38

I have several different RGB LEDs and bicolour red/green LEDs, from different suppliers.

I find that, with the appropriate resistors, they can produce a perfect orange and cyan; and a barely acceptable magenta. One thing they cannot produce, however, is a tolerable yellow. It looks awful.

Given that our eyes see yellow simply as a combination of red and green; and given that TVs and computer monitors successfully render a lovely yellow on a daily basis using nothing more than red and green pixels; why is it that my LEDs are utterly incapable of the feat? I have tried googling for answers, and come up empty.

Of course, each colour will have its own forward voltage, but I have allowed for this by providing different resistances; or, in some cases, adjustable PWM signals. But no matter how I vary the amount of red vs. the amount of green, at no point is a decent yellow produced.

Sadly, due to the limitations of cameras, I am unable to provide a realistic illustration of the pathetic yellow they produce. But suffice it to say that it's a sickly, unpleasant colour.

Now, maybe I just have low quality LEDs. But still, any shade of yellow should be possible from the correct ratio of red and green. Even if the green is a bit red-shifted already, surely I would simply need less of it?

winny
  • 13,064
  • 6
  • 46
  • 63
Sod Almighty
  • 1,295
  • 13
  • 22
  • 6
    This will be much less opinionated and much more answerable if you could post datasheet to RGB and RG LEDs you have tried and seen. Personally I have seen RG LEDs that generates a pleasant yellow so I couldn't answer it. – pipe May 10 '19 at 13:30
  • 1
    One interesting reflection I have made which could be related: I can't make a white LED gray! No matter how dim I make it, my brain interprets it as just a less bright white instead of gray. – pipe May 10 '19 at 13:36
  • 3
    "Given that our eyes see yellow simply as a combination of red and green;" -- have you never seen a color chart? – Scott Seidman May 10 '19 at 13:40
  • You either need better LEDs, you need to drive them better, or you need a half-way decent diffuser to mix the colors better. The fact is - architectural LED RBG fixtures exist, and they produce yellow. OLED displays exist, and they produce yellow... – brhans May 10 '19 at 14:21
  • 4
    @pipe Try brown! When I'm not an EE, I do concerts and shows. Do you have gingerbread-mans during Christmas where you live? Have to tried to light a set in gingerbread-brown? I had normal halogen light and the full LEE color filter catalogue at my disposal, but good brown, never. – winny May 10 '19 at 15:59
  • 1
    @winny Interesting, now that you mention it I can't imagine how a "brown" LED would look like, so I don't doubt you. – pipe May 10 '19 at 17:44
  • 3
    @pipe brown is nothing more than dark yellow-orange, but easier said than done. – hobbs May 10 '19 at 22:29
  • The problem is your assertion that LEDs cannot create a decent yellow. That is not true. It more about economics. – Misunderstood May 10 '19 at 22:31
  • 2
    @winny can an LED monitor produce brown without magic? – Misunderstood May 10 '19 at 22:40
  • @ScottSeidman Yes I have. And I stand by my assertion. The eyes can see only three colours. – Sod Almighty May 10 '19 at 23:19
  • 1
    @pipe Difficult to post datasheets for random LEDs bought from shady sellers in China. Otherwise I would have done so. – Sod Almighty May 10 '19 at 23:20
  • https://upload.wikimedia.org/wikipedia/commons/thumb/9/94/1416_Color_Sensitivity.jpg/300px-1416_Color_Sensitivity.jpg. this is not as clear as you believe. There is substantial sensitivity spread for each type of photoreceptor – Scott Seidman May 11 '19 at 01:08

2 Answers2

58

Most probably because of short wavelength of your green LED and not as monochromatic green as you might expect (x and y coordinates closer to the center). If you take a look at the CIE 1931 curve and plot your red and green x and y coordinates (listed in the datasheet from serious manufacturers, otherwise assume the pure wavelength on the outer rim or move it in slightly), the only colors you are able to produce are along that line. For RGB, it's the same but you are forming a triangle instead and you can only mix color within that triangle.

