Efficient data conversion in time-domain spectroscopy

In applications of time-domain dielectric spectroscopy, in which frequency-domain information is obtained by means of Fourier transformation of step-response data, results are often required at a number M of evenly spaced frequencies, where M is considered smaller than N, the number of equidistant samples from which the discrete Fourier transform (DFT) is formed. The object of this paper is to point out a method for reducing the computing effort required by fast Fourier transformation (FFT) techniques used for collecting spectral data