Discussion in 'Digital Darkroom' started by emre, Apr 28, 2005.
What is the difference between 5000K/6500K and D50/D65? I think it's
about time I learned!
Color Temperature is a number (in Kelvin) that describes the color of light emitted by what is called a blackbody radiator. This is a theoretical object (the closest item in the real world that behaves as a blackbody is our sun). It's somewhat dangerous to use color temperature to define what you want because the reality is if all light sources were true blackbodies a particular color temperature would produce the same color of light. Because natural materials are not theoretical blackbodies, heating them to a specific temperature creates deviates from the theoretical color from magenta to green. It's really much safer to use the term correlated color temperature (CCT) because many colors of white may correlate to the same blackbody color temperature. Different illuminants can have the same correlated color temperature. This is one reason why the CIE defined the Standard Illuminants.These illuminants are defined spectrally meaning a certain amount of energy at each wavelength across the spectrum. This is an exact and non ambiguous description of color. D65 is an exact color, it is not a range of colors. If you have a color meter that reports color temperature of a light source many light sources that appear different could read the same, that's kind of a problem! Rather than say 6500K, it's much safer to say a CCT of 6500K. The color you get from a 6500K light source can vary due to the shift of magenta/green. D65 is an exact, non ambiguous color that we can fingerprint exactly by viewing it's Spectral Power Distribution (SPD).
5000K is approximately "soft daylight" and it's the standard used to match prints to slides to "reality." "Daylight" film is optimized for approx 5000K, morning light from your window is probably about 5000K, "daylight" fluorescent lights try to approximate 5000K, and traditional photographic color labs use 5000K slide and print viewing systems.
6500K is part of a more purely digital system that has only indirect non-digital relationship to "reality" (6500K not experienced except perhaps at Himalayan altitudes)...it's not appropriate for visually evaluating photographic prints because nobody lights galleries or homes with acetylene torches. I hope.
Fascinating...so D50 and D65 are CCTs? It appears that a range of CCTs (the lines) map to the same color temperature. In general, is Dxx the CCT of xx00K, or are D50 and D65 the only defined standard illuminants? How can I visualize D50 and D65 on this graph; are they the intersection points? I assume the numbers around the border are wavelengths.
-->Fascinating...so D50 and D65 are CCTs?
No. They are exact colors that fall on that CIE chromatisity diagram. If you look at this,
you'll see the CCT lines run up and down (magenta to green). So 6500K can be anywhere
on that line! That's why it's better to use CCT. D65 is an exact spot (location) on the graph.
-->It appears that a range of CCTs (the lines) map to the same color temperature.
Look at one kelvin value on one line. Your color temp meter can tell you 6500K and the
color could be anywhere on that line. That's the variation. That's why I say 6500K is a
range of colors, D65 is an exact color.
-->In general, is Dxx the CCT of xx00K, or are D50 and D65 the only defined standard
D50 has a correlated color temperature of 5000K, D65 has a CCT of 6500K. The difference
again is the D illuminants are exact colors as defined spectrally, not a range of colors. This
is why when you are asked to calibrate a display and offered either a CCT or D illuminant,
always pick the D illuminant. There's no ambiguity in the target value although this doesn't
mean the display produces D50 (that's simply not possible).
-->How can I visualize D50 and D65 on this graph; are they the intersection points? I
assume the numbers around the border are wavelengths.
No, they can be exactly plotted. Examine this plot. It's from the Sony Artisan software and
shows up close where the D Illuminants fall on the CIE chart in reference to the black body
Andrew: In your pretty graph of CCTs, what are the X and Y axes..?
John: 5000K is a very warm daylight. If you set your monitor's white to 5000K it will look quite yellowish to most people. I believe most people who actually bother to set it, set the temperature to 6500K.
As an aside, I believe 5000K is the "temperature" of pure sunlight without including any light coming from the blue sky -- e.g. if you have a ray of light coming through a hole into a building. Standing outside on a cloudless day you should get a reading of 5500K or so.
As to 6500K being "unnatural" and "not experienced except perhaps at Himalayan altitudes" let me point out that this is the color temperature of outside light on an overcast day. Not even a day with heavy clouds (which will give you 7000K or even higher). As such it's perfectly suitable for examining photographs...
-->Andrew: In your pretty graph of CCTs, what are the X and Y axes..?
The core color model is called CIE XYZ (1931). This is the color model from which all other
device-independent color models are created. Like the RGB color model with three additive
primaries, CIE XYZ uses three spectrally defined imaginary primaries: X, Y, and Z. These X,
Y, and Z primaries may be combined to describe all colors visible to the standard observer
(us humans based on science done by the CIE in 1931). From that came a synthetic space
called CIE xyY which itself is derived from CIE XYZ. This allows us to plot colors 2
dimensionally on a graph which is seen in the first illustration I posted. You'll notice all
around the horseshoe shape are the actual wavelengths of visible light (from about 440
-700nm). So we are simply plotting a 3 dimensional color space in two dimensions which
are x and y.
