Mixed-mode common mistakes¶
| Mistake | What you see | Fix |
|---|---|---|
| Computing before port roles are confirmed | Plausible SDD21, wrong phase / swapped P-N | Confirm mapping; see port mapping |
| Arithmetic on dB traces | Magnitude in the right ballpark, physically meaningless mix | Convert to complex linear, combine, dB at the end |
| Dropping the 0.5 prefactor | Everything 6 dB too hot | Every mixed-mode formula has 0.5 |
Assuming [out][in] vs [in][out] |
PN is the opposite path |
Ask; do not infer from one filename |
| One S2P named “SDD21” | A single-ended trace | Need four port-pairs or an S4P |
| Treating 2-port SDD11 as exact | Return-loss plots that will not match a 4-port VNA | Call it approximate; measure S4P |
| Mixing frequency grids | Interpolated junk at the band edges | Align x before combining |
| 50 Ω vs 100 Ω confusion | Impedance “error” that is only a definition | 50 Ω SE is the Touchstone default; mixed-mode pair is usually 100 Ω differential — see Differential reference impedance |
| Legacy sign-flipped scripts | dB overlay looks fine, phase and SDC/SCD do not | Use the identities in conversion formulas |
If a result is surprising, check mapping and units before checking the DUT.