[time-nuts] GPS multipath and SNR (was: Precision Time Protocol – Windows 10 implementation)

[Attila Kinali]

On Mon, 12 Aug 2019 12:05:14 +0200
Adam Kumiszcza <akumiszcza at gmail.com> wrote:

> I cannot see any difference between these positions. I've made 10
> measurement for each position, getting 12 SNR values for satellites for
> each measurement and then calculated the average of these.
> Hanging antenna had 25.18 SNR average (2.21 standard deviation). Antenna on
> a ground plane had 26.10 SNR average (2.08 standard deviation).
> https://docs.google.com/spreadsheets/d/1lvGFTQS62MqiQALjHmWFZRY3hb5O1OqANFGBhnCaEUo/edit?usp=sharing

SNR is a bad indicator of multipath. Multipath changes the SNR by
at most 3dB. And the change can go in either direction. For GPS L1 C/A
the multipath can affect the timing solution by up to 30ns (IIRC, I don't
have a reference at hand). That's much more than you'd expect from just
a 3dB degradation, and definitely not what you would expect from a 3dB
increase in signal strength. The reason for this is that the multipath
signal is correlated with the signal and affects it differently than just
random noise.

If you want to see what the effect really is, you have to measure the
output of your GPS to a more stable standard (at the very least a good
rubidium vapor cell standard) and look at the variations with a tau 
up to 24h. Note that multipath is usually correlated on a 12h/24h basis,
but it is not necessarily a strong correlation as the signal path changes
slightly due to changing satellite orbit and changing lensing effects
of the atmosphere. How much correlation you get depends on a lot of
factors that are not easy to quantify in an uncontrolled environment.

			Attila Kinali

-- 
<JaberWorky>	The bad part of Zurich is where the degenerates
                throw DARK chocolate at you.