Displays use the wrong colours indigo and violet due to the lack of elements able to display violet. Colour spaces generally take this into consideration by omitting short wavelengths. A prism is able to produce a continuous wide range of visible light at a given luminance level and high saturation. Using prisms alone would lead to variation in hue according to viewing angle.

To get round these problems, i suggest a display made up of pixels constructed thus. There are two sources of white light whose brightness can be varied. One is passed through a prism whose angle can be varied by rotation, the other through a polarised filter and a layer of monochrome liquid crystal. These two lights, one coloured, the other white, are combined through a short length of optic fibre ending in a hemisphere visible as part of an array of similar pixels on a flat surface.

The result would be a display consisting of pixels whose hue can be varied throughout the visual spectrum via the spectrum, along with saturation by proportion of white light passing through the liquid crystal, and brightness by how intense the light source is. The second light source needs to vary in order to saturate the first to the right degree. The display would also be close to Lambertian - similar from a wide range of angles - due to the hemispheres and the fact that light from the prisms cannot be viewed directly.

The colour of the light from each pixel would not correspond to most colour spaces because of the absence of non-spectral colours and the presence of genuine indigo and violet. However, non-spectral colours could be provided in the conventional way by adjacent pixels. Actually, i'm not aware of any colour space that incorporates spectral indigo and violet. An alternative genuine spectral colour space could be defined by wavelength, saturation and brightness.

One thing i don't like about this is the fact that there are moving elements involved. I wonder if there is a way of obtaining a spectral colour through elements of the order of a wavelength of visible light, but i know we don't like nanotechnology here so i won't go down that route.
-- nineteenthly, Jan 01 2008

Color Gamut http://en.wikipedia.org/wiki/Gamut
A nice writeup about the range of possible colo(u)rs [csea, Jan 02 2008]

DLP / DMD info http://www.dlp.com/tech/what.aspx
Digital Light Processing / Digital Mirror Device [csea, Jan 02 2008]

Materials with refractive indices varying according to voltage http://www.patentst...7151736-claims.html
They exist, examples here. [csea, Jan 02 2008]

Reposted link to CD photo http://i122.photobu...y/falsespectrum.png
Spectacularly awful rendition of violet in this image [nineteenthly, Dec 22 2008]

OLPC XO-1 http://en.wikipedia.org/wiki/OLPC_XO-1
Display which uses a diffraction grating rather than filters to produce RGB; may be a good starting point. [spidermother, Dec 24 2008]

Another kind of spectral monitor http://uk.youtube.com/watch?v=X-iJbFvFD4c
Another unconvincing display [nineteenthly, Dec 24 2008]

//The display would also be close to Lambertian due to the hemispheres and the fact that light from the prisms cannot be viewed directly.// This definitely needs explaining for the likes of I.

But, I get the general idea. I'm pretty sure there are materials whose refractive index depends on an applied voltage - would these help?

But also, I don't entirely understand why these short-wavelength colours can't be generated by conventional approaches. Ultraviolet LEDs exist, so is it not possible to produce violet/indigo LEDs? Or, if not, use UV LEDs with suitable fluorescent coatings to give indigo/ violet. Or have I missed the point?
-- MaxwellBuchanan, Jan 01 2008


Thanks for the point [Maxwell]. A violet LED would have to be right at the limit of human vision to provide the whole range of the spectrum, but a prismatic approach would be simpler because only one device is required to provide the hue element.

I think refractive indices varying according to voltage would definitely be a big help. This system just uses a single transparent substance such as glass to control the actual hue though.
-- nineteenthly, Jan 02 2008


"Lambertian" is just a fancy way of saying "diffused," though it has some rather more technical meanings.

I've provided a [link] to a good way of considering what colors are visible to the human eye.

If I understand the idea correctly, it could be achieved using a variant of TI's DLP / DMD technology [link2] using nano-prisms.

Probably way too complicated to be commercially feasible (and therefore prime HB material!) +

I don't think we have to be too circumspect about nanotechnology that is a relatively minor tweak to existing technology. It's 2008, after all!
-- csea, Jan 02 2008


(+) nice.
-- ConsulFlaminicus, Jan 02 2008


Sorry, i tend to make a fetish out of long or technical-sounding words. Thanks for the links: interesting stuff.

If nanotechnology isn't tabu, what about some variation on nanoflake paint?

When does nanotech become magic?
-- nineteenthly, Jan 02 2008


I still don't see why indigo/violet LEDs can't be made or, if not, why you can't use UV LEDs and coat them with indigo/violet fluorophores.
-- MaxwellBuchanan, Jan 02 2008


// When does nanotech become magic ? //

Strictly speaking, it doesn't. It's a grey area in the HB.

The criteria is precision. If you propose an idea which uses genetic engineering, nanotechnology, or other cutting-edge techniques in a vague and hopeful way i.e. "Make Unicorns by genetic engineering" you are likely to get slapped down good and hard, possibly getting the idea m@rk3d f0r de13t1on.

But if your are explicit enough, for instance, "Give white horses horns on their forehead by modifing Site 23A17-Alpha in Chromosome 14, splicing in Gene Z4-kk23-Blue from a North Atlantic Narwhal" they you are quite likely to get a fair hearing.

Others may differ in their view. You could ask She Who Must Be Obeyed directly for a ruling.
-- 8th of 7, Jan 02 2008


All hues can be displayed at a reasonable level of saturation using RGB. Adding one or two more primary colors (e.g. cyan and violet) would increase the saturation that could be achieved for hues near the added colors.

Your approach as described would be incapable of displaying some hues in the red-magenta range at any level of saturation, since the eye will not perceive such hues when exposed to any single wavelength of light; adding white light won't help.

In theory, if you were to double up pixels so that each would feature an arbitrary mix of two arbitrarily-chosen wavelengths, that could work (note that one can perceive white from the right combination of two wavelengths). In practice, though, I suspect that it would be hard to avoid significant perceptual differences between colors whose saturation was low enough that hue should be almost irrelevant.
-- supercat, Dec 22 2008


Currently, there's a problem with displaying violet. There would be sacrifices with magenta, but the reverse approach to RGB could be taken - violet replaces magenta rather than the other way round. It could also be dithered.
I think losing magenta is a price worth paying for true violet. I posted a link to a crappy photo i took of a CD, where the violet was way off. It clashes horribly with blue. Red to blue looks OK, but beyond blue the hue on RGB simply veers off in a weird direction which bears no relation to human colour perception. It's a glaring inconsistency.
-- nineteenthly, Dec 22 2008


So a picture of a spectrum was taken with a camera whose spectral response at shorter wavelengths is weird. What does that prove? That you have a dodgy camera?
-- supercat, Dec 24 2008


I´ve never seen a non-weird picture of a spectrum. The camera is indeed dodgy, but the "colour bar" appearence of the "violet" is telling. There´s something horribly wrong there.
-- nineteenthly, Dec 24 2008


So, not something for watching ghosts with?
-- coprocephalous, Dec 24 2008


Sadly no, but see link.
-- nineteenthly, Dec 24 2008



random, halfbakery