First flight crash

[DanMan32]

If it can manage to compensate for compass deviation by leveraging GPS, this staying out of ATTI mode, wind would not be much more of an issue than "normal" flight.

 

[BudWalker]

........

One question to you deep analyzing guys ... This magnetic area ... should that have triggered a message to calibrate the compass?

More on my first response to this. A compass error will be triggered if the magnetometer modulus (i.e. geomagnetic field strength) falls outside the normal range. @sar104 has previously determined the field strength normal range to be [1400, 1600] for the M2. That's probably the same for the P3 as well as other DJI platforms.

The field strength normal range for the MM appears to be wider. If it were [1400, 1600] then a compass error would have existed at the launch point chosen by @Enrico BBB . At -70 secs the MM was placed in a location where the magMod was about 1350 but the compass error was false. Had this been an M2 the compass error would be true.


From the GPS data it appears that the MM was being moved around in the interval where compass error was true. @Enrico BBB were you searching for a spot where the compass error would go away?


@Enrico BBB you had asked who was at fault. IMHO, the normal field strength range is too wide - it needs to be narrower. Had it been [1400, 1600] the compass error would have been shown and motor start would have been prevented.

 

[Enrico BBB]

More on my first response to this. A compass error will be triggered if the magnetometer modulus (i.e. geomagnetic field strength) falls outside the normal range. @sar104 has previously determined the field strength normal range to be [1400, 1600] for the M2. That's probably the same for the P3 as well as other DJI platforms.

The field strength normal range for the MM appears to be wider. If it were [1400, 1600] then a compass error would have existed at the launch point chosen by @Enrico BBB . At -70 secs the MM was placed in a location where the magMod was about 1350 but the compass error was false. Had this been an M2 the compass error would be true.
View attachment 88072

From the GPS data it appears that the MM was being moved around in the interval where compass error was true. @Enrico BBB were you searching for a spot where the compass error would go away?
View attachment 88073

@Enrico BBB you had asked who was at fault. IMHO, the normal field strength range is too wide - it needs to be narrower. Had it been [1400, 1600] the compass error would have been shown and motor start would have been prevented.

I rotate the drone about 180 degrees before launch because the app asked me to put the back side of the drone in my direction (I have been asked to be in the back side of the drone before launch). There was absolutely no wind! I don't know if this could be helpful but the drone got crazy after it reached a fly quote higher than the surrounding buildings height. But no wind!

 

[Prismatic]

If it can manage to compensate for compass deviation by leveraging GPS, this staying out of ATTI mode, wind would not be much more of an issue than "normal" flight.

I think you underestimate the difficulty wind poses with respect to "dynamic compass calibration".

The wind anywhere near the ground consists of a swirling stream of knotted vorticies and gusts. The instantaneous directional vector of any drone-sized volume of the atmosphere is unpredictable. Aircraft movement on a given vector is a function of applied thrust and wind.

The flight control system (FCS) uses a reference vector provided by the compass to "subtract out" the wind component and calculate the thrust-induced aircraft movement. If the FCS doesn't know the true direction in which thrust is applied--if that variable is an unknown quantity--the navigation equations are unsolvable.

In other words, without the reference vector provided by the compass, the FCS can't even guess how the thrust it applies relates to the observed change in the aircraft's position.

In perfectly still air, dynamic compass calibration might be possible, because the FCS could assume a wind vector of zero. But that does not remotely approximate a general solution.

 

[Prismatic]

... But no wind!

You detected no wind at the launch point, I am sure. However, the wind changes with altitude--sometimes a little, sometimes a lot. Even an imperceptible breeze will move the drone and require compensating thrust. And away it goes.

I'll note that the drone went out of control just about the time it reached the height of the surrounding structures, where it was no longer protected by them. That is, just as it may have first needed to apply position-stablizing thrust.

 

[Enrico BBB]

You detected no wind at the launch point, I am sure. However, the wind changes with altitude--sometimes a little, sometimes a lot. Even an imperceptible breeze will move the drone and require compensating thrust. And away it goes.

I'll note that the drone went out of control just about the time it reached the height of the surrounding structures, where it was no longer protected by them. That is, just as it may have first needed to apply position-stablizing thrust.

It is possible... This is the reason why I talked about the height comparing to the surrounding buildings. I don't know if wind started all that mess but I saw the drone starting to the right full throttle even if, as we know, I didn't give any kind of input. When I saw this ( we are talking about 4 sec) I tried to give some input, especially go home function but the drone didn't accept any of them.

