Realization and Optimization of Inverse Algorithm of QR Decomposition of Arbitrary Order Signal Autocorrelation Matrix Based on HLS

The computation core of all kinds of adaptive beamforming algorithms lies in the inverse operation of signal autocorrelation function matrix. However, the design of all kinds of matrix inverse modules in FPGA development platform has some problems, such as difficulty in timing design, weak scalability and poor portability. This paper proposes a matrix inverse module of autocorrelation function of arbitrary order signal based on HLS (High Level Synthesis), completes the QR decomposition algorithm based on Housholder method, and encapsulates it into a large complex matrix inverse IP that meets the requirements of highspeed calculations, low resource consumption, and customizable configuration order. Finally, the synthesis results of RTL are analyzed from the aspects of timing delay, resource consumption, program configurability, etc. According to the synthesis results, the IP can improve the module speed to the magnitude of 2ms under the condition of 64 order complex matrix, and realize the advantages of configurable order, high data precision. It can well meet the high demand of real time in practical engineering and can be flexibly configured in various types of FPGA development platforms.