Hello Everyone,

I'm glad to find this list as I begin using Csound. I've been trying to get linear predictive coding to work in Csound 5.08.  In Windows XP, the coefficient method works fine, but the pole/zero method of lpanal is not very stable and very dependent upon the number of poles and the hop size.  Is this normal?

My real concern is the use of lpanal in Ubuntu Linux.  I used instrument 2709 from the Csound Book.  I'm aware that I have to re-analyze a3.aif due to operating system issues, but no matter which lpanal method I use (-a, -p, -h flags) the output blows up, as if the filter is always unstable.

Is this a known problem?  I changed nothing about 2709.orc or 2709.sco, only the lpanal method and operating system vary.

Thanks!

-Nik Francis

on 2008-11-10 at 15:03 Nik Francis wrote:

>My real concern is the use of lpanal in Ubuntu Linux.  I used
>instrument 2709 from the Csound Book.  I'm aware that I have to
>re-analyze a3.aif due to operating system issues, but no matter which
>lpanal method I use (-a, -p, -h flags) the output blows up, as if the
>filter is always unstable.

¿are you using the doubles version of csound? try the floats version.
the implementation of LPC in csound is not particularly good to begin
with, but in the doubles version it's completely broken. the assumption
of int32 is hard-coded in some parts of the source code.

best,

lj