Gianni Pavan wrote:
> The MP3 works both in frequency and in time; by using algorithms based on
> both frequency and time masking, it can cut a short segment of a signal if
> it is just after a strong sound. This happens with sperm whale clicks: if
> they are very loud, after each click the spectrogram shows a short segment
> of silence (it is not noticeable by hearing, but it may cut some important
> features we measure in the 10 ms after a click!).
Something you may want to investigate as part of this sort of detailed analysis:
Are you doing a sonogram or just looking at the raw waveform? If you are
doing a sonogram are you sure what you are seeing is not an anomaly of
the particular FFT routine your software uses. At 10 ms you are not
talking all that many samples. The FFT routines have their own
anomalies, which are much more severe than those produced by
compression. This seems to be particularly true right around loud sounds.
Have you tested your analysis using different software? There is quite a
bit of variability in the display you get from different software. Or
even within different software at different settings.
Walt
________________________________________________________________________
________________________________________________________________________
|