Professional Documents
Culture Documents
SJSU ScholarWorks
Master's Theses
2008
This Thesis is brought to you for free and open access by the Master's Theses and Graduate Research at SJSU ScholarWorks. It has been accepted for
inclusion in Master's Theses by an authorized administrator of SJSU ScholarWorks. For more information, please contact scholarworks@sjsu.edu.
A Thesis
Presented to
The Faculty of the Department of Electrical Engineering
San Jose State University
In Partial Fulfillment
of the Requirements for the Degree
Masters of Science
by
Vincent Wall
December 2008
INFORMATION TO USERS
The quality of this reproduction is dependent upon the quality of the copy
submitted. Broken or indistinct print, colored or poor quality illustrations and
photographs, print bleed-through, substandard margins, and improper
alignment can adversely affect reproduction.
In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if unauthorized
copyright material had to be removed, a note will indicate the deletion.
UMI
UMI Microform 1463411
Copyright 2009 by ProQuest LLC.
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
ProQuest LLC
789 E. Eisenhower Parkway
PO Box 1346
Ann Arbor, Ml 48106-1346
2008
Vincent Wall
ALL RIGHTS RESERVED
t '"-V
l-v,'V
4-
Date
7/^
//
' ^:JetC&4fa- ^
Date
i1
\ A
/7/20S
Date
yijijti
associate Dean
Date
ABSTRACT
By Vincent Wall
Transistors do not operate like ideal current sources when they are in the
saturation region, using the following equation id = ^
The reason for this in MOS devices is channel length modulation, modeled
approximately by (1 + AVds). Channel length modulation is caused when the voltage on
the drain is increased, which increases the width of the drain depletion region, thereby
shortening the channel, and increasing the current through the drain. This directly
decreases the output resistance of the transistor. Standard 0.18u MOS transistors have a
lambda, in the saturation region, between 75mV"' and 21mV"1. A Super MOS transistor
circuit negates this effect by using negative feedback to stabilize the drain current,
thereby reducing channel length modulation. Multiple Super MOS circuits and regular
MOS transistors have been designed, fabricated, and tested using TSMC's 0.18 process.
The results have shown that X is reduced up to a factor of 4 with minimal reduction in
drive when compared to a minimally sized 0.18u process transistor (W=270nm,
L=180nm).
DEDICATIONS
To my mother who has drilled into me since I was a toddler the importance of an
education. And to my son, Jacob, who is the reason that I did not give up three fourths of
the way through my master's degree.
ACKNOWLEDGMENTS
I would like to thank Professor Parent for being more than a Professor and both
Professors Lili He and Sotoudeh Hamedi-Hagh for being a part of my thesis committee.
In addition I would like to thank the following Professors: Papalias for answering all my
dumb questions, Freeman for letting me into the program, and Singh for believing that I
would finish the program. Lastly I would like to thank Savander Parker for being the
man to go to when a little help was needed and Giri and Liz Venket for giving me
support.
VI
Table of Contents
Page
List of Tables
List of Figures
xi
1. Introduction
1.1. Background
2. Development
9
9
13
15
19
19
19
21
2.2.1.3. AC Response
22
vii
Page
2.2.2. P-Type MOS
24
24
27
2.2.2.3. AC Response
28
3. Applications
30
30
30
32
36
3.2.1. Super3tOpAmp
37
3.2.2. Superl3tOpAmp
39
41
41
47
50
51
53
55
58
60
64
Page
4.4.2. Comparison Super PMOS Current Mirrors
69
74
Works Cited
75
IX
List of Tables
Page
Table 1: Average Lambda for NMOS, Super3tNMOS, & Super 13tNMOS
20
21
23
24
27
29
36
37
50
52
Table 11: Average Lambda for Testing Results of PMOS, Super3t PMOS,
&Superl3tPMOS
55
58
List of Figures
Page
Figure 1: Cascode Amplifier
Figure 5: Leafcell Simulation Results: Super3t NMOS vs. NMOS; w=6.4u, l=6.4u
10
11
14
16
20
Figure 10: Id vs. Vg & Gm for NMOS, Super3tNMOS, & Super 13t NMOS
22
Figure 11: Frequency Response for NMOS, Super3t NMOS, & Superl3t NMO
23
25
26
Figure 14: Id vs. Vd for PMOS, Super3t PMOS, & Super 13tPMOS
27
Figure 15: Id vs. Vg & Gm for PMOS, Super3t PMOS, & Super 13t PMOS
28
Figure 16: Frequency Response for PMOS, Super3t PMOS, & Superl3t PMOS
29
31
31
xi
Page
Figure 19: lout vs. Iref of Super3t NMOS vs. Superl3tNMOS
32
Figure 20: lout vs. Iref of Super3t PMOS vs. Superl3t PMOS
33
34
35
38
40
42
43
44
45
46
48
49
Figure 32: Id vs. Vd Testing Results for NMOS, Super3t NMOS, & Superl3t NMOS.. 51
Figure 33: Id vs. Vg & Gm Testing Results for NMOS, Super3t NMOS,
&Superl3tNMOS
52
53
54
Figure 36: Id vs. Vd Testing Results for PMOS, Super3t PMOS, & Superl3t PMOS.... 56
Figure 37: Closeup - Id vs. Vd Testing Results for PMOS, Super3t PMOS,
&Superl3tPMOS
57
xii
Page
Figure 38: Id vs. Vg & Gm Testing Results for PMOS, Super3t PMOS,
&Superl3tPMOS
59
60
61
62
63
Figure 43: lout vs. Iref- Testing Results - Super3t NMOS Current Mirror
65
Figure 44: Delta (Iref/Iout) - Testing Results - Super3t NMOS Current Mirror
66
Figure 45: lout vs. Iref- Testing Results - Superl3t NMOS Current Mirror
67
Figure 46: Delta (Iref/Iout) - Testing Results - Superl3t NMOS Current Mirror
68
Figure 47: lout vs. Iref- Testing Results - Super3t PMOS Current Mirror
70
Figure 48: Delta (Iref/Iout) - Testing Results - Super3t PMOS Current Mirror
71
Figure 49: lout vs. Iref- Testing Results - Superl3t PMOS Current Mirror
72
Figure 50: Delta (Iref/Iout) - Testing Results - Superl3t PMOS Current Mirror
73
XIII
CHAPTER 1
INTRODUCTION
1.1. Background
With transistor gate lengths one micron and smaller, the gain of analog circuits
is severely degraded due to large channel conductance. Cascoding of transistors can
and has been used to overcome the large channel conductance leading to greater gain
and accuracy in the circuits [1]. The gain produced by this technique of cascoding
transistors is too small to be used in several applications, Op Amps being one of these
[2]. Taking a look at the small signal equation for a common source amplifier,
equation 1.0, we see that there are two ways to increase the gain of the circuit.
