October, 2009

LEDs Part II: Color Rendering

Sunday, October 25th, 2009

I just posted the second  in a series of articles on LEDs on the subject of color rendering in LEDs and how it can be improved.

LEDs Part II: Color Rendering

Sunday, October 25th, 2009

This is the 2nd part of our LED article series.

Last time, we discussed misunderstandings about LEDs in general and why they really are a great source for use in light fixtures despite some of their misused and misquoted specs.  In this article we will discuss another misunderstood aspect of LEDs:  the subject of color rendering index (CRI) and LEDs, why they can’t obtain high CRI by themselves (internally) today, and external ways to improve the CRI while waiting for the ultimate LED that has high CRI without external help.  If you want to learn more about color rendering, you can see this article here as well.


Lets start off by saying that it’s pretty much impossible to get a normal daylight (5600K is media production’s standard for daylight) LED at this time that has a high CRI (above 80 and even 80 is really pushing it).  There are a couple of discreet LED components available in the “warm white” (3000K) range that have CRI approaching 85 to 90 but that’s not what people in media production are looking for today.

Increasingly, they are dumping lights in the 3200K (tungsten colored) range for pennies on the dollar and going for all daylight output, electronic instruments that are cool in output, complimentary to the CCD or CMOS sensors their cameras use and economical to operate.  Especially those folks producing media in their own studios, paying for their own electricity and air conditioning—tungsten is not really what they want for a lot of reasons.  After all that’s why Cool Lights was originally created—the cost of electricity gets ridiculous when using conventional lighting and you’re paying for it yourself.  LEDs are more and more being sought after in this “daylight” craze because they have that cool and phenomenal output, energy efficient, easy to keep small and portable, operate off DC sources like batteries, and the list of benefits goes on.  Now if we could just get their color rendering capabilities to be on par with other “electronic” lighting like fluorescents and HMIs/metal halide which are more than suitable in that regard.

You’ve maybe seen me post somewhere on the internet about this CRI “shortcoming” of LEDs before.  So, with that said, anyone advertising a “high CRI” or “90 CRI” single color, daylight LED panel is stretching the truth.  There is simply too much missing in the spectral output of that kind of mono-color LED to be able to do it right.  In the meantime, we’ve gotten around it the same way they did with fluorescent in the old days (before high CRI tubes), by simply adding external magenta “minus green” filters.  That counteracts the overabundance of green and levels off the spectrum enough that the LEDs then play well with other light and render colors more than adequately.

It’s simple to verify what I’ve said here about daylight LEDs.  As I’ve said before, go to every major LED manufacturer’s (Cree, Lumileds, Nichia, Seoul Semiconductor, etc.) website.  Look at their LED offerings in the 5000K to 6000K range.  What is the CRI?  Do they even list CRI as a specification?  If they don’t that in itself will tell the story.  If they don’t list it, it’s because they don’t have a good story to tell there, yet.  So where did these companies producing a “High CRI” LED panel get their high CRI LEDs?  The answer is simple: they didn’t get any at all and the panel is not really high CRI.  You can also verify that for yourself where you ask opinions of people that own these panels.  The common answer is it’s “a bit green”.  One such panel “manufacturer” proclaimed to a prospective customer that their panel was high CRI and when asked if it needed a minus green they said yes.  These two statements are in complete opposition to each other and if one is true the other cannot be true.


Once you get to the level of products using LEDs though, it’s not all “snake oil” out there.  Those that are making a bicolor or RGB type LED panel CAN approach higher CRIs in the 90 range if they do it right.  Those panels fill in what’s missing in the spectrum and become a much better approximation of the kind of light that is capable of higher CRI, like real daylight.  Products from several manufacturers have gone to this approach and while it works well, it can be quite expensive.

Some have even used the simple approach of building in a minus green so to speak.  One well known “1×1” panel seemingly obtains higher CRI by adding magenta tint to the 5mm LEDs they use.  This makes for a weaker LED though!  Adding any kind of filter takes away some of the light.  Whether you do it internally or externally it doesn’t matter—it’s still a filter and filters take away something to get you what you want.

Is it really necessary that multi-color or RGB panels be so expensive if you want to use that method?  The poor man’s approach is to simply take at least two colors of LED and control them independently of each other with manual dimmers so that you can custom mix your color temperature result. Let’s call that the “bi-color” approach and I believe this will be the common method in the near future to give the best CRI with the least amount of sacrifice where LEDs are to be used. When you do this, you level enough elements in the light spectrum which make a mono-color LED daylight panel so lacking in its color rendering capabilities.  You lose some light but not as much (hopefully) as with the filtering (minus green) method.  So, we no longer need external minus green filters with such a panel.  And, if we’ve done our job right and we can select just about any color temperature in the range that most normal videography / photography / filmmaking will use (through the use of the two dimmers) then we don’t need any other filters at all to correct to whatever light color temperature we need.  You should be able to dial in 3200K, 4000K, 5000K, 5600K, 6000K and probably even higher if you like but you need to use your eye or a color meter to help identify when you’ve arrived at the right “mix”.  A full RGB panel should be even better, for color rendering—but not enough to justify the extra cost in money or light output (very weak) in most people’s mind or budget.


