Visually measuring CRI

[Holly Snyder]

The other night I was setting up my two cheap compact fluorescent lights for taking photos of my paintings. With the lights reflecting off our white wall, it was obvious that one appeared to have a pinkish cast and the other had a greenish cast. The lights are from two different manufacturers, one is 5000 K, 82 CRI, and the other 5100 K, 84 CRI. Even though the color temperatures are very close, with the CRI being so low, we figured that's why you could see the difference in color. My husband and I were talking (both of us have an electrical engineering background), and wanted to make a spectrometer for artists to measure the color of their light source. (They do exist but they're upwards of $1500.)

Then we realized that we had a diffraction grating, which separates white light into visual wavelengths. We started playing, and were excited at what we found. With the grating in between the camera and our light source, we were able to record a visual spectrum of different light sources. An incandescent bulb, which I think is generally around 100 CRI, showed a smooth spectrum of color on the grating, like a rainbow. A fluorescent, whose coated phosphors emit varying amounts of light at different wavelengths (i.e. lower CRI), showed discrete images of the bulb corresponding to each of the red, orange, green, cyan, and violet phosphors. An image of the two is posted.

I should note that the digital camera images are a bit different than what our eyes see, as they are also subject to the filters in the digital camera sensor. In the left side of image #1, the pictured spectrum of the incandescent source shows bands of red, green and blue wavelengths (the camera's filters). In reality its a beautiful continuous rainbow. Similarly in the right side of image #1, the spectrum of the fluorescent bulb in reality also clearly shows orange and purple discrete images (that don't show up in the digital image since they are between or beyond camera filters). If you look at image #2, and wonder why you can see orange and violet in the images, it's because the grating was close to the camera (to take our setup shot), and the camera is mixing the bands together.

You can probably guess that if a paint color has a high reflection in one of the dark areas of the spectrum of your light source, it will look substantially different than natural light. When I get some higher CRI bulbs, I'll post some new pictures. For now, I'll take pictures with incandescent bulbs, and color correct the images with a gray card in PhotoShop.

The upshot to all of this is, if you want an inexpensive qualitative way of measuring the CRI of your light source, all you have to do is hold up a diffraction grating in front of your light. A higher CRI should have more continuous colors. By the way, you can order diffraction gratings from www.edmundoptics.com, part # 40267, two 6" x 12" sheets for $9.60. It's pretty cool.

Holly

[Garth Herrick]

Hey Holly,

Thanks for the tip! I ordered my diffusion grid sheets from EdmundOptics.com a couple of days ago and bingo, they had already arrived yesterday! A lot of packaging for a little sheet. The cost was $19.27 with shipping.

Anyway, I have had some interesting revelations as to the shortcomings of my studio lighting. I can't believe how fluorescent lights are composed of just a few simple steps on the spectrum scale. It is no wonder that so many colors can look so dead. The so-called full-spectrum incandescent bulbs seem to completely lack the production of anything in the yellow range (as do the fluorecents). My 500 watt photographic halogen lamp has a full spectrum 100 CRI, but it is 3200 K, rather than daylight. I wonder how the so-called 98 CRI full spectrum fluorescent tubes look?

Looking at a television or computer monitor, it is obvious there are only three phosphers, i.e.: RGB.

Thanks again, whenever I look at new lighting, I take one of these sheets along for a quick analysis.

Garth

[Holly Snyder]

Garth,

Thanks for posting your results with the diffraction grating sheets, they're interesting. I'm hoping to purchase the Lumichrome 5000K, 96 CRI fluorescents pretty soon, I'm curious to see their spectrum too.

You can also take two pieces of cardboard, hole punch one, then tape them together (with the hole now in the center). No light should get through the tape joint (black duct tape works great), and preferably the cardboard would be black on one side. Hold this between your light source and the grating, with the black side facing you. On the grating you'll see a smaller, but easier to look at spectrum, with a nice black background to set it off. You'll probably have to move the grating up and down and front and back, until you find where the spectrum image appears.

I'm glad this was useful to you.

Cheers,

Holly

[Holly Snyder]

Well I finally took some photos of the spectrums of my new Lumichrome T8 fluorescent tubes and my Sunwave compact fluorescent. Overall I would have expected the spectrums to be a bit smoother for these higher CRI bulbs, closer to the incandescent spectrum in the first image I posted. But this is all been a learning experience. Again with the previous posted images, my camera is only recording the red, green and blue, in reality there are purple and yellow visible on the diffraction grating for all the spectrums shown.

In any case, the first image is a comparison of the Lumichrome T8, 5000K, 96 CRI bulb to the Lumichrome 6500K, 98 CRI bulb. For these images, the grating was 10' from the light, and the camera was 14'3" from the light. The red phosphors on the 6500K have a smoother gradient than the 5000K, and I think the green is smoother and doesn't have the black gap (missing light) between the green and red. That's to be expected, as the 6500K has a higher CRI. However the blue gradient of the 5000K appears to be smoother, which I don't really understand. If I had a general purpose, low CRI, T8 fluorescent around the house I would test that also, but we have all T12's.

The second image compares a 5000K, 82 CRI compact fluorescent to Sunwave's 5500K, 93 CRI compact. In this case the diffraction grating was 4' from the light, and the camera was 7'9" from the light. It's hard to tell much difference between these images, so I'm not sure what conclusion to draw.

My husband may have access to a spectrometer at some point, which would display a spectral distribution curve. If that happens, I'll post some more results.

Holly

[Garth Herrick]

Holly,

Thanks for your thorough and enlightening report on the Lumichrome fluorescent lamps. You have given me the courage to share my own light tests for comparison.

I am including a similar shot of my new Just Normlicht Color Control daylight 5000 fluorescent lamps, and for comparison the earlier Phillips lamps I had installed before. I was not able to get as far away from the lamps with my camera as you. My camera was about eight feet away, and the diffusion screen was about two feet closer. For the Just Normlicht shot, I had only one lamp lit, while for the Phillips shot, two were lit, which is obvious in the image. I too am surprised that the digital camera can only see the red, green and blue light in the spectrum. To my eye there was much more. What the camera cannot see are the deep intense blue-purples, the brilliant turquoise, the yellow range, and the deepest garnet reds.

Despite the shortcomings of my photography, the Just Normlicht T-8 5000 K lamp seems roughly comparable in its CRI index to the Lumichrome T-8 lamps in your test. The salesman claimed the Just Normlicht has a CRI of 98. Compared to the Lumichrome 5000 K lamp, it appears to be at least as good, as it looks to have the smoother spectrum graduations like the Lumichrome 6500K lamp.

Again thanks,

Garth

[Holly Snyder]

Garth,

Cool. The Just Normlicht spectrum does look quite similar to the Lumichrome's, but a bit better (smoother). Could it be that the spectrums are just a bit out of focus? If you used autofocus and focused on the diffraction grating edge, the image would be slightly out of focus, since these are kind of virtual images that exist somewhat behind the actual grating. I manual focused on the images as best I could.

Interesting comparison to the Philips. The Philips is much less smooth, that is, a lot of black gaps. It would be a fairer comparison however if you only had one Philips light on. Maybe we shouldn't have done this, but our ballast is for two lights, and my husband just rotated one light off briefly while I took the Lumichrome pictures.

In any case, these images don't do justice to the spectrum in reality, as the blue-purples, turquoise, yellows and deepest reds are missing. I think we have the same camera Garth, the D100? It is disturbing the camera can't capture these colors, but not completely surprising. The camera separates the light through red, green and blue filters before reaching the CCD. Maybe someday the technology will change and the digital filter system will be redone and based on the Munsell system? I'm not sure how these images would look if taken on a film-based camera?

For comparison's sake, I took a picture of the spectrum of northeast light (not that is really matters the direction of light) seen through a pinhole. The pinhole was just a piece of cardboard with a hole from a hole punch. You can see the smoothness of the gradient in the image below, but once again, in reality the image is a smooth (and perfect) rainbow of colors. At least this allows a benchmark comparison of the artificial lights against northlight, seen through the skewed eyes of a camera. Below is also a picture of the setup, with the pinhole in the cardboard surrounded by several pieces of dark cloth held in place with clothespins, so no light will interfere with the image. The grating was 5'4" from the point source of light and the camera ~9' from the light source.

The other artificial light images probably should have been taken using this point source method, as you can see the gradient better. Maybe I'll redo all of them, and place them side by side along with the northlight picture for comparison's sake.

Holly

[Garth Herrick]

Holly,

This is interesting and revealing. Even the pinhole spectrum of natural north light seems to lack yellow in the digital camera capture.

Holly, I was afraid to show my diffraction grid pictures until you showed yours first. The camera really didn't do the spectrum any justice. Seeing that your pictures looked about the same as mine made me realize the fluorescent lamps I bought really weren't all that bad after all.

Your photos are better because you were farther from the light source than I was and you had a black backdrop to photograph the spectrum against. Last night I did't have a handy black background, and I took the left half of the photo against a foil insulation ceiling in the shadow of the light. I took the right half several weeks earlier before I changed the lamps.

Actually I did manually focus on the spectrum, but for some reason it never quite appeared to be able to be in focus as well as my eyes. I was in a rush to add a post to your thread last night, so my photo was quick and dirty. I held the loose diffraction grid in one hand slightly arched so it would not collapse, and tried to focus and shoot the picture with my other hand. all while laying on the floor. So no wonder my photo looks worse than yours. I also slightly rotated one tube to shut it off for the photo. I hope that in the short term the dual lamp ballast is not harmful to just one lamp lit.

Thanks for initiating this spectrum analysis of (nearly) full spectrum lighting.

Garth

[Richard Monro]

Has anyone done any spectrum work on the Philips F32T8 TL950 fluorescent bulbs? They have a CRI of 98 and seem to be used extensively in the print industry. They also appear to be less expensive than other high CRI lights.