Please select Into the mobile phone version | Continue to access the computer ver.
Spark Compass Error Firmware Fix?
2124 19 2017-10-18
Uploading and Loding Picture ...(0/1)
o(^-^)o
InTheReeds
Second Officer
Flight distance : 347103 ft
United States
Offline

Does anyone know or suspect that the Compass Error a lot of us see will be fixed in v01.00.0700?

I'm asking specifically about the issue found in this thread where Spark is seeing a massive angular error of +/-350º when in reality it's closer to +/-15º.
2017-10-18
Use props
djiuser_Kf4iPA3
First Officer
Flight distance : 441001 ft
Israel
Offline

I don't know if there's an upcoming fix, I don't think anyone does.
But this error clearly shows that something is rotten in the kingdom of the Spark compass... or at least in its data processing in the firmware.
2017-10-18
Use props
Gunship9
Captain
United States
Offline

I know the posters in the thread figured the spark was seeing 350 degrees of yaw to get the compass error, but I doubt the spark uses a graphical chart to figure out compass error.  It also wouldn't use lines to connect data points it has sampled from the magnetic compass.  

I bet the engineers know that -179 and +179 are close to each other and used a software algorithm to account for those data points.  They be smart about navigation programming or good at reverse engineering long established solutions for calculating compass error.

The chart is just a way for people to easily see what a list of data points mean.  It isn't perfect.  I think what the chart shows is a compass starting to swing more just before the error.  You see it (data sample points) getting farther from yaw data then settle down again.  What have the engineers stated is the thresh hold for difference before it throws an error?  10 degrees for five data samples or 5 degrees for 20 data samples?
2017-10-18
Use props
djiuser_Kf4iPA3
First Officer
Flight distance : 441001 ft
Israel
Offline

I don't think this is relevant or points to the wisdom of the engineers. Having been in the hi-tech business for over 25 years, I've seen some pretty stupid bugs created by very wise engineers.
They just forgot a simple sanity check in the firmware. An error of 350 should have been an exception for the programmers.
An error of 10 degrees - I doubt whether it should alert an error at all, especially given that with Spark's current implementation the consequence is rather heavy-weighted - because a compass error leads, unfortunaltely, to the loss of data from other navigation systems as well.
2017-10-18
Use props
Gunship9
Captain
United States
Offline

djiuser_Kf4iPA3 Posted at 2017-10-18 07:10
I don't think this is relevant or points to the wisdom of the engineers. Having been in the hi-tech business for over 25 years, I've seen some pretty stupid bugs created by very wise engineers.
They just forgot a simple sanity check in the firmware. An error of 350 should have been an exception for the programmers.
An error of 10 degrees - I doubt whether it should alert an error at all, especially given that with Spark's current implementation the consequence is rather heavy-weighted - because a compass error leads, unfortunaltely, to the loss of data from other navigation systems as well.

If it was a bug, wouldn't we and the beta testers, see errors every time any Spark did a 360 rotation?  And lots of errors if flying in that direction?  Flying near -+179?

I don't think the chart means anything and is deceptive graphically.
2017-10-18
Use props
Kyokushin
Captain
Flight distance : 296378 ft
Poland
Offline

And workaround would be simple - just rotate a bit - 45-95 dg (it should just bring gps back) when you feel a copter is flyingaway or got compass error. It also should be easy to replicate then.
2017-10-18
Use props
djiuser_Kf4iPA3
First Officer
Flight distance : 441001 ft
Israel
Offline

Gunship9 Posted at 2017-10-18 08:03
If it was a bug, wouldn't we and the beta testers, see errors every time any Spark did a 360 rotation?  And lots of errors if flying in that direction?  Flying near -+179?

I don't think the chart means anything and is deceptive graphically.

You don;t trust the chart? Fine, forget the chart. Look at the event at the left and just consider the numbers. A difference, 10 to 20 degrees, and it lasts at least as long - and that does not count as an error.
So it's clear that incident shouldn't have been an error.

And it doesn't mean it has to happen every time the Spark turns south. Could be a combination of conditions that leads to some dark corner in the firmware, anyone who ever wrote code knows about events like this. And a glitch it is, that shouldn't have been en error. I hope they can reproduce it and fix it.
2017-10-18
Use props
hallmark007
Captain
Flight distance : 7732080 ft
  • >>>
Ireland
Offline

Gunship9 Posted at 2017-10-18 08:03
If it was a bug, wouldn't we and the beta testers, see errors every time any Spark did a 360 rotation?  And lots of errors if flying in that direction?  Flying near -+179?

