Know Thy LED

The invention of the LED is one of the most important discoveries of our times. They are everywhere, from our flashlights to household lighting and television sets. We don’t need to tell you that a project with more blinkies is better than a project with fewer blinkies. But an LED is not simply an LED; the sheer variety of LEDs is amazing, and so in this write-up, we’ll take a closer look at how to choose the right LED for your next masterpiece.

The LED Family Tree

The first official LED was created in 1927 by Russian inventor Oleg Losev, however, the discovery of electroluminescence was made two decades prior. British experimenter H. J. Round of Marconi Labs was the first to report the phenomenon in 1907. He found that silicon carbide would glow with a yellowish light when a potential of ten volts was applied to it. This set off years of experimenting with materials such as silicon carbide, gallium arsenide, gallium antimonide, indium phosphide, and silicon-germanium in an attempt to create a practical device.

In 1955, Rubin Braunstein reported infrared emission from gallium arsenide, however James R. Biard and Gary Pittman of Texas Instruments presented the first IR lamp (PDF) in 1961 which was the first practical LED to be patented in the August of the same year. Consequently, the first commercial LED was an IR LED with 890 nm light output and was called the SNX-100.

The era of the visible LED began in 1962 by Nick Holonyak, Jr. who was working at General Electric at the time. He discovered the red LED and published the results in the Applied Physics Letters on December 1, 1962 and currently holds around 41 patents to his name. He is known as the father of the visible LED and is also responsible for the laser diode commonly used in CD and DVD players. A decade later came the discovery of the yellow LED, M. George Craford, who happens to be a former graduate student of Holonyak.


The LED that Won the Nobel Prize

In 2014, three scientists, Isamu Akasaki, Hiroshi Amano, and Shuji Nakamura won the Nobel prize for inventing the blue LED in the early 1990s. Although RGB LEDs are obviously not possible without the “B”, the invention of the blue LED was important beyond the color. Blue LEDs are bright and efficient, and were the last stepping stone towards producing the white LED that illuminates the world today.

There are two methods to create white light from LEDs. The obvious method involves mixing three primary colors in the correct proportions to produce white illumination. The second method which is used to make white LEDs is the phosphor method where the blue LED shines onto a yellow phosphor coating.

In this method, the blue LED is used in conjunction with a yellow phosphor coating. The idea is to have part of the blue light converted into yellow light and leave a part of it in its original wavelength. When both these lights combine, they form a white beam which is far more efficient and pure than that from the first method.

Believe it or not, this discovery of color combination was made by Sir Isaac Newton in the early 1700s.

Behind their Glowing Personality

Regardless of the color, LEDs are all electroluminescent. Electroluminescence is the phenomenon wherein a material emanates light when an electrical current is passed through it. The underlying process involves the recombination of electrons and holes in the material. Check out this video for a quick summary and visualization.

An LED is a diode, or a PN junction. The types of materials used in the junction determines the color and intensity of light emitted. Voltage applied across the junction provides the energy to break electrons free from their parent atoms. The free valence electrons later recombine and release that energy as a photon. A typical LED construction is shown below. The semiconductor die is where the recombination happens and the photons are emitted. In order to channel this light, a reflective conical cavity is made and the epoxy lens on top allows for further collimation or diffusion of the light.

If the physics of the LED interests you then I suggest starting with The First Practical LED (PDF) as a reading resource.


Different materials used in the die preparation are as follows.

  • Aluminum gallium arsenide (AlGaAs) and gallium arsenide LEDs emit red and infrared
  • Gallium nitride LEDs emits bright blue
  • Indium gallium nitride (InGaN) emits blue, green and ultraviolet high-brightness light
  • Aluminum gallium nitride (AlGaN) LEDs emits ultraviolet
  • Yttrium aluminum garnet LEDs emits white
  • Gallium phosphide (GaP) LEDs emits reddish, yellow and green
  • Aluminum gallium indium phosphide (AlGaInP) yields yellow, orange and red high-brightness LEDs
  • Aluminum gallium phosphide(AlGaP) LEDs emits green


An LED for Every Reason

Beyond different colors, different LED materials lend themselves to different applications. Lower-intensity LEDs are typically employed as equipment indicators like in the case of router blinking lights. There could be as many as 100 LEDs on typical rack mount equipment and these should draw as little power as possible, but don’t need to be blindingly bright. Seven segment displays can have a luminous intensity (LI) as low as 260 ucd at 15 mA.

