Setting WB in Kelvin vs. other methods...
An article by Magne Nilsen (Feb. 4, 2004). Used with permission.
|Setting the camera to 5000 Kelvin is even better than using manual White balance, right? Unfortunately - No. There are several other factors than can contribute in messing up the very important aspect of getting both colors and neutrals to look "right" in your images. Wrong White Balance is one problem area, having a not-so-neutral light source could be an other.|
The Kelvin scale in color theory is based on the blackbody curve locus, which indicates what color a perfect blackbody radiator would reflect, if heated up at different temperatures measured in Kelvin. As a single parameter this is too limiting, or even outright misleading for all possible light situations, so the actual color temperature should also be annotated with a second parameter indicating how far away. and in which direction from the blackbody locus the actual value is. This is in the literature defined as a set of iso-temperature lines and a measurement in +/- delta uv units. As you see from the above linked image these offsets can go in either a more greenish/yellow direction or in a reddish/purple direction. Capture One indicates this with the "+/- Tone balance" slider. PS CS and ACR implements an equivalent function as a "+/- Tint" slider. When only a single Kelvin number is reported or used - that could actually be anywhere on the given Kelvin temperatures iso-temperature line, and that would not carry enough precision. Setting manual WB with a single Kelvin value could actually be quite erroneous, unless the temperature is the one exactly on the blackbody curve, unfortunately it seldom is.
If the +/- value becomes big (at least for the +/- uv values), this would mean you are moving further away from a natural light source, and when you get too far away you will have problems getting the whites and grays and even some colors to appear natural to your brain, and many manmade light sources can be quite far away from the blackbody curve.
To make it even more complex, even a a lamp marked as 4000 Kelvin does not then necessarily mean the exact neutral spot on the blackbody curve - but on top of that i.e. most Cool White Fluorescent lamps can have huge spikes in their spectral response, meaning that certain colors will reflect totally wrong. The only lamp I've seen and used that handles this is the Solux lamp. http://www.soluxtli.com/colorproofing.htm
Anyway, setting WB by single-clicking is always best performed in a RAW converter, since then all the rules can be obeyed in the underlying linear world of both the RAW data and the XYZ or Lab colorspaces where the theories are meant to be applied. Setting it later in i.e. Photoshop will often give inferior results, since the linearity is long gone. If you don't shoot RAW, your only real trustworthy option for critical work is to set WB manually before shooting, and always try to use use lights without strange spectral behavior.