The question of antenna modeling convergence testing has arisen several
times in the past month, so I put some notes specifically on the subject
at my website--too long to post. Basically, do not trust any
model--especially those you may generate with automated
segmentation--until you have tested it for convergence. A model may be
considered converged when the significant output data (for hams, gain,
front-to-back ratio, feedpoint resistance and reactance, and current
magnitudes and phases) does not change significantly when the number of
segments per wire is increased by a significant number (for example,
doubling the number of segments per wire). What counts as a significant
change in output data varies with the purpose of the model, but less than
1.5 to 2% per data item may be a usable general figure. Some data may be
stable, while other output numbers may wander, indicating a model that
needs more segments per wire for reliability. However, every method of
moments program will prove unreliable for some types of models. Some
types of models, such as antennas with only linear elements in the
region of 1/2 wl each, are converged at the minimum number of elements;
others, such as those using complex geometries, may require many more
segments per half wavelength to achieve convergence; and some may never
converge at all. The Web page contains some examples, which may grow in
number with time. I hope the information proves useful to some.
-73-
LB, W4RNL
L. B. Cebik, W4RNL /\ /\ * / / / (Off)(423) 974-7215
1434 High Mesa Drive / \/ \/\ ----/\--- (Hm) (423) 938-6335
Knoxville, Tennessee /\ \ \ \ / / || / (FAX)(423) 974-3509
37938-4443 USA / \ \ \ \ || cebik@utk.edu
URL: http://funnelweb.utcc.utk.edu/~cebik/radio.html
--
FAQ on WWW: http://www.contesting.com/towertalkfaq.html
Submissions: towertalk@contesting.com
Administrative requests: towertalk-REQUEST@contesting.com
Problems: owner-towertalk@contesting.com
|