R
Member
Every VNA measurement is only as good as its calibration, and the calibration is only as good as the kit. I learned this the hard way chasing a 0.2 dB ripple that turned out to be connector repeatability, not the DUT.
First, understand what repeatability really costs. Each mate and demate of a connector changes its electrical length slightly and its contact resistance a little more. At low frequencies this is negligible; above a few GHz it shows up directly in S11 and in phase. If your calibration standards are mated dozens of times a day, their effective characteristics drift from the values stored in the cal kit file.
Second, separate kit uncertainty from instrument uncertainty. A good practice is to measure a verification device — a stable airline or a known attenuator — right after calibration, and again after a batch of DUT measurements. If the verification result drifts, suspect the connectors before you suspect the DUT. I keep a log of verification measurements; a slow trend is far more informative than a single pass/fail.
Third, handle connectors like measurement instruments, not hardware. Use a torque wrench, keep caps on when not in use, clean with approved methods only, and retire standards that show visible wear. I also avoid mixing connector grades: a precision standard mated to a worn production connector contaminates the standard.
Fourth, budget the uncertainty honestly. For critical measurements, include connector repeatability in the uncertainty budget rather than quoting only the VNA datasheet number. It makes the result look worse on paper, but it stops arguments later when two labs disagree.
How do you track cal kit health on your bench? Do you re-verify after every calibration, or only when something looks wrong?
First, understand what repeatability really costs. Each mate and demate of a connector changes its electrical length slightly and its contact resistance a little more. At low frequencies this is negligible; above a few GHz it shows up directly in S11 and in phase. If your calibration standards are mated dozens of times a day, their effective characteristics drift from the values stored in the cal kit file.
Second, separate kit uncertainty from instrument uncertainty. A good practice is to measure a verification device — a stable airline or a known attenuator — right after calibration, and again after a batch of DUT measurements. If the verification result drifts, suspect the connectors before you suspect the DUT. I keep a log of verification measurements; a slow trend is far more informative than a single pass/fail.
Third, handle connectors like measurement instruments, not hardware. Use a torque wrench, keep caps on when not in use, clean with approved methods only, and retire standards that show visible wear. I also avoid mixing connector grades: a precision standard mated to a worn production connector contaminates the standard.
Fourth, budget the uncertainty honestly. For critical measurements, include connector repeatability in the uncertainty budget rather than quoting only the VNA datasheet number. It makes the result look worse on paper, but it stops arguments later when two labs disagree.
How do you track cal kit health on your bench? Do you re-verify after every calibration, or only when something looks wrong?
