Dept of Elect Eng, Fac. of Eng., Port Said, Suez Canal Univ., Egypt. EM: Sherif_sharroush2003@yahoo.com
Dept of Electricity, Fac. of Industrial Edu., Suez, Suez Canal Univ., Egypt. EM: yasser@alumni.uwaterloo.ca
3
Dept of Elect Eng, Fac. of Eng., Port Said, Suez Canal Univ., Egypt. EM: dessouki2000@yahoo.com
Alex Higher Inst. Of Eng. and Tech & Fac. of Eng., Alex. Univ., Alexandria, Egypt. EM: sbadawy@ieee.org
Key Words: MOSFET transistor, ultra low power, subthreshold region, amplifier.
I.
INTRODUCTION
In recent years, the demand for the battery-operated
portable applications such as notebook and laptop computers, personal digital assistants, cellular phones, and other
portable communication devices has been increased. So, the
need increases also to reduce the power consumption. There
are several methods to reduce the power consumption such
as constant voltage scaling [1, 2], constant electric-field
scaling [3], switching activity reduction [4, 5], architectural
techniques such as pipelining and parallelism [6], and computer-aided design (CAD) issues of device sizing, interconnect [7, 8], and logic optimization [9, 10].
However, there are some applications such as portable
computing gadgets, medical electronic equipments, and
hand watches where ultra low-power consumption with low
or medium frequencies (tens or hundreds of megahertz) is
the primary requirement. As a solution to this problem,
energy recovery or quasi-adiabatic techniques [6] can be
used. However, this involves the use of high-quality inductors which is difficult to integrate [11]. So, in these applications, the use of the MOSFET transistor in the subthreshold
region, where the power-supply voltage, VDD is lower than
the threshold voltage, Vth, seems to be suitable as it reduces
the power consumption considerably. The subthreshold leakage current of the MOSFET transistor in this case will be
used as the operating current to perform the computations in
logic circuits. However, the use of the MOSFET transistor
in the subthreshold region to perform high-speed operations
is not possible since the operating current in the subthreshold region is much lower than that in the triode or saturation regions, thus requiring longer intervals of time to
charge or discharge the parasitic capacitances and thus performing the computations in digital circuits. We will,
throughout this paper, refer to the MOSFET transistor operating in the subthreshold region (with VDD<Vth) as the subthreshold transistor and refer to the MOSFET transistor operating in the triode or saturation regions (with VDD>Vth) as the
superthreshold transistor. Thus, the two transistors can be
made in the same fabrication process and with the same
dimensions, but the two terms subthreshold and superthreshold simply refer to their regions of operation.
As another benefit of operating the MOSFET transistor
in the subthreshold region, the transistor input capacitance
in the subthreshold region is much less than that in the superthreshold region [12]. On the other hand, the input capacitance in the superthreshold operation is dominated by the
gate-oxide capacitance. Due to the smaller input capacitance
and the lower-supply voltage in the subthreshold region, the
power consumption will be much less than that in the superthreshold region. The subthreshold region operation was
investigated and utilized especially in digital-circuit applications. The interested reader can refer to [12-15] for these
applications.
Impact of Scaling on the Subthreshold Operation
There is no doubt that the MOSFET transistor dimensions are scaled down in order to squeeze more devices in
the same area, thus performing more functions at the same
time interval and increasing the speed. At earlier versions of
the CMOS technologies, the power-supply voltage is maintained constant at 5 V [3], thus causing the electric field to
increase and decreasing the reliability of the devices due to
the increased probability of breakdown. So, the need arises
to reduce the power-supply voltage, VDD in order to increase
the device's reliability. Also, due to the increasing demand
for the portable devices which are power-sensitive, VDD
scales down in order to decrease the dynamic power consumption. This scaling scenario, where the power supply
and the dimensions are scaled down with the same ratio, is
known as the constant electric-field scaling.
Fig. 1 The ratio of the threshold voltage, Vth to VDD for the 0.35 m (on the
left) and for the 0.13 m (on the right) CMOS technologies.
ID
versus VGS.
Fig. 3 shows the energy-band diagram of a MOS structure with a p-type substrate biased so that Is 2I fp . At the
same time, the Fermi level is closer to the conduction band
than to the valence band at the surface, so the semiconductor
surface develops the characteristics of a lightly doped n-type
material. We would expect, then, to observe some conduction between the n+ source and drain contacts through this
weakly inverted channel. This conduction is due to the diffusion of minority carriers in the channel. The condition of
I fp I s 2I fp is known as weak inversion. A detailed study
of the conduction in this region reveals that the drain current
depends exponentially on VGS.
The following Equation is the current-voltage relationship for the subthreshold transistor [17]
vGS Vth0 JvSB KvDS
v
DS
nVT
VT
sub
0
(1)
with
Ie
1 e
I0
W 2 1.8
VT e
L
P 0 C ox
(2)
isub
I 0e
vGS Vth
nVT
v
DS
1 e VT
isub
W
P 0Cox
L
2 1. 8
VT e e
v
DS
VT
vGS Vth
nVT
(5)
III.
W
L
P 0 C ox
2 1 .8
V T e e
W 2 1 .8
V T e e
L
V GS v gs V th
V GS V th
P 0 C ox
nV T
nV T
(7)
v gs
e nV T
where isubt is the total (dc+ac) subthreshold current. Expressing the exponential term evgs in a Taylor-series expansion,
we obtain
i subt
W 2 1 .8
V T e e
L
P 0 C ox
1 v gs v gs
nV
nV T
T
V GS V th
nV T
v gs
nV T
P 0C
I sub
? isubt
......
