Concept

A colour reference is not a swatch. It is the one definition that every stage of a job answers to — and the reason that matters is arithmetic: in a chain of conversions the errors add up, and against a shared reference they do not.

Two ways to move a colour through a workflow

A colour has to reach a proof, a press, a laminate, a screen. There are two ways to get it there, and they behave completely differently.

Two ways of getting a colour to four devices In a chain each device converts from the previous device's output, so every conversion adds its own error and the errors accumulate. With a central reference every device converts from the same definition, so each error is its own and none of them compound. A chain — each device copies the last design +e proof +e press +e the label ERROR ACCUMULATES A reference — each device reads the same definition THE DEFINITION design proof press the label ERROR STAYS WHERE IT IS MADE

The same four stages, wired two ways. On the left each stage converts from what the previous one produced. On the right each converts from the same definition.

In a chain, every conversion inherits everything before it. The proof is a conversion of the design; the press is matched to the proof; the laminated result is judged against the press sheet. Each step is defensible on its own and each adds a little, and by the end the accumulated difference is larger than anything anyone would have accepted in a single step.

Against a shared reference, each stage converts once. The proofing system aims at the definition. The press aims at the definition. The screen aims at the definition. Each has its own error, and the errors do not compound, because nothing downstream is copying anything upstream.

Errors are not repeated through conversions, because each new device refers to the record rather than to its predecessor.

Which means each device may be honest about its own limits

A press cannot reach every colour a screen can show, and a screen cannot show every colour an ink can print. Given the definition, each device can make its own best approach to it — a gamut mapping suited to what that device actually is — instead of inheriting a compromise somebody else made earlier for different reasons.

That is only possible if the definition is richer than the compromise. Three Lab numbers already contain a decision about the light and the observer; a spectrum does not.

Why spectral, and why not tied to a substrate

ChromaSpot records what each colour does to light, wavelength by wavelength. From that, the colorimetry can be computed for any lighting condition and either standard observer — so the same definition serves a print shop working at D50 and an interior specification working at D65 and 10°, without either of them holding a converted copy of the other’s numbers.

It also means the influence of ultraviolet, and of the brighteners in paper that respond to it, is something the definition can express rather than something that quietly happens to it.

And the colour is not tied to the material it was first made on. That is a separate argument and it has its own page: Substrate Independent.

Where the colours came from

The samples are real: printed, with pigments and colorants the industry uses, and measured. Not computed points on a grid — which matters, because a colour that no ink can make is a colour that will be specified by somebody and then argued about for a month.

Different printing technologies reach them with different pigments, so metamerism arises. That is not a flaw to be apologised for; it is a fact of the industry, and it is predictable from spectral data and from nothing else.

The whole idea in one line. Keep one definition, rich enough that every stage can convert from it independently — and then nobody is copying anybody else’s mistakes.

Read next