Beruflich Dokumente
Kultur Dokumente
1
DFT2017
Agenda:
Introduction
Background
Motivations
Modelling
Proposed Architecture
Results
Conclusion and future work
References
2
Introduction DFT2017
Power Density Trend
CMOS challenges: 1E+03
Active Power Density
Normalized Energy/cycle
20
15
Functionality challenge 10
Etotal
This attributed to their zero-leakage energy compared to that of CMOS power transistors as
shown in Fig. 1.
Fig. 2 shows the FEA simulation of the adopted MEMS
To model and capture the physical behaviour of MEMS accurately, COMSOL multiphysics
tool has been used in our work. Parameteric sweep analysis Has been performed to calculate
the following points:
Structure stiffness
Damping analysis
Energy-Latency Analysis of N/MEMS
5
Modelling DFT2017
Fig. 5: Illustrates that: (a) switching energy based FEA at g0=200nm and A=450um2 as a function of gd and resonant frequency
(w); (b) Tmech: as a function of gap ratio and resonant frequency obtained from 3D FEA at A=450um2; (c) Tmech: of the scaled
relay at g0=40nm and A=45um2 as a function of gd and resonant frequency; (d) switching energy of the scaled relay.
6
Modelling DFT2017
Device
o dimensions COMSOL
o geometry multiphysics
o Material 3D FEA
o etc. simulator
VHDL code
-Parametric sweep o ARM-M0
Contact -Transient analysis microprocessor
modelling -Frequency analysis o FIR filter
o FB (DSP, RF, ….)
o Stiffness (k)
Extract Device o Damping (b)
parameters o Mass(m) Synopsis-
o Pull-in Voltage (Vpi) simulator
o Mechanical time synthesis
o Switching energy
o etc.
Verilog-AMS
Optimize
(Electrical and o Dynamic and
mechanical static power
lumped model) [1] o Area
Results
o Total energy Cadence
o Timing simulator
o etc.
7
Switch Simulator Based on (FEA) DFT2017
VS0
Fele 0
Source
Lumped mechanical parameters, 3D (FEA) based
Fvdw 0
𝑆𝑘 Z0
Scaling
factors
Vg1 VS1
θ𝑏
Φ𝑚 Fele 1
Body
fabricated/predicted
Fvdw 1
Verilog-AMS
on COMSOL
Verilog-AMS
Z1
Based on
Vg2
Gate
MEMS/ NEM
VSN
Fele N
Fvdw N
ZN
Vg2
Mechanical contact
Drain
Fig. 7: Graphic illustration of the hierarchical model of the switch simulator. The highlighted regions represent the electrical and
mechanical lumped parameters, which is highlighted regions represent the electrical and mechanical lumped parameters, which is
8
written in Verilog-AMS and co-simulated in The Spectre spice tool.
Reliability Characterizations DFT2017
Time(us) Time(us)
τ𝑚𝑒𝑐ℎ.τ𝑠𝑡𝑎𝑏𝑙𝑒τ𝑒𝑣𝑎𝑙𝑢𝑎𝑡𝑒 τ𝑚𝑒𝑐ℎ. τ𝑠𝑡𝑎𝑏𝑙𝑒
(a) (b)
Fig. 7: Resonant frequency of the 2-spring N/MEMS
Fig. 6: Contact bouncing impact when using ramped Vdd (a) safe bouncing;
(b) unstable bouncing cause operation failure of power-gated circuits.
9
Proposed Reliable Approach DFT2017
Combinational
Sleep=0 circuits
Clock
(a) (c)
Fig. 8: Illustrates that (a) proposed architecture; (b) 3D schematic of N/MEMS 10
Results: DFT2017
Table III: Energy consumption for 32-tap FIR filter at various asynchronous PG setups .
In our experiment a complex combinational circuit, such as 32-tap FIR filter, is used due to the fact that, unlike CMOS
power switch N/MEMS favour a complex design architecture coupled with low duty cycle
implementations
It should be noted that our proposed approach only differs from [3] by using a ramp supply voltage with
various T.
These results indicate that at low data rate our approach can achieve greater energy savings by about 9%
compared to the previously reported results
This can, firstly, be attributed to completely cutting off the leakage dissipation during the idle state. Secondly, the dynamic
energy of the FIR filter is significantly minimized due to mitigation of the contact bouncing and current rushing by slow
charging of the load capacitance.
However, increasing the data rate will lead to increase switching energy of the MEMS-based power gating which outweigh
its leakage power savings, as illustrated in Table III.
It should be noted that in this experiment the damping coefficient was chosen to be equal to 50 ×10−8 N.s/m. Reducing
the damping coefficient to 50 ×10−8 N.s/m has led the simulator to diverge, thereby no useful results are obtained due
to the large bouncing time (floating). 11
Results: DFT2017
Vdd Vdd
32 64 32 32 32 32
Multiplier Accumulator
D Flip-flop D Flip-flop D Flip-flop
Data_in Data_out
12
Conclusion: DFT2017
Conclusion:
This study demonstrated the limitations at which N/MEMS can be adopted without
impact any potential failure of the power-gated circuits.
Furthermore, this study showed the threshold at which N/MEMS can achieved
greater energy saving compared to that of CMOS transistor.
13
DFT2017
References
[1] H. Alrudainy, A. Mokhov, and A. Yakovlev, “A scalable physical model for nano-electro-mechanical
relays,” in Power and Timing Modeling, Optimization and Simulation (PATMOS), 2014 24th International
Workshop on, pp. 1–7, Sept 2014
[2] D. Lee, W. Lee, C. Chen, F. Fallah, J. Provine, S. Chong, J. Watkins, six-terminal nem relays,” Computer
Computer-Aided Design of Integrated Circuits and Systems, IEEE Transactions on, vol. 32, pp. 653–666,
May 2013.
[3] M. Spencer, F. Chen, et al., "Demonstration of integrated micro-electro-mechanical relay circuits for
Source
VLSI applications," IEEE Journal of Solid-State Circuits, vol. 46, pp. 308-320, // 2011.
[4] M. Henry and L. Nazhandali, “From transistors to mems: Throughputaware power gating in cmos
circuits,” in Design, Automation Test in rope Conference Exhibition (DATE), 2010, pp. 130–135, March
2010.
[5] B. Amelifard and M. Pedram, “Design of an efficient power delivery network in an soc to enable
dynamic power management,” in Low Power Electronics and Design (ISLPED), 2007 ACM/IEEE
International Symposium on, pp. 328–333, Aug 2007.
14