I don't think the chart means anything and is deceptive graphically.

The reality is compass problems are by and large users fault, I don’t see to many compass problems around here lately, and there is not a problem that needs to be fixed through firmware or hardware.
2017-10-18
Use props
djiuser_Kf4iPA3
First Officer
Flight distance : 441001 ft
Israel
Offline

Seems I have angered the gods once again...
What's presented up here is a miscalculation. A smaller difference (in reality) is presented as an error; A larger difference is not. How can that be right?
It doesn't have anything to do with conditions, because the numbers were collected and the real difference is small. It's not a real error; It's a false alarm, a misinterpretation of the numbers.
2017-10-18
Use props
Anuvis
First Officer
Flight distance : 1976736 ft
Greece
Offline

I am getting lately a lot false atti mode warnings.....It just says for one sec atti mode and there is no such warning in the flight record....So i guess it is false alarm....I think that is problem of dji go app not firmware
2017-10-18
Use props
djiuser_Kf4iPA3
First Officer
Flight distance : 441001 ft
Israel
Offline

From your description, it seems to me those are two separate issues. You say your error indication doesn't appear in the flight data; The error (false error, actually) that started this thread appeared in the flight data.
2017-10-18
Use props
hallmark007
Captain
Flight distance : 7732080 ft
  • >>>
Ireland
Offline

Anuvis Posted at 2017-10-18 12:28
I am getting lately a lot false atti mode warnings.....It just says for one sec atti mode and there is no such warning in the flight record....So i guess it is false alarm....I think that is problem of dji go app not firmware

What flight record are you looking at?

You can upload your log to link below see if it shows up there.

http://www.phantomhelp.com/LogViewer/Upload/

2017-10-18
Use props
Gunship9
Captain
United States
Offline

djiuser_Kf4iPA3 Posted at 2017-10-18 11:15
Seems I have angered the gods once again...
What's presented up here is a miscalculation. A smaller difference (in reality) is presented as an error; A larger difference is not. How can that be right?
It doesn't have anything to do with conditions, because the numbers were collected and the real difference is small. It's not a real error; It's a false alarm, a misinterpretation of the numbers.

Not sure about the gods.  

Your data chart shows the compass tracking the yaw data until midway through the chart.  At that point it starts swinging one way then the other.  I could imagine an error pop up if that is the software threshold.  Later the compass starts tracking along with the yaw data,  I could see the compass error warning turning off at that point.  Did the Spark fly past a metal light pole?

I don't think the chart supports your hypothesis that the Spark is confused by compass movement across the -+179 heading.  The GPS health went to zero (low) when the magnetic yaw first got far enough off of the yaw readings.  That was before when it swung back across the -+179 heading.  Then when the compass stabilized, the health data bit came back up.
2017-10-18
Use props
djiuser_Kf4iPA3
First Officer
Flight distance : 441001 ft
Israel
Offline

Gunship9 Posted at 2017-10-18 16:07
Not sure about the gods.  

Your data chart shows the compass tracking the yaw data until midway through the chart.  At that point it starts swinging one way then the other.  I could imagine an error pop up if that is the software threshold.  Later the compass starts tracking along with the yaw data,  I could see the compass error warning turning off at that point.  Did the Spark fly past a metal light pole?

It's not my chart, i didn't produce it. I just read the numbers.
But you're making the same mistake that apparently the firmware does... if you say the compass reading "swings one way and then the other", you're misinterpreting the numbers.
The yaw scale on the chart is between -180 and + 180, but in fact on a circle these two points are one and the same. What you see as a "swing one way or the other" is in fact a transition across 180. it's not a real "swing", and if you look at the original thread with the chart and scroll a message down - you will see that when the scale is corrected to reflect the fact that +180 and -180 are the same point, the real difference between yaw and magyaw is actually small. It's not a real error.

The GPS health drops to zero, because Spark's algorithm throws away valid GPS data when it has a compass error. That's another issue, doesn't make any sense to me, but that's the way it is.

So in fact what we have here is a false alarm that causes not only a false indication of a compass error - but in fact, the way Spark's algo works, it causes throwing away GPS data as well...
2017-10-18
Use props
djiuser_xLE6O03
Second Officer
Flight distance : 1198346 ft
Romania
Offline

Gunship9 Posted at 2017-10-18 16:07
Not sure about the gods.  

Your data chart shows the compass tracking the yaw data until midway through the chart.  At that point it starts swinging one way then the other.  I could imagine an error pop up if that is the software threshold.  Later the compass starts tracking along with the yaw data,  I could see the compass error warning turning off at that point.  Did the Spark fly past a metal light pole?

