Theory explaining how Mavic could drop down from skies... Your thoughts, please...

[Anthony Viscomi]

I thank you for your support and apologize if I might have made it look a "wide-spread issue"... That is not, definitely, but I was just making reference to the reports that I have seen... If only ONE that I could discard pilot error, I still would be worried about...

Many thanks!

Feel free to use this debacle as an example:
Spark falling out of sky

 

[Alex ROV]

Feel free to use this debacle as an example:
Spark falling out of sky

Many thanks! I did not know about the issues on Spark model...

I might be completly wrong but I believe it matches with my theory of "false landing detection and shutdown"...
Spark being designed to be hand catched would have her system "more flexible", "more likely" to detect a landing event or catching... Therefore more likely to suffer of "false landing indication"...

Lets say that is a combination of barometer, sonar and accelerometers to detect movement or resistance and than assume tha she has landed... Spark being hand catched would require a more flexible, wider range of movements and still detecting a real "landing" or "catching"...

 

[Alex ROV]

That is one link... This one was in a thread here...

Mavic pro simply fell out of the sky while flying

Airdata UAV - Flight Data Analysis for Drones

It does not show anything anormal... It was a regular flight according to the log, just like all the others... No "lost connection", nothing...

I'm still looking for some others in my spare time and will keep posting... The last one I saw was about 2 weeks ago, with a very experienced pilot... He had the Mavic hovering in front of him, then he looked down to RC to switch to Sport Mode and suddently heard her crashing the ground a few meters ahead...

Many thanks!

 

[geigy]

I too think that its admirable work done by Alex ROV [snip] In regards to DJI providing us (the consumer) with any data pertaining to failure rates is ludicrous. What other consumer product(s) manufacturer publicly publishes failure rate information.

healthydrones aka AirData is not part of DJI, who I agree other than manufacturing a great product, seems to be falling down a bit lately with support. It seems to me that AirtData would be in a good position to keep stats on crashes and other failures - I'm not a member of that site but thought maybe someone else might know.

 

[gnirtS]

The .dat file will be far more useful than the .txt that airdata uses as its got all the internal diagnostics.
As i said though, ive not seen any data to suggest mavics ARE falling out of the sky (and the ones that do arent user error).

 

[BudWalker]

Very good tech contribution! If we forget about the "name" of the device or subsystem in charge for providing the indication that a landing has happened, lets say, in the rare case or possibility of this subsystem glitches for a few seconds, just enough to produce a "false landing" indication... Would the motors shut down? What catches my attention is the logs showing a regular end of flight, without "lost connection" warnings or messages...

Many thanks!

You're right. But, I thought I didn't have to point out that your logic depends on the IMU glitching at just the right moment and never at any other time. I can assure you that if the IMU glitches in this way we would know about it.

In post #21 @Anthony Viscomi referred to a Spark incident where it fell out of the sky. I looked at the flight referenced here
Spark falling out of sky

In fact it did fall out of the sky on the third flight contained in the .DAT. The first two times it landed with a normal motor shutdown. The third time the motors were running right up to the very end.


Also, if the motors are shutdown because the AC has landed that fact will be recorded in the eventLog stream of the .DAT. E.g.this shows the entries for the second landing

261.603 : 1405161386 : 15394 [L-RC]Start landing. rc vib!
262.303 : 1408309059 : 15429 [L-FMU/MOTOR] Stop. reason:ACT.landing
262.303 : 1408309349 : 15429 [L-FMU/MOTOR]Total start times: 2, time: 238.76

There are no such entries for the end of the .DAT

 

[Alex ROV]

You're right. But, I thought I didn't have to point out that your logic depends on the IMU glitching at just the right moment and never at any other time. I can assure you that if the IMU glitches in this way we would know about it.

In post #21 @Anthony Viscomi referred to a Spark incident where it fell out of the sky. I looked at the flight referenced here
Spark falling out of sky

In fact it did fall out of the sky on the third flight contained in the .DAT. The first two times it landed with a normal motor shutdown. The third time the motors were running right up to the very end.
View attachment 17990

Also, if the motors are shutdown because the AC has landed that fact will be recorded in the eventLog stream of the .DAT. E.g.this shows the entries for the second landing

261.603 : 1405161386 : 15394 [L-RC]Start landing. rc vib!
262.303 : 1408309059 : 15429 [L-FMU/MOTOR] Stop. reason:ACT.landing
262.303 : 1408309349 : 15429 [L-FMU/MOTOR]Total start times: 2, time: 238.76

There are no such entries for the end of the .DAT

Very nice tools! I'm going to read the weblinks for the tools you have created, it looks great...

Please, could you explain me how the AC descends on the first flight while the motor's speed are looking in the same range as for the max height? And on the second flight, the motors speed up but the relativeHeight almost does not change?

I am a submarine pilot (ROV - Remotly Operated Vehicles) and we have 3 or 4 sets of vertical thrusters and 4 horizontal units on the sub and the working algoritm has always amazed me when we operate in auto modes like auto positioning, autodepth, autoheading, autoaltitude... Wish we could have the same graphics and tools...

Many thanks!

 

[BudWalker]

Very nice tools! I'm going to read the weblinks for the tools you have created, it looks great...

Please, could you explain me how the AC descends on the first flight while the motor's speed are looking in the same range as for the max height? And on the second flight, the motors speed up but the relativeHeight almost does not change?

I am a submarine pilot (ROV - Remotly Operated Vehicles) and we have 3 or 4 sets of vertical thrusters and 4 horizontal units on the sub and the working algoritm has always amazed me when we operate in auto modes like auto positioning, autodepth, autoheading, autoaltitude... Wish we could have the same graphics and tools...

Many thanks!

If I understand your question the plot needs to be zoomed some in the vertical scale to see that the motorSpeeds increase when the AC ascends.

 

[Alex ROV]

Ok... I am going to download your tools and use my flights to study!

Many thanks and congratulations for your work and helpful assistance!

 

[stratos]

Can you post one log that shows an example of this?

Well. honestly I experienced something like this long time back, early Dec 2016.

Short story, I took of from a public golf Green. It was around 6 or 7pm so it was not too bright (but being summer it was still very clear).
Now, I was landing as always and from about 1meter pushing the stick down and it just shut off the engines and crashed into the green with a lot of force.

I was lucky that the surface was really soft so I just broke 2 props and scared a bit the back my quad.
Shared the logs with DJI support and they siad that because of the colour - (being very similar) the sensors "though" that the ground was closer, the ultra sonic sensors got confused as well for the soft surface and well yeah, to the mavic the floor was 1 meter closer so it decided to shut down the engines.

But this was not related to "Lost connection"

 

[Brojon]

A few things, firstly I'm very unconvinced that mavics are falling from the sky with any sort of regularity. Of the few incidences you can find online most boil down to user error it them failing to understand how the mavic works.

Secondly, a stuck barometer would be very very easy to see in the txt logs or the. Dat file and I haven't heard of a single example of it.
It's a theory but so far from what I can see lacks any evidence (or evidence of an actual problem in the first place)

This certainly sounds like a rehash of a current conversation on Facebook. Likely the same guy.
I pointed out the barometer too but it went in one ear and out the other.

 

[charliesRig]

Pilot error Pulled both stick back (CSC)
Mavic respond and Fell out of the sky.!

Rookie move! totally My fault and I own it......C.....
.


.

I did Panic Once when Loss of signal with controller....... (windy & sports mode)
Failsafe mode Kicked in and RTH activated Mavic came back and Hung in the air above home port.?
Low on battery Less the 20%
I gained control late battery was 8% and I panic at around 30ft
I pulled back both stick (CSC)
All motors Stop.!

again set controller to" Emergence" Not "Always" activate CSC
(should be by default)
and just the other day it happen again But with More then 20% battery I could take my time Troubleshoot and
Found by toggling between Sport and "P" I could Descend and land No Damage
.
.
sorry if I sound like a broken Record (Pieces of Vinyl)
This lesson cost me $350 Hopefully I can help some one on here not to make the same Mistake
......................C
.

 

[imnotreallyscott]

I'm whatever is the farthest thing from a drone or avionics expert. But what I can see here is the OP completely filled with confirmation bias. I haven't read the entire thread but he's been repeatedly asked for logs to demonstrate that a significant problem actually exists and follows the contour of his "theory." None have been provided (that I've seen so far), just stories "about a guy..." A theory has evidence to support it. This is not that.

 

[imnotreallyscott]

When I find anyone again of them I shall post it, not a problem... But if you find any showing "lost connection" or anything diferente than a regular end of flight for this "drop down" reported events you can post it as well, I believe your database is much bigger than mine... At end of the day that is what I'm kindly asking for, contributions for the case...

Confirmation bias case in point.

 

[Michael Ainsworth]

That's also a problem with stats and other issues.
As a nurse of 40+ years experience I know that you can virtually make any issue into what you perceive is the problem.
Not saying OP is completely wrong but his viewpoint is compromised.
So far his stats do not add up.
Keep on though who knows you may discover a trend somewhere.

 

[Alex ROV]

I'm whatever is the farthest thing from a drone or avionics expert. But what I can see here is the OP completely filled with confirmation bias. I haven't read the entire thread but he's been repeatedly asked for logs to demonstrate that a significant problem actually exists and follows the contour of his "theory." None have been provided (that I've seen so far), just stories "about a guy..." A theory has evidence to support it. This is not that.

I have posted one example a few messages ago... A completely normal flight according to the logs, but the log/flight ends BEFORE the aircraft (apparently) shutdown and fall... It could not be done by the pilot...

Mavic pro simply fell out of the sky while flying

It's happening much more with Sparks... What I believe that matches with my theory, as Spark would be more suitable to detect a "false landing" as they are designed to be hand catched...

Some DJI Spark drones are falling from the sky, and that’s to be expected

It doesn't happen that much with Mavic's and I have already apologized if I had made it looks like a frequent issue...

 

[Alex ROV]

And here is another log... A very experienced pilot with the Mavic hovering a few meters from him... Another normal flight log without CSC or any issues... The log/flight ends, Mavic shut down and fall...

Airdata UAV - Flight Data Analysis for Drones

The whole report and details are in a private Facebook Mavic owners group.

 

[msinger]

And here is another log... A very experienced pilot with the Mavic hovering a few meters from him... Another normal flight log without CSC or any issues... The log/flight ends, Mavic shut down and fall...

Airdata UAV - Flight Data Analysis for Drones

The whole report and details are in a private Facebook Mavic owners group.

Where do you see the "MAVIC HAS LANDED" message in the logs you are posting above?

 

[Alex ROV]

Where do you see the "MAVIC HAS LANDED" message in the logs you are posting above?

The "Mavic has landed" message was actually on the last column of the flight's DAT files, they show more details than the logs... I will keep searching and posting, specially now that I have found the tools created by user BudWalker, that's a real piece of GREAT work!

Here is another one... Airdata UAV - Flight Data Analysis for Drones
And on its comments there are reports of 3 more users with the same issue... Should find them again and post here...

But right now it looks like that Sparkland is a much more fertile terrain for studying... Which at the end might bring benefits for us, as well...

 

[BudWalker]

The "Mavic has landed" message was actually on the last column of the flight's DAT files, they show more details than the logs... I will keep searching and posting, specially now that I have found the tools created by user BudWalker, that's a real piece of GREAT work!

Here is another one... Airdata UAV - Flight Data Analysis for Drones
And on its comments there are reports of 3 more users with the same issue... Should find them again and post here...

But right now it looks like that Sparkland is a much more fertile terrain for studying... Which at the end might bring benefits for us, as well...

That's great! But, we need to see the data so that your theory can be verified. Can you please provide that .DAT by posting a Dropbox or GoogleDrive link? Or, at least provide a link where someone else has presented the results of looking at the .DAT.