need advice on calibration, mainly issues with sim3 --mag

[Arne]

Hi,

I'm trying to get a usable result out of the calibration process for quite some time now, but so far I'm still stuck at the sim3 --mag part. The MAE keeps shooting up in the air. Every time one or two IMU_MAG_ALGN_XX parameters gets larger the 10^5, sometimes up to 10^9.
I did both calibrations on two completely different places (more the 5 miles away from each other). My copter is a heavy octo, which makes the dancing part pretty hard, hence I even build a calibso machine... but I still can't get a valid parameter set . Can you please take a look at my graphs and give me a hint what to try next?

static temp

static rate
looks pretty noisy, but I think considering the y-scale it's ok!?

static acc

static mag

that's a log file from a machine run, the graphs are much much smoother than the manual ones
dyn rate

dyn acc

dyn mag

[JussiH]

Looking at the data, it is pretty smooth, but probably going too fast...try and slow down your machine.

Rates are OK, the noise is very low magnitude, so it should work.

Do show us a photo or video of your calibso machine - Im sure a lot of guys will find it very interesting!

[Arne]

No fancy photos/videos yet, but there is a sneak preview:

DSCN0340_klein.JPG (497.22 KiB) Viewed 14614 times

OK, i'll try a slower one.

[kinderkram]

Hi Arne,
this is a very nice machine. I've planned something similar but never got to it.
Looks like a simple and cheap solution - thx for showing it off!

Norbert

[Kisssys]

Arne,

I believe each axis needs to reach a minimum value and maximum value at the same time with each other axis. Without rotating the center also I don't believe you are achieving that. Maybe if you spin it long enough it works out.

You can display 2 axis at a time on the graph and make sure that occurs.

I have a harness I hang from a tree to calibrate my big machines, I had to add a slip ring in the middle to achieve that goal.

Cheers

[JussiH]

[Lindbaergh]

[Arne]

[kinderkram]

...or rotate the given frame completely.

[Arne]

I'll add a third frame to the calibso machine to cover the all dimensions, then it hopefully works as expected.
Anyway, I have a question about the IMU_MAG_INCL param. It has to be set after the cal -acc stuff, afterwards you run cal -mag and that will output a new IMU_MAG_INCL param, which will be copy'n'pasted to the param file. Ones transmitting the param file to the AQ the later IMU_MAG_INCL param will be write into the ROM. Is that correct?

My actual IMU_MAG_INCL is about -67 for Berlin, the calculated param is about -3.7e-004. I took a look at the great video tutorial made by terramax and the behavior is the same at first, but in the video somehow the correct IMU_MAG_INCL is transmitted to the AQ. Do I miss something here?