Performance Evaluation of Viterbi Decoders by Multicanonical Monte Carlo Simulations

We propose a novel simulation-based method to evaluate the performance of Viterbi decoders. In particular, we address scenarios where the error probability is very low, that is, scenarios where classical Monte Carlo simulations would require impractical execution times before producing reliable results. As other recent fast-simulation approaches, the proposed method relies on the multicanonical Monte Carlo technique, but, unlike the existing general-purpose methods, it is specifically designed for Viterbi decoders, the algorithm being driven by a control variable that depends on the state metrics of the various survivors over the trellis. In simple scenarios for which analytical tools are available, the simulation results agree with them, while, in the most common scenarios, no analytical tool is available and the proposed method gives the fastest way for the estimation of low error probabilities.