Records and formats
Converra cases declare requirements, geometry, operating conditions, material data, limits, sampling, and cost assumptions. Run records bind the inputs and model identities to candidate outcomes and their cost ledgers.
Formats and supported workflows change over time. Use the current handbook for the active schemas, import steps, and result interpretation.
What verification checks
Different checks support different claims. A verified ledger does not establish that the physical assumptions fit a particular magnet.
- Input identity.Case and dataset hashes bind a result to the exact files used.
- Numerical comparisons.Some field calculations are compared with an independent implementation under a fixed probe deck and declared tolerances.
- Acceptance recomputation.A separate checker recomputes selected result fields and declared sampling checks. The handbook identifies when a study’s acceptance outcome is inconclusive or failed.
- Reproduction.Published cases, records, and checkers are available for the studies linked below. They do not cover every geometry, material, or operating point.
Selected benchmark results
The current benchmark guide describes each result with its inputs, status, and scope. These examples illustrate the range of evidence, including unresolved outcomes.
The closest coarse-screened candidate uses 184 m against an approximate 190 m reference. Its refined check remained INCONCLUSIVE; the study did not produce a passing recommendation.
A fixed geometry and probe deck compare Converra with a separate Biot–Savart implementation. This checks the field kernel on those inputs, not design qualification.
The reported cost reduction uses synthetic prices and is conditional on its baseline and screening model.
The record identifies material-coverage and current-capacity blockers for that case.
Selected artifacts
These files support reproduction of the included historical cases. They are not a complete archive of current Converra formats or benchmarks.
A fixed coupled case and reference output for the declared self-field model. This frozen fixture supports parity checks for that case only.
The OC-010 Feather-M2 case, run record, model and checker versions, and dataset hash bound together.
A standard-library checker for artifact hashes, the embedded case, ledger arithmetic, and optimum selection in the included record.
The OC-010 input case associated with the included manifest.
The OC-014 reference output is available as a separate file: oc-014.reference.json.
Inputs and prices
- Material data.Studies distinguish measured conductor characterization from any declared model extension. A result outside measured coverage should be read with that extension and its assumptions in view.
- Cost inputs.Some published benchmarks use synthetic prices for arithmetic demonstrations; those figures are not supplier quotes or commercial savings estimates.
- Mechanical checks.First-order screening bounds are not a structural analysis or manufacturing assessment.
Boundaries
These are synthetic or benchmark studies; no customer design result is represented here. A passing screen applies only to its declared checks. Engineering review, material qualification, and manufacturing evidence remain outside the result.