Do you know of any good compact fluorescent bulbs? By "good", I mean that the CFL bulb doesn't make colors look all weird, especially prints!
I posted a comparison of the X-Rite Color Checker under different lights recently; the composite image is at [a href=\"http://www.pbase.com/andy_fraser/lighting_comparisons]http://www.pbase.com/andy_fraser/lighting_comparisons[/url] The objective of the test was to see if the spectrum spikes of various CFLs would be visible on the colour patches, and to compare their overall colour rendition to the reputedly good Solux lamps.
The image is a collection of exposures of the chart under the specified lights (flash, household CFL, photographic CFL, daylight CFL, household halogen, Solux 4100, Solux 4700). Once the total image was put together it was white balanced using the white square with flash exposure.
The colour temperature differences can be compared by examining the first few "gray" columns, and between any two lamps, compare for abnormal colour shifts in the colour patches along their rows. The colour balance between rows is certainly apparent. The colour errors on patches are sometimes subtle.
The 60W daylight halogen manufactured by Globe (6th row) is the most obvious weird lighting to start with, especially it's effect on the purple patch.
Some notes on how this total image was generated:
- The photos were taken autoexposed with a Canon XTi
- Each raw file was converted to a tiff in LR without any change in whitebalance
- The rows of all the tiffs were cut&pasted into one total chart in CS3
- The first white square of each row was replicated and placed at the left end, and that square was blurred with the Average tool (I really only needed this for the flash, but did it for every row anyway)
- The image was flattened, saved as tiff, and brought back into LR
- The white balance was set using the averaged white square under flash lighting
- The resulting file was saved as a jpeg
- The lamp was mounted on the ceiling, 30" back from the wall, and the top of the chart was 28" down from the ceiling.
- An averaged white square had to be used to set the WB, since the R,G,B values are not exactly the same at every point across the original square.
- There is a small exposure variation between any two particular rows that has not been adjusted. This is at most 5%. Since the R,G,B of each light are not in the same proportion, I hadn't decided how to correctly calculate what exposure compensation to apply. However, the rows can still be compared since what we're looking for are any individual patches that deviate more than the rest of the row.
- A test image containing a larger number of more specific colours patches than the X-Rite chart, possibly using paints or pigments would be good for a second version of this test since I expect the difference between a flourescent and a good halogen or flash should be apparent.
- The $5 Globe CFL compares reasonably well with the $15 Cameron "Digital Imaging" CFL. It's distributed by Cameron, sold at Henry's and Vistek in Canada. It's package is marked 5000K, true daylight, 26W, 1600 lumen, type 5071015.
- The Canon flash is of course neutral since that's how I set it in Lightroom.
Some personal opinions and conclusions:
- MR16 halogens provide a better light due to more intensity in beam than CFLs which spill light everywhere
- The solux have a more even intensity across the width of the beam than the generic MR16 (not on the chart) I bought from a hardware store.
- The 4700K Solux is too bluish for residential use, it's too cold or "technical" a light. I like the 4100 better. There is a 3500, I have not tested it.
- The overall effect of light from multiple MR16 beams on the wall is more pleasant than the CFL lamps used previously.
- A minor requirement of artwork lighting in the home is that the wall itself is evenly illuminated, and the 24 degree beam (shown in the setup photo on pbase) is too narrow for that.
- I've ordered eight of the Solux 4100K, 35W, wide (36 degree) floods.
- Until the Soluxs arrive, I have $2 Ikea MR16s installed. They're not really "warm white", they're more like "beige" compared to the 4100.
- It would be interesting to redo this test, but with a test chart providing a larger number of very specific colour patches that is still a valid representation of the colours encountered in nature, but would also be capable of showing any real effect of flourescent spectrum spikes.
(The pbase site allows comments to be left; probably not useful since I don't want to admin that ro maintain a discussion over there)