Help with MA flyaway/crash

[JimPD]

Been very useful being on here the last few weeks since getting my Mavic Air. Gutted that my first post is looking for help after my MA crashed.

On holiday in Scotland. 6 weeks flying experience with no issues. Set up to fly out over water to take photos of the town we were visiting.
Compass calibrated as demanded (mine nearly always does) and took off, climbed and turned towards where I wanted to go.
Only a few seconds from takeoff, the a/c suddenly became unresponsive and the a/c flew off in a curved flightpath more or less opposite the way I was heading, at maybe 5-8m height. There were some messages on screen but I had little time to read / react before it hit trees and fell into water (sea). Did press RTH button before impact but no response.
Recovered the Air and thoroughly dried out over several days. It will now power up again but there are numerous errors reported.

I am struggling to understand what went wrong other than the possibility of aircraft malfunction. I do however understand that lots of people will think the same and often it is pilot error at work.

I have raised it with DJI but hoped I could get some ideas or guidance off the experts on here before I send it off. I attach the flight records from the flight.

 

[Simmo]

Hmmm @sar104
Dunno! Your errors are here....


Your flight log here.....

It seems that you were in Atti mode, but suprisingly, from 15 seconds on, you made no attempt to fly or controll the aircraft.
It doesn't seem that RTH was pressed.
Sorry for your loss.
I cant give you a full analysis, perhaps @sar104 can decipher the log?

 

[JimPD]

Looking through the flight record on my phone, I pressed the RTH button at 16 seconds and that was the last thing I did as impact was at 18/19 seconds.
I attach a screenshot at 16 seconds where you can see the RTH icon is blue (normally black) as I press it.

The first list you posted show the mode as P-GPS until the aircraft changes to ATTI at 22s. This is after impact.

The warning stating 'IMU encounters heading exception' at 7 seconds correlates to the a/c becoming unresponsive and heading away on its own course.
It seems likely that this is the cause of the flyaway, but what caused the heading exception?

You can see the initial curved flightpath and the second, uncontrolled, flightpath

Note also the zig-zag path recorded. I dont think it looked so erratic from where I was standing.

 

[sar104]

This looks like another case of IMU error caused, most likely, by magnetic interference at the takeoff point. Comparing the actual motion of the aircraft with the motion indicated by the logged IMU velocity (integrated w.r.t. time to get position) , which should be the same, shows the following familiar situation:



As you can see, after takeoff (0, 0) the initial motion of the aircraft disagrees with what the FC thinks it should be. Typically this happens because the IMU yaw was initialized in the wrong direction due to an incorrect compass heading on the ground.

Question: do you recall exactly which direction the aircraft was facing at takeoff?

This really needs the mobile device DAT file to see the actual compass data.

How to retrieve a V3.DAT from the tablet

I'm also concerned by your statement that the compass asks to be calibrated nearly every flight. That's not normal and suggests that something is wrong, although not necessarily related to this incident.

 

[JimPD]

Thanks for the response

The MA was facing approx ENE at takeoff and I turned to face approx SSE by the time the problem occurred.
I will have a go at the DAT file & upload shortly if I can

 

[JimPD]

I have the DAT file but the site won't allow upload due to extension type. Ideas?

 

[sar104]

Thanks for the response

The MA was facing approx ENE at takeoff and I turned to face approx SSE by the time the problem occurred.
I will have a go at the DAT file & upload shortly if I can

Right - unfortunately the log indicates that the yaw was WNW at takeoff - hence the track discrepancy.



To post the DAT file just change the extension from .DAT to .txt.

 

[JimPD]

Here it is

 

[Irish-apple]

Sorry pal but @sar104 is the dude when these things come around .. Some say hes not even human !! Fact

 

[Dave A]

Sorry for the loss of your MA. Interested to see what more sar104 has to say.

 

[sar104]

Sorry pal but @sar104 is the dude when these things come around .. Some say hes not even human !! Fact

Thanks, but don't forget that @BudWalker was figuring this stuff out well before me, and he wrote the DAT file conversion software that we all use.

 

[sar104]

Here it is

OK - that tells a tale. The initial data in the file, around one minute before takeoff, show that the IMU yaws were aligned with the compass magnetic yaw. That was probably an incorrect heading, and was before you attempted to calibrate the compass:



The compass calibration should not have been accepted, because it didn't work. IMU1, in particular, was hopelessly wrong:



If you had restarted the aircraft after the calibration then the yaw might have reinitialized correctly, but you didn't, instead going straight to takeoff:



Neither of the IMUs was even close to correct at this point. The compass (there is only one - the two traces are the interpretation of the compass data by the two IMUs) was indicating a takeoff heading of -80° - nearly due west, so if it was actually pointing ENE then there was significant magnetic interference. IMU0 had the heading as -20°, and IMU1, which was the active IMU for the flight, had the -57° that I mentioned previously.

The end result is that the FC is completely confused - when forward pitch is applied the GPS data tells the FC that it is not moving in the expected direction (forwards with respect to the calculated heading). That causes the FC to start applying in appropriate corrective attitude that just makes things worse, and leads to uncontrolled flight.

The two IMUs have mostly good agreement on pitch and roll, which are the two parameters that have to be correct to convert the 3-axis magnetometer data to a magnetic heading, so I don't think that this is an IMU problem. It's most likely magnetic interference at the takeoff point or, since the MA only has one compass, it could be a compass problem. The data during the calibration do not look correct to me - did you follow the instructions precisely, and were you well away from any potential sources of magnetic interference, such as vehicles, railings etc.?

@BudWalker - there appears to be a tick frequency error or similar - the timebases in the txt log and the DAT file conversion are slightly different.

