CRI of White LEDs

U

UnknownVT

Flashlight Enthusiast
Joined
Dec 27, 2002
Messages
3,672
CRI (Color rendering index) of white LEDs have been low in the region of CRI=70, compared to fluorescent lights even with the average consumer grade CFL from GE manages CRI=82.

Some consider that white LEDs seem better than the low CRI may indicate -
so much so that the CIE (International Commission on Illumination) had a paper out since 2007 suggesting that white LEDs should be assessed visually using the Gretag Macbeth ColorChecker chart.

The Macbeth chart is the photographic industry standard for testing the color accuracy of digital cameras and film - digital camera tests these days use software to analyze the test photo results - some criticize the Macbeth chart for not having a wide enough gamut to use visually - yet that is precisely what the CIE are proposing - to use the Macbeth chart for visual assessment for the CRI of white LEDs.

The abstract of the paper is on the CIE website:
CIE TC 1-62

on searching for a copy of CIE TC 1-62 I also came across these:

Labsphere
" What is the TC 1-62 Colour Rendering of White LEDs?

The present color rendering system gives a poor rating for white LEDs yet the color appearance of white LEDs is better than color rendering index would suggest. This is a potential barrier to introduction of white LEDs into main stream applications so TC 1-62 was established to investigate, by visual experiments, color rendering properties of white LED light sources and to test the applicability of the CIE color rendering index to white LEDs.
"

Then -
pdf VISUAL OBSERVATION OF COLOUR RENDERING
from Colour and Multimedia Laboratory of the University of Veszprém, Hungary
- where they followed the CIE TC 1-62 recommendation and used the Macbeth chart for the experiment
the visual experiments look very interesting -
their comments on the 4000K tests

" As can be seen, there are huge differences in ordering the lamps according to visual or calculated colour difference. The exceptionally good visual performance of the traditional Cool White lamp is hard to understand, one reason might be that most of the test samples contained only very little long wave radiation. It is also of interest that the small peak wavelength difference between the two clusters produced a large difference in Ra. The rank order for the 6500 K series is the same for all four methods of evaluation. "

I am not too sure if they have taken into account the human eye/brain behavior with different light levels -
as shown in the Kruithof curve -
could this have any bearing on their 4000K results?

But best of all -
I have found a copy of the CIE TC 1-62 paper that is currently viewable on line using embedded FlashPlayer:

CIE 177 2007 - CIE TC 1-62 paper
 
C

Chauncey Gardner

Enlightened
Joined
Dec 21, 2009
Messages
397
Location
Near the foot of Mt Shasta, CA
Interesting.

Thanks for putting this up.

I hate flourescent lights and find it hard to believe the CRI would be higher than even a warm white LED when compared.

I have returned a couple of suits after taking them out of the garment bag at home only to find out they were a substantially different color / shade than what they appeared to be in the store.

Just anecdotally it's no contest in what has the more accurate color rendering between the flourescents in my kitchen & warm white of an LF2XT & Quark mini aa (WW). Outside those two lights do a pretty good job of mimicking daylight against the trees & surrounding vegetation.

Flourescents make everthing look rather "flat" to my eye.

I'll check out the links.
 
H

HarryN

Flashlight Enthusiast
Joined
Jan 22, 2004
Messages
3,941
Location
Pleasanton (Bay Area), CA, USA
Hi, it is a fascinating and complicated area, isn't it.

IMHO, the main challenge with the CRI indication is that it is based on a rather poor reference - the incan bulb. As a practical matter, this means the noon day sun is also has a relatively poor CRI rating, in spite of the tendency of our eyes to be designed for its light.

The whole area gets really interesting when you start putting in CCT preferences, which are somewhat genetically influenced. ( for instance, I prefer 5-6000 K, high spectral content light). Obtaining such an LED is nearly impossible, so I have to lower my goal to more like 4000K.

The MacBeth / Munsel system is not ideal, but not all that bad either. Certainly for LED general lighting and flashlights, it is a marginal issue. When it comes to viewing fine photographs, life can be dicey.

I saw a presentation on the possible influence of the MacBeth system, and it basically allowed a lot of discreet wavelengths to beat it, vs a nice continuous spectrum. Is that your interpretation?

BTW - if you don't like tube FL lights, there is a good chance that you are just buying the wrong tube. Buy the ones that are daylight, high CRI and it is entirely a different experience.
 
C

Curious_character

Flashlight Enthusiast
Joined
Nov 10, 2006
Messages
1,211
Below 5000K, the CRI is based on comparison to a black body at the same color temperature. This means that a light with a good 2700K CRI, for example, would look and render colors quite differently than a light with good 4000K CRI. Above 5000K, a "phase" of daylight is used as the reference, so lights considered to have a color temperature in that range look different yet.

Even for a single color temperature, lights with a particular CRI can have an infinite number of spectral shapes and correspondingly different appearance and color rendering. Trying to reduce an entire spectrum to a single number inevitably results in losing a lot of information. It's a mistake to consider CRI an absolute measure of anything, when at best it's a crude grouping based on somewhat arbitrary criteria.

c_c
 
C

Chauncey Gardner

Enlightened
Joined
Dec 21, 2009
Messages
397
Location
Near the foot of Mt Shasta, CA
Hi, it is a fascinating and complicated area, isn't it.

IMHO, the main challenge with the CRI indication is that it is based on a rather poor reference - the incan bulb. As a practical matter, this means the noon day sun is also has a relatively poor CRI rating, in spite of the tendency of our eyes to be designed for its light.

