Transcript Document

DIFFERENTIAL AMPLIFIERS
DIFFERENTIAL AMPLIFIER
1. VERY HIGH INPUT IMPEDENCE
2. VERY HIGH BANDWIDTH
3. DIFFERENTIAL INPUT
4. DC DIFFERENTIAL INPUT ACCEPTED
5. HIGH COMMON MODE REJECTION
6. SIGNAL INTEGRITY AT THE OUTPUT
Let us start with a simple amplifier that can give us
at the output a signal proportional to the difference
between two signals, each with reference to a ground.
In a MOSFET under small signal conditions the
output current is proportional to the signal voltage
across the gate and source. As is obvious any signal
applied to the source needs to be preferably applied
through a buffer to reduce loading, while that at the
gate could be applied directly.
It is evident from the circuit diagram that when we
consider the output at transistor M2, vo2, we are looking
at a cascade of CD amplifier followed by CG amplifier
from vs1 to vo2 and CS amplifier from vs2 to v02. Similarly
considering the output at transistor M1, vo1, we are
looking at a cascade of CD amplifier followed by CG
amplifier from vs2 to vo1 and CS amplifier from vs1 to v01.
Assuming a total symmetrical circuit, it is sufficient to
analyze the effect of any one signal input to analyze the
total circuit. In analyzing the circuit we will use
superposition theorem.
Let us now look at the equivalent circuit and its analysis.
We represent the current source ISS by its output
impedance 1/go in the equivalent circuit.
In the equivalent circuit
vgs1  vgs2  vs1
vo1   gm1 vgs1 RL and


vgs1  1 go gm1 vgs1  gm1 vgs2  vs1
Solving for vo1/vs1 from the above three equations we get
gm1 

1


vo1
g

o 
  gm1 RL  
and

vs1
1  2gm1 

go 
 gm1 
 g

vo2


o
  gm1 RL  

2
g
vs1
1  m1 

go 
Let us define the following:
vo1
 A1 1 ;
vs1
vo1
 A1 2 ;
vs2
vo2
 A21
vs1
and
vo2
 A22
vs2
For a symmetric network A11 = A22 and A12 = A21.
We can then write
vo1  A1 1 vs1  A1 2 vs2 and
vo2  A21 vs1  A22 vs2
We can now recast the equations as
v  v 
vo1  Ad vs1  vs2   ACM s1 s2
2
v  v 
vo2   Ad vs1  vs2   ACM s1 s2
2
where
and
A11  A12 
 gm1RL 
g R 
    m1 L 
  
2
 2 
 1  2gm go  
 go RL 
ACM  A11  A12    

 2 
Ad 
The signal (vs1 – vs2) is called the differential signal
and the signal (vs1 + vs2) is called the Common Mode
signal. This leads us to define a very important
parameter defining a differential Amplifier, the
Common Mode Rejection Ratio, CMRR. CMRR is
defined as
CMRR 
Ad
ACM

1  2gm1
2
go 
 gm1 

 

g
 o 
Having evaluated the gains of differential and
common mode gain an interesting fall out is what is
popularly known as half circuit equivalent.
This has been reduced to two half circuits to evaluate
differential and common mode gain. These are
Differential
Common Mode
Half Circuit
Half Circuit
Ad 
and
A11  A12 
 gm1RL 
g R 
    m1 L 
  
2
 2 
 1  2gm go  
 go RL 
ACM  A11  A12    

 2 
The load Resistance used in a CMOS circuit could be
each a PMOS transistor in saturation with a constant
Gate to Source Voltage or Gate tied to Drain.
In both these cases the load resistance is the same for M1
and M2.
Let us now consider an nonsymmetric load on M1 and
M2 and look at the outputs
vo1 and vo2. For this circuit
we will have
g
A1 1    m1
 gm3
  1  gm go

1 2g g
m
o

g
A1 2   m1
 gm3

gm go

1 2g g
m
o





g
A22   m1
 gd
 3
  1  gm go

  1  2 gm go





g
A21   m1
 gd3

gm go

1 2g g
m
o









This will give us the outputs vo1 and vo2 as
vo1  Ad1 vs1  vs2   ACM1
vs1  vs2 
2
vo2   Ad2 vs1  vs2   ACM2
A1 1  A1 2 
where
Ad1 
and
ACM1  A1 1  A1 2 
2
;
;
vs1  vs2 
Ad2 
2
A22  A21 
2
ACM2  A22  A21 
Assuming that gm1 >> go, the values of Ad1, Ad2, ACM1
and ACM2 reduce to
 g

