Those analysis window sizes (and 50% overlap - the absolute minimum 
possible) are rather small - have you tried larger ones (e.g. 1024, 256, 
1024 or 2048)? Unless the source pitch is fairly high (say over 400Hz), 
any pitch-tracker will struggle to resolve things, as the whole bass 
register is covered by only a handful of bins.

Richard Dobson


Bruce McKinney wrote:
> I have been attempting to call a diskgrain instrument in real-time from 
> an instrument that
> uses the pvsanal and pvspitch opcodes to analyse an incoming audio 
> signal and then
> eventually send it to the called instrument using diskgrain. In the 
> example below, I
> have simplified everything in my attempt to find out what is not 
> working. The called instrument,
> i 5, uses the CSound manual example verbatim, which works--obviously, I 
> have substituted my
> own sound file. However, when I include the line of code :
> 
> k1, k2 pvspitch f1, kthresh; in the calling instrument, the diskgrain 
> output includes
> a second invariant pitch. The pitch changes based on my choices for 
> ksmps or overlap
> in the pvsanal, but I can not get rid of it.
> 
> Is there some way in which it is not possible to use these two opcodes 
> in this manner,
> or am I doing something stupid!
..
> 
> sr = 44100
> ksmps = 64
> nchnls = 2
..
> f1 pvsanal a98, 512,256,1024,0;; this is bruce !!!
>