DJI Mavic, Air and Mini Drones
Friendly, Helpful & Knowledgeable Community
Join Us Now

Mavic 2 Pro out of control, crashed into lake... help?

It has nothing to do with correcting magnetic to true heading, and rotating the aircraft gives it no information on declination anyway. It converts magnetic to true by adding the declination that it computes from a global magnetic model in the aircraft firmware.

Im, sorry but this is BS. There is no such thing as a global magnetic model in the aircraft firmware.

Rotating the model allow it to get an accurate calibration to magnetic north with respect to the aircraft orientation and the anomalies of the Aircraft magnets in its motors.. Why do you think you need to rotate the model otherwise? In multiple axis?

Please provide some evidence of this fictitious magnetic model in the Aircraft.

You make too many assumptions, Sir.

So why does DJI make you recalibrate the compass if you travel > preset distance. (By the way that distance reflects the distance when the deviation becomes substantial.)

You persist in this fantasy and we have had this argument before. Enough for me to stop coming to this site.

I appreciate you knowledge on other subjects, particularly your diagnostic skills on reviewing flight logs. But please don't assume that you are an expert in everything. You don't know, what you don't know, and assumption is not a valid scientific process unless you can validate and prove that assumption with some facts. I did this before in our discussions and provided many reference sources. You to date have provided none.

Maybe I will come back to this site in a year or so.
 
Im, sorry but this is BS. There is no such thing as a global magnetic model in the aircraft firmware.

Rotating the model allow it to get an accurate calibration to magnetic north with respect to the aircraft orientation and the anomalies of the Aircraft magnets in its motors.. Why do you think you need to rotate the model otherwise? In multiple axis?

.....
The aircraft has a magnetometer and 2 inertial measuring units in it. If you rotate the aircraft around and around, the magnetometer will figure out the direction of magnetic north. However, finding true north (the point on the surface of the earth that is on the axis of rotation) can only be found by a full inertial alignment (which fine tunes the local vertical, then the latitude, and the direction of true north relative to the drone) BUT an inertial alignment takes a long time to be anywhere near accurate. The reason for this is that the direction of true north can only be found by observing the change in the gravity vector as the earth rotates - and that only changes by 15 degrees an hour.

So we dont do inertial alignments with our drones - with the grade of inertial in these drones it would take longer than the battery lasted. It relies on knowing its heading, and a rough level (startup level is key).

So the mag declination is most likely to be in a lookup table ie get GPS latitude / longitude -> lookup = done.

Can I prove the firmware has a lookup table? No. But I know these drones don't have the time to find true north through inertial alignment.
 
Im, sorry but this is BS. There is no such thing as a global magnetic model in the aircraft firmware.

Rotating the model allow it to get an accurate calibration to magnetic north with respect to the aircraft orientation and the anomalies of the Aircraft magnets in its motors.. Why do you think you need to rotate the model otherwise? In multiple axis?

Please provide some evidence of this fictitious magnetic model in the Aircraft.

You make too many assumptions, Sir.

So why does DJI make you recalibrate the compass if you travel > preset distance. (By the way that distance reflects the distance when the deviation becomes substantial.)

You persist in this fantasy and we have had this argument before. Enough for me to stop coming to this site.

I appreciate you knowledge on other subjects, particularly your diagnostic skills on reviewing flight logs. But please don't assume that you are an expert in everything. You don't know, what you don't know, and assumption is not a valid scientific process unless you can validate and prove that assumption with some facts. I did this before in our discussions and provided many reference sources. You to date have provided none.

Maybe I will come back to this site in a year or so.
Well, mine doesn’t ask me unless it’s a long way... only had it on 1000+ mile trips. However it asks me.

