PCB manufacturing PCB manufacturing
Home > Blog

Two-Port Shunt-Through PDN Impedance: Trust the Milliohms

November 17th, 2026

Before trusting a milliohm reading from a two-port shunt-through fixture, measure a known low-impedance reference and check the return path. A low, smooth S21 trace can include ground-loop error or fixture feedthrough. A short check helps reveal the residual, but it does not replace validation with a reference near the impedance you need to report.

Start at the intended rail measurement point, define the band and power state, then qualify the fixture before applying a target impedance.

Two-port PCB test coupon with decoupling capacitors and coaxial launches
Define the rail and return at the measurement plane. The pictured coupon is illustrative, not a fixture wiring specification.

PDN Impedance Measurement: What the VNA Sees

The power delivery network (PDN) presents impedance between a rail and its return. The result depends on the selected location, frequency and circuit state. Measuring near the regulator does not necessarily describe the path at the load. A powered device may also present a different response from an unpowered board.

A VNA sweep measures small-signal behavior around the test condition. It does not replace a time-domain load step or a DC voltage-drop measurement. State which evidence the requirement needs before setting up the bench.

2 Port Shunt Through Measurement: Convert Complex S21 to Ohms

In the ideal lumped circuit, Port 1 and Port 2 connect to the same signal node at the calibrated plane, and the DUT impedance Z connects that node to the common return. With equal real reference impedances Z0, S21 = 2Z ÷ (2Z + Z0) and Z = (Z0/2) × S21 ÷ (1 − S21). Apply the conversion to complex S21, not a dB value. With 50 Ω references, 10 mΩ gives approximately −67.96 dB and 5 mΩ gives −73.98 dB in this model.

Ideal circuit with two port signal nodes joined and the PCB rail impedance from the shared node to its return
At the ideal measurement plane the shunt DUT is a branch to the return, not a series element in the signal path.

Impedance Measurement Methods: Reflection vs. Shunt Through

Method selection for PCB impedance measurements
Method When it helps Check before use
One-port reflection Direct input impedance when complex S11 is adequately resolved. At very low impedance the reflection approaches a short; demonstrate that the setup can resolve the difference.
Two-port shunt through Milliohm-class rail-to-return measurements with suitable dynamic range. Verify shared return, ground-loop treatment, fixture residual, and a comparable reference.
Two-port series through A suitable floating device with higher series impedance. A tiny series impedance causes very little change.

For the opposite case, see the VNA high impedance measurement guide to validate a series fixture.

VNA Ground Loop: Why the Low End Can Look Wrong

Two cables can connect the board return back to the instrument through more than one shield path. Part of the source return current can flow along the receiver-side shield, adding a voltage the receiver interprets as signal. This can dominate certain low-frequency measurements even when the board is unchanged.

Compare a short and a known low-impedance reference with the manufacturer’s supported ground-loop mitigation and the approved bench setup. Use a suitable isolation or common-mode accessory only within its specified band and ratings, and include it in the qualified measurement setup. Never defeat protective earth or casually float test equipment. A change on the reference as well as on the board points to the measurement setup, not the PCB alone.

VNA Dynamic Range: Compare the Residual With the Signal

Under a simple additive error model, 10 mΩ in the ideal fixture gives |S21| ≈ 0.000400. A coherent residual of −80 dB has magnitude 0.000100, or around 25% of that signal. If it adds in phase, the converted value is about 12.50 mΩ; if it subtracts in phase, about 7.50 mΩ. These are calculated cases, not a VNA accuracy specification. Random noise and coherent errors require different treatment.

Example 10 milliohm result spans approximately 7.50 to 12.50 milliohms under a coherent residual model
A residual below the DUT trace can still change the pass/fail outcome. Do not subtract two dB magnitudes to remove it.

At 5 mΩ the same residual would be about half the nominal transmission amplitude. A narrower receiver bandwidth may improve random noise, but it does not remove a coherent shield or leakage signal. Measure repeated residual and reference traces over the entire decision band.

VNA Calibration Procedure: Verify at the Board Plane

Follow the analyzer and fixture method for reference calibration and any documented compensation, including accessories used during the test. A calibration at remote connectors does not automatically remove the board trace or probe contact. If a low-impedance reference was used as a calibration standard, use a separate characterized part to verify it afterward.

Use the fixture’s specified residual check and separate reference to confirm different failure modes. Record sample location, contact repeatability, source level, frequency grid, IF bandwidth and state. For a powered rail, confirm the method and components are rated for the bias and transients before any measurement; an unpowered sweep must not be relabeled as a powered result.

PDN Target Impedance: Apply It Only to Valid Data

Decoupling or stackup comparison: Keep the same plane, fixture, state and band on both designs. Confirm that apparent changes in the board’s peaks exceed the repeatability and residual uncertainty. If they do not, validate the measurement before crediting a layout or capacitor change.

Production or development acceptance: Compare qualified complex impedance or magnitude, as specified, against the approved band and limit. If a hypothetical upper limit is 12 mΩ, the earlier 7.50–12.50 mΩ example cannot prove a pass even though its nominal value is 10 mΩ. Where relevant, verify the actual voltage response under the approved load step as separate evidence.

Low-Impedance Test Report: What to Sign Off

Setup and release worksheet
Step Record Decision
Worked residual check 10 mΩ modeled DUT, −80 dB coherent residual, example upper limit 12 mΩ. Worst illustrated result is about 12.50 mΩ. Improve or independently bound the setup before deciding.
Fixture verification Measurement plane, return, isolation, residual traces and characterized check device. If either check fails or the result changes excessively between repeat connections, do not sign off.
Product disposition Actual state, band, uncertainty, target and independent test evidence. Use the product’s approved decision rule. No universal IPC milliohm limit applies.

Two-Port Shunt-Through PDN Impedance FAQs

Can I convert S21 in dB directly into impedance?

No. Use complex linear S21 and the correct topology and reference impedances.

Is an excellent short trace enough?

No. Also verify a known low-impedance reference near the range you intend to report.

Why is the low end sensitive to the cable setup?

Shared return paths can create a systematic receiver signal. Verify the return and approved mitigation on an independent reference.

Can I remove a ground loop by lifting protective earth?

No. Keep the safety earth intact and use an approved measurement accessory and setup.

Does a narrower IF bandwidth eliminate the error?

It can improve random noise. It does not reliably remove a coherent ground-loop or fixture signal.

Is powered behavior the same as a bare-board sweep?

Not necessarily. Record bias, the operating state and applicable instrument protection.

Can I subtract the short baseline in dB?

No, not as a general correction. Use a qualified complex calibration or model and verify independently.

Does meeting a PDN target prove the rail never droops?

No. Verify the actual load-step voltage and operating requirements separately when they govern release.

Technical References

Custom High-Speed PCBs With Power Distribution Review

Need a high-speed multilayer PCB with a power distribution network you can verify? Ask EBest Circuit to review the proposed stackup, return geometry and coupon or probe access before fabrication. Documenting those features up front helps keep the build and validation plan aligned.

Send your Gerbers, stackup, power-rail and inspection requirements, and quantities to sales@bestpcbs.com. Request a custom PCB quote and identify the rail locations your team needs to verify.

VNA High Impedance Measurement: When S11 Misleads

November 17th, 2026

When a VNA sees an almost open circuit, a small error in the measured reflection can become a large error in the calculated impedance. Start with a calibrated one-port measurement at the DUT plane. If the impedance changes too much when you repeat the setup, try a validated series-through fixture for a suitable floating component. If neither method resolves the required range, report a bound or repeat with a better fixture instead of quoting a precise value.

The goal is to distinguish the device from an open, not just to display an impedance on the Smith chart. Use complex measurements and an independent reference near the range you intend to report.

Two-port PCB test coupon with a series component between coax launches
On a series-through coupon, the component interrupts the signal path. This view is illustrative, not a fixture design specification.

VNA Impedance Measurement: What Counts as High?

High means large relative to the reference impedance and hard to distinguish with the particular fixture, frequency and uncertainty. There is no universal cutoff in ohms. A 1 kΩ resistor measured against 50 Ω already brings the reflection close to +1, but whether it is reportable depends on how accurately the actual setup can resolve the difference from an open.

Define the device, band, and reference plane. A high-impedance node connected to ground elsewhere cannot simply be placed between two VNA ports as a floating series element. A component measured in isolation is also not necessarily equivalent to its impedance in the assembled circuit.

Reflection Coefficient to Impedance: Keep the Complex Sign

For a one-port measurement with real reference impedance Z0, Z = Z0 × (1 + Γ) ÷ (1 − Γ), where Γ is complex S11 at the chosen plane. Use both magnitude and phase or real and imaginary data. In a 50 Ω system, Γ = +0.90476 represents about 1 kΩ, but Γ = −0.90476 represents about 2.50 Ω. Identical reflection magnitude cannot distinguish them.

An ideal open has Γ = +1 and an ideal short Γ = −1. Do not convert a magnitude-only return-loss chart to ohms by assuming one sign. Also check the file’s reference impedance and the actual measurement planes before comparing values from two setups.

S11 Measurement Using Network Analyzer: A Near-Open Error Example

Consider an ideal 1 kΩ resistor in a 50 Ω system. Its calculated reflection is approximately +0.90476. If the real part were read 0.01 lower or higher with no imaginary error, the conversion would give 900 Ω or 1.123 kΩ. These are hypothetical endpoints for one error assumption. Both phase and magnitude can move in a real sweep.

Reflection coefficient perturbations around 0.90476 convert to approximately 900, 1000, and 1123 ohms
Equal changes in the real reflection coefficient do not create equal changes in calculated impedance. The denominator shrinks as Γ approaches +1.

For a primarily resistive load, the local sensitivity dZ/dΓ is 2Z0 ÷ (1 − Γ)2. That explains why residual calibration errors, an imperfect open or a shifted fixture plane can dominate the reading. Near an open, admittance may be better behaved numerically, but it does not recover information your measurement did not resolve.

Reflection vs. Series Through vs. Shunt Through

Select the fixture by impedance range and the physical node
Method Use when Validate before use
One-port reflection You need input impedance at one defined port and can resolve its complex reflection. Check open, fixture and a separate known reference. Near +1, verify that uncertainty still permits the required decision.
Two-port series through The device can be inserted as an isolated series element and transmission is above the reliable floor. Check an open gap for residual feedthrough and use an independently characterized series reference.
Two-port shunt through A low impedance to the common return is the quantity of interest. Qualify return paths and fixture residuals. A very high shunt impedance barely alters transmission.

Series Through Impedance Measurement: Recalculate With S21

For equal, real 50 Ω reference impedances and a single lumped series impedance Z between the planes, S21 = 100 Ω ÷ (100 Ω + Z). Invert the complex value to obtain Z = 100 Ω × (1/S21 − 1). A 1 kΩ pure resistance gives S21 ≈ 0.09091, or −20.83 dB.

If only the transmission magnitude varies by ±0.10 dB in this ideal example, with phase fixed, the reported resistance becomes about 987–1013 Ω. Do not compare that range directly with the earlier reflection example as if the two assumed errors were measured on the same setup. At sufficiently high impedance, feedthrough, receiver noise or a parasitic path around the DUT can dominate S21. Use the measured residual and required uncertainty, not a universal impedance boundary.