The whole area gets really interesting when you start putting in CCT preferences, which are somewhat genetically influenced. ( for instance, I prefer 5-6000 K, high spectral content light). Obtaining such an LED is nearly impossible, so I have to lower my goal to more like 4000K.

The MacBeth / Munsel system is not ideal, but not all that bad either. Certainly for LED general lighting and flashlights, it is a marginal issue. When it comes to viewing fine photographs, life can be dicey.

I saw a presentation on the possible influence of the MacBeth system, and it basically allowed a lot of discreet wavelengths to beat it, vs a nice continuous spectrum. Is that your interpretation?

BTW - if you don't like tube FL lights, there is a good chance that you are just buying the wrong tube. Buy the ones that are daylight, high CRI and it is entirely a different experience.


I will take a look next time at what is available, maybe different tubes would be an improvement.
I had planned on sticking in a skylight in the kitchen during a planned remodel, but even better rendering from higher CRI tubes would be an improvement.

I had assumed (incorrectly) the higher the CRI the closer to natural sunlight of a given source.

Thanks for getting into a bit more detail, it was an "ah ha!" moment.
I'll finish checking out the linked papers before commenting much further, but when it comes to well done photos I can see the same issues (complaints) that a serious foodie / cook would in his kitchen.

Both cases you are less than pleased with the way colors are represented with what is before you & it is less than satisfying to a discerning eye when looking at subject matter you really enjoy.

Hopefully this isn't viewed as a minor "threadjacking" :) by UnkownVT.
 
Last edited:
Y

Yavox

Enlightened
Joined
Mar 7, 2009
Messages
206
Location
Poland
What is the highest CRI available LED at the moment, suitable for a flashlight which could produce 200 or more OTF lumens and who produces them? Are they more expensive than "normal" LEDs?
 
H

HarryN

Flashlight Enthusiast
Joined
Jan 22, 2004
Messages
3,941
Location
Pleasanton (Bay Area), CA, USA
What is the highest CRI available LED at the moment, suitable for a flashlight which could produce 200 or more OTF lumens and who produces them? Are they more expensive than "normal" LEDs?

That is a surprisingly difficult question. From prior attempts at making flashlights, I use "rated lumens / 2" to get OTF lumens. Using that derating factor, that means we need to look for an LED package with something like 400 lumens, or 4 each x 100 lumens.

In the 4 each x 100 lumens, probably a high CRI Lumileds Rebel.

In the 1 x 400 lumens, perhaps LEDEngin neutral or warm tint.
 
H

HarryN

Flashlight Enthusiast
Joined
Jan 22, 2004
Messages
3,941
Location
Pleasanton (Bay Area), CA, USA
Below 5000K, the CRI is based on comparison to a black body at the same color temperature. This means that a light with a good 2700K CRI, for example, would look and render colors quite differently than a light with good 4000K CRI. Above 5000K, a "phase" of daylight is used as the reference, so lights considered to have a color temperature in that range look different yet.

Even for a single color temperature, lights with a particular CRI can have an infinite number of spectral shapes and correspondingly different appearance and color rendering. Trying to reduce an entire spectrum to a single number inevitably results in losing a lot of information. It's a mistake to consider CRI an absolute measure of anything, when at best it's a crude grouping based on somewhat arbitrary criteria.

c_c

CC - Thanks for that clarification. I will go back and look at those definitions again.

Harry
 
Archie Cruz

Archie Cruz

Banned
Joined
Aug 1, 2007
Messages
205
Much of these new 'findings' and demand for a revamped metric for rating visually pleasing CRI/Colro Temp was pretty much confirmed by research done by the 'quack' Aten Imago, a former CPF member.
Here's the original article Presented to the IEEE, LRI :twothumbs Fully digested by HarryN ;)
--
Theory & Method of LED Color Blending For Enhanced Nocturnal
(Scotopic & Mesotopic) Subject Rendering

Authors:

Äten Imägo
Imägo Metrics - Bethesda, MD - USA

M.P.El-Darwish
Geneva, Switzerland

Version 2.0
October 24, 2005


“Optimizing Human / Technology Interaction Performance”




Abstract

This document lays out the theoretical and research basis for the optimization of LED color blending for the purpose of better aligning the requirements of night adapted human vision (AKA Scotopic/Mesotopic Vision) with the artificial LED source illuminations prevalent in personal illumination devices such as flashlights, bicycle lights and personal reading lamps.

Keywords: Scotopic,Mesotopic,Photopic, Human Factors, Usability, Performance Optimization, Dichromasty, Night Adaptedness, Luxeon, LED, Flashlight, Reading Lamp, Vision, Amber, Straw, White


CONTENTS
EXECUTIVE SUMMARY 3
A BRIEF HISTORICAL OVERVIEW OF THE CASE 4
THEORETICAL BASIS 5
Illumination brightness and contrast as influencers 6
Conditions for optimized supplemental illumination 6
Shortcomings of current LED emitters 8
Methods for creating Amber/Straw/White light from combined beams 8
FUTURE OPTIONS 9
Design community barriers to blended LED theory 9
Current design solutions in development 10
REFERENCES 12



Executive summary
This document sets forth the theory and method associated with the blending the light of two or more high flux LEDs into a combined beam color that has a Color Rendering Index (CRI) as well as wavelength that is harmonized to the requirements of persons performing tasks in low or no ambient light conditions. The need for such a method is due to the lack of a suitable single emitter LED from the available manufactured stock.
Who will benefit from combined color LED arrays?
o Individuals that have benefited from the wavelength of light produced by incandescent lights in the 3,000ºK to 4,000ºK range.
o Individuals seeking variable combination beam colors
o Stakeholder communities that are developing large scale environmental lighting - using LEDs to replace incandescent lights.
o Manufacturers looking for ways to expand the markets for their products.

