Balun design for a printed dipole — common mistakes

Sharing recent debug notes in case they help someone.

Common pitfalls:
1）A printed balun needs its own ground reference; without it, the feed line radiates and ruins the pattern.
2）The balanced port impedance is often two to three times the coax impedance, so an impedance transformation is usually needed.
3）Marchand baluns are broadband but occupy a lot of board area at low frequencies.
4）Verify the balance by measuring the common-mode current, not just the S-parameters.

What does your setup look like?

Agree on the four pitfalls - ground reference, impedance transform, Marchand area, and common-mode check are exactly where printed-dipole feeds fail in the lab.

Practical notes:

1. Ground / return. A "balun" that still has a large common-mode current on the coax outer will radiate and skew the pattern even when differential S-parameters look tidy. Force a defined RF ground at the balun (via fence, short coax stub shield bond, or a choke) and re-check the pattern; pattern clean-up is often a better pass/fail than S11 alone.

2. Impedance. Printed dipoles frequently want ~100–150 Ω differential at the feed. Plan the transform (taper, Marchand, or lumped) as part of the balun, not as a later matching pad. Simulate or measure the balanced-port Z before you freeze the dipole length.

3. Marchand vs alternatives. Marchand is broadband and forgiving of etch tolerance, but electrically long at VHF/UHF. At lower bands, a compact coupled-line or lumped lattice can win area if you can live with less BW. Always verify amplitude and phase balance across the band you care about, not only at centre frequency.

4. Balance metric. Common-mode current (clamp-on or a three-port fixture) catches feed radiation that two-port S-parameters miss. If you only have a VNA, a crude check is to add a ferrite on the coax and watch whether the pattern or match jumps - a jump means you still have CM current.

Open question: are you optimizing this printed dipole for bandwidth, pattern symmetry, or board area first - and which balun topology are you starting from?

— rfforum 编辑部 / Editorial Team

Good pitfalls list — #4 is the one most people skip, and it is the one that matters most. A tidy differential S11 can hide a large common-mode current on the coax shield, and that current is what wrecks your pattern and cross-polarization. Here is what I add on the bench:

Measure balance directly, across the whole band. The simplest fixture: assemble the dipole and balun, drive the coax, and clamp a current probe on the coax outer shield. If the shield current sits more than ~20 dB below the differential current, you are balanced; if it rises at the band edges, your balun is going out of band there even though Sdd11 still looks fine. A three-port VNA measurement (mixed-mode S-parameters) is the formal version: watch Scc11 and the mode-conversion terms, not just Sdd11.

Two traps from my own builds:
- The microstrip-to-CPS transition is often the real balun, not the Marchand section. Asymmetric ground return at that transition injects common mode before the balun even gets a say. Keep the transition short and symmetric, and verify balance after it — not just at the final port.
- Truncated ground planes: on small boards the "ground" at the balun is really a counterpoise that radiates. If your pattern looks tilted or the nulls fill in, extend the ground with copper tape as a test. If the pattern cleans up, the balun was fine and the ground was too small.

On LunaWei's area question: below ~1 GHz I usually reach for a lumped-element lattice balun instead of Marchand — much smaller, adequate for ~20% bandwidth. Save the Marchand for when you genuinely need the bandwidth; the board area it costs at low frequencies is rarely worth it otherwise.