Abstract

The design and optimization of modern optical fiber transmission systems relies on an increasingly close interaction between computer modeling and laboratory experiments. However, the full potential of modeling tools will not be realized as long as random processes, such as amplifier noise and fiber birefringence, are modeled using time-consuming Monte Carlo simulations. In practice Monte Carlo simulation is simply not capable of accurately computing error rates or outage probabilities, both of which are strongly influenced by extremely rare, worst-case events. By an outage probability we mean the probability that a system penalty exceeds an allowed margin. Even when Monte Carlo simulation is used to compute average measures of the system performance, such as the Q-factor, it can be too inefficient to be useful for optimizing the design of long-haul wavelength-division multiplexed (WDM) transmission systems. Moreover, a system that has been optimized to produce the best average behavior may not necessarily have the lowest possible error rate or outage probability [1].

© 2002 Optical Society of America