Most of us learned in school that we do not actually “see” objects around us, but instead the light reflected by the objects. While the reflective properties of the objects are important in determining their perceived color, equally important is the light source illuminating the objects.
A recent article on CNET, “Shining a light on high-CRI LEDs,” examined this with respect to certain LED lamp products (Cree TW and GE Reveal) and indeed shed some “light” on the issue, but there are a few more aspects, not investigated, that are also very important to consider. So, I feel compelled to shine some “more light” on the subject here:
Color Rendering: The CNET article illustrates what many of us have known for a long time, that is, the standard Color Rendering Index (CRI) does not give the full story when it comes to color rendering. Indeed, the article states that while the Cree lamp claimed a higher CRI, the perception was that the GE lamp had better performance.
The main reason is that the number of “test colors” for CRI (based on only eight pastels) is simply too low and the metric can be “gamed”. When the CNET folks used a spectrometer to look at more test colors, a more accurate color rendering capability comparison was revealed, consistent with their perception (that the GE lamp outperformed the Cree lamp). An especially important test color to consider is “R9”, which corresponds to deep red colors which are important for human complexion, for example.
Until a better, more widely accepted metric for color rendering is established, one should ask for and keep R9 in mind (>50 ok, >90 preferred) when comparing light sources for illumination. A proposed method for considering all real-world colors was recently published by Soraa.
Whiteness Rendering: The CNET crew did not provide a detailed comparison of white objects. If they had, and compared either Cree or GE’s lamp against a standard incandescent, they would have observed a lack of whiteness in any objects including optical brightening agents (OBAs), such as clothing, paper, plastics (even human teeth).
These objects would appear comparatively dull and dingy under the Cree and GE lamps, which are based on blue-emitting LEDs, compared to the bright white we are used to under incandescent light. Short wavelength light (violet or shorter) is required to excite OBAs, and blue-based white-emitting LEDs cannot produce it, presenting a serious limitation in the vast majority of LED based illumination products on the market today.
The visual whiteness effect is very strong, as was demonstrated in recent experimental work by Prof. Kevin Houser’s group at Penn State University.
Color Point: The color of the light source also has an impact in lighting perception. The Cree TW and GE Reveal lamps each use neodymium-glass filters to notch yellow emission and improve green/red color discrimination. In addition to introducing unnecessary loss into the lamp, the other issue is that the color point of the lamp itself is shifted off-white (about 3 points in u’v’, the CIE 1976 color space).
This shift (for both lamps) is towards lower v’ which gives the lamps a pinkish tint. This is not necessarily a problem in lighting design, if all lamps are designed the same, but in a real-world application where one needs to mix and match different lamp types, having tinted lamps as a part of the design can be distracting, at least, and an eyesore, in the extreme.
Many light sources introduced after the incandescent lamp achieved their higher efficacy in part by “tricking” the human eye with partial / discontinuous and truncated spectra that the eye interprets as white when looking into the source. However, the deficiency of this spectral trickery is revealed when the light is reflected by objects and shortcomings in color become apparent. In the end, truly natural color rendering can only be achieved with a spectral distribution that is full and without discontinuities.