Talk:Color temperature

Long term lack of sources

This article contains some useful information, but sadly contains a large amount of text that has had "citation needed" tags for 11 years. We really need find sources for this text or remove the text from the article (no matter how reasonable or accurate it seems). Ashmoo (talk) 13:36, 22 November 2023 (UTC)

White point = 6500K?

White point is assumed as 6500K in both scale picture and Template:Color temperature, but 6500K is a really obviously bluish light in fact. I think 5500K will be a more accurate white point. For example, my all screens are obviously less bluish than 6500K lamps but also less warm than 4000K lamps, i don't have a 5000K/5500K lamp to compare but i think they are about 5500K. Can anyone update this image and template to make the white point 5500K (or more preferably make the template allow custom white points, and even gammas)? (The lowest and highest values given must also be changed, and even the current lowest value is wrong (it does not give pure red). This discussion is more about the image and template but i think discussing it there can be make it more centralized.) RuzDD (talk) 13:18, 6 February 2024 (UTC)

Addition: Colors of File:CFBulbs.jpg looks very realistic to me with my screens, and i think most of LCD (and probably ?LED) screens have ~5500K white points like mine displays. RuzDD (talk) 13:26, 6 February 2024 (UTC)

The use of 6500K is very much intentional here. Due to color constancy, there is no "true white", and even standard illuminant E (equal energy white), which has a uniform spectrum, is not special when it comes to white points. Like the D-series standard illuminants on the daylight locus, like Illuminant D65, E is also not on the planckian locus, and is not directly defined by temperature, but only by correlated color temperature, where it is approximated by 5455K. However, most color science uses D65 (6504K) as the default illuminant, since it is generally under this illuminant that human color vision evolved (average midday light in Western Europe... so mind the eurocentricity), and therefore is expected to represent a true neutral illuminant, if such a thing can even be said to exist. You could almost say that if D65 looks blue, then you may have normalized to the color temp of (typically warmer) indoor lighting! Curran919 (talk) 20:23, 6 February 2024 (UTC)

Isn't D65 deliberately bluish because it includes illumination from the blue sky rather than being a definition of white? Where the sun alone in that condition is around 5700K? Sincerely, North8000 (talk) 20:47, 6 February 2024 (UTC)

@North8000 essentially, and when something is illuminated under daylight, that includes direct and scattered sunlight. You have to add direct sunlight and the blue sky light integrated over the sky's visible solid angle (ideally, a hemisphere) to get the daylight standard illuminant, which is estimated by D65. If you block out the scattered, indirect skylight at noon, you will get close to 5700K with direct sunlight alone (though I don't know if that changes after passing through the atmospheric window). At sunset, the direct sunlight passes through so much atmosphere, that it drops to 2000K, which is why so many people characterize the sun as being yellow instead of white, as it is (almost) by definition. That's why it's called your golden hour. Curran919 (talk) 22:43, 6 February 2024 (UTC)

I know the physics better than the related terminology and standards. The question is does daylight standard (illuminant) define "white"? But I guess the more specific question need to start by exactly defining the proposed change. Sincerely, North8000 (talk) 23:12, 6 February 2024 (UTC)

The temperature of sun is 5772K. I know this is for photosphere, and we need the temperature of it because we are seeing it, not the core of the sun. Due to rayleigh scattering, temperature of sunlight will be change for most parameters but the average temperature of the sunlight will be about 5772K. So, if we get sun as the reference, we must assume the whitepoint is about 5772K otherwise our value will not be average. RuzDD (talk) 20:56, 6 February 2024 (UTC)

I can see that (sunlight is warmer than 6500K indoor lighting even in noon) when i turn on a 6500K lamp at noon. RuzDD (talk) 20:59, 6 February 2024 (UTC)

@RuzDD CCT of ambient light changes significantly depending on latitude, season, time of day, portion of indirect to direct illumination (how much blue sky is seen) and atmospheric content. D65 is just an "average". Like most standards, it is to an extent arbitrary, but it is the standard now and that's why it's used when defining the graphics. Curran919 (talk) 22:33, 6 February 2024 (UTC)

If the average is D65, where is the warmer part of the light? If temperature of the sun is 5772K, the average must be 5772K otherwise this means sun gives warmer lights to some directions and colder lights to some directions.