VNA Calibration Procedure: Put the Reference at the DUT

Set the frequency range, power and receiver settings for the intended test. Calibrate to the needed planes or apply a justified fixture compensation or de-embedding procedure, then verify with a device the calibration did not use. A calibration at the coax faces does not automatically remove the launch, traces or gap pads. If the task is instead to verify a rail-to-return value in milliohms, use the two-port shunt-through PDN impedance guide to check the ground return and fixture residual. Make sure a recommended series or shunt workflow is supported by the instrument.

For a series fixture, check the open gap and then a characterized component near the expected impedance. Keep the fixture and component plane consistent, and inspect the residual transmission over the whole required band. If the measured DUT response approaches that residual, a simple subtraction of magnitudes is not a valid correction. For related modeling work, see the S-parameter DC extrapolation guide for why a finite-frequency result does not establish DC behavior.

Fixture Parasitic Capacitance: A High-Impedance Trap

Suppose a nominal 1 kΩ resistor has an additional 0.2 pF across it in the measurement fixture. In a simple parallel model, Z = 1 ÷ (1/R + j2πfC). At 100 MHz, that produces approximately 984 − j124 Ω. This calculation is not a claim about common fixture capacitance or a particular resistor; it shows how an unmodeled path can turn an apparently resistive result into a complex one.

One-kilohm resistor and an illustrative fixture capacitance of 0.2 picofarads in parallel across the measurement gap
Both branches are across the same two points. A real fixture may also have series parasitics and distributed coupling.

For a component tolerance check, compare impedance with a valid independent reference over the same band and bias. For a high-impedance board node, verify that mounting, nearby copper and bias conditions represent the actual use. If the intended result is DC resistance, avoid reporting a VNA-derived extrapolation as a direct DC measurement.

High Impedance Test Report: Decide Whether the Number Is Defensible

Method choice and reporting worksheet
Checkpoint Record Decision
Example one-port decision Target 1 kΩ ±5%; assumed real Γ uncertainty ±0.01 gives 900–1123 Ω. That range cannot prove 950–1050 Ω. Qualify another method with an independent reference.
Independent validation Planes, fixture ID, grounding, open-gap response, known device response, reconnection repeats and band. If the DUT is indistinguishable from the residual, qualify a bounded statement or remeasure. Do not invent precision.
Release format Complex Z or admittance versus frequency, actual limits, bias, applicable uncertainty and decision rule. Apply the agreed electrical criterion; no generic IPC requirement sets the usable VNA high-impedance range.

VNA High Impedance Measurement FAQs

Can I get impedance from S11 in dB alone?

No. Include phase or both real and imaginary parts, and use the stated reference impedance.

Why does S11 change very little near an open?

Different large impedances can all reflect nearly all the incident signal with similar phase. The inverse conversion magnifies the small differences.

Can I use series through for a grounded board node?

Not as the simple floating series fixture shown here. Use a method compatible with its actual return and bias paths.

Does a low-noise trace prove the result is correct?

No. Fixture feedthrough, reference-plane errors and calibration residuals can remain stable.

Can I remove the open fixture trace by subtracting dB?

Not as a general correction. Use the approved complex fixture model or calibration method and confirm it on an independent reference.

Does series through always improve the high end?

No. It depends on fixture isolation, usable received level, device grounding, and the chosen band.

What if a one-port conversion reports a negative resistance?

For an expected passive resistor, check residual errors, calibration and uncertainty before assigning a physical meaning to the result.

Should I report one nominal value or a sweep?

Use the specified frequency or band. Report complex impedance, the plane and uncertainty rather than extending a single value to untested conditions.

Technical References

Custom RF PCBs With Stackup and Coupon Review

Need a new RF PCB or impedance test coupon? Ask EBest Circuit to review the stackup, coax launches and coupon geometry before quoting. Marking the actual measurement plane and intended fixture early helps avoid a coupon that cannot distinguish the component from its surroundings.

Send your Gerbers, stackup, operating band, required coupon and quantities to sales@bestpcbs.com. Request a custom RF PCB quote and tell us what the coupon needs to verify.

VNA Calibration Drift: Which PCB Results Can You Trust?

November 17th, 2026

A VNA calibration pass at the start of a shift does not prove it stayed valid for every board. Measure an independent check standard at the start and at defined checkpoints. If the last check fails, stop release, use the test timestamps to identify the potentially affected boards, and restore and verify the setup before retesting. Do not subtract one reference error from all PCB measurements.

Use a check device that exercises the measurement that drives the decision. A well-matched attenuator can help track transmission loss, but it cannot automatically verify a weak reflection, a deep notch, or a different connector plane. Establish separate limits where those measurements matter.

Coaxial attenuator, RF test leads, and a two-port PCB coupon on a test bench
Verify the same measurement system that will test the boards, at the same planes and settings. The pictured layout is illustrative.

Calibration vs Calibration Verification: What Each One Proves

Calibration estimates correction terms using defined connections and standards. Verification asks whether the corrected system can measure an independent, suitable artifact as expected. A third step, tracking drift, compares repeated verification readings against an accepted baseline over time. It is possible for the first verification to pass and a later one to fail.

Compare both with the reference characterization at the start and with the baseline later. A stable but wrong reading can pass a relative drift check. A reference check also measures the combined setup: the analyzer, adapters, cables, contacts and settings. It does not prove which component caused a shift.

VNA Calibration Standards: Pick a Separate Check Device

Do not use the same open, short, load or thru assumed by the calibration as your sole evidence that the correction works. An independent two-port with characterized insertion loss and phase can check transmission. Use a suitable mismatch artifact when reflection matters, and an independently characterized lower-level path when a filter stopband approaches the system floor. A single device does not cover every error term or uncertainty.

Make sure the reference planes, connector family and orientation match the setup or are explicitly accounted for. Compare on a supported frequency grid and log the selected metrics, band and reference uncertainty. For a narrowband filter, the worst point in the passband may matter more than the center frequency.

VNA Calibration Procedure: Establish the Start-of-Run Baseline

  1. Prepare the analyzer as specified by its manufacturer. Secure the cables, identify the calibration planes, and settle the artifact and setup in the conditions specified for your test plan.
  2. Calibrate on the intended frequency grid with the required kit definition and test settings. Save the correction state and identify any adapter already included.
  3. Attach the independent standard with the approved connection practice. Verify it against its characterized data and allowed uncertainty. If it fails now, do not make it the baseline.
  4. After it passes, save the raw complex readings, the initial reference trace, temperature, time and setup ID. Check repeatability with the expected reconnection procedure when the artifact must be removed for board testing.

If the ports cannot mate, use the suitable workflow and independent check described in the Unknown Thru calibration guide before establishing this baseline.

Measurement Drift: Separate Time From Physical Movement

Leave a check artifact connected and repeat a short series without moving the setup. A gradual, reproducible shift suggests a time-dependent setup or environment effect; it does not identify the analyzer as the cause. If the shift appears only after moving a cable or reseating a connector, isolate that action using the same artifact and settings. If one artifact changes while another suitably chosen one remains stable, inspect the first artifact and verify that both checks exercise the relevant measurement.

For a phase requirement, compare the shortest signed angular difference at each frequency rather than treating +179° and −179° as 358° apart. Check the shape across the band too if a delay change matters. For amplitude, define whether the report uses signed S21 in dB or positive insertion loss. In this article’s examples, insertion loss is positive and an increase means the measured path looks lossier.

Check standard drift at 0, 30, 60, and 90 minutes; only the 90-minute point exceeds the example 0.10 dB limit
Four hypothetical checkpoint values, not a continuous drift measurement. A failed check flags the interval after the last known good check, not the exact instant drift began.

VNA Calibration Verification: Interval Checks vs. Triggered Checks

Use these checks together when the consequence of releasing a bad lot warrants it. Set the schedule from observed stability, reconnection repeatability and the size of an acceptable retest window. A vendor warm-up recommendation is not a calibrated test-window length.

Choose verification timing by the failure it catches
Check Best use Limits and next action
Start and end of run Short, stable runs with an acceptable whole-run retest cost. A failed end check can put the entire run on hold if no intermediate evidence exists.
Periodic checkpoint Long RF coupon or filter runs where an intermediate check can locate affected work. More handling and test time. Compare against a stable fixture and documented repeatability.
Triggered re-verification After an adapter change, cable movement, unexpected temperature shift, or other event that can invalidate the setup. Do not rely on this alone; gradual drift can occur without a recorded trigger.

Failed VNA Verification: Find Which Boards to Hold

Use this worksheet for a serialized lot with reliable time stamps. In the hypothetical example at 10 GHz, the verified start value is 10.02 dB, and the maximum allowed absolute change from it is 0.10 dB. At 60 minutes, the reference reads 10.09 dB; at 90 minutes, 10.16 dB. The first is 0.07 dB from baseline and still passes; the second is 0.14 dB away and fails.

Retest worksheet: VNA drift example and decision
Record Example entry Action
Start baseline 0 min: 10.02 dB; independent absolute check already accepted. Save the state and baseline. Apply the agreed 0.10 dB drift limit only to this example.
Last known good 30 min: 10.06 dB (+0.04). 60 min: 10.09 dB (+0.07). Use the 60-minute pass as the initial containment boundary only if setup and timestamps are reliable.
First failed check 90 min: 10.16 dB (+0.14); the 0.10 dB allowance is exceeded. Hold units measured after the last good check until the system passes again. Include earlier units if logs indicate intermittent failure or setup changes.
Restored setup Record adjustment, new calibration ID if applicable, independent pass and rerun results. Retest the identified units. Expand scope if timestamps or calibrated-state records do not support the boundary.

VNA Measurement Uncertainty: Leave Room at the Limit

Set the allowable drift before testing, not after seeing the results. For illustration, a drawing allows a maximum of 1.00 dB loss at one frequency. A PCB measures 0.88 dB. If the validated decision rule allows for up to 0.10 dB of drift and another bounded 0.06 dB of error from distinct contributors, a conservative additive upper result is 1.04 dB. That result cannot be released under this example. These are assumed bounds, not a universal uncertainty formula; follow the contractual decision rule and avoid counting the same contribution twice.

Measured insertion loss 0.88 dB and conservative upper result 1.04 dB exceeding an example 1.00 dB limit
The nominal reading is below the example limit, but the upper decision bound crosses it. This calculation assumes a currently valid setup.

If a verification check has already failed, this allowance is no longer evidence that the affected board result is valid. Retest on a verified setup. Also do not transfer a measured offset on a well-matched attenuator directly to an unmatched filter: their interactions with residual system errors can differ.

After a Drift Failure: Restore the Setup Before Retest

Record the failing trace, setup and serial numbers before any change. Check the connector condition and seating, cable position and instrument status. If the approved correction is invalid, recalibrate at the intended planes and run a separate verification again. Some analyzers offer a model-specific calibration update; follow the documented restrictions and still verify with an independent device.

