Fast Modulation of Filter Parameters

[S0lo]

Let me ask a question here. Why can’t a click or pop be part of correct filter behavior?

In other words, if you change a filter parameter (say cutoff) almost instantaneously to a far value, then shouldn’t the filter naturally click or pop at that moment? I would think that at least some filter designs would.

[mystran]

Let me ask a question here. Why can’t a click or pop be part of correct filter behavior?

In other words, if you change a filter parameter (say cutoff) almost instantaneously to a far value, then shouldn’t the filter naturally click or pop at that moment? I would think that at least some filter designs would.

Well, this goes into to the question of how do we define a "click or pop" as basically any "well behaved" filter will still have a discontinuity in the first derivative when you apply instantaneous modulation. That's a 12dB/oct pop. Perhaps in some structures it might be even steeper decay, but that's what you usually expect. If you have a discontinuity in the waveform itself, that's a 6dB/oct pop/click. If you have a transient of some sort that results in an amplitude spike, then at "low frequencies" (depending on the shape of the spike) you can have a flat 0dB/oct click (in addition to potential other problems).

[S0lo]

Thanks for answering mystran. But (I should have said) that I was kinda aiming that question to soundmodel as a way for him to not get locked into thinking that all pops and clicks are inherently wrong. Just because too many papers out there talk about it as being so.

I just ran a sin wave into a TPT based low pass while modulating the cutoff with a square wave LFO (slow). The output has obvious clicks at the cutoff changes. I’m not surprised at all, that’s how it should sound like. I would expect the analog counterpart to do the same.

If someone wants to eliminate that, without filtering the modulator and without modifying the filter, then I wouldn’t know what to do.

[mystran]

I would expect the analog counterpart to do the same.

Some designs might have a lowpass filter (eg. typically a single cap to the ground) on the cutoff input side to limit the modulation bandwidth. You see this perhaps more often with hybrid designs where the filter is analog but the modulation comes from a digital source, like when envelopes/LFOs are implemented with a microcontroller that is then output with a simple DAC producing a stepped waveform, but you could do it in a pure analog too (reduce high freq noise, perhaps improve servo stability, etc)... although obviously if you want to have filter FM then you have to have enough modulation bandwidth to let the FM modulation signal through.

[soundmodel]

Let me ask a question here. Why can’t a click or pop be part of correct filter behavior?

In other words, if you change a filter parameter (say cutoff) almost instantaneously to a far value, then shouldn’t the filter naturally click or pop at that moment? I would think that at least some filter designs would.

It can, but it's not desirable in a musical context. Ideally, we'd obviously want a filter that acts continuously across the parameters. Or that has a "cross-over" behavior built-in.

However, if you're talking about the mere perception of two separate transients very closely to each other, then obviously it is audible as a click, but it's a corner case, because no-one really uses filters that way.

However, the original reason for the artifacts seems to be that people deal with transfer functions, even though it's guaranteed that transfer functions do not have any certain kind of continuity between them, since they're LTI.

And yes, it doesn't sound mathematically odd at all that there's a function that's parametrizable so that the the cut-off moves the filter linearly in frequency. Like the Moog ladder:

https://soundspear.com/blog/5-different ... io-filters
http://www.blaukraut.info/temp/moog_transfer.png

[EmTee]

Even when doing filter modulation with TPT's you might need some easing on the changes to prevent clicks. They might not blow up but they still don't like instantaneous changes

[S0lo]

Let me ask a question here. Why can’t a click or pop be part of correct filter behavior?

In other words, if you change a filter parameter (say cutoff) almost instantaneously to a far value, then shouldn’t the filter naturally click or pop at that moment? I would think that at least some filter designs would.

It can, but it's not desirable in a musical context. Ideally, we'd obviously want a filter that acts continuously across the parameters. Or that has a "cross-over" behavior built-in.

However, if you're talking about the mere perception of two separate transients very closely to each other, then obviously it is audible as a click, but it's a corner case, because no-one really uses filters that way.

Sorry, but thats just not true. What happened to all these musicians using square LFOs, stair steps, sample and hold.. etc. hundreds of use cases, you definitely don't want to smooth that. And even if you do, then just lowpass (slew limit) the modulator. Otherwise, if you forcefully modify the filter to give you a smoother response, then it's not the same filter you designed or emulated any more. It's a different filter now. If thats OK with you by design, then it's OK. But I'm sure that this can very much be undesirable in some musical contexts as it is desirable in others.

[soundmodel]

You mean that a square LFO, stair steps, sample and hold etc. are not smoothed or the process (e.g. a filter) is not smoothed?

I don't know, but doing e.g. a <= 5 ms cross-fade rarely affects the perception of continuity.

[S0lo]

You mean that a square LFO, stair steps, sample and hold etc. are not smoothed or the process (e.g. a filter) is not smoothed?

I don't know, but doing e.g. a <= 5 ms cross-fade rarely affects the perception of continuity.

What I meant is, relative discontinuity in the modulator, and therefore potentially in the filter output can have musical value in many contexts.