Ad1    m1 
 2gm3 
;
 g 
Ad2    m1 
 2gd3 
;
 g 
ACM1    o 
 2gm3 
;
 g 
ACM2    o 
 2gd3 
Now since we are interested only in single ended output
(say) vo2, we will device a useful method to use the other
output to our advantage. What we would like to do is to
connect the output vo1 to the gate of M4. We will then
get the most commonly used single ended differential
amplifier structure overleaf.
Solving for the circuit assuming gm1 , gm3 >> go, gd1, gd3
we obtain the differential and common mode gain as
Ad 
gm1
gd1  gd3 
Giving us
and


go gd1


ACM   

 2gm3 gd1  gd3  
 gm1 gm3 

CMRR  2 

 go gd1 
In our discussions so far we had considered NMOS
input devices with PMOS load devices. It is equally
likely that we may use PMOS input transistors and
NMOS load transistors.
The relative advantages of the NMOS input
differential Amplifier and PMOS input Differential
Amplifier will be seen as we go down the course.
NMOS input pair:
M5
Common Mode Input Range
Lowest common mode input voltage at gate of M1(M2)
v G1(min) = V SS + v GS3 + v SD1 - v SG1
for saturation, the minimum value of v SD1 = v SG1 - |V T1 |
Therefore, v G1(min) = V SS + v GS3 - |V T1 |
or
ISS
VG1 min  VSS 
 VTO3 | VTO1 |

v G1(max) = V
or
DD
- v SD5 - v SG1
VG1 max  VDD  VSD5
ISS

| VTO1 |

Thermal Noise


16k T
2
veq th   

'
3
2
K
P I1 W1 L1 



1
2
1



 KN' W3 L3   2 
 
  1   '
 


K
W
L

P
1
1




To reduce thermal noise we choose
 KN' W3 L3  

  1
 K' W L  
 P 1 1 
and large value of gm1.
Assume that V DD = 3V and that V SS = -3V. Using K’N = 2K’P
18 A/V2, 0.8V <VTO3 , VT1< 1.2V, find the common mode range
for worst case conditions. Assume that ISS = 100A, W1/L1
= W2/L2 = 5, W3/L3 = W4/L4 = 1, and vSD5 = 0.2V.
ISS
VG1 min  VSS 
 VTO3 | VTO1 |

100
  3
 1.2  0.8   0.25V
18
VG1 max  VDD  VSD5
ISS

| VTO1 |

100
 3  0.2 
 1.2  0.6V
5 x9
The input common mode range is -0.25V to 0.6V.
Slew Rate:
This defines the rate at
which the load capacitor
charged. In other words it
defines the rate dv/dt at
the output. Slew rate is a
measure of the output to
follow the input signal. This
is normally associated with
large
signal
property.
Under large signal, only one
of M1 or M2 will be ON and
the charging current will
be I5. This gives a slew
rate CL(VDD- VSS)/I5.
Parasitic elements in the Differential Amplifier:
CT = tail capacitor (common mode only)
CM = mirror capacitor = Cdg1 + Cdb1 + Cgs3 + Cgs4 + Cdb3
COUT = output capacitor » Cbd4 + Cbd2 + Cgd2 + CL
Noise Sources in Differential Amplifiers:
Noise can be normally modeled as a current source in
parallel to iD. This current source represents two sources
of noise, thermal noise and flicker noise. The mean square
current noise source is defined as


 8kT gm 1  
(KF) ID 
2
in  

 f
2
3

f Cox L 
where
f  Bandwidth at frequency f
KF  Flicker Noise Coefficient
  gmbs gm
The mean square noise reflected to the gate giving
mean square voltage noise at the gate



(KF)
in2  8kT 1  
2
 f
veq 


2
'
gm  3gm
f Cox WLK 
The total output noise current, is obtained by summing
each of the noise current contributions.
2
2 2
2 2
2
2
2
2
2 2
ito
 gm
v

g
v

g
v

g
v

g
v
1 eq1
m1 eq2
m3 eq3
m3 eq4
m1 eq
where
 g2 
 m3   v 2  v 2 
2
2
2
veq
 veq

v

1
eq2
eq4 
 g2   eq3
 m1 


Assu min g Identical N and P Transistors
 g2 
 m3  2v 2
2
2
veq
 2 veq

1  2 
eq3
g
 m1 


The total 1/f and thermal noise contributions can be
written as
2
'




B
2



B
K
L
p 
2
 1 N N 1 
veq 1 / f  

 K' B   L  
 f W1 L1  
 P P  3 


16k T
2
veq th   
1
 3 2K' I W L  2
1 1
P 1


where
BN,P 

KF
2 Cox KN' ,P
1


'
2


 KN W3 L3   
 
  1   '


 KP W1 L1   


1/f Noise
2
'




2
B




K
B
L
p
2

N
N
1
  
 1
veq 1 / f  

 K' B   L  
f
W
L
1 1

 P P  3 
To get the input noise for NMOS input stages
interchange BP for BN, KN’ for KP’ and vice versa.
Since BN = 5BP it is preferable to use PMOS input stage
to reduce 1/f noise with large area for M1 and M2 and
2
 KN' BN   L1 

    1
'
 K B  L 
 P P  3