| With the booming development of communication technology,some services such as cloud computing,ultra-high-definition streaming video,Internet of Things,social networking and mobile connectivity have gradually emerged,making the number of users,network bandwidth and node capacity has been in explosive growth.In order to meet the needs of the current information technology industry,the transmission capacity of optical fiber communication system needs to be greatly improved.It is reported that the maximum single fiber capacity can be realized 50T at present.According to Huawei’s latest forecast,the single-fiber capacity of the backbone network is expected to reach above 100T in 2030.Multidimensional multiplexing and higher-order coding modulation techniques are effective in increasing transmission capacity,however there is still a gap achieved with these techniques between the channel capacity and the Shannon limit.Constellation shaping can narrow the gap with Shannon limit,so as to further improve the transmission capacity of communication system,which has become a hot issue in the field of optical fiber communication.In this dissertation,based on an in-depth study of the generation of constellation shaping optical signals for high-speed fiber optic communication,the research focuses on generating 16QAM optical signal based on geometric-probabilistic hybrid shaping in order to solve the problem of how to achieve probabilistic shaping of asymmetric constellations and improve the gain of hybrid shaping.In addition,the research focuses on K-means cluster algorithm applied for geometric shaping based on iterative polar modulation in order to solve the problem of how to resist fiber linear and nonlinear effects using machine learning algorithms.Furthermore,the research focuses on the power optimization algorithm of probabilistic shaping optical signal based on bit level distribution matcher in order to solve the problem of how to improve the throughput of probabilistic shaping systems and optimize the power of signals.The main research work and innovations of this dissertation are as follows.1.A geometric-probabilistic hybrid shaping optical signal generation scheme based on probabilistic subset shaping algorithm is proposed.A geometric-probabilistic hybrid shaping 16QAM optical signal generation scheme based on probabilistic subset shaping algorithm is proposed for the problem of probabilistic shaping applied on asymmetric structure constellation and improving the hybrid shaping gain.The proposed scheme first generates a training sequence based on a predefined optimal source distribution,and then acquires a GS-16QAM constellation after iterations.Secondly,the probabilistic subset shaping algorithm is applied to GS-16QAM to realize probabilistic shaping according to the principle of "selecting the subset first and then determining the signal points in the subset".The BER performance and generalized mutual information performance of the proposed scheme are studied experimentally.The experimental results show that the proposed scheme improves the OSNR gain by 2.4 dB,2.6 dB and 3 dB in terms of BER compared to CAP-GS-16QAM,CAP-PAS-16QAM and CAPSquare-16QAM schemes,respectively.Moreover,the CAP-GPS-16QAM scheme has a high GMI over the entire OSNR range,indicating that the proposed scheme can transmit a larger capacity.2.A geometrically shaped K-means algorithm based on iterative polarization modulation is proposed.K-means cluster algorithm applied for geometric shaping based on iterative polar modulation is proposed for the problem of the adaptation of machine learning algorithms to the constellation shaping optical signals.The proposed scheme uses an iterative quantization process to generate the geometrically shaped optical signal.And a split-step Fourier algorithm is used to simulate the propagation of geometrically shaped optical signals in optical fibers,and a K-means algorithm is used to complete the reconstruction of the signals at the receiving end.Simulation results show that the geometric shaping scheme based on iterative polarization modulation can effectively improve the transmission capacity of the communication system in both long/short distance optical transmission systems.Under the hard decision BER threshold,the SNR gain of the IPM signal receiving scheme using the K-means algorithm is 0.9 dB and 1.7 dB respectively compared with the receiving scheme of IPM signal and QAM signal using BPS algorithm when the modulation order is 16;the SNR gain of the proposed scheme is 0.3 dB and 1.1 dB respectively compared with the other two schemes when the modulation order is 64;the SNR gain of the proposed scheme is 0.4 dB and 0.8 dB respectively compared with the other two schemes when the modulation order is 256.3.A power optimization algorithm of probabilistic shaping signal based on bit level distribution matcher is proposed.A power optimization algorithm of probabilistic shaping signal based on bit level distribution matcher is proposed for the problem of how to improve the throughput and optimize the power performance of probabilistic shaping systems.The proposed scheme has a shorter output codeword length and higher throughput than the conventional probabilistic amplitude shaping based on symbol-level distribution matcher for higher-order QAM modulation.In order to optimize the power performance of the scheme,a bit allocation power optimization algorithm implemented by interleavers is proposed.The algorithm can effectively reduce the transmit power of probabilistic shaped optical signals and improve the robustness of optical signals to fiber nonlinear noise.The simulation results show that the SNR gains of the proposed scheme compared with the PAS-256QAM scheme are 0.79 dB,0.82 dB and 0.6 dB respectively when the entropy is 5.4 and the rate of LDPC is 3/4,5/6 and 8/9.Under the same conditions,the SNR gains obtained by the proposed scheme are 0.3 dB,0.27 dB and 0.22 dB respectively when the entropy is 6.At the same time,the proposed scheme can also obtain higher achievable information rate. |
