Coaxial Stub Isolation at 1.5 kW: Design, Measurement and Safety

Michel Spelier, ON7EH/OP3T, developed a two-stub network after a multi-operator IOTA station experienced interband interference while running roughly 1 to 1.5 kW from closely spaced antennas. The station already used band-pass filters and a triplexer. The stub adds targeted rejection on the remaining troublesome band pairs; it does not replace the rest of the isolation system.

Where the Stub Fits

A stand-alone 20 dB notch is nowhere near enough to protect a receiver connected to a nearby antenna from a 1.5 kW transmitter. It can still be decisive after the antennas, band-pass filters, triplexer and switching system have already supplied most of the required isolation.

Twenty decibels corresponds to a 100:1 power ratio; 25 dB corresponds to about 316:1. If the unwanted signal arriving at the stub is −10 dBm, another 23 dB ideally reduces it to −33 dBm. That arithmetic is valid for the measured path. It does not prove that every passive element’s catalog number can be added without accounting for mismatch, interaction and alternative coupling paths.

Topology and Measurement Planes

The double-stub network is a through coaxial path with two branches connected in shunt at a tee: one branch is short-circuited and the other is open-circuited. At the wanted frequency both branches transform toward high impedance at the junction. At selected unwanted frequencies one branch transforms toward a low shunt impedance and reflects energy away from the through path.

Three different quantities are useful, and they must not share one unlabeled “isolation” number:

An “additional 23 dB” result is strongest when it is the ratio between two end-to-end station measurements made at the same planes, frequency, antenna geometry and switch state, with only the installed stub state changed. A stand-alone S21 notch can explain that improvement, but it does not include alternate coupling around the coaxial path.

Why Isolation Numbers Do Not Always Add Cleanly

A first planning estimate is often written as:

The approximation assumes that all terms refer to the same unwanted frequency and the same forward signal path, and that each device sees the impedance used when its rejection was measured. Real filters reflect much of their stopband power. Cascading reflective networks creates multiple reflections, so their combined response depends on all relevant S-parameters and the source and load impedances—not only separate S21 magnitudes.

Keysight’s network de-embedding note explains why a two-port is defined by S11, S21, S12 and S22 and why cascaded networks are analysed through signal-flow or transfer matrices. The most dependable contest-station number is therefore the measured transmitter-output-to-receiver-input coupling with the complete installed chain in place.

The sum also stops improving when another path dominates:

• antenna-to-antenna field coupling;
• relay, switch or triplexer leakage;
• common-mode current on feedline exteriors;
• cabinet, control-cable or mains coupling;
• transmitter noise already inside the victim band; or
• receiver-generated products after front-end overload.

The 1.5 kW Budget Is Arithmetic, Not a Receiver Specification

For 1 kW and 1.5 kW:

The field case used S9+40 dB as an operational marker. The familiar IARU recommendation places S9 below 30 MHz at 50 µV into 50 Ω, equivalent to −73 dBm; a nominal “40 dB over S9” is therefore −33 dBm. The linked IARU Region 1 HF Manager’s Handbook remains useful context, but an S-meter indication is not a receiver-protection limit.

That table says only what isolation reaches the chosen marker. It does not say that −33 dBm is safe for a particular transceiver. Blocking, reciprocal mixing, ADC headroom, preselector state, AGC behavior and protection circuitry differ substantially. A receiver may tolerate more at one frequency and suffer measurable desensitization from less at another.

The right acceptance test keeps a calibrated weak wanted signal at the receiver while a second source applies the blocker. Measure sensitivity, SINAD, decoded error rate or another defined performance metric. Rohde & Schwarz’s blocking-test overview uses the same principle: receiver performance is evaluated in the presence of a controlled wanted and interfering signal.

Worked Example—With the Assumptions Visible

Suppose an end-to-end measurement without the stub shows 72 dB coupling loss from a +61.76 dBm transmitter to the receiver input:

If inserting the installed stub changes that same end-to-end measurement by 23 dB at the exact interfering frequency:

This is a valid before-and-after system result if the receiver port, antenna geometry, switch state, transmit frequency and power remain controlled. It is not valid to construct the 72 dB from optimistic catalog maxima measured under unrelated terminations.

