uncommanded descent

[FoxhallGH]

The mystery isn't what happens ... it's why.

Check out my post #19 - as we were both posting about the same time I think ...

 

[RogerDH]

You are saying you calibrated the IMU and it still listed. I have a couple of ideas. Just checking if there was possibly some slight wind and it was tilting to hold it's position? Maybe double check by flying indoors.
Also any chance when changing the props that any one of the split props were mixed up and aren't providing the proper lift and are compensating by spinning faster? Any power warnings when giving full thrust?

If it was both listing and holding position, I think it must have been fighting a breeze.

 

[slup]

Check out my post #19 - as we were both posting about the same time I think ...

Yeah ... have seen that clip earlier. Unfortunately I don't think it's only a prop issue, even though I'm sure the props are involved, but it's more to it. Have seen to many examples where pilots aware of the stowage swoop to new props ... it goes well for a while but then the problems resurfaces.

 

[timinator]

What's happening during a "uncommanded descent" incident is pretty well explored ...everything starts with a sudden up pitch & then stays near horizontal angle, meantime one or both rear motors are at or very near max rpm's, heading speed goes down (as no down pitch can be achieved) ... the AC lose lift & down speed goes positive (descend).

In below DAT log example from February the AC have just started a failsafe RTH ...

-All goes well to start with. At 704,8sec the AC have turned around & start to gain heading speed (green).
-The pitch is around 15 degrees down which is according to spec. (dark green).
-During the whole RTH process the rear motors have been revving pretty high (Black & Blue).
-Then suddenly at 709,7sec (where the markers are in the chart) the AC pitch up to 14 degrees then down again to very near horizontal level ...
-There the right rear motor are at max (around 16000rpm), the heading speed goes down and the down speed (dark yellow) goes up (meaning descend) ---> the AC have lost lift totally.
-Close there after the RC-AC connection is lost & the AC is later found on ground.

The mystery isn't what happens ... it's why.

(Click on the chart to make it larger)
View attachment 99025

Great analysis. It's putting itself into an aerodynamic stall but not solely because of the prop. The aircraft has an aft CG and maybe that's why people have had success flying it home backwards when this happens. I'm wondering if reverse flight decreases engine RPM, relatively speaking, between front and back engines which is what stops the runaway rear engines?

 

[timinator]

This is an interesting post on YouTube, from a guy in Idaho (Idaho Quadcopter), on the 29th March. His video is rather long, but it 'shows' the Mavic Mini problems as per this thread. He has videoed his observations, and also comments about reading posts here on this forum that led him to diagnose that he had to change the rear props to cure the problem. That did in fact work, and he had a completely different (positive) experience with his MM after that prop change.
He has interesting conclusions right at the end of the video, and shows that when the MM is packed into the FlyMore case, the rear props (which end up folded up under the front half of the MM), should cross & point directly left & right, and sit in an open area of the case - that is - if they are aligned as per the DJI label stuck to the bottom of the case. If however, the props swing back toward the tail of the drone as it is stowed, then the props end up sandwiched between the drone's body, and a support ridge immediately behind the cut-out area for the props. He believes that it was this misalignment of the props as the MM was put into the case, that caused the prop blades to be put under pressure and squeezed flat (reducing their pitch). It is notable that there is a rectangular block shape in the lid of the Flymore case that applies positive pressure onto the top of the drone's body when it is stowed, and this would in turn, apply compression to the blades if they were stowed incorrectly and sitting on top of the support piece in the case.

About 6:03 you can actually see it enter the aerodynamic stall.

 

[rscheff]

As my Mavic Mini increasingly has problems with keeping altitude when flying at full speed I want to ask you to do a little test with yours to check if this problem is affecting all Mavic Minis more or less:

1. Go to a flat location with at least 200m free space in one direction.
2. Take off in P mode and go to a preferred altitude, keep the altitude as shown within the Fly app in mind. Don't touch the throttle stick anymore from now on.
3. Fly straight away approx. 200m with full speed, then slow down for hovering.
4. Read the altitude there, then wait about 10s and check for uncommanded altitude loss.
5. Turn 180° and fly back to your home point with full speed.
6. When back, check the altitude as displayed by the Fly app. Is it still the same as was at the beginning, or how far is it off?
7. Switch to Sports mode and repeat from step 2.

Please let me know your results.

