Flashlight Tint/Color Reference

ElectronGuru

Flashaholic
Joined
Aug 18, 2007
Messages
6,055
Location
Oregon
Flashlight Tint/Color Reference


Introduction
I've been working on a new technique, photographing the color (aka tint) of light, as projected from flashlights and other sources. These are photos, not of flashlights, beamshots or bounceshots, but of the color of the actual light being projected.

I want to make choosing a flashlight tint as easy as choosing a color at the paint store. View the profile for a given bulb or emitter and instantly know its color. Then compare 2+ profiles and see (first hand) which is warmer or cooler and by how much. The details are quite abstract, but the results are not. Read on or jump to the Use section to get started. And see post 30 if you start to feel :thinking:



Theory
Have a look at this graphic:

Kelvincolortemp.png


This is the spectrum of natural color (see note 3), from red to blue, by which we see. Not the color of objects but the color of the light that illuminates objects. Look at a red chair outside at noon, then put the same chair inside, next to a 100 watt bulb. It will look that much more red. The closer to "white" a given light source, the easier it is to see the "true" colors in the space we are viewing.

LEDs tend toward blue. The older/cheaper an LED is, the bluer it tends to be. The holy grail of LEDs is to be as white as possible. Incans tend toward yellow. The lower the voltage, the yellower they tend to be. The holy grail of incans is to be as white as possible. Looking at the complete spectrum above, there is but one point in the middle, one white that both technologies strive for - balanced white.

Have a look at the CCT column on the Welch Allyn bulb site:
http://www.walamp.com/lpd/webstore/searchbylamp.tpl?SKU=17345948104771&cart=12413984541267753

CCT (Correlated Color Temperature) is a specific measurement form of Kelvin, in this case, the color temperature of a specific bulb at the standard voltage. See how most of the CCT values are in the 2500-3500 range? Notice how this correlates to the yellow section of the spectrum above. Driven at their spec (not overdriven) voltage, this is the color each bulb is expected to produce.



Practice
This isn't like the boiling point of water at sea level or the atomic weight of X element. There is no true white, measured or otherwise. What looks white to people only is, because our eyes (genetically speaking) have been looking at light from the sun for many many years. As much as we try to control this with fancy measures, color comes down to simple perception. And what looks white to one person may not look white to another. So selecting a single point that represents all light sources with a single common scale is tough.

Have a look at the grid of samples below. Each column is one of three flashlights, selected for their nearly neutral, slightly blue, and slightly yellow properties. Each row is a different Kelvin setting on the camera, the color the camera is expecting to see. The result is a tint, representing of how far off (away) from that setting a light source's color is. Working from my computer screen (and eyes), 4300K is the most representative of what each flashlight is projecting (see Limitations for more information):


2vi4fmr.jpg


And here are the same three lights, with straight 4300K calibration and photoshop'd exaggerations on either side:


j9srr7.jpg



Method
In a dark room, shine the flashlight at an (ExpoDisc covered) lens and move it about, adjusting to completely bathe the Disc in even light while being as intense as possible, then take the picture. 1-3 feet away is best, depending on the width of the beam/hotspot being projected. Light blurs to a single color (because of the ExpoDisc) on the way into the camera. F11 is small enough to reduce blotching while still 'fast' enough for easy exposure lengths. Shutter speed is light metered (Av - aperture priority) so profiles [generally] have the same brightness (dim lights appear equal to brighter ones), leaving color as the remaining variable.



Limitations
  • Measurements are based on samples. Incans may change color with age and certainly change color with voltage. LEDs, while more consistent during operation, vary more from unit to unit. The make/model I have may be newer/older than what you have or are considering. Even two LEDs out of the same batch have differences. Binning/sorting minimizes this, but plus or minus 100 degrees K "sees" even tiny variations. Companies improve their products as supplies improve and costs go down. Some of my samples are new, others are 3-4 years old.
  • 4300K is the setting at which these photographs most closely resemble my real world "this is what I'm seeing" experience, but there are no absolutes. 4300K works with my camera and my computer and this file format (JPEG). Were this on a different system (with a different color space like AdobeRGB) or printed on paper, there would likely be a different "standard." And even then, there are differences in monitor (CRT vs LCD), its calibration (or miscalibration), platform (Mac vs PC), software, and of course us. I double checked my findings with Mrs Guru, who has extraordinary color perception, but even if all your equipment was the same as all my equipment, you may still see differently (and not be wrong).
  • Knowing the simple color of a light does not necessarily predict its ability to render colors. Just because a light source is balanced (ie white), does not mean it also has blue, green, and red components ample enough to reveal colors as well as the sun. The sun produces a broad mixture of energy, including colors (and of those, only some are visible). Some of the efficiency offered by LEDs, for example, comes from the narrow range of colors they produce (not wasting energy on less useful frequencies).


