Академический Документы
Профессиональный Документы
Культура Документы
Page 1.0-1
Page 1.0-2
Chapter 10
D/A and A/D
Converters
Systems
Chapter 6
Simple CMOS &
BiCMOS OTA's
Chapter 7
High Performance
OTA's
Chapter 8
CMOS/BiCMOS
Comparators
Complex
Simple
Chapter 4
CMOS
Subcircuits
Chapter 5
CMOS
Amplifiers
Chapter
Chapter10
2
CMOS/BiCMOS
D/A and A/D
Technology
Converters
Chapter
Chapter11
3
CMOS/BiCMOS
Analog
Modeling
Systems
Circuits
Devices
Introduction
CMOS Analog Circuit Design
Fig. 1.0-01
Page 1.1-1
Fig. 1-1
Page 1.1-2
Electrical
Design
Comparison
with design
specifications
Implementation
Comparison
with design
specifications
Simulation
Physical Definition
Physical
Design
Physical Verification
Parasitic Extraction
Fabrication
Fabrication
Testing and
Product
Development
Product
Fig. 1.1-2
Page 1.1-3
;;
L
W/L ratios
Circuit or
systems
specifications
Analog
Integrated
Circuit Design
M3
vin
+
VDD
M6
M4
M1
Cc
vout
CL
M2
+
VBias
-
M7
M5
VSS
Topology
DC Currents
Fig. 1.1-3
Physical Aspects
Implementation of the physical design including:
- Transistors and passive components
- Connections between the above
- Busses for power and clock distribution
- External connections
Testing Aspects
Design and implementation for the experimental verification of the circuit after
fabrication
CMOS Analog Circuit Design
Page 1.1-4
Digital Circuits
Signal are discontinuous in
amplitude and time - binary
signals have two amplitude states
Designed at the systems level
Component have fixed values
Standard
CAD tools have been extremely
successful
Timing models only
Programmable by software
Regular blocks
Easy to route automatically
Dynamic range unlimited
Page 1.1-5
Page 1.2-1
3.0V
2.5V
2.0V
Desktop Systems
1.5V
1.0V
Portable Systems
1995
1998
2001
2004
Year
2007
2010
Fig. 1.2-1
P.E. Allen - 2004
Page 1.2-2
2005-2006
10
5
2
Analog
Headroom
1
0.5
VDD
VT (scenario 2)
0.2
VT (scenario 1)
0.1
0.01
0.02
0.05
0.2
0.1
0.5
MOSFET Channel Length, m
1
Fig. 1.2-2
Page 1.2-3
Trends in IC Technology
Technology Speed Figure of Merit vs. Time:
ft
HEMTs, HBTs
300GHz
SiGe
100GHz
30GHz
10GHz
3GHz
GaAs
Bipolar
1GHz
1m
0.25m
0.35m
0.5m
0.8m 0.6m
1.5m
2m
3m
Carrier Frequency of RF
Cellular Telephony
77 79 81 83 85 87 89 91 93 95 97 99
CMOS
0.09m
0.13m
0.18m
01 03
05
Year
Fig. 1.2-3B
ft
0.35 micron
0.25 micron
0.18 micron
25GHz
40GHz
60GHz
fmax
40GHz
60-70GHz
90-100GHz
P.E. Allen - 2004
Page 1.2-4
Actual
?
1950
1960
Discrete
Transistors
1970
Bipolar
Analog IC
1980
1990
MOS
Analog IC
2000
Fig. 1.2-4
Page 1.2-5
Innovative
Solution
Generic
Function
Application driven circuit innovation:
Standard
Technology
Innovative
Solution
New
Application
CMOS Analog Circuit Design
Fig. 1.2-5
P.E. Allen - 2004
Page 1.2-6
Page 1.3-1
Video
RF
Acoustic
Imaging
Seismic
Sonar
Microwave
Radar
Audio
Optical
AM-FM radio, TV
Telecommunications
10
100
1k
10k
100k
1M
10M
Signal Frequency (Hz)
100M
1G
10G
100G
Fig. 1.3-1
Page 1.3-2
;
;
;
;
;
;
;
;
;
;
Mostly analog
implementation
Fuzzy boundary,
keeps moving to
the right
Surface acoustic
waves
MOS digital logic
MOS analog
10
100
1k
10k
100k
1M
10M
Signal Frequency (Hz)
100M
Optical
GaAs
1G
10G
100G
Fig. 1.3-2
P.E. Allen - 2004
Page 1.3-3
Page 1.3-4
Physical
Sensors
Actuators
Transmission
Media
VLSI
DIGITAL
SYSTEM
Imagers &
Displays
Audio
I/O
Storage
Media
Analog/Digital
Interface
Electronics
Fig. 1.3-3
Page 1.3-5
A/D
DSP
System
Trend:
ASP
A/D
DSP
System
Fig. 1.3-4
Page 1.3-6
Page 1.3-7
Page 1.4-1
Drain Current
Signal Definition
Quantity
Subscript Example
Total instantaneous value of the signal Lowercase Uppercase
qA
DC value of the signal
Uppercase Uppercase
QA
AC value of the signal
Lowercase Lowercase
qa
Complex variable, phasor, or rms value Uppercase Lowercase
Qa
of the signal
Example:
Idm
id
ID
iD
t
Fig. 1.4-1
Page 1.4-2
Enhancement
PMOS with
VBS = 0V.
S
Enhancement
B NMOS with
VBS 0V.
Enhancement
B PMOS with
VBS 0V.
S
Simple
NMOS
symbol
S
CMOS Analog Circuit Design
Simple
PMOS
symbol
D
P.E. Allen - 2004
Page 1.4-3
Differential amplifier,
op amp, or comparator
+
A vV1 -
V1
I1
V
-
Independent
current source
I2
Independent
voltage sources
+
V2
V1
Voltage-controlled,
voltage source
GmV1
I1
Voltage-controlled,
current source
I2
+
RmI1 +-
Ai I1
V2
-
Current-controlled,
voltage source
Current-controlled,
current source
Page 1.4-4
QABC
A = Terminal with the larger magnitude of potential
B = Terminal with the smaller magnitude of potential
C = Condition of the remaining terminal with respect to terminal B
C = 0 There is an infinite resistance between terminal B and the 3rd terminal
C = S There is a zero resistance between terminal B and the 3rd terminal
C = R There is a finite resistance between terminal B and the 3rd terminal
C = X There is a voltage source in series with a resistor between terminal B
and the 3rd terminal in such a manner as to reverse bias a PN junction.
Examples
I DSS
S
VGS
CDGS
+
G
(a.)
IDS
S
(b.)
BVDGO
G
(c.)
(a.) Capacitance from drain to gate with the source shorted to the gate.
(b.) Drain-source current when gate is shorted to source (depletion device)
(c.) Breakdown voltage from drain to gate with the source is open- circuited to the gate.
CMOS Analog Circuit Design
Page 1.5-1
1.5 - SUMMARY
Analog IC design combines a function or application with IC technology for a successful
solution.
Analog IC design consists of three major steps:
1.) Electrical design Topology, W/L values, and dc currents
2.) Physical design (Layout)
3.) Test design (Testing)
Analog designers must be flexible and have a skill set that allows one to simplify and
understand a complex problem
Analog IC design is driven by improving technologies rather than new technologies.
Analog IC design has reached maturity and is here to stay.
The appropriate philosophy is If it can be done economically by digital, dont use
analog.
As a result of the above, analog finds applications where speed, area, or power have
advantages over a digital approach.
Deep-submicron technologies will offer severe challenges to the creativity of the analog
designer.
Page 2.0-1
CMOS
Technology
and
Fabrication
CMOS
Transistor and Passive
Component
Modeling
Fig. 2.0-1
Page 2.0-2
Bipolar
Junction
Isolated
Dielectric
Isolated
SiliconGermanium
Fig. 150-01
Bipolar/CMOS
Oxide
isolated
Silicon
CMOS
Aluminum
gate
MOS
PMOS
(Aluminum
Gate)
Silicon
gate
NMOS
Aluminum
gate
Silicon
gate
Page 2.0-3
BJT
MOSFET
100 GHz
50 GHz (0.25m)
Less 1/f
More 1/f
9 decades of exponential 2-3 decades of square law
current versus vBE
behavior
Slightly larger
Smaller for short channel
Poor
Good
Voltage dependent
Reasonably good
Page 2.0-4
n+
n+
Source/drain
extensions
Deep p-well
STI
p+
p+
Source/drain
STI
extensions
Deep n-well
p-substrate
031211-02
Page 2.0-5
70
60
NMOS
Slow, 70C
fT (GHz)
50
Typical, 25C
40
30
Slow, 70C
PMOS
20
10
0
100
200
300
|VGS-VT| (mV)
400
500
030901-07
The upper frequency limit is probably around 40 GHz for NMOS with an fT in the vicinity
of 60GHz with an overdrive of 0.5V and at the slow-high temperature corner.
Page 2.1-1
0.5-0.8mm
Fig. 2.1-1r
P.E. Allen - 2004
Page 2.1-2
Oxidation
Description:
Oxidation is the process by which a layer of silicon dioxide is grown on the surface of a
silicon wafer.
Original silicon surface
tox
Silicon dioxide
0.44 tox
Silicon substrate
Fig. 2.1-2
Uses:
Protect the underlying material from contamination
Provide isolation between two layers.
Very thin oxides (100 to 1000) are grown using dry oxidation techniques. Thicker
oxides (>1000) are grown using wet oxidation techniques.
Page 2.1-3
Diffusion
Diffusion is the movement of impurity atoms at the surface of the silicon into the bulk of
the silicon.
Always in the direction from higher concentration to lower concentration.
Low
Concentration
High
Concentration
Fig. 150-04
Gaussian
ERFC
N(x)
N0
t1 < t2 < t3
N(x)
t1 < t2 < t3
NB
NB
t1
t2
t3
Depth (x)
Infinite source of impurities at the surface.
t1
t2
t3
Depth (x)
Finite source of impurities at the surface.
Fig. 150-05
Page 2.1-4
Ion Implantation
Ion implantation is the process by which
impurity ions are accelerated to a high
velocity and physically lodged into the
target material.
Path of
impurity
atom
Fixed Atom
Fixed Atom
Fixed Atom
Fig. 150-06
Concentration peak
Depth (x)
Fig. 150-07
Deposition
Deposition is the means by which various materials are deposited on the silicon wafer.
Examples:
Silicon nitride (Si3N4)
Silicon dioxide (SiO2)
Aluminum
Polysilicon
There are various ways to deposit a material on a substrate:
Chemical-vapor deposition (CVD)
Low-pressure chemical-vapor deposition (LPCVD)
Plasma-assisted chemical-vapor deposition (PECVD)
Sputter deposition
Material that is being deposited using these techniques covers the entire wafer.
Page 2.1-6
Etching
Mask
Film
Underlying layer
(a) Portion of the top layer ready for etching.
a Selectivity
Mask
Film
c
b
Selectivity
Anisotropy
Underlying layer
Important considerations:
(b) Horizontal etching and etching of underlying layer.
Fig. 150-08
Anisotropy of the etch is defined as,
A = 1-(lateral etch rate/vertical etch rate)
Selectivity of the etch (film to mask and film to substrate) is defined as,
film etch rate
Sfilm-mask = mask etch rate
A = 1 and Sfilm-mask = are desired.
There are basically two types of etches:
Wet etch which uses chemicals
Dry etch which uses chemically active ionized gases.
CMOS Analog Circuit Design
Page 2.1-7
Epitaxy
Epitaxial growth consists of the formation of a layer of single-crystal silicon on the
surface of the silicon material so that the crystal structure of the silicon is continuous
across the interfaces.
It is done externally to the material as opposed to diffusion which is internal
The epitaxial layer (epi) can be doped differently, even oppositely, of the material on
which it grown
It accomplished at high temperatures using a chemical reaction at the surface
The epi layer can be any thickness, typically 1-20 microns
Gaseous cloud containing SiCL4 or SiH4
Si +
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
- Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Si
Fig. 150-09
Page 2.1-8
Photolithography
Components
Photoresist material
Mask
Material to be patterned (e.g., oxide)
Positive photoresist
Areas exposed to UV light are soluble in the developer
Negative photoresist
Areas not exposed to UV light are soluble in the developer
Steps
1. Apply photoresist
2. Soft bake (drives off solvents in the photoresist)
3. Expose the photoresist to UV light through a mask
4. Develop (remove unwanted photoresist using solvents)
5. Hard bake ( 100C)
6. Remove photoresist (solvents)
Page 2.1-9
Photoresist
Polysilicon
Fig. 150-10
Page 2.1-10
Develop
Polysilicon
Photoresist
Etch
Photoresist
Polysilicon
Remove
photoresist
Polysilicon
Fig. 150-11
Page 2.1-11
Underlying Layer
Photoresist
SiO2
Underlying Layer
SiO2
SiO2
Underlying Layer
CMOS Analog Circuit Design
Fig. 150-12
Page 2.1-12
Page 2.1-13
yy
;;
Gate Ox
Oxide
Substrate
p+
p-
n-
n+
Poly
Salicide Polycide
Metal
031231-13
Page 2.1-14
yy
;;
Gate Ox
Oxide
n-well
p-well
Substrate
p+
p-
n-
n+
Poly
Salicide Polycide
Metal
031231-12
Page 2.1-15
Shallow
Trench
Isolation
yy
;;
Gate Ox
Oxide
Shallow
Trench
Isolation
n-well
Shallow
Trench
Isolation
p-well
Substrate
p+
p-
n-
n+
Poly
Salicide Polycide
Metal
031231-11
Page 2.1-16
p threshold implant
p threshold implant
Shallow
Trench
Isolation
yy
;;
Gate Ox
Oxide
Shallow
Trench
Isolation
n-well
Shallow
Trench
Isolation
p-well
Substrate
p+
p-
n-
n+
Poly
Salicide Polycide
Metal
031231-10
Page 2.1-17
; ;
Thin Oxide
Shallow
Trench
Isolation
yy
;;
Gate Ox
Oxide
Shallow
Trench
Isolation
n-well
Shallow
Trench
Isolation
p-well
Substrate
p+
p-
n-
n+
Poly
Salicide Polycide
Metal
031231-09
Page 2.1-18
; ;
Shallow pImplant
Shallow
Trench
Isolation
yy
;;
Gate Ox
Oxide
Shallow pImplant
Shallow nImplant
Shallow nImplant
Shallow
Trench
Isolation
n-well
Shallow
Trench
Isolation
p-well
Substrate
p+
p-
n-
n+
Poly
Salicide Polycide
Metal
031231-08
Page 2.1-19
; ;
Sidewall
Spacers
Shallow
Trench
Isolation
yy
;;
Gate Ox
Oxide
Sidewall
Spacers
Shallow
Trench
Isolation
n-well
Shallow
Trench
Isolation
p-well
Substrate
p+
p-
n-
n+
Poly
Salicide Polycide
Metal
031231-07
Page 2.1-20
; ;
p+
implant
n+
implant
p+
n+
Sidewall
p+ Spacers
p+
implant
implant
yy
;;
Gate Ox
Oxide
n+
implant
p+
implant
n+
n+
p+
p+
p+
Shallow
Trench
Isolation
n+
implant
Shallow
Trench
Isolation
n-well
Shallow
Trench
Isolation
p-well
Substrate
p+
p-
n-
n+
Poly
Salicide Polycide
Metal
031231-06
Page 2.1-21
Step 9 Siliciding
Siliciding and polyciding is used to reduce interconnect resistivity by placing a lowresistance silicide such as TiSi2, WSi2, TaSi2, etc. on top of the diffusions.
; ;
Sidewall
Spacers
Salicide
Salicide
p+
n+
yy
;;
Gate Ox
Oxide
Salicide
p+
p+
Shallow
Trench
Isolation
Polycide
Salicide
n+
n+
p+
Shallow
Trench
Isolation
n-well
Shallow
Trench
Isolation
p-well
Substrate
p+
p-
n-
n+
Poly
Salicide Polycide
Metal
031231-05
Page 2.1-22
Intermediate
Oxide
Layer
Salicide
Salicide
p+
n+
yy
;;
Oxide
Polycide
Salicide
p+
p+
Shallow
Trench
Isolation
Gate Ox
; ;
Sidewall
Spacers
Salicide
n+
n+
p+
Shallow
Trench
Isolation
n-well
Shallow
Trench
Isolation
p-well
Substrate
p+
p-
n-
n+
Salicide Polycide
Poly
Metal
031231-04
Page 2.1-23
Tungsten
Plugs
Salicide
p+
Salicide
n+
yy
;;
Oxide
Polycide
Salicide
p+
p+
Shallow
Trench
Isolation
Gate Ox
; ;
Sidewall
Spacers
Tungsten
Plug
Salicide
n+
n+
p+
Shallow
Trench
Isolation
n-well
First
Level
Metal
Shallow
Trench
Isolation
p-well
Substrate
p+
p-
n-
n+
Poly
Salicide Polycide
Metal
031231-03
Page 2.1-24
Intermediate
Oxide
Layers
Tungsten
Plugs
Tungsten
Plugs
Salicide
p+
Salicide
n+
yy
;;
Oxide
Salicide
Tungsten
Plug
Salicide
p+
p+
Shallow
Trench
Isolation
Gate Ox
; ;
Tungsten Plugs
Polycide
Sidewall
Spacers
n+
n+
p+
Shallow
Trench
Isolation
n-well
Second
Level
Metal
First
Level
Metal
Shallow
Trench
Isolation
p-well
Substrate
p+
p-
n-
n+
Poly
Salicide Polycide
Metal
031231-02
Page 2.1-25
Completed Fabrication
After multiple levels of metal are applied, the fabrication is completed with a thicker toplevel metal and a protective layer to hermetically seal the circuit from the environment.
Note that metal is used for the upper level metal vias. The chip is electrically connected
by removing the protective layer over large bonding pads.
; ;
Metal Vias
Intermediate
Oxide
Layers
Tungsten
Plugs
Salicide
p+
Metal Via
Tungsten Plugs
Polycide
Sidewall
Spacers
Tungsten
Plugs
Salicide
n+
Salicide
p+
p+
Shallow
Trench
Isolation
Top
Metal
Tungsten
Plug
Salicide
n+
n+
p+
Shallow
Trench
Isolation
n-well
Second
Level
Metal
First
Level
Metal
Shallow
Trench
Isolation
p-well
Substrate
Gate Ox
Oxide
p+
p-
n-
n+
Poly
Salicide Polycide
Metal
031231-01
Page 2.1-26
Tungsten Plug
TEOS
SOG
Polycide
Sidewall
Spacer
TEOS/BPSG
Poly
Gate
Fig. 2.8-20
CMOS Analog Circuit Design
Page 2.2-1
Metal 3
Aluminum
Vias
Metal 2
Tungsten
Plugs
Metal 1
Transistors
CMOS Analog Circuit Design
Fig.180-11
Page 2.2-2
SUMMARY
Fabrication is the means by which the circuit components, both active and passive, are
built as an integrated circuit.
Basic process steps include:
1.) Oxide growth
2.) Thermal diffusion
3.) Ion implantation
4.) Deposition
5.) Etching
6.) Epitaxy
The complexity of a process can be measured in the terms of the number of masking
steps or masks required to implement the process.
Major CMOS Processing Steps:
1.) p and n wells
2.) Shallow trench isolation
3.) Threshold shift
4.) Thin oxide and gate polysilicon
5.) Lightly doped drains and sources
6.) Sidewall spacer
7.) Heavily doped drains and sources
8.) Siliciding (Salicide and Polycide)
9.) Bottom metal, tungsten plugs, and oxide
10.) Higher level metals, tungsten plugs/vias, and oxide
11.) Top level metal, vias and protective oxide
Page 2.2-3
p-type semiconductor
iD
+vD Depletion
region
n-type
semiconductor
p-type
semiconductor
iD
+ v -D W
-W1
W2
Fig. 06-01
1. Doped atoms near the metallurgical junction lose their free carriers by diffusion.
2. As these fixed atoms lose their free carriers, they build up an electric field, which
opposes the diffusion mechanism.
3. Equilibrium conditions are reached when:
Current due to diffusion = Current due to electric field
CMOS Analog Circuit Design
Page 2.2-4
0
xd
xp
n-type
semiconductor
p-type
semiconductor
iD
-NA
xn
+ v-D -
qND
-W1
0
Fig. 06-02
W2
-qNA
iD
+v Built-in potential, o:
D
iD
NAND
o = Vt ln n 2 ,
i
+v D
kT
where Vt = q and ni2
is the intrinsic concentration of silicon.
CMOS Analog Circuit Design
Chapter 2 Section 2 (5/02/04)
E0
Potential (V)
Fig. 06-03
x
xd
Fig. 06-04A
dv qNA
E0
5.) The electric field, = - dx = - x + C1
CMOS Analog Circuit Design
Page 2.2-6
qNA x 2
-W1
W2
v = 2 + W1x + C2
xd0
8.) A second boundary condition is obtained by assuming that the potential of the neutral
p-type region is zero. This boundary condition is,
v = 0 for x = -W1
Substituting in the expression above gives,
W12
qNA x 2
v = 2 + W1x + 2
CMOS Analog Circuit Design
Chapter 2 Section 2 (5/02/04)
12.) The depletion region width on the p-side of the pn junction is given as
2(o -vD)
2(o -vD)
and
W
=
W1 =
2
NA
ND
qNA1 + ND
qND1 + NA
Page 2.2-8
2si(o-vD)ND
qNA(NA+ND)
2si(o-vD)NA
qND(NA+ND)
1
N
Depletion capacitancesiA
siA
siA
Cj = d = W 1+W 2 =
2si(o-vD) ND
q(ND+NA) NA +
=A
siqNAND
2(NA+ND)
1
o-vD =
Cj
NA
ND
Cj0
Cj0
vD
1 - o
Fig. 06-05
0 vD
Example 1
An abrupt silicon pn junction has the doping densities of NA = 1015 atoms/cm3 and ND =
1016 atoms/cm3. Calculate the junction built-in potential, the depletion-layer widths, the
maximum field and the depletion capacitance with 10V reverse bias if Cj0 = 3pF.
Solution
At room temperature, kT/q = 26mV and the intrinsic concentration is ni = 1.5x1010 cm-3.
10151016
Therefore, the junction built-in potential is o = 0.026 ln2.25x1026 = 0.637V
4
max = - W 1 =
= -5.38x10 V/cm
1.04x10-12
3pF
The depletion capacitance can be found as Cj =
= 0.734pF
1 + (10/0.637)
CMOS Analog Circuit Design
Page 2.2-10
VD (V)
-1
Breakdown
-2
-3
Fig. 6-06
P.E. Allen - 2004
Page 2.2-11
Example 2
An abrupt pn junction has doping densities of NA = 3x1016 atoms/cm3 and ND = 4x1019
atoms/cm3. Calculate the breakdown voltage if crit = 3x105 V/cm.
Solution
2
si
si(NA+ND) 2
1.04x10-129x1010
VR = 2qNAND Emax 2qNA Emax = 21.6x10-193x1016 = 9.7V
Page 2.2-12
0
-NA
Fig. 6-07
W 1 m
2si(o-vD)NA m
N
W 2 = qND(NA+ND)
m
vD m
o-vD
1 o
Page 2.2-13
vD
iD = I exp Vt - 1
25
20
iD 15
I s 10
5
0
-5
-4
-3
-2
-1
vD/Vt
16
10 x10
16
8x10
16
iD 6x10
Is
16
4x10
2x1016
0
-40
CMOS Analog Circuit Design
-30
-20
-10
0
vD/Vt
10
20
30
40
P.E. Allen - 2004
Page 2.2-14
Metal-Semiconductor Junctions
Ohmic Junctions: A pn junction formed by a highly doped semiconductor and metal.
Energy band diagram
IV Characteristics
I
1
Vacuum Level
;;;;
;;;;
Thermionic
or tunneling
qm
qB
n-type metal
qs
Contact
Resistance
EC
EF
EV
n-type semiconductor
Fig. 2.3-4
;;;;
;;;;
;;;;
qB
n-type metal
Forward Bias
EC
EF
Reverse Bias
Forward Bias
Reverse Bias
EV
n-type semiconductor
Fig. 2.3-5
SUMMARY
Characterized the reverse bias operation of the abrupt pn junction
pn junction has a barrier potential o
Depletion region widths are proportional to N-0.5
The pn junction depletion region acts like a voltage dependent capacitance
Applications of the reverse biased pn junction
Isolate transistors from the material they are built in
Variable capacitors - varactors
Page 2.3-1
n-channel transistor
SiO2
(n+
Substrate tie
dra
in
sou
rce
FOX
FOX
)
(n+
)
(p+
FOX
p+
FOX
n+
sou
rce
Well tie
dra
in
(p+
Polysilicon
FOX
n-well
p- substrate
Fig. 2.4-1
Qb QSS
VT = MS + -2F - Cox + Cox
where
MS = F(substrate) - F(gate)
F = Equilibrium electrostatic potential (Femi potential)
kT
F(PMOS) = q ln(ND/ni) = Vt ln(ND/ni)
kT
F(NMOS) = q ln(ni/NA) = Vt ln(ni/NA)
Qb 2qNAsi(|-2F+vSB|)
QSS = undesired positive charge present between the oxide and the bulk silicon
Rewriting the threshold voltage expression gives,
Qb0 QSS Qb - Qb0
VT = MS -2F - C - C - C
= VT0 + |-2F + vSB| - |-2F|
ox
ox
ox
where
2qsiNA
Qb0 QSS
and
=
VT0 = MS - 2F - Cox - Cox
Cox
CMOS Analog Circuit Design
Page 2.3-3
N-Channel
p-type
P-Channel
n-type
+
+
+
+
+
+
+
Page 2.3-4
The equilibrium electrostatic potential for the n+ polysilicon gate is found from as
41019
F(gate) = 0.0259 ln 1.451010 = 0.563 V
Page 2.3-5
1/2
= 0.577 V1/2
Page 2.3-6
Drain
ne
lW
id
th
,W
Bulk
Ch
an
Polysilicon
p+
Fig.
4.3-4
n+
n+
n-channel
Channel
Length, L
p substrate (bulk)
The threshold voltage for a depletion mode NMOS transistor will be negative (a negative
gate potential is necessary to attract enough holes underneath the gate to cause this
region to invert to p-type material).
;;;
;;;
yyy
Page 2.3-7
VGS
n+
n-channel
n+
Diffusion Current
p-substrate/well
log iD
Diffusion Current
Drift Current
10-6
10-12
VT
VGS
Page 2.4-1
Page 2.4-2
Page 2.4-3
PN Junction Capacitors
Generally made by diffusion into the well.
Anode
;;;
;;;
Cj
n+
Cathode
rD
Cj
p+
Rwj
n-well
Substrate
Cw
VA
p+
n+
Rwj
Rw
Depletion
Region
Anode
VB
Rwj
Cathode
Rs
p- substrate
Layout:
Minimize the distance between the p+ and n+ diffusions.
Two different versions have been tested.
1.) Large islands 9m on a side
2.) Small islands 1.2m on a side
Fig. 2.5-011
n+ diffusion
p+ diffusion
n-well
Fig. 2.5-1A
Page 2.4-4
Cmin
Qmin
120
Qmax
100
3
2.5
Small Islands
Large Islands
80
QAnode
CAnode (pF)
4
3.5
Small Islands
2
1.5
60
40
Large Islands
1
20
0.5
0
0
0
0.5
1
1.5
2
2.5
Cathode Voltage (V)
3.5
0.5
1.5
2
2.5
Cathode Voltage (V)
3.5
Fig2.5-1B
Summary:
Terminal Small Islands (598 1.2m x1.2m)
Large Islands (42 9m x 9m)
Under Test Cmax/Cmin
Qmin
Qmax
Cmax/Cmin
Qmin
Qmax
Anode
1.23
94.5
109
1.32
19
22.6
Cathode
1.21
8.4
9.2
1.29
8.6
9.5
Electrons as majority carriers lead to higher Q because of their higher mobility.
The resistance, Rwj, is reduced in small islands compared with large islands higher Q.
CMOS Analog Circuit Design
Page 2.4-5
Vcontrol
n+
p+
n+
p+
p+
VS
n+
n-well
Vcontrol
VS+
VS-
Vcontrol
n+
p+
p+
p+
p+
n+
VS+
VS-
Vcontrol
n-well
Fig. 2.5-015
An examination of the electric field lines shows that because the symmetry inherent in the
differential configuration, the path to the small-signal ground can be shortened if devices
with opposite polarity alternate.
;;
Page 2.4-6
;;;;
D,S,B
p+
n+
p+
Charge
carrier path
n- well
p- substrate/bulk
Capacitance
Cox
Cox
Weak
Inv.
Accumulation
VSG
Moderate
Inversion
Depletion
Strong
Inversion
Fig. 2.5-012
;;
;;;;
G
p+
p-channel
p+
D,S
VDD
B
n+
p+
n- well
p- substrate/bulk
Capacitance
Cox
B=D=S
Cox
Inversion
Mode MOS
VT shift due
to VBS
0
VSG
Fig. 2.5-013
Page 2.4-8
Simulation Results for Standard and Inversion Mode 0.25m CMOS Varactors
n-well:
Page 2.4-9
-0.65V
CG
0.4
vB
0.2
0.0
-1.5 -1.4 -1.3 -1.2 -1.1 -1.0 -0.9 -0.8 -0.7 -0.6 -0.5
vB (Volts)
Fig. 2.5-3
Cmax/Cmin 4
Interpretation:
Capacitance
Cmax
Cox
VBS = -1.50V
VBS = -1.05V
VBS = -0.65V
Inversion
Mode MOS
Cmin
0
CMOS Analog Circuit Design
0.6
Volts or pF
1.0
CG
0.65V
VSG
Fig. 2.5-34
P.E. Allen - 2004
;;;
Page 2.4-10
;;;
;;;;;;;;
;;;
;;;;;;;;
Bulk Rsj Cj
p+
Cov
n+
p- substrate/bulk
Rd
D,S
Cov
Cox
Csi Cd
Cd
Rsi
D,S
Rd
n+
n- LDD
Fig. 2.5-2
Cmax
4.5
34
VG = 2.1V
3.5
VG = 2.1V
32
QGate
CGate (pF)
Qmin
Qmax
38
36
VG = 1.8V
2.5
VG = 1.5V
26
VG = 1.5V
VG = 1.8V
30
28
24
22
1.5
0
0.5
1
1.5
2
2.5
Drain/Source Voltage (V)
3.5
0.5
1
1.5
2
2.5
Drain/Source Voltage (V)
3.5
Fig. 2.5-1c
VG =1.8V: Cmax/Cmin ratio = 2.15 (1.91), Qmax = 34.3 (5.4), and Qmin = 25.8(4.9)
CMOS Analog Circuit Design
Chapter 2 Section 4 (5/02/04)
;;
Page 2.4-11
;;;;
G
p+
n+
n+
n- well
p- substrate/bulk
Capacitance
Cox
B=D=S
Cox
Accumulation
Mode MOS
0
VSG
Fig. 2.5-014
Page 2.4-12
;;;
Bulk
p+
Cov
Cw
Rs
n+
Rw
;;;
;;;;
Rd
D,S
Cov
Cox
Rd
Cd
Cd
n- LDD
D,S
B
n+
n- well
p- substrate/bulk
Fig. 2.5-5
Cmin
VG = 0.9V
3.2
VG = 0.6V
VG = 0.6V
35
2.8
VG = 0.9V
30
VG = 0.3V
2.4
VG = 0.3V
40
QGate
CGate (pF)
3.6
Qmin
Qmax
45
25
0
0.5
1
1.5
2
2.5
Drain/Source Voltage (V)
3.5
0.5
1
1.5
2
2.5
Drain/Source Voltage (V)
3.5
Fig. 2.5-6
VG = 0.6V: Cmax/Cmin ratio = 1.69 (1.61), Qmax = 38.3 (15.0), and Qmin = 33.2(13.6)
CMOS Analog Circuit Design
Page 2.4-13
Differential Varactors
Vcontrol
Vcontrol
A
B
Diode Varactor
Vcontrol
B
VDD
Inversion-PMOS Varactor
Accumulation-PMOS Varactor
Fig. 040-01
Varactor
Diode
I-MOS
A-MOS
fL fH
fC
Tuning
Range
(GHz) (GHz)
1.731.83 10.9%
1.93
1.711.81 11.0%
1.91
1.701.80 10.6%
1.89
P. Andreani and S. Mattisson, On the Use of MOS Varactors in RF VCOs, IEEE J. of Solid-State Circuits, Vol. 35, No. 6, June 2000, pp. 905910.
CMOS Analog Circuit Design
P.E. Allen - 2004
Page 2.4-14
VSB1
M2
VSB2
Fig. 040-02
T. Tille, J. Sauerbrey and D. Schmitt-Landsiedel, A 1.8V MOSFET-Only Modulator Using Substrate Biased Depletion-Mode MOS Capacitors
in Series Compensation, IEEE J. of Solid-State Circuits, Vol. 36, No. 7, July 2001, pp. 1041-1047.
CMOS Analog Circuit Design
P.E. Allen - 2004
Chapter 2 Section 4 (5/02/04)
Page 2.4-15
M1
VS/D
M2
B
Keep the S/D at the lowest
potential to avoid forward
biasing the bulk-source.
Fig. 040-03
Page 2.4-16
B
IOX
IOX
Polysilicon II
IOX
Polysilicon I
FOX
FOX
substrate
Top Plate
Bottom Plate
Bottom Plate
Page 2.4-17
M1
Poly
T
M3
T
M2
M1
M2
B
B
M1
Poly
M2
M1
Fig. 2.5-8
Page 2.4-18
Metal
Metal 3
Metal 2
Metal 1
Side view:
Fig2.5-9
These capacitors are sometimes called fractal capacitors because the fractal patterns are
structures that enclose a finite area with a near-infinite perimeter.
The capacitor/area can be increased by a factor of 10 over vertical flux capacitors.
CMOS Analog Circuit Design
Page 2.4-19
030909-01
Lateral View
030909-02
Top View
R. Aparicio and A. Hajimiri, Capacity Limits and Matching Properties of Integrated Capacitors, IEEE J. of Solid-State Circuits, vol. 37, no. 3,
March 2002, pp. 384-393.
CMOS Analog Circuit Design
P.E. Allen - 2004
Page 2.4-20
Vias
030909-03
Vias
030909-04
Lateral View
Top View
Page 2.4-21
Cap. Density
(aF/m2)
fres.
Q @ Rs () Breakdown (V)
(GHz) 1 GHz
VPP
158.3
18.99
103
0.0054 3.65
14.5
0.57
355
PW
101.5
33.5
315
0.0094 1.1
8.6
0.55
380
HPP
35.8
6.94
427
0.0615 6.0
21
1.1
690
Experimental results for a digital CMOS process with 7 layers of metal, Lmin =0.24m, tox
= 0.7m and tmetal = 0.53m for the bottom 5 layers of metal. All capacitors = 1pF.
Structure Cap. Density Caver. Area Cap.
(aF/m2)
(pF) (m2) Enhanc
(1 pF)
ement
VPP
1512.2
1.01 670
7.4
VB
1281.3
1.07 839.7
6.3
HPP
203.6
1.09 5378
1.0
MIM
1100
1.05 960.9
5.4
CMOS Analog Circuit Design
Std.
Dev.
(fF)
5.06
14.19
26.11
-
Q @ Breakfres.
Caver. (GHz) 1 GHz down
(V)
0.0050 >40
83.2
128
0.0132 37.1 48.7
124
0.0239 21
63.8
500
11
95
P.E. Allen - 2004
Page 2.4-22
12
HPP
VPP
PW
10
8
6
4
2
0
94
96
98
100
102
104
106
030909-05
Caver
Experimental results for a digital CMOS process with 7 layers of metal, Lmin =0.24m, tox
= 0.7m and tmetal = 0.53m for the bottom 5 layers of metal. All capacitors = 10pF.
Structure Cap. Density Caver. Area Cap.
(aF/m2)
(pF) (m2) Enhanc
(10 pF)
ement
VPP
1480.0
11.46 7749
8.0
VB
1223.2
10.60 8666
6.6
HPP
183.6
10.21 55615
1.0
MIM
1100
10.13 9216
6.0
CMOS Analog Circuit Design
Chapter 2 Section 4 (5/02/04)
Std.
Dev.
(fF)
73.43
73.21
182.1
-
Q @ Breakfres.
Caver. (GHz) 1 GHz down
(V)
0.0064 11.3 26.6
125
0.0069 11.1 17.8
121
0.0178 6.17 23.5
495
4.05 25.6
P.E. Allen - 2004
Page 2.4-23
Capacitor Errors
1.) Oxide gradients
2.) Edge effects
3.) Parasitics
4.) Voltage dependence
5.) Temperature dependence
Page 2.4-24
A2
A1
A2
x1
x2
x1
No common centroid
layout
Common centroid
layout
0.2% matching of poly resistors was achieved using an array of 50 unit resistors.
CMOS Analog Circuit Design
Page 2.4-25
C
A
C
A
Page 2.4-26
C1 C1 C1 C1
C1 C1
C2
C1 C1
C2 +
1
1 C1
If
C2 C1
C2 C2
=
,
then
C2
C1
C1 = C1
Therefore, the best matching results are obtained when the area/periphery ratio of C2 is
equal to the area/periphery ratio of C1.
CMOS Analog Circuit Design
Page 2.4-27
Relative Accuracy
0.01
Unit Capacitance = 4pF
0.00
4
8
16
Ratio of Capacitors
32
64
Page 2.4-28
Desired
Capacitor
Bottom
plate
parasitic
Bottom Plate
Page 2.4-29
A'
B'
A'
B'
B
Sensitive to edge varation in
upper plate only.
Fig. 2.6-13
Top Plate
of Capacitor
Bottom plate
of capacitor
Fig. 2.6-14
CMOS Analog Circuit Design
Page 2.4-30
Page 2.4-32
RESISTORS
MOS Resistors - Source/Drain Resistor
Metal
p+
SiO2
FOX
FOX
n- well
p- substrate
Fig. 2.5-16
Diffusion:
Ion Implanted:
10-100 ohms/square
500-2000 ohms/square
Absolute accuracy = 35%
Absolute accuracy = 15%
Relative accuracy=2% (5m), 0.2% (50m) Relative accuracy=2% (5m), 0.15% (50m
Temperature coefficient = +1500 ppm/C
Temperature coefficient = +400 ppm/C
Voltage coefficient 200 ppm/V
Voltage coefficient 800 ppm/V
Comments:
Parasitic capacitance to substrate is voltage dependent.
Piezoresistance effects occur due to chip strain from mounting.
CMOS Analog Circuit Design
Page 2.4-33
Polysilicon Resistor
Metal
;;;;;
Polysilicon resistor
;;
FOX
p- substrate
Fig. 2.5-17
Page 2.4-34
N-well Resistor
Metal
n+
FOX
FOX
FOX
n- well
p- substrate
Fig. 2.5-18
1000-5000 ohms/square
Absolute accuracy = 40%
Relative accuracy 5%
Temperature coefficient = 4000 ppm/C
Voltage coefficient is large 8000 ppm/V
Comments:
Good when large values of resistance are needed.
Parasitics are large and resistance is voltage dependent
Page 2.4-35
Page 2.4-36
Page 2.4-37
INDUCTORS
Inductors
What is the range of values for on-chip inductors?
;;;;;;
12
Inductance (nH)
10
8
6
L = 50
;;;;;;
4
Interconnect parasitics
are too large
0 0
10
20
30
40
Frequency (GHz)
50
Fig. 6-5
Page 2.4-38
Fig.6-6
Page 2.4-39
C1
C2
R1
R2
Fig. 16-7
Design Parameters:
Inductance,L = (Lself + Lmutual)
L
Quality factor, Q = R
1
Self-resonant frequency: fself =
LC
Trade-off exists between the Q and self-resonant frequency
Typical values are L = 1-8nH and Q = 3-6 at 2GHz
CMOS Analog Circuit Design
Page 2.4-40
;;;;;
SiO2
ID
Silicon
I
Nturns = 2.5
Fig. 6-9
Typically: 3 < Nturns < 5 and S = Smin for the given current
Select the OD, Nturns, and W so that ID allows sufficient magnetic flux to flow
through the center.
Loss Mechanisms:
Skin effect
Page 2.4-41
C1
C2
R1
R2
CLoad
Fig. 12.2-13
where:
L is the desired inductance
R is the series resistance
C1 and C2 are the capacitance from the inductor to the ground plane
R1 and R2 are the eddy current losses in the silicon
Guidelines for using spiral inductors on chip:
Lossy substrate degrades Q at frequencies close to fself
To achieve an inductor, one must select frequencies less than fself
The Q of the capacitors associated with the inductor should be very high
CMOS Analog Circuit Design
Page 2.4-42
Fig. 2.5-12
Q = 5-6, fSR = 30-40GHz. Q = 10-11, fSR = 15-30GHz1. Good for high L in small area.
1
The skin effect and substrate loss appear to be the limiting factor at higher frequencies of self-resonance.
CMOS Analog Circuit Design
Page 2.4-44
Inductors - Continued
Self-resonance as a function of inductance. Outer dimension of inductors.
Page 2.4-45
Solenoid Inductors
Example:
Upper Metal
ent
;;
;;;;;;;;
;;
Coil
Coil
Cur
Contact
Vias
rent
Curr
Magnetic Flux
Lower Metal
SiO2
Silicon
Fig. 6-11
Comments:
Magnetic flux is small due to planar structure
Capacitive coupling to substrate is still present
Potentially best with a ferromagnetic core
Page 2.4-46
Transformers
Transformer structures are easily obtained using stacked inductors as shown below for a
1:2 transformer.
Method of reducing the
inter-winding capacitances.
8 turns
Transformers Continued
A 1:4 transformer:
Structure-
3 turns
Page 2.4-47
Secondary
Page 2.5-1
n+
Base
Emitter
Collector
p+
n+
n+
p-well
n-substrate
Page 2.5-2
n+
Base
n+
p+
n+
p-well
n-substrate
Page 2.5-3
n-well
p-diffusion
contact
p-substrate
diffusion
Base
n-well
contact
Lateral
Collector
Emitter
31.2
m
71.4
m
Base
Gate
Lateral
Collector
V SS
Emitter
V SS
Gate
(poly)
84.0 m
33.0 m
Page 2.5-4
Base
Lateral
Collector
Vertical
Collector
( V SS )
V CE =
4.0 V
150
1.0
Lateral Efficiency
130
Lateral
VCE =
4.0 V
0.8
110
V CE =
0.4 V
90
VCE =
0. 4V
0.6
0.4
0.2
70
50
1 nA
0
1 nA
10 nA
100 nA
1 A
10 A
Lateral Collector Current
100 A
1 mA
10 nA
100 nA
1 A
10 A
Emitter Current
100 A
1 mA
Page 2.5-5
0.006 mm2
90
0.70
150
En (midband)
1.92 nV / Hz
3.2 Hz
fc (En)
In @ 5 Hz
2.46 nV / Hz
3.53 pA / Hz
In (midband)
0.61 pA / Hz
162 Hz
85 MHz
16 V
fc (In)
fT
Early voltage
Page 2.5-6
;;
Polysilicon
n+
Source
Channel
n+
p+
n-well
p-substrate
Fig. 190-07
Page 2.5-7
;;
;;
;;
;;
;;
;;
;;;;;;;;;;
;;
;;
;;;;
;; ;;
;;;;
;;
;;;
y;; ;;;
y;;;;
VDD
n+
p+
p+
p-well
n+
VSS
p+
n+
RN-
RP-
n- substrate
Fig. 190-08
VDD
VDD
RNA
Vin VSS
Vout
B
RP-
VSS
VSS
CMOS Analog Circuit Design
Fig. 190-09
Page 2.5-8
p-channel transistor
guard bars
VDD
FOX
p+
n-channel transistor
guard bars
VSS
FOX
FOX
FOX
FOX
p-well
FOX
FOX
n- substrate
Figure 190-10
For more information see R. Troutman, CMOS Latchup, Kluwer Academic Publishers.
Page 2.5-9
p+ resistor
Bonding
Pad
VSS
Implementation in CMOS technology
Metal
FOX
n+
FOX
p+
FOX
n-well
p-substrate
Fig. 190-11
CMOS Analog Circuit Design
Chapter 2 Section 5 (5/02/04)
Page 2.5-11
Noise in Transistors
Shot Noise
i2 = 2qIDf (amperes2)
where
q = charge of an electron
ID = dc value of iD
f = bandwidth in Hz
i2
Noise current spectral density = f (amperes2/Hz)
Thermal Noise
Resistor:
v2 = 4kTRf (volts2)
MOSFET:
8kTgmf
iD 2 =
(ignoring bottom gate)
3
where
k = Boltzmanns constant
R = resistor or equivalent resistor in which the thermal noise is occurring.
gm = transconductance of the MOSFET
CMOS Analog Circuit Design
Page 2.5-12
Noise power
spectral density
1/f
log(f)
CMOS Analog Circuit Design
Fig. 190-12
P.E. Allen - 2004
Page 2.6-1
1.0
1.5
0.5
0.5
0.5
1.5
0.5
0.5
Fig. 2.6-01
Fig. 2.6-02
4.) Minimize the ratio of the perimeter to the area (a circle is optimum).
5.) For parallel plates make one larger than the other to eliminate alignment problems.
CMOS Analog Circuit Design
Page 2.6-2
Total perimeter
is 25 units
Total perimeter
is 36 units
Etch compensation
Total area is
12.5 units
Fig. 2.6-03
Total area is
18 units.
Page 2.6-3
FOX
;;
;;
;;
;;
;;
Polysilicon
gate
L
Active area
drain/source
Fig. 2.6-04
Metal 1
Comments:
Make sure to contact the source and drain with multiple contacts to evenly distribute
the current flow under the gate.
Minimize the area of the source and drain to reduce bulk-source/drain capacitance.
CMOS Analog Circuit Design
Page 2.6-4
;;
;
;;
;
;;
;
;;;
Mirror Symmetry
;;;;
;;
;;;;
;;
Photolithographic Invariance
Fig. 2.6-05
Page 2.6-5
;;;
;;;
;;
;
;;;
Metal 2
Via 1
;;;;;;;;;
;;;;;;;;;
;;;;
;
;;
;;
;
;;;;;;;;
Metal 1
Fig. 2.6-06
Page 2.6-6
;;
;;;
;;
;;
;;;
;;
;;
;;;
;;
;;;;;;;
Via 1
Metal 2
Metal 1
Fig. 2.6-07
Page 2.6-7
Resistor Layout
Direction of current flow
W
T
Area, A
L
Fig. 2.6-15
Page 2.6-8
Metal
FOX
FOX
FOX
Substrate
Active area (diffusion)
FOX
Contact
FOX
Substrate
Well diffusion
Contact
Cut
Cut
L
Metal 1
Metal 1
Well resistor
Fig. 2.6-16
Corner corrections:
0.5
1.45
1.25
Fig. 2.6-16B
Page 2.6-9
s = T = 3000 10-8 cm = 30 /
The number of squares of resistance, N, is
20m
L
N = W = 0.8m = 25
giving the total resistance as
R = s = 30 25 = 750
Page 2.6-10
Capacitor Layout
Double-polysilicon capacitor
Metal
Polysilicon 2
Metal 3
FOX
Substrate
;;;;
;;;;
;;;;
;;;;
;;;;
;;;;
FOX
Substrate
Polysilicon gate
Polysilicon 2
Cut
Metal 2 Metal 1
Polysilicon gate
Metal 3
Via 2
Cut
Metal 2
Via 1
Metal 1
Metal 1
Fig. 2.6-17
Page 2.6-11
Design Rules
Design rules are geometrical constraints that guarantee the proper operation of a circuit
implemented by a given CMOS process.
These rules are necessary to avoid problems such as device misalignment, metal
fracturing, lack of continuity, etc.
Design rules are expressed in terms of minimum dimensions such as minimum values of:
- Widths
- Separations
- Extensions
- Overlaps
Design rules typically use a minimum feature dimension called lambda. Lambda is
usually equal to the minimum channel length.
Minimum resolution of the design rules is typically half lambda.
In most processes, lambda can be scaled or reduced as the process matures.
Page 2.7-1
Page 2.7-2
TOP
VIEW
p substrate
p-
p+
ni
n-
n+
Metal
Fig.2.7-1
Page 2.7-3
TOP
VIEW
n collector
Epitaxial
Region
SIDE
VIEW
n+ buried layer
p substrate
p+
p-
ni
n-
n+
Metal
Fig.2.7-2
Page 2.7-4
TOP
VIEW
n collector
p+
isolation
p+
isolation
p+
isolation
n collector
SIDE
VIEW
n+ buried layer
p substrate
p+
p-
n-
ni
n+
Metal
Fig.2.7-3
Page 2.7-5
TOP
VIEW
n collector
p base
p+
isolation
pisolation
p base
n collector
SIDE
VIEW
n+ buried layer
p substrate
p+
CMOS Analog Circuit Design
p-
ni
n-
n+
Metal
Fig.2.7-4
Page 2.7-6
TOP
VIEW
p+
isolation
n+
p+
isolation
n+ emitter
p base
n collector
SIDE
VIEW
n+ buried layer
p substrate
p+
p-
n-
ni
n+
Metal
Fig.2.7-5
Page 2.7-7
TOP
VIEW
p+
isolation
n+
p+
p+
isolation
n+ emitter
p base
n collector
SIDE
VIEW
n+ buried layer
p substrate
p+
CMOS Analog Circuit Design
p-
ni
n-
n+
Metal
Fig.2.7-6
Page 2.7-8
;;;;;;;;;;;;
;;;;;;;;;;;;
;;;;;;;;;;;;
;;;;;;;;;;;;
;;;;;;;;;;;;
;;;;;;;;;;;;
;;;;;;;;;;;;
;;;;;;;;; ;;;;;
TOP
VIEW
Dielectric Layer
p+
isolation
SIDE
VIEW
n+
p+
p+
isolation
n+ emitter
p base
n collector
n+ buried layer
p substrate
p+
p-
n-
ni
n+
Metal
Fig.2.7-7
Page 2.7-9
TOP
VIEW
;;;;
;;
;;
;;;;
;;;;;;;; ;;;;
p+
isolation
SIDE
VIEW
n+
p+
n+ emitter
p base
p+
isolation
n collector
n+ buried layer
p substrate
Fig.2.7-8
Page 2.7-10
TOP
VIEW
;;;;;;;;;;;;
;;;;;;;; ;;;;
;;;;;;;; ;;;;
Passivation
p+
isolation
SIDE
VIEW
p+
n+
p+
isolation
n+ emitter
p base
n collector
n+ buried layer
p substrate
Fig.2.7-9
Page 2.7-11
n+
1020
Substrate Doping Level
1019
1018
Epitaxial
collector
doping level
1017
1016
1015
1014
1013
1012
Emitter
1021 n+
Base Collector
Buried Layer
10
11
Substrate
12
Page 2.7-12
TOP
VIEW
;;;;
;;;;
;;;;;;
p+
isolation/
collector
p+
isolation/
collector
p+
n+
p emitter
n base
SIDE
VIEW
p collector/substrate
p+
p-
n-
ni
n+
Metal
Fig.2.7-11
Page 2.7-13
TOP
VIEW
;;;;;;;;;;;;;
p+
isolation
SIDE
VIEW
p+
p+
p collector
p emitter
n+
p+
isolation
n base
n+ buried layer
p substrate
p+
p-
ni
n-
n+
Metal
Fig.2.7-12
Page 2.7-14
Page 2.8-1
Page 2.8-2
1m
p-substrate
BiCMOS-01
5m
p+ buried
layer
PMOS Transistor
NMOS Transistor
n+ buried layer
p+ buried layer
1m
p-substrate
BiCMOS-02
5m
Page 2.8-3
Epitaxial Growth
NPN Transistor
n-well
p-well
n+ buried layer
p+ buried
layer
PMOS Transistor
NMOS Transistor
n-well
p-well
n+ buried layer
p-type
Epitaxial
Silicon
p+ buried layer
1m
p-substrate
BiCMOS-03
5m
Comment:
As the epi layer grows vertically, it assumes the doping level of the substrate beneath it.
In addition, the high temperature of the epitaxial process causes the buried layers to
diffuse upward and downward.
CMOS Analog Circuit Design
Page 2.8-4
Collector Tub
NPN Transistor
Original Area of
CollectorTub Implant
Collector Tub
n+ buried layer
PMOS Transistor
p-well
p+ buried
layer
n-well
n+ buried layer
NMOS Transistor
p-well
p-type
Epitaxial
Silicon
p+ buried layer
1m
p-substrate
BiCMOS-04
5m
Comment:
The collector area is developed by an initial implant followed by a drive-in diffusion to
form the collector tub.
Page 2.8-5
PMOS Transistor
NMOS Transistor
Nitride
-Silicon
Collector Tub
n+ buried layer
p-well
p+ buried
layer
n-well
n+ buried layer
p-well
p-type
Epitaxial
Silicon
p+ buried layer
1m
p-substrate
BiCMOS-05
5m
Comment:
The silicon nitride is use to impede the growth of the thick oxide which allows contact
to the substrate
-silicon is used for stress relief and to minimize the birds beak encroachment
CMOS Analog Circuit Design
Page 2.8-6
Field Oxide
FOX
NPN Transistor
FOX
Collector Tub
PMOS Transistor
Field Oxide
p-well
n+ buried layer
p+ buried
layer
NMOS Transistor
Field Oxide
n-well
Field Oxide
p-type
Epitaxial
Silicon
p-well
n+ buried layer
p+ buried layer
1m
p-substrate
BiCMOS-06
5m
Comments:
The field oxide is used to isolate surface structures (i.e. metal) from the substrate
Page 2.8-7
FOX
NPN Transistor
Collector Sink
FOX
Collector Tub
n+ buried layer
PMOS Transistor
Anti-Punch Through
Threshold Adjust
Field Oxide
Field Oxide
Field Oxide
n-well
p-well
p+ buried
layer
NMOS Transistor
Anti-Punch Through
Threshold Adjust
n+ buried layer
p-well
p-type
Epitaxial
Silicon
p+ buried layer
1m
p-substrate
BiCMOS-07
5m
Page 2.8-8
Base Definition
FOX
NPN Transistor
FOX
Collector Tub
PMOS Transistor
Field Oxide
Field Oxide
p+ buried
layer
Field Oxide
n-well
p-well
n+ buried layer
NMOS Transistor
p-well
n+ buried layer
p-type
Epitaxial
Silicon
p+ buried layer
1m
p-substrate
BiCMOS-08
5m
Page 2.8-9
PMOS Transistor
NMOS Transistor
FOX
FOX
Sacrifical Oxide
Field Oxide
Field Oxide
Field Oxide
n-well
p-well
p-well
Sub-Collector
n+ buried layer
p+ buried
layer
n+ buried layer
p-type
Epitaxial
Silicon
p+ buried layer
1m
p-substrate
BiCMOS-09
5m
Page 2.8-10
Emitter Implant
FOX
NPN Transistor
Emitter Implant
FOX
Collector Tub
PMOS Transistor
Field Oxide
NMOS Transistor
Field Oxide
Field Oxide
n-well
p-well
p-well
Sub-Collector
n+ buried layer
p+ buried
layer
n+ buried layer
p-type
Epitaxial
Silicon
p+ buried layer
1m
p-substrate
BiCMOS-10
5m
Comments:
The polysilicon above the base is implanted with n-type carriers
Page 2.8-11
Emitter Diffusion
FOX
FOX
NPN Transistor
PMOS Transistor
Field Oxide
NMOS Transistor
Field Oxide
Field Oxide
n-well
p-well
p-well
Emitter
n+ buried layer
p+ buried
layer
n+ buried layer
p-type
Epitaxial
Silicon
p+ buried layer
1m
p-substrate
BiCMOS-11
5m
Comments:
The polysilicon not over the emitter window is removed and the n-type carriers diffuse
into the base forming the emitter
Page 2.8-12
FOX
NPN Transistor
FOX
PMOS Transistor
Field Oxide
Field Oxide
p+ buried
layer
Field Oxide
n-well
p-well
n+ buried layer
NMOS Transistor
n+ buried layer
p-well
p-type
Epitaxial
Silicon
p+ buried layer
1m
p-substrate
BiCMOS-12
5m
Comments:
The surface of the region where the MOSFETs are to be built is cleared and a thin gate
oxide is deposited with a polysilicon layer on top of the thin oxide
The polysilicon is removed over the source and drain areas
A light source/drain diffusion is done for the NMOS and PMOS (separately)
Page 2.8-13
FOX
FOX
NPN Transistor
PMOS Transistor
Field Oxide
Field Oxide
p+ buried
layer
Field Oxide
n-well
p-well
n+ buried layer
NMOS Transistor
n+ buried layer
p-well
p-type
Epitaxial
Silicon
p+ buried layer
1m
p-substrate
BiCMOS-13
5m
Comments:
The sidewall spacers prevent the heavy source/drain doping from being near the
channel of the MOSFET
Page 2.8-14
Siliciding
FOX
NPN Transistor
Silicide TiSi2
FOX
PMOS Transistor
Silicide TiSi2
Field Oxide
Field Oxide
p+ buried
layer
Field Oxide
n-well
p-well
n+ buried layer
NMOS Transistor
Silicide TiSi2
n+ buried layer
p-well
p-type
Epitaxial
Silicon
p+ buried layer
1m
p-substrate
BiCMOS-14
5m
Comments:
Siliciding is used to reduce the resistance of the polysilicon and to provide ohmic
contacts to the base, emitter, collector, sources and drains
Page 2.8-15
Contacts
Tungsten Plugs
FOX
FOX
Tungsten Plugs
TEOS/BPSG/SOG
Field Oxide
Field Oxide
p+ buried
layer
n+ buried layer
TEOS/BPSG/SOG
Field
FieldOxide
Oxide
n-well
p-well
n+ buried layer
Tungsten Plugs
TEOS/BPSG/SOG
p-well
p-type
Epitaxial
Silicon
p+ buried layer
1m
p-substrate
BiCMOS-15
5m
Comments:
A dielectric is deposited over the entire wafer
One of the purposes of the dielectric is to smooth out the surface
Tungsten plugs are used to make electrical contact between the transistors and metal1
Page 2.8-16
Metal1
FOX
Metal1
FOX
Metal1
TEOS/BPSG/SOG
TEOS/BPSG/SOG
Field Oxide
Field Oxide
p+ buried
layer
n+ buried layer
TEOS/BPSG/SOG
Field
FieldOxide
Oxide
n-well
p-well
n+ buried layer
Metal1
p-well
p-type
Epitaxial
Silicon
p+ buried layer
1m
p-substrate
BiCMOS-16
5m
Page 2.8-17
Metal1-Metal2 Vias
FOX
FOX
Tungsten Plugs
TEOS/BPSG/SOG
TEOS/BPSG/SOG
Field Oxide
Field Oxide
p+ buried
layer
n+ buried layer
TEOS/BPSG/SOG
Field
FieldOxide
Oxide
n-well
p-well
n+ buried layer
Oxide/
SOG/
Oxide
p-well
p-type
Epitaxial
Silicon
p+ buried layer
1m
p-substrate
BiCMOS-17
5m
Page 2.8-18
Metal2
Metal 2
FOX
FOX
Oxide/
SOG/
Oxide
TEOS/BPSG/SOG
TEOS/BPSG/SOG
Field Oxide
Field Oxide
n+ buried layer
p+ buried
layer
Field
FieldOxide
Oxide
n-well
p-well
n+ buried layer
TEOS/BPSG/SOG
p-well
p-type
Epitaxial
Silicon
p+ buried layer
1m
p-substrate
BiCMOS-18
5m
Page 2.8-19
Metal2-Metal3 Vias
TEOS/
BPSG/
SOG
FOX
FOX
Oxide/
SOG/
Oxide
TEOS/BPSG/SOG
TEOS/BPSG/SOG
Field Oxide
Field Oxide
n+ buried layer
p+ buried
layer
Field
FieldOxide
Oxide
n-well
p-well
n+ buried layer
TEOS/BPSG/SOG
p-well
p-type
Epitaxial
Silicon
p+ buried layer
1m
p-substrate
BiCMOS-19
5m
Comments:
The metal2-metal3 vias will be filled with metal3 as opposed to tungsten plugs
CMOS Analog Circuit Design
Page 2.8-20
Completed Wafer
Nitride (Hermetically seals the wafer)
Oxide/SOG/Oxide
Metal3
Metal3
Vias
TEOS/
BPSG/
SOG
FOX
FOX
Oxide/
SOG/
Oxide
TEOS/BPSG/SOG
TEOS/BPSG/SOG
Field Oxide
Field Oxide
n+ buried layer
p+ buried
layer
Field
FieldOxide
Oxide
n-well
p-well
n+ buried layer
TEOS/BPSG/SOG
p-well
p-type
Epitaxial
Silicon
p+ buried layer
1m
p-substrate
BiCMOS-20
5m
P.E. Allen - 2004
Page 2.8-21
SUMMARY
This section has illustrated the major process steps for a 0.5micron BiCMOS
technology.
The performance of the active devices are:
npn bipolar junction transistor:
F = 100-140
BVCEO = 7V
fT = 12GHz,
n-channel FET:
K = 127A/V2
VT = 0.64V
N 0.060
p-channel FET:
VT = -0.63V
P 0.072
K = 34A/V2
Although todays state of the art is 0.25m or 0.18m BiCMOS, the processing steps
illustrated above approximate that which is done in a smaller geometry.
Page 2.9-1
Page 3.0-1
Analog
Integrated
Circuit
Design
CMOS
Transistor and Passive
Component
Modeling
Fig.3.0-1
Page 3.0-2
Comparison of
simulation with
expectations
Thinking Model
Simple,
10% to 50% accuracy
Updating Technology
Extraction of Simple
Model Parameters
from Computer Models
Expectations
"Ballpark"
Computer Simulation
Refined and
optimized
design
Fig.3.0-02
This chapter is devoted to the simple model suitable for design not using simulation.
CMOS Analog Circuit Design
Page 3.0-3
Small-signal, midband
Rin, Av, Rout
(.TF)
Time Dependent
Small-signal frequency
response-poles and zeros
(.AC)
Large-signal transient
response - Slew rate
(.TRAN)
Page 3.1-1
;;;
;;;
Metal-Oxide-Semiconductor Structure
Bulk/Substrate
Source
Gate
Drain
Polysilicon
p+
n+
Thin Oxide
(10-100nm
100-1000)
n+
p- substrate
Heavily
Doped p
Lightly
Doped p
Intrinsic
Doping
Lightly
Doped n
Heavily Metal
Doped n
Fig.3.1-01
Terminals:
Bulk - Used to make an ohmic contact to the substrate
Gate - The gate voltage is applied in such a manner as to invert the doping of the
material directly beneath the gate to form a channel between the source and drain.
Source - Source of the carriers flowing in the channel
Drain - Collects the carriers flowing in the channel
CMOS Analog Circuit Design
Page 3.1-2
;;;
;; ;;;
;; ;;;
;;;
;;;;
;;
;;;
;; ;;;
;;;
;;;
;;;;;;
;;;
Subthreshold (VG<VT)
VB = 0
VS = 0
VG < VT
VD = 0
Polysilicon
p+
n+
p- substrate
Threshold (VG=VT)
VB = 0
n+
Depletion Region
VG =VT
VS = 0
VD = 0
Polysilicon
p+
n+
n+
Inverted Region
p- substrate
VG >VT
VD = 0
Polysilicon
p+
n+
n+
Inverted Region
p- substrate
Fig.3.1-02
Page 3.1-3
;;;
;;
;;;
;; ;;;
;;;
;;;;;;;;;
;; ;;;
;;;
;;;
;;;;;;
;;;
VS = 0
VD = 0.1V
iD
vG =VT
Polysilicon
p+
n+
p- substrate
VGS=2VT:
VB = 0
n+
Depletion Region
VS = 0
VG = 2VT
Polysilicon
p+
VB = 0
0 VT 2VT 3VT
vGS
iD
n+
Inverted Region
VS = 0
VD = 0.1V
iD
n+
p- substrate
VGS=3VT:
iD
VG = 3VT
0
0 VT 2VT 3VT
VD = 0.1V
vGS
iD
Polysilicon
p+
p- substrate
n+
n+
Inverted Region
0
0 VT 2VT 3VT
vGS
Fig.3.1-03
P.E. Allen - 2004
Page 3.1-4
;;;
;;
;;;
;;;;
;; ;;;
;;;
;;
;;;;
;;;
;;
;;;;
;;;
VS = 0
vG =2VT
Polysilicon
p+
n+
VS = 0
VG = 2VT
n+
VB = 0
VD = 0.5VT
iD
;;;
;;;
;;
;;;;
;;;
VS = 0
VG = 2VT
n+
p- substrate
0.5VT
vDS
VT
iD
VGS = 2VT
n+
Polysilicon
p+
Channel current
p- substrate
VDS=VT:
VGS = 2VT
n+
Polysilicon
p+
iD
Inverted Region
p- substrate
VDS=0.5VT:
VB = 0
VD = 0V
iD
0
0
VD =VT
iD
0.5VT
vDS
VT
iD
VGS = 2VT
n+
A depletion region
forms between the drain and channel
0
0
0.5VT
vDS
Fig.3.1-04
VT
Page 3.1-5
;;;
;;
;;;
;; ;;;
;;;
;;
;;;;
;;;
;;;;;;;;;
;;;
;;;
;;;;;;
;;;
VS = 0
vG =VT
Polysilicon
p+
n+
VD = 2VT
iD
n+
p- substrate
VGS=2VT:
VB = 0
VS = 0
VG = 2VT
Polysilicon
p+
n+
VB = 0
VS = 0
p- substrate
n+
0 VT
vDS
2VT 3VT
iD
VGS =2VT
n+
VG = 3VT
Polysilicon
p+
VGS =VT
VD = 2VT
iD
p- substrate
VGS=3VT:
iD
0
0 VT
VD = 2VT
iD
iD
vDS
2VT 3VT
VGS =3VT
n+
Further increase in
VG will cause the FET to become active
0
0 VT
2VT 3VT
vDS
Fig.3.1-05
Page 3.1-6
iD(A)
1500
VGS = 2.5
1000
VGS = 2.0
500
VGS = 1.5
VGS = 1.0
vDS (Volts)
Fig. 3.1-6
Page 3.1-7
VDS = 5V
5000
VDS = 4V
VDS = 3V
iD(A)
4000
3000
VDS = 2V
2000
VDS = 1V
1000
0
0
2
3
vGS (Volts)
5
Fig. 3.1-7
Page 3.2-1
;;;
;;;
vDS
iDdy = - WoQI(y)dv =
vDS
WoCox[vGS-v(y)-VT] dv
iD =
WoCox
vDS2
(v
-V
)v
GS
T
DS
L
2
Page 3.2-2
iD
vDS = vGS-VT
Active Region
Saturation Region
Increasing
values of vGS
vDS
Fig. 110-04
The saturation voltage for MOSFETs is the value of drain-source voltage at the peak of
the inverted parabolas.
vDS
diD oCoxW
[(vGS-VT) - vDS] = 0
L
dvDS =
Cutoff Saturation
vG
S=
0
0
vD
Useful definitions:
oCoxW KW
= L =
L
Active
SV
T
vDS(sat) = vGS - VT
VT
vGS
Fig. 3.2-4
Page 3.2-3
0.75
Channel modulation effects
0.5
0.25
Cutoff Region
0
0
0.5
1.0
1.5
2.0
= 1.0
vGS-VT
= 0.867
VGS0-VT
vGS-VT
= 0.707
VGS0-VT
vGS-VT
= 0.5
VGS0-VT
vGS-VT
=0
VGS0-VT
vDS
VGS0-VT
2.5
Fig. 110-05
Page 3.2-4
Illustration of the Need to Account for the Influence of vD S on the Simple Sah Model
Compare the Simple Sah model to SPICE level 2:
25A
K' = 44.8A/V
k = 0, vDS(sat)
= 1.0V
20A
2
2
K' = 44.8A/V
k=0.5, vDS(sat)
= 1.0V
15A
iD
SPICE Level 2
10A
K' = 29.6A/V 2
k = 0, v (sat)
= 1.0V DS
5A
0A
0
0.2
0.4
0.6
vDS (volts)
0.8
Page 3.2-5
D
0
vD S
vD S
Assume that the threshold voltage varies across the channel in the following way:
VT(y) = VT + kv(y)
where VT is the value of VT the at the source end of the channel and k is a constant.
Integrating the above gives,
WoCox
v2(y)vD S
(v -VT)v(y) - (1+k) 2
iD = L
GS
0
or
WoCox
v2DS
(vGS-VT)vDS - (1+k)
iD = L
2
To find vDS(sat), set the diD/dvDS equal to zero and solve for vDS = vDS(sat),
vGS - VT
vDS(sat) = 1 + k
Therefore, in the saturation region, the drain current is
WoCox
iD = 2(1+k)L (vGS - VT)2
For k = 0.5 and K = 44.8A/V2, excellent correlation is achieved with SPICE 2.
CMOS Analog Circuit Design
Page 3.2-6
;;;
;;;;;;;
VG > VT
VD > VDS(sat)
Depletion
Region
Polysilicon
Illustration:
p+
;;;;;;;
n+
n+
eff
Note that Leff = L - Xd
Therefore the model in saturation
Xd
p- substrate
Fig110-06
becomes,
diD
dL
i dXd
KW
KW
2 eff = D
iD = 2L (vGS-VT)2 dv = (v
V
)
2Leff2 GS T dvDS Leff dvDS iD
eff
DS
Page 3.2-7
iD
iD3(0)
iD2(0)
iD1(0)
VGS3
VGS2
VGS1
vDS
-1
Page 3.2-8
;;;
;;
;;;;
;;;
;;;;;;;;;
;;;
;;;;
;;
;;;
;;
;;;;
;;;
;;;
;;;;;;
;;;
VD > 0
iD
Polysilicon
p+
n+
n+
Channel current
p- substrate
Fig.110-07A
VSB1 > 0V
VSB1>0V:
VS = 0
VG > VT
VD > 0
iD
Polysilicon
p+
n+
n+
Channel current
p- substrate
Fig.110-07B
VSB2 >VSB1 V = 0
S
VG > VT
VD > 0
iD = 0
Polysilicon
p+
;;;;;;;;;
n+
n+
p- substrate
Fig.110-07C
Page 3.2-9
Influence of the Bulk Voltage on the Large Signal MOSFET Model - Continued
Bulk-Source (vBS) influence on the transconductance characteristicsiD
Decreasing values
of bulk-source voltage
VBS = 0
ID
vDS vGS-VT
VT0
VT1
VT2
vGS
Fig. 110-08
VT2
In general, the simple model incorporates the bulk effect into VT by the previously
developed relationship:
VT(vBS) = VT0 + 2|f| + |vBS| - 2|f|
Page 3.2-10
iD
D
+
B
v
Non-saturation+
+ DS
vGS vBS
WoCox
vDS2
- (vGS - VT)vDS iD =
L
2 (1 + vDS)
S Fig. 110-10
SaturationWoCox
vDS(sat)2
WoCox
2
(vGS-VT)vDS(sat) (1+vDS) =
iD =
L
2
2L (vGS-VT) (1+vDS)
where:
o = zero field mobility (cm2/voltsec)
Cox = gate oxide capacitance per unit area (F/cm2)
Page 3.2-11
Silicon Constants
Constant
Symbol
VG
k
ni
0
si
ox
Constant Description
Silicon bandgap (27C)
Boltzmanns constant
Intrinsic carrier
concentration (27C)
Permittivity of free space
Permittivity of silicon
Permittivity of SiO2
Value
Units
1.205
1.381x10-23
1.45x1010
V
J/K
cm-3
8.854x10-14
11.7 0
3.9 0
F/cm
F/cm
F/cm
Page 3.2-12
MOSFET Parameters
Model Parameters for a Typical CMOS Bulk Process (0.8m CMOS n-well):
Parameter
Parameter
Symbol
Description
VT0 Threshold Voltage
(VBS = 0)
Transconductance ParaK'
meter (in saturation)
Bulk threshold
parameter
Channel length
modulation parameter
2|F| Surface potential at
strong inversion
Units
V
110.0 10%
50.0 10%
A/V2
0.4
0.57
(V)1/2
0.04 (L=1 m)
0.01 (L=2 m)
0.7
0.05 (L=1 m)
0.01 (L=2 m)
0.8
(V)-1
V
Page 3.2-13
where,
rG, rS, rB, and rD are ohmic and contact
resistances
and
vBD
iBD = Is exp Vt - 1
rG
CGD
rD
CBD
vBD
+ iBD
vBS
+ -
iD
rB
iBS
CBS
CGS
and
vBS
iBS = Is exp Vt - 1
CGB
rS
S
Fig. 3.2-10
Page 3.3-1
Page 3.3-2
VELOCITY SATURATION
What is Velocity Saturation?
