Problem with step 3

Static, Dynamic, Log File Analysis, Troubleshooting

Re: Problem with step 3

Postby Andrej » Mon Apr 01, 2013 1:38 pm

I had same problems before, after I made all exactly like described in wiki. After each step do the copy/paste all #defines *** to "all parametrs section" and save parametr file. From step 3 every step use the previouse step parametrs from file. In step 3 it is normal that mae swing ... in time it come lower and lower. You have to wait for finish step 3, for my files it took ~15min every next step you can stop at any time, copy-paste-save and make next.

I made express dynamic calibration ~2min in the room and relatively short calculations. Now my quadra flying, it is lazy and has huge inertion but seems it have to be PID tuned.
Posts: 79
Joined: Fri Mar 22, 2013 5:10 pm

Re: Problem with step 3

Postby leopold1 » Mon Apr 01, 2013 3:51 pm

Did some overtime last night just to try understand the calculation procedure.
It seems that when rerunning step3 it takes longer and longer to get to a lower MAE.
And the longer it takes it seems that it goes down, and then increases again to finally, after a couple of hours, decrease untl the MAE is allost the lowest value again.
So I'm pleased to say I've reached my lowest point in step 3 and continueing with step4. :)

I also didnt quite know that you could/should add multiple dynamic logs. I thought one was enough in conjunction with the three statics.

ps, need to get a faster computer ...
Posts: 311
Joined: Tue Jul 31, 2012 11:43 pm
Location: Belgium

Re: Problem with step 3

Postby chschmid » Mon Apr 01, 2013 4:22 pm

Hi Vincent

Good to hear that you succeeded. My I7 was calculating for many days to get nice results. So a fast computer helps.

Posts: 1800
Joined: Wed Jul 11, 2012 7:41 pm
Location: Herrliberg, Switzerland

Re: Problem with step 3

Postby afernan » Fri Jul 05, 2013 11:08 am

Similar problems here.

I did already a lot of succesfull calibrations, but I´m doing now a new HEXA and I can´t converg on Step3, I think i did all what you can imagine

- checked that all files used have no checksum errors (by exporting with logdump or by running cal.exe on command-line)
- 3 static files have been used several times succesfully in previous copters. So this indicates that the new dyn.log should be the problem (I guess)
- repeated 2 different dynamic logs more
- Used 1stat (did will all three stats)+ 1dyn file. Also with 2 stat + 1 dyn.
- Trimmed noise at start/end of files.
- leaved all night long running > at the end oscillate, no convergence

The behaviour of MAE is always similar:
- if MAE always decrease alogn the calculation, then it will converg
- if MAE first increase abit, then decrease a bit, and then increase slowly oscilating at the end: no convergence

Many times it arrives to an acceptable value (like MAE = 0.04). At this point we should be able to stop manually and get the params. This could be a simple solution for this non-convergence problem that many of us have experience.

Any idea is wellcome (I know... meanwhile I´ll keep dancing Calipsos to get more dyn logs)

UPDATE: now´s working.
This is what I´ve did (same static than before):
- changed the batery from 4S 5000mA to 3S 2200mA
- changed the rotation sequnce. Before I did exactly what´s in wiki page. Now I´v changed to the following sequence, starting with:
- copter top side facing up: I did the four lateral sides
- 1 x copter top down
- 1 x front side pointing up

Got a convergence in 10 min about (mae = 0.07)

- "abort" button doest´t work in step3, (QGC 1.3beta2)
- can´t make work two or more dyn logs at the same tine. Is it possible?. Is it useful?
Posts: 1214
Joined: Sat Jul 07, 2012 5:24 pm
Location: Madrid, Spain


Return to Calibration

Who is online

Users browsing this forum: No registered users and 2 guests