Yellow is tricky here because you need your green to be almost yellow to be able to mix green and red to a good yellow, at which point you have sacrificed a lot of green and cyan colors. To make things worse, the production spread in green LEDs is very high compared to other colors, so it will vary unless you calibrate each one, at least from batch to batch.

enter image description here

Pardon my paint skills here, but let me give you three examples. A "yellow-optimized" RGB LED will have high wavelength for the green LED but you are missing out a large area of light blue, cyan and blue-green:

Yellow-optimized

A cyan-optimized RGB LED has short wavelength on the green LED and will sacrifice the yellow colors to get good cyan.

Cyan-optimized

Commercial RGBA (A for amber, thanks for pointing it out Ilmari Karonen) LEDs exists for this reason, at which point you will have a trapezium (thanks Wildcard!) with four points you can draw all the colors within, at the added expense of one more LED and associated driver channel.

RGBA

winny
  • 13,064
  • 6
  • 46
  • 63
  • 8
    The color sensitivity of eyes also varies. The pigments (essentially filters) in the (usually) three cones are controlled by genes, and some people are far more sensitive to one primary color than the other two. Lighting spectra are based on average sensitivities, so some people see a particular color very differently. I see the red/green part of the spectrum slightly differently with each of my eyes, particularly around the yellow region. Whether that's due to a mechanical difference in my retina, or how I process the signals is unknown. https://en.wikipedia.org/wiki/Tetrachromacy#Humans – Phil G May 10 '19 at 16:31
  • 15
    For anyone else curious about it, in this context the A in RGBA apparently stands for "amber". (I was pretty sure it couldn't be ["alpha"](https://en.wikipedia.org/wiki/RGBA_color_space), but I couldn't figure out what else it might stand for, so I had to Google it. Now you know.) – Ilmari Karonen May 10 '19 at 17:14
  • 2
    @PhillG the CIE 1931 is an average of human vision among non-colorblind people. – winny May 10 '19 at 17:31
  • 8
    I learned this as a side effect from researching why the green in some RGB LEDs required such a high voltage compared to the "normal" green LEDs. Turns out there are such things as "cyan optimized green" to get an arguably better coverage, and they have a different chemistry and thus a higher voltage drop. – pipe May 10 '19 at 17:53
  • 9
    @pipe To some extent, green LED does not exist in nature. You either need to push red AlGaInP to become green or blue InGaN to appear green by doping. I would suspect this is the reason to the wide production spread. – winny May 10 '19 at 18:01
  • A spectrophotometer captures the spectra and you compute three integrals, x, y, and z. These are then normalized and combined into (x,y) coordinates on the CIE plane. But these coordinates need a reference point. For LEDs, it's the D65 point (located at [.312726866, .3290235126].) Draw a line starting at D65 and through the CIE coordinate pair (x,y), then out to the perimeter curve where you see the numbers. If you look very closely at the CIE curves you have shown, you can see this D65 point at the center of the whitish ribs right at the coordinate mentioned. – jonk May 10 '19 at 21:06
  • 2
    Yellow doesn't represent a wide span along the perimeter curve. So it takes some very carefully placed (x,y) points to achieve the perception of yellow. – jonk May 10 '19 at 21:09
  • Hey, wait. How does my LED monitor reproduce that fine looking yellow in your images? I especially like the #fffc19 yellow. – Misunderstood May 10 '19 at 22:42
  • @Misunderstood Mainly psychologically since it limits your perception. Try turning your screen all yellow and illuminate something with it and compare to a lemon under white light. – winny May 10 '19 at 22:45
  • @winny The assertion is that yellow cannot be adequately produced (perceived) with RGB LEDs. My monitor uses 3 monochrome LEDs and it can perfectly reproduce as perceived by my vision. A reflection of light off a surface is a completely different thing with additional photonic properties. If RGB LEDs could not reproduce yellow then I would not see yellow in an RGB image. – Misunderstood May 10 '19 at 22:57
  • 2
    @Misunderstood, again, a limited color gamut will distort your vision. If you change your color temperature on your screen, you will adapt to it. Same goes for limited yellow, unless way overboard, you will adapt to the “new yellow” within your limited scope. You need an external source to compare it with. I’m not aware if additive RGB and subtractive CMYK would have any second order effect on your perception here, but perhaps play around with an incandescent lamp with a yellow filter just next to your screen. – winny May 10 '19 at 23:00
  • @winny why would I care if my vision is distorted. As long as I perceive a fine looking yellow, it's a fine looking yellow. Sure I can make adjustments to my monitor and a eliminate its ability to reproduce what I perceive to be a fine yellow, but why would I want to do that? How do I even know that the yellow I see is the same yellow you see. Light is (photons are) invisible. It's how my eyeballs respond to the various energy levels of the photons and how my mind interprets those sensations with what I have learned. – Misunderstood May 10 '19 at 23:12
  • 1
    @winny Thank you for such a detailed answer. I tend to imagine the spectrum as a horizontal line of colours, and it's difficult to get my head around the two dimensional representation. But you're saying that the green chip may have some hidden blue in it, and as such it's impossible to add sufficient green to the overall colour without also adding enough blue to make it closer to white than I would like? Or, in other words, if imagined as a horizontal spectrum, the green chip covers far more area in the green region than desired, and thus pollutes the yellow when mixed? – Sod Almighty May 10 '19 at 23:31
  • Incidentally, a yellow pixel on my screen will look yellow even if the rest of the screen is black. So screens are, clearly, able to produce nice yellows; reinterpretation by the brain notwithstanding. As an average consumer, is there any way to know which LEDs are likely to produce nice yellows? LEDs don't usually come with datasheets. Maybe if you bought a stupid amount of them for a stupidly high price from a major distributor, but that's not going to happen. – Sod Almighty May 10 '19 at 23:35
  • 3
    @SodAlmighty -- any LED you get from a decent distributor (whether in stupidly high amounts or mere onesies) will have a datasheet (it may not *ship* with the part, but it will definitely be available online via the distributor's website!) – ThreePhaseEel May 11 '19 at 03:25
  • 1
    there are multi-color led-packages out there to fill in the gaps: some have up to 7 colors. for example the [LZ7](http://www.ledengin.com/products/emitters#LZ7) package with 7 different colors. there are also lights for the event/theater/show/stage usage that use more than 3 colors to get better coverage. – Stefan Krüger s-light May 11 '19 at 14:17
  • @StefanKrüger Perhaps, but £13.84 a unit is taking the piss. – Sod Almighty May 11 '19 at 17:51
  • 1
    TIL that Brits and Americans define "trapezoid" and "trapezium" differently. By the American definition, that last diagram is not a trapezoid, it's a trapezium. – Wildcard May 11 '19 at 17:57
  • @Misunderstood , monitors work totally differently, it's unrelated. (Explained by IceGlasses below.) – Fattie May 11 '19 at 19:37
  • @Wildcard. Didn’t know that. I’ll edit accordingly. Thanks! – winny May 11 '19 at 20:02
19

