Vorbis - Sep 2002 - c't listening test: Ogg problem at 128kbps

Hi there,

I'm new to this list, and in fact rather new to the Ogg/Vorbis codec
altogether. I recently downloaded the test files from the test
conducted by the German computer magazine c't (this was mentioned by
someone here recently).

The idea was that you were given seven WAV files each for 64kbps and
128kbps, and had to rate them. The files were randomly chosen for each
participant. Also supplied was an (another) uncompressed sample for
reference. Now that the test is over, I got an email telling me which
file corresponds to which codec. Here are my results:

At 64kbps, Ogg turned out to be the best, along with the uncompressed
sample (the difference was negligible compared to the other codecs --
all others were lots worse). 64kbps was rather easy, because some
(MP3, MP3Pro, Real) had frequency response problems, AAC had
background noise, and WMA lacked impulse response. So I don't think I
should be concerned about my hearing...

At 128kbps, however, most of the frequency response problems were gone
-- except this time, I picked Ogg as the worst, based on an overly
exacerbated treble. Others were still noisy and lacked impulse detail,
but Ogg/Vorbis was the one I decided first was definitely not close to
the original. Listening to it again, I can confirm the difference I
heard, although maybe I wouldn't put it in last place this time.

The passage I based the Ogg/Vorbis 128bps decision on is from "Love at
First Sight" by Kylie, and it's simply so that the high percussions
seem to have too long a decay time compared to the original. It's thus
probably not a frequency response problem but effectively sounds like
one.

I tested Ogg/Vorbis again simply by encoding the reference file with
oggenc from 1.0, at default quality. It has exactly the same
problem. If I go to quality 10 (I haven't tried anything in between),
the problem vanishes.

If you like, I can put the reference file on the web (assuming it's
copyright-cleared, given that c't has distributed them to basically
anyone), so that you can try encoding and listening for yourself.

Personally, I found the c't test files badly chosen -- the samples
were from the mentioned pop song (compressed to basically constant
envelope), some slow jazz, and an excerpt from an opera. No music with
"stereo effect", which would have revealed many problems with a number
of codecs, large dynamics, or music that is really critical in all
respects like piano solo. I would have had so many suggestions...

Please let me know what you think and whether I shall upload the
file. I'm concerned about this, it's probably the only weak point of
Ogg/Vorbis at the moment.

  Andras

==========================================================================Major
Andras
    e-mail: andras@users.sourceforge.net
    www:    http://andras.webhop.org/
==========================================================================---
>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@xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is
needed.
Unsubscribe messages sent to the list will be ignored/filtered.

> -----Original Message-----
> From:	Major A [SMTP:andras@users.sourceforge.net]
> Sent:	Tuesday, September 03, 2002 4:41 AM
> To:	vorbis@xiph.org
> Subject:	[vorbis] c't listening test: Ogg problem at 128kbps
> 
> At 128kbps, however, most of the frequency response problems were gone
> -- except this time, I picked Ogg as the worst, based on an overly
> exacerbated treble. Others were still noisy and lacked impulse detail,
> but Ogg/Vorbis was the one I decided first was definitely not close to
> the original. Listening to it again, I can confirm the difference I
> heard, although maybe I wouldn't put it in last place this time.
> 
> The passage I based the Ogg/Vorbis 128bps decision on is from "Love at
> First Sight" by Kylie, and it's simply so that the high
percussions
> seem to have too long a decay time compared to the original. It's thus
> probably not a frequency response problem but effectively sounds like
> one.
> 
I've listened last weekend to the test files from the vorbis
homepage on a friend's hifi installation and our conclusion
was that the Vorbis 128kbps files sounded less "flat" than
the original file as if it tended to overamplify slightly
high freqs.

André
--- >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@xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is
needed.
Unsubscribe messages sent to the list will be ignored/filtered.

> At 128kbps, however, most of the frequency response problems were gone
> -- except this time, I picked Ogg as the worst, based on an overly
> exacerbated treble. Others were still noisy and lacked impulse detail,
> but Ogg/Vorbis was the one I decided first was definitely not close to
> the original. Listening to it again, I can confirm the difference I
> heard, although maybe I wouldn't put it in last place this time.
It's caused by the lossless coupling at hi frqs and the effect is
stronger if you listen the music with headphones. I don't know if
it's possible to fix the problem without removing lossless coupling,
but the problem should be addressed in some way, because most music
this day have that kind of sound.
> If you like, I can put the reference file on the web (assuming it's
> copyright-cleared, given that c't has distributed them to basically
> anyone), so that you can try encoding and listening for yourself.
If you can, I'm curious to try.
> Please let me know what you think and whether I shall upload the
> file. I'm concerned about this, it's probably the only weak point
of
> Ogg/Vorbis at the moment.
Yes, I think so.

<p>Bye.

--- >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@xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is
needed.
Unsubscribe messages sent to the list will be ignored/filtered.