DJI Mini 2, Spiral out of control, 2022-02-15

[Wolooloo]

Hi guys,

I have (should say "had") a DJI Mini 2 that I have been using over the past 6 months. It has about 275 flights with 50 hours of flight time and 650 kilometers.

Today as the drone took off, it flew forward properly for a few seconds before suddenly reporting a rotor motor error and subsequently spiraling out of control. I stumbled for a few seconds and then I thought to keep it as high as possible, meanwhile trying to use the directional joystick in all directions to see if any would bring it back towards me. I was able to keep it high up until I thought to click return to home when it very quickly lost all control and went for a free fall into the nearby pond... thus ending its life.

I was wondering if you could have a look at the flight path data to see if you spot anything. I loaded the log file into PhantomHelp and AirData and notice that the height of the drone is all screwed up (it went into negative values before my attempt to raise the drone higher to buy time). I didn't even see that at the time and I have no idea how that could happen, but my hypothesis is that one of the propellers must've come loose and fell off?

Attached is the TXT log file and the only DAT file existing on my phone (not sure if it is the correct one, but the time stamp kind of matches).

Thanks,
John

 

[Wolooloo]

One more thought as I was researching the issue with the DAT files. From my understanding, if you keep "Auto-sync Flight Records" ON, the system automatically deletes the files when it connects to the internet and updates the DJI database or whatever. I think I must've turned it ON in the very beginning when it sounded like a good idea, but what are really the advantages of doing this? Moving forward, what is the consensus: is it best to keep it ON or OFF, and why?

 

[Meta4]

my hypothesis is that one of the propellers must've come loose and fell off?

The trouble started at 26.5 sec when the drone started to rotate anti-clockwise, pitch forward and roll to the left.
This suggests the front right motor was not providing lift.

You had been climbing at full throttle so the altitude continued to increase for 0.3 sec after this but then started losing altitude.

Because of the extreme and complicated joystick inputs given, it's difficult to tell how the drone responded to your controls.
The false negative altitude reading is puzzling and means that altitude data is unreliable for this incident.

If you had lost a prop or motor, the drone would not be able to maintain or recover altitude.
That the drone did recover and maintain height (for a time) suggests an erratic malfunction of either the motor or the ESC providing power to the motor.
Multiple Motor is Blocked warning messages after 94 seconds support this.

 

[slup]

...I was wondering if you could have a look at the flight path data to see if you spot anything... my hypothesis is that one of the propellers must've come loose and fell off?

As the mobile device .DAT log from a Mini 2 is encrypted & can only be read by DJI staff ... this will be a bit of guess work, this as the telemetry aren't conclusive when it comes to losing thrust from one corner.

The Mini starts suddenly to yaw CCW uncommanded at 26,4sec into the flight ... this yawing is alone typical when a prop or motor fails as it will be a rotational torque imbalance as it is 2 CW torquing + 2 CCW torquing usually.

But when a motor or prop fails the yawing is always coupled with big changes in both the pitch & roll axis ... but we don't see that in this case. Yeah, the pitch & roll moves ... but not as much as they usually do when a corner loses thrust. Furthermore both pitch & roll stays relative constant during roughly 1,5sec which aren't typical for a thrust loss (have red framed that in the chart).

Later in the flight you also manage to gain height ... that shouldn't be possible if one corner isn't performing ...

Here below the telemetry ... (the chart marker where the incident starts & check the graph legend under the chart for what the graphs represent)

(Click on the chart to make it larger)

If checking off the velocity differences between the GPS & IMU (which should be very near 0mph if all is healthy) ... we see big deviations, especially in an easterly direction ... big deviations usually indicate IMU problems.

(Click on the chart to make it larger)


Going into the .TXT log message stream shows below ... it's a couple motor errors initially yes, but the majority concerns thing's that make your Mini 2 fly stable ... Gyro failures & GPS mismatch messages. (All the gimbal motor errors comes from the craft rotation ... so doesn't indicate more than that).

All this ... together with the very wrong barometric height readings strongly suggest that this is a IMU failure involving the gyro.

(Click on the picture to make it larger)

... and right in the end of the log your Mini 2 falls into ATTI mode ... with between 13-18 satellites locked. This indicates that the flight controller have given up & left all stabilization over to you.

 

[Meta4]

... and right in the end of the log your Mini 2 falls into ATTI mode ... with between 13-18 satellites locked. This indicates that the flight controller have given up & left all stabilization over to you.

This happens when the drone is tumbling/spinning fast and mostly upside down, preventing proper GPS reception.

 

[slup]

This happens when the drone is tumbling/spinning fast and mostly upside down, preventing proper GPS reception.

Yep, could have been ... but this time the GPS reception still was good, and the FC had before kept the GPS level up even though the craft had tumbled for a long time earlier. Together with the suspected gyro failure it's quite typical for the FC to eventually give up if the craft is airborne long enough.

 

[Deleted member 103366]

Jeepers what a flight! Do you by any chance has a video cached on the phone? Do you think it went anywhere near the report height, 618ft ?

