . . 118
7.4.1 Stabilizing the Output . . . . . . . . . . . . . . . . . . 120
7.4.2 The Final Schematic . . . . . . . . . . . . . . . . . . . 121
8 Closing Remarks 122
8.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
8.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
9 Bibliography 124
Preface
The purpose of this document is to familiarize the reader with the Cadence
set of tools in order to do analog micro-electronic circuits. The design process
will use TSMCs CMOSP35 technology and as a result requires access to the
restricted technology les.
iv
Chapter 1
Introduction
This tutorial assumes that the user is working in the CMC supplied envi-
ronment for CMOSP35 design.
1.1 Review of CMOS FETs
The Complimentary Metal Oxide Semi-Conductor Field Eect Transistor is
a four terminal device: Base, Emitter, Collector and Substrate. Unlike the
Bipolar Transistor the MOSFET is a symmetrical device: the source and
drain can be interchanged.
The models which SPICE uses for the CMOSP35 technology are very
accurate, however, are too complex to be used by humans. A quick overview
of the device operation follows, for a more complete discussion please refer
to an appropriate text book. The gate of the device is insulated from the
rest of the device, meaning that no current will ow into the gate of a
MOSFET. The substrate connection, for an N-channel transistor, is always
connected to V
SS
(which usually means ground). When the voltage on the
gate of the device is large enough an inversion layer forms under the gate,
between the drain and source. This means that a channel of charge carriers
exists between the two N-type regions. For an N-channel transistor these
charge carriers will be electrons. This allows current to ow from the source
terminal to the drain terminal. Some generalized relationships are stated
below.
1
CHAPTER 1. INTRODUCTION 2
I
ds
=
eff
C
ox
W
L
__
V
g
V
fb
2
B
V
ds
2
_
V
ds
_
eff
C
ox
W
L
_
2
2
si
qN
a
3C
ox
_
(2
B
+V
ds
)
3
2
(2
B
)
3
2
_
_
(1.1)
1.2 Creating a New Library in Cadence
Once Cadence has been started the icfb window is shown. This is the main
window for Cadence; all messages, including error and warning messages,
will be displayed in this window.
Figure 1.1: Cadence icfb Window with CMOSP35 technology
In order to create a new library select the following menu options:
Tools Library Manager
From the Library Manager window it is possible to access all available
libraries and to create new libraries. In order to create a new library:
File New Library
This will pop-up a new window where the library name can be specied.
In this example we will use the name mylib.
Once you have entered the name of the library which is to be created,
press OK. The next dialog will ask information about the technology le
which is to be associated with the new library. For this tutorial we will be
using CMOSP35, therefore, select Attach to existing techle: this will bring
another dialog window which lets us select which technology le we will use.
Select cmosp35.
Figure 1.2: Cadence Library Manager
Figure 1.3: New
Library Dialog
Window
Figure 1.4: Technol-
ogy File For New Library
Window
Figure 1.5: Attach Design Library Window
1.3 Schematic Capture
In this section we will be showing step by step how to enter a circuit
schematic into Cadences Virtuoso Schematic Editor and how to create a
symbol view for the schematic using the Virtuoso Symbol Editor. This sec-
tion will not give a circuit directly, but rather leave the reader to use some
of the schematics presented later in the document (Section 1.4 is a good
starting circuit for a novice user).
In order to create a new Cell View, open the Library Manager and click
on the new library which has been created, followed by:
File New Cell View
A dialog window will appear asking for the name of the new Cell View:
enter OpAmp. This will automatically open the schematic capture session.
1.3.1 Virtuoso Schematic Editor
Figure 1.6) is the main window from which all of the schematic capture is
performed. Along the left boarder of the window are icon-buttons which
allow for easy access to the more common commands. Moving the mouse
cursor over these windows will show a tool-tip which explains which com-
mand is executed with each button. All commands are also available via the
menus at the top of the window. Most commands will also have a short-cut
key associated with them.
When a schematic is entered the Insert Instance but is used (alterna-
tively Add Instance may be used) is used in order to place components
into the schematic. We will mostly be using transistors, resistors and ca-
pacitors (which are found in the library cmosp35) and power supplies and
grounds (which are found in the library analogLib). If the exact name of
the desired cell is not known the Browse button may be used to open the
Library Manager and graphically select the component. Components are
connected with narrow wires which may also be added via the icon-buttons
or the menus.
When an instance of a component is added to a schematic all of the
available parameters to the model may be set. These parameters may be
changed later using the Properties option. Some parameters are mandatory
to be entered (e.g. power supply voltage) whereas some parameters will
default to certain values if they are not entered. Values may also be set to
variables, by entering a string instead of a numeric value, which can be set
Figure 1.6: Cadence Virtuoso Schematic Editor
in the simulation stage.
Once a design has been entered, it can be saved with a Check and Save
icon-button. This will do a general check of the circuit in order to make sure
that all circuits have and ground and that all device terminals are connected
to something.
An example of schematics for generating the characteristic curves is
shown in Section 1.4. We will not discuss all of the options available for
this window since they are very numerous and are mostly self-explanatory.
1.3.2 Virtuoso Symbol Editor
Once the schematic for the circuit is done, a symbol view must be created.
This is the view which is used when an instance of the circuit is put into
another schematic (e.g. a test bench circuit). Create a default symbol by
clicking:
Design Create Cell View From Cell View
Simply except the defaults and this opens the Symbol Editor. On start-
up the symbol editor will have a plain looking rectangle with terminal pins
for each I/O pin inserted in the schematic. Once again we will simply accept
the defaults and
In order to use the circuit which has been created is it necessary to
create a test bench circuit. This test bench circuit is created similarly to
the original circuit, except that no symbol is generally required. The symbol
of the newly created circuit may be inserted as any other sub-circuit.
1.3.3 Arma Analog Circuit Design Environment
Once a test bench circuit has been created and saved it is possible to start
the simulation environment:
Tools Analog Environment
The Arma window, gure 1.7, is the window from which all simulations
are congured and executed.
Figure 1.7: Arma Analog Environment Simulation Window
In order to run a simulation there are several things which must be
dened:
Selecting a simulator. The Avant Star-HSPICE simulator is the best
simulator available; in order to select the simulator: Setup Simu-
lator. In the dialog window which appears change the simulator se-
lection to hspiceS; click OK. It is also important to dene the en-
vironment for the simulator so that all of the correct models les
are used by the simulator: Setup Environment. The dialog box
which appears contains a eld entitled Include File. Set this eld to
/CMC/kits/cmosp35/models/hspice/icdhspice.init,click OK.
Setting variables from the schematic. If there are any parameter set-
tings which are set to variables, these variables must be copied from
the cell view to the analog environment: Variables Copy From Cell
view. All variables which appear in the cell view will now be listed in
the bottom left of the Arma window. By double-clicking the vari-
ables it is possible to change the value of the variable for the next
simulation run. All variables must be assigned values before a simula-
tion may be performed.
Selecting and conguring the simulations. There are several types of
simulations which may be performed at the same time. In order to
choose the simulation settings: Simulation Choose. Most of the
parameters available in the simulation settings are relatively straight
forward and are left to the reader to lookup.
Selecting which variables are saved/plotted from the simulation. The
last step required to run a simulation is to set which variables should
be plotted and saved. By default not all variables are saved since
this could lead to vary large amounts a data being generated by the
simulation, this becomes more important for larger designs. In order
to plot/save a set of values: Outputs To Be Plotted Select on
Schematic. This allows the user to click on all nodes (voltages) and
device terminals (currents entering/leaving) in the schematic. The
schematic should reect which nodes/terminals have been selected.
Once the selection is done press ESC.
Simulation may now be started by using the menu system or with the
icon-buttons located on the right side of the Arma window.
1.3.4 The Waveform Window
Once the simulation is complete all outputs which were selected to be plot-
ted, will be in Waveform Window, Figure 1.8.
Figure 1.8: Waveform Window
The Waveform Window allows may customizations in order to generate
desired plots; it is possible to add annotations, titles, modify plot ranges
& axis and add additional plots. Most of these options are fairly straight
forward and will not be discussed in detail, the reader is encouraged to
experiment with the various display options.
One of the more advanced features available from the window is the
calculator which is discussed in the next section.
1.3.5 The Cadence Calculator
The Cadence Calculator is an extremely powerful tool for analyzing data
generated by the simulation. Some of the many features of the Calculator
are: basic arithmetic operations on waveforms, Discrete Fourier Transforms,
Total Harmonic Distortion analysis, and many more.
Figure 1.9: Cadence Calculator
The calculator allows entry of waveform information by a variety of ways.
One way of easily accessing the data is to use the wave button and followed
by selecting one of the waveforms displayed in the waveform window. If the
simulation contains a larger number of simulated values than the browser
button may be employed to browse through all of the information stored
from the simulation.
The order in which operations are entered into the calculator may be
counter intuitive to new users: The calculator uses a Postx notation. The
table below shows some examples of the default order of operations:
a +b a, b, +
a b +c a, b, , c, +
sin a + cos b b, cos, a, sin, +
In the above table each letter represents one expression/waveform from the
simulation. The display stack options is useful for evaluating large expres-
sions. In order to learn more about the many functions available in the
calculator the reader should refer to the Cadence documentation.
1.4 Generating the Characteristic MOSFET Curves
The enhancement-type MOSFET (Metal-Oxide Semiconductor Field-Eect
Transistor) is the most widely used eld-eect transistor in the FET family,
which signicance is on par with that of the bipolar junction transistor,with
each having its own areas of application. The current-control mechanism is
based on an electric eld established by the voltage applied to the control
terminal. And the current is conducted by only type of carrier(electrons or
holes) depending on the type of FET (N channel or P channel).
1.4.1 N-channel Enhancement-Type MOSFET
The transistor is fabricated on a P-type substrate, which is a single-crystal
silicon wafer that provides physical support for the device. Two heavily
doped n-type regions, the source and the drain regions, are created in the
substrate. A thin (about 0.1um) layer of silicon dioxide(SiO2) is grown
on the surface of the substrate, covering the area between the source the
drain regions. Metal is deposited on top of the oxide layer to form the gate
electrode of the device. Metal contacts are also made to the source region,
the drain region,and the substrate, also known as the body. Thus, four
terminals are brought out: the Gate(G), the Source(S), the Drain(D), and
the Body(B).
Observe that the substrate forms PN junctions with the source and drain
regions. In normal operation these PN junctions are kept reverse-biased at
all time. Since the drain will be at a positive voltage relative to the source,
the two PN junctions can be eectively cut o by simply connecting the sub-
strate terminal to the source terminal.Here, the substrate will be considered
as having no eect on device operation, and the MOSFET will be treated
as a 3-terminal device, with the terminals being the gate(G), the source(S),
and the drain(D). We applied a voltage to the gate controls current ow
between source and drain. This current will ow in the longitudinal direc-
tion from drain to source in the region called channel region. Note that
this region has a length L and a width W, two important parameters of the
MOSFET. Typically, L is in the range 1 to 10 m, and W is in the range 2
to 500 m.
The operation with V
ds
With no bias voltage applied to the gate, two back-to-back diodes exist
in series between drain and source. They prevent current conduction from
drain to source when a voltage V
ds
is applied. In fact, the path between
drain and source has a very high resistance (of the order of 10
12
)
With a positive voltage, which exceed the threshold voltage V
t
, applied
to the gate, the transistor induced a n-channel. When applying a positive
voltage V
ds
between drain and source, as shown in the gure below,
Figure 1.10: N-Channel Test Circuit
The voltage V
ds
cause a current i
D
to ow through the induced N chan-
nel. Current is carried by free electrons travelling from source to drain. The
magnitude of i
D
depends on the density of electrons in the channel, which in
turn depends on the magnitude of V
GS
. As V
GS
exceeds V
t
, more electrons
are attracted into the channel. We may visualize the increase in charge
carriers in the channel as an increase in the channel depth. The result is a
channel of increased conductance or equivalently reduced resistance.
Let V
GS
be held constant at a value greater than V
t
(for example 2V),
and increase the V
DS
from 0 to 3.3V. As V
DS
is increased, the I
D
V
DS
curve is shown in
Eventually, when V
DS
is increased to the value that reduces the voltage
between gate and channel at the drain end to V
t
, that is :
V
GS
V
DS
= V
t
or
V
DS
= V
GS
V
t
the channel depth at the drain end decreases to almost zero, and the channel
is said to be pinched o. Increase V
DS
beyond this value has little eect
(theoretically, no eect) on the channel shape, and the current through the
channel remain constant at the value reached for V
DS
= V
GS
V
t
. The
drain current thus saturates at this value, and the MOSFET is said to
Figure 1.11: I
D
V
DS
Curve
have entered the saturation region of operation. The voltage V
DS
at which
saturation occurs is named V
DS,sat
V
DS,sat
= V
GS
V
t
Obviously, for every value of V
GS
V
t
, there is a corresponding value of
V
DS,sat
. The device operates in the saturation region if V
DS
V
DS,sat
. The
region of the I
D
V
DS
characteristic obtained for V
DS
< V
DS
,sat is called
the triode region.
The I
D
V
DS
Characteristics
The Figure above shows a typical set of I
D
V
DS
characteristics, which
are a family of curves, each measured at a constant V
GS
. We can see that
there are three distinct regions of operation: the cuto region, the triode
region, and the saturation region. The saturation region is used if the FET
is to operate as a amplier. For operation as a switch, the cuto and triode
regions are utilized.
1. Triode
If V
GS
> V
t
and V
DS
V
GS
V
t
, then the n-channel is continuous
all the way from S to D. The S and D are connected by a conductor
(or a resistor) of a given resistance. The drain current increases if
the voltage drop between S and D increases. The channel resistance
Figure 1.12: I
D
V
DS
Curve
depends on how much charge is injected at the S-end, which in turn
is controlled by v
GS
. The Drain current I
d
depends on both v
GS
and
V
GD
(or V
DS
). The I
D
V
DS
characteristics can be approximately
described by the relationship
I
D
= K [2(V
GS
V
t
)V
DS
V
DS
V
DS
] (1.2)
in which K is a device parameter given by
K = 0.5U
n
C
ox
_
W
L
_ _
A
V
2
_
(1.3)
U
n
physical constant known as the electron mobility(its value in
this case applies for the electrons in the induced n channel)
C
ox
oxide capacitance, the capacitance per unit area of the
gate-to-body capacitor for which the oxide layer serves as
dielectric.
L,W the length and the width of the channel.
Since for a given fabrication process the quantity (0.5U
n
*C
ox
) is a
constant, approximately 10A/V
2
for the standard NMOS process
with a 0.1m oxide thickness. So the aspect ratio of
W
L
determines its
conductivity parameter K.
If V
DS
is suciently small so that we can neglect the V
DS
V
DS
in
equation 1.2, then the I
D
V
DS
characteristics near the origin the
relationship
I
D

