On Mar 16, 12:21*am, Roy Lewallen wrote:
Hi Owen,

I suppose that R.W.P. King disagrees with the "common explanation."
He makes it quite clear that there is interaction of the antenna field
with the stub perpendicular to the axis of the antenna wire, and that
the coaxial stub does not interact in the same way and the antenna
performance is therefore different. *(Antennas chapter of Transmission
Lines, Antennas and Wave Guides, King, Mimno and Wing.) *This is why I
like using a feedline to guarantee the phasing. *It can be done by
driving collinear dipoles with equal lengths of transmission line, or
by using an arrangement like the "coaxial collinear," where the
radiating elements are outer conductors of coaxial transmission lines
used to insure that the multiple feedpoints are at least fed in-phase
voltages (and you have to consider that the currents are not exactly
in phase).

Cheers,
Tom

In most phased arrays, the objective is to get the fields from the
elements to be in some particular ratio. Driving them with currents in
that same ratio doesn't always accomplish the desired field ratio,
though, when elements have different current distributions as they often
do. (Seehttp://eznec.com/Amateur/Articles/Current_Dist.pdf.) The
difference between field ratio and feedpoint current ratio is
particularly great when base feeding half wave elements. As it turns
out, you'll often get better field ratios by feeding with voltages
having the desired magnitude ratio and phase difference than feeding
with properly ratioed currents, when dealing with end fed half wave
elements. The coaxial collinear requires a pretty delicate balance of
outer and inner velocity factors as well as the effects of mutual
coupling, particularly when there are more than a couple of elements. So
I suspect that the current distribution can either help or hinder
depending on how the factors are traded off. I wouldn't be surprised,
though, if ratioing the voltages rather than currents is actually helpful..

As an illustration, open the EZNEC example file Cardioid.EZ. Change the
number of segments to 10 per wire for better accuracy. (It can still be
run with the demo program.) Click FF Plot and note the nice cardioid
pattern. Then change the Z coordinates of End 2 of the two wires to 0.47
m to make them nearly anti-resonant, and click FF Plot again. The
pattern deterioration is due to the elements having different current
distributions. Finally, change the source types from I to V. This will
force the voltages, rather than currents, at the antenna bases to be in
the desired ratio. Run FF Plot again. You still won't have the nice
cardioid back, but it's quite an improvement over the pattern with
"correctly" ratioed base currents. The bottom line is that the element
currents are more closely related to the base voltages than the base
currents, when the elements are near anti-resonance (parallel, or half
wave, resonance).

Roy Lewallen, W7EL

Thanks for the clarifications, Roy. Indeed, with my last slightly
cryptic comment about considering that currents might not be in phase,
I was wanting to communicate that you always want to check the
currents on the elements to make sure they do what you want. That's
true no matter how you feed the antenna, though as you say the feed
you use may aid in insuring that the currents stay the way you want.

I'm a bit surprised about your comment about the coaxial (fed)
collinear requiring a "pretty delicate balance" between coax
propagation velocity and (presumably) radiating element geometry.
What I've found in my simulations is that I could change the coax vf,
keeping the elements a transmission-line half wave long so that the
feedpoints were all the same in-phase voltage, and the net gain of the
antenna for a given physical length was only slightly affected. I'd
typically see a couple of the elements in a ten element array with
considerably lower current magnitude, but the currents were nearly in-
phase on all the elements, and the pattern was always the desired
"flat pancake". On the other hand, I wasn't trying for any up or down
slope to the pattern, and I can see that things might change in that
case. With the propagation velocities I was using, between 0.66 and
about 0.9, and the element diameters I was using, I suppose the
elements were always shorter than resonance, and the self and mutual
impedances were not changing in any dramatic fashion.

Or, perhaps my model was all screwed up! ;-)

Cheers,
Tom

Owen Duffy wrote:

I think that is what I had done, but I used the same diameter top to
bottom.

Sorry, my mistake when looking at the source. Your model is just as I
described. I apologize for the error.


