36
\$\begingroup\$

Color doesn't have actual temperature. Try putting a blue square and a red square up on your monitor and hold a thermometer against both regions. If you find that there's a difference, you're doing it wrong. You probably know this already.

So why is color temperature measured in Kelvin? Kelvin is a measurement of the heat in a substance from absolute zero. That means, when there is actually no heat whatsoever in a substance and the molecules in it are absolutely still, that's 0 K. 0 K may not actually be possible, but that doesn't stop us from measuring relative to it, and this is a digression anyhow.

Is there some substance that emits different colors at different temperatures, which has been used as a reference to map temperature to color temperature? Or is it more complex than that? Or is the choice to use Kelvin completely arbitrary, with no relation to heat at all?

\$\endgroup\$
6
  • 2
    \$\begingroup\$ All substances emit light at 2000K or 4000K! For example, the wire in your light bulb does.The light of this red-hot or white-hot glow has that color temperature (2000K, or 4000K, or whatever). And the sun's surface temperature is ~5800K which is therefore the sun light's color temperature, cum grano salis because of atmosphere and such. \$\endgroup\$ Jan 14, 2016 at 9:11
  • 3
    \$\begingroup\$ Related: How is temperature related to color? and Can two 2500 K light bulbs replace one 5000 K bulb for growing plants indoors? (and likely quite a few others) on Physics. \$\endgroup\$
    – user
    Jan 14, 2016 at 10:15
  • 1
    \$\begingroup\$ Color doesn't have a temperature, but temperature has a color. It's called black-body radiation. I agree it's a strange way of measuring hue, but it's as good as any other scale. \$\endgroup\$ Jan 14, 2016 at 19:59
  • 1
    \$\begingroup\$ Slight nit-pick- I made a very sensitive surface measuring thermometer and you could see noticeable differences in temperature for different printed colors- the emissivity was different so the ceiling lights warmed some colors more than others. If the air and other things were not sucking heat away eventually that paper would heat to the 2700K that it was 'seeing'. \$\endgroup\$ Jan 16, 2016 at 8:19
  • 1
    \$\begingroup\$ @scottbb Back in Uni (in Toronto) the massive new library building had a graffito on the sidewalk in front- 233°C. Oh, how we laughed. \$\endgroup\$ Jan 17, 2016 at 18:54

5 Answers 5

38
\$\begingroup\$

It is related to a heated substance, albeit in a somewhat theoretical way. The substance is an ideal incandescent black body, which would radiate a given color within a given color space at a given temperature. The location within the color space vs. temperature is called the Planckian locus, and I don't claim to understand everything in that article, but explore it to whatever depth you'd like.

For a more general "light reading" explanation of color temperature and it's correlation to black body radiators, see Wikipedia's Color temperature article.

\$\endgroup\$
0
28
\$\begingroup\$

Wikipedia's introductory statement on color temperature relates them quite well:

The color temperature of a light source is the temperature of an ideal black-body radiator that radiates light of comparable hue to that of the light source.

Black body radiators are an idealized concept, that radiate an energy spectrum with a peak intensity at a frequency that depends on the the temperature of the black body radiator. The higher the black body's temperature, the higher the peak frequency of the emission spectrum of the black body radiator. Any emission from an ideal black body radiator is purely from heat energy. Thus a 6500 K black body emits photons whose frequency spectrum peaks at what we have called 6500 K color temperature (in the blue-white, "daylight", color temperature range).

While there are no actual black body radiators, there exist several decent approximations that act quite a bit like black bodies. Stars, incandescent light bulbs, and electric range stoves are examples. That is why 5500 - 6500 K is called daylight color temperature — we measure the sun's black body temperature at around 5780 K. Similarly, because incandescent light bulbs are not light emitters so much as heat emitters in the visible light spectrum, the "indoor" color temperature of about 2500 K is the nominal black body radiation temperature and spectral peak of incandescent bulbs.

Related questions here at Photography.SE:

This Physics.SE question also addresses the current question: How is temperature related to color?

