A brief tour of LEDs

One might say that figuring out how to make a diode junction emit visible light was a bright idea. A real lightbulb moment… LEDs are magical things!

In 1968, they started being available for non-absurd amounts of money. Between then and 1989, they were just little handy indicator things… little red or amber or green indicators that could replace tiny incandescent or neon bulbs.

In 1989, blue LEDs hit the market. The earliest blue LEDs were expensive dim little things that had a forward voltage of 4.9 volts, but a few years later, more varieties started to appear.

Now, LEDs were always amazing. The big appeal was that they’d last nearly forever and they were mostly immune to mechanical damage. But, in the early days, they were only available in a few colors, mostly red and yellow and green… and they weren’t very bright. So there were hard limits as to what they could accomplish.

In the mid 90s, blue LEDs were not only available but they started to get accessible and bright and efficient.

In 1999, this random Internet personality cooked up the LED Museum and, for a period of time, it became full of fascinating new LED products before LEDs started to be all forms of light generation from little indicator lights to giant displays to light fixtures.

I can assure you that I checked that page a lot when it was new, which I guess explains some things.

Parameters and graphs to care deeply about

(Sorry, this can get a bit dry but at least now you should be able to read a datasheet)

• I_f: Maximum forward current. How much current the LED can handle. There may be several numbers. For example, there might be a design I_f value that all of the specs are based on, plus an absolute maximum I_f that represents the maximum power before the whole thing might blow up. Either way, you want to prevent the LED from having more than this current flowing through it. You will find LEDs that are only 5 mA and this can go up to several amps for LEDs instended to be part of light fixtures
• I_fp: Maximum pulsed power, usually with a duty cycle and frequency attached. This is how much power above I_f you can run through the LED but only briefly. This is helpful because you might be doing some sort of matrixed display where you are powering a bunch of LEDs with a single controller or if you are making a persistence-of-vision toy to increase the perceived brightness of the dots. You might not get this, but some LEDs, like the IR LEDs used in remote controls, can pulse pretty hot for their package size.
• V_f: Forward voltage If you want to drive the LED at I_f, this is what the voltage across the LED will be. This is often times a range, not a single value. This is usually something over 2 volts, if it’s much over 4 volts, you might be looking at a LED that contains a series of LED dies in a single package.
• V_r: Reverse voltage limit LEDs are diodes, so they work as such. However, a lot of LEDs meant for light aren’t actually very good diodes so they can only block a small amount of reverse voltage before they blow up.
• T_j: Junction temperature. The temperature at the chip itself, generally specified as a maximum. A lot of the above paramaters are specified so as to keep T_j below the maximum, so if you exceed I_f, you are probably going to exceed T_j.
• λ_p: Peak wavelength. If you look at the wavelengths produced by the LED, this is the one that’s highest. This is part of you can determine the color of a given LED.
• λ_d: Dominant wavelength. The center of the wavelengths produced by the LED, another way to determine the color of a given LED.
• θ: Half intensity beam angle. This is the angle of coverage of the light emitted by the LED as you received it, without any extra optics added. A lot of LEDs have a tiny epoxy lens on top that narrows the beam angle.
• Φ_v: Luminous flux. Measured in lumens, this is the total energy emitted from the LED, weighted by how it’s perceived by the eye. This ignores the effects of the lens and instead measures the toal light output, so if you have two LEDs with identical Φ_v, the one with the narrower θ will look brighter if it’s pointed directly at you.
• Lifespan: LEDs don’t so much burn out as an incandescent bulb or go blinky like a fluroescent bulb, they just slowly fade to nothingness. If it’s specified, it’s generally rated as the time it takes for them to get to half-brightness and it’s usually also at a given T_j.
• R_a: CRI. This is a somewhat misleading scale, but the closer you get to 100 the better quality the white is, which I’ll talk more about below. This does not apply for colored LEDs, just white LEDs.
• CCT: The color temperature of a white LED, compared to a black-body radiator, where 2000k is really yellow-ish and 6500k is about daylight and 20,000k is pretty darn blue.
• T_j (or sometimes ambient temperature) vs V_f or Φ_v: Junction temperature vs forward voltage or luminous flux. As LEDs heat up, their properties change. The forward voltage and light output both drop. Sometimes the LED manufacturer will give you some plots for these values.
• I_f vs λ_p or λ_d: Forward current vs wavelength. Sometimes the manufacturer will give you a graph of how the color shifts between when the LED is at the lowest power to when it’s at higher power.

