Link to Flight Logs and supporting videos    https://1drv.ms/u/s!ArAx9oM6899Rgec7dYUMu4fz3YT5-g?e=M6UvJy

As others here know, I was not convinced that magnetic interference is a trigger for the GENUINE flyaways we hear about on the forum. It is my personal opinion that it is a hardware or software glitch. It is possible to detect these glitches with something bench-bound like a computer or similar. You can’t do this with a flying drone. Therefore, a glitch isn’t provable.
  What is observable is the result of a drone (Mavic Pro) that has been deliberately flown from an unsafe surface. The logic being that the magnetic interference will cause a measurable result in the log file. It should cause the drone to malfunction in some way that can sensibly be attributed to the interference. If no error shows up, even under extreme conditions, then it casts doubt on Magnetic interference being the predictable cause of a flyaway.
  I took certain precautions to limit the risk to people or buildings, should the flyaway happen. The reported behaviour of an uncontrollable drone is that it flies off at a constant altitude. I picked a spot where a very high hill screens the populous so the drone would crash into it or fly off over the sea.
  I considered the loss of the drone carefully. Although I was not convinced Mag Inf was a drone killer, I accepted that it was a serious risk and goodbye Mavic was acceptable.
  The manual and common sense warn drone users NOT to fly if the drone warns of magnetic interference.
  So, don’t do it!
  I tried using a normal compass to detect magnetic force, but it was insensitive. Instead, I used the built-in App on a Samsung S9+.
  A WW2 reinforced concrete bunker is akin to a large car park or flat roof building. There is a strong magnetic field produced by the reinforcement and solid steel beams but if it has been built for at least a year or so, the steel loses its random field and takes on the earth’s alignment within about 15 degrees max. Ships are known for this also. A modern car, like mine, can have different polarity in the width of a door panel and be random. The deviation there is as much as 180 degrees. The same for the roof and bonnet or hood and be very strong.
  I decided to use the readings from the drone to determine the most powerful magnetic field. It is the drone that is affected, so there can’t be any disagreement there. I also looked for a yellow warning triangle on the GO4 App screen, but there wasn’t one. I placed the drone pointing (magnetic) east on the car bonnet because the north pole on the steel was (Magnetic) west. It was suggested by a wise member here, that if the magnetic field at the TO point coincided with magnetic north, the drone would warn of interference, but still be properly aligned. I thought this was sound reasoning.
  During the test flight, I landed once facing magnetic north on a bunker with a 12 deviation from the north, then took off again. It will be interesting to see the flight log analysis at this point.
  Readers should also know that I use old firmware that predates the NFZ system. It is a modified version from No Limit Drones, the only differences are user-adjustable parameters, beyond DJIs settings. I use all those DJI standard settings. I decided to use the old firmware because Mavic Pro s hit a bad patch with broken firmware. It was the only time I had a near flyaway and at that time, there was no way to revert to earlier firmware. Current firmware is 1.03.07 and has been stable in all respects for 18 months flying.
  I used a fixed video camera to record the flights, as much as it could. There will always be someone who thinks it’s a scam. The drone was recording as well. I confess to a mistake. I accidentally set focus to close up, so everything is blurred as well as low resolution. Sorry.
  Having taken off from the car bonnet and not wishing to lose the drone, I ran through simple checks to verify the drone was stable at this point. I was aware that flyaways usually happen later. I circled, increasing the radius and also flew very low over the bunkers to upset the compass system. I flew up (

altitude)

to 121 metres and generally in line of sight. All this I do normally anyway when I take photos and video.
  We now await comment from those reviewing the flight logs. I am most grateful to them for their valuable time.
  

It will be interesting to see the flight log analysis at this point.
How about posting an abbreviated post that just addresses whatever the point is?
I've read your essay twice and can't see what you are trying to say.
Did you see any effect that should be analysed?
You've also got 4 flight logs in there ... which one is relevant?

We now await comment from those reviewing the flight logs. I am most grateful to them for their valuable time.
You could save some of that valuable time by just getting to the point.
Is what a drone killer?

Neither I will go in and dig around all your linked files ... it's considered kind of good manner to pick out those ones with a specific questionmark that you want others to look at.

