Mavic Pro doesn't fly right, crabs left.

[Danr-62]

That didn't change much - the periodic variation is still there.

View attachment 84679

Firstly - are you sure that you fully demagnetized the aircraft? I'm very surprised not to see a more significant difference, even if you hadn't properly fixed it.

Secondly - I suggest another flight test. Take off to 50 ft or so and then slowly rotate the aircraft (rudder only), first CW through two revolutions, then CCW. I'd like to see what the resulting DAT file shows.

The first time I ran the fixer on just the areas as directed at the fixer website for a mavic, then took a test flight. Flight I posted the data on.

Then I ran the fixer all over the dang thing, hit everything but the motors and camera gimbal. Then took it for another test flight and still no change.

I will do as you requested, see if I can do it a minute between rain showers today.

I would like to add that I was pleasantly surprised. The Cfixer I received has the look and feel of a quality built piece of equipment exceeding my expectations.

 

[Danr-62]

Two files one leading with left rudder, then the other leading with right rudder. Both files including those two full cw ccw revolutions you requested.

AC .dat

I hope I got em slow and steady enough, had to work kinda fast as had a very brief break in the rain. If not let me know and I'll doer again.

 

[BudWalker]

Two files one leading with left rudder, then the other leading with right rudder. Both files including those two full cw ccw revolutions you requested.

AC .dat

I hope I got em slow and steady enough, had to work kinda fast as had a very brief break in the rain. If not let me know and I'll doer again.

The problem appears to be in the Y axis magnetometer, probably the front one. When the M1 is facing west there is a noticeable difference between the two Y axis magnetometers. But, not when facing other directions. In particular, there isn't a noticeable difference when facing east.

The Z and X axis magnetometers don't have this problem.

I think this means there is something on the front right side of the M1 that is distorting the geomagnetic field. Can you tell us more about your calibration technique? Do you turn the M1 in your hands while remaining fixed? Or, do you hold the M1 fixed and rotate your body? If the later are your wrists locked such that a watch, etc effectively becomes part of the M1?

 

[Danr-62]

The problem appears to be in the Y axis magnetometer, probably the front one. When the M1 is facing west there is a noticeable difference between the two Y axis magnetometers. But, not when facing other directions. In particular, there isn't a noticeable difference when facing east.
View attachment 84690
The Z and X axis magnetometers don't have this problem.

I think this means there is something on the front right side of the M1 that is distorting the geomagnetic field. Can you tell us more about your calibration technique? Do you turn the M1 in your hands while remaining fixed? Or, do you hold the M1 fixed and rotate your body? If the later are your wrists locked such that a watch, etc effectively becomes part of the M1?

When I calibrate the compass I stand out in the yard well away from all power lines and metal.
I take everything out of my pockets, no metal zippers even take off my belt, I do the actual calibration several feet from the RC/tablet.

I rotate the drone only, holding it out in front of me as in I do not do the "calibration dance". I rotate the drone CCW as directed by the app.

I had a strobe on the nose if it but took it off, the problem started before the strobe was on it.
I only flew it once with it stuck there removed it rethinking placement.

 

[sar104]

Two files one leading with left rudder, then the other leading with right rudder. Both files including those two full cw ccw revolutions you requested.

AC .dat

I hope I got em slow and steady enough, had to work kinda fast as had a very brief break in the rain. If not let me know and I'll doer again.

The problem appears to be in the Y axis magnetometer, probably the front one. When the M1 is facing west there is a noticeable difference between the two Y axis magnetometers. But, not when facing other directions. In particular, there isn't a noticeable difference when facing east.
View attachment 84690
The Z and X axis magnetometers don't have this problem.

I think this means there is something on the front right side of the M1 that is distorting the geomagnetic field. Can you tell us more about your calibration technique? Do you turn the M1 in your hands while remaining fixed? Or, do you hold the M1 fixed and rotate your body? If the later are your wrists locked such that a watch, etc effectively becomes part of the M1?

Those data are pretty definitive - the two compasses also show a 2π (360°) periodic disagreement of the magnitude of the heading offset that you have been seeing. This is as classic a dataset of that effect that I've seen. Below are the data with a sinusoidal fit - the functional form that you would expect for an aircraft with an uncompensated magnetic field that is different at the two compass locations:



I think it is more generalized than just a problem with one axis. If you plot the difference between the x, y and z field compensated field components from the two magnetometers, then the x and y data have similar magnitude disagreements (as seen from the LS fit A values), with the z axis not far behind:



It still looks like a magnetization issue on the aircraft that either hasn't been calibrated out or is too large to calibrate out.

 

[Danr-62]

I can try some more with the Cfixer, that failing to provide results would you guys suggest replacing the #1 compass/GPU unit? Send it in for repair?

