Finally it's my Mini's turn - flew straight into a wall and crashed

[new2mavic]

I thought it wont happen to me and I was totally wrong.

Flew it at the usual spot, between tall buildings, weak GPS signal. Never had any problem in the past but today it suddenly banked to the left while hovering and accelerated towards a wall. I applied full right aileron just before it hit the wall but it was too late ?.

I could not see any abnomality in the flight log data ( attached ) except for the big discrepancies between the gyro / vio yaw and the magnetic yaw. The difference was about 180 degrees .

My big question is this discrepancy happened all the time because of the steel in the surrounding buildings. The Mini was able to cope with it perfectly in all previous flights at this spot and in the first 1.5 minutes of this last flight but why it suddenly went crazy ?

 

[new2mavic]

OK, after examining the data on PhantomHelp. The Mini was shown to be banking to the left but flying to the right. The reality is that it flew to the left. This means the compass was wrong. The nose of the craft was pointing to NE when it happened or about 45 degree but yaw360 reads 232 degree so the compass was 180 degree off !

My question still is, with such a big compass error right from the beginning, how come it could behave so normally in the first 1.5 minutes ?

 

[slup]

My quick guess to that is that it wasn't until then the AC needed counteracting to maintain position ... there's there the "snowball started to roll" & the counteraction got bigger & bigger together with a position error that got bigger & bigger ... & yeah, the rest you already know.

Haven't time to dig in into the log right now, perhaps later ... but based on my guess you perhaps you self can check it off (as I know you have log reading skills )?

Sorry for your Mini

 

[new2mavic]

My quick guess to that is that it wasn't until then the AC needed counteracting to maintain position ... there's there the "snowball started to roll" & the counteraction got bigger & bigger together with a position error that got bigger & bigger ... & yeah, the rest you already know.

Haven't time to dig in into the log right now, perhaps later ... but based on my guess you perhaps you self can check it off (as I know you have log reading skills )?

Sorry for your Mini

I am afraid I am in such a deep sorrow that I will not be able to analyze objectively for quite some time ???

 

[Phantomrain.org]

My quick guess to that is that it wasn't until then the AC needed counteracting to maintain position ... there's there the "snowball started to roll" & the counteraction got bigger & bigger together with a position error that got bigger & bigger ... & yeah, the rest you already know.

Haven't time to dig in into the log right now, perhaps later ... but based on my guess you perhaps you self can check it off (as I know you have log reading skills )?

Sorry for your Mini

Its an excellent guess, and this is the Reason why everyone needs to understand that both compass errors and magnetic interference to your drone is like a sledge hammer to its brain.

Yes it will fly with the initial impact but the swelling is going to happen and it will no longer be able to think straight , this Is why its so important not only to move your drone to another take off point , but you must Restart the Drone , just because the Warning message went away on the controller the drone is still most likely compromised

Phantomrain.org
Coal

 

[Gkinghrn]

if the op had simply removed his fingers off the sticks would the drone have stopped trying to 'get somewhere that was wrong'? Newbie question I know but curious to understand more..

 

[steve_lyt]

could a gust of wind , not caught it. buildings can turn into a wind tunnel

 

[new2mavic]

could a gust of wind , not caught it. buildings can turn into a wind tunnel

There was no wind at all ....... It was just 20 meters away from me when it happened.

 

[sar104]

I thought it wont happen to me and I was totally wrong.

Flew it at the usual spot, between tall buildings, weak GPS signal. Never had any problem in the past but today it suddenly banked to the left while hovering and accelerated towards a wall. I applied full right aileron just before it hit the wall but it was too late ?.

I could not see any abnomality in the flight log data ( attached ) except for the big discrepancies between the gyro / vio yaw and the magnetic yaw. The difference was about 180 degrees .

My big question is this discrepancy happened all the time because of the steel in the surrounding buildings. The Mini was able to cope with it perfectly in all previous flights at this spot and in the first 1.5 minutes of this last flight but why it suddenly went crazy ?