Benefits
What benefits can be expected?
o Abandonment of incandescent bulbs
o Reduced cost of bulb replacement
o Elimination of color interaction issues resulting from the use of current LEDs
o Introduction of nuanced white combination arrays
o Increased sales opportunities
Productivity Enhancements
What benefit in productivity will result from this process?
o A competitive edge: Manufacturers, Distributors and Customers will flock to this solution, until a single optimum emitter bin color is introduced
o Enables end-users to experience the same advantages from an combined LED array that is currently available to users of incandescent bulbs
o This is a gap filling solution that will bridge a potentially long cycle in the High Flux design process


Improvements Realized in Actual Use
Improvements will be realized as a result of:
o Reduced eye-strain
o Reduced user errors that can slow productivity down
o Increased speed of object identification as well as spatial navigation

The argument for an Amber/Straw/White VS Daylight LED source
A brief historical overview of the case
Taken from a layman's point of view and then underscored by empirical science, the argument for an Amber/Straw/White light for night-adapted illumination, versus a Daylight white is thus:
The current human visual sensory array of Retinal Rods & Cones evolved from a biologically adapted array that deliberately took advantage of the wavelength of light produced by flames. Flames such as those from camp fires, torches and later- oil lamps, candles and gas burning street lamps- invariably produced light that was Amber/Straw/White in wavelength.
This represents an adaptive evolution of the retinal sensory array to coincide with the available wavelength of light. The result is that our visual system became calibrated to optimize night vision to this wavelength-that in layman's terms we have called Amber/Straw/White.
The scientific deconstruction of this evolutionary trend leads to what we currently know about scotopic and mesotopic vision. What is known about scotopic vision is that Retinal cells called Rods are sensitive to very low levels of illumination and are responsible for our ability to see in dim light (scotopic vision). Rods contain a pigment with a maximum sensitivity at about 510 nm, in the green part of the spectrum. The rod pigment is often called visual purple since when it's extracted by chemists in sufficient quantities the pigment has a purple appearance. Scotopic vision is completely lacking in color; a single spectral sensitivity function is color-blind and thus scotopic vision is monochromatic.
The situation becomes more complex when nocturnal vision combines night adapted or scotopic vision, with full color or photopic vision. Mesotopic vision is the scientific term for a combination between photopic vision and scotopic vision in low but not quite dark lighting situations.
For example: the combination of the higher total sensitivity of the rods in the eye for the blue range with the color perception through the cones results in a very strong appearance of blueish colors (eg. flowers) around dawn.

The human eye uses pure scotopic vision in the range below 0.034 cd/m2,and pure photopic vision in the range above 3.4 cd/m2.
In the case of a general subject field that is a field of low contrast darkness that is then pierced by a smaller and localized region of illumination, the visual system works to resolve the brightness and color contrasts by combining the use of Rods and Cones. The resulting compromise illumination color that supports this effort is Amber/Straw/White - light near the 580nm region of the spectrum. This is the most productive supplementary illumination color space for both scotopic and mesotopic vision.
Caveat Emptor - Just as the anthropological record depicts divergent geographical trends in the evolution of human sensory perception, we do not contend that the generalized statements made about human scotopic, mesotopic or photopic vision are absolute or universal. The nomadic peoples of the arctic circle, for example, seem to have evolved an ability to perceive color nocturnal contrast differently from those living on at equatorial latitudes.

Theoretical basis
The foregoing historical and scientific statements pertain to circumstances in which there is no ambient environmental light and the sole source of illumination is a highly localized one, such as might be generated by a flashlight, torch, candle or spotlight. The situation becomes more complex when one considers the following examples:
o Low level and general blueish illumination such as may be produced by the full moon on a clear night
o Low level and general Amber illumination such as might be produced by certain sodium vapor street light arrays
In such cases, the visual system resolves these 'casts' so that they are not perceived as casts at all. When a bright and localized beam is then projected onto a target subject that is surrounded by a diffuse base illumination color, the visual system is shocked and the newly introduced light appears to have a cast that is typically the spectral compliment of the prevailing light. As an example- the introduction of an otherwise neutral and spectrally pure white beam of light from a flashlight into a scene that is bathed is the amber hue of sodium vapor street lights (AKA 'Crime Lights') creates a discontinuity in signals sent by Rods & Cones to the brain- resulting in a perception that the Daylight beam is actually Blue in hue. This discontinuity is greater when the localized beam is smaller. Expressed differently- as the field of supplemental illumination widens and exposes more retinal cells (Cones) to its wavelength, so in turn will the resolution or 'leveling' be more in favor of this localized beam color. In practical terms this means that in a situation of a low level of general illumination plus a localized supplemental light- the supplemental light wavelength should match that of the general illumination. Interestingly, this statement holds true in broad daylight as well!
In cases where the general illumination is so low as to be negligible, then the sole and localized illumination source is optimally in the 580nm region of the spectrum.
Illumination brightness and contrast as influencers
Thus far we've only talked about the interactions of illumination source wavelengths in affecting perceived subject color and ease of night-adapted vision. Another part of the equation relates to the relative contrast and brightness of the scene and the target or 'focus' region within that scene.
I've already touched on the discontinuity that occurs when a bright light of a different wavelength is projected into the scene. I'd like to comment on the added effect of brightness on the visual system's interpretation of colors.
The prevalent thinking has been that the desirable torch or flashlight illumination is one that is closest to neutral daylight in color (approximately 5,400ºK) and brightness is as high as technically possible. Some manufacturers have gone so far as to insist that a bright narrow spotlight that is 5,400ºK in hue is the best solution to most personal illumination challenges. I beg to differ.
Introducing a highly localized (spot) beam of light into a scene bathed in darkness creates havoc for fully night-adapted vision. Choosing to make that beam Daylight balance by default, compounds the shock and reduces night adaptedness which results in increased perceived contrast between the spotlit target and the surrounding scene. This predicament not only makes effective object identification more difficult and less effective- it also 'bleaches' Rods and renders them null- making perception of details in the surrounding scene very difficult if not impossible. In summary, we can make the following statements with regard to localized light brightness, focus and color.

