Florescent and LED spectrum is nowhere near flat

[Oren Arieli]

I'm with Jad. I hate walking into a new room and praying that I can use more than 3 tungsten lights at a time without:
A) cooking the talent
B) blowing a fuse
C) frying a bulb
D) melting the fake Ficus tree

Tungsten has had a great run. Three cheers for Thomas Edison. My future lighting budget will probably not spent on hot filament technology.

[Dean Sensui]

I was shooting our cooking segments in a room that had lots of cool daylight but needed additional lighting to provide some directional qualities.

The supplemental lighting was from two CoolLights 600s, fitted with "minus 1/2 green" filters, plus one small Kinoflo that I got via Craigslist for $65.

I put a tungsten "hair light" in the back to provide some additional warmth to the stainless steel environment.

There has never been a problem with a green spike. And the food comes out looking good.

Attached is a frame grab from the Sony EX1.

http://hawaiigoesfishing.com/images/...ok_&_print.jpg

[Tom Morrow]

Hey, that CD spectrometer was a fun little project. I'm not getting anywhere as nice results as the web, but I do feel like I'm getting a sense of the spectrum of my lights with it.

I've been watching video more carefully since this discussion, and often noticing sickly color casts in some of the newer TV shows that might be using flourescent lighting (House, 24). I'm guessing this is one of those "you'll get used to it" things, kinda like CD sound in the 1980's.

I did some experiments and determined that 1/8 minusgreen on the LEDs and CFLs, and 1/4 minusgreen on the 5000k linear tubes looks like it will come closest to neutral. Next task is to try them out and see how close we came.

I've always liked incandescent light, and now I really know why... there's just more information there.

[Charles Papert]

Hmmmm...probably not. Kino's have been in use now for 25 years. You've seen plenty of movies and TV shows shot with them without "sickly" color casts. That particular look is deliberate and often dialed in in post as a choice. If there is a difference between corrected fluorescent and tungsten light, it's pretty subtle. I've always meant to shoot a head to head comparison, might still be an interesting procedure and maybe include a quality LCD unit as well.

[Tom Morrow]

You know, that's the one thing I haven't seen on the kino site is a side by side flor vs tungsten photo to show how well they can render light. If anyone has good examples I'd like to see. Obviously it's not trivial getting the light shaped the same from different bulb sizes, but a really accurate test could be done illuminating a frosted window indirectly to illuminate a scene.

[Charles Papert]

It would be fairly easy to duplicate the effect of a Diva 400 or a 2 ft 4 bank with diffusion over it by using a tungsten unit and shooting it through the same size diffusion (to minimize variables). The color temperature may shift as a result but at least it will be the same shift in both. Don't know if I'll ever get around to it but it would be an interesting experiment, especially to put the images up here as a blind test.

[Guy Holt]

Tests recently performed by The Academy of Motion Picture Arts and Sciences (AMPAS) has revealed a lot of the information in this thread to be untrue. The AMPAS tests clearly demonstrate that all LED panels that use Phospher White LEDs (Litepanels, Cool Lights, etc.) are so deficit in certain parts of the color spectrum that they are inherently incapable of rendering colors accurately and so don’t mix well with other light sources. It also reveals that the high CRI claims of LED manufacturers to be nothing more than the “emperor’s new clothes”.

If you look at the Spectral Power Distribution graphs of Phosphor White LEDs (available at ScreenLight & Grip's E-Newsletter) you see that, as Tom noted at the outset, they have a discontinuous spectral quality that is unlike that of natural daylight, HMI, or incandescent light. In the case of the 3200K Phosphor White LEDs, the phosphors added shape the spectral distribution by enhancing certain colors in the spectrum to simulate the spectral distribution of incandescent light. As a result, the spectral distribution of Phosphor White LEDs resembles a series of peaks and valleys with a big spike at about 465nm (the blue “pump” LED) and a broader bump between 500 and 700nm produced by the phosphors. While the discontinuous spectral distribution of high CRI White Phosphor LEDS may appear white to the eye, and the color of objects illuminated by it appear natural to the eye (hence their high CRI), to film emulsions and digital imaging systems designed to reproduce accurate color under continuous spectrum light sources (like daylight or incandescent lamps), the color of the same objects appear slightly off on screen. For example, in a company newsletter on this subject, I have posted a split Macbeth chart showing the visible effects of studio tungsten light in the top half of each color patch, and a high CRI White LED (that shall remain nameless) in the bottom half.
It doesn’t take a calibrated reference monitor to see the wide variations in color patch hue caused by the discontinuous spectral distribution of the Phosphor White LED.

