Variability Aware Performance Evaluation of Low Power SRAM Cell
Hansel Dsilva, Julian Pinto, Arzan Elchidana and Sudhakar Mande
Department of Electronics and Telecommunication,
Don Bosco Institute of Technology, Mumbai University, India.
email: dsilvahansel@ieee.org
Abstract— Till today CMOS scaling is considered as
the best option to achieve higher density, high performance
and low power integrated circuits. However, scaling of con-
ventional planar MOSFET in the sub-45nm regime leads
to many undesirable short channel effects. FinFET is con-
sidered as the suitable candidate for the replacement of
conventional planar MOSFETs. In this work, suitability of
FinFETs for replacement of planar bulk technology in sub-
20nm regime has been verified using Predictive Technology
Models. For this purpose, the performance of the FinFET
based SRAM cell is compared with conventional planar
Bulk based SRAM cell. Moreover, robustness of FinFET
based SRAM cell against process, temperature and power
supply variations is evaluated and compared with conven-
tional planar based SRAM cell. Our simulation results
confirms the suitability of FinFETs for the replacement of
conventional planar CMOS technology.
Keywords—Predictive Technology Models,independent-
gate FinFET based 6T SRAM, planar Bulk based 6T
SRAM, PVT analysis, Static Noise Margin, Access Time.
I. Introduction
T
HE impact of process variation on performance and
yield of integrated circuits is emerging as one of the
major problems for the continued CMOS scaling in the sub-
22nm regime. In order, to combat the variability in such
aggressively scaled CMOS technology, research focuses on
new device architectures.
FinFET is considered as the most promising candidate
for replacement of the conventional bulk MOSFET [1–3].
It is expected that ability of FinFET to be able to control
the channel by its three sided gate configuration results
in improvement in short channel performance and reduced
variability. As compared to the planar Bulk device the
FinFET device exhibits 100X greater I
on
/I
off
-ratio. In lit-
erature, [4, 5] it is seen that the FinFET process flow is
similar to that of a conventional Bulk device.
For the emerging SRAM applications the FinFET plays
a crucial role as it promises high noise margins and im-
proved timings. In this work, FinFET application is fo-
cused around SRAM cell as it is estimated to occupy nearly
90% of the chip area in the near future. But there exists
area limitation in VLSI design. Hence, SRAM design needs
to involve minimum sized devices which will provide at the
same timings and equal stability as their large scale device
based counterparts.
As we approach the 22nm node for design of SRAM, we
need to take into account device fluctuations induced due
to random variations. In literature [4, 6], which focus on
impact of process variations on the characteristics of the
FinFET device. In this paper, impact of process, volt-
age and temperature variations on SRAM performance is
studied at the 16nm technology regime. The main reason
behind including temperature variations is to see the im-
pact of temperature inversion on the FinFET based SRAM.
Also, the impact of supply voltage is studied to get a better
idea of noise margin dependency on supply voltage. This
will enable better definitions of design corners for sub-22nm
FinFET technology.
The models used for simulation were those developed by
the PTM-group at Arizona State University. These mod-
els are predictive and are designed in accordance to the
ITRS- roadmap and are based on the BSIM-MG model.
The PTM- group choose to target an I
off
= 0.1 nA/um for
the PTM- CMG as given in [7,8]. To enable a compara-
tive study between planar Bulk and FinFET technology,
we have tuned the PTM-Bulk high-k 16nm model to I
off
=
0.1 nA/um. The rest of the paper is organized as follows.
In the first section, we have characterized the FinFET
based 6T SRAM cell at the 20nm, 16nm, 14nm, 10nm and
7nm technology nodes. The SRAM cell hold, read and
write static noise margin followed by access timing was
tabulated and a study was conducted to make predictions
on the trend of the SRAM characteristics for further scaling
beyond the 22nm technology node.
In the second section, a comparative study between the
PTM-Bulk high-k and PTM-CMG was conducted at the
16nm technology node. A study on the impact of beta-ratio
on SRAM characteristics followed by a comparison with
that of the PTM- Bulk high-k 16nm. This gives a more
clear idea into the stability of the FinFET based SRAM
cell and its advantages and disadvantages over the planar
Bulk based SRAM.
In the third section, variability analysis is explored by
inducing process, voltage and temperature variation in the
PTM-CMG and PTM-Bulk High K at 16nm technology
node. This is summarized as seen in Fig.3 .This study will
help in better understanding the yield of the FinFET based
SRAM during fabrication. To the best knowledge of the
authors, this is the first paper to include PVT-analysis of
an FinFET based 6T-SRAM cell and compare the results
to its counterpart planar Bulk based SRAM cell at 16nm
technology node.
978-1-4799-1314-5/13/$31.00 ©2013 IEEE 183 5th Asia Symposium on Quality Electronic Design