DIMU with very short full calibration results

[chestnut]

Thank all, and Menno. Now I'm testing. I'll report it for all.

[chestnut]

Dear all,

Thanks for support full calib. It's working good. This is some video first test with full calib
http://www.youtube.com/watch?v=MMVSpjLOdWQ
http://www.youtube.com/watch?v=8de9_nkN ... e=youtu.be

[afernan]

I´m just received a board with DIMU inside. I have none experienc on that.
Could someone summarize the steps for a short/full calibration? What QGC to use?

I guess is something like:
1. add the static files from DIMU
2. add the dynamic files from DIMU
3. Check "using DIMU" box.
4.- start the step's.
5. On some steps, you have to abort it by Hand, if the lowest values reached . Which ones?


Thanks

[kinderkram]

First of all, click this: http://autoquad.org/wiki/wiki/autoquad- ... us-upgrade

[afernan]

I´ve read the wiki and say that for DIMU V1.01 board to add following line in config.h

#include "board_dimu_v3.h"

But in r297 that file v3 was renamed to v1_1, so I guess now we should ads that v1_1 file instead v3. How can be sure that we´re using DIMU instead AIMU?

Is that correct?

Thanks

Angel

[joebar.rc]

Yes, correct. Once you add the dimu header file in the config.h, the FW will use the dimu. (There are checks in the imu.c for that)
If you btw do that without having a dimu, the AQ will not continue with the boot

[JussiH]

Regarding cal step 2 and 3 with DIMU, this is what worked for me:

For the cal runs 2 and 3, look at the averages. For Acc you need it close to 9.806 and for mags the average needs to be close to 2. This produced the best results for me in the 3 calibration runs I have done with DIMUs so far.

Let Cal step 2 and 3 run too far and the averages will go up and eventually the calc will crasj.

[afernan]

@Menno:
thanks. The AltHold is AWESOM! I´m testing indoor (lot of room turbulences) and I can´t believe the precission! This is a great improvement.

Also the SIMPLE leveling is very useful. Unfortunally is not kept after switch off. For me this leveling works better that the one obtained by the calibration. Anyway I had lot of problems with calibration.
This is the best I´ve got (3xstatic files, three DIN tries):
STEP 1: MAE = 0.0006
STEP 2: MAE = 0.02
STEP 3: MAE = 0.52
STEp 4a: MAE = 0.064
STEp 4b: MAE = 0.064
STEp 5: MAE = 0.064
STEp 6: MAE = 0.23

I repeated several times step 2 and 3. Sometimes MAE went to very high values. Repeating, then went better. I bit extrange behaviour.

Angel

[afernan]

#Jussi:

i´m tring to do what you propose. Using DIMU box If I choose all three static files, no good results.Using only one (the one with Z vertical) I get this (what avrg should I look?)
*****STEP 2*****
MAE: 0.0241243
AVG: 9.79335
Dynamic:
MAE: 0.0544428
AVG: 9.80436
Dynamic acc stats:
MAE: 0.0544428
AVG: 9.80436
Static 0 acc stats:
MAE: 0.0241243
AVG: 9.79335

****STEP 3****
MAE: 0.533174
AVG: 1.67327


Angel

[JussiH]

Yes, DIMU cal seems to work best with just one good level static file

You are looking at the static average. Its ok at 9.79335. Was this a file that was generated with the board in a orthogonal position (not level)? I have noticed that static files made the board level usually has a closer to ideal average.

Your dynamic acc averages are very close to ideal.

For cal-mag you need to look at the average as the solution runs, when its as close to 2 as you can get, abort and use the numbers printed. You may need a few tries and pick the closest one.