What accounts for these results? Phosphor White LEDS, compared to continuous light sources, have no output at wavelengths shorter than about 425nm, which means that violet colors don't render well. Second, there is minimal output in the medium blue-cyan-turquoise range from about 465-510nm, which is why the aqua-type colors don't render well either. Third, with the long-wavelengths cutoff in the high-600 nm range, pinks, reds, oranges, and other long wave-length colors tend to look a little dull under Phosphor White LEDs, compared with how they look under a continuous spectrum light source like Tungsten which extends all the way out on the long-wavelength end. Lacking these colors within the spectrum, skin tones also look pale under Phosphor White LEDs.

For side-by-side comparisons between Phosphor White LED fixtures and Tungsten see a newsletter article I wrote on this subject that is available at ScreenLight & Grip's E-Newsletter.

Guy Holt, Gaffer, ScreenLight & Grip, Lighting Rental and Sales in Boston

[Guy Holt]

Originally Posted by Bill Davis

Yes, there are spectrum issues. However, the judicious use of filtering (subtractive magenta for counteracting some of the green) - and more important, the ability of modern editing suite software to "re-color balance" in post - makes it perfectly possible to generate very good on-screen results with these amazingly efficient light sources.

Good luck trying to correct all the deficiencies of LED panels noted above with gels. Even the better LED light panels, like the Litepanel 1x1s (forget about the Cool Lights LEDs), are so deficit in certain parts of the color spectrum that by the time you came up with a color gel pack to match them to a continuous light source like a tungsten or HMI light, the LED panel would put out very little light with all those gels on it.

While you can white balance out/time out the characteristic green/magenta bias of Phosphor White LEDs in digital video cameras/digital film intermediate, the camera/timer is not able to replace the parts of the spectrum that are missing all together. And since gels only rebalance the spectral distribution of a light source by passing the wavelength of the color that they are, gels cannot correct for these deficiencies either because there is not light of those wavelengths to pass in White Phosphor LEDs to begin with.

This inability of Phosphor White LEDs to render color is very visible in the tests Tom mentioned that The Academy of Motion Picture Arts and Sciences (AMPAS) recently performed. In one a model was photographed wearing a dress that had a number of different blue/cyan tints. Footage was shot with both a tungsten source and a Phosphor White LED source (see ScreenLight & Grip's E-Newsletter for the results.). The tungsten-lit footage displayed all of the subtle differences in blue/cyan tones in the fabric, while the LED-lit footage, lacking cyan output, showed just a nice blue dress, without the same richness of hue. Since the light doesn’t put out much cyan, the camera/film simply can’t record it. You can also see in the side-by-side comparison that, the skin tones of the model don’t stand out because of the rapid drop off of long wavelength colors within the spectrum of White Phosphor LEDS. The bottom line is that, simply by nature of their discontinuous spectral distribution, all high CRI Phosphor White LEDs (Cool Lights and Litepanels included) will never accurately reproduce colors on screen regardless of how much gelling or post production correction you do.

While they are less than perfect at reproducing parts of the color spectrum, the color rendering of Phosphor White LEDs may, as Bill points out, be adequate in certain situations. For a specific application, say where lights must be operated off of batteries, a LED fixture offers the unique advantage of greater power efficiency over conventional lights, which may out weigh its shortcomings in color rendering. However, a Phosphor White LED is clearly not the best choice in applications where color rendition is critical (food/product shots) or mixed with a uniform continuous light source, such as a studio lit with tungsten fixtures, where its color deficiencies will be quite noticeable and unacceptable in comparison. For this reason, to pick the right LED luminary for a particular job it helps to have a thorough understanding of the technology. For our company newsletter I have put together an overview of the technology and what products are available for motion picture lighting (available at ScreenLight & Grip's E-Newsletter.)

