Abstract
An artificial intelligence of things (AIoT) chip is a critical hardware component in edge devices that supports data acquisition and processing in the artificial intelligence (AI) era. In this paper, we introduce the architecture and circuit techniques for AIoT systems as well as the design challenges associated with them. To address the design challenges, we describe our design approaches to improve energy efficiency from the standpoints of system architecture, AI computing, and data acquisition circuits. Meanwhile, we present our most recent research progress and state-of-the-art AIoT chip demonstrations at a prestigious conference. Our design techniques improve power and energy efficiency by orders of magnitude on corresponding state of the arts (SOTA), allowing for an advanced AIoT chip design.
Similar content being viewed by others
References
Ye L, Wang Z, Liu Y, et al. The challenges and emerging technologies for low-power artificial intelligence IoT systems. IEEE Trans Circ Syst I, 2021, 68: 4821–4834
Bell C G, Chen R, Rege S. Effect of technology on near term computer structures. Computer, 1972, 5: 29–38
Warden P, Situnayake D. TinyML: Machine Learning with TensorFlow Lite on Arduino and Ultra-Low-Power Microcontrollers. Sevastopol: O’Reilly Media, 2019
Lin J, Zhu L G, Chen W-M, et al. On-device training under 256KB memory. 2022. ArXiv:2206.15472
Wang X, Magno M, Cavigelli L, et al. FANN-on-MCU: an open-source toolkit for energy-efficient neural network inference at the edge of the Internet of Things. IEEE Int Things J, 2020, 7: 4403–4417
Liu Y, Wang Z X, He W, et al. An 82 nW 0.53 pJ/SOP clock-free spiking neural network with 40 µ,s latency for AloT wake-up functions using ultimate-event-driven bionic architecture and computing-in-memory technique. In: Proceedings of IEEE International Solid-State Circuit Conference (ISSCC), San Francisco, 2022. 372–374
Li H, Tan Z, Bao Y, et al. Energy-efficient CMOS humidity sensors using adaptive range-shift zoom CDC and power-aware floating inverter amplifier array. IEEE J Solid-State Circ, 2021, 56: 3560–3572
Jing Y, Wang Z, Shen L, et al. An infomation-aware adaptive data acquisition system using level-crossing ADC with signal-dependent full scale and adaptive resolution for IoT applications. In: Proceedings of IEEE International Symposium on Circuits and Systems (ISCAS), 2023
Alioto M. Enabling the Internet of Things: From Integrated Circuits to Integrated Systems. Berlin: Springer, 2017
Wang Z X, Ye L, Zhang H, et al. 20.2 A 57 nW software-defined always-on wake-up chip for IoT devices with asynchronous pipelined event-driven architecture and time-shielding level-crossing ADC. In: Proceedings of IEEE International Solid-State Circuit Conference (ISSCC), San Francisco, 2020. 314–315
Wang Z X, Ye L, Liu Y, et al. 12.1 A 148 nW general-purpose event-driven intelligent wake-up chip for AIoT devices using asynchronous spike-based feature extractor and convolutional neural network. In: Proceedings of IEEE International Solid-State Circuit Conference (ISSCC), San Francisco, 2021. 436–438
Raychowdhury A, Tokunaga C, Beltman W, et al. A 2.3 nJ/frame voice activity detector-based audio front-end for context-aware system-on-chip applications in 32-nm CMOS. IEEE J Solid-State Circ, 2013, 48: 1963–1969
Price M, Glass J, Chandrakasan A P. A low-power speech recognizer and voice activity detector using deep neural networks. IEEE J Solid-State Circ, 2018, 53: 66–75
Badami K, Lauwereins S, Meert W, et al. Context-aware hierarchical information-sensing in a 6 µW 90 nm CMOS voice activity detector. In: Proceedings of IEEE International Solid-State Circuits Conference, 2015
Yang M H, Yeh C H, Zhou Y Y, et al. A 1 µW voice activity detector using analog feature extraction and digital deep neural network. In: Proceedings of IEEE International Solid-State Circuits Conference, 2018. 346–348
Cho M C, Oh S C, Shi Z, et al. A 142 nW voice and acoustic activity detection chip for mm-scale sensor nodes using time-interleaved mixer-based frequency scanning. In: Proceedings of IEEE International Solid-State Circuits Conference, 2019. 278–280
Chen Y, Cho M C, Jeong S, et al. A dual-stage, ultra-low-power acoustic event detection system. In: Proceedings of IEEE International Workshop on Signal Processing Systems (SiPS), 2016. 213–218
Jeong S, Chen Y, Jang T, et al. Always-on 12-nW acoustic sensing and object recognition microsystem for unattended ground sensor nodes. IEEE J Solid-State Circ, 2018, 53: 261–274
