[time-nuts] Positional accuracy of the M12+T

[Poul-Henning Kamp]

In message <20070103183054.8B5F9BE01 at ip-64-139-1-69.sjc.megapath.net>, Hal Murr
ay writes:
>
>> Position hold mode allows them to solve N(sat) equations with one
>> unknown instead of N(sat) equations with four unknowns. 
>
>That assumes you have good data for your location.
>
>How good is the find-my-location code?

It averages 10000 position samples and that is appearantly
good enough.

I played with plotting the per satelite residuals as a function
of a deliberately distorted set of coords for position hold
and it is quite insentitive.

The main effect of a slightly off position seems to be a diurnal
component to the estimated uncertainty more than an actual
component to the PPS output.

If you record the per sat residuals along with long+lat for the
sats and average the residual per az/elev gridpoint for a day
or two, you will be able to see a systematic effect but translating
this into a tangible delta to the held coordinates has so far
escaped me.

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.