# RF S-parameter and de-embedding evidence replay model card

- Version: 2026-08-24-v100
- Experiment: SA-KNOW-RF-SPARAM-001
- Execution: precomputed browser CSV replay only
- Rows: 6048

## Purpose

Teach how an intrinsic small-signal two-port, passive fixtures, reference impedance, correction assumptions, residual errors, stability/passivity screens, cutoff-frequency proxies, and noise parameters fit into one evidence chain.

## Model

The generator uses three original assumed common-source small-signal Y-parameter states. It converts Y to S at real 50 ohm and 75 ohm references, converts S to ABCD, and cascades original passive positive-delay fixture boxes. Four correction classes are replayed: raw, lossless port extension, an averaged symmetric split-fixture proxy, and exact declared-fixture inversion.

The split-fixture proxy is not an IEEE 370 algorithm. No TRL, SOLT, LRM, VNA calibration, instrument operation, Touchstone ingestion, time gating, DC extrapolation, or uncertainty framework is executed.

## Screens

The packet computes fixture sigma-max, Rollet K and Delta, first-order fT and fmax proxies, short-circuit current-gain magnitude, and the standard four-noise-parameter equation from assumed inputs. These are equation-literacy outputs, not extracted device specifications, stability signoff, noise characterization, or matching guidance.

## Replacement evidence

Replace every assumed device, fixture, residual-error, and noise field with owned bias- and temperature-tagged complex data; exact port and impedance definitions; calibration and de-embedding provenance; propagated uncertainty; passivity, reciprocity, and causality review across an adequate band; model provenance; and accountable RF signoff authority.

## Rights

NIST, IEEE, MIT OCW, UC Berkeley, scikit-rf, and BSIM sources are linked as official or rights-holder anchors. No paper, standard text, course file, code example, plot, table, Touchstone file, or external device value is mirrored.