Fixing DJI's Compass Problem

[ianwood]

The way I see it, the compass just tells the brains, the direction it is pointed.. Rotating the compass shouldn't really screw up anything, other then deciding which direction is front.. I am extremely careful when I calibrate to make sure the compass is parallel to the Earth.. So with my compass rotated, when I do the second part of the calibration, by turning the copter on edge, I position is so one leg is slightly dipped down (to keep the compass in the same position, in relation to the ground, if it was attached in the factory set position) .. I know my fellow copter flyer is plagued with the major left hook, after a straight run, but he doesn't experience the TBE.. My toilet bowl effect, was upwards of 70 feet..

You're missing the point. No matter how you calibrate the compass, the FC is making incorrect calculations based on the fact your compass is off axis by 30 degrees. So, you've substituted one set of inaccurate calculations for another.

One thing I have noticed, is that if I do a straight run, then let go of the toggle, the copter will keep going straight ahead for a period of time.. This is fine, if there's no swing, one way or the other, but becomes very noticeable, when you're behind the copter and instead of carrying on straight, it makes a dramatic left or right hook, depending on declination.. Maybe there's something in their programing, that momentarily shuts the compass off, when you let go of the toggle.. Otherwise, wouldn't the craft naturally arc left, as you were flying in a straight line? I was thinking maybe precession plays a part, but counter rotating props should null that aspect.. However, I'm not sure if the craft may have a built in gyroscope, which could be sending the wrong message, because of precession.. I just don't know enough about how the electronics function on this craft, other than what I can observe for myself

Nothing shuts the compass off, ever. The FC would have no way to know which way to go. In GPS mode, the FC is pretty much always going somewhere even when you're not moving the sticks. And it has both a 6DOF gyroscope and a 3 axis accelerometer. And gyroscopic precession has no bearing on this issue otherwise everyone would see it evenly which is not the case. This issue is very simply a declination calculation error. I have seen it in the data. Let's stop second guessing it. We pretty much have the smoking gun. The guilty party is identified and it is up to DJI to put them in a corrective facility.

And on that note... For everyone reading and posting in this thread: If you want to continue the debate on rotating the compass you are welcome to do so in another thread. This thread is focused on getting a solution to the declination issue. This thread is not the place to discuss "it kind of works" jerry rigged kludges that compromise the integrity of the sensor and the FC algorithms. DJI themselves have said the Phantom compass is not meant to be rotated. I said that in my first post. I've obliged the debate but I now want the focus to return to the specifics of getting DJI to fix this issue in the only way it can truly be fixed.

 

[sar104]

The way I see it, the compass just tells the brains, the direction it is pointed.. Rotating the compass shouldn't really screw up anything, other then deciding which direction is front.. I am extremely careful when I calibrate to make sure the compass is parallel to the Earth.. So with my compass rotated, when I do the second part of the calibration, by turning the copter on edge, I position is so one leg is slightly dipped down (to keep the compass in the same position, in relation to the ground, if it was attached in the factory set position) .. I know my fellow copter flyer is plagued with the major left hook, after a straight run, but he doesn't experience the TBE.. My toilet bowl effect, was upwards of 70 feet..

Don't forget that the magnetic field is not tangential to the earth's surface - it has a significant vertical component in most locations. Even if the compass sensor is only measuring the horizontal components (relative to the compass sensor), if the compass has been rotated then as the Phantom pitches and rolls the compass axes will not move in the planes expected by the FC, based on its understanding of attitude from the IMU, and the measured field components will produce an incorrect heading value. That could be detected and corrected for by the compass calibration procedure, but only if the software is programmed to expect a compass rotation, which I would guess it is not.

I would further guess that this is at least part of the reason why compass rotation is not a complete fix.

 

[PCDan]

Sorry for asking, but what is TBE? Like most of you I am getting some drift and hook.

 

[obiwan_pierogi]

I can see the reasoning behind, not moving the compass, because everything about the copter -- the toggles etc. are based on the copter design.. Ie: if you push forward on the right toggle, you naturally expect the craft to move, in the direction the camera is pointed.. So by moving the compass, you disrupt this fundamental programing.. I also understand that rotating the compass, doesn't correct the problem.. I'm just at a loss to figure out why if you were to bring your copter over here, it would, most likely,hover in place, yet mine instantly goes into the toilet bowl effect?. ..And why (and how), does my rotating the compass, prevent this from happening?