Vout
The first way is to increase the transconductance, and the second would be to
increase its output resistance. Transconductance (gm) is calculated by equation 1.1b
and it is the ratio of the output current to input current.
_ lout
<10b)
gm
~~m
1
r =
Super MOS circuits greatly improve the output resistance of the transistor by
decreasing lambda. This is done by using a third transistor to provide feedback to the
1
drain transistor. Both types of Super MOS discussed in this paper, Super 3t and Super
13t, improve channel length modulation.
1.2. Gain Boosting Technique
One way to increase the gain of a MOS device is to create a cascode circuit
(Figure 1) yielding a gain of [2]:
AV
= gmirol(gm2ro2+1)
(1-1)
Vdd
Vss
Continuing to use the cascode method to increase the dc-gain of the circuit
would quickly destroy the output swing of the circuit, since each transistor would need
to have a Vt across it. To overcome this, one needs to increase the output resistance
(decreasing lambda) of the circuit. This can be done by using a third transistor as an
2
amplifying feedback loop to the cascading transistor as was done in the Super3t MOS
model in Figure 2. This creates a gain [2] as described in equation 1.2 with an output
resistance as described in equation 1.3.
A v
= gmirol(gm2ro2(gm3ro3
out = rol(gm2ro2(gm3ro3
+ l) + l)
(1.2)
+ ! ) + !) + ro2
(L3)
drain
Iin
Nl
K2 4
qnd
\7
gate
NO
source
X7
One could get a higher dc-gain by adding an Op Amp to the gate of the cascoding
transistor connecting the drain of the gate transistor to the OpAmp's negative input.
The positive input of the OpAmp is connected to a Vref, and its output connects to the
gate of the cascading transistor. This circuit uses the same equations as the Super3t
MOS, 1.2 and 1.3, substituting Avopamp for gni3ro3 [2].
With the gain-boosting principle one can overcome the inherent limitations of a
MOS device by increasing the output resistance and dc gain of the cascaded circuit.
The new limitations on the gain of the transistor are now set by the following factors:
leakage current, weak avalanche, and thermal feedback [4].
1.3. Low Output Conductance Composite MOSFET's
To increase the output resistance (Rout) of the transistor, one could use the circuit
design technique that C. Galup-Montoro, M. C. Schneider, and I. J. B. Loss
demonstrated in their paper "Low Output Conductance MOSFET's for High
Frequency Analog Design." They demonstrated a technique to create transistor arrays
that were wider at the drain than at the source [5]. Their proposed transistors could be
designed to increase the ratio of transconductance-to-output conductance over that of a
short channel transistor. The trade-off in their design would be a slight penalty in its
signal swing [5]. The Super MOS transistors incorporate their design technique by
making the width of the drain transistor larger than the width of the gate transistor.
1.4. Laterally Diffused Implanted MOS Transistor
Based on the finding of Basham [6], a non-circuit way to reduce lambda would
be to use another family of transistors. The family that shows the greatest potential to
increase RoUt is the laterally diffused implanted MOS transistors (LDMOS). LDMOS
is created by an additional well or shifted well that, when diffused, creates an
asymmetric device. The devices were first introduced as a method of exceeding the
limits of photolithography. At that time its analog capabilities were sparsely
researched. This was in part due to the complexity in determining the position of the
junction and its threshold voltage. These difficulties have been largely alleviated by
the tight process controls required for submicron devices. It can be shown that the
4
initial channel current, in LDMOS, is significantly reduced, and that Lambda, channel
length modulation, is reduced. Future studies should compare and contrast the
performance and trade-offs of LDMOS transistors versus Super MOS transistors [6].
CHAPTER 2
DEVELOPMENT
2.1. Super MOS
Super MOS circuits are one way to decrease the effect of channel length
modulation, Lambda (X) is used to model channel length modulation. Lambda can be
calculated by using the approximate calculation, equation 2.1, or by the more exact
formula equation 2.2 [7].
