I've been comparing the results for an antenna I modded [1] with the demo
version of EZNEC and 4NEC2. Although EZNEC warned that the segments were
a little long measurement with an MFJ analiser shows a good match with
the predicted results. Being naturally tight I thought I'd try one of the
other modeling programs 4NEC2 which allows more segments. However I noticed
a discrepancy of about 1dbi in the predictions and a different predicted
SWR. These differences remain even when the 4NEC2 model is constructed with
the same number of segments as the EZNEC version and both models using
no ground (free space).

As both models are offering results to 2 decimal places I would expect
to be able to trust them to 1dp. Am I just kidding myself that modeling is
this accurate (Ignoring all the real world variations)?



[1] 3 element UK VHF FM broadcast radio to 3 element 2M
--
Rod
M0DTG
"I could be wrong, I could be right"

On Fri, 01 Apr 2005 20:08:18 +0100, wrote:

Am I just kidding myself that modeling is
this accurate (Ignoring all the real world variations)?

Hi Rod,

Usually the difference lies in the modeler (the loose nut behind the
steering wheel).

Do both models have the same diameter wire/rod for elements? Do they
both observe real world (copper/aluminum/aluminium) losses?

73's
Richard Clark, KB7QHC

Never confuse resolution with accuracy. The two decimal resolution
provided by EZNEC is sometimes useful in detecting which way the gain is
going when making changes. It's not meant or claimed to represent the
accuracy of the results. You should never expect accuracy of a model to
be within a hundredth of a dB, or anywhere near that. And I seriously
doubt the possibility of measuring an actual antenna to that accuracy.
In a lot of cases, 1 dB is pretty good for either modeling or measurement.

EZNEC, and I'm sure 4NEC2, uses NEC-2 for calculations. Both should
produce results essentially identical to NEC-2. (I say essentially,
because at least with EZNEC, there are always minor differences due to
such things as consolidation of constants scattered around the original
code, differences in calculation order and variable precision, and so
forth.) In general, if a model can be constructed that's identical to
the real antenna and its environment, the results will be strikingly
good. But that's not always possible. And there are a number of
situations where NEC-2 and therefore EZNEC does have difficulty
producing a good answer. One is when wires of different diameters are
connected together. This is covered in detail in the EZNEC manual
(integral to the demo program as well as the full programs) -- see
"Stepped Diameter" in the index. Other problems are often related to
source placement, which NEC-2 is very fussy about. These can sometimes
be fixed or corrected for, but sometimes not. The Average Gain test (see
"Average Gain" in the EZNEC manual index) allows you to spot and correct
for those kinds of problems.

The most common cause of reported discrepancy between EZNEC and NEC-2
and some other programs is that in EZNEC, you specify wire diameter
while for NEC-2 you specify radius. People often miss that when
comparing the two. Another common cause of differences -- not present in
your situation -- is when analyzing antennas near antiresonance (e.g.,
half wave vertical or full wave dipole), small changes in source
placement cause large changes in feedpoint impedance. In those cases,
differences in segmentation can produce substantially different
impedance results.

I have never seen a significant difference between EZNEC and NEC-2
results (except where I've unintentionally introduced an obvious bug
into the modified NEC-2 code incorporated in EZNEC), and it's my intent
that there not be any. I'd like very much to hear from anyone who thinks
he sees a difference. So far, all apparent differences have turned out
to be due to unintentional differences in the models, but if any real
discrepancy is present, I want to know about it so it can be tracked
down and corrected.

Roy Lewallen, W7EL

wrote:
I've been comparing the results for an antenna I modded [1] with the demo
version of EZNEC and 4NEC2. Although EZNEC warned that the segments were
a little long measurement with an MFJ analiser shows a good match with
the predicted results. Being naturally tight I thought I'd try one of the
other modeling programs 4NEC2 which allows more segments. However I noticed
a discrepancy of about 1dbi in the predictions and a different predicted
SWR. These differences remain even when the 4NEC2 model is constructed with
the same number of segments as the EZNEC version and both models using
no ground (free space).

As both models are offering results to 2 decimal places I would expect
to be able to trust them to 1dp. Am I just kidding myself that modeling is
this accurate (Ignoring all the real world variations)?



[1] 3 element UK VHF FM broadcast radio to 3 element 2M

Roy Lewallen wrote:

snip useful info

The most common cause of reported discrepancy between EZNEC and NEC-2
and some other programs is that in EZNEC, you specify wire diameter
while for NEC-2 you specify radius. People often miss that when
comparing the two. Another common cause of differences -- not present in
your situation -- is when analyzing antennas near antiresonance (e.g.,
half wave vertical or full wave dipole), small changes in source
placement cause large changes in feedpoint impedance. In those cases,
differences in segmentation can produce substantially different
impedance results.

I have never seen a significant difference between EZNEC and NEC-2
results (except where I've unintentionally introduced an obvious bug
into the modified NEC-2 code incorporated in EZNEC), and it's my intent
that there not be any. I'd like very much to hear from anyone who thinks
he sees a difference. So far, all apparent differences have turned out
to be due to unintentional differences in the models, but if any real
discrepancy is present, I want to know about it so it can be tracked
down and corrected.

Roy Lewallen, W7EL


Thanks for the information and offer. As a previous poster pointed put it
normally the operator. I'm going to spend some time this weekend looking to
see if I have done something wrong (likely) before wasting anyones times.

--
Rod
M0DTG

remove -- to reply

wrote in message ...
These differences remain even when the 4NEC2 model is constructed with
the same number of segments as the EZNEC version and both models using
no ground (free space).

As both models are offering results to 2 decimal places I would expect
to be able to trust them to 1dp. Am I just kidding myself that modeling is
this accurate (Ignoring all the real world variations)?

Just noticed above posting. To be 100% sure nothing went wrong, I
would love to take a further look at the cause of the differences
between both models. If possible could you send me the EZnec and (and
the (4)nec2) model...

Arie.

wrote:

snip

Just noticed above posting. To be 100% sure nothing went wrong, I
would love to take a further look at the cause of the differences
between both models. If possible could you send me the EZnec and (and
the (4)nec2) model...

Arie.

Thanks for the offer. As a previous poster pointed put it normally
the operator. I'm going to spend some time this weekend looking to see if I
have done something wrong (likely) before wasting anyones times.

--
Rod
M0DTG

remove -- to reply

I said

However I
noticed
a discrepancy of about 1dbi in the predictions and a different predicted
SWR. These differences remain even when the 4NEC2 model is constructed
with the same number of segments as the EZNEC version and both models
using no ground (free space).


And I can now confirm the discrepancy was 100% pilot error.

--
Rod
M0DTG

remove -- to reply