MIT Subthreshold Operation

6.
012 - Microelectronic Devices and Circuits
Lecture 12 - Sub-threshold MOSFET Operation - Outline
Announcement
Hour exam two: in 2 weeks, Thursday, Nov. 5, 7:30-9:30 pm
Review
ALSO: sign up for an iLab account!!
MOSFET model: gradual channel approximation

iD
(Example: n-MOS)
0
for (vGS VT)/ 0 vDS (cutoff)
K(vGS VT)2 /2
for 0 (vGS VT)/ vDS (saturation)
K(vGS VT vDS/2)vDS for 0 vDS (vGS VT)/ (linear)
with K (W/L)eCox*, VT = VFB 2p-Si + [2Si qNA(|2p-Si| vBS)]1/2/Cox*

and = 1 + [(Si qNA/2(|2p-Si| vBS)]1/2 /Cox (frequently 1)
The factor : what it means physically
Sub-threshold operation - qualitative explanation

Looking back at Lecture 10 (Sub-threshold electron charge)
Operating an n-channel MOSFET as a lateral npn BJT
The sub-threshold MOSFET gate-controlled lateral BJT
Why we care and need to quantify these observations
Quantitative sub-threshold modeling

iD,sub-threshold((0)), then iD,s-t(vGS, vDS)
[with vBS = 0]
Stepping back and looking at the equations
Clif Fonstad, 10/22/09
Lecture 12 - Slide 1
Final comments on
The Gradual Channel result ignoring and valid for v BS " 0, and v DS # 0 is:
iG (vGS ,v DS ,v BS ) = 0, iB (vGS ,v DS ,v BS ) = 0, and
0
iD (vGS ,v DS ,v BS ) =
K
2
[vGS " VT (v BS )]
2
#
v &
K $vGS " VT (v BS ) " DS ' v DS
%
2 (
for
!
for
[vGS " VT (v BS )] < 0 <v DS

0 < [vGS " VT (v BS )] <v DS
for
0 <v DS < [vGS " VT (v BS )]
with K )
W
*
*
e Cox
and Cox
) *ox t ox
L
We noted last lecture that these simple expressions without are easy to
remember, and refining them to include involves easy to remember
substitutions:
v DS " # v DS
L "#L
K "K #
!
What we haven't done yet is to look at itself, and ask what it means.
What is it physically?
1/xDT(VBS)
G
!
*
Cox + $Si qN A 2$Si 2% p&Si & v BS
ox
1
$SiqN A
ox/tox
" # 1+ *
=
*
Cox 2 2% p&Si & v BS
Cox
Si
Si/xDT
*
*
"Si x DT
"Si t ox
CDT
CDT
B
= 1+
= 1+
= 1+ * = *
"ox t ox
"ox x DT
Cox CGB
Look back at Lec. 10.
Foil 7 from Lecture 10
MOS Capacitors: the gate charge as vGB is varied

qG* [coul/cm2]
"2
2Cox
vGB $ VFB ) (
#SiqN A %'
(
"
qG =
1+
$1**
"
'
Cox &
#SiqN A
)
"
qG" = Cox
(vGB # VT )
Inversion
Layer
Charge
+ qN AP X DT
qNAPXDT
Depletion
Region
Charge
VFB
"
qG" = Cox
(vGB # VFB )
The charge expressions:

"
, Cox
(vGB # VFB )
.
"2
. %SiqN A &
2Cox
vGB # VFB ) )
(
"
( 1+
qG (vGB ) = #1++
"
(
%SiqN A
. Cox '
*
. C " (v # V ) + qN X
/ ox GB
T
A DT
vGB [V]
VT
Accumulation
Layer Charge
"
Cox
#
for
vGB $ VFB
for
VFB $ vGB $ VT
!
VT $ vGB
for
$ox
t ox
MOS Capacitors: How good is all this modeling?

