DNV 418-2 PDF

This standard describes the hearing model and psychoacoustic metrics dependent on the hearing model. The input to the hearing model are sound signals recorded using the procedures of ECMA-74. The hearing model expresses specific loudness ^[1]. Psychoacoustic models use the specific loudness to express the strength of any tonalities or roughness in the sound generated by Information Technology and Telecommunications (ITT) equipment. While developed for ITT equipment, the psychoacoustic methods of this standard may be relevant to other applications like automobiles, consumer appliances, etc.

The tonality metric of this standard uses the auto-correlation function to describe causes of perceived tonality such as individual or multiple steady or time-varying discrete tones, individual or multiple spectrally elevated bands or slopes of noise, and combinations of these phenomena. A similar approach was published in 1998 to determine “pitch salience” ^[2].

The roughness metric presented in this standard uses a spectrum of the sound signal envelope, refined by a quadratic fit estimator, to describe roughness arising from sound signal envelope variations within a critical band at modulation rates between 20 and around 300 Hz. For steady sounds, roughness perception peaks at modulation rates of 70 Hz.

The loudness metric of this standard uses a nonlinear combination of tonal and noise loudness calculated as intermediate results of the tonality algorithm to achieve a very good match of perceived loudness, especially for sounds with a subcritical bandwidth (sounds containing tonal and noise components).