LED color temperature issue

[King Cheung]

Dear All,

I used Sekonic C500 color meter to check the color temperature of several LED lighting. However, the result is quite different from what the manufactures state. For example, for one of the very famous and trusted manufacturers' LED light, I have to adjust it all the way to 6700K in order for my meter read 5600K. It is not a single incidence.

Then someone told me that it is because LED light inevitably has some missing wavelength, so for checking the color temperature, we need to use a dedicated LED color meter. Anyone heard about it? Anyone has used this kind of dedicated meter?

Thanks in advance.
King

 

[Robert Hart]

Having a dedicated light meter for LED lighting sounds a bit like revising history to make it more palatable to modern sensitivities. It may seem nicer but does it remain a statement of the truth? My comment should be tempered by my confession that I do not consider myself competent in lighting.

Practice, testing, more practice and testing as the best people do it in building up their knowledge and experience. We are lucky in that the modern digital cinema cameras show us that very second, within limits what we can expect. We can experiment for the best result in a fraction of the time and cost that previous chemical imaging generations had to invest just to find out the things which did and did not work. We live in remarkable times. I for one am enjoying the privelidge.

 

[Mark Britton]

LED lights don't generate a continuous spectrum of light, at least to date. A color meter essentially samples three wavelengths of light (a red, a green and a blue) and makes assumptions based on the ratios it receives compared to a known/calibrated reference.

A LED light might have a large valley where little light is emitted in a narrow band of colors. If that valley happens to line up with one of the color meter's sampling wavelengths, then the LED light could easily read way off from what the "actual" color temperature of the LED light is emitting (approximating?). The color meter could 'miss' seeing the blue output of a particular LED fixture, for example.

The manufacturing and science behind LED fixtures coming out today are evolving rapidly, and different methods of compensating for the current limitations of LED lighting varies significantly. A 'dedicated' LED color meter would have to be made differently than a traditional color meter, and each LED light would probably have profiled and programmed into the specialized meter. I'm not aware of any such device currently.

My guess is that by the time someone got around to making a practical 'LED color meter', LED lights would have evolved to the point where a specialized device won't seem warranted.

And we won't stress out so much about if a new LED light that unexpectedly shows up on set is going to give us horrible color rendition in, say, skintones with that sophisticated color meter we call "the camera" :laugh:

 

[King Cheung]

About miss seeing the blue output, the difference of tungsten setting between the meter reading and Kevin stated on the light fixture is somehow less than daylight setting. Also, Arri still doesn't have L7D on market, and it may be the reason.
Thanks.
King

LED lights don't generate a continuous spectrum of light, at least to date. A color meter essentially samples three wavelengths of light (a red, a green and a blue) and makes assumptions based on the ratios it receives compared to a known/calibrated reference.

A LED light might have a large valley where little light is emitted in a narrow band of colors. If that valley happens to line up with one of the color meter's sampling wavelengths, then the LED light could easily read way off from what the "actual" color temperature of the LED light is emitting (approximating?). The color meter could 'miss' seeing the blue output of a particular LED fixture, for example.

The manufacturing and science behind LED fixtures coming out today are evolving rapidly, and different methods of compensating for the current limitations of LED lighting varies significantly. A 'dedicated' LED color meter would have to be made differently than a traditional color meter, and each LED light would probably have profiled and programmed into the specialized meter. I'm not aware of any such device currently.

My guess is that by the time someone got around to making a practical 'LED color meter', LED lights would have evolved to the point where a specialized device won't seem warranted.

And we won't stress out so much about if a new LED light that unexpectedly shows up on set is going to give us horrible color rendition in, say, skintones with that sophisticated color meter we call "the camera" :laugh:

 

[Peter Berger]

A 'dedicated' LED color meter would have to be made differently than a traditional color meter, and each LED light would probably have profiled and programmed into the specialized meter. I'm not aware of any such device currently.

This issue was recently discussed at length on the Cinematographer’s Mailing List (CML.) It was reported in that forum that Konica Minolta, manufactures a light meter, the 200A, that can accurately measure the color output of any LED emitter. But, because it is not intended for measurement for photographic corrective purposes but rather for LED billboard quality control and maintenance, it is of no use to us. Here are the reasons given in the CML forum.

