ELEN-457

OPERATIONAL AMPLIFIERS

ELEN-457

Jose Silva-Martinez and Edgar Sanchez-Sinencio

Spring 2003

Jose Silva-Martinez -1- Texas A&M University

 ELEN-457

OPAMPS (Buffered OTAs)

Main parameters:

•DC gain (104-106) V/V)

•Frequency Limitations
•Bandwith (Few Hertz~1kHz)
•Gain-Bandwidth product (1~100 Mhz)
•Output resistance (r0 < 1kOhm)
•Input Resistance (ri > 100 kOhms) +
•Linear range
•Common-mode input range
•Output swing -
•Stability
•DC Offsets
•Noise Level
•CMRR, PSRR
•… and more!!!

Jose Silva-Martinez -2- Texas A&M University

 ELEN-457

VDD
Two-stage Operational
M2 M2
i01 2i01
M3
i02 Transconductance
-v0 Amplifier (OTA)
i01 − i01 R1 C1
RL CL OPEN-LOOP DC gain is very high
vd -vd
M1 M1
but not precise!

æ g öæ g ö æ g ö
ADC = çç m1 ÷÷çç m3 ÷÷ • •çç mn
VSS è g1 è g 2 è gL

Depends on temperature variations,

+ transistor mismatches, process
parameter tolerances, aging, etc.
Variations could be easily greater than
300 %!
-
Open-loop OPAMPS are used as sign
detectors (COMPARATORS) but not
for linear operation!!

Jose Silva-Martinez -3- Texas A&M University

 VDD ELEN-457

M2 M2 M3
i01 i02
2i01
-v0
i01 − i01 R1 C1
RL CL
vd -vd
M1 M1

VSS
Main equations

g m1 g m3
A VDC = −
g1 g L
Phase Margin < 45 degrees
g AVDC
ωp1 = 1
C1
gL Phase
ωp 2 =
CL
( (
GBW = (A VDC )* min ωp1 ,ωp 2 ))
æω ö æω ö ω
Phase _ m arg in =180 − tan −1 ç u ÷ − tan −1 ç u ωp1 ωp2 ωu
ç ωp1 ÷ ç ωp 2 GBW
è è

Jose Silva-Martinez -4- Texas A&M University

 VDD ELEN-457

M2 M2 M3
i02
Two-stage OTA
i01 2i01
-v0
i01 − i01 R1 C1 CM
RL CL
vd -vd
M1 M1

Phase compensation ➨ Pole splitting!!

After compensation
VSS
AVDC ➨ Phase Margin > 45 degrees
➨ Bandwidth is reduced!!!
g m1 g m3
AVDC = −
g1 g L
g1
ω p1 ≅
g Phase
C1 + m 3 C M
gL GBW
g m3 ω
ω p2 ≅
C1 + C L ωp1’ ωp1 ωp2
g m1 GBW’ ωp2’
GBW = AVDC * ω p1 ≅
CM
æ
− GBW ÷
ö æ GBW ö
Phase _ m arg in = 180 − tan 1 çç ÷
− tan −1 ç
ç ω
è ω p1 è p2

Jose Silva-Martinez -5- Texas A&M University

 ELEN-457

Compensated Two-stage OTA

VDD

M2 M2 M3 Dominant pole is not precise as well!

i01 i02
2i01
g1
-v0 ωP1 =
æ g m3 ö
i01 − i01 R1 C1 CM çç Cm
vd -vd
RL CL g
è L
M1 M1
Depends on temperature variations,
transistor mismatches, process
parameter tolerances, aging, etc.
VSS
Variations could be greater than 300
%.

Gain-Bandwidth product is more precise, variations are usually less than 50%

g m1
GBW ≅
Cm

Jose Silva-Martinez -6- Texas A&M University

 ELEN-457

VDD
M2 M2 M3
i01 2i01 iout
-v0
i01 − i01 R1 C1 CM
RL CL
vd -vd
M1 M1
IB2
IB1 VB M4

VSS

Spectral noise density at low frequencies (capacitors are neglected)

8 kT é æ g ö
2 ù
2
i eq _ out = ê(2g m1 + 2g m 2 )ç m 3 ÷÷ + g m3 + g m 4 ú Output referred current noise density
3 ê ç ú
ë è g1 ø
é ù
2 ê
i eq _ out 8kT ê g m 2 g m3 + g m 4
v eq _ in 2 = =
2 3g ê 2 + 2 + 2
Input referred voltage noise density
æ g m1g m3 ö m1 ê g m1 æ g m3 ö
çç ÷÷ ê g m1 çç ÷÷
è g1 ø ë è g1 ø