Conditions for optimized supplemental illumination
o In cases where effective perception of shape and form in the prevailing and surrounding darkness is important:
o Supplemental illumination level or brightness should only be just as bright as is necessary in order for the user to perceive sufficient detail in the subject for effective identification
o The transition from focus (spot) to corona, penumbra and then surrounding darkness- should be as gradual as possible
o If there is negligible or no ambient illumination, then a localized torch or flashlight beam color of Amber/Straw/White (580nm) is ideal
o If there is a predominant ambient color, then matching that color will mitigate against a chromatic discontinuity that will likely result in the perception of a cast
o The axis of illumination should be slightly off dead-center from the user's direct line of sight
o In cases where the ambient illumination is broad daylight and the subject is located in a pocket of darkness:
o The supplemental illumination level can be as bright as or slightly brighter than the ambient brightness level
o The transition from focus (spot) to corona, penumbra and then surrounding brightness- may be abrupt
o The supplemental illumination may be daylight or as close a match in color to that of the surrounding ambient scene
o In such cases, there is no advantage in using light that is the same as that used for night and indeed doing so would be counterproductive
o The axis of illumination should be slightly off dead-center from the users direct line of sight
o In cases where the subject area is a significant portion of the overall scene
o The supplemental illumination should cover as much of the target area as possible
o The light should be as diffuse as possible
o The axis of illumination should be slightly off dead-center from the users direct line of sight
o In cases where the entire scene is an enclosure such a small room, passageway or container with reflective walls and a high degree of subject color differentiation is important
o The supplemental illumination is more effective if it's a neutral white or slightly warm white in the 5,200ºK to 5,400ºK range with a Color Rendering Index (CRI) of between 90 and 100
o The transition from focus (spot) to corona, penumbra and then surrounding darkness- should be as gradual as possible
o The light should be as diffuse as possible
o The axis of illumination should be slightly off dead-center from the users direct line of sight
o The supplemental illumination level can be as bright as or slightly brighter than the ambient brightness level

Shortcomings of current LED emitters
Currently, LEDs are available in a wide range of colors. A narrow range of white LEDs have filled market needs for illumination that supplements or supplants the most neutral and highest color rendering Daylight. This is a boon to many industries that depend on repeatable, consistent and long-lived Daylight balanced artificial light.
Unfortunately, the need for an Amber/Straw/White LED has been neglected for lack of a scientific basis for such a need. In addition, research and development efforts have understandably been focused on serving the growing demand for large-scale end uses, rather than the smaller demands of the personal illumination markets.
The nearest to an ideal flashlight LED color for applications such as the outdoors markets, law enforcement, security and general consumer using such lights outdoors (as differentiated from high color differentiation applications), has existed in the form of an Amber Lumileds™ Luxeon™ emitter. A warmish white X bin LuxeonV emitter has also been a near proxy color. The Amber Lumileds™ Luxeon™ emitter is far too 'Orange' in cast to be useful for general applications. As a result, it is only through mixing of light from an Amber Lumileds™ Luxeon™ emitter with light from other White Lumileds™ Luxeon™ emitters that an output brightness and color optimized blended color can be achieved. This is the core of our theory.
Methods for creating Amber/Straw/White light from combined beams
There are two approaches for creating the target Amber/Straw/White beam at 580nm from Luxeon LEDs
Approach 1 - Co-located flashlights
The most obvious approach, and one that I used in initial experiments, simply entailed binding together three or four identical flashlights- one using a Lumileds Amber Luxeon LED and the remainder containing White Luxeon LEDs of the same color or BIN and the same output. The axes of the resulting cluster of lights were adjusted so that all the beams would overlap- creating the single blended beam.
Approach 2 - Co-located emitters
A more efficient and effective approach is to cluster either two, three or four white emitters around a single amber one and to then insert this array into a large flashlight. This latter method is currently popular in the design of products that seek to multiply the generated output from a fixed physical space inside a flashlight's head. Most typically, each emitter is contained in it's own reflector.

The benefits of using Approach 2 include:
o Ability to nuance the relative output between the Amber and White emitters
o Ability to create massive banks of light for large-scale applications
o Ability to replace one or more emitters to create special effects
o Smoother beam mixture
o Resultant beam is more akin to that from an incandescent bulb flashlight
Future Options

In the future, Lumileds may introduce an LED or series of LEDs that emit beam colors that are comparable to those currently produced by incandescent bulbs. In the interim, another evolutionary stage may involve the use of the new K2 series of Luxeon III emitters.

This K2 series of emitters efficiently dissipates heat, a shortcoming of the prior generation of Luxeon III.

