Display that supports full CIE xyY color space

Right now, all current monitors use three (or four) primary colors to create their color, which could never allow them to produce all colors that can be seen by the human eye; in fact, their colors will always be inside a triangle on the xyY diagram of all possible colors.

But we could decide to replace these color filters with something such as Lyot filters, which can be tuned, similarly to LCDs, to allow only light whose wavelength is higher/lower than a certain value to pass through.

This would mean that we could basically create color filters that can be adjusted to produce any color, by replacing each pixel with two stacked filters, which allow a range $[a, b]$ to pass, and a third filter next to it which allows the range $[c,\infty)$ to pass through.

Using Mathematica, I have tested that a mapping between these three values is enough to create any possible $x$, $y$ and $Y$ in the CIE xyY color space:

So, in other words, the range of $$\text{XYZ}(a,b,c)=\int_{[a,b]\cup[c,\infty)}\text{I}(\lambda)\cdot(\overline{x}(\lambda),\overline{y}(\lambda),\overline{z}(\lambda))$$

, where $\text{I}$ is the illuminant, is enough to produce the color of any object that can be illuminated by that illuminant.

So, my questions are:

• Are there already any such displays that can produce any visible color, and if not, then why not?
• Could this actually be done in practice?
• Fascinating question, though I don't know that we have the expertise on this site to answer it. :) I'm not aware of any displays that work on principles other than additively combining a set of primaries, which as you mentioned only gives a convex polygonal subset of the CIE space. The liquid crystal tunable filter sounds like amazing tech, but I'd guess that it's not in consumer displays just due to some set of engineering tradeoffs vs standard LCDs (like cost, brightness, field of view, switching speed, and suchlike). Dec 27, 2016 at 0:35
• One issue I see with this, what content would you play through such a display if it existed? All currently existing digital media is encoded as RGB (or YCrCb, which gets converted to the same thing). So to take advantage of this new colour gamut, graphics programmers would have to completely change the way they write colour code, and CCD manufacturers would also have to create new sensors that capture the full range of colours.
– russ
Dec 27, 2016 at 9:54
• @russ There's not much content for HDR displays yet either, and they are slowly being adapted. I think that if such a display produced images that look significantly better than what regular RGB displays can do, it would have a chance to eventually see widespread use. That's another big if, though. Dec 27, 2016 at 14:34
• @Quinchilion yes but most devices are capable of capturing higher dynamic range but not color outside the tristimulus triangle. Besides HDR is just upwards in the chart so and can easily be reranged accetably. Sure no problem dont display color you dont measure. Dec 27, 2016 at 16:28

No, most likely these kinds of displays do not exist, at the time of writing this post. Although the negative is hard to prove. And in this case i don't have deep enough understanding of the lore.

Reason for me saying this is that usable displays need some serious miniaturization. The miniaturization process is quite expensive and requires some serious investment. There should be quite massive patent trail if this was indeed true. So we should look in patent databases for this trail.

Second, having such tech probably will not yield full color space, at least initially. Because the filter can only filter out colors that the back light can generate. Most probably therefore this is some kind of projector technology at first. Also filters probably have a practical limit to how much they can be tuned.

Third, if/when such tech is available it must recuperate the development cost by selling enough units to overcome the capital investment. There is a likelihood that this is one of those things that can not be produced without the miniaturization step like digital cameras. So keep a lookout for a killer application for this tech.

Some major downsides of this kind of tech is the fact that you filter out light. This means it eats up part of the energy input. This leads to a heat dissipation problem, which for perfect narrow band filters is actually pretty huge part of the emitted light. Lot of energy goes wasted, and needs to be engineered around.

