[time-nuts] nuts about position

Poul-Henning Kamp phk at phk.freebsd.dk
Tue Jun 5 15:33:00 EDT 2018


--------
In message <SN1PR11MB10248E0077E3055B2286E92ECE660 at SN1PR11MB1024.namprd11.prod.outlook.com>, Mark Sims writes:

>The next thing I want to try is a receiver self-survey vs the post processed results...

I fell into that trap many years ago as well :-)

The post processed result eliminates and compensates for a lot of
effects which are invisible to the receiver, that is the entire
point of post processing.

But if your goal is to get good 1PPS out of your receiver, forcing
it to hold a "perfect" position, means asking it to deal with effects
it cannot measure, and as a result you get *worse* 1PPS performance
than if you use the "wrong" self-survey position.

That said, there are certainly a lot of issues with how the self-survey
position should be calculated *and* maintained, because it really needs
to change over time for optimal performance.

Almost any suboptimal antenna-position means that you should vary
the position-hold coordinates over the orbital period (but see
below), but over longer periods also environmental factors come
into play.

For instance if you put the antenna on a wood-construction, be it a
pole or a house, it coordinates will vary at the cm level over the
seasons as the wood expands and contracts.

The good news is that the receiver gives you the input data to work
and model these biases, the bad news you don't get a lot of data.

I did some experiments when ten Oncore M12+T's passed through my lab
many years ago:  I plotted the reported "residuals" vs.  satellite
position, and I tried to update the held position on a daily basis
using the trend, and the results were measurable at the several ns
level:

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

I didn't persue this further, because my latitude is particular
unfavourable to the simple (N/S+E/W) approach (no birds north of
me) and with only a single receiver the amount of input data is too
low for such a crude algorithm.

A more complex geometrical model could probably get more mileage
out of the sparse data, as would tracking more birds than the
12 the M12 could track.

-- 
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