You can thank the DoE and its expanding Energy Star standards for outdoor lighting for a lot of the recent progress in high-brightness LEDs. Add to that an earlier DoE call for 160 lumens/W from LEDs by 2011 and a lifetime exceeding 10 years (approaching 100,000 hours). No wonder, then, that IC makers are quickly mobilizing with new reference designs for LED controller-drivers, new on-line design tools, and special chips for driving industrial, commercial, automotive, and outdoor applications.
International regulations and business models for retooling city streetlights and the like, however, are yet to standardize. The total-cost-of-ownership for LED-based systems creates a fair degree of skepticism in some arenas. But the “payback” time, a major factor in those market segments, may well drop to just two or three years by 2011 or 2012 thanks mainly to rapid gains in white-LED technology and volume production. At least that's the way more chip makers are betting.
Market watchers think that forward lighting for vehicles (i.e., headlight systems), on the other hand, will struggle a bit. Large-area lighting systems for the home versus outdoors also face a mixed reception. But higher-power RGB projection applications look promising, with anticipation of great improvements to the traditional color-wheel system.
Typical LED efficacy is perhaps five times that of traditional incandescent lightbulbs and by some accounts now approaches or exceeds sodium- and tungsten-based emitters. The hitch is that LEDs are basically semiconductor devices that dissipate the bulk of their heat through conduction, not radiation. Thermal management is thus the main priority for chip and system designers; close behind are maximizing the reliability and efficiency of the supply powering the system.
“There's two terms people are going to be starting to look out for: L70 and L50,” says Rick Zarr, National Semiconductor's (Santa Clara, Calif.) PowerWise brand technologist. “In LED-speak, the L70 is the life of the LED to the point where it diminishes to 70% of its original brightness. When you were driving a little 1-3/4 indicator LED with a 10mA, it would last 50,000 hrs never got warm. But, if you're going to spend the money to install a fixture that is more efficient, you'll need to manage heat so the LED lasts for the life of the fixture. One of the things that is absolutely destructive to the LED is a rise in temperature above its recommended operating point.” Thus in practical terms, he says, the chip designer wants to limit the maximum junction temperature of the typical LED to 100°C.
Electrically, the power source/driver needs to provide a constant-current for the LED and a dimming function. A distributed LED architecture and a modular supply architecture are the key to success for the higher-power systems.
“For street lighting, you might have, say, 60 to 80 LEDs in a single fixture, each driven at 350 mA,” says Bernie Weir, ON Semiconductor Computing and Consumer Group Director of Applications (Phoenix). “And it needs to be a modular approach where they can expand the number of LEDs, depending on the pole height and so on.” Physically, the old fixture can be retained and retooled, or the head is pulled off and the streetlight replaced with an entirely new fixture. If the shell of, for instance, a sodium-vapor installation remains in place, the old lamp is removed and replaced with serial/parallel strings of LEDs. The ballast might be left in the fixture to cut the expense of removing it. Inductors are often retained to maintain good EMI performance.
Given the physical and thermal issues, “what that means for IC makers is that it's important to have high efficiency drivers,” says Weir. A typical power supply example might be a PFC (power factor correction) stage, followed by a fixed-frequency dc output at 24 or 48 V, and then multiple strings of LEDs powered from that. Beyond these considerations, designers want to catch the benefits of ambient light-sensing circuitry, particularly for streetlighting. Light-sensing is a way of implementing LED dimming for free, notes Alfred Hesener, Fairchild Semiconductor Europe marketing director.
Several new products have arrived for simplifying general lighting tasks. One on the ac/dc side is ON Semiconductor's NCP-1652 single-stage power factor correction and step-down controller, for which a new 100-W reference design (AND8394/D) is available. Also, STMicroelectronics developed its L6562A transition mode PFC controller for primary-side control, an improved version of its L6562 suited to LED applications when used with the company's TSM1052 constant-current driver.
Other circuitry includes a new protection IC for multi-LED applications in streetlights, headlights, and runway lighting. ON Semi's NUD4700 sits in parallel with each LED to shunt the path of any LED that fails open.
National Semiconductor says its LM3424 constant-current controller is the industry's first to offer thermal foldback circuitry. It drives two external FETs for up to 20-A applications. The company introduced its Webench design tool for lighting last year. Enhancements for thermal foldback design are getting ready to launch. This section will be suited for any general thermal design application.
On the road again
LEDs are old hat when it comes to automotive signal lights and relatively low-power (e.g., 20 to 50 mA) applications. “We already see 200 to 400 LEDs in vehicles,” says Alfred Hesener. “HIDs currently in automotive dimmers are close to extinction. It's awkward, they need more than 400 V and complex electronics.”
But so-called forward lighting systems, i.e., the headlights, are another matter. Indeed, major headway is likely five to six years off. “There are some quite good reasons not to use LEDs in headlights at the moment,” says Peter Green, Senior Systems Engineer for the Lighting Group at International Rectifier (El Segundo, Calif.) “Headlamps need to generate a lot of light, meaning you're drawing a lot of power, which means you might need to use a fan in the headlight assembly.”
