RGB, RGBW and Why White Is the Hardest Colour to Mix

mix

Stage lighting starts from a simple fact: an LED emits one narrow band of wavelengths. A red die is not “red paint” — it is an emitter that sends out light between roughly 615 and 635 nanometres and nothing else. You cannot filter it into green, and you cannot dim it into a different hue. The only way to make a colour is to add emitters together.

That is why every colour-mixing fixture you own is really a small addition problem. Three numbers go in — how hard to drive red, green and blue — and one colour comes out. The console shows you a wheel or an XY pad, but underneath it is always running the same three sliders.

Here is the first surprise. To make a neutral 6500K white, the three channels come out almost equal in radiated power — about 35% red, 34% green, 31% blue. Even split, which sounds reassuringly fair.

But the emitters are nowhere near equally good at turning power into light. Green sits right at the peak of human sensitivity and returns about 556 lumens for every watt of radiated power. Blue sits out at the far edge of vision and returns about 53. The ratio is more than ten to one.

Put those two facts together and the picture flips. Blue eats 31% of the power and delivers under 6% of the light. Green eats 34% and delivers two thirds of it. The fixture is spending most of its energy on the channel your eye can barely see.

grey

Now the actual question. If three channels can hit the white point exactly, what is left to get wrong? colour has three dimensions, and mixing controls only two of them.

The chromaticity diagram gives you x and y — the hue and the tint. A white point is a single coordinate on that map, and RGB can reach it. But a coordinate is not a spectrum. An infinite family of different spectra all land on the same coordinate, and your eye does not see the coordinate. It sees the spectrum, filtered through the reflectivity of whatever you are lighting.

So the useful test is not “does the white match?” but “is there any light coming out at the wavelengths this object reflects?” Drop three narrow peaks onto a skin tone and the answer depends entirely on where the peaks fall.

This is the whole answer in one picture. The RGB white is not dim and its colour is correct, but there is a hole exactly where yellow should be. Measured per nanometre, the 570-590 nm band carries 0.005 of the energy in the blue band — half a per cent. Widen the window to 560-610 nm and an RGB white still holds only 1.7% of its total energy, against 26.0% for a white built with a broadband emitter.

Yellow is not a minor band. Human vision is most sensitive right around 555 nm, and skin, wood, brass, warm-painted walls, and nearly every material with a warm tint reflect strongly in exactly the region the RGB white is missing. The light hits the object, the object has nothing to send back, and what reaches the camera is the grey.

Measured against a threshold of 20% of peak, an RGB white at 6500K has four separate empty regions: 380-449 nm, 481-509 nm, 551-609 nm, and 641-780 nm. That four-hundred-nanometre span covers most of what we call a warm surface.

w

The fix is not a better mixing algorithm. It is a different ingredient. Put a dedicated white emitter on the board — a phosphor-converted die, the same kind used in ordinary white lighting — and you are no longer trying to synthesise a continuum out of three spikes. You are simply dimming a broadband source and tinting it.

The physics explains why this works so well. A phosphor white is a blue pump with a broad emission band built on top of it, and that band fills the yellow and orange region natively. Adding one white channel takes the 560-610 nm energy share of a 6500K white from 1.7% to 26%, without touching the white point at all.

One caution about the word “brighter”. Adding a white channel does not raise the ceiling of the fixture — it changes what the fixture is good at. Look at what happens across colour temperatures:

At 2700K a white die saves about 4.5% and the RGB mix is already comfortable. At 6500K it saves nearly 16%, because a daylight white forces a large blue component and blue is where the efficiency goes to die. If your work is tungsten-warm, the extra channel buys you convenience. If your work is daylight, it buys you real output.

layouts

Once you accept that extra emitters are about spectrum rather than brightness, the naming conventions stop being marketing and start being a specification. Each added letter is an answer to a specific rendering problem.

An A adds amber, which does for the orange-red region what the white die does for yellow. Because amber is a broadband-ish emitter sitting at 590 nm, it lets the fixture reach a warm white without dragging the blue channel up to compensate. This is why an RGBA layout can outperform RGBW on skin tone at 3200K even though it has no white die at all.

An L adds lemon, a yellow-green emitter around 575 nm. It is the highest-leverage single channel on the board — lemon returns roughly 615 lumens per radiated watt, the best of any emitter here. A lemon channel lets a fixture build a bright warm white without asking the green channel to fake yellow, and it extends the reachable colour gamut into the saturated yellows that plain RGB simply cannot express.

