Maximum resources per chip: audio error with synth containing multiple generators

[inscripttech]

link to the relevant synth file (via Dropbox)

The above synth file contains two audio-rate noise generators and four oscillators producing low frequency sine waves. The resulting output when a note is triggered produces dropouts that sound like buffer underruns, but I'm not too sure.

Is there an upper limit on the amount of CPU that a chip can use? This patch with a few envelopes would be akin to a duophonic synth voice and it appears that the audio engine is struggling to produce even the generation portion.

Perhaps I am missing the use of another module that could optimize patches of this nature?

[kometbomb]

I think it might be the filter constantly getting a new cutoff value - how it works is that changing the filter parameters causes a small precalc to happen, which could technically cause more load even though simply using the filter is not that heavy. Or the whole filter might get unstable or something? It is not the best filter out there. I have no audio at the moment, will check later.

[kometbomb]

I just checked and it seems the CPU use is just so high that the single thread (each channel runs in its own thread) cannot process the synth fast enough. On my desktop it seems to run fast enough but it IS heavy.

[inscripttech]

Interesting. Can you explain a bit about why that would be? I'm running a quad-core 16gb machine so it must be some particularity of the audio engine. I have nothing against the constraints, for the record. Just trying to figure out where exactly the limit is.

[kometbomb]

It is quite inefficient because of how you can have the feedback loop back to the signal chain so it goes through the whole "tree" of modules for each sample. A different way of doing it would be to render longer buffers of audio and apply e.g. the filter over the buffer, then a distortion etc. but that makes it impossible to have the feedback (you can't send back signals at every sample because the modules don't run in paraller as they are running now). I think it might be interesting to have some kind of CPU monitor/profiler kind of thing for the synths to find the hot spots. I think at least the filter might be one heavy-ish part, since the cutoff changes constantly. pe 29. toukok. 2020 klo 9.52 inscripttech ([email protected]) kirjoitti:

…

Interesting. Can you explain a bit about why that would be? I'm running a quad-core 16gb machine so it must be some particularity of the audio engine. I have nothing against the constraints, for the record. Just trying to figure out where exactly the limit is. — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#43 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAIIV2OXIAWMRNB6Q7YSOF3RT5LUJANCNFSM4NMXTQBA> .