I took off just after sunset, and by the time I went to land, it was too dark to clearly see the MP.  So, I initiated a RTH.    I've attached a recording of the DJI app's log player, showing stick inputs, warnings, and flight path.

What concerns me is the circular pattern at the very beginning of the video - the point where I initiated the RTH.   Keep in mind, I did do a calibration (despite not being promted to do so) shortly after receiving it.  I figured, coming from CN, it's probably worth doing.  I did so in a large open grassy field - no metal anywhere.

So, the RTH brings the MP to just above the homepoint, when suddenly the 16 satellite GPS signal goes from full to a single red tick.  The MP drifts off to the east, and then south, at 10-12 MPH.  Because of how dark it was, and me looking at the screen trying to figure out why I don't hear the descent, the drifting was nearly 1000' south before I starting to apply directional inputs on the sticks.  At that point, I had completely lost my bearings, and panic had set in.

Fortuantely, I was able to get it pointed in the right direction, and manually guided it back in for landing.

Admittedly, my fault was with taking off after sunset (though still light), and for relying so confidently on the RTH capabilities.    Given that it was a cloudless, windless night, I had no reason to expect a sudden GPS/compass failure that would not only render the RTH feature useless, but would literally send the MP off yonder.

I sent the log to support, and after several rounds, they concluded that this was acceptable behavior given the conditions.   My primary concern was with the unexplainable movement of the aircraft without any stick input.   During the day, this would have been easy to identify and correct.  But in the dark, it proved nearly catastrophic.

That is scary.... I'm wondering what is the issue with Mavic and the sudden loss of GPS.  You are not the only one who has experienced this problem.  I also wonder why DJI thinks that this is acceptable behaviour given the conditions if it's a clear windless night and the Mavic is up at 400 feet..  Are there power lines or big antennas nearby?  Seeing the weird "toilet bowl" at the beginning of the video already indicates that there is a problem, but the only warning I see is "Strong interference to Remote Controller", no warning about any GPS issue...

Did you by any change have the VPS on? My gathering after seeing so many similar incident is that a bug in the fw swich mavics to ATTI mode when its too dark( or too bright) for the VPS to work properly. Few others fly at location under conditions that are perfect for GPS signal. ANd they did not really lost GPS as it still show 10+ of sats and the distance meter still work properly o. Just the Mavic seems to think that the GPS are lost and switch itself to ATTI. Hopefully Ken will be able to ask the devs

MAvicMinh Posted at 2016-11-26 03:08
Did you by any change have the VPS on? My gathering after seeing so many similar incident is that a  ...

That, in a nutshell is what support said - that since I was flying in the dark, VPS couldn't see the ground and combined with the ATTI mode, it simply started to drift.   They suggested that I switch to dynamic homepoint, which would have the MP land wherever the controller is.   That sounds good if I'm in an open field, but the reality is, that just causes other unintended consequences - like standing in the shade of a tree while flying, and on RTH, it tries to land in the tree.

The closest transmitting antenna is about 1.5 miles away (a cell tower).  Power lines (69kV) are about 0.5 miles south of the home point, but only about 50' up - and I was at/near 400.

I realize GPS is an imperfect science, but surely whatever 'plan b' routine is invoked when there is a GPS-compass discrepancy out not cause the unit to drift at such a fast rate, without warning.  

For me, I'm done with flying in the dark (or even dawn/dusk).  Without a usable video feed to obtain visual cues, the lack of any indication/warning of drift, and ultimately the seemingly random movement w/o stick input, makes for a disastrous recipe.

MavicUser Posted at 2016-11-25 23:07
That is scary.... I'm wondering what is the issue with Mavic and the sudden loss of GPS.  You are no ...

Yes, and if you look at the initial approach to the homepoint, you can see a big, unusual circle as well.  Then, as I'm bringing it back manually on the last leg, you can see it's constantly fighting my input - it wants to loop.  There are several curly cues in that return path.   Very disturbing.

Had the same issue!
I was flying at dusk with permanent 15+ GPS satellites. After flying for about 5 minutes in the same area, suddenly it lost GPS signal, but still had 15 satellites(!!), and it switched to ATTI mode, where it drifted away very fast and lost height.
Since I was flying LOS, I was able to recover it after some seconds, but this happened twice.

Sudden and unexplained loss of previously solid GPS signal followed by erratic and sometimes uncontrollable behaviour in ATTI mode is a major worry. Is this also a known issue in the Phantom drone lineup? I'm considering cancelling my Mavic order and waiting on the Phantom 4 Pro. Or cancelling and waiting for new firmware that fixes the serious issues like this one.