To be used for color correction purposes in photography, a color meter must be calibrated to the film or digital camera sensor in order to measure the color of light and present the data in terms that a photographer can use. As you may recall, you set the earlier Minolta Color Meters (the II and III) for the type of film being used (Type A, B, or D) and it generated readings in Kelvin and CC (Color Compensating) and LB (Light Balancing) scales, and provided on the back an index that directly correlated the reading to corrective gels. While the Correlated Color Temperature readings with Δuv specifications generated by the 200A will no doubt accurately place a light source (including LEDs) in color space, the readings do not translate to CC (green/magenta) or LB (red/blue) measurements. In short, for a color meter to work requires complementary imaging systems, light sources, and corrective gels calibrated for both. To use the available color correction gels (listed above) to correct LEDs is a misapplication of a finely tuned system of correction designed for continuous spectrum light sources only.

A good example of this is what happens when you try to convert the 5500K out-put of Phosphor White LEDs to 2900k with Full CTO gel. Where you can do it with some success with HMIs because there are long wavelengths in it’s continuous spectrum to pass disproportionately to the blue part of the spectrum to achieve a nominal 2900K, since LEDs don’t put out much beyond 625nm, there is not much for a filter to pass to rebalance the light output to 2900K, so the “corrected” light is too cool. Another undesirable consequence comes from the fact that Full CTO is designed to pass extra green (there is a bump in the spectral transmission curve of Full CTO in the green portion of the spectrum) and so it creates, given the amount of green inherent in Daylight LEDs to begin with, a disproportionate amount of green (creating an overall green bias) to the “corrected” light when used on Phosphor White LEDs.

DP Michael Morian recently conducted some camera tests in which he attempts to correct “daylight” LED 1×1 light panels to 3200k. The results demonstrate clearly the unexpected and undesirable consequences of using CTO gels on a light source for which they were not designed. You can see the results for yourself on his website at http://talltalepictures.com/blog/2011/12/06/are-led-1x1-panels-ready-for-prime-time/.

Peter Berger

 

[Mark Britton]

Thanks Peter, excellent explanation and correction!

 

[neale-anderson]

Solution = buy one of these http://www.rotolight.com/

 

[King Cheung]

Solution?? You sure??? Or it is just an advertisement???? Please explain to us.

Solution = buy one of these http://www.rotolight.com/

 

[neale-anderson]

Solution?? You sure??? Or it is just an advertisement???? Please explain to us.

Hi I take delivery tomorrow,so I will post some test's.In the meantime if you research as much as you can about these light's,you should find as I did they are very accurate and consistant .Ps I have no affiliation with the company Im very impressed so far

 

[Sergei Franklin]

I find color meters are useful to match same tyoe of lights, ie match HMI lights. I do not find they work for matching different light types such as HMI to Kinoflo or LED.

 

[Peter Berger]

if you research ... you should find as I did they are very accurate and consistant....

If in your research, you searched Rotolight in past Reduser thread topics, you would have found that is simply not true. As was pointed out in one - http://reduser.net/forum/showthread.php?82878-Rotolight-LED - if the color is so accurate and consistent, why does the ANova kit come with ½, ¼, & 1/8 minus green gels? Because, like the output of all Bi-color LEDs using 5500 and 3200 White Phosphor LEDs, the output of the ANova is not accurate or consistent. While it intuitively makes sense that you can achieve a nominally correct intermediate color balance by balancing the output of 3200K & 5500K LED diodes as the Rotolight website claims, as Harry Box points out in the 4[SUP]th[/SUP] edition of his handbook this approach can not generate light that is green/magenta neutral and so color correction is required regardless of what the website claims – which is why the kit comes with an assortment of minus green gels.

