Today I was flying my copter (GAUI 330/APM1/v.2.5.5 from MP) and suddenly it made a very strange flip/roll. Luckily I had some height and the copter stabilised after e few seconds but this was quite scary. Before and after this moment everything was as normal.
I had just taken off and was doing a fairly high speed forward flight agains the wind. It was quite windy (10-15 m/s).
If you look in the attached log, you'll see that at around lin 15740 there is a large spike in the roll value.
Can anyone comment on what might have caused this?
I reply to my own message, I got the same thing today again. I never had this before 2.5.5, so could this be a problem with the new software?
I've started test flights todayon a R550 hex. The first couple of takeoffs were nice and the default values were stable enough to watch for a while before tuning. I then got a flip on takeoff. I went back out this evening and got another flip on the first takeoff attempt. My takeoff is a slowly increasing throttle. Takeoff occurs at about half throttle as expected. Any ideas would be appreciated. I don't like the idea of Russian Roulette each time I advance the throttle.
OK, i have some footage of a flip now. This basically happens everytime I suddenly increase the throttle to full. When I don't do this, there seems to be no problem at all. If I hold the quad in my hands and give full throttle it doesn't do this. I have attached the log of this flight.
It's a little bit hard to see but the copter is flying with the back (blue legs) toward the camera. It flips backward (yellow legs up) everytime I engage full throttle. The actual flip you are seeing on the attached video is the third peak in the log.
this is the logfile that was produced during this flip. The flip in the video is around lines 27900.
BTW, is there some way of easily splitting out the relevant parts of a logfile?
Speculation, but I suspect it might be because the ESC/motors are pegged at max and have no more left to give to stabilise if an arm dips.
I see this with my Tri. It flys with the tail slightly off vertical to compensate for Yaw. So, it hasn't got quite the max vertical thrust from the tail as from the other arms. If I give it max throttle, it starts to dip in the tail until I back it off a little.
Perhaps try recalibrating the ESCs. Otherwise, it might just be a difference in the motors/ESCs.
I was thinking about ESC problems too, but in that case i would expect the effect to be there when I hold the quad in my hands too. I recalibrated but saw no difference in the 'hand test'. However I didn't fly after recalibration yet. This takes some time because I find trying this out in the back yard too risky.
Apart from that, I flew the same quad with almost all versions from the 2.x.x series and never observed this effect before. My next step will be to dowgrade to 2.5.3 and try again. (probably this weekend).
Just a question here, but i identified an issue in most stock ESC firmwares where the digital filter on the input attempts to de-glitch the signal. This means that it does a throttle hold for short periods of time when noise or loss of input signal occurs. You can test this for yourself by spinning one up to mid throttle, then just unplug the ESC signal lead and every ESC I tested continued to run on for maybe a second or two. This would be more than enough to flip a quad in the air.
The recommendation is to move to a multirotor specfic ESC firmware such as the SimonK, which removes the input filter, removes all the calibration and battery cutoff mess (yet another possibility to cause the flip). Here is a link to a thread for the same reasons and ESCs that support the firmware.
The problem here is that we use off the shelf componets in a way they were never intended. I would say most ESC firmware stock has a number of issues when used in a multi-rotor platform. The firmware interaction only adds another layer for failure, and thus the SimonK firmware attempts to remove those features and let the stabilization system have a more direct motor control function.
I have had this happen to me too... I was not even at full throttle but at "fast climb" and I would get a hard roll over.
gonna see if I downgrade and try it see if that clears anything up.
I found that the problem was gone after downgrading to 2.5.3 and back again after i upgraded to 2.5.5, so I have posted a replay to the 2.5.5 discussion since I now feel this is really a problem of the 2.5.5. software.
Gerrit: I can't replicate this and none of the dev team has seen it either, but I suspect that it has something to do with your ESCs and their timing. If you've got the timing too tight, at high throttle they can start to stutter because the timing has advanced to the wrong pole. Normally ESCs should be fine with their default settings, however. Have you changed yours? If not, perhaps they got their timing accidentally changed in some bootup noise. Might be worth resetting them to their default settings.
As for why you see this in 2.5.5 and not in 2.5.5, it may be that the throttle curves are different in the two and the newer one just pushes the ESCs harder.
I use the standard ESC's that come with the GAUI 330x AFAIK at their default settings. However I will reset them to default and try again.
However I would expect that if that is really the problem that if I hold the quad in my hand and give full throttle that I would notice a 'back flipping' torque. However I don't feel anything like that.
As for the changes in curves, that should be easy to find in the software isn't it, do you know where in the software that would roughly be?
just got to fly with 2.5.3 and the problem is gone. full throttle no roll.
gonna test some more this weekend when I get a chance. will try upgrading to 2.5.5 and see if it comes back