By way of background. I am a very serious pilot. I am a video professional and I have been flying DJI platforms for more than 3 years accident free. I am currently in pilot school, because I believe in doing it right and that knowledge is safety. Lastly I calibrated my IMU and compass early that afternoon and had 2 epic flights without incident.

Fast forward 2 hours:
Right at dusk I wanted to do a quick bay flight. Shortly into the flight I got a warning that said: “Warning:]data exception (In Flight) (Using)”

Shortly there after on an attempt to get back the Phantom became erratic, losing and gaining large altitudes and alternating between being sluggish and then hypersensitive to stick input. Altitude and vertical speed telemetry became unreliable. After a long series of dramatic assents and descents I got the bird back within 1500 ft of me before it went down. I have heard a lot about these unexplained altitude issues and (hopefully for the common good) decided to spend the afternoon putting together a video of the visual flight data and the fpv to show the forum in detail what I experienced.

I would sure appreciate it if a few of you would put your eyes on it and give me your feedback. Ultimately I would like the group’s opinion on if this was truly unavoidable due to mechanical failure and if I will have a leg to stand on with DJI.

PS so as to be accurate I did not compress time on this 3min 8 sec flight so you can watch it all or the action begins a 1 minute 24 secs if you just want to cut to the chase.




Password: djip4

Frightening!!  You did well under the circumstances.  In what mode were you flying?  Did you try other modes?

Mate, really sorry to see this:-(
But - are you sure it's a barometer failure?
You wanna say that the aircraft altitude is actually driven by barometric information?
I'm a newbee but this seems to me impossible, I think that barometer should only provide you with information but any craft can't be ruled by this information though (if flying manual and not RTH or so).
When the air pressure suddenly drops or rises (suden change of air pressure matching altitude change of let's say 20m is nothing exceptional) the craft altitude follows this change?? Does not make sense, even GPS would provide more accurate altitude in changing weather or difficult barometric conditions which are always near any coast...

Sorry for your loss. This really give me the chills.
One thing I notice is that you took too long to press the RTH or to change the flight Attitude of the P4. Another thing is that I believe DJI should make that kind of warnings more noticeable on screen. I feel very sorry for your loss.

WOW, that's terrible.
It does look there there was a barometer issue. I would contact support and send the aircraft in for evaluation.

Did you recover the aircraft?

... (repeat post make by my iPad) ...

DJI-Ken Posted at 2016-9-15 12:57
WOW, that was a lucky save.
It does look there there was a barometer issue. I would contact support  ...

It wasn't a save it all, but a complete loss. The video shows the bird going into the water. Whether the issue is barometric pressure sensor failure,  or an IMU Failure. . .   Something went seriously wrong.  

There have been plenty of documented cases where the DJI phantom 4's altimeter is grossly in accurate.    This makes no sense to me. Barometric pressure is a useful tool, but is slow, and prone to inaccuracies. There are two GPS units built into this device. Why is it not pulling altitude data from these units?    (Note:  on a different discussion board, someone stated "there is no altitude data in GPS". All I can say is that is totally wrong. Even my car can give me my altitude on my single unit GPS when it has at least three satellites )

Man. . . sorry for your loss. Hopefully your bird  can be replaced.

Of course GPS gives altitude just by logic how does it work, it's defined like point in 3D space, but altitude is less accurate, think about basic geometry - this dimension is defined by satellites all above your head at the best 150 degrees sector but the latitude and longitude is defined by 360 degrees resection.

I'm trying to understand what tool determines positioning at what time and what went wrong here - air pressure change and phantom follows it to original measurement? Non sense, 20 m AGL and 50 m away from motorway would be different air composition/density/pressure than 500m away in the same altitude. So I imagine that the Phantom's brain gets information of any vertical movement by -barometer, gps, gyroscope and accelerometer (even your bloody phone knows if you moved it half a metre down or up) and compass and from some combination determines its position and what's happening...
Just couple days ago I've seen a video of accident here when phantom was flown from the roof, made a circle and crashed into the roof, the conclusion was that it was due to compass calibration in interfered environment, but I still don't understand it - when the pilot stops touching the sticks why is the craft moving at all?? Gyroscope itself knows about change of position. GPS? So why is it moving and not hoovering in place?

