Measurement of the attributes of complex tonal components commonly found in product sound

All rotating components in machinery produce sounds that contain tonal components, and the presence of these tones can significantly affect the quality of the product sound. Tone corrections for metrics based on weighted, average sound pressure level have been used since the late 1960s to assess annoyance due to aircraft noise and to rate climate control machinery. Much research has also been focused on measuring the strength of well separated tones in noise. Metrics such as the Prominence Ratio, the Tone-to-Noise Ratio, and variants, as well as more complex models such as the Joint Nordic Method, Aures' Tonalness, and Virtual Pitch, produce values that often correlate well with subjects' judgments on the level of the tonal features that they perceive when listening to the sound. However, when sounds are more complex, these metrics do not always work well. Models for the tonalness of two types of tonal sounds are considered here: narrow-band random noise and tones with random frequency fluctuations. The influence of bandwidth and roll-off rate on perceived tonalness are explored for the narrow-band sounds, and the effect of the range and the rate of change of the frequency variation on perceived tonalness is explored for frequency modulated sounds. It was found that roll-off rate affected the perception of tonalness, and that when frequency variations could be tracked or were very small, metrics derived from averaged spectra produced inaccurate predictions of tonalness. Based on the results of these two investigations modifications to tonal metrics are proposed.