dB(A) has come to be viewed as a convenient single number description of noise emission. That definition is overly optimistic and may lead to unpleasant surprises.
History
In the 1930s Bell Laboratories was engaged in fundamental research on human hearing for application to telephone communications. One area of interest was the perception of loudness. Participants at the 1939 World’s Fair in NYC rated the loudness of tones of various frequencies relative to that of a 1 kHz tone.
One thing that was discovered was that low frequency tones of the same energy produced a smaller loudness sensation than the reference tone. But for quieter sounds the difference was more dramatic than louder ones.
Bell Labs systematized their findings into three simple four-pole electrical networks that simulated the loudness response vs. frequency for quiet sounds (the “A” network), speech-level sounds (the “B” network), and loud sounds (the “C” network).
The sound pressure level of a tone, after filtering by the appropriate network, was labeled dB(A), dB(B) or dB(C), and correlated decently with the perceived loudness. The A-network strongly de-emphasizes sound energy below 500 Hz.
What it’s actually good for.
The practical uses of dB(A) are very different from its original intent:
- dB(A) levels correlate well with hearing loss in industrial environments.
- dB(A) levels correlate well with annoyance of certain types of transportation noise sources.
- dB(A) level differences give an indication of relative loudness for two sounds of similar spectral content, unless there is significant energy below 500 Hz.
What it’s not good for
- dB(A) levels do not correlate reliably with perceived loudness for non-tonal sounds.
- dB(A) levels does not correlate with perceived loudness for sounds with strong low-frequency content.
- dB(A) does not flag the presence of prominent tones.
- dB(A) level differences do not correlate with perceived loudness for dissimilar sounds.
- dB(A) is not a good predictor of annoyance, which has several perceptual dimensions.
Discussion
dB(A) was never intended to be a universal metric for annoyance. It came into wide use because it had some relationship to perceived loudness and was easily incorporated into sound level meters.
There are situations in which, by virtue of special-case limiting assumptions, the dB(A) metric works well. Unfortunately the limiting assumptions are largely forgotten.
So its primacy is largely an accident of history.
Instead a range of Sound Quality metrics have been developed that give a more precise indication of perceptual components of the sound. Combinations of these metrics are correlated with subjective responses in order to study the factors determining acceptability of a product sound.