Regarding the other... you must know how it works. Just tell us and use it to explain a drone’s actions. TIA
 
  • Like
Reactions: charliesRig
Im, sorry but this is BS. .
I'm sorry but whatever field your expertise lies in, you are not correct on this.
Rotating the model allow it to get an accurate calibration to magnetic north with respect to the aircraft orientation and the anomalies of the Aircraft magnets in its motors.. Why do you think you need to rotate the model otherwise? In multiple axis?
The only thing that calibrating the compass does is to identify the magnetic fields that are part of the drone.
You make too many assumptions, Sir.
It would appear that you are the one making incorrect assumptions.
So why does DJI make you recalibrate the compass if you travel > preset distance. (By the way that distance reflects the distance when the deviation becomes substantial.)
You'd have to ask DJI that because they've never explained it.
What is known is that most DJI drones don't need it at all
(I've been flying my Phantom for >2.5 yrs and never calibrated a thing and it still flies as well as it did on day 1 - and I've travelled thousands of miles with it in multiple countries.
Why DJI make some M2 drones ask for recalibration of the compass is a complete mystery as there is no physical reason for it.
I appreciate you knowledge on other subjects, particularly your diagnostic skills on reviewing flight logs. But please don't assume that you are an expert in everything.
You don't know, what you don't know, and assumption is not a valid scientific process unless you can validate and prove that assumption with some facts.
If you only knew the qualifications of the person you are addressing, you might rethink that.
 
Im, sorry but this is BS. There is no such thing as a global magnetic model in the aircraft firmware.

Rotating the model allow it to get an accurate calibration to magnetic north with respect to the aircraft orientation and the anomalies of the Aircraft magnets in its motors.. Why do you think you need to rotate the model otherwise? In multiple axis?

Please provide some evidence of this fictitious magnetic model in the Aircraft.

You make too many assumptions, Sir.

So why does DJI make you recalibrate the compass if you travel > preset distance. (By the way that distance reflects the distance when the deviation becomes substantial.)

You persist in this fantasy and we have had this argument before. Enough for me to stop coming to this site.

I appreciate you knowledge on other subjects, particularly your diagnostic skills on reviewing flight logs. But please don't assume that you are an expert in everything. You don't know, what you don't know, and assumption is not a valid scientific process unless you can validate and prove that assumption with some facts. I did this before in our discussions and provided many reference sources. You to date have provided none.

Maybe I will come back to this site in a year or so.

I've no interest in convincing you of any of this - just correcting the misinformation that you keep posting in case anyone else is confused by it. I don't know exactly which model DJI uses, but it will be one of the common ones. For example:


It might be a lookup table based on a global model, but computationally that's not necessary.

If you want to try to understand why rotating the aircraft cannot get declination then consider the following. All the magnetometers detect is the magnetic field, from which the FC can determine the direction of magnetic north. To measure the declination it would also have to know the direction of true north, which it does not - it has no way to detect true north.

As I and others have explained multiple times before, the purpose of the rotation is for the FC to observe which components of the magnetic field change with the rotation (those are the external field - assumed to be the earth's undistorted field) and which remain constant during rotation (the internal magnetic field of the aircraft itself). That allows it to store the internal field in its calibration (which you can read in the event stream) and subtract that from the measured field during flight, leaving just the earth's field that it needs for navigation. It's not even remotely complicated, so I've no idea at all why you seem to be incapable of understanding it.
 
The aircraft has a magnetometer and 2 inertial measuring units in it. If you rotate the aircraft around and around, the magnetometer will figure out the direction of magnetic north. However, finding true north (the point on the surface of the earth that is on the axis of rotation) can only be found by a full inertial alignment (which fine tunes the local vertical, then the latitude, and the direction of true north relative to the drone) BUT an inertial alignment takes a long time to be anywhere near accurate. The reason for this is that the direction of true north can only be found by observing the change in the gravity vector as the earth rotates - and that only changes by 15 degrees an hour.

So we dont do inertial alignments with our drones - with the grade of inertial in these drones it would take longer than the battery lasted. It relies on knowing its heading, and a rough level (startup level is key).

So the mag declination is most likely to be in a lookup table ie get GPS latitude / longitude -> lookup = done.

Can I prove the firmware has a lookup table? No. But I know these drones don't have the time to find true north through inertial alignment.

An inertial alignment would also require an actual gyrocompass, since the drift and bias in the MEMS rate gyros would easily exceed the rate of change of the gravitational field vector.
 
