Sample not triggered if MIDI overlaps

[Dan Korneff]

@Christoph-Hart

Synth.noteOff(storeNoteNumber);

This line is giving me an error.
Illegal operation in audio thread: String creation

[d.healey]

@dustbro said in Sample not triggered if MIDI overlaps:

This line is giving me an error.
Illegal operation in audio thread: String creation

That error message usual means you have a typo in a variable name

[Christoph Hart]

Not sure what went wrong here (probably something trivial), but use the event ID obtained by Message.getEventId() and the Synth.noteOffByEventId() method which is far more reliable.

[Christoph Hart]

BTW, a hint about the issue would be just adding Console.print(Message.getNoteNumber()); into the onNoteOn callback to see whether the second overlapping MIDI event reaches the scripting engine or if the problem lies before this (either the DAW itself or the way how HISE processes and prepares these messages).

[Dan Korneff]

@Christoph-Hart @d-healey Illegal operation in audio thread is being caused by the timer interval.
When Synth.startTimer(); is set to anything less than 1, I get the error.

[Dan Korneff]

@Christoph-Hart said in Sample not triggered if MIDI overlaps:

Console.print(Message.getNoteNumber());

hmm,.... that sounds way too logical. :grinning_cat_face_with_smiling_eyes:
Great idea! the overlapping midi note doesn't make it to the scripting engine. :(

[Christoph Hart]

And you can confirm that this is not a problem with other plugins?

Can you test what happens if you overlap notes with different MIDI channels?

[Dan Korneff]

@Christoph-Hart This is for a drum sampler. I can confirm that overlapping notes do trigger samples in SSD, Superior Drummer, Addictive drums, and Groove Agent.

[Dan Korneff]

@Christoph-Hart said in Sample not triggered if MIDI overlaps:

Can you test what happens if you overlap notes with different MIDI channels?

Ok... changing the midi channel for an overlapped note will trigger the sampler.

[Dan Korneff]

so what does this mean??

[Christoph Hart]

HISE processes incoming MIDI messages and creates the event IDs, I suspect there is something fishy going on in this logic.

So just to make sure we use the same conditions:

• HISE 2.0.0 (the app, not compiled plugins). Also is this behaviour new (=> is it not occurring with HISE 1.6.0 or earlier?)
• any DAW / OS
• happens with the minimal example here:

HiseSnippet 743.3oc6W0saSCCE1osFXhsNPBw04E.oF13m6V+eah0QUSY.WMYbNq0hD6PhSgdKOY7HvixdC.6jrkDsntlKXrIHWkym8I4yemexIiCDTHLTDfL1X5Re.YrI1dIWNu2bBiiNrOxnIdDITBAlIPcW5SBCAGjgQ880.Faz.Eec9dcItDNExfPnSDLJbDyiIyPG29MLW2gDGXJyK2t2s8gTAumvUDo3ScbKjOg9YxL3Xhda0vHi6MvgIEA1RhDBQFM5JbVZOW7Udx9OgEx9jKnMrP1pGTB7Pgqilw56Q8lybcFew4NDgLviyTg5IpvSviXNrKwyTiGEufYlG40CiZqhdV4oWqRomUIzCkicMRXmhDBmHWhrHyzgizEXhhQpC4RfGxjKyGt9qR2GiGyjz4ky2ZkvWUf5OMeSC9agGb1Y.UlQ1F3ge3lMR2LgJODay3PbEYLQdZrs46IK.y8ANDnEUqUTT9y0snzesKJeKUpd8SCH7PeQXgGrM3wlJ3ZEIC7cgvv.3KSzQ4738DjfRW578FpNjk5iMQFEDmtzwSDwkEROpu1wj7MGJV0ja+kzonw+6TTEo8Z56Zjiuamv26i6bPe6IocHhu2b.eA3J7gho4iDbg+bAmQymCLAjArYyff7ruziTGoT8wkdQAKJjA2GnjkWAMYyWE4HPQs7vGnNwY16zN9wcIPqu21NJTVn.8i+XB3BjvBkbCnBU3AVUHn95EBtlO2spzjsvIGRy3FB2t+ZRyK3ZRD41MY2DGmVbWPW2FmlvdmPXahSqktKHsqb.n5+aN.zDQjjwmMhn5h+MUmpii7rUycPAES3bvU8hvF0zyNjX2RamL3A2I13WpqzEsz1FoKZcwh2HuCOBMPbpZtIYfvU2y8AwHpyMO9WY1P8OUJaSKzBhaTpJ4oFW3TJUK1OSoNk6wyqrG6TYO1sxd7hJ6wKqrGupxd75U3gd1rNQRgWRomBX7f9DIQkaOfSTYtwY4neCPOMjSO

