Mavic Air taking off on its own after manual landing

[airsnobu]

Can someone please help me understand what happened during the 3 landing attempts of this flight?

DJI Flight Log Viewer | Phantom Help

Upload and view flight logs from your DJI drones

www.phantomhelp.com

I landed my Mavic Air 1 manually, and, while keeping the left stick downwards for engine cut off... the drone decided to spin up the engines instead and fly away. For a few seconds there i could not control the aircraft, trying to put the drone down resulting in erratic behavior. I then regained control and attempted landing #2, which had the same outcome, it landed for 1 or 2 seconds and then flew away again, in both occasions the propellers never stopped spinning. I then tried RTH which i had to immediately cancel since it was heading straight for some power lines. On my 3rd manual attempt i crashed it into a window, broke 3 props, yet surprisingly, the drone is unharmed. These birds are tough!

This isn't the first time this landing-flyaway drama has happened to me, happened twice before but managed to land it safely on the second attempt each time. I guess a crash was imminent since hope seems to have been my only strategy.

I looked at the settings and Landing Protection seems to be off, downward sensor (VPS) is On.

Could it be a problem with the downward sensor? Could it be my controller giving bad inputs?

I tried reading the flight log but i can't really draw any meaningful conclusion. I would be more than grateful if someone more experienced in flight forensics could take a look. I am terrified to fly this thing again until i understand the root cause of this erratic landing behavior, just not to hurt someone inadvertedly.

Many thanks.

 

[Yaros]

I'm no expert at logs so I'll leave this to others, but I can see that the drone went into ATTI mode because it lost GPS, and flew away as it didn't have GPS, it was still perfectly controllable, but you didn't make any inputs with the controller to manually prevent the flyaway.

 

[RodPad]

Welcome to the Forum!
The flight viewer link works for me, so much new stuff I can't catch up.

I'm surprised no members have replied yet. There is like 12+ members that is good with this stuff, REALLY GOOD.

For you to post a link back to this thread instead of posting duplicate stuff in many places

Rod ..

Added: I hit the save button and I got an alert!

 

[airsnobu]

I'm no expert at logs so I'll leave this to others, but I can see that the drone went into ATTI mode because it lost GPS, and flew away as it didn't have GPS, it was still perfectly controllable, but you didn't make any inputs with the controller to manually prevent the flyaway.

Ignore the ATTI part at the end, that's the crash. The crash i understand, what i don't understand is the two flyaways before it.

 

[Yaros]

Ignore the ATTI part at the end, that's the crash. The crash i understand, what i don't understand is the two flyaways before it.

I can't spot the flyaway, I'm not very savvy when it comes to log inspection, but I can see that during the auto landing you are pushing the throttle up preventing it from landing. You should have cancelled the landing, moved the drone to an appropriate position, then trigger landing again.

Also, the Mavic Air 1 is pretty prone to flyaways, in my experience I had 2 of them in about 5 hours of flying. Make sure to calibrate the Compass (and maybe the IMU) to avoid any further problems.

 

[Meta4]

I landed my Mavic Air 1 manually, and, while keeping the left stick downwards for engine cut off... the drone decided to spin up the engines instead and fly away.

The reason the motors didn't shut off was that the flight data doesn't show any time where you held the left stick fown for the three seconds that's required to shut down the motors.
The data shows the left stick being held down for <2 seconds, then briefly not held fully down, then held down again for >2 seconds several times.

 

[Meta4]

Also, the Mavic Air 1 is pretty prone to flyaways,

It's not
What you are calling a "flyaway is just a flight incident for which you don't know the cause.

Make sure to calibrate the Compass (and maybe the IMU) to avoid any further problems.

That's completely unnecessary and won't do a thing to prevent problems or "flyaways".

 

[airsnobu]

The reason the motors didn't shut off was that the flight data doesn't show any time where you held the left stick fown for the three seconds that's required to shut down the motors.
The data shows the left stick being held down for <2 seconds, then briefly not held fully down, then held down again for >2 seconds several times.

Thanks for pointing this out. Interesting find. I was holding down the left stick while watching in shock how the motors revved up instead of down. Strange how this gets reflected in the logs but i never let go of that left stick pointing down.

 

[Meta4]

This isn't the first time this landing-flyaway drama

In each case, the reason your drone flew away after each failed landing attampt was because you pushed the right stick forward and flew it away.

 

[Meta4]

Thanks for pointing this out. Interesting find. I was holding down the left stick while watching in shock how the motors revved up instead of down. Strange how this gets reflected in the logs but i never let go of that left stick pointing down.

The data doesn't show any period where the left stick was fully down for >2 seconds.
The small intervals where the joystick isn't showing as fully down appear a little unusual.
There may be a problem with the controller?

 

[airsnobu]

I'm starting to suspect that as well. The gimbal tilt wheel never worked on this controller, so maybe it has degraded further. Still, i landed this drone hundreds of times with the same controller with no trouble, but yes, hard data is difficult to dismiss.

 

[slup]

Just briefly looked over this thread & noticed a failed landing due to a inconsistent throttle down command.... Have myself a Mavic Air 1 & have during my ownership been forced to clean the sticks 2 times due to ghost inputs, especially when I have applied throttle down. The ghost inputs have been such that when I have only applied throttle down (left stick down) another axis also have been activated... in my case a rudder command affecting the yaw. The solution have been to clean the sticks according to below YT video. This yaw "ghost command" have in my case been visible in the stick calibration screen, so you can check this before attempting a cleaning...

I f you see that you have stick commands that affect another axis then the intended... it can be high time to disassemble the RC & clean the sticks sensor surfaces.

It's a bit fiddly so take your time... it will fix possible ghost inputs, so it's worth the effort.

 

[WanderLost]

Thanks for pointing this out. Interesting find. I was holding down the left stick while watching in shock how the motors revved up instead of down. Strange how this gets reflected in the logs but i never let go of that left stick pointing down.

Is it possible the potentiometers on your control sticks are dirty? If that happens, the controller won't see your control inputs and the drone will see some random inputs coming from them.

 

[airsnobu]

In the Calibration screen i see no crosstalk on either axis. Even so, i disassembled the controller and gave it a good clean, fixed the gimbal tilt wheel in the process (one of the wire leads was loose).

Not yet sold on the controller giving ghost inputs but i'll take a few test flights to try to understand the behavior better.

Many thanks for all your help so far, you are a top notch community.

One more thing, just for me to understand, if let's say the downward sensor is malfunctioning, would the drone hover in place with the throlle stick pulled down or would it fly up and/or move position as well?

 

[Meta4]

One more thing, just for me to understand, if let's say the downward sensor is malfunctioning, would the drone hover in place with the throttle stick pulled down or would it fly up and/or move position as well?

Here's the relevant part of the data:
The red highlighted cells show the throttle pulled fully downward.
You can see two brief intervals where the throttle input jumped from fully down (364) to the centre position (1024).
The blue highlighted areas show the barometric height data and VPS data (in feet).
From the data you can see the drone rise slightly each time.