 

[JimPD]

There was a metal structure nearby but not for the first time. For example, I have set up & taken off within a few feet of my car several times.
Like I said in the OP, compass calibration is required just about every time I set the a/c up for a flight. Even happens most times if I land, change battery and switch on for another flight from the same point. I have read in forums (maybe this one) that some MAs are like this and some aren't. This is probably my biggest frustration with the thing, although I think its a brilliant bit of kit.

Did the calibration as usual. Seemed to work OK as far as I could tell so went for the take-off. As far as I am aware, you can't take off if the compass is requiring calibration (I may be wrong).

From memory, another guy you helped on here with a similar problem had his MA repaired / replaced under DJI warranty rather than his cost.

 

[UAVNV]

Only a few seconds from takeoff, the a/c suddenly became unresponsive and the a/c flew off in a curved flightpath more or less opposite the way I was heading, at maybe 5-8m height. There were some messages on screen but I had little time to read / react before it hit trees and fell into water (sea). Did press RTH button

@sar from an analytics standpoint when the IMU fails, what would be the best course of action in this scenario?

I have in my SERE footnotes:

press pause button, check magnetic environment, check yaw, check drift, test yaw opposite, test drift opposite, decend, check crash environment, decend, emergency shut down.

 

[Tripner]

Relative to this thread, I have a question about avoiding the problem that JimPD has faced. When preparing to take off one can check the small compass of the DJI Go4 app that appears in the lower left of ones phone or tablet. North shows up as “N” and an arrow displays the orientation of the Mavic relative to N. You can then rotate the AC to test if the arrow is correct and you can also make sure if “N” matches with your own understanding of which direction is correct for north. Additionally one would want to have a good launch point such as grass instead of concrete with hidden re-bar in it.

So…is there anything else one can do to be absolutely sure your bird won’t take off with seriously screwed up IMU and compass calibrations. Last year I lost a Mavic Pro to a combination of IMU failure and operator panic and I would like to hold onto my M-Air for as long as possible.

Thanks guys

 

[sar104]

There was a metal structure nearby but not for the first time. For example, I have set up & taken off within a few feet of my car several times.
Like I said in the OP, compass calibration is required just about every time I set the a/c up for a flight. Even happens most times if I land, change battery and switch on for another flight from the same point. I have read in forums (maybe this one) that some MAs are like this and some aren't. This is probably my biggest frustration with the thing, although I think its a brilliant bit of kit.

Did the calibration as usual. Seemed to work OK as far as I could tell so went for the take-off. As far as I am aware, you can't take off if the compass is requiring calibration (I may be wrong).

From memory, another guy you helped on here with a similar problem had his MA repaired / replaced under DJI warranty rather than his cost.

Obviously it's not possible to tell from the log what caused the magnetic interference, if that is what it was. I'd suggest some follow up tests to try to narrow it down, but I doubt the aircraft is in any condition to do that.

I've also seen reports about frequent requests to calibrate, but that's definitely not normal behavior. I would return an aircraft for replacement if it did that. For comparison, I've never calibrated my original Mavic Pro, which has flown in both the US and Europe, and it has never requested calibration.

Given that you have the aircraft and the logs showing that behavior, I would guess that there is a reasonable chance that DJI will replace it.

 

[sar104]

@sar from an analytics standpoint when the IMU fails, what would be the best course of action in this scenario?

I have in my SERE footnotes:

press pause button, check magnetic environment, check yaw, check drift, test yaw opposite, test drift opposite, decend, check crash environment, decend, emergency shut down.

The essential test, before takeoff, is to ensure that the aircraft heading arrow in the DJI GO 4 app is pointing in the same direction that the aircraft, on the ground, is actually facing. In this case there will have been an obvious and large discrepancy. Never take off under those conditions - it will lead to this kind of uncontrolled flight. It's not IMU failure per se - it's sensor fusion failure. And, on an aircraft with no manual ATTI switch, there is absolutely no way to recover the situation.

 

[JimPD]

Many thanks @sar104 for your help on this. Absolutely invaluable.
I will send the MA off & wait for DJIs response. Fingers crossed!

Two questions:
The MA controller has customisable buttons on it - is there any way of making one of these a manual ATTI mode switch?
Surely, if the IMU detects a conflict like this, there should be a better reaction than to do this. Wouldn't it be better to trust its GPS signal and go into some kind of 'limp mode' where the pilot can manually bring it back in or go into RTH mode based just on the GPS location?

Thanks again

 

[ryanolivas]

Been very useful being on here the last few weeks since getting my Mavic Air. Gutted that my first post is looking for help after my MA crashed.

On holiday in Scotland. 6 weeks flying experience with no issues. Set up to fly out over water to take photos of the town we were visiting.
Compass calibrated as demanded (mine nearly always does) and took off, climbed and turned towards where I wanted to go.
Only a few seconds from takeoff, the a/c suddenly became unresponsive and the a/c flew off in a curved flightpath more or less opposite the way I was heading, at maybe 5-8m height. There were some messages on screen but I had little time to read / react before it hit trees and fell into water (sea). Did press RTH button before impact but no response.
Recovered the Air and thoroughly dried out over several days. It will now power up again but there are numerous errors reported.

I am struggling to understand what went wrong other than the possibility of aircraft malfunction. I do however understand that lots of people will think the same and often it is pilot error at work.

I have raised it with DJI but hoped I could get some ideas or guidance off the experts on here before I send it off. I attach the flight records from the flight.

You are so lucky that you have retrieved your drone after it crashed to the sea. All electronic components got wet and still the GPS of the drone is functioning?

 

[JimPD]

I dont think GPS is still functioning.
It still powers up, links to controller and receives a picture from the camera but there are faults on IMU, accelerometer, 2 of the ESCs, compass, gimbal. Can't recall GPS status.

JIm