OBDD-based optimization of input probabilities for weighted random pattern generation

Abstract

Numerous methods have been devised to compute and to optimize fault detection probabilities for combinational circuits. The methods range from topological to algebraic. In combination with OBDDs, algebraic methods have received more and more attention. Recently, an OBDD based method has been presented which allows the computation of exact fault detection probabilities for many combinational circuits. We combine this method with strategies making use of necessary assignments (computed by an implication procedure). The experimental results show that the resulting method leads to a decrease of the time and space requirements for computing fault detection probabilities of the hard faults by a factor of 4 on average compared to the original algorithm. By this means it is now possible to efficiently use the OBDD based approach also for the optimization of input probabilities for weighted random pattern testing. Since in contrast to other optimization procedures this method is based on the exact fault detection probabilities we succeed in the determination of weight sets of superior quality, i.e. the test application time (number of random patterns) is considerably reduced compared to previous approaches.<>