Guy Holt, Gaffer, ScreenLight & Grip, Lighting Rental and Sales in Boston

[Guy Holt]

Originally Posted by Charles Papert

I can't tell you what the technical difference is between high and low end flos and LED's, but the bottom line is that the products packaged for the film industry photograph just fine. Most features and TV shows utilize both technologies and the skin tones are obviously acceptable there.

Unfortunately this is not true. In fact, it was a bad experience that Daryn Okada, ASC, (a AMPAS Science and Technology Council member) had mixing a Phosphor White LED lite panel with tungsten lights on a studio set that lead to AMPAS undertaking their Solid State Lighting Project tests that have revealed the high CRI claims of LED manufacturers to be nothing more than the “emperor’s new clothes”. According to Matt Hurwitz’s article “It’s a Phosphorous World” that appeared in the June issue of ICG Magazine, Daryn Okada, first noticed color rendering problems mixing Phoshpor White LEDs with tungsten sources when he used a “Tungsten” LED panel to touch up a face on one talent mark of a tungsten lit set. As the story goes, he hid a small LED unit behind a chair, to add some glow to an actress’s face when she reached a mark where the keys had fallen off. “The manufacturer claimed the unit to be a ‘tungsten LED source,’” he recounts. “She stopped right in the doorway, where I had this LED, and looked fine. But when I got the dailies back, her face was totally magenta.” What’s worse, Okada says the image could not be repaired in post, because there wasn’t enough of the right color of light in the scanned negative for a color timer to bring out.

It is a common mistake to think that a custom camera white balance can correct for the deficiencies of LEDs in every situation. Take Daryn Okada’s situation above. Had Daryn Okada been shooting with a digital video camera, he would have noticed the off color of the LED source immediately. But, given the limited spectral output of LEDs, his ability to remedy the problem would have been limited. If he white balanced the camera for the LED source, the background of the room beyond the doorway that was lit by tungsten lights would turn very green. In a mixed light situation such as this, the only alternative is to match the LED source to the prevalent tungsten source with a custom gel pack on the LED head. But, since gels rebalance the spectral distribution of a light source by passing the wavelength of the color that they are, it is not practical to use gels to correct for these deficiencies either because there is not light of those wavelengths to pass in White Phosphor LEDs to begin with.

To understand why this is so, we need only look at a similar situation: the conversion of tungsten light to daylight using full CTB gel. As you can see in the Spectral Power Distribution graph of tungsten light, tungsten light is so deficient in the blue part of the spectrum that it takes a quite saturated blue gel to balance it to daylight. In fact, the transmission coefficient of full CTB gel is only 36%, which means that it passes only 36% of the source. That is why gelling tungsten lights is a very inefficient way to create a daylight source (a tungsten 1000w gelled with CTB becomes a 350W daylight source.)

If you look at a Spectral Power Distribution graph of a White Phosphor LED light, a gel pack that would match it to tungsten light would have to include a violet gel to extend its’ spectral output below 425nm. It would have to include medium blue, cyan, and turquoise gels to fill in the missing wavelength from 465-510nm. Finally it would have to include pink, red, and orange gels to extend its’ spectral output beyond its’ 600nm cut-off. All of these gels would have to be quite saturated, since there is very little, if any, output of these wavelengths in White Phosphor LEDs to begin with. Imagine how much light you will get out of a LED light panel with such a gel pack (LED light panels put out barely enough to begin with, and have no output to waste to such accurate color correction.) In other words, White Phosphor LEDS are so deficit in certain parts of the color spectrum that by the time you came up with a color gel pack to match them to a continuous light source like a tungsten light, the LED panel would put out very little light with all those gels on it.

The bottom line is that color gel packs, camera white balance, or digital intermediate timing can’t bring out a color if it isn’t there to begin with. Simply by nature of their discontinuous spectral distribution, even high CRI Phosphor White LEDs will never accurately reproduce colors on screen regardless what you do on set or in post
Kino Flos, on the other hand, have a much more continuous spectrum. Once you filter out their green spike with minus green (magenta gel), they are able to render flesh tones more accurately than LED lights. For details, see my newsletter article available at ScreenLight & Grip's E-Newsletter.)

Guy Holt, Gaffer, ScreenLight & Grip, Lighting Rental and Sales in Boston

[Guy Holt]

Originally Posted by David W. Jones

With using fluorescent or LED lights, I simply use my old trusty Minolta Color meter II to quickly determine the appropriate gels needed to bring the lighting instrument into spec. It may be a 1/8 minus green along with a 1/4 CTO for example.

