[vorbis] Archival quality for music

[John Morton]

On Fri, 02 Aug 2002 05:47, Olaf Marzocchi wrote:
> This mail depends upon the fact that I don't have a couple of good
> earphones ;-)
>
> I read in the site that q=6 is a very high quality, but does it contain
> perceivable differencies from the original? (for 95% of people, of course).

I can hear differences between original and encoded applause segments at q6 
that go away at q7, but no one else on the list has claimed to hear what I'm 
hearing, so I expect I'm the 5% minority for that particular kind of 
artifact. For anything else, something over q4 is fine. 

What this illustrates is that even if you belive the claim that q6 is good 
enough for 95% of people (and there's not substantial evidence for this, 
though intuition would suggest somewhere in the region of q4-5 is more 
accurate), you still need to find out where in that 95% you are, and for what 
sort of music.

> I also found q=6 to produce files slightly bigger (1/10 bigger) than those
> produced with lame VBR q=2 (about 192 bps on average). I always thought
> LAME VBR q=2 produced files with very little (if audible) differencies. May
> OGG obtain better results than mp3 only at low bitrates? (I don't think so,
> but I don't have enough skills to be sure).
> I also think that ogg q=6 is the minimum quality level to use with ogg
> because it has a better way to encode stereo than q=5.x.

I believe lame's VBR mode tends to use joint stereo, while vorbis switchs to 
a lossless channel coupling method over q5-q6 region - this may account for 
the difference in size. I haven't heard of anyone claiming that lame vbr 2 is
indistinquishable, quality-wise from the original, though that claim might be 
made for whateve the --r3mix switch sets these days. 

Certainly fatboy.wav sounds is pretty easy to ABX against lame at any vbr 
setting, the last time I checked.

> To summarize, I need an advice about the quality level to set in order to
> achieve comparable or better results than mp3s (LAME) at about 192 kbps or,
> as preferred alternative, to produce files with very little perceivable
> differencies form originals.
> If this value has a nominal bitrate of more than 256 kbps please give me an
> idea of the level of distorctions at 192 kbps nominal (OGG of course).

I'm going to use q7 for 'hard to encode' stuff, which, for me is just a few 
live ablums, right now, and q4.99 for everything else - I can't really ABX 
anything past q4, but I'm adding a bit to account for future improvements in 
my hearing and equipment, and staying shy of 5.00 as there's still a large 
bitrate hop between 4.99 and 5.00. 

I suggest you do some ABX testing against the usual set of difficult to 
encode samples, and music you're familar with at various different quality 
levels to find one or two that suit you. If they happen to produce files with 
a better ABR than lame, use ogg, otherwise stick with lame. 

John

<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.