[time-nuts] Crude Survey Technique

Poul-Henning Kamp phk at phk.freebsd.dk
Sun Nov 24 13:58:10 EST 2013


In message <8D0B752C7051AAF-1670-690FB at webmail-d157.sysops.aol.com>, johncroos@
aol.com writes:

>I realize that any number to the right of the 5th decimal point
>reported by T-Bolt mon is likely bogus. The data I reported earlier
>pretty much supports that. Now I am running tests with 3 units set
>for 10K, 20K, and 40K fixes to see how long each takes and to compare
>results obtained. 

Been there, done that, your result will depend on the time of day.

No, seriously.

Back when I had 10 GPS receives sitting on my roof:

	http://phk.freebsd.dk/raga/

I collected positions from all 10 for over a month, and tried all
sorts of ways to see what you could do with post-processing.

The primary result is that you need to average over N*12 hours,
since that is the major periodicity of the GPS orbits.

Once you do that, your result will have reasonably reproducibility,
as long as you start at the same time of day, not too many days
from each other.

The problem here is that the local reflection environment repeats
on the orbital 12 period, and it introduces biases into your
position.

The next experiment I tried, was more successful:  I adjusted the
averaged position, based on the position hold mode residuals:

	http://phk.freebsd.dk/raga/sneak/

This was an incredibly slow process, but after some months it
did in fact converge.

Because these are "remote" GPS receivers, where the downcable
signal is RS-422, temperature effects totally negated any
beneficial effect on the PPS signal, and I dropped it from
the project.

If somebody wants a interesting little project, try taking
that up again, and see if the 12h wiggles in the PPS signal
can be reduced that way.

-- 
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.


More information about the time-nuts mailing list