Aaaah, the light bulb. An eminent monument to human ingenuity. With the movement of insignificant particles, man has created light for use in those times when God wouldst take it away. There is so much we can say about the light bulb. And yet there is so much we really shouldn't say, lest feelings be injured. I'm in the LED camp. I'm a guy who firmly thinks LEDs will one day overtake other technologies as the lighting elements of the future. Well, only of the future, anyway. Currently, LEDs are basically not worth anywhere as much as they cost. The incandescent bulb for which we revere Thomas Edison is on crutches. It's weak and pathetic and hopelessly inefficient. The fluorescent will fall in time. Halogens and HID lights are all doomed to failure for the long run. They're not solutions in this world of everything in high cost. Petroleum is high cost, health care is high cost, internet access is costly, and sure enough, electricity is costly... especially for a country currently in recession.
Well, I make no promises that the LED will save us from all of these things. Petroleum costs are Dubya's fault. Internet access is due to companies wanting to milk the cash cow and telecom companies not willing to upgrade equipment because they suffered during that brief period when dot-coms could whip their butts. Well, electricity will still be expensive because so much of our electricity is powered by petroleum in spite of what most people might think. But, hey, I'm on the topic of saving power here. And this applies not just to our own homes and apartments... I'm also talking about our street lamps and automotive headlights and office buildings and camera flash bulbs and monitor/projector backlights... everything where you would need a source of illumination.
I'm generally a fan of simplicity. I like approaches that are simpler in favor of excessively complex technically involved solutions. That's why so much of my research in video compression involves avoiding motion compensation methods. That's why among all global illumination schemes, I like Monte Carlo Path Tracing. That's why I think all internal combustion piston engines need electromechanical valve actuators instead of camshafts... especially not multiple cams for the sake of variable valve timing. That's why I think all current-day cars should use Anderson CVTs, which are devilishly simple and can give you everything... although in the long run, if one day, electric vehicles become even remotely feasible, the whole direct-drive electric motor at each wheel approach would be the best solution. One of Burt Rutan's philosophies, which I happen to like a lot, is the idea that the best solution to a problem is often a *low-tech* solution. Well, LEDs are not necessarily the absolute simplest or lowest tech in comparison to the likes of filament incandescent bulbs or even fluorescent. But even so, they aren't plagued by many of the problems that we might have with other lightsources and are far more flexible in so many ways, and ultimately hold the crown in technical superiority.
To call an LED an LED or light-emitting diode is technically redundant. All diodes emit some sort of electromagnetic radiation. It just so happens that LEDs are made of specific materials and doping concentrations that their emissions would actually be in the visible light spectrum. Typically for LEDs, this is some sort of Aluminium-Gallium-Arsenide combination, but there are several others depending on the color. The diode is basically the simplest of all semiconductor devices -- it's a sort of switch that lets current only flow in one direction. You have an N-doped semiconductor, and a P-doped semiconductor. You create a bond between them and free electrons in the N-field join with free holes in the P-field. Actually the holes don't move per se, it's more like electrons move so that the hole concentration shifts, but it's a lot easier in mathematical terms to think about holes as positively-charged particles that can move. Anyway, the closer electrons and holes meet and lead to a region of space around the P-N junction becoming depleted of electrons and holes. That depleted region is basically an insulator, and prevents the further-away electrons and holes from meeting. Anyway, when you apply a voltage to this such that the anode meets the P-side and the cathode meets the N-side, the electrons and holes are repelled away from their terminals, and start to shrink the depletion region. Eventually, the depletion region is shrunken to virtually nil, and the electrons and holes are free to move, and you get a conductor. It takes a certain minimum voltage to reach this point, but once that happens, the majority of electrons in the diode reach conduction band energies, which is an upper valence band where electrons are free to move. When an electron of conduction band energy meets a hole, it drops to a lower valence band, which means you emit a photon. With most diodes, this is a pretty small hop, so the emitted photon is in a low-frequency band like infrared (and that's the band that we usually use for remote controls). In an LED, we simply have semiconductor materials and dopants that create a bigger band energy difference. An interesting effect of this is that overdriving some LEDs can actually increase its luminous efficiency, albeit at the cost of lifespan. This is because when you're at the typical drive strength, you have only a certain percentage of electrons that are actually high enough energy to be *well* within the conduction band. Having more drive strength can put more electrons well within the conduction band. Still, there are other LEDs that can actually have more luminous efficiency when underdriven because of the materials and the effective rate of diminishing returns. LEDs are specifically constructed in terms of crystalline structure to let as many photons out as possible, else they would just be absorbed by the semiconductors, and some of these materials may not respond well to overdriving.
