Fixing DJI's Compass Problem

[ElGuano]

Not to be a wet blanket but...

When the poster above says it clears up in 2nd or 3rd pass, isn't that still undesirable?

Shouldn't the declination be looked up in a table (and kept fresh each release) based on GPS coordinates and applied immediately?

A couple of thoughts:

1. I don't think we know yet if it clears up after a couple of passes per boot, or whether it also saves learned declination between battery changes. If the latter, it may be a non-issue until you move regions. And in either case, it seems way better, faster.

2. The lookup table is an easy to understand example many have latched onto, but how would it be built? What resolution would you use? The big world map has only three easy-to-digest bands, but in reality it's a more or less smooth gradient with hundreds of different declination figures between points. Would an area set as -8 be enough if an area is actually -10.2, or -14.9?

Keeping a dynamic system that learns and corrects over time is beneficial in a number of cases, including "between" zones on a low resolution table, or if there are areas with local anomalies not generally captured in a global map.

Just brainstorming...

 

[wkf94025]

i particularly agree with el guano's last paragraph. it will be what it will be, but if anyone is asking this consumer, i would trade a minute of warmup laps each battery for better precision over the remaining flight.

kelly

 

[sar104]

Not to be a wet blanket but...

When the poster above says it clears up in 2nd or 3rd pass, isn't that still undesirable?

Shouldn't the declination be looked up in a table (and kept fresh each release) based on GPS coordinates and applied immediately?

A couple of thoughts:

1. I don't think we know yet if it clears up after a couple of passes per boot, or whether it also saves learned declination between battery changes. If the latter, it may be a non-issue until you move regions. And in either case, it seems way better, faster.

2. The lookup table is an easy to understand example many have latched onto, but how would it be built? What resolution would you use? The big world map has only three easy-to-digest bands, but in reality it's a more or less smooth gradient with hundreds of different declination figures between points. Would an area set as -8 be enough if an area is actually -10.2, or -14.9?

Keeping a dynamic system that learns and corrects over time is beneficial in a number of cases, including "between" zones on a low resolution table, or if there are areas with local anomalies not generally captured in a global map.

Just brainstorming...

The tables exist, as do full-earth models that estimate declination from lat/long, including time-dependent drift. Either of those is simple. I cannot believe that the problem was ever with knowing the declination, but rather using it properly.

 

[SilentAV8R]

The tables exist, as do full-earth models that estimate declination from lat/long, including time-dependent drift. Either of those is simple. I cannot believe that the problem was ever with knowing the declination, but rather using it properly.

As was noted above, it all depends on the resolution you want to use. That "lookup table" will get frightfully large in no time. Talk about a memory hog!! So at some point a compromise needs to be struck because I doubt there is enough memory space to allow a dataset that covers every single lat/long down to 30-second resolution. I'm not even sure ultimately what resolution would be need to achieve an adequate result.

In the end it may be moot, at least here in the US, since BLOS will be a thing of the past once the FAA gets around to publishing the sUAS Rule.

 

[AnselA]

Declination for a flying site can be found on http://magnetic-declination.com/ . An option to manually enter it in app settings would be OK for me.

 

[ianwood]

You can put a full world model with better than one degree accuracy in less than 3Kb. Not big at all.

 

[SilentAV8R]

You can put a full world model with better than one degree accuracy in less than 3Kb. Not big at all.

Ok, I did not realize that. That's a lot of data in a vey small chunk!!!

 

[sar104]

The tables exist, as do full-earth models that estimate declination from lat/long, including time-dependent drift. Either of those is simple. I cannot believe that the problem was ever with knowing the declination, but rather using it properly.

As was noted above, it all depends on the resolution you want to use. That "lookup table" will get frightfully large in no time. Talk about a memory hog!! So at some point a compromise needs to be struck because I doubt there is enough memory space to allow a dataset that covers every single lat/long down to 30-second resolution. I'm not even sure ultimately what resolution would be need to achieve an adequate result.

In the end it may be moot, at least here in the US, since BLOS will be a thing of the past once the FAA gets around to publishing the sUAS Rule.

As Ianwood noted, such a table does not need to be large, although I make it a bit bigger than 3kB. Certainly a 1 degree resolution (lat & long) is ample with interpolation, and that is going to be no more than 64 kB.

EDIT: I realized that the 3 kB referred to is a model description such as the IGRF, so that would be a reasonable size.


And I'm not sure why you think the problem only applies when the aircraft is out of visual range.

 

[AnselA]

You can put a full world model with better than one degree accuracy in less than 3Kb. Not big at all.

That's pretty compact!

