Research On Receiver Of NFDM System With Discrete Spectrum Modulation

In recent years,to overcome the limitation of Kerr nonlinearity which limits the transmission capacity of optic fiber communication systems,researchers have proposed a revolutionary scheme,namely nonlinear frequency division multiplexing(NFDM)system based on nonlinear Fourier transform(NFT).The system takes nonlinear as the inherent property of the system and theoretically avoids the influence of nonlinear effects.Although researchers have carried out a large number of confirmatory experiments based on discrete spectrum modulated NFDM system,there are still many challenges.One of the challenges is to find the optimal receivers for multiple-eigenvalue NFDM systems.The spectral efficiency(SE)of NFDM system increases with the increase of the number of eigenvalues,and the demodulation complexity also increases sharply,which reduces the accuracy of demodulation significantly.This may result in the failure of the receiver's NFT algorithm to demodulate multiple eigenvalues due to the problem of searching false roots or insufficient number of eigenvalues when recovering eigenvalues.Further,the amplifier spontaneous emission(ASE)noise,the phase noise(PN)introduced by the lasers,and the application of the polarization division multiplexing(PDM)technique make the demodulation even more difficult.In this paper,a receiver based on regression neural network(NN)is proposed,which takes the received time-domain waveform as the input and the constellation coordinate as the output for training,avoiding the process of searching for eigenvalues.By selecting and optimizing the internal structure of neural network reasonably,the NN receiver can realize the prediction of continuous variables and obtain a higher robustness against impairments,which is more suitable for practical NFDM transmission systems.The main research work is as follows:1.By studying the basic theory of discrete spectral NFDM system,an NN receiver based on regression neural network model is proposed,which is used to replace NFT algorithm to realize the demodulation of multiple eigenvalues.The inevitable ASE noise and phase noise in the system are modeled,and the selection of training data,the design of concrete structure and the setting of hyperparameters of the NN receiver are analyzed in the presence of damage effects.2.The operating principle and complexity of three NFDM receivers are analyzed systematically,including NFT receiver,NN receiver and Minimum Euclidean Distance(MD)receiver.The simulation platform of discrete spectrum NFDM coherent optical communication system is built to verify the performance of the three receivers according to relevant theories.3.It is verified by simulation that the NN receiver has strong robustness,and has a certain tolerance to the damage of communication system.It can transmit 1000 km with low bit error rate(BER)in SP-NFDM system with 2 GBaud 16QAM modulation format and four eigenvalue modulation.The performance of three receivers(NFT,MD and NN)in NFDM system with two eigenvalue modulation is compared.The NN receiver can achieve lower bit error rate and error vector amplitude(EVM),and the prediction speed is much lower than other receivers.The overall performance of NN receiver is the best,and it is more suitable for high-order modulation scheme.