[NLRS] Computing distances [LONG]

[Donn Baker]

At 22:53 24-09-05 -0500, John P. Toscano wrote:
>
SNIP
>So which algorithm is "correct"?  Beats the heck out of me.  Donn puts 
>more faith in the method used by GL, and I'd be curious to know the 
>basis of that.  (I'm not trying to be argumentative or disputing his 
>choice, I'm trying to learn something.)

*******
Actually, the "reason(s)" are subjective more than qualitative.  Yes, GL
uses calculations for the earth as an oblate ellipsoid, not a sphere. So do
a lot of methods.  BUT, frankly, I prefer GL because it 1) runs on my Palm
Pilot, and 2) does things "right." The author, Rex, KK6MK, made it into a
real program. Excellent man-machine interface, error handling, the correct
options, and so on.  

As John says, the program uses the WGS84 data set as default.  This is an
international standard. The NAD27 data set (North American Datum, 1927
version), is for North America (optimized for NA, not the world.  There is
also a NAD83, but GL doesn't use that).  A spherical model is simple (you
can easily do the arithmetic with a pencil and paper), and for many cases
doesn't make much of a difference.  For hundreds of years, ships used a
spherical model and seemed to find their way with it OK.
*******

SNIP
>Now, if it was a random difference, sometimes + and sometimes -, then I 
>guess it would be nothing to worry about.  The problem is that the 
>method I used has a clear bias of calculating a smaller number than 
>BD2000, BD2004, or GL.  In my own log, the difference over 500 QSO's 
>(including one dupe that needs to come out), is 92,257 points (my 
>spreadsheet based on the AMSAT program) vs. 92,457 points (using the GL 
>method), or a full 200 points.  

*********
I'll bet a nickel that the difference is due primarily from the East-West
nature of the QSOs. The distance around the earth at the equator is larger
relative to the distance around the poles.  If you use a sphere as the
model, your E-W numbers will be smaller, and N-S distances will be smaller.
 I'd also bet a nickel that, with a random data set (i.e., more or less
equal numbers of N-S and E-W QSOs), the difference between the two models
would be a LOT smaller.
*********

73 Donn
WA2VOI/0