Research On Nonlinear Effect Equalization Technology For Few-mode Fiber Mode Division Multiplexing System

In optical communication system,multiple orthogonal optical waveguide modes in optical fiber are used as independent channels for transmission which is called mode division multiplexing(MDM)technology,which can realize the doubling of system capacity.Compared with multi-mode fiber,few-mode fiber has strong enough anti-mode coupling ability and larger system capacity than single-mode fiber,so using few-mode fiber is a good compromise method.But in the actual transformation process,the nonlinear effect,dispersion and attenuation link damage and manufacturing defects caused by the orthogonality of mode coupling between modes will reduce the transmission performance of system,including the linear damage in the fiber link which can be compensated by digital signal processing,therefore,how to carry out the nonlinear damage compensation,how to realize nonlinear equalization is the focus in few-mode fiber mode division multiplexing system.In this paper,the nonlinear equalization problem of the mode multiplexing system is studied.Firstly,the link damage and mode coupling in the mode multiplexing system are introduced.Based on the transmission theory,the transmission model of the multiplexing system is established.The main work of this paper includes the following aspects:(1)The link damage in the few-mode fiber is analyzed,including nonlinear effect,linear effect and mode coupling.Based on the Generalized multi-mode Nonlinear Schrodinger Equation(GMM-NLSE),the equation is simplified in a 2×2 MDM module division multiplexing system.The model of 2×2 MDM system is built based on the design of weakly coupled few-mode fiber which can be demultiplexed.(2)The nonlinear equalization technique of mode division multiplexing system is studied.At first,the classical back propagation(BP)algorithm is introduced.Aiming at the problem that this algorithm leads to poor real-time performance of signals,this paper adopts the nearest neighbor algorithm based on machine learning,including K-nearest neighbor(KNN)algorithm,distance-weighted KNN(DW-KNN)algorithm and artificial neural network(ANN)algorithm for nonlinear equalization.The basic principle and flow of the three algorithms are introduced in detail.(3)The algorithm performance is compared from three aspects,including bit error rate,constellation map and nonlinear tolerance after nonlinear equalization in the 2×2 MDM system.Simulation results show that KNN,DW-KNN and ANN algorithms can reduce bit error rate in a 2×2 56Gb/s 16QAM mode division multiplexing system,Q factor of ANN algorithm is 0.5db higher than that of DW-KNN algorithm and 0.8dB higher than that of KNN algorithm.When OSNR is 18dB and bit error rate is 10-3,the nonlinear tolerance of ANN algorithm is 1.4dB than that of DW-KNN algorithm and 2.1dB higher than that of KNN algorithm.Therefore,ANN algorithm has better nonlinear equalization performance.