I will try this tomorrow, but I think I know what you are getting at. I have noticed, especially in sport mode, that as I fly low to ground for a good distance, the drone goes down, when I stop it goes up to what I believe to be a higher altitude than when I started. I never really documented anything or really paid too much attention to it, but I will now.

 

[sar104]

Pure conjecture but it smells like a software problem to me. I was, still am, a programmer.

If it is software the obvious rhetorical question would be where's the reboot button on this thing? In my fantasy world a quick reboot command issued to drone and controller would force the drone to hover in place until everything came back up and reconnected. Maybe it can already do that, I must try switching off the controller and Fly app and see what happens, does anyone know?

But not so fast... it's likely impossible to force a reboot of the vehicle's onboard control systems without switching off the motors/GPS/IMU. I guess it would depend on how tightly-bound those subsystem modules are, and/or whether there's a fail safe built-in to keep it in the air while the control system is restarting. If it's a bug in one of the subsystems required to hover restarting wouldn't work. Sorry, just thinking out loud here.

How can it be just a software problem if the motors are spinning at maximum rpm but the aircraft is descending?

 

[rehkram]

Software is telling the motors to spin faster. So in the chain of causation the command issued is the start of the problem.

The inability of the props to provide lift at overspeed is, in my opinion, initiated by the (erroneous?) command to spin them up to overspeed. MM's blades are very thin, it's easy to see how the extreme speed would flatten out the aerofoil section, or perhaps distort the aerofoil in other ways, and so lose lift despite the increasing RPM

Sure the blades need to be beefed-up, but the resulting lack of lift is a secondary effect, not the cause. There's a logic error in there I believe.

BTW this has happened to me. It was windy, hovering before landing. In my case gusting wind seemed to be implicated in it going into this strange mode. No big deal, I just forced it to land in a hurry commanding left stick down. The noise was striking; could have been the props either flattening, warping or fluttering, I don't know.

 

[rehkram]

So, in my opinion it's not so much an 'uncommanded descent' as a 'falling out of the sky' thing as the props flex or do whatever they are doing that causes them to lose lift at high RPMs when wind is buffeting the plane.

The brain should not be telling the motors to exceed the design limits of the props when the wind is gusting. See my post elsewhere on the forum about limiting my MM takeoffs to light wind conditions. In my case the MM was hovering right in front of me 4 feet off the ground. If it had happened over water a mile away I wouldn't be quite so blasé about it.

 

[FoxhallGH]

There are two things that can happen with prop blades as motor rpm increases:
1) As the rpm increases, the torsional (twisting) stresses along the length of the blade increase, and the flexible plastic blade will tend to 'feather' into the airflow - especially out toward the tips - which reduces the 'pitch' and lifting ability of the prop.
2) In addition, the inner part of the prop with the coarsest pitch will hit a speed where it stalls. Not only is the blade-stall going to massively reduce the lifting ability of that blade, but it will upset the airflow behind it so that its partner on the other side of the hub, will be trying to gain lift in dead air. If not the same as - this would be similar to the effect of 'cavitation' on a ship's propellor.

It appears that the software in the Mavic Mini is written to simply increase rpm when additional lift is required - and in their 'normal' shape, the MM DJI prop's will accommodate. However, if the Mav' has been stowed in a way that warps the blades (i.e. reduces their pitch - see my post #19), then those prop's are going to be more effective at lower rpm than they are at high! The reason for this is that the flattened tips provide almost no lift, and the course-pitched inner part of the prop will do all the work right up to the point where it stalls and suddenly stops providing any lift at all. In the stalled & over-rev'd condition, the prop's are going to act like a flat disk of plastic with some parachute-like facilities - which explains why the MM doesn't just plummet but tends to lower itself using the front motors to maintain horizontal stability.

This also explains why 'flying backwards' is going to be an effective way to rescue a MM in this condition ... It's because when the MM is flying backwards, the prop's providing the 'power' are the unaffected front set, while the rear set don't have to rev' so high - and are simply providing a bit of lift.

My conclusion is that if the Mavic Mini is stowed incorrectly with pressure placed on the rear prop blades in such a way that they deform and 'flatten' - the Mavic will need to use higher motor rpm to get an equivalent lift from the prop-set. Under 'no-stress' flight conditions, the MM should fly as normal, but higher prop rpm (higher pitch sound) will probably be noted. However, if the MM needs 'power', the motor rpm maxes out and the blades will stall and become almost ineffective. The Mavic's motor controllers will keep the rpm high in an attempt to regain lift, which will maintain the stall until downward drift stops (e.g. the MM bounces off the ground into a hover). I believe that the flight data analysis that has gone on in this post, backs that up ... I also believe that it is going to be impossible to replicate this problem on a Mavic Mini that has been stowed correctly in the FlyMore case, or is usually stowed some other way.

