Noisy setpoint by analog input

[dlmed]

Hi, I’m sending a torque setpoint to the GPIO15 of the ODrivePro from another device. This is the GPIO config:

odrv.can.config.protocol = Protocol.NONE
odrv0.config.gpio15_mode = GpioMode.ANALOG_IN
odrv0.config.gpio15_analog_mapping.min = -0.284
odrv0.config.gpio15_analog_mapping.max = 0.284
odrv0.config.gpio15_analog_mapping.endpoint = odrv0.axis0.controller._input_torque_property

The min and max values correspond to the limit torque given by the current limit and torque constant. When I plot the setpoint in the GUI Inspector I see it’s very noisy and I already checked from the device that sends the analog signal that the noise is not coming from there (stable current in the line). Here’s how it looks:

This is the output value coming from the device that is sending the signal:

So at max it should oscillate of about 0.0001 Nm and not 0.026 Nm. Is it expected to be this noisy the analog inputs of the ODrive Pro? would it change if I use another GPIO? Is there another way to reduce the noise? Thanks in advance.

[solomondg]

I’m immediately thinking of ground loops here – the analog input isn’t isolated, so ground bounce between the sender and the ODrive could definitely be causing this inaccuracy. Could you sketch a quick diagram of your system – how things are wired together, etc?

[dlmed]

Here is a rough diagram of the connections:

When I turn off the 48v power supply the noise reduces:

Then increases when I turn it on:

I tried connecting the 2 grounds of both power supplies:

But very similar results… (internally the bus and AO GND have around 1mV difference)

[dlmed]

According to the AO Module datasheet, the bus and the AO should be isolated:

[solomondg]

Interesting. I’d add a very small RC filter, maybe 1kohm + 100nF (tau = 1.6kHz, ODrive samples analog at 1kHz). There may also be an additional consideration with how we do analog reads on the Pro – if you’re still getting noise with the RC filter, I can look into that.

Generally, I would recommend against doing analog input – it’s a more nice-to-have than a strongly recommended method of control. Any sort of digital input (USB, UART, or CAN) is definitely recommended over analog, or maybe R/C PWM (1-2ms pulse width) if possible from the PLT?

[dlmed]

It was the ODrive’s power supply the culprit… I changed it to another one of better quality and there’s way less noise in the GPIO analog input. Idk why I assumed that the Regen Clamp would smooth the power supply voltage ripple, my bad. I also tested the RC filter and it seems to smooth out the signal, but gets rid of some high frequency signals (not noise), so I prefer to leave it without it as with the new power supply works fine. Thanks for the support.

[solomondg]

Fully understood! Yes, the regen clamp will definitely smooth supply ripple, but many supplies (at light load) will expand the acceptable tolerance of the output voltage, which is a bit of a different ball game than high-frequency noise filtering. Very curious the old and new power supplies, if you’re able/willing to share.