Spooky Action at a Distance

[jackgarner]

Hello! Had a similar issue where my mavic was returning to home with iffy connection to the remote when it lost its gps signal and switched to ATTI (still showing ~18 satellites!) It then proceeded to autoland itself and crash =(

I was hoping someone would be able to look at the .dat file from the drone to see if this is another instance of the faulty gps unit
@BudWalker do you think you could help me out? =)

https://1drv.ms/u/s!AhkHQqoTVHqJrTaKUG2Hqfo9lh3Y

 

[BudWalker]

Hello! Had a similar issue where my mavic was returning to home with iffy connection to the remote when it lost its gps signal and switched to ATTI (still showing ~18 satellites!) It then proceeded to autoland itself and crash =(

I was hoping someone would be able to look at the .dat file from the drone to see if this is another instance of the faulty gps unit
@BudWalker do you think you could help me out? =)

https://1drv.ms/u/s!AhkHQqoTVHqJrTaKUG2Hqfo9lh3Y

I took a look. In summary there was a gpsHealth problem during the RTH that caused a switch to ATTI. This, in turn, cancelled the RTH and the Mavic landed where it was (without returning to home).


Just for grins I looked at the final moments. It doesn't look like it hit that hard. Here is the accelerometer data.

The motors had all stopped but it looks the left front was still being commanded to turn. From the gyro data it looks like it might have flipped over.



How bad was it? What was the damage?

 

[jackgarner]

@BudWalker It's kind of a funny story... =)

I only had it for 2 days so I was still a little unsure of its intentions when I lost connection. I saw the autoland and chased it down with my car only to find that when I should have been in range I could not connect to it. So I headed back to the home point to see if it had somehow gotten past me - nope..

Round 2 - Driving back to the spot where it was last recorded and I regain connection. On my display, I see a note with an address of where it had landed! lol

When my mavic crashed it did flip as you said (I took a shot at combing through the .dat file myself using your program! Took me awhile to find something that would read the .dat from the drone =) it hit hard enough to knock the battery out! The person who found it somehow had the knowledge and the morality to put the battery in and turn my little drone back on so I could find it!

I am unsure of the height that it fell from and I was hoping there was some sort of ultrasonic / vision sensor data that I was missing but I think this will just remain a mystery (they did have a car in the driveway that I suspect might have been the culprit ~2.5 feet high). The damage is limited to a large dent in the back of the craft, stiff front left/right motors, and a full set of broken props

While it was flyable I sent it out to dji hoping they replace the unit due to the gps issue. I'd rather not chance it autolanding if I fly out of range again.

 

[BudWalker]

@BudWalker It's kind of a funny story... =)

I only had it for 2 days so I was still a little unsure of its intentions when I lost connection. I saw the autoland and chased it down with my car only to find that when I should have been in range I could not connect to it. So I headed back to the home point to see if it had somehow gotten past me - nope..

Round 2 - Driving back to the spot where it was last recorded and I regain connection. On my display, I see a note with an address of where it had landed! lol

When my mavic crashed it did flip as you said (I took a shot at combing through the .dat file myself using your program! Took me awhile to find something that would read the .dat from the drone =) it hit hard enough to knock the battery out! The person who found it somehow had the knowledge and the morality to put the battery in and turn my little drone back on so I could find it!

I am unsure of the height that it fell from and I was hoping there was some sort of ultrasonic / vision sensor data that I was missing but I think this will just remain a mystery (they did have a car in the driveway that I suspect might have been the culprit ~2.5 feet high). The damage is limited to a large dent in the back of the craft, stiff front left/right motors, and a full set of broken props

While it was flyable I sent it out to dji hoping they replace the unit due to the gps issue. I'd rather not chance it autolanding if I fly out of range again.

Looks to me like it didn't fall. Maybe it got caught up in something at the landing site. Here's some entries from the eventLog
689.121 : 3349300843 : 36980 [L-FMU/FSM]near ground
694.042 : 3371447508 : 37226 [L-FMU/MOTOR]safe_near_grd:0
699.366 : 3395405167 : 37492 [L-GPS]<GPS INFO>[monitor][0][0]:lce:0,sfe:0,dit:79,fe:2,dynseed 150 cnt 15000
700.720 : 3401498641 : 37560 [L-FMU/MOTOR]fault on , esc(3) is stall
700.722 : 3401507455 : 37560 [L-FMU/MOTOR] Stop. reason:motor stall near ground
The vpsHeight is 0.1 meters. The accelerometers and gyros are stable. Then at about 700 accelerometers and gyros act like one of the props caught something. If it hadn't been for that it would've landed safely, stop the motors and be waiting for you. Just my theory.

 

[jackgarner]

@BudWalker Wow I thought it would have to fall further than that to put such a large dent in the back! yikes haha

Too bad dji doesn't have the vps video recording to the internal sd alongside the sensor data. It would be nice to know what it hit and if I need to go back to pay for any damage I may have caused.

