Professional Documents
Culture Documents
fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMSCS.2018.2827388, IEEE
Transactions on Multi-Scale Computing Systems
Abstract— The low-power multi-sensor system with power management and nonvolatile memory access control for IoT
applications are proposed, which achieves almost zero standby power at the no-operation modes. A power management scheme
with activity localization can reduce the number of transitions between power-on and power-off modes with rescheduling and
bundling task procedures. In addition, autonomously standby mode transition control selects the optimum standby mode of
microcontrollers, reducing total power consumption. We demonstrate with evaluation board as a use case of IoT applications,
observing 91% power reductions by adopting task scheduling and autonomously standby mode transition control combination.
Furthermore, we propose a new nonvolatile memory access control technology, and estimate the possibility for future low-power
effect.
—————————— ——————————
1 INTRODUCTION
2332-7766 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMSCS.2018.2827388, IEEE
Transactions on Multi-Scale Computing Systems
puting system is to maximize the no operation period Vcc Vcc Vcc Vcc Sensor
Input
and/or to minimize the BET by following two approaches. n (b) (C)
Vdd Peripheral
network
Sensor Controller
Sensor 0 Power
IF
(MCUlow)
2332-7766 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMSCS.2018.2827388, IEEE
Transactions on Multi-Scale Computing Systems
that, the processing data is sent out to sensor networks (a) Example task
through the sensor-net interface. In the conventional sys- Task A
Input Interval
1 3 5 1 3 5
Required MCU performance:
Task A Task-A: MCUlow,
tem, the standby power is consumed in both active mode Input interval
Deadline
Deadline Task-B:
T MCUhigh
Input Interval
and standby (waiting) mode. Meanwhile, a system dia- Task BB
Task 2 4 2 4
ments are summarized below. Minimal execution of MCUhigh so as not to cause a deadline violation
T
2332-7766 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMSCS.2018.2827388, IEEE
Transactions on Multi-Scale Computing Systems
Total Energy
Nonvolatile memory access controller NVM
Consumption Energy Consumption
while active state Memory Block
Energy Consumption Control register
Energy Overhead
while standby state Block 1
when active state Power control Block 2
returning from
Memory block control
2332-7766 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMSCS.2018.2827388, IEEE
Transactions on Multi-Scale Computing Systems
Fig. 6 shows the concept of proposed NVM architecture. access instruction continues for the specified number of
Fig. 6 (a) is the conventional NVM architecture, which has times or more. Set the power supply to the standby mode.
memory access all the time during CPU operation and the
whole memory is powered on. Fig. 6 (b) shows NVRAM (d) BET Judgement
divided into the blocks and independently control the The breakeven time (BET) is defined as the time during
power supply, and only the blocks necessary for access are which the energy consumption becomes equal when the
turned on. Unnecessary blocks are turned off immediately. power is controlled and when it is not controlled. BET is
By reading the access address exchanged between the CPU determined by a preset power parameters which are the
and the NVM and the necessary block by reading the read ratio of the power-on overhead energy and standby
code, it is possible to shut down the power while avoiding (idling) energy. Only when the idle time is longer than BET,
the access delay. the power supply of the memory block is turned off.
Fig. 7 shows the block diagram of newly proposed non-
volatile memory (NVM) access controller. NVM controller (e) Power setting (Memory block control)
has below function for memory power control. Provide registers for controlling the power supply of the
specified NVM to always be on and displaying the power
(a) Control register setting status of the NVM. Whether or not to turn ON the
A hardware setting register for setting the type and capac- power supply of the NVM of each bank is determined by
ity of the NVM, a memory use area setting register for set- each setting register and the access state. There are the fol-
ting whether or not the mounted memory is used for each lowing two factors for setting the power supply state.
task, and various registers for setting the operation of the 1) ON setting by instruction address determination
power control function in detail. 2) ON setting by instruction decode judgment
2332-7766 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMSCS.2018.2827388, IEEE
Transactions on Multi-Scale Computing Systems
500
Intruder detection Distance calculation
(Period=30ms, fixed) 400 Reduction
from IR sensor data
of 91%
300
Set IR sampling period 200
and Intruder detection period Reduction
100 of ~1%
Intruder detection 0
(Period=50/100/500ms) Conventional Proposed-1 Proposed-2
(Estimated)
Proposed-1: Task scheduling + Autonomous standby mode transition technology
Proposed-2: Porposed-1 + Nonvolatile memory access control technology
(a) Conventional (b) Proposed
Fig. 13. Evaluation results.