How can we know?
Poisson's Equation in MOS
As we argued when starting, Jh and Je are zero in steady
state so the carrier populations are in equilibrium with
the potential barriers, (x), as they are in thermal
equilibrium, and we have:
n(x) = n ie q" (x ) kT
and
p(x) = n ie#q" (x ) kT
Once again this means we can find (x), and then n(x) and
p(x), by solving Poisson's equation:
! d 2" (x)
q
#q" (x )/ kT
q" (x )/ kT
=
#
n
e
#
e
+ N d (x) # N a (x)
(
)
i
2
dx
$
This version is only valid, however, when |(x)| -p.

When |(x)| > -p we have accumulation and inversion layers,
and we assume them to be infinitely thin sheets of charge,
i.e. we model them as delta functions.
Poisson's Equation calculation of gate charge

Calculation compared with depletion approximation
model for tox = 3 nm and NA = 1018 cm-3:
tox,eff 3.2 nm
tox,eff 3.3 nm
We'll look in this

vicinity today.
We've ignored
sub-threshold
charge in our
MOSFET i-v
modelling thus
far.
Plot courtesy of Prof. Antoniadis
MOS Capacitors: Sub-threshold charge

Assessing how much we are neglecting
Sheet density of electrons below threshold in weak inversion
In the depletion approximation for the MOS we say that the
charge due to the electrons is negligible before we reach
threshold and the strong inversion layer builds up:
*
qN (inversion ) (vGB ) = "Cox
(vGB " VT )
But how good an approximation is this? To see, we calculate
the electron charge below threshold (weak inversion):
qN (sub"threshold ) (vGB ) = " q
$ ne
x d ( vGB )
q# (x )/ kT
dx
(x) is a non-linear function of x, making the integral difficult,
" (x) = " p +

!
qN A
2
x
x
( d)
2#Si
but if we use a linear approximation for (x) near x = 0,

where the term in the integral is largest, we can get a very
good approximate analytical expression for the integral.
Sub-threshold electron charge, cont.
We begin by saying
2qN A [" (0) % " p ]

d" (x)
" (x) # " (0) + ax where
=%
where a $
dx x= 0
&Si
With this linear approximation to (x) we can do the integral
and find
qN (sub"threshold ) (vGB ) # q
kT n(0)
kT
= "q
q a
q
$Si
n ie q% (0) kT
2qN A [% (0) " % p ]
To proceed it is easiest to evaluate this expression for various

values of (0) below threshold (when its value is -p), and to
also find the corresponding value of vGB, from
vGB " VFB = # (0) " # p +
t ox
2$SiqN A [# (0) " # p ]
$ox
This has been done and is plotted along with the strong
inversion layer charge above threshold on the following foil.
Sub-threshold electron charge, cont.
6 mV
Neglecting this charge in the electrostatics calculation resulted in

only a 6 mV error in our estimate of the threshold voltage value.
Today we will look at its impact on the sub-threshold drain current.
MOSFETs: Conventional strong inversion operation,

VGS > VT
vGS > V T +G
S
n+
vDS > 0 D i
D
n+
p-Si
vBS
+
B
High concentration of
electrons in a strong inversion
layer drifting to the drain
because of field due to v DS .
n-type surface channel; drift flux from source to drain

In our gradual channel approximation modeling we have assume a
high conductivity n-type channel has been induced under the gate.
MOSFETs: Sub-threshold operation, VGS <~ VT

vGS ~
< V T +G
S
n+
vDS > 0 D i
D
n+
p-Si
A small number of electrons
surmount the barrier and
diffuse to drain.
vBS
+
B
The electrons diffuse and do not

feel v DS until they get to the
edge of the depletion region.
No surface channel; diffusion flux from source to drain when vDS > 0
For any vGB > VFB some electrons in the source can surmount the
barrier and diffuse to the drain. Though always small, this flux can
become consequential as vGS approaches VT.
MOSFETs: Sub-threshold operation, VGS <~ VT

What do we mean by "consequential"?
When is this current big enough to matter?
There are at least three places where it matters:
1. It can limit the gain of a MOSFET linear amplifier.

