LED and CRI?

[Carl King]

Hi,

 

I'm trying to help a client of mine. A videographer (who works for the same client) is sending in footage that he's shooting indoors with some LEDs, and he's white balancing at 5500k. The footage looks very yellow and skin tones are zombie-like. Creepy. He's being a bit defensive, so I'm trying to guess what is causing this.

 

Is it possible that his LED lights are poor CRI and he's trying to make it warmer via color correction, but the spectrum is missing certain frequencies?

 

If a camera is white balanced to 5500k and the lights are 5500k, will it look drastically different from a camera white balanced at 3200k with lights at 3200k?

 

Thanks!

[Christoph Helms]

Do you have a sample shot?

It could also have something to do with the color profile... if you shot with the Technicolor Cinestyle Profile, Skin Tone Colors will be on the pale side! Were there other Tungstens in the room that resulted in mixed light conditions?

[Carl King]

I asked about the color profile and he says it is all flat. But these skin tones are not just pale, they are yellow / grey zombie looking. Everything in the shots has a yellow color to it. Now that you mention it: there is, in fact, a tungsten light that he claims to be using as a backlight. But I don't see any backlighting in the shots, so I don't know. I'm really starting to suspect it's just low-CRI LEDs.

 

Here is something on Wikipedia: http://en.wikipedia.org/wiki/Color_rendering_index#Film_and_video_high-CRI_LED_lighting_incompatibility

 

"Problems have been encountered attempting to use otherwise high CRI LED lighting on film and video sets. The color spectra of LED lighting primary colors does not match the expected color wavelength bandpasses of film emulsions and digital sensors. As a result, color rendition can be completely unpredictable in optical prints, transfers to digital media from film (DI's), and video camera recordings. This phenomenon with respect to motion picture film has been documented in an LED lighting evaluation series of tests produced by the Academy of Motion Picture Arts and Sciences scientific staff.^[29]"

[Adrian Sierkowski]

LEDs are very bad for color across the board. Even the expensive ones leave skin sickly to my own eye. On the feature we're shooting I'm using a lot of LED heads (litepanels) out of necessity for power in remote locations where you just can't get a genny, so battery power is a plus, and they look pretty bad. Not horrible, and as I'm keeping them cooler in every shot I can live with it. But whenever I balance the camera to the LED it just looks bad.

If you have a color meter, maybe you can find a quick correction gel for it to get a little less sickliness.

[Guy Holt]

LEDs are very bad for color across the board. Even the expensive ones leave skin sickly to my own eye.

 

 

There is good reason to be wary about LED's ability to render a good skin tone. The inability of Phosphor White LEDs, even those used in the Litepanel and Coolights 1x1, to render color accurately has been well established in tests recently performed by The Academy of Motion Picture Arts and Sciences (AMPAS) as part of their “Solid State Lighting Project Technical Assessment.” (see http://www.screenlightandgrip.com/html/emailnewsletter_generators.html#anchorHigh%20Output%20AC%20LEDs for details.) In one (below) a model was photographed wearing a dress that had a number of different blue/cyan tints. Footage was shot with both a true tungsten source and a White Phosphor LED source. The tungsten-lit footage displayed all of the subtle differences in blue 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 because those wavelengths are not reflected by the dress.

Left: Tungsten source, Right: White Phosphor LED source.

 

The same holds true of flesh tones illuminated by LED light. As is also evident in the pictures above, skin tones don’t reproduce well under White Phosphor LED lights because of the steep drop off of high frequency colors (above the 600nm cut off) such as pinks, reds, oranges, and other long wave-length colors. As the illustration below, comparing the reflected spectral distribution of a Caucasian skin tone under theoretical pure white light (an even distribution of all wavelengths) to that of a Phosphor White LED demonstrates, absent these wavelengths the skin tones look pale under LEDs because light reflected by the skin tone is likewise absent these critical long wavelength colors.

 

 

Reflected Spectral Distribution of Caucasian skin tone under theoretical White Light and Phosphor White LED Light

 

In the picture above illuminated by the Phosphor White LED, both the cyan/blue dress and the skin tone, don’t reproduce well because you can't get accurate color reflected from an object unless that color is in the light in the first place. In other words, if the light source doesn’t generate the color (cyan), it is not reflected by the object (the dress) and so the camera/film simply can’t record it.

 

For more details regarding the issues surrounding the use of LED lights in motion picture lighting see http://www.screenlightandgrip.com/html/emailnewsletter_generators.html#anchorHigh%20Output%20AC%20LEDs

 

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

Edited August 13, 2013 by Guy Holt

[Phil Rhodes]

Small errata.

 

 

 

skin tones don’t reproduce well under White Phosphor LED lights because of the steep drop off of high frequency colors (above the 600nm cut off)

 

Er, low frequency colours, surely.

[Carl King]

Thanks for the replies. This is great stuff, very helpful.

[Wyatt Garfield]

Great post, Guy.