RF coupon or filter acceptance: Use the verified setup to repeat the affected transmission and reflection checks over the contract band. Record both the original result and the retest, including which procedure authorizes final disposition. Comparative development tests: Re-run the reference and both board revisions at the same settings. Do not credit a design change if the verification shift could account for the observed improvement. No general PCB workmanship class sets a universal VNA drift allowance; use the instrument workflow and the agreed RF acceptance plan.

VNA Calibration Drift FAQs

How often should I recheck the VNA?

Base the interval on stability evidence, required decision margin and how many units you can retest. Add triggered checks when the setup changes.

Do I need to recalibrate every time the check is repeated?

No. If the approved check still passes and the setup is unchanged, keep the correction under the agreed plan. An out-of-limit result needs investigation and a verified correction.

Can I simply reuse the calibration thru as my check?

It is a useful sanity check, but not independent evidence on its own. Use a suitable artifact not used to fit the correction.

Does a stable reference trace prove it is accurate?

No. It can stay stable while being biased. Establish the baseline against independently characterized data first.

What if only phase drifts?

Check the phase allowance, reference plane, and cable or adapter handling. A passing magnitude reading does not override a phase requirement.

Can I subtract the reference shift from every result?

Not without an approved model and independent verification showing that correction applies to the DUT and measurement. Ordinarily, hold and retest affected boards.

Does the first failed check tell me when the drift started?

No. It provides a detection time. Use the last valid check and the event history to define a conservative affected interval.

What must the test report keep?

Keep verification traces, baseline, settings, test and event times, serial numbers, limits and disposition so the affected interval can be reconstructed.

Technical References

Custom RF PCBs With Stackup and Test Coupon Review

Need a custom RF PCB or new test coupon? Ask EBest Circuit to review the stackup, launch geometry and coupon location before fabrication. Calling out the intended measurement plane and fixture at quoting gives the design and fabrication team a common basis for the RF evidence.

Send your files, operating band, impedance targets, coupon requirements and quantities to sales@bestpcbs.com. Request a custom RF PCB quote and include the measurements your team must verify.

S-Parameter DC Extrapolation: Find the Right 0 Hz Limit

November 17th, 2026

Set the DC limit from the circuit you actually measured, not from the first row of the Touchstone file. A direct copper path and a path with a series DC-blocking capacitor can both look almost lossless at 1 MHz. At 0 Hz, the first still conducts and the second does not. If you copy the first frequency point to DC, a transient simulation can show a voltage that never drains away.

Save the original sweep. Before adding a 0 Hz row or fitting a transient model, identify the circuit and reference planes, calculate the expected DC limit, and test a result that depends on it. If all you need is an in-band RF comparison, leave the unmeasured region unclaimed.

Coaxial DC-block module alongside a two-port RF PCB evaluation coupon
Check whether a DC-block module sits inside the recorded reference planes or will be added separately in the simulation.

DC Extrapolation: What the Missing Point Controls

DC extrapolation supplies the zero-frequency behavior that a sweep starting above 0 Hz does not measure. In a stable, linear time-invariant system, the DC transfer determines the settled output of a step, provided the simulated source, load, and initial conditions match the intended circuit. It is not a way to recover unmeasured resonances or prove that the fitted transient is accurate.

Use the whole two-port matrix, not just S21. S21 describes forward transmission with port 2 matched to its reference; S11 describes the reflection seen from port 1 under that same condition. A missing match or a simulator load that differs from the reference changes the actual voltage. Start by documenting the file’s reference impedance, physical port map, and whether coupling components and any bias network were included.

S-Parameter DC Extrapolation: Start With the DC Circuit

Draw the circuit between the actual measurement planes. Carry resistances and DC terminations into that drawing; in a first-pass ideal capacitor model, replace each series capacitor with an open. A through path, an isolated series break, and a through path with a shunt to ground have different DC limits. Include real leakage or bias paths if they matter to the system. Resistance between signal pins alone does not establish the shunt paths to ground.

Two equivalent circuits: a one-ohm series through path and an open series capacitor at DC
Ideal DC equivalents for two simple structures with no other shunt or leakage path. The port reference impedance is not a physical part in either drawing.

For equal, real 50 Ω reference impedances and one series impedance Z between the ports, S21 = 100 Ω ÷ (100 Ω + Z) and S11 = Z ÷ (100 Ω + Z). A 1 Ω through path gives S21 = 0.9901 (−0.086 dB) and S11 = +0.00990 (about −40.09 dB). An ideal series open gives S21 = 0 and S11 = +1. Keep the latter sign: an open and a short can both have reflection magnitude 1, yet they are not the same circuit. When a measured load is close to an open rather than exactly open, see the VNA high impedance measurement guide for deciding whether the fixture can resolve a finite value.

AC Coupling Capacitor: Why 1 MHz Is Not DC

In this calculated example, an ideal 100 nF series capacitor works between a 50 Ω source and a 50 Ω load. Its time constant is (50 + 50) Ω × 100 nF = 10 µs, corresponding to a −3 dB point of roughly 15.9 kHz relative to the high-frequency transmission of this ideal circuit. At 1 MHz the reactance magnitude is about 1.59 Ω and |S21| is 0.99987, or roughly −0.0011 dB. That first measured frequency is well above the corner and cannot identify the DC behavior on its own.

With a 1 V ideal source step ahead of the 50 Ω source resistance, the matched load starts at 0.50 V and decays to 0 V. The ideal load voltage is 0.50 e−t/10 µs V: about 184 mV at 10 µs and 3.37 mV at 50 µs. Compare that with about 0.495 V settled for the separate 1 Ω DC-connected path. These are calculations for simple circuits, not measured board data or recommended global limits. A real dielectric, bias resistor, source waveform, or termination can change the result.

Calculated matched step response for the 100 nanofarad DC-blocked path versus the one ohm DC-connected path
The DC-blocked load voltage decays over several time constants. A steady-state voltage is a different result, even if the two paths look similar across a high-frequency sweep.

DC Measurement vs. Extrapolation vs. Circuit Modeling

Choose the method by what you can independently verify
Method When to use it Trade-off and check
Independent DC and low-frequency measurement You can access the correct planes and the device remains in its operating state. Confirm terminations and parasitic effects. A four-wire resistance check can support a simple through path; it does not measure the low-frequency knee or characterize a biased active network.
Topology-constrained circuit or fit The topology is known but the VNA cannot reach the region that controls the transient. Constrain the DC limit and compare with overlap or separate low-frequency data. Do not add a coupling capacitor again outside the model if it is already inside.
Automatic DC extrapolation Exploratory plots, or a verified near-DC sweep with justified behavior. Convenient but algorithm-dependent. Inspect the imported zero-frequency matrix and compare against topology and a reference response. Restrict use if those checks fail.

Touchstone File: Keep the Measured and Derived Files Separate

Archive the original file as measured. Make a separate model identified as DC-extended. Record the original start frequency, point spacing, first measured point, derived DC matrix, reference impedance, and whether the analysis tool resamples onto a uniform grid. Verify the simulator imports 0 Hz as intended; some workflows synthesize their own endpoint or require a particular grid for a low-pass transform. A scope-like time plot is not proof that the software used your DC row.

For a passive real circuit with consistent real port references, a finite zero-frequency matrix should be real. A fit or extrapolation should not introduce gain or an impossible low-frequency shunt. For a cascaded path, check that each section agrees on the interconnection and impedances before assembling it; see cascading S-parameters without losing mismatch information. Keep the component and de-embedding boundaries explicit rather than forcing a through limit on the entire assembly.

Time Domain Step Response: Check the Tail and the Transition

Use a step suited to what you need to verify. For a DC-blocked link, verify that a long pulse decays toward the expected baseline and that the droop during the intended observation window meets the link budget. A stimulus far shorter than 10 µs in the example above can miss much of the decay. If your data is too coarsely spaced to support the duration of the record for an FFT-based transform, collect a suitable sweep or use an independently tested circuit fit; do not treat zero padding as extra measured data.

For a DC-connected coupon, verify the DC drop with an independent resistance or loaded transfer measurement and then check the early transition against a scope or a valid in-band model. For a fixture with a bias tee, match the actual bias-port termination and operating bias and verify the port that is supposed to block DC. A wrong final value points first to the DC path, termination, or simulator settings; an accurate final value with the wrong decay or transition still needs a low-frequency or bandwidth check.

S-Parameter Model Validation: Use a Release Worksheet

Use this worksheet for a model that will feed an eye diagram, step analysis, or a long-pulse check. Set tolerances from the project and the measurement uncertainty, then capture a falsifiable check. A chart label such as “verified” is not a substitute for the topology, target value, and supporting measurement.

Model release checks with one worked entry
Test Record or compare Decision
DC block: worked example 100 nF, 50 Ω each side, 1 V source step, export starts at 1 MHz. Ideal load at 50 µs: 3.37 mV. If the simulator instead predicts a settled voltage near 0.50 V, reject the DC continuation. Inspect the imported matrix and AC coupling topology.
Resistive through path Port IDs, continuity and resistance at the planes, DC matrix, actual source and load. Investigate any mismatch between measured drop and modeled final value. Check shunt paths as well as series loss.
Shape and band fidelity First measured frequency, independently checked low-frequency points or decay, residual across the original band. Reject a model that matches either endpoint or measured band but misses the required decay, droop, or transition.
Release boundary Original and derived filenames, limits on frequency, bias, source/load, and time window. Release only for intended uses supported by those checks. If needed, label it in-band only.

S-Parameter Simulation: When to Leave DC Unmodeled

If you are comparing insertion loss only from 2 to 8 GHz and the sweep covers that band, a guessed DC row adds no evidence. Keep the decision inside the measured band. If correcting the endpoint changes the predicted eye baseline or long-pulse droop, the result is sensitive to information that the measured band did not provide. Obtain a lower-frequency reference or a verified circuit description before using the model for a pass/fail decision.

No general PCB workmanship criterion specifies the DC entries in a Touchstone file. Use the product’s actual DC and transient requirements. When discussing a board with your fabricator, specify which path they build and where the measurement ends; agree separately on what any coupon or characterization report must demonstrate.

S-Parameter DC Extrapolation FAQs

Is copying the first data point to DC safe?

Only if independent evidence supports a flat continuation and the circuit’s DC limit. A 100 nF series block can look nearly lossless at 1 MHz while transmitting no DC at all.

Can a VNA measure an actual 0 Hz S-parameter point?

A normal RF sweep starts above zero. Some systems offer separate low-frequency or DC characterization modes or a derived endpoint; check the instrument workflow rather than treating a synthesized DC entry as an RF sample.

Should S11 be zero for a series DC block?

No. An ideal series open with no other paths gives a reflection coefficient of +1 at DC in this 50 Ω example. Zero would represent a matched port.

Does zero DC transmission mean there is no output pulse?

No. The output can jump and then decay because a step contains changing voltage as well as a steady component.

What if the designer already modeled the AC coupling capacitor?

Find the measurement planes. If the capacitor is inside both the imported model and the external schematic, remove the duplicate representation or use a correctly de-embedded model.

Can I check a copper through path with a DMM?

Yes, a suitable four-wire measurement can confirm small DC series resistance at the same planes. It does not by itself establish the complete low-frequency two-port response.