 

[Bigbells]

Sorry for asking, but what is TBE? Like most of you I am getting some drift and hook.

It's Toilet Bowl Effect, as when the Phantom hunts in circles instead of hovering in a tight zone.

Now, maybe somebody can tell me what FC is.

Edit: ah! Flight Controller.

 

[obiwan_pierogi]

PCDan.. Here is my copter in the TBE.. Note: all I did was rotate the copter on its axis.. The rest of the movement was the copter acting on its own.. You even see me drop the controls to my thighs, to show I'm not touching the controls.. https://www.youtube.com/watch?v=VJBcAhdm5Uw

 

[PCDan]

PCDan.. Here is my copter in the TBE.. Note: all I did was rotate the copter on its axis.. The rest of the movement was the copter acting on its own.. You even see me drop the controls to my thighs, to show I'm not touching the controls.. https://www.youtube.com/watch?v=VJBcAhdm5Uw

Thanks, and wow that's scary. I've had a similar effect but on a much smaller scale/circle, and the P2 seemed to figure out where it was and stabilized. I will just fly in really open spaces until they get the compass fix out.

 

[obiwan_pierogi]

PCDan... This is how my copter normally flies, with the compass in the factory set position.. This is why I think there's something more to the story, than just re-writing the program.. My God, if every copter flew like this around my declination no.s 17.5 positive, everyone would be sending their unit back.. ..But that isn't the case.. A friend North of me, with a slightly higher declination, has no problem with going into the TBE.

 

[Babyyak]

I live in California Bay Area, I have an original phantom 2 shipped to me in December some of the first batches. I confirm I have this j hook issue. Left drifting and j hook after throttle off, also toilet bowl effect. I never really noticed this issue until after I upgraded my p330cb firmware, but It could have been happening before that. I also noticed my phantom is very unstable after the p330cb update with 3.02 and I had to turn my gains down significantly after this to stable out my phantom. Was running gains at 145/145/140/140 atti 110/110, am now at 80/80/70/70 atti 110/110.

 

[SilentAV8R]

I live in southern California. The declination here is around 12 degrees. I have not had any issues, compass related or otherwise.

 

[chrisono]

I live in Victoria B.C and the issue is present in my phantom as well

 

[SilentAV8R]

I don't really notice TBE with my Phantoms, but the j-hook is VERY pronounced. When you guys say TBE, you mean that it hovers in a circle, right? If mine do, it is a very small circle, but like I said, my Phantoms fly "straight" about 15 degrees off and at the end of a run, drift about 20 feet to the left (at the beginning of a battery). I'm curious too....

As far as hovering accuracy, DJI states that the accuracy is 0.8 m and 2.5 m horizontal. That is 2.62 feet vertically and 8.2 feet laterally.

So that means yours Phantom can wander around inside a cylinder that is 5 meters across and 1.6 meters tall (16 feet across x 5.25 feet tall) and still be in spec. That is the stated accuracy.

 

[ianwood]

I live in southern California. The declination here is around 12 degrees. I have not had any issues, compass related or otherwise.

So you have a P2x and when you start with a new battery take off and fly straight out only using elevator, it flies perfectly straight and doesn't hook to the left when you let go?

 

[xtonex]

I live in southern California. The declination here is around 12 degrees. I have not had any issues, compass related or otherwise.

You might wanna go recheck and look for it... your in a high declination area

 

[SilentAV8R]

I live in southern California. The declination here is around 12 degrees. I have not had any issues, compass related or otherwise.

So you have a P2x and when you start with a new battery take off and fly straight out only using elevator, it flies perfectly straight and doesn't hook to the left when you let go?

That is exactly what I am saying. I take off, establish a stable hover and push forward on the "elevator" stick. Flies straight out. I have tested this in zero wind and not had an observable issue. Mine operates within the stated limits of the GPS accuracy.

 

[SilentAV8R]

I live in southern California. The declination here is around 12 degrees. I have not had any issues, compass related or otherwise.

You might wanna go recheck and look for it... your in a high declination area

As I have said, mine operates within the stated accuracy of the GPS. It appears to me that you all want it to stop instantly with zero movement. That is beyond the capabilities of the GPS. The best it can do is to stay within 2.5 meters (8.2 feet) of where you stop laterally and 0.8 meters (2.6 feet) vertically

I know some of you are saying "Ah-Ha!! He has the issue and does not even know it". My response is that I know the limitations of the equipment and I do not expect it to exceed those inherent design limitations.