Stopped reading your long post after you write:

"-I was not convinced that magnetic interference is a trigger for the GENUINE flyaways we hear about on the forum. It is my personal opinion that it is a hardware or software glitch"

Will show you two different DAT logs, one of yours & one other from a Mini FlyAway due to a Yaw-magYaw disagreement...

The facts first ...

1. When the AC is powered on the magYaw angle will set the IMUYaw angle the same.
2. If these 2 angles disagree the flight controller don't have a correct information of the AC heading.
3. If the AC heading information is wrong the FC will because of that apply counter actions in the wrong direction (due to wind influence for instance).
4. Applying counter actions in the wrong direction will lead to that the FC will be notified that the desired position hold didn't succeeded ... and again apply counter actions ...
5. The AC will again fail to reach the desired position ... instead actually increase the position error ... again leading to even more counter actions but only makes the error larger for every try.
6. All this leads to a rolling snowball effect & the AC going into the famous "Toilet Bowl flight path"

This is from one of your flights ...

Grey = Magnetic field modulus
Green = IMUYaw angle (-180/+180 degree with 0=north)
Blue = magYaw angle, compass (-180/+180 degree with 0=north)

As seen both green & blue are at the same angle at power on in the beginning of the log & the magmod is stable. Then at the green/blue field in the chart the motor start & auto take off happens. The magmod is a bit disturbed due to motor amp draw but no major spikes. The Blue & Green graphs agrees then fine & the flight goes without any problems.



Then below ... similar chart with the same graphs from a flight with a Mini launched from a metal reinforced concrete balcony floor 70m up in a building in Hong Kong.

As seen here ... again a full agreement between green & blue having the same angle at power on. The motor start & auto take off again happens at the green/blue field & the Mini then quickly are steered away from the building. Here is seen a big change in the magmod (Grey), indicating a sudden change in the magnetic field & with that the magYaw, compass (Blue) makes a big change ... but leaving the IMUYaw at the same angle as before ... the magYaw & IMUYaw now disagrees with 172 degrees.

Just moments later in this flight the Mini are effected by winds & put in counter actions to hold position ... and goes into that "Toilet Bowl" flight path with an ever increasing speed, totally uncontrollable, smashing into the building wall & falls 70m down.



To avoid this the pilot have a very easy control procedure that always should be included in the pre-flight checks.

Once the AC is powered on & place at it's take off spot ... bring up the map view in your app & check so that the AC icon there is pointing equal to reality considering the map details like roads, buildings. If the icon isn't pointing correctly, power down the AC, move to another spot well away from the first & power on & repeat the procedure ...

So far DJI only have one model that have a counter measure to this magYaw/IMUYaw disagreement due to a magnetic disturbed place during power on ... the M2 serie. Both the M2P & the M2Z can correct the IMUYaw angle to the magYaw when a disagreement is recorded ... but ONLY if the AC isn't moved after power on. In this way an event as described can be avoided.

There were two flights. The first had a landing and takeoff without motor shutdown. It is this first complete flight that needs to be analysed.
I have no experience in reading the graphical output, so I leave that to someone who is interested in such things. I wasn't sure which files you use for analysis, because there are several folders with possible logs in, so I just sent the logs for both flights.
I included as much info as possible in the link for those who are interested to pick what they wanted. Far too often insufficient info is provided requiring to OP to keep adding. Then the thread gets messed up with guesses and opinion...
All I have tried to do is provide evidence by test for someone with an open mind to look into.
I am not forcing my opinion on anyone, am I?
No one has to do anything, do they?
I fully understood your explanation about starting and field alignment, That has always made sense.
Have you read the right logfile?
Thank you for that.

AlansDronePics Posted at 3-17 04:55
There were two flights. The first had a landing and takeoff without motor shutdown. It is this first complete flight that needs to be analysed.
I have no experience in reading the graphical output, so I leave that to someone who is interested in such things. I wasn't sure which files you use for analysis, because there are several folders with possible logs in, so I just sent the logs for both flights.
I included as much info as possible in the link for those who are interested to pick what they wanted. Far too often insufficient info is provided requiring to OP to keep adding. Then the thread gets messed up with guesses and opinion...

Then the thread gets messed up with guesses and opinion...
So far I'm having to guess what this thread is about.

All I have tried to do is provide evidence by test for someone with an open mind to look into.
I am not forcing my opinion on anyone, am I?
No one has to do anything, do they?
Have you read the right logfile?

Is what a drone killer?
What did you observe in that flight that I should look for in the data?
It would help if you used less words but said more instead of wandering around in circles without being specific.

AlansDronePics Posted at 3-17 04:55
There were two flights. The first had a landing and takeoff without motor shutdown. It is this first complete flight that needs to be analysed.
I have no experience in reading the graphical output, so I leave that to someone who is interested in such things. I wasn't sure which files you use for analysis, because there are several folders with possible logs in, so I just sent the logs for both flights.
I included as much info as possible in the link for those who are interested to pick what they wanted. Far too often insufficient info is provided requiring to OP to keep adding. Then the thread gets messed up with guesses and opinion...

I just took one of the DAT logs randomly ...

My advice is to provide both the mobile device TXT & DAT log from the flight were you have question marks. Then describe the flight & circumstances short and concise.

Want to get some understanding of flight analyzis & different log types read here --> Flight log retrieval and analysis guide

?? Summary: In take-off and flight conditions which should cause compass problems, you made several flights, and didn't loose your drone.  ??

slup Posted at 3-17 04:23
Neither I will go in and dig around all your linked files ... it's considered kind of good manner to pick out those ones with a specific questionmark that you want others to look at.

Stopped reading your long post after you write:

Here's the relevant flight:
https://www.phantomhelp.com/LogViewer/HOP5DIXQC0IUC7HHXX69/

If it's the compass error message (below from Airdata) in the flight log that's the question mark, it's easily explained , and I can add that in this case that wouldn't risk a AC flyaway ... & will not in anyway prove that magnetic interference is a HW or SW glitch.



As I said earlier it's the power on location that matter most ... in this flight the AC was powered on well outside of the magnetic interference from that wreck shown on the map ... & then launched in an other place. This is clearly shown from the DAT log & the difference in GPS coordinates at power on & auto take off.

Here plotted in GE ...



Note that magYaw & IMUYaw agrees at power on ... but then later at take off the magYaw (compass) are thrown away from IMUYaw by 55 degrees shortly and by that triggers the messages in the log. But soon there after the compass is out of the interference & everything agrees again & the flight can continue without hiccups.

Labroides Posted at 3-17 05:12
Here's the relevant flight:
https://www.phantomhelp.com/LogViewer/HOP5DIXQC0IUC7HHXX69/

Thank you, without the graphics, it doesn't tell me anything.
However, a later post has answered that with graphics.

slup Posted at 3-17 06:25
If it's the compass error message (below from Airdata) in the flight log that's the question mark, it's easily explained , and I can add that in this case that wouldn't risk a AC flyaway ... & will not in anyway prove that magnetic interference is a HW or SW glitch.

[view_image]

Point taken. Thank you for explaining.
It does raise the question though why some drones (the pilots have claimed) have become unstable flying near steel bridges.
It seems steel soon reverts to the earth's magnetic field.
Do you have an opinion about that?

slup Posted at 3-17 04:23
Neither I will go in and dig around all your linked files ... it's considered kind of good manner to pick out those ones with a specific questionmark that you want others to look at.

Stopped reading your long post after you write:

Hi Slup,

Interesting to read your explanation!
Your text in red, yes always check your compass before take-off, i do always.

Question ; In the DAT file graph i see the 2 lines (blue+green) are in a full agreement at start (before time zero and at takeoff)
If this is the case, can i than see a compass showing different heading in map view than the actual heading?

If yes, than the compass will show actual heading after take-off?  (out of interference ect), is it possible to return to the actual heading after many minutes flying in your opinion?

cheers
JJB

cheers
JJB

If you take a compass, in this case one in a drone ... put it near something magnetic, then the compass will be off, right ... Power in this stage on the drone, with a wrong pointing compass & the IMUYaw will lock on to that wrong direction. All look good agreement wise between IMUYaw & magYaw (Green/Blue) ... but if you here look at the drone icon on the map the wrong direction will be reflected there.

If you now take off ... the drone sooner then later will come away from what's disturbing the compass (magYaw) & it will start to show the correct direction & by that start to disagree to the angle the IMUYaw show. The flight controller now have a wrong sense of heading ... let's say a full half turn (180 degrees), what actually is west the AC think's is east. All good until the FC needs to do a position correction due to for instance wind.