And, as Harry Box also points out, White Phosphor LEDs, like the ones used in this fixture, don’t emit frequencies above a 625nm cut-off, so blending 3200 and 5500 White Phosphor LEDs is not going to correct the deficiency of long wavelength colors in LED heads like the Rotolight ANova. Because the missing long wavelength colors are critical to rendering a good flesh tone, the kit also includes a cosmetic peach gel as well. Don’t buy into all the hype surrounding LEDs. As the gel pack for this light indicates, their color output is pretty poor. And using standard color correction gels like CTOs and CTBs on them lead to unexpected results. For more details on how poor the color output of these LEDs are, I would suggest you go to the companion website Focal Press has set up for the fourth edition of Harry Box’s “Set Lighting Technician’s Handbook.” On it, he has links to several very good articles on LED color output as well as trouble shooting HMIs and generators. You can log onto the site at http://booksite.focalpress.com/box/setlighting/ with the pass-code "setlighting."

Peter Berger

 

[neale-anderson]

If in your research, you searched Rotolight in past Reduser thread topics, you would have found that is simply not true. As was pointed out in one - http://reduser.net/forum/showthread.php?82878-Rotolight-LED - if the color is so accurate and consistent, why does the ANova kit come with ½, ¼, & 1/8 minus green gels? Because, like the output of all Bi-color LEDs using 5500 and 3200 White Phosphor LEDs, the output of the ANova is not accurate or consistent. While it intuitively makes sense that you can achieve a nominally correct intermediate color balance by balancing the output of 3200K & 5500K LED diodes as the Rotolight website claims, as Harry Box points out in the 4[SUP]th[/SUP] edition of his handbook this approach can not generate light that is green/magenta neutral and so color correction is required regardless of what the website claims – which is why the kit comes with an assortment of minus green gels.

And, as Harry Box also points out, White Phosphor LEDs, like the ones used in this fixture, don’t emit frequencies above a 625nm cut-off, so blending 3200 and 5500 White Phosphor LEDs is not going to correct the deficiency of long wavelength colors in LED heads like the Rotolight ANova. Because the missing long wavelength colors are critical to rendering a good flesh tone, the kit also includes a cosmetic peach gel as well. Don’t buy into all the hype surrounding LEDs. As the gel pack for this light indicates, their color output is pretty poor. And using standard color correction gels like CTOs and CTBs on them lead to unexpected results. For more details on how poor the color output of these LEDs are, I would suggest you go to the companion website Focal Press has set up for the fourth edition of Harry Box’s “Set Lighting Technician’s Handbook.” On it, he has links to several very good articles on LED color output as well as trouble shooting HMIs and generators. You can log onto the site at http://booksite.focalpress.com/box/setlighting/ with the pass-code "setlighting."

Peter Berger

So he has tested these ?.Can you point me to his tests

 

[Peter Berger]

So he has tested these ?.Can you point me to his tests

He doesn’t have to – the fact that Rotolight includes minus green gels in the kit is an admission that the color output of their ANova is not accurate and consistent. Why would you need ½, ¼, & 1/8 minus green gels if it was? Any LED light that purports to offer a continuous range of nominally correct color temperatures between 3200K & 5500K by mixing only warm and cool LEDs is going to require green/magenta correction – whether its CoolLights, Litepanels, LED312, or Rotolight Bi-colors.

According to Harry Box, if you were to plot on a chromaticity diagram all the colors that are possible by mixing two colors of light it would form a straight line between the two points. In contrast, if you were to chart the colors generated by heating a black body radiator (the basis of the CCT scale) as it turns red, orange, yellow, and finally blue as it is heated, it is not a straight line but an arc. So it is simply not possible to create light that remains neutral in terms of a green/magenta bias, while mixing only two LED colors. Any “bi-color” panel, even Rotolight’s, will require green/magenta correction as you dial between the two points – this is why they include the ½, ¼, & 1/8 minus green gels in the kit. And, this is why Arri correctly incorporated “green” and “magenta” LEDs in their Locaster panel.