The problems of CRI and lighting were most prevalent in use with real film where there is no white balance circuitry to fix color temperature or CRI issues.  As digital mediums have taken the forefront, with their advanced white balance capabilities, this problem becomes less an issue especially when you are not mixing low CRI light like that from LEDs with other higher CRI light.  When you are mixing, your camera’s white balance can’t do its job as effectively.  How can this work?

Most people think that the white balance only deals with color temperature and they also confuse a “too green” issue with color temperature but there is no relationship.  You can have a 5600K light with a bias toward green but its still 5600K color temperature just as another one that’s more balanced in its spectrum, has a higher CRI and less green is also still 5600K.  How is this possible? It has to do with CRI which is not related to color temperature.

Color temperature is the measurement of bias in white light on a scale between red and blue.  Both still white light but one more reddish and the other more bluish.  I have a whole other article on the subject so we’ll only touch on this here.  Color temperature is not used to measure anything else including single wavelength colors like red, blue or green.  On the red bias end of white, we have the lower color temperatures like 3200K or tungsten and lower.  On the blue bias end we have the higher color temperatures like 5600K or daylight and higher.  Where does green fit into this though?

There is another bias or axis used in light measurement and its known as the green/magenta axis.  Many people don’t realize it but the green / magenta axis is actually an indirect measure of CRI which has nothing to do with color temperature at all.  A good color temperature meter (RGB type) will not only give you readings along the blue/red axis (color temperature of the light being measured) but also along the green/magenta axis (which gives what type of correction filter you may need to use–if you need to use one at all it means the light is lower in CRI and probably has a green or magenta imbalance).  Thus, while normally CRI is only measured in an integrating sphere which is an expensive type of testing hardware, you can indirectly measure it and extrapolate what the CRI is by how much out of balance it is on the green / magenta axis. CRI of 100 would be perfectly in balance.

Your digital stills or video camera has a custom white balance which not only reads color temperature of light on the scene in order to correct and make white appear as white, but also it can correct along the green / magenta axis as well when the light isn’t too complex or mixed with multiple sources.  Many that have an offending green fixture can verify this fact easily.  Simply make a custom white balance on a scene lit by the overly green light (lower CRI) and see the results.  Not only does it balance the color temperature so all the colors look correct, it also does a great job in correcting CRI or abundance of green too.  Also, I’m not talking about the preset white balances of 5600K or 3200K–custom white balance is necessary to kick in the magenta/green axis correction.


Back to the original question, why is it so hard to make a daylight mono-color LED with high color rendering?  To answer that question, let’s use the venerable fluorescent tube as example and why we can get a higher CRI tube now which doesn’t require external minus green.  If we can do it with the flo tube, why not the LED?  Let’s see.

Quite a while ago, when fluorescent tubes were less mature, if people wanted to use them in media production either as “practicals” or as production lighting, they needed to add minus green externally to the tube otherwise you got the results of the dreaded “green spike” in your film or video.

As they started to mature and people started demanding not just more light output but also a higher quality light, the simplest way was to add more magenta to the phosphor mix inside the tube to counteract the green.  The filter is now inside the tube rather than outside.  Of course, as in adding any filter to a light, this reduced the output.  Doesn’t matter whether it’s inside or outside the tube the effect is the same.  This is why all things being equal except the tubes, a high CRI and a lower CRI tube (being driven by the same ballast) the low CRI will beat the higher CRI tube in output.


Still don’t understand?  Let’s use another simple analogy.

Light and sound are both waves and subject to control through filters.  The bass and treble control on your stereo is a filter too.  Listen to the stereo with “flat” EQ (no bass or treble added—just the original signal as it was mixed into the sound) and the volume at a certain level.  Then turn down or up the bass knob. You may find yourself reaching for the volume knob afterwards to either turn it up or down as well.  Reducing the bass frequencies in the sound takes some of the “volume” out.  And so it is with reducing the green levels in electronic lighting such as fluorescent or LED.  The wavelength of the color green is also the area where our eyes have the most sensitivity, and thus it’s no coincidence that reducing the green wavelengths in the light through filtering reduces our perception of strength of the light.  To keep with the analogy to sound, where “bass” adds more fullness or perceived volume for our ear’s way of hearing, green is the “bass” of light or what gives it so much “volume” to our eye’s way of seeing.  Reduce the green with a magenta filter (internal or external) and you’ll find yourself wanting to increase the “volume” of the power supply or ballast supplying the light to drive it harder and make up for the loss.

Not a bad analogy if I do say so myself and that should illustrate the issues involved which go far beyond the needs of filmmakers to the needs of people using commercial lighting which is what truly drives lighting development—not filmmakers.