= 2K(V
GS
V
t
)V
DS
(1.4)
This linear relationship represents the operation of the MOS transistor
as a linear resistance R
DS
R
DS
=
V
DS
I
D
=
1
2K(V
GS
V
t
)
(1.5)
2. Saturation
If V
GS
> V
t
and V
DS
V
GS
V
t
, then N-channel is induced at the S-
end, but the channel is depleted at the D-end. That is, the N-channel
is pinched o at the Drain-end. Increasing V
ds
beyond V
ds(sat)
, or
equivalently decreasing V
GD
below V
t
, creates a fully depleted region
between the inversion n-channel and the drain region. An electric
eld is set up in this region, pointing from the Drain region toward
the inversion channel. Carrier electrons in the N-channel that reach
the depletion boundary are swept across the depletion region into the
Drain. This is similar to PN junction diode where the minority carrier
electrons of the P-side are swept to the n-side by the built-in eld
whenever they reach the depletion boundary. Once the drain-end of
channel is pinched o, the current no longer depends on the voltage
apply between S and D.
The boundary between the triode region and the saturation region is
characterized by
V
DS
= V
GS
V
t
(1.6)
Substituting it into Equation 1.2 gives the saturation value of the
current I
D
is
I
D
= K(V
GS
V
t
) (V
GS
V
t
) (1.7)
Thus in saturation the MOSFET provides a drain current whose value
is independent of the drain voltage V
DS
and is determined by the gate
voltage V
GS
according to the square-law relationship.
The complete independence of I
D
on V
DS
in saturation and the corre-
sponding innite output resistance at the drain is an idealization based
on the premise that once the channel is pinched o at the drain end,
further increases in V
DS
have no eect on the channels shape. In prac-
tice, increasing V
DS
beyond v
DS
,sat does aect the channel somewhat.
Specically, as V
DS
is increased, the channel pinch-o point is moved
slightly away from the drain toward the source. Thus the eective
channel is reduced, a phenomenon called channel-length modulation.
Since the channel resistance is proportional to the channel length, the
channel resistance is decreased. This results in the slight increase of
the drain current beyond the saturation level. Now since K is in-
versely proportional to the channel length (Equation 1.3), so, K and,
correspondingly, I
D
, increases with V
DS
. Mathematically, the channel
length modulation introduces a V
DS
-dependent term in I
D
:
I
D
= K(V
GS
V
t
)(V
GS
V
t
)(1 + V
DS
) (1.8)
the channel-length modulation parameter:0.005 < < 0.03
From Fig.nid.ps we extrapolated the straight-line I
D
V
DS
character-
istics in saturation, intercept the V
DS
-axis at the point V
DS
=
1
=
V
A
. So v
A
is in the range 200 to 30 volts. It should be obvious that
channel-length modulation makes the output resistance in saturation
nite. Let the output resistance Rout as
R
out
=
1
K(V
GS
V
t
) (V
GS
V
t
)
V
GS
= constant (1.9)
approximated by
R
out

=
1
I
D
(1.10)
substituted by =
1
V
A
R
out

=
V
A
I
D
(1.11)
Thus the output resistance is inversely proportional to the DC bias
current I
D
.
3. Cuto
If V
GS
< V
t
(and of course, V
GD
< V
t
), then the no n-channel is present
and no current ows.
1.4.2 P-channel Enhancement-Type MOSFET
A P-channel enhancement-type MOSFET (PMOS transistor) is fabricated
on an N-type substrate with p+ regions for the drain and the source, and
holes as charge carriers. The device operates in the same manner as the
N-channel device except the V
GS
and V
DS
are negative and the threshold
Figure 1.13: P-Channel Test Circuit
voltage V
t
is negative. Also the current i
D
enters the source terminal and
leaves through the drain terminal.
To induce a channel we apply a gate voltage that is more negative than
V
t
, and apply a drain voltage that is more negative than the source voltage
(i.e. V
DS
is negative or, equivalently, v
SD
is positive). The current i
D
is
given by the same equation as for NMOS, and the K is given by
K = 0.5 U
p
C
ox
_
W
L
_
(1.12)
where U
p
is the mobility of holes in the induced p channel. Typically, U
p

