Realization And Optimization Of Long-reach Passive Optical Network Based On Power Multiplexing

In recent years,with the continuous improvement of social information,a variety of new high bandwidth services emerge rapidly,such as 3D games,virtual reality,online education and so on,which put forward higher requirements for the bandwidth of the access network system.At the same time,the access network is also expected to cover a wider range and support larger capacity.Next generation passive optical network(NG-PON2)is well in the way to provide 10 Gb/s bandwidth for each user,which promotes the research and development of long-reach PON.Compared with the coherent detection receiver,the direct detection receiver,with the advantages of low cost and simple structure,is still the major scheme in long-reach PON.In order to expand the transmission distance and improve the system capacity in PON,it's mainly considered from the following two points:one is to improve the spectrum utilization and power budget to accommodate more users at low cost;the other is to improve the transmission performance and receiver sensitivity to resist the damage caused by long-distance transmission.To achieve the above goals,a power domain non-orthogonal multiplexing scheme is proposed for long-reach PON in this paper.It has the advantages of simple implementation and good compatibility and scalability,and can be combined with any coding/multiplexing technology.Therefore,each user can share the same block of resources.In terms of system scheme optimization,the influence of power allocation ratio and transmit power on BER performance and fairness index is discussed.In terms of transmission performance,neural network is dedicated to replace the traditional SIC receiver,which can eliminate the error propagation effect and enhance the sensitivity of the receiver.In this paper,VPI,MATLAB and Tensor Flow are used to verify and study the above scheme,and offline experiments are applied to verifying the concept of PD-NOMA-PON.The results are as follows:1.Optimizing PDR can adjust user's access performance,and when the transmission power is controlled within the range of weak nonlinear effect of the system,it's effective to increase the transmission power for improving the user's power budget.2.The receiver scheme based on neural networks can effectively avoid the influence of error propagation effect.Compared with the traditional successful interference cancellation(SIC)demodulation algorithm,when=10^-3 and each user with 4QAM modulation,the sensitivity of Convolutional neural network(CNN)receiver for user 1 with transmission distance of 60 km is improved by 0.6 d B@PDR=10,0.5 d B@PDR=8 and 0.6d B@PDR=6,and the sensitivity of CNN(Convolutional neural network)receiver for user2 with transmission distance of 20 km is improved by 1.7 dB @PDR=10,1.8 dB @PDR=8and 0.6 dB @PDR=6.