So why do I love LEDs? Well, they're a lot more durable than any other type of bulb because they're so small and dense. You can fit them into dense circuitry more easily. They don't have a filament or any open electrodes or phosphors (well, actually some of them do), so they last a lot longer and never really burn out unless you severely overdrive them -- and it's also possible to make an LED in certain colors that is designed to exhibit safe breakdown cycles so as not to incur heavy damage from overdriving or reverse driving. Longer than fluorescent by 10-fold, and longer than incandescent by 100-fold. Even those specific white-light LEDs that happen to use phosphor materials can outlast fluorescents by 2:1 or 3:1. But where LEDs really kick ass is efficiency. Typical incandescent bulbs are around 10% luminescent efficiency -- that's because the filament has to get really hot, and therefore out of your 60 watt incandescent bulb, 54 watts are wasted as heat. In fact, lower wattage incandescents are generally less efficient than higher wattage ones because the higher wattage ones often have thicker filaments that evaporate less slowly and can work at higher temperatures. Only the remaining 6 W actually comes out as light. A halogen is only a little better at around 15%. Fluorescents get pretty impressive at 30% efficiency and some well above that. Part of the efficiency loss with fluorescents is due to the way they work -- you have mercury vapor emitting ultraviolet light which is converted to visible light by a phosphor coating. This is not a perfectly efficient process, and there is still quite a bit lost to heat because the mercury vapor is not an excellent conductor of electricity by any means. Although, in terms of lumens, fluorescents still exhibit the best numbers -- that will change in time. Ah, but the LED... the LED can easily achieve luminous efficiencies around 90%. And there's even research going on with quantum dot methods to raise LED efficiencies to over 99%. They still waste some power as heat, but interestingly enough, because it's semiconductor based, they actually have a negative temperature coefficient of resistivity. That is, the resistance goes down with temperature. So what does this mean? It means that you can make a screw-in bulb made of multiple LEDs that can be as bright as a 30 W incandescent bulb and burn less than 4 W. Not only that, but you can burn at less than 4 W with a lifespan approaching 100,000 hours (11 years). Moreover, LED lifespans are not really to the point where they simply don't work anymore -- they're actually measurements of time before the LED's brightness for the standard drive strength is down to about half.
So for all that, why haven't we all switched to 100% LED based lighting? Well, cost is really the biggest factor. I mean, the aforementioned 30 W-equivalent replacement bulbs can easily go for around $80 US. Now, while that works out cheaper in terms of electricity savings and not having to replace bulbs for 20 years or so, how in the world can you EVER justify spending $80 on a lightbulb? Bear in mind that higher luminous efficiency doesn't just save you money on your lighting bill. There's a reason that stores use fluorescent lighting. They produce so much less heat than incandescents that it also saves them money on air conditioning. It makes it possible to have open-shelf freezers that successfully keep frozen foods frozen. Now imagine the same thing on a scale of 3-4 W instead of 25+ W. And also, fluorescent bulbs last long enough that you can justify the extra cost of the bulbs themselves, which is very small compared to the price of LED bulbs. Also, fluorescents last a fair bit longer in stores than they do in the home because they're generally left on every hour that the store is open, and see very few on-off cycles, which is one thing that adversely affects the lifespan of fluorescent bulbs. On-off cycles make very little difference to incandescents and basically no difference to LEDs. This would be true of LED replacement bulbs too, but they're not available in high enough volume to feed entire franchises. And either way, it would far too high a one-time cost because they're just so expensive right now. Either way, there is a difference between luminous efficiency and luminous efficacy. That's why for now, fluorescents are still king. Typically, for incandescent bulbs, you get a luminous efficacy around 15 lumens/watt. For a superbright white LED, it's often around twice that -- 30 lumens/watt. There are exceptional lab cases from companies like Nichia and Cree where they get around 50-74 lumens/watt. But with fluorescents, the range goes around 60-100 lumens/watt, with 85 being the typical.
