【作者】 王勇；

【导师】 潘炜；

【作者基本信息】 西南交通大学 ， 通信与信息系统， 2005， 硕士

【摘要】 拉曼光纤放大器(FRA)是一种光纤基的全波段放大器。由于具有可分布式放大、增益高、输出功率大、响应速度快、可降低非线性影响、噪声特性好等诸多优点,因而,随着DWDM长距离传输系统的大量应用,FRA的重要性日渐显露。 本文利用理论分析与实验结合的方式对拉曼光纤放大器的增益平坦特性进行了研究。首先,在简化功率耦合方程的基础上,引入遗传算法优化抽运光数目与抽运功率,同时实现了信号增益谱和噪声谱的平坦。计算结果表明:双向抽运方式下,系统噪声谱在C+L波段比反向抽运更为平坦;通过多次测试,对于10波抽运,100信道的DWDM系统(带宽为80nm),增益波动小于1dB的优化时间可控制在400s以内。与国内同类优化算法相比,优化时间及精度处于最优状态;其次,首次采用反向传播(BP)神经网络研究拉曼光纤放大器(FRA)中增益带宽、增益平坦度与抽运光数目配置的非线性依赖关系,利用其对于输入、输出参量已知,但解析关系未知的函数具有良好逼近能力的这一特点,构造了2-5-1型网络。同时得到,在增益平坦度确定的条件下,不同增益带宽处,抽运光数目必存在下限值。 最后,实验验证本文采用的平均功率法及遗传算法的正确性与稳定性;同时验证了对于同一组输入信号,在不同抽运光功率条件下,开关增益在抽运光功率较小时随着抽运光功率的线性增加而线性增加,即在此情况下拉曼光纤放大器不容易出现饱和,饱和增益较高;但当抽运光功率继续增大超过200mW时,开关增益成非线性增加,出现了饱和现象。对于两组输入光,在小信号低增益的情况下,拉曼光纤放大器的增益几乎与信号光的功率无关,只与抽运光的功率有关。更多还原

【Abstract】 Fiber Raman amplifier (FRA) is full communication windows amplifier based on the fiber. The importance of FRA appears gradually when long distance transmission system of DWDM adopted largely because of its excellent features of distributed amplification, high gain, high output power, quick response speed, small non-linear influence, high signal noise ratio, et al.The paper research features of gain ripple utilized theory analysis and experiment measure method. Firstly, after optimize the coupling differential equation of multi-wave pumped fiber Raman amplifier, the genetic algorithm is used to approach the flatness of signal gain and noise spectra simultaneously. The calculated results show that spectra of noise are flatter under the case of bidirectional pump than backward pump in C and L wave band. A fiber Raman amplifier pumped by 10-wavelength source is optimized and the gain ripple less than ldB in 100-channel DWDM system of 80nm gain spectra is obtained within 400s through many tests. The results of elapsed time of optimization and precision are optimized state in our country. Secondly, the nonlinear relations of gain bandwidth, gain flatness to number of pumping wave in broad-band fiber Raman amplifier are researched by using neural network technology for the first time. Utilized its peculiarity of good approaching ability to an unknown function with given input data and corresponding output data, a double layer back propagation neural network is formed. The simulation results show that under the condition of gain flatness to be determined the lowest bounds of number of pumping waves exists at different gain bandwidth.Finally, the experiment verified exactness and stability of average power analysis method and the genetic algorithm. Proved to the same group’s input signal under different pump power condition, on-off gain is increased linearly with accordance to pump power increased linearly when pump power is small. Namely fiber Raman amplifier is not apt to the saturation under this situation. The saturation gain is gained highly. But as pump power increased at exceeding 200mW, on-off gain increased non-linearly. The saturation phenomenon has appeared. For two更多还原