Can a faulty GPS module cause the drifting problem on Phantom 4 Pro? Drone is drifting east /or turning West in Autonomous missions

[Fly Dawg]

@timmydjr Per Bud's analysis above, a slight modification to the test mission proposed in Post #35. Add 2 additional Waypoint's to that plan. Waypoint 1 is only a starting point. After the aircraft reaches Waypoint 2 and stays for 10 seconds, proceed to Waypoint 3. At Waypoint 3, pause and take a Photo ( As shown below ) Set Waypoint 4 exactly the same, keeping the aircraft heading at 0°. Set no actions for Waypoint 5. Once this flight is completed, first it will eliminate the possibility of Pilot RC inputs to the aircraft. Second, we can look at the data from this flight to see if the magnetometer spikes still exist. Third, between Waypoint 4 and Waypoint 5, we should certainly be able to notice deviation in the aircraft flight direction. Should this change between Waypoint 2 and Waypoint 3, that will also be noticeable since there is no photo taken for that portion of the flight.
What do you think @BudWalker ?

 

[BudWalker]

@timmydjr Per Bud's analysis above, a slight modification to the test mission proposed in Post #35. Add 2 additional Waypoint's to that plan. Waypoint 1 is only a starting point. After the aircraft reaches Waypoint 2 and stays for 10 seconds, proceed to Waypoint 3. At Waypoint 3, pause and take a Photo ( As shown below ) Set Waypoint 4 exactly the same, keeping the aircraft heading at 0°. Set no actions for Waypoint 5. Once this flight is completed, first it will eliminate the possibility of Pilot RC inputs to the aircraft. Second, we can look at the data from this flight to see if the magnetometer spikes still exist. Third, between Waypoint 4 and Waypoint 5, we should certainly be able to notice deviation in the aircraft flight direction. Should this change between Waypoint 2 and Waypoint 3, that will also be noticeable since there is no photo taken for that portion of the flight.
What do you think @BudWalker ?

View attachment 108885

If I'm not mistaken in a Litchi waypoint mission the heading value for a waypoint is just a camera direction. It's not the direction of flight to the next waypoint

Maybe the explanation I gave in post #33 needs some help. There was no problem or anomalous behavior in FLY393. The P4P hit each Image Capture Point (ICP) exactly on target. If the path from the south to north traversal was bent or crooked that's because the mapping software laid out the ICPs that way. It's possible that the path from one ICP to the next may not have been exactly straight. That's to be expected since there is noise and things like wind to contend with. The direction of flight is constantly being updated to adjust for any errors. But, that's irrelevant. All that matters is to make it to the next ICP.

I doubt if the Litchi waypoints will have the same magnetometer spikes as the ICPs.

 

[Fly Dawg]

I doubt if the Litchi waypoints will have the same magnetometer spikes as the ICPs.

Maybe not, but if it is an aircraft issue with the magnetometer that may be causing the presumed drift don't you think that you would see the same sort anomaly with any autonomous flight ware? Looking at the log file, unless I missed a few there were 59 photo's taken in FLY393. 2 were failures. As you noted they were roughly 4-5 seconds apart. In the mission I proposed only uses a single photo at two waypoints. The reason I used 0° for this is that the street is practically facing due north. Mainly my thought was to eliminate an actual aircraft issue, during photos, if that spike causes some sort of anomaly and to see if it shows up during an automated easy mission with photo's involved. Granted, if this does indeed have an effect, it wont be that noticable with only two photos as far as "drift" goes. But if the spikes are there, that should be a clue that there may be an issue. I could of coarse be completely wrong, but that is what I would think. A photo, is a photo reguardless of the flight software used. The spikes should show up, if indeed that is an issue, wouldn't you think?

 

[Fly Dawg]

If I'm not mistaken in a Litchi waypoint mission the heading value for a waypoint is just a camera direction. It's not the direction of flight to the next waypoint

Just as a note as well. The Heading direction in Litchi is the direction of flight, and /or rotation of the aircraft. The gimbal angle can be changed to interpolate the pitch angles.

 

[BudWalker]

Maybe not, but if it is an aircraft issue with the magnetometer that may be causing the presumed drift don't you think that you would see the same sort anomaly with any autonomous flight ware? Looking at the log file, unless I missed a few there were 59 photo's taken in FLY393. 2 were failures. As you noted they were roughly 4-5 seconds apart. In the mission I proposed only uses a single photo at two waypoints. The reason I used 0° for this is that the street is practically facing due north. Mainly my thought was to eliminate an actual aircraft issue, during photos, if that spike causes some sort of anomaly and to see if it shows up during an automated easy mission with photo's involved. Granted, if this does indeed have an effect, it wont be that noticable with only two photos as far as "drift" goes. But if the spikes are there, that should be a clue that there may be an issue. I could of coarse be completely wrong, but that is what I would think. A photo, is a photo reguardless of the flight software used. The spikes should show up, if indeed that is an issue, wouldn't you think?

But, there wasn't any anomaly in FLY393, drift or otherwise. The magnetometer spikes didn't cause the drift because there was no drift.

