M.H.

Perrott
Analysis and Design of Analog Integrated Circuits
Lecture 16
Subthreshold Operation and g
m
/I
d
Design
Michael H. Perrott
April 1, 2012
Copyright 2012 by Michael H. Perrott
All rights reserved.
M.H. Perrott
A Closer Look at Transconductance
Assuming device is in strong inversion and in saturation:
2
I
d
V
gs
I
d_op
V
V
TH
V
gs_op
V
ds
> V
M
1
I
d
V
gs
NMOS
g
s
d
g
m
=
V
gs
I
d
V
gs_op
I
D
=

n
C
ox
2
W
L
(V
gs
V
TH
)
2
(1 +V
ds
)
g
m
=
I
d
V
gs

n
C
ox
W
L
(V
gs
V
TH
)
s
2
n
C
ox
W
L
I
d
g
m

I
d
q
2
n
C
ox
W/L

I
d

2I
d
(V
gs
V
TH
)
M.H. Perrott
Unity Gain Frequency for Current Gain, f
t
Under fairly general conditions, we calculate:
f
t
is a key parameter for characterizing the achievable
gainbandwidth product with circuits that use the device
3
|I
d
/I
in
|
f 0dB
M
1
I
d
I
in
NMOS
g
s
d
f
t
V
ds
M.H. Perrott
Current Density as a Key Parameter
Current density is defined as the ratio I
d
/W:
- Well assume that current density is altered by keeping
I
d
fixed such that only W varies
Maintains constant power
r
o
(i.e., 1/g
ds
= 1/(I
d
)) will remain somewhat constant
4
I
d
W
M
1
I
d
I
d
W
L
W
M.H. Perrott
Investigating Impact of Current Density
For simplicity, let us assume that the CMOS device
follows the square law relationship
- This will lead to the formulations:
- These formulations are only accurate over a narrow
region of strong inversion (with the device in saturation)
- However, the general trends observed from the above
expressions as a function of current density will provide
useful insight
5
I
D


n
C
ox
2
W
L
(V
gs
V
TH
)
2
g
m

2I
d
V
gs
V
TH
V
gs
V
TH

s
2L

n
C
ox

I
d
W

M.H. Perrott
10
7
10
6
10
5
10
4
10
3
0
0.5
1
1.5
x 10
6
Gate Overdrive versus Current Density
V
g
s

V
t
h

(
V
o
l
t
s
)
10
7
10
6
10
5
10
4
10
3
0
2
4
6
8
x 10
3
Transconductance versus Current Density
g
m

(
1
/
O
h
m
s
)
10
7
10
6
10
5
10
4
10
3
0
1
2
3
4
x 10
11
ft versus Current Density
f
t

(
H
z
)
Current Density Id/W (Amps/micron)
6
Investigate the Impact of Increasing Current Density
g
m
decreases
f
t
increases
Gate overdrive increases
g
m

2I
d
V
gs
V
TH
f
t
=
1
2
g
m
C
gs


W
W
V
gs
V
TH

s
2L

n
C
ox

I
d
W

W decreased with fixed I

V
t
h

(
V
o
l
t
s
)
10
7
10
6
10
5
10
4
10
3
0
2
4
6
8
x 10
3
Transconductance versus Current Density
g
m

(
1
/
O
h
m
s
)
10
7
10
6
10
5
10
4
10
3
0
1
2
3
4
x 10
11
ft versus Current Density
f
t

(
H
z
)
Current Density Id/W (Amps/micron)
7
Transconductance Efficiency Versus f
t
g
m
decreases
f
t
increases
Gate overdrive increases
g
m

2I
d
V
gs
V
TH
f
t
=
1
2
g
m
C
gs


W
W
V
gs
V
TH

s
2L

n
C
ox

I
d
W

Higher f
t
(faster speed)
M.H. Perrott
Weak
10
7
10
6
10
5
10
4
10
3
0
0.5
1
1.5
x 10
6
Gate Overdrive versus Current Density
V
g
s

