[TowerTalk] tool to retrieve data and build HFTA profiles

Ian White GM3SEK gm3sek at ifwtech.co.uk
Tue Mar 6 00:14:05 PST 2012

Jim Lux wrote:
>i was thinking that once I run a profile to the edge of the dataset, I 
>could go through and "prune" it with some simple algorithm. Obviously, 
>you could eliminate ticks that vary less than some delta.  If someone 
>has a clever way, you could try to fit straight lines to "collections" 
>of points and replace them with the endpoints. But that's getting 
>pretty fancy..
>I'll be happy when I can just retrieve the files and generate the 
>profiles in ANY form.
When a friend delivered 360 files of detailed radial profiles out to 
100km from this site, it certainly concentrated the mind about the 
editing problem! This is messy terrain, with a lot of local detail that 
needs to be inserted, so I had to edit some of the files quite 
drastically to stay below the 150-line limit.

In my case the editing was done by hand in Notepad, periodically 
reloading the file into HFTA to view the changes. The algorithm was 
roughly as follows...

The profile only needs to run to the visible horizon (or perhaps a 
little beyond, where the horizon ends at a ridge) or to some arbitrary 
distance limit, whichever is the smaller.

Far distant horizons below 1deg elevation can be ignored.

If total number of data points >140 then:

   Start at the greatest distance (where small errors matter the least)

   Scan through groups of 3 lines, eliminating line #2 where height 
differs by <5m from average of #1 and #3

   Stop scanning at about 5km from tower.


In most terrains, that one-pass algorithm should free up enough data 
lines to allow for user insertion of closer-in detail that SRTM may have 
missed. From that viewpoint, it would be worth the effort of coding.

If not, then it's over to the user to edit the file manually. Viewing 
the profile in HFTA will often show what else can be removed.

The basic rule is that distant detail matters much less then detail 
closer to the base of the tower.

For example, level terrain and uniform slopes can usually be simplified 
to the first and last points. This particularly applies to large lakes 
or the ocean.

Terrain in the shadow of a low ridge can often be simplified to just a 
few points on the far side, until it becomes visible again at a greater 
distance. But leave the roll-over of the ridge at full detail so that 
HFTA can account for diffraction.

The acid test of what can be safely ignored is: does it affect the 
predicted elevation lobes? You may be surprised at what can safely be 
ignored at a far distance; and also surprised by the effects of some 
'minor' terrain features closer in.

If you see some unexpected effects, then save a copy of your original 
profile and try deleting/changing various terrain features until you 
find the one that is making a difference. The numerical difference may 
be open to debate, but HFTA will probably identify the correct terrain 
features that need to be bulldozed and hauled away [cue song].


73 from Ian GM3SEK

More information about the TowerTalk mailing list