M2E - Tilting forward under takeoff

[sar104]

@sar104 ... any ideas too what it can be, seen anything similar before, can the ESC's cause this?

Also... +20 degrees pitch at take-off, is that the normal position standing flat for a M2... my MA1 only stand at +7. Can a placement in a up-slope for take-off cause the rears rev up to much making it flip forward?... Or isn't the +20 degrees true & something is actually wrong with the IMU calibration there?

Not without more information. The IMU acceleration data agree that the aircraft was pitched before takeoff, which makes it unlikely that it's an IMU problem, unless they were both improperly calibrated previously.

 

[BudWalker]

In a normal flight the actual gyro data pretty much follows what the FC commands. In this case the gyroZ and gyroX data was normal in this regard but the gyroY data was not.

While the motors were idling before 3.775 secs the FC was commanding -3.39 gyroY but the feedback was showing gyroY to be essentially 0.0. The FC was attempting the bring the nose down from the 20° nose up attitude. But, the motors weren't providing enough thrust to change the pitch.

At 3.767 secs the motors spun up and the M2E began to move forward. The FC then began to command +4° gyroY to cancel the forward movement. But, the actual gyroY decreased until the motors were shut off and the M2E settled back down.

I can think of two possibilities.

1) If the pilot had provided more thrust as suggested by @sar104 the M2E would have been able to level itself. Starting from +20° pitch and minimal launch thrust was a bad combination.
2) This incident reminds me of
Mavic Pro behaving strangely
where flotation balls were added to the legs. This caused a pitch attitude before launch seen here. But, I'm sure that @BobbyXo would have mentioned any unusual aspects like this.

 

[sar104]

In a normal flight the actual gyro data pretty much follows what the FC commands. In this case the gyroZ and gyroX data was normal in this regard but the gyroY data was not.
View attachment 129672
While the motors were idling before 3.775 secs the FC was commanding -3.39 gyroY but the feedback was showing gyroY to be essentially 0.0. The FC was attempting the bring the nose down from the 20° nose up attitude. But, the motors weren't providing enough thrust to change the pitch.

At 3.767 secs the motors spun up and the M2E began to move forward. The FC then began to command +4° gyroY to cancel the forward movement. But, the actual gyroY decreased until the motors were shut off and the M2E settled back down.

I can think of two possibilities.

1) If the pilot had provided more thrust as suggested by @sar104 the M2E would have been able to level itself. Starting from +20° pitch and minimal launch thrust was a bad combination.
2) This incident reminds me of
Mavic Pro behaving strangely
where flotation balls were added to the legs. This caused a pitch attitude before launch seen here. But, I'm sure that @BobbyXo would have mentioned any unusual aspects like this.

The gyros just reflected the pitch change. With the motors idling the aircraft was not rotating, so all rate gyros should be zero. The subsequent negative gyro_Y values correspond to it pitching forwards. It pitched through approximately 58° until the front props hit the ground, at which point they stopped rotating which exacerbated the flip. That's a little high - I measured the pitch excursion required to bring the front props in contact with the ground to be just under 45°, but we don't know what the surface was, so it's not impossible.

 

[BobbyXo]

Hi! Thank you for helping.
This log is with manual takeoff. I can attach a file that has auto takeoff if you want to take a look.
Also, i tried holding the M2E in my hand an take off. Can send dat files for that also.
Files attached. ( I think it is the correct ones)

The other file is too large, se link instead: Automated takeoff.DAT

 

[sar104]

Hi! Thank you for helping.
This log is with manual takeoff. I can attach a file that has auto takeoff if you want to take a look.
Also, i tried holding the M2E in my hand an take off. Can send dat files for that also.
Files attached. ( I think it is the correct ones)

The other file is too large, se link instead: Automated takeoff.DAT

Any chance you could comment on the observations above - particularly on the data that showed the aircraft far from level (pitched nose up by 22°) on the ground before takeoff? It would be really helpful if we could validate the IMU accelerometer calibration. And can you confirm whether or not that was a manual takeoff? That was FLY022.DAT.

