S-Parameter Basics for RF PCB Engineers

S-Parameter Basics: What Are S-Parameters?

S-parameters (scattering parameters) describe how RF energy enters and exits the ports of a network—a filter, a trace, a connector, an entire PCB—by relating incident and reflected travelling waves. Understanding s-parameter basics is essential for every RF and microwave PCB engineer because these parameters are the primary language for verifying that a layout meets its performance targets. Unlike Z- or Y-parameters, S-parameters are measured with every port terminated in a reference impedance (almost always 50 Ω), so they can be captured directly on a vector network analyzer (VNA) without open- and short-circuit conditions that are impractical at microwave frequencies.

The Two-Port S-Matrix: S11, S21, S12, S22

Most PCB structures you will measure or simulate are two-port networks. The 2 × 2 scattering matrix relates the outgoing waves b to the incoming waves a:

b₁ = S11 · a₁ + S12 · a₂
b₂ = S21 · a₁ + S22 · a₂

Each element has a direct physical meaning, and grasping these four parameters is the core of s-parameter basics:

For a passive, reciprocal structure such as a microstrip line or a PCB filter, S21 = S12. An amplifier or circulator breaks this symmetry, so S12 and S21 differ—sometimes by 20 dB or more. Understanding these roles is closely related to how circulators and isolators enforce signal direction on an RF board.

Magnitude in dB: The Numbers You See Every Day

S-parameters are complex (magnitude and phase), but day-to-day RF specs are stated in decibels of magnitude:

|Sᵢⱼ| (dB) = 20 · log₁₀(|Sᵢⱼ|)

Because S-parameters are voltage ratios (not power ratios), the factor is 20, not 10. Some practical benchmarks that every engineer learning s-parameter basics should memorize:

Return loss is often quoted as a positive number by convention (e.g., “15 dB return loss” means S11 = −15 dB). Be careful with sign conventions in datasheets—always check whether the author means the magnitude of S11 or the negative of it.

Relating S11 to VSWR

VSWR and S11 carry the same information. The conversion is:

VSWR = (1 + |Γ|) / (1 − |Γ|), where |Γ| = 10^(S11 dB / 20)

An S11 of −10 dB corresponds to VSWR ≈ 1.92:1; −15 dB gives 1.43:1; −20 dB gives 1.22:1. Most RF system specs call for VSWR ≤ 2:1 (S11 ≤ −9.5 dB) at each interface. Seeing these relationships on a Smith chart makes the impedance behaviour at each frequency immediately visible.

Phase and Group Delay: Why They Matter on a PCB

The phase of S21 tells you the electrical length of the network. A uniform 50 Ω microstrip that is exactly one wavelength long at 5 GHz shows −360° of S21 phase at that frequency. Phase is critical for:

• Matching networks—a quarter-wave transformer must be exactly 90° at the design frequency.
• Phased-array feed networks—element-to-element phase error must stay within ±5° to keep sidelobe levels under control.
• Balanced / differential pairs—intra-pair phase skew converts differential signal into common mode.

Group delay, τ = −dφ/dω, reveals dispersion. A flat group delay means the structure passes all frequencies with the same time delay; a peaking or rippling group delay warns of resonances or impedance discontinuities that will distort wideband signals. In high-frequency versus high-speed design, group-delay flatness often separates an acceptable channel from a failing one.

Reading an S-Parameter Touchstone File

S-parameter data is exchanged in Touchstone (.s2p) format, defined by the IBIS Open Forum Touchstone 2.1 specification. A typical header line looks like:

# GHz S MA R 50

This declares frequency in GHz, S-parameters, magnitude-angle format, 50 Ω reference. Each subsequent row lists frequency, then S11 mag, S11 angle, S21 mag, S21 angle, S12 mag, S12 angle, S22 mag, S22 angle. Some files use RI (real-imaginary) or dB-angle format instead—always check the option line before importing.

When you drop a vendor-supplied .s2p model into your EDA tool—whether in an Altium RF design workflow or in ADS—the simulator treats the file as a frequency-domain black box and cascades it with the rest of your circuit. Garbage data in the file means garbage simulation results, so always sanity-check passivity (|S11|² + |S21|² ≤ 1 for a lossless or lossy passive) and causality before trusting the model.

How to Measure S-Parameters on a PCB

Equipment and Calibration

You need a VNA (two-port minimum), coaxial cables rated to your frequency of interest, and a calibration kit matched to your connector type—SMA to 18 GHz, 2.92 mm (K) to 40 GHz, 1.85 mm (V) to 67 GHz. Perform a full two-port SOLT (short-open-load-thru) or TRL calibration to move the reference plane to the cable tips. Mastering this calibration step is one of the most practical aspects of s-parameter basics.

