Transient Simulation for High-Speed Channels with Recurrent Neural Network

Thong Nguyen, Tianjian Lu, Ju Sun, Quang Le, Ken We, Jose Schut-Aine

Research output: Chapter in Book/Report/Conference proceedingConference contribution

22 Scopus citations

Abstract

Recent success of recurrent neural network (RNN) in modeling time sequence has drawn a lot of attentions across multiple fields. Prior works have shown that RNN modeling can be suitable and even powerful for macro-modeling in circuit simulation. In this work, we propose using RNN for high-speed channel simulation in the time domain, which can handle the nonlinear behaviors of the IO buffers. The numerical example has demonstrated the capability and the accuracy of the proposed approach. Through the numerical example, we investigate the multiple well-known RNN structures on their capability of accurate transient channel simulation. We also examine the tunable parameters in the RNN model such as the optimization method in dealing with nonlinearities.

Original languageEnglish (US)
Title of host publicationEPEPS 2018 - IEEE 27th Conference on Electrical Performance of Electronic Packaging and Systems
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages303-305
Number of pages3
ISBN (Electronic)9781538693032
DOIs
StatePublished - Nov 13 2018
Externally publishedYes
Event27th IEEE Conference on Electrical Performance of Electronic Packaging and Systems, EPEPS 2018 - San Jose, United States
Duration: Oct 14 2018Oct 17 2018

Publication series

NameEPEPS 2018 - IEEE 27th Conference on Electrical Performance of Electronic Packaging and Systems

Conference

Conference27th IEEE Conference on Electrical Performance of Electronic Packaging and Systems, EPEPS 2018
Country/TerritoryUnited States
CitySan Jose
Period10/14/1810/17/18

Bibliographical note

Publisher Copyright:
© 2018 IEEE.

Fingerprint

Dive into the research topics of 'Transient Simulation for High-Speed Channels with Recurrent Neural Network'. Together they form a unique fingerprint.

Cite this