Anatomy of a DJI Flyaway

Wow - fascinating to have captured one of these events with a CAN bus recorder. I hope this helps DJI understand what causes these flight controller failures and prevent them from happening in the future. Surely all the idiots who thought these fly-aways are caused by "pilot error" will now be silenced forever.
 
Ianwood,

You are earning my respect with every post of yours I read!!! Hands down!

I have some 20 years with electronics of all kind including developing of elecronic devices with microcontrollers. I've never opened NAZA but suspect that the PCB may be multilayer PCB. If the inner layers are done with not much of precision and/or has fine traces, inner vias or not laminated good they can start to fail and this is extreme duficult to diagnose even if you have the full and detailed schematics.

I also doubt it very much that sine P3 is out and based on different hardware and P2 is no longer manifactured DJI will not bother to investigate and/or fix the problem.

Ianwood, did you suffer damages from the last crash?

I am sorry to hear that you had second crash in such short time frame :(
 
Last edited:
There are a number of different types of "flyaway", but when considering the sort Ian has been unlucky enough to encounter (i.e. suicide glideslope from a hovering position); is it possible to identify any common threads.

Two which occured to me were :

- Always a crash from a GPS Hover (rather than active flight, or ATTI mode)
- Always relatively old aircraft (e.g. number of flights >100)

Any truth in this ?
 
Last edited:
Thanks for the feedback, guys. I agree on a lot of the assessments. The rareness of the event makes it intriguing especially considering that the two times it happened, there was nothing obviously different than any of the other successful flights.

Both hardware and firmware defect seem unlikely for their own reasons yet it has to be one of them. We've found the murder weapon but not the culprit who pulled the trigger.

Now I need to go back and look for similar issues in other flight logs to see if there were hints of this same issue that didn't cause the NAZA to down to the bird.
 
ian,

what flight mode were you using and how would you describe the flying you were doing when the faults occurred?
 
Last edited:
Thanks for the feedback, guys. I agree on a lot of the assessments. The rareness of the event makes it intriguing especially considering that the two times it happened, there was nothing obviously different than any of the other successful flights.

Both hardware and firmware defect seem unlikely for their own reasons yet it has to be one of them. We've found the murder weapon but not the culprit who pulled the trigger.

Now I need to go back and look for similar issues in other flight logs to see if there were hints of this same issue that didn't cause the NAZA to down to the bird.
Very interesting data, but maybe it would be of help if you also remind me what upgrades are installed. Have you done any changes lately? The fact that this happened after 12 minutes could be related to hardware (heat buildup) but if you get it back in the air and the problem is gone than I can only agree with Trumpie's software issue.
 
i remember when my naza just took off at full speed when i had no hands on the sticks at the time, i nearly made me give this whole thing up as it smashed into a house roof, the only house for miles may i add, its likely a bug in the firmware somewhere,


looking at it i be looking at the COG and speed code, sometimes these things are never resolved as they can't replicate it, its so rare in reality DJI could put on one a rig all year round and never get one to do it, often manufactures if they can't find the problem will try to put in safeguards to work around it but it don't always work,

very very interesting stuff fella very impressed :)
 
Last edited:
Firmware is 3.08. Not that it matters much. If it is a firmware issue, it occurs infrequently enough that it is probably in all versions.
 
This sounds familiar reported by a Taiwanese P2V+ pilot back to February 2015 with crazy speed, distance and satellite for a split second. LAT down shifted from 25 to 13, which is about 1,200km away.

https://www.facebook.com/notes/fpv-...e-flyaway-of-phantom-2-vision/552946811474796

10991068_552947838141360_5425218136934232895_n.jpg
 
Last edited:
  • Like
Reactions: ianwood
204 indeed! Very interesting. One interesting difference is that one decided to fly off whereas mine decided to simply give up being able to balance itself. Agree on the hover. Both of mine were from a hover.
 
There are a number of different types of "flyaway", but when considering the sort Ian has been unlucky enough to encounter (i.e. suicide glideslope from a hovering position); is it possible to identify any common threads.

Two which occured to me were :

- Always a crash from a GPS Hover (rather than active flight, or ATTI mode)
- Always relatively old aircraft (e.g. number of flights >100)

Any truth in this ?
The other, and more commonly, reported 'flyaway' is where the craft heads up and away at speed.

But perhaps they are the same- it's simply a case of the shortest route to wherever it thinks it wants to go. The fact that the planet Earth is in the way matters not.
 
Ian this is a really interesting discovery with the data you pulled from the bird. Thinking back to the many here saying that similar failures were user caused puts many things and people into perspective. Thank you for sharing the data. You da'man!
 
Would anyone happen to know which GPS chip is being used? I found some info it might be uBlox, but I'm not sure.
 
Last edited:
Thanks!
 
Mine didn't have a have a far flung destination to go to. It simply didn't want to fly anymore. Here's a visualization I've made. I'm still working to get more of the video of the crash but at least there's a few seconds of it losing control.

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

EDIT: It could be related to the uBlox. But I would guess it is more DJI's implementation of it. It seems DJI has had trouble with their GPS units. The A2 has had random dropouts (the zero satellites issue). I think Inspire has had them too.
 
Don't forget that as DJI uses a CAN bus for everything, all the uBlox GPS does is feed a microcontroller that runs DJI code to interface with that bus. I frankly wouldn't suspect the uBlox module itself (they're used in zillions of devices and highly regarded), but whatever happens next.
 
Agree on the uBlox itself not being a likely cause. I think it would be much more likely how the NAZA collects and uses the uBlox data.
 

Members online

No members online now.

Forum statistics

Threads
143,066
Messages
1,467,358
Members
104,935
Latest member
Pauos31