Aid
A=
x =
@Vg
Id-AVd^
did
dVds
(did \
n u
(21)
(12)
id-\^L
\dVds jvds
The channel length shrinks because the drain voltage increases, which increases
the drains depletion width, and it forms a junction with the substrate. This shortens the
"effective" channel length [8]. When the channel length shrinks and the drain current
increases, lambda becomes 1/Vd at a specific gate voltage. This effect can clearly be
seen in current mirrors. In a current mirror, Figure 3, the biasing voltage Vg is tied to
Vd, and any changes in Vd will result in a shift in the output current (Id) of the device,
even though Iref is at a constant current. In Figure 4 one can see that even though Iref
has a constant value Iout's value fluctuates with Vd.
^7
K?
lref=500uA
500u
IrefMOOuA
400U
lref=300uA
300U
lref=200uA
200U
lref=100uA
100U
0.0
1.0
2.0
3.0
Vds ( V )
5.0
This is quite different for the Super MOS. The Super MOS has a smaller value
for X, which means that it is stable for all Vds in the saturation region. This translates
into a stable Id for all values of Vds in the saturation region. These curves and their
properties will be discussed below.
There are several different types of Super MOS's. They have been named
according to the number of transistors that are needed to create the circuit. These
circuits are 3t, 13t novel, 13t patented, and 12t, to name a few. This thesis will be
focusing on the 3t and 13t novel (hereinafter 13t) [2], [3]. All of the circuits come in
both P-type and N-type varieties. A comparison and analysis of these two Super
MOS's versus a regular 0.18u MOS will be shown.
The design of the two Super MOS's was done to show how well they compare
to a regular transistor of minimal sizing. To address this it was decided to design the
Super MOS's so that their current driving capabilities matched, as closely as possible,
the curves of a regular MOS; W=270nm and L=180nm. They could have been
designed to produce the best Super MOS by focusing on a greater gm and lower Vt,
but it was felt that it would have provided a false comparison since their Id vs. Vds
curves would no longer match. For fabrication, TSMC's 0.18u deep process, provided
by MOSIS, was used.
A side benefit of the Super MOS circuits is that an analog designer will be able
to use these Super MOS designs as a standard cell. To show how this can be done,
several current mirrors and two OpAmps were designed and tested with no changes to
their Super MOS components.
8
700u
3.0m ,
Super3t NMOS
Vgs=5V
600u
500 u
'
+00U
"
Vgs=5V
2.0m
Vgs=4V
Vgs=4V
-X,
Vr-
300u
Vgs=3V
Vgs=3V
1,0m
200u
; //_.-''
: / '
0.00
Vgs=2V
Vgs=2V
100U
i'/'
*
'
r',--f-!--.
" "
Vgs=1V
,--!?-,-
t--.-Jh-i--r-S--.--r-,;-,
Vgs=1V
,
't-T-|--j"|--
3.0
Vdf v )
3.0
Vd ( V )
6.0
Figure 5. Leafcell Simulation Results: Super3t NMOS vs. NMOS; w=6.4u, l=6.4u
Not limiting the voltage on I;n led to an interesting result when the circuit was
fabricated and tested: when Vds approached 3v the current rose in an exponential
fashion. Graph 2 shows this behavior.
The discrepancies seen between simulation and testing of this Super3t NMOS
are twofold. First, the exponential rise in Id at the tail end of Vds did not match he
simulation and led to an increase of A,. During testing, Vds was extended out to five
volts (these results were not included) and the exponential rise in Id continued. One
notes the rise in Lambda followed it. The second discrepancy was the inverted slope in
the linear region as seen in simulation, which was not seen in the testing results of
10
Figure 6. Both of these issues were resolved with the addition of the current source
onto the Super 3t MOS, which limited the maximum voltage at the I;n node.
,,
| m
^.,,,..,,, , , , , , , , , , ,
,,
. S
Vds(V)
|
VG=0V
VO0.33V
VG=0.E6V
VG=1.DV
VG=1.33V
VG=1.B7VVG=2 .OV*VG=2.33V V Q 2 6 7 V * V G = 3 . 0 V |
(2.3)
Since the goal is to mimic a regular MOS, Id could not be a design variable. In
addition (Vgs-Vt) is not available for to design with, since it will not be known what
Vgs a future designer will use. This leaves only the width and Iin, of the Super MOS,
11
available as a design variable: if one wanted a larger length a different process would
have been used.
In Figure 2, transistors Nl and N2 create a feedback loop. Their interactions
with each other can be viewed in the following equations:
Id = ^^^-{Vgsl-Vt)\\
Iin =
X{Vdsl)
junCox w.
f(Vgs2-Vty(l+A2Vds2)
Vgs2 = Vds0
(2.4)
(2.5)
(2.5b)
/jnCox
Wj
-f(Vgsl-Vty(l
+ Al(ydd-Vgs2))
(2.6)
(2.7)
I
W2
1
-1
(Vgs2 - Vi)
(2.8)
If we take a look at equation 2.8 we can see that width of transistor N2, W2, is
inversely proportional to Vds2, and Vds2 is a squared component of Id. The more we
increase the width of W2 the lower Id will be by the square of Vds2. W2, has the most
direct effect on the slope of the Id vs. Vd curve of the system in the saturation region,
and is it is directly responsible for lambda's value. Care must be taken in choosing its
12
value because if it is increased too much then it will start to take over as the current
limiting variable of the system. Transistor NTs width, Wl, has the most effect on the
curve in the linear region and the Vt of the circuit. Looking at equation 2.9, if the
width, W, is increased significantly, then Nl's overdrive voltage will shrink, thereby
affecting the bias voltage.
