Should a synth never "click"?

[Leslie Sanford]

In certain scenarios, my synth will click. For example, if the amplitude envelope's attack time is set to zero, you can get a pop or click. Ditto for instant release time. How noticeable this is depends on a lot of things, but you can especially hear it with a sine waveform.

I could rig the envelopes so that you can't set them to instant time (make the minimum, say, 3ms), but I'd rather give the user freedom to get those tight sounds.

Another situation I've just run into is modulating the filter's cutoff frequency with an LFO set to a down ramp (sawtooth). With certain frequency cutoff settings and modulation levels, a click occurs. This seems to happen when the FM modulation extends beyond the range of the filter. The filter has clipping built-in so that the cutoff can never be modulated beyond half the sample rate. So if the modulation level is set high enough, an LFO's down ramp gets clipped. I'm not certain, but this seems to be the source of clicking with some settings, especially when the resonance is turned up.

The LFO outputs naive waveform shapes, which may be part of the problem. However, if you play with the filter's cutoff frequency and the modulation level, you can hit that sweet spot where no clicking occurs and you still get the sound you're looking for.

So my question is whether or not clicking should ever occur? And why we're at it, if anyone has tips for avoiding clicking in the first place.

[Chris Walton]

Sorry for not getting back to you on the synth yet, it's been my intention but I was swamped, and then I finally had some time but I forgot

anyway, the clicking was one of the things I intended to mention. The easiest solution is to run things that are "clickable" through an RC filter set to ~3-4ms (so that means sawtooth LFO, square LFO, but not sine LFO or triangle LFO (although that does "click" theoretically)) ... I'd say with the envelope times just put them at a minimum of 3-4ms as well.

I don't think it matters if the cutoff gets clipped ... I can't think of any reason for that to cause "clicks".

[mystran]

Envelopes artificially slowed such that they are unable to produce "clicks" are IMHO waste of time (as in, I wouldn't waste time with a synth that I know can't give me the click when I want it) but YMMV. That said, a synth shouldn't click unless that was what the user obviously wanted.

As for filter cutoff.. if the filter goes all the way to Nyquist (or preferably slightly above) then nobody's going to notice if you clip the cutoff or not, since it passes everything anyway..

[Music Engineer]

I could rig the envelopes so that you can't set them to instant time (make the minimum, say, 3ms), but I'd rather give the user freedom to get those tight sounds.In certain scenarios, my synth will click. For example, if the amplitude envelope's attack time is set to zero, you can get a pop or click.

of course. zero time implies an instant jump. that's how is has to be. but are your oscs free-running? i'd recommend to set them to a zero crossing on note on to avoid clicks. in aggressor, i let the user choose the start-phase - with zero attack times, one can intentionally create or supress clicks at the transient by choosing a proper start-phase.

Another situation I've just run into is modulating the filter's cutoff frequency with an LFO set to a down ramp (sawtooth). With certain frequency cutoff settings and modulation levels, a click occurs.

i suppose you filter is IIR? then this:

http://www.kvraudio.com/forum/viewtopic.php?t=164146

might be of interest.

[MadBrain]

Hmm, you can deal with clicks on note attacks by setting up the oscillator phase to 0 when a key is pressed - some yamaha FM synths do this. But that gives a "less analog" feel to the synth, so dunno.

Depends on who you target/how you "advertise" your synth. If you're going after the "analog" thing, slower attacks are probably a good idea, it even helps the feeling as far as I can hear.

You might also want to set your amplifier before the filter, so that when the level changes very fast, the click is filtered, which adds to the feeling... In particular, saw and square waves are already series of clicks, so this adds to the thing. But changing the amplifier/filter order has other effects.

[Leslie Sanford]

Hmm, you can deal with clicks on note attacks by setting up the oscillator phase to 0 when a key is pressed - some yamaha FM synths do this. But that gives a "less analog" feel to the synth, so dunno.

Yeah, I really like the feel randomizing the phase of the oscillators gives me. Of course, one thing I learned is that you have to becareful when you randomize the phase, or you'll get clicks there, too.

You might also want to set your amplifier before the filter, so that when the level changes very fast, the click is filtered, which adds to the feeling... In particular, saw and square waves are already series of clicks, so this adds to the thing. But changing the amplifier/filter order has other effects.

Interesting suggestion thanks.

My thought on the envelopes is to leave them be and document in the manual that certain settins will cause pops. But I can still see getting complaints even with this disclaimer.

[Leslie Sanford]

i suppose you filter is IIR? then this:

http://www.kvraudio.com/forum/viewtopic.php?t=164146

might be of interest.

Interesting that you discovered that clicks can be caused by the direction of the filter modulation depending on the filter form. That's what I've discovered. Going up suddenly causes a click, where as going down doesn't.

As an experiment tonight, I was able to mitigate the clicking in the filter by changing the LFO's to use phase distortion for generating sawtooth waveforms. The distortion is set so that the sawtooth's discontinuity is softed a bit. I haven't made up my mind if I'll use this solution.

[Gavin@FAW]

Any one who complains about clicks in the amp envelope with fast times need to get out more!

If anything clicks add abit of character, all analog synth envelopes click away. Sounds nice if you have an arp going and you start from a clicking rhythmic pattern and turn the knob to a longer decay time and it starts playing a melody.

What you want to watch out for is creeking in the filter with a fast envelope. Lots of synths (digital) do this and it is quite audible and annoying. Especially on snappy basses where its not masked by higher frequency content.

[nollock]

i suppose you filter is IIR? then this:

http://www.kvraudio.com/forum/viewtopic.php?t=164146

might be of interest.

Surely if you have a lowpass instantly drop from 10khz to 200hz, then some kind of click should be expected? You're instantly attenuating a whole bunch of frequencys, putting a step in the frequency domain.

or is this somthing else?

[xoxos]

depends on the application.. a tool for people who know what they're doing, yes.. a simple format beginner for beginners, perhaps tuck the corners in for them.

[disturb]

clicks = good

[mystran]

clicks = good

Agree. Don't blame the synths, it's almost always crappy reverbs that make clicks sound bad.

[duncanparsons]

As for filter cutoff.. if the filter goes all the way to Nyquist (or preferably slightly above) then nobody's going to notice if you clip the cutoff or not, since it passes everything anyway..

..cutoff going above Nyquist would be an interesting one... I'll let you read up on the theory for yourself

DSP

[mystran]

As for filter cutoff.. if the filter goes all the way to Nyquist (or preferably slightly above) then nobody's going to notice if you clip the cutoff or not, since it passes everything anyway..

..cutoff going above Nyquist would be an interesting one... I'll let you read up on the theory for yourself

I mean what I said.

I can't see why anyone would want a completely linear filter in a synth, and since you need to oversample anyway once you make it non-linear (or it'll sound crap), you can go beyond the final Nyquist just fine. My point was that for a typical filter the sound won't change almost any once the resonant frequency has gone above what will become the final Nyquist when resampled to external rate, there is little reason not to clip the cutoff. For certain types of non-linearities this might not hold if the filter is in self-oscillation, but...

[duncanparsons]

But if you oversample, you change Nyquist... A digital filter in itself can never go over nyquist, without getting imaginary numbers that can never resolve back to reality.

I agree with what you mean, in that you can keep a certain response curve over a number of frequencies; but you would be putting the cutoff in the 2nd quadrant of a different filter, rather than the 3rd quadrant of the original filter.

DSP