That circular motion looks like a compass problem, it's what would happen with a bad calibration, but it impossible to know, did it go to Atti mode, flying at night look for the red lights this is the front of the aircraft, it might help if get disoriented,
All is well and ended well, good job..

My problem was I couldn't see the red lights.  I didn't realize it had drifted and by the time I looked up to see where it was (in the dark sky) it had already moved about 500' south of my position.  I was simply looking in the wrong place.  By the time I actually caught a glimpse of it in the distance, it was over 1000' from the home point - and speeding away.   I fly in ATTI by choice with my P2V+ all the time, so it's not an issue of familiarity.  When I heard the buzzing props, I assumed it was over my coming down.  It was instead speeding away in the dark.

I think DJI needs to update the firmware so that if it drops to ATTI mode, the Mavic should IGNORE/DISABLE all input from the GPS/Compass instead of taking those into account.  I have a feeling that all these weird flight oscillations are because of some kind of problem with the compass and/or the GPS..   The whole point of ATTI mode is to enable the pilot to fly the aircraft manually, not to fight the controls of the aircraft to correct for whatever problems that are causing the aircraft to drop to ATTI mode in the first place.  

However, what is more worrying is the fact that erkme73 sent his logs to DJI, and they have concluded that this is ACCEPTABLE behaviour, and there is nothing to fix...

@ erkme73, did DJI ever tell you what was the cause of the problem?  

Had similar problem twice and from what I read it happens while coming back from flight. Can it be some interference with remote at short distance that is throwing gps signal away while is getting close to remote... maybe some wifi interference? (Pretty sure it was on airplane mode) battery levels 40-30% ... if compass problem Why is that happening only when about to land. Just guessing

MavicUser Posted at 2016-11-26 21:38
@ erkme73, did DJI ever tell you what was the cause of the problem?  

@mavicuser - yes and no.  They said that what I experienced was due to the VPS system not being able to see anything (since it was dark).  In fact, here's the reply I got from them:

Here is what happened to you, you set your home point and your GPS is working good but when you initiated the RTH and when the unit was descending you had a weak GPS signal that is why it switch to ATTI mode which detects the VPS and detected other home point, what we best recommend is before flying your unit always do the Compass calibration and to set the Dynamic home point in your App, using this dynamic homepoint the unit will detect and land where the remote controller is.

I pushed back a bit saying that I didn't want to forego precision homepoint by choosing the dynamic option.  Too often I'm standing with the controller in the shade of a tree.  The last thing I want is for the MP to follow me to the shady spot.

They replied:  

Thank you for providing these information's. With regards to your concern, one factor that caused the GPS signal dropped is the interference within the area. The unit shifted to Atti mode when GPS is low or lost and use VPS Barometer to stabilize but it is still GPS assisted and will continue to Return to your set home point, however, you have mentioned on the initial email that it was dark and did not notice how far the unit drifted away. While the unit is in Atti Mode it uses VPS barometer and Vision sensor on the bottom of the Aircraft to locate the homepoint, since it is  dark the VPS Vision Sensor cannot locate the recorded or exact Home point Location.

We recommend you to fly in a different location without any interference and check the flight mode every flight.

So, if I read that correctly, it was the GPS signal that failed (vs. the compass data being bad).  And, the fix is to fly in a different location.  That is, don't fly it when you're at home.  Great.   Not really an answer.

erkme73 Posted at 2016-11-27 18:24
@mavicuser - yes and no.  They said that what I experienced was due to the VPS system not being abl ...

Thanks for attaching the DJI replies...

It's weird that VPS and Downward sensor would seem to lead the Mavic astray by making it drift off that quickly... I'm wondering if turning off VPS and Downward sensors could mitigate this problem, so that in ATTI mode, it would only be drifting because of the wind, and not because of VPS/Downward sensors..

One issue I still have regarding this, is the toilet bowl effect even in ATTI mode... It's obvious that ATTI mode still does not disregard compass/GPS, which is still a big problem in such cases if you have a bad compass calibration...  

MavicUser Posted at 2016-11-27 22:41
Thanks for attaching the DJI replies...

It's weird that VPS and Downward sensor would seem to lea ...

No, and I call BS on that explanation.  When at 400' VPS doesn't do any good - never mind at night.   

I just had another TBE incident - it was the wildest, fastest one yet.   When I finally got it on the ground, I attempted to calibrate the compass.  I had only 15% of remaining battery, but it didn't tell me not to proceed...