The main issue with LEDs insofar as luminous efficacy is the fact that they're very directional emissions. This is unlike an incandescent or fluorescent bulb where you can have a full 360-degree emission of light. Also, it's currently not easy to make a good white LED. Because of the fact that LED lights illuminate only a small angular range, you need several of them to illuminate an entire room. And that's why, even though you can get a 4 W bulb that can replace a 30 W incandescent bulb, it's very hard to go beyond that no matter how much power you provide. Sure, there are a few bulbs out there that have diffusers on them to send the light out radially, but when you diffuse, you will lose brightness, and hence, luminous efficacy goes down.
If you've ever looked at listings for various LED-based screw-in replacement bulbs, you might see something that will make you think you're being lied to. For instance -- a bulb made with 18 LEDs that is equivalent to a 15-W incandescent draws some 30 mA at 120V. So the bulb as a whole draws about 3.6 W. And then, you can also see a bulb made with 36 LEDs that is equivalent to a 30-W incandescent, and it draws... 30 mA at 120V for a 3.6 W power consumption. Now that sounds odd... how can twice as many LEDs consume the same amount of power? Well, it's not quite a lie, but it is a feature that doesn't scale up ad infinitum. It has to do with the drive circuit behind the LEDs. The thing to realize is that a diode doesn't really consume a significant amount of power. If the voltage across the diode is less than the "on" voltage, it's essentially an insulator, and if above, it is essentially a conductor. The light that is emitted is not really a power-drain, but rather a result of the moving current -- aside from turning the diodes on, the voltage doesn't really matter, only the current does. There is only a small amount of actual power consumed by the LEDs themselves which is wasted as heat. There is some small internal resistance, which actually goes up as the power sent through the LED goes down. That means that it's safe to wire several LEDs together in series so long as you have enough voltage to power them into the on state. Let's say, for instance that you have LEDs that turn on at 3 V (which is actually pretty normal for white-light LEDs), and based on the previous example, lets say they were designed for 30 mA of current... now that means that if you have 40 of these LEDs in series, you need 120 V to turn them all on. Since our power source only goes up to 120 V, we need to go a little lower, so 36 LEDs... Now 36 * 3 = 108 and 120V - 108V = 12 V, which means we need a pulldown resistor that will give us a 12 V drop and keep the current down to 30 mA, so we need a 12 / 0.030 = 400 ohm resistor. Pretty simple -- high school physics stuff for me. And similarly, for the 18-LED example, we'd simply have a 2.2k resistor (120 - (18 * 3) = 66V, 66 V/ 0.03A = 2200). Of course, we do need to make sure that the resistor can handle the wattage it consumes. For instance, our 400 ohm resistor would use 12V * 0.03 A = 0.36 W. For reliability, it's generally a good idea never to exceed 60% of a resistor's wattage rating, which in this case means we'd need a resistor rated for at least 0.6 W. In the 18-LED case, we'd need at least a 3.3 W rated resistor. Generally, widely available resistors are not rated very high -- on the order of a 0.25-0.5 W. But it is reasonable to use more than one resistor in series to get the same total effect, and companies that sell them could just as well use not-so-widely available resistors and freely charge some $140 for a bulb.
Actually, the 36-LEDs at 3 V each is not that good of an idea when connected to an AC source. You'd actually want to keep the total voltage of a series of LEDs not much greater than 80% of the total voltage or, in this case, 96 V, which means an on voltage of around 2.7 V. Also, given the 30 mA drive current, it's probably a mixed color variety. There are multiple ways to get white LEDs, and the most common is to use a blue LED and a phosphor coating that partially absorbs the blue light and emits a spectral emission ranging from red to green. These work all right, but they have less lifespan than the typical LED because the coating doesn't last forever. There are methods that use a single yellow phosphor, as well as methods that use multiple phosphors of various colors. Obviously, using multiple phosphors gives you a wider range and coverage over the visible spectrum, but is more costly to make. Yet more costly and broader in spectrum is done using red, green, and blue phosphors (unlike the former cases, they're dense enough not to allow any source leakage) excited by an ultraviolet LED emission. Many manufacturers tend to believe, though, that the future lies in mixed emitters. That is, a single LED device would contain multiple smaller P-N junctions of various color emissions that mix to create a white light. For instance, using a blue and a yellow emitter or red+green+blue emitters, or even red+green+blue+yellow emitters. This is in general, more efficient because it doesn't experience the loss that results from the wavelength conversion. The only real problem here is the fact that it is very hard to get a correct mixture of different colors to make a "true white light" result. Aside from that, multiple emitters generally means higher on-voltage, so it's hard to get as many of them in the same circuit. And yes, they're yet more expensive.
