分组并行的交叠式Allan方差快速算法

The overlapping fast Allan variance algorithm with grouping parallel

针对标准式Allan方差计算效率低、稳定性差的问题,该文构建了一种局部交叠的快速Allan方差算法。该算法采用分组并行运算的方法,通过局部的交叠式处理,得到指数变步长的数据簇。然后利用标准Allan方差公式获得对应数据簇的Allan方差值,最后绘制双对数曲线分析误差类型与大小。仿真与实测结果表明:与标准式Allan方差相比较,该算法不仅能够更精确地分离和计算噪声系数,而且大幅度提高了解算效率,长度为720000个样本点的数据解算时长从26.71 min降低到0.35 s。

To deal with low calculation efficiency and unstable numerical value of standard Allan variance,a fast algorithm for partial overlap Allan variance was constructed in this paper.By adopting the method of grouping parallel operation,this algorithm obtained data clusters of variable step-size via partial overlapping.Then,standard Allan variance formula was used to reach the Allan variance value corresponding to data clusters,and a double logarithm curve was drawn for the analysis of deviation type and size.According to simulation and actual results,compared with standard Allan variance,this algorithm could not only separate and calculate noise figure more accurately,but also improve calculation efficiency greatly.As the result,in terms of the data with 720,000 sample points,the calculation time was reduced from 26.71 min to 0.35 s.