Here is a revised deck with different diameters:

CM
CE
GW 10 1 0 -2 2 0 -2 2.1 0.005
GW 1 15 0 0 0 0 0 5 0.015
GW 2 30 0 0 5 0 0 15 0.005
GE 1
GN 1
EK
EX 0 1 1 1 0
TL 10 1 2 1 50 5 1e+99 1e+99 0.0001
FR 0 0 0 0 15 0
EN

In the above, the lower conductor is three times the diameter of the
upper conductor. The TL is wired into the lowest segment of the upper
conductor. Again, I have shunted the TL with 10k R to represent loss in a
real TL.

This model does not show in phase currents in upper and lower parts of
the vertical.

I've been running your model without the loss, and I'm seeing currents
in the upper and lower wires which are nearly 180 degrees out of phase.

between the wire stub and the antenna which doesn't exist between the
ideal transmission line and the antenna, so performance is different.

For sure -- maximum gain is about 46 degrees above the horizon.

You might as well leave your source open circuited as to connect it to
the shorted end of the transmission line stub. The current into one

I don't think I did that.

You're right, you didn't. My mistake.

. . .

In playing with the model, I noticed something surprising -- length and
Z0 of the transmission line have very little effect on the pattern, even
over wide ranges (5 to 5000 ohm Z0, lengths from essentially zero to one
wavelength). In fact, try removing the transmission line altogether,
leaving the wires connected directly together and look at the pattern.
Then try changing one wire end slightly to break the connection between
them -- again, very little change in the pattern. The fact is that the
junction of the two wires is at a point of very little current, so you
can connect or disconnect them with almost no change. Likewise, you can
insert just about anything (of zero physical size), including an ideal
transmission line of any length, without any real effect. So the
transmission line stub doesn't really do anything significant at all.
What I don't understand yet is exactly why the wire stub does what it
does. It sure doesn't work like the simplified explanations imply.

Roy Lewallen, W7EL

Hi Roy,

Roy Lewallen wrote in
treetonline:

....

Thanks, all noted.

What I don't understand
yet is exactly why the wire stub does what it does. It sure doesn't
work like the simplified explanations imply.

Returning to my diagram a), below is an expansion of the detail at the
junction of the stub and vertical sections.



|
|
|
|
|
|
|
| A
B |
---------------------|



--------------------|
|
C |
| D
|
|
|
|
|
|
|
|

It strikes me that if we omit the stub all together, and leave a gap in
its place, we have two unconnected resonant elements, the top half wave,
and the bottom quarter wave with a driving source. The two elements are
field coupled to some extent, and currents will setup in each section out
of phase. NEC models support this, and I think they are correct in doing
so.

Returning now to a) with the stub connected and very close to resonance,
and with reference to the diagram above, for A, B, C and D very close to
the corners, I(A)=I(B) and I(C)=I(D).

If the desired outcome of using the stub is that the upper and lower
sections are in phase, then I(A)~=I(D). That implies common mode current
in the stub, so to cause I(A)~=I(D), the stub must have common mode
current (equal to (I(A)+I(D))/2 per conductor).

If that is true, then reduction of the physical stub to a pure
differential mode TL element is discarding part of what makes it "work".
That implies that replacement of the stub with a two terminal equivalent
impedance, eg by insertion of a load in an NEC segment, or insertion of
one port of a TL network in an NEC segment is an inadequate model.

Am I on the wrong track here?

Owen

On Mar 16, 2:33*pm, Owen Duffy wrote:
Hi Roy,

Roy Lewallen wrote ystreetonline:

...

Thanks, all noted.

What I don't understand
yet is exactly why the wire stub does what it does. It sure doesn't
work like the simplified explanations imply.

Returning to my diagram a), below is an expansion of the detail at the
junction of the stub and vertical sections.

* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *| A
* * * * * * * * * * * *B * * |
* * * * ---------------------|

* * * * *--------------------|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * *C * * |
* * * * * * * * * * * * * * *| D
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|

It strikes me that if we omit the stub all together, and leave a gap in
its place, we have two unconnected resonant elements, the top half wave,
and the bottom quarter wave with a driving source. The two elements are
field coupled to some extent, and currents will setup in each section out
of phase. NEC models support this, and I think they are correct in doing
so.

Returning now to a) with the stub connected and very close to resonance,
and with reference to the diagram above, for A, B, C and D very close to
the corners, I(A)=I(B) and I(C)=I(D).

