ECSS Commercial EEE Components -- Class 3 Lot Homogeneity (space-systems/ecss/q6013-class-3-lot-homogeneity)
Use when the task is the clause 6.5.5 uniformity precondition of ECSS-Q-ST-60-13C at the lowest assurance class: a commercial delivery is about to be sampled, the manufacturer publishes no wafer or assembly lot behind it, and the only evidence of one production run is what the procurement paperwork and the parts themselves carry.
Domain quick reference
- Uniformity here is assembled out of supply evidence, not out of a traceability tuple. Part number, package code, manufacturer, marking style, origin and the distributor reference are what a commercial delivery actually offers, and the question is how many of them hold across the whole box.
- Two of those attributes are not scored at all. A delivery carrying two part numbers or two package codes is two different articles, and no weighting of the remainder makes one sample speak for both. That is a stop, taken before any score is computed.
- The rest are weighted and summed into a score held as an exact integer ratio of matched weight to total weight. Weights are a declared input, so a programme that values origin above marking style says so in the policy rather than in the reading of the result.
- A supplier batch declaration is the relief this class admits. Where the supplier names one production batch and that declaration covers every unit in the delivery, the score threshold drops to the declared relaxed figure. A declaration missing even one unit gives no relief at all; it is a statement about a different population.
- The date-code window is an independent axis. Parts that agree on every supply attribute can still have been made months apart, so the window in whole weeks is compared with the limit whatever the score says.
- The reduced sample this class permits is sized by exact integer square root of the delivery, bounded by a floor, a cap and the delivery itself, and it has to reach both ends of the date-code window. A sample drawn entirely from the middle of the window has tested the middle of the window.
Workflow
- Validate every unit: a non-empty identifier, the two decisive supply attributes, the weighted attributes, and a four-digit date code. A duplicate identifier is an input error.
- Refuse a population whose two-digit years straddle a century rollover, because the window cannot be ordered from the codes alone.
- Compare the decisive attributes across the population and stop on any that carries more than one value.
- Score the weighted attributes: sum the weights of those holding one value across every unit, against the sum of all weights.
- Validate any supplier batch declaration and check it names every unit before letting it lower the threshold.
- Take the date-code window in whole weeks between the oldest and newest unit and compare it with the admitted limit.
- Size the reduced sample by integer square root of the delivery, bounded by the floor, the cap and the delivery size.
- Map the drawn units onto the window: count them, and confirm one sits at the oldest code and one at the newest.
- Return one verdict in precedence order: a decisive attribute conflict, a window wider than admitted, a score below the applicable threshold, a sample that is undersized or does not span the window, otherwise a uniform batch.
Pitfalls
- Averaging a package-code difference into the score. Two package codes are two articles; a score of ninety per cent over two articles is a number about nothing.
- Taking a supplier batch declaration at face value. The relief comes from the declaration covering the delivery, so a declaration that names ninety of a hundred units is evidence about ninety units and the delivery is still judged on its own score.
- Reading a high score as a narrow date-code window. The axes are independent: one distributor, one marking and one origin say nothing about how far apart the parts were made.
- Sizing the reduced sample with a floating-point square root. A delivery of exactly a hundred sits on the boundary, and that boundary is the one number two sites compare.
- Drawing the sample from whatever was on top of the box. Units near one date code cluster together in a reel, so a sample that never reaches the far end of the window has left the oldest material untested.
- Reporting uniformity as a single yes. The useful output is which attributes held, which did not, and how wide the window was, because that is what the next delivery gets compared against.
Behavior contract (gate 3)
The unit validation, century-rollover refusal, decisive-attribute stop, weighted uniformity scoring, declaration relief, date-code window, integer-root sample sizing, window endpoint coverage and verdict precedence are exercised by the gate 3 contract test: scripts/test_q6013_class_3_lot_homogeneity.py against scripts/q6013_class_3_lot_homogeneity_logic.py (stdlib unittest, offline). Run: python3 scripts/test_q6013_class_3_lot_homogeneity.py
Compliance
- ECSS standards are freely downloadable (ESA); cite the source and paraphrase per standards-map.yaml.
- compliance: STANDARDS-REF, gated: false.
