color temperature

[Jim Nelson]

hi,

 

can someone please help me out?

 

I know that 3200K is orange and 5600K is blue. And some of my friends told me that in between 3200K and 5600K you have yellow then green. However, I saw on the internet so many Kelvin scales that say that in between 3200K and 5600K you have white.

 

In Blaine Brown's Cinematography Theory and Practice book it says green. But in most of the color temperature charts on the internet it says white. So I'm really confused.

 

So what color exactly is in between 3200K and 5600K?

 

 

Thanks for your help

[Matthew Parnell]

Kelvins are used to effectively measure the colour temperature of white light.

 

While difficult to explain, the easiest way to think about it is to imagine colors on a compass. North is Blue, South is Red, East is Magenta, West is Green. Along the north-south axis is a band of white formed by the convergence of all the colours to white. Kelvins are measured on the north south axis, so from red to blue through white. Colour obviously isnt this linear, but that gives you the general idea.

 

Wiki will answer you question in more detail: http://en.wikipedia.org/wiki/Color_temperature

[Jim Nelson]

hi,

 

I checked wikipedia but it still hasn't answered my question unfortunately :(

[Jim Nelson]

hi I just don't understand why you have white on the kelvin scale while in Blain Brown's book and also on color wheels you have green. And + white isn't a color.

[David Mullen ASC]

A color temp isn't "white" -- "white" is only relative to what color temp the film stock is balanced for. If the film stock is balanced for 3200K, then 3200K is "white" and if balanced for 5500K, then that's "white" to the film stock.

 

It also only describes a full spectrum source that contains all the visible colors.

 

Something with a partial spectrum or a strong green spike is not described by color temp.

 

In other words, you can have a 3200K or 5500K flourescent bulb that has a green spike in it.

 

3200K has more redder wavelengths and 5500K has more bluer wavelengths and the color temps in between are just in between those two ends. So on 5500K balanced stock, a 4300K light would look warm, and on 3200K balanced stock, 4300K would look bluer.

 

The greenishness of a light has to be measured with a color meter that usually gives you an answer in terms of how green or magenta the light has a bias towards.

 

In this case, you aren't measuring color temp so much as you are the balance of RGB or YCM relative to what you consider "white".

[Tom Jensen]

Also you can look at it this way. You have two stocks, Tungsten and Daylight. If you take Tungsten stock out of doors and use it, it will have a bluish tint to it. This is why we put an orange filter (85) in front of the lens. You wouldn't normally use Daylight balanced stock indoors and use them under tungsten lighting conditions because the film will come out yellow. You wouldn't want to put a blue filter in front of the lens because you would lose light and the image would be harder to see due to the dark filter and you are decreasing the amount of light that hits the film in an already dark situation. There are TONS of books on the subject. It's best that you read one of those.

[Jim Hyslop]

And + white isn't a color.

That is one of those well-known "facts" which have no basis in truth. White and black are both colours. Colour is how we perceive the effects of the visible portion of the electromagnetic spectrum, and in particular the three primary colours of light, red, green and blue. White is a mixture of all three colours. Black is the absence of all three.

 

Another definition I like is that light is the property by which we can distinguish two objects which are otherwise identical. Consider a set of snooker balls. They are all identical (without resorting to high-precision measuring tools), except for their colour.

[Chris Keth]

You're trying to make color temperature fit a normal color spectrum or color wheel. Those happen when we mix pure colors of light together. We're not talking about pure colors of light nor are we talking about mixing. We're talking about spectra of light that are biased toward orange and biased toward blue. The scale is created by mapping those spectra with the temperature black-body radiator that created them.

 

There never really is just blue light and just orange light. There is "more blue and less orange" and "less blue and more orange."

[Jim Nelson]

I'm sorry but I'm still a bit confused.

 

I understand that you get white by mixing the colors. And I understand that when you use tungsten film then light that has a color temperature of 3200K will be white. But I don't understand why most kelvin scales have white in between 3200K and 5600K?

