Hello all,

I'm a new Phantom 4 owner (since December), happy and all flying here and there, if it weren't for some issues

Long story short, in mid-January I had an incident with my Phantom 4, it kind of crashed from about half a meter (2ft). I was testing the Litchi app and at the same time trying to make sense of how the distance to objects look in the video feed. Anyhow, because it was in a radius of less than 20m and the battery was under 20%, the drone decided by itself to land there (a meter off from where I left it) without me being aware that it does that. The obstacle avoidance system didn't see the small branches in the vineyard and it got stuck in them and then it fell upside down in the snow. In the process a propeller and a propeller mounting plate (the plastic one) were destroyed. I've uploaded the crash:



Anyhow, there's no official DJI Repair Service Center in my region (somewhere in Eastern Europe) so I took it to an unoficial repair center (a guy in a basement knowing a bit about how to fix stuff). Since it was fixed, I've noticed that sometimes it doesn't want to come down or if it does, it descends very slowly; even worse, sometimes it gains in altitude without any of my input. I've analyzed the logs and recording and I've discovered that the VPS is acting odd, randomly showing a VPS calculated altitude of 0.0-0.2m although obviously I'm flying freely at a height of at least 50m without any danger in sight.

An example here of the drone acting by itself, gaining in altitude without any input based on false vps data:

It happens in almost every flight (since the crash), minus the first two after it was fixed. So, if anyone is still with me, what do you guys think:
- If I deactivate the downward VPS is there anything I should take into account, any disadvantage? I would guess it would be safer for sure than the Phantom flying by itself because it wants to avoid air.
- Is this something that can be fixed?
- Any idea what could  I try/test to better understand why is the VPS fooled?

Thanks in advance!

The VPS may have been damaged. Connect it to the Assistant2 and restore to factory defaults to see if it helps.
Also a IMU calibration may help.And a sensor calibration.

I agree with Ken on the IMU calibration.  Also calibrate gimbal and compass.  And you can expect vps not to be accurate at high altitude.

DJI-Ken Posted at 2017-3-29 12:25
The VPS may have been damaged. Connect it to the Assistant2 and restore to factory defaults to see if it helps.
Also a IMU calibration may help.And a sensor calibration.

Thanks for the advice! I'm going to try in the weekend a VPS calibration. The IMU should be fine, it was calibrated after the crash. I don't hold much hope that the VPS calibration would improve the situation, most probably it is a hardware problem following the crash. I will fly without VPS (downward) from now on, to avoid any further issue.

indicibil Posted at 2017-3-29 13:14
Thanks for the advice! I'm going to try in the weekend a VPS calibration. The IMU should be fine, it was calibrated after the crash. I don't hold much hope that the VPS calibration would improve the situation, most probably it is a hardware problem following the crash. I will fly without VPS (downward) from now on, to avoid any further issue.

Sounds good, let us know how it turns out.

Kneepuck Posted at 2017-3-29 13:11
I agree with Ken on the IMU calibration.  Also calibrate gimbal and compass.  And you can expect vps not to be accurate at high altitude.

Thanks!

I don't expect the vps to be accurate at high altitude, I actually expect not to show anything because it is out of range. In this case it sees something very close below (0.1m - 0.7m). Sometimes it shows 0m, which scares me because it might be a risk for the motors to stop if I try to lower the drone.

The fact that it gains altitude as it tries to avoid a crash based on false vps data is also dangerous, imo. If I'm in a forest between the trees and it starts to ascend quickly I'll have little control over the situation and it will crash again.

I would be worried about descending with the throttle full down with it acting like that - it might decide that it has landed and turn the motors off?   There have been a few crashes with it landing mid-air but I don't know if this is the same situation.

Presumably this is a problem with the ultrasonic sensors, not the vision sensors, so calibration of the vision system in DJI Assistant wont help?

Does it ever give good values, eg when within 15m of the ground, or is it always showing close to zero?   Is there anything obstructing the ultrasonic sensors?

You could upload a log file of a short flight to: http://www.phantomhelp.com/LogViewer/Upload/ so that we can see more detail...

