What it shows
The lab’s own entry states the result this way:
The published record says, word for word (an excerpt)
On 10 of 10 de-embedded coupons, scikit-rf's unconstrained VectorFitting emits σ_max>1 (formally non-passive under IEEE-P370) while the passive-by-construction Foster/Cayley synthesis emits σ_max≤1 on 10 of 10
In plain words: in the recorded run, the public library’s default fit failed the standard’s passivity test on 10 of 10 coupons, and our synthesis failed on 0. Our construction is passive for any parameter values, and a separate check of 20000 random parameter sets found 0 violations.
Each run draws its coupons from the commit it runs at, so the per-coupon numbers change from run to run. This page relies only on the class result in the recorded run of 10 coupons: every coupon failed for the default fit and passed for ours.
Why it matters
A non-passive circuit model can make a downstream simulation unstable or wrong. The standard exists to catch this. On the 10 coupons tested, the library’s default fit, used without an extra enforcement step, did not pass it.
What is ours, and what is not
The fitting method and the passivity test are public (see the prior art below). What is ours is the passive-by-construction synthesis and this side-by-side test against a public tool’s default output.
Who should care
- Signal-integrity engineers who export fitted models into circuit simulators.
- Reviewers. This grades a public tool’s default output under a public standard’s test, not against our own solver.
The limits, in the record’s words
The published record says, word for word (an excerpt)
sim-vs-sim: about the S-synthesis object (Re Y ⪰ 0 ⇒ Cayley contraction), not measured silicon.
The published record says, word for word (an excerpt)
Grivet-Talocia post-hoc enforcement (the repo's own Gate-82 path) does not reliably converge on multi-port MTL, disclosed.
The published record says, word for word (an excerpt)
the script derives its eval seed from the commit sha, so every re-run resamples the 10 coupons
The published record says, word for word (an excerpt)
the 'system' is a library's default fit, not a vendor product, and no counterparty has been shown it.
In plain words: this is about the fitted model, in simulation, not measured silicon. The violations are marginal. The test is of a library’s default setting, not of a vendor product, and enforcement after the fact, which our own export also uses, does not always converge on larger models.
Outside comparison
Compared with: the VectorFitting class of scikit-rf, an open-source Python library for radio-frequency engineering. Retrieved 2026-10-05. scikit-rf, its public page · the retrieval record
Open source for this step
Tools and datasets we publish for the package step of building a multi-chip package. They are the checkers around this work, not a copy of the result itself.
- physics-lint: One command that checks a folder of physics models against a fixed set of named physical rules, with findings straight into CI.
- maxwell-lint: Flags a coupling extractor whose answers no passive set of conductors could produce.
- sparam-lint: Is your signal-response model physically possible? Five physical laws checked from the command line.
- interval-core: The interval arithmetic core behind our proofs over whole families of layouts.
- touchstone-tools: Read, write and convert Touchstone files, the standard text files that record how signals pass through a package's connections, and refuse to write one that cannot be read back.
- physics-lint-mcp: The physics checks, callable by an AI agent.
- physics-lint-action: A GitHub Action that fails the build when a model breaks one of a fixed set of named physical rules.
- Signal-response validity corpus: A labelled corpus of physically invalid signal-response networks, and a scorer that grades any checker against it.
- screening-ceiling: The screening-ceiling family as an open dataset.