One advantage of this to the flashlight design engineer may include the ability to cluster a group of three, four and five emitters is very near proximity and in such a fashion that they may share one reflector. Though not necessarily envisioned for traditional flashlight designs, such arrays could be a boon to designers of illumination for other fixtures such as reading lamps, bicycle lights, camcorder lights, area lights and so forth. This would have the following advantages

o Individual emitter/reflector pairs would not require as precise an alignment to remaining emitter/reflector pairs
o A variable focus relationship between the cluster and single reflector could be established.
Design community barriers to blended LED theory
The design community in any domain of industry is responsible for driving changes that ultimately benefit the user/consumer. Understandably, barriers to change exist within the flashlight community. The following is a list of possible design community barriers and how they may be overcome
o Cultural: The power of mass communication in disseminating information can result in 'mob mentality' or 'group-think', which can in turn result in sub optimal designs being adopted by masses of manufacturers. Once the concept of the LED as a light source for flashlights took hold, it was only a matter of time before cultural norms emerged that would set the stage for 'standards' of excellence in all aspects of LED use in flashlights. Changing the tone of these standards is critical to winning community adoption of a better design solution
o Technical: Design is often constrained by technical limitations. Before the advent of user-centered design that took human factors into considerations, nearly all efforts in the design of LED based flashlights have been concentrated on increasing brightness, achieving as neutral a white beam as possible and optimizing battery usage. By changing the attitude to beam colors to include Amber/Straw/White as an option; previous constraints cease to be an issue- paving the way for wider adoption of the blended beam solution
o Strategic: Manufacturers, while slow to making changes on a large scale- do notice design trends in the user community. Up-scaling adoption at the user level should drive changes up the value chain to the manufacturers- resulting in exponential changes across the entire flashlight landscape.

Current design solutions in development
Based solely on the 'prior art' trial experiments of imago Metrics's test of the multiple-lights with one being a Lumileds Luxeon III Amber, Andrew Wynn Rouse of Wynn Bright flashlights -with a mult-emitter light already in production immediately saw the possibility of replacing one of the four white emitters with an amber emitter which was furnished courtesy of Future Electronics - distribution partners for Lumileds, manufacturers of the Luxeon brand of LED emitters.

The website containing the original pictures is here: http://rouse.com/RT4 (BAM+)

There are BAM+ beam shots here: http://rouse.com/beams
In the current design prototype, we did not have the physical space for a symmetrical array as is specified for the working prototype. Even so, the slight misalignment of the amber emitter in the array turned out not to be conspicuous in real-world use- though for aesthetic reasons we would like to continue development of a symmetrical array such as one amber surrounded by three white emitters (three points of an equilateral triangle) with a fourth, the amber emitter at the center.
One symmetrical design currently in alpha stage, is a flashlight using one amber emitter flanked by two white ones. This design is named the MiniMighty flashlight.
The results of preliminary trials show that a high-power LED array can match the illumination quality of a medium/high power 'hotwire' incandescent lights, with all the advantages of using LEDs and none of the disadvantages associated with incandescent lights.
Prototypes in progress shall serve as proof of concept and as envisioned, will fill a gap currently unfilled by any solution other than filtration.
In addition, future trials will include the ability to variably nuance the relative output of amber versus white emitters, so as to affect a range of color temperatures from 4,000ºK to 3,200ºK in addition to permitting variable brightness levels. Such fixtures could have far ranging applications in interior architectural illuminations, where nuanced illumination hues can have a bearing on the intended mood of the interior design setting as well as in supporting effective task illumination.

References
Ware, C (2000) Information Visualization- Perception for Design 103-149.
Salvendy,G (1997) Handbook of Human Factors and Ergonomics (672,875-876,1737,1699,62-64,669,870,1735,862-863,176,872,59)
Morris,A (1950)Visual acuity at Scotopic Levels of Illumination
______________________________________________
© 2005 Imago Metrics LLC - All Rights Reserved

Finally folks are coming around to understanding that interactions of localized ( ie flashlight, display light, area light) illuminations with ambient illumination, context of use and level of light are the key drivers for determining ideal CRI and not blind engineering combined with manufacturing capacity.
 
U

UnknownVT

Flashlight Enthusiast
Joined
Dec 27, 2002
Messages
3,672
Much of these new 'findings' and demand for a revamped metric for rating visually pleasing CRI/Colro Temp was pretty much confirmed by research done by the 'quack' Aten Imago, a former CPF member.
Here's the original article Presented to the IEEE, LRI :twothumbs Fully digested by HarryN ;)

Yes, well, there is more than one mechanism at play here -
first I do not claim to be any expert on this and have found/learnt some of this as I went along.

As mentioned CRI is very dependent on CCT (Correlated Color Temperature)
eg: it may surprise many to learn that an incandescent/tungsten bulb rates CRI=100 (perfect) by definition, yet anyone who has taken a photo under tungsten lighting with daylight white balance (or daylight slide film) will know immediately that that the light is very yellow/amber - and will have difficulties seeing yellow on white or distinguishing navy from black - so how can it possibly be CRI=100 perfect?

It is the way that CRI (color rendering index) is defined comparing to a reference black-body of the same color temperature - so an incandescent bulb is its own reference by the definition...... that's why it rates CRI=100.

Whereas a 2700K CFL (compact fluorescent spiral) may only rate CRI=82 on average - but may actually contain more blue component so that it may be a tiny bit easier to see the difference between navy and black - but its CRI rates lower - because it is not exactly like a true tungsten bulb..... how's that for circular logic?

A good white LED rates about CRI=70 and not much higher - and since there is a huge push to use LEDs for general lighting and a CRI=70 is very poor -
one might suspect that the CIE (an accredited international body on illumination) proposal to use visual assessment for CRI of white LEDs might just be yielding a little to the pressure to use white LEDs for general lighting?
eg: we might find that a white LED formerly rated at CRI=70 all of a sudden under the new CRI measurement guidelines rate in the 80's?
So has that white LED improved?
or just made more acceptable so that big corporations will start to use them to save energy......
don't get me wrong saving energy is obviously a very good thing - but will our eyes appreciate this?

