How thin can you go? Wire gauge for POTA

Every portable operator has packed this decision. The 28 AWG magnet-wire dipole weighs nothing and disappears into a tree; the 18 AWG version survives being yanked out of one. Between them sits a real three-way tradeoff — weight vs. SWR bandwidth vs. radiated power — that mostly gets settled by folklore (“thin wire is lossy”, “thick wire is broader”). This example settles it with a solver instead.

One catalog design carries the whole comparison: dipoles.pota_invvee — a half-wave 20 m inverted-V on a 7 m telescoping pole over average ground, where the wire itself is a knob. The wire_type dropdown selects from a wire catalog the same way the station designs pick a feedline from the cable catalog: 28, 22, 18, 16, 14 or 12 AWG copper, bare or PVC-insulated, each entry carrying its real radius, conductivity, jacket, and grams per meter.

What the wire actually does in the model

This isn’t a lookup-table correction. Since momwire’s distributed-loading release, the wire’s material enters the impedance matrix itself as a per-meter series impedance:

Skin-effect resistance. At 14 MHz current flows in the outer ~17 µm of the copper, so RF resistance runs 5–15× the DC value — about 1 Ω/m for 28 AWG, 0.3 Ω/m for 18 AWG. The solver uses the exact solid-round-conductor law (the same one NEC’s LD 5 card implements), valid from DC through full skin effect.
Insulation loading. A dielectric jacket stores E-field energy next to the wire, which slows the wave on the wire, and the antenna tunes a few percent lower than bare wire cut to the same length. That’s the velocity factor behind “my cut-to-formula insulated dipole came out long.”

Both engines model both effects. Conductor loss: the momwire loading and PyNEC’s native LD 5 card agree on the added resistance to half a percent. Insulation: NEC has no insulated-wire card, so the NEC deck spells the jacket the way momwire models it, as a pair. The wire card carries a slightly larger equivalent radius (the jacket’s effect on the charge), a distributed series load (LD 2, henries per metre) puts back the inductance the larger radius removed, and LD 5 is rescaled so the copper loss stays the real conductor’s. Each pairing is a cross-engine oracle in the test suite — two independent implementations of the same physics, kept honest against each other.

The numbers

The stock design, PVC-insulated wire, with the dropdown flipped and everything else held — the same length for every wire. The 2:1 SWR window is on 50 Ω over average ground; the weight is for the full ~10 m of wire; the loss is relative to a perfect conductor:

wire	weight	resonance	2:1 SWR window	gain vs lossless wire
28 AWG PVC	17 g	14.00 MHz	590 kHz	−0.35 dB (92 %)
22 AWG PVC	53 g	14.08 MHz	640 kHz	−0.17 dB (96 %)
18 AWG PVC	111 g	14.17 MHz	680 kHz	−0.11 dB (97.5 %)
16 AWG PVC	162 g	14.21 MHz	700 kHz	−0.09 dB (98 %)
14 AWG PVC	249 g	14.22 MHz	710 kHz	−0.07 dB (98.5 %)
12 AWG PVC	377 g	14.25 MHz	740 kHz	−0.05 dB (99 %)

Re-cutting each wire to resonance at 14.1 MHz moves its window by 10 kHz or less and its loss not at all, so the table reads the same either way.

Four readings worth taking home:

The loss argument is real but small. Going from 18 AWG all the way down to 28 AWG costs a quarter of a decibel — nobody on the other end will hear it. The power budget in the workbench shows exactly where it went: a wire loss (I²R) row that grows from ~2.5 % to ~8 % of the input power as the wire thins.
Bandwidth favors thick wire, for the geometric reason. Two effects stack here and pull the same way with different price tags: a fatter conductor is intrinsically broader (lower ln(L/a) Q — free), and a lossier conductor is also broader (resistance flattens the SWR curve — paid for in watts). The table shows the honest sum: 18 AWG is ~90 kHz broader than 28 AWG and radiates more. Thin wire’s loss does not buy its bandwidth back.
Past 18 AWG the returns flatten. From 18 to 12 AWG buys 0.06 dB and 60 kHz of window for 3.4× the weight. The case against thin wire is mostly made by 22 AWG; the heavy gauges are for strength, not signal.
The insulation is a bigger tuning effect than the gauge. Flip any PVC entry to its bare twin and the resonance jumps 370–670 kHz, to about 14.65 MHz — several times the entire gauge-to-gauge bandwidth difference. Even among the PVC wires, most of the 250 kHz the resonance walks from 28 to 12 AWG is the jacket: a thin conductor carries a proportionally thicker one, and bare wire of every gauge resonates within 50 kHz of the others. Cut your antenna for the wire you’ll actually hang.

And the weight column is the one the solver can’t argue with: the 28 AWG antenna is 17 grams. For a summit activation where every antenna is a compromise, −0.36 dB for a 6× lighter pack is a fine trade — now it’s a number instead of a guess.

Try it

Open dipoles.pota_invvee in the simulator, put the sweep on 20 m, and flip wire_type through the catalog. Watch the wire weight row in the Info pane, the wire loss (I²R) row in the power budget, and the SWR trace breathe as the gauge changes. Then flip 22 AWG PVC to bare 22 AWG and watch the whole resonance walk up the band — the velocity factor, live.

The design also carries the comparison as charts, under Wire gauge in the chart picker, each with one curve per PVC gauge:

Z vs length by gauge — R and X against length_factor. Each wire’s X = 0 crossing is its resonant length, from 0.934 for 28 AWG to 0.951 for 12 AWG.
resonant length by gauge — the length re-held at resonance as the mast goes from 4 to 10 m, with the held length_factor drawn beside it.
patterns by gauge — the elevation and azimuth cuts and the pattern table. The shapes lie on top of each other; only the peak gain moves, by the loss column above (4.94 dBi for 28 AWG, 5.23 dBi for 12 AWG).
band-tuned SWR by gauge — each wire with its length and droop angle tuned for the band edges (the worse of the SWRs at 14.0 and 14.35 MHz as low as it goes), swept wide enough to show the whole 2:1 window.

That last chart answers the question the table above leaves open: what if you cut each wire for the band instead of keeping one length? Every gauge then lands at about 42–43° of droop, bottoms out near 1.07 at 14.18 MHz, and covers all of 20 m:

wire	2:1 SWR window	worst band-edge SWR
28 AWG PVC	603 kHz	1.51
22 AWG PVC	644 kHz	1.47
18 AWG PVC	673 kHz	1.45
16 AWG PVC	689 kHz	1.44
14 AWG PVC	708 kHz	1.42
12 AWG PVC	726 kHz	1.41

Tuned for the band, gauge is worth about a tenth of an SWR unit at the edges. On this antenna, bandwidth alone is no reason to carry heavier wire. The tunings are fixed numbers for the stock 7 m mast; move the mast and they no longer apply.

The same four run from the command line:

python -m antennaknobs analyze --builder dipoles.pota_invvee --list
python -m antennaknobs analyze --builder dipoles.pota_invvee \
    --analysis "Z vs length by gauge" --fn gauge.png