Cut for one band, worked on another

Real antennas get cut once and worked everywhere. The 10 m inverted-L that was perfect in October is what you have when the propagation moves to 12 m; the 80 m skyloop is what’s in the air when 17 m opens. You don’t re-saw the wire — you reach for the tuner. Two catalog designs model exactly that situation, and they lean on a distinction the workbench keeps carefully: the frequency the antenna is built for is not the frequency you measure it at.

Two frequencies, two controls

Every design carries both:

• design_freq sizes the geometry. It’s the saw and the tape measure — change it and the wires themselves change length. In the workbench it’s the band tabs row.
• freq is where the measurement happens: the impedance readout, the SWR, the Smith chart, the pattern. In the workbench it’s the dial.

For most designs the two travel together — the dial is locked to the design frequency, and retuning the antenna drags the measurement along. That lock is a convenience, not a law. Open it and you can park the dial anywhere while the geometry stays frozen: that’s “checking what my 10 m antenna looks like on 12 m” as a single click.

Off-band designs open with the lock already open: geometry parked on the band they’re cut for, dial parked on the band they’re operated on. (They have to — snapping both controls to one frequency would silently re-saw the antenna for the operating band and dissolve the design’s whole premise.)

The hard case: a short antenna and a T-network

verticals.inverted_l_tmatch is a 10 m inverted-L (cut at 28.57 MHz) worked on 12 m. At 24.9 MHz the riser is electrically short and the feed sees roughly 11 − 117 j Ω — low resistance, a big capacitive reactance, nowhere near 50 Ω. The classic fix is the ham T-tuner: two series capacitors flanking a shunt inductor.

The stock knobs are the solved tuner — the design opens as the “after” picture, and the interesting move is turning the knobs and watching the match fall apart. Three things to look at:

1. The power budget. The tuner is a composite box, so its rows sit indented under tuner in the readout: series C1, shunt coil, series C2. The coil row is the story — a T-network matches a short antenna by riding a virtual resistance of ~2 kΩ through the tee, which runs high circulating current through the inductor. At the stock coil Q of 200 (a good air-wound coil) that burns ~9 % of everything you feed it, before a single watt reaches the wire. This is the classic hidden cost of the T-match, and it’s why the design ships a real coil: set coil_q to 0 (ideal) and the loss vanishes — along with the entire budget display, because a lossless network has nothing to report.
2. The loaded Q. The match is narrow — sweep the band and the SWR notch is sharp (loaded Q ≈ 13). That’s the real-world “retune every 50 kHz” behavior of tuners on short antennas, reproduced from first principles.
3. The touchiness. The stock values land SWR ≈ 1.0 in the workbench, but they only stay there because they were tuned against the exact solve the workbench runs. Solve the same stock design on a different basis (the test suite’s sinusoidal reference, say) and the bare antenna moves by only a few percent — which the ~2 kΩ virtual-resistance ride magnifies to SWR ≈ 1.4 at the input. A high-Q match amplifies every small difference, in models and in hardware alike: it’s the same reason the physical version of this tuner needs re-dipping when the feedline is re-routed or the ground dries out. Nudge series_c1_pF a hair off stock and watch how fast the notch walks away — then walk it back, which is precisely the bench experience.

The easy case: a big loop and an L-network

loops.skyloop_lmatch is the opposite corner: an 80 m full-wave triangular loop (~85 m of wire) worked on 17 m, where its perimeter is ~4.7 λ and the corner feed sits around 225 − 70 j Ω. That’s a moderate mismatch, not a desperate one, and a two-element L-network handles it: series inductor to the source, shunt capacitor across the feed.

Same stock coil Q of 200, very different bill: the L-match runs modest circulating current, so the coil burns only ~1 % and the design opens at SWR ≈ 1.15 in the workbench. Comparing the two budget readouts side by side is the cleanest illustration in the catalog of why “it matched” isn’t the same claim as “it matched cheaply”: the SWR meter reads ~1 in both shacks while the tuner eats nine times more of your power in one of them.

Authoring one

If you’re writing your own design, an off-band design is three decisions:

class Builder(SomeAntenna):
    default_params = MappingProxyType({
        **SomeAntenna.default_params,
        # 1. Keep the inherited design_freq (the band it's CUT for)
        #    and set freq to the band it's WORKED on.
        "freq": 24.9,
        # 2. Give the matchbox coil a real Q — an ideal matchbox
        #    dissipates nothing and hides the whole power budget.
        "coil_q": 200.0,
        ...
    })

    def build_network(self):
        return Network(
            ports={"feed": PortOnWire("feed"), "in": PortVirtual("in")},
            branches=[
                # 3. The matchbox is a station-stdlib composite, so its
                #    budget rows group under the instance name.
                Instance("tuner",
                         t_network_tuner(c1_pF=..., c2_pF=..., l_uH=...,
                                         ql=self.coil_q or None),
                         rig="in", out="feed"),
            ],
            sources=[Driven(port="in", voltage=1 + 0j)],
        )

The workbench does the rest from the two frequency defaults: it opens the design with the geometry on design_freq, the dial on freq, and the lock open. Tune your stock matchbox values at freq — that’s where the design will be judged — and remember the T-match lesson above: the higher the virtual resistance your match rides, the more sensitive the stock tune is to everything, so quote the reference you tuned against. A ui_params["budget_labels"] map turns the structural row names into friendly ones ("tuner: Shunt m" → "shunt coil").

The network vocabulary (ports, branches, boxes) is introduced in Station modelling; the full watt-by-watt methodology is worked in Coax vs. ladder line.