What a Bi Color LED Costs You for the Dial

This page contains affiliate links. If you buy through one we may earn a commission, at no cost to you.

They will work in the sense that the camera will read them as roughly neutral and you can white balance the rest out. They will not work in the sense the question implies, which is that setting a Kelvin number tells you what the light is. A correlated colour temperature is one coordinate, and a white source needs two. The second one has a name, Duv, almost nothing publishes it, and the panel architecture most likely to have a large one is the bi colour panel with the dial that everybody, including this site, recommends for its flexibility.

What the dial actually is

A bi colour LED panel is not a light whose colour changes. It is two lights in one housing. One set of chips is a warm white, typically around 3200 K, and the other set is a cool white, typically around 5600 K, and the dial is a crossfade between the two. Our own NEEWER two pack bi colour 660 LED kit states the construction in the specification in so many words: 330 warm and 330 cool white LEDs, 40 W, a range of 3200 K to 5600 K, and up to 3300 lux at one metre. Six hundred and sixty chips, half of each.

Read that as an engineer rather than as a buyer and two consequences drop out immediately.

The first is output. At either end of the dial you are using 330 of 660 chips. The other half is dark. The panel is at its brightest somewhere in the middle, where both banks are running, and at its dimmest at the two settings you are most likely to want, which are matched daylight and matched tungsten. The listing publishes one output figure, 3300 lux at one metre, and does not say at which dial position it was taken, so that single number is ambiguous by close to a factor of two. A factor of two in light is exactly one stop, the same size step as going from f/2.8 to f/4, and if you want to read that arithmetic properly it is in why ND8 and ND64 stacked make ND512. Recovering a stop by moving the light closer means coming in to one over the square root of two of the distance, which is 0.71 of it, so a panel that gave you what you needed at one metre has to be at 71 centimetres instead. That is not nothing: it changes the apparent size of the source relative to the subject and therefore the hardness of the shadows, and it puts the panel in shot more often.

The second consequence is colour, and it is the one no number on the box reports.

A straight line across a curved reference

Colour temperature is defined by reference to a black body radiator, an idealised hot object whose colour depends only on its temperature. Plot the colour of that object as it heats up and you get a single curve through colour space, the Planckian locus. A real light source is almost never exactly on that curve, so correlated colour temperature is defined as the temperature of the point on the curve that is closest to the source. Which means a CCT is a distance measured along the curve, and by construction it discards the distance to the curve.

Now add the fact that makes a crossfade a problem. When you add two light sources together, the colour of the result lies on the straight line joining their two colours. That is not a rule of thumb, it is how additive mixing works in chromaticity coordinates. So the dial on a bi colour panel traces a straight chord from its warm end to its cool end. The reference it is calibrated against is a curve. A straight line between two points on a curve does not follow the curve.

The arithmetic is checkable with a calculator and the published black body coordinates. A black body at 3200 K sits at x 0.4243 and y 0.3987. At 5600 K it sits at x 0.3304 and y 0.3392. Average those, because that is what a half and half mix gives you, and the midpoint is x 0.3774, y 0.3690. Now ask where the curve itself is at x 0.3774: between the published 4000 K point at y 0.3768 and the 4500 K point at y 0.3636, which interpolates to about y 0.3747. So the mix sits roughly 0.0058 of y below the curve, and below the curve is the pink and magenta direction.

That number is small and it matters anyway. The ANSI specification that defines nominal white bins for lamps allows a Duv of roughly plus or minus 0.006, so an offset of this order is inside the box that lets a product be sold as white, and it is still an offset you see the moment you put the panel next to a source that is on the curve. It is also, crucially, invisible to every control you have: the dial reads 4400 K, the camera's Kelvin setting goes to 4400 K, and the magenta stays exactly where it was, because the camera's correction runs along the same curve the measurement did. This is why the mid positions on cheap tunable fixtures look faintly pink or, with a different phosphor pairing, faintly green, and why the remedy on set is a gel rather than a number.

The output figures our own listings publish, and why you cannot race them

Our Godox ML60IIBi, a 70 W bi colour COB, publishes its output in a form worth dismantling. The listing reads Lux 6500K, 1M at 23,300 lumens, 0.5M at 86,600 lumens, which contains two separate problems. The units are wrong: a figure quoted at a stated distance is an illuminance and belongs in lux, which is what the label at the front of the same line says, and lumens are a total output that has no distance attached. And the figure is quoted at 6500 K only, which is the extreme cool end of its 2800 K to 6500 K range and therefore the single most favourable point on its own dial. Run the pair of numbers against the inverse square law and they come out interestingly: 23,300 at one metre predicts 93,200 at half a metre, and the listing says 86,600, which is about 7 percent short. That shortfall is in the direction a real reflector produces, because at half a metre a COB with a reflector is no longer small compared with its distance and the inverse square law starts to overestimate. So the two numbers are plausibly a real measurement, badly labelled, at the one CCT that flatters the product. Our record is the Godox ML60IIBi review.

