Hi Folks, This is my first posted halfbaked idea, so please forgive any mistakes in protocol. (No on second thought tell me or I'll keep on doing it.)

I think I need to start with some background. Current TVs and computer monitors have three 'colour guns', one each for Red, Green and Blue. Human eyes have 3 different types of colour detector (plus one which detects all colours which I'll ignore) called 'cones'. These colour detectors detect light over a range of wavelengths, but are most sensitive at a specific wavelength. Each type of cones detect light most efficiently at a certain wavelength, or less strongly the further away from that it is. So intermediate colours are perceived if light stimulates 2 types of cone. Sorry for that waffle, the point is that although 'Blue' is the peak of detection of the cone type at one end of the visible spectrum, the green cone also detects this colour a bit. So one can perceive as different colours beyond this (ie, indigo and violet), because they don't stimulate green as much.

My idea is therefore that one could have a fourth gun (and coloured phosphors in the screen) producing 'violet'. Such monitors could produce an even better display. Also, this has a fringe benefit in saving on wasted memory. I don't know if x86 based PCs are the same, but my RiscPC wastes 1/4 of the screen memory in order to retain memory word alignment for each pixel. Currently (in 16 million colour modes) 8 bits are used for each of red, green and blue, and another 8 are ignored.

(Thanks Jutta, Waugsqueke, everyone.)
-- Loris, Jul 26 2002

Colourblind monitors http://www.halfbake...lorblind_20displays
some relevant discussion with this related idea. [Loris, Jul 26 2002, last modified Oct 04 2004]

Differentiating Wavelengths http://webexhibits....usesofcolor/1C.html
Background: How the eye works, different cones, frequencies, where in the brain this is processed, ... [jutta, Jul 26 2002]

Diffraction-based display technologies http://www.eetimes..../951news/heart.html
EE Times article on display technology capable of producing pixels with any specified wavelength [wiml, Jul 26 2002, last modified Oct 04 2004]

squant http://www.negativl.../squant/photos.html
This is cute. Oh, and the idea fascinates me. But for this to be visually beneficial wouldn't humans have to differentiate purple much the same way that birds differentiate green or eskimos white? [reensure, Jul 26 2002]

Color Space FAQ http://www.faqs.org...ics/colorspace-faq/
Pretty much all you need to know about colors in computer displays and printing. [pottedstu, Jul 29 2002, last modified Oct 04 2004]

Cone sensitivity chart http://www.photo.ne...dscott/vis00010.htm
also has lots of other interesting bits. [Loris, Jul 29 2002, last modified Oct 04 2004]

J Scruggs Color Theory http://www.bway.net/~jscruggs/add.html
Describes color synthesis. [pottedstu, Jul 30 2002, last modified Oct 04 2004]

Color Space FAQ http://www.cs.uu.nl...colorspace-faq.html
Alternative location of above document. [pottedstu, Jul 30 2002, last modified Oct 04 2004]

link to fundamental colours on dictionary.com http://www.dictiona...undamental%20colors
[Loris, Jul 31 2002, last modified Oct 04 2004]

YARK's Color Tutorial http://www.yarc.com/colortut.htm
Has some gamut charts comparing human vision vith television colour models. Also some information about print colour models. . [amadeus, Jul 31 2002, last modified Oct 04 2004]

In your last sentence, you mean bits, not bytes. Your system likely uses 8 bits each for red, green, and blue, yielding 2^24 = roughly 16 million colours.

If I understand correctly, you are trying to more directly target the specific receptors in the eye and eliminate "spray". The details depend on the frequency response of the cells, but if it is improvable at all, I think you should be able to do that by moving the three existing frequencies around:

Now:
--- red --- green --- blue ---

Improved:
red ------- green ------- blue

rather than by adding entirely new frequencies.
-- jutta, Jul 26 2002


Alright, since you asked. Please don't start your ideas with "OK". I hate that.

Otherwise, I don't really understand much about this. It's not clear to me how having a violet gun helps.
-- waugsqueke, Jul 26 2002


I'm fairly certain that present monitor technology comes pretty darn close to tickling the photoreceptors of the eye in every (distinguishible) way that a true full-spectrum output could. Adding more guns and phosphors would not change things noticeably. Even [jutta]'s suggestion of spreading the wavelengths out (presumably so that the red and blue wavelengths are long and short enough not to tickle the green receptors at all) would only affect the most extreme reds and blues, and only negligibly at that. About the only way left to improve the output is to increase the dynamic range of intensity (brightness).

