S-Parameter Theory — Physical Intuition, VNA, Smith Chart

// 01 — The Problem They Solve

Why S-Parameters?

At low frequencies (audio, power electronics) we describe circuits with voltage (V) and current (I). We connect a voltmeter, measure a voltage, and life is simple. At RF and microwave frequencies, this breaks down completely.

Here's the problem: at microwave frequencies, the probe you use to measure a circuit becomes part of the circuit. A voltmeter has some capacitance — at 10 GHz that capacitance is a short circuit that completely changes what you're measuring. And even if you could build a perfect probe, what does "voltage" even mean at a point where the wavelength is smaller than your circuit? Voltage is a path-dependent integral — it changes depending on which route you take from point A to point B.

S-parameters solve all of this by measuring something that behaves perfectly at any frequency: travelling waves of power. Instead of voltage and current, we measure how much of a wave gets reflected and how much gets transmitted. These are well-defined, measurable, and don't change based on how you make the measurement.

The key insight: At RF frequencies, we stop thinking about voltages at nodes and start thinking about waves travelling through pipes. S-parameters describe how your device handles those waves — how much it reflects, how much it passes, how much it amplifies. Everything else follows from that.

// 02 — Waves, Not Voltages

The Core Concept — Travelling Waves

Imagine your RF device — amplifier, filter, connector — sitting in the middle of two coaxial cables. Signals travel down those cables as waves. When a wave hits your device, three things can happen: some of it reflects back the way it came, some of it transmits through, and some of it gets converted to heat (loss). S-parameters simply measure the ratios of these waves.

Figure 1 — Waves at a two-port device. Incident wave a₁ enters Port 1. Some reflects back as b₁ (ratio = S11). Some transmits through to Port 2 as b₂ (ratio = S21). The rest is lost to heat. For a passive device the total must equal the incident power. An amplifier can have |S21| > 1 because it adds energy from its DC supply.

The S-Parameter Matrix — Physically

[ b₁ ] [ S11 S12 ] [ a₁ ]
[ b₂ ] = [ S21 S22 ] [ a₂ ]

S11 = b₁/a₁ when a₂=0 (wave reflected from port 1, with port 2 matched)
S21 = b₂/a₁ when a₂=0 (wave transmitted port 1→2, forward gain/loss)
S12 = b₁/a₂ when a₁=0 (wave transmitted port 2→1, reverse isolation)
S22 = b₂/a₂ when a₁=0 (wave reflected from port 2, output match)

Note: S-parameters are complex numbers — they have magnitude AND phase. The magnitude tells you how much; the phase tells you when it arrives.

// 03 — The Vector Network Analyser

How a VNA Actually Works

A Vector Network Analyser (VNA) is the instrument that measures S-parameters. It's one of the most important pieces of RF test equipment. Understanding how it works helps you understand what S-parameters actually are — and why calibration is so important.

Figure 2 — Inside the VNA. A signal source sweeps through frequencies. A directional coupler splits off the incident wave (reference receiver A) and the reflected wave (receiver B). The transmitted wave goes to receiver C. S11 = B/A. S21 = C/A. Taking ratios cancels out any cable loss or source variation — which is why VNA measurements are so accurate.

Calibration — The Most Important Thing Nobody Explains

Every VNA measurement you see has been calibrated. Without calibration, the numbers are meaningless. Here's why, and what calibration actually does.

The problem: The VNA's cables, connectors and internal switches have their own S-parameters. They add delay, reflection and loss that the VNA can't distinguish from your device. If the measurement reference plane is at the VNA port, but your device is 30 cm of cable away, you're measuring cable + device, not just device.

The solution (SOLT calibration): Before measuring the device, you measure three known standards at the reference plane — Short (total reflection, Γ=−1), Open (total reflection, Γ=+1), and Load (perfect 50 Ω, Γ=0). Then you measure the Thru (both ports connected). The VNA uses these to compute the error model and mathematically remove everything between its internal reference and the calibration plane. After calibration, the measurement plane is wherever you put your calibration kit.

Physical interpretation of calibration: The Short, Open and Load standards are three known points on the Smith chart (−1, +1, and 0). They're spread across the chart, which means they give the VNA enough information to triangulate where every other impedance sits — and to correct for the distortion that the cables and connectors were adding.

