Yoshimi and Poly Aftertouch

[BJG145]

I'm impatient to get cracking with this now. Turns up Tuesday!

[BJG145]

For some years now, we've had poly (key mode) aftertouch in Yoshimi (classed as experimental)

I just tried this with an Exquis. It's not very melodious and I only got as far as testing pitchbend, but I can confirm that it's working.

Three cats in a sack

[Martin Walker]

Well done - thanks for confirming that this works on Yoshimi

[Folderol]

Thanks a lot for this. Very much appreciated

[BJG145]

One of the reasons for the dodgy demo was that I was having problems with hanging MIDI notes. This turned out to be a problem with the Pisound card, and I had to dig out an RPi 3 with Bullseye in order to fix it. But it's all done, and Yoshimi on RPi 5 with an Exquis is now working great.

So here's a slightly better demo of poly aftertouch on filter cutoff. (I still haven't got the hang of recording audio directly on the RPi yet, so I'm slumming it with a battery amp and a Tascam.)

https://on.soundcloud.com/5676Tumnt86kFmhj7

(I'm finding that the pressure modulation range is towards the top of the keypress rather than the bottom, so you have to be quite careful to control it without releasing the note; need to tweak the settings somewhere on either the Exquis or Yoshimi to get them to play nice together. This wouldn't be a problem with a regular keyboard which isn't going to send aftertouch at all until you hit the end of the key travel.)

[Martin Walker]

Well done Ben!

[Folderol]

Thanks again Ben. This is sounding pretty conclusive

[BJG145]

I've now got this set up with a browser-based remote desktop connection called Raspberry Pi Connect which was already installed and ready to go in my Pi OS build.



This is the kind of integration I wanted...controlling the Yoshimi GUI on one monitor with the DAW on the other, got my JV30 plugged into the RPi and the output routed back to Sonar via the PC's audio interface.

Sorted. I'm good to go.

[ajay_m]

When I developed a plugin for the kmi qunexus, I found that poly AT is hard to control without an envelope that slews over time, responding relatively quickly to increased pressure but then when pressure is released, being slower to ramp down. This makes it much more natural to control. The whole code for the qunexus plugin is on the reaper stash (search qunexus unleashed) and contains full documentation. Having some of that midi processing in the synth would be awesome as nobody really does this now. Code is open source so feel free to pinch any of the ideas, but I promise you the techniques implemented make poly AT much more musically useful in my opinion.

[The Elf]

You make a good point. I could imagine this would also be good for mono aftertouch.

[BJG145]

Going back to the Exquis, the difference is that its keys can send continuous expression data as soon as you start pressing them, which could be an asset, but you might not always want that when using it in poly-AT mode as opposed to MPE mode. Eg it made it impossible to control pitchbend sensibly in that first Yoshimi demo.

I just came across an interesting approach to this in the Lumatone docs, called Lumatouch. The way this works is that you can configure a key for full-range expressive potential and then trigger it by digging into the aftertouch. I think this is what Mike Battaglia's using at the end of his rendition of Scarborough Fair.

[BJG145]

You make a good point. I could imagine this would also be good for mono aftertouch.

...yes, interesting idea. I've rarely used aftertouch myself but I think I can see how that might improve the response and I'm thinking about why. Perhaps because it takes more effort to press the key into aftertouch than it does to release it, as it naturally springs back.

To be fair to software developers, this probably belongs in the hardware, but I take your point that if the hardware developers aren't going to do it then the synth developers might as well.

[ajay_m]

Yes. For mono aftertouch it is traditional to average out the pressure across all keys for which aftertouch is activated. This of course is necessary if the triggering device really only has a single pressure sensor. However poly AT devices want to be able to output channel after touch because many receiving synths ignore poly AT. In this scenario you want to ensure that when a chord is held and aftertouch is triggered, adding or removing notes from the chord causes aftertouch to ramp via an envelope to a new average value, again slowly if the value is less than before, fast if it is greater. This makes it much easier for the player to manage dynamics. Natural sound sources like an orchestra tend to be able to crescendo faster than they decrescendo, and so the control problem with aftertouch without the extra processing i have implemented tends to cause unnatural decreases in volume or cutoff as the player tries to release notes smoothly. A synth that implemented my algorithms would be far more "musical" with regard to performance. I know this because of course I fine tuned the playing experience for me using the qunexus, which also has virtually no key travel as well. But the problem exists even in conventional keyboards fitted with poly AT such as my hydrasynth, which has some rudimentary pressure envelope processing but this is not sufficient for true control.

[Folderol]

Interesting - I'll discuss this with the other main Yoshimi developers (although we have more important priorities at the moment).

[BJG145]

...for mono aftertouch it is traditional to average out the pressure across all keys for which aftertouch is activated...

...right, good point.

(I posted about Exquis/Yoshimi on the Exquis forum and they've pointed out that you can edit the curves on the Exquis, so I'll look at that.)