There are brighter LEDs that are designed for fog lamps and traffic lights and have LI of 34 cd at 350 mA (2.15V). It does not stop there either. LED grow lights are specifically targeted towards horticulture and indoor farming. There, a mix of blue and red light is usually used for growing plants in artificial lighting, although some companies claim that targeted lighting with 730 nm, 660 nm, and 450 nm provides the best balance of growth and efficiency.

And just when you thought things could not get any more complicated, we have the case of white LEDs. The color of light produced is measured on the Kelvin scale, where a lower number equates to a warmer light—the higher the number, the whiter, (and, yes, eventually bluer) the light will be.

Say What Watt?

LEDs intended for lighting, especially those sold on eBay, are often specified by how many watts they consume. For instance, this LED ad says 12 V and 20 W which by Ohm’s Law means a current consumption of 1.66 A. The wattage is the amount of power which the lamp ‘should’ consume at a particular supplied voltage, but isn’t a good measure of light output — for that you should be concerned with LI.

But wattage does matter. Assuming that an LED’s efficiency is roughly 50%, running that LED at 20 W means that around 10 watts will be dissipated as heat. Because the efficiency of an LED gets worse as it heats up, this kind of LED absolutely requires a heat sink if run at high currents. We’ll discuss this topic more in a future article on driving LEDs.

The Future Of LED Research

There is a lot of ongoing research in the field of LED manufacturing as well as basic material science that underlies it. In the case of manufacturing, work is focused on creating LEDs that are smaller so that they can be used in higher resolution displays. Patents such as this one for micro-reflectors on a substrate for high-density LED array are being filed every day. With wearables becoming more popular, a recent patent on flexible LED substrate devices is proof that we are on the path to flexible electronics.

There is also scope for higher efficiency LED designs as well as LED with better thermal management. UV LED design is still evolving and there is room to improve. published results suggest UV LED of up to 75 Watts are on the horizon.

The future is really bright and hopefully a lot more efficient.

80 thoughts on “Know Thy LED

  1. I bought a PILE of bog standard 5mm red LEDs from Digi-Key a while back, and was shocked to discover that the long lead on them was the *cathode*. Fortunately, they had the usual flat spot properly on the cathode side. I send these out in kits and have to remember to trim the cathode to meet everyone’s expectations.

    1. I can confirm (those were red 3mm LEDs from unknown source), Except of that, I also found diodes that had the internal anvil part on anode, not cathode as usual. It was red 5mm LED too, I checked a dozen of them.

  2. A very nice article. I think it would have been really informative if you could have gone into the E-k diagrams a bit to illustrate why some semiconductors have direct bandgaps that create photons and others have indirect bandgaps that create phonons. I always thought that was a pretty cool way to visualize what is going on…except understanding where the E-k diagrams come from is pretty involved.

  3. I have always referred to the larger metal part in the LED as the anvil. I am sure I called it that because of its shape and relative size. I was unaware that it was really called that.

    1. It really is surprising how many times in electronics we call something what it looks like and that’s actually what it’s called. Even more surprising is when you realize radio buttons on computer forms are called such because they behave like the exclusive buttons in old radios. Took me 15 years of writing code that used radio buttons to even question it and find it was so simple. Ever since been looking up a lot of those kind of things and finding it’s pretty much because that’s what it looks like or behaves like.

        1. Because it looks like handheld film cameras from the era before camcorders. If you run a Google Image Search on “Bell & Howell movie camera”, you’ll see all but the bellows in the first line of results.