Vod-
2U
<2'9>
/mCoxf
13
vps
drain
Plfvps
vpsJP0
H d18P
1 R
18P ^ 1 net102 net 1021 ^
w=540,0n
w=540.0n
C*-
+C
l = 180.0n
1=180.0n
n
e
t
1
0
6
l
m
=5
dra InjN'1
J
6
l
m:Unet102
d18N
net106
w=540.0n
= 180.0n
I I fe I
IOW
*
n e t 0 1 0 l l m = 10
m
N2^net106
d18N
' net0101
w=540.0n
P2fnet102
l=180.0n
d18P
_ m=5 vnps
w=540.0n
1
net010lfN0
l = 180.0n
ga i e gate ^ d18N
" 1H
1
m:1 iivv rn p s
w=270.0n
= 180.0n
-elm
sourceinn:1
vnps^L
source
Figure 7. Super3tNMOS
14
15
(2.10)
(2 .11)
16
Biasing is achieved with transistors N5-N6 and P0-P1. If one takes a look at
transistors, Nl, N2, and N5 form a loop. This loop ensures that the drain transistor,
NO, is always biased at 500mV-600mV, allowing the output to swing almost rail to
rail. The loop equations are below [2].
Vdsx=Vgs5-Vgs2
(2.12)
If+r'-ft+vw
k--= fmCox
|2/ 5
J2/ 2
V*5
V*2
J2/2
(2.13b)
(2.14)
|2/ 2
(2.15)
V*2
(2/ 2
\ * 2
V*2
(2.16)
s 0.414
s=Ii
(2.17)
and k5 =^ki
(2.18)
Thus, the ratio of the currents 10 and 12 can be found by solving for Vdsl
(equations 2.19 - 2.24), comparing equation 2.24 to equation 2.17, and solving for 12.
17
(2.19)
Fds, > Vgsx -Vt
-
(2.20)
VdSi=Vgsr -Vt
(2.21)
= Vgsx-Vt
(2.22)
= \2LL + Vt-Vt
(2.23)
Hi
(2.24)
/c 0
0.414 p . = >2I
k2 y K0
(2.25)
0A14yfc = j2I^,k2=k0
(2.26)
V^ = # T
(2-27)
0.414
0-17._/ 0
2
/2
(2.28)
0.085=^
(2.29)
From what has been found from numerous simulations of this circuit, the
biasing stage does not affect X as viewed from the drain of NO. However, it does affect
the max current out of that drain.
18
*10
4K#S*? if"'
H9S3JL~..
f.S
Rout =
1
A*Id
(2.30)
Table 1. Average Lambda for NMOS, Super3t NMOS, & Superl3t NMOS
Regular NMOS
Super3tNMOS
Superl3tNMOS
Ave Lambda (sat)
0.149768V"1
0.00080065V"1
0.004941V"1
66.769 Kohm
12.489 Mohm
2.024 Mohm
20
1,8
Superl3tNMOS
0.573V
0.630V
0.685V
Vt @ Vsb=0V
gm max @Vsb=0V
2.53*10" S
2.45*10" S
21
1.67*10"5S
x10
2.5
Super3tNMOS
Figure 10. Id vs. Vd &Gm for NMOS, Super3t NMOS, & Superl3t NMOS
2.2.1.3. AC Response
Simulation of the AC response was done with the transistors in a common
source configuration. From the Figure 11, we can see that both the Super 13t and
Super3t transistors have a unity gain of approximately 775 MHz, while a standard
transistor has a unity gain of approximately 46.5 GHz (the y-axis was calculated by the
following equation: 20*log(Vout/Vin)). Table 3 lists the relevant information for the
three transistor's AC response.
22
46.5 GHz
775 MHz
775 MHz
39.87 degrees
43.46 degrees
39.13 degrees
8.04 GHz
134 MHz
335.5 MHz
Max Gain
12.78 dB
15.32 dB
14.02 dB
10
\ \
"""""N
20
\
approx 775 M.hz
'I
\^
s
S-20
\ \
\
-40
-60
1
.an
,m , ,
'
J.
10
15
Log(Freq)
.... AvMMOS
Av SuperSl NMOS
25
30
L_
20
Av Super) 3t NMOS
Figure 11. Frequency Response for NMOS, Super3t NMOS, & Superl3t NMOS
23
2.2.2. P-TypeMOS
The P-Type MOS's that will be discussed are the same types as the N-Type
MOS. These are the standard PMOS, Super3t PMOS, and the Superl3t PMOS.
Figures 12 and 13 are schematics of the Super3t PMOS and Super 13t PMOS
respectively.
2.2.2.1. Id vs. Vd -Lambda
Figure 14 shows the simulation results for the PMOS variety of the three
transistors. The most notable difference between the PMOS's Graph and NMOS's
Graph is that the Super 13t PMOS transistors Id curves do not follow close at all to the
regular PMOS's curves. The Super3t PMOS also does not follow as closely to the
regular PMOS curves as does it Super3t NMOS counterpart. Looking at Table 4 we
see that the lambdas are not as low as the NMOS circuits; meaning that the PMOS
output resistance is slightly higher than the NMOS. Table 4 lists the values Lambda
and Rou,.
