This is with pepper I assume? Is it an issue only with the audio outputs? You could try adding a pull down resistor (maybe 1k?) to ground from each of the audio out pins (R25 and R56). There's something else that could be done in software which leaves the audio codec always running ... but I guess a hw solution as suggested above would be overall more reliable (assuming it works)

    EDIT: Yes, with Pepper.
    giuliomoro It is a bit unclear to me. Which sides of R25 and R56 are the correct ones to connect to 1K resistors, and where is the next suitable ground on the board?
    alt text

      @giuliomoro Example for the left channel. Does R1K need to attach between IC and R56, or on the other side of R56?
      alt text

      Sorry for late response. Either side of R56 should be fine, although on second thought I am thinking that maybe none will work. I need to have a look at what happens to the input voltage of the opamp but I am in the middle of a move and haven't unpacked the the oscilloscope yet. We may need a software solution but also should wait until I am able to work properly again

        11 days later

        sorry not yet, turns out that moving is more complicated than it may seem !

          a month later

          giuliomoro sorry not yet, turns out that moving is more complicated than it may seem !

          I'll give you anytime you want, hehe.
          In the meantime: Bela/Pepper emits CV on main audio output, even when running software doesn't produce any sound.
          My own tiny osci (eurorack, vpme.de) shows
          - a flat DC+ of approx. 0.1 - 0.2 Volts on the right channel. Left channel shows same, but with added noise floor.
          Muting this after outputs produces click exactly like it does, when Bela shuts down a patch to load another. That then also clicks upon patch loading/start.

          It still feels like there could be a software solution soft-fading out internal levels before audio engine shutdown and vice versa upon patch load. What do you think?

          Would soft-mute be more easy, if i converted finished PD patches to C++?

              jayrope - a flat DC+ on the of approx. 0.1 - 0.2 Volts on the right channel. Left channel shows same, but with added noise floor.

              That's not so unexpected: we use the NE5532 opamp with a decent amount of gain there and it has no offset correction, so a (relatively) small offset is to be expected. This could be compensated for in software if needed.

              jayrope when Bela shuts down a patch to load another

              What happens there is that the codec is powered off and then powered back on again. You'd expect that the differential amplifier configuration would make that result in a quiet fade to 0, but it seems it's not the case. The alternative is to leave the codec running throughout, in which case you get a click only when the codec is started about 10 seconds after power up. With the new product launch next week I will merge into the master branch some code that allows to do that.

                    8 days later

                    jayrope Every time i stop a patch and load a new one i get a pretty loud click upon stop, and another one of those upon start of the next patch.

                    The cause remains, but it is smaller now. I had set Pepper's output jumpers (j21/j22) inccorrectly and so was using Bela's built-in outputs instead of Peppers opamp.
                    Now a DC offset is mostly gone, which was the main cause for the clicking. There is still a low volume clicking, however much more tolerable for live or recording purposes (noise is politics, right?)