 

http://www.mediacollege.com/lighting/colou...emperature.html

 

And some kelvin scales and all color charts have green in between.

 

http://www.colourenergy.com/img/misc/kelvin.jpg

 

http://www.gofishsc.com/Articles/images/Pe...r_wheel.svg.png

[David Mullen ASC]

It's because you can't draw a bar chart like that, that goes from orange at one end to blue at the other... without making up a transitional color like green or white in between. It has nothing to do with the light itself, it's just an illustration problem in using a chart like that. Just ignore it.

 

In color temperature, the higher Kelvin numbers are bluer than the lower Kelvin numbers, that's all you need to understand. There is no "green" or "white" in between. You're being too literal.

 

If your eyes or film stock are balanced for daylight 5500K, then that's "white" so how would you draw that chart? White at the top getting redder and redder below that?

 

What you have to understand is that a chart that goes from infrared to red thru green to blue and then ultraviolet is showing you not color temperature but the color wavelengths of visible light.

 

But a full spectrum source like the sun or a tungsten light bulb has ALL the visible wavelengths in it. It's not just blue wavelengths for 5500K lamps and red wavelengths for 3200K lamps. Both types of lights have all those wavelengths in them, it's just that they have more blue or redder wavelengths mixed in depending on their color temperature.

 

The chart that transitions to white in the middle is probably the more accurate way to think about color temp as long as you realize that it doesn't mean that "white" is 4300K or so... you can take that scale from blue to orange with white in the middle and slide it around to whatever you consider the "white point" to be. For example, if you are shooting on daylight film, then the white portion of the scale would be in the 5500K range and what's above that would be bluer and what's below that would be more orange.

[Chris Keth]

I'm sorry but I'm still a bit confused.

 

I understand that you get white by mixing the colors. And I understand that when you use tungsten film then light that has a color temperature of 3200K will be white. But I don't understand why most kelvin scales have white in between 3200K and 5600K?

 

http://www.mediacollege.com/lighting/colou...emperature.html

 

And some kelvin scales and all color charts have green in between.

 

http://www.colourenergy.com/img/misc/kelvin.jpg

 

http://www.gofishsc.com/Articles/images/Pe...r_wheel.svg.png

 

Simply put: because those diagrams are wrong.

 

"White" in film is only as good as what we tell the camera is white through white-balancing, filtration, or a combination of the two. With a video camera or an extensive set of filters, we could make the camera record 10,000K light as white or 1000K as white and who is to argue? It's on film as white light.

 

Green shouldn't be considered as part of color temperature. It can be added-to or subtracted-from any color temperature of light. Therefore we should represent the hue of light in 2 dimensions, rather than just along one line. On one axis there is color temperature and on the other axis we have green/magenta balance. They cross at the point where we have defined white to be (for repetition sake, there is no inherent white light), and where the image tint is neither green nor magenta.

 

 

Here's a quick thing I threw together for illustration. I intentionally left numbers off of the chart because those can change, depending on the filmstock/white-balance, and filtration. One could tweak the position and opacity of those gradient layers so that the center is truly neutral grey. I didn't take the time.

 

Edited April 13, 2009 by Chris Keth

[jeff woods]

And some kelvin scales and all color charts have green in between...http://www.gofishsc.com/Articles/images/Pe...r_wheel.svg.png

 

All that everyone has said, plus c o m p l e t e l y ignore that third one; it has no bearing on what you are asking.

 

And it's hosted by a fishing website. That made my brain hurt...

 

-j

Edited April 13, 2009 by jeff woods

[Chris Keth]

All that everyone has said, plus c o m p l e t e l y ignore that third one; it has no bearing on what you are asking.

 

And it's hosted by a fishing website. That made my brain hurt...

 

-j

 

It's nice for painting but it has no bearing on the physics of how light is produced by a radiating element.

[Dominic Case]

You are confusing colour with colour temperature.