          It’s a reference to a speed camera for another reason. Press photographers in the era before SLRs used the Speed-Graphic brand the most. Those definitely had bellows.

          1. That doesn’t answer the question.

            Using an unrecognizably abstract symbol of a bellows camera in a road sign is like using a picture of two flags to mean “traffic lights ahead”. You know, because there used to be a man with flags like a hundred years ago.

    1. True. The maximum luminous efficacy for a monochromatic green LED is about 600 lm/W whereas the maximum for a continuous spectrum white light is about 250 lm/W at 100% conversion of electricity to light.

      Practical broad spectrum diodes are about 25% efficient all things considered, so the maximum you can expect out of a quality white LED is about 60 lm/W and 75 lm/W if you allow for less than ideal spectrum. They’re about the same as compact fluorescent tubes in terms of light quality vs. power consumption.

      When you see advertisements for LEDs boasting things like “replaces 60W incandecent” and the lamp is rated at 6-7 Watts, you know they’re bullshitting because a standard 60 Watt bulb puts out around 840 lm and the diode would therefore have to do over 120 lm/W.

      The claim depends on the idea that the standard lightbulb “wastes” light by radiating in all directions whereas the LED is directional. However, a room looks very gloomy if you don’t light up the ceiling as well.

      1. Before you decide that both pronunciations are correct, take a look at the historical aspect from which the Light Emitting Diode was abbreviated as L.E.D. which was pronounced like the individual letters L. E. D
        Even today it is still written in All Caps. LED. and not led.

        1. Dictionaries already show both pronunciations. Language evolves and led is easier to pronounce quickly and often than el-ee-dee.The single syllable word led has become an acceptable alternative, whether we like or agree with it or not.

        1. But we aren’t saying “a light”; we’re saying “an El”. There are two hard vowels interfacing, so we introduce the ‘n’.

          Try saying “a L” (uh-ehl) aloud. It sounds clumsy. Now say “an L” (aen-ehl) aloud. It flows. This is called a liaison.

          1. Unless you say: a ‘lead’.

            And incidentally, I understand that if you use an acronym as a word you have to only upper case the first letter, which is something that annoys me too I have to add, when the BBC for instance does it with ‘NATO’ it makes me grind my teeth.
            But anyway then it would be An Led, Grrrr

    1. I’ve alway heard it pronounced “L” “E” “D”, that is until LCD TVs came out and marketing started calling the led TVs.

      Interestingly though, I pronounce OLED, O-led.

  4. “In this method, the blue LED is used in conjunction with a yellow phosphor coating. The idea is to have part of the blue light converted into yellow light and leave a part of it in its original wavelength. When both these lights combine, they form a white beam which is far more efficient and pure than that from the first method.”

    I’ve noticed the color is still “off” in these times. Two things that LEDs enable is low voltage lighting aka 4PPoE and light communication (VLC) in one device.

    1. Ultraviolet LED’s are also a thing, and my understanding is that a lot of high power white LED’s are actually ultraviolet emitters converting all their output to visible being phosphors, with the phosphor mix determining things like the color temperature.

    2. The article is in error. blue-yellow LEDs have worse color rendering index than RGB-LEDs. While it’s more efficient to produce just blue and yellow, it’s far from pure white.

      The efficiency and efficacy of LEDs are two different things. The light output in lumens is relative to the wavelenght of light, such that the most “efficient” LED is monochromatic greenish yellow – it produces the most lumens per watt. This has nothing to do with the diode itself – lumens are normalized according to human vision.

      Red diodes get the least lumens per watt, which is why the blue-yellow diodes omit red colors entirely and produce an unacceptably poor spectrum for general lighting. Pure white is supposed to have all the colors, and all the in-between hues as well in a continuous spectrum.

