Re: [Gtkpod-devel] MP3 Gapless Logic Possibly Incorrect?

[John R. <sax...@gm...>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Sun, Jan 24, 2010 at 02:23:35PM -0600, Michael Tiffany wrote:
> On Sun, 2010-01-24 at 13:54 -0500, John Regan wrote:
> 
> > I think you may be on the right track here, I had no idea --nogap was 
> > deprecated!  When I pull up the man page for LAME, it's in there with no 
> > mention of anything being deprecated.
> 
> It may not even be really deprecated, but it just doesn't really have to
> do with what we're doing, despite the name.  I believe that if you
> encode multiple mp3s on the same command line (otherwise nothing
> happens), then it takes what would usually be the postgap of track 1,
> and fills the rest of the sample with the start of track 2 and so on.
> The problem is if you ever listen to the tracks not exactly in order,
> you'll have very small samples of tracks on the wrong track.

For me, that's never been much of a problem - I only have a few albums 
that I've really cared about being gapless (because the tracks run into 
eachother), and I tend to listen to those in a single run-through.  So, 
I had never noticed a problem before!

> 
> > I re-encoded the album with my usual script, loaded it up into gtkpod, 
> > and it added the gapless track flag.  I'll attach the script I use, it's 
> > nothing fancy.
> 
> OK, I finally replicated the zero postgap.  I needed to use both the
> --nogap and the --nogaptags flags for the lame tag to be written, and
> then encode multiple mp3s on the same line.
> 
> > I should mention, I went ahead and un-applied my change to gtkpod, so 
> > I'm running off of what's currently in the git repo again.
> > 
> > I haven't tested this on my iPod yet, and won't be able to until later 
> > tonight, but if gtkpod is checking the gapless track flag, I'll take 
> > that as a pretty good indication it'll work.  I'll follow-up on that, 
> > though.
> > 
> > I do have some concerns, though:
> > When I ran eyeD3 --lametag on the new MP3, there was no mention of 
> > gapless, unlike my old MP3s.  Here's the output:
> > 
> > [john@slim mp3]$ eyeD3 --lametag 01\ -\ Once\ Again.mp3
> > 
> > 01 - Once Again.mp3     [ 3.66 MB ]
> > - -------------------------------------------------------------------------------
> > Encoder Version     : LAME3.98r
> > LAME Tag Revision   : 0
> > VBR Method          : Variable Bitrate method2 (mtrh)
> > Lowpass Filter      : 18500
> > Radio Replay Gain   : -6.6 dB (Set automatically)
> > Encoding Flags      : --nspsytune --nssafejoint
> > ATH Type            : 4
> > Bitrate (Minimum)   : 32
> > Encoder Delay       : 576 samples
> > Encoder Padding     : 1632 samples
> > Noise Shaping       : 1
> > Stereo Mode         : Joint
> > Unwise Settings     : False
> > Sample Frequency    : 44.1 kHz
> > MP3 Gain            : 0 (+0.0 dB)
> > Preset              : V2
> > Surround Info       : None
> > Music Length        : 3.66 MB
> > Music CRC-16        : A615
> > LAME Tag CRC-16     : 80B6
> > 
> > Also, I tried reloading MP3s from an album that I know had track gaps, 
> > and they showed up as gapless, too - is that correct behavior?  Maybe 
> > I'm misunderstanding what "gapless" is supposed to mean, but I always 
> > assumed that I shouldn't see it on anything that I didn't specifically 
> > encode as gapless.
> 
> I believe it is a misunderstanding.  The only time the lame tag reports
> No Gap data is if it was encoded with the --nogap flag.  Encoded mp3s
> have a pregap to account for encoder and decoder delay, and they have a
> postgap to pad the length of the data to an full frame (multiple of 1152
> samples).  For music to be played gaplessly, these values must be known
> so the player knows exactly where the real data stops and starts.  This
> is the case whether the tracks actually have music that runs together or
> not (though you'll really only notice it if they do).  Therefore,
> gapless playback can work for any pair of songs where the pregap and
> postgap are known, but for tracks with silence between them, it just
> means that the silence from track 1 cuts off exactly when it should and
> the silence from track 2 starts exactly when it should rather than there
> being a small delay or pop.  lame will add the pregap and postgap data
> to *every* encoded file unless it's told to not write the lame tag, so
> every file should have the gapless flag set in gtkpod.
> 
> That hopefully isn't too confusing.  I did the best I could :)
> 
> > Both these albums were ripped with Exact Audio Copy and encoded to FLAC, 
> > then I'll use my lil' script to convert to MP3 for my iPod/sharing with 
> > friends.
> 
> Let me know how the ipod handles it.  With the gapless flag set I think
> you'll be ok.  I wasn't convinced that your tracks should have played
> gaplessly before, but now I think they will, but only when played in the
> proper order.  Otherwise, you lose the advantages of true gapless
> playback.  I'm inclined to say we should leave the code as is, but I
> could see removing the check for postgap as long as people understand
> what it means.
> 
> Michael
> 
> 
> -- 
> Michael Tiffany
> ti...@ti...
> http://www.tiffman.com
> 

So I did the final test - does it play gaplessly on the iPod?

The answer is yes :)

So, I think the problem then, this whole time, was that I was encoding 
with those --nogap and --nogaptags switches, which creates a zero-length 
padding.  I'd say leave the code as-is, but maybe in some manual or some 
wiki somewhere, specify that those --nogap and --nogaptags are not 
needed, and will in fact not work with gtkpod?  I can't be the only one 
that sees the --nogap and --nogaptags in the man page of LAME and think 
"Oh, I should use this switch, then, if the tracks run into eachother."

Alternatively, and this seems like it'd be more work, gtkpod could 
recognize these MP3s encoded with --nogap and figure out what to do.  
The warning on a wiki or manual seems to be the much quicker and simpler 
solution.

Anyway, I have a much better understanding of what "gapless" is, and I'm 
very content with gtkpod now - it seems to be the only Linux iPod 
manager that does everything 100% correctly... just gotta make sure 
*I'm* doing everything 100% correctly, too. :-D

- -- 
John Regan
(m) +1 904 417 8788
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.10 (GNU/Linux)

iEYEARECAAYFAktdL0gACgkQFIrb5Z8QrLGuZQCeJUERH7G28BlvOyslYi+/ed9p
L2MAnAyJKVvjV/eKn1BizEtXkUhdRjh0
=57Kv
-----END PGP SIGNATURE-----