Mid Air "Compass Error, Exit P-GPS Mode"

[blade strike]

My only gripe and it's a big one with the P3 compass, is that while it shares the same compass safety feature as the inspire in order to prevent flyaways, the 2 inspire pilots at out club have never had one compass error thrown. There are 3 of us flying P3's and we have all had them while trying to take of near the club house about 10 to 15 Mtrs away, while we were trying these take of points and testing in order to see if P3's were safe to fly at our club, one inspire pilot took of about ten times getting closer and closer each time and he ended up taking of a Mtr away with not one compass error. Now anyone flying a P3 has to take of from a dedicated helipad at the other end for safety reasons of course. This testing we did the other month has made me believe the P3 compass is more sensitive than that of its big brother the inspire.

That's because the inspire has more room to get the compass away from the things that generate interference. I have already put in to R&D question regarding the sensitivity of the compass logic. So they will be looking into it. But the thing is we had a few hundred beta testers and we didn't see any increase of compass errors.

 

[leekiya]

Yea that was the conclusion we came up with. Good to know they will look into it tho, we guessed it might be something they might figure out with the next generation P3, but like you say it's room and distance that's the problem. Better it's to sensitive tho to prevent flyaways like you pointed out the other day, and I feel people should learn how to fly Atti anyway just in case. Since those tests were done the other month i have practised and thought a lot what I would do if I got a compass error 1 or 2k out and then lost FPV with a disconnect on the way back, it would take some pretty tricky flying and staying calm but I would like to think I had it covered. Lol

 

[isopro]

That's because the inspire has more room to get the compass away from the things that generate interference. I have already put in to R&D question regarding the sensitivity of the compass logic. So they will be looking into it. But the thing is we had a few hundred beta testers and we didn't see any increase of compass errors.

Im at over 200 flights on my P3, i`ve add compass error dropout in average every 30 flights it happen mostly if i dont wait 20 sec after take-off and or randomly mid flight, i've no problem flying in atti mode, BUT 2 flight ago, i was filming a house close to a lake and after my usual POI flight over and arround the house, a Sea-Doo passed by on the lake and it looked to me like a great ending for the video i was producing, so i started chassing the said seadoo, it was a sunset, the sun add huge reflection on the lake, space weather was calm, no wire or antenna in the next 1 mile at least, 10-15KM wind over the lake, last time i callibrated compass was weeks ago and it always fly flawlessly, when suddenly i add compass error dropout, BUT this time it looks like i also add partial lost of control of my craft, Gyroscope went nuts, Barometer also, the craft and gimbal were banking abnormally, i was sudenlly gaining speed in random direction and loosing altitude because banking angle was to high, I was able to put the throttle to the max and guide the craft back over land, once over land i gained back proper control, and landed, resseted craft checked all sensors, reviewed footage, rebooted and did the rest of the shot over land only...But i wonder what could of caused me to get such abnormal lost of Compass and Gyro, at same time.. ???
Now im in bed but later maybe i can get flight log and upload it here+video recording...
Would you know what could of caused this ?

 

[snowghost]

Taking my dji hat off....

Imho, people calibrating their compass at every time they fly is probably the root cause for this.

Once you have a solid compass calibration, no need do it again. Unless you travel a few hundred miles or apply a firmware update.

You have to remember some soil has a high amount of metals in it.. not saying that's the cause but just because you're in an open field doesn't mean much.

So what I do is calibrate my compass and if it's a solid calibration I leave it, period.

Dji hat back on....

I don't touch it even after firmware updates. I check the values every time I fly though.

 

[isopro]

I don't touch it even after firmware updates. I check the values every time I fly though.

Yup i also check values before every take off and after every landing, value of compass, barometer, interference, battery cells before, and about every 2 minutes into each flight...I wonder what caused the issue i add..

 

[Foosy]

Sorry but you're 100% wrong.. without a compass the craft can't fly in gps mode..This is what causes a flyaway, the system listening to bad compass data.

Maybe I was not clear in my original post:
Compass is used in navigation, but is not always needed. For example it should be able to hover without a compass. If there is a compass error, the last thing I want is for it to also ignore gps. If it were to maintain its position it would not fly away.

My near flyaway occurred when it exited GPS mode mid flight for no reason. Well, it claimed a compass error, but I don't believe it. And when it did, it started to drift away.