According to it it doesn't have a compass problem, degauss and or re-calibrations has made no real change in readout before or after Cfixer. I suppose it is possible I am reading that wrong?

 

[Danr-62]

If you guys would look at this one and let me know if you see any change at all, I'd really appreciate it.

AC .dat 293

It is another 50 odd feet up rotate twice cw and ccw file.

 

[sar104]

If you guys would look at this one and let me know if you see any change at all, I'd really appreciate it.

AC .dat 293

It is another 50 odd feet up rotate twice cw and ccw file.

You changed the phase of the disagreement, and the magnitude reduced slightly, but the basic problem hasn't changed.

 

[Danr-62]

You changed the phase of the disagreement, and the magnitude reduced slightly, but the basic problem hasn't changed.

View attachment 84732

View attachment 84738

Ok thank you very much now if I could impose on you one more time.

This link includes the same up cw ccw rotation as well as the a flight directly following it. I fly the northeast and southwest tracks again as before and more importantly? I also include direct north, south, east and west flight paths.

The rotate cw ccw file will probably tell the tail, I included the flight just in case it gives more to go on but doubt it.

Fly283 is the rotate cw ccw file and fly284 is of the flight that goes with it if you feel like looking at it, probably won't be necessary as 383 will reflect what the flight data will show.

AC .dat (2)

I feeling like I'm asking to much of you, taking up a lot of your free time. Well without going on and on etc I just wanted to say please know that I do not take it for granted.

 

[Danr-62]

Well I rolled the firmware back from .500 to .400, the result of the reduction above sar's post #28. A previous "refresh" had zero effect. Rolling back is the only thing that had any effect at all.
Then I updated back to .500 last night and I believe it is even worse now.

I flew it around a for a minute this morning, it was bad. I flew out to where it would fly straight at me and hit the RTH button, while taking a straight line as it should to HP the dang thing came in flying near sideways in flight it's right front prop leading the way.

I give up, sure wasn't for lack of trying and it appears as done everything we can for it.

At this point it looks like it is a hardware problem of some sort. I am sending it off to DJIdroneservice so they can have a look at it.

Many thanks to everyone that posted offering help and support, is very much appreciated.

 

[sar104]

Ok thank you very much now if I could impose on you one more time.

This link includes the same up cw ccw rotation as well as the a flight directly following it. I fly the northeast and southwest tracks again as before and more importantly? I also include direct north, south, east and west flight paths.

The rotate cw ccw file will probably tell the tail, I included the flight just in case it gives more to go on but doubt it.

Fly283 is the rotate cw ccw file and fly284 is of the flight that goes with it if you feel like looking at it, probably won't be necessary as 383 will reflect what the flight data will show.

AC .dat (2)

I feeling like I'm asking to much of you, taking up a lot of your free time. Well without going on and on etc I just wanted to say please know that I do not take it for granted.

I'll take a look at that later today. Before you give up - try one more sanity check. Take a compass - either analog or even a cell phone compass app - and run it over the aircraft. There should be essentially no detectable magnetic field associated with the aircraft.

 

[Danr-62]

I'll take a look at that later today. Before you give up - try one more sanity check. Take a compass - either analog or even a cell phone compass app - and run it over the aircraft. There should be essentially no detectable magnetic field associated with the aircraft.

I did that my post #10 in this thread, I found that the needle of a compass (an actual compass not digital) would react to the body of the mavic being near it. It would have to pass within 1/4" of the compass to move the needle.

Cfixer didn't really change that, nothing I would consider strong and any metal moves a compass needle to some degree. it should be able to account for it as part of itself during calibration?
I don't have another mavic for comparison value unfortunately.

I wanted to see how close the fixer had to be small piece of metal to have a solid effect, fully removing magnetization. I dragged a very small screwdriver shaft across a magnet making it magnetic.

Then I tired demagnetizing it with the fixer at decreasing distances starting from 2" out.
2, 1, 1/2, and then 1/4 of an inch away each in turn until it was successfully degaussed. It didn't demagnetize it until I laid the screwdriver shaft right against the fixer ring.

Based on that test the last file I posted "Fly283" in post #28, that is the result of me taking the top cover of the mavic off, removing compass 1 and laying it right on the Fixer and cycling it.
I also laid it directly on compass 2 while I had the the top cover off.

I could tell by the following flight that that doing so had no effect, still crabbing left.

I posted the FLY283 just to verify it post #29,

The only thing that did have any effect at all and verified per your findings post #28, was rolling the firmware back one version "FLY293" was the resulting file.

 

[Mossiback]

...I could tell by the following flight that that doing so had no effect, still crabbing left...

