Measuring the Color of Light

If you've been using a camera for a while, you may have run into the fact that different kinds of white light are actually composed of differing amounts of the colors of the spectrum. The light emitted by an incandescent tungsten bulb is composed of more yellows and reds than blues while the light emitted by the sun is composed of more nearly equal parts of the color spectrum. Unless you've worked with different kinds of light for a while, however, it's something that you may never have noticed to any substantial degree. Our human visual cortex is extremely good at processing out the image data it receives so that we don't normally perceive these differences much unless we train ourselves to notice them.

Unfortunately, the digital sensors in our cameras don't have the same level of visual latitude as our brains and the software that interprets the sensor data needs to know the color composition of the light illuminating the subject in order to produce an accurate image. This can be done either manually or automatically, with varying levels of success. If this value is way off—say you create a photograph outdoors with your camera set to a tungsten color temperature setting—your images will have a massive color shift to them and appear ultra-blue toned. Vice versa, if you make a photograph indoors at night on fully automatic mode and your camera doesn't dial in a correct setting for the light used, which happens more often than we'd like, everything will appear orange-tinted.

To talk about the color composition of light, we use what seems at first to be an odd term: color temperature. And, since we're talking about temperature, we use the SI unit of the Kelvin as a unit of measurement. I know the first few times I delved into this subject, I thought this was rather strange. After all, light itself doesn't have a temperature. Since it's radiation, when it hits a surface, it can cause that surface to heat up. But the light itself is neither hot nor cold. So, why the heck do we use temperature to describe the color composition of light?

You can thank science for this. Science defines things in terms of other things. A meter is defined by how far light travels in a vacuum in 1/299,792,458th of a second. A second is defined as the duration of 9,192,631,770 periods of the radiation corresponding to the transition between the two hyperfine levels of the ground state of Cesium 133. So, it should come as no surprise that the color composition of light is measured by comparing it to something. That something happens to be the light radiated by a theoretical black body when heated.

"Cesium say wha? Black body? What the heck is a black body?", you ask. Good question.

A black body is an object that absorbs all light that falls on it. It doesn't reflect any light, so when it's cold, it appears as black as black can be. Because it is non-reflective, any light it emits when it is heated up is strictly the result of thermal radiation. The properties of a black body are fundamental to Planck's Law of black-body radiation and are part of the foundation that supports quantum electrodynamics.

"Um, yah, you're losing me with that quantum talk. Don't you dare start throwing equations around!"

You're not alone. I don't follow it all either, especially when the equations really start flowing. But, it's important to note that a black body is a theoretical construct. To the best of my knowledge, nobody has one in their pocket that they can pull out and show the class. Instead, lets use a proxy that you've at least seen on television. Think of a chunk of iron for a moment.

When you heat up a chunk of iron in a furnace, at some point it gets so hot that it starts to glow a dull red. Heat it up some more, and its glow turns a bright red and then starts to turn orange. At some point the iron melts, but keep heating it up even more and its glow will continue to change color. The change in the color that the iron glows corresponds to its temperature. This correspondence works so well, that you can actually perform the reverse and derive the temperature of hot iron by measuring the color it is glowing. This is especially handy if you are measuring the temperature of things that are hot enough to melt a traditional contact thermometer.

Back to black bodies. Imagine you have a handy pocket-sized theoretical black body at your disposal, along with a nicely efficient pocket torch. When you fire up your torch and heat up the black body, it will start to glow at some point. When it reaches a temperature of 2900K, or about 4800ºF, the color composition of its glow will roughly match the that of light emitted by your typical 100W tungsten lightbulb, with a strong bias towards the red and orange part of the spectrum. Keep on going to 3200K, or 5300ºF, and the glow will match most photoflood lights. Now, crank it up to 5500K, or about 9300ºF, and your black body is now glowing with roughly the same color composition as the light from the sun as it appears at noon in the summer.

The graph to the right shows the relative composition of colors of the light emitted by our handy portable black body at various temperatures. As you can see, the light emitted by a 3200K source contains more reds and oranges than blues, which closely matches the spectrum emitted by a photoflood tungsten light. The light emitted at 5500K contains a more even set of colors, typical of daylight. And the light emitted at 9000K, typical of the light illuminating shady areas or that you may find on extremely overcast days, contains more blues than reds and oranges.

Of course, long before you heat up your black body to the point where you can compare it to sunlight, almost ten-thousand degrees Fahrenheit and close to the temperature of the surface of the sun, you've most likely burned your fingers, dropped the black body on the floor, and set your house on fire. Please only attempt this in a properly equipped laboratory setting.

All fun aside, a black body, or even a decent approximation of one, that can operate at all the temperatures that we're talking about isn't something that you or I can probably play around with. You can apparently buy a black body simulator, but the most of the ones that I have found online, such as the SBIR 4100 Cavity Blackbody, max out at around 1000C (1273K), which is a bit wimpy for comparing to the color composition of light. As well, there's not a price listed, which means it's expensive.

Probably the best approximation of a black body for our purposes, and one that we have actually do easy access to, is the common household lightbulb. When turned on, the tungsten filament of a lightbulb is heated by electricity to between 2000K and 3000K, depending on the design of the bulb and how much current is used, and emits light with a color temperature of about the same value. This works, of course, because the filament is protected from oxygen by either a vacuum in older designs, or by a mix of inert gases.

Really, all of this is a bit academic. In the real world outside of the lab, people don't think about the details of how meters and seconds are defined and photographers don't think about light in terms of black body radiators. You know from experience that a meter is about that long, while holding out your arms. Well, some of us that are stuck with a system of archaic units know it as being about a yard long, or 3 feet. Similarly, there's no real reason to worry about heating up black body radiators, or comprehending Planck's law of black body radiation, when dealing with the color of light.

Instead, you can just use the values of color temperature and keep it in your head that daylight has a color temperature of around 5500K, standard photo tungsten lights is 3200K, and a 100W lightbulb comes in at around 2900K. When you set your camera to use the daylight or tungsten preset, you're using one of these values. Once you get the feel for the various color temperature values, you'll be able to master setting of the color temperature value for the making or processing of a photograph. More importantly, you'll have control over the fourth variable of photography along with exposure time, aperture, and sensor sensitivity. It's like knowing that you need an exposure of 1/1000th of a second, er, I mean 9,192,632 periods of Cesium 133 radiation, in order to freeze the some kinds of movement.

Related Posts:

• Color Temperature and White Balance, which builds on this post and discusses how color temperature is used in white balance.

Related Links:

• Timeline of Atomic Clock History, via Josh Susser.
• What Color is the Sun?
• Pyrometry, the non-contact measurement of temperature, on Wikipedia.
• A discussion of Planck's Law, on Wikipedia.
• The NIST Reference on Constants, Units, and Uncertainty
• Wikipedia's Color Temperature article

Updated 4/9 to add a graphic of the composition of colors in light at various color temperatures.