[Arny Krueger]

"Sonnova" wrote in message

On Mon, 10 Dec 2007 15:41:10 -0800, Arny Krueger wrote
(in article ):

"MC" wrote in message

"Doug McDonald" wrote in
message ...
Arny Krueger wrote:

It is true that the essence of perceptual coding is
eliminating tones that will not be heard anyway,

This is only true at very low, very poor quality,
bitrates, and possibly for very low level high
frequency content.

For good encoders (such as LAME VBR at = 160K) no
"tones are left out". The only "perceptual" feature
is how the encoder distributes highly colored noise.
At very high bitrates the noise becomes pretty much
high frequiency rolled off white noise.

I tested this with the FHG pro MP3 coder and found that
the essence of the loss due to lossy coding was
something like a loss of dynamic range. Sort of what
like what analog tape or the LP format do all the time.
Even the quantities were similar. The rise in background
noise looked something like pink noise.

And I'll bet that it rides the envelope of the music as
the dynamics change, doesn't it, causing a nasty
breathing sound?

Not at all. Not even 0.1 dB changes.

I did some tone burst tests, and there was no trace of dynamics compression.
All tone burst levels in the reconstructed wave were within 0.1 dB of the
originals.

In fact I suppose it can be said that a high proportion
of the "fine detail," in either images or sound, is
actually noise and can be dispensed with. Right?

In moderation, either the ear or eye can be fooled into
perceiving no loss of details.

As long as that same sensory apparatus doesn't notice or
mind the artifacts which accompany "being fooled", I'm
sure that's true.

When the artifacts are below the thresholds of sensation, there is no choice
about being fooled. You are simply fooled, and blind tests make that very
clear when it is true. With sighted evaluations, people can be fooled by
their expectations, and think they are hearing what they can't hear at all.