| In recent years,with the continuous development of communication technology,emerging Internet technologies such as 4k/8k video,augmented/virtual reality,and cloud computing have emerged,and the global IP data traffic has shown an explosive growth trend.At present,for scenarios such as 5G fronthaul network,data center optical interconnection,and optical fiber access network,optical fiber transmission systems are developing in the direction of large capacity,wide coverage,and low cost and Intensity Modulation/Direct Detection（IM/DD）optical fiber transmission system has been widely used due to its simple transmission technology and low cost.However,with the continuous increase of system transmission rate,serious nonlinear distortion will occur after the signal passes through the IM/DD system.Therefore,researching advanced digital signal processing technology to suppress signal damage is of great significance for the improvement of IM/DD system capacity.On the basis of studying the theory of signal damage suppression methods in IM/DD optical fiber transmission system,this thesis focuses on the bidirectional Volterra signal filtering method based on decision kernel,the simplified nonlinear maximum likelihood sequence estimation（MLSE）signal receiving method based on decision threshold and the deep neural network decision receiving method based on feature construction.The main contents and innovate achievements of this thesis are as follows:（1）In order to improve the ability of Volterra signal filtering method to suppress signal impairment,a bidirectional Volterra signal filter method based on decision kernel is proposed.In this method,forward symbols and backward symbols are both considered at the input of each order in the Volterra series,and a decision kernel item is added at the back,which can improve its performance without increasing its computational complexity.The performance of the method used in single-carrier 50 Gbit/s PAM-4 IM/DD optical fiber transmission system is experimentally studied.The experimental results show that the proposed method can effectively suppress various signal impairments.Compared with the Volterra signal filter,a link power gain of about 1.3 dB can be obtained at the bit error rate of 3.8×10-3 which is the threshold of hard decision-forward error correction（HD-FEC）with the help of the proposed method.（2）In order to solve the problem that the calculation complexity of MLSE based on PAM-4 is too high,a simplified nonlinear MLSE signal receiving method based on decision threshold is proposed.The method adopts the cascade structure of pre-filter,post-filter and MLSE.By setting the decision threshold for the output symbol of the pre-filter,the constellation points within the decision threshold range use direct hard decision,and the constellation points outside the decision range use direct hard decision.The joint decision receiving method of MLSE reception is adopted,thereby effectively reducing the computational complexity of MLSE.The performance of the method used in single-carrier 54 Gbit/s Faster-Than-Nyquist（FTN）PAM-4 optical fiber transmission system is studied by simulation.The simulation results show that the proposed method can flexibly adjust the decision threshold according to actual needs,thereby effectively reducing the computational complexity of MLSE.When the received optical power is-24 dBm,compared with MLSE,the proposed simplified method satisfies that the system bit error rate is lower than the HD-FEC threshold,and the computational complexity is reduced by 19.7%.（3）In order to solve the problem of insufficient input information features in the traditional neural network decision receiving method,a deep neural network decision receiving method based on feature construction is proposed.This method uses the second-order Volterra series to construct features of the input information of the neural network,which not only improves the performance of the neural network but also accelerates the training convergence process.The performance of the method used in the single-carrier PAM-8 IM/DD optical fiber transmission system is studied by simulation.The simulation results show that the proposed method can effectively speed up the convergence speed of the neural network training process and has better decision receiving performance.Compared with the traditional neural network decision receiving method,the proposed method can achieve a link power gain of about 0.3 dB at bit error rate is 1×10-2,and a link power gain of about 1.1 dB at the HD-FEC threshold. |
PDF Full Text Request