\$\endgroup\$
11
  • 7
    \$\begingroup\$ @JDługosz Maybe you just haven't made fotos in a room with light bulbs as the only light source. I've seen it very often. Many modern LED-lamps have 2700K. \$\endgroup\$
    – Zenit
    Jan 14, 2016 at 8:35
  • 3
    \$\begingroup\$ @JDługosz And you are right, the color can't be corrected very well. Human sight doesn't care much, but I haven't seen a lot of cheap cameras that would be able to correctly white-balance photos taken in rooms with such lighting - they are always very yellow-reddish. But they are one of the main categories of indoor lighting, and some people prefer them (they are supposedly more relaxing). \$\endgroup\$
    – Luaan
    Jan 14, 2016 at 9:54
  • 2
    \$\begingroup\$ @JDługosz I regularly use color temperatures in the high 2000s and low 3000s in Lightroom. Using 4000K for a scene lit mainly by light bulbs would look way too orange. \$\endgroup\$
    – JohannesD
    Jan 14, 2016 at 10:20
  • 9
    \$\begingroup\$ Just to make it absolutely clear - an incandescent bulb is actually a pretty fair approximation of a black body and the actual, physical, temperature of the filament when running is typically around 2250°C (or ~2500K). The sun is also a pretty decent black body and its actual, physical, temperature at the surface is about 6000K. \$\endgroup\$
    – J...
    Jan 14, 2016 at 13:56
  • 3
    \$\begingroup\$ And, this is getting slightly astray, but if you're wondering why daylight color temperature is a range when the sun's temperature is pretty much fixed, it's because the sky is blue — that is, the atmosphere scatters more of the sun's blue light than its red, so the color temperature of a scene is lower or higher depending on how much direct and how much indirect sunlight it gets. \$\endgroup\$
    – hobbs
    Jan 15, 2016 at 0:26
18
\$\begingroup\$

The colour temperature is related to the black-body radiation produced by hot objects. The black-body radiation curve, shown below, shows the approximate intensity* curves at each wavelength for the radiation emitted by bodies at 5000K, 4000K and 3000K.

* It actually shows the spectral radiance curve, which is a kind of flux. But you can think of it as an intensity if it helps. The two quantities are closely related.

Black-body radiation curves for objects at different temperatures

Image Source: Wikipedia

Note how the curves pass through the visible spectrum. Depending on how much of the (area under the) curve is in the visible spectrum, the colour will look different. This is described by the Planckian locus when talking about colour temperature.

Blackbody radiation curve on CIE chromaticity diagram

Image Source: Wikipedia

The CIE diagram above shows the visual colour of bodies at various temperatures. Bodies with temperatures around 3000K tend to look red, while bodies around 5000K or 6000K will look whiter. Bodies that are hotter than this will tend to look blue.

\$\endgroup\$
9
\$\begingroup\$

As the other answers note, colour temperature corresponds to blackbody radiation at that temperature.

But why do we care about that? To understand that, you must first ask yourself "What is white?"

Physically, white isn't a colour. There's no wavelength of light that corresponds to "white", just like there is none that corresponds to "black" or "grey" or "pink" — all of those colours are just "artifacts" of human perception. Physically, they are a mix of a many different wavelengths (in natural light in particular, white is by definition the mix of all the visible wavelengths of the Sun).

Human colour perception depends on mixing the intensity of three different light-receptors. Now, each of those actually covers a broad range of wavelengths ("physical colours"), so this is a bit more complicated, but each of them has a peak at a different wavelength — we usually call them red, green and blue respectively. This is how computers can display all the colours we can see with just a mix of three different wavelengths — some intelligent alien with a different sight would just think we're all full of nonsense, because our pictures look nothing like the real thing. Basically, we tweak the intensities of the three wavelengths (that roughly correspond to the peaks) to produce the same excitation in the photoreceptors that real light would.

In this model, "white" means "100% red + 100% green + 100% blue". However, as I've already noted, natural white light doesn't really work like that — it's a composite of many different wavelengths without such pretty ratios. Now we come to evolution: white is the colour that doesn't change the hue. Colour perception is balanced to allow us to still see the same colours even when ambient lighting conditions change — for example, when walking under a forest canopy, or when dealing with scattered light (e.g. "in a shadow"). This also means that the natural colour temperature corresponds to the temperature of the sun's photosphere - basically, sun is white by definition, because that's what evolution adapted us to (the reason it looks yellowish to the eye is because some of the blue light is scattered away by the atmosphere — our sight adapted to see objects illuminated by the Sun (and the atmosphere), not to see the Sun itself).

The fun part is that this also allows us to use light sources that aren't as hot as the Sun. The simplest examples are incandescent bulbs which tend to have lower temperature, but use the same basic principle — make the wire hot enough so that it radiates enough visible light to make the white balancing work for humans. LED lights use a principle more like your computer screen - three distinct (well, not exactly three, but "three narrow bands") wavelengths to produce any colour. The good thing is that this is much more efficient. The bad thing is that it can actually produce visibly different light effects, so it doesn't really map to natural light at all.

But the core is: LED lights are nowhere near their "colour temperature", so what meaning does colour temperature have in that case? The main point is that under different temperatures, the intensity of signals produced at each of the three photoreceptors is different (for the same "colours"). When you change colour temperature on your monitor, you're basically tweaking how intense each of those three channels is in relation to the others — that's what gives you the "reddish" or "blueish" hues. You're simulating the effect of a different blackbody temperature on human sight — and since human sight ignores so much of the information in light, it actually works quite well most of the time. When doing the setting on your camera, you're doing the exact opposite - you're trying to map the "shifted" colours to the "objective" Red+Green+Blue data. The reason the setting usually uses colour temperature is simply because that's what's used everywhere - you can have a look at the colour temperatures of your lighting and use that on your camera as well.

