[time-nuts] Position accuracy and Thunderbolt performance

[Poul-Henning Kamp]

In message <BLU125-W5F8BE19CBFB55CAB16B54CE170 at phx.gbl>, Mark Sims writes:


>The graphs are scaled to the same scale factors.  The jumps in the 200 foot
> graph are NOT correlated to satellite constellation changes, etc.  The s
>elf-reported performance is degraded by a factor of around 10.  

Are you 100% sure it is not correlated to sat's entering and leaving the
GPS solution ?

It looks exactly like what I saw when I played with something similar
(see last email). 

Remember that the GPS may not include the sat just because it clears
the mask angle, in particular, when you fudged the pos-hold, and have
TRAIM enabled, the sats will be thrown out near the horizon.

If you can correlate the jumps with sats being accepted and thrown out,
and know their azimuth at the time, you can tell which way you need
to move your pos-hold coords.

-- 
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.