Best Colour Spectrum of lights

[Pickwick]

Hi all,

I take photographs in some very dark areas, and was wondering what sort of lights provide the most accurate colour spectrum.

I saw one post on here that said incan's provide it better than LED's (I use incans). Is there another type that I should know about?

Thanks in advance.

[jtr1962]

I made this a while back but maybe it can answer your question:



Despite the 100 CRI of incandescent some of the colors are hard to distinguish from one another. In all cases I set the white balance on the camera to sunlight so that the difference in color temperatures was apparent. In all cases colors are distorted a bit from natural daylight. The GE Chroma 50 appears best at rendering colors but the Paralite Maxum 5000 is almost as good. The Ushio 850s, which are basically commodity tubes costing only a little over $2 in lots of 25, are quite decent. It actually takes a trained eye to distinguish them from the more expensive "full spectrum" GEs and Paralites. The Philips TL741 is noticeably poorer than the other tubes. The incandescent is just awful in all respects. Many colors just don't show up at all. You don't even really have a white point. As you can see CRI isn't everything when it comes to lighting. Also note that the camera makes the distortions appear worse than in person because the eye compensates somewhat for the short comings of each type of light.

[winny]

Originally Posted by Pickwick

I saw one post on here that said incan's provide it better than LED's (I use incans). Is there another type that I should know about?

That's correct. Incans (halogens included) have a CRI-value of 100 and thus have perfect color rendering. LEDs have between a CRI-value of 70-80 and therefor some colors do not render properly.

[brickbat]

Couple things to keep in mind. First, as JTR mentions, although any incandescent has a CRI of 100 does not mean it's a perfect source, or that it will render colors more accurately than, say, a 4100K source with a CRI of 78. CRIs are only to be compared when the source has the same CCT.

Thing 2, assuming your're using a digital camera, you have color correction at your disposal. Take advantage of it. I wouldn't be surprised if sources that have spectra without bigs peaks yield the fewest 'surprises'. IOW, although halogen or incandescent will definitely yield yellowish pics, correcting the color may be pretty straightforward.

Thing 3. Why not use a xenon flash? They have a high CRI and 'natural' CCT.

[2xTrinity]

That's correct. Incans (halogens included) have a CRI-value of 100 and thus have perfect color rendering. LEDs have between a CRI-value of 70-80 and therefor some colors do not render properly.

At this point though, as shown in the image above, CCT is not the whole story. An ideal hotwire would run at around 5000k, at which point it would look exactly light sunlight. However, that's about 1200k over the melting point of tungsten. As it is, incandescent put out a lot more red and yellow, but less blue. White LEDs right now are generally fairly poor for color rendition -- they use a blue LED to activate a yellow phosphor, and have big gaps around the blue-green, and red portions of the spectrum. However, that type of source has the potential to be the best, by either combining muiltiple emitters (this would in theory be the most efficient) to fill in the gaps, or by using multiple phosphors like the higher end fluorescent tubes.

Although for lighting photographs, one benefit about using incan is that it will vary from natural sunlight in a specifically predictable way, if you know the color temp, then you can use software to correct it, while fluorescent sources can vary dramatically from one to the other (at this point, I'd consider white LEDs by definition a fluorescent source) However, for lighting a workspace, I'd definitely go with a cool fluorescent.

[winny]

Originally Posted by brickbat

CRIs are only to be compared when the source has the same CCT.

I understand what you are saying, but we need to separate two things.
CRI does not take into account what CCT you have, and it shouldn't. That would be like "I'm pulling 5 A though this wire, but it's only from a 2 V source, so it's only 5 small/light/soft amperes.". That's perhaps true, but the cable will certainly not mind. It will run just as hot.
The same thing goes with photography. Once you have calibrated your equipment for your CCT of X, the only thing that matters is how well it render colors (i.e. (or is it e.g. here? I'm bad at both English and Latin) CRI).

Accurate colors doesn't have anything to do with color temperature. I would agree that some things look terrible when lighten with too high or too low color temperature even with a CRI of 100, but the colors will be accurate. But let's try to separate accurate colors (which is excatly what the CRI-figure is trying to show) and nice-looking colors, which is quite arbitrary.


The CRI-system does have its flaws, mainly as it only checks the color rendering of eight colors, and disregards any over rendering. If you are really interested in getting accurate colors, you should get your sources CRD (color rendering diagram). Then you can really see what properties the light have.

[brickbat]

So, Winny, are you saying that every source with a 100 CRI renders colors with more accuracy than a source with a lower CRI? Hard to fathom, but I'm not a color scientist. Just can't see how, for example a 2200K 100 CRI source is more 'accurate' than a 4200K 78 CRI source. There are real-world, practical limits on how much correction you can do in a digital camera or the human vision system, for that matter, right?

I suppose we'd have to agree on what we mean by 'accuracy' before going much farther, eh?

BTW, I think your english is pretty good. (e.g or i.e ? I can never keep them straight either...)

[jtr1962]

One thing I've noticed is that many cameras don't have particularly good auto color balance. Sometimes they key in on the wrong point and that throws everything off. I remember a picture I took under incandescent of a Christmas present in red wrapping. The wrapping paper was a vibrant red under both 5000K fluorescent and sunlight. It looked even redder under incandescent. However, in the picture it ended up looking brick red because the autocolorbalance corrected incorrectly. Even setting the color balance to "incandescent" doesn't necessarily guarantee a good picture since incandescents vary considerably in their color temperatures (a 100K difference when you're in the high 2000s is a much greater color shift than the same 100K difference when you're at 5000K).

For the reasons in the last paragraph I always take pictures with the color balance set to "sunlight". I can then correct in software with much more favorable results. Once color corrected, the incandescent colors are as good as anything else:



However, despite the CRI of 100, the corrected colors don't really appear noticeably better than either of the two full spectrum fluorescents without correction.