 

Colour temperature is - simply - the temperature something is heated to when it glows a particular colour. At around 1800K a poker in the fire is glowing "red-hot". It is radiating a lot of heat, some infra-red, and a little red light. (All forms of electromagnetic radiation). As it gets hotter, some of the shortewr wavelengths of visible light get added into the mix, giving an orange then a yellow glow. At around 5000K all wavelengths of visible light are there so the object is "white hot", and if the object gets hotter still, there are more bluish wavelngths than red ones.

 

This chart shows the distribution of different wavelengths (colours) at different colour temperatures.. Don't worry about the formulae.

 

To put your question in a different way - did you ever hear of something being "green-hot"?

 

Green light has wavelengths in the 500-550nm range (reds are in the high 600s, blues in the low 400s). Incandescent light (the sort that is described by colour temperature) never has more light in the green range than at the red or blue ends of the scale.

 

The filament in a tugsten lamp is actually as hot as its colour temperature. When it comes to sunlight it's a bit more complicated., The surface of the sun is around 18,000K (blue hot) but the earth's atmosphere scatters quite a lot of the bluish wavelengths making it appear much cooler than it really is.

[Chris Pritzlaff]

is there a scale for the green/magenta shift? I know there has to be some measurable scale but am not sure what it is called

[Chris Keth]

is there a scale for the green/magenta shift? I know there has to be some measurable scale but am not sure what it is called

 

Not as far as I know, but if anyone knows this is the place. When I've measured it, it was with a minolta color meter which gives a readout in unitless numbers which you cross reference to a table which gives you correction in strength of + or - green.

[Richard Andrewski]

is there a scale for the green/magenta shift? I know there has to be some measurable scale but am not sure what it is called

 

Yes there is. In practical terms its color rendering index or CRI for short. We know that the CRI of those things closest to the theoretical black body radiating light (sun or real tungsten light) are pretty darn close to 100 CRI or said another way, they are balanced on the magenta green axis.

 

Its when we get to electronic lighting or things measured with a correlated color temperature (CCT) that we are mimicking the black body and the farther away from that ideal, the less the source is full spectrum and also the more the CRI is potentially lower. The more green, the more color rendering is affected and the lower the CRI. But, you can have a 5600K CRI 90 light and a 5600K CRI 70 light. The CRI 70 will have a noticeable green spike, the CRI 90 shouldn't have. They are both 5600K though and the absence or presence of green has nothing to do with the color temperature.

 

Its no accident that most electronic lighting (Fluorescent, LED and HMI) all have a green spike. Its because our eyes are most sensitive to green and thus when we're trying to make a super efficient light source, why not use a predominance of green to help out with the efficiency. Its true in all of those electronic types including LEDs.

 

So, in these electronic sources, we have the potential of a green spike and the more you have of that, the lower the CRI. When we need to correct for this green spike, we counteract it with the color opposite to green on the color wheel--magenta. This can be done in the form of a correction filter, an algorithm in white balance for digital mediums, and even by adding more magenta to the phosphor of an LED or flo bulb.

 

Thus you have the magenta / green axis someone spoke of which indicates the balance of the electronic light and also indirectly its CRI.

 

On the issue of whether its white in the scale or not for color temperature, it is indeed white. Color temperature does not describe colors but rather describes the shift in white between red and blue. So you have "warm white" which is more toward the 3000k (red) end, neutral white (4100K or so in the middle area) and "daylight white" which is more toward the 6500k (blue) end and higher and lower of course. Those were commercial terms applied to those color temperature ranges for the sake of simplicity in commercial bulb descriptions. A lot of people refer to the red end as orange but its really red if you want to get technical.

 

Again, any green present in the white is simply an artifact of the way light is produced in that light and is measured by the best RGB type color temp meters like the Kenko KCM-3100 as the two axis RED/BLUE (color temperature) and GREEN/MAGENTA (correction) or as I said its indirectly referring to CRI. To get a real CRI number you either have to guess or go to a photospectrometer and integrating sphere to give you the real number. Nothing else I know of reports CRI directly like that.

