The difference between A weighting and an unweighted measurement depends on=
the
bandwidth and the spectrum of the noise, as has been said previously. The=
amount of the difference can be calculated by integrating the weighting cur=
ve
over the bandwidth of the measurement. You don't have to pull out your cal=
culus
text book because I just did the calculation.
for white noise in a 48 kHz bandwidth, A weighting reduces the measurement =
by
Message: 2.
Subject: 72 dB
for pink noise in a 48 kHz bandwidth, A weighting reduces the measurement b=
y
Message: 3.
Subject: 70 dB
As shown by Robin below, the difference in level between ITU-R 468 weightin=
g and
A-weighting for a microphone tends to run about 10 to 12 dB. The A-weighti=
ng
curve and the 468 curve are different, but they're not that different, othe=
r
than the offset.
I maintain that A-weighting is better than 468 weighting for determining th=
e
audibility of noise, although neither of them is particularly accurate. I =
can
demonstrate why this is if anyone is interested. But the important point i=
s
that if one were to order a group of microphones in terms of their A-weight=
ed
noise or in terms of their 468 weighted noise, the ordering is almost the s=
ame.
That range of 10 to 12 dB for the observed difference between A-weighting =
and
468 weighting has mostly to do with whether the microphone is an omni or a=
pressure-gradient type. Because pressure-gradient microphones are resistan=
ce
controlled (the condenser ones, that is), they tend to have considerably mo=
re
low-frequency noise than do omnis. That accounts for the 2 dB difference.
One final point. I don't think that this has been published anywhere, but =
the
threshold of audibility for noise is about 6 dB SPL. If the reproduction i=
s
done at a natural level, if a microphone having a self noise of less than 6=
dB
SPL was used then the noise in the recording will be inaudible.
Eric
|