The most important short-channel
effect in MOSFETs is the velocity
105
saturation of carriers in the channel.
A plot of electron drift velocity
5x104
versus electric field is shown below.
2x104
104
5x103
105
106
Electric Field (V/m)
107
Fig130-1
Page 3.3-3
E
JD = JS = W = QI(y)vd(y) iD = WQI(y)vd(y) = 1 + E/Ec iD 1+ Ec = WQI(y)nE
1 dv
dv
iD 1 + Ec dy= WQI(y)n dy
vDS
1 dv
i 1 + Ec dydy = WQI(y)ndv
2[1+(vGS-VT)] L [2(vGS-VT)vDS-vDS2]
1 vDS L
21 + Ec L
Page 3.3-4
Saturation Voltage
Differentiating iD with respect to vDS and setting equal to zero gives,
(VGS-VT)
1
+
VDS(sat) = 1 + 2(VGS-VT -1 (VGS-VT)1 2
or
(VGS-VT)
VDS(sat) VDS(sat) 1 +
2
Note that the transistor will enter the saturation region for vDS < vGS - VT in the
presence of velocity saturation.
Therefore the large signal model in the saturation region is,
(VGS-VT)
K
W
2
GS T
Page 3.3-5
iD/W (A/m)
800
= 0.4
600
= 0.6
400
= 0.8
= 1.0
200
0
0.5
1.5
2
vGS (V)
2.5
3
Fig130-2
Note as the velocity saturation effect becomes stronger, that the drain current-gate
voltage relationship becomes linear.
Page 3.3-6
D
G
iD
+
+
vGS' RSX
vGS
Fig130-3
- S
Page 3.3-7
700
600
NFET VGS=1.8V
Leff =
0.08m
PFET
Leff =
0.11m
VGS=1.4V
500
400
VGS=-1.8V
300
VGS=-1.4V
200
100
VGS=1.0V
VGS=-1.0V
VGS=0.6V
VGS=-0.6V
0
-1.8
-1.2
-0.6
0.6
0.0
Drain Voltage (V)
1.2
1.8
Fig130-4
Su, L., et.al., A High Performance Sub-0.25m CMOS Technology with Multiple Thresholds and Copper Interconnects, 1998 Symposium on
VLSI Technology Digest of Technical Papers, pp. 18-19.
CMOS Analog Circuit Design
P.E. Allen - 2004
Page 3.3-8
Page 3.3-9
n
2
1
0
L
Lmin
Fig.130-5
2.) Note that the value of varies with channel length, L. The data below is from a
0.25m CMOS technology.
0.6
0.5
0.4
PMOS
0.3
0.2
NMOS
0.1
0
0
0.5
1
1.5
Channel Length (microns)
2.5
Fig.130-6
P.E. Allen - 2004
Page 3.3-10
;;;
;;;
yyy
VGS
n+
n-channel
n+
Diffusion Current
p-substrate/well
Fig. 140-01
Page 3.3-11
log iD
Diffusion Current
Drift Current
10-6
10-12
VT
V
Fig. 140-02 GS
Page 3.3-12
iD = qADn
=
qXD
n
exp
1
exp
- V
n
po
V
L
L
t
Page 3.3-13
Define It as
k2
It = qXDnnpo expVt
to get,
vGS-V T
vDS
W
iD = L It exp nVt 1 - exp- Vt
where n 1.5 3
iD
If vDS > 0, then
1A
v
vGS-VT
W
DS
iD = It L exp nVt 1 + V
A
VGS=VT
VGS<VT
1V
vDS
Fig. 140-03
P.E. Allen - 2004
Page 3.3-14
vGS-V T
vDS ID qID ID Cox
W It
= It L nV exp nV 1 + V = nV = nkT = V C +C
t
t
t
t ox js
A
diD | ID
gds = dvDS Q VA
The boundary between nonsaturated and saturated is found as,
Vov = VDS(sat) = VON = VGS VT = 2nVt
Page 3.3-15
vGS
ID(M1)
1A
100nA
10nA
1nA
100pA
0V
0.4V
0.8V
VGS
1.2V
1.6V
2V
Fig. 140-05
Y. Cheng and C. Hu, MOSFET Modeling & BSIM3 Users Guide, Kluwer Academic Publishers, Boston, 1999.
CMOS Analog Circuit Design
Page 3.3-16
1
1 25.75S
fT = |2 T = 2 11.5fF 360MHz
Cgs = 11.5fF
(Equivalent transistor operating in strong inversion has an fT = 3.4GHz)
CMOS Analog Circuit Design
Page 3.3-17
;;;
;;;;;;;;
VG > V T
VD > VDS(sat)
Polysilicon
p+
Depletion
Region
;;;;;;;;
;;;;;;;;
A
n+
Fixed
Atom
p- substrate
Free n+
electron
Free
hole
Fig130-7
Page 3.3-18
iDB
G
B
S
Fig130-8
Small-signal model:
iDB
IDB
gdb = vDB = K2 VDS - VDS(sat)
This conductance will have a negative influence on high-output resistance current
sinks/sources.
CMOS Analog Circuit Design
Page 3.3-19
Page 3.3-20
Quiet Circuits
VDD(Analog)
;;
vin
vout
VGS
p+
n+
vin vout
VDD(Digital)
p+
n+
p+
Substrate Noise
VDD(Analog)
vin
;;;
;;;;;
;;;
;;;
;;;;;
;;;
Digital Ground
RL
vout
vin
n+
;;;
;;;
RL
vout Analog Ground
VGS
n+ channel
stop (1 -cm)
p+ channel stop (1-cm)
Hot
Carrier
Put substrate connections
as close to the noise source
as possible
n+
n+
Back-gating due to a
momentary change in
reverse bias
iD
ID
"AC ground"
p+
iD
iD
vGS
VGS
Heavily
Doped p
Lightly
Doped p
Intrinsic
Doping
Lightly
Doped n
Fig. SI-01
Heavily Metal
Doped n
Page 3.3-21
Quiet Circuits
VDD(Analog)
;;
vin
vout
VGS
p+
n+
vin vout
n+
p+
VGS
n+
Substrate Noise
VDD(Analog)
;;;
;;;
VDD(Digital)
p+
vin
;;;
;;;;;
;;;
;;;
;;;;;
;;;
Digital Ground
RL
vout
vin
RL
vout Analog Ground
n+
p-epitaxial
layer (15 -cm)
n+
p+
Reduced back
gating due to
smaller resistance
"AC ground"
Heavily
Doped p
Lightly
Doped p
Intrinsic
Doping
Lightly
Doped n
Heavily Metal
Doped n
Fig. SI-02
Page 3.3-22
;;
vin
VDD
vout
Digital Ground
L1
Cs1
n- well
VDD(Digital)
;;;
;;;
;;;;;
;;;;;;;;
;;;
;;;;;
vin vout
p+
n+
p+
n+
Hot
Carrier
Hot
Carrier Coupling
p+
Lightly
Doped p
Intrinsic
Doping
Cs3
vout Rs1
Cs2
Substrate
Rs2
n+
n- well
Cs4
Coupling
Rs3
Cs5
L3
L2
Coupling
p- substrate
Heavily
Doped p
vin
Lightly
Doped n
Heavily Metal
Doped n
Fig. SI-06
Page 3.3-23
VGS
;
RL
vout
vin
Substrate Noise
VDD
VDD(Analog)
vin
;;;
;;;
VGS
n+
n+
Analog Ground
Lightly
Doped p
L4
vout
CL
L6
p+
VGS
Substrate
Cs6 Rs4
Cs5
Cs7
L5
p- substrate
Heavily
Doped p
RL
RL
vout
Intrinsic
Doping
Lightly
Doped n
Heavily Metal
Doped n
Fig. SI-07
P.E. Allen - 2004
Page 3.3-24
p+
Emitter
;;;;;;;;
Collector
n+
n+
p- well
Fig. SI-04
Heavily
Doped p
Lightly
Doped p
Intrinsic
Doping
Lightly
Doped n
Heavily Metal
Doped n
Also, there is coupling from power supplies and clock lines to other adjacent signal lines.
CMOS Analog Circuit Design
Page 3.3-25
C2
t=0
-
Vin
C1
Vout
VDD
VSS
P.E. Allen - 2004
Fig. SI-05
Page 3.3-26
Page 3.4-1
Gate
Source
C1
Drain
C2
C3
FOX
FOX
C4
CBS
CBD
Bulk
Fig120-06
Page 3.4-2
vBS
vBSMJSW
1
1
PB
PB
and
2.) vBS> FCPB
CJAS
CBS =
1+MJ
1- FC
CJSWPS
1 - FC
Polysilicon gate
H
1+MJSW
D
Source
Drain
F
E
A
SiO2
VBS
1 - (1+MJ)FC + MJ PB
B
Bulk
Fig. 120-07
VBS
1 - (1+MJSW)FC + MJSW PB
CBS
where
AS = area of the source
vBS FCPB
PS = perimeter of the source
CJSW = zero bias, bulk source sidewall capacitance
MJSW = bulk-source sidewall grading coefficient
For the bulk-drain depletion capacitance replace "S" by "D" in the above.
vBS FCPB
PB
FCPB
vBS
Fig. 120-08
Page 3.4-3
LD
Actual
L (Leff)
Oxide encroachment
Mask
W
Actual
W (Weff)
Gate
Drain-gate overlap
capacitance CGD (C3)
Source-gate overlap
capacitance CGS (C1)
Gate
FOX
Source
Gate-Channel
Capacitance (C2)
Bulk
FOX
Drain
Channel-Bulk
Capacitance (C4)
Overlap capacitances:
C1 = C3 = LDWeffCox = CGSO or CGDO
(LD 0.015 m for LDD structures)
Channel capacitances:
C2 = gate-to-channel = CoxWeff(L-2LD) =
CoxWeffLeff
C4 = voltage dependent channelbulk/substrate capacitance
Fig. 120-09
Page 3.4-4
Overlap
FOX C5
Gate
Source/Drain
C5 FOX
Bulk
Fig120-10
C5 = CGBO
Capacitance values based on an oxide thickness of 140 or Cox=24.7 10-4 F/m2:
Type
CGSO
CGDO
CGBO
CJ
CJSW
MJ
MJSW
P-Channel
220 10-12
220 10-12
700 10-12
560 10-6
350 10-12
0.5
0.35
N-Channel
220 10-12
220 10-12
700 10-12
770 10-6
380 10-12
0.5
0.38
Units
F/m
F/m
F/m
F/m2
F/m
Page 3.4-5
Cutoff
VB = 0
;;;
;;;
;;;
;;;
VS = 0
CGS
VG < VT
Polysilicon
p+
p- substrate
Saturated
VB = 0
p+
p- substrate
Active
VB = 0
n+
VS = 0
CGS
n+
CGB
VG >VT
Polysilicon
n+
p- substrate
VD >VG -VT
CGD
n+
Inverted Region
VS = 0
CGS
VG >VT
Polysilicon
p+
VD > 0
CGD
n+
VD <VG -VT
CGD
n+
Inverted Region
Fig120-1
Page 3.4-6
CGS
C1+ 0.67C2
C1+ 0.5C2
C1, C3
2C5
0
CGS, CGD
vDS = constant
vBS = 0
C4 Small
CGS, CGD
CGD
CGB
Off
Saturation
VT
vGS
NonSaturation
vDS +VT
Fig120-12
Page 3.5-1
Q 2 2|F| - VBS
CMOS Analog Circuit Design
Page 3.5-2
where
id
+
vgs
+
vbs
gmvgs
rds
gmbsvbs
+D
vds
-
Fig. 120-01
diD |
gm dvGS Q =
diD |
iD
gds dvDS Q = 1 + vDS iD
(VGS-VT) = 2ID
and
D iD vGS iD vT
gm
gmbs = v Q = v v = - v v =
= gm
BS
GS BS Q
T BSQ 2 2|F| - VBS
An extremely important
assumption:
gm 10gmbs 100gds
+
vgs
gmvgs
+D
vds
rds
Fig. 120-02
Page 3.5-3
i
AC Resistance
v
V
DC resistance = i = I = RDC
Q
Useful for biasing - creating current from
voltage and vice versa
Small-Signal Load (AC resistance):
D
G
S
G
B
VT
VDS
Fig. 120-03
id
DC Resistance
ID
+
vgs
+
vbs
-
gmvgs
gmbsvbs
rds
+D
vds
-
Fig. 120-01
Page 3.5-4
V DS
gm = vGS Q =
(1+
V
)
DS
L
L
iD |
KW VDS
gmbs = vBS Q = 2L 2 - V
F
BS
iD |
ID
KW
KW
gds = vDS Q = L ( VGS - VT - VDS)(1+VDS) + 1+VDS L (VGS - VT - VDS)
Note:
While the small-signal model analysis is independent of the region of operation, the
evaluation of the small-signal performance is not.
Page 3.5-5
Page 3.5-6
Cgb
id
gmvgs
rds
gmbsvbs
vbs
+
vds
S
Cbd
Cbs
+
B
Fig120-13
3 o
fT = 4 L2 (VGS-VT)
Page 3.6-1
Page 3.6-2
0.4
0.2
0
50
100
Symbol
O
Min. L
6m
5m
4m
2m
150
200
Temperature (C)
NA (cm-3)
2x1016
1x1016
2x1016
3.3x1016
250
300
Fig. 3.6-1
(mV/C)
-3.5
-2.5
-2.3
-1.8
)
tox (A
1000
650
500
275
Page 3.6-3
0.4
0.2
0
Symbol
O
50
Min. L
6m
5m
4m
2m
100
150
200
Temperature (C)
NA (cm-3)
2x1015
2x1015
2x1016
1.1x1016
)
Tox (A
1000
650
500
275
250
300
Fig. 3.6-2
(mV/C)
+3.5
+2.5
+2.3
+2.0
P.E. Allen - 2004
Page 3.6-4
800
600
(T)
(cm2/Vs)
400
Theory
matched
at 25C
200
50
100
150
200
Temperature (C)
250
300
Fig. 3.6-3
Page 3.6-5
800
600
(T)
(cm2/Vs)
400 Theory
matched
at 25C
200
50
Data
Symbol Min. L
6 m
5 m
4 m
2 m
100
150
200
Temperature (C)
250
300
Fig. 3.6-4
Page 3.6-6
T
dID -1.5oCox T -2.5
-1.5
2
=
[V
-V
(T-T
)]
+
C
[VGS-VT0-(T-To)] = 0
T
T
GS
T0
o
o
ox
dT
2To
o
o
where
-4T
VGS VT0 - (T-To) = 3
VGS(ZTC) = VT0 - To - 3
Page 3.6-7
100
VDS = 6V
80
ID (A)
275C
60
300C
40
150C
275C 250C 200C
Zero TC Point
20
25C
100C
0
0
0.6
1.2
1.8
VGS (V)
2.4
3
Fig. 3.6-065
Page 3.6-8
300C
40
ID (A)
VDS = -6V
30
150C
20
275C
VSG(ZTC) -1.95V
10
150C
100C
25C
250C
0
0
-0.6
-1.2
-1.8
-2.4
-3.0
VGS (V)
Fig. 3.6-066
Zero temperature coefficient will occur for every MOSFET up to about 200C.
CMOS Analog Circuit Design
Page 3.6-9
;;;
;;;;;;;
VG > V T
VD > VDS(sat)
Depletion
Region
Polysilicon
p+
;;;;;;;
n+
n+
p-well
n- substrate
Fig.3.6-5
Page 3.6-10
;;;
;;; ;;
VG <VT
VD > VDS(sat)
Polysilicon
p+
;;; ;;
n+
Depletion
Region
n+
p-well
n- substrate
Fig.3.6-6
Page 3.6-11
iD Is = qA Lp + Ln L N = KT 3exp V t
Differentiating with respect to temperature gives,
VGo
VGo
qKT 3VGo
3Is Is VGo
dIs 3KT 3
=
=
exp
exp
2
dT
T
KT
T + T Vt
Vt
Vt
dIs
3
1 VGo
TCF = I dT = T + T V t
s
Example
Assume that the temperature is 300 (room temperature) and calculate the reverse
diode current change and the TCF for a 5 increase.
Solution
The TCF can be calculated from the above expression as
TCF = 0.01 + 0.155 = 0.165
Since the TCF is change per degree, the reverse current will increase by a factor of 1.165
for every degree (or C) change in temperature. Multiplying by 1.165 five times gives
an increase of approximately 2. Thus, the reverse saturation current approximately
doubles for every 5C temperature increase. Experimentally, the reverse current doubles
for every 8 C increase in temperature because the reverse current is part leakage current.
CMOS Analog Circuit Design
Page 3.6-12
10-6
10-7
10-8
10-9
250C
50m
Lmin
10-10
10-11
10-12
1.8
IR
Data
Symbol Min. L
6 m
5 m
4 m
2 m
100C
1V
Theory
matched
at 150C GenerationDiffusion
Recombination
Leakage
Leakage
Dominant
Dominant
2
2.2
2.4
2.6
2.8
Fig.
3.6-7
1000/T (K-1)
Theory:
V (T)
G
Is(T) T 3 exp kT
Page 3.6-13
=
V
=
=
t
dT
T
T T
T
T - T
Is dT
Assuming that vD = VD = 0.6 V the temperature dependence of the forward diode voltage
at room temperature is approximately -2.3 mV/C.
CMOS Analog Circuit Design
Page 3.6-14
MOSFET NOISE
MOS Device Noise at Low Frequencies
D
eN2
B
G
S
in2
G
Noise
Free
MOSFET
B
S
*
Noise
Free
MOSFET S
where
8kTgm(1+) KF ID
+ fSCoxL2 f (amperes2)
3
f = bandwidth at a frequency, f
gmbs
= gm
k = Boltzmanns constant
KF = Flicker noise coefficient
S = Slope factor of the 1/f noise
in2 =
Page 3.6-15
in2
KF
+ 2fC WL K f (volts2)
en2 = gm2 = 3gm
ox
KF
It will be convenient to use B = 2CoxK for model simplification.
in2
1/f noise
Thermal noise
log10(f)
Page 3.6-16
ID(A)
25/25
25/25
25/25
1.2/1.2
1.2/1.2
1.2/1.2
0.8/0.8
0.8/0.8
0.8/0.8
25/2
25/2
25/2
25/1
25/1
25/1
25/0.6
25/0.6
25/0.6
90
50
20
90
50
20
90
50
20
90
50
20
90
50
20
90
50
20
Thermal Noise
Noise Voltage at
100Hz (nV/ Hz ) Voltage (nV/ Hz )
360
40
360
35
360
25
10,000
350
10,000
200
10,000
180
70,000
1800
60,000
1500
50,000
1200
900
30
850
28
1000
38
850
33
1000
30
950
50
750
42
700
35
Page 3.6-17
G
vin
vgs
gmvgs
Cgs
rds
io2
in2
S
S
Circuit 1: Frequency Dependent Noise Model
ei2
Cgd
G
D
vin
ii2
Cgs
vgs
gmvgs
rds
io2
S
S
Circuit 2: Input-referenced Noise Model
Page 3.6-18
in2
e = g 2 + (C )2
m
gd
2
i
ii2:
Open-circuit the input and find io2 of both models and equate to get ii2 .
Ckt. 1: io2 = in2
(1/Cgs)
gm2ii2
2
Ckt. 2: io2 = (1/Cds) + (1/Cgs) ii2 + 2(Cgs+Cds)2
2Cgs2
gm2
2
2
2Cgs2 in if Cgd < Cgs ii = gm2 in2
Page 3.7-1
iC
vBC+
+
vCE
+
vBE iE
iC
vBC+
+
vCE
+
vBE iE
E
npn
iB
B
E
pnp
Fig.07-02
Base - The base is a region which physically separates the emitter and collector and
has an opposite doping (holes for the npn and electrons for the pnp BJTs). The word
base comes from the way that the first transistors were constructed. The base was
the physical support for the whole transistor.
Collector - The collector serves to collect those carries injected from the emitter into
the base and which reach the collector without recombination.
Page 3.7-2
;;;;;
;;;
;;;;;
;;;
E
Depletion
Region
n+
C Depletion
Region
;
;
;;
B
n+
A
Fig.070-03
n
A'
Depletion Depletion
Region Region
Comments:
The emitter-base depeletion region is generally smaller in width because the doping
level is higher and base-emitter junction is generally forward-biased.
The next slide will examine the carrier concentrations see looking into the above A-A
cross-section.
CMOS Analog Circuit Design
Page 3.7-3
;;
;;
;
Carrier
Concentration
nnE
pp(x)
Depletion
Region
pnE
A
pnE(0)
Emitter
;;;
;;;
;;;
;;;
Depletion
Region
np(0)
x=0
NA
np(x)
Base
np(WB)
x =WB
nnC
ND
pnC
Collector
x
A'
Fig.070-04
Comments:
The above carrier concentrations assume that the base-emitter junction is forward
biased and the base-collector junction is reverse biased.
The above carrier concentration will be used to derive the large signal model on the
next slide.
CMOS Analog Circuit Design
Page 3.7-4
Derivation of the BJT Large Signal Model in the Foward Active Region
1.) Carrier concentrations in the base on the emitter side. The concentration of electrons
in the base on the emitter side (x = 0) is
np(0) = npo exp(vBE/Vt)
The concentration of electrons in the base on the collector side (x = WB) is
np(WB) = npo exp(vBC/Vt) 0 because vBC is negative and large.
2.) If the recombination of electrons in the base is small, then the minority-carrier
concentrations, np(x), are straight lines and shown on the previous page. From
charge-neutrality requirements for the base,
NA + np(x) = pp(x) np(x) - pp(x) = NA
3.) The collector current is produced by minority-carrier electrons in the base diffusing in
the direction of the concentration gradient and being swept across the collector-base
depletion region by the field existing there. Therefore, the diffusion current density
due to electrons in the base is
dnp(x)
Jn = qDn dx
where Dn is the diffusion constant for electrons. The derivative is the slope of the
concentration profile in the base which gives,
np(0)
Jn = -qDn W
B
Page 3.7-5
Derivation of the BJT Large Signal Model in the Foward Active Region - Continued
3.) Continued
If the collector current is defined as positive flowing into the collector terminal, then
np(0)
vBE
qADnnpo
iC = qADn W B = W B exp Vt
Page 3.7-6
and
vBE
qADp ni2
iB2 = Lp ND exp Vt
1 n W qA qAD n 2
vBE
po B
p i
+ Lp ND exp Vt
iB = iB1 + iB2 = 2
b
Page 3.7-7
Derivation of the Current Gain from Emitter to Collector in Forward Active Region
iC
1.) Emitter to collector current gain is designated as, F = iE .
2.) Since sum of all currents flowing into the transistor must be zero, we can write that
iC
iC
iE = -(iC+iB) = - iC + F =- iC 1+ F = - F
F
F = 1 + F =
1
=
1
W 2 D W N T
1 + F 1 + B + p B A
2bDn Dn Lp ND
where
1
W B2
1 + 2bDn
and
Emitter injection efficiency
1
Dp W B NA 1
1 + Dn Lp ND
P.E. Allen - 2004
Page 3.7-8
Large Signal Model for the BJT in the Forward Active Region
Large-signal model for a npn transistor:
iB
B
vBE
FiB
E
v
iB = Is exp BE
F
Vt
Assumes vBE is a
constant and iB is
determined externally
+
VBE(on)
E
FiB
E
Fig.070-05
iB
C
vBE
E
FiB
E
iB = - Is exp -vBE
F
Vt
Assumes vBE is a
constant and iB is
determined externally
VBE(on)
+
E
Fi B
E
Fig.070-06
Page 3.7-9
;;
;;
;;
;;
Carrier
Concentration
Emitter
;;;
;;;
;;;
;;;
Collector depletion
region widens due to a
change in vCE, VCE
Initial
Depletion
Region
Collector
WB
Fig.070-07
Note that the change of the collector-emitter voltage causes the amount of charge in the
base to change slightly influencing the collector current.
CMOS Analog Circuit Design
Page 3.7-10
V
=
-W
A
B W B
vCE W B vCE
VA
where VA is called the Early voltage.
Page 3.7-11
VBE4
VBE3
VBE2
VBE1
VA
vCE
Fig.070-08
vBE
vCE
iC = IS 1 + VA exp Vt
Page 3.7-12
Saturation Region
BE forward biased
BC forward biased
vBC
Cutoff Region
BE reverse biased
BC reverse biased
E
Fig.080-01
Note: While the back-to-back diode model is appropriate here, it is not a suitable model
for the BJT in general because it does not model the current gain mechanism of the BJT.
Essentially, the back-to-back diode model has a very wide base region and all the injected
carriers from the emitter recombine in the base (F = 0).
CMOS Analog Circuit Design
Page 3.7-13
Saturation Region
In the saturation region, both the base-emitter and base-collector pn junctions are
forward biased.
Consequently, there is injection of electrons into the base from both the emitter and
collector.
The carrier concentrations in saturation are:
;;
;;
;;
;;
;;
Carrier
Concentration
nnE
pp(x)
np(0)
Electrons
pnE
pnE(0)
Emitter
np1(x)
np(x) iC
np2(x)
WB
Base
;;
;;
;;
;;
np(WB)
nnC
pnC
Electrons
x
Collector
Fig.080-02
Page 3.7-14
4
3
0.02mA
Saturation
2
Cutoff
-8
-6
IB=0.01mA
IB=0
Forward
active
region
0.01mA
1
-4
-2
10
-0.02
Saturation
-0.04
0.02mA
0.03mA
Inverse
active
region
20
30
Cutoff
IB=0
40
VCE(V)
BVCEO
-0.06
0.04mA
-0.08
-0.10
Fig.080-04
Page 3.7-15
B
VBE(on)
E
C
VCE(sat)
E
VCE(sat)
VBE(on)
npn
where VBE(on) 0.6 to 0.7V and VCE(sat) 0.2V
pnp
Fig.1.3-11
v
BE
np(x) iC
iEF = -IES exp V - 1
t
np1(x)
where IES is a constant called saturation current
iCR np2(x)
2.) The collector injected current in the base resulting
from np2(x) is,
WB
v
Base
BC
iCR = -ICS exp Vt - 1
vBE
vBC
iC = iCR + FiEF = FIES exp V - 1 -ICS exp V - 1
t
t
vBE
vBC
iE = iEF + RiCR = IES exp Vt - 1 +RICS exp Vt - 1
Page 3.7-16
np(WB)
iEF
x
Fig.080-06
Page 3.7-17
vBE IS vBC
iC = IS exp V - 1 -R exp V - 1
t
t
and
IS vBE
vBC
iE =-F exp Vt - 1 +IS exp Vt - 1
These equations are valid for all four regions of operation of the BJT.
Page 3.7-18
IE=1.5mA
1.0
IE=1.0mA
iC
IE
VCB
IE=0.5mA
0.5
IE=0
Fig.080-08
20
40
60
80
100
VCB(V)
BVCBO
As the collector-base voltage becomes large, the collector current can be written as,
i C = - F i E M
where
1
M=
vCB n
1 - BV
CBO
Page 3.7-19
iC = -(iE + iB)
iC 1- M = -iB
iC = 1- M iB
F
F
where,
1
vCB n
1 - BVCBO
=
BVCEO n
BVCBO
F1/n
1 - BVCBO
Note that BVCEO is less than BVCBO. For F = 100 and n = 4, BVCEO 0.5BVCBO.
M=
Page 3.7-20
Region I
Region II
Region III
T=125C
300
T=25C
200
T=-55C
100
0
0.1A
1A
10A
100A
1mA
10mA
iC
Fig.080-09
1 F
TCF = F +7000ppm/C (ppm = parts per million)
Page 3.7-21
ln FH
iC ln FM
ln FL
vBE
vBE
due to recombination, m 2
vBE
iC
IS
1
FL = iBX = ISX exp Vt 1 - m
ln Is
Region I
iB
Region
II
Region III
Fig.080-10
vBE
(linear scale)
IS iC[1-(1/m)]
for m = 2, FL iC
SX Is
Region III:
vBE
iC = ISH exp 2Vt due to the high level injection and
vBE
IS
iB FM exp Vt
vBE
ISH
ISH2
1
FH IS F exp- 2Vt IS FM iC
Page 3.7-22
iC = IC + ic
iB = IB + ib
vi
VCC
VBE
;;;;;;
;;;
;
;;;;;;
;;;
;
;
;;;;;;
;;;
;
;;;;;;
;;;
Carrier
Concentration
Qh
Collector
Depletion
Region
Qe
IC+ic
IC
WB
Base
Emitter
x
Collector
Fig.090-02
An increase in vBE (vi) causes more electrons to be injected in the base increasing the
base current iB by an amount ib. The increased base current causes the collector current
iC to increase by an amount ic.
vi i b i c
Page 3.7-23
ic = gmvi
BE
BE vbe vi
The large signal model for iC is
vBE
iC = IS exp V
t
gm =
vBE
IS
VBE IC
d
I
exp
=
exp
dvBE S
V t Q V t
Vt = Vt
IC
gm = V t
vi 1 vi 2 1 vi 3
iC = IS exp Vt = IS exp Vt expVt = IC expVt IC1 + Vt + 2 Vt + 6 Vt +
But
iC = IC + ic
vi IC vi
IC vi
IC
ic IC Vt + 2 Vt2 + 6 Vt3 + Vt vi = gmvi
Page 3.7-24
o
r = gm
Page 3.7-25
ic = govce
CE
CE vce
The large signal model for iC , including the influence of vCE, is
vCE
vBE
iC = IS 1 + V exp V
A
t
diC |
VBE IC
1
go dvCE Q = IS VA exp Vt VA
VA
ro = I C
Page 3.7-26
B
+
vi
ib
ic
gmvi
ro
C
+
vce
E
Fig. 090-03
Note that the small signal model is the same for either a npn or a pnp BJT.
Example:
Find the small signal input resistance, Rin, the output resistance, Rout, and the voltage
gain of the common emitter BJT if the BJT is unloaded (RL = ), vout/vin, the dc collector
current is 1mA, the Early voltage is 100V, and at room temperature.
o
IC 1mA
1
gm = V = 26mV = 26 mhos
R
=
r
=
in
gm = 10026 = 2.6k
t
VA 100V
vout
Rout = ro = IC = 1mA = 100k
vin = -gm ro = - 26mS100k = -2600V/V
CMOS Analog Circuit Design
Page 3.7-27
;;
;;
;;
;;
Carrier
Concentration
Emitter
;;;
;;;
;;;
;;;
Collector depletion
region widens due to a
change in vCE, VCE
Initial
Depletion
Region
Collector
WB
Fig.3.7-6
We noted that an increase in vCE causes and increase in the depletion width and a
decrease in the total minority-carrier charge stored in the base and therefore a decrease in
the base recombination current, iB1.
This influence is modeled by a collector-base resistor, r, defined as
vCE vCE iC
iC
r = i
=
=
r
iC i1 o i1 oro (if base current is primarily recomb.)
1
In general, r 10 oro for the npn BJT and about 2-5 oro for the lateral pnp BJT.
CMOS Analog Circuit Design
Page 3.7-28
iC = IC + ic
iB = IB + ib
vi
VCC
Qh
VBE
;;;;;;
;;;
;
;;;;;;
;;;
;
;
;;;;;;
;;;
;
;;;;;;
;;;
Carrier
Concentration
Collector
Depletion
Region
Qe
IC+ic
IC
x
Collector
WB
Base
Fig.3.7-16
The vBE change causes a change in the minority carriers, Qe = qe, which must be equal
to the change in majority carriers, Qh = qh. This charge can be related to the voltage
across the base, vi, as
qh = Cbvi
where Cb is the base-charging capacitor and is given as
qh F ic
IC
Cb = vi = vi = F gm = F Vt
W B2
The base transit time F is defined as 2Dn
CMOS Analog Circuit Design
Page 3.7-29
Base
Emitter
n+ emitter
p+
n+
p- isolation
rex
Cje
rb
rc3 C p base
Ccs
n collector
Cje
p- isolation
Ccs
rc1
rc2
n+ buried layer
p- substrate
p+
p-
ni
n-
n+
Metal
Fig.3.7-18
m
1 -
0
Resistances are all bulk ohmic resistances. rb, rc, and rex are important. Also, rb = f(IC).
CMOS Analog Circuit Design
Page 3.7-30
rb
C
B'
r
+
v1
-
gmv1
rc
ro
Ccs
rex
E
E
Fig. 3.7-19
Page 3.7-31
Example
Derive the complete small signal equivalent circuit for a BJT at IC = 1mA, VCB = 3V, and
VCS = 5V. The device parameters are Cje0 = 10fF, ne = 0.5, 0e = 0.9V, C0 = 10fF, nc =
0.3, 0c = 0.5V, Ccs0 = 20fF, ns = 0.3, 0s = 0.65V, o = 100, F = 10ps, VA = 20V, rb =
300, rc = 50, rex = 5, and r = 10oro.
Solution
Because Cje is difficult to determine and usually an insignificant part of C, let us
approximate it as 2Cje0.
Cje = 20fF
C0
Ccs0
10fF
20F
C = V =
=
5.6fF
and
C
=
=
cs
3 0.3
5 0.3 = 10.5fF
VCSn
CBne
s
1+
1+
1+
1+
0.5
0.65
0c
0s
IC 1mA
Cb = F gm = (10ps)(38mA/V) = 0.38pF
gm = Vt = 26mV = 38mA/V
C = Cb + Cje = 0.38pF+0.02pF = 0.4pF
o
VA 20V
r =g =10026=2.6k, ro= I =1mA =20k, and r=10ro=1010020k =20M
m
C
CMOS Analog Circuit Design
Page 3.7-32
ii
+
v1
-
gmv1
ro
Ccs
io
Fig.3.7-20
Page 3.7-33
100
10
1
0.1T
(log scale)
Fig.3.7-21
Note that the product of the magnitude and frequency at any point on the 6dB/octave
curve is equal to T.
For example,
0.1 T x10 = T
In measuring T, the value of |(j)| is measured at some frequency less than T (say
x) and T is calculated by taking the product of |(jx)| and x to get T.
CMOS Analog Circuit Design
Page 3.7-34
Current Dependence of fT
C C Cb Cje C
Cje C
1
Note that T = = gm + gm = gm + gm + gm = F + gm + gm
At low currents, the Cje and C terms dominate causing T to rise and T to fall.
At high currents, T approaches F which is the maximum value of T.
For further increases in collector current, T decreases because of high-level injection
effects and the Kirk effect.
Typical frequency dependence of fT:
fT (GHz)
10
8
6
4
2
0
10A
1mA
100A
IC
10mA Fig.3.7-22
Page 3.7-35
+
v in
-
Add all internal noise sources to the BJT small signal model to get:
B
rb
B'
v2b = 4kTrb f
+
r
i2b
E
I
= 2qIBf + K1 B f
f
v gmv
ro
io
i2c = 2qICf
Noise-free BJT
where
vb2 = thermal noise of the base resistance
ib2 = base shot and flicker noise currents
and
ic2 = collector shot and flicker noise currents
CMOS Analog Circuit Design
Page 3.7-36
B
vin
rb
B'
r
ieq2 = ii2
C
C
v
gmv
ro
Noise-free BJT
io2
E
Fig. 3.7-25
ic2
veq2 = vi2 = vb2 + gm2
Page 3.7-37
1 2
i 2
2
2
2
2 = ic2 + gm2r2
io = ic + gm
i
b
b
1
srC+1
r + sC
1 2
ib2 = ic2 + |(j)|2 ib2
io2 = ic2 + o2 s
+ 1
Circuit 2:
Page 3.7-38
i2eq = i2i
10-23
1/f
10-24
Thermal
Influence of
a decreasing
10-25
f
100 1k 10k 100k 1M 10M 100M1G
log f
Page 3.7-39
rb
B'
v2b = 4kTrb f
*
C
v gmv
rc
+
r
I
i2b = 2qIBf + K1 B f
f
v2c = 4kTrcf
ro
CCS
v2e = 4kTref
i2c = 2qICf
*
re
E
CMOS Analog Circuit Design
Page 3.7-40
MOS Transistor
2KW
2LID
Intrinsic Gain
T
Input Noise Voltage
(V2/Hz)
1
F
VT +
4kTrb +
VGS-VT R (W/L)
- W/L
2
R
1
D
Rin
2qIC
gm2
IC
IB a
2qIB + K1 f +
|(j)|2
Rout
gm
1
ID
gm
2KID
3
=
Cgs 2Cox
WL3
8kT
K
+
3gm WLCoxf
Bipolar Transistor
VA
Vt
kT R AE QB
q - R - AE - QB
VA
IC
r
IC
Vt
2KWID
L
Page 3.8-1
;;;
;;;;;;;
Depletion
Region
Polysilicon
p+
;;;;;;;
n+
n+
v(y)
dy
p- substrate
Fig.3.8-1
Assume, the charge in the depletion region between the channel and bulk is no longer
constant and is dependent on v(y). Therefore, we model the dependence of threshold
voltage, VT, on y as
VT(y) = VT0 + 2|F| + vCB - 2F
where vCB is the voltage across the depletion region at y and is expressed as
vCB = vS + v(y) vB = v(y) + vSB
VT(y) = VT0 + 2|F| + v(y) + vSB - 2F
CMOS Analog Circuit Design
Page 3.8-2
iDdy = nWQI(y)dv(y)
Integrating this result over the channel from source to drain gives,
vDS
iD = dy = nWCox
0
vDS
2
2
1.5
1.5
iDL = nWC vGS-VT0+ 2|F|- 2 vDS-3 2|F| +vSB+vDS +3 2|F|+vSB
nWCox
vDS
2
2
1.5+ 2| |+v 1.5
v
-V
+
2|
|v
2|
|+v
+v
or iD =
T0
F 2 DS 3 F SB DS
3 F SB
L
GS
These results agree with the first edition of the text if the following definitions are made:
si
VBIN VT0 - 2|F| ,
1
and 4C W 1
ox
ox
Page 3.8-3
sWCox
vDS
2
1.5
1.5
iD = Lmod vGS -VBIN - 2 vDS - 3s 2|F| +vSB+vDS + 2|F| +vSB
Page 3.8-4
Mobility Degradation
The degradation of the surface mobility o can be written as
UCRITsi
UEXP
s = o C [v -V -UTRAv ]
ox
GS
T
DS
where
UCRIT = Critical field for mobility degradation (Volts/cm)
UTRA = Transverse field coefficient for mobility degradation
UEXP = Critical field exponent for mobility degradation
Normally, s o
Page 3.8-5
;;
Page 3.8-6
;;;;
;;;;
;;;
;;;;;
;;;;
;;;
;;;;;
Polysilicon Gate
SourceXJ
WS
WS W
XJ+WS
Gate oxide
XJ Drain
XJ+WD
WD
2WS
XJ
S = 2L 1+ XJ - 1
2WD
XJ
D = 2L 1+ XJ - 1
where
XJ = metallurgical junction depth (meters)
2si
WS = source depletion width =
qNSUB (2F + |vSB|) )
2si
WD = Drain depletion width =
qNSUB (2F + |vSB| + vDS) )
CMOS Analog Circuit Design
Page 3.8-7
2si
qNSUB
vDS-vDS(sat)
+
4
v
-v (sat)2
DS DS
1 +
4
and
vGS-VBIN s2
+ 22 1vDS(sat) =
2 vGS-VBIN
1+ 2
+
2
+
|v
|
F
BS
Page 3.9-1
Minimum Minimum
Model
iD Accuracy in iD Accuracy in
L (m) Tox (nm) Continuity Strong Inversion Subthreshold
MOS1
5
50
Poor
Poor
Not Modeled
MOS2
2
25
Poor
Poor
Poor
MOS3
1
20
Poor
Fair
Poor
BSIM1
0.8
15
Fair
Good
Fair
BSIM2
0.35
7.5
Fair
Good
Good
BSIM3v2
0.25
5
Fair
Good
Good
BSIM3v3
0.15
4
Good
Good
Good
Poor
Fair
Poor
Fair
Fair
Good
Good
Page 3.9-2
Page 3.9-3
Page 3.9-4
Page 3.9-5
BSIM2 Model
Generic composite expression for the model parameters:
LX WX
X = Xo + Leff + W
eff
where
Xo = parameter for a given W and L
LX (WX) = first-order dependence of X on L (W)
Modeling features of BSIM2:
Mobility
Mobility reduction by the vertical field
Mobility reduction by the lateral field
Drain Current
Velocity saturation
Linear region drain current
Saturation region drain current
Subthreshold current
oCoxWeff kT evGS-Vt-Voff
iDS =
q
1 - eqVDS/kT
Leff
n
where
Voff = VOF + VOFB vBS + VOFD vDS and
n = NO +
NB
+ ND vDS
PHI - vBS
Page 3.9-6
Saturation
(DIBL)
Linear
Region
(Triode)
Channel
length
modulation
(CLM)
0
vDS(sat)
Drain
current
Substrate
current
induced
body
effect
(SCBE)
5V
vDS
(3.1-2)
Page 3.9-7
10
11
12
Weff,2
Weff,1
Leff,1
Leff,2
Leff,3
Leff,4
Leff
A long, wide device is used as the base to add geometry effects as corrections.
Procedure:
1.) Oxide thickness and the differences between the drawn and effective channel
dimensions are provided as process input.
2.) A long, wide device is used to determine some base parameters which are used as
the starting point for each individual device extraction in the second phase.
3.) In the second phase, a set of parameters is extracted independently for each device.
This phase represents the fitting of the data for each independent device to the intrinsic
equation structure of the model
1.) In the third phase, the compiled parameters from the second phase are used to
determine the geometry parameters. This represents the imposition of the extrinsic
structure onto the model.
CMOS Analog Circuit Design
Page 3.9-8
Page 3.9-9
iD
+
vGS
ID(M1)
1A
100nA
10nA
1nA
100pA
0V
0.4V
0.8V
VGS
1.2V
1.6V
2V
Page 3.9-10
BSIM3 Model
The background for the BSIM3 model and the equations are given in detail in the text
MOSFET Modeling & BSIM3 Users Guide, by Y. Cheng and C. Hu, Kluwer Academic
Publishers, 1999.
The short channel effects included in the BSIM3 model are:
Normal and reverse short-channel and narrow-width effects on the threshold.
Channel length modulation (CLM).
Drain induced barrier lowering (DIBL).
Velocity saturation.
Mobility degradation due to the vertical electric field.
Impact ionization.
Band-to-band tunnelling.
Velocity overshoot.
Self-heating.
1.) Channel quantiztion.
2.) Polysilicon depletion.
Page 3.9-11
W eff
AbulkvDS
1
v
-V
vDS < VDS(sat)
iDS = effCox Leff
2 vDS ,
vDS GS th
1+ EsatLeff
vDS - VDS(sat)
iDS = WeffvsatCox[vGS Vth AbulkVDS(sat)]1+
vDS > VDS(sat)
,
VA
where
EsatLeff(vGS-Vth)
VDS(sat) = AbulkEsatLeff + (vGS-Vth)
Leff = Ldrawn 2dL
W eff = W drawn 2dW
Esat = Electric field where the drift velocity (v) saturates
vsat = saturation velocity of carriers in the channel
2vsat
eff
eff = Esat
= 1+(Ey/Esat)
Note: Assume Abulk 1 and extract Vth and VA.
CMOS Analog Circuit Design
Page 3.9-12
VENDOR: TSMC
FEATURE SIZE: 0.25 microns
INTRODUCTION: This report contains the lot average results obtained by MOSIS from measurements of MOSIS
test structures on each wafer of this fabrication lot. SPICE parameters obtained from similar measurements on a
selected wafer are also attached.
COMMENTS: TSMC 0251P5M.
TRANSISTOR PARAMETERS
MINIMUM
Vth
SHORT
Idss
Vth
Vpt
WIDE
Ids0
LARGE
Vth
Vjbkd
Ijlk
Gamma
K (Uo*Cox/2)
CMOS Analog Circuit Design
W/L
N-CHANNEL
P-CHANNEL
UNITS
0.36/0.24
0.54
-0.50
volts
557
0.56
7.6
-256
-0.56
-7.2
uA/um
volts
volts
6.6
-1.5
pA/um
0.47
5.8
-25.0
0.44
112.0
-0.60
-7.0
-1.1
0.61
-23.0
volts
volts
pA
V0.5
uA/V2
20.0/0.24
20.0/0.24
50.0/50.0
Page 3.9-13
Page 3.9-14
Page 3.9-15
x1
x2
Circuit
Level
Simulators
x3
Process
Simulators
"Extraction" Methodology
I-V characterisitcs
Capacitances
Transconductances
Measurement
Methods
Objective
Develop models having adjustable precision in ac and dc perfomrance using table
lookup models.
Advantages
Usable at any level device, circuit, or behavioral
Quickly developed from experiment or process simulators
Faster than analytical device models (BSIM)
Disadvantages
Requires approximately 10kbytes for a typical MOS model
Cant be parameterized easily
CMOS Analog Circuit Design
Page 3.9-16
Page 3.10-1
(1)
2
Weff
vDS
(v - V )v (1 + vDS)
iD = K
Leff GS T DS 2
(2)
where
VT = VT0 + [ 2|F| + vSB 2|F| ]
(3)
Page 3.10-3
(7)
(8)
y = iD
x = vGS
K' Weff
1/2
m = 2Leff
(9)
and
K' Weff
1/2
b = - 2Leff
VT0
(10)
Page 3.10-4
Mobility degradation
region
vDS >VDSAT
Weak inversion
region
0
0
b =VT0
1/2
K Weff
m=
2L eff
vGS
AppB-01
Comments:
Stay away from the extreme regions of mobility degradation and weak inversion
Use channel lengths greater than Lmin
Page 3.10-5
xi yi - ( xi yi)/n
(12)
xi - (xi)2/n
1/2
determine VT0 and K W/2L. The data in Table B-1 also give I D as a function of VGS.
Table 3.10-1 Data for Example 3.10-1
VGS (V)
1.000
1.200
1.500
1.700
1.900
ID (A)
0.700
2.00
8.00
13.95
22.1
ID (A)1/2
0.837
1.414
2.828
3.735
4.701
VSB (V)
0.000
0.000
0.000
0.000
0.000
Page 3.10-6
ID2 - ID1
y
=
x VGS2 - VGS1
Gives
m1 =
1.414 - 0.837
= 2.885
0.2
m2 =
2.828 - 1.414
= 4.713
0.3
m3 =
3.735 - 2.828
= 4.535
0.2
m4 =
4.701 - 3.735
= 4.830
0.2
These results indicate that the first (lowest value of VGS) data point is either bad, or at a
point where the transistor is in weak inversion. This data point will not be included in
subsequent analysis. Performing the linear regression yields the following results.
K'Weff
2
and
VT0 = 0.898 V
2Leff = 21.92 A/V
Page 3.10-7
Page 3.10-8
VT0
VT1
VT2
VT3
vGS
FigAppB-02
By plotting VT versus x of Eq. (13) one can measure the slope of the best fit line from
which the parameter can be extracted. In order to do this, VT must be determined at
various values of vSB using the technique previously described.
Page 3.10-9
VSB =2V
VSB =1V
m=
VSB =0V
0.5
(vSB +2 F ) (2 F )
0.5
FigAppB-03
Page 3.10-11
xi = 0.9423
(xi)2 = 1.764
These values give m = 0.506 = .
Page 3.10-12
Nonsaturation
region
i'D
m = i'D
AppB-03
CMOS Analog Circuit Design
Saturation region
vDS
P.E. Allen - 2004
Page 3.10-13
Page 3.10-14
Page 3.10-15
(A/V2)
VT(V)
(V-1)
294.1x10-6
1.4564
0.4190
0.1437
5.) The results for a NMOS and PMOS transistor are shown on the following pages.
Anurag Kaplish, Parameter Optimization of Deep Submicron MOSFETS Using a Genetic Algorithm, May 4, 2000, Special Project Report, School
of ECE, Georgia Tech.
CMOS Analog Circuit Design
P.E. Allen - 2004
Chapter 3 Section 10 (5/2/04)
Page 3.10-16
Page 3.10-17
Page 3.10-19
Page 3.11-1
Page 4.0-1
Blocks or circuits
(Combination of primitives, independent)
Sub-blocks or subcircuits
(A primitive, not independent)
Chapter 4
Fig. 4.0-1
Page 4.0-2
Biasing
Circuits
Input
Differential
Amplifier
Second
Gain
Stage
Inverter
Source
Current
Coupled Pair Mirror Load
Current
Sink Load
Output
Stage
Source
Follower
Current
Sink Load
Fig. 4.0-2
Page 4.1-1
IOFF
rOFF
VOS
rON
A
B
CAB
CAC
IA
CA
CBC
IB
CB
VC
Fig. 4.1-1
Actual switch:
VC = controlling terminal for the switch (VC high switch ON, VC low switch OFF)
roff = resistance of the switch when OFF
ron = resistance of the switch when ON
VOS = offset voltage when the switch is ON Ioff = offset current when the switch is OFF
IA and IB are leakage currents to ground
CA and CB are capacitances to ground
CAC and CBC = parasitic capacitors between the control terminal and switch terminals
CMOS Analog Circuit Design
Page 4.1-2
(S/D)
(D/S)
C (G)
Fig4.1-2
vDS
1
RON iD = CoxW
L (vGS - VT)
Page 4.1-3
Circuit
1
(0 to 1V)
(0 to 1V)
(S/D)
(D/S)
Circuit
2
Gate
Fig.4.1-3
To insure that the bulk-source and bulk-drain pn junctions are reverse biased, the bulk
voltage must be less than the minimum analog signal for a NMOS switch.
To insure that the switch is on, the gate voltage must be greater than the maximum
analog signal plus the threshold for a NMOS switch.
Therefore:
VBulk 0V
and
VGate(on) > 1V + VT
Also,
VGate(off) 0V
Unfortunately, the large value of reverse bias bulk voltage causes the threshold voltage to
increase.
CMOS Analog Circuit Design
Page 4.1-4
50A
VGS=3.0V
VGS=3.5V
VGS=4.0V
VGS=4.5V
VGS=5.0V
VGS=2.5V
VGS=2.0V
VGS=1.5V
iD 0A
VGS=1.0V
-50A
-100A
-1V
-0.5V
0V
vDS
0.5V
1V
Fig. 4.1-4
VGS 2 0 DC 0.0
VBS 3 0 DC -5.0
.DC VDS -1 1 0.1 VGS 1 5 0.5
.PRINT DC ID(M1)
.PROBE
.END
P.E. Allen - 2004
Page 4.1-5
MOSFEET On Resistance
100k
10k
W/L = 1m/1m
1
k
W/L = 5m/1m
W/L = 10m/1m
100
W/L = 50m/1m
10
1V 1.5V 2V 2.5V 3V 3.5V 4V 4.5V 5V
Fig. 4.1-5
VGS
Page 4.1-6
vin=2.5V
VGate
v
+ C vin>0
RON
Fig. 4.1-6
Example
Initially assume the capacitor is uncharged. If VGate(ON) is 5V and is high for 0.1s,
find the W/L of the MOSFET switch that will charge a capacitance of 10pF in five time
constants.
Solution
The time constant must be 100ns/5 = 20ns. Therefore RON must be less than
20ns/10pF = 2k. The ON resistance of the MOSFET (for small vDS) is
1
W
1
1
=1.06
RON = KN(W/L)(VGS-VT) L = RONKN(VGS-VT) =
2k110A/V24.3
Comments:
It is relatively easy to charge on-chip capacitors with minimum size switches.
Switch resistance is really not constant during switching and the problem is more
complex than above.
CMOS Analog Circuit Design
Page 4.1-7
gON(0)
t=0
t=0
vDS(0)
VGS=5V
VDS
Fig. 4.1-7
VGate
+
vGS(t)
-
VGS=5V
vIN
VGS=5V-vIN
gON(0)
+
C
vC(0) = 0
gON()
t=
vDS()
vDS(0)
VDS
Fig. 4.1-8
KWVDS(0) KW
KW
gON = 2L [VGS(0)-VT] + 2L [VGS()-vIN-VT]
4L
CMOS Analog Circuit Design
Page 4.1-8
C2 = 10pF
0V
vout(t)
+ +
Fig.4.1-9
Solution
Note that the source of the NMOS is on the right and is always at ground potential so
there is no bulk effect as long as the voltage across C1 is positive. The voltage across C1
can be expressed as
-t
vC1(t) = 5expR C
ON 1
At 10ns, vC1 is 5/100 or 0.05V. Therefore,
-103
ln(100)
-10-8
exp(G 103)=100 G
=
5exp
=
=0.0046S
0.05=5exp
ON
ON
RON
RON10-11
103
KWVDS(0)
KW
110x10-65W
W
-6
-6
110x10 4.3
0.0046 = L (VGS-VT) =
=
356x10
4L
4
L
L
W
0.0046
Thus, L =
= 13.71 14
356x10-6
CMOS Analog Circuit Design
Page 4.1-9
RBulk
CH
+
vCH
-
vout
Fig. 4.1-10
Typically, no problems occur unless capacitance voltages are held for a long time. For
example,
vout(t) = vCH e-t/(RBulkCH)
If RBulk 109 and CH = 10pF, the time constant is 10910-11 = 0.01seconds
Page 4.1-10
CL
+
vCL
-
Cchannel
CGS0
CGD0
Rchannel
CL
VS
A distributed model of
the transistor switch.
+
vCL
-
Cchannel
2
2
CGS0
VS
Cchannel
CGD0
Rchannel
CL
VS
A lumped model of
the transistor switch.
+
vCL
-
Fig. 4.1-11
Page 4.1-11
Page 4.1-12
A
B
Switch OFF
vin+VT
C
Charge
injection
vin
CL
vin
CL
Fig. 4.1-13
1.) During the on-to-off transition time from A to B, the charge injection is absorbed by
the low impedance source, vin.
2.) The switch turns off when the gate voltage is vin+VT (point B).
3.) From B to C the switch is off but the gate voltage is changing. As a result charge
injection occurs to CL.
Page 4.1-13
Charge
injection
vin
vin
CL
CL
Fig. 4.1-14
Page 4.1-14
Voltage
Discretized Gate Voltage
vGATE
vin+VT
vin
vin+VT
vin
vCL
Slow Transition
vCL
Fast Transition
Charge
injection
due to fast
transition
t
Fig 4.1-15
The time constant of the channel, RchannelCchannel, determines whether or not the
capacitance, CL, fully charges during each voltage step.
B.J. Sheu and C. Hu, Switched-Induced Error Voltage on A Switched Capacitor, IEEE J. Solid-State Circuits, Vol. SC-19, No. 4, pp. 519-525,
August 1984.
CMOS Analog Circuit Design
P.E. Allen - 2004
Page 4.1-15
VHT
1.) Slow transition occurs when 2CL >> U.
C
WCGD0 + channel
UCL WCGD0
2
(VS+2VT -VL)
Verror = -
CL
2 CL
VHT
2.) Fast transition occurs when 2C << U.
L
Cchannel
3
WCGD0 +
V
2
HT WCGD0
(VS+2VT -VL)
Verror = -
CL
CL
VHT - 6UCL CMOS Analog Circuit Design
Page 4.1-16
vin VS VD
VS +VT
C
VT
VL
COL
Charge
injection
CL
COL
+
VS +V T
Circuit at the VL
instant gate
reaches VS +VT
CL
vCL
VS
Fig. 4.1-16
The switch decrease from B to C is modeled as a negative step of magnitude VS +VT - VL.
The output voltage on the capacitor after opening the switch is,
COL
COL
CL COL
Page 4.1-17
VHT 110x10-63.1132
= 2.66x109 < 25x109
VHT =VH-VS-VT = 5-1-0.887=3.113V 2CL =
2200fF
which corresponds to the fast transition case. Using the previous expression gives,
Verror =
176x10-18+0.5(1.58x10-15)
3.32x10-3 176x10-18
-
3.113- 30x10-3 - 200x10-15(1+1.774-0) = -16.94mV
200x10-15
CMOS Analog Circuit Design
Page 4.1-18
Verror = -
200x10-15
314x10-6 176x10-18
(1+1.774-0) = -8.21mV
220x10-6 200x10-15
Comment:
These results are not expected to give precise answers regarding the amount of
charge feedthrough one should expect in an actual circuit. Rather, they are a guide to
understand the effects of various circuit elements and terminal conditions in order to
minimize unwanted behavior by design techniques.
Page 4.1-19
W1
L1
WD = W1
LD 2L1
M1
MD
Fig. 4.1-19
Page 4.1-20
1d
Page 4.1-21
VDD
Clock
Clock
Fig. 4.1-21
Advantages:
Feedthrough somewhat diminished
Larger dynamic range
Lower ON resistance
Disadvantages:
Requires a complementary clock
Requires more area
Page 4.1-22
5V
vin-|VTP|
0.8m
0.8m
M2
0.8m
0.8m
0V
M1
5V
0V
CL =
1pF
+
vCL
vin+VTN
Fig. 4.1-18
PVHTP 50x10-6(1.8)2
= 8.10x107 . Thus, the NMOS transistor is in the
we have 2CL =
210-12
slow transition and the PMOS transistor is in the fast transition regimes.
Error due to NMOS:
176x10-18 + 0.5(1.58x10-15)
10810-12 176x10-18
(2.5+1.4-0)
Verror(NMOS) = -
10-12
2110x10-6
10-12
= -1.840mV
CMOS Analog Circuit Design
Page 4.1-23
Verror(PMOS) =
1.8- 610810-12 + 10-12 (5+1.4-2.5)
10-12
= 1.956mV
Net error voltage due to charge injection is 116V. This will vary with VS.
Page 4.1-24
VDD
M1
A
B
VDD
VA,B
10k
Switch On Resistance
1A
VDD
=1V
VDD=1V
8k
VDD
=1.5V
VDD=1.5V
6k
VDD=2V
4k
VDD
=2V
2k
VDD=2.5V
VDD=3V
M2
0
Fig. 4.1-22
Spice File:
Simulation CMOS transmission switch resistance
M1 1 3 2 0 MNMOS L=1U W=10U
M2 1 0 2 3 MPMOS L=1U W=10U
.MODEL MNMOS NMOS VTO=0.7, KP=110U,
+LAMBDA=0.04, GAMMA=0.4, PHI=0.7
.MODEL MPMOS PMOS VTO=-0.7, KP=50U,
+ LAMBDA=0.05, GAMMA=0.5, PHI=0.8
0V
0.5V
1V
1.5V
2V
2.5V
3V
VA,B (Common mode voltage)
Fig. 4.1-22A
VDD 3 0
VAB 1 0
IA 2 0 DC 1U
.DC VAB 0 3 0.02 VDD 1 3 0.5
.PRINT DC V(1,2)
.END
Result:
Low ON resistance over a wide voltage range is difficult as the power supply decreases.
CMOS Analog Circuit Design
Page 4.1-25
VDD
M3
Analog
Signal
Input
M4
Analog
Signal
Output
M5
V SS
M2
V Control
High State
M1
M1
Analog
Signal
Input
VSS
Analog
Signal
Input
Analog
Signal
Output
VDD
Analog
Signal
Output
M2
M2
High State
Low State
Page 4.1-26
(Prevents latchup)
To a
single
NMOS
switch
Vhi 5V
Vsub_hi
M1
0V
C2
C1
CL
3.3V
0V
0V
C2
Vhi = 2VDDCgate,NMOS switch + C2 + CL
CMOS Analog Circuit Design
Page 4.1-27
3.3V
C1
C2
CBulk
CStorage
0V
Fig. 4.1-225
Page 4.1-28
M1
M2
M3
M5
C1
C1
M4
M6
CLK_in
VDD
CLK_out
VSS
Fig. 4.1-23
Simulation:
3.0
Output
2.0
Volts
Input
1.0
0.0
-1.0
0.0
1.0
2.0
3.0
4.0
5.0
6.0
Time (s)
7.0
8.0
9.0
10.0
Fig. 4.1-24
T.B. Cho and R.R. Gray, A 10b, 20 Msample/s, 35mW Pipeline A/D Converter, IEEE J. of Solid-State Circuits, Vol. 30, No. 3m March 1995, pp.
166-172.
CMOS Analog Circuit Design
P.E. Allen - 2004
Page 4.1-29
M1
M3
vg
Boosted Clock
M4
VDD
M8
C1
M7
C2
C3
M5
M12
VDD
Input Signal
vg
M13
M10
M9
S
t
M11
Fig. 4.1-26
low: M7 and M10 make vg=0 and C3 charges to VDD, high: C3 connected to vGS11.
M7 reduces the vDS and vGS of M10 when = 0. M13 ensures that vGS8 VDD.
The parasitics at the source of M11 require this node to be driven from a low impedance.
A.M. Abo and P.R. Gray, A 1.5V, 10-bit, 14.3-MS/s CMOS Pipeline Analog-to-Digital Converter, IEEE J. of Solid-State Circuits, Vol. 34, No. 5,
May 1999, pp. 599-605.
CMOS Analog Circuit Design
P.E. Allen - 2004
Page 4.1-30
Page 4.2-1
i
vSG = v
vGS = v
i
-
VT
Fig. 4-2-1
Note that when the gate is connected to the drain of an enhancement MOSFET, the
MOSFET is always in the saturation region.
vDS vGS - VT
vD - vS vG - vS - VT vD - vG -VT
vDG -VT
Since VT is always greater than zero for an enhancement device, then vDG = 0 satisfies
the conditions for saturation.
Works for NMOS or PMOS
Note that the drain could be VT less than the gate and still be in saturation
CMOS Analog Circuit Design
Page 4.2-2
KW
i = iD = 2L (vGS - VT)2 = 2 (vGS - VT)2 and v = vGS = vDS = VT +
Small-Signal Characteristics:
The small signal model is a linearization of the large signal model at an operating point.
iD = 2 (vGS-VT)2(1+vDS) d + ID = 2 [vgs+(VGS-VT)]2[1+(vds+VDS)]
Page 4.2-3
i
AC Resistance
v
V
DC resistance = i = I
Q
DC Resistance
ID
VDS
VT
Fig. 4-2-2B
id
+
vgs
+
vbs
gmvgs
gmbsvbs
rds
+D
vds
-
Fig. 4.2-4
vds
1
1
AC resistance = i = g + g g
d
m
ds
m
where
gm = (VGS-VT) = 2ID
Page 4.2-4
id
B
S
G
B
+
vgs
+
vbs
-
gmvgs
gmbsvbs
rds
+D
vds
-
Fig. 4.2-4
where
iD vT
D iD vGS
gm
gmbs =
= gm
2 2|F| - VBS
It is very useful to simplify the small signal model when possible. The following are
reasonable guidelines for this simplification:
gm 10gmbs 100gds
CMOS Analog Circuit Design
Page 4.2-5
Fig. 4.2-5
2I
2100
VGS =
+ VT0 =
11010 + 0.7 = 1.126V
Thus let us guess at a gate-source voltage of 1.3V (to account for the bulk effect) and
calculate the resulting gate-source voltage.
Page 4.2-6
id
G,D,B
gmvgs
rds
gmbsvbs
+
vds = vgs
rac
vac
106
rac = 469 + 46.33 + 4 = 1926
If we had used the previous approximations of gm 10gmbs 100gds, then we could have
simply let
rac 1/gm = 1/469S = 2132
Probably the most important result of this approximation is that we would not have to
find VBS which took a lot of effort for little return.
CMOS Analog Circuit Design
Page 4.2-7
VDD
M2
ID
P
N (VDD-|VTP|)
VBias =
and ID = N(VBias-VTN)2
P
1 + N
Use the ratio of P/N to design VBias and the value of N to design
+
VBias
VTN +
M1
Fig. 4.2-7
VDD
2IBias
1
+ VT1 +
2IBias
2
IBias
M2
+ VT2
M1
VBias
Fig. 4.2-8
Page 4.2-8
VBias
B
RAB
Fig. 4.2-9
100A
VGS=10V
VGS=9V
60A
VGS=8V
VGS=7V
20A
VGS=2V
-20A
VGS=3V
VGS=4V
-60A
VGS=5V
VGS=6V
-100A
-1V
CMOS Analog Circuit Design
-0.6V
-0.2V
0.2V
0.6V
1V Fig. 4.2-95
P.E. Allen - 2004
Page 4.2-9
VDD
IBias
IBias
+
+
M1
VC
M2
iAB
A
+
VC
-
B
-
iAB
A
+
RAB
vAB
vAB
B
-
Fig. 4.2-10
vAB2
iD1 = 1 (vAB + VC - VT)vAB - 2
vAB2
iD2 = 2 (VC - VT)vAB - 2
vAB2
vAB2
iAB = iD1 + iD2 = vAB2 + (VC - VT)vAB - 2 + (VC - VT)vAB - 2
iAB = 2(VC - VT)vAB
1
R AB = 2(VC - VT)
CMOS Analog Circuit Design
Page 4.2-10
Vc=7V
6V
5V
I(VSENSE)
1mA
4V
W=15u
L=3u
VBS=-5.0V
3V
-1mA
-2mA
-2
-1
VDS
2 Fig. 4.1-11
VDS 10 0
VSS 5 0 DC -5
.DC VDS -2.0 2.0 .2 VC 3 7 1
.PRINT DC I(VSENSE)
.PROBE
.END
Page 4.2-11
R
R
i1
i2
iD2
VC2
iD3
v2
iD4
M1
i1
VSS
M2
M3
VSS
i2
M4
VC1
Fig. 4.2-12
Page 4.2-12
150uA
V C2 = 6V
5V
100uA
VBC =-5V
V3 =0V
VC1 =7V
I(VSENSE)
50uA
4V
3V
2V
- 50uA
-100uA
- 150uA
-3
-2
-1
V1-V2
Comments:
Good linearity and tunability.
Can be used as a multiplier.
CMOS Analog Circuit Design
Page 4.2-13
Linearity
Good
How
Controlled
VGS or
W/L
VC or W/L
Parallel MOSFET
Very
Good
VC1 - VC2
or
W/L
Double-MOSFET,
differential resistor
Poor
Restrictions
v (VC - VT)
Page 4.3-1
VDD
iOUT
VGG
and
1
rout = di /dv =
D DS
+
vOUT
-
1+VDS
D
;;
;;
;;
0
0 VGG-VT0
VGG
Slope = 1/rout
VDD
vOUT
Fig. 4.3-1
Page 4.3-2
Slope = 1/rout
+
iOUT
vOUT
-
;;
;;
;;
VMIN
VDD-VGG
VGG
VGG+|VT0| VDD
vOUT
Fig. 280-02
Page 4.3-3
10W/L W/L
0.1W/L
ID
Enhance
Channel
0
Provide
Current
VT
VGS
vGS
Fig. 280-03
VGS = VT0 + VON = Part to enhance the channel + Part to cause current flow
where
VON = VDS(sat) = VGS - VT0
2ID
K(W/L) for the simple current sink.
Note that VMIN can be reduced by using large values of W/L.
VMIN = VON = VDS(sat) =
Page 4.3-4
;
;
;
;
;
;
120
iOUT (A)
100
Slope = 1/Rout
80
+
vOUT
-
VGS1 =
1.126V
40
20
iOUT
10m
1m
60
Vmin
2
3
vOUT (Volts)
Comments:
VMIN is too large - desire VMIN to approach zero, at least approach VCE(sat)
Slope too high - desire the characteristic to be flat implying very large output resistance
(KN = 110A/V2, VT = 0.7Vand = 0.04V-1)
rds = 250k
Page 4.3-5
Loop
gain
=
M2
iOUT = gmR
iOUT
rout(w.fb.) = rout(w/o fb.)x [1+gmR] = rds(1+gmR)
M2'
+ vOUT
VGG
vS
-
Fig. 280-09
Page 4.3-6
iout
+
rds2
vout
VGG
Loop equation:
vs2
R
R
vg2 = vb2 = 0
vout = (iout-gm2vgs2-gmbs2vbs2)rds2
Fig. 280-10
+ ioutR
= iout(rds2+R) - gm2rds2vgs2 - gmbs2rds2vbs2
But,
vgs2 = 0 - vs2 = -ioutR and
vbs2 = 0 - vs2 = -ioutR
Therefore,
vout = iout[rds2 + R + gm2rds2R + gmbs2rds2R]
or
vout
rout = i
= rds2 + R + gm2rds2R + gmbs2rds2R gm2rds2R = 2R
( = gmrds)
out
A general principle emerges:
The output resistance of a cascode circuit R x (Common source voltage gain of the
cascoding transistor)
Page 4.3-7
iOUT
M2
M1
+
gm2vgs2
gmbs2vbs2
vOUT
VGG2
gm1vgs1
VGG1
-
rds1
rds2
+
vs2
-
vout
Fig. 280-11
Page 4.3-8
;;
;;
;;
;;
;;
;;
Example
Use the model parameters
KN=110A/V2, VT = 0.7 and N =
0.04V-1 to calculate (a) the smallsignal output resistance for the simple
current sink if IOUT = 100A and (b)
the small-signal output resistance for
the cascode current sink with IOUT =
100A. Assume that all W/L values
are 1.
Slope = 1/Rout
iOUT (A)
100
80
60
vOUT
40
20
VGG1 =
1.126V
Vmin
2
3
vOUT (Volts)
5
Fig. 280-12
Solution
(a) Using = 0.04 V-1 and IOUT = 100A gives rds1 = 250k = rds2. (b) Ignoring the
bulk effect, we find that gm1 = gm2 = 469S which gives rout = (250k)(469S)(250k)
= 29.32M.
