作者: D. Fraser
DOI: 10.1109/29.17559
关键词: Algorithm 、 Mathematics 、 Wavelength 、 Digital filter 、 Convolution 、 Linear interpolation 、 Discrete mathematics 、 Interpolation 、 Fast Fourier transform 、 Nyquist frequency 、 Root-mean-square deviation
摘要: A numerical investigation into the accuracy of interpolation by, fast Fourier transform (FFT), using a sinusoidal test signal, is described. The method precisely defined, including previously unnoticed detail which makes significant difference to result. experiments show that, with no input windowing, almost independent wavelength very close Nyquist limit. resulting RMS error inversely proportional sequence length and low for lengths likely be encountered in practice. As passes through limit, there sudden increase error, as expected from sampling theory. If ends are windowed by short, cosine half-bells, further improved at longer wavelengths. In comparison, small-kernal convolution methods, such linear cubic convolution, perform badly wavelengths anywhere near >