S21 Choke Measurements: When the Series Model Breaks

My objection is not to the series-through equation. It is to treating a smooth S21 curve as proof of a choke’s universal common-mode rejection. G3TXQ’s practical series method and K6JCA’s later fixture analysis help show why the network behind that curve matters.

What S21 Actually Measures

S21 is the complex forward travelling-wave ratio from port 1 to port 2 for the entire network between calibrated reference planes. It contains magnitude and phase. It is not inherently “common-mode rejection,” and its dB magnitude alone is not impedance.

A full description includes:

• port reference impedances and topology;
• calibration type and reference-plane locations;
• fixture compensation or de-embedding;
• the exact common-mode connection of the DUT;
• complex S11, S21, S12 and S22 where required; and
• instrument dynamic range, noise floor and uncertainty.

The Valid Two-Terminal Series Equation

For one two-terminal impedance ZDUT in series between two equal real port impedances Z0:

port 1, Z0 ──── ZDUT ──── port 2, Z0

This is the series-through measurement approach used by G3TXQ to obtain a choke’s resistance and reactance from vector transmission. K6JCA’s account of that method credits G3TXQ and writes its conversion compactly as ZDUT = Z0 × ((2/S21) − 2). It is the same equation above, not a competing formula or a different factor of two.

Keysight's current impedance-measurement guidance gives the same series-through conversion and explicitly calls for calibration or fixture compensation at the DUT terminals.

The conversion must use complex linear S21. Substituting a negative dB number into the equation is invalid because phase has been discarded.

Not Every Parasitic Invalidates the Formula

Terminal-to-terminal winding capacitance does not by itself create a measurement error. If the finished choke is a two-terminal black box and all of its inductance, resistance and terminal-to-terminal capacitance lie between the two reference nodes, the series formula returns the total two-terminal impedance of that assembly. That is normally the quantity wanted.

For example, terminal-to-terminal capacitance in parallel with the ferrite winding changes the finished choke impedance and self-resonance. It is real DUT behaviour, not a fixture error to be removed.

The simple model breaks when additional terminals or paths matter:

• capacitance from port 1 or port 2 to VNA chassis, bench or a ground plane;
• fixture shields or connector bodies completing another return path;
• radiative or conductive leakage bypassing the intended two reference nodes;
• mutual coupling to nearby conductors not included in the intended DUT boundary;
• electrically long structure that cannot be represented by one lumped impedance; or
• measured transmission approaching the residual isolation or noise floor.

Calibration corrects systematic error to the reference planes. It does not change a three-terminal object into a two-terminal one.

Why a Few Picofarads Can Matter

A shunt capacitance to the common reference has reactance magnitude 1/(2πfC). For 2 pF:

Whether that capacitance is an error depends on its endpoints. Across the DUT terminals it belongs to the finished choke. From a port node to an unintended common reference, it makes the network three-terminal and contaminates a simple series extraction.

The problem is not automatically “higher frequency.” It begins wherever fixture paths become comparable with the impedances and accuracy being claimed.

Fixture Leakage Can Produce a Beautiful Floor

As DUT impedance rises, intended series-path transmission falls. Eventually direct electromagnetic coupling, cable leakage, adapter isolation or receiver noise can dominate. The trace may remain smooth and repeatable because the leakage mechanism is stable.

Check by:

• replacing the DUT with an open circuit and comparing the apparent transmission;
• changing fixture spacing and orientation without changing the DUT;
• using absorptive or shielded separation where appropriate;
• measuring multiple known standards across the claimed impedance range; and
• reporting a usable range rather than plotting beyond the validated floor.

S21 dB Is a Fixture Result, Not Intrinsic Rejection

Even when the ideal series equation is valid, a 50 Ω/50 Ω fixture fixes a particular voltage divider. For a 1 kΩ real impedance, it yields approximately −20.8 dB transmission. That is not universal installed current reduction.

With an unchanged external-mode Thevenin source and path:

A fixture can therefore measure the choke impedance accurately while its raw transmission dB remains unrelated to an antenna's suppression. Convert to a portable component model first, then state the installed circuit assumptions.

What Y21 Can Correct

For equal real port references, a complete two-port S matrix can be converted to an admittance matrix:

Y = (I − S) × inverse(I + S) / Z0

For a lumped reciprocal π network with a series branch impedance Zseries between the two port nodes and independent shunt admittances from each node to the common reference:

Y21 = −1 / Zseries

so:

Zseries = −1 / Y21

This is genuinely useful: independent port-to-reference shunts appear in the diagonal admittances and do not change the ideal off-diagonal series term.

That is the specific problem addressed in K6JCA’s explanation of the Y21 method, which credits Dick Benson, W1QG, with introducing the approach. K6JCA compared the G3TXQ series extraction with the π-network extraction: small port-to-reference capacitances gave similar results, while increasing those shunts in a simulation made the series-only result diverge. The point is not that G3TXQ’s equation is wrong. It is that applying it to a network with significant independent shunts asks it to solve a different circuit.

