Direct Diode Green Lasers (Part 2, Chromaticity)

In my last post on Direct Green Lasers (DGL) I wrote about electrical power to lumens conversion.  In this post, I am going to talk about the color range/space and how it is affected by the green wavelength used.   The color chart above is a standard CIE chromaticity chart, often referred to as a “horseshoe plot” due to its shape,  that plots all the possible colors in terms of an “x” and “y” chromaticity coordinate.  The color wavelengths in nanometers (nm) are labeled on the outside of the horseshoe.

First of all, I am not going to give a deep scientific definition to color space, but rather try and give the reader some practical information to help in understanding the significance of the wavelength spec for green lasers.  I apologize in advance to the serious color scientists as I am probably going to butcher some terms.

Humans are not very good at absolute color measurement as the human visual system is adaptive and relativistic.  A wide range of wavelengths of light will look “green,” and this is particularly if you don’t see them side by side.

Laser light has very narrow bandwidth which puts it on the edge of the CIE horseshoe so to speak.  It turn out that diode LED’s  while as pure as lasers are still comparatively very saturated and would plot very near the edge of the horseshoe as well.   A big exception to this would be so called “white LEDs” (and some other colors) that are actually made with blue LEDs stimulating phosphors).

Usually 3 primary colors with wavelengths that are considered, red, green, and blue are mixed to form any of the other possible colors (some systems use more than 3 color primaries).   The RGB phosphors used in old CRTs were not pure color wavelengths and so these “primaries” where inside the horseshoe rather than on the edge like lasers. The TV standards for broadcasting grew up with these limitations set how what color could be represented.  If you plot the 3 primaries used for standard definition television you the SDTV triangle.   Also ploted is the newer HDTV standard which defined a slightly larger color space triangle.

For more on the color space concept I would suggest reading the on-line article by Matthew S. Brennesholtz on expanded color gamuts and the Wikipedia article on CIE 1931.

The the lasers, I have also plotted the color spaces (triangles) assuming a 640nm red and a 460nm blue and then a triangle for each of 510nm, 525nm, or 532nm for the green.    For a given set of RGB wavelengths only colors “inside the triangle” can be represented.  Also if you follow an edge of the triangle it shows what color can be reproduced in between two of the colors assuming the 3rd color is off when mixing the the other two primary colors.

One thing immediately obvious is that the laser primaries are way outside the color gamut/triangles for SDTV and HDTV.  While this means a wider range of color could be represented, it also poses a problem when using existing standard video and still image standards.  For example if you have bright green grass in a video, the video signal will call for nearly 100% green, but if you use a 532nm green laser at 100%, the grass will look like it is glowing green rather than green grass.   So if you want the grass to look right, you actually have to desaturate the green by adding red and blue to it to get to a point on or inside the HDTV/SDTV triangle if you want the image to look like it is indended.

If you have a “wide color gamut display” using lasers or LEDs then you need content that matches your gamut to take advantage of it.  If you use the commonly available video and photos formats which were coded/compressed for small color gamuts you can’t take advantage of the full color gamut if you want the images to look like they were intended (and not an over-saturated glowing look)

Consider particularly the triangle made by the 510nm green lasers and notice how it cuts off the bright yellows and yellow-greens of the SDTV and HDTV color spaces.  There is no way to mix 510nm “green” with 640nm “red” to give a good yellow.  You have to have at least a 520nm green to fully represent the yellow within the standards.    This a major reason why there is the push to have direct green laser wavelengths of at least 520nm or longer.

You may notice that any of the greens from 520nm to 545nm (much more than 545nm and it starts cutting off some of the blue-green areas) will give a larger color space than HDTV.    But if you go back and look at the photopic response curve from part one (copied below) you will see that as the wavelength goes from 510nm to 555nm, the lumens per Watt improves.  For example, if the wall plug efficiency (WPE) was the same you would get nearly double the lumens per Watt at 532nm that you would get at 510nm.  Since 532nm is the common wavelength targeted by frequency double green lasers I tend to “derate” the WPE of shorter wavelength green lasers by their difference in lumens per Watt.   So a 525nm greens efficiency would be multiplied by 542/603=90% to get its effective WPE compared to a 532nm green laser.

