[LINK] limits of technology in finding someone

[markus.buchhorn at anu.edu.au]

Stewart wrote:

> I remember when the USA had dithering on the system 
> (to deliberately reduce
> the accuracy to a few hundred metres, maybe even more).  
> At that time they
> boasted that their military accuracy, with no dithering, was a matter of
> centimetres.  
> 
> In fact, I distinctly remember one claim that it was measurable in
> millimeters, and that it was more accurate for house-block surveying
> than the conventional terrestrial triangulation methods.

It isn't exagerated, just misunderstood. Bear in mind that we are dealing 
with a radio signal (and all its signal/noise issues) of a clock ticking 
away on a satellite (and its accuracy issues). The underlying GPS 
system's signals provide positional XYZ information accurate to better 
than 10metres (see below though) for a handheld receiver. You can do 
better with a really decent antenna and amplifier, which makes it less 
portable, but fits in a backpack or a missile. Those systems can measure 
a position to within a meter or two. I've seen council staff using them 
in a few places.

You then have something called differential GPS (dGPS) which correlates 
variations in measured positions against a known location, i.e. a post in 
the corner somewhere which reads the same GPS satellites as your current 
mobile receiver, and there is an additional bit of stuff that does 
correlation between the two. Those things are very accurate, a fraction 
of a centimeter is possible. Various folks use them for geosciences, like 
earthquakes and fault movements. In those cases usually a jeep is parked 
in one corner of the valley and provides the reference point, and the 
mobile receiver is carted around within its radio range. The same system 
can be used for site surveys, precision agriculture, etc. 

Apart from the signal/noise issues within a small receiver, and having 
only some parts of the sky visible (especially if in only one direction) 
you're right that the shape of the earth makes life complicated. It's 
pretty well understood beyond the basic oblate-spheroid, against which a 
mean sea-surface level is applied (your altitude shouldn't vary with the 
tides...), but sure it's not perfect. The bigger issue is that for the 
vertical altitude you only have a small variation across most of the 
surface (most of us spend most of our time from -100m to +1500m in 
Australia) against a distance from the centre of the Earth of over 
6,000,000 meters. So a small error in the timing signals translates into 
a larger error in the vertical direction than horizontal. So you get sub-
10m in lat/long, but only sub-40m-ish in the vertical (with most non-dGPS 
receivers). (The maths is over my head here; it's all in the frequency 
domain).

Hope that clarifies things a bit?

Cheers,
      Markus