The FC think's it's heading is north (south in reality) & sense that it's moving away from the position in a direction towards east (goes west in reality) ... to correct & maintain position FC put increased thrust on the front/rear left props to to regain the position located in a westerly direction ... but due to the 180 degree error what's actually have happened is that the new, corrected position now is even more to the east ... so the FC again perform the same maneuver ending up even more off the correct position ... & the snow ball are now rolling. Of course this happens quickly & the speed sky rockets & all pilot inputs just further mess thing's up for the FC. Most events like this ends up in a crash pretty quick but if thing's are cleared the AC eventually will go into ATTI mode when the FC just plain gives up ... but as said few drones will last that long, and if they do the pilots in general isn't trained to handle ATTI mode.

Of course this disagreement with two different heading angles fed to the FC can occur even though the start spot was healthy ... when you during flight comes to close to magnetic objects. Smaller thing's will be quickly cleared by the increase of distance to them & all goes back to normal. But if the magnetic object is large enough like a steel bridge, a large steel ship or flying into/under a steel roof with metal beams held up by steel pillars then you most likely doesn't get away enough quick in order to regain control & avoid a crash into something.

Lastly ... if you have got a IMUYaw/magYaw disagreement at height due to a wrongly initiated IMUYaw from a magnetic interfered take off spot that disagreement will not go away ... with one exception, that is if you have a Mavic 2 Pro or Zoom AND didn't move it once it was powered up(or actually the IMUYaw was initiated). Those 2 drones are so far the only ones that can correct the IMUYaw to the magYaw once a disagreement is recorded.   

I’m not convinced either.
It’s wrong to suggest that because your test showed something different than normally seen in drones that suffered magnetic interference on the ground.

To save much time just put your drone on a steel manhole start up and wait for interference to show, I’m also convinced that it’s so much easier to say something doesn’t happen when you haven’t a clue what does .
It’s very confusing .

hallmark007 Posted at 3-17 10:03
I’m not convinced either.
It’s wrong to suggest that because your test showed something different than normally seen in drones that suffered magnetic interference on the ground.

The test itself from the OP wasn't useful at all to prove or even indicate that magnetic interference & the effect on drones just is some odd HW & SW glitch ...

First the OP don't have full knowledge over how drone FC work & react in relation to magnetic disturbance ... then due to this, powers on the AC in a healthy environment & after a correct IMUYaw initialization moving in the AC into magnetic disturbance with a following take off. All this is shown in the DAT log. Up & away the drone have an equal good & consistent heading as when it was powered on.

A proper test would be to put the AC in a magnetic interfered spot, power it on ... then check off on the drone icon on the app map that it's for certain are pointing wrong compered to reality ... then take off.

Even without this test I can predict the outcome ...

slup Posted at 3-17 10:19
The test itself from the OP wasn't useful at all to prove or even indicate that magnetic interference & the effect on drones just is some odd HW & SW glitch ...

First the OP don't have full knowledge over how drone FC work & react in relation to magnetic disturbance ... then due to this, powers on the AC in a healthy environment & after a correct IMUYaw initialization moving in the AC into magnetic disturbance with a following take off. All this is shown in the DAT log. Up & away the drone have an equal good & consistent heading as when it was powered on.

I’m in full agreement with that, I’m not sure why anyone finds the need to try show by testing something doesn’t happen, a much more progressive test would be to show what is happening, but most of us already know this,  as it’s pretty logical .

hallmark007 Posted at 3-17 10:24
I’m in full agreement with that, I’m not sure why anyone finds the need to try show by testing something doesn’t happen, a much more progressive test would be to show what is happening, but most of us already know this,  as it’s pretty logical .

Mmm ... but I can understand/respect what's logical for those that have knowledge perhaps isn't for those that doesn't have  ... yet, (one way for the OP to gain knowledge is to learn reading logs for instance).

But it's lethal (for the drone) to just think that you know ... spread a wrong message to other's & in that way cause unnecessary crashes.  

slup Posted at 3-17 09:49
If you take a compass, in this case one in a drone ... put it near something magnetic, then the compass will be off, right ... Power in this stage on the drone, with a wrong pointing compass & the IMUYaw will lock on to that wrong direction. All look good agreement wise between IMUYaw & magYaw (Green/Blue) ... but if you here look at the drone icon on the map the wrong direction will be reflected there.