[Chris Brown]

I'm sorry but I'm still a bit confused.

I'll add to the pile.

 

The color of light is measured along two axes: the yellow/blue axis and the green/magenta axis.

 

The yellow/blue axis is measured in degrees Kelvin (˚K). This unit of measurement is used because Lord Kelvin determined, among other things, that as a black object (Carbon in his experiments) was heated, it exhibited a warm, red glow at high temperatures (about 2000˚K). As the object became hotter, it glowed more yellow. Even hotter and it started to exhibit a bluish tint at about 5000˚K. Even hotter and it became even more cool/blue. So, film & camera makers decided years ago to balance films to certain "color temperatures". This would allow the film to record neutral colors when used in the proper lighting environment (e.g., indoors with tungsten lights, or outside under the sun).

 

This yellow/blue axis is sometimes referred to as a "warm/cool" axis, and judicial use of CTO and CTB gels can give the illusion of an indoor or outdoor scene, or can help in setting a tone for a certain time of day.

 

The green/magenta axis was incorporated by color scientists (and film & camera makers) to account for the color shift of light at a specific color temperature. Color scientists determined this green/magenta axis to be perpendicular to the yellow/blue axis. The green/magenta color shift is independent of any color temperature. It's usually a result of other causes such as weak or strong phosphors in a lamp, or a light manufacturers bulb coating, or debris in the atmosphere (e.g., smog), or a lens manufacturer's glass. The green/magenta axis unit of measurement is a "mired".

 

To get neutral color, a good color meter and accurate use of CTB and CTO gels can be used to great advantage when using film in achieving neutral color. When shooting digital, a neutral white board (one without optical brighteners) can be used to set the white balance in-camera. (However, additional color balancing or grading is almost always done in post).

[Dominic Case]

The green/magenta axis unit of measurement is a "mired".

This isn't how I understand Mireds.

 

If the colour temperature of a light source is quoted in Kelvins, then it can also be quoted in Mireds (short for Micro Reciprocal Degrees), being 1000000/(col temp).

 

The advantage of Mireds over Kelvins is that you can calculate the filter you need. For example, to convert 5,000K to 3,000K, you express each light source in Mireds (1000000/5000 = 200Mireds; 1000000/3000 = 333Mireds). The difference (333-200 = 133) is the Mired shift filter you need.

 

I don't think Mireds are any use for measuring the green/magenta axis, which is more to do with the extent to which a light source doesn't match an incandescent (or Planckian) light source, and is also used to measure how accurately colours match each other under different light sources. You don't want to even think about how the CRI (Colour Rendering Index) is derived. Wikipedia covers it in excruciating detail.

[Richard Andrewski]

That's right, mireds don't have anything at all to do with green/magenta measurement. My color meter gives out a LB or light balancing index and a CC or color compensating index. CRI is just derived indirectly from that by guessing how much off a balance in the green/magenta shift you really are. Or derived with no guessing in an integrating sphere and spectrographic computer setup.

[David Bradley]

As far as my understanding goes colour temperature doesn't account for green / magenta spikes. Colour temperature refers to the light emitted by a black radiator which when heated to incandescence emits a fairly consistent colour of light (emitting relatively similar amounts of energy throughout the visible spectrum (see Plancks Law)). Light emitted from sources other than black bodies (such as flourescent lamps) may feature 'spikes' of magenta or green. The spike is visible if we plot a graph showing energy on the Y axis represented by KJ/nm against visible spectrum on the X axis represented by nm. A Black body creates a smooth curve on the graph where as a flourescent lamp shows a smooth curve with a spike.

 

Flourescents have a 'correlated colour temperature' which is also read in kelvin. Correlated colour temperatures are based more on human perception in that they are measured relative to our perception of the colour of the light a black body emits when heated to incandescence.

 

3 channel colour temp meters can be used to measure the green channel to figure out how much you will need to compensate. Best bet is just to eliminate flourescent fixtures from your lighting kit - they look horrible anyway. Thankfully kino flos were invented!

 

PS has anyone had problems with their kino flos? I bring at least 2 x 3200K 4ft 4 banks when I shoot and I find that if I use a CTB to bring me up to 5500K I get a magenta cast to the light? I've started using a 1/8 plus green to compensate but I was under the impression that this wouldn't be necessary?

[David Mullen ASC]

It's so much more light efficient to add warming gels to daylight Kinos than adding blue gels to tungsten Kinos, so I've never tried to get a K32 tube up to 5000K, not when it's easier to start with a K55 tube.

 

Generally I find the opposite -- adding gel tends to make the bulb warmer, which increases the green spike. The only time I've felt Kinos looked pinkish is when using the 2900K tubes, which I find to be an odd color.

[John Sprung]

Its no accident that most electronic lighting (Fluorescent, LED and HMI) all have a green spike. Its because our eyes are most sensitive to green ....

 

No, flourescents have a green spike because they're mercury arc lamps with a phosphor coating inside. Mercury has a strong emission line at 546.1 nanometers, which is a nice middle green. Except it's not so nice when it shows up on film. Mercury has much stronger lines in the ultraviolet, which is what drives the phosphors to flouresce. The others don't have a green spike.

 

Peak sensitivity for our eyes is 550 - 555 nanometers, but it's not a steep spike.

 

********

 

The whole color temperature thing is a very long story. Objects glow if they're hot enough. Billions of years ago, stars were the available light sources, especially the nearby one we call the sun. Life evolved. Seeing turned out to be very useful for survival, so most life forms on this planet can do it. That's why we're specifically adapted to see using the 5600 K light of the sun, as filtered by the atmosphere. Particularly at the beginning and end of the day, the atmosphere gives us a lower effective color temperature.

 

After millions of years of human evolution, we developed intelligence, and along came Isaac Newton. He found out how to sort light out into pure wavelengths. Then came Max Planck. He figured out the relationship between the temperature of a glowing hot object and how much light it emits at each wavelength.

 

Color temperature only applies to light that has a Planckian distribution of energy vs. wavelength. Green doesn't have a color temperature.

 

Our eye/brain combination can adapt to use anything in a range from roughly 2600 to 6500 Kelvin as a normal "white" light source. But only one color temperature at a time. Suppose you're reading a book made of white paper lit by a 2850 Kelvin incandescent desk lamp, and suddenly through a tiny hole in the blinds, a 1/2" diameter spot of sunlight falls on the page. You're adapted to tungsten, the sunlight will look blueish to you. It won't be an extreme Microsoft Blue Screen of Death blue, just blueish. Likewise, if you were adapted to daylight, a tiny spot of equal intensity tungsten light would look red/orangeish to you.

 

Around 1000 - 1200 K, things are just hot enough to start seeing the light, and they look quite red. A typical kitchen toaster has wires in it that get at least that hot. At 60,000 K, they look quite blue to us. (Nothing ever really gets that hot, but you can still do Planck's math with that number, and get a theoretical distribution that can be produced in real light by filtering.)

 

Our cameras can also be adapted to a similar range of color temperatures. With video, it's often just a matter of filling the screen with white and pushing a button to tell it to set the white balance. With film and some digital data cameras, we hang glass filters to get in the right ballpark, and make the final adjustments in post.

 

****************

 

Another very long story, but worth looking into, is CIE 1931 chromaticity. It abstracts all the color range of human vision to a two dimensional diagram, with pure wavelengths around the outside and a small curve of color temperatures in the middle.

 

 

 

 

 

-- J.S.

[Jim Nelson]

thank you very very much everyone. Things got much more clear :)

[Jim Nelson]

"At around 5000K all wavelengths of visible light are there so the object is "white hot""

 

 

One more thing though, my cinematography teacher said that 3200K is orange and 5600K is blue. why do some people say that 5600K is white light?

 

Of course now I understand that when you white balance to 5600K it'll look white. But if you don't. isn't 5600K blue?

 

 

Thanks