It's even a little worse than winny indicates:

Green LEDs are finicky. One result is that green LEDs emit over a broader distribution of wavelengths instead of being a nearly laser-like single wavelength. And when you map that range of greens to the xy colorspace you're not on the spectral locus anymore. So even your yellow-optimized RGB LED may not get as close to the yellow as you'd hope.

From What is red? On the chromaticity of orange-red InGaN/GaN based LEDs: Achievable InGaN colors

For those who keep noting their LED monitors can show a nice yellow, your LED monitor almost certainly doesn't use three colors of LEDs. It instead is going to use white LEDs and color filters, and the white LEDs get their yellow from a phosphor instead of the bandgap of the semiconductor.

IceGlasses
  • 341
  • 1
  • 5
  • My slightly in “otherwise assume the pure wavelength on the outer rim or move it in slightly” may have been an understatement. Nice answer! – winny May 11 '19 at 07:27
  • OLED displays do use direct emissivity, do they have a closer to monochromatic green than normal LEDs? – Dan Is Fiddling By Firelight May 11 '19 at 16:40
  • I'm afraid I don't know very much about OLEDs. In a quick web search I can only find one mention of the spectrum of an OLED, so don't take too much from this, but it looks like they may be similar. (And to be preemptive, I don't know much about quantum dots either, but those are another way to get a color different, and possibly more specific, than the underlying emitter.) – IceGlasses May 11 '19 at 23:38