Dan wrote:
> Correct in concept but one has to be very careful in using a tone. I've t=
ried it with 1 kHz, and found that if I hand-held the meter the reflection =
from my body will change the reading up or down as much as 2 dB.
Yep, this this can be a problem. So, ideally, the calibration should be don=
e in an environment with as little reflections as possible. I believe that =
it is generally an illusion to get very precise sound level measurements wi=
th errors less than 3 dB. There are just too many other parameters that inf=
luence the results. Sources of error are for instance the polar pick-up pat=
tern of the microphone and the temporal structure of the noise to be measur=
ed.
>I suggest using band-limited pink noise, 200 Hz to 2 kHz, for the stimulus=
, at a level of 64 dBA (slow) measured by the sound level meter at the mic =
location, the SLM on your field mic stand at the height you use. Then put t=
he capsule in the same spot, with field windscreen, and the recorder set to=
a precisely repeatable record gain.
Yes, but this would unfortunately require to take into account the A-weight=
ing filter of the sound level meter (which was not the case with a 1 kHz te=
st signal). If the sound level meter however also had a C-weighting option =
(http://en.wikipedia.org/wiki/Weighting_filter), then a band-limited noise =
from 200 Hz to 2 kHz would be fine.
> > The final noise measurements could then be carried out directly on the =
recorded .wav files.
>
> Using an A-weighted software meter for the level calibration.
I'm afraid that this additional A-weighting filter might make the calibrati=
on procedure a bit too complicated. But it one used the C-weighting option,=
then there would be no further weighting curve compensation required for s=
uch a 200 Hz to 2 KHz signal.
Note that the animals don't care about A-weighting at all (which is adopted=
to the human auditory system). So, I would recommend to deal entirely with=
unweighted noise levels in this application.
>
> > This would also allow you to determine the spectral distribution of the=
ambient noise, which may also significantly influence the vocalization par=
ameters of the singing birds. A number of studies have shown that many anim=
al species shift their songs towards higher frequencies in order to avoid m=
asking effects by low-frequency ambient noise.
>
> Third-octave analysis would surely be enough for this purpose, maybe even=
octave analysis to reduce the quantity of data to be compared.
Yep. Unfortunately, sound analysis software that supports (third-) octave a=
nalysis is usually expensive. One could instead also use simple FFT spectra=
to compare the different noise characters.
Regards,
Raimund
|