MePLER: A 20.6-pJ Side-Channel-Aware In-Memory CDT Sampler

Abstract

This work presents a MePLER, an in-Memory cumulative distribution table (CDT) sampler, featuring custom cell derived from NAND-Type CAM for range-matching, pipelined and segmented array for reduced energy, and suppressed timing and power side-channel leakage. The precision and sample range are configurable for different sampling requirements. A 65nm prototype achieves constant 85.9-MSps, 1-sample/cycle throughput, 20.6-pJ/sample efficiency, and 0.03-mm2 footprint.

Publication
2021 Symposium on VLSI Circuits
Avatar
Dai Li
PhD 2021, now at Google
Avatar
Yan He
PhD 2023, now at Nvidia
Avatar
Akhil Pakala
MECE 2024, now at Apple
Avatar
Kaiyuan Yang
Associate Professor of ECE