RF test automation - VNA calibration drift between runs?

Automated VNA measurements drift between calibration runs, and re-calibrating every run kills throughput. How much drift do you tolerate, and how do you track it?

VNA calibration drift between automated runs is usually environment + connector wear + residual standards error, not the analyzer "forgetting" math.

A triage order that works in production cells:
1. Temperature: bridge, cables, and eCal blocks drift with a few C. Log ambient (and eCal case temp if you can) next to every cal; a cal done cold and a DUT run warm looks like the DUT moved.
2. Connectors: torque, pin depth, and mating cycles. A robot that over-mates or under-torques will walk the residual. Inspect / gauge the working port on a schedule; keep a golden thru as a check standard after each cal.
3. Cal interval vs check: full SOLT/TRL less often; insert a quick known thru / Beatty / airline check every N DUTs. Recal only when the check fails a guard band - recalibrating every run can add more connector wear than it removes.
4. Cable / fixture: strain relief and fixed bend radius; mark cable positions. Phase at high frequency is mostly geometry.
5. eCal / mechanical standards: characterize aging of the eCal against mechanical standards quarterly; do not assume the eCal is drift-free forever.

If you share frequency, cal type (eCal vs mechanical), and whether the fail is mag, phase, or both, people can usually say temp vs connector vs fixture.

— rfforum 编辑部 / Editorial Team