Ok, so I've not been that impressed with the whole compassmot thing, given it ruined a perfectly good quad. I did the whole compassmot thing, and eventually got 148%. It flew and loiters ok (eventually) but one pack to the next the toilet bowl kept coming back, and I had yaw issues and loiter took ages to stabilise.
So, I admitted defeat, and vowed to resort back to 2.9.1.
Or I did, if I wasn't so stubborn. Never one to admit defeat gracefully, I decided to give the compassmot one last chance, after a few mods.
Here are my mods to my quad (QUAD heavy from www.quadframe.com)...
My setup:
1. Mumetal on the underside of the APM board, in a single 8x8cm sheet. This overhangs the board slightly, but that's intentional - I wanted some screw tops covered by it.
2. Ferrite rings on the 4 servo leads from the ESC's, wrapped twice each.
3. Errr....that's it. And I refuse to do more! I want to fly, not be constantly tweaking!
Result? With no other mods, Compassmot has dropped from 148% to 43%, and I did the compassmot to 100%, not the 50% suggested in the instructions. Once I've got the kids down later tonight, I'll re-run live compass calibration then have a test flight. Hopefully it will be the last damned time I have to do this, and I can actually get back to flying...
If this still fails...well, I'm going back to 2.9.1
Comments
As an update to this - I bit the bullet and bought a DJI GPS stand, raises the external GPS/compass (and some rewiring), and my compassmot is now 4%.
I wiped them in my haste to "upgrade" to 2.9.1. Facepalms all round.
It is just my curiosity on magnetic interference.
Actually, compass interference is not my problem with AC3.0, no TB effect now.
I thought that magnetic interference is the reason to uncommanded yaw issue.
But it might not be the reason.
It's possible Jiro. But I can't be sure (wiping the logs when "upgrading" down to 2.9.1 doesn't help).
However, I need to remain open to suggestion, that it might not be mag related. Afterall, I had a bizzare situation where the quad refused to arm (unless it was via MP) once. I had just completed a loiter / flight test (that "failed", naturally), landed ok on Stabilise without incident. But when I went to take off again...nothing. When I armed manually via MP, it flew as if nothing had happened (loiter and auto flights excluded of course). A full power on/off fixed it later, but again, this sort of thing didn't happen in 2.9.1 on an unchanged airframe.
Just annother 3.0.x "feature" - failsafes that kick in for no reason, and without any notice or alert which one has just been activated. Would be good to which failsafe was running via mavlink maybe?
So my problems with 3.0.x aren't just due to compassmot (though this is the main problem). So yes, I'll be waiting until we're a good few releases into 3.x.x until I try again. But I may try the new 2.6 external compass from 3DR, which is supposedly due out tomorrow.
Euan, sorry about your crash.
You said that switching from alt hold to stabilize involved fall out the sky with no power.
What do you think about magnetic field difference?
Anyway, recent stability patch for AC3 is not good for stability other than loiter.
So, I agree with your thought that wait for the next release.
Just looking at your video compassmot method. How high did you push the throttle during compassmot, because it seems you did not even reach hover point. To correctly apply compassmot you should,push to at least 70% max theottle by holding down the quad with your hand (careful not to cut yourself) or weighing it down.
Just a thought.
Ok, I'm back to 2.9.1. I hate admitting defeat, but there you go.
I've given 3.0.x a fair crack of the whip, but I'm simply not prepared to sacrifice so much time and money getting compassmot to work. 2.9.1 gave me perfectly good flight, it was reliable, and it was safe - the failsafes worked. Loiter was even better as well.
The job to lift the APM up by 1-3 inches is a job I am simply not prepared to undertake. Both out of principle, and out of practicality. The process of compassmot is also too complex, it's too dangerous, the values fluctuate wildly, it takes too much time for some who is time pressed, and you have no real way of know if 3%, 43% or 143% are any good, or even how to compare the result in an imperical manner.
It involves so much learning, and understanding of mag fields and electronics - none of which should be relevant. I should be learning PID's, flight dynamics, route planning, using the APM auto features to give better video framing. It should fly out the box using all it's features and failsafes, so I can focus on actually using it and customising it from there. Right now, I have a standard off-the-shelf frame and motors that I can only fly on sticks, or alt hold. If I wanted that I'd have got the basic Naza.
I know theres a ton of work done on this by Randy et al, and I've love to help, but I can't help thinking I'd be going in a completely different direction with this. By focuing on usability, and simplifying the configuration process for newbs. Yeah, old hands might say "but PID's and stuff are easy"....but they're not. But compared to compassmot, they're easy. I can probably guaranteed any newb buying today are thinking "wtf is compassmot???". I really do think we're heading into dangerous waters with the current compassmot process; people are going to crash quads because of it. I did - and I know all about compassmot. Worse - someone or there child is going to get injured doing it.
What's scary is the variability of it - I got 40% variability in results, even within the same 10 minute session - and we use it in RTL, one of the main safe modes. That does worry me. Enough for me to go back to 2.9.1? Definitely.
The light at the end of the tunnel is 2.6 and the external compass. But I'm still not touching it until compassmot works consistently (maybe via a preset throttle program in MP?), and gives repeatable, reliable results (I should be able to run back to back tests, and only see 2-3% variance). And when it works with common airframes without modification.
Sorry about the rant guys - I've been trying to fly for 2 weeks now, and get some good video shots with the gimbal (now damaged btw) in the beautiful swiss summer. But I can't. I don't want to be dismantling my airframe and runnning calibration after calibration down in the basement. I've gone back to 2.9.1, I did *all* the calibrations in 15 minutes, and I *know* I can fly it tomorrow for pack after pack.
Sorry.
Ok, my test flight took place. How did it go?
Try and guess:
I'm royally pissed off with 3.0.1 right now. Let me calm down, get my logs downloaded, and try to do some analysis.
But lets just say that switching from alt hold to stabilize involved a plummet out the sky with no power. I do this activity 5-6x a pack. For the last 20 packs on 2.9.1, this never had a problem. My fourth flight on 3.0.1 (I'm not even "flying" yet, after installing 3.0.1 2 weeks ago, I'm still "testing"), and...
-
1
-
2
of 2 Next