This is my attempt at proper and simple explanation of the light color subject. The confusion I have seen in this forum regarding light color is so widespread and so horrendous, that I simply couldn’t take it anymore.
Usually such explanations start with “color temperature” represented by values like 5000K, 3500K, etc. I happen to think this fact alone is the main reason for massive amount of confusion which exists in this area. So I am going to ask you to forget everything you think you know about “color temperature”. Forget the term itself. In fact, just forget everything you think you know about light color. Let’s start from the beginning.
The visible light consists of electro-magnetic waves with different wavelengths, each wavelength representing a different color:
Chart 1. Visible spectrum. (*)
If we take a theoretical light source which emits equal amount of energy at each wavelength in the visible spectrum, this is how it’s energy distribution chart would look like:
Chart 2. Illuminant E - equal-energy radiator.
But real life light sources around us do not emit equal amount of energy at each wavelength, they emit different amount of energy at different wavelengths, in other words - different amount of each color. So spectral energy distribution charts for real life light sources are very different from the chart above (called equal-energy radiator). Here, for example, is the chart for a typical noon sunlight:
Chart 3. Noon sunlight.
Although color characteristics of noon sunlight are not perfectly even, they are close enough – all colors are well represented. Let’s add a graph for typical sunset light:
Chart 4. Sunset.
As you can see, color characteristics of sunset light are greatly uneven. There is little green and even less blue. Now let’s add a chart for typical incandescent bulb:
Chart 5. Incandescent bulb.
Notice something? Incandescent bulb has light color properties similar to sunset light. Not exactly the same, but close enough. Now let’s add another theoretical light source – an incandescent bulb whose wire filament is made of non-melting metal heated up to higher temperature than regular incandescent bulb:
Chart 6. Non-melting high temp. incandescent bulb.
This one is similar to noon sunlight! If only we had non-melting metal, we could have this great light source! But alas, we don’t have non-melting metal. So we have to get by with LEDs. Let’s see how a typical “white” LED spectral energy distribution looks like:
Chart 7. LED.
What the hell is this?! Is this a joke?! This looks nothing like a proper light source! A lot of colors grossly underrepresented, some are grossly overrepresented, anyway you look at it – this is a mess! But LED manufactures have been straggling to produce more color accurate emitters, here is a graph for typical “warm high CRI” LED:
Chart 8. “warm high CRI” LED.
This is better. It somewhat accurately approximates sunset light, but remember – sunset light itself does not represent many colors sufficiently. Now the last graph for today, an attempt by LED manufacturers to approximate noon sunlight – typical “neutral high CRI” LED:
Chart 9. “neutral high CRI” LED.
Considering what we’ve been through, this is much better! Still has some way to go, but at least most colors are there, more or less.
Now that you understand how things actually work, we can look at industry standard parameters, like color temperature and CRI, which (in many cases very inaccurately) are used to describe color rendering properties of light sources.
I hope you understand by now, to accurately describe color rendering properties of a light source – you need a graph. The is no way to describe that crazy wave of irregular light sources like LEDs using a single magic number. But that doesn’t stop people from trying…
Correlated color temperature (CCT) tells us which of the following graphs the graph of our light source most closely resembles, while CRI tells us how closely it resembles the corresponding graph:
This leads us to a couple of amazing discoveries:
- Two light sources with EXACTLY the same CCT and CRI parameters could have WASTLY different color rendering properties. Their graphs could be NOTHING ALIKE, but still average out to EXACTLY the same CCT and CRI.
- When it comes to color rendering capabilities, 2000K CRI 100 and even 3000K CRI 100 – are pretty bad lights, they don’t cover a lot of colors properly. While 5000K CRI 100 and even 6000K CRI 100 are great lights, they cover all the colors very well.
I hope you found this explanation helpful. Questions, suggestions and donations in the form of flashlights are most welcome.
(*) All charts in this tutorial are gross approximations. If someone can create better charts for this tutorial (gnuplot?), that would be great.