Does a correct DC matrix guarantee a correct transient?

No. The decay, transition shape, losses, and timing also depend on the rest of the model and on the source and load.

What should I send with a DC-extended S2P file?

Send the original export, the derived file, the DC assumptions and basis, the reference planes and port impedances, and the independent check that supports its intended use.

Technical References

Custom RF PCBs With Stackup and Coupon Review

Planning a new RF PCB or evaluating a prototype before production? Ask EBest Circuit to review the proposed stackup, RF transitions, and coupon placement. Reviewing the board layout before quoting gives you a chance to put the measurement planes where they serve the design, instead of leaving them to a last-minute fixture.

Send your layout files, stackup, working band, impedance targets, and quantities to sales@bestpcbs.com. Request a custom RF PCB quote and tell us what the coupon needs to verify.

Cascading S-Parameters: When Adding Loss in dB Fails

November 17th, 2026

Do not add the insertion loss of two PCB blocks in dB unless the reflections at their joint are small enough for your error budget. Join the full complex two-port models first. In the example below, the same two forward-transmission magnitudes predict 6.20 dB of loss if you add them, but the complete model gives 5.38 dB or 6.94 dB depending on the reflection phase. None of those figures is measured board data.

Two separate RF PCB test coupons connected in a physical cascade
The modeled joint must be the joint on the bench. A connector or adapter at that joint must be represented once, not twice.

Cascading S-Parameters: What Are You Connecting?

A cascade puts the output of network A into the input of network B. The wave transmitted by A travels toward B, and a wave reflected from B travels back toward A. If both internal ports reflect, some of that wave makes repeated trips across the interface. The combined transmission depends on both the loss through each network and the complex size and phase of those returning waves.

Draw the port order, joint and exterior reference planes before opening either S2P file. If A and B were measured with the same removable coax launch attached, their separate measurements might include duplicate hardware. Do not join those files as though the shared launch were a zero-length joint. Decide whether the physical connector or transition belongs inside A, inside B, or in a separate measured block.

S-Parameter Cascade Formula: Include the Interface Reflections

For two directly joined two-ports at the same frequency, with connected waves on compatible real reference impedances and consistent normalization, the total forward transmission is:

S21,total = A21B21 ÷ (1 − A22B11).

With no incident wave at the output of B, the wave coming back into A equals the wave entering B multiplied by B11. At A’s output, the forward wave is the direct contribution from A21 plus that returning wave multiplied by A22. Substituting one relation into the other gives the denominator. If one internal port is perfectly matched, the product A22B11 is zero and the simple S21 multiplication is exact under these assumptions.

Two connected S-parameter blocks with internal reflections A22 and B11 and forward signal crossing from A to B
Use A’s output reflection and B’s input reflection at the connected planes. Keep any separate electrical length in its own modeled block.

Insertion Loss Calculation: How Much Can Adding dB Miss?

In a hypothetical calculation at one frequency, let both forward-transmission magnitudes be 0.70, and let the two internal reflection magnitudes be 0.30, or about 10.46 dB return loss each. The product estimate is 0.49, giving S21 = −6.20 dB in magnitude notation (positive insertion loss of 6.20 dB).

Now include phase. If A22B11 = +0.09, the magnitude is 0.49 ÷ 0.91 = 0.5385, giving 5.38 dB insertion loss. If the reflection product is −0.09, its magnitude is unchanged, but the result is 0.49 ÷ 1.09 = 0.4495, giving 6.94 dB insertion loss. The shortcut missed by about 0.82 dB in the first case and 0.75 dB in the second, using unrounded results. That is material in a 0.50 dB modeling budget but may be acceptable for a rough estimate with a looser budget.

Example comparison of 5.38 dB for in-phase reflection, 6.20 dB for adding loss, and 6.94 dB for opposite-phase reflection
Calculated results for the same magnitudes and different interface phases. A larger return-loss number means a smaller reflection magnitude.

If the phase is not known and the reflection-product magnitude is r < 1, the denominator magnitude lies between 1 − r and 1 + r. Bound the shortcut error across the whole band before using it. If that bound consumes your allowed margin, use the complex cascade and verify it.

S-Parameters to ABCD Conversion vs. Direct Network Connection

Do not multiply the raw S matrices as though they were transfer matrices. The variables sit on different sides of the network. Chain matrices such as ABCD can be multiplied in physical order when the input, output and current conventions match. Or use a library’s direct network connection with explicit ports, which avoids manually choosing a matrix convention.

Three approaches to cascading or estimating a two-port chain
Approach Use when Check before trusting it
Add loss in dB Getting a first estimate for well-matched stages Bound reflection error against the modeling budget. No return-loss prediction.
ABCD or chain matrices Cascading compatible two-ports in a documented convention Confirm order and conversion, especially near very small S21, where common conversion formulas can be poorly conditioned.
Direct complex connection Using a library with known port and impedance handling Verify the connection map, normalization and frequency alignment. Compare with an independent assembly measurement.

Touchstone File: Align Frequency, Port Order and Reference Impedance

Confirm the file’s frequency unit, port mapping, complex representation and reference impedance. Decide what reference planes the measured blocks include. A 50 Ω file and a 75 Ω file do not establish the geometry of a real transition; normalize or connect the full networks with a method that handles their reference impedances. Include a physical transition separately if it is part of the assembly you intend to predict.

Use the shared measured band and a documented frequency grid. Avoid extending a file outside its valid band to make the points line up. If a solver needs new points, apply the S-parameter interpolation checks before joining the models. Near a notch, a sweep that misses the feature cannot be repaired by adding more display points.

How to Verify a Cascaded S-Parameter Model on the Bench

Measure the physically joined chain with the same exterior planes, port directions and frequency band as the model. Plot the complex residuals for forward and reverse transmission and the relevant reflections. A close S21 magnitude at one frequency can hide a wrong phase slope, a flipped block or a fixture error. Save the independent assembly file with the model.

Worked assembly comparison; values and limits are illustrative
Item Example residual Interpretation and next step
Insertion loss Model 5.38 dB; assembly 5.56 dB; difference 0.18 dB Within the 0.25 dB example limit if the agreed uncertainty rule is satisfied.
Transmission phase Model −70°; assembly −61°; difference +9° Outside the 5° example limit. Check the reference planes, connector delays and file orientation.
Reflection or reverse path Compare relevant S11, S22 and S12 across the critical band Not proven by the first two rows. Use the actual limits and retain the measured file.

Where to Use a Two-Port Cascade in PCB Engineering

Cable plus RF board: Use the cable and board files to estimate loss and phase at the system interfaces. Include any adapter once. Verify total response at the intended cable position across the operating band; a changed cable route can invalidate the comparison.

Connector launch plus PCB trace: Place the join at defined geometry, not at a vague connector boundary. Check the return loss and phase as well as insertion loss. If adjacent structures couple directly across the supposed boundary, measure or simulate the combination instead of assuming independent two-ports.

Filter plus another RF stage: Check passband ripple and stopband rejection as separate tasks. A good two-port fit in the passband does not prove a deep stopband result near the instrument’s noise floor. Verify any notch that determines whether the design passes.

How to Release a Cascaded Network for Simulation

Record the original files, the model generator and version, frequency range, grid, port directions, reference impedances, joins and operations. Sanity-check with a known low-reflection case. If the assembled measurement differs, first check planes, order and repeated hardware, then compare at original sample points before blaming interpolation. For passive blocks check the resulting model as appropriate; for transient use review the required causality and bandwidth assumptions separately. If the cascade will drive a step or eye simulation, use the S-parameter DC extrapolation guide to check that the combined path has the correct low-frequency limit.

Cascading S-Parameters FAQs

Can I multiply the S21 magnitudes?

Only as an approximation unless the internal reflection product is zero. Bound the error before making a decision.

Can two lossy stages have less loss than a scalar estimate suggests?

Yes. Interface reflection phase can increase or decrease transmission relative to the scalar product. That does not imply a passive assembly creates gain.

Can I multiply the two raw S matrices?

No. Use an appropriate connection calculation or a compatible chain representation.

What if the port references are different?

Use a method that handles the reference impedances and wave conventions for both full matrices. Do not treat that as a model of unspecified physical hardware.

Can I cascade files with different frequency grids?

Only after aligning them with a justified method in their common band. Check independent points where sharp features matter.

Does agreement in dB prove the model is correct?

No. Compare the relevant complex transmission and reflection terms at matching planes.

What if there is a deep null near the conversion?

Common chain conversions can amplify numerical and measurement errors when forward transmission is tiny. Check the method’s conditioning and the measured floor.

When is a two-port model insufficient?

Use a wider model when coupling, common-mode paths, or additional ports matter and cannot be represented independently with the intended terminations.

Technical References

Custom High-Speed PCBs With Stackup and Coupon Review

Need a custom RF or high-speed PCB? Ask EBest Circuit to review the proposed stackup and test coupon before fabrication. Send the layout or Gerbers, impedance targets, test interfaces, quantity and delivery needs to sales@bestpcbs.com for a project review and quotation.

VNA Unknown Thru Calibration: When Ports Cannot Mate

November 17th, 2026

When your VNA test ports cannot mate, do not add a barrel and tell the analyzer it is a zero-length thru. The barrel can introduce loss, delay and mismatch. Use a properly defined thru, a suitable reciprocal Unknown Thru, or the instrument’s adapter-removal workflow. For an Unknown Thru, pass an independent amplitude and phase check before using the setup to judge an RF board.

Machined coaxial through adapter joining two phase-stable VNA test cables
The thru is a temporary calibration connection. Document what remains on each test port when the board replaces it.

VNA Calibration: When the Test Ports Cannot Mate

Test ports are non-insertable in this setup if their mating ends cannot directly connect at the intended reference planes. Common examples include two same-sex connectors and two different connector families. The board may fit the two cables perfectly while the cables cannot mate with each other for a flush thru. Do not force the connection or mate connector interfaces that look similar but are not compatible. Confirm the interface, inspect the connectors and use the maker’s specified mating and torque procedure.

Draw the measured board, both mating faces and every adapter you plan to leave in place. The cut between test hardware and the board defines what the data will include. A calibration that verifies well at the wrong cut still describes the wrong network.

SOLR Calibration: What Does an Unknown Thru Solve?

In an Unknown Thru or SOLR calibration, establish the reflection standards at each port, then measure a reciprocal transmission connection in both directions. Rohde & Schwarz uses the name UOSM for a related unknown-through workflow. The instrument constrains the forward and reverse transmission of the thru to be equal as complex quantities (S21 = S12 under consistent references). The thru can still have loss and delay; its two reflections do not have to be identical.

It is the transmission connection that is “unknown,” not the port standards or setup. Follow the calibration wizard and use the appropriate standards at each port. Do not use an isolator, amplifier or other unverified directional device as the assumed reciprocal connection. Equality of magnitude at one frequency is not enough to establish reciprocity.

Adapter Removal Calibration vs. Unknown Thru vs. Defined Thru

Start with the standards you have and the planes the drawing requires. A suitable characterized thru can be the simplest choice; SOLR helps where an accurately defined thru is unavailable. If your platform cannot support SOLR with the required geometry, use a supported adapter-removal workflow or arrange an appropriate calibration kit.