Page 4.3-9
B. If the drain currents of two or more transistors are equal and the transistors are matched and operating in the saturation region, then the gatesource voltages are related by the W/L ratios (ignoring bulk effects).
If iD1 = iD2, then vGS1 = VT1 +
W 2/L2
W 1/L1 (vGS2 - VT2)
M2
+
vGS2
iD2
W2
L2
Fig. 290-02
iD1
+
vGS1
W1
L1
-
M2
+
vGS2
iD2
W2
L2
Fig. 290-03
or
if W2/L2 = W1/L1, then vGS1 = vGS2
CMOS Analog Circuit Design
Page 4.3-10
iOUT
2VT+2VON
M2
M4
iOUT
+
+
vDS2
+
+
VT+VON VT+VON vOUT
+
M3
VT+VON
+
VT+VON
M1
0
vOUT
Fig. 4.3-10
VT+2VON
Page 4.3-11
Since
2ID
VON =
K(W/L) ,
then if L/W is
quadrupled, then
VON is doubled.
VMIN = 2VON.
IREF
VDD
IREF
iOUT
M2 +
1/1
VON
+
+
VT+VON VT+2VON M3
M1 +
1/1
VON
+
VT+VON 1/1 -
M4
1/4
iOUT
VMIN
+
vOUT
0
2VON
vOUT
Fig. 290-04
Example
Use the cascode current sink configuration above to design a current sink of 100A
and a VMIN = 1V. Assume the device parameters of Table 3.1-2.
Solution
With VMIN = 1V, choose VON = 0.5V. Assuming M1 and M2 are identical gives
2IOUT
2100x10-6
W
L = KVON2 = 110x10-6x0.25 = 7.27
CMOS Analog Circuit Design
W 1 W2 W3
W4
=
=
=
7.27
and
L1 L2 L3
L4 = 1.82
P.E. Allen - 2004
Page 4.3-12
VDD
IREF
iOUT
+
M5
VT
1/1
-
M2 +
1/1
VON
1/4
+
+
VT+VON + M3
M1 +
VT+2VON
VON
VON
+
V
+V
T
ON
1/1
1/1
M4
+
vOUT
-Fig. 290-05
Design Procedure
1.) Since VMIN = 2VON = 2VDS(sat), let VON = 0.5VMIN.
W 1 W 2 W 3 W 5 2IREF
8IREF
2IREF
=
=
=
=
=
2.) VON =
K(W/L)
L1 L2 L3 L5 KV 2 KV
ON
MIN2
W4
2IREF
2IREF
IREF
3.) L4 =
=
=
K(VGS4-VT)2 K(2VON)2 2KVON2
CMOS Analog Circuit Design
Page 4.3-13
Output Only
D
+
Input
G
Only
Input
Only B
+
+
+
S
+
+
+
Fig. 4.3-12B
VDD
+
E
IREF
M5
M3
VT +2VON
+
VGS3
Fig. 4.3-12A
P.E. Allen - 2004
Page 4.3-14
iOUT(A)
80
60
40
20
0
VMIN
0
3
vOUT(V)
Page 4.3-15
VDD
IREF
Design procedure:
Same as the previous except
VON VMIN
R = IREF = 2IREF
For the previous example,
0.3V
R = 2100A = 1.5k
5
Fig. 290-06
VT+2VON
+
VON R
- VT+VON
iOU
+ M3
M4
VT
+ M1
VON
-
M2 +
VON
Fig. 290-0
Observation:
Note that the last several slides have been devoted to just getting the MOS cascode
current sink/source to have the same minimum voltage as the BJT!
T.L. Brooks and A.L. Westwick, A Low-Power Differential CMOS Bandgap Reference, Proc. of IEEE Inter. Solid-State Circuits Conf., Feb.
1994, pp. 248-249.
CMOS Analog Circuit Design
P.E. Allen - 2004
Page 4.3-16
iD3
M5
M7 M6
IREF
IREF
VO1
M1
Increasing vGS3
VGS3(norm)
M3
vOUT
M4
IREF
VGS3(max)
iOUT
M2
VDS3(min)
VDS3(sat)
vDS3
Fig. 290-08
Comments:
Achieves very high output resistance by increasing the loop gain due to the M4-M5
inverting amplifier.
gm4 gm3rds2gm4rds4
rds3gm3rds2gm4rds4
Loop gain = gm3rds2gds4+gds5
if
r
ds4
ds5
out
2
2
E. Sackinger and W. Guggenbuhl, A Versatile Building Block: The CMOS Differential Difference Amplifier, IEEE J. of Solid-State Circuits, vol.
SC-22, no. 2, pp. 287-294, April 1987.
CMOS Analog Circuit Design
P.E. Allen - 2004
Chapter 4 Section 3 (5/2/04)
Page 4.3-17
Page 4.3-18
IREF
+IB
VDD
VDD
iOUT
ID4A
IB
M4A M4B
+ +
VGS4AVGS4B
M1
M3 +
+
VDS2
IB
vOUT
IREF+IB
M2 +
VDS2
-
2ID4
KN(W4A/L4A) -
2IB
KN(W4B/L4B) =
2IB+2IRE
KN(W2/L
or
ID4
IB
IB+IREF
W 4A/L4A W 4B/L4B =
W 2/L2
A number of solutions exist. For example, let IB = IREF. This gives ID4A = 5.824IREF
assuming all W/L ratios are identical.
Fig. 290-10
Page 4.3-19
=
110A/V2(W/L) 1 + 1.1
+5V
+5V
+5V
Therefore,
2100A
110A 186A
10A iOUT
(2.049)
0.3V =
110A/V2(W/L)
+
M3
W 2100A2.0492
85/1
L = 110A/V20.32 = 84.8 85.
M4A M4B
85/1
85/1
With IB = 10A, then ID4A =
10A
10 + 110 = 186A
M1
85/1
vOUT
M2 110A
85/1
Fig. 290-11
Page 4.3-20
Comparison of the MOS Cascode Current Sink and Regulated Cascode Current
Sink
Close examination in the knee area reveals interesting differences.
Simulation results:
110
105
MOS Cascode
iOUT (A)
100
BJT Cascode
Regulated
MOS
Cascode
95
90
85
80
0.1
0.2
0.3
vOUT (V)
0.4
0.5
Fig. 290-12
Comments:
The regulated cascode current is smaller than the cascode current because the drainsource voltages of M1 and M2 are not equal.
The regulated cascode current sink has a smaller VMIN due to the fact that M3 can
have a drain-source voltage smaller than VDS(sat).
CMOS Analog Circuit Design
Page 4.3-21
rOUT
1
rds = D
VA
ro = C
gm2rds2rds1
Fro
gm2rds2rds1
VMIN
VDS(sat) =
VON
VCE(sat)
0.2V
VT + 2VON
2VCE(sat)
2VON
Cascode MOS
Cascode BJT
Minimum VMIN Cascode Current
Sink
Regulated Cascode Current Sink* rds3gm3rds2gm4(rds4||rds5) VT +VON
Minimum VMIN Regulated
VON
rds3gm3rds2gm4(rds4||rds5)
Cascode Current Sink*
* Unfortunately, the regulated cascode current sink has a dominant pole in the feedback
loop which can cause a pole-zero doublet which leads to a combination of fast and slow
time constants. For this reason, the regulated cascode circuit should only be used in
biasing applications unless the impact of this dynamic is understood.
Page 4.4-1
iin
iin
+
vin
-
iout
Current
Mirror
iout
Slope = 1/Rout
Slope
= 1/Rin
vout
-
Ai
1
VMIN (in)
iin
vin
VMIN (out)
vout
Output Characteristics
Fig. 300-01
Therefore, Rout, Rin, VMIN(out), VMIN(in), and Ai will characterize the current mirror.
CMOS Analog Circuit Design
Page 4.4-2
iO
+
vDS1
-
M2
M1
+
vGS
-
+
vDS2
Fig. 300-02
Page 4.4-3
1 + vDS2
Ratio Error
1 + vDS1
1 100 %
8.0
7.0
0.02
0.015
0.01
6.0
5.0
4.0
3.0
2.0
1.0
0.0
0.0
Fig. 300-03
CMOS Analog Circuit Design
1.0
2.0
3.0
vDS2 - vDS1 (volts)
4.0
5.0
Page 4.4-4
GS
T1
How do you analyze the mismatch? Use plus and minus worst case approach. Define
K = K2-K1 and K = 0.5(K2+K1)
K1= K-0.5K and K2= K+0.5K
VT = VT2-VT1 and VT = 0.5(VT1+VT2) VT1 =VT -0.5VT and VT2=VT+0.5VT
Substituting these terms into the above equation gives,
VT 2
K
1 +
1
2
2K 2(vGS-VT)
iO (K+0.5K)(vGS - VT - 0.5VT )
iI = (K-0.5K)(v - V + 0.5V )2 = K
VT 2
GS
T
T
1 1+
2(vGS-VT)
2K
Assuming that the terms added to or subtracted from 1 are smaller than unity gives
VT 2
VT 2
iO
2 VT
K
K
K
1 +
1 +
1
1
1
+
2K
2K 2(vGS-VT) 2(vGS-VT)
K (vGS-VT)
iI
Assume K/K = 5% and VT/(vGS-VT) = 10%.
Page 4.4-5
i
Ratio Error O 1 100 %
ii
16.0
14.0
12.0
10.0
iI = 3A
8.0
iI = 5A
6.0
iI = 10A
4.0
2.0
iI = 100A
0.0
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
10
9.0
Fig. 300-4
VT (mV)
Key: Make the part of VGS causing the current to flow, VON, more significant than VT.
Page 4.4-6
iI
;;;;;;;;;;
M2
GND
iO
M1
VDS1
-
M1
M2
+
VGS
+
VDS2
-
Fig. 300-5
Solution
We note that the tolerance is not multiplied by the nominal gain factor of 4. The ratio of
W2 to W1 and consequently the gain of the current amplifier is
iO W 2 20 0.1 1 (0.1/20)
0.1 0.1
0.1 0.4
4 1
1
4 1
=
=
=
4
iI W 1 5 0.1
20
5
20 - 20 = 4 - (0.03)
1 (0.1/5)
where we have assumed that the variations would both have the same sign (correlated). It
is seen that this ratio error is 0.75% of the desired current ratio or gain.
CMOS Analog Circuit Design
Page 4.4-7
;;
;;
;;
;
;;
;;
;;
;
;;
;;
;;
;
;;
;;
;;;;;; ;
iI
M2a
M2b
M1
iO
M2c
iI
M2d
GND
iO
M1
M2
GND
Fig. 300-6
Page 4.4-8
Ib
iI
iI
VT
+ M1
VON
-
iO
+
VT+VON
-
M2
VDD
Ib
iO
VT
M1
VON
+
- VT+VON
-
M2
Ib
Fig. 300-7
Page 4.4-9
+
gm3v3
iO
M3
M4
M1
S3=G2 -
vin
iin
gm1v1
-
M2
D3=G3=G4
+
rds3 v3
D4
+
rds4
gm4vgs4
D1=G1 +
rds1 v1
gm2vgs2
S1
-
S4
vout
iout
D2
rds2
-
S2
Fig. 310-018
Rout:
vout = rds4(iout-gm4vgs4) + rds2(iout-gm2vgs2)
vgs4 = -vs4 = -ioutrds2 and
But, iin = 0 so that v1 = v3 = 0
vout = iout[rds4 + rds2 + gm4rds2rds4] rds2gm4rds4
Rin:
1
1
1
1
2
Rin = g ||rds3 + g ||rds1 g + g g
m3
m1
m1
m3
m
VMIN(out) = VT + 2VON
VMIN(in) = 2(VT +VON)
Current gain match: Excellent since vDS1 = vDS2
CMOS Analog Circuit Design
vgs2 = 0
Page 4.4-10
IREF
M4
1/4
VDD
VDD
VDD
IO
1/1
M5
D5=G3
M2
+
1/1
M3
1/1
io
gm5vgs5
iin
M1
1/1
rds5
vin
gm3vgs3
= gm3vin
-
D3=S5 +
rds3 vs5
S3=G5 Fig. 310-02
Rout gm2rds2rds1
Rin = ?vin = rds5(iin-gm5vgs5)+vs5 = rds5(iin +gm5vs5)+vs5 = rds5iin+(1+gm5rds5)vs5
But, vs5 = rds3(iin - gm3vin)
vin = rds5iin + (1+gm5rds5)rds3iin - gm3rds3(1+gm5rds5)vin
vin rds5 + rds3 + rds3gm5rds5 1
Rin = iin = gm3rds3(1+gm5rds5) gm3
VMIN(out) = 2VON
VMIN(in) = VT + VON
Current gain is excellent because vDS1 = vDS3.
CMOS Analog Circuit Design
Page 4.4-11
I1
iin
VDD
I2
iout
iin
+
R
gm3vgs3
Rin = ?
+
+
M4
M3
vin = iinR + rds3(iin-gm3vgs3)
rds3
vin
+
vin
v2
M2
M1
v1
gm1vgs1
rds1
+ rds1(iin-gm1vgs1)
But,
vgs1 = vin-iinR
Self-biased, cascode current mirror
Small-signal model to calculate Rin.
Fig. 310-03
and
vgs3 = vin-rds1(iin-gm1vgs1) = vin-rds1iin+gm1rds1(vin-iinR)
vin = iinR+rds3iin-gm3rds3[vin-rds1iin+gm1rds1(vin-iinR)]+rds1[iin-gm1(vin+iinR)]
vin[1+gm3rds3+gm1rds1gm3rds3+gm1rds1]
= iin[R+rds1+rds3+gm3rds3rds1+ gm1rds1gm3rds3R]
R + rds1 + rds3 + gm3rds3rds1 + gm1rds1gm3rds3R 1
Rin =
g
+R
1 + gm3rds3 + gm1rds1gm3rds3 + gm1rds1
m1
Rout gm4rds4rds2
VMIN(in) = VT + 2VON VMIN(out) = 2VON Current gain matching is excellent
CMOS Analog Circuit Design
Page 4.4-12
iO
M3
+
gm3vgs3
iin
M1
M2
+ vgs3
+
vin gm1vgs1
rds1
gm2vgs2
rds3
vout
+
rds2
vgs2=vgs1
-
Fig. 310-09
1+gm3rds2+gm1rds1gm3rds3
1+gm3rds2+gm1rds1gm3rds3
gm1rds1gm3rds3
Rout = rds3+rds2
1 + gm2rds2
gm2
Page 4.4-13
gm2
m2 ds2 ds3
gm1gm3vgs3
vgs3 = vin - vgs1= vin vgs3 =
gm2
vin
gm1gm3
1 + gm2
gm1gm3 vin
gm2 +gm3
iin gm2 +gm3
Rin = g g
m1 m3
VMIN(in) = 2(VT+VON)
VMIN(out) = VT + 2VON
Current gain matching - poor, vDS1 vDS2
Page 4.4-14
Evolution of the Regulated Cascode Current Mirror from the Wilson Current
Mirror
iI
iO
M3
iI
iO
M3
M1
M1
M2
M2
VBias2
Page 4.4-15
I
ii I
VDD
IBias
VDD
VDD
IO
io
M3
M1
M2
M4
FIG. 310-11
Rout gm2rds3
1
Rin
gm4
VMIN(out) = VT+2VON (Can be reduced to 2VON)
(Can be reduced to VON)
VMIN(in) = VT+VON
Current gain matching - good as long as vDS4 = vDS2
CMOS Analog Circuit Design
Page 4.4-16
SUMMARY
Summary of MOS Current Mirrors
Current
Mirror
Accuracy
Output
Resistance
Input
Resistance
Simple
Poor
rds
Cascode
Excellent
gmrds2
Wide Output
Swing
Cascode
Self-biased
Cascode
Wilson
Excellent
Regulated
Cascode
CMOS Analog Circuit Design
Minimum
Input
Voltage
VT+VON
VT+2VON
2(VT+VON)
gmrds2
1
gm
2
gm
1
gm
Minimum
Output
Voltage
VON
2VON
VT+VON
Excellent
gmrds2
1
R + gm
2VON
VT+2VON
Poor
gmrds2
2
gm
2(VT+VON) VT+2VON
GoodExcellent
gm2rds3
1
gm
VT+2VON
(min. is
2VON)
VT+VON
(min. is
VON)
P.E. Allen - 2004
Page 4.5-1
Noise
Temperature
Powe
r Sup
ply
Fig. 4.5-1
Page 4.5-2
Ideally, we want SV
DD
Page 4.5-3
VCC
IIN
IIN
IOUT
IC1
IOUT
ID1
IB1 IB2
Q1
Q2
M1
M2
Fig. 360-02
IOUT
VCC-VBE 1
R
1+ 2
F
IOUT
IREF
VDD-VGS
=
R
VDD -
2IIN
1 - V T
R
IREF
S VCC = 1
S VDD = 1
Page 4.5-4
2
K'(W2/L2) + 4R2VON1
2R2
2IIN
K'(W1/L1)
Differentiating IOUT with respect to VDD gives,
dIOUT
dVON1
1
1
=
2 IOUT dVDD
2/(K' W2/L2)+ 4R2VON1 dVDD ,
where
VDD
IIN
R1
ID1
M1
IOUT
M2
R2
Fig. 360-04
VON1 =
IREF
IOUT
S VDD =S VDD
IIN
IIN
IIN
VON1
VON1
= V
S 4VON12 SVDD = 0.5SVDD
ON22+4 IOUTR2VON1 VDD
P.E. Allen - 2004
Page 4.5-5
Example 4.5-1
For the MOS Widlar current reference, find IOUT if IIN = 100A, R2 = 4k, K =
200A/V2, and W2/L2 = W1/L1 = 25. Assume the temperature is 27C and that n = 1.5.
Find the sensitivity of IOUT with respect to VDD.
Solution
2IIN
2100
VON1 =
=
20025 = 0.2V
K'(W1/L1)
2
20025 + 4(0.004)0.2
IOUT =
A = 5 A IOUT = 25A
20.004
Note that VON2 = VON1 - IOUTR2 = 0.2-(25)(0.004) = 0.1V > 2nVt = 78mV so both
transistors are in strong inversion.
For the sensitivity calculations, assume that VDD >> VGS1. Therefore IIN VDD/R1.
-
2
20025 +
IIN
VON1
VON1
S VDD = 4VON22 SVDD 4VON22 = 0.5
Therefore, a 10% variation in VDD causes a 5% variation in IOUT.
IREF
Page 4.5-6
IIN
M2
Fig. 360-7
Transfer Characteristics:
2IIN
K'(W1/L1)
Weak Inversion Operation:
IIN
VGS2 VT nVt ln (W1/L1)IT IINR
IOUT
M1
VON1 =
VDD
Circuit:
1.6
1.4
Weak Inversion
1.2
1.0
IOUT(A)
0.8
0.6
0.4
Strong Inversion
0.2
0
6
IIN(A)
10
Fig. 360-8
Page 4.5-7
VDD
IIN
IOUT
R1
Operation:
2IIN
K'(W1/L1)
R2
VGS1 VT +
IOUT = R2 =
VT
R2 if VT > VON1
The sensitivity of IOUT with respect to VDD is
IOUT
S VDD
V ON1
= IOUTR2
IIN
VON1
M1
R2
IIN
Fig. 360-10
DD
IOUT
S VDD
M2
ID1
1, then
0.1
= 21.1 = 0.045
Page 4.5-8
VDD
R1
M2
+
+
R2
M1 V
VREF
R2
VREF = R +R VDD
1 2
VREF
S VDD =1
REF
VREF =
VDD (P/N)
VTN + (P/N) (VDD-|VTP|)
Fig. 370-01
VREF
S VDD
=1
Page 4.5-9
VDD
IBias
VREF
Fig. 370-02
Page 4.5-10
VDD
R
R
R1
vout
VREF
VREF
-
R2
Fig. 370-03
VREF = VGS = VT +
2(VDD-VREF)
R
1
or VREF = VT - R +
2(VDD-VT)
1
+
R
(R)2
VREF
S VDD
VDD
= 1 + (VREF-VT)R VREF
VDD
SVREF = 0.283
P.E. Allen - 2004
Page 4.5-11
VCC
R
R
+
R1
VREF
VREF
vout
R2
Fig. 370-04
kT
VREF = VEB = q ln Is
VCC VEB VCC
R
R
kT VCC
VREF q ln RIs
VREF
1
1
SVCC = ln[VCC/(RIs)] = ln(I/Is)
If VCC=5V, R = 4.3k and Is = 1fA,
then VREF = 0.719V.
I=
VREF
Also, S V
CC
= 0.0362
S VCC
CC
Page 4.5-13
VDD
Temperature coefficient of VBV (mV/C)
R
i
VDD
IQ
VBV
5
4
3
2
1
-1
10
VBV
-2
-3
VDD
Fig. 370-05
VREF = VBV
VREF
SVDD
VREF VDD
vref VDD
rZ VDD
= V V
=
DD
REF vdd VBV rZ + R VBV
Page 4.5-14
M4
I2
I1
Startup
I5
I6
M6
M1
+
VGS1 R
-
0V
2
K'NW (V
GS1 - VT)
2L
Desired
operating
V
point
I2 = GS1
R
I1 =
M5
M3
M2
M8
IQ
Undesired
operating
point
VQ
v
Fig. 370-06
Principle:
If M3 = M4, then I1 I2. However, the M1-R loop gives VGS1 = VT1 +
2I1
KN(W1/L1)
2I1
VGS1 VT1 1
=
+
R
R
KN(W1/L1)
R
2VT1
VT1
1
1
1
The output current, Iout = I1 = I2 can be solved as Iout = R +
+
+
1R ( R)2
1R2 R
1
Page 4.5-15
VDD
Fig. 370-07
RB 1 6 100KILOHM
.OP
.DC VDD 0 5 0.1
.MODEL N NMOS VTO=0.7 KP=110U
GAMMA=0.4 +PHI=0.7 LAMBDA=0.04
.MODEL P PMOS VTO=-0.7 KP=50U
GAMMA=0.57 +PHI=0.8 LAMBDA=0.05
.PRINT DC ID(M1) ID(M2) ID(M5)
.PROBE
.END
Page 4.5-16
120A
M4
M5
M3C
MC4
MC5
RB
M7
80A
I1
RON
I2
M2
M8
Startup
ID2
100A
M1
+
VGS1 R
0V
I5
ID1
60A
40A
20A
0
VDD
5
Fig. 370-
M7 6 6 5 9 N W=20U L=1U
RB 1 6 100KILOHM
.OP
.DC VDD 0 5 0.1
.MODEL N NMOS VTO=0.7
KP=110U GAMMA=0.4 PHI=0.7
LAMBDA=0.04
.MODEL P PMOS VTO=-0.7
KP=50U GAMMA=0.57 PHI=0.8
LAMBDA=0.05
.PRINT DC ID(M1) ID(M2) ID(M5)
.PROBE
.END
P.E. Allen - 2004
Page 4.5-17
M3
M4
M5
i2
M6 I
1
I2
M1
Desired
operating
point
I5
M2
+
VEB1
Q1
M7
i2=i1
Undesired
operating
point
Startup
i2=VTln(i1/Is)/R
VR
i1
Fig. 370-09
VEB1
Iout = I2 = R
BJT can be a MOSFET in weak inversion.
Page 4.5-18
I1
M4
VT
VT
VT+VON
I2
VON
M1
M2
VT+VON
R
VSS
VR
Fig. 4.5-8A
Page 4.5-19
VREF
SV
PP
1
<1
<<1
<<1
<<1
Page 4.5-20
i IsexpVt
(ln Is) 3
VGO
VGO
1 Is
=
=
+
-VGO
TVt
TVt
Is T
T
T
Is = KT3exp Vt
Resistors:
(1/R)(dR/dT) ppm/C
CMOS Analog Circuit Design
Page 4.5-21
RIs
VDD
R
+
VREF
dR
VREF
Vt
dVREF
dIs VREF-VGO
Vt
dVREF 3Vt Vt dR
VREF-VGO
dR 3Vt
V
t
dVREF
T
RdT - T VREF-VGO
dR 3Vt
dT =
V
t RdT - T
T
Vt
1 + V -V
DD
REF
3Vt
1 dVREF VREF-VGO Vt dR
TCF = V
=
VREFT VREF RdT VREFT
REF dT
If VREF = 0.6V, Vt = 0.026V, and the R is polysilicon, then at 27K the TCF is
0.6-1.205 0.0260.0015 30.026
- 0.6300 = 33110-6-65x10-6-433x10-6 =-3859ppm/C
TCF = 0.6300 0.6
CMOS Analog Circuit Design
Page 4.5-22
1
VREF = VT - R +
2(VDD-VT)
1
+
R
(R)2
R dT
2R
T
1
1 + 2R (V V )
DD
REF
VDD
R
+
VREF
Fig. 380-02
R dT
2R
T
TCF =
1
VREF(1 + 2R (V V ))
DD
REF
+
Page 4.5-23
2(5 0.7) 1 2
+ 22 = 1.281V
22
5 1.281 1.5
-6
1500
10
dVREF
222
300
=
= -1.189x10-3V/C
Now,
1
dT
1+
222 (5 - 1.281)
The fractional temperature coefficient is given by
1
TCF = 1.189 10-3 1.281 = 928 ppm/C
2.310-3 +
R
R
1R
1
1R)2
Although we could grind out the derivative of Iout with respect to T, the temperature
performance of this circuit is not that good to spend the time to do so. Therefore, let us
assume that VGS1 VT1 which gives
VT1
dIout 1 dVT1 1 dR
dT = R dT - R2 dT
Iout R
In the resistor is polysilicon, then
1 dIout
1 dVT1 1 dR - 1 dR -2.3x10-3
-1.5x10-3 = -4786ppm/C
TCF = Iout dT = VT1 dT - R dT = VT1 - R dT = 0.7
Base-emitter referenced source:
VBE1
The output current was given as, Iout = I2 = R
1 dVBE1 1 dR
The TCF = V
- R dT
BE1 dT
1
If VBE1 = 0.6V and R is poly, then the TCF = 0.6 (-2x10-3) - 1.5x10-3 = -4833ppm/C.
CMOS Analog Circuit Design
Page 4.5-25
VDD
Technique to Make gm Dependent on a Resistor
Consider the following circuit with all transistors having a
W/L = 10. This is a bootstrapped reference which creates a M3
M4
Vbias independent of VDD. The two key equations are:
I3 = I4 I1 = I2
M2D
and
M1
+
VGS1 = VGS2 + I2R
M2A
M2B M2CVBias
Solving for I2 gives:
R=5k
2I2
2I1 1
VGS1-VGS2 1 2I1
= R 1 I2 =
Fig. 4.5-11
R
2 = R 1 1 - 2
1
1
1
I2 = R 2 I2 = I1 =
=
= 18.18A
1
21R2 2110x10-61025x106
Now, Vbias can be written as
2I2
1
1
+ 0.7 = 0.1818+0.7 = 0.8818V
Vbias=VGS1= 1 +VTN = 1R+VTN =
-6
110x10 105x103
Any transistor with VGS = Vbias will have a current flow that is given by 1/2R2.
1
1
2
=
g
=
Therefore,
gm = 2I =
m
R
2R2 R
(This means that the temperature dependence of gm will be that of 1/R which can be used
to achieve temperature controlled performance.)
CMOS Analog Circuit Design
Page 4.5-26
VREF
S VDD
TCF
<1
>1000ppm/C
<<1
>1000ppm/C
<<1
Can be very small
Good if currents >1000ppm/C
are matched
Good if currents >1000ppm/C
are matched
Comments
BV too large
Requires startup circuit
Requires startup circuit
Page 4.6-1
I1
T
+
-VBE Vt
+0.085mV/C
T
Vt = kT
q
KVt
Fig. 390-01
JC = W B
exp Vt
Page 4.6-3
where, = 3
JC = N AW B DT3 exp Vt
Vt
=
exp
JC0
T
T0
T0
k
or
JC
T
q
T
ln JC0 = lnT0 + kT VBE - VGO - T0 (VBE0 - VGO)
where VBE0 is the value of VBE at T = T0.
5.) Solving for VBE from the above results gives,
kT T0 kT JC
T
T
VBE(T) = VGO1 - T0 + VBE0T0 + q ln T + q lnJC0
CMOS Analog Circuit Design
Page 4.6-4
VBE VGO T VGO VBE0 kT ln(T0/T) T0 (kT/q) kT lnJC0 k JC
+ln T T + q +q lnJC0
T = T 1-T0- T0 + T0 + q
T
T=T
0
T
T0 + T0 + q
T
T
T0 T
T0 T2 T
T
T
Therefore,
VGO VBE0 k k
VBE0 - VGO
VBE |
VBE |
k
=
+
+
or
=
+
(
)
T=T
T=T
0
0
T
T0
T0
T
T0
q q
q
Typical values of and are 1 and 3.2. If VBE0 = 0.6V, then at room temperature:
VBE |
0.026
0.6-1.205
0.6-1.205-0.1092
=
=
+
(1-3.2)
= -1.826mV/C
T=T
0
T
300
300
300
Page 4.6-5
VDD
IC2
IC1
+ VBE Q1
AE1
Q2
AE2
Fig. 390-02
kT JC1
VBE = VBE1- VBE2 = q lnJC2
- Find (VBE)/T,
Page 4.6-7
VREF
=0
T
1.280
VREF= 0
T0 = 300K
T
1.270
1.260
1.250
VREF
=0
T
1.240
T0 = 200K
TC
-60
-40
-20
20
40
60
80
100
120
Fig. 4.6-3
Bandgap curvature correction will be necessary for low ppm/C bandgap references.
R.J. Widlar, New Developments in IC Voltage Regulators, IEEE J. of Solid-State Circuits, Vol. SC-6, pp. 2-7, February 1971.
CMOS Analog Circuit Design
P.E. Allen - 2004
Page 4.6-9
R3
M3
IREF
R4
R2
VREF Q1
I1
Q2
R1
M1
VSS
I2
M2
Fig. 390-05
I2
Is2
AE2
VBE1 - VBE2
V t I1
Vt
Vt
ln
- ln
=
=
=
ln
ln
R2
R2 Is1
R2
R2 AE1
Is2
Is1
if I1 = I2 which is forced by the current mirror consisting of M1 and M2.
R1
AE2
VREF = VBE1 + I1R1 = VBE1 + R2 lnAE1 Vt = VBE1 + KVt
While an op amp could be used to make I1 = I2 it suffers from offset and noise and leads to
deterioration of the bandgap temperature performance.
VREF is with respect to VDD and therefore is susceptible to changes on VDD.
I2 =
Page 4.6-10
VDD
M7
M8
M9
M5
M6
M10
M3
M4
+
M1
M2
x1
xn
VSS
CMOS Analog Circuit Design
Q3
Q2
Q1
VREF
I2
I1
kR
xn
Fig. 390-06
P.E. Allen - 2004
Page 4.6-11
VDD
+
VR1 R1
-
=
exp
ID3
W 4/L 4
Vt
where
W 1W 4L 2L 3
VR1
VR1 = Vt lnL 1L 4W 2W 3 and IR1 = R1
CMOS Analog Circuit Design
Chapter 4 Section 6 (5/2/04)
ID6
M6
+
+
VR2
VREF
Q5
Fig. 390-07
T
T0
W6L3 R2
T
VREF = L 6W 3 R1 Vt lnL 1L 4W 2W 3 + VGO1 - T0 + VBE0 T0 + 3Vt ln T
To achieve VREF/T = 0 at T = T0, we get
VREF k R2W 6L 3 W 1W 4L 2L 3 VGO
VBE0
3k
ln
=
+
+
T
T0
T0
q
q R1 L 6W 3
L 1L 4W 2W 3
Therefore,
R 2W 6L 3 W 1 W 4 L 2 L 3 q
R1L6W3 lnL 1L 4W 2W 3 = kT0 (VGO - VBE0) - 3
Under the above constraint, VREF has an zero TCF at T = T0 and has a value of
T0
3kT
3kT
VREF = VGO + q 1 + ln T = VGO + q
Practical values of VREF/T for the weak inversion bandgap are less than 100 ppm/C.
CMOS Analog Circuit Design
Page 4.6-13
Voltage
VBE
VPTAT
VPTAT2
VRef = VBE + VPTAT + VPTAT2
Temperature
Fig. 400-01
compensation
I. Lee et. al., Exponential Curvature-Compensated BiCMOS Bandgap
References, IEEE Journal of Solid-State Circuits, vol. 29, no. 11, pp. 1396-1403,
Nov. 1994.
Nonlinear cancellation
G.M. Meijer et. al., A New Curvature-Corrected Bandgap Reference, IEEE
Journal of Solid-State Circuits, vol. 17, no. 6, pp. 1139-1143, December 1982.
CMOS Analog Circuit Design
Page 4.6-14
Vt 2IPTAT
VREF = R2 + R3 lnINL + Iconstant + IPTAT R1
VDD
IVBE+INL
IPTAT
VDD VDD
Page 4.6-15
BT
BT
Vin
VREF = VBE + AT + (1+) R VBE + AT + R
where
I=AT
I=BT
A and B are constant
T = temperature
The temperature dependence of is
R BT
1+
VREF
Fig. 400-03
BTe1/T
C
Not good for small values of Vin.
Vin VREF + Vsat. = VGO + Vsat. = 1.4V
VREF = VBE(T) + AT +
Page 4.6-16
VCC
IPTAT
Q8
Q4
IConstant
V
= REF
R2
Q7
IPTAT
Note that,
IPTAT Ic T 1 = 1
Q6
Q3
0
and
Iconstant Ic T = 0,
Q2
Q5
VBE
Previously we found,
T
T
VREF
Q1
VBE
VBE(T) VGO - T0 VGO-VBE(T0) -( -)Vt lnT0
R2 VREF
VPTAT
R1
VPTAT R1
so that
T
T
VBE(IPTAT) =VGO-T VGO-VBE(T0)-(-1)Vt lnT
Conventional
Curvature Corrected
0
0
Bandgap Reference
Bandgap Reference
Fig. 400-04
and
T
T
VBE(IConstant) =VGO - T0 VGO -VBE(T0) -Vt lnT0
Page 4.6-17
VDD
IPTAT
IVBE(PTAT)
IVBE(Const)
Iconst
IVBE
VDD
KIPTAT
Iconst
+
VGS(ZTC)
+
Q1
Q2
VREF
R1
R0
R2
Constant Current
Generator
040629-01
VBE(PTAT)
VBE(Const)
R
0
R1
R2
R0
T R0
T
T
T
= R1 VGO-T0VGO-VBE(T0)-(-1)Vt lnT0- R2 VGO-T0VGO-VBE(T0)-Vt lnT0
R0 R0
R0
R0
R1 -1
Let R1 - R2 = 1 and R1 (-1) = R2 R2 = to cancel the nonlinear curvature term.
CMOS Analog Circuit Design
Chapter 4 Section 6 (5/2/04)
VOS
VR1 = -VOS + Vt lnR3AE21 + iC1R2
V R1
R2
VREF = VBE2 - VOS + iC1R2 = VBE2 - VOS + R1 R2 = VBE2 - VOS + R1 VR1
R2AE1
R2 R2
VOS
VREF = VBE2 - VOS1+ R1 + R1 V t ln R3AE21 - iC1R2
CMOS Analog Circuit Design
Page 4.6-19
IR1
IREF
IR2
VREF
R2
+
VGS
Fig. 400-06
V
R
VGS = IR2R2 = R R = R VREF
dVGS R2 dVREF VREF dR2 R2 dR1 R2 dVREF dR2 dR1
dT = R1 dT + R1 dT - R 2 dT = R1 dT + dT - dT
1
dR
dR2
1
REF
Page 4.6-20
M7
M8
M9
M5
M6
M10
Bandgap
Voltage,
R4
VBG
R3
M11
M13
M15
M17
M19
M12
M14
M16
M18
M20
M4
M3
M1
M2
IPTAT
R1
Q2
Q1
R2
Q3
IREF
To Slave
Bias Ckt.
To Slave
Bias Ckt.
Rext
xn
Fig. 400-07
Constant current:
VBG
IREF = R
where
ext
CMOS Analog Circuit Design
VT
VBG = VBE3 + IPTATR2 = VBE3 + R1 ln(n)R2
P.E. Allen - 2004
Page 4.6-21
Ib
VPBias2
VNBias2
VNBias1
Fig. 400-08
We will examine bandgap voltage references once again when we consider low
voltage circuits in Section 6 of Chapter 7.
Page 4.7-1
CHAPTER 4 - SUMMARY
This chapter covered the analysis and design of sub-blocks or subcircuits including:
- Switches
- MOS diode and floating resistor realizations
- Current sinks and sources
- Current mirrors (amplifiers)
- Current and voltage references - Bandgap reference
Subcircuits represent primitives of circuit design and do not stand alone
The current sink/source is an important subcircuit which is used for biases and ac loads
A current sink/source is characterized by
1.) The independence of the current on the voltage across it (rout)
2.) The voltage range over which the current is not independent of the voltage (VMIN )
A current mirror is characterized by
1.) The independence of the output current on the voltage across it (rout large)
2.) The output voltage range over which output current is dependent (VMIN (out))
3.) The independence of the input voltage on the input current (rin small)
4.) The range of input voltage over which the input current is independent (VMIN(in))
5.) The accuracy of the current out as a function of the current in ratio.
A voltage or current reference is independent of power supply and temperature
The bandgap reference is the best realization of a voltage reference
CMOS Analog Circuit Design
Page 5.0-1
Blocks or circuits
(Combination of primitives, independent)
Chapter 5
Sub-blocks or subcircuits
(A primitive, not independent)
Fig. 5.0-1
CMOS Analog Circuit Design
Page 5.0-2
Biasing
Circuits
Input
Differential
Amplifier
Second
Gain
Stage
Inverter
Source
Current
Coupled Pair Mirror Load
Current
Sink Load
Output
Stage
Source
Follower
Current
Sink Load
Fig. 5.0-2
Page 5.0-3
VDD
+
V T+
2VON
+
VT+VON
-
+
VEB +
VEB +
VEC(sat)
-
MOS Loads
IBias
BJT Loads
IBias
IBias
IBias
+
VBE +
VCE(sat)
-
V T+
2VON
+
VBE
-
- V ++V
T ON
MOS Transconductors
BJT Transconductors
Fig320-01
Page 5.1-1
Page 5.1-2
Inverters
The inverting amplifier is an amplifier which amplifies and inverts the input signal.
The inverting amplifier generally has the source on ac ground or the common-source
configuration.
Various types of inverting CMOS amplifiers:
M2
VDD
M2
M2
ID
vIN
ID
vOUT
M1
vOUT
vIN
M1
ID
vIN
vOUT
M2
VGG2
ID
vIN
M1
Active
Active
Depletion
NMOS Load PMOS Load NMOS Load
Inverter
Inverter
Inverter
M2
I
vOUT vIN D
vOUT
M1
Current
Source Load
Inverter
M1
Pushpull
Inverter
Fig. 5.1-1
We will consider:
Active PMOS Load Inverter (active load inverter)
Current Source Load Inverter
Push-pull Inverter
Page 5.1-3
5V
vIN=2.5V
ID (mA)
M2
0.3
M1
vIN=2.0V
0.2
+
vIN
M2
0.1
vIN=1.5V
0.0
vOUT
A,B
vOUT
+
vOUT
W1 = 2m
L1 1m
-
vIN=1.0V
5
M2 cutoff
M2 saturated
3
2
0
0
d
ate
r
u
D
at ve
1 s cti
M 1a
M
E
F
1
Fig. 320-02
W2 = 1m
L2 1m
ID
2v
3
IN
J K
Page 5.1-4
2
vDS1
(vOUT)2
iD = 1(vGS1 VT)vDS1 2 = 1 (VDD VT)(vOUT ) 2
Page 5.1-5
gm2vgs2
M2
ID vOUT G1
D1=D2=G2
+
vIN
vin gm1vgs1
M1
S1=B1
S2=B2
Rout
rds2
rds1
+
vout
-
+
vin
gm1vin
-
rds1
gm2vout
rds2
+
vout
Fig. 320-03
1/2
=
K' L W
vin gds1 + gds2 + gm2
gm2
P 1 2
The small-signal output resistance can also be found from the above by letting vin = 0 to
get,
1
1
Rout = gds1 + gds2 + gm2 gm2
CMOS Analog Circuit Design
Page 5.1-6
Vin
CM
+
Vin gmVin
-
+
Rout
Vout
-
Cout
sCM(Vout-Vin) + gmVin
Cgs1
+ GoutVout + sCoutVout = 0
Vout(Gout + sCM + sCout) = - (gm sCM)Vin
Fig. 320-04
s
sCM
g
R
1
1- gm
m out
z1
Vout
-(gm sCM)
1 - p1
where
gm1
and z1 = CM
1
p1 = Rout(Cout+CM) ,
gm = gm1,
and
1
Rout = [gds1+gds2+gm2]-1 gm2 , CM = Cgd1 , and Cout = Cbd1+Cbd2+Cgs2+CL
CMOS Analog Circuit Design
Page 5.1-7
dB
20log10(gmRout)
0dB
z1
|p1| -3dB
log10f
Fig. 5.1-4A
Observation:
The poles in a MOSFET circuit can be found by summing the capacitance connected
to a node and multiplying this capacitance times the equivalent resistance from this node
to ground and inverting the product.
Page 5.1-8
Page 5.1-9
5V
vIN=2.5V
2.5V
ID (mA)
0.3
0.2
KJIH F
M1
vIN=2.0V
M2
0.1
D
C
0.0
0
M2
A,B
vOUT
vIN=1.5V
+
vOUT
W1 = 2m
L1 1m
-
vIN=1.0V
5 A B C
D
4
vOUT
+
vIN
W2 = 2m
L2 1m
ID
M2 active
3 M2 saturated
ed
rat
u
t
a ve
1 s ti
M 1 ac
M
2
E
1
0
0
2v
3
IN
J K
Fig. 5.1-5
vOUT 3.2V
P.E. Allen - 2004
Page 5.1-10
= (VDD - VT1)1 -
1-
2
1
Page 5.1-11
ID
VGG2
vIN
S2=B2
M2
vOUT G1
+
vin
M1
-
rds2
D1=D2
gm1vgs1
rds1
Rout
+
vout
-
+
vin
gm1vin
-
rds1
S1=B1=G2
rds2
+
vout
Fig. 5.1-5B
Midband Performance:
2K'NW 1
vout
gm1
1
1
1
1/2 1
=
=
!!!
and
R
=
out
vin gds1 + gds2 L1ID 1 + 2
D
gds1 + gds2 ID(1 + 2)
log|Av|
Amax
Amax
10
Amax
100 Weak
inversion
Amax
1000
0.1A
1A
Strong
inversion
10A
100A
1mA
10mA
ID
Fig. 5.1-6
Page 5.1-12
gmRout 1 - z
1
Vout(s)
=
Vin(s)
s
1 - p1
M2
Cgd2
Cgd1
Vin
CM
Cbd2
Vout
Cbd1
M1
CL
+
Vin gmVin
-
+
Rout
Cout
Vout
-
Fig. 5.1-4
gm
1
gm = gm1,
p1 = Rout(Cout+CM) , and z1 = CM
1
and Cout = Cgd2 + Cbd1 + Cbd2 + CL CM = Cgd1
and Rout = gds1 + gds2
Therefore, if |p1|<|z1|, then the 3 dB frequency response can be expressed as
gds1 + gds2
-3dB 1 = C
gd1 + Cgd2 + Cbd1 + Cbd2 + CL
where
Page 5.1-13
1101 3-0.7
Page 5.1-14
v =4.5V
vIN=5.0V IN
vIN=4.0V
0.8
vIN=2.0V
ID (mA)
vIN=3.5V
5V
vIN=0.5V
vIN=1.0V
vIN=1.5V
M2
0.6
vIN=2.5V
0.2
H
I
0.0
0
F
vIN=3.0V
J,K 1
vOUT
vIN=2.0V
D
+
vOUT
W1 = 1m
L1 1m
-
M1
vIN
vIN=3.5V vIN=4.5V
vIN=2.5V
0.4
G
W2 = 2m
L2 1m
ID
vIN=3.0V
vIN=1.5V
vIN=1.0V
4 CA,B 5
A B C
D
E
vOUT
ed
rat
u
t
a ve
1 s ti
M 1 ac
M
ve
cti ted
2 a tura
M sa
2
M
2
1
F
G
0
0
Note
the railto-rail
output
voltage
swing
2v
3
IN
J K
Fig. 5.1-8
Page 5.1-15
CM
M2
+
M1
+
vin
-
gm1vin
rds1
gm2vin
rds2
Cout
+
vout
-
vout
vin
Fig. 5.1-9
=
=
(2/I
)
D
vin
1 + 2
gds1 + gds2
1
Rout = g + g
ds1
ds2
gm1+gm2 gm1+gm2
z = CM = Cgd1+Cgd2
and
(gds1 + gds2)
p1 = Cgd1 + Cgd2 + Cbd1 + Cbd2 + CL
If z1 > |p1|, then
gds1 + gds2
-3dB = C + C + C + C + C
gd1
gd2
bd1
bd2
L
CMOS Analog Circuit Design
Page 5.1-16
Page 5.1-17
M2
VDD
M2
eout2
en12
vin
Noise
Free
MOSFETs
M1
eout2
eeq2
vin
Noise
Free
MOSFETs
M1
Fig. 5.1-10
Approach:
1.) Assume a mean-square input-voltage-noise spectral density en2 in series with the
gate of each MOSFET.
(This step assumes that the MOSFET is the common source configuration.)
2.) Calculate the output-voltage-noise spectral density, eout2 (Assume all sources are
additive).
3.) Refer the output-voltage-noise spectral density back to the input to get equivalent
input noise eeq2.
4.) Substitute the type of noise source, 1/f or thermal.
CMOS Analog Circuit Design
Page 5.1-18
gm2 en2
2
2
eout2
eout2
3.) eeq2 = en12 1 + gm1 en1
en12
eeq2
vin
M1
M1
Up to now, the type of noise is not defined. vin
*
*
Fig. 5.1-10
1/f Noise
KF
B
Substituting en2= 2fCoxWLK = fWL , into the above gives,
B1
K'2B2 L1 1/2
1/2
2
eeq(1/f) = fW 1L 1 1 + K'1B1 L2
To minimize 1/f noise, 1.) Make L2>>L1, 2.) increase the value of W1 and 3.) choose M1
as a PMOS.
Thermal Noise
8kT
Substituting en2= 3gm into the above gives,
W L K'
8kT
21/21/2
2 1
Page 5.1-19
+
vgs2
gm2vgs2
rds1
rds2
eout2
_
Fig. 5.1-11
Page 5.1-20
VGG2
M2
M2
Noise
Free
MOSFETs
eout2
eeq2
vin
M1
VDD
eout2
en12
vin
Noise
Free
MOSFETs
M1
Fig. 5.1-12.
2
gm2 2 en2
(gm2rout)2
Page 5.1-21
*
vin
M2
eout2
en12
Noise
Free
MOSFETs
M1
Fig. 5.1-13.
The equivalent input-voltage-noise spectral density of the push-pull inverter can be found
as
eeq =
g e
g +g
m1 n1 2
m2
m1
g e
+g
m2 n2 2
m2
m1
+ g
If the two transconductances are balanced (gm1 = gm2), then the noise contribution of
each device is divided by two.
The total noise contribution can only be reduced by reducing the noise contribution of
each device.
(Basically, both M1 and M2 act like the load transistor and input transistor, so
there is no defined input transistor that can cause the noise of the load transistor to be
insignificant.)
CMOS Analog Circuit Design
Page 5.1-22
AC Voltage
Gain
AC Output
Resistance
Bandwidth (CGB=0)
Equivalent,
input-referred,meansquare noise voltage
p-channel
active load
inverter
n-channel
active load
inverter
-gm1
gm2
1
gm2
gm2
CBD1+CGS1+CGS2+CBD2
en12
gm2
+ en22g 2
-gm1
gm2+gmb2
1
gm2+gmb2
gm2+gmb2
CBD1+CGD1+CGS2+CBS2
en12
gm2
+ en22g 2
Current
source load
inverter
-gm1
gds1+gds2
1
gds1+gds2
gds1+gds2
CBD1+CGD1+CDG2+CBD2
en12
gm2
+ en22g 2
n-channel
depletion
load inverter
-gm1
~ gmb2
1
gmb2+gds1+gds2
gmb2+gds1+gds2
CBD1+CGD1+CGS2+CBD2
-(gm1+gm2)
gds1+gds2
1
gds1+gds2
gds1+gds2
CBD1+CGD1+CGS2+CBD2
Push-Pull
inverter
en1
m1
m1
m1
gm2
2
en2 g 2
m1
gm1en1 gm1en1
2
g
+ g
g
g
m2 m1
m2
m1
Page 5.2-1
Page 5.2-2
Page 5.2-3
VDD
vG1 M1
;y ;y
v
IBias ID
n+
S2 G2 D2
p+
n+
n+
VDD
M4
vGS1
iD1
iD2
M2 v
G2
+
vGS2
-
M3 ISS
VBulk
Fig. 5.2-2
n+
p-well
n-substrate
Fig. 5.2-3
1.) Bulks connected to the sources: No modulation of VT but large common mode
parasitic capacitance.
2.) Bulks connected to ground: Smaller common mode parasitic capacitors, but
modulation of VT.
If the technology is n-well CMOS, there is no choice. The bulks must be connected to
ground.
CMOS Analog Circuit Design
Page 5.2-4
4ISS
4ISS
and
iD/ISS
1.0
0.8
iD1
0.6
0.4
iD2
0.2
-2.0
0.0
-1.414
1.414
2.0
vID
(ISS/)0.5 Fig. 5.2-4
K'1 I SS W 1
diD1
1/2
gm = dvID(VID = 0) = (ISS/4)1/2 = 4L1
(half the gm of an inverting amplifier)
Page 5.2-5
2m
1m
M4
iD4 iOUT
M3
iD3
2m
1m
M1
vGS1
vG1
-
iD1
2m
1m
2m
1m
M2
ISS
M5
VBias
iD2
vGS2
VDD
2
vOUT
vG2
-
Fig. 5.2-5
Note that output signal to ground is equivalent to the differential output signal due to the
current mirror.
The short-circuit, transconductance is given as
K'1 I SS W 1
diOUT
1/2
gm = dvID (VID = 0) = (ISS)1/2 = L1
Page 5.2-6
Voltage Transfer Function of the Differential Amplifer with a Current Mirror Load
VDD = 5V
2m
1m
iD1
2m
1m
M1
vGS1 -
vG1
2m
1m
M2
ISS
M5
iD2
+
- vGS2 vOUT
vG2
-
M4 active
M4 saturated
M4 iD4 iOUT
M3
vOUT (Volts)
iD3
2m
1m
2m
1m
VIC = 2V
2
M2 saturated
M2 active
0
-1
VBias
-0.5
0
vID (Volts)
0.5
1
Fig. 330-01
Page 5.2-7
IDD
vSG1
+
M1
iD1
iD3
vG1
M3
-
M2
vSG2
+
iD2 iOUT
iD4
M4
+
vOUT
-
+
vG2
Fig. 5.2-7
Page 5.2-8
VDD
2m
1m
2m
1m
M4
iD4 iOUT
M3
iD3
2m
1m
2m
1m
iD1
M1
vGS1
vG1
-
iD2
M2
-
2m
1m
ISS
vGS2
VDD
2
vOUT
vG2
M5
VBias
Fig. 330-02
Page 5.2-9
iD3
iD2
iD1
M1
M2
vid
vout
M5
VBias
ISS
C3
D1=G3=D3=G4
M4
iD4 iout
M3
G1
+ vid
+
vg1
-
G2
+
vg2
C1
-
rds1
i3
1
gm3
rds3
S1=S2
rds5
gm2vgs2
gm1vgs1
D2=D4
rds2
i3
S3
G2
G1
+ vid +
+
vgs2
vgs1
gm1vgs1
-
S4
D1=G3=D3=G4
i3
1
rds1 rds3 gm3
rds4 C2
+
vout
D2=D4
C3
C1 gm2vgs2
S1=S2=S3=S4
i3
rds2
rds4 C2
iout'
+
vout
Fig. 330-03
Differential Transconductance:
Assume that the output of the differential amplifier is an ac short.
gm1gm3rp1
iout = 1 + g r vgs1 gm2vgs2 gm1vgs1 gm2vgs2 = gmdvid
m3 p1
where gm1 = gm2 = gmd, rp1 = rds1rds3 and i'out designates the output current into a short
circuit.
It can be shown that the current mirror causes this requirement to be invalid because the drain loads are not matched. However, we will continue to
use the assumption regardless.
CMOS Analog Circuit Design
P.E. Allen - 2004
Page 5.2-10
ISS
Note that the small-signal gain is inversely
vout
proportional to the square root of the bias
vin
Stong Inversion
current!
Weak
InversExample:
ion
If W1/L1 = 2m/1m and ISS = 50A
log(IBias)
1A
(10A), then
Fig. 330-04
Av(n-channel) = 46.6V/V (104.23V/V)
Av(p-channel) = 31.4V/V (70.27V/V)
1
1
rout = gds2 + gds4 = 25A0.09V-1 = 0.444M (2.22M)
CMOS Analog Circuit Design
Page 5.2-11
Common Mode Analysis for the Current Mirror Load Differential Amplifier
The current mirror load differential amplifier is not a good example for common mode
analysis because the current mirror rejects the common mode signal.
VDD
-M3-M4
1
M
M3
M4
M2
+
v 0V
M2 out
M1
vic
+
VBias
-
M5
Fig. 5.2-8A
mode Output = output due to - output due to
due to vic
M1-M3-M4 path
M2 path
Therefore:
The common mode output voltage should ideally be zero.
Any voltage that exists at the output is due to mismatches in the gain between the two
different paths.
Page 5.2-12
M3
vo1
M1
vid
2
VDD
M4
vo2
M2
vo1
v1
M3
M1
vid
2
VDD
M4
M2
vo1
vo2
v2
ISS
2
ISS
M5
M4
vo2
M2
M1
1
M5x 2
ISS
2
vic
VBias
Differential-mode circuit
M3
vic
VBias
General circuit
Common-mode circuit
Fig. 330-05
Differential-Mode Analysis:
gm1
vo2
gm2
vo1
and
+
vid
2gm3
vid
2gm4
Note that these voltage gains are half of the active load inverter voltage gain.
CMOS Analog Circuit Design
Page 5.2-13
+ vgs1 2rds5
gm1vgs1
rds1 rds3
+
1
gm3
vo1
-
Fig. 330-06
1 + 2gm1rds5
1 + 2gm1rds5
vic
2gm3
Common-Mode Rejection Ratio (CMRR):
|vo1/vid| gm1/2gm3
CMRR = |v /v | = g /2g = gm1rds5
o1 ic
ds5 m3
How could you easily increase the CMRR of this differential amplifier?
Page 5.2-14
VDD
Cgs3+Cgs4
M4
M3
Cgd4
Cgd1
vid
Cbd4
Cgd2 +
vout
CL
-
Cbd1
Cbd2
M2
M1
M5
G2
G1
D1=G3=D3=G4
+ vid +
+
i3
C3
vgs2
vgs1
1
gm1vgs1
gm3 C1 gm2vgs2
S1=S2=S3=S4
VBias
iout'
D2=D4
i3
rds2
rds4 C2
+
vout
-
Fig. 330-07
Ignore the zeros that occur due to Cgd1, Cgd2 and Cgd4.
C1 = Cgd1 + Cbd1 + Cbd3 + Cgs3 + Cgs4,C2 = Cbd2 + Cbd4 + Cgd2 + CL and C3 = Cgd4
If C3 0, then we can write
2
gm3
ggs2 + gds4
gm1
2
If we further assume that gm3/C1 >> (gds2+gds4)/C2 = 2
then the frequency response of the differential amplifier reduces to
Vout(s) gm1 2
(A more detailed analysis will be made in Chapter 6)
Vid(s) gds2 + gds4 s + 2
CMOS Analog Circuit Design
Page 5.2-15
VDD
R1
gm1vin
vin
M2
vo1 gm2vo1
M1
vout
R2
Fig. 5.2-10C
vout = (gm1R1gm2R2)vin
P.E. Allen - 2004
Page 5.2-16
M3
gm1vid gm1vid
2
2
M1 gm1vid gm2vid
2
2
+
vid
2 M5
rout
+
M2
vid
vout
+ 2
vid
VBias
Fig. 5.2-11
Page 5.2-17
gm2(eff)vin
M2
gm2vgs2 iout
+ vgs2 -
M2
rds1
vin
vin
M1 vin
VBias
rds1
Small-signal model
Fig. 5.2-11A
vin
vgs2 = vg2 - vs2 = vin - (gm2rds1)vgs2 vgs2 = 1+gm2rds1
gm2vin
Thus, iout = 1+gm2rds1 = gm2(eff) vin
CMOS Analog Circuit Design
Page 5.2-18
VDD
iD3
M1
vGS1
vG1
-
VBias
M2
-
+
CL
vSG1
+
M1
+
iD1
vOUT
vG1
vGS2 +
vG2
ISS
IDD
iD2
iD1
+
M5
M4
iD4 iOUT
M3
M2
VBias
iD2 iOUT
iD3
iD4
M4 CL
M3
M5
vSG2
+
+
vG2
vOUT
Fig. 5.2-11B
Note that slew rate can only occur when the differential input signal is large enough to
cause ISS (IDD) to flow through only one of the differential input transistors.
ISS IDD
SR = C = C
If CL = 5pF and ISS = 10A, the slew rate is SR = 2V/s.
L
L
(For the BJT differential amplifier slewing occurs at 100mV whereas for the MOSFET
differential amplifier it can be 2V or more.)
CMOS Analog Circuit Design
Page 5.2-19
VDD
M5
M5
M5
VBias
VBias
en12
M1
M2
en22
eeq2
M1
M2
ito2
M3
en32
en42
M4
vOUT
Vout
M3
M4
Fig. 5.2-11C
2BP
16kT
1/2
eeq(1/f)= fW L
1 + K B L eeq(th)= 3[2K' (W/L) I ]1/2 1+ L3W 1K'1
1 1
P P 3
1
1 1
CMOS Analog Circuit Design
Page 5.2-20
M3
X1
M7
v3
v1
IBias
M4
X1
X1
I3
I4
I1
I2
M1 X1
M5
M6
X1
v4
v2
X1 M2
I5
X2
Fig. 5.2-12
Current
0
0
I3
I1
I3
VDD
VDS1<VDS(sat)
(a.) I1>I3.
vDS1 0
I1
VSD3<VSD(sat)
(b.) I3>I1.
VDD
vDS1
Fig. 5.2-13
Page 5.2-21
MC1
v3
MC4
IC4
IC3
M4
I3
I4
MC2A
v1
VCM
MC2B
MC5
M1
M2
v4
Selfresistances
of M1-M4
v2
M5
MB
VSS
Fig. 5.2-14
Operation:
Common mode output voltages are sensed at the gates of MC2A and MC2B and
compared to VCM.
The current in MC3 provides the negative feedback to drive the common mode output
voltage to the desired level.
With large values of output voltage, this common mode feedback scheme has flaws.
CMOS Analog Circuit Design
Page 5.2-22
MC4
IC4
IC3
MC1
MC2
v3
MC5
I3
I4
RCM1
RCM2
v1
VCM
M4
M1
M2
v4
Selfresistances
of M1-M4
v2
M5
MB
VSS
Fig. 5.2-145
Note that RCM1 and RCM2 must not load the output of the differential amplifier.
Page 5.2-23
vout
CL
ALA20
Page 5.2-24
Max. ICMR
VSG4
-
M3
VDD
M4
vout
gm1Rout
+
vin
M2
M1
Min. ICMR
+
VBias
-
CL
I5
I5 = SRCL,
-3dB, Pdiss
M5
VSS
ALA20
Procedure:
1.) Pick ISS to satisfy the slew rate knowing CL or
the power dissipation
2.) Check to see if Rout will satisfy the frequency
response, if not change ISS or modify circuit
3.) Design W3/L3 (W4/L4) to satisfy the upper ICMR
4.) Design W1/L1 (W2/L2) to satisfy the gain
5.) Design W5/L5 to satisfy the lower ICMR
6.) Iterate where necessary
Page 5.2-25
Example 5.2-2 - Design of a MOS Differential Amp. with a Current Mirror Load
Design the currents and W/L values of the current mirror load MOS differential amplifier
to satisfy the following specifications: VDD = -VSS = 2.5V, SR 10V/s (CL=5pF), f-3dB
100kHz (CL=5pF), a small signal gain of 100V/V, -1.5VICMR2V and Pdiss 1mW.
Use the parameters of KN=110A/V2, KP=50A/V2, VTN=0.7V, VTP=-0.7V,
N=0.04V-1 and P=0.05V-1.
Solution
1.) To meet the slew rate, ISS 50A. For maximum Pdiss, ISS 200A.
2
2.) f-3dB of 100kHz implies that Rout 318k. Therefore Rout = (N+P)ISS 318k
ISS 70A Thus, pick ISS = 100A
3.) VIC(max) = VDD - VSG3 + VTN1 2V = 2.5 - VSG3 + 0.7
250A
VSG3 = 1.2V =
50A/V2(W3/L3) + 0.7
W3 W4
2
L3 = L4 = (0.5)2 = 8
gm1
W1 W 2
2110A/V2(W1/L1)
W1
= 23.31
4.) 100=gm1Rout=gds2+gds4 =
L1
L1 = L2 =18.4
(0.04+0.05) 50A
Page 5.2-26
250A
110A/V2(18.4) + 0.7
W5
2ISS
VDS5(sat) = 0.3 - 0.222 = 0.0777 L5 =
KNVDS5(sat)2 = 17.35
We probably should increase W1/L1 to reduce VGS1 and allow for a variation in VTN. If
we choose W1/L1 = 40, then VDS5(sat) = 0.149V and W5/L5 = 9. (Larger than specified
gain should be okay.)
Page 5.3-1
VGG2
RS
vS
+
vIN
-
Cgd1
M2
Rs2
+
v1
M1 -
vOUT
Fig. 5.3-1
The Miller effect causes Cgd1 to be increased by the value of 1 + (v1/vin) and appear in
parallel with the gate-source of M1 causing a dominant pole to occur.
The cascode amplifier eliminates this dominant pole by keeping the value of v1/vin
small by making the value of R2 to be approximately 2/gm2.
Page 5.3-2
vIN=4.5V
0.5
0.3
vIN=2.5V
ID (mA)
0.4
vIN=4.0V
vIN=3.5V
vIN=3.0V
K
0.2
5V
M3 W3 2m
=
L3 1m
ID
2.3V
M2
G F
E
JIH
M3
0.1
0.0
0
vOUT
M1
vIN=1.5V
+
vIN
A,B
vIN=2.0V 3.4V
vIN=1.0V
+
W2 2m
=
L2 1m
vOUT
W1 2m
=
L1 1m
-
5 A B C
D
4
vOUT
M3 active
M3 saturated
M2 saturated
M2 active
2
F
1
Fig. 5.3-2
0
0
M1 saturated
G H
M1
active
2v
3
IN
J K
Page 5.3-3
vDS1
(vOUT - vDS1)2
iD2 = 2 (VGG2 - vDS1 - VT2)(vOUT - vDS1)
2
and
3
iD3 = (VDD VGG3 |VT3|)2
2
where we have also assumed that both vDS1 and vOUT are small, and vIN = VDD.
Solving for vOUT by realizing that iD1 = iD2 = iD3 and 1 = 2 we get,
3
1
1
vOUT(min) = 22 (VDD VGG3 |VT3|)2 VGG2 VT2 + VDD VT1
Page 5.3-4
Example 5.3-1 - Calculation of the Min. Output Voltage for the Cascode Amplifier
(a.) Assume the values and parameters used for the cascode configuration plotted in the
previous slide on the voltage transfer function and calculate the value of vOUT(min).
(b.) Find the value of vOUT(max) and vOUT(min) where all transistors are in saturation.
Solution
(a.) Using the previous result gives,
vOUT(min) = 0.50 volts.
We note that simulation gives a value of about 0.75 volts. If we include the influence of
the channel modulation on M3 in the previous derivation, the calculated value is 0.62
volts which is closer. The difference is attributable to the assumption that both vDS1 and
vOUT are small.
(b.) The largest output voltage for which all transistors of the cascode amplifier are in
saturation is given as
vOUT(max) = VDD - VSD3(sat)
and the corresponding minimum output voltage is
vOUT(min) = VDS1(sat) + VDS2(sat) .
For the cascode amplifier of Fig. 5.3-2, these limits are 3.0V and 2.7V.
Consequently, the range over which all transistors are saturated is quite small for a 5V
power supply.
CMOS Analog Circuit Design
Page 5.3-5
G1
D2=D3
rds2
+
+
+
vin =
v1
r
v
ds3
out
vgs1 gm1vgs1
rds1
S1=G2=G3
Small-signal model of cascode amplifier neglecting the bulk effect on M2.
C1
rds2
D1=S2
D2=D3
G1
+
+
+
vin
1
v1
g
v
gm1vin
C
r
C
v
m2
1
2
ds3
3
out
rds1 gm2
Simplified equivalent model of the above circuit.
Fig. 5.3-3
gm1
gds3 =
2K'1W 1
L1ID23
Page 5.3-6
=
2
Page 5.3-7
(gm1 sC1)(gds2 + gm2)
Vout(s)
1
Vin(s) = 1 + as + bs2 gds1gds2 + gds3(gm2 + gds1 + gds2)
where
C3(gds1 + gds2 + gm2) + C2(gds2 + gds3) + C1(gds2 + gds3)
a=
gds1gds2 + gds3(gm2 + gds1 + gds2)
and
C3(C1 + C2)
b = gds1gds2 + gds3(gm2 + gds1 + gds2)
CMOS Analog Circuit Design
D2=D3
rds3
C3
+
vout
-
Fig. 5.3-4A
Page 5.3-8
1
s
s
1
s2
P(s) = 1 + as + bs2 = 1 p1 1 p2
= 1 s p1 + p2 + p1p2
p2 =
C1 + C2
C3(C1 + C2)
Assuming C1, C2, and C3 are the same order of magnitude, and gm2 is greater than gds3,
then |p1| is smaller than |p2|. Therefore the approximation of |p2| >> |p1| is valid.
Note that there is a right-half plane zero at z1 = gm1/C1.
CMOS Analog Circuit Design
Page 5.3-9
Vin(s) G3 1+[R1(C1+C2)+R3(C2+C3)+gm1R1R3C2]s+(C1C2+C1C3+C2C3)R1R3s2
Assuming that the poles are split allows the use of the previous technique to get,
gm1C2
1
1
p1 = R1(C1+C2)+R3(C2+C3)+gm1R1R3C2 gm1R1R3C2 andp2 C1C2+C1C3+C2C3
The Miller effect has caused the input pole, 1/R1C1, to be decreased by a value of gm1R3.
CMOS Analog Circuit Design
Page 5.3-10
R1(C1+C2)+ gm(C2+C3)+gm1R1gmC2
1
1
2
R1(C1+3C2)
R1(C1+C2)+ gm(C2+C3)+2R1C2
Page 5.3-11
M1
+
gmbs3v4 rds3
vin
gm1vin
-
v1
rds1
G2=G3=G4=S1=S4
D4=S3
+
v4
vout
rds4
-
Fig. 5.3-6
rout [gm2rds1rds2][gm3rds3rds4] =
ID
12
34
+
2K'2(W/L)2
2K'3(W/L)3
2K'1(W/L)1ID
Av = gm1rout gm1{[gm2rds1rds2][gm3rds3rds4]}
12
34
+
2K'2(W/L)2
2K'3(W/L)3
CMOS Analog Circuit Design
Page 5.3-12
Page 5.3-13
I=
VDD
K PW 3
2
2L3 (VDD - VGG3-|VTP|)
M3
VGG3
+
M2
VGG2
Fig. 5.3-7
vOUT
I = Pdiss = (SR)Cout
VDD
M1
-
g
1/L1)
|Av| = g m1 = 2KN(W
2
ds3
P I
Page 5.3-14
Page 5.4-1
ii
Ai
iS
Current
Amplifier
RS
Single-ended input.
RS >> Rin
iS
RL
RS
ii
+
-
io
Ai
Current
Amplifier
Differential input.
RL
Fig. 5.4-1
Page 5.4-2
io = Ai(i1-i2) = Ai R1 - io
Ao
If Ai(s) = s
, then
A + 1
vout R2 1 R2
R2 Ao
Ao
1 R1 s
s
vin R1
R1 1+Ao
A +(1+Ao)
1+ Ai(s)
A(1+Ao) +1
-3dB = A(1+Ao)
Ai
io
vout
Fig. 5.4-2
Page 5.4-3
R
Voltage Amplifier, R2 > K
R2 Ao
1
dB
R2
R1 1+Ao
Voltage Amplifier, R = K >1
1
Ao
dB
K
1+Ao
Current Amplifier
Ao dB
(1+Ao)A
0dB
GBi
GB1 GB2
log10()
Fig. 7.2-10
Page 5.4-4
VDD
I1
iin
M1
I2
R
iout
M2
iin
+
vin
gm1vin
-
iout
C2
rds1
and
gm2vin
rds2
RL
0
C3
Fig. 5.4-3
Current Amplifier
1
1
Rin = gm1 Rout = 1Io
C1
W2/L2
Ai = W 1/L 1 .
Frequency response:
-(gm1+gds1)
-(gm1+gds1)
-gm1
p1 = C1+C2 = Cbd1+Cgs1+Cgs2+Cgd2 Cbd1+Cgs1+Cgs2+Cgd2
Note that the bandwidth can be almost doubled by including the resistor, R.
(R removes Cgs1 from p1)
Page 5.4-5
Page 5.4-6
I1
iin
VDD
I2
iout
+
R
M3
M4
vin
vout
M1
M2
Current Amplifier
1
Rin R + g ,
m1
Rout rds2gm4rds4,
and
Fig. 5.4-4
W2/L2
Ai = W /L
1
Page 5.4-7
Page 5.4-8
VDD
I2
iout
iin
I3
+
vin
- gm1vgs1
M2
M1
i=0
VGG3
rds1
vgs3
+
gm3vgs3
+
vgs1
-
rds3
Fig. 5.4-5
Current Amplifier
Feedback concept:
Input resistance without feedback rds1.
gm1 gm3
Loop gain gds1gds3 assuming that the resistances of I1 and I3 are very large.
Rin(no fb.)
rds1
1
Rin = 1 + Loop gain g r g r
=
m1 ds1 m3 ds3 gm1gm3rds3
Small signal analysis:
iin = gm1vgs1 - gds1vgs3
and vgs3 = -vin vgs1 = vin - (gm3 vgs3rds3) = vin(1+gm3rds3)
1
iin = gm1(1+gm3rds3)vin + gds1vin gm1gm3rds3vin
Rin gm1gm3rds3
CMOS Analog Circuit Design
Page 5.4-9
iID
iO
i2 iIC
2
Fig. 5.4-6
i1+i2
iO = AIDiID AICiIC = AID(i1 - i2) AIC 2
Implementations:
VDD
VDD
I
2I
i1
i2
i2
M1 M2
M3 M4
M3
VDD
VDD
M4
iO
iO
M1
i1
i1-i2
VGG1
M2
i2
M6
M5
VGG2
Fig. 5.4-7
Page 5.4-10
Summary
Current amplifiers have a low input resistance, high output resistance, and a defined
output-input current relationship
Input resistances less than 1/gm require feedback
However, all feedback loops have internal poles that cause the benefits of negative
feedback to vanish at high frequencies.