How the Two 20-Metre Stubs Create Three Stopbands

The field design uses one short-circuited quarter-wave line and one open-circuited half-wave line at the 20-metre design frequency f₀. For an ideal lossless line:

At f₀, both lines transform their terminations toward a high impedance at the tee, so the wanted 20-metre path is disturbed relatively little. At selected harmonic ratios, one stub transforms toward a shunt short:

The physical lengths are shorter than free-space values by the cable’s phase velocity, and the velocity factor is frequency- and construction-dependent. The tee, connectors, end termination and trimming method add electrical length. Real cable loss prevents an infinite passband impedance or a zero-ohm stopband, which is why the final network must be tuned and swept as assembled.

For a first cut using a cable velocity factor VF:

These are electrical-length estimates, not cut dimensions. Start long and trim the assembled network while measuring from its final connectors. The first-order relationship Δf/f ≈ −Δl/l shows why a small electrical-length error moves the notch center in the opposite direction. Phase-velocity tolerance, connector repeatability, bending and temperature also move it. IEC 61196-1-108:2025 provides current coaxial-cable test methods for phase, group delay, propagation velocity and electrical length; IEC 61196-1-111:2024 covers phase stability with temperature, bending and twisting.

Notch depth and notch width trade against loss and loaded Q. Cable loss, tee loss and imperfect terminations limit the deepest rejection; intentional or incidental damping can broaden a notch while reducing its peak depth. Record the full response for the worst cable temperature and assembly state instead of trimming only the center marker.

Interaction with Band-Pass, Cavity and Triplexer Networks

Band-pass filters, cavity filters, triplexers and shunt stubs are generally reflective in their stopbands. The impedance presented by one network changes the response of the next. Cable length between them can rotate the reflected impedance, moving a notch, creating ripple or increasing voltage at a connector even when each device looks satisfactory in a separate 50 Ω test.

Characterize the intended order and connecting cables as one multiport or cascaded network. Measure all four two-port S-parameters of each passive stage, then verify the installed chain with the real source and load. In the wanted band, check insertion loss, return loss and heating. In every unwanted band, check minimum rejection and the voltage/current stress created by mismatch.

A cavity or BPF should remain the broad source- or receiver-selectivity layer; the stub adds targeted rejection where the band geometry supports it. If the stub is switched or bypassed, every state—including transition and unpowered states—must preserve a defined 50 Ω path or a safely inhibited transmitter.

Protection Works in Both Directions—but Not for Every Interferer

Installed in the 20-metre branch, the reciprocal network reduces coupled 40-, 15- and 10-metre fundamentals entering the 20-metre receiver. When the 20-metre transmitter is active, it also attenuates its output near 7, 21 and 28 MHz before the common antenna system.

The 28.2 MHz stopband is relevant to the second harmonic of a 14.1 MHz transmitter. The stub can supplement the transmitter’s normal output filtering, but it should not be the only evidence of spectral compliance. Check the transmitter and amplifier with a suitable sample, attenuation chain and spectrum analyzer under load.

A 20-metre receive-path filter cannot remove interference that already lies inside its passband. The second harmonic of 7.05 MHz is 14.10 MHz. To the 20-metre branch that energy is indistinguishable by frequency from a wanted signal and must be reduced at the 40-metre source, by antenna isolation, or both.

The same limitation applies to broadband phase noise and switching products generated inside the victim band. A narrow victim-side notch cannot repair source noise that shares the wanted frequency range.

What “At 1.5 kW” Must Prove

In an ideal matched 50 Ω line carrying 1.5 kW of sinusoidal RF:

These are through-line matched-load values, not the maximum stress everywhere in the stub assembly. The open end can support a voltage maximum; the shorted end supports a current maximum. Tee geometry, mismatch, finite stopband power, cable loss and resonance can increase local heating or electric-field stress. An unspecified “10 mm coax” or an N connector family name does not establish a rating.

A small-signal VNA sweep proves linear S-parameters at milliwatt power. It cannot by itself prove:

• connector voltage and current margin;
• dielectric heating or loss at 1.5 kW;
• temperature drift of notch frequency;
• corona or arcing at an open end;
• heating at the short-circuit bond;
• performance at the actual duty cycle and ambient temperature; or
• survival under the worst passband VSWR magnitude and phase.

A credible high-power statement needs the exact cable and connectors, assembly geometry, frequency, load, VSWR, waveform, average power or PEP definition, duty cycle, key-down duration, ambient temperature and acceptance limit. Monitor the cable, tee, open end, short and connectors until thermal equilibrium or through a clearly defined intermittent cycle.

IEC 61169-1-6:2022 provides RF-power test methods for connectors at specified frequency, temperature and altitude. IEC 61196-1-119:2023 provides RF power-rating and power-withstand methods for coaxial cables and cable assemblies under the same kinds of declared conditions. Use the lower applicable limit of the cable, every connector, tee, switch, termination and enclosure—not a connector-family or cable-diameter assumption.

