Research On Training Sequences Aided Extended Kalman Filter For Optical Signal Processing

The recent evolutions in information technology,such as 5G,Internet of Things,and big data,are promoting the growing demands on bandwidth and capacity of data transmission.According to the IEEE 802.3bs standard,the industry standard for single-carrier long-haul optical communication has been raised to 400 Gbit/s,which requires higher spectrum efficiency.The traditional quadrature phase shift key signal has no ability to meet these needs.One way to solve this problem is by adopting higher order modulation formats,such as 16 quadrature amplitude modulation and 64 quadrature amplitude modulation.However,higher order modulation formats demand more advanced digital signal processing algorithms.One way to improve the performance of these algorithms is by using the training sequences,this research is mainly focusing on improving the performances of Kalman filter algorithm by using training sequences.The convergence performance of the extended Kalman algorithm with single measurement equation depends on the modulation format,and the algorithm often fails to converge under high-order signals such as 64 quadrature amplitude modulation,introducing the ordinary training sequences can solve these problems.In this article,the performances of ordinary training sequences are verified by simulation and experiment.Accurate frame synchronization algorithm is the precondition of ordinary training sequences aided Kalman algorithm,therefore in this paper,the m-sequence is used for frame synchronization.The simulation results show that this scheme is efficient under different modulation formats.Then,the performances of algorithm stability,convergence speed,polarization tracking rate,laser linewidth tolerance are investigated under different format ratio.For 64 quadrature amplitude modulation signals,the stability of Kalman algorithm is greatly increased by using ordinary training sequences.In terms of convergence speed,this scheme can achieve fast convergence speed and be independent of the modulation format.Compared with the case without training sequences,the convergence speed is improved by nearly 50 times.In terms of polarization tracking rate ability and linewidth tolerance,the performances are associated with format ratio of ordinary training sequences,the higher the ratio,the better the performance.In view of the shortcomings of the ordinary training sequences that they cannot provide useful information to the Kalman algorithm,this paper explores the special training sequences,which adopting special structural designs in structure,in this way the special training sequences can calculate the inverse Jones matrix directly.On the one hand,it can achieve depolarization quickly,on the other hand,it can provide more information to the Kalman algorithm and improve the convergence speed of the Kalman algorithm.This part is mainly investigated by simulation.Similar to the investigation of ordinary training sequences,the special training sequences also need to locate accurate position of the training sequences.The simulation results show the accuracy of the frame synchronization of the m-sequence under special training sequences.Then,the efficacities of the depolarization under different polarization cross-talk angles are investigated,and the effects of training sequences length,optical signal-to-noise ratio and other factors on the depolarization accuracy are verified.It can be concluded that the depolarization can be achieved at different cross-talk angles.The depolarization accuracy will be higher with the longer length of the training sequences and higher signal-to-noise ratio.In terms of convergence speed,Kalman filter can achieve fast convergence within ten date points with the help of special training sequences.This paper summaries training sequences aided kalman filter under different modulation formats,and the corresponding parameter configuration suggestions are given in this paper.