Is a Power Divider Necessary for a Phased Array?
Patrick Tatro
I’m putting together a 2 element phased array
for 160 meters. I’ve used ELNEC to model center
loaded elements 30 feet tall and spaced 30 feet
apart with very interesting results. I’m
developing an auto-tuner for each element to allow
maximum band coverage and plan to use lengths of
coax to create the phase shift.
I’m afraid that the whole thing wont work
because of power imbalance in the two elements.
Your comments and suggestions would be greatly
appreciated.
73’s
Pat N0WCG
patric...@stortek.com
Roy Lewallen
In article <557s5l$adg...@stortek.com>,
Power balance isn't the criterion for making a phased array work. In
general, what you want is the _currents_ to be in the right ratio (both
magnitude and phase-wise). Except for some special cases (such as two
elements in phase or 180 degrees out of phase), elements with equal
currents will have unequal powers. (In a properly designed 4-square
array with very low ground loss, one element actually feeds power _into_
the feed system. It gets this power by coupling to the other elements.)
See The ARRL Antenna Book, Chapter 8, for more explanation.
Roy Lewallen, W7EL
W8JI Tom
>
> I’m putting together a 2 element phased array
>for 160 meters. I’ve used ELNEC to model center
>loaded elements 30 feet tall and spaced 30 feet
>apart with very interesting results. I’m
>developing an auto-tuner for each element to allow
>maximum band coverage and plan to use lengths of
>coax to create the phase shift.
Remember the model won't include any losses, such as matching and ground
losses, unless you put them in. Any ground connection is considered
perfect.
Expect to have loss, not gain, with 30 feet of spacing and a 30 ft tall
element. Especially if the verticals are base loaded or tuned.
> I’m afraid that the whole thing wont work
>because of power imbalance in the two elements.
The elements have to have equal amounts of radiation, not power. If the
elements are identical equal radiation will happen with equal current IN
EACH ELEMENT, not equal power in the feedline to each element.
But before you go far, consider losses carefully. Without proper
modelling, you might invent a magical antenna that just won't live up to
expectations.
73 Tom
Roy Lewallen
In article <55ajcs$9...@newsbf02.news.aol.com>,
w8j...@aol.com (W8JI Tom) wrote:
>In article <557s5l$adg...@stortek.com>, patric...@stortek.com
>(Patrick Tatro) writes:
>
>>
>> I’m putting together a 2 element phased array
>>for 160 meters. I’ve used ELNEC to model center
>>loaded elements 30 feet tall and spaced 30 feet
>>apart with very interesting results. I’m
>>developing an auto-tuner for each element to allow
>>maximum band coverage and plan to use lengths of
>>coax to create the phase shift.
>
>Remember the model won't include any losses, such as matching and ground
>losses, unless you put them in. Any ground connection is considered
>perfect.
>
>Expect to have loss, not gain, with 30 feet of spacing and a 30 ft tall
>element. Especially if the verticals are base loaded or tuned.
Actually, the gain relative to a single element can in some cases be
greater when loss is greater.
The 90-degree phased, 90-degree spaced antenna is an interesting special
case. For this antenna, the gain relative to a single element will always
be 3 dB, regardless of the loss. This assumes that both elements are
identical and have the same amount of loss.
(Cf. Jasik's description of "coupled resistance" in chapter 20. This array
has a total "coupled resistance" of zero, so the overall efficiency is the
same as for a single element. Or see it for yourself with EZNEC or ELNEC.)
If you have two identical elements which are spaced far enough apart so the
fields can totally reinforce in some direction (see Fig. 12 in Chapter 8 of
the ARRL Antenna Book), and the two elements are driven with equal
magnitude currents, the gain relative to a single element will always
approach 3 dB relative to a single element as the loss per element gets
very large. This is because the gain is 3 dB in the absence of mutual
coupling, and when loss is large, the effect of mutual coupling is swamped
by the loss. So, the gain of a 2-element array with elements spaced 1/2
wavelength and fed in phase will decrease from 3.8 dB to 3 dB as loss goes
from zero to a high value. If they're fed out of phase, the gain will
increase from 2.3 dB to 3 dB as loss increases. (All gains relative to a
single element with the same loss.)
