A week ago I went in the countryside to get some footage consisting mainly of snowy hills and valley. Everything ok.
Saturday I went back to get more video since there was much more snow than before.
---
First Episode.
As always, after the checklist, I took off and I keep hovering the P4 at about 3 mt for one or two minutes before sending it away.The P4 suddenly moved forward itself.... OMG !!!! A flyaway !!! ..
I pushed backward the throttle and the P4 stopped.
I thought I could have accidentally touched the right stick (but, No, that wasn't the cause).
Ok... wait a bit more and... the P4 started to fly backwards itself... OMG*2 !!!! now it wants to kill the owner !!!
I pushed forward the throttle and the P4 stopped close enough to be hand grabbed. End of the flight.

I attempted to think what could be happened. I recalled about VPS sensors getting "drunk" when they cannot "see" a clear pattern, but the issue was mainly related to altitude drops, not horizontal movements.
So, doubts start to arise.... What about Ultrasound sensors ? GPS signal was good during the takeoff.
Do VPS and ultrasound sensors override GPS signal ?
I read many threads about behavior of P4 under and over 10 mt of altitude; about issues when flying at low altitude over water surfaces, but once again I expected these issues to related to altitude, not to the horizontal position.
Then I have searched in the device the .txt file for an analysis, but the folder flightrecords seems to contain only logs until november 2016.
BTW.. I attempted a second flight sending the p4 over 10 mt, and everything went smooth this time.
---
Second Episode
Once again, I wanted to fly at low altitude over a little river.
As you can notice in the video, the P4 suddenly had an altitude drop consisting in about 1-2 meters. When you see the P4 regaining altitude, it was because I immediately pushed the left stick up to regain some altitude...
Once again, I took the P4 back to me and aborted the flight.
It was a crash 100% if I wasn't quick enough.
I recalibrated the IMU in cold environment, then I repeated the take off.
This time I put a pair of gloves under the home point in order to give the VPS some pattern "to see". All went ok this time.
But I wish to understand more about VPS, Ultrasounds sensors, barometric sensors etc etc... I understand the basics (i.e. if the height is > 10 mt then the barometric sensor overrides VPS, or that the VPS doesn't work well in low light conditions or in uniform patterns and so on),  but it could be useful to understand more clearly how those sensors cooperate togheter, who overrides who, and above all, how to prevent damages and flyaways in situations similar those I have mentioned in this post.
BTW... I think events similar those mentioned here may explain pretty well why expert pilots loves ATTI mode so much.

The two images above  refer to the first episode (P4 moving itself forward and backwards).The first image is a screenshot of the dji go app showing the P4 moving forward (and the RC right stick pulled backwards to contrast the flyaway).
The second image is specular to the first, with the RC right stick pushed forward in the attempt to stop the P4 that was moving backwards itself.

The video refers to the second episode (altitude loss)
Here is the link to the video :

Have you completed all necessary important calibrations, including the RC sticks?
Did a proper compass calibration, away from any ferrous metal or electrical interference?


RedHotPoker

RedHotPoker Posted at 2017-1-23 03:38
Have you completed all necessary important calibrations, including the RC sticks?
Did a proper compass calibration, away from any ferrous metal or electrical interference?

Yes, I did a compass calibration in a field, far away from power lines, pipes or any metallic objects.
I performed the compass calibration after I came back from Thailand. By that time I also performed an IMU calibration in a flat surface (checked with a bubble level instrument). Since then, I flew about 10 times without any issue.
The RC  and VPS calibrations, instead, were done several months ago.
I forgot to mention that after a second IMU calibration I made two further tests :
1 - take off from a concrete surface. Result : perfect hoovering.
2 - take off from a snowy surface, but with a couple of black gloves left on the snow. I don't know if it was because of them, but the P4 kept hoovering in a solid-rock position.

If it can't see any detail in the snow then the Visual Positioning System can't keep the horizontal position steady and it will drift with the drift of the GPS signal which can jump around by a few meters - for warmup, and other occasions, always hover high enough that it can't hit anyone's head if it moves!

The ultrasonic sonar is only for height sensing, not horizontal position.  But normally vertical position will be kept by the barometer.  If you are flying over cold snow and you then fly over a warm river then the difference in air temperature could cause a drop in pressure sufficient to cause the aircraft to drop a bit to maintain the barometric altitude.

I don't think the ultrasonic sensors are ever used to maintain altitude except when ground following in the smart modes.  Their main use is for landing, auto land will slow the vertical decent speed as the ultrasonic sensors detect the ground approaching so that it doesn't crash into the ground but can descend at a decent speed when higher up.