If you are really bent on using tetrachromatic color representations, perhaps you should first invest in genetic research to improve human sight by adding color receptors at other wavelengths.
-- BigBrother, Jul 26 2002


Actually, the "red" cones are stimulated somewhat by wavelengths between blue and ultraviolet. Interestingly, this means that while most hues can be perceived as a single wavelength of light somewhere in the visible spectrum, magenta and other hues between violet and red require two distinct wavelengths of light. BTW, although "white light" is generally composed of red, green, and blue, the proper combination of monochromatic yellow and blue will also appear white.
-- supercat, Jul 27 2002


This would have no effect. You can synthesis any colour using a combination of any 3 arbitrary "primary" colours, as long as the primary colours are reasonably distinct.

One variation would be if you consider printing. Normal printing uses 4 colours, cyan, magenta and yellow, plus black, because it's hard to mix the three colours to get pure black. However, some printers will use an additional colour for special effects: a bright orange, or gold or silver ink, for example. You could have a special screen with an additional phosphor which might sparkle or exhibit other optical properties (maybe be slightly inset or projecting), and the fourth beam could excite that.
-- pottedstu, Jul 29 2002


A lot of computer printers now use 6 colors in addition to black.

I just want a monitor that can do orange.
-- DrCurry, Jul 29 2002


"This would have no effect. You can synthesis any colour using a combination of any 3 arbitrary "primary" colours, as long as the primary colours are reasonably distinct."

Pottedstu, I believe this is incorrect. Or rather, the qualification does not apply to current monitors. (and to be pedantic the generalisation is false.)

Jutta has also made the very good point that one could space out the gun-colours more (ie move the blue to violet) and this would work fine. Have a look at the link I've just posted (Cone sensitivity chart) The top diagram shows a spectrum. I must admit I haven't checked, but I'm willing to bet that the colours to the right of blue are not in fact indigo and violet, but rather purple (ie red+blue). The payoff diagram is further down and shows the cones sensitivities at different wavelengths. Can you see that at the top of the blue peak there is still some activity from the green sensor? This is where the blue phosphors emit.

One further point I should have made before is that having violet as an extra gun, rather than a replacement of blue would allow backwards compatibility. When a program didn't know about the violet data it could work OK just using the other three.
-- Loris, Jul 30 2002


// This would have no effect. You can synthesis any colour using a combination of any 3 arbitrary "primary" colours, as long as the primary colours are reasonably distinct. //

Loris: do you have any evidence this is incorrect? All the sources I have suggest that the RGB system is sufficient to produce any perceivable colour.

As the links show, the actual perception of colour doesn't arise from separate red, green and blue components, but the red cones are sensitive to some green, etc. However we are still able to perceive a pure red colour and a pure green colour.

I think you are arguing that if more than one cone is excited, we are unable to perceive pure colour. Therefore, you claim that a system which excited only one cone would be superior to a system that excited more than one cone. However, the flaw in this is that we never perceive pure colour through only one sort of cone. All systems for generating colours, whether in inks or lights, work by mixing basic colours in the correct proportions. What matters is not individual wavelengths, but the blending of different components.

As to the claim // the colours to the right of blue are not in fact indigo and violet, but rather purple (ie red+blue). // what is the difference between purple and indigo and violet? Is it that purple is for you made of a mix of colours, while indigo and violet are pure? Because the human eye cannot perceive a difference between a single-frequency light and a blend. Although they differ in optical properties, it doesn't make sense to give them different colour names.

The link I posted doesn't seem to be working, so I'll link to another version of the same document.
-- pottedstu, Jul 30 2002


I also direct you to J Scruggs's short article on color synthesis. I think the crucial passages are:

// Additive Color Synthesis is the method of creating color by mixing various proportions of two or three distinct stimulus colors of light. These primary colors are commonly red, green, and blue, however they may be any wavelengths to stimulate distinct receptors on the retina of the eye. //

// All color sensations can be produced this way, including those red-blue mixes (purples and magentas) not found at any wavelength band in the spectrum. //

Is it this that Loris disputes?
-- pottedstu, Jul 30 2002


"Loris: do you have any evidence this is incorrect? All the sources I have suggest that the RGB system is sufficient to produce any perceivable colour."

Well, yes.. I've already linked to it (See the 'cone sensitivity chart link') The point is that the colour phosphors are not frequency seperated enough (distinct enough) to perceive the full range.

Thought experiment: Move the blue colour slightly 'inward' towards the green. Now it is impossible to generate the 'pure blue' obtainable before, because you stimulate the 'green cones' more.

