Design And Implementation Of High Frequency Resolution Solar Radio Dynamic Spectrum Analyzer Based On GPU

Observation of the dynamic spectrum of solar radio has important scientific significance in space physics,and it is helpful to predict the space weather environment of the earth.Solar radio dynamic spectrum analyzer is an important tool for receiving solar radio bursts.It can provide data accumulation for radio basic research.Constructing a high frequency resolution solar radio dynamic spectrum analyzer can observe the fine structure of solar radio burst frequency resolution.However,there are few solar radio dynamic spectrum analyzers with high frequency resolution at home and abroad.Most of the currently popular FPGA-based spectrum analyzers pay attention to the improvement of time resolution,and once this spectrum analyzer is completed,the frequency resolution cannot be changed.The GPU has parallel and high-density computing capabilities.It can efficiently use large points and multiple batches of fast Fourier transform operations to achieve high frequency resolution.Its low cost,easy development,and high performance have great advantages.Therefore,this paper combined with the project requirements to achieve a GPU-based high-frequency-resolution solar radio dynamic spectrum analyzer.Through a large amount of literature study and research on the existing spectrum analyzer of the observation station,a high frequency resolution solar radio dynamic spectrum analyzer was designed and realized by CPU+GPU architecture.The spectrum analyzer system is mainly composed of a high-speed data acquisition part,a high-frequency resolution spectrum realization part,and a dynamic spectrum display part.The physical connection between the two hosts in the spectrum analyzer system was realized using 10 Gigabit fiber,the network address mapping realized the virtual connection between the two hosts,and the virtual hard disk reduced the transmission time by means of packet buffering.The GPU is developed by Compute Unified Device Architecture(CUDA)programming to achieve parallel accelerated processing of data.The system's performance indicators are mainly reflected in high frequency resolution.For the reception and processing of solar radio signals in the frequency range of 150 MHz to 500 MHz,a frequency resolution of 2 kHz can be achievedThe main work completed in this paper includes:designing the overall structure of the spectrum analyzer,and analyzing and calculating the important parameters of the spectrum analyzer such as frequency resolution.The high-speed data acquisition card's dynamic link library function interface was used to program the board to achieve data collection and storage.On VS2012,CUDA7.5 platform,a data processing program is designed to complete the spectrum analysis work such as fast Fourier transform and power spectral density.Among them,the CPU and GPU computing tasks are reasonably allocated to achieve asynchronous parallel execution of the CPU and GPU;the CUFFT library is used to implement fast Fourier transform of time-domain signal data;and the overall execution of software is implemented using CUDA streams to achieve task-level parallel optimization;An improved parallel reduction algorithm is applied to evaluation operations such as power and accumulation.A high frequency resolution dynamic spectrum display program was written using MATLAB's powerful visualization operation capabilities,and real-time monitoring and offline observation of dynamic spectrum were realized.The existing conditions were used to complete the test of the spectrum analyzer in a laboratory environment,and the reliability,advancedness,and real-time performance of the spectrum analyzer system were verified through a large number of comparative experiments.After a long period of performance testing of the entire machine,the designed and implemented spectrum analyzer meets the needs of solar radio observation with high frequency resolution.