The classic DJI flyaway
Despite doing everything reasonable to avoid it, I am already a member of the esteemed DJI Flyaway club, but today I reconfirmed my membership by being the victim of a new mystery flyaway, which happened for no appreciable reason and without any sort of warning, following 2 minutes of perfectly normal flight control.
Location is everything
Ever since my first DJI flyaway, I have been extra cautious about where I fly, so as to minimize as much as possible the risk to people, animals and property if the worst should happen. Today I was up on the high hills of Ditcham, Hampshire in the setting sun, trying to get some stunning high aerial views of the interesting clouds and sweeping landscape. If we are going to crash, this is at least the perfect place to do it, miles away from anything that isn’t fields or trees.
All systems ready ?
This was the first flight of the day, and I had completed all the normal ground checks (including GPS mast stability), and had waited, powered up on the ground for at least a minute, to allow the satellites and home point to lock in (confirmed).
There was nothing in the surroundings that I considered a threat to my control signal or compass input. No power masts or large metal structures around the launch site, only farm buildings about 1-2 Km away, and just nothing around that might be able to interfere with my control. No people, no cars, no fencing, nothing. Just trees and fields, and one narrow road with only my car on it.
I had newly recharged batteries in my Transmitter, there were green flashing lights on the Naza (home point locked in, full satellites), so launched cautiously and proceeded with my flight, not expecting any problems, but still keeping my initial moves low while I checked that all was normal, and I had full control.
2 minutes of normal
As the video shows, I gradually increased height and began framing up shots for the video, and this continued normally for about 2 and a half minutes, during which I had perfect control, and the craft allowed me to precisely position it wherever I wanted. However, just as I was adjusting my gimbal angle, and preparing for a bit of serious height, suddenly all that changed…
It was about 50 m away from me, and maybe 40 m up when suddenly the flyer went loco, and I lost control. Video reveals that at this moment it suddenly went full throttle, and dragged hard right, both without my input. Control signal was still present as the craft responded to my Yaw input as I tried to turn it back towards me, but the throttle wasn’t listening, and I certainly wasn’t dragging it right on the cyclic.
My emergency reactions kicked in 2 seconds after this manoeuvre started and I switched to manual and ramped the throttle. Unfortunately my flyer was now somewhere between on its side and upside down as I did this, so it had the effect of powering me ground-wards even faster. I had time to manually cut the motors just as it ploughed into a corn-field at devastating speed.
Finding the wreckage
There was no visible damage to the corn in which I crashed, but the same cannot be said for the TBS Discovery, which sustained some comprehensive injuries.
I also had the problem of not knowing where the flyer was – I had watched where it landed, but it was now buried, utterly invisibly somewhere in a cornfield that is absolutely vast, and the surface of which looks identical all the way into the distance. The quad had landed perhaps 200m from my location in the corner of the field and at that distance, it’s hard to be precise. The best I could do was ‘probably about 100 yards further than that tree over there and maybe 50m in from the rightmost hedge’.
To scour all that area on foot would have taken hours, so I was very grateful that I had a Loc8tor Lite tracker attached to my craft, and had recently replaced its batteries ! This allowed me to walk to roughly the area in which it crashed, and then scan for the tracker’s signal, which would lead me directly to it. Didn’t get a signal at all initially, leading me to think the tracker transmitter was damaged or inoperable, but as I walked further up the field I was thrilled to see a light and hear a minimum signal beep which told me the tracker had found something.
Guided by my locator I walked on through the field (keeping to the plough lines so I didn’t disturb or crush it) and I eventually found what used to be my fully functioning quadcopter. Motors were off, as I had input in the final seconds of its flight, and it sat silently broken with various bits of it detached.
Well, the GoPro 4 survived with only a tiny scratch to its lens, something of a miracle considering that nearly everything surrounding it was severely mangled, or missing. Here’s the depressing damage from one angle following its recovery from the field…
The gimbal and left hand side of the craft took the brunt of the impact, the former being a mangled mess of twisted steel, and both left side legs smashed off entirely. This designed-in flaw looks like it worked as intended; by breaking like this they may have saved the top and bottom frames of the Discovery, and the internal FC electronics, which look fine, although of course I won’t know this until the rebuild.
UPDATE: It didn’t save the top board, which has split at one edge, and a few components got knocked off it.
Of course the left-most props broke, and I would be unwise to trust the motors that hit the dirt either, so this means I need 2 new arms, motors and possibly ESCs, though I do at least have all those, and they look undamaged.
The power pack was badly dented and puffed, and although it didn’t catch fire I’ve retired it, as there is a risk that it will later. I have also lost 2 landing gear extensions (which I couldn’t find), and had smashed my external battery voltage alarm into many pieces. The main wreck, and hardest thing to rebuild will be the camera gimbal and its amazing vibration absorption components. I have yet to work out which bits I must replace, but hopefully the TBS website will provide me with all the spares I could need, which was was one of the main factors involved in choosing the Discovery Pro as an AV platform.
What caused that then ?
I don’t know is the short answer. Perhaps DJI do. I am fairly certain that the only thing that failed here was the Flight Controller IMU, probably something to do with GPS or attitude functionality, and I have no idea why it did so. None of the conditions here ring any warning bells based on previous flyaway reports (mine or other people’s), of which I have viewed a great many. All ground checks were passed, the battery didn’t fail, nor did anything come loose or any connection fail (as far as I can see).
One suspect is perhaps the Devo 10 TX – did I lose control signal to the craft ? I don’t think so, as it responded to my yaw input, and my cutting of the motors milliseconds before it hit the ground. I wouldn’t have yaw control if I had lost signal and triggered DJI’s Return To Home failsafe, which was set to be active on control loss.
Did the GPS disc come off its mast ? It was certainly off after the crash ! Again, I don’t think so. The craft wasn’t particularly high, and I had line of sight constantly, and I didn’t see the disc flapping around. This also wouldn’t explain the throttle to 100% that the video shows as control is lost.
Another factor might be compass calibration. I haven’t calibrated my compass since the beginning of this season’s flying – perhaps 10 flight sessions ago. DJI say you don’t need to do it before every flight, although some users think you should. I don’t think it’s necessary, or a cause of today’s flyaway based on the perfect control I had in the 2 minutes before it all went wrong; if the compass calibration was off, why did I have such perfect control initially and on every flight beforehand since the craft was built ?
Could it have been the result of geomagnetic variation, some 2.4GHz interference, or even some sort of solar-based turbulence (I didn’t check that on this occasion), or some other external influence on my control signal ? All seem unlikely.
I will probably never know precisely what caused this crash, but it displays all the symptoms of the classic DJI flyaway that users have experienced with all versions of DJI’s flight controllers since that company began. The sudden acceleration, the hard drag in one direction are typical of what flyaway victims report. Some say that a switch into manual mode can save them, and in a few cases I have seen evidence it can, but in this instance I simply did not have enough time in manual mode to even attempt to correct the orientation of the craft.
If you think you know what happened, please comment on the video (on Youtube) – happy to discuss all ideas and thoughts my fellow DJIers may have…
Aerocam down ?
It’s going to be a while before I am back in the air with the Disco. This being my second flyaway, my trust in DJI flight controllers is at an all-time low. I still have a 550 hex with a Naza M-Lite in it (never crashed or flown away so far), and the Naza M V2 in the crashed Discovery has probably survived this impact and could possibly fly again. It did survive the first flyaway, and completed maybe 20 or more perfectly controlled flight sessions since, in a variety of craft.
So I have to decide if I am happy to accept the ongoing flyaway damage risk and restriction of fly-sites imposed by using these DJI FC’s, or do I try and rebuild my Discovery Pro with a different flight controller? There are other options, but I don’t know enough about them yet.
Let the research and rebuild commence !
See you when it’s done.