Don’t buy into all the hype surrounding LEDs. As was also pointed out in the Rotolight thread, the gels included in the kit for this light clearly demonstrates just how meaningless high CRI numbers are when it comes to photography (CRI was developed for merchandize display not photography.) A CRI of 91 can’t mean much when you need cosmetic peach and minus green gels to get decent color rendering on screen. As the gel pack for this light indicates, their color output is pretty poor. If you don’t believe me, check out the articles Mike Wood has written for Protocol Magazine that Harry Box includes on the companion website for the fourth edition of his “Set Lighting Technician’s Handbook.” You can log onto the site at http://booksite.focalpress.com/box/setlighting/ with the pass-code "setlighting."

Peter Berger

 

[neale-anderson]

He doesn’t have to – the fact that Rotolight includes minus green gels in the kit is an admission that the color output of their ANova is not accurate and consistent. Why would you need ½, ¼, & 1/8 minus green gels if it was? Any LED light that purports to offer a continuous range of nominally correct color temperatures between 3200K & 5500K by mixing only warm and cool LEDs is going to require green/magenta correction – whether its CoolLights, Litepanels, LED312, or Rotolight Bi-colors.

According to Harry Box, if you were to plot on a chromaticity diagram all the colors that are possible by mixing two colors of light it would form a straight line between the two points. In contrast, if you were to chart the colors generated by heating a black body radiator (the basis of the CCT scale) as it turns red, orange, yellow, and finally blue as it is heated, it is not a straight line but an arc. So it is simply not possible to create light that remains neutral in terms of a green/magenta bias, while mixing only two LED colors. Any “bi-color” panel, even Rotolight’s, will require green/magenta correction as you dial between the two points – this is why they include the ½, ¼, & 1/8 minus green gels in the kit. And, this is why Arri correctly incorporated “green” and “magenta” LEDs in their Locaster panel.

Don’t buy into all the hype surrounding LEDs. As was also pointed out in the Rotolight thread, the gels included in the kit for this light clearly demonstrates just how meaningless high CRI numbers are when it comes to photography (CRI was developed for merchandize display not photography.) A CRI of 91 can’t mean much when you need cosmetic peach and minus green gels to get decent color rendering on screen. As the gel pack for this light indicates, their color output is pretty poor. If you don’t believe me, check out the articles Mike Wood has written for Protocol Magazine that Harry Box includes on the companion website for the fourth edition of his “Set Lighting Technician’s Handbook.” You can log onto the site at http://booksite.focalpress.com/box/setlighting/ with the pass-code "setlighting."

Peter Berger

So let me understand your direction,no test's,have you picked up the light ?,has harry ?,get back to me when you have

 

[Peter Berger]

So let me understand your direction,no test's,have you picked up the light ?,has harry ?,get back to me when you have

Rotolight is not immune to the laws of physics? There are inherent limitations to the “Stokes Shift” process by which the White Phosphor LEDs used in the Rotolight transform the shorter wavelengths of their pump color to longer wavelengths by exciting phosphors. One is that there is a drop out of colors in the medium blue-cyan-turquoise range from about 465-510nm. A second is that there is a steep drop off in colors over 600nm. This is true to one degree or another of all LED fixtures using White Phosphor LEDs, and they all show the same deficiencies in color rendering. Why would the Rotolights be any different?

Some think that because of the rapid pace in improvement of LED technology that each new fixture has better output and color rendering than the previous. As was clearly demonstrated in a recent thread on Cinema 5D, that is simply not the case. To summarize that thread, a close look at the improvement in lumens per watt efficiency of the Cree Xlamp line of LEDs reveals that it, in fact, comes at the expense of color rendering. According to Cree’s website, the MT-G2 “pushes performance limits to redefine lumen levels and efficacy (it is 25% brighter then the MT-G).” But when you compare the spectral output of the two Cree Xlamp emitters side by side, as they do in the Cinema 5D thread, you see that the greater lumen output of the MT-G2 comes at the expense of color rendering capability. The reason for this is that there are limitations inherent in the remote phosphor approach that both the Cree XLamps and the Rotolight ANova use.


