中子脉冲序列核信号相关和功率谱实时计算

Real-time correlation and power spectrum calculation for neutron pulse series

为解决基于PC机平台、高达1 GHz采样率下之相关计算和频谱分析对实时性要求的关键技术问题,针对中子脉冲序列核信号本身所具有的特殊的"0,1"结构特点,采用快速移动的方法,借助于内存管理及SSE优化设计,创建了优化频谱分析的流程,构造了高速、实时的相关计算和功率谱分析算法,实现了1 GHz采样率下的中子脉冲序列核信号的实时相关计算和频谱分析。性能测试结果表明,在计数率为3×106s-1,单个块(长度为1 024)时,研究的相关算法的计算时间为0.29μs,相应的英特尔数学内核库的相关计算时间为129.95μs。现场实际试验表明,该算法达到了对中子脉冲序列核信号进行相关计算和频谱分析的实时性要求。

The real-time algorithm of correlation calculation and frequency spectrum analysis is presented for neutron pulse series up to 1 GHz sampling rates based on PC. The fast moving method is proposed based on the special ＂0＂ and ＂1＂ structure of the series and spectrum analysis process is optimized. Memory management and streaming SIMD（single-instruction multiple-data） extensions are adopted in the algorithm for optimizing. Simulation demonstrates that the presented correlation algorithm consumes 0.29 μs when data block length is 1 024 points and count rate is 3 ×10^6 s^-1 , while the correlation algorithm of Intel Math Kernel Library consumes 129.95μs, Practical experiment shows that the algorithm meets the demand of real-time correlation calculation and frequency spectrum analysis.