Another episode of un-commanded YAW to the left - ruined my timelapse

[Deleted member 94047]

I am getting frustrated by this. Took the MA for a morning spin. First battery spent making a timelapse of the sunrise. Second battery in, ascend to 120m, and I start taking interval pictures of the morning fog dissipating. I had what it seemed like an uncommanded YAW to the left on the ascent but I dismissed it as a possible slight push to the left of the left stick as I was pushing it up. I put the MA on hover, frame the shot, adjust the camera settings and start snapping away. Somewhere towards the third quarter of the flight, the MA starts yawing to the left; I counter with a push on the left stick to the right, the MA starts yawing again as soon as I let go of the stick; I counter again ..... this went on for about 30 or so seconds and the MA stabilized again. I just want to know what caused it. If anyone could take a look, here is the DAT file.

 

[silverdragon]

I understand your frustration. I go Through that with the Spark and Mavic Air. The MA did that one time and headed for a car valued @ about 80k and week before that it Yawed 45° at this $124k car. I won't fly at these meets anymore. Both times I had LOS on both but I was moving forward so I stopped in time in both situations.

 

[sar104]

This doesn't have a compellingly obvious explanation, as far as I can see. Looking at the IMU and magnetic yaw values:



Note that since the MA has only one compass, the two different magnetic yaw traces just represent the interpretation of the measured magnetic field by the two IMUs, based on their individual measurements of pitch and yaw. They agree, in this case.

The IMU yaw values are initialized correctly to the initial compass heading, but I cannot explain why IMU1 doesn't record the CW turn after takeoff. It is seen by both IMU0 and IMU1 rate gyros, and by the compass. Anyway - that leaves the IMU0 yaw approximately correct and the IMU1 yaw nearly 30° off. That situation persists until 640 seconds, when the uncommanded CW turn starts:



That maneuver is suggestive of the FC suddenly attempting to reconcile the IMU1 yaw by rotating the aircraft while holding the IMU1 yaw value constant. CCW rudder is applied before the maneuver completes, and both IMUs record that result. Over the next 20 seconds the IMU1 yaw converges with the IMU0 yaw.

I think that explains the immediate cause of the initial uncommanded yaw, but it leaves much else unexplained. @BudWalker?

 

[Deleted member 94047]

Thanks sar, I read that several times over (no, not because of the way you wrote it). I really appreciate you taking time to look at this. For what it's worth, it happened again in my first flight this morning. The AC started yawing to the left slightly while hovering and taking interval pjotos. And having read your post before heading out, I just let it be. The yaw was not as excessive as the one yesterday and it settled after just a few seconds. I returned it back to where it was before the yaw movement started and the rest of the flight completed without a hitch. My second flight after that did not have any issues.

So, to the important question, is there anything I can do? I recalibrated the RC sticks yesterday just in case. Will an IMU calibration help? Thanks a lot again

 

[silverdragon]

I must of mentioned at least 5 times of my un-commanded YAW MOVEMENT and there has been numerous times when other has mentioned it here on the forum and not to mention the amount of times DJI has got wind about the issue and not mention that there has been 2 updates since the issue. I wonder why DJI has not done anything about it.... ohhhh, I forgot to mention that I have the same issue with my SPARK....2 DJI Products with the above issue.

 

[Deleted member 94047]

Yes, there have been many posts about this here on this forum. I have also encountered the occasional yaw immediately after takeoff every now and then. But this is the first time I am encountering uncommanded yaw movements late into the flight. And you are absolutely right about DJI, they should get their act together and fix this.

 

[silverdragon]

Yes, there have been many posts about this here on this forum. I have also encountered the occasional yaw immediately after takeoff every now and then. But this is the first time I am encountering uncommanded yaw movements late into the flight. And you are absolutely right about DJI, they should get their act together and fix this.

You are right about the movement later in flight I am experiencing that myself now with the MA. The Spark still does it when I first take off with the height of 3 feet or so.

 

[sar104]

Thanks sar, I read that several times over (no, not because of the way you wrote it). I really appreciate you taking time to look at this. For what it's worth, it happened again in my first flight this morning. The AC started yawing to the left slightly while hovering and taking interval pjotos. And having read your post before heading out, I just let it be. The yaw was not as excessive as the one yesterday and it settled after just a few seconds. I returned it back to where it was before the yaw movement started and the rest of the flight completed without a hitch. My second flight after that did not have any issues.

