Tech note: why shallow scratches on blasted surfaces are hard — a sharper camera can mean more false calls

Heat spreaders are key parts of thermal modules, with a sand-blasted surface. The customer needed to find scratches, abrasion, dents and blasting defects; the hardest were bright scratches and shallow scratches. We ran five rounds of tests, and this note collects what we learned.

Light angle decides what you see

• Exposed copper: highlighted by both high- and low-angle ring light
• Scratches: in early tests, only low-angle light showed them
• Blasting defects, bright scratches, bright abrasion: shown by high-angle ring light, which also makes dust and dirt stand out
• Dents and scratches: with a dome light, most were detected, including dents at the four corners

The side effect of higher resolution

Switching from a 5 MP camera (about 50 μm) to a 12 MP monochrome camera (34.5 μm) made defects clearer — but also the blasting texture, so false calls became easier. Resolution should be chosen together with surface texture, not simply maximised.

The hidden advantage of colour cameras

A colour camera offers R, G and B channels, and the same defect has different contrast in each. In these samples the G channel showed defects best.

Data and limits

• Defects detected in the on-site demo were about 7–23 pixels, or 0.35–1.15 mm
• Minimum detection: about 0.15 mm with clear colour difference, about 0.25 mm without
• Shallow scratches still cause false calls; with a large field of view, ambient light interferes and a hood is needed
• A texture-comparison method for uniformity took about 130 ms per part and found uneven areas, though more samples are needed to confirm stability

Takeaway

No single light solves blasted surfaces. Validate samples with several lights and cameras first, find the best condition for each defect type, then decide between one multi-light station and separate stations — before false calls hurt mass production.