Table 4. Average Lambda for PMOS, Super3t PMOS, & Superl3t PMOS
Regular PMOS
Super3t PMOS
Ave lambda (sat)
Ave Rom (sat) @ 100 uA
Superl3tPMOS
-0.36208V"
-0.00010664V"
-0.09160 V"1
27.618 Kohm
93.773 Mohm
100.17 Kohm
24
source
vps
vsg
sourceTP0
vs.- ' "* dl8P
w=270.0n
1=180.0n
net15m;
P2Tvps
, net15
w=540.0n
l = 180.0n _ ,
m:llnet9
d18P
N 2 ^ vvps
p
d1SN
w=540.0n
l=180.0n ^_i
vps
net15' 'PI
d18P
w=540.0n
l=180.0n
d r a i n l m = 10
Nlfnet4
d18N
net4
w = 540.0n
1=180.0n
m:1 vnps
$
net9' N 0
drain
net4
w=540.0n
^ l = 1S0.0n
vnpsi m:1
vnps
Figure 12. Super3t PMOS
25
vnps
N2 vnps
net069
dl3N
net078
neta79: m=2
w=270.0n
l=180.0n
N51
d1SN
net 12
net076
270.0n
l=1Ba.0n
m:1, net076
1
vnps N3
| | diaN
net069 ~ 18N
1.35u
ILi=ia
iae.0n
net069irn=3
net069
net076
drain
ISM
=1.35u
180.0n
m=2
net023.
drain
irainTPI
3 ll^dlSP
net023.
O
w=1.35u
l l _ , l=180.0n
m
=3
neS86
P4 net079
et079
*=270.0n
d18P
Ij^
l=1S0.ai
m=3
P5
'5 Tnet078
d1BP
w=270.0n
l = 180.0n
m::1 J|Source
P3 net076
et079
w=270.0n
l=180.0n >
m:1 l n e t 0 8 6
d1SP
net023
net079
P2
tSP
Nil!
-270,0n
lae.csn
net017im:1
17I
net017TP6
gatej r dl8P
-2|K3
w=270.0n
1 L l=180,0n
source Xm:1
P0 Tnet0l
net0S6
d1SP "*1 I n ,ate
w=270.0n
t?
I=180,0n J r
m:llsour<
gate
Figure 14. Id vs. Vd for PMOS, Super3t PMOS, & Superl3t PMOS
SuperBtPMOS
-0.751V
-0.720V
-0.621V
-6.51*10-6S
-6.38*10"6S
-2.07*10"6S
27
x10
**"*'**
_*--**-**-*-
Jf
-3f
jr
-5
.*...X
&-&
-1.8
f
JC
-1.0
-1.4
-1.2
-1
Vgs(V)
-0.8
- gm PMOS
-0.6
-
-0.4
-0.2
gm Sup3tPMOS -
Figure 15. Id vs. Vg & Gm for PMOS, Super3t PMOS, & Superl3t PMOS
2.2.2.3. AC Response
The AC Response of a Super PMOS is different from those of a Super NMOS.
Looking at Table 6, one of the first differences is that the unity gain frequency is 175
MHz lower for the Super 13t PMOS and 475 MHz lower for the Super3t PMOS than
for their Super NMOS counterparts. The maximum gain possible is also about half has
much as N-type MOS. But, if we compare the Super PMOS AC response to that of a
regular PMOS we can clearly see that the major differences are in the max frequencies.
This can clearly be seen in Figure 16. If we compare Figure 11 and Figure 16, it is
seen that the Super3t NMOS has a higher gain. In Figure 16 however, the Superl3t
28
PMOS has a higher gain. In Figure 16, the Super 13t PMOS has a unity gain frequency
300 MHz higher than the Super3t PMOS. In contrast, the N-type Super MOS had
approximately the same unity gain frequency. The difference in the unity gain
frequencies is due to the different size of the N-type and P-type Super MOS. This is
because the Id vs. Vds curves of the N-type and P-type Super MOS are not the same.
Table 6. AC Characteristics for PMOS, Super3t PMOS, & Superl3t PMOS
Regular PMOS
Super3t PMOS
Superl3tPMOS
Unity Gain Freq
14.4 GHz
300 MHz
600 MHz
47.31 degrees
46.84 degrees
42.00 degrees
8.04 GHz
187.5 MHz
336.5 MHz
6.82 dB
6.68 dB
7.09 dB
14.4 GHZ
15
"" to PMCS -
"*waiptfl34FMe6|
Figure 16. Frequency Response for PMOS, Super3t PMOS, Superl3t PMOS
29
CHAPTER 3
APPLICATIONS
3.1. Current Mirror Comparisons
The most important property of a current mirror is the ability of the output
current (Iout) to mirror a reference current (Iref). The Super3t type MOS was better in
both of these characteristics.
