VOICE Home Page: http://www.os2voice.org
[Previous Page] [Next Page]
[Features]
|
VOICE Home Page: http://www.os2voice.org |
[Previous Page] [Next Page] [Features] |
|
By Christian Hennecke ©March 2001 |
You may ask why one should use Ogg Vorbis when there are already MP3 and others
around, with MP3 being a de facto standard. Well, Xiphophorus.Org,
the creators of Ogg, have designed it to be completely free and Open Source with
parts released under GPL and parts under LPGL. First, there are no royalties to
pay for using it and that's a big advantage over MP3. In case you don't know: The
Fraunhofer institute holds patents for MP3 in at least the US and Germany. While
nobody really seems interested in charging private users for using free MP3 encoders
and decoders, commercial users have to pay, e.g., for including MP3 sound with their
games or offering music on their site. But this may change in the future. Just think
of Unisys and the LZW patent. Work is already under way to include Ogg Vorbis support
into the ICECAST streaming server. And second, Ogg Vorbis is designed to offer better
quality. More on that later.
At the moment Brian Havard's site
is the one to go if you are interested in obtaining Ogg Vorbis for OS/2. Always
be sure to get the latest release, since beta 4 offers huge improvements over beta
3 (and beta 5 in its time is going to offer a lot more than beta 4). Also keep in
mind that the original plug-in was based on beta 2 and so it probably won't play
files created by later versions properly (later versions are backwards compatible
though).
At the moment six different encoding modes are supported. Each of these is a
full variable bitrate (VBR) mode, though they are a bit different in that you select
the average desired bitrate. If you think of the LAME MP3 encoder, the result
would be that of LAME's --abr option, not
the -v option. The available modes are 112,
128, 160, 192, 256, and 350 kbps.
Another feature is support for tags and naming. Using commandline options you
can specify a comment, the date, track number, title, album and artist to include
in the file. The naming feature let's you use this data for automatically naming
the resulting file(s).
The following tables show the results for encoding the tracks with OggEnc from
Vorbis beta 4 and LAME 3.86 running on an AMD K6-III 450. Options used for LAME
were: -ms --abr <bitrate>
Tab.1: Results for encoding track 1 (playing time: 9:24 min,
size: 97.238kB)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Tab.2: Results for encoding track 2 (playing time: 6:09 min,
size: 63.607kB)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
While OggEnc produces files of about the same as similar MP3s, it takes much
longer to accomplish the job than LAME. Currently that is. As I pointed out before,
Vorbis is still in beta phase and therefore it hasn't been optimised in terms of
CPU usage.
To play Ogg Vorbis encoded files you can use the commandline utilities from the
Vorbis distribution or PM123 with the above mentioned plug-in. Generally speaking,
playing Ogg Vorbis files imposes more CPU load on your system than playing MP3s.
However, significant improvements can be expected. This already shows in Brian Havard's
tweaked version of Sofiya's plug-in for PM123. While the latter created a load of
about 17-18% on my system with the analyzer turned off, the latest version has arrived
at 15%. The commandline tools perform slightly better at 14% load. Compared to a
rate of 11-12% for MP3s on PM123 this looks very promising. (Note: Other MP3 players
for OS/2 like WarpAMP, z! (both 10% load) or the QU player (8% load) are less resource-hungry,
but aren't suitable for comparison here.)
Of course all the above would be completely obsolete if the sound wasn't good.
My machine is connected to a nice stereo and I cross-checked the results of Ogg
Vorbis and MP3 against the original, all played via a Terratec Maestro 32/96 soundcard.
I did the MP3 encoding with LAME, because it is considered as the best MP3 encoder
around, meanwhile better than Fraunhofer's original MP3ENC.
As it turned out, an average bitrate of 128 kbps is not the way to go if you
really want to listen to the music, neither with MP3 nor Ogg Vorbis. However,
I think that Ogg Vorbis produces the better result, if you use an up to date version
- the old ones from last year showed some audible artifacts. Choosing an average
bitrate of 256 kbps gives much better results in both cases. To me Ogg Vorbis sounds
superior to MP3 at 256 kbps, too, though the difference is smaller than at 128 kbps.
Generally the following can be observed: MP3 seems to literally compress the
sound more and adds a certain harshness while taking away more details. This especially
showed if I turned my attention to the celli and the flute in the tracks. The celli's
snaring and the sound of the flute's mechanics was more present with Ogg Vorbis.
MP3 made the singer sound more strained than on CD. Ogg Vorbis seems to preserve
more room, i.e. instruments are not only located in the plane between the speakers,
but also in the space behind.
Please note that these descriptions are somewhat exaggerated to clarify
the differences. MP3 is not as bad as the above sentences "sound". Nethertheless
I think that Vorbis is better, but the difference is not that huge.
If you are interested, always be sure to get beta 4 binaries and beta 5 when
it's available, since they offer huge improvements over older versions.
Hopefully we'll see more players for OS/2 with Ogg Vorbis support soon. I have
asked Dink about including it in z!, but he's not interested in doing so. Of course
a plug-in for MMOS2 would be nice, since it would enable all MMOS2 players to use
this format.
Article References:Ogg Vorbis homepage: http://www.vorbis.com |