On Mar 12, 11:01 am, Art Unwin wrote:
Scott, you are introducing sharp corners to your new antenna so NEC
will stray from accuracy.

Is NEC4 better in this regard?

Here is the NEC2 reported performance of my current design. Its a 24
foot boom, with 5 bands. 4 elements on 15-10 and 3 elements on 20-17.
Optimization is for F/R around mid-band and bandwidth (under 2.5:1 at
edges).

http://remote.wu2x.com:8888/WU2X/

Click each band.html file to see the graphs on each band. The non-
active driven elements are shorted.

Here is the stock Cubex 4 element quad, which is 4 elements on all
bands (on a 24' boom).

http://remote.wu2x.com:8888/Cubex/

-Scott, WU2X

On Mar 12, 10:22 am, wrote:
On Mar 12, 11:01 am, Art Unwin wrote:

Scott, you are introducing sharp corners to your new antenna so NEC
will stray from accuracy.

Is NEC4 better in this regard?

Here is the NEC2 reported performance of my current design. Its a 24
foot boom, with 5 bands. 4 elements on 15-10 and 3 elements on 20-17.
Optimization is for F/R around mid-band and bandwidth (under 2.5:1 at
edges).

http://remote.wu2x.com:8888/WU2X/

Click each band.html file to see the graphs on each band. The non-
active driven elements are shorted.

Here is the stock Cubex 4 element quad, which is 4 elements on all
bands (on a 24' boom).

http://remote.wu2x.com:8888/Cubex/

-Scott, WU2X

I would suggest that you check with what Cebik says about it.
At the same time you are also now dealing with vertcal polarisation
which also has a lot of characteristics that are different from planar
yagis where the elements are not in equilibrrium as compared to a quad
design.
One can discharge corona where the other doesn't soto use the same
calculations
for designs that behave differently raises a lot of questions besides
the sharp corners.
Niether of the programs are perfect but better than shooting in the
dark.
Art

wrote:
On Mar 12, 11:01 am, Art Unwin wrote:
Scott, you are introducing sharp corners to your new antenna so NEC
will stray from accuracy.

Is NEC4 better in this regard?


Neither NEC-2 nor NEC-4 has any problem in this regard. Art's statement
is incorrect as quoted.

Roy Lewallen, W7EL

On Mar 12, 2:27 pm, Roy Lewallen wrote:
wrote:
On Mar 12, 11:01 am, Art Unwin wrote:
Scott, you are introducing sharp corners to your new antenna so NEC
will stray from accuracy.

Is NEC4 better in this regard?

Neither NEC-2 nor NEC-4 has any problem in this regard. Art's statement
is incorrect as quoted.

Roy Lewallen, W7EL

That really is good news. Capacitive coupling between close spaced
wires,
progressive change in capacitive coupling to ground from a
verticle ,or wires
at different spacings and closeness must be a devil to calculate tho I
suppose you can also
insert fudge factors for alignment from impirical results which I
presume is the reason
for program modifications despite the inflexibility to change of
Maxwells laws.
Well you highlight one of the answers the gentleman was asking for so
a forthcoming
complete answer from you should give him closure on this subject
Art.

Art Unwin wrote:
On Mar 12, 2:27 pm, Roy Lewallen wrote:

wrote:

On Mar 12, 11:01 am, Art Unwin wrote:

Scott, you are introducing sharp corners to your new antenna so NEC
will stray from accuracy.

Is NEC4 better in this regard?

Neither NEC-2 nor NEC-4 has any problem in this regard. Art's statement
is incorrect as quoted.

Roy Lewallen, W7EL


That really is good news. Capacitive coupling between close spaced
wires,
progressive change in capacitive coupling to ground from a
verticle ,or wires
at different spacings and closeness must be a devil to calculate

tedious, but not complex. Like all Finite Element approaches, you
basically break the problem up into little tiny chunks and rigorously
apply basic laws of physics. Charge is charge.

The challenge is not in the theory of operation, but in the practical
implementation.


There's also the practical implementation detail of taking a high level
description of something and turning it into many, many smaller pieces.
(For instance, turning an arbitrary 3-d shape into lots of little plane
triangles or quadrilaterals). You want a small number of pieces so the
computational work is less (many of these techniques have "work" that
goes as the cube of the number of pieces, so going from 10 segments to
100 segments takes 1000 times as much computation), but also you want
the pieces small enough that they approximate the original continuous
curve to an adequate level of accuracy (the calculus problem)


As you make the chunks smaller, round off errors and numerical precision
become a bigger issue (e.g. on a computer with finite precision, summing
a million millionths might not equal one). So even if you have a 1000
processor Beowulf cluster, it might not help.



tho I
suppose you can also
insert fudge factors for alignment from impirical results which I
presume is the reason
for program modifications despite the inflexibility to change of
Maxwells laws.


