Comparison of two biasing Monte Carlo methods for calculating outage probabilities in systems with multisection PMD compensators

We evaluate the performance of single-section and three-section polarization-mode-dispersion (PMD) compensators using the biasing Monte Carlo methods of importance sampling (IS) and multicanonical Monte Carlo (MMC). We show that standard IS that biases only first-order PMD is insufficient to compute penalties in most compensated systems, while multiple IS that biases both first- and second-order PMD and MMC work well with all the compensators that we investigated. We show that multiple IS works well even in a system with a three-section compensator, when both first- and second-order PMD are compensated. The applicability of IS in these systems is consistent with the existence of a large correlation between first- and second-order PMD of the transmission line and higher orders of PMD after compensation, so that the first two orders, even when compensated, remain highly correlated with the residual penalty. We directly demonstrate the existence of this correlation.