Quite a while back, before I learned the proper terminology, I wrote about the metameric failures here and here that occur when using non-tungsten light sources. These sources have become increasingly common and popular due to their low power consumption and low heat output, and solid state — LED — lighting forms the basis of my DIY lighting project exactly because of those reasons.

Since cheap LED light strips are the focus of this project, it seemed useful to make at least some rudimentary tests on the color quality of light. Since I don’t yet have the neutral or cool white LEDs that I intend to use for the final panels, I’ll have to revise this in the future to deal with that, but rudimentary tests now at least make some considerations to the quality of light that can be achieved.

For these tests I used my i1 Display Pro’s ambient light reading capabilities to make brightness and color temperature measurements of various sources using the spotread program that’s part of Argyll CMS. The Edison based lamps were mounted in a Smith Victor A80 UltraCool and positioned such that the front of the lamp was 36“ (914mm) from the front surface of the i1 Display Pro. The LED test panel was placed with the front surface of the LEDs at 36” as well.

The lights were allowed to warm up for approximately 1 minute before measuring. Admittedly, 1 minute may not have been enough time for all of the lights to reach a warmed-up steady-state condition before testing, so some more research is warranted there.

Spotread provides information on the illuminance, correlated color temperature, and Delta-E (color difference) between the measured color and the actual color temp associated with the CCT. These values are summarized in the following table — only the 1 minute warmed up value is used.

Before I go any further I want to make a couple of comments. The light fixture that I was using was designed fro ECA photo floods, as such the filament is properly placed in the reflector so the fixtures focuses the light much better when being used with those lamps. Compact fluorescents and screw in LEDs don’t have the same design and so don’t focus quite as well. Compounding the situation, the Westcott 50 W CFL is so large, that it physically protrudes from the front of the fixture. Ultimately what this means is that the illuminance numbers should be taken as a rough guide and nothing more.

That said, one thing that’s always bugged me about CFLs and LEDs is that while they claim to produce the same lumens as a tungsten lamp, they almost never appear to actually do when I’m looking at a scene lit by them. For example, Westcott markets the 50 W CFL used here as being equal to a 250 W tungsten in lumens output. However, in this test, I saw less than 1/4 the lumens.

Now don’t get me wrong, this was far from a well controlled scientific test so the difference may not be nearly as pronounced if for example the fixtures were optimized for each of the light sources — or I used a bare bulb that was equally poorly optimized for all the sources.

Of all the lights, I was very impressed by the performance of the KinoFlo CFL. Though the color temp measured quite a bit higher than the 5500K that they claimed, the Delta-e was very good for the color temp it actually produced.

I was also quite impressed by the LED strip lights. For being $1/foot lights, they produces a pretty good amount of light; the 4’ on my test panel almost matched the 26 W KinoFlo CFL in terms of light at the sensor. Moreover, while they were a bit low in color temp, 2849K instead of 3000K, the delta-e was second only to the KinoFlo in this test.

That said, from looking at tons of spec sheets for LEDs, warm white LEDs do tend to produce pretty good results and run at pretty high CRIs. Neutral (4000K) and daylight (5000K) LEDs both tend to run much cooler (some of Cree’s daylight LEDs can hit CCTs as high as 10,000K) and there tends to be a fall off in CRI from mid-to-high 90s to mid 70s to low 80s with the cooler LEDs.

There are two other comments I want to make on the test. First, is that I used a colorimeter not a spectrophotometer. With a spectrophotometer, like X-Rite’s i1 Pro, I could start looking at spectral power emission graphs which paint a much better picture of the actual light output. Unfortunately, the costs for the i1 Pro are currently prohibitive, and while the ColorMunki Photo may be a workable substitute, it has a UV cut filter which has other implications in their primary role.

The other test, though much less accurate than generating a spectral power distribution, is testing with a calibrated color target like a X-Rite Color check card^{(Affiliate Link)} along with a verity of natural and synthetic materials. It certainly would be cheaper than a spectrophotometer.

I’m going to wrap this up here for the time being. As I said at the start, this wasn’t an exhaustive or strongly controlled test, just a cursory look. At the cursory level though, I don’t see any overt issues that preclude moving forward. That said, I ultimately would like to conduct further testing on this, and will at a minimum need to quantify the performance of the neutral and cool LED strips to make any kind of informed decision on them, let alone properly filtering them to tungsten if I should choose to do so.