Decision & bench craft · theory and measurements live in the research site

Baluns & Ununs
Types, Build, Best Practice

A builder's page, not a physics page. It answers four questions: which magnetic device you are actually holding, which one to pick for your power and bands, how to wind and seal it without destroying it, and what to check before you blame the antenna. Every "why" — core materials, loss measurements, the 49:1 versus 64:1 argument, VNA methodology — is owned by the EFHW Deep Research site and linked, never copied here.

Five devices, three jobs

"Balun" is a job description, not a part. Three jobs get conflated: impedance transformation, balanced-current enforcement, and common-mode suppression. A device usually does one and a half. Read the middle column as "which job this actually performs", and the last column as the reason people burn money.

Label discipline on this page. [C] = engineering consensus with a citable source, or arithmetic you can redo from the formula shown. [M] = measured by us, with the raw data file in the repository. Anything we have not measured is written as and pointed at the research site. We do not print numbers we cannot derive or show.
Device Job it really does Working bands Pick it when Do not pick it when
49:1 unun C Steps 50 Ω up to ~2450 Ω (50 × 49). An autotransformer: it does not balance anything. 1.8–30 MHz, one wire cut for the lowest band End-fed half-wave where the feedpoint really is a few kΩ. You need symmetry, or you run S-metre power through a small core and call the resulting heat “normal”.
64:1 unun C Same job, 3200 Ω step-up (50 × 64). As above Your measured feedpoint sits nearer 3–5 kΩ than 2.5. You want a decision made for you — the 49 vs 64 argument is a measurement argument, see the research site.
Guanella / TLT current balun C Transforms by carrying power in a transmission line wound on the core; forces equal and opposite conductor currents. Ratio per section is small (1:4 typical, two sections for 1:16) You want balance and transformation, wideband, with headroom on heat. You need 1:49 from one small core — the section count and size grow with the ratio.
1:1 current balun C No transformation. High series impedance to common mode on both conductors. Wherever the core’s impedance peak is Feeding a balanced antenna (dipole, loop) with coax. An end-fed wire — it fixes balance, not the 2–5 kΩ mismatch.
Common-mode choke (CMC) C Suppresses shield current. Not a balun at all. Broadest of the five Always, on an end-fed. This is the TV/RPI noise item. You expect it to improve match — it acts on what leaves the antenna system, not on Zₚ.
LTA / distributed C Lossy transmission-line section; no core, no saturation. VHF/UHF mostly; HF designs get physically long Very wideband duty, or you refuse to trust ferrite thermally. HF legal limit — the loss you accept is real heat in a box.
Dipole + 1:1 current balun C Nothing transforms; the balanced antenna presents ~50Ω and the balun only keeps the shield quiet. Per band (or a trap/fan dipole) You can actually put up two halves. This is the honest baseline. You have one support and one wire — then you are back on the row above.

Decide by constraint, not by brand

Four constraints, applied in order. Each leaf names the device and links the measurement argument that backs it — we do not re-print their numbers here.

Do this arithmetic before you buy a core. Feedpoint RMS voltage on an end-fed half wave is V = sqrt(P x Z): 100 W into 2450 ohm gives about 495 V RMS, roughly 700 V peak. C That is why arc-over, turn spacing and the first few turns of insulation are the failure point at power, not the ferrite. Write the number down for your own power and load; it decides your wire, your form of winding and your enclosure.

Where good designs die

Same core, same turns, same radio — one builder gets 20 dB of return loss, another gets a warm brick that deades on the low bands. The difference is almost entirely process.

Inspect before contact

Chamfer and deburr the bore of the core with a fine file or tape. A sharp inner rim chafes enamel under tension on the far side, where you cannot see it.

Failure mode: a turn-to-core short that appears only when cold, or an internal short after one season of thermal cycling.

Wind uniformly, light tension

Spread the turns evenly over as much circumference as you can, with constant tension. Pile-ups raise local capacitance and local flux density; wide gaps leak field. Both cost you the low bands first.

Failure mode: fine on 20 m, dead on 80 m.

Mark the phase dots before you cut

Wrong phasing on a transmission-line transformer turns a balun into a plain choke with no transformation; wrong phasing on an autotransformer gives you a shorted turn. Mark start and finish before trimming the wire.

Failure mode: high SWR everywhere, hot core at low power, builder convinced the circuit is bogus.

Short leads, no pigtails

At a 2–5 kΩ node every millimetre of stray lead is added inductance and a small antenna. Solder to the terminal, not to a loop of stranded wire.

Failure mode: return loss that moves when you rotate the box in your hand.

Impregnate, do not glaze

Void-free impregnation is what makes a measured prototype survive a rooftop. Warm the assembly, flood with low-viscosity epoxy or polyurethane varnish, then let it outgas in vacuum so trapped air does not stay as an insulated bubble around a lossy turn.

Failure mode: the bench numbers were right; the resin hid a void that became a hot spot at duty cycle, and a foam-like layer that thermally insulates the very thing that gets hot.

Seal against water, not against curiosity

Drain hole at the lowest point of the enclosure, gasketed lid, UV-stable body, and strain relief on the feedline that clamps the jacket rather than the conductors.

