Fast fault coverage estimation of sequential tests using entropy measurements
Michael S. Hsiao, Sarmad Tanwir · 2018
In this paper, we explore the relationship between fault coverage for sequential circuits and the observed entropy during test application. We find that the entropy measured at a subset of the primary and pseudo-primary outputs is highly and significantly correlated to the regions of the design that are exercised by the test vectors. Consequently, they can be used for estimating the coverages for the newly generated test sets. As computing the entropy requires only logic simulation, it can greatly speed up the fault grading process. In our experiments on the ISCAS'89 benchmarks, we are able to achieve an average root mean squared error of 2.1% in fault coverage estimation for test sets of varying lengths, together with an order of magnitude speedup over fault simulation.