      Using a phosphor more like that of fluorescent tubes generates a spectrum that spreads a little bit into the green and orange colors, and by adjusting the relative ratios it produces the typical 2500K “warm” light with Ra=87 which is the bog standard for energy saving lights: it still makes colors look like shit but most people don’t mind.

      That however reduces the efficacy of the bulb significantly: the lm/W figure is no longer any better than a regular CFL – meaning you’re paying more for the LED and it makes less visible light – and it’s much worse than standard T5 or T8 fluorescent tubes.

      1. Thank you! Was about to ask about it.
        Where can i find real blue-yellow LEDs so i can see the difference? I should look for very low Ks?
        And are there LED based lights that produce very good white?

        1. Buy any regular LED flashlight. They’re predominantly blue-yellow diodes. The color temperature is typically very high 5500 – 7500K as anything lower would just look yellow with blue fringes.

          There are very high Ra LED based lights, such as the Philips L-Prize and its followers, but they’re hybrid constructs with a blue and/or ultraviolet diodes coupled with red diodes inside a plastic dome that is coated with a phosphor that produces the rest of the spectrum by conversion out of the blue/UV light. These achieve Ra=99 which is almost perfect and you couldn’t tell the difference from a real incandecent bulb without special meters.

          The problem with those however is, that the Philips L-Prize cost $60 when it was introduced in 2012, and subsequently suffered poor demand, and as far as I can tell they pulled it off the market.

          1. Many of the glass bulb – LED filament bulbs you get at the discount store here (“Hofer”) have Ra /CRI values up to 97 and cost about 3 to € for a 7W (60W replacement) bulb. Of course nobody wants to pay 60$ for a bulb. But we have 2019 now :-) Also the phosphors have become better than 2012

        2. Board censorship ate my comment.

          In any case, there are 2700K warm white bare diodes from luxeon, cree etc. that achieve a CRI of 90 at 140 lumens per Watt, but that’s measured at the industry standard 25 C junction temperature, and the luminous efficacy drops instantly as the diode warms up, and it also drops with any sort of diffuser or optics in front of the light.

      2. No. phosphor based white LEDs have a CRI between 80 and 97 which is really good. I have RGB LEDs and most yellows look very bad, up to unrecognizable. There is just no yellow in the RGB spectrum, about 50nm of spectral width are more or less missing.

    1. The anvil holds the actual chip, so it has to sit underneath it, so that’s why its profile broadens itself to accommodate. The other simply takes up as much of the remaining space as it can so that it can be held immobile in the plastic matrix . There’s a tiny wire that’s bonded between an electrode on the top of the chip and the top of the post (the piece that’s not the anvil).

      1. The anvil also provides heatsinking to the substrate, which is important even at fairly low power, and in many packages is also the reflector, as many designs emit at the edge of the die, not through the surface.

  5. Surprised for led lighting there wasn’t mention of the “lumen per watt” to clarify efficiency… although for the life of the buib led should not need to be replaced, there are more efficient light producers that use less electricity for the amount of lumens.

    1. Possibly because lumens doesn’t measure total energy output, only luminous density within a prescribed arc of a sphere. You can have a device that outputs more total energy in the form of light per watt electricity isotropically than a denser patterned beam, or both could be the same as far as energy output per unit input.. so the lumens/watt doesn’t really have meaning outside of specific applications of the device(as opposed to general properties)… many of which you would have secondary lensing to focus the beam anyway.

  6. Another theory about why LED materials emit light, is that light is actually made of material particles, just as electric current is, and when particles in electric current collide with material in the LED, light particles are knocked loose. Kind of another interesting, if not popular theory- but oh well, there are many theories out there.

      1. Interesting the relation of LED light emission to maser/laser emission too- does a maser represent what has for centuries been called “luminescence”? In addition- luminescence seems to not be in accord with the so-called black body laws developed by Plank and others where the frequency of light emitted from a body is only related directly to the temperature of an object- which is clearly not the case in fluorescence. oh well- my ‘tree cents :)

      1. As I understand the word photon, which I think was coined by Arthur Compton, a photon is a quantum, and a quantum depends on a frequency (E=hv), so in that sense, no, I am saying that there is a theory that that light is made of material particles – as Newton and Descartes before him hypothesized. It seems like a valid theory- and is only a theory. But perhaps people may eventually transition to a quantum theory that is based on material particles- or I guess in theory the interpretation of quantum theory as being based on material particles could have always been the intent- but was certainly never vocalized or recorded publicly if true.

