[vorbis] APPLAUD.WAV problems

[Bostjan]

<<<<<<<<
----- Original Message -----
From: "Gian-Carlo Pascutto" <gcp at sjeng.org>
To: <vorbis at xiph.org>
Sent: Thursday, February 14, 2002 5:01 PM
Subject: Re: [vorbis] APPLAUD.WAV problems
> The ironic fact is that Gian-Carlo seems to have noticed this
> lossy-stereo-coding problem already and has tuned his OGG encoders to
> eliminate it.
I didn't eliminate the lossy-stereo-coding problem; I eliminated the
lossy stereo modes :)
>>>>>>>>

It isn't at all a bad idea. Is it possible to add a switch to oggenc that
eliminates the lossy stereo modes?

<<<<<<
Several fixes went in after RC2 to fix the stereo issues, and as far
as we were able to verify back then, they fixed the problem clips
we had at that point.
If LAME is doing better at the same bitrate, that might indicate there
is still a problem somewhere.
>>>>>>>>>

Yes. I don't use lame otherwise, but the difference is HUGE. Please, take a
moment to hear by yourself. Try to encode  applaud.wav with lame at 128 kbit
CBR and compare it with oggenc -q2 (which produces a file of similar size).
The ogg file has A LOT of high freq. noise added (details are described
earlier in this thread).

<<<<<<<
Then again, applaud is a known problem
case for encoders, and it might just as well be the encoder is doing
things perfectly everywhere else. I don't really like getting impressions
by using that kind of clips, as I think they are giving a skewed view.
>>>>>>>>

I agree, but many songs contain applauds and a premium encoder such as OGG
shouldn't produce such CLEARLY EVIDENT artifacts.

<<<<<<<
On the other hand, they are good for doing regression testing or for
checking for known pitfalls. But they are not fit to make any kind of
conclusion about general quality.

What I'm trying to say is that Ogg doing badly on applaud _doesn't_
mean there is a problem with stereo encoding or that Ogg is boosting
high frequencies at levels lower than -q6. You can't tell that from _one_
problem case.
>>>>>>>

You are right. But if a person hears such an evident artefact on a common
sound like an applaud, he would think again before re-using the encoder
below -q 5.

<<<<<<<
 And from the fact that you've used a known problem clip
to come to report it leads me to believe it probably isn't, since you would
have detected it on your own music already wouldn't you?
>>>>>>>>

You are right again. I consider OGG to be an excellent performer,
otherwise I wouldn't have chosen it for my music collection. I wrote this
message just to make sure that Monty (and others) know of this problem.

<<<<<<
I'm not trying to downplay that RC3 might have problems on applaud,
I'm just writing this overly long mail because yours left (at least for me)
the impression that GT2 at 160kbps gives better results than RC3 -q5.
I certainly doub't that is true for the general case, see my answer to
the next question.
>>>>>>

I'm sorry to have created confusion. What I actually ment was that GT2 -b
128 compresses APPLAUD.WAV better than RC3 oggenc -q 4.9 at the same
bitrate.

<<<<<<<<
(I haven't done any rigid listening tests of Ogg RC3 at less than -q6,
so I'll refrain from giving any impression about quality or possible
problems at those levels myself.)

> Gian Carlo, would you mind explaining how did you tune your encoders - at
> least to the guys that hold on this project? I believe that your
suggestions
> would help OGG making one step ahead towards being the best lossy encoder
> available.

This already happened - Monty has copies of my modes and I have discussed
the differences with him. RC3 incorporates the improvements in GT2. However,
I have also noticed that some of the problem clips for RC2 that GT2 solved
are again problem cases for RC3. That is unexpected - I've already mailed
a few of those clips to Monty, and I'd like to discuss what might be the
cause.
>>>>>>>>>

Thank you for you effort. I really enjoy using OGG and I often reccomend it
to my friends. I posted this just to report a flaw and help OGG become
better.

Regards,
Bostjan

<p><p><p><p><p><p><p><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.