NEC-LIST:NEC 4 Glitch

Bob,

I tried a single plate with 1526 patches, writing a NGF. I ran codes
compiled for MAXMAT of 3600 and 7500. The 3600 size stores the
matrix on disk, since the number of unknowns is 3052, and when the
NGF is written the code saves half of the disk space for future use.
Runs with both code sizes completed successfully. Don't know if
there is some problem with the specific MAXMAT size that you used.
You can change the size of the matrix reserved for future use with
the NGF by changing the value of IRESRV in the code. Also, if a XQ
command comes before the WG the code will fill and factor the matrix
using the whole MAXMAT**2 space, and then write the NGF. That may
not leave any space for adding to the model, but you can run a code
with larger MAXMAT for that.

Of course, the single flat plate is not a valid model, since the the
back side of the patches is exposed. The waveguide would not be
valid either unless the ends are closed. Even then I don't think it
would give good results when wires are connected inside. The NEC
treatment of a wire connected to a patch surface is not good. Dick
Adler has gotten results for a half wire loop on top of a patch
pedestal that give a capacitive impedance for the loop when it should
be inductive. I compared NEC with EIGER on that, and NEC is off by a
large capacitive shift. The reason is that, due to the integration
method for the patches in NEC, the field of the point charge on the
end of the wire is not accurately canceled by the numerical
integration of the 1/rho current on the patches about the connection
point. I would use a wire grid model for the waveguide, but an
actual EFIE patch model would be better.

Jerry Burke
LLNL

-- 
The NEC-List mailing list <nec-list_at_gweep.ca>
http://www.gweep.ca/mailman/listinfo.cgi/nec-list