Switching and Interlocks Must Fail Safe

Never change a stub, cavity, BPF, triplexer or antenna path while RF is present unless the complete switch assembly has a documented hot-switch rating for the actual frequency, forward power, mismatch, duty cycle and load. Contact bounce or an open transition can expose a relay gap, receiver port or stub end to far higher voltage than the matched-line value.

A safe switched system inhibits keying before any relay moves and restores transmit enable only after the commanded state is confirmed. Use break-before-make sequencing where the topology requires it, auxiliary contact or independent position feedback, and a defined unpowered state. A stale, contradictory or missing state must leave the transmitter inhibited.

• Interlock the amplifier and exciter, not only a software band label.
• Prevent two transmitters from selecting incompatible shared paths.
• Include relay transition time, bounce, control loss and restart behavior.
• Provide a rated load or isolated safe state where the switching system requires one.
• Recheck receive-path protection and transmitter output match in every valid state.

IEC 62368-1:2023 treats safety by identifying energy sources and providing safeguards for ordinary, instructed and skilled persons. For this RF assembly, enclosure, access control, interlocking, earthing and fire/thermal containment are part of the system—not optional accessories to the coaxial design.

DC Ground Is Not Automatically Lightning Protection

A shorted quarter-wave stub provides a DC continuity path while transforming toward high impedance at its RF design frequency. Purpose-designed quarter-wave coaxial surge-protective devices do exist. The current IEC 61169-1-3:2026 test standard explicitly includes quarter-wave short-stub devices among coaxial SPDs and defines surge-withstand performance requirements and test methods.

IEC 61643-21:2025 gives current requirements and test methods for SPDs connected to telecommunications and signalling networks. IEC 62305-3:2024 covers the structural lightning-protection system and protection against touch and step voltages. A coaxial SPD must be selected and coordinated within that larger site design.

A Safe Station Validation Sequence

1. Finish low-power characterization first. With rated 50 Ω terminations, calibrate at the stub connectors and record S11, S21, S12 and S22 across the wanted band and every stopband.
2. Verify every switch state without RF. Confirm relay position feedback, unpowered behavior, interlock truth table and continuity to the intended load before enabling an exciter.
3. Measure installed paths at low level. Record transmitter-sampling-plane-to-receiver-input coupling for every band pair, antenna direction and stable switch state.
4. Separate interference mechanisms. Measure fundamentals, harmonics, discrete spurs and broadband transmitter noise with an appropriately rated sampler and attenuation chain.
5. Test receiver blocking. Maintain a calibrated weak wanted signal while increasing a controlled interferer; record sensitivity, SINAD, decode rate or another declared degradation metric.
6. Enclose and clear the high-power area. Guard exposed conductors and the open stub end, establish the required RF-exposure exclusion area, use remote temperature and RF sampling, and prevent access while energized.
7. Raise power progressively. Confirm linear scaling, stable match and correct interlock action before approaching the authorized station power. Stop on unexpected reflected power, non-linear leakage, odor, sound, arcing or abnormal temperature rise.
8. Run the thermal condition. Use the intended frequency, waveform, average power or PEP, duty cycle, ambient, key-down time and worst declared mismatch. Do not infer temperature by touch.
9. De-energize before adjustment. Inhibit and physically isolate all transmitters, follow the applicable hazardous-energy procedure, verify the safe state and allow stored energy to discharge before opening or trimming the assembly.
10. Repeat after stabilization. Re-sweep notch centers, passband return loss and minimum rejection; repeat relevant states with practical cable routing, weather exposure and antenna rotation.

The ICNIRP 2020 RF guidelines cover human exposure from 100 kHz to 300 GHz. Apply the locally required exposure limits, assessment method and access controls; the power at the transmitter connector does not by itself define a safe distance.

Bottom Line

A shorted quarter-wave and open half-wave section referenced to 20 metres can add rejection near 40, 15 and 10 metres while leaving 20 metres usable. An installed 20–25 dB improvement can be extremely valuable after the station already provides roughly 70 dB of installed isolation.

The operating decision is conditional. The stub’s rejection must be measured across each operating segment, the complete station must be tested for receiver desensitization, and 1.5 kW power handling must be demonstrated separately. The system result—not the depth of one VNA marker—is what decides whether the contest position is protected.

Questions, antenna-factor records or height trials to share? Contact RF.Guru.