Sean Costello wrote:

> What sort of filters? Maybe higher-order filters might need this sort of precision, but it seems
> like these would be better achieved using cascaded second-order sections.

Well I think the problem comes when one implements slightly unstable recursive filters. Believe it or
not they´re necessary in the modelling of waveguide bores.

> More importantly: Which opcodes need to have the higher precision? As someone pointed out, the
> inputs and outputs are always floats, so would this create a problem? All the internal precision in
> the world won't help if it gets rounded down upon exiting the opcode. Do you need higher precision
> in filters, delay lines, oscillators, or...?

Well filters must use the highest numerical precission available IMHO. Also all the mathematic
operations, data passed internally of course and interpolating delay lines. Maybe with sound generators
it could be user selectable. But notice, for example, that Buzz evaluates a closed form expression
which divides eventually by a very small quantity and the associated numerical errors can be greatly
reduced with higher precission opcodes. I think the current implementation is not very optimal in this
respect. Even if only the filter routines use more precission but the output is rounded, it should make
a big difference with those unstability problems!
Hey guys, it´s time to work and put all this on practice! Any volunteers? ;-)

--
Josep M Comajuncosas
C/ Circumval.lacio 75  08790 Gelida - Penedes
Catalunya - SPAIN
home phone : 93 7792243 / 00 34 3 7792243

Csound page at http://members.tripod.com/csound/