Hi guys,
I rolled back to 3.1 because I was sick of my crafts (2 3dr quads) flipping in the middle of auto missions for what seem to be no reason. That was the only time I saw a problem with them.
Anyway, I was just hovering while doing a 90 degree yaw every 30 seconds or so. I seem to have a motor thats performing less than the others so I was trying to sort out which it is. After the fourth yaw, the craft just dropped to one side. Thankfully I was only about 5 ft off the ground so only a prop break.
I'm getting pretty frustrated with not being able to have ANY trust in this code! Honestly with my eyes it looked like a motor failure, but I've done a few full power run ups and everything seems to be performing just fine.
It looks like motor 1 and 2 had some weird interaction there. 2 quit so 1 tried to compensate? Or 1 started ramping up so 2 tried to compensate too, but then 2,3, and 4 all shut off at the same time?
Could anyone help me figure out what's going on here? I've done the compass_mot test everyone talks about nowadays and it was well below the suggested max.
Any help would be great. Even just some discussion on how to start building confidence in the system again after too many unexplained failures in a row would be great.
Happy flying!
Replies
You had some mechanical or ESC/prop failure.
line 15678 : ATT/Roll starts to get very different from NTUN/DRol
line 15708 roll is now 53.86degree - if throttle were working properly, it would shoot off at great speed.. so:
line 15714 logs a throttle reduction (CTUN/TrhOut)
when trottle is all the way down, you are at a uncommanded 75degree roll that flips later on.
If I only got $1 when people complain about confidence when the fault is outside ArduPilot/Copter :)
-At least you had complete logs :)
loitercrash.png
Haha, hold on hold on, give me credit where credit is due! I said it looked mechanical first, I didn't jump straight to blaming the code! And I said rebuilding trust in the system, not just the ardupilot code!
Shoot... Ok maybe I was frustrated typing. My fault. Sorry!
Alright, so you did give me some great info there, which is the difference between ATT/Roll and NTUN/DRol. I've read through the list of parameters a few times but somehow have missed the desired roll rate out from the autopilot. Thanks for pointing that out!
I could definitely stand to learn more about motors though. As far as I can tell, my ESC connectors are all solid, my connection to the PDB is solid, and all motors at flight speeds. Is there some sort of failure mode for motors that have a good number of flight hours on them (50 or so I'd guess? I have all flights logged though so I could total it up for a more accurate estimate), or maybe motors that got too hot once? How often should I be replacing motors? Seems like it would vary by quality, but are there datasheets available with suggestions somewhere?
Thanks for the help and sorry again I jumped to blaming code. I swear I don't normally do that. I have been having a few crashes lately (on my other craft) that I couldn't find the cause of where it was in auto, doing fine, then just falls. It seemed like it wasn't mechanical because they can still fly just fine and everything is connected up. I'll go back and look again though with the new ATT vs NTUN roll info.
Thanks again!
I commented for fun, you were not "blaming" too much :)
Motors *should* hold "forever", except for bad ball bearings there are no wear inside.
That said: Some sloppy manufacturers cut/flatten windings whilse assembling the base of a outrunner, this can give defective windings, usually detected by lower efficiency / heat.
With correct propellers, motors should not be hot after flight.
A motor that have been too hot, may have internal short circuit between windings, resulting in wrong impedance, and bad efficiency.
ESC's can have bad (cold) solderings, I've seen myself ESC's where Source pin became loose. this is a nasty fault that may come and go as it can get fused together again (poorly) by the current.
I've also seen sloppy soldering between windings and main motor leads.