On 22 Dec 2003 02:46:50 -0800
(KikeG) wrote:

Ian Molton wrote in message
...

Pure nonscience.

yeah right.

if the average for the tracks is 128kbit, the avergae for the album
is 128kbit and vice-versa

But the samples used at the test are not even full tracks, are very
short pieces of music somewhat hard to encode. Full tracks will
average closer to 128 kbps. Go try yourself.

You dont know they would be. the rest of the tracks may be substantially
easier or harder to encode, and thus the average bitrate would eb
significantly lower or higher.

The target of VBR codecs is to
maintain quality, dynamically increasing bitrate if there's need for
it. 128 Kbps is the AVERAGE bitrate.

Again, you have missed the point. VBR does NOT mean the average bitrate
is constant over a nominal period. It simply means the bitrate is not
constant, and there are (at least) two strategies in common use:

1) the one you (seem to be) describing, quality based VBR, where the
bitrate is the lowest possible to maintain a constant 'q' factor
(quality) throughout the track. This type does indeed use less bits for
easy areas and more on demand.

2) AVERAGE based VBR, which does the same as above, but keeps the
average, over a noominal period, as close to a specified value as
possible. thus any easy to encode parts of the track 'waste' bits, and
difficult parts may not get enough bits, or potentially can starve bits
from nearby portions of the track.

Strategy 2 is used where streaming is being used and a constant rate
bitstream is needed. Strategy 1 is used to keep filesize down without
compromising quality.

This bitrate was calculated averaging the result of a large number of
albums.

Assuming they were all encoded as VBR then whoever was doing the
encoding did an apalling job of matching the quality settings.

Nothing to to with CBR, where 128 Kbps is a CONSTANT bitrate.

Duh.

Please inform yourself about what VBR means.

I think its you that needs informing.

either compare all at the same average bitrate for a given piece, or
dont bother at all.

You are wrong, that would be highly unrealistic. If you can't
understand this, there's little I can do to help you.

If you had two hypothetical 'mp3 type' encoders that can do fixed and
variable bitrate, you should be comparing them at a fixed bitrate.
Variable bitrate encoding should be done by selecting the correct
bitrate for a 'frame' in the track based on quality, and thus in a fair
test, all encoders should use the same quality model for determining
quality.

--
Spyros lair: http://www.mnementh.co.uk/ |||| Maintainer: arm26 linux

Do not meddle in the affairs of Dragons, for you are tasty and good with
ketchup.