Our reference request needs genuine reject faceplates alongside the accepted bright-band examples. The captions still make the reflections sound more certain than they are. How do I obtain and preserve the physical defect evidence without letting the next lighting comparison become a contest for the cleanest-looking image?
Have quality select retained rejects against the actual appearance requirement and identify their physical regions. Don't make a bright caption or an artificial-looking line stand in for that decision.
Quality is checking retained production rejects. The accepted examples are already identified, and the uncertain captions remain in the history rather than the current reference answers. I have not assumed a rejected whole part means every pictured line on it is a defect.
Keep the normal surface variation in that selection too. A single dramatic scratch may be easy to photograph without representing the required limit, while the strongest band on an accepted faceplate is still useful evidence about false rejection.
And keep each region linked through a full view before cropping. That will help the next person compare the same physical feature instead of whichever bright patch catches their eye
Quality found two traceable reject faceplates with different required defect regions, one faint score near the edge and one broader scratch on the face. Full views link their decisions. Both accepted references remain in the comparison under the same recorded conditions.
Has the comparison kept both defects visible across that range, Emma? I would state any hidden region directly rather than let an overall example pass conceal it behind a strong result elsewhere on the faceplate.
The face scratch remains visible, but the faint edge score is lost at one intended presentation. Vision is assessing a separate edge view while retaining the face coverage. We have repaired the reject references, not completed the inspection capability assessment.