Three choices for a non-insertable test setup
Method What you need Trade-off and check
Defined thru A compatible thru and its valid kit definition Check the actual standard and definition match. Do not substitute an unknown barrel for a flush standard.
Unknown thru (SOLR/UOSM) Suitable reciprocal path, correct port standards, sufficient received signal and supported phase-selection workflow The analyzer can resolve loss and delay without a full thru model, but verify the phase solution independently.
Adapter removal Compatible standards and supported calibration or characterization steps The classic two-calibration method needs more reconnections. Control repeatability and confirm the resulting planes.

Group Delay Measurement: Why the Thru Phase Estimate Matters

At 10 GHz, a quarter of a period is 25 ps, or 90°. Keysight’s cited M9485A workflow asks for the thru phase estimate to be within a quarter wavelength; refer to the actual instrument’s implementation. The estimate helps choose a phase solution; the two possible transmission solutions can differ by 180° without differing in magnitude. The quarter-cycle example does not mean the thru itself must have less than 25 ps of delay or that errors under 25 ps are acceptable for the final measurement.

In this ideal nondispersive example, an independently established thru delay is 145 ps. A 125 ps guess differs by 20 ps: |Δφ| = 360° × 10 GHz × 20 ps = 72°. A 110 ps guess differs by 35 ps, giving 126°. The latter is outside the quarter-cycle example. For an independently measured and nearly linear unwrapped phase trace, estimate one-way delay as −Δφ ÷ (360° × Δf), with frequency in Hz and phase in degrees. For dispersive media, where phase does not follow one constant delay across the band, follow the analyzer’s phase-selection procedure instead of entering a guessed coax delay.

At 10 GHz, a delay estimate 20 ps off corresponds to 72 degrees and 35 ps off corresponds to 126 degrees
This is a calculation for phase-guess proximity, not an instrument accuracy chart or fabrication limit.

VNA Reference Plane: Set It Before Connecting the Thru

Set the frequency range, point count, power and any other settings required by the calibration workflow. Keep the cables in their measurement position. Capture the one-port standards at the actual intended mating faces, then connect the reciprocal thru and follow the wizard. Record the thru ID, connection orientation and delay or phase-selection setting. If a transition stays on the measurement cable, account for its location in the calibration design rather than silently moving it after verification.

The same two VNA reference planes bound a temporary reciprocal thru and then the RF PCB under test
Use the same mating planes when the board replaces the thru. Adding or moving hardware can invalidate that assumption.

Check transmission through the chosen connection across the whole band. A deep notch near the receiver floor makes the estimate unreliable there. For example, Keysight cautions against 40 dB or more combined loss in the thru and calibration path on the cited workflow; use your instrument and measured signal rather than assuming 40 dB is universal. A plain VNA port extension may correct an accessible path’s delay, but it does not replace a two-port calibration that accounts for adapter mismatch.

Insertion Loss Measurement: Verify Phase, Not Just dB

Measure a separately characterized two-port that was not used to solve the calibration. Use the same frequency grid or a justified reference comparison, correct port order and specified impedance. Check both directions and compare phase as well as magnitude. When the application depends on return loss, verify the reflection response with a suitable independent mismatch standard too. The artifacts and results should cover the ranges that matter to the job. If that check must stay valid across a production run, use this VNA calibration drift guide to set checkpoint intervals and identify boards that need retesting.

Example release worksheet at 10 GHz; figures and limits are illustrative
Check Reference → indication Decision
Forward insertion loss 6.00 dB → 6.09 dB; error +0.09 dB Passes the example 0.15 dB magnitude-error limit, assuming the agreed uncertainty rule is satisfied.
Forward transmission phase −70° → −62°; difference +8° Fails the example 5° error limit. Hold calibration and review the settings and reference planes.
Reverse transmission and reseat Independent reverse reference plus readings after a controlled reconnect Required checks still open. Compare complex results to approved limits; neither earlier row releases the station.

Where Unknown Thru Helps on an RF PCB Bench

Same-sex ports on a coupon: Select a reciprocal thru that mates to the cables without forcing them into a new position. Qualify the setup in both directions before comparing coupon insertion loss or phase. Keep the board launches in the measurement unless your documented reference planes and separate fixture-removal method say otherwise.

Filter with mixed connector types: Use the correct reflection kit for each connector and a supported thru that mates both ends. Verify return loss and the loss and phase through the passband. Separately demonstrate usable receiver margin if stopband rejection is an acceptance requirement; a passband reference cannot by itself prove the stopband result.

Coax-to-waveguide fixtures: Use appropriate standards and a verified reciprocal transition. If the analyzer requires a dispersive phase selection, follow that process and verify with a distinct known connection over the operating band. Otherwise hold phase-sensitive conclusions.

VNA Calibration Procedure: What to Record Before Release

Record the analyzer and firmware, frequency grid, test settings, port connector types and planes, kit revision, thru ID and orientation, estimated delay or branch selection, and independent verification results. Include units and the governing decision rule. For troubleshooting, first inspect the raw receiver response and repeatable seating; then verify the cal kit and phase choice. Recheck the independent artifact after any change before retesting boards. A corrected plot that looks smooth is not evidence that the calibration passed.

VNA Unknown Thru Calibration FAQs

Does the Unknown Thru need to be lossless?

No. It must meet the instrument’s reciprocity, signal and phase-selection conditions. Its loss must leave enough received signal across the calibration band.

Can the Unknown Thru have different S11 and S22?

Yes. Reciprocity constrains the complex forward and reverse transmission under consistent references, not equality of the two reflection responses.

Can I enter a zero delay for a barrel?

Only when a procedure and verified phase range justify it as an estimate; do not define a real barrel as a perfect flush thru to avoid characterizing it.

Can I use the board itself as the Unknown Thru?

Some analyzers allow it if reciprocity, phase-selection and transmission conditions are met. Avoid using that same board as the only verification artifact.

Why does my loss pass but phase fail?

A magnitude check can miss a wrong phase branch or wrong plane. Check the phase against an independent known connection, then check connection repeatability.

Does a port extension remove adapter mismatch?

Not by itself. A phase or loss model does not recover the adapter’s full reflections and interaction with the device.

What if the two test ports use different connector families?

Use the correct standards for both connector families and a compatible thru or characterized transition within the instrument’s supported workflow.

When should I repeat verification?

Recheck when a cable moves substantially, a connector or adapter changes, a cal setting changes materially, or an independent reference fails. Use the approved interval for ongoing production checks.

Technical References

Custom RF PCBs With Stackup and Coupon Review

Need a custom RF PCB? Share your proposed stackup, layout and connector launches with EBest Circuit for review before fabrication. Include your operating band, impedance targets, coupon needs, quantity and delivery date. Send the files to sales@bestpcbs.com to request a project review and quotation.

S-Parameter Interpolation: Avoid False Insertion Loss

November 17th, 2026

Two S-parameter files can agree at every measured frequency and still predict different insertion loss between those frequencies. Before you call the PCB lossy, plot the original measured points over the solver trace. If the extra loss appears only between samples, check how the solver interpolated the complex data. In the worked example below, the interpolation mode alone introduces 3.01 dB of apparent extra loss. That is a calculated example, not a measured board result.

RF PCB transmission-line coupon with two coaxial launches and the S-Parameter Interpolation title
A two-port VNA sweep measures discrete points. A solver can generate more points, but it cannot measure what happened between them.

What Is S-Parameter Interpolation?

Interpolation estimates a complex S-parameter at a frequency between measured samples. For example, if the VNA recorded S21 at 4.00 and 4.10 GHz, the simulator may need an estimate at 4.05 GHz. S21 describes forward transmission under the file’s reference conditions; it has both magnitude and phase. A trace can therefore look smooth in dB while giving the wrong phase, or have the right phase but predict too much loss.

Keep the original export as the reference. Tag derived files as interpolated, and record the source sweep, target grid, method, port order and reference impedance. Treat guesses beyond the measured band, including a missing DC point, as extrapolation and review them separately. If your next step is to join two verified models, see this guide to cascading S-parameters with interface reflections before adding their losses.

What a Touchstone File Tells You About Interpolation

The Touchstone option line specifies frequency units and data representation. RI stores real and imaginary values; MA stores linear magnitude and angle; DB stores magnitude in dB and angle. These describe how the original points are encoded. They do not guarantee how the simulator fills in gaps. Re-exporting the same samples as RI does not, by itself, select Cartesian interpolation.

For a useful A/B comparison, keep the measured file, reference impedances, ports and target frequency grid fixed. Change only the interpolation setting. Compare both traces at the original points before investigating what they do between them. If they already differ at the original samples, check renormalization, reference-plane changes and any automatic fitting or smoothing.

Linear Interpolation Formula: Why the Midpoint Loses 3.01 dB

In this calculated example, a path has constant transmission magnitude of 0.800 and delay of 2.5 ns. Between 4.00 and 4.10 GHz its phase rotates 90°. At 4.05 GHz, polar interpolation yields −1.94 dB; Cartesian interpolation yields −4.95 dB. The PCB did not change. The latter method cut across the complex-plane arc.

Let t be the fractional position between the two frequencies. Cartesian interpolation computes (1 − t)S1 + tS2; polar interpolation applies the same weights separately to linear magnitude and unwrapped phase. Here t = 0.5, S1 = 0.800 + j0 and S2 = 0 − j0.800. The Cartesian midpoint is 0.400 − j0.400, with magnitude 0.566. The polar midpoint is 0.800∠−45°. Apply 20 log10|S21| to each magnitude to reproduce the numbers. Insertion loss, stated as a positive loss, is 1.94 dB versus 4.95 dB.

Complex-plane diagram comparing magnitude of 0.800 on the polar arc with 0.566 at the Cartesian chord midpoint
Same endpoints, different midpoints. The arc preserves magnitude in this idealized delay example; the chord does not.

Cartesian vs. Polar S-Parameter Interpolation: Which Should You Use?

Try polar interpolation for a smooth delay-dominated path and test it against measured points you did not give the model. Try Cartesian when the data pass near a null, where phase is difficult to interpret. If the raw sweep missed a narrow feature, neither method deserves a pass until a denser measurement resolves it.

S-parameter interpolation methods and the checks that decide between them
Method Try it when Watch for Verify with
Cartesian linear (real, imaginary) Phase becomes unreliable near a null; the complex trajectory is adequately sampled Chords create loss when phase rotates greatly per step Midpoint complex residual and S21 loss error
Polar linear (magnitude, phase) Transmission is smooth and delay dominates phase Phase unwrap can choose the wrong turn; phase near a null is unstable Measured phase trend, notch depth and ripple
Re-measure a finer sweep A notch, resonance or rapid phase rotation falls between samples Extra sweep time and drift between passes Convergence as spacing decreases, with unchanged calibration and fixture

Phase Unwrapping and Frequency Step Size

A 2.5 ns delay rotates the phase by 90° per 100 MHz, or 180° per 200 MHz. At 180°, a wrapped phase difference alone cannot tell the two directions apart. Even well below that figure, a Cartesian chord can still create unacceptable loss. Near a resonance or deep null, phase can change much faster than the nominal through-path delay suggests.

For a constant-amplitude, constant-delay path, the Cartesian midpoint amplitude ratio is cos(|Δφ|/2) when the actual phase step |Δφ| is below 180°. If the example’s Cartesian midpoint error budget is 0.10 dB, that model gives a maximum step of about 17.36° or 19.3 MHz at 2.5 ns. This is a starting sweep setting for the example, not a VNA specification. Check it with an offset sweep and tighten it where the DUT changes more quickly.

How to Check S-Parameter Interpolation Against Measured Points

Take a calibrated sweep on a fine grid. Save alternating measured points as a separate verification file, then feed only the retained points to each interpolation method. Compare the predicted results at the withheld points against the saved measurements. Keep VNA settings, fixture, reference planes and signal path unchanged. Repeat near critical frequencies if the measured trace is close to the acceptance limit.

Holdout validation diagram showing retained 4.00 and 4.10 GHz points and the independent 4.05 GHz verification point
Do not give the 4.05 GHz verification value to the interpolator. Replotting a generated point on a denser display is not independent validation.
Holdout worksheet: what to enter, how to judge it, and what to do next
Check Record Decision and next step
Through loss at 4.05 GHz (example) Calculated reference −1.94 dB; Cartesian −4.95 dB; residual 3.01 dB Fails a hypothetical 0.10 dB budget. Try polar, then confirm against an actual held-out sample.
Delay-sensitive band Wrapped phase residual at each held-out point; group-delay residual if required If outside the design budget, verify phase unwrap and retest with a finer grid.
Filter notch or resonance Measured and predicted feature frequency, depth and critical-band residual If the fitted shape or frequency fails, measure locally at finer spacing. Near the noise floor, assess uncertainty before classifying depth errors.

S-Parameter Passivity Check After Interpolation

For a passive network, check passivity over the generated grid, not only at the original points. Where ports use fixed positive real references and power-wave normalization, the largest singular value of the full S-matrix must not exceed one for ideal passive data. An S21 magnitude below one does not prove the network is passive. Evaluate numerical or measurement tolerance separately from the ideal bound.

With a shared linear weight and identical normalizations, Cartesian interpolation of two passive whole matrices preserves the singular-value bound between them. That result does not carry over to arbitrary cubic interpolation, element-wise polar interpolation or causality. For impulse or step simulations, use the solver’s grid requirements and review the S-parameter causality and time-domain checks separately.

S-Parameter Simulation: What to Check Before You Release the File

Channel-loss budget: keep the full band needed for the channel model, identify the frequency or band where spec margin is tightest, and allocate an explicit dB error budget to interpolation. If the residual at an independent point exceeds that budget, do not use the dense model for sign-off.

RF filter model: check that the generated file preserves the passband ripple, stopband rejection and notch location at the frequencies in the filter specification. Add measured points near features rather than relying on a smooth trace or an interpolated null.

Time-domain use: log the measured band and any assumed DC or high-frequency behavior. Repeat the analysis after a finer measurement sweep; compare arrival times and feature amplitudes against the project tolerance, not just the software’s display. Hand off the original file, the derived file and the holdout report together so the next engineer can reproduce the decision.

S-Parameter Interpolation FAQs

Does a denser Touchstone export improve the original measurement?

No. A denser export adds estimates. To establish what the DUT actually did between points, take a finer sweep or independently measure selected frequencies.

Can I interpolate only the dB trace?

For a magnitude-only plot it is a display choice. For a complex simulation, also preserve phase and specify whether the chosen algorithm operates in linear magnitude or dB; these are not equivalent operations.

What phase step is too large?

At 180° the two wrap directions are ambiguous from an isolated pair. Much smaller spacing can already be necessary for your error budget; the equal-amplitude 2.5 ns example needs about 19.3 MHz for a 0.10 dB Cartesian midpoint target.

Can I use polar interpolation across a deep notch?

Only if measurements confirm the result. Phase at very small amplitudes is sensitive to noise, and a coarse sweep may miss the notch depth or position. Re-measure around the feature before deciding.

Should I use a spline instead of a straight line?

Use it only if independent data show that it lowers relevant errors. A smooth spline may overshoot or distort narrow peaks; recheck loss, phase and passivity on the resulting grid.

Is adding a DC point interpolation?

Not unless the measured band already brackets the point. A missing DC value requires a physical low-frequency model, independent measurement, or an explicit extrapolation decision.

Why do two simulators disagree with the same S2P file?

Confirm they use the same ports, reference impedance, reference planes, interpolation domain and out-of-band policy. Then compare their values at the original points and at one withheld midpoint.

Should every S-parameter use the S21 error limit?

No. Set limits for the actual task. A channel model may prioritize forward loss and delay, while a matching problem may require a tight complex reflection error, especially in the bands of interest.

Technical References

Custom High-Speed PCBs With Stackup and Coupon Review

Need custom high-speed or RF PCBs? Ask EBest Circuit to review your proposed stackup and impedance-coupon requirements before fabrication, so you can agree on which layers and structures need to be represented. Send your layout or Gerber files, stackup target, impedance requirements, quantities and requested delivery date to sales@bestpcbs.com for a project review and quotation.

VNA Isolation Calibration: Measure the DUT, Not Port Leakage

November 17th, 2026

Use VNA isolation calibration only when a stable leakage path is large enough to affect the result and the correction improves an independent check. The optional step measures transmission with the two test ports terminated separately. It can reduce stable additive leakage in the calibration model. It cannot recover a signal buried in random noise or separate physical PCB crosstalk from the PCB you intend to measure.

Start with each port on its own matched load. Save the complex trace, then verify the result with a known attenuator in the range you intend to claim. If the corrected trace looks deeper but a check standard disagrees, do not use the nicer-looking stopband.

VNA Isolation Calibration: What the Optional Step Removes

There should be no intentional transmission path between independently terminated ports. A repeatable response can still appear because part of the source signal reaches the receive path. On a VNA that implements an isolation term, that leakage becomes an additive error estimate. A two-port calibration can have separate forward and reverse isolation estimates.

Two separate coaxial test cables, each capped with its own RF termination
RF connection illustration. The loads are separate terminations at the calibration planes, not a through connection.

Thinking in complex numbers explains the limit. If M is the measured response, D the wanted response, and L an additive leakage term, the simplified model is M ≈ D + L. The correction estimates and removes a vector, not a number of decibels. This picture isolates one error mechanism; it is not the complete two-port error model.

VNA Crosstalk Measurement: Separate Three Leakage Paths

First decide where the unwanted signal travels. Measure with the cable ends separately terminated, then compare a fixture isolation witness with the intended launches, cable routing, switch states and unused ports. A witness can be an intentionally isolated coupon designed to preserve the fixture geometry. Its result characterizes that witness; it does not automatically equal the leakage when a real PCB is installed.

Diagram comparing terminated cable ends, a fixture isolation witness and the real PCB path
These are three different measurement boundaries. Do not use a trace taken at one boundary to subtract everything at another.
Path Check Action
Analyzer and cables Two independent loads. Try isolation correction if the residual is stable.
Fixture or switch path Matched isolated witness. Fix or validate the fixture separately.
PCB coupling Actual board and required terminations. Keep it if this coupling is the thing being measured.

VNA Calibration Procedure: Terminate Both Test Ports

  1. Define the planes and conditions. Record the port map, frequency grid, source power, receiver settings and reference impedance. Use 50 Ω loads only for a setup whose calibration and test plan call for them.
  2. Follow the instrument’s calibration sequence. At the isolation prompt, put a separate matched load on each indicated test port. Do not use an open cable, a thru, or the DUT in place of an isolation standard.
  3. Acquire a clean estimate. Use the approved IF bandwidth or averaging setting for the calibration. Hold cable positions and switch states stable. Save the calibration and record whether both directions were acquired.
  4. Keep a comparison calibration. Where the instrument supports it, compare equivalent calibrations with and without the isolation term. Do not deliberately change receiver range or source power at the same time; it defeats the comparison.

VNA Noise Floor: Decide Whether Isolation Correction Helps

Repeat the terminated sweep and look at complex consistency. If a response is repeatable at the same frequencies and keeps phase across sweeps, a stable term may be present. If the response varies strongly in phase and magnitude near the floor, subtracting a noisy estimate can add error. A lower displayed trace after calibration is not by itself a reason to keep the term.

Use the VNA IF bandwidth guide when the problem may be random noise and sweep time rather than a stable leakage term. Verify a new calibration if instrument settings or the physical setup change. Never assume that a narrower bandwidth removes coherent leakage.

S21 Measurement: Convert Leakage Separation to an Error Bound

A leakage residual 20 dB below the wanted signal can still shift its reading by about −0.92 to +0.83 dB. For a calculated example at one frequency, assume S21 for the DUT is −80 dB and the remaining leakage bound is −100 dB at the same plane and normalization. The residual-to-wanted amplitude ratio is 10−20/20 = 0.100.

For a coherent residual with unknown phase, the extreme magnitudes are 0.900× and 1.100× the wanted amplitude. Convert those ratios with 20 log10. If you allocate no more than ±0.5 dB to this one error source, the more conservative cancellation case requires a ratio no greater than 1 − 10−0.5/20, or about 0.0559. That corresponds to 25.1 dB separation when rounded conservatively. Neither figure includes random noise, calibration-match error or repeatability.

Calculated amplitude bounds for a coherent residual 20 decibels below the wanted transmission: from minus 0.92 to plus 0.83 decibels
Calculated worst-phase limits for one additive residual term. The diagram does not represent measured VNA performance.

Use a justified upper bound on residual leakage, not the easiest single terminated sample. If you do not know the wanted amplitude independently, work from the product limit as a design target for the measurement or use the verified complex error bound and the applicable decision rule. Do not label the example separation as a full uncertainty budget.

VNA Calibration Standards: Verify With an Attenuator You Did Not Fit

Choose a characterized attenuator whose attenuation, reference plane, frequency range and uncertainty are appropriate to the decision. Measure it in both directions if both are in scope. Compare the result using each candidate calibration, with the same cables, band and test settings. Reconnect once or more under the approved repeatability procedure to see whether the claimed improvement survives re-mating.

A thru can verify basic tracking, but it cannot establish deep-attenuation accuracy. A low-loss reference is not a substitute for verification near −80 dB if that is what you intend to report. Do not assume a chain of nominal pads has the precise sum of its labels across the band. If the verification does not cover the required level, improve the setup or report the unverified band explicitly.

After both directions have enough verified margin, use the S21 vs. S12 guide to compare their complex responses. Do not interpret a direction-to-direction difference below the validated floor as nonreciprocal PCB behavior.

RF Isolation Measurement: Apply the Check to Three PCB Tasks

Filter stopband

Keep the filter and any shielding that belongs to the product in the measurement path. Compare the isolated fixture witness and attenuator before the sweep. Release only the frequencies at which the verified leakage and noise allowance supports the specified stopband limit; do not call the display floor the filter response.

