Research On Nonlinear Frequency Division Multiplexing Optical Fiber Communication Systems

With the maturity and wide application of the fifth-generation mobile communication technology,optical fiber communication has gradually become the backbone of the global telecommunication network.However,the birth of various new applications and new services also requires a rapid increase in the capacity of the communication system,which puts forward higher requirements on the transmission performance and spectrum resource utilization of the optical fiber communication system.Optical fiber is the medium of signal transmission in the system,and its nonlinear effect is the most important factor limiting long-distance information transmission in optical fiber communication systems.Therefore,nonlinear compensation has always been a research hotspot in the field of optical fiber communication.Traditional nonlinear compensation schemes,including digital back propagation(DBP),optical phase conjugation(OPC)and phase conjugated twin waves(PCTW),have basically reached the bottleneck of compensation effect.It is difficult to bring about substantial performance improvements.In recent years,a new optical fiber communication system transmission scheme called nonlinear frequency division multiplexing(NFDM)has attracted extensive attention from researchers,which transforms the nonlinear problem into a linear problem by mapping the dispersion and nonlinearity to a linear phase rotation in the nonlinear frequency domain.Based on the research and analysis of Nonlinear Fourier Transform(NFT),this paper realizes the NFDM transmission system schemes of nonlinear continuous spectrum and discrete spectrum respectively.And based on the machine learning algorithm,in view of the weakening of system integrability in the actual transmission environment,two equalization schemes at the receiving end are proposed to further optimize the system performance.The main research work is as follows:(1)The basic theory of NFT and INFT is studied and the algorithm is verified.In this paper,the nonlinear Schr(?)dinger equation is used as the theoretical model of optical signal propagation in optical fiber,and the problem of invariant eigenvalues of the model is solved based on the mathematical method Lax pair.The mathematical process of NFT is theoretically studied,and the specific calculation methods of nonlinear continuous spectrum and discrete spectrum are obtained.At the same time,the NFT and INFT algorithms of continuous spectrum and discrete spectrum are realized in MATLAB,and the above algorithm is further verified by multi-angle simulation to ensure its accuracy and robustness.(2)The scheme of NFDM optical fiber transmission system based on nonlinear spectrum modulation is realized.The scheme is based on NFT and INFT algorithms,and the NFDM transmission system of continuous spectrum and discrete spectrum modulation is respectively completed through joint debugging of MATLAB and VPI.For the NFDM system with continuous spectral modulation,under the same noise condition,the bit error rate performance of the system under different modulation formats and subcarrier numbers is compared.In addition,it is also found that the performance of NFDM system with continuous spectral modulation is better than that of the traditional Orthogonal Frequency Division Multiplexing(OFDM)system.And the Q factor under the optimal transmission condition is 0.6dB higher than that of the OFDM system.For the NFDM system with discrete spectral modulation,this paper implements a multi-soliton transmission scheme that multiplexes 7 eigenvalues,and obtains its simulation results under QPSK and 16QAM,where the Q factor of the system reaches 7.64 dB at 16QAM.(3)A feature engineering based model parameter transfer convolutional neural network equalization scheme for NFDM system with continuous spectral modulation is proposed.It is verified by simulation that this scheme can effectively improve the Q factor of the system.And under the 16QAM modulation format,according to the data at the receiving end in different transmission environments,the generation rate of the FE-CNN equalizer adapted to the new scene is increased by more than 50%through the method of transfer learning,which greatly improves the equalization efficiency at the receiving end of the system.It can effectively deal with the damage of system integrability caused by factors such as equipment noise and high-order dispersion in the actual transmission environment.(4)A equalization scheme based on Gaussian mixture model clustering for NFDM system with discrete spectral modulation is proposed.The Q factor of the system under the condition of different eigenvalues is compared by experiments.It is shown that the scheme can improve the bit error rate performance of the system to a certain extent,and is suitable for the NFDM system with discrete spectral modulation that multiplexes different eigenvalues.By comparing with the K-means equalization scheme under different transmission powers,it is further verified that the GMM scheme has a better optimization effect on the NFDM system with discrete spectral modulation.