Antennas/Antennas_CD-ROM_Walnut_Creek_September_1996.iso/articles/interact/wiregrid

Text File | 1996-06-30 | 17KB | 354 lines

ANTENNA-STRUCTURE INTERACTION
Modeling with MININEC
R. P. Haviland, W4MB



One of the problems of the real world is that antennas must be
mounted on some type of a structure, a boom, a mast, or a tower.
And often even larger structures are involved, an auto or an
airplane.

Over the years a number of rules of thumb have developed giving
good mounting practices. But these have always been suspect,
sometimes without reason, but other times as a result of flaws
such as poor performance in some direction, or an unusually high
SWR. Too often, a nagging uncertainity remained- is this the best
that can be done.

The advances of the past several years in antenna modeling by
computer has opened the way to look at these antenna to structure
interactions in considerable detail. This article is intended to
be an introduction to these techniques for Amateur type antenna
problems, using equipment, techniques and software now easily
available to individuals.

It is assumed that the software is one of the versions of MININEC
3 which are available (see ads in the Amateur magazines). Those
with change 12 are the best, since this change eliminated a
restriction on angle between wires. However, any after change 6
should be satisfactory.

There is a version of NEC available for the PC family. However,
it is relatively expensive, and available only to members of The
Applied Computational Electromagnetics Society. See the authors
article in thefirst issue of Communications Quarterly for Society
information.

CONCEPTS OF STRUCTURE MODELLING

In doing this work, it is convenient to consider that there are
two types of structures. The first can be called "antenna- like
structures". This includes many items commonly found close to the
antenna. Booms, towers and guys are of this class. Because their
dimensions are not greatly different than antenna dimensions,
they are easy to model, essentially with the same rules as used
for the antennas themselves.

There hasn't been a good name for the second class of structure.
Relatively large areas of metal are involved, usually as curved
surfaces, as found in the body of an auto or the skin of an
aircraft. The large areas mean that the standard antenna modeling
techniques of straight, thin wires can't be used directly.
Instead, the structure must be approximated by a model composed
of the thin, straight wires which the antenna models can handle.
The usual name for these are "wire- frame models". Examples are
now found in almost every issue of electronic and computer
journals, often in ads for drafting software. The pre- computer
name would have been "Tinker- Toy models" after the childs toy.

Here, it must always be remembered that the performance output of
a computer antenna model is only an approximation to the real
antenna performance. In fact, this is true for most antenna
calculations, because the equations representing conditions are
so complex that they can only be solved by approximation. The
approximations are even poorer when large areas are analyzed.
This does not say to throw out the model results, however. The
results are useful guides, to design and to operation. As the
saying goes, at least the starting point is in the right
ball-park, if not exactly at the right plate. Use the model
results for what they are- guides to the real situation, accurate
enough to give relative performance and a good starting point for
final adjustment.

ANTENNA-LIKE STRUCTURES

The class of structures made of long thin elements are easily
modelled using the rules standard for antennas. To review, the
rules are;

-Neglect undriven elements if less than about 0.1 wavelength
long:
-Use 4 segments per half-wave for an indication of performance, 8
for reasonable accuracy, 16 or more if high accuracy is
important:
-Check the geometry, especially the end points, to be certain the
elements are properly positioned.

A boom is treated exactly like a dipole element, except that it
normally is not driven but parasitic. Guys are also simple, being
just inclined long wires. One end of grounded guys must end at
the earth's surface, z-dimension of zero. Poor earth can be
simulated by introducing a loss, a load resistance at the
junction point.

Over most common frequencies, the tower is simple, just a
grounded vertical antenna. But the tower size-frequency relation
must be watched. For example, a common tower leg width is 12
inches, about 0.3 meter. At 14 Mhz, this represents .014
wavelength, well within the "wire size" limit for good accuracy.
However, at 144 Mhz, the tower width is about 1/8 wavelength, and
accuracy will be poor because the thin-wire approximation is not
valid.

This does not say that the results are worthless, however. In a
typical situation the tower currents will be low if the antenna
is symmetrically mounted on the tower. A typical value of maximum
tower current will be one percent of the antenna driving point
current. Since power varies as the square of current, the tower
radiation will be around one hundredth percent of the antenna
radiation, or 40 DB below the drive power. About the only effect
of the tower current will be to fill in the nulls somewhat.

Suppose the calculation of this current is off by a factor of
ten, so the actual current is 10 percent of the antenna drive.
Now the power radiated will be 20 DB down. Such an error will not
affect the main-lobe calculations at all, but may mean that the
secondary lobes and nulls are very much in error. In addition,
there may be some error in the driving point impedance. However,
the calculated results will still give an overall estimate of
performance.

