CFA Top Hat Effects
In last month’s experiments described in my article, LAB NOTES: CFA Looking for Poynting Synthesis, in the August 2000 issue of antenneX experiment, I combined a loop-based magnetic field source with a capacitor E field radiator and investigated the properties of the radiation from exciting the two sources at various relative amplitudes and phases. It was found that the total radiation was the linear superposition of the separate radiation from each source.
This month’s experiment was to drive a “top hat” configuration CFA using the two-channel drive. I found that superposition still explains the total radiation, and that if the E plate voltage and D plate current relative phases are 0 degrees (in phase) the radiated field is a maximum, but if the phasing is 180 degrees (exactly out of phase) a null in the radiation can be observed without a corresponding null in either drive voltage and current. The fact that the E plate radiation can be made to cancel the D plate radiation reinforces the idea that the CFA does not synthesize the poynting vector, but instead is an unusual form of a 2-element antenna. The ability to squeeze two radiators into essentially the same space suggests that the CFA can be twice as effective as a single short monopole under the correct conditions.
The size of the CFA used in this experiment is:
E plate cylinder: diameter = 4.875 inches, length = 6.75 inches
E plate height above D plate = 4 inches
D plate diameter = 12 inches, hole = 1.93 inches
D plate spacing = 0.4 inches
Ground plane = 19 x 22 inches
Figure 1 shows the circuit diagram. Each plate drive is via a 50-ohm cable and a stepped voltage attenuator (20 dB = 10 x power change) from a power splitter. An oscilloscope is externally triggered from the RF signal generator to start the traces at the same point on the waveform regardless of the amplitude and phase variation of the waveform being viewed on the screen. A resistive load terminates the cables at the D plate and the E plate; both plates add a capacitance across these loads. Extra isolating fixed attenuators damp out any mismatch reflections and keep the voltages and currents stable at all frequencies and adjustments.
The D plate is fed on its edge from a connector about an inch away from the plate. The terminating resistor is connected at the connector end of the 1-inch feed wire, and the wire itself is led once through the current sensing toroid. The current sensor is calibrated in RF amperes by the appropriate loading of the secondary of the toroid with a resistor. The phasing of the current sensor is checked by grounding the D plate with a 20-ohm resistor while observing the voltage and current on the two scope channels. (The toroid secondary output can be inverted or not, determined by which output terminal is grounded. Much of the data was taken with this coil inverted, a fact that became obvious when I analyzed the data and checked this connection)

The CFA was first tested connected in the minimum lead length tuner configuration described in the June issue of antenneX, driven by the D plate channel of Figure 1. The 3.6 MHz radiated signal at the sense antenna 55 feet away was -47 dBm. The tuner was then disconnected and the D and E plates driven separately through the attenuator channels. The CFA had a 500 pF high capacitance D plate with small spacing and a lucite dielectric to increase the capacitance. The two-channel test showed that the D plate radiation was very small compared to the E plate radiation.
Table 1 shows radiation of this CFA configuration. The D plate signal level was in the receiver noise, and did not have enough dynamic range to allow any adjustment of amplitude, so the high capacitance plate was disassembled, the spacing increased to 0.4 inches, and the dielectric removed. The resulting D plate has a capacitance of about 50 pF.

Using the spectrum analyzer, it was easy to track the radiation as the frequency was increased. Heikki Antman, OH2BGC, suggested that my CFA was too small for 80 meters and this test seemed to support that observation. The radiation increased with frequency until about 5 MHz, then leveled off to above 15 MHz. The rest of the experiment was done at frequencies around 5.8 MHz.
Phasing between the channels is done by inserting a cable in one or the other channel, or removing the cable to get a 90-degree phase between D plate current and E plate voltage. The frequency can be adjusted in order to scan the phase a few degrees very precisely and smoothly (as opposed to tuning a resonant circuit, which changes phase very rapidly). Because the cable that was moved to go from 0 degrees phase to 180 degrees phase was not the whole channel cable length, and because the channel residual cable lengths were slightly different, the frequency had to be reset about 200 kHz when making this change.
Table 2 shows the radiation for the 90-degree condition. Note that superposition is indicated by the radiated signal change with attenuation. Most of the radiation comes from the E plate: a 20 dB voltage attenuation change (10 x) produces a 10 dB (10 x power) change in the radiated signal in the E plate channel, but only a small change when the D plate drive is changed. The D plate induced change increases as the E plate drive is reduced.

