Influence of WiFi on GPS readings (200m error)

[Marcus Bauer]

On Mon, 2008-07-21 at 07:54 -0700, C R McClenaghan wrote:
> I'm not sure where the fault lies, but with tangogps, I seem to get a  
> scatter plot displayed for my position. Initially the fix seems  
> accurate, then it starts to wander, although I may not be. If I am  
> moving, the trial doesn't seem to follow the route I've taken. I think  
> it may be tangogps, but I thought I'd try using my garmin bluetooth in  
> a "side by side" kind of test to see whether it is tangogps or the gps  
> radio.

tangogps just uses whatever comes from the GPS chip. No mangling
whatsoever. You can test it by just dumping the output of /dev/ttySAC1
to a logfile, convert it to GPX or KML and use it with Google maps.

The u-blox chip is quite jittery (extremly so when used with an SD card
without the fixes, be it kernel or additional capacitor).

Additionally in a city the accuracy of any GPS while driving or walking
will deviate up to 50m from your real position. Just have a look at
openstreetmap GPX traces for various cities...

The performance of the u-blox chip (without an SD card) is impressive -
apart from that flaw the hardware guys at openmoko have done an
excellent job.

Last word about the jitter: most GPS chips support different modes, one
where they flatten out the GPS signal and a 'straight' one. Have a look
into the u-blox binary protocol and you will most likely find something
there to switch - I know that SIRF has something to this account. The
Globallocate chip in the Neo 1973 does flatten by default. However, you
need to go between 10m and 20m before it starts to 'believe' that you
move. The u-blox tells you after 3-4m.

Marcus

> 
> Chris
> 
> On Jul 21, 2008, at 12:02 AM, Russell Sears wrote:
> 
> > I've had this problem when the initial fix was obtained while I'm  
> > inside
> > a vehicle (such as a bus or car) or near a lot of metal.  Going to a
> > complete stop, then moving a few times (in the car) seems to give the
> > chipset a better chance to realize it's got a bogus fix.
> >
> > -Rusty
> >
> > Yaroslav Halchenko wrote:
> >> Hi All,
> >>
> >> Finally I got to play with FR. Flushed todays (0721) dev image and  
> >> kernel.
> >> Running TangoGPS. While WiFi is on (WEP encrypted. wpa_supplicant
> >> powered ;)) I am having location error (never hit the right spot)  
> >> around
> >> 200-400m with reported speed from 0.5 to 20 km/h (while I am siting
> >> steadily in 1 place).
> >>
> >> When I turn WiFi Off (just on my FR, without touching access point  
> >> or a
> >> laptop from which I am writing) -- location moves to the right spot  
> >> with
> >> a bias of 10m or so.
> >>
> >> Therefore, the question: is that expected? ie that we can't rely on  
> >> GPS
> >> readings while WiFi is on?  Or from the other side: what is
> >> 'documented' precision in GPS readings while WiFi is enabled (and not
> >> actually very actively used, if used at all  since I guess TangoGPS
> >> already had those tiles from OSM downloaded).
> >>
> >> Or may be it is just my FR which behaves that way? Did anyone observe
> >> any similar behavior?
> >>
> >
> >
> > _______________________________________________
> > Openmoko community mailing list
> > community at lists.openmoko.org
> > http://lists.openmoko.org/mailman/listinfo/community
> 
> 
> _______________________________________________
> Openmoko community mailing list
> community at lists.openmoko.org
> http://lists.openmoko.org/mailman/listinfo/community