Research On Hardware Implementation Of Impairment Eoualization Algorithms Of Rotation Of State Of Polarization Based On Kalman Filter

With the fast development of cloud computing,online high definition video and mobile Internet,the demand for communication system bandwidth is also increasing.Coherent optical communication systems have become the main solutions for long distance communication due to their advantages of high transmission rate and large transmission capacity.Around 2010,single wavelength 100Gbit/s high-speed coherent optical communication systems were gradually commercialized in backbone networks.In high-speed coherent optical systems,the transmitter uses Quadrature Amplitude Modulation(QAM)and Polarization Division Multiplexing(PDM)to increase the transmission efficiency of a single wavelength.Coherent receivers rely on coherent detection and Digital Signal Processing(DSP)to demodulate signals.The former can perform phase and polarization diversity reception on the received optical signal,while linearly converting the received optical signal to the electrical domain.The latter can perform signal impairment equalization based on digital signals obtained by coherent detection,thereby achieving signal recovery,decision and decoding in the electrical domain.Optical signals in optical fibers inevitably suffered from all kinds of impairments.The signal impairments mainly include channel impairments such as loss,chromatic dispersion(CD),polarization mode dispersion(PMD),rotation of state of polarization(RSOP),polarization dependent loss(PDL),carrier frequency offset(CFO)and carrier phase noise(CPN).The last two were introduced by the laser.In face of such complex signal impairments,it is necessary to design an effective and practical DSP algorithm.In practice,DSP algorithms are integrated on application specific integrated circuit(ASIC)chips.The specially designed ASIC chip can meet the needs of high-speed optical communication to process optical signals in real time,and flexibly utilize hardware resources for layout and wiring.However,the simulation of DSP algorithms in most literatures is based on offline implementation.From offline implementation to hardware implementation,there are still many issues to consider,including feedback delay,fixed-point algorithms,parallelization algorithms,and limited hardware resources.The feedback delay mainly refers to that due to the calculation delay of the hardware unit,the feedback information at the current moment cannot be directly transmitted to the next moment,thereby generating a feedback delay.Using fixed-point numbers instead of floating-point numbers in algorithms,the coefficients will have a certain degree of rounding error.The parallelization of serial algorithms is equivalent to increasing the degree of impairments.Limited hardware resources limit the complexity of the algorithm.These problems limit the performance of DSP algorithms in hardware implementation.Therefore,it is important to analyze the performance of DSP algorithms from the perspective of hardware implementation.This paper focuses on the subject of analyzing the RSOP equalization algorithm from the perspective of hardware implementation,and studies the system structure,impairments model and equalization algorithm of coherent optical communication.Based on Matlab and FPGA platform,we make the performance comparison and result analysis between RSOP equalization algorithm based on Extended Kalman Filter(EKF)and CMA algorithm.The main research work and innovations are as follows:(1)The basic flow of signal modulation and demodulation in the coherent optical communication system,as well as the principle and structure of related optical devices are studied in detail.We study the physical mechanism and mathematical model of impairments of optical fiber communication system,as well as the principles and processes of the corresponding equalization algorithm.(2)The basic principle of the RSOP equalization algorithm based on EKF is studied.Based on Matlab,the pseudo code is used to compare the performance of the EKF algorithm and the CMA algorithm from the perspective of hardware implementation.It mainly includes three aspects:serial floating-point algorithm,fixed-point algorithm,and parallelized algorithm.The results show that the compensation accuracy of the EKF algorithm is the same as that of the CMA algorithm.But the compensation range of the RSOP is far beyond that of the CMA algorithm.The performance advantage is more obvious especially in the parallel structure.(3)Based on the FPGA hardware simulation platform,the above simulation results are further verified.With 128 channels in parallel,the RSOP equalization range of the EKF algorithm reaches 0.4Mrad/s.At this time,the BER after the CMA algorithm equalization exceeds the 7%FEC threshold requirement.And the complexity of the algorithm can be analyzed according to the consumption of hardware resources.The EKF algorithm slightly exceeds the CMA algorithm in LUT consumption,but is far lower than the CMA algorithm in DSP48 and FF resource consumption.It can be concluded that the hardware resource utilization of the EKF algorithm is less and its algorithm complexity is lower.