~2.7KHz noise when Bela is connected to guitar pedal ground

lilwashington

My team and I are having a strange issue only when the Bela is connected to our ground. The guitar pedal we are making is a custom design so as not to confuse with proprietary designs and we have control over it. It has an AD823 opamp for signal mixing between an analog passthrough and the Bela output. There is a persistent 2.7KHz ring on the input and output when connected, regardless if the Bela is outputting a signal, but if the Bela is receiving power from a separate source/ground and the analog electronics are powered from their own source/ground, with only the ADC and DAC connected between both, the ring is nonexistent. Since we are making a self contained guitar pedal, this is where the issue lies.

From testing we are very sure the Bela is almost "poisoning" the ground and the ground noise that is injected is audible on the output jacks from the sinusoid generated. It follows the CPU thinking pattern underneath the ring as well. This ground is isolated to one single power adapter throughout, but ground loops may still be the issue.

We are filtering for LPF 19KHz on both the ADC and DAC, and have tried filtering 5V -> GND, but nothing has changed the ring at all. The output is sent to a regular fender guitar amp.

If anyone has had a similar issue before I could not find it and any help would be appreciated as this is the last step to completing the project.

giuliomoro

What Bela device are you using?

lilwashington

Bela Rev C

giuliomoro

One easy test is to try cutting solder bridges JP2 and JP3 to separate GND and GNDA. These are the two solder bridges labelled "GND" on the top layer. You can attempt different configurations: none bridged, only JP2 bridged, only JP3 bridged and see if any of those configurations works for you. The objective is to avoid ground loops: if you are bridging GND and GNDA externally, you'll want to avoid doing that on the board as well.

frothedoatmilk

Hey Giulio, I'm working on the same project. I contacted you earlier this year around the issue regarding single-message SPI communications not using chip-select correctly. We are attempting to build a guitar pedal using the Bela.

These solder bridges were not connected on our board initially? We tried all combinations joining them and it did not affect our hum issue.

We are powering the Bela from a 9V power supply which we are stepping down to 5V for the BBB, and using a voltage divider to create a reference for an external ad823 op-amp in a summing configuration.

We appear to have cleaned up the noise on the input side by isolating our input filtering circuit entirely from our power and output circuits -- the input filter is grounded to the Bela's ADC ground. The only thing which helps the noise is grounding our external op-amp to the Bela's DAC ground.

If you'd like I can send you KiCAD files for our external circuitry. We have tried the following and it has not helped:

Changed the decoupling capacitance between SYS_5V and GND
Introduce a LPF on main power supply
Notch filter on our reference voltage divider for the external op-amp
Tried measuring with no signal going into/out of the ADC/DAC lines
Referenced audio output to both the DAC ground and the power ground. There was a slight reduction when referencing to the power ground
Swapped the voltage divider for a Zener diode

giuliomoro

These solder bridges were not connected on our board initially

The design files show that they are connected ... actually there's a third one to try (JP5), even though that doesn't have exposed copper, so it may need some scraping.

We appear to have cleaned up the noise on the input side by isolating our input filtering circuit entirely from our power and output circuits -- the input filter is grounded to the Bela's ADC ground. The only thing which helps the noise is grounding our external op-amp to the Bela's DAC ground.

So did this solve part of the issue? What part is left then?

frothedoatmilk

The design files show that they are connected ... actually there's a third one to try (JP5), even though that doesn't have exposed copper, so it may need some scraping.

I think we got confused as there is solder on these pads but no visible bridge. Not really sure why I didn't just use a multimeter to check continuity until now. It's been a week!

We're very afraid of tampering with the Bela board too much and breaking it, so I'm just confirming: you're saying that if we remove the solder from these bridges we will separate the grounds? Or do we need to scrape out the underlying copper on all of these pads?

Here is a photo of those pads on our board, just so we're on the same page about what they look like:

So did this solve part of the issue? What part is left then?

The noise has been reduced somewhat, but it's still fairly loud.

giuliomoro

if we remove the solder from these bridges we will separate the grounds?

Solder is just on top of the exposed copper, but there's a trace on the top layer joining the two exposed copper pads, see the design files:

You need a sharp knife to cut the trace, then to undo the cut you'd place solder on the exposed pads. JP2 and JP3 are ready to be cut as they are, but before you attempt cutting JP5, scrape the soldermask off its pads to make sure you can join it back together later.