At 09:23 AM 7/28/99 , you wrote:
>Correct me if I'm wrong, but I was under the impression that the gps
receivers
>were set up to take their position at 1 second intervals.  Since their
time is
>set by one of the gps receivers in the sky, then they should have taken the
>readings at the beginning of each second, which will be within msec of each
>other.

There really is nothing that says a GPS receiver has to make 1s updates-
the PN
code that is used for SA is just 1ms long, so conceivably, you could make 1ms
position updates with a receiver that just uses code phase.

>Aside from that, the gps receivers select 4 satellites, one to get the
time, and
>three the triangulate the position.  Both receivers would have to select the
>same satelite for time, and the same three for triangulation, in the same
area
>for very good correlation.  There is a great chance that they would select
the
>same four satellites, but not necessarily assign the same roles to the same
>sats.

It is a bit more complicated- time is derived from all the satellites.  The
receiver
adjusts its clock to best fit the signal coming down from all the SVs (GPS
term for
satellite).  It is an iterative process- it makes a good guess, and
iterates to minimize
the time errors.  4 is the absolute minimum for a full 3-D solution, but
the DOP term
(dilution of precision) can shoot up if the SVs are in poor geometry (all
directly above,
for example).  If you use more SVs, you can make your time estimate more
accurate, plus
you can reduce the errors that occur when you lose or gain a SV. This one
reason why 12 channel receivers are becoming so popular.

>So, yes, I know I'm not going to get great positioning.  However, it will be
>significantly better than the 100M the SA affords.

True differential GPS is pretty hard to cobble up- because you have to have
continuous
code (or carrier) phase information.  You can do a pretty good, cheap
pseudo differential
GPS if your two receivers are close enough that they are using the same SVs
to generate
their solution.  At that point you can just fix one station and subtract
the two positions. This will only work well if you are using *the exact
same* SVs.  This becomes problematic in the real world because trees or
buildings may be blocking one or more of the SVs.

>Secondly, 1M resolution is unrealistic(ie, low probability of being within 1
>meter of the real position given by the mean of the plots) even if you are
>willing to stay in one place for an hour or longer.  You can average and
get to
>5M accuracy if you take one reading every ten minutes for 24 hours, and
after 36
>hours you can assume 1M.  After 36 hours the results don't improve much
UNLESS
>you have a receiver which is significantly better than mine.

To my knowledge the DOD has never published the actual statistics of the SA
clock
dithering.  With this knowledge you would be able to get an analytic answer
of how
good a position you can get with averaging.  All of the results I've seen are
from observations.

I second the fact about receiver quality.  Most consumer GPS receivers
don't have precision components, since SA limits the precision much more
than their receivers.  Long term averages are going to be skewed pretty
badly by the lack of ionospheric corrections unless you happen to happen to
have a dual channel (L1/L2) receiver with a really big memory.  The effect
of the ionosphere can be as much as a 40m error in the distance to the
SV, which varies with elevation angle above the horizon.  The Ionospheric
time delay is maximum at about 1400 local time, but it can vary by as much
as 25% (RMS) from the monthly
mean.  There are some corrections made, but it can still have a signifigant
effect.

Matt Bennett
mjb@arlut.utexas.edu
--
Matt Bennett
mjb@arlut.utexas.edu
512-835-3867