Rudy Severns, N6LF

PO Box 589, Cottage Grove, OR 97424;[email protected]

Experimental Determination of
Ground System Performance for
HF Verticals­
Part I
Test Setup and Instrumentation
This description of the test setup used by the author for a series of experiments sets
the stage for a series of articles describing his results.

HF verticals located on or near ground are

a perennial topic among amateurs. Over the
past several years this discussion has been
illuminated (and in some cases obscured!) by
the advent of really good modeling software
based on NEC (numerical electromagnetic
code). This has resulted in a vast literature
on antennas using the results of modeling.
However, these results are not without some
controversy. In particular the relative merits
of a large number of buried radials versus a
few elevated radials has been especially con-
tentious. What has been missing from the dis-
cussion are careful field measurements done
with good instrumentation and technique to
see if the NEC predictions actually hold up
in the real world. To address this problem I
performed a series of field experiments, over Figure 1— This drawing illustrates the traditional measurement scheme.
a period of a year, to examine how different
ground system arrangements affected the
behavior of a vertical antenna and to see if
field measurements on a real antenna would
correlate with NEC modeling. that in sparse (<10 radials) radial systems look at the effect of radial numbers on the
The results of these experiments will be lying close to ground, there can be a substan- characteristics of ¼ wavelength and several
presented in a series of QEX articles. There tial increase in ground loss when the radials shorter loaded antennas. Part 5 will take a
is no pretence that these experiments will are made much longer than 1⁄8 wavelength. look at the problems of ground systems for
answer all questions or even definitively This is a case of more copper = more loss, multiband verticals, where a range of 7 to
settle some of the arguments, but at least they which is not at all intuitive! Part 3 will com- 30 MHz must be accommodated. Finally in
should give us something to think about. pare verticals with a large number of ground Part 6, I will report on some experiments with
In Part 1, I will discuss the test range, test surface radials to verticals with four elevated a full size ¼ wavelength vertical on 160 m. In
instrumentation and test procedures used for radials. This part will directly address the addition, because this series will take many
all the experiments. Part 2, which is included elevated radial controversy. Part 3 will also months to be published, there will be lots of
in this issue of QEX, discusses an earlier and have comparisons between several different time for feedback. I plan to include some of
apparently overlooked prediction from NEC, elevated radial configurations. Part 4 will this in Part 6.

QEX – January/February 2009 21

 another transmission line.
Test Setup Amplitude measurements with a pro-
The physical layout of the test range, fessional VNA are typically displayed to
the instrumentation employed and the test 0.001 dB, but of course nothing else in the
procedures were all key elements in obtain- system is stable to that level. In practice I
ing reliable results. The following discus- found that measurements made over a short
sion provides descriptions of these elements period of time (2-3 hours) were repeatable to
which remained essentially constant for within 0.05 dB. That is more than adequate
the experiments. The majority of measure- for these experiments. A weakness of this
ments were done at 7.2 MHz although there measurement method is that as the separation
was some work at 160, 30, 20, 17, 15 and between the test antenna and the receiving
10 meters. The information given here is antenna is increased, the attenuation around
intended to provide information common to the transmission loop becomes quite large,
all the experiments. –40 to –60 dB. For instrumentation and a
Test Concept physical setup with a noise floor and stray
The traditional test procedure for these coupling below –110 dBm, this is accept-
kinds of measurements is well known. As able but it did limit the separation distance
shown in Figure 1, a test antenna is excited on 40 m to about 2.25 wavelengths for the
with a known power, and the resulting signal particular receiving antenna employed. This
is measured at a remote point. A change is is in the far field but not by much. Another
then made in the test antenna and the mea- limitation was that ± 0.05 dB repeatability
surement is repeated. The difference between was possible only when the antenna under
the two measurements is a measure of the test and the receive antennas were actually
effect of the change in the antenna and/or stable to that level. This usually meant that
ground system on performance. The signal measurements had to be made in early morn-
transmission to antenna 2 from the excita- ing when the test range was in the shade or
tion of antenna 1 (S21) will be proportional late in the day when things had reached ther-
to the radiation efficiency of the antenna. In mal equilibrium. It was very easy to detect
other words, S21 ~ input power × Rr / (Rr + a cloud passing over by the small changes
Rg) where Rr is the radiation resistance and due to temperature changes in the antennas. I
Rg is the ground loss. For our purposes we could readily detect the effect of the wind on
can assume that losses due to conductors are the vertical, causing it to move slightly. In the
small. Both Rr and Rg will vary as we change end the A-B comparison measurements were
the ground system but the final goal is to see probably within a few tenths of a dB but only
the effect on the transmitted signal.1 when I carefully attended to all the details.
The standard way to make these measure- This brings us to an important point. The
ments is to use a transmitter combined with purpose of the experiments was to determine
forward and reflected power meters to excite the effect of different ground system arrange-
the test antenna (antenna 1) with a known ments from their effect on S21. All the mea-
power. A calibrated receiver is connected to a surements were relative A-B comparisons. In
remote receiving antenna (antenna 2) to mea- Figure 2 — This diagram shows the vector other words, they were comparisons between
network analyzer approach for measuring
sure the resulting signal. In my initial tests antenna performance. two different configurations. There was no
I used both an HP3586C and an HP3585A
spectrum analyzer for the receiver. I wished
to measure the performance differences
between configurations to within 0.1 dB if
possible, and these instruments were capable
of that. However, the limiting factor turned
out to be my ability to measure the excita-
tion power; 0.1 dB corresponds to about 2%.
To make repeatable measurements to 0.1 dB
you would need to measure power to better
than 1%.
To get around that problem I decided to
use the instrumentation scheme illustrated in
Figure 2. I chose to make the measurements
with a vector network analyzer (VNA) in
the transmission mode (S21 is the response
at port 2 due to the excitation at port 1).
The transmission path was from the VNA
output port, out to the test antenna via a
transmission line, from there to the receive
antenna and back to the VNA input port via
Figure 3— A view of the test antenna area and test equipment shelter. The receiving antenna
Notes appear on page 25.
1
is at the far end of the pasture.

