When the math runs out

An 8-bit image has 256 levels per channel, and a 10-bit image has 1,024. Stretch a narrow 8-bit range across a wide one and those levels pull apart into visible steps, which is banding. It shows worst in the smooth gradients such as skies, fog, and skin. No amount of precision further down the chain puts the missing steps back, because the source only ever held 256 of them.

```sketch { "kind": "range", "scale": { "min": 0, "max": 1024, "unit": "levels" }, "bars": [ { "label": "8-bit · 256", "from": 0, "to": 256, "accent": true }, { "label": "10-bit · 1,024", "from": 0, "to": 1024 } ], "caption": "Same range, a quarter the levels. The 8-bit bar reaches only 256 and stretches into visible gaps, the banding, while 10-bit's 1,024 levels fill it smooth." } ```

The vectorscope graticule marks hue directions and calibrated color-bar targets. Those boxes are not saturation limits for natural images. Oversaturation is a picture failure: color variation flattens, a channel or output boundary clips, skin becomes implausible, or noise is amplified. Several operations can change perceived or measured chroma, so isolate the contribution of each move instead of diagnosing from distance alone.

Take the same diagnosis to the silk below. The crimson has lost internal variation after both saturation and contrast were pushed. Neither the picture nor the scope names the responsible control on its own. Take one move back, watch whether color detail returns, then test the other. The changing evidence tells you which operation caused which part of the damage.

Answer the two checks and grade the silk, then set the plan for a plate that came in with all of it.