Coping with imperfect Design and Manufacturing Defects

Fault diagnosis plays a very important role in enhancing the yield of the VLSI manufacturing process. A number of researchers have proposed various techniques for fault diagnosis over the years, but a number of practical problems still remain. In particular, the problem of multiple fault diagnosis and characterization of test sets from the point of view of diagnosis did not receive much attention. The first part of the talk will discuss a framework to analyze multiple faults, and diagnose the same. Based on multiple fault simulation in a particle swarm optimization environment, experimentation on benchmarks reveal that up to ten faults can be diagnosed within a reasonable amount of time. The second part of the talk will discuss a new metric to characterize test sets based on their diagnostic power. The metric can also be utilized to increase the diagnosability of incompletely specified test sets using don't care filling. The don't care filling approach can be integrated with test pattern generation tools to aid in better diagnostic pattern set generation.