Common Mode Radiation of E-Field antennas
Updated 20170827
I’m convinced there's something useful in the concept of E-field antennas. These use a step-up transformer to feed a plate or cylinder of large area, and give unexpectedly good performance. This page lays out my thoughts with some measurements.

The EH antenna and its forerunner Crossed-Field antenna are often dogged by accusations the feed-lines radiate, and not the antenna itself. Prof. Maurice Hately (callsign G3HAT, originator of crossed field antenna) went to his grave arguing the point. The
EH antenna and more recent Poynting vector antenna also fail to acknowledge common mode effects. On this page I tackle the issue head-on.

After making several EH antennas and many hours tapping off coils, I never measured a reduction to near zero, of common mode (CM) current. My measurement instrument is a simple transformer current meter.

The antenna I'm testing has ancestry to the EH antenna without the lower cylinder and uses a plate for the high voltage element. It has a step-up transformer feeding a plate. The lower plate (or cylinder) of an EH antenna can be removed with little effect on the radiated signal, and this is also proved here. This page looks at the feed line question, and explores some other performance issues. A photo of AUT1 (antenna under test 1).
E-field antenna

Note the Isotron and MicroVert antennas are not the same as the antennas under discussion here. They’re just loading coils with fat cylinders or plates, no step-up transformer involved.
A photo of AUT2, constructed according to instructions in the
PVA Book.
EH 2017-1

Construction notes

The antennas are made of commonly available materials. The coax is RG58; the ferrites FT140-61; the “paddle” is perforated 1mm thick aluminium; the coil former is 50mm plain PVC pipe. The only unusual item is the coil winding, which is made of 4mm wide copper strip. This strip is available from specialised metalwork suppliers. This photo shows the antenna at 120cm long (compare my foot!), the large coil used for tests without ground-grid/plane, small coil with it.
E-field ant

The advantage of copper strip is it minimises inter-turn capacitance, for lower losses and higher voltage generation. Also, the strip can be easily soldered to for reliable contact with cables without deforming.

Aluminium mesh was laid down for some of these tests. The mesh is termed a “ground-grid” here, as some may dislike the term ground-plane. The term ground-plane is used loosely in amateur circles. The ground-grid is 2.4m x 0.5m for tests on this page, see photos further down.

For tests with the ground-grid connected to the top of the common mode choke (i.e., the “hot side”) I had to use a smaller coil to get a good match. Tapping off a coil which is too wide a diameter is a problem, because the wires connecting the end of the coax have stray inductance. The ground-grid connected to the top of the common mode choke needs much less inductance, leading to a smaller diameter coil to maintain the resonant frequency.

The ground-grid is made of aluminium mesh. This is sold for craft modellers to use as a base for papier-mâché or the like. A 2m x 0.6m piece is used as the ground-grid. The mesh tears easily, so needs re-enforcement with at least some metal bars to keep it in place.

Accuracy of measurement
Comparing antennas is more difficult than you expect. There are many variables such as:
  1. Near field effects, measurements should be in far field
  2. Far field zone is many metres away for HF antennas
  3. Signal resolution of the measuring receiver (ideally 0.1dB)
  4. Frequency dependency of the receiver itself and its antenna
  5. Polarisation effects
  6. Effects of weather conditions, moisture on Rx or Tx antenna
  7. Propagation changing over time
  8. Ground effects
  9. Directional effects
  10. Background noise/signals adding to received amplitude
  11. Variations in transmitter output from temperature
  12. Moving objects close to the receiver or transmitter antenna
  13. SWR mismatch as either antenna is adjusted
  14. Effects of the above on any reference antenna when compared to the AUT (antenna under test)
  15. Take-off angle
  16. Losses of common mode current chokes
  17. Directional effects (azimuth)
There will be other factors I haven't thought of. Trying to control all these is in my opinion impossible for those with limited means. Advancing technology does help. For point 13, network analysers have substantially reduced in price. There is no need to use £100,000+ of test equipment for these tests.

For peer reviewed science, calibrated test equipment would be necessary. But I’m not trying to explain physics of how the antenna radiates, but doing a practical experiment.

A far field receiver with accurate relative signal readings, plus instant feedback to the transmitter solves several measurement problems. It allows instant results from adjustments. Points 1,2,3,5,6,7,9,10,13 can be eliminated or compensated for with this method. For example, near immediate comparisons of adjustments basically eliminates changes in propagation which happen quite slowly. Also, background signals (10) can be avoided by moving a few kHz.

I have a transmitter with a test antenna, and an online receiver a few miles away. The transmitter antenna being the one under test and adjusted. The receiving one is a 35m long inverted-V, or the metamaterial-dipole described elsewhere on this site. The online receiver used was the
KiwiSDR.

The major factors not solved by the VNA and online receiver are points 5,9,11,13,14. Polarisation can be addressed by a rotatable receive antenna. Variations in Tx power can be monitored by a power meter. Take-off angle at a particular test site is impossible to control, but influenced by height and grounding.

