Signal Processing Issues in FPGA Implementation of 3D Beamformer Using Adaptive Algorithms

Three dimensional beamforming (3DBF) is a famed technology that enhances spectral efficiency, capacity and coverage area of the 5G networks compared to conventional two dimensional (2D) beamforming by implementing both user specific horizontal and vertical beamforming. In this paper, three LMS based adaptive spatial filtering algorithms, namely, Ang's, Mathew's and Fixed step-size LMS have been implemented and evaluated on FPGA for 3DBF scenarios. The associated signal processing issues, specifically for FPGA implementation have been critically investigated. A prototype design of a 3DBF system deploying these adaptive algorithms has been demonstrated using Xilinx Zynq®-7000 family based FPGA platform. Coexistence of Xilinx System Generator
^™
and MATLAB Simulink® environment has been suitably utilized for the purpose. The performance metrics in terms of convergence properties like Mean Square Error (MSE) and Step-size along with beam pattern of all algorithms have been evaluated. Finally, hardware resource utilization of these algorithms has also been analyzed. Investigations reveal the superiority of Mathew's algorithm over the other two for 3DBF.