Strange,sudden uncommanded left yaw and motion during a flight...

[gnirtS]

Wondering if any log file analysers can help with with this.

Standard drone flight (2nd of the day), perfectly normal behaviour for the first 10mins of the flight, light winds, remote area with no interference.

Just taking a normal video when the drone starts moving increasingly rapidly full left (flight direction, not yaw). I correct with yaw but no matter where the nose is pointing it continues to fly 90 degrees or so left.
No errors show on screen.
I point the nose back to the launch boat (180 degree roughly from direction of filming) and the left horizon is visibly tilted some 30degrees on the screen and still showing large left sideways flight.

I manage to crab it back towards the boat and descend (in case it was wind but the 180 degree heading change and still full left drift hints no), eventually the roll stops somewhat and am eventually able to hand catch it but for a 45 second or so period it was nearly totally out of control requiring large inputs just to stop it flying off.

The video is here where it started (problem persists until about 55 seconds):

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Accept third party cookies

The .TXT file log is here:-

Dropbox - File Deleted - Simplify your life

www.dropbox.com

The .DAT file log is here:-

Dropbox - File Deleted - Simplify your life

www.dropbox.com

AirData flight log:-


There are some not enough power ESC errors at roughly the time it started (nothing showed on screen that i could see) and these vanished.
Look at record numbers roughly 6200 to 6500 onwards for the onset of the problem. There is a high wind warning later in the event but im suspecting this is more of a false report due to the problem than actual wind.

A few points:-
- Compass was calibrated the day previously
- Compass and IMU showed all green and low errors prior to take off
- Second flight of the day within a few miles of that location
- I dont think it was wind because the direction of travel was always left even when the nose was pointed in a completely different direction
- There was a Mavic Pro in the air within a few hundred metres at all time and that didn't log any odd flight behaviour or wind warnings
- Winds were low (observed) and low according to AirData (in the single figures MPH range and no gusts)
- Props are all undamaged, drone hasnt been crashed etc
- Compass looks fine as there was no yaw associated with this event at all
- A new flight was conducted the next day near to that location and that had no issues at all

So, any ideas what the hell happened there? I very nearly lost the drone due to proximity of cliffs and not being about to initially get it heading back.

 

[MAvic_South_Oz]

I once flew my MP at the same location as my son was flying his P4.
No problems or me controlling mine, but his lost control slightly a couple of times, then totally came down without input, uncontrollable, and had a minor crash with no damage thankfully.
Somehow my system upset his, but not the other way around.

Not sure if the nearby MP could have affected ours, but just thought I'd share.
Hopefully your data above can shed some light.

 

[Meta4]

Wondering if any log file analysers can help with with this. ...
So, any ideas what the **** happened there?

can you make it a little easier?
I'm looking at a spreadsheet with 39 columns and 7200 lines of data and a flight pattern that's all over the map.
If you could point me to what part of the flight it would help a lot.

 

[gnirtS]

Look at record numbers roughly 6200 to 6500 onwards for the onset of the problem. There is a high wind warning later in the event but im suspecting this is more of a false report due to the problem than actual wind.

That bit.

 

[Meta4]

That bit.

It would be nice if you could pinpoint the time in the flight that you are concerned about.
From 6200-6500 you made 12 rudder movements, some of them max stick while you were pushing the right stick full forward.
Your flight data looks like this:

DJI Flight Log Viewer | Phantom Help

Upload and view flight logs from your DJI drones

www.phantomhelp.com

Where's the problem?

 

[gnirtS]

On the video the first full left motion starts at about 6 seconds (can be seen with me initially trying to adjust the rudder to keep the video centred before giving up and realising something was wrong).

Then after deciding to come back ASAP at 15 seconds in the video you can see the increasing left roll and flight as i do that (notice horizon tilt, none of that is RC commanded).
This continues until about 29 seconds into the video where i dial in enough opposite direction roll to stop the drift and then start gingerly working it back towards the boat for recovery.

This corresponds to pretty much data point 6060 onwards with 6217 showing the heavy left roll and horizon 90 degrees from the nose direct with no corresponding RC input.

6240 onwards shows the Y airspeed increasing steadily from the normal P-GPS max of 32mph up to 42mph.

That points towards a strong wind gust EXCEPT that i first observed the leftwards flight facing in totally the opposite direction to this (its a constant 90 degrees left of nose regardless of orientation). That and the other drone in the area wasn't affected and no wind was logged by the ship in excess of 8mph or so.

 

[Meta4]

I think I'll pass on this (I'm getting seasick).
It would be a lot easier to analyse if it was a simpler flight pattern.
The speed looks like a strong tailwind - I can't think of anything else that would cause that.
Having sailed around steep islands like that, I'm well aware that the wind experienced on a boat can vary a lot in that terrain with wind being funnelled through valleys at double (or more) the local wind strength .
In my part of the world they call that bullets.
Could it be that you encountered some bullets?

