Drone crash - what did I do?

[DanMan32]

I had a weird and scary issue last Saturday with my P3. Had at least one good flight, maybe two. A good way into a flight at about 250-300ft, I started seeing a tremendous roll in my FPV view. A few seconds latter it went into ATTI mode, and after that a banner about compass error.
I regained composure and started flying ATTI (fortunately winds were very light) when after a couple minutes the error went away and I was back in P-GPS mode.

P3's tend to have long "warm-up" time. Perhaps it's doing IMU/compass yaw agreement.

 

[sar104]

Simple: GPS delta. If the AC flies a few feet in what it thinks is north based on compass, but GPS says it moved east, then the compass is wrong and flight control should offset the compass dynamically.

If one of my old phones not having a compass/magnetonometer could figure out heading just by GPS when moving, so can the flight control.

In no case should the AC go flying 50mph without the pilot directing it to do so.

That's not going to work either. Heading doesn't define direction of motion for a quadcopter, so just because it moves in a certain direction doesn't mean that it is facing that way. And in any case, you are talking about detecting the error after launch. The better way to do that is just to reset the IMU yaw to the correct compass yaw.

 

[Meta4]

wow who would think a wood screw in a 2x4 would cause compass errors, thats crazy.

It's physics.
Try it with a simple magnetic compass and a nail.
Bring the nail in close to the compass and watch the needle deflection.

 

[Bigbird48]

Right I understand that but having it so sensitive on a drone that a screw stops it from taking off I think is over kill

 

[Meta4]

Right I understand that but having it so sensitive on a drone that a screw stops it from taking off I think is over kill

It's just physics, that's how a compass works - any compass.
It's not something you can turn down.

 

[Bigbird48]

It's just physics, that's how a compass works - any compass.
It's not something you can turn down.

got it thanks

 

[Kilrah]

Simple: GPS delta. If the AC flies a few feet

We were talking of determining error BEFORE takeoff.

If the AC flies a few feet in what it thinks is north based on compass, but GPS says it moved east, then the compass is wrong and flight control should offset the compass dynamically.

The way it moves relative to ground says nothing about where it is facing since a quad can move in any direction just the same unlike a fixed wing aircraft that always moves "forward" with regard to the air.
The only way it can attempt to correct itself is by giving "some" somewhat random input in the form of tilting to one side, and see what effect that has on ground speed. It already does that, but it's very slow, imprecise (wind is throwing a moving wrench into any calculations), and requires excursions that are so huge it's often too far or down before it's got a chance to use the info it was trying to gather.

 

[laurens23]

That's not going to work either. Heading doesn't define direction of motion for a quadcopter, so just because it moves in a certain direction doesn't mean that it is facing that way. And in any case, you are talking about detecting the error after launch. The better way to do that is just to reset the IMU yaw to the correct compass yaw.

Most (all) of these compass error appear due to interference on the ground. why does the drone not do an automatic (re)set the IMU during/after takeoff (for example after reaching 5m altitude)? seems like that would avoid a lot of problems.

 

[sar104]

Most (all) of these compass error appear due to interference on the ground. why does the drone not do an automatic (re)set the IMU during/after takeoff (for example after reaching 5m altitude)? seems like that would avoid a lot of problems.

I agree - that would be my strategy.

 

[Simmo]

Lets think of the down sides of doing so??
Why would staff of DJI getting paid $45m not implement this??

 

[Meta4]

Lets think of the down sides of doing so??
Why would staff of DJI getting paid $45m not implement this??

The upside for DJI would be no more incidents that could legitimately be called a flyaway.

 

[naiku]

2 - Wait until the gpsHealth is 4 or 5. This means that the FC has confidence in the Lat/Long it's computing from the GPS, IMU and magnetometer data. It's shown in the bar graph display on the Go App display and on the controller.

For those of us with a Mavic Air that does not have a screen on the controller, would it make sense to wait until at least 10 satellites are locked before taking off? Usually I'll wait until there are 12 at a minimum, trying to determine the correlation between GPS health of 4 or 5 and number if satellites.



I'm curious if DJI could implement something in the firmware to prevent the suicidal fly away. Essentially something where if the compass and IMU are fighting each other, that instead of going nuts, the drone hovers in place and a warning appears on screen requiring some kind of user input to clear it and land the drone safely. While I'm sure it's not that simple, if the flight logs show issues with the compass, surely the firmware could detect that and force the drone into some kind of failsafe mode. It may not help with all instances, but I'd imagine it would cut down on them.

 

[BudWalker]

It's just physics, that's how a compass works - any compass.
It's not something you can turn down.

Actually, it could be turned down. But, then it wouldn't hover or navigate properly.

 

[sar104]

For those of us with a Mavic Air that does not have a screen on the controller, would it make sense to wait until at least 10 satellites are locked before taking off? Usually I'll wait until there are 12 at a minimum, trying to determine the correlation between GPS health of 4 or 5 and number if satellites.



I'm curious if DJI could implement something in the firmware to prevent the suicidal fly away. Essentially something where if the compass and IMU are fighting each other, that instead of going nuts, the drone hovers in place and a warning appears on screen requiring some kind of user input to clear it and land the drone safely. While I'm sure it's not that simple, if the flight logs show issues with the compass, surely the firmware could detect that and force the drone into some kind of failsafe mode. It may not help with all instances, but I'd imagine it would cut down on them.