Then there is this whole question of how our eyes/brain adapt to lighting levels - empirically our eyes follow the
Kruithof curve
that's why I questioned whether that experiment using the new CIE guidelines for CRI in the pdf VISUAL OBSERVATION OF COLOUR RENDERING (from some of the same authors who originated the CIE TC 1-62 paper)
may have missed the eyes' reaction to different levels of lighting.

Although noonday sunny daylight might be held as the ideal lighting - our eyes do not see that as "white" under all lighting levels - at lower lighting levels most people actually will persist in seeing 2700degK as "white" (please see the Kruithof curve for an explanation)

One should read this paper:

The Color of White

paper published by the WAAC - Western Association for Art Conservation -
specifically on illumination for displaying art/paintings -
their findings fit well in the Kruithof curve.
 
H

HarryN

Flashlight Enthusiast
Joined
Jan 22, 2004
Messages
3,941
Location
Pleasanton (Bay Area), CA, USA
Much of these new 'findings' and demand for a revamped metric for rating visually pleasing CRI/Colro Temp was pretty much confirmed by research done by the 'quack' Aten Imago, a former CPF member.
Here's the original article Presented to the IEEE, LRI :twothumbs Fully digested by HarryN ;)
--
Theory & Method of LED Color Blending For Enhanced Nocturnal
(Scotopic & Mesotopic) Subject Rendering

Authors:

Äten Imägo
Imägo Metrics - Bethesda, MD - USA

M.P.El-Darwish

Hi Archie, thanks for including the info.

Aten's presentation and book are more or less a rehash of what was well known and widely published in the 5 years prior. I don't fault him for that, as it happens commonly at many technical conferences.

What I do credit him and Andrew for is taking the effort to really build a light to test the work with real life parts.

More or less, they pointed out the obvious - the phase diagram we all use to "trick our eyes into seeing a color using mixing" only tricks it some of the time. A house is not complete by just putting up the frame and exterior, and lighting is not complete by just throwing up 2 or 3 colors into a phase diagram.

The challenge that they faced then in actually using the technique still exists today - no one makes an LED in the desired wavelength that they qualitatively described as "straw" color.

The reason is that most Red and amber leds are made using AlInGaP technology, which is great for red, and just barely works for amber. For blue and green, the common technology uses InGaN, which just barely works down to a yellow green. This means there is not a commercially viable way to make yellow directly.

The solution is now getting closer with the intro of Lumileds "amber" made from phos converted blue. This means that making "yellow" in a similar approach is no longer a technical problem, just a market size question.
 
B

blasterman

Flashlight Enthusiast
Joined
Jul 17, 2008
Messages
1,802
The present color rendering system gives a poor rating for white LEDs yet the color appearance of white LEDs is better than color rendering index would suggest.

I have two really good opinions on this.

First, comparative CRI light sources like fluorescent tubes have horridly spikey spectrums. Cool White LEDs spike on blue, but are fairly smooth after that. Low CRI fluorescent tubes however have multiple spikes and valleys which tend to bother our eyes. Hence why cool-white LEDs (good ones in low CCT bins) appear to have better color rendition than their published CRI would indicate.

Next, the current CRI standard is near useless, as evidenced by the fact Philips is claiming it gets 98 CRI from a rather cheap fluorescent tube. If the sun is a 100 CRI.....whatever.

Regardless of how cool-white LEDs might be considered better than other mediocre CRI light sources, I still consider them totally inadequate for interior lighting.
 
J

jtr1962

Flashaholic
Joined
Nov 22, 2003
Messages
6,511
Location
Flushing, NY
Although noonday sunny daylight might be held as the ideal lighting - our eyes do not see that as "white" under all lighting levels - at lower lighting levels most people actually will persist in seeing 2700degK as "white" (please see the Kruithof curve for an explanation)
I've always had trouble wrapping my head around this because to me 2700K looks distinctly yellow and horrid regardless of lighting level. I could be in a room for hours but not acclimate enough so that 2700K appears white. I recently made a reading light so I can read in bed using 5 SSC high-CRI neutral whites. While the color rendering is great, to me the tint still looks more yellow than I'd like, and this is 4000K. Personally, I need to get to around 5000K before a light looks white to me, regardless of color level.

A good white LED rates about CRI=70 and not much higher - and since there is a huge push to use LEDs for general lighting and a CRI=70 is very poor -
one might suspect that the CIE (an accredited international body on illumination) proposal to use visual assessment for CRI of white LEDs might just be yielding a little to the pressure to use white LEDs for general lighting?
eg: we might find that a white LED formerly rated at CRI=70 all of a sudden under the new CRI measurement guidelines rate in the 80's?
Not at all true that cool white LEDs only have a CRI of 70. One of the pdfs you linked to had a blue plus phosphor LED rated at 80. Actually, high 70s to 80 seems to be where most phosphor whites fall, provided the CCT is reasonable ( i.e. 5000K to 6500K ). The 8000K to 10000K ones are those which might have a CRI of 70, but no sane person would consider using them for general lighting. Subjectively, cool white LED always seems better than most fluorescent of similar CRI, likely because the spectrum is more or less continuous. This is telling me we need a new measurement of color quality simply because what our eyes are telling us is way different from what the CRI says. Right now I'm lighting my bedroom with some LED screw-base lamps I had received for testing. One was retrofitted with cool-white Rebels, another with Cree XP-Gs. The 6500K CCT seems just fine, and a lot more bearably than a 6500K CFL. Truth is, I probably wouldn't want much lower CCT using LEDs even though I lean towards 5000K fluorescents. The LEDs seem more natural than the CCT and CRI numbers might indicate.
 
