posted 01 July 2011 22:53            

The basic structure of the patch is ...

MemRec -> TimeScaling -> delay
audio input / \ mixer
\ threshold ----> trigger ---- /

The delays are stretched to twice the original length, which means that sometimes it's still playing at a strong volume when a new trigger is activated by the live input. When that happens, of course, there is an obnoxious click.

I can think of 4 basic approaches that would eliminate the clicks, but I'm not sure which would be the most expeditious, nor how to implement it. For future reference, I wouldn't mind know how to do all of them... so here they are.

1) Have the sound ignore new triggers if the amplitude of the stretched sound is non-zero (or above a given threshold). (Probably the least desirable solution, but maybe useful in some settings.)

2) Have the stretched playback portion of the sound respond to all triggers by very quickly "ducking" to zero and ramping up to track newly triggered input.

3) Have the new triggered be delayed until the stretched playback has stopped playing. The worry with this one is that delays might begin to compound and pile up, so the stretched playback would get further and further behind the original input. (But again, it might be useful in some future situation.)

4) Trigger a new polyphonic voice if the first voice is busy. This one seems like it would produce the most natural and organic output in this situation. I think it might only require 2 voices (maybe 3) to ensure that there was always a free voice available. I know how to handle polyphony when MIDI is involved, but not in a situation like this.

=m.p.=

posted 02 July 2011 15:12            

{'recording' & ?VoiceNumber}

posted 03 July 2011 11:36            

Thanks for this syntax. (I wonder when/if usage of curly brackets will become 2nd nature for me.) I'm still not sure how use it to generate polyphonic output, since the Sound as no MIDI component (nor will there even be a MIDI interface used in the performance. I'm assuming I'll need to use a MIDIVoice, click on the "script" option, and then (glup!) write a smalltalk script that includes {'recording' & ?VoiceNumber}, yes?

posted 03 July 2011 12:32            

If you want to control everything from the VCS, you could use a Replicator to get multiple copies (see the Prototype named "Replicator creates polyphony" for an example). In that case, you could use something like the following expression to gate the recordings:

posted 03 July 2011 13:33            

re your 4 basic aproches

1) Zero crossing will not stop the click. If a signal jumps in level even if there is a perfect join at zero it will still click.

2) This will give you holes which may sound click like.

3) As you say, you may loose control and build up un-handleable data.

4) Is the closest, but the best solution to clicks is normally cross fade. To achieve a good cross fade you must have the outgoing and incoming signals playing at the same time (during the crossover) so you need two copies you your algo and you have to alternate between them. The trigger will start the new sound and start that sound fading in while making the old sound start fading out at the same rate.

So the trigger needs to make a 0,1 toggle, (change state on each trigger) type square wave. This squarish wave needs to be smoothed (trapezium shaped) and feed into a VCA on the output of one algo. The inverse (centered about 0.5) of the same smoothed square wave feeds a VCA of the other algo. The triggers can also be gated by the same squareish wave form (and the inverted version) by using multiplication, to generate the two alternating streams of triggers.

If you can get the new signals zero to line up with the old signals zero as well, you should have an even cleaner sound but the cross fader will probably give you the greatest improvement.

I hope I've understood what you are trying to do.

Thanks

Pete

posted 03 July 2011 21:48            

ForumEx.kym

In addition to the Trigger window of the MemoryWriter, I tried pasting your suggested expression:

(!Gate countTriggersMod: ?NumberVoices) eq: (?VoiceNumber - 1)

into every conceivable combination of parameter windows, but to no avail.

posted 04 July 2011 12:48            

Yes, I know Sound is "working" to the extent that I've got polyphony. But the additional voices are more-or-less simultaneous (with just a touch of chorusing). But what I want is for them to trigger and play sequentially: i.e. first trigger activates voice 1, second trigger activates voice 2, 3rd trigger activates voice 3, then continues to recycle voice 1, 2, 3 etc.

In addition to my ears telling me that the polyphonic voices are triggering simultaneously, I can see in my VCS that all of the Replicator-created gates (_1Gate, _2Gate, _3Gate) are all firing each and every time the input crosses the threshold and generates a trigger. I assume that when I get this working, that these numbered gates will respond sequentially, too.

posted 04 July 2011 13:24            

ForumEx2.kym

I've uploaded a renamed version of my Sound.

posted 04 July 2011 17:17            

ForumEx3.kym

Also ... now the MemoryRecordings no longer updates with new recordings as the audio source plays new motifs, which is a key feature of my original idea. I've uploaded a new example with all three Sounds for comparison.

Don't worry about responding anymore today ... it's a holiday for pete's sake. I've got to go fire up the grill and get ready for dinner guests. Happy 4th of July! ;-) We can try to tackle this sometime on Tuesday. Cheers!!!

posted 15 July 2011 14:26            

ForumEx5.kym

Unfortunately, instead of the rare occasional click, I now have "ghost notes" all over the place, which sound like short, false starts. Can't quite figure out why. I've spent the entire morning (and beyond) messing around with TimeConstant (RMSSquared), Threshold settings, AttackTime and ReleaseTime (PeakDector) ...all to no avail. Making these changes does change the playback (different response and different "ghost" notes). But nothing seems like a solution yet. ;-(

You can hear the problem in both of the attached files, if you listen closely. One is a copy of the one you sent posted -- modified to make it easier to hear what I'm talking about. The other is a portion of the actual project I'm working on. I'm using an audio file that will eventually be replaced by a live performer, but it's too large to upload.

posted 15 July 2011 15:28            

ForumEx6.kym

The Gate was staying on until the beginning of the next phrase, so some of the material that preceded the next gate would be captured in this gate.