节点文献

100Gbps PM-(D)QPSK相干光传输系统DSP算法研究

Research on DSP Algorithms of 100Gbps PM-(D)QPSK Optical Coherent Transmission System

【作者】 周娴

【导师】 陈雪;

【作者基本信息】 北京邮电大学 , 电磁场与微波技术, 2011, 博士

【摘要】 随着以高清视频、IPTV、P2P等为代表的高速率高质量视频、数据业务的日益流行,电信骨干网络的传输带宽需求迅猛增长,实现单波长信道100Gbps信号的高效传输,为超大容量DWDM光传输网的建设提供支撑技术,对国家信息基础设施的建设意义重大。100Gbps PM-(D)QPSK相干光传输系统结合了偏振复用技术、多进制调制技术、相干检测技术,具有较高的频谱效率、OSNR灵敏度,在满足当前50GHz DWDM波长间隔链路设计规则的同时,可以利用DSP算法灵活地补偿信号传输损伤,成为了100Gbps传输系统商业部署最具可行性的技术方案之一。本文对100Gbps PM-(D)QPSK相干光传输系统的电域DSP核心算法进行了深入研究,并取得了一定的创新性研究成果。本文的主要创新点和研究工作包括:1、针对目前相干光接收机中前馈式全数字时钟同步算法复杂度高、对ADC采样率要求高,而混合时钟同步环路非全数字实现、用于偏振复用系统存在同频不同相等问题,本文将Gardner提出的反馈式全数字时钟同步算法运用于112Gbps PM-(D) QPSK相干光传输系统中,并通过仿真分析了光传输系统中各种效应对算法同步性能的影响。为使该算法易于硬件实现,解决在对符号速率高达28Gsym/s的信号进行实时处理时所面临的电子速率瓶颈问题,本文在上述研究基础上,提出了一种基于并行处理的反馈式全数字时钟同步算法,仿真结果表明该算法可支持160个并行处理单元同时对320个异步采样值进行定时同步调整,使硬件时钟频率从串行处理时的56GHz下降到175MHz左右。2、针对Gardner定时误差检测算法的低色散容限极大限制112Gbps PM-(D)QPSK相干光传输系统实用性这一问题,本文提出了一种高色散容限的全数字时钟同步与自适应均衡解复用联合处理方案,通过仿真验证了该联合方案能够将系统残留色散容限从±200ps/nm增大到±800ps/nm以上,且在不增加任何额外计算复杂度情况下,同时完成时钟同步,CD、PMD补偿和偏振解复用。同样为克服电子速率瓶颈问题,本文在提出一种支持大规模并行处理自适应均衡解复用器的基础上,设计了基于并行处理的时钟同步与均衡联合处理方案。仿真结果表明该方案可以在处理性能基本保持不变的情况下,对硬件处理时钟频率要求下降到百MHz量级。3、为抑制残留频偏所造成的信号相位跳变,本文提出一种基于训练序列的频偏估计算法。该算法利用训练序列准确地去除了估偏符号的调制相位,提高了频偏估计精度,避免了传统频偏估计算法中的M次方运算,在降低算法复杂度的同时,使频偏估计范围与信号相位调制方式和调制级数无关,理论估偏范围达到[-RS/2;+RS/2] (RS为符号速率)。仿真结果表明,基于训练序列频偏估计算法对于112Gb/s PM-(D)QPSK信号实际频偏估计范围可达[-13GHz,+13GHz],而且训练符号开销仅为0.000011%以下。4、本文利用VPItransmissionMaker8.5和MATLAB软件搭建了112Gbps PM-(D)QPSK相干光传输系统仿真平台,并对传输系统整体性能进行了综合性仿真分析。在仿真验证DSP核心算法可行性的基础上,进行了112Gbps PM-(D)QPSK相干光传输系统的背靠背实验和1240km传输实验。在1240km长距离传输实验中,光纤链路上没有进行光域色散补偿,链路累计色散完全依靠电域DSP算法进行补偿,实现的色散补偿量超过20000ps/nm,且在信号入纤光功率小于+2dBm时,系统BER性能优于1E-3所需要的OSNR低于17dB。

