gate off event

How can I extract note off information to send an event from a module's polyphonic pitch output (when shows 'off' in debugger)? The voices are reassigned earlier in the patch so I can't simply use a Gate input without setting up a whole new MIDI output. There's probably an easier solution that I'm overlooking but I could use some perspective here, too deep in the weeds..

Find more posts tagged with

Accepted answers

All comments

colB

https://community.native-instruments.com/discussion/47970/gate-off-event

Hard to say for sure without more information for context.

Just remember that MIDI is not polyphonic. It's just a serialised stream of events.

The obvious thing would be to just use a MIDI note Pitch/Gate module that will give note on and off…

If it's some polyphonic process that internally generates events… like some sort of poly sequencer, maybe you could just extract note off info polyphonically, then use an event voice combiner to mix all those note off events into the same stream. And output using a Channel Message module.

Worst case is that you are in core, and have multiple poly events on the same tick. In that case you might need to buffer them with some sort of FIFO buffer…

Like I said, hard to tell without more context.

silvermoats

I'm using a wraparound process for dynamic voice allocation in Reaktor to reassign pitches ('to Voice' modules) which is why directly using 'gate in' wouldn't work. I was thinking since the re-assigned pitches still show 'note off' it might be possible to extract that information, but it seems like it might just need to route the gates through a parallel chain…

silvermoats

voice reassign test.ens

here's an example .ens of my current method (thanks for the latest voice core cell, lol) - it seems to work ok using parallel pitch/gate signal flow. the # of voices knob is a stand-in for a more complicated voice allocation process I've got operating correctly with note values, didn't want to muddy up the demo here but this is the basic idea. any thoughts on how I could streamline this? I'd like to avoid stuck notes when the # of voices is changed & preserve the original routing when possible (diverges at full polyphony as in screenshot). for hairy structures it would be nice if there was a clean way to allocate pitch+gate with a single process, which was the thrust of my original question. even sharing the first step of pitch/gate processing with the single core cell on the bottom gives me stuck notes for some reason I'm possibly overlooking

I'm also open to the possibility that there's some easier way to dynamically change the # of voices outside of the properties menu with built-in Reaktor modules that's been hiding from me all along.. 🙃

colB

Hmm, still trying to understand the requirements…

Q. Are you trying to manually control number of active voices…

If so, why?

and what is it you need note-off info for?

I suppose what i'm asking is: tell me about why the project needs dynamic allocation, not about how you are trying to implement it

Manual midi coding in Reaktor is not easy, so it's good to really nail down the requirements before even starting to think about implementation.

silvermoats

yes manual control of # active voices, but also potentially automated / generative control, just looking to do it without actually directly changing the box in properties. I'd like to use the polyphony for parallel signal processing that isn't necessarily related to traditional 'synth with #x voices' architecture, but being able to implement some homebrew version of that would keep things open to being routed along those terms. how I'm dealing with this now uses two instruments, one with a small number of 'real' synth voices, and another instrument housing the DSP guts to which those note values are routed - each of those values, respectively goes into a different per-voice macro, which are functionally identical, but can be switched off along with the number of voices. this polyphonic processing kind of thing can get pretty computationally expensive but aside from the cost there, having the ability to kill voices on the fly would be more intuitive from an interface perspective & I could see it making a big difference in an audio effect context for input signal that has a particularly dynamic range. it's not an absolute requirement by any means but I figured I'd check in here for some perspective before I sink too much time into it

for my purposes the MIDI part is only really relevant insomuch as that's how the voices are arriving into Reaktor, although I realize it's the source of the 'off' I'm seeing in debug mode later in the signal chain - purpose of the thread was to see whether there's anyway of extracting those directly even though they don't trigger an event..

it would be great to extract note offs from the re-routed voices to do something simple like an ADSR envelope for starters, but not something I'll get hung up on if its impractical. it is mildly frustrating though because I can manage this with pitch+gate ons pretty easily, so I feel like I'm halfway there - firing triggers rather than gates would work in a pinch but I feel like there's probably a workaround.

figuring out how it ties into the big picture is usually the last step for me!

silvermoats

Next go I have at this I'm thinking it might be easier to match the gate events to their original voice by using a comparator or some kind of conditional logic relative to MIDI pitch # instead of setting up a dedicated rerouting chain for gates to match that of pitch. I had some notion that could be done polyphonically using voice shift modules but it's too much hassle to sort binary on/off logic across multiple voices without missing any or getting a false reading

colB

Using polyphony for parallel processing is quite different to using it for regular poly synth duties.

To control polyphony for parallel processing, just set up a core cell (assuming core use case) with some max number of voices, and send the voice info to an input, then at the top level of the core cell, switch the SR clock off if the voice number is greater than your dynamic maximum. You might have to set the output to zero (or one) for those inactive voices, and whatever other housekeeping might be required depending on the context.

That way you can have 30 voice, but a dynamic count at e.g. 12, then any voice with a number greater than 12 will be inactive.

For synth duties its more tricky because you need to reuse e.g. 'oldest' voices to make the voice stealing less obvious. And if you want to 'release' unused voices, you need to do it based on some clever envelope follower logic, or internal knowledge from the synth engine, not 'note-off', otherwise voice with long release setting will be audibly chopped.

silvermoats

Yeah I'm looking at using a combination between those two approaches - the parallel processing is a nested layer controlled by more traditional synth polyphony - & the latter is definitely what I've been finding more tricky. For my purposes I'm generally ok with artificial-sounding voice stealing, I just don't want any stuck/delayed notes, and the logistics of routing note offs to their correct voice has proven a bit difficult for me to conceptualize when they could potentially be sent from the MIDI after a voice steal has already happened in Reaktor.

That's a great tip for faking max voice # in core, thanks for sharing! Will have to give that a shot soon

Quick links

Unanswered questions