Given these limitations, the remote phosphor approach to generating photographic white light with LEDs has been all but abandoned by the more serious manufacturers of LED luminaries for motion picture production – leaving only manufacturers of LEDs for the general illumination market and those manufacturers that repurpose them for motion picture lighting applications. In order to get from an LED fixture the consistent color and quality of light that one gets from an inexpensive Tungsten Fresnel, requires a multi-color LED array and a sophisticated micro-processor controlled light engine. For example, the Arri L7-C LED Fresnel overcomes the inevitable green/magenta shift of bi-color LED lamp heads like the Rotolight by blending the output of five different color LEDs. This approach assures more consistent color output that is green/magenta neutral – something that “dumb” Bi-color panels like the Rotolight simply cannot do.

Peter Berger

 

[Marc-André Gray]

According to Harry Box, if you were to plot on a chromaticity diagram all the colors that are possible by mixing two colors of light it would form a straight line between the two points. In contrast, if you were to chart the colors generated by heating a black body radiator (the basis of the CCT scale) as it turns red, orange, yellow, and finally blue as it is heated, it is not a straight line but an arc. So it is simply not possible to create light that remains neutral in terms of a green/magenta bias, while mixing only two LED colors.

Thanks for sharing Harry Box's insights. This type of information is great.

I did a little reading over at Wikipedia and it seems there's a tonne of problems with the CRI.
http://en.wikipedia.org/wiki/Color_rendering_index#Criticism_and_resolution

They also go on to talk about various other indices people are working on to improve on the CRI.

 

[Nick Gardner]

While all of the above is true, I use LEDs almost everyday to light people, and with the appropriate gells, they look great. Last time I checked, HMIs weren't full spectrum either, and people use them all the time. I find LEDs much more flattering than kinos, I always thought they rendered weird skin tones.

It's a different quality of light, but once you get a handle on how to use them, they can look great. Nothing is ever going to beat a 6x6 silk with a 5k shot into it, but a 1x1 LED weighs a couple of pounds, doesn't get hot, and I've done night exteriors lit off of AB batteries and the 300 watt inverter in my van ;-)

 

[King Cheung]

Ok, maybe, I misunderstood you, and I cannot wait to see your independent test, which surely benefits all of us in here.
Thanks.
King

Hi I take delivery tomorrow,so I will post some test's.In the meantime if you research as much as you can about these light's,you should find as I did they are very accurate and consistant .Ps I have no affiliation with the company Im very impressed so far

 

[King Cheung]

It is because light meters respond different light types differently, right? Then in the situation of mixed lighting types, what can we do? Any experience to be shared?
Thanks.
King

I find color meters are useful to match same tyoe of lights, ie match HMI lights. I do not find they work for matching different light types such as HMI to Kinoflo or LED.

 

[Peter Berger]

Last time I checked, HMIs weren't full spectrum either, and people use them all the time.

A distinct advantage (and it is a big one) to HMIs over LED panels is that the “discontinuous” color spectrum of HMIs is a lot more continuous than that of the White Phosphor LEDs used in LED light panels like the Rotolight ANova and so can be easily corrected with gels when LED panels can not.

The problem with White Phosphor LEDs is that they generate a very discontinuous spectrum resembling peaks and valleys and drop off appreciably above 600nm. Where it is the long wavelength colors above 600nm that are critical to rendering a realistic flesh tone, skin will always look better under HMI light than LED light because the output of HMIs extend well beyond the 600nm cut-off of the White Phosphor LEDs used in LED fixtures like the Rotolight ANova (use this link for details.)

I find that once I correct for the minor green spike of HMIs with minus green (magenta) gels that have been designed specifically for that purpose, I get a much better flesh tone than what I get with the White Phosphor LED panels. I think it is because the spectral distribution of the corrected HMIs is much more continuous and contains a greater proportion of the long wave length colors that are so critical to rendering flesh-tones accurately than a LED panel that will still lack those colors even after correction.

Another advantage to HMIs is that they provide enough output that you can waste some to diffusion material or a Chimera bank and get a much softer and more flattering light output by diffusing them. I find that with LEDs if you don’t want the look of a thousand points of light (with a thousand tiny shadows) you have to put heavy diffusion on them – knocking down the output of an already weak source.

Peter Berger