Thanks 
I was browsing around for posts on filters doing some research into where Rack 2 is at, and found this topic and the quite interesting paper Raph Levien that was linked. I had to investigate a little further to see what he was doing. I thought even though this thread is 2 years old perhaps it could be helpful in the future to have a bit more insight for anyone interested, especially since I had, without reading any of this had already gotten in contact with Andrew Belt about improving the high pass of the Fundamental Moog based RK4 VCF.
7 Likes
there is a design for a stupid simple real filter (24 db resonant etc). first saw it modeled using vcv fundamental modules by Jakub Ciupinski. (https://www.youtube.com/channel/UCVN95kHPHk--fSwvTCRJmdw) the idea is that if you feed back a slightly delayed signal and mix it with the original signal you get filtering. I coded it up as a module and it worked.
3 Likes
But why would this work, some might ask?
The humble delay will create a phase shift (relative to the original signal). So…we’re into phase shift territory.
Mixing the phase shifted signal with the original will cause a combfilter effect. Due to the phasing in/out of all frequencies in the spectrum. In the harmonic series, the frequency of each harmonic is simply an integer mutiple of the root frequency. An octave is simply a doubling of frequency. For any ‘harmonic’ signal, we can reconstruct it using these ‘harmonic’ frequencies (summing sines at their specific amplitudes, additive synthesis, for now ignoring relative phase per harmonic within the spectrum).
So…going up the spectrum:
Combining both statements means that there are simply ever more harmonics per octave (1–>2, none in between, 2–>4, one in between, 4–>8, 3 in between, etcetera). So…ever more frequencies, are phasing in/ouout. Higher frequencies will phase in/out faster (more often), simply because their wavelength is ever shorter (relative to lower frequencies with their longer wavelengths).
So…generally relatively more higher frequencies will phase each other out when shifting relative phase a little. So…it’s not so much Low Pass filtering, but more a specific case of phase cancellation. Actually you could better call it a Comb Filter, not specifically (or only) attenutaing higher frequencies, just relatively more (all over the spectrum).
Note that the amount of phase shift very much determines where and which frequencies will phase cancel in the spectrum.
You can hear the phase shift / comb filter effect range all over the spectrum in action by simply detuning 2 oscillators.
You can of course experiment with high precision delays / phase control to see/hear/control all this in detail.
E.g. by using the Sckitam Waveguide Delay (which also supports audiorate Phase Shifting / Phase Modulation)
2 Likes
Thanks! I’m still interested in the topic although I haven’t had time to think about VCV for a while, hopefully I’ll get back into it this summer. Although I have mostly concluded that I have more novel ideas in other areas, I am interested in using some dynamical systems ideas for oddball filters.
KRTPluginA/F.cpp at master · jackokring/KRTPluginA · GitHub is linear but with an inter pole pair (2+2) clip distortion. It’s linear so zero delays are about as simple as it gets. There is much linear interpolation of coefficients.
But it is quick as it does no non-linear maths except one tanh on control frequency change.
Another warning in digital filters is that BiQuads (of which this isn’t) can behave very strange on fast modulation of the cut-off.
1 Like
Just using this example to mention that necro-bumping is allowed and encouraged on the VCV Community. The askers of questions may be long-gone, but answers and facts can still be helpful to others. The exception is when the question is obsolete, such as involving an old version or a bug that has been fixed years ago.
3 Likes