Hi folks!

Long time pilot (P2V, P2V+, P3P, Mavic Air, Mavic 2 Pro), first time poster! I'm hoping one or more of you could offer any insight or expertise with my situation. I was on a family trip in Pontiac, MO last weekend, and took the Mavic 2P up for some sunset shots the first night with zero issues. Have always felt the Mavic 2 was a super stable, reliable drone. The next day, I took it up to view the young cousins out tubing on the boat, and all conditions were great. Weather was clear, minimal wind, clear LOS to the drone, IMU and compass calibrated before the flight. Never lost GPS or radio signal, but 5 minutes into the flight, flying to the left, the drone picked up a mind of its own. It tilted at an extreme angle on its own, and within literally 2 seconds (verified on the flight logs), it ascended from roughly 32ft to 62ft (with no left stick command), and speed jumped from 24mph to 48mph+ (in Positioning mode).

From there, it was only a matter of 5 or 6 more seconds before the drone headed down sideways into the water, not responding to any commands. The drone threw no warnings or errors until it hit the water (loss of signal, check antenna).

I have DJI Care Refresh, but the lake is over 135 ft deep where it went own, and diving isn't allowed at that depth, even with the experts, since the lake feeds an enormous dam.... obviously, if I can't recover the drone, the Refresh plan doesn't help me. Is anyone here well-versed in looking at the flight data to shed some light on what caused this to happen? I've already reached out to DJI and provided them the info, and while I'm hopeful and confident that they'll stand behind the product malfunctioning, it would be helpful to know what could have caused this for future reference... it's scary to think if that type of loss of control happened over land where it could potentially have affected people on the ground, or if it had struck one of the kids tubing at such a high rate of speed...

Thank you in advance!!
Sorry to hear what happened, that water sure is a killer for us.
 
  • Like
Reactions: codaroommedia
I've no interest in convincing you of any of this - just correcting the misinformation that you keep posting in case anyone else is confused by it. I don't know exactly which model DJI uses, but it will be one of the common ones. For example:
Usual model is to have an array of values in the code to provide declination value for a given coordinates , after it received from the gps signal. Now, if gps is not on the model it becomes a problem, and one needs either specify a location in settings or recalibrate mag offsets.
 
I have done a fair bit of open water sailing, and besides staying at a Holiday Inn Express last night, also have several open water navigation courses under my belt. Maybe this is well known to folks, but in case not, maybe this will help the discussion. In sailing terms the magnetic variation is the difference between magnetic north and true north. I think the more formal term is magnetic declination. And magnetic deviation is the difference between magnetic north and where the boat's compass actually points.

Nautical charts have magnetic variation printed on them, along with a date, and an annual correction to the variation to apply for each year past the chart reference. This is needed because the position of the magnetic north pole is not static, it actually moves with time.

Apart from the chart data, there are several computational models that take latitude, longitude and time as input and spit out a variation number, two well known ones are The World Magnetic Model, a joint US & UK development, and The International Geomagnetic Reference Field. So yes there ARE publicly available models to get variation (declination) from lat/lon data (and date/time). Websites here for anyone interested: World Magnetic Model | NCEI & IAGA V-MOD Geomagnetic Field Modeling: International Geomagnetic Reference Field IGRF-12.

And for deviation - this is the result of things on the boat that have an influence on where the compass actually points, because it rarely is magnetic north. There are many things that can influence the magnetic field in the vicinity of the compass, including anything metal, batteries, electrical circuitry, other gear like GPS, depth sensors, etc, etc, etc. The deviation if unique to a particular vessel, and is constant wherever that vessel is throughout the world.

The process by which the deviation for a particular vessel is obtained is called "swinging the compass" (hmm, sound familiar?). The vessel is steered to a set of known headings in different directions, typically 8 or more. At each known heading the compass indication is recorded. This separates the influence on the compass that remains constant at all headings (the earth's magnetic field) and the part that changes with changing heading (the impact of the "other stuff"). The result is a deviation card with the correction the navigator needs apply at each heading. Generally the deviation is not constant for all headings because the influence of the "other stuff" varies depending on its orientation to the earth's magnetic field & the compass. To me this is analogous to the compass calibration called for by the DJI Go software.