To be honest I think that these accident can be caused by RC interference - air is full of various wavelengths, can't the craft get non existing RC signals? - or by failure of the brain itself.

Ben Tichy Posted at 2016-9-17 19:06
Of course GPS gives altitude just by logic how does it work, it's defined like point in 3D space, bu ...

'To be honest I think that these accident can be caused by RC interference - air is full of various wavelengths, can't the craft get non existing RC signals? - or by failure of the brain itself.'

Not really, for the very reason you have to bind your aircraft to your controller, another source of interference is not going to take control of your aircraft. A bad compass calibration causes exactly the scenario you described, the aircraft flies in an arc under its own control.

I wonder why, at no time you switched to Atti mode which is, in my point of view, the first thing to do in such cases ?
RTH will never be a solution when the craft encounter such behavior as this procedure relies on perfect data acquired by the quad : at the time you engage RTH, the height value is way wrong.
Of course, it is always easy to tell anything after a crash and I don't want to give any lesson here (I never had any problems with my P3P nor my P4 so I don't know how I would react in such case) but : when you saw for the first time the error message, the quad was perhaps high enough to switch to Atti mode and get it close to you manually ?
Hope DJI will do something for your P4...

DJI-Ken Posted at 2016-9-15 21:57
WOW, that was a lucky save.
It does look there there was a barometer issue. I would contact support  ...

Did you look the video from the beginning to the end ?

"It does look there there was a barometer issue" : obviously as this is what the error message shows !
"...send the aircraft in for evaluation" : why not if there is a postal service working underwater...

Ulysse Posted at 2016-9-17 03:29
I wonder why, at no time you switched to Atti mode which is, in my point of view, the first thing to ...

Having Multiple Monitors for the same variable is key in Aviation.  The Airbus has three of everything for critical system. . . there will be a HORIZON for the Pilot, a Horizon for the 1st Officer, and a third HORIZON on the central console as a breaker of any ties. . . 99.99% they all agree. . . but when they don't, 2 out of 3 wins.

The Phantom has several inputs but chooses to use only a select few.   SONAR (which appears limited to Near-Ground Operations. . . at it reads "8.3 feet" when at 390 feet). . .  IMU's which measure acceleration/inertial changes which, when integrated, can give an every increasing error distance from Home-Point. . . GPS, which is converted into 2D Long/Lat coordinates for display.  In very small distances, the SONAR is probably the most accurate. . . then GPS, then Barometer. . .

Whatever Formula DJI is using it would be completely unacceptable in commercial Aviation.
When combining the Altititude Errors with a WAYPOINT MISSION you are in for serious trouble.   Easily reproduced in my Bird... fly around a park... set a few waypoints with reference to KNOWN heights (my son and I took a 100 FT measuring tape to a local park and ran this experiment)... then run the mission.  The Long/Lat mission was nearly flawless... the ALTITUDE was off as much as 16 feet when the highest point of the mission was only 55 feet.  At one point, the bird was 9 feet off the ground (I could almost touch it) saying it was 22 feet.  The Backstop of a Baseball park, measured with tape at 15' 11" was coming in at 28 feet.  But, as the Bird hovered... the altitude slowly dropped... 25'.. then 22'... then 19'.. .then 17' then finally 16'.    The BAROMETER IS LAGGY.

This absolutley reinforces the axiom... "YOU MUST MAINTAIN VISUAL LINE-OF-SIGHT AT ALL TIMES... in order to recovery the bird should it go awry..."  Yes... safe... but the buggy Altimeter severly limits my options of AUTOMATED missions.

Ulysse Posted at 2016-9-17 18:39
Did you look the video from the beginning to the end ?

"It does look there there was a barometer  ...

Now I see, what did support say, you've created a ticket correct?

TheodoreRex Posted at 2016-9-17 14:36
It wasn't a save it all, but a complete loss. The video shows the bird going into the water. Whethe ...

I see now and I edited my post.
I do not know what caused the failure but the user will have to see what support says after they see the video and analyze the flight record.

Sorry for you loss!
I am curious if flying over water has anything to do with this P4 behavior or just a hardware failure; keep usposted.  I fly over water all the time, this is not a confidence booster.

