Pattern Perturbation Method for Multicanonical Monte Carlo Simulations in Optical Communications

In this letter, we present a novel method to efficiently account for pattern dependence in the bit-error-ratio (BER) estimation of an optical communication system through multi-canonical Monte Carlo (MMC) simulations. The proposed method considers the transmitted pattern as random, extending the MMC random-walk to the overall state space comprising noise and pattern. The resulting complexity is only linear in the intersymbol interference length, allowing us to account for long pattern dependence in a feasible computational time. The accuracy of the method is confirmed by the agreement of simulations with the exact probability density functions and BER values evaluated through the Karhunen-Loeve series expansion method.