Hi,

I just took my apparently new Mavic Air (replaced under warranty) out of the box. Did a firmware upgrade, charged all batteries full and went for a flight in a forest. I did compass calibration as prompted, didn't do any other calibrations (no prompts or warnings). Took on the prop guards and switched off forward/backward vision sensors due to the type of surrounding (a path in the forest to narrow to use them).

Then (about 3 mins from start) the bird started going down. I was able to counter that falling with the left stick, however it kept being stubborn and  it actually touched the ground a few times (lightly, and the ground was soft, covered with  leaves). I didn't let it land there as I didn't really see what's the surface below the leaves and was afraid to damage the props (so I don't know what would have happened - wheather it would finally realize that it had landed and stop the motors or not).


Some time later (after it got stabilized in about 40 secs) I noticed that the displayed height was incorrect (way too high). Then it was stable again and I continued the flight without problems - which I now think was a bad idea, anyway.


I looked at the logs, they show that IMU altitude was rising fast (incorrect), while VPS altitude shows generally correct measurements (lile IMU: 29.9ft, VPS: 0.3ft at 3m 8.1sec). That's a WHY #1.


The WHY #2 is: why the MA didn't listen to its downward vision sensor and actually let it hit the ground (it could actually hit it harded if I didn't counter it with the rudder or if the ground was not that soft). It's true it was just before sunset, but it didn't complain about insufficient light for the vision sensors to work even during the next flight later.


https://www.phantomhelp.com/LogViewer/DW9WAPHI872M9F19U4Q0

https://streamable.com/kdejk

Thanks for any advice!

Hi,

Had a look at your log and video. If i look at the log the rc inputs are inline with the baro height indication of the MA.

Baro height is relative to the takeoff position height, so if you climb a few feet in the beginning and your drone remain at the same relative height above the ground the VPS indication is correct.

See the chart for RC input and baro and vps height.

cheers
JJB

Hi, thank you for reaching out and we do apologize for the inconvenience that this has caused. Can you please try to recalibrate your IMU and try to fly in a different location wherein there is no possible interference with a wide-open area to see if the problem persists? Also, please make sure that you have updated the firmware to the latest version to avoid any possible problems. Please keep us posted. Thank you for your understanding and support!

Hi,

JJB, thank you for looking at the logs. Yes, it looks like rc inputs are roughly in line with the baro height. But this is due to the fact that I was trying to counter the falling by steering up, while in the same time the baro height measurement was rising with no reason. There was no climbing involved, the area is flat.
The real problem is how the VPS and IMU (baro?) height clearly started to drift apart, till about 20 metres' difference. That's weird. At first I thought it might be the moving leaves beneath the MA confusing the downward vision sensor, but here we see that it's the vision sensor that was correct...

DJI Gamora, thanks for your suggestions. For sure, I'll recalibrate the IMU before the next flight and watch it carefully. However, that strange phenomenon happened to the MA only once, lasted for about 40 seconds, after that it was fully stable and predictable. So the problem already is not persisting. I would only like to understand, why such thing can happen - are there any other possible reasons than IMU calibration? All firmware was updated.

Thanks!

Hi,

regardless of the "thanks" in my above email, I still don't have aswers to my 2 WHYs.

In particular, the WHY #2: why the MA didn't listen to its downward vision sensor and actually let it hit the ground?

Any ideas, friends?

Avoid Gravity Posted at 11-18 16:35
Hi,

regardless of the "thanks" in my above email, I still don't have aswers to my 2 WHYs.

Just a guess, but the drone may prioritize the IMU information over the vision position sensor information.  The VPS can be affected by low light and dirty lenses so it would make sense to listen to the IMU first.