De-Embedding Launch Connectors

The coaxial-to-microstrip launch adds its own S11 and insertion loss. To see only the PCB structure, you must de-embed the launches. Common approaches:

• Fixture removal (AFR / port extension)—uses a thru standard on the same board to mathematically remove the launch.
• TRL on-board cal—requires thru, reflect and line standards patterned on the DUT board; accurate but consumes board space.
• Simulated de-embedding—model the launch in an EM solver such as HFSS and subtract its S-matrix.

Without de-embedding, a 50 Ω line can appear to have S11 = −12 dB at 10 GHz simply because of the connector transition—misleading you into chasing a layout problem that does not exist.

Practical Measurement Tips

1. Set the VNA IF bandwidth to 1 kHz or narrower for low-noise traces; wider (10 kHz) is fine for broadband sweeps where speed matters.
2. Use averaging (8–16 sweeps) when measuring high-isolation paths (S21 < −40 dB).
3. Verify calibration quality: after calibrating, measure the thru standard and confirm S21 ≈ 0 dB ± 0.1 dB and S11 < −40 dB across the band.
4. Record the ambient temperature—Dk of PTFE laminates shifts roughly +125 ppm/°C, enough to move a narrowband filter by several MHz over a 40 °C swing.

Documenting these measurement conditions belongs in every RF PCB documentation package you hand off to manufacturing or your customer.

Using S-Parameters in RF PCB Simulation

Most RF EDA tools let you cascade, interpolate and optimise using .s2p or .snp files. Key workflows:

• Co-simulation—combine EM-extracted S-parameters of a PCB structure with SPICE or harmonic-balance analysis of active devices.
• Stability analysis—compute the Rollett stability factor K and |Δ| from the four S-parameters of a transistor to confirm unconditional stability (K > 1 and |Δ| < 1).
• Mixed-mode conversion—convert single-ended 4-port .s4p data into differential (Sdd) and common-mode (Scc) parameters to evaluate a differential pair or balun.

When choosing passive components, always prefer vendor-supplied S-parameter models over ideal lumped-element values. A 10 nH inductor that looks perfect at 900 MHz may self-resonate at 2.5 GHz and behave as a capacitor above that—something only visible in the S-parameter model. This is a core principle of sound RF component selection.

Common S-Parameter Pitfalls

• Confusing return loss sign—some tools plot S11 as a negative dB value, others flip it positive. A “return loss of 20 dB” and “S11 = −20 dB” are the same thing; “S11 = 20 dB” would mean gain from the reflected port, which is impossible for a passive device.
• Ignoring reference impedance—S-parameters are defined relative to a reference impedance. A 75 Ω system’s S-parameters cannot be compared directly to a 50 Ω system’s without renormalisation.
• Trusting extrapolated data—a .s2p file measured to 6 GHz tells you nothing about behaviour at 10 GHz. Never let a simulator silently extrapolate beyond the file’s frequency range.
• Neglecting port parasitics in simulation—if your EM port definition does not match the physical launch, the simulated S11 will not correlate with measurement.

Quick-Reference: S-Parameter Basics Checklist for Layout Review

1. Confirm S11 ≤ −15 dB (VSWR ≤ 1.43) at every interface across the operating band.
2. Verify S21 insertion loss is within budget: allocate loss per section (connector, trace, via transition, filter) and sum.
3. Check S21 phase linearity / group-delay flatness for wideband or modulated signals.
4. Compare EM-simulated S-parameters against vendor models for connectors and passives—flag any discrepancy > 2 dB.
5. Include .s2p files and measurement conditions in the design release package.

FAQ

What is the difference between S-parameters and Z-parameters?

S-parameters describe wave scattering with terminated ports and are measured directly by a VNA. Z-parameters describe voltage-current relationships with open-circuit ports. Both carry equivalent information and can be mathematically converted, but S-parameters are preferred above a few hundred MHz because open-circuit conditions are hard to maintain at high frequencies.

Can I use S-parameters for a three-port device like a circulator?

Yes. A three-port device has a 3 × 3 S-matrix (nine parameters), stored in a .s3p Touchstone file. The same s-parameter basics of return loss and isolation apply at each port.

How many frequency points should an S-parameter file have?

Enough to capture every resonance and transition. For a wideband passive up to 20 GHz, 2 001 points (10 MHz steps) is a practical minimum. Narrowband filters may need finer steps (1 MHz or less) around the passband edges. As noted in the Touchstone 2.1 specification, the file itself imposes no limit on point count.

Do S-parameters include connector effects?

Measured S-parameters include everything between the calibration reference planes. If you calibrate at the cable tips, the connector launch is inside the measurement. De-embed the launches to isolate the PCB structure alone.

What S11 value is “good enough” for an RF PCB trace?

For a 50 Ω transmission line, S11 ≤ −20 dB across the operating band is a reasonable target. At millimeter-wave frequencies (above 24 GHz), achieving −15 dB consistently is often considered acceptable due to tighter manufacturing tolerances and connector limitations.