Did you try -double ? It sounds like you may be pushing fdelayltv farther than anyone else to date, so that could be where the trouble is. We may need to protect it from small denominators.
- Julius On Tue, Feb 19, 2019 at 2:32 PM Dario Sanfilippo <sanfilippo.da...@gmail.com> wrote: > Hello, list. > > I am experiencing some problems with NANs and I sometimes don't understand > where they come from. > > I have replaced all divisions with the condition that if the divisor is 0, > the output is 0, otherwise it is x1/x2. > > I'm working with a self-oscillating FDN which uses tanh function for > stability processing. There are also 6th-order fdelayltv units and the > other processing in the system is mainly about 1-pole and 1-p-1-z filters > that I have implemented in Faust. So there is arithmetic, some > trigonometric functions and some other math like abs, min, max, sqrt and > pow. > > I've tested the tanh function by putting it within a feedback loop with a > huge feedback coefficient and it simply clips to 1. > > Are the fdalyltv likely to produce NANs is the signal setting the delay is > not within the expected range? > > In your experience, what else is likely to produce NANs? > > Thanks, > Dario > _______________________________________________ > Faudiostream-users mailing list > Faudiostream-users@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/faudiostream-users > -- Julius O. Smith III <j...@ccrma.stanford.edu> Professor of Music and, by courtesy, Electrical Engineering CCRMA, Stanford University http://ccrma.stanford.edu/~jos/ <http://ccrma.stanford.edu/>
_______________________________________________ Faudiostream-users mailing list Faudiostream-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/faudiostream-users
