gapless in Winamp - not gapless in Rockbox?

[nieroster]

I have a question regarding gapless playback in Rockbox. I've downloaded a live album from http://www.coldplay.com/lrlrl/lr.html and played it back in Winamp: no problem! I played it back on my H140 using Rockbox: small gaps between the tracks.

I investigated a litte and found out that the mp3 files have been encoded by FhG. They play gapless in Winamp, but not in foobar2000. When I edit the gapless information in foobar2000 and set the encoder delay to 576, they also play gapless in foobar, but not on my H140.

Does Rockbox ignore the encoder delay tag when a file is not encoded by lame? Why does Winamp play the tracks gapless if they aren't?

[yapper]

See http://www.rockbox.org/wiki/GaplessHowTo which includes the following warning:

If your format of choice remains MP3, then you can still have gapless playback by using any recent version of the LAME encoder, which will encode gapless by default. From LAME v3.90.3 upwards Rockbox will decode gapless, files encoded by older versions or other encoders are not guaranteed to playback gapless in Rockbox.

[nieroster]

See http://www.rockbox.org/wiki/GaplessHowTo which includes the following warning:

If your format of choice remains MP3, then you can still have gapless playback by using any recent version of the LAME encoder, which will encode gapless by default. From LAME v3.90.3 upwards Rockbox will decode gapless, files encoded by older versions or other encoders are not guaranteed to playback gapless in Rockbox.

I've read this of course and I know that I can reencode the files, but this is not an elegant solution.

If for whatever reason a player can play them gapless, why can't rockbox?

[Llorean]

Because nobody has implemented any other *accurate* gapless methods for Rockbox.

Many players will attempt "gap removal" type functions by attempting to detect "silence" (not always true silence) at the end of a track, or other methods to approximate near-gapless or gapless-sounding playback.

Rockbox uses special tags lame adds to identify where the encoded audio actually ends and gets as near true gapless as possible on MP3.

If you know the exact methods these other programs use to handle gapless playback in non-lame MP3 files, it's possible one could be implemented, but it's somewhat unlikely the "approximately gapless" methods many use will be accepted.

[markun]

The files are encoded with iTunes which also adds tags for gapless playback. We already support these tags in AAC. Since ID3 tag handling has been moved to apps/metadata in the source tree it should be easier to add support for it.

EDIT: good news I manged to add support for iTunes gapless. The code is not clean enough to be committed, but I can post a patch on the patch tracker lateron. I tested it with your coldplay album and there are no gaps.

EDIT: patch added: http://www.rockbox.org/tracker/task/10629

[nieroster]

Well thanks a lot! I did not expect that the issue will be treated so quickly!

Thanks again, I will test.

EDIT: It works great! I've also tested another album encoded in iTunes: no gaps!

!!GREAT WORK!!

[markun]

I made a second patch (don't know if you tried that one or the first one) and it's now in SVN.

[zfox]

A popular AAC codec w/out gapless playback support by Rockbox is Nero AAC http://www.nero.com/eng/downloads-nerodigital-nero-aac-codec.php

Adding support to Rockbox would be nice. The Apple firmware supports gapless playback with this Nero codec.

[markun]

Do you have some test files for me?

[saratoga]

I asked one of the Nero developers ages ago about how they stored their gapless info and he said the documentation is the faad source code.

[zfox]

Do you have some test files for me?

http://users.auth.gr/~kyriak/temp/

[markun]

Thanks for the files.

I see 4 Nero specific atoms in the mp4 files: "ndrm", "flags" and "chpl". Not sure of any of those is used to store the gapless information. I can't find much information online.

[zfox]

Making a question I got the following reply...

The gapless data is in the chapter list "chpl". MP4Box can handle this, so that's an example of how to read it. Basically, the chapter describes when the track starts in the decoded data and when it ends.

[markun]

I made a test file and get the following info:

with mp4box -info:

Code: [Select]

... Duration 00:00:31.059
... 1338 samples

with mp4box -diso:

Code: [Select]

  <ChapterListBox>
  <BoxInfo Size="26" Type="chpl" /> 
  <FullBoxInfo Version="1" Flags="0" /> 
  <Chapter name="" startTime="00:00:00.059" /> 
  </ChapterListBox>

in foobar2000:

Code: [Select]

Duration : 0:31.000 (1367100 samples)

Do you think these 0.059 seconds should be trimmed from the end or the beginning? Well, we can just implement it and test both of course.

[zfox]

Decoding my files to wav with winamp (no gapless playback for nero files) you see the gap (with a wav editor) at the beginning.