Roy Lewallen, W7EL
W8JI Tom
In article <55at7i$6...@nadine.teleport.com>, w7...@teleport.com (Roy
Lewallen) writes:
>Actually, the gain relative to a single element can in some cases be
>greater when loss is greater.
>
>The 90-degree phased, 90-degree spaced antenna is an interesting special
>case. For this antenna, the gain relative to a single element will always
>be 3 dB, regardless of the loss. This assumes that both elements are
>identical and have the same amount of loss.
Hi Roy,
I noticed he was planning on using 30 ft spacing, which is .055 wl on 160.
Unless I'm missing something, efficiency of both elements will drop
drastically at that spacing.
The array would be very valuable for receiving, even at that close
spacing, but not for transmitting unless the matching system and ground
system was very well constructed. My receiving array uses four elements
spaced .125 WL on 160.
But with my pair of full size verticals on 160 and a full complement of
radials, I kept the spacing almost 2/10 wl to minimize headaches from
mutual coupling.
In any event, a power divider is unnecessary.
73 Tom
Anthony Severdia
In <55d8jq$9...@newsbf02.news.aol.com> w8j...@aol.com (W8JI Tom)
writes:
Hi Tom-
OK, if not a Wilskenson power divider (or similar), then what steps
do you take to assure equal element currents? Yes, mutual coupling
does make the whole problem difficult but, it seems to me, that
changing the element spacing just sets up a new set of problematic
conditions with mutuals. This is not mentioned in this and your prior
posts.
-=Tony=- W6ANV
you
W8JI Tom
In article <55eenh$s...@sjx-ixn9.ix.netcom.com>,
anth...@ix.netcom.com(Anthony Severdia) writes:
>
> Hi Tom-
>
> OK, if not a Wilskenson power divider (or similar), then what steps
>do you take to assure equal element currents? Yes, mutual coupling
>does make the whole problem difficult but, it seems to me, that
>changing the element spacing just sets up a new set of problematic
>conditions with mutuals. This is not mentioned in this and your prior
>posts.
>
> -=Tony=- W6ANV
>
>
Hi Tony,
I use about 100 ft spacing.
Current is controlled by changing the impedance presented at the common
point (where the two lines are combined) by adjusting the T networks (they
also adjust phase slightly). The common point is then matched to 50 ohms.
This is a little tricky, because the networks need to be swapped like
this:
E ant------90 deg line-----e 1------T
net---===delay===180==cp---Tnet---2
W ant------90 deg line-----w
cp is common point
delay line is 75 ohm 20 degrees shorter than S with a 180 degree flip
lines to antennas are 50 ohms
The T's are adjusted for deep back nulls (listening over the air) with
minimum SWR at the CP feed (it's touched up with a L net at CP added after
initial adjustment).
I do MOST of the phase in the delay line, the rest and power division in
the T's. Whatever is connected to 2 is the front, whatever connected to 1
is the rear. I just swap the antenna feeders around at 1 and 2. This does
it all, and is usable about 100 kHz wide on 160 with 1/4 wl verts. About
30-40 dB f/b at optimum freq and direction, maybe 20dB or so typical.
I mix this antenna through a variable phase and level network with my
receiving arrays on RX to notch QRN or peak desired signals. A little more
complex than normal, but optimum and adjustable.
73 Tom
Anthony Severdia
In <55eo24$r...@newsbf02.news.aol.com> w8j...@aol.com (W8JI Tom)
writes:
Tom:
This all goes far beyond my question which hoped for a direct
answer in more fundamental terms. (Field practice is separate issue).
Can you rephrase all of this ignoring the side issues of phase
in directional reversal and "notches" by added means? Also, the
ascii depiction needs some clarification, I think.
Thanks much.
-=Tony=- W6ANV
W8JI Tom
> Hi Tom-
>
> OK, if not a Wilskenson power divider (or similar), then what steps
>do you take to assure equal element currents? Yes, mutual coupling
>does make the whole problem difficult but, it seems to me, that
>changing the element spacing just sets up a new set of problematic
>conditions with mutuals. This is not mentioned in this and your prior
>posts.
>
> -=Tony=- W6ANV
Let me try again.
I watch the null depth. That is the true test because no combination of
phase shift that will produce a deep null without nearly equal radiation
from each element. If the elements are perfectly symmetrical that means
equal current.
