Stratified Sampling Monte Carlo Algorithm for Efficient BER Estimation in Long-Haul Optical Transmission Systems

We propose an efficient stratified sampling (SS) algorithm for estimating the bit error rate (BER) of a digital communication system. Our algorithm efficiently exploits the observations of an approximate, but usually fast, model of the system under investigation to drive a clever Monte Carlo (MC) estimation based on SS. The proposed method is faster than standard MC even at BER in the range 10^-3 to 10^-5 . Moreover, it is possible to evaluate the estimated standard deviation of the measured BER, such as in an MC simulation, so that it is possible to associate a confidence to the results. We test the algorithm both in a simple optical system distorted by group velocity dispersion (GVD) and in more complex differential quadrature phase shift keying (DQPSK) systems. In the last case, we measured computational savings up to 70% compared with standard MC.