3.1.1. Comparison of Super3t NMOS vs. Superl3t NMOS
Figure 17 and 18 are the block diagrams of the Super3t NMOS and Superl3t
NMOS current mirrors. In simulation, the Super3t NMOS was able to track Iref
through a range of OA to 200uA, whereas the Super 13t NMOS cutoff at lOOuA (Figure
19). This occurs because of the transresistance component of the circuit. The issue
lies with the transistors P2-P4 in Figure 8. As more current passes through transistor
Nl, the voltage on the source of N2 rises. This allows less current to flow through
transistors P2, P5, and N2, reducing the voltage at the gate of NO, and turning off the
circuit. Output current to reference current linearity of the Super 13t NMOS is also
less than the Super3t NMOS. Looking at Figure 19, we see that the Super3t NMOS
tracks extremely well except for when Vds goes below 0.72V.
30
iref
iout
vps
10i
drain
vps
super_3t_nmos
super_3t_nmos
gate
source
rill
drain
vps
gate
vnps
vnps
source
vnps
iref
vps
tout
,10
11
o
gate
gate
X>
super_13t_nmos
super_13t_nmos
^
a.
vnps
^ ^
31
x10
,...
1.7SH
1,5
-iref
*.-- Nm0s3Tlout@wl$=O,36
. ^ - Nmos3Tlout@vds=0.72
*-'-- Nmas3Tlout@vds=1.08
^^.
- Nmo$3TIout@wl8=1.44
- * - Nm&s3Tlout@vds=1 8
-~~- * Nmosl 3Tteut@vds=0.36
- Nmos13Tlouvds=0,72
-> *
Nmosl 3Ttout@vds*1,08
., . , Nmosl 3TIout@vds= 1,44
# Nmosl 3Ttoirt@vds=1.8
s-
1.25
. S&..;.
0.78
0.25
Figure 19. lout vs. Iref of Super3t NMOS & Superl3t NMOS
32
21 and 22 are block diagrams of the Super3t PMOS and Super 13t PMOS current
Mirrors.
ft 10
<
r*
-0.25
-0,5
-0.75
-1
Pme3tlQut@vds=Q
..._ pmos3llout@vds=0,38
-- * Pmos3tloul@vds=0.72
--*-- Pmos3tlout@vds=1.08
..___ pmo83tlout@vd*=1.44
* Pmos3tlout@vdS=1,8
Pmos 13tlout@vds=0
Pmos13tlout@vds=0.36
* Pmos13tloutvds=0.72
* Pmos13ttou*@vds=1,Q8
* Pmos13tlout@vds=1 44
* Pmos13ttoyt@vds=1.8
^JK
-1,25
-1.6
-1.75
J>
sir
Jr
r
-1.75
t
-1.5
i
-1,25
i
-0.75
-05
xW
Figure 20. lout vs. Iref of Super3t PMOS & Superl3t PMOS
33
vps
rr
101
source vps
vps source
vag
veg
super_3Lpmos
super_3t_pmos
drain vnps
vnps drain
vnriis
tout
fref <
34
11
iref
iout
vps
lI0
11
CO
"O
CL
a.
>
>
gate
gate
super_13Lnmos
super_.13t- nmos
CD
U
urc
CD
0]
>
CO
"a
O-
c
>
vnps
35
01
Unity Phase
(degrees)
VoutdB@Cl=lpF
1.25E+07
6.55E+07
86.2
VoutdB@Cl=10pF
1.58E+07
3.03E+07
67.56
VoutdB@Cl=100pF
1.10E+05
2.38E+06
86.46
VoutdB@Cl=100uF
1.56E-01
1.74E+00
66.64
>1
>1
13.14
Vout dB@Cl=lmF
36
Unity Phase
(degrees)
VoutdB@Cl=lfF
1.14E+08
5.98E+08
64.43
VoutdB@Cl=10fF
1.14E+08
5.98E+08
64.55
VoutdB@Cl=50fF
1.14E+08
4.37E+08
70.30
VoutdB@Cl=100fF
1.14E+08
4.37E+08
69.28
Vout dB@Cl=lpF
6.55E+07
1.98E+08
67.71
VoutdB@Cl=10pF
7.18E+06
3.77E+07
65.75
37
vnps
source vps
vin_n
vln_p ^ ^
vin_p
vps source
vps source
V
*uper_3t w pmo
s*ipHjr_3t_pmtj5
vout
drain vnps
vnps drain
vnps drain
IjnwlBriP l - * ) |
l=1B9.0n
nrf25l'
vps source
5uper_J5Lpmu5
vnps drain
drain
vps
vps
drain
super_3t_nmoE
gate
Euper_3t_nmoB
super_3t_nmos
gate
vnps
source
00
39
tttw
Figure 24. Superl3 OpAmp
40
CHAPTER 4
TESTING RESULTS and COMPARISONS
The testing of the extracted values and figures of the all the circuits was done
by measuring five of every fabricated circuit and averaging the results.
4.1. Testing Station.
All of the circuits were measured by an HP4156 Parameter Analyzer. A
custom probe card, see Figure 25, along with a Signatone probe state completed the
interface to the circuit containing die. Figure 26 shows the complete testing Station,
while Figure 27 shows a picture of the die. The die measured 1.5mm x 1.5mm and
contained all 14 circuits. The biggest structures visible in figure 27 are the pads that
the probe attaches to. Figure 28 is the layout of the pads that the test card interfaced
with. Each pad measures 72um by 72um. The entire five pad layout measures
287.19um by 242.865um. Figure 29 is the layout of Figure 26.