I think what many people don't realize is that human skin is not a solid color, so gelling lights cannot make up for spectral spikes/gaps. Imagine you have an LED source that appears to output pure 5600K, but is in fact a discontinuous spectrum of wavelengths which are mixed to appear white. When you shine this light on a subject, a percentage of the light is reflected off the immediate surface of their skin, sending back specular highlights which are uncolored by the skin, and which read on camera as neutral. The remainder of the light passes into the many translucent layers of the epidermis, where various wavelengths are scattered, absorbed, or reflected back. This is where the trouble begins (as well as some conjecture on my part). The pigment in their skin cells may absorb some of the spectrum but reflect a green spike, while their skin pores, blood vessels, and creases may absorb the green spike but reflect a red spike with a vengeance. If any combination of these phenomenon is occurring, you get strange skin tones mottled with pink and green or other colors (hence the zombie effect). You can try to compensate with gels, but the relationship between the various colors in the skin has been shattered. If you add plus or minus green, you tint the highlights, potentially accentuating the difference between the highlights and absorbed skin tone.

If you're in a bind, makeup can help unify the mottled skin tones, and diffusing sources to soften the specular highlights can help (again helping to unify the discrepancy between highlights and skin tones). Also, I've learned to avoid gelling daylight LEDs to tungsten, as that seems to hurt the color rendition dramatically. I read somewhere that color dimmable LEDs have smoother curves because the multiple colored LEDs fill in some of the gaps, but I'm not sure if this is true. The worst LEDs I've used were some "Cool Lights" daylight 10x10s a few years ago, which lead to all these thoughts on how skin responds to uneven light emissions. Litepanels seem better, but still cause some problems (especially when gelled from 5600 to 3200). The color of Daylight Rosco Pads read much different on digital cameras than to the eye, which makes me think their CRI is not good, but I haven't seen anything too strange going on with them. Of course, test any new products before you use them, as these opinions are drawn from experiences with specific applications, and are not universal truths!

[Phil Rhodes]

I think what many people don't realize is that human skin is not a solid color, so gelling lights cannot make up for spectral spikes/gaps.

 

That's nothing specifically to do with skin tone. You can't fix spectral deficiencies with gels because gels can only remove specific colours, not add it. If the light source simply doesn't have any (for instance) turquoise in it, you can't add any with filtration and colours which reflect any (for instance) turquoise can never look right. As Guy's post makes clear, combining the reflectance profile of the subject with the emission profile of the lighting device can be quite alarming, but that's not solely a skin tone problem. Those shots in Guy's post demonstrate that turquoise is a huge problem too.

 

I'm also trying to encourage people to avoid discussing CRI. It's notoriously inadequate for characterising discontinuous-spectrum sources such as LEDs. Manufacturers who don't publish TLCI data - which may not be perfect, but it's very certainly better - need to be heckled until they do.

 

If by "color dimmable LEDs" you mean RGB clusters designed to allow colour mixing, they're a complete disaster from a colour quality perspective when producing white light. There are movie lights designed to emit only white light which include amber, red, and other colours of LED, which sometimes seem to provoke better CRI numbers. Whether this helps in reality I'm not so sure - coloured LEDs are often quite monochromatic, which means you're adding spikes to a spectral output that's already problematically uneven. If the spikes happen to be in a place where the CRI test patches happen to be quite reflective, that might help you pass the test more readily, but it's still going to be easy to find things that reflect monochromatically and may fall between the spikes.

 

The best current white LED designs use a collection of white emitters using different phosphor blends on each, with the result that the lion's share of the output consists of light produced by phosphors, which emit a broad spectrum of radiation, not LEDs, which emit a sharp spike. Arri tell me the L7 series work like this.

 

P

[Wyatt Garfield]

 

That's nothing specifically to do with skin tone.

 

Agree 100%.

 

However, skin seems to be particularly revealing of spectral deficiencies. This is due in part to the fact that we know how healthy skin should appear, but it also has a lot to do with the complex quality of skin. Polluted skin tones cause a lot of strife on set and in post, and I think it's helpful to recognize that a subject's face contains many tones (specular highlights, lips, skin, creases, bloodvessels, etc), all of which react differently to spectral deficiencies. With each tone reacting differently, the natural relationship between them is lost, causing an unnatural appearance. While solid opaque colors (like the chips on a macbeth chart) can be easily keyed and corrected in post, the complex detail of skin is nearly impossible to make true in post.

 

As for my vague reference to "color dimmable LEDS", I was thinking of units like the Arri LoCaster, which I have not used. I would love to hear if anyone has done any spectral analysis of those units. Also, I would imagine you're completely right about RGB LED clusters being a complete disaster for color rendition.

 

Interesting to think about spikes passing the CRI test swatches, only to hit a monochromatic surface in practical application... I didn't realize that's how the CRI rating system worked. The more I learn about the CRI system, the less I like it! I'll only refer to it only when discussing it's shortcomings from here on out.