• Thank you for your answer, but don't LCD color filters filter light out too? In fact, the larger the color range a display has, the more narrow the transmittance spectrum of the filter will be, meaning that there will be large losses even for white. With tunable color filters, though, there would be large losses only for saturated colors (which matches what objects do in reality; saturated objects absorb more of the illuminant), but pure white would be created by letting all wavelengths pass through (a < 385 nm, b, c > 745 nm), meaning less losses than LCDs. Dec 27, 2016 at 13:30
• Also, if we pick a backlight whose spectrum is the same as the LEDs that light up the room, we could produce all the colors that can possibly be found in objects of that room. I do agree that minaturization would be difficult (although it would probably be easier than other recent display technologies such as OLEDs since Lyot filters use liquid crystals to be tuned). Dec 27, 2016 at 13:41
• @d9584 yes but the LCD has the luxury of having only 3 bands to contend to so it can survive with a light that has less band than full spectrum and it can pass things in wider band. Dec 27, 2016 at 14:04

In addition to what the others have said, there are some practical issues with making such a display. It was mentioned in the comments that all current content is made for current display technologies, so having extra display capabilities wouldn't get you anything useful. Of course, if such displays were common, content would eventually come, right? Well, not necessarily. Where would it come from? Photos and video are produced either with film cameras or digital cameras. These have wider gamuts and higher dynamic range than current monitors, but I don't believe any of them cover the entire color space you're describing.

Even if you did have ways of capturing the whole range or generating it you'd run into the problem that you'd need more bits per channel, and you'd need data structures and applications to interpret that data appropriately. Photoshop, for example, has supported working in 32-bit per channel float for a while, but many of its functions are unavailable in that mode. Next would be the playback apps. Then all the services that stream stuff to our devices would need to be updated, too. That would mean new codecs, too.

The fact is, this is happening. Apple and other manufacturers are starting to use wider gamut displays on their consumer devices (the iMac and iPhone, for example). UHD content is being made for both televisions and consoles, and the display hardware is now available for them. As display gamuts get wider color gamuts (currently most wide displays are DCI-P3 or a variation thereof) and higher dynamic range the software will get written to display properly on them, then codecs will get written store and stream such data. I predict that after P3 displays are the norm, we'll make the next step to something wider like Rec. 2020, and then onto something even wider like ACES-CG. Will it ever cover the whole gamut, and the dynamic range of the eye? Probably someday. Someday soon? Probably not. But in 20-50 years, I could see it being the norm. In the meantime there's a lot of infrastructure that needs to be updated to support it, and it's going to be slow to get it all working properly.

The answer should be a theoretical no for any tri-light system, or even four light.

The reason is, if you examine the original CIE RGB primaries, you will see that the positions of the three CIE 1931 lights correspond to the tips of the triangle in the spectral locus image here:

If you look closely, you will see that along the edges of the green red axis, and the red blue, that there is some "escape" of the values where the primaries mixed cannot replicate the range of visible lights that the experimental observers were attempting to replicate.

What we learn from this is that any combination of tristimulus physical lamps can never encompass the entire spectral locus shape.

In order to fully enclose the unique shape of the spectral locus, it would require using primaries that consist of imaginary visual lamps that would form a triangle well beyond the spectral locus tips. Examples include the imaginary XYZ lamps, or the ACES lamps. Those lamps of course do not exist in the physical sense.

So with all of that said, is it plausible that such a display could exist in the future? Absolutely. If we switch to a 40 lamp display, or something close to that, we could replicate the various wavelengths that enclose the entire shape of the spectral locus and fully encapsulate the CIE standard observer response.

• ...but the system I'm describing is not a tri-light system. Instead of using a normal color filter, it uses a filter that can be adjusted using liquid crystals to allow only a certain wavelength range to pass. Dec 27, 2016 at 19:00
• That's sort of what we already have. The point is, given that the CIE model is based on a standard observer, it is impossible with any given tristimulus system and the visible light range. Dec 27, 2016 at 19:07
• No, it isn't. Currently the liquid crystals just change the opacity of the color filters without changing their actual colors. What I'm talking about is using three filters that change the transmittance range of the color filter (like in this image: rp-photonics.com/img/edge_filter.png). This would mean that the three points in the chromacity diagram are not fixed, and can be moved aroud to produce any color. Dec 27, 2016 at 19:15
• @d9584 I can't speak to the hypothetical idea of a wavelength based system, and the outline I tried to answer for was why no full spectral locus display currently exists as per your first question. I also suspect that any such system would need an absolutely absurd emission both in intensity and uniform spectral output to be feasible. Given the constraints on the latter, it would need a light-year jump in technology to be even remotely feasible. Dec 27, 2016 at 22:52