I originally suggested Cfixer because I misread your OP. Cfixer has been shown to correct a drone that does not fly straight (most state that it seems to drift to the right) or it slowly rotates in a hover. In extreme cases of a magnified part it will drop out of GPS mode due to conflicts between the compass and other systems. What I missed is that yours flies straight while pointed a few degrees off (crabbing). I apologize for that but if you keep your MP long enough you will probably need it. Mine still flies straight but will very slowly rotate a few degrees in a hover, depending on the direction it is pointed, and could probably use a degaussing.

 

[sar104]

I did that my post #10 in this thread, I found that the needle of a compass (an actual compass not digital) would react to the body of the mavic being near it. It would have to pass within 1/4" of the compass to move the needle.

Cfixer didn't really change that, nothing I would consider strong and any metal moves a compass needle to some degree. it should be able to account for it as part of itself during calibration?
I don't have another mavic for comparison value unfortunately.

I wanted to see how close the fixer had to be small piece of metal to have a solid effect, fully removing magnetization. I dragged a very small screwdriver shaft across a magnet making it magnetic.

Then I tired demagnetizing it with the fixer at decreasing distances starting from 2" out.
2, 1, 1/2, and then 1/4 of an inch away each in turn until it was successfully degaussed. It didn't demagnetize it until I laid the screwdriver shaft right against the fixer ring.

Based on that test the last file I posted "Fly283" in post #28, that is the result of me taking the top cover of the mavic off, removing compass 1 and laying it right on the Fixer and cycling it.
I also laid it directly on compass 2 while I had the the top cover off.

I could tell by the following flight that that doing so had no effect, still crabbing left.

I posted the FLY283 just to verify it post #29,

The only thing that did have any effect at all and verified per your findings post #28, was rolling the firmware back one version "FLY293" was the resulting file.

I ran that check on a Mavic Pro and Mavic 2 - I get pretty much no compass needle displacement at all from anywhere on the body of the aircraft - only the motors. Unmagnetized metal, even ferromagnetic metal, shouldn't affect a compass needle. If you are seeing a response from a compass then I would still suspect that magnetized components are the problem here.

 

[sar104]

I originally suggested Cfixer because I misread your OP. Cfixer has been shown to correct a drone that does not fly straight (most state that it seems to drift to the right) or it slowly rotates in a hover. What I missed is that yours flies straight while pointed a few degrees off (crabbing). I apologize for that but if you keep your MP long enough you will probably need it. Mine still flies straight but will very slowly rotate a few degrees in a hover, depending on the direction it is pointed, and could probably use a degaussing.

Both those behaviors may be caused by residual magnetism, so it's an appropriate fix in either case.

 

[Mossiback]

Both those behaviors may be caused by residual magnetism, so it's an appropriate fix in either case.

Thanks for that. I hate giving inappropriate wrong advice.

 

[Danr-62]

No amount of deguassing has an effect on this, the lower area of its tail end.

And a very small needle reaction to the forward area of its heatsink, almost nothing.

I have nothing to compare with so have no idea if that is "normal" or not.

 

[Danr-62]

I originally suggested Cfixer because I misread your OP. Cfixer has been shown to correct a drone that does not fly straight (most state that it seems to drift to the right) or it slowly rotates in a hover. In extreme cases of a magnified part it will drop out of GPS mode due to conflicts between the compass and other systems. What I missed is that yours flies straight while pointed a few degrees off (crabbing). I apologize for that but if you keep your MP long enough you will probably need it. Mine still flies straight but will very slowly rotate a few degrees in a hover, depending on the direction it is pointed, and could probably use a degaussing.

I don't regret buying the Cfixer at all, handy to have around.

That is the odd part of it at least in my mind, it does not continue to rotate during a hover.

I can put it in the air and then put the cross hairs on a tree and it will pretty much remain there. Very ever so slightly right and left due to wind but remain on target.

 

[Danr-62]

I ran that check on a Mavic Pro and Mavic 2 - I get pretty much no compass needle displacement at all from anywhere on the body of the aircraft - only the motors. Unmagnetized metal, even ferromagnetic metal, shouldn't affect a compass needle. If you are seeing a response from a compass then I would still suspect that magnetized components are the problem here.

Maybe the Cfixer I have doesn't work then or at least not as well as it should?

I used it on a paper clip, the compass needle has a slight reaction when the paper clip is move across in front of it. No change, same reaction before and after using the fixer.

I am powering the Cfixer with the charger for my P4.

 

[Danr-62]

I took it apart and checked it was/is the rear arm spring pivot assemblies that cause the compass needle to move, nothing else on the body has that effect on it.

No amount of using the fixer on them completely removed needle reaction but did reduce it some, positioning the pivots so as to slide into the hole in the fixer body.

I'm not sure if it made any difference or not. I take it out and do a couple rotations each way and post the file. Perhaps it will show a change?

Anyone have a good arm with spring pivot assy that could run past a compass needle a minute for me!