Distortion Woes Made Worse by Screen Graphics and Certain Mixers

tony10000 · March 6, 2025, 4:31am

Really bad distortion that isn’t there in the VCV Rack file.

PCB Version p11
Firmware version v2.0.0-dev-12.2
Bogaudio 2.0.0-dev-12.0

Distorted.vcv (3.2 KB)
Distorted.yml (12.5 KB)

tony10000 · March 6, 2025, 5:12pm

Can anyone duplicate or confirm? @danngreen?

danngreen · March 6, 2025, 6:20pm

Yes I hear it. If I run the FM-Op output directly to a panel jack, then I don’t hear any of that noise, so it must be coming from the Mix8. Also, playing around with it a bit, I can get that type of noise to happen on channel 1 of the mixer (the VCO->VCF voice). So that also points to the problem being the Mix8.

tony10000 · March 6, 2025, 6:47pm

Hi Dann:

Unfortunately, that does not seem to be the problem. I changed out the BogAudio Mix 8 for the NANO Performance mixer and am having the same issue.

Distorted 2.vcv (3.2 KB)
Distorted 2.yml (12.4 KB)

tony10000 · March 6, 2025, 6:56pm

Hi @danngreen

I think I figured it out. It has something to do with the screen display. If the display shows the instruments, I hear the distortion. If I hit the white back button and go back to the “load patch/new patch/settings” screen, the distortion disappears. Try it and see if you can repeat it with the “Distorted 2” file I uploaded.

tony10000 · March 6, 2025, 7:09pm

@danngreen

The display introduces a strange IM distortion into the signal. Very strange. It seems to increase with amplitude increase according to my tests. Also, when I exited the display by hitting the back button, the MM crashed and locked and played a single note. I have had that happen before. Hope that helps!

tony10000 · March 6, 2025, 7:18pm

@danngreen

Went back to the original “Distorted” file. I can repeat the behavior there. The distortion seems to be worse on the first channel with the BogAudio VCO/VCF and increases with amplitude. It is greatly reduced when not showing the instrument display.

danngreen · March 6, 2025, 7:44pm

Interesting, yes when I go to the Main Menu then the distortion gets less. But it’s still there, I can clearly hear it if I isolate one channel.

I’ve re-done how displays work so they’re much more efficient now, and will be releasing that version shortly. But I still get that noise… Looking into it…

tony10000 · March 6, 2025, 9:15pm

@danngreen Yeah, it is still there. For some reason, it is much worse with the BA VCO/VCF…not sure why. Could be the plugins or just the frequency range.

danngreen · March 7, 2025, 1:06am

Yeah, this is strange. I need to test more to see if it happens with any other modules. It seems like changing a small thing makes it go away (even saving the patch from within MM, and then reloading it will make it go away).

tony10000 · March 12, 2025, 2:53am

I was playing around with the Plaits clone with BogAudio Mix 8 and got really bad noise and distortion…so, there may be a problem with the BA mixers after all.

danngreen · March 12, 2025, 4:55pm

Could be. I didn’t find a bug in the mixers, but I did find and report a memory leak in Bogaudio modules (technically all of them, but practically speaking it only effects ones with large memory buffers like CVD). Matt fixed it right away in the main Bogaudio repo, so that fix will be present in the MM version soon:

github.com/bogaudio/BogaudioModules

Memory leak due to virtual function call in destructor

opened 09:28PM - 10 Mar 25 UTC

closed 04:39PM - 11 Mar 25 UTC

danngreen

I saw some evidence of memory leakage when using the MetaModule port, where memo…ry is more constrained and leaks more noticeable. I then confirmed it also is happening in VCV Rack with the latest Bogaudio plugin (2.4.45) from the VCV Library. To reproduce: 1. Open memory monitor (e.g. `top` on Linux, or Activity Monitor on macos) 2. Open VCV Rack and make a patch with 64 CVD modules 3. Select all the CVD modules and do Copy 4. Delete all the CVD modules 5. Do a Paste to put the modules back 6. Repeat steps 4 and 5 a few dozen times (delete, paste, delete, paste, etc..) 7. The memory usage should go up by roughly 100MB with each paste, and will not go down when you do the Delete. Making a new patch instead of deleting will have the same effect. The math for the ~100MB with each paste makes sense, as CVD has a 10 second buffer at 48kHz, which is about 1.8MB, times 64 modules is around 117MB. So, looking into the code, I think I found the issue (and I learned some trivial about virtual destructors in C++) In the virtual base class BGModule, we have a destructor that calls removeChannel(): https://github.com/bogaudio/BogaudioModules/blob/b45d6b93453d0ff72018acf23b28e4f874713a40/src/module.hpp#L31-L36 This is supposed to call the derived class’s `removeChannel()`, which is where the memory is supposed to get cleaned up. For example, CVD: https://github.com/bogaudio/BogaudioModules/blob/b45d6b93453d0ff72018acf23b28e4f874713a40/src/CVD.cpp#L23-L26 But, as I learned today, in c++ when you call a virtual function from a base class destructor, the **base class** function is run, not the derived class function. Source: https://isocpp.org/wiki/faq/strange-inheritance#calling-virtuals-from-dtors So this means when the CVD module is deleted, its `removeChannels()` is never called, and the memory is never freed. I didn't believe this at first and had to put some `printf`s in the `CVD::removeChannel()` and in the `BGModule::removeChannel()` to see that indeed, the CVD removeChannel is not called, but instead the empty BGModule one is run. One solution would be to put the call to `removeChannels()` into the derived class destructor. But this would need to be done for every Bogaudio module. I haven’t been able to come up with a solution that doesn’t involve changing every module. But it could maybe be made easier with helper function or something that every derived class destructor calls. Anyways I'm happy to help implement a solution to this, as I need to make it work on the MetaModule platform. Right now I'm exploring the helper function idea and using some regex to "patch" all the modules, but if you have any preference for a method of handling this, then I'm happy to do it that way so my fork can stay more-or-less in sync.

tony10000 · March 12, 2025, 5:36pm

OK…good to hear. Looking forward to it!

tony10000 · March 21, 2025, 5:42pm

Had some more problems with noise and distortion, this time with the Plaits clone. Isolated the problem to the Bogaudio MIX8. Substituted the Befaco mixer and the noise and distortion disappeared. Not sure why this is happening.

synthetictheory · March 27, 2025, 9:34am

The issue also occurs on my end when using the Plaits module with a Befaco STMix. The problem improves slightly when I exit the module’s screen. However, when I connect Plaits directly to the MM outputs, there is no distortion or clipping. Interestingly, I tested several mixer modules (each connected to a different set of MM outputs) within the same patch, routing Plaits into each one. After doing so, the signal on the Befaco STMix no longer showed distortion. It seems there’s something unusual with the signal flow. I’m running MM on v2.0.0-dev-12.9.

tony10000 · April 16, 2025, 11:37am

@danngreen Still getting that crunchy distortion, this time with the Befaco Mixers.

Firmware version v2.0.0-dev-13.8 32K/64

It has something to do with the screen graphics during a busy sequence. It is really bad when the module graphics are displayed and less noticeable when you change to another screen. Test file below.

crunchy graphics.yml (12.4 KB)

tony10000 · April 16, 2025, 9:40pm

I have spent some time doing some tests this afternoon. The problem seems to be specific to how the Energy plugin interfaces with Befaco STMix.

It is very similar to the problem with how the Plaits clone and FM8 sometimes interact with the Bogaudio Mix4 and Mix8.

I tested Energy with Mix 4 , and there is no noise at all.

Very strange bug.

Best practice until this is fixed is to try and isolate the noise source and try another mixer.

danngreen · April 16, 2025, 10:28pm

Yes, this is a strange one. It’s on my radar – just am trying to tackle the plugin unloading bug now

tony10000 · April 16, 2025, 11:19pm

OK…I will keep testing and let you know my observations.

tony10000 · April 22, 2025, 5:00am

I have been doing more testing on this issue…this time with the BogAudio FM Op and the Mix 8 combination. I believe that the noise and distortion that I am hearing is quantization noise. I am not sure how both of these plugins are coded, but they seem to run fine in VCV rack with no quantization noise. So, there is something going in with the code conversion and/or the audio subsystem in MM. I am not sure what bit depth the MM uses, but the noise can be reduced by increasing the sample rate. It is also exacerbated if the video is on the module display but not sure why.