 

[BudWalker]

Hi! Thank you for helping.
This log is with manual takeoff. I can attach a file that has auto takeoff if you want to take a look.
Also, i tried holding the M2E in my hand an take off. Can send dat files for that also.
Files attached. ( I think it is the correct ones)

The other file is too large, se link instead: Automated takeoff.DAT

The "Automated takeoff.DAT" wasn't an actual flight but it provided the most interesting data. Like the other two .DATs it showed an initial pitch up of about 20°. But, at 221 secs the accelerometer data then became strange and almost certainly non-physical. In brief it looks like the FC decided to calibrate the IMU(s) by first turning off the IMU heater to cool down to 31.2° C. This was then followed by turning on the heater. When the IMU reached the operating temp of 65° C the accelerometers were then set to values consistent with the M2E resting level on a flat surface. In particular the pitch was then 0°

This also happened in the "Manual Takeoff .DAT" but long after the launch attempts were made and the M2E was at rest on the ground.

It appears to me that the FC performed an IMU calibration because it decided the IMU was not calibrated properly. If not then it's probably a HW malfunction and needs to be repaired. @BobbyXo can you try an IMU calibration? And, then power off/on the M2Eand let sit at rest for 300 secs. Then post both .DATs

Here the plots from the "Automated Takeoff .DAT".

The top 2 plots show the composite acceleration not equal to 1.0 during the IMU warm up interval. This is clearly not physical if the M2E is at rest which is shown by the mag and gyro data in plots 3 and 4. Finally, the bottom plot shows a correct pitch of 0° after the IMU warm up.

The above shows the IMU(0) data but the same is true for the IMU(1) data. This would argue for a faulty IMU calibration. That, or even though the M2 has "dual IMUs", there is only one set of accel and gyro sensors. The dual IMU records come from either dual circuitry excluding the sensors or two different fusing algorithms using the same IMU data.

 

[slup]

The "Automated takeoff.DAT" wasn't an actual flight but it provided the most interesting data. Like the other two .DATs it showed an initial pitch up of about 20°. But, at 221 secs the accelerometer data then became strange and almost certainly non-physical. In brief it looks like the FC decided to calibrate the IMU(s) by first turning off the IMU heater to cool down to 31.2° C. This was then followed by turning on the heater. When the IMU reached the operating temp of 65° C the accelerometers were then set to values consistent with the M2E resting level on a flat surface. In particular the pitch was then 0°

This also happened in the "Manual Takeoff .DAT" but long after the launch attempts were made and the M2E was at rest on the ground.

It appears to me that the FC performed an IMU calibration because it decided the IMU was not calibrated properly. If not then it's probably a HW malfunction and needs to be repaired. @BobbyXo can you try an IMU calibration? And, then power off/on the M2Eand let sit at rest for 300 secs. Then post both .DATs

Here the plots from the "Automated Takeoff .DAT".
View attachment 129702
The top 2 plots show the composite acceleration not equal to 1.0 during the IMU warm up interval. This is clearly not physical if the M2E is at rest which is shown by the mag and gyro data in plots 3 and 4. Finally, the bottom plot shows a correct pitch of 0° after the IMU warm up.

The above shows the IMU(0) data but the same is true for the IMU(1) data. This would argue for a faulty IMU calibration. That, or even though the M2 has "dual IMUs", there is only one set of accel and gyro sensors. The dual IMU records come from either dual circuitry excluding the sensors or two different fusing algorithms using the same IMU data.

Finally, a plausible explanation ... nice digging with a sharp eye @BudWalker

 

[BobbyXo]

The "Automated takeoff.DAT" wasn't an actual flight but it provided the most interesting data. Like the other two .DATs it showed an initial pitch up of about 20°. But, at 221 secs the accelerometer data then became strange and almost certainly non-physical. In brief it looks like the FC decided to calibrate the IMU(s) by first turning off the IMU heater to cool down to 31.2° C. This was then followed by turning on the heater. When the IMU reached the operating temp of 65° C the accelerometers were then set to values consistent with the M2E resting level on a flat surface. In particular the pitch was then 0°

This also happened in the "Manual Takeoff .DAT" but long after the launch attempts were made and the M2E was at rest on the ground.

It appears to me that the FC performed an IMU calibration because it decided the IMU was not calibrated properly. If not then it's probably a HW malfunction and needs to be repaired. @BobbyXo can you try an IMU calibration? And, then power off/on the M2Eand let sit at rest for 300 secs. Then post both .DATs

Here the plots from the "Automated Takeoff .DAT".
View attachment 129702
The top 2 plots show the composite acceleration not equal to 1.0 during the IMU warm up interval. This is clearly not physical if the M2E is at rest which is shown by the mag and gyro data in plots 3 and 4. Finally, the bottom plot shows a correct pitch of 0° after the IMU warm up.