View attachment 95352

This had nothing to do with steel buildings during flight. The compass error is at power up, resulting in the IMU yaw being initialized as 55° (NE) when it was actually -117° (SW). That's the classic compass error - the compass corrects once it takes off and leaves the location of the magnetic distortion, but the iMU yaw remains wrong by the amount of the original error. It's easier to visualize in unwrapped angle space:



That's why it is so important to verify that the aircraft arrow on the map is pointing in the correct direction before you take off.

The reason that the flight was stable for 90 seconds is that you also took off without a GPS position lock - navHealth was zero until 44 seconds, and the aircraft was in P-GPS under VPS control. In that mode it doesn't need a correct yaw value since it is navigating visually.



At 49 seconds it switched to GPS-based P-GPS, and now it is in trouble as soon as it starts to move. As you pointed out, there was very little wind and you didn't apply any further elevator or aileron inputs and so it hovered precariously until a disturbance finally caused it to move slightly at 88 seconds. At that point the FC attempted to correct for the motion, but in exactly the wrong direction due to the yaw error, leading to immediate uncontrolled flight, illustrated by both the attitude and velocity data:

 

[sar104]

if the op had simply removed his fingers off the sticks would the drone have stopped trying to 'get somewhere that was wrong'? Newbie question I know but curious to understand more..

No - it was trying to hold position, not go anywhere. The ~ 180° compass error is fatal in P-GPS mode.

 

[sar104]

My quick guess to that is that it wasn't until then the AC needed counteracting to maintain position ... there's there the "snowball started to roll" & the counteraction got bigger & bigger together with a position error that got bigger & bigger ... & yeah, the rest you already know.

Haven't time to dig in into the log right now, perhaps later ... but based on my guess you perhaps you self can check it off (as I know you have log reading skills )?

Sorry for your Mini

That's what happened, together with 45 seconds of initial stable flight in vision mode.

 

[new2mavic]

Thanks sar104. Very thorough analysis as usual.

Although the Mini hit the wall with such a big bang and fell some 15 meters before hitting concrete ground. It's still flyable !

The gimbal was broken into several pieces but still linked by wires. Miraculously, the camera is still working ! I replaced all the props and glued everything back in place. Now I have basically a working Mini with the gimbal in FPV mode.

I then tried to hover it inside my apartment. It was just as stable as it was before the crash. Much to my surprise, the yaw360 reading is about correct ! That leaves me wonder what had caused the fatal 180 degree compass error in the previous flight
....

 

[sar104]

Thanks sar104. Very thorough analysis as usual.

Although the Mini hit the wall with such a big bang and fell some 15 meters before hitting concrete ground. It's still flyable !

The gimbal was broken into several pieces but still linked by wires. Miraculously, the camera is still working ! I replaced all the props and glued everything back in place. Now I have basically a working Mini with the gimbal in FPV mode.

I then tried to hover it inside my apartment. It was just as stable as it was before the crash. Much to my surprise, the yaw360 reading is about correct ! That leaves me wonder what had caused the fatal 180 degree compass error in the previous flight ....

The compass error is caused by powering up the aircraft in a locally distorted magnetic field. It could be a nearby ferrous object, buried rebar in concrete, a vehicle, etc.. You can see on the yaw graph above that the compass corrects after climbing less than a meter or so from the takeoff surface, so it was very local. If you had looked at the orientation arrow you would have seen that it was wrong.

 

[new2mavic]

The compass error is caused by powering up the aircraft in a locally distorted magnetic field. It could be a nearby ferrous object, buried rebar in concrete, a vehicle, etc.. You can see on the yaw graph above that the compass corrects after climbing less than a meter or so from the takeoff surface, so it was very local. If you had looked at the orientation arrow you would have seen that it was wrong.