Algorithmically, if there is a compass error there is an offset from where true north is. I see it as if the phantom is pointed in a different angle than it really is. Not ideal for flying but still manageable, if it maintains its GPS.

But I have for dji an optimization that will prevent flyaway a if there is a compass error. Even if the compass is going bonkers and is jumping all over the place, the best thing it can do is return home, and for this a compass is not needed: both bird and home GPS coordinates are known, hence angle of flight can be derived. It should pick a direction and continuously calculate distance and orientation. It can modify its orientation until it reaches a state where the angle stays constant and the distance diminishes. (Trig: shortest path between two points).

 

[ianwood]

Maybe I was not clear in my original post:
Compass is used in navigation, but is not always needed. For example it should be able to hover without a compass. If there is a compass error, the last thing I want is for it to also ignore gps. If it were to maintain its position it would not fly away.

My near flyaway occurred when it exited GPS mode mid flight for no reason. Well, it claimed a compass error, but I don't believe it. And when it did, it started to drift away.

Algorithmically, if there is a compass error there is an offset from where true north is. I see it as if the phantom is pointed in a different angle than it really is. Not ideal for flying but still manageable, if it maintains its GPS.

But I have for dji an optimization that will prevent flyaway a if there is a compass error. Even if the compass is going bonkers and is jumping all over the place, the best thing it can do is return home, and for this a compass is not needed: both bird and home GPS coordinates are known, hence angle of flight can be derived. It should pick a direction and continuously calculate distance and orientation. It can modify its orientation until it reaches a state where the angle stays constant and the distance diminishes. (Trig: shortest path between two points).

Compass is essential in all forms of GPS based flight. Period. This includes position hold, RTH and everything else.

 

[Foosy]

Compass is essential in all forms of GPS based flight. Period. This includes position hold, RTH and everything else.

My explanations must not be getting to you.
While the current implementation may always use compass, it is not necessary for all forms of movement. Specifically, if the coordinates of the destination are known.

As I pointed out, you can have an algorithm that navigates relationally. That means if you know your current position and your destination, you do not need a compass. You just use feedback from any random movement to calculate if your movement is correct or incorrect and you continue to change your movement until your calculation shows that your movement is optimal.

Since this requires a converging algorithm, it is not as good as flying with a compass, but better than a flyaway.

 

[blade strike]

With current GPS hardware and refresh rates this is not an option. A compass allows for immediate direction correction.

 

[ianwood]

My explanations must not be getting to you.
While the current implementation may always use compass, it is not necessary for all forms of movement. Specifically, if the coordinates of the destination are known.

As I pointed out, you can have an algorithm that navigates relationally. That means if you know your current position and your destination, you do not need a compass. You just use feedback from any random movement to calculate if your movement is correct or incorrect and you continue to change your movement until your calculation shows that your movement is optimal.

Since this requires a converging algorithm, it is not as good as flying with a compass, but better than a flyaway.

I understood what you said. There's three reasons it won't work:

1. Wind.
2. Varying levels of GPS precision.
3. Mild to moderate compass errors that return plausible values.

 

[shane0926]

With current GPS hardware and refresh rates this is not an option. A compass allows for immediate direction correction.

What is the current gps refresh rate?

 

[blade strike]

What is the current gps refresh rate?

Sorry not allowed to get that technical...Ian might be able to share though

 

[ianwood]

I would guess it's >10Hz. Even with a high refresh rate, you still can't overcome GPS accuracy limitations. If you can guarantee <1m accuracy with little to no wind, you could do it in conjunction with the gyro.

Increase wind and/or HDOP and the amount of space needed goes up quickly. The Phantom would have to keep increasing the test radius until it gets a high confidence in the heading. It could use that for a number of seconds and then as the gyro drift increases, it would need to re-correct using another test radius. With some blustery wind and less than perfect HDOP, you might never get home.

Ironically, finding and maintaining a heading only from GPS course could look kind of like TBE.

Finally, factor in that inaccurate compass readings can often be within the plausible range and thus not identified as an error. So you would have to manually trigger a no-compass GPS mode or risk entering this mode for no reason.

 

[Bret Lucas]

Where to start?