        1. First define “material particle”.

          Material implies mass, which could not be the case as then it wouldn’t be able to move at the speed of light. Conversely, if you remove the speed limitation, all physics breaks down and no longer describes the observable universe because then it should be commonplace for effects to appear before their causes.

          1. There is definitely evidence that the claim that light always travels at a constant velocity is not accurate- the Pound-Rebka experiment is the main one, but Arthur Compton showed that the frequency of light is made larger on reflection. But beyond that, in my mind, galaxies, stars, planets, atoms are all particulate and material in nature, it seems logical to conclude the same property for light, as Lucretius did, Robert Grossteste, Antoine Lavoisier- many great thinkers of science history believed that light is basically an atom. Another great theory is that all matter is made of light- which also seems logical to me. It’s all in the realm of free thought and we must be free to entertain compelling, although, non-popular, theories- otherwise we might languish as we did for centuries under the Ptolemaic cosmology.

          2. Light always travels at the speed of light. It’s just that the speed of light changes depending on the medium.

            The speed of light in vacuum cannot be overcome by anything, and not reached by anything other than light itself, because doing so would cause reverse causality and break all observed physics.

          3. Good point about light moving more slowly depending on the medium it is moving through- I think that is clear evidence that material light particles are colliding with other particles in the medium- and again, another solid piece of evidence that the speed of light is not constant- it certainly changes in a medium. So what about in empty space? That’s a good point- I think that the concept of conservation of motion that goes back to Hipparchos and John the Grammarian- and was echo’d by Rene Descartes and Newton is the case for light particles too- so, for example, the speed of a light particle in empty space, depends on the velocity the particle had when it entered empty space- which can be very variable. The theory that the size of space and/or speed of time depends on the speed of light, I think is doubtful. A clock in water might move more slowly in water, but that does not mean that time itself slows down. Herbert Dingle argued against the famous “twin paradox” by saying that it is impossible for the two twins to be moving at a velocity other than the same velocity relative to each other. i do think that there probably is some maximum speed limit that any particle can obtain at any scale- but I think that is the result of particle collision. Great to read that others have thoughts about this.

          4. >”A clock in water might move more slowly in water, but that does not mean that time itself slows down. ”

            You don’t understand that time doesn’t exist. Time is the relative order of events and a dimension in relativistic physics – it doesn’t have a speed, there is no universal time. The speed of light in vacuum is the maximum speed at which information, which means causal events, can propagate in the local reference frame, and anything faster is violating causality.

            Photons themselves may move faster or slower depending on local space, gravity, and medium, but that has nothing to do with light being material particles or material particles being light.

          5. I think we are going to have to respectfully disagree, because I think that time does exist if anything exists. In my view, time is the same throughout the universe- in this theory 5 o’ the clock here is 5 o’ the clock in M13 and M31 too. In terms of the twin paradox being “experimentally confirmed”, that claim is not as curious to me knowing that there is more than one explanation to any experiment, and all the evidence suggests that we are living in a very corrupt and dishonest age.

          6. >”Herbert Dingle argued against the famous “twin paradox” by saying that it is impossible for the two twins to be moving at a velocity other than the same velocity relative to each other.”

            It’s curious then that the twin paradox has been experimentally confirmed.

  7. I would suggest reading on how Shuji Nakamura had to fight an uphill battle to even have the ability to conduct research on GaN LEDs, without that kind of determination, who knows how long it would take for the poweful blue LED to be discovered…

  8. I recall way back in the late 70s we would put a red LED and a green LED in series on the top of a 9v transistor radio battery with a small switch to make a yellowish low intensity flashlight. You can imagine how exciting it was when we found out about blue LEDs and even more excited when you could get them at the local Radio Shack. RGB!