Nigel_ Posted at 2017-1-23 05:54
If it can't see any detail in the snow then the Visual Positioning System can't keep the horizontal position steady and it will drift with the drift of the GPS signal which can jump around by a few meters - for warmup, and other occasions, always hover high enough that it can't hit anyone's head if it moves!

The ultrasonic sonar is only for height sensing, not horizontal position.  But normally vertical position will be kept by the barometer.  If you are flying over cold snow and you then fly over a warm river then the difference in air temperature could cause a drop in pressure sufficient to cause the aircraft to drop a bit to maintain the barometric altitude.

Thank you for the explanations, Nigel,
Reading your explanation regarding the VPS is now clear why I had two evident drifts.
I have no idea if they would be more than a couple of meters if I did not try to contrast them with the throttle.For safety reasons I always keep hoovering at about 3 mt to prevent injuries in case of drifts, but maybe a couple of meters more is better.
By the way, having some footprint on the snow plus a couple of gloves left on the snowy ground were apparently enough to get a perfect hoovering (see the picture). No drifts in this case, which can confirm your statements about VPS.

If barometric sensors are used to control the height even at 2-3 meters of altitude (of course excluding ATTI mode), then your statement about difference of pressure caused by gradients of temperature explains pretty well the fall noticed in the video. Indeed, the P4 flew over a river with flowing water (therefore surely above 0°C), while the surrounding borders could be a couple of degrees under.

Ok, well, not sure what else to mention at this point.  With a proper IMU and compass calibration already done.
Hopefully someone will have more info to share.

Oh by the way: "Hovering" only has one o. ;-)


RedHotPoker

I have no idea if they would be more than a couple of meters if I did not try to contrast them with the throttle.For safety reasons I always keep hoovering at about 3 mt to prevent injuries in case of drifts, but maybe a couple of meters more is better.

If you don't have many satellites and poor reception or interference then GPS can drift by 100 meters, but I think the Phantom switches to atti mode if accuracy gets worse than a few meters, so it then just drifts with the wind instead of making high speed unexpected adjustments.

We don't see many issues reported here on this subject so I guess the VPS works well in most circumstances and the GPS must have a good receiver for when VPS isn't working, we do need to be careful when flying after sunset though - don't get too close to people, buildings, trees if VPS may not be working.

Mine goes up and down about a meter while at low altitude over the snow, sometimes slightly horizontally but not much at all. Assumed the snow was a big factor. Better then I expected though

Silvestrone Posted at 2017-1-23 03:37
The two images above  refer to the first episode (P4 moving itself forward and backwards).The first image is a screenshot of the dji go app showing the P4 moving forward (and the RC right stick pulled backwards to contrast the flyaway).
The second image is specular to the first, with the RC right stick pushed forward in the attempt to stop the P4 that was moving backwards itself.

VPS is not magic and has certain limitations that are listed in the manual.
Over snow ticks several of the VPS limitation baxes and probably accounts for what you saw.
Did your Phantom work well above VPS height?

RedHotPoker Posted at 2017-1-23 10:54
Ok, well, not sure what else to mention at this point.  With a proper IMU and compass calibration already done.
Hopefully someone will have more info to share.

Thank you for the info RedHotPoker,
ah... and thank you also for correcting my english. This forum is great. In a single place I can improve my drone knowledge and improve my english as well .

Silvestrone Posted at 2017-1-24 23:51
Thank you for the info RedHotPoker,
ah... and thank you also for correcting my english. This forum is great. In a single place I can improve my drone knowledge and improve my english as well .

That's swell of you, thanks. ;-)
Where are you situated? Your location isn't listed underneath your forum handle.

Yes, it's a good place to be associated, and spend some time, on the forum..


RedHotPoker

Labroides Posted at 2017-1-24 18:15
VPS is not magic and has certain limitations that are listed in the manual.
Over snow ticks several of the VPS limitation baxes and probably accounts for what you saw.
Did your Phantom work well above VPS height?

Above 10 meters the P4 was keeping its position in a fantastic way.
Coming back to 6 months ago, when I started to use my p4, I recall that I had a similar issue on a pebbled pavement in very low light conditions. By that time the DJI Go App gave me a warning, and the drift was pretty slow. The last week, on the snow, the drift was really quick.

Silvestrone Posted at 2017-1-25 00:09
Above 10 meters the P4 was keeping its position in a fantastic way.
Coming back to 6 months ago, when I started to use my p4, I recall that I had a similar issue on a pebbled pavement in very low light conditions. By that time the DJI Go App gave me a warning, and the drift was pretty slow. The last week, on the snow, the drift was really quick.

If you get into a similar situation and need to be down in VPS range, it would be a simple matter to switch off the VPS and things should settle down.