One more thing on the CIE chart at the top, you will the “black body curve” in the middle of the chart numbers on it like 6500 or 10,000; these are the so called “color temperatures” of a black body is heated to the given temperature (in Kelvin).    A 6500 “white” is a little on the red side (also known as “warmer”) where a 10,000 “white” is a little slightly blue (also known as colder) which to the human looks “whiter than white” (and why some detergents put “bluing agents” in them).    “D65” is a common standard “white” that is very close to 6500 but slightly off the black body curve.   In the industry it is known that most westerners tend to prefer warmer colors toward D65/6500, whereas people living in Asia seem to prefer the cooler colors such as D93/9300 or even 13,000 Kelvin where the “white” has a clearly blue tint to it (I haven’t seen a study as to why).

It turns out that the target color temperature and the wavelengths of the red, green and blue will set how much of each color in Watts you will want.   If for example the color temperature is set for 6500, it requires will need somewhat more red but if you want 10,000, it requires somewhat more blue and green.

Karl Guttag
Karl Guttag
Articles: 243


  1. Mike

    This is a great topic and your analysis is right.
    I am an imaging specialist and agree with you that 520nm is shortest limit to reproduce good colors in my experience, too.

    I have a big question. How about 515nm?
    As far as I know, current available samples of DGL are around 515nm.
    Can you commrcialize pico-projectors with some compromise by using 515nm, or use 532nm SGL(SHG LD) to have good quality?


    • 515nm is about the minimum you could use from a decent, if not exceptional, color spaces perspective. Note that 515nm gives only about 415 lumens per Watt of light, versus 532nm that has 603 lumens/Watt. This means that 515nm has only about 70% of the lumens per Watt of light as 532nm. So when you look at the wall plug efficiency, you have to derate it when comparing to a 532-laser by about 0.7X. Until you have the 532nm “equivalent” WPE of about 10% (515nm at 13%), they will not be efficient enough for embedding in my opinion.

      The obvious problems with using both is that is that it would be expensive as you have the cost of both and the complexity of getting a 4th laser aimed in the case of beam steering. You have something that is already too expensive and you would be making it more expensive.

      As I have pointed out multiple times, particularly in the first article in this series, there is a huge difference between “samples” and production ready products. Usually when they make an engineering sample with the right wavelength at least at first is has issues with efficiency and/or lifetime and/or yield=cost, and or temperature stability and/or a number of other things. This is the nature of the basic material science they are working on. Eventually they will solve everything, its just going to take longer than some people want.

      If I take 7 different laser samples from multiple companies, I might be able to get all the spec’s I want, the problem is that I can’t get it in any single laser. Even if I take the best of the spec’s from 7 different lasers, it the efficiency would be at best marginal and the cost would be too high today.

    • Your comment makes no sense technically. You can have 10-bits or more of color space with or without lasers. The bits of color space are determined by the number of levels between off and full-on. BTW, Microvsion has fewer bits per color than most of the others as they can’t control the lasers that well.

      • the info came from Mitsubishi Laser TV.
        i didn’t know MVIS lasers could not handle the on off speeds needed for 10 bit.

      • It sounds like you picked up some “marketing physics” which is devoid of any technical reality. There are several concepts which are subject to misuse/abuse.

        In my December 26th blog I wrote about “chromaticity” or the range of colors. This is where lasers and LED have a big advantage over phosphors. Contrary to what some marketing people put out promoting lasers, LEDs also support very wide (wider than the commonly used standards support) color gamuts (the triangles in the CIE chart).

        What is being confused is the “dynamic range,” “color gamut,” and “color/bit depth.” Dynamic range is the measuring of light (of a given color) from its darkest to its lightest/brightest point. Color gamut is the range of colors and is often expresses as an area function of the CIE chart; but note you could have a triangle that covers a large area but still cuts off some desirable colors (see the 510nm green triangle in my blog). Color/bit depth is expressed in binary power of 2 terms (such as 8-bits). If you have 8-bits per pixel then there are 2 to the 8th or 256 possible levels over the dynamic range (darkest to brightest level) and 10-bits means there are 1024 different levels.

        But within bit depth you have to understand if the different levels are “linear/binary weighted” or follow some function known as a gamma curve (for a definition of gamma click here). The human visual system is non-linear so having linearly, evenly space steps, between each level is not as useful as having them weighted to follow a visual function. Your eye is much more sensitive to small changes at low intensity levels so you want to start off with small steps in the dark region and have progressively larger steps as you go brighter. A common gamma curve is about 2.2 which came from the natural functions of CRTs using phosphors and it got built into many standards. So you need to know if the 8-bits is “gamma corrected” or “linear/binary weighted.” To a first approximation, 8-bits linear is about as useful as 6-bits linear/binary weighted. Manufactures often don’t tell you if they mean “linear” or “gamma corrected” particularly ones that are “challenged” in providing a lot of levels.

