Maximum GPS Error = FLYAWAY?

[CapnBob]

I have been doing some experiments logging GPS signals and accuracies.
It seems there really is a monster lurking in the shadows..

GPS accuracy numbers are always stated with the caveat of "95% of the time"

It made me wonder: what about the other 5% ?

I fired up a BU-353 GPS puck on a laptop and monitored it with "VisualGPS" software for a period of 8 hours.
The results are in the image below. The main cluster of positions is slightly Southwest of where the unit fired up.
The damning data is in the "tentacles" that shoot out from the main cluster.
The distance to the Northwest extreme and the Southeast extreme is ~1,600' each, from the main cluster. (See the number in the lower left of the image)
This is an envelope of 3200', or over a half mile.
I have seen far larger deviations, bet never took a screenshot.

I am going to to a longer term study and will post the results.



I started searching for Horror stories, and there they were. Random HUGE errors...
Here are a few snippets:

"The maximum error I ever recorded in a track, was about 600 kilometers. This
was obtained two years ago with an Etrex Vista. It was working well for some
time, then the position jumped 600 km to the east, where it wandered around
for 8 minutes, then it jumped back."

"Last week I had a similar track recorded with the GPSmap60CS. After a while
with good precision, it jumped 62 km to the north, recorded 3 track points
near to each other, then it jumped 84 km to the south-west where it wandered
around for one minute, then it jumped back to the correct position."

"While I haven't had quite such large errors, I HAVE had the GPS suddenly
"think" my course had reversed itself! Also I've seen quite large
groundspeeds while sitting at a stop light, speeds on the order of 600 MPH!"

"That doesn't account for the observations, like Heinrich's, of errors of
hundreds of kilometers. If the distribution were Gaussian I'd never
expect to see errors of 1000 times sigma since the probability would be
miniscule. But I've actually seen such errors a number of times. The
largest error I've seen so far was a track that continued for several
minutes on my eMap showing my position as over the north Pacific
approaching Seattle at over 700 mph while I was actually about 1500
miles away heading south on a bike path near San Francisco at 15 mph."

"In either of these cases, the position errors can be very large and are
not at all Gaussian in character (e.g. an error of 1000 miles is
probably just about as likely as an error of only 100 miles).
Fortunately such errors are both rare and generally easy to spot"

Well, "rare" and "easy to spot" certainly apply here.
No many how many flights you have, all it takes is once, and it's "Bye Bye Birdie".

Unless dji has an algorithm to discard ridiculously large position changes over extended lengths of time, it seems to me that we have the smoking gun.

 

[bbfpv]

Wow. That's very interesting. You would assume these would be accounted for in the algo because DJI would have noticed similar anomalies during testing... but since we're clearly the beta testers, maybe not?

 

[robinb]

I did have a gps tracker on my cat until I found he could run a mile in a couple of seconds....

 

[rbhamilton]

Wow. Well that explains a lot. (The GPS data - not the cat thing).

 

[mad in nc]

Good data CaPnBob, Don't know if this this is even related as your data shows fixed location with drift I believe but I just found this site last week and I haven't had time to dive into it, correlate the predictions etc. The forward predictions of the number of sats overhead (based on your location that you enter i.e. address, city, state, zip, country etc.) that the program shows to me means "it" would guide me not to fly now at time of posting or in certain other time windows if the Phantom baseline is 6 minimal for GPS lock.

I should go out and fire up the Phantom as the chart shows below 6 but its just to **** cold for us here in the South after the ice storm we had last night....

site link is http://satpredictor.navcomtech.com

 

[flyNfrank]

The flytrex use to display that same rats nest looking mess on all their devices at one point. It use to never be seen on anything and just appeared after an update of their own. I complained to them about it because it also meant you had to do an extra process in dashware to clean it up from being seen in with the dashware gauges. Anyway it still does it everytime while the quad is on and not running. You can see it on all of the current flights in my flytrex profile that no flying took place. I can't remember ever hearing an explanation from them on what was taking place. I do remember the words "GPS Drift" being used though.

So from what I learned from researching something similar to this in the past is, there are a number of clocks on the satellite. If these clocks are not dead-set with one another, this is when you will have "satellite drift". To best explain it would be, say each clock settings are off from each other by 750 milliseconds. Then suggest that 3 processes are to start at 10:00am. and when that time comes each process did take place as it was programmed to. But because the clocks were off in time from each other, it caused the satellite to do some unexpected stuff which creates a drift. When the satellite you are using to obtain stats with moves, it causes the GPS to take the hit as a result.

I can't say that this event would be what causes some of the known fly-aways or not. But I will say I do not think it is possible unless the drift was a large one. Keep in mind, it's not like no data is being received from the satellite while this happens.

 

[flyNfrank]

I want to add that the highest percentage of fly-aways on the vPlus take place from those that do not take the time to read the manual. The next percentage come from mechanical failures of some kind.

