Performance analysis of Viterbi decoding of convolutional coded data transmitted over a non-linear channel using MODSIM. Volume 1: Theory and test cases

A method is described for computing the upper bound of the event and bit error probabilities for convolutional coded data in the presence of correlated or signal dependent noise. Preference probabilities for error paths are computed by a quasi Monte Carlo technique known as importance sampling and extrapolated for longer paths. The required upper bounds are computed as sums of preference probabilities weighted by the corresponding numbers of error paths or the error counts on them. Examples are given for hard and soft decision decoding metrics.