View Single Post
Old 01-18-2019, 06:05 AM   #106
Human being with feelings
sai'ke's Avatar
Join Date: Aug 2009
Location: Germoney
Posts: 330

Argh. I wrote a rather large reply, but REAPER forum's autologout feature just tanked it. I dislike that it doesn't store what you typed if you are logged out ;__;.

I'll try to rewrite more briefly what I have written before.

1. Stability and self-oscillation (oscillation without input) are two separate things. Although they're related, when I speak about instability, I mean cases where the sound just stops. The signal goes to positive or negative infinity and that's it. These cases are the ones that should really be 'fixed' if they are still there.

2. Self-oscillation is not necessarily an undesirable property to me. Self-oscillation happens in analog gear as well. What I could make is a panic shortcut on the keyboard that resets the filter state (if this would help). Currently, what you can do is just set it to that filter again. Any filter change (even to the same one) will reset the internal filter state.

3. You can clip the output with the output clipper to protect your ears and equipment. Once you settle in on a tone, turn it off. The clipping I was talking about in the filter context to avoid stability has nothing to do with output clipping. Instability arises when the feedback starts to increase the signal indefinitely. (Soft)-clipping the feedback can often tame this behaviour, and actually, in a lot of analog circuits, feedback clipping is a thing (diode clippers are often used).

4. Ducking might be possible, and could be based on the RMS estimate of the input. It'd always introduce some weird transient effects though and is probably not that useful (over say clipping) since you wouldn't want ducking effects in your actual audio.

5. Some synths / filter designs lower the output gain when increasing the resonance. This is easy to do in a linear filter, but becomes complicated when there are non-linear elements in the filter. For cases where there is only a fixed non-linearity, one can make input/output curves and correct for gain increases. However, considering that there is a completely freely selectable non-linear elements in a fair few of these filters (the spline waveshaper), it becomes impossible to calculate such a gain compensation analytically for quite a few of these filters. Flexibility comes at a price.

6. Resonance should go down again (even in the case of self oscillation) if you lower the resonance. In some cases, you may have to lower it a lot (potentially even zero if you're using the spline wave-shaper). If you can give me a case where it doesn't, I'd be interested in studying it.

7. I can see one thing that you may want to avoid if you don't like loud bitey resonance; namely high slopes in the spline waveshaper. Large slopes there could lead to a high gain for some of the filters where the waveshaper is in the feedback mechanism (such as the SWR ones). Ending the curve with a high slope could even lead to instability if it wasn't for feedback clipping.

I am planning on making a testfile where I run the filter through various types of audio (noise, low freq, high freq, bandpass sections,
nasty squares etc) to do some more testing. If you have audio clip and setting combinations where you spot misbehaviour, please let me know.
They'd be useful as case studies.

As for 'misbehavior' detection, by the time you have enough data to detect it, you're already too late. :P
[Tunes] | [Tracker Plugin: Thread|Github|Reapack] | [Routing Plugin: Thread|Github|Reapack] | [Filther: Thread|Github|Reapack]
sai'ke is offline   Reply With Quote