[time-nuts] GPS location inaccuracies from a cell phone

[jimlux]

On 2/2/20 9:35 AM, Chris Wilson wrote:
> 
> 
>    02/02/2020 17:28
> 
> Hopefully not too off topic a question, but GPS experts abound here...
> 
> I am running a tracking device server on one of my PC's and an option
> is to use an app on a cell phone and the phone acts as a tracking
> device. But it shows seemingly random anomalies in position. For
> example I walked the dogs around the wood earlier. 99% of the tracking
> of the phone is correct, but I see occasional abnormalities where the
> track veers off into the distance to a "dead end" where I have
> certainly not taken the phone.
> 
> Could it be because I am not keeping the phone in a constant
> orientation? I do not see such anomalies with a "proper" tracking
> device, say in a vehicle? Where the trace veers off I may have been
> bending down burning some rubbish. Te phone would have been in the top
> pocket of my overalls. Any idea why these anomalies occur please? The
> track can be seen at http://www.chriswilson.tv/phone.jpg
> 
> 

GPS (and time derived from GPS) on a phone is a funny thing.  They 
obsess about energy consumption ( how many picojoules/fix kind of 
things).  Therefore, they do fixes on a sort of "as needed" basis and 
feed that to the API.  A couple things can screw this up:
The *legal* requirement driving GPS is the E-911 service, but they only 
need good accuracy when you're making a 911 call.
Most phones make heavy use of Assisted-GPS - the cellsite gives them an 
estimate of position and the approximate code phase and timing, so that 
acquisition can be fast (and consume small power).
A mapping application (or the phone API, I don't know) could also do 
some sort of "forward estimation" of position (i.e. you were heading 130 
degrees at 5 m/s so after 10 seconds, your position is X meters north 
and Y meters East.

Note also that the mapping applications target people *driving* so they 
do interesting things like snapping to likely positions (i.e. you're not 
likely to be in the middle of a river, so they snap to the road) and 
they filter out small velocity variations.  If you're walking, the 
application might shift to a different position filter (particularly in 
urban areas, where multipath is a reality, but position is aided by 
things like known WiFi access points, etc.)