All and all, I am rather impressed with the vps landing ability - I was flying over a pretty tight area and it managed to find someone's driveway to land on using only vps. Obviously, it did crash but seeing that it was only ~10cm off the ground makes me think that it has something to do with the vortexes generated when landing / hovering so close to the ground.

Do you think DJI will warranty something like this? It isn't normal for the gps signal to disappear like this, is it?

 

[BudWalker]

@BudWalker Wow I thought it would have to fall further than that to put such a large dent in the back! yikes haha

Too bad dji doesn't have the vps video recording to the internal sd alongside the sensor data. It would be nice to know what it hit and if I need to go back to pay for any damage I may have caused.

All and all, I am rather impressed with the vps landing ability - I was flying over a pretty tight area and it managed to find someone's driveway to land on using only vps. Obviously, it did crash but seeing that it was only ~10cm off the ground makes me think that it has something to do with the vortexes generated when landing / hovering so close to the ground.

Do you think DJI will warranty something like this? It isn't normal for the gps signal to disappear like this, is it?

I don't have any special insight about what DJI will do. Except to say they seem to be reasonable and accept fault if DJI HW or SW contributed to an incident. That said DJI is a large company and you may not get a knowledgeable person on first contact. You may have to try a different channel if you don't at first succeed. Good luck.

 

[Mavicsailor]

Have there been any conclusions as to what causes GPS Health to drop while maintaining satellite count? (Causing it to drop into ATTI)

 

[Vlakvark]

I'm not the expert to really answer that, but, I haven't experienced a drop into ATTI since December 2016. I'll tend to ascribe it to a bug that has been removed by DJI in their recent updates. My total distance flown, since getting my Mavic early November, is currently reaching for 500km, without any physical mishap (I'm keeping my thumbs crossed)!


Sent from my iPad using MavicPilots

 

[storemaster]

@BudWalker
Hi, any chance you could also have a look on my logfile? Today experienced my firts gps outtakes while flying.
Had also some OA-messages during flight, but they could come from the sun I think. Also some birds were around. But the biggest problem was switching from GPs to ATII during flight, but I also had 15+ sattelites available.

logfile --> Dropbox - FLY028.DAT

p.s.: I'm on latest official FW.. v01.03.0400

Cheers,
Wolfgang

 

[BudWalker]

@BudWalker
Hi, any chance you could also have a look on my logfile? Today experienced my firts gps outtakes while flying.
Had also some OA-messages during flight, but they could come from the sun I think. Also some birds were around. But the biggest problem was switching from GPs to ATII during flight, but I also had 15+ sattelites available.

logfile --> Dropbox - FLY028.DAT

p.s.: I'm on latest official FW.. v01.03.0400

Cheers,
Wolfgang

I took a quick look. There isn't much to say except that the incident was caused by gpsHealth dropping to 1 from 5. This is just like all the other gpsHealth-drops-from-5-to-1 incidents. It's being discussed here and a few other threads.

 

[HaydenMR]

Props to @BudWalker for taking so much time to help everyone, and providing so much detail to those who don't have the means to do so.
Good to see someone here to help and not bash others.

 

[Strafe1]

I second that.

 

[Mavicsailor]

I second that.

The old saying "No good deed goes unpunished" comes to mind when I think of BudWalker. Folks ask for help because he has proven himself as a forum expert, and he just keeps helping. DJI needs to hire him as a lead engineer! If they are smart, they will listen to his analysis and suggestions.

 

[storemaster]

I took a quick look. There isn't much to say except that the incident was caused by gpsHealth dropping to 1 from 5. This is just like all the other gpsHealth-drops-from-5-to-1 incidents. It's being discussed here and a few other threads.
View attachment 7165

Hi, thanks for looking at it. I did a complete calibration of IMU, compass, and Vision sensor to be sure. Had 3 unenventful flights after that. Seems that a clean solid calibration did it for me. *knockonwoods*

 

[misskimo]

I looked at the .DAT for a incident that @erkme73 provided. I'm posting the results here since that incident is very similar to the ones being discussed in this thread. I'll describe what happened but since this is my first Mavic incident I can only speculate why it happened.

This was a ToiletBowling incident which happened here.
View attachment 1711


It appears to me that switching to Sport mode caused the Z axis magnetometer to become noisy.

The magZ noise is actually pretty severe with a minimum of 35 and a maximum of 1800. I've not seen anything like this on the P3.

DatCon has a computed field called magYaw that is the heading computed from the magnetometers and is different than Yaw which comes from the Flight Controller. Usually, magYaw is the same as Yaw. When the Z magnetometer became noisy magYaw began to separate from Yaw.

