Mavic Mini uncommanded descent tests

digit3l · May 3, 2020

I just had this same error as well. New MM owner, been on less than 15 flights probably, using the Fly More Case. So, the only effective solution is to replace the props at this point and make sure you store it in a safe manner..i.e. not in the FM case?

Thanks for doing all the research on this!

lee82gx · May 3, 2020

So we know that the props are the first level of root cause, but the 2nd level of root cause is still undetermined. As in why the front props do not experience so much RPM even in hover, while the weight balance or COG does not suggest very uneven distribution. (I know the popular theory is the case or storage, but many already show that the case has plenty of space to avoid stress in storage).
Here I suggest the 2nd level of root cause.

due to the orientation of the rear motor axis being too tilted to the sides. Perhaps this is the reason the motor has to spin more to generate nett vertical lift, because of the cancellation of forces happen from rear left and rear right motor due to the tilt angle.

also, is it possible that the rear props when rotated at tilt angle also causes uneven bending stress which accelerate the prop failure?

rehkram · May 3, 2020

Interesting observation. Looking at my MM down the centerline from the tail I see what you mean. The obvious question would be does that outward tilt manifest itself while in flight, or only on the ground.

In other words it could be as-designed compensation for upward flexing of the arms under flight loadings. But while that compensation would work for power climbing, how about descending (he asked rhetorically) ?

It would be interesting to know if the other Mavic recent designs have this, I don't own one that I can check, sadly.

lee82gx · May 3, 2020

rehkram said:
Interesting observation. Looking at my MM down the centerline from the tail I see what you mean. The obvious question would be does that outward tilt manifest itself while in flight, or only on the ground.

In other words it could be as-designed compensation for upward flexing of the arms under flight loadings. But while that compensation would work for power climbing, how about descending (he asked rhetorically) ?

It would be interesting to know if the other Mavic recent designs have this, I don't own one that I can check, sadly.

Many Mavics and most phantoms have motors that are not axially pointed vertically upward but have a certain angle. I've not see them drop though.

sar104 · May 3, 2020

lee82gx said:
Many Mavics and most phantoms have motors that are not axially pointed vertically upward but have a certain angle. I've not see them drop though.

That's correct. And in this case the problem is clearly caused by deformed props, not the motor axis.

JSM · May 3, 2020

Does any one know if there are better props available as yet?

sar104 · May 3, 2020

JSM said:
Does any one know if there are better props available as yet?

There are a few listed, but whether they are any better is anyone's guess.

Amazon.com: 8 Pairs Foldable Carbon Fiber Propellers Noise Reduction Quick-Release CW & CCW Blades Props Accessories Kit for DJI Mavic Mini Drone (16 pcs Yellow) : Toys & Games

Buy 8 Pairs Foldable Carbon Fiber Propellers Noise Reduction Quick-Release CW & CCW Blades Props Accessories Kit for DJI Mavic Mini Drone (16 pcs Yellow): Propellers - Amazon.com ✓ FREE DELIVERY possible on eligible purchases

www.amazon.com

Amazon.com: RCGEEK 8PCS Compatible with DJI Mavic Mini Propellers 4726 F Props Foldable Quick Release Blades Fit for Mavic Mini Drone Replacement Accessories;: Toys & Games

Buy RCGEEK 8PCS Compatible with DJI Mavic Mini Propellers 4726 F Props Foldable Quick Release Blades Fit for Mavic Mini Drone Replacement Accessories;: Propellers - Amazon.com ✓ FREE DELIVERY possible on eligible purchases

www.amazon.com

Amazon.com: AxPower Upgraded Version Mavic Mini Propellers for DJI Mavic Mini Replacement Low-Noise and Quick-Release Blades(Silver): Toys & Games

Buy AxPower Upgraded Version Mavic Mini Propellers for DJI Mavic Mini Replacement Low-Noise and Quick-Release Blades(Silver): Toys & Games - Amazon.com ✓ FREE DELIVERY possible on eligible purchases

www.amazon.com

bigswifty · May 4, 2020

Thank you @sar104 for doing this, it's been really interesting to follow. I might actually buy a Mini on the strength of these findings. Yes it's a weak piece of design, but the workarounds are pretty straightforward - store it carefully/properly.

Anyway, thank you!

RichR · May 4, 2020

Again many thanks for doing the work to properly investigate this issue, rather than let speculation and guesswork form a consensus. I would like to think that DJI have done similar work during pre-production but perhaps scaling up to full production has resulted in poorer quality plastic mouldings. It's clear they know there is an issue with storage in the Fly More case, as newer ones get the sticker advising on how to store the props whereas at launch and early on they gave the impression it was more robust than it turns out to be.

I've started looking at my rear props when I get it out of the case to see if there's any sign of the tips being deformed, and also on takeoff I'm checking the back goes up at the same rate as the front.

Hughes820 · May 4, 2020

I noticed my back motors were “flapping up” while hovering. When I went to replace the blades I noticed play in both back motors. No play in the front I have reason to believe the fly more case is causing the shaft in motor to get tweaked a bit causing a wobble. Just made sure nothing was broken or missing and checked the screws. Replaced all props gonna recalibrate everything and test it out but I’m convinced that fly more case is putting pressure on the back motors that’s just my take on this issue

BudWalker · May 4, 2020

One way of looking at the UD problem is that the FC isn't properly allocating the available forces. I.e. more force should be allocated to lift and less to lateral motion. The record type CtrlAllocation has data that shows how the FC is allocating these forces and what happens during a UD. Here is an example, taken from FLY061, where a UD was successfully arrested.

