Vorbis dev - May 2000 - Branch stuff

I've been playing with the current Monty branch and I've found a case
where it performs poorly. I havn't had a chance to verify the mainline
(network at home isn't working; System with good headphones isn't on
net,
and I've only sneakernetted the branch)..

The gspi35_1 sounds like it's suffering from preecho. I changed the
preecho threshold to make more shortblocks, but it sounds worse. I had
believed that short blocks were fixed in hte branch..

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/

> 
> I've been playing with the current Monty branch and I've found a
case
> where it performs poorly. I havn't had a chance to verify the mainline
> (network at home isn't working; System with good headphones isn't
on net,
> and I've only sneakernetted the branch)..
The current monty branch is using experimental 'B' codebooks, and I
found
during verification last night that these codebooks (although they work) are
generated very suboptimally (ie, latticepare.c's paring heuristics are
broken
almost to worst case ;-)

If you replace the 'b' codebooks with the 'a' codebooks, you
should get better
results (in modes.h, replace occurences of 'res0b' with
'res0a').

Now, if *that*'s having trouble, I definately want to hear it.  I don't
think I
have a copy of the test sample you mention.

Monty

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/