22 QEX – January/February 2009

 Figure 5 — Here is the test antenna base at ground level, with 64 radials.

Figure 4 — This photo shows a typical test

antenna and center post support.

attempt to measure absolute signal strengths

or radiation patterns. The separation distance
between the test antenna and the receiving
antenna was sufficient to place the receiving
antenna outside the reactive near field but the
groundwave was still significant. This was
not a problem for the type of measurements
being made. The presence of a metal pump
house and a travel trailer, both of which are
small in terms of a wavelength might have
had an impact on pattern measurements but
should not have affected the type of A-B
measurements being made in this series of
experiments.
Physical Arrangement
The test range was set up in a field as
shown in Figure 3, with an area for the test
antennas (including ground systems), a
remote receiving antenna (in the far distance)
and a small travel trailer to provide shelter for
the instrumentation. Figure 6 — The base plate is in position for elevated radials.
The eight poles, in an 80 foot diameter
circle around the test antenna, were used to
support elevated radials as needed. When of the base plate at ground level with 64 radi- ground. The radials lying on the ground in
more than eight elevated radials were needed, als attached is shown in Figure 5. Figure 6 were not present during elevated
a ½ inch Dacron line was stretched around The base plate is isolated from ground but radial tests!
the posts at the desired height and tightened there are three ground stakes (4 foot copper- As shown in Figures 5 and 6, a coaxial
with a turn-buckle. Each post has a backstay clad steel rods) close to the plate for those common mode choke (balun) was used to
to a buried deadman anchor so the radials tests where grounding is desired. The ground isolate the transmission line from the test
could be well tensioned. Radial heights on stakes have short pig-tail leads to connect to antenna. This was done for all measurements
each post were located using a laser level to the base plate when desired. whether or not ground stakes were engaged.
keep the radial fan flat around the circle. Figure 6 shows an example of the base The choke has an impedance of >3 kΩ at
In the center of the circle there is a support plate positioned for elevated radial tests. 7.2 MHz. For those tests in which the
post (PVC pipe) as shown in Figure 4, with The base plate, the radials and the entire test SteppIR vertical was employed, the balun
Dacron support lines attached to the top. This antenna are elevated by sliding them along that comes with that antenna was used in lieu
post is intended to hold the antenna under test the support pipe. This arrangement made it of the choke shown in the photos.
and allow it to move up and down to vary the very easy to change the height of the radi- The receiving antenna was a 3-turn dia-
height for elevated radial tests. An example als in small increments up to 4½ feet above mond loop with a diagonal dimension of

QEX – January/February 2009 23

24 inches, as shown in Figure 7. The loop factor for the change in cable loss with varia- on top! Critical for maximum accuracy! The
was resonant at 8.2 MHz. This loop was tions in SWR. common mode choke in the photo is undergo-
installed at the top of a 40 foot mast, as ing characterization for transmission loss and
shown in Figure 8. series impedance at 7.2 MHz. It turned out
Test Instrumentation
The distance from the base of the test however, that the impedance of the choke was
Feed point impedance, transmission gain
antenna to the receiving loop is a little over much greater than the 50 Ω reference imped-
(S21) and radial current measurements were
300 feet, about 2¼ wavelengths at 7.2 MHz. ance of the VNA. Above about 2 kΩ even
all made using a VNA. Two analyzers were
The elevation angle from the base of the test an HP VNA becomes inaccurate for a direct
available: an HP3577A with an HP35677A
vertical is about 8°. measurement. For choke measurements,
S-parameter test box and an N2PK analyzer
The coax from the VNA output port to the I used an HP4815 analyzer, which is well
with dual fast detectors. Figures 9 and 10 are
base of the test antenna was ½ inch Andrews suited for high-impedance measurements.
photos of these instruments.
Heliax with N connectors. The coax from After careful comparisons between the HP
Note the organic automatic heating unit
the receiving antenna back to the VNA was and N2PK VNAs, the N2PK was selected for
LMR400. Low loss coax was used because it
provided better shield attenuation to reduce
coupling and in the case of the heliax run-
ning out to the test antenna, the very low loss
removed the need for an additional correction

Figure 7 — This photo shows the

loop receiving antenna.