41
42
n
13
:?
j,
*
J
* *
f inn
-*
iff
**'
43
i D D'D
do do
44
111111T" II lllfTFiili
45
S-Btn 6-gtn
GKO
D-6tp G-Btp
S-Btp
sTBfp"
Smfiol
GTKp
<*D
DTHp
46
v*d
Vr
0T
vpSFE
4.2. N-TypeMOS.
The results of testing the N-Type MOS along with how each of the three
circuits compares to each other will be discussed in the below sections. Figures 30 and
31 are the circuit layouts for the Super3t NMOS and Super 13t NMOS that were tested.
The layout for the Super3t NMOS measures 8.505um by 12.6 urn. The layout for the
Superl3tNMOS measures 9.27umby 13.995 urn.
47
48
49
0.215648 V"1
0.058099 V 1
0.059602 V 1
46.371 Kohm
172.120 Kohm
167.780 Kohm
50
x10
2.25
2
1.75
1.5
1.25
1
0.75
0.5
0.25
0
0.2
0.4
0.6
NMOS
1.2
0.8
VD
Super3tNMOS
1.4
1.6
1.8
- Super13tNMOS
Figure 32. Id vs. Vd Testing Results for NMOS, Super3t NMOS, & Superl3t NMOS
51
still shows that the Regular NMOS and Super 13t NMOS are about equal while the
Superl3t NMOS is shown to be noticeable smaller. This is clearly shown in Figure 33.
Table 10. Vt & Gm for Testing Results of NMOS, Super3t NMOS, & Superl3t NMOS
Regular NMOS
Super3tNMOS
Superl3tNMOS
Vt @ Vsb=0V
0.648V
0.5760V
gmmax @ Vsb=0V
0.648V
3
151.967*10" S
157.051*10" S
106.242*10-3 S
x1(T
N M 0 S
GmNMOS
Super3tNMOS - 0 - Gm Super3tNMOS
Super13tNMOS
Gm Super13tNMOS |
Figure 33. Id vs. Vg & Gm Testing Results for NMOS, Super3t NMOS, & Superl3t NMOS
52
4.3. P-TypeMOS.
The results of testing the P-Type MOS along with how each of the three
circuits compares to each other will be discussed in the below sections. Figures 34 and
35 are the circuit layouts for the Super3t PMOS and Super 13t PMOS that were tested.
The layout for the Super3t PMOS measures 12.195um by 7.785 um. The layout for the
SuperBt PMOS measures 12.285um by 11.025 um.
53
54
(sat) @ 100 uA
-1.739 V"1
-2.200 V"1
-1.410V"1
5.75 Kohm
4.545 Kohm
7.092 Kohm
55
-x- PMOS
Super3tPMOS * S u p e r ! 3tPMOS
Figure 36. Id vs. Vd Testing Results for PMOS, Super3t PMOS, &Superl3t PMOS
56
-1.8
-1.6
-1.4
-1.2
-1
-0.8
_ l
-0.6
VD
| ....x.
P M 0 S
Super3tPMOS * S u p s r 1 3 t P M O s |
Figure 37. Closeup - Id vs. Vd Testing Results for PMOS, Super31PMOS, & Superl3t PMOS
57
-0.864V
-1.08V
_6
-0.936V
6
-95.976*10 S
-35.449*10" S
-32.319*10_6S
Table 12 also shows that the Vt of all three circuits is extremely high. When Table 12
is compared to Table 5, we see that the regular PMOS has a Vt lOOmV greater than
simulation. For the Super3t PMOS it is 260mV greater, and the Super 13t PMOS is
315mV greater than simulation predicted.
58
x10
-1.8
-1.6
-1.4
-1.2
-1
-0.8
-0.6
-0.4
-0.2
VG
I
_,,__
----$- Gm Super3tPMOS
Figure 38. Id vs. Vg & Gm Testing Results for PMOS, Super3t PMOS, & Superl3t PMOS
59
60
vWv&V^W'vWO
61
Ii!
