Liao K, et al. Sci China Inf Sci April 2016 Vol. 59 042404:2
advantages of fast speed and ultralow power consumption have been a popular research topic, and numer-
ous different types of flip-flops have been invented and investigated over the past decades [5–9]. Recently,
a static contention-free single-phase-clocked (SPC) flip-flop (S
2
CFF) with low-power consumption has
been presented in [10] by Kim et al., which to our knowledge is one of the best edge-trigged flip-flops
implemented by traditional planar MOSFET technology in the literature.
It is well known that the total power dissipation can be divided into three major parts: dynamic power
dissipation, static power dissipation, and short-circuit power dissipation. However, with the technology
scaling down, leakage power consumption has become an increasingly important part of the total power
consumption. Since the states of flip-flops only switch at the rising edges of the clock signal and remain
the same at other times in the clock period, this inherent trend could be an unpleasant restriction
for advanced flip-flops to further reduce the power consumption. To solve this problem, several novel
multigate devices aiming at implementing ultralow leakage current have been proposed, such as ultrathin
body devices [11], fully depleted silicon on insulator (FDSOI) [12], and fin-type field-effect transistors
(FinFETs) [13]. So far, FinFET seems to be the most promising option because of its superior electrical
properties and timing performance, and commercial chips based on FinFETs have already been released
by Intel, TSMC, and other global foundries.
In this article, we chose the original S
2
CFF based on planar MOSFET as our research object and
reconstructed it with FinFETs to achieve high speed and ultralow power consumption. At the same
time, we discovered an intrinsic defect (unnecessary energy loss) of S
2
CFF by analyzing the simulation
results and proposed an improvement approach by utilizing the excellent design flexibility of multimode
FinFET. The verification results show that the multimode FinFET flip-flop offers a good solution for the
discovered defect, and the additional penalty on timing performance is acceptable.
2 FinFET devices
With the scaling down of transistor conductive channels, short-channel effects (SCEs) become intoler-
able, and many multigate devices have been proposed to overcome SCEs. Due to its relatively simple
manufacturing process and good compatibility with bulk CMOS, FinFET is considered to be the most
feasible new multigate device. Through the stronger control of the conductive channel by the double
gate (front gate and back gate), the FinFET device has the advantages of higher on-state current, lower
off-state current, and faster switching speed [14, 15].
Furthermore, according to whether the front gate and back gate are tied together or not, FinFET
circuits can be divided into three different operating modes, namely, shorted-gate (SG-mode), low-power
(LP-mode), and independent-gate (IG-mode) [16]. Different circuit operating modes have different char-
acteristics, which increases the design flexibility. For example, the SG-mode FinFET has high on-state
current and fast switching speed to achieve high performance, while the LP-mode FinFET has low off-
state current to reduce the leakage power dissipation. Figure 1(a) and (b) illustrate the three-dimensional
diagram and cross-sectional top view of a FinFET transistor, and Figure 1(c) shows the electrical model
schematic of the three operating modes of N-FinFETs and P-FinFETs, respectively.
In our study, the simulations are based on the predictive technology model (PTM) for 32-nm Fin-
FETs [17]. PTM is a theoretical mode that ignores the actual process parameters. It covers sufficient
physical effects, and excellent scalability of PTM across process and design conditions has been shown in
the published results. Considering there is no FinFET Spice model officially offered by foundries, PTM
was selected for our study without loss of generality. The primary parameters of the devices are listed in
Table 1, which are typical for manufactured 32-nm FinFETs. The parameter L
gate
denotes the length of
the gate, while H
fin
and W
fin
denote the height and width of the silicon fin, respectively. For a common
SG-mode FinFET, the equivalent gate width W
gate
can be calculated by
W
gate
=2× H
fin
+ W
fin
. (1)
Note that, W
fin
is usually much smaller than H
fin
, and for LP- and IG-mode FinFETs, W
fin
is eliminated
because the top silicon of the gate is polished to separate the double gates. In addition, the thickness of