 

[BudWalker]

Just as a note as well. The Heading direction in Litchi is the direction of flight, and /or rotation of the aircraft. The gimbal angle can be changed to interpolate the pitch angles.

direction of flight, and /or rotation of the aircraft
These are not the same. Apply some aileron and the AC will fly in one direction with the gimbal pointed in a different direction. I don't have a P4P but the way I read the specs the gimbal is aligned with the X axis.

 

[Fly Dawg]

These are not the same. Apply some aileron and the AC will fly in one direction with the gimbal pointed in a different direction. I don't have a P4P but the way I read the specs the gimbal is aligned along the X axis.

Agreed. Nor do I have a P4P. But from my flights with Litchi, which I use exclusively. The directional rotation, and/or flight direction is the heading direction. Aileron and Rudder aside, the flight should be in a near perfectly straight line, at least in an autonomous flight. It is not unusual at all in manual flight to give slight rudder with the throttle, or aileron with the elevator, but autonomously this should not happen. Assuming you give no manual inputs to the gimbal, it should be perfectly alligned with the front of the aircraft and should not vary, assuming your flight profile is correct and the gimbal is perfectly centered.

We may be discussing Apples and Apples actually. More info from the OP I think is needed. The data says one thing and the OP says otherwise. Personally and this is neither here nor there, but I am old school telemetry with analog and 16-32 channel chart recorders and magnetic tape for 128 channels........I am still working on deciphering the newer data....I really don't know why I said that......

 

[Fly Dawg]

@BudWalker Bud, I had some time today and went through every signal possible from the FLY393 Dat file, specifically looking for crosstalk, and found nothing. The spikes you referenced are only there on the magnetometer signals and no where else that I can find. The only thing I can think of is a bad ground reference someplace, but impossible to locate from the data. Is it possible that the quantity of these spikes could cause drifting? I agree that the data does not show any signs of that, but being that it is a part of the directional orientation, I am curious due to the sheer number of spikes, only when the camera is taking a still shot, could be related to the issue that the OP "states" that is occurring? I am kind of grasping at straws at the moment without additional input from the OP, and have not seen anything as of yet from @sar104 as to his opinion. At the moment I am at a complete loss as to where the origin may lie. This is quite intriguing to locate a possible source, but where??

 

[BudWalker]

@BudWalker Bud, I had some time today and went through every signal possible from the FLY393 Dat file, specifically looking for crosstalk, and found nothing. The spikes you referenced are only there on the magnetometer signals and no where else that I can find. The only thing I can think of is a bad ground reference someplace, but impossible to locate from the data. Is it possible that the quantity of these spikes could cause drifting? I agree that the data does not show any signs of that, but being that it is a part of the directional orientation, I am curious due to the sheer number of spikes, only when the camera is taking a still shot, could be related to the issue that the OP "states" that is occurring? I am kind of grasping at straws at the moment without additional input from the OP, and have not seen anything as of yet from @sar104 as to his opinion. At the moment I am at a complete loss as to where the origin may lie. This is quite intriguing to locate a possible source, but where??

Is it possible that the quantity of these spikes could cause drifting?
Just to be clear I'm assuming that you're referring to the Yaw drift seen iIn FLY393. There was no drifting off course. IMO the yaw "drift" isn't what I would call drift - it's a deliberate action by the FC to reconcile a Yaw/magYaw separation. And, it starts before the magnetometer spikes.

In the top graph the FC is holding the Yaw value (blue plot) constant. But, as can be seen from the mag data (green plot) and an integrated Z axis gyro (red plot) the P4P is actually rotating. The rotation continues until the magYaw value is equal (more or less) to the geoDeclination value.

This method of reconciling a Yaw/magYaw separation can be seen in many flights and several of the DJI platforms. In fact, it can be seen in FLY406.

 

[sar104]

Apologies for the rather delayed help here, but it's taken me a long time to reconcile the various observations. In the end I focussed on FLY406, although the same problem persists in all the flights, I think.

In 406 the OP described aligning the aircraft with the street, which is around half a degree east of north. As shown previously, that's visible in the record at 186 s. At 207 s he has the aircraft aligned, and at 210 s he applies forward elevator. The aircraft flies straight, but on a track of 15°.

The problem appears to be with compass 0, which is the compass in use in all these flights. While compass 1 records a correct heading of approximately 0°, compass 0 thinks that the aircraft is pointing on a heading of 15°, and so that is the direction that the aircraft flies, crabbing right while actually still facing due north.

The issue is likely caused by compass 0 not being properly calibrated to remove the aircraft's magnetic field. If you plot compass 0 against compass 1 you see the following:

Note the 15° offset at 0° and 180°, with the compasses agreeing at 90° and 270°. Looking at the delta as a function of yaw makes it clearer:

That 2π periodic behavior indicates a stationary magnetic field associated with the aircraft distorting the compass 0 measurements.

 

[Fly Dawg]

@sar104 You are correct. It is in FLY394 as well ~ 15° difference.