So, to the important question, is there anything I can do? I recalibrated the RC sticks yesterday just in case. Will an IMU calibration help? Thanks a lot again

If my hypothesis was correct then the problem was caused by the strange behavior of IMU1 in response to yaw. It appeared to be seeing approximately the same inertial data as IMU0 but did not adjust the yaw accordingly, leading to the discrepancy that the FC later attempted to fix. I have no explanation for that and so it's difficult to suggest how to stop it from happening. You certainly could try an IMU calibration. There are other possible bandaid-type fixes, such as always taking off and then having the aircraft do a few full rotations to try to ensure that all the yaw values agree, but that's not a real fix.

 

[Deleted member 94047]

Great, will try those

 

[BudWalker]

I am getting frustrated by this. Took the MA for a morning spin. First battery spent making a timelapse of the sunrise. Second battery in, ascend to 120m, and I start taking interval pictures of the morning fog dissipating. I had what it seemed like an uncommanded YAW to the left on the ascent but I dismissed it as a possible slight push to the left of the left stick as I was pushing it up. I put the MA on hover, frame the shot, adjust the camera settings and start snapping away. Somewhere towards the third quarter of the flight, the MA starts yawing to the left; I counter with a push on the left stick to the right, the MA starts yawing again as soon as I let go of the stick; I counter again ..... this went on for about 30 or so seconds and the MA stabilized again. I just want to know what caused it. If anyone could take a look, here is the DAT file.

Don't mean to be picky but it might be significant. Your description is opposite the data. The data shows uncommanded Yaw to the right (CW) and then your response being to push the left stick to the right - not left.

 

[BudWalker]

This doesn't have a compellingly obvious explanation, as far as I can see. Looking at the IMU and magnetic yaw values:

View attachment 82021

Note that since the MA has only one compass, the two different magnetic yaw traces just represent the interpretation of the measured magnetic field by the two IMUs, based on their individual measurements of pitch and yaw. They agree, in this case.

The IMU yaw values are initialized correctly to the initial compass heading, but I cannot explain why IMU1 doesn't record the CW turn after takeoff. It is seen by both IMU0 and IMU1 rate gyros, and by the compass. Anyway - that leaves the IMU0 yaw approximately correct and the IMU1 yaw nearly 30° off. That situation persists until 640 seconds, when the uncommanded CW turn starts:

View attachment 82022

That maneuver is suggestive of the FC suddenly attempting to reconcile the IMU1 yaw by rotating the aircraft while holding the IMU1 yaw value constant. CCW rudder is applied before the maneuver completes, and both IMUs record that result. Over the next 20 seconds the IMU1 yaw converges with the IMU0 yaw.

I think that explains the immediate cause of the initial uncommanded yaw, but it leaves much else unexplained. @BudWalker?

It's not too surprising that IMU(1):Yaw didn't get updated when the MA rotated at around 37 secs. IMU(0) was active. I think we've seen this before.

I don't recall exactly where but I think we've also seen the situation where the switch from one IMU to the other began with reconciling a Yaw/magYaw separation on the inactive IMU. That would explain the behavior here. But, then there was no indication that the switch from IMU(0) to IMU(1) actually took place......

 

[sar104]

It's not too surprising that IMU(1):Yaw didn't get updated when the MA rotated at around 37 secs. IMU(0) was active. I think we've seen this before.

I don't recall exactly where but I think we've also seen the situation where the switch from one IMU to the other began with reconciling a Yaw/magYaw separation on the inactive IMU. That would explain the behavior here. But, then there was no indication that the switch from IMU(0) to IMU(1) actually took place......

I know that we've seen what looked like an IMU switch followed by an attempted yaw reconciliation, but even the non-active IMU should track attitude changes - so why do you think it's not surprising that it didn't?

As for the switch, I've still been unable to find a consistently reliable indicator of the active IMU either in the DAT data or the event stream.

 

[BudWalker]

I know that we've seen what looked like an IMU switch followed by an attempted yaw reconciliation, but even the non-active IMU should track attitude changes - so why do you think it's not surprising that it didn't?

As for the switch, I've still been unable to find a consistently reliable indicator of the active IMU either in the DAT data or the event stream.

I'll have to find it but I thought the Yaw/magYaw reconciliation took place in preparation for the IMU switch. I think that was the same one where the FC didn't keep the computed values current in non-active IMU.

