[sdiy] Do digital filters still explode?

[Gordon JC Pearce]

On Sun, 2011-02-06 at 04:02 +0000, karl dalen wrote:
> Well, i do remeber the word explode but if it was
> Collin who said it, well i can be completely wrong.
> 
> The statement then included a explosion of overtones
> and undertones, explosion such as in sudden boom of
> aliasing, perhaps something similar to too deep and
> drastic change of modulation index in a FM oscillator
> ,burst of squelchy noise.

Aha, I know this one.

I had a crack at modifying xsynth-dssi to include a filter crossmod
control, where I mixed the output of one oscillator with the filter
cutoff.  It didn't really work in a musically useful way.

Then I realised something.  I was taking the oscillator value at the end
of the jack period and mixing that with the filter cutoff - so what I
was actually doing was modulating it with the "VCO" output *passed
through a sample-and-hold at samplerate/periodsize Hz* or given my
typical settings about 180Hz!  Oops.  No wonder it sounded hellish.

Rewriting the voice callback so it calculated a periodsize-length array
of filter cutoff values and worked out the coefficients for each frame
in the period really improved matters, and brought the CEM3394-like
noise I was looking for.

Gordon MM0YEQ