If the desired outcome of using the stub is that the upper and lower
sections are in phase, then I(A)~=I(D). That implies common mode current
in the stub, so to cause I(A)~=I(D), the stub must have common mode
current (equal to (I(A)+I(D))/2 per conductor).

If that is true, then reduction of the physical stub to a pure
differential mode TL element is discarding part of what makes it "work".
That implies that replacement of the stub with a two terminal equivalent
impedance, eg by insertion of a load in an NEC segment, or insertion of
one port of a TL network in an NEC segment is an inadequate model.

Am I on the wrong track here?

Owen

For what it's worth, I think you're on exactly the right track, Owen.

Some things to ponder: does it make any significant difference if the
stub is, say, 2mm wires spaced 20mm apart or 1mm wires spaced 10mm
apart (that is, the same impedance line, but different physical size),
and does it make any significant difference if the wires are kept in a
plane that includes the antenna elements, or if they are twisted near
their attachment point so they lie in a plane perpendicular to the
antenna wire, or if they are twisted throughout their length? What if
they are coiled in a spiral out from the antenna wire, so their
shorted end lies much closer than a quarter wave from the axis of the
antenna? I don't have any answers to these questions; they just seem
like an interesting and reasonable extension of your original
question.

Cheers,
Tom

On Mar 16, 2:33*pm, Owen Duffy wrote:
Hi Roy,

Roy Lewallen wrote ystreetonline:

...

Thanks, all noted.

What I don't understand
yet is exactly why the wire stub does what it does. It sure doesn't
work like the simplified explanations imply.

Returning to my diagram a), below is an expansion of the detail at the
junction of the stub and vertical sections.

* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *| A
* * * * * * * * * * * *B * * |
* * * * ---------------------|

* * * * *--------------------|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * *C * * |
* * * * * * * * * * * * * * *| D
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * *|

It strikes me that if we omit the stub all together, and leave a gap in
its place, we have two unconnected resonant elements, the top half wave,
and the bottom quarter wave with a driving source. The two elements are
field coupled to some extent, and currents will setup in each section out
of phase. NEC models support this, and I think they are correct in doing
so.

Returning now to a) with the stub connected and very close to resonance,
and with reference to the diagram above, for A, B, C and D very close to
the corners, I(A)=I(B) and I(C)=I(D).

If the desired outcome of using the stub is that the upper and lower
sections are in phase, then I(A)~=I(D). That implies common mode current
in the stub, so to cause I(A)~=I(D), the stub must have common mode
current (equal to (I(A)+I(D))/2 per conductor).

If that is true, then reduction of the physical stub to a pure
differential mode TL element is discarding part of what makes it "work".
That implies that replacement of the stub with a two terminal equivalent
impedance, eg by insertion of a load in an NEC segment, or insertion of
one port of a TL network in an NEC segment is an inadequate model.

Am I on the wrong track here?

Owen

I'm sorry...perhaps I don't understand your notation. Don't you
expect that the current at A will be (rather roughly) out of phase
with the current at D? If I think about a collinear with three half-
wave elements end to end, and drive the center of the center element,
if it's going to act like I want, I'll have high current near the
middle of each element, and those three will be in-phase. Because of
the mutual impedances among the elements, things get a bit funny at
the ends. I suppose there is a fairly large voltage across the gap
between adjacent elements, and therefore there will be moderately high
current near those ends to account for the capacitive current in the
air between them. That's what I'm seeing in the EZNEC model I just
hacked, and it's as I'd expect. The currents near the ends of the
central element are considerably higher than the currents near the
open ends of the outer elements.

(Now to spend a few minutes playing with changing the length of the
stubs through resonance...)

Cheers,
Tom

Hi Tom,

K7ITM wrote in
:

....
I'm sorry...perhaps I don't understand your notation. Don't you

I am taking a convention that the sense of currents in segments is from
bottom to top. That means that I defined all segments in order from bottom
to top.

My notation ~= is to mean approximately equal.

Does that clarify things?

Cheers
Owen

On Mar 16, 11:39*pm, Owen Duffy wrote:
Hi Tom,

K7ITM wrote :

...

