论文部分内容阅读
目的 以TMS320C62x为例,针对定点快速傅里叶变换(FFT)进行研究方法借助复平面单位圆,对采用原码截断算法的字长误差进行理论分析,对原码、补码格式下截断与舍入算法进行仿真比较;并讨论噪声对字长误差的影响结果 对线性相位信号作r级FFT运算时,原码截断算法产生的有限字长误差会使频域输出绝对值减小,其误差均值和方差都与r成线性关系;对信号加入适当的白噪声,误差的均值会明显降低结论TMS320C62x仿真器上得到的结果验证了理论分析的正确性
Aim To take TMS320C62x as an example, aiming at fixed point fast Fourier transform (FFT), the paper uses the flat unit circle to theoretically analyze the word length error of the original code truncation algorithm, The results show that when the linear phase signal is r-level FFT operation, the finite word length error generated by the original code truncation algorithm will decrease the absolute value of the output frequency in the frequency domain. The error mean value And variance are linear with r; adding the appropriate white noise to the signal, the average error will be significantly reduced Conclusion The results obtained on the TMS320C62x simulator verify the correctness of the theoretical analysis