[vorbis] Vorbis determined to be as good as MPC at 128 kbps!

[Stephen So]

John did the first merge and I did the second merge independently, and 
both of us compared and found out it was exactly the same (in terms of 
code).  Garf's tunings did not conflict with 1.0.1 except in one 
instance with the noiseguard values, where there was a change between 
1.0 and 1.0.1.  We decided that they probably would help in terms of 
quality so we left it that way (kept the 1.0.1 values).  Of course, that 
was just an assumption but they were very minor changes (4, 4, 
somethign) to (3, 3, something).  Since John and I worked independently 
and got to the same code, then the margin for error is quite low, I think.

Steve.

owner-vorbis at xiph.org wrote:

>From: "Age Bosma" <agebosma at home.nl>
>  
>
>>>A few months back when somebody tried to merge oggenc 1.0.1 with his
>>>tunings, I think he made a few comments about the tunings he did, but I
>>>don't think he did any coding or helping.  (Eventually that project was
>>>abandoned, but that wasn't due to Garf.)
>>>
>>>      
>>>
>>The project wasn't abandoned, they succeeded merging the two which
>>resulted into GT3b2 (http://www.rarewares.org/ogg.html).
>>    
>>
>
>They tried on two seperate occasions, I think.
>
>They did succeed in producing binaries that mostly worked, but if I remember
>right, in both cases it wouldn't generate output files that were bit-for-bit
>identical with the original unmerged versions.
>
>In other words, in you used a Q=3 setting, the output would be slightly
>different from what the official 1.01 oggenc would give.  At that setting it
>should have been identical.
>
>If you used Q=6, the output would be slightly different from what Garf's
>tunings would give.  And at that setting, it too should have been identical.
>
>They would sound basically the same, but the output wasn't exactly what they
>were expecting.  And they couldn't explain why.
>
>Also, I'm not sure they came up with an decent solution to what happens when
>you use a fractional quality value between 1.01 tables and the Garf tuned
>tables.  In that case, the encoding was influenced by both.  Garf was tuned
>against an earlier version and apparently it didn't match as well with
>v1.01.  (Or some such.  I'm not an encoder developer.)
>
>The last I heard, although they did produce working binaries (on several
>occasions), they never got to a point where they were satisfied with the
>results.
>
>Maybe I missed them finally succeeding, but the last I heard, they gave up
>but were still going to distribute the binaries (in case anybody wanted
>them) but indicate them as being "not recommended".  It sounded okay, but
>just didn't give the output file that was expected.
>
>
>
>--- >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.
>
>  
>


-- 
-------------------------------------------------
Stephen So PhD Student
Signal Processing Laboratory
School of Microelectronic Engineering
Faculty of Engineering and Information Technology
Nathan Campus Griffith University
Brisbane QLD Australia 4111.

Phone: +61-7-387 53754
Fax:   +61-7-387 55198
E-mail: s.so at griffith.edu.au
-------------------------------------------------

--- >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.