G3TXQ Common-Mode Chokes Revisited: What the Chart and Y21 Really Show

The historical G3TXQ common-mode-choke page did something important: it compared measured complex impedance across 1–30 MHz and showed that material, core count, turns and geometry change the useful band. Its colour bars reported impedance-magnitude ranges; black bars marked frequencies where the measured series resistance exceeded the magnitude of reactance. Those data remain a useful starting point for reproducing the named constructions.

What the Historical Work Establishes—and What It Does Not

Hunt documented a VNA2180, a short two-BNC fixture with clips, a through calibration made with the clips shorted, and complex S21 conversion. He estimated the jig’s parallel capacitance at 0.2 pF or less. His accompanying derivation of R and X from S21 models the choke as a series impedance between two 50 Ω ports. That derivation includes both 50 Ω terminations and is algebraically correct for that circuit.

The published page does not provide traceable calibration standards, raw Touchstone data, a full uncertainty budget, a de-embedded fixture model, sample-to-sample statistics or large-signal thermal tests for every charted build. Consequently, the bars document those measured examples; they do not certify every nominally similar toroid, winding, coax type or power level.

Start with the Complex Quantity

Magnitude predicts current reduction only after the rest of the common-mode circuit is known. Phase matters because choke reactance can add to or cancel the loop reactance. Resistance can damp a resonance, but it is also the small-signal term associated with dissipation: for sinusoidal common-mode current, the first-order loss is P ≈ I_CM,RMS²R_CM. Choke voltage and insulation stress can still be large, approximately I_CM,RMS|Z_CM| in the lumped model.

That is why “always prefer resistive” and “only |Z| matters” are both incomplete. A broad, high minimum impedance with controlled phase, acceptable loss and thermal margin is often desirable. The required balance depends on the actual source, return path, placement, operating band and common-mode drive.

Two Correct Conversions—and Their Boundaries

The pure-series S21 result

For one complex impedance Z between equal real reference impedances Z₀, with no other path between or around the ports:

S21 here is the calibrated complex linear quantity, not only its magnitude in decibels. With 50 Ω ports, the coefficient is 100 Ω. Hunt’s magnitude-and-phase derivation is this same result expressed in components. Keysight’s VNA documentation gives the equivalent transmission-impedance conversion for arbitrary port impedances and reduces to the same formula for equal 50 Ω ports.

The full S-to-Y conversion

For equal real Z₀, convert the complete calibrated two-port matrix—not S21 alone:

For a reciprocal π network with a physical series-arm impedance Z_s, the transfer admittance is Y21 = −1 / Z_s, so Z_s = −1 / Y21. Independent shunt admittances from either port to the fixture reference do not appear in Y21. That is the practical reason this representation can outperform an S21-only pure-series extraction when port shunts matter.

For a pure series DUT, the complete Y conversion and the correct S21-only equation return the same Z. Y21 is therefore not the only mathematically valid method. It is a useful model choice that separates independent port shunts and exposes all four measured S-parameters.

Where the 6.02 dB Question Actually Comes From

There is no inherent 6 dB loss in Y21. There is, however, a factor of two in the matched two-port conversion. At high |Z|, the pure-series relation approaches S21 ≈ 2Z₀/Z. Replacing 2Z₀ with Z₀ underestimates |Z| by two, which is 20 log₁₀(2) = 6.0206 dB.

That error can arise when a one-way voltage-divider intuition is mixed with power-wave S-parameter definitions, or when linear S21 and S21 in dB are confused. It is not present in Hunt’s published formula, in Keysight’s documented transmission conversion, or in the complete matrix conversion to Y. A claimed 6 dB disagreement should therefore be audited by writing both complex equations, port impedances, calibration normalization and reference planes—not resolved by declaring one method immune.

The Fixture Decides What “The Choke” Means

A 50 Ω VNA measures waves at its calibrated ports. The fixture must launch the current mode whose impedance is being claimed. For a coax choke, that target is normally current on the cable exterior; the interior differential transmission mode should not be silently mixed into the result. Document exactly which conductors are bonded at each DUT terminal, how connector shells and the bench return are arranged, and where the reference planes lie.

NIST’s RF measurement guidance distinguishes calibration at accessible coaxial planes from fixture de-embedding to the device planes and notes that accuracy depends on the standards or fixture model. IEEE 370-2020 likewise treats fixture design, consistency and measured-data quality as part of electrical characterization. Those principles remain useful at HF even though IEEE 370 addresses high-frequency interconnects rather than amateur-radio chokes.

