A Readout Scheme for PCM-Based Analog In-Memory Computing With Drift Compensation Through Reference Conductance Tracking

A Readout Scheme for PCM-Based Analog In-Memory Computing With Drift Compensation Through Reference Conductance Tracking

This article presents a readout scheme for analog in-memory computing (AIMC) based on an embedded phase-change memory (ePCM). Conductance time drift is overcome with a hardware compensation technique based on a reference cell conductance tracking (RCCT). Accuracy drop due to circuits mismatch and va...

Full description

Saved in:
Bibliographic Details
Main Authors: Alessio Antolini, Andrea Lico, Francesco Zavalloni, Eleonora Franchi Scarselli, Antonio Gnudi, Mattia Luigi Torres, Roberto Canegallo, Marco Pasotti
Format: Article
Language:English
Published: IEEE 2024-01-01
Series:IEEE Open Journal of the Solid-State Circuits Society
Subjects:
Analog in-memory computing (AIMC)
artificial intelligence
drift compensation
matrix-vector multiplication (MVM)
phase-change memory (PCM)
Online Access:https://ieeexplore.ieee.org/document/10609348/
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This article presents a readout scheme for analog in-memory computing (AIMC) based on an embedded phase-change memory (ePCM). Conductance time drift is overcome with a hardware compensation technique based on a reference cell conductance tracking (RCCT). Accuracy drop due to circuits mismatch and variability involved in the computational chain are minimized with an optimized iterative program-and-verify algorithm applied to the phase-change memory (PCM) devices. The proposed AIMC scheme is designed and manufactured in a 90-nm STMicroelectronics CMOS technology, with the aim of adding a signed multiply-and-accumulate (MAC) computation feature to a Ge-Rich GeSbTe (GST) embedded PCM array. Experimental characterizations are performed under different operating conditions and show that the mean MAC decrease in time is approximately null at room temperature and reduced by a factor of 3 after 64-h bake at <inline-formula> <tex-math notation="LaTeX">$85~{^{\circ }}$ </tex-math></inline-formula>C. Based on several MAC operations, the estimated <inline-formula> <tex-math notation="LaTeX">$512\times 512$ </tex-math></inline-formula> matrix-vector-multiplication (MVM) accuracy is 97.4%, whose decrease in time is less than 3% in the worst case.
ISSN:2644-1349

Similar Items