 

[rob levine]

I can only interpret what I see ... according to your video the drone crashes to the right but according to the flight log it crashes with the left side of the drone first (look at the drone triangle in my earlier attached pic). For me that indicates a 180 degree compass error. You launched in a place surrounded by a high steel fence ... & under that rubber on the ground it perhaps was a steel plate or reinforcements, all highly magnetic.

How to get an compass error ...

Perhaps you stored the drone close to an magnetic object & by that magnetized it ... or you tried to launch in a magnetic interfered area, steel reinforcement or electrical installation close under ground or steel constructions near by. In this case the drone will give you a message to calibrate the compass in order to launch ... but if you do that every thing goes heywire once airborne & away from the interference. The drone can't anymore hoover depending on that it doesn't know the direction anymore & by that it can't correctly compensate to stay in place as it doesn't know in what direction it shall compensate. Usually this means that the drone starts to bank & turn quicker & quicker, stick inputs in this situation risk to make things even worse ... as the drone doesn't know the direction it is pointing. Things usually escalates pretty quick with a crash as a result. Try to google "toilet bowl effect" ...

One way to avoid this is to change launch location to be sure of that you aren't in a magnetic interfered area before you calibrate the compass due to a message in the DJI fly app. Also look on the drone marked as a triangle in the app ... is it facing to the right direction, if not this is the first sign that something is wrong with the compass, either that you actually shall calibrate or that you are in an area with interference.

My bet so far is compass error due to interfered area & you actually did calibrate (but as you say, you can't remember) ... and by that a pilot error. But perhaps more people will join in that can examine the flight log more in depth & point out a different reason (& explanation to way the heading differs 180 degrees between the video & log).

That's interesting about the pointing of the drone on the interface. Never thought of that. So I could point the drone in a particular direction I knew, i.e. due north, and if the nose doesn't point that way on the interface that would be a problem? Can you post a screencap?

 

[DanMan32]

I think you underestimate the difficulty wind poses with respect to "dynamic compass calibration".

The wind anywhere near the ground consists of a swirling stream of knotted vorticies and gusts. The instantaneous directional vector of any drone-sized volume of the atmosphere is unpredictable. Aircraft movement on a given vector is a function of applied thrust and wind.

The flight control system (FCS) uses a reference vector provided by the compass to "subtract out" the wind component and calculate the thrust-induced aircraft movement. If the FCS doesn't know the true direction in which thrust is applied--if that variable is an unknown quantity--the navigation equations are unsolvable.

In other words, without the reference vector provided by the compass, the FCS can't even guess how the thrust it applies relates to the observed change in the aircraft's position.

In perfectly still air, dynamic compass calibration might be possible, because the FCS could assume a wind vector of zero. But that does not remotely approximate a general solution.

There's still gyroscope, accelerometer and GPS. I will concede though that unless the quad knows exactly how to apply forward thrust with respect to the quad's nose, the quad might not know direction of travel with respect to nose. But I'd rather it fly sideways with it thinking its flying straight but stable than totally out of control.

Anyway, regardless if wind caused GPS delta or AC propulsion, AC would be able to know where north is from the GPS delta, and adjust the compass accordingly.

 

[slup]

That's interesting about the pointing of the drone on the interface. Never thought of that. So I could point the drone in a particular direction I knew, i.e. due north, and if the nose doesn't point that way on the interface that would be a problem? Can you post a screencap?

If you look at my post #56 I try to explain one easy way to determine if something fishy is going on with the drone compass that you easily can implement in your pre-flight procedure. The drone "triangle" shall in relation to the map objects be correct vs. the reality ... & this has nothing to do with if your mobile device compass are true to direction or not, the effect from that shows only by that the map itself isn't aligned with reality. Your mobile device compass error effect though the "Radar" view in GO4 ... & by that makes it hard for you to point the RC towards the drone if the drone are outside VLOS.

Try it by yourself ... power up GO4 & your drone and turn the drone by hand, then the "triangle" also turns.

 

[Prismatic]

... Anyway, regardless if wind caused GPS delta or AC propulsion, AC would be able to know where north is from the GPS delta, and adjust the compass accordingly.

If @sar104 is so inclined on his return to this forum, perhaps he'll weigh in (though I'm sure he'll have plenty of bigger fish to fry). I had this conversation with him some time ago, in which I was defending exactly your idea for dynamic compass calibration. He nicely countered my points, until I realized that the idea just wasn't going to work.
And FWIW, if it were possible to algorithmically detect, contain, and correct flyaways due to compass errors, I'm pretty sure DJI has the engineering team to do it. That they have not suggests that the problem is not as simple as it superficially appears.
To avoid further hijacking of this thread, I'll leave it at that.

