SuperQ;444676 Wrote: 
> When I said implementation, I was talking about things like metaflac.  I
> guess I was wrong, metaflac will write tags that could clip tracks, but
> provides a hint to newer software that there is a peak to consider.
> 
> This seems to be a bad backwards compatibility design in the tag
> system.  If you designed the decoder before the peak tag existed, you're
> kinda screwed.
AFAIK the peak tag has always existed. Playback systems which implement
the GAIN tag but ignore the PEAK tag are in principle broken.

It seems you would rather that there were no PEAK tags, but simply have
GAIN tags that were scaled back at analysis time so as to avoid
clipping. This achieves nothing other than to allow for lazy, broken
implementations to get away with it.

There are good reasons why this is not a good idea:

Suppose the suggested boost would be +6dB, and the peak level is -2dB.
I take it that you would prefer the GAIN tag to be set to +2dB so as to
prevent clipping. What happens is that when played back, the file will
be perceived as being 4dB quieter than others - and rightly so, because
of course we do want to avoid clipping.

But now suppose the playback system incorporates other digital volume
control, and the listener happens to have set his digital control to
-9dB. If the GAIN and PEAK tags are +6dB and -2dB (ie. the way that it
does actually work in replaygain), we can see that within the context of
an overall volume level of -9dB, we are able to apply the full +6dB
adjustment. The perceived loudness of the track is now in line with
other tracks. (As it happens, the Squeezebox system does have digital
volume control but does NOT take advantage of this additional
possibility, which is a shame. I decided not to push this issue in the
bug report - getting the clipping fixed was far more important).

There are other non-audiophile perspectives to consider. It may be that
in a background listening environment, we really do want to have equal
perceived loudness at all times, and in this case some kind of peak
limiter can be applied to prevent hard clipping but still allow the full
amount of overall boost.

But under your desired approach, the playback system has no idea what
the original boost would have been, and the track will remain 4dB
quieter than other tracks in all circumstances.

I hope you can now see why the way that replaygain tags work currently
is actually the optimal solution.


-- 
cliveb

Transporter -> ATC SCM100A
------------------------------------------------------------------------
cliveb's Profile: http://forums.slimdevices.com/member.php?userid=348
View this thread: http://forums.slimdevices.com/showthread.php?t=65911

_______________________________________________
audiophiles mailing list
[email protected]
http://lists.slimdevices.com/mailman/listinfo/audiophiles

Reply via email to