Performance of Simulated Annealing Algorithm in Equalized Multimode Fiber Links With Linear Equalizers

Abstract

This paper reports on the suitability of the simulated annealing (SA) optimization algorithm to equalize multimode fiber (MMF) links using linear equalizers through simulation tools and experimental validation. Several variations of the SA algorithm and the selection of initial tap weights through the use of a predefined set of tap weights were examined. The authors also experiment with several eye mask shapes that can be used as a cost function and found that the symoctagon offers the best performance. Transmission experiments and simulations at 10 Gb/s were carried out on a large number of 300-m-long MMF links with a feedforward equalizer at the receiver using both legacy (OM1 and OM2) and next-generation (OM3) fibers. Simulation and experimental results show that the SA algorithm in its various forms is well suited to equalize MMF links with a linear equalizer. The algorithm showed relatively fast convergence with a tradeoff between convergence speed and final value for the eye opening. The SA algorithm converges well even in the presence of noise, and the link performance is limited by the signal-to-noise ratio rather than the adaptation algorithm