I must have tried it 10 different times - arms wide, near, twisting it in my hands, standing, kneeling, everything.  When I finished with the second (vertical) rotation, the LED blinked red and the calibration failed.  EVERY TIME.  I literally cannot calibrate my compass now.    The thing was smoking hot.  So maybe there is a temperature/batter voltage component?

By the way, is there a way to turn on some form of email notification of a new reply on this forum?  I have a hard time remembering to check.

Since you had sent the log to support. Could you please provide me with your ticket number?
I'll help to check it.

erkme73 Posted at 2016-11-28 11:39
No, and I call BS on that explanation.  When at 400' VPS doesn't do any good - never mind at night ...

That sounds bad... Maybe a hardware problem?

Several points:

1. You should not be flying at night or if you choose to, caveat emptor. Most countries have laws making night flying illegal and you gave some reasons to support why (no visibility of craft)
2. VPS does not operate from an altitude greater than say 10m or so. So that's irrelevant.
3. The Mavic has dual compass dual IMU redundancy. The margin of error in a GPS sensor loss situation displayed in your original post looks narrower than the width of a house (~10m). Its pretty darn good by any measure. Those of us who flew P1's and P2's experienced larger TBE's in broad daylight.

Bottom line. Don't sweat it. Next year, they will have a multiple redundancy inertial navigation system that will dead reckon the flight path for the Mavic 2 -

hallmark007 Posted at 2016-11-27 04:22
That circular motion looks like a compass problem, it's what would happen with a bad calibration, bu ...

That circular motion looks like a compass problem, it's what would happen with a bad calibration, but it impossible to know...
I would not be so quick to blame the compass. Remember a circular motion can occur with a rock solid constant heading. The mad spiral I posted over at the Litchi ATTI Problems looks like a compass issue when you observe the ground path. But a review of the video and the flight log clearly shows the aircraft heading was stable throughout.

One fortunate thing these ATTI events seem to have in common is the GPS module retains a satellite fix throughout even though the device seemingly cannot cannot correctly calculate speed or sensibly locate itself for navigation. Therefore your RC will still display a distance downrange which is a godsend for recovery if the aircraft is difficult to see. Just pitch in each of the four directions until you fiind which one counts down the distance and you will know it is on its way home.

Logger Posted at 2016-11-29 11:43
That circular motion looks like a compass problem, it's what would happen with a bad calibration, ...

I guess your right, I wasn't blaming it on compass, just merely pointing out what some of the pitfalls could be, it's never easy to draw one conclusion,
Have you come to your own conclusion as to what may have happened to your own drone, was it Mavic or was it litchi,
I haven't received my Mavic yet it arrives today and I'm away until Friday , but will be flying litchi, as I have with no problems, with P4 ,
I'm am interested to know what your thoughts are on what caused your problems.
Good. Luck

No I saw you said it was impossible to know. I realise you were only speculating, as am I.  I submitted the issue to Litchi. Whether they look at it or not remains unknown. Perhaps Litchi thinks it is a DJI
caused issue and DJI reckon it is a Litchi issue.
OTOH if these events were unknown to DJI you would think they would be interested in my Logs. I have publicly said I have all the files of my flights with two ATTI events in two days. Their lack of interest probably means they are well aware of the issue and are hopefully working on a solution.
In my case I currently think the fault is in the GPS module firmware, while my events are getting triggered by something Litchi is doing when the mavic passes way-points. But this is pure speculation of course.

erkme73 Posted at 2016-11-26 22:04
That, in a nutshell is what support said - that since I was flying in the dark, VPS couldn't see t ...

So basically the Mavic's VPS isn't as good as the Phantom's (which I never had trouble landing in the dark while VPS was enabled)?

It's possible that switching to Sports mode is the cause for the TBE issue.  See the post over at mavicpilots.com at: http://mavicpilots.com/threads/s ... 8/page-2#post-30253

I would suggest that everyone stay off Sports mode for now...

MavicUser Posted at 2016-11-30 11:18
It's possible that switching to Sports mode is the cause for the TBE issue.  See the post over at ma ...

There has been an update to this. Please see There has been an update to this. Please see http://mavicpilots.com/threads/s ... 8/page-3#post-30944

BudWalker Posted at 2016-11-30 06:55
There has been an update to this. Please see There has been an update to this. Please see http://m ...

This is the correct link: http://mavicpilots.com/threads/s ... 8/page-3#post-30944