Abstract
Accurate evaluation of four-wave-mixing (FWM) contributions is one of the most time consuming problems in numerical simulations of wavelength-division-multiplexed light-wave systems. A too large constant step size can lead to widely overestimate the output FWM power. In this letter for the first time to our knowledge, I analyze the maximum allowed constant step size to be utilized in split-step Fourier method simulations, for obtaining a given accuracy in the FWM power evaluation. A very general rule is given, valid for any possible value of fiber dispersion, channel separation, channel power and number of channels considered in the simulation.