Clicky vs your mom
category: general [glöplog]
So I'm working on the software synth and I have this neat fm-synth in it. I've found that changing the fm-synths configuration while in sustain sound pretty awesome. Problem is I get a 'click' sound half the time because the waveform has an ugly cut in it. You know what I'm talking about? So I'm wondering how I should go about solving this. It's waveform is a little too complex to figure out how much to shift the phase until there is no click, so I don't think I'm going to do that. I think what I'll do is store the last sample out and when the synth's configuration is changed clear a counter that will be used to blend from the last sample since the change and the new waveform. Still, I think that's a little too complex. Does anyone have any nice tips for getting rid of or avoiding the "clickies"
fade down the amplitude to the operator disappearing and fade up the new operator appearing, in a short time? (just a random idea)
generally, just avoid changing values drastically from one sample to another. a 100-sample interpolation is usually good enough.
gargaj, he wants to change the synth topology, not the parameters...
Random ideas (from somebody who's never touched any synth code):
- add a filter to remove the worst of the click. You'll probably lose high-end, and it won't be perfect, but it's better than a loud click.
- interpolate between old/new states like gargaj said
- hold the new waveform until the amplitude + direction of the old waveform roughly matches the start of the new one
- add a filter to remove the worst of the click. You'll probably lose high-end, and it won't be perfect, but it's better than a loud click.
- interpolate between old/new states like gargaj said
- hold the new waveform until the amplitude + direction of the old waveform roughly matches the start of the new one
ahh, hold a wave form. That's interesting idea... hmm. I don't think I'll do that but I'll certainly keep that idea in mind, that's interesting
gargaj, 100-sample interpolation is usually good? Ok, thanks
gargaj, 100-sample interpolation is usually good? Ok, thanks
ok.. yeah that works. I'm interpolating 50 samples and I find, for my use, that's enough. Adds a little bit of code but overall not too much
go fekk a demo about it