In my opinion, the two ways to remove this problem (after changing the rear props), are:
1) very careful stowage to ensure that the prop-blades stowed under the MM have not swung backwards and are being compressed under the body of the drone while it is in the FlyMore case.
2) and/or - use of a 3rd party prop made from a stiffer composite material which will not deform.

 

[GerdS]

In my case it's not a prop problem for sure. I never get "max. power reached" messages and my MM sounds same as in the video after changing the props.
I also secure my props with a 3D printed lock (Mavic Mini Propeller Protector by Gerd_S) before putting the MM into the Fly More box and they are not deformed or bent.
For me, both uncommanded ascending and descending are mainly caused by a serious firmware bug which is there from the very beginning. For sure, bended props may also lead to uncommanded descending, but don't explain uncommanded ascending.

Please do the test I asked for before, if you will see any changes in altitude this will proof my suspicion.

 

[Meta4]

For sure, bended props may also lead to uncommanded descending, but don't explain uncommanded ascending.

Uncommanded ascending only shows up in a few reports but when it does, it's always with uncommanded descending.
The whole issue is quite complex.
The problem shows up in windy conditions but also in light breezes.
It usually shows up half way through a flight, after the drone has been flying normally.
All of these suggest that the cause is not as simple as just the props being warped in storage.

 

[GerdS]

Uncommanded ascending only shows up in a few reports but when it does, it's always with uncommanded descending.

Not true, it is actually fully reproducable to me with every flight while uncommanded descending is occuring quite rarely.
Please do the test with yours according to my previous post (uncommanded descent).

 

[sar104]

Software is telling the motors to spin faster. So in the chain of causation the command issued is the start of the problem.

The inability of the props to provide lift at overspeed is, in my opinion, initiated by the (erroneous?) command to spin them up to overspeed. MM's blades are very thin, it's easy to see how the extreme speed would flatten out the aerofoil section, or perhaps distort the aerofoil in other ways, and so lose lift despite the increasing RPM

Sure the blades need to be beefed-up, but the resulting lack of lift is a secondary effect, not the cause. There's a logic error in there I believe.

BTW this has happened to me. It was windy, hovering before landing. In my case gusting wind seemed to be implicated in it going into this strange mode. No big deal, I just forced it to land in a hurry commanding left stick down. The noise was striking; could have been the props either flattening, warping or fluttering, I don't know.

Ah - understood. That's a similar hypothesis to one that I've been trying to test. I'm not sure that it really counts as a software problem though, because the props are presumably supposed to maintain their aerodynamic characteristics at least up to the maximum motor speed, hopefully with some margin. So I think that they have to be somewhat deformed to begin with for this to manifest itself.

 

[sar104]

As my Mavic Mini increasingly has problems with keeping altitude when flying at full speed I want to ask you to do a little test with yours to check if this problem is affecting all Mavic Minis more or less:

1. Go to a flat location with at least 200m free space in one direction.
2. Take off in P mode and go to a preferred altitude, keep the altitude as shown within the Fly app in mind. Don't touch the throttle stick anymore from now on.
3. Fly straight away approx. 200m with full speed, then slow down for hovering.
4. Read the altitude there, then wait about 10s and check for uncommanded altitude loss.
5. Turn 180° and fly back to your home point with full speed.
6. When back, check the altitude as displayed by the Fly app. Is it still the same as was at the beginning, or how far is it off?
7. Switch to Sports mode and repeat from step 2.

Please let me know your results.

I've tried that a couple of times. No descent. I'm pretty sure now that the props have to be already deformed to some extent before this problem manifests itself.

 

[sar104]

There are two things that can happen with prop blades as motor rpm increases:
1) As the rpm increases, the torsional (twisting) stresses along the length of the blade increase, and the flexible plastic blade will tend to 'feather' into the airflow - especially out toward the tips - which reduces the 'pitch' and lifting ability of the prop.
2) In addition, the inner part of the prop with the coarsest pitch will hit a speed where it stalls. Not only is the blade-stall going to massively reduce the lifting ability of that blade, but it will upset the airflow behind it so that its partner on the other side of the hub, will be trying to gain lift in dead air. If not the same as - this would be similar to the effect of 'cavitation' on a ship's propellor.