Failure mode: the commonest field failure of all is a box that collected rain over winter. Water in a high-field region does not just corrode, it adds loss that grows with frequency.

Give the cold ground a place to live

The unun’s ground terminal and the coax shield must join at one short, low-inductance point, and that point is where your counterpoise or radial attaches. Long earth studs do not help; a copper plate and a bolted join do.

Failure mode: an “antenna that works fine indoors and terrible outside” is usually a ground/counterpoise story, not a transformer story.

Verify before you close the lid

Measure before potting, not after. A 1-port sweep with a known-resistive load substituted at the far end tells you whether the build matches the design; a sweep of the finished antenna tells you nothing about the transformer specifically.

Failure mode: sealed box, surprising SWR, no way to know whose fault it is, new box built from the same mistake.

What a durable feedpoint actually costs

Generic parts, no brand endorsement, and no price claim — prices move. The point of this table is the parts people omit, then wonder why the box failed.

Item Why it is on the list What gets skipped
Toroid core, material per Q4 above The design’s thermal and impedance budget lives here. Buying one core for every experiment.
Enamelled copper, two gauges (fine for the ratio, heavy for the high-current turn) An autotransformer’s low-Z turn carries the whole antenna current. Using one thin wire throughout.
PTFE or FEP insulated wire for the first turns That is where several hundred volts peak appears. Assuming enamel is an insulation system at QRO.
Low-viscosity impregnating resin + vacuum bag or pump Voids, not resin, are what kill the build. Brushing on surface varnish and calling it potted.
IP-rated enclosure, gasket, drain hole, UV-stable Rain and sun out-ferrite any HF design. A 3D-printed lid with no gasket.
Chassis SO-239 or N panel mount + short internal links Keeps the ground node short and measurable. Feeding wire straight to a PCB pad.
Copper plate or bus bar for the cold-ground / radial stud One low-inductance join for shield and counterpoise. “It is only ground.”
Ferrite beads or a spare core for the CMC The CMC is a separate device; you need both. Expecting the unun to stop shield current.

Install it, then argue with the numbers

Optimisation targets you can check on paper, plus the four field practices that move real-world results more than any winding tweak.

Four practices, in the order they pay

  1. Put a CMC on the feedline, close to the feedpoint. It does not change your match; it stops the shield from becoming the other half of the antenna, which is what actually causes the TV interference, the hot coax and the “SWR changes when I stand near it” report. C
  2. Give the counterpoise a deliberate length and route. On a single-ended feed the radials are part of the radiator, not housekeeping. C Length choice and its effect on pattern are adjudicated in Counterpoise — Balance Ground.
  3. Keep the feedline away from the wire for a wavelength or so. Coupling between the last metre of coax and the antenna end invalidates everything you measured at the bench.
  4. Measure the system twice: once at the rig, once at the box. If you only ever measure at the rig, coax loss is hiding inside your conclusion about the transformer.
Target How to check it on paper Our measured value
Flux density headroom Bmax = V / (4.44 · fmin · Nprimary · Ae). Raise turns or core area, never just the wire gauge. C — see research site
Insertion loss budget Core loss grows with frequency and with flux density; it is a trade, not a spec sheet number. C 49:1 loss measurements
Thermal rise at duty cycle Estimate dissipated power, then check it against the enclosure’s surface area. FT8 and AM are duty-cycle tests, SSB is not. C — not measured by us
Self-resonance above the top band Winding capacitance plus magnetising inductance sets a ceiling; more turns buys low-band performance and spends high-band headroom. C — see research site
Power derating point The cliff is material-dependent and abrupt on Mn-Zn. Do not extrapolate from a QRP success. C stacked-core comparison
The honest row at the bottom of the table. Not one value above is measured by us. This page is deliberately silent about dB and degrees Celsius: means “we have not measured it”, not a number we trust.

When to stop optimising the balun

Optimisation has an exit condition. Choose a different antenna rather than a better transformer when:

  • You need a symmetrical pattern, low take-off angle on the low bands, or real rejection of common mode — a balanced antenna with a 1:1 current balun is a different, better solution, not a worse balun.
  • Your tuning loop costs more loss than the transformer you are trying to perfect. The research site’s transformer work is excellent; a mismatched ATU will still eat the difference.
  • The band you want puts the feedpoint above what one wire and one core can hold. Then it is a physics limit, not a craft limit.

Where the numbers on this page come from

This page publishes no measurement of its own. Anything quantitative here is either arithmetic from a stated formula, or a pointer to the work that measured it.

G. Guanella — transmission-line transformer principle H. Guggenbüll — broad-band matching of balanced lines W8JI — balun & transformer measurements PA3HHO — EFHF / EFHW transformer research W. Chew — RF Transmission Line Transformers ARRL Handbook — matching networks

Owned by this site’s research hub (theory, VNA methodology, measured comparisons):

M rows reserved: any value we measure ourselves will cite its raw data file under the repository logs/ directory.

Build one, then measure it

This page ends where a claim would start. If you do wind and pot one, bring a VNA, bring a 2.5 kΩ non-inductive load, and bring the methodology the research site already argues about.