前视声纳中多波束形成算法的FPGA实现

FPGA implementation of multibeam forming algorithm in forward looking sonar

在阵元数量和波束数量较多的情况下,多波束形成算法对FPGA的计算资源以及数据传输能力要求极高。文中使用Xilinx公司的Kintex-7系列FPGA对96通道、256波束的二维前视声纳中正交多波束形成算法进行研究。针对正交多波束形成算法资源消耗过大的问题,整体算法采用12个子模块并行逐级求和的方法实现,该方法可对算法结构进行显著优化,每个子模块结构包括数字正交混频、低通滤波抽取和多波束形成三个部分。子模块工作的最高频率为128 MHz,故提出通过8通道复用的方法实现数字正交混频和滤波抽取,通过乒乓存储操作以及2048倍复用复数乘法器实现多波束形成;再将三个算法子单元串行处理,实现数据不间断输出,从而解决传统全并行多波束形成算法导致FPGA资源消耗严重以及不足的问题。最后,通过Matlab仿真和标准信号源测试验证所提方案的可行性。

In the case of a large number of array elements and beams,the multi beam forming algorithm requires extremely high computational resources and data transmission capabilities of FPGA.The Xilinx′s Kinex-7 series FPGA is used to study the orthogonal multi beam forming algorithm in 96 channel and 256 beam 2D forward looking sonar.In allusion to the problem of excessive resource consumption of orthogonal multibeam forming algorithm,the overall algorithm is implemented by means of parallel and step-by-step summation of 12 sub-modules,which can significantly optimize the algorithm structure.Each sub-module structure includes three parts:digital orthogonal demodulation,low-pass filter decimation and multibeam forming.As the maximum frequency of the submodule operation is 128 MHz,an 8-channel multiplexing method is proposed to achieve digital orthogonal demodulation and filtering extraction.Multibeam forming is realized by means of ping-pong storage operation and 2048 fold multiplexed complex multiplier.The serial processing of three computational subunits is conducted to achieve uninterrupted data output,thereby solving the problem of severe and insufficient FPGA resource consumption caused by traditional fully parallel multibeam forming algorithms.The feasibility of the proposed scheme is verified by means of Matlab simulation and standard signal source testing.