Heading values changing wildly

[CapnKernel]

pitch and roll should both be zero sitting level once you level the sensors

I just want to check, by "level the sensors" do you mean to hold the module level, or does it mean to run a levelling process?  If the latter, is it the one I posted in the screenshot?

You can plot the raw sensor values using "pypilot_scope_wx" or just dump them to a terminal with something like

"pypilot_client -c imu.compass"

Thanks I'll try this.

Are these values corrected somehow, or are they directly from the chip?

Anyway,   you may need to calibrate the accelerometers.  about 20-30% of mpu9255 I received need this calibration to be able to work properly.

Just to confirm, by "need to calibrate" do you mean do what I showed in the screenshot above?

(A bit strange, the MPU9255 I first tried worked fine at the start of 2020).

As for the compass heading changing wildly..  this indicates either the magnetic sensors are defective, or there is a stray magnetic field nearby.   Perhaps there is ac power or something?

I'll do some more tests.  The problem has occurred in my shed at home, and on the boat.  I have a sensor app in my phone, the heading doesn't have the same problem when the phone is in the same position as the autopilot, so I don't think it's environmental.  The heading is a bit off in the shed, but it's steel.  If it's a signal, perhaps it's coming from the autopilot itself.  But the problem is happening when the motors are off.  I'll try just the RPi and MPU9255, nothing else.

Otherwise you may need a different mpu9255

Thanks, I'll try some different ones.