[vorbis] Ogg Vorbis Quality Analisis (+ bug) (+ misc)

AGS sound_AT_sinderman.com at spamm.off
Wed Jan 16 09:22:25 PST 2002



I have run an ogg analisis using EAQUAL from -q 0 till
-q 10 evaluating bitrate and quality. The results are
shown in a nice graph, and can be seen here:

http://audio.sinderman.com/

One thing I have found is a bug, as it is cleary shown
by the blue lines. Other discoveries are written in that
page, and others are still evolving in my mind :)

The goals of this analisis is to help the developers
(if this is help at all for them), an to help people decide
which  -q is most suitable for them.

But I have some more questions:

1) RC3 files can be peeled. Is peelability going to further improved?

2) Is there a cost for peeling instead of encoding directly with lower
quality? I now is less than reencoding but, is negletable?

3) What happens with a file encoded with lossless channel coupling when
it is peeled to a quality with lossy coupling?

4) Can next version (I understand there will be a kind of RC3b before RC4
with the -b removed) have a (hidden?) option for turning on lossless channel
coupling at different q values for testing?

5) We are trying to educate people to think in quality terms instead of
bitrate. But q 5 for most of the world means nothing, and 128 or 256 quiet
a lot. Shouldn't we do ie as Microsoft decided (probably after spending
millions of dollars in research :) to make it (also) as simple as "radio
quality" "almost CD quality" "CD quality" etc?

6) What about an idependent codec table comparison? Because if we take 5)
and  said -q 6 is almost CD quality when Microsoft comes with its shitty
codec claiming also almost CD quality with half of the size will not be
fair. A common messurement of CD (and others) quality should be used.
EAQUAL? Eventhough it is prune to errors, it is independent :), and as an
average comparison sould be quiet secure.

7) What about building a FAQ? :) (Developers please ignore, you have other
priorities) I talk exclusively to frecuent people in the lists. Any
volunteers? I have seen some questions coming once and again all the time.
I hope I am not one of those :)

8) I would like to know also what do you think about the issues I have
rised in the URL mentioned above. Or if the graph is showing anyone further
facts.

9) Do you think further graphs could be useful for developers/users? I am
open for comments.

<p>                        Well, thats all. Too long for my forst post.
                        Cheers, AGS.

<p>--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.



More information about the Vorbis mailing list