Here is original thread with more data . https://forum.dji.com/forum.php? ... mp;page=1#pid979616
2017-10-18
Use props
Anuvis
First Officer
Flight distance : 1976736 ft
Greece
Offline

hallmark007 Posted at 2017-10-18 13:01
What flight record are you looking at?

You can upload your log to link below see if it shows up there.

yes there is nothing about the alarm in flight records on phantom help or in flight log in my phone after i replay the flight....So that's i guess is a false alarm....
2017-10-18
Use props
Gunship9
Captain
United States
Offline

djiuser_Kf4iPA3 Posted at 2017-10-18 21:17
It's not my chart, i didn't produce it. I just read the numbers.
But you're making the same mistake that apparently the firmware does... if you say the compass reading "swings one way and then the other", you're misinterpreting the numbers.
The yaw scale on the chart is between -180 and + 180, but in fact on a circle these two points are one and the same. What you see as a "swing one way or the other" is in fact a transition across 180. it's not a real "swing", and if you look at the original thread with the chart and scroll a message down - you will see that when the scale is corrected to reflect the fact that +180 and -180 are the same point, the real difference between yaw and magyaw is actually small. It's not a real error.

Difference is yaw and magyaw is actually small?  Small is 50 degrees, 15 degrees, 5 degrees, or 0.5 degrees?  Looks like firmware thinks 3 degrees is small (jitter) and 15 degrees is large.  The forum thinks 20 degrees is small (not navigators returning to home).

The compass didn't stay jittering on the course the yaw data had.  Instead it swung fifteen degrees to the left of the yaw data's recorded course and then fifteen degrees to the right of yaw data's recorded course.  It did this rapidly.  

GPS faulted as soon as compass went 15 degrees off (maybe 19 degrees).  Fault went away as soon as the compass settled down on the same course as yaw data.  Compass shouldn't swing if gyros and accelerometers didn't feel the aircraft turn.  fault.

I know it didn't travel 170 degrees off off of the yaw data's course (that is just chart graphics that look like that).

No biggie.  But, I think  It is reading the data correctly and that is why you won't see a firmware fix.
2017-10-19
Use props
djiuser_Kf4iPA3
First Officer
Flight distance : 441001 ft
Israel
Offline

I think we're looking at two different charts here, hence the misunderstanding.
I was looking at the chart presented in the thread pointed to by the first message here: https://forum.dji.com/thread-114937-1-1.html

"How small is small" is a good question. I would answer "smaller than a bigger difference that didn't produce an error"... that appears to be the case in teh chart I was looking at. That can't be right.

And the algo that throws away GPS data when there is a compass error is pretty much known. GPS data remains intact, but is being ignored, and the craft exits P-GPS mode. That's a whole different issue, a behavior that doesn't make sense to me, but it's unrelated to the question of why there's a compass error indication to begin with.
2017-10-19
Use props
Gunship9
Captain
United States
Offline

Gunship9 Posted at 2017-10-19 00:36
Difference is yaw and magyaw is actually small?  Small is 50 degrees, 15 degrees, 5 degrees, or 0.5 degrees?  Looks like firmware thinks 3 degrees is small (jitter) and 15 degrees is large.  The forum thinks 20 degrees is small (not navigators returning to home).

The compass didn't stay jittering on the course the yaw data had.  Instead it swung fifteen degrees to the left of the yaw data's recorded course and then fifteen degrees to the right of yaw data's recorded course.  It did this rapidly.  

The guy with the chart just picks a set of data points where the compass is swinging back across the yaw data course.  I think the fault (indicated by GPS health bit) happens earlier because of the compass swinging towards 145 degrees while yaw is at 169 degrees.

There has to be more than we know for the spark to turn off the compass fault (second dip to 149ish degrees).  I didn't think that was any better but the compass seems to return to steady course after that.
2017-10-19
Use props
hallmark007
Captain
Flight distance : 7732080 ft
  • >>>
Ireland
Offline

I think it’s fair to say there are very little compass error problems around here lately, so maybe compass problems where more down to unnecessary calibration bad environments and general misuse by users, maybe correct preparation and knowing the the pitfalls and generally users becoming more familiar with their AC is helping reduce the number of reported incidents.
So I would think this is not a software/Firmware/Hardware issue but more likely a learning curve. I’m sure we haven’t seen the last of them but if it was SW/FW/HW problem then we should be seeing more and more of them coming up here and yet we are seeing much less,  go figure.
2017-10-21
Use props
Advanced
You need to log in before you can reply Login | Register now

Credit Rules