In this part of the experiment, the interaction between the E plate drive and the D plate current was very visible in the D plate current. Coupling from E to D plate resulted in current from the E plate through the coupling capacitance between the plates to the D plate, then through the 1-inch connecting wire to the D plate 50-ohm termination, then to ground through the resistor. The coupling was much larger than I expected. It acted like a fraction of the E plate voltage was impressed across the terminating resistor, which limited the current. At high E plate voltages (0 dB attenuation) the phase of the current through the sensor was near 0 degrees; at low E plate voltages (20and 40 dB attenuation in Table 2) the current indicated was phased 90 degrees to the E plate voltage. At this setting all current came from the D plate drive. This strong interaction may greatly complicate tuner design, and some experimenters have called it a form of SWR reflection. In this experimental configuration, the interacted power is dissipated in the loads and attenuators.
Table 3 shows the phase values in terms of the position on the horizontal scale of the oscilloscope where the RF sinewave crosses in the upward direction. The number of scale marks in one cycle is shown, along with the resulting phase calculation.

Table 4 shows the detected radiation at 5.8 MHz when E plate voltage and D plate current are in phase. The characteristic that indicates superposition – the small effect of the d plate variation at high E plate drive levels — is clearly evident. The change in D plate current is not 40 dB (100 x). I believe this is because the E plate to D plate coupling contributes to the current flow measured at this point at high E plate drives to reduce the measured value.

Table 5 shows the detected radiation change as frequency is adjusted. The phase changes about 15 degrees over this frequency range. Note that the radiation changes are small and show no sharp peaks and valleys, and the radiated power roughly follows the change in E plate attenuation.
Table 6 shows the same data as Table 4, except the phase between E plate voltage and D plate current is 180 degrees.
Note that the radiated signal is lower in all cases than the corresponding point in phase data of Table 4, and for E plate attenuation levels between 16 and 20 dB and D plate attenuation levels between 3 and 6 dB, a deep radiation null is visible (indicated in red background). The spectrum analyzer noise floor is between -100 and about -95 dBm.


Table 7 shows the detected power versus frequency for the out of phase condition, corresponding to Table 5. In addition, two extra sweeps with an E plate attenuation of 18 dB are inserted to compare directly the in and out of phase conditions near the best null values. The phase and amplitude dependence of these nulls are clearly shown in Tables 6 and 7. The system would only balance in this fashion if superposition holds instead of Poynting Vector Synthesis, which would require either E voltage or D current to go to zero for a null to occur. The indicated behavior is the same as would be found in a wheatstone bridge or other linear nulling device.

These tables show a possible way to optimize the CFA: tune it for a radiation null, then, reverse the phase of the drive in one channel. But, notice that the largest in phase radiated signal at the reversed null setting is smaller than that achieved when the E plate drive is raised to 0 dB attenuation. So, it appears that the best way to adjust the CFA is to separately maximize the D plate and E plate radiation by tuning for highest E plate voltage and highest D plate current and making sure the two are in phase. But phase is not too critical when D plate current and E plate voltage are adding in phase. The orientation in space of the E fields 90 degrees to the H field is also critical for a deep null, which really depends on both the E fields and H fields in space cancelling simultaneously.
I downloaded the inventors’ articles in wireless world. These can be found at: http://metalab.unc.edu/modena/cfa.html
In particular, the article “CFA: working assumption “ December 1990 describes the CFA well, except for the poynting vector synthesis assertion. In Figure 9 of that article, a curve of “efficiency” vs. phase is shown. That illustration shows an extreme sensitivity with about a 2 to 4 degree range (depending on frequency) between the 25% efficiency points. I found no such behavior in my experiments.
I believe it would be very helpful if the inventors of the CFA dusted off their notebooks and described their experiments that proved to them that poynting vector synthesis occurs. “Science” implies, among other things, that the particular effects or experiments can be duplicated. Being secretive about the technology of an invention is only useful for a little while for soon someone else will figure out the “secret”, simply because they know it exists – it’s the way of the world. Surely, if the inventors have duplicable experimental proof of the poynting synthesis theory, it would be advisable to make this public. It would eliminate a lot of controversy, and make a good article in antenneX!
Until then, I think my next experiment will be to duplicate the Tables above at 14 or 21 MHz, to test the effect of size on the null and radiation ability, and check the loop based H-element “CFA” for the null behavior. Then, on to a tuner that separately maximizes radiation.
Originally posted on the AntennaX Online Magazine by Joel C. Hungerford, KB1EGI
Last Updated : 21st May 2024