I can't speak to exactly how DJI generates its variation and deviation data, but it seems likely that they use one of the readily available models to determine variation (declination), and then they rely on the results from the "calibration dance" for deviation.

BTW - I don't have a horse in this particular scrap, just wanted to contribute something that might help clarify things based on something unrelated to drones. My apologies if this is not considered constructive.
 
Last edited:
I have done a fair bit of open water sailing, and besides staying at a Holiday Inn last night, also have several open water navigation courses under my belt. Maybe this is well known to folks, but in case not, maybe this will help the discussion. In sailing terms the magnetic variation is the difference between magnetic north and true north. I think the more formal term is magnetic declination. And magnetic deviation is the difference between magnetic north and where the boat's compass actually points.

Nautical charts have magnetic variation printed on them, along with a date, and an annual correction to the variation to apply for each year past the chart reference. This is needed because the position of the magnetic north pole is not static, it actually moves with time.

Apart from the chart data, there are several computational models that take latitude, longitude and time as input and spit out a variation number, two well known ones are The World Magnetic Model, a joint US & UK development, and The International Geomagnetic Reference Field. So yes there ARE publicly available models to get variation (declination) from lat/lon data (and date/time). Websites here for anyone interested: World Magnetic Model | NCEI & IAGA V-MOD Geomagnetic Field Modeling: International Geomagnetic Reference Field IGRF-12.

And for deviation - this is the result of things on the boat that have an influence on where the compass actually points, because it rarely is magnetic north. There are many things that can influence the magnetic field in the vicinity of the compass, including anything metal, batteries, electrical circuitry, other gear like GPS, depth sensors, etc, etc, etc. The deviation if unique to a particular vessel, and is constant wherever that vessel is throughout the world.

The process by which the deviation for a particular vessel is obtained is called "swinging the compass" (hmm, sound familiar?). The vessel is steered to a set of known headings in different directions, typically 8 or more. At each known heading the compass indication is recorded. This separates the influence on the compass that remains constant at all headings (the earth's magnetic field) and the part that changes with changing heading (the impact of the "other stuff"). The result is a deviation card with the correction the navigator needs apply at each heading. Generally the deviation is not constant for all headings because the influence of the "other stuff" varies depending on its orientation to the earth's magnetic field & the compass. To me this is analogous to the compass calibration called for by the DJI Go software.

I can't speak to exactly how DJI generates its variation and deviation data, but it seems likely that they use one of the readily available models to determine variation (declination), and then they rely on the results from the "calibration dance" for deviation.

BTW - I don't have a horse in this particular scrap, just wanted to contribute something that might help clarify things based on something unrelated to drones. My apologies if this is not considered constructive.

That's correct, and in fact "variation" and "declination" are synonyms. Magnetic deviation, due to the internal magnetic field of the aircraft components in much the same way as the steel hull of a ship, is exactly what the calibration is detecting and correcting for, and so the calibration process is directly analogous to your "swinging the compass" process. Thanks for pointing that out.
 
Generally the deviation is not constant for all headings because the influence of the "other stuff" varies depending on its orientation to the earth's magnetic field & the compass. To me this is analogous to the compass calibration called for by the DJI Go software.

I can't speak to exactly how DJI generates its variation and deviation data, but it seems likely that they use one of the readily available models to determine variation (declination), and then they rely on the results from the "calibration dance" for deviation.

Yes. And as a sailor I can add that circling with the boat is also the way to calibrate autopilot compass. Any modern autopilot displays either magnetic North or geographic "true North" obtained after computation relative both to the date and to your actual position. No reason for a different process with our drones. And with a boat, you calibrate once except if you have introduced new magnetic objects in the vicinity of the autopilot compass.
 