All that can be done to reduce objects near either antenna is to site the Rx one where it is less likely to be disturbed. For the Tx, large objects near the antenna (e.g. bodies, ladders) should be moved away or into the same position during successive tests. The effect of such objects can be seen immediately by readings of the online receiver.

Azimuth directionality is probably not a big uncertainty, because small antennas are inherently omni-directional, “non-beam” by nature.

Where comparative signal plots are shown on this page, they were taken within an hour of each other except where noted. This measure, in addition to testing in dry conditions only, mitigates any propagation influences. Testing was done in mid-summer, and not during any major sporadic-E events. The receiving antenna was not substantially altered during the test period.

If you think my attempts at reducing measurement errors are insufficient, be free to add 3dB uncertainty to most numbers on this page.
With these hopefully more rigorous methods, perhaps some objections about previous measurements of E-field antennas can be dispelled or reduced. Such measurements are found in this book.

Antennas at receive side

Frequency used approximately (+/- QRM) 28MHz. Inverted-V dipole, total length 35m, 11m high at centre, approx 1m high at ends. Situated at edge of open arable field. Photo follows.
Inverted-V

“Meta-material” mini-dipole. Situated in loft at same site as the inverted-V. Length 1.5m, width 0.75m. Efficiency similar to standard dipole, but narrow bandwidth. Small size makes it easy to rotate for polarisation testing.
MMI dipole

Common mode chokes
To control how much feed-line radiates, or rather define where its "cut off”, can be done with a high impedance ferrite choke. Also called a Guanella choke or balun, this is familiar to antenna designers the world over. In tests near the 28MHz frequency, I used 4 turns of coaxial cable through five FT140-61 toroids. There is plenty of information on construction of baluns elsewhere on the Internet, not to repeat it here. The balun of AUT1 sitting on the roof.
Coax choke

The balun of AUT2
EH_Balun
Either balun can be "shuffled" along the coax cable by pushing the turns through. By this method the length of radiating cable below the antenna can be varied. To verify the balun does stop common mode current, a simple clip-over current meter was used.
CM meter

Above the balun, the meter slams over to full scale indicating high common mode current. Below the balun only a small indication is produced. To back up this result, the network analyser reading swings sharply if the cable above the balun is touched. The reading changes a tiny amount, if at all, when the cable below the balun is touched.

With an effective balun, and a way to move it along the cable, a way of defining the length of radiating cable is achieved.

As side note, air-cored baluns/chokes are difficult in this application. They have too narrow a bandwidth, and are easily detuned by nearby objects. The result is their CM attenuation is poor, and the antenna is difficult to tune. An air-core choke only works effectively at a current minima, such as a quarter wave from the AUT. Having quarter wave then lays the experiment wide open to radiation by CM current. A possible future experiment is to use a tuned trap and not a ferrite choke.

Feed-line length reduction (AUT1)
It was decided to begin with 1.5m of cable and antenna length of 1.0m. The frequency used for all these tests was near 28.0MHz (λ/4 = 2.7m). So the starting point comparison was for a total length close to a quarter wave. The KiwiSDR S-meter extension was used to obtain a trace, from the inverted-V
Screen Shot 2017-07-01 at 12.20.00

The method was to take measurements within a few minutes of each other to minimise weather effects. It was found the rotation of the paddle and position of the coax on the roof made no difference to the signal. Reducing the cable to 1.0m:
Screen Shot 2017-07-02 at 19.49.41

The first plot omits the -90 cal line but was the same level by inspection. Unfortunately the S-meter extension doesn’t give a numerical readout or 1dB calibration lines. Then the paddle was moved down to 30cm and the cable reduced to 0.6m.
Screen Shot 2017-07-04 at 21.31.08
The same trace was produced. The total length was now 130cm. To compensate for length reduction the coil tap off positions had to change. Next at 100cm, which is 18% of a dipole length.
Screen Shot 2017-07-05 at 19.04.38
This at last produces a definite reduction in signal strength. There seems to be some kind of threshold point at 0.1λ total length (0.05λ coax length), and when the feed-line is shorter than the antenna itself. Scaling up for lower bands, a 1.0m cable on 20m may be the same threshold, or 2.0m cable on 40m.

Effects of ground-grid (AUT1)
It was found from previous tests (and textbooks) the local ground can have a large effect on antenna performance. A short helical whip outperformed a full-length half-wave horizontal antenna in a field test. This was put down to:
  1. A car body having lower ground losses than actual dirt, compensating for the low efficiency of the short whip
  2. Possible take-off angle reduction with the car body ground-plane/reflector
The transmitter site is on a slate tile roof, with a layer of metal foil covered plasterboard underneath. A photo of the site from 2015:
TLC and roof
The roof window allows easy access to modify the antenna, and the KiwiSDR provides instant signal readings from 3 miles away. An aluminium mesh was placed on the roof. The mesh was also connected to the roof tile bracket and roof window frame.