(8)
L
I
sub
nV T
ox
2 1 .8
V T e e
V GS V th
nV T
v gs
1
nV T
v gs
I sub g m vgs
(9)
I sub
.
nVT
gm
(4)
fact,
(10)
v gs
nV
T
v
and gs ,etc can be neglected only if the
nV
iC
ISe
v BE
,
nVT
(12)
isubt
nVT
I 0e
e nVT
(13)
1
VGS Vth
nVT
I 0e
Kv DS
e nVT
nVT
KI sub
1
Kgm
nVT
This seems to be an interesting result; the output resistance, r0 of the subthreshold transistor is interrelated to the
transconductance, gm through the factor, K . This can be
expected since the subthreshold current depends exponentially on both vGS and vDS. Depending on the values of K ,
r0, the subthreshold transistor can be used as a poor or a
good current source whose value can be controlled by VGS.
This is in contrast to the current source implemented by the
superthreshold device where its stability is related to ro only
of the transistor.
Modeling the body effect
The effect of the source-to-substrate voltage, vSB on the
subthreshold current can be investigated by defining the
body transconductance factor, gmb as
wisubt
atvGS andvDS
wvBS
g mb
constan t .
Now, since
i subt
I 0e
? g
mb
I 0e
I 0e
vGS Vth K v DS
nV T
VGS Vth K V DS
nV T
V GS V th K V DS
nV T
J v BS
J v BS
e
T
nV
(a)
(b)
(c)
Fig. 4 (a) A simple small-signal equivalent circuit of the subthreshold transistor, (b) the small-signal equivalent circuit of the subthreshold transistor
taking into account the output resistance, (c) the small-signal equivalent
circuit of the subthreshold transistor taking into account the output resistance and the body effect.
J v S v B
nV T
nV T
J
nV
(15)
(16)
I sub
Jg m
nV T
The body-transconductance parameter is also interrelated to the transconductance, gm. Now, refer to Fig. 4 for
the small-signal equivalent circuit of the subthreshold transistor. Fig. 4 (a) illustrates the equivalent circuit without
taking into account the effects of vDS or the substrate, Fig. 4
(b) illustrates the equivalent circuit taking into account the
effect of vDS through augmenting the resistor, ro, and Fig. 4
1
// r0
g m g mb
Jg m , r0
, g mb
1
g m K J 1
?R
(a)
IV.
Av
(b)
Fig. 5 (a) The diode-connected subthreshold transistor and (b) the corresponding small-signal equivalent circuit.
Since the voltage, vgs is applied on the dependentcurrent source, gm vgs, we can apply the source-absorption
theorem [18] and thus replacing this source with a resistance
equal to 1/gm. The voltage, vbs is zero, so the dependent
source, gmbvbs will be replaced with an open circuit. The
resistance seen at the drain end is thus
1
1
//
Kg m g m
1
.
g m1 ro1 //
gm 2 K J 1
1
K 1g m
(17)
The resistance given by Eq. 17 represents the equivalent resistance of the diode-connected subthreshold transistor. The resistance seen at the source end is [19]
(18)
It can be concluded that the voltage gain of the subthreshold transistor amplifier is less that that using the superthreshold transistor due to the fact that gm of the subthreshold transistor is less than that of the superthreshold one.
g m r0
1
r0 //
gm
1
Kg m
gm
Kg m
1
(19)
This is to be compared with VA/VT for the superthreshold transistor [18]. The common-gate, the sourcefollower, and the cascade stages can be studied in a similar
manner.
V.
CONCLUSIONS
Due to the increasing demand of portable applications,
the need arises to reduce the power consumption as possible
as we can. In applications where the low or medium performance can be tolerated, the subthreshold region can be utilized. In long idle-time applications, the subthreshold leakage current is considered the main cause of static power
consumption. However, in ultra low-power applications, the
subthreshold leakage current is the operating current. In this
paper, the use of the subthreshold transistor as an amplifier
was investigated and the ac small-signal equivalent circuit
of the subthreshold transistor was developed. Throughout
the paper, the differences between the subthreshold and superthreshold transistors were illustrated.
VI.
FUTURE WORK
In this Section, we will propose two points for the future work to be completed by the reader. First, the currentvoltage relationship of Eq. 5 of the subthreshold transistor
reveals the similarity between the subthreshold transistor
and the BJT transistor in that the current in both devices
depends exponentially on the voltage difference between
two terminals. This suggests that the subthreshold transistor
can be used to implement the functions performed by the
BJT transistor but in ultra low-power applications. Among
these functions is the logarithmic amplifier illustrated in Fig.
8 [20].
The reader is encouraged to investigate the implementation of the logarithmic function using the subthreshold
transistor and compare the performance, power consumption, and area overhead with that using the BJT transistor.
Besides the ultra low-power consumption of the subthreshold transistor, the ease of fabrication of the MOSFET
transistor is also a motivation for utilizing the MOSFET
transistor in these types of applications. The exponential
dependence of the current through the subthreshold transistor on the gate-to-source voltage seems to open the door for
a new type of applications that can be performed using only
the BJT transistor. Among these applications is the logarithmic amplifier.
The second point is the effect of the gate-tunneling current, IG of the subthreshold transistor. Throughout this paper, we have neglected the gate-tunneling current, IG of the
subthreshold transistor. IG increases exponentially as the
gate-oxide thickness, tox decreases due to the increased
probability of tunneling. Keeping in mind that tox decreases
with CMOS technology scaling in order to alleviate short-
REFERENCES