The phosphors of modern fluorescent lamps are a tricky thing. We can’t control their spectrum at will. We can only take some phosphor that suits our needs out of the set of phosphors chemistry knows. The best one we can take has a 100-unit peak at a wavelength of 575nm (it’s yellow). And the 620nm red filter has a transmission factor of 1/10 of the maximum at 575nm.
What does this mean, ultimately? It means we’ve got two wavelengths instead of one at the filter’s output: 620nm with an intensity of 100 units and 575nm with an intensity of 100*1/10 = 10 units (the intensity in the lamp’s spectrum is multiplied by the filter’s transmission factor at this wavelength). That’s not a negligible value.
As a result, the “extra” peak in the lamp spectrum that passes partially through the filter yields a polychromatic color (red with a tincture of yellow) instead of a monochromatic red. It means that in the CIE diagram the corresponding vertex of the color gamut triangle is shifted from the bottom border upwards, closer to yellow hues, thus reducing the total area of the triangle.
Seeing is believing, so I’d like to illustrate my point. I asked for help from the plasma physics department of Skobeltsyn Nuclear Physics Research Institute and they offered me an opportunity to use their automatic spectrographic system. The system was designed to research and control the process of growth of artificial diamond films in SHF plasma by the emission spectrums of the plasma, so it should deal with an ordinary LCD monitor with ease.
So I turned the system on (the big angular box is a monochromator Solar TII MS3504i; its input port is on the left and a LED with an optical system is fastened next to that input; the orange cylinder of the photo-sensor can be seen on the right – it is fastened on the monochromator’s output port; the system’s power supply is at the top)…
I adjusted the height of the input optical system and connected the other end of the LED to it:
And finally I put it in front of the monitor. The system is PC-controlled, so the whole process of taking the spectrum in the range I was interested in (from 380 to 700nm) was accomplished in a couple of minutes.
The X-axis shows wavelengths in angstrom (10A = 1nm), the Y-axis shows intensity in some imaginary units. For better readability, the diagram is painted according to how our eye perceives the different wavelengths.
This test was performed on a Samsung SyncMaster 913N monitor, a rather old low-end model on a TN matrix, but that doesn’t matter. The same lamps with the same spectrum are employed in a majority of modern LCD monitors.
What does the spectrum show? It illustrates what I’ve said above: besides the three peaks corresponding to blue, red and green sub-pixels, there is some junk in the area of 570-600nm and 480-500nm. It is these unnecessary peaks that shift the vertexes of the color gamut triangle up to the center of the CIE diagram.