 

[ThorDar]

Here's the problem. Bad launch site. GPS was blocked and reflected by the surrounding apartment blocks and trees. A good GPS signal lock requires an open view of the sky down as close to the horizon as possible.

 

[Fly Dawg]

Here's the problem. Bad launch site. GPS was blocked and reflected by the surrounding apartment blocks. A good GPS signal lock requires an open view of the sky down as close to the horizon as possible.

GPS was not the issue in this flight. Sat's and Navhealth were Ok. This is an issue with the IMU/Compass.

 

[ThorDar]

GPS was not the issue in this flight. Sat's and Navhealth were Ok. This is an issue with the IMU/Compass.

View attachment 88123

I disagree. Regardless of how the logs were affected, he had too few clean satelite

GPS was not the issue in this flight. Sat's and Navhealth were Ok. This is an issue with the IMU/Compass.

View attachment 88123

I respectfully disagree. A poor GPS lock with multipath signals at the start of a flight affects everything that follows. I've been flying GPS drones since 2015 and I've done thousands of flights. I learned early on how critical a good GPS lock is and how too many obstacles to a clear view of the sky affects getting a good, accurate lock.

The start of this was too few satellites locked and reflections off the nearby trees and buildings for those it could detect. The obstacles would also affect the compass. Look closer at that screen shot. This flight never got past poor and dangerous on the GPS lock.

 

[Fly Dawg]

I respectfully disagree. A poor GPS lock with multipath signals at the start of a flight affects everything that follows. I've been flying GPS drones since 2015 and I've done thousands of flights. I learned early on how critical a good GPS lock is and how too many obstacles to a clear view of the sky affects getting a good, accurate lock.

I as well have been flying since 2015, however, you should go back and look at the IMU data. Airdata contains mostly overviews unless you extract the actual data to see the values. As was previously mentioned everything is pointing to an IMU and/or Compass issue. As I mentioned. The GPS data, and hdop look ok at takeoff. However, the directional data does not.

 

[Meta4]

Here's the problem. Bad launch site. GPS was blocked and reflected by the surrounding apartment blocks and trees.

Regardless of how the logs were affected, he had too few clean satelite

A poor GPS lock with multipath signals at the start of a flight affects everything that follows. I've been flying GPS drones since 2015 and I've done thousands of flights. I learned early on how critical a good GPS lock is and how too many obstacles to a clear view of the sky affects getting a good, accurate lock.

The start of this was too few satellites locked and reflections off the nearby trees and buildings for those it could detect. The obstacles would also affect the compass. Look closer at that screen shot. This flight never got past poor and dangerous on the GPS lock.

Interesting idea but it's not supported by the actual data.
Airdata is not the place to see real data.
Have a look at his GPS numbers and GPS Lavel as judged by teh flight controller here:

http://www.phantomhelp.com/LogViewer/DQ8EAKG08224991RUZPI/DJIFlightRecord_2019-12-10_[16-04-53]csv.zip

Not having good GPS does not cause a drone to zip off in an uncontroller fashion.
Lack of GPS was not the problem but the data shows a real issue that was.

 

[sar104]

There's still gyroscope, accelerometer and GPS. I will concede though that unless the quad knows exactly how to apply forward thrust with respect to the quad's nose, the quad might not know direction of travel with respect to nose. But I'd rather it fly sideways with it thinking its flying straight but stable than totally out of control.

Anyway, regardless if wind caused GPS delta or AC propulsion, AC would be able to know where north is from the GPS delta, and adjust the compass accordingly.

If @sar104 is so inclined on his return to this forum, perhaps he'll weigh in (though I'm sure he'll have plenty of bigger fish to fry). I had this conversation with him some time ago, in which I was defending exactly your idea for dynamic compass calibration. He nicely countered my points, until I realized that the idea just wasn't going to work.
And FWIW, if it were possible to algorithmically detect, contain, and correct flyaways due to compass errors, I'm pretty sure DJI has the engineering team to do it. That they have not suggests that the problem is not as simple as it superficially appears.
To avoid further hijacking of this thread, I'll leave it at that.

@Prismatic is correct - knowing change in position alone does not allow the aircraft to compute heading, because it doesn't know whether the change in position was due just to applied thrust or applied thrust plus external force (i.e. wind).

