Research On Intelligent Nonlinearity Equalization Technology In High-speed Coherent Optical Fiber Communication System

Optical fibers currently underpin the global telecommunication infrastructures and services,which carry over 95% of the digital data traffic.Kerr fiber nonlinearity is one of the most important factors to constrain the spectrum efficiency and the information capacity of the next-generation long-haul high-speed coherent optical fiber communication networks.Aiming at this problem,this thesis is focused on the Kerr nonlinear effects and their compensation in long-haul high-speed coherent optical fiber communication systems.The intra-channel nonlinear equalization network based on the first-order perturbation theory has been developed in the single-channel dualpolarization 16-level quadrature amplitude modulation(DP-16QAM)transmission system,which can compensate for the intra-channel nonlinearity effectively.Besides,the inter-channel nonlinearity has been analysed and compensated in DP superchannel optical transmission system.The main works in this report are described as follows:1.The expression characterizing the intra-channel nonlinearity in the polarization multiplexed coherent optical fiber communication system has been derived using the first-order nonlinear perturbation theory based on the Manakov equation.The predistortion scheme at the transmitter and the symmetric dispersion compensation scheme for nonlinearity equalization have been investigated.The computational complexities of considered algorithms have also been analyzed.2.A single-channel DP-16 QAM optical fiber transmission system has been developed,and a perturbation-based neural network(P-NN)arising from the perturbation theory of nonlinear distortions has been designed.The operation and the analysis of the P-NN nonlinearity equalization are elaborated.It is found that the P-NN equalizer can obtain peak signal-to-noise ratio(SNR)gains of 1.37 d B and 0.86 d B compared to the electronic dispersion compensation(EDC)and the single step per span(St PS)digital back propagation(DBP)algorithm,respectively.The P-NN has a lower computational complexity than the DBP method.The developed P-NN can also be applied in the optical communication systems with different modulation formats.3.The influence of equalization enhanced phase noise(EEPN)on digital nonlinearity compensation in Nyquist-spaced superchannel transmission systems,with a numerically simulated 820 km optical field link,is also comprehensively studied.It is found that the peak SNR is decreased by 5.11 d B in the nonlinear compensation using the full-bandwidth DBP algorithm due to the EEPN noise when the 3-d B laser linewidth is 500 k Hz.The influence is also aggravated as the laser linewidth increases.The dualpolarization Nyquist spaced superchannel optical fiber communication system and the deep neural network equalizer for the compensation of interchannel nonlinearity have been implemented.It is found that the neural network equalizer can obtain a SNR gain of 0.76 d B under 2000 km transmission distance.Correspondingly,fiber nonlinearity can be alleviated over the entire transmission span.Finally,the robustness of the interchannel nonlinearity equalization neural network to EEPN noise has also been investigated.