CRI - marketing ploys

[Dr Evil]

I've known what CRI for quite a while. The one thing I've always wondered is how companies claim 90+ CRI through a broad range of color temps. I always thought you'd have to have a light close to daylight to get 100 CRI. My mind was going crazy trying to comprehend how it would be possible to have 100 CRI across the light temp range. Especially since "white" lights are yellowish at one end and blueish at the other end of the kelvin range. That seems to be partially true. Daylight will give you the broadest range of colors. It would be nice if light manufacturers based the CRI on true daylight.

Anyway, my search for some of my CRI questions took me to http://www.fullspectrumsolutions.com/cri_explained.htm. It seems like a very good CRI explanation to me.

[AnAppleSnail]

Anyway, my search for some of my CRI questions took me to http://www.fullspectrumsolutions.com/cri_explained.htm. It seems like a very good CRI explanation to me.

There is some temperature limit on the CRI range, but if your light source "Approximates the distribution of light of a given-temperature blackbody radiator" then it can have a high CRI. If it weren't for that limit, then my thumb has a CRI of 100...for a CCT of 300K or so. With some LED structures it's easier to claim a high CRI by doping the phosphor to produce a lower color temperature. This makes sense, since the blue LED produces a blue spike, and the phosphor a broad spread much like a black-body radiator. Check out the relative spectral distributions of cold, neutral, and warm-white LEDs compared to a black-body radiator.

[Kinnza]

CRI is a measure of the color rendition fidelity of the testlight with respect to either a blackbody radiator of the same CCT (for CCT<5000K) or a daylight phase of equal CCT (for CCT>5000K).

Thus is not actually meaningfull to compare CRI of different CCTs, something that most people are not aware of. CRI not uses a fixed reference, but a variable one depending of the CCT you are analyzing. It allows lights of any CCT to have high CRI, which is a great marketing tool, but actually informs little about the absolute color rendering of the light.

CRI not says anything about subjetive sensation of fidelity except if you consider the CCT your mind sets as the "ideal" color rendering, something that changes from person to person.

In order to get better info about absolute color rendering, taking into account how different CCTs renders colors, you need to use different color metrics than CRI. For pure fidelity, the CRI-GAI metric (which averages CRI score and GAI -Gamut Area Index- score) works better as some scientific studies has shown: although it has the same problem about depending of a variable CCT reference, CRI and GAI usually works on opposite directions so averaging both usually gets a more meaningful figure.

The metric which has shown to correlate better with absolute color rendition, especially for attractiveness and appealing effect is the Memory Color Quality Index, as it dont use any reference source light, but an average of perception of some common objects.

[Cavannus]

Hi, I have a question about colour rendering index.

For a given warm colour temperature, e.g. 2700K, is CRI related to the emitted light colour/tint as well as to lighted object colour rendering?

Let me explain what I mean:

- I have some 2700K fluorescent tubes and led bulbs whose CRI is rated at 80. Colours render well, i.e. browns have a nice warm colour, etc. -- BUT the light tint is somewhat pinkish compared to an incandescent bulb, so that you immediately notice that this is not incandescent.

- I also have an AmbientLED A-type bulb with remote phosphor, whose CRI is 80. The tint is yellowish as an incandescent bulb but colours render badly: browns and oranges look bluish, skin looks sick, etc.

- I own a couple of led flashlights that use the Cree 90 XP-G: the CRI is 90+ and both tint and colour rendering are very close to a halogen bulb.

=> SO: if I get a 90+ CRI led bulb (such as the new Philips L-Prize Winner or any 90 CRI rated bulb), should I expect both tint colour and colour rendering close to a heated filament? I.e. should I expect that this bulb with mimic an incandescent light as the 90 XP-G does?

Thanks for your attention

[Kinnza]

CRI figure is completely unrelated to color temperature/tone/tint. Color tone of cool lights (high K) tend to be somewhat bluish while warm whites tend to be yellowish, and that is valid irrespective of the CRI, as the CRI is calculated against a reference of the same CCT (Correlated Color Temperature in K).

CRI not informs about how good is the color rendering of a given light, but how close it renders colors as compared to a reference light of the same CCT. So you cant compare the color rendering of lights with different CCT based on CRI, as each are using a different reference. And of course CRI dont say nothing about the tint of the light, a factor that affects a lot to perceived color, so depending of each person preference about color temperature, some people may think a light with lower CRI render colors more "naturally" than other with higher CRI but different CCT because they prefer the effect that lower CCT light does to object's color appearance

If you get a LED with CCT close to your preferred halogen, then the higher its CRI, the closer it renders colors than the halogen. On the other hand, heated filaments emits light along a wide range of CCTs. Just check how changes the color tone as you dim an halogen light.

[notrefined]

For a given CCT <5000k, the tint of any light source will have to approach the tint of an incandescent source at the same CCT as the CRI approaches 100. That is, the closer you are to completely mimicking the appearance of a black body radiator, the less room there is to stray from its spectrum. So a very high CRI source at 3000k will have to look pretty yellow like an incandescent body at that temperature, with only a little room to stray to pink or green or whatever. But a lower CRI source can be more biased in any given direction and still score the specified CRI. Does this help?

edited to add: for example, I have a number of led lights using the minimum 90 CRI, 3000k XP-G. There is a definite variation in tint, with some being more yellow and some being more pink. The range of tints overall is much smaller, however, than the equivalent 80CRI part, which may appear much more pink or yellow at the same specified 3000k color temperature.

Hi, I have a question about colour rendering index.

For a given warm colour temperature, e.g. 2700K, is CRI related to the emitted light colour/tint as well as to lighted object colour rendering?

Let me explain what I mean:

- I have some 2700K fluorescent tubes and led bulbs whose CRI is rated at 80. Colours render well, i.e. browns have a nice warm colour, etc. -- BUT the light tint is somewhat pinkish compared to an incandescent bulb, so that you immediately notice that this is not incandescent.

- I also have an AmbientLED A-type bulb with remote phosphor, whose CRI is 80. The tint is yellowish as an incandescent bulb but colours render badly: browns and oranges look bluish, skin looks sick, etc.

- I own a couple of led flashlights that use the Cree 90 XP-G: the CRI is 90+ and both tint and colour rendering are very close to a halogen bulb.

=> SO: if I get a 90+ CRI led bulb (such as the new Philips L-Prize Winner or any 90 CRI rated bulb), should I expect both tint colour and colour rendering close to a heated filament? I.e. should I expect that this bulb with mimic an incandescent light as the 90 XP-G does?

Thanks for your attention

[Cavannus]

Kinnza: thanks for your reply, however my question was not about the difference between CRI and CCT, it was about CRI parameters for a GIVEN CCT (but if I publish a post on my blog about colour rendering and colour temperature, I will base my text on yours!!).


Notrefined: yes it does help! (So mmhhhhh... shoud I buy the L-Prize Winner bulb??... ).

[Kinnza]

I was actually answering to that question, but probably I wasn't clear enough. English is second language for me, sometimes I have it difficult to explain nuances.


Color rendering and color tone are related, but are not the same thing:


-Color tone is the "tint" of the light itself, as seen viewed directly or reflected on a white wall. It is defined by its chromaticity coordinates on a color space. The most used are the CIE 1931 (x,y) and the CIELUVW (u',v'). The coordinates defines the color appearance of the light. A given color appearance may be obtained with miriads of different spectrums, so while a given spectrum of light defines an unique set of coordinates, you can't associate directly a set of coordinates to any concrete spectrum.


Thus you can get same coordinates with different spectrums.


The planckian Locus is a line that crosses the color space, being each point of it the chromaticity coordinates of all black body radiators (for each temperature in K). As for lights with CCT below 5000K, the reference for CRI is black body radiators, by definition BB radiators have CRI 100.


But you can get same chromaticity coordinates than a BB radiator with a different spectrum. In this case, the tint of the light is the same for both. But they wont render colors the same because their spectrum is different.

On the other hand, you can obtain a very similar color rendering with lights that have different chromaticity coordinates.

-Color rendering is a very complex topic. The thread started about the CRI marketing ploys: how lighting manufacturers uses the flaws on the CRI calculation to releases lamps with high CRI figures but that it not always result on the excellent color rendering it should credit.

So when you ask

SO: if I get a 90+ CRI led bulb (such as the new Philips L-Prize Winner or any 90 CRI rated bulb), should I expect both tint colour and colour rendering close to a heated filament? I.e. should I expect that this bulb with mimic an incandescent light as the 90 XP-G does?

the answer is more complex that you may think initially, requiring different explanations over the many questions in that phrase. As tint color and color rendering are different things, answer is different for both:

-Tint color of a 90+ CRI should be similar to a incan/halogen of same color temperature (CCT)?

Not necessarily.

It will be very similar if the distance of the 90+CRI light to the planckian locus is very small or nill (in which case, the tint will be exactly the same). This distance is called Duv and you can get it on the reports of quality brands lamps. In the example cited, the Phillips L-Prize lamp has a Duv of 0.00135 with a CCT of 2810K. That Duv correspond to a little more than 2M (McAdam ellipse), where 90% of persons notice the difference, although it is small and very acceptable for most people.

However, ANSI areas for SSL covers 7M which is a different of tint very noticeable for anybody. So you dont have any guarantee for many products than their 90+CRI LEDs actually have a tint very close toa incan of same CCT.

Lately many LED manufacturers are offering tighter bin areas with 3M diameter. With product using that binning and that provides good color stability along lifetime, you can expect a very close tint to incans.

-Color rendering of a 90+CRI part should be very close to an incan of same CCT?

Theoretically, it should, as it is what supposedly CRI measures.

In the practice, it is not so clear due CRI metric has many flaws and manufacturers uses them to get high CRI figures to lights which not render colors as close as that figure should guarantee.

The most apparent flaw of the CRI calculation is the inadequate samples. No any of them represent saturated colors, and red tones are mostly missed. Thus you can get a very high CRI for a light which renders red tones very poorly or which makes very poorly for saturated colors (along all the spectrum).

And actually, probably the most valued characteristic of incans is how they enhance the red tones. Due that, many manufacturers currently report R9, the figure of CRI for the 9th sample, which is not included in the general CRI (Ra) figure. R9 is a good figure to know how the light renders red tones.

The shorter answer to the question is you can expect a 90+CRI light render colors very close to incandescent if its R9 figure is very high too. Unfortunately, little LEDs achieves it for the moment. Best ones are those that reach a R9 of 80. The Phillips L-Prize has R9=78. Good but not so much, you will notice the difference on red tones and specially, on saturated ones.

[notrefined]

Wow...Kinnza, that was the single best and most succinct comment on this topic I have ever heard or read.

[Cavannus]

Kinnza, English is not my my first language either, so I really appreciate your precise and long explanation.



Let me read it one more time (or several times), since I'm not a technical guy; you will help me figure out what CRI means and why some high CRIs are not as close as natural lighting as claimed (I personally clearly see that the XP-G 90 is NOT incandescent, while I tend to ignore it in real outdoor use such as caving).

[AnAppleSnail]

Let me read it one more time (or several times), since I'm not a technical guy; you will help me figure out what CRI means and why some high CRIs are not as close as natural lighting as claimed (I personally clearly see that the XP-G 90 is NOT incandescent, while I tend to ignore it in real outdoor use such as caving).

The "Color Rendering Index" does not answer "How well does this light source let me perceive color?". Instead, it answers "How much is this source like a black-body radiator of a given temperature (CCT)?" Color perception is a different question from light source similarities. As an example:

Consider the glowing coils in a toaster. That red-orange glow comes from an object of about 1000K. The 'spectral distribution' of such a source would be heavily weighted towards the red and infrared. Such a source would produce few green or blue photons, with relatively more yellow, red and infrared. This is not a good light source to perceive color. This is a 100 CRI light source, but you could not judge colors very well with it.

A weak mini mag lite bulb is around 3000K. It is a yellowish white that can let you distinguish many colors, although blues will be muted. If you are a designer looking for just so-and-so a shade of hair, such light would not be very good - but better than a toaster. This 100CRI filament is acceptable for seeing color. An LED could be produced to closely match the spectral output of this light - it might be 80 or 90 CRI with a CCT of 3000. It may also be acceptable for seeing color.

Sunlight is around 5000K - unsurprisingly, the surface of the sun is close to this temperature. This is the light source human eyes probably evolved to handle best, and is excellent to see color with. An LED with this CCT and a high CRI (There is a very expensive, very rare Nichia like this) would be quite good for perceiving colors.

There are different tests for color perception, but they boil down to sitting down groups of people to see what color cards they can or cannot differentiate between under given lighting conditions. The group sizes and methods are governed by the standard test, and gives a good handle on how well light works for a given task, but that's a more onerous test than a spectrograph to get the CCT and CRI.

[kaichu dento]

AppleSnail, as usual an excellent post, but I just wanted to make one correction regarding a common misconception about the human eye and the 5000k output of the sun. While it's easier to zero in on a particular number like this, it is so obvious as to defy need to point out, but very few hours of human waking have ever followed midday sun, and as such have also the ability to work well during the early morning hours, late evening, not to mention different phases of moonlight - all normal light sources that our eyes have been working with for millennia.

So once again, the importance of personal preference comes in as far more important than the search for the '5000k holy grail', or whichever other combination one feels compelled to strive for. It's even possible for people to like many different tints and light sources, just in the same way as I can love being under direct sunlight at the beach when the CCT would be highest, or late evening when it gets warmer and warmer, or full moonlight with it's eerily bluish magical lighting, I also love my Nichia 119 Haiku, my hCRI V10R and my Titan.

Think about it guys - don't you actually have a range of favorites, as opposed to one, and one only?

[mvyrmnd]

AppleSnail, as usual an excellent post, but I just wanted to make one correction regarding a common misconception about the human eye and the 5000k output of the sun. While it's easier to zero in on a particular number like this, it is so obvious as to defy need to point out, but very few hours of human waking have ever followed midday sun, and as such have also the ability to work well during the early morning hours, late evening, not to mention different phases of moonlight - all normal light sources that our eyes have been working with for millennia.

So once again, the importance of personal preference comes in as far more important than the search for the '5000k holy grail', or whichever other combination one feels compelled to strive for. It's even possible for people to like many different tints and light sources, just in the same way as I can love being under direct sunlight at the beach when the CCT would be highest, or late evening when it gets warmer and warmer, or full moonlight with it's eerily bluish magical lighting, I also love my Nichia 119 Haiku, my hCRI V10R and my Titan.

Think about it guys - don't you actually have a range of favorites, as opposed to one, and one only?

I agree with your entire post, expect the last bit. To my eye, 3000-3200K is perfect for seeing at night. The 90-CRI 7A3 XP-G meets my needs and preferences perfectly. Above 4000K I can't see as well, or as far, even with a brighter light.

I believe personal preference is a huge part of the equation. Luckily for me, my preference has already been catered for

[kaichu dento]

I agree with your entire post, expect the last bit. To my eye, 3000-3200K is perfect for seeing at night. The 90-CRI 7A3 XP-G meets my needs and preferences perfectly. Above 4000K I can't see as well, or as far, even with a brighter light.

I believe personal preference is a huge part of the equation. Luckily for me, my preference has already been catered for

You have the same tastes as I do, but I've been finding that I love the cool, moonlight tint from the floody beam of my Titan a lot when at lower levels - it's like the moon is shining, and having always seen the moonlight as romantic and magical, I like it too.

[TEEJ]

You have the same tastes as I do, but I've been finding that I love the cool, moonlight tint from the floody beam of my Titan a lot when at lower levels - it's like the moon is shining, and having always seen the moonlight as romantic and magical, I like it too.

So, instead of a high CRI light, maybe we need a high RMI. (Romantic and Magical Index)

[kaichu dento]

So, instead of a high CRI light, maybe we need a high RMI. (Romantic and Magical Index)

Not that the two are mutually exclusive, but yeah, I think so.