In addition, these feedback loops can have a slow time constant from a pole-zero pair.
Voltage amplifiers using a current amplifier have high values of gain-bandwidth
Current amplifiers are useful at low power supplies and for switched current
applications
Page 5.5-1
VDD
f1(vIN)
i1
f2(vIN)
Buffer
Class A
i2
vIN
iOUT
RL
+
vOUT
-
VSS
Current
i1
t
i2=IQ
iOUT
Class AB
i1
Current
iOUT
t
i2
Class B
Current
i1
iOUT
t
i2
Fig. 5.5-005
Page 5.5-2
Class A Amplifiers
Current source load inverter:
VDD
VGG2
IQ
VDD+|VSS|
RL
M2
iOUT
vOUT
iD
RL dominates
as the load line
IQ
iD1
RL
vOUT
M1CL
A Class A circuit has current vIN
IQRL
IQRL
VDD
VSS
flow in the MOSFETs during
Fig. 5.5-1
VSS
the entire period of a
sinusoidal signal.
Characteristics of Class A amplifiers:
Unsymmetrical sinking and sourcing
Linear
Poor efficiency
vOUT(peak)2
vOUT(peak)2
vOUT(peak)
2RL
2RL
PRL
2
Efficiency = PSupply = (VDD-VSS)IQ =
(VDD-VSS) = V DD -V SS
(VDD -VSS) 2RL
Maximum efficiency occurs when vOUT(peak) = VDD = |VSS| which gives 25%.
CMOS Analog Circuit Design
Page 5.5-3
Minimum RL for
maximum swing
IQ
0
VSS
IQ R L
IQ R L
Larger RL
vDS1
VDD Fig. 040-03
Page 5.5-4
Page 5.5-5
gm1vin
rds1
rds2
RL
C2
+
vout
Fig. 5.5-2
Page 5.5-6
L2 = K (V -V
2m
P DD GG2-|VTP|)2 50(2.3)2
The small-signal performance is Av = -8.21 V/V (includes RL = 20k) and rout = 50k
The roots are, zero = gm1/Cgd1 .59GHz and pole = 1/[(RL||rout)CL)] -11.14kHz
CMOS Analog Circuit Design
Page 5.5-7
Page 5.5-8
VDD
Triode
VDD-VGS1
vIN M1
IQ
VSS
M3
M2
VSS
VSS
iOUT
|VSS|+VON2+VGS1
VGS1
vOUT
VDD-VON1+VGS1
RL
Fig. 040-01
vIN
IQRL<|VSS|+VON2
Triode
or
or
|VSS|+VON2
|VSS|
Fig. 040-02
Page 5.5-9
Page 5.5-10
M2
RL
Page 5.5-11
2
(VDD-VSS) = V DD -V SS
PSupply = (VDD-VSS)IQ =
(VDD -VSS) 2RL
Maximum efficiency occurs when vOUT(peak) = VDD = |VSS| which gives 25%.
Comments:
Maximum efficiency occurs for the minimum value of RL which gives maximum swing.
Other values of RL result in less efficiency (and smaller signal swings before clipping)
We have ignored the fact that the dynamic Q point cannot travel along the full length of
the load line because of minimum and maximum voltage limits.
CMOS Analog Circuit Design
Page 5.5-12
C1
gm1vgs1
+
+
vin
-
vgs1
gmbs1vbs1
rds1
RL
rds1
rds2
C2
+
vout
-
C1
gm1vin
rds2
gm1vout
gmbs1vout
RL
C2
+
vout
-
Fig. 040-04
gm1
gm1RL
Vout
gm1
=
Page 5.5-13
Page 5.5-14
VDD
VDD
M1
vIN
M6
VGG
VSS
iOUT
VBias
VDD
M5 M1
VSS
iOUT
VSS
vOUT
VBias
vOUT
VDD RL
VDD
RL
M4 M2
Efficiency:
VDD
M2
Depends on how the
vIN
M3
transistors are biased.
VSS
Fig. 060-01
VSS
VSS
Class B - one transistor
has current flow for only 180 of the sinusoid (half period)
vOUT(peak)2
PRL
2RL
vOUT(peak)
Efficiency = P
=
=
2 VDD -VSS
1 2vOUT(peak)
VDD
(VDD -VSS)2
RL
Page 5.5-15
vG1
1V
1mA
iD1
0mA
vout
vG1
vG2
0V
0mA
vout
-1V
iD2
-2V
-1mA
-2
0
2
1
Vin(V)
Class B, push-pull, source follower
-1
1mA
iD1
1V
0V
-1V
2V
vG2
iD2
-2V
-2
-1mA
0
2
1
Vin(V)
Class AB, push-pull, source follower
-1
Fig. 060-02
Comments:
Note that vOUT cannot reach the extreme values of VDD and VSS
Page 5.5-16
+
+
vin
+
vin
-
C1
gm1vgs1
vgs1
gmbs1vbs1
rds1
gm2vgs2 gmbs2vbs2
rds2
RL
C2
+
vout
-
C1
gm1vin
1
RL
g
gm1vout gmbs1vout rds1 gm2vin gm2vout m2gmbs2vout rds2
C2
+
vout
Fig. 060-03
vout
gm1 + gm2
vin = gds1+gds2+gm1+gmbs1+gm2+gmbs2+GL
1
Rout = gds1+gds2+gm1+gmbs1+gm2+gmbs2 (does not include RL)
If VDD = -VSS = 2.5V, Vout = 0V, ID1 = ID2 = 500A, and W/L = 20m/2m, Av = 0.787
(RL=) and Rout = 448.
A zero and pole are located at
-(gds1+gds2+gm1+gmbs1+gm2+gmbs2+GL)
-(gm1+gm2)
p
=
.
z=
C1
C1+C2
These roots will be high-frequency because the associated resistances are small.
CMOS Analog Circuit Design
Page 5.5-17
iOUT
vIN
vOUT
VTR1
M1CL
VSS
RL
Fig. 060-04
Comments:
The batteries VTR1 and VTR2 are necessary to control the bias current in M1 and M2.
The efficiency is the same as the push-pull, source follower.
Page 5.5-18
M5
VGG3
M1 M3
iOUT
vIN
vOUT
VGG4
M2 M4
M7
RL
CL
M8
VSS
Fig. 060-05
VGG3 and VGG4 can be used to bias this amplifier in class AB or class B operation.
Note, that the bias current in M6 and M8 is not dependent upon VDD or VSS (assuming
VGG3 and VGG4 are not dependent on VDD and VSS).
Page 5.5-19
M7
Ib
vin+
M1
M3 M4
M8
M9
vin-
M2
M6
Ib
I=2Ib
VSS
M10
Fig. 060-055
In steady-state, the current through M5 and M6 is 2Ib. If W4/L4 = W9/L9 and W3/L3 =
W8/L8, then the currents in M1 and M2 can be determined by the following relationship:
W 1 /L 1
W 2 /L 2
I1 = I2 = Ib W /L = Ib W /L
7
7
10 10
If vin+ goes low, M5 pulls the gates of M1 and M2 high. M4 shuts off causing all of the
current flowing through M5 (2Ib) to flow through M3 shutting off M1. The gate of M2 is
high allowing the buffer to strongly sink current. If vin- goes high, M6 pulls the gates of
M1 and M2 low. As before, this shuts off M2 and turns on M1 allowing strong sourcing.
CMOS Analog Circuit Design
Page 5.5-20
2V
iD1
1V
vG1
iD2
iD1
0V
2mA
2V
1mA
1V
0mA
iD2
-1V
vOUT
-2V
-2V
vG1
iD1
-1V
-2mA
-2V
1mA
iD1
0V
-1mA
2mA
vG2
0mA
iD2
vOUT
iD2
-1mA
-2mA
-2V
0V
1V
2V
vIN
Class AB, push-pull, inverting amplifier. Fig.060-06
-1V
0V
1V
2V
vIN
Class B, push-pull, inverting amplifier.
-1V
Comments:
Note that there is significant distortion at vIN =0V for the Class B inverter
Note that vOUT cannot reach the extreme values of VDD and VSS
IOUT+(max) and IOUT-(max) is always less than VDD/RL or VSS/RL
For vOUT = 0V, there is quiescent current flowing in M1 and M2 for Class AB
CMOS Analog Circuit Design
Page 5.5-21
VDD
M3
VDD
Q1
iB
M2
vout
vout
iB
Q1
M2
M3
CL
VSS
p-well CMOS
VSS
CL
VSS
n-well CMOS
Fig. 5.5-8A
Comments:
Can use either substrate or lateral BJTs.
Small-signal output resistance is 1/gm which can easily be less than 100.
Unfortunately, only PNP or NPN BJTs are available but not both on a standard CMOS
technology.
In order for the BJT to sink (or source) large currents, the base current, iB, must be
large. Providing large currents as the voltage gets to extreme values is difficult for
MOSFET circuits to accomplish.
If one considers the MOSFET driver, the emitter can only pull to within vBE+VON of the
power supply rails. This value can be 1V or more.
We will consider the BJT as an output stage in more detail in Sec. 7.1.
CMOS Analog Circuit Design
Page 5.5-22
Error
Amplifier
vIN
Error
Amplifier
M2
+
iOUT
vOUT
+
CL
M1
RL
Fig. 060-07
VSS
rds1||rds2
Rout = 1+Loop Gain
Comments:
Can achieve output resistances as low as 10.
If the error amplifiers are not balanced, it is difficult to control the quiescent current in
M1 and M2
Great linearity because of the strong feedback
Can be efficient if operated in class B or class AB
CMOS Analog Circuit Design
Page 5.5-23
R1
R2
iOUT
vIN
vOUT
CL
M1
VSS
RL
Fig. 060-08
R1 gm1+gm2
Loop gain R1+R2gds1+gds2+GL
rds1||rds2
R1
gm1+gm2
1+R1+R2gds1+gds2+GL
1+0.5 25+20
Rout =
Page 5.5-24
Q2
G +
G
VGS
M1
VGS
VSG
ID
ID1
G
- ID
S
S
+
VSG
M1
ID1
Q2
ID
D
ID
D
Fig. 5.5-11
PMOS Equivalent:
NMOS Equivalent:
K P W 1
KPW1
2
ID=(1+2)ID1=(1+2) 2L1 (VGS-VT) ID=(1+2)ID1=(1+2) 2L1 (VSG-VT)2
Page 5.5-25
Page 5.6-1
+
xi
x = either voltage or current
xs
A
xo
- xf
xo
A = xi = high-gain amplifier
F = feedback network
F
Fig. 5.6-1
Closed-loop gain:
xo
A
Af = xs = 1+AF
If AF >> 1, then,
xo 1
Af = xs F
Therefore, to precisely define the closed-loop gain, Af, we only need to make A large and
Af becomes dependent on F which can be determined by passive elements.
Page 5.6-2
Types of Amplifiers
The gain of an amplifier is given as
xo
A= x
i
Therefore, since x can be voltage or current, there are four types of amplifiers as
summarized below.
Types of
VoltageVoltageCurrentCurrentAmplifers
controlled,
controlled,
controlled,
controlled,
current-source voltage-source current-source voltage-source
xi variable*
Voltage
Voltage
Current
Current
xo variable
Current
Voltage
Current
Voltage
Desired Ri
Large
Large
Small
Small
Desired Ro
Large
Small
Large
Small
* The xi , xs, and xf must all be the same type of variable, voltage or current.
Page 5.6-3
io
+
vi R i
-
vs
Ro
Gmvi
+
vi
-
RL
Differential
Amplifier
VCCS
Second
Stage
io
VCCS
Fig. 5.6-2
io
GmRoRi
=
G
=
M
vs
(Ri + RS)(Ro + RL)
This amplifier is sometimes called an operational transconductance amplifier (OTA).
Page 5.6-4
+
vi Ri
-
Ro
Avvi
+
vo
-
RL
+
vi
-
Differential
Amplifier
VCVS
Second
Stage
VCVS
Output
Stage
vo
Fig. 5.6
vo
AvRiRL
=
A
=
V
vs
(RS + Ri)(Ro + RL)
This amplifier is normally called an operational amplifier.
Page 5.6-5
RS
io
Ri
Gmvi
Ro
i1
ii
RL
i2
Current
Differential
Amplifier
CCCS
Second
Stage
io
CCCS
Fig. 5.6-4
io
AiRSRo
=
A
=
I
is
(RS + Ri)(Ro + RL)
Page 5.6-6
RS
+
vi Ri
-
i1
Ro
Rmvi
+
vo
RL
CCVS
ii
i2
Current
Differential
Amplifier
Second
Stage
Output
Stage
vo
CCVS
Fig. 5.6-5
vo
RmRSRL
=
R
=
M
is
(Ri + RS)(Ro + RL)
Page 5.7-1
Page 6.0-1
Blocks or circuits
(Combination of primitives, independent)
Sub-blocks or subcircuits
(A primitive, not independent)
Fig. 6.0-1
Page 6.1-1
v1
v2
+ Differential
Transconductance
Stage
High
Gain
Stage
Bias
Circuitry
vOUT
Output vOUT'
Buffer
Fig. 110-01
Page 6.1-2
Ideal Op Amp
Symbol:
i1
+
v1
VDD
+
i2 vi
- -
+
v2
- -
+
VSS vOUT = Av(v1-v2)
Fig. 110-02
Null port:
If the differential gain of the op amp is large enough then input terminal pair becomes a
null port.
A null port is a pair of terminals where the voltage is zero and the current is zero.
I.e.,
v1 - v2 = vi = 0
and
i1 = 0 and i2 = 0
Therefore, ideal op amps can be analyzed by assuming the differential input voltage is
zero and that no current flows into or out of the differential inputs.
CMOS Analog Circuit Design
Page 6.1-3
- R2
+
vinn
vinp
+
v2
v1
- -
+
vout
Fig. 110-03
R1+R2
vout = R1 vinp
R2
vout = -R vinn
Page 6.1-4
i2 R2
+ ii
vi
-
+
vin
-
+
vout
Virtual Ground
Fig. 110-04
Solution
If Av , then vi 0 because of the negative feedback path through R2.
(The op amp with fb. makes its input terminal voltages equal.)
vi = 0 and ii = 0
Note that the null port becomes the familiar virtual ground if one of the op amp input
terminals is on ground. If this is the case, then we can write that
vin
vout
and
i2 = R2
i1 = R1
vout
R2
Since, ii = 0, then i1 + i2 = 0 giving the desired result as vin = - R1 .
CMOS Analog Circuit Design
Page 6.1-5
CMRR
Ricm
IB2
en2
v2
*
VOS
in2
v1
Cid
Rid
Rout
vout
Ideal Op Amp
Ricm
IB1
Fig. 110-05
where
Rid = differential input resistance
Cid = differential input capacitance
Ricm = common mode input resistance
VOS = input-offset voltage
IB1 and IB2 = differential input-bias currents
IOS = input-offset current (IOS = IB1-IB2)
CMRR = common-mode rejection ratio
e2n = voltage-noise spectral density (mean-square volts/Hertz)
i2n = current-noise spectral density (mean-square amps/Hertz)
CMOS Analog Circuit Design
Page 6.1-6
Differential-frequency response:
Av0
Av0 p1p2p3
=
Av(s) = s
s
s
1
1
1
p
p
p
1
2
3
where p1, p2, p3, are the poles of the differential-frequency response (ignoring zeros).
|Av(j)| dB
Asymptotic
Magnitude
20log10(Av0)
Actual
Magnitude
0dB
-6dB/oct.
GB
2 3
1
-12dB/oct.
-18dB/oct.
Fig. 110-06
Page 6.1-7
Final Value +
vIN
vOUT
Final Value
Final Value -
Lower Tolerance
Settling Time
0
Ts
Fig. 110-07
Page 6.1-8
Conversion
Voltage
to Current
Classic Differential
Amplifier
Current
to Voltage
Differential-to-single ended
Load (Current Mirror)
Voltage
to Current
Transconductance
Grounded Gate
Modified Differential
Amplifier
Source/Sink
Current Loads
Current
Stage
Transconductance
Grounded Source
Second
Voltage
Stage
Class B
(Push-Pull)
Class A (Source
or Sink Load)
Current
to Voltage
MOS Diode
Load
First
Voltage
Stage
Table 110-01
Page 6.1-9
vin
vin
+
VBias
vout
M1 M2
vout
M7
M5
VSS
VI
I V
VI
IV
Fig. 6.1-8
Page 6.1-10
M3
M10 M11
+
vin
-
M1
M2
M8
M6
M7
VBias
M4
M5
M9
vout
vin
vout
VBias
V I
I I
VSS
IV
Fig. 6.1-9
Page 6.1-11
Page 6.1-12
Page 6.1-13
Requirement
See Tables 3.1-1 and 3.1-2
2.5 V 10%
100 A
0 to 70C
Value
70 dB
5 MHz
1 sec
5 V/sec
1.5 V
60 dB
60 dB
1.5 V
N/A, capacitive load only
10 mV
100nV/ Hz at 1KHz
10,000 min. channel length2
P.E. Allen - 2004
Page 6.1-14
Page 6.2-1
Page 6.2-2
Vout(s)
Fig. 120-01
Page 6.2-3
|A(j)F(j)|
-40dB/decade
Arg[-A(j)F(j)]
0dB
180
135
90
45
0dB
Fig. Fig. 120-02
Frequency (rads/sec.)
A measure of stability is given by the phase when |A(j)F(j)| = 1. This phase is called
phase margin.
Phase margin = M = Arg[-A(j0dB)F(j0dB)] = Arg[L(j0dB)]
0
Page 6.2-4
1.2
1.0
60
65
70
vout(t) 0.8
Av0
0.6
0.4
0.2
0
0
5
10
ot = nt (sec.)
15 Fig. 120-03
A good step response is one that quickly reaches its final value.
Therefore, we see that phase margin should be at least 45 and preferably 60 or larger.
(A rule of thumb for satisfactory stability is that there should be less than three rings.)
Note that good stability is not necessarily the quickest risetime.
CMOS Analog Circuit Design
Page 6.2-5
M3 M4
VCC
Q3
M6
Q4
Q6
vout
vin
+
M1 M2
+
VBias
-
vin
+
+
VBias
-
M7
M5
Q1
vout
Q2
Q7
Q5
VSS
VEE
Fig. 120-04
Small-Signal Model:
D1, D3 (C1, C3)
+
g v
gm1vin
v1 m2 in
R1
C1
2
2
gm4v1
R2
C2
R3
C3
+
vout
Fig. 120-05
Note that this model neglects the base-collector and gate-drain capacitances for purposes
of simplification.
CMOS Analog Circuit Design
Page 6.2-6
R2
C2
+
v2
- gm6v2
R3
C3
+
vout
gm1Vin
RI
CI
+
VI
- gmIIVI
RII
CII
+
Vout
Fig. 120-06
The locations for the two poles are given by the following equations
1
1
p1 = RICI and p2 = RIICII
where RI (RII) is the resistance to ground seen from the output of the first (second) stage
and CI (CII) is the capacitance to ground seen from the output of the first (second) stage.
CMOS Analog Circuit Design
Page 6.2-7
|A(j)|
-20dB/decade
GB
log10()
0dB
Phase Shift
-40dB/decade
-45/decade
Arg[-A(j)]
180
135
-45/decade
90
45
0
|p1'|
|p2'| 0dB
log10()
Fig. 120-07
If we assume that F(s) = 1 (this is the worst case for stability considerations), then the
above plot is the same as the loop gain.
Note that the phase margin is much less than 45.
Therefore, the op amp must be compensated before using it in a closed-loop
configuration.
CMOS Analog Circuit Design
Page 6.2-8
VCC
M4
Q3
M6
CM
Cc
vin
+
M1
Cc
M2
vin
+
CII
Q1
vout
Q2
CII
CI
+
VBias
-
M7
M5
Q6
vout
CI
+
VBias
-
Q4
CM
Q7
Q5
VSS
VEE
Fig. 120-08
Page 6.2-9
1
rds1||rds3 CM gm3
v2
gm2vin
2
v2
+
vin gm1vin
-
CI
Cc
rds2||rds4
rds6||rds7 CL
+
vout
-
Cc
gm6v2
rds6||rds7
CII
+
vout
Fig. 120-09
Same circuit holds for the BJT op amp with different component relationships.
CMOS Analog Circuit Design
Page 6.2-10
1
s s
1 s2
s
s2
In general, D(s) = 1-p 1-p = 1-s p + p +p p D(s) 1-p + p p , if |p2|>>|p1|
1
2
2
1 2
1
1 2
1
gmII
-1
-1
p1 = RI(CI+CII)+RII(CII+Cc)+gmIIR1RIICc gmIIR1RIICc , z = Cc
p2 =
-[RI(CI+CII)+RII(CII+Cc)+gmIIR1RIICc]
-gmIICc
-gmII
Page 6.2-11
p2'
p1'
p1
z1
Fig. 120-11
Page 6.2-12
|A(j)F(j)|
Uncompensated
-20dB/decade
Compensated
GB
log10()
0dB
Phase Shift
-40dB/decade
Arg[-A(j)F(j)|
Uncompensated
180
-45/decade
135
-45/decade
90
45
0
Compensated
|p1|
Phase
Margin
log10()
No phase margin
|p2'| |p2|
|p1'|
Fig. 120-12
GB = Avd(0)|p1| = (gmIgmIIRIRII)g
CMOS Analog Circuit Design
Page 6.2-13
VDD
Cc
1
|p1| R (g R C )
I m6 II c
RII
vout
RI
M6
vI
gm6RIICc
Fig. 120-13
VDD
Cc
gm6
|p2| CII
VDD
RII
RII
vout
M6
CII
vout
1
GBCc 0
M6
CII
Fig. 120-14
m6
-R
1
II sC
-gm6RII(1/sCc)
R
c
II
VDD
Cc
RII
vout
v''
M6
v'
Fig. 120-15
Page 6.2-14
i3
1
rds1 rds3 gm3
gm2Vin
2
C3
i3
+
Vo1
rds2
rds4
Fig. 120-16
The transfer function from the input to the output voltage of the first stage, Vo1(s), can be
written as
sC3 + 2gm3
gm3+gds1+gds3
-gm1
-gm1
Vo1(s)
Page 6.2-15
F=1
Avd(0) dB
-6dB/octave
Cc 0
GB
log10()
0dB
Phase Shift
Magnitude influence of C3
0
45
90
135
180
CMOS Analog Circuit Design
Cc = 0
Cc 0
Cc = 0
-45/decade
Cc 0
-45/decade
Phase margi
ignoring C3
Cc = 0
Phase margin due to C3
|p1|
-12dB/octave
|p3||z3||p2|
log10()
Fig. 120-17
P.E. Allen - 2004
Page 6.2-16
1
1
GB = Av(0)|p1| = (gmIgmIIRIRII)g R R C = C = (gm1gm2R1R2)g R R C = C
c
c
mII I II c
m2 1 2 c
The requirement for 45 phase margin is:
180 - Arg[AF] = 180 - tan-1|p1| - tan-1|p2| - tan-1 z = 45
Let = GB and assume that z 10GB, therefore we get,
GB
GB
GB
180 - tan-1|p1| - tan-1|p2| - tan-1 z = 45
GB
GB
135 tan-1(Av(0)) + tan-1|p2| + tan-1(0.1) = 90 + tan-1|p2| + 5.7
GB
GB
39.3 tan-1|p2| |p | = 0.818 |p2| 1.22GB
2
The requirement for 60 phase margin:
|p2| 2.2GB if z 10GB
If 60 phase margin is required, then the following relationships apply:
gm6 10gm1
gm6 2.2gm1
>
g
>
10g
and
Cc > 0.22C2
m6
m1
Cc
Cc
C2 > Cc
CMOS Analog Circuit Design
Page 6.2-17
j1
z1
Fig. 430-01
Page 6.2-18
Use of Buffer to Eliminate the Feedforward Path through the Miller Capacitor
Model:
Cc
+1
VI
Cc
The transfer
+
V
function is given
Vout
CI
v
in gmIvin
RI
OUT
Inverting
RII
High-Gain
gmIIVI
by the following
Stage
equation,
Vo(s)
(gmI)(gmII)(RI)(RII)
=
Vin(s) 1 + s[RICI + RIICII + RICc + gmIIRIRIICc] + s2[RIRIICII(CI + Cc)]
Using the technique as before to approximate p1 and p2 results in the following
1
1
p1 RICI + RIICII + RICc + gmIIRIRIICc gmIIRIRIICc
and
gmIICc
p2 CII(CI + Cc)
Comments:
Poles are approximately what they were before with the zero removed.
For 45 phase margin, |p2| must be greater than GB
For 60 phase margin, |p2| must be greater than 1.73GB
CII
+
Vout
-
Fig. 430-02
Page 6.2-19
Use of Buffer with Finite Output Resistance to Eliminate the RHP Zero
Assume that the unity-gain buffer has an output resistance of Ro.
Model:
Cc
Ro
Inverting
High-Gain
Stage
+1
VI
vOUT
+
Vin gmIvin
-
CI
RI
Cc
Ro
Vout
Ro
gmIIVI
RII
CII
+
Vout
Fig. 430-03
It can be shown that if the output resistance of the buffer amplifier, Ro, is not neglected
that another pole occurs at,
1
p4 Ro[CICc/(CI + Cc)]
and a LHP zero at
1
z2 RoCc
Closer examination shows that if a resistor, called a nulling resistor, is placed in series
with Cc that the RHP zero can be eliminated or moved to the LHP.
CMOS Analog Circuit Design
Page 6.2-20
Use of Nulling Resistor to Eliminate the RHP Zero (or turn it into a LHP zero)
Cc
Rz
VI
Inverting
High-Gain
Stage
vOUT
+
Vin gmIvin
-
CI
Cc
RI
Rz
gmIIVI
RII
CII
+
Vout
Fig. 430-04
Nodal equations:
VI
sCc
gmIVin + RI + sCIVI + 1 + sCcRz (VI Vout) = 0
Vo
sCc
Solution:
Vout(s) a{1 s[(Cc/gmII) RzCc]}
Vin(s) =
1 + bs + cs2 + ds3
where
a = gmIgmIIRIRII
b = (CII + Cc)RII + (CI + Cc)RI + gmIIRIRIICc + RzCc
c = [RIRII(CICII + CcCI + CcCII) + RzCc(RICI + RIICII)]
d = RIRIIRzCICIICc
W,J. Parrish, "An Ion Implanted CMOS Amplifier for High Performance Active Filters", Ph.D. Dissertation, 1976, Univ. of CA., Santa Barbara.
CMOS Analog Circuit Design
P.E. Allen - 2004
Chapter 6 Section 2 (5/2/04)
Page 6.2-21
Page 6.2-22
Rz
RII
Vout
V''
M6
V'
Fig. Fig. 430-05
gm6
1
-gm6RIIRz + sC
-R
g
R
+
II m6 z sCc - 1
RII
c
V
Vout =
1 V +
1 V =
1
RII + Rz + sCc
RII + Rz + sCc
RII + Rz + sCc
when V = V = V.
Setting the numerator equal to zero and assuming gm6 = gmII gives,
1
z1 = Cc(1/gmII Rz)
Page 6.2-23
A Design Procedure that Allows the RHP Zero to Cancel the Output Pole, p2
We desire that z1 = p2 in terms of the previous notation.
Therefore,
gmII
1
j
=
CII
Cc(1/gmII Rz)
-p
z
-p
-p
1
Fig. 430-06
1
4
2
The value of Rz can be found as
Cc + CII
Rz = Cc (1/gmII)
With p2 canceled, the remaining roots are p1 and p4(the pole due to Rz) . For unity-gain
stability, all that is required is that
gmI
Av(0)
|p4| > Av(0)|p1| = gmIIRIIRICc = C
c
and
(1/RzCI) > (gmI/Cc) = GB
Substituting Rz into the above inequality and assuming CII >> Cc results in
gmI
Cc >
gmII CICII
This procedure gives excellent stability for a fixed value of CII ( CL).
Unfortunately, as CL changes, p2 changes and the zero must be readjusted to cancel p2.
CMOS Analog Circuit Design
Page 6.2-24
M12
Cc
M11
M7
vOUT
Cgd6
+
Iin
R1
VBias
M8
Cc
rds8
+
V1
Vs8
- gm8Vs8 - gm6V1
R2
C2
+
Vout
-
M6
Cgd6
M10
M9
VSS
Fig. 6.2-15B
Iin
R1
V1
- gm8Vs8
Cc
+
gm8
Vs8
- gm6V1
R2
+
Vout
-
C2
gm8sCc
B.K. Ahuja, An Improved Frequency Compensation Technique for CMOS Operational Amplifiers, IEEE J. of Solid-State Circuits, Vol. SC-18,
No. 6 (Dec. 1983) pp. 629-633.
CMOS Analog Circuit Design
P.E. Allen - 2004
Chapter 6 Section 2 (5/2/04)
Page 6.2-25
sCc
1
+
gm8
Vout -gm6
Cc
CcC2
C
C
g
Iin G1G2
C
c
2
m6
c
Using the approximate method of solving for the roots of the denominator gives
-1
-6
p1 = Cc Cc C2 gm6Cc g r 2C
m6 ds c
gm8 + G2 + G2 + G1G2
and
gm6rds2Cc
gm8rds2G2 gm6 gm8rds
6
p2
=
C =
CcC2
6
3 |p2|
2
gm8G2
where all the various channel resistance have been assumed to equal rds and p2 is the
output pole for normal Miller compensation.
Result:
Dominant pole is approximately the same and the output pole is increased by gmrds.
CMOS Analog Circuit Design
Page 6.2-26
-gm6gm8rds -gm8
Cc
3C2
-1
gm6rdsCc
gm6
Cgd6
Fig. 6.2-16A
Page 6.2-27
VDD
Cc
gm8rds8
3
rds7
rds7
vout
vout
1
GBCc 0
M8
M6
CII
M6
CII
3
3
Rout = rds7||gm6gm8rds8 gm6gm8rds8
Page 6.2-28
C2
-A
R1
-A
R1
C1
C1 C3(1+A)
RootID01
C2
+A
R1
+A
R1
C1
C1 C3(1-A)
RootID02
Page 6.2-29
F(s)
vin
+
vout
A(s)
RootID03
s
A(s)p +1
1
Vout
A(s)
A(s)
1
RII
vout
v''
M6
v'
Fig. 120-15
Page 6.2-30
Feedforward Compensation
Use two parallel paths to achieve a LHP zero for lead compensation purposes.
RHP Zero
LHP Zero
Cc
Cc
-A
Vi
Vout
Inverting
High Gain
Amplifier
CII
Cc
Vi
Vout
Inverting
High Gain
Amplifier
RII
CII
Vi
Vout
+1
RII
Cc
+
Vi
-
+
gmIIVi
CII
RII
Vout
-
Fig.430-09
ACc
Vout(s)
s + gmII/ACc
Page 6.2-31
Self-Compensated Op Amps
Self compensation occurs when the load capacitor is the compensation capacitor (can
never be unstable for resistive feedback)
|dB|
Rout(must be large)
+
Gm
-
vin
+
Av(0) dB
-20dB/dec.
vout
Rout
Fig. 430-10
CL
Increasing CL
0dB
Voltage gain:
vout
vin = Av(0) = GmRout
Dominant pole:
-1
p1 = RoutCL
Unity-gainbandwidth:
Gm
GB = Av(0)|p1| = C
L
Stability:
Large load capacitors simply reduce GB but the phase is still 90 at GB.
CMOS Analog Circuit Design
Page 6.2-32
M3
VDD
M4
C c I5
M6
I6 ICL
M3
M4
C c I5
vout
vin>>0
+
M1
Assume a
virtural
ground
M2
CL
I7
vin<<0
+
M1
I5
+
VBias
-
Assume a
virtural
ground
M2
M6
I6=0
ICL
vout
CL
I7
I5
M7
M5
+
VBias
-
VSS
Positive Slew Rate
I5 I6-I5-I7 I5
, CL = Cc because I6>>I5
Cc
SR+ = min
M7
M5
VSS
Negative Slew Rate
Fig. 140-05
I5 I7-I5 I5
,
= C if I7>>I5.
c
Cc CL
SR- = min
Therefore, if CL is not too large and if I7 is significantly greater than I5, then the slew rate
of the two-stage op amp should be,
I5
SR = C
c
CMOS Analog Circuit Design
Page 6.3-1
M4
Cc
vout
vin
+
M1
+
VBias
-
CL
M2
M7
M5
VSS
Fig. 6.3-1
Notation:
Wi
Si = Li = W/L of the ith transistor
Page 6.3-2
S7
S7
S6 2S7
called the balance conditions
4.) For balance, I6 must equal I7
S4 = S5
5.) So if the balance conditions are satisfied, then VDG4 = 0 and M4 is saturated.
CMOS Analog Circuit Design
Page 6.3-3
2IDS
Page 6.3-4
Op Amp Specifications
The following design procedure assumes that specifications for the following parameters
are given.
1. Gain at dc, Av(0)
Max. ICMR
and/or p3
2. Gain-bandwidth, GB
VDD
Vout(max)
+
+
V
3. Phase margin (or settling time)
SG6
VSG4
M6
gm6 or
4. Input common-mode range, ICMR
M3
M4
Proper Mirroring
Cc
I6
5. Load Capacitance, CL
VSG4=VSG6
g
GB = m1
vout
Cc
6. Slew-rate, SR
CL
Cc 0.2CL
vin M1
M2
7. Output voltage swing
(PM = 60)
+
8. Power dissipation, Pdiss
I5
Min. ICMR
I5 = SRCc
Vout(min)
+
VBias
-
M5
M7
VSS
Fig. 160-02
Page 6.3-5
I5 = SR .Cc
or
I5 10
2 .Ts
3. Design for S3 from the maximum input voltage specification.
I5
S3 = K'3[VDD Vin(max) |VT03|(max) + VT1(min)]2
4. Verify that the pole of M3 due to Cgs3 and Cgs4 (= 0.67W3L3Cox) will not be dominant by
assuming it to be greater than 10 GB
gm3
2Cgs3 > 10GB.
CMOS Analog Circuit Design
Page 6.3-6
S
=
gm6 = 2.2gm2(CL/Cc) and gm4 =
6 gm4S4
S4I4 S4
2KP'S4I4
8. Calculate I6 from
gm62
I6 = 2K'6S6
Check to make sure that S6 satisfies the Vout(max) requirement and adjust as necessary.
9. Design S7 to achieve the desired current ratios between I5 and I6.
S7 = (I6/I5)S5
(Check the minimum output voltage requirements)
Page 6.3-7
Page 6.3-8
(W/L)3 = (W/L)4 = 15
(W/L)3 = (50x10-6)[2.5 2 .85 + 0.55]2 = 15
Page 6.3-9
(W/L)5 = 4.5
(W/L)5 = (110x10-6)(0.35)2 = 4.49 4.5
Page 6.3-10
Page 6.3-11
(W/L)7 = 14
Let us check the Vout(min) specification although the W/L of M7 is so large that this is
probably not necessary. The value of Vout(min) is
295
11014 = 0.351V
which is less than required. At this point, the first-cut design is complete.
10.) Now check to see that the gain specification has been met
(92.45x10-6)(942.5x10-6)
Av = 15x10-6(.04 + .05)95x10-6(.04 + .05) = 7,697V/V
which exceeds the specifications by a factor of two. .An easy way to achieve more gain
would be to increase the W and L values by a factor of two which because of the
decreased value of would multiply the above gain by a factor of 20.
11.) The final step in the hand design is to establish true electrical widths and lengths
based upon L and W variations. In this example L will be due to lateral diffusion only.
Unless otherwise noted, W will not be taken into account. All dimensions will be
rounded to integer values. Assume that L = 0.2m. Therefore, we have
Vout(min) = VDS7(sat) =
Page 6.3-12
M3
VDD = 2.5V
M4
15m
1m
M6
Cc = 3pF
M1
30A
vin
+
4.5m
1m
3m
1m
3m
1m
vout
M2
CL =
10pF
95A
30A
14m
1m
4.5m
M5 1m
VSS = -2.5V
M8
94m
1m
M7
Fig. 6.3-3
Page 6.3-13
Incorporating the Nulling Resistor into the Miller Compensated Two-Stage Op Amp
Circuit:
VDD
M11
VA
M3
M4 V
B
CM
M10
VC
vin-
M6
M8
Cc
vout
vin+
M1
M2
CL
IBias
M12
M9
M5
VSS
M7
Fig. 160-03
Page 6.3-14
The resistor, Rz, is realized by the transistor M8 which is operating in the active region
because the dc current through it is zero. Therefore, Rz, can be written as
vDS8
1
Rz = iD8
= KPS8(VSG8-|VTP|)
VDS8=0
The bias circuit is designed so that voltage VA is equal to VB.
W 11
I10 W 6
S10S6I6
Cc
Page 6.3-15
3pF
(W/L)8 = 3pF+10pF
19495A
= 5.63 6
15A
Page 6.3-16
2I10
KPS10 + |VTP| =
215
501 + 0.7 = 1.474V
Page 6.3-17
To cancel p2,
Cc+CL
1
z1 = p2 Rz = gm6ACC = gm6B
Which gives
C
gm6B = gm6ACc+CL
M11
M3
M6
vout
vin
+
M1
M2
M6B
M10
M4
CL
Cc
M8
VSS
M9
M7
Fig. 6.3-4A
Page 6.3-18
VDD
M6
M4
vout
M2
K2IBias
K1IBias
M5
VSS
M7
Fig. 6.3-04D
10-2
10-3
1
10
IBias
100
IBias(ref)
Fig. 160-05
Page 6.3-19
L1 L2 L3
Poly
Diffusion
Diffusion
L
CMOS Analog Circuit Design
Fig. 6.3-5
P.E. Allen - 2004
Page 6.3-20
;;;;;
;;;
;
;;;;
;
;;
;
;;
;
;;;;
;;;
; ;
;;;;
;
;;
;
;;
;;;;;
;;;;;;;;;;;;
;;;;;;
Metal 1
Poly
Diffusion
Contacts
Ground
(a)
Input
Output
Ground
(b)
Figure 6.3-6 The layout of a 5-to-1 current mirror. (a) Layout which minimizes
area at the sacrifice of matching. (b) Layout which optimizes matching.
Page 6.3-21
;;;;
;;;;
;;;;
;;
;;;;
;;;;
;;;;
;;
;
2
;;
;;
;;
;
Drain 2
Gate 2
Source 2
Drain 1
Gate 1
Source 1
Metal 2
Metal 1
Poly
Diffusion Contacts
Figure 6.3-7 The layout of two transistors with a 1.5 to 1 matching using
centroid geometry to improve matching.
Page 6.3-22
Reduction of Parasitics
The major objective of good layout is to minimize the parasitics that influence the design.
Typical parasitics include:
Capacitors to ac ground
Series resistance
Capacitive parasitics is minimized by minimizing area and maximizing the distance
between the conductor and ac ground.
Resistance parasitics are minimized by using wide busses and keeping the bus length
short.
For example:
At 2m/square, a metal run of 1000m and 2m wide will have 1 of resistance.
At 1 mA this amounts to a 1 mV drop which could easily be greater than the least
significant bit of an analog-digital converter. (For example, a 10 bit ADC with VREF =
1V has an LSB of 1mV)
Page 6.3-23
;;
;;
;;
;;
Source
Metal 1
Poly
Gate
Source
Drain
Source
Diffusion
Contacts
Note: Can get more W/L in less area with the above geometry.
Page 6.3-24
M7
M5
M8
M6A
M9
M6
VPBias1
Bandgap
Voltage,
V
- BG
M10
VDD
M5A
VPBias2
R4
M11
M13
M12
M14
R2A
R3
Master
Voltage
Reference
Circuit
M4
M3
M1
M2
IPTAT
Q2
Q1
R2
R1
Q3
Slave
Bias
Circuit
M15
M16
R1A
M3A
IREF
VNBias2
M1A
Rext
VNBias1
xn
M4A
M2A
Page 6.4-1
Vdd
Av(Vdd=0)
PSRR = A (V =0)
dd
in
Vin
V2
V1
VDD
Vout
Vss
VSS
Fig.180-01
Vdd
V2
V2
V1
Av(V1-V2)
VDD
Vout
Vss
V1
VSS
Vout
AddVdd
Fig. 180-02
=
(Good for frequencies up to GB)
dd 1+Av Av PSRR+
CMOS Analog Circuit Design
Page 6.4-2
M3
Cc
M4
VDD
M6
M2
CII
CI
M5
gm6(V1-Vdd)
I3
rds1
rds2
I3
gm1V5
gm2V5
Vout
Vdd
M1
1
gm3
rds4
Vdd - I3 gm1Vout
gm3
rds5
+
M7
V5 -
V1
rds6
Cc
+
rds7
CII
CI
Vout
VBias
V5 0
gm6(V1-Vdd)
VSS
rds4
Vdd
gds1Vdd
rds2
Fig. 180-03
gm1Vout
+
V1
-
rds6
Cc
CI
CII
+
Vout
rds7
Page 6.4-3
sg
C
mII
c
+ 1
G g
I
ds6
where gmII > gmI and that all transconductances are larger than the channel
conductances.
s
s
sCc sCII
+
1
+
1
+
1
+
1
g
GIIAvo GB
|p2|
Vdd gmIgmII gmI
mII
+
PSRR = V
=Gg
=
gds6 sGIIAvo
sgmIICc
out I ds6
GIgds6 + 1
+ 1
g
ds6GB
Page 6.4-4
|PSRR+(j)| dB
gds6
gds6GB
GIIAv0
GB |p2|
Fig. 180-04
At approximately the dominant pole, the PSRR falls off with a -20dB/decade slope and
degrades the higher frequency PSRR + of the two-stage op amp.
Using the values of Example 6.3-1 we get:
PSRR+(0) = 68.8dB,
z1 = -5MHz, z2 = -15MHz
and p1 = -906Hz
Page 6.4-5
M3
M1
Cc
M4
M2
Vout
CII
CI
M5
M7
Vout
Vdd
VDD
M6
Cc
Vdd
Rout
Vout
0dB
1
RoutCc
Other sources
of PSRR+
besides Cc
VBias
VSS
Fig. 180-05
Page 6.4-6
M4
Cc
VDD
M6
Vout
Cc
M1
M2
M5
VBias
CII
CI
M7 V
ss
VBias grounded
gmIVout
RI
CI
gmIIV1
CII
RII
gm7Vss
+
Vout
-
VSS
Fig. 180-06
Page 6.4-7
s(C
+
C
)
c
I
GI + 1
s
sCc sCII
s
+
1
+1
+
1
+
1
GB
|p
|
g
g
G
A
Vss gmIgmII mI
mII
II v0
g
g
sC
s
c
m7
mI
+1
G + 1
G
GB
I
I
Comments:
PSRR- zeros = PSRR + zeros
DC gain Second-stage gain,
PSRR- pole (Second-stage gain) x (PSRR+ pole)
Assuming the values of Ex. 6.3-1 gives a gain of 23.7 dB and a pole -147 kHz. The dc
value of PSRR- is very poor for this case, however, this case can be avoided by correctly
implementing VBias which we consider next.
CMOS Analog Circuit Design
Page 6.4-8
M1
M4
M2
Cc
VDD
M6
Vout
CII
CI
Cc
Vss
M5
VBias
M7 V
ss
rds5
gmIVout
CI
RI
+
V1 gmIIV1
-
Cgd7
rds7
CII
rds6
+
Vout
-
VSS
Fig. 180-07
If the value of VBias is independent of Vss, then the model shown results. The nodal
equations for this model are
0 = (GI + sCc + sCI)V1 - (gmI + sCc)Vout
and
(gds7 + sCgd7)Vss = (gmII - sCc)V1 + (GII + sCc + sCII + sCgd7)Vout
Again, solving for Vout/Vss and inverting gives
Vss s2[CcCI+CICII+CIICc+CICgd7+CcCgd7]+s[GI(Cc+CII+Cgd7)+GII(Cc+CI)+Cc(gmIIgmI)]+GIGII+gmIgmII
Vout =
(sCgd7+gds7)(s(CI+Cc)+GI)
Page 6.4-9
sC
s(C
+C
)
gd7
I
c
+1 G
+ 1
g
I
ds7
PSRR- = V g
sCc
gds7 +1 GI + 1
Comments:
DC gain has been increased by the ratio of GII to gds7
Two poles instead of one, however the pole at -gds7/Cgd7 is large and can be ignored.
Using the values of Ex. 6.3-1 and assume that Cds7 = 10fF, gives,
PSRR-(0) = 76.7dB
CMOS Analog Circuit Design
and
Page 6.4-10
Frequency Response of the Negative PSRR of the Two-Stage Op Amp with VBias
Connected to VSS
;;
;;
;;
;;
;;
GIIAv0
gds7
|PSRR-(j)| dB
Invalid
region
of
analysis
GI
Cc
GB |p2|
Fig. 180-08
Page 6.4-11
M1
M4
M2
M5
VBias
Cc
VDD
M6
Vout
VBias
CII
CI
M5 or M7
iss
Vss
VSS
M7 V
ss
VSS
VBias grounded
Fig. 180-09
V = gm7Zout = gm7Rout sR C +1
ss
out out
Vout
Vss
20 to
40dB
0dB
CMOS Analog Circuit Design
1
RoutCout
Fig.180-10
P.E. Allen - 2004
Page 6.4-12
M1
M4
M2
Cc
Vout
rds7
CII
CI
M7
VBias
vout
Vss
rds7
Vss
M5
What is Zout?
Vt
Zout = I
t
VDD
M6
Zout
VSS
Fig. 180-11
gmIVt
It = gmIIV1 = g GI+sCI+sCc
GI+s(CI+Cc)
Thus, Zout = gmIgMII
mII
It
Cc CII+Cgd7
gmIVout
CI
RI
+
V1 gmIIV1
-
rds6||rds7
+
Vout
-
Vt
Fig.180-12
rds7
1+
Vss
Zout s(Cc+CI) + GI+gmIgmIIrds7
-GI
V
= 1 =
Pole at C +C
s(Cc+CI) + GI
out
c I
The two-stage op amp will never have good PSRR because of the Miller compensation.
CMOS Analog Circuit Design
Page 6.5-1
Page 6.5-2
VDD
M4
VDD
Implementation of the
floating voltage VBias.
M3
M4
MB3
MB4
MC3
MC3
MC4
MC4
vo1
vo1
R
MB5
MC2
MC1
M1
M2
VBias
+v
vin
+ 2
in
2+
VBias
-
MC1
M1
M5
VSS
+v
+
VBias
MB1
MC2
M2
MB2
in
2+
VBias
-
vin
+ 2
M5
VSS
Fig. 6.5-1
Page 6.5-3
Page 6.5-4
M4
MT2
MB3
MB4
MC3
M6
MC4
MT1
Cc
vo1
vout
MB5
MC1
M1
+v
+
VBias
MB1
MC2
M2
MB2
vin
+ 2
in
2-
M7
M5
+
VBias
-
VSS
Fig. 6.5-2
p1
z1
Fig. 6.5-2
Page 6.5-5
M4
Rz
vin
+
M1
+
VBias
-
M2
VBP
Cc
MC6
VBN
MC7
vout
CL
M7
M5
VSS
Fig. 6.5-3
Av = gmIgmIIRIRII
where gmI = gm1 = gm2,
gmII = gm6,
1
2
RI = gds2 + gds4 = (2 + 4)ID5 and RII = (gmC6rdsC6rds6)||(gmC7rdsC7rds7)
Comments:
The second-stage gain has greatly increased improving the Miller compensation
The overall gain is approximately (gmrds)3 or very large
Output pole, p2, is approximately the same if Cc is constant
The zero RHP is the same if Cc is constant
CMOS Analog Circuit Design
Page 6.5-6
vout = g
+ g
R
m3 2
m4 2 II
M6
M4
M15
M8
M3
vin
+
M1
M2
M14
R1
M9
M7
R2
vout
M12
M10
M5
+
VBias
-
CL
M11
M13
VSS
Fig. 6.5-4
m1 gm2
= 2 + 2 kvin RII = gm1kRII vin
where
RII = (gm7rds7rds6)||(gm12rds12rds11)
and
gm8 gm6
k = gm3 = gm4
This op amp is balanced because the drain-to-ground loads for M1 and M2 are identical.
TABLE 1 - Design Relationships for Balanced, Cascode Output Stage Op Amp.
gm1gm8
1 gm1gm8 gm2gm6
Iout
GB = gm3CL
Av = 2 gm3 + gm4 RII
Slew rate = CL
I5 1/2
I5 1/2
Vin(max) = VDD 3 |VTO3|(max) +VT1(min)
Vin(min) = VSS + VDS5 + 1 + VT1(min)
CMOS Analog Circuit Design
Page 6.5-7
Page 6.5-8
2I6
2I7
500 A
=
=
K'PS6
K'PS7
(50 A/V2)S6
which gives S6 = S7 = S8 = 40 and S3 = S4 = (40/2.5) = 16.
5.) Next the values of R1 and R2 are designed. For the resistor of the self-biased cascode
we can write
R1 = VDS12(sat)/250A = 2k and R2 = VSD7(sat)/250A = 2k
0.5 =
Page 6.5-9
Page 6.5-10
Page 6.5-11
Page 6.5-12
;;;
;;
;;;;;;;
Thin
oxide
Poly I
Poly II
n-channel
n+
n+
p substrate/well
Fig. 6.5-5
Minimum Poly
separation
;;;;;;;;
Thin
oxide
Poly I
Poly I
n+ n-channel n+ n-channel
n+
p substrate/well
Fig. 6.5-5A
Input Common Mode Range for Two Types of Differential Amplifier Loads
VDD-VSG3+VTN
+
VSG3
Input
- M3
Common
Mode
M1
Range
VSS+VDS5+VGS1
+
VBias
-
VDD-VSD3+VTN
VDD
+
VSD4
M4 M2
M5 vicm
VSS
Differential amplifier with
a current mirror load.
VDD
+
V
Input SD3
Common - M3
Mode
Range
M1
VSS+VDS5+VGS1
+
VBias
-
VSD4
M4 VBP
M2
M5 vicm
VSS
Differential amplifier with
Fig. 6.5-6
current source loads.
In order to improve the ICMR, it is desirable to use current source (sink) loads without
losing half the gain.
The resulting solution is the folded cascode op amp.
Page 6.5-14
I2
I1
+
vin
-
M13 M6 I6
M1
M2
M7 I7
vout
R1
R2
I3
+
VBias
-
M3
M8
CL
M9
M12
M10
M11
VSS
Fig. 6.5-7
Comments:
I4 and I5, should be designed so that I6 and I7 never become zero (i.e. I4=I5=1.5I3)
This amplifier is nearly balanced (would be exactly if RA was equal to RB)
Self compensating
Poor noise performance, the gain occurs at the output so all intermediate transistors
contribute to the noise along with the input transistors. (Some first stage gain can be
achieved if RA and RB are greater than gm1 or gm2.
CMOS Analog Circuit Design
Page 6.5-15
gm7vgs7
RB
rds2 rds5
i7
vgs7 rds7 i
10
+
RII
+
vout
-
Fig. 140-07
rds6+R2+(1/gm10) 1
rds7 + RII
RII
and
R
=
B
1 + gm7rds7 gm7rds7 where RII gm9rds9rds11
gm6
1 + gm6rds6
The small-signal voltage transfer function can be found as follows. The current i10 is
written as
-gm1(rds1||rds4)vin -gm1vin
i10 = 2[RA + (rds1||rds4)] 2
and the current i7 can be expressed as
gm2(rds2||rds5)vin
gm2vin
gm2vin
RII(gds2+gds5)
i7 = RII
=
=
where
k
=
RII(gds2+gds5) 2(1+k)
gm7rds7
2g r + (rds2||rds5) 21 + g r
m7 ds7
m7 ds7
The output voltage, vout, is equal to the sum of i7 and i10 flowing through Rout. Thus,
vout gm1
gm2
2+k
=
+
R
=
Page 6.5-16
Page 6.5-17
Page 6.5-18
R
Vss
Cgd11
VGSG9
Cgd9
M9
Vss
Vss
VGS11
rds9
Vout
Cgd9
Vss
Cout
rds11
M11
Vss
Rout
+
Vout
-
Fig. 6.5-9A
This model assumes that gate, source and drain of M11 and the gate and source of M9 all
vary with VSS.
We shall examine Vout/Vss rather than PSRR-. (Small Vout/Vss will lead to large PSRR-.)
The transfer function of Vout/Vss can be found as
sCgd9Rout
Vout
Page 6.5-19
|Avd()|
1
Cgd9Rout
Dominant
pole frequency
0dB
Cgd9
Cout
GB
Vout
Vss
Other sources of Vss injection, i.e. rds9
log10()
Fig. 6.5-10A
We see that the PSRR of the cascode op amp is much better than the two-stage op amp.
Page 6.5-20
Maximum input
CM
2I4
2
S4 = S5 =K V -V (max)+V
P DD in
T1
Differential
Voltage Gain
gm2
vout gm1
2+k
=
+
2(1+k)Rout = 2+2k gmIRout
vin 2
Power dissipation
Pdiss = (VDD-VSS)(I3+I12+I10+I11)
10
Comments
RII(gds2+gds4)
gm7rds7
Page 6.5-21
Page 6.5-22
= 91.6
I3
100
2
2 110x10-6-1.5+2.5
-0.7
KNVin(min)-VSS- K S - VT1
11035.9
N 1
We need to check that the values of S4 and S5 are large enough to satisfy the maximum
input common mode voltage. The maximum input common mode voltage of 2.5 requires
2I4
2125A
S4 = S5 K [V -V (max)+V ]2 =
-6
50x10 A/V2[0.7V]2 = 10.2
P
DD in
T1
which is much less than 80. In fact, with S4 = S5 = 80, the maximum input common mode
voltage is 3V. Finally, S12, is given as
125
S12 = 100 S3 = 114.53
The power dissipation is found to be
Pdiss = 5V(125A+125A+125A) = 1.875mW
CMOS Analog Circuit Design
Page 6.5-23
gm = 2755080 = 774.6S
1 1
RII gm9rds9rds11 = (774.6S)3S 3S = 86.07M
1
1
Rout 86.07M||(774.6S)3.75S 2S+6.25S = 19.40M
RII(gds2+gds4) 86.07M(2S+6.25S)(3.75S)
=
= 3.4375
774.6S
gm7rds7
The small-signal, differential-input, voltage gain is
2+k
2+3.4375
Avd = 2+2k gmIRout = 2+6.875 0.628x10-319.40x106 = 7,464 V/V
The gain is larger than required by the specifications which should be okay.
k=
Page 6.5-24
Page 6.6-1
Page 6.6-2
vOUT
VDD
Simulation:
RL
This circuit will give the voltage transfer
CL
VSS
function curve. This curve should identify:
Fig. 240-01
1.) The linear range of operation
2.) The gain in the linear range
3.) The output limits
4.) The systematic input offset voltage
5.) DC operating conditions, power dissipation
6.) When biased in the linear range, the small-signal frequency response can be
obtained
7.) From the open-loop frequency response, the phase margin can be obtained (F = 1)
Measurement:
This circuit probably will not work unless the op amp gain is very low.
CMOS Analog Circuit Design
Page 6.6-3
vOUT
vIN
CL
C
RL
VDD
VSS
Fig. 240-02
Op Amp
Open Loop
Frequency
Response
Av(0)
0dB
1
RC
Av(0)
RC
log10(w)
Fig. 240-03
Page 6.6-4
Magnitude, dB
80
vOUT = Av(s)R+(1/sC)vOUT +vIN
|Av(j)|
60
-1/RC
= Av(s)s+(1/RC)vOUT +vIN
40
Vout(j)
vOUT
-[s +(1/RC)]Av(s)
Vin(j)
20
vIN = s +(1/RC)+Av(s)/RC
0
-(s+0.01)
-[s +(1/RC)]
= s +0.01
= s +(1/RC)
-20
0.001
0.1
10
1000
105
107
Av(s) +0.01
Av(s) +1/RC
S01E2S2
Radian Frequency (radians/sec)
Substituting, Av(s) gives,
-2x106s -2x104
-2x106s -2x104
-2x106(s +0.01)
vOUT
=
=
=
vIN (s+0.01)(s+500)+2x104 s2+500s +2x104 (s+41.07)(s+1529.72)
CMOS Analog Circuit Design
Page 6.6-5
R
+
vi
-
vOUT
CL
RL
VDD
VSS
Fig. 240-04
Make R as large and measure vout and vi to get the open loop gain.
Page 6.6-6
vOUT=VOS
VDD
VOS
CL
VSS
RL
Fig. 6.6-4
Page 6.6-7
V
+ OS-
vout
VDD
vcm
-
CL
RL
VSS
Fig. 6.6-5
Make sure that the output voltage of the op amp is in the linear region.
100k
+
-
Page 6.6-8
vOS
vSET
100k
10k
vOUT
VDD
10
vI
-
CL
VSS
RL
Fig. 240-07
Note:
1.) PSRR- can be measured similar to
PSRR+ by changing only VSS.
2.) The 1V perturbation can be
replaced by a sinusoid to measure
CMRR or PSRR as follows:
1000vdd
1000vss
PSRR+ = vos , PSRR- = vos
1000vcm
and CMRR = v
os
Page 6.6-9
100k
+
-
vos
100k
vicm
10k
vOUT
VDD
10
vicm
vi
-
CL
RL
VSS
Fig. 240-08
vos
But vid = vi and vos 1000vi = 1000vid vid = 1000
Page 6.6-10
VDD
V2 -
Av(V1-V2)
Vout
V1
V1 +
Vcm
VSS
Vcm
Vout
AcVcm
Fig. 6.6-7
V1+V
Vout = Av(V1-V2) A
2 = -AvVout AcmVcm
Acm
Acm
Vout = 1+A Vcm A Vcm
v
v
Av Vcm
cm
Page 6.6-11
80
150
Arg[CMRR] Degrees
85
|CMRR| dB
75
70
65
60
55
50
100
50
0
-50
-100
-150
45
-200
10
100
1000
104
105
106
Frequency (Hz)
107
108
10
100
1000
104
105
106
Frequency (Hz)
107
108
Fig. 240-10
Page 6.6-12
V1
Av(V1-V2)
VDD
V1
VSS
Vss
Vout
Vss = 0
AddVdd
Fig. 6.6-9
Page 6.6-13
IDD
vOUT
VDD
1
1
vIN
vIN
ISS
CL
RL
VSS
ICMR
Also, monitor
IDD or ISS. Fig.240-11
Page 6.6-14
+
vI
-
VDD
vOUT
Without RL
With RL
VO1
VO2
VOS
RL
vI(mV)
VSS
Fig. 240-12
V 01
Rout = RL V02 1
Method 2:
R
100R
- VDD
vIN
Rout
VSS
Ro
Fig. 240-13
1
Av -1 100Ro
1
Page 6.6-15
Settling Error
Tolerance
VDD
+SR -SR
vin
1
-
CL
RL
VSS
vout
1
Settling Time
Feedthrough
t
Fig. 240-14
If the slew rate influences the small signal response, then make the input step size small
enough to avoid slew rate (i.e. less than 0.5V for MOS).
Page 6.6-16
100
60
10
50
Phase Margin
40
Overshoot
30
Overshoot (%)
70
1.0
20
10
0
0.2
0.4
= 1
2Q
0.6
0.8
0.1
Fig. 240-15
Page 6.6-17
Page 6.6-18
8 VDD
6 vout
+
9 VSS
Fig. 240-17
.SUBCKT OPAMP 1 2 6 8 9
M1 4 2 3 3 NMOS1 W=3U L=1U AD=18P AS=18P PD=18U PS=18U
M2 5 1 3 3 NMOS1 W=3U L=1U AD=18P AS=18P PD=18U PS=18U
M3 4 4 8 8 PMOS1 W=15U L=1U AD=90P AS=90P PD=42U PS=42U
M4 5 4 8 8 PMOS1 W=15U L=1U AD=90P AS=90P PD=42U PS=42U
M5 3 7 9 9 NMOS1 W=4.5U L=1U AD=27P AS=27P PD=21U PS=21U
M6 6 5 8 8 PMOS1 W=94U L=1U AD=564P AS=564P PD=200U PS=200U
M7 6 7 9 9 NMOS1 W=14U L=1U AD=84P AS=84P PD=40U PS=40U
M8 7 7 9 9 NMOS1 W=4.5U L=1U AD=27P AS=27P PD=21U PS=21U
CC 5 6 3.0P
.MODEL NMOS1 NMOS VTO=0.70 KP=110U GAMMA=0.4 LAMBDA=0.04 PHI=0.7
+MJ=0.5 MJSW=0.38 CGBO=700P CGSO=220P CGDO=220P CJ=770U CJSW=380P
+LD=0.016U TOX=14N
.MODEL PMOS1 PMOS VTO=-0.7 KP=50U GAMMA=0..57 LAMBDA=0.05 PHI=0.8
+MJ=0.5 MJSW=.35 CGBO=700P CGSO=220P CGDO=220P CJ=560U CJSW=350P +LD=0.014U TOX=14N
IBIAS 8 7 30U
.ENDS
Page 6.6-19
Page 6.6-20
VOS
vOUT(V)
1
0
-1
-2
-2.5
-2
-1.5
-1.0 -0.5
0
0.5
vIN(mV)
1.5
Fig. 240-18
Page 6.6-21
80
150
Magnitude (dB)
60
40
20
0
100
50
0
-50
-100
-20
-150
GB
-40
Phase Margin
GB
-200
10
100
1000
105
106
104
Frequency (Hz)
107
108
10
100
1000
104
105
106
Frequency (Hz)
107
108
Fig. 6.6-16
Page 6.6-22
VDD
Subckt.
+
vout
vin
VSS
Fig. 6.6-16A
40
ID(M5)
30
20
vOUT (V)
Input CMR
10
ID(M5) A
-3
-3
-2
-1
0
vIN(V)
Fig. 240-21
Page 6.6-23
100
Arg[PSRR+(j)] (Degrees)
|PSRR+(j)| dB
80
60
40
20
0
-20
10
100
1000
10
10
10
Frequency (Hz)
10
10
50
-50
-100
10
100
1000
104
105
106
Frequency (Hz)
107
108
Fig. 240-22
Page 6.6-24
120
Arg[PSRR-(j)] (Degrees)
150
|PSRR-(j)| dB
100
80
60
PSRR+
40
20
10
100
50
0
-50
-100
-150
-200
100
1000
10
10
10
Frequency (Hz)
10
10
10
100
1000
104
105
106
Frequency (Hz)
107
108
Fig. 240-23
Page 6.6-25
0.15
0.1
0.05
vout(t)
Volts
Volts
0.5
vin(t)
-0.5
vout(t)
-0.05
vin(t)
-1
-0.1
-1.5
-0.15
0
2
3
Time (Microseconds)
2.5
3.0
3.5
4.0
Time (Microseconds)
4.5
Fig. 240-24
VDD
M6
Cc iCc
iCL
vout
dvout
dt
CL
95A
+
VBias
-
M7
VSS
Fig. 240-25
Page 6.6-26
Design
(Ex. 6.3-1)
>5000
5 MHz
-1V to 2V
>10 (V/sec)
< 2mW
2V
60
-
Simulation
(Ex. 1)
10,000
5 MHz
-1.2 V to 2.4 V,
+10, -7(V/sec)
0.625mW
+2.3V, -2.2V
87
106
65
122.5k
Page 6.6-27
Example 6.6-3
Why is the negative-going overshoot
larger than the positive-going overshoot
on the small-signal transient response of
the last slide?
Consider the following circuit and
waveform:
VDD = 2.5V
94/1
M6 i6
iCc
iCL
0.1V
vout
Cc
95A
CL
-0.1V
VBias
M7
0.1s
0.1s
Fig. 240-26
VSS = -2.5V
During the rise time,
iCL = CL(dvout/dt )= 10pF(0.2V/0.1s) = 20A and iCc = 3pf(2V/s) = 6A
i6 = 95A + 20A + 6A = 121A gm6 = 1066S (nominal was 942.5S)
During the fall time, iCL = CL(-dvout/dt) = 10pF(-0.2V/0.1s) = -20A
and iCc = -3pf(2V/s) = -6A
i6 = 95A - 20A - 6A = 69A
gm6 = 805S
The dominant pole is p1 (RIgm6RIICc)-1 but the GB is gmI/Cc = 94.25S/3pF =
31.42x106 rads/sec and stays constant. Thus we must look elsewhere for the reason.
Recall that p2 gm6/CL which explains the difference.
p2(95A) = 94.25x106 rads/sec, p2(121A) = 106.6 x106 rads/sec, and p2(69A) =
80.05 x106 rads/sec. Thus, the phase margin is less during the fall time than the rise time.
Page 6.7-1
Page 6.7-2
vo
v2
R id
v1
Ro v o
R id
Avd (v 1 -v 2 )
v2
v2
(a.)
(b.)
Avd (v -v )
Ro 1 2
(c.)
vo
Ro
Fig. 010-01
Figure 1 - (a.) Op amp symbol. (b.) Thevenin form of simple model. (c.) Norton form of
simple model.
SPICE Description of Fig. 1c
RID 1 2 {Rid}
RO 3 0 {Ro}
GAVD 0 3 1 2 {Avd/Ro}
Page 6.7-3
R1 =
1k
vin
Figure 2 Noninverting
voltage amplifier for Ex. 1.
R2 = 100k
Fig. 010-02
SMALL-SIGNAL CHARACTERISTICS
V(2)/VIN = 1.009E+02
INPUT RESISTANCE AT VIN = 9.901E+08
OUTPUT RESISTANCE AT V(2) = 1.010E-01.
Page 6.7-4
Ric1
Rid
Avd(v1 -v2 )
Ro
Avc v1
2Ro
Avc v2
2Ro
+
Ro
vo
-
Ric2
Linear Op Amp Macromodel
Fig. 010-03
Figure 3 - Simple op amp model including differential and common mode behavior.
SPICE File:
.SUBCKT LINOPAMP 1 2 3
RIC1 1 0 {Ric}
RID 1 2 {Rid}
RIC2 2 0 {Ric}
CMOS Analog Circuit Design
GAVD/RO 0 3 1 2 {Avd/Ro}
GAVC1/RO 0 3 1 0 {Avc/2Ro}
GAVC2/RO 0 3 2 0 {Avc/2Ro}
RO 3 0 {Ro}
.ENDS LINOPAMP
Page 6.7-5
vo
Rid
v2
Avd(0)
(v1 -v2 )
R1
R1
C1
Fig. 010-04
Figure 4 - Macromodel for the op amp including the frequency response of Avd.
Model Using Passive Components with Constant Output Resistance:
v1
Rid
v2
Avd(0)
(v1 -v2 )
R1
R1
C1
v3
Ro
vo
Ro
Fig. 010-05
Page 6.7-6
GAVD/RO 0 3 1 2 1000
R1 3 0 100
C1 3 0 100UF
.ENDS
.AC DEC 10 100 10MEG
.PRINT AC V(21) V(22) V(23)
.PROBE
.END
Page 6.7-7
30dB
20dB
Gain of 10
10dB
0dB
Gain of 1
-10dB
15.9kHz
-20dB
100Hz
1kHz
159kHz
10kHz
1.59MHz
100kHz
1MHz
10MHz
Fig. 010-06
Page 6.7-8
Page 6.7-9
Op Amp Macromodel
Avc v1
2Ro
Ric1
Avc v2
2Ro
C2
R2
Rid
5
3
2
+
Ric2
Avd(v1 -v2 )
Ro
C1
v3
Ro
R1
v4
Ro
Ro
vo
-
Fig. 010-07
Page 6.7-10
Ro
Avd(v1 -v2 )
Ro
L1
vo Avd(v1 -v2 )
R1
C1
R1
kAvd
(v1 -v2 )
Ro
Ro
vo
-
(a.)
v3
Ro
(b.)
Fig. 010-08
Figure 8 - (a.) Independent zero model. (b.) Method of modeling zeros without
introducing new nodes.
Inductor:
Avd(0)
s
Vo(s) = Ro (sL1 + Ro) [V1(s)-V2(s)] = Avd(0)Ro/L1 + 1 [V1(s)-V2(s)] .
Feedforward:
Avd(0)
Vo(s) = (s/1) +11+k(s/1)+k [V1(s)-V2(s)] .
1
The zero can be expressed as
z1 = -11 + k
where k can be + or - by reversing the direction of the current source.
CMOS Analog Circuit Design
Page 6.7-11
GV3/RO 0 4 3 0 0.01
GAVD/RO 4 0 1 2 0.01
RO 4 0 100
.ENDS
.AC DEC 10 1 100MEG
.PRINT AC V(2) VDB(2) VP(2)
.PROBE
.END
Page 6.7-12
VDB(2)
80dB
60dB
40dB
15.9Hz or 100rps
20dB
1.59MHz or 10Mrps
0dB
1Hz
10Hz
Figure 9 - Asymptotic magnitude frequency response of the op amp model of Ex. 6.7-3.
CMOS Analog Circuit Design
Page 6.7-13
Ric1
Ric2
D2
D1
VIH1
RLIM
VIL1
D4
VIL2
Rid
Avc v4
2Ro
D3
+
VIH2
Nonlinear Op
Amp Macromodel
Avc v5
2Ro
Avd(v -v )
4 5
Ro
D5
Ro
VOH
+
D6
+ 10
+ 11 vo
VOL
-
Fig. 010-10
Figure 10 - Op amp macromodel that limits the input and output voltages.