About the cheapest incandescent replacement LED bulbs I've seen online are Mule's DynaLUX LEDison series (referred to as LEDison because they plug into Edison screw lamps). And those run for about $50 a bulb for the white versions. There are some that get cheaper technically, but many of them are excessively small (e.g. made to replace 15 W), or have poor angular spread (less than 90 degrees). At least the LEDison series is supposedly valid to replace everything from 10-60W incandescent bulbs and give you a wide near 360-degree field. I've seen bulbs that could theoretically do a better job of room lighting, but they go for close to $170 a bulb. I still don't find it feasible to pay $50 for a light bulb, but at the very least, it's a sign that prices could come down at some point to acceptable ranges.
As much as having a very directional emission is not suitable for room lighting, it's very suitable for other tasks. As it is right now, a large number of states are switching over to LEDs for traffic lights. You can easily recognize them by the fact that you'll see a traffic light made up of several dots that are uniform in brightness across the entire light rather than bright in the center and fading out towards the edges. It proves to be a major savings in power consumption to use 15-20 W worth of LED clusters as opposed to 165 W bulbs that have to be replaced 10x as often. And for 1 million traffic lights, that yields almost $2 million in electrical savings per year, and they're much brighter and visible from greater distances to boot.
Within the traffic itself, you're most likely to see LEDs in taillights. This works out nicely because they'll generally be a single color (red), and they will project light very brightly to the drivers directly behind. And of course, use a lot less power than bulbs and last many times longer than the vehicle. For headlights, there's still more research to be done (with the Japanese automakers putting the most effort into it). More luminous efficacy is needed along with the ability to project the light into patterns that provide suitable coverage of the road ahead. The difference with taillights and headlights is that taillights are meant to be visible to others while headlights also need to help your own visibility. When you have a lightsource like an incandescent bulb, the directional output is predictable, and therefore, it's a standard problem as far as creating a mirror or a lens that focuses the light the way you want. With an LED, you need a big array of LEDs to get the appropriate brightness, but when the light is coming very directionally off of a large surface area rather than a tiny filament, it's a much more difficult optics problem. And in fact, this is also why you don't see fluorescent bulbs in car headlamps... the lightsource is not a tiny filament, but the surface of the bulb itself. And to get a decent brightness, you need a large surface area. Unfortunately, there aren't very many usable tricks that allow you to minimize size for surface area the way you can take a meter-long filament and make it into a double coil and shorten it down to an inch. And again, it saves power and it can potentially be a lot brighter. Of course, when and if electric cars really become feasible products, the power cost of headlights is not that big a deal -- saving 100 watts on head and tail lamps isn't all that significant compared to the 150,000 watts (about 201 hp) you'd need to drive the vehicle itself... but every little bit helps when you really need to be stringent with your electrical power. Besides which, the age of viable electric cars is still way down the line.
In some cases, you can get more effective brightness out of LEDs by pulsing them very quickly. One of the things about LED dice is that their low current limitations are actually average current budgets. Often times, the highest luminous efficiency can come at much higher currents, even though that might be bad for the lifespan of the device. Fortunately, though, you can use pulsing to use very high currents momentarily and keep the average current low. The effect takes advantage of the non-linearity of human visual perception to smooth out the end result. One of the things about the human eye's nonlinearity is an effect that comes from its very accurate time-integrator. So as long as the pulsing is fast enough (say 100 Hz or something), we see it as continuously bright at a brightness about even with the average brightness. However, this doesn't really do much good with current day "ultrabright" LEDs because they often have maximum efficiency currents and maximum rated currents that are pretty close. For it to really work well, you need a significant difference in luminous efficacy between the high current and rated current to really see a difference. Say hypothetically, that you have an LED rated at a drive current of 10 mA. But the LED happens to be most efficient at around 25 mA. Well, what we could do is pulse the LED on and off at 20 mA, and thus, maintain an average current of 10 mA. Of course, we also have an average brightness (or perceived brightness) around half that we'd see at 20 mA drive. So as long as the LED shows more than twice the *apparent* brightness at 20 mA (which if the efficiency is higher, then it should), then the overall apparent brightness of the pulsed LED is higher than standard constant-current drive. Well, nowadays, this isn't as useful because the superbright LEDs are designed for high drive currents, and usually the max efficiency is found only at a minor increase in current over the drive strength. Typically, when the drive currents are above 20 mA, pulsing isn't worth it.
