Mavic Air Fly-away

[laurens23]

An interesting test, for anyone who has a MA, might be to disturb the local magnetic field with a magnet while the aircraft is powered up and on the ground, without moving the aircraft, and then check the two "magnetometer" sets of data to see if both m0 and m1, or only m0, detected the change.

Yes with a mavic pro you can wave a magnet in front and behind and see either compass 1 or compass 2 go crazy. It would be easy to do this magnet test with an Air, unfortunately I don't have an Air myself to test it.

 

[sar104]

Yes with a mavic pro you can wave a magnet in front and behind and see either compass 1 or compass 2 go crazy. It would be easy to do this magnet test with an Air, unfortunately I don't have an Air myself to test it.

Nor do I. Volunteers?

 

[dbyantis]

Interesting thought. But as you'll see above, I didn't attack anyone - merely stated the truth. Then it's him who attacked my statement with falsehood. Doesn't matter if someone is highly skilled or even the president, if they use falsehood to make incorrect claims, they are still wrong.

It is for posters like this they provide an ignore option, antagonistic troll.

 

[sar104]

It is for posters like this they provide an ignore option, antagonistic troll.

Well while he had an unfortunate manner, he may have been correct in his statement regarding the number of magnetometers. I've been discussing the question with @BudWalker; he has a couple of aircraft DAT files from a Mavic Air. Those record data at the full sample rate of the IMUs, and he noted that in those files the two magnetometer data sets are actually identical - the resulting hypothesis being that the differences in the mobile device DAT files may be that the sample rate is lower and that the data points are not at exactly the same times, rather that there being two separate magnetometers. That is just a hypothesis but, if correct, it may mean that there is only one 3-axis magnetometer on the MA. Now I'm hunting through the DAT-defined fields for data from the elusive "vision compass".

 

[ac0j]

It is for posters like this they provide an ignore option, antagonistic troll.

Agree and disagree. Just because he is fixed in his belief, be it right or wrong, this has created a positive flow of information I am curious to see it play out.

 

[laurens23]

I was working from mobile DATs in all the recent MA cases - they contain all the necessary data.

The vision-based compass is interesting - I guess they have traded components for added computational complexity, which suggests that processing power is not a limiting factor. But the issue remains as to the nature of the m1 data and whether or not it is from a "vision compass" or a second magnetometer - why would the IMU or FC go to the trouble of disguising vision-based yaw data as 3-D magnetic field data? I'll see if I can make some enquiries directly with DJI.

If you get an answer from DJI let me know . I looked around but did not find anything.

 

[sar104]

Agree and disagree. Just because he is fixed in his belief, be it right or wrong, this has created a positive flow of information I am curious to see it play out.

Indeed. See my post above.

 

[Travisimo]

DJI has responded in at least a couple of forum threads on their site stating that the Mavic Air has dual IMU and a compass. If it had two actual compass modules, you can be sure they'd be mentioning it somewhere in the specs or in their forum replies. Now the elusive "vision compass" is a bit of an unknown, but it seems clear that it's not an actual standard compass module replicant. Maybe it's just as useful, maybe it's better than having two identical compasses, or maybe it's not. Maybe the vision compass is an improvement over the Mavic Pro's, or maybe the Pro's is still better. We can speculate about a lot of things.

I cannot vouch for the authenticity of DJI's responses or the knowledge of their employees (or forum administrators), but here are a couple of threads where it seems clear they are acknowledging there is but one actual compass:

Dual compass Mavic Air (DJI Thor)
Mavic Air - Dual IMU and Compass? (DJI Elektra)
How many IMU's and Compass has the new Mavic Air? (Same as above)

Now then, whether or not the Mavic Air has two compass modules or not is a great investigative exercise, but isn't the important question whether or not having two compass modules would have made a difference in this particular case? Did the compass actually fail, or was the compass simply wrong because of magnetic interference? If the compass failed (meaning, either a hardware failure or a software error), then having a secondary compass may have prevented the incident. However, if the compass was influenced by an external force (as evidenced by an improper compass heading), wouldn't it be safe to assume it likely would have affected both compasses in a similar (if not identical) way? I'll leave that one up to the experts because I honestly don't know.

 

[Helgegustav]

This thread blew up a bit, and led to some interesting discussions. I thought I'd jump in and let you know that after some correspondence with DJI, the case is going to be covered by their warranty.