 

[Meta4]

Yep, could have been ... but this time the GPS reception still was good, and the FC had before kept the GPS level up even though the craft had tumbled for a long time earlier. Together with the suspected gyro failure it's quite typical for the FC to eventually give up if the craft is airborne long enough.

The GPS system seems to delay dropping for momentary loss of signal.
The loss of sat numbers and GPS status doesn't show immediately.

 

[slup]

The GPS system seems to delay dropping for momentary loss of signal.
The loss of sat numbers and GPS status doesn't show immediately.

Without a .DAT we will never know for sure & it's not vital for the incident either, but pairing it together with likely HW problems we know that the FC sooner or later reverts to ATTI... so in that sense it fits, if it was the case here will remain unknown.

 

[Deleted member 103366]

The lake appears to be between 3m and 4m deep, boatable and fishable. The drone's last indicated position is about 18m off shore.
A fishing expedition perhaps?

 

[Wolooloo]

You guys are on top of this! The explanations you provided make sense and I think I follow most of it.

My original hypothesis of losing the propeller mid-flight was just a guess, mostly based on how it seemed "imaginable" for a propeller to come off (instead of a motor just suddenly failing!) and it superficially accounted for the spinning. Anyways, it seems like the more likely explanation goes something like this (correct me if I misinterpreted it, though):

One of the motors malfunctioned erratically, but didn't completely die and that would explain why it didn't just collapse (and in fact it was able to get higher again later). That unexpected behavior threw off the stabilization mechanisms of the drone (IMU vs GPS) and was never able to recover control. Finally, after a little while of spinning unconstrained and a bunch of overlapping warnings, it went into "recovery mode" (ATTI) and at that point it just tumbled down from the sky.

Also, yes, altitude data is unreliable, but very roughly matches what happened. When the first error occurred, the drone starting spinning and was coming down relatively slowly (not sure how fast, but nothing like free fall). I was mostly paralyzed for a few seconds, but based on the camera I think it got pretty close to the water (although with all the spinning, hard to say how close). Unfortunately, at that point it had drifted over the water... otherwise I could've let it land.

Anyways, a few seconds later I realized how dire the situation was and I thought to at least direct the drone upwards, which worked for a while. 150-200 meters of height seems very plausible to me. I really wish I would've thought of recording video at some point! After all the craziness, the controls timed out (possibly triggered by me pressing return home button out of desperation?) and quickly collapsed. I am also very disappointed about how close to land on the other side it got! So close and very shallow, but I am on vacation here in South Africa and it is impossible for me to retrieve :/

Finally, but really quickly back to the log files, I didn't realize the DAT files are unworkable on the Mini 2. Only by DJI, in which case, I should just keep the Auto-sync Flight Records" ON, right? I reported the case as a "flyaway" on DJI's Repair Center website. I'm not sure about the procedure and whether they'll get back to me with more information?

 

[slup]

...One of the motors malfunctioned erratically, but didn't completely die and that would explain why it didn't just collapse (and in fact it was able to get higher again later). That unexpected behavior threw off the stabilization mechanisms of the drone (IMU vs GPS) and was never able to recover control. Finally, after a little while of spinning unconstrained and a bunch of overlapping warnings, it went into "recovery mode" (ATTI) and at that point it just tumbled down from the sky.

It's not common that gyros fail during prop or motor failures ... even though the craft is spinning & tumbling. Something went really wrong there & the FC commanded a gyroscope re-initialization ... but failed & instead advised a craft restart.

What also separates this from a ordinary prop or motor failure is that you could gain height ... that will nearly never happen.

Adding in the other 2 pieces in the puzzle regarding the very negative barometric height & the ATTI in the end ... strongly suggests that this was something else than a loss of thrust from some of the corners.

... I reported the case as a "flyaway" on DJI's Repair Center website.

I wouldn't do that ... as you had auto sync on, DJI already have access to the .DAT log & may find further info on this & together with the clear messages from the .TXT log about HW failures ... this could be a warranty case instead.

... I should just keep the Auto-sync Flight Records" ON, right?

Doesn't matter ... you will not, without DJI's assistance, get out any info from the .DAT log, so in this sense it makes no difference. If you use Airdata.com & find it nice to get your flights up there without manual .TXT log offloading & uploading ... you keep the sync to the DJI cloud service on.

 

[Chaosrider]

I thought to keep it as high as possible, meanwhile trying to use the directional joystick in all directions to see if any would bring it back towards me.

Knowing what you know now, it might have been better to have tried to steer it to somewhere soft, rather than up.

Sorry about the loss of your bird.

TCS

 

[gogordy]

I had a mavic mini lose a prop and still come home. When it got close I could hear a strange vibration, but it still came in for a nice landing close to the take off spot.

 

[kodiak]

As the mobile device .DAT log from a Mini 2 is encrypted & can only be read by DJI staff ...

Do you happen to know why DJI did this? Are they concerned about third-party interpretations conflicting with their own?

 

[slup]

Do you happen to know why DJI did this? Are they concerned about third-party interpretations conflicting with their own?

Possibly ... but don't have any way to know, I'm not connected to DJI in any way.

 

[Meta4]

Do you happen to know why DJI did this? Are they concerned about third-party interpretations conflicting with their own?

