sine waves with envelope in poly~ clipping


    Jun 19 2013 | 3:48 am
    Hi!
    I am having troubles with the poly~ object. Basically I built a sampler for sustaining sine waves. I have a cycle~ with envelope inside a poly~ and I am giving it different frequencies through a $1 message. This works fine for a few minutes then suddenly it starts clipping and sometimes even the sound stops. I then have to turn off the audio and switch it back on to continue playing.
    Anyone any idea why this is happening? (see attachment)
    Thanks,
    Lilian

    • Jun 19 2013 | 9:21 am
      I think it's getting too loud - running up to 1000 voices of anything typically requires quite controlled envelopes.
      The breaking up you're hearing might also be caused by heavy cpu load? I know that if I max out your patch my 2.66ghz i7 mbp reaches at least 140% cpu.
      It seems as if the only thing that'll truly stop a note in the patch is note stealing... and that happens @ 1000 poly. Reducing that to 100 seems to completely solve the problem.
    • Jun 19 2013 | 10:34 am
      Thanks for your answer!
      I had reduced the voices to 100 before - without success though. Even if I reduce to 20 I still get the hearing knock out, while my CPU (in Max) shows no more than 1 or 2 %. The overdrive / knockout happens faster when I am using higher frequencies (like above 8000Hz). It seems as if they are "ganging up" among themselfs like a feedback loop. So my guess would be that they somehow create intermediate tones. I don't understand it though.
    • Jun 19 2013 | 10:51 am
      I definitely am not seeing that response here.
      The ganging up, does that still happen if you deactivate the adc~-triggered notes? As in, starting some notes, and then not letting more get triggered?
      I'd also start looking at the envelope - that function looks very odd indeed, all the values are bunched up off to the side, and if I'm reading it correctly it's a very fast envelope, too? - try swapping it out with a straight "0, 0.05 1000" message, perhaps?