V
t
h

(
V
o
l
t
s
)
10
7
10
6
10
5
10
4
10
3
0
2
4
6
8
x 10
3
Transconductance versus Current Density
g
m

(
1
/
O
h
m
s
)
10
7
10
6
10
5
10
4
10
3
0
1
2
3
4
x 10
11
ft versus Current Density
f
t

(
H
z
)
Current Density Id/W (Amps/micron)
8
Transistor Inversion Operating Regions
g
m
decreases
f
t
increases
Gate overdrive increases
g
m

2I
d
V
gs
V
TH
f
t
=
1
2
g
m
C
gs


W
W
V
gs
V
TH

s
2L

n
C
ox

I
d
W

Moderate Strong
M.H. Perrott
Key Insights Related to Current Density
Current density sets the device operating mode
- Weak inversion (subthreshold): highest g
m
efficiency
Achieves highest g
m
for a given amount of current, I
d
- Strong inversion: highest f
t
Achieves highest speed for a given amount of current, I
d
- Moderate inversion: compromise between the two
Often the best choice for circuits that do not demand the
highest speed but cannot afford the low speed of weak
inversion (subthresholdoperation)
Key issue: validity of square law current assumption
- The above is only accurate over a narrow range of strong
inversion (i.e., the previous plots are inaccurate)
General observations above are still true, though
9
I
D
=

n
C
ox
2
W
L
(V
gs
V
TH
)
2
(1 +V
ds
)
M.H. Perrott
A Proper Model for Subthreshold Operation
Drain current:
- Where:
- Note: channel length modulation, i.e., , is ignored here
10
I
D
= I
D0
W
L
e
V
gs
/(nV
t
)

1 e
V
ds
/V
t

n =
C
ox
+C
depl
C
ox
1.5
S D
G
V
GS
<V
TH
V
DS
I
D
V
t
=
kT
q
26mV at T = 300K
I
D0
=
n
C
ox
(n 1)V
2
t
e
V
TH
/(nV
t
)
M.H. Perrott
Saturation Region for Subthreshold Operation
Saturation occurs at roughly V
ds
> 100 mV
11
I
D
= I
D0
W
L
e
V
gs
/(nV
t
)

1 e
V
ds
/V
t

I
D0
W
L
e
V
gs
/(nV
t
)
0 0.1 0.2 0.3 0.4 0.5
0
20
40
60
80
100
120
140
Drain current Versus Vds and Vgs
I
d

(
n
A
)
Vds (Volts)
Vgs = 0.4V
Vgs = 0.42V
Vgs = 0.44V
M.H. Perrott
Transconductance in Subthreshold Region
Assuming device is in subthreshold and in saturation:
12
I
d
V
gs
I
d_op
V
ds
> 100mV
M
1
I
d
V
gs
NMOS
g
s
d
g
m
=
V
gs
I
d
V
gs_op
V
gs_op
g
m
=
I
d
V
gs
I
D0
W
L
e
V
gs
/(nV
t
)
1
nV
t
=
I
d
nV
t
Recall for strong inversion : g
m

2I
d
(V
gs
V
TH
)
I
D
I
D0
W
L
e
V
gs
/(nV
t
)
g
m
purely a
function of I
d
!
M.H. Perrott
Comparison of Strong and Weak Inversion for g
m
Assumption: I
d
is constant with only W varying
Strong inversion formulation predicts ever increasing
g
m
with reduced overdrive voltage
- Reduced current density leads to reduced overdrive
voltage and therefore higher g
m
Weak inversion formulation predicts that g
m
will hit a
maximum value as current density is reduced
- Note that the area of the device no longer influences g
m
when operating in weak inversion (i.e., subthreshold)
13
g
m
=
I
d
nV
t
g
m

