Mavic Uncommanded wandering

[Jbourne210]

Morning y'all. I had something happen the other day, and it's got me a little spooked.

I did a quick test flight the other day (color management settings), and had my Mavic just start wandering around on me after a non aggressive turn. I never got an ATTI warning (or any other warning for that matter). It didn't freak out or anything, but it was definitely just wandering around up there for a few seconds. I've flown it twice since, and it hasn't repeated, but I was hoping some of you guru's would take a look at the vid and let me know what you think. I'm also attaching the associated TXT file (not sure if that's the best, or if I should extract the DAT?).
Details of the flight:

• Calm winds
• Good sat coverage (15 if I remember correctly) on take off
• Sun at my back
• Temps were within nominal limits

Here's the vid:

Also, I do remember that I got a video cache error, but that shouldn't be related.

I'm an old RC heli guy, so I'm okay with flying her in ATTI mode. And I'm not too concerned if it happens again, but I'd like to know why. I'd also like to know if I have an issue with my imu system as well.
Thanks in advance.

Take offs are optional. Landings are mandatory...

 

[Anthony Viscomi]

Here's your log:

DJI Flight Log Viewer - PhantomHelp.com

 

[Jbourne210]

Thank you Anthony. I didn't know that site existed, I'll be sure to add it to my list of tools.

Question though, what does the little black and white dot mean around the place that it started wandering? Any ideas?

Thanks again!

 

[Anthony Viscomi]

Thank you Anthony. I didn't know that site existed, I'll be sure to add it to my list of tools.

Question though, what does the little black and white dot mean around the place that it started wandering? Any ideas?

Thanks again!

@msinger can surely shed more light

 

[Jbourne210]

This pic shows where the 'wandering' happened....

 

[sar104]

Assuming that you are referring to the period from 130 - 150 seconds in the flight, the behavior is clearly visible in the pitch and roll data:





GPS was good and the aircraft was not in ATTI mode. The magnitude of those excursions was of the order of a couple of meters. AirData thinks it was wind gusts around 15 mph, but it's hard to tell for sure. The DAT file would give more information.

 

[Jbourne210]

sar104,
Thanks for that! Can I ask which program you used for the analysis?

Regarding the DAT files, I'm getting a warning indicating that the selected file types are not authorized to be uploaded. I realize that I can just change the extension to TXT, then upload them, but I don't want to bust any forum rules.

Suggestions?

You guys rock!

 

[sar104]

sar104,
Thanks for that! Can I ask which program you used for the analysis?

Regarding the DAT files, I'm getting a warning indicating that the selected file types are not authorized to be uploaded. I realize that I can just change the extension to TXT, then upload them, but I don't want to bust any forum rules.

Suggestions?

You guys rock!

The DAT file will also be too large to upload - you will need to use Dropbox or Google Drive and then post the link.

On the data analysis question - take a look at this thread where I just reposted a summary of the options:

Analyzing data / flight files

 

[Jbourne210]

Awesome, thanks sar!

 

[Jbourne210]

Okay, so I've uploaded the DAT files to my OneDrive account. I've actually loaded three files, as they all seem to have different times, all closely related in time.

Anyway, here's the link:

Mavic DAT Files

In the meantime, I'll be looking into the file analysis software you mentioned.

Thanks again

 

[BD0G]

Could this be a high KP Index issue?

 

[Jbourne210]

I thought so too, but when I saw it, and later saw the vid, it just seemed weird. I would think that if it was getting bad GPS data, it would make obvious commanded movements (even wrong ones). This just seemed like it was floating around. Like the IMU's had a brain fart. Just real lazy. Oh, and the video is real time. I didn't slow it down or anything... Strange...

 

[BD0G]

There is a new app from Polar Pro to determine what ND filter to use and it also provides important data for the area you are flying in or plan to fly in. Like Wind Speed (ground level of course) , Golden Hour Times , KP index , temperature , etc. All on one screen and readily accessible.

It also tells you what ND filter to use to get cinematic video.

Its called Polar Pro Filter app.

 

[Mossiback]

The video also shows that the gimbal appears to be locked, like it was in Fixed Wing mode.

 

[Jbourne210]

The video also shows that the gimbal appears to be locked, like it was in Fixed Wing mode.

Exactly. Weirded me out when it happened. As I said, I felt comfortable just watching her float, and when I told her to move she did. It was just odd. I'd hate for that to happen at more than 80 feet from me though...

 

[sar104]

Okay, so I've uploaded the DAT files to my OneDrive account. I've actually loaded three files, as they all seem to have different times, all closely related in time.

Anyway, here's the link:

Mavic DAT Files

In the meantime, I'll be looking into the file analysis software you mentioned.

