AACY Posted at 2017-4-25 23:52
You refer to this ?

"The log shows the Mavic with a heading of 164 at launch but it was actually closer to 0 as shown by the video. Also, direction of flight was close to 0 but with no aileron or rudder input. These are clear signs that the compass was compromised by the geomagnetic distortion. After reaching a height of 1 meter the Mavic was clear of the geomagnetic distortion and the compass then had the correct heading. However, the IMU still has the incorrect value. It's at this point the compass error is declared."

Not sure what you mean by axis lock. Maybe you're referring to ATTI mode where the AC isn't trying to navigate; it's just listening to the IMU and will drift with the wind. If so, you're right the Mavic didn't switch from GPS+ATTI to ATTI mode. This appears to be "normal" with the Mavic. The Phantom3 would have switched to ATTI. The stratgey with the Mavic appears to be to stay in GPS+ATTI mode and attempt to fix the problem. I've looked at several flights where this has occurred. It actually works sometimes. That's what happened in this flight.
The beige/tan background is the interval where there is a compass error. It starts at about 14 secs where magYaw and Yaw differ by about 150 degrees. Following this Yaw remains constant but magYaw is changing to become closer to Yaw. What's actually happening is that the Mavic is being rotated so that the compass (magnetometers) will become aligned with the Yaw value. I.e., rather than the Flight Controller changing it's Yaw value to agree with magYaw the Mavic is rotated until magYaw shows the same value as the Yaw value. This has been confirmed with other flights where a video is available. I haven't done that confirmation with this flight.

Finally, at about 27 secs the Mavic has rotated enough so that Yaw and magYaw are close enough that the compass error is no longer necessary. Normal, stable flight should have been possible at this point. But, I'm sure the pilot was only trying to recover the Mavic without further incident.

Oleh Artyukh Posted at 2017-4-25 23:31
I don't say Mavic didn't reacted to all inputs. It did reacted to the left stick, but didn't to the right one, and you can definitely see it in the video:
https://www.youtube.com/watch?v=pYR6kD7FrKM

It may have seemed like the Mavic wouldn't respond to aileron input. I think it's probably more like you and the Flight Controller were both trying to control the Mavic. I tried to confirm by looking at the effect on the horizontal velocity. But couldn't becasue of the number of control inputs. In addition to full left aileron there was some negative elevator and throttle inputs. Add onto this what the Flight Controller was inputting. Time 26.147 is where you started applying left aileron.

rnrnrn Posted at 2017-4-26 05:10
Hi SkunkWerxs,

I read your previous discussions on this thread about the magnetic compass and thought I could offer a bit of a clarification on the matter.

rnrnrn ,
           I'm starting to love this op's thread it's making people Think of the operating circuitry of the Mavic that gives a better understanding of there system that they paid for loll, now for my opinions of your post
           "Declination "  Is used to determine the area your in, to get True North cowardanat along with GPS for the GO4 app Telemetry especially when you use any autonomous mode.
            As for something you don't know ---- When you are flying manually-- Declination is totally unimportant to the pilot because he is using his stick inputs to control the direction his Mavic goes -- But still the compass declination is used along with GPS for the GO4 app for that areas visual telemetry information .
              The log analysis --- I believe your right along with DroneFlying who pointed this out.
But this doesn't explain the op's Mavic behavior ---- I am in Total agreement with you !!
               It was pointed out that right B4 taking off without noticing a compass error would cause this
Behavior!!! But I never tookoff with a compass error B4
               I have experienced Compass error B4 and never lost manual control of my Mavic !!!
The only way I can see a compass error causing this behavior is if the Mavic was in a autonomous mode
This is why this thread is so interesting to me!!!
               I have used the Litchi program it has what is called a Waypoint Mission Mode , it helped me to better understand the Mavic's autonomous system --- Litchi mission mode does let you put in longitudes
and latitudes for full autonomous flight -- the only thing you need to know that is very important is your
altitude setting , This is how many of the Litchi users Crash , but if your are familiar with the area your flying in you should be alright --- I would recommend this program to better understand the Mavic system and capabilities --- You do have to pay for it but to me it was worth it .     
FlySafe-FlyFree

BudWalker Posted at 2017-4-26 05:55
It may have seemed like the Mavic wouldn't respond to aileron input. I think it's probably more like you and the Flight Controller were both trying to control the Mavic. I tried to confirm by looking at the effect on the horizontal velocity. But couldn't becasue of the number of control inputs. In addition to full left aileron there was some negative elevator and throttle inputs. Add onto this what the Flight Controller was inputting. Time 26.147 is where you started applying left aileron.