There's no reason to think they would be concerned about third-party interpretations conflicting with their own.
It's more likely to ensure that flight data isn't tampered with and remains reliable.

 

[kodiak]

Possibly ... but don't have any way to know, I'm not connected to DJI in any way.

Understood.

There's no reason to think they would be concerned about third-party interpretations conflicting with their own.
It's more likely to ensure that flight data isn't tampered with and remains reliable.

That makes sense. Is there a list or reference anywhere that breaks down which machines have encrypted .DAT logs?

 

[sar104]

This is a very interesting case, in fact I don't recall ever seeing another quite like it.

Pilot observations:

1. The aircraft climbed normally but then transitioned to a spinning/spiral descent;
2. Aircraft recovered and climbed, but then did not respond to directional control;
3. After 90 seconds of flight, aircraft entered uncontrolled descent again and crashed into the water.

Telemetry anomalies:

1. Uncommanded spinning descents that exceeds MM2 specification envelope for yaw rate and vertical velocity;
2. Height is incorrect after initial climb;
3. 3-D inertial discrepancies.

These are worth digging deeper into since, as noted by others, it's hard to see a single failure that lead to the resulting flight characteristics.

Firstly the overall behavior:



The first and second spinning descents are clearly shown in the attitude data. Based on the pilot account the pitch, roll and yaw readings are credible. The relative height is clearly wrong, however, since it goes negative by around 25 m.

Looking at the unwrapped heading and yaw rates more clearly shows the persistent CCW spin:



The first spin has a mean rate of 280°/s, the second is 350°/s before control is fully lost and the rates exceed 1000°/s. For comparison, a lost prop or motor generally causes a spinning descent at around 1000°/s. More on that later.

At this point it's hard to distinguish hardware failure from FC issues.

The next obvious step is to try to compare the rate values from the IMU with absolute values from the compass, barometer and GNSS. Those values are not directly recorded in the log, and so the method is to compare rate fields with absolute fields by integrating the rates with respect to time. For example we can integrate vertical velocity to get relative height, and compare that with OSD_height:



This gives a clear clue as to why the height is wrong - the time integral of vertical velocity agrees perfectly with the recorded height (and also with the VPS height) in the initial climb, but thereafter the vertical velocity rapidly puts the aircraft below ground level, all the way to -340 m. Since the primary measurement of velocity comes from the IMU accelerometers, that suggests a problem with the accelerometers.

Looking in the x-y plane, and integrating the north and east velocities to get distance on those axes in the earth FOR:



That shows some disagreement, especially at the end, but it is only around 20 m or so - much less than the 350 m error in the z-axis. From that it is fairly simple to conclude that the z-axis accelerometer went bad (the errors from that alone while the aircraft tilt is non-zero are enough to give the horizontal errors) or that it was being incorrectly read by the IMU.

While that alone accounts for the height errors, it is really difficult to see how it caused the spin and descent, since erroneous descent solutions always lead to uncommanded climbs.

The spin behavior - which appears to have been real, is almost certainly a different hardware issue. It actually looks like the result of the loss of half of the front right prop at 26 seconds, which allowed unstable flight, followed by loss of the other half at 94 seconds that resulted in uncontrolled descent, but that doesn't explain the accelerometer problem. It's always rather unsatisfactory when multiple failures are required to explain an event. It will be interesting to see what DJI support concludes.

 

[Wolooloo]

I'm sorry about my tardy response; our travel schedule got really hectic our last week here in South Africa and just now got back on track! Such awesome analysis!!! I'm an engineer and I wish I got to do this type of stuff at work hahaha. It's very interesting to see how "special" my accident was haha. Also, and I didn't mention this before, but it was an officially refurbished drone by DJI. I didn't think much of it at first, but coincidence?

Anyways, I don't know much about the inner workings of drone components and their interactions, so I don't know if the IMU misbehavior that "slup" was talking about earlier would account for the [Integral of z-speed over time] difference vs [OSD_height] (and how is that measured independently? GPS?), but it is clearly out of whack. I can imagine how that one malfunction can happen, but it seems almost impossible that a propeller would come off exactly during the same flight? Or that the defect in the sensors was enough to trigger an unusually aggressive behavior that unscrewed the propeller?

One more thing that is probably unimportant (since no one commented on it), but on the original post I said in passing: "I was able to keep it high up until I thought to click return to home [the H button on the remote] when it very quickly lost all control and went for a free fall into the nearby pond...". I just want to make sure this comment wasn't missed, but it felt like as soon I as pressed that button (which seems really dumb in retrospect), it went tumbling down.

Good news to finish up, though! After the initial conversation here on the forum, I reported the incident to DJI, they got back to me within a few hours and confirmed it was a malfunction of some sort and they would send me a new one. I tried digging a little deeper for answers, but they just provided an empty bureaucratic response...

"Based on the analysis result, it seems that the incident was caused by a possible propulsion issue. However, as the aircraft was lost and we could not get it for further analysis, that was only one of our speculations. Since there was no pilot error discovered, we determined to cover this case under warranty."

Anyways, I just received a new one in the mail today!