 

[gnirtS]

Im doubting its wind for a few reasons - the fact the movement was ALWAYS nose left 90 degrees independent of actual nose heading (a wind would be a specific direction and would change).
Also in the heavily banked part of the flight on the ocean the bank is in the direction of flight - if the drone was trying to fight wind then it'd be rolled into the wind not with it as is the case here.
Local wind strength is in the region of 5-6mph with an 8mph peak. No convective currents or swirling, no storm cells around. No wind observed on the trees, water or foliage on the cliffs before or during the event and nothing reported on the other mavic pro very nearby.
im almost leading towards a gyro upset or similar in this.

 

[BudWalker]

Wondering if any log file analysers can help with with this.

Standard drone flight (2nd of the day), perfectly normal behaviour for the first 10mins of the flight, light winds, remote area with no interference.

Just taking a normal video when the drone starts moving increasingly rapidly full left (flight direction, not yaw). I correct with yaw but no matter where the nose is pointing it continues to fly 90 degrees or so left.
No errors show on screen.
I point the nose back to the launch boat (180 degree roughly from direction of filming) and the left horizon is visibly tilted some 30degrees on the screen and still showing large left sideways flight.

I manage to crab it back towards the boat and descend (in case it was wind but the 180 degree heading change and still full left drift hints no), eventually the roll stops somewhat and am eventually able to hand catch it but for a 45 second or so period it was nearly totally out of control requiring large inputs just to stop it flying off.

The video is here where it started (problem persists until about 55 seconds):

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Accept third party cookies

The .TXT file log is here:-

Dropbox - File Deleted - Simplify your life

www.dropbox.com

The .DAT file log is here:-

Dropbox - File Deleted - Simplify your life

www.dropbox.com

AirData flight log:-


There are some not enough power ESC errors at roughly the time it started (nothing showed on screen that i could see) and these vanished.
Look at record numbers roughly 6200 to 6500 onwards for the onset of the problem. There is a high wind warning later in the event but im suspecting this is more of a false report due to the problem than actual wind.

A few points:-
- Compass was calibrated the day previously
- Compass and IMU showed all green and low errors prior to take off
- Second flight of the day within a few miles of that location
- I dont think it was wind because the direction of travel was always left even when the nose was pointed in a completely different direction
- There was a Mavic Pro in the air within a few hundred metres at all time and that didn't log any odd flight behaviour or wind warnings
- Winds were low (observed) and low according to AirData (in the single figures MPH range and no gusts)
- Props are all undamaged, drone hasnt been crashed etc
- Compass looks fine as there was no yaw associated with this event at all
- A new flight was conducted the next day near to that location and that had no issues at all

So, any ideas what the **** happened there? I very nearly lost the drone due to proximity of cliffs and not being about to initially get it heading back.

The posted .DAT (FLY023) doesn't match the .txt. FLY024.DAT is the correct .DAT

 

[sar104]

It wasn't wind - it looks like a compass or IMU issue. The DAT file will confirm which. The flight ran into problems at around 590 seconds, when the IMU velocities began to disagree with the flight path:



Looking in more detail - the GPS position data clearly show the aircraft begin to move left, but the IMU doesn't see that. There is already significant disagreement before the video starts at 607 seconds.



EDIT - I forgot to note that the track arrows are the recorded yaw value - i.e. the direction that the aircraft is recorded to have been facing.

 

[BudWalker]

Wondering if any log file analysers can help with with this.

Standard drone flight (2nd of the day), perfectly normal behaviour for the first 10mins of the flight, light winds, remote area with no interference.

Just taking a normal video when the drone starts moving increasingly rapidly full left (flight direction, not yaw). I correct with yaw but no matter where the nose is pointing it continues to fly 90 degrees or so left.
No errors show on screen.
I point the nose back to the launch boat (180 degree roughly from direction of filming) and the left horizon is visibly tilted some 30degrees on the screen and still showing large left sideways flight.

I manage to crab it back towards the boat and descend (in case it was wind but the 180 degree heading change and still full left drift hints no), eventually the roll stops somewhat and am eventually able to hand catch it but for a 45 second or so period it was nearly totally out of control requiring large inputs just to stop it flying off.

The video is here where it started (problem persists until about 55 seconds):

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Accept third party cookies

The .TXT file log is here:-

Dropbox - File Deleted - Simplify your life

www.dropbox.com

The .DAT file log is here:-

Dropbox - File Deleted - Simplify your life

www.dropbox.com

AirData flight log:-


There are some not enough power ESC errors at roughly the time it started (nothing showed on screen that i could see) and these vanished.
Look at record numbers roughly 6200 to 6500 onwards for the onset of the problem. There is a high wind warning later in the event but im suspecting this is more of a false report due to the problem than actual wind.

