The Problem of Frequency Range in Speech and Music Reproduction

Abstract

The specification of a frequency range for a speech or music reproducing system is a complicated matter, with immediate implications as to quality, efficiency, cost, and bulk. The boundaries of the spectra of speech and music are well established but the cost of reproducing these extreme ranges in a near-ideal manner is prohibitive for all ordinary applications. Studies of masking by noise and of hearing ability in the population indicate that statistically there is no real loss in restricting the band width to the order of half that of the original spectrum. Furthermore, it is the opinion of many experienced observers that something less than the complete spectrum yields optimum listening enjoyment and satisfaction. Further restriction of the frequency range entails loss of quality for essentially distortion-free sources with low noise content. The situation is examined for both speech and music with the aid of frequency diagrams for the audio frequency range which show the relationships between cut-off frequencies, band widths and the various criteria of quality and intelligibility. A rational system for departing from the complete frequency range can be developed in the form of a preferred series of reproduction bands based on the liminal intervals of Gannett and Kearney and aural balance relationships. This provides a practical approach to a system of bands with definite but just-perceptible quality differences between them, which is a most useful concept and one which would greatly aid in the engineering of sound systems. In the case of amplified speech communication, where the emphasis shifts from quality to effectiveness from a communications viewpoint, articulation, power economy, directivity and similar considerations, different bands are necessary and possible frequency ranges are discussed.