TY - GEN
T1 - A Compact Low-Power Time-Domain Winner-Take-All Circuit
AU - Seol, Min Seok
AU - Kong, Bai Sun
N1 - Publisher Copyright:
© 2023 IEEE.
PY - 2023
Y1 - 2023
N2 - In this paper, a compact low-power time-domain winner-take-all (WTA) circuit is proposed. The proposed WTA circuit has a global inhibition signal to inhibit the inputs except the one exceeding the threshold voltage first. The inhibition signal has another important role in preventing redundant transitions in the circuit, resulting in a substantially reduced switching power consumption. Furthermore, the inhibition signal allows the proposed WTA circuit to be configured as a fully parallel structure even for a large number of inputs, achieving compact area and lowlatency operation. The proposed WTA circuit was designed in a 2Snm CMOS process. The performance evaluation results indicated that the proposed WTA circuit improved energy consumption by 88% and layout area by 73.2% as compared to conventional WTA circuits.
AB - In this paper, a compact low-power time-domain winner-take-all (WTA) circuit is proposed. The proposed WTA circuit has a global inhibition signal to inhibit the inputs except the one exceeding the threshold voltage first. The inhibition signal has another important role in preventing redundant transitions in the circuit, resulting in a substantially reduced switching power consumption. Furthermore, the inhibition signal allows the proposed WTA circuit to be configured as a fully parallel structure even for a large number of inputs, achieving compact area and lowlatency operation. The proposed WTA circuit was designed in a 2Snm CMOS process. The performance evaluation results indicated that the proposed WTA circuit improved energy consumption by 88% and layout area by 73.2% as compared to conventional WTA circuits.
KW - convolutional neural networks
KW - spiking neural networks
KW - time-domain WTA
KW - Winner-take-all
UR - https://www.scopus.com/pages/publications/85183578251
U2 - 10.1109/ICECS58634.2023.10382779
DO - 10.1109/ICECS58634.2023.10382779
M3 - Conference contribution
AN - SCOPUS:85183578251
T3 - ICECS 2023 - 2023 30th IEEE International Conference on Electronics, Circuits and Systems: Technosapiens for Saving Humanity
BT - ICECS 2023 - 2023 30th IEEE International Conference on Electronics, Circuits and Systems
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 30th IEEE International Conference on Electronics, Circuits and Systems, ICECS 2023
Y2 - 4 December 2023 through 7 December 2023
ER -