[Openstreetmap] Best time of day for urban canyon performance

[SteveC]

Hi schuyler,

As a back of the envelope guide to see how useful satellite availability
might be, I grabbed the data from

http://www.ngs.noaa.gov/GPS/GPS.html

and looked at the satellites visible within the vertical to some degree
over days and weeks. Results below, wanted to compare notes and see if
your stuff agrees, or there is another source to compare against?


Quick plot of an earth-neutral orbit (from that noaa data) and a plot
explained below here:

http://www.fractalus.com/steve/stuff/urbancanyon/


* within 30 degrees (60 degree cone looking vertically up from its point
on an idealised sphere) there is perhaps a 30-45 minute window
(depends on latitude) with 3 in view, per day

* this window precesses at about 15-30 minutes per week(!), depending on
latitude

* below 30 degrees it starts to get extremely unlikely you'll see more
than 2 sats. Given that, and a quick walk in a built up area, it doesn't
seem super useful to know when there will be 2 above as even low rise
center of london restricts to this range. Maybe when there will be
satellites east/west if you're on an east/west road is more useful.

* The number of satellites visible on average over a day decays
exponentially (the fit is quite nice) with drop in angle from vertical.
It looks like this:

degrees from vert | No. of sats
10     .281
20     .865
30     1.57
40     2.41
50     3.44
60     4.70
70     6.10
80     7.83
90     10.2   (90, eg can see all the sky)

This data was for someone at 45 degrees latitude, higher can be slightly
helped by the orbit plane.  Of course, visibility != locked on, nice
signal.

* The orbits are extremely stable and FAPP can be predicted from last
weeks data off that site by figureing the precess in. So long as they
don't change the orbit on purpose.

* You could test this by putting a receiver in a large shielded cone
pointing at the sky for 24 hours, for different sized cones and plot the
average and timewise number of locks. Cone might need to be a bit big,
error range would increase with angle.


That other plot shows number of sats visible for a 60 degree cone over a
week (animation). Each graph in the animation is 24 hours along the x
axis, animated to show the precess.


My conclusion is that it isn't worth looking at unless you take in to
account all sorts of error sources and really have difficulty with some
spot getting reception, and don't want to average data.

My other conclusion is that the GPS orbits are very well designed, and
have graceful degredation with visibility :-) I'm guessing all of this
is in a book somewhere, or on the web but I couldn't find it.

But, I have caveats:

I havn't looked at many months of data, the precess may be a wobble that
waxes and wanes. My Earth was an idealised sphere, the angle measured
using the dot product of vectors {surface_point ->
some_point_on_the_normal} and {surface_point -> satellite}. And of
course, I might have screwed up somewhere...

have fun,

SteveC steve at fractalus.com http://www.fractalus.com/steve/