Strange drone behaviour resulting in a crash

[BudWalker]

Thanks for your reply. I'll get the DAT file later today when I get home. I don't think I started the drone until I got to the beach as I'm usually very careful about battery usage.

The .DAT we want is FLY009.DAT,

 

[maccboy]

The only DAT files I can find for that day (13/09/19) have different timestamps to the time I flew, so I've attached both.
I really appreciate any information you can give me - and for your help!

 

[maccboy]

The .DAT we want is FLY009.DAT,

Ah, sorry. Just seen this! Thanks.

 

[sar104]

I don't think I've ever seen that many compass errors in such a short event stream, but I don't think it was a compass problem:



Anyone used to looking at this plot will notice a couple of unusual features. The compass indicates that the aircraft was pointing SW (-123°) at launch, rather than 117° from IM(1), which was recorded in the txt log. What happens at 30 seconds is very interesting - the compass data indicate a CW turn, IMU1 records a CCW turn, and IMU0 doesn't notice much at all. Nor do the z-axis rate gyros, which at least agree with each other - not much else does.



The x and y rate gyros, in contrast, disagree from the start:



None of which completely explains anything. While this obviously needs a closer look, my initial hypotheses would be either hardware (IMU) failure or the FC lost its mind.

 

[hiflyer201]

Sorry, bummer for sure but ..sport just to climb?...not enough advantage and it’s impossible to get good video or frame decent shots in sport..but then you know that. I’ve had two “no damage” but uncontrollable crashes with my Spark after taking off from a smooth rock and learning later it was concrete full of rebar.
In both cases it just flew off instantly and luckily hit a wall before getting too high. Both accidents were carbon copies and six months apart in a foreign country. I guess you can say I’m a slow learner

 

[BudWalker]

I don't think I've ever seen that many compass errors in such a short event stream, but I don't think it was a compass problem:

View attachment 81889

Anyone used to looking at this plot will notice a couple of unusual features. The compass indicates that the aircraft was pointing SW (-123°) at launch, rather than 117° from IM(1), which was recorded in the txt log. What happens at 30 seconds is very interesting - the compass data indicate a CW turn, IMU1 records a CCW turn, and IMU0 doesn't notice much at all. Nor do the z-axis rate gyros, which at least agree with each other - not much else does.

View attachment 81898

The x and y rate gyros, in contrast, disagree from the start:

View attachment 81900

None of which completely explains anything. While this obviously needs a closer look, my initial hypotheses would be either hardware (IMU) failure or the FC lost its mind.

I think you might have meant that IMU(1):Yaw (not IMU(0):Yaw) didn't change much? I think what might have happened is that navHealth reached 4 and the FC started to reconcile the Yaw/magYaw separation. If, as you pointed out, IMU(1) was active then the FC would hold IMU(1):Yaw constant while it rotated the AC. IMU(1):totalGyroZ and IMU(1):magYaw both increase while IMU(1):Yaw is held constant.

There are some significant problems with this though. There is that magYaw dip at 30 secs that doesn't appear in the totalGyroZ and the totalGyroZ rate of change is about 4 times that of magYaw. Also, Yaw isn't held exactly constant as we've seen in other incidents - I attribute this to the fly away that was in progress.

There is something odd about the magnetometer data. MagMod starts around 1500 but then settles around 880 when the MA is placed on the ground prior to the launch. There is a small return towards a normal value as the MA ascends at about 5 secs. It's not until the MA reaches 26 meters altitude before the magMod returns to normal.

It's almost as if the entire area had a lower field strength close to the ground, except for the fact the field strength was normal while the MA was being handled prior to launch.

@maccboy was something attached to the MA after it was powered up and prior to launch?

This quite the puzzler.

 

[Northwood Mediaworks]

I struggle to think that the sensor state was normal before actually lifting off.

 

[maccboy]

I think you might have meant that IMU(1):Yaw (not IMU(0):Yaw) didn't change much? I think what might have happened is that navHealth reached 4 and the FC started to reconcile the Yaw/magYaw separation. If, as you pointed out, IMU(1) was active then the FC would hold IMU(1):Yaw constant while it rotated the AC. IMU(1):totalGyroZ and IMU(1):magYaw both increase while IMU(1):Yaw is held constant.
View attachment 81927
There are some significant problems with this though. There is that magYaw dip at 30 secs that doesn't appear in the totalGyroZ and the totalGyroZ rate of change is about 4 times that of magYaw. Also, Yaw isn't held exactly constant as we've seen in other incidents - I attribute this to the fly away that was in progress.

There is something odd about the magnetometer data. MagMod starts around 1500 but then settles around 880 when the MA is placed on the ground prior to the launch. There is a small return towards a normal value as the MA ascends at about 5 secs. It's not until the MA reaches 26 meters altitude before the magMod returns to normal.
View attachment 81929
It's almost as if the entire area had a lower field strength close to the ground, except for the fact the field strength was normal while the MA was being handled prior to launch.

@maccboy was something attached to the MA after it was powered up and prior to launch?

This quite the puzzler.

Thanks for all for your replies. There wasn't anything attached to the drone. It's a complete mystery. I did take off from a large rock, though, which mught sort of tie in with the post about doing the same thing.

