SMALL - espritmodel.com SMALL - Telemetry SMALL - Radio
Reply
Thread Tools
Old Mar 22, 2014, 09:57 PM
Registered User
Joined Oct 2013
699 Posts
Discussion
Never thought it would happen, but I've been hit with a Naza fly-away

My Background
I've been flying for over a year this point. Not a grizzled old veteran, but I consider myself experienced (when I get into something, I'm a bit obsessive/neurotic about it and really go deep.) Started with a Phantom, built a tricopter, hex, long-flight quad and X8. I've crashed and rebuilt parts of most of those craft as well. I own and fly with most flight controllers: Naza, KK2, Multiwii, Megapirate, and multiple APM boards. Most of my crashes are either pilot error (controlled flight into trees mostly) or equipment failure (losing a motor clip and having the motor casing fly off, bad ESC firmware not syncing with motor and cutting out mid-flight.)

The Fly-Away
When the fly-away happened, it scared the sh*t out of me. You really don't have a lot of time to react, so it's a harrowing minute or so of sheer terror. Here's the basic background and sequence of events:

Happened with the Naza V2 on my X8 with FrSky module (with telemetry) on 9XR radio using ER9X firmware. I've flown this craft many times in the same area and have had no prior problems. Good GPS before arming and taking off. I had 8 satellites reported by my OSD before I armed it. Armed the craft and took off in GPS mode. At first the X8 seemed to drift a little bit more than normal, but nothing far out of the ordinary. Felt like it was drifting in the wind a bit backward, so I had to give it a bit of forward pitch from time to time to get it to stay in place. Slowly, over the course of probably 10 seconds, the drifting turned into an absolute sprint. It basically took off in a single direction and no amount of stick could change its course. It was headed straight for a tree, so I gave it a bunch of throttle and it dutifully shot up and avoided the tree. It was basically going like a bat out of hell in one singular direction. At this point, I flip it to ATTI mode. Absolutely no change, still could not control it at all. When ATTI was not doing anything, I flipped it to manual. The X8 immediately started reacting to all stick input. Unfortunately, I could not recover it to stable flight after losing orientation, so it eventually crashed, coming down from about 20-30 meters up.

Had I been able to reorient myself and get the craft into a hover, I would have been able to recover completely with manual.

I was extremely lucky that I didn't damage anyone else's property. My X8 took a ridiculous high speed powered flight straight to the ground. I was fortunate that a tree broke the fall and I only ended up with a few broken props, broken arm mount and damaged my gimbal's yaw motor.

So, what DIDN'T cause it?
Props, motors and ESCs were not at fault. Apart from some broken props after I rebuilt it, everything was attached still and worked after replacing the broken props and arm mount.

Flight controller was still solidly stuck dead center of the CoG. The GPS unit and mast surprisingly was not broken from the crash. So, I could tell that GPS/compass module was still oriented correctly.

Interference from some other 2.8Ghz source? Unlikely. No wifi enabled on my GoPro. The FrSky would beep at me if it was dropping the signal. Besides, the craft was only about 5 meters in the air and only 10 meters away from me when it decided to fly away. There were no cellular towers anywhere nor any large metal pipes or power lines. I was flying in a greenbelt area surrounded by water, though it is located in a residential area. Also, the craft responded to _some_ stick (like throttle), it didn't respond to pitch/roll at all. As soon as I flicked it to manual, everything responded. Even after the crash, I was able to disarm the motors from over 100 meters away, with the craft obscured behind trees and a fence. I only heard the FrSky module beep at me briefly when I ran into my car to drive to the crash site.

Power dropout? Also unlikely. I fly with two batteries in parallel and have a separate BEC for the gimbal+video transmitter, and another BEC for charging the GoPro. The only thing that runs from the Naza PMU is the Naza FC itself and the FrSky receiver. I fly with a battery alarm on each of the 4s leads. Good power at takeoff and it flew away after only about 30 seconds. After the crash, the batteries were still reporting good voltage.

