Colour measurement from scratch — no prior knowledge needed. 12 short slides, swipe or tap through. Forgot a term? Check the glossary.
A panel comes off the line. Somebody holds it up next to the master, and two people look at exactly the same thing:
Neither of them is lying. They genuinely see it differently. And there's no way to settle it — you can't hand someone your eyes.
It's not that anyone is careless. Human colour judgement has four built-in problems:
An instrument doesn't have opinions, doesn't get tired and doesn't care what it saw five minutes ago. Shine a known light at the panel, measure what bounces back, and turn it into numbers. Now the argument has an answer both sides can check.
But which numbers? Not plain RGB — the same numeric step there can look huge in one part of the colour range and invisible in another. CIELAB (1976) was built specifically so that equal numeric distance ≈ equal perceived difference:
Both pairs are the same numeric distance apart. That's the point of a perceptually uniform space.
Three numbers place any colour in a 3D space. Each axis runs between two opposites:
Two more — C* and h — aren't new information. They're the same point described the way a colorist would say it out loud:
Drag the sliders — this is the exact plane the calculator draws under every result:
L* left = darker, right = lighter · a* left = greener, right = redder · b* left = bluer, right = yellower
Remember problem three — the light changes everything? That doesn't go away just because you bought an instrument. Colour is what happens when a light source, an object and an eye meet. Change any one and the answer changes.
Illustrative shift only, not a spectral simulation.
Every spectrophotometer does the same three things: shine a light, catch what comes back, do the maths. What separates them is geometry — where the light comes from and where the sensor sits.
45/0 — light in at 45°, sensor reads straight on
This is where geometry stops being theory. Pick up the wrong instrument and your numbers are meaningless — no matter how carefully you measured.
Named as common examples, not recommendations — in practice your customer's specification usually dictates the geometry, and sometimes the exact model.
The instrument is honest, but it can only report what you point it at. Four habits separate a number you can defend from one you can't:
You have master and sample as numbers. Now: how far apart are they? That single answer is ΔE — and there are four accepted ways to calculate it, each a better approximation of human perception than the last.
Same colour, same numeric tolerance — four different "acceptable" zones
ΔE tells you how much. It says nothing about which way — and you can't tint a batch with "how much". For that you need the components:
Sometimes every number passes and the part still gets rejected. Usually the culprit is gloss — measured in gloss units (GU), typically at 60°.
You started with two people disagreeing over a panel and no way to settle it. Now you have a chain that anyone can follow and check:
Next: test yourself on the basics, try the colour vision self-check from Part 1, or go straight to the calculator.