BER simulation of digital communication systems with intersymbol interference and non-Gaussian noise using improved importance sampling

The authors present an efficient and robust solution to the problem of simulating the bit error rate (BER) performance of digital communication systems with intersymbol interference and nonGaussian noise. BER simulation results are shown for the conventional importance sampling (CIS) and improved importance sampling (IIS) techniques. In each case a bounding technique was applied to ensure that the BER estimates simulated were obtained with a specified estimate variance. The simulation improvement ratio over a direct Monte Carlo approach was derived and showed that the IIS technique gives a constant improvement for any size of system memory length. The CIS technique was shown to have a detrimental dimensionality effect, such that its effectiveness decreases as the system memory length increases.<>