Subcircuit Description
.SUBCKT NONLINOPAMP 1 2 3
RIC1 1 0 {Ricm}
RLIM1 1 4 0.1
D1 4 6 IDEALMOD
VIH1 6 0 {VIH1}
D2 7 4 IDEALMOD
VIL1 7 0 {VIL1}
RID 4 5 {Rid}
RIC2 2 0 {Ricm}
RLIM2 2 5 0.1
D3 5 8 IDEALMOD
VIH2 8 0 {VIH1}
D4 9 5 IDEALMOD
VIL2 9 0 {VIL2}
GAVD/RO 0 3 4 5 {Avd/Ro}
GAVC1/RO 0 3 4 0 {Avc/Ro}
GAVC2/RO 0 3 5 0 {Avc/Ro}
RO 3 0 {Ro}
D5 3 10 IDEALMOD
VOH 10 0 {VOH}
D6 11 3 IDEALMOD
VOL 11 0 {VOL}
.MODEL IDEALMOD D N=0.001
.ENDS
Page 6.7-14
VIL1 7 0 -5V
RID 4 5 1MEG
RIC2 2 0 100MEG
RLIM2 2 5 0.1
D3 5 8 IDEALMOD
VIH2 8 0 5V
D4 9 5 IDEALMOD
VIL2 9 0 -5v
GAVD/RO 0 3 4 5 1000
GAVC1/2RO 0 3 4 0 0.05
GAVC2/2RO 0 3 5 0 0.05
RO 3 0 100
D5 3 10 IDEALMOD
VOH 10 0 10V
D6 11 3 IDEALMOD
VOL 11 0 -10V
.MODEL IDEALMOD D N=0.0001
.ENDS
.DC VIN -15 15 0.1
.PRINT V(2)
.PROBE
.END
Page 6.7-15
-5V
0V
VIN
5V
10V
Fig. 010-11
Page 6.7-16
Io
2
Io
ILimit
2
D1
D3
D2
ILimit
D4
Io
ILimit
2
Io
2
Io
2 Fig. 010-12
v1
4
Rid
2
Avd (v -v )
Ro 1 2
v2
Ro
D3
D4
ILimit
D1
D2
6
D5
+
VOH
D6
7
VOL
vo
Fig. 010-13
Page 6.7-17
Page 6.7-18
V(3)
5V
0V
-5V
-10V
-15V
-10V
-5V
0V
VIN
5V
10V
15V
Fig. 010-14
Page 6.7-19
D3
Rid
v2
Avd(0)
(v1 -v2 )
R1
R1
D1
C1
D4
ISR
D2
7
v4 -v5
Ro
vo
Ro
Fig. 010-15
Page 6.7-20
D2 7 0 IDEALMOD
D3 5 6 IDEALMOD
D4 7 5 IDEALMOD
ISR 6 7 1A
GVO/R0 0 3 4 5 0.01
RO 3 0 100
.MODEL IDEALMOD D N=0.0001
.ENDS
Page 6.7-21
5V
Output
Voltage
0V
-5V
-10V
0s
Input
Voltage
2s
4s
Time
6s
8s
10s
Fig. 010-16
Page 6.7-22
Page 6.8-1
Page 7.0-1
Buffered
High Frequency
Differential
Output
Two-Stage
Op Amp
Low Power
Low Noise
Low Voltage
Fig. 7.0-1
Page 7.1-1
Page 7.1-2
Page 7.1-3
M6
IBias
M12
M2
M18
Cc
M15
R1
M1
VSG18 +
-
M11
M7
M4
I17
M10
M5
R1 M8
+
vin
M17
R1
VSS
vout
CL
VSS
VSG21 +
-
VGS19
I20
M16
M14
VSS
+
VDD V
GS22
-
M19
M13
M3
M22
VDD
M21
M20
Buffer
Fig. 7.1-1
1
Rout = g
1000, Av(0) = 65dB (IBias=50A), and GB = 60MHz for CL = 1pF
m21+gm22
Output bias current?
M18-M19-M21-M22 loop VSG18+VGS19 = VSG21+VGS22
2I18
2I19
2I21
2I22
which gives
+
=
+
KPS18
KNS19
KPS21
KNS22
CMOS Analog Circuit Design
Page 7.1-4
240
14
M7
144
M3 14
100A
C1=8pF
M4
240
14
M6
2400
7.5
C2=5pF
vout
RL
vin +
+
-
vin'
M2
M1
M5
360
7.5
460
7.5
Cross over stage VSS
Input
stage
1400
14
Output Stage
Fig. 7.1-2
This op amp is capable of delivering 160mW to a 100 load while only dissipating 7mW
of quiescent power!
Page 7.1-5
240
14
M7
144
M3 14
240
14
M4
C1=8pF
C2=5pF
M1
M2
vout
100A
vin'
460
7.5
VDD
M6
M5
360
7.5
RL
vout
M6 Active
M1-M3
M6 SaturInverter
M5 Saturated ated
M2-M4
Inverter
M5 Active
0
VSS
VA
VB
VDD
VSS
vin'
Fig. 7.1-3
Page 7.1-6
Performance
6V
7 mW
8.1 Vpp
78.1 dB
260kHz
1.7 V/ Hz
55 dB
42 dB
10 mV
Page 7.1-7
Vin(s) = s
op amp
p - 1 p - 1 p - 1
1
2
3
Pole p3'
vout
x1
Output
stage
CL
RL
Fig. 7.1-4
p2
p3'
Compensated poles
Uncompensated poles
p2'
p 1' p 1
p2
p3=p3' p2'
p1' p1
p3
Miller compensation applied around
both the second and the third stage.
CMOS Analog Circuit Design
Page 7.1-8
Error
Gain
Amplifier
Amplifier
viin
Error
Amplifier
M2
+
iout
vout
+
CL
M1
RL
Fig. 7.1-5A
VSS
Comments:
The output resistance will be equal to rds1||rds2 divided by the loop gain
If the error amplifiers are not perfectly matched, the bias current in M1 and M2 is not
defined
Page 7.1-9
M6
M16
-A1+
Cc
VOS
M9
vout
Error Loop
M8
Unbuffered
op amp
+
VBias
-
M17
+
A2
M10
M8A
M6A
M12
M13
VSS
M11
Fig. 7.1-6
The feedback circuitry of the two error amplifiers tries to insure that the voltages in
the loop sum to zero. Without the M9-M12 feedback circuit, there is no way to adjust the
output for any error in the loop. The circuit works as follows:
When VOS is positive, M6 tries to turn off and so does M6A. IM9 reduces thus
reducing IM12. A reduction in IM12 reduces IM8A thus decreasing VGS8A. VGS8A
ideally decreases by an amount equal to VOS. A similar result holds for negative
offsets and offsets in EA2.
CMOS Analog Circuit Design
Page 7.1-10
M4
Cc1
MR1
vin
M1
vout
M2
+
M5
VBias
A1 amplifier
M6A
VSS
Fig. 7.1-7
Page 7.1-11
+
vin
-
+
-
M16
+
VBiasP
-
M4H
M3H
M3
M5A
MP3A
MP4
MP5
M4
MP4A
Cc
M6
MP3
MR1
M9
M8A
Cc2
Cc1
MR2
M10
M1 M2
M2A M1A
MN3A
M6A
M8
M17
M5
MN3
MN4
MN5A
M13
M12
+
VBiasN
-
M3A
M4A
M11
M3HA
MN4A
M4HA
VSS
Compensation:
Uses nulling Miller compensation.
Short circuit protection:
MP3-MN3-MN4-MP4-MP5
MN3A-MP3A-MP4A-MN4A-MN5A
(max. output 60mA)
CMOS Analog Circuit Design
vout
RC
gm1
Fig. 7.1-8
CC
gm6
R1
C1
RL
CL
Page 7.1-12
Simulated Results
7.0 mW
82 dB
500kHz
0.4 mV
85 dB/104 dB
81 dB/98 dB
Measured Results
5.0 mW
83 dB
420 kHz
1 mV
86 dB/106 dB
80 dB/98 dB
0.03%
0.08%
0.13%(1 kHz)
0.32%(4 kHz)
0.05%
0.16%
3 s
0.8 V/s
-
0.13%(1 kHz)
0.20%(4 kHz)
<5 s
0.6 V/s
130 nV/ Hz
49 nV/ Hz
rds6||rds6A 50k
Rout Loop Gain 5000 = 10
CMOS Analog Circuit Design
Page 7.1-13
Transistor/Capacitor
M8A
M13
M9
M10
M11
M12
MP3
MN3
MP4
MN4
MP5
MN3A
MP3A
MN4A
MP4A
MN5A
m/m or pF
481/6
66/12
27/6
6/22
14/6
140/6
8/6
244/6
43/12
12/6
6/6
6/6
337/6
24/12
20/12
6/6
P.E. Allen - 2004
Page 7.1-14
M8
M3
200A 10/1 1/1
M4
M6
1/1 10/1
vout
+
vin
-
CL
M2
M1
10/1
10/1
1/1
M5
M10
10/1
M9
1/1
10/1
M7
VSS
Fig. 7.1-9
Output Resistance:
Ro
Rout = 1+LG
where
1
Ro = gds6+gds7
and
gm2
|LG| = 2gm4 (gm6+gm7)Ro
gm2
.
(gds6+gds7)1 + 2gm4(gm6+gm7)Ro
Page 7.1-15
Example 7.1-1 - Low Output Resistance Using the Simple Shunt Negative Feedback
Buffer
Find the output resistance of above op amp using the model parameters of Table 3.1-2.
Solution
The current flowing in the output transistors, M6 and M7, is 1mA which gives Ro of
1
1000
Ro = ( + )1mA = 0.09 = 11.11k
N P
To calculate the loop gain, we find that
gm2 = 2KN10100A = 469S
gm4 = 2KP1100A = 100S
and
gm6 = 2KP101000A = 1mS
Therefore, the loop gain is
469
|LG| = 100 1211.11 = 104.2
Solving for the output resistance, Rout, gives
11.11k
Rout = 1 + 104.2 = 106 (Assumes that RL is large)
CMOS Analog Circuit Design
Page 7.1-16
;;;
;;
;
;;;
;;;
Emitter
n+ (Emitter)
Base
p+
p- well (Base)
Collector
(VDD)
Collector (VDD)
n+
Base
n- substrate (Collector)
Fig. 7.1-10
Emitter
Comments:
gm of the BJT is larger than the FET so that the output resistance w/o feedback is lower
Can use the lateral or substrate BJT but since the collector is on ac ground, the
substrate BJT is preferred
Current is required to drive the BJT
Page 7.1-17
M8
Q10
vout
M9
CL
RL
M11
Output Buffer
VSS
Fig. 7.1-11
Small-signal output resistance :
r10 + (1/gm9)
1
1
=
+
Rout
gm10 gm9(1+F)
1+F
= 51.6+6.7 = 58.3 where I10=500A, I8=100A, W9/L9=100 and F is 100
Maximum output voltage:
Ic10
2KP
V
ln
vOUT(max) = VDD - VSD8(sat) - vBE10 = VDD t Is10
I8(W 8/L8)
Voltage gain:
gm10RL
vout gm1 gm6
gm9
vin gds2+gds4gds6+gds7gm9+gmbs9+gds8+g101+gm10RL
Compensation will be more complex because of the additional stages.
CMOS Analog Circuit Design
Page 7.1-18
Page 7.1-19
gmbs9 =
2 2F + VBS9 2 0.7+2
where we have assumed that the value of VSB9 is approximately 2V.
300S
Page 7.1-20
M10
+
vin
-
M7
Q8
95A
M1
M2
133A
Cc
vout
IBias
M3
M9
M4
CL
RL
M6
VSS
Output
Buffer
Fig. 7.1-12
Slew Rate:
+
IOUT (1 + F)I7
SR+ = CL =
CL
and
SR- =
Page 7.1-21
R1
V1
gmIIV2
-
+ V +
r
R2 V 2
- gm8V
R3
gm9V1
Vout
Fig. 7.1-13
Nodal equations:
gmIVin = (GI + sCc)V1 sCcV2 + 0Vout
0 = (gmII sCc)V1 + (GII + g + sCc + sC)V2 (g + sC)Vout
0 gm9V1 (gm13 + sC)V2 + (gm13 + sC)Vout
where g > G3
The approximate voltage transfer function is:
(s/z1) 1 (s/z2) 1
V9(s)
Vin(s) Av0(s/p1) 1 (s/p2) 1
where
gmIgmII
gm13 gmII
gm9
1
z1 = Cc
z2 = C + Cc 1 + gmII
Av0 = GIGII
C
gm9
Page 7.1-22
vOUT (Volts)
RL = 100
RL =50
-1
-2
-3
-2
-1.5
-1
-0.5
0
0.5
vIN (Volts)
1.5
Fig. 7.1-14A
Page 7.1-23
Page 7.1-24
SUMMARY
A buffered op amp requires an output resistance between 10 Ro 1000
Output resistance using MOSFETs only can be reduced by,
- Source follower output (1/gm)
- Negative shunt feedback (frequency is a problem in this approach)
Use of substrate (or lateral) BJTs can reduce the output resistance because gm is
larger than the gm of a MOSFET
Adding a buffer stage to lower the output resistance will most likely complicate the
compensation of the op amp
Page 7.2-1
Page 7.2-2
20dB
0dB
-3dB GB
log10()
Fig. 7.2-1
Page 7.2-3
20dB
0dB
-40dB/dec
Next higher pole
A
-3dB GB
log10()
Fig. 7.2-2
Page 7.2-4
-p2 = z1
|Avd(0)| dB
0dB
Fig. 7.2-3
GB
Increase
Old New
|p1|
|p1|
-p1
New
Old
GB
-p1
Old
New
GB |p ||p |
4 3
log10()
|p2|
-40dB/dec
-60dB/dec
-80dB/dec
Before cancelling
p2 by z1 and
increasing p1
Page 7.2-5
Page 7.2-6
This value of GB is designed from the relationship that GB = gm1/Cc. Assuming gm1 is
constant, then Cc = gm1/GB = (94.25x10-6)/(0.330x109) = 286fF. It might be useful to
increase gm1 in order to keep Cc above the surrounding parasitic capacitors (Cgd6 =
20.7fF). The success of this method assumes that there are no other roots with a
magnitude smaller than 10GB.
CMOS Analog Circuit Design
Page 7.2-7
Example 7.2-2 - Increasing the GB of the Folded Cascode Op Amp of Ex. 6.5-3
Use the folded-cascode op amp designed
VDD
in Example 6.5-3 and apply the above
approach to increase the gainbandwidth as
M14 M4 I4 M5 I5
much as possible.
Assume that the
A
B
drain/source areas are equal to 2m times the
RB
RA
width of the transistor and that all voltage
I2
I1
dependent capacitors are at zero voltage.
M13 M6 I6 M7 I7
vout
+
R1
M1 M2
Solution
vin
R2
CL
The poles of the folded cascode op amp are: I3
-1
M9
M8
pA RACA (the pole at the source of M6 )
+
M12
VBias M3
M11
M10
-1
pB R C
(the pole at the source of M7)
VSS
Fig. 6.5-7
B B
-1
p6 (R2+1/gm10)C6 (the pole at the drain of M6)
-gm8
-gm9
p8 C (the pole at the source of M8 )
p
Page 7.2-8
Page 7.2-9
Page 7.2-10
4.) Next, we consider the pole, p8. The capacitance connected to this node is
C8= Cbd10 + Cgd10 + Cgs8 + Cbs8
These capacitors are given as,
Cbs8 = Cbd10 = (770x10-6)(36.4x10-62x10-6) + (380x10-12)(238.4x10-6) = 85.2fF
Cgs8 = (220x10-1236.4x10-6) + (0.67)(36.4x10-610-624.7x10-4) = 67.9fF
and
Cgd10 = (220x10-12)(36.4x10-6) = 8fF
The capacitance C8 is equal to
C8 = 67.9fF + 8fF + 85.2fF + 85.2fF = 0.246pF
Using the gm8 of Ex. 6.5-3 of 774.6S, the pole p8 is found as, -p8 = 3.149x109 rads/sec.
5.) The capacitance for the pole at p9 is identical with C8. Therefore, since gm9 is also
774.6S, the pole p9 is equal to p8 and found to be -p9 = 3.149x109 rads/sec.
6.) Finally, the capacitance associated with p10 is given as
C10 = Cgs10 + Cgs11 + Cbd8
These capacitors are given as
CMOS Analog Circuit Design
Page 7.2-11
Page 7.2-12
MOSFET Op
Amp
gm dependence
BJT Op Amp
IC
IC
W
2K L ID
kT/q = Vt
Maximum gm
1 mA/V
20 mA/V
GB for 10pF
15 MHz
300 MHz
GB for 1pF
150 MHz
3 GHz
Note that the power dissipation will be large for large GB because current is needed for
large gm.
Assumption:
All higher-order roots are above GB.
The larger GB, the more difficult this becomes.
Conclusion:
The best CMOS op amps have a GB of 10-50MHz
The best BJT op amps have a GB of 100-200MHz
Page 7.2-13
Switched Amplifiers
Switched amplifiers are time varying circuits that yield circuits with smaller parasitic
capacitors and therefore higher frequency response. Such circuits are called dynamically
biased.
Switched amplifiers require a nonoverlapping clock
Switched amplifiers only work during a portion of a clock period
Bias conditions are setup on one clock phase and then maintained by capacitance on the
active phase
Switched amplifiers use switches and capacitors resulting in feedthrough problems
Simplified circuits on the active phase minimize the parasitics
Typical clock:
1
t
T
0
CMOS Analog Circuit Design
0.5
1.5
t
2 T Fig. 7.2-3B
P.E. Allen - 2004
Page 7.2-14
CB
M2
1 1
2
vin
ID
vout
COS
M1
1
VSS
Fig. 7.2-4
During phase 1 the offset and bias of the inverter is sampled and applied to COS and CB.
During phase 2 COS is connected in series with the input and provides offset canceling
plus bias for M1. CB provides the bias for M2.
(This circuit illustrates the concept of switched amplifiers but is too simple to illustrate
the reduction of bias parasitics.)
Page 7.2-15
M8
+
VB2
-
M4
M7
vin-
IB
C1
1 vin+
C2
M6
M3
vout
M2
M5 +
VB1
-
M1
VSS
Fig.7.2-5
Page 7.2-16
VDD
+
VB2
-
M8
M7
C1
IB
C2
M6
VDD-VB2-(vin+-vin-)
+
VDD-VB2-vin+
vin+
+
vin+-VSS-VB1
-
M5 +
VB1
-
+
VDD-VB2-vin+
vin- -
C1
+
vin+-VSS-VB1
-
C2
VSS+VB1-(vin+-vin-)
VSS
Equivalent circuit during the 1 clock period
M4
M3
vout
M2
M1
VSS
Equivalent circuit during the 2 clock period.
Fig. 7.2-6
Page 7.2-17
+
VB2
M8 -
M7
1C2
vin-
IB
M6
M5 +
VB1
-
C1
M4
M3
C4 2
1
2
C3
vout
vin+
M2
VSS
M1
Fig. 7.2-7
S. Masuda, et. al., CMOS Sampled Differential Push-Pull Cascode Op Amp, Proc. of 1984 International Symposium on Circuits and Systems,
Montreal, Canada, May 1984, pp. 1211-12-14.
CMOS Analog Circuit Design
P.E. Allen - 2004
Page 7.2-18
io
Ro
+
Current
Amplifier
Fig. 7.2-8A
Requirements:
io = Ai(i1-i2)
Ri1 = Ri2 = 0
Ro =
Ideal source and load requirements:
Rsource =
RLoad = 0
CMOS Analog Circuit Design
Page 7.2-19
-R2Ao
Av(0) = R1(1+Ao)
and
-3dB = A(1+Ao)
P.E. Allen - 2004
Page 7.2-20
R
Voltage Amplifier, R2 > K
R2 Ao
1
dB
R2
R1 1+Ao
Voltage Amplifier, R = K >1
1
Ao
dB
K
1+Ao
Current Amplifier
Ao dB
(1+Ao)A
0dB
GBi
GB1 GB2
log10()
Fig. 7.2-10
Page 7.2-21
M4
vin
R1
M5
R2
M6
M7
M3
vout
IBias
M1
M2
VSS
M8
M9
Fig. 7.2-11
An inverting amplifier with a gain of 10 is achieved if R2 = 20R1 assuming the gain of the
current mirror is unity.
What is the GB of this amplifier?
R2Ao
Ao
1
1
GB = |Av(0)|-3dB = R (1+A ) R C = (1+A )R C = 2R C
1
o
2 o
o 1 o
1 o
where Co is the capacitance seen at the output of the current mirror.
If R1 = 10k and Co = 250fF, then GB = 31.83MHz.
Limitations:
R2
R1>Rin = 1/gm1 and R2 < rds2||rds6
R1 << gm1(rds2||rds6)
CMOS Analog Circuit Design
Page 7.2-22
M7
M13
M8
M9
M6
M5
R4
R1
vin
M12
vout
R2
IBias
M3
M4
M15
M1
M2
VSS
M10
M11
Fig. 7.2-12
New limitations:
R2
1
R1 > gm1 and R2 < gm4rds4rds2||gm6rds6rds8 R1 << gm1(gm4rds4rds2||gm6rds6rds8)
CMOS Analog Circuit Design
Page 7.2-23
Page 7.2-24
Example 7.2-3 - Design of a High GB Voltage Amplifier using Current Feedback Contd
The information required to calculate these capacitors is found from Table 3.2-1. The
various capacitors are,
Cgd4 = Cgd6 = CGDOx10m = (220x10-12)(20x10-6) = 4.4fF
Cbd4 = CJxAD4+CJSWxPD4 = (770x10-6)(20x10-12)+(380x10-12)(44x10-6)
= 15.4fF+16.7fF = 32.1fF
Cbd6 = (560x10-6)(20x10-12)+(350x10-12)(44x10-6) = 26.6fF
Cgs12 = (220x10-12)(20x10-6) + (0.67)(20x10-610-624.7x10-4) = 37.3fF
Therefore,
Co = 4.4fF+32.1fF+4.4fF+26.6fF+37.3fF = 105fF
Note that if we had not reduced the W/L of M2, M4, M6, and M8 that Co would have
easily exceeded 100fF. Since 105fF is close to our original guess of 100fF, let us keep the
values of R1 and R2. If this value was significantly different, then we would adjust the
values of R1 and R2 so that the GB is 500MHz. One must also check to make sure that
the input pole is greater than 500MHz.
The design is completed by assuming that IBias = 100A and that the current in M9
through M12 be 100A. Thus W13/L13 = W14/L14 = 20m,/1m and W9/L9 through
W12/L12 are 20m/1m.
CMOS Analog Circuit Design
Page 7.2-25
R1 = 1k
20
|vout/vin| dB
-20dB/dec
10
R1 = 3.2k
0
-40dB/dec
-10
-20
f-3dB
GB
-30
105
106
107
108
Frequency (Hz)
109
1010
Fig. 7.2-13
Simulation Results:
GB 300MHz
Closed-loop gain = 18dB
f-3dB 38MHz
(Loss of -2dB is attributed to source follower and R1)
Note second pole at about 1GHz. To get these results, it was necessary to bias the input
at -1.7VDC using 3V power supplies.
If R1 is decreased to 1k results in:
Gain of 26.4dB, f-3dB = 32MHz, and GB = 630MHz
CMOS Analog Circuit Design
Page 7.2-26
M1
M2
R2
+1
+ vout -
vin-
R2
+1
VBias
x4
=1/8
x2
= 1/4
x1
=1/2
VSS
x1
=1/2
x2
= 1/4
x4
=1/8
Fig. 7.2-135A
R1 and the current mirrors are used for gain variation. R2 is fixed.
Can cascade this amplifier for higher gains
BW = BW i 21/n-1
Page 7.2-27
Page 7.2-28
Simulation Results
Output voltage swing is 1.26V for a 2.5V power supply.
Voltage gain is 0 to 60dB in 2dB steps (gain error = 0.17dB)
Maximum GB is 1.5GHz
Total current: 3.6mA
Page 7.2-29
vout
VBN
0dB
2dB
0dB
M2
VB1
M2dB
M0dB
vin
2dB
M2
M3
VB1
vin
M0dB
M2dB
Fig. 7.2-137A
P. Orsatti, F. Piazza, and Q. Huang, A 71 MHz CMOS IF-Basdband Strip for GSM, IEEE JSSC, vol. 35, No. 1, Jan. 2000, pp. 104-108.
CMOS Analog Circuit Design
P.E. Allen - 2004
Page 7.2-30
CMFB
2dB
0dB
0dB
2dB
-10dB
VBP
M5
M6
vout
M6
vout
VBN
M4
0dB M2
Load
M3
vin
M5
M4
M2
M2
M2
-10dB
Load
M3
M0dB
M0dB
M2dB
vin
M2dB
-10dB
Fig. 7.2-137A
Page 7.2-31
vin
-
+
-A1
+
- A2
|Avd1(0|) dB
vo1
+
vout
+
vo2
Fig. 7.2-14
|Avd1(s)|
|Avd2(s)|
|Avd2(0|) dB
0 dB
|p3|
|p1|
log10(f)
|p2|
GB
Comments:
Op amp will be conditionally stable
Compensation will be challenging
Page 7.2-32
gm3
gm2
-gm1
vout
gm4
Fig. 7.2-15
Comments:
All Miller capacitances must be around inverting stages
Ensure that the RHP zeros generated by the Miller compensation are canceled
Avoid pole-zero doublets which can introduce a slow time constant
R.G.H. Eschauzier and J.H.Huijsing, Frequency Compensation Techniques for Low-Power Operational Amplifiers, Kluwer Academic publishers,
1995, Chapter 6.
CMOS Analog Circuit Design
P.E. Allen - 2004
Page 7.2-33
R.G. H. Eschauzier et. al., A Programmable 1.5V CMOS Class-AB Operational Amplifier with Hybrid Nested Miller Compensation for 120dB Gain and 6MHz UGT, IEEE J.
of Solid State Circuits, vol. 29, No. 12, pp. 1497-1504, Dec. 1994.
Page 7.2-34
p3
p2
p1
Cm1
j
p4
p3 p2
p1
Cm2
j
p4
p3
p2
p1
Cm3
j
p4
p3
p2
p1
Fig. 7.2-17
Page 7.2-35
SUMMARY
Normal op amps limited by gm/C
Typical limit for CMOS op amp is GB 50MHz
Other approaches to high frequency CMOS op amps:
Current amplifiers (Transimpedance amplifiers)
Switched amplifier (simplifies the circuit reduce capacitances)
Parallel path op amps (compensation becomes more complex)
What does the future hold?
Reduction of channel lengths mean:
* Reduced capacitances Higher GBs
* Higher transconductances (larger values of K) Higher GBs
* Increased channel conductance Lower gains (more stages required)
* Reduction of power supply Increased capacitances
In otherwords, there should be some improvement in op amp GBs but it wont be
inversely proportional to the decrease in channel length. I.e. maybe GBs 100MHz
for 0.2m CMOS.
Page 7.3-1
Page 7.3-2
v1-v2
A
t
-A
A
t
-A
v2
2A
t
-2A
Fig. 7.3-1
vin
+
+
-
vout
Fig. 7.3-1A
Page 7.3-3
VSS
CM output voltage = 0
vod
VDD
VSS
vod
VDD
VSS
Fig. 7.3-2
Page 7.3-4
VDD
+
VBP
-
M8
Cc
Rz
vo1
vi1
M9
M3
M6
M4
M1
Rz
Cc
vi2
M2
M5
+
VBN
-
vo2
M7
VSS
Fig. 7.3-3
+
vid
- -
+
-
Fig. 7.3-4
CMOS Analog Circuit Design
+
vod
-
CL
+
vid
- -
+
-
+
vo2
-
+
vo1
-
2CL
2CL
P.E. Allen - 2004
Page 7.3-5
M4
M14
M5
M13
M6
M7
R1
vo2
vi1
M1
R2
M16
vi2
M2
vo1
M3
M12
M8
M9
VBias
M11
M10
M17
VSS
Fig. 7.3-5
Page 7.3-6
M3
M4
M13
M7
vo1
Cc
M6
M14
Rz Cc
Rz
vi1
M1
M9
M10
VBN
-
M2
M5
VSS
vo2
vi2
M12
M8
Fig. 7.3-6
Comments:
Able to actively source and sink output current
Output quiescent current poorly defined
Page 7.3-7
M20
M4
M5
M19
M6
M14
Rz
M15
M7
R1
M10
vo2
M13
vi1
Cc
M1
M11
vi2
M2
Rz
Cc
vo1
M18
M16
M8
M3
M17
M9
VBias
VSS
Fig. 7.3-7A
Note that the followers M11-M13 and M10-M12 are necessary for level translation to the
output stage.
Page 7.3-8
M3 M4
M5
M8
M6
M20
M21
M9
M10
R2
vo1
vi1
M1 M2
M22
vo2
R1
M16
M15
M17
M23
M18
VBias
M12
M13
M11
M14
VSS
vi2
M19
Fig. 7.3-8
Page 7.3-9
vi1
VSG3
i2
+ M1 M2 +
vGS2
vGS1
+
+
vSG3
vSG4
M3 M4
i2
i1
VGS2
vi2
VSG4
Fig. 7.3-9
Operation:
Voltage loop vi1 - vi2 = -VGS1+ vGS1 + vSG4 - VSG4 = VSG3 - vSG3 - vGS2 + VGS2
Using the notation for ac, dc, and total variables gives,
vi2 - vi1 = vid = (vsg1 + vgs4) = -(vsg3 + vgs2)
If M1 = M2 = M3 = M4, then half of the differential input is applied across each transistor
with the correct polarity.
gm1vid gm4vid
gm2vid
gm3vid
and
i2 = - 2 = - 2
i1 = 2 = 2
CMOS Analog Circuit Design
Page 7.3-10
M10
M7 M8
M9
M25
M26
M13
vi1
M1 M2
M21
vo1
M22
M19
M15
M24
vi2
R1
M14
vo2
M20
M3 M4
R2
M16
M27
M11
M17
M18
M28 +
VBias
-
M5
M6
VSS
M23
M12
Fig. 7.3-10
Quiescent output currents are defined by the current in the input cross-coupled
differential amplifier.
CMOS Analog Circuit Design
Page 7.3-11
M1
Common-mode
M2
feedback circuit
Ro1
vo1
Ro2
io2(source)
io1(source)
io1(sink)
Ro3
Ro4
vo2
io2(sink)
VSS
Model of output of differential
output op amp
Fig. 7.3-11
Operation:
M1 and M2 sense the common-mode output voltage.
If this voltage rises, the currents in M1 and M2 decrease.
This decreased current flowing through Ro3 and Ro4 cause the common-mode output
voltage to decrease with respect to VSS.
CMOS Analog Circuit Design
Page 7.3-12
M10
M7
Cc
vo1
Rz
vi1
M9
M11
M6
M3
M4
M1
+
VBN
-
M2
M5
VSS
Rz
Cc
vo2
vi2
M8
Fig. 7.3-12
Comments:
Simple
Unreferenced
Page 7.3-13
vo1
M1 M2
I5
vo2
M3 M4
Vocm
Iocm
To correction
circuitry
I6
M5
M6
VSS
Fig. 7.3-13
Operation:
1.) The desired common-mode output voltage, Vocm, creates Iocm.
2.) The actual common-mode output voltage creates current I5 which is mirrored to I6 .
3.) If M1 through M4 are matched and the current mirror is ideal, then when Iocm = I6
the actual common-mode output voltage should be equal to the desired commonmode output voltage.
4.) The above steps assume that a correction circuitry exists that changes the commonmode output voltage in the correct manner.
CMOS Analog Circuit Design
Page 7.3-14
MC1
v3
MC4
IC4
IC3
M4
I3
I4
MC2A
v1
VCM
MC2B
MC5
M1
M2
v4
Selfresistances
of M1-M4
v2
M5
MB
VSS
Fig. 7.3-13A
Page 7.3-15
M4
IBias MC3
Commonmode feedback circuit
MC1
I3
v3
MC4
IC4
IC3
v1
VCM
v4
RCM2
RCM1
MC2
I4
M1
M2
v2
M5
MC5
MB
Fig. 7.3-15New
VSS
Page 7.3-16
+
vid
- -
2
vo1
+
-
CMbias
vo2
1
Ccm
1
Ccm
Vocm
Fig. 7.3-14
Operation:
1.) During the 1 phase, both Ccm are charged to the desired value of Vocm and CMbias
= Vocm.
2.) During the 2 phase, the Ccm capacitors are connected between the differential
outputs and the CMbias node. The average value applied to the CMbias node will be
Vocm.
Page 7.3-17
SUMMARY
Advantages of differential output op amps:
- 6 dB increase in signal amplitude
- Cancellation of even harmonics
- Cancellation of common mode signals including clock feedthrough
Disadvantages of differential output op amps:
- Need for common mode output voltage stabilization
- Compensation of common mode feedback loop
- Difficult to interface with single-ended circuits
Most differential output op amps are truly balanced
For push-pull outputs, the quiescent current should be well defined
Common mode feedback schemes include,
- Unreferenced
- Referenced
Page 7.4-1
Page 7.4-2
Subthreshold Operation
Most micropower op amps use transistors in the subthreshold region.
Subthreshold characteristics:
iD
;;;;
Square Law
1A
100nA
iD
Strong Inversion
vGS =VT
Transition
Exponential
100nA
vGS VT
Weak Inversion
vGS
qv
W
GS
vDS
2V Fig. 7.4-0A
1V
VT
qID
gm = nkT
and gds ID
Operation with channel length = Lmin also will normally be in weak inversion.
Page 7.4-3
M4
Cc
vout
vin
+
M1
+
VBias
-
CL
M2
M7
M5
VSS
Fig.7.4-1
ID
GB and SR:
ID5
ID1
ID1
kT
and
SR = C = 2 C = 2GB n1 q = 2GBn1Vt
GB = (n1kT/q)C
CMOS Analog Circuit Design
Page 7.4-4
Page 7.4-5
M3
M4
M8
M6
vi2
M1
M2
vout
+
VBias
Cc
M5
M7
VSS
Fig. 7.4-2
Page 7.4-6
VDD
M4
M3
M6
VT+2VON
-
M13
M1
M2
M14
vi1
M10
vout
Cc
M5
I5
gm1+gm2
+ M11
M12 M15
+
R
Av =
2
out
VT+2VON
VBias
M9
gm1
M7
= gds6gds10 gds7gds11
Fig. 7.4-3A
VSS
+
gm10
gm11
I5
I5
2nnVt
= I 2 2 I 2 2 = 2I
7 nnVt2(nnn2+npp2)
7 n
7 p
+ I7
I7
nnV t
npV t
Can easily achieve gains greater than 80dB with power dissipation of less than 1W.
Page 7.4-7
M18
M20
i1v
i1
i2
M3
M4
i1
i2
M2
M1
A(i2-i1)
I5
M5
M22
M24
M28
M26
VSS
M19
i
vi1 2
M21
i2
A(i1-i2)
+
M25
VBiasM29
M27
-
M23
Fig. 7.4-4
Page 7.4-8
Let
L28 = A L26 and L29 = A L27
The output current available can be found by assuming that vin = vi1-vi2 > 0.
i1 + i2 = I5 + A(i2-i1)
The ratio of i2 to i1 can be expressed as
vin
i2
=
exp
nV
i1
t
Defining the output current as iOUT = b(i2-i1) and combining the above two equations
gives,
vin
bI5expnV - 1
vin
t
i
=
when
A
=
2.16
and
iOUT =
OUT
vin
nVt = 1
(1+A) - (A-1)expnVt
where b corresponds to any current gain through current mirrors (M6-M4 and M8-M3).
CMOS Analog Circuit Design
Page 7.4-9
A=2
A = 1.5
A=1
A = 0.3
A=0
1
vIN nVt
2
Fig. 7.4-5
Page 7.4-10
i1
M1
i2
M2 +
Vds2
-
Current
531A
100A
0.1Vds1(sat)
Vds1(sat)
Volts
Fig. 7.4-6
Page 7.4-11
Page 7.4-12
M8
M17
M10
M7
M9
M18
M21
M13
i1
vi1
i2
M1 M2
M29
ki1
vo1
i1
M14
M30
M27
i2
M22
vi2
ki2
M28
M3 M4
i1
i2
vo2
ki1
ki2
M25
M26
i2
M5
M23 VBias
M15
i1
M24
M11
M16
M20
M19
M12
-
M6
VSS
Fig.7.4-7
Page 7.4-13
iin
IB
kiin
M3
50/1
M5 M4
1/1
M1
1/1
1/1
M2
210/1
Fig. 7.4-7A
Page 7.4-14
SUMMARY
Operation of transistors is generally in weak inversion
Boosting techniques are needed to get output sourcing and sinking currents that are
larger than that available during quiescent operation
Be careful about using circuits at weak inversion, i.e. the self-biased cascode will cause
the resistor to be too large
Page 7.5-1
;;;;
Page 7.5-2
Introduction
VDD
Why do we need low noise op amps?
Dynamic range:
Dynamic Range = 6dBx(Number. of bits)
Signal-to-noise ratio (SNR)
Maximum RMS Signal
=
Noise + Distortion
Noise
Fig. 7.5-0B
(SNDR includes both noise and distortion)
Consider a 14 bit digital-to-analog converter with a 1V reference with a bandwidth of
1MHz.
0.5V
= 0.3535 Vrms
Maximum RMS signal is
2
A 14 bit D/A converter requires 14x6dB dynamic range or 84 dB or 16,400.
0.3535
The value of the least significant bit (LSB) = 16,400 = 21.6Vrms
If the equivalent input noise of the op amp is not less than this value, then the LSB
cannot be resolved and the D/A converter will be in error. An op amp with an equivalent
input-noise spectral density of 10nV/ Hz will have an rms noise voltage of approximately
(10nV/ Hz)(1000 Hz) = 10Vrms in a 1MHz bandwidth.
;;;;
Page 7.5-3
M1
M1
2
in1
G
M1 is
noisy S
M1 is
noiseless S
Fig. 7.5-0A
(KF)ID
2 8kTgm(1+) (KF)ID
if vBS 0
or
i
+
n =
3
3
fCoxL2
fCoxL2
gmbs
Recall that = g
m
Gate voltage model assuming common source operation:
2
8kTgm
in =
D
2
en1
M1
G
*
M1 is
noiseless S
M1 is
noisy S
2
i
8kT
KF
N
2
en = 2 = 3gm + 2fCoxWLK
gm
M1
or
Fig. 7.5-0C
8kT(1+)
KF
+
if vBS 0
WLK
2fC
3g
m
ox
en =
Page 7.5-4
Page 7.5-5
VDD
M7
M10
I5
M5
+
vin
-
2
en1
2
en2
M1
*
M1
M2
Cc
2
en8
vout
M11
+
VBias
-
M8 M9
M3
M2
2
en6
M6
M4
M3
2
en4
VBias
M6
*
M4
VSS
eto2
VSS
M7
M8
2
en3
2
M9 en9
VBias
2
en7
VSG7
Fig. 7.5-1
2
eto, is as follows where gm8(eff) 1/rds1,
2
Divide by (gm1RIgm6RII)2 to get the eq. input-noise voltage spectral density, eeq, as
2
2
eeq
eto
2en6
en8
2 gm32en3
2 gm32en3
= (gm1gm6RIRII)2 = gm12RI2 + 2en11+gm1 2 +
2 2en11+gm1 2
en1
en1 gm12rds12en1
2
where en6 = en7, en3 = en4, en1 = en2 and en8 = en9 and gm1RI is large.
CMOS Analog Circuit Design
Page 7.5-6
Comments;
2
Because we have selected PMOS input transistors, en1 has been minimized if we
choose W1L1 (W2L2) large.
Make L1<<L3 to remove the influence of the second term in the brackets.
Page 7.5-7
2
2 gm32en3
2en11+gm1 2
en1
2
2en1 1 +
KNW3L1
KPW1L3 (V2/Hz)
Comments:
The choices that reduce the 1/f noise also reduce the thermal noise.
Noise Corner:
Equating the equivalent input-noise voltage spectral density for the 1/f noise and the
thermal noise gives the noise corner, fc, as
3gmB
fc = 8kTWL
CMOS Analog Circuit Design
Page 7.5-8
Example 7.5-1 Design of A Two-Stage, Miller Op Amp for Low 1/f Noise
Use the parameters of Table 3.1-2 along with the value of KF = 4x10-28 FA for
NMOS and 0.5x10-28 FA for PMOS and design the previous op amp to minimize the 1/f
noise. Calculate the corresponding thermal noise and solve for the noise corner
frequency. From this information, estimate the rms noise in a frequency range of 1Hz to
100kHz. What is the dynamic range of this op amp if the maximum signal is a 1V peakto-peak sinusoid?
Solution
1.) The 1/f noise constants, BN and BP are calculated as follows.
KF
4x10-28FA
BN = 2CoxKN = 224.7x10-4F/m2110x10-6A2/V = 7.36x10-22 (Vm)2
and
KF
0.5x10-28FA
BP = 2CoxKP = 224.7x10-4F/m250x10-6A2/V = 2.02x10-22 (Vm)2
2.) Now select the geometry of the various transistors that influence the noise
performance.
2
To keep en1 small, let W1 = 100m and L1 = 1m. Select W3 = 100m and L3 =
20m and letW8 and L8 be the same as W1 and L1 since they little influence on the noise.
Page 7.5-9
1 + 502.02 20 =
1.1606 =
(V2/Hz)
eeq = 2x
f
f
f
21.562x10-171 +
1101001
-17
-17
2
5010020 = 3.124x10 1.33= 4.164x10 (V /Hz)
2
4.) The noise corner frequency is found by equating the two expressions for eeq to get
4.689x10-12
fc = 4.164x10-17 = 112.6kHz
This noise corner is indicative of the fact that the thermal noise is much less than the 1/f
noise.
CMOS Analog Circuit Design
Page 7.5-10
-12
4.689x10
2
df = 4.689x10-12[ln(100,000) - ln(1)]
Veq(rms) =
f
1
= 0.540x10-10 Vrms2 = 7.34 Vrms
The maximum signal in rms is 0.353V. Dividing this by 7.34V gives 48,044 or 93.6dB
which is equivalent to about 15 bits of resolution.
6.) Note that the design of the remainder of the op amp will have little influence on the
noise and is not included in this example.
Page 7.5-11
Lateral BJT
Since the 1/f noise is associated with current flowing at the surface of the channel, the
lateral BJT offers a lower 1/f noise input device because the majority of current flows
beneath the surface.
Vertical
Collector
(VDD)
n+
Lateral
Emitter Collector
;;;;;;;
Symbol.
Base
p+
n+
n+
n+
p-well
n-substrate
Base
Lateral
Collector
Emitter
Fig. 7.5-3
Comments:
Base of the BJT is the well
Two collectors-one horizontal (desired) and one vertical (undesired)
Lateral collector current
Collector efficiency is defined as Total collector current and is 60-70%
Reverse biased collector-base acts like a photodetector and is often used for lightsensing purposes
CMOS Analog Circuit Design
Page 7.5-12
Base
Gates
Emitter Lateral
Collector
;;;;
;;;;;
n+
p+
p-well
n-substrate
n+
n+
n+
Gate
Base
Emitter
Symbol.
Fig. 7.5-4
Page 7.5-13
;;
;;;;;;
;;
;;;;;;
;;;;
;;
;;;;
;;;;;;
;;;;
;;
;;
;;
;;;;;;
;;;;
;;
;
;;;;
;;
;;;;;;
;;
;;;;;;;;;;
;;
n-substrate
p-well
n-diffusion
p-diffusion
Polysilicon
90
70%
150
2.46nV/ Hz
1.92nV/ Hz
3.2Hz
3.53pA/ Hz
Fig. 7.5-7A
Metal
0.61pA/ Hz
fc(in)
162 Hz
Generally, the above structure is made as small as fT
85 MHz
possible and then paralleled with identical geomet- Early voltage
16V
ries to achieve the desired BJT.
1.2m CMOS with n-well
Vertical Base Lateral Emitter Gate
Collector
Collector
Page 7.5-14
M13M14
46.8
3.6
vi2
M15M16
58.2
7.2
M5
Q1
58.2
7.2
M7
Q2
VSS
M3 M4
480
R1 18
=34k
D1
511
3.6
1296
3.6
vi1
81.6
3.6
M10 M11
Rz = 300 Cc = 1pF vout
M12
M8 M9
M6
480
18
270
1.2
130 43.8
3.6 6.6
45.6
3.6
VSS
Experimental noise
performance:
384
1.2
Fig. 7.5-6
10
8
6
4
Voltage noise of lateral BJT at 170A
2
0
10
100
1000
Frequency (Hz)
104
105
Fig. 7.5-7
Page 7.5-15
0.211 mm2
2.5 V
2.1 mA
11.1 MHz
23.8 nV/ Hz
3.2 nV/ Hz
55 Hz
5.2 pA/ Hz
0.73 pA/ Hz
50 Hz
1.68 A
14.0 nA
1.0 mV
99.6 dB
67.6 dB
73.9 dB
39.0 V/S
42.5 V/S
P.E. Allen - 2004
Page 7.5-16
Vu(f)
Vin(f)
Clock
+1
t
-1
vu
f
VA(f)
VB(f)
VC(f)
vB
vA
vin
A1
T =1
fc
vB
vout
A2
fc
2fc
3fc
fc
2fc
3fc
fc
2fc
3fc
Fig. 7.5-8
Page 7.5-17
Chopper-Stabilized Amplifier
VDD
Chopper-stabilized Amplifier:
VDD
M3 M4
+
vin
-
M7 M8
1
2
1
2
M1 M2
2
1
M5 M6
2
1
IBias
IBias
V
Circuit equivalent during 1 phase: SS
vu1
vueq
VSS
vu2
+ A1
+ A2
v
vueq = vu1 + u2
A1
vu2
u1
vueq
+ A1
-
+ A2
+
v
u2
v
vueq = -vu1 + u2 , vueq(aver) =
A1
A1
CMOS Analog Circuit Design
+
Fig. 7.5-10
P.E. Allen - 2004
Page 7.5-18
nV/ Hz
Without chopper
With chopper
fc = 16kHz
100
10
10
20
30
Frequency (kHz)
40
50
Fig. 7.5-11
Comments:
The switches in the chopper-stabilized op amp introduce a thermal noise equal to kT/C
where k is Boltzmanns constant, T is absolute temperature and C are capacitors
charged by the switches (parasitics in the case of the chopper-stabilized amplifier).
Requires two-phase, non-overlapping clocks.
Trade-off between the lowering of 1/f noise and the introduction of the kT/C noise.
CMOS Analog Circuit Design
Page 7.5-19
SUMMARY
Primary sources of noise for CMOS circuits is thermal and 1/f
Noise analysis:
1.) Insert a noise generator for each transistor that contributes to the noise.
(Generally ignore the current source transistor of source-coupled pairs.)
2.) Find the output noise voltage across an open-circuit or output noise current into a
short circuit.
3.) Reflect the total output noise back to the input resulting in the equivalent input
noise voltage.
Noise is reduced in op amps by making the input stage gain as large as possible and
reducing the noise of this stage as much as possible.
The input stage noise can be reduced by using lateral BJTs (particularily the 1/f noise)
Doubly correlated sampling can transfer the noise at low frequencies to the clock
frequency (this technique is used to achieve low input offset voltage op amps).
Page 7.6-1
Page 7.6-2
Introduction
While low voltage op amps can be easily designed in weak inversion, strong
inversion leads to higher performance and is the focus of this section.
Semiconductor Industry Associates Roadmap for Power Supplies:
Feature Size
0.35m 0.25m 0.18m 0.13m 0.10m 0.07m
3.0V
2.5V
2.0V
Desktop Systems
1.5V
Single
Cell
Voltage
1.0V
Portable Systems
1995
1998
2001
2004
Year
2007
2010
Fig. 7.6-2
Threshold voltages will remain about 0.5 to 0.7V in order to allow the MOSFET to be
turned off.
CMOS Analog Circuit Design
Page 7.6-3
Approach
Low voltage input stages with reasonable ICMR
Low voltage bias and load circuits
Low voltage op amps
Page 7.6-4
+
VBias
-
VSD3(sat)
M3
M4
-VT1
vicm
VGS1
VDS5(sat)
M1
+
VBias
-
M2
M5
Fig. 7.6-3
Example:
If the threshold magnitudes are 0.7V, VDD = 1.5V and the saturation voltages are
0.3V, then
and
Vicm(lower) = 0.3 + 1.0 = 1.3V
Vicm(upper) = 1.5 - 0.3 + 0.7 = 1.9V
giving an ICMR of 0.6V.
Page 7.6-5
VDD
MN3
MN4
MP5
MP1
Vicm
MP2
MN2
MN1
MP4
MP3
MN5
Fig. 7.6-4
gm(eq) = gmN
gm(eq) = gmN + gmP
gm(eq) = gmP
where gm(eq) is the equivalent input transconductance of the above input stage, gmN is
the input transconductance for the n-channel input and gmP is the input transconductance for the p-channel
gm(eff)
input.
gmN+gmP
gmP
VSDP5(sat)+VGSN1
Vonp
n-channel on
p-channel off
VDD-VSDP5(sat)+VGSN1 VDD
gmN
Vicm
Fig. 7.6-5
Page 7.6-6
3:1
Ip
Page 7.6-7
vicm
MN1
MB1
Inn
MN2
Vonn
Ipp
In
Ib
Ib
Ip
vicm MP1
MB2
MP2 v
icm
Vonp
Fig. 7.6-6A
Result:
gm(eff)
gmN=gmP
Vicm
VDD Fig. 7.6-7
Vonp
Vonn
The above techniques and many similar ones are good for power supply values down to
about 1.5V. Below than, different techniques must be used or the technology must be
modified (natural devices).
CMOS Analog Circuit Design
Page 7.6-8
;;
Bulk-Driven MOSFET
A depletion device would permit large ICMR even with very small power supply voltages
because VGS is zero or negative.
When a MOSFET is driven from the bulk with the gate held constant, it acts like a
depletion transistor.
Cross-section of an n-channel
vBS
VDD
VGS
VDS
bulk-driven MOSFET:
;;
;;
;;;;;
;;
;;
;;
;;;;;;
;;
;;;;;
;;
Bulk
p+
Gate
Drain
n+
Channel
Depletion p-well
Region
Source
n+
Substrate
n+
QP
QV
n substrate
Fig. 7.6-8
Page 7.6-9
2000
Bulk-source driven
1500
1000
Saturation: VDS > VBS VP gives,
VBS = VP + VON
500
VBS2
IDSS
Gate-source
iD = IDSS 1 - VP
driven
0
Comments:
-3
-2
-1
0
1
2
3
Fig. 7.6-9
gm (bulk) > gm(gate) if VBS > 0
Gate-Source or Bulk-Source Voltage (Volts)
(forward biased )
Noise of both configurations are the same (any differences comes from the gate versus
bulk noise)
Bulk-driven MOSFET tends to be more linear at lower currents than the gate-driven
MOSFET
Very useful for generation of IDSS floating current sources.
Page 7.6-10
Page 7.6-11
Bulk-Source Current
200nA
150nA
100nA
50nA
-50nA
-0.50V -0.25V
0.00V 0.25V 0.50V
Input Common-Mode Voltage Fig. 7.6-10A
Comments:
Effective ICMR is from VSS to VDD -0.3V
The transconductance of the input stage can vary as much as 100% over the ICMR
which makes it very difficult to compensate
CMOS Analog Circuit Design
Page 7.6-12
VDD
-5
6 10
iin
iout
M1
+
+
VGS VBS
-
M2
-
+
VGS
Simple bulk-driven
current mirror
+
VGS3
+
VGS1
-
2m CMOS
iout
M4
M3
+
+
VBS3 VGS4
-M2
M1
+
+
VBS1
V
GS2
-
Cascodebulk-driven
current mirror. Fig.7.6-11
5 10-5
Iout (A)
iin
Iin=50A
4 10-5
Iin=40A
3 10-5
Iin=30A
2 10-5
Iin=20A
1 10-5
Iin=10A
0
0
0.2
0.4
0.6
Vout (V)
0.8
1
Fig. 7.6-12
The cascode current mirror gives a minimum input voltage of less than 0.5V for currents
less than 100A
CMOS Analog Circuit Design
Page 7.6-13
IBias
iin
VEB +
- Q3
iout
M2
M1
Fig. 7.6-13
Page 7.6-14
I1-IB
iin
VDD
I2
IB
IB
I1
M4
M7
iout
M7
M3
M4
or
M6
M6
M2
M1
M5
I2
IB1
iin
iout
M3
IB2
IB1
M5
M1
IB2
M2
Fig. 7.6-13A
Page 7.6-15
VDD
VDD
IPTAT
VDD
IVBE
VRef
VDD
VDD
VDD
INL
IVBE
IPTAT
VRef
VRef
VPTAT
IPTAT
INL
VBE
R2
R3
R1
Page 7.6-16
IVBE
Buss
IPTAT
Buss
M7 M8
IVBE
M3
M6
VBE
R3
-
IVBE
M5
M4
IPTAT
Q1
M9
Q2
+
R1 VPTAT R2
-
IPTAT
R4 Vout2
+
Vout1
-
Figure 7.6-15A
V PTAT
R2
Vout1 = IPTATR2 = R1 R2 = VPTAT R1
V BE
R4
Vout2 = IVBER4 = R R4 = VBE R
3
3
Page 7.6-17
1:K3
M2 M3
INL
I2
IVBE
M2 active
M3 off
M4
Current
1:K2
K3INL
M2 sat.
M3 on
K1IPTAT
K2IVBE
INL
K1IPTAT
Temperature
Illustration of the various currents.
Fig. 7.6-16
0,
K2IVBE > K1IPTAT
INL = K I
1 PTAT - K2IVBE, K2IVBE < K1IPTAT
The combination of the above concept with the previous slide yielded a curvaturecorrected bandgap reference of 0.596V with a TC of 20ppm/C from -15C to 90C using
a 1.1V power supply. In addition, the line regulation was 408 ppm/V for 1.2VDD10V
and 2000 ppm/V for 1.1VDD10V. The quiescent current was 14A.
G.A. Rincon-Mora and P.E. Allen, A 1.1-V Current-Mode and Piecewise-Linear Curvature-Corrected Bandgap Reference, J. of Solid-State
Circuits, vol. 33, no. 10, October 1998, pp. 1551-1554.
CMOS Analog Circuit Design
P.E. Allen - 2004
Page 7.6-18
M3
M15
IBias
vin
+
M1 M2
VON
-
VT+2VON
+
VT+VON
-
M12
+
VT+VON M7 M11
+
R1 VON
M6
-
M13
Cc
vout
CL
M5
M16
M8
M9
M14
M10
Fig. 7.6-17
Page 7.6-19
Page 7.6-20
VDS5(sat) = 0.0909 =
KN(W 5/L5)
L5 = 110(0.0909)2 = 44
The design of M3 and M4 is accomplished from the upper input common mode voltage:
Vicm(max) = VDD-VSD3(sat)+VTN = 2-VSD3(sat)+0.75 = 2.5V
Solving for VSD3(sat) gives 0.25V. Assume that the currents in M6 and M7 are 20A.
This gives a current of 30A in M3 and M4. Knowing the current in M3 (M4) gives
W 3 W4
230
230
L3 = L4 (0.25)250 = 19.2
VSD3(sat) 50(W3/L3)
Next, using the VSD(sat) = V ON of M3 and M4, design M10 through M12. Let us
assume that I10 = I5 = 20A which gives W10/L10 = 44. R1 is designed as R1 =
0.25V/20A = 12.5k. The W/L ratios of M11 and M12 can be expressed as
2I11
W 11 W 12
220
=
=
=
L11 L12 KPVSD11(sat)2 50(0.25)2 = 12.8
CMOS Analog Circuit Design
Page 7.6-21
Page 7.6-22
Page 7.6-23
6000/6
6000/6
M12
M8 M9
vin-
M10
2000/2
M11
vin+
vout
Cc=30pF
M1 M2
IBias
Rz=1k
Q5
Q6
M3 M4
400/2
400/2
CL
400/2
M7
Fig. 7.6-18
Page 7.6-24
Page 7.6-25
;;
;;
IBB
Reduced Threshold MOSFET
IE
Gate
p+
p+
n+
D
IBB
ICD
ICS
Source
Drain
n-well
p- substrate
Layout
Parasitic BJT
Problem:
Want to limit the BJT current to some value called, Imax.
Therefore,
Imax
IBB = CS + CD + 1
Fig. 7.6-19
T. Lehmann and M. Cassia, 1V Power Supply CMOS Cascode Amplifier, IEEE J. of Solid-State Circuits, Vol. 36, No. 7, 2001
CMOS Analog Circuit Design
P.E. Allen - 2004
Chapter 7 Section 6 (5/2/04)
Page 7.6-26
VDD
VBias1
M7
M3
IS,E
M6
Note:
ID,C = ICD + ID
IS,E = ID + IE + IR
ID,C
M8
VBias2
R
IBB
M5
M1
M2
M4
IR
VBias
-
Page 7.6-27
M12
M9
M10
M13
Cx
M1
M11
M17
M2
vout
CL
M7
VBiasN M3
M5
M4
M14
M15
VSS
M8
M16
Fig. 7.6-21
Page 7.6-28
Page 7.6-29
SUMMARY
Integrated circuit power supplies are rapidly decreasing (today 2-3Volts)
Classical analog circuit design techniques begin to deteriorate at 1.5-2 Volts
Approaches for lower voltage circuits:
- Use natural NMOS transistors (VT 0.1V)
- Drive the bulk terminal
- Forward bias the bulk
- Use depeletion devices
The dynamic range will be compressed if the noise is not also reduced
Fortunately, the threshold reduction continues to allow the techniques of this section to
be used in todays technology
Page 7.7-1
CHAPTER 7 - SUMMARY
This chapter has considered improved op amp performance in the areas of:
Op amps that can drive low output load resistances and large output capacitances
Op amps with improved bandwidth
Op amps with differential output
Op amps having low power dissipation
Op amps having low noise
Op amps that can work at low voltages
The objective of this chapter has been to show how to improve the performance of an op
amp.
We found that improvements are always possible
The key is to balance the tradeoffs against the particular performance improvement
This chapter is an excellent example of the degrees of freedom and choices that
different circuit architectures can offer.
We also illustrated further the approaches to designing op amps
The next chapter begins the transition from analog to digital with the introduction of the
comparator.
CMOS Analog Circuit Design
Page 8.0-1
CHAPTER 8 COMPARATORS
Chapter Outline
8.1 Characterization of Comparators
8.2 Two-Stage, Open-Loop Comparators
8.3 Other Open-Loop Comparators
8.4 Improving the Performance of Open-Loop Comparators
8.5 Discrete-Time Comparators
8.6 High-Speed Comparators
8.7 Summary
Page 8.1-1
Page 8.1-2
What is a Comparator?
The comparator is essentially a 1-bit analog-digital converter.
Input is analog
Output is digital
Types of comparators:
Open-loop (op amps without compensation)
Regenerative (use of positive feedback - latches)
Combination of open-loop and regenerative comparators
Page 8.1-3
vP
vN
+
-
vO
Fig. 8.1-1
Static Characteristics
Gain
Output high and low states
Input resolution
Offset
Noise
Dynamic Characteristics
Propagation delay
Slew rate
Page 8.1-4
vo
VOH
VOH
vP-vN
vP-vN
VOL
VOL
Noninverting Comparator
Inverting Comparator
Fig. 8.1-2A
Page 8.1-5
Fig. 8.1-2
Model:
vP
+
vP-vN
vN
f0(vP-vN)
+
vO
Comparator
VOH for (vP-vN) > 0
f0(vP-vN) =
VOL for (vP-vN) < 0
Fig. 8.1-3
VOH-VOL
where V is the input voltage change
V
V 0
Gain = Av = lim
CMOS Analog Circuit Design
Page 8.1-6
vP-vN
VIH
VOL
Fig. 8.1-4
+
vP-vN
vN
f1(vP-vN)
+
vO
VOH VOL
The voltage gain is Av = VIH VIL
Comparator
VOH for (vP-vN) > VIH
f1(vP-vN) = Av(vP-vN) for VIL< (vP-vN)<VIH
VOL for (vP-vN) < VIL
Fig. 8.1-5
CMOS Analog Circuit Design
Page 8.1-7
VOS
VIL
vP-vN
VIH
VOL
Fig. 8.1-6
VOH+VOL
VOS = the input voltage necessary to make the output equal
when vP = vN.
2
Model:
vP
+vP'
VOSv '-v '
P N
vN
-v '
N
f1(vP'-vN')
Comparator
+
vO
Fig. 8.1-7
Page 8.1-8
;;
VOH
Rms Noise
vP-vN
VOL
Transition Uncertainty
Fig. 8.1-8
Noise leads to an uncertainty in the transition region which causes jitter or phase noise.
Page 8.1-9
V +V
vi = IH IL
2
t
Fig. 8.1-9
Page 8.1-10
Page 8.1-11
1
-t
/
c]V
p
=
A
(0)
[1
e
t
=
ln
v
in
p
c
VOH -VOL
2
1 - 2Av(0)Vin
Define the minimum input voltage to the comparator as,
VOH -VOL
tp = c ln V (min)
Vin(min) = Av(0)
in
1- 2Vin
Define k as the ratio of the input step voltage, Vin, to the minimum input voltage, Vin(min),
Vin
2k
k = Vin(min)
tp = c ln 2k-1
Thus, if k = 1, tp = 0.693c.
vout
Illustration:
VOH
+
-
vout
VOH+VOL
2
Vin = Vin(min)
VOL
0 t t (max)
0 p p
t
Fig. 8.1-10
P.E. Allen - 2004
Page 8.1-12
Page 8.1-13
Page 8.2-1
Page 8.2-2
Two-Stage Comparator
An important category of comparators are those which use a high-gain stage to drive
their outputs between VOH and VOL for very small input voltage changes.
The two-stage op amp without compensation is an excellent implementation of a highgain, open-loop comparator.
VDD
M3
vin
+
M1
M6
vout
M2
CL
+
VBias
-
M4
M7
M5
VSS
Fig. 8.2-1
Page 8.2-3
8I7
1 - (V -V (min)-|V |)2
VOH = VDD - (VDD-VG6(min)-|VTP|)1 6 DD G6
TP
Minimum output voltage
VOL = VSS
Small-signal voltage gain
gm1 gm6
Av(0) = gds2+gds4gds6+gds7
Poles
Input:
Output:
-(gds2+gds4)
-(gds6+gds7)
p
=
p1 =
2
CI
CII
Frequency response
Av(0)
Av(s) = s
s
p - 1 p - 1
1
2
Page 8.2-4
8234x10-6
= 2.2V
1 - 50x10-6
2
38(2.5-0-0.7)
The value of VOL is -2.5V. The gain was evaluated in Ex. 6.3-1 as Av(0) = 7696.
Therefore, the input resolution is
VOH-VOL 4.7V
Vin(min) = Av(0) = 7696 = 0.611mV
Next, we find the poles of the comparator, p1 and p2. From Ex. 6.3-1 we find that
gds2 + gds4
15x10-6(0.04+0.05)
p1 = =
= -6.75x106 (1.074MHz)
CI
0.2x10-12
and
gds6 + gds7
95x10-6(0.04+0.05)
=
= -1.71x106 (0.272MHz)
p2 = CII
5x10-12
CMOS Analog Circuit Design
Page 8.2-5
p2etp1 p1etp2
vout(t) = Av(0)Vin1 + p -p - p -p
1 2
1 2
Normalizing gives,
p2
vout(t)
m
1
vout(tn ) = A (0)V = 1 - m-1e-tn + m-1e-mtn where m = p 1 and
v
in
1
t
p
t
p
-t
1
1
If p1 = p2 (m =1), then
vout(tn) = 1 - e + tp1e = 1 - e n - tne-tn
tn = -tp1
m=4
0.8
m=2
m = 1 m = 0.5
m = 0.25
0.6
0.4
p2
m= p
1
0.2
0
2
4
6
Normalized Time (tn = -tp1 )
10
Fig. 8.2-2
Page 8.2-6
ln(m)
exp
- exp -m
=
dtn
m-1 m-1
m-1
For the two-stage comparator using NMOS input transistors, the slew rate is
I7
SR- = CII
SR+
I6-I7 0.56(VDD-VG6(min)-|VTP|)2 - I7
= CII =
CII
Page 8.2-7
Page 8.2-8
m
1
vout(tn) = Av(0)Vin 1 - m-1e-tn + m-1e-mtn
cant be easily solved so approximate the step response as a power series to get
tn2
m2tn2 mtn2Av(0)Vin
m
1
2
2
or
Vin(min)
1
=
mVin
mk
This approximation is particularly good for large values of k.
tpn
VOH+VOL
mAv(0)Vin =
Page 8.2-9
Page 8.2-10
i3
M3
i1
vG1
M1
M4 vo1
M6
CI
i2
M2
vG2
vout
Page 8.2-11
7ISS
1
56(VDD-vo1-|VTP|)2
4.) Assume vG2 = VREF and vG1 << VREF, therefore i2 = ISS and i1 = 0.
Same as in 3.) but now as vo1 approaches vS2 with ISS/2 flowing, the value of vGS2
becomes larger and M5 becomes active and ISS decreases. In the limit, ISS 0,vDS2 0
and vDS5 0 resulting in
vo1 VSS and
vout = VDD - (VDD-VSS
7ISS
|VTP|)1 1
56(VDD-VSS-|VTP|)2
CMOS Analog Circuit Design
Page 8.2-12
7ISS
vout = VDD - (VDD-vo1-|VTP|)1 1
56(VDD-vo1-|VTP|)2
6.) Assume that vG1 = VREF and vG2 >> VREF. When the source voltage of M1 or M2
causes M5 to be active, then ISS decreases and
7ISS
vo1 VSS and vout = VDD - (VDD-VSS-|VTP|)1 1
56(VDD -VSS-|VTP|)2
7.) Assume vG1 = VREF and vG2 < VREF and i1 <ISS and i2 > 0. Consequently, i4>i2
which causes vo1 to increase. When M4 becomes active i4 decreases until i2 = i4 at
which vo1 stabilizes at (M6 will be off under these conditions and vout VSS).
vG2 < VREF, i1 < ISS and i2 > 0
VDD - VSD4(sat) < vo1 < VDD,
CMOS Analog Circuit Design
Page 8.2-13
VDD
VSS
VSS
VDD
VSS
VSS
Page 8.2-14
VDD
M6
i6
+
vin
-
vout
i7
M7
VSS
Fig. 8.2-4
KN(W7/L7)
KP(W6/L6) (VBias- VSS -VTN)
Example:
If W7/L7 = W6/L6, VDD = 2.5V, VSS = -2.5V, and VBias = 0V the trip point for the
circuit above is
VTRP = 2.5 - 0.7 - 110/50 (0 +2.5 -0.7) = -0.870V
Page 8.2-15
Page 8.2-16
vG2
2.5V
0V
0.2
-2.5V
0.4
0.6
t(s)
M8
4.5m
M5 1m
VSS = -2.5V
35m
1m
M7
Fig. 8.2-5A
Fig. 8.2-5
Solution
1.) Total delay = sum of the first and second stage delays, t1 and t2
2.) First, consider the change of vG2 from -2.5V to 2.5V at 0.2s.
The last row of Table 8.2-1 gives vo1 = +2.5V and vout = -2.5V
3.) tf1, requires CI, Vo1, and I5. CI = 0.2pF, I5 = 30A and V1 can be calculated by
finding the trip point of the output stage/
CMOS Analog Circuit Design
Page 8.2-17
2342
2 = 234A V
(V
-|V
|)
=
0.7
+
SG6
5038 = 1.196V
2 SG6 TP
Therefore, the trip point of the second stage is VTRP2 = 2.5 - 1.196 = 1.304V
Therefore, V1 = 2.5V - 1.304V = VSG6 = 1.196V. Thus the falling propagation time
delay of the first stage is
1.196V
tfo1 = 0.2pF 30A = 8 ns
5.) The rising propagation time delay of the second stage requires CII, Vout, and I6. CII
is given as 5pF, Vout = 2.5V (assuming the trip point of the circuit connected to the
output of the comparator is 0V), and I6 can be found as follows:
VG6(guess) 0.5[VG6(I6=234A) + VG6(min)]
215
VG6(min) = VG1 - VGS1(ISS/2) + VDS2 -VGS1(ISS/2) = -0.7 - 1103 = -1.00V
VG6(guess) 0.5(1.304V-1.00V) = 0.152V
6
3850
Therefore VSG6 = 2.348V and I6 = 2 (VSG6-|VTP|)2 = 2 (2.348 - 0.7)2 = 2,580A
CMOS Analog Circuit Design
Page 8.2-18
2.5V
trout = 5pF 2,580A-234A = 5.3 ns
Thus the total propagation time delay of the rising output of the comparator is
approximately 13.3 ns and most of this delay is attributable to the first stage.
7.) Next consider the change of vG2 from 2.5V to -2.5V which occurs at 0.4s. We shall
assume that vG2 has been at 2.5V long enough for the conditions of Table 8.2-1 to be
valid. Therefore, vo1 VSS = -2.5V and vout VDD. The propagation time delays for the
first and second stages are calculated as
3V
vout
1.304V-(-1.00V)
= 15.4 ns 2V
tro1 = 0.2pF
30A
V
= 1.304V
TRP6
2.5V
tfout = 5pF 234A = 53.42ns
8.) The total propagation time delay of the
falling output is 68.82 ns. Taking the
average of the rising and falling propagation
time delays gives a propagation time delay
for this two-stage, open-loop comparator of
about 41.06ns.
1V
0V
-1V
-2V
vo1
Rising prop.
delay time
-3V
200ns
300ns
Falling prop.
delay time
400ns
Time
500ns
600ns
Fig. 8.2-6
Page 8.2-19
Design Relationships
|pII|CII
1
|pI| = |pII| =
,
and I7 = I6 = +
N P
tp mk
W6
2I6
W7
2I7
and
=
L6 = K (V
L
2
7 KN(VDS7(sat))2
P SD6(sat))
2C I
Guess CI as 0.1pF to 0.5pF
I5 = I7 C
II
I5
W3 W 4
=
=
L3 L4 K (V
2
P SG3-|VTP|)
Av(0)(gds2+gds4)(gds6+gds7) W 1 W 2 gm12
L1 = L2 = KNI5
gm6
Find CI and check assumption
gm1 =
CI = Cgd2+Cgd4+Cgs6+Cbd2+Cbd4
2I5
W5
VDS5(sat) = Vicm--VGS1-VSS L =
5 KN(VDS5(sat))2
Comments
Choose m = 1
VSD6(sat) = VDD-VOH
VDS7(sat) = VOL - VSS
A result of choosing m = 1.