In the case of LEDs, color meters, like the Minolta III F, are completely useless. The meter makes its calculation of the color temperature based on an assumption that the light source has a continuous spectrum – LEDs don’t. Color readings of an LED have been shown to be misleading for both correlated color temperature and green/magenta shift because of their discontinuous spectral output.

Another drawback to White Phosphor LEDs is that their color output is very inconsistent. That is because their color output is affected by a number of factors: the binning and manufacturing tolerance of their blue pump, the thermal management of the fixture, the ageing of the phosphors, and even the ambient temperature. For example, a one degree shift in the junction temperature of the blue InGaN LED (pump color) in remote phosphor LEDs, will cause a +/- 2nm shift in the dominant wavelength. If compounded by the average wavelength variation of +/- 2nm of blue InGaN LEDs, a 5nm divergence from the prescribed 455nm wavelength of the pump color will create color inconsistency of 5 MacAdams ellispses. While not readily apparent to the eye, image capture systems will easily see this variation.

And, as broadcast studios lit exclusively with Phosphor White LEDs are finding out their output depreciates overtime and their color shifts much faster than the manufacturers say. The useable lamp life of Phosphor White LED luminaries designed for motion picture lighting applications is probably no more than 1500 hrs. To put that in perspective, it is the equivalent of burning through two HMI globes. Since these types of LED fixtures have no interchangeable parts that can be replaced after reaching low-light failure, after 1500 hrs the fixture can only be thrown away while an HMI head can be lamped with another bulb.

You have to be wary of all the claims made by LED head manufacturers because they all put a little spin on the scientific data which has a tendency to cloud issues. For this reason, it helps to have a thorough understanding of the technology. For our company newsletter I have put together an overview of the technology and what products are available for motion picture lighting (available at ScreenLight & Grip's E-Newsletter.)

Guy Holt, Gaffer, ScreenLight & Grip, Lighting Rental and Sales in Boston

[Charles Papert]

Originally Posted by Guy Holt

Unfortunately this is not true.

But Guy, your always incredibly thorough research and white papers notwithstanding, what I was saying that IS true is that these units are out there working and producing results all over the industry without causing consistent problems. Obviously if they were, we wouldn't be using them. Daryn's issue as described could just as easily have been duplicated with, say, a Diva; I've had plenty of issues over the years with those going magenta on me. We all know that the green/magenta shift is not going to be anywhere near as noticeable to the eye as on film (video will tend to reproduce it similarly to the eye). I am regularly mixing tungsten, kino and LED's and I don't see ill effects. I just finished a pilot where I used the Litepads from Litegear as a beauty fill, either under lens or near lens axis. Their "tungsten" ribbon has a green cast and I use 1/4 minusgreen to correct it out. I watch at the monitor and dial the Litepad up and down to the appropriate level. If there was something "evil" happening in the spectrum, I'm not seeing it on the monitor--and in a WYSIWYG world of digital acquisition, what else do I need?

I'm not saying that LED is the be-all and end-all, I use it as a tool along with everything we've always used. I'm not giving up on tungsten yet. I'm just not sure what to do with this information being that empirically speaking, I'm satisfied with what I'm getting. It's a little bit like the naysayers with the DSLR's--on paper the specs are substandard, but the proof is in the pudding; a ton of great work has been shot on them. At some point you just have to get the work done, and in this day and age, speed and flexibility is key. I can (and do) stick the Litepads just about anywhere and get a good result far quicker and easier than I used to with non-LED technology.

This all said, it's interesting to learn about and I will keep an eye on color rendition in the tones you mentioned to see if I can detect issues.

[Bruce Watson]

Originally Posted by Charles Papert

... but the proof is in the pudding...

Sorry, pet peeve of mine. The "correct" quote seems to be: All the proof of the pudding is in the eating which at least makes sense. That quote is often attributed to William Camden, who wrote it in his Remaines of a Greater Worke Concerning Britaine, 1605. An old quote very often misused.

[Charles Papert]

Excellent Bruce! Consider me corrected and I will spread the gospel!

