Lmms: VST pitch automation not working

[Cangle50]

If you try to automate the pitch in vestige with a Automation track, it makes a nasty unacceptable crackling/static noise

this is on a Windows 10 OS PC.

[SecondFlight]

What VSTs have you tested this with?

[Cangle50]

Take your pick of vst it doesn't appear to matter but you could try it on synth1.

[DomClark]

I've experienced this too. I haven't investigated too far, but I think it's caused by #4460. VST automation is currently sent to the plugin from the main thread, and if it's still holding the mutex to send the pitch event, audio processing is skipped for the plugin. I think my fix for #4543 should also fix this, as it changes VST automation to be sent from the processing thread.

[SecondFlight]

Thanks for the report, @Cangle50!

[Cangle50]

Your Welcome SecondFlight

[PhysSong]

I think it's caused by #4460.

I think so. However, I thought it wouldn't matter. See https://github.com/LMMS/lmms/pull/4460#issuecomment-404756725
If I can reproduce it, I'll probably investigate the noise part.

[musikBear]

If you by PITCH mean the marked one. I can automate that fine in win32 rc7

I also threw one dial from the matrix in a AUT-block. Automates perfectly

[Cangle50]

To musikBear if you can, try it in the 64 bit ver. that's the ver. I use.

[DomClark]

I think the issue is that the automation doesn't always block the processing; sometimes it gets sent in time, sometimes it doesn't. The result is a stuttering sound where sometimes the buffer contains audio and sometimes it is cleared to zero. Here's the waveform of some recorded playback - the left hand side had pitch automation, the right hand side didn't.

[musikBear]

@Cangle50
I cant run x64

[Cangle50]

@musikBear
So your saying you can't hear a unacceptable crackling/static noise when automating the pitch on a continuous note in VeSTige with any vst's run on win x32 rc7?

[musikBear]

@Cangle50

on a continuous note

Ohhh so on a continuous note
thats new..
Yes it cracles

[Cangle50]

@musikBear
Thank you for the response.

On a side note to the Dev's. The crackling noise may not be limited to the automation of the pitch.
If you scrub the keyboard keys up and down in quick sequence you can hear crackling noise, just not as
prevalently as automating the Pitch knob.

[tresf]

@DomClark is this worth milestoning for the 1.2.0 branch, or is this a regression we'll have to live with on 1.2.0? Please classify and milestone appropriately.

[DomClark]

The automation can be fixed quite easily I think, I'll make a PR sometime soon. The keyboard scrubbing might be trickier. I'd like to make some changes to the IPC code on the master branch, outlined in https://github.com/LMMS/lmms/pull/4540#issuecomment-415910129, which would probably fix it, but that's a future thing.

[PhysSong]

Should we revert #4460? It seems like the change introduced some tricky bugs. I'm not sure it's worth reverting #4460 because it has some positive effects.

[PhysSong]

After some thinking, now I think reverting #4460 is an overkill. I was able to hear the noise with manual changes as well, but it was quite slight.
I think the timeout in #4460 which is zero for playbacks can be increased a bit. 1~2 ms should be fine since most changes don't take too long. My concern was long-blocking jobs such as reloading external plugins.
@DomClark Anyway, I think https://github.com/LMMS/lmms/issues/4543#issuecomment-415904784 should be a part of the next RC.