Will check CI later
VSG3 = VDD-Vicm++VTN
gm6 =
2KPW6I6
VOH-VOL
A
(0)
=
v
L6
Vin(min)
Page 8.2-20
Page 8.2-21
W3 W 4
L3 = L4 = 4
W1 W 2
L1 = L2 = 6
Page 8.2-22
10A 2/1
M10
M8
2/1
M11
10A
M5
M9
2/1
40A
40A
8/1
VSS
CMOS Analog Circuit Design
3/1
M12
M7
400A
29/1
Fig. 8.2-7
P.E. Allen - 2004
Page 8.2-23
Design Relationships
dvout CII(VOH-VOL)
I7 = I6 = CII dt =
tp
Comments
Assume the trip point of the output is (VOHVOL)/2. Let tp1 = tp2 = 0.5tp
W6
2I6
W7
2I7
and L =
L6 = K (V
2
7 KN(VDS7(sat))2
P SD6(sat))
Guess CI as 0.1pF to 0.5pF
VSD6(sat) = VDD-VOH
Typically 0.1pf<CI<0.5pF
dvo1 CI(VOH-VOL)
I5 = CI dt
tp
W3 W 4
I5
L3 = L4 = K (V
2
P SG3-|VTP|)
VSG3 = VDD-Vicm++VTN
Av(0)(gds2+gds4)(gds6+gds7) W1 W 2 gm12
L1 = L2 = KNI5
gm6
Find CI and check assumption
CI = Cgd2+Cgd4+Cgs6+Cbd2+Cbd4
VOH-VOL
Av(0) = V (min)
in
If CI is greater than the guess in step 3, increase
the value of CI and repeat steps 4 through 6
W5
2I5
VDS5(sat) = Vicm--VGS1-VSS L =
5 KN(VDS5(sat))2
1
2
gm1 =
7
8
gm6 =
and
2KPW6I6
L6
Page 8.2-24
L3 L4 50(1.2-0.7)2
L3 = L4 = 2
CMOS Analog Circuit Design
Page 8.2-25
W 3 W 4 (81)2
L3 = L4 = 11040 = 1.49
W1 W 2
L1 = L2 = 2
Page 8.2-26
SUMMARY
The two-stage, open-loop comparator has two poles which should as large as possible
The transient response of a two-stage, open-loop comparator will be limited by either
the bandwidth or the slew rate
It is important to know the initial states of a two-stage, open-loop comparator when
finding the propagation delay time
If the comparator is gainbandwidth limited then the poles should be as large as possible
for minimum propagation delay time
If the comparator is slew rate limited, then the current sinking and sourcing ability
should be as large as possible
Page 8.3-1
Page 8.3-2
Push-Pull Comparators
Clamped:
VDD
M6
M4
M8
M3
vin
+
M1
vout
M2
CL
M5
+
VBias
-
M9
VSS
M7
Fig. 8.3-1
Comments:
Gain reduced Larger input resolution
Push-pull output Higher slew rates
Page 8.3-3
M4
M15
M8
M3
vin
+
M1
M2
M14
M7
R2
R1
M9
vout
M12
M10
M5
+
VBias
-
CII
M11
M13
VSS
Fig. 8.3-2
Comments:
Can also use the folded cascode architecture
Cascode output stage result in a slow linear response (dominant pole is small)
Poorer noise performance
CMOS Analog Circuit Design
Page 8.3-4
M8
M3
M10
M4
M6
vin
+
M1
vout
M2
CII
+
VBias
-
M5
M7
VSS
M9
M11
Fig. 8.3-3
Comments:
Slew rate = 3V/s into 50pF
Linear rise/fall time = 100ns into 50pF
Propagation delay time 1s
Loop gain 32,000 V/V
CMOS Analog Circuit Design
Page 8.3-5
VDD
M6
M6
M4
M3
M4
M3
vin+
vin-
vout
vin+
M1
M1
Extremely
large sourcing
current
vinM2
M2
M5
VBias
M5
VSS
VSS
Fig. 8.3-4
Advantage:
Large sink or source current with out a large quiescent current.
Disadvantage:
Poor common mode range (vin+ slower than vin-)
M. Bazes, Two Novel Full Complementary Self-Biased CMOS Differential Amplifiers, IEEE Journal of Solid-State Circuits, Vol. 26, No. 2, Feb.
1991, pp. 165-168.
Page 8.4-1
Page 8.4-2
Autozeroing Techniques
Use the comparator as an op amp to sample the dc input offset voltage and cancel the
offset during operation.
Ideal
Comparator
Ideal
Comparator
vIN
VOS
Ideal
Comparator
VOS
vOUT
+ VOS
VOS
VOS
+
-C
CAZ
AZ
Model of Comparator.
Autozero Cycle
Comparison Cycle
Fig. 8.4-1
Comments:
The comparator must be stable in the unity-gain mode (self-compensating comparators
are good, the two-stage op comparator would require compensation to be switched in
during the autozero cycle.)
Complete offset cancellation is limited by charge injection
Page 8.4-3
vIN-
Comparator
vIN+
2
vOUT
VOS
CAZ 1
+
VOS
-
vOUT = VOS
VOS
Comparator during 1 phase
vIN
vOUT
+
vIN+ + VOS VOS
Comparator during 2 phase
Fig. 8.4-2
Page 8.4-4
1 CAZ
- 1
+
vOUT
1
Fig. 8.4-3
Inverting:
CAZ
vIN
2
vOUT
- 1
+
Fig. 8.4-4
Comment on autozeroing:
Need to be careful about noise that gets sampled onto the autozeroing capacitor and is
present on the comparison phase of the process.
CMOS Analog Circuit Design
Page 8.4-5
vin
t
VOH
vout
t
VOL
Fig. 8.4-6A
VTRP+
VTRPVOH
vout
t
VOL
CMOS Analog Circuit Design
Fig. 8.4-6B
P.E. Allen - 2004
Page 8.4-6
vOUT
VOH
R1 (V -V )
R2 OH OL
0
0
VTRP+
vIN
VTRP-
VOH
VTRP+
vIN
VTRP-
VOL
VOL
Counterclockwise Bistable
Clockwise Bistable
Fig. 8.4-5
Page 8.4-7
OH
R2
vIN
R1
vOUT
Fig. 8.4-7
-R1VOH
R2
R1 (V -V )
R2 OH OL
0
0
R V
- 1 OL
R2
vIN
VOL
V
TRP
R2 VOL
2
2
1
1
Lower Trip Point:
Assume that vOUT = VOH, the lower trip point occurs when,
R1
R2
R1
VTRP- = - R2 VOH
0 = R1+R2VOH + R1+R2VTRP
Page 8.4-8
vOUT
R1
R2
VOH
R1 (V -V )
R1+R2 OH OL
0
0
R1VOL
R1+R2
VOL
vIN
R1VOH
R1+R2
Fig. 8.4-8
2
1
+
Vin = VTRP -VTRP = R +R VOH -VOL
1
2
2
1
Page 8.4-9
R2
R1
vIN
vOUT
R1+R2
R2 VREF
0
0
VREF
Fig. 8.4-9
R1 (V -V )
R2 OH OL
R1VOH
R2
VOL
vIN
R1|VOL|
R2
R1+R2
R1
VTRP+ = R VREF - R VOL
2
2
R1+R2
R1
VTRP- = R2 VREF - R2 VOH
Shifting Factor:
R1+R2
V
R
2 REF
Page 8.4-10
vOUT
vOUT
R1
R2
VOH
R1 (V -V )
R1+R2 OH OL
R2
R1+R2VREF
0
0
VREF
VOL
R1|VOL|
R1+R2
vIN
R1VOH
R1+R2
Fig. 8.4-10
Shifting Factor:
R
2
R +R VREF
2
1
CMOS Analog Circuit Design
Page 8.4-11
and
R1
R2
0 = R1+R2 (-2) + R1+R2VREF
Page 8.4-12
IBias
M6
M4
M7
vo1
vi1
vo2
M2
M1
M8
vi2
M5
Fig. 8.4-11
VSS
Page 8.4-13
VDD
vo1
vo2 is high.
M3 M6
M1
M7 M4
vo2
M2
i1 = i3
i2 = i6
W6/L6
M6 would like to source the current i6 = W L i1
vin
M5
3 3
I5
As vin begins to increase towards the trip point, the
current flow through M2 increases. When i2 = i6,
Fig. 8.4-12A
VSS
the upper trip point will occur.
W 6 /L 6
W6/L6
i5
Also, i2 = i5 - i1 = i5 - i3
Knowing i1 and i2 allows the calculation of vGS1 and vGS2 which gives
2i2
2 + VT2 -
2i1
1 - VT1
P.E. Allen - 2004
Page 8.4-14
W7/L7
i5
i2 = i4 = 1 + [(W /L )/(W /L )]
7 7
4 4
Also, i1 = i5 - i2 = i5 - i4
Knowing i1 and i2 allows the calculation of vGS1 and vGS2 which gives
VTRP- = vGS2 - vGS1 =
2i2
2 + VT2 -
2i1
1 - VT1
Page 8.4-15
Page 8.4-16
Page 8.4-17
IBias
M6
M4
M7
M9
M8
vi1
M2
M1
M10
M8
vout
M11
M5
VSS
vi2
Fig. 8.4-14
Page 8.4-18
Schmitt Trigger
The Schmitt trigger is a circuit that has better defined switching points.
Consider the following circuit:
How does this circuit work?
VDD
Assume the input voltage, vin, is low and the output
M5
voltage, vout , is high.
M3, M4 and M5 are on and M1, M2 and M6 are off.
When vin is increased from zero, M2 starts to turn on causing
M4
M3
vout
vin
M3 to start turning off. Positive feedback causes M2 to turn
on further and eventually both M1 and M2 are on and the
M6
output is at zero.
M2
The upper switching point, VTRP+ is found as follows:
When vin is low, the voltage at the source of M2 (M3) is
M1
vS2 = VDD-VTN3
Fig. 8.4-15
VTRP+ occurs when the input voltage causes the currents in M3 and M1 to be equal.
Page 8.4-19
VTRP- =
1 + 5/6
vout
VDD
0 0
VTRP-
VTRP+ VDD
vin
Fig. 8.4-16
CMOS Analog Circuit Design
Page 8.4-20
SUMMARY
Open-loop, continuous-time comparators can be improved in the areas of:
- Current sinking and sourcing
- Removal of offset voltages
- Removal of the influence of a noisy signal through hysteresis
Comparators with hysteresis (positive feedback)
- External
- Internal
Page 8.5-1
Page 8.5-2
VC -
+
C
Cp
+
VOS
V1 - VOS
+ -
Vout
A
V2
Cp
VOS
-
Vout
VOS
-
Fig. 8.5-1
1 Phase:
The V1 input is sampled and the dc input offset voltage is autozeroed.
VCp(1) = VOS
VC(1) = V1 - VOS and
2 Phase:
V 2C
(V1-VOS)C VOSCp
Vout(2) =-A C+Cp - C+Cp + C+Cp + AVOS
C
Cp
C
C
= -A (V2-V1) C+C + VOSC+C + C+C + AVOS = -A(V2-V1) C+C A(V1-V2)
p
p
p
p
if Cp is smaller than C.
CMOS Analog Circuit Design
Page 8.5-3
1
- +
2
1
+-
vin
2
1
vout
-
1
Fig. 8.5-2
Comments:
Reduces the influence of charge injection
Eliminates even harmonics
Page 8.5-4
Regenerative Comparators
Regenerative comparators use positive feedback to accomplish the comparison of two
signals. Latches have a faster switching speed that the previous bistable comparators.
NMOS and PMOS latch:
VDD
VDD
I1
I2
vo1
vo2
vo1
vo2
I1
M2
M1
M2
M1
NMOS latch
I2
PMOS latch
Fig. 8.5-3
Page 8.5-5
VDD
I1
VDD
I2
vo2
vo1
M1
M2
C1
Vo2
V '
gm1Vo2 R1 so1
-
C2
Vo1
Vo2
V '
gm2Vo1 R2 so2
Fig. 8.5-4
Vo1
gm1Vo2+G1Vo1+sC Vo1- s = gm1Vo2+G1Vo1+sC1V o1-C1Vo1 = 0
Vo2
gm2Vo1+G2Vo2+sC Vo2- s = gm2Vo1+G2Vo2+sC2V o2-C2Vo2 = 0
Page 8.5-6
= 1-gmR
Taking the inverse Laplace transform gives
if gmR >>1.
C
L = || gmR = gm =
= 0.67Cox
2K(W/L)I
if C Cgs.
Vout(t) = et/L Vi
WL 3
2KI
Page 8.5-7
0.5
0.4
0.3
0.8
Vout
VOH-VOL
0.2
0.1
0.05
0.6
Vi
VOH-VOL
0.03
0.01
0.4
0.005
0.2
0
0
t
L
5
Fig. 8.5-5
VOH- VOL
Page 8.5-8
Page 8.5-9
W1
W2
1
=
K
L (vin+ - VT) + L (VREF- - VT)
VREFN
R1
and
W
W2
1
1
+
R2 = KN L (vin - VT) + L (VREF - VT)
3.) The input voltage which causes R1 and R2 to be equal is given by
vin(threshold) = (W2/W1)VREF
W2/W1 = 1/4 generates a threshold of 0.25VREF.
Performance 20Ms/s & 200W
VDD
1
M10 Latch
/Reset
1
M6
M8
M4
voutM2
R2
M1
vin-
VREF+
Fig. 8.5-6
T.B. Cho and P.R. Gray, A 10b, 20Msamples/s, 35mW pipeline A/D Converter, IEEE J. Solid-State Circuits, vol. 30, no. 3, pp. 166-172, March
1995.
Page 8.5-10
M10
M5
M6
vout+
voutM4
M3
vin+
M8
M7
M1
vin-
M2
Fig. 8.5-7
A. Coban, 1.5V, 1mW, 98-dB Delta-Sigma ADC, Ph.D. dissertation, School of ECE, Georgia Tech, Atlanta, GA 30332-0250.
Page 8.5-11
Dynamic Latch
Circuit:
VDD
Latch
M8
VREF
M6
M4 +
vout
M3
vin
vout-
M7
M1
Latch
M2
M5
Fig. 8.5-8
Number
of Samples
;;
;;;
;;;
;;
;;;
;;;
= 5.65
-15
L = 1.2m
(0.6m Process)
-10 -5
0
5
10
Input offset voltage (mV)
15
Fig. 8.5-9
Page 8.5-12
SUMMARY
Discrete-time comparators must work with clocks
Switched capacitor comparators use op amps to transfer charge and autozero
Regenerative comparators (latches) use positive feedback
The propagation delay of the regenerative comparator is slow at the beginning and
speeds up rapidly as time increases
The highest speed comparators will use a combination of open-loop comparators and
latches
Page 8.6-1
Page 8.6-2
A0
sT+1
A0
sT+1
A0
sT+1
A0
sT+1
A0
sT+1
Linear
small
signal
Linear
small
signal
Linear
& large
signal
Large
signal
small C
Large
signal
bigger C
Large
signal
big C
Fig. 8.6-1
Page 8.6-3
Page 8.6-4
Cv1
Reset
C1
Cv3
+ -+
Cv2
C2
FB
FB
Cv5
+ -
+ -+
-+
Reset
Reset
Cv4
FB
+
Latch vout
-
Reset
Cv6
FB
FB
Clock
+ vin -
Fig. 8.6-3
Comments:
Autozero and reset phase followed by comparison phase
More switches are needed to accomplish the reset and autozero of all preamplifiers
simultaneously
Can run as high as 100Msps
Page 8.6-5
VDD
KN(W1/L1)
Kp(W3/L3)
Dominant Pole:
gm3 gm4
|pdominant| = C = C
where C is the capacitance seen from the
output nodes to ground.
M3
FB
M4
Q
Reset
Q
FB
M1
M5
M6
M2
Latch
Enable
If (W1/L1)/(W3/L3) = 100 and the
bias current is 100A, then A = -3.85
Latch
Preamplifier
and the bandwidth is 15.9MHz if C =
0.5pF.
VBias
Comments:
Fig. 8.6-4
If a buffer is used to reduce the output
capacitance, one must take into account the loss of the buffer.
The use of a preamplifier before the latch reduces the latch offset by the gain of the
preamplifier so that the offset is due to the preamplifier only.
CMOS Analog Circuit Design
Page 8.6-6
An Improved Preamplifier
Circuit:
VDD
VBiasP
vout- M5
M3
VBiasP
M4
M6
vout+
Reset
M12
M10
FB M11
FB
M8
M7
VBias
vin+
vin-
M2
M1
VBiasN
M9
Fig. 8.6-5
Gain:
KN(W1/L1)I1
KN(W1/L1)
=
KP(W3/L3)
KP(W3/L3)I3
If I5 = 24I3, the gain is increased by a factor of 5
gm1
Av = - gm3 = -
I5
1+I3
Page 8.6-7
vin=VREF
vin=VREF VPR
VREF-VT+V
S2
vin-VT
S1
CT
CO
+
vout
-
CT
vin = VREF+V
CO
+
vout
=VPR
-
Precharge phase.
CT
CO
+
vout =VPR -CT V
CO
Amplification phase.
Fig. 8.6-6
Comments:
Only positive values of voltage will be amplified.
Large offset voltages result as a function of the subthreshold current.
Page 8.6-8
VDD
CT
VPR
S1
M2
S3
S2
vout
vin
CT
S3
S1
M1
CO
Fig. 8.6-7
Comments:
NMOS and PMOS allow both polarities of input
CMOS switches along with dummy switches reduce the charge injection
Switch S3 prevents the subthreshold current influence
Used as a preamplifier in a comparator with 8-bit resolution at 20Msps and a power
dissipation of less than 5W
Page 8.6-9
A High-Speed Comparator
Circuit:
VDD
Self-biased
diff amp Output
Driver
Preamp
vout
vin+
vin-
IBias
Latch
Fig. 8.6-8
Comments:
Designed to have a tp = 10ns with a 5pF load and a 10mV overdrive
Not synchronous
Comparator gain is greater than 2000V/V and the quiescent current was 100A
CMOS Analog Circuit Design
Page 8.7-1
CHAPTER 8 - SUMMARY
Types of Comparators Presented
High-gain, open-loop
Improved high-gain, open-loop, comparators
Hysteresis
Autozeroing
Regenerative comparators
Discrete-time comparators
Performance Characterization
Propagation delay time
Binary output swing
Input resolution and/or gain
Input offset voltage
Power dissipation
Important Principles
The speed of the comparator depends on the linear and slewing responses
The dc input offset voltage depends on the matching and is reduced by autozeroing.
Charge injection is the limit of autozeroing
The comparator gain should be large enough for a binary output when vin = Vin(min)
Cascaded comparators, the first stages should large GB and the last stages high SR
CMOS Analog Circuit Design
5/2/04
Page 9.0-1
9.0 - INTRODUCTION
Organization
;;;
;;;
;;;
;;;;
;;;
;;;;;;;;;;
;;;
;;;
;;;;;;
Chapter 10
D/A and A/D
Converters
Chapter 9
Switched Capacitor Circuits
Systems
Chapter 6
Simple CMOS &
BiCMOS OTA's
Chapter 7
High Performance
OTA's
Chapter 8
CMOS/BiCMOS
Comparators
Complex
Simple
Chapter 4
CMOS/BiCMOS
Subcircuits
Chapter 5
CMOS/BiCMOS
Amplifiers
;;;
;;;;
;;;
;;;;;;;;;;
Circuits
Chapter10
1
Chapter
Introduction
D/ to Analog CMOS Design
Devices
Chapter
Chapter11
2
Analog
CMOS
Technology
Systems
Chapter 3
CMOS
Modeling
Page 9.0-2
Page 9.1-1
vC (t)
i1(t)
i2 (t)
v2 (t)
v1(t)
i2 (t)
v2 (t)
Fig 9.1-01
1
t
0
2
1
0
0
T/2
3T/2 2T
Fig. 9.1-02
CMOS Analog Circuit Design
Page 9.1-2
1
i1(average) = T i1(t)dt = T i1(t)dt
2
0
0
Charge and current are related as,
v (t)
v1(t)
vC (t)
C 2
dq1(t)
i1(t) = dt
Fig. 9.1-03
Substituting this in the above gives,
1 T/2
q1(T/2)-q1(0) CvC(T/2)-CvC(0)
i1(average) = T dq1(t) =
=
T
T
0
T
T
For the continuous time circuit:
v1(t)
v2 (t)
V1-V2
T
i1(average) = R
RC
Fig. 9.1-04
For v1(t) V1 and v2(t) V2, the signal frequency must be much less than fc.
i1(average) =
Page 9.1-3
Page 9.1-4
i1(t)
i2 (t)
v1(t)
v2 (t)
Fig 9.1-01
(V1-V2)2 T
Power = TRon
e -t/(RonC)dt
(V1-V2)2
= (T/C)
-e -T /(RonC) + 1
(V1-V2)2
(T/C) if T >> RonC
Thus, if R = T/C, then the power dissipation is identical in the continuous time and
discrete time realizations.
CMOS Analog Circuit Design
Page 9.1-5
S1
S2
C
vC (t)
Series
i1(t)
i2 (t)
v2 (t)
C1
v1(t)
S1
vC1 (t)
i2 (t)
i1(t)
S2
v2 (t)
vC2(t)
C2
Series-Parallel
v1(t)
S1 C S2 i2 (t)
vC (t)
S1
S2
v2 (t)
Bilinear
Fig. 9.1-05
Series-Parallel:
The current, i1(t), that flows during both the 1 and 2 clocks is:
T
q1(T/2)-q1(0) q1(T)-q1(T/2)
1T
1 T/2
+
i1(average) = T i1(t)dt = T i1(t)dt + i1(t)dt =
T
T
0
0
T/2
Therefore, i1(average) can be written as,
C2 [vC2(T/2)-vC2(0)] C1 [vC1(T)-vC1(T/2)]
+
i1(average) =
T
T
The sequence of switches cause,vC2(0)=V2, vC2(T/2)=V1, vC1(T/2)=0, and vC1(T)= V1-V2.
Applying these results gives
C2[V1-V2] C1[V1-V2- 0] (C1+C2)(V1-V2)
+
=
i1(average) =
T
T
T
T
Equating the average current to the continuous time circuit gives: R = C1 + C2
CMOS Analog Circuit Design
Page 9.1-6
Page 9.1-7
Parallel
v1(t) C
v2 (t)
Series
v1(t)
C
1
Series-Parallel
v1(t)
T
C
v2 (t)
T
C1 + C2
C1
C2
2
C
v1(t)
v2 (t)
Bilinear
T
C
v2 (t)
T
4C
Page 9.1-8
R1
v2
v1
C2
The transfer function of this simple circuit is,
V2(j)
1
1
Fig. 9.1-06
H(j) = V (j) = jR C + 1 = j + 1
1
1 2
1
where 1 = R1C2 is the time constant of the circuit and determines the accuracy.
Continuous Time Accuracy
Let 1 = C. The accuracy of C can be expressed as,
dC dR1 dC2
C = R1 + C2 5% to 20% depending on the size of the components
Discrete Time Accuracy
T
1
Let 1 = D = C1 C2 = fcC1 C2. The accuracy of D can be expressed as,
dD dC2 dC1 dfc
D = C2 - C1 - fc 0.1% to 1% depending on the size of components
The above is the primary reason for the success of switched capacitor circuits in CMOS
technology.
CMOS Analog Circuit Design
Page 9.1-9
v(t)
A sampled-data
voltage waveform
for a two-phase
clock.
1
t/T
vO(t)
v(t)
A sampled-data
voltage waveform
for the odd-phase
clock.
1
A sampled-data v (t)
voltage waveform
for the even-phase
clock.
t/T
v(t)
Fig. 9.1-065
t/T
P.E. Allen - 2004
Page 9.1-10
z-domain Relationships:
Consider the one-sided z-transform of a sequence, v(nT), defined as
Page 9.1-11
Vi (z) =
o
Vi (z)
Switched
Capacitor
Circuit
+ Vi (z)
Fig. 9.1-07
V o (z)
H ij (z) = i
V i(z)
e
where i and j can be either e or o. For example, Hoe(z) represents Vo (z)/ V i (z) .
Also, a transfer function, H(z) can be defined as
e
Page 9.1-12
Page 9.1-13
v1
C1
C2
v2
1
1
2
2
2
t
n- 23 n-1 n- 21 n n+ 21 n+1 T
Clock phasing for this example.
Fig. 9.1-08
Solution
1: (n-1)T< t < (n-0.5)T
Equivalent circuit:
C2
v1o(n-1)T C1
C2
v2e(n- 23 )T v2o(n-1)T
Equivalent circuit.
v1o(n-1)T C1
v2e(n- 23 )T v2o(n-1)T
Fig. 9.1-09
(1)
P.E. Allen - 2004
Page 9.1-14
C1
e
v1(n-1/2)T
C2
C1 vo(n-1)T
1
v2e(n- 21 )T
v2o(n-1)T
Fig. 9.1-10
The output of this circuit can be expressed as the superposition of two voltage
sources, vo1 (n-1)T and vo2 (n-1)T given as
C1
C2
ve2 (n-1/2)T = C1+C2 vo1 (n-1)T + C1+C2 vo2 (n-1)T.
(2)
(3)
(4)
Page 9.1-15
o
z
V2(z)
z-1
C2
C1+C2
oo
-1 , where =
H (z) = o =
=
(7)
C2
1 + - z
C1 .
V 1(z)
-1
1 - z C1+C2
Page 9.1-16
Discrete
time frequency
response
=0
-1
Imaginary Axis
+j1
r=1
=
= -
=0
+1 Real
Axis
= -
Continuous Frequency Domain
-j1
Discrete Frequency Domain
Fig. 9.1-11
Page 9.1-17
(3)
ArgHoo = - tan-1(1+)cos(T)- = - tan-1
cos(T) - 1+
CMOS Analog Circuit Design
Page 9.1-18
Page 9.1-19
Page 9.1-20
0.8
0.707
0.6
Magnitude
oo jT
|H (e
)|
0.4
0.2
0
|H(j)|
= 1/1
0
0.2
0.4
/c
0.6
0.8
50
oo jT
Arg[H (e
= 1/1
0
-50
-100
)]
Arg[H(j)]
0
0.2
0.4
/c
0.6
0.8
Fig. 9.1-12
Page 9.1-21
SUMMARY
Resistance emulation is the replacement of continuous time resistors with switched
capacitor approximations
- Parallel switched capacitor resistor emulation
- Series switched capacitor resistor emulation
- Series-parallel switched capacitor resistor emulation
- Bilinear switched capacitor resistor emulation
Time constant accuracy of switched capacitor circuits is proportional to the
capacitance ratio and the clock frequency
Analysis of switched capacitor circuits includes the following steps:
1.) Analyze the circuit in the time-domain during a selected phase period.
2.) The resulting equations are based on q = Cv.
3.) Analyze the following phase period carrying over the initial conditions from the
previous analysis.
4.) Identify the time-domain equation that relates the desired voltage variables.
5.) Convert this equation to the z-domain.
6.) Solve for the desired z-domain transfer function.
7.) Replace z by ejT and examine the frequency response.
CMOS Analog Circuit Design
Page 9.2-1
R1
vIN
vOUT
vIN
R2
R1
vOUT
Fig. 9.2-01
Gain and GB = :
Vout R1+R2
Vin = R1
Gain , GB = :
Avd(0)R1
R1+R2
Vout(s) R1+R2
=
Avd(0)R1
Vin(s) R1
1 + R1+R2
Gain , GB :
GBR1
R1+R2 H
Vout(s) R1+R2 R1+R2
GBR1 = R1 s+H
Vin(s) = R1
s + R1+R2
Vout
R2
=
Vin
R1
R1Avd(0)
R2
Vout(s)
R1+R2
Vin(s)
Avd(0)R1
R1
1 + R1+R2
GBR1
R2 H
Vout(s) R2 R1+R2
=
=
GBR1 - R1 s+H
Vin(s) R1
s + R1+R2
Page 9.2-2
Page 9.2-3
Page 9.2-4
CHARGE AMPLIFIERS
Noninverting and Inverting Charge Amplifiers
C2
C1
vIN
vOUT
C2
C1
vOUT
Gain and GB = :
Vout C1+C2
Vin = C2
Gain , GB = :
Avd(0)C2
Vout C1+C2 C1+C2
Vin = C2
Avd(0)C2
1 + C1+C2
Gain , GB :
GBC2
Vout C1+C2 C1+C2
Vin = C2
GBC2
s + C1+C2
CMOS Analog Circuit Design
vIN
Vout
C1
=
Vin
C2
Avd(0)C2
Vout C1 C1+C2
Vin = -C2
Avd(0)C2
1 + C1+C2
GBC2
Vout C1 C1+C2
Vin = -C2
GBC2
s + C1+C2
P.E. Allen - 2004
Page 9.2-5
vin
+
C1
vC1
+
vC2
-
C2
vout
vin
C1
+
-
vC2
vout
+
C2
vC1
Analysis:
Find the even-odd and the even-even z-domain
transfer function for the above switched capacitor
inverting amplifier.
1: (n -1)T < t < (n -0.5)T
n- 23 n-1 n- 21
n+ 21 n+1 T
o
vC1
(n -1)T = vino (n -1)T
and
o
vC2(n -1)T = 0
CMOS Analog Circuit Design
Page 9.2-6
C1
+ o
vin (n-1)T
-
C1
vino (n-1)T
vC2 = 0 e
- + vout (n-1/2)T
C2
vC1 = 0
- +
+
Simplified equivalent circuit.
C1
e
o
vout (n-1/2)T = - C2 vin (n-1)T
e
V out
(z)
o
Vin (z)
C1
= -C2 z -1/ 2
Page 9.2-7
Assume that the applied input signal, vin (n-1)T, was unchanged during the previous
2 phase period(from t = (n-3/2)T to t = (n-1)T), then
o
V out(z) = -C z -1 Vin(z)
2
or
e
Hee (z) =
V out(z)
e
Vin(z)
C1
= -C2 z -1
Page 9.2-8
Hoe (e jT)
V out( e jT)
e
Vout( e jT)
C1
= - C2 e -jT/2
and
e
Hee (e jT) =
V out(e jT)
o
Vout( e jT)
C1
= -C2 e -jT
If C1/C2 = R2/R1, then the magnitude response is identical to inverting unity gain amp.
However, the phase shift of Hoe(e jT) is
Arg[Hoe(e jT)] = 180 - T/2
and the phase shift of Hee(e jT) is
Arg[Hee(e jT)] = 180 - T.
Comments:
The phase shift of the SC inverting amplifier has an excess linear phase delay.
When the frequency is equal to 0.5fc, this delay is 90.
One must be careful when using switched capacitor circuits in a feedback loop
because of the excess phase delay.
CMOS Analog Circuit Design
Page 9.2-9
v1(t)
CP
v1(t)
CP
C
2
CP
CP
=
= T
i2(average) = T i2(t)dt =
T
T
T/2
Substituting this expression into the one above shows that
RT = -T/C
Page 9.2-10
C
o
e
1
vout(n -1/2)T = C2 vin(n -1)T
Noninverting Switched Capacitor Voltage Amplifier.
C1
C1
o
e
V out(z) = C2 z -1/2 Vin(z) Hoe(z) = C2 z-1/2
If the applied input signal, vin(n -1)T, was unchanged during the previous 2 phase, then,
C1
e
e
V out(z) = C2 z-1 Vin(z)
C1
Hee(z) = C2 z-1
Comments:
Excess phase of H oe(e jT) is -T/2 and for H ee(e jT) is -T
CMOS Analog Circuit Design
Page 9.2-11
vC1(n -1)T = 0
and
vC1(t)
o
o
vout(n
-1)T = 0 .
vC2(n -1)T =
2: (n -0.5)T < t < nT
The voltage across C2 is
C1 e
e
vout(n -1/2)T = - C2 vin(n -1/2)T
C1 e
e
V out(z) = - C2V in(z)
vC2
vout
+
C2
C1
Hoe(z) = - C
Comments:
The inverting switched capacitor amplifier has no excess phase delay.
There is no transfer of charge during 1.
CMOS Analog Circuit Design
Page 9.2-12
vo
(1)
(2)
(3)
(4)
(5)
Page 9.2-13
Page 9.2-14
Page 9.2-15
COL
COL
COL
COL
COL
vC(t)
vin
Verror =
10-12
2110x10-6 - 10-12 (1+1.4+0)
M1
COL
M4
COL
COL
M5
vC2
+
C2
-
vout
+
Fig. 9.2-9
Page 9.2-16
10910-12 220x10-18
220x10-18+(0.5)(24.7)(10-16)
Verror =
Page 9.2-17
2 turning off
Finally, as switch M4 turns off, there will be feedthrough onto C2. Since, M4 has one
of its terminals at 0V, the feedthrough is the same as before and is 5.19mV. The final
voltage across C2, and therefore the output voltage vout, is given as
vout(2off) = vC2(2off) = 1.00022V + 0.00519V = 1.00541V
It is interesting to note that the last feedthrough has the most influence.
CMOS Analog Circuit Design
Page 9.2-18
C2
e
vout
(n-1/2)T
Avd(0)
+
e
vout
(n-1/2)T
e
out
Vout(z) C1 -1/2
1
.
= C2 z
Hoe(z) = o
+
C
C
1
2
V in(z)
1 - Avd(0)C2
Comments:
The phase response is unaffected by the finite gain
A gain of 1000 gives a magnitude of 0.998 rather than 1.0.
CMOS Analog Circuit Design
Page 9.2-19
Page 9.2-20
SUMMARY
Continuous time amplifiers are influenced by the gain and gainbandwidth of the op amp
Charge amplifiers are also influenced by the gain and the gainbandwidth of the op amp
Switched capacitor amplifiers replace the resistors of the continuous time amplifier with
switched capacitor equivalents
The transresistor SC amplifiers can be inverting and noninverting with the positive
input terminal of the op amp on ground
The nonidealities of the SC amplifier include:
- Switches
- Capacitors
- Op amp finite gain
- Op amp finite GB
Page 9.3-1
R1
C2
Vout
Vin
C2
R1
Vout
+
Inverter
(b.)
(a.)
Vin(j) j R 1C2 j
Vin(j) j R 1C2 j =
Frequency Response:
|Vout(j)/Vin(j)|
Arg[Vout(j)/Vin(j)]
40 dB
90
20 dB
0 dB
-20 dB
I I I
100 10
10I 100I
0
log10
log10
-40 dB
(a.)
(b.)
Page 9.3-2
=
=
Vin
Avd(s) (s/)
Avd(s) sR1C2 s
sR1C2
1 + (s/) + 1
1 + sR1C2+1
Avd(0)a GB GB
where
Avd(s) = s+a = s+a s
Vout
Case 1: s 0 Avd(s) = Avd(0) Vin - Avd(0)
GB I
GB
Vout
Case 2: s Avd(s) = s
V - s s
in
Vout
I
V - s
Case 3: 0 < s < Avd(s) =
in
|Vout(j)/Vin(j)|
Avd(0) dB
Eq. (1)
0 dB
Eq. (3)
I
x1 = I
Avd(0)
(2)
(3)
Arg[Vout(j)/Vin(j)]
I
180
10Avd(0) 10I
135
Avd(0)
x2 = GB
log10
Eq. (2)
(1)
90
45
0
GB
10
10GB
I
Avd(0)
GB
log10
Page 9.3-3
Example 9.3-1 - Frequency Range over which the Continuous Time Integrator is
Ideal
Find the range of frequencies over which the continuous time integrator
approximates ideal behavior if Avd(0) and GB of the op amp are 1000 and 1MHz,
respectively. Assume that I is 2000 radians/sec.
Solution
The idealness of an integrator is determined by how close the phase shift is to
90 (+90 for an inverting integrator and -90 for a noninverting integrator).
The actual phase shift in the asymptotic plot of the integrator is approximately 6 above
90 at the frequency 10I/Avd(0) and approximately 6 below 90 at GB /10.
Assume for this example that a 6 tolerance is satisfactory. The frequency range can be
found by evaluating 10I/Avd(0) and GB/10.
Therefore the range over which the integrator approximates ideal behavior is from 10Hz
to 100kHz. This range will decrease as the phase tolerance is decreased.
Page 9.3-4
vC2
vout
+
C2
vc2(n-1)T = vout(n-1)T .
2: (n -0.5)T < t < n T
Equivalent circuit:
o
vC2 = vout
(n-1)T e
vout (n-1/2)T
- +
C2
- o2
vin (n-1)T
+
+
t=0
vC1 = 0
+ -
t=0
C1
2
vino(n-1)T
C1
o
vout
(n-1)T C e
2 vout (n-1/2)T
+
- +
vC2 = 0
-
C1 o
e
o
We can write that, vout(n -1/2)T = C2 vin(n -1)T + vout(n -1)T
Page 9.3-5
oo
H (e ) = o j = C2 e j -1 = C2 e j/2 - e-j/2
V in( e )
Replacing ej/2 - e-j/2 by its equivalent trigonometric identity, the above becomes
o
Hoo(e
T
C1
V out(e j) C1
e-j/2
T/2
) = o j = C2 j2 sin(T/2) T = jTC2 sin(T/2) e-j/2
V in( e )
Page 9.3-6
1 /c
1
H(j) = 10j/ and Hoo(e j) = 10j/ sin(/ ) e-j/c
c
c
c
Plots:
0
5
Magnitude
-50
Arg[H(j)]
-100
3
oo j T
|H (e
)|
2
1
-200
|H(j)|
-150
-250
-300
0
0
0.2
0.4
/ c
0.6
0.8
0.2
0.4
0.6
/
c
0.8
Page 9.3-7
vin
vC1(t)
S1
o
c1
v (n -1)T = 0
C1
S2
S4
S3
vC2
+
C2
vout
and
3
o
o
e
vc2(n -1)T = vout(n -1)T = vout(n -2)T.
Equivalent circuit:
C1
t=0
t=0
2 vC1 = 0 2
vine(n-1/2)T
vC2 =
e
vout
(n-3/2)T e
- + vout (n-1/2)T
C2
+
vine (n-1/2)T
e
vC1 = 0 vout (n-3/2)T C2 e
vout (n-1/2)T
+
- + - +
vC2 = 0
+
C1
-
Page 9.3-8
ee
j
H (e ) = e j = - C2 e -1 = - C2 e j/2 - e-j/2
V in( e )
j/2
Replacing e - e-j/2 by 2j sin(T/2) and simplifying gives,
e
j
out(e )
C1
V
T/2
j
ee
H (e ) = e j = - jTC2 sin(T/2) e j/2
V in( e )
Same as noninverting integrator except for phase error.
Consequently, the magnitude response is identical but the phase response is given as
Arg[Hee(e j)] = 2 + 2 .
Comments:
The phase error is + for the inverting integrator and - for the noninverting integrator.
The cascade of an inverting and noninverting switched capacitor integrator has no
phase error.
CMOS Analog Circuit Design
Page 9.3-9
A Sign Multiplexer
A circuit that changes the 1 and 2 of the leftmost switches of the stray insensitive,
switched capacitor integrator.
1
2
x
VC
To switch connected
to the input signal (S1).
VC
This circuit steers the 1 and 2 clocks to the input switch (S1) and the leftmost switch
connected to ground (S2) as a function of whether Vc is high or low.
Page 9.3-10
e
out
o
vout
(n-1)T
Avd(0)
e
C2 vout (n-1/2)T
+
- +
vC2 = 0
-
o
vout
(n-1)T -
vC1 = 0
- +
e
C1 vout (n-1/2)T
Avd(0)
+
-
Fig. 9.3-10
-1
H (z) = o = 1 - z
C1
C1
1
1
V in(z)
1 - Avd(0) - Avd(0)C2(1-z-1) 1 - Avd(0) - Avd(0)C2(1-z-1)
CMOS Analog Circuit Design
Page 9.3-11
Hoo(e ) =
1
C1
C1/C2
1 - Avd(0) 1 + 2C2 - j
T
2Avd(0) tan 2
where now HI(e jT) is the integrator transfer function for Avd(0) = .
The error of an integrator can be expressed as
HI(j)
H(j) = [1-m()] e-j()
where
m() = the magnitude error due to Avd(0)
() = the phase error due to Avd(0)
If () is much less than unity, then this expression can be approximated by
HI(j)
H(j) 1 - m() - j()
Comparing Eq. (1) with Eq. (2) gives m() and ()due to a finite value of Avd(0) as
1
C1/C2
C1
m(j) = - A (0) 1 + 2C and (j) = 2A (0) tan(T/2)
vd
2
vd
(1)
(2)
Page 9.3-12
Example 9.3-3 - Evaluation of the Integrator Errors due to a finite value of Av d(0)
Assume that the clock frequency and integrator frequency of a switch capacitor
integrator is 100kHz and 10kHz, respectively. If the value of Avd(0) is 100, find the value
of m(j) and (j) at 10kHz.
Solution
The ratio of C1 to C2 is found as
C1
210,000
=
T
=
I
100,000 = 0.6283 .
C2
Substituting this value along with that for Avd(0) into m(j) and (j) gives
0.6283
m(j) = - 1 + 2 = -0.0131
and
0.6283
(j) = 2100tan(18) = 0.554 .
The ideal switched capacitor transfer function, HI(j), will be multiplied by a value of
approximately 1/1.0131 = 0.987 and will have an additional phase lag of approximately
0.554.
In general, the phase shift error is more serious than the magnitude error.
Page 9.3-13
Inverting Integrator
C2
C2
() -e-k C +C cos(T)
1
2
1
C2 GB
k1 C1+C2 fc
K. Martin and A.S. Sedra, Effects of the Op Amp Finite Gain and Bandwidth on the Performance of Switched-Capacitor Filters, IEEE Trans. on
Circuits and Systems, vol. CAS-28, no. 8, August 1981, pp. 822-829.
CMOS Analog Circuit Design
P.E. Allen - 2004
Page 9.3-14
vR2on
12d 2kTRon1 kT
2kTRon
= 2+2 = 2 = C Volts(rms)2
1
(2)
where 1 = 1/(RonC). Note that the switch has an effective noise bandwidth of
1
fsw = 4RonC Hz
which is found by dividing Eq. (2) by Eq. (1).
CMOS Analog Circuit Design
(3)
Page 9.3-15
SUMMARY
The discrete time noninverting integrator transfer function is
o
C1
V out(e j)
T/2
j
ee
H (e ) = e j = - jTC sin(T/2) ej/2
2
V in( e )
In general the integrator transfer function can be expressed as
H(ejT) = (Ideal)x(Magnitude error)x(Phase error)
Note that the cascade of an noninverting integrator with a inverting integrator has no
phase error
A capacitor C and a switch (or switches) has a thermal noise given as kT/C where T is
the clock period
Page 9.4-1
vout(t)
Dependent
Switched
Independent
Unswitched Voltage
Capacitor
Voltage
Capacitor
Source
Circuit
Source
Figure 9.4-1 - Two-port characterization of a general switched capacitor circuit.
Approach:
Four port - allows both phases to be examined
Two-port - simplifies the models but not as general
CMOS Analog Circuit Design
Page 9.4-2
v(t)
Ve(z)
Phase Dependent
Voltage Source
Vo(z)
1
t/T
z-1/2Vo(z)
Phase Independent
Voltage Source for
the Odd Phase
Vo(z)
1
t/T
Ve(z)
Phase Independent
Voltage Source for
the Even Phase
z-1/2Ve(z)
2
t/T
Page 9.4-3
+
v1(t)
-
C
2
+
v2(t)
-
Negative SC Transresistance
C
+ 2
v1(t)
1
-
2 +
1 v2(t)
-
Positive SC Transresistance
C
+
v1(t)
-
+
v2(t)
-
+
o
V1
e
V1
+
+
o
V1
-e
V1
+
+
o
V1
-e
V1
+
C(1-z-1)
-Cz-1/2
Cz-1/2
C
Cz-1/2
-Cz-1/2
-Cz-1/2
+
v2(t)
-
Cz-1/2
+
v1(t)
-
+
o
V1
e
V1
+
Switched Capacitor, Two-Port Circuit Four-Port, z-domain Equivalent Model Simplified, Two-Port z-domain Model
+
o
V2
e
V2
+
+
o
V2
e
V2
+
+
o
V1
-
Cz-1/2
+
e
V2
-
-Cz-1/2
+
e
V2
-
+
o
V2
-e
V2
+
C
+
+
e
e
V1
V2
(Circuit connected between
defined voltages)
+
o
V2
-e
V2
+
C(1-z-1)
+
+
e
e
V1
V2
(Circuit connected between
defined voltages) Fig. 9.4-3
K.R. Laker, Equivalent Circuits for Analysis and Synthesis of Switched Capacitor Networks, Bell System Technical Journal, vol. 58, no. 3,
March 1979, pp. 729-769.
CMOS Analog Circuit Design
P.E. Allen - 2004
Page 9.4-4
C
+
+
v2(t)
v1(t)
1
Capacitor and
Shunt Switch
C
+
o
V1
-e
V1
+
+
o
V2
e
V2
+
+
o
V1
e
V1
+
+
o
V1
-e
V1
+
+
o
V2
-e
V2
+
+
o
V2
e
V2
+
-Cz-1/2
Cz-1/2
-Cz-1/2
V1
+
-Cz-1/2
+
+
v2(t)
v1(t)
Unswitched
Capacitor
+
o
V1
-
Cz-1/2
Simplified Four-port
z-domain Model
-Cz-1/2
Switched Capacitor
Circuit
+
o
V2
-e
V2
+
Fig. 9.4-4
Page 9.4-5
+
vi(t)
-
+
vo(t) = Avvi(t)
-
+
Vio(z)
-
+
Voo(z) = AvVio(z)
-
Vie(z)
+
Voe(z) = AvVie(z)
+
Page 9.4-6
Cz-1/2
-Cz-1/2
Cz-1/2
Next, let us sum the currents flowing away from the positive V 2 node of the fourport z-domain model in Fig. 9.4-3. This equation is,
e
-Cz-1/2(V 2 - V 1 ) + Cz-1/2V 2 + CV 2 = 0.
e
Page 9.4-7
v1
v2
1
v1
v3
vo
v4
2
1
+
Fig. 9.6-4a
V2(z)
V3(z)
V4(z)
Vo(z)
V1(z)
1
2
2
2
1
V1(z)
e
V2(z)
V4(z)
V3(z)
+
-
Vo(z)
Fig9.4-6b
V2(z)
V3(z)
V4(z)
e
1
o
V1(z)
2
2
Vo(z)
2
1
+
Fig. 9.4-7
Page 9.4-8
z-1/2V
o
i (z)
e
C2(1-z-1)V o (z)
=0
Hoe(z)
C1
C2
2
2
vi(t)
vo
2
-
+
-C1z-1/2
(a.)
C2(1-z-1) V e(z)
o
Vo(z)
Vi(z)
z-1/2Vo(z)
(b.)
Figure 9.4-8 - (a.) Modified equivalent circuit
of Fig. 9.3-4a. (b.) Two-port, z-domain model
for Fig. 9.4-8a.
e
o
(V o (z)/V i (z))
C1z-1/2
= C2(1-z-1) .
Hoo(z) is found by using the relationship that V o (z) = z-1/2V o (z) to get
C1z-1
o
o
Hoo(z) = (V o (z)/V i (z)) = C2(1-z-1)
which is equal to z-domain transfer function of the noninverting SC integrator.
CMOS Analog Circuit Design
Page 9.4-9
C1
2
e
C2(1-z-1)V o (z)
+
=0
which can be rearranged to give
e
Hee(z) =
V o (z)
e
vi(t)
C2
2
vo
2
+
C1
e
Vi(z)
(a.)
C2(1-z-1) V e(z)
o
Vo(z)
z-1/2Vo(z)
(b.)
Figure 9.4-9 - (a.) Modified equivalent circuit of
inverting SC integrator. (b.) Two-port, z-domain
model for Fig. 9.4-9a
-C1
= C (1-z-1) .
2
V i (z)
which is equal to inverting, switched capacitor integrator z-domain transfer function.
o
Heo(z) is found by using the relationship that V o (z) = z-1/2V o (z) to get
o
Heo(z)
V o (z)
e
V i (z)
C1z-1/2
= C (1-z-1) .
2
P.E. Allen - 2004
Page 9.4-10
C1
1
o
V 2 (z)
and
for the summing, switched capacitor
integrator of Fig. 9.4-10a.
Solution
This circuit is time-variant because C3 is
charged from a different circuit for each phase.
Therefore, we must use a four-port model. The
resulting z-domain model for Fig. 9.4-10a is
shown in Fig. 9.4-10b.
v1(t)
C3
vo
2
-
+
C1
1
1
2
v2(t)
Page 9.4-11
(1)
e
o
C2V 2 (z)
o
o
o
+ C3V o (z) - C1z-1V 1 (z) - C3z-1V o (z) = 0 (3)
o
Solving for V o (z) gives,
o
o
C1z-1V 1 (z) C2V 2 (z)
o
V o (z) = C3(1-z-1) - C3(1-z-1)
Multiplying Eq. (1) by z-1/2 and adding it to Eq. (2) gives
o
Vo(z)
Vi(z)
- C3
V1(z)
-C1z-1/2
V2(z)
C2
e
Vi(z)
C3
Voe(z)
-C3z-1/2
Fig. 9.4-10b - Four-port, z-domain
model for Fig. 9.4-10a.
e
V o (z)
Page 9.4-12
i(t)
Time-domain:
+
v1(t)
-
T
T
i(t) = C v t - 2 - v2t - 2
1
Cv3(t)
i(t)
+
v2(t)
+v3(t) -
Delay of T/2
Rin =
T
2
+
Fig. 9.4-11b
SPICE Primitives:
2
1
CV4
LosslessTransmission Line
V1-V2
TD = T/2, Z0 = R
Fig. 9.4-11c
B.D. Nelin, Analysis of Switched-Capacitor Networks Using General-Purpose Circuit Simulation Programs, IEEE Trans. on Circuits and Systems, pp. 43-48, vol. CAS-30,
No. 1, Jan. 1983.
Page 9.4-13
C2z-1/2
-C2z-1/2
-C2z-1/2
C2z-1/2
C1
C1z-1/2
-C1z-1/2
C1z-1/2
C1
T
=
I
C2
fc = 0.6283
AssumeC2 = 1F C1 = 0.6283F.
C2
1
3
5
Next we replace the switched
+
+o
106V3
o
Vi
Vo
capacitor C1 and the unswitched
0
0
0
capacitor of integrator by the z-e
-e
Vo
Vi
domain model of the second row of
+
+
106V4
Fig. 9.4-3 and the first row of Fig.
2
4
6
C2
9.4-4 to obtain Fig. 9.4-12. Note Figure 9.4-12 - z-domain model for noninverting switched capacitor integrator.
that in addition we used Fig. 9.4-5
for the op amp and assumed that the op amp had a differential voltage gain of 106. Also,
the unswitched Cs are conductances.
As the op amp gain becomes large, the important components are indicated by the
darker shading.
P.E. Allen - 2004
Page 9.4-14
VIN 1 0 DC 0 AC 1
R10C1 1 0 1.592
X10PC1 1 0 10 DELAY
G10 1 0 10 0 0.6283
X14NC1 1 4 14 DELAY
G14 4 1 14 0 0.6283
R40C1 4 0 1.592
X40PC1 4 0 40 DELAY
G40 4 0 40 0 0.6283
X43PC2 4 3 43 DELAY
G43 4 3 43 0 1
R35 3 5 1.0
X56PC2 5 6 56 DELAY
G56 5 6 56 0 1
R46 4 6 1.0
X36NC2 3 6 36 DELAY
Page 9.4-15
200
150
Magnitude
4
3
Both H
oe
and H
oo
2
1
100
50
0
oo
Phase of H (jw)
-50
oe
Phase of H (jw)
-100
-150
20
40
60
Frequency (kHz)
(a.)
80
100
-200
20
40
60
Frequency (kHz)
(b.)
80
100
Comments:
This approach is applicable to all switched capacitor circuits that use two-phase,
nonoverlapping clocks.
If the op amp gain is large, some simplification is possible in the four-port z-domain
models.
The primary advantage of this approach is that it is not necessary to learn a new
simulator.
CMOS Analog Circuit Design
Page 9.4-16
K. Suyama, Users Manual for SWITCAP2, Version 1.1, Dept. of Elect. Engr., Columbia University, New York, NY 10027, Feb. 1992.
Page 9.4-17
Page 9.4-18
Page 9.4-19
Logic
Threshold
+
...
...
+
-
+
v
-
Av
Page 9.4-20
SWITCAP - Resistors
RQ
RQ
Ceq
R= T
4Ceq
RQ
RQ
RQ
t
RQ
The clock, RQ, for the resistor is run at a frequency, much higher than the system clock in
order to make the resistor model still approximate a resistor at frequencies near the
system clock.
Page 9.4-21
RON
MQ
S
S
Cgd
Cbs
Cbd
Frequency
Higher than
MQ clock
More information:
SWITCAP Distribution Center
Columbia University
411 Low Memorial Library
New York, NY 10027
suyama@elab.columbia.edu
CMOS Analog Circuit Design
Page 9.4-22
SUMMARY
Can replace various switch-capacitor combinations with a z-domain model
The z-domain model consists of:
- Positive and negative conductances
- Delayed conductances (storistor)
- Controlled sources
- Independent sources
These models permit SPICE simulation of switched capacitor circuits
The type of clock circuits considered here are limited to two-phase clocks
Page 9.5-1
Page 9.5-2
Page 9.5-3
2
1
1C1
1
- C1
+
2
C
1 2 1 1
C1
vo(t)
vo(t)
2
vi(t)
vo(t)
1
vi(t)
1C1
2
1
(a.)
(b.)
Figure 9.5-1 - (a.) Noninverting, first-order low pass circuit. (b.) Equivalent circuit of Fig. 9.5-1a.
Transfer function:
Summing currents flowing toward the inverting
op amp terminal gives
e
2C1V o (z)
Vo(z)
C12
-C11z-1/2 C1(1-z-1) V e(z)
o
Vo(z)
z
e
+
1
o
z-1/2Vo(z)
V o (z)
1z-1
1+2
=
=
o
-1
z-1
Figure 9.5-2 - z-domain model of Fig. 9.5-1b.
V i (z) 1 + 2 - z
1 - 1+2
Equating the above to H(z) of the previous page gives the design equations as
1 = A0/B0
and
2 = (1-B0)/B0
CMOS Analog Circuit Design
Page 9.5-4
2
1
1C1
2
vo(t)
- C1
C1
vo(t)
vo(t)
2
vi(t)
2
2C1
1
1
2
vi(t)
1C1
1
1
Equivalent circuit.
e
V o (z)
e
Page 9.5-5
1z-1
= 2 + 1- z-1
o
V (z)
V o (z)
(2)
Next, we note from Eq. (1) that 1-z-1 sT. Furthermore, if sT<<1, then z-1 1.
Making these substitutions in Eq. (2), we get
o
V o (s)
1
1/ 2
2 + sT = 1 + s(T/2)
(3)
V
Equating Eq. (3) to the specifications gives 1 = 102 and 2 = -3dB/fc
2 = 6283/100,000 = 0.0628 and 1 = 0.6283
o
i (s)
Page 9.5-6
2C
2 1
1
1C
1C
vo(t)
Transfer function:
Summing currents at the
inverting input node of the op
amp gives
e
vo(t)
2
- C
vi(t)
- C
vi(t)
(b.)
(a.)
Figure 9.5-3 - (a.) Switched-capacitor, high pass circuit. (b.) Version of Fig. 9.5-3a
that constrains the charging of C1 to the 2 phase.
2
1(1-z-1)
(1-z-1)
Voe(z)
Vo(z)
Vi(z)
z-1/2Vo(z)
Page 9.5-7
3C
2C
1
1
3C
2C
1
1
vo(t)
1
vi(t)
1C
2
1
2
vo(t)
C
vi(t)
C
2 1 1
Transfer function:
(a.)
(b.)
Summing the currents
Figure 9.5-5 - (a.) High or low frequency boost circuit. (b.) Modification of (a.) to simplify
flowing into the
the z-domain modeling
inverting input of the
3(1-z-1)
2
op amp gives
o
-1z-1/2 (1-z-1) V e(z)
-1z-1/2Voi (z)+3(1-z-1)Vei (z)+2Veo(z)+(1-z-1)Veo(z) = 0 Vie(z)
Vo(z)
o
Since Voi(z) = z-1/2Vei(z), then the above becomes
o
Vi(z)
e
+
z-1/2Vo(z)
Veo(z) 2+1-z-1 = 1z-1Vei(z) - 3(1-z-1)Vei(z)
Solving for Hee(z) gives
Figure 9.5-6 - z-domain model for Fig. 9.5-5b.
1+3
1-B0
1z-1-3(1-z-1) -3 1- 3 z-1
A1+A0
- A0
=
and
=
Hee(z) = +(1-z-1) = 2+1
1
2 B0
3
B0
B0
z-1
2
1-2+1
Page 9.5-8
From Fig. 9.5-7, we see that the desired response has a dc gain of 10, a right-half
plane zero at 2 kHz and a pole at -200 Hz. Thus, we see that the following
relationships must hold.
1
1
2
=
10
,
=
2000
,
and
2
T 3
T = 200
From these relationships we get the desired values as
2000
200
1 = fc , 2 = fc , and 3 = 1
CMOS Analog Circuit Design
Page 9.5-9
1C
1
2
vi(t)
C1
1
1C
-+
+-
2
+
2
1
1C
2C
-+
+-
vo(t) vi(t)
vo(t)
2C
1C
2C
1
2
+
vo(t)
2C
1C
2
1
1C
1
2
3C
2 2C
1
2
2
1
vo(t) vi(t)
-
3C
-+
+-
vo(t)
-
vo(t)
2 2C
1
1
(c.)
(a.)
(b.)
Figure 9.5-8 - Differential implementations of (a.) Fig. 9.5-1, (b.) Fig. 9.5-3, and (c.) Fig. 9.5-5.
2
1
Comments:
Differential operation reduces clock feedthrough, common mode noise sources and
enhances the signal swing.
Differential operation requires op amps or OTAs with differential outputs which in turn
requires a means of stabilizing the output common mode voltage.
Page 9.5-10
SUMMARY
Examined first-order SC circuits (lowpass, highpass, allpass)
Illustrated design by assuming the clock frequency is higher than the signal frequency
(s-domain design)
Illustrated direct design by equating coefficients between the desired and design in the
z-domain (requires the specifications in the z-domain)
Page 9.6-1
SecondOrder
Circuit
Stage 1
SecondOrder
Circuit
Stage 2
(a.)
SecondOrder
Circuit
Stage n
Vout
FirstOrder
Circuit
Stage 1
SecondSecondVout
Order
Order
Circuit
Circuit
Stage 2
Stage n
(b.)
Figure 9.6-1 - (a.) Cascade design when n is even. (b.) Cascade designwhen n is odd.
Vin
Ladder design
Also uses first- and second-order circuits
There are also other applications of first- and second-order circuits:
Oscillators
Converters
CMOS Analog Circuit Design
Page 9.6-2
=
K
Ha(s) = Vin(s) =
(s-p )(s-p )
o
1
2
s2 + Q s+ o2
j
o
2Q
Page 9.6-3
-1
1
Vout(s) = s (K1 + K2s)Vin(s) + Q Vout(s) + s (K0Vin(s) +o2Vout(s))
If we define the voltage V1(s) as
-1 K0
-1
o
Vout(s) = s (K1 + K2s) Vin(s) + Q Vout(s) - oV1(s)
Synthesizing the voltages V1(s)
Vin(s) K2
CA=1
and Vout(s), gives
Vout(s) 1/o
Vin(s) o/K0
V1(s)
CB=1
Vin(s) 1/K1
Vout(s) Q/o
Vout(s)
V1(s) -1/o
(a.)
(b.)
Figure 9.6-2 - (a.) Realization of V1(s). (b.) Realization of Vout(s).
CMOS Analog Circuit Design
Page 9.6-4
Vin(z)
2C1
Vout(z)
C1
2
1
e
Vin(z)
2
1
C2 V e (z)
out
4C2
Vin(z)
1C1
2
1
V1(z)
5C2
V1(s)
1
2
6C2
Vout(z)
2
1
(a.)
(b.)
Figure 9.6-3 - (a.) Switched capacitor realization of Fig. 9.6-2a. (b.) Switched
capacitor realization of Fig. 9.6-2b.
Note that we multiplied the V 1 (z) input of Fig. 9.6-3b by z-1/2 to convert it to V 1 (z).
CMOS Analog Circuit Design
Page 9.6-5
2C 1
e
V1(z)
5C 2
1
2
6C 2
C2
1
e
Vout(z)
+
If we assume that T<<1,
4C 2
e
1
1
then 1-z-1 sT and V1(z)
3C 2
e
andVout(z) can be
Figure 9.6-4 - Low Q, switched capacitor, biquad realization.
approximated as
1 e
2 e
2 e
-1 1 e
e
V 1 (s) - sT Vin(s) - sT Vout(s) = s T Vin(s) + T Vout(s)
and
5 e
6 e
-1 4
e
e
V out(s) s ( T + s3)Vin(s) - T V 1(s) + T Vout(s) .
These equations can be combined to give the transfer function, Hee(s) as follows.
s4 15
-3s2 + T + T2
Hee(s)
s6 25
s2 + T + T2
CMOS Analog Circuit Design
Page 9.6-6
s4 15
-3s2 + T + T2 -(K s2+ K s + K )
2
1
0
Equating Hee(s) to Ha(s) gives
o
s6 25 =
s2 + Q s+ o2
s2 + T + T2
oT
K0T
which gives,
1 = o , 2 = |5| = oT, 3 = K2, 4 = K1T, and 6 = Q .
Largest capacitor ratio:
If Q > 1 and oT << 1, the largest capacitor ratio is 6.
For this reason, the low-Q, switched capacitor biquad is restricted to Q < 5.
Sum of capacitance:
To find this value, normalize all of the capacitors connected or switched into the
inverting terminal of each op amp by the smallest capacitor, minC. The sum of the
normalized capacitors associated with each op amp will be the sum of the capacitance
connected to that op amp. Thus,
1 n
C = min i
i =1
where there are n capacitors connected to the op amp inverting terminal, including the
integrating capacitor.
CMOS Analog Circuit Design
Page 9.6-7
Page 9.6-8
Page 9.6-9
Voltage Scaling
It is desirable to keep the amplitudes of the output voltages of the two op amps
approximately equal over the frequency range of interest. This can be done by voltage
scaling.
If the voltage at the output node of an op amp in a switched capacitor circuit is to
be scaled by a factor of k, then all switched and unswitched capacitors connected to that
output node must be scaled by a factor of 1/k.
For example,
1C1
C1
v1
2C2
C2
Page 9.6-10
1 K0 K1
s
Vout(s) 1/o
CB=1
Vin(s) K2
Vout(s) 1/Q
Vin(s) K1/o
V1(s)
V1(s) -1/o
Vout(s)
Vin(s) o/K0
Realization of Vout(s).
Realization of V1(s).
Page 9.6-11
4C1
Vin(z)
3C1
2C1
Vout(z)
C1
2
1
e
Vin(z)
1C1
2
1
V1(z)
e
6C2
5C2 2
Vin(z)
V1(s)
1
2
C2 V e (z)
out
+
1
(a.)
(b.)
Figure 9.6-6 - (a.) Switched capacitor realization of Fig. 9.6-5a. (b.) Switched
capacitor realization of Fig. 9.6-5b.
Note that we multiplied the V 1 (z) input of Fig. 9.6-6b by z-1/2 to convert it to V 1 (z).
CMOS Analog Circuit Design
Page 9.6-12
2
1
4C1
1
2
C2
2
Vout(z)
e
andVout(z)
5C2
1 Ve (z)
1
1 1
Vin(z)
2C1
6C2
can be
Figure 9.6-7 - High Q, switched capacitor, biquad realization.
approximated as
e
1 1
1 2
e
e
V 1 (s) - s T + s3V in(s) - s T + s4V out(s)
and
5 e
-1
e
e
V out(s) s (s6)Vin(s) - T V 1(s) .
These equations can be combined to give the transfer function, Hee(s) as follows.
s35 15
-6s2 + T + T2
Hee(s)
s45 25
s2 + T + T2
CMOS Analog Circuit Design
Page 9.6-13
s35 15
2
-6s + T + T2 -(K s + K s + K )
2 2
1
0
o
s45 25 =
s2 + Q s+ o2
s2 + T + T2
which gives,
K0T
K1
1
1 = o , 2 = |5| = oT, 3 = o, 4 =Q, and 6 = K2 .
Largest capacitor ratio:
If Q > 1 and oT << 1, the largest capacitor ratio is 2 (5) or 4 depending on the
values of Q and oT.
Page 9.6-14
Page 9.6-15
Page 9.6-16
C
2
2
e
Vin(z)
2
V1(s)
Vout(z)
+
I
H
1
J
2
1
2
L
2
e
V out(z)
(DJ^ - AH^)z-2 - [D( I^ + J^) - AG^]z - D I^
=
e
-2 -1
V in(z) (DB - AE)z [2DB - A(C + E) + DF]z + D(B +F)
e
^) + FH^ - E( I^+J^) - CJ^]z-1 - [ I^(C+E) - G
^(F+B)]
V 1(z) (EJ^ - BH^)z-2+[B(G^+H
=
e
(DB - AE)z-2 - [2DB - A(C + E) + DF]z-1 + D(B +F)
V in(z)
G^ = G+L,
where
H^ = H+L ,
I^ = I+K
and
J^ = J+L
Page 9.6-17
G
e
Vin(z)
-Hz-1
B(1-z-1)
e
(z)
Vout
+
I
-Jz-1
L(1-z-1)
Figure 9.6-9 - z-domain equivalent circuit for the Fleischer-Laker biquad of Fig. 9.6-8.
Type 1E Biquad (F = 0)
e
V out
z-2(JD - HA) + z-1(AG - DJ - DI) + DI
e = -2
-1
V in z (DB - AE) + z (AC + AE - 2BD) + BD
and
e
Page 9.6-18
G
e
Vin(z)
-Hz-1
B(1-z-1)
e
Vout
(z)
+
I
-Jz-1
L(1-z-1)
Figure 9.6-9 - z-domain equivalent circuit for the Fleischer-Laker biquad of Fig. 9.6-8.
Type 1F Biquad (E = 0)
e
Page 9.6-19
Page 9.6-20
Page 9.6-21
SUMMARY
The second-order switched capacitor circuit is a very versatile circuit
The second-order switched capacitor circuit will be very useful in filter design
Low-Q biquad is good for Qs up to about 5 before the elements spreads become large
Design methods:
- Assume that fc>>fsig and using continuous time specifications and design
- Direct design equate the z-domain transfer function to a z-domain specification
and solve for the capacitor ratios
Page 9.7-1
Stopband
fcutoff =
fPassband
Frequency
Phase
0 0
Frequency
Slope =
-Time delay
Page 9.7-2
Characterization of Filters
A low pass filter magnitude response.
Tn(jn)
T(j)
1
T(jPB)/T(j0)
T(j0)
T(jPB)
T(jSB)
0
0
T(jSB)/T(j0)
0
0
PB SB
1 SB/PB=n
(b.)
(a.)
Figure 9.7-1 - (a.) Low pass filter. (b.) Normalized, low pass filter.
Page 9.7-3
An(jn) dB
1
log10(n)
T(jPB)
A(jSB)
T(jSB)
A(jPB)
0
0
log10(n)
1
n
(b.)
(a.)
Figure 9.7-2 - (a.) Low pass filter of Fig. 9.7-1 as a Bode plot. (b.) Low pass filter of
Fig. 9.7-2a shown in terms of attenuation (A(j) = 1/T(j)).
Therefore,
Passband ripple = T(jPB) dB
Stopband gain = T(jSB) dB or Stopband attenuation = A(jPB)
Transition frequency is still = n = SB/PB
Page 9.7-4
(j ) =
0.8
A
0.6
|T LPn (j n)|
0.4
Butterworth Magnitude
Approximation:
LPn
N=5
N=4
N=6
1
1+
0.2
N=3
N=2
N=8
2N
N=10
1 + 2 n
0
0
0.5
1
1.5
2
where N is the order of the
Normalized Frequency, n
approximation and is defined in
the above plot.
The magnitude of the Butterworth filter approximation at SB is given as
j SB
1
T
= |T
(j
)|
=
T
=
LPn
LPn
n
SB
2N
PB
1 + 2 n
This equation in terms of dB is useful for finding N given the filter specifications.
2.5
2N
Page 9.7-5
Page 9.7-6
Poles
-0.70711 j0.70711
-0.50000 j0.86603
-0.38268 j0.92388
-0.92388 j0.38268
-0.30902 j0.95106
-0.80902 j0.58779
-0.25882 j0.96593
-0.96593 j0.25882
-0.70711 j0.70711
-0.22252 j0.97493
-0.90097 j0.43388
-0.62349 j0.78183
-0.19509 j0.98079
-0.83147 j0.55557
-0.55557 j0.83147
-0.98079 j0.19509
-0.17365 j0.98481
-0.76604 j0.64279
-0.50000 j0.86603
-0.93969 j0.34202
-0.15643 j0.98769
-0.89101 j0.45399
-0.45399 j0.89101
-0.98769 j0.15643
-0.70711 j0.70711
a1 coefficient
1.41421
1.00000
0.76536
1.84776
0.61804
1.61804
0.51764 1.93186
1.41421
0.44505 1.80194
1.24698
0.39018 1.66294
1.11114 1.96158
0.34730 1.53208
1.00000 1.87938
0.31286 1.78202
0.90798 1.97538
1.41421
P.E. Allen - 2002
Page 9.7-7
Example 9.7-2 - Finding the Butterworth Roots and Polynomial for a given N
Find the roots for a Butterworth approximation with =1 for N = 5.
Solution
For N = 5, the following first- and second-order products are obtained from Table
9.7-1
1
1
1
2
1.5
TLPn(jn )
1
T1(jn )
0.5
T3 (jn)
0
0.5
1
1.5
2
Normalized Frequency, n
2.5
Page 9.7-8
1
A
0.8
1
1+2
N=2
0.6
TLPn(jn )
0.4
N=3
N=4
0.2
0
N=5
0.5
1
1.5
2
Normalized Frequency, n
2.5
1
(jn) = 1 + 2 cosh2[Ncosh-1( )] , n > 1
n
where N is the order of the filter approximation and is defined as
1
|TLPn(PB)| = |TLPn(1)| = TPB = 1+2 .
N is determined from 20 log10(TSB) = TSB (dB) = -10log10{1 + 2cosh2[Ncosh-1(n)]}
T
LPn
Page 9.7-9
Page 9.7-10
Normalized Pole
Locations
-0.54887 j0.89513
-0.24709 j0.96600
-0.49417
-0.13954 j0.98338
-0.33687 j0.40733
-0.08946 j0.99011
-0.23421 j0.61192
-0.28949
-0.06218 j0.99341
-0.16988 j0.72723
-0.23206 j0.26618
-0.04571 j0.99528
-0.12807 j0.79816
-0.18507 j0.44294
-0.20541
a0
a1
1.10251
0.99420
1.09773
0.49417
0.98650
0.27940
0.98831
0.42930
0.27907
0.67374
0.17892
0.46841
0.99073
0.55772
0.12471
0.99268
0.65346
0.23045
0.12436
0.33976
0.46413
0.09142
0.25615
0.37014
P.E. Allen - 2002
Page 9.7-11
0.9883
0.4293
2
sn+0.1789sn+0.9883sn+0.4684sn+0.4293
Other Approximations
Thomson Filters - Maximally flat magnitude and linear phase1
Elliptic Filters - Ripple both in the passband and stopband, the smallest transition region
of all filters.2
An excellent collection of filter approximations and data is found in A.I. Zverev,
Handbook of Filter Synthesis, John Wiley & Sons, Inc., New York, 1967.
W.E. Thomson, Delay Networks Having Maximally Flat Frequency Characteristics, Proc. IEEE, part 3, vol. 96, Nov. 1949, pp. 487-490.
W. Cauer, Synthesis of Linear Communication Networks, McGraw-Hill Book Co., New York, NY, 1958.
ECE 6414 - Analog Integrated Systems Design
Page 9.7-13
Normalized
LP Filter
Root
Locations
Frequency
Transform the
Roots to HP,
BP, or BS
Cascade of
First- and/or
Second-Order
Stages
Normalized
Low-Pass
RLC Ladder
Realization
Frequency
Transform the
L's and C's to
HP, BP, or BS
First-Order
Replacement
of Ladder
Components
Denormalize
the Filter
Realization
All designs start with a normalized, low pass filter with a passband of 1 radian/second and
an impedance of 1 that will satisfy the filter specification.
1.) Cascade approach - starts with the normalized, low pass filter root locations.
2.) Ladder approach - starts with the normalized, low pass, RLC ladder realizations.
