Testing a Nanocrossbar for Multiple Fault Detection

This paper proposes an approach for testing a nanocrossbar switch; fault detection is considered in the presence of faulty switches and nets (of a permanent nature only) in the crossbar. To ensure detection, a one-to-one (onto) relationship in the setting (programming) of the switches is established in each of the configurations of the crossbar. This is accomplished using a constant-sum transformation of the characteristic matrix of the crossbar by utilizing different graph algorithms in O(N^4.5) where N is the matrix dimension. Matrix properties are related to graph algorithms to generate permutation matrices as corresponding to the configurations (phases) of the crossbar. The conditions by which multiple faults are detected by the modified counting sequence (as test set), are proved. Simulation results are provided to further substantiate the validity of the proposed approach to test nanocrossbars of very large dimension and with different switch distribution.