If you now take off ... the drone sooner then later will come away from what's disturbing the compass (magYaw) & it will start to show the correct direction & by that start to disagree to the angle the IMUYaw show. The flight controller now have a wrong sense of heading ... let's say a full half turn (180 degrees), what actually is west the AC think's is east. All good until the FC needs to do a position correction due to for instance wind.

Thanks for your answer Slup.

I fully understand it, if start flying with a compass not showing the actual heading ; yes, fly away in a TBE move.
Seen few 'great; examples on ths forum.
But see also flightrecords where compass out of the blue changes heading!, not direct afer start but later in flight. Not because of flying close to iron etc  but flying high up in the sky.
For me this is a true puzzle, the answer is often ; your drone picked up some interference at your takeoff point.
If interference is not only seen as a compass offset to the actual heading; wich part is than interfered??

cheers
JJB

JJB* Posted at 3-17 11:57
Thanks for your answer Slup.

I fully understand it, if start flying with a compass not showing the actual heading ; yes, fly away in a TBE move.

Well, all man made thing's are for sure not fail proof ... once one value unexpectedly goes nuts the only thing you have to judge what's wrong is to look at other components giving the same or equivalent values/data.

When it comes to Yaw errors you have several components giving similar information, IMUYaw (initialized by) magYaw (the compass), then VIO Yaw (a form of optical reading registering yaw change) and then the rate gyros  ... if 3 agrees & only one disagrees you most likely have nailed the culprit. Then of course most AC have more then one of these HW from where you also can get comparisons.

To look at these values the mobile DAT log or the AC DAT log (from those AC's were they aren't encrypted) are a must, the TXT log doesn't contain neither raw sensor data or motor data ...

slup Posted at 3-17 12:33
Well, all man made thing's are for sure not fail proof ... once one value unexpectedly goes nuts the only thing you have to judge what's wrong is to look at other components giving the same or equivalent values/data.

When it comes to Yaw errors you have several components giving similar information, IMUYaw (initialized by) magYaw (the compass), then VIO Yaw (a form of optical reading registering yaw change) and then the rate gyros  ... if 3 agrees & only one disagrees you most likely have nailed the culprit. Then of course most AC have more then one of these HW from where you also can get comparisons.

In simple words ; possible interference picked up at start position is an easy compass check.
If compass is showing the correct heading before/at takeoff, and in flight after some period of time a compass/yaw/imu error gets there; not likely from some interference at start point wich was asleep for many minutes....  ;-)

cheers
JJB

JJB* Posted at 3-17 12:56
In simple words ; possible interference picked up at start position is an easy compass check.
If compass is showing the correct heading before/at takeoff, and in flight after some period of time a compass/yaw/imu error gets there; not likely from some interference at start point wich was asleep for many minutes....  ;-)

"...not likely from some interference at start point..."

Agree ... that wouldn't be my first thought.

hallmark007 Posted at 3-17 10:03
I’m not convinced either.
It’s wrong to suggest that because your test showed something different than normally seen in drones that suffered magnetic interference on the ground.

If a theory can't be reproducible under test, at the least, it shows the theory is incomplete or unreliable.
In my test, I accept that in the heat of the moment, I forgot to switch off the drone and restart. The test was flawed. I did try and find the strongest piece of magnetised metal in the area. There are no manhole covers nearby, and those in town I tested had very little magnetism in them. My car was highly magnetic with every panel showing the north in a different direction.
I still think it is worth another try.
If it proves the divergent mag fields do cause the loss of control or a flyaway, then it supports that conclusion.
I would have thought that would please the people who believe magnetism to be the cause.

slup Posted at 3-17 10:19
The test itself from the OP wasn't useful at all to prove or even indicate that magnetic interference & the effect on drones just is some odd HW & SW glitch ...

First the OP don't have full knowledge over how drone FC work & react in relation to magnetic disturbance ... then due to this, powers on the AC in a healthy environment & after a correct IMUYaw initialization moving in the AC into magnetic disturbance with a following take off. All this is shown in the DAT log. Up & away the drone have an equal good & consistent heading as when it was powered on.