 

[ianwood]

As far as hovering accuracy, DJI states that the accuracy is 0.8 m and 2.5 m horizontal. That is 2.62 feet vertically and 8.2 feet laterally.

So that means yours Phantom can wander around inside a cylinder that is 5 meters across and 1.6 meters tall (16 feet across x 5.25 feet tall) and still be in spec. That is the stated accuracy.

TBE is not related to the accuracy of the hover. It is a specific behavior as described a few posts up. The accuracy of the DJI hover has to do with drift caused by GPS shifts, IMU precision and the corrective ability of the FC.

As I see it, TBE typically occurs in one of two forms:

Larger error TBE is when the compass heading and GPS course differ enough that the corrective action of the FC actually moves the Phantom further away from where it is trying to put itself. This is manifested in TBE that grows in circumference and requires intervention to stop.

Small error TBE is when the compass heading differs enough from the GPS course that the corrective action misses the target but it is still close enough that it brings the Phantom closer to the target. This results in corrective actions that close in on the target in a diminishing radius until it is close enough to stop correcting or the wind disrupts the radial nature of it. This is the kind of TBE that reduces in size and stops on its own.

I just made both of those terms up but the distinction is valid.

 

[ianwood]

That is exactly what I am saying. I take off, establish a stable hover and push forward on the "elevator" stick. Flies straight out. I have tested this in zero wind and not had an observable issue. Mine operates within the stated limits of the GPS accuracy.

This has very little to do with GPS accuracy. It has to do with a conflict between GPS course and compass heading. If what you say is true, yours would be the second anomaly. And by anomaly, I mean that everyone but you and one other has the issue in the affected areas.

If your Phantom is indeed immune to the compass disease, we need to find out why as that could tell us what needs to be fixed. Can you post a video? It should be at the beginning of flight, Phantom taking off, flying out straight (ideally towards a reference point) and then releasing the elevator and allowing it to settle on its own.

Also if you can post a screenshot of your firmware page. When did you buy your Phantom?

obiwan, if you can have your friend do similar. The goal is to validate that these are true anomalies and if they are, we need to use them to further diagnose and fix the issue.

 

[xtonex]

I live in southern California. The declination here is around 12 degrees. I have not had any issues, compass related or otherwise.

You might wanna go recheck and look for it... your in a high declination area

As I have said, mine operates within the stated accuracy of the GPS. It appears to me that you all want it to stop instantly with zero movement. That is beyond the capabilities of the GPS. The best it can do is to stay within 2.5 meters (8.2 feet) of where you stop laterally and 0.8 meters (2.6 feet) vertically

I know some of you are saying "Ah-Ha!! He has the issue and does not even know it". My response is that I know the limitations of the equipment and I do not expect it to exceed those inherent design limitations.

Or maybe your not understanding whats actually happening... i can fly straight too. its when i let go of the stick after a straight flight forward it hooks to the right and i mean hooks well beyond the "stated gps accuracy" and then on the same path in reverse it hooks in the opposite direction. chances are you dont realize thats happening as many of us didnt at first (you correct it b4 it gets out of hand or maybe dont fly straight runs all that often). after a few mins it hooks somewhere around the stated accuracy. and then right b4 its time to land i have barely any hook at all. this is not a design or equipment limitation! Go try your course lock and tell us if its actually perfectly on course in the BEGINNING of the flight. im pretty sure you'll realize its not. Please show us your video of doing these things we speak of and make sure u do it in a very open space!

 

[ianwood]

Or maybe your not understanding whats actually happening... i can fly straight too. its when i let go of the stick after a straight flight forward it hooks to the right and i mean hooks well beyond the "stated gps accuracy" and then on the same path in reverse it hooks in the opposite direction. chances are you dont realize thats happening as many of us didnt at first (you correct it b4 it gets out of hand or maybe dont fly straight runs all that often). after a few mins it hooks somewhere around the stated accuracy. and then right b4 its time to land i have barely any hook at all. this is not a design or equipment limitation! Go try your course lock and tell us if its actually perfectly on course in the BEGINNING of the flight. im pretty sure you'll realize its not. Please show us your video of doing these things we speak of and make sure u do it in a very open space!

I'm 90% sure this is the case. Most people don't even know it's happening until they have a reason to fly straight (tight spots, lining up shots, etc.). And even then, many of us don't know what's normal until we get familiar with it.