Container modulators not working

Christoph Hart

No, polyphonic mode is supposed to be polyphonic - all voices are rendered independently. If you can reproduce your issue, I'll take a look.

d.healey

@christoph-hart So far it seems to be playing well so maybe I did something. If it happens again I shall try to capture it in a bottle. Do you get the glitchyness with gain modulators? I'll make a video of the issue tomorrow.

d.healey

Here be the video: https://youtu.be/eWLQJlKH7E4

d.healey

@d-healey Any progress on this issue? Is it something that's just happening on my system or are you getting it too?

d.healey

Bumpy bump... the is the last road block I've come to that's preventing me releasing some libraries so I'm keen to understand if the issue is on my end or not.

Christoph Hart

Ah sorry David I‘ll look into it after the weekend.

d.healey

@christoph-hart Thank you :)

Christoph Hart

Whoah, that's a strange one that doesn't happen on my system (it's so elementary I would have noticed it). Can you check if it happens on a Windows system too? So far I can describe this problem:

if a time variant modulator is active, there is a burst of modulation values for the first buffer in the release phase of a note.
just happens on Linux (?)

Also can you render a audio file of it, then I can take a look at it. Tools -> Record one second audio file was added exactly for this purpose (although it's lying and actually records 8 seconds). Also please supply your buffer size and sample rate as these specs are important numbers to figure out what's wrong.

d.healey

@christoph-hart Could you provide a Windows binary? I don't have a build system for Windows at the moment.

d.healey

@christoph-hart Doesn't matter, I got my Windows using brother to build and test it for me. The result is the same.

As soon as he added the LFO the message Master Chain: Illegal call in audio thread detected I get the same message on my system.

I got him to make the 8 second audio file, it's silent though, is that how it's meant to be?

Here's the link to the audio files (Linux and Windows) and the Windows build we used - https://www.dropbox.com/s/hitn7c6t9ip0ffv/HISE.zip?dl=0

We both had buffer set to 512 and Sample Rate set to 48. He also tried 1024 and 256 but it didn't change anything.

Christoph Hart

The error message is a result of the over-verbosity of the new audio thread guard system, which monitors unwanted operations in the audio calculation, but sometimes it fires (mostly during initialisation of modules) without a serious reason. I tried to get rid of false positives, but some slip through. TLDR: shouldn't be the cause of the problem.

Weird that the audio file is silent - silly question: have you played something :)?

Start the recording by selecting the menu item
During the next seconds, hit random keys and make sound
Get the HISE audio output as .wav file

If this doesn't work, can you load HISE into a sequencer on your Linux system, and export the glitch as audio file (or any other way of recording the HISE output if the inbuilt tool fails for whatever reason)? Also if you run the Debug build, does it fire a assertion when the glitch appears?

Christoph Hart

Also maybe I didn't follow your exact steps:

Add a sine generator
Add a LFO modulator
Play notes -> glitch at the note release

Have I missed something?

d.healey

@christoph-hart I did not play any notes, that explains that issue :) I shall make a new file.

Your steps are correct for creating the glitches. The other issue can be demonstrated just by adding an LFO to the master chain, it doesn't move at all and there is no display in the plot. I haven't checked the debug build, I will do that today.

Christoph Hart

Don't worry about the LFO in the master chain, this one I can replicate :)

d.healey

@christoph-hart I've updated the files in dropbox now. No glitches on Windows. Going to test debug mode on Linux now. There was a jassert juce_file.ccp line 134 on Windows in debug mode.

d.healey

Debug build on linux:

JUCE assertion failure in ModuleBrowser.cpp:158 - this just happens on initialisation
I get a lot of nasty noise in the debug build - I've added another 8 second audio file to the dropbox zip file demonstrating this - but no assertion messages related to this.

With more careful listening I think the glitches are actually present even without the LFO. I'm not hearing them as strongly in my sampled instruments as I do with just the sine wave generator though.

I also tested the Windows build in a virtual machine on my Linux system and there was no glitches so this does seem to be a Linux issue.

Christoph Hart

Hmm, it seems as it sets every 8th sample, but the rest is silent, which correlates with the downsampling factor of the modulator system.

Can you change the downsampling factor and try again? It's a preprocessor macro in hi_core/LibConfig.h:

#ifndef HISE_EVENT_RASTER
#define HISE_EVENT_RASTER 8
#endif

Around Line 66. Change it to #define HISE_EVENT_RASTER 1, recompile and try again. If this solves the issue, we're still not sure what caused it, but on the right track :)

Christoph Hart

BTW, the best way to debug issues in the modulator system is by creating a dummy wave file with a fixed value of 0dB and loading it in the Loop player:

DC Wave file

(Make sure you turn your speakers down, it's "loud"!)

d.healey

@christoph-hart That seems to have fixed it

Christoph Hart

Found it. It was indeed a code path that was just executed on Linux and contained low-level SSE optimizations that didn't work apparently :)

I've just pushed the change. Please check if it solves it - and make sure you roll back the manual change you did otherwise you won't be able to test it...

Container modulators not working

15

1.8k

10.4k

90.7k