A test-application-count based learning technique for test time reduction

Abstract

One popular adaptive test approach is to reorder the test patterns according to their fault detection performance - by applying the more effective patterns first, the total test time can be significantly reduced. While very effective, the detection performance based approach fails to identify some high-quality test patterns and leaves them unused throughout the test application process. In this paper, we propose a test-application-count based learning technique to help identify high-quality test patterns. By ensuring that all patterns are applied for at least the specified number of times, the proposed technique finds more high-quality test patterns and moves them to the front of the test pattern list. Experimental results show that the proposed test-application-count based learning technique achieves 52% test time reduction (TTR) in average - a 12% improvement compared to the detection performance based approach.