Mixing LEDs to make full color lighting seems straight forward on the surface.  Turn on red for red, green for green, and both for yellow.  Many color fixtures work this way in fact.  There is a separate DMX channel for each color under control and the fixture turns each color on in proportion to its DMX value.  This method works if the application doesn’t require a particular repeatable color.

If you do need a particular color, adjusting color sliders and guessing at the output isn’t a great way to get there.  What makes for better control is a color picker as is used in photo editing where you can adjust easily understandable hue, saturation and brightness values.  It would also be a benefit if the color a light fixture produced with a particular value set was the same as is produced on your device or computer display with these values.

For a fixture to operate from hue, saturation and brightness it has to calculate the color target from these values then calculate the drive level for each of its sources to achieve the color target.  The fixture can use information about each of the emitters efficiency and color at different operating currents and temperatures to calculate the required drive level.  This method works with more than just RGB fixtures.  If the fixture also has white and amber emitters the color target defined by the hue, saturation and brightness values doesn’t change.  The fixture will combine its additional sources into the drive level calculations to utilize all the different color LEDs it might have.

You might want color translation in your light fixture if you want your fixture to emit a particular color, you want the color to be consistent between fixtures, and you want to set the color in a convenient way.

To achieve a CIE color coordinate target with LEDs requires mixing accurate light amounts from the LED sources and knowledge of the precise color that an LED will generate under the given excitation and ambient conditions.

The requirements for accurate color illumination with LEDs are derived from the sensitivity of human color vision.  Human vision translates color by the ratio of red, green, and blue cone simulation.  The red, green, and blue cones do not only respond to red, green, and blue light.  There is continuous overlap in cone excitation.  The red cone is excited at some level from deep red through blue, the green cone is stimulated by red-orange through blue-violet, and the blue cone is stimulated by yellow-green to deep violet.  The total of all three cone stimulation amounts follows a smooth bell curve that peaks at green.  The sensitivity of the eye to color differences depends on how much the cone stimulation ratio changes in that area of color.  The eye is most sensitive around the area of warm whites and least sensitive in the saturated green area.  A green mix that has a 5% error in the amount of light coming from the green source may not look off target but a warm white mix with that amount of error would be obviously off target.

Human brightness perception is also helpful to understand.  Like many other senses brightness perception is not linear.  For a perceived doubling of light intensity the number light light photons coming from the source must increase by much more than double.  An easy approximation mathematically is the use the square so that a perceived doubling requires a four times increase in the source photon emission.  If this is translated into dimming, achieving a perceived 10% dimming level from full brightness requires the source to emit 1% of the photons that it does at full intensity.

To design a precise color illumination system it is necessary to understand how an LED behaves.  LEDs do not make light photons at any current level.  There is a minimum current at which the LED will make photons and around this level the color of the light of the LED changes a lot making it difficult to compensate for.  This level is typically below 1% of the maximum LED current.  When operated above the minimum current level the LED’s color shifts with the current and operating temperature an amount that is dependent on the LED’s color.  Some red LEDs change in intensity by a factor of two as the operating temperature changes from minimum to maximum.  Some Green LEDs change in color several perceptible steps as the operating temperature changes.  All LEDs decrease in efficiency as the current increases.

To generate an accurate color illuminant from LED sources an LED drive level is calculated based upon the LEDs’ intensity and color characteristics over the operating temperature and current levels.  Some of these characteristics like color drift are consistent enough for a particular LED that the same data can be used for all fixtures if the color accuracy requirement isn’t too great.  Other characteristics like LED intensity can change more between parts or lots of parts.  This data would need to be collected for each fixture produced as a calibration step in the production process.  If the color accuracy requirement is sufficiently loose calibration wouldn’t be necessary.