I viewed these logs and compared many times from different PoW.
At 18.4sec I released elevator stick but the drone was still increasing speed, and applied it down several times before the crash with no affect. That's what makes me confused.

@BudWalker, as you are very strong in logs analysis, probably you can explain these errors in log messages (https://www.dropbox.com/s/5rt6b0 ... ogMessages.log?dl=0):

------------------------------------------------------------------------------------------------------------
      0 [L-GPS]<GPS ERROR>m8 uid not received!
------------------------------------------------------------------------------------------------------------
     41 [L-GPS]<GPS ERROR>[0][0] frequency error 38
------------------------------------------------------------------------------------------------------------
    300 [L-FMU]FATAL ERROR fail to get data 95 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 130 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 131 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 148 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 184 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 196 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 197 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 198 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 199 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 200 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 201 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 202 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 203 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 204 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 205 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 206 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 207 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 208 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 209 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 210 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 214 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 215 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 226 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 227 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 228 length for invalid src
    300 [L-FMU]FATAL ERROR fail to get data 229 length for invalid src
------------------------------------------------------------------------------------------------------------


And what these "0" and "300" mean in the start of each line in the log messages file?

BTW, at the midnight I sent email with all the descriptions, videos and logs to support.eu@dji.com, still no reply.

Oleh Artyukh Posted at 2017-4-26 06:32
I viewed these logs and compared many times from different PoW.
At 18.4sec I released elevator stick but the drone was still increasing speed, and applied it down several times before the crash with no affect. That's what makes me confused.

The point I tried to make is that it is both you AND the Flight Controller are trying to control the Mavic. It  only seems like you don't have any control. You have some control but, in this case, it's difficult to separate the two. You only showed elevator input. There is also aileron to consider. In addition, you need to consider direction of flight and horizontal speed.

Before u do compas calibration check first in sensor if no interference around you. When u do compas calibration move away form your remote and keys out your pocket. Always check first in the dji go app if no inferences.

Oleh Artyukh Posted at 2017-4-26 06:55
@BudWalker, as you are very strong in logs analysis, probably you can explain these errors in log messages (https://www.dropbox.com/s/5rt6b0gx0ryxwvc/FLY088.DAT.LogMessages.log?dl=0):

------------------------------------------------------------------------------------------------------------

I'm curious. How did you obtain this log. This is the same as the eventLog stream produced by DatCon. But, DatCon attaches the time at the beginning of each entry. The numbers at the beginning of each entry are sequencing numbers generated by the AC.

I don't know what those entries are about. Some of them are seen in most .DATs. It seems that the content changes with every firmware update.

BudWalker Posted at 2017-4-26 07:25
I'm curious. How did you obtain this log. This is the same as the eventLog stream produced by DatCon. But, DatCon attaches the time at the beginning of each entry. The numbers at the beginning of each entry are sequencing numbers generated by the AC.

I don't know what those entries are about. Some of them are seen in most .DATs. It seems that the content changes with every firmware update.

I've got them from converter here: http://www.djilogs.com/

SkunkWerxs Posted at 2017-4-26 06:15
rnrnrn ,
           I'm starting to love this op's thread it's making people Think of the operating circuitry of the Mavic that gives a better understanding of there system that they paid for loll, now for my opinions of your post
           "Declination "  Is used to determine the area your in, to get True North cowardanat along with GPS for the GO4 app Telemetry especially when you use any autonomous mode.

Hi again SkunkWerxs,

This thread is really cool There's a load of information to draw upon and some good analysis of the situation.

I mentioned that declination should be unimportant because the compass doesn't care where true north is. In any case when you want to get the true north you calculate course + declination + deviation. In autonomous mode you are oriented around a GPS position and in this case the position should be reflected by the GPS coordinates (which have a tendency to be not accurate enough especially the elevation) but I'm pretty sure that the course will be oriented around the compass. Let's say we take POI mode. The distance from the chosen point and the speed will be GPS oriented but the course doesn't have to be like it is on the chart. It can, sure, but doesn't have to be. You can just take the path of the device between the chosen POI and then extrapolate by the course it took to get to the edge where you want the craft to fly. I do believe that's how it is since the camera does not - from my experience - correct well enough to point properly and that would be the result of the compass instability.

