High-Performance FPGA-Based QR Decomposition

This paper discusses two high-performance implementations of QR decomposition suitable for demanding adaptive beamforming radar applications, where both processing throughput and latency are important. In a beamforming application, the radar typically collects a small set of samples in training mode to build an internal mathematical model of the environment. The adaptive weights are then computed from the model using the QR decomposition process in a split-second time frame. Then, in radar processing mode, the weights are applied to realtime data streaming through at very high rate.