No fudge factors in NEC. What changes there are between versions do
things like:

handle insulated wires, or wires embedded in a medium other than free
space. If you accept the constraint of uninsulated wires in a vacuum,
you can simplify the equations, which takes less computation (i.e. you
don't have to take epsilonr or sigma into account). Since air is pretty
close to a vacuum, and most people build antennas out of uninsulated
wires, the first version of NEC did the easy case.

better approximations of the charge distribution on the segment from a
numerical analysis standpoint. i.e. rather than using sin(x) for values
of x near pi/2, where small changes in x result in very small changes in
sin(x), you use 1-cos(x). You could have also just used a zillion
digit sin calculation, but that gets back to the computational
efficiency thing.

More accurate calculations of the interaction between chunks. NEC
essentially calculates the coupling between every possible pair of
segments in your model. Calculating coupling between two segments some
distance apart assuming the segment is very much smaller in diameter
than the spacing and where they are parallel is fairly straightforward.
Calculating coupling between two conductors of diameter d, separated
by a distance close to d, with them at an angle, is a bit tougher. (a
lot of it is back to the issue of precision of trig functions)

Easier ways to define a model. NEC4 includes a function to enter a wire
with a catenary curve. For NEC2, you'd have to do that outside, and
then enter the wire as a series of smaller wires.


There's a readily available paper out there that explains all the
improvements from NEC2 to the later versions, and how they were
experimentally validated.

{https://e-reports-ext.llnl.gov/pdf/210389.pdf}

Well you highlight one of the answers the gentleman was asking for so
a forthcoming
complete answer from you should give him closure on this subject
Art.


Here's the reference to the paper by Burke:
Title: Recent advances to NEC (Numerical Electromagnetics Code):
Applications and validation
Authors: Burke, G. J.
Affiliation: Lawrence Livermore National Lab., CA.
Journal: Presented at the Conference on Modern Antenna Design Using
Computers and Measurement Application To Antenna Problems of Military
Interest, Ankara, Turkey, 19-20 Oct. 1989
Publication Date: 03/1989
Category: Communications and Radar
Origin: STI
NASA/STI Keywords:
ANTENNA DESIGN, COMPUTER AIDED DESIGN, COMPUTER PROGRAMS, ELECTROMAGNETISM,
MATHEMATICAL MODELS, ELECTRICAL INSULATION, INTEGRAL EQUATIONS, WIRING
Bibliographic Code:
1989STIN...9011917B

Abstract

Capabilities of the antenna modeling code NEC are reviewed and results are
presented to illustrate typical applications. Recent developments are
discussed that will improve accuracy in modeling electrically small
antennas, stepped-radius wires and junctions of tightly coupled wires, and
also a new capability for modeling insulated wires in air or earth is
described. These advances will be included in a future release of NEC,
while for now the results serve to illustrate limitations of the present
code. NEC results are compared with independent analytical and numerical
solutions and measurements to validate the model for wires near ground and
for insulated wires.

On Mar 12, 6:13 pm, Jim Lux wrote:
Art Unwin wrote:
On Mar 12, 2:27 pm, Roy Lewallen wrote:

wrote:

On Mar 12, 11:01 am, Art Unwin wrote:

Scott, you are introducing sharp corners to your new antenna so NEC
will stray from accuracy.

Is NEC4 better in this regard?

Neither NEC-2 nor NEC-4 has any problem in this regard. Art's statement
is incorrect as quoted.

Roy Lewallen, W7EL

That really is good news. Capacitive coupling between close spaced
wires,
progressive change in capacitive coupling to ground from a
verticle ,or wires
at different spacings and closeness must be a devil to calculate

tedious, but not complex. Like all Finite Element approaches, you
basically break the problem up into little tiny chunks and rigorously
apply basic laws of physics. Charge is charge.

The challenge is not in the theory of operation, but in the practical
implementation.

There's also the practical implementation detail of taking a high level
description of something and turning it into many, many smaller pieces.
(For instance, turning an arbitrary 3-d shape into lots of little plane
triangles or quadrilaterals). You want a small number of pieces so the
computational work is less (many of these techniques have "work" that
goes as the cube of the number of pieces, so going from 10 segments to
100 segments takes 1000 times as much computation), but also you want
the pieces small enough that they approximate the original continuous
curve to an adequate level of accuracy (the calculus problem)

As you make the chunks smaller, round off errors and numerical precision
become a bigger issue (e.g. on a computer with finite precision, summing
a million millionths might not equal one). So even if you have a 1000
processor Beowulf cluster, it might not help.

tho I

suppose you can also
insert fudge factors for alignment from impirical results which I
presume is the reason
for program modifications despite the inflexibility to change of
Maxwells laws.

No fudge factors in NEC. What changes there are between versions do
things like:

handle insulated wires, or wires embedded in a medium other than free
space. If you accept the constraint of uninsulated wires in a vacuum,
you can simplify the equations, which takes less computation (i.e. you
don't have to take epsilonr or sigma into account). Since air is pretty
close to a vacuum, and most people build antennas out of uninsulated
wires, the first version of NEC did the easy case.

better approximations of the charge distribution on the segment from a
numerical analysis standpoint. i.e. rather than using sin(x) for values
of x near pi/2, where small changes in x result in very small changes in
sin(x), you use 1-cos(x). You could have also just used a zillion
digit sin calculation, but that gets back to the computational
efficiency thing.

