Reducing the computation time in simulations of fading channels (mobile radio system design)

Monte Carlo simulations require large amounts of computation time for reliable results when the bit-error rate (BER) is small. This is particularly true for fading channels since errors are clustered, and it is the number of fades, more than the number of bits, which determine the accuracy. It is shown that importance sampling can reduce the run times of such simulations. It is demonstrated that biasing the simulation program to produce more fades can reduce the variance of the estimates of the BER. An efficient method for creating channel gain sequences associated with fades, according to an arbitrarily specified fading spectrum, is presented.<
>