This is a weak answer, but I believe that a PC has no idea of true =
"time" in an MP3 file. I think it only knows the number of samples, and =
the sample rate, and infers time based on the sound card (?)'s clock =
crystal. Perhaps the accuracy of that crystal can be called into =
question. It's also possible that PCs measure time by CPU clock cycles; =
if that were the case, timing would be less than perfect since the CPU =
is interrupt-driven (IRQs) and doesn't always run smoothly in real time. =
Finally, since MP3 is a compressed file format, it's possible that the =
compression is not linear in time, and that decoding relies on some sort =
of interpolation of the original samples in the time domain, resulting =
in drift. I am basing these statements on my background understanding =
of audio and computers (don't laugh), and want to admit that I have no =
specific information on this subject. But it's pretty interesting to =
me, so I look forward to reading more informed posts than mine. You do =
seem to have stumbled into a fairly arcane area.
My interest lies in my desire to assemble a multi-track recording rig on =
an extremely low budget. One of my latest thinking exercises was to =
simultaneously use between 2 and 4 little portable stereo MP3 recorders =
(such as my iRiver iFP8xx) to create a 4 or 8-track digital recorder. I =
abandoned this idea because I decided that clock drift would prevent me =
from accurately synchronizing the tracks when I import the separate MP3 =
recordings into Audacity.
Rich
----- Original Message -----=20
From: qukza=20
To: audacity-users@...
Sent: Monday, November 21, 2005 10:06 AM
Subject: [Audacity-users] Time discrepency between MP3 and WAV
Hi,
I have a question about the time representation of audio files. I have =
an MP3 file (22050Hz, 16 bit, mono) that is a recording of an interview. =
The recording is about 62 minutes long. two to three minutes before the =
end of the recording, the interview ends and there is silence until the =
end of the recording. When I load the recording into Audacity the =
program indicates the point at which the interview ends is about 59:37. =
I converted the MP3 to a PCM file and opened it in Audacity and the =
interview end point is now about 59:33:25, an almost 4 second =
difference. Cooledit 2000 behaves in exactly the same way. If I copy =
both the MP3 and WAV files onto a Marantz PMD670 Digital Recorder and =
play back each file, the time display on the recorder indicates that the =
interview ends just after 59:33 for both files--so the Marantz is =
consistent. I have another audio program and the discrepancy when using =
MP3 is even greater, the audio ends at 59:54 o r a 21 second difference =
between the wav and mp3 file.=20
So what's the real time? I'm guessing the times I am getting on the =
Marantz and the WAV files. If that's the case why is the time line of =
MP3s misrepresented in PC software and is there any way to adjust or =
work around this issue when using MP3s? It also seems to be an issue of =
drift. That a couple of minutes into the file one might not notice a =
discrepancy but at an a hour or more the discrepancy is very apparent.
Thanks, Alan.
-------------------------------------------------------------------------=
-----
Yahoo! FareChase - Search multiple travel sites in one click.
|