Looks like it wasn't entirely my fault after all.

In the next few days I'll receive a shipping label to send the aircraft to DJI, for diagnostics and repair.

Thank you again sar, I provided them with some of your calculations, and it seems to have worked.

Hope I get the drone back soon so I can get up in the air again.

 

[JoostGT3]

Happy to hear that Helge!
To be honest, DJI can be a pain in the butt sometimes, but I feel like in cases like these, mostly it is being taken care of more than properly. Sure, we would probably like to receive our own drone repaired and well, but on the other hand, I also have received a refurbished unit back, and I have to say the service was excellent.
Hope you'll be back in the air soon and able to enjoy the fun of flying one of these awesome toys!

 

[Stimmy]

As a consistent lurker I just wanted to peek out of the shadows and say thanks to sar104 for some great info here. I also have a Mavic Air, so if you need any specific tests done with flight logs attached please let me know. For science of course.

 

[Simmo]

As a consistent lurker I just wanted to peek out of the shadows and say thanks to sar104 for some great info here. I also have a Mavic Air, so if you need any specific tests done with flight logs attached please let me know. For science of course.

Stimmy @sar104 is having difficulty with a FCT, so if you could conduct a quick one and upload the data....most appreciated... ()

 

[Stimmy]

Alrighty having no idea what an FCT actually was I did a quick Google search and came up with this.
Mavic Tours, FCT, Nigeria | Phone: +234 811 373 7525

You trying to send me on Mavic tours of FCT, Nigeria?

 

[Simmo]

Nah mate, I'm sorry! Twas a little jest....
https://www.google.com.au/search?q=...-ab&gfe_rd=cr&dcr=0&ei=qbTEWruVFNDN8geiza3IDA

 

[sar104]

DJI has responded in at least a couple of forum threads on their site stating that the Mavic Air has dual IMU and a compass. If it had two actual compass modules, you can be sure they'd be mentioning it somewhere in the specs or in their forum replies. Now the elusive "vision compass" is a bit of an unknown, but it seems clear that it's not an actual standard compass module replicant. Maybe it's just as useful, maybe it's better than having two identical compasses, or maybe it's not. Maybe the vision compass is an improvement over the Mavic Pro's, or maybe the Pro's is still better. We can speculate about a lot of things.

I cannot vouch for the authenticity of DJI's responses or the knowledge of their employees (or forum administrators), but here are a couple of threads where it seems clear they are acknowledging there is but one actual compass:

Dual compass Mavic Air (DJI Thor)
Mavic Air - Dual IMU and Compass? (DJI Elektra)
How many IMU's and Compass has the new Mavic Air? (Same as above)

Now then, whether or not the Mavic Air has two compass modules or not is a great investigative exercise, but isn't the important question whether or not having two compass modules would have made a difference in this particular case? Did the compass actually fail, or was the compass simply wrong because of magnetic interference? If the compass failed (meaning, either a hardware failure or a software error), then having a secondary compass may have prevented the incident. However, if the compass was influenced by an external force (as evidenced by an improper compass heading), wouldn't it be safe to assume it likely would have affected both compasses in a similar (if not identical) way? I'll leave that one up to the experts because I honestly don't know.

What's interesting about this event (at least to me) is that there is no evidence that the compass provided incorrect data.

In the DAT file records for the two IMUs, computing pitch, roll and yaw from the individual quaternions gives the orientation of the aircraft as determined by those IMUs.



The pitch and roll data from the two IMUs are in broad agreement for the first 40 s or so of the record, although they diverge somewhat after that. However, even though they are receiving identical magnetometer data, they initialize at different yaw values - IMU0 at 168° and IMU1 at 57°. And neither of those is the orientation recalled by the OP (220° = -140°). Comparison with the recorded flight yaw values shows that IMU0 was the active IMU.

However, if we take the actual recorded magnetometer data and compute yaw from that, using the pitch and roll data from the two IMUs, then we can compare the magnetic yaw from the two IMUs with the IMU yaw. Since they are using the same magnetometer data and agree, initially, on pitch and roll, those two computed magnetic yaws should be similar.



And that is what we see - the two IMU magnetic yaws agree until the pitch and roll values diverge. More notably, however, the computed magnetic yaw is completely different from both IMU0 yaw and IMU1 yaw and, in fact, agrees with the OP's recollection of the orientation and the orientation that explained the aircraft's flight path. The magnetic yaw starts at -130° (= 230°).