A Design Procedure for the Low Pass, SC Filters Using the Cascade Approach
1.) From TPB, TSB, and n (or APB, ASB, and n) determine the required order of the
filter approximation, N.
2.) From tables similar to Table 9.7-1 and 9.7-2 find the normalized poles of the
approximation.
3.) Group the complex-conjugate poles into second-order realizations. For odd-order
realizations there will be one first-order term.
4.) Realize each of the terms using the first- and second-order blocks of the previous
lectures.
5.) Cascade the realizations in the order from input to output of the lowest-Q stage first
(first-order stages generally should be first).
More information can be found elsewhere1,2,3,4.
K.R. Laker and W.M.C. Sansen, Design of Analog Integrated Circuits and Systems, McGraw Hill, New York, 1994.
P.E. Allen and E. Sanchez-Sinencio, Switched Capacitor Circuits, Van Nostrand Reinhold, New York, 1984.
R. Gregorian and G.C. Temes, Analog MOS Integrated Circuits for Signal Processing, John Wiley & Sons, New York, 1987.
3
4
L.P. Huelsman and P.E. Allen, Introduction to the Theory and Design of Active Filters, McGraw Hill Book Company, New York, 1980.
ECE 6414 - Analog Integrated Systems Design
P.E. Allen - 2002
Page 9.7-15
Example 9.7-5 - Fifth-order, Low Pass, SC Filter using the Cascade Approach
Design a cascade, switched capacitor realization for a Chebyshev filter approximation
to the filter specifications of TPB = -1dB, TSB = -25dB, fPB = 1kHz and fSB = 1.5kHz. Give
a schematic and component value for the realization. Also simulate the realization and
compare to an ideal realization. Use a clock frequency of 20kHz.
Solution
First we see that n = 1.5. Next, recall that when TPB = -1dB that this corresponds to =
0.5088. We find that N = 5 satisfies the specifications (TSB = -29.9dB). Using the results
of previous lecture, we may write TLPn(sn) as
0.2895
0.9883
0.4293
2
TLPn(sn) = sn+0.2895 2
(1)
sn+0.1789sn+0.9883sn+0.4684sn+0.4293
Next, we design each of the three stages
1
individually.
21C11
11
11
2
Vin(ej)
V2(ej)
Stage 1 - First-order Stage
1
2
- C11
2
1
Let us select the first-order stage shown. We will
+
assume that fc is much greater than fBP (i.e. 100) and use
transfer function shown below to accomplish the design.
Stage 1
11/21
(2)
T1(s) 1 + s(T/21)
ECE 6414 - Analog Integrated Systems Design
Chapter 9 Section 7 (5/2/04)
Page 9.7-17
2
1
1
2
1
+
+
T(0) n
(5)
= 2 n
Stage 2
2
sn + Q sn + n
where T(0) = 1, n = 0.9941 and Q = (0.9941/0.1789) = 5.56. Therefore, select the lowpass version of the high-Q biquad. First, apply the normalization of Eq. (3) to get
sn3252 1252
-62s n2 + Tn
+
T n2
.
(6)
T2(sn)
sn4252 2252
2
s n + Tn
+
T n2
To get a low pass realization, select 32 = 62 = 0 to get
-(1252/T n2)
(7)
T2(sn)
sn4252 2252 .
2
+
s n + Tn
T n2
ECE 6414 - Analog Integrated Systems Design
Chapter 9 Section 7 (5/2/04)
C
63
23
j
13
13
53
23
2
V3(e )
2
Vout(ej)
sn + 0.4684sn + 0.4293
2
1
2
2
- C13
- C23
1
1
2
1
T(0) n2
+
+
= 2
(8)
2
s n + (n/Q)sn + n
Stage 3
where we see that T(0) = 1, n = 0.6552 and Q = (0.6552/0.4684) = 1.3988. Therefore,
select the low pass version of the low-Q biquad. Apply the normalization of Eq. (3) to get
ECE 6414 - Analog Integrated Systems Design
Page 9.7-19
sn43 1353
-33s n2 + Tn +
2
Tn
T3(sn)
.
(9)
sn63 2353
2
s n + Tn +
T n2
To get a low pass realization, select 33 = 43 = 0 to get
- (1353/T n2)
(10)
T3(sn)
sn63 2353 .
2
s n + Tn +
T n2
Equating Eq. (10) to the last term of TLPn(sn) gives
0.4293PB2 0.429342
2
1353 = 2353 = 0.4293T n =
=
= 0.04184
fc2
400
and
63 = 0.4684Tn = (0.4684PB/fc) = (0.46842/20) = 0.1472
Choose a13 = a23 = 53 to get optimum voltage scaling. Thus , 13 = 23 = 53 = 0.2058
and 63 = 0.1472. The third-stage capacitance is
3rd-stage capacitance = 1+(3(0.2058)/0.1472)+(2/0.1472) =18.78 units of capacitance
The total capacitance of this design is 13 + 17.32 + 18.78 = 49.10 units of capacitance.
ECE 6414 - Analog Integrated Systems Design
Page 9.7-20
2
2
21C11
11C11
C11
Stage 3
2
2
1
23C13
13C13
1
C13
63C23
53C23
1
2
2
Stage 2
22C12
12C12
1
42C12
52C22
2
1
C23
C12
Vout(ej)
2
2
C22
Page 9.7-21
200
Stage 1 Output
-20
Stage 3 Output
-30
-40
Stage 2 Output
(Filter Output)
-50
150
Phase (Degrees)
Magnitude (dB)
-10
100
50
0
-50
-100
-60
-150
-70
-200
500
1000
500
1000
Comments:
There appears to be a sinx/x effect on the magnitude which causes the passband
specification to not be satisfied. This can be avoided by prewarping the specifications
before designing the filter.
Stopband specifications met
None of the outputs of the biquads exceeds 0 dB (Need to check internal biquad nodes)
ECE 6414 - Analog Integrated Systems Design
Page 9.7-22
1 0 DC 0 AC 1
XNC7 17 18 19 20 NC7
XAMP5 19 20 21 22 AMP
XUSCP5 19 20 21 22 USCP
XUSCP6 21 22 15 16 USCP1
XPC8 21 22 15 16 PC6
SUBCKT DELAY 1 2 3
ED 4 0 1 2 1
TD 4 0 3 0 ZO=1K TD=25US
RDO 3 0 1K
.ENDS DELAY
.SUBCKT NC1 1 2 3 4
RNC1 1 0 11.0011
XNC1 1 0 10 DELAY
GNC1 1 0 10 0 0.0909
XNC2 1 4 14 DELAY
GNC2 4 1 14 0 0.0909
XNC3 4 0 40 DELAY
GNC3 4 0 40 0 0.0909
RNC2 4 0 11.0011
.ENDS NC1
.SUBCKT NC3 1 2 3 4
RNC1 1 0 4.8581
XNC1 1 0 10 DELAY
GNC1 1 0 10 0 0.2058
XNC2 1 4 14 DELAY
GNC2 4 1 14 0 0.2058
XNC3 4 0 40 DELAY
GNC3 4 0 40 0 0.2058
RNC2 4 0 4.8581
Ends NC3
P.E. Allen - 2002
Page 9.7-23
.SUBCKT NC7 1 2 3 4
RNC1 1 0 3.2018
XNC1 1 0 10 DELAY
GNC1 1 0 10 0 0.3123
XNC2 1 4 14 DELAY
GNC2 4 1 14 0 0.3123
XNC3 4 0 40 DELAY
GNC3 4 0 40 0 0.3123
RNC2 4 0 3.2018
.ENDS NC7
.SUBCKT PC1 1 2 3 4
RPC1 2 4 11.0011
.ENDS PC1
.SUBCKT PC2 1 2 3 4
RPC1 2 4 4.8581
.ENDS PC2
.SUBCKT PC4 1 2 3 4
RPC1 2 4 6.7980
.ENDS PC4
.SUBCKT PC6 1 2 3 4
RPC1 2 4 3.2018
.ENDS PC6
.SUBCKT USCP 1 2 3 4
R1 1 3 1
R2 2 4 1
XUSC1 1 2 12 DELAY
GUSC1 1 2 12 0 1
XUSC2 1 4 14 DELAY
GUSC2 4 1 14 0 1
XUSC3 3 2 32 DELAY
GUSC3 2 3 32 0 1
XUSC4 3 4 34 DELAY
GUSC4 3 4 34 0 1
.ENDS USCP
.SUBCKT USCP1 1 2 3 4
R1 1 3 5.5586
R2 2 4 5.5586
XUSC1 1 2 12 DELAY
GUSC1 1 2 12 0 0.1799
XUSC2 1 4 14 DELAY
GUSC2 4 1 14 0 .1799
XUSC3 3 2 32 DELAY
GUSC3 2 3 32 0 .1799
XUSC4 3 4 34 DELAY
GUSC4 3 4 34 0 .1799
.ENDS USCP1
.SUBCKT AMP 1 2 3 4
EODD 3 0 1 0 1E6
EVEN 4 0 2 0 1E6
.ENDS AMP
.AC LIN 100 10 3K
.PRINT AC V(5) VP(5) V(13)
VP(13) V(21) VP(21)
.PROBE
.END
Page 9.7-24
.SUBCKT NC7 1 2 3 4
RNC1 1 0 3.2018
XNC1 1 0 10 DELAY
GNC1 1 0 10 0 0.3123
XNC2 1 4 14 DELAY
GNC2 4 1 14 0 0.3123
XNC3 4 0 40 DELAY
GNC3 4 0 40 0 0.3123
RNC2 4 0 3.2018
.ENDS NC
.SUBCKT PC1 1 2 3 4
RPC1 2 4 11.0011
.ENDS PC1
.SUBCKT PC2 1 2 3 4
RPC1 2 4 4.8581
.ENDS PC2
.SUBCKT PC4 1 2 3 4
RPC1 2 4 6.7980
.ENDS PC4
.SUBCKT PC6 1 2 3 4
RPC1 2 4 3.2018
.ENDS PC6
.SUBCKT USCP 1 2 3 4
R1 1 3 1
R2 2 4 1
XUSC1 1 2 12 DELAY
GUSC1 1 2 12 0 1
XUSC2 1 4 14 DELAY
GUSC2 4 1 14 0 1
XUSC3 3 2 32 DELAY
GUSC3 2 3 32 0 1
XUSC4 3 4 34 DELAY
GUSC4 3 4 34 0 1
.ENDS USCP
.SUBCKT USCP1 1 2 3 4
R1 1 3 5.5586
R2 2 4 5.5586
XUSC1 1 2 12 DELAY
GUSC1 1 2 12 0 0.1799
XUSC2 1 4 14 DELAY
GUSC2 4 1 14 0 .1799
XUSC3 3 2 32 DELAY
GUSC3 2 3 32 0 .1799
XUSC4 3 4 34 DELAY
GUSC4 3 4 34 0 .1799
.ENDS USCP1
.SUBCKT AMP 1 2 3 4
EODD 3 0 1 0 1E6
EVEN 4 0 2 0 1E6
.ENDS AMP
.AC LIN 100 10 3K
.PRINT AC V(5) VP(5) V(13)
VP(13) V(21) VP(21)
.PROBE
.END
P.E. Allen - 2002
Page 9.7-25
S6
S7
S8
S9
S10
CL12
CL22
CL42
C12
CL52
C22
E1
E2
END;
OPTIONS;
NOLIST;
GRID;
END;
TIMING;
PERIOD 50E-6;
CLOCK CLK 1 (0 25/50);
END;
SUBCKT (1 100) STG1;
S1
(1 2)
CLK;
S2
(2 0)
#CLK;
S3
(3 4)
#CLK;
S4
(3 0)
CLK;
S5
(5 100)
#CLK;
S6
(5 0)
CLK;
CL11 (2 3)
0.0909;
CL21 (3 5)
0.0909;
E1
(100 0 0 4)1E6;
END;
(6 0)
#CLK;
(7 0)
CLK;
(7 8)
#CLK;
(300 9)
#CLK;
(9 0)
#CLK;
(2 3)
0.3123;
(3 9)
0.3123;
(4 300)
0.1799;
(4 5)
1;
(6 7)
0.3123;
(8 300)
1;
(5 0 0 4) 1E6;
(300 0 0 8)1E6
CL53
C23
E1
E2
END;
CIRCUIT;
X1
(1 100)
STG1;
X2
(100 200) STG3;
X3
(200 300) STG2;
V1
(2 0);
END;
ANALYZE SSS;
INFREQ 1 3000 LIN 150;
SET V1 AC 1.0 0.0;
PRINT vdb(100) vp(100);
PRINT vdb(200) vp(200);
PRINT vdb(300) vp(300);
PLOT vdb(300);
END;
(6 7)
0.2058;
(8 200)
1;
(5 0 0 4) 1E6;
(200 0 0 8)1E6
END;
Page 9.7-26
TBP(j)
1
TPB
One possible
filter realization
TSB
0
0
THP(j)
Transition
Region
1
TPB
1
TPB
TSB
A
D
0
0 SB1 PB1 PB2 SB2
(c.)
(rps)
Transition
Region
(rps)
SB PB
TBS(j)
One possible
filter realization
TSB
(rps) 0
0
PB SB
(a.)
Lower
Transition
Region
One possible
filter realization
(b.)
One possible
Lower
filter realization
Transition
A Region
C
Upper TransiTSB
B
D
tion Region
0
(rps)
0 PB1 SB1 SB2 PB2
(d.)
Practical magnitude responses of (a.) low pass, (b.) high pass, (c.) bandpass, and (d.)
bandstop filter.
We will use transformations from the normalized, low pass filter to the normalized
high pass, bandpass or bandstop to achieve other types of filters.
ECE 6414 - Analog Integrated Systems Design
Page 9.7-27
p1lnp2lnp3lnpNln
shn
THPn(shn) = 1
1
1
1
=
1
1
1
1
+p
+p
+p
+p
s
+
s
+
s
+
s
+
1lnshn 2lnshn 3ln shn Nln
shn
hn p1ln hn p2ln hn p3ln hn pNln
N
shn
= shn+p1hnshn+p2hnshn+p3hnshn+pNhn
Page 9.7-29
shn
shn
shn
= (shn+p1hn)(shn+p6hn)(shn+p2hn)(shn+p5hn)(shn+p3hn)(shn+p4hn)
2
2
2
shn
shn
shn
.
= 2
2
2
shn+0.5176shn+1shn+1.4141shn+1shn+1.9318shn+1
63C23
1
e 33C23
Vin(z)
53C23
1
2 C23
2
1
C13
23C13
V3(z)
Stage 2
e
V2(z)
62C22
52C22
C12
21C12
2
1
Stage 1
61C21
C22 2 32C22
2
1
Stage 3
Page 9.7-30
51C21
31C21 2 C21
2
1
C11
21C11
2
e
Vout(z)
2
1
Page 9.7-31
1
BW
where b = r .
sln' = r sln = bsln = bPB = sbn + sbn
6.) Solve for sbn in terms of sln' from the following quadratic equation.
2
sbn - sln' sbn + 1 = 0 sbn = sln' /2 sln' /22 - 1 .
ECE 6414 - Analog Integrated Systems Design
Page 9.7-32
TLPn(jln )
1
0
-1 0
Bandpass
Normalization
b s ln = BW s ln s 'ln
r
1
(a.)
0
-b 0 b
r ln (rps)
PB
TBPn(jbn )
-r
BW
0
0
(d.)
'ln (rps)
Normalized
2
s 'ln
low-pass to s 'ln
-1
normalized 2
2
bandpass
transformation
s bn
TPBn (jb )
BW
1
(b.)
Bandpass
Denormalization
sb b s bn = BW sbn
r
b (rps)
1
b
b
-1
0
0
(c.)
bn (rps)
Figure 9.7-10 - Illustration of the development of a bandpass filter from a low-pass filter.
(a.) Ideal normalized, low-pass filter. (b.) Normalization of (a.) for bandpass
transformation. (c.) Application of low-pass to bandpass transformation. (d.)
Denormalized bandpass filter.
ECE 6414 - Analog Integrated Systems Design
Page 9.7-33
4.) The normalized bandpass poles can be found from the normalized, low pass poles, pkln
using
jbn
pkln
pkln 2
'
p
jln
jbn
pkbn = 2
2 -1 .
For each pole of the low-pass filter, two poles
result for the bandpass filter.
Figure 9.7-11 - Illustration of how the
normalized, low-pass, complex conjugate
poles are transformed into two normalized,
bandpass, complex conjugate poles.
pkbn
p'jln
'ln
p'kln
= p'jln*
Low-pass Poles
Normalized by PB r
BW
bn
p*jbn
p*kbn
Normalized
Bandpass Poles
sbn
T
(
)
k
kon
Kk sbn
Qk
= s2bn+(2kbn)sbn+(2bn+2kbn) =
kon
s2bn + Qk sbn + 2kon
where j and k corresponds to the jth and kth low-pass poles which are a complex
conjugate pair, Kk is a gain constant, and
2bn+2kbn
kon =
and
Qk = 2bn
.
6.) Realize each second-order product with a bandpass switched capacitor biquad and
cascade in the order of increasing Q.
2
kbn
+2kbn
Page 9.7-35
Page 9.7-36
j'ln
j1
p1ln
p1ln
p1bn jbn
p5bn
j1
j0.8660
j1
p3bn
3 zeros
at j
p'1ln
p2ln
-1
ln
-0.5000
p2ln
p'2ln
-1
'ln
p'3ln
bn
-1
p 2bn
p3ln
-j0.8660
-j1
(a.)
p3ln
-j1
(b.)
p6bn
p4bn
-j1
(c.)
Pole locations for Ex. 9.7-8. (a.) Normalized low-pass poles. (b.) Bandpass
normalized low-pass poles. (c.) Normalized bandpass poles.
Page 9.7-37
s
10.0410 bn
=
1.0904
s2bn+10.0410sbn+1.09042
K2sbn
K2sbn
T2(sbn) = (s+p2bn)(s+p3bn) = (sbn+0.0457+j0.9159)(sbn+0.0457-j0.9159)
0.9170
s
10.0333 bn
.
=
0.9170
2
2
sbn+10.0333sbn+0.9159
and
K3sbn
K3sbn
T3(sbn) = (s+p5bn)(s+p6bn) = (sbn+0.1000+j0.9950)(sbn+0.1000-j0.9950)
1.0000
s
5.0000 bn
.
=
1.0000
s2bn+5.0000sbn+1.00002
ECE 6414 - Analog Integrated Systems Design
Chapter 9 Section 7 (5/2/04)
Page 9.7-39
1
2
e
Vin(z)
23C13
43C13
33C13
53C23
C13
1
2
C23
2
V3(z)
Stage 3
1
42C12
Stage 2
1
2
32C12
C12
52C22
C22
e
V2(z)
1
2
22C12
21C11
41C11
31C11
51C21
C11
1
2
C21
2
Vout(z)
Stage 1
ECE 6414 - Analog Integrated Systems Design
Page 9.7-40
L2n
LN,n
CN-1,n
C1n
C3n
+
Vout (sn )
-
(a.)
+
LN,n
CN-1,n
Vin (sn )
L3n
C2n
L1n
1
Vout (sn )
-
(b.)
Figure 9.7-12 - Singly-terminated, RLC prototype filters. (a.) N even. (b.) N odd.
ECE 6414 - Analog Integrated Systems Design
Page 9.7-41
R0n
L1n
C2n
I5
I3
I1
+
-
L3n
V2
C4n
+
-
L5n
V4
R 6n
Vout (s n )
-
Page 9.7-43
V (s)
3
V3:
V 4:
V (s)
V1(s)
3
sC
V
(s)
= 0
2n
2
R
R
V3(s)
Vout(s)
sC
V
(s)
4n
4
R
R6n = 0
and
Vout:
V4(s) -
sL5nVout(s)
- Vout = 0
R6n
R'
V1'(s) = sL1n Vin(s) - V2(s) - R' V1'(s)
1
V2(s) = sR'C2n [V1'(s) - V3'(s) ]
R'
V3'(s) = sL3n [V2(s) - V4(s)]
R'
1
V4(s) = sR'C4n [V3'(s) - R6n Vout(s)]
R6n
Vout(s) = sL5n [V4(s) - Vout(s)]
Note that each of these functions is the integration of voltage variables and is easily
realized using switched capacitor integrators.
Page 9.7-45
Page 9.7-46
Example 9.7-8 - Fifth-order, Low Pass, Switched Capacitor Filter using the Ladder
Approach
Design a ladder, switched capacitor realization for a Chebyshev filter approximation
to the filter specifications of TBP = -1dB, TSB = -25dB, fPB = 1kHz and fSB = 1.5 kHz. Give
a schematic and component value for the realization. Also simulate the realization and
compare to an ideal realization. Use a clock frequency of 20 kHz. Adjust your design so
that it does not suffer the -6dB loss in the pass band. (Note that this example should be
identical with Ex. 1.)
Solution
From previous work, we know that a 5th-order, Chebyshev approximation will
satisfy the specification. The corresponding low pass, RLC prototype filter is
L5n =2.1349 H L3n =3.0009 H L1n=2.1349 H
+
Vin (sn)
-
+
C 4n=
1.0911 F
C2n =
1.0911 F
Vout(s n)
-
Next, we must find the state equations and express them in the form of an integrator.
Fortunately, the above results can be directly used in this example.
Finally, use switched-capacitor integrators to realize each of the five state functions
and connect each of the realizations together.
ECE 6414 - Analog Integrated Systems Design
Page 9.7-47
R'
1
L1n: V1'(sn) = sn L1n Vin(sn) - V2(sn) - R' V1'(sn)
(1) Vin(ej) 1
2
2
21C1
This equation can be realized by the switched capacitor
+
integrator of Fig. 9.7-17 which has one noninverting input V2(ej) 2
1
and two inverting inputs. Therefore,
31C1
1
j)
V'
(e
1
2
(2)
V1(z) = z-1 11Vin(z) - 21zV2(z) - 31zV1(z)
1
1
However, since fPB < fc, replace z by 1 and z-1 by sT.
Figure 9.7-17 - Realization of V1'.
1
(5)
V2(z) = z-1 12V1 (z) - 22zV3(z) .
Simplifying as above gives
1
(6)
V2(sn) snTn 12V1 (sn) - 22V3(sn) .
Equating Eq. (4) to Eq. (6) yields the design of the capacitor ratios for the second
integrator as
PB
Tn
2000
12 = 22 = RC2n = RfcC2n = 120,0001.0911 = 0.2879.
The second integrator has a total capacitance of
1
Second integrator capacitance = 0.2879 + 2 = 5.47 units of capacitance.
ECE 6414 - Analog Integrated Systems Design
Page 9.7-49
(9)
V 3(sn) snTn 13V2(sn) - 23V4(sn) .
Equating Eq. (7) to Eq. (9) yields the capacitor ratios for the third integrator as
RTn RPB
12000
13 = 23 = L3n = fcL3n = 20,0003.0009 = 0.1047.
The third integrator has a total capacitance of
1
Third integrator capacitance = 0.1047 + 2 = 11.55 units of capacitance
V'
(e
)
1
2
C4n: V4(sn) = sn R'C4n [V3 ( sn)- R6nVout(sn)] (10)
2
2
Eq. (10) can be realized by the switched capacitor
24C4
+
Vout(ej)
integrator of Fig. 9.7-20 with one noninverting and
2
1
1
one inverting input. As before we write that
1
Page 9.7-51
Page 9.7-52
31C1
Vin(ej)
1
211C1
C1
2
21C1
1
2
C2
22C2
1
C3
13C3
2
23C3
1
2
C4
24C4
1
+
15C5
C5
2
25C5
1
V'3(ej)
14C2 1
1
2
V4(ej)
V'1(ej)
12C2 1
V2(ej)
1
2
1
2
Vout(ej)
Page 9.7-53
200
150
Magnitude (dB)
V1' Output
-10
V2 Output
-20
V3' Output
-30
-40
V4 Output
-50
Filter Output
Filter Phase
V2 Phase
100
V3' Phase
50
0
-50
V4 Phase
-100
V1' Phase
-150
-60
-200
-70
0
500
1000
3000
3500
500
1000
3000
3500
Comments:
Both passband and stopband specifications satisfied.
Some of the op amp outputs are exceeding 0 dB (need to voltage scale for maximum
dynamic range)
Page 9.7-54
XAMP3 13 14 9 10 AMP
**************************
*V4 STAGE
XNC41 9 10 25 26 NC2
XPC41 15 16 25 26 PC2
XUSC4 11 12 25 26 USCP
XAMP4 25 26 11 12 AMP
**************************
*VOUT STAGE
XNC51 11 12 17 18 NC1
XPC51 15 16 17 18 PC1
XUSC5 15 16 17 18 USCP
XAMP5 17 18 15 16 AMP
*************************
.SUBCKT DELAY 1 2 3
ED 4 0 1 2 1
TD 4 0 3 0 ZO=1K TD=25US
RDO 3 0 1K
.ENDS DELAY
.SUBCKT NC1 1 2 3 4
RNC1 1 0 6.7934
XNC1 1 0 10 DELAY
GNC1 1 0 10 0 .1472
XNC2 1 4 14 DELAY
GNC2 4 1 14 0 .1472
XNC3 4 0 40 DELAY
GNC3 4 0 40 0 .1472
RNC2 4 0 6.7934
.ENDS NC1
P.E. Allen - 2002
Page 9.7-55
GNC3 4 0 40 0 0.1047
RNC2 4 0 9.5521
.ENDS NC3
.SUBCKT NC4 1 2 3 4
RNC1 1 0 3.4730
XNC1 1 0 10 DELAY
GNC1 1 0 10 0 .2879
XNC2 1 4 14 DELAY
GNC2 4 1 14 0 .2879
XNC3 4 0 40 DELAY
GNC3 4 0 40 0 .1472
RNC2 4 0 6.7955
.ENDS NC4
.SUBCKT PC1 1 2 3 4
RPC1 2 4 6.7934
.ENDS PC1
.SUBCKT PC2 1 2 3 4
RPC1 2 4 3.4730
.ENDS PC2
.SUBCKT NC3 1 2 3 4
RNC1 1 0 9.5521
XNC1 1 0 10 DELAY
GNC1 1 0 10 0 0.1047
XNC2 1 4 14 DELAY
GNC2 4 1 14 0 0.1047
XNC3 4 0 40 DELAY
.SUBCKT PC3 1 2 3 4
RPC1 2 4 9.5521
.ENDS PC3
Page 9.7-56
CIRCUIT
OPTIONS;
NOLIST;
GRID;
END;
TIMING;
PERIOD 50E-6;
CLOCK CLK 1 (0 25/50);
END;
SUBCKT (1 4) NC (P:CAP);
S1
(1 2)
CLK;
S2
(2 0)
#CLK;
S3
(3 0)
CLK;
S4
(3 4)
#CLK;
C11
(2 3)
CAP;
END;
SUBCKT (1 4) PC (P:CAP1);
S1
(1 2)
#CLK;
S2
(2 0)
CLK;
S3
(3 0)
CLK;
S4
(3 4)
#CLK;
C21
(2 3)
CAP1;
END;
Page 9.7-57
ln
Lhn = 1
Cln
Design Procedure:
Normalized LowNormalized High1.) Identify the appropriate RLC
Pass
Network
Pass Network
prototype, low pass circuit to meet
the specifications.
2.) Transform each inductor and capacitor by the normalized, low pass to high pass
transformation.
3.) Choose the state variables and write the state functions.
4.) Realize the state functions using switched capacitor circuits.
The problem: The realizations are derivative circuits.
Page 9.7-58
C1
1
C2
Vin(z) C1
Vout(z)
Vin(z) C1
1
-
Vin(z) C1
Vout(z)
Vout(z)
(a.)
C2
C2
(b.)
(c.)
Figure 9.7-26 - (a.) Switched capacitor differentiatior circuit. (b.) Stray insensitive version of (a.).
(c.) Modification to keep op amp output from being discharged to ground during 1.
Transfer function:
1: (n-1)T < t < (n -0.5)T
e
o
vc1(n -0.5)T = vin(n -1)T
and
o
vc2(n -0.5)T = 0
C1
vin(n-1)
C2
e
vout
(n)
vin(n)
C1 e
C1 e
e
vout(n )T = - C2 vin(n )T + C2 vin(n -1)T
e
Vout(z)
C1 e
C1 e
C1
C1
e
e
Vout(z) = C V in(z) - z-1 C V in(z) = - C (1-z-1)Vin(z) Hee(z) =
=
e
C2 (1-z-1)
2
2
2
V in(z)
ECE 6414 - Analog Integrated Systems Design
Page 9.7-59
C2 2j sin(T/2)e
ejT/2
2
2
or
jTC1 sin(/2)
-j sin(/2)
= - C2 /2 (-j/2) = o /2 e-jT/2
= (Ideal)x(Mag. Error)x(Phase Error)
where o = C2/(C1T).
Frequency Response for C2 = 0.2C1:
|Hee(ejT)
5
10
1
0
Continuous
Time
Discrete
Time
0 o= 10c
c
2
Phase
0
-90
c
c
2
Continuous
Time
-180
-270
Discrete
Time
Page 9.7-61
Page 9.7-62
12C2
V1'
Vout
C2
2
1
V2
1
-
1 2
22C2
V2
13C3
V2
Vout
C3
2
1
1 2
23C3
Vout
1
+
Page 9.7-63
sn
r
s bn + 1
sbn
BW
Cbn =
r
Cln
BW
Design Procedure:
1.) Identify the appropriate RLC
Normalized
Lbn = BW 1
Low-Pass
r Cln
prototype, low pass circuit to meet
Network
Normalized Bandpass Network
the specifications.
2.) Transform each inductor and capacitor by the normalized, low pass to bandpass
transformation.
3.) Choose the state variables and write the state functions.
4.) Realize the state functions using switched capacitor circuits.
In this case, the state functions will be second-order, bandpass functions which can
be realized by switched-capacitor biquads.
ECE 6414 - Analog Integrated Systems Design
Page 9.7-64
R0n
Vin(sn)
C1bn =
r C
1ln
BW
L2bn =
C2bn =
r
L2ln
1/L2bn
BW
+
I2 C
=
V1 L1bn = 3bn
rC
3ln
1/C
1bn
BW
L4bn =
+
V3
-
r
L4ln C4bn =
BW
1/L4bn
I4
L3bn =
1/C3bn
R5n
+
Vout(sn)
-
Ex.6-B
The state equations for this
circuit can be written as illustrated below.
V1(s)
Z1bn
Vin(s) = I2(s) + Z1bn R0n + V1(s) V1(s) = R0n [Vin(s) - I2(s)R0n - V1(s)]
sL1bn(1/sC1bn)
s/C1bn
s/C1bn
= 2
where
Z1bn = sL1bn + (1/sC1bn) = 2
s + (1/L1bnC1bn) s +1
s/R0nC1bn
R0n
V1(s) =
Vin(s) - R V 2(s) - V 1(s)
2
s +1
(1)
P.E. Allen - 2002
Page 9.7-65
(2)
V 2 (s) Vout(s)
s/RC3bn
R
V3(s)=Z3bn(I2(s)-I4(s))=Z3bn R - R
V
(s)V
V3(s)=
(3)
2
out
R
2
5n
s +1
5n
and
I4(s) =Y4bn[V3(s)-Vout(s)] Vout(s) = R5nY4bn[V3(s)-Vout(s)]
sR5n/L4bn
[V3(s)-Vout(s)]
or Vout(s) =
s 2+1
How to realize? Consider the bandpass form of the low-Q and high-Q biquads:
2C1
2C1
1
2
C1
e
V1(z)
5C2
6C2
4C1
2
1
e
Vout(z)
4C2
Vin(z)
C2
1
2
(4)
5C2
C1 Ve (z)
1
Vin(z) 3C1
-
1
2
C2
2
Vout(z)
Page 9.7-66
s + (25/T2)
sn2+ (25/Tn2)
All the 2s and 5s will be given as: 25 = Tn2 = n2T 2 = r2/fc2 = (2)2(fr/fc)2
2fr 23x103
2 = |5| = f =
= 0.1473
c
128x105
Now all that is left is to design 4 for each stage (assuming R0n = R5n = R = 1).
Therefore, let
Page 9.7-67
=
=
=
41
Tn R0nC1bn
R0nC1bn fcrC1ln 128x1030.7658 = 0.03848
There will be one noninverting input (Vin) and two inverting inputs (V2 and V1).
2(0.1437)
2
capacitances = 0.03848 + 0.03848 + 3 = 62.44 units of capacitance
Stage 2
42
rBW
TnBW
R
2600
=
=
=
=
42 rL2ln fcrL2ln 128x1031.8478 = 0.01594
Tn L2bn
There will be one noninverting input (V1) and one inverting input (V3).
2
2(0.1437)
capacitances = 0.01594 + 0.01594 + 2 = 145.50 = units of capacitance
Stage 3
Same as stage 2. 43 = 0.01594
There will be one noninverting input (V2) and one inverting input (Vout).
capacitances = 145.50 units of capacitance
Stage 4
Same as stage 1 except capacitances = 61.44 units of capacitance. 44 = 0.03848.
There will be one noninverting input (V3) and one inverting input (Vout).
ECE 6414 - Analog Integrated Systems Design
Page 9.7-68
C1
1
2
1
2
C2
2
2
1
2C1
2,5
+
Ex.9.7-13B
Vin 1 41C21
2
41C21
21 =
51 =
0.1473
V'2
22 =
52 =
0.1473
143C23 43C23
2
23 =
53 =
0.1473
Vout
24 =
52 =
0.1473 2
1 2
1
2
V1 C
V3 C C
2
42 22 42C22 1
2
44 42 44 42 1
P.E. Allen - 2002
Page 9.7-69
Low pass
Prototype
RLC Ckt.
Write
State
Equations
Use SC
Integrators to
Design Each
State Equation
Low Pass
Switched
Capacitor
Filter
Normalized LP
to Normalized
High pass
Transformation
Choose
State
Variables
Write
State
Equations
Use SC
Differentiators
to Design Each
State Equation
High Pass
Switched
Capacitor
Filter
Normalized LP
to Normalized
Bandpass
Transformation
Choose
State
Variables
Write
State
Equations
Use SC
BP Ckts. to
Design Each
State Equation
Bandpass
Switched
Capacitor
Filter
Eliminate
L-cutsets
and
C-loops
Normalized LP
to Normalized
Bandpass
Transformation
Normalized LP
to Normalized
High pass
Transformation
Choose
State
Variables
Write
State
Equations
Use SC
BS Ckts. to
Design Each
State Equation
Bandstop
Switched
Capacitor
Filter
Page 9.7-70
Anti-Aliasing Filter
Baseband
c-PB
c+PB
2c-PB
2c+PB
0
c
2c
-PB 0 PB
Figure 9.7-28 - Spectrum of a discrete-time filter and a continuous-time
anti-aliasing filter.
The primary problem of aliasing is that there are undesired passbands that contribute
to the noise in the desired baseband.
Page 9.7-71
;;
;;
;;
fc-fsw
fc+fsw
f
fc-fB
-fB
fc+fB
0.5fc
fc
0 fsw fB
Figure 9.7-31 - Illustration of noise aliasing in switched capacitor circuits.
It can be shown that the aliasing enhances the baseband noise voltage spectral density by
a factor of 2fsw/fc. Therefore, the baseband noise voltage spectral density is
kT/C 2fsw
2kT
2
eBN = fsw x fc = fcC volts2/Hz
Multiplying this equation by 2fB gives the baseband noise voltage in volts(rms)2.
Therefore, the baseband noise voltage is
2kT
2kT 2fB 2kT /C
Page 9.7-73
Vin (s)
R1
R3
K=1
C4
(a.)
Vout (s)
K=1
Voltage
Amplifier
(b.)
Transfer function:
K
TLP(0) o2
R1R3C2C4
Vout(s)
=
=
1
o
1
1
K
1
Vin(s)
s 2 + s R3C4 + R1C2 + R3C2 - R3C4+ R1R3C2C4
s 2 + Q s + o2
We desire K = 1 in order to not influence the passband gain of the SCF. With K = 1,
1
R1R3C2C4
1/mn(RC)2
Vout(s)
=
=
1
s 2 + (1/RC)[(n+1)/n]s + 1/mn(RC)2
1
1
Vin(s)
s 2 + s R1C2 + R3C2 + R1R3C2C4
where R3 = nR1 = nR
and C4 = mC2 = mC.
ECE 6414 - Analog Integrated Systems Design
Chapter 9 Section 7 (5/2/04)
Design Equations for The Unity Gain, Sallen and Key Low Pass Filter
Equating Vout(s)/Vin(s) to the standard second-order low pass transfer function, we
get two design equations which are
1
o =
mnRC
m
1
Q = (n +1) n
The approach to designing the components of Fig. 9.7-29a is to select a value of m
compatible with standard capacitor values such that
1
m 4Q 2 .
Then, n, can be calculated from
1
1
n = 2mQ 2 - 1 2mQ 2 1-4mQ 2 .
This equation provides two values of n for any given Q and m. It can be shown that
these values are reciprocal. Thus, the use of either one produces the same element
spread.
Incidentally, these filters have excellent linearity because the op amp is in unity gain.
ECE 6414 - Analog Integrated Systems Design
Page 9.7-75
Page 9.7-76
R1=
1
2|TLP(0)|oQC
Vin
C4=
4Q2(1+|TLP(0)|)C
R2=
1
2oQC
C5=C
R3=
1
2(1+|TLP(0)|)oQC
Vout
This gain of this circuit in the passband is determined by the ratio of R2/R1.
Page 9.7-77
Page 9.8-1
SUMMARY
Switched capacitor circuits have reached maturity in CMOS technology.
The switched capacitor circuit concept was a pivotal step in the implementation of
analog signal processing circuits in CMOS technology.
The accuracy of the signal processing is proportional to the capacitor ratios.
Switched capacitor circuits have been developed for:
Amplification
Integration
Differentiation
Summation
Filtering
Comparison
Analog-digital conversion
Approaches to switched capacitor circuit design:
Oversampled approach clock frequency is much greater than the signal frequency
z-domain approach the specifications are converted to the z-domain and directly
realized. Such circuits can operate to within half of the clock frequency.
SPICE or SWITCAP permits frequency domain simulation of switched capacitor ckts.
Clock feedthrough and kT/C noise represent the lower limit of the dynamic range of
switched capacitor circuits.
ECE 6414 - Analog Integrated Systems Design
5/2/04
Page 10.0-1
10.0 - INTRODUCTION
Organization
;;;
;;;
;;;
;;;;
;;;
;;;
;;;;
;;;
;;;
;;;
;;;;;;
;;;
;;;;
;;;
;;;;;;;;;;
Chapter 10
D/A and A/D
Converters
Chapter 9
Switched Capacitor Circuits
Systems
Chapter 6
Simple CMOS &
BiCMOS OTA's
Chapter 7
High Performance
OTA's
Chapter 8
CMOS/BiCMOS
Comparators
Complex
Simple
Chapter 4
CMOS/BiCMOS
Subcircuits
Chapter 5
CMOS/BiCMOS
Amplifiers
Circuits
Chapter10
1
Chapter
Introduction
D/ to Analog CMOS Design
Devices
Chapter
Chapter11
2
Analog
CMOS
Technology
Systems
Chapter 3
CMOS
Modeling
Page 10.0-2
ANALOG
SIGNAL
(Speech,
sensors,
radar,
etc.)
DIGITAL
PROCESSOR
(Microprocessor)
PRE-PROCESSING
(Filtering and analog
to digital conversion)
POST-PROCESSING
(Digital to analog
conversion and
filtering)
ANALOG
OUTPUT
SIGNAL
CONTROL
ANALOG
A/D
DIGITAL
D/A
ANALOG
Page 10.0-3
DIGITALANALOG
CONVERTER
Reference
Filter
Amplifier
Analog
Output
Fig. 10.1-01
Page 10.0-4
VREF
b0
b1
b2
DigitalAnalog
Converter
bN-1
vOUT
b0
b1
b2
DigitalAnalog
Converter
Latch
Sample
and
Hold
VOUT*
bN-1
Clock
Asynchronous
Synchronous
Fig. 10.1-02
Page 10.0-5
Voltage
Reference
VREF
DVREF
Scaling
Network
Output
Amplifier
vOUT =
KDVREF
Binary Switches
b0 b1 b2
bN-1
Figure 10.1-3
Page 10.1-1
0.875
Infinite Resolution
Characteristic
0.750
0.625
1 LSB
0.500
Vertical Shifted
Characteristic
0.375
0.250
0.125
0.000
000
001
010
011
100
101
Digital Input Code
110
111
Fig. 10.1-4
Page 10.1-2
Definitions
Resolution of the DAC is equal to the number of bits in the applied digital input word.
The full scale (FS):
FS = Analog output when all bits are 1 - Analog output all bits are 0
VREF
1
FS = (VREF - 2N ) - 0 = VREF1 - 2N
000
001
010
011
100
101
110
Digital
Input
111
Code
Fig. 10.1-5
Page 10.1-3
More Definitions
Dynamic Range (DR) of a DAC is the ratio of the FSR to the smallest difference that
can be resolved (i.e. an LSB)
FSR
FSR
DR = LSB change = (FSR/2N) = 2N
or in terms of decibels
DR(dB) = 6.02N (dB)
Signal-to-noise ratio (SNR) for the DAC is the ratio of the full scale value to the rms
value of the quantization noise.
T
rms(quantization noise) =
1
LSB FSR
2 - 0.5 2dt =
LSB
= N
T
T
12
2 12
0
vOUT(rms)
(FSR/ 12 2N)
Maximum SNR (SNRmax) for a sinusoid is defined as
vOUTmax(rms)
FSR/(2 2)
6 2N
=
= 2
SNRmax =
(FSR/ 12 2N) FSR/( 12 2N)
or in terms of decibels
N
62
SNRmax(dB) = 20log10 2 = 10 log10(6)+20 log10(2N)-20 log10(2) = 1.76 + 6.02N dB
SNR =
Page 10.1-5
7/8
Gain
Error
6/8
Actual
Characteristic
5/8
7/8
6/8
5/8
4/8
Offset
Error
2/8
Infinite
Resolution
Characteristic
3/8
2/8
Ideal 3-bit
Resolution
Characteristic
1/8
0
4/8
Infinite
Resolution
Characteristic
3/8
Ideal 3-bit
Resolution
Characteristic
1/8
Page 10.1-6
where Vcx is the actual voltage change on a bit-to-bit basis and Vs is the ideal LSB
change of (VFSR/2N)
8
8
Infinite Resolution Characteristic
7
Example of a 3-bit DAC:
8
6
8
5
8
4
8
3
8
2
8
1
8
0
8 000
-1 LSB INL
A
-1.5 LSB DNL
Ideal 3-bit Characteristic
Actual 3-bit Characteristic
001
010
011 100
101
Digital Input Code
110
111
Fig. 10.1-7
Page 10.1-7
13.16
12/16
-2 LSB DNL
11/16
-1.5 LSB
INL
10/16
9/16
8/16
-2 LSB DNL
7/16
6/16
5/16
4/16
3/16
2/16
1/16
0/16
b0
b1
b2
b3
0
0
0
0
0
0
0
1
0
0
1
0
0
0
1
1
0
1
0
0
0 0 0 1 1
1 1 1 0 0
0 1 1 0 0
1 0 1 0 1
Digital Input Code
1
0
1
0
1
0
1
1
1
1
0
0
1
1
0
1
1
1
1
0
1
1
1
1
Fig. 10.1-8
Page 10.1-9
Page 10.1-10
H = GB R +R or GB C +C
2
2
1
1
To avoid errors in DACs (and ADCs), vout(t) must be within 0.5LSB of the final value by
the end of the conversion time.
Multiple-pole response:
Typically the response is underdamped like the following (see Appendix C of text).
vOUT(t)
Upper Tolerance
Final Value +
vIN
vOUT
Final Value
Final Value -
Lower Tolerance
Settling Time
0
CMOS Analog Circuit Design
Ts
t
Fig. 6.1-7
Page 10.1-11
H = C +C GB = (2)(0.5)(106) = 3.141x106
1
2
and ACL = 1. Assume that the ideal output is equal to VREF. Therefore the value of the
output voltage which is 0.5LSB of VREF is
1
1 - 2N+1 = 1 - e-H T
or
2N+1 = eH T
Solving for T gives
N+1
N+1
9
T = H ln(2) = 0.693 H = 3.141 0.693 = 1.986s
Page 10.1-12
TESTING OF DACs
Input-Output Test
Test setup:
Digital
Word
Input
(N+2 bits)
N-bit
DAC
under
test
ADC
ADC with
Output
Digital
Vout
more resolution
Subtractor
than DAC
(N+2 bits)
(N+2 bits)
Digital
Error
Output
(N+2 bits)
Fig. 10.1-9
Comments:
Sweep the digital input word from 000...0 to 111...1.
The ADC should have more resolution by at least 2 bits and be more accurate than the
errors of the DAC
INL will show up in the output as the presence of 1s in any bit.
If there is a 1 in the Nth bit, the INL is greater than 0.5LSB
DNL will show up as a change between each successive digital error output.
The bits which are greater than N in the digital error output can be used to resolve the
errors to less than 0.5LSB
CMOS Analog Circuit Design
Page 10.1-13
Spectral Test
Test setup:
1
0
0
0
1
0
0
0
1
0
0
1
0 1 1 1
|Vout(j)|
Vout
1
1
1
1
Noise floor
due to nonlinearities
VREF
Comments:
fsig
Digital input pattern is selected to
Digital
N-bit V
have a fundamental frequency which Pattern
out Distortion
Spectral
DAC
has a magnitude of at least 6N dB
Analyzer
Generator
Output
under
above its harmonics.
(N bits)
test
Length of the digital sequence
Fig. 10.1-10
determines the spectral purity of the
Clock
fundamental frequency.
All nonlinearities of the DAC (i.e. INL and DNL) will cause harmonics of the
fundamental frequency
The THD can be used to determine the SNR dB range between the magnitude of the
fundamental and the THD. This SNR should be at least 6N dB to have an INL of less than
0.5LSB for an ENOB of N-bits.
Note that the noise contribution of VREF must be less than the noise floor due to
nonlinearities.
If the period of the digital pattern is increased, the frequency dependence of INL can be
measured.
CMOS Analog Circuit Design
Page 10.2-1
Serial
Charge
Current
Voltage
Charge
Fast
Fig. 10.2-1
Page 10.2-2
VREF
RF
I2
IN-1
vOUT
Fig. 10.2-2
Page 10.2-3
S0
S1
I0
RMSB
I1
2R
SN-1
S2
I2
4R
IN-1
2N-1R
RLSB
vOUT
-
IO
Fig. 10.2-3
Comments:
1.) RF can be used to scale the gain of the DAC. If RF = KR/2, then
b
bN-1
b1 b2
bN-1
-KRb0 b1 b2
0
vOUT = -RFIO = 2 R + 2R + 4R ++ 2N-1RVREF vOUT = -K 2 + 4 + 8 ++ 2N V REF
where bi is 1 if switch Si is connected toVREF or 0 if switch Si is connected to ground.
RMSB
R
1
2.) Component spread value = RLSB = 2N-1R = 2N-1
3.) Attributes:
Insensitive to parasitics fast
Large component spread value
Trimming required for large values of N
Nonmonotonic
CMOS Analog Circuit Design
Page 10.2-4
I0 2R
I1 2R
S0
S1
S2
VREF
2R
I2
IN-1
SN-1
RF = KR
+
IO
vOUT
Fig. 10.2-4
8I
4I
2I
R
4I
R
2I
2R
2R
I
2R
I
2R
Fig. 10.2-4(2R-R)
Attributes:
Not sensitive to parasitics (currents through the resistors never change as Si is varied)
Small element spread. Resistors made from same unit (2R consist of two in series or R
consists of two in parallel)
Not monotonic
CMOS Analog Circuit Design
Page 10.2-5
S0
N-1
b0 2 I
SN-3
bN-3
SN-2
SN-1
4I
I
2I
bN-2 bN-1
Transistor
Array
A2
+
V
A
-
R2
A1
vOUT
VA
-
Fig. 10.2-5
Operation:
vOUT = R2(bN-1I + bN-22I + bN-34I + + b02N-1I)
VREF
If I = IREF = N ,
2 R2
then
b0
b1 b2
bN-3 bN-2 bN-1
Attributes:
Fast (no floating nodes) and not monotonic
Accuracy of MSB greater than LSBs
CMOS Analog Circuit Design
Page 10.2-6
Voltage
Scaling
Network
V2
V3
Decoder
Logic
vOUT
V2N
Fig. 10.2-6
Operation:
Creates all possible values of the analog output then uses a decoding network to
determine which voltage to select based on the digital input word.
Page 10.2-7
b2 b2
Input = 101
b1 b1
b0 b0
VREF
vOUT
Attributes:
Guaranteed
monotonic
Compatible with
CMOS
technology
Large area if N is
large
Sensitive to
parasitics
Requires a buffer
Large current can
flow through the
resistor string.
vOUT
7VREF
8
6VREF
8
5VREF
8
4VREF
8
3VREF
8
2VREF
8
VREF
8
0
000 001 010 011 100 101 110 111
(a.)
Figure 10.2-7 - (a.) Implementation of a 3-bit voltage scaling DAC. (b.) Input-output
characteristics of Fig. 10.2-7(a.)
CMOS Analog Circuit Design
Page 10.2-8
b2
b1
b0
3-to-8 Decoder
vOUT
R/2
Fig. 10.2-8
- R 2n
=
Ri+1
R
Therefore,
i+1
R V REF
VREF R
2n-1(R+R)VREF
=
2n-2 Vi
INL = 2n-1(R+R) + 2n-1(R-R) - 2 = 2R VREF
R 2n
2n-1
Therefore,
2n-1
n
V
R
2
REF
R
R2n
INL=2n 2R VREF=2n-1 R 2n =2n-1 R LSBs
DNL
=
R LSBs
2n
Fig. 10.2-085
Page 10.2-10
Page 10.2-11
VREF
Charge
Scaling
Network
vOUT
Fig. 10.2-9
C1
+
VREF
C2
Vout
Fig. 10.2-9b
Page 10.2-12
C
2
+
C
4
C
2N-2
C
2N-1
C
2N-1
vOUT
Operation:
1
S0
S1
S2
SN-2
SN-1
Terminating
1.) All switches
2
Capacitor
2
2
2
2
connected to ground
V
Fig. 10.2-10
REF
during 1.
2.) Switch Si closes to VREF if bi = 1 or to ground if bi = 0.
Equating the charge in the capacitors gives,
b1C b2C
bN-1C
VREFCeq = VREF b0C + 2 + 22 + ... + 2N1 = Ctot vOUT = 2C vOUT
which gives
vOUT = [b02-1 + b12-2 + b22-3 + ... + bN-12-N]VREF
Equivalent circuit of the binary-weighted, charge
Ceq.
scaling DAC is:
+
Attributes:
vOUT
2C - Ceq.
VREF
Accurate
Sensitive to parasitics
Fig. 10.2-11
Not monotonic
Charge feedthrough occurs at turn on of switches
CMOS Analog Circuit Design
Chapter 10 Section 2 (5/2/04)
Page 10.2-14
Page 10.2-15
C+C
C+C
and
1
C)
(C(C-C) -Cterm
2n-1(C-C)
vOUT(0111...) = (C+C)+(C-C) VREF = (C+C)+(C-C) V REF
C-C
2
= 2C 1 - 2nV REF
vOUT(1000...) - vOUT(0111...)
C+C
C-C
C
2
-2n
1- -1 = (2n-1)
n
-1
LSBs
=
2
n
LSB
C LSBs
2C
2C 2
Therefore,
C
DNL = (2n - 1) C LSBs
Page 10.2-16
Example 10.2-2 - DNL and INL of a Binary Weighted Capacitor Array DAC
If the tolerance of the capacitors in an 8-bit, binary weighted, charge scaling DAC are
0.5%, find the worst case INL and DNL.
Solution
For the worst case INL, we get from above that
INL = (27)(0.005) = 0.64 LSBs
For the worst case DNL, we can write that
DNL = (28-1)(0.005) = 1.275 LSBs
Page 10.2-17
C
INL = 2N-1
= N .
C
C
2
2
From the data presented in Chapter 2, it is reasonable to assume that the relative
accuracy of the capacitor ratios will decrease with the number of bits. Let us assume a
unit capacitor of 50 m by 50 m and a relative accuracy of approximately 0.1%.
Solving for N in the above equation gives approximately 10 bits. However, the 0.1%
figure corresponds to ratios of 16:1 or 4 bits. In order to get a solution, we estimate the
relative accuracy of capacitor ratios as
C
C 0.001 + 0.0001N
Using this approximate relationship, a 9-bit digital-analog converter should be
realizable.
Page 10.2-18
1
VREF
CF = 2NC/K
2N-2C
2N-3C
2C
vOUT
-
Fig. 10.2-12
Attributes:
No floating nodes which implies insensitive to parasitics and fast
No terminating capacitor required
With the above configuration, charge feedthrough will be Verror -(COL/2CN)V
Can totally eliminate parasitics with parasitic-insensitive switched capacitor circuitry
but not the charge feedthrough
Page 10.2-19
Advantage
Fast, insensitive to
switch parasitics
Monotonic, equal
resistors
Charge
Scaling
Disadvantage
Large element spread,
nonmonotonic
Large area, sensitive
to parasitic
capacitance
Large element spread,
nonmonotonic
Page 10.3-1
Page 10.3-2
m+k-bit DAC.
bits
+
DAC
VREF
k-LSB
bits
k-bit
LSB
DAC
2m
Fig. 10.3-1
b
b1
bm-1
bm+1
bm+k-1
1 bm
0
vOUT = 2 + 4 + + 2m VREF + 2m 2 + 4 + + 2k V REF
b b1
bm-1
bm bm+1
bm+k-1
vOUT = 2 + 4 + + 2m + 2m+1 + 2m+2 + + 2m+k V REF
Accuracy?
VREF 2n
Weighting factor of the i-th bit = 2i+1 2n = 2n-i-1 LSBs
0.5 LSB 1 100
Accuracy of the i-th bit = 2n-i-1 LSB = 2n-i = 2n-i %
Page 10.3-3
Page 10.3-4
vOUT
vOUT(act.)
- vOUT
0/32
2/32
4/32
6/32
8/32
10/32
12/32
14/32
16/32
18/32
20/32
22/32
24/32
26/32
28/32
30/32
0/32
1/32
2/32
3/32
0/32
1/32
2/32
3/32
0/32
1/32
2/32
3/32
0/32
1/32
2/32
3/32
Change in
vOUT(act) 2/32
1/32
1/32
1/32
-3/32
1/32
1/32
1/32
-3/32
1/32
1/32
1/32
-3/32
1/32
1/32
1/32
Page 10.3-5
The worst case value of x occurs when both b2 and b3 are 1. Therefore, we get
3
1
1
x4 32 x 24 .
The scaling factor, x, can be expressed as
1 1
6
1
x x = 4 24 = 24 24
Therefore, the tolerance required for the scaling factor x is 5/24 to 7/24. This corresponds
to an accuracy of 16.7% which is less than the 25% (100%/2k) because of the
influence of the LSB bits. It can be shown that the INL will be equal to 0.5LSB or less
(see Problem 10.3-6 of text).
CMOS Analog Circuit Design
Page 10.3-6
m-bit
MSB
DAC
vOUT
VREF/2m
k-LSB
bits
k-bit
LSB
DAC
Fig. 10.3-2
bm
b0 b1
bm-1
bm+1
bm+k-1VREF
vOUT = 2 + 4 + + 2m VREF + 2 + 4 + + 2k 2m
b0
b1
bm-1
bm
bm+1
bm+k-1
Page 10.3-7
b6
b5
b4
15R
i2
io
b3
b2
MSB
LSB
I
16
Current
I Divider
2
I
4
I
8
I
16
LSB subDAC
I
4
I
8
vOUT
b0
b1
RF
I
2
Fig. 10.3-3
MSB subDAC
b b1 b2 b3 1 b4 b5 b6 b7
vOUT = R I 2 + 4 + 8 + 16 + 16 2 + 4 + 8 + 16
0
F
Page 10.3-8
+
C
2
C
4
C
8
b7
2
VREF
MSB Array
Cs
b6
b5
2
Scaling
Capacitor
C
8
LSB Array
b4
2
C
8
C
4
b3
b2
2
C
2
b1
vOUT
b0
2
2
Fig. 10.3-4
Page 10.3-9
2C
vOUT
V1
V2
16 b0 b1 b2 b3
V1 = 15/8 b0 + 15/8 b1 + 15/8 b2 + 15/8 b3VREF = 15 2 + 4 + 8 + 16 VREF
V2 = 2 b4 + 2 b5 + 2 b6 + 2 b7VREF = 2 + 4 + 8 + 16 VREF
Combining the elements of the simplified equivalent circuit above gives
1 +15 8
2 2 15
15+1515
16
15
1
vOUT=1 15 8 V1+1 15 8 V2 = 15+1515+16V1+15+1515+16V2 = 16V1+16V2
2 + 2 +15 2 + 2 +15
7 bV
b
b1 b2 b3 b4 b5 b6
b7
i REF
0
Page 10.3-10
Charge Amplifier DAC Using Two Binary Weighted Charge Amplifier SubDACs
Implementation:
C/8
b4
b4
b5
b5
+
VREF -
b6
C
1
C/2
C/4
b7
b7
b0
b1
2C
1
b6
b0
1
vO1
C/8
1
LSB Array
b1
A1
VREF
b2
C
C/2
2C
C/4
b3
A2
vOUT
-
b2
b3
C/8
1
MSB Array
Fig. 10.3-6
Attributes:
MSB subDAC is not dependent upon the accuracy of the scaling factor for the LSB
subDAC.
Insensitive to parasitics, fast
Limited to op amp dynamics
No ICMR problems with the op amp
CMOS Analog Circuit Design
Page 10.3-11
R2 R3
R2m-2 R2m-1
vOUT
SF
Ck =
2k-1C
Bus A
Sk,A
R2m
Sk,B
VREF
m-to-2m Decoder B
Ck-1 =
2k-2C
C1
=C
C2
=2C
Sk-1,A
S2A
S1A
Sk-1,B
S2B
S1B
Bus B
SF
m-MSB bits
Operation:
1.) Switches SF and S1B through Sk,B discharge all capacitors.
2.) Decoders A and B connect Bus A and Bus B to the top and bottom, respectively, of
the appropriate resistor as determined by the m-bits.
3.) The charge scaling subDAC divides the voltage across this resistor by capacitive
division determined by the k-bits.
Attributes:
MSBs are monotonic but the accuracy is poor
Accuracy of LSBs is good
CMOS Analog Circuit Design
Page 10.3-12
Voltage Scaling MSB SubDAC And Charge Scaling LSB SubDAC - Continued
Equivalent circuit of the voltage scaling (MSB) and charge scaling (LSB) DAC:
Ck =
Bus A 2k-1C
Ck-1 =
2k-2C
C2
=2C
C1
=C
Bus A
Ceq.
C
vOUT
2-mVREF
Sk,A
Sk-1,A
S2A
S1A
Sk,B
Sk,B
S2B
S1B
Bus B
V'REF
2-mVREF
2kC - Ceq.
Bus B
v'OUT
vOUT
V'REF
Fig. 10.3-8
where,
VREF = V
b b1
bm-2 bm-1
2 + 22 + + 2m-1 + 2m
0
REF 1
and
b
VREF bm bm+1
bm+k bm+k-1
bm+1
bm+k bm+k-1
m
vOUT = 2m 2 + 22 + + 2k-1 + 2k = VREF2m+1 + 2m+2 + + 2m+k-1 + 2m+k
Adding VREF and vOUT gives the DAC output voltage as
b0
b1
bm-2 bm-1 bm
bm+1
bm+k bm+k-1
Page 10.3-13
C2 =
2m-1C
Cm-1
=21C
Cm
=C
R1
Cm
=C
R2
vOUT
VREF
S1,A
S1,B
S2,A
S2,B
Sm-2A
Sm-1A
Sm-2B
Sm-1B
vk
kto2k
Decoder
R3
VREF
R2k-2
R2k-1
Fig. 10.3-9A
k-LSB bits
k-bit,
R2k LSB
voltage
scaling
subDAC
b
b b1
bm-2 bm-1
vk
bm+1
bm+k bm+k-1
m
vOUT = 2 + 22 ++ 2m-1 + 2m VREF + 2m where vk = 21 + 22 ++ 2k-1 + 2k V REF
b
b1
bm-2 bm-1 bm bm+1
bm+k bm+k-1
0
vOUT =21 + 22 + + 2m-1 + 2m + 2m+1 + 2m+2 + + 2m+k-1 + 2m+k VREF
Attributes:
MSBs have good accuracy
LSBs are monotonic, have poor accuracy - require trimming for good accuracy
0
1
R
C
INL = INL(R) + INL(C) = 2n-1 R + 2k-1 C LSBs
R
C
and
DNL = DNL(R) + DNL(C) = 2k R + (2k-1) C LSBs
MSB Charge Scaling SubDAC and LSB Voltage Scaling SubDAC
R
C
INL = INL(R) + INL(C) = 2k-1 R + 2n-1 C LSBs
R
C
and
DNL = DNL(R) + DNL(C) = R + (2n-1) C LSBs
CMOS Analog Circuit Design
Page 10.3-15
Example 10.3-3 - Design of a DAC using Voltage Scaling for MBSs and Charge
Scaling for LSBs
Consider a 12-bit DAC that uses voltage scaling for the MSBs charge scaling for the
LSBs. To minimize the capacitor element spread and the number of resistors, choose m =
5 and k = 7. Find the tolerances necessary for the resistors and capacitors to give an INL
and DNL equal to or less than 2 LSB and 1 LSB, respectively.
Solution
Substituting n = 12 and k = 7 into the previous equations gives
R
C
R
C
and
1 = 27 R + (27-1) C
2 = 211 R + 26 C
Solving these two equations simultaneously gives
C
C
25-2
=
=
0.0154
11
6
5
C 2 -2 -2
C = 1.54%
and
R 2 - 26(0.0154)
R
=
=
0.0005
11
R
R = 0.05%
2
We see that the capacitor tolerance will be easy to meet but that the resistor
tolerance will require resistor trimming to meet the 0.05% requirement. Because of the
2n-1 multiplying R/R in the relationship, it will not do any good to try different values of
m and k. This realization will consist of 32 equal value resistors and 7 binary-weighted
capacitors with an element spread of 64.
CMOS Analog Circuit Design
Chapter 10 Section 3 (5/2/04)
Example 10.3-4 - Design of a DAC using Charge Scaling for MBSs and Voltage
Scaling for LSBs
Consider a 12-bit DAC that uses charge scaling for the MSBs voltage scaling for the
LSBs. To minimize the capacitor element spread and the number of resistors, choose m =
7 and k = 5. Find the tolerances necessary for the resistors and capacitors to give an INL
and DNL equal to or less than 2 LSB and 1 LSB, respectively.
Solution
Substituting the values of this example into the relationships developed on a previous
slide, we get
R
C
R
C
and
1 = R + (212-1) C
2 = 24 R + 211 C
Solving these two equations simultaneously gives
C
C
R 3
R
24-2
=
=
0.000221
=
0.0221%
and
=
0.0968
C 216-211-24
C
R 25-1
R = 9.68%
For this example, the resistor tolerance is easy to meet but the capacitor tolerance will
be difficult. To achieve accurate capacitor tolerances, we should decrease the value of m
and increase the value of k to achieve a smaller capacitor value spread and thereby
enhance the tolerance of the capacitors. If we choose m = 5 and k = 7, the capacitor
tolerance remains about the same but the resistor tolerance becomes 2.36% which is still
reasonable. The largest to smallest capacitor ratio is 16 rather than 64 which will help to
meet the capacitor tolerance requirements.
CMOS Analog Circuit Design
Page 10.3-17
Page 10.4-1
S2
VREF
S3
C1
C2
S4
vC2
Fig. 10.4-1
Operation:
Switch S1 is the redistribution switch that parallels C1 and C2 sharing their charge
Switch S2 precharges C1 to VREF if the ith bit, bi, is a 1
Switch S3 discharges C1 to zero if the ith bit, bi, is a 0
Switch S4 is used at the beginning of the conversion process to initially discharge C2
Conversion always begins with the LSB bit and goes to the MSB bit.
CMOS Analog Circuit Design
Page 10.4-2
vC1/VREF
vC2/VREF
Page 10.4-3
Pipeline DAC
Implementation:
0
1/2
bN-1
= 1
z-1
1/2
bN-2 = 1
VREF
z-1
1/2
z-1
vOUT
b0 = 1
Fig. 10.4-3
Page 10.4-4
A
B
+1
+1
Sample
and
hold
1
2
vOUT
FIG. 10.4-4
Page 10.4-6
Page 10.4-7
Figure
10.2-3
Primary Disadvantage
Large element spread, nonmonotonic
Current-scaling, active
devices
Voltage-scaling
10.2-5
10.2-7
Charge-scaling,
binary weighted capacitors
Binary weighted, charge
amplifier
Current-scaling subDACs
using current division
Charge-scaling subDACs
using charge division
Binary weighted charge
amplifier subDACs
Voltage-scaling (MSBs),
charge-scaling (LSBs)
Charge-scaling (MSBs),
voltage-scaling (LSBs)
Serial, charge redistribution
Pipeline, algorithmic
10.4-4
10.2-4
Primary Advantage
Fast, insensitive to parasitic capacitance
Page 10.5-1
Page 10.5-2
x(t)
Prefilter
Sample/Hold
Quantizer
Digital
Processor
y(kTN)
Encoder
Fig.10.5-1
Prefilter - Avoids the aliasing of high frequency signals back into the baseband of the
ADC
Sample-and-hold - Maintains the input analog signal constant during conversion
Quantizer - Finds the subrange that corresponds to the sampled analog input
Encoder - Encoding of the digital bits corresponding to the subrange
Page 10.5-3
fS+fB 2fS-fB
fS
fS
2
Use of an antialiasing filter to avoid aliasing.
-fB
2fS
2fS+fB
2fS
Antialiasing
Filter
-fB
fB f S
2
fS
Fig. 10.5-2
Page 10.5-4
Nyquist ADCs
Oversampled ADCs
Integrating (Serial)
Successive
Approximation1-bit
Pipeline Algorithmic
Flash Multiple-bit
Pipeline Folding and
interpolating
Page 10.5-5
000
001
010
011
100
101
110
111
0000000
0000001
0000011
0000111
0001111
0011111
0111111
1111111
Gray
000
001
011
010
110
111
101
100
Twos
Complement
000
111
110
101
100
011
010
001
Page 10.5-6
Input-Output Characteristics
Ideal input-output characteristics of a 3-bit ADC
111
Infinite Resolution
Characteristic
110
101
1 LSB
100
Ideal 3-bit
Characteristic
011
010
1 LSB
001
Quantization
Noise LSBs
000
1.0
0.5
0.0
-0.5
vin
VREF
0
8
2
3
4
5
6
7
8
8
8
8
8
8
8
8
Analog Input Value Normalized to VREF
Figure 10.5-3 Ideal input-output characteristics of a 3-bit ADC.
CMOS Analog Circuit Design
1
8
Page 10.5-7
Definitions
The dynamic range, signal-to-noise ratio (SNR), and the effective number of bits
(ENOB) of the ADC are the same as for the DAC
Resolution of the ADC is the smallest analog change that distinguishable by an ADC.
Quantization Noise is the 0.5LSB uncertainty between the infinite resolution
characteristic and the actual characteristic.
Offset Error is the difference between the ideal finite resolution characteristic and
actual finite resolution characteristic
Gain Error is the
Gain Error = 1.5LSBs
111
111
difference between
110
110
the ideal finite
Ideal
Ideal
101 Characteristic
resolution charact101 Characteristic
eristic and actual
100
100
Actual
finite resolution
011
011
Characteristic
characteristic
010
010
Offset = 1.5 LSBs
measured at full001
001
scale input. This
vin
vin
000
000
0 1 2 3 4 5 6 7 8 VREF
0 1 2 3 4 5 6 7 8 VREF
difference is
8 8 8 8 8 8 8 8 8
8 8 8 8 8 8 8 8 8
proportional to the
(a.)
(b.)
analog input
Figure 10.5-4 - (a.) Example of offset error for a 3-bit ADC. (b.) Example of gain
voltage.
error for a 3-bit ADC.
P.E. Allen - 2004
Page 10.5-8
Page 10.5-9
111
Ideal
Characteristic
110
INL =
+1LSB
101
Actual
Characteristic
100
DNL =
+1LSB
INL =
-1LSB
011
010
DNL =
0 LSB
001
vin
000
0
1
2
3
4
5
6
7
8 VREF
8
8
8
8
8
8
8
8
8
Example of INL and DNL for a 3-bit ADC.) Fig.10.5-5
CMOS Analog Circuit Design
Page 10.5-10
Monotonicity
A monotonic ADC has all vertical jumps positive. Note that monotonicity can only be
detected by DNL.
Example of a nonmonotonic ADC:
111
Actual
Characteristic
110
101
100
DNL =
-2 LSB
011
Ideal
Characteristic
010
001
000
0
8
1
8
2
8
3
8
4
8
5
8
6
8
7
8
8
8
vin
VREF
Fig. 10.5-6L
Page 10.5-11
Characteristic
DNL =
+1 LSB
001
000
0
8
1
8
2
8
3
8
4
8
5
8
6
8
7
8
8
8
vin
VREF
Fig. 10.5-6DL
DYNAMIC CHARACTERISTICS
The dynamic characteristics of ADCs are influenced by:
Comparators
Sample-hold circuits
Circuit parasitics
Logic propagation delay
Page 10.5-13
Comparator
The comparator is the quantizing unit of ADCs.
Open-loop model:
VOS
V1
Vi Av(s)Vi
Ri
Comparator
Ro
Vo
V2
Fig.10.5-7
Nonideal aspects:
Input offset voltage, VOS (a static characteristic)
Propagation time delay
- Bandwidth (linear)
Av(0)
Av(0)
= s + 1
Av(s) = s
c
c+1
- Slew rate (nonlinear)
CV
T = I
(I is constant)
CMOS Analog Circuit Design
Page 10.5-14
-t
/
p
c
p
c
= Av(0) [1- e ] vin(min) = Av(0) [1- e ] Av(0)
2
Therefore, tp is
tp(max) = c ln(2) = 0.693c
If vin is greater than vin(min), i.e. vin = kvin(min), then
vout
2k
tp = c ln2k -1
vin > vin(min)
VOH
Illustration of these results:
v
V +V
vin
+
-
out
OH
vin = vin(min)
VOL
0 t t (max)
0 p p
OL
2
t
Fig.10.5-8
Page 10.5-15
Page 10.5-17
Sample-and-Hold Circuit
Waveforms of a sample-and-hold circuit:
Amplitude
Definitions:
Acquisition time (ta) = time required
to acquire the analog voltage
Settling time (ts) = time required to
settle to the final held voltage to within
an accuracy tolerance
vin*(t)
Sample
Hold
ta
ts
Output of S/H
valid for ADC
conversion
vin*(t)
vin(t)
vin(t)
Time
Fig.10.5-9
1
Maximum sample rate = fsample(max) = Tsample
Tsample = ta + ts
S/H Command
Hold
Other consideratons:
Aperture time= the time required for the sampling switch to open after the S/H
command is initiated
Aperture jitter = variation in the aperture time due to clock variations and noise
Types of S/H circuits:
No feedback - faster, less accurate
Feedback - slower, more accurate
CMOS Analog Circuit Design
Page 10.5-18
CH
vin(t), vout(t)
vout(t)
Amplitude
vin(t)
Switch
Closed
(sample)
vout(t)
vin(t)
Switch
Open
(hold)
Time
vin(t), vout(t)
Switch
Closed
(sample)
Fig.10.5-10
Attributes:
Fast, open-loop
Requires current from the input to charge CH
DC voltage offset of the op amp and the charge feedthrough of the switch will create dc
errors
Page 10.5-19
Settling Time
Assume the op amp has a dominant pole at -a and a second pole at -GB.
