MVDR Adaptive Beamforming Using oneAPI SYCL HLS Targeting Direct RF FPGA

The robust MVDR adaptive beamformer provides advanced spatial filtering capabilities by determining the weight vectors for beam steering dynamically. This direct method of adaptive weight computation based on sample covariance matrix of the noise field results in very rapid convergence. The proposed optimal MVDR beamformer is obtained by replacing the matrix inverse operation using QR decomposition (QRD). The oneAPI SYCL HLS flow provides significant productivity advantages to accelerate the MVDR beamforming application from SYCL to synthesizable code on the FPGA platform. This paper presents the design and implementation of the proposed QRD based MVDR adaptive beamformer running on the Intel FPGA Stratix® 10-AX FPGA platform with integrated data converters. The advantages of the proposed adaptive beamformer method over conventional technique are exemplified in a synthetically simulated scenario running on the FPGA platform. Other beamforming issues like channel synchronization, including latency and phase alignment, are also investigated.