Now observe that there are colours perceivable beyond blue (Indigo and violet - look at the top image in the 'cone sensitivity chart link' - the author has found it necessary to draw purples on the blue end. While this looks sort-of right, it doesn't match up to real violet. To see that you'll either have to look at a rainbow - or a UV lamp, which generally have some visible high frequency light. If you do that be sure to wear eye protection.

"I think you are arguing that if more than one cone is excited, we are unable to perceive pure colour"

No, I'm not saying this.

"As to the claim // the colours to the right of blue are not in fact indigo and violet, but rather purple (ie red+blue). // what is the difference between purple and indigo and violet? Is it that purple is for you made of a mix of colours, while indigo and violet are pure?"

The difference is that they look different. Like red and green look different. This is why a rainbow on TV looks different to a rainbow perceived directly. I'm sorry if you've never seen one, you've missed out. (You need a fairly strong one to see the full range red orange yellow green blue indigo violet - otherwise the colours in question are lost against the blue sky.) You could try and get hold of a triangular prism and make a spectrum using sunlight.

Look at the wavelength vs cone sensitivity chart on the link I gave. You can see that on the left side the green cones don't detect any light, while the red has another little peak then tails off before the blue. So a single wavelength of light at these positions eg violet is distinguishable from red+blue, because the latter also excite the green cones.

The three fundamental colours are red, green and violet, not red green and blue as sometimes claimed. It appears that 'primary colours' are just those you mix to give your palette, whereas 'fundamental colours' are those needed for full coverage of the human visual spectrum. (I'll add a link to dictionary.com).
-- Loris, Jul 31 2002


The gamut of colours produced by present day monitors are much smaller than the gamut that can be seen by the eye, so this idea may have some merit. The colourspace faq seems to provide good information about what can be represented by particular colour models. I gather that the RGB model is rather limited, and there are models that are much better representations (such as the LAB colour model [which is subtractive iirc, but anyway ...]). See links.
-- amadeus, Jul 31 2002


Very interesting. I'll have to do some staring at very colorful objects, play with various lenses, toy with photoshop, then get back to you guys.

So... What is the frequency of a "magenta" photon? If you shine monochromatic "magenta" photons in your eyes, will they look *exactly* the same as the mixed colors that your monitor uses? Will those receptors tire after five minutes, and make the color look different?

Off to do some serious staring... WATCH OUT!
-- Crazy Bastard, Jul 31 2002


loris: The primary colors are red, green, and blue. Each of these colors stimulates one of the three types of cone significantly while stimulating others only minimally. Light which is of a slightly shorter wavelength of blue light stimulates the "red" cones moreso than blue light, and is thus less suitable for use as a primary color.

While there are a number of reasons for using more than three colors of ink, they are generally not applicable to luminous displays; there's one exception (see below).

Some reasons for using multiple ink colors:

-1- The apparent color of a reflective object may vary substantially with illumination. An object printed with a three-color process may look identical to one printed with a many-color process when viewed under daylight, and yet look very different when viewed under fluorescent or incandescent light.

-2- Pigments which accurately produce the subtractive primary colors (cyan, magenta, and yellow) tend to be very expensive; most common pigments are a bit "off" [this is especially true of magenta]. Since a common magenta and cyan don't combine to produce an exact blue, it's cheaper to use an additional blue ink than to use perfect magenta and cyan inks.

-3- Using more colors of ink may reduce the total amount of ink needed on a page. The standard four-color printing process uses black ink for this purpose (cyan, magenta, and yellow combine to produce black, but at the expense of using a lot of ink; the amount of black ink required is less than the amount of each other color that would be required, and black ink is cheaper per unit volume to boot.

Note that most of these issues are totally non-applicable to luminous displays. The only thing that might be a factor would be the use of an additional "white" color on color LCD displays to improve brightness at the expense of color accuracy and saturation. This approach, btw, is baked in some color incandescent lamp-based signs. The signs use four types of light: red, green, blue, and white. A combination (red+white), (blue+white), or (green+white) will appear brighter but less saturated than any color by itself; indeed, the white lights by themselves are probably as bright as the other three colors combined. While color fidelity will be compromised by this approach, the brighter image may be worth it in applications not requiring photorealistic color.
-- supercat, Aug 01 2002


I'm starting to wonder whether people are defining magenta in different ways. To me it means the colour my BBC computer called magenta, which was red+blue. I also call this purple. To the 'little Oxford English Dictionary' I've just looked at it is "shade of crimson; aniline dye of this colour." To me crimson is bright red (and suprisingly the OED doesn't have an entry), so there appear to be at least 2 definitions. Americans may have a third.

