[Sonnova]

On Fri, 23 Nov 2007 10:09:11 -0800, Arny Krueger wrote
(in article ):

"Sonnova" wrote in message

On Tue, 20 Nov 2007 14:45:33 -0800, Steven Sullivan wrote
(in article ):

Sonnova wrote:

Depends on the music. I can always tell an MP3 on
classical,

Color me skeptical. 'Classical' isn't necessarily harder
to encode than nonclassical music.

Why don't I color you "not thinking about it enough"
instead? No, classical isn't necessarily harder to encode
than pop. But because of the much larger dynamic range of
classical music (ppp to fff) it's easier to hear the
artifacts than with pop and rock which tends to run the
gamut from ff to fff! While this isn't always the case
with pop, it is with the vast majority of it. The limited
dynamic range (read that LOUD) that most pop music has
masks most of the audible artifacts.

Doesn't wash. Dynamic range is often the enemy of the audibility of
artifacts.

I'm sorry. You're wrong. It's during soft passages and during changes from
loud to soft (and vise-versa) where compression artifacts are most likely to
be heard.

Also, the idea that pop music necessarily has limited dynamic range is yet
another old high end audiophile's tale.

Frankly I don't care. I don't listen to pop and rock ever. I just know that
loud rock masks the effects of compression almost completely.

Classical music has some built-in
limitations on dynamic range. While there might be some creschendos and a
few sonic spectaculars, most of it is pretty tame from a dynamic range
standpoint.

Pure bull. There are thousands of works in the standard repertoire that go
from triple pianissimo to triple fortissimo (or vice-versa) rather
instantaneously. and many more cases of the music going from a single
instrument playing softly one moment, to the entire orchestra in crescendo
the next.

Because orchestral music has to be recorded in large rooms with
something like 100 people in attendance, and distant micing is the style,

It's not everybody's recording style, but that's not what I'm talking about
anyway. Unless the original recording was compressed in dynamic range when it
was captured, the difference in dynamic contrasts between a single flute
playing softly and the entire 80-100 piece orchestra playing an fff crescendo
can be as much as 40dB, that's 100:1 in terms of voltage - more than enough
to make these artifacts apparent.

the noise floor is higher than what can be achieved in a well-isolated
studio with a few closely-miced musicians.

And that's relevant, how? We're discussing the change in signal level between
the loudest point in a piece of music and the softest, not where the noise
floor is located.

If I could ensure you wouldn't use any wav analysis
tricks to identify the mp3 from source, I'd be happy to
test your hearing on this.

Since this poster uses a handle, and not a legal name, he would probably not
to want to submit to any proctored tests.

And since we likely live thousands of miles apart, how would one establish
such a test? You're talking nonsense.

It's not that hard. Believe me, when the dynamic range
changes suddenly there is an accompanying, uncorrolated
artifact that is as unmistakable as it is unpleasant that
you cannot miss once you've heard it.

I've heard that story many times, too. It's another thing we hear right
before the random guessing starts! ;-)

So you'er saying that you can't hear it? Hmmmm. Very interesting.

You don't actually think that a lossy compression
algorithm could throw portions of the waveform away
without it being noticeable at least occasionally, do you?

Sure why not? Or, don't you believe in masking?

Under certain circumstances, I certainly do believe in masking, but an
algorithm by itself can't apply a lossy compression scheme to music in a
foolproof manner. Here's an analogy - don't take it too literally, but it is
illustrative. Do you watch digital TV? You know, DVDs, digital cable or
satellite? If so, then I'm sure that when watching a TV program that you have
noticed the picture momentarily break-up into a screen full of little boxes
with some scenes but not with others. I'm sure that you have also noticed
that this almost NEVER happens on purchased DVDs of hollywood movies. Reason?
The TV station or satellite provider uses an automatic compressor to encode
the video into mpeg2 or mpeg4-H.264. This compressor is utilizing an
algorithm to do the compression. The algorithm can apply very limited
intelligence to the compression process, following as it does, a set of
rules. Certain types of scene changes, changes in lighting level, speed of
motion in the scene, etc, catch the algorithm out and compression artifacts
in the form of pixillation occur. The reason why you don't see this in DVD
releases of, for instance, Hollywood feature films, is that they don't use an
algorithm - at least not by it itself. They have a human being watching the
film frame-by-frame as it's being transferred to adjust the amount of
compression being applied on a scene-by-scene basis. At a point where the
picture would pixillate using an algorithm, the human compression engineer
will cut-back on the amount of compression until the pixillation disappears.
In some scenes like the lightbulbs popping and flashing when they hit the
cold Atlantic water that was quickly filling the ship in "Titanic". They had
to transfer those scenes to DVD with NO compression at all during the frames
where the screen goes completely white during light-bulb "explosions" to keep
them from pixillating. If you watch the movie on HBO, you'll see that the
algorithm can't handle those scenes and the picture breaks up for several
frames every time a lightbulb pops. That's what's wrong with algorithms. They
have very limited range over which they can make compression "decisions".

Now, I'm very sure that given complete control over the MP3 encoding process,
someone who is very familiar with the failings of the compression algorithm
and very familiar with the music could make a compressed MP3 of, say, Ravel's
"Daphne et Chloe" ballet in such a way that NOBODY could tell that it was
compressed using a lossy compression scheme even when ABXed with the original
source material. Unfortunately, MP3s are all compressed using an algorithm
and artifacts do show up and they show up in some kinds of music more than
they show up in others.