MCM interconnect test scheme based on adaptive genetic algorithm

Abstract

Interconnect test technology has become a bottleneck in the application of multi-chip module (MCM), so study on new methods of test generation to acquire better test set is significant. This paper presents a novel optimization approach of adaptive genetic algorithm (AGA) for the MCM interconnect test generation problem. By combing the characteristics of MCM interconnect test, an accurate fitness function is designed to compute the fitness of each candidate vector. AGA is composed of populations of chromosomes and three evolutionary operators: selection, crossover and mutation. The international standard MCM benchmark circuit was used to verify the approach. Comparing with not only the evolutionary algorithms, but also the deterministic algorithms, experimental results demonstrate that the hybrid approach can achieve high fault coverage, short CPU time and compact test set, which shows that it is a novel optimized method deserving research.