Estimating Bit Error Rates and Noise Margins via Simulation

Monte Carlo simulations, with their inherent uncertainty, are often used to determine the bit error rate of digital communication systems operating in the presence of noise. To assess the noise margin in a system operating with an acceptable bit error rate, this noisy functional relationship must be inverted. We describe a novel technique with improved speed and comparable accuracy for estimating the minimum Eb/N0, given the acceptable bit error rate, which can be used with almost any Monte Carlo simulator. This technique has been used to speed up simulation runs by a factor of more than 20 while designing a commercial RF chip. We show that computing the grand average of multiple runs to estimate the bit error rate, when the underlying error process is Poisson, is the best one can do in the absence of any other information. Lastly, we mention areas for future investigation.