When i copy the hexcode of 3000K and paste it to make my screen glow in that color under a 3000K light, the screen is looked more yellow than my white objects. When i copy the hexcode of 4000K and do the same thing, my screen and white objects looked more smilar (i also have a 4000K light source and it's obviously less yellow). So, the whitepoint of my screens are more near to 5500K than 6500K and the white of my screens are really white to me.. RuzDD (talk) 23:21, 6 February 2024 (UTC)

Let me try out an analogy that came to me as a shower thought. Trying to define "white" as a specific spectrum is like trying to define room temperature as a specific temperature. There is not one real answer, because what feels like the ideal neutral ambient temperature is going to change depending on circumstance. It can be useful to define it at 25*C to have a standard, but that doesn't mean it is a ground truth. Someone can easily come and say, "actually, I feel like room temp should be 24 or 26*C", which may feel more neutral to them given their local humidity, windchill, etc. as well as them subjectively due to their base body temperature... but if they try to write an exam by saying "I'm going to assume ambient temperature as 26*C instead of the standard ambient 25*C", its going to confused the prof and probably get marked wrong. Thats why we have standards. Not only is your perception of the white point highly dependent on all of your environmental conditions (as I explained), its also quite subjective, given the variance in color vision, such as the ratio of blue to green/red cones, which determines the blue-yellow color axis that is approximately parallel to the daylight locus near the white point. Of course, that cone ratio shouldn't matter much due to color constancy and chromatic adaptation (you haven't acknowledged these terms, which are critical to the understanding of this concept, so I don't know if I'm screaming into the void), but that's exactly the point, that white is... fluid... and we just need to agree on a standard, which the world has largely treated as defaut D65 for most color appearance models. Curran919 (talk) 15:55, 7 February 2024 (UTC)

I know those, but i i think there's a better way to define the ideal "room temperature". We find the average air temperature of the earth for all places and all times, and this will be a more good standard. Same way, i think getting the average sunlight temperature (average, in all places and all times, for example not only noon's) of the earth will give a more accurate average (and, physics says this must be same with the photosphere temperature of the sun otherwise sun must have very different photosphere temperatures in different positions). That's 5772K, and i think it will be better to round it to 5500K so the difference will be very rounded. RuzDD (talk) 16:12, 7 February 2024 (UTC)

@RuzDD well then you need to take it up with CIE, not Wikipedia! Curran919 (talk) 16:19, 7 February 2024 (UTC)

According to standard illuminant CIE already created many different whitepoints including D55. RuzDD (talk) 16:33, 7 February 2024 (UTC)

@RuzDD That article is a bit outdated. See the discussion on its talk page: Talk:Standard_illuminant#There_are_only_two_standard_illuminants:_A_and_D65 Curran919 (talk) 22:06, 7 February 2024 (UTC)

D55 is not standard but it exists (if i understood it correctly), and i think that'll not require move this discussion to CIE (i think we don't have to watch the standards those are given by CIE in all circumstances). RuzDD (talk) 23:43, 7 February 2024 (UTC)

By the way, looks like kinda accepts me because 5000K-5500K temperatures have a very flat curve. RuzDD (talk) 01:45, 8 February 2024 (UTC)

@RuzDD Maybe to reiterate how arbitrary white is and how equal energy white (based on the spectrum, related to "white" noise) is a completely separate concept from perceptual white, imagine the star Vega, which has an average photosphere temp of 9600K. The star looks blue to us now, but if we evolved on a planet around Vega, the most useful perceptual white point, that which would allow us to see the most colors on Vega would be 9600K and not equal energy white (~5405K). We'd likely evolve to have that as a "default" white point, and equal energy white would look very yellow to us. We probably wouldn't even call it equal energy "white" because by the time we invent light spectra measurement, the equal energy point would look nothing like the "white" we had been using as our term for a perceptually neutral color. Curran919 (talk) 10:48, 8 February 2024 (UTC)

To add a note, this only applies to an illuminant or self-luminous object. The definition of white for an object (surface color, in terms of reflectance spectrum) is still uniform, as you'd expect, regardless of where your eye evolved. Curran919 (talk) 11:04, 8 February 2024 (UTC)

Of course "white" is just a word with varying definitions. Maybe clear day noon sun is a good one. But I do agree that 5500 is closer to the overall norm than 6500. North8000 (talk) 15:28, 6 February 2024 (UTC)

Strike a part of my post. Maybe I'm wrong. North8000 (talk) 20:51, 6 February 2024 (UTC)

@RuzDD: I think we need to focus this discussion. To do that, could you specify exactly what your proposed change is? Sincerely, North8000 (talk) 03:39, 7 February 2024 (UTC)

Mathematical equations those used in this template and for making that gradient photo. RuzDD (talk) 12:48, 7 February 2024 (UTC)

There is a good description for that graphic you can read on its wikicommons page that explains the derivation. I see nothing wrong with it. If you still disagree, you should post about it on the color wikiproject where there are smarter people than myself when it comes to color models. Curran919 (talk) 15:44, 7 February 2024 (UTC)

One problem with this picture is that the device-dependent lightness artifacts are misleading: 6500k is shown as significantly lighter than the colors further away because that's what an sRGB display can show the brightest, when the intention of the diagram is ostensibly to show the differences in color. It might work better to put a narrower strip on a black background and then at least roughly equalize the lightness across the gradient. Even better might be to limit it to, say, 3000–9000K CCT, and just show a circular spot of the same lightness for each example, against a black background, instead of a gradient. If these are all representing "white" then it's to some extent inherently misleading to show a smooth gradient at all. –jacobolus (t) 17:31, 7 February 2024 (UTC)

@Jacobolus I think gradient is better bacause it allows users to select much more colors, and i think narrower things will be less understandable. And, i think lightness operation is not needed right now but if it should be made i think it'll be better to doing it in a new image and looking at the blackbody curve to make brightness increasing instead of staying the same (for example, 8000K is lighter than 6500K, just like the incendescent bulbs does). I think this kelvin scale (1000-12000) is good and even not enough, i don't think narrower will be better. But i think we must make different images for different gammas (1.0, 1.2, 1.4, 1.6, 1.8, 1.9, 2.0, 2.1, 2.2, 2.3, 2.4, 2.5, 2.6, etc). RuzDD (talk) 18:35, 7 February 2024 (UTC)

Addition: but i think putting colors on black will be better as you say. RuzDD (talk) 18:36, 7 February 2024 (UTC)

Any particular CCT does not involve any inherent intensity or perceived lightness. It only describes chromaticity. I have no idea what you are getting at with "different images for different gammas". The output device is always presumed to be an sRGB display with prescribed characteristics. –jacobolus (t) 18:41, 7 February 2024 (UTC)

Yes, but blackbodies involves and i think creating some grays is not a very great idea unless they are representating something (blackbody in my case). RuzDD (talk) 18:46, 7 February 2024 (UTC)

@Jacobolus the spectrum here [1] appears to be more isoluminant, if that's what you mean. I see the benefit. I wonder if turning the white to gray would lead to confusion? I do like the divergent nature of the "emphasized" white, but maybe after this whole conversation of white not being fixed, deemphasizing it would convey that idea. Curran919 (talk) 00:43, 8 February 2024 (UTC)

This photo is a lot harder to understand for me (you didn't pinged me but i want to answer anyway). But, creating a photo with increasing brightness (higher kelvin = brighter color) may be good for some circumstances (like an incendescent bulb, both brightness and color temperature increases as you force it to work harder and harder). RuzDD (talk) 00:51, 8 February 2024 (UTC)

Note: this photo is even not very color balanced for me, for example yellow is significiantly brighter than white/gray for me. RuzDD (talk) 00:52, 8 February 2024 (UTC)

Yeah, something like that, but it ideally needs to have as much black or dark gray background as possible for the gradient to really read as something like "white". –jacobolus (t) 01:08, 8 February 2024 (UTC)