There is yet another side to the LED coin. The Organic LED, AKA the Polymer LED. Actually there are more than a few forms of this item -- the FOLED (Flexible), TOLED (Transparent), SOLED (Stacked), PHOLED (Phosphorescent), QD-OLED (Quantum Dot)... All of which have their uses. OLEDs will be the flat screen of the future. LCDs and Plasmas are utterly inferior... well, they will be. OLEDs still have a long way to go. Unlike their bulb counterparts, OLED screens currently still have horrible lifespans. The worst of these are the blue OLED elements, which only seem to last around 2,000 hours. Typically, these types of screens are not considered economically feasible for public sale until all three components (red, green, and blue) last over 10,000 hours. This may not sound like very much, but when you consider a car lasting for 250,000 miles, that's only 10,000 hours at an average speed of 25 mph. Similarly, we wouldn't really watch TV 20 hours out of a day. Granted, for them to be usable in an office environment, we'd need some more lifespan because a computer screen will probably be left on for at least 8 hours out of a day. Currently, the red and green OLED components are already above that 10,000 hour mark, but blue is a long way from that.
The OLED is destined to take the display crown away from LCDs, Plasmas, CRTs, LCoS, etc. Currently, you may see them on the occasional cell phone or digital camera where the expected usage life of the product may not be very long -- conveniently a lot of people first heard about OLEDs from the Philips-Norelco electric razor used in the James Bond film 'Die Another Day'. My own cell phone screen is an LCD with an OLED backlight (and it shows with the way I can almost use the screen as a flashlight). Cell phones may be ubiquitous, but with very expensive cell phones (which are most likely to have true OLED screens), people tend to buy the phones without any intention of keeping them for years and years (the typical buyers of expensive phones tend to make 6 figures and have money to burn). Conversely, while you might keep a digital camera for quite some time, the occasion to take photographs doesn't present itself very often unless you're a professional. For OLEDs to take over the monitor and TV area is really a matter of time... specifically, time to further improve on the materials and effective lifespan. Better plastics and better barriers still need to be worked out, as do better drive circuits. Bear in mind that it doesn't take a whole lot for color shifts to really become apparent. There are cases where a company had its product at an expo while, at the time, only having blue panel lifespans of 2,000 hours (~83 days). While that sounds all right for its time (early 2002), the end result was that they had to swap out panels 4 times a day because the blue fade was visible within a single day. With OLEDs like any other LED, these lifespan figures correspond to the panel losing half of its brightness, so it's only mildly surprising that a group of industry experts who are probably well trained to notice minor color shifts would notice something within the course of a few hours of continuous running... the problem would really be exaggerated by the fact that the difference in lifespan between the blue and R/G panels was probably very significant. If all three panels had similar lifespans, the end effect may not have been as noticeable.
So what advantages do OLEDs bring to the table? Of course, power consumption... that should come as no surprise. OLED screens are inherently emissive like Plasma or CRT, so they don't require backlights, polarizers, color filters, etc. And because of this, they also exhibit very high contrast and have wider viewing angles than any LCD. The theoretical color reproduction range of OLED screens is beyond that of any other display (although the lifespan issue still makes that difficult from a color balance standpoint). Still, being an emissive plastic panel means that they don't need any extra electronics for actual capability to display, so they can be made into the thinnest and lightest imaginable displays. Also, like other LEDs, the actual switching time is very quick... on the order of microseconds, so scan speeds are completely a non issue, and this feature is relatively unaffected by temperature. Being a completely solid state device means no open cell, and no need to store extra components within a vacuum. They still need some sort of protection from the elements, but this can be achieved through coatings just as easily as through some sort of sealed housing. And unlike any other display, OLEDs can be made flexible. A few firms have demoed screens that can be rolled and unrolled like a projector screen and function normally.