In Lecture 21 we will learn that we achieve maximum gain from
MOSFETs operating in strong inversion when we bias as close to
threshold as possible. This current limits how close we can get.
2. It is a major source of power dissipation and heating in

modern VLSI digital ICs.
When you have millions of MOSFETs on an IC chip, even a little bit
of current through the half that are supposed to be "off" can add up
to a lot of power dissipation. We'll see this in Lecture 16.
3. It can be used to make very low voltage, ultra-low power

integrated circuits.
In Lecture 25 we'll talk about MIT/TI research on sub-threshold
circuits with 0.3 V supplies and using W's of power.
Sub-threshold Operation of MOSFETs: finding iD
Begin by considering the device illustrated below:

G
-tox
0
n+
n+
tn+
0
- Set vGS = VFB, and vDS = vBS = 0.
- The potential profile vs. y, (y) at any x between 0 and tn+ is then:
-tox
0
vGS = VFB
n+
vDS = 0
n+
tn+
x
(y)
n+
vBS = 0
B
L
0
p
Sub-threshold Operation of MOSFETs, cont.

- Now consider (y) when vGS = VFB, vBS = 0. and vDS > 0:
-tox S
0
vGS = VFB
vDS > 0
(y)
n++ vDS
n+
n+
vDS
n+
tn+
x
vBS = 0
0
B
L
- So far this is standard MOSFET operating procedure. We could

apply a positive voltage to the gate and when it was larger than VT
we would see the normal drain current that we modeled earlier.
Rather than do this, however, consider forward biasing the
substrate-source diode junction, I.e, vBS > 0

- Apply vBS > 0, keep the same vDS > 0, and adjust vGS such that
the potential at the oxide-Si interface, (0,y), equals p

+ vBS.
- Now consider (x,y):
vGS s.t. (0,y) = p + vBS
-tox
S
Electron
Injection
and
Diffusion
n+
(y)
vDS > 0
n+
n+
tn+
x
n++ vDS
vBS > 0
0
B
L
vBS
p',n'
- With this biasing the structure is
being operated as a lateral BJT!
The drain/collector current is:
De
iD /C " W t n +qn i2
e qv BS / kT #1
N Ap Leff
p+ vBS
p
n'(0+) = npo(eqvBS/kT-1)
- This is not sub-threshold operation yet.
y

- Now again make vBS = 0, but keep the same vDS and vGS so that
the potential at the oxide-Si interface, (0,y), is still > p.
- Now (x,y) is different for 0 < x < xD,
(0,y)
and xD < x < tn+:
n++ vDS
vGS s.t. (0,y) > p
G
vDS > 0
D
-tox S
-p n+
0
(x)
xD
Injection
n+
n+
tn+
x
(x)-p
p
vBS = 0
0
B
L
- Now there is lateral BJT action

only along the interface.
- The drain current that flows in
this case is the sub-threshold
drain current.
(y)
n++ vDS
n+
0
- This is sub-threshold operation!
- The barrier at the n+-p junction is lowered near the oxide-Si

interface for any vGS > VFB.
- The barrier is lowered by (x) - p for 0 < x < xD.
Plot vs y
at xed x,
0 < x < xD.
(This is the effective vBE on the lateral BJT between x and x + dx.)
-tox S
VFB < vGS < VT G
vDS > 0
(0,y)
D
n++ vDS
Injection
xD
n+
-p
(x)
n+
tn+
x
vBS = 0
0
(x)
B
L
-
The barrier lowering
vGS
(effective forward bias)
(1) is controlled by vGS,
and (2) decreases
quickly with x.
vBE,eff(x)
= [(x) p]
n+
y
(x)-p L
p
Plot vs x
at xed y,
0 < y < L.
- p
(0)
-tox
xd
p
Injection occurs over this range.