\$\endgroup\$
6
  • \$\begingroup\$ This is a good answer that approaches the question from a practical, rather than purely technical standpoint. The point about artificial light sources not being as physically hot as their colour temperature is also a good one. \$\endgroup\$
    – Andrew
    Jan 14, 2016 at 15:07
  • \$\begingroup\$ @Andrew Except for incandescent lamps - they aren't as hot as the sun, sure, but their color temperature is also much lower to match. \$\endgroup\$
    – Random832
    Jan 14, 2016 at 16:46
  • 3
    \$\begingroup\$ @Random832 of course. Different light emission mechanisms work in different ways. The colour temperature is a concept that comes from black-body radiation, but the perceived colour (and white balance effects) aren't always due to black-body radiation. \$\endgroup\$
    – Andrew
    Jan 14, 2016 at 16:48
  • 2
    \$\begingroup\$ @Luaan Why do you need some alien? You can use mantis shrimps too consider us to be "too color limited" :) theoatmeal.com/comics/mantis_shrimp \$\endgroup\$ Jan 14, 2016 at 17:46
  • 2
    \$\begingroup\$ @woliveirajr Well, mostly because of the "communication" part. Pidgeons might think we're retarded for ignoring all that UV light, but they're the ones running into windows all the time, so... :D \$\endgroup\$
    – Luaan
    Jan 14, 2016 at 20:05
1
\$\begingroup\$

Before the thermometer smiths and potters and glassblowers and the like depended on the color of the glowing material to monitor progress. It was believed that most minerals had a unique color at different stages as heated. It was also known that objects expand and contract as their temperature changed. Daniel Fahrenheit (German 1686-1736) devised a mercury thermometer. He used the number 180 as the number of steps (degrees) between freezing and boiling water, 180 being a highly divisible number. Anders Celsius (Swedish (1701 – 1744) thought the 180 business was crazy. Celsius put 100 steps between freezing and boiling water.

Mercury, alcohol, and other liquids were common used in thermometers, however none expand or contract linearly thus the markings on the tubes have different spacing at different regions. In 1802 Joseph Louis Gay-Lussac (French 1778 – 1850) showed that the coefficient of air and various common gasses are about the same. A tube with a float atop a column of hydrogen falls and rises uniformly with temperature. If cooling continues the float should hit the bottom at -273C. Scientists abhor negative temperatures, and named this bottoming out as “absolute temperature”. Thus the Absolute Scale now called the Kelvin scale to honor William Thomson 1st Baron Kelvin (Irish 1824 – 1907 Nobel Laureate) for his work on Black Body Radiation).

A temperature in the Kelvin scale can be converted to the Celsius scale by adding 273. Metallurgists commonly used the Kelvin scale as did many other branches of science. Lightbulb designs evolved to use the metal tungsten as their glowing filament. The lighting industry adopted the Kelvin scale to describe the color that lamps produced. The photo industry, highly dependent on artificial illumination, adopted the Kelvin scale to classify color.

Table of some selected practical sources of illumination and their color temperatures.

Sunlight Noon 5400K

Skylight 120,000K to 18,000K

Photographic Daylight 5,500K (agreed to by film makers)

Flash Cube - Flip Flash 4,950K

Clear Flashbulb (zirconium wire filled) 4,200K

Clear aluminum wire filled flashbulb 3,800K

500 watt Photographic lamp 3,200K

100 watt household tungsten light bulb 2,900K

60 watt household tungsten light bulb 2,820K

\$\endgroup\$
5
  • \$\begingroup\$ Do you have a citation for the Fahrenheit scale being initially defined as 180° from freezing to boiling? I learned it as 96° from the freezing point of salt water to human body temperature. \$\endgroup\$
    – mattdm
    Mar 15, 2018 at 18:09
  • \$\begingroup\$ @ mattdm -- From a text book by Asimov and Zimmermann "Fahrenheit: Facts, History & Conversion Formulas". Retrieved 16 September 2017. \$\endgroup\$ Mar 15, 2018 at 18:52
  • \$\begingroup\$ Thanks! From what I found online, it appears the scale was initially as I said but recalibrated later. livescience.com/39916-fahrenheit.html \$\endgroup\$
    – mattdm
    Mar 15, 2018 at 19:13
  • \$\begingroup\$ @ mattdm -- The 180 degree spread is significant as it is an aid to that helps students understand better, the conversion formula i.e. one degree C larger by 1.8 steps (degree) than a Fahrenheit step. \$\endgroup\$ Mar 16, 2018 at 0:31
  • \$\begingroup\$ Yeah — I'm kind of amazed that I never knew that, having only learned the origin story and not about the later adjustment. \$\endgroup\$
    – mattdm
    Mar 16, 2018 at 0:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.