Request advice: Compass error, Yaw error, Weak GPS signal

[JJErrico]

Nearly lost control of my P4A yesterday shortly after takeoff. Would appreciate help from fellow owners to understand the cause. Here's some pertinent facts:

• At 32 sec, I had VLOS to the AC approx 20 ft from Home point and 115 ft AGL.
• Message appeared "exiting GPS mode. Warning:Compass Error. Exit P-GPS Mode".
• Simultaneously and instantly, the GPS view displayed a straight line to a point approximately 1000 feet from the Home point. You have to see this on the video to believe it.
• The AC headed off on its own, seeming to pursue that point and to "get to it" in a hurry.
• It went from GPS_Atti to Atti mode.
• I lost VLOS at that point and wasn't sure if it was on a permanent trip.
• Never hit RTH since I wasn't confident the AC really knew where Home was at this point. Decided to fly manually.
• To get it Home, control was extremely difficult. Wind at ground level was <10 mph, but must have much higher at 200 ft. It drifted alot and required many tweaks. I have some tall trees nearby and had to inch my way between them and put it down on my deck.
• The trip back was plagued with messages including "Compass error", "Yaw error", and "Exiting GPS mode".
• The log indicates it was popping into Atti mode part of the time.

A video of DJI GO 4 playback is here...

Airdata shared link... Airdata UAV - Flight Data Analysis for Drones
Airdata UAV - Flight Data Analysis for Drones

Flight data log attached.

 

[Meta4]

Nearly lost control of my P4A yesterday shortly after takeoff. Would appreciate help from fellow owners to understand the cause. Here's some pertinent facts:

• At 32 sec, I had VLOS to the AC approx 20 ft from Home point and 115 ft AGL.
• Message appeared "exiting GPS mode. Warning:Compass Error. Exit P-GPS Mode".

Airdata isn't much use for investigating flight incidents.
It only shows a brief summary rather than the actual data.
To better see what happened, go to DJI Flight Log Viewer
Follow the instructions to upload your flight record.
Come back and post a link to the report it provides.

Looking at what you've written, something has caused a compass error soon after takeoff.
The first thing I'd ask about the surface you launched from.
Was it steel or reinforced concrete or close to a lot of steel?
Did you do any compass calibration?

When your Phantom has a compass error, it can't deal with the conflicting information it hets from the compass and the GPS.
To deal with this, it ignores the GPS info which means it drops into atti mode.
In atti mode it has no horizontal position holding ability and will drift with any wind.
RTH will be useless in that situation as it requires GPS.
The entire flight from 0:31 was in atti mode.

 

[sar104]

First obvious observations from the log. The cause of the yaw error at 31 seconds is not clear, but it resulted, as those always do, in the aircraft switching out of P-GPS to ATTI mode. At that point it was not trying to fly anywhere in particular - it was drifting on the wind which was 10 -12 mph out of the NNW.

To try figure out why the yaw errors occurred we would need to see the onboard DAT file for that flight.

How to retrieve a V3 .DAT File

 

[sar104]

Further inspection reveals that the initial yaw error is probably misnamed, and is due to a very poor GPS lock leading to inconsistent position and speed data at 31 seconds. The graph below illustrates the problem by the comparison between the FC speed reporting (green) and the speed implied by the rate of change of aircraft location.

 

[JJErrico]

Airdata isn't much use for investigating flight incidents.
It only shows a brief summary rather than the actual data.
To better see what happened, go to DJI Flight Log Viewer
Follow the instructions to upload your flight record.
Come back and post a link to the report it provides.

Looking at what you've written, something has caused a compass error soon after takeoff.
The first thing I'd ask about the surface you launched from.
Was it steel or reinforced concrete or close to a lot of steel?
Did you do any compass calibration?

When your Phantom has a compass error, it can't deal with the conflicting information it hets from the compass and the GPS.
To deal with this, it ignores the GPS info which means it drops into atti mode.
In atti mode it has no horizontal position holding ability and will drift with any wind.
RTH will be useless in that situation as it requires GPS.
The entire flight from 0:31 was in atti mode.

Thank you for taking the time to review and help me understand. I really appreciate this scrutiny.

I agree that the visualization part of Airdata is limited in diagnosing problems. However, in the original post, I attached the TXT log downloaded from Airdata via "Download CSV". That has 36 parameters recorded at 100 millisecond intervals. I believe sar104 did his plot with that data.

I have not calibrated the compass since unboxing. My understanding for P4A is that you calibrate only when the app asks you to. If anybody has a different take on that, I'd like to hear.

Several reviewers have suggested that I consider that the area I launched from may have magnetic disturbance issues. The takeoff point is on my deck and I've taken off from there about 20 times with no issues before yesterday 9/10. There is a barbeque grill about 20 feet away ! To try to eliminate the magnetic question, I did 2 short test flights today out in the middle of a soccer field. I did get the same set of errors again. This was flying at <20 ft AGL and within 200 feet of Home point. I will post one of those flight logs as it narrows the problem quite a bit.