Thanks again

Interesting. Whenever you see that kind of circling it tends to be indicative of yaw issues. And that is the case here:



The timing on the DAT file was different, so here the offending period is 190 - 210 seconds. As you can see, during that time the compass heading diverged, temporarily, from the IMU heading computed from the rate gyros. Not enough to cause any errors, but enough to confuse the FC for ten seconds or so. In this case the IMU heading was clearly correct (the aircraft was not rotating) which indicates a magnetic issue. I'm not sure what might be causing that - perhaps @BudWalker could give a second opinion.

 

[BudWalker]

I don't really have a compelling answer, just some rambling. Let me know if you find anything more useful.

For the benefit of others who might be reading this -
The magYaw signal is computed by DatCon and doesn't come directly from the Mavic's .DAT. It provides a heading value that's independent of the FC's Yaw value. Normally, magYaw and Yaw move together. It's clear the FC must use either magYaw or something similar to help determine the confidence in the Yaw value. The P3 will switch to ATTI when it sees a magYaw/separation. The Mavic, and other post P3 platforms, will either switch to ATTI and/or try to take corrective action to eliminate the magYaw/Yaw separation. It seems that's what happened here. Here is the magYaw/Yaw separation and the uncommanded wondering.


But what caused the magYaw/Yaw separation. magYaw is computed using magnetometer data but it has to be corrected for any pitch and roll. There is a well known formula for this used by DatCon and is also seen in the record definitions seen within the .DAT file itself. In this case the magYaw/Yaw separation appears to be correlated with roll. But, roll itself doesn't track totalGyro:X value.



So, I give up. Maybe the next thing to try is using totalGyro:X instead of roll when computing magYaw.

@Jbourne210 got all that? If you're interested you could do just an IMU calibration and then try it for a few flights. Do a compass calibration only if the problem persists. We're trying to make just one change at a time to know which change, if any, makes a difference.

 

[Nikon1]

I don't really have a compelling answer, just some rambling. Let me know if you find anything more useful.

For the benefit of others who might be reading this -
The magYaw signal is computed by DatCon and doesn't come directly from the Mavic's .DAT. It provides a heading value that's independent of the FC's Yaw value. Normally, magYaw and Yaw move together. It's clear the FC must use either magYaw or something similar to help determine the confidence in the Yaw value. The P3 will switch to ATTI when it sees a magYaw/separation. The Mavic, and other post P3 platforms, will either switch to ATTI and/or try to take corrective action to eliminate the magYaw/Yaw separation. It seems that's what happened here. Here is the magYaw/Yaw separation and the uncommanded wondering.
View attachment 28906View attachment 28907

But what caused the magYaw/Yaw separation. magYaw is computed using magnetometer data but it has to be corrected for any pitch and roll. There is a well known formula for this used by DatCon and is also seen in the record definitions seen within the .DAT file itself. In this case the magYaw/Yaw separation appears to be correlated with roll. But, roll itself doesn't track totalGyro:X value.
View attachment 28912
View attachment 28913

So, I give up. Maybe the next thing to try is using totalGyro:X instead of roll when computing magYaw.

@Jbourne210 got all that? If you're interested you could do just an IMU calibration and then try it for a few flights. Do a compass calibration only if the problem persists. We're trying to make just one change at a time to know which chanfe, if any, makes a difference.

That is one big post full of great explanation! Thank you - and extended Thanks to all the other “mad scientists” who help us to understand what’s occurring with our birds via these “dat” files.

 

[Jbourne210]

Wow....

**** good work there Budwalker! I've never experienced ATTI on the mavic, but I flew many early FC quads many moons ago, as well as some modded RC helis, and it reminded me of that. Just...wandering.

It's interesting you mention performing an IMU calibration. I always check sensor status prior to flight, and had noticed prior to THAT flight, that one of the IMU's was quite a bit higher that the others. Not out of the green, but higher nonetheless. So I performed a calibration, and this happened on the next flight. With all of the analysis performed, and the details you uncovered, I wonder now if I didn't get a bad calibration.

On that note, I have flown it a few times since, and haven't had an issue, but I still wonder why I didn't get an ATTI warning if, for all intents and purposes, the aircraft entered that mode....

Thanks for all y'alls help!!

 

[sar104]

Wow....

**** good work there Budwalker! I've never experienced ATTI on the mavic, but I flew many early FC quads many moons ago, as well as some modded RC helis, and it reminded me of that. Just...wandering.

It's interesting you mention performing an IMU calibration. I always check sensor status prior to flight, and had noticed prior to THAT flight, that one of the IMU's was quite a bit higher that the others. Not out of the green, but higher nonetheless. So I performed a calibration, and this happened on the next flight. With all of the analysis performed, and the details you uncovered, I wonder now if I didn't get a bad calibration.

On that note, I have flown it a few times since, and haven't had an issue, but I still wonder why I didn't get an ATTI warning if, for all intents and purposes, the aircraft entered that mode....

Thanks for all y'alls help!!

It didn't enter ATTI mode - it takes significantly more provocation in terms of IMU/compass discrepancies to do that. This was just a temporary divergence that was later reconciled.