Why S21 Can Look Right for Chokes but Still Be Wrong

The short version is this: S21 is not the common-mode rejection of a choke. It is the forward transmission through a particular two-port fixture, usually with two 50 ohm VNA ports. That fixture may be useful, and at lower frequencies it may even correlate well with the choke’s common-mode impedance. But as frequency rises, the fixture, parasitic capacitance, return path, direct coupling, connector bodies, and cable geometry increasingly become part of the measurement.

A VNA S21 measurement is a measurement of forward transmission from port 1 to port 2 in a defined RF network. S-parameters are complex quantities, with magnitude and phase, and they describe reflection and transmission between ports of a network. They do not automatically describe the behavior of a real cable’s common-mode current path.

What the Simple S21 Method Assumes

The classic two-port S21 choke test assumes the device under test is a single series impedance between two ideal 50 ohm ports:

VNA port 1        DUT/common-mode choke        VNA port 2
  50 Ω   ─────────────── Zcm ───────────────   50 Ω

For that ideal circuit:

S21 = 2Z0 / (2Z0 + Zcm)

and therefore:

Zcm = 2Z0 × (1/S21 - 1)

With Z0 = 50 Ω, this becomes:

Zcm = 100 × (1/S21 - 1)

This is why the method can look convincing. If the choke really behaves like a lumped series impedance, the fixture is short, the calibration is good, and parasitic capacitance is small, the result can be plausible. The commonly used G3TXQ-style S21 formula is often written as Zdut = Z0 × ((2/S21) - 2), which is the same expression for equal 50 ohm ports.

But that derivation has a hidden condition: the DUT must be the only relevant RF element in the path.

In reality, the measurement is closer to this:

                 Cstray1                Cstray2
                   │                      │
VNA port 1 ────────┴──── Zcm ─────────────┴──── VNA port 2
  50 Ω                 fixture, leads, core, cable       50 Ω
                   │                       │
                VNA body / bench / ground / air

The moment shunt capacitance, lead inductance, cable length, fixture fields, connector bodies, or bench coupling become significant, the simple S21-to-impedance formula is no longer measuring only the choke.

“Rejection” Is Not an Intrinsic Property of the Choke

A common-mode choke mainly has a frequency-dependent impedance:

Zcm(f) = R(f) + jX(f)

The actual current reduction depends on the common-mode source impedance, load impedance, and return path:

Iwithout = Vcm / (Zsource + Zload)

Iwith = Vcm / (Zsource + Zload + Zcm)

So the attenuation is approximately:

Acm = 20 log10 |(Zsource + Zload + Zcm) / (Zsource + Zload)|

The VNA S21 setup silently fixes this world to:

Zsource + Zload = 50 Ω + 50 Ω = 100 Ω

But a real common-mode cable environment is not necessarily 100 ohms. It may be 20 ohms, 150 ohms, 600 ohms, capacitive, inductive, resonant, or changing with hand position, cable routing, enclosure bonding, antenna geometry, and nearby objects.

Why the Error Gets Worse at Higher Frequency

At low frequency, a few picofarads look like a very large impedance. At higher frequency, the same capacitance becomes a bypass path.

A few picofarads may seem harmless. But if the choke impedance is hundreds or thousands of ohms, that capacitance is now in the same impedance range as the DUT. At that point the VNA is no longer seeing “a choke in series.” It is seeing a distributed RF network.

This is why the simple S21 method often remains believable at lower HF but becomes questionable as frequency increases. The curve may still look smooth. It may even look textbook. But it can be the fixture behaving nicely, not the choke being measured correctly.

K6JCA’s analysis of common-mode choke measurements shows that the S21 method can be affected by parasitic shunt capacitances, while the Y21 method separates the series DUT impedance from those shunt paths in a π-network model. In practice, S21 and Y21 can agree when shunt capacitance is small, but diverge when parasitic capacitance becomes significant.

The Dangerous Part: S21 Can Look Correct

The S21 method can deceive because it has all the signs of a “good” RF measurement:

• the VNA is calibrated;
• the trace is smooth;
• the result has units or dB;
• the curve resembles ferrite behavior;
• repeated measurements agree;
• the low-frequency part matches expectations.

But calibration only corrects the measurement system to the calibration reference planes. It does not make the DUT a pure series impedance. It does not remove the physics of the fixture unless you explicitly de-embed, model, or control those effects.

Typical high-frequency failure mechanisms include:

• stray capacitance from the fixture nodes to the VNA bodies or ground;
• capacitance across the choke or winding;
• direct electric-field coupling from input fixture to output fixture;
• magnetic coupling around the DUT instead of through the DUT;
• coax shields and connector bodies becoming part of the common-mode return;
• fixture length becoming electrically significant;
• the DUT no longer behaving as a lumped component;
• S21 reaching the leakage floor of the fixture, so the measured attenuation is fixture isolation.

This is why a bad S21 method can produce a beautiful graph. The measurement is not necessarily noisy. It is worse than noisy: it is confidently measuring the wrong network.

What the Y21 Method Fixes

If the goal is to measure the common-mode impedance of the choke as a component, the better VNA method is the Y21 method.

Instead of pretending the DUT is the only series element, the Y21 method treats the measured network as a two-port admittance matrix. For equal real Z0, the measured S-parameters can be converted to Y-parameters:

Y = (I - S) × inverse(I + S) / Z0

Then, for the series branch of the π-network:

Zcm = -1 / Y21

This is the key advantage: shunt parasitics at port 1 and port 2 are represented in Y11 and Y22, while the through series branch is represented by Y21. In the lumped π-network model, Zcm = -1/Y21 extracts the series DUT impedance without being fooled in the same way by external shunt capacitances.