  • Like
Reactions: RWD and sar104
@Peio64270 , I go thinking about your "new magnetic objects" this a bit. At least when I do the "calibration dance", I have my controller and either iPad or iPhone in my left hand, less than 3 feet away from my M2P. I wonder how much impact that big chunk of magnetic object has on the quality of my calibration, and whether I should be putting the controller on the ground and then doing the dance...

EDIT: just saw the post from @conchman in General Discussions saying exactly this.

I also got to thinking that my "Holiday Inn Express" joke probably does not play well to an international audience, sorry about that. The joke is based on a series of television commercials here in the States that jokingly implied that staying at a Holiday Inn Express on a business trip made you smarter. Which does not sound very funny at all as I explain it - I guess you kind of had to be there.
 
Last edited:
  • Like
Reactions: Brandaloo and rfc
**UPDATE** DJI found this to be a malfunction of the equipment and provided me with a code for a new Mavic 2 Pro, body only, in their store. Very satisfied with their response.

For some of the veterans here... when I received the new drone and unboxed it, there were definite indications that it wasn’t a “new” drone, though it was sold and packaged as one. To the naked eye, just some scuffs on the top, a couple scrape marks on the bottom of the gimbal behind the camera (under the protective sticker), etc. None of that bothers me so long as it flies and it’s covered. Updated firmware, and everything functioned fine.

That said, what concerns me are the back camera sensors, both side and rear facing. The lenses themselves are either off center, recessed into the body, or both. On the other camera sensors, the lens butts up to a foam ring on the inside of the body... I’m assuming that’s a moisture seal. On these back sensors, they don’t contact that foam ring and you can see into the drone. My question— am I overthinking this, or is this something I should reach out to DJI about? My worry would be even ambient moisture affecting the internal components and breaking the down/causing failure over time.

Thanks again!!
 
**UPDATE** DJI found this to be a malfunction of the equipment and provided me with a code for a new Mavic 2 Pro, body only, in their store. Very satisfied with their response.

For some of the veterans here... when I received the new drone and unboxed it, there were definite indications that it wasn’t a “new” drone, though it was sold and packaged as one. To the naked eye, just some scuffs on the top, a couple scrape marks on the bottom of the gimbal behind the camera (under the protective sticker), etc. None of that bothers me so long as it flies and it’s covered. Updated firmware, and everything functioned fine.

That said, what concerns me are the back camera sensors, both side and rear facing. The lenses themselves are either off center, recessed into the body, or both. On the other camera sensors, the lens butts up to a foam ring on the inside of the body... I’m assuming that’s a moisture seal. On these back sensors, they don’t contact that foam ring and you can see into the drone. My question— am I overthinking this, or is this something I should reach out to DJI about? My worry would be even ambient moisture affecting the internal components and breaking the down/causing failure over time.

Thanks again!!
None of these drones are hermetically sealed, as they need cooling air. However, if you think the build quality of what they sent you isn't up to standard and might impact the performance of a sensor, you should send it back for another one.
 
UPDATE/further help from the experts??

Hi folks! As stated, DJI replaced this under warranty by providing a coupon code for a new "body only" Mavic 2 Pro on dji.com. A slow process, as expected, but a great result. Here's the issue: the brand new M2P body ordered on DJI.com shipped from Atlanta, in new packaging, but was clearly not new when opening the package (scratches on the body and gimbal). Incredibly minor stuff, and it didn't bother me in the least, so long as it worked and was still covered by DJI! After a few days of getting all the batteries/drone/controller synced on the right firmware, etc... I took her up. On the first flight, I kept getting "obstacle avoided, change course" messages, even though I wasn't near anything. I reflashed firmware and tried again, this time resulting in the M2P stopping itself mid-air. I noticed that the drone vision system was showing intermittent graphics like something was at the read of the drone/in the proximity. However, the drone was immediately above me with nothing surrounding it, no birds, insects, etc. I tried again in a few different areas, just to try to eliminate a particular flying area, and it would intermittently happen, especially with drone movement. When I brought the drone down, I noticed that the rear vision sensors/lenses were a bit off center and recessed into the body. There is a small foam gasket on the inside of the body lens "holes," and the lenses were not flush. Shining a light from an angle, you could see right into the body.