Fig. 12. Task flow of intruder detection system
2332-7766 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMSCS.2018.2827388, IEEE
Transactions on Multi-Scale Computing Systems
2332-7766 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMSCS.2018.2827388, IEEE
Transactions on Multi-Scale Computing Systems
[3] L. Zhao et al., “Gayser-2: The second prototype CPU with fine- Proc. 13th International New Circuits and Systems Conference
grained run-time power gating,” 16th Asia and South Pacific De- (NEWCAS), pp.1-4, 2015.
sign Automation Conference (ASP-DAC), pp. 87–88 (2011).
[4] K. Usami et al., “Design and Control Methodology for Fine Grain
Power Gating based on Energy Characterization and Core Pro-
filing of Microprocessors,” 19th Asia and South Pacific Design
Automation Conference (ASP-DAC), 2014. Masanori Hayashikoshi was born in Ehime,
[5] L. Benini et al., “A survey of design techniques for system-level Japan, on December 7, 1961. He received the
dynamic power management,” IEEE Transactions on Very Large B.S. and M.S. degrees in electronic engineer-
ing from Kobe University, Kobe, Japan, in
Scale Integration (VLSI) Systems, Vol. 8, No. 3, pp. 299–316
1984 and 1986, respectively.
(2000). In 1986 he joined the LSI Research and
[6] M. Weiser et al., “Scheduling for Reduces CPU Energy,” Pro- Development Laboratory, Mitsubishi Electric
ceeding of the 1st USENIX Conference on Operating Systems De- Corporation, Hyogo, Japan. Since then he has
been engaged in the research and develop-
sign and Implementation, OSDI ’94, Berkeley, CA, USA, USENIX
ment of nonvolatile semiconductor memories
Association (1994). and high density DRAM’s. Now, he is a Chief Professional of Core
[7] F. Yao et al., “A scheduling model for reduced CPU energy, Technology Business Division in Renesas Electronics Corporation.
Foundations of Computer Science,” 1995. Proceedings., 36th An- He has been engaged in the research and development of Flash
memory, high density DRAM, low power SDRAM, embedded MRAM
nual Symposium on, pp. 374–382 (1995).
for MCUs, and now engaged in the Normally-Off computing architec-
[8] W. Huang et al., “An optimal speed control scheme supported ture as for further low-power solution with NVRAMs.
by media servers for low power multimedia applications,” Mul-
timedia Systems, Vol. 15 No. 2, pp. 113–124 (2009). Hideyuki Noda received B.E. and M.E. de-
grees in electronics engineering from Waseda
[9] M. E. T. Gerards et al., “Optimal DPM and DVFS for Frame-
University in 1995 and 1997, respectively, and
based Real-time Systems,” ACM Trans. Archit. Code Optim., Vol. Ph.D. degree in electronics engineering from
9, No. 4, pp. 41:1–41:23 (2013). Hiroshima University in 2007.
