The screen was full.
Sixteen sub-windows. Each showing a different parameter. Green traces sweeping from 10 MHz to somewhere near 67 GHz. Smith charts scattered across the grid — circular, precise, each one a compressed description of how the device under test handles impedance at every frequency in the sweep.
The Rohde & Schwarz ZNA Vector Network Analyzer — rated 10 MHz to 67 GHz — was connected to a multi-port RF test fixture via four phase-stable orange/yellow coaxial cables. In the center of the fixture: a gold MMIC die, bare, seated in a precision carrier, its bond wires connecting it to the fixture's microstrip launch structures.
Four ports. Sixteen S-parameter combinations. All running simultaneously.
It looked complete.
It was complete — for what a VNA measures.
The question worth asking: what does a VNA actually measure?
What the screen showed
The ZNA display was arranged in a 4×4 grid of trace windows. Reading across the visible softkey menu on the right: S-Params, Noise Figure, Gain Compression, Y/Z-Parameters, Time Domain, Power Spectrum.
The active mode was S-parameters. The visible traces included:
→ S11, S21, S31 — reflection from port 1, forward transmission to ports 2 and 3 → S12, S22, S32 — reverse transmission from port 2, reflection at port 2 → Multiple Smith charts showing impedance loci — the circular plots that compress frequency-swept impedance into a single 2D representation
The sweep start: 10 MHz. The upper frequency — readable from the ZNA front panel label: 67 GHz.
That frequency range covers everything from the lower end of cellular bands through 5G mmWave at 28 GHz, E-band at 60–90 GHz approaching, and all of the intermediate microwave bands in between.
The timestamp in the lower right corner: 9/25/2024, 9:15:52 AM. A morning measurement session, live at the show.
The DUT on the left: a multi-port MMIC in a test fixture. The gold die visible at the center — approximately 2–3 mm on a side, judging by the connector scale — had multiple bond wire connections to the fixture. This is the kind of device that ends up inside phased array antenna modules, radar front ends, or wideband receiver chains.
What S-parameters are — and what they are not
The ZNA was measuring S-parameters. Most RF engineers know what S-parameters are. Fewer think carefully about what they are not.
An S-parameter is a ratio.
S21 is the ratio of the signal coming out of port 2 to the signal going into port 1, with all other ports terminated in the reference impedance (typically 50 Ω). It describes transmission gain or loss between those two ports under those specific conditions.
S11 is the ratio of the signal reflected back from port 1 to the signal injected into port 1. It describes how much of the input signal bounces back — the reflection coefficient, expressed as a complex number carrying both magnitude and phase information.
The Smith chart is a graphical representation of the reflection coefficient mapped onto a normalized impedance plane. Every point on the chart corresponds to a specific complex impedance at a specific frequency. The trace sweeping across the chart as frequency increases tells you how the device's impedance changes across the measurement band.
All of this is precisely defined. Rigorously calibrated. Reproducible.
But it is all conditional.
Every S-parameter measurement assumes: → The device is linear — it behaves the same regardless of signal level → The device is time-invariant — its characteristics don't change during the measurement → The system is reciprocal — or if not, the non-reciprocity is correctly characterized → The port terminations are exactly 50 Ω → The calibration is valid at the measurement temperature and cable configuration
In the ZNA demo, these conditions were controlled and met.
In your application, some of them will not be.
The gap between measurement and reality
The MMIC in the fixture on the demo bench operates linearly at the small signal levels used by the VNA — typically −10 to −30 dBm per port.
In a real system, that same MMIC might be driven at 0 dBm, +10 dBm, or higher — into gain compression, where S21 starts to depend on input power, and where the simple linear S-parameter model begins to break down.
The ZNA has a Gain Compression measurement mode — visible in the softkey menu. It can characterize the 1 dB compression point and the AM-AM / AM-PM distortion that occurs when the device leaves its linear region. But that is a different measurement, requiring a different setup. It does not appear automatically in the S-parameter sweep.
Similarly: the Smith chart shows impedance at the test fixture's reference plane. In a real PCB, the reference plane is wherever the board trace connects to the device. The transformation from fixture reference to board reference requires de-embedding — a calibration step that accounts for the electrical length and impedance of the launch structure between the measurement reference and the actual device terminals.
Get the de-embedding wrong by a fraction of a millimeter at 67 GHz — the wavelength is approximately 4.5 mm in free space — and the Smith chart moves measurably. What looked like a well-matched input becomes a mismatch that costs you 1–2 dB of insertion loss in your final design.