Never skip calibration. A VNA that hasn't been calibrated might show S11 = −15 dB for a perfect load (0 dB in reality). Cable quality, connector wear, and temperature drift all corrupt the measurement. Calibrate immediately before your measurement session, at the actual frequency range and power level you'll use.

// 04 — S11: Input Reflection

S11 — The Reflection Story

S11 is the ratio of the wave coming back out of Port 1 to the wave going in. It tells you how well your device "accepts" the incident wave. A good device has a low S11 — it absorbs most of the wave rather than bouncing it back.

S11 is a complex number — it has magnitude (how much reflects) and phase (at what angle the reflected wave comes back). Both matter. The magnitude alone gives you Return Loss. The full complex number gives you the impedance.

S11 — Key Conversions

Return Loss = −20·log₁₀(|S11|) dB (positive number, high = good match)
VSWR = (1 + |S11|) / (1 − |S11|) (>1, ideally close to 1.0)
Mismatch Loss = −10·log₁₀(1 − |S11|²) dB (power lost to reflection)

S11 = −10 dB → |Γ| = 0.316 → VSWR = 1.93 → 10% power reflected, 0.46 dB mismatch loss
S11 = −20 dB → |Γ| = 0.100 → VSWR = 1.22 → 1% power reflected, 0.04 dB mismatch loss
S11 = −30 dB → |Γ| = 0.032 → VSWR = 1.07 → 0.1% reflected — near-perfect match

S11 on the Smith Chart

The Smith chart is a map of all possible complex impedances, plotted in the Γ (reflection coefficient) plane. Every point on the Smith chart corresponds to a specific impedance AND a specific complex S11 value. They're the same thing, just expressed differently.

The geography of the Smith chart:

Figure 3 — The Smith Chart geography. Every impedance maps to one point. Centre = perfect 50 Ω match. Rim = total reflection (lossless reactive element). Left point = short circuit. Right point = open circuit. Upper half = inductive. Lower half = capacitive. Moving clockwise = adding transmission line length.

S11 for Different Devices — What to Expect

AMPLIFIER

S11 of an Amplifier

Typically −10 to −20 dB in the operating band. Has a pronounced dip at the frequency where the input matching network resonates. Outside the band, S11 rises back toward the rim (poor match).

On Smith Chart: Traces a loop — starting near the rim at low frequency, diving toward the centre near the design frequency, then spiralling back out at high frequency as the transistor's capacitance dominates.

Watch for: If S11 goes outside the unit circle at any frequency, the amplifier is potentially unstable at that frequency.

FILTER

S11 of a Filter

In the passband: S11 should be < −15 dB — filter is matched and passing power. In the stopband: S11 → 0 dB — filter is reflecting everything back (it's a reactive wall, not an absorbing load).

Physical meaning: When you put a signal into a filter's stopband, the filter doesn't absorb it — it throws it back at you. This has implications for what's upstream of the filter.

On Smith Chart: In passband, S11 near centre. In stopband, S11 hugs the rim (rim = total reflection).

CONNECTOR / CABLE

S11 of a Cable or Connector

Should be very small — < −25 dB for a good connector. Any resonance (series or shunt) shows up as a dip or bump in the S11 trace. A bad SMA connector might show −15 dB at the resonant frequency of the poorly-made outer conductor interface.

Use this: If your system S11 looks worse than the sum of its parts, plug in just the cable and measure it. Often the culprit is a damaged connector or a barrel adapter.

ANTENNA

S11 of an Antenna

An antenna S11 shows a deep dip at the resonant frequency — this is where the antenna is actually radiating efficiently. The dip represents the antenna impedance crossing through the 50 Ω circle.

S11 = −10 dB is the standard "antenna works" threshold. Better than −15 dB is good. The bandwidth of the dip tells you the antenna's operating bandwidth.

Key insight: A perfect short circuit or perfect open circuit doesn't radiate. An antenna works by being "just imperfect enough" to dump power into the radiation resistance rather than reflecting it back.

// 05 — S21: Forward Transmission

S21 — The Signal That Made It Through

S21 is the ratio of the wave coming out of Port 2 to the wave going into Port 1. It describes the forward transmission through the device. For an amplifier it's the gain. For a filter it's the passband response. For a cable it's the insertion loss.

