Since a discussion has been started on HFTA, I'll throw in a few observations.
The terrain around my QTH is complex, with many steep gullies and rapidly
changing elevations. We're not talking cliffs or mountains, but rolling terrain
intersected by many 30 to 50 ft deep gullies.
I once downloaded terrain data from the USGS web site. I download both DEM and
3rd arc-second NED data and generated terrain profiles for both using MICRODEM.
When plotting these two elevation profiles, they essentially lay on top of each
other. But if you look closely, there are differences of up to 4 feet at
various distances from the tower.
When I generate the HFTA takeoff profiles, however, there are large differences
between these two "essentially identical" elevation profiles. Some 14 MHz
examples: 7.5 db difference at a 6 deg takeoff angle for a 30 ft antenna height
(3 element Yagi) and 8 db difference at a 2.5 deg takeoff angle for a 40 ft
antenna height.
Sort of a butterfly effect ... small changes in elevation profiles cause large
changes in computed takeoff angle.
Interestingly, however, the HFTA Figure of Merit for these two elevation
profiles isn't very different.
I've had discussions with Dean on these observation and the preliminary
conclusion is that my terrain may cause a lot of spacial aliasing in the HFTA
algorithms. As we know, one must sample a sinusoid at a sample rate at least
twice the sinusoid's frequency (Nyquist rate). Similarly, when you sample
spacial variations (elevations along a radial, for example) one needs to sample
more frequently in range if the elevation changes a lot and quickly with range.
In my case, the 30 meter "range sample" may be too large to capture the "high
frequency" terrain variations between sample points. Or there could be
artifacts introduced by the terrain smoothing algorithms within the HFTA code.
USGS now has 10 meter range resolution data available for many locations, but
unfortunately HFTA can only accept 150 range points (including the tower
coordinate) along a given azimuth radial. So 10 meter data doesn't "reach out"
far enough in range to make a comparison of results with 30m and 10m range
sets.
I think HFTA is a wonderful program, but like any modelling code, it's
important to understand the limitations of that code and situations were
problems may arise. As Dean mentions in his HFTA instructions, trust the
results to +/- 3db. And if something looks funny or unrealistic. make a small
change in antenna height (a foot or two). If this small change in antenna
height causes large differences in takeoff angle plots, there's probably some
spacial aliasing and /or unrealistic diffraction going on along that azimuth
radial. Change the antenna height until you find one where a +/- 1 ft change in
height doesn't significantly change results.
I do think it's time for someone to pick up the cloak and try to improve on
HFTA. Being able to use 10m resolution data would be one improvement, and
perhaps some adjustments of the internal terrain profile smoothing filters.
As a practicing engineer, I prefer to know what's going on "under the hood" of
a model before I fully trust the results, or alternatively have its results
compared with a widely accepted model code's output. I don't know if there is a
"professional grade" electromagnetic ray tracing code out there one could use
for verification, but I'd love to see how HFTA stacks up against one so we can
avoid any pitfalls. Any takers?
The Navy's Advanced Refractive Effects Prediction System (AREPS) package may be
such an animal. See
http://www.public.navy.mil/spawar/Pacific/AP/Documents/sofAREPS_36.pdf and
http://www.public.navy.mil/spawar/Pacific/AP/Pages/SoftwarePrograms.aspx I'm
not sure since my professional area of expertise is not electromagnetic
modelling codes.
If anyone wants to see my HFTA results and elevation profiles behind this
discussion, send me an email off-reflector.
N3AE
_______________________________________________
_______________________________________________
TowerTalk mailing list
TowerTalk@contesting.com
http://lists.contesting.com/mailman/listinfo/towertalk
|