Improving the accuracy of defect diagnosis by test score based on fault free

Abstract

In the initial phase of production improvement for new technology, a fault chip may contain multiple defects. Owing to the interaction between defects, the state-of-the-art defect diagnosis procedure based on the simulation of single-modelled faults may generate a large number of candidate faults. Early methods based on test scores removed the correct candidate faults when narrowing the set of candidate faults, which caused a drop in diagnostic accuracy. This Letter proposes a new method combined with test scores to make better use of diagnostic information to reduce the number of candidate faults and improve diagnostic accuracy. The approach computes a score to every test by using fault-free output responses of a circuit, which can estimate its ability of fault identification. This results in a set of candidate faults with a reduced number of redundant faults and an increased number of real faults. Thus, the diagnosis resolution and precision are improved. The method can reduce the running time of the procedure for it uses the fault-free information to calculate the test scores offline. The experimental results on benchmark circuits demonstrate the efficacy of the proposed method.