Research On Key Technologies Of Damage Estimation In Large Capacity Optical Fiber Communication System

With the explosive growth of the number of network users,the demand for communication capacity will show a rapid growth trend.Large-capacity optical communication systems are the trend of future development.The flexible and efficient system damage estimation can promptly reflect the faults and performance of the network to ensure the safety and stability of the optical communication system.At the same time,high performance,low-cost damage compensation algorithms can improve the transmission performance of the communication system,reduce communication system DSP resource consumption and power consumption.Therefore,this thesis mainly researches from the aspects of intelligent system damage estimation and link damage compensation,and the main research work and results are as follows:Aiming at the optical signal-to-noise ratio(OSNR)noise of optical fiber communication systems,a multi-input multi-output multi-task convolution based on migration learning is proposed in the large-capacity multi-core optical fiber space division multiplexing system.Neural network to realize OSNR estimation of space division multiplexing optical fiber transmission system.The performance and complexity of the traditional convolutional neural network and the proposed scheme are compared.The results show that the proposed multi-input multi-output multi-task convolutional neural network has higher OSNR estimation efficiency.Experimental results show that the root mean square error(RMSE)of OSNR is less than 0.6 d B.The transfer learning reduces the number of training samples by 48% and the training epochs by 50%,respectively.Aiming at the nonlinear noise of optical fiber communication system,a multi-task artificial neural network based on asynchronous complex histograms is proposed to realize modulation format recognition,OSNR and fiber nonlinear noise power estimation.Compared with the performance of the scheme based on the amplitude histogram,the average RMSE of OSNR based on the asynchronous complex histogram scheme can reach0.37 d B,which has better estimation performance.In terms of nonlinear noise estimation performance,the average RMSE can reach 0.25 d B.The results show that the estimation scheme has superiority in nonlinear noise estimation,compatible with higher-order modulation format signals,and the scheme can monitor more optical network parameters at the same time.Aimed at the carrier interference noise of the optical fiber communication system.In a high-spectrum efficiency frequency division multiplexing(SEFDM)system that reduces the data transmission bandwidth through bandwidth compression technology,a simplified Tomlinson-Harashima precoding(THP)algorithm is proposed,which can reduce the complexity of the user-side algorithm.We have realized the intensity modulation-direct detection 40 km 16 quadrature amplitude modulation(QAM)SEFDM optical system transmission based on the precoding scheme.The results show that the precoding scheme achieves good transmission performance under the premise of ensuring higher spectrum efficiency of the system.When the compression ratio is 0.8,compared with fixed sphere detection based on iteration detection,the algorithm complexity of THP is reduced by an order of magnitude,and the performance of the system is improved.Regarding the phase noise of the optical fiber communication system,the high-performance and high-order star QAM's ability to resist phase noise and the tolerance to the laser linewidth are analyzed.At first,in the case of low signal-to-noise ratio,the blind phase estimation algorithm will lead to system performance degradation.In this part,the phase jump damage problem existing in star 16 QAM is explained and explained in detail for the first time.On the basis of low signal-to-noise ratio and phase noise probability model analysis,single-differential and double-differential encoding schemes are used to alleviate the bit error problem caused by phase jump in blind phase estimation.In the large-capacity high-performance high-order star QAM optical system,the performance of the blind phase estimation based on single difference and the blind phase estimation scheme based on double difference is compared in detail.Including time-varying phase processing capability,transmission performance and laser line tolerance,etc.,and the establishment of system experiments to verify the performance of the proposed phase estimation scheme.