A UV adds a genuine ultraviolet emitter around 405 nm. This one is categorically different: it contributes almost nothing to visible white — about 1.5 lumens per radiated watt, effectively zero. It exists for blacklight effects and for fluorescing paint and costumes, and it should never be counted as part of a white mix. A real UV channel is not a purple tint on RGB; it is a separate emitter that makes white fabric glow.

split

Extra emitters buy you spectrum. They cost you something physical: the emitters have to sit somewhere, and they cannot all sit in the same place.

This matters whenever colour arrives at a surface from more than one point. If the red die and the blue die are 90 mm apart on a fixture face, then a hard object lit by that fixture receives its red from one angle and its blue from another. The result is a visible coloured edge along the shadow boundary, and it is worst when the object is close to the fixture.

The angle that matters is atan(s / d), where s is the separation between emitters and d is the distance to the lit object. Divide the numbers out and the pattern is clear.

Two conclusions come straight out of that table. A single COB package, where all the colours come out of one small die area, has no separation to begin with and cannot produce this artefact at any distance. And within a multi-emitter face, the problem shrinks with distance — a 24-emitter PAR puts its opposite dies 170 mm apart, which is 9.65° on a subject one metre away and 1.22° at eight metres.

Practically: for close-up work — an interview, a product shot, a performer at the stage lip — keep the multi-emitter fixture back, or use a single-emitter source. For a wash from the truss at six metres and beyond, the split is under two degrees and nobody will ever see it.

sat

There is a trade-off that catches out anyone who has ever pushed a console fader to full and wondered why the stage got darker.

A fully saturated colour means only one emitter channel is on. A pastel means all three are on together. Since the three channels sum, the pastel is always the brighter of the two — and not by a little.

Run the numbers on radiated power alone. Green alone returns about 556 lumens per watt. All three channels at equal power return about 829. So a saturated blue — blue only — delivers around 6% of the light that the same three channels produce when run together. The dimmer colour is not a fault in the fixture. It is the arithmetic of addition.

The practical rule that follows: when a cue needs to read as bright, mix toward white and carry the colour in the tint, not in the saturation. A saturated blue on a dark stage will always lose to a pale blue at the same fader level, and it will lose by a factor, not a notch.

choose

Putting it all together, the choice is mostly about what you light and where the fixture sits — not about which fixture is “best”.

The short version. If you light people, or anything with a warm surface, and you are choosing between an RGB fixture and one with a white or amber emitter, take the extra emitter. The white point will be identical and the faces will not be.

If your work is saturated colour effects — beams, chases, washes on a backdrop — plain RGB already covers the widest saturated gamut for the money, and the spectrum gap does not matter because you are not trying to render a surface faithfully.

And if you need genuine blacklight, that is a separate physical capability that only a real UV emitter provides. No combination of RGB achieves it, regardless of how purple the colour looks in the air.

Fixtures that make the choice concrete

All four of the following are built around a genuine white or broadband-warm emitter rather than an RGB synthesis, which is exactly the distinction this article is about.

Can a good console fix the grey white?

No. The console can only drive the emitters it has. A mixing algorithm moves you around the chromaticity map — it cannot create energy at a wavelength no emitter produces. The white point was never the problem.

Does RGBW mean the fixture is brighter?

Not in general. It is more efficient at producing a given white, by roughly 4.5% at 2700K and about 16% at 6500K in our model, because a broadband emitter is a better use of power than synthesising white from three narrow peaks. A saturated colour does not benefit at all.

What does the extra A actually do?

Amber is a broadband-ish emitter at about 590 nm. It lets a fixture reach a warm white without pushing the blue channel hard, which is why RGBA can beat RGBW on skin tone even though it has no dedicated white die.

Is a 6-in-1 always better than a 4-in-1?

It is more capable, not automatically better. Lemon is the highest-leverage channel on the board at roughly 615 lumens per radiated watt, and it is worth paying for if you need saturated yellows or bright warm whites. But a genuine UV emitter contributes essentially nothing to visible white, so you are paying for a separate capability.

Why does my saturated blue look so dim?

Because blue is the least efficient emitter in the mix. It returns roughly 53 lumens per watt of radiated power against green at about 556 — a ratio above ten to one. A fully saturated colour uses only that one channel, so it loses the brightness the other two would have added.

What causes coloured fringes on edges?

Emitter separation. If different-coloured dies sit apart on the fixture face, a nearby subject receives each colour from a slightly different angle. The angle is atan(s/d), so it grows as the subject gets closer and vanishes as it moves away. A single COB package has no separation and cannot produce the effect.