 

[JJErrico]

First obvious observations from the log. The cause of the yaw error at 31 seconds is not clear, but it resulted, as those always do, in the aircraft switching out of P-GPS to ATTI mode. At that point it was not trying to fly anywhere in particular - it was drifting on the wind which was 10 -12 mph out of the NNW.

To try figure out why the yaw errors occurred we would need to see the onboard DAT file for that flight.

How to retrieve a V3 .DAT File

Thanks much for taking the time to look at my flight. Unfortunately, I did not have much luck with DJI Assistant 2...couldn't find it in the App store for either iPhone or iPAD. I do have DJI Assistant (no 2) installed on my iPhone but it didn't recognize the P4A.

 

[JJErrico]

Further inspection reveals that the initial yaw error is probably misnamed, and is due to a very poor GPS lock leading to inconsistent position and speed data at 31 seconds. The graph below illustrates the problem by the comparison between the FC speed reporting (green) and the speed implied by the rate of change of aircraft location.

View attachment 88053

sar104, this is a great analysis. Thank you. I've been wrestling with the chicken and egg problem as to whether a "compass error" caused the trip into Atti mode or if the "weak GPS signal" which is really no signal was the root cause and caused the AC to report the "compass error". On the Youtube video at 34 sec, you can see the app reports 11 satellites but the signal bars drop off to none. I assume the signal bars indication is what the drone supposedly sees as satellites locked and determines how many bars are displayed. Perhaps that's not what it's saying. I used to think a greater number of satellites locked in meant GPS was solid...not anymore.

 

[JJErrico]

Several reviewers have suggested that I consider that the area I launched from may have magnetic disturbance issues. The takeoff point is on my deck and I've taken off from there about 20 times with no issues before yesterday, 9/10. There is a barbeque grill about 20 feet away !

To try to eliminate the magnetic question, I did 2 short test flights today out in the middle of a soccer field. I did get the same set of errors again. This was flying at <20 ft AGL and within 200 feet of Home point. I've attached the flight data log for this short flight. The AC stayed in GPS_Atti mode for the whole flight even though there was reports of compass error and yaw error on and off. The app reported "weak GPS signal" and lost the signal bars despite the fact that the log shows it never had less than 15 satellites nor does it show "weak GPS signal". Not sure which data to trust.

I raised the question of when to calibrate compass, and another member, in a separate conversation, weighs in that this must be done and not wait for the app to direct it. What does this group think?

At this point, I think I'm obliged to re-calibrate and re-do the test flights. If I get similar results to today, then I think we'll have further narrowed it down (as not calibration-related). If not, I will have learned a great lesson about calibration and to not trust the app.

 

[BudWalker]

Thanks much for taking the time to look at my flight. Unfortunately, I did not have much luck with DJI Assistant 2...couldn't find it in the App store for either iPhone or iPAD. I do have DJI Assistant (no 2) installed on my iPhone but it didn't recognize the P4A.

The DJI Assistant can be found here
Mavic – Specs, FAQ, Tutorials and Downloads

In the absence of the .DAT file it would be helpful if you could provide the .txt for your flights. It may seem like the AD generated .csv is OK, but, there is a lot left out of that .csv that will be in the .txt.

 

[Meta4]

I agree that the visualization part of Airdata is limited in diagnosing problems. However, in the original post, I attached the TXT log downloaded from Airdata via "Download CSV". That has 36 parameters recorded at 100 millisecond intervals. I believe sar104 did his plot with that data.

I asked you to upload to Phantomhelp because it is very good for investigating and Airdata is next to useless for investigating.

I have not calibrated the compass since unboxing. My understanding for P4A is that you calibrate only when the app asks you to. If anybody has a different take on that, I'd like to hear..

That's good.
I asked to eliminate some potential causes of malfunction.

Several reviewers have suggested that I consider that the area I launched from may have magnetic disturbance issues. The takeoff point is on my deck and I've taken off from there about 20 times with no issues before yesterday 9/10. There is a barbeque grill about 20 feet away ! To try to eliminate the magnetic question, I did 2 short test flights today out in the middle of a soccer field. I did get the same set of errors again. This was flying at <20 ft AGL and within 200 feet of Home point. I will post one of those flight logs as it narrows the problem quite a bit.

That does sound like there may be a problem with the machine.
Please don't send Airdata info this time.

 

[JJErrico]

The DJI Assistant can be found here
Mavic – Specs, FAQ, Tutorials and Downloads

In the absence of the .DAT file it would be helpful if you could provide the .txt for your flights. It may seem like the AD generated .csv is OK, but, there is a lot left out of that .csv that will be in the .txt.

Thanks BudWalker. I successfully installed the Windows app and am copying the files now. Will provide. Meanwhile, I did attach Airdata-derived TXT files to my previous posts.

