Importance Sampling Simulation of Orthogonal Space-Time Block Coded Systems

Evaluation of bit error rates (BER) of a digital communication system is usually done via simulation using the Monte Carlo (MC) method. For low BERs, this method requires very large sample sizes to produce the rare error events. To overcome this limitation, importance sampling (IS) technique can be used to reduce simulation runtime. This reduction is achieved by modifying the input distribution and subsequently scaling the number of errors to make the estimator unbiased. In this paper, we illustrate the use of IS for simulating orthogonal space time block coded (OSTBC) systems over frequency nonselective Rayleigh fading channels. In particular, we show how to bias the Rayleigh and additive white Gaussian noise (AWGN) processes using the variance scaling and mean translation methods, respectively, in order to maximize the computational efficiency gain. It is demonstrated that the proposed method requires much smaller sample sizes to achieve the same accuracy required by a conventional MC estimator, especially at high signal to noise ratios