I spent a morning chasing a Q value that moved every time I re-measured the same cavity resonator. The resonator was fine — my sweep settings were not. Here is what I learned.

First, sweep step size matters more than most people expect. A high-Q resonance is narrow, sometimes only a few hundred kHz wide at 3 GHz. If your step is wider than a tenth of the 3 dB bandwidth, the peak gets sampled off-center and the measured Q comes out low. I now set the number of points so I get at least 20 points across the resonance, then narrow the span around it.

Second, IF bandwidth trades noise against sweep speed, and both affect the result. A wide IF bandwidth adds noise to the trace and rounds off the peak; too narrow and the sweep slows down enough that drift creeps in during the measurement. On my bench, 1 kHz IF bandwidth with averaging turned on gives repeatable Q within 2% on a Q~5000 resonator.

Third, check the source power. Drive the resonator too hard and nonlinear effects shift the resonance and flatten the peak. I sweep power down until the measured Q stops changing, then back off another 3 dB for margin.

Fourth, calibrate at the resonator plane. A long cable between the calibration reference and the device under test adds phase ripple that distorts the resonance shape. Port extension helps, but a proper calibration at the device plane is worth the effort.

What is your rule of thumb for VNA settings on high-Q devices? I am curious how others trade off sweep time against repeatability.