Performance analysis of implementation of radar signal processing algorithms on FPGA and GPU

Many civil and military application which requires real time and parallel processing is implemented on FPGA. FPGAs are used to process and transfer high speed real time data in radar applications. Besides, it is a big advantage to program FPGA again and again while it is on hardware. Users started to use GPUs in data parallel applications besides doing graphical processing. Many application are accelerated by running on GPUs and GPUs give better performance results than CPUs in applications that requires parallel processing. In this thesis, radar signal processing algorithms; digital down conversion, pulse compression, doppler processing and digital beam forming are implemented on both FPGA and GPU. And the results are compared. These algorithms are implemented on Xilinx System Generator for FPGA and OpenCL for GPU. In the designed simulation model, the direction of arrival, range and the doppler frequency of the simulated target can be detected. The results show that while processing real time data, FPGA has advantage over GPU. On the other hand, in the design used for detecting the direction of arrival, increasing number of antenna and look angle cause to increase the number of multiplication unit. Thus the total design cannot be fit on a single FPGA.