More accurate calculations of the interaction between chunks. NEC
essentially calculates the coupling between every possible pair of
segments in your model. Calculating coupling between two segments some
distance apart assuming the segment is very much smaller in diameter
than the spacing and where they are parallel is fairly straightforward.
Calculating coupling between two conductors of diameter d, separated
by a distance close to d, with them at an angle, is a bit tougher. (a
lot of it is back to the issue of precision of trig functions)

Easier ways to define a model. NEC4 includes a function to enter a wire
with a catenary curve. For NEC2, you'd have to do that outside, and
then enter the wire as a series of smaller wires.

There's a readily available paper out there that explains all the
improvements from NEC2 to the later versions, and how they were
experimentally validated.

{https://e-reports-ext.llnl.gov/pdf/210389.pdf}

Well you highlight one of the answers the gentleman was asking for so
a forthcoming
complete answer from you should give him closure on this subject
Art.

Here's the reference to the paper by Burke:
Title: Recent advances to NEC (Numerical Electromagnetics Code):
Applications and validation
Authors: Burke, G. J.
Affiliation: Lawrence Livermore National Lab., CA.
Journal: Presented at the Conference on Modern Antenna Design Using
Computers and Measurement Application To Antenna Problems of Military
Interest, Ankara, Turkey, 19-20 Oct. 1989
Publication Date: 03/1989
Category: Communications and Radar
Origin: STI
NASA/STI Keywords:
ANTENNA DESIGN, COMPUTER AIDED DESIGN, COMPUTER PROGRAMS, ELECTROMAGNETISM,
MATHEMATICAL MODELS, ELECTRICAL INSULATION, INTEGRAL EQUATIONS, WIRING
Bibliographic Code:
1989STIN...9011917B

Abstract

Capabilities of the antenna modeling code NEC are reviewed and results are
presented to illustrate typical applications. Recent developments are
discussed that will improve accuracy in modeling electrically small
antennas, stepped-radius wires and junctions of tightly coupled wires, and
also a new capability for modeling insulated wires in air or earth is
described. These advances will be included in a future release of NEC,
while for now the results serve to illustrate limitations of the present
code. NEC results are compared with independent analytical and numerical
solutions and measurements to validate the model for wires near ground and
for insulated wires.

Jim, that was a comprehensive answer for the gen tleman. Probably more
information
that he can deal with.Seems like it is quite easy to have a yagi
behave when using NEC2
but I still hear stories of the diffuculty of tuning them after
following the nec2 instructions.
Never had a quad before so I am really out in left field on that one
Regards
Art

Jim Lux wrote:

tedious, but not complex. Like all Finite Element approaches, you
basically break the problem up into little tiny chunks and rigorously
apply basic laws of physics. Charge is charge.

The challenge is not in the theory of operation, but in the practical
implementation.

. . .

The last time I read any of Art's postings, he was using AO, which is
MININEC-based. MININEC has a number of peculiarities due to choices made
in its implementation, including problems with close spaced parallel
wires, wires connected at an angle, and very simplified ground model.
(See "MININEC: The Other Edge of the Sword", _QST_, Feb. 1991 for more
detail.) NEC-2 and NEC-4 also have peculiarities due to their
implementations, but ones which are different from MININEC and in some
cases from each other.

Roy Lewallen, W7EL

Jim Lux writes:

Here's the reference to the paper by Burke:
Title: Recent advances to NEC (Numerical Electromagnetics Code):
Applications and validation
Authors: Burke, G. J.
Affiliation: Lawrence Livermore National Lab., CA.
Journal: Presented at the Conference on Modern Antenna Design Using
Computers and Measurement Application To Antenna Problems of Military
Interest, Ankara, Turkey, 19-20 Oct. 1989
Publication Date: 03/1989

Thanks.

Recent advances, dated 1989.

What's been going on during the 18 years since then? Polishing NEC-4,
and developing other codes for specialized purposes, presumably. I
know it's a big question, but maybe somebody here can tell us
something.

73
LA4RT Jon, Trondheim, Norway

Jon Kåre Hellan wrote:
Jim Lux writes:


Here's the reference to the paper by Burke:
Title: Recent advances to NEC (Numerical Electromagnetics Code):
Applications and validation
Authors: Burke, G. J.
Affiliation: Lawrence Livermore National Lab., CA.
Journal: Presented at the Conference on Modern Antenna Design Using
Computers and Measurement Application To Antenna Problems of Military
Interest, Ankara, Turkey, 19-20 Oct. 1989
Publication Date: 03/1989


Thanks.

Recent advances, dated 1989.

What's been going on during the 18 years since then? Polishing NEC-4,
and developing other codes for specialized purposes, presumably. I
know it's a big question, but maybe somebody here can tell us
something.

That's about it. NEC-4 is sort of about as good as you can get for a
"wires" based MoM code. The math isn't changing, the implementation does
the math, etc.


These days, it's other modeling techniques that are being developed:
various FDTD schemes
various voxel based schemes
all manner of harmonic balance

lots of work on automatic meshing (i.e., take the mechanical solid model
from Unigraphics or Pro/E or Solidworks or whatever and turn it into
something that a EM code can process)







73
LA4RT Jon, Trondheim, Norway