Which brings us to the comparison that does not work, and this is where this post has to argue against its own thesis. A Godox SL60IID, also 70 W, at a fixed 5600 K, publishes 18,600 lux at one metre with the standard reflector. That is less than the bi colour's 23,300, not more. If the clean engineering story were the whole story, the fixed panel should win on output at equal wattage, because it spends every chip on one spectrum. It does not win here, and pretending otherwise would be dishonest.

The reason is that the two figures are not comparable, and that is the real lesson. They were taken at different colour temperatures, through different reflectors, on different optical designs. Reflector geometry alone can move a lux at one metre figure by more than the difference in question, because lux is light per unit area and a narrower beam concentrates the same lumens onto less of it. And colour temperature has its own effect on efficiency: a white LED is a blue die with a phosphor on top, the phosphor loses energy every time it converts a blue photon into a longer wavelength one, and a cooler white converts less of its blue, so lumens per watt generally rise with CCT at a given fidelity. A 6500 K figure and a 5600 K figure are not the same test. So the honest version of the output argument is narrower than the usual one: a fixed panel spends all its chips on one spectrum and therefore does not lose anything to the dial, and whether that makes any particular fixed panel brighter than any particular bi colour one is a question neither listing lets you answer. Our record is the Godox SL60IID review, and the field is in best LED video lights.

One light on the site publishes its colour in the format that would actually settle an argument. The Aputure amaran 100x S, a 100 W bi colour, states SSI figures of 89 or better against tungsten and 87 or better against D56 daylight, which is to say it publishes a spectral similarity score at each end of its range rather than one number for the whole dial. That is the right shape of disclosure and it is rare. It also happens to be a quieter claim than the CRI 95 plus on the same listing, which is what you would expect once you know how the two metrics are built, and we take that apart in why a CRI 95 light still renders red badly. Our record is the Aputure amaran 100x S review. What even that listing does not publish is an output figure at each end of the dial, or a Duv anywhere.

What we got wrong in our own guide

Our LED video lights guide says that "bi-color models let you adjust color temperature from warm to cool, which is essential for matching ambient light or creating mood", and concludes that "for most creators, a bi-color COB light with good CRI and portable design offers the best balance."

The conclusion is defensible. The word "essential" and the word "matching" are not. Matching a Kelvin figure to an ambient source does not match the light to it, because the Kelvin figure is one of the two coordinates and the crossfade is the mechanism most likely to put a difference in the other one. You can match the number exactly and still have a visible green or magenta split between your panel and the window, and when that happens there is no dial position that fixes it. The guide also never mentions that the dial costs output, which is the part a buyer would most want to know before choosing between a fixed and a tunable panel of the same wattage.

The honest limit

Not one listing on this site publishes a Duv, a chromaticity coordinate, or an output figure at more than one point on the dial. So there is no version of this post that ends with a better product to buy. The numbers that would let you choose on this basis are not printed, and asking for them is a reasonable thing to do and will usually fail.

What is left is a set of free moves, and they are better than the purchase.

Use the ends of the dial, not the middle. At 3200 K or 5600 K one bank is running on its own, the panel is emitting a single phosphor's spectrum, and the result sits on or very near the curve. The middle is where both the output penalty and the off curve tint are at their worst, which is the exact opposite of how the dial is usually described.

Then stop mixing families of source, because that is the problem the dial was sold as solving and it cannot. Your camera applies one white balance to the whole frame. A single source sitting off the curve is therefore completely fixable: grey card, custom white balance, done, and the tint disappears along with the cast. Two sources sitting in different places, a panel at 4400 K and a household bulb across the room, are not fixable at all, because one correction cannot serve both and you will be choosing which half of the frame to sacrifice. Turning the room light off costs nothing and solves more than any dial setting. Closing the curtain costs nothing. Moving the subject so one source dominates costs nothing.

And the limit worth keeping in view: for most of what most people shoot, none of this is visible. A single panel, one source in the frame, a custom white balance off a grey card, and the tint goes with the cast. The dial earns its keep when you genuinely cannot control the ambient light and need to get close to it, which is a real situation and a narrower one than the marketing suggests. If you can control the room, a fixed daylight panel spends all its chips on one spectrum and takes one decision off your hands, and the decision it takes away is the one you were most likely to get wrong.