- To calculate iD, we first find the current in each dx thick slab:
G
VFB < vGS < VT
vDS > 0
-tox
0
x
x+dx
n+
n+
xD
vBS = 0
vDS > 0
y
n' ( x,0) = n i (e q" (x,vGS )/ kT #1) $ n ie q" (x,vGS )/ kT
diD (x) = q De
n' ( x,L) " n ie q# (x,vGD )/ kT
n'(x,0) " n'(x,L)

W
W dx #
De q n ie q$ (x,vGS )/ kT (1" e"qv DS )/ kT ) dx
L
L

- Then we add up all the contributions to get iD:
'
W $0
q" (x,vGS )/ kT
iD =
De & # q n ie
dx)(1* e*qv DS / kT )
L &% x d
)(
- This is what we called qN(sub-threshold) in Lecture 9 and today on Foil 7.
Substituting the expression we found for this (see Foil 7), we have:
%
W
kT
iD(sub"threshold ) =
De 'q
L
'& q
!
(
#Si
n ie q $ (0,vGS ) kT *(1" e"qv DS / kT )
2qN A [$ (0,vGS ) " $ p ]
*)
- Using the Einstein relation and replacing ni with NAeqp/kT, we obtain:

2
#
&
W
kT
1
*
iD(sub"th ) = e Cox
% (
*
L
$ q ' 2Cox
2q)Si N A
q {* (0,vGS )"[ "* p ] }
e
[* (0,vGS ) " * p ]
kT
(1" e
"qv DS / kT
- To finish (we are almost done) we need to replace (0,vGS) with vGS
since we want the drain current's dependence on the terminal voltage.
- The relationship relating (0,vGS) and vGS is:
vGS = VFB
1
+ [" (0) # " p ] + * 2$SiqN A [" (0) # " p ]
Cox
- From this we can relate a change in vGS to a change in (0), which
is what we really need. To first order the two are linearly related:
"vGS
')
dvGS
1
#
"$ (0) = (1+
*
d$ (0)
)* 2Cox
+
2%SiqN A )
, "$ (0) . n "$ (0)
[$ (0) & $ p ] )"n
- In the current equation we have the quantity {(0,vGS) - [-p]}. - p is

simply (0,VT), the potential at x = 0 when the gate voltage is VT, so
{" (0,v
GS
) # [#" p ]} = {" (0,vGS ) # " (0,VT )} = {vGS # VT } n
- Using this and the definition for n, we arrive at:
iD(sub"threshold ) #
W
*
e Cox
L
$ kT ' 2
q { vGS "VT } n kT
n
"1
e
1" e"qv DS / kT )
)
(
& ) (
% q(

- To fully complete our modeling, we must add two more points:
1. The dependences on vBS and vDS:
vBS: The threshold voltage depends on vBS. (0,VT) does also,
i.e. (0,VT) = - p-vBS, and so do the junction barriers. Taking
this all into account we find that the only change we need to
make is to acknowledge that n and VT both depend on vBS.
vDS: The drain to source voltage introduced a factor (1 - e-qvDS/kT) 1.
This is discussed in the handout posted on Stellar.
The complete expression for iD is:
iD,s"t (vGS ,v DS ,v BS ) #
W
*
e Cox
L
$ kT ' 2
q { vGS "VT (v BS )} n kT
1" e"qv DS / kT )
(
& ) [ n(v BS ) "1] e
% q(
2. The factor n:
The value of n depends on (0,vGS). Notice, however, that the subthreshold current is largest as (0,vGS) approaches -p-vBS, so it
makes sense to evaluate it there and take that as its value for all
vGS:
&
* &
*
(
1
n " '1+
*
() 2Cox
2#SiqN A ( (
1
+ - '1+ *
[$ (0) % $ p ] (, () Cox
(
#SiqN A
+
2[%2$ p % v BS ] (,
** Notice that this is exactly the same expression as that for !