Course lock - definitely compass related but with a touch of GPS to keep on a certain line in space as opposed to a floating course line due to wind drift. Home lock - definitely GPS and compass based - GPS data would give the location and the compass the orientation of the craft.

But... in all above cases - whether the true north can be ascertained makes no difference. It would only make a difference if you had to follow a route on the chart - this was not the case in this accident and will rarely happen. I'm just wondering how the calculation of the desired route takes place in the software. Anyhow it's only software - same as for the GPS position accuracy - it's calculated onboard based on the satellite ephemeris ;-)

Cheers!

BudWalker Posted at 2017-4-26 05:32
Not sure what you mean by axis lock. Maybe you're referring to ATTI mode where the AC isn't trying to navigate; it's just listening to the IMU and will drift with the wind. If so, you're right the Mavic didn't switch from GPS+ATTI to ATTI mode. This appears to be "normal" with the Mavic. The Phantom3 would have switched to ATTI. The stratgey with the Mavic appears to be to stay in GPS+ATTI mode and attempt to fix the problem. I've looked at several flights where this has occurred. It actually works sometimes. That's what happened in this flight.
The beige/tan background is the interval where there is a compass error. It starts at about 14 secs where magYaw and Yaw differ by about 150 degrees. Following this Yaw remains constant but magYaw is changing to become closer to Yaw. What's actually happening is that the Mavic is being rotated so that the compass (magnetometers) will become aligned with the Yaw value. I.e., rather than the Flight Controller changing it's Yaw value to agree with magYaw the Mavic is rotated until magYaw shows the same value as the Yaw value. This has been confirmed with other flights where a video is available. I haven't done that confirmation with this flight.

BudWalker,

Exactly, when I refer to axis lock I mean ATTI. And that's what bothers me about this behavior of the Mavic.

Thanks to your great analysis now I understand what the AC is trying to do. But I personally think is not the best way to avoid the issue, the flight computer taking over and trying to match actual yaw with  magnetic heading. The problem is that this could cause crashes as already happened, because most of people as I until now, will not expect this behavior and try to correct it.

Sure, now being aware of this I will just let the AC do its thing and even gain some altitude for it to be clear and have enough space, but as you can see this is quite counter-intuitive and anyone in such situation will try to fight it back and find a place to land ASAP, deeming normally in a crash.

As you mentioned the Phantom3 does this automatically, rolls back into ATTI when encounters this kind of discrepancy on Yaw and Magnetic Heading. This is a behavior that I would expect, since I would still have full control of the AC and be able to manage a safer landing.

Cheers!  

AACY Posted at 2017-4-26 17:59
BudWalker,

Exactly, when I refer to axis lock I mean ATTI. And that's what bothers me about this behavior of the Mavic.

I agree. The P3 behavior is more desirable. But, there are problems with the P3 behavior. The existence of a compass error has to be inferred; it can't be directly measured. This means there is the possibility of either a false positive or not detecting an actual compass error. It seemed that with each firmware update the P3 compass error detecting algorithm got more sensitive. This would then produce more false positives in order to detect more real compass errors. It got to the point that almost every flight would have a compass error detected when none existed. Maybe I'm exaggerating, but only to make a point. :-)

I'm guessing that the Mavic behavior is DJI's attempt to fix this. It actually works pretty well in some cases. I just analyzed an incident where it worked. Take a look here.
http://mavicpilots.com/threads/slow-yaw-at-start-of-flight-without-control-inputs.14824/#post-173571

It would seem that if the magYaw/Yaw difference is not too great then this approach works. Maybe a Mavic/P3 hybrid would be best.

