At the end of a string of posts with Pete, N4ZR, we find ourselves down in the
deep dodo of a stub's actual behavior vs model or conceptual device behavior.
The discussion originated around evaluating a lazy V, four vertical doublets
distributed around the tower, pulled out from the top of a tower to a
feedpoint, and back to the tower, with four coaxial feedlines from the doublet
centers back to the tower and the switch. One feedline fed at a time, the rest
passive and providing gain and front to back. Then add the issue of leaving the
unfed feedlines' shields connected to the switch vs. isolating them.
Any of the following may be flawed and review and comment is appreciated.
Conceptually, a stub is usually thought of as a lump device. Therein lie the
problems.
At a simple glance it would seem the case of parallel line stubs (450 ohm
ladder line, etc) that modeling the wires is roughly sufficient, noting
velocity factor, and there are no huge surprises, assuming close parallel
caveats are respected.
The overlooked aspects seem to occur most with coax stubs, where we are tempted
to treat them as a lumped constant device. Particularly, the EZNEC style
conductorless transmission lines do not provide for shield current that surely
exists with the balunless connections to the doublets in the lazy vee. Nor do I
know of a way EZNEC can model a coaxial gamma match, other than by substituting
the equivalent separate rod and lumped capacitor device.
The completely sufficient lumped state evaluation of an actual coaxial stub
does not occur unless we do work to obtain it: in the case of the lazy vee
placing a longitudinal current suppression device where the stub connects to
the doublet wire elements; coiling the stub, or placing it in some
configuration that significant induction of longitudinal current along the
shield does not occur; and for open stubs, not extending one or the other of
the conductors alone from the open end.
For the last, it's certainly not just a stub any more. It looks like the rod
part of a coaxial gamma match, minus the shorting bar. We model a coaxial gamma
match as a rod equal to the shield dimensions and any extensions, and a series
capacitor between the center conductor connection point and the rod. (That's
where the look like a capacitor reference I posted to N4ZR comes from.)
To tune a coaxial gamma match, one chooses the length of the rod the same as
the regular version. You set the position of the shorting bar the same way. You
adjust the amount of the center conductor shoved inside the rod to get the
capacitance needed.
When it's all over with one can yank the center conductor out and replace it
with a regular cap and get a match without changing the settings of the rod or
shorting bar. Or seen another way, the "stub" and shield extension from the
stub is replaced with the entire shield and a lumped capacitor to achieve the
same effect.
I'm not trying to make any grand statements about all stubs. The gamma match
"stub" IS a short one.
It would seem that the exact effect for any length stub, which might be more
than a 1/4 wavelength, is a complex issue that has to be worked into the model
somehow.
In the case of the common implementation of a lazy vee, the stub shield is
neither isolated from the doublet element nor placed in a guaranteed
no-current-induced configuration. At the very least the shield needs to be
added as a wire element. Further the current on the outside of the shield will
not be subject to the same velocity factor as the interior currents, and will
be somewhat out of phase with the current on the interior where they connect at
the end.
I suspect the EZNEC requirement for modeling is a very short small single
segment wire at the tower, with a feedpoint, a very large load (500K or such),
the transmission line from that wire to a smiliar small wire at the doublet
center. The model will need wire of shield diameter and physical length from
one end of the small wire in the doublet to one end of the feedpoint small
wire, and that same end of feed point wire attached to both the tower and the
like end of the other corresponding feedpoint wires.
To model the detached feedpoints, the connections to the tower and other
feedlines are removed, but nothing else is removed.
What I do not know is whether the connected version requires a capacitive load
across the short center wire to account for behavior like the coaxial gamma
match which is not provided in the modeling program's transmission line device.
One probably needs a two part shield conductor to account for the droop in the
feedline and to get the actual feedline shield length into the calculations.
The droop will cause the wire to be closer to one wire as it pulls away, and
therefore subject to some induction from the lower doublet half not balanced
out from the top.
Since a formula open stub evaluation assumes that the only emf approaching the
end is on the interior, and is strictly balanced, coax out in the open subject
to longitudinal current induction could not depend on modeling devices
depending on the balance.
Pete has some experiments going which induce reactance changes that I think are
significant, but he can post these himself.
Comments, wisdom, references, whatever?
73, Guy.
|