I do this with a delay line and impedance and phase shifting circuits. A
power divider is a waste of time, it gurantees nothing and allows no
adjustment.
With end fire, closer spacing increases loss, and wider spacing decreases
directivity. I split the difference.
73 Tom
Anthony Severdia
In <55ol2e$s...@newsbf02.news.aol.com> w8j...@aol.com (W8JI Tom)
writes:
Dammit! Why do you COMPLETELY avoid what I said and asked
in my previous follow-up post? By _neatly_ (or is it "cutsily"
obliterating the key questions of my previous post.
I will take time to cool-off, sir ...sir ...sir!
-=Tony=-
Roy Lewallen
I frequently see newsgroup postings out of order, and haven't seen the
original posting yet. But I saw a posting with some of the exchange
between Tony and Tom.
The Wilkinson Divider is a port-to-post isolation device. When two (or
more) loads are 50 ohms, equal powers are delivered to them. If one load
changes impedance, it gets less power, and the difference is delivered to
the resistor which is part of the Wilkinson Divider. The second load's
power stays unchanged. A Wilkinson Divider doesn't insure equal currents in
two elements, which is usually what you want. (It doesn't even insure equal
powers, for that matter.)
Two or more elements of an array seldom have the same impedance, even if
they exhibit the same impedances when fed individually, because of mutual
coupling. (Exceptions are an array of only two elements with the elements
fed in phase or 180 degrees out of phase.) The problem then is to deliver
equal currents to two elements of unequal impedance. There are ways to do
this. Some are described in Chapter 8 of the ARRL Antenna Book. A simple
method is in Vol. 2 of the ARRL Antenna Compendium, "The Simplest Phased
Array Feed System -- That Works". Forrest Gehrke, K2BT described a more
general method in a series of articles in Ham Radio magazine some years
ago. I believe that all these methods are summarized in John Devoldere's
(ON4UN) book Low-Band DXing. A book by Paul Lee, N6PL, Vertical Antenna
Handbook, describes the method used by broadcasters. Although it's complex
and can be tedious to adjust, it should result in the correct currents.
Roy Lewallen, W7EL
FORREST GEHRKE
Tony,
I think you are asking whether a Wilkenson power
divider may be used for driving a two element
vertical array to assure equal driving current.
I have not seen your original post and am going
by the succeeding posts.
I am guessing where you are coming from. You
probably are assuming that the two elements are
identical impedance and therefore you expect to
feed these elements with an equal power divider
to assure that they are driven with equal currents.
However, due to mutual coupling between these
two elements, and when fed with equal magnitude
currents but 90 degrees out of phase, those
elements do not look anything like their self
impedances.
With a good ground screen under each element and
assuming they are 1/4 wavelength high and 1/4
wavelength spaced the driving impedances of these
elements will approximate 21 - j 15 for the
reference element and 51 + j 15 for the element
being fed 90 degrees lagging, with equal magnitude
currents.
As you can see the two elements are far from
identical in driving impedance and consequently
a Wilkenson power divider isn't the right
prescription by a long shot. W7EL, Roy Lewallen
has described a simple method of feeding this
2 el in-line array. I believe you will find
it in the ARRL antenna manual or else in their
compendium. ON4UN's "Low Band DX-ing" is another
source for this information and this book is also
published by ARRL.
//
k2bt
* RM 1.3 02583 *
dick sander
>> Hi Tom-
>>
>> OK, if not a Wilskenson power divider (or similar), then what steps
>>do you take to assure equal element currents? Yes, mutual coupling
>>does make the whole problem difficult but, it seems to me, that
>>changing the element spacing just sets up a new set of problematic
>>conditions with mutuals. This is not mentioned in this and your prior
>>posts.
>>
>> -=Tony=- W6ANV
>
>Tom W8JI writes:
>
>Let me try again.
>
>I watch the null depth. That is the true test because no combination of
>phase shift that will produce a deep null without nearly equal radiation
>from each element. If the elements are perfectly symmetrical that means
>equal current.
>
>I do this with a delay line and impedance and phase shifting circuits. A
>power divider is a waste of time, it gurantees nothing and allows no
>adjustment.
