Flat, washed-out scans, and a roll whose frames disagree.

A density needs a denominator

A colour negative is read in density: D = −log10(frame / light). Without a measurement of the bare light, and of the bare film base, the numbers have no absolute scale at all — so a conversion has to infer both from the picture. Inferred frame by frame, they land frame by frame: a conversion that infers the base from each frame on its own tends to need a nudge on every frame, and can put two frames of one roll in different places. How trichromatic (narrowband RGB) scanning works has the rest of the physics.

The usual causes

Exposed for the film, not for the empty gate

Meter on the negative and the empty gate goes past the top of the sensor, where nothing can be measured against it. Set the exposure once on the empty gate, film out, and shoot every frame on it — the frames will look dark, and that is right. Exposure, step by step.

No shot of the bare light, or one taken at other settings

Every density is divided by the light, so the empty gate has to be shot with the roll, at the roll's settings. Frames taken at other settings than their empty gate carry the difference into every density; a calibration split across two shutter speeds carries about 90 counts of error into every density on the roll while looking perfectly ordinary.

The film base guessed instead of measured

The orange mask sits under every frame. Measured once, off the clear rebate around the frames, it is a property of the roll — the same stock, the same development — so every difference between frames is measurement error rather than signal. Guessed per frame, it becomes a different guess on every frame.

The holder read as film

A sliver of black holder inside what is read as base makes the base read too dense. The whole roll then grades wrong, and every frame still looks plausible on its own — which is why it goes unnoticed.

White light, and a camera's own filters

Under white light the sensor's red, green and blue filters overlap, so each channel carries some of the other two dyes, and the orange mask has to be separated afterwards. It can be done: over a full 36-frame roll on one measured station it costs about five counts a channel against three narrow bands. Narrowband RGB or white.

A scanning program that levels each file

From a film scanner, an export with auto-levels, a tone curve or colour conversion is no longer a measurement of the film: it has been levelled frame by frame before any converter sees it. What a converter wants is the scanner's own numbers — linear, 16 bits a channel, nothing stretched.

What Trichroma Lab does about it

It does not adjust per frame what it can measure once per roll. Under three narrow bands with a flat, the film base is measured off your own rebate, per roll, in density, and the light is measured from the empty gate; the prescan shows how far the frames disagreed, and calls them agreeing when every channel is within 10 counts. Scene contrast is measured during the roll pass — a reading, not a switch.

Six negatives beside a Frontier SP‑3000's own scans show where that lands.