So the magnetometer data, even though just from one magnetometer, look to be correct, but both IMUs are ending up with wildly incorrect yaw values. That looks potentially like a problem, or possibly a bug(?), in the IMUs themselves.

 

[Prismatic]

Nor do I. Volunteers?

I would, gladly, but I'm packing today for a 12-day trip to southern UT & northern AZ. If no one does it sooner, I'll do the test on my return.

 

[RayOZ]

Just a thought on vision compass...
I wonder on start-up the two IMUs are synced to the magnetic compass, as long as the magnetic compass is healthy. I assume the software can use the cameras to determine roll and pitch if they can locate the horizon in the images. The bottom camera compare images for changes on the ground for yaw.
I suppose a quick test is to have the MA hover in front of a pic with a horizon, and then roll the pic to see if the IMU1 value changes.
But then, if you are flying between trees, the vision compass would be compromised.

 

[sar104]

Just a thought on vision compass...
I wonder on start-up the two IMUs are synced to the magnetic compass, as long as the magnetic compass is healthy. I assume the software can use the cameras to determine roll and pitch if they can locate the horizon in the images. The bottom camera compare images for changes on the ground for yaw.
I suppose a quick test is to have the MA hover in front of a pic with a horizon, and then roll the pic to see if the IMU1 value changes.
But then, if you are flying between trees, the vision compass would be compromised.

I think that it is extremely unlikely that a vision system could be used to initialize pitch and roll with any suitable level of accuracy, and that's not really needed - the accelerometers do that. In this case the compass demonstrably was healthy, and yet the IMUs were not initialized to it, at least within the span of the log file.

 

[Tinmania]

It is for posters like this they provide an ignore option, antagonistic troll.

His approach might have been caustic, and of course unwelcome in a MA forum, but regarding the MA only having one actual compass, he seems to have been correct.

So let's please refrain from shooting the messenger.



Mike

 

[BudWalker]

What's interesting about this event (at least to me) is that there is no evidence that the compass provided incorrect data.

In the DAT file records for the two IMUs, computing pitch, roll and yaw from the individual quaternions gives the orientation of the aircraft as determined by those IMUs.

View attachment 35064

The pitch and roll data from the two IMUs are in broad agreement for the first 40 s or so of the record, although they diverge somewhat after that. However, even though they are receiving identical magnetometer data, they initialize at different yaw values - IMU0 at 168° and IMU1 at 57°. And neither of those is the orientation recalled by the OP (220° = -140°). Comparison with the recorded flight yaw values shows that IMU0 was the active IMU.

However, if we take the actual recorded magnetometer data and compute yaw from that, using the pitch and roll data from the two IMUs, then we can compare the magnetic yaw from the two IMUs with the IMU yaw. Since they are using the same magnetometer data and agree, initially, on pitch and roll, those two computed magnetic yaws should be similar.

View attachment 35065

And that is what we see - the two IMU magnetic yaws agree until the pitch and roll values diverge. More notably, however, the computed magnetic yaw is completely different from both IMU0 yaw and IMU1 yaw and, in fact, agrees with the OP's recollection of the orientation and the orientation that explained the aircraft's flight path. The magnetic yaw starts at -130° (= 230°).

So the magnetometer data, even though just from one magnetometer, look to be correct, but both IMUs are ending up with wildly incorrect yaw values. That looks potentially like a problem, or possibly a bug(?), in the IMUs themselves.

I was speculating the same over on

How many times do you have to calibrate the compass?

specifically that the FC Yaw value wasn't initialized correctly. It was further speculated that it was related to the FW version. But, I don't think so because these incidents have occurred using FW dated Feb 8 2018 (@Helgegustav ) while other other flights using FW dated Feb 8 2018 didn't have a problem.

Just spit balling here. Is it possible that incorrect Yaw initialization happens when the MA is powered up while being moved about by the pilot? (Shouldn't make any difference; I do this all the time with my MP) If there is a slight delay between reading magYaw and initializing Yaw this could cause the error. OTOH, a stationary MA wouldn't have the incorrect Yaw initialization problem. The .DAT from the MA itself would show if this happened for these incidents. Unfortunately, the tablet .DAT doesn't have initial 1 or 2 secs of data required to determine this.