基于GPU的高动态信号快速捕获设计与实现

Design and Implementation of Rapid Acquisition for High-dynamic Signal Based on GPU

传统的软件接收机均是由CPU处理器实现,然而针对多普勒频率偏移较大的卫星信号捕获问题,基于FFT的并行频率空间搜索捕获方法可捕获较大的频偏并跟踪一定的频率变化率,但其计算重复次数较多,耗时过长。为此本文使用高性能计算平台,设计了一种适用于GPU的高动态扩频信号捕获的FFT并行频率空间搜索捕获方法,较好地平衡了其在高动态扩频信号捕获的优越性和高捕获耗时之间的矛盾,实现对高动态扩频信号的快速捕获。并充分利用GPU加速、高性能并行计算架构,测试结果表明:将FFT的并行频率空间搜索捕获按本文的设计方案在GPU平台实现后,捕获单颗卫星信号时间加速约96倍,捕获全部卫星时间加速46倍。

Traditional software receiver are implemented by a CPU processor.However,parallel FFT frequency space search acquisition method can capture a large frequency offset and search for a large frequency range for the large doppler frequency problem,but it needs large amount of computation,and takes too long time on CPU.And the traditional method takes too long time.So this paper using high-performance computing platform,designed a rapid acquisition for high-dynamic signal base on GPU by parallel FFT frequency space search acquisition method.It fully inherited the advantages and effectively shorten the time of capture.By taking full use of GPU acceleration,and high-performance parallel computing architecture,the test results showed that：the parallel FFT frequency space search acquisition on GPU platform acceleration time of about 96 times to capture a single satellite signal,and accelerated 46 times to through all satellites.