=
0.5U
n
, with the result that for the same W/L ratio a PMOS transistor has
half the value of K as the NMOS device.
The I
D
V
DS
characteristics is shown above. The current I
D
is given
by the same equation used for NMOS.
I
D
= K(V
GS
V
t
)(V
GS
V
t
)(1 + V
DS
)
where V
GS
, V
t
, , and V
DS
are all negative.
PMOS technology was originally the dominant one. However, because
NMOS devices can be made smaller and thus operate faster, and because
NMOS requires lower supply voltages than PMOS, NMOS technology has
virtually replaced PMOS. Nevertheless, it is important to develop the PMOS
Figure 1.14: I
D
V
DS
Curve
transistor for two reasons: PMOS devices are still available for discrete-
circuit design, and more importantly, both PMOS and NMOS transistors
are utilized in CMOS circuits!
Chapter 2
An Introduction to Op-Amps
2.1 Parameters of an Op-Amp
This section will discuss Op-Amp parameters. The designer of an Op-Amp
must have a clear understanding of what Op-Amp parameters mean and
their impact on circuit design. The selection of any Op-Amp must be based
on an understanding of what particular parameters are most important to
the application. In the next section, we will discuss the method of measure-
ment of these dierent parameters
2.1.1 Oset Voltage
All Op-Amps require a small voltage between their inverting and noninvert-
ing inputs to balance mismatches due to unavoidable process variations. The
required voltage is known as the input oset voltage and is abbreviated V
os
.
V
os
is normally modelled as a voltage source driving the noninverting input.
Generally, Bipolar input Op-Amps typically oer better oset parameters
than JFET or CMOS input Op-Amps. There are two other parameters re-
lated to and aect V
os
: the average temperature coecient of input oset
voltage, and the input oset voltage long-term drift. The average temper-
ature coecient of input oset voltage, a V
os
, species the expected input
oset drift over temperature. Its units is
_
mV
o
C
. V
os
is measured at the tem-
perature extremes of the part, and a V
os
is computed as
V
os
o
C
. Normal aging in
semiconductors causes changes in the characteristics of devices. The input
oset voltage long-term drift species how V
os
is expected to change with
time. Its units are
mV
month
. Input oset voltage is of concern anytime that
DC accuracy is required of the circuit. One way to null the oset is to use
external null inputs on a single Op-Amp package (2.1). A potentiometer is
17
CHAPTER 2. AN INTRODUCTION TO OP-AMPS 18
Figure 2.1: Oset Voltage Adjust
connected between the null inputs with the adjustable terminal connected
to the negative supply through a series resistor. The input oset voltage
is nulled by shorting the inputs and adjusting the potentiometer until the
output is zero. However, even if the V
os
is nulled at the beginning, it will
change with temperature and some other conditions.
2.1.2 Input Current
The input circuitry of all Op-Amps requires a certain amount of bias current
for proper operation. The input bias current, I
IB
, is computed as the average
of the two inputs:
I
IB
=
(I
N
+I
P
)
2
(2.1)
CMOS and JFET inputs oer much lower input current than standard bipo-
lar inputs. The dierence between the bias currents at the inverting and
noninverting inputs is called the input oset current, I
os
= I
N
+I
P
. Oset
current is typically an order of magnitude less than bias current.
Input bias current is of concern when the source impedance is high. If
the Op-Amp has high input bias current, it will load the source and a lower
than expected voltage is seen. The best solution is to use an Op-Amp with
either CMOS or JFET input. The source impedance can also be lowered by
using a buer stage to drive the Op-Amp that has high input bias current.
In the case of bipolar inputs, oset current can be nullied by matching
the impedance seen at the inputs. In the case of CMOS or JFET inputs,
the oset current is usually not an issue and matching the impedance is
not necessary. The average temperature coecient of input oset current,
Figure 2.2: Output Voltage Swing
I
os
, species the expected input oset drift over temperature. Its units are
_
mA
o
C
.
2.1.3 Input Common Mode Voltage Range
The input common voltage is dened as the average voltage at the inverting
and noninverting input pins. If the common mode voltage gets too high or
too low, the inputs will shut down and proper operation ceases. The com-
mon mode input voltage range, VICR, species the range over which normal
operation is guaranteed. For instance, Rail to rail input Op-Amps use com-
plementary N and P channel devices in the dierential inputs. When the
common-mode input voltage nears either rail, at least one of the dierential
inputs is still active, and the common-mode input voltage range includes
both power rails.
2.1.4 Maximum Output Voltage Swing
The maximum output voltage, V
OM
, is dened as the maximum positive
or negative peak output voltage that can be obtained without waveform
clipping, when quiescent DC output voltage is zero. V
OM
is limited by
the output impedance of the amplier, the saturation voltage of the output
transistors, and the power supply voltages. This is shown pictorially in 2.2.
This emitter follower structure cannot drive the output voltage to ei-
ther rail. Rail-to-rail output Op-Amps use a common emitter (bipolar) or
common source (CMOS) output stage. With these structures, the output
voltage swing is only limited by the saturation voltage (bipolar) or the on
resistance (CMOS) of the output transistors, and the load being driven.
2.1.5 Output Impedance
Dierent data sheets list the output impedance under two dierent condi-
tions. Some data sheets list closed-loop output impedance while others list
open-loop output impedance, both designated by Z
o
. Z
o
is dened as the
small signal impedance between the output terminal and ground. Generally,
values run from 50 to 200 .
Common emitter (bipolar) and common source (CMOS) output stages
used in rail-to-rail output Op-Amps have higher output impedance than
emitter follower output stages. Output impedance is a design issue when
using rail-to-rail output Op-Amps to drive small resistive or large capacitive
loads. If the load is mainly resistive, the output impedance will limit how
close to the rails the output can go. If the load is capacitive, the extra phase
shift will erode phase margin. 2.3 shows how output impedance aects the
output signal assuming Z
o
is mostly resistive.
Figure 2.3: Eect of Output Impedance
2.1.6 Common-Mode Rejection Ratio
Common-mode rejection ratio, CMRR, is dened as the ratio of the dif-
ferential voltage amplication to the common-mode voltage amplication,
A
dif
A
com
. Ideally this ratio would be innite with common mode voltages being
totally rejected.
The common-mode input voltage aects the bias point of the input dif-
ferential pair. Because of the inherent mismatches in the input circuitry,
changing the bias point changes the oset voltage, which, in turn, changes
the output voltage.
2.1.7 Supply Voltage Rejection Ratio
Supply voltage rejection ratio, kSVR (AKA power supply rejection ratio,
PSRR), is the ratio of power supply voltage change to output voltage change.
The power voltage aects the bias point of the input dierential pair.
Because of the inherent mismatches in the input circuitry, changing the bias
point changes the oset voltage, which, in turn, changes the output voltage.
For a dual supply Op-Amp, KSV R =
V CC
V OS
or KSV R =
V
DD
V OS
. The term
V
CC
means that the plus and minus power supplies are changed symmetri-
cally. For a single supply Op-Amp, KSV R =
V CC
V OS
or KSV R =
V
DD
V OS
. Also
note that the mechanism that produces kSVR is the same as for CMRR.
Therefore kSVR as published in the data sheet is a DC parameter like
CMRR. When kSVR is graphed vs. frequency, it falls o as the frequency
increases.
2.1.8 Slew Rate
Slew rate, SR, is the rate of change in the output voltage caused by a step
input. Its units are V/ms or V/ms. 2.4 shows slew rate graphically.
Figure 2.4: Slew Rate
The primary factor controlling slew rate in most amps is an internal
compensation capacitor CC, which is added to make the Op-Amp unity
gain stable. Referring to 2.5, voltage change in the second stage is lim-
ited by the charging and discharging of the compensation capacitor CC.
The maximum rate of change is when either side of the dierential pair is
conducting 2IE. Essentially SR =
2IE
CC
. However, that not all Op-Amps
have compensation capacitors. In Op-Amps without internal compensation
capacitors, the slew rate is determined by internal Op-Amp parasitic capac-
itances. Uncompensated Op-Amps have greater bandwidth and slew rate,
but the designer must ensure the stability of the circuit.
Figure 2.5: Op amp schematic simplied
In Op-Amps, power consumption is traded for noise and speed. In order
to increase slew rate, the bias currents within the Op-Amp are increased.
2.1.9 Unity Gain Bandwidth and Phase Margin
Unity-gain bandwidth (B1) and gain bandwidth product (GBW) are very
similar. B1 species the frequency at which A
V D
of the Op-Amp is 1.
GBW species the gain-bandwidth product of the Op-Amp in an open loop
conguration and the output loaded:
GBW = AV D f (2.2)
Phase margin at unity gain (fm) is the dierence between the amount of
phase shift a signal experiences through the Op-Amp at unity gain and
180
o
:
fm = 180
o
f@B1 (2.3)
Gain margin is the dierence between unity gain and the gain at 180
o
phase
shift:
Gain margin = 1 Gain@180
o
phase shift (2.4)
In order to make the Op-Amp stable, a capacitor, C
C
, is purposely fabricated
on chip in the second stage (2.5). This type of frequency compensation is
termed dominant pole compensation. The idea is to cause the open-loop
gain of the Op-Amp to roll o to unity before the output phase shifts by
180
o
. 2.5 is very simplied, and there are other frequency shaping elements
within a real Op-Amp. 2.6 shows a typical gain vs. frequency plot for an
internally compensated Op-Amp.
Figure 2.6: Voltage Amplication and Phase Shift vs. Frequency
Phase margin and gain margin are dierent ways of specifying the sta-
bility of the circuit. Since rail-to-rail output Op-Amps have higher output
impedance, a signicant phase shift is seen when driving capacitive loads.
This extra phase shift erodes the phase margin, and for this reason most
CMOS Op-Amps with rail-to-rail outputs have limited ability to drive ca-
pacitive loads.
2.1.10 Settling Time
It takes a nite time for a signal to propagate through the internal circuitry
of an Op-Amp. Therefore, it takes a period of time for the output to react
to a step change in the input. In addition, the output normally overshoots
the target value, experiences damped oscillation, and settles to a nal value.
Settling time, t
s
, is the time required for the output voltage to settle to
within a specied percentage of the nal value given a step input. Settling
time is a design issue in data acquisition circuits when signals are changing
rapidly.
Figure 2.7: Settling Time
2.2 Methodology of Choosing Op-Amp Parame-
ters
The methodology of choosing the parameters of the transistors, and their
relationships, then it will be possible to get the desired quiescent point to
ensure the ideal wave output.
Here we will to present a method of choosing the parameter in an Op-
Amp circuit, and in the same way we can get a very ecient optimization
method. Further more in this way we can learn how to combine dierent
parts of a circuit together, and deal with more complex circuits. First of all
we must clarify the relationship among dierent parameters, then we can
nish our job orderly. The quiescent point is very important for our design.
2.3 How to Adjust the Parameters
To simplify the discussion, we will concentrate on the simplest OTA Op-
Amp 2.8 and demonstrate how to adjust its parameters to get the proper DC
gain. At the same time we try to extend the method to other sophisticated
architectures.
2.3.1 Specication
Here we are asked to design an Op-Amp, whose V
DD
= 3.3 V , V
SS
=
0 V , input bias voltage = 1.65V (Input V
0
= 1.65 V ). Swing of output
(Considering the current source V
DD
will occupy some voltage, the swing
should be 0 3 V .)
Figure 2.8: Circuit with Default Parameters
2.3.2 Procedure of Optimization
1. Draw a circuit with default parameters.(See Figure 2.8).
2. Adjust the current source value I
0
. The current value takes highest
priority of all of the parameters, all of the other parameters will be
chosen to match the current value.
From the circuit, we know that when the Op-Amp is working in the
common mode (which means the V
1
= V
0
= 1.65 V ), then the current
through the two (2) dierential ampliers is symmetric, so the value
of I
1
=
1
2
I
0
. V
out
= 1.65 V , V
ref
= 1.65 V , from here we can
decide what is the maximum and minimum value we can get through
a PMOS. We design a very simple circuit to test the current value
from the drain of a PMOS (Figure 2.9). When changing the width of
the channel of the PMOS, we can get a set of values of the current
value of the drain(Figure 2.10). Considering the width of the active
load should not to be too large so we can get better gain. The range
of the length should be 400nm 2000 nm, then I
1
should be from
86.4366A 800.00A. So the I
0
should be double I
1
it should be
from 170A 1700A.
Figure 2.9: Simple Test Circuit for Drain Current
We can choose any value inside the range, so we regard the 400m as
the initial value of our design.
3. Decide the width of the channel of the active load. From Figure 2.9 we
perform a parametric analysis. The value of the width of the channel,
through which we can get the proper value for the quiescent point. It
is about 1.25m.
4. We can assign an arbitrary big value to the width of the dierential
Figure 2.10: Simple Test Circuit - Parametric Analysis of Channel Width
amplier, suppose we choose a value 5 times bigger than the width of
the active load.
5. Do the simulation and compare the gain we get with the gain we want.
Figure 2.11: Simple Test Circuit - DC Response/Gain
2.3.3 Optimize the Parameters of the Op-Amp
1. According to the gain of the specication, decide the width of the
dierential amplier.
G
m
=
_
2K
n
I
0
_
W
L
__1
2
R
out
= V
e
L
1
I
0
A
V
= G
m
R
out
= V
e
_
2K
n
W
1
L
1
I
0
_
frac12
There are three (3) variables that will aect A
V
, W, L and I
0
.
If there is small dierence between the two (2) gains, then what we
need to do is just to adjust the value of the length and the width of
the channel of dierential amplier, otherwise we must reduce the I
0
.
2. Reduce I
0
; according to the fabricating technology, the width of the
channel of the transistor cannot be less than 350 m, which means we
cannot reduce the I
0
less than 170A. As mentioned above, otherwise
the quiescent point cannot be at the middle of the curve. (See Figure
2.12), this will distort the input waveform. Each time I
0
is adjusted
the width of the active load should be changed accordingly. Thus
another way to adjust I
0
must be devised.
Figure 2.12: Waveform Distortion
3. Further reducing I
0
. First, we can cascade active loads together, be-
cause the voltage is on the cascade active load, there is less voltage
across each transistor, so the I
0
can be reduced dramatically. However,
it becomes more complex to further optimize the circuit. See gure
2.13
Figure 2.13: Active Loads Cascaded Together
Figure 2.14: Cascaded Active Loads - DC Response/Gain
4. Cascade Op-Amp - adjusting the quiescent point: If after all of the
above eort, we still cannot get the proper gain, we have to use a
cascade Op-Amp. The key to optimizing the cascade Op-Amp is the
quiescent point, which guarantees the proper operation of the circuit.
Please refer the Figure 2.15 (Output curve of rst the stage).
The output of the rst stage is linear in the area of (0.5 V 2.15 V ).
So we should adjust the quiescent point of the input of the second
stage.
Figure 2.15: Output Curve of the First Stage
Checking the 2 stage Op-Amp, as in Figure 2.16
Figure 2.16: Two Stage Op-Amp Circuit
In the second stage amplier, we get the waveform as shown in Figure
2.17 and Figure 2.18: We notice that the operating point of the am-
plier begins to work is about 0V, actually at this point the output of
Figure 2.17: Second Stage Waveform
the rst stage is in the non-linear area. That means the wave we get
isnt the best.
Figure 2.18: Second Stage Circuit
We can change the parameters of the two (2) transistors, however,
it will have little aect on the non-linear problem. (See Figure 2.19
Non-Linear Problem). Thus we must use another topology. (Figure
2.20).
Figure 2.19: Non-Linear Problem
Figure 2.20: Dierent Topology to overcome Non-Linear Problem
We then get the resulting waveform shown in Figure 2.21.
Figure 2.21: Resulting Waveform
We can adjust the circuit in another way, to change the input linear
area of the second stage. To simplify the problem we can change the
output stage.
Figure 2.22: Circuit with Modied Output Stage
2.3.4 How to get the Quiescent point in a complex circuit
Keeping a good quiescent point is very important and can be easily forgotten
as the circuits become more complex. There are certain ways to design with
the help of the computer, and establish models for dierent stages.
2.4 Target Op-Amp Specications
The table below denes the best and worst cases for several key Op-Amp
parameters, in addition the target Op-Amp parameters for this design are
listed. These values are based upon textbook values, fabricated Op-Amp
datasheets and from other research. During its design these key parameters
were always kept at close hand.
Specication Worst Case Target Case Best Case
Gain 100 1, 000 1, 000 100, 000 1,000,000+
Frequency Range (Hz) 10-20,000 10-200,000 5-500,000
Bandwidth 5.0 KHz 20.0 KHz 50.0 KHz
Input Voltage 1 mV 1 V 0.1 V
Output Resistance 500 K 1 M 10 M
Power Consumption mW W nW
CMRR > 40 dB > 50dB > 60dB
Slew Rate 3 V/sec 2 V/sec 0.5 V/sec
THD 1% 0.1% 0.01%
Parasitic Capacitance 1.0 F 1.0 nF 1.0 fF
Rise Time 1 sec 0.1 sec 0.001 sec
Settling Time 10 sec 1 sec 0.1 sec
Chapter 3
Current Mirrors and Biasing
Networks
One of the most important parts of an analog design is the biasing circuity.
The purpose of the bias circuity is establish an appropriate DC operating
point for the transistor. With the correct DC operating point established a
stable and predictable DC drain current I
D
and a DC drain-source voltage
ensures operation in the saturation region for all input signals that may be
encountered. This component forms the basis for an operational amplier
whereby various circuits like the dierential pair, gain stage and output
stage rely on its awless stable operation.
For the Operational Amplier design ve dierent types of current mir-
rors were examined; Basic Current Mirror, Cascade/Cascode Current Mir-
ror, Wilson Current Mirror, Modied Wilson Current Mirror and Reduced
Cascade/Cascode Current Mirror. The advantages and disadvantages of
each type of current mirror will be outlined later.
The ve current mirrors which were examined were designed in Cadence
and were placed into a standard test circuit consisting of a basic dieren-
tial pair with active load and basic common-source amplier output stage
with a 10K load. These current mirrors were then subjected to several
tests including; DC Sweep, current mirror output impedance and stability
of current supplied across dynamic voltage range.
The ability of a current mirror to hold current constant, the number of
transistors used and their sizes are the general dening factors on whether
a current mirror is Good or not. These factors were considered when de-
ciding on the current mirror to be used in the Op-Amp Design.
35
CHAPTER 3. CURRENT MIRRORS AND BIASING NETWORKS 36
3.1 Ideal Characteristics of a Current Mirror
1. Output current linearly related to the input current. I
out
= A I
in
.
2. Input Resistance is zero.
3. Output resistance is innity.
3.2 Basic Current Mirror Derivation
Below is the derivation of the simple current mirror.
Q1 is operating in the saturation region since its drain is shorted to its
gate.
Thus,
I
D1
=
1
2
K
n
_
W
L
_
1
(V
GS
V
t
)
2
(3.1)
Note we neglect channel-length modulation and assume = 0.
The drain current of Q1 is supplied by V
DD
through a resistor, R.
Assuming gate currents to be approximately 0.
I
D1
= I
ref
=
V
DD
V
GS
R
(3.2)
Now looking at Q2,
It has the same V
gs
as Q1, and assuming it is operating in saturation, its
drain current, which is the output current I
o
of the current source will be,
I
O
= I
D2
=
1
2
K
n
_
W
L
_
2
(V
GS
V
t
)
2
(3.3)
Again neglecting channel-length modulation.
Using equations 1,2 and 3 we are able to relate the output current I
o
to
the reference current I
ref
.
Rearranging equation 1 and substituting I
ref
= I
d1
I
ref
_
W
L
_
1
=
1
2
K
n
_
W
L
_
1
(V
GS
V
t
))
2
(3.4)
We know,
I
O
=
_
1
2
K
n
(V
GS
V
t
)
2
_ _
W
L
_
(3.5)
So substituting into equation 3.5,
I
O
=
I
ref
_
W
L
_
1
_
W
L
_
2
I
O
I
ref
=
_
W
L
_
2
_
W
L
_
1
(3.6)
Thus we have a relationship whereby modifying the width and length we
can change the output current.
Thus if the transistors were matched, i.e. width and lengths equal and
other parameters the same we have,
I
O
I
ref
= 1 (3.7)
I
O
= I
ref
(3.8)
This is called a current mirror since the reference current is mirrored
or held constant at the output.
3.3 Benchmark Test Circuit
The purpose of the test circuit is to establish a benchmark which could be
used to evaluate the performance and design of each of the dierent types
of biasing circuits. This test circuit was kept the same for all tests and the
transistor sizes were set to the smallest possible optimum values. The test
circuit itself consists of a basic dierential pair (W=1900nm, L=350nm), a
basic two transistor active load (W=800nm, L=350nm) and common source
amplier (W=800nm, L=350nm).
Figure 3.1: Benchmark Test Circuit Schematic
The widths of the transistors were adjusted so that the current supplied
to the dierential pair was between 6.50A and 7.00A. Then the output
impedance was measured, as well as, the overall size of the transistors once
the desired current was achieved. These results were recorded out would
help to choose which current mirror would be the Winner.
3.4 Examined Current Mirrors
3.4.1 Basic Current Mirror
We will now examine the most fundamental and simple type of current
mirror, the Basic Current mirror (See Figure 3.2). This type of current
mirror uses a minimum of (3) three transistors. The derivation shown earlier
shows the general operation of a current mirror whereby current is held
constant regardless of the voltage being supplied.
Figure 3.2: Basic Current Mirror Schematic
This circuit is very simple and does a very good job of supplying constant
current, however, it does not supply absolutely stable current. See gure 3.3,
the output current supplied to the active load and the output impedance.
Figure 3.3: Basic Current Mirror Simulation Results
The main advantage of this current mirror is its simplicity and ease
of implementation, however, the major disadvantage is that the current
supplied is not completely stable.
3.4.2 Cascade/Cascode Current Mirror
The second current mirror examined was the Cascade/Cascode Current Mir-
ror. This current mirror uses a minimum of (5) ve transistors, these tran-
sistors can be seen in the Figure 3.4.
Figure 3.4: Cascade/Cascode Current Mirror Schematic
This circuit is a little bit more complex than the simple current mirror
with (2) two extra transistors and would be more than enough for any design.
The main disadvantage to this current mirror is that it is not very good at
supplying higher amounts of current, in particular to the output stages. It
was given very high consideration when deciding which current mirror to
use for the Op-Amp design.
Figure 3.5: Cascade/Cascode Current Mirror Simulation Results
The main advantage to this design is that it provides stable current
and it has relatively small transistor sizes. In addition to this we have a
higher output resistance compared to the basic current mirror. The main
disadvantage to this type of current mirror is a reduced dynamic range.
Thus, it scored high in our choice for current mirrors when testing.
3.4.3 Wilson Current Mirror
The third current mirror examined was the Wilson current mirror. This
current mirror uses a minimum of (4) four transistors. In the Figure 3.6 you
can see the circuit layout.
Figure 3.6: Wilson Current Mirror Schematic
This circuit is not as complex as the cascade current mirror and does
provide good stable current, however, to provide the benchmark 6.50
7.00A very large transistors (200m) had to be used, and thus, given that
the current best, cascade current mirror provided similar qualities but
with much smaller transistor sizes this current mirror was ranked very low.
Figure 3.7: Wilson Current Mirror Simulation Results
As stated above this design was not considered for the Op-Amp design
and the cascade current mirror was the current best.
3.4.4 Modied Wilson Current Mirror
The fourth current mirror examined was the Modied Wilson current mirror.
This current mirror uses a minimum of (5) ve transistors and has a similar
layout as the cascade current mirror. This current mirror was expected to
perform similar to the regular Wilson current mirror. You can see the circuit
schematic in Figure 3.8.
Figure 3.8: Modied Wilson Current Mirror Schematic
Upon testing the results revealed that the initial idea that it would per-
form similarly to the regular Wilson current mirror were conrmed. It had
similar results with the output current as output resistance and the transis-
tor sizes needed to attain the benchmark current we also large (200u).
Figure 3.9: Modied Wilson Current Mirror Simulation Results
Upon seeing these results, the Modied Wilson current mirror was not
to be chosen as the current mirror for the Op-Amp design.
3.4.5 Reduced Cascade Current Mirror
The fth and nal current mirror examined was the Reduced Cascade cur-
rent mirror. This current mirror used a minimum of (6) six transistors and
had a similar layout to that of the basic cascade current mirror. The word
Reduced in the name refers to the reduced voltage at which the current
reaches a stable output, it is usually about (1/2) one-half of the usual volt-
age.
Figure 3.10: Reduced Cascade Current Mirror Schematic
This current mirror is a bit more complex and the transistor sizes are
comparable to the basic cascade current mirror. However, the reduced cas-
cade current mirror oers a reduced voltage and is able to provide higher
amounts of current. This is very useful in the output stages were higher
amounts of current are needed to bias the output stages.
Figure 3.11: Reduced Cascade Current Mirror Simulation Results
The results of this current mirror appeared to be the most promising
for the Op-Amp design, it provided stable current, it was able to provide
higher amounts of output current if necessary, as well as, it reduced the volt-
age at which the current was stable compared to the other current mirrors
examined.
3.5 Conclusion
The decision to use the Reduced Cascade/Cascode Current Mirror in the de-
sign of this Op-Amp are now shown here. First, the reduced cascade/cascode
current mirror provided stable and linear current in all test situations. Sec-
ondly, it provided very high output resistance and a very low input resis-
tance. Third, it was able to supply larger amounts of current which were
needed to drive the output stage. Lastly, given it had a few extra transistors
their sizes were small compared to other current mirror setups looked at. In
the table below you can see the current mirrors ranked from best to worst
(top to bottom) as well as a few of the ranking criterion and their evaluation.
The Basic Cascade/Cascode current mirror would have also done a very
good job in the design, however with the reduced cascades reduced voltage
it oered a bit of an advantage over the basic cascade current mirror.
Current Mirrors Min # of FETs Current Output Input Complexity Transistor Sizes
Ranked: (Best to Worst) including I
ref
Stability Resistance Resistance needed to achieve
Transistors Ref 7.0A Current
Reduced Cascade/Cascode 6+2N Excellent Very High Very Low Very High Small
Basic Cascade/Cascode 3+2N Very Good Very High Very Low Very High Very Small
Basic Wilson 2+2N Good High Very Low Moderate Very Large
Modied Wilson 3+2N Good High Very Low Moderate Very Large
Basic 2+N Poor Low Very Low Simple Small
N = Number of times current is mirrored. (In these designs N = 2,
dierential Pair Biasing and Output Stage)
Chapter 4
Dierential Input Stage
4.1 The Unbuered Op-Amp
Figure 4.1: Two Stage Op-Amp
The amplier shown in gure 4.1 can be segregated into:
1. Biasing block.
2. Dierential input stage.
3. Output stage.
48
CHAPTER 4. DIFFERENTIAL INPUT STAGE 49
The biasing block, consists of M8, M9, and M5. The bias current greatly
aects the performance of the amplier. The dierential input stage is
composed of M1, M2, M3, M4 with: M1 matching M2, M3 matching M4.
The output stage consists of M6 and M7.
4.2 Small Signal Equivalent Circuits
Figure 4.2: NFET and Its Small Signal Model
In the analysis of a FET amplier circuits, the FET can be replaced by
the equivalent circuit model, 4.2. Where:
r
o
=
|V A|
I
D
Where V A =
1
g
m
is the transconductance parameter = K
n
W
L
(V
GS
V
t
)
Therefore, the small signal model parameters:
g
m
and r
o
depends on the DC bias of the FET.
Ideal constant DC voltage sources are replaced by short circuits, this is a
result of the fact that the voltage across an ideal constant DC voltage source
does not change and thus there will always be a zero voltage signal across
a constant DC voltage source. The signal current of an ideal constant DC
current source will always be zero thus an ideal constant DC current source
can be replaced by an open circuit.
For the gate drain connected FET device, the model is as shown in gure
4.3.
Since v
ds
= v
gs
(the eective resistance =
v
gs
v
gs
g
m
=
1
g
m
)
Figure 4.4 shows only the dierential input stage of the circuit. In gure
4.1 M5 are replaced by current source I
ss
. The small signal analysis of the
Figure 4.3: MOSFET Resistor and Its Small Signal Model
dierential input stage can be accomplished with the assistance of the model
shown in gure 4.5 which is only appropriate for dierential analysis when
both sides of the amplier are assumed to be perfectly matched. If this
condition is satised, then the point where the two sources of M1 and M2
are connected can be considered at AC ground. The body eect is neglected.
Figure 4.4: CMOS Dierential Amplier using n-channel Input Devices
The model in gure 4.6 is the simplied to the model shown in gure
4.5. Since: M1 matches M2, M3 matches M4 (The point where S1, S2 of
M1, M2 are connected can be considered at AC ground.) Since S3, S4 are
Figure 4.5: The Exact Model for the CMOS Dierential Amplier
AC ground, therefore S1, S2, S3, S4 can be joined up in a node.
Figure 4.6: The Simplied Equivalent Model
Referring to gure 4.6:
C
1
= c
gd1
+c
gs3
+c
gs4
C
2
= c
gd2
C
3
= c
gd4
Any small signal on gate of M1 will result in a small signal current i
d1
, which
will ow from drain to source of M1.
i
d1
= g
m1
v
gs1
and also be mirrored from M3 to M4
g
m4
v
gs4
= i
d1
Now since S1, S2, S3, S4 have the same potential (one node)
i
d1
will also ow from the source to drain of M2. g
m2
v
gs2
= i
d1
i
out
= i
d1
(i
d1
) = 2i
d1
(4.1)
r
out
= r
o2
r
o4
(4.2)
i
d1
= g
m1
v
gs1
Since v
gs1
= v
gs2
v
id
= v
gs1
+v
gs2
Therefore, v
id
= 2v
gs1
i
d1
v
id
=
g
m1
v
gs1
2
vgs
=
g
m
1
2
v
id
=
2i
d1
g
m1
(4.3)
Now the small signal output voltage is simply:
v
out
= i
out
r
out
(Substitute 4.1 and 4.2)
v
out
= 2i
d1
(r
o2
r
o4
) (4.4)
Now dividing 4.4 over 4.3 :
v
out
v
id
=
2i
d1
(r
o2
r
o4
)
2i
d1
g
m1
v
out
vid
= g
m1,2
(r
o2
r
o4
) (4.5)
Now:
g
m1,2
= (2
1,2
I
D1,2
)
1
2
(r
o2
r
o4
)
=
1
2I
D,2
(since r =
1
I
)
Therefore,
v
out
v
id
= (2
1,2
I
D1,2
)
1
2
_
1
2I
D1,2
_
v
out
v
id
= K
_
W
1,2
L
1,2
I
D1,2
_1
2
_
1
_
K is a constant, uncontrollable by the designer. The eect of on the
gain diminishes as L increases, such that
1
is directly proportional to the