BudWalker Posted at 2017-4-26 05:32
Not sure what you mean by axis lock. Maybe you're referring to ATTI mode where the AC isn't trying to navigate; it's just listening to the IMU and will drift with the wind. If so, you're right the Mavic didn't switch from GPS+ATTI to ATTI mode. This appears to be "normal" with the Mavic. The Phantom3 would have switched to ATTI. The stratgey with the Mavic appears to be to stay in GPS+ATTI mode and attempt to fix the problem. I've looked at several flights where this has occurred. It actually works sometimes. That's what happened in this flight.
The beige/tan background is the interval where there is a compass error. It starts at about 14 secs where magYaw and Yaw differ by about 150 degrees. Following this Yaw remains constant but magYaw is changing to become closer to Yaw. What's actually happening is that the Mavic is being rotated so that the compass (magnetometers) will become aligned with the Yaw value. I.e., rather than the Flight Controller changing it's Yaw value to agree with magYaw the Mavic is rotated until magYaw shows the same value as the Yaw value. This has been confirmed with other flights where a video is available. I haven't done that confirmation with this flight.

BudWalker,   Correct me if I'm wrong ,,,,, According to your annalist , your saying that the Manual control
                   of the Mavic is Dependent on the Compass within the Mavic for it's manual course that
                   that the Pilot is flying ????
          I do agree with you that the op was trying to compensate with control stick inputs --- no need to analyze Data for this , you can see it on the op's video - half way through the tracking video you can see when the Mavic reached the top of the arch and when he explained , the Mavic was trying to change
course, note the erratic tracking Path line.
                    I would agree with your annalist of the op's flight data if the op was flying in a
Autonomous Mode but he wasn't , he did state he did panic ( We all would have ) explaining all of his control inputs. The Data also shows he was trying to control the Mavic while at the same time something else was over riding him and was not recorded.        

rnrnrn Posted at 2017-4-26 11:38
Hi again SkunkWerxs,

This thread is really cool  There's a load of information to draw upon and some good analysis of the situation.

rnrnrn ,
          GPS coordinates are much more accurate then a compass this is why they are used in Land surveying and guidance systems these days they are also used for topographic mapping instead of using a inclination meter from the ground. I only know of three types of GPS one is for General Public use accurate
within 2 to 3 feet /// The other GPS (I don't know the name of it offhand ) is the type that is paid for by businesses like construction and surveying companies the accuracy is within a couple of centimeters
And a third  that would be for Military use, accuracy is probably classified , all have different names.      
           This all just brings us back to where we started --- one thing I believe for sure is that the compass is always in use for the GO4 app Radar heading direction ( the little red arrow ) GPS is used for tracking the Mavic in the GO4 app using the map or satellite screen view and both among other components.
Oh , the Mavic's altitude is calculated by using a onboard barometer , this is Fact it is in the users manual.      

SkunkWerxs Posted at 2017-4-27 07:49
BudWalker,   Correct me if I'm wrong ,,,,, According to your annalist , your saying that the Manual control
                   of the Mavic is Dependent on the Compass within the Mavic for it's manual course that
                   that the Pilot is flying ????

"According to your annalist , your saying that the Manual control
                   of the Mavic is Dependent on the Compass within the Mavic for it's manual course that
                   that the Pilot is flying ????"


I don't believe I said this. At least I didn't intend to.

BudWalker Posted at 2017-4-27 09:05
"According to your annalist , your saying that the Manual control
                   of the Mavic is Dependent on the Compass within the Mavic for it's manual course that
                   that the Pilot is flying ????"

BudWalker ,
                 If you read what you posted you will see that what was said would be true if the op was in a autonomous mode ----  it's ok there's a lot to this system to fully grasp
FlySafe-FlyFree

BudWalker Posted at 2017-4-27 04:54
I agree. The P3 behavior is more desirable. But, there are problems with the P3 behavior. The existence of a compass error has to be inferred; it can't be directly measured. This means there is the possibility of either a false positive or not detecting an actual compass error. It seemed that with each firmware update the P3 compass error detecting algorithm got more sensitive. This would then produce more false positives in order to detect more real compass errors. It got to the point that almost every flight would have a compass error detected when none existed. Maybe I'm exaggerating, but only to make a point. :-)

I'm guessing that the Mavic behavior is DJI's attempt to fix this. It actually works pretty well in some cases. I just analyzed an incident where it worked. Take a look here.

Sadly I do not own nor flew a P3, but thanks for your input on that. Regarding the slow yaw that's great as you said, only when the difference is not so great.

But it would be great to have a middle ground where, when the difference is so great the MAVIC uses the protocol used on the P3 and enter in full Axis Lock (ATTI). And when the difference is not so great to have it slowly yaw and compensate and have a pop up in the screen so the user is aware of what the AC is doing.

Cheers!