The Yaw/magYaw separation had been 5 degrees but slowly grew to 30 degrees at 850 secs when ToiletBowling was obvious.

With the P3 this amount of Yaw/magYaw separation often leads to a nonGPSCause error (presented as a compass error in the DJI Go App) and a switch to ATTI mode. In this incident there was no switch to ATTI mode. I'll speculate that the ToiletBowling actually started at 823 but didn't become apparent until 848 secs when the pilot removed control stick input. At this point the Mavic should've just hovered without stick input but was moving (ToiletBowling).

I'm speculating the magZ noise was caused, at least partly, by the switch to Sport mode. I've asked @erkme73 if the other ToiletBowling incidents happened only after a switch to Sport mode.

A compass calibration issue can be seen by looking at the behavior of Yaw and magYaw. There was no indication of a compass calibration issue in this flight.

The .csv used here has been attached here as a .zip for those interested in looking this incident.

Wow. Im reading this
Got less than an hour on my mavic
3rd flight normal mod. Went out to 1800 m and back. No problems even inside my car because its freaking cold in the arctic
Then switched it to fly fast. Lost gps
Starts flying off in the wind. And never switched it to Sport mode. Blamed it on my car.
A days later. Tried it again in sport mode. Went up 30 m. Pushed forward on the stick
It was listed to the left and front by 20° and going that way then no GPS

 

[misskimo]

Also noticed while in course lock mod when hovering its starts going around and getting bigger to around 50'

 

[ScrappyMavic]

I took a look at the RTH ATTI event from @jackgarner and the ATTI event from @storemaster. One thing not mentioned here, and that I'd like to highlight @BudWalker, is that the drop in gpsHealth from 5 to 1 is preceded by > 30 deg magYaw to Yaw separation and roll/pitch oscillations in both cases. Therefore again what is happening is that there is no issue with GPS position sensing at all, no issue with gpsHealth. The problem lies with the flight controller's Yaw estimate not matching reality. Attempting to do 2D X-Y positioning based on GPS data relies on a fairly accurate Yaw estimate. Too many degrees of error (~20-30) and the control loop goes unstable in two dimensions. (typically TBE) Since the flight controller really doesn't know if it's a GPS problem or Yaw problem, it just drops gpsHealth to 1 for safety.

The way I think it works can best be described in terms of frequency response. The compass is lowpass filtered and has a very small gain when applied to FC Yaw estimate. This can been seen as taking 10's of seconds to have significant impact on aircraft. The gyros are very fast but have low freq DC drift. So, the gyros are high pass filtered and have a large gain when applied to FC Yaw estimate, and work on the ms timescale. Due to the large number of reported ATTI events, TBE's, etc., there is obviously some systemic weakness in the Mavic. My personal theory is the early rear compass noise issue lead DJI to reduce the compass gain and/or bandwidth even more than usual for the Mavic to cope with the high levels of noise. This could lead to more magYaw to Yaw divergence with 1) fast maneuvers 2) gyro error/drift 3) poor compass calibration. Since most Mavic's now have the rear compass noise fixed, I'm hoping a future firmware update will return the compass gain/bandwidth to normal.

jackgarner issue with ATTI mode during RTH



storemaster issue with ATTI out of the blue

 

[misskimo]

So this problem can be fixed by downgrading?
Im with you on the yall and gps computing getting messed up like my ole dell with a virus! When I flew mine yesterday and it went nuts as soon as I got to 30 or so meters and flipped it in sport mode and pushed the stick forward looking at the screen on my phone and noticed I wasnt going the way it was pointed , looked out the window at the mavic and it was at full right and full forward then the GPS locked out . Didnt like my comands but got it back . Also noticed a delay in right stick comands. Like a half a second it would respond with full tilt and motor over responding
Oh. Been flying electric helicopters for 17 years

 

[BudWalker]

So this problem can be fixed by downgrading?
Im with you on the yall and gps computing getting messed up like my ole dell with a virus! When I flew mine yesterday and it went nuts as soon as I got to 30 or so meters and flipped it in sport mode and pushed the stick forward looking at the screen on my phone and noticed I wasnt going the way it was pointed , looked out the window at the mavic and it was at full right and full forward then the GPS locked out . Didnt like my comands but got it back . Also noticed a delay in right stick comands. Like a half a second it would respond with full tilt and motor over responding
Oh. Been flying electric helicopters for 17 years

Would it be possible for you to retrieve the .DAT file for this flight so we can take a look? For info on how to retrieve it look here. It'll be large so you'll need to Dropbox it and post a link. Also, since there was a delay in the stick commands could you retrieve the .txt for the flight. Since it's a .txt file it can be attached to a post without having to .zip it.

 

[misskimo]

Ok. Doing it now. Thanks
Sure sucks worrying when its ganna do it again! And still debating asking DJI to look at it. Then get a big bill and a old used mavic