There are several ways to check the importance of such stray
currents. One is to compare the pattern and drive resistance of
the antenna alone and with the tower present. A second is to
introduce a large resistance at the point of maximum stray
current, and make the same comparison. And still another is to
move the drive point from the antenna to the stray current point.
If the antenna-alone pattern appears to be badly degraded, try
another structure. Or consider introducing insulators, which is
the equivalent to the large resistance trial.

In this connection, it is no longer necessary to physically cut a
guy or even a tower to stop the current flow. Several ferrite
beads or even pieces of old TV H-oscillator or deflection ferrite
cores at the point of maximum current will reduce the current to
a negligible value. See the book by W2DU and the ARRL Antenna
Handbook for methods of estimating and measuring the effects.

If the stray currents on a thick tower do seem to be important,
the tower interaction should be analyzed as a wire-frame model,
see below.

EXAMPLE OF ANTENNA-LIKE STRUCTURE ANALYSIS

Suppose that you have been asked to advise a friend on the use of
his sailboat rigging as an antenna. The common way of doing this
is to insulate the backstay, and to use this as a sloping
vertical, fed against the ballast, motor and other metal hull
parts. Is this the best way to go?

For simplicity, assume that the boat is 12 meters LOA, 4 meters
beam, with a 10 meter high mast, about the dimensions of common
35 footers. Also for simplicity, assume that the mast is stepped
on the ballast, and that the stay and shroud chain-plates make a
straight line with their wires, effectively connecting to the
earth at water level. The spreaders are at 5 meters, and are
non-conducting. Then the ship model is as shown in the two views
of Fig. 1. Assume that the shroud/stay diameter is .01 meter, and
the mast diameter is 0.2 meters. Use 16 segments for the longer
elements, and 8 for the shorter.

The solution speed will depend on the computer. 15-20 minutes per
frequency will be typical. Making a paper copy of results will
add to the time, but is worthwhile for future reference.

Assuming that the feed is at the base of the backstay, with other
stay/shroud/mast connections intact, the following is found:
                      Drive, Ohms        Radiation   Ratios, DB
     Band      Resistance   Reactance     F/Back     F/Side
     160          0.4         260           0         6.6
      80         80          1240         -.5         1.5
      40        335          -450        -1.6         -.8
      20         30            80         3.3         6.5
      15        170          -270        10.7         0.6

With a parallel resonance at 5.24 Mhz, and a series resonance at
13.28 Mhz. With a good antenna match box, this will be a good
antenna from 80 to 15 meters, and a usable but inefficient one on
160. There is some signal variation as heading changes, but this
is less than the normal fading range.

Additional runs can be made with the feed at other locations, and
with various insulation trials, at the top and/or bottom of
shrouds and stays. Other antennas can be introduced, for example,
the often used "Hustler Whip" on the quarter. Since this is an
example rather than a real problem, these are not explored here.

LARGE-SURFACE STRUCTURES

To show the approach to using thin-wire antenna analysis in
modelling surfaces, assume that an antenna is to be mounted on a
square metal plate, a ground plane. The plate can be represented
by a grid of crossed wires, two forms being shown in Fig. 2.

One limiting condition shows up if it is necessary to mount the
antenna at the center of the plate. Since there is no wire
junction at the center on Fig. 1A, there is no "ground" to feed
the antenna against. This point does exist in Fig. 1B. It is
evident that the planning of the wire grids must include antenna
location as a factor.

The usual "rule of thumb" is that the wire surface area should
equal the area of the surface it is to represent. But there are
limits to this. One is that the wire radius must be small
compared to a wavelength: a radius greater than 1/30 wavelength
is likely to give errors. Also, the segmentation of wires is
important: a segment should not be longer than 1/5 wavelength.

The ususal arrangement is a square mesh, i.e., four 90 degree
angles. However, there are situations where other angles are
necessary to model the surface geometry, and even places where
different angles give better results. For example, the flat
plates of Fig. 2 can also be represented by the radial wires of
Fig. 3. These give at least as good a representation as the
square mesh, and have the merit of requiring fewer wires and
therefore less computation time. Probably the best rule is to use
the square mesh for the general body, and radial connections at
and close to the antenna mounting point.

One problem with the models of Figs. 2 and 3 is that the current
is necessarily the same on the top and bottom of the plate. This
may not be accurate, for example with a ground-plane antenna
mounted atop a mast. In such situations two plates separated by a
small distance are necessary, as shown in cross-section in Fig.
4.

It is evident from this that there is a considerable amount of
art in the design of the model for analysis. Also, any attempt at
accurate modelling is likely to end up with a large number of
wires, which means a large computer and long run times. In
Amateur use with no more than a PC available, models will have to
be crude, and run times long.

It is very easy to make an error in laying out and entering the
model data. Use a consistent approach to numbering wires, and use
symmetry to help keep the numbers and dimensions correct. A
program which assembles antennas from wires or elements is a big
help, especially if it will allow mirror-imaging the elements, an
easy way to get symmetry. In a complex model, check andrecheck is
the procedure.