 

[RichWest]

Tag to watch...because I didn't understand a word he said. CapnBob, when you do derive an answer could you convert it to English for us TSTL people... (Technical Speak as a Third Language).

I'd add, could random/quick dis-connections (poorly connected GPS plug) contribute to the data inaccuracies? I understand there is a battery to minimize some of those type of effects, loss of sat references. I'd assume most, if not all, of the GPS processing is done on the puck and then the GPS location is fed down to the Naza-Lite controller.

Feel free to ignore my post if you feel it is not on topic to your conversation.

 

[robinb]

Remember to switch to ATTI mode if you start to get a problem.

 

[Steeleagle]

I would think the errors are mostly device related. Too many things rely on accurate GPS data now - things like instrument approaches for aircraft - which must work with a high degree of accuracy and consistency. GPS signals do occasionally lose accuracy and that can actually be predicted in advance but it's pretty rare.

 

[flyNfrank]

Tag to watch...because I didn't understand a word he said. CapnBob, when you do derive an answer could you convert it to English for us TSTL people... (Technical Speak as a Third Language)

Don't you mean....convert it to English for us "Too Stupid To Learn" arrogant people?

Real pathetic that you make fun of someone like this publicly. Whats wrong with keeping your mouth shut after reading it and just going on with your day?

 

[bbfpv]

I didn't think he was making fun of anyone but himself for not being technical, and simply asking Capn for an explanation in layman's terms once the research was complete. I could be wrong...

 

[MapMaker53]

I thought the same. He certainly wasn't making fun of anyone -- and I, too, would appreciate a dumbed-down explanation of CapnBob's findings when he finishes his research.

 

[CapnBob]

I want to add that the highest percentage of fly-aways on the vPlus take place from those that do not take the time to read the manual. The next percentage come from mechanical failures of some kind.

I agree. However, there are flyaways experienced by veteran flyers that have been unexplained.
I believe this could be a possible cause

 

[flyNfrank]

I thought the same. He certainly wasn't making fun of anyone -- and I, too, would appreciate a dumbed-down explanation of CapnBob's findings when he finishes his research.

I thought he was talking about my post because I didn't see anything out of whack with CapnBob's OP. If he was not speaking of my post, then I apologize for the mistake.

 

[CapnBob]

Well, if you consider that the speed of light is 186,000 miles per second, A 1 millisecond error would introduce
an error of 186 miles. The precision required for an accurate GPS fix is orders of magnitude more precise.

The flytrex use to display that same rats nest looking mess on all their devices at one point. It use to never be seen on anything and just appeared after an update of their own. I complained to them about it because it also meant you had to do an extra process in dashware to clean it up from being seen in with the dashware gauges. Anyway it still does it everytime while the quad is on and not running. You can see it on all of the current flights in my flytrex profile that no flying took place. I can't remember ever hearing an explanation from them on what was taking place. I do remember the words "GPS Drift" being used though.

So from what I learned from researching something similar to this in the past is, there are a number of clocks on the satellite. If these clocks are not dead-set with one another, this is when you will have "satellite drift". To best explain it would be, say each clock settings are off from each other by 750 milliseconds. Then suggest that 3 processes are to start at 10:00am. and when that time comes each process did take place as it was programmed to. But because the clocks were off in time from each other, it caused the satellite to do some unexpected stuff which creates a drift. When the satellite you are using to obtain stats with moves, it causes the GPS to take the hit as a result.

I can't say that this event would be what causes some of the known fly-aways or not. But I will say I do not think it is possible unless the drift was a large one. Keep in mind, it's not like no data is being received from the satellite while this happens.

 

[CapnBob]

Here is a 4 hour capture from this morning.
You can see a large anomaly to the northwest of approximately .55 miles, as indicated in the lower left area of the image.
The entire envelope of the capture is nearly 3/4 mile.
Also, as evidenced by the jagged appearance of the anomaly, it is not simply a single spike point, it is a progression out to maximum error and back again.

I am going to capture a log to see what the duration of these things are.

 

[flyNfrank]

I want to add that the highest percentage of fly-aways on the vPlus take place from those that do not take the time to read the manual. The next percentage come from mechanical failures of some kind.

I agree. However, there are flyaways experienced by veteran flyers that have been unexplained.
I believe this could be a possible cause

Well, if you really think about it, if someone has never experienced a fly-away, they are noob's as well in that area. I have had one, but I wouldn't say I was a veteran in that area.

Btw, what I posted on the Satellite Drift is indeed caused by clock time on the satellite. Also my Nexus 7 I use processes at 200 milliseconds. At any rate I have a feeling we're off on subject matter. I'm just sold more on the idea our gps issues come from a connection break, or humidity, condensation, or?

 

[aartsf]

Remember to switch to ATTI mode if you start to get a problem.

If one loses gps-lock, the bird will automatically swith to att-mode .... not??

 

[flyNfrank]

See here from my flytrex profile this one is almost 1 1/2hrs long. And this shows up on every non moving flight. It's common.