S21 in dB: if positive → amplification. If negative → loss. The sign change is the clearest way to tell immediately whether you're looking at an active or passive device.

S21 for Different Devices

AMPLIFIER S21

Gain Flatness and Bandwidth

S21 should be positive dB (gain), flat across the bandwidth, and rolling off predictably outside. The 3 dB bandwidth of S21 defines the amplifier's usable frequency range.

What roll-off at high f means: The transistor's fT is being approached. The transistor can't respond fast enough to the RF signal — gain falls at −6 dB/octave (20 dB/decade) for a simple single-pole device.

What ripple in S21 means: Resonances in the matching network or printed circuit board. A small notch can indicate a parasitic resonance (via inductance + pad capacitance).

FILTER S21

Insertion Loss and Stopband

S21 for a filter shows the passband (S21 near 0 dB with small insertion loss) and stopband (S21 deeply negative dB). The transition between them is the filter's roll-off.

Insertion loss (IL): How much the filter attenuates the signal even in the passband. A SAW filter might have IL = 2 dB. An LC filter on FR4 might have IL = 0.5 dB at 100 MHz but 5 dB at 5 GHz.

Stopband rejection: How deep S21 goes in the stopband — typically measured at a specific offset frequency. The steeper the transition, the higher the filter order required.

CABLE S21

Insertion Loss vs Frequency

S21 for a cable shows a gradual rolloff — the cable is lossiest at high frequencies because of skin effect (loss ∝ √f) and dielectric loss (loss ∝ f). Plot S21 vs frequency and the slope tells you the dominant loss mechanism.

Rule of thumb for coax: RG-58 at 1 GHz: ~0.6 dB/m. At 10 GHz: ~2 dB/m. LMR-400 is much better: 0.23 dB/m at 1 GHz.

POWER DIVIDER S21

Split Loss

An ideal 3-way (power split) power divider has S21 = −3.01 dB. This isn't insertion loss — it's the unavoidable consequence of splitting the power equally between two outputs. You get half the power at each port.

Don't confuse split loss with insertion loss. A good divider has S21 = −3 dB ±0.5 dB across bandwidth, and S11 < −20 dB — meaning the input is matched even though only half the power exits each port.

// 06 — S12 and S22

S12 and S22 — The Ones Engineers Forget to Check

S12 — Reverse isolation: the leak backwards.
S12 is how much signal leaks from Port 2 backwards to Port 1. For an amplifier, a good S12 is very small — typically −20 to −40 dB. This matters because:

1. LO leakage: The local oscillator in your receiver is a strong signal. If it leaks back through S12 of the LNA and out of the antenna, you've turned your receiver into an unintentional transmitter — a regulatory violation.

2. Stability: If S12 is large and the output reflection is large (high S22), the amplifier can form an oscillation loop. S12 growing at high frequencies is always a stability warning sign.

S22 — Output reflection: what bounces off the output.
S22 is the match at Port 2. For a PA driving an antenna, S22 matters because reflected power from a mismatched antenna bounces back into the PA. Most PAs have protection circuits against high reflected power (SWR foldback), but keeping S22 small is still best practice.

Parameter	What it measures	Good value	Bad value — consequence
S11	Input reflection, input match	< −15 dB (VSWR <1.43)	> −10 dB: significant reflected power lost
S21	Forward gain or loss	As specified for application	Ripple > ±1 dB: resonances in design
S12	Reverse isolation	< −20 dB (amplifier)	> −10 dB: LO leakage, stability risk
S22	Output match	< −10 dB	> −5 dB: power loss, PA protection triggers

// 07 — Phase and Group Delay

Phase and Group Delay

S-parameters are complex — they have magnitude AND phase. The phase of S21 tells you how much the signal is delayed as it passes through the device. This is usually unimportant for simple signals, but becomes critical for wideband modulations like OFDM (WiFi, 5G, LTE) which require constant group delay across the channel bandwidth.

Group Delay — The Physical Meaning

Group Delay τ = −dφ/dω = −dφ/(2π·df) [seconds]

where φ is the phase of S21 in radians and f is frequency in Hz.