>
>With end fire, closer spacing increases loss, and wider spacing decreases
>directivity. I split the difference.
>
>73 Tom
>
Roy W7EL writes:
>I frequently see newsgroup postings out of order, and haven't seen the
>original posting yet. But I saw a posting with some of the exchange
>between Tony and Tom.
>
>The Wilkinson Divider is a port-to-post isolation device. When two (or
>more) loads are 50 ohms, equal powers are delivered to them. If one load
>changes impedance, it gets less power, and the difference is delivered to
>the resistor which is part of the Wilkinson Divider. The second load's
>power stays unchanged. A Wilkinson Divider doesn't insure equal currents in
>two elements, which is usually what you want. (It doesn't even insure equal
>powers, for that matter.)
>
>snip....
>
>Roy Lewallen, W7EL
Hello Roy, Tom, Forrest and all:
I understand that a "purists" believes that the current-forced divider with an
90 degree L-network (hereafter called L-net) is the optimum way to feed two phased arrays.
But before I give up my simple broadband Toridal splitter with a 100 ohm
resistor tied across it, and using coax lengths for phase delay, I'd like to
ask a guru a couple of questions pertaining to my antennas.
I'm using two hytowers on 80m, 40m, 20m spaced 62 ft (1/4wl on 80m). Will the
same single L-net feed all three bands without having to change anything?
The hytower is resonant on 3.8 MHz. When I want to operate on 3.5 MHz CW I add
a bit of base loading. If I'm using an L-net will anything (BW, SWR) change
because the 1/4 wl coaxes feeding the antennas were cut for 3.8 MHz.
If I decide to use the hytower only on 80m, roughly what will the SWR appear
to be at the feed point of the L-networK? Will I need some additional matching
(another L-network) at that point?
If I decide to load my two hytowers on 160m for a while, will I need to do
something to the L-net that I built for 80m to make it work?
I see in the pattern plots that due to mutual coupling I loose some front-to-
back (F/B) using a Wilkinson. Is that loss of a deep null really that critical?
On the same pattern plots I don't see any loss of gain over a single antenna.
Am I correct to assume while using a Wilkinson that my ERP (enter rough pileups)
is the same an an L-net? Because winning pileups is the bottem line.
On the pattern plots shown in the ARRL Antenna Handbook, there are no plots for
anything other than 0 degree elevation. Is there any diminished F/B at 60 degrees?
What I'm really asking is what about a multiband system or short loaded verticals
that we little pistols use?
FYI: Using the two hytowers described above I typically see 2-3 S-units F/B on 80m
(90 deg endfire); 2 S-units between endfire/broadside on 40m; and various nulls
on 40m using my broadband splitter-combiner.
73, Dick -K5QY -These are my opinions; not those of my employer
Roy Lewallen
In article <55tgg3$8...@news01.aud.alcatel.com>,
san...@aud.alcatel.com (dick sander) wrote:
>Hello Roy, Tom, Forrest and all:
>
>I understand that a "purists" believes that the current-forced divider
with an
>90 degree L-network (hereafter called L-net) is the optimum way to feed
two phased arrays.
If that's the case, then I'm not a "purist". There are lots of ways to get
the desired currents into the array elements. I listed references to
several in my posting, and ON4UN's book summarizes several.
>But before I give up my simple broadband Toridal splitter with a 100 ohm
>resistor tied across it, and using coax lengths for phase delay, I'd like
to
>ask a guru a couple of questions pertaining to my antennas.
>
>I'm using two hytowers on 80m, 40m, 20m spaced 62 ft (1/4wl on 80m). Will
the
>same single L-net feed all three bands without having to change anything?
Nope. Neither will any other network. (By "feed", I assume that you mean it
will cause the correct current ratio in the elements. If your only goal is
a low SWR, networks containing a resistor are the best bet. Of course, some
fraction of the applied power -- maybe a large fraction -- will be
dissipated in the resistor.)
>The hytower is resonant on 3.8 MHz. When I want to operate on 3.5 MHz CW I
add
>a bit of base loading. If I'm using an L-net will anything (BW, SWR)
change
>because the 1/4 wl coaxes feeding the antennas were cut for 3.8 MHz.