Fail-safe was triggered and it was just flying home to a bogus home point? Unlikely. I didn't have my Fatsharks on after the fly-away so didn't look at the OSD, so I can't be 100% sure that it didn't trigger the fail-safe. I don't think this was the case though, given the fact that I could still control the throttle. I've triggered fail-safe before on this craft and have not been able to give it any sort of control input until I toggled flight modes. I also don't remember the craft flying this fast when the RTH fail-safe was triggered. If RTH is triggered, it usually stops where it is, gains altitude if it's too low (it should have, it was only 5 meters high at fly-away), then flies home at a leisurely pace. This fly-away was just "I'm going to fly in this direction whether you like it or not."

So, speculate wildly then about the cause then
The Naza FC could be getting spurious GPS/compass data. There's nothing to indicate this actually happened, but I personally think it's not enough to explain the behavior. Even if you get bad GPS data, eventually, I would expect it to get valid data and hopefully recover. The X8 was also just flying away in one straight direction, if the GPS data were garbled, I would expect some gyration and change in flight direction.

To me, it feels like a bug in the Naza FC firmware. It gets some bad data then locks itself into a loop where it flies in a fixed direction. I was still able to control the altitude with the throttle, so it was not a complete shutdown and the FC was still running its control loop. It just seems like whatever GPS or ATTI routine that it gets stuck in there's really no recovering without disregarding the GPS and compass data entirely.

The other possibility is that the bug in the Naza FC is in hardware or the software where it's getting a stuck command. It felt like the Naza thought my roll/pitch stick was stuck in one direction.

OK, so what can DJI do to improve things?
First and foremost, I'd love to see an new flight mode that was just flight stabilized and did not rely on GPS or compass. Basically, something like Arducopter's stabilize mode. At least with this mode available, it would make recovery much easier. It would be an improved manual mode where if you release the sticks, it would still level itself. I really don't think manual mode is very valuable to the majority of pilots. If you want to do flips or crazy maneuvers, you're not flying a DJI FC.

I would love it if DJI could find and fix this bug, but I'm not sure they've ever acknowledged there being any sort of problem with fly-aways. Probably owing to legal issues. If they admit that they have a bug in their FC, all those lost dollars and damaged property (and maybe injured people) could sue them.

Short of them admitting to a possible bug, they should build in some more fail-safe to prevent this type of behavior. They tried to do that with the advanced receiver protection thing. But I think they didn't fix the right problem. They assumed that the receiver was inundated by a strong signal that swamped actual radio controls. In my case, that was almost certainly not the case. There's a bug in their FC that imitates this supposed condition, but the bug is with the FC itself (either in the hardware or the software.)

I almost didn't write this post because I know that it will inevitably become a contentious point. If you reply, please keep it constructive.
LoneStriker is offline Find More Posts by LoneStriker
Reply With Quote
Sign up now
to remove ads between posts
Old Mar 22, 2014, 10:09 PM
Registered User
_Milo_'s Avatar
United States, ID, Pocatello
Joined Apr 2013
512 Posts
Did you check the IMU values, anything unusual?
_Milo_ is online now Find More Posts by _Milo_
Reply With Quote
Old Mar 22, 2014, 10:17 PM
Registered User
Joined Aug 2013
128 Posts
There is no perfect answer to your question. It happens. It even happens to Boeing aircrafts. As an engineer, you can't expect a perfect system so it's you own decision to take the risk of flight. I am glad that it did not caused an unwanted damage, i hope it did not happen again.
mustafa is offline Find More Posts by mustafa
Reply With Quote
Old Mar 22, 2014, 10:25 PM
My RC builds in my profile.
daign's Avatar
United States, CA, Los Angeles
Joined Feb 2011
2,568 Posts
flipping from GPS/RTH back to ATTI would have easily regained control.

I suspect the info provided. Had you fully tested both the Naza RTH failsafe successfully multiple times and also tested it by killing your transmitter. I see no mention of the naza calibration dance.