Anyway, Crazy Bastard, assuming magenta=mid purple it is not possible to have a magenta photon. Because the eye detects photons between red and blue as green. (Theres probably a joke somewhere in this - the light detector cells detect light using a pigment called visual purple).

However, suppose one took red and green light and mixed it. This would look identical to an intermediate wavelength colour (ie yellow or orange depending on the proportions).

Supercat: I disagree with your first statement, and so do all the dictionaries I've looked at (which state anything). In fact I used to believe they (primary colours) were RGB until I was shown otherwise. While I don't want to fall into the trap of 'argument with authority', I think one can make a logical case for either:

RGB (as you have it) Blue stimulates red least

RGV Violet doesn't stimulate green.

This suggests that maybe both blue and violet are in fact primary colours (ie, both are necessary to get full spectrum perception). Albeit the effect quite subtle.

I'm quite happy with this because this was my thesis to begin with. :-)

I am not going to write much about coloured inks: they are not particularly relevent to additive light displays. Although I'd like to make the observation that multiple pigments are basically used because of chemistry. Pigments just don't do what we'd ideally want them to do.

I am interested in your comment on white light in fluorescent light displays. It occurs to me that a fringe benefit could be that having 4 different components would allow for the pixel unit to be square, with gridlike position of each colour element. This might give a slightly better perception of even colour.
-- Loris, Aug 01 2002


I've still not seen any evidence that a monitor with a violet component would be perceptibly different from one without. All the references I can find on colour synthesis claim that an RGB display can produce all possible colours in the CIE colour model, which is supposed to give a very good approximation of human colour perception.

From this it's not certain that a violet gun would have no effect -- it may have a small one -- but the added complexity of implementing and tuning the extra colour element would detract from any improvements on extreme frequencies. I'm actually more worried by my monitor's inability to represent blacks, which a violet component would not improve.

Another problem is that colour perception doesn't just depend on the wavelengths but on surrounding colours, on ambient light, on the texture of the object being perceived, stereoscopic effects, and other factors. These account in some measure for the fact that images on a monitor will not resemble real images. (In fact, since we are used to seeing photographic images with a glossy surface, I think adding a glossy surface to a monitor would do more to give an impression of realism than anything else.)
-- pottedstu, Aug 01 2002


Loris: looking at the sensitivity chart, you may have a point about the green sensitivity not falling off completely at the wavelength used for "blue". What you are suggesting, then, is that although there are only three kinds of color receptors it may take more than three wavelengths of light to obtain all possible combinations of stimulation because there's no such thing as "anti-light" [i.e. mathematically it would be possible to produce any combination of RGB stimulation using RGB light except that some combinations of stimulation would require negative amounts of green].

Interesting notion. I wonder what would be the effect of experiments to test whether subjects could distinguish a saturated violet hue from the closest-approximating blend of red and blue. Violet would only be needed as a 'primary color' for producing a saturated color; any shade of violet that was even the slightest bit paler than pure violet could be produced quite adequately with existing RGB technology.

I would still posit, though, that RGB is a much better set of colors than would be RGV. It may be that RGBV is better still, though I don't think the extra expense of a violet gun would be in most cases justified. Additionally, I'm not sure whether everyone would see the new hues the same way. A physics professor of mine was surprised that his students could only see seven lines in the mercury spectrum, since he saw nine. Turns out his cataract surgery extended his visible range into the very near ultraviolet.

BTW, as for the notion that using RGBW on an LCD display would allow 2x2 pixel mosaics, that was in fact very much part of the thinking. A number of the RGBW light boards I've seen use pixel-splitting to enhance resolution on certain images. Normal color LCD's allow 3:1 pixel splitting horizontally but none vertically; 2:1 in each direction might be better.
-- supercat, Aug 01 2002


Hmmm.... Well, I'm concerned on a more simplistic level. I'm simply not certain that any mix of separate, limited colors can accurately reproduce the perception of true color, unless they match the detail level of that found in nature, in effect, the way that molecules of different materials reflect light of different frequencies.

Then you still have "mental" color to deal with, this is all only pertinent to the eye's perception of color. The fact that the brain can "imagine" a color that we call magenta is amazing! It leads me to believe that things might get yet more complex in this discussion of what color really is.
-- Crazy Bastard, Aug 11 2002



random, halfbakery