OK, I have checked the screen capture video and it did show that the orientation arrow was pointing to a very wrong direction. This is the moment before the crash. In reality the drone was generally pointing away from me but the arrow showed that it was pointing in the opposite direction :



The crash :

A lesson learnt the hard way .....

So if the craft was powered up near ferrous objects, there will be no chance for the error to get corrected for the entire flight ? Do higher-end drones like M2P behave in the same way ?

 

[sar104]

OK, I have checked the screen capture video and it did show that the orientation arrow was pointing to a very wrong direction. This is the moment before the crash. The drone was pointing away from me but the arrow showed that it was pointing towards me :

View attachment 95395


A lesson learnt the hard way .....

So if the craft was powered up near ferrous objects, there will be no chance for the error to get corrected for the entire flight ? Do higher-end drones like M2P behave in the same way ?

As far as I'm aware, only the Mavic 2 has been programmed to detect this kind of error. The Mavic 2 IMU yaw is initialized as usual from the compass, but if the magnetic yaw changes on takeoff without the rate gyros detecting any rotation, then the FC resets the IMU yaw to match the new compass heading. That only applies directly at takeoff, and it only works if the Mavic 2 is powered up and then not moved prior to takeoff.

This post shows the process in action:

Compass error demonstration

 

[slup]

OK, I have checked the screen capture video and it did show that the orientation arrow was pointing to a very wrong direction. This is the moment before the crash. In reality the drone was generally pointing away from me but the arrow showed that it was pointing in the opposite direction :

View attachment 95395

The crash :

A lesson learnt the hard way .....

So if the craft was powered up near ferrous objects, there will be no chance for the error to get corrected for the entire flight ? Do higher-end drones like M2P behave in the same way ?

I know it hurts now & your thought are filled with self accusations ... but my advise to you now is to slow it down & really think through & understand the indications you get instead of just hoping for the best & continue.

You have been in this situation for just over a week ago --> Mavic Mini sensor error ... when you took off during very similar circumstances, with magnetic interference & bad GPS reception, you barely slipped away that time.

All this said with the very best intentions ...

 

[new2mavic]

You have been in this situation for just over a week ago --> Mavic Mini sensor error ... when you took off during very similar circumstances, with magnetic interference & bad GPS reception, you barely slipped away that time.

In fact that flight was conducted at exactly the same spot - the drone took off from the balcony of my apartment. I got the sensor error warning in that flight but there was no problem at all.

In this flight there was not any warning message other than weak GPS. Still can't figure out why I got off in the previous flight ( and several others at exactly the same location ).

 

[sar104]

Still can't figure out why I got off in the previous flight ( and several others at exactly the same location ).

Pure luck.

 

[Prismatic]

Always hate to see someone lose their drone. ?

You were puzzled by the previous successful flights in the same place, between tall structures. I imagine you are taking off from an artificial surface. Concrete, in particular, often has a metal reinforcing mesh within it. This creates a grid of danger, but until now you had been lucky to launch from a "safe square".

But metal can be buried almost anywhere. In old war zones, industrial districts, and mining areas, there's iron and steel scattered everywhere, but I have seen it in unexpected places that don't have an obvious reason like that.

To assure that your flight will be safe, verify--before takeoff--that the drone is truly oriented as it's depicted on the map in the app. If it isn't, you must not launch. To do so practically guarantees disaster, as you've seen.

 

[new2mavic]

OK, just to make sure that I have really learnt it. I just did an experiment with the (crashed) Mini inside my apartment. I powered it on close to the floor , hovered it and check the yaw360 and magyaw figures in the .DAT file. Then I repeated but with the Mini powered on 1.5 m above the floor. The orientation of the drone was about the same in the two flights. Here are the results :



It is very obvious that powering the drone close to the floor can cause an error of about 180 degrees in yaw360 which I believe is used for position correction during flight.

I must have powered the Mini on close to the floor in today's flight but somewhere else in previous flights so I got off in the past but not today.