Dear list, Here; http://electro-music.com/forum/viewtopic.php?t=37921 is a discussion on instability in the "convenience filters", particularly LPF. It seems like LPF sometimes goes unstable while I feel that a filter with this purpose (quick&easy filtering with no need to do z-plane transforms ourselves) shouldn't go unstable. Could rounding errors in 64bit floats really cause issues this bad? Are we verifying that the coefficients we get are indeed stable after rounding? That last one is what I'd bet on as a cause for this, but I couldn't do a z-plane transform to save my life. If one of our residential DSP wizards could kindly have a look at that? Maybe Perry? Yours, Kas.
2009/11/11 Kassen
Dear list,
Here; http://electro-music.com/forum/viewtopic.php?t=37921 is a discussion on instability in the "convenience filters", particularly LPF. It seems like LPF sometimes goes unstable while I feel that a filter with this purpose (quick&easy filtering with no need to do z-plane transforms ourselves) shouldn't go unstable. Could rounding errors in 64bit floats really cause issues this bad? Are we verifying that the coefficients we get are indeed stable after rounding? That last one is what I'd bet on as a cause for this, but I couldn't do a z-plane transform to save my life.
If one of our residential DSP wizards could kindly have a look at that? Maybe Perry?
Yours, Kas.
I wonder if ChucK has done the same things with respect to denormalization that had to be done in the Ardour/LADSPA world a year or two ago. They were having many problems with things like filters and other LADSPA plugins that are reminiscent of this description. Seems like they went through a big round of fixes, caught most of them and there haven't been many problems since. Cheers, Mark
participants (2)
-
Kassen -
Mark Knecht