90% of naza 'flyaways' are setup errors. I know because I've had it happen myself. I had a flyaway because I had yet to do the naza gps calibration on a new hexa. A simple switch from failsafe/GPS back to ATTI and voila, full control again. Calibrated the GPS and it flew home perfectly. DESPITE this naza being from a similar quad in the exact same flight location.

So what eliminated you from flipping a switch and flying home in ATTI? Not trying to be antagonistic, just reading the facts.
daign is online now Find More Posts by daign
Reply With Quote
Old Mar 22, 2014, 10:31 PM
FPV Racer - N4FPV
Eddie James's Avatar
United States, GA, Kennesaw
Joined Sep 2009
858 Posts
I don't know if we are reading the same thing. I just read above that he said he tried ATT to no avail. Did you read that?
Eddie James is offline Find More Posts by Eddie James
RCG Plus Member
Latest blog entry: CavCam, Inc.
Reply With Quote
Old Mar 22, 2014, 10:50 PM
My RC builds in my profile.
daign's Avatar
United States, CA, Los Angeles
Joined Feb 2011
2,568 Posts
Quote:
Originally Posted by Eddie James View Post
I don't know if we are reading the same thing. I just read above that he said he tried ATT to no avail. Did you read that?
I read it. It must be flipped from RTH to ATTI to GPS and back to ATTI if I recall. I myself experienced that same panic, it required that extra flip to cancel RTH. It is described in the manual. A simple one click flip of the switch from GPS to ATTI won't do it while the controller is in RTH failsafe.

http://wiki.dji.com/en/index.php/Pha...ss_Calibration
daign is online now Find More Posts by daign
Reply With Quote
Old Mar 22, 2014, 10:52 PM
My RC builds in my profile.
daign's Avatar
United States, CA, Los Angeles
Joined Feb 2011
2,568 Posts
Also depending on how OP has his failsafe setup, if its on another switch, that failsafe/panic switch would override the flight mode switch, making his controller stuck in RTH. I had this setup on my aurora 9, until I realized that issue. On my taranis I have it set to one switch now.
daign is online now Find More Posts by daign
Reply With Quote
Old Mar 22, 2014, 10:57 PM
Registered User
Germany, HE, Darmstadt
Joined Aug 2013
421 Posts
I dont understand why you used the trim buttons on the remote, when the remote is calibrated with the software there is no need of that and it makes no sense in my mind because the quadcopter will drift randomly in any direction maybe cause of wind or decalibrated compass etc.. So that means a minute later the copter could be drifting in a totally different direction and you had to constantly trim the whole flight? In the naza software you can calibrate the remote and i think you can calibrate the remote as well with a software, i calibrate before every flight
Daninho is offline Find More Posts by Daninho
Reply With Quote
Old Mar 22, 2014, 11:46 PM
Registered User
Joined Oct 2013
699 Posts
Quote:
Originally Posted by _Milo_ View Post
Did you check the IMU values, anything unusual?
I have never had to check IMU values. Again, the craft flies fine normally. It flew fine before the fly-away, it flies fine now after the fly-away. I always let the IMU warm up and the GPS lock before flying. The Naza assistant has never indicated that I needed to do any IMU calibration after the initial setup.

Quote:
Originally Posted by mustafa View Post
There is no perfect answer to your question. It happens. It even happens to Boeing aircrafts. As an engineer, you can't expect a perfect system so it's you own decision to take the risk of flight. I am glad that it did not caused an unwanted damage, i hope it did not happen again.
Can't really compare passenger aircraft and personal multirotors. Of course things fail, but in passenger aircraft, there are multiple redundant systems and backups to those systems. For FCs, we know that there is a single point of failure with most of the current setup. I go in knowing all of this. What I don't expect, though, is to have a bug that's been around for a while not being addressed.

Quote:
Originally Posted by daign View Post
flipping from GPS/RTH back to ATTI would have easily regained control.