 

[BudWalker]

Apologies for the rather delayed help here, but it's taken me a long time to reconcile the various observations. In the end I focussed on FLY406, although the same problem persists in all the flights, I think.

In 406 the OP described aligning the aircraft with the street, which is around half a degree east of north. As shown previously, that's visible in the record at 186 s. At 207 s he has the aircraft aligned, and at 210 s he applies forward elevator. The aircraft flies straight, but on a track of 15°.

View attachment 108957

The problem appears to be with compass 0, which is the compass in use in all these flights. While compass 1 records a correct heading of approximately 0°, compass 0 thinks that the aircraft is pointing on a heading of 15°, and so that is the direction that the aircraft flies, crabbing right while actually still facing due north.

The issue is likely caused by compass 0 not being properly calibrated to remove the aircraft's magnetic field. If you plot compass 0 against compass 1 you see the following:

View attachment 108958

Note the 15° offset at 0° and 180°, with the compasses agreeing at 90° and 270°. Looking at the delta as a function of yaw makes it clearer:

View attachment 108959

That 2π periodic behavior indicates a stationary magnetic field associated with the aircraft distorting the compass 0 measurements.

I was speculating the compass discrepancy might be the cause of the P4P Yaw drift that several pilots have seen lately. Unfortunately, the most recent one that I know about has only the table .DAT which doesn't have the magnetometer data. Phantom 4 Pro V2 Fell out of the sky...

But, I did come across this .DAT
Dropbox - FLY231.DAT

It appears to have the same 15° discrepancy. It's an autonomous mapping mission with the magnetometer spikes coincident with the images being captured.

 

[sar104]

I was speculating the compass discrepancy might be the cause of the P4P Yaw drift that several pilots have seen lately. Unfortunately, the most recent one that I know about has only the table .DAT which doesn't have the magnetometer data. Phantom 4 Pro V2 Fell out of the sky...

But, I did come across this .DAT
Dropbox - FLY231.DAT

It appears to have the same 15° discrepancy. It's an autonomous mapping mission with the magnetometer spikes coincident with the images being captured.
View attachment 108974

The range of yaws recorded in that log is too small to be definitive, but it potentially looks like a similar situation.

 

[Newt59]

Greetings. I can deal with North Avenue Trade school folks. It's Bam's and Gators that get under my skin sometimes.

Yeah, well my trade school wasn’t even on North Avenue. I had hoped to transfer after Coosa Valley Tech Electronics Technology which came after a few years in the Navy. But I had 13 good years in the office equipment field and a lot of fun. Then went back to commercial construction. I think I bumped my head. And now I’ve been selling fertilizer and chemicals for 16 years. Still can’t figure what I was thinking on that one either.

And I work for Bobby Howard of Howard Fertilizer and Chemical. One of the original Bull Gators. So I have to be careful... UGA makes us print our invoices on plain paper for them because our logo is a gator. No joke. And I’m the only sales rep to wear a Ga Tech hat into the UGA golf course superintendents Office. One day he might let me back in. Maybe..

Bama and that other school over there with the identity problem... yep. Tough to tolerate.

 

[Blade4]

OP did you ever get this fixed? I am having same isssue

 

[Fly Dawg]

@Blade4 You might try and share your flight data on the thread you just created. It can be compared to this one to see if there is any relevance.

 

[timmydjr]

OP did you ever get this fixed? I am having same isssue

No I did not. As a matter of fact, I purchased another new P4P body and the same thing happened. So I have 3 P4P's that all do this. I would imagine that this is happening to many people and they simply do not realize it. It becomes very evident then you fly a perfectly straight line over a decent distance (ie .75 mile or more). When you run a "corridore" mission to follow the perfectly straight street, you can see that your photos are angled. the further you fly the more angled they get.

I come to the realization that 1. It is not going to be fixed. 2. it is most likely happening to more people who do not realize it. and 3. It does not seem to cause failures in the mapping process. I have to believe that it causes the maps to not be as good as they could be but I could not spend any more time on it and since no-one else thinks there is a problem, I decided to join them

 

[timmydjr]

@Blade4 You might try and share your flight data on the thread you just created. It can be compared to this one to see if there is any relevance.

You can free fly the P4P all day long with no issues. This only occurs with autonomous missions.

 

[Fly Dawg]

You can free fly the P4P all day long with no issues. This only occurs with autonomous missions.

I assume this is only occurring with Litchi? I have been flying Litchi for years and have never seen this type of issue before and still have not. Granted this is with a P3A, so I really don't see where the P4 would be any different. Although evidently it is.

 

[timmydjr]

I assume this is only occurring with Litchi? I have been flying Litchi for years and have never seen this type of issue before and still have not. Granted this is with a P3A, so I really don't see where the P4 would be any different. Although evidently it is.

Flydawg, this is using Pix4D or DroneDeploy, and I think Map Pilot. For mapping missions, you can not use Litchi. I use and love Litchi but I have not tested this. I would create a mission that follow a straight road about .75 miles then turn around and come back. I could have 4 legs up and down the road and see if there is any rotation.