On improved importance sampling simulations for CDMA communication systems with memory

The performance of the improved importance sampling simulation method is analyzed with respect to system memory and applied to the simulation of a multi-user mobile CDMA communication system over multipath fading channels. To combat the well-known inefficiency of Monte-Carlo simulations of complex multi-user CDMA systems, importance sampling methods have already been applied and shown to be very efficient, but they were subject to quite restrictive assumptions or to a greatly simplified simulation setup. We investigate the application to systems with memory in conjunction with improved importance sampling. Two approaches are analyzed depending on the knowledge of the probability density function of the noise at the output of the system, namely the input version and the output version. The application shows that, when both methods can be applied, the output version is to be prefered in terms of greatest sample savings (about 4 orders of magnitude at a signal to noise ratio of 10 dB)