A vector network analyzer measures how a circuit responds to a signal swept across a frequency range: how much comes back (reflection), how much gets through (transmission), and the phase of both. Everything else, from antenna tuning to filter design to finding a bad connector, is a specific application of that one measurement.
Most of the value is lost if the calibration is wrong. Start there.
Calibration Is Not Optional
A VNA measures at its ports, then math (error correction) moves that measurement to your device under test. That math only works if the analyzer knows exactly what a short, an open, a load, and a through connection look like at the plane where your device actually connects; not at the analyzer’s front panel, not at the end of a cable, but at the reference plane you’ll actually be measuring from.
SOLT calibration (short, open, load, thru) is the standard for most bench work:
- Connect the short standard, sweep, and let the analyzer capture it.
- Repeat for open and load.
- For two-port measurements, connect a thru between both ports and capture that too.
- The analyzer now has a full error model for everything between its measurement hardware and your reference plane.
The calibration plane is wherever you calibrated, not wherever you think it is. If you calibrate at the end of a cable, then add an adapter before connecting your device, that adapter’s loss and mismatch are now uncorrected error sitting in every measurement. Calibrate as close to the actual device connection as the calibration kit allows, and if you must add an adapter after calibrating, know that you’ve moved the reference plane and account for it.
Recalibrate after any cable change. A cable that’s been flexed, coiled differently, or swapped for a “identical” one has a different phase response. VNAs are sensitive enough that this matters for accurate S-parameters, not just for the paranoid.
Check the calibration before you trust the data. After calibrating, measure the load standard again as if it were a device. A well-calibrated setup should show return loss well below -40 dB. If it doesn’t, something in the cal was off, a bad connection, a worn standard, a loose torque wrench setting, and every measurement after it is suspect.
Return Loss and S11: Is the Match Good?
S11 is the ratio of reflected to incident power at port 1, usually shown as return loss in dB (more negative is a better match) or as impedance on a Smith chart.
This is the single most useful VNA measurement for hardware work. An antenna, a filter input, an amplifier input, anything with a target impedance (almost always 50 ohms) can be checked this way. -10 dB return loss is a common “good enough” threshold for many RF applications, corresponding to about 90% of power actually getting into the device. -20 dB or better is a tight match.
Common failure pattern: a board passes on the bench with a short test cable and fails intermittently in the field. Often the culprit is an impedance mismatch that a short cable partially masks (standing waves need distance to build up meaningfully) and a longer real-world cable run exposes. Measure S11 with the actual cable length and connector types you’ll ship with, not the shortest one on your bench.
Insertion Loss and S21: What Gets Through
S21 measures transmission from port 1 to port 2, the basis for characterizing filters, checking cable loss, and verifying an amplifier’s gain.
For a cable, S21 should be a small, smoothly decreasing negative number as frequency increases (more loss at higher frequency is normal; a sudden dip or ripple usually means a connector problem or an unintended resonance). For a filter, S21 across frequency is the entire point: it’s the passband, stopband, and rolloff, measured directly rather than inferred from a simulation.
Checking a suspect cable or connector: measure S21 through the cable in question against a known-good reference cable of similar length. A gap of more than a dB or two at your operating frequency, beyond what the datasheet loss-per-meter predicts, points to a bad connector or damaged cable rather than a design problem.
A Representative Example: Debugging a Mismatched Antenna Feed
Here’s the kind of case this measurement is built for. A board’s 2.4GHz antenna passes bench range tests but fails at the edge of spec in the field. S11 on the antenna feed, measured at the actual PCB connector with a cal kit good to that reference plane, shows -6 dB return loss at 2.4GHz instead of the expected -15 dB or better; a genuinely poor match, not a measurement artifact, confirmed by rechecking the cal against the load standard first.
Sweeping S11 across a wider band shows the actual resonance shifted about 80MHz low, consistent with a matching network component value drifting from nominal or a layout parasitic not accounted for in the original match calculation. Adjusting the matching network’s series capacitor value and re-measuring brings the resonance back on frequency and return loss to -18dB. The VNA measurement takes twenty minutes and points straight at the fix. The alternative, trial-and-error component swaps checked against range tests, runs a full day per iteration.
What a VNA Won’t Tell You
It measures linear behavior at the frequencies you sweep. It won’t catch nonlinear effects (intermodulation, compression) that only show up at real signal power, and it won’t tell you anything about a frequency you didn’t sweep across. A narrow sweep around your target frequency can miss a resonance or spur sitting just outside the range, which is why the debugging example above started with a wide sweep, not a narrow one centered on the expected frequency.
If you’re chasing an RF or impedance-matching problem and want a second set of eyes on the measurement, not just the design, that’s exactly what an independent review covers.