ive been stuck on a few aliasing issues, so to keep making some progress i went back to look at the low frequency bump. at first i didnt think it was too critical, as its subaudible. but today i hit the 303 filter with a slow square wave, and that low frequency resonance is really strong. it gave a 400mVpp ring at 7.5Hz and 1.2s ringdown time. you can see the effect when frequency changes, and for constant notes there can be a slight 7.5Hz AM on the signal, as the sharp edges of the waves re-excite the low end. i also realized i had a decent amount of processing time, so adding the third highpass filter in there wouldnt be an issue. below are plots of the real vs the microcontroller (i have dashed and solid swapped from the graphs directly above - on these dashed is ideal, solid is uC):
there is some slight mismatch, but definitely an improvement over the previous version. i also measured the ringdown of the uC, and set it to 1.2s.
basically, i removed the old single pole highpass (ms/(ms+1)) and added a second order highpass (as^2/(as^2+bs+1)) with a = 0.0156 and b = 0.313, this is approximately an Fc of 8Hz and Q of 0.4. it will probably get tweaked later to deal with the tradeoff of how Q and Fc vary across the various ladder filter cutoff settings.
basically, i removed the old single pole highpass (ms/(ms+1)) and added a second order highpass (as^2/(as^2+bs+1)) with a = 0.0156 and b = 0.313, this is approximately an Fc of 8Hz and Q of 0.4. it will probably get tweaked later to deal with the tradeoff of how Q and Fc vary across the various ladder filter cutoff settings.
Statistics: Posted by guest — Sun Feb 09, 2025 1:14 am