It appears that the software in the Mavic Mini is written to simply increase rpm when additional lift is required - and in their 'normal' shape, the MM DJI prop's will accommodate. However, if the Mav' has been stowed in a way that warps the blades (i.e. reduces their pitch - see my post #19), then those prop's are going to be more effective at lower rpm than they are at high! The reason for this is that the flattened tips provide almost no lift, and the course-pitched inner part of the prop will do all the work right up to the point where it stalls and suddenly stops providing any lift at all. In the stalled & over-rev'd condition, the prop's are going to act like a flat disk of plastic with some parachute-like facilities - which explains why the MM doesn't just plummet but tends to lower itself using the front motors to maintain horizontal stability.

This also explains why 'flying backwards' is going to be an effective way to rescue a MM in this condition ... It's because when the MM is flying backwards, the prop's providing the 'power' are the unaffected front set, while the rear set don't have to rev' so high - and are simply providing a bit of lift.

My conclusion is that if the Mavic Mini is stowed incorrectly with pressure placed on the rear prop blades in such a way that they deform and 'flatten' - the Mavic will need to use higher motor rpm to get an equivalent lift from the prop-set. Under 'no-stress' flight conditions, the MM should fly as normal, but higher prop rpm (higher pitch sound) will probably be noted. However, if the MM needs 'power', the motor rpm maxes out and the blades will stall and become almost ineffective. The Mavic's motor controllers will keep the rpm high in an attempt to regain lift, which will maintain the stall until downward drift stops (e.g. the MM bounces off the ground into a hover). I believe that the flight data analysis that has gone on in this post, backs that up ... I also believe that it is going to be impossible to replicate this problem on a Mavic Mini that has been stowed correctly in the FlyMore case, or is usually stowed some other way.

In my opinion, the two ways to remove this problem (after changing the rear props), are:
1) very careful stowage to ensure that the prop-blades stowed under the MM have not swung backwards and are being compressed under the body of the drone while it is in the FlyMore case.
2) and/or - use of a 3rd party prop made from a stiffer composite material which will not deform.

I'm currently trying to induce the rear prop deformation by simulating that storage condition. I'd really like to show at least significantly increased rear motor speeds relative to undeformed props.

 

[FoxhallGH]

I'm currently trying to induce the rear prop deformation by simulating that storage condition. I'd really like to show at least significantly increased rear motor speeds relative to undeformed props.

That will be interesting ... It's times like this you realise just how useful that ability for the Mav Pro and 2 Pro to display the motor rpm on the controller LCD really is ...

 

[rehkram]

Gerd, I will fly your test profile later today if the weather cooperates. I'll also be interested to compare the results. I seldom need an excuse to fly the MM but flying your test is a good one.

And thanks for the great input folks. Much to absorb there. I'm quite sure there are multiple factors in play, it's a classic case of James Reason's 'swiss cheese' model of accident causation;
when the holes all line-up an accident or upset may occur. Software, hardware, weather and perhaps human factors are the slices here. This makes it hard to duplicate a problem at will. I'm now thinking the quickest temporary fix would likely be stiffer prop blades.

 

[Meta4]

Software, hardware, weather and perhaps human factors are the slices here. This makes it hard to duplicate a problem at will.

We've seen enough cases to eliminate human factors and weather from the equation by now.

I'm now thinking the quickest temporary fix would likely be stiffer prop blades.

If it was that simple .... the issue would have been solved long ago.

 

[sar104]

Gerd, I will fly your test profile later today if the weather cooperates. I'll also be interested to compare the results. I seldom need an excuse to fly the MM but flying your test is a good one.

And thanks for the great input folks. Much to absorb there. I'm quite sure there are multiple factors in play, it's a classic case of James Reason's 'swiss cheese' model of accident causation;
when the holes all line-up an accident or upset may occur. Software, hardware, weather and perhaps human factors are the slices here. This makes it hard to duplicate a problem at will. I'm now thinking the quickest temporary fix would likely be stiffer prop blades.

I'm pretty sure that stiffer blades would fix the problem, but DJI haven't left much headroom to stay under 300 250 g. That might even be the genesis of the problem.