- Comparing current levels above and below threshold:
The ranges of the two models do not overlap, but is it still
interesting to compare the largest possible value of the subthreshold drain current model (vGS - VT = 0 V),* with the strong
inversion model at vGS - VT = 0.06 V, 0.1 V, and 0.2 V:
iD(sub"threshold )
K
$ kT ' 2
q v "V
n kT
# & ) ( n "1) e { GS T }
% q(
(0.025)2
iD(strong inversion )
!
"
1
2
v
$
V
( GS T )
2#
0.4
0.25
(0.06)2
(0.1)2
(0.2)2
vBS = 0
= 1.56 x 10-4 V2
= 1.5 x 10-3 V2
= 4 x 10-3 V2
= 1.6 x 10-2 V2
We see that the current in strong inversion drift current quickly

becomes much larger, although only grows quadratically.
* This is pushing the model, particularly with regard to the

diffusion current model, beyond it's range of strict validity,
and is probably somewhat of an over-estimate.
- Plotting our models for the earlier device: NA = 1018 cm-3, tox = 3 nm:
vBS = 0
- Zooming into a lower current scale: NA = 1018 cm-3, tox = 3 nm:
vBS = 0

- Repeating the plot with a log current scale: NA = 1018 cm-3, tox = 3 nm:
Slope = 60 x n mV/decade*
* n = 1.25 here so 75 mV/decade
vBS = 0
Sub-threshold Output Characteristic

- We plot a family of iD vs vDS curves with (vGS - VT) as the family
variable, after first defining the sub-threshold diode saturation
2
$
'
current, IS,s-t:
W
kT
*
2
IS,s"t # e Cox
& ) [ n "1] = K o Vt [ n "1]
L
% q(
kT
Note : Vt "
,
q
log iD,s-t
!
(vGS-VT) = -0.12xn Volts
10-2 IS,s-t
10-3 IS,s-t
IS,s-t
vDS
iD,s"t (vGS ,v DS ) # IS,s"t e

W
*
e Cox
L
10-1 IS,s-t
!
10-4
Ko "
q { vGS "VT } n kT
"qv DS / kT
1"
e
(
)
Note: The device we modeled had n = 1.25, so it

follows a "75 mV rule" [i.e. 60 x n = 75].
vBS = 0
Sub-threshold Output Characteristic, cont.

- To compare this with something we've already seen, consider the
BJT and plot a family of iC vs vCE curves with vBE as the family
variable
log iC
vBE = 0.66 Volts
10+11 FIES
vBE = 0.60 Volts
10+10 FIES
10+9
10+8
vBE = 0.54 Volts
FIES
vBE = 0.48 Volts
FIES
vCE
iC (v BE ,vCE ) " # F IES e qv BE
kT
(1$ e
$qvCE / kT
- The two biggest differences are (1) the magnitudes of the IS's,
and (2) the factor of "n" in the MOSFET case. The totality of vBE
reduces the barrier, whereas only a fraction 1/n of vGS does.
!
- A third difference is that a BJT has a base current.*
* This is the price paid for having n = 1 in a BJT.
Large Signal Model for MOSFET Operating Sub-threshold

- The large signal model for a MOSFET operating in the weak
inversion or sub-threshold region looks the same model as that
for a device operating in strong inversion (vGS > VT) EXCEPT
there is a different equation relating iD to vGS, vDS, and vBS:
We will limit our model to
vGS " VT , v DS > 3kT /q and v BS = 0.
D
!
iD
iD (vGS , vDS )
G
iG (= 0)
iG,s"t (vGS ,v DS ,v BS ) = 0
S,B
iD,s"t (vGS ,v DS ,0) # IS,s"t (1" $v DS ) e
!
Early effect
q { vGS "VTo } n kT
(1" e
"qv DS / kT
1 for vDS > 3 kT/q

6.012 - Microelectronic Devices and Circuits
Lecture 12 - Sub-threshold MOSFET Operation - Summary
Sub-threshold operation - qualitative explanation

