On Accuracy and Efficiency of Monte-Carlo BER Simulations for Fading Channels

Accuracy and efficiency of Monte-Carlo (MC) bit error rate (BER) simulation for a fading channel is discussed. While the simulation error is minimized for given total number of trials by using one symbol/noise realization per channel realization (M = 1), it does not result in the best efficiency. The latter can be further improved by vectorizing the "noise part" of the simulation and using M > 1. Extensive numerical simulations validate these conclusions. Efficient numerical procedure of finding optimum M is suggested. The proposed technique preserves the problem-independence of the original MC method, and can also be combined with any known importance sampling method for better simulation efficiency.