[isw]

In article ,
Audio Empire wrote:

On Tue, 28 Sep 2010 06:39:46 -0700, Sebastian Kaliszewski wrote
(in article ):

Audio Empire wrote:
On Mon, 27 Sep 2010 04:00:45 -0700, Romy the Cat wrote
(in article ):

What I found the most amassing in this story is that presentation was
made for Audio Engineering Society and it looks like they were
AMAZED!!!


I think that they were amazed by the sound of the difference signal
between
the unaltered master and the compressed copy. It was that so much
"extraneous" info was removed from the master that it was apparently
possible
to still tell what the music was supposed to be and who was singing it.
That's a lot of loss.


But that's the whole point of psychoacustic compression! Remove what
psychoacustic model deems unhearable (because it's masked by the other
parts of the signal, and our brain could not preceive it).

Wether that psychoacustic model is right or wrong is another story,
though. And that's why telling that you hear artifacts with 320bps mp3's
without disclosing encoder used is pretty useless. In lossy compression
world 320bps does not necessarily equal 320bps (from another encoder).

rgds
\SK


I'm not withholding the encoder used, Other than the fact that it's the one
used in Audacity, and the one used in Apple iTunes, I don't know what encoder
it is. I assume that since audio that's encoded with these plays back on any
MP3 player, that these encoders follow the MP3 standard (whatever that might
be).

To be in compliance with the standard, an encoder must (1) produce data
which is in compliance with the proper syntax, and (2) not break a
"reference" decoder when played through it. That's all. Note that there
is nothing there about how good it sounds...

Since I eschew MP3 as much as possible, and do not rip music using it, I
haven't spent any time learning anything other than a cursory amount about
the subject. Lossy compression simply doesn't interest me except as something
to avoid when practicable.

How do you feel about hugely inefficient encoding methods which require
vastly more data than is necessary for "transparent" reproduction? A lot
of folks think *those* are "something to avoid when practicable".

Isaac