非平稳信号实时谱分析算法及其FPGA实现

Real-time spectrum analysis algorithm for non-stationary signal and it's implementation on FPGA platforms

针对传统谱分析仪不具有实时谱检测功能且非平稳瞬态信号分析能力不足的问题,提出一种实时谱分析方法并利用FPGA(field-programmable gate array)平台硬件实现。该方法采用长度逼近平稳信号的观察窗、多相滤波器组、线性调频Z变换(chirp Z transform,CZT)频谱细化算法和谱分析算法实现信号实时谱分析。根据系统时延分析和用户输入参数对FPGA各模块时钟频率与运算参数进行配置。仿真结果表明,该方法克服了传统基于快速傅里叶变换(fast fourier transform,FFT)算法全景谱分析和短时傅里叶变换时频相互制约的缺陷;对于平稳信号频率测量误差小于0.6%和功率误差小于4.5%,系统最大时延小于37μs;对于长度为32.768μs的非平稳信号最大时间测量误差和频率测量误差分别为0.836μs和94 k Hz。该设计除有22.558μs的初始计算延时外,对连续数据处理具有实时性能。

Given to the problems that traditional spectrum analyzer does not have the function of real-time spectrum detection and non-stationary transient signal has inadequate analysis capacity,a real-time spectrum analytical approach is presented and implemented on FPGA hardware platform.In this approach,an observation window whose window length is close to the stationary signal,poly-phase filter,spectrum zooming of Chirp Z Transform and spectrum analysis are adopted in this method to achieve real-time spectrum analysis of signal.We can give the configuration of FPGA by the clock frequency and the parameters which can be calculated according to the system delay analysis and the input arguments.The simulation result proves that this method can not only overcome the problem of traditional panoramic spectrum analysis based on the fast Fourier transform algorithm,but also overcome the shortcoming of the mutual restriction between temporal domain and frequency domain in short-time Fourier transform.For stationary signal,it has a frequency measurement error less than 0.6%and power error less than 4.5%,the maximum system time delay is less than 37μs.For non-stationary signal whose length is 32.768μs,the maximum time measurement error and frequency measurement error are 0.836μs and 94 kHz,respectively.Except for having 22.558μs initial time delay in this design.it has the property of real-time processing of continuous data.