 

[dragonash]

add me to the list

FC40
Brooklyn,NY

If I go forward, it curves to the right
If I hover, it does a clockwise TBE

Calibrating the compass onsite make it a little better.
This just started too... like today. After I put on my 9 inch props from dji. Weird

 

[SilentAV8R]

And I'm not sure why you think the problem only applies when the aircraft is out of visual range.

My reference to that was that when within visual range it is much easier to make stick corrections to stay on track. BLOS it starts to get more difficult and more important to have the flight control system do more of the work of maintaining track.

 

[Andy T]

add me to the list

FC40
Brooklyn,NY

If I go forward, it curves to the right
If I hover, it does a clockwise TBE

Calibrating the compass onsite make it a little better.
This just started too... like today. After I put on my 9 inch props from dji. Weird

Wow, that's actually really interesting. It flew perfectly before?

 

[landonkk]

add me to the list

FC40
Brooklyn,NY

If I go forward, it curves to the right
If I hover, it does a clockwise TBE

Calibrating the compass onsite make it a little better.
This just started too... like today. After I put on my 9 inch props from dji. Weird

Considering that it happened without a firmware upgrade or a change of location I would guess that it is a different issue. You might want to initiate an advanced calibration in the Assistant.

 

[skyjoel]

I am in the Sacramento, Ca area directly in the middle of one of the affected zones and have dealt with this issue since day one of flying 60 days ago...I have just corrected manually while in GPS mode by pushing forward and right on my sticks while rotating...it is a hassle but a workaround. i hate when i release my sticks and expect a hover but instead get a rotation of about six foot diameter circle! If I go to ATTI mode it gets worse!
My main use of the P@ with h33d gimbal and gopro is to provide video inspections of roofs and properties for absentee landlords so my range is only 100-200 feet from the tx. I would love to have more reliability and stability in my flights that is what i paid for?

 

[verisimilitude]

i hate when i release my sticks and expect a hover but instead get a rotation of about six foot diameter circle! If I go to ATTI mode it gets worse!

Why are you surprised that your Phantom doesn't maintain position when hovering in ATTI mode?

 

[yachtsman]

So far, I have 21 people confirming this issue with the P2/P2V. I want 50 or more. This is a bug in the Naza and it's time to get DJI to fix it (or at least tell us how to).

If you are in western or northeastern US and Canada, you need to test for this and chime in here! Same for NZ, ZA, eastern OZ. Also, northeast of South America. All you Europeans and American mid-westerners consider yourself lucky.

Hi Ian Wood, I've read your document "DJI P2 and magnetic declination" and I'm pretty sure you have identified the problem correctly. Working at sea I've done my compass corrections.
I only had time to do a short test flight of my P2 in Scandinavia, magnetic variation (as I call it being a ships officer) there is about 5 degrees East (positive). Over the last couple of weeks I've been flying in southern Brazil, where magnetic variation is 19 degrees west (negative) and I see very different behavior in flight pattern/hovering.
I've been calibrating the compass on each flight, but I see exactly what you have been describing, right hand hooking and TBE. None of this was apparent during my test flights in Scandinavia, there the P2 flew straight as an arrow.

You can add me to your list of owners experiencing this problem. I will carry out some more flights in Scandinavia when returning in two weeks and it should be interesting to see if the problems disappear.

Brgds,

yachtsman

 

[SilentAV8R]

Any more feedback from the beta testers??

 

[onfourblades]

Any more feedback from the beta testers??

follow progress here viewtopic.php?f=19&t=11555

 

[obiwan_pierogi]

I wish they had a fix for me.. Just got in touch with the dealer, after purchasing a pv2+.. Why do they always harp about compass calibration? Don't they know that's the first thing you do? I'm waiting to hear back from him again, as I'm concerned if I rotate the compass the 30 degrees, as on my vision 2, it'll void the warranty.. As it is right now, it's almost un-flyable.. 30' left swing, after a straight 40 foot run.. Then it instantly flops into a counter-clockwise TBE ... My declination is 17.2+ .. I tell you, if everyone experienced the left hook, and toilet bowl effect as I do, They'd either crash, or ship their copter back..

 

[Andy T]

I wish they had a fix for me.. Just got in touch with the dealer, after purchasing a pv2+.. Why do they always harp about compass calibration? Don't they know that's the first thing you do? I'm waiting to hear back from him again, as I'm concerned if I rotate the compass the 30 degrees, as on my vision 2, it'll void the warranty.. As it is right now, it's almost un-flyable.. 30' left swing, after a straight 40 foot run.. Then it instantly flops into a counter-clockwise TBE ... My declination is 17.2+ .. I tell you, if everyone experienced the left hook, and toilet bowl effect as I do, They'd either crash, or ship their copter back..

I think you live near me.. it's pretty insane how those things fly here. hopefully the firmware update wont take too long to be released... it's driving me nuts