In this one I looked at the IMU cnt fields to determine which IMU is active.

That's been a pretty reliable indicator of which IMU is active. But, I agree there have been situations where it isn't clear.

 

[sar104]

I'll have to find it but I thought the Yaw/magYaw reconciliation took place in preparation for the IMU switch. I think that was the same one where the FC didn't keep the computed values current in non-active IMU.

Maybe, but if the FC is watching the values from only one of the IMUs then one reason to have dual IMUs is defeated - the FC cannot compare them for consistency. And in almost all the logs I've seen the recorded IMUs agree. And in this case it's clear that IMU1 yaw is being updated - it just isn't recording all the rotations.

 

[Deleted member 94047]

Don't mean to be picky but it might be significant. Your description is opposite the data

I will look in to that. I remember it as a CCW yaw but I could be recalling it wrong. I was taking pictures every 5 seconds at the time so there should be enough to give an indication as to which direction the uncommanded yaw was. Will look into it.

Edit: You are right, it yawed to the right (CW). I remembered it wrong.

 

[BudWalker]

Maybe, but if the FC is watching the values from only one of the IMUs then one reason to have dual IMUs is defeated - the FC cannot compare them for consistency. And in almost all the logs I've seen the recorded IMUs agree. And in this case it's clear that IMU1 yaw is being updated - it just isn't recording all the rotations.

Here is the example that I was referring to.
Mavic 2 Pro with a mind of it's own!

For about 300 secs prior to the IMU switch the IMU(1):Yaw value wasn't being updated.


I suspect the FC is smart enough to prioritize the calculations that it does. The active IMU calcs go first followed by the non-active IMU calcs. If it manages to do all the non-active IMU calcs before it has to return to the active IMU calcs then the cnt field is updated. Just a theory...

 

[Tiago Miranda]

I've experienced the same issue after take off quite frequently . It was solved after an IMU calibration.

 

[sar104]

Here is the example that I was referring to.
Mavic 2 Pro with a mind of it's own!

For about 300 secs prior to the IMU switch the IMU(1):Yaw value wasn't being updated.
View attachment 82053

I suspect the FC is smart enough to prioritize the calculations that it does. The active IMU calcs go first followed by the non-active IMU calcs. If it manages to do all the non-active IMU calcs before it has to return to the active IMU calcs then the cnt field is updated. Just a theory...

Yes - I remember that one. But note that in that case, the IMU1 values were simply not being updated at all. In this case they are being updated, but apparently incorrectly. In fact when the initial uncommanded 30° CCW yaw occurred, IMU1 recorded a rapid 10° CW yaw followed by a rapid 10° CCW yaw.

 

[BudWalker]

Yes - I remember that one. But note that in that case, the IMU1 values were simply not being updated at all. In this case they are being updated, but apparently incorrectly. In fact when the initial uncommanded 30° CCW yaw occurred, IMU1 recorded a rapid 10° CW yaw followed by a rapid 10° CCW yaw.

View attachment 82060

Looks like the IMU(1):Gyro data is OK - it's just the Yaw computed from the gyro data that is incorrect.

Maybe, that's your point? The same kind of thing happens at 643 secs where it appears that gyro data isn't used to compute IMU(1):Yaw


I suspect that IMU(1) sensor data is OK. It's just the computed values that aren't valid whether they are missing or incorrect. I'm speculating that the IMU(1) Yaw/magYaw separation was being reconciled in preparation for a switch to IMU(1). Sounds great except for the fact that it didn't switch.

 

[PoonDog]

I am getting frustrated by this. Took the MA for a morning spin. First battery spent making a timelapse of the sunrise. Second battery in, ascend to 120m, and I start taking interval pictures of the morning fog dissipating. I had what it seemed like an uncommanded YAW to the left on the ascent but I dismissed it as a possible slight push to the left of the left stick as I was pushing it up. I put the MA on hover, frame the shot, adjust the camera settings and start snapping away. Somewhere towards the third quarter of the flight, the MA starts yawing to the left; I counter with a push on the left stick to the right, the MA starts yawing again as soon as I let go of the stick; I counter again ..... this went on for about 30 or so seconds and the MA stabilized again. I just want to know what caused it. If anyone could take a look, here is the DAT file.

Gremlins, a good spray with windex should kill them..