if
62
fflSSSll
CD
CO
63
64
1.8
x10
^-0-0-^0
- Ave3tNMOSIout@vds=0.36
Ave3tNMOSIout@vds=0.72
Ave3tNMOSIout@vds=1.08
- $ Ave3tNMOSIout@vds=!44
O
0.75
Ave3tNMOSIout@vdB=1.8
1
Iref
Figure 43. lout vs. Iref - Testing Results - Super3t NMOS Current Mirror
65
x1(f
0.95
0.9
3tNMOSDelta@Vds=0.36
3tNMOSDelta@Vds=0.72
3tNMOSDelta@Vds=1.08
0 - 3tNMOSDelta@Vds=1.44
e
3tNMOSDelta@Vds=1.8
=2 0.85
8
0.8
0.75
0.7
0.25
0.5
0.75
1
lref(A)
1.25
1.5
1.75
2
xlC4
Figure 44. Delta (Iref/Iout) - Testing Results - Super3t NMOS Current Mirror
66
x10
.^^-i^^-^^X)^*a^fiwfiuQwa-g>
Figure 45. lout vs. Iref - Testing Results - Superl3t NMOS Current Mirror
67
13tNMOSDelta@Vds=0.36
13tNMOSDelta@Vds-0.72
13tNMOSDeta@Vds=1.08
13tNMOSDelta@Vds=1.44
e 13tNMOSDelta@Vds-1.8
0.25
0.5
0.75
1
iref(A)
1.25
1.5
1.75
x10
Figure 46. Delta (Iref/Iout) - Testing Results - Superl3t NMOS Current Mirror
68
69
x10
Ave3tPMOSIout@vds=0.36
Ava3tPMOSIOut@vds=0.72
- Ave3tPM0Sl0Ut(gvds=1.08
- 0 Ave3tPMOSIout@vds=1.44
O
Ave3tPMOSIoutvds=1.8
-1.75
-1.5
-1.25
-1
lref(A)
-0.75
-0.5
-0.25
x10"
Figure 47. lout vs. Iref - Testing Results - Super3t PMOS Current Mirror
70
1.2r
Figure 48. Delta (Iref/Iout) - Testing Results - Super3t PMOS Current Mirror
71
x10
Or
Ave13tPMOSIout@vds=0.36
Ave13tPMOSl0Ut@vds=0.72
Ave13tPMOSl0Ut@vds=1.08
Ave13tPMOSIout@vds=1.44
- e Ave13tPMOSIout@vds=1.8
-
-1.5
-1.25
-1
-0.75
-0.5
-0.25
lref(A)
Figure 49. lout vs. Iref - Testing Results - Superl3t PMOS Current Mirror
72
0
x1tf4
21
-2
-1.75
-1.5
-1.25
-1
lref(A)
-0.75
-0.5
-0.25
0
10-4
Figure 50. Delta (Iref/Iout) - Testing Results - Superl3t PMOS Current Mirror
73
CHAPTER 5
CONCLUSIONS AND FUTURE WORK
It has been shown that a Super NMOS circuit will improve lambda when
compared to a standard transistor; improving it up to a factor of 4. This then increases
the output resistance of the transistor by that same factor. Depending on the specific
application, the design trade-offs may not be worth it. Those trade-offs are: a severe
loss in the max frequency that can be designed for, an increase of Vt, and an increase
of area on the die.
There are several more aspects that need to be studied for these designs.
Primarily the Super PMOS circuits will need to be reevaluated as they performed
poorly during test. This was reflected in both the transistor and current mirror circuits.
Improvements in layout will increase their tested performance. The second point
would be to create real world OpAmps. The two OpAmps that were designed were for
academic purposes and would be impractical for use in a real-world application. They
were made to demonstrate how the Super MOS circuits would function and could be
used in a design. Thirdly, one would need to fully take advantage of the gain boosting
stage of the Super 13t MOS circuits. If this is done, they will be of much greater
benefit to future circuit designers. By comparison, the Super3t MOS circuits have one
characteristic that makes them superior to the SuperBt MOS: they are easier to design,
and therefore faster to implement in a final circuit.
74
WORKS CITED
[1] E. Tiiliharju , S. Zarabadi, M. Ismail, and K. Halonen. "A Novel Very-HighOutput-Impedance High-Swing Cascode Stage and Its Applications," Proceedings
of 1997 IEEE International Symposium on Circuits and Systems, 1997, pp.19761979.
[2] A. Nguyen "Super-MOS Transistor and Their Applications. A study and
Comparison of Three Recent Circuits," M.S. Thesis, Ohio State University,
Columbus, Ohio, 1997.
[3] F. Serra-Graells "Problemes de Circuits Integrats Analogies," Universitat
Autonoma de Barcelona, October 2004. [Online]. Available:
http://www.cnm.es/~pserra/uab/cia/ciaex.pdf. [Accessed: March 31, 2006].
[4] M. Ismail and T. Fiez, Analog VLSI: Signal and Information Processing,
Singapore: McGraw-Hill Book Co., 1994.
[5] C. Galup-Montoro, M. C. Schneider, and I. J. B. Loss. "Low Output Conductance
Composite MOSFET's for High Frequency Analog Design," Proceedings of 1997
IEEE International Symposium on Circuits and Systems, 1994, pp. 783-786.
[6] E.J. Basham, D. W. Parent. "Evaluation of a Double Implanted Diffused MOSFET
for Analog Operation," University/Government/Industry Microelectronics
Symposium, 200616th Biennial, 2006, pp. 125-130.
[7] R.A. Zane, "Cadence Tools: Design Example #lb: Estimating Lambda using
Variable Sweep & Waveform Calculator," September 2005.[Online]. Available:
http://ece.colorado.edu/~ecen5007/cadence/schexlb.html.[Accessed: March 16,
2006].
[8] T.H. Lee, "A Review of MOS Device Physics," September 2002. [Online].
Available:http://people.deas.harvard.edu/~jones/esl54/lectures/lecture_4/pdfs/MO
S_review.pdf. [Accessed: February 27, 2006].
[9] K. Bult and G. J. G. M. Geelen. "The CMOS gain-boosting technique," Analog
Integrated Circuits and Signal Processing, vol. 1, pp. 119-135, October 1991.
75
[10] Tiiliharju, Esa, Seyed Zarabadi, Mohammed Ismail, and Kari Halonen.
"Application Notes: A Novel Very-High-Output-Impedance High-Swing Cascode
Stage and Its Applications," Proceedings of 1997 IEEE International Symposium
on Circuits and Systems, 1997, pp. 1976-1979.
[11] J.D. Conway and G.G. Schrooten. "An Automatic layout Generator for Analog
Circuits," IEEE Proceedings [3rd] European Conference on Design Automation,
1992, pp. 252-256.
76