BudWalker Posted at 2017-4-27 04:54
I agree. The P3 behavior is more desirable. But, there are problems with the P3 behavior. The existence of a compass error has to be inferred; it can't be directly measured. This means there is the possibility of either a false positive or not detecting an actual compass error. It seemed that with each firmware update the P3 compass error detecting algorithm got more sensitive. This would then produce more false positives in order to detect more real compass errors. It got to the point that almost every flight would have a compass error detected when none existed. Maybe I'm exaggerating, but only to make a point. :-)

I'm guessing that the Mavic behavior is DJI's attempt to fix this. It actually works pretty well in some cases. I just analyzed an incident where it worked. Take a look here.

Hey BudWalker,

Here is a post regarding this, have a look at the last messages!

http://forum.dji.com/thread-94313-1-1.html

Maye we could get more understanding out of it, and use your expertise to compare the .DAT data

Cheers!

Ok, finally I received official reply from DJI

"It appears that there was a defect found. It will be warranty if you send it in. "

But I won't send it from Ukraine to USA - it's cheaper and quicker to repair it here for my own cost

Oleh Artyukh Posted at 2017-5-10 03:43
Ok, finally I received official reply from DJI

"It appears that there was a defect found. It will be warranty if you send it in. "

Did they tell what the defect was, good news you got sorted hope your up and running soon.

hallmark007 Posted at 2017-5-10 04:02
Did they tell what the defect was, good news you got sorted hope your up and running soon.

No, they didn't

Oleh Artyukh Posted at 2017-5-10 03:43
Ok, finally I received official reply from DJI

"It appears that there was a defect found. It will be warranty if you send it in. "

Oleh Artyukh , Great News -- Good for You !!!!!

                 Dear DJI can you explain what  Oleh Artyukh defect was ????
Was it Software?? was it Hardware?? the way I see it , if there's a defect
there will be more to follow, just look at the Mavic's past history or do you
DJI want to remain silent on this so called defect that was found , which in
my opinion is totally unfair to your supporting customers  

SkunkWerxs Posted at 2017-5-10 04:37
Oleh Artyukh , Great News -- Good for You !!!!!

                 Dear DJI can you explain what  Oleh Artyukh defect was ????

The only text I received at this moment is:

"HI Oleh,

It appears that there was a defect found. It will be warranty if you send it in.

Thanks for your patience. "

Waiting for their reply in email or here from DJI representatives.

Oleh Artyukh Posted at 2017-5-10 04:52
The only text I received at this moment is:

"HI Oleh,

Thanks Oleh please let us know if you find out more --- Your case was one of the most interesting
                   cases I've seen or heard of B4
Thanks again, hope you get your bird back real soon !!

SkunkWerxs Posted at 2017-5-10 04:58
Thanks Oleh please let us know if you find out more --- Your case was one of the most interesting
                   cases I've seen or heard of B4
Thanks again, hope you get your bird back real soon !!

Sure, I'm trying to get this info from DJI.

And I'm already flying, I had to pay for local repair service ((

Just throwing in my 2 cents here. Have you heard of kP index? Maybe check how high that was for your location the day you crashed and the day before.
It may have charged the metal poles in the bench, causing the compass to fail...

K4Unl Posted at 2017-5-12 03:39
Just throwing in my 2 cents here. Have you heard of kP index? Maybe check how high that was for your location the day you crashed and the day before.
It may have charged the metal poles in the bench, causing the compass to fail...

K4Unl , Yes I heard of the Kp index, but have never heard of a Mavic Crash because of it
            This was not the case here

The answer to this problem is in the first 4 seconds of the video he posted..............

DroneFlying Posted at 2017-4-24 02:10
Now I see I had to go up as fast as possible, probably I could avoid hit into the tree, but I was in panic  and tried to make Mavic go back, but it continued flying like a crazy and like with course lock (flight direction was the same while head as turning around).

Yes, it probably would have recovered eventually, but I understand what you mean: it's a little scary when your expensive drone suddenly stops responding to the controls.

The same thing happened to me. Was up and doing a stationary circle and all was good. Then flew it out and did a a 360 and near the end it stop responding with the right stick. So I tried to land it and it wasn't come down quick enough. So I shot it up in the air. In the video it lost compass and at the end all just became good. If someone tells me how to get all the info out th really for everyone to view. I had no poles and launched it from my driveway as I always do.