-->As an aside, I believe 5000K is the "temperature" of pure sunlight without including
any light coming from the blue sky --
Again, try to stay away from the use of K without specifying it's a correlated value. D50 is
the exact color of daylight spectrally as measured by the CIE when they originally defined
the standard illuminants. This is the exact color (of course, depending on when and where
you measure the color, you'll get some differences in the spectral measurement).
As I said, the closest object we know of that can produce a true color temperature like the
blackbody is our sun. But that light has to go through our atmosphere and thus the color
is changed. And in reality, even our sun doesn't exactly behave like a blackbody unless
we're looking at a black hole (which could be very dangerous). A blackbody emits nor
absorbs light until it begins to be heated. Clearly a very theoretical, non-real object.
-->e.g. if you have a ray of light coming through a hole into a building. Standing outside
on a cloudless day you should get a reading of 5500K or so.
Yes but again, that 5000K reading could be on either side (magenta or green) on the CIE
chart from the true blackbody curve. We need to stop using the term color temperature in
a way that we assume is describing a specific, exact color. If you say CCT of 5000K, you're
making it clear that the actual spectral color falls anywhere within the line seen in my
illustration. When you define a true illuminant like D50, you're describing an exact color;
no range, no ambiguity.
If you had say an Eye-One Spectrophotometer, you could go outside and measure the
actual spectral makeup of the light. You can get the CCT or you can show the exact
spectral curve that is making up the wavelength of that light. A color temp meter can't do
this, it's giving you a CCT which as you can see, is somewhat over the map.
I should make clear that on this CIE Chromatisity diagram you're seeing just the xy
chromatisity values (just colors, not luminance; the 3rd dimension). So in CIE xyY, the big
Y is describing luminance and the xy describe the "color". This is how we can plot these 2
dimensionally on this graph (and the sole reason for this color model based on CIE XYZ).
The x and y values used to plot a color on this diagram are referred to as the chromaticity
coordinates or sometimes, chromaticity values. The most saturated colors (the pure
spectral hues of the rainbow) are plotted around the edge of this horseshoe-shaped
diagram. The colors become progressively less saturated as you move toward the center of
the diagram. We are only missing the luminance here (but heck, we're using 2
By plotting the three primary colors of a color space (let's say those in sRGB) we can see all
the colors that make up that space and thus it's color gamut. Plot any three primary colors
in xy and all colors that fall within are the color gamut of this color space. We see this
within the context of human vision (the entire shape). Pretty cool!
KAA, you're right about 5000 being warm(er) (6500 being cool(er), but you're wrong about "most people" unless you mean "most people who use Macs and are dependant upon Spyders and the like." What I've heard is that PC people tend to use 5000, which is closer to the 5200 Macbeth commonly provided for photolabs.
I don't agree about the "yellowish" assertion either...that perception may to do with aging CRT monitors, overly warm ambient light (my workspace is lit at approx 5000K), or due to comparison to the much more blue 6500, which is of course an irrelevant comparison: yes, a 6500 monitor is bluer than a 5000 monitor, or a 5000 monitor is yellower than a 6500 monitor...
...what mostly matters is to use the same sort of color to illuminate prints when making comparisons ...and neither matters if it doesn't approximate the ultimate display space, if display's the intention.
I don't think 5000 looks inherently yellow in a properly set large LCD of moderate quality. It may well look yellow in aging CRTs, especially if contrast is set low.
Much of the "look" of any monitor has to do with its ambient lighting and with its physical surface. The appearances of monitors are inherently more affected by ambient light than others. Another factor.
When viewing prints for photolab purposes the standard has always been 5200K. The 6500 number relates most specifically to lithography (printers ink) and, I think, Macs.
The fact is that if we're making prints we're probably mostly making them for viewing under tungsten or mixed lighting: tungsten plus daylight or daylight plus fluorescent office lighting etc.
I've been talking about printmaking, not the circular, self-referential, digital version of reality
One reason why a CCT of 5000K on an LCD doesn't look so dim and yellow is those units can
produce significantly higher luminace levels. Out of the box, you might get a CRT to produce
100 cd/m2 but you're not going to get more than a year or two out of the unit. An LCD can
easily produce 130+ cd/m2 (some will happily produce 200). When you attempt to drive a
CRT to a CCT of 5000K, you have to greatly reduce the green and blue electronics (max out
->Andrew Rodney: I would like to thank you for your easy to understand explanation. I am currently working on my diploma thesis and I wanted a brief chapter about color management. I found out that many internet sources are not true and I ended up reading about color science.
So, thanks for your insight.
Separate names with a comma.