Micro and Nanosystems
Aloke Saha
1,*
, Rahul Pal
2
and Jayanta Ghosh
3
1
Department of Electrical and Computer Engineering, Dr. B. C. Roy Engineering College, Durgapur, India;
2
Department of Electrical and Computer Engineering, Bengal College of Engineering, Durgapur, India;
3
Department of
Electrical and Computer Engineering, NIT Patna, Patna, India
Abstract: Background: The present study explores a novel self-pipelining strategy that can enhance
speed-power efficiency as well as the reliability of a binary multiplier as compared to state-of-art reg-
ister and wavepipelining.
Method: Proper synchronization with efficient clocking between the subsequent self-pipelining stages
has been assured to design a self-pipelined multiplier. Each self-pipelining stage consists of self-latching
leaf cells that are designed, optimized and evaluated by TSMC 0.18μm CMOS technology with 1.8V
supply rail and at 25°C temperature. The T-Spice transient response and simulated results for the de-
signed circuits are presented. The proposed idea has been applied to design 4-b×4-b self-pipelined Wal-
lace-tree multiplier. The multiplier was validated for all possible test patterns and the transient response
was evaluated. The circuit performance in terms of propagation delay, average power and Power-Delay-
Product (PDP) is recorded. Next, the decomposition logic is applied to design a higher-order multiplier
(i.e., 8-bit×8-bit and 16-bit×16-bit) based on the proposed strategy using 4-bit×4-bit self-pipelined multi-
plier. The designed multiplier was also validated through extensive TSpice simulation for all the required
test patterns using W-Edit and the evaluated performance is presented. All the designs, optimizations and
evaluations performed are based on BSIM3 device parameter of TSMC 0.18μm CMOS technology with
1.8V supply rail at 25°C temperature using S-Edit of Tanner EDA.
Results: The reliability was investigated of the proposed 4-b×4-b multiplier in the temperature range -
40°C to 100°C for maximum PDP variation.
Conclusion: A benchmarking analysis in terms of speed-power performance with recent competitive
design reveals preeminence of the proposed technique.
A R T I C L E H I S T O R Y
Received: May 05, 2019
Revised: July 27, 2019
Accepted: August 06, 2019
DOI:
10.2174/1876402911666190916155445
Keywords: Decomposition logic, Power-Delay-Product (PDP), reliability, self-latching, self-pipelining, parallel multiplier.
1. INTRODUCTION
Speed-power efficiency, as well as reliability, is an im-
portant design constraint for modern digital VLSI designer
[1-3]. A multiplier acts as a major bottleneck for most digital
computing systems in terms of operating-speed, power-
dissipation as well as reliability [4-21]. High throughput
characteristic of parallel multiplier has earlier attracted re-
searchers for a long time [22, 23]. After the introduction of a
parallel multiplier by C. S. Wallace in 1964 [22] and
subsequently by L. Dada in 1965 [23], a number of structural
modifications have been investigated to achieve high-speed
multiplication along with low-power dissipation. Rapid
technological advancements in modern nano-scale MOS
devices and an increase in function/device density in a single
monolithic IC make the reliability also a major design issue.
Increase in function density with complex device
interconnection leads to generation of localised heat (hot-
spot) and is responsible to increase the rate of IC failure [24].
*Address correspondence to this author at the Department of Electrical and
Computer Engineering, Dr. B. C. Roy Engineering College, Durgapur,
India; Tel: +91 8967627938; E-mail: saha81@gmail.com
Conventional Register pipelining [18] is a well known
and globally accepted design strategy at the structural level
that can improve throughput performance of a parallel multi-
plier. However in this approach, a large number of interme-
diate latches are used to achieve the pipelining process. As a
result, the throughput efficiency of a pipelined microelec-
tronic circuit increases at the cost of more power, area and
latency as compared to the non-pipelined system. Wave-
pipelining was investigated by Burleson et al. in 1998 [25]
as a feasible alternative to overcome problems associated
with conventional register pipelining. Wave-pipelining [12,
13, 25] improves the overall efficiency by removing inter-
mediate latches/flip-flops from the circuit. The maximum
throughput of a wave-pipelined system depends not on the
critical path delay but on the delay difference between the
worst and the best path delay [25]. Hence, the delay-
equalization among all the data-paths from input to output is
a critical design challenge for effective wave-pipelining.
However the wave-pipelined system suffers from reliability
issue and its efficiency depends mostly on the designer’s
skills. For more detail, readers are directed to references [12,
13, 25].
1876-4037/20 $65.00+.00 ©2020 Bentham Science Publishers
Send Orders for Reprints to reprints@benthamscience.net
Micro and Nanosystems, 2020, 12, 149-158
149
RESEARCH ARTICLE
Novel Self-Pipelining Approach for Speed-Power Efficient Reliable Binary
Multiplication