[tlug] GPS is accurate

[John Fremlin <john@example.com>]

Jon Povey wrote:
> Curt Sampson wrote:
> > > Is GPS important to you -- for instance, would you want to track a
> > > child?  If so, you may want to NOT use softbank.  Apparently  GPS
> > > satellites (of all providers) shift to reposition themselves
> > > occasionally.  During that time, the GPS information will not be
> > > accurate.
> > Someone's fed you a line of total rubbish here.
>
>  From what I know about GPS this may not be rubbish. (I'm not an expert, 
> but I used to work with some).

(1) The GPS satellites are independent of your mobile telecoms service 
provider.

(2) It is true that the satellites are sometimes repositioned slightly. 
You can read all about it at 
http://www.navcen.uscg.gov/gps/gps_status_explained.htm

(3) However, the original GPS has a SATELLITE IS TEMPORARILY OUT bit in 
the Navigation Message (sent every 30s) and a warning SATELLITE WILL BE 
TEMPORARILY OUT bit to indicate when the satellite should not be 
trusted. The newer GPS signals apparently send a status bit every 6s.

So in practice, I would imagine it would be extremely rare for the GPS 
reading to be incorrect because of trusting a wrongly positioned 
satellite, My conclusion is that this idea is indeed probably rubbish 
and needs to be supported with some evidence.

> > GPS satellites are constantly shifting position; they're in orbit, after
> > all. But even when they move from one orbit to another, they always know
> > where they are, and will always provide accurate information which, if
> > you have it from four satellites (or three, if you have an accurate time
> > source), can be used to accurately compute your location. (See Wikipedia
> > or whatever if you want more details on this.)
>
> GPS signals have two components; one is the timed signal that you 
> measure delay to work out distance to the sat. The other is almanac data 
> which tells the receiver where the sats are. Normally the almanac data 
> takes a while to completely receive (low bandwidth) which is why cold 
> starts of most GPS receivers take so long to get a position.
>
> I *think* cellphones download the almanac over the phone data network to 
> allow faster GPS lock. 

The almanac and the ephemeris data can both be downloaded over the 
network. At least my Nokia phone does it and I guess Android will as 
well as it is much newer. http://en.wikipedia.org/wiki/Assisted_GPS

> I was certainly surprised how fast mine got a position when I tried it. 
> So, maybe the information you get through the 
> phone network can sometimes be out of date compared to what you would 
> get from the sats.. if they just adjusted orbits or somesuch?

I highly doubt that, unless someone actually goes up there and sneakily 
gives the satellites a shove that the operators weren't expecting or 
something (extremely improbable) like that.

> So, I'm not sure if this is the case, but maybe if you have an invalid 
> almanac you can get a bad position from GPS.

If you have an almanac or ephemeris data that you trust and that is 
wrong then definitely you will get a bad position but I believe that the 
engineering safeguards are solid and the chance of this happening is 
effectively negligible.

In practice, Google Maps on my Nokia phone tends to be quite happy to 
use the Softbank basestations for navigation and doesn't need a GPS lock 
so this whole discussion is focusing on the wrong thing entirely. . .