The principle is reasonably straight forward. Your GPS receiver gets 'almanac data' from the constellation which defines where each satellite is in space.
A time distance speed of light calc yields a range from the satellite. So, you now know that you are somewhere on the surface of a sphere, of the calculated radius, from that known point in space that the satellite is currently at.
Fix another satellite, and this gives you another position sphere about another point is space.
The intersection of 2 spheres will yield a position circle, (think of 2 soap bubbles blown together).
Get another range from another (3rd) satellite. This gives you another sphere. Intersect this with our circle, and it will yield 2 points.
Normally, this is enough to give you a fix, because 1 of the 2 points will be unfeasable (say 500 miles underground, or 8,000 miles into space).
For best accuracy, another satellite (or an altimeter input, which provide a position sphere of its own) is required. The reasons for this are tricky to explain in words alone, and lend themselves to a diagram or 2. Suffice to say its because the clock inside your receiver is not accurate enough, but an extra fix provides a way of mathematically figuring out the clock inaccuracy.
So 4 sats = good fix.
3 sats = approx fix.
3 sats + 1 alt = better than 3, worse than 4.
Does that help?
CPB.
[This message has been edited by Capt Pit Bull (edited 14 December 2000).]