[TowerTalk] Re: linear loading

w8ji.tom w8ji.tom@MCIONE.com
Wed, 18 Nov 1998 18:05:57 -0500

Hi Cebik,

I was hoping you'd say something, since I know you modeled this type of
thing a lot.
I appreciate your work, some of it is really detailed.

> NEC (both -2 and -4) have a weakness in modeling two different
> both of which enter into modeling linear loads.  1.  NEC does not handle
> angular junctions of wires having dissimilar diameters accurately.  Most
> linear load arrangements use smaller diameter wire for the linear load
> than for the main element.  2.  NEC does not handle accurately closely
> spaced wires of different diameters.  

It is also my understanding NEC also does not allow for the normal
"pushing" or "pulling" (depending on phase) of current to one side of a
conductor when it is placed close to another conductor. That largely
ignored effect increases the resistance beyond normal skin effect
calculations when conductor diameter approaches the conductor spacing. It
is very evident in a wide strap or tightly coiled conductor for example,
but only increases resistance a few percent in a round conductor spaced a
few wire diameters from another current carrying conductor.

> layout and load positioning.  They raise some interesting questions of
> their own.  For example, consider a linear load that ostensibly begins at
> mid-element and is brought toward the dipole center and then returns to
> mid-element.  Now consider a pair of linear loads that begin at the
> and proceed outward, turn around and return to center to join the
> so-called main element.  When one models all wires with the same
> there is no electrical difference between the two systems, except for the
> collinear or non collinear placement of the outer end of the element, and
> the offset is relatively insignificant.

I think you are saying is it makes no difference if the linear loading
wires start out on the element and fold back to the center (let's call that
the "feedpoint" so we have a good mental picture), or if they start at the
"feedpoint" and fold outwards along the element to the same distance.

That makes sense, since common mode current distribution would be
essentially the same.

> When linar loads are centered and the wires are equispaced from the main
> element to follow, they answer very closely to stub reactance equations.

That's the way I've seen it work in "real antennas". Stubs are good loading
devices, but I can often do better with a lumped inductor of good form
factor (especially when a small hat is included). The more reactance
required, the more results skew in favor of the lumped inductor **if** that
inductor is placed at a location where common mode current distribution
remains essentially the same as with the stub.

> When the wires are not symmetrical with respect to main element, current
> along the load wires differ and the requisite length differs from stub
> transmission line calculations.  As linear loads are moved outward,
> currents on the wires differ, regardless of symmetry with respect to the
> main element, and stub calculations also fail to coincide with requisite
> modeled lengths.

That's because the stub becomes part of the radiating system, it radiates
and so is both an antenna and a stub. The current and voltage does not
follow "stub rules" when it is both a stub and radiator or when a third
conductor interacts with the stub through coupling. 

>  However, models of physical stubs (when no modeling
> limitation is exceeded) do correspond within shop limitations with
> constructed antennas.  Using TL models can be misleading relative to both
> linear load length and losses (since TL models are lossless).

But linear loading stubs, contrary to folklore, are NOT lossless. They
might well be lossier than another loading method, like an end-hat and a
lumped inductor (generally the most efficient system).
> Modeling stubs as physical wires tends to provide reasonably accurate
> theoretical figures for such matters as bandwidth, source impedance,
> current levels and phases, etc., but only for models that do not exceed
> program limitations.  However, every commercial implementation of a
> load of which I am aware fails to model accurately due to exceeding one
> more limitations of existing programs.

That's sure the truth! The "modeler" generally includes all losses in the
lumped inductance by modeling it as a certain resistance in series with a
certain reactance (defined as "Q") while the linear load model violates
rules and ignores losses. Fractally loaded antennas are one example where
incorrect modeling has resulted in "pie-in-the-sky" efficiency claims for
loaded antennas, and an avalanche of pathological science caused by flawed

In the real world linear loading and lumped loading are essentially the
same in performance ***unless something is "done wrong" in one of the
systems***. I've found it easier to optimize lumped loading, and have had
better results with lumped loading systems.

There is too much unfiltered pathological science presented to amateurs,
like the 5 dB gain increase claimed by adding a second driven element in a
quad. Lossless loading is another one. If something makes a significant
difference in a similar size antenna, it is because something was seriously
wrong with the other comparison system.

That still doesn't stop the best antenna from being the one each of us
like, no matter how it really works.

73 Tom

FAQ on WWW:               http://www.contesting.com/towertalkfaq.html
Submissions:              towertalk@contesting.com
Administrative requests:  towertalk-REQUEST@contesting.com
Problems:                 owner-towertalk@contesting.com
Search:                   http://www.contesting.com/km9p/search.htm