Notes
  1. All CPF members may use any profile anywhere within CPF and CPF Marketplace. Please contact me directly before using profiles elsewhere.
  2. K stands for Kelvin: The color emitted by a theoretical piece of black material at [given] degrees Kelvin (its temperature). 4300K = 7280.6 ºF. Make iron hot enough and it will glow blue!
  3. Developed to measure the color of hot, glowing objects, the Kelvin scale is very poor at describing the variety of colors now possible with LEDs. There is, for example, no green on the Kelvin scale. Here I use it only to measure the degree/strength of a given color. See post 8 for more information.
  4. All colors presented are darker than actual to enhance viewability, so perfect gray here represents perfect white.
  5. Colors are very much contextual. The background color your CPF profile is set to may effect how the colors appear. Framed in light blue, for example, profiles may appear less blue than they actually are. For highest accuracy, zoom in on a profile or open it up in its own window (right/command-click).


Extra Reading




Use
Here are several familiar (non flashlight) examples to get you started and refer back to later. As you study them (and the profiles below), remember that this is counter intuitive to color rendering ability. Profiles show the 'distance' from white, so the less tinted the source of your light (the closer it is to neutral gray here), the more 'real' the colors of what you are viewing will appear.



2uot3s8.jpg



wm0y6p.jpg



2mwe7lz.jpg



14wx3ky.jpg



xnvgg5.jpg



213rt6h.jpg



2ymeg76.jpg





:devil:​
 
Last edited:

ElectronGuru

Flashaholic
Joined
Aug 18, 2007
Messages
6,055
Location
Oregon
Re: Flashlight Color Temperature Reference


Bulb & Emitter Profiles



2wcr0uf.jpg



15ey92q.jpg



qp5d1d.jpg



fbmkd1.jpg



2l52fr.jpg



10cvjut.jpg



2zsyafo.jpg



2utnfw9.jpg



s1sqy9.jpg



14wx3ky.jpg



2hn3ml3.jpg



166elag.jpg



2mnejie.jpg



2mqkzu9.jpg



6nwhlc.jpg



2l1ffc.jpg



30lc3u0.jpg



2wguah2.jpg



ixb4au.jpg



optwjq.jpg



5ml7yw.jpg



2ugf59u.jpg



fmi7ih.jpg



i4qzkn.jpg






:devil:​
 
Last edited:

Tekno_Cowboy

Flashlight Enthusiast
Joined
Apr 2, 2008
Messages
1,680
Location
Minnesota
Re: Flashlight Color Temperature Reference

Nice Idea :twothumbs:

I hope this idea catches on. I would be glad to help you add more lights, as I get quite a few through here.

What would be great is to get someone with a ton of well-sorted led's of the most common bins used in lights, and get them to photo the led's at a given current. :thumbsup:
 

greenLED

Flashaholic
Joined
Mar 26, 2004
Messages
13,263
Location
La Tiquicia
Re: Flashlight Color Temperature Reference

Very interesting concept, GE!
Love seeing this type of innovative ideas on CPF. :twothumbs

:thinking: Let me get this straight, these are pics looking into the beam of a light (through the ExpoDisc), right?
 
Last edited:

McGizmo

Flashaholic
Joined
May 1, 2002
Messages
17,290
Location
Maui
Re: Flashlight Color Temperature Reference

What is the color temperature of a green tinted white LED? When a LED is far enough off the planck curve, do you think CCT has much merit?

Is Green LED warm or cool? (I mean the guy I posted after :nana:)
 

ElectronGuru

Flashaholic
Joined
Aug 18, 2007
Messages
6,055
Location
Oregon
Re: Flashlight Color Temperature Reference

Thanks Guys,

I would be glad to help you add more lights, as I get quite a few through here.

I do only have so many lights and I'd happily add other collections to the lineup. I personally would love adding Ra Lights and more Malkoff's. And of course McGizmo!


Let me get this straight, these are pics looking into the beam of a light (through the ExpoDisc), right?

Correct. The ExpoDisc sits on the front of the lens, and I aim the beam directly through it, into the lens, at the camera sensor. The Disc lets cameras record the ambient color temperature of a room/place for the purposes of calibrating the following shot, by neutralizing/canceling the tint from light sources within that room. Half way through the process, the camera is left with a (normally discarded) image of that tint. This is just a focus on and enhancement of that step.