channel length.
Then a proportionality can be established between
W
1,2
L
1,2
and the drain
current versus the small signal gain such that:
v
out
v
id
_
W
1,2
L
1,2
I
D1,2
_1
2
Therefore,
small signal gain
_
W
1,2
L
1,2
I
D1,2
_1
2
(4.6)
The constant was not included since the value is not dependent on anything
the designer can adjust.
Conclusions:
1. Increasing W
1,2
, L
1,2
or both increases the gain.
2. Decreasing the drain current through M1, M2 which is also
1
2
I
D6
,
increases the gain.
4.3 The Frequency Response
Referring back to the model of the input stage: gure 4.6. We will work to
eliminate all low impedance nodes (having high RC time):
If,
1
c
1
1
g
m3
1
[c
2
(r
o2
r
o4
)]
Then the node (D1=D2=G3=G4) is a low impedance node (with large RC
time) Now: c
3
is assumed to be zero. In most applications of dierential
ampliers, this assumption turns out to be valid. Then the model to be
considered for the high frequency response analysis is turned to be as shown
in gure 4.7.
In the conguration where the small signal is applied to the gate of M4
while the gate of M2 is grounded, v
gs2
= 0, v
id
= v
gs1
. Now:
g
m4
v
gs4
= i
d1
= g
m1
v
gs1
= g
m1
v
id
Figure 4.7: High Frequency Small Signal Model with Parasitic Capacitors
Now redrawing the model in Figure 4.7 with the new parameters will
result in the model shown in Figure 4.8.
Figure 4.8: Model of the Input Stage used to Determine the Frequency Response
The high frequency output can be given by:
v
o1
= g
m1
v
id
(r
o2
r
o4
)
1
_
1 +S
1
c2(r
o2
r
o4
)
_
The freq. Response =
v
o1
v
i
d
= g
m1
(r
o2
r
o4
)
1
_
1 +S
1
c
2
(r
o2
r
o4
_
Now let us consider the input and the output stage shown in gure 4.9.
The capacitor C
c
in Figure 4.8 is removed for the purpose of the rst
analysis. C
1
, C
2
represent the total lumped capacitance from each ground.
Since the output nodes associated with each output is a high impedance,
these nodes will be the dominant frequency dependent nodes in the circuit.
Since each node in a circuit contributes a high frequency pole, the frequency
response will be dominated by the high impedance nodes.
Figure 4.10 is the model derived for Figure 4.9 making use of Figure 4.8
to determine the frequency response of the two-stage Op-Amp.
To determine the exact value for c
1
, c
2
, Figure 4.11 shows the parasitic
capacitors explicitly for the input and the output stage which include the
bulk depletion capacitors (c
gb
, c
sb
, c
db
) and the overlap capacitors (c
gs
, c
gd
).
Figure 4.9: Two Stage Op-Amp with Lumped Parasitic Capacitors
Figure 4.10: Model Used to Determine the Frequency Response of the Two-Stage
Op-Amp
Miller theorem was used to determine the eect of the bridging capacitor
c
gd6
connected from the gate to drain of M6. Miller theorem approximates
the eects of gate-drain capacitor by replacing the bridging capacitor with
an equivalent input capacitor of value c
gd
(1+A2) and the equivalent output
capacitor with a value of c
gd
(1 +
1
A2
). Where:
Figure 4.11: Two Stage Op-Amp with Parasitic Capacitors Shown Explicitly
A2: gain across the original bridging capacitor and is, from gure 4.10:
A2 =
v
o
v
o1
= g
m6
v
o1
r
o6
r
o7
v
o1
A2 = g
m6
(r
o6
r
o7
) (4.7)
Thus c1, c2 for g(10) can be determined by examining gure 4.11.
c
1
= c
db4
+c
gd4
+c
db2
+c
gd2
+c
gs6
+c
gd6
(1 +A2)
c
2
= c
db6
+c
db7
+c
gd7
+c
gd6
_
1 +
1
A2
_
+c
L
Now assume c
1
< c
2
(the pole associated with the di-amp output
_
1
c
1
(r
o2
r
o4
)
_
will be lower in frequency than the pole associated with the output of the
output stage
_
1
c
2
(r
o6
r
o7
)
_
.
Also from the high frequency model gure 4.10:
v
o
v
id
=
_
v
o
v
o1
_
_
v
o1
v
id
_
_
1
1 +
S
c
2
(r
o6
r
o7
)
_
_
1
1 +
S
c
1
(r
o2
r
o4
)
_
v
o
v
o1
= g
m6
(r
o6
r
o7
)
1
_
1 +
S
c
2
(r
o6
r
o7
)
_
v
o1
v
id
= g
m1
(r
o2
r
o4
)
_
1
1 +
S
c
1
(r
o2
r
o4
)
_
Therefore, the frequency response
v
o
v
id
= [g
m6
(r
o6
r
o7
)] [g
m1
(r
o2
r
o4
)]
_
1
1 +
S
c
1
(r
o2
r
o4
)
_
_
1
(1 +
S
c
2
(r
o6
r
o7
)
_
(4.8)
And the poles are:
P
1
=
1
c
1
(r
o2
r
o4
)
(4.9)
P
2
=
1
c
2
(r
o6
r
o7
)
(4.10)
4.4 Phase Margin
The phase margin is the dierence between the phase at the frequency at
which the magnitude plot reaches 0dB and the phase at the frequency at
which the phase has shifted 180
o
. It is recommended for stability reasons
that the phase margin of any amplier be at least 45
o
(60
o
is recommended).
A phase margin below 45
o
will result in long settling time and increased
propagation delay. The Op-Amp system can also be thought of as a simple
second order linear control system with the phase margin directly aecting
the transient response of the system.
Phase margin measurement procedure with the Cadence design tool, the
phase margin can be measured by applying the following steps:
1. Obtain the AC response simulation
2. Delete all the outputs in the Arma window
3. Select the AC response curve (it will turn to yellow).
4. From the Arma window: Result direct plot AC magnitude and
phase
5. Follow the prompt at the bottom of the schematic window and select
the output node
6. From the Arma window : output to be plotted
7. You will get the magnitude and the phase frequency response. Split
the graphs.
Figure 4.12: Miller Phase Margin Measurement
4.5 Compensation
The goal of the compensation task is to achieve a phase margin greater than
45
o
. Now we will include C
c
and the model will be as in Figure 4.13
Keeping in mind that the two poles of the system without compensation
as determined previously are:
P
1
=
1
c
1
(r
o2
r
o4
)
Figure 4.13: Model Used to Determine the Frequency Response of the Two-Stage
Op-Amp
P
2
=
1
c
2
(r
o6
r
o7
)
Two results come from adding the compensation capacitor C
c
:
1. The eective capacitance shunting r
o2
r
o4
is increased by the additive
amount of approximately g
m1
r
o2
r
o4
C
c
. This moves P
1
down by
a signicant amount. C
c
will dominate the value of c
1
and will cause
the pole P
1
to roll o much earlier than without C
c
to a new location.
2. P
2
is moved to a higher frequency.
Let r
1
= r
o2
r
o4
, r
2
= r
o6
r
o7
,
v
o
v
id
=
g
m1
g
m6
r
1
r
2
_
1 S
C
c
g
m6
_
1 +S [r
1
(c
1
+C
c
) +r
2
(c
2
+C
c
) +g
m6
r
1
r
2
] +S
2
r
1
r
2
[c
1
c
2
+C
c
(c
1
+c
2
)]
(4.11)
A general second order polynomial can be written as:
P(S) = 1 +aS +bS
2
=
_
1
S
P
1
__
1
S
P
2
_
= 1 S
_
1
P
1
+
1
P
2
_
+
S
2
P
1
P
2
If | P
2
|| P
1
| then:
P(S) = 1
S
P
1
+
S
2
P
1
P
2
Therefore, P
1
, P
2
may be written in terms of a, b as:
P
1
=
1
a
P
2
=
a
b
The key in this technique is the assumption that the magnitude of the root
P
2
is greater than the magnitude of the root P
1
P
1
=
1
r
1
(c
1
+C
c
) +r
2
(c
2
+C
c
) +g
m6
r
1
r
2
C
c
P
1

=
1
g
m6
r
1
r
2
C
c
P
2
=
r
1
(c
1
+C
c
) +r
2
(c
2
+C
c
) +g
m6
r
1
r
2
C
c
r1
r
2(c
1
c
2
+C
c
(c
1
+c
2
))
P
2

