SerOpt: Transistor Sizing Algorithm and Optimization Utility for Minimizing Soft Error Rate

Yehuda Kra, Yoav Weitzman, Adam Teman

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

This paper presents a transistor sizing algorithm and utility for soft-error (SER) protection circuits. The algorithm applies a gradient-descent convergence scheme for rapidly calculating an optimized sizing configuration per specific protection circuit instance. The utility efficiently interacts with commercial SPICE circuit-level tools for highest accuracy. The analysis and optimization flow is distributed over parallel compute resources to provide a scalable and feasible solution for large designs. The solution is demonstrated on two configurations of the Muller C-element circuit, a widely used component for SER protection across designs. By sizing the test circuits with the proposed utility, a SER protection improvement of as much as two orders-of-magnitude is achieved over a baseline design within minutes of compute runtime.

Original languageEnglish
Title of host publicationDCIS 2022 - Proceedings of the 37th Conference on Design of Circuits and Integrated Systems
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781665459501
DOIs
StatePublished - 1 Jan 2022
Event37th Conference on Design of Circuits and Integrated Systems, DCIS 2022 - Pamplona, Spain
Duration: 16 Nov 202218 Nov 2022

Publication series

NameDCIS 2022 - Proceedings of the 37th Conference on Design of Circuits and Integrated Systems

Conference

Conference37th Conference on Design of Circuits and Integrated Systems, DCIS 2022
Country/TerritorySpain
CityPamplona
Period16/11/2218/11/22

All Science Journal Classification (ASJC) codes

  • Hardware and Architecture
  • Energy Engineering and Power Technology
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'SerOpt: Transistor Sizing Algorithm and Optimization Utility for Minimizing Soft Error Rate'. Together they form a unique fingerprint.

Cite this