• KP index has nothing to do with compass errors and anything under 8 is not a concern.
• Power lines will not cause magnetic interference unless you land on them.
• Calibrating your IMU will not help a compass error.

OP, suggest you find a safe place to do a compass calibration away from any and all metal and magnetic sources. Hopefully that should keep it from reoccurring.

Solid advice Ian. In addition, I think DJI have included the "Compass error - switch to ATTI" as a red flag course of action which wasn't available in previous versions. So in a way this is a good thing for the operator.

 

[Foosy]

I would guess it's >10Hz. Even with a high refresh rate, you still can't overcome GPS accuracy limitations. If you can guarantee <1m accuracy with little to no wind, you could do it in conjunction with the gyro.

Increase wind and/or HDOP and the amount of space needed goes up quickly. The Phantom would have to keep increasing the test radius until it gets a high confidence in the heading. It could use that for a number of seconds and then as the gyro drift increases, it would need to re-correct using another test radius. With some blustery wind and less than perfect HDOP, you might never get home.

Ironically, finding and maintaining a heading only from GPS course could look kind of like TBE.

Finally, factor in that inaccurate compass readings can often be within the plausible range and thus not identified as an error. So you would have to manually trigger a no-compass GPS mode or risk entering this mode for no reason.

Great data!
So now we have a great conversation going.

So , if I understand you, you are saying that in an ideal world with no winds, and a faster processor, we could use navigation without a compass in certain situations. I agree with this.

Also, you bring a great point: when should an algorithm trigger this mode of navigation (GPS without compass), suggesting that compass errors are always there and we should be careful of not entering this mode for no reason.
The answer is easy: it would trigger this mode at the same instance as now the attitude mode is triggered due to compass error.
There is already a decision process, and as happened to me and a few others, after an unexpected compass error, well off the ground, the system switched off GPS mode. So it does not do it for small compass errors. And I could argue that what it did was less safe than what I am suggesting it does. Of course we would have to put the system to test, but wouldn't we all agree that if we could optimize this algorithm of no-compass GPS RTH then we would have a safer platform?

We could still allow users to switch it to attitude mode and try to take control the way we do now, but, in my opinion, we would have provided a safer system.

Of course I am speaking all theoretically. Would have to code it and measure performance to see how it performs.

 

[Foosy]

One more idea: if the processor on board is not fast enough to calculate the test radius, we could use the mobile device to do that.
As long as the phantom is communicating with the controller and app, if we enter a 'catastrophic situation, due to compass issues' we could have the app take control and send the navigation data to the phantom. All the phantom needs to do is send its current coordinates, which it already does now.

 

[Recce2]

I had the same issue last evening over water. Started toilet bowl, auto switched to atti, I flew it back in atti. Did a compass calibration and it was fine after that.
Never happened before.
I did not do a compass calibration before flight because it was close to home and I like doing it where I know things are clear. I guess I will start doing the compass cal before every flight now and be ready for atti mode.

 

[Foosy]

I'll add my $0.02 on this:
...
Like what? DJI is using off-the-shelf MEMS components. The compass is the only MEMS component to provide absolute orientation reference to the Earth. If you've got a better idea as to what they should use, I'd love to hear it.

Don't know which MEMS dji is using now. But the BOSCH BNO055 looks promising. http://www.bosch-sensortec.com/en/homepage/products_3/9_axis_sensors_5/ecompass_2/bno055_3/bno055_4

 

[Foosy]

I had the same issue last evening over water. Started toilet bowl, auto switched to atti, I flew it back in atti. Did a compass calibration and it was fine after that.
Never happened before.
I did not do a compass calibration before flight because it was close to home and I like doing it where I know things are clear. I guess I will start doing the compass cal before every flight now and be ready for atti mode.

You should only do compass calibration if you change your location by more than 100 miles. There is no reason to recalibrate. In other words, if there is a need to recalibrate in the same area, it means that the sensor is losing its calibration which would indicate a bug.

I don't think we or dji knows yet why certain compass errors occur. Or if they do, they have not shared them with us. I do not believe that the sensor loses its calibration. I think that the situation where compass errors are occurring are so intermittent, that they are difficult to troubleshoot.

 

[Bob Denny]

Thank you so much. A voice of reason.

Sent from my Nexus 7 using Tapatalk