Chen F F, Un K F, Yu W H, et al. A 108nW 0.8mm2 analog voice activity detector (VAD) featuring a time-domain CNN as a programmable feature extractor and a sparsity-aware computational scheme in 28nm CMOS. In: Proceedings of IEEE International Solid-State Circuit Conference (ISSCC), San Francisco, 2022. 368–369
Wang A, Chen C, Shi C R. Design and analysis of an always-ON input-biased pA-current sub-nW mV-threshold hysteretic comparator for near-zero energy sensing. IEEE Trans Circ Syst I, 2017, 64: 2284–2294
Bong K, Choi S, Kim C, et al. A low-power convolutional neural network face recognition processor and a CIS integrated with always-on face detector. IEEE J Solid-State Circ, 2018, 53: 115–123
Wannamaker R A, Lipshitz S P, Vanderkooy J, et al. A theory of nonsubtractive dither. IEEE Trans Signal Process, 2000, 48: 499–516
Fredenburg J A, Flynn M P. A 90-MS/s 11-MHz-bandwidth 62-dB SNDR noise-shaping SAR ADC. IEEE J Solid-State Circ, 2012, 47: 2898–2904
Um J Y, Kim Y J, Song E W, et al. A digital-domain calibration of split-capacitor DAC for a differential SAR ADC without additional analog circuits. IEEE Trans Circ Syst I, 2013, 60: 2845–2856
Swindlehurst E, Chiang S W. Histogram-based calibration of capacitor mismatch in SAR ADCs. Electron Lett, 2015, 51: 2096–2098
Schreier R, Temes G C. Understanding Delta-sigma Data Converters. Hoboken: John Wiley & Sons, 2017
Shu Y S, Kuo L T, Lo T Y. An oversampling SAR ADC with DAC mismatch error shaping achieving 105 dB SFDR and 101 dB SNDR over 1 kHz BW in 55 nm CMOS. IEEE J Solid-State Circ, 2016, 51: 2928–2940
Chae Y, Han G. Low voltage, low power, inverter-based switched-capacitor delta-sigma modulator. IEEE J Solid-State Circ, 2009, 44: 458–472
Lee I, Han G, Chae Y. A 2 mW, 50 dB DR, 10 MHz BW 5× interleaved bandpass delta-sigma modulator at 50 MHz IF. IEEE Trans Circ Syst I, 2015, 62: 80–89
Shen L X, Lu N S, Sun N. A 1V 0.25µW inverter-stacking amplifier with 1.07 noise efficiency factor. In: Proceedings of Symposium on VLSI Circuits, Kyoto, 2017. 140–141
Shen L, Mukherjee A, Li S, et al. A 0.6-V tail-less inverter stacking amplifier with 0.96 PEF. In: Proceedings of Symposium on VLSI Circuits, Kyoto, 2019. 144–145
Shen L, Lu N, Sun N. A 1-V 0.25-µW inverter stacking amplifier with 1.07 noise efficiency factor. IEEE J Solid-State Circ, 2018, 53: 896–905
Shen L X, Shen Y, Tang X Y, et al. 3.4 A 0.01mm2 25µW 2MS/s 74dB-SNDR continuous-time pipelined-SAR ADC with 120fF input capacitor. In: Proceedings of IEEE International Solid-State Circuits Conference (ISSCC), San Francisco, 2019. 64–66
Shen L, Sun N, Shen Y, et al. A two-step ADC with a continuous-time SAR-based first stage. IEEE J Solid-State Circ, 2019, 54: 3375–3385
Park K, Yeom S, Kim S Y. Ultra-low power CMOS image sensor with two-step logical shift algorithm-based correlated double sampling scheme. IEEE Trans Circ Syst I, 2020, 67: 3718–3727
Liu J, Tang X, Zhao W, et al. A 13-bit 0.005-mm2 40-MS/s SAR ADC with kT/C noise cancellation. IEEE J Solid-State Circ, 2020, 55: 3260–3270
Rovere G, Fateh S, Benini L. A 2.2µW cognitive always-on wake-up circuit for event-driven duty-cycling of IoT sensor nodes. IEEE J Emerg Sel Top Circ Syst, 2018, 8: 543–554
Wang Z, Ye L, Liu Y, et al. A 148nW general-purpose event-driven intelligent wake-up chip for AIoT devices using asynchronous spike-based feature extractor and convolutional neural network. In: Proceedings of IEEE International Solid-State Circuit Conference (ISSCC), San Francisco, 2021. 436–438
Dong Q, Sinangil M E, Erbagci B, et al. A 351TOPS/W and 372.4GOPS compute-in-memory SRAM macro in 7nm FinFET CMOS for machine-learning applications. In: Proceedings of IEEE International Solid-State Circuit Conference, San Francisco, 2020. 242–244
Biswas A, Chandrakasan A P. Conv-RAM: an energy-efficient SRAM with embedded convolution computation for low-power CNN-based machine learning applications. In: Proceedings of IEEE International Solid-State Circuit Conference, San Francisco, 2018. 488–490
Yang J, Kong Y, Wang Z, et al. Sandwich-RAM: an energy-efficient in-memory BWN architecture with pulse-width modulation. In: Proceedings of IEEE International Solid-State Circuit Conference, San Francisco, 2019. 394–396
Valavi H, Ramadge P J, Nestler E, et al. A 64-tile 2.4-Mb in-memory-computing CNN accelerator employing charge-domain compute. IEEE J Solid-State Circ, 2019, 54: 1789–1799