Jose Silva-Martinez -7- Texas A&M University

 ELEN-457
VDD
M2 M2 M3
i01 i02
2i01
-v0 Design considerations:
i01 − i01 R1 C1 CM
RL CL ➨ |VSS | > VGS1+VDSAT_current source+∆
∆
vd -vd
➨ VGS2 = VGS3 (low offset)
M1 M1
➨ VDSAT2 , VDSAT3 , VDSAT4 have to be
IB2
minimized (max swing)
IB1 VB M4

VSS
Cancels the zero
g m1 g m 3
A VDC =
Design Equations: g1 g L 1/gm3 CM
g1 g m3 g
ω p1 ≅ − , ωp2 = − , ωZ = m3
g m3 C1 + C L CM
CM
gL
g m1
GBW = A VDC * ω p1 ≅
CM
16 kT é gm2 ù
v eq _ in 2 ≅ ê1 +
3g m1 ë g m1

−1 ç
æ GBW ö æ GBW ö
Phase _ m arg in = 180 − tan ÷ − tan −1 ç
ç ω p1 ÷ ç ωp2
è è
Jose Silva-Martinez -8- Texas A&M University
 ELEN-457
Buffered OTA = Operational Voltage Amplifier (OPAMP)
VDD SIMPLE MACROMODEL:
r0
M2 M2 M3 vy
i01 i02 v+
2i01 M4 vout
v-
A2(vy)
R1 C1
i01 − i01 R1 C1 CM gm(v+-v-)
v0
vd -vd
M1 M1

IB1 IB2 IB3 Internal pole Buffer

A VDC = (g m R 1 )A 2
VSS
A 2 ~ 1 − 0 .5
AVDC gm
GBW ≅
C1
R out = r0 ≅ 1 / g m 4
➨Notice that load capacitors must satisfy the
following condition, otherwise phase margin
ω is not good enough
ωp1 1
GBW ωp2 GBW <
r0 C L

Jose Silva-Martinez -9- Texas A&M University

 ELEN-457

OPAMP Characterization
Main parameters to be measured:

•DC gain (104-106) V/V)

•Frequency Limitations
•Bandwith (Few Hertz~1kHz)
•Gain-Bandwidth product (1~100 Mhz)
•Input and Output Impedance
•Signal Swing
•Common-mode input range
•Output swing
•Stability
•DC Offset
•Slew-rate For this section, see:
•CMRR CMOS Analog design, Allen & Holberg
•PSRR 2nd edition, HPR, 2002.

Jose Silva-Martinez -10- Texas A&M University

 ELEN-457

OPAMP Characterization

DC gain (104-106) V/V) : How to measure/characterize it?

•Very difficult to measure in open- •Stabilize for DC

loop due to DC offsets.
Vi +
V0
+
V0≠0 -
-

CC RC

•For DC, the OPAMP operates in closed loop!!

•For frequencies higher than 1/RCCC, the
OPAMP operates in open-loop with a
grounded load given by RC.

Jose Silva-Martinez -11- Texas A&M University

 ELEN-457

OPAMP: DC Characterization
Vi + How to measure/characterize it?
A(s) V0
- •At DC vo A(s)
= ≅1
v i 1+ A(s)
CC RC
•If A(s)B(s) <<1 then the measured gain is
dominated by the OPAMP transfer function!
vo A (s ) vo
= ≅ A(s)
v i 1+ A(s)B(s)
AVDC vi
1 OPAMP
(open-loop)
sC C 1
B(s) = =
1 1+ sR C C C
RC +
sC C

é 1 ù ω
ωP1 > [A DC ]ê ωP1
ë R CCC 1
GBW
R CCC
Jose Silva-Martinez -12- Texas A&M University
 ELEN-457

OPAMP Characterization
DC Offset

Ioffset
•Practical circuit
+ V0 Ioffset
A(s) + 1
- +
~VDD, VSS A(s) + 1 V0~Voffset
- -Ioffset

Offset free
Offset free
Ioffset
+ V0 ~0 Voffset
Voffset
A(s) + 1 +
- -Ioffset V0~Voffset
A(s)
-

