[CrackMonkey] Vorbis to replace MP3 on Windows XP!

[Seth David Schoen]

Ben Brockert writes:

> Monkey Master wrote:
> >         Some other reasons why teh mp3->vorbis thing isn't as simple
> > as gif->png:
> > 
> >         1: both mp3 and vorbis are lossy, so doing a conversion from
> > one to another is not as simple as the gif2png stuff.
> 
> There shouldn't be anything lost in the conversion, though, as ov is a
> better codec. If someone would just make a free drag and drop converter
> for Windows and MacOS that just converted and kept the quality the same
> and the bitrate the same, it'd help a lot.

I don't think it's that easy.

Here's a simple hypothetical example: it could be that lossy codec X
drops some information in the time-domain which psychoacoustic models
say listeners won't miss much.  And lossy codec Y drops some
information in the frequency-domain which psychoacoustic models say
listeners won't miss much.  Now, if you compress something with codec
X and then uncompress it, you get an uncompressed version that's
_still missing_ that time-domain information.  If you play it, it may
still sound OK, because you didn't need that time-domain information.
But if you now compress it with codec Y, it may sound _awful_, because
now both the time-domain and frequency-domain information have been
stripped out, and perhaps it happens that you _do_ need one or the
other, at least, to get good sound.

The general problem is that different lossy codecs are lossy in
different ways, and yet the particular losses due to a given codec
can't normally be undone upon decompression (that's why it's called
lossy).  So if you apply two different lossy codecs, the total loss of
quality might be _substantially_ greater than if you just applied one,
because the particular information they strip out could be completely
different.

Another simple example: if you have compression method A which is
"drop the low-order bit of even-numbered samples" (and is patented)
and compression method B which is "drop the low-order bit of
odd-numbered samples" (and is not patented), then if you had something
compressed with A, and then uncompressed, if you compress with B, you
have an additional loss of quality which is every bit (no pun
intended) as bad as the original compression with A.  And the
information that you're losing is different information, and you don't
get back what you lost the first time.

-- 
Seth David Schoen <schoen at loyalty.org>  | And do not say, I will study when I
Temp.  http://www.loyalty.org/~schoen/  | have leisure; for perhaps you will
down:  http://www.loyalty.org/   (CAF)  | not have leisure.  -- Pirke Avot 2:5