What is the color temperature of a green tinted white LED? When a LED is far enough off the planck curve, do you think CCT has much merit?

Great question.

CCT (K) is linear (two dimensions). Every color measure I can think of (RGB, CMYK, color wheels, etc) is 3+ dimensions, and with good reason. While specific colors translate directly into accurate K values, the opposite is not true - K values do not contain and cannot be translated into specific colors. Here is the greenest light source I have, a simple night light:


2jfljzm.jpg


Notice how its K is less than the bluest LED above? I don't have cool matching temp examples, but imagine a pair of lights with the same high K value, one with a green tint and one with a blue tint. And another pair with the same low K value, but one orange and one red. With color, the usefulness of K is as a measure of distance, in this case, showing how far away from neutral (white/gray) a given light source is. In this guide, the color sample itself is the primary tool, the K value only measures a given color's strength (degree of tintedness).

Edit: Following your question, I've been looking deeper into color rendition. There is some funny thinking going on here. Here's a paragraph from a wonderful article they sheds a bit o light on why:

Typical phosphor-based white LEDs have color rendering index (CRI) values comparable to discharge lamps (fluorescent and HID lamps). Many people mistakenly believe that a high CRI means high color rendering properties. Actually, the CRI is merely an index of how similar a light source makes colors appear in comparison to a reference source such as incandescent. That's why an incandescent lamp has a CRI near 100. Recent studies show that mixed-color "white" LED systems with a CRI in the 20s can result in higher color preferences than systems with a CRI in the 90s.

So a high quality LED is 90% as good as an incan (CRI=90) at rendering color, but how good is the incan and how do you measure that? To your original question, the only reason CCT made sense as a description of incan color is because incans only have one color question, "how yellow is it?" Using CCT to describe the variety and complexity of possible LED colors makes only slightly more sense as bringing black and white film to the Rose Parade.
 
Last edited:

MichaelW

Flashlight Enthusiast
Joined
Dec 8, 2007
Messages
1,788
Location
USA
Re: Flashlight Color Temperature Reference

Great work.

I was hoping to bring peace in out time to the LED & incan world.
LEDs should adopt roughly three 'colors'
3000K warm white (hey that looks like a mini-mag)
4500K neutral white (it looks like malkoff will have to change that w to an n)
6000K cool white


any chance to verify HIR & HID automotive headlights?
 
Last edited:

McGizmo

Flashaholic
Joined
May 1, 2002
Messages
17,290
Location
Maui
Re: Flashlight Color Temperature Reference

Electron Guru,

Since the LED is not a black body radiator, the further from the Planck curve its tint lies, the further from any relevance assigning a CCT to it would have. As you have stated, CRI is also based on black body radiators and keyed to a specific Kelvin temperature. From what I can gather, the CRI number that is typically given is actually an average of perhaps a dozen or more CRI numbers based on the rendering ability of different bands of light. It is posible to have one LED source that has a CRI of 90 that is based on a CCT of 3500k and it is an average of some CRI numbers that are 100 and some that are 80 let's say. You could have another LED that also has a CRI of 90 but it is based on 5500k and all of its constituent CRI values are all 90. The latter LED would likely be much much more acceptable to folks in most color rendering applications and likely appear closer to white when viewing a white surface with it.

We experience various tints with these LED sources and of course we want a means of quantifying and qualifying them and sorting them. Unfortunately they vary in a "n" dimensional matrix and we seek a one dimensional assignment of number for them or in some cases a single letter designation. Unfortunately their unique position may be so far from the line we want to place them on that we are wasting our time.

If you look at data sheets from the LED manufacturers you will often find much more information about the LED's and how the bins are defined. It is clear that you can have obvious variations in tint and color rendering ability from samples within the same bin. The x,y coordinates of the color measure allow for a LED to be plotted with some information beyond the planck curve but even then, two LED's can "earn" the same x/y values yet still be quite different in terms of their full spectrum output and distribution.

It has been my experience that knowing both the CRI and the CCT it is based on has given me the best "feel" for what to expect from a LED in terms of color when I use the LED to illuminate objects of color.

Because the industry has been involved in incandescent for most of its history, the inclination is to valuate and qualify the LED using the same means and measures. Things are changing but it seems that the marketing arms of the luminaire manufacturers are more than willing to pass on misleading and insufficient information in terms of identifying their wares. What is a warm white LED? What is a cool white LED? The assumption is based on off white tints ranging from yellow to blue as you get cooler yet the reality may involve tints you have never experienced in an incandescent source?!?!