Offset free
Offset free

Jose Silva-Martinez -13- Texas A&M University

 ELEN-457

OPAMP Characterization: GBW and stability

Vin
Vin
+
+ V0
V0 A(s)
A(s) -
-

AVDC
Vi
OPAMP
(open-loop)

ω
ωP1
GBW
Enough phase margin

Jose Silva-Martinez -14- Texas A&M University

 ELEN-457

OPAMP Characterization: GBW and stability

Vin
Vin
+
+ V0
V0 A(s)
A(s) -
-

AVDC
Vi
OPAMP
(open-loop)

ω
ωP1

Not enough phase margin

Jose Silva-Martinez -15- Texas A&M University

 ELEN-457

OPAMP Characterization: Slew-Rate (max speed)

Vin
+
A(s) V0
-

Max output level

VDD

Slew-Rate
d
max Vo( t ) t
dt

VSS
Min output level

Jose Silva-Martinez -16- Texas A&M University

 ELEN-457

OPAMP Characterization: Output Swing

Vin
+
A(s) Low-Frequency Gain
-
∆Vo
Vmax A low−freq ≅
∆Vi

vi vo
Use a slow triangular input t Output
signal such that the raising swing
and falling edges are not
determined by slew rate Vmin
limitations
Small signal (~ 10 mV)

vo
vi

DC Offset

Jose Silva-Martinez -17- Texas A&M University

 ELEN-457

OPAMP Characterization: Input and Output impedance

iin
iin
+ +
A(s)
Vin A(s) -
Vin
-

CC RC
Zin

R
RC and CC as large as possible!!

At Low frequencies:
C R C Zmeasured ≠ Zo (Why????)

At medium frequencies:
ω Zmeasured = Zo||RC

Be sure that the OPAMP (all internal

transistors) is properly biased during
characterization!!

Jose Silva-Martinez -18- Texas A&M University

 ELEN-457

If the OPAMP is not precise, how we can design accurate systems?

Answer is FEEDBACK!!!

Are we using feedback during our daily activities?

•Can you shave yourself closing your eyes?

•Can you drive your car closing your eyes?

•Can you adjust the supply voltages without a voltage

indicator?

•Can you measure (without any equipment) the

magnetic field generated by your cellular phone?

•Can you control properly your daily activities without

feedback?

•Feedback gives you better control

•Benefits?
•Drawbacks?
Jose Silva-Martinez -19- Texas A&M University
 ELEN-457

r0
+ vout
v+
+
AV(s) *(v+-v-)
v-
-
-
A VDC
AV ≅
Typical values for low-frequency opamps: s
1+
AV~105 ωP
ωP~100 rad/sec gm
GBW = A VDC * ω p ≅
r0 ~ < 100 Ohms i2 FEEDBACK!! CM
R out = r0

i1
Z2
If AV~∞ then
vi - v- = 0 VIRTUAL GROUND
Z1 v0 i1=-i2
-v0/AV vo Z
+
=− 2
vi/Z1 = -v0/Z2 vi Z1

Jose Silva-Martinez -20- Texas A&M University

 i2
ELEN-457

i1 If AV is finite then v- ≠ 0
Z2
vi - i1 = −i 2 if Zin = ∞
Z1 v0 or

-v0/AV +
(vi − (− v 0 / A V ) )Y1 = (v 0 − (− v 0 / A V ) )Y2

é ù é ù
ê ú ê
A VDC v0 Y1 ê 1 Z
ú=− 2 ê 1
AV = =−
1+
s vi Y2 ê 1 æ Y1 ö ú Z1 ê 1 æ Z2 ö
ωP ê1+ çç1+ ÷÷ ú ê1+ çç1+ ÷÷
êë A Vè Y2 øú êë A Vè Z1 ø
R out = 0 and Z in = ∞

Av(s)
s
1+
1 æ Z2 ö ωP æ Z 2 ö
Error = − çç1+ ÷÷ = − çç1+
Error A V è Z1 A DC è Z1
ω

Jose Silva-Martinez -21- Texas A&M University

 ELEN-457

vi + v0
A(s) FEEDBACK:
-
•If you measure the output at the time
B(s) you apply the stimuli you can have
v i − B (s ) v 0 better control on the system!!

Definitions:
A(s): Amplifier gain (very large but not very well controlled)
B(s): Feedback Factor (Very well controlled)
T(s)=A(s)B(s): Loop Gain (Extremely important parameter!!)

