Research On Spectral Analysis And Peak Search Of Continuous Wave Radar Based On GPU Hard Ware

Continuous-Wave(CW) radar because of its compact structure, speed without blur, high accuracy and easy-to-digital implementation, etc. has been widely used in the measurement of the shooting range.With the rising requirements of high precision parameters measurement of the target, data quantity of the signal analyzer used in analysing increases quickly. The way of processing only by CPU cannot adapt to the requirement of the shooting range.For the requirement of shooting range, this thesis describes a way of using CPU + GPU heterogeneous model to carry out the signal analyzer of CW radar. The spectrum analysis and searching spectral-peak are calculated by GPU. It provides a good way to solve the problem, makes man-machine conversation more convenient, also reduces the production of high-precision measurement data delay.The main research contents include:The Stockham formulations of the FFT based on GPU parallelize transplant is discussed in this thesis. Performance of FFT algorithms can depend heavily on the design of the memory subsystem and how well it is exploited. Cooley-Tukey can be quite expensive due to incoherent memory accesses. In this thesis, we avoid the index-shuffling stage using Stockham formulations of the FFT.Two way including "step by step" and matrix transpose to carry out Stockham formulations of the FFT is discussed in this thesis. The first way is to push through each Stage of the wing-operation mapping of several hundred threads in order to make use of the strong parallel computing power of the GPU. As it has to access global memory every time and cannot make use of the performance of high storage bandwidth. The second way exploits shared memory by decomposing large FFTs into a sequence of smaller ones.The way of peak Search based on GPU is discussed in this thesis. Peak-Search can be considered as maximum search. We make optimization of parallel programming by thread quantity, sharing memory access optimization, launched follow-up circulation in the algorithm on the GPU using parallel reduction algorithm.The structure of the signal analyzer and software used in the system are improved. As the CPU and GPU have different advantages, CPU is used to compute when the quantity of the date is small, otherwise we use GPU. In this way, the executing time reduces.