Chen P, Wu M, Zhao W, et al. A 22-nm delta-sigma computing-in-memory (ΔΣCIM) SRAM macro with near-zero-mean outputs and LSB-first ADCs achieving 21.38TOPS/W for 8b-MAC edge AI processing. In: Proceedings of IEEE International Solid-State Circuit Conference, San Francisco, 2023. 140–141
Ghosh-dastidar S, Adeli H. Spiking neural networks. Int J Neur Syst, 2009, 19: 295–308
Liu Y, Chen Z Y, Wang Z X, et al. A 22nm 0.43pJ/SOP sparsity-aware in-memory neuromorphic computing system with hybrid spiking and artificial neural network and configurable topology. In: Proceedings of IEEE Custom Integrated Circuits Conference (CICC), 2023
Wang D, Kim S, Yang M, et al. A background-noise and process-variation-tolerant 109nW acoustic feature extractor based on spike-domain divisive-energy normalization for an always-on keyword spotting device. In: Proceedings of IEEE International Solid-State Circuit Conference (ISSCC), San Francisco, 2021. 372–374
Shan W W, Yang M H, Xu J M, et al. 14.1 A 510nW 0.41V low-memory low-computation keyword-spotting chip using serial FFT-based MFCC and binarized depthwise separable convolutional neural network in 28 nm CMOS. In: Proceedings of IEEE International Solid-State Circuit Conference, San Francisco, 2021. 230–232
Park J, Lee J Y, Jeon D, et al. 7.6 A 65-nm 236.5 nJ/classification neuromorphic processor with 7.5% energy overhead on-chip learning using direct spike-only feedback. In: Proceedings of IEEE International Solid-State Circuit Conference, 2019
Zhang Y, Xue C, Wang X, et al. Single-mode 6T CMOS SRAM macros with keeper-loading-free peripherals and row-separate dynamic body bias achieving 2.53fW/bit leakage for AIoT sensing platforms. In: Proceedings of IEEE International Solid-State Circuit Conference, San Francisco, 2022. 184–186
Osada K, Saitoh Y, Ibe E, et al. 16.7fA/cell tunnel-leakage-suppressed 16Mb SRAM for handling cosmic-ray-induced multi-errors. In: Proceedings of IEEE International Solid-State Circuit Conference, San Francisco, 2022. 302–494
Wang Z, Ye L, Zhang H, et al. A 57nW software-defined always-on wake-up chip for IoT devices with asynchronous pipelined event-driven architecture and time-shielding level-crossing ADC. In: Proceedings of IEEE International Solid-State Circuit Conference, 2020. 314–315
Wu T F, Ho C R, Chen M S W. A flash-based non-uniform sampling ADC with hybrid quantization enabling digital antialiasing filter. IEEE J Solid-State Circ, 2017, 52: 2335–2349
Weltin-Wu C, Tsividis Y. An event-driven clockless level-crossing ADC with signal-dependent adaptive resolution. IEEE J Solid-State Circ, 2013, 48: 2180–2190
Park K S, Khashaba A, Abdelrahman A, et al. A second-order temperature compensated 1µW/MHz 100MHz RC oscillator with ±140ppm inaccuracy from −40°C to 95°C. In: Proceedings of IEEE Custom Integrated Circuits Conference (CICC), Austin, 2021
Jiang H, Wang P H P, Mercier P P, et al. A 0.4–V 0.93-nW/kHz relaxation oscillator exploiting comparator temperature-dependent delay to achieve 94-ppm/°C stability. IEEE J Solid-State Circ, 2018, 53: 3004–3011
Savanth A, Myers J, Weddell A, et al., A 0.68nW/kHz supply-independent relaxation oscillator with ±0.49%/V and 96ppm/°C stability. In: Proceedings of IEEE International Solid-State Circuits Conference (ISSCC), San Francisco, 2017. 96–97
Zhang Y, You Y, Ren W, et al. A 0.954nW 32kHz crystal oscillator in 22nm CMOS with Gm-C-based current injection control. In: Proceedings of IEEE International Solid-State Circuit Conference, San Francisco, 2023. 68–70
Xu L, Jang T, Lim J, et al. A 510-pW 32-kHz crystal oscillator with high energy-to-noise-ratio pulse injection. IEEE J Solid-State Circ, 2022, 57: 434–451
Kim K M, Kim S, Choi K S, et al. A sub-nW single-supply 32-kHz sub-harmonic pulse injection crystal oscillator. IEEE J Solid-State Circ, 2021, 56: 1849–1858
Hsiao K J. A1.89nW/0.15V self-charged XO for real-time clock generation. In: Proceedings of IEEE International Solid-State Circuit Conference, San Francisco, 2014. 298–300
Acknowledgements
This work was supported by National Natural Science Foundation of China (Grant Nos. 92164301, 62225401) and 111 Project.
Author information
Authors and Affiliations
Corresponding authors
Rights and permissions
About this article
Cite this article
Ye, L., Wang, Z., Jia, T. et al. Research progress on low-power artificial intelligence of things (AIoT) chip design. Sci. China Inf. Sci. 66, 200407 (2023). https://doi.org/10.1007/s11432-023-3813-8
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s11432-023-3813-8