Fault diagnosis using state information

Repeated fault diagnosis on large integrated circuits may often be computationally prohibitive due to expensive fault simulation requirements. Fault dictionaries can help alleviate this problem, but they may be infeasible to store because of their large sizes, and more importantly, they typically provide only a black box view of the circuit and hence almost no diagnostic flexibility. The problem occurs because dictionaries usually only store primary output information. A new approach to fault diagnosis based on state information is presented. The selective storage of state information is shown to significantly improve the time for diagnostic fault simulation. We also describe a method to reduce the amount of information stored by choosing only a subset of the state space. This approach is shown to be ideally suited for partial scan circuits whose simple structure is exploited to reduce storage requirements. Experiments on the ISCAS 89 benchmark circuits are performed to demonstrate the efficiency of the state information based diagnosis technique.