Dan Margulis Applied Color Theory -- Color Temperature and Metamerism

From: Ron Bean
Date: Fri, Dec 31, 1999, 7:16 PM
RE: Color Temperature and Metamerism

I've been reading two of the books Dan recommended a while back:
"Principles of Color Reproduction" by Yule, and "Principles of Color Technology" by Billmeyer & Saltzman. They're both excellent (BTW you can skip the math and still get a lot out of them). I learned a couple of things that I haven't seen elsewhere, and I thought I'd mention them here.

Billmeyer & Saltzman's discussion of metamerism is especially good-- they talk about "invariant" vs "conditional" (metameric) matches between two colors, as a result of the spectral content of the light vs the spectral reflectance of the pigments. In the past I've heard that metamers are colors that match under 5000K lighting but not under some other light source, but they talk about it in terms of *any two light sources*.

I had gotten the idea that 5000K lighting produces more "accurate" color, but I think they would prefer to say it's more *standard* rather than more accurate. They mention the idea of a "color rendering index" as being *relative to some standard*. 5000K is more "accurate" in the sense that it's the most common standard. But you could pick some other standard, if you were designing something to be viewed under a specific light source. In fact, I recently saw a color swatch form used by a silkscreen printer that includes a space to specify which light source is to be used for each job (including "customer supplied light source").

I've seen books that implied (or even stated directly) that all light sources have a color temperature, but that's clearly not the case. Color temperature refers specifically to black body radiation, and therefore only applies to light sources that behave like black bodies at some temperature. Daylight, incandescent light, and arc lights are reasonable (though not exact) approximations of black body radiation, but most garden-variety fluorescents and gas-discharge lamps (ie, most commercial lighting) are not, unless they've been specifically engineered to behave that way.

Fluorescents can have wierd peaks and dips in their spectral graphs (in addition to the mercury lines), depending on the phosphors they use. They show the spectral power graph of a type of "5000K" fluorescent called a "prime color" lamp that has several sharp discontinuities in it, which is not what you'd see with a true 5000K light source. And they point out that this spectral content must be accounted for when making calculations to be used with this light source.

Yule also has a comment about cheap fluorescents with oddball spectral contents:

"Unfortunately, the old type of fluorescent tube is still used where merchandise is displayed, so that a printed package originally evaluated under the standard light source looks very different to the consumer. In particular, the reds appear dull and brownish. But if the package were printed to look satisfactory in fluorescent light, its color would appear exaggerated in daylight or tungsten light. It is to be hoped that the old type of fluorescent tube will eventually become obsolete, for this appears to be the only satisfactory way out of the difficulty."

That was in 1967, and they haven't gone away yet...


From: Chris Murphy,
Date: Sat, Jan 1, 2000, 2:02 PM
RE: Re: Color Temperature and Metamerism

> In the
>past I've heard that metamers are colors that match under 5000K
>lighting but not under some other light source, but they talk
>about it in terms of *any two light sources*.

Metamers are color samples that match under one kind of illuminant but not under another because their spectral properties are different. Under one kind of lighting the differences may cancel each other out, or just not be affected enough for a visible difference to occur. In other words, you can have metamers that don't match under 5000K but do match under tungsten for example.

>I had gotten the idea that 5000K lighting produces more
>"accurate" color, but I think they would prefer to say it's more
>*standard* rather than more accurate.

Yes, but I would add to this. 5000K is rather vague. The standard illuminant for graphic arts is actually D50, not 5000K. They aren't the same. D50 is an illuminant, 5000K is not. D50 has a very specific "spectral power distribution" which is a fancy way of describing what amounts of each visible wavelength that light source outputs.

Some problems with ANSI's "standard lighting source" or whatever they refer to it as, and the ICC specification is that the luminosity is not specified. The amount of light absolutely affects perception of a color.

So it's possible to specify D50 in two locations yet get different results because of luminosity.

There is also a problem with CIE colorimetry. That is, CIE is the International Commission on Illumination and among many other things, they define standard illuminants such as D50, D65, A, B, F2, F7, F8 (many, many others). The problem is that they never intended CIE colorimetry to apply to cross media applications. So D50 for a monitor isn't the same as D50 from a viewing booth on paper.

> They mention the idea of a
>"color rendering index" as being *relative to some standard*.

The standard illuminant to which the light source is attempting to simulate. What is the standard illuminant 5000K fluorescents are trying to simulate? I don't know but I find it difficult to believe it's D50 given the CRI numbers they advertise. Daylight fluorescents (such as Phillips TL-90 and GE's Chroma 50) both are more similar to the F8 illuminant than they are D50. They really aren't the best solution for standard lighting (D50 in the U.S.), and in some application (such as photography) they can be problematic.

>5000K is more "accurate" in the sense that it's the most common
>standard. But you could pick some other standard, if you were
>designing something to be viewed under a specific light source.

Sure. One of the benefits of spectral measurement data only profiles (i.e. the profile contains only measured data, not some profile manufacturer's interpretation of them) is that the appearance of specific RGB, CMYK (or whatever primary system is being used) file can be predicted under any of the standard illuminants (or for which a custom illuminants spectral power distribution has been defined).

So for Women's Day magazine, the publisher's marketing people might conclude most readers of the magazine view under warm halogen lighting. They could then create separations that account for that lighting condition. Readers now see lipstick ads with closer matches to the actual color of the product. This can be a big deal.

A large format printer might decide to make separations to compensate for use of metal halide or mercury vapor light sources used in trade shows. It is a common complaint in the trade show industry that what they saw at the printer to approve the booth signs looks different in the trade room. Too late once it gets there.

> but most
>garden-variety fluorescents and gas-discharge lamps (ie, most
>commercial lighting) are not, unless they've been specifically
>engineered to behave that way.

Fluorescent bulbs contain mercury. Excited mercury is how the phosphor coating fluoresces and emits light - it's the actual inner surface coating where the bulk of the light comes from. However, the mercury itself gives fluorescent bulbs very spiky distribution; that is, fluorescent lights give off massive emission at some wavelengths of the visible spectrum.

For analog proof to press sheet, since they are made of ink pigments, they are not metamers so metamerism doesn't occur even though 5000K fluorescent isn't really very close to D50. But with the increase of proofs that don't use actual ink pigments, this will become an increasing problem because inkjet proofs (for example) and press sheets ARE metamers. A D50 light source will be more crucial in cases where proofs not using ink pigments will be used as contract proofs.

>That was in 1967, and they haven't gone away yet...

Of course not. They're making too much money; especially on "full spectrum" bulbs. Guess what? A candle is also a full spectrum light source, so that's one of those B.S. marketing terms I was referring to before. I'm waiting for everyone to have a hissy fit over this too. It would be very worthwhile - get rid of your fluorescents as you have the opportunity. Besides, the mercury content makes them unfriendly to the environment. Some states now have specific laws regarding their disposal because of this.

Chris Murphy


From: "Ron Bean"
Date: Tue, Jan 4, 2000, 11:32 PM
RE: Re: Color Temperature and Metamerism

Chris Murphy writes:

>5000K is rather vague. The standard
>illuminant for graphic arts is actually D50, not 5000K. They aren't the
>same. D50 is an illuminant, 5000K is not. D50 has a very specific
>"spectral power distribution" which is a fancy way of describing what
>amounts of each visible wavelength that light source outputs.

A "black body" heated to 5000K should also have a very specific spectral power distribution (if you don't believe me, ask a physicist). If that's not the way it's used in the graphic arts, then I think they're mis-using the terminology. BTW the same math describes radiant heating; it's just in a different part of the spectrum.

>What is the standard illuminant 5000K fluorescents are trying >to simulate? I don't know but I find it difficult to believe it's D50 >given the CRI numbers they advertise.

I think the problem is trying to construct a lamp that behaves like a theoretical ideal, instead of the other way around. As a practical matter, it would have made more sense to define the standard in terms of the behavior of a specific type of lamp. If someone used a different type of lamp, that would be non-standard by definition.

That way you'd *know* that everyone was using the same thing, instead of having everyone approach the ideal from different directions, and in ways that don't match each other.

In other words, it would be a real-world standard, instead of a theoretical standard that everyone approaches but nobody actually meets.

Maybe that's what happens when standards are set by theoreticians rather than engineers...


From: Chris Murphy, Date: Wed, Jan 5, 2000, 7:29 PM
RE: Re: Color Temperature and Metamerism

>A "black body" heated to 5000K should also have a very specific
>spectral power distribution

Of course it does. You're saying a black body at 5000K. 5000K by itself isn't very helpful and it's not the same thing as D50 or daylight 5000K.

>I think the problem is trying to construct a lamp that
>behaves like a theoretical ideal, instead of the other way
>around.

Illuminant E is the theoretical ideal. D50 isn't a theoretical ideal, it just happens to be the standard. The problem is fluorescent technology isn't suitable for simulating an illuminant with a smooth spectral power distribution like D50. Halogen can do this with a special filament and reflector, and such a bulb exists:

http://www.soluxtli.com

>Maybe that's what happens when standards are set by theoreticians
>rather than engineers...

I don't know the history of where D50 came to be or why ANSI selected it as the standard. If I had to guess, it would be alone the lines of the human visual system being well adapted to sunlight more than an artificial light source, and they basically had to pick a standard illuminant that didn't have a color cast (like most fluorescent, all incandescent, most tungsten, halogen, and all metal halid, mercury vapor, and sodium I forget what based lights. Those are poor lighting choices. Incandescent is arguably the most common but it's extremely red and has very little purple/blue emission. So basing the standard on a bulb instead of a bulb on a standard I think is backward because with the exception of the SoLux bulb I mention above, there isn't one suitable in terms of consistency (between bulbs and as the bulb ages), and reasonably neutral and a defined illuminant. In order to get a true D65 you need to include a certain amount of UV radiation whereas D50 is very weak in UV. Getting UV emission in a light source that is ALSO well behaved (smooth spectral power distribution instead of spiky) is exceptionally difficult and I'm not aware of it being done or worked on. If you want UV, you usually end up with some fluorescent bulb designed for that purpose and it also has very spiky "other" emission other than UV which those buying the bulb don't care about. In the graphic arts, we would care about that behavior.

The point of bringing up this info about D65 is that it represents something not attainable at this point for a well behaved artificial lighting source. D50 is something that is attainable to within a very decent tolerance of true D50, and it isn't fluorescent technology.

Chris Murphy


From: Ron Bean
Date: Wed, Jan 5, 2000, 10:32 PM
RE: Re: Color Temperature and Metamerism

Chris Murphy writes:

>You're saying a black body at 5000K. 5000K by itself
>isn't very helpful and it's not the same thing as D50 or daylight 5000K.

Which is why it's not helpful to assign a "color temperature" to something that is not black body radiation. It's sloppy terminology, and it confuses the issue by collapsing it into a single term that doesn't describe what's happening.

It would be much better to skip that and talk about the spectral content of the actual light source (my original point was that various textbooks describe this stuff in a way that is misleading).

>D50 isn't a theoretical ideal, it just happens to be the
>standard.

So, how *exactly* do you build a D50 light source, and why isn't it defined in terms of something everyone builds the same way?

>...they basically had to pick a standard
>illuminant that didn't have a color cast (like most fluorescent, all
>incandescent, most tungsten, halogen, and all metal halid, mercury vapor,
>and sodium I forget what based lights. Those are poor lighting choices.

IMHO the best lighting choice would be the one that the product will actually be viewed under (if you know what that is). Your trade show example is a good one; I would argue that D50 is a poor choice in that situation.

>Incandescent is arguably the most common...

I'm not so sure about that these days. I have about half incandescents and half fluorescents at home. Commercial buildings are overwhelmingly fluorescent.

The real problem is that incandescents have too much red, and fluorescents usually have too little, so there's no way a single standard can accomodate them both.

>So basing the standard on a bulb
>instead of a bulb on a standard I think is backward because with the
>exception of the SoLux bulb I mention above, there isn't one suitable in
>terms of consistency (between bulbs and as the bulb ages), and reasonably
>neutral and a defined illuminant.

The problem is that you need a standard that you can hold two objects under to see if they match. That means it has to be a physical device. Yes, it will age, and you have to deal with that somehow. But first you have to agree on *how* to build it, so you don't have radically different light sources claiming to be standard. If you can't do that, then it's unsolvable.

In fact, if everyone used identical so-called "5000K" fluorescents, then that would be a defacto standard, even though it's technically wrong. If they all use *different* so-called "5000K" fluorescents, then it's hardly worth having a standard in the first place.

You could still have other standards for other purposes, but when the question is "do they match", it has to be under an existing light source (however imperfect), and one that the people who buy the lightbulbs can agree on.

>D50 is something that is attainable to within a very
>decent tolerance of true D50, and it isn't fluorescent technology.

So if it were defined as something other than fluorescent, then fluorescents would clearly be wrong, by definition, and there would be no way for anyone to claim otherwise.


From: Chris Murphy Date: Thu, Jan 6, 2000, 1:44 AM
RE: Re: Color Temperature and Metamerism

>Which is why it's not helpful to assign a "color temperature" to
>something that is not black body radiation. It's sloppy terminology,

Yes exactly.

>It would be much better to skip that and talk about the spectral
>content of the actual light source

Yes exactly. That is the definition of an illuminant. An illuminant is a defined spectral power distribution. So an incandescent lightbulb in my home is a light source, but it corresponds to illuminant A as defined by the CIE. Illuminant A has a specific spectral power distribution.

>So, how *exactly* do you build a D50 light source, and why isn't
>it defined in terms of something everyone builds the same way?

Spectral power distribution curves do not define HOW to make a light source produce a given spectral power distribution :) they simply define the AMOUNTS (usualy in % reflectance or normalized output) of each wavelength in the visible spectrum, usually in 10nm increments.

To build such an artificial light source takes a lot of R&D, lots of trial and error, and a good spectroradiometer to ascertain the spectral power distribution of your light source. Once you determine the spectral power distribution for that light source, you can call it an illuminant (light source + spectral power distribution curve = illuminant). Once you do this, you can patent it, and someone has done this and they have patented it.

>IMHO the best lighting choice would be the one that the product
>will actually be viewed under (if you know what that is). Your
>trade show example is a good one; I would argue that D50 is a
>poor choice in that situation.

I disagree. I think the issue is that it would be difficult to get enough "D50" type lights in such a room to provide adequate lighting. D50 is the single standard. People don't follow the single illuminant standard for lighting conditions so I don't see how having more than one is going to help matter.

Despite that, I'm not saying I do not support multiple illuminant lighting condition standards. For trade shows there would likely be three standards, mercury (or mercury blended), metal halide, and one other I'm not able to remember at the moment. Anyway, you could have to make custom separations for each of these lighting conditions in order to get the color you are expecting.

I guarantee you that you're going to have a problemm making separations that compensate for non-standard (non-D50) lighting without using some kind of color management system to assist you. All proofing systems are based on D50, or D65, so conventional proofing systems aren't going to even help you. Plus, the service provider of these products will need to invest in multiple proofing "booths" or areas in order to provide suitable lighting for each standard you decide on.

>I'm not so sure about that these days. I have about half
>incandescents and half fluorescents at home. Commercial buildings
>are overwhelmingly fluorescent.

I have compact fluorescent outside, and primarily incandescent indoors with a couple of halogen lights. Even if it turns out the situation is mixed, you realize that for distributed advertising, you really can't target a standard lighting condition. Even within fluorescent lighting there are more than 12 different standard illuminants defined by the CIE, probably more independent (non-standard) fluorescent based illuminants exist as well.

>The real problem is that incandescents have too much red, and
>fluorescents usually have too little, so there's no way a single
>standard can accomodate them both.

Yes. Fluorescents also have massive (MASSIVE) spikes in orange-yellow and another at yellow-green. I'm sure there's another but I don't have any spectral power distribution curves for any fluorescent bulbs since I'm at Macworld at the moment.

>The problem is that you need a standard that you can hold two
>objects under to see if they match.

If it's an analog proof and a press sheet, both of which contain actual press ink pigments (i.e. they are NOT metamers) they will look the same and MATCH under all lighting conditions. They won't look correct (per the standard), but they will look wrong in the same way.

With the increasing use of proofs that do not use actual press ink pigments, they are metamers. The farther you deviate from D50 which is what separations are based on (Photoshop separation tables, curves, and ICC Profiles all assume D50 lighting), the bigger the potential different between press sheet and proof under non-D50 lighting conditions.

>Yes, it will age, and you have to deal with
>that somehow.

The SoLux bulbs essentially die before their color temperature (spectra power distribution) is significantly atered (like maybe 30 to 50 K). Fluorescents can last a very long time well after they are no longer qualified for operation, as much as 500K. Most wait until they flicker before they replace their lights.

> But first you have to agree on *how* to build it,
>so you don't have radically different light sources claiming to
>be standard. If you can't do that, then it's unsolvable.

They don't claim to be the standard. No light claims to have a CRI of 100. SoLux is the closest it gets at 99+.

>In fact, if everyone used identical so-called "5000K" fluorescents,
>then that would be a defacto standard, even though it's technically
>wrong. If they all use *different* so-called "5000K" fluorescents,
>then it's hardly worth having a standard in the first place.

Then the companies aren't differentiating themselves in their own minds. Why don't printers follow standards exactly and force their presses to behave exacly alike?

Chris Murphy


From: Ron Bean
Date: Sun, Jan 9, 2000, 1:10 PM
RE: Re: Color Temperature and Metamerism

Chris Murphy writes:

[lotsa stuff I basically agree with, just a few comments...]

>To build such an artificial light source takes a lot of R&D, lots of
>trial and error, and a good spectroradiometer to ascertain the spectral
>power distribution of your light source.

It would take a lot less trial and error if the standard were based on an existing light source.

>People don't follow the single illuminant standard for
>lighting conditions so I don't see how having more than one is going to
>help matters.

They don't follow the standard because it was too hard to make a lightbulb that follows it closely. Now that the SoLux bulbs are available, it's a lot easier.

Unfortunately, people are used to using the fluorescents, and they're not likely to change unless they become dissatisfied with them for some reason.

>Despite that, I'm not saying I do not support multiple illuminant
>lighting condition standards...

>I guarantee you that you're going to have a problem making separations
>that compensate for non-standard (non-D50) lighting without using some
>kind of color management system to assist you.

Well, we have those now...

>Plus, the service provider of these products will need to
>invest in multiple proofing "booths" or areas in order to provide
>suitable lighting for each standard you decide on.

Probably one booth with several light sources.

[BTW Billmeyer & Saltzman claim you should specify the entire booth, not just the lightbulbs. That sounds like overkill to me, but they know more about color than I do.]

>Even if it turns out the situation is
>mixed, you realize that for distributed advertising, you really can't
>target a standard lighting condition.

Yeah, that's the paradox. Even if you get a perfect color match, it's never going to be viewed under those lighting conditions again.

>With the increasing use of proofs that do not use actual press ink
>pigments, they are metamers. The farther you deviate from D50 which is
>what separations are based on (Photoshop separation tables, curves, and
>ICC Profiles all assume D50 lighting), the bigger the potential different
>between press sheet and proof under non-D50 lighting conditions.

Which is why the separations should be based on whatever they're using in the pressroom. If that's the SoLux bulbs, no problem. If it's the usual fluorescents, most likely the printer will blame the designer for any problems.


From: Chris Murphy,
Date: Sun, Jan 9, 2000, 7:41 PM
RE: Re: Color Temperature and Metamerism

>It would take a lot less trial and error if the standard were
>based on an existing light source.

I'd like to hear what existing light source that is NOT PATENTED is also suitable as a proofing light source.

>They don't follow the standard because it was too hard to make a
>lightbulb that follows it closely. Now that the SoLux bulbs are
>available, it's a lot easier.

Before SoLux there was 5000K fluorescent; even when those were available people were either uninformed or ignored the standard lighting condition and still complained about color, and could not believe lighting could make that much difference.

The point of this is that if one standard is impossible for a majority of people to follow, multiple standards will be also.

Chris Murphy


From: Chris Murphy
Date: Mon, Jan 10, 2000, 1:50 AM
RE: Re: Color Temperature and Metamerism

>If it's patented, you license it.

I would not support making a patented technology the standard. It would be unfair for ANSI to say, "this company gets to win the lottery."

>Since this would generate quite a bit of business for the patent
>holder, they'd probably give you a pretty good rate (and patents
>do expire).

Why would they give a good rate? If ANSI says, "Joe Blow's technology is the standard, anyone who wants to make it has to go to them," Joe Blow has every incentive to ramp up production and maintain a monopology so no one can compete. That's how markets work. The only time you get a good rate is when there is competition.

>It's not *that* unusual for standards to specify patented
>technology. If you want to build a standard 56K modem, you have
>to license technologies from several companies.

yes and remember what it was like when company A couldn't use company's b's technology or vice versa. It basically took 2 years for them to figure out they needed to co-develop and patent the v.90. There are v.90 modem choices today low cost because more than one company has licensing authority for v.90. Your lighting example is nothing like this.

> The standard way
>of measuring octane for gasoline uses a special test engine made
>by a particular company.

There are other formulas for computing octane. That's competition. They do not generate the same results, but they still compete. If this one company abuses their patent, the state or federal agencies could mandate their own standard for sake of consistent terminology for consumers.

>So what's your solution for making people follow the standard?

Send Christmas cards to all their customers telling them that their vendor doesn't use standard lighting. I don't have all the answers. Ultimately it comes down to education, so I'm doing what I can to educate people on the problem. There are limits to what any one individual can do.

Chris Murphy


From: Ron Bean
Date: Sun, Jan 9, 2000, 10:31 PM
RE: Re: Color Temperature and Metamerism

Chris Murphy writes:

>I'd like to hear what existing light source that is NOT PATENTED is also
>suitable as a proofing light source.

If it's patented, you license it.

Since this would generate quite a bit of business for the patent holder, they'd probably give you a pretty good rate (and patents do expire).

It's not *that* unusual for standards to specify patented technology. If you want to build a standard 56K modem, you have to license technologies from several companies. The standard way of measuring octane for gasoline uses a special test engine made by a particular company.

>The point of this is that if one standard is impossible for a majority of
>people to follow, multiple standards will be also.

So what's your solution for making people follow the standard?

I don't think there's a technological solution to that one. But I'd like to think there might be a non-technological one.

Adobe Photoshop training classes are taught in the US by Sterling Ledet & Associates, Inc.