Carbon-Friendly / Eyeball-FriendlyMichael Herf, November 2009
In the rush to manage carbon emissions, one of the first victims is the incandescent lightbulb.
After seeing the green glow of the hardware store CFL, many wince. And some ask, "What is this doing to my artwork?"
My wife's an artist and illustrator. One night several years ago, she was up late working on a contest submission. To help her work at night, I gave her a "full-spectrum" Ott-Lite. "Just paint with this. It'll be like sunlight!"
After working through the night, Lorna woke up the next morning to find that the painting that had looked good under that light looked quite pink in the light of day.
Like many CFL-based sources, this "full-spectrum" light had accentuated the greens in her paint, and she'd managed to compensate by mixing her paint in the opposite direction, a sort of orange/pink color.
For many color-sensitive jobs, CFLs, LEDs, and other new light sources are proving difficult to work with. And for others they just don't look right.
For the past few years, we wondered how do we do this right? How do we pick the right color temperature? How do we find full-spectrum lights that are actually good enough? And why do even 90CRI lights look a little strange sometimes?
Along the way, this obsession, talking about lighting at all hours, led to making f.lux, a color temperature tool for computer screens, and the purchase of lots of lighting gear, and after all that, a pretty workable art studio with great lighting, day or night.
CRI and color quality
Most lighting gear advertises a "CRI" number, the "Color Rendering Index". To describe how well a light source works when you point it at a bunch of different color swatches, CRI is the best tool we've got to work with today. Most light sources are rated this way.
CRI attracts quite a bit of criticism. Most notably, it gives too high a rating when one color is out of whack but the rest of the colors are fine. Also, it uses a small number of swatches to do its rating, so manufacturers can "cheat" and have inaccurate colors that sneak between the swatches. See the new CQS proposal site for ideas about how to fix all this.
Despite the cricisms, CRI works pretty well: a CRI=98 source is really good, and while you might occasionally like the way a CRI=86 source looks better than a CRI=90 source, you can usually trust that a higher CRI gives better portrayal of color.
Also, the scale is really "distance from 100" so it can be helpful to think, "CRI=96 is twice as good as CRI=92, not 4% better."
The better CFLs you can buy today at a hardware store have CRI=85, and specialized CFLs can achieve CRI=91. If it's not marked, it's probably 80 or lower.
CRI rules of thumb
For general room lighting, 85-90CRI is usually adequate. Lights in this range may have a slight green cast. For task lighting or color matching, it's preferable to look for 95CRI or better. Most people won't tell you this is necessary, but it really helps, and it's possible to do without spending too much money.
Light generated by "burning stuff" is measured using "color temperature" and the quality of the light is compared with the blackbody radiation curve curve, which usually we perceive as "warm" (like a candle) or "cool" colors (like north-facing sunlight).
Daylight has a "correlated color temperature" of 6500K ("6500 kelvin"), and the north-facing light that artists love is at least 7500K, a more blue-balanced color.
Incadescent bulbs emit 2700K, and halogen sources around 3400k. (These look more "orange".) But all of these sources have CRI=100, since we measure them against the ideal planckian radiator, which they match pretty closely.
While you might think that a 2700K light causes color distortions, it does so in a consistent way, which your eye can compensate for (see "Von Kries" adaptation below). CFLs and other sources distort colors in a way that your eye doesn't easily fix, for instance by creating "spikes" of extra green that distort particular colors.
Painting In Daylight
Some artists prefer to paint in general room lighting, using incandescent or halogen lights. These can give a more flattering appearance to skin tones.
However, most artists I know prefer to paint in light from 4700K-5500K, to approximate daylight. At first, making a transition from a 2700K room to a 5000K studio may feel "blue", but with a high-quality light, color reproduction is maximized at these color temperatures. Also, using true daylight sources in your studio will help you paint during the day and at night without having colors change at sunset.
So-called "daylight" incandescents like the GE Reveal (neodymium coating on an incandescent) are actually purple-coated 2700K bulbs, and their CRI can be as low as 70-80, so they are not good sources for most art applications.
