Characterization algorithm of failure distribution for LSI yield improvement

Abstract

This paper describes an improvement in our algorithm, which can efficiently characterize a process-induced random failure distribution and a design-induced systematic failure distribution from unknown-induced failure distributions of a memory LSI, to predict a reason for yield degradation in it. The algorithm analyzes a function "T(f) isn't greater than 1or not" related to kind and content of "f". The "f" is a divisor of distances between failure-pairs. We have expanded the algorithm, which can pick out 7 characteristic failure distributions by using relationship between the failure densities and the function "T(f)".