Mavic Mini Flyaway

[ghost.its]

Hi, Pilots. It is my first post. My friend lost control as soon as his drone was in the air. There are video and logs.

I'd like to ask you for help to explain what had happend.
Thanks in advance.

 

[ATP]

Without data my first guess is that the drone was too close to the water surface.

 

[ghost.its]

When VPS is affected by flying close to the water is it normal that drone does not response to RC signal? I woluld expect rather sth like drifting or kind of atti mode. I'm just a beginner, just asking.

 

[slup]

From knowing that the AC took off from a boat and comparing your video, showing a characteristic flight path usually coming from a near 180 degree magnetic disturbed compass at AC power on & the direction the AC is pointing in the video ... with the direction the IMU thought the AC had directly from the log ... I say that this is a yaw error coming from powering on the AC in a magnetic disturbed area.

If you can provide the mobile device .DAT log also we can get access to the sensor data & confirm this ... the .DAT log for this flight ends with FLY021.DAT, read here how to retrieve it --> Mavic Flight Log Retrieval and Analysis Guide (read under section 3.)

 

[ghost.its]

Thanks for help and great feedback. We are searching for .DAT file. I will let you know about results asap.

 

[bmwbob]

That's a lot of metal around takeoff!

 

[ghost.its]

This is what we found. MavicESC - Google Drive

 

[slup]

This is what we found. MavicESC - Google Drive

So ... when the flyaway starts 4,5sec into the flight imuYaw & magYaw (compass) disagrees approx 148 degrees. With this difference the flight path makes sense ... & it's easy to think that the disagreement is due to a power on in a magnetic disturbed area, especially as it happens on a boat.

Below the disagreement at 4,5sec pure visually on a sat. picture ... the yellow bar is the heading direction the IMU think's the AC have, and the blue shows the heading from the compass.



But if we go back a second earlier ... the AC was then also airborne. The big disagreement isn't there at that point it's only healthy 8 degrees ... and your video also confirms that both the compass & IMU have the correct info about the heading at that point which is equal to reality as the flight path goes ahead of the bow according to your video.

If this had been a yaw error coming from a magnetic disturbed power on location both the yellow & the blue bar (below) had been pointing as the yellow bar in the pic. above.



Here from the video just when the AC takes off ...



So what happened ... firstly it's strange that it's two FLY021.DAT logs (possibly a shorter disconnect between the mobile device & the RC can generate two DAT's with the same ending).
The end of the first DAT is position wise 8,87 meters away from the start of the second DAT ... the distance can perhaps be explained by that the boat is moving. Any how ... this have nothing to do with the incident, just an observation.



Looking into the sensor data in the DAT logs gives us this ... first the earliest FLY021.DAT

This should be when the AC was powered on ... we see that imuYaw (red) have been initialized to the same degree value as the magYaw (green) ... the black graph shows the difference between them & it's really close to zero degrees. Adding to this I've added the gyro (blue) which value wise always start up on zero & than adds or subtracts depending on CW or CCW turn of the AC ... but shape wise it confirms all turns that the IMU & compass indicate. When this log ends we have a short gap in time until the second DAT starts ...



Here below we have the second DAT ... same graphs again but with 2 additional sensor values, the grey that shows the magmod (magnetic field) & the purple that shows another yaw measure coming from the VPS sensors on the belly of the AC ... called VIOyaw. the VIOyaw is here really short lived, just a couple seconds is recorded just about where the flyaway starts, probably due to being over water.

What we see here until 4,5sec is a pretty normal behavior, to the left in the chart we see that the disagreement between imuYaw & magYaw is healty, approx 13 degrees according to the black graph ... that continues through the take off phase (the light brown/blue fields). The magmod (magnetic field) change abit after take off as seen from the grey graph & becomes a bit more irregular but the differences is well within the normal. If this incident had been due to a magnetic disturbed compass we would have seen a much bigger change of the magmod then we see here.

Comparing the gyro (blue) with the VIOyaw (purple) & the magYaw (green, compass) both shape wise & value wise where the pointers are in the chart says us that those 3 data agrees well ... but if we look at the imuYaw (red) we see that it suddenly jumps from -25,6 degrees to -163,27 degrees ... none of the other graphs register any movement there, & nor does your video ... & this is what sets off the disagreement to dangerous 148 degrees.

At 12 sec you apply a rudder input (left stick to the right) ... and all 3, imuYaw, magYaw & Gyro record an equal change ... but the imu & compass disagreement is still stuck at 150 degrees.



If we compare the difference in GPS & IMU speed in north & easterly direction, we should see graphs with a value very near zero if everything is healthy ... but if we look how thigs develop here after 4,5sec when the incident starts. What we see here usually indicate either a yaw error coming from a power on in a magnetic disturbed spot or IMU problems.



So my conclusion of all this is that this was a IMU failure, either a computational error or a break down ... with this you should be able to make a case towards DJI as it wasn't a pilot error. The consequence was the same as for a set off compass at power on as it generated a near 150 degrees disagreement which made the flight controller believe that the AC was rotated a halv turn & by that applied thrust to the wrong motors to keep position when the AC was affected by the wind.

 

[ghost.its]

That's great news.Thanks. I really appreciate your deep and professional analysis.

 

[ghost.its]

Thanks a lot once again. There is a great new from DJI:

 

[boblui]

That's very generous of DJI.

My Mini has suffered from the same issue because I powered it up on the balcony floor which has a lot of steel reinforcement beneath the surface. As the result, the compass was off by 180 degree. Here is the video. Note the tilting of the horizon which also shows up in your video. It was due to the craft tilting so much in it's attempt to correct for shift in position that the mechanical limit of the glimbal was exceeded so it could no longer keep the camera horizontal :

 

[slup]

That's very generous of DJI.

My Mini has suffered from the same issue because I powered it up on the balcony floor which has a lot of steel reinforcement beneath. As the result, the compass was off by 180 degree. Here is the video. Note the tilting of the horizon which also shows up in your video. It was due to the craft tilting so much in it's attempt to correct for shift in position that the mechanical limit of the glimbal was exceeded so it could no longer keep the camera horizontal :

The difference here compered to your case was that this most probably was a IMU failure creating the same type of flyaway as a 180 degree disturbed compass at power up... so not a pilot error.

 

[jglowka]

So ... when the flyaway starts 4,5sec into the flight imuYaw & magYaw (compass) disagrees approx 148 degrees. With this difference the flight path makes sense ... & it's easy to think that the disagreement is due to a power on in a magnetic disturbed area, especially as it happens on a boat.

Below the disagreement at 4,5sec pure visually on a sat. picture ... the yellow bar is the heading direction the IMU think's the AC have, and the blue shows the heading from the compass.

View attachment 108489

But if we go back a second earlier ... the AC was then also airborne. The big disagreement isn't there at that point it's only healthy 8 degrees ... and your video also confirms that both the compass & IMU have the correct info about the heading at that point which is equal to reality as the flight path goes ahead of the bow according to your video.

If this had been a yaw error coming from a magnetic disturbed power on location both the yellow & the blue bar (below) had been pointing as the yellow bar in the pic. above.

View attachment 108490

Here from the video just when the AC takes off ...

View attachment 108491

So what happened ... firstly it's strange that it's two FLY021.DAT logs (possibly a shorter disconnect between the mobile device & the RC can generate two DAT's with the same ending).
The end of the first DAT is position wise 8,87 meters away from the start of the second DAT ... the distance can perhaps be explained by that the boat is moving. Any how ... this have nothing to do with the incident, just an observation.

View attachment 108492

Looking into the sensor data in the DAT logs gives us this ... first the earliest FLY021.DAT

This should be when the AC was powered on ... we see that imuYaw (red) have been initialized to the same degree value as the magYaw (green) ... the black graph shows the difference between them & it's really close to zero degrees. Adding to this I've added the gyro (blue) which value wise always start up on zero & than adds or subtracts depending on CW or CCW turn of the AC ... but shape wise it confirms all turns that the IMU & compass indicate. When this log ends we have a short gap in time until the second DAT starts ...

View attachment 108493

Here below we have the second DAT ... same graphs again but with 2 additional sensor values, the grey that shows the magmod (magnetic field) & the purple that shows another yaw measure coming from the VPS sensors on the belly of the AC ... called VIOyaw. the VIOyaw is here really short lived, just a couple seconds is recorded just about where the flyaway starts, probably due to being over water.

What we see here until 4,5sec is a pretty normal behavior, to the left in the chart we see that the disagreement between imuYaw & magYaw is healty, approx 13 degrees according to the black graph ... that continues through the take off phase (the light brown/blue fields). The magmod (magnetic field) change abit after take off as seen from the grey graph & becomes a bit more irregular but the differences is well within the normal. If this incident had been due to a magnetic disturbed compass we would have seen a much bigger change of the magmod then we see here.

Comparing the gyro (blue) with the VIOyaw (purple) & the magYaw (green, compass) both shape wise & value wise where the pointers are in the chart says us that those 3 data agrees well ... but if we look at the imuYaw (red) we see that it suddenly jumps from -25,6 degrees to -163,27 degrees ... none of the other graphs register any movement there, & nor does your video ... & this is what sets off the disagreement to dangerous 148 degrees.

At 12 sec you apply a rudder input (left stick to the right) ... and all 3, imuYaw, magYaw & Gyro record an equal change ... but the imu & compass disagreement is still stuck at 150 degrees.

View attachment 108494

If we compare the difference in GPS & IMU speed in north & easterly direction, we should see graphs with a value very near zero if everything is healthy ... but if we look how thigs develop here after 4,5sec when the incident starts. What we see here usually indicate either a yaw error coming from a power on in a magnetic disturbed spot or IMU problems.

View attachment 108495

So my conclusion of all this is that this was a IMU failure, either a computational error or a break down ... with this you should be able to make a case towards DJI as it wasn't a pilot error. The consequence was the same as for a set off compass at power on as it generated a near 150 degrees disagreement which made the flight controller believe that the AC was rotated a halv turn & by that applied thrust to the wrong motors to keep position when the AC was affected by the wind.

Very useful analysis. Thank you!
Just had an analogue issue. The only difference - the IMU failure was in the middle of longer fly. The drone finished grounded some tens of meters away with only a few propeller blades broken (pure luck).
After replacement, restart, and compass calibration it seems to be working normally - but I've tested it only on two shorter flights... Should I continue to use it or report it to DJI?

 

[slup]

Very useful analysis. Thank you!
Just had an analogue issue. The only difference - the IMU failure was in the middle of longer fly. The drone finished grounded some tens of meters away with only a few propeller blades broken (pure luck).
After replacement, restart, and compass calibration it seems to be working normally - but I've tested it only on two shorter flights... Should I continue to use it or report it to DJI?

Hi there & welcome to the forum ?

If you want to know what really happened in your flight I recommend that you start up your own thread under the section --> Mavic Crash & Flyaway Assistance. There explain what happened & attach your flight logs (the mobile device stored .TXT log but also the .DAT if you know which is the correct one).

Read up here on how to retrieve --> Mavic Flight Log Retrieval and Analysis Guide (read under section 3.)

In this way it's easier to relate all posts to only one case & make it more understandable for others that visits the thread later on.