Figure 9 — HP3577A with an HP35677A S-parameter test box.

Figure 8 — Here is the receiving antenna Figure 10 — Here is my test bench, showing the N2PK VNA with the associated laptop
atop a 40 foot mast. N7MQ assisting! computer and HP calibration loads.

24 QEX – January/February 2009

 Figure 11 — This photo shows a typical
shielded current transformer.

most of the measurements because its perfor-

mance was very close to the HP and had the
advantage of direct readout to a computer,
which made data reduction much easier. The
N2PK VNA was also much lighter than the Figure 12 — Here is the test setup for a typical radial current measurement.
HP (70+ pounds!) and much more suitable
for field measurements.
On several occasions it was necessary to current division between radials and the rela-
measure the current division ratios between tive current distribution along a radial. Acknowledgement
the radials and in some cases, the relative This experimental work was inspired by
current distribution along a radial. To make Comments on test procedures the earlier work of Jerry Sevick and Arch
these measurements a set of shielded current A good physical setup and professional Doty.2, 3 Some of my experiments were
transformers, like the one shown in Figure instrumentation are a very good start, but to a repeat of their earlier work with more
11 were used. obtain reliable data great care must be exer- advanced instrumentation. I would also
To make a current measurement, a radial cised in using and calibrating this equipment. like to thank Mark Perrin, N7MQ and Paul
wire was passed through the current trans- For feed point impedance measurements, at Thompson, W8IEB for the many hours of
former, as shown in Figure 12. Current trans- the beginning and end of every test run an help they provided during the experiments.
formers were placed in the same location OSL (open, short, reference load) calibration Without their help, I would still be out in the
simultaneously on all the radials during a was performed with the calibration plane at field taking measurements!
measurement. The transformer being used to the test antenna feed point. At the beginning
sense current was terminated in 50 Ω by the and end of each test run a transmission cali- Notes
instrumentation, so all of the dormant current bration was also performed. 1
R. Severns, N6LF, “Radiation Resistance
transformers were also terminated in 50 Ω. In addition, before beginning a series Variation With Radial System Design,” Sep
This was done to compensate for any interac- of measurements a measurement of stray 2008. Available at: antennasbyn6lf.com.
tion introduced by the current transformer. At coupling and possible interference was per- 2
J. Sevick, W2FMI, The Short Vertical
the very least, the effect of the current trans- formed. The procedure was to disconnect the Antenna and Ground Radial, CQ
Communications, Inc, 2003. Jerry’s work
former would be the same on all radials. The feed line from the base of the test antenna, ter- also has appeared in a number of QST
active current transformer was isolated with minate the feed line with a 50 Ω load and then articles.
a choke as shown in Figure 12. measure the transmission gain of the entire 3
A. Doty, K8CFU, “Improving Vertical Antenna
system in this state. Throughout the series of Efficiency, A Study of Radial Wire Ground
Even with this degree of care, the current Systems,” CQ Magazine, April 1984, pp
measurements were still a bit tricky because of experiments, this transmission level was never 24-31. This article also has a very nice
the residual interaction between the cable from higher than –110 dBm and usually –115 dBm list of earlier references related to ground
the current transformer and nearby radials. In or lower, at 7.2 MHz. As a further check on systems for verticals.
some cases I actually used four identical cables results, most experiments were run several
in a symmetrical layout to try to minimize times to verify consistency and repeatability.
All of this was very time consuming but abso- Rudy Severns, N6LF, was first licensed as
imbalance due to this interaction. I believe the
WN7WAG in 1954 and has held an Extra class
resulting measurements were reasonable and lutely necessary to assure the best possible
license since 1959. He is a consultant in the
useful but not especially precise! measurements. I did not delude myself, how- design of power electronics, magnetic compo-
The relative value of the current was ever, into thinking the measurements were per- nents and power-conversion equipment. Rudy
determined by using the VNA in the trans- fect and cannot be improved on. I do believe holds a BSE degree from the University of
mission mode, measuring S21 for the loop the results make sense, fit well with NEC California at Los Angeles. He is the author of
from the VNA output port to the base of the modeling predictions, give useful insights into two books and over 80 technical papers. Rudy is
antenna, out along the radial to the current vertical antenna/ground system behavior, and an ARRL Life Member, and also an IEEE Fellow.
transformer and back to the VNA input port. potentially can be of practical help in optimiz-
This was a convenient way to measure the ing a given antenna installation.

QEX – January/February 2009 25