I contacted DJI immediately, and they asked for pictures, which I provided right on the phone. The agent was fantastic, and after consulting with her team, told me it was a manufacturing error, and needed to be sent back for replacement. No problem. I sent it back (UPS ground 4 business days there, 4 days back), and DJI swapped it with a new serial number and shipped it back to me. Upon unboxing the new drone, the rear vision sensors/lenses were again recessed, on top of a faulty gimbal/overload error right out of the box. Never flown. Contacted DJI again, and they "expedited" my case and offered to rush the shipping back. Send it back to California again, they swapped the serial number/unit again, and even after promising multiple times, did not rush the shipping back... (missed a deadline as a result, but that's a different issue). Upon receiving now my FOURTH M2P a few days ago, I opened the box to a new or refurbed unit and immediately noticed the rear vision sensors are AGAIN recessed, but this time it's seemingly just the top one. I thought that perhaps I was being overly picky or just insanely unlucky, so I activated the drone. It seems to fly well, as did the others, but I got a couple false "rear proximity" graphics in the first flight, and now I'm INCREDIBLY nervous. I want to be able to trust my drone (as I did the sunken original). Can anyone offer some insight here? Does anyone have the same manufactured gaps in the rear sensors or false rear readings (pics attached)? It seems exceedingly unlikely that I'd receive 3 straight manufacturing defects, but I'm confident this is not the design of the aircraft. In addition, if this needs to be replaced yet again (which I'm sure it does), does anyone have any advice on getting the process sped up? Each time I've had to send these in, through no fault of my own, it's a solid 2 weeks at minimum without my M2P.

Thanks in advance!!!!
 

Attachments

  • IMG_0690.JPG
    IMG_0690.JPG
    1.5 MB · Views: 30
  • IMG_0694.JPG
    IMG_0694.JPG
    1 MB · Views: 29
  • IMG_0695.JPG
    IMG_0695.JPG
    965.7 KB · Views: 30
  • IMG_0698.JPG
    IMG_0698.JPG
    1.2 MB · Views: 29
Wow, although always replaced, that gives me a headache and I certainly understand your frustration and anxiety of the quality.
However, I would still put her up in the sky and see if these strange errors come to show. DJI however will have the upper hand in this process.

It has been unfortunately proven, that replacements are often done with refurb units, which per se is not a bad thing but if there's no real QC, you are at someone's mercy.
Please report back, a very interesting case.

Hope this unit is doing ok. Give it a try.
I've checked mine and they seem to be a bit recessed too, as I can see some black between the outer case and the grey ring. However you upper one is clearly more into the shell.

dji_drone_rear.jpg
 
Last edited:
Yes, the errors still show. The drone seems to think there’s something behind it when there isn’t from time to time. My guess is shadows caused by the improper positioning of the lenses on the rear. Again, I have nothing against a refurb, so long as it functions properly. All I know is, since the first M2P (which I loved, and is the original cause of this thread) took a dive in the middle of June, I haven’t had a fully functional drone that I’m confident in, and it’s almost October. Does anyone at DJI on here, or anyone with high level DJI experience have any recommendations?
 
Wow, although always replaced, that gives me a headache and I certainly understand your frustration and anxiety of the quality.
However, I would still put her up in the sky and see if these strange errors come to show. DJI however will have the upper hand in this process.

It has been unfortunately proven, that replacements are often done with refurb units, which per se is not a bad thing but if there's no real QC, you are at someone's mercy.
Please report back, a very interesting case.

Hope this unit is doing ok. Give it a try.
I've checked mine and they seem to be a bit recessed too, as I can see some black between the outer case and the grey ring. However you upper one is clearly more into the shell.

View attachment 82604
Are you able to take similar shots of your rear lenses? I’m curious to see if yours are indeed different. Thanks!!!
 
Lycus Tech Mavic Air 3 Case

DJI Drone Deals

New Threads

Forum statistics

Threads
130,586
Messages
1,554,112
Members
159,586
Latest member
DoubleBarS