基于Systolic阵的IQRD-SMI算法的研究与FPGA优化实现

Research and FPGA Optimization Implementation of IQRD-SMI Algorithm Based on Systolic Array

摘要:

针对自适应波束形成系统中权值求解计算量大的问题,研究了基于LCMV的逆QR分解算法(IQRD-SMI算法) 及其传统脉动流水的GR-TSA实现结构,在Matlab中对算法以脉动阵搭建并进行算法仿真；分析了CORDIC技术的原理以及如何利用它在FPGA上实现Systolic阵中的处理单元；在此基础上,给出一种基于Systolic阵列的复用IQRD-SMI算法实现结构GR-ITSA,比较了它与已有文献中结构的异同；基于ISE软件,在Xilinx的xc7k325t型FPGA上实现了基于脉动阵列的IQRD-SMI算法的底层设计和硬件仿真,并将FPGA定点仿真数据回传Matlab与计算机浮点仿真数据对比,进而进行系统的性能与误差分析。

Abstract:

For the problem of the weight values need a large amount of calculations in the adaptive beam-forming system, studied the inverse QR decomposition algorithm based on LCMV (IQRD-SMI algorithm) and the structure of traditional Systolic array GR-TSA, built with systolic array structures and simulated in Matlab. The principle of CORDIC technology is analyzed and how to use it on FPGA implement systolic array processing unit. On the basis of this, Based on the reuse of systolic arrays IQRD-SMI algorithm structure is given, Compared it with existing in the literature. Underlying design based on IQRD-SMI algorithm and hardware simulation based on systolic array be realized with xc7k325t in Xilinx FPGA, And the FPGA fixed-point simulation data and the computer floating-point simulation data contrast. The system and error is analysed.