Re: [Paparazzi-devel] More questions about AR.Drone2 magnetometer scaling

[Christophe De Wagter]

Dear Ori,

From this graph you can't tell a lot. Scaled is not the correct term, they are bias+scaled (scaled=calibrated), and yes indeed it is not uncommon that xy are scaled differently than z.

The only plot that does tell if something is totally off is if you make the same plot while turning the quadrotor in all directions, not just horizontal but also up and down and upside down to have the mag vector point in all possible directions. The calibrated plot of vectors should afterwards be a nice sphere. The raw mag is typically a slightly squeezed sphere with an offset (and different scale). 

-Christophe 

On Mon, Feb 9, 2015 at 4:21 AM, Ori Pessach <address@hidden> wrote:

Hi,

I'm still seeing weird behavior from the AR.Drone when it comes to magnetometer readings, and I was hoping someone could help me understand what I'm looking at.

I logged some magnetometer readings while the drone was on the ground, pointing approximately to the north. Then, I plotted the raw and scaled magnetometer readings. I would expect the 3D vectors described by the readings to point roughly in the same direction, but instead, they look like this:

https://www.youtube.com/watch?v=TewLXoCXcSQ

(The green vectors are the scaled values, red are raw. I scaled both datasets to be visible in the same chart, and flipped the scaled values' Y axis because I think I saw the Paparazzi code do that.)

The Z axis seems to have a much larger multiplier than the X and Y axes. I have to admit I don't understand how this is supposed to work. The calibration process seems pretty ad-hoc - it observes the measured values and then fits them to a range based on the min/max values it saw during the calibration... Right?

Anyway, from my reading of the AK8975C datasheet I don't get the impression that the Z axis should be scaled so differently from the other two axes. So I'm probably missing something, but I'm not sure what. Does this look normal to anyone?

Thanks,

Ori

_______________________________________________
Paparazzi-devel mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/paparazzi-devel