Differential reference impedance

Touchstone files from a VNA are almost always 50 Ω single-ended per port (R 50 in the header). A differential pair of two 50 Ω lines is 100 Ω differential and 25 Ω common-mode. That is a change of basis, not a measurement error.

What to keep 50 Ω

Leave the four S2Ps (or the S4P) at 50 Ω while you convert to mixed-mode with the 0.5 formulas or scikit-rf. The conversion already produces mixed-mode waves referenced to the odd/even-mode impedances that go with those 50 Ω ports.

Renormalizing the single-ended file to 100 Ω before mixed-mode conversion is a different experiment. Do not do it by accident.

When Z0 actually matters

• Comparing mixed-mode SDD11 to a 100 Ω TDR or a 100 Ω simulation port.
• A DUT that is not 100 Ω differential (some RF pairs, some probes). Then renormalize after mixed-mode conversion, or measure in the DUT’s reference.
• Headers that are not 50 Ω (R 75, etc.) — believe the header.

CloudSprite

Upload the file the instrument wrote. Record the intended differential Z0 as a dataset parameter if it is not 100 Ω (zdiff=90). Analysis scripts should read that parameter rather than assuming.