Yes. Pattern and SWR will both change. The amount can be calculated for an
individual case. It's hard to make a general statement about the effect,
except that typically the gain will change only slightly but the null depth
can change quite a bit.
>If I decide to use the hytower only on 80m, roughly what will the SWR
appear
>to be at the feed point of the L-networK? Will I need some additional
matching
>(another L-network) at that point?
I don't know the mutual impedance between two hi-towers. You may be able to
model them with ELNEC, EZNEC, or a similar program and derive the
information from that. Then, you can find the impedance at the input of the
L network using the equations in Chapter 8 of the ARRL Antenna Book. Any
matching done on the transmitter side of the point where the element
feedlines are connected has no effect on the pattern.
>If I decide to load my two hytowers on 160m for a while, will I need to do
>something to the L-net that I built for 80m to make it work?
You betcha.
>I see in the pattern plots that due to mutual coupling I loose some
front-to-
>back (F/B) using a Wilkinson. Is that loss of a deep null really that
critical?
That's a question only you can answer. You might care, you might not.
Consider this about a Wilkinson divider: If you're not losing a significant
amount of power in the resistor, it's not doing anything and you can remove
it. If the resistor is doing anything, it's dissipating power. Whether or
not this is critical is also up to you.
>On the same pattern plots I don't see any loss of gain over a single
antenna.
What pattern plots are these?
>Am I correct to assume while using a Wilkinson that my ERP (enter rough
pileups)
>is the same an an L-net? Because winning pileups is the bottem line.
As I said, if the Wilkinson isn't dropping your signal level by dissipating
power, it isn't doing anything.
>On the pattern plots shown in the ARRL Antenna Handbook, there are no
plots for
>anything other than 0 degree elevation. Is there any diminished F/B at 60
degrees?
Generally, yes. You can see this easily with a modeling program, including
the free ELNEC demo.
>What I'm really asking is what about a multiband system or short loaded
verticals
>that we little pistols use?
I don't know of any way to build a multiband system which has a good
pattern on more than one band.
>FYI: Using the two hytowers described above I typically see 2-3 S-units
F/B on 80m
>(90 deg endfire); 2 S-units between endfire/broadside on 40m; and various
nulls
>on 40m using my broadband splitter-combiner.
If that's what you're after, go for it.
Roy Lewallen, W7EL
w8j...@aol.com
In article <55ttmq$m...@nadine.teleport.com>, w7...@teleport.com (Roy
Lewallen) writes:
>>What I'm really asking is what about a multiband system or short loaded
>verticals
>>that we little pistols use?
>
>I don't know of any way to build a multiband system which has a good
>pattern on more than one band.
>
>
Roy answered everything very well. I especially enjoyed his truism about
the resistor in the Wilkenson not getting hot or even being necessary when
the system is working, and getting hot and being required only when
something is out of whack.
Roy part of the splitter fantasia stems from the fact if 1/4 wl lines are
fed equal powers from the same impedance , the currents at the far end are
equal regardless of load. I never could figure out how that applies to
arrays with different reactaances at the load, since that changes the
electrical line length. But I've never looked at that system more than a
few minutes because I try to avoid resistors in an antenna system unless
they are in receiving antennas.
You can use a system (I use here) called cross fire phasing, but the
impedance problem in the elements still needs addressed. With cross-fire
feed the phasing is correct over octaves of BW, only element impedance
changes are the problem.
73 Tom
dick sander
Dick Sander, K5QY wrote:
>I understand that a "purist" believes that the current-forced divider with a
>90 degree L-network (hereafter called L-net) is the optimum way to feed two
>phased arrays.
>.....snip......
Roy Lewallen, W7EL responded:
>If that's the case, then I'm not a "purist". There are lots of ways to get
>the desired currents into the array elements. I listed references to
>several in my posting, and ON4UN's book summarizes several.
>
>I don't know of any way to build a multiband system which has a good
>pattern on more than one band.
>
>If that's what you're after, go for it.
>
>Roy Lewallen, W7EL
Tom, W8JI wrote:
>Roy answered everything very well. I especially enjoyed his truism about
>the resistor in the Wilkenson not getting hot or even being necessary when
>the system is working, and getting hot and being required only when
>something is out of whack.