=
g
m6
C
c
c
1
c
2
+c
2
C
c
+c
1
C
c
P
2

=
g
m6
c
2
(4.12)
The second pole should not begin to aect the frequency response until after
the magnitude response is below 0dB.
It is of interest to note that a zero occurs in the right-hand-plane (RHP)
due to the feed forward path through C
c
. The RHP zero is located at:
Z
1
=
g
m6
C
c
(4.13)
This RHP zero has negative consequences on our phase margin, causing the
phase plot to shift 180
o
more quickly. To avoid the eect of RHP zero, one
must try to move the zero well beyond the point at which the magnitude
plot reaches 0dB (suggested rule of thumb: factor of 10 greater)
4.6 Adding R
z
in series with C
c
One remedy to the zero problem is to add a resistor R
z
in series with C
c
Z
1
=
1
_
C
c
_
1
g
m6
R
z
__ (4.14)
And the zero can be pushed into the LHP where it adds phase shifts and
increases phase margin if:
R
z
>
1
g
m6
(4.15)
The zero location is in the RHP, when R
z
= 0. As R
z
increases in
value, the zero gets pushed to innity at the point at which R
z
=
1
g
m6
.
Once R
z
>
1
g
m6
, the zero appears in the LHP where its phase shifts adds to
the overall phase response. Thus improving the phase margin. This type of
compensation is referred to as lead compensation and is commonly used as a
simple method for improving the phase margin. One should be careful about
using R
z
, since the absolute values of the resistors are not well predicted.
The value of the resistor should be simulated over its max and min values to
ensure that no matter if the zero is pushed into the LHP or RHP, that value
of the zero is always 10 times greater than the gain bandwidth product.
4.7 Gain Bandwidth Product
GBW for the compensated Op-Amp is the open loop gain multiplied by the
bandwidth of the amplier (as set by P
2
).
GBW = g
m1
r
1
g
m6
r
2
_
1
g
m6
r
1
C
c
r
2
_
GBW

=
g
m1
C
c
(4.16)
GBW
_
I
D1,2
W
1,2
L
1,2
_1
2
C
c
Therefore, the most ecient way to increase GBW is to decrease C
c
.
The value of C
c
must be large enough to aect the initial roll-o frequency
as larger C
c
improves phase margin. Therefore, to know the value of C
c
:
1. We need to know the GBW specication.
2. Iteratively choosing the values for
W
1,2
L
1,2
and I
D1,2
and then solving for
C
c
.
Conclusions:
P
2
should be > GBW Therefore,
g
m6
c
2
>
g
m1,2
C
c
C
c
> C
2
_
g
m1,2
g
m6
_
(4.17)
Practically speaking, the load capacitor usually dominates the value of c
2
,
so c
2
= c
L
C
c
> c
L
_
g
m1,2
g
m6
_
(4.18)
Therefore, the eect of c
L
on phase margin is as follows: Minimum size of
Cc directly depend on the size of c
L
. For example if the zero is 10 times
larger than GBW, then in order to achieve a 45
o
phase margin, P
2
must be
least 1.22 times higher than GBW. And to get a phase margin of 60
o
P
2
must be 2.2 times greater than GBW.
4.8 Large Signal Consideration
Analysis and design can be greatly simplied by separating DC or bias
calculations from small signal calculations. That is once a stable operating
point has been established and all DC quantities are calculated, we may
then perform AC analysis. Figure 4.14 shows the large signal equivalent
circuit model for n-channel MOSFET in saturation. The model for p-channel
CMOS is similar except for K
p
instead of K
n
.
Figure 4.14: Large-Signal Equivalent Circuit Model in Saturation
For N-channel MOSFET to be in saturation:
1. V
GS
V
t
2. V
DS
(V
GS
V
t
)
And the drain saturation current is given by:
I
D
=
1
2
K
n

W
L
(V
GS
V
t
)
2
If = K
W
L
Then,
I
D
=
1
2
(V
GS
V
t
)
2
(4.19)
For P-channel MOSFET to be in saturation:
1. V
GS
V
t
2. V
DS
(V
GS
V
t
)
Equation 4.19 can be written as:
V
GS
=
_
2I
D
_1
2
+V
t
(4.20)
For the N-channel:
1. V
DS
(V
GS
V
t
). Now by substituting V
GS
:
2. V
DS

_
2I
D
_1
2
The large signal characteristics that are important include the Common-
mode range, slew rate, and output signal swing
4.9 Slew Rate
The slew rate is dened as the maximum rate of change of the output volt-
age due to change in the input voltage. For this particular amplier, the
maximum output voltage is ultimately limited by how fast the tail current
device M5 can charge and discharge the compensation capacitor, the slew
rate can then be approximated as:
SR =
dV
o
dt
SR
I
D5
C
c
Typically, the di-amp is the major limitation when considering slew
rate. However, the tradeo issues again come into play. If I
D5
is increased
too much, the gain of the di-amp may decrease below a satisfactory amount.
If Cc is made too small then the phase margin may decrease below an
acceptable amount.
4.10 The Common-Mode Range
Common-mode range is dened as the range between the maximum and
minimum common-mode voltages for which the amplier behaves linearly.
Referring to 4.15, suppose that the common mode voltage is DC value and
that the dierential signal is also shown. If the common mode voltage is
swept from ground to V
DD
, there will be a range for which the amplier
will behave normally and where the gain of the amplier is relatively con-
stant. Above or below that range, the gain drops considerably because the
common-mode voltage forces one or more devices into the triode region.
The maximum common-mode voltage is limited by both M1 and M2
going into triode. This point can be dened by a borderline equation in
which
(For M1, M2 to stay in saturation)
V
D1,2
(V
G1,2
V
t1
)
Now:
V
D1
= V
DD
V
SG3
Therefore,
V
G1
V
DD
V
G3
+V
t1
Substitute the value of V
GS
from equation 4.20 and V
t3
= ve value because
it is p type CMOS:
V
G1
V
DD
_
2I
D3
3
_1
2
V
t3
+V
t1
Now since I
D3
=
1
2
I
D5
:
V
G1
V
DD
_
I
D5
3
_1
2
V
t3
+V
t1
(4.21)
V
G1
V
DD
_
L
3
I
D5
W
3
K
3
_1
2
V
t3
+V
t1
(4.22)
The minimum voltage is limited by M5 being driven into non-saturation
by the common-mode voltage source.
V
D5
V
SS
+V
G5
V
t5
V
D5
= V
G1,2
V
GS1,2
Therefore,
V
G1,2
V
GS1,2
V
SS
+V
G5
V
t5
V
G1,2
V
SS
+V
G5
V
t5
+V
GS1,2
Since, V
GS
=
_
2I
D
_1
2
+V
t
(From 4.20)
Therefore,
V
G1,2
V
SS
+
_
2I
D5
5
_1
2
+
_
I
D5
1
_1
2
+V
t5
V t5 +V
t1
V
G1,2
V
SS
+
_
2I
D5
5
_1
2
+
_
I
D5
1
_1
2
+V
t1
V
G1,2
V
SS
+
_
2L
5
I
D5
W
5
K
5
_1
2
+
_
L
1
I
D5
W
1
K
1
_1
2
+V
t1
(4.23)
Determining the CMR for the two-stage Op-Amp (Figure 4.15).
Figure 4.15: Determining the CMR for the Two-Stage Op-Amp
4.11 Important Relationships for The Design
Relate W
3
, W
4
to W
5
M3, M4 carry half of I
tail
, then the widths of M3 and
M4 can be determined by assuming that v
sg3
= V
sg4
= V
gs5
. Using 16:
I
d3,4
I
d5
=
0.5K
P
_
W
L
_
3,4
_
V
gs3,4
V
2
t
_
0.5K
n
_
W
L
_
5
(V
gs5
V
t
)

K
P
_
W
L
_
3,4
K
n
_
W
L
_
5
And since L3 = L5 and Kn = 3Kp, then that leads to the conclusion
that:
W
3,4
= 1.5W
5
(4.24)
Therefore, the width of M3, M4 can be determined in terms of W
5
.
Relate W
6
, W
7
to W
5
. The values for M6 and M7 are determined by the
amount of load capacitance attached to the output. If a large capacitance is
present, the width of M6 and M7 will need to be large so as to provide enough
sinking and sourcing current to and from the load capacitor. Suppose it was
decided that the current needed for M6 and M7 was twice that of M5. Then:
W
7
= 2W
5
(4.25)
W
6
= 6W
5
(Because W
6
is P-Type so to account for the dierence K values)
4.12 Tradeos for Increasing the Gain of the Two
Stage Op-Amp.
Using the equations in the summary above, the tradeo issues for increasing
the gain of the two stage Op-Amp are summarized in the following table.
We wish to... Thus we could... Some Secondary eects are...
Increase DC gain Increase
W
L
1,2
Decrease phase margin
Increase GBW
Increase CMR
Decrease ID5 Decrease SR
Increase CMR
Increase CMRR
Increase phase margin
Increase
W
L
6
Increase phase margin
Increase output swing
Decrease I
D6
Decrease output currents drive
Decrease phase margin
4.13 Design Methodology for the Two Stage Op-
Amp
Design methodology is a topology dependent subject and it is highly depen-
dent on the analysis of the circuit. The purpose on the design tool is not
to replace the analysis completely. The analysis rather guides the designers
thoughts and while doing the design and helps to make the results of the
design tool make sense.
The following design methodology is developed for the two stage Op-Amp
analyzed above.
1. To dene the requirements, set the specications, and dene the bound-
ary conditions. Boundary conditions The TSMC 0.35 CMOS Technol-
ogy is used. Process specications: V
t
, K
, C
ox
, ..etc. Because there
are 12 models for N-FET and 12 models for P-FET automatically
selected within Cadence tool so the process parameters cannot be de-
termined at this stage.
Supply voltage: 0 3.3 V Operating temperature: 0 to 70
o
C,
Requirements, Gain, GBW, CMRR Slew rate, Input common mode
range: V
in
(min), V
in
(max), Output voltage swing: V
out
(max), V
out
(min),
PSRR, Oset, Output Load.
2. Choose the smallest length that will keep the channel modulation pa-
rameter ( constant and give good matching for current mirrors. The
value used is 350nm
3. Design the two stages without C
c
, R
z
, or C
L
, for the best dc response.
Design the devices sizes for proper dc performance. It is important
that before any AC analysis is performed, the values of the DC points
in the circuit be checked to ensure that every device is in saturation.
Failure to do so will result in very wrong answers. This step is accom-
plished as follows:
(a) Calculate the minimum value for the compensation capacitor C
c
.
From 40: C
c
0.22 C
L
. Get C
L
from the requirements
(b) Get the slew rate from the requirement and then calculate the
minimum value for the tail current I
D5
from: ID5 = SR. C
c
(c) Using cadence design tool, simulate the two stage amplier with
starting widths of: 800nm for n type devices, 1.9m for p type
devices. Get the circuit to work and test the DC response.
(d) Measure I
D5
. Now try to optimize the Op-Amp for maximum
gain by:
i. Measure the existing I
D5
. To get the desired value of I
D5
calculated in B, simultaneously change W5 and the width of
the related devices:
W
3,4
= 1.5W
5
W
6
= 6W
5
W
7
= 2W
5
ii. Check the operating points of all the devices (M1, M2, M3,
M4, M5, M6, and M7) to make sure that the devices are in
saturation. All design parameters can be known from Ca-
dence by doing the following:
Arma Window: Results Print DC Operating Points
iii. Increase the gain by increasing the size of W
1,2
. You may
perform parametric analysis for dierent width size. You
can not decide on the nal value of W1,2 at this point even
you may get high gain, because of the trade o issues and
the secondary eect, particularly on the phase margin which
can not be measured till the ac analysis is performed.
iv. At this point you can extract the process parameters: V
t
, K, C
ox
...etc. Knowing K for the devices is now important to calcu-
late V
G
(min) and V
G1
(max) to make sure that M1, M2, M3,
M4 and M5 are not operating near the boundary, 4.22,4.23
causing one of the devices to operate outside the saturation
region.
v. If any of the devices are brought out of saturation or are
working at the minimum condition required for saturation,
then it should be brought back to saturation.
vi. Repeat 1-4 until all the transistors are in saturation while
the dc gain is increased.
(e) Measure the CMR and the gain. Compare with the required
specications.
4. (a) Add up C
c
and C
L
to the amplier. The value of C
L
is from
the specications while the value of C
c
is previously calculated in
Step3-A
(b) Check the phase margin condition (4.18), as now you can know
all the process parameter from cadence.
(c) Verify that the phase margin is > 45
o
by measuring it using the
procedure in 4-4.
(d) If the phase margin is not > 45
o
then it should be corrected
by changing the value of C
c
or R
2
, referring to (4.15, 4.18). A
parametric analysis may be preformed for dierent C
c
sizes.
(e) Calculate the slew rate according to the new Cc value that makes
the phase margin > 45
o
. If it is less than the specication then
perform the following steps:
i. Calculate the required I
D5
from (22). Substitute the last
value of C
c
and the required SR.
ii. Perform DC analysis only and return back to Step3-D, with
the desired value of I
D5
as calculated above.
5. (a) Calculate the required R
z
using (4.14)
(b) Add R
z
to the circuit and perform an AC analysis. Check the
dierence in the phase response.
6. Measure the slew rate, GBW, P2 and the phase margin. Compare
with the specications.
4.14 Design Example
Figure 4.16 shows a two stage op-amp designed using the above design
methodology and optimized for maximum gain. Figure 4.17 shows the gain
for the large signal consideration(DC response). Figure 4.18 shows the fre-
quency response. Figure 4.19 shows the phase margin measurements.
Figure 4.16: Two-Stage Op-Amp using design Methodology
Figure 4.17: Gain for the Large Signal Analysis
Figure 4.18: Frequency Response
Figure 4.19: Phase Margin
4.15 Limitations of the Two Stage Op-Amp
1. Insucient gain
2. Limited stable bandwidth caused by the instability to control the
higher order poles of the op-amp.
3. Poor power supply rejection ratio.
4.16 The Cascode Op-Amp
The motivation for using the cascode conguration to increase the gain can
be seen by examining how the gain of the two-stage op-amp could be in-
creased. There are three ways in which the gain of the two stage op-amp
could be increased:
1. Add additional gain stage
2. Increase the transconductance of the rst or the second stage.
3. Increase the output resistance seen by the rst or second stage.
Due to possible instability, the rst approach is not attractive. Of the
latter two approaches, the third is the more attractive way because the out-
put resistance increases in proportion to a decrease current [ r
o
= 1/ ],
whereas the transconductance increases as the square root of the increase in
bias current [ g
m
= (2 B I
D
)
1
/2]. Thus it is generally more ecient to
increase ro rather than g
m
.
Figure 4.20 shows a cascode dierential stage. Figure 4.21 shows the
DC response for an unoptimized conguration to serve as indication for
the whole project and to help in deciding about the best input stage to be
considered in the nal op-amp design. The transistors Mc1 and Mc2 perform
the resistance multiplication, while Mc3 is used to keep the drain voltage of
the input transistor matched, which helps to reduce the voltage oset.
Rout ( [gmc2 . roc2 . ro4 ] [ gmc1. roc1. ro2 ]. A = gm1.R
out
One of the disadvantages of this design is the requirement for the addi-
tional bias voltages V
B1
, V
B2
. Further more, the common mode input range
is reduced due to the extra voltage drop required by the two cascode devices,
Mc1, Mc2. In many cases, the CMR limitation is not important since the
non-inverting input of the op-amp will be connected to ground.
Figure 4.20: Cascode Dierential Stage
Folded Cascode Op-Amp Figure 4.22 shows a folded cascode dierential
stage. Figure 4.23 shows the DC response for an unoptimized conguration
to serve as indication for the whole project and to help in deciding about
the best input stage to be considered in the nal op-amp design.
This circuit uses a current-folding circuit technique to permit direct con-
nection of the drains of the p-channel dierential amplier to the sources of
the cascode devices. This requires two additional transistors (M5 and M6),
operate in much the same manner as the previous cascode circuit. Here
however, the input common-mode range is larger because only three tran-
sistors are now stacked in the input chain between the two power supplies
(as compared to ve in the conventional cascode circuit). The folded cas-
code circuit is frequently used as a single-stage op-amp. Its voltage gain can
be determined as: A = gm1 R
o
Where Ro is the output resistance: R
o
= [
gmc1. roc1(ro6ro2)] [ gmc2. roc2. ro3)]
Figure 4.21: DC Response for Unoptimized Circuit
Figure 4.22: Folded Cascode Dierential Stage
Figure 4.23: DC Response for Unoptimized Circuit
Chapter 5
Inverting Ampliers
In this section, we introduce an important conguration that is extensively
employed in the design of integrated circuits: the inverting amplier. The
inverting amplier is also called common-source amplier or inverter. It is
usually used as the basic gain stage for CMOS circuits. Later we will see
that inverter is used in the gain stage of the nal circuit.
Typically, the inverter has three kinds of congurations. The rst one
uses the common-source conguration with an active resistor as a load. The
second one uses a current sources/sink as an active load. The third one uses
the input voltage to control both the amplifying transistor and the load
transistor, this conguration is also known as push-pull amplier. In this
section, the three congurations will be discussed respectively in terms of
their output swing, small-signal voltage gain, input and output resistance
and -3 dB frequency.
76
CHAPTER 5. INVERTING AMPLIFIERS 77
5.1 Inverter with Active Resistor Load
The typical circuit of the rst type has been shown in Figure 5.1. This
circuit uses a common-source, N-channel transistor M1 with a P-channel
transistor M2 as the load of M1.
Figure 5.1: Simple Inverter Circuit
The parameters of the circuit are as follows: V
DD
= 3.3V, V
SS
= 0V ;
W
1
= 10.8m, L
1
= 10m, V
TN
= 0.5V ; W
2
= 28m, L
2
= 10m, V
TP
=
0.7V . We set the operating point at 1.65V and DC current at 106.7A,
so V
gs1
= 1.65V , I
d1
= 106.7A, I
d2
= 106.7A Under these conditions,
V
ds1
= 1.365V V
ds2
= V
gs2
= 1.935V . Thus we can calculate
1
and
2
which we will need for future calculations. Since
I
d1
=

