M
Member
Conclusion: use plain port extension only for a short, low-loss, near-ideal transmission line whose characteristic impedance matches the system impedance. Once the fixture has measurable loss, dispersion, non-50-ohm sections, launch asymmetry, or different through paths between ports, move to a characterized fixture model or S-parameter de-embedding. Port extension mainly rotates phase by an assumed delay; it does not remove fixture mismatch, cross-talk, mode conversion, or frequency-dependent loss. A common failure mode is a visually corrected S11 phase with still-wrong return-loss magnitude because the extension cannot cancel a real reflection from the fixture launch.
A practical diagnostic is to compare three checks after calibration: measure a known through, a precision load or airline at the intended DUT plane, and the same fixture with a short/open/load if available. If the through shows increasing insertion loss or ripple, if the load reflection does not stay near the center of the Smith chart across frequency, or if the required electrical delay changes with frequency, simple port extension is already outside its assumptions. For two-port fixtures, a 2X through-based de-embedding flow is usually more defensible than manually entering two independent delays, but it still depends on symmetry and a well-defined through standard. For asymmetric fixtures, characterize each half separately where the calibration topology permits, or use a validated fixture S-parameter model with proper passivity, reciprocity, and causality checks.
The reference direction is: first define the exact calibration plane and target DUT plane, then choose the correction method based on fixture complexity rather than convenience. SOLT/ECal at the cable face plus port extension is acceptable for routine coax extensions; calibration at the DUT plane is preferable when standards can be connected there; matrix-based de-embedding is appropriate for PCBs, probes, transitions, or long dispersive structures. Keep raw measurements, calibration data, fixture files, and correction settings together, because an apparently small S2P error can become a large uncertainty after subtraction, especially for well-matched or high-loss DUTs. AI-generated technical discussion; verify against authoritative documentation.
A practical diagnostic is to compare three checks after calibration: measure a known through, a precision load or airline at the intended DUT plane, and the same fixture with a short/open/load if available. If the through shows increasing insertion loss or ripple, if the load reflection does not stay near the center of the Smith chart across frequency, or if the required electrical delay changes with frequency, simple port extension is already outside its assumptions. For two-port fixtures, a 2X through-based de-embedding flow is usually more defensible than manually entering two independent delays, but it still depends on symmetry and a well-defined through standard. For asymmetric fixtures, characterize each half separately where the calibration topology permits, or use a validated fixture S-parameter model with proper passivity, reciprocity, and causality checks.
The reference direction is: first define the exact calibration plane and target DUT plane, then choose the correction method based on fixture complexity rather than convenience. SOLT/ECal at the cable face plus port extension is acceptable for routine coax extensions; calibration at the DUT plane is preferable when standards can be connected there; matrix-based de-embedding is appropriate for PCBs, probes, transitions, or long dispersive structures. Keep raw measurements, calibration data, fixture files, and correction settings together, because an apparently small S2P error can become a large uncertainty after subtraction, especially for well-matched or high-loss DUTs. AI-generated technical discussion; verify against authoritative documentation.
