Occasional erroneous values from i2c magnetometer

[jschall]

Issue details

This is a plot over 15 hours:

There are occasionally these huge outliers from the first magnetometer.

Version

Based on dab0d8e

Platform

[ ] All
[ ] AntennaTracker
[x] Copter
[ ] Plane
[ ] Rover

Airframe type

Quad

Hardware type

Pixhawk 2

[gmorph]

Did you notice anything else happen in the dataflash logs at the time of the blips?
Thanks, Grant.

[jschall]

Nothing at the same time.

[tridge]

i2c doesn't have any checksums, so this could just be bus noise.
We could try and include some sort of spike filter, but really the EKF is already doing that for us.

[tridge]

if we double sampled we may be able to detect these, but it would halve the sample rate and that could cause aliasing issues

[OXINARF]

Can this be closed?

[WickedShell]

I'm closing it and pointing to the I2C DNF as the fix. I could reproduce this problem before with noise on the I2C bus, #5935 if anyone disagrees this can be reopened, but at a minimum would need to prove that it still happens now after that patch.