Using GPS would be the way to go for altitude adjustment, but how often does the P4 get a GPS calculation? Combined with the barometer these things shouldn't happen. This looks like could be HW *and* SW issue.
Definitely lowers confidence, be good if DJI could let US know their analysis of the flight, but I doubt they will :/

GPS is normally so far off in regards to altitude that I hope it is never used for such.     I have seen many GPS units jump over 200' +/- in altitude when moved only a few inches.

calls4u2 Posted at 2016-9-18 07:17
Using GPS would be the way to go for altitude adjustment, but how often does the P4 get a GPS calcul ...

GPS is never used for calculating altitude by anyone, as it is wildly inaccurate. And the aircraft has a chip dedicated to receiving GPS signals and making positional calculations hundreds of times per second, otherwise it could never stabilise the aircraft in a hover.

TheodoreRex Posted at 2016-9-17 17:38
Having Multiple Monitors for the same variable is key in Aviation.  The Airbus has three of everyt ...

I know all of this but :
1. A P4 is not an Commercial airplane with a lot of redundancy on every system.
2. this doesn't answer my question about not going to Atti to try to save the P4, moreover after you was aware there was a problem with the barometer (and while still receiving the video from the quad).

This has been written several times : the key when the quad behaves erraticaly is to swith to Atti mode (and it is more or less the only thing to do)
No blame at all, you ask for comments and these are mine...

Ulysse Posted at 2016-9-18 00:07
I know all of this but :
1. A P4 is not an Commercial airplane with a lot of redundancy on every s ...

It's difficult to know what to do if the barometer develops a fault in flight.

ATTI mode is literally an altitude hold mode and given that the manual states that in ATTI mode it only uses the barometer for height measurements, it might make the situation worse to force the Phantom to rely totally on the very sensor it is having issues with.

Wonder if there's a case for dual barometers in future designs....

Yes mine crashed straight down . I sent it DJI and they replaced it.

Geebax Posted at 2016-9-17 23:57
GPS is never used for calculating altitude by anyone, as it is wildly inaccurate. And the aircraft ...

I use RTK (real time kinematics) GPS at my work every day and required accuracy is +/- 50mm! It receives about one signal per second.
Of course it's a survey tool and you get on-line corrections but even without these is the accuracy +/- couple metres.
I was working with RTK GPS since ten years ago, when there were no Glonass for wide use so I often worked with just five sattelites, slow but it still worked with accuracy of couple cm.

Just wanna say that decent commercial GPS receiver in an open area (what's better than to be in the air) with 18 sattelites must get you within 5m altitude, enough to save the craft in case other systems fail... +/- 200m is out thought, I've never seen such a thing :-))

davidson_g Posted at 2016-9-18 07:22
It's difficult to know what to do if the barometer develops a fault in flight.

ATTI mode is liter ...

I wonder when there's gonna come a tool when in emergency you could just turn off all bloody altimeters and GPS's and manually save your craft just by FPV flying :-(

Ben Tichy Posted at 2016-9-18 21:02
I wonder when there's gonna come a tool when in emergency you could just turn off all bloody altim ...

I was going to mention that too but having had some interesting times mastering throttle control on nano hexes to maintain altitude, I wonder how a heavy drone like the Phantom would handle under totally manual throttle control. I'd imagine vortex ring state would be a very real risk whilst trying to figure out the correct throttle power in an emergency.

I'm sure there must be a way to use the IMU to avoid large swings in altitude in the event of barometer failure, perhaps by deliberate bobbing up and down so as to cycle between positive and negative vertical acceleration and thus hopefully maintain some sort of altitude control.

My reasoning behind the deliberate bobbing up and down is that the accelerometers only measure acceleration and would not detect a steady ascent/descent however fast. On final ground descent, hopefully the ultrasonic sonar could then be brought into play for a controlled landing.

Ben Tichy Posted at 2016-9-19 06:00
I use RTK (real time kinematics) GPS at my work every day and required accuracy is +/- 50mm! It rec ...

RTK is a completely different system to normal GPS reception. It produces the accuracy because it has a receiver/transmitter unit on the site where the measurements are being made, and that device is retransmitting the carrier from one of the GPS satellites. This is not the case for simple GPS receivers. A good GPS receiver can still produce very accurate rsults, but not in meauring altitude.