I'm sorry...perhaps I don't understand your notation. *Don't you

I am taking a convention that the sense of currents in segments is from
bottom to top. That means that I defined all segments in order from bottom
to top.

My notation ~= is to mean approximately equal.

Does that clarify things?

Cheers
Owen

Yes--and then if they were exactly equal, would that not imply only
transmission line current on the stub? Obviously, they are exactly
equal if you simply connect the ends of the elements together...but
that isn't what gets us to in-phase currents at the centers of each
element (in the case of the symmetrical 3 element design; or the base
current in the bottom quarter wave in phase with the center current in
the top half wave...), and (nearly) equal currents at those current
maxima. To the extent that the currents A and D in your diagram
differ, there is common-mode or "antenna" current on the stub.

Cheers,
Tom

K7ITM wrote in
:

....
Yes--and then if they were exactly equal, would that not imply only
transmission line current on the stub? Obviously, they are exactly

Thinking some more about it, my current thinking is that my analysis was
flawed. I was using the standing wave currents, when I should be using
the travelling wave components.

I suspect that when NEC models the conductor arrangement at my fig a), it
correctly accounts for propagation delay and the phase relationships
compute correctly.

If we replace the stub with a TL element, I suspect that NEC reduces the
TL to a two port network and loads a segment of the vertical with an
equivalent steady state impedance of the s/c stub network. If that is
done, the reduction to a lumped load means that there is zero delay to
travelling waves, and the computed currents (amplitude and phase) in the
vertical will be incorrect. This means that you cannot replace a resonant
stub with a high value of resistance, it doesn't work.

If that is the case, it suggests that NEC cannot model such phasing
schemes using TL elements.

Owen

Owen Duffy wrote:
Thinking some more about it, my current thinking is that my analysis was
flawed. I was using the standing wave currents, when I should be using
the travelling wave components.

That's exactly the flaw committed by w8ji and w7el when
they tried to measure the delay through a 75m loading
coil using standing wave current which doesn't appreciably
change phase through a loading coil or through the entire
90 degree length of a monopole. Using standing wave
current, w8ji measured a 3 nS delay through a 10 inch
long coil, a VF of 0.27.

http://www.w8ji.com/inductor_current_time_delay.htm

W7EL reported: "I found that the difference in current
between input and output of the inductor was 3.1% in
magnitude and with *no measurable phase shift*, despite
the short antenna... The result from the second test was
a current difference of 5.4%, again with *no measurable
phase shift*."

Of course, phase shift is not measurable when one is
using standing wave current with its almost unchanging
phase. EZNEC supports that assertion. Bench measurements
support that assertion.

When traveling waves are used to measure the delay through
a 75m loading coil, the correct delay through w8ji's 10
inch coil turns out to be about 26 nS (~37 degrees) at 4 MHz
with a more believable VF of 0.033.

http://www.w5dxp.com/current2.htm
--
73, Cecil http://www.w5dxp.com
"Government 'help' to business is just as disastrous as
government persecution..." Ayn Rand

Owen Duffy wrote:
K7ITM wrote in
:

...
Yes--and then if they were exactly equal, would that not imply only
transmission line current on the stub? Obviously, they are exactly

Thinking some more about it, my current thinking is that my analysis was
flawed. I was using the standing wave currents, when I should be using
the travelling wave components.

I suspect that when NEC models the conductor arrangement at my fig a), it
correctly accounts for propagation delay and the phase relationships
compute correctly.

If we replace the stub with a TL element, I suspect that NEC reduces the
TL to a two port network and loads a segment of the vertical with an
equivalent steady state impedance of the s/c stub network. If that is
done, the reduction to a lumped load means that there is zero delay to
travelling waves, and the computed currents (amplitude and phase) in the
vertical will be incorrect. This means that you cannot replace a resonant
stub with a high value of resistance, it doesn't work.

If that is the case, it suggests that NEC cannot model such phasing
schemes using TL elements.

Owen

Why would NEC reduce a TL two-port to a lumped load? Two-port
parameters can handle transmission line problems quite well without
the simplifying assumption that all components are of zero length.
73,
Tom Donaly, KA6RUH