It can't hover in place when those errors occur - that's the problem. The failsafe action is to switch to ATTI mode, which often happens, but doesn't always help if the aircraft is then blown away by the wind or, as is often the case, the pilot has no idea how to control the aircraft in ATTI mode.

 

[BudWalker]

For those of us with a Mavic Air that does not have a screen on the controller, would it make sense to wait until at least 10 satellites are locked before taking off? Usually I'll wait until there are 12 at a minimum, trying to determine the correlation between GPS health of 4 or 5 and number if satellites.
......

A gpsHealth of 4 or 5 requires an adequate number of satellite. But, that's not sufficient - just having 16 satellites doesn't guarantee gpsHealth of 4 or 5. There can be other issues.

The Go App also has the bar graph display that shows the gpsHealth.

 

[naiku]

It can't hover in place when those errors occur - that's the problem. The failsafe action is to switch to ATTI mode, which often happens, but doesn't always help if the aircraft is then blown away by the wind or, as is often the case, the pilot has no idea how to control the aircraft in ATTI mode.

Ah OK. So, ultimately when the error occurs, the drone switches to ATTI mode but at the same time is already off on it's suicide mission, which typically by then it is too late for the pilot to regain control (or as you said, knows how to control it in ATTI mode). Surely though, if it knows to switch to ATTI mode, can it not also act as if the control sticks were in the neutral position? Take OP's crash for instance, it took off, switched to ATTI mode and zoomed off at full throttle before crashing, what triggered it to fly off at full throttle? (is this a result of the compass / magnetic error and ATTI mode causing the drone to want to get to where it thinks it should be?) Clearly not the pilot, so what is preventing the software from as it switches to ATTI mode also preventing full throttle?

Apologies if this is an obvious answer, I would definitely fall under one of those pilots who has no idea how to use ATTI mode. I have never used it, only had the drone a couple weeks and won't fly it unless I get a good GPS lock. Just trying to figure out a way that may prevent some of these type of crashes.

 

[BudWalker]

Most (all) of these compass error appear due to interference on the ground. why does the drone not do an automatic (re)set the IMU during/after takeoff (for example after reaching 5m altitude)? seems like that would avoid a lot of problems.

The strategy used to be something like this. The P3 would switch to ATTI mode at the first indication of a "compass error". The quotes are because the FC can't know if it's the compass or the IMU which is incorrect. Simply resetting the Yaw value to the heading derived from the magnetometers isn't a good idea because the magnetometers could be incorrect.

This strategy prevented a lot of fly aways. But, it also caused a lot of false positives. Seemed like every FW update would produce more false positives in an attempt to reduce fly aways. I can recall incident after incident where the answer was "switched to ATTI mode for no apparent reason, don't worry, keep flying".

About the time the Mavic Pro came out the strategy was changed to try and reconcile these errors. Sometimes it works and sometimes it doesn't.

 

[BudWalker]

Ah OK. So, ultimately when the error occurs, the drone switches to ATTI mode but at the same time is already off on it's suicide mission, which typically by then it is too late for the pilot to regain control (or as you said, knows how to control it in ATTI mode). Surely though, if it knows to switch to ATTI mode, can it not also act as if the control sticks were in the neutral position? Take OP's crash for instance, it took off, switched to ATTI mode and zoomed off at full throttle before crashing, what triggered it to fly off at full throttle? (is this a result of the compass / magnetic error and ATTI mode causing the drone to want to get to where it thinks it should be?) Clearly not the pilot, so what is preventing the software from as it switches to ATTI mode also preventing full throttle?

Apologies if this is an obvious answer, I would definitely fall under one of those pilots who has no idea how to use ATTI mode. I have never used it, only had the drone a couple weeks and won't fly it unless I get a good GPS lock. Just trying to figure out a way that may prevent some of these type of crashes.

I've seen incidents where it's clear that the FC is trying to recover even after it's switched to ATTI. Sometimes the switch was to a mode where it's using the vision system data and not the GPS data. BTW, flyCState is still GPS+ATTI - how confusing is that? Anyway, I think this is a deliberate strategy on DJI's part to reduce the number of unexplained compass error/switch to ATTI incidents.

I'm of the opinion that the old way is really better. Maybe a case of I'm-old-it's-new-I-don't-like-it. But, I'm also thinking that here is a world class tech and engineering behemoth and in comparison I don't know all that much about drone FC.

 

[DanMc]

...that when the Unit crashed the battery flew about 15ft from point of impact, which was a poor souls granite tombstone....that by the time I was able to run and start picking up the parts...the battery had ignited in flames and started a small fire.

Wow, crash into a tombstone and burn. There has to be a spiritual lesson in there for all of us. Know where you're going!

 

[Lonjsnyder]

I am positive my AC was in ATTI mode - probably should not have been, but I used it like that previously with no issues.
The concern I have now with this Mag interference issue - and as Meta4 noted, that it's probably too late once you notice it....
This is a bit concerning with a $1,400 unit and a crash a near certainty. If the lack of control I had is the result of the interference, then what would we be looking for at take -off? I only ask that because it appeared I had a "all green lights"
and ready for take-off. Again, it got in the air and it was gone until it hit the tombstone.
would anyone send this back to DJI? Or do I have an operator issue and contemplate purchase of another unit?

Personally, I prefer to check for magnetic interference independently of the Go4 App .... see my post #37 in this thread:

Phantom Rain : Magnetic Interference is the Phantom Killer in disguise:

Here is the final answer from DJI for the difference between P4P 1.0 and 2.0. In conclusion, my P4P v2.0 is working as it should.

r.tapatalk.com