Nigel_ Posted at 2017-3-29 13:27
I would be worried about descending with the throttle full down with it acting like that - it might decide that it has landed and turn the motors off?   There have been a few crashes with it landing mid-air but I don't know if this is the same situation.

Presumably this is a problem with the ultrasonic sensors, not the vision sensors, so calibration of the vision system in DJI Assistant wont help?

Yes, that's what I'm worried about too, that the motors will stop in flight. The vps still works sometimes, especially at low altitudes but the random errors appear everywhere.

I've uploaded the flight log for the flight where it exhibited a lot of vps errors (the second video in the first post):

http://www.phantomhelp.com/LogViewer/O3GUKWOFJFAKJM964BXY/

If I compare the IMU Altitude with the VPS Altitude in a graph, it becomes quite scary:

Regarding the obstruction of the ultrasonic sensor, I've checked again and again, it looks clean, nothing in the way, I carefully cleaned some more the fine grid covering the sensors. It should be something visible, large and solid in order to bounce back the sound wave so quickly.

indicibil Posted at 2017-3-29 13:49
Yes, that's what I'm worried about too, that the motors will stop in flight. The vps still works sometimes, especially at low altitudes but the random errors appear everywhere.

I've uploaded the flight log for the flight where it exhibited a lot of vps errors (the second video in the first post):

Looks good at take off up to 15m where the ultrasonics are supposed to stop working.  After that it seems to have bad readings, maybe when you are using most power?  And the readings at landing are rubbish.  Could just be a loose connector inside.

Your's isn't the only one I've seen readings on the ultrasonics at altitude, it may be worth looking at logs from before the crash, I wouldn't be surprised if it was nothing to do with the crash and you just noticed it when checking for damage, in which case it could be a warrantee issue...

indicibil Posted at 2017-3-29 14:00
Regarding the obstruction of the ultrasonic sensor, I've checked again and again, it looks clean, nothing in the way, I carefully cleaned some more the fine grid covering the sensors. It should be something visible, large and solid in order to bounce back the sound wave so quickly.

Yes, I was really wondering if you had mounted something extra on the aircraft.
But the bad readings in that log do not look like an obstruction, more like a bad connection.

Definitely, I'm in the process of analyzing all the flight logs before and after. So far I've analyzed most post-crash flight logs and only one before-crash, which looks very good.

The same type of graph for a flight a week before the crash (I removed the zero values):



The first two flights after the crash also don't exhibit the problem. First post-crash flight, IMU vs VPS altitude:

indicibil Posted at 2017-3-29 14:22
Definitely, I'm in the process of analyzing all the flight logs before and after. So far I've analyzed most post-crash flight logs and only one before-crash, which looks very good.

The same type of graph for a flight a week before the crash (I removed the zero values):

Do you have a gimbal guard installed?

Mobilcams Posted at 2017-3-30 11:21
Do you have a gimbal guard installed?

No guard installed, no custom part on the drone whatsoever.

I've calibrated right now the VPS via DJI Assistant 2, don't know if it will help. Also I did another IMU calibration just in case. I will test it in the weekend and let you know if there's any improvement.

I also did a short flight today with downwards VPS disabled, a bit shaky at take off, seems to be moving a bit more when close to the ground and the first few seconds it has a tendency to descent. Otherwise it is fine.

Nigel_ Posted at 2017-3-29 13:27
I would be worried about descending with the throttle full down with it acting like that - it might decide that it has landed and turn the motors off?   There have been a few crashes with it landing mid-air but I don't know if this is the same situation.

Presumably this is a problem with the ultrasonic sensors, not the vision sensors, so calibration of the vision system in DJI Assistant wont help?

This is exactly what happend to me, it turned itself off and fell from 700+ feet... See my post here http://mavicpilots.com/threads/h ... .19439/#post-224490

My p4 is having the same issues. No crashes though. Sometimes the vps will show false readings even within the vps range. I have to either put it in sport mode or disable vps to bring it down. I heard the same issue with the mavics were solved via firmware.

This is what mine looks like.