A few points:-
- Compass was calibrated the day previously
- Compass and IMU showed all green and low errors prior to take off
- Second flight of the day within a few miles of that location
- I dont think it was wind because the direction of travel was always left even when the nose was pointed in a completely different direction
- There was a Mavic Pro in the air within a few hundred metres at all time and that didn't log any odd flight behaviour or wind warnings
- Winds were low (observed) and low according to AirData (in the single figures MPH range and no gusts)
- Props are all undamaged, drone hasnt been crashed etc
- Compass looks fine as there was no yaw associated with this event at all
- A new flight was conducted the next day near to that location and that had no issues at all

So, any ideas what the **** happened there? I very nearly lost the drone due to proximity of cliffs and not being about to initially get it heading back.

My working hypothesis is that the M2 got caught up in the wind close to those cliffs. The tablet M2 tablet .DAT has wind calculations done by the FC. Please retrieve FLY024.DAT and post it.

Throw us a bone here. There are several .csv files to chose from, each having different row numbers. The fact that the posted .DAT was incorrect helped some as that reduced the number of .csv files that had to be looked at. The YouYube start corresponds to about 613 secs into the flight.

 

[sar104]

My working hypothesis is that the M2 got caught up in the wind close to those cliffs. The tablet M2 tablet .DAT has wind calculations done by the FC. Please retrieve FLY024.DAT and post it.

Throw us a bone here. There are several .csv files to chose from, each having different row numbers. The fact that posted .DAT was incorrect helped some as that reduced the number of .csv files that had to be looked at. The YouYube start corresponds to about 613 secs into the flight.

Winds would not explain the IMU velocity discrepancy though. And the wind direction was generally out of the NW - i.e. away from the cliffs.

 

[gnirtS]

Correct DAT is here:-

Dropbox - File Deleted - Simplify your life

www.dropbox.com

I dont think its wind purely because the drift was always in the same direction relative to the nose regardless of the direction the nose was pointing. To me that points more at an IMI/accelerometer issue.

 

[sar104]

It was a yaw error on the active IMU (IMU1):



That causes a sensor fusion error:

603.370 : 34902 [L-FDI]NS(0) FUSION(1): fault on , magn_heading_err_large

What's not clear is the cause of the error, since the pitch and roll values are good (hence IMU1 magnetic yaw is good) and the IMU1 rate gyros agree with IMU0.

 

[BudWalker]

It was a yaw error on the active IMU (IMU1):

View attachment 67672

That causes a sensor fusion error:

603.370 : 34902 [L-FDI]NS(0) FUSION(1): fault on , magn_heading_err_large

What's not clear is the cause of the error, since the pitch and roll values are good (hence IMU1 magnetic yaw is good) and the IMU1 rate gyros agree with IMU0.

But, the rate gyros don't agree, sometimes by 30°.


Also, it was 13 secs from the Yawerror to the fault seen in the eventLog stream. That seems long to me.

Why do you think that IMU1 is active?

 

[sar104]

But, the rate gyros don't agree, sometimes by 30°.
View attachment 67678

Also, it was 13 secs from the Yawerror to the fault seen in the eventLog stream. That seems long to me.

Why do you think that IMU1 is active?

Those are rate gyro degrees per second rather than degrees, and I think you are looking at noise and digitization effects. If you back out a little it looks better:



And if you integrate the rate gyro data to get degrees:



There are some differences, but they don't seem large enough to account for the 90° yaw error. Looking at the Z-axis gyro data specifically, the difference is only 11° when the yaw error is 88°.



IMU1 is active because the OSD_yaw value follows the IMU1 yaw, not IMU0 yaw:

 

[Meta4]

So, any ideas what the **** happened there? I very nearly lost the drone due to proximity of cliffs and not being about to initially get it heading back.

If the issue was a yaw error as mentioned above, you'd have to look at your boat launch technique.
Hand launch or launch from the deck?
You were very lucky it was a small one as many yaw errors are unrecoverable.

 

[sar104]

If the issue was a yaw error as mentioned above, you'd have to look at your boat launch technique.
Hand launch or launch from the deck?
You were very lucky it was a small one as many yaw errors are unrecoverable.

In this case it had nothing to do with the launch - and although this was a yaw error, it wasn't a compass error.

 

[Meta4]

In this case it had nothing to do with the launch - and although this was a yaw error, it wasn't a compass error.

How often do you see that?

 

[sar104]

How often do you see that?

Very rarely. It usually means it was an IMU problem - accelerometers or rate gyros, but I can't see enough disagreement between the IMUs to account for it. We have one compass and two IMUs interpreting the magnetometer data similarly - as expected since they agree on pitch and roll. But they disagree on yaw - IMU0 yaw agrees with the magnetic yaw computed by both IMUs, but IMU1 yaw diverges. That suggests

A bit of a mystery to me. Maybe @BudWalker will be able to figure out what I must have missed.