Try Googling 'gps altitude' and you will get a mass of hits, all saying that it is not accurate. This from Wiki: 'Global Positioning System (GPS) receivers can also determine altitude by trilateration with four or more satellites. In aircraft, altitude determined using autonomous GPS is not reliable enough to supersede the pressure altimeter without using some method of augmentation. In hiking and climbing, it is not uncommon to find that the altitude measured by GPS is off by as much as 400 feet depending on satellite orientation.'

In our experiment of the Phantom 4's GPS accuracy. . . we found the Long/Lat readings are incredibly accurate.  We measured 100' by tape. . . then flew the bird from Origin to 100 feet and it was accurate to within a few inches. . . We even flew it about 3 minutes all over a local park. . . and again. . . it's X-Y (Long/Latitude) reading were within inches.  The ALTITUDE, however, was widely inaccurate.

That begs the question. . . how could it be so accurate in 2 Axis's (X & Y translation) and so bad in the 3rd axis (Z - Altitude)?  A few inches compared to a many yards. . . the answer, I think, lays in our Imagination and Trigonometry. . . We envision GPS like This:



The Geometry would imply that X/Y/Z error should all be about the same... so each axis would have similar errors.

But in reality... the Geometry is more like this:

Shallow angles

The SATELLITES are relatively low in the sky (100 miles?) but are very far away (possibly thousands of miles).   This makes a really long/narrow triangle... the SIN/COS of Trigonometry... the LONG side should be fairly accurate... but the SHORT side could have a much higher error of measurement...  All Satellite signals are LINE-OF-SIGHT... so it a fairly shallow plate...

DO WE HAVE ANY SURVEYORS on this discussion?  They would have more to say...

sorry for youre loss, Please keep us updated because this is not boosting my confendence in my p4

I still seeying the image of th p4 going down the water... SMH

TheodoreRex Posted at 2016-9-21 20:19
In our experiment of the Phantom 4's GPS accuracy. . . we found the Long/Lat readings are incredibly ...

Yep, I'm a surveyor (who recently bought P4 and trying to get a head around :-) )

You hit the nail with the second picture and your explanation (and perhaps you're better with describing things) - the receiver gets a signal from satellite, in that signal is information about exact position of the satellite and a time. From that the receiver gets a length/distance to that satellite (and nothing - or nearly nothing - else).
So when it gets several such a distances from various directions it's a simple resection (simple for computer, haha) to get a position.
But the point is in that directions, geometry - for an altitude, like on your second picture, it gets signals from at the best 150 degrees sector as all the signals comes from above, but for lat/long it gets signals from all around 360 degrees (ideal conditions of course - but these are common for flying thing) and considering that there is always some mistake in any measurement you can see why altitude is much less accurate - if you tape measure known distances from two known points, your point will be more accurate if one of those points on your left and one on your right, then if these points are both ahead just a bit apart.
Number of satellites also counts of course.

Your first picture - the signal/view of satellite must be direct, it never gets reflected signal which might be technically perhaps possible, but it's useless - no chance to get exact time what was needed to reach distance sat-receiver.

This discussion got me to idea to examine common GPS accuracy - I will compare elevation what I get by basic Garmin satnav in my car to the point I survey with RTK GPS (+/-50mm in elevation). Now I parked on a known point, satnav signal is obstructed with a two storey building from one side and still difference is just 2-3m ;-) Although I don't know if the altitude information comes from GPS measurement or it's based on 2D position and terrain map information.....

It's not gonna take long time and I will go for Suunto watches to play also with barometer, I wanna get deeper into meteorology but also want to find out about affection of altimeter to weather changes....

Ben Tichy Posted at 2016-9-21 15:31
Yep, I'm a surveyor (who recently bought P4 and trying to get a head around :-) )

You hit the nail ...

Humidity for one...   Water Vapor has a lower density than the Oxygen/Nitrogen mix we call Air... you can know this by watching water vapor rise from a pond... or why Clouds form up there instead of down here... when water vapor moves in... the Barometric Pressure drops... since the increased water vapor displaces the drier "higher pressure" oxygen/air.

The Phantom 4 uses GPS to fix it's point in space then references the Barometric Pressure... and makes adjustments to the implied Altitude ... I say, IMPLIED ALTITUDE because Barometric Pressure is NOT a measure of Altitude... but calculates an estimate... based upon a reference (HOMEPOINT GPS) value.

It will lag.  I wonder if the FAN inside the drone is a factor in the lagging Altitude... perhaps a small modification can result in a more rapid equilibration...