EXAMPLE ANALYSIS- A CAR ANTENNA

Suppose there is a proposal to use a diplexer with the
factory-mounted whip antenna af a typical small two door auto.
The basic questions are, what is the drive characteristic, and
what kind of a pattern is obtained.

Figs. 5, 6, and 7 show three views of a wire-frame model of such
a car. Note that the top is supported at three points along the
sides, with no other top-body connection, to represent windows.
The top and body sections do have a wire along the center-line,
and the body does have three side to side side straps across the
bottom.

Despite it's relative crudity, the model approaches the limit of
typical PCs. Only one current pulse per wire segment can be used
with standard MININEC.

Even with this crude model and the limits of the computer, the
results are worthwhile and interesting. At 146 mhz, the factory
mounted 30 inch whip on the right front fender at the windshield
front shows a drive impedance of 95 -j18 ohms, an easy match. The
pattern is surprisingly good. As seen by the car driver, the
front-back ratio to the horizon is 0.8 DB. The right side
radiation is 0.6 DB below the forward, and the left side 0.9 DB
lower. The pattern is essentially circular. Diplexer design
should be relatively easy.

The built-in antenna could be used at 28.5 Mhz. As with all short
antennas, the drive impedance is poor, 2.4 ohms resistive and
-415 ohm reactive. The pattern remains quite good, with a F/B
ratio of -0.9 DB. The front-right ratio is -2.2 DB and the front
left is -3 DB. The poorest radiation is to the front.

Having the basic car model, it is easy to move the antenna
location and to change length. The absolute values calculated
from this coarse model are likely to be in error. But the
relative change will be better, and it is easy to see if a
"relatively best" location has been found. (It will take some
time to run the range of possibilities, however.)

More wires and therefore better accuracy would be possible with a
late model computer with more memory, provided the arrays used in
the program are redimensioned. Compiling the program (as in Quick
Basic) is a good time saver. If much work is to be done, a
numeric co-processor saves time.

DATA ON MODELLING

There is considerable data on the relation between model design
and accuracy in the paper, Verifying Wire-Grid Model Integrity
with Program "Check", C.W.Trueman and S.J.Kubina, Applied
Computational Electromagnetics (ACES) Journal,V5N2, Winter, 1990.
This includes a table of warning/error conditions, and many
sketches showing problem situations.

The same issue includes Selecting Wire Radius for Grid/Mesh
Models, by L.A.Oyckanmi and J. Watkins. The previous issue
includes Modeling Electrically Small, Thin Surfaces with Wire
Grids, by T.H. Hubing and J.F. Kauffman.

Wire Grid modeling of Surfaces, by A.C. Ludwig, IEEE Trans.
Antennas and Propagation, vol. AP-35, Sept. 1987 will also be
helpful. The references to these will lead to other papers.




Captions


Fig. 1.

RF model of a typical sailboat. Only key metallic elements are
shown, keel, engine and chain plates are assumed to be at water
level or below. a) View from port or starboard. b) View from bow
or stern.


Fig. 2.

Wire-grid models of a square ground-plane plate. a) A coarse
grid, which requires large conductor size, and limits antenna
locations. However, it conserves computer memory, and requires
less time for solution. b) Fine grid, suitable for use at higher
frequencies, with less restriction on antenna location. Wire size
can be one-half that of a) for the same accuracy. Grids are shown
as squares, the prefered arrangement, but may be 3, 5 or more
sides, with angles other than 90 degrees.

Fig. 3.

An alternate to an equal-sided grid, especially useful at the
antenna mounting location. Such radials may be used as diagionals
to ensure good distribution of current and better accuracy.
However, such refinements increase memory demand and run time.

Fig. 4.

Accurate models sometimes require a closed surface, to allow currents on the top and bottom to be independent. The mast mounted ground plane is an example. The ground plane may be a grid as in Fig. 2, or radial as in Fig. 3.

Fig 5.

Modeling flat or curved surfaces requires many wire sections. This side view of a typical small auto ignores many features, such as wheels and bumpers. The relatively coarse grid is adequate for reasonable accuracy at 28 MHz, but the number of wires could be increased for better accuracy at 144 MHz.

Fig. 6.

Top view of the auto of Fig. 5. In numbering wires for computer entry, time can be saved by making use of the symmetry of the structure. It is also easier to avoid errors if the wires are numbered in mirror-image pairs.

Fig. 7.

Front view of the auto of Fig. 5. This and additional views are not usually necessary to see all of the model wires, but are helpful in avoiding numbering, dimension and entry errors. Some CAD programs produce this third view automaticlly, and also output a table of line (wire) end points. If much work is to be done, a small program can be written to transfer the table to RF analysis format.