Multicanonical Monte Carlo Method Applied to the Investigation of Polarization Effects in Optical Fiber Communication Systems

Abstract

Polarization-mode dispersion (PMD) is a major source of impairments in optical fiber com‐ munication systems. PMD causes distortion and broadens the optical pulses carrying infor‐ mation and lead to inter-symbol interference. In long-haul transmission systems it is necessary to limit the penalty caused by polarization effects [1], so that the probability of ex‐ ceeding a maximum specified penalty, such as 1 dB, will be small, typically 10-5 or less. This probability is referred as the outage probability. Since PMD is a random process, Monte Car‐ lo simulations are often used to compute PMD-induced penalties. However, the rare events of interest to system designers, which consists of large penalties, cannot be efficiently com‐ puted using standard (unbiased) Monte Carlo simulations or laboratory experiments. A very large number of samples must be explored using standard unbiased Monte Carlo simu‐ lations in order to obtain an accurate estimate of the probability of large penalties, which is computationally costly. To overcome this hurdle, advanced Monte Carlo methods, such as importance sampling (IS) [2], [3] and multicanonical Monte Carlo (MMC) [4] methods, have been applied to compute PMD-induced penalties [5], [6] using a much smaller number of samples. The analytical connections between MMC and IS are presented in [7], [8], [9], [10]. The MMC method has also been used to estimate the bit-error rate (BER) in optical fiber communication systems due to amplified spontaneous emission noise (ASE) [11], for which no practical IS implementation has been developed, and to estimate BER in spectrum-sliced wavelength-division-multiplexed (SS-WDM) systems with semiconductor optical amplifier (SOA) induced noise [12]. More recently, MMC has been used in WDM systems, where the