Drift of Combinational Circuits Failure Rates with a Probabilistic Model Approximated by Partitioning

This paper addresses applications subject to stringent safety constraints. The study focuses on comparing two analytical models for assessing fault propagation in combinational circuits. The first model is exhaustive and exact. The second approximates the first one as it partitions netlists into sub circuits, that are separately analyzed with the first model. Thanks to the clustering approach, the second model is more scalable than the reference model. We demonstrate through analytical arguments and experimental results that the approximating model is not optimistic compared to the reference probability values. The experimental comparisons also show that the results of the second model remain close to the reference ones.