定点FFT量化误差模型及性能分析

摘要

介绍了快速傅里叶变换(FFT)的基本原理,针对硬件实现中的定点运算,分析推导出了不同FFT长度和不同量化位数带来的误差模型,并进行了实验验证.结果表明,相同量化位数条件下,FFT长度越长误差越大；相同FFT长度条件下,量化位数越多,误差越小.实验结果为FFT设计提供了参考.%The basic theory of fast Fourier transform (FFT) is introduced. An error model with different FFT length and different quantification bit width is analyzed according to the fixed-point calculation of hardware realization. The experimental results indicate that the longer FFT length is, the bigger error will be in the condition of the same quantification bit width, and the wider quantification bit width is, the smaller error will be in the condition of the same FFT length. The results provide a reference for FFT design.