Understanding the forward voltage of an LED

LEDs are diodes. Diodes, in general, are really excited to conduct in one direction, so if you take a 9V battery and an LED, the LED will happily burn itself into oblivion conducting all of that power.

Below a voltage threshold, an LED will not conduct at all, like a switch that’s turned off. Once you reach the voltage threshold, however, it will start conducting. At this point, the LED will also start to emit light. The light might come with a bit of a color shift at this level.

As the voltage increases, the LED will start to conduct more and more current. At some point, you reach the rated V_f for the LED, where it’s hitting it’s performance targets like specified color and emitted intensity.

V_f is usually specified as a range because different LEDs will have slightly different forward voltages at the same amperage. Some will be slightly more efficient, some will be a little less.

As the voltage increases past that point, the LED will be emitting more heat than the packaging can transfer away from it and the LED will start to burn up. There is a threshold point where the tiny wires connecting the LED package to the pins will burn up and the LED will never light up again, but before that point, the LED will age rapidly and dim.

Yep, your Light Emitting Diode can become a Smoke Emitting Diode.

Because the forward voltage goes down instead of up as the LED heats up, the LED cannot be self-regulating. You need to use some mechanism to control the amount of power flowing through the LED.

The simplest LED circuit is a “throwie” made by taping a little coin cell to an LED. The coin-cell is inherently current-limited and has some internal resistance and between those two properties, it won’t zap the LED.

What you really want is a resistor, something that looks like this schematic:

You can calculate a resistor value that will limit the current flowing through a circuit, which is the next easiest way to limit the forward power.

You can emulate the self-regulating effects of a throwie by using a constant-current power supply that lets the voltage swing up and down but ensures that there’s always a consistent amperage flowing through the circuit, either by a switching DC/DC power supply or a circuit that acts as a variable resistor. Which I’ll get to later on. Just know that, just like in Star Wars, you don’t have to join the resistance!

LED chains and arrays

You can drive LEDs in series from the same power source, pretty much up to the limits of the power source. If you’ve got a set of LEDs with a 3V forward voltage drop and a 12V power source, you can run roughly 4 LEDs in series. You can even mix-and-match LEDs in this situation; not only is it OK to use a series of identical LEDs, all of which have a slightly different V_f, it’s OK to drive a chain of different colored LEDs as long as they are all rated for the amperage that’s flowing through the chain.

Running LEDs in parallel is a problem, however. Unless you put current control, say a resistor or LED driver, on each LED, the differences in V_f just between identical devices can be enough to zap the LEDs and the second one LED fails, the rest will go.

There are, however, devices that contain parallel and series arrays of LEDs. Presumably the manufacturer is “binning” the LED chips, presumably either measuring them carefully or taking slices from the same die, to reduce the variance in V_f, at which point you can treat the larger device as if it’s really just a single LED and use a single resistor or LED driver.

I wouldn’t bother doing it to save resistors, but it should be possible to save on LED driver chips by building your own parallel/series array of LEDs, especially if you target all of the LEDs running below their specified power and also use constant-current drive or maybe use tiny resistors for each of the series chains to fix any imbalances.

A few packages of note

Discrete “indicator” LEDs

These were the first LEDs you saw. For the longest time, they were all you could get.

Generally between 5mA and 100 mA depending on the package. Generally starting around 2V for the red ones, going up past 3V for interesting colors, sometimes you can end up in the 4-5V range for odd and old LED formulations.

There is a wide variety of LEDs in a few “standard” packages like the 3mm, 5mm, and 8mm through-hole LEDs.

There’s a smattering of common footprints for surface-mount parts.

Surface mount resistors, ceramic capacitors, and LEDs tend to come in a variety of numbered parts, where the LEDs generally range from the smallest 0201 package up to 1206 packages. The numbers represent the length and width, more or less, measured in hundredths of an inch. So a 0201 is 0.02 inches × 0.01 inches, which is extremely tiny, whereas an 0805 is a more reasonably size of 0.08 inches × 0.05 inches. 0805, maybe 0603, most folks can eventually get to the point where they can hand-solder with careful technique and a soldering iron, below that is either mad crazy soldering technique or reflow soldering.