A practical Y21 workflow is:

• configure the choke for common-mode measurement; for a two-wire choke, tie the two wires together at each end;
• calibrate the VNA at the fixture reference planes with a proper SOLT calibration and a real thru reference;
• measure the full two-port S-parameter set: S11, S21, S12, and S22;
• convert S-parameters to Y-parameters;
• calculate Zcm = -1 / Y21;
• plot |Zcm|, Rcm, and Xcm versus frequency.

A major practical warning: the Y21 method requires a real full two-port measurement. A simple transmission-only setup that only measures S11 and S21 is not enough unless additional reversed measurements are performed and merged correctly. Many low-cost T/R VNA setups do not provide all four calibrated S-parameters directly.

The Y21 method is therefore the better answer when you want to know:

“What is the common-mode impedance of this choke?”

But it still does not directly answer:

“How much common-mode current will this choke remove in my actual cable system?”

For that, you need to inject and measure current.

The Old-School EMC Method: Inject Current and Measure Current

The EMC way is closer to the real problem. Do not infer common-mode rejection from a 50 ohm S21 fixture. Instead, inject a common-mode RF disturbance onto the actual cable, or onto a defined cable fixture, then measure the common-mode current before and/or after the choke.

That is the philosophy behind IEC/EN 61000-4-6. The 2023 edition covers conducted RF immunity from 150 kHz to 80 MHz, with product committees allowed to use the described methods up to 230 MHz in some cases. The standard also makes an important point: the method is structured for repeatability between facilities, not because real-world coupling can be determined with perfect quantitative exactness.

A practical choke test inspired by IEC 61000-4-6 looks like this:

RF generator / tracking generator
        │
 attenuator / amplifier
        │
 injection clamp
        │
================== cable under test ==================
        │
      choke
        │
 monitor current probe
        │
 spectrum analyzer / receiver

For a coaxial cable, the current probe goes around the entire coax, not only the center conductor. The differential-mode current inside the coax cancels magnetically. The current probe measures the net current, which is the common-mode current flowing on the outside of the cable shield and returning through the environment, ground plane, fixtures, and equipment capacitance.

The basic measurement is:

Acm(f) = Iref(f) - IDUT(f)

where:

• Iref(f) is the common-mode current measured without the choke;
• IDUT(f) is the common-mode current measured with the choke;
• Acm(f) is the current reduction in dB.

This method directly answers:

“How much common-mode current reduction do I get in this cable setup?”

That is much more meaningful than calling a 50 ohm S21 insertion-loss curve “common-mode rejection.”

Current Probe Correction

A current probe does not directly output current. It outputs a voltage proportional to current. You correct the analyzer reading using the probe’s transfer impedance:

I[dBµA] = V_analyzer[dBµV] + cable/attenuator corrections - Zt_probe[dBΩ]

For relative choke measurements, many correction terms cancel if the same monitor probe, analyzer, cables, and settings are used for the reference and DUT sweep. But if you want absolute current, apply the transfer-impedance correction from the probe documentation.

A Better EMC-Style Setup Uses Two Current Probes

The cleanest practical version uses two monitor probes:

injection clamp ── probe 1 ── choke ── probe 2 ── termination/load

Then you measure:

• Iin(f): the common-mode current entering the choke;
• Iout(f): the common-mode current leaving the choke.

The current reduction is:

Acm(f) = Iin(f) - Iout(f)

Even better, use software to control the generator:

• set frequency;
• adjust generator power until Iin reaches a target current;
• measure Iout;
• calculate Iin - Iout;
• step to the next frequency.

This removes a lot of uncertainty caused by changing injection efficiency with frequency.

The EMC Method Is Better, but Not Magic

The current-injection method also has pitfalls. The common-mode return path must still be defined. Cable length, cable height, ground-plane bonding, clamp location, auxiliary-equipment impedance, and termination all matter.

IEC-style testing pays so much attention to ground reference planes, coupling and decoupling networks, clamp position, common-mode impedance, and current monitoring because the common-mode path is the test.

So the EMC method does not remove all RF complexity. It simply measures the right physical quantity: current on the cable.

Recommended Wording

Do not write:

This choke has 35 dB common-mode rejection.

That sounds like an intrinsic property, and it is usually not true.

Better:

Using the Y21 method, the choke common-mode impedance is |Zcm| = 1.2 kΩ at 10 MHz, with R = ... and X = ...

or:

In the defined current-injection fixture, with this cable length, ground plane, termination, and probe spacing, the choke reduced the measured common-mode current by 28 dB at 10 MHz.

or, if you used the simple S21 method:

In a 50 Ω / 50 Ω two-port fixture, the measured S21 insertion loss was 28 dB. This is fixture-dependent and should not be interpreted as general common-mode rejection.

Conclusion

The S21 method can look correct because, under ideal low-frequency conditions, it is based on a valid equation for a series impedance between two 50 ohm ports. But common-mode choking is not a generic 50 ohm two-port attenuation problem. It is a current-flow problem in a messy RF return path.

As frequency increases, the simple S21 method becomes increasingly fragile because picofarads, millimeters, connector bodies, cable shields, and bench capacitance stop being negligible.

Use the Y21 method when you want the choke’s common-mode impedance.

Use IEC 61000-4-6-style current injection and current monitoring when you want the choke’s actual common-mode current reduction in a defined cable setup.

The simple S21 method is not useless. It is just very easy to overinterpret. Its real result is the transmission of the fixture you built, not necessarily the common-mode rejection of the choke.

Interested in more technical content? Subscribe to our updates for deep-dive RF articles and lab notes.

Questions or experiences to share? Feel free to contact RF.Guru for practical RF and antenna support.