HF Horizontal-U Wire: Why We Prefer EF-OCF Feeding Over EFHW
HF Horizontal-U Wire: Why We Prefer EF-OCF Feeding Over EFHW
Jean-Claude Ducasse, F1QM, fitted a long HF wire into his garden as a folded, fractal-like Horizontal-U. That geometry is the reason for this comparison. The question is not which transformer label creates a magic pattern; it is which feed architecture lets the installed wire, return path and choke boundary do useful work across the intended bands.
F1QM's installation is a fine example of making the available garden part of the antenna design. His Horizontal-U photograph shows the physical idea: fold a long wire around the site instead of pretending the site is an empty rectangle. I would keep that useful geometry. I would then choose the source, transformer, intentional return branch and choke as one circuit.
My position: for this kind of multiband folded wire, I prefer an EF-OCF architecture when its measured complex feed impedance supports a moderate transformation ratio. A 4:1 UNUN can perform that transformation, while a separately specified choke establishes the common-mode boundary. That is a design preference with conditions—not proof that every 4:1 implementation beats every EFHW.
Start With F1QM's Wire, Not With a Transformer Label
“Fractal” is useful visual shorthand here, not a claim of a mathematically self-similar antenna. Electrically, this is a long conductor folded into a Horizontal-U with changing height, spacing and proximity to soil, buildings, vegetation and other conductors. Those details change segment coupling, electrical length, impedance and current phase.
A NEC model solves conductor currents for the geometry, environment, source and loads supplied to it, then derives the field from those currents. The outline matters enormously, but it is not the whole model. Moving the source, adding a second wire branch, using the outside of the coax as a branch, changing the choke boundary or omitting the feedline can all change the solved current distribution.
That is why I do not accept either shortcut: “the box label creates the lobe” is wrong, but “the same wire outline must keep the same pattern after the feed architecture changes” is also too strong. The complete driven structure decides.
EFHW, OCF Dipole and EF-OCF Are Different Circuits
The names are often mixed together, so define the conductors before comparing results:
| Architecture | Driven wire and return path | Matching and common-mode boundary |
|---|---|---|
| EFHW | A wire is driven near a high-impedance end region. A real return path still exists through a counterpoise, capacitance, feedline exterior, support structure or a combination. | A high-ratio network may transform the measured end impedance. Common-mode control is a separate installed-current question. |
| Conventional OCF dipole | Two wire legs meet at an off-centre feedpoint and are intended to carry the differential antenna current. | The network must suit the measured load. A choke may be needed when the installed system drives feedline-exterior current. |
| EF-OCF / EFOC | RF.Guru's one-end-access interpretation makes the asymmetric branches explicit. One branch is the main wire; the declared coax-exterior section may form the other branch until the choke boundary. | When the measured load calls for it, a 4:1 UNUN performs impedance transformation and a separate choke defines where intentional exterior current should end. |
A coax data-sheet velocity factor applies to the internal differential mode between centre conductor and shield. It is not a cut-length constant for current on the outside of the shield. The exterior branch belongs to the installed single-conductor environment, so its electrical boundary is confirmed from current and impedance behaviour on every intended band.
Why the Moderate-Ratio Route Appeals to Me
An end-fed half-wave mode can present a high, frequency-dependent complex impedance. A folded multiband wire can move that impedance substantially as height, bends and coupling change. A nominal 49:1 network may be appropriate for one declared load and poor for another; the number alone does not establish match, loss, voltage or temperature.
Moving to an off-centre source can place the feedpoint in a more moderate impedance region on useful modes. When measurements show that relationship, a nominal 4:1 network asks for less impedance transformation and may be easier to realise over the required bandwidth. It also lets me separate the jobs: the UNUN transforms the differential load; the choke controls the unwanted continuation of common-mode current.
That separation covers the unbalanced reality of most amateur installations. Feedline routing, unequal capacitance, supports, buildings and station wiring seldom preserve textbook symmetry. It can also serve the uncommon installation that behaves as a balanced load, but only when the complete UNUN-plus-choke arrangement is deliberately configured and verified. A current balun does not automatically cover the normal unbalanced case merely because its ratio is 4:1.