Applying Mason Rule: v o Direct trayectory A(s) A(s)

= = =
vi 1− loop gain 1− (− T (s) ) 1+ A (s) B(s)

Jose Silva-Martinez -22- Texas A&M University

 ELEN-457

vi + v0
STABILITY: v o = A (s)
A(s) v i 1+ A(s) B(s)
-
If: A0
B0
vo s 2 + BWs + ω0 2
=
v i 1+ A 0 B0
A(s) OPAMP s 2 + BWs + ω0 2
(open-loop)
A(s)B0
Real and negative
Imaginary and negative H(s)
Real and positive (=1)
ω
0 dB 0 dB ω
ωP1
B0

Enough phase margin

Jose Silva-Martinez -23- Texas A&M University

 ELEN-457

vi + v0
A(s) STABILITY ISSUES:
-
A0
B0 vo s 2 + BWs + ω0 2
=
v i 1+ A 0 B0
2 2
OPAMP
s + BWs + ω 0
A(s)
(open-loop)
Real and negative
A(s)B0
ω
s=jω
Imaginary and negative
H(s)
0 dB ω
Unstable (denominator
ωP1
B 0 dB ω
goes to zero and overall 0
gain goes to infinity)
-∞ (0 in linear scale)!!!
Not Enough phase margin!!!
Jose Silva-Martinez -24- Texas A&M University
 ELEN-457

+ vo A(s)
vi A(s) v0 =
-
v i 1− (− T(s) )
If T (s) >>1, then
B(s) v o A (s) 1 1
v
vi − 0 ≅ = Error ≈
B (s ) v i T (s) B(s) T (s)

Key points:
If you want to amplify your signal:B(s) must be an attenuator (voltage divider!!)
The error is determined by the overall loop gain: T(s)=A(s)B(s)

Z1 1 A(s)
B(s) = =
Z1 + Z 2 1+ Z 2
Z1 Z1 Z2 T(s) Error
-
Max
v o (s ) Z A(s) v0 Error
≅1+ 2 vi +
v i (s) Z1 ωmax
ω

Jose Silva-Martinez -25- Texas A&M University

 ELEN-457

Inverting Amplifier: Apply superposition Z1 Z2

vi -
A(s) and B(s) are sharing some elements!! A(s) v0
A(s)= ? B(s) = ? T(s) = ? +

Why not to open

the loop here? Loop can also
Z2 be opened here

Z1
vi - Z1 Z2
-
AV(s) v0
+ A(s) v0
Z2 A V +
A (s) = −
Z1 + Z 2 Z1A V
T (s) =
Z1 + Z 2
æ ö
ç
Z2 A V ç
v0 A(s) Z1 + Z 2 Z2 ç 1
=− =− =− ç
vi 1+ T (s) 1+ Z1A V Z1 ç 1+ Z 2
Z1 + Z 2 ç Z1
ç 1+
è AV

Jose Silva-Martinez -26- Texas A&M University

 ELEN-457

Inverting Amplifier: consider the non-zero output impedance!!

Z1 Z2
vi -
Z2 r0
A(s) v0
+
Z1
vi v-

V0’ Z2 r0

+ Z1A V
v+ T (s) =
AV (v+-v-) Z1 Z1 + Z 2 + r0
- v-

V0’
A ' (s ) = −
(Z 2 + r0 )A V v+ +
AV (v+-v-)
Z1 + Z 2 + r0
-

Similarly to previous analysis é ù

ê r0
(Z 2 + r0 )A V vo
ê 1−
v0 ' Z +Z +r (s )= − Z 2 êê A VZ2
Z 2 + r0
(s) = − 1 2 0 vi Z1 1 +
vi Z1A V ê
1+ ê 1+ Z1
Z1 + Z 2 + r0 êë AV
Jose Silva-Martinez -27- Texas A&M University
 ELEN-457

Inverting Amplifier: consider the non-zero output impedance!!

é ù
ê r0
ê 1− Z1 Z2
vo Z2 ê A VZ2 vi
(s ) = −
-
vi Z 1 ê 1 + Z 2 + r0 A(s) v0
ê
ê1+ Z1 +
êë AV
The error can be approximated as:

Z 2 + r0 IDEAL GAIN
1+
r0 Z1
Error ≅ − − Determined by the OPAMP open-loop gain
A V Z2 AV
Determined by the OPAMP output impedance

What are the effects of the finite OPAMP input impedance and load
(externally connected) impedance?
These could be exam questions!!

Jose Silva-Martinez -28- Texas A&M University