Generation of optimized single distributions of weights for random built-in self-test

This paper presents a probabilistic-based fault-model approach to the generation of optimized single distributions of weights for random built-in self-test. Many techniques use multiple sets of weights to obtain an important reduction in the test length. However, these strategies consume large memory areas to store the different distributions. In order to obtain a highly optimized set of weights, which reduces area overhead, a global optimization procedure is used to minimize a testability cost function that projects random fault coverage. Important reduction in test length, using only a highly optimized single distribution of weights, is reported