Coupling between RF traces

Drive the defined aggressor and measure the victim at the specified end, with the other ports in their required terminations. Compare a fixture witness using the same cables and orientation. Do not subtract the coupling between PCB traces you came to measure, or report a matched-port result as a guarantee for an unmatched deployed load.

RF switch off-state

Keep the intended off-state, bias and unused-port terminations. Characterize or eliminate leakage around the fixture and through the switch matrix used for measurement. Verify representative off-state attenuation and repeat the state transition at the approved power; do not assume an open bench condition verifies off-state isolation.

VNA Dynamic Range: Decide What the Report Can Claim

Use the worksheet for the weakest relevant part of the band. Include both port directions when required and record the path used to establish the leakage bound. Document the product limit and chosen guard band separately from the calibration steps.

Check Example Decision
Leakage allowance −80 dB wanted, −100 dB residual bound. 20 dB gap is insufficient for the example ±0.5 dB allowance.
Independent reference Enter known attenuation, band and uncertainty. Verify both directions required by the test plan.
Reporting limit Save settings, witness, repeats and unsupported band. Report unresolved points as inconclusive, not a precise pass.

For RF PCB test acceptance, use the drawing, contract and qualified test plan for limits and uncertainty. There is no universal PCB isolation-calibration separation to substitute for them. If you cannot distinguish the DUT from the setup, retain the raw and corrected data and state exactly what remains unproven.

VNA Isolation Calibration FAQs

Do I need two matched loads?

Use the termination procedure for your model. The conventional two-port isolation step described here uses separate matched loads at the indicated planes.

Can I skip isolation calibration?

Yes if your method allows it and independent verification shows that residual leakage and uncertainty meet the measurement need.

Is S21 in dB the same as a positive isolation number?

Not in sign: for the same transmission path, S21 of −80 dB is often described as 80 dB of isolation. Record both the path and sign convention.

Why did my trace become noisier after calibration?

The correction can include noise from its estimate. Repeat at documented settings, check an independent reference and compare a calibration without isolation.

Can I subtract the terminated plot in dB?

No. Coherent leakage adds as a complex amplitude and may reinforce or cancel the wanted signal.

Can I use the same correction after moving a cable?

Only if your procedure verifies it still applies. Movement can alter both transmission tracking and leakage.

Will an isolation correction remove fixture crosstalk?

Not generally. Leakage that changes with the installed DUT or setup is not a stable matched-load term. Verify the actual fixture separately.

How much separation should I require?

Derive it from the allowed error and the verified type of residual, then include the other uncertainty sources. Do not treat the example as a specification.

Technical References

Custom RF PCBs With Stackup and Test Coupon Review

Planning a filter, switch or multi-channel RF PCB? Send EBest Circuit your stackup, Gerbers or ODB++, launch parts, frequency band and the locations of the isolation checks. Reviewing the launch and coupon geometry before fabrication helps make board results and fixture results comparable.

Request an RF PCB fabrication quote with your quantity, required date and acceptance plan.

VNA Port Extension: Move the Plane, Check the Result

November 17th, 2026

VNA port extension moves the electrical reference plane past a section of transmission line. Use it when the section behaves like a nearly matched line with predictable delay. A flat phase trace after correction tells you that the fitted slope is gone. It does not tell you that the connector launch, fixture ripple or loss has been removed. If the two test ports cannot mate to establish the initial two-port calibration, see the VNA Unknown Thru calibration guide before making any port-extension correction.

For a two-port PCB fixture, enter the one-way delay at each port, not the round-trip delay measured on S11 or S22. Then measure a standard you did not use to fit the correction. If that check fails the phase or amplitude budget, stop and change the measurement method before signing off the DUT.

VNA Port Extension: What Actually Moves?

A reference plane is where the instrument reports the incident and reflected waves. If you calibrate at the coax connectors, a PCB launch and its feed line normally lie inside the measured network. If you want the response at the component pads instead, define the new planes at those pads. Moving the planes changes which electrical length you report; nothing physically moves. If the production spec is defined at the board’s connector, keep that connector in the result.

RF PCB test fixture with two coaxial launches and a device position between them
RF fixture illustration. Identify the connector planes and device planes separately on the actual fixture drawing.

For reflection, the wave goes down the test section and back along the same path. For transmission, it passes once through the section at each end. This is why one entry on port 1 affects S11 by twice its delay but S21 by only its one-way contribution. Compare data at the same planes and reference impedance; otherwise raw and corrected traces describe different networks.

VNA Port Extension Calculation: One-Way vs. Round-Trip Delay

If the unwrapped S11 phase falls by 64.8° over a 1 GHz interval, the reflection takes 180 ps round trip and the port extension entry is 90 ps one way. This is a calculated example for a smooth, nearly matched fixture with a suitable reflection standard. Use several clean points across the intended band, rather than fitting two noisy endpoints.

Use one-way delay = −Δφ/(720 × Δf) for a reflection phase fit, with Δφ in unwrapped degrees and Δf in hertz; the result is seconds. At 3 GHz, removing 180 ps of round-trip delay adds 194.4° of phase to S11 before wrapping. If port 2 has a 110 ps one-way delay, S21 crosses 90 + 110 = 200 ps of fixtures and its correction at 3 GHz is +216°. Neither correction changes magnitude in delay-only mode.

PCB fixture showing 90 ps one-way delay at port 1, 110 ps at port 2, 180 ps round-trip reflection and 200 ps forward delay
In this calculated example S11 returns through the left fixture; S21 crosses both fixture sections. The instrument entries are 90 ps and 110 ps.

A useful sanity check: entering 180 ps for port 1 leaves 180 ps of excess round-trip correction. That is +64.8° per GHz of residual S11 slope in the example. If the corrected slope changes sign and is close to the original size, check for a factor-of-two error before changing the fixture.

VNA Electrical Delay vs. Port Extension vs. De-Embedding

Choose according to what must be measured. A display delay is useful for inspecting a trace; a port extension is a plane correction based on a line model; network de-embedding can remove fixture behavior that the line model cannot describe. They are not interchangeable.

Method When to use it When to switch
Electrical delay Check one trace for non-linear phase. You need to move a shared reference plane.
Port extension Move the plane along a matched line. The check short ripples or the phase slope varies sharply.
Network de-embedding Use validated fixture networks to remove mismatch, loss and delay. The model is unreliable or you can calibrate at the DUT plane.

Validate at the same device plane regardless of method. For a simulation deliverable, too much delay removal can make the response arrive before its defined time origin. See the S-parameter causality guide for how to tell an origin error from finite-band ringing or time wrap.

Manual Port Extension: Verify the Phase Slope First

  1. Define the measurement first. Mark the coax calibration planes and DUT planes on the fixture drawing. Keep the sweep band, cable position, calibration settings and raw traces in the record.
  2. Measure a known reflection at the intended DUT plane. Unwrap the phase over usable points and account for the standard’s own phase if the manual method supports it. Do not fit across an unexplained notch, wrap or abruptly changing slope.
  3. Set the one-way delay on the correct port. Use the actual unit, such as ps rather than ns. If converting a physical length, use the line’s propagation velocity, not the speed of light in air; for a dispersive line, check that a constant delay is acceptable over the band.
  4. Start with delay only. Compare raw and corrected phase and magnitude at identical sweep settings. If magnitude moved, check whether loss compensation or another operation is enabled. Then verify with a different standard.

Automatic Port Extensions: When an Open or Short Is Good Enough

On the applicable Keysight implementation, the automatic function estimates delay from an open, a short or both, with an optional loss fit. It assumes an ideal standard in this operation. A populated footprint is not an open. Even an unpopulated footprint can have fringing capacitance; a short with a long return loop adds inductance. A seemingly clean fit can include the electrical length of the standard itself.

Use standards with reproducible geometry at the intended device plane. Record the open and short estimates before averaging or accepting the result. If they differ enough to exceed your phase error allowance, inspect the termination geometry and launch mismatch, or change the fit band with justification and verify across the full required band. Do not use the same open both to fit and independently approve the correction.

Loss Compensation: What the VNA Can and Cannot Correct

For an ideal matched-line example with 0.18 dB one-way fixture loss at port 1 and 0.25 dB at port 2 at 3 GHz, the correction is +0.43 dB to S21 and +0.36 dB to S11. Those sums explain a line loss adjustment; they are not fixture specifications. A mismatched launch can interact with the DUT reflection. A smooth loss curve cannot cancel that interaction, even if it makes the average amplitude look right.

Keep the loss fit disabled unless you can verify its improvement separately. If the corrected open or short shows substantial magnitude ripple around its expected response, investigate the fixture mismatch, standard and calibration before using the data. Avoid accepting a trace simply because an instrument option shifted every sample below 0 dB. That does not show that the two-port network is physically correct.

VNA System Verification: A Check Standard You Did Not Fit

Use the following worksheet after changing a reference plane. Fill in the band, known standard, separate limits for phase and amplitude, and the measured residual for each. Use an independent check short or load for reflection and a characterized verification line or through for transmission, as appropriate. A load can check reflection magnitude and residual match; its small reflection makes it a poor phase reference. A nonzero-length through must be compared against its known delay, not a flat zero-degree trace.

Corrected short phase comparison showing an example in-budget trace and a residual ramp that exceeds the example phase budget
Illustrative phase residual comparison, not measured data. Judge the corrected trace against an independent reference and the applicable budget.
Field Example Decision
Band and test 1–10 GHz; independent short at port 1 DUT plane. Save raw and corrected S11 and the short model.
Phase residual Example limit: ±5°. Example result: +7° at 10 GHz. Hold data. Refit or change method and verify again.
Other checks Enter magnitude error and reconnection spread. Do not accept on phase alone.

If the entire ±5° phase allowance at 10 GHz were assigned to a pure delay error, S21 would permit only 1.39 ps of total delay error; S11 would permit 0.694 ps of one-way error. In a real budget, allocate part of the allowance to the calibration, standard and repeatability as well. The example is not an IPC acceptance limit; use the applicable drawing or agreed measurement procedure.

PCB Test Fixture: Where Port Extension Is Useful

Filter coupon: preserve the filter, remove the feeds

Put the DUT planes at the filter’s port pads when comparing against a pad-referenced model. Determine the one-way delay of both feed sections. Verify with a known through or line before installing the filter. Accept the result only if the reference agrees with phase and loss budgets and the measured filter passband, stopband and group delay meet the project limits. Do not fit the fixture correction on the filter passband and then use the same fit as proof of filter delay.

PCB antenna: report at the feed you intended to tune

If the matching network is specified at its antenna feed, set the port 1 plane there, leaving the antenna and intended matching elements in the measured network. Verify with a separate short or load on a matching coupon, then check the frequency and depth of S11 minima and the usable band against the antenna requirement. A delay-only correction rotates the impedance display but cannot improve return-loss magnitude. If the connector is the specified product test plane, measure and report there instead.

Mismatched connector launch: distinguish the two failure modes

If the launch is itself the product feature under test, leave it inside the DUT boundary and compare to the connector-plane mask. If it is only a fixture and its ripple dominates the error budget, validate a fixture model or calibrate at a closer plane. In both cases retain raw and processed files, calibration identifiers, port directions and the verification result. If there is no check standard or equivalent validation, label the trace as diagnostic instead of releasing it as verified device data.