N
C
ox
W
1
2L
1
(V
gs1
V
TN
)
2
106.7 10
6
=

N
C
ox
W
1
2L
1
(1.65 0.5)
2
So
1
=

N
C
ox
W
1
2L
1
= 171.6A/V
2
, K
p
=
p
C
ox
= 52.9A/V
2
Use the same way, we have:
2
=

P
C
ox
W
2
L
2
= 148.2A/V
2
, K
p
=
P
C
ox
= 52.9A/V
2
1. Transfer Characteristics
Figure 5.2 illustrates the Cadence simulation result of the large-signal
characteristics of this circuit. This plot shows the I
d
versus V
ds
char-
acteristics of M1 and the load line plotted on the same graph. The
voltage-transfer curve is shown in Figure 5.3.
Figure 5.2: Inverter Type 1: Parametric Analysis sweeping V
in
Figure 5.3: Voltage-Transfer Curve
From these gures, we can observe that this type of inverting amplier
has low gain since the slope of the V
out
versus V
in
is small.
2. Output Swing
From these curves, it is obvious that this inverter has limited output
voltage range (point A and point B in Figure 5.2). We will nd the
maximum output voltage and minimum output voltage.
With the gate and drain connected together, M2 always works in the
saturated region as long as | V
ds2
|>| V
TP
|, and V
ds2
= V
gs2
=| V
TP
|
+
_
2I
d2
2
. When V
in
is zero, M1 is cuto, and the drain current of M1
I
d1
is also zero, V
ds2
=| V
TP
|. So V
out
(max) is approximately equal
to:
V
out
(max) = V
DD
| V
TP
|= 3.3 0.7 = 2.6(V )
When V
in
increases until V
in
> V
TN
, M1 conducts, and V
out
begins to
drop. When V
out
> V
in
V
TN
, M1 enters saturated region. When V
out
further drops until V
out
< V
in
V
TN
, M1 enters non-saturated region.
This is the region where the minimum output voltage happens. Since
M1 is in non-saturated region, the drain current of M1 should be:
I
d1
=
1
_
(V
DD
V
SS
V
TN
)(V
out
V
SS
)
(V
out
V
SS
)
2
2
_
Because M2 is in saturated region, the drain current of M2 should be:
I
d2
=

2
2
(V
DD
V
out
| V
TP
|)
2
Equating the above two equations, and simplifying V
TN
=| V
TP
|= V
T
in order to simplify the calculation. We can get the approximate value
of V
out
(min).
V
out
(min) V
DD
V
T

V
DD
V
SS
V
T
_
1 +

2
1
= 3.3 0.5
3.3 0.5
_
1 +
148.2
171.6
= 0.7 V
You may notice that in Figure 5.3, the V
out
(max) is around 3V, dif-
ferent from our calculation. This is because the subthreshold eect of
transistors. We usually suppose that no current ows in transistors
when V
in
< V
TN
. But in reality, there is current conducting through
transistors when V
in
< V
TN
, and I
d
has an exponential relationship
with V
ds
at this region. This is called subthreshold eect. It is this
eect that results in the dierence between our calculation and the
simulation result.
3. Small-Signal Voltage Gain.
There are two methods to calculate the small-signal voltage gain.
(a) The principle of the rst method is this: the gain is found when
M1 works in saturated region, and the drain current of M1 and
M2 are always equal. Therefore, I
d1
is:
I
d1
=

1
2
(V
in
V
SS
V
TN
)
2
and since M2 always operates in saturation, I
d2
is:
I
d2
=