 

[sar104]

I think you might have meant that IMU(1):Yaw (not IMU(0):Yaw) didn't change much? I think what might have happened is that navHealth reached 4 and the FC started to reconcile the Yaw/magYaw separation. If, as you pointed out, IMU(1) was active then the FC would hold IMU(1):Yaw constant while it rotated the AC. IMU(1):totalGyroZ and IMU(1):magYaw both increase while IMU(1):Yaw is held constant.
View attachment 81927
There are some significant problems with this though. There is that magYaw dip at 30 secs that doesn't appear in the totalGyroZ and the totalGyroZ rate of change is about 4 times that of magYaw. Also, Yaw isn't held exactly constant as we've seen in other incidents - I attribute this to the fly away that was in progress.

There is something odd about the magnetometer data. MagMod starts around 1500 but then settles around 880 when the MA is placed on the ground prior to the launch. There is a small return towards a normal value as the MA ascends at about 5 secs. It's not until the MA reaches 26 meters altitude before the magMod returns to normal.
View attachment 81929
It's almost as if the entire area had a lower field strength close to the ground, except for the fact the field strength was normal while the MA was being handled prior to launch.

@maccboy was something attached to the MA after it was powered up and prior to launch?

This quite the puzzler.

Yes - thanks for spotting and pointing out that I reversed the IMUs in that part of the description. I hadn’t checked the magnetic modulus - that is curious. The discrepancies between the x and y rate gyros is the biggest problem, in my view. It leads to huge pitch and roll discrepancies between the two IMUs, including one of the IMUs indicating that the aircraft was flying upside down later in the flight.

 

[Skywalker]

Thanks for all for your replies. There wasn't anything attached to the drone. It's a complete mystery. I did take off from a large rock, though, which mught sort of tie in with the post about doing the same thing.

Were you in a foreign country? Was the large rock smooth?

 

[rfc]

Were you in a foreign country? Was the large rock smooth?

OP is in Ravenglass, in Cumbria, UK...NW UK, on the Irish Sea. Don't know about the rock. Maybe the takeoff point was over buried Roman weaponry.

 

[DanMan32]

Rocks can contain metal.

 

[Skywalker]

OP is in Ravenglass, in Cumbria, UK...NW UK, on the Irish Sea. Don't know about the rock. Maybe the takeoff point was over buried Roman weaponry.

For clarification, the comment was in jest based off hiflyer201’s post above. I enjoyed his/her humor.

 

[maccboy]

Thanks again. As said, Ravenglass is in the Lake District in England. The rock was smooth - and I will certainly avoid taking off from a large rock again! I've really impressed with the information that can be recreated from the files by you posters. It's amazing!

 

[rob levine]

Thanks for all the replies. I did find it with just a broken prop. I was on a beach so there was nothing obvious that could have caused a problem with interference. As I said, I was aiming to fly vertically upwards, take some shots, and fly back down. It wasn't really that big a deal, in my view.
I was hoping someone could discover what the cause of the strange flight path was from the data. I did nothing to create the problem. Flying without GPS wasn't the cause, inference wasn't the cause, flying GoVision wasn't the cause. The sensors were fine. That's it, really!

Yeah but there is no RTH without GPS

 

[maccboy]

Yeah but there is no RTH without GPS

I realise that but, to be honest, it's not essential for a straight up and straight down flight

 

[wsteele]

I realise that but, to be honest, it's not essential for a straight up and straight down flight

There was so much else going on with this flight I don't think my comment are that relevant I disagree with you that GPS is not essential to straight up and straight down flight. I guess to a degree you would get vision positioning; but, no GPS and any wind at altitude would cause the drone to drift off and with no homepoint, no reference to bring it back to, it easy to get into a lot of trouble even if the rest of your drone systems are working perfectly. I heard very few experienced OPs not state they wait for a green GPS and a homepoint established before they get more than 10 feet or so above the ground.

 

[kidroc]

I strongly disagree!! Not to be rude or anything I see your ac had a lot going on with it but lets say it was fine and had no issues.. The vision system at best works till what 30ft if you had no gps and you knew it then I would expect it to go at some point into atti mode and be at the mercy of the wind and my skills as a pilot. Like i said not trying to be mean but that's what I would expect to happen to a experienced pilot live and learn and buy a cheap quad and at least learn to fly it without any aids besides self leveling.

 

[Kazzaaz]

I strongly disagree!! Not to be rude or anything I see your ac had a lot going on with it but lets say it was fine and had no issues.. The vision system at best works till what 30ft if you had no gps and you knew it then I would expect it to go at some point into atti mode and be at the mercy of the wind and my skills as a pilot. Like i said not trying to be mean but that's what I would expect to happen to a experienced pilot live and learn and buy a cheap quad and at least learn to fly it without any aids besides self leveling.

I frequently take off without GPS and note the home point when as locks more satellites in the air. Never had any problems.

Everyone should have learned how to fly in ATTI mode in Drones 101.
Unless you're flying in a windstorm, taking the drone up and down under VLOS should be a piece of cake.
And if you aren't confident in your ability to do this then I honestly recommend to practice right now before you lose GPS lock mid-flight and drift into a tree.

 

[Maverick23]

Even going straight up without gps is a bad idea. Unless you are super confident in flying and dealing with all the correcting movements needed for the wind, you need GPS to stay locked in place or else the copter will just drift away. The computer needs either the GPS coordinates or expert manual input from you initiate the required movements to keep the copter staying in place. It’s always fighting some force of the wind