VNA Port Extension FAQs

Is port extension a replacement for SOLT calibration?

No. Calibrate first at an accessible plane. The extension then moves the reference through a section modeled separately; it does not solve the same systematic error terms as calibration.

Does port extension improve S11?

Delay-only correction changes phase, not magnitude. A reported magnitude change may come from enabled loss compensation or a different processing step, not an improvement of the hardware.

Can port 1 and port 2 use different delays?

Yes. Fit and record each separately. S21 and S12 cross both sections; each reflection uses its own section twice.

Why did my open and short give different delays?

The standards may be nonideal or located at different planes, or fixture mismatch may change the fit. Inspect the geometry and compare the estimates against the phase budget.

Can I use the delay shown by a NanoVNA directly?

Check the software and firmware manual for whether the displayed value or setting represents one-way line delay, round-trip delay or a trace-level phase correction. Verify it on a known length before transferring settings between instruments.

Should I average delay across a wide band?

Only where the fixture behaves as the assumed model and the independent reference meets the required full-band residual. A narrower fitting band does not by itself validate operation outside it.

Does time gating solve the same problem?

No. Gating suppresses selected time events in the processed response. Port extension moves the reporting plane; it does not selectively remove an echo. Keep the unprocessed data either way.

What goes into the Touchstone delivery notes?

Include the original and output planes, port order and reference impedance, sweep band, one-way delay by port, loss and other processing settings, and the verification evidence.

Technical References

Custom RF PCBs With Stackup and Test Coupon Review

Ordering an RF PCB or a separate test coupon? Send EBest Circuit your intended stackup, controlled-impedance targets, launch drawing and test-plane locations before releasing the fabrication data. An early review gives your fabrication and lab teams the same geometry to reference when they compare board and coupon results.

Request an RF PCB quote from EBest Circuit with your Gerbers or ODB++, quantity and required delivery date.

S-Parameter Causality: Is the Early Pulse in Your Model Real?

November 17th, 2026

If a PCB channel model seems to respond before the input, check the reference plane, measured band and frequency step before repairing the file. A real transmission response cannot precede its excitation, but a finite-band plot can show small oscillations around a valid arrival. A late echo can also fold into negative time on a centred display. Change one setting at a time and confirm the result on an independent reference before deciding the model is noncausal.

An RF PCB coupon with coaxial connections in front of a network analyzer
Capture the measurement setup and reference planes before comparing transformed or fitted time-domain data.

What Is S-Parameter Causality?

A causal network does not respond to a future input. For a correctly referenced linear time-invariant model, the physical impulse response is zero before the excitation at the input reference plane. For a transmission path a known physical delay can give a more specific expected onset, although estimating it requires the actual path, propagation model and reference planes. A reflection at the calibrated input plane can legitimately begin near zero.

Passivity and reciprocity test different properties. The S-parameter passivity check examines whether a passive model creates net power, not whether it responds too early. A small negative group delay over a limited band is likewise not, by itself, proof of noncausality. Use the full measurement and modelling context, rather than one plot attribute.

S-Parameter Impulse Response: Why a Causal Path Can Show Pre-Ringing

An inverse Fourier transform of a measured sweep works with finite bandwidth. Limiting the spectrum multiplies it by a frequency-domain window; in time this convolves the actual impulse with the window’s time response. A sharp truncation can create sidelobes before and after a causal pulse. A softer window often suppresses sidelobes but broadens the main pulse, so nearby features become harder to separate.

Keep the time origin and original data fixed. Compare the early response with a known line or thru using the same processing and record how it changes under reasonable windows and measured spans. Small features near the main arrival that vary as expected with the window are consistent with truncation. Do not call them real PCB events or declare the file good solely because one strong window hides them.

Illustrative impulse showing sidelobes on both sides of the expected main arrival after band limiting
Schematic only, not measured data. Finite-band processing can put sidelobes ahead of the main arrival without making the real device noncausal.

Frequency Step Size: A Late Echo Can Appear at Negative Time

On a uniformly spaced frequency grid, the inverse-transform time record repeats with period T = 1 ÷ Δf. Span and step do different jobs: a larger measured bandwidth can separate closer events, while a smaller frequency step extends the record. Zero-padding or interpolating the same measured samples does not supply independent frequency points or prove that a late event is inside the measured record. If your solver needs a different frequency grid, use this S-parameter interpolation guide to compare Cartesian and polar methods against withheld measured points.

Worked example: assume a delayed echo at +8.8 ns. With a 100 MHz step, T is 10 ns and a centred display covers −5 ns up to +5 ns. The echo appears at 8.8 − 10 = −1.2 ns in that period. Re-measure the same band on a 25 MHz grid and T becomes 40 ns; the centred record goes from −20 ns up to +20 ns and can show the echo at +8.8 ns. These times illustrate ideal periodic mapping, not measured results.

A plus 8.8 nanosecond example echo appears at minus 1.2 nanoseconds in a 10 nanosecond centred period and at its positive time in a 40 nanosecond period
Keep bandwidth comparable and physically reacquire the finer grid. Do not infer additional measured time range from display interpolation.

VNA Time Domain: Three Lookalike Faults and Their Checks

Put separate tests around three questions: does the feature track the frequency window, does it move with the record period, or does it first appear during model processing? Perform the changes on a saved copy and verify instrument calibration on the actual sweep grid. Every new data record needs an identifiable setup and reference.

What you see Check that separates it What the result supports
Small features clustered either side of the main arrival Compare two reasonable windows and a known line at the same reference planes. A change in spacing or sidelobe amplitude consistent with the reference supports a finite-band artifact.
Isolated early pulse Re-measure at the same span with a finer frequency step, keeping the setup comparable. A feature moving by an amount consistent with the repeat period supports a folded response.
Entire response moves after port extension or fixture removal Compare raw and processed stages against a known path and check removed delay and sign. Hold the processed model until time origin and fixture assignment agree.
Early feature persists in a physically implausible location after controlled changes Compare independent reference measurements and validate the model method and settings. Do not dismiss it as ringing; isolate the first failing stage and do not release unverified data.

S-Parameter Reference Plane: Check the Delay You Removed

Keep a sketch from the VNA calibration planes to the DUT planes, with port numbers and the fixture on each side. Compare the untouched sweep with each processed file. On a transmission path a simple delay correction changes phase slope and shifts time. If the applied correction removes more delay than the physical fixture contributes, a correctly measured arrival can appear too early in that processed file.

For example, if an arrival is 1.4 ns from the raw reference and the documented combined fixture delay is 0.6 ns, a delay-only correction would leave about 0.8 ns in this illustrative case. Removing 1.6 ns instead would place it at about −0.2 ns. Real de-embedding can also change amplitude and reflections, so do not treat every fixture as a pure delay. Verify on a reference, and check units and port extensions before editing the DUT model.

If the delay was removed at the instrument, use the VNA port extension guide to check each port’s one-way delay and verify the shifted plane with an independent standard before judging a time-domain arrival.

DC Extrapolation: Use the Physical Low-Frequency Path

A sweep starting above DC does not establish what happens at DC. A through connection with a resistive path and an otherwise isolated series-capacitor path have different low-frequency limits. Use the schematic, DC resistance when relevant, and a suitable low-frequency reference to constrain the extrapolation. Do not assume S21 = 1 or extend the first sample flat for every PCB network.

For a serial interconnect, compare transient results under justified low- and high-frequency continuations, including any actual coupling components. For a DC-blocking filter, use a low-frequency model that preserves the open series path; a convenient flat extrapolation can create a misleading step baseline. In both cases, report which bands were measured and which were estimated. If the outcome changes beyond the project tolerance within plausible assumptions, limit what the model may be used for or remeasure.

Vector Fitting S-Parameters: Compare the Fit With Original Measurements

A well-formed stable, proper rational model with a physically appropriate nonnegative delay can be a useful causal representation. But a fit forces structure on the data. It cannot prove the real PCB matches the result, especially outside the measured band. Do not “fix” a large early pulse by zeroing the response before an arbitrary time; changing the time waveform also changes the frequency response.

For a backplane or cable launch, verify the fit against complex S-parameters, expected delay, and the transient or eye metric used by the system team. For a narrowband RF filter, preserve the wanted passband transmission and phase as well as the stopband range your fixture can establish. For a passive channel, run a separate passivity check on the resulting model and inspect the errors after any enforcement. Reject or limit a fit that meets numerical criteria only by losing a response the application depends on.

S-Parameter Quality Check: What to Sign Off Before Handoff

Use this worksheet when sending a model to a designer or accepting it from a vendor. Fill in the measured band, actual point spacing, reference planes, expected arrival, and the early-time interval used for review. State how acquisition, repeatability and processing affect the decision. There is no universal percentage of pre-arrival energy that applies regardless of bandwidth and window.

Record Evidence to include Action if missing or inconsistent
Measured sweep and origin Device and fixture IDs, planes, grid, settings, and reference device trace. Hold timing comparisons until another engineer can reproduce the origin and band.
Early feature example 100 MHz grid shows −1.2 ns; a qualified 25 MHz remeasurement confirms a corresponding echo at +8.8 ns. Record the original reading as a display wrap candidate. Confirm the actual echo on the extended record before releasing the original timing claim.
Processing and intended use Window, extrapolation, fixture removal, fit method, residuals and intended application and band. Do not generalize the fit to unqualified conditions. Refit, remeasure or set limits on use.

IEEE 370-2020 covers electrical characterization and measured-data quality for printed circuit boards and related interconnects up to 50 GHz. If invoked, apply the agreed method and derived acceptance criteria; the case values above are teaching examples, not IEEE thresholds or PCB workmanship limits.

S-Parameter Causality FAQs

Does every sample before time zero mean a noncausal PCB?

No. Check finite bandwidth, record wrap, reference plane and the measurements before judging the model.

How do I tell ringing from time wrap?

Compare windows on a fixed data set for ringing. Re-measure with a finer frequency step and similar span to test whether an event moves as a folded echo should.

Does a wider span always increase time range?

No. Time-record period is set by frequency step. At fixed point count, a wider span generally increases the step and shortens the record.

Does zero-padding fix missing measurements?

No. It may increase displayed sample density, but does not add information or replace actual low-frequency or finer-grid measurements.

Can a wrong port extension make a response appear early?

Yes. Excess delay removal can shift an arrival too far left. Recheck the fixture map and a known reference before judging the network.

Must the file contain a measured DC sample?

No, but a simulation sensitive to the DC limit needs a physically justified low-frequency continuation and a stated range of validity.

Does a causal fit prove a passive channel model is acceptable?

No. Separately check passivity, complex in-band error, physical delay and performance in the intended circuit.

Should I trim all pre-arrival samples before simulation?

Not as a first step. Find the cause before changing the waveform and verify any processed result against the original complex data.

Technical References

Custom High-Speed PCBs With Stackup and Coupon Review

Have a high-speed PCB build that depends on a trustworthy channel model? Send EBest Circuit your layout files, stackup, target impedance, connector or fixture assumptions and the coupon requirements. Request a fabrication quote and stackup review to define the build and verification scope before ordering.