 

[JSM]

I can only interpret what I see ... according to your video the drone crashes to the right but according to the flight log it crashes with the left side of the drone first (look at the drone triangle in my earlier attached pic). For me that indicates a 180 degree compass error. You launched in a place surrounded by a high steel fence ... & under that rubber on the ground it perhaps was a steel plate or reinforcements, all highly magnetic.

How to get an compass error ...

Perhaps you stored the drone close to an magnetic object & by that magnetized it ... or you tried to launch in a magnetic interfered area, steel reinforcement or electrical installation close under ground or steel constructions near by. In this case the drone will give you a message to calibrate the compass in order to launch ... but if you do that every thing goes heywire once airborne & away from the interference. The drone can't anymore hoover depending on that it doesn't know the direction anymore & by that it can't correctly compensate to stay in place as it doesn't know in what direction it shall compensate. Usually this means that the drone starts to bank & turn quicker & quicker, stick inputs in this situation risk to make things even worse ... as the drone doesn't know the direction it is pointing. Things usually escalates pretty quick with a crash as a result. Try to google "toilet bowl effect" ...

One way to avoid this is to change launch location to be sure of that you aren't in a magnetic interfered area before you calibrate the compass due to a message in the DJI fly app. Also look on the drone marked as a triangle in the app ... is it facing to the right direction, if not this is the first sign that something is wrong with the compass, either that you actually shall calibrate or that you are in an area with interference.

My bet so far is compass error due to interfered area & you actually did calibrate (but as you say, you can't remember) ... and by that a pilot error. But perhaps more people will join in that can examine the flight log more in depth & point out a different reason (& explanation to way the heading differs 180 degrees between the video & log).

Just an observation, regardless of a possible compass issue and the Drone in Atti Mode; theoretically it still should have been able to be controlled manually which gets down to pilot skills in these sorts of situations, stand to be corrected on this !!!!

 

[slup]

Just an observation, regardless of a possible compass issue and the Drone in Atti Mode; theoretically it still should have been able to be controlled manually which gets down to pilot skills in these sorts of situations, stand to be corrected on this !!!!

An explanation to why you can't overtake a flyaway due to compass error is given here --> Post #76
Atti mode wasn't a reason for the crash ... it occurred after the crash when the drone were on the ground.

 

[Bob4131]

First flight, take off, vertical climb to aprox 18mt, light rotation (C mode all the time). Suddenly the drone start going left at high speed. Controller completely unuseful because no command has obtained any change. I tried with "come home" option, nothing changed. Drone crashed into a block. Propellers and gimbal destroyed. May I upload a video or the LOG txt? Maybe someone could help me to understand what has really happened....thank you

Let me relate my own experience. I had flown DJI drones for over a year and lifted off from a new location but similiar to previous areas flown so I did not recalibrate the compass. After getting 11 GPS satellites, I lifted straight up. Everything when normally for 4 seconds until, at 49 feet (15 meters), the drone started to fly away at high speed, in a curled flight path. I attempted to make the drone come down but for the next 20 seconds it orbited erratically but did finally get close enough to the ground to auto land. I got a compass warning only after the erratic flight had begun. This all occurred in a parking lot with an asphalt surface. I latter learned, that before the parking lot was there, this area is where a small mountain of iron ore had been stored for a steel mill for ~ 100 years. Much of that ore would have sunk into the ground over the years and was merely paved over.
Your experience seems similar to mine in that when you were high enough, the material in the ground influencing the compass no longer overloaded the earth's natural magnetic field and at that point it went haywire. I think the experts here have given you the cause of the crash. I do not know DJI's warranty policy in this case. It might be interesting to understand the history of your particular patch of ground, if you do not think that rebar in concrete at your takeoff spot was the issue.
The following is speculation on my part, but I have gotten magnetic interference warnings before prior to taking off from concrete with rebar (and I heed the warning and move take off spots), but I wonder if a defused result from my old iron ore spot might have been interpreted differently by the drone and that explains why I got no warning prior to takeoff.

 

[Bob4131]

An explanation to why you can't overtake a flyaway due to compass error is given here --> Post #76
Atti mode wasn't a reason for the crash ... it occurred after the crash when the drone were on the ground.

See my post #99, but I will say I was lucky to have landed without it hitting a light post. If I had been near a building (<40 meters), I would have hit it. The parking lot was nearly empty, otherwise I would have crashed into a vehicle.