1. Calibrate at or move to the DUT planes. Record calibration method, port impedance, cables, adapters and any port extension. A through normalization helps, but it does not characterize every changed fixture field.
2. Measure all four complex S-parameters. Convert the calibrated matrix to Y with software whose convention is documented. Check reciprocity; a large S12/S21 disagreement is a diagnostic, not a value to average away.
3. Characterize residual paths. Measure through, open and representative short/fixture states; move cables and nearby metal. A direct capacitive bridge can dominate when Y21 is very small.
4. Control dynamic range. High impedance produces low transmission. Verify receiver noise floor with source power, IF bandwidth and averaging chosen to stay linear and repeatable; do not heat or bias the DUT during a small-signal sweep.
5. Repeat the build. Change fixture orientation, reconnect the DUT and measure multiple cores/windings. Report spread and a usable upper-impedance boundary instead of a smooth line beyond the evidence.
6. Validate the installed system. Measure exterior current before and after the choke at relevant locations. The bench impedance is a component result, not proof of station-level suppression.

As |Y21| approaches the measurement residual, inversion makes a small absolute admittance error a large impedance error. The right response is an uncertainty or reporting limit—not ever larger plotted kilohms.

How Much Choking Is Required?

There is no universal “20 dB choke.” In a simplified lumped loop with existing complex impedance Z_loop and an added series choke Z_choke, the common-mode current reduction is:

The simplified expression 20 log₁₀(1 + Z_choke/R_loop) is valid only for positive real loop and choke impedances. If both are assumed real and Z_loop = 100 Ω, its conditional examples are:

A real feedline exterior, mast, antenna, ground, station wiring and nearby conductors form a distributed network. Z_loop is not generally 100 Ω or purely real, and choke placement changes the network. Specify the required result—exterior-current reduction, pattern stability, received-noise reduction, equipment immunity or accessible-voltage control—then measure it across the operating conditions.

Ferrite, Turns and Power Are a Coupled Design

Fair-Rite describes suppression ferrite with complex permeability and a frequency-dependent series R+jX. Its selection guidance explicitly includes source/load impedance, frequency, geometry, field strength and temperature. Current material pages also publish small-signal conditions and temperature information; those curves are not transmitter-power ratings.

• Material and geometry: mix number alone does not define impedance. Core dimensions, number of cores, conductor position and winding geometry matter.
• Turns: low-frequency inductance and impedance often rise approximately with N² in the lumped small-signal region. More turns also add conductor length and inter-turn/environment capacitance, moving resonance and sometimes reducing upper-band impedance.
• Heat: resistive common-mode impedance dissipates I_CM,RMS²R. Temperature and field can change permeability and impedance; Fair-Rite warns that suppression performance derates with temperature and bias.
• Transmitter stress: ideal equal-and-opposite differential coax currents largely cancel core flux. “Transmitter watts” still cannot rate the assembly: imbalance/common-mode current, frequency, modulation, mismatch, duty cycle, ambient, cooling, coax loss, connectors, bend radius and RF voltage all matter.

Test the exact assembly at low power first. Then raise power in controlled steps with the representative band, waveform, mismatch and duty cycle while observing core, cable and connector temperatures remotely. Stop at the lowest qualified component limit. “No saturation” is not a thermal design method, and Curie temperature is not a permissible operating target.

Air-Core Coax Chokes Are Real Distributed Components

A coax solenoid can be useful, particularly as a measured band-specific solution. Well below its first important self-resonance, a lumped approximation Z ≈ R + jωL may be adequate. It is not purely inductive or lossless: coax conductor and dielectric loss, connectors, distributed capacitance, turn coupling, radiation and the installation all remain.

Above or near resonance, impedance does not simply rise linearly with frequency. Geometry and nearby objects can move the response; a sharp bench peak may not survive mounting. An air-core design avoids ferrite permeability drift and ferrite core loss, but it does not avoid cable heating, insulation voltage, connector limits or thermal change. Low-band size and achievable bandwidth are design-specific—not a universal failure of the topology.

Measure an air-core choke with the same calibrated, uncertainty-aware process and verify it in place. Y21 can help separate independent port shunts, but it cannot remove direct electric-field coupling or turn a distributed resonator into an intrinsic lumped inductor.

What a Reproducible Modern Chart Should Publish

• manufacturer, material, part number, lot where relevant, dimensions, cable, connectors, turns, winding pitch and photographs;
• fixture drawing, common-mode terminal definition, calibration/de-embedding method, reference planes, port impedance, source power, IF bandwidth and averaging;
• the four complex S-parameters or Touchstone file, the exact conversion and the sign/port convention;
• R, X and |Z| with a repeatability/uncertainty band and a declared high-impedance reporting limit;
• sample spread, cable-position sensitivity and environmental sensitivity;
• large-signal current, voltage and temperature results with frequency, waveform, mismatch, duty cycle, ambient and stop limits;
• installed exterior-current validation, kept separate from intrinsic component characterization.

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