System Of Virtual Spectrum Analysis Based On DSP

Digital signal processing almost relates to all engineering fields and spectrumanalysis is one of the most important techniques in it. Since the first Digital SignalProcessor (DSP) could be used in 1970's, DSP brought significant chance tospectrum analysis technology for its grestest speed speed and efficiency.Nowadays, most top grade apparatus of spectrum alalysis are depended onimporting and their price is very expensive. The appearance of Virtual Instrumentprovides our country an opportunity to develop our own low cost equipment ofspectrum analysis.Just for the reason of above, combining with technology of Virtual Instrumentand DSP, we developed the system of virtual spectrum analysis to study thetechnique of spectrum analysis.The system can be separated into two parts: Virtual Instrument part and DSPpart.In Virtual Instrument part, we finished the front panel design of the system;relized controlling DSP's operation and transferring parameters to DSP;communicated with DSP and display the spectrum analysis results of DSP.The paper in this part simply introduces the operational principle ofLabVIEW—the Virtual Instrument developing platform and detailedly introducesevery section's function of the system's front panel.We also expatiate how torelize controlling and communicating with DSP, and provide the important APIfunction.In DSP part, the paper also is divided two parts.One is how to sample the signal by using the 6701EVM board of TI Corp.Thepaper in this part specializes every hardware element concerning sample,includingcodec,McBSP,DMA and HPI. It also gives every part's configuration and flowdiagram and analyses them.The other is how to complete the data processing algorithm in DSP.The paper in this part firstly classifies the signal that can be processed in thesystem.The signal mainly is divided into deterministic signal and stationaryrandom signal.Every kind of signal has its special spectrum analysis way.It's necessary to truncate data in course of analyzing spectrum and need touse window function.The paper introduces the five kinds of window functionsproviding in the system.To deterministic signal,it's discrete Fourier transform (DFT) that make itfrom time domain to frequency domain.Just because of fast Fourier transform(FFT) algorithm,DFT can be realized to real-time processing.In this part of thepaper ,we anatomy the FFT algorithm in detail and complete the split-radixalgorithm.In this part of the paper ,we also discuss resolution of frequency.Bydistinguishing the concept of " resolution of practice " and " resolution ofcalculation ",we point out that resolution of frequency is inverse ratio with theactual length of the data and zerofill can not improve it.To stationary random signal, it's not analysed by FFT but by power spectrumon account of that it is power signal.The system completes classical powerspectrum method and modern power spectrum method.The classical powerspectrum method is divided into periodogram method and ameliorationmethod.the spectrum characteristic of periodogram method is not verygood,whose spectrum line is too fluctuated when N is large.So we complete welchalgorithm of amelioration method to improve it.The polestar of welch algorithm isto make the signal segmentation,superposition and averaging,to smooth thespectiom.The variance characteristic of classical power spectrum method is bad and itsresolution is low.Modern power spectrum method solves the defect well.Wecomplete burg algorithm of AR model,its smoothness and resolution are all betterthan classical power spectrum method.In the end of the paper, we provide the test of the system and compared thethe characteristic of all kinds of algorithms.