I am no expert by any stretch of the term and I don't claim to even have a good working understanding of color and the spectrums of light but with the aid of a spectrometer and integrating sphere, I have seen enough to appreciate the overall complexity of the beast and realize that much is lost in any attempt to simplify these light sources by trying to treat them as a black body radiator. Apples VS Oranges or perhaps even more appropriate Cats VS Dogs.

I would point out that an integrating sphere integrates while a spectrometer differentiates. Some numbers assigned are based on averaging and integrating whereas other characteristics require a two or three or more dimensional matrix to be well defined and identified.

To quantify a LED source even by CCT and CRI might be no more illusive than qualifying members of this forum by their weight and age. :shrug:

Granted we are much more complicated than a LED but the LED has more dimensions than some of us give it credit for.

4500k is one number and just what does this tell you? In my youth, there was a magic set of three numbers a male might hear on occasion; 36 24 36. Well that was three times as much information but beyond being recognizable to some as to what these numbers related to, it was still insignificant information on any real meaningful level.

I feel I have thrown a monkey wrench in your endeavor here but I think you can understand my confusion and concern or at least relate to it.
 

McGizmo

Flashaholic
Joined
May 1, 2002
Messages
17,290
Location
Maui
Re: Flashlight Color Temperature Reference

Electron Guru,
I want to quote here, some of what I am including in an e-mail response I am sending to you:

"There are members who speak with authority on the lights and yet it is clear to me that they don't really understand the difference between lux and flux. Once you enter the realm of color, it gets way way more complicated!!

On a white wall, I admit that I have my own color temperature preference (based on black body spectrum) but I understand that a LED source of that color temp may or may not be pleasing to my eye. Once you get into illuminating a full spectrum landscape, I need a whole bunch more parameters to identify in describing my preferred light source and much of it is a guess based on limited understanding and not empirical or measured experience.

In the real world, I have found that in many cases, flux trumps CCT as well as CRI since sufficient light, across the spectrum may be more important than the actual distribution or spectrum itself. Contrast is also critical and it may or may not have color in its makeup; the distribution of intensity (lux) alone may be at issue.

Ultimately, my need to understand the light source is back seat to actually using the light source. If I can get the source into the ballpark then I can go about viewing the ball park."

I admit that these comments are just my take on things and I claim no expertise nor do I consider myself appointed or qualified to speak for anyone else. Whatever floats your boat.
 

Gene43

Enlightened
Joined
Sep 2, 2006
Messages
835
Location
South Alabama
Re: Flashlight Color Temperature Reference

A real world frustration often encountered is the described "warm" LED's which in addition to being "warm" are also an unexpected green. I've found this to be a common fault with incan "replacement" bulbs intended for household use.
 

ElectronGuru

Flashaholic
Joined
Aug 18, 2007
Messages
6,055
Location
Oregon
Re: Flashlight Color Temperature Reference

Finally! Proof that my conscience is not black.
Thought you might notice that :thumbsup:


I was hoping to bring peace in our time to the LED & incan world.

They're both so much fun, I would be seriously :mecry: if I had to choose just one. Even forgetting results for a moment, it would be half as much cool stuff to learn and get excited about. Perhaps we can at least calm the waters here.


LEDs should adopt roughly three 'colors'
3000K warm white (hey that looks like a mini-mag)
4500K neutral white (it looks like malkoff will have to change that w to an n)
6000K cool white

With a few exceptions, these labels are coming out of the marketing departments of LED makers and seem to be relative to older LEDs, not the spectrum as a whole. Malkoff and others probably continue the convention to avoid confusion (rightly so) but some meaning is lost in the process. I'm hoping the presence of a new complete scale will help clear some of our confusion that results, seeing how everything looks relative to everything else.

+/- 1500 is pretty wide for an ideal. I'm hoping to narrow it down to +/- 1000, something closer to 3500-5500. Part of the challenge is that there seems to be an advantage to a slightly yellow tint. I'm not sure there is a similar benefit to a slightly blue one.


any chance to verify HIR & HID automotive headlights?

I'll see what I can do.
 
Last edited:

Tekno_Cowboy

Flashlight Enthusiast
Joined
Apr 2, 2008
Messages
1,680
Location
Minnesota
Re: Flashlight Color Temperature Reference

McGizmo:

While I agree that accurately describing the true output of LED's is very complicated, are you saying that the complexity of it makes this attempt to simplify it pointless?
 

