User Patch Noise Update

markemorse 2018-06-06 10:59:33 UTC #21

Just to add another voice to the chorus, because sometimes this matters
I'm on v12, I've been working with different power supplies all week hoping there was one that would reduce the noise enough for me to use the OWL onstage this weekend, because I'd kind of (eh, stupidly) built an arrangement around it. I bought an Ojai (before reading this thread, which amazingly works sometimes) as a last resort, and it's absolutely no different, this +15-20db F# whine. So today I'm going to buy some cheap shimmer verb to use instead, point being that it remains a bummer it's not stage-ready.

bfabricius 2018-10-09 17:07:59 UTC #22

Hi @mars,

bumping this thread since ive looked into OwlWare a bit recently. A friend of mine builds eurorack modules and he has produced a general purpose (programmable) module (small batch, hand assembled) and he doesnt really know what to do with it now. It's pretty much the same target architecture that the Owl Modular has, with a few minor differences. I'm thinking we might be able to get OwlWare to run on it.

While i was diving into the FW code, I actually remembered this noise issue I am experiencing with my Owl Modular. I was interested why the noise goes away when i run the Owl in remote control mode with OwlControl.

Just to confirm, I can see you are doing this when in remote control

github.com

pingdynasty/OwlWare/blob/495cc9aea1e4ecec2aef531fdb9e3a21dba6fdd4/Source/MidiHandler.hpp#L103


  break;
case PATCH_PARAMETER_D:
  midi_values[PARAMETER_D] = value<<5;
  break;
case PATCH_PARAMETER_E:
  midi_values[PARAMETER_E] = value<<5;
  break;
#endif /* OWLMODULAR */
case PATCH_CONTROL:
  if(value == 127){
	memcpy(midi_values, getAnalogValues(), NOF_PARAMETERS*sizeof(uint16_t));
	setParameterValues(midi_values, NOF_PARAMETERS);
  }else{
	setParameterValues(getAnalogValues(), NOF_PARAMETERS);
  }
  break;
case PATCH_BUTTON:
  if(value == 127){
	togglePushButton();
	midi.sendCc(LED, getLed() == GREEN ? 42 : 84);
  }

which explains that the paramter values are not taken from the ADC but replaced by the ones transferred via midi cc from OwlControl. Is that right? I havent checked this yet, but Im assuming the OwlControl sensitivity defines the parameter smoothing in OwlControl. At least its more noisy when i have it running at Low than at High sensitivity.

If it is, then I can see that if not in remote control mode you take the parameter values from ADC and smooth them in the leaky integrator in PatchProcessor of OwlWare:

github.com

pingdynasty/OwlWare/blob/495cc9aea1e4ecec2aef531fdb9e3a21dba6fdd4/ProgramSource/PatchProcessor.cpp#L58


 * y(n) = (1-alpha)*y(n-1) + alpha*y(n)
 * with alpha=0.5, fs=48k, bs=128, then w0 ~= 18hz
 */
for(int i=0; i<NOF_ADC_VALUES; ++i){
#ifdef SMOOTH_HYSTERESIS
  if(abs(params[i]-parameterValues[i]) > 7)
#endif
    // 16 = half a midi step (4096/128=32)
#ifdef OWLMODULAR
    if(i<4){
	parameterValues[i] = (parameterValues[i]*SMOOTH_FACTOR + 4095 - params[i])/(SMOOTH_FACTOR+1);
    }else{
	parameterValues[i] = (parameterValues[i]*SMOOTH_FACTOR - params[i])/(SMOOTH_FACTOR+1);
    }
#else /* OWLMODULAR */
    parameterValues[i] = (parameterValues[i]*SMOOTH_FACTOR + params[i])/(SMOOTH_FACTOR+1);
#endif /* OWLMODULAR */
}
// for(int i=NOF_ADC_VALUES; i<NOF_PARAMETERS; ++i)
//   // todo!
}

Again Im running an old v14 fw build and i can actually see the parameter values jumping around in OwlControl when not in remote control. This could explain the prominent noise when running a majority of patches (e.g. Bank A / Overdrive ). Im wondering what would happen if i build OwlWare for OWLMODULAR from head in master (since the leaky integrator code is definitively active there). Hopefully that will solve the mystery of the noisy parameters

Thanks for the help ( I know you probably have other things to do regarding OpenWare but this little experiment is a nice way to get into OwlWare ).

Cheers,
Ben

mars 2018-12-03 12:08:43 UTC #23

Hi Ben, sorry for late reply!

Yes the smoothing was removed in a code revision, the reasoning was that a new parameter handling method in the patch code would take care of it instead. But this has not turned out as planned: some newer patches use the new FloatParameter constructs (with smoothing) while all older patches and many newer ones still don't.

Also in hindsight it makes more sense for the device-specific firmware to smooth the parameters it knows comes from the ADC, rather than for the device-agnostic patch to do so.

mars 2018-12-03 12:19:35 UTC #24

Note by the way that the PatchProcessor in OwlWare only processes the parameters for the statically compiled factory patches. Dynamic patches have their own PatchProcessor in the OwlProgram repo.

bfabricius 2018-12-04 08:19:27 UTC #25

Not a problem

Im actually not quite sure what fw version i was using on my OWL modular last, but I traced the noise down to most probably being ADC control values that werent smoothed causing it since when control was delegated to midi the noise disappeared. Could be the filter was not working or that I tried a newer firmware without smoothing and old patches that didnt use FloatParam.