面向高速光通信系统的电域色散直接动态综合补偿方法研究

In optical communication systems with increasing transmission capacity, compensation for chromatic dispersion (CD) and polarization mode dispersion (PMD) is helpful to maintain the long distance transmission of optical signals, and enables the existing fiber cables can be reused in system expansion. For the special issue of electronic dispersion compensation of high-speed optical communication systems, based on our previous research works, this project focuses on studying and improving a direct dynamic comprehensive compensation method for CD and PMD using continuously distributed delays recurrent neural networks (CDD-RNN), including: (1) to construct the Lyapunov-Krasovskii function of the CDD-RNN with complex multi-level activation function, and obtain its stability criterion; (2) to verify the applicability of the CDD-RNN algorithm in the case of noisy weight matrix; (3) to design the memristive neuron circuit with accurately adjustment of those weights and stable memory, and verify its enhancement of the self-organizing and adaptive abilities for the CDD-RNN networks; (4) to solve the fixed rotation problem of the existing RNN electronic dispersion compensation method; (5) to design the method for generating initial state vector of the CDD-RNN to speed up the convergence. As a result, a new direct dynamic comprehensive electronic dispersion compensation method for CD and PMD of high speed optical communication systems will be developed. The method would not depend on the statistic information of received data, but could track instantly the fast time-varying characteristics of dispersion, suitable for multi-level modulations, detect signal directly without designing any filter, and complete signal recovery without using decision function.

在传输容量不断提升的光通信系统中，色度色散和偏振模色散的有效补偿是保持光信号长距离传输的关键，也有利于已铺设的光缆再利用。针对高速光通信系统电色散补偿的特定问题，研究连续分布时滞递归神经网络（CDD-RNN）框架下的色散直接动态综合补偿方法：（1）建立复数多值激励CDD-RNN的Lyapunov-Krasovskii泛函，获得其稳定性判据；（2）验证CDD-RNN色散补偿系统的权值矩阵“含噪”时的算法适用性；（3）设计具备权值精确调整、存储功能稳定的忆阻神经电路，验证其对网络自组织自适应的增强能力；（4）解决原RNN色散补偿方法所存在的输出信号星座固定旋转问题；（5）设计CDD-RNN色散补偿系统的初始状态向量生成方法，加速算法收敛。最终形成一种“不依赖输入信号统计信息、可跟踪色散快速时变、适用于多进制调制、无需设计滤波器、码元恢复和判决一体完成”的色散直接动态综合补偿方法。

在传输容量不断提升的光通信系统中，色度色散和偏振模色散的有效补偿是保持光信号长距离传输的关键，也有利于已铺设光缆的再利用。项目针对高速光通信系统调制方式复杂化、色散快速时变等特点，考虑到现有电色散补偿方法普遍存在“对数据量依赖严重、应对时变色散能力保守、色散补偿方式间接”等不足，研究并形成一种“可不依赖输入信号的统计量信息、能实时跟踪色散快速时变特性、适用于多进制调制、无需设计滤波器实现信号直接检测、码元恢复和判决一体完成”的高速光通信色散直接动态综合补偿方法。建立了多值激励函数条件下CDD-RNN的Lyapunov-Krasovskii泛函，获得其稳定性判据；验证CDD-RNN色散补偿系统的权值矩阵“含噪”时的算法适用性；解决了通过控制通路增益和多值激励函数陡峭度保障算法的正确收敛性；解决了Lyapunov-Krasovskii泛函基础上，结合多稳定分析理论和值空间信息来区分真伪平衡点问题。建立了一种基于相位激励反馈结构网络的高速光通信系统电色散补偿方法； 建立了基于概率密度函数拟合和模糊逻辑的相干光通信系统盲均衡方法；建立了适用于极化复用正交幅相调制相干光通信系统的双判决盲均衡方法；建立了正交频分复用60千兆赫毫米波光载无线电系统的均衡补偿方法。将研究成果应用到海岛岸基分布式电源系统，保障了了系统微妙级脉宽窄脉冲的无失真传送，解决了驱动信号远距离、准确、无中断传输的难题。