Adapting to varying distribution of unknown response bits

Traditionally, test patterns that are generated for a given circuit are applied in an identical manner to all manufactured devices until each device under test either fails or passes each test. With increasing process variations, the statistical diversity of manufactured devices is increasing, making such one-sizefits-all approaches increasingly inefficient. Adaptive test techniques address this problem by tailoring the test decisions for the statistical characteristics of the device under test. In this article, we present several adaptive strategies to enable adaptive unknown bit masking for faster-than-at-speed testing so as to ensure no yield loss while attaining the maximum test quality based on tester memory constraints. We also develop a tester-enabled compression scheme that helps alleviate memory constraints further, shifting the tradeoff space favorably to improve test quality.