Lost control, then 2 seconds later lost connection, on my Mavic Pro what happened ??

[mikeyway]

Hi guys, new member but long time pilot here.

So I must sadly report my Drone is gone (for now). Due to legal reasons it has a sign with my name and address, but since it droppen into a valley about 400m from my starting position there was no way to get there.

I watched a lot of "fly away" videos and read some articles and posts, but none really describe what happened to me. I'm very careful when using expensive equipment and before flying I go through my checklist and when I'm at a new place I calibrate the compass. I even go as far as to take off my watch to not disturb the process (probably overkill but gives me peace of mind).

So I took off, flew about 120m high and 400m out to get a good view. Suddenly I saw that the camera was going crazy. I assumed that it was the gimbal resetting. But then the aircraft lost height and a second later the feed cut off. I waited for 30m, maybe the RTH kicks in and it was just a serviere bug where multiple systems failed at once. But nothing happened. I sadly had my back to the drone to do a flyover and when I turned around the drone was already gone.
The weather was amazing, about 20 celsius so icing could not have been the issue. I was about 1800m above sea level, which far below the drones max altitude.
The only idea I have is a battery fault, but would the motors cut out before the video feed ?
Since at 1:10:90 the cell deviation is at 0.001 V, at 11:00:00 it's at 0.056V with Cell 2 loosing 0.075 V.

I'm running Go 4 at v.4.3.12 and the remote at 01.04.0500 (recent), I'm very sure that the aircraft was on the newest version as well, since I always update them at the same time.

Attached you'll find the video clip, the .txt and .dat flight records.
The online viewer is here: DJI Flight Log Viewer - PhantomHelp.com

What could have been the problem ?

 

[sar104]

What happened is that the FC killed the motors because it thought that it was on the ground.

​

95.291 : 8465 [L-FMU/FSM]state changed. cur: on ground​

95.320 : 8467 [L-FMU/MOTOR] Stop. reason:landing.RC_Thr​

95.320 : 8467 [L-FMU/MOTOR]Total start times: 1, time: 95.32​

95.320 : 8467 [L-PWM]set actuator mode:IDLE​

95.320 : 8467 [L-FLYMODE]CTRL reset all by motor stopped​

95.321 : 8467 [L-FMU/FSM]state changed. cur: motor off​

95.337 : 8468 [L-IMU]set imu switch level = 1​

95.398 : 8471 [L-TAKEOFF]alti: 1801.617188 tors: 91.809647​

That led to the inverted fall that you saw in the video:



As for why the FC thought it had landed, the only thing I see is that when you applied full down throttle at 94.4 seconds there was no descent recorded for over 1 second. I thought that it required 3 seconds of no descent to stop the motors, but perhaps not on this firmware.

 

[mikeyway]

What happened is that the FC killed the motors because it thought that it was on the ground.

​

95.291 : 8465 [L-FMU/FSM]state changed. cur: on ground​

95.320 : 8467 [L-FMU/MOTOR] Stop. reason:landing.RC_Thr​

95.320 : 8467 [L-FMU/MOTOR]Total start times: 1, time: 95.32​

95.320 : 8467 [L-PWM]set actuator mode:IDLE​

95.320 : 8467 [L-FLYMODE]CTRL reset all by motor stopped​

95.321 : 8467 [L-FMU/FSM]state changed. cur: motor off​

95.337 : 8468 [L-IMU]set imu switch level = 1​

95.398 : 8471 [L-TAKEOFF]alti: 1801.617188 tors: 91.809647​

View attachment 70077

That led to the inverted fall that you saw in the video:

View attachment 70078

As for why the FC thought it had landed, the only thing I see is that when you applied full down throttle at 94.4 seconds there was no descent recorded for over 1 second. I thought that it required 3 seconds of no descent to stop the motors, but perhaps not on this firmware.

View attachment 70079

Thank you for the fast reply.
I still don’t get why the connection was lost. And wouldn’t the motors turn back on if the IMU recognizes fast downward motion ?

 

[sar104]

Thank you for the fast reply.
I still don’t get why the connection was lost. And wouldn’t the motors turn back on if the IMU recognizes fast downward motion ?