I suspect the info provided. Had you fully tested both the Naza RTH failsafe successfully multiple times and also tested it by killing your transmitter. I see no mention of the naza calibration dance.

90% of naza 'flyaways' are setup errors. I know because I've had it happen myself. I had a flyaway because I had yet to do the naza gps calibration on a new hexa. A simple switch from failsafe/GPS back to ATTI and voila, full control again. Calibrated the GPS and it flew home perfectly. DESPITE this naza being from a similar quad in the exact same flight location.

So what eliminated you from flipping a switch and flying home in ATTI? Not trying to be antagonistic, just reading the facts.
I've done the calibration dance many times. I don't do it before every flight like some people do. There have been reports of fly-aways even after religious dancing.

I never said I activated the RTH feature for the fly-away. In fact, I discounted that as a possibility. I have my RTH/fail-safe setup in two ways:

1. On a switch I can trigger manually on my radio.
2. As the default flight mode if I turn off my radio.

I've tested both. They work. I know how to regain control after RTH is triggered. I've also done this many times and have tested it as well. I explicitly stated that the fly-away WAS LIKELY NOT RTH. It didn't behave like RTH. I listed it as a possibility, but one that was almost entirely discountable.

Quote:
Originally Posted by Eddie James View Post
I don't know if we are reading the same thing. I just read above that he said he tried ATT to no avail. Did you read that?
This is correct. I flipped it to ATTI and it behaved the same as GPS. Only manual kicked it out of the fly-away mode.

Quote:
Originally Posted by daign View Post
I read it. It must be flipped from RTH to ATTI to GPS and back to ATTI if I recall. I myself experienced that same panic, it required that extra flip to cancel RTH. It is described in the manual. A simple one click flip of the switch from GPS to ATTI won't do it while the controller is in RTH failsafe.

http://wiki.dji.com/en/index.php/Pha...ss_Calibration
Again, I know what RTH is. I've triggered it and have had it trigger by itself. I know how to regain control after RTH, This fly-away was not RTH.

Quote:
Originally Posted by daign View Post
Also depending on how OP has his failsafe setup, if its on another switch, that failsafe/panic switch would override the flight mode switch, making his controller stuck in RTH. I had this setup on my aurora 9, until I realized that issue. On my taranis I have it set to one switch now.
Not the case. I have it setup on a switch as well as my failsafe if the receiver loses power. RTH would have been useless here if both GPS mode and ATTI mode were not behaving correctly, there's no chance RTH would actually work as it relies on the same subsystems as GPS.

Believe it or not, there are rational, clueful people on these boards (in addition to the newbies and idiots.) The reason I hesitate to post yet another fly-away thread is that everyone basically jumps to the conclusion that it's pilot error. I've already gone through all the conceivable alternatives and there's nothing that I suspect other than a problem with the FC itself.

I program and work with complex systems as my day job, so I'm very technically savvy. To me, the root of this problem seems to be an error in the FC itself.
LoneStriker is offline Find More Posts by LoneStriker
Reply With Quote
Old Mar 22, 2014, 11:51 PM
Registered User
Joined Oct 2013
699 Posts
Quote:
Originally Posted by Daninho View Post
I dont understand why you used the trim buttons on the remote, when the remote is calibrated with the software there is no need of that and it makes no sense in my mind because the quadcopter will drift randomly in any direction maybe cause of wind or decalibrated compass etc.. So that means a minute later the copter could be drifting in a totally different direction and you had to constantly trim the whole flight? In the naza software you can calibrate the remote and i think you can calibrate the remote as well with a software, i calibrate before every flight
Not sure who this is directed at, but I certainly would never use trim on my radio for Naza or APM controllers. About the only time I've ever used trim was with something like the KK2 board as there is no compass or GPS for it to keep the craft's orientation stable. If your ESCs and radio are properly calibrated, and the FC setup correctly along with its sensors, there's no reason for trim.