The above shows the IMU(0) data but the same is true for the IMU(1) data. This would argue for a faulty IMU calibration. That, or even though the M2 has "dual IMUs", there is only one set of accel and gyro sensors. The dual IMU records come from either dual circuitry excluding the sensors or two different fusing algorithms using the same IMU data.

Hi, will try this tonight, and post the DATs. Thank you!

 

[BobbyXo]

See attached files

 

[BobbyXo]

Did you have a chance to take a look @BudWalker ?

 

[BudWalker]

Did you have a chance to take a look @BudWalker ?

Sorry, I didn't see that you had posted the .DATs. Often I don't get a notification unless I'm explicitly tagged. Anyway, I took a quick look and it appears to me that the problem didn't go away. Both before and after the calibration the M2E shows a +22° pitch. This was also the case in the subsequent two .DATs.

Unless someone can think of something else it's time to send it off for repair.

The contents of the eventLog stream aren't always that conclusive - it seems to contain debug statements useful to DJI engineers. But, I did notice this entry

311.396 : [L-GYRO_ACC][imu multi side cali(0)] acc bias:[-0.426 -0.077 -0.095]

I'm betting this means that the bias applied to the X axis accelerometer is -0.426 - more than 4 times the bias of the Y and Z axis accelerometers. A faulty X axis accelerometer would cause the pitch up situation seen in the M2E.

 

[sar104]

Sorry, I didn't see that you had posted the .DATs. Often I don't get a notification unless I'm explicitly tagged. Anyway, I took a quick look and it appears to me that the problem didn't go away. Both before and after the calibration the M2E shows a +22° pitch. This was also the case in the subsequent two .DATs.

Unless someone can think of something else it's time to send it off for repair.

The contents of the eventLog stream aren't always that conclusive - it seems to contain debug statements useful to DJI engineers. But, I did notice this entry

311.396 : [L-GYRO_ACC][imu multi side cali(0)] acc bias:[-0.426 -0.077 -0.095]

I'm betting this means that the bias applied to the X axis accelerometer is -0.426 - more than 4 times the bias of the Y and Z axis accelerometers. A faulty X axis accelerometer would cause the pitch up situation seen in the M2E.

An x bias of -0.426 seems like the wrong direction though, because the upward pitch of 22° represents an x acceleration of around +0.4. The accelerometer calibration should be self-consistent, so if you look at the total magnitude it should be 1 g in unaccelerated conditions. That looks correct:



Comparing the pitch calculated from a (probably incorrect) initial angle of 22° by IMU1 fusion, gyroY integration, and accelerometer values gives this, which shows reasonable agreement in the initial rotation:



So it certainly could be that the calibration of the accelerometers has effectively simply rotated the aircraft positively around the y axis by 22°, which presumably could happen with a bad IMU calibration, but it's very strange that both IMUs have almost exactly the same offset.

 

[BudWalker]

....

So it certainly could be that the calibration of the accelerometers has effectively simply rotated the aircraft positively around the y axis by 22°, which presumably could happen with a bad IMU calibration, but it's very strange that both IMUs have almost exactly the same offset.

Maybe there is only one set of sensors..
or
@BobbyXo can you confirm that the M2E doesn't have leg extenders or anything else that would affect the orientation? And, that the IMU calibration procedure is as per the DJI instructions?

 

[sar104]

Maybe there is only one set of sensors..
or
@BobbyXo can you confirm that the M2E doesn't leg extenders or anything else that would affect the orientation? And, that the IMU calibration procedure is as per the DJI instructions?

I'm pretty sure that each IMU has a full 6-axis MEMS package.

 

[BobbyXo]

Maybe there is only one set of sensors..
or
@BobbyXo can you confirm that the M2E doesn't have leg extenders or anything else that would affect the orientation? And, that the IMU calibration procedure is as per the DJI instructions?

No leg extenders under calibration. Followed the pictures from the app while calibrating. Can make a video to the calibration If needed. ?

 

[BudWalker]

No leg extenders under calibration. Followed the pictures from the app while calibrating. Can make a video to the calibration If needed. ?

No need. Good luck with DJI repair.

 

[BobbyXo]

No need. Good luck with DJI repair.

Hi!
Thank you for good help. Unfortunately there are no DJI repair service center in Norway