→ The measurement is precise. → Its applicability to your specific situation depends on how carefully you've mapped the measurement conditions onto your design conditions.
Why 4 ports, and why the full matrix
The ZNA in this demo was running a full 4-port measurement — all 16 S-parameters simultaneously. This capability matters for several reasons that go beyond convenience.
Crosstalk characterization: In a multi-channel MMIC — a 4-port device might be a 2×2 MIMO front end, a dual-channel LNA, or a 4-port switch — the off-diagonal S-parameters (S31, S41, S32, S42) describe the isolation between channels. These are often the most important parameters for system performance, and the hardest to measure accurately, because they involve small signals in the presence of large ones.
Full reciprocity verification: A passive device should satisfy Sij = Sji (reciprocity). Measuring the full matrix lets you verify this — and deviations from reciprocity often indicate calibration problems, cable movement, or actual non-reciprocal behavior in the device (ferrite components, active devices).
Time domain analysis: The ZNA's Time Domain measurement mode (visible in the softkey menu) applies an inverse Fourier transform to the frequency-domain S-parameter data to produce a time-domain reflectometry (TDR) view. This lets you localize impedance discontinuities along a transmission line — seeing, in the time domain, exactly where a mismatch is occurring in a multi-section structure.
The ZNA running 16 traces simultaneously is not showing off. It is completing the characterization in one measurement pass, eliminating the possibility that cable movement or temperature drift between sequential single-parameter measurements introduces errors into the comparison.
The Rohde & Schwarz ZNA in context
The ZNA is R&S's mid-to-high-end VNA platform — positioned between the more accessible ZNL/ZNLE series and the top-tier ZNA67 with optional extensions to higher frequencies.
The 67 GHz upper frequency of this unit covers: → All 5G FR2 bands (24–52 GHz) → E-band lower edge (60 GHz) → V-band (40–75 GHz) for satellite and point-to-point links → Automotive radar at 77 GHz — reachable with the optional extension
The 4-port configuration with direct-access port architecture (the labeled covers on each port) allows connection of external bias tees, switches, or signal combiners directly at the port, without the additional cable length that would degrade calibration accuracy.
The phase-stable test cables in the photo — orange/yellow, semi-rigid style — are a necessary part of the measurement chain. At 67 GHz, a cable that flexes between calibration and measurement introduces a phase error that shifts every point on every Smith chart. Phase-stable cables maintain their electrical length regardless of mechanical position.
What the full screen of traces actually tells you
Back to the 16-window display.
It tells you, precisely and reliably:
How this device, under small-signal linear conditions, at these specific port terminations, with this calibration, on this day at 9:15 AM, transforms RF signals between each pair of its ports across the frequency range 10 MHz to 67 GHz.
That is an enormous amount of useful information.
It does not tell you:
How the device behaves when driven hard. How it behaves when the power supply voltage varies. How it behaves when temperature shifts by 30°C. How it behaves when the adjacent channel is transmitting and the isolation is finite. How it behaves when the PCB ground plane has a resonance at 43 GHz that nobody expected.
None of that is a criticism of the ZNA or of S-parameter measurement. All of those additional behaviors can be characterized — with different measurement modes, different setups, different conditions.
But they require you to ask the right question first.
The ZNA gives you a complete answer to the question you asked.
Whether you asked the right question is on you.
The Smith chart as philosophy
One more observation about the Smith charts on the screen.
A Smith chart is one of the most information-dense visualizations in engineering. A single point encodes two numbers — real and imaginary impedance — normalized to a reference. A swept trace across the chart encodes the full frequency-dependent impedance behavior of a device or network.
But a Smith chart is also a projection.
It maps 3-dimensional information (real part, imaginary part, frequency) onto a 2-dimensional surface. Frequency is implicit — you know it changes as the trace moves, but you can't read the frequency directly from the chart position without cursor markers.
This is the general condition of all S-parameter measurement:
You are seeing a projection of the device's behavior onto a specific measurement framework.
The framework is well-defined, mathematically rigorous, and enormously useful.
But it is a framework. Not the device itself.
The device exists in physical space, with real materials, real temperatures, real nonlinearities, and real interactions with everything around it.
The ZNA shows you the shadow.
A very precise, very well-calibrated shadow.
All photos: Thomas · @SignalByThomas