However: considering that LED's deliver such an apparently disorganized spectrum, is it not perhaps accurate to apply such a misquote to them?!

[Richard Andrewski]

Here's some thoughts that come up for me as I see this "revived" thread.

1). I can't particularly speak for other manufacturers but we don't claim LEDs have a high CRI. I just looked at the Litepanels site and I don't even see mention of CRI. There are probably some that do call it high CRI. Its real simple, if there is a color rendering problem they aren't high CRI. Most people using digital mediums today and LED fixtures are not having a problem, are white balancing and using filters and not seeing any big issues. They seem to want to say in the study that filters don't work and yet people using real film have been using this filter method for decades to correct on location lighting like parking lot or office flos, street lights, etc. How was it they got along till this study came out to tell them they're wrong? Filters may not correct a spectrum but they do correct bias and that's more than enough for most uses.

2). Far from trying to cover up anything, we've been at the forefront of education on this subject. We published at the time of release of the LED 600 some articles on LEDs in our site blog in early 2009, long before the AMPAS study came out in late 2010. We've also talked about it many times on sites. Here's one recent thread in particular which even links back to threads that go back to like 2008:

http://www.dvinfo.net/forum/photon-m...weirdness.html

This AMPAS study is primarily showing the response of LEDs and other sources on film. They barely scratched the surface of talking about what happens with digital. How many here, other than Charles, are still using real film a lot (or have ever used real film stock for that matter). I think in the study, they used a Dalsa Genesis (defunct digital motion picture camera) and a Nikon D7000 in a couple of tests and they barely talked about it at all--not particularly widely discussed or used cameras either. Why is that relevant? This is a site primarily talking about mainstream digital cameras motion and still. Digital is used more and more everyday and film less and less. That's a very elite group now and getting smaller all the time. That population here that deals with real film stock is very small compared to the users that use nothing but digital. So the response of lighting to film isn't very relevant to talking about the WYSIWYG world of digital except as an academic topic--and to definitely note, that there are differences between the two mediums. For example, digital is modeled after human perception and has a built-in mechanism for compensating for over-bias in lighting. In other words, a white balance takes care of most issues and if there is some issue of missing color in the spectrum it doesn't come up as an issue for most productions except the most critical. Whatever shift there is is mostly taken care of by taking out the bias of the light in the white balance. On film (which is not even close to human perception and the results had better be well engineered by people that understand how to make it work) the "white balance" so to speak is built in so its appropriate to talk about these issues for film people as they have other problems they have to deal with that digital people don't. I would even go so far as to say there could be a CRI rating for film and a CRI rating for digital. Again, digital is modeled after human perception and so is the CRI rating system. Thus CRI is very relevant to digital users. Not so much for film users. They totally have to understand how light affects the film stock they are using or they will have bad surprises in store when the dailies come back. If a light that is called high CRI isn't working on digital, then it simply isn't high CRI as claimed. The effect of the same light will be even more pronounced on real film given the way it reacts to discontinuous lighting like flos and LEDs (among others). Real film users should already know this however. Seasoned pros have had to deal with these issues for decades. And who would be using film that isn't a seasoned pro? The non-WYSIWYG nature of film doesn't tolerate or suffer fools very well. Mistakes are expensive. With digital you can see the results immediately if you want and correct anything right away. These issues with lighting and real film have existed as long as color film and discontinuous lighting have both existed at the same time which is pretty much since the 50s but the 60s in a more mainstream way with filmed color television dramas. Flos and LEDs are basically the same technology so why would you expect anything different? Knowledgeable film people would already be treating LEDs like they did the lower CRI flo and metal halide counterparts they've had to deal with on location for decades.

3). In the study they used 3200K tungsten lighting as the "reference" for their film shots of the color charts. Flipping it around to a more relevant subject here on this primarily "digital" site, there is no proof that using 3200K lighting with digital would be a good reference. Most sensors I know about were optimized for the daylight range. In a perfect world, color rendering is not affected by color temperature and is more related to spectrum. There is no perfect world however so tungsten lighting may not be all that optimal for digital given the way the sensors are optimized for daylight. Many have said they feel that 3200K (heavier on the RED side of the spectrum than on the BLUE) lighting which lacks in exciting the blue channel of a digital sensor leaves something to be desired. The point of this is, if you were to do some tests yourself using your digital camera, what would you use as a "perfect" reference? Where is the proof that your reference shot (with tungsten) is the real color to compare against other lighting with your digital camera? I hope you won't use tungsten. Even high CRI flos aren't all that perfect in their spectrum, there are always spikes of one kind or another. Probably need to use real sunlight, but then how would you light the other Macbeth chart with the test fixture using this split chart idea where both reference and test are shot at the same time?

