复合重要抽样理论及其在一阶PMD模拟器中的应用

Multiple Important Sampling and Its Application in First Order Polarization Mode Dispersion Emulator

复合重要抽样可以减小计算方差,提高计算效率。介绍了复合重要抽样的基本理论,并将该理论应用于一阶偏振模色散模拟器仿真中。仿真结果证明,采用复合重要抽样技术,仿真次数为10万次时,可获得概率为1×10-13的极小概率事件,对应的差分群时延为其均值的4倍多,可以达到可能的最大值。与传统的Monte-Carlo仿真方法相比,大大提高了仿真效率。

Calculation variance can be reduced by multiple important sampling technique, and then calculation efficiency can be improved. Multiple important sampling theory and its application in the first order polarization mode dispersion （PMD） emulator have been introduced. The simulation result shows that low probability（1×10^-13） events whose different group delay （DGD） is more than four times of its mean value can be obtained and the DGD can hit its possible maximum value with only 1.0×10^5 samples by multiple important sampling technique. It proves to be a much more efficient method than the traditional Monte-Carlo one.