What Y21 Cannot Correct

Y21 is model-based, not magic. It does not automatically distinguish the wanted choke path from every other path between the two port nodes. Direct port-to-port capacitance or coupling appears in the transfer admittance and is part of Y21. A distributed, non-reciprocal or multi-conductor structure may not have one unique lumped series branch.

Y21 also does not fix:

• bad calibration or wrong reference planes;
• fixture leakage that bypasses the intended DUT but still couples port to port;
• insufficient dynamic range;
• an incorrectly configured common-mode connection; or
• large-signal voltage, current, nonlinearity and heating.

Validate the π model, inspect reciprocity such as Y12 versus Y21, and perturb the fixture geometry. If the extracted series impedance changes when an unintended shunt alone changes, the model or calibration is incomplete.

A Full Two-Port Measurement Really Is Required

Converting S to Y generally requires S11, S21, S12 and S22. A transmission/reflection instrument that measures only one forward direction does not directly supply the complete matrix. Reverse measurements can be merged only with controlled port reversal, calibration and reference planes.

The Y21 measurement guide gives the practical workflow. The important result is complex R, X and |Z|, not a slogan that one parameter method “loses” a fixed dB amount.

Current Injection Answers a Different Question

A defined external-current fixture can test the system quantity directly: how much current changes when the choke is inserted into that cable setup. This is not the same as extracting the choke's small-signal impedance.

The current IEC 61000-4-6:2023 standard is an equipment-immunity method for conducted RF disturbances. Its controlled injection devices, reference conditions and monitoring illustrate why common-mode fixtures require careful return-path definition. A home-built choke fixture inspired by it is not automatically an IEC-compliant test.

RF source ─ injection device ─ defined cable/return path ─ choke ─ termination
                                      │
                              calibrated current probe

For coax, a probe around the complete cable responds to net longitudinal current. The intended internal equal-and-opposite mode largely cancels in the probe aperture.

Current-Probe Corrections

A current probe produces voltage according to its transfer impedance Zt. In logarithmic units:

I[dBµA] = Vreceiver[dBµV] + path-loss corrections − Zt[dBΩ]

For a relative before/after comparison at the same position, many constant corrections cancel. For absolute current, use the probe calibration over frequency and include cable loss, attenuators, receiver impedance, linearity and uncertainty.

Probe placement matters because external current forms a standing-wave distribution. A measurement centimetres before and after a choke is not automatically the choke's intrinsic “attenuation”; those are different positions in a distributed circuit.

Why Two Probes Are Not Automatically Cleaner

Using one probe on each side can be convenient for monitoring, but simply subtracting Iin and Iout does not remove the fixture. The probes have different transfer impedances and loading, and the cable section can radiate, couple or store energy between the sample points.

A defensible two-probe ratio requires:

• individual calibration and cross-checking of both probes;
• fixed cable geometry, termination and return path;
• defined sample planes and a reference measurement without the DUT;
• proof that source control has not changed the comparison quantity; and
• uncertainty that includes probe loading and fixture repeatability.

For many ham comparisons, one calibrated probe used at the same physical plane before and after insertion is easier to interpret, followed by a multi-position scan to see how the complete current distribution moved.

Use the Wording That Matches the Test

A bare claim such as “35 dB common-mode rejection” omits the essential model.

Validation Workflow

1. Define the DUT boundary. Include intended winding capacitance and exclude unintended bench paths.
2. Calibrate at the terminals. Use an appropriate full two-port calibration or validated compensation.
3. Measure standards. Confirm the claimed impedance range with known opens, shorts and impedances.
4. Acquire the full complex matrix. Preserve all four S-parameters when using Y conversion.
5. Test the model. Compare series S21 and Y21 results, reciprocity and fixture perturbations.
6. Establish the leakage floor. Use an open and geometry changes rather than trusting a smooth trace.
7. Perform current verification. Use a defined return path and calibrated probe.
8. Check the installed system. Scan multiple coax positions on every relevant band and verify voltage and temperature separately.

The Practical Verdict

I would keep the G3TXQ-style series measurement where the calibrated network really is a two-terminal DUT over the range in question. Where independent port shunts matter and a lumped π model fits, I would use the full-matrix Y21 extraction: it solves that particular fixture problem instead of hiding it inside the choke’s reported impedance.

That is why an S21 plot can look entirely convincing and still answer the wrong question. Recover the choke impedance using the circuit actually present; then use that impedance in the antenna’s current path. Neither a beautiful trace nor its dB label is a portable promise of common-mode suppression. Installed current, voltage and thermal checks still determine whether the chosen choke does the job.

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