Firstly I connected the coax outer 40mm below the balun to the mesh. Aside from a very small downward shift in frequency, the performance was unchanged. There was no measurable increase in far field signal strength. But, this test does prove again there is very little common mode current getting through the balun. This plot shows a 2:1 SWR bandwidth of 4%, taken with the ground-grid connected, before the coil tap was moved to shift back to 28.0MHz. There is little difference in bandwidth without the ground-grid.

SWR with gnd below choke

Then the coax just above the balun was connected to the ground-grid. The effect was substantial, with a rebuild required to retune. I had to change the coil for one with a smaller diameter to get back to a low SWR at 28MHz, as explained above in the construction paragraph. The following two signal readings are with the ground-grid connected below the balun (cold side) and then above the balun (hot side). It should be noted these plots were done on different days, though the weather conditions were similar, and from the inverted-V.
Screen Shot 2017-07-08 at 19.06.01

Screen Shot 2017-07-24 at 19.21.01
A substantial increase of approximately 9dB is made by connecting the ground-grid to the hot side. But the bandwidth is similar.
Bandwidth with GP

On the possibility of the ground-grid/plane radiating, it can be disproved. By putting my (lossy) hand on the mesh, little reduction was noted in the transmitted signal. So the ground-grid is not acting purely as one half of an unbalanced “dipole”. Additionally, the ground-grid is arranged sloping at 30degrees off horizontal, so should give tilted polarisation. Yet the antenna gives vertical polarisation, as shown next.

Polarisation Testing

The E-field antenna with ground-grid was tested using the KiwiSDR S-meter extension for vertical polarisation using the meta-aerial min dipole.
Metamaterial vertical
Then for horizontal. 96dBm - 72dBm = 24dB, so the antenna is strongly vertically polarised.
Metamaterial horzontal

“Star” EH Antenna Retest (AUT2)
Moving to the second AUT, there are three tests to do:
  • Signal strength vs. AUT1
  • Polarisation measurement
  • Lower cylinder removal
A picture of “AUT2”.
EH_2017

In
the PVA Book the recommendation is to run the wires from the step-up coil to the cylinders outside the cylinders, which is supposed to be better than running them inside, as was done in my previous builds. Changing to AUT2 gives the field strength result to the inverted-V:
EH to inverted-V 2017-08-10 at 21.46.09

Now to the MMI-dipole vertical
EH MMI-dipole vertical 2017-08-05 at 16.38.44
…and the MMI-dipole horizontal
EH MMI-dipole horizontal 2017-08-05 at 16.22.21

Disconnecting the lower cylinder made the resonance increase from 28.0 to 31.6MHz. To get back to 28.0MHz, the radiating coax section was lengthened by about 1.0m, and an extra turn put on the coil. The result from the inverted-V.
EH cylinder removed to inv-V 27.801MHz 2017-08-13 at 14.27.55

The signal produced was within 1dB of AUT2 with the lower cylinder. For the MMI dipole vertical, the signal went up by about 1dB.
Pasted Graphic

The 2:1 VSWR bandwidth of AUT2 was about half that of AUT1, as shown next.
EH 2017 SWR

As a side observation, the sky noise at the transmitting side (which is with a transceiver) was markedly lower with AUT2 than AUT1.

Conclusions

The active length of the whole antenna was reduced from 2.5m to 1.5m with no measurable reduction in effectiveness. This is 60%, which should have a large effect according to the conventional theory of cable radiation. Reducing to 0.1λ (0.11m) for the whole antenna did start to reduce effectiveness.

The 60% length reduction of the total antenna is a definitive test. Strong common mode current on the feed line suggests there is a contribution from it, but reducing from 150cm to 60cm (2.5x) with no effect indicates the opposite. So,
I conclude the cable is not contributing much radiation, but needs to be equal or longer than the antenna itself for efficient operation. It could be that the coil losses become dominant, or the length is reduced to some fundamental limit.

AUT2 performed 1-2dB better than AUT1 on the inverted-V, but worse to the vertical. The top of AUT2 was 0.5m higher than AUT1, and it’s cable was the same as the starting length with AUT1. AUT2 had about half the 2:1 SWR bandwidth of AUT1. Removing the lower cylinder of AUT2 produced about 1dB drop in signal strength. So, the lower cylinder did make minor improvement, but nowhere near so much as the ground-grid.

The EH antenna (AUT2) is reported as having mixed, or elliptical polarisation. A vertical bias of about 13dB was indicated, not so much as AUT1. But, if it was mixed polarisation the horizontal and vertical results should be similar, and they are not. With the lower cylinder disconnection making little difference to signal strength, I conclude
the EH antenna with 2 close cylinders does not work by Poynting Vector Synthesis. The lower cylinder is little more than cosmetic and can be dispensed with.

A
further webpage is for tests beyond feed line radiation and EH antennas.