Yes, despite the fact that incandescent sources all have a CRI of 100 there are limits to color correction with both cameras and the eye. I find that under about 3500K or over about 7500K there is just no white point no matter how long I stare at an object. Also, once a significant part of the spectrum shifts outside visible wavelengths no amount of color correction will help because the reflectance of objects will be different for these wavelengths. Or put another way, color balancing is basically taking a picture, and shifting the colors by weighting different parts of the spectrum differently. However, if the source is deficient in certain wavelengths it may not work entirely. For example, you should in theory be able to color balance a picture taken under something glowing a cherry red and make it look as if it were taken under sunlight. In practice there just aren't enough blues from the source to get any kind of response from the camera so these areas will appear black in the original and black in the color corrected version. Additionally, the presence of a great deal of infrared in the source may reflect off objects and cause some sort of response in the camera. Depending upon how the objects reflect infrared, they could end up color corrected to completely different colors than they really are. The same type of reasoning applies to color correcting even under incandescent. Some colors may well appear off, although not by a huge amount. Color correction works well most of the time but one must be aware of its limits.

[winny]

Originally Posted by brickbat

So, Winny, are you saying that every source with a 100 CRI renders colors with more accuracy than a source with a lower CRI?

Appart from the fact that the CRI system only test eight colors so it can miss the rest 1+ million, yes. They are chosen with care so most of the time, the CRI-figure does tell you a lot. If we can play with the thought for a moment that the CRI-figure would test all the colors humans can see, then to answer your question, yes.

Originally Posted by brickbat

Hard to fathom, but I'm not a color scientist. Just can't see how, for example a 2200K 100 CRI source is more 'accurate' than a 4200K 78 CRI source.

That's a perfectly normal thought. It's bang on target was we discussed about good-looking colors vs. accurate colors.
I'll try to elaborate some here. 2200 K might be on the edge for this example it has very very little blue in it, but I think we can proceed anyway. If you would take JTR's color chart and measure all colors what they look like, and compare that with what they should look under 2200 K, you'll notice that they are all perfect. You will also notice that the further into the blue spectrum of colors on the chart you look, the fainter they look. You might even have to crank up the power of that 2200 K source (while still having the same CCT of course) in order to see them properly, but they will look exactly as they should.
With the 4200 K source (my favorite BTW ), the blue ones will not look as dim, but some colors does not look like they should under 4200 K, they might appear too green, lilac, orange or whatever, depending on what the CRD of the source looks like. Now, in your eyes, that might very well be better as you almost couldn't see the blue shades with the 2200 K lamp, and your brain can somewhat compensate for the sometimes strange colors, but if you wanted to photograph it, your camera can only do what you tell it to do. Using for example Photoshop, you can easily compensate for just about any color temperature (although 2200 K will be difficult as you don't have a dynamic range big/good enough to catch the blue shades well enough compared to the red ones, but with an hyper-proffesstional/ideal camera, no problem) very easily, whereas missing colors can be impossible to compensate for sometimes. If for example shade A of green would look as shade B of green, and shade B would look like shade B (a common CRI-error), you are impossible to tell on the film/picture what appears as shade B is in reality. Is it shade A or B? In other words, you are screwed.

To your defense (I'm actually good at heart you know ), you don't always need to tell the difference between shade A and B, depending on how close they are.

Originally Posted by brickbat

There are real-world, practical limits on how much correction you can do in a digital camera or the human vision system, for that matter, right?

Yes indeed! If your example would have been 3000 K and CRI=100 versus 4200 K and CRI=78, it would have been much easier.

Originally Posted by brickbat

I suppose we'd have to agree on what we mean by 'accuracy' before going much farther, eh?

That's a very good comment on just about anything. If we don't speak about the same thing, we might just as well be quiet. I'm under the impression that what the light industry refers as accurate colors is how good CRI-value or CRD something has. You are of course entitled to have another opinion on the matter.

Originally Posted by brickbat

BTW, I think your english is pretty good. (e.g or i.e ? I can never keep them straight either...)

Thanks! I do my best, but I still have problems with -s -'s and -s', not to mention -ys and -ies.


I hope that clarified some things for you!

[snakebite]

i have a few phillips f17t8/tl950 left from rebuilding light tables.
they claim 98cri 5000k
these were very expen$ive.
wonder how they would compare?
have some sylvania 950 that seem close but a bit lower in deep red.
it would be nice to get all these together for a test.

Originally Posted by jtr1962

I made this a while back but maybe it can answer your question:



Despite the 100 CRI of incandescent some of the colors are hard to distinguish from one another. In all cases I set the white balance on the camera to sunlight so that the difference in color temperatures was apparent. In all cases colors are distorted a bit from natural daylight. The GE Chroma 50 appears best at rendering colors but the Paralite Maxum 5000 is almost as good. The Ushio 850s, which are basically commodity tubes costing only a little over $2 in lots of 25, are quite decent. It actually takes a trained eye to distinguish them from the more expensive "full spectrum" GEs and Paralites. The Philips TL741 is noticeably poorer than the other tubes. The incandescent is just awful in all respects. Many colors just don't show up at all. You don't even really have a white point. As you can see CRI isn't everything when it comes to lighting. Also note that the camera makes the distortions appear worse than in person because the eye compensates somewhat for the short comings of each type of light.

[winny]

Originally Posted by snakebite

i have a few phillips f17t8/tl950 left from rebuilding light tables.
they claim 98cri 5000k

If you look at the CRD of that tube here , you'll notice that they blow the competition away. The price you pay for this is lower efficiency and high cost of the tubes themself. They are used in places where color accuracy is at an absolute premium.