Research On Nonlinear Phase Noise Monitoring Method In Optical Fiber Communication System

With the development of coherent detection and digital signal processing technology,high-speed optical communication system has become the mainstream scheme to achieve large capacity transmission,and plays a very important role in the field of modern communication.With the development of 5G technology and various emerging Internet applications,it is urgent to build a high-speed optical fiber communication network with long distance,large capacity,high flexibility,high reliability and low delay.For this purpose,it is necessary to establish a reliable real-time optical communication monitoring system.One of the most difficult problems in long-distance optical communication is the nonlinear phase noise accumulated by Kerr effect,which can cause signal distortion.Therefore,it is an important subject to monitor the nonlinear effect of damaged signal in the development of optical network.Based on the deep learning method,this paper focuses on the convolution neural network algorithm to achieve nonlinear phase noise monitoring in optical fiber communication system.By building the simulation coherent system of the actual optical fiber link,the constellation images are preprocessed and used as the inputs of the research algorithm of CNN model.Then,an expectation layer is added after the CNN model outputs the probability vector corresponding to the predicted value to realize the continuous and more accurate output of the predicted value.The main research work of this paper is as follows:(1)Starting from the discussion of the emerging Internet technology,this paper describes the importance of the research on nonlinear phase noise monitoring in optical fiber link system,briefly introduces research status of the current optical performance monitoring methods,and analyzes the feasibility of machine learning methods,especially deep learning,in realizing different parameter monitoring,as well as the advantages and disadvantages of existing neural network models.(2)In view of the shortcomings of traditional CNN model,a lightweight convolutional neural network,MobileNet,is introduced,and an effective monitoring model for nonlinear phase noise estimation is proposed.At the same time,based on the QPSK partition method,the prediction model is extended to the optical fiber link system with higher-order modulation format.(3)Through the simulation platform,the coherent communication link system is built to get the training data.Then we use Tensorflow to build a convolution neural network model.The results show that the average relative error of nonlinear phase estimation is less than 5.0%,2.0%and 1.0%respectively when the optical signal-to-noise ratio is greater than 18dB,22dB and 27dB.Simultaneous interpreting the influence of different transmission rates on CNN monitoring results,three different transmission rates were used(56 Gbit/s,70 Gbit/s and 112 Gbit/s).The test results showed that the nonlinear phase estimation error fluctuated with the theoretical value at different transmission rates,but the overall results were less than 2%.So under the tolerance of 2%,we can get the result that the algorithm is insensitive to the change of data rate.In order to evaluate the performance of the new method,this paper also compares it with the pilot assisted DA-ML algorithm.The results show that the proposed method is better than DA-ML algorithm in performance.It fully shows the effectiveness of applying CNN method to nonlinear phase estimation.