Research On Demultiplexing Technology Of Few Mode Fiber Space Division Multiplexing System

With the commercialization of 5G technology,various high-bandwidth and low-latency application scenarios such as autonomous driving,cloud computing,and high-definition video live broadcasts continue to emerge,and the demand for network bandwidth has shown explosive growth.In order to face the upcoming "bandwidth crisis",new technologies that can expand the capacity of optical fiber communication networks are urgently needed.At present,time,wavelength(frequency),phase,amplitude and polarization dimension multiplexing technology has gradually approached the capacity limit of single-mode fiber.Inspired by multiple-input multiple-output(MIMO)multi-stream parallel transmission,mode division multiplexing(MDM)technology based on few-mode fiber(FMF)has recently been proposed,and the use of different orthogonal modes to independently transmit multiple signals greatly improves fiber transmission capacity has become a potential candidate for the next-generation long-distance optical fiber transmission system,which has attracted widespread attention from scholars at home and abroad.The non-ideal components and transmission environment of the MDM system lead to the destruction of orthogonality between different modes,resulting in interferences such as mode coupling(MC),differential mode group delay(DMGD),and mode dependent loss(MDL).Among them,MDL directly affects the system capacity,and even causes communication interruption in severe cases,so compared to MC and DMGD,it has a more serious impact on the performance of the MDM system.In this paper,by studying the related theory of the pattern,building an FMF-based MDM simulation system,starting from the MIMO signal processing,researching the technology that can effectively suppress the MDL in the MDM system.Specifically,the main research work is summarized as follows:First,a simulation model of a few-mode fiber mode division multiplexing system is built.Considering the classic MIMO detection algorithms,the zero-forcing algorithm(ZF)and the minimum mean square error algorithm(MMSE)have poor performance,while the maximum likelihood detection(ML)has excellent performance but the exponential complexity is unbearable,a new MDM system detection algorithm based on machine learning is proposed.Different from the classic MIMO detection algorithm,it is an end-toend network model.The process is mainly divided into offline training and online deployment.The available network model is obtained through offline training,and then the signal detection and recovery can be realized by online deployment.Four different channel models(strong coupling without scrambler,weak coupling without scrambler,strong coupling with scrambler,and weak coupling with scrambler channel models)are used for simulation to verify the effectiveness of the proposed algorithm.The simulation results show that the performance of the MDM detection algorithm based on machine learning is close to the optimal ML detection.Second,starting from the Deep Neural Network(DNN),a DNN-based MDM system integrated signal detection algorithm is proposed,which integrates channel estimation and signal detection to form a deep neural network.At the receiving end of the MDM system,the received signal is directly input into the DNN model for detection,without the need for tedious channel estimation,signal detection,and decoding processes,and the decoded transmission signal is directly obtained.The DNN algorithm optimizes the signal processing process at the receiving end from the system structure,so compared to the classic MDM system detection algorithm,the DNN-based integrated detection algorithm has better performance.The simulation results show that whether or not TAST coding is used,the DNN-based MDM system detection algorithm has achieved excellent detection performance in different channel environments.