2
2
(V
DD
V
out
V
TP
)
2
Equating the above equations. Because A
V
is the derivative of
V
out
V
in
, we have
A
V
=
V
out
V
in
=
_
2
_1
2
=
_
K
N
W
1
L
1
K
P
W
2
W
1
_1
2
(b) Another method to nd A
V
is from the small-signal model. The
small-signal model of the inverter with an active-resistor load is
shown in Figure 5.4.
Figure 5.4: Inverter: Small-Signal Model
The gain can be expressed as:
A
V
=
v
out
v
in
=
g
m1
g
ds1
+g
m2
+g
ds2
=
_
K
N
W
1
L
1
K
P
W
2
L
2
_1
2
where g
m2
g
ds1
, g
ds2
.
We get the same results from the two methods. According to the
Cadence simulation result, g
m1
= 172, g
m2
= 155.5, so A
V
=
1.1, which is a fairly low gain. So this type of inverter is suitable
for situations where low-gain inverting stage is desired. Besides,
we can see from the A
V
expression that in order to improve the
gain, we can either increase the ratio of
W
1
L
1
or decrease the ratio
of
W
2
L
2
.
4. Input and Output Resistance
From the small-signal model, we can nd that the input and output
resistance is
r
in
=
r
out
=
1
g
ds1
+g
m2
+g
ds2
=
1
g
m2
=
1
155.5 10
6
= 6.4 k
We can see that the output resistance of the active-resistor load in-
verter is low because of the diode-connected transistor M2. The re-
sulting low output resistance can be very useful in situations where a
large bandwidth is the main expectation from an inverting gain stage.
5. Upper -3 dB Frequency
When discussing the frequency response of the inverter, we need to
take the internal capacitors into consideration. Figure 5.5 illustrates
the resulting small-signal model.
C
1
= C
gs1
C
2
= C
bd1
+C
gs2
+C
bd2
C
3
= C
gd1
r
2
= r
out
=
1
g
out
r
1
= r
in
=
Using nodal analysis, we may write:
V
out
(s)(g
out
+sC
2
) +C
3
[V
out
(s) V
in
(s)] +g
m
V
in
(s) = 0,
where s = j
Figure 5.5: Inverter: Small-Signal Model
So the transfer function
V
out
(s)
V
in
(s)
is:
A
V
(s) =
g
m
_
1 s
_
C
3
g
m
__
g
out
_
1 +s
_
(C
2
+C
3
)
g
out
__ = A
V M
1
s
Z
1
1 +
s
P
1
(5.1)
where A
V M
is the midband gain, Z
1
is the zero and P
1
is the pole and
A
V M
=
g
m1
g
out
=
g
gm1
g
ds1
+g
m2
+g
ds2
, Z
1
=
g
m1
C
2
, P
1
=
g
out
C
2
+C
3
From this formula, we nd that inverter has a rst-order transfer func-
tion. So the -3 dB frequency can be written as:
H
=
g
out
C
2
+C
3
=
g
m2
+g
ds1
+g
ds2
C
db1
+C
gd1
+C
gs2
+C
db2
(5.2)
From this equation, we can get to the conclusion that the -3 dB fre-
quency of the active resistor load inverter is approximately propor-
tional to the square root of the drain current. So in order to increase
the bandwidth, we can increase the drain current because r
out
will
decrease. Unfortunately, this will also decrease the gain.
Lacking of accurate values of these capacitors, we cannot calculate an
accurate value of the -3 dB frequency. But it still can be found from
the frequency response because -3 dB frequency is the point where the
gain drops to 0.707A
V
. Simulation results of the frequency response
of the active-resistor load inverter is shown in Figure 5.6. From this
curve, we determine the -3 dB frequency to be 26.12MHz, which is a
fairly wide bandwidth.
Figure 5.6: Frequency Response of the Active-Resistor Load Inverter
5.2 Inverter with Current Source/Sink Load
The second type of inverter uses active load. There are a number of ways to
congure active load, such as cascade current sink, current mirror, reduced
cascade current sink and so on. Here we will use current mirror. The circuit
is shown in Figure 5.7.
In order to compare the performances of this type with the rst one,
set the operating point the same as the rst one, i.e. V
gs1
= 1.65V , I
d1
=
106.7A, V
gs2
= 1.935V , I
d2
= 106.7A, I
d3
= I
d4
= 106.7A.
From
I
d
=
C
ox
W
2L
(V
gs
V
T
)
2
, K
N
= 158.9A/V
2
, K
P
= 52.0A/V
2
We can calculate out:
W
1
= 10.8m; L
1
= 10m; W
2
= 28m; L
2
= 10m;
W
3
= 28m; L
3
= 10m; W
4
= 18.5m; L
4
= 10m
Figure 5.7: Inverter Circuit with Current Source/Sink Load
1. Output Swing.
The large-signal characteristics and voltage-transfer curve are respec-
tively shown in Figure 5.8 and Figure 5.9.
Figure 5.8: Inverter Type 2: Parametric Analysis sweeping V
in
Observing these curves, we can nd that this type of inverter still
experiences the limitation for output swing(point A in Figure 5.8).
This limitation can be found by a method similar to that used for the
active resistor inverter.
Basically, we can divide the transfer curve into four distinct segments.
In segment 1, V
in
< V
TN
, M1 is cuto. M2 is conducted, but there is
no current. So V
out
(max) is equal to V
DD
since M2 can pull output
voltage up to V
DD
. In segment 2, V
in
increases to V
in
> V
TN
, M1
begins to conduct, and V
out
decreases as V
in
increase. M2 works in
non-saturation region. In segment 3, V
in
further increases, making
both M1 and M2 operates in saturation. V
out
dropped quickly. The
transfer curve is almost linear and very steep, indicating large voltage
gain. In segment 4, V
in
increases to V
in
> V
out
+V
TN
, and M1 enters
non-saturation region. This is the segment where V
out
(min) happens.
From the above analysis, we can see that V
out
(max) = V
DD
= 3.3 V ,
and the lower limit can be found when M1 is in the non-saturation
region. V
out
(min) can be written as
V
out
= (V
DD
V
TN
)
_
_
_
1
_
1
_
1
__
V
DD
(V
DD
V
gs2
) | V
TP
|
V
DD
V
TN
_
2
_1
2
_
_
_
= (3.3 0.5)
_
_
_
1
_
1
_
148.2
171.7
__
3.3 1.365 0.7
3.3 0.5
_
2
_1
2
_
_
_
= 0.3 V
(5.3)
The small-signal voltage gain can be found from its small-signal model
Figure 5.10.
Figure 5.10: Inverter Type 2: Small Signal Model V
in
The gain can be expressed as
A
V
=
v
out
v
in
= g
m1
(r
ds1
r
ds2
) =
g
m1
g
ds1
+g
ds2
Because g
ds1
=
1
I
d
,g
ds2
=
2
I
d
and g
m1
=
2
1
I
d
A
V
=
_
2
N
C
os
W
1
L
1
I
d
_
1
1
+
2
_
Here we notice a signicant result. The gain increases as the DC
operating current decreases. This occurs because the output conduc-
tance r
ds1
and r
ds2
is inversely proportional to the current whereas the
transconductance g
m1
is proportional to square root of the bias cur-
rent. So we can increase the gain of this type of inverter by decreasing
the biasing current I
d
. According to the Cadence simulation result,
g
m1
= 172.3, g
ds1
= 392.3n, g
ds1
= 855.9n, so
A
V
=
g
m1
g
ds1
+g
ds2
=
172.3 10
6
392.3 10
9
+ 855.9 10
9
= 137.6
3. Output Resistance.
From the small-signal model, we can nd that the output resistance is
r
out
=
1
g
ds1
+g
ds2
=
1
392.3 10
9
+ 855.9 10
9
= 0.8M
From these calculation results, we can see that under the same biasing
conditions, this current source load inverter has a much higher gain and
output resistance when comparing to the active resistor load inverter.
But you will see soon that the high output resistance results in a
narrow bandwidth.
4. -3 dB Frequency. The frequency response can be found by using Figure
5.5. But in this case, C is
C
1
= C
gs1
+C
gs2
C
2
= C
db1
+C
db2
C
3
= C
gd1
+C
gd2
So -3 dB frequency is
H
=
g
out
C
2
+C
3
=
g
ds1
+g
ds2
C
db1
+C
gd1
+C
gd2
+C
bd2
(5.4)
Simulation result of the frequency response of this inverter is shown
in Figure 5.11. From this curve, we measure the -3 dB frequency is
3.75MHz, which is a lower frequency compared to the active resistor
load inverter. This is due to the higher output resistance.
5.3 Push-Pull Inverter
One of the disadvantages of the current source inverter introduced above is
that it requires a biasing voltage. If the gate of M2 in Figure 5.7 is taken
to the gate of M1, the push-pull CMOS inverter is achieved. In this circuit,
Figure 5.11: Frequency Response of the Inverter with Current Source/Sink Load
Figure 5.12: Push-Pull Inverter Circuit
both biasing and amplifying transistors are driven by the input signal. The
typical circuit is shown in Figure 5.12.
Still, like the above two circuits, set the operating point at 1.65V, i.e.
V
gs1
= 1.65V , I
d1
= 106.7A, V
gs2
= 1.65V , I
d2
= 106.7A From
I
d
=
C
ox
W
2L
(V
gs
V
T
)
2
, K
N
= 158.9A/V
2
, K
P
= 52.9A/V
2
We can get,
W
1
= 10.8m, L
1
= 10m, W
2
= 48.1m, L
2
= 10m;
1. Output Swing.
The large-signal characteristics and voltage-transfer curve are respec-
tively illustrated in Figure 5.13 and Figure 5.14.
Figure 5.13: Inverter Type 3: Parametric Analysis Sweeping V
in
Comparing the large-signal voltage transfer characteristics between
the current-mirror and push-pull inverter, it is seen that this type
of inverter shows two advantages. First, the push-pull inverter will
have a higher gain. This is due to the fact that both transistors are
being driven by V
in
. Second, the output swing of push-pull inverter
is capable of operating from V
DD
to V
SS
, whereas a current mirror
inverter cannot drive all the way to V
SS
.
The small-signal voltage gain can be found from its small-signal model
Figure 5.15.
Figure 5.15: Inverter Type 3: Small Signal Model
The small-signal voltage gain can be expressed as
A
V
=
v
out
v
in
=
g
m1
+g
m2
g
ds1
+g
ds2
Because g
ds1
=
1
I
d
, g
ds2
=
2
I
d
, g
m1
=
2
1
I
d
and g
m2
=
2
2
I
d
:
A
V
=
v
out
v
in
=
g
m1
+g
m2
g
ds1
+g
ds2
=
(2
1
)
1
2
+ (2
2
)
1
2
(
N
+
P
)
I
d
(5.5)
From this equation, we can get to the conclusion that the small-signal
voltage gain of pull-pull inverter is inversely proportional to the square
root of DC current. In order to increase gain, we can decrease the DC
operating current. This is the same as the current mirror inverter.
According to the Cadence simulation result, g
m1
= 172.1, g
ds1
=
547.7n, g
m2
= 208.2, g
ds1
= 761.4n, so
A
V
=
v
out
v
in
=
g
m1
+g
m2
g
ds1
+g
ds2
=
172.1 10
6
+ 208.2 10
9
547.7 10
9
+ 761.4 10
9
= 290.5
This result shows that under the same DC current, push-pull inverter
has a much higher gain than the other two types.
3. Output Resistance and -3 dB Frequency
H
.
From the small-signal model, we can nd that the output resistance
and -3 dB frequency are the same as the current mirror inverter, which
are
r
out
=
1
g
ds1
+g
ds2
and
H
=
g
out
C
2
+C
3
=
g
ds1
+g
ds2
C
bd1
+C
gd1
+C
gd2
+C
bd2
In this case,
r
out
=
1
547.7 10
9
+ 761.4 10
9
= 0.76 M
Simulation result of the frequency response of the push-pull inverter
is shown in Figure 5.16, and the -3 dB frequency is 2.98MHz.
Figure 5.16: Frequency Response of the Push-Pull Inverter
5.4 Comparison
We have nished the analysis of the three types of inverters: active load
inverter, current source inverter and push-pull inverter. Main performances
of the three inverters are concluded in the following table.
Type 1 Type 2 Type 3
Active Resistor Load Current Source Load Push-Pull
Operating Point V
gs1
= 1.65, I
d1
= 1.06.7A
Output V
out
(max) = 2.6V V
out
(max) = 3.3V V
out
(max) = 3.3V
Swing V
out
(min) = 0.8V V
out
(min) = 0.3V V
out
(min) = 0V
Gain 1.1 137.6 290.5
Output Resistance 64k 0.8M 0.76M
-3dB Frequency 26.12MHz 3.75MHz 2.98MHz
The comparison shows that under the same DC current, the active resistor
load inverter has the lowest gain but the best bandwidth. So this type of
inverter is suitable for situations where low gain and wide bandwidth is
desired. Push-pull inverter has the highest gain and output swing. But we
all know that this type of inverter may cause cross-over distortion. The
performance of the current source load inverter is between the other two
types, which is a modest gain and output swing.
5.5 Application
In the nal circuit, the second type of inverter is used to make a crossover
stage. The purpose of this stage is to provide gain and drive to the two out-
put transistors. Figure 5.17 illustrates the basic conguration of a crossover
stage.
Figure 5.17: Application 1
In this diagram, the two inverters consisting of M1, M3 and M2, M4
are the crossover stage. You can see that these two inverters are the cur-
rent mirror load inverter introduced above. Figure 5.18 shows the practical
circuit in the our groups design.
Figure 5.18: Application 2
In this circuit, the crossover stage is achieved with transistors M9, M22,
M23 and M10, M24, M25. Here the current mirror is replaced by reduced
cascade current source to bias the inverting amplier, and it comes from
our biasing subcircuit. The input of the inverters is from the rst stage, i.e.
dierential stage. The output of the rst inverter consisting of M9, M22,
M23 will drive transistor M11 of the output stage, and the second inverter
consisting of M10, M24, M25 will drive M12 of the output stage.
The performance of the two inverters is similar to that of the current
mirror load inverter introduced above. This is the specic application of
the current source load inverter. This stage can provide high gain for the
circuit. Its high input and output resistance make it match well with both
the dierential stage and the output stage.
Chapter 6
Control Network and Output
Stage
The primary objective of the CMOS output stage is to function as a current
transformer. Most output stages have a high current gain and a low voltage
gain. The specic requirements of an output stage might be:
1. Provide sucient output power in the form of voltage or current.;
2. Avoid signal distortion.;
3. Be ecient. (The eciency can be dened as the ratio of the power
dissipated on the load against the power delivered from the supply);
4. Provide protection from abnormal conditions. (Short circuit, over tem-
perature and so on).
The second requirement results from the fact that the signal swings are large
and the non-linearity normally not encountered in small-signal ampliers
will become important. The third requirement is born out of the need to
minimize power dissipation in the driver transistors themselves compared
with that dissipated in the load. The fourth requirement is normally met
with CMOS output stages since MOS devices are self-limiting.
An important function of the output stage is to provide the amplier
with a low output resistance so that it can deliver the output signal to the
load without loss of gain. Since the output stage is the nal stage of the
amplier, it usually deals with relatively large signals. Thus the small-signal
approximations and models either are not applicable or must be used with
care. Nevertheless, linearity remains a very important requirement. In fact,
95
CHAPTER 6. CONTROL NETWORK AND OUTPUT STAGE 96
a measure of quality of the output stage is the total harmonic distortion
(THD) it introduces.
6.1 Classication of Output Stage
Several approaches to implementing the output amplier will now be in-
troduced here. Output stages are classied according to the drain current
waveform that results when an input signal is applied. For example, in a
Class-A amplier, the transistor conducts for the entire cycle of the input
signal, while in a Class-B amplier, a transistor only conducts for half of the
cycle of the input sine wave. And the Class-AB amplier, involves biasing
the transistor at a nonzero DC current much smaller the peak current of the
sine-wave signal. As a result, the transistor conducts for an interval slightly
greater than half a cycle. The resulting conduction angle is greater than
180
o
but much less than 360
o
. The Class-AB amplier has another transis-
tor that conducts for an interval slightly greater than that of the negative
half-cycle, and the currents from the two transistors are combined in the
load.
6.2 Class-A Output Stage
Here we introduce two kinds of Class-A amplier output stages:
6.2.1 Simple output amplier using a Class-A, current-source
inverter
The circuit in Figure 6.1 will reduce the output resistance and increase the
current driving capability. The output resistance can also be seen in Figure
6.2.
Figure 6.1: Reduced Output Resistance/Increased Current Driving Circuit.
Figure 6.2: Output Resistance for a Class-A Output Stage.
6.2.2 Common-Drain (Source-Follower) Output Amplier
The circuit conguration shown in Figure 6.3 has both large current gain
and low output resistance. But since the source is the output node, the MOS
device becomes dependent on the body eect. The body eect causes the
threshold voltage V
t
to increase as the output voltage is increased, creating
a situation where the maximum output is substantially lower than V
DD
. It
is seen that two N-channel devices are used, rather than a P-channel and an
N-channel.
The eciency of the source follower can be shown to be similar to the
class-A amplier. The distortion of the source follower will be better than
the Class-A amplier because of the inherent negative feedback of the source
follower. Eciency is dened as the ratio of the power dissipated in R
L
(load
resistor) to the power required from the power supplies.
Figure 6.3: Class-A Output Stage - Source Follower Circuit.
6.2.3 Power Analysis
Next, we see that if the bias current, I, is properly selected the output voltage
can swing from 0 to V
DD
with the quiescent value being 1.65V (the absolute
maximum value should be
V
DD
2
). Now, assuming that the bias current, I,
is selected to allow a maximum negative load current of
V
DD
2R
L
, the drain
current of the upper transistor will swing from 0 2I with the quiescent
value being I. So the instantaneous power dissipation in the transistor should
be: PD
1
= V
ds1
i
d1
. The maximum instantaneous power dissipation in the
upper transistor is V
DD
I
2
. So the upper transistor must be able to withstand
a continuous power dissipation of V
DD
I
2
(I =
V
DD
2R
L
).
The power conversion eciency of an output stage is dened as
E
ff
=
load power(P
L
)
supply power(P
S
)
(6.1)
For the Class-A output stage, the average load power will be
P
L
=
V
2
o
2 R
L
(6.2)
Since the current in the lower transistor is constant (I), the power drawn
from the negative supply is V
DD

I
2
. The average current in the upper
transistor is equal to I, and thus the average power drawn from the positive
supply is V
DD

