Pure Data Audio Thread question

Genzo3

Hi! I am pretty new to the Bella world, but I've almost finished my first project. When I play it alone without the computer, I get some glitches in the sound. I've read a lot of post about the "audio thread" and I have some questions ... I've read that "reading files and sockets at every audio callback from within the audio thread is bad practice in general" but, what does it mean? Do I have to put in the _main.pd file only the audio sums and in separate PD files the calculations? I'm not getting a clear view on what's the most efficient way of programing in PD for the Bela platform. In my computer the PD file (only one) works perfectly ...

Could someone throw some light tothese questions? Thanks!

giuliomoro

Long story short: objects that perform expensive, infrequent computations may cause underruns even when the average CPU load is low. This is because such expensive computation takes longer than the length of one audio buffer. The real solution in this case is to re-write the object so that the computations are split evenly across buffers. However, this is a huge effort that would change some core principles of Puredata. The practical solution to this (which is what Pd does on your computer with the "delay(ms)" parameter in "Audio settings") is to increase the size of the buffer. On Bela you can do this by increasing the blocksize of the project (Project settings in the IDE).

Some more in-depth considerations on this topic can be found at the links below and in many other old posts on this forum:
https://lists.puredata.info/pipermail/pd-list/2016-09/116224.html
https://forum.bela.io/d/490-underrun-detected-errors

Genzo3

Thaks Giuliomoro for your answer ... I've increased the size of the buffer and I got rid of the underrrun errors with the Bela connected to the computer throught the IDE window, and the CPU is under control, but when I unplug the Bela from the computer and run it with an external battery I inmediatly hear the clicks ... Why is that? Sorry if my questions seems to be stupid ... I'm pretty lost ... And by the way, the problem happens with the implementation of the PD created by Johannes Kreidler for pitch shifting / granular synthesis ...
alt text

giuliomoro

I think this could be because the project settings (including blocksize) of the program when it runs at startup are those that were active in the IDE at the time when you first set the program to run at boot, so your auto-boot program is probably still running with the old blocksize (16?).

To fix this:
- make sure you have the correct block size selected in the IDE and the program is running fine
- go to the IDE settings and select "Run project on boot" > "None"
- go to the IDE settings and select "Run project on boot" > "YourProjectName"

this way the settings for run at boot will be updated.

Genzo3

That was it!!!! I'm hearing it now without any glitches! Thanks a lot! I was almost crazy because in the IDE I didn't heard any problems, but in "external battery mode" the problems were back!
I'm so happy now that I'm going to buy a second Bela for my next project RIGHT NOW! I was still waiting for a second one because with the glitches the program and the hardware were useles for live performance, but now it's a different game ...

Thanks again!

giuliomoro

Genzo3 I was almost crazy

Sorry about that, I think we should avoid this confusion, so I created an issue here

giuliomoro

@Genzo3 incidentally, I would be curious to see the patch that was causing you trouble: the one you pasted above does not seem to have anything critical in it.

Genzo3

It seems I can not attach a PD file ... With a screen capture you won't be able to se the pd subpatches ...

giuliomoro

If they are just subpatches, you can just open the _main.pd file with a text editor and paste the content here. If there are abstractions involved, maybe you have a git repo somewhere?

Genzo3

Here it goes ...

giuliomoro

Thanks (I edited your post by hiding the URL to the github gist, otherwise that would show the 2000+ lines in this page!) , it's maybe the [env~] that is causing troubles, similarly to what was happening here.