Last things first - I just read John's later posting, and rescued this
message from the out-tray. I hope this message will supply the extra
detail you need, John.

Just one final thing:
I trust this is not another one of those situations where there is an
attempt by vendors to "reinterpret" Maxwell's equations (or explain
things that Maxwell "left out").

Oh no. On that topic, I am an ironclad hardliner!

If you remember where we came in, Roy was mentioning a few types of
antennas that it is acknowledged cannot be modeled with NEC-based
programs. Cecil then inquired if the Lattin was one of those... and,
subject to correction, I think it may be (if you require one model that
covers all frequencies).

But every one of this small number of exceptions is for a clear and
understandable reason, so they don't change the big picture, which is
that "almost" all types of wire/rod antennas CAN be modeled accurately
by NEC. If anyone thinks NEC doesn't work for their own pet antenna, the
burden of proving that is entirely on them.


We now hand you back to the original reply...

J. B. Wood wrote:
In article , Ian White GM3SEK
wrote:

That isn't a complete model of this particular antenna. The missing part
is the velocity factor of the twin-lead when acting as a stub, which
means that the electrical length of the stub is different from the
physical length. Which of those two lengths would you use in the NEC
model?

The answer is easy for a single-band model; but it's not so easy to
create one NEC model that will be valid for all the bands this antenna
is designed to cover.

Hello, Ian. You would use the physical length for all wires that are
interconnected and/or separated by free space. After all, that's what
we're trying to model.

Certainly... but most of this antenna consists of pairs of parallel
wires that are physically interconnected, but are *not* separated by
free space - the wires that are part of the twin-lead.

You still must decide how many electrically-small
segments would constitute, say, a 1 foot length of conductor. The higher
the frequency, the more segments you will need. If transmission line is
to be connected between segments, NEC has tools for doing that. BTW, my
experience is with LLNL's NEC-4 (FORTRAN-77 source code) rather than the
commercially-available packages. Sincerely,

Sorry, that model still wouldn't work (unless I've misunderstood the
principle of this antenna).

The whole point of modeling a multiband antenna is to get one model that
is good for all its operating frequencies. That allows us to check that
the SWR dips at all the right places, and to find out what's really
happening in the supposedly "non-operative" parts of the antenna.

AIUI, the central part of the Lattin antenna is a half-wave dipole at
the highest operating frequency - call it 30MHz, so the wavelength is a
nice round number, 10.0m. Outside each end of this 5m long dipole is a
quarter-wave stub made of twin-lead. These stubs are resonant at 30MHz,
so they cut off the rest of the antenna (much like a trap) leaving just
the central half-wave dipole as the only functional part at of the
antenna.

The normal differential-mode velocity factor of the twin-lead applies to
this stub, so its correct physical length is not a quarter-wavelength
(2.5m) but about 0.8*2.5m = 2.0m.

Moving to the next lower operating frequency, there will be another pair
of quarter-wave resonant stubs isolating the ends of a half-wave
resonant dipole. But part of the physical length of this longer dipole
is the 30MHz stub. If you model it at its true physical length of 2.0m,
this will be correct for the lower frequency, but if you ignore the
differential-mode velocity factor, the stub won't be resonant at 30MHz
any more.

So the question remains: how can we model this "simplest" case of a
two-band Lattin antenna, in a way that will be accurate at both
frequencies? If we can solve that one, then extending it to the full
5-band Lattin should be child's play :-)




--
73 from Ian GM3SEK
http://www.ifwtech.co.uk/g3sek