Look back at Lecture 10
(Sub-threshold electron charge)
BJT action in depletion/weak inversion layer along oxide
the interface
MOSFET gate-controlled lateral BJT
Important in/for
1. power dissipation in normally-off logic gates

2. limiting the gain of strong inversion linear amplifiers
3. realizing ultra-low power, very low voltage electronics
Quantitative sub-threshold modeling

This gives us a precise description of the voltage dependence
It also gives us the information on IS,s-t and n we need for
device design
iD,s"t (vGS ,v DS ,v BS ) # IS,s"t e

with:
IS,s"t
2
$
'
W
kT
*
# e Cox
& ) [ n "1]
L
% q(
q { vGS "VT (v BS )} n kT
&(
1
and n " '1+ *
() Cox
"qv DS / kT
1"
e
(
)
*(
#SiqN A
+ =2[$2% p $ v BS ] (,
MIT OpenCourseWare
http://ocw.mit.edu
6.012 Microelectronic Devices and Circuits

Fall 2009
For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.

MIT Subthreshold Operation

Uploaded by

Copyright:

Available Formats

MIT Subthreshold Operation

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

MIT Subthreshold Operation

Uploaded by

Copyright:

Available Formats

6.

012 - Microelectronic Devices and Circuits

Lecture 12 - Sub-threshold MOSFET Operation - Outline

ALSO: sign up for an iLab account!!

MOSFET model: gradual channel approximation

with K (W/L)eCox*, VT = VFB 2p-Si + [2Si qNA(|2p-Si| vBS)]1/2/Cox*

Sub-threshold operation - qualitative explanation

Operating an n-channel MOSFET as a lateral npn BJT

The sub-threshold MOSFET gate-controlled lateral BJT

Why we care and need to quantify these observations

Quantitative sub-threshold modeling

Stepping back and looking at the equations

Clif Fonstad, 10/22/09

[vGS " VT (v BS )] < 0 <v DS

0 <v DS < [vGS " VT (v BS )]

remember, and refining them to include involves easy to remember

Look back at Lec. 10.

Foil 7 from Lecture 10

MOS Capacitors: the gate charge as vGB is varied

The charge expressions:

Foil 8 from Lecture 10

MOS Capacitors: How good is all this modeling?

This version is only valid, however, when |(x)| -p.

Foil 9 from Lecture 10

Poisson's Equation calculation of gate charge

Clif Fonstad, 10/22/09

We'll look in this

Plot courtesy of Prof. Antoniadis

Foil 10 from Lecture 10

MOS Capacitors: Sub-threshold charge

But how good an approximation is this? To see, we calculate

the electron charge below threshold (weak inversion):

qN (sub"threshold ) (vGB ) = " q

(x) is a non-linear function of x, making the integral difficult,

" (x) = " p +

but if we use a linear approximation for (x) near x = 0,

Clif Fonstad, 10/22/09

Foil 11 from Lecture 10

Sub-threshold electron charge, cont.

2qN A [" (0) % " p ]

With this linear approximation to (x) we can do the integral

To proceed it is easiest to evaluate this expression for various

vGB " VFB = # (0) " # p +

Foil 12 from Lecture 10

Sub-threshold electron charge, cont.

Neglecting this charge in the electrostatics calculation resulted in

MOSFETs: Conventional strong inversion operation,

n-type surface channel; drift flux from source to drain

MOSFETs: Sub-threshold operation, VGS <~ VT

The electrons diffuse and do not

MOSFETs: Sub-threshold operation, VGS <~ VT

1. It can limit the gain of a MOSFET linear amplifier.

2. It is a major source of power dissipation and heating in

3. It can be used to make very low voltage, ultra-low power

Sub-threshold Operation of MOSFETs: finding iD

Begin by considering the device illustrated below:

- Set vGS = VFB, and vDS = vBS = 0.

Clif Fonstad, 10/22/09

Sub-threshold Operation of MOSFETs, cont.

with K (W/L)eCox, VT = VFB 2p-Si + [2Si qNA(|2p-Si| vBS)]1/2/Cox