Hello!

I'm encountering a behavior with the vcf~ object that I don't understand, and I think it's contributing to a lot of popping noises being produced by my patch.

The behavior is this: even when the first and second vcf~ inputs stabilize at 0, I'm seeing an extremely low-amplitude, low-frequency signal continue to be output by the the vcf. It appears to have something to do with the middle inlet, as if vcf stops accepting inputs just before they reach 0.

I've included a simple patch here that illustrates the problem. The test frequency is defined by a phasor~, and both the test frequency amplitude and the center frequency itself are controlled by a simple ADSR envelope. The two inlets to vcf~ are snapshotted, as is the output, so you can see that even when the inputs go to 0, the output oscillates, very slowly and quietly.

I would massively appreciate any insight anyone has into this!

Thanks,
Charlie

vcf_weird_output.pd

@chvolow24 said:

It appears to have something to do with the middle inlet, as if vcf stops accepting inputs just before they reach 0.

Filter math in general strongly disagrees with a cutoff frequency of 0.

FWIW, in SuperCollider, if I do something similar (gradually pull a filter frequency down to 0 while silencing the input), I do not get zeroes in the output.

a = {
    var line = Line.kr(1, 0, 10);
    RLPF.ar(LFSaw.ar(200) * line, line * 100, 0.5).poll(10)
}.play;

... wait 20-30 seconds...
UGen(RLPF): -0.000123828
UGen(RLPF): -0.000124234
UGen(RLPF): -0.00012464
UGen(RLPF): -0.000125046
UGen(RLPF): -0.000125451

So I think you simply can't expect pure silence out of pretty much any filter at 0 Hz. You would need to find another way. Probably put the envelope-controlled amplifier after the filter, rather than before.

hjh

@ddw_music Undesirable behavior at a 0Hz cutoff makes sense, though I think what's most surprising to me is that the filter still produced a signal when the filtered signal (the leftmost inlet) was also at 0. But there's a lot I don't understand about the mechanics of how these filter work, so I shouldn't be surprised. Moving the amp env downstream is an easy fix that should work. thanks so much for your help!

@chvolow24 said:

@ddw_music I think what's most surprising to me is that the filter still produced a signal when the filtered signal (the leftmost inlet) was also at 0.

Digital filters can be built to depend only on input samples (variously called "feed-forward," "non-recursive" or "finite impulse response" filters), or they can depend on both input samples and previous output samples ("feedback," "recursive" or "infinite impulse response").

Typical digital filters with resonance are of the feedback type (because resonance without feedback requires a high filter order, which is CPU-expensive). That's certainly true of vcf~.

So you silenced the input but the feedback components still had a small amount of energy.

hjh