2I
d
(V
gs
V
TH
)
M.H. Perrott
v
gs
g
m
v
gs
C
gs
r
o
g
mb
v
s
d
g
s
d
g
s
Hybrid- Model in Subthreshold Region (In Saturation)
Looks the same in form as for strong inversion, but
different expressions for the various parameters
- We can use the very same Thevenin modeling
approach as in strong inversion
We just need to calculate g
m
and g
mb
differently
g
m

1
n

I
d
V
t
g
mb

n 1
n

I
d
V
t
r
o

1
I
d
14
M.H. Perrott
v
gs
g
m
v
gs
C
gs
i
nd
2
r
o
g
mb
v
s
d
g
s
d
g
s
Noise for Subthreshold Operation (In Saturation)
i
2
nd
= 4kTg
dso
f +
K
f
f
g
2
m
WLC
2
ox
f
Recall transistor drain noise in strong inversion:
In weak inversion (i.e., subthreshold):
Thermal noise 1/f noise
i
2
nd
= 2kTng
m
f +
K
f
f
g
2
m
WLC
2
ox
f
Thermal noise 1/f noise
15
M.H. Perrott
Strong Inversion Versus Weak Inversion
Strong inversion (V
gs
> V
TH
)
- Poor g
m
efficiency (i.e., g
m
/I
d
is low) but fast speed
- Need V
ds
> (V
gs
V
TH
) = V to be in saturation
- Key device parameters are calculated as:
Weak inversion (V
gs
< V
TH
)
- Good g
m
efficiency (i.e., g
m
/I
d
is high) but slow speed
- Need V
ds
> 100mV to be in saturation
- Key device parameters are calculated as:
Moderate inversion: compromise between the two
16
g
m

2I
d
(V
gs
V
TH
)
g
mb

g
m
2
q
2|
F
| +V
SB
r
o

1
I
d
g
m

1
n

I
d
V
t
g
mb

n 1
n

I
d
V
t
r
o

1
I
d
Thevenin Modeling Techniques Can Be Applied to All Cases
M.H. Perrott
g
m
/I
d
Design
g
m
/I
d
design is completely SPICE based
- Hand calculations of g
m
, r
o
, etc. are not performed
Various transistor parameters are plotted in terms of
g
m
/I
d
- Low g
m
/I
d
corresponds to strong inversion
- High g
m
/I
d
corresponds to weak inversion
Once a given value of g
m
/I
d
is chosen, it constrains
the relationship between W, L, f
t
, etc. such that the
sizing of devices becomes a straightforward exercise
17
M.H. Perrott
Useful References Related to g
m
/I
d
Design
Prof. Bernhard Bosers Lecture:
- B. E. Boser, "Analog Circuit Design with Submicron
Transistors," IEEE SSCS Meeting, Santa Clara Valley, May
19, 2005,
http://www.ewh.ieee.org/r6/scv/ssc/May1905.htm
Prof. Boris Murmanns Course Notes:
- https://ccnet.stanford.edu/cgi-
bin/course.cgi?cc=ee214&action=handout_view&V_sect
ion=general
See Slides 45 to 67 in particular
Prof. Reid Harrisons paper on a low noise instrument
amplifier:
- http://www.ece.utah.edu/~harrison/JSSC_Jun_03.pdf
18
M.H. Perrott
19
Summary
CMOS devices in saturation can be utilized in weak,
moderate, or strong inversion
- Each region of operation involves different expressions
for drain current as a function of V
gs
and V
ds
- It is best to use SPICE to calculate parameters such as
g
m
, g
mb
, r
o
due to the complexity of the device model in
encompassing these three operating regions
g
m
/I
d
methodology is one such approach
- Weak inversion offers large g
m
/I
d
but slow speed, and
strong inversion offers fast speed but lower g
m
/I
d
- Moderate inversion offers the best compromise between
achieving reasonable g
m
/I
d
and reasonable speed
Thevenin modeling approach is valid for all operating
regions once g
m
, g
mb
, and r
o
are known