Aside from the materials problems, there's still one more that makes OLEDs difficult to handle. Unlike LCDs, it's not the voltage that controls brightness/darkness. With OLEDs, you need to maintain at least the on voltage, but control the brightness via current. With large screens, this is not easy. It's hard enough to swing voltages up and down along these long wires leading from addressing multiplexors to pixels (a bunch of long wires close together separated by insulators makes for a lot of capacitance), but when you have to swing current... Larger screens = longer wires = more resistance. Because the on voltage is more or less a hard-line, the actual voltage swing doesn't need to be very large, so the capacitance is not as much of an issue, but trying to control current with nondeterministic resistance is just as hard if not harder. This often translates to the inability to provide sufficient drive strength far away from the addressing multiplexors, and thereby having to underdrive every pixel so that everything looks even. While underdriving does improve lifespan in general, it makes for a duller, undersaturated color reproduction, which effectively kills the marketability.
The usual Indium-Tin Oxide electrodes are fairly low resistivity, but are far too rigid, and in turn, far too brittle for flexible displays. The PEDOT conductive plastic electrodes are certainly flexible enough for extremely flexible displays, but are highly resistive. Moreover, PEDOT is currently not suitable for active matrix TFT drivers. This is because TFT drive transistors are largely polysilicon, which is processed at several hundred degrees centigrade. Work is in progress to utilize quick laser pulses to anneal amorphus silicon into polysilicon, which would lower the effective temperature (high temperatures are only instantaneously achieved). Visually, OLEDs don't really show that much difference between passive and active matrix like LCDs do; it's just much harder to get sufficient drive current to achieve high brightness. Also, because of the small voltage swings and the fast switching times, passive matrix OLED screens need to refresh at extremely high frequencies, which can reduce the apparent brightness in terms of human visual perception either way. Regardless, active matrix displays require less current because line addressors simply have to signal control transistors, which can hold a current for some time until a future refresh.
SOLEDs (Stacked OLEDs) are also a nice methodology, especially for medium to small size screens. It is simply where each of the three R,G, and B portions of a pixel are on separate panels. The panels are transparent (except maybe the rearmost... all three are transparent only if totally transparent displays are desired). One of the main advantages is that it raises the effective resolution of the panel because we don't need to have neighbouring RGB cells. Also, the color of each pixel will actually appear to be the true color rather than an rgb triplet -- similar to the effect of a dye sublimation printer. They're not really necessary on large screens because large screens don't need the extra resolution enhancement, but more importantly, each panel needs to be separately addressed and driven, which is somewhat more of a power drain than a panel full of triplets -- the difference is small but grows with larger and larger screens. However, SOLEDs are cheap to produce simply because you can make entire panels of a single wavelength.
Quantum Dots or CdSe/PbSe nanocrystals could theoretically help on the OLED front as well. Quantum dots can be easily tuned to emit light of certain wavelengths, and the luminous efficiency is typically outstanding. Not that long ago, a team at MIT had worked out a process that only requires a single layer of nanocrystal phosphors, and hence, achieve extremely high luminous efficiencies approaching 100% (key word being 'approaching'). Of course, a small team of academic researchers who don't have access to true fab facilities cannot really produce a proof of concept on large scale devices, so it's still yet to be seen how the endgame would really be played out with QD-OLEDs for displays. Using QD phosphors is something that lends itself well to SOLEDs, though. The use of inkjet processes similar to what Epson uses for their big OLED screens could be made even easier without having to worry about cell separation. Inkjet processes do tend to be rather fussy in general, but when your entire phosphor range is the same material, it's not really a concern. Inkjet techniques are also cheap from the standpoint of cost of operation, because they don't require clean rooms and the like. A friend of mine at RIT works with inkjet IC fabrication and actually is able to keep the entire process running on the desk of his office.
Well, it all sounds really great with a whole lot of problems that sound solveable in the long run. I'm also not an electrical engineer, so I can't really For now, though, it's all just a few people rambling on about how great the future of technology will be with all too few people actually doing something about it. That's the problem with the future -- it's never near enough.