    1. Orange and yellow LEDs were around fairly early, 1974 or 75. And green were there too. But I also remember that a lot if “surplus” LEDs at the time were pretty bad. Weak output, but also not so great casing. I suspect some of the low output was because the casing wasn’t the right color so it acted as a filter. Some if it was awful, but it was cheap, and it was right before easy access to distributors came long, a change that meant if you ordered a specific LED you could rely on it.

      It was is when I got into the habit of soldering a 1K resistor to a 9V battery, so I could always be sure I had the right polarity, and an LED with decent light, before soldering it into a circuit.

      I’m pretty sure I don’t have any of those scrap LEDs around, memory says at one point I realized they would never see real use, so I scrapped them.


  9. I remember watching a documentary some time back that LEDs were problematic in exactly a single context: accident investigation.

    If you deliver a sharp blow to a fixture holding an incandescent bulb (like simulating a vehicular crash), you can examine traces of the filament after the fact and determine whether the bulb was illuminated or not at the time of the impact.

    So far as I am aware, nothing of the sort can be learned from solid-state lighting.

  10. I worked for Veeco in the MOCVD (Metal Organic Chemical Vapor Deposition) division, final test department. MaxBright and Epik700 are machines that deposit several atoms thick layers thick of Gallium, Indium, Aluminum, etc on to wafers in a reactor to make LEDS. I still find it amazing how thees large power hungry multi-million dollar tools, make such small, but important devices, that we take for granted today. In thees tools the metals are bound to an organic molecule in solution which is vaporized at low temperatures using a bubbler, mixed with other metal compounds and silane in precise proportions and fed into a reactor at 1100 degrees C along with H2 and or N2(depends on the recipe). Depending on the tool there can be a dozen metal blends available.
    In the reactor the organic portion is broken free of the metal and is exahusted, while the metal is deposited on the silicon.
    Thats the simplified jist of it anyway..

    1. It follows from Ohm’s law.

      U = Joules per Coulomb J/C = Watt-seconds per Coulomb Ws/C
      I = Coulombs per second C/s
      U = RI
      Ws/C = RC/s
      W = RC^2/s^2
      W = R (C/s)^2 = RI^2
      R = W/(C/s)^2
      substituting R back to Ohm’s law:
      U = W/(C/s)
      which is equivalent to
      U = P/I

      Ohm’s and Joule’s laws are essentially the same. They’re just different ways of writing the same principle.

  11. Efficient and bright blue LEDs are a byproduct of a failure to make a solid state blue LASER. The Japanese guys credited with the blue LED were working on a side firing blue LASER but were stumped because the silicon kept cracking.

    Then one of them noticed that the cracks *glowed blue* and had an “ah ha” insight. If the chip was made to crack a lot, it would make a bright blue LED that could be combined with red and green dies to make RGB/white LEDs.

    1. Except with the small unimportant detail, that blue LEDs are not made from silicon – and were never made from it. Early ones were made from SiC but inefficient, now they are GaN.

  12. The most exotic (and dangerous!) LED I’ve used was a bunch of 275nm LEDs. I don’t know what exotic semiconductor material they used. They are like the least efficient LEDs. With an output in the 275nm region of about 1-2mW, and an input power of about 100mW, they have an efficiency of 1-2%. And the cost around 20-25USD a piece. But 1-2mW of 275nm is still enough to give you cancer in less than a minute of exposure (that is the same for both eyes and skin). Don’t try this at home kids!

    1. It probably won’t give you cancer, but a serious sunburn or cataracts if pointed at the eyes.

      The occupational hazard limit for UVC is 0.2 μW/cm² so a 2 mW diode becomes “safe” to look at some 3-4 ft away. You don’t want to hold it up to your eye at arm lenght.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.