McGizmo

McGizmo

Flashaholic
Joined
May 1, 2002
Messages
17,282
Location
Maui
I am no expert on CRI and I can only hope not to make false or misleading comments regarding CRI. The following comments are what I have gleaned and believe to be true; corrections are welcome.

When a CRI number is given or stated it is a number for the CRIa which is an average of a dozen or so CRI indexes measured at specific spectral positions. It is also based on a specific CCT. In simple terms, CRI is used to evaluate how well an artificial light source renders colors compared to a black body source of light at the same CCT. An incandescent source is essentially the standard against which other sources are compared.

I am in the camp of folks who believe that sunlight should be the standard against which other lights sources are measured and although direct sunlight may have a spectral distribution of a black body source of a specific CCT, indirect sunlight often does not.

I think there are specific spectral areas of output from the various artificial sources of light that can be identified as the problem areas. These would be the regions in the spectrum where a source either has excess or limited output, relative to the whole. If some single color rendering index number is to be used as a measure of a lights ability to render color in general then it might be more illuminating to base this average on the problem indexes and not dilute it by averaging in other indexes which are typically adequate. In other words, a white LED for instance might be measured in its ability to render blue and red and green (in the spectral bands where we see the peaks and troughs).

One white LED may have a CRI of 87 and another of 94. The immediate assumption is that the LED having a CRI of 94 is a better source for color rendition. Well what if the one having CRI of 94 has a CCT of 2200 and the one with CRI of 87 is at a CCT of 5000. What if the one with CRI of 87 has all of its index numbers being 87 where as the one with a CRI of 94 has some indexes at 99 and a couple in the low 80's? I personally would prefer the LED having the CRI of 87 in this example and this preference is based on presumed color rendering even though it doesn't score as high as the other LED. But it is a personal preference.

Color rendering can be a very significant aspect in artificial illumination and being somehow able to quantify a light sources ability to render color certainly has merit. I would think that if a standard is to be established then this standard should be as ideal as possible and widely accepted as a standard. I would also suspect that CCT should be integral and not set aside as a qualifier but perhaps I am wrong. Perhaps some measure of significant deviation from such a standard would tell us more than the present CRIa index. :shrug:

In fixed lighting, the subject of illumination and its reflective nature is just as important as the source of illumination and the source should be selected with this in mind. In portable illumination, the nature of the subject may or may not be known in advance and this can be significant if choices in the illumination device are available. To what ever extent a measure or index of color rendering can aid us in making the best choice for the illumination task at hand such a measure or index has merit.
 
B

blasterman

Flashlight Enthusiast
Joined
Jul 17, 2008
Messages
1,802
While the color rendering is great, to me the tint still looks more yellow than I'd like, and this is 4000K

Ditto on better cool white LEDs being superior than cheap 'daylight' CFLs.

One thing I've noticed with commercial quality 4100k T8 tubes of good subjective color is they cheat a little and have just a smudge of pink in them. However, light sources that claim to be neutral and have a bit of yellow or green in them tend not to look as good, even though the CRI and even CCT might be identical. This is especially true of neutral white LEDs where even a slight variation to the yellow or green side causes complaints.

I've noticed that my latest batches of Bridgelux LEDs aren't nearly as warm as prior ones, likely in an attempt to get efficiency up. However, the character is distinctly 'rosy' more than yellow, and this is certainly by intent given their focus on commercial use.
 
Y

Yavox

Enlightened
Joined
Mar 7, 2009
Messages
206
Location
Poland
Has anyone tried to bulild a multi-LED flashlight that would give extactly the same spectrum as an incan? I mean putting a few high-CRI LEDs inside to act as a main light source and supplement them by a few auxilaty LEDs which could fill the gap in the spectrum (like a red LED) or make the spectral curve more smooth. This would probably require not only boosting the red frequencies but also absorbing some blues, but we could provide an optic to each led and paint some of the lens with a glass paint a bit, to filter out what is not needed.

With some luck, maybe even a quad P60 dropin would do the job? The main problem would probably be the driver, because if we would like to have 3 or 5 modes, those different colored LEDs should probably be powered in non-linear way (related to each other) in order to keep the spectral curve the same and only change the brightness, not the frequency distribution.

I am not sure if I can describe this in english clear enough, but I hope it is more or less understandable what I mean :)
 
Greta

Greta

Administrator
Staff member
Joined
Apr 8, 2002
Messages
16,008
Location
Arizona
Much of these new 'findings' and demand for a revamped metric for rating visually pleasing CRI/Colro Temp was pretty much confirmed by research done by the 'quack' Aten Imago, a former CPF member.

Is that disassociative identity disorder kicking in again, Aten? Don't play people for fools here. You've been allowed to post and be part of this community again because you have refrained from your old assinine ways. I suggest you get yourself back under the radar... :ironic:
 
C

Curious_character

Flashlight Enthusiast
Joined
Nov 10, 2006
Messages
1,211
I recently added onto my house, and lit it with Cree LR6C recessed lights. They consist of several white LEDs, a number of red ones, and I believe a few yellow ones as well to get a CRI of 90. (The LR6 is "warm white" 2700K and the LR6C is "bright white" 3500K -- both are 90 CRI.) The thing that sold me is that the light includes a color sensor and feedback system to keep the color constant by automatically adjusting the colored LEDs as necessary. Without this feedback system, the color would change with time as LEDs age at different rates. I like the color (and efficiency -- the light output of a 60 watt incandescent for about 10 watts), although when compared to CFL, the lights look a bit on the purplish side. In various other places in the house I have some cool white fluorescent tubes and warm and bright CFLs, and get some really interesting effects going from one environment to another, or looking from one lit area into another. For example, the cool white fluorescents look greenish when viewed from an area lit with the LED lights, and yellow walls in the LED lit area look peach colored when viewed from the fluorescent lit area. A new bathroom has the LED lights, and I still have CFL just outside. Looking from the bathroom, the floor seems to change color at the point where the incident light changes.

