IMPORTANCE SAMPLING METHODOLOGIES FOR SIMULATION OF COMMUNICATION-SYSTEMS WITH TIME-VARYING CHANNELS AND ADAPTIVE EQUALIZERS

Formula-based analysis of communication links with time-varying channe ls and adaptive equalizers is not always tractable-Monte Carlo simulat ion must be used to obtain bit error rate (BER) estimates of the syste m. Conventional Monte Carlo simulation will not provide estimates for low BER's due to excessive run time. Although importance sampling (IS) techniques offer the potential for large speedup factors for BER esti mation using MC simulation, IS techniques have not been used for simul ating communication links with adaptive equalizers. In this paper, we present (for the first time) two IS methodologies for MC simulation of communication links characterized by time-varying channels and adapti ve equalizers. One methodology is denoted as the ''twin system'' (TS) method. A key feature of the TS method is that biased noise samples ar e input to the adaptive equalizer, but the equalizer is only allowed t o adapt to these samples for a time interval equal to the memory of th e system. In addition to the TS technique, we also present a statistic ally biased, but simpler, technique for using IS with adaptive equaliz ers which is based on the independence assumption between equalizer in put and equalizer taps (the ''IA'' method). Experimental results are p resented that show run-time speedup factors of two to seven orders of magnitude for a static linear channel with memory, and of two to almos t five orders of magnitude for a slowly-varying random linear channel with memory for both the IA and TS methods.