On Mixing Colors of Light

[pinball1970]

Metamer is how I learned to spell it
Wikipedia also does a pretty good explanation (graphs and all!)

And metamerism in the phenomena. Trying to explain the difference between colour constancy and metamerism has been something of a feature at work for a while.

 

[pinball1970]

Metamer is how I learned to spell it
Wikipedia also does a pretty good explanation (graphs and all!)

And metamerism in the phenomena. Trying to explain the difference between colour constancy and metamerism has been something of a feature at work for a while.

 

[sophiecentaur]

colour constancy

Never heard of of that one. I guess it means how much difference in the RGB values is noticeable(?). I remember the JND (just noticeable difference) which is different in various regions of the CIE diagram.

 

[pinball1970]

Never heard of of that one. I guess it means how much difference in the RGB values is noticeable(?). I remember the JND (just noticeable difference) which is different in various regions of the CIE diagram.

We use it for assessment of samples under different lights. Clients have different requirements and point of sale.
In short if they want a particular colour, they want that colour not to be different depending on where that setting is.
It presents a lot of challenges, as an example, say a car has interior grey, upholstery, seat belt and an exterior that matches.
Requirement is the car looks that colour in the show room and when it out in daylight.
The showroom will have a particular SPD and daylight has a different one.

 

[sophiecentaur]

Paints and pigments are a real pain. TV engineering is a piece of cake in comparison.

 

[pinball1970]

Paints and pigments are a real pain. TV engineering is a piece of cake in comparison.

Yes, three different sets of colourants for metal, leather and upholstery.
My old lab dealt with the interior side in the 90s so metamerism and colour constancy was a major part.

Life is simpler these days, but I miss it!

 

[AlexB23]

Don't worry about your optics. Colourimetry is way outside mainstream optics; it's mainly to do with how the sensors in your eye work and how the brain recognises the result of mixtures of wavelengths and maps it onto 'colour space'.

When you say "average out" that would only be right for 'equal' levels of the two primaries you have selected. Controlling the levels of those two independently can produce a colour that looks the same as (a metemer) an approximate spectral colour. That colour will lie on a straight line between the two primariesUse the formulae in some of the posts higher up this post. Of course, the really interesting bit is how three primaries, in the appropriate relative levels can produce a 'match' of any colour within the triangle (nowhere near the spectral curve).

Imo it is risky to use the words 'colour' and wavelength together without being very aware that they are actually different things. Colour is what your brain makes of a range of different colours which only 'might' be monochromatic (spectral lines) Most of the colours we see are not from monochromatic sources. It is sad that so many 'experts' misuse the two terms. How can anyone be expected to get understanding of this topic in this desert of colourimetric misinformation?

Fascinating stuff. I do agree, as colorimetric stuff is not optics. Yeah, a lot of people including myself conflate color with wavelength or frequency, when color is just how we perceive these light waves.

 

[blue raven]

What does that mean? As you move between areas on the CIE chart, why shouldn't more than one value change?

Did you notice that there are very few objects in a scene with one of RGB being near zero? Saturated colours are rare in everyday scenes.

In the real world, when we made a change to one color, actually the other colors are affected as well by this change, but not in the same manner. Imagine we have sliders (values) for each color, and when you move one slider, the other sliders (values) will also start to move/change. So in the real world it is impossible to change just one color. For example, if you want to make a color redder, by adding red the green will be strongly affected, the violet will be a bit affected, the only one that won't be affected is the black, which has a kind of color inertia. When you look at such color charts, if you think you can move on one axis, in reality you can't, and you will move on two or more axes, creating a bit of a mess.

@Charles Link
I think the CIE graph has one big hole, namely the lack of black (or perhaps "darkness"? ) that appears in CIE Lab space as negative L, which reduces virtually any color, see this relatively accurate representation (others are not so good)

Why black (color) matters? Because using white and black you can produce 3D effects (and this might be quite interesting for physics, even thinking about black holes) and without black it does not work very well.
So many posts, and you haven't even scratched the surface of this interesting topic. You should be bolder.

 

[Charles Link]

@blue raven The Physics Forums rules only allow for mainstream physics, and do not allow for any personal theories where we are creating things that are outside of the realm of what has been established as credible science.

 

[sophiecentaur]

In the real world, when we made a change to one color, actually the other colors are affected as well

I think you need to re-think this. You would need to specify which part of the 'real world' you are discussing. Are you talking in terms of Analysis or Synthesis?