It's weird that I can't replicate this because this is nothing that is depending on the DAW or OS.

[Dan Korneff]

@Christoph-Hart I've been experiencing the audio dropout effects for about a year now. I assume it's related and has been happening on all versions.
What's the easiest way to download pre 1.6 to test?
I'm goin to check the 2.0 app on all the Mac + PC combinations I have available here now.

[Dan Korneff]

Overlapping example confirmed not working on:

macOS 10.13.6
Reaper 5.961
HISE 2.0 AU crashes GarageBand on load

windows 10
Reaper 5.961
Cubase 9.5.41

[Dan Korneff]

Hey guys. I need a little help writing a script to get around this issue.
I'm trying to send a noteOff command directly after a noteOn is triggered.

Here's what I'm doing:

function onNoteOn()
{
	noteID = Message.getEventId();
	Synth.noteOffByEventId(noteID);
}

In the console I'm getting an error:

Script Processor1:! onNoteOn() - Line 3, column 31: Illegal operation in audio thread: String creation 

¯_(ツ)_/¯

[Dan Korneff]

This post is deleted!

[Dan Korneff]

@Christoph-Hart said in Sample not triggered if MIDI overlaps:

HISE processes incoming MIDI messages and creates the event IDs, I suspect there is something fishy going on in this logic.

@Christoph-Hart This is correct.
I've made a more simple example that contains only:

function onNoteOn()
{
    Console.print("on");
}

and using this midi file with an overlapping note:
midi tes.MID

If you overlap midi events of the same note and same channel, the 2nd noteOn command doesn't even make it into the Main Container.
If you look at this example, the midi event contains 4 midi notes, but only 3 make it into HISE.

[Dan Korneff]

Chatted with some code slingers the other day who pointed me to here:

[Christoph Hart]

The synthesiser class in JUCE is only partly relevant, I think the culprit is the way I am generating event IDs from the MIDI input. The problem is that there is no "correct" solution, as soon as MIDI notes with the same channel and note number overlap, things get into undefined territory.

Currently I have only a single slot per note number, the first way would be to change that to a FIFO queue, but there are performance implications that have to be evaluated against the edge case of people not using their piano roll correctly :)

[Dan Korneff]

I've just started going thru all of the JUCE tutorials. Very interesting stuff. Also thumbing thru the HISE code to learn how you're handling these situations. It's all over my head at the moment, but I'm interested to see the solution.

@Christoph-Hart said in Sample not triggered if MIDI overlaps:

there are performance implications that have to be evaluated against the edge case of people not using their piano roll correctly

man... I think all of my beta testers are the "edge case". :face_with_tears_of_joy: They draw in midi for basic drum beats, then drop a midi note right on top of another note to create their flams. Every. Single. One.

@Christoph-Hart said in Sample not triggered if MIDI overlaps:

as soon as MIDI notes with the same channel and note number overlap, things get into undefined territory.

I can definitely understand this issue. Once there is an overlap, how does the buffer determine which noteOn/noteOff belongs to which event ID.

[Christoph Hart]

how does the buffer determine which noteOn/noteOff belongs to which event ID.

Well, FIFO means First in First out, so the earlier note would automatically get the note off. The other model would be a LIFO (Last in, first out" stack), but I think the result would be super weird.

They draw in midi for basic drum beats, then drop a midi note right on top of another note to create their flams. Every. Single. One.

It will probably come down to a preprocessor definition that customizes this behaviour on a per project basis. This problem is more apparent on drum based projects, because you don't need your super VA style Saw synth to flam :)