Mavic 2 Pro self ascend

[Proton2]

I'am a bit puzzled. This is a third time that it happened with my Mavic 2 Pro. Without any stick input it started to gain altitude very fast. As i can recall I was flying backwards everytime that it happend.
I recorded screen from my flight logs (see attachment).
Logs from this flight you can find here It's at 9:28 flight time.
I start to wonder if sending it to DJI for replace/repair isn't the best option ?. Maybe someone can help mi to clarify this.

 

[slup]

Yes, your observation was correct ... indeed you have an uncommanded ascent starting from 568sec & that takes the AC from 67m height to 117m as highest, and this even though you shortly apply negative throttle to descend. The incident stops then at around 586sec when the AC finally listens to your full throttle input & descends. Looking at the green graph representing the Z-speed (speed down positive, & up negative) it have a oscillating look which could indicate some kind of hardware oddity.

Think the mobile device .DAT log ending with FLY026.DAT is needed for this ... Read up here on how to retrieve (see section 3.) --> Mavic Flight Log Retrieval and Analysis Guide attach it then in a new post in this thread.

 

[Proton2]

Thank You
I attached required dat file to this post.
I also wonder - on all of my batteries 4th cell is always weeker than others. Those are all from fly more combo kit and have the same production time. Is it a common phenomenon ?

 

[slup]

This looks really odd ...

As already seen in the .TXT log the ascend is uncommanded ... without corresponding stick inputs the Z-speed first goes slightly positive (descend) starting about at 568sec but then when negative throttle input is given to stop the ascend the Z-speed goes negative (ascend) ... during the whole time weather it's z-speed neg. or pos. the height increases.



Checking the motor rpm's show a sudden increase for all 4 motors ...



And the current draw increases in a similar way ...



And all this seems commanded by the flight controller as a rise in motor command % also is seen during the same period ...



Unfortunately this is above my knowledge, can't find a reason for this ... it's not at all as the AC should behave.

@sar104 hopefully see this & jump in with a deeper explanation.

 

[boblui]

Is it possible that the 3D camera at the bottom made the FC think that there was some obstacles approaching from beneath so it commanded the craft to ascend to avoid it ? Not sure if the flight logs has related info

 

[sar104]

This looks really odd ...

As already seen in the .TXT log the ascend is uncommanded ... without corresponding stick inputs the Z-speed first goes slightly positive (descend) starting about at 568sec but then when negative throttle input is given to stop the ascend the Z-speed goes negative (ascend) ... during the whole time weather it's z-speed neg. or pos. the height increases.

View attachment 105679

Checking the motor rpm's show a sudden increase for all 4 motors ...

View attachment 105680

And the current draw increases in a similar way ...

View attachment 105681

And all this seems commanded by the flight controller as a rise in motor command % also is seen during the same period ...

View attachment 105682

Unfortunately this is above my knowledge, can't find a reason for this ... it's not at all as the AC should behave.

@sar104 hopefully see this & jump in with a deeper explanation.

The problem is the accelerometer data, particularly on the z axis. However, both IMU1 and IMU0 show the same spurious data, which makes me think it's another of those strange Mavic 2 FC processor lapses.



The pressure altitude are data (black solid line) are noisy but agree in delta with the GPS height data (blue solid line), and when differentiated with respect to time (black dashed line) agree with the GPS vertical velocity (blue dashed line). Those all clearly show the aircraft climbing. However, transforming the IMU accelerometer data into the earth frame of reference, and then integrating the z-axis data with respect to time to get vertical velocity (red and green solid lines), indicates why the FC thought that the aircraft was descending. When combined with the other vertical data via the sensor fusion algorithm, the FC ended up with very incorrect vertical velocity (red and green dashed lines). It thought that the aircraft was descending, which is why it commanded the motors to speed up.

 

[Proton2]

Thank You slup and sar104 for the analysis. I'm sendig it to DJI for repair/replace. I don't want it to happen when i will be flying in a controlled airspece.

 

[sar104]

Thank You slup and sar104 for the analysis. I'm sendig it to DJI for repair/replace. I don't want it to happen when i will be flying in a controlled airspece.

As I mentioned, this may well not be a hardware issue. The M2 has a documented, though rare, history of randomly exhibiting IMU/FC computational problems. It will be interesting to see what you get from DJI.

 

[Proton2]

I've sent my mavic to DJI, and they said that Downward Infrared Sensing System is falty and Gimbal Axis Arm Module (RH). Repair of course free of charge.
I wonder how this can be related to this problem ?

 

[Not A Speck Of Cereal]

I've sent my mavic to DJI, and they said that Downward Infrared Sensing System is falty and Gimbal Axis Arm Module (RH). Repair of course free of charge.
I wonder how this can be related to this problem ?

I can imagine how the Downward Infrared Sensing System could cause, if it sends a lot of messages about being too close to the ground when it's not.

 

[slup]

I've sent my mavic to DJI, and they said that Downward Infrared Sensing System is falty and Gimbal Axis Arm Module (RH). Repair of course free of charge.
I wonder how this can be related to this problem ?

If that sensor doesn't function as it should & influence the FC in some way it doesn't sound impossible ... but

I can imagine how the Downward Infrared Sensing System could cause, if it sends a lot of messages about being too close to the ground when it's not.

... it's only strange that the TXT or the DAT doesn't show any thing being registered from the bottom sensors when the uncommanded ascent happens.

 

[sar104]

I've sent my mavic to DJI, and they said that Downward Infrared Sensing System is falty and Gimbal Axis Arm Module (RH). Repair of course free of charge.
I wonder how this can be related to this problem ?

Neither of those things were the cause of your problem, as explained previously. Whether or not they were faulty is a different issue entirely. Either they didn't look at the flight DAT file or they didn't understand what they saw. As I mentioned, it looks to me like a processor error, which has been seen a number of times with the M2. It's very uncommon and intermittent and so they most likely would not see it in testing.

 

[boblui]

I've sent my mavic to DJI, and they said that Downward Infrared Sensing System is falty and Gimbal Axis Arm Module (RH). Repair of course free of charge.
I wonder how this can be related to this problem ?

Obivously DJI is just making it up. When the self asend occurred, the IR sensor was not giving any readings as the ground was wayyyyy of out it's range.

 

[Andris]

Any further update on this issue? I am experiencing the same