I
2
. Thus, the maximum eciency attainable is 25% when
V
o
=
1
2
V
DD
. Because this is a rather low value, the Class-A output stage
is rarely used in large power applications. Note, that in practice the output
voltage is limited to lower values in order to avoid transistor saturation and
associated nonlinear distortion. Thus, the eciency achieved is usually in
the 10% to 20% range.
6.3 Class-B Output Stage
6.3.1 Push-Pull, Inverting CMOS amplier
The Push-Pull amplier (Figure 6.4) has the advantage of better eciency.
It is well known that a Class-B, push-pull amplier has a maximum eciency
of 78.5% which means that less quiescent current is needed to meet the
output-current demands of the amplier. The circuit operates in a push-
pull fashion: N-channel transistor pushes current into the load when V
in
is
positive, and P-channel transistor pulls current from the load when V
in
is
negative. Note that there exists a range of V
in
centered around zero where
both transistors are cut o and V
out
is zero. This Dead Band results
in the crossover distortion. The crossover distortion of a Class-B output
stage can be reduced substantially by employing a high-gain Op-Amp and
overall negative feedback. A more practical method for reducing and almost
eliminating crossover distortion is found in the Class-AB amplier. The
output voltage swing is limited to a threshold voltage below V
DD
and above
V
SS
.
Figure 6.4: Class-B Output Stage - Push-Pull Circuit.
Figure 6.5: Class-B Output Stage - Transfer Characteristics.
Figure 6.6: Output Resistance for a Class-B (and Class-AB) Output Stage.
6.3.2 Power Analysis
To calculate the power-conversion eciency, E
ff
, of the Class-B stage, we
neglect the crossover distortion and consider the case of an output sinusoid
of peak amplitude V
o
. So P
L
=
V
2
o
2R
L
The current drawn from each supply
will consist of half sine waves of peak amplitude. Thus, the average current
drawn from each of the two power supplies will be
V
o
R
L
.
E
ff
=
load power(P
L
)
supply power(P
S
)
It follows that the maximum eciency is obtained when V
o
is at its
maximum
V
DD
2
. The power-conversion eciency is 78.5%. This value is
much larger than that obtained in the Class-A stage.
6.4 Class-AB Output Stage
Crossover distortion can be virtually eliminated by biasing the complemen-
tary output transistor at a small, nonzero current. The result is the Class-
AB output stage. (See Figure 6.7). The Class-AB stage operates in much
the same manner as the Class-B circuit, with one important exception: for
small input, both transistors conduct, and as input is increased or decreased,
one of the two transistors takes over the operation. Since the transition is a
smooth one, crossover distortion will be almost totally eliminated.
Figure 6.7: Class-AB Output Stage Schematic.
This type of amplier is sometimes called an operational transconduc-
tance amplier (OTA). It is very useful in driving capacitive loads. And
because it based on a oating current source, the output voltage can swing
from V
SS
to V
DD
. And the power relationships in the Class-AB stage are
almost identical to those derived for Class-B circuit in the previous section.
6.5 Short Circuit Protection
The circuit shown in Figure 6.8 shows a Class-AB output stage equipped
with protection against the eect of short-circuiting the output while the
stage is sourcing current. The large current that ows through M1 in the
event of a short circuit will develop a voltage drop across R1 of sucient
value to turn M2 on. The drain of M2 will then conduct most of the current
I
bias
, robbing M1 of its base drive. The current through M1 will thus be
reduced to a safe operating level. This method of short-circuit protection
is eective in ensuring device safety, but it has the disadvantage that un-
der normal operation about 0.5V drop might appear across each R. This
means that the voltage swing at the output will be reduced by that much,
in each direction. On the other hand, the inclusion of the resistors provides
the additional benet of protecting the output transistors against thermal
runaway.
Figure 6.8: Short-Circuit Protection Circuit Schematic.
6.6 Conclusion
After having studied several dierent kinds of the output stages, and based
on the analysis above, the decision to use the Class-AB output stage was
nal. Again, the most important factors being:
1. It has high power-conversion eciency (Maximum 78.5%).
2. It has low output resistance.
3. It can eliminate the crossover distortion in Class B.
Figure 6.9 shows the nal output stage used; the Crossover-Class-B output
stage. It can be divided to two stages: one is the normal Class-B stage,
the other is the crossover stage. It provides the gain, bias, compensation
and drive to the two transistors in the nal stage. These two stages form
the Class-AB stage, and provide good performance. The total harmonic
distortion in this circuit is very small, only 0.05%. Figures 6.10, 6.11 and
6.13 show the three main types of analysis performed, AC Response, DC
Response and Transient Response, respectively.
Harmonic Analysis: The distortion of an amplier is always identied by
its response to the sinewave input signal. It is generated by the non-linearity
of the amplier transfer characteristics. So if we know that the input signal
is:
V
in
() = V sin t (6.3)
Figure 6.9: Final Output Stage with Crossover + Class-B.
Figure 6.10: AC Response of The Final Output Stage.
Then the output distortion should be:
V
out
() = a
1
V sin (t)+a
2
V sin (2t)+a
3
V sin (3t)+...+a
n
V sin(nt)
(6.4)
So the i
th
harmonic distortion (HD) is dened as the magnitude of the i
th
harmonic to the magnitude of the 1
st
harmonic. For example, the second
harmonic distortion is HD
2
=
a
2
a
1
And the total harmonic distortion (THD)
Figure 6.11: DC response of The Final Output Stage.
Figure 6.12: Transient Response of The Final Output Stage.
is dened as:
THD =
_
_
a
2
2
+a
2
3
+... +a
2
n
a
1
(6.5)
Normally, the harmonic distortion of an ideal amplier should be less
than 1.0%.
6.7 Design Considerations
While working on the output stage, an important phenomena was seen:
when the gain is increased the slope is increased and the linear range will
decrease. So if the gain is very high and the input signal is not small enough,
there will be harmonic distortion. In the nal circuit, the magnitude of
the input signal is 1, so there is almost no harmonic distortion. If the
magnitude of the input signal is increased harmonic distortion will begin to
appear. For example, when the magnitude is 100V , the harmonic distortion
is around 10%. Is there a way that could deal with the harmonic distortion
when the input signal is increased to get higher output swing?
6.7.1 Negative Feedback
Negative feedback is always employed in the amplier design to eect one
or more of the following properties:
Desensitize the gain: that is, make the value of the gain less sensitive
to variations in the value of circuit components, such as variations that
might be caused by the changes in temperature.
Reduce nonlinear distortion: that is, make the output proportional to
the input, make the gain constant independent of signal level.
Reduce the eect of noise: that is, minimize the contribution to the
output of unwanted electric signals generated by the circuit compo-
nents and extraneous interference.
Control the input and output impedances: that is, raise or lower in-
put and output impedances by the selection of appropriate feedback
topology.
Extend the bandwidth of the amplier.
All of the above desirable properties are obtained at the expense of a re-
duction in gain and at the risk of the amplier becoming unstable (that is
oscillating).
6.7.2 Frequency Compensation
From the bode plot, we know that if the closed-loop-Amplier has one or two
poles, it is stable because the maximum phase shift of output is always less
than 180 degrees. But if there are more than 2 poles, it will become unstable.
That means the amplier will oscillate. So in order to keep it stable, we
must use the frequency compensation. A popular method for frequency
compensation involves connecting a feedback capacitor to an inverting stage
in the amplier. This causes the pole formed at the input of the amplier
stage to shift to a lower frequency and thus become dominant, while the pole
formed at the output of the amplier stage is moved to a very high frequency
and thus becomes unimportant. This process is known as pole splitting. And
there are two parameters that are very important in the feedback amplier
design: gain margin and phase margin. The gain margin represents the
amount by which the loop gain can be increased while stability is maintained.
Feedback ampliers are usually designed to have sucient gain margin to
allow for the inevitable changes in loop gain with temperature, time and
so on. On the other hand, for a stable amplier, the phase lag must be
less than 180
o
. And feedback ampliers are normally designed with a phase
margin of at least 45
o
. The amount of phase margin has a profound eect
on the shape of the close-loop magnitude response.
Figure 6.13: Phase of the Crossover Output Stage.
Chapter 7
Integrating the Sub-Circuits
The last steps in the design ow are to combine the circuits and test the
full schematic in order to ensure that there are no unexpected interactions
between the various sub-circuits. The larger the design the more dicult
these interactions can become to understand. Figure 7.1 shows the full
schematic representations.
Figure 7.1: Schematic of Entire Op-Amp
108
CHAPTER 7. INTEGRATING THE SUB-CIRCUITS 109
7.1 Overall Performance
A simulation of the completed circuit was performed with a purely resistance
load of 100 Ms (shown in gure 7.2) and an input signal of 1KHz with
a magnitude of 1 V . It can be seen from the transient analysis that the
output waveform is linearly amplied around the operating voltage of 1.65
V. The gain is approximately 95 10
4
(easily seen on the AC Response). The
dierential input voltage, in the DC Sweep, is being measured relative to
the operating voltage of 1.65 V.
Figure 7.2: Overall Performance with 1 V ,1KHz Input
A close up of the DC Sweep transition region can be seen in gure 7.3.
The input voltage swing range can be seen to be approximately 16 V
Figure 7.3: DC Performance with 1 V ,1KHz Input
7.2 The Measurement of Some Main Parameters
In this part, we introduce some methods to measure parameters and then
apply some of them into the design of the Op-Amp to evaluate it.
7.2.1 Input Oset Voltage
Figure 7.4: Measurement of Oset Voltage
Method: in 7.4, DUT is the Op-Amp measured. We presume its an
ideal Op-Amp which has its non-inverting pin connected to an oset voltage
source. Although there is no input signal, V
out
doesnt equal to zero because
of the oset voltage (V
os
).
For an ideal Op-Amp: V
+
= V
For this circuit: V

+
= V
os
V
= V
out
_
R
2
R
1
+R
2
_
Then V
os
= V
out
_
R
2
R
1
+R
2
_
The result curve shows that when V
in
= 0, V
out
= 2.173V , because
R
2
R
1
+R
2
=
100
100 + 1M
= 10
4
so
V
os
= 2.173 10
4
= 0.217mv
Figure 7.5: Schematic for Measuring Oset Voltage
Figure 7.6: Result Plot of Oset Voltage Test
7.2.2 Common-Mode Rejection Ratio (CMRR)
Figure 7.7: Measurement of CMRR
Figure 7.8: Schematic for Measuring CMRR
Figure 7.9: Test Results of CMRR Simulation
Method:
CMRR(dB) = 20 log
A
d
A
c
(7.1)
A
d
is the dierential mode signal gain of Op-Amp.
A
c
is the common mode signal gain of Op-Amp.
For the 7.11, signal source is a low frequency AC signal. We get the following:
Figure 7.10: Equation
By calculating it:
Generally, A
d
A
c
, A
d
1, and (1 +A
d
) R
2
R
1
, so approximately,
CMRR =
A
d
A
c
=
R
1
R
2
V
out
V
in
(7.2)
In the test circuit 7.11,
R
1
R
2
= 10
4
, so
CMRR(dB) = 20 log 10
4
V
in
V
out
(7.3)
From the curve of 7.9, the amplitude of V
o
ut is about 0.55mV and the
amplitude of the input signal has been set to 1mV. So we can get the result
of our designs CMRR is
CMRR(dB) = 20 log
_
10
4
1
0.55
_
= 85.2 dB (7.4)
7.2.3 Output Resistance - R
o
Method: Mr.Sedra and Mr.Smith provide us a method to get the close-loop
output resistance. (microelectronic circuits. Third edition. Pages 94 to
95). The schematic is below:
Figure 7.13: Derivation of the closed-loop output resistance
In order to nd the output resistance of a closed-loop-Amplier, we short
the signal source, which makes the inverting and non-inverting congura-
tions identical and apply a test voltage V
x
to the output as shown in Figure
7.14. Then the output resistance R
out
=
V x
I
can be obtained by straightfor-
ward analysis of the circuit.
We get the result directly(procedures omitted):
R
out
= (R
1
+R
2
)||(
R
o
1 +Ab
) (7.5)
Note that R
out
is the closed-loop resistance of Op-amp and R
o
is the
open-loop resistance. The former one varies according to dierent feed-
back networks but the latter is a very important inherent characteristic of
the Op-amp. From equation 7.5, Normally R
o
is much smaller than R
1
+R
2
,
so we get:
R
out
=
R
o
1 +Ab
(7.6)
b is dened as: b =
R
1
R
1
+R
2
Consider the situation described as the schematic below:
Figure 7.14:
At this situation, R
1
0 and R
2
, then b = 0, R
out
= R
o
. So
by calculating R
out
=
V
x
I
, we should get the open-loop resistance simultane-
ously.
Simulation result of our Op-amps output resistance is some curves like:
Figure 7.15: Schematic for Measuring Output Resistance
Figure 7.16: Result of Output Resistance Test
7.3 Delivering Power to the Load/Instantaneous
Power
The maximum amount of power which can be delivered to the load can
easily be tested by decreasing the resistance of the load. We shall see that
the if the load becomes too small the Op-Amp will not be able to maintain
the correct output voltage. See Figure 7.17.
Figure 7.17: AC & DC Simulation for a Parametric Load with 1 V ,1KHz Input
It is obvious that the amount of current which the circuit can supply is
not enough. In order to increase the driving power which the circuit has, it is
necessary to increase the strength
_
W
L
_
of the push-pull output transistors.
Figure 7.18 shows the instantaneous power of the Op-Amp at 1KHz through
a 10M load.
Figure 7.18: Instantaneous Power
7.4 Improving the Output Buer
A simplied schematic for testing the improved output buer is shown in
gure 7.19. By increasing the transistor widths to: 15m for the N-channel
and 32.84m for the P-channel; we can improve the amount of power the
Op-Amp can supply. Two capacitors have been added to the control network
in this schematic. These Miller Capacitors are required for stability.
Figure 7.19: Output Test Schematic
Figure 7.20: AC & DC Simulation with larger Output Transistors for 1 V ,1KHz
Input
The size of the output transistors may be increased further if the design
specications require more current to be delivered. The larger the output
transistors the more care must be taken in making sure that the transistors
are equally matched.
7.4.1 Stabilizing the Output
Without the new Miller Capacitors the Op-Amp becomes unstable. This
was found in preliminary results; shown in gure 7.21. It should be noted
that the noise on the output is increasing with time, however, this instability
is not seen in the frequency simulation. This seems to be a limitation of the
HSPICE AC simulation.
Figure 7.21: Instability due to lack of Miller Capacitors
The frequency domain representation of the transient signal was calcu-
lated from the 3 msec 4 msec, using a rectangular data window. The
noise can be seen clearly at approximately 100KHz. Once the Miller Ca-
pacitors are in place (with a value of 0.1 pF) this harmonic is completely
removed.
7.4.2 The Final Schematic
The nal Schematic for the Op-Amp design is in gure 7.22.
Figure 7.22: Final Op-Amp Schematic
Chapter 8
Closing Remarks
8.1 Conclusion
This project was a huge success from its beginning to its tentative com-
pletion. By using a team oriented approach to the design methodology it
was possible to bring together such an intricate project. Many challenges
were incurred and resolved soundly and swiftly through the cooperation and
input of all of the group members. Each member of the group has become
a Mini-expert in their area of their design and since these designs all t
together in a larger puzzle knowledge of the other areas has also been gained
my each group member.
122
CHAPTER 8. CLOSING REMARKS 123
8.2 Future Work
Due to the time limitations of this course, (5 Months), only the design
methodology was covered in this tutorial. The next step in the realization
of this Operational Amplier is its layout and fabrication. Layout being the
placing of the actual transistors down to the metal and poly layers. This
next step will have its own design methodology and its own set of challenges,
including troubleshooting, simulating and software testing.
Once the layout step has been completed the next step would be to have
the Operational Amplier fabricated. The fabrication can take several weeks
to several months once it has been sent to the fabricator. This time delay
is a result of the fabricator waiting to get enough designs to fabricate them
on a whole wafer of silicon, to fabricate (10) Operational Ampliers on a
whole piece of silicon would not be cost ecient. Thus, once the fabricator
completes this step the fabricated chips would be set back and would go to
the next stage which is actual hardware testing of the Op-Amp.
The hardware testing of the Operational Amplier would involve phys-
ically connecting the Op-Amp chip to a power supply and putting in an
input signal and then viewing its output signal on an oscilloscope. Here the
DC and AC tests could be run as well as the protection of the circuit tested.
Chapter 9
Bibliography
1. A.S. Sedra And K.C. Smith, Microelectronic Circuits - 4
th
Edi-
tion, Oxford University Press, New York, 1998 ISBN: 0-19-511663-1,
Pages 221-874
2. Allan R. Hambley, Electronics - 2
nd
Edition, Prentice-Hall, Upper
Saddle River, New Jersey, 2000, ISBN: 0-13-691982-0, Pages 211-328
3. Wai-Kai Chen, The VLSI Handbook, CRC Press, Boca Raton,
Florida, 1999, ISBN: 0849385938
4. Ashok Ambardar, Analog and Digital Signal Processing, Brooks-
Cole Publishing Company, Pacic Grove, California, ISBN: 0-534-
95409-X
5. James W. Milsson And Susan A. Riedel, Electric Circuits - 5
th
Edi-
tion, Addison-Wesley Publishing Company, Reading, Massachusetts,
1996, ISBN: 0-201-55707-X, Pages 177-211
124

Cadence Op-Amp Schematic Design Tutorial For TSMC CMOSP35

Uploaded by

Copyright:

Available Formats

Cadence Op-Amp Schematic Design Tutorial For TSMC CMOSP35

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Cadence Op-Amp Schematic Design Tutorial For TSMC CMOSP35

Uploaded by

Copyright:

Available Formats

Cadence Op-Amp Schematic Design Tutorial for

is directly proportional to the

For this circuit: V

You might also like