Thus many have gotten around this by simply driving the tube harder and sacrificing overall tube life to do it.  A 55w tube in the range of 70 CRI may put out as high as 3600 to 4000 lumens but as we know is very “green”.  A 55w tube in the 85 to 90 CRI range can give about 3000 lumens output and is more than acceptable in CRI performance.  Take it up to 95 CRI which you may need if you are shooting real film (remember that real film has no ability to correct imbalances along either color axis) and it will drop quite a bit to 2500 lumens.  Drive the bulb harder and you get back to 3000 lumens or higher.  Who knows how much life you give up on the tube to do this however.

Therefore, filtering to counteract the green spike in fluorescent tubes was the easiest way to get the color spectrum in balance enough to improve the CRI and give a better quality of light.  So it will be in the LED world when they finally get around to making a higher CRI mono-color LED.  We’ve already seen the benefit of adding minus green externally to our own Cool Lights LED 256 and 600. Others have done it by simply adding it to the LED itself.  Why haven’t other LED component manufacturers simply added more magenta tint to their LED to solve this issue?  The answer is two fold:

  1. Doing any thing that might reduce light output is not a subject they want to discuss.  LED manufacturers are doing everything they can to increase output and efficiency and are not at the stage yet where they want to reduce it for the sake of quality of light.
  2. It’s still a young industry just like fluorescent was at one time.  They’re still learning to walk.  Once they master “walking” they’ll start looking at running, skipping, hopping, jumping, etc.  After fluorescent matured, they had the engineering bandwidth to look at improving other aspects and so it will be with LEDs.

In the meantime, if filtering isn’t your preferred method, albeit the inexpensive way, to increase CRI at the loss of some light, then we have to look back to the bi-color or RGB methods.  At Cool Lights, we’ve ruled out the RGB approach at least for now. We think the output is too weak, the cost too high, and you have to consider that as well as color rendering—those criteria are also important.  Plus it’s more difficult to dial in the usable color temperatures that are considered paramount to media production.  A microprocessor and interface will be necessary in that case and thats part of the higher cost.

To our way of thinking and for a long time, we’ve considered the bi-color approach to be the best possible of all worlds including economy if you want to get away from magenta filters.  I first approached the subject of a bi-color panel back in 2007 on two different filmmaking message boards and it wasn’t well understood what the benefit would be.  Most viewed it as a sacrifice in light output if you have half high and half low color temperature.  If you wanted only the higher one then all the lower are off.  I think people are starting to understand now and the time is right for bi-color products with higher CRI. They won’t be as economical as the single color panels and you sacrifice a bit of output in the name of CRI (not as much as in RGB types or with magenta filters though) but it’s a good overall compromise, so we have started down that road.

My first small panel from the summer of 2007 literally had 3200K and 5600K LEDs in alternating rows with two dimmers.  So if you wanted 3200K, you did turn off all the 5600K and vise versa for the opposite end.  To have both on at the same time would have given something in the range of 4500K which isn’t a standard for media production.  So you would sacrifice half the panel to get one common range or the other.


We later came to the conclusion that the best overall panel will produce daylight in the range of 5600K to 6000k when all the color temperatures are active, thus giving the best output and in the most commonly used range today.  To do this requires a custom mix of LEDs both of higher and lower color temperature than normal media production standards.  In such a panel, all the low color temperature LEDs will be something in the range below 3000K and all high color temperature LEDs will be something in the range well above 6000K.  Thus, in the extremes when you use only the low or high color temperature range those will be used for “special effects” or rare cases and while you lose half the panel, it’s okay because those are just special cases and not often encountered.  In the normal cases of needing 5600K, which is in high demand today,  you have the entire panel contributing to that so all LEDs are firing–nothing is wasted.  In the case of needing 3200K, you have much of the high color temperature dimmed but not entirely so while you lose some of the entire output you still have quite a bit.  In the case of needing something like 4100K to match practical fluorescents in an office or commercial location, you would turn down the high color temperature by say 50% or so. And so on…

As previously mentioned, a panel like this can be controlled simply by two manual dimmers to custom mix the color temperature you need.  You can do it either by eye or if you have a reliable, RGB type color meter, you can use that as well to custom mix until you have the color temp you need on a shoot.  You could also use a microprocessor controller to dial in a color temperature through some kind of control and display, but that starts getting back to an expensive approach.  Just as tooling and mold costs must be amortized into a fixture to recover that as you go along, so would be software/programming costs so amortized and they would be considerable.  The two dimmers is a good starting approach and we can ease into a microprocessor approach over time as we get access to economical programming resources.


While waiting for the ultimate daylight, high CRI LED components to appear on the market, we have other choices to make a high CRI LED panel!  We can continue to use an economy mono-colored daylight or tungsten LED panel and add minus green filters as necessary or we can go to a bi-color, two manual dimmer approach and not need filters of any kind at all.  So, we don’t need to wait for that day, we have our solutions and they are both doable now.

(c) Copyright 2009 All rights reserved. May not be used in part or total without the express permission of the author in writing. Waiver of damages:This information and all included material are provided as is, and Cool Lights USA can not under any circumstances be made responsible for any damage, injury or losses caused directly or indirectly by implementation of the information in this article.