r.los, or looking for other line of sight analysis

[Tom Poindexter]

I'm trying to do some line of sight analysis over a large area, and find that
r.los doesn't seem to work well.  My situation is trying to estimate
coverage for VHF/UHF radio propogation (largely line of sight).  In my
local area (Denver, Colorado), I use amateur radio repeaters that are
situated at 10,500 ft. MSL.  From my house, at roughly 5,700 ft, I can see
the mountain top where the repeater is located, 40 or so miles away.  

When I run r.los, with the observer point at the repeater, the lower elevation
areas are never selected as visible on the output raster.  In addition, r.los 
takes hours to run with my data, 1:250,000 DEMs converted to GRASS rasters.
I've browsed through the r.los code, and it appears to me that it's geared
primarily for short distances, and assumes a flat earth (?).

Does anyone know of alternate line of sight analysis for GRASS?  Or possibly
a decent algorithm that I might be able to implement?

-- 
Tom Poindexter
tpoindex at nyx.net
http://www.nyx.net/~tpoindex/