An efficient AVF estimation technique using circuit partitioning

Abstract

Soft errors induced by radiation particles are increasingly becoming a source of concern for reliable design of VLSI systems. An important parameter to quantify the soft error response of a system is Architectural Vulnerability Factor (AVF), which is the probability that a fault in a part of the system will result in an error. One of the most common techniques to estimate AVF of a system is by using statistical fault injection (SFI). Traditional SFI techniques can be computationally inefficient with increasing design complexity. In this work, a novel technique has been developed to enable computationally efficient AVF estimation using enhanced node observability. Possible target applications for this technique have been identified with results showing orders of magnitude improvement over traditional SFI techniques.