Weird pitch shift when playing filter in ACE

[DorianDay]

I was messing around in ACE, playing the filter and I noticed when I fed back the filter into itself, the note shifts back by a semi-tone. Does anyone know what's going on inside the filter to cause that.

I uploaded the patches, one with feedback, one with out. They're pretty simple, VCF1 -> VCA1, min cutoff, max res, max key following. If you plug the VCF output back into VCF mix and feed it back, the note shifts back a semitone.

I used abletons spectrum analyzer to assess the shift and could hear it.

No Feedback:
http://www50.zippyshare.com/v/93555657/file.html

Feedback:
http://www50.zippyshare.com/v/81574979/file.html

Thanks in advance

[Ch00rD]

Don't ask me for the details, but afaik it's entirely to be expected that pitch changes with feedback at various settings (you can even play scales over multiple octaves this way in Diva with all oscillators at zero volume with some careful parameter modulation ). A perhaps intuitive example is to compare it to the feedback between an electric guitar and an amplifier at high volume, where the distance between the amp speaker and guitar affects the pitch of the feedback.

[Urs]

This is to be expected, I'm afraid. Depending on the way the cables take through the modules there's always going to be a delay of a quarter sample or more. I guess that a really unfortunate routing may cause a delay of up to two samples. Any such delay will have an effect on the tuning of self oscillating loops.

Also, feedback from a filter into itself is typically positive, thus it *lowers* the effect of Resonance within a range of frequencies. In highly non-linear ( = distorting) filters you'll hear that the amount of resonance also has an effect on the tuning of self-oscillation, hence more feedback changes the tuning!

- Urs

[DorianDay]

Thank you both so much for looking at my question and taking the time to answer it.

I have two follow up questions that hopefully Urs will have the patience to answer.

1) The only reason I noticed this phenomenon is because I was trying to deconstruct one of the presets, HS Church Bells, under One Point One. The main sound source are two resonant VCF, tuned to each other using cutoff. Both filters are fed back into themselves, I don't notice that this has any impact on the patch when I drop the feedback loop. As stated earlier in my experiment I noticed the pitch drop due to the feedback loop. Is the feedback loop serving any kind of function I'm not noticing in this patch?

2) ACE is my first foray into modular programming, and I must say I feel like I've entered a whole new, very exciting world. I'm interested in learning more in-depth what is going on inside the synthesizer, and synthesis in general. Can any recommend any literature that would familiarize me with these principles so hopefully I could answer questions like this for myself in the future. I studied math in university, but not much engineering or physics so if I need a primer on any of that a recommendation would be appreciated.

Thank you so much Urs for taking the time to answer questions and for creating this synth in the first place, it's given me new inspiration in the music creation process.

[Howard]

1) The only reason I noticed this phenomenon is because I was trying to deconstruct one of the presets, HS Church Bells, under One Point One. The main sound source are two resonant VCF, tuned to each other using cutoff. Both filters are fed back into themselves, I don't notice that this has any impact on the patch when I drop the feedback loop.

The feedback is just to ensure good self-oscillation without any input (could have used +5V instead). If you remove the feedback cables, a bit of the VCOs is fed into the filters despite setting the Gains to minimum - the signal does get a bit "warbly"!

[DorianDay]

Thank you Howard for the response, that patch and several others make much more sense now.

Urs, today I was messing around with resonance/self-oscilation and I noticed what you mentioned about increasing resonance causing a detuning of the cutoff. I see it has something to do with non-linear/distorting filter models. I was wondering if you could go into more detail on this.

I tried to replicate the situation on several other synths, Sylenth, Strobe, Massive, Arturia Modular and Ohmforces resonant filter Quadfromage. None of these had this resonance/detuning phenomenon. Are these all linear filters to the best of your inference?

Thank you in Advance
Dorian Day

[Urs]

Every analogue filter has this. A Juno 60 for instance turns to self oscillation on the last 3 millimeters of the Resonance fader. During those 3 millimeters the pitch goes down about a fifth IIRC.

In a naive example, the cutoff frequency is determined by a set of variable resistors. The more resistance typically the lower the cutoff frequency. Non-linear elements such as waveshapers pose a resistance that depends on level. Typically a higher current gets attenuated more than a lower one. This, in a strictly naive sense, contributes towards the attenuation by the resistors (OpAmps/transistors/biased diodes). The higher the level of self oscillation, the stronger the overall attenuation by non-linear elements and thus the lower the resulting frequency.

I can't comment on other people's synths. Maybe they have compensated for this, maybe they use linear filters. No idea. A proper, uncompromising analogue model of any of the classic filters should however display gain dependent frequency of self oscilation.

[DorianDay]

Thank you so much for your reply

ACE has inspired me to reach under the hood as much as I can, I recently read Allen Stranges book on synthesizers and am trying to pick up on the nuances of synthesis, so thank you for contributing to my knowledge.

Dorian Day

[TomTwohy]

Ha, I had Allen Strange as a Professor in Jr College. Back in 1979-81

As a matter of fact the class was in a "Roland System 100" which is the model (with obvious differences) used for ACE's architecture.

I haven't thought of Allen in years