Validation

Hardened audit system around the calculator suite. Every case has a formula, an expected value with provenance, an actual value computed live, a trust label, a tolerance, and a receipt hash.

Trust label legend

Every validation case carries exactly one trust label declaring what the case proves. These labels are emitted into the certificate and the JSON report so downstream reviewers know what kind of evidence a PASS represents.

Exact analytic

Formula identity / closed-form expected value.

Textbook benchmark

Reproduces a known textbook value within the published precision.

Public experimental support

Validates a coefficient or effect against a published measurement, not an exact scenario.

Public scenario anchor

Public mission facts as scenario context — NOT exact validation.

Adversarial fail-closed

Invalid input triggers a warning or rejection.

Cross-tool consistency

Two independent code paths agree on the same physics term.

Implementation regression

Confirms the code path has not drifted from its earlier value.

Summary

Cross-tool consistency checks

These cases compute the same physics term through two independent code paths in shared/shared.js / shared/validation.js and assert that the results agree to floating-point precision. They detect drift between calculators rather than validating any external reference.

ID	Case	Expected	Actual	Status

Authoritative constants

Every constant used by the calculators with its public source. Edit a constant in shared/shared.js only with a matching source change here.

Symbol	Value	Units	Source	Note

Validation cases

Analytic textbook cases — exact closed-form expected values

ID	Case	Formula	Expected	Actual	Rel. err	Tol	Status

JPL Horizons public geometry — publicly tabulated ephemeris snapshots

Tolerances are relaxed to absorb published rounding, NOT to admit drift. Anchor values are recorded explicitly in shared/validation.js.

ID	Case	Formula	Expected	Actual	Rel. err	Tol	Status

NASA/JPL DSOC public scenario anchors — scenario anchors only

Public scenario anchors, not internal mission validation. These cases verify that the publicly disclosed Psyche/DSOC parameters (transmit power, aperture, wavelength, Hale-receiver aperture, transverse velocity) produce a sensible order-of-magnitude link budget in our uniform-spot model. DSOC internal parameters — modulation, coding, detector quantum efficiency, AO loops, pointing-control loops, scheme — are not public and are not modeled. These are DSOC public scenario anchors only.

ID	Case	Formula	Expected	Actual	Rel. err	Tol	Status

Public-data experimental cross-checks — canonical experiments and published measurements

Compares the calculators' outputs against published values from canonical GR experiments (Eddington, GPS / Ashby, Pound–Rebka, Vessot–Levine GP-A, Cassini superior-conjunction, NASA DSOC public records). Tolerances reflect published precision plus the model's approximation order.

ID	Case	Formula	Expected	Actual	Rel. err	Tol	Status

Adversarial / fail-closed tests — calculators must error or warn on bad inputs

ID	Case	Failure condition	Expected	Actual	Rel. err	Tol	Status

Cross-tool consistency — same physics through two code paths must match

ID	Case	Formula	Expected	Actual	Rel. err	Tol	Status

Source registry

Every formula and constant points to a public source. Source IDs in the validation cases above resolve to entries here.

ID	Name	Role

Deployment-blocking test rules

Every case marked required: true in shared/validation.js must PASS.
If any required case fails, the deployment gate above goes to BLOCK.
Adversarial cases test that the calculator fails closed on bad inputs; their PASS condition is that a NaN, Infinity, or warning is produced.
The receipt hash is recomputed on every page load. Two reviewers running this page should obtain the same receipt hash if the calculator module is unchanged.

Certificate format

Every calculator on the lab also exports a JSON certificate. Schema:

{
  "tool":          "<tool-name>",
  "version":       "0.1.0",
  "timestamp_utc": "ISO-8601",
  "inputs":        { ... },
  "constants":     { ... },
  "formulas":      [ "Δt = (1+γ) GM/c³ ln((r1+r2+R)/(r1+r2-R))", ... ],
  "assumptions":   [ ... ],
  "outputs":       { ... },
  "tolerance_tests": [ { "term": "...", "rank": "KEEP|WATCH|IGNORE", ... } ],
  "warnings":      [ ... ],
  "source_notes":  [ ... ],
  "generator":     "Relativistic Optics Lab (...)",
  "hash":          { "sha256": "<64-hex>", "short": "<12-hex>" }
}

Download report

The full validation report can be downloaded as JSON. This is the result that should be archived alongside any engineering note that cites a calculator output.

Download last committed snapshot Download validation_badge.json

Static badge (shields.io-style):