SICE: Design-dependent statistical interconnect corner extraction under inter/intra-die variations

Z. Feng, P. Li, Z. Ren

Research output: Contribution to journalArticlepeer-review

2 Scopus citations

Abstract

While traditional worst-case corner analysis is often too pessimistic for nanometer designs, full-blown statistical circuit analysis requires significant modelling infrastructures. In this study, a design-dependent statistical interconnect corner extraction (SICE) methodology is proposed. SICE achieves a good trade-off between complexity and pessimism by extracting more than one process corners in a statistical sense, which are also design dependent. Our new approach removes the pessimism incurred in prior work while being computationally efficient. The efficiency of SICE comes from the use of parameter dimension reduction techniques. The statistical corners are further compacted by an iterative output clustering method. Numerical results show that SICE achieves up to 260X speedups over the Monte Carlo method.

Original languageEnglish
Pages (from-to)248-258
Number of pages11
JournalIET Circuits, Devices and Systems
Volume3
Issue number5
DOIs
StatePublished - 2009

Fingerprint

Dive into the research topics of 'SICE: Design-dependent statistical interconnect corner extraction under inter/intra-die variations'. Together they form a unique fingerprint.

Cite this