ElectronGuru

Flashaholic
Joined
Aug 18, 2007
Messages
6,055
Location
Oregon
Re: Flashlight Color Temperature Reference

A real world frustration often encountered is the described "warm" LED's which in addition to being "warm" are also an unexpected green. I've found this to be a common fault with incan "replacement" bulbs intended for household use.

I don't know how typical it is, but I've so far found zero green in your M60W MC-E, just the slightest hint (100K worth) of blue. I've yet to see an LED in the < 4000 range in person.


It has been my experience that knowing both the CRI and the CCT it is based on has given me the best "feel" for what to expect from a LED in terms of color when I use the LED to illuminate objects of color.

Granted we are much more complicated than a LED but the LED has more dimensions than some of us give it credit for.

Being much closer to the source end of things, hearing your and Gene's perspective adds greatly to our understanding, thank you both.


Once you get into illuminating a full spectrum landscape, I need a whole bunch more parameters to identify in describing my preferred light source and much of it is a guess based on limited understanding and not empirical or measured experience.

In the real world, I have found that in many cases, flux trumps CCT as well as CRI since sufficient light, across the spectrum may be more important than the actual distribution or spectrum itself. Contrast is also critical and it may or may not have color in its makeup; the distribution of intensity (lux) alone may be at issue.

Ultimately, my need to understand the light source is back seat to actually using the light source. If I can get the source into the ballpark then I can go about viewing the ball park."

I'm working on a color rendition test and will include shadows in the criteria. I agree, whatever the measurement is suppose to reveal, there is no substitute for real testing. Now its a matter of building a test with enough variation to challenge the lights, but with enough stability to be a useful comparison.


Edit: In recognition of the broader meaning of "color", I am removing "temperature" from the thread's title
 
Last edited:

MichaelW

Flashlight Enthusiast
Joined
Dec 8, 2007
Messages
1,788
Location
USA
Re: Flashlight Color Temperature Reference

I was thinking long term with a 3000K CCT for warm-white LEDs.
In the future the LEDs will be so bright... that they will be brighter AND warmer than incans. It seems the harder they get pushed [for more light/efficiency] the white point keeps rising-3500K+
Well what if you want to keep it warm while having your brightness too (without filtering), then you'd be left with leds {note this time frame might be ten years out}

and a 6000K CCT is usually the most efficient LEDs, they are for people who like bragging rights, or where the situation requires quantity over quality (if you are old and have a yellowing of the eye)

and 4500K for the rest of us.

What ever came of this?
http://www.led-professional.com/content/view/1038/29/
It is nice that the granularity is smaller. There are 5 bins between 4500-7000 instead of 4, 4 instead of 3 from 3500-4500, and 6 vs. 5 from 3500-2540.
 

Tekno_Cowboy

Flashlight Enthusiast
Joined
Apr 2, 2008
Messages
1,680
Location
Minnesota
Re: Flashlight Color Temperature Reference

McGizmo:

While I agree that accurately describing the true output of LED's is very complicated, are you saying that the complexity of it makes this attempt to simplify it pointless?
Reading it over again it looks more like you were just providing information to help make the testing more accurate.

I think I spent a little too much time out in the sun today :ohgeez:
 

ElectronGuru

Flashaholic
Joined
Aug 18, 2007
Messages
6,055
Location
Oregon
A few tips if you are considering your own ExpoDisc:


Get the Neutral version, not the Warm/Portrait version.

Getting the size that matches the filter size of your biggest lens means you can use it on all your lenses. But for purposes here, match the size of your intended lens so you can hold the camera/disc with one hand and light with the other. The focal length you choose is not important. I use a 35mm, but a telephoto should work just as well.

Learn to take perfectly calibrated, indoor, flash-free photos first to familiarize yourself with how it and your camera work together.


Be sure your camera allows specific Kelvin settings!

It appears here (Canon) as Color Temp - 5200K:
http://a.img-dpreview.com/reviews/CanonEOS20D/Images/Captures/anim_menu1.gif

Cameras confirmed to have this feature
Canon 10D
Canon 20D
Canon 30D
Canon 40D
Canon 50D
Canon 5D
Canon 5D Mark II
Nikon D2H
Nikon D2X
Nikon D3
Nikon D3X


Edit: It may be possible, shooting in raw, to calibrate to 4300K in software after taking the picture, not needing the fancy camera to do it during shooting.
 
Last edited:
Top