Synthesis: It's easy to vary the R primary signal to the display without changing the G and B primary voltages and you would get a result that lies along a line between the white point and one side of the triangle.

Analysis: It would be very difficult to change the pigment / paint / filter to change just one of the tristimulus values.

namely the lack of black

Your worry is misplaced. The whole point of the CIE (chrominance)chart is that it is independent of the Luminance. I mentioned earlier that 'skin tones' are virtually identical for pale skinned Nordics and very dark skinned Africans. The RGB signals have more or less the same ratios for both skins. You can have a lot of fun with a colour picker, writing down the RGB values for various faces in a crowd.

If you look at the old PAL / NTSC coding you see the levels of luminance , varying over each tv line andthe phase carrying a high frequency chrominance subcarrier signal which varies in amplitude (saturation) and

'hue (the phase of the subcarrier). This colour bar signal uses very saturated bars.

 

[blue raven]

@blue raven The Physics Forums rules only allow for mainstream physics, and do not allow for any personal theories where we are creating things that are outside of the realm of what has been established as credible science.

I don't recall such a topic "Mixing light colors" in theoretical physics, but in 2014 the Nobel Prize in Physics was awarded for the invention of the blue LED, and there you might find some related topics that can be explored without being accused of drifting apart from mainstream physics, such as how a photon is produced, the energy of a photon in quantum mechanics, valence, conduction and gap bands, see this video


You might find also interesting the topic about fluorescent and phosphorescent colors
https://en.wikipedia.org/wiki/Luminous_paint

 

[sophiecentaur]

I don't recall such a topic "Mixing light colors" in theoretical physics

Colourimetry is not 'theoretical' it's to do with very much applied Physics.

accused of drifting apart from mainstream physics

Who would be 'accusing' anyone? PF has a foot in all fields.

 

[hutchphd]

The "physics" of colorimetry is rather straightforward. The eyes report the "color" based upon the integrated intensity from several differently sensitive chromophors in the eye. (Note that this depends upon both the illuminant and the object under study, thus allowing metamerism) The assignment of color to an object is not therefore unique. In addition the reduction of a spectrum to a handfull of data is very "lossy". These are "known unknowns".
In addition there are a host of "physiologic" unknowns that have to do with processing this data into what we call "color". Many of these are "unknown unknowns".( https://en.wikipedia.org/wiki/There_are_unknown_unknowns)
But it represents a relatively arcane corner of physics (which paid my bills for quite a span)

 

[blue raven]

Synthesis: It's easy to vary the R primary signal to the display without changing the G and B primary voltages and you would get a result that lies along a line between the white point and one side of the triangle.

Analysis: It would be very difficult to change the pigment / paint / filter to change just one of the tristimulus values.

If you look at the old PAL / NTSC coding you see the levels of luminance , varying over each tv line andthe phase carrying a high frequency chrominance subcarrier signal which varies in amplitude (saturation) and

'hue (the phase of the subcarrier). This colour bar signal uses very saturated bars.View attachment 355873View attachment 355874

And have you wondered how accurate these colors were reproduced on the screen? Back in time, it was common knowledge that screens didn't accurately reproduce colors to match a quality print. There was quite a difference between the image on the screen and the print.
Any change at the input must be measured at the output, using a device such as a spectrophotometer like this one, X-rite i1 or cheaper ones. Nowadays, good monitors can be calibrated in the range of an acceptable error Delta E, where this error is given by the deviations from the target color on the color map/space, see what Delta E is. A calibrated monitor comes with a report that shows the values of this Delta E error for each color, if I remember correctly. So the challenge is to keep all errors below Delta E<1.5, this error is good, in this case you can be sure that all colors displayed on the screen will match a quality print, which also can be verified using a spectrophotometer.

 

[pinball1970]

And have you wondered how accurate these colors were reproduced on the screen? Back in time, it was common knowledge that screens didn't accurately reproduce colors to match a quality print. There was quite a difference between the image on the screen and the print.
Any change at the input must be measured at the output, using a device such as a spectrophotometer like this one, X-rite i1 or cheaper ones. Nowadays, good monitors can be calibrated in the range of an acceptable error Delta E, where this error is given by the deviations from the target color on the color map/space, see what Delta E is. A calibrated monitor comes with a report that shows the values of this Delta E error for each color, if I remember correctly. So the challenge is to keep all errors below Delta E<1.5, this error is good, in this case you can be sure that all colors displayed on the screen will match a quality print, which also can be verified using a spectrophotometer.

1.5 is huge EDIT: Big

 

[Charles Link]

@blue raven The video in post 191 is interesting, but there is a big step I am missing that I mentioned also in post 176=how do we go from having just one of these blue LED's to suddenly being able to manufacture millions of them on a computer screen? There seems to be some way of copying things atom by atom that was never taught in the classroom to students of my generation. That is in a way getting off topic, but at the same time it is relevant to this topic and would bring it up to date to how things are presently being done.

 

[sophiecentaur]

Back in time, it was common knowledge that screens didn't accurately reproduce colors to match a quality print.

It's an on-going problem I think. A "quality " print is a bit of an oxymoron. A printed image of a familiar object (or have the object and print side by side) will often be a disappointment and that's even when not through a TV. I have used the word "illuminant" frequently in this thread and it's essential to include the lighting in reliable colour matching. Most quality prints on people's walls have to take what they get in the way of room lighting.
Colour films seen in the cinema had dreadful colour fidelity. But they used to get away with it because cinema lights were switched off for viewing so your eye had nothing to judge the colour against except the 'integrates to grey' of the whole scene. And people were a lot less fussy than when sitting in their home with some lights on, watching TV. Nowadays, the owner of a $1k tv will never criticise what they paid that money for.

 

[sophiecentaur]

That is in a way getting off topic,

It certainly is. lol. Why not start a new thread? I can see you have already found a Google page for information.

 

[sophiecentaur]

all colors displayed on the screen will match a quality print,

You are comparing apples with oranges here. A "quality print" will have the benefit of human intervention and a choice of many pigments; it's a one off. A TV / camera system has to deal with everything that's thrown at it, once the illuminant has been added into the mix.

However many times it's mentioned that Colour is not Wavelength there seems to be a background routine in people's minds that rejects that idea. Primaries - whether pigments or phosphors- are colours; each manufacturer can use any spectra to produce a given 'colour' of primary. Two layers of 'primary' filters can produce black (if there is no overlap in responses) or something like the primary colour that sits between them. Colour wheel diagrams seem to ignore this entirely. Printed results can be exquisite but it all a depends on skill and experience of the expert printer (person).

The rot starts to set in at Primary School in Art classes. But. hell, teacher training can't have a colourimetry class squeezed into the teaching course curiculum.

 

[.Scott]

It's always risky to pipe in this late in a discussion. But I think that there are two items that have not been covered.
First, when color/intensity comparisons are done, they are most precise when the two samples adjoin each other. So, in the original experiment, I suggest that the both the pure tone color and the two-tone metamer be projected side by side with a hard transition between them - and with mechanisms to adjust the brightness of all three tones so that a very close match can be perceived. It would then be interesting to make comparisons among several observers - perhaps we would find "metamer twins".

The other issue is relates to color perception. It isn't just the "pixel readout" of the rods and cones that makes the color. Edges detected in the visual field by the brain are categorized as either "illumination edges" or "reflection edges". Illumination edges are caused by changes in those "pixel values" across the visual field due to differences in the lighting and orientation of the subjects being viewed. So, for example, if you are looking at a red car on a sunny day, the actual "pixel values" across the car will be different depending on whether they are facing the sun, directly reflecting the suns image, or in a shaded spot. But, as long as we recognize that all of those changes are due to illumination, we will see only a single color.

A better example is looking at the same object in ambient lighting with different color tints. To the extent possible, the brain will attempt to compensate. Of course, there are limits. If you look at a parking lot illuminated only by Sodium or Xenon lighting, you'll likely notice that some car colors are missing from the lot.

There were a lot of these kind of experiments in the 1980s. (Sorry for the lack of citations. I don't have access to that kind of library here). One experiment is to look through a tube (like a cardboard paper towel tube) at one spot in the visual field - so that the brain doesn't have enough context to recognize illumination factors. So, for example, looking through a blue-tinted bottle without the tube allows you to correctly ID colors. But looking through the tube at the same spot forces you into "pixel" mode.

 

[sophiecentaur]

so that the brain doesn't have enough context

Absolutely. That's why people were happy with Technicolor at the time in a darkened cinema. Watching TV in a room with the Sun streaming in is a much more stringent requirement.

People should make a point of reading your post. Your points are well made but not everyone is aware. I have to respect the methods used to produce those original colour response curves in humans.

 

[Charles Link]

Very interesting for the previous two inputs. Most of my inputs have probably been at the 101 level, but the subject was really new to me when I wrote the OP. It was only after that, when the CIE chart was posted, that I discovered from a google that the chart has a 3rd coordinate for the blue response of $z=1-x-y$. I do think the subject of mixing colors of light and our perception, etc. is rather specialized with even the regular ones who are on Physics Forums. There might be a half dozen who know this subject at the 400+ level, and then many who might not know it even at the 101 level.

One additional comment I would like to make is that others have said I put too much emphasis on the pure spectral single wavelength types of color. The CIE chart is good IMO, but the spectral sources get what I think is somewhat underrepresented when they only get a line of zero area on the perimeter, leaving the whole chart with its tongue-shaped area to represent all the possible colors. The colors in any case are all made up of light from these pure spectral colors of individual wavelengths. Without these contributors, there wouldn't be any light.

It would be interesting if we had tunable type monochromatic LED's, so that the picture we see on a TV screen could be made from many monochromatic sources all on the perimeter of the CIE chart, rather than the broadband LED's (whose components also come from the perimeter) whose color coordinates are then a point near the perimeter, with a color picture typically being made from 3 of them=red, green, and blue.

 

[sophiecentaur]

It would be interesting if we had tunable type monochromatic LED's,

What would be the point? I already said that monochromatic sources (by definition) are very inefficient so why bother to try to develop a system with (how many??) sources? How many sources with different output spectrum would you propose? The layout of standard style display is pretty jam packed with LEDs so how would get a reliably bright picture for any wanted display colour.

whose components also come from the perimeter

Ye gods, what does that mean? The primaries are at the vertices of the triangle.
Could you estimate the sort of range a"tunable monochromatic source" would have. Could you fit enough on a high definition screen. Thing is that there is a very adequate gamut of possible display colours using wide band primaries so what would be the advantage of your proposed unobtainatrons?

 

[sophiecentaur]

many monochromatic sources all on the perimeter of the CIE chart,

Before you get carried away with spectral colour (again) then just look at the CIE colour chart below. The gamut of the usual primaries misses out nearly half of the perceivable colours. The reason for that is that all those cyans/greens/blues are not needed, i.e. not frequently seen in critical situations etc. They are more uniform than what the triangle includes and perceptual differences are low. Keeping the Primary Triangle the way it is has a noise advantage in the B and G channels because the range of signal levels only contains the more relevant xy values. Monochrome / spectral components between 480 and 540nm could be produced but the comittee decided that a well engineered system would not need them. BTW equal spaces between colour points on the chart do not represent equal perceptual differences - which is another way of saying what I mean.

 

[Charles Link]

@sophiecentaur Whether my explanations and interpretations are the best or not is subject to debate, but you did post the CIE color map a couple of times, and I did find that to be a very good thing, especially when I learned that the chart has what is a hidden coordinate, being actually a 3 dimensional map with the blue response of $z=1-x-y$. In that sense, I think the thread served it's purpose whether you would rate my posts with a $B+$ or a $C-$ or less.

 

[sophiecentaur]

@sophiecentaur Whether my explanations and interpretations are the best or not is subject to debate, but you did post the CIE color map a couple of times, and I did find that to be a very good thing, especially when I learned that the chart has what is a hidden coordinate, being actually a 3 dimensional map with the blue response of $z=1-x-y$. In that sense, I think the thread served it's purpose whether you would rate my posts with a $B+$ or a $C-$ or less.

You just introduced a fresh idea into the thread. You should be prepared to justify it and argue where you think my criticism is wrong. Your running critique of the thread is not actually appropriate.

 

[Charles Link]

You just introduced a fresh idea into the thread. You should be prepared to justify it and argue where you think my criticism is wrong. Your running critique of the thread is not actually appropriate.

If you are referring to post 202, it isn't worth getting excited about it. I'm out of new ideas at the moment, and what I presented in post 202 wasn't meant to be taken too seriously. Meanwhile it's cold here in Chicago, but at least I have a Starbucks a block away. Cheers. :)

 

[sophiecentaur]

it isn't worth getting excited about it.

It was hardly worth your posting it if you didn't want to discuss it. Non-sequiturs can damage a perfectly good thread.