 

[BudWalker]

Thanks BudWalker. I successfully installed the Windows app and am copying the files now. Will provide. Meanwhile, I did attach Airdata-derived TXT files to my previous posts.

OK, glad you can provide the .DAT files. They will be large so you'll have to use something like Dropbox or GoogleDrive and post a link.

I saw the AD files. They really don't provide adequate information. Eg., the AD .csv does not contain gpsHealth (sometimes called gpsLevel). This is a much better indication of how good the GPS data is. numSats isn't really all that informative. The .txt file will contain the gpsHealth data.

 

[JJErrico]

OK, glad you can provide the .DAT files. They will be large so you'll have to use something like Dropbox or GoogleDrive and post a link.

I saw the AD files. They really don't provide adequate information. Eg., the AD .csv does not contain gpsHealth (sometimes called gpsLevel). This is a much better indication of how good the GPS data is. numSats isn't really all that informative. The .txt file will contain the gpsHealth data.

The .DAT files are here... P4A data files - Google Drive

Thanks for your help !

 

[Meta4]

Meanwhile, I did attach Airdata-derived TXT files to my previous posts.

As mentioned above ... forget Airdata it's not very helpful.

 

[BudWalker]

The .DAT files are here... P4A data files - Google Drive

Thanks for your help !

I've looked at both the .DAT files. I can only offer a half-baked theory; probably just make it more confusing. In summary, there seems to be something strange about the IMU. Please calibrate only the IMU so we can see if that makes a difference.

FLY071 has three instances of the following (FLY074 has one)
1. the eventLog stream reports
[L-FDI]ns req:fdi,0to0,reason:fusion.gps_yaw_err,result:fail for same index

2. a YAW_ERROR_LARGE error is declared

3. the eventLog stream reports
[L-FMU/LED]action changed. imu error:ns_abnormal(3)

4. gpsHealth drops to 0
Here is an example from FLY071

I don't know what to make of 1 and 3; the contents of the eventLog stream is always changing. But, suspect that gpsHealth doesn't really drop to 0; it's just set to 0 as a means to cause a switch to ATTI.

I'd be interested to see if calibrating the IMU will have any effect.

 

[JJErrico]

Thank you Budwalker. That will be my next step. I'll then repeat the safe low level test flight and report back.

 

[sar104]

I've looked at both the .DAT files. I can only offer a half-baked theory; probably just make it more confusing. In summary, there seems to be something strange about the IMU. Please calibrate only the IMU so we can see if that makes a difference.

FLY071 has three instances of the following (FLY074 has one)
1. the eventLog stream reports
[L-FDI]ns req:fdi,0to0,reason:fusion.gps_yaw_err,result:fail for same index

2. a YAW_ERROR_LARGE error is declared

3. the eventLog stream reports
[L-FMU/LED]action changed. imu error:ns_abnormal(3)

4. gpsHealth drops to 0
Here is an example from FLY071
View attachment 88069

I don't know what to make of 1 and 3; the contents of the eventLog stream is always changing. But, suspect that gpsHealth doesn't really drop to 0; it's just set to 0 as a means to cause a switch to ATTI.

I'd be interested to see if calibrating the IMU will have any effect.

Irrespective of the GPS health parameter - notice that the aircraft GPS location changes at a physically unreasonable rate after 31 seconds, which I think is conclusive evidence that the GPS data are incorrect. If there are other IMU issues then those would have to be in addition to that problem. But, since that problem alone could have caused the switch into ATTI, the question becomes - are there any other conclusive data that there were additional issues?

I haven't yet had a chance to look at the DAT files but the log indicated that the aircraft maintained a fairly steady heading of 62° over the initial period of the flight, so the magnetometer and gyro data should reflect that.

 

[BudWalker]

Irrespective of the GPS health parameter - notice that the aircraft GPS location changes at a physically unreasonable rate after 31 seconds, which I think is conclusive evidence that the GPS data are incorrect. If there are other IMU issues then those would have to be in addition to that problem. But, since that problem alone could have caused the switch into ATTI, the question becomes - are there any other conclusive data that there were additional issues?

I haven't yet had a chance to look at the DAT files but the log indicated that the aircraft maintained a fairly steady heading of 62° over the initial period of the flight, so the magnetometer and gyro data should reflect that.

Vel:GPS-H is a signal computed by DatCon that is GPS derived speed - the horizontal speed from the FC. Usually, it's close to 0.0. But, here it's not

I think this makes your conclusion more likely.

 

[JJErrico]

I presume this means we're not really expecting the IMU calibration will fix the root cause, but that it's still a good idea to do that and verify with a test flight to take it off the table. Am I reading you guys correctly?

 

[BudWalker]

I presume this means we're not really expecting the IMU calibration will fix the root cause, but that it's still a good idea to do that and verify with a test flight to take it off the table. Am I reading you guys correctly?

Yeah, that'd be good. It's not really clear what's causing what - at least to me it's not.