Full elevator (green) caused some altitude (red) deviations but, over all, there was no descent. Instead, speed (blue) was sacrificed.

This was achieved by limiting tilt (red).

Also shown is the lift_diff that DatCon computes from the raw_lift and fix_lift data in the CtrlAllocation record.

Finally, the CtrlVelVert vel_cmd and vel_fdbk data provide a slightly different perspective. The UD is indicated when the vel_cmd > 0.0 and vel_fdbk < 0.0.

About 1 sec later the FC starts adjusting the tilt_scale and lift_diff so that vel_cmd and vel_fdbk become more or less equal.

Later in FLY061 there was a UD that wasn't arrested

sar104 · May 4, 2020

BudWalker said:
One way of looking at the UD problem is that the FC isn't properly allocating the available forces. I.e. more force should be allocated to lift and less to lateral motion. The record type CtrlAllocation has data that shows how the FC is allocating these forces and what happens during a UD. Here is an example, taken from FLY061, where a UD was successfully arrested.
View attachment 100470
Full elevator (green) caused some altitude (red) deviations but, over all, there was no descent. Instead, speed (blue) was sacrificed.

This was achieved by limiting tilt (red).

View attachment 100472
Also shown is the lift_diff that DatCon computes from the raw_lift and fix_lift data in the CtrlAllocation record.

Finally, the CtrlVelVert vel_cmd and vel_fdbk data provide a slightly different perspective. The UD is indicated when the vel_cmd > 0.0 and vel_fdbk < 0.0.
View attachment 100474
About 1 sec later the FC starts adjusting the tilt_scale and lift_diff so that vel_cmd and vel_fdbk become more or less equal.

Later in FLY061 there was a UD that wasn't arrested
View attachment 100475

Completely agree. It appears that the FC could prevent the problem by prioritizing vertical control rather than pitch, even if that results in rather comically slow forward progress.

BudWalker · May 4, 2020

sar104 said:
Completely agree. It appears that the FC could prevent the problem by prioritizing vertical control rather than pitch, even if that results in rather comically slow forward progress.

But, to be fair, the FC doesn't know about the reduced force in the back props. If the allocation algorithm were modified to give more priority to lift then lateral speed would be reduced in some situations when it's not necessary.

sar104 · May 4, 2020

BudWalker said:
But, to be fair, the FC doesn't know about the reduced force in the back props. If the allocation algorithm were modified to give more priority to lift then lateral speed would be reduced in some situations when it's not necessary.

But presumably it should know that the rear props are underperforming, since it can see pitch as a function of motor speed. I assume that's why the pitch kicks up to 30° on the application of full forward elevator and then immediately falls back to around 15° with the deformed props.

BudWalker · May 4, 2020

sar104 said:
But presumably it should know that the rear props are underperforming, since it can see pitch as a function of motor speed. I assume that's why the pitch kicks up to 30° on the application of full forward elevator and then immediately falls back to around 15° with the deformed props.

More to the point, rear prop reduced force is an undeniable fact of life. The allocation algorithm should be expanded to include this even if the MM couldn't make it home on a windy day.

rumitcu · May 4, 2020

sar104 said:
CONCLUSION

The Mini's problems appear to be due to a shockingly bad propeller design, presumably adopted to save mass. I find it really hard to conceive that DJI didn't discover this problem in pre-release testing, unless it really does only happen with a particular batch or batches of props. That seems unlikely though, given the number of reports. Or perhaps it only happens if the props are pre-deformed in storage, and DJI never let that happen. It certainly seems to be predominantly reported by users with the Fly More combo case. Very careless, either way.

Here is an example of a pre-release unit that appears to be experiencing the typical uncommanded descents (jump to 4:10): .

According to the YouTuber, DJI was made aware of the issue, came out to try to understand the cause and claimed it would be fixed in the production units due to improvements in: 1) firmware updates, and 2) more efficient propellers. Clearly DJI did not fix the problem that they knew about before the production units shipped.

Steve Mavic · May 4, 2020

Capt KO said:
Since then I have them stored (securely in the case) using plastic caps that keep the blades folded, as shown in the diagram. No problem since...

Any chance of you posting a couple a pictures of your improvised prop storage solution?
Thanks.?

Max67 · May 6, 2020

BudWalker said:
One way of looking at the UD problem is that the FC isn't properly allocating the available forces. I.e. more force should be allocated to lift and less to lateral motion. The record type CtrlAllocation has data that shows how the FC is allocating these forces and what happens during a UD. Here is an example, taken from FLY061, where a UD was successfully arrested.
View attachment 100470
Full elevator (green) caused some altitude (red) deviations but, over all, there was no descent. Instead, speed (blue) was sacrificed.

This was achieved by limiting tilt (red).

View attachment 100472
Also shown is the lift_diff that DatCon computes from the raw_lift and fix_lift data in the CtrlAllocation record.

Finally, the CtrlVelVert vel_cmd and vel_fdbk data provide a slightly different perspective. The UD is indicated when the vel_cmd > 0.0 and vel_fdbk < 0.0.
View attachment 100474
About 1 sec later the FC starts adjusting the tilt_scale and lift_diff so that vel_cmd and vel_fdbk become more or less equal.

Later in FLY061 there was a UD that wasn't arrested
View attachment 100475

Which analysis software do you use?