Constant group delay: All frequency components of the signal experience the same time delay → signal shape preserved (zero distortion)
Varying group delay: Different frequencies arrive at different times → pulse spreading, ISI, degraded EVM in digital modulations

Typical: Good LNA: ±0.1 ns GD variation. SAW filter: ±2 ns. Problematic: >5 ns variation across channel BW for 256-QAM

Physical picture of group delay: Imagine a drum beat (wideband pulse) going through a device. If all frequencies arrive together, you hear a clean beat. If low frequencies arrive early and high frequencies late (or vice versa), the beat is smeared into a "whoosh". That smearing is group delay variation. For a data signal, it means bits bleed into each other — inter-symbol interference.

// 08 — Touchstone Files

Touchstone Files — The Format Behind the Datasheet

When a component manufacturer measures their device on a VNA, they export the S-parameter data as a Touchstone file (.s2p for two-port, .s1p for one-port). This is the universal RF data format — every RF simulation tool and VNA in the world reads and writes it.

Touchstone .s2p File Format

# Hz S MA R 50    ← Option line: Hz or GHz, S-params, MA=magnitude-angle (or RI=real-imag, DB=dB-angle), 50Ω reference
! Comment line
freq S11_mag S11_ang S21_mag S21_ang S12_mag S12_ang S22_mag S22_ang
1e9   0.245    −142.3 5.012    21.4   0.007    −55.1 0.189   −168.4
2e9   0.198    −161.5 4.867    19.8   0.009    −62.3 0.201   −174.1
...

At 1 GHz: |S11|=0.245 (−12.2 dB), |S21|=5.012 (+14.0 dB gain), |S12|=0.007 (−43.1 dB isolation)

Every component on the RFLab S-Parameter tools page accepts Touchstone files. You can download .s2p files directly from manufacturer websites (Qorvo, Skyworks, Minicircuits, Keysight have them all) and drop them straight into the plotter.

Upload .s2p and plot S11/S21/S12/S22 → Plot S11 on Smith Chart → Cascade two .s2p files →

// 09 — Real-World Examples

What Good vs Bad Looks Like

When you load a Touchstone file into a plotter, here is exactly what you're looking for — and what should make you nervous:

What you see in the plot	Physical meaning	Action
S21 flat ±0.5 dB across band	Excellent gain flatness — device works evenly across the band	Use it — this is what you want
S21 has a sharp notch (dip)	Resonance — parasitic LC in the package or PCB. Often a via or bond wire	Avoid that frequency or redesign the PCB land pattern
S11 < −20 dB across band	Excellent input match — VNA (or source) sees 50 Ω cleanly	No matching network needed
S11 approaches 0 dB at some frequency	Total reflection at that frequency — could be oscillation	Run stability analysis immediately
S11 > 0 dB at any frequency	Active reflection — the device is generating power back toward Port 1. Oscillation risk.	Dangerous — add stabilising resistor or pad
S12 rising steeply with frequency	Capacitive feedthrough growing — at some frequency S12 will be large enough to cause oscillation	Check K-factor at those high frequencies. Add series gate/base resistor.
S21 phase not linear (not a straight line)	Non-constant group delay — signal will be distorted for wideband modulations	Check group delay variation. Acceptable for narrowband; problematic for OFDM.
S22 goes above −5 dB	Poor output match — PA will see high VSWR from mismatched load	Add output matching network or circulator/isolator
S-params change between measurements	Loose connector, damaged cable, or device heating up	Re-torque connectors, check for damaged cables, recalibrate

// 10 — Try the Tools

Try the S-Parameter Tools

Every concept on this page has a live tool on RFLab. Upload any Touchstone .s2p file from a component datasheet and visualise everything described above.

📡

S-PARAMS

S-Parameter Plotter

Upload any .s2p Touchstone file and instantly plot S11, S21, S12, S22 vs frequency on the same chart. Hover any frequency to read off exact magnitude and phase values.

📡

S-PARAMS

Smith Chart Plotter

Plot S11 from any .s2p on an interactive Smith chart. See exactly where the impedance sits vs frequency. Find the matching frequency, check if S11 is inside or outside the unit circle.

📡

S-PARAMS

Amplifier Stability Analyser

Compute Rollett K factor, |Δ| and μ stability factor across all frequencies from your .s2p file. Identifies exactly which frequencies are potentially unstable.