Need help desperately with pinpointing the reason (/reasons?) for crashing my Pixhawk Quadcopter hard from high altitude. (Log file included)

Link to the log file

I have tried asking around on Reddit, but could not get any help there at all!

Hi everyone, I am flying a large Quadcopter with PixHawk (it is a Chinese clone) which I slowly built myself over the period of the past few months. I have been experiencing some crashes that appear suddenly out of the blue. My Quadcopter suddenly starts swerving to one direction, and remote inputs have very limited effect, though they seem to be able to change the direction very slightly. It generally ends with my quad crashing into the ground, and changing flight modes has no effects. Using a FlySky i6 transmitter with modded firmware, so if that's an issue, I can consider moving over to a good TX

Yesterday's crash in detail:

So for the past few days, after I had flashed the firmware again and reset everything, I was flying crash free. That gave me enough confidence to get adventurous and I moved to high altitudes, high enough to stay clear of nearby buildings (my home is in a developing area, and not a lot of people live here, so we don't have a lot of high buildings). I felt it must be around hundred feet high but Pixhawk logs say the altitude was something like 10-20 metres, less than 50 feet. I suck at estimating distances.



Altitude vs Time Graph

I tried some fast turns and circles in Alt Hold, and then changed to circle flight mode. I exited Circle in "Land" mode, lost some altitude while moving ahead and then proceeded to move straight with AltHold. Suddenly, the quad started moving away from me, and I thought it must be the wind. But then it didn't stop moving when I tried with the transmitter, and tilted to one side. It went really fast, and before I knew it, it was racing away. I tried to switch to RTL, but that didn't work. I tried Land, but it also didn't seem to work (Logs show that the mode was changed but it didn't Land).

It raced to something like 300metres away. Disarming didn't cross my mind, since I didn't even have time to consider it, and it might have caused it to fall over some person. I was lucky since some guys who were crossing had seen it go vroom.. It had crashed in a field. I have broken just one prop, bent one, and pulled a GPS connecter wire, which looks like a miracle to me. Guess these are the benefits of a Carbon fibre and Nylon Polyamide frame, lol. The LiPo was pulled off and an ESC had detached connectors from the PDB. Actually I feel very grateful there wasn't any injury to someone or some serious damage to my quad (I feel this is why you should avoid flying near people).

I am at my wits ends trying to figure out what causes these issues to pop up randomly. I don't have logs corresponding to the old crashes, since I had cleared the SD card some days back.





Desired Roll/Pitch/Yaw seem to be pretty close to the actual Roll/Pitch/Yaw measured by the sensors!

There's a GPS glitch error, but EKF failsafe is set to Land too, so I don't think that's the issue.

I admittedly have high vibrations, but I have had safe flight with even higher vibrations, so I think that isn't the issue either.



Any help would be greatly appreciated. I am desperate for help.

Using Ardupilot, Copter4.0.5. Now updated to 4.0.6

Generic_Redditor said:

Yesterday's crash in detail:

Click to expand...

You lost me about here...

Not sure what the issue is but would do some process of elimination, would take off the GPS control and just attempt angle mode flight LOS since it is the simplest option with the least things that can fail and lead to it going wrong (just requires accelerometer and gyro to be working and ESCs/motors/receiver). If it still occurs without GPS hooked up then you know it isn't GPS or compass to blame or other autonomous flight software it is something physical. Best shot in the dark is checking the "soft mounting" of the FC and if it isn't already soft mounted do that and tighten up any other frame screws to reduce the vibrations all that noise needs to be ignored by the quad which means it might be ignoring real high level motion of the quad as well.