Failure Probability and Fault Observability of SRAM-FPGA Systems

We describe a simulation-based fault injection technique for failure probability and fault observability assessment of SRAM-FPGA systems. Our approach relies on a model of FPGA netlists realised with the Stochastic Activity Networks formalism. Faults can be injected into the model either stochastically or exhaustively one at a time. Fault propagation is traced to the output pins, using a four-valued logic that enables faulty signals to be tagged and recognized. We considered some of the ITC'99 benchmarks as examples