The best sources available for painting or viewing art today at 4700-5000K are from SoLux, which uses a special coating with halogen bulbs. Their products are available mostly in the MR-16 format (a halogen with two small pins commonly found in track lighting.) SoLux achieves a CRI=98 at 5000K.
Replacement bulbs for daylight
For bulbs you can put into a regular socket, we've had moderately good luck with the CRI=90 Westcott Spiderlite bulbs: available at Adorama. They are big, but they work reasonably well.
Replacement bulbs for "warm white"
It is difficult to find a 90+CRI "warm white" CFL bulb. The newer LED bulbs appear to be taking the lead here. They are extremely expensive right now, but prices are coming down quickly.
For newer LED "solid state" lamps, you should avoid almost all the LED bulbs you find at the hardware store. Lemnis Pharox (owned by Philips) appear to have reasonable light output (like a dim 40-watt bulb), and reasonable color quality.
For "downlights" that can be used in your ceiling, the Cree LR6 appears to be unmatched. 92CRI, and dimmable too. $90 a pop.
Other form factors
If you are serious about producing or viewing art, you may consider installing special lighting gear for just this purpose. Most of the interesting products we've found are based on tube-fluorescents and modified MR-16 sources like the SoLux.
All of the products listed here are for daylight color temperatures (usually 5000K), though some products offer cool white equivalents.
Solutions we've tried at 4700-5500K:
If you prefer warmer lights, there are 2900-4100K models available with similar CRI:
Most of the recent legislation and effort around carbon emissions has been in maximizing the light emitted per unit of energy (with the rest going to heat). For instance, CFLs can produce 60 lumens per watt easily, and incandescent sources may only produce 15.
But what we find is a sort of "cheating CRI" with the goal of improving lumens/watt. Many high-efficiency CFL sources produce an excess of green light. According to Wikipedia's chart, the theoretical maximum efficiency for 555nm green light is 683 lumens/watt, whereas "white light" can only achieve 251 lumens/watt, so the payoff in efficiency for distorting green response is big: about 2.7x.Some efficiency numbers briefly (from wikipedia):
Legislation and Future Tech
Europe has already begun the transition away from traditional incandescent sources, and the US has signed the The Energy Independence and Security Act of 2007, which effectively bans traditional bulbs starting in 2012.
Also there is a new contest in the style of the "X-prize" (called the "L-Prize") that encourages manufacturers to produce a 90+CRI source that achieves at least 90 lumens per watt. L-Prize details here.
Many of the promising new technologies are LED-based, with several manufacturers moving efforts away from CFL technologies entirely. The US Energy Department called CFLs a "debacle". A NYTimes article noted that adoption of CFLs in some areas is down 35-50% since 2007, despite promotions and advertising.
For examples of future technologies, see Philips' entry for the L-prize, a new LED lamp from Cree that looks very promising, and even this project in Los Angeles to replace street lights with LED sources.
Computer Display Color
Why do the CCFL and LED backlights used for computer displays not cause similar color distortions to their "general lighting" counterparts?
Spectral errors (e.g., a "spike" in a particular green frequency) are most noticeable when a light reflects off a particular surface and combines with the colors there. For instance, reds may be rendered properly by a light, but greens may be 5x brighter than expected.
On a computer display, we display RGB channels separately, and we can measure and compensate for these errors, having the display and computer work together to compensate. The resulting colors match the ones that your eyes' three color sensors expect.
Von Kries Adaptation
Why do "whites" look about the same at different times of day, even though the actual colors vary quite a lot? The "von Kries hypothesis" explains this, suggesting that the three receptors in the eye can change their "gain" in different environments in order to compensate for different lighting conditions.
In particular, Von Kries explains why we can perceive colors reasonably accurately under a wide range of black body light sources. A corollary is that we can't perceive colors accurately when particular single frequencies are emphasized in a non-uniform way, as we see with fluorescent sources.