K7ITM wrote:
. . .
But a better way to do a broadband vertical collinear is to feed
several dipoles, stacked end-to-end (with some gap from one to the
next), each fed with the same electrical length of feedline, with the
far ends of all the feedlines paralleled. If the gap from one dipole
to the next is enough that the mutual impedances among the dipoles are
all small, then each dipole will have current very nearly in phase
with the others and the radiation pattern will be perpendicular to the
axis of the dipoles. It's a messier feed arrangement, but it's much
better for keeping the antenna currents in phase along the whole
antenna across a relatively wide frequency range.

You can avoid the problem of different feedpoint impedances due to
mutual coupling by using lines of an odd number of quarter wavelengths
to feed the elements. If you use lines of those lengths all going back
to a common point, the currents in the elements will forced to be equal
in amplitude and phase regardless of differences in their feedpoint
impedances. There's more about this in Chapter 8 of the _ARRL Antenna
Book_. It's become known as the "current forcing" method.

Roy Lewallen, W7EL

On Feb 6, 6:11 pm, Roy Lewallen wrote:
K7ITM wrote:
. . .
But a better way to do a broadband vertical collinear is to feed
several dipoles, stacked end-to-end (with some gap from one to the
next), each fed with the same electrical length of feedline, with the
far ends of all the feedlines paralleled. If the gap from one dipole
to the next is enough that the mutual impedances among the dipoles are
all small, then each dipole will have current very nearly in phase
with the others and the radiation pattern will be perpendicular to the
axis of the dipoles. It's a messier feed arrangement, but it's much
better for keeping the antenna currents in phase along the whole
antenna across a relatively wide frequency range.

You can avoid the problem of different feedpoint impedances due to
mutual coupling by using lines of an odd number of quarter wavelengths
to feed the elements. If you use lines of those lengths all going back
to a common point, the currents in the elements will forced to be equal
in amplitude and phase regardless of differences in their feedpoint
impedances. There's more about this in Chapter 8 of the _ARRL Antenna
Book_. It's become known as the "current forcing" method.

Roy Lewallen, W7EL

Yes, I thought about mentioning this, except that in this case it
won't (or at least may not) work very well. I like to think that a
practical antenna of this sort is nicely built with nominally one wave
long doublets; it saves on feedpoints. Then, assuming the the
feedlines go perpendicular at least 1/4 wave away from the feedpoints,
that means the feedlines are at least 3/4 wave long, for just a two-
dipole antenna. Taking into account the fact that the velocity factor
in the line is likely going to be noticably less than 1, it's probably
5/4 wave minimum we're faced with, and more if there are going to be
more elements than four half-waves. But if Dave99 wants to cover
100MHz centered around 500MHz, or maybe even more, as I got from one
of his postings in this thread, and we make the lines 5/4 wave long on
500MHz, then 10% removed in frequency from that, they'll be 5/40 or
1/8 of a wave off from 5/4. If the lines were 11/4 wave long, a 10%
change in frequency would result in more than a quarter wave change in
electrical line length. At least this is an issue to be aware of. In
general, lines that are long in terms of number of wavelengths change
length by electrical quarter waves rather rapidly with changes in
frequency, and it's easy to forget about that till it bites you and
leaves a scar for you to remember.

But after all that, I don't think the mutual impedance thing is all
that much of a problem for vertically stacked antennas, if you provide
even a little space between them. My recollection from modeling this
sort of antenna (and fairly careful modeling of the coaxial collinear)
is that it's not much of an issue in a practical antenna. "YMMV," but
it's easy enough to model in Roy's kindly provided free version of
EZNEC, so long as you don't have to go to too many elements, and then
I think the licensed-for-a-fee version with way more capability than
you'll need for this is still a pretty economical solution for the
time it saves.

Cheers,
Tom

K7ITM wrote:

Yes, I thought about mentioning this, except that in this case it
won't (or at least may not) work very well. I like to think that a
practical antenna of this sort is nicely built with nominally one wave
long doublets; it saves on feedpoints. Then, assuming the the
feedlines go perpendicular at least 1/4 wave away from the feedpoints,
that means the feedlines are at least 3/4 wave long, for just a two-
dipole antenna. Taking into account the fact that the velocity factor
in the line is likely going to be noticably less than 1, it's probably
5/4 wave minimum we're faced with, and more if there are going to be
more elements than four half-waves. But if Dave99 wants to cover
100MHz centered around 500MHz, or maybe even more, as I got from one
of his postings in this thread, and we make the lines 5/4 wave long on
500MHz, then 10% removed in frequency from that, they'll be 5/40 or
1/8 of a wave off from 5/4. If the lines were 11/4 wave long, a 10%
change in frequency would result in more than a quarter wave change in
electrical line length. At least this is an issue to be aware of. In
general, lines that are long in terms of number of wavelengths change
length by electrical quarter waves rather rapidly with changes in
frequency, and it's easy to forget about that till it bites you and
leaves a scar for you to remember.

But after all that, I don't think the mutual impedance thing is all
that much of a problem for vertically stacked antennas, if you provide
even a little space between them. My recollection from modeling this
sort of antenna (and fairly careful modeling of the coaxial collinear)
is that it's not much of an issue in a practical antenna. "YMMV," but
it's easy enough to model in Roy's kindly provided free version of
EZNEC, so long as you don't have to go to too many elements, and then
I think the licensed-for-a-fee version with way more capability than
you'll need for this is still a pretty economical solution for the
time it saves.

Thanks for the kind words about EZNEC. I agree that the problems, if
any, can be identified and probably overcome by modeling, whether with
EZNEC or some other program. I honestly haven't done it for a group of
collinear dipoles, so don't know how much feedpoint impedance alteration
takes place due to coupling in that sort of array. I'll certainly defer
to your recollection. In any case, if the effect is significant at all,
it would only affect the top and bottom elements to any degree, and very
possibly not enough to have much of an impact on the overall pattern.
You're absolutely right about the potentially severe bandwidth reduction
due to using long feedlines. EZNEC and other programs allow you to
include the feedlines in the model, so you can see exactly what the
impact would be on both the pattern and feedpoint impedance, for
whatever lengths you choose.

Roy Lewallen, W7EL