Filter instability?
-
So I’ve roughly modeled a JUNO-6 filter in Faust but comparing to the original shows much more instabilities than its software version
Just introducing random lfnoise on the cutoff frequency seems fishy so I’m wondering how you’d approach this on a more professional level. Also this only appears at maximum resonance.
https://instagram.com/morphoice - 80s inspired Synthwave Music, Arcade & Gameboy homebrew!
-
@Morphoice Sounds great so far!
-
@Morphoice said in Filter instability?:
Just introducing random lfnoise on the cutoff frequency seems fishy
It seems fishy because you didn't mention smoothing.
-
@aaronventure said in Filter instability?:
@Morphoice said in Filter instability?:
Just introducing random lfnoise on the cutoff frequency seems fishy
It seems fishy because you didn't mention smoothing.
? whatever do you mean by that?
-
Not sure I can help: my approach would've been component modelling. That's probably one of the easiest ways to get the instability implicitly. Modelling that kind of thing manually can get complex if accuracy is the goal. I think at somepoint you hit limits with what you can do with simpler dsp
-
@griffinboy that OTA filter has wuite a shitload of transistors I doubt component modeling would be cpu effective, I have the schematics somewhere. Besides the instability is usually imperfections in the material or some sort of transistor drift.
Eg at close to max resonance value the self oscillation changes pitch downward for no reason at all, i can only measure it and define it in a table or sth
At max cutoff self oscillation just vanishes but thats probably just it going iut of hearing range I have yet to scope it
Filters in code dont self resonate unless you inject some sort of signal to start with, so I’ll pribably have to inject a sine at some point
https://instagram.com/morphoice - 80s inspired Synthwave Music, Arcade & Gameboy homebrew!
-
True. A state space model is probably ideal for such a filter.
These things are a pain to derive though. -
@griffinboy yeah probably not a doable approach for my budget
-
@griffinboy i dont even think faust can do three dimensional tables
-
@Morphoice because it'll get the job done, you just might need to also apply smoothing to the noise osc to get the correct curve.
Instead of using just noise, there's a faust function for outputting a random value at a given interval, so check that out maybe. I forgot what it's called, and I'm on the phone.
-
@aaronventure yes thats actually the one I was referring to it should be smooth already.