I accept the error of starting in a clean spot. I was anxious that this was goodbye Mavic.
However, thanks to your excellent explanation, I am prepared to try again, starting with the drone sensing north but is actually south at switch on. The car roof edge is the strongest magnetic field, so I hope to try again.
Obviously, this pole reversal field is far worse than a user will find on the ground. It is the north/south edges of a metal area that have the strongest influence, so the centre of a roof or car park will be far weaker, if not already aligned with the earth.
Remember, this should prove your theory is right, so I wonder why the resistance to finding the answer. Your help so far has been gratefully received.

JJB* Posted at 3-17 12:56
In simple words ; possible interference picked up at start position is an easy compass check.
If compass is showing the correct heading before/at takeoff, and in flight after some period of time a compass/yaw/imu error gets there; not likely from some interference at start point wich was asleep for many minutes....  ;-)

That explanation is a bit convoluted isn't it  JJB.

slup Posted at 3-17 12:33
Well, all man made thing's are for sure not fail proof ... once one value unexpectedly goes nuts the only thing you have to judge what's wrong is to look at other components giving the same or equivalent values/data.

When it comes to Yaw errors you have several components giving similar information, IMUYaw (initialized by) magYaw (the compass), then VIO Yaw (a form of optical reading registering yaw change) and then the rate gyros  ... if 3 agrees & only one disagrees you most likely have nailed the culprit. Then of course most AC have more then one of these HW from where you also can get comparisons.

I understood the Mavic Pro AC log is readable, being old and in my case, with old firmware.
Hopefully this time I can provide the right files.
Do I download from the AC using DJI assistant?
Thanks

slup Posted at 3-17 10:34
Mmm ... but I can understand/respect what's logical for those that have knowledge perhaps isn't for those that doesn't have  ... yet, (one way for the OP to gain knowledge is to learn reading logs for instance).

But it's lethal (for the drone) to just think that you know ... spread a wrong message to other's & in that way cause unnecessary crashes.

That isn't fair!
I have made it quite clear that my test is against all the rules, but most experiments probe beyond known boundaries.
Where have I said that anyone should disregard magnetic interference?
There was a time when MPs went through a period where so-called flyaways happened frequently. Then after several firmware revisions, those issues stopped. There are a remarkable number of Mavic Minis with failures at the moment.
It isn't unreasonable to suspect a cause and test for it.
Come on, don't say things that haven't been said..

slup Posted at 3-17 09:49
If you take a compass, in this case one in a drone ... put it near something magnetic, then the compass will be off, right ... Power in this stage on the drone, with a wrong pointing compass & the IMUYaw will lock on to that wrong direction. All look good agreement wise between IMUYaw & magYaw (Green/Blue) ... but if you here look at the drone icon on the map the wrong direction will be reflected there.

If you now take off ... the drone sooner then later will come away from what's disturbing the compass (magYaw) & it will start to show the correct direction & by that start to disagree to the angle the IMUYaw show. The flight controller now have a wrong sense of heading ... let's say a full half turn (180 degrees), what actually is west the AC think's is east. All good until the FC needs to do a position correction due to for instance wind.

That does make sense. I am happy with that.
Thanks.

AlansDronePics Posted at 3-18 01:32
If a theory can't be reproducible under test, at the least, it shows the theory is incomplete or unreliable.
In my test, I accept that in the heat of the moment, I forgot to switch off the drone and restart. The test was flawed. I did try and find the strongest piece of magnetised metal in the area. There are no manhole covers nearby, and those in town I tested had very little magnetism in them. My car was highly magnetic with every panel showing the north in a different direction.
I still think it is worth another try.

I think they’re should always be a reason to believe something is wrong, and it should be backed up with both reasons and also reasonable explanation as to what could be actually causing the problem.

If we try hard enough to prove something wrong because we believe it to be, we will always find something that convinces us we’re right, but it’s always much more convincing if we have another argument that shows what we believe to be the true problem, “and not just bug problems”

AlansDronePics Posted at 3-18 01:50
I understood the Mavic Pro AC log is readable, being old and in my case, with old firmware.
Hopefully this time I can provide the right files.
Do I download from the AC using DJI assistant?

I understood the Mavic Pro AC log is readable...

The mobile device DAT log is perfectly sufficient to get hold of the data regarding Yaw errors ...