# Variable slope filters

Member Posts: 449 Advisor

Would be nice to see it done. Seems it's always a choice between 1 2 or more poles. I've never seen a continuous 3db or 9 db per octave filter. Always 6 12 18 etc. Are they out there?

Tagged:

• Member Posts: 7 Member

The most basic filter, like a RC filter, is 6db/oct.

• Member Posts: 56 Advisor

I once built a "variable slope filter" by cascading four 1-pole filters in series and using a 5-input linear selector to crossfade across them. I suppose it wasn't mathematically sound, but to my ears and my spectrum analyzer it behaved as I thought it might. It wasn't resonant, of course.

• Member Posts: 292 Advisor

fir filters can be designed for arbitrary slope but that can't really be modulated

if you go mixing poles in iir filters it'll introduce phase problems. you could use linear phase but then you'll run into latency and the inability to modulate

I would probably go for chet's method, tho I also read something about using shelving filters for this as well

• Member Posts: 16 Member

This is a variable slope filter built with 16 shelving filters - the phase response could be a bit a mess (because of the 32 poles) and it leaves a floor on the highend, it is near the indicated ideal cutoff line up to ~ 20 dB / oct, then it starts getting too flat

With some 'true' lowpass filters for covering the 'integer' pole part there is certainly much room for improvements

---

Another straight forward method is doing fft-filtering, but i am not really convinced by this because of the introduced latency and a bit of an odd phase behaviour

• Member Posts: 449 Advisor

I guess it's the nature of the beast. Thought I could use allpass filters to align the output phase of the filter with the input, or match the input phase with the filter output after going thru an all pass. If the phase was linear the a simple crossfade might work. Don't know much about allpass , maybe the phase shift isn't the same as a filter.

Interesting, I'll check it out. I'm still thinking an all pass filter might actually match the phase relationship of the low pass filter. If so, averaging the two filters might actually produce -10db per decade if the low pass is -20db per decade. Or close to -3.01db per octave. If I remember the all pass approaches 0 degrees on the low end and -180 degrees on the high end. It's been a long time, but when I think about it, just subtracting an all pass for the straight signal would produce a high pass filter. If added then it would make a low pass filter. Perhaps averaging those two signals would produce a flat curve. If so, then a crossfade of some sort might swing both ways.

• Member Posts: 16 Member

Could also be interesting to produce a symmetric, ideal linear-phase impulse response for such a slope to see how it looks and changes with the rolloff steepness

• Member Posts: 449 Advisor

I guess I'll check the phase of the all pass filter and see whats going on there. I'm not sure they even have one yet, I've only seen the response on an oscilloscope of an analog one with an op amp.

• Member Posts: 16 Member

sounds like a good plan :)

• Member Posts: 292 Advisor
edited September 2022

so uhhh... this video isn't exactly the same subject but it covers a lot of stuff that sorta illuminates the issue and helps it make sense

• Member Posts: 97 Advisor
edited September 2022

Mixing between 'pickup' macros connected to the factory 4-pole filter seems to work well to get 'in-between' slopes, without any drastic / noticeable phase issues:

• Member Posts: 449 Advisor

I came up with something using the all pass filter, not exactly sure what it is but it does seem to shift from no filtering to a 6db per octave. The switch changes from a low to high pass. It has a sweep from 8 hz to 20k but the filter is nothing but a crossfade between the input and the output of a 6db all pass phase shifter.

• Member Posts: 449 Advisor

Yeh, I see, I made three variable slopes so far. A high pass, low pass and notch. It seems to work and I assume more depth would accumulate with two of these in series. Haven't really got the hang of it but in a way they are pretty cool. The flat part actually reduces in volume as the high or low end rolloff is flattened out. So when mixing added highs is adding a sense of overall volume. But reducing the low end might balance that out. Just a cheap theory but it's worth trying out. This has a sweep generator in it and a scope to show it working but you can apply any signal in place of the sweeping sine wave. I'm gonna have a listen and see if I like it. In a way it's what I was looking for. It kinda resembles a shelving filter as you flatten it out.

• Member Posts: 449 Advisor

This is actually pretty nice. Have a listen on a guitar or some real instrument. It's almost ideal in a mixing session.