The connection was lost when it dropped out of sight below the ridge line:




The motors will not restart after shutdown without a CSC, any more than they would start to begin with if you threw the aircraft off a cliff.

 

[sar104]

By the way - this was clearly an aircraft control failure - if the aircraft were under warranty, rather than two years old, then this would be a warranty case.

 

[mikeyway]

Was there and way this could have been prevented ? Or did the FC just interpreted the data incorrectly ?
I was always under the impression that for the FC to shut down the motors for a landing it would at least need one of the bottom facing sensors (sonar or camera) to provide confirmation.

 

[Keule]

the only thing I see is that when you applied full down throttle at 94.4 seconds there was no descent recorded for over 1 second.

This area is well known by Paragliders and the good thermal conditions, especially this time of the year.
Could it be the case that for the exact 1 second upwind of 3 m/sec and downthrottle with -3 m/sec the FC sensed a false landing condition?

 

[sar104]

This area is well known by Paragliders and the good thermal conditions, especially this time of the year.
Could it be the case that for the exact 1 second upwind of 3 m/sec and downthrottle with -3 m/sec the FC sensed a false landing condition?

That's exactly what I was wondering too - it's perfect terrain for those conditions. I actually know that valley pretty well and I would not be at all surprised if there had been significant thermals on that side. The wind was out of the east and appeared to be rather variable in speed, which is what you would expect in that situation.

 

[sar104]

Was there and way this could have been prevented ? Or did the FC just interpreted the data incorrectly ?
I was always under the impression that for the FC to shut down the motors for a landing it would at least need one of the bottom facing sensors (sonar or camera) to provide confirmation.

As far as I know on the Mavic Pro it's just lack of descent as indicated by the barometric sensor. The VPS system did not see the ground.

 

[mikeyway]

This area is well known by Paragliders and the good thermal conditions, especially this time of the year.
Could it be the case that for the exact 1 second upwind of 3 m/sec and downthrottle with -3 m/sec the FC sensed a false landing condition?

It’s true I know some people who go paragliding there as well. That truly could have been the cause, but that makes me loose a bit of hope in DJI‘s Design philosophy.
Has someone tried to hold the drone just by the arms without covering the sensors and throttle down ?

 

[sar104]

It’s true I know some people who go paragliding there as well. That truly could have been the cause, but that makes me loose a bit of hope in DJI‘s Design philosophy.
Has someone tried to hold the drone just by the arms without covering the sensors and throttle down ?

That's a hand catch and yes - the motors will stop. I thought it took more than 1 second though.

 

[mikeyway]

That's a hand catch and yes - the motors will stop. I thought it took more than 1 second though.

Usually with hand catches you cover the bottom sensors though (at least I do).
When I land it’s usually more then 1 sec before it powers down though

 

[sar104]

If you look at the motor speeds and tilt there is clear evidence of a strong updraft towards the end of the flight as the tilt remains basically constant but the motors slow down.

 

[mikeyway]

Let’s hope future models rely more on the VPS or at least requires a “forced landing” when no ground is detected. I’ll send the results to DJI when they (hopefully) answer my job application.

 

[Keule]

when they (hopefully) answer my job application

Nice - i hope it'll work out!
Are you with DJI in contact or a dealer nearby?

Good luck!

 

[mikeyway]

It’s a marketing position at DJI. Showing new products to the press, explaining the highlights etc and also giving feedback to the R&D department. One would be to not ******* turn off the props 120m above the homepoint without consulting any of the onboard sensors. And after you turned them off if the aircraft starts falling down and they’re still spinning from the wind turn them back on. Since it has rpm sensors it should be able to detect that.

 

[sar104]

@BudWalker - any thoughts on this event?

 

[sar104]

Here's a more accurate depiction of the flight, using just GPS location and height.

 

[mikeyway]

But you’re assuming it hit the ground exactly below where the last contact was. I‘m not well versed in drone physics but couldn’t there be a lot of deviation, especially with the wind ?

 

[sar104]

But you’re assuming it hit the ground exactly below where the last contact was. I‘m not well versed in drone physics but couldn’t there be a lot of deviation, especially with the wind ?

No - the flight track ends in mid-air. It will have drifted. That's fairly easy to estimate too, if you have plans to try to find it.