        My understanding is that Microvision uses analog control of the level so it is not so much a question of speed but of the ability to change the drive levels of the laser. Note controlling the intensity for beam scanning is particularly difficult (one of the other things Microvision doesn’t like to talk about but you can find in some of their patents). The laser beam is not swept by the mirror at a constant velocity (goes back to the physics of the way the mirror moves) but rather speeds up and slows down as it sweeps. When it slows down if there was not compensation, the pixels would get thinner and brighter and when it speeds up they get darker and wider. Additionally the laser output can vary base on how much it has been used on the prior pixels (what Microvision in their patents called “pattern-dependant” variation). Part of the processing and power consumption problem with controlling the “bit depth” for LBS is compensating for the varying speed and light output for a given drive of the laser beam.

    • If you take say an OSRAM Ostar set of LEDs, they have nominal wavelengths of about 617 for red, 525 for green, and 465 for blue which are pretty close to an ideal triangle. LEDs typically have about 30 to 50 nanometers of spectral width so that pulls them in only slightly from the edge of the horseshoe so to speak, but still well outside any of the current color standards. A classic problem we found is that if you have bright green grass and the image calls for nearly 100% green, if you give 100% with LEDs, the grass will almost look like it is glowing it is so much more saturated than what the standard was expecting.

      Using so called white LEDs is a whole other matter. They often have a strange color spectrum. You also cannot talk about color space without talking about the color filters that are used to turn the “while light” into the primary colors. If they use narrow band filters, you get rich saturated colors but block/waste a lot of light. If you use wider filters you get more light but the color space becomes smaller. An additional big problem is that the color filters can only block like, they don’t covert it to a different color. So for example if the only “red” spectrum the white LED puts out does not make the red wavelength. You may note that the flat panel TV makers make two types of “LED” illuminated TVs. Most of them have “while LEDs” and rely on color filters, these TVs usually don’t have a wider color gamut or more saturated colors. Some of the better/more expensive “LED” TVs use separate red, green, and blue LEDs and these will have much more saturated colors.

  2. Hello Karl and thanks for your blog on these diodes. I’ve recently tested the Corn ing G-10000 SHG laser and iy makes wonderful eep mode free holograms (less than 2m linewidth),can we expect the same from thes green direct diodes?

    • Frequency doubled lasers ala Corning (I don’t know the specific specs for this laser) usually have a very short coherence length, on the order of 0.1nm which is a big problem projectors as coherence causes speckle which is very hard to reduce so we want several nanometers of spectral width. The current crop of direct diode green lasers have about 2nm spectral width which is a big help with speckle.

      The requirements of lasers for projectors is probably adverse to the interest of other applications. We want a “sloppy laser” with as little coherence as possible. One way to break up coherence is to increase the bandwidth.

  3. Hello Karl and thanks for your blog on these diodes.I’ve recently tested the Corn ing G-1000 SHG laser and it makes wonderful deep mode free holograms(less than 2nm linewidth),can we expect the same from thes green direct diodes?

  4. I find the absence of greens from the RGB gamut painful.

    I was encouraged to see how much more green − particularly emerald greens, and blue-greens − can be covered by this green laser.

    Instead of mixing three primary colors RGB, it seems possible to use opponent color theory (see CIE LAB, Natural Color System, and so on). This way the primary colors are on x and y axes: Red versus Green (R/G) plus Blue versus Yellow (B/Y), forming a quadrilateral rather than a triangle, thus covering much more of the human vision color range. Here the truer green from the laser opens up the other half the human colors.

    I wouldnt worry too much about looking at low-tech images ‘as originally intended’. These days, computers are fast enough to convert the old technology to the new technology whenever desirable.

    Thanks for bringing this green laser to my attention.

    • I think it is still a problem getting direct green lasers much longer than 525nm. Generally, the longer the green wavelength, the less efficient and less stable/reliable the laser becomes. Also, the perception of lumens/nits goes down as the wavelength is shorter (as the second chart in the article points out).

Leave a Reply

%d bloggers like this: