Re: NEC-LIST: Quick method on Complex Exponential Integration

Thanks for the very interesting information, pointers, and references!

I find the NEC4 modification of the cosine factor (C) of the current
expansion from C cos(...) to C (cos(...)-1) intriguing. I'll look it
up.

At first glance, the change in the C basis function seems to make the
obtention of more accuracy for the constant term essential, because of
the subtraction of nearly equal quantities when small segments are
concerned. Am I correct?

I'm curious to know whether you've done any formal analysis regarding
the effect on the accuracy resulting from the use of the improvement
in the computation method for the integral discussed. NEC2's INTX
performs the numerical integration to about 4 significant digits, if I
understand well the role of the test variable RX.

This apparently low numerical accuracy of one third of the matrix
entries makes me wonder why NEC2 was transformed into NEC2D. Is there
a reference justifying the use of double precision in NEC2D? Is this
strictly for numerical matrix solution reasons, resulting from the
matrix conditioning, or are there other benefits as well?

I think I remember a column by E.K. Miller on the subject of SP vs. DP
in the A&P Magazine, but can't find it at this time.

Also, have you applied the same kind of treatment to other similar
computations, such as the H field computation of uniform current
filament performed in routine HFK? (This integral is similar to the
INTX one, but with factors in R^2 and R^3 in the denominator).

Thanks, and Arrividerci!

Alexandre Kampouris

ps: I'll be away for 15 days starting tomorrow.
Received on Fri Oct 16 1998 - 09:48:06 EDT