The drift I mentioned in my post was the X8 moving around like it wasn't locked in GPS-wise after I took off. There is sometimes a minute amount of drifting for the X8, but usually only if there's enough wind to push it around. In this case, there was little to no wind, so the drifting was likely due to GPS errors slowly accumulating (my speculation.)
LoneStriker is offline Find More Posts by LoneStriker
Reply With Quote
Old Mar 23, 2014, 05:36 AM
Registered User
United Kingdom, Bozeat
Joined Jan 2007
249 Posts
Quote:
Originally Posted by LoneStriker View Post

"Felt like it was drifting in the wind a bit backward, so I had to give it a bit of forward pitch from time to time to get it to stay in place. Slowly, over the course of probably 10 seconds, the drifting turned into an absolute sprint. It basically took off in a single direction and no amount of stick could change its course. It was headed straight for a tree, so I gave it a bunch of throttle and it dutifully shot up and avoided the tree. It was basically going like a bat out of hell in one singular direction."
It's unlikely and extremely rare, but I would not completely rule out an intermittent Tx fault. I have witnessed similar situations a couple of times in the last 2 years, when the problem turned out to be a mechanical or electrical Tx malfunction. (In neither case was the Tx FrSky or Futaba BTW)
Martin Bell is offline Find More Posts by Martin Bell
Reply With Quote
Old Mar 23, 2014, 06:21 AM
Registered User
Greenland
Joined Dec 2013
126 Posts
flying march 22 2014 (2 min 1 sec)
in my video min 1:20 you see my Naza M v2 run off.
It started to run off when changed from GPS to ATT soon after lift off. Killed it with manual mode and then back to ATT. No problems after that and continued flying until battery end.
rotomotor is offline Find More Posts by rotomotor
Reply With Quote
Old Mar 23, 2014, 06:24 AM
Registered User
Joined Aug 2013
128 Posts
Can't really compare passenger aircraft and personal multirotors. Of course things fail, but in passenger aircraft, there are multiple redundant systems and backups to those systems. For FCs, we know that there is a single point of failure with most of the current setup. I go in knowing all of this. What I don't expect, though, is to have a bug that's been around for a while not being addressed.

--> Well of course my intention is not to compare the passenger aircrafts and multi rotors but to mention that even the million dollar FCs can fail. I must say, this devices and their FCs are really really complex than we think. So it's not really something like address the bug
mustafa is offline Find More Posts by mustafa
Reply With Quote
Old Mar 23, 2014, 07:37 AM
R/C Addict
United States, MA, Boston
Joined Feb 2014
331 Posts
Yep always the naza. It has a bug that dji refuses to fix
EcoAerialImaging is offline Find More Posts by EcoAerialImaging
Reply With Quote
Old Mar 23, 2014, 08:26 AM
jab
Unregistered User
Joined Aug 2007
1,081 Posts
Quote:
Originally Posted by Martin Bell View Post
It's unlikely and extremely rare, but I would not completely rule out an intermittent Tx fault. I have witnessed similar situations a couple of times in the last 2 years, when the problem turned out to be a mechanical or electrical Tx malfunction. (In neither case was the Tx FrSky or Futaba BTW)
Why would it then suddenly start responding normally to stick input again, at the exact moment he switched to manual mode?
jab is offline Find More Posts by jab
Reply With Quote
Reply


Thread Tools

Similar Threads
Category Thread Thread Starter Forum Replies Last Post
Discussion I've been away from the forums and RC for 8 years. So many questions! Ampnoise Electric Plane Talk 17 Feb 24, 2014 03:39 PM
Discussion I've been a bad boy......Again... But this one has more room for boats pops52 Dock Talk 32 Oct 28, 2013 12:31 AM
Discussion Did you get rain? (would've been "Got Rain?!" but the rain stopped, thank God) spacephrawg Dock Talk 10 Mar 17, 2010 08:38 AM
I never thought this would happen to me.. snowball1288 Electric Plane Talk 33 Aug 08, 2005 01:35 AM
I never thought THIS would happen Bleriot Foamies (Kits) 18 Apr 26, 2001 02:57 PM