Data recording and processing for high-resolution frequency analysis

This paper discusses the practical considerations associated with analyzing analog data signals to very high resolution in frequency. In the problem considered here, the data set is recorded on analog tape, digitized, and processed on a digital computer. A primary concern is that recorder speed variations, or flutter, can produce artifacts which might become evident when high-resolution spectral analysis is performed. This study concludes that fractional millihertz analysis resolution can be achieved with existing analog tape recorder technology. Another problem which must be considered is the efficient storage of the large data sequences resulting from processing long-time signal records. The use of two-stage frequency analysis is described as an approach to addressing this problem. Finally, the implementation issues associated with the periodogram and AR (autoregressive) estimation algorithms are outlined. It is anticipated that the discussion of practical considerations given here will be useful to researchers who must address problems that are similar to those cited in this paper.