The problem is, of course, that all the spectra are different, even when light types have identical color temperature and CRI ratings. And that will inevitably lead to different color renditions and different light color appearances. All equal-temperature heated filaments have very nearly the same spectrum, so defining a spectrum required only the single number of color temperature. With the variety of spectral shapes we get from the various types of lights and phosphors, it's now much more complicated.

I had another thought that further illustrates a shortcoming of CRI as a defining parameter -- Do you have an incandescent light hooked to a dimmer? Turn the dimmer all the way up. The full brightness incandescent light has a CRI of 100. Now turn the dimmer down until the lights are reddish. What's the CRI now? Answer: 100. How about when they're turned down to make a dim, really red light? Still 100. So in what way is that really red light with a CRI of 100 better than an LED with CRI of 70? Which would you prefer to light things up with? Which do you think shows colors more accurately?

c_c
 
U

UnknownVT

Flashlight Enthusiast
Joined
Dec 27, 2002
Messages
3,672
and get some really interesting effects going from one environment to another, or looking from one lit area into another. For example, the cool white fluorescents look greenish when viewed from an area lit with the LED lights, and yellow walls in the LED lit area look peach colored when viewed from the fluorescent lit area. A new bathroom has the LED lights, and I still have CFL just outside. Looking from the bathroom, the floor seems to change color at the point where the incident light changes.

I had another thought that further illustrates a shortcoming of CRI as a defining parameter -- Do you have an incandescent light hooked to a dimmer? Turn the dimmer all the way up. The full brightness incandescent light has a CRI of 100. Now turn the dimmer down until the lights are reddish. What's the CRI now? Answer: 100. How about when they're turned down to make a dim, really red light? Still 100. So in what way is that really red light with a CRI of 100 better than an LED with CRI of 70? Which would you prefer to light things up with? Which do you think shows colors more accurately?

Interesting input.
I think our eyes tend to be able to distinguish differences well -
whereas without anything to compare to,
I think our eyes/brain adapt to "accept" or optimize to that ambient lighting.

CRI is mostly misunderstood - and because it is quantitative - ie: there is a hard figure attach to it - many people just take it as some kind of gospel - ie: CRI=100 has to reproduce colors more accurately than CRI=82.

Of course without the corresponding CCT (color temperature) this is kind of meaningless. Just to cite the same example already given -
an ordinary tungsten household bulb is rated CRI=100 by definition - yet most of us know that we have a really hard time being able to see yellow on white or distinguishing navy from black.
Whereas even a cheapo daylight 6500K CFL with a CRI=82 does not have that problem.

However that says nothing about which is more pleasant to the eye -
at lower levels of illumination the daylight 6500K CFL would look too cool-blue and probably pretty unpleasant - whereas 2700K tungsten (or even CFL) actually looks quite nice - since our eyes/brain are conditioned through both physiology and evolution to see well under that color temperature.

So CRI is absolutely linked with the color temperature.

BUT even color temperature may be misunderstood - we almost always assume that color temperature is on the Planckian locus - ie: that of a radiating blackbody. BUT what happens when the light source falls off that locus?

Classic example is "white" light that's made up of just yellow and blue LEDs - its spectrum is obviously mainly the two peaks in yellow and blue - but to out eyes that light is "white" - it is only when we look at specific colors that we realize there is something amiss. This also applies to using RGB LEDs to make white - with spectrum of Red Green and Blue peaks - the light again does look "white" but specific colors may have some difficulties.

These may seem extreme examples but this is one area where CRI becomes very applicable - as "white" that is daylight has CRI=100, yellow/blue LEDs would have a very low CRI and similarly RGB LEDs would not have a good CRI either. (for real practical RGB application and some difficulties please see: LED Stage lighting )

This also shows that CRI is not any real indication of how pleasant a light source is to our eyes.

Our eyes/brain does adapt to different light levels -
that is why I questioned the newer recommended visual assessment of CRI as used in
pdf VISUAL OBSERVATION OF COLOUR RENDERING
from Colour and Multimedia Laboratory of the University of Veszprém, Hungary
(by some of the same authors of the CIE TC 1-62 paper that recommends the newer visual assessment for CRI or white LEDs)
They were surprised to find that the "cool white" lighting at 4000K did well -
because of that I am not sure if they have taken into account the human eye/brain behavior that follows the Kruithof curve -
the illumination level and the surroundings are important - otherwise the results depending on human observation/comparison would vary wildly depending on the environment and surrounding ambient lighting -
eg: experiment illuminated at 10,000lux, 1,000lux and 100lux would yield very different results -
also conducted during daylight hours in a natural daylight lit lab vs one held in the dark, or under tungsten lighting......
there are just too many variables, and I am not too sure if the eye/brain behavior has been rigorously taken into account, to have any new CRI measurements, based on human visual assessment, to be truly meaningful

I very strongly suggest please reading these two links:

The Color of White

and

Kruithof curve
 
Last edited:

Similar threads

J
Replies
4
Views
4K
bshanahan14rulz
bshanahan14rulz
W
Replies
2
Views
2K
windmill
W
Anders Hoveland
Replies
4
Views
2K
Anders Hoveland
Anders Hoveland
Top