A lower ratio is an opportunity, not an efficiency certificate. Core material, winding topology, conductor length, leakage inductance, capacitance, flux, load phase, frequency and cooling still determine loss and stress. Fewer turns or a lower nominal ratio does not prove that one transformer is better.
The Non-Harmonic Bands Expose the Real Design
On a long 80-metre-class wire, some higher amateur bands lie near useful harmonic modes while 60, 17 and 12 metres do not fall into the same simple sequence. Folding the conductor around a garden changes every mode again. Those bands are valuable because they expose whether the system was designed from its actual loads or from a convenient label.
I do not promise that EF-OCF feeding puts every band into one impedance window. The chosen feedpoint can make several loads more moderate and tuner friendly, or it can trade one awkward band for another. A tuner can transform the impedance presented at its own port; it cannot remove transformer loss, feedline loss, unintended common-mode current or an unwanted far-field pattern.
That is the practical reason for preferring the EF-OCF route here: it gives us feedpoint position and intentional return-path length as design variables, while allowing transformation and common-mode control to be specified separately. We still have to earn the result band by band.
Ten Metres Makes the Pattern Question Impossible to Ignore
At 10 metres, an 80-metre-class conductor contains many electrical wavelengths. Small changes in bends, height, nearby objects, source location and return branch can move current maxima and nulls. The far field can contain many narrow lobes and deep directions of weak response.
Neither a low SWR nor a preferred transformer ratio proves that those lobes point toward the desired path. The only defensible pattern statement comes from a model that includes the complete driven geometry and a field check suitable for the claim. On 80 or 40 metres the two feed choices may produce broadly similar useful coverage; on the upper bands it is unsafe to assume that they do.
Turn the Garden Layout Into a Defined Antenna
I would develop F1QM's Horizontal-U in this order:
- Preserve the available geometry: record every leg, bend, height, support and nearby conductor rather than replacing the garden with a straight-wire sketch.
- Model complete alternatives: include source position, both intended branches, feedline exterior, choke boundary, realistic ground and loads. Compare accepted-power patterns, not only normalized plots.
- Measure the complex load: use consistent reference planes at the intended EFHW and EF-OCF feedpoints on every band.
- Select the network from that load: verify insertion loss, return loss, temperature and voltage margin across the operating range instead of selecting 49:1 or 4:1 by antenna name.
- Map exterior current: place and specify the choke from the intended branch boundary, then confirm that current beyond it is acceptably small.
- Verify what matters on air: compare accepted power, calibrated field or restored-baseline A/B/B/A results on declared paths. SWR alone cannot rank the antennas.
Conversion is therefore not always “move the box and shorten the wire.” A conventional OCF dipole needs two defined wire legs. A one-end EF-OCF needs a deliberately declared second branch and choke boundary. Reuse the conductor where sensible, but redraw and remeasure the complete circuit.
Bottom line: F1QM's folded Horizontal-U is the valuable idea. For a multiband implementation, I prefer EF-OCF feeding when the installed loads support it because it exposes the return path and separates a moderate-ratio UNUN from the common-mode choke. Geometry, source position and return current still decide the pattern, and measurement—not the ratio printed on a box—decides whether the preference delivered an advantage.
Model and current-path references
Mini-FAQ
- Is the Horizontal-U a true fractal antenna? The name is visual shorthand for a long wire folded around the site. Its electrical behaviour comes from the actual segments, coupling, heights, source and return path—not from the word “fractal.”
- Does the same wire outline guarantee the same pattern? No. The outline is important, but source position, the second branch, feedline exterior, choke and environment change the conductor currents from which the pattern is formed.
- Is a 4:1 UNUN always better than a 49:1 EFHW network? No. The measured complex load, bandwidth, loss, voltage, temperature and current paths decide. A lower ratio is useful only when it suits the installed load.
- Are 17 and 12 metres automatically poor on an EFHW? No. They are not simple harmonic modes of an 80-metre fundamental, so the installed load may be awkward, but geometry and matching decide the actual result.
- Can a tuner make the Horizontal-U efficient? A tuner can transform impedance at its reference plane. It cannot erase feedline or transformer loss, suppress exterior current by itself or choose the far-field lobes.
- May the coax exterior be an intentional EF-OCF branch? Yes, when that branch is explicitly designed to end at a separately specified choke. Its length and current must be verified on each band rather than inferred from coax internal velocity factor.