GB2
The unity-gain response can be approximated as,
A(s) s2 + GBs + GB2
4
3
-0.5GBt
sin
The resulting step response is,
vout(t) = 1 3e
4 GBt +
Defining the error as the difference between the final normalized value and vout(t), gives,
4
Error(t) = = 1 - vout(t) = 3 e-0.5GBt
In most ADCs, the error is equal to 0.5LSB. Since the voltage is normalized,
4
2N
3
Solving for the time, ts, required to settle with 0.5LSB from the above equation gives
2 4
1
ts = GB ln 2N = GB [1.3863N + 1.6740]
3
Thus as the resolution of the ADC increases, the settling time for any unity-gain buffer
amplifiers will increase. For example, if we are using the open-loop, buffered S/H circuit
in a 10 bit ADC, the amount of time required for the unity-gain buffer with a GB of 1MHz
to settle to within 10 bit accuracy is 2.473s.
1
2N+1
4
-0.5GBts
3e
e0.5GBts =
Page 10.5-20
vin(t)
1d
2
vout (t)
vin(t)
+
vout (t)
1d
1d
C
Differential switched-capacitor S/H
Fig.10.5-11
Page 10.5-21
VDD
VDD
Clock
IB
IB
D1
vin(t)
D3
Clock
D2
D4
vout(t)
IB
D5
D1
D2
D3
D4
vin(t)
CH
vout(t)
CH
IB
M1 M2
Hold
VSS
D6
Sample
2IB
VSS
Practical implementation of the diode bridge S/H.
Attributes:
Fig.10.5-12
Fast
Clock feedthrough is signal independent
Sample uncertainty caused by the finite slope of the clocks is minimized
During the hold phase the feedthrough from input to hold node is minimized because of
D5 and D6
CMOS Analog Circuit Design
Page 10.5-22
vin(t)
vout(t)
vin(t)
CH
Closed-loop S/H circuit. 1 is the sample
phase and 2 is the hold phase.
+
-
CH
-
vout(t)
An improved version.
Fig.10.5-13
Attributes:
Accurate
First circuit has signal-dependent feedthrough
Slower because of the op amp feedback loop
Page 10.5-23
CH
2
vin(t) 1d CH
1
-
vout(t)
1d
2d
CH
1d
2d
2
1
+
-+
+-
vin(t)
vout(t)
2d 1
1d
CH
2
2
CH
CH 1d
2d
A differential version that avoids
large changes at the op amp output
Fig.10.5-14
Attributes:
Accurate
Signal-dependent feedthrough eliminated by a delayed clock
Differential circuit keeps the output of the op amps constant during the 1 phase
avoiding slew rate limits
CMOS Analog Circuit Design
Page 10.5-24
VDD
IB
iin
1
iout
2
CH
Fig.10.5-15
Attributes:
Fast
Requires current in and out
Good for low voltage implementations
Page 10.5-25
vin
Clock
Analog-Digital
Converter
Digital
Output
vin(to)
to
Aperature Jitter = t
Figure10.5-14 - Illustration of aperature jitter in an ADC.
dvin
V = dt t = Vpt .
The rms value of this noise is given as
dv
V p t
in
V(rms) = dt t =
.
2
The aperature jitter can lead to a limitation in the desired dynamic range of an ADC. For
example, if the aperature jitter of the clock is 100ps, and the input signal is a full scale
peak-to-peak sinusoid at 1MHz, the rms value of noise due to this aperature jitter is
111V(rms) if the value of VREF = 1V.
Page 10.5-26
TESTING OF ADCs
Input-Output Test for an ADC
Test Setup:
Vin
N-bit
ADC
under
test
Digital
Word
Output
(N bits)
DAC with
Vin'
more resolution
than ADC
+
(N+2 bits)
Qn =
Vin-Vin'
Fig.10.5-17
Page 10.5-27
2.0 LSB
1.5 LSB
+2LSB
INL
1.0 LSB
0.5 LSB
-2LSB
DNL
0.0 LSB
-0.5 LSB
+2LSB
DNL
-1.0 LSB
-1.5 LSB
-2.0 LSB
0
16
-2LSB
INL
1
16
2
16
3
16
4
16
5
6
7
8
9 10 11 12 13 14 15 16
16 16 16 16 16 16 16 16 16 16 16 16
Fig.10.5-18
Analog Input Normalized to VREF
Page 10.5-28
fsig
t
VREF
Harmonic
Vin
free
sinusoid
1
0
0
0
1
0
0
0
1
0
0
1
1
1
1
1
0 1 1 1
Vout(DAC)
|Vout(j)|
DR
t
VREF
fsig
N-bit
Distortion
DAC Vout(DAC)
or
with N+2
Spectrum
bits
Analyzer
resolution
N-bit
ADC
under
test
Noise floor
due to nonlinearities
Spectral
Output
Fig. 10.5-19A
Clock
Comments:
Input sinusoid must have less distortion that the required dynamic range
DAC must have more accuracy than the ADC
CMOS Analog Circuit Design
Page 10.5-29
Pure
Sinusoidal
Input, fin
AnalogDigital
Converter
Fast RAM
Buffer
FFT
Postprocessor
Frequency
Spectrum
Fig.10.5-19B
Comments:
Stores the digital output codes of the ADC in a RAM buffer
After the measurement, a postprocessor uses the FFT to analyze the quantization noise
and distortion components
Need to use a window to eliminate measurement errors (Raised Cosine or 4-term
Blackmann-Harris are often used)
Requires a spectrally pure sinusoid
Page 10.5-30
Number of
Occurances
Sinusoidal Input
Triangular Input
Mid
Output
Full Code
Scale
Comments:
Scale
Fig.10.5-20
Emphasizes
the time spent at a given level and can show DNL and missing codes
DNL
H(i)/Nt
Width of the bin as a fraction of full scale
DNL(i) = Ratio of the bin width to the ideal bin width -1 = P(i) -1
where
H(i) = number of counts in the ith bin
Nt = total number of samples
P(i) = ratio of the bin width to the ideal bin width
INL is found from the cumulative bin widths
CMOS Analog Circuit Design
Page 10.5-31
DNL
Histogram
or
Code Test
Yes (spikes)
No
Yes (Peaks in
distribution)
Yes (Offset of
distribution average)
FFT Test
Yes (Elevated
noise floor)
Yes (Elevated
noise floor)
Yes (Harmonics in
the baseband)
Yes (Elevated
noise floor)
Yes (Elevated
noise floor)
No
Sinewave
Curve
Fit Test
Yes
Beat
Frequency
Test
Yes
Yes
Yes
Yes
Yes
Yes
No
Yes
No
No
No
No
Yes (Measures
analog bandwidth)
No
No
No
No
P.E. Allen - 2004
Page 10.5-32
Page 10.6-1
vin*
VREF
nT
Interval
Counter
nT
+
-
vT
Ramp
Generator
vT
*
v
in
Reset
nT
Attributes:
Clock
f =1/T
Simplicity of operation
Subject to error in the ramp generator
Long conversion time 2NT
t
Output
Counter
Output
nN
Fig.10.6-1
Page 10.6-2
Dual-Slope ADC
Block diagram:
vin*
-VREF
Waveforms:
1
Positive
Integrator
vint
+
-
Vth
Digital
Control
Carry
Output
Counter
vin
VREF+Vth
NREFT
t1 = NREFT
vin'''
vin''' > vin'' > vin'.
vin''
vin'
Binary
Output
Vth
0
Fig.10.6-2
Reset
t0(start)
t2'
t
t2''
t2'''
t2= NoutT
Fig.10.6-3
Operation:
1.) Initially vint = 0 and vin is sampled and held (vIN* > 0).
2.) Reset the positive integrator by integrating a positive voltage until vint (0) = Vth.
3.) Integrate vin* for NREF clock cycles to get,
NREFT
Page 10.7-1
0.75VREF
0.50VREF
0.25VREF
0
t
6 T
Fig.10.7-2
Page 10.7-2
Vin*
VREF
+
Comparator
Digital-Analog
Converter
Output
Register
Output
Shift
Register
Conditional
Gates
Clock
Fig.10.7-1
R. Hnatek, A User's Handbook of D/A and A/D Converters, JohnWiley and Sons, Inc., New York, NY, 1976.
CMOS Analog Circuit Design
Page 10.7-3
vOA
CompAnalog
arator
In
+ MSB
Analog
Switch
1
Analog
Switch
2
Analog
Switch
3
VREF
LSB
Analog
Switch
5
Analog
Switch
4
LSB
MSB
Gate
0 1
FF1
R RD S
0 1
FF2
R RD S
0 1
FF3
R RD S
0 1
FF4
R RD S
0 1
FF5
R RD S
G1
G2
G3
G4
G5
-1
Delay
Delay
Clock pulses
SR1
Start pulse
SR2
SR3
SR4
SR5
Shift Register
Fig.10.7-3
Page 10.7-5
Page 10.7-6
VDAC
S2 precharge
Serial
DAC
(Fig.
10.4-1)
VREF
Start
Clock
Data storage
register
DAC control
register
S3 discharge
S1 charge share
S4 reset
Sequence and
control logic
Fig.10.7-7
Conversion Sequence:
Digitalanalog
Conversion
Number
1
2
3
.
.
N
d2
d3
...
aN
aN-1
.
.
a2
aN
.
.
a3
d1
...
Comparat
or
dN-1
dN
.
.
aN-1
.
.
aN
Output
aN
aN-1
aN-2
.
.
a1
Number
of
Charging
Steps
2
4
6
.
.
2N
Page 10.7-7
vc1/VREF
1.00
1xxx
11xx
111x
1101
13/16
0.75
0.50
0.25
0.00
0 1 2 0 1 2 3 4 0 1 2 3 4 5 6 0 1 2 3 4 5 6 7 8
1 bit
2 bits
3 bits
t/T
4 bits
vc2/VREF
1.00
0.75
0.50
0.25
0.00
0 1 2 0 1 2 3 4 0 1 2 3 4 5 6 0 1 2 3 4 5 6 7 8
t/T
Fig.10.7-8
Page 10.7-8
LSB
MSB
+ -
Vin*
z-1
VREF
+ -
z-1
Stage 1
+ -
Vi-1
z-1
Vi
z-1
i-th stage
Stage 2
Stage N
Fig.10.7-9
Operation:
Each stage muliplies its input by 2 and adds or subtracts VREF depending upon the sign
of the input.
Vi/VREF
i-th stage,
1.0
Vi = 2Vi-1 - biVREF
where bi is given as
+1 if Vi-1>0
bi = -1 if Vi-1<0
bi+1
=+1
bi+1
=-1
[bi,bi+1]
-1.0
-0.5
-1.0
bi = -1
[0,0]
[0,1]
0.5
bi = +1
[1,0]
[1,1]
1.0
Vi-1/VREF
Fig.10.7-10
Page 10.7-9
1 1 1 1
Vanalog = 5 2 4 + 8 + 16
= 5(0.4375) = 2.1875
0.8
0.6
Stage 3
0.4
0.2
0
Stage 2
-0.2
-0.4
Stage 1
-0.6
-0.8
-1
-1
0.4 0.6
0.8
1
P.E. Allen - 2004
Page 10.7-10
MSB-1
SR
SR
SR
SR
SR
SR
i-th Bit
SR
LSB
+ -
Vin*
VREF
z-1
Stage 1
+ -
z-1
Stage 2
+ -
Vi-1
z-1
i-th stage
Vi
z-1
Stage N
Fig.10.7-9B
Page 10.7-11
N-1 N
N
i=1
where Ai (Aj) is the actual gain of 2 for the i-th ( j-th) stage.
Errors include:
1.) Gain errors - x2 amplifier or summing junctions
2.) Offset errors - comparator or summing junctions
i-th stage including errors,
Vi = AiVi-1 + VOSi - biAsiVREF
= +1 if Vi-1>VOCi
bi = = -1 if Vi-1<VOCi
where
Ai is the gain of 2 amplifier for the i-th stage
VOSi is the system offset errors of the i-th stage
Asi is the gain of 1 summer for the i-th stage
VOCi is the comparator offset voltage of the i-th stage
;;;;
;; ;
Page 10.7-12
2Ai
1Vi/VREF -1
-1
2VOSi
Vo/VREF
1
2VOSi
1 Vi/VREF -1
-1
1 Vi/VREF
-1
2VOCi
Page 10.7-13
The most robust values of Vin will be near -VREF , 0 and +VREF. or
when each stage output is furthest from the comparator threshold, 0V.
Page 10.7-15
Voi
x2
Sample
and
Hold
+1
+VREF
+1
Vb
Sample
and
Hold
-VREF
+
-
Va
x2
+1
Vo
VREF
Vo ="1"
+1
Vo ="0"
S1
-VREF
Vin*
Different version of iterative algorithm ADC implementation
Fig. 10.7-13
Operation:
1.) Sample the input by connecting switch S1 to Vin*.
2.) Multiply Vin * by 2.
3.) If Va > VREF, set the corresponding bit = 1 and subtract VREF from Va.
If Va < VREF, set the corresponding bit = 0 and add zero to Va.
4.) Repeat until all N bits have been converted.
CMOS Analog Circuit Design
Page 10.7-16
2.0
1.6
1.2
0.8
0.4
0.0
;;;;
;;;;
;;;;
;;;;
;;;;
;;;; ;;;;
REF
REF
2.0
1.6
1.2
0.8
0.4
t/T
0.0
t/T
Fig. 10.7-14.
Page 10.7-17
m-bit subDAC
C1
C2
C3
Cm-1
Cm
Cm
To Successive
Approximation
Register
VREF
Register
m control lines
Successive
Approximation
Register
m+2-bit
k-bit Calibration
subDAC DAC
k-bits
Adder
Data Register
Control
Logic
m+k-bits
Data Output
V1
V2
Fig.10.7-15
Comments:
Self-calibration can be accomplished during a calibration cycle or at start-up
In the above scheme, the LSB bits are not calibrated
Calibration can extend the resolution to 2-4 bits more that without calibration
CMOS Analog Circuit Design
Page 10.7-18
C1
Vx1
C1
VREF
Connection of C1 to VREF.
Fig.10.7-16
C -C
1
1
3.) The result will be Vx1 = C + C VREF. If C1 = C1 , then Vx1 = 0.
1
1
4.) If Vx1 0, then the comparator output will be either high or low. Depending on the
comparator output, the calibration circuitry makes a correction through the calibration
DAC until the comparator output changes. At this point the MSB is calibrated and the
MSB correction voltage, V1 is stored.
5.) Proceed to the next MSB with C1 out of the array and repeat for C2 and C2 . Store
the correction voltage, V2, in the data register.
6.) Repeat for C3 with C1 and C2 out of the array. Continue until all of the capacitors of
the MSB DAC have been corrected.
Note that for any combination of MSB bits the calibration circuit adds the correct
combined correction voltage during normal operation.
CMOS Analog Circuit Design
Page 10.7-19
Advantage
High resolution
Disadvantage
Requires considerable
digital control
circuitry
Simple
Slow
Accuracy depends on
input
Requires other digital
circuitry
Successive approximation ADCs also can be calibrated extending their resolution 2-4 bits
more than without calibration.
Page 10.8-2
Comments:
Fast, in the first phase of the clock the
analog input is sampled and applied to the
comparators. In the second phase, the
digital encoding network determines the
correct output digital word.
Number of comparator required is 2N-1
Can put a sample-hold at the input or can
used clocked comparators
Typical sampling frequencies can be as
high as 400MHz for 6-bits in sub-micron
CMOS technology.
R
0.750VREF
R
0.625VREF
R
0.500VREF
R
0.375VREF
R
0.250VREF
R
0.125VREF
+
+
+
+
+
+
+
1
1
0
0
0
2N-1
to N
encoder
Output
Digital
Word
101
0
0
R
Fig.10.8-1
Page 10.8-4
Page 10.8-5
Comparators
CMOS Analog Circuit Design
10
11
12
13
14
15
16
Fig.10.8-2B
Page 10.8-6
V *-V
Ao
in
Ri
L -1Vout(s) = (s/103) + 1 s vout(t) = Ao(1 - e-103t)(Vin* - VRi).
The worst case occurs when
Vin*-VRi = 0.5VLSB = VREF/27 = 5/128
Page 10.8-8
VREF V
DD
+A2
V2a
V2b
VREF
2
VDD
+A1
V1
V1b
V2
V2a
V2b
V2c
-8
-7
-6
V2c R
V1a
R
Volts
VDD
Vth
V2
8 to 3
encoder
-5
-4
3-bit
digital
output
1 2 3 4 5 6 7 8 Comparator
Threshold
Vth
-3
V1c
V1b
V1a
V1
-2
V1c R
-1
Fig.10.8-3
0.5VREF
VREF
Vin
Fig.10.8-4
Comments:
Loading of the input is reduced from 8 comparators to two amplifiers.
The comparators no longer need a large ICMR
V1 and V2, are interpolated through the resistor string and applied to the comparators.
Because of the amplification of the input amplifiers and a single threshold, the
comparators can be simple and are often replaced by a latch.
If the dots in Fig. 10.8-4 are not equally spaced, INL and DNL will result.
CMOS Analog Circuit Design
Page 10.8-9
VREF V
DD
+A2
Vth
V2
V2a
V2b
R
VDD
+A1
VREF
2
R R/4
V1
V1b
-8
-7
V2c R R/4
V1a
R
R R
-6
-5
R R/4
-4
8 to 3
encoder
3-bit
digital
output
-3
V1c R R/4
R
-2
-1
Fig.10.8-6
Page 10.8-10
Coarse
Quantizer
Vin
Folding
Preprocessor
Fine
Quantizer
n1
bits
n2
bits
Encoding
Logic
n1+n2
bits
Digital
Output
Fig.10.8-7
Operation:
The input is split into two or more parallel paths.
First path uses a coarse quantizer to quantize the signal into 2n1 values
The second path maps all of the 2n1 subranges onto a single subrange and applies this
analog signal to a fine quantizer of 2n2 subranges.
Thus, the total number of comparators is 2n1-1 + 2n2-1 compared with 2n1+n2-1 for a
parallel ADC.
I.e., if n1 = 2 and n2 = 4, the folding ADC requires 3 + 15 = 18 compared with 63
comparators.
CMOS Analog Circuit Design
Page 10.8-11
Folding Preprocessor
Illustration:
FS/2
2n1
subranges
-FS/2
FS/F
2n2
subranges
Fig.10.8-8
Comments:
Folding is done simultaneously or in parallel so that only one clock cycle is needed for
conversion.
Folding will tend to increase the bandwidth of the analog input by a factor of F.
Folding can reduce the power consumption and require less chip area.
CMOS Analog Circuit Design
Page 10.8-12
After Analog
Preprocessing
VREF
No
Folding
32
n1 = 2
n2 = 3
VREF
4
Folding
0
MSBs =
00
Analog Input
01
10
VREF
11
Fig.10.8-9
Problems:
The sharp discontinuities of the folder are difficult to implement at high speeds.
Fine quantizer must work at voltages ranging from 0 to VREF/4 (subranging).
Page 10.8-13
Vin
VREF
Vin
VREF
Fig.10.8-10.
Page 10.8-14
Folder 2
+
-
Vin
Folder 7
5-bit
digital
output
Decoder
Folder 1
+
-
2 bits
+
Comparators
3 bits
LSBs
Fig.10.8-11
Comments:
Number of comparators is 7 for the fine quantizer and 3 for the course quantizer
The zero crossings of the folders must be equally spaced to avoid linearity errors
The number of folders can be reduced and the comparators simplified by use of
interpolation
CMOS Analog Circuit Design
Folding Circuits
Implementation of
a times 4 folder:
Page 10.8-15
VDD
+VREF
R
RL
Folding
Outputs
+V
- out
RL
V8
I
V7
V2
Comments:
Horizontal shifting is
achieved by modifying
the topmost and
bottom resistors of the
resistor string
Folding and
interpolation ADCs
offer the most
resolution at high
speeds (8 bits at
200MHz)
CMOS Analog Circuit Design
V1
V2
V7
V8
V1
R Vin
Vout
+IRL
-IRL
V1
V2
V3
V4
V5
V6
V7
V8 VREF
Vin
Fig. 10.8-12A
Page 10.8-16
Page 10.8-18
Page 10.8-20
Clock
Vi-1
S/H
VREF
VREF
k-bit
ADC
k-bit
DAC
Vi
Av =2k
i-th stage
k-bits
Fig.10.8-13
b0
bk-2 bk-1
b1
Residue voltage = Vi-1 - 2 + 22 + + 2k-1 + 2k VREF
Page 10.8-21
Stage 2
Stage 3
111
110
101
VREF
100
2
011
010
001
000
0
Clock 1
Digital output = 011
111
110
101
100
011
010
001
000
Clock 2
111
111
110
101
100
011
010
001
000
Clock 3
001
Voltage
VREF
MSB
LSB
Time
Fig.10.8-14
Page 10.8-22
Comments:
Resolution of the
comparators for the
following stages increases
but fortunately, the
tolerance of each stage
decreases by 2k for every
additional stage.
Removes the frequency
limitation of the amplifier
Voltage
10
0.5000VREF
0.4375VREF
0.3750VREF
0.3125VREF
0.2500VREF
01
11
10
01
00
00
Time
Clock 1
Digital output word = 01
Clock 2
10
Fig.10.8-15
Page 10.8-23
0
R
ON
1
R
OFF
0
R
OFF
0
0
Fig.10.8-16
CMOS Analog Circuit Design
Page 10.8-24
Page 10.8-25
Vin(2)
VREF
0
1/16
2/16
3/16
0
1/16
2/16
3/16
0
1/16
2/16
3/16
0
1/16
2/16
3/16
b2 b3 Ideal Ouput
b0 b1 b2 b2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
1
0
0
1
1
0
0
1
1
0
0
1
1
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
0
0
0
0
1
1
1
1
0
0
0
0
1
1
1
1
0
0
1
1
0
0
1
1
0
0
1
1
0
0
1
1
0
1
0
1
0
1
0
1
0
1
0
1
0
1
0
1
1111
1110
1101
1100
1011
Vin(1) b
0
VREF
0 0
1/16 0
2/16 0
3/16 0
4/16 0
5/16 0
6/16 0
7/16 0
8/16 1
9/16 1
10/16 1
11/16 1
12/16 1
13/16 1
14/16 1
15/16 1
1010
INL=0LSB
1001
-DNL=0LSB
1000
0111
0110
-INL=2LSB
0101
0100
0011
+DNL=2LSB
0010
0001
0000
0
16
1
16
2
16
3
16
4
16
5
16
6
7
8
9 10
16 16 16 16 16
Analog Input Voltage
11
16
12 13 14 15 16
16 16 16 16 16
Comparing the actual digital output word with the ideal output word gives the following:
+INL = 0LSB, -INL = 0111-0101 = -2LSB, +DNL = (1000-0101) - 1LSB = +2LSB, and DNL = (0101-0100) - 1LSB = 0LSB.
CMOS Analog Circuit Design
Page 10.8-26
b
b1
0
The input to the second ADC is vin(2) = k vin(1) - 2 + 4 . If vin(2) is vin(2) when k = 4,
then the |vin(2) - vin(2)| must be less than 1/8 or the LSB bits will be in error.
b
b
b1
b1 1
0
0
Therefore, vin(2) - vin(2) = k vin(1) - k 2 + 4 - 4 vin(1) + 4 2 + 4 8
b
b
b
b1
b1
b1 1
0
0
0
If k = 4+k, then 4 vin(1) + k vin(1) - 4 2 + 4 - k 2 + 4 - 4 vin(1) + 4 2 + 4 8
b b
b
b
1
1
1
0
or kvin(1)- 2 + 4 8 where the largest value of vin(1)- 2 + 4 is 1/4 for any vin(1).
k 1 1
k 1
Therefore,
4 8 k 1/2. The tolerance of k is k = 24 = 8 12.5%
CMOS Analog Circuit Design
Page 10.8-27
Vin*
MSB
ADC
MSBs
DAC
Vr1
LSB
ADC
Fig.10.8-21
Increment
by 1
LSBs
Vr2
DAC
Features:
Requires only 2n/2-1 comparators
LSBs decoded using 31 preset charge redistribution capacitor arrays
Reference voltages used in the LSBs are generated by the MSB ADC
No op amps are used
CMOS Analog Circuit Design
Page 10.8-28
VREF
R32
R31
Operation:
1.) Sample Vin* on
each 32C
capacitance
autozeroing the
comparators
R30
Vr2
Analog
MUX
Vr1
Vin*
Vin*
VRi
Ri
Vin*
+ -
Vin*
32C
Vin*
+ 32C
+
+
Vin*
+ -
32C
Latch
Bank
and
Binary
Encoder
MSB
Output
VRi-Vin*
R2
R1
MSBs
Vin*
Vin*
+ -
32C
+
Fig.10.8-22
Page 10.8-29
and
Binary
Encoder
Output
Vr1
Vr2
Vin*
C 16C 8C
4C
Vr2
Vr1
2C
Switches
set to
"Code"
Fig.10.8-23
Comments:
Requires two full clock cycles
Reuses the comparators
Accuracy limited by resistor string and
its dynamic loading
Accuracy also limited by the capacitor
array
Comparator is a 3-stage, low-gain,
wide-bandwidth, using internal
autozeroing
P.E. Allen - 2004
Page 10.8-30
For an input of
0.4VREF the
output should
be 0110.
Stage 1
VREF
Stage 2
VREF
11
0.75VREF
10
0.50VREF
11
111
110
101
100
011
010
001
000
0.75VREF
11
10
01
00
Error
10
0.50VREF
101
100
11
10
01
00
-01
-10
Vin* = 0.4VREF
Comments:
Vin* = 0.4VREF
01
01
Add a cor0.25VREF
0.25VREF
recting bit to
00
00
the following
stage to
0
Time
0
Time
Clock 1 Clock 2
Clock 1 Clock 2
correct for
00
10
Digital output word = 10
Digital output word = 01
Fig.10.8-19
errors in the
previous stage.
The subranging or amplification of the next stage does not include the correcting bit.
Correction can be done after all stages of the pipeline ADC have converted or after each
individual stage.
CMOS Analog Circuit Design
Page 10.8-31
S/H
2N1
S/H
-
ADC-N1 bit
2N2
S/H
DAC-N1 bit
ADC-N2 bit
DAC-N2 bit
2N3
ADC-N3 bit
DAC-N3 bit
V REF
3 bits
N3 bit REG
3 bits
N2 bit REG
N2 bit REG
3 bits
N1 bit REG
N1 bit REG
4 bits
S/H
N1 bit REG
Comments:
The ADC of the first stage uses 16 equal capacitors instead of 4 binary weighted for
more accuracy
One bit of the last three stages is used for error correction.
CMOS Analog Circuit Design
Page 10.8-32
ADC
4 bits
12 bits
Fig. 11-30
Page 10.8-33
S/H
T2
S/H
Digital
word
out
Vin
TM
S/H
T1
T2
TM T1+TC T2+TC
TM+TC
Comments:
Can get the same throughput with less chip area
If M = N, then a digital word is converted at every clock cycle
Multiplexer and timing become challenges at high speeds
CMOS Analog Circuit Design
Fig.10.8-20
Page 10.8-34
Page 10.8-36
Primary Disadvantage
Area is large if N > 6
Requires accurate
interpolation
Fast
Bandwidth increases if
no S/H used
Increased number of bits Slower than flash
Small area with large
throughput
Typical Performance:
6-8 bits
500-2000 Msamples/sec.
The ENOB at the Nyquist frequency is typically 1-2 bits less that the ENOB at low
frequencies.
Power is approximately 0.3 to 1W
CMOS Analog Circuit Design
Chapter 10 Section 9 (5/2/04)
Page 10.9-2
ADC0
Digital
Multiplexer
ADC1
x(t)
x(nT)
ADCM-1
fs
fs/M
fs=1/T FigEx1-01
Problems
- Offset mismatch of interleaved channels
- Gain mismatch of interleaved channels
- Aperture error between channels
Solutions - Digital-background calibration is used to overcome the offset, gain, and
sample-time errors between channels. Digital-background calibration is a tradeoff in
overhead versus enhanced performance.
Only two channels are given in this example to illustrate the method.
S. Jamal, D. Fu, C.J. Chang, P. Hurst and S. Lewis, A 10-b, 120-Msample/s Time-Interleaved Analog-to-Digital Converter With Digital
Background Calibration, IEEE J. of Solid-State Circuits, vol. 37, no. 12, Dec. 2002, pp. 1618-1627.
CMOS Analog Circuit Design
P.E. Allen - 2004
Chapter 10 Section 9 (5/2/04)
Page 10.9-3
Amplifier
Chopping SHA
C[m]
VOS
ADC
S[m]
+
a[m]
V[m]
o
C[m]
Accumulator
Fig. Ex1-01
C[m] is a pseudo-random binary signal = 1 where m is the discrete time index. C[m] is
white with zero mean.
VOS models the input-referred offset of the sample-hold amplifier and the ADC.
How does it work?
1.) The chopped analog signal is sampled and digitized by the ADC producing S[m].
2.) A variable offset, V[m], is subtracted from S[m] and the result multiplied by C[m] to
produce the channel output, a[m].
3.) Since the analog signal has been chopped twice, it is unaffected by the chopping.
4.) Because of the chopping process, the only dc component in the accumulator is due to
differences between the analog offset from the SHA and the ADC in the channel and
the accumulator output V[m].
5.) In steady state, the negative feedback forces the average of the accumulator input to be
zero. o controls the bandwidth of the notch, the speed of convergence, and the
variance of V[m] at convergence.
CMOS Analog Circuit Design
Page 10.9-4
Gain Calibration
ADC output spectrum for two time-interleaved channels with a sinusoidal input at fo and
a gain mismatch between channels.
Y(f) = Output
YC(f) = ChoppedOutput
Input
Image
f
Fig Ex1-03 o
f
fi = s -fo
2
Chopped Input
Chopped Image
fs
2
f
fi = s -fo
2
fo
fs
2
The image amplitude is proportional to the gain mismatch between the two channels.
How does it work?
1.) The ADC output is chopped by multiplying it by a signal that alternates at the channel
sampling rate.
2.) This multiplication causes the image to shift to fo and the input to fi.
3.) Next, the output and chopped output signals are multiplied in the time domain.
4.) The result has a dc component that is proportional to the gain mismatch between the
two channels.
Page 10.9-5
1[n]
a2[m]
2
G[n]
z-1
Sample
Time
2[n] Calibration
(-1)n
1+z-2
y[n]
yc[n]
Gain-error Detector
Accumulator
Fig. Ex1-04
Operation:
1.) a1[m] and a2[m] are upsampled by a factor of two by inserting zero samples to
produce a signal at the ADC sample rate of fs.
2.) a2[m] is delayed so that 1[n]2[n] = 0.
3.) At the input of the gain-error detector the signal is passed through a short FIR filter.
4.) The output of the FIR filter is y[n] which is chopped to produce yc[n].
5.) The image at 0.5fs -fo turns out to be in phase with the input at fo.
6.) Therefore, multiplying y[n] with yc[n] produces a signal with a dc component that is
proportional to the gain mismatch.
7.) g scales the product of y[n] and yc[n] to produce the accumulator output.
8.) The feedback on the lower ADC channel causes the accumulator input to converge to
zero in the steady state eliminating the gain mismatch between the channels.
The FIR filter is used notch out fs/4 to prevent generation of unwanted dc component.
Page 10.9-6
(-1)n
Fixed Delay
y
1+z-2
2
Adaptive FIR
yc
Phase Detector
Calculate or
Look-up
Coefficients
z-1
t
Accumulator
Fig. Ex1-05
Operation:
1.) 1 goes through a fixed delay that equals the delay through the adaptive FIR filter
when t= 0.
2.) The sum of 1 and 2 are applied to a phase detector.
3.) Except for the unit delay in the phase detector, the same method of calculating the
correlation between the input and the image can be used in the time calibration.
4.) The feedback system is designed to adjust the delay of the adaptive filter so that the
delays experience by both 1 and 2 are identical
Based on simulations, a 21-tap FIR filter with 10-b coefficients is sufficient to correct any
timing error between 200ps to 10-bit accuracy for frequencies as high as 54 MHz.
Page 10.9-7
Output Registers
1.5 bits
Input Input
S/H
1.5 bits
Stage
1
1.5 bits
Stage
i
Stage
13
Code (Di)
1.5-bit
ADC
Vm,i
1.5-bit
DAC
S/H
Vout,i
Fig. Ex1-06
1.5 bit stages permit digital error correction for every stage after the first.
Page 10.9-8
Implementation - Continued
Circuits:
Chopping Amplifier
Op Amp
VDD
Vip
Vin
CF
BIAS4
1''
Chopper
C
I
c
c
c
BIAS4
-+
1'
CI
VCM
M9
Voutn
M7
+-
Voutp
CF
1
1''
C[m]=1 gives c = 1 and c = 0.
C[m]=-1 gives c = 0 and c = 1.
VB1
M10
VB2
M8
Von
Vip
M1
VCM
Vop
M5
VB3
M3
VB5
M11
M6
M2
Vin
M4
M12
M13
VCMFB
M14
VB5
Fig. Ex1-07
Stage capacitors:
First three stages CI = 0.5pF, remaining stages CI = 0.125pF
Op Amp:
50dB gain and settling time of 7ns
CMOS Analog Circuit Design
Chapter 10 Section 9 (5/2/04)
Experimental Results
ADC output spectrum (fs = 120 Msps,
Vin = 3Vpp, and fo = 0.99MHz)
Page 10.9-10
Page 10.9-11
Performance Summary
Process
Resolution
Sampling Rate
Active area
Power Dissipation (Analog/Total)
Full-Scale Input
o = g=t
Without Calibration
0.35m double-poly CMOS
10 bits
120 Megasamples/s
5.2 mm2
171 mW/234 mW
3V peak-to-peak
0
With Calibration
0.35m double-poly CMOS
10 bits
120 Megasamples/s
5.2 mm2
171 mW/234 mW + External
3V peak-to-peak
2-22
+0.75/-0.41 LSB
+0.44/0.36 LSB
-62.4 dB
-62.4 dB
42.5 dB
56.8 dB
46.6 dB
70.2 dB
67.0 dB
67.0 dB
68.0 dB
68.0 dB
43.1 dB
61.5 dB
Page 10.9-12
Yun-Ti Wang and Behzad Razavi, An 8-Bit 150-MHz CMOS A/D Converter, J. of Solid-State Circuits, vol. 35, no. 3, March 2000, pp. 308-317.
CMOS Analog Circuit Design
P.E. Allen - 2004
Chapter 10 Section 9 (5/2/04)
Page 10.9-13
VR,j+1
Preamplifiers
++
Vy
- -
Interpolating Amplifiers
++
- -
Vx
++
Vo2
Aj+1
- -
Vy
VR,j
VR,j+1
Vin
++
Vo3
Vo1
- ++
- -
++
Vo3
21+1
Vo2
VR,j
VR,j+1
- -
Vin
++
- 22+1
++
VR,j
- Aj
2
++
Vx
- 21+1
++
Vo1
VR,j
VR,j+1
Vin
- 22+1
Fig. E2-01
Problem with this scheme is the exponential growth of power and hardware.
Page 10.9-14
VR,j+1
Preamplifiers
++
Vy
- -
Interpolating
Amplifiers
++
Vo2
- -
MUX
Sliding
Stage
++
- -
Stage 1
Stage 2
Stage 3
Vmax
Aj+1
++
- -
++
VR,j
- Aj
Vo3
++
Vx
- -
++
VR,j+2
- -
Vin
VR,j-1
++
Vo1
Sub-ADC
Vmin
- Fig. E2-02
In this example, the first stage employs 16 preamplifiers to generate 16 zero crossings.
If the analog input lies between VR,j and VR,j+1, then a 4-bit coarse ADC and a 16-to-4
MUX route the outputs of the preamplifiers sensing VR,j-1,, VR,j+2 to the next
interpolating stage
The sub-ADC detects 4-bits, 2 of which are used for subsequent digital error correction.
CMOS Analog Circuit Design
Chapter 10 Section 9 (5/2/04)
Page 10.9-16
Embedded Pipelining
Where to apply the pipelining?
Interfacing the stages at
the input of the MUX has
two advantages.
1.) Multiplexing switches
can function as the sampling
switches.
2.) The interconnect
capacitance serves as the
S/H capacitors.
Note that each stage in the pipeline operates in the sample mode for half of the clock
period and in the hold mode for the other half. Since the sub-ADC only operates
during the hold mode, the possibility of interleaving exists to increase the throughput.
CMOS Analog Circuit Design
Chapter 10 Section 9 (5/2/04)
Interleaving
The multiplexers, distributed
sampling circuits, and 2xinterpolation amplifiers
are duplicated for the even
and the odd channels
whereas the front-end
buffer, the preamplifiers, and
all of the sub-ADCs are
shared between the 2
channels.
Difficulties with the first sub-ADC:
1.) Kickback noise disturbs the analog signals at the inputs of the multiplexers.
2.) Sub-ADC must wait until the front-end SHA, the buffer, and the preamplifiers
have settled.
3.) Sub-ADC is in the critical delay path.
Page 10.9-18
Interleaving Continued
Note that one bit of overlap and digital correction suppress errors due to mismatches
between the main path and replica path.
CMOS Analog Circuit Design
Page 10.9-19
Clock scheme which provides both rising and falling edges for sample and hold
operations.
Page 10.9-20
Punctured Interpolation
Implementation of punctured interpolation.
In this scheme, the original inputs (VA1, VA2 and VA3) are used to generate a second set of
interpolated outputs (VB1and VB2). If the offset components of the adjacent VAs are
uncorrelated, then the standard deviation of the offsets of the corresponding VBs are,
A1+A2
B1 =
B1 = 2
2 +2
A1
original
A2
=
2
2
Page 10.9-21
Circuits
Realization of a slice of the signal path in the
first stage.
Page 10.9-22
Circuits Continued
Triple-channel interleaved SHA circuit.
Layout Floorplan
Experimental Performance
SNDR and SFDR at fsample = 150 MHz:
Page 10.9-24
Page 10.9-25
Technology
Resolution
DNL
INL
Sampling Rate
SNDR @ fin = 1.8MHz
fin = 70MHz
1.6 Vp-p
1.5 pF
1.2 mm2
3.3V
330 mW
53 mW
12 mW
395 mW
Page 10.9-26
M. P. Flynn and B. Sheahan, A 400-Msample/s, 6-b CMOS Folding and Interpolating ADC, IEEE J. of Solid-State Circuits, vol. 33, no. 12, Dec.
1998, pp. 1932-1938.
CMOS Analog Circuit Design
P.E. Allen - 2004
Chapter 10 Section 9 (5/2/04)
Page 10.9-27
ADC Architecture
Block diagram:
A folding factor of 4 is chosen requiring 16 folders that produce 16 offset folding signals
and drive 16 comparators. A separate 1-bit coarse ADC determines the MSB.
Page 10.9-28
Folder
Folder block:
The first stage (preamplifier) of the folder uses resistive loads for better speed and
linearity.
The outputs of the second stage are current that permits the current mode operation.
Interpolation
Intepolation is used to eliminate half or more of the folder blocks.
The output from one folder is fed into the split-in-4 blocks.
A quarter of the folder 1 output is added to a quarter of the folder 2 output to give F2R
Two quarters of folder 1 output are summed to form F1R and so forth.
If four parallel MOSFETs are used, a quarter of the current flows through each
device. This causes two problems, 1.) adds an extra node in the signal path and 2.) it
does not allow low supply voltages.
CMOS Analog Circuit Design
Page 10.9-30
Page 10.9-31
Comparators
Tracking
Latching
Comparator advantages:
Because the currents are summed to drive the latching devices, the input signal has very
little effect after latching begins.
Since there is always a path for the current to flow, the folders are not disturbed when
the comparators change from tracking to latching.
Since the output voltage swing is small, the comparator is fast.
CMOS Analog Circuit Design
Page 10.9-32
A cyclic thermometer code is used which is more complex than a flash thermometer code.
The cyclic code along with the decoding logic can surpress the bubbles in the cyclic
code.
The reduced number of comparators does not cause a size penalty for using the cyclic
code.
CMOS Analog Circuit Design
Test Results
FFT for 1 MHz sinusoid sampled at
400 Msamples/s (decimated).
Performance summary:
Technology
SNDR (1MHz sinusoid)
Supply voltage
Power
Area
Input capacitance
CMOS Analog Circuit Design
Page 10.9-33
Page 10.9-34
P.C.S. Scholtens and M. Vertregt, A 6-b 1.6-Gsample/s Flash ADC in 0.18 m CMOS Using Averaging Termination, IEEE J. of Solid-State
Circuits, vol. 37, no. 12, Dec. 2002, pp. 1599-1609.
CMOS Analog Circuit Design
P.E. Allen - 2004
Chapter 10 Section 9 (5/2/04)
Page 10.9-35
Clock
T/H
Resistor Averaging
Vin+
Vin-
Preamplifiers
Clock
Clock
Resistor Averaging
What is Averaging?
Averaging is a technique that connects the outputs of adjacent amplifiers to obtain more
accuracy and more speed.
Digital
Encoding
Network
Digital
Output
Comparators
Fig. Ex4-01
Results:
1.) With no averaging, the standard deviation of the offset voltage is 11mV.
2.) With averaging of the preamplifiers, the standard deviation is 9mV.
3.) With averaging of the preamplifiers and comparators, the standard deviation is 3.7mV.
CMOS Analog Circuit Design
Page 10.9-36
(WL)input
It can be shown that averaging will reduce the value of VOS by approximately 3.
Therefore, the transistors can be made 9 times smaller to achieve the same VOS.
This means that the capacitances are reduced by a factor of 9 while the resistances are
only increased by a factor of 3. As a result, we find that,
1
BWsingle = RO(Cwire + Cload + Cj))
and
1
BWaveraging =
C
Cj
wire
M. Choi and A. Abidi, A 6-b, 1.3-Gsamples/s A/D Converter in 0.35m CMOS, IEEE J. of Solid-State Circuits, vol. 36, no. 12, Dec. 2001, pp.
1847-1858.
CMOS Analog Circuit Design
P.E. Allen - 2004
Chapter 10 Section 9 (5/2/04)
Page 10.9-37
Page 10.9-38
Averaging Termination
In addition to the above concepts of averaging, by applying the concept of averaging
termination, the number of over-range amplifiers can be reduced leading to reduced
power consumption.
The over-range amplifiers are the amplifiers which are outside the usable voltage
range for the purposes of making the averaging network look like an infinite array.
Some results:
No averaging termination:
Averaging termination:
Page 10.9-40
Experimental Results
ENOB versus fsample:
Measured linearity:
Summary of results:
Measured Quantity
ENOB
fsample = 1.6Gsamples/s, fsignal = 263 MHz
fsample = 1.5Gsamples/s, fsignal = 300 MHz
fsample = 1.6Gsamples/s, fsignal = 660 MHz
Power consumption (1.95V analog and 2.25V digital)
At 1.6 Gsamples/s the digital is increased to 2.35V
Value
5.6 bits
5.7 bits
5.0 bits
328mW
340mW
Page 10.10-1
Page 10.10-2
Digital
Processor
Filtering
Sampling
Quantization
y(kTN)
Digital Coding
Fig.10.9-01
Filtering
Sampling
Modulator
Decimation
Filter
Quantization
Digital Coding
y(kTN)
Fig.10.9-02
Components:
Filter - Prevents possible aliasing of the following sampling step.
Sampling - Necessary for any analog-to-digital conversion.
Quantization - Decides the nearest analog voltage to the sampled voltage (determines
the resolution).
Digital Coding - Converts the quantizer information into a digital signal.
CMOS Analog Circuit Design
Page 10.10-3
;;;;
;;;;
;;
Amplitude
Anti-aliasing filter
Signal Bandwidth
fB 0.5fN = 0.5fS
fS =fN
fS =MfN
Amplitude
0
CMOS Analog Circuit Design
Signal Bandwidth
Anti-aliasing filter
Transition band
0 fB = fN
0.5fN
0.5fS
Fig.10.9-03
Page 10.10-4
-6
-4
Ideal curve
-2 1
Input, x
-1 2
-3
-5
Quantization error, e
The quantized signal y can be
1
represented as,
y = Gx + e
-1
where
G = gain of the ADC, normally 1
e = quantization error
The mean square value of the quantization error is
2
1 /2
2 = SQ =
2
e(x)
dx
=
erms
12
Input, x
Fig.10.9-04
-/2
Page 10.10-5
no =
M
0
What does all this mean?
One way to increase the resolution of an ADC is to make the bandwidth of the signal,
fB, less than the clock frequency, fS. In otherwords, give up bandwidth for precision.
However, it is seen from the above that a doubling of the oversampling ratio M, only
gives a decrease of the inband noise, no, of 1/ 2 which corresponds to -3dB decrease or
an increase of resolution of 0.5 bits
The conclusion is that reduction of the oversampling ratio is not a very good method of
increasing the resolution of a Nyquist analog-digital converter.
CMOS Analog Circuit Design
Page 10.10-6
Input+
2.) Predictive oversampling - Uses noise shaping
plus oversampling to reduce the inband noise to
a much greater extent than the straightoversampling ADC. Both the signal and noise
quantization spectrums are shaped.
Quantizer
Output
Loop
Filter
Fig.10.9-05
fS
Loop
Filter
Quantizer
Output
Fig.10.9-06
The noise-shaping oversampling ADCs are also known as delta-sigma ADCs. We will
only consider the delta-sigma type oversampling ADCs.
CMOS Analog Circuit Design
Page 10.10-7
Analog f
B
Input
x(t)
Modulator
(Analog)
fD<fS
Decimator
(Digital)
2fB Digital
PCM
Lowpass Filter
(Digital)
Fig.10.9-07
Page 10.10-8
fS
x +
Integrator v
A/D
u
Components:
Integrator (continuous or discrete time)
Fig.10.9-08
Coarse quantizer (typically two levels)
- A/D which is a comparator for two levels
- D/A which is a switch for two levels
First-order modulator output for a sinusoidal input:
D/A
Quantizer
1.5
1
Volts
0.5
0
-0.5
-1
-1.5
50
100
150
200
Tme (Units of T, clock period)
250
Fig.10.9-09
P.E. Allen - 2004
Page 10.10-9
x[nT ]
Integrator
y[nT ]
v[nTs] +
s +
s +
s
Writing the following relationships,
Delay
+
Quany[nTs] = q[nTs] +v[nTs]
tizer
Fig. 10.9-10
v[nTs] = w[(n-1)Ts] + v[(n-1)Ts]
y[nTs] = q[nTs]+w[(n-1)Ts]+v[(n-1)Ts] = q[nTs]+{x[(n-1)Ts]-y[(n-1)Ts]}+v[(n-1)Ts]
But the first equation can be written as
q[(n-1)Ts] = y[(n-1)Ts]} - v[(n-1)Ts]
y[(n-1)Ts] = q[(n-1)Ts] +v[(n-1)Ts]
Substituting this relationship into the above gives,
y[nTs] = x[(n-1)Ts] + q[nTs] - q[(n-1)Ts]
Converting this expression to the z-domain gives,
Y(z) = z-1X(z) + (1-z-1)Q(z)
Definitions:
Y(z)
Signal Transfer Function = STF = X(x) = z-1
Y(z)
Noise Transfer Function = NT F= Q(x) = 1-z-1
CMOS Analog Circuit Design
Page 10.10-10
Higher-Order Modulators
A second-order, modulator:
q[nTs]
x[nTs]+
Integrator 1
Delay
+
-
Integrator 2
Delay
+
-
y[nTs]
Quantizer
Fig.10.9-11
Page 10.10-11
-1
-j
T
s x
(1-z )= 1 - e
2j e-jf/fs = sin(f/Ts) 2j e-jf/fs
2j
2j x ejf/fs =
|NTFQ(f)| = (2sinfTs)L
Magnitude of noise shaping function
|1-z-1| = (2sinfTs)
Magnitude of the noise transfer
function,
10
L=3
8
6
L=2
;;
;;
L=1
LPF
fb
Frequency
fs/2
Fig.10.9-12
P.E. Allen - 2004
Page 10.10-12
2L
SB = fs (2sinfTs) 12 df 2L+1M2L+112
-fb
L 1 L 1
SB =
L+0.5
L+0.5 erms
=
M
M
2L+1
2L+1
12
Note that as the is a much more efficient way of achieving resolution by increasing M.
erms
Page 10.10-13
erms = 2L+1ML+0.5
0
L=0
-20
L=1
-40
(dB)
erms
-60
n0
L=2
L=3
L=4
-80
-100
8
16 32
64 128 512 1024
Oversampling Ratio, M
Fig.10.9-15
P.E. Allen - 2004
Page 10.10-14
2L+1 2L+1 12
Nyquist Converter:
The dynamic range of a N-bit Nyquist rate ADC is (now becomes VREF for large N),
Maximum signal power (VREF/2 2)2 3
=
=
= 2 22N
DR = 1.5 2N
SQ
2/12
Expressing DR in terms of dB (DRdB) and solving for N, gives
DRdB - 1.7609
or DRdB = (6.0206N + 1.7609) dB
N=
6.0206
Example: A 16-bit ADC requires about 98dB of dynamic range. For a second-order
modulator, M must be 153 or 256 since we must use powers of 2.
Therefore, if the bandwidth is 20kHz, then the clock frequency must be 10.24MHz.
DR2
Page 10.10-15
Multibit Quantizers
A single-bit quantizer:
= VREF
Advantage is that the DAC is linear.
v<0
v>0
Multi-bit quantizer:
Fig. 10.9-13
Consists of an ADC and DAC of B-bits.
VREF
= 2B-1
Disadvantage is that the DAC is no longer perfectly linear.
v
VREF
VREF
2
VREF
2
fS
A/D
D/A
Fig. 10.9-135
Quantizer
Fig. 10.9-14
Page 10.10-16
M = 3 2L+1 (2B-1)21/(2L+1)
Converting the dynamic range to 79,433 and substituting into the above equation gives a
minimum oversampling ratio of M = 48.03 which would correspond to an oversampling
rate of 64. Using the definition of M as fc/2fB gives fB as 10MHz/264 = 78kHz.
(b.) and (c.) For part (b.) and (c.) we obtain a minimum oversampling rates of M = 32.53
and 96.48, respectively. These values correspond to oversampling rates of 32 and 128,
respectively. The bandwidth of the converters is 312kHz for (b.) and 78kHz for (c.).
Page 10.10-17
w[nTs] +
x[nTs] +
-
y[nTs]
v[nTs] +
Quantizer
Q(z)
X(z) +
Integrator
z-1
W(z) +
-
V(z)
Y(z)
Quantizer
Q(z)
X(z) +
-
z-1
1-z-1
Y(z)
Fig.10.9-16
z-1
Y(z) = Q(z) + 1-z-1 [X(z) - Y(z)]
z-1
1
Y(z) 1-z-1 = Q(z) + 1-z-1 X(z)
Page 10.10-18
X2(z) +
z-1
z-1
= 1-z-1 X(z) - 1-z-1 [(1-z
1)Q1(z) + z-1X(z)]
z-1
1-z-1
Y2(z)
Q1(z)
X(z) +
z-1
1-z-1
Y1(z)
z-1 +
z-1 +
Y(z)
Fig.10.9-17
z-2
z-2
Y2(z) = (1-z-1)Q2(z) + z-1X2(z) = (1-z-1)Q2(z) + 1-z-1 X(z) - z-2Q1(z) - 1-z-1 X(z)
= (1-z-1)Q2(z) - z-2Q1(z)
Y(z) = Y2(z) - z-1Y2(z) + z-2Y1(z) = (1-z-1)Y2(z) + z-2Y1(z)
= (1-z-1)2Q2(z) - (1-z-1)z-2Q1(z) + (1-z-1)z-2Q1(z) + z-3X(z) = (1-z-1)2Q2(z) + z-3X(z)
Page 10.10-19
1
1-z-1
Q1(z)
+ +
+
+
Y1(z)
Y(z)
z-1
+ Q2(z)
+ +
1
-1
1-z
-Q1(z)
+
-
z-1
+ Q3(z)
+ +
1
1-z-1
-Q2(z)
+
-
z-1
+
+
1-z-1
Y2(z)
1-z-1
1-z-1
Y3(z)
Fig. 10.9-17A
Comments:
The above structures that eliminate the noise of all quantizers except the last are called
MASH or multistage architectures.
Digital error cancellation logic is used to remove the quantization noise of all stages,
except that of the last one.
CMOS Analog Circuit Design
Page 10.10-20
a1z-1 Y1
1-z-1
a2z-1 Y2
1-z-1
+
a3z-1 Y3
1-z-1
a4z-1 Y4 +
1-z-1
1-bit
A/D
1-bit
D/A
Fig.10.9-20
1.25
1.00
0.75
0.50
0.25
0.00
-1.00
y3
y4
-0.60
-0.20
0.20
0.60
input signal amplitude / VREF
1.00
Page 10.10-21
X +
-
1-bit
A/D
1-bit
A/D
Fig.10.9-20
1.50
y1
y2
1.25
y3
y4
1.00
0.75
0.50
0.25
0.00
-0.60
-0.20
0.20
0.60
input signal amplitude / VREF
-1.00
1.00
Page 10.10-22
X +
-
a1z-1 +
1-z-1 -
+
-
Fig.10.9-21
a2z-1
1-z-1
a3z-1
1-z-1
q1
+
+
q2
+
+
Comments:
The stability is guaranteed for cascaded structures
The maximum input range is almost equal to the reference voltage level for the
cascaded structures
All structures are sensitive to the circuit imperfection of the first stages
The output of cascaded structures is multibit requiring a more complex digital
decimator
Page 10.10-23
z-1
Vo(z)
Vi(z)
Vo(z)
az-1
1-z-1
Fig.10.9-22
Fully-Differential, Switched
Capacitor Implementation:
2
+
vin
-
Cs
2Cs
Ci
+ - +
+
vout
-
Ci
Fig.10.9-23
T
Vout(z) Cs z-1
V out(e j) C1
e-j/2
T/2 -j/2
C1
o
=
=
e
Vin(z) Ci 1-z-1
V in( e j) C2 j2 sin(T/2) T jTC2 sin(T/2)
o
I
V out(e j)
C1
o
=
(Ideal)x(Magnitude
error)x(Phase
error)
where
Ideal
=
I
TC2
j
V in( e j)
CMOS Analog Circuit Design
Page 10.10-24
VDD2/2
VDDMCs
DR = 4kT/MC = 8kT
s
2.) Lower bound on the sampling capacitor, Cs:
8kTDR
Cs =
2
VDDM
3.) Static power dissipation of the integrator:
Pint = IbVDD
4.) Settling time for a step input of Vo,max:
Vo,max Ci Cs
CsVDD
Ib = C i T
= T
= CsVDD(2fs) = 2MfNCsVDD
C VDD = T
settle
settle
settle i
Page 10.10-25
Implementation of Modulators
Most of todays delta-sigma modulators use fully differential switched capacitor
implementations.
Advantages are:
Doubles the signal swing and increases the dynamic range by 6dB
Common-mode signals that may couple to the signal through the supply lines and
substrate are canceled
Charge injected by the switches are canceled to a first-order
Example:
First integrator
dissipates the
most power and
requires the most
accuracy.
X +
0.5z-1
1 - z- 1
VRef+
VRef-
YB
VRef+
1
C
1d
1d
YB
2C
2
2
0.5z-1
1 - z- 1
VRef+
VRef-
YB
YB
VRef-
2C
VRef+
Q1
2C
1
1d
1d
2
2
YB
Y
VRef-
2C
Fig.10.9-24
Page 10.10-26
a1
z - 1 y1
b1
a2
z - 1 y2
a3
z - 1 y3
a4
1-bit
z - 1 y4 A/D
b2
1-bit
D/A
(6-1)
where a1 = 1/3, a2 = 3/25, a3 = 1/10, a4 = 1/10, b1= 6/5, b2= 1 and = 1/6
Advantages:
The modulator combines the advantages of both DFB and DFF type modulators:
Only four op amps are required. The 1st integrators output swing is between VREF
for large input signal amplitudes (0.6VREF), even if the integrator gain is large (0.5).
A local resonator is formed by the feedback around the last two integrators to further
suppress the quantization noise.
The modulator is fully pipelined for fast settling.
A.L. Coban and P.E. Allen, A 1.5V, 1mW Audio Modulator with 98dB Dynamic Range, Proc. of 1999 Int. Solid-State Circuits Conf., Feb.
1999, pp. 50-51.
CMOS Analog Circuit Design
P.E. Allen - 2004
Page 10.10-27
DR = 98 dB
BW = 20 kHz
Cs = 5 pF
0.5 m CMOS
CMOS Analog Circuit Design
Page 10.10-28
OSR = 64
OSR = 32
OSR = 16
OSR = 8
Page 10.10-29
Capacitor
Cs
Capacitor Values
Integrator 1 Integrator 2 Integrator 3 Integrator 4
5.00pF
0.15pF
0.30pF
0.10pF
Ci
Ca
15.00pF
1.25pF
3.00pF
1.00pF
1d
2
0.05pF
2d
Cb1
0.12pF
Cb2
0.10pF
Fig.10.9-25
Page 10.10-31
Page 10.10-33
frequency, (kHz)
Page 10.10-34
Supply voltage
Die area
Supply current
analog part
digital part
Reference voltage
Clock frequency
Oversampling ratio
Signal bandwidth
Peak SNR
Peak SNDR
Peak S/D
HD3 @ -5dBv 2kHz input
DR
1.5 V
1.02 mm x 0.52 mm
660 A
630 A
30 A
0.75V
2.8224MHz
64
20kHz
89 dB
87 dB
101dB
-105dBv
98 dB
Page 10.10-35
Page 10.10-36
fs
L+1-th
order
2fN
First-half
band filter
fN
fN
Second-half
band filter
Droop
correction
Fig.10.9-26
1.) For modulators with (1-z-1)L noise shaping comb filters are very efficient.
Comb filters are suitable for reducing the sampling rate to four times the Nyquist
rate.
Designed to supress the quantization noise that would otherwise alias into the
signal band upon sampling at an intermediate rate of fs1.
2.) The remaining filtering is performed by in stages by FIR or IIR filters.
Supresses out-of-band components of the signal
3.) Droop correction - may be required depending upon the ADC specifications
Page 10.10-37
Comb Filters
A comb filter that computes a running average of the last D input samples is given as
1 D-1
y[n] = D x[n-i]
i=0
Page 10.10-38
|HD(f)| dB
-20
K=1
-40
K= 2
-60
K=3
-80
-100
0
fb
fs
D
2 fs
D
Frequency
3 fs
D
4 fs
D
Fig.10.9-27
Page 10.10-39
z-1
-
Numerator Section
+
+
z-1
-
z-1
fs/D
K = L +1 Integrators
+
z-1
Denominator Section
+
z-1
Y
-
z-1
K = L +1 Differentiators
Fig.10.9-28
Comments:
1.) The L+1 integrators operating at the sampling frequency, fs, realize the denominator
of HD(z).
2.) The L+1 differentiators operating at the output rate of fs1 (= fs/D) realize the
numerator of HD(z).
3.) Placing the integrator delays in the feedforward path reduces the critical path from
L+1 adder delays to a single adder delay.
Page 10.10-40
h(0)
Input
z-1
y(n)
Output
y(n)
h(0)
z-1
h(1)
z-1
h(2)
z-1
h(2)
z-1
h(3)
z-1
Input
Output
z-1
h(1)
x(n)
h(3)
z-1
h(N-1)
h(N-1)
Direct-form structure
for an FIR digital filter.
Transposed direct-form
FIR filter structure.
Fig.10.9-29
S.R. Norsworthy, R. Schreier, and G.C. Temes, Delta-Sigma Data Converters-Theory, Design, and Simulation, IEEE Press, NY, Chapter 13, 1997.
CMOS Analog Circuit Design
P.E. Allen - 2004
Page 10.10-41
Magnitude (dB)
-20
-50
-80
-110
4000
Frequency (Hz)
Page 10.10-42
analog fB
input
Time
Frequency
CMOS Analog Circuit Design
fD
LOW-PASS
2f B
FILTER
digital
PCM
DECIMATOR
MODULATOR
Time
Frequency
Frequency
P.E. Allen - 2004
Page 10.10-43
Bandpass Modulators
Block diagram of a bandpass modulator:
fS
x +
Resonator v
A/D
Components:
u
Resonator - a bandpass filter of order
D/A
2N, N= 1, 2,....
Quantizer
Fig.10.9-27A
Coarse quantizer (1 bit or multi-bit)
The noise-shaping of the bandpass oversampled ADC has the following interesting
characteristics:
Center frequency = fs (2N-1)/4
Bandwidth = BW = fs /M
dB
Illustration of the Frequency Spectrum
(N=1):
BW
BW
Attenuation
fs
4
3fs
4
Frequency
fs
Fig. 11-32
Application of the bandpass ADC is for systems with narrowband signals (IF
frequencies)
CMOS Analog Circuit Design
Page 10.10-44
z-1
V(z)
z-1
Fig. 10.9-27C
Q(z)
X(z) +
-
z-1
1+z-2
Y(z)
Fig.10.9-27B
z-1
Y(z) = Q(z) + [X(z) - Y(z)] -2
1+z
1+z-2
z-1
Q(z) +
Y(z) =
1+ z-1-z-2X(z)
1+ z-1-z-2
1+z-2
NTFQ (z) =
1+ z-1-z-2
Page 10.10-45
Resonator Design
Resonators can be designed by applying a lowpass to bandpass transform as follows:
X(z) +
+
V(z)
z-1
z-1
1 - z-1
V(z)
z-2
-z-2
1 + z-2
Fig. 10.9-27D
Result:
Simple way to design the resonator
Inherits the stability of a lowpass modulator
Center frequency located at fs/4
Page 10.10-46
0.5
z-2
1 + z-2
+
+
0.5
z-2
1 + z-2
Y(z)
Fig. 10.9-27E
Comments:
Designed by applying a lowpass to bandpass transform to a second-order lowpass
modulator
The stabilty and SNR characteristics are the same as those of a second-order lowpass
modulator
The z-domain output is given as,
Y(z) = z-4X(z) + (1+z-2)2Q(z)
The zeros are located at z = j which corresponds to notches at fs/4.
Page 10.10-47
z-1
z-1
V(z)
Fig. 10.9-27F
Page 10.10-49
IF
RF
BP
Filter
BP Approach
Digital
LP
Filter
Nyquist
ADC
LP
Filter
Nyquist
ADC
sinLO2t
Analog
Digital
Mixer
LO1
RF
IF
LPF
LPF
sinLO2t
Digital
VCO
cosLO2t
BP
Mod.
Mixer
Fig10.9-27G
Page 10.10-50
DAC
MfN
Analog
lowpass
filter
Output
Analog Section
Fig10.9-29
Operation:
1.) A digital signal with N-bits with a data rate of fN is sampled at a higher rate of MfN by
means of an interpolator.
2.) Interpolation is achieved by inserting 0s between each input word with a rate of
MfN and then filtering with a lowpass filter.
3.) The MSB of the digital filter is applied to a DAC which is applied to an analog
lowpass filter to achieve the analog output.
CMOS Analog Circuit Design
Page 10.10-51
Interpolation
fS +
fS
fS
Digital
Filter
fS MSB
VRef
-VRef
DAC
Digital Code
Conversion
fS=MfN
y(k)
0100000000000000 =-1 fS
1011111111111111 = 1
Analog
Analog Output
Lowpass
Filter
Fig10.9-31
Operation:
1.) Interpolate a digital word at the conversion rate of the converter (fN) up to the sample
frequency, fs.
2.) The word length is then reduced to one bit with a digital sigma-delta modulator.
3.) The one bit PDM signal is converted to an analog signal by switching between two
reference voltages.
4.) The high-frequency quantization noise is removed with an analog lowpass filter
yielding the required analog output signal.
Sources of error:
Device mismatch (causes harmonic distortion rather than DNL or INL)
Component noise
Device nonlinearities
Clock jitter sensitivity
Inband quantization error from the - modulator
CMOS Analog Circuit Design
Page 10.10-52
IRef
1
y(k)
2
R
C
1
y(k)
Analog
2
lowpass Analog y(k)
Output
filter
with -3dB
2
frequency
y(k)
of 0.5fN
-VRef
Voltage-driven DAC with a
passive lowpass filter stage.
Analog
lowpass Analog
filter with Output
-3dB
frequency
of 0.5fN
-IRef
Current-driven DAC with a
passive lowpass filter stage.
Fig10.9-32
Page 10.10-53
v(t)
y(k)
Vref1 - vref2
Vref1 + Vref2
y(k) +
2
2
This results in a gain or an offset error, but the output is still linear.
v(t) = Vref +
Page 10.10-54
Page 10.10-55
VRef
1
y(k)
C2
C1
2
1
y(k)
To analog
lowpass
filter
Fig10.9-34
-VRef
Page 10.10-56
-0.5fN 0
Interpolation
filter output
0.5fN
-0.5fN 0
0.5fN
MfN Frequency
0.5fN
MfN Frequency
fN
(M-1)fN
MfN Frequency
Delta-sigma
modulator
output
-0.5fN 0
Lowpass
filter
output
-0.5fN 0
CMOS Analog Circuit Design
MfN Frequency
Fig10.9-33
Page 10.10-57
Page 10.11-1
Dual Slope
Successive Approximation
with self-correction
Speed
Area Dependence
Maximum
(Expressed in terms on the number of
Practical Number of T a clock period) bits, N, or other
of Bits (1)
ADC parameters
12-14 bits
Independent
2(2NT)
N
12-15 bits
NT
1-Bit Pipeline
10 bits
T (After NT delay )
Algorithmic
12 bits
NT
Independent
Flash
6 bits
2N
Two-step, flash
10-12 bits
2T
2N/2
Mulitple-bit, M-pipe
12-14 bits
MT
2N/M
- Oversampled (1-bit, L
loops and M= oversampling
ratio = f clock/2fb)
15-17 bits
MT
Page 10.11-2
20
15
10
5
1
Flash
Pipelined
Algorithmic
Successive approximation
Dual-slope
Delta-sigma
Folding/Interpolating
Bandpass delta-sigma
102
104
106
Conversion rate, (samples/sec.)
1010
108
Figure 10.10-1
Page 10.11-3
100
10
0.1
0.01
100
10 6
10 4
Conversion Rate (Samples/second)
10
10 10
Figure 10.10-2
P.E. Allen - 2004
Page 10.11-4
References - Continued
[19] A 5-V Single-Chip Delta-Sigma Audio A/D Converter with 111 dB Dynamic Range. Fujimori, I., IEEE J-SC, vol. 32, no.
3, Mar 97 329-336
[20] A 256 x 256 CMOS Imaging Array with Wide Dynamic Range Pixels and Column-Parallel Digital Output. Decker, S.,
IEEE J-SC, vol. 33, no.12, Dec 98 2081-2091
[21] A 400 Msample/s, 6-b CMOS Folding and Interpolating ADC. Flynn, M., IEEE J-SC, vol. 33, no.12, Dec 98 1932-1938
[22] An Analog Background Calibration Technique for Time-Interleaved Analog-to-Digital Converters. Dyer, K. C., IEEE J-SC,
vol. 33, no.12, Dec 98 1912-1919
[23] A CMOS 6-b, 400-Msample/s ADC with Error Correction. Tsukamoto, S., IEEE J-SC, vol. 33, no.12, Dec 98 1939-1947
[24] A Continuously Calibrated 12-b, 10-MS/s, 3.3-V ADC. Ingino, J. M., IEEE J-SC, vol. 33, no.12, Dec 98 1920-1931
[25] A Delta-Sigma PLL for 14b, 50 ksamples/s Frequency-to-Digital Conversion of a 10 MHz FM Signal. Galton, I., IEEE JSC, vol. 33, no.12, Dec 98 2042-2053
[26] A Digital Background Calibration Technique for Time-Interleaved Analog-to-Digital Converters. Fu, D., IEEE J-SC, vol. 33,
no.12, Dec 98 1904-1911
[27] An IEEE 1451 Standard Transducer Interface Chip with 12-b ADC, Two 12-b DACs, 10-kB Flash EEPROM, and 8-b
Microcontroller. Cummins, T., IEEE J-SC, vol. 33, no.12, Dec 98 2112-2120
[28] A Single-Ended 12-bit 20 Msample/s Self-Calibrating Pipeline A/D Converter. Opris, I. E., IEEE J-SC, vol. 33, no.12, Dec
98 1898-1903
[29] A 900-mV Low-Power A/D Converter with 77-dB Dynamic Range. Peluso, V., IEEE J-SC, vol. 33, no.12, Dec 98
1887-1897
[30] Third-Order Modulator Using Second-Order Noise-Shaping Dynamic Element Matching. Yasuda, A., IEEE J-SC, vol.
33, no.12, Dec 98 1879-1886
[31] R, G, B Acquisition Interface with Line-Locked Clock Generator for Flat Panel Display. Marie, H., IEEE J-SC, vol. 33,
no.7, Jul 98 1009-1013
[32] A 25 MS/s 8-b - 10 MS/s 10-b CMOS Data Acquisition IC for Digital Storage Oscilloscopes. Kusayanagi, N., IEEE J-SC,
vol. 33, no.3, Mar 98 492-496
[33] A Multimode Digital Detector Readout for Solid-State Medical Imaging Detectors. Boles, C. D., IEEE J-SC, vol. 33, no.5,
May 98 733-742
[34] CMOS Charge-Transfer Preamplifier for Offset-Fluctuation Cancellation in Low Power A/D Converters. Kotani, K., IEEE JSC, vol. 33, no.5, May 98 762-769
[35] Design Techniques for a Low-Power Low-Cost CMOS A/D Converter. Chang, Dong-Young, IEEE J-SC, vol. 33, no.8,
Aug 98 1244-1248
CMOS Analog Circuit Design
Page 10.11-1
CONCLUSION
Key aspects:
1.) Square law relationship:
KW
iD = 2L (vGS - VT)2
2.) Small-signal transconductance formula:
2KWID
L
3.) Small-signal simplification:
gm 10gmbs 100gds
4.) Saturation relationship:
gm =
2ID
K(W/L)
Remember to think and understand the problem before using the simulator.
Any questions concerning the course can be e-mailed to pallen@ece.gatech.edu
Other analog resources can be found at www.aicdesign.org
VDS(sat) =