I’ve found that the “PLCC-2” package is reasonably universal.

Otherwise, there’s an absurd variety of specialized packages to be found that will probably require you to make your own footprints.

Larger “lighting” LED packages

Right on the boundary seems to be 5050 LEDs, commonly seen as the basis for a LED strip. There are also 3528, 5630, 5730, and maybe a few other chip LEDs that seem to be popular for LED strips. Unlike the chip LEDs I mentioned previously that are measured in hundredths of an inch, all of these are metric sizes, so a 5050 LED is 5mm x 5mm and, if you were to put it in resistor terms, it’s a 2020.

One standard form-factor that I’ve seen is the 1W / 3W / 5W LED package. I’ve got a strip of really old Lumileds LEDs from the early days of high-powered LEDs and it’s the same size. These LEDs available from all sorts of random merchants on Amazon and Aliexpress.

I’ve also found that there’s a decent number of 2835 and 3535 LEDs that do seem to be roughly equivalent if you look at the datasheet.

Otherwise, the LED manufacturers have decided that they really don’t want you to be able to switch vendors easily, so they tend to make custom packages that might require you to make your own footprint.

These sizes are where you start to see real effort put into thermal constraints and dealing with the power, so these may have a tiny LED array of chips and the datasheet might show the soldering area but suggest that you connect it to a much larger heat-sink.

Some of these packages will have pads on the bottom such that you need to reflow it or other interesting thermal requirements.

A bunch of manufacturers have done a stellar job making a fairly wide variety of these lighting-grade high-power LEDs available, already soldered down to a star-shaped aluminum-core PCB.

Yes, I made that pun on purpose.

Chip-on-board modules

Instead of putting the LED in a package that is then soldered to a board or module, you can deliver it directly mounted to a board or module.

This solves a few problems in a few different ways.

First, every layer you have between the actual LED die and the heatsink means that your thermal transfer is less efficient.

Second, if you want to get a really bright LED, it’s easier to take little LED chips and make an array of them than it is to make one giant LED chip.

So, pretty much, if you take one of the beefy lighting-oriented LEDs from the previous section that’s already coming mounted to an aluminum-core PCB and just deliver it as a that sort of module with everything that doesn’t need to be there deleted, that’s a COB module. Pretty much, there’s a flat aluminum or copper surface on the bottom, the LEDs are mounted atop that, and then there’s some sort of plastic-lens-package thing that goes on top. They don’t need a circuit board underneath them, you can smack them directly atop the heatsink.

These are the beefyboi LEDs that can melt your eyeballs if you aren’t careful.

Some of these require a fairly GIANT heatsink to get rid of all of the heat. Generally these go to some fairly high voltages, 36V or sometimes even more. There’s 100W COB modules floating around that are a 10 x 10 array of LEDs.

LED strips

Flexible PCBs made from a plastic substrate with etched copper on top have been around for a long time, but mostly only as a piece of a commercial product. LED strips are much more recent thing.

There’s a ton of these around, using 5050 LEDs or some of the 5630/5730 LEDs on a flexible PCB. Some newer strips are made as COB modules instead of discrete modules. Pretty much all of these are designed to be driven at either 12v or 24v, with no need for you to add resistors or constant-current drive or anything.

Multi-color RGB/RGBW LED modules

It’s a lot easier to plan on different colors of LEDs right next to each other in the same package than it is to scoot individual packages next to each other.

This means you have a lot more pins to deal with, which the manufacturer might deal with by using a common anode or common cathode.

Also, I mentioned above the graph of T_j vs V_f or Φ_v. Generally, different colors react differently, where the red is usually the color that dims the most as the LED heats up but also has the lowest forward voltage. So, if you don’t have enough voltage, an RGB LED will probably (but not always) turn red and if it’s over-heating it’ll tend to go blue.

Again, you can get these in strip form, as little 5mm LEDs, as beefyboi COB modules, etc.

RGB is a lie

Without getting bogged down too deeply into details, it’s important to know that RGB colors are an extremely convenient lie.

There are colors that you can see that mixing red, green, and blue light won’t let you reach.

Furthermore, the photo-receptors in your eye for color vision are a specific set of wavelengths that your brain then processes in various stages, where you end up with spikes such that the colors red, green, and blue kinda fall out as a useful intermediate approximation.

Furthermore, even if your eye is able to perceive white from an RGB LED, the way the light shines on a surface can distort things.

I’m going to explain this a bit further below because in order for all of this to make sense, you need to understand phosphors and that RGB is a lie at the same time.

Phosphor quality

This is mostly relevant to white LEDs, but there are some LEDs that use phosphors to reach other colors.

Before everybody was obsessing over energy efficient LED lighting (and complaining about how it didn’t look as nice as incandescent), they were obsessing over energy efficient fluorescent lighting (and complaining about how it didn’t look as nice as incandescent).

It turns out that these the same exact problem.

Incandescent lights (and related lights like halogens) work well because they are, for the most part, glowing at a black-body temperature, same as the sun. This means that you’ve got a wide dispersal of light across the spectrum.

Fluorescent bulbs have argon and mercury on the inside that are excited by high voltage. The natural color of one is a blue-ish color that comes from the emission spectrum of mercury and argon, as predicted by physics, which is not a very natural color. Conversely, LED lighting also naturally produces a set of wavelengths based on the properties of the particular materials in the LED semiconductor junction, which are also not natural colors.

Thus, for both fluorescent bulbs and LEDs, to get white or pink or a purple-that-isn’t-actually-violet or other “complicated” colors, you need a phosphor that’s going to absorb some set of wavelengths coming in and re-emit them at different frequencies.

The first fluorescent bulbs had some really nasty color tinges from relatively simple halophosphate-based phosphor formulations, where they’d look white, maybe a bit greenish-white, but the spectrum had huge gaps. Later on, rare-earth phosphors that could blend to make a smoother spectrum became popular, except that they are a lot more expensive both because they use rare-earth elements (which are not as rare as the name would lead you to believe but still fairly expensive) and also just because there’s a lot of variables to control in order to make the process of producing them consistently go right. Furthermore, a lot of folks hated fluorescent bulbs because they were used to incandescent bulbs that pretty much came at a single color temperature and thus end up with a cool-white bulb that they hated instead of a nice warm-white bulb that matched what they had already.

Conversely, there are very similar phosphor formulations that react to blue and emit yellow that exist for LEDs. When you use the cheapest and simplest formulations, you also tend to get a cool-white of relatively poor color quality that isn’t necessarily pleasing. There are better phosphor color formulations for LEDs, although at least some of them are still covered by patents.

Thus, bad LEDs and bad fluorescent bulbs are actually for the exact same reason: the simplest possible phosphor formulation purchased by someone who has other things going on in their life and wants to replace a burnt-out incandescent bulb quickly.

So far, the best whites come from LEDs that, instead of using blue, use a near-visible UV to excite a phosphor formulation, such that the entire visible spectrum comes from a blend of fancy expensive phosphors. This avoids the problems from a fluorescent bulb having a small number of really intense emission spectrums poking out from the smooth phosphor spectrum and also avoids the need to work around the visible blue wavelengths of the LED creating a hump.

Thus, if you tell me that it’s impossible to make a good white LED, I’ll be in ultra-violet disagreement with you about that point. On the other hand, these LEDs with excellent whites come at an efficiency cost as well as a financial cost, which is why we don’t see them everywhere.

For white and white-ish lights, the ability of a white light to accurately illuminate the colors of a surface is usually measured as CRI, where 100 is as-good-as-the-sun, as measured by looking at a set of color samples illuminated by the light source being tested and comparing that to a color sample from a known light source.

RGB LEDs are generally 25-ish, cheap fluorescent bulbs are 50-60ish, the simplest white LED phosphors will get you in the 60-70 range, the cheap LED light fixtures you find in the store tend to target being around 80 these days, and 95-99 LEDs are generally not absurdly priced anymore.

Obviously, if you are setting up a crafting/painting/arting/sewing space, 95+ CRI lights and/or really good incandescents are a really good investment.

In theory, a 100 CRI LED bulb is perfect, however, there are some known methodological issues with the way CRI is designed. Primarily, the color samples that are used are mostly pastel colors. Thus, even a high-CRI LED might not be appropriate for movie lighting and thus there are even more specific scales intended for the those cases and corresponding extremely expensive LED fixtures.

Phosphors aren’t just for white or almost-white. There are pink and purple phosphor LEDs. Neon signage tubes are still available in a wide variety of interesting phosphor formulations that represent all kinds of fun colors. Furthermore, there are “lime green” LEDs that use a phosphor to get something that’s a more-efficient green for color mixing.

I guess I can summarize a few overall observations:

First, there’s a lot of bad LED phosphors out there and this is arguably quite fine for a lot of applications, most especially when you viewing the color of the LED directly instead of using as illumination.

Second, I also think there are interesting LEDs that use phosphors that aren’t white that could be made.

But, mostly, note that if you are buying LEDs in any form and they aren’t trumpeting their high CRI values, they probably have a mediocre CRI from using the least patent-encumbered and cheapest formulation and if you try to solder up a zillion of them to light your room you won’t like it.

Wavelengths and why a rainbow of LEDs in interesting wavelengths is fun

I have a collection of LEDs in all sorts of wavelengths.

In theory, you can mix any visible color with just a little bit of red, green, and blue, but that’s theory, not reality. One easy way to see this is to get a violet or UV LED and watch things fluoresce but it’s not confined to blacklight.

Ergo, and the details are much more complicated, an orange LED is more than just red-plus-green. A spectral-violet LED is going to light things differently than a phosphor-violet LED. The dye on a fabric or pigment on a publication reflects light based on a continuous spectrum and you might notice that the colors have gone weird even if you can’t tell the difference when you look at the LED.

RGB vs RGBW vs RGBWW/RGBYW/et al.

As I pointed out previously, the CRI for an RGB LED is like 25, which counts as a pretty awful white, whereas even bad white LEDs generally start higher than that. Furthermore, white LEDs with phosphor in them are really efficient, generally far more efficient than the equivalent collection of RGB LEDs.

Obviously, there’s a lot more to an RGB LED than making a pretty mediocre white. The CRI of 25 is a bit deceiving because it’s not a bad white to look at if it’s diffused so you can’t pick out the individual LED chips, it’s just that you’ll realize how inaccurate the color is if you try and use it to light a room. After all, your computer screen is an RGB device not an RGBW device and it doesn’t feel wrong or anything.

There are a lot of RGBW LEDs out there these days, and that’s a good thing.

If you just want red or green or blue, there’s no real difference from an RGB LED. However, if you want a pale blue, you can raise the level of the white LED instead of raising the level of the green and red LEDs. This results in lower energy consumption (or higher brightness) and also means that the white LED is “filling in” some of the wavelengths not represented by the green and red LED to allow it to work better as illumination.

There are even more complex LEDs to be found.

If you want white but within a range of color temperatures, you will get a better version of white fading between two different white phosphor formulations than you will by adding blue to cool the white and red to warm the white. RGBWW LEDs with two whites are quite popular, especially for intelligent lightbulbs, for this reason.

There’s absolutely merit to having three whites, maybe an extremely cool (20,000K) ice-blue white on one side and something that’s either a very warm light (Maybe 2000K) or even a phosphor-based amber (generally sold as “PC Amber”) on the other.

Obviously, adding 1 or more whites isn’t going to help you reach saturated pure colors, just really good desaturated colors.

On the other hand, there are also saturated intermediate colors out there, so you can sometimes find or construct your own fixtures with intermediate colors like yellow or orange or cyan or violet.

All of this comes at a bit of a complexity cost, because you need more LED driver channels and may have to write some library code to make sense out of the color mixing problems.

The usefulness of RGBW/RGBWW/et al. and other more sophisticated LED configurations is dependent upon what you are working with and what sort of results you expect. If you just want orange light, maybe you just want an orange LED, which will give you a better orange than you’d get from RGB mixing. If you are just dealing with intense saturated colors, RGB is probably all you need. But getting good at the software and hardware problems of mixing the light from 4-6 LEDS might be a good thing to experiment with.

Which terminal is which?

Here’s a few ways to… well, positively identify… the anode:

For surface mount LEDs, I’ve definitely run into some surprises here and there. There’s usually a mark and it’s usually as I’m showing, but sometimes I’ve had some nasty surprises that required some cursing and very fiddly desolderng.

Also, some of the smallest LEDs don’t have enough space for a mark or notch. Good luck there.

Where to go from here?

I guess this wasn’t very brief after all?

Either way, here’s some more avenues for investigation:

• Intelligent LEDs
• LED strips
• Driving LEDs
• Watt do I do with the heats?