>
>73 Tom
Roy and Tom thanks for the responses. I do have access to the references
listed. Your answers verified what I knew. After nearly 20 years with my
splitter-combiner using 1300 W output, I've never seen any burned or charred
spots on my 100 ohm, 50 W resistor.
I posted my questions because I felt you guys were giving that poor power
resistor more bad press than it deserves. I just wanted to present a multiband
application where a resistor does have a use.
It seems to me that it boils down to monoband phase arrays with proper phasing
feed systems, or a Toriodal Wilkinson for multiband verts such as a Hytower,
Butternut, MFJ, etc. The only publication that I'm aware of describing a toriodal splitter-combiner for a "multiband array" is in the Antenna Compendium Vol 2,
page 10.
Thanks for the discussion.
73, Dick -K5QY
w8j...@aol.com
In article <55vu0v$c...@news01.aud.alcatel.com>, san...@aud.alcatel.com
(dick sander) writes:
>I posted my questions because I felt you guys were giving that poor power
>resistor more bad press than it deserves. I just wanted to present a
>multiband
>application where a resistor does have a use.
>
>It seems to me that it boils down to monoband phase arrays with proper
>phasing
>feed systems, or a Toriodal Wilkinson for multiband verts such as a
Hytower,
>Butternut, MFJ, etc. The only publication that I'm aware of describing a
>toriodal splitter-combiner for a "multiband array" is in the Antenna
>Compendium Vol 2,
>page 10.
>
>Thanks for the discussion.
>
>73, Dick -K5QY
Hi Dick,
I assume were are talking about a two element end fire array, since that
was the original subject. If you play with some numbers on paper, you'll
see equal voltages at the common point of two 90 degree lines results in
equal currents in the load regardless of load resistance as long as the
loads aren't reactive.
I suspect someone took off with this fact, misapplied it in an article or
on the air by adapting it to phased arrays, and it became accepted as
fact.
I might be missing something, but I see no reason for the splitter. In
every case I can think of, it makes the problem of equal current at the
load worse, not better. So I'll take a risk and go one step further, I'll
say it use of a splitter makes the problem of equal element currents
worse, not better.
Here are the problems:
1.) We need equal voltages at the common point, not equal currents or
power. A power splitter is the last thing we want at that junction. We
need equal voltage, and that means a direct parallel connection.
2.) The lines only provide equal load current when they are 90 electrical
degrees long, that rules out reactive loads or any phase shift system
being inserted in any line.
The only application I can think of where the splitter would work, is
where the port impedances are exactly equal and line lengths are exactly
90 degrees or odd multiples of 90 degrees. The only case I can think of
where line impedances are equal in reactance and impedance is in a
bi-directional system with end-fire or broadside arrays, using either zero
or 180 degree phasing.
The problem isn't one of heat, it's just a system that has no effect when
it isn't needed and bad effects when it is.
73 Tom
w8j...@aol.com
As I look at the results (on
paper) with and without a splitter in a phased array, they actually made
systems
looked at worse instead of better.
Has anyone analyzed arrays in detail with and without them, or measured
them?
I am also puzzled by the use of the name "Wilkinsen" to describe what we
always referred to as a "magic T" in CATV product design. Am I missing
something here?
73 Tom
w8j...@aol.com
In article <8CC5336.02CF...@cencore.com>,
forrest...@cencore.com (FORREST GEHRKE) writes:
> think this got started back in 1978 when Dana Atchley
>published an article in QST using Wilkenson power
>splitters for a 4square array. Dana mistakenly
>overlooked the effects of mutual impedance between
>verticals. He assumed all verticals were at their self-Z.
>
>This has been picked up in other articles and books
>and like reflected power being lost power, will
>never die.
>//
> k2bt
Hi Forrest,
Yes, evey thing I've seen makes me agree with you. The closer I look, the
less merit in a power divider there is.
As a matter of fact, when I use 1/4 wl transmission lines and vary the
load, using a T or wilkensen actually makes current distribution and
radiated power LESS equal, not better.
One reason is the 1/4 wl lines between the loads and phase shift
network/common point need equal voltage at the source, not equal current.
A second problem is if the lines are 1/4 wl, adding any delay line changes
the length and impedance transformation of one line.
The whole scheme seems to make little sense.
73 Tom