RF Measurement and VNA Workflow

A repeatable workflow for defining reference planes, calibrating a VNA, and preserving credible RF evidence.

RF · Measurement

A radio-frequency (RF) result belongs to the device under test (DUT) and the complete measurement system: setup, calibration, stimulus, receiver, processing, environment, and uncertainty. Preserve enough context to reproduce the result.

Start with the decision

Before connecting the DUT, record:

  • the question and pass/fail condition,
  • DUT serial number, revision, firmware, and operating state,
  • stimulus level, frequency range, waveform, and source impedance,
  • electrical reference planes,
  • sweep, bandwidth, detector, averaging, and correction settings,
  • expected result and damage limits,
  • calibration or de-embedding method,
  • important uncertainty contributors.

Instrument screenshots are useful context, but raw data and saved instrument state are the evidence.

Choose the instrument for the quantity

Instrument Primary question
vector network analyzer (VNA) How does a linear network reflect and transmit magnitude and phase?
Spectrum or signal analyzer What power exists versus frequency, time, or modulation state?
Power meter What is traceable average or peak RF power?
Oscilloscope What voltage exists versus time within the probe and bandwidth limits?
Noise-figure analyzer How much SNR degradation does the device add?
Signal generator How does the DUT respond to controlled RF stimulus?

No single instrument replaces the others. A VNA result is not automatically valid for a nonlinear or time-varying DUT.

Prove safe input and usable dynamic range

Predict power at every test port before connecting the DUT. The wanted response must remain above the configured noise floor with margin, below receiver compression with margin, and below every absolute damage limit:

\[P_{\text{noise}} + M_{\text{low}} < P_{\text{rx}} < P_{\text{compression}} - M_{\text{high}}\]

Suppose a driven DUT can produce +10 dBm at the VNA receiver, but the selected receiver path must remain below -5 dBm for the required linearity. At least 15 dB attenuation is needed at the receiver reference plane, plus deliberate headroom for gain tolerance, mismatch, transients, and configuration mistakes. Enter the attenuator in the calibration or correction model and verify that the weaker response still clears the noise floor.

Check the operating window experimentally:

  1. Terminate or isolate the receiver and record the configured noise floor.
  2. Use a known thru or check device to verify calibration against a predefined tolerance.
  3. Repeat the DUT measurement at a second source power. A linear DUT’s corrected $S$-parameters should remain stable within the uncertainty and repeatability allocation.
  4. Add attenuation or external protection before testing gain, conversion, or unknown output power.

A visually clean trace can still be receiver noise, source leakage, or compression. Dynamic range is a property of the configured measurement, not only the instrument data sheet.

VNA workflow

  1. Select ports, connectors, adapters, cables, sweep, source power, IF bandwidth, and calibration method.
  2. Inspect and clean mating surfaces. Use the specified connector torque.
  3. Let the setup reach thermal stability and arrange cables in their measurement position.
  4. Calibrate at the intended reference planes.
  5. Verify with a check standard or known device that was not used to solve the calibration.
  6. Connect the DUT without twisting connectors or moving the calibrated cable path more than necessary.
  7. Check linearity by repeating at a second source power.
  8. Repeat at least one disconnect and reconnect when connector repeatability can affect the result.
  9. Save Touchstone data, instrument state, calibration identity, and setup photographs.

Verification catches bad standards, incorrect cal-kit definitions, damaged cables, and connection mistakes that a successful calibration routine cannot detect by itself.

Convert reflection quantities

For a one-port measurement with reflection coefficient $\Gamma = S_{11}$:

\[\mathrm{RL} = -20\log_{10}|\Gamma|\] \[\mathrm{VSWR} = \frac{1 + |\Gamma|}{1 - |\Gamma|}\] \[\mathrm{ML} = -10\log_{10}\left(1 - |\Gamma|^2\right)\]
If $ \Gamma = 0.1$, return loss is 20 dB, VSWR is approximately 1.22:1, and mismatch loss is approximately 0.044 dB. These quantities describe the same mismatch from different viewpoints. They do not include dissipative insertion loss or uncertainty from the measurement setup.

Insertion loss for a passive two-port path is commonly reported as:

\[\mathrm{IL} = -20\log_{10}|S_{21}|\]

Illustrative two-port decision

Suppose a 2.4 GHz interconnect has these requirements at the calibrated reference planes:

  • insertion loss no greater than 1.0 dB,
  • input return loss at least 15 dB,
  • no compression across the intended stimulus range.

The setup uses a two-port calibration at the cable ends, a verification thru not used in the calibration, fixed cable placement, and one reconnect. The illustrative result is:

Quantity Result Expanded uncertainty Guarded result Decision
Insertion loss 0.62 dB 0.12 dB upper bound 0.74 dB pass against 1.0 dB
Return loss 18.4 dB 0.8 dB lower bound 17.6 dB pass against 15 dB
Reconnect delta 0.05 dB maximum included above within repeatability allocation pass
Source-power check less than 0.03 dB change not a compliance result no observed compression linearity check passes

This is a measurement-decision example, not DUT data. A real report must state frequency span, IF bandwidth, point count, power, averaging, calibration kit, connector type, uncertainty method, and raw Touchstone file.

Know the calibration boundary

Calibration corrects the systematic errors represented by its error model. It does not remove:

  • incorrect calibration-standard models,
  • drift after calibration,
  • cable flexure,
  • connector contamination, wear, or repeatability,
  • random receiver noise,
  • receiver compression or DUT nonlinearity,
  • fixture variation or an invalid de-embedding model,
  • temperature change.

Short-open-load-thru (SOLT) calibration is convenient when accurate coaxial standards are available at the reference plane. Thru-reflect-line (TRL) calibration can be better suited to planar fixtures when suitable line standards and reference impedance are controlled. Choose the method from the fixture and uncertainty problem, not habit.

Spectrum measurements

  • Set attenuation and preamplifier state to avoid both front-end overload and an unnecessary noise penalty.
  • Use span, resolution bandwidth (RBW), video bandwidth (VBW), detector, and sweep time that resolve the phenomenon.
  • Check how the displayed noise floor changes with RBW before treating it as DUT noise.
  • Measure source and instrument residuals before assigning a spur to the DUT.
  • Add external filtering when strong out-of-band energy can compress the analyzer input.
  • Record correction factors for cables, attenuators, amplifiers, and probes.

Build an uncertainty budget

At minimum, consider:

  • residual instrument error,
  • calibration-standard uncertainty,
  • cable and connector repeatability,
  • source and load mismatch,
  • drift and environmental change,
  • noise and trace processing,
  • external component calibration,
  • fixture and de-embedding error,
  • DUT repeatability.

For consequential claims, repeat calibrations and connections. A smooth trace does not imply low uncertainty.

Preserve the measurement package

measurement/
  README.md
  setup/
  instrument-state/
  calibration/
  raw/
  corrected/
  analysis/

The README should identify who ran the test, when it ran, DUT pedigree, firmware, temperature, cables, fixtures, calibration, reference planes, processing, and known anomalies.

Measurement and Instrumentation, RF PCB Layout and Signal Integrity, and Design Verification and Test.

Sources