4). Cool Lights LED fixtures weren't a part of the AMPAS test--not sure why we get several mentions in Guy Holt's post like we were there somehow. I saw clearly in one of their BTS scenes a Litepanels 1x1 was there along with some Arri tungsten fixture. I don't know what else they used, they didn't even bother to mention it. How scientific is that? Its basically over-generalization in the packaging of a "study" and they barely scratch the surface of how digital reacts to light. Perhaps relevant to their small group of film users but not so relevant to most of the rest of the world which is now using digital mediums. Probably the spectrums of most LEDs are similar but there will be some that are better than others. Certainly I've done tests recently with real, high CRI LED phosphors and I think they stack up every bit as well as the high CRI fluorescents that many have been using with great results in the film and TV world for about 21 years or so. Now we just need the price of said phosphors to come down so we can enjoy them in LED lighting like we do in fluroescent.

[Guy Holt]

Originally Posted by Charles Papert

But Guy, your always incredibly thorough research and white papers notwithstanding, what I was saying that IS true is that these units are out there working and producing results all over the industry without causing consistent problems. Obviously if they were, we wouldn't be using them.

I don’t deny that LEDs are a valuable tool in certain circumstances and can’t be mixed with other sources. For a specific application, say where lights must be operated off of batteries, a LED fixture offers the unique advantage of greater power efficiency over conventional lights, which may out weigh its shortcomings in color rendering. However, a Phosphor White LED is clearly not the best choice in applications where color rendition is critical. For instance I wouldn’t try to light a table top food/product shot with them. Because of their limited color rendering capability, food presentation that will look vibrant and colorful to eye under LEDs, will tend to look a little dull under on camera. By comparison a full spectrum daylight source such as HMI or LEP will capture the vibrant colors.

Or, imagine you were shooting an Air India spot and the uniform of the first class stewardesses were the blue/cyan dress in the AMPAS test I mentioned above. One would think LED light panels would the perfect choice because they are small, flat, soft, and hence ideal for lighting the inside of an airplane. But, in such a situation the Air India reps would be concerned by the fact that the subtle blue/cyan hues of their stewardesses’ uniforms are not being accurately rendered on screen where they would be if they were lit with true tungsten sources (side by side comparison available at ScreenLight & Grip's E-Newsletter.)

I also wouldn’t try to mix them with a uniform continuous light source, such as a studio lit with tungsten fixtures, where if caught in isolation as in Daryn’s case, their color deficiencies will be quite noticeable and unacceptable in comparison.

Originally Posted by Charles Papert

Daryn's issue as described could just as easily have been duplicated with, say, a Diva; I've had plenty of issues over the years with those going magenta on me.

But, with a Diva, once you have corrected for the magenta shift, you have a nearly full spectrum source. With LEDs, it is nearly impossible to correct for their deficiencies with gels.

Originally Posted by Charles Papert

I am regularly mixing tungsten, kino and LED's and I don't see ill effects. I just finished a pilot where I used the Litepads from Litegear as a beauty fill, either under lens or near lens axis. Their "tungsten" ribbon has a green cast and I use 1/4 minusgreen to correct it out. I watch at the monitor and dial the Litepad up and down to the appropriate level. If there was something "evil" happening in the spectrum, I'm not seeing it on the monitor--and in a WYSIWYG world of digital acquisition, what else do I need?

I wouldn’t think you would see the color rendering deficiencies of LEDs when used as a fill in a mixed light situation. I am talking about situations where LEDs are used uniformly or when seen in contrast to other sources in the same frame. I am not saying that you can’t use LEDs successfully. As long as you understand the limitations of the tools you have at your disposal you can do good work. But, in the case of LEDs it helps to have a thorough understanding of the technology to avoid bad situations like those discussed above.

Guy Holt, Gaffer, ScreenLight & Grip, Lighting Rental & Sales in Boston