ECSS Solar Cell Assemblies — Failed Cell Assemblies (space-systems/ecss/e2008-failed-cell-assemblies)
Use when the task is the clause 6.5.2 treatment of ECSS-E-ST-20-08C: a solar cell assembly has been inspected against the listed failure modes, and whether that unit counts as failed has to be settled on the evidence rather than on an impression of how bad it looked.
Domain quick reference
- The rule is disjunctive, not a score. One listed mode found on the unit is the whole answer. The modes are not weighted against each other, a mode is not offset by the seven that came out clean, and there is no threshold count of modes below which a unit survives.
- That makes the verdict short-circuit. A unit with a present mode is failed whatever the remaining modes did, so an examination that stopped early because the first mode was found is still a complete answer for this clause.
- The same short circuit does not run the other way. A unit with no present mode and a mode nobody looked at has not been shown to be sound; it is unevaluated. Reading an absent observation as a clean result is how an incomplete inspection passes as a pass.
- The catalogue mixes two bases. Some modes are seen and take a yes or no; others exist only as a quantity crossing a limit and take a measurement. Handing one basis the other's evidence is an input defect, because a truthy number would condemn every unit that was measured at all.
- Measured modes face the limit in a direction. A degradation and a delamination are modes by rising above a limit; an insulation resistance is a mode by falling below one, and a single comparison cannot serve both.
- Limits are inclusive. A measurement landing exactly on the limit is not the mode, so the comparison absorbs representation error instead of the limit being nudged to make the arithmetic tidy.
- The lot tally is a separate arithmetic. The failed share is taken over the units actually settled, so units left unevaluated neither dilute the share nor quietly count as good.
Workflow
- Resolve the mode catalogue and merge any project limit overrides onto the declared defaults, refusing an unrecognized limit, a non-positive one, or a fraction above unity.
- Normalize every observation onto a listed mode, refusing an unlisted mode and a mode observed twice under two spellings.
- Reduce each mode in the catalogue to present, absent or never examined, validating the evidence against the mode's basis in both directions.
- Grade each measured mode against its limit in the direction that mode crosses it, treating a value exactly on the limit as absent.
- Settle the unit: failed the moment any mode is present, unevaluated when nothing is present but something was never examined, sound only when every listed mode was examined and came out absent.
- Report every present mode and every unexamined mode, so the disposition names what drove it rather than delivering a bare verdict.
- Roll the units into the lot, taking the failed share over the settled units and carrying the unevaluated units through as their own list.
Pitfalls
- Scoring the modes. Counting three minor modes as worse than one major one reintroduces a judgement the clause does not have; the count is reported, but the verdict turns on the first present mode alone.
- Reading an unexamined mode as a clean one. Nothing was submitted, so nothing was found, so the unit looks sound -- and the one mode that mattered is the one nobody looked for.
- Waiting for a complete examination before condemning. A found mode is enough, and holding the disposition open for the remaining modes keeps a known-bad unit in the flow.
- Passing a boolean to a measured mode. It reads as a small number, sits under every rising limit, and turns a mode that was never measured into a pass.
- Using one comparison for both directions. An insulation resistance is a mode by being too low; graded as though it were a degradation, a dead unit reads clean.
- Widening a limit so a measurement exactly on it counts. The edge is already inclusive; the tolerance belongs inside the comparison.
- Taking the lot failed share over every unit presented. Units that were never settled belong outside the denominator, not silently in the good column.
Behavior contract (gate 3)
The catalogue split by basis, the limit resolution and its refusals, the three per-mode statuses, the basis-mismatch refusals in both directions, the inclusive limit edges on a rising and a falling mode, the one-mode-condemns rule with its short circuit past unexamined modes, the unevaluated verdict for a clean but incomplete unit, and the lot tally with its settled-unit denominator are exercised by the gate 3 contract test: scripts/test_e2008_failed_cell_assemblies.py against scripts/e2008_failed_cell_assemblies_logic.py (stdlib unittest, offline). Run: python3 scripts/test_e2008_failed_cell_assemblies.py
Compliance
- ECSS standards are freely downloadable (ESA); cite the source and paraphrase per standards-map.yaml.
- compliance: STANDARDS-REF, gated: false.
