FFT Algorithm plays an important role in engineering applications.The requirement of precision and resources of FFT’s hardware structure become more and more important.In this paper,whole calculating process of radix-2 FFT was analyzed,and a developed method to value the error of radix-2 FFT was proposed. Taking radix-2 FFT of 16 points as an ex-ample,we analyzed the source and the transaction of errors during whole calculating.Finally,the signal-noise ratio can be cal-culated by means of the error analysis method, which is equal to 25623 ×22b.%FFT算法在工程领域有着大量的应用,精度要求也越来越高,对系统的硬件资源占用也越来越大,对FFT算法进行误差分析,精确定位每级FFT舍入的原则可更高效地指导开展FFT硬件架构设计。以基-2 FFT算法为例,详细分析了基-2 FFT算法的实现结构及误差产生来源,提出了一种基-2 FFT的误差分析模型,并以16点基-2 FFT为例,采用该误差分析方法进行误差分析,得到信噪比25623×22b(b为第1级输入位宽)。
展开▼