Pixhawk GyrX inconsistency => crash

Hi all,

First of all, I’m a newbie to multicopter but I think I have a not so bad understanding of it :wink: I have a x550 frame from Hobbyking (quadcopter) and Pixhawk (chinese copy). I already did several flights with it and it flights pretty well. But today, I had a really bad crash (my first).

After takeoff, I switched to AltHold, moved a bit forward/backward and left/right but suddenly, my quadcopter started to roll to the right (more and more). Full left on my transmitter did not help. Switching back to stabilize didn’t allow my to save it and it crashed at ~60 km/h on a tennis court. Fortunately, it was late and nobody was playing but this could have turned quite bad :cry:

Looking at the logs, I see that just before the copter starts to drift in an uncontrollable way, GyrX of IMU and IMU2 show inconsistent behaviour (around log line 7600): one increase, the other decrease!? Does anyone have an idea how and/or why this could happen?

Also at around 8600, AccZ starts to go crazy. I have seen this already once in one of my first flight in Auto Mode. At that time, the consequence was that the copter started to put full throttle and climbed pretty fast. Switching to stabilize, I was able to take control again. This could be because of high vibration on my quad at full throttle since plotting CTUN.ThrOut and IMU.AccZ show some correlation.

What worry me most is the reason why my quadcopter suddenly started to roll to the right in an uncontrollable way. Since I’m new to all of this, I don’t have enough experience to find a good explanation. Can anyone give me some pointers?

Here is my dataflash log: drive.google.com/open?id=0Bw0Ex … authuser=0

P.S. I remember having a similar issue during my first or second flight. The quad started to roll to the right. I was at very low altitude and was able to land without issue. I remember having unplugged/plugged the battery to reset the board and everything was fine again.

P.P.S. Today, I plugged in the battery with the quad on stable ground. Once the blue/red/white led blinking was over and the quad was acquiring GPS, I pressed the arm button of the Pixhawk and the I moved it to a different place, some meters away. Once GPS was acquired (green led), I armed and took off. Can it be that the gyroscopes were not completely calibrated when I moved the quad? I don’t think so since the first part of the flight does not show any inconsistencies.

Forgot to mention that I’m using stable 3.2.1.

Logging could be set higher so we can see the RCIN and RCOUT values.

Looks like you have bad vibrations in this build and you need to mount the Flight Controller on some foam to cut down on the vibrations.

As soon as the unit started becoming unstable you should have switched to Stabilize mode to try and recover the unit.


Thank you Mike for your answers.

Yes, I will try to increase logging next time. I was a bit concerned about the CPU cycles this would eat since I already see some slow loops in the logs. On a Pixhawk, can I set logging to all without worrying about slow loops too much? But first, I have to check in more details what got damaged and what I need to replace/re-order. First crash and a bad one :cry:

Do you say this because of IMU.AccZ at around 8600?

More damping was planned but since in PosHold or AltHold the vibration level was pretty low and in healthy ranges, I postponed it. Now I think that at full or high throttle my quad may start to vibrate badly (some resonance effect?) in the Z axis. This explains I think the high vibrations seen on this axis when the quad was at full throttle. Still the reason why it started to drift more and more to the right is still to be answered and probably not related to this (I guess).

Yes, agreed. Still learning (the hard way this time) :wink: I guess I could have crashed earlier if I had switched directly to Stabilize. I think the AHRS just went crazy for some reason. Could this be due to the inconsistent GyrX values from IMU and IMU2 around 7600? Any explanation for this?

I think when I had this similar issue during one of my first flight, I plugged in the battery with the quad on its side (because it is easier for me to connect the plugs). Then I put it on a leveled surface and started the flight. After some times it started drifting also but at very low altitude so I could safely land. For some reason, I had expected that the gyro calibration would not completed until the quad is stable for some time. At least this was the behavior I had on my small Hubsan X4 clone. Since that day, I always plug the battery in with the quad on a more or less level surface (a soccer field). As soon as the led is green, I arm and flight. Everything went fine until yesterday evening. Should I wait more before taking off maybe? I’ve read about people waiting at least 3 minutes but this seems to be more related to the acquisition of a stable GPS signal.

Also another question: does the compass(es) have anything to with roll or pitch behavior? I thought that they were only used for yaw but I’m not so sure anymore :confused:

Thank you for any help that can help me understand what happened. I’m not worried about my damaged quad but I’m worried about the same happening again.

The good thing is that I will now expect the thing to go wrong at any time and flight with enough free space (I did my maiden flight in front of my house with my daughter playing next to me :blush:).


Careful here. Large tall objects effect GPS signal. I have had several bad things happen because of GPS drift.
What happens is the signal for the GPS bounces off the roof of the house and causes the GPS position to suddenly move either towards the house or away from the house. This causes the copter to correct for that movement by flying in the opposite direction very quickly hitting the house, your car or people in the street.
This is only a concern when flying in a GPS assisted mode like Loiter, RTL, Position Hold and even Land. So watch out for failsafe as I was flying in Stabilize mode thinking everything was good when the battery failsafe kick in and it headed for my car.

If you are using a Pixhawk the logging turned up should be fine. APM boards need to be turned back down after testing. They are under powered.


Thank you Mike. I disassemble my quad yesterday evening and I think I just lose the frame, 3 propellers, 2 motors and the battery (ok the GPS mast also). It will take me some time to rebuild it but I will take your comments into account the next time I flight.


Just for anyone interested, I found this (new) post on drones-discuss today and I think it looks quite similar to the issue I had:

groups.google.com/d/msg/drones- … qIOJEfNFAJ

Hi all,

I repaired my quad, improved vibration dampening and made a first flight this evening and… crashed again after ~15 minutes. And the interesting thing is that the crash looks exactly as the last time. Fortunately, I was at low altitude and only one propeller died this time!

Few seconds before the crash, I see inconsistent GyrX values between IMU and IMU2. One shows a positive value and the other a negative one. Also as you can see, I tried to fight the drift but eventually crashed.

Today, I flighted with one of the latest daily build of 3.3. Of course, I redid compass and accel calibrations. I had one slow drift during this flight but was able to land nicely, disarm, re-arm and everything was fine again!? I also had this problem 2-3 times with 3.2.1.

Since my Pixhawk is a Chinese clone, could it be related (at least the GyrX inconsistencies) to bad hardware?

My logs are available here.

Thank for any advices. I don’t feel like I won’t flight this quad as long as I haven’t found the root cause of these crashes.


Also how can AHR2 and ATT start to diverge? I see this at several places in my logs. Here is an example with AHR2.Roll and ATT.Roll.

Just before I crashed, it seems that AHR2 decided to converge again in a quite sudden way!? However, I’m not able to find any correlation/explanation for this behaviour in my log!? Anyone has an idea? Thank you.


I manage to replace the MPU-6000 on my Pixhawk (actually my brother did it. He has all the required tooling at work). And since then… everything is fine :sunglasses: My Pixhawk seems to be rock-stable now, no more gyro inconsistencies and no more leans during flight.

So I guess that my original MPU-6000 was bad but no bad enough to completely make it unusable. I also think that EKF makes a good work and made my copter usable up to a point where the leans manifested because of bad/inconsistent sensors values. Actually, the IMU was reported as 100% healthy in the new 3.3-rcX builds.

I can flight again without worrying too much :wink: