Fast and Hardware-Efficient Variable Step Size Adaptive Beamformer

Constant step size least mean square (CSS-LMS) is one of the most popular adaptive beamforming algorithms. However, for varying channel signal-to-noise ratios (SNRs), the CSS algorithms are not effective, and there is a need for variable step size (VSS) algorithms. The VSS algorithms provide extremely deep nulls for the interferences; however, they are complex to implement on hardware. Hence, this paper proposes two hardware efficient variable step size algorithms namely, efficient variable step size LMS (EVSS-LMS) and reduced complexity parallel LMS (EVSS-RC-pLMS). The proposed EVSS algorithms eliminate the complex operations of VSS algorithms like division and exponential and approximates them to simpler operations. Further, MATLAB simulations demonstrate accelerated convergence, deep nulls, lower error floor, and better performance in varying SNR environments for the proposed algorithms. Additionally, the finite precision radiation patterns are similar to infinite precision. Hardware synthesis results show the outstanding performance of EVSS in terms of resource utilization on the Xilinx Artix-7 FPGA.