[10] Takashi Nakada et al., “Design Aid of Multi-core Embedded Sys- In 1997, he joined the ULSI Laboratory,
Mitsubishi Electric Corporation, Hyogo, Japan,
tems with Energy Model,” IPSJ Transactions on Advanced Com-
where he has been engaged in the design and
puting Systems, vol.7, No.3, Aug., 2014. development of embedded memory for sys-
[11] T. Nakada, et al.: “Design Aid of Multi-core Embedded Systems with En- tem LSI solutions. In 2003, he transferred to
ergy Model,” IPSJ Transactions on Advanced Computing Systems, Vol.7, Renesas Technology Corp. (currently,
Renesas Electronics Corporation). Since then
No.3, pp. 37–46, (2014)
he has been engaged in the architecture design and development of
[12] T. Nakada, et al., “Energy-Efficient Continuous Task Scheduling for application-oriented IPs such as image processing IPs, and motor
Near Real-Time Periodic Tasks,” IEEE International Conference on Data control IPs for SoCs and MCUs designed for consumer and automo-
Science and Data Intensive Systems, 2015, pp. 675 – 681. tive application. He is currently a Section Manager of CPU system de-
partment, Core Technology Business Division in the 1st Solution Busi-
[13] M. Hayashikoshi, et al., “Low-Power Multi-Sensor System with
ness Unit, Renesas Electronics Corporation, Japan. Dr. Noda is also
Task Scheduling and Autonomous Standby Mode Transition a Visiting Associate Professor of Graduate School of Natural Science
Control for IoT Applications,” 2017 IEEE Symposium in Low- and Technology, Kanazawa University, Japan.
Power and High-Speed Chips (COOL CHIPS), pp.1-3, Apr. 2017.
Hiroyuki Kawai received the B.S. and M.S.
[14] Hiroshi Nakamura, “Normally-Off Computing Project: Chal-
degrees in control engineering from Osaka
lenges and Opportunities,” in Proc. 19th Asia and South Pacific University, Osaka, Japan, in 1984 and 1986,
Design Automation Conference (ASP-DAC), pp. 1-5, Jan. 2014. respectively, and received the Ph.D. degree in
[15] Masanori Hayashikoshi et al., “Normally-Off MCU Architecture electronics, information and communication
engineering from Waseda University, Tokyo,
for Low-power Sensor Node,” in Proc. 19th Asia and South Pa-
Japan, in 2005. In 1986, he joined the LSI La-
cific Design Automation Conference (ASP-DAC), pp. 12-16, Jan. boratory, Mitsubishi Electric Corporation, Hy-
2014. ogo, Japan where he worked on the research
[16] M. Hayashikoshi et al., " Normally-Off MCU Architecture and and development of digital image signal pro-
cessors including graphics and image recog-
Power Management Method for Low-Power Sensor Network,"
nition. He was transferred to Renesas Technology Corporation in
in Proc. 12th International SoC Design Conference (ISOCC), 2003 and Renesas Electronics Corporation in 2010. He was engaged
pp.151-152, Nov. 2015 in the development of reconfigurable circuits and communication IP’s
[17] M. Hayashikoshi, et al., “Low-Power Multi-Sensor System with in the Logic IP development Department of Renesas Electronics Cor-
poration, Hyogo, Japan. Since 2015, he has been a professor of De-
Normally-off Sensing Technology for IoT Applications,” in Proc.
partment of Electronic Engineering and Information Science, Faculty
13th International SoC Design Conference (ISOCC), pp.195-196, of Science and Engineering, Tokushima Bunri University, Sanuki, Ka-
Oct. 2016. gawa, Japan. He currently works on the research of sensor network
[18] Renesas RX63N User’s manual: Hardware, pp. 293-295, for IOT applications and computational intelligence for attentive com-
puter-based systems. He is a member of the IEEE, IEICE and SICE.
http://www.renesas.com/en-us/doc/prod-
ucts/mpumcu/doc/rx_family/r01uh0041ej0180_rx63n631.pdf
[19] T. Kawahara, et al., “Scalable Spin-Transfer Torque RAM Tech-
nology for Normally-Off Computing,” IEEE Design & Test of
Computers, Vol. 28, No. 1, pp.52-63, 2011.
[20] C. Layer, et al., “Low-Power Hybrid STT/CMOS System-on-
Chip Embedding Non-Volatile Magnetic Memory Blocks,” in
2332-7766 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TMSCS.2018.2827388, IEEE
Transactions on Multi-Scale Computing Systems
2332-7766 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.