【Abstract】 With the rising popularity of high-speed and high-quality videos, data traffics such as HDTV, IPTV and P2P, the bandwidth requirement for the backbone network has increased dramatically. The implementation of 100Gbps efficient transmission per wavelength channel can provide supporting technologies for the construction of super capacity DWDM transmission network and is significant to the construction of national information infrastructure. The 100 Gbps PM-(D)QPSK optical coherent transimission system combines the polarization multiplexing, multi-level modulation and coherent detection technologies, which not noly has the high OSNR sensitivity and frequency spectrum efficiency for the 50GHz wavelength interval of current DWDM link design, but also can employ digital signal processing (DSP) algorithms to compensate for transmission impairments. Therefore it is a suitable candidate for future high-speed optical transmission system. This paper mainly researches on the key DSP algorithms for 100Gbps PM-(D)QPSK optical coherent transmission system and gains a number of innovative research achievements. The main work and innovative contributions are listed as follows.1、In optical coherent receiver, the feed-forward digital timing recovery has the high computing complexity and requirement for ADC sampling rate, and the hybrid timing recovery loop can not compensate the sampling phase errors for two polarization signal simultaneously because of non-all-digital implementation. To solve these problems, we propose to use a Gardner back-feed digital timing recovery algorithm in the 112Gbps PM-(D)QPSK optical coherent receiver and analyze the influences of the various effects induced in optical transmission processing to the timing recovery algorithm’s performance. Furthermore, in order to make it easy to hardware implementation and break the limitation of electronic processing rate, we propose a parallel digital timing recovery algorithm based on the above research. Simulation results show that this new algorithm can support 160 parallel processing units to adjust time synchronization for 320 asynchronous samples at the same time. Therefore, the hardware clock frequency is reduced from 56GHz in serial processing to 175MHz in parallel.2、To solve the problem that the practicability of 112Gbps PM-(D)QPSK optical coherent transmission system is greatly limited by the low dispersion tolerance of Gardner timing error detecting algorithm, we propose a joint scheme that embeds an adaptive equalizer into the all-digital timing recovery loop. The simulation results demonstrate that the joint scheme can not only improve the tolerance of system toward residual dispersion form±200ps/nm to above±800ps/nm, but also accomplish synchronization and compensation linear transmission impairments simultaneously. Furthermore, in order to cope with the speed bottleneck of hardware, we propose an adaptive equalization scheme and design a parallel joint scheme of timing recovery and equalization. Finally, we demonstrate that the proposed parallel scheme allow the hardware to process 112G bit/s POLMUX-DQPSK signal at the hundreds MHz range.3、In order to reduce the probability of phase cycle slip caused by residual frequency offset, we propose a carrier frequency offset estimation algorithm based on training sequence. The new algorithm can use training symbol to remove the modulated phase information of estimated signal, which improves the estimation accuracy and avoids the Mth power operation of the classical algorithm. Therefore, the proposed algorithm reduces the computing complexity and has a constant theoretical estimated range for carrier frequency offset ([-RS/2,+RS/2], RS is symbol rate) which is independent of the signal phase modulation level and format. The simulation results demonstrate that its estimated frequency offset range for 112Gb/s PM-(D)QPSK signal can reach [-13GHz,+13GHz] and the training overhead can less than 0.000011%。4、We build a simulation platform of 112Gbps PM-(D)QPSK optical coherent transmission system by using VPItransmissionMaker 8.5and MATLAB softwares. Base on the feasibilities of the key DSP algorithms demonstrated by simulation, we carry out the back-to-back and 1240km transmission experiments of 112Gbps PM-(D)QPSK singal. In the 1240km long haul transmission experiment, the dispersion compensating fibers (DCFs) are not employed in transmission link. All the accumulated dispersions are compensated by DSP algorithms in electric field. The compensation quantity of CD is large than 20000ps/nm and BER performance of system can superior to 1E-3 at 17dB of OSNR when the launch power of signal is no more than +2dBm.

节点文献中: