High CRI and its significance

[Icebreak]

Erasmus -

I haven't seen a spectrum chart on Cree WW but I've seen one for a Luxeion IV WW. It had a much smaller blue spike than most white LEDs.

I can't tell a alot from the photos other than it looks nice. Next time you do shots would you consider something like strawberries or roses? Anyway you seem to like it and your perspective is important.

Like the unit quite a bit. I wouldn't mind having several of them.

mejesster -

Most development for better CRI in LEDs shy away from filters. I saw an LED lantern at Academy Sports with an amber filter. My rough caveman experiments have shown it doesn't work well. This is because most white LEDs have a blue spike, a cyan dip and weak red. Filters are subtractive so they won't create color that's needed.

Krisp -

Interesting combination. As you know several folks have tried different combinations but I've never heard of yours. Somebody, I think Milky, built one for somebody using a P7 and R/O. I'd like to see that one too. I saw a strange one on Craig's LED Museum site that had a cyan with a with a white. Spectra looked like a mountain with no blue spike. It looked symetric.

Photos of strawberries and flowers would help to see what your light really looked like but again, perspective counts a lot and if you like it it's probably a good CRI beam.

 

[2xTrinity]

Here is one point I'd like to clarify about CRI which has been asked repeatedly in this thread:

CRI is a measure of how closely an emitter spectrum will render a pallete of test colors, compared to a blackbody radiator of the same color temperature. This means the CRI numbers are only valid for comparison of lights at similar color temp, for example, comparing a 3500k incandescent to a 3500k LED.

Saying that a 100CRI long-life incan at 2700k is "better" than a 80CRI fluorescent at 4000k is not a valid comparison.

There are some cases where a warm or a cool tint may be desired independently of color rendition -- this is done to enhance contrast in a particular situtation.

For example, wearing blue-blocking sunglasses outside while sailing will greatly increase contrast between ships, rocks etc. and the surroudnign blue sea and blue sky. Even though color rendition will be worse with the glasses on, as they are subtracting from a "perfect" daylight source.

Another example is the case brought up by BabyDoc, performing oral exam, a warm light source is preferable because he is looking for subtle differences in shades of red, so having a light source that exaggerates those differences is desired -- EVEN if that light is technically less accurate compared natural sunlight.

Another example is that of people preferrng incandescent light to ~4200k HID light outdoors for spotting animals. In my experience, the latter renders scenery closer to how it looks under natural light, but with the incan, spotting animal is easier, because the contrast between brown fur and green foliage is exaggerated. Grass for example often looks "Deader" under incndescent because reddish-brown components are boosted. This is purely a function of color temperature and is really not the same as CRI.

So IMO the best approach is to FIRST find the optimum color temp for your applicaiton, then find the light source that has the highest CRI. I personally almost never like to depart from the 3500k-5500k range, regardless of CRI.

One other quick question or two.
does CRI correspond directly to daylight color balance?

Can filters make for better CRI numbers?

It seems to me that an emitter which is filterd so the color tempature is corrected looses some of it's OOMPH.
It seems that an emitter which comes "pre corrected" (for lack of a better term) comes having lost some of it's OOMPH to begin with.

I was just wondering if the loss of lumens due to filtering is similar to the amount of loss (if you want to call it that) that the hi CRI emitters come with when compared to non high CRI emitters?
If this doesn't make any sence please just ignore.
Thank you.
Yaesumofo

Any time a filter is used, output is going to be lost as you are simply blocking portions of the spectrum. Also, in the case of filtering cool white LEDs, that's definitely the wrong approach, as those are actually producing a pretty decent spectrum expect for the fact they're MISSING the deep red wavelengths. No mount of subtractive filters will add red to the output. Applying CTO filters for example, will just end up giving you a yellow-green appearance, rather than a warm white, unless you somehow actually add red.

While it is true the present time that warm white LEDs are less efficient than cool white, this is only a marginal difference, and this won't necessarily be true in the future -- as the phosphor process is improved, it will be possible to make warm and neutral white LEDs that actually produce MORE lumens per watt than their cool counterparts, due to the fact that the human eye has its sensitivity peak around the yellow-green portion of the spectrum, and very low sensitivity to blue light.

Warm White Crees for example are around in Q2 bin, whcih is only about 16% lower output thn the best cool whites. Filtering a 6500k cool LED to have the same spectrum as a warm LED on the other hand, you'd lose at least 60% of the output output.

Also, note that I'm talking strictly about color temperature, not necessarily maximizing CRI, though the warm white LEDs do have higher CRI numbers than their cool counterparts too, because they have a different phosphor blend that doesn't lack as much for deep red (R9).

I know this is distracting from the high CRI LEDs coming out, but I thought I'd add a bit of info from a different discipline.
It's fairly common practice in stage lighting to use lights of two colors, one warm and one cool, often amber and blue, to balance color rendition. Has anyone considered or tried multi-LED/color filtering solutions to get better color reproduction?

This is a perfectly good and relevant point. I have had good success mixing white LEDs with a greenish tint to them (Cree WH bin) with Red LED with overlapping beam patterns. Mixing Cool white LEDs with bluish tints, and incandescent light also produces similar result -- in both cases a 4500k light with high CRI.

If you look at the cool white LEDs, just about all the CRI-indices are above 70, except R9, which is -10. That corresponds to deep red. Supplementing that with an external red emitter (it doesn't have to be high power at all) will greatly imrprove CRI without harming efficiency. The amount of power consumed by the red LED relative to a white LED in an optimm array is about 10% at most. So that means the overall hit to lumens/watt using that method is only around 6%.

However, I haven't use this method much these days. Most of the applications where I want LEDs are small applications where only single emitters are workable. That's why I've been despearately wanting someone to do a group buy on the neutral and warm white Crees, which are only marginally lower lumens/watt but dramatically better light quality.

Color aside, the bigest advantage of the LED light for me as a pediatrician is the bright and even illumination of the oral cavity. The average physician either uses a simple incadescent flashlight or uses a Welch Allen otoscope to look at the mouth. Either of these devices have too narrow a hotspot and a less intense spill area, making it necessary to reaim the light a few times as you look around the light looking for details.

High CRI warm LEDs have a number of advantages compared to incans, particularly for dimming:

• More power efficient
• Less Waste Heat
• Efficiency actually INCREASES as the light is dimmed
• Color temperature remains the same as the light is dimmed
• Possible to dim to much lower levels

 

[Stillphoto]

Great lesson on the applications of CRI in everyday life Don. This is a very exciting thread for me to check out, being a photographer/working in film.

I would have to agree that yes, the sunlight shot you took in the shade would be under the influence of some extra blue. Similar to when its slightly cloudy out, hence why the cloudy white balance bumps up the warm tones to compensate.

I must also mention that some materials/fabrics just don't reflect light in a way that our cameras can properly render them, especially synthetics. Hence why alot of times black nylon bags and such are rendered with a much higher red content to them. From experience, even the best zeiss optics and $30k+ digital large format systems are thrown off by it.

 

[err0r]

Great lesson on the applications of CRI in everyday life Don. This is a very exciting thread for me to check out, being a photographer/working in film.

I would have to agree that yes, the sunlight shot you took in the shade would be under the influence of some extra blue. Similar to when its slightly cloudy out, hence why the cloudy white balance bumps up the warm tones to compensate.

I must also mention that some materials/fabrics just don't reflect light in a way that our cameras can properly render them, especially synthetics. Hence why alot of times black nylon bags and such are rendered with a much higher red content to them. From experience, even the best zeiss optics and $30k+ digital large format systems are thrown off by it.

Typically due to reflected infrared light, as digital sensors can be rather sensitive to IR, even if the camera contains internal IR filters (as is the norm). For an extreme example of what this can do, look out for photos taken with the Leica M8 (see this page) which lacks an internal IR filter all together.

 

[McGizmo]

2xTrinity,
Thanks for the clear explanation and expanding on the CRI and how it is dependent on color temp. I also appreciate how you have cut to the meat of what really matters and how to consider our light options and how we might want to consider them.

ErrOr,

I recently posted about a IR PD flashlight I built and showed how the camera in my phone could actually use it as a light source. :green: Strong UV will also show up in a digital image without some filtering I have also found.

There is some photography discussion now in the SunDrop thread and I took a shot of possible interest I will post in that thread instead of here.

 

[jtr1962]

The spec sheet for the Nichia 083 says 80 lumens typical at 350 mA. This is better than expected given the color rendering. I've recently been playing around with RGB Rebels. Very interesting to say the least. I tried a good approximation of 6500K white and reds rendered quite well. This just proves you don't need a warm emitter to get good color rendering. In fact, with the RGB approach colors look more vivid than under sunlight. I've heard that technically the CRI of RGB is often poor, even though nearly everyone finds the results more appealing than anything else. Anyway, interesting topic. Now that LEDs are finally starting to get bright enough, it's time we worked on their color. I think we need high CRI sources in warm, neutral, and cool tints for the general lighting market. Everyone has their own perferences, and we shouldn't have to trade, for example, high CCT for good CRI.

 

[Stillphoto]

Typically due to reflected infrared light, as digital sensors can be rather sensitive to IR, even if the camera contains internal IR filters (as is the norm). For an extreme example of what this can do, look out for photos taken with the Leica M8 (see this page) which lacks an internal IR filter all together.

Yep, exactly the photos I was thinking of when I typed this.

 

[2xTrinity]

The spec sheet for the Nichia 083 says 80 lumens typical at 350 mA. This is better than expected given the color rendering. I've recently been playing around with RGB Rebels. Very interesting to say the least. I tried a good approximation of 6500K white and reds rendered quite well. This just proves you don't need a warm emitter to get good color rendering. In fact, with the RGB approach colors look more vivid than under sunlight. I've heard that technically the CRI of RGB is often poor, even though nearly everyone finds the results more appealing than anything else.

CRI will be bad because more vividly saturated colors are still a departure from a 6500k blackbody source, even though it's arguably a good departure. That's the main weakness of the CRI system -- it doesn't specify HOW a source differs from an ideal source. This is also what makes fluorescent lamps from different brands difficult to match. For example, in the room I am in I have three different lamps of the same CRI and CCT (3500k, 84CRI). Each kind of lamp is a different tint compared to each other, and each s different compared to a 3500k hotwire (ideal source).

Oddly enough, the one that looks the most like the incan on a white wall actually is actually the most different from the incan when it comes to actual color rendition (looking at a painting in the room). But when all are on at the same time, the slight differences all "average out" and the end result is pretty good.

Anyway, interesting topic. Now that LEDs are finally starting to get bright enough, it's time we worked on their color. I think we need high CRI sources in warm, neutral, and cool tints for the general lighting market. Everyone has their own perferences, and we shouldn't have to trade, for example, high CCT for good CRI.

Agreed. High CRI High CCT is possible with fluoros, but not in a concentrated beam like you can get with LEDs. Finding 5000k in high CRI (>90) is quite easy. Low CCT high CRI is possible with incans. However, there is just about nothing that is neutral CCT (between 3500k and 5000k) that is also high CRI.

I personally am afraid that market forces will cause LEDs to really only be available in warm color temp with high CRI, in order to mimic existing incan fixtures, and a neutral high CCT source--which I believe is actually the ideal for most situations--will only be made in small quantity simply because people aren't used to anything like that existing.

 

[BabyDoc]

Here is one point I'd like to clarify about CRI which has been asked repeatedly in this thread:

CRI is a measure of how closely an emitter spectrum will render a pallete of test colors, compared to a blackbody radiator of the same color temperature. This means the CRI numbers are only valid for comparison of lights at similar color temp, for example, comparing a 3500k incandescent to a 3500k LED.

Saying that a 100CRI long-life incan at 2700k is "better" than a 80CRI fluorescent at 4000k is not a valid comparison.

There are some cases where a warm or a cool tint may be desired independently of color rendition -- this is done to enhance contrast in a particular situtation.

For example, wearing blue-blocking sunglasses outside while sailing will greatly increase contrast between ships, rocks etc. and the surroudnign blue sea and blue sky. Even though color rendition will be worse with the glasses on, as they are subtracting from a "perfect" daylight source.

Another example is the case brought up by BabyDoc, performing oral exam, a warm light source is preferable because he is looking for subtle differences in shades of red, so having a light source that exaggerates those differences is desired -- EVEN if that light is technically less accurate compared natural sunlight.

Another example is that of people preferrng incandescent light to ~4200k HID light outdoors for spotting animals. In my experience, the latter renders scenery closer to how it looks under natural light, but with the incan, spotting animal is easier, because the contrast between brown fur and green foliage is exaggerated. Grass for example often looks "Deader" under incndescent because reddish-brown components are boosted. This is purely a function of color temperature and is really not the same as CRI.

So IMO the best approach is to FIRST find the optimum color temp for your applicaiton, then find the light source that has the highest CRI. I personally almost never like to depart from the 3500k-5500k range, regardless of CRI.


Any time a filter is used, output is going to be lost as you are simply blocking portions of the spectrum. Also, in the case of filtering cool white LEDs, that's definitely the wrong approach, as those are actually producing a pretty decent spectrum expect for the fact they're MISSING the deep red wavelengths. No mount of subtractive filters will add red to the output. Applying CTO filters for example, will just end up giving you a yellow-green appearance, rather than a warm white, unless you somehow actually add red.

While it is true the present time that warm white LEDs are less efficient than cool white, this is only a marginal difference, and this won't necessarily be true in the future -- as the phosphor process is improved, it will be possible to make warm and neutral white LEDs that actually produce MORE lumens per watt than their cool counterparts, due to the fact that the human eye has its sensitivity peak around the yellow-green portion of the spectrum, and very low sensitivity to blue light.

Warm White Crees for example are around in Q2 bin, whcih is only about 16% lower output thn the best cool whites. Filtering a 6500k cool LED to have the same spectrum as a warm LED on the other hand, you'd lose at least 60% of the output output.

Also, note that I'm talking strictly about color temperature, not necessarily maximizing CRI, though the warm white LEDs do have higher CRI numbers than their cool counterparts too, because they have a different phosphor blend that doesn't lack as much for deep red (R9).


This is a perfectly good and relevant point. I have had good success mixing white LEDs with a greenish tint to them (Cree WH bin) with Red LED with overlapping beam patterns. Mixing Cool white LEDs with bluish tints, and incandescent light also produces similar result -- in both cases a 4500k light with high CRI.

If you look at the cool white LEDs, just about all the CRI-indices are above 70, except R9, which is -10. That corresponds to deep red. Supplementing that with an external red emitter (it doesn't have to be high power at all) will greatly imrprove CRI without harming efficiency. The amount of power consumed by the red LED relative to a white LED in an optimm array is about 10% at most. So that means the overall hit to lumens/watt using that method is only around 6%.

However, I haven't use this method much these days. Most of the applications where I want LEDs are small applications where only single emitters are workable. That's why I've been despearately wanting someone to do a group buy on the neutral and warm white Crees, which are only marginally lower lumens/watt but dramatically better light quality.



High CRI warm LEDs have a number of advantages compared to incans, particularly for dimming:

• More power efficient
• Less Waste Heat
• Efficiency actually INCREASES as the light is dimmed
• Color temperature remains the same as the light is dimmed
• Possible to dim to much lower levels

I agree with Trinity x 2 that the ideal color output of an LED not only depends on personal preferences, it also depends on personal needs. Again, the problems of light reproduction remind me of the problems with sound reproduction. Somebody who primarily listens to rock music will probably want sound that emphasizes the bass. People who listen to vocals, will want the midrange emphasized to a certain extent. Those with classical tastes will want a more neutral sound. What is neutral, real sound, however, is still not easily defined, mainly because the same orchestra will sound different in different accoustical environments even more you consider reproducing the sound.

I don't think it is much different with lighting. What is the standard? Even if we say the Sun is the standard in nature by which we should judge neutrality, even that standard changes depending on the time of day, season, cloud conditions, etc. With indoor lighting we have accustomed ourselves to incadescents and its emphasis on yellows, similar to the way the sun emphasizes yellow much of the time. Furthermore, I think women dictate our adhering to this standard because they need to apply makeup in a way that won't make them look ghostly if the light is too blue, or jaundiced if the light is too yellow. Their roses have to be red, and their room colors have to appear as they were when they or their decorator first chose them. As McGizmo has pointed out, meat that needs to look red in order to be appetizing will not be acceptal appearing purple. Although this may be the norm for most people, as Trinity x 2 pointed out, there are times when special needs would dictate a different color spectrum to illuminate those needs. In bright light, for example, don't most of us wear tinted glasses, yet never complain that the colors are altered by the glasses? Somehow, the need over rides any concern about color alteration. We go to a rock bar, and the lights are often flashing purple and nobody is surprised by it. That same light would be totally unacceptable in a medical examination light where as I mentioned before, you have to be able to distinguish shades of pink and red easily.

In sumary, we have different standards for different needs, and makeing a one-light-fits-all is an impossibility. Still, I think it is possible to come up with GENERAL PURPOSE lights most people would prefer for most situations. I believe that that light will have to approach the specturm of the incadescent light for all the reasons I enumerated earlier. Whether that will be best accomplished by an array of different colored LED emitters, or by a single emitter like the Nichia 083 remains to be seen. From what I have seen so far, the Nichia 083 does a better job than any other single LED out there. If not perfect, it is certainly a big step in the right direction.

 

[McGizmo]

Color is but one means of identification and discrimination. It is a piece of information that may or may not be needed or even desired, for that matter.

There seems to be a strong demand in the lighting industry for specific CCT but independent of CRI. I don't think the casual observer gets much beyond the perceived tint or CCT.

I think I may have mentioned that the daylight sampling I took the other day earned a CRIa in the mid 80's if I recall correctly.

As stated, CRI is based both on CCT and black body emissions. LED's don't necessarily play by these rules. That is not to say though that the concept of color rendering is not important or significant in many situations.

Until I recognized the real and significant advantage of including a red LED with an array of white LED's I had not given the samples of these Nichia high CRI LED's much thought. They were warmer than I prefer tint wise and not up to par in term of flux.

With an array of various white and some colored LED's and the ability to drive them independently I can imagine a graphic equalizer where one dials in the LED's to fit a preselected spectral curve. It is likely much more of a challenge to tweak the phosphor of a single emitter in such a manner to fit such a curve. It seems to me that Nichia has done a nice job with these High CRI's and I base this on what I see more than what I attempt to measure.

There has been some interest expressed to me in LED lighting in 6000 CCT but I suspect that also assumed if not understood, is a desire for good color rendition.

I keeping with the concern expressed by 2xTrinity, and to use a weak analogy, I think the lighting industry is in a role of a young parent guiding a young market. The parent can and probably should know what is best and guide the child along accordingly. This is not likely the path of least resistance though. If it gets left to marketing departments, I suspect some great potential will never be realized.

 

[yaesumofo]

http://phoenix.lpl.arizona.edu/news.php
Hey guys. this is an image of martian soil in the scooper on the phoenix craft. One thing which I call interesting about this is the fact that RGB Leds were used to illuminate the subject. How cool.
I do not know if it is three images one of green one of blue and one in red which have been put together. Or if it was just one shot with all three leds illuminating the shot.
NASA has gone to a lot of trouble to insure their ability have accurate color reproduction.
Maybe the HI CRI 083's weren't available yet or maybe they are using other white emitters or maybe they aren't using white but tuned RGB emitters designed for the martian mission...
Who knows? I am sure if we dig a bit we could learn at lot...
Yaesumofo

 

[mahoney]

RGB was probably state of the art when the lander was designed/built, and the receptors in the camera are probably well optimized to the specific wavelengths produced.

However, it turns out RGB is not the best possible way to illuminate something for viewing by the human eye. Your eyes are amazingly adaptable and can make RGB (sort of) work, but if you can compare a colored object side-by-side in RGB light, RGBA, a mix using all 7 available Luxeon colors, a tungsten halogen light source, a ceramic metal halide light source, and sunlight of some sort, you will see a significant difference. A white object will not show as much of difference between the sources.

In general, the fuller and more balanced the spectrum is, the better, at least for accurate color perception. And yes, your perception of color relates to the illumination levels and types you are used viewing things in to as well as the degree of yellowing in the structure of your eyes.

We do an occasional workshop on color and light for our design students. One of the points we demonstrate is this; if your object needs to reflect a particular wavelength of light in order to appear the correct color, then light of that wavelength had better be produced by your light source. Fortunately most things don't reflect light of just one wavelength.

Purple items are actually quite good to demonstrate this, as Don's images show. Purple is a color "created" by our mind in response to blue and red light getting into our eyes at the same time. There is no purple light wavelength. So our perception of the color of a purple object can change dramatically in response to the spectra of the light that's illuminating it. In contrast, a green object could appear green because it's reflecting green wavelengths, or blue and yellow wavelengths, or possibly all of the above. Depending on the reflected spectra coming from a particular green object, our perception of it's color might, or might not, be dramatically altered by a change in the spectra of the light illuminating it.

I am very interested in these Nichia LEDs, and the warm white Crees. I have been happy with the brightness of LED flashlights for a while, but I have not been completely happy with the color rendering. I can't light a paint sample or fabric swatch in a dark corner and really have a good idea what it will look like under a fuller spectrum light source.

 

[GVNCflash]

Determining CRI is done by using a test strip of colors and comparing it's color rendering to the rendering of a reference light. It should never be mistaken for a measurement of light quality or the appeal of the light output.

Adding a "red" LED source into the mix of works well specifically because we don't generally desire a light with good CRI. We don't like lighting with high CRI because it is unflattering. This is why bathrooms sometimes have pink filters in front of lights, and why bathrooms are usually done in pinkish/amber tile/cabinets/etc, it reflects pink/warm tones onto the skin which are more flattering.

Lights with a high CRI are really only useful in specialized situations. They have almost zero usefulness in the home. Ask your girlfriend/wife if she wants candles or fluorescent lights.

Also, color temperature assumes a scale where a blackbody source is radiating(like a piece of iron heated up). Since lamps like HID and fluorescent don't "radiate" like a filament does, the scale has limited usefulness.

NIST is developing a different "CRI" scale for solid-state sources called the [FONT=Arial, Helvetica, sans-serif][SIZE=-1][FONT=Arial, Helvetica, sans-serif][SIZE=-1]Color Quality Scale (CQS). That might be more useful in the future. Hard to say.

I guess the moral of the story is "Be careful what you wish for".
[/SIZE][/FONT][/SIZE][/FONT]

 

[Codeman]

NISAT-CQS

Although it's a few years old, this has some interesting info as well that is pertinent in understanding the limitations of CRI.

 

[McGizmo]

Codeman,
Thanks for the link! It is nice to read a pretty straight forward explanation of what we are up against in trying to quantify and qualify color and color rendition. It is also nice to read about the industry and their awareness of the shortcomings of categorizing SSL by standards based on black body sources.

GVNCflash,

Thanks for your input and insight. In terms of color rendition, as you suggest, the harsh reality of "true" color may not be as pleasant or acceptable to the viewer as a flattering warm wash of light. I think this may be part of the reason behind an apparent interest in CCT with little focus on CRI as expressed or seemingly expressed by marketing and even bin designations.

In the case of a Dr. examining a patient, the truth is more important than a fantasy or pleasantry.

Slightly off topic but in terms of "desire" I think there is a correlation between color temp of light and ambient temperatures. Higher ambient temperatures and climates seem more conducive to higher color temperatures and lower ambient temperatures and cooler climates seem to bring out a desire for the warmer color temperatures; fire for when it's cold?

Below is a data sheet from Nichia in reference to the High CRI 083:

Now you can see they have used 15 color samples which I wonder about. My integrating sphere also has an extend range of CRn numbers.

 

[2xTrinity]

RGB was probably state of the art when the lander was designed/built, and the receptors in the camera are probably well optimized to the specific wavelengths produced.

This is why. Those probes are designed many years before they're actually built an launched, so most of the computer equipment etc. is already obsolete compared to the general market. In the case of LEDs, there's been a huge explosion in the efficiency and output of white LEDs beyween the time when that probe was likely designed, and now.

However, it turns out RGB is not the best possible way to illuminate something for viewing by the human eye. Your eyes are amazingly adaptable and can make RGB (sort of) work, but if you can compare a colored object side-by-side in RGB light, RGBA, a mix using all 7 available Luxeon colors, a tungsten halogen light source, a ceramic metal halide light source, and sunlight of some sort, you will see a significant difference. A white object will not show as much of difference between the sources.

This is just as true for a camera as it is for the human eye, as like the eye, the camera detects color as a combination of three primary colors, using color filters over separate black-and-white CCDs to create the different color responses.

So that's why RGB and blackbody white will appear the same when looked at head on (such as looking at a computer monitor) however, the color of an object is determined by the reflectiveneess of the object at different frequencies. This means for accurate results, a light source similar to our sun (continuous spectrum) must be used. Differences in color temp can be corrected for easily with whit ebalance, but voids in the specturm cannot.

Imagine for example an RGB sceme using 630nm (red), 530nm (green) and 430nm (blue) sources, with narrow spectral line width. Then imagine that there is an object that strictly reflects light at around 590nm. Under a continuous white light, it will appear yellow. Under the hypothetical RGB light I described, it will appear black.

Determining CRI is done by using a test strip of colors and comparing it's color rendering to the rendering of a reference light. It should never be mistaken for a measurement of light quality or the appeal of the light output.

Adding a "red" LED source into the mix of works well specifically because we don't generally desire a light with good CRI. We don't like lighting with high CRI because it is unflattering.

In this case it actually DOES improve CRI. There is a red deficiency in the white LEDs compared to an ideal soure (the CRI9 is -10)

Only a tiny amoutn of red light is actually needed to correct the CRI, and doing this doesn't really change the apparent color temperature much. This is the way I use red mixed with white, in very small ratios. Beyond that point though, you'd be right, the purpose of your red LED would no longer be to correct a deficiency, but to "warmify" the output for aesthetics.

Lights with a high CRI are really only useful in specialized situations. They have almost zero usefulness in the home. Ask your girlfriend/wife if she wants candles or fluorescent lights.

Bad example. The CRI of a fluorescent light is usually around 80. The CRI of a candle is 100 by definition, as it is blackbody radiation.

You are right though in many cases, accurate color reproduction is actually not preffered for aesthetic reasons, for much the same reason as many people prefer EQ settings on music that boost bass or treble, rather than a "flat" EQ.

Also, color temperature assumes a scale where a blackbody source is radiating(like a piece of iron heated up). Since lamps like HID and fluorescent don't "radiate" like a filament does, the scale has limited usefulness.

The sun which is by far the most significant light source is a blackbody radiator, and the effect of rayleigh scattering where different amounts of blue light are filtered away at different latitudes and time of day behaves is simialr to changing the color temperature. There are no natural light sources that produce light outside of the blackbody line (ie, the sun or moon never appear purple or green)

Also, for the sake of computing CRI, true blackbody spectrum is only used up to 3700k (melting point of tungsten), above that, real world daylight samples at different times of day are used as test sources, then assigned the number of the theoretical blackbody temp that would appear to be the same color.

 

[jtr1962]

Looking at that data sheet, one thing I'm wondering about is why it's for a warm white LED (based on the x,y coordinates) when the 083 used in your photos and in the Sundrop appears to be neutral, perhaps around 4500K to 5000K, with no discernable tint either way. Nichia does have a bunch of 083 part numbers in both warm and cool tints. Is the part you have a cool tint, high CRI version? If so, hopefully this means more focus on high CRI not just for the warm tints but the cooler ones as well.

 

[McGizmo]

jtr1962,

I would guess that the subjective term of "warm" was used to best cover the full population of these high CRI LED's. I personally don't like the warm tints and prefer the cooler tints of the higher color temps. It may be due to 56 year old eyeballs that have seen more than their share of abuse and it may be due to the warm climate I live in. The sample high CRI 083's I had were indeed warm in tint and that was another reason I left them in their bag until I became more aware of some need and advantages in better color rendition. I was able to pay a premium and make some bin preferences known in the reel that I purchased. I purchased color rank d2 which is the region below d1 (red dot) in the spectrum graph above. I sought the highest color temperature available within the 083-H1 grouping. Relative to all of the artificial light sources I have around my house which are LED or t8 fluorescent in the garage (high color temp), the Nichia does look warm but only initially. As soon as I focus within the beam of light, I loose the sense of warm tint.

I need to build some lights now but I think one thing I want to do in the near future is remove the integrating sphere from my spectrometer and use a fiber optic sampling head to capture the spectrum off some reflected surfaces of color. I think it would be interesting to put that purple pack pack in bright direct sunlight and capture the spectrum of light reflecting off of the pack.

Having a feel for the emissive qualities of our light sources is one thing but perhaps a better understanding of the reflective qualities and natures of typical targets might be illuminating as well. We only see that which comes to us and in the case of an artificial light source it is the reflected light that matters and not the light sent forth.

 

[jtr1962]

Thanks for the explanation. CCT of the d2 bin based on the diagram would be in the low 4000s. This accounts for its warmish tint relative to the other light sources in your house. I share your preference for cooler tints in general lighting, with 5000K to 5500K being about optimal to my eyes. I just wish somebody would specifically focus on cool LEDs with good color rendering, not that the color of the present ones is horrible. I've also experimented with adding a small bit of red to improve the color rendering of cool white LEDs. One problem is that it tends to make the light a touch pink. Sure, the eyes adjust but I think a better way to go about it might be to mix a RGB source of the same CCT as the main LED. It only needs to be a small percentage of the total output, perhaps 10% or 15%, to improve the overall CRI without affecting the white balance. My theory is that the red of course fills in the missing red while the green fills in the valley. The blue really doesn't fill in anything missing, but it serves to balance out the red and green to prevent "yellowing" of the final result.

 

[GVNCflash]

>>In this case it actually DOES improve CRI. There is a red deficiency in the white LEDs compared to an ideal soure (the CRI9 is -10)<<

You are technically correct, it does improve CRI, however since again CRI is a scale established for blackbody sources, the comparison is not as mathematically easy as this.

>>Only a tiny amount of red light is actually needed to correct the CRI, and doing this doesn't really change the apparent color temperature much. This is the way I use red mixed with white, in very small ratios. Beyond that point though, you'd be right, the purpose of your red LED would no longer be to correct a deficiency, but to "warmify" the output for aesthetics.<<

Aesthetics here is what we are really discussing. I'm not an engineer, so I don't have the training in the math or physics.

>>Bad example. The CRI of a fluorescent light is usually around 80. The CRI of a candle is 100 by definition, as it is blackbody radiation.<<

My bad.

You are right though in many cases, accurate color reproduction is actually not preffered for aesthetic reasons, for much the same reason as many people prefer EQ settings on music that boost bass or treble, rather than a "flat" EQ.

>>The sun which is by far the most significant light source is a blackbody radiator, and the effect of rayleigh scattering where different amounts of blue light are filtered away at different latitudes and time of day behaves is simialr to changing the color temperature. There are no natural light sources that produce light outside of the blackbody line (ie, the sun or moon never appear purple or green)<<

Yes, there are no natural sources that produce light outside the blackbody line because all natural sources produce light by heating something up (although this is also technically untrue, as certain fish/insect/etc produce light through chemical interaction).

I think what might help would, instead of thinking about it in terms of CRI, might be spectral power distribution. Of course it's relative between sources. If I'm not mistaken, the reason we generally prefer incandscent sources is their almost straight diagonal line through the spectrum, while other sources have "spiked" SPD.

Again I think the important thing to remember is that the aesthetics of it is more important than the science of it. Just my feeling.