Spectral measurements made from finite duration observations
Abstract
The present article is concerned with the consequences of calculating the spectra of continuous signals from finite length observations using Fourier Transform methods. It is shown that when this is done, the frequency resolution of the resultant spectrum varies inversely with the length of the observation period. The form of the spectrum is modified by the introduction of spurious frequency information, which may be suppressed by in turn modifying the observed data by means of 'windows.' In the special case where the signal is periodic, it is beneficial to perform the analysis over an observation period equal to an integral multiple of the signals period. In the general case, the accuracy of the spectrum can be improved by averaging. Averaging reduces resolution and requires that further observations up to the available observational period are made.
 Publication:

Society of Environmental Engineers Journal
 Pub Date:
 March 1975
 Bibcode:
 1975SEEJ...14...11B
 Keywords:

 Fourier Transformation;
 Power Spectra;
 Signal Analysis;
 Spectrum Analysis;
 Periodic Variations;
 Time Response;
 Windows