Indeed, in these applications (often specified to a few dozen LEDs and upwards of 100 W) it's that pesky fan that has designers backing off. “LEDs will take over, but for headlamps slower than sooner,” says Brian Blackburn, Senior Field Applications Engineer, ON Semiconductor. “LEDs for a given lumen output compared to HIDs take only about 40 percent of the electrical power. But it's a two-dimensional thermal problem, and a lot of these LED headlamp assemblies have a fan in them. LED headlamps will get there but will not catch on until we can eliminate the fan - it's got to go.” The dc/dc converter design also needs to cut EMI - almost as important an issue as thermals and cost.
Most qualification processes in vehicles take the long road. But it's been done, notes Steve Pietkiewicz, Linear Technology Corp. VP and general manager of power products (Milpitas, Calif.). Consider the Cadillac Escalade SUV, reportedly the first sports utility vehicle (but not the first automobile) to use a LED headlamp assembly. It's built by Hella (Germany) and driven by Linear Technology's LTC1871 boost/flyback/SEPIC controller (with external MOSFET), which took five years to qualify for the application. Cold crank and peak starting current requirements are among the specs the controller-with-external driver MOSFET arrangement needs to meet.
Ultimately, the lighting issues extend as much into the car's physical design and product identification as anything else. “While we talk about energy efficiency, customers use LEDs as much for styling as for reliability,” says Bernie Weir. “Almost all the Audis have a special daytime LED running light design such than when anybody sees that car coming, they know it's an Audi. It goes beyond lighting.”
In the meantime, manufacturers are trying to address a growing number of mid-range applications. “An example of an application that has broken out on higher currents is the rear combination stop lamps (RCLs),” says Weir. “The primary reason is styling. The secondary reason is reliability. These are typically from 12 to 24 LEDs that are pixellated along the entire surface area of the stop lamp. We're talking from 250 mA to 500 mA. Those LEDs are typically divided in several strings of three.”
ON Semiconductor plans to introduce a linear current-regulator and controller later this year that will be suited for automotive rear combination lamps. The chip will control up to 24 LEDs directly from the battery. An integrated 800 Hz oscillator and PWM controller will allow precise tail lamp dimming from zero to 75% . In addition the device contains a diagnostic signal for the detection of any open LED string, thus allowing vehicle signaling and latch-off control for lamp functions such as the turn signal.
In another midrange application, the firm is also looking closely at its NCV3065 DC/DC controller for HB-LEDs, which is suited to buck, boost, or buck-boost operation, for use in fog lamps (15 to 22 W) and daytime running lamps (7 to 10 W).
The video projection market, another sector of the mid-power application arena for LEDs, is also poised for a leap. Most of the recent excitement in LEDs has come from Luminus Devices Inc., Billerica, Mass. Color accuracy in such projection and other RGB systems is a function of the ability to finely mix three colors. Ultimately, that's a measure of the driver configuration to quickly respond from zero to full current in controlling the red, green, and blue channels. Linear Technology Corp. says its just-released LTC3743 controller chip, using two separate control loops, has a response time of just 2 nsec for a nominal current change from 2 to 20 A. Beyond its use in projection systems, Linear says the chip's high efficiency (90 to 95%) and small solution size (625 mm2) suits it to general lighting and other high-brightness white LED tasks.
The home stretch
Residential lighting with LEDs likely faces an extended period of development. Outside lighting, versus what's installed in the home, looks more economical. The major issues are cost and aesthetics. Some recent product arrivals include Fairchild's RD226 reference design for lower-power ac-input lighting systems using HB-LEDs. It features the FAN102 device used as a primary-side-control PWM controller.
When LEDs do go in, they'll tend to be installed outside in the front yard, notes Peter Green. “It coming, but it'll take five years,” he says. Cost is the problem, with industry estimates pegging LEDs at from two to seven or eight times that of traditional installations. “You may see them now indoors in small little candle lamps, but they don't give much light. Very few LEDs will go inside the house initially,” he says.
Yet the residential market is potentially a high volume one, notes Vipin Vothra, STMicroelectronics Applications Manager for Power Supply Applications, (Schaumburg, Ill.).
“It's starting to take effect,” he says. “We're approaching the point where one can buy an LED-based lamp and put it in the same fixture as a fluorescent tube, thus extending lifetime from about 2,000 hours to 10 years.” But the debate in various European countries about staying with present bulbs is actually between incandescent versus fluorescent lamps, not versus the LED. “Fluorescents have mercury content and a big environmental impact,” says Vothra. While the so-called total cost of ownership in a residential application doesn't enter, “right now the price of LEDs is not low enough to abolish incandescent bulbs.”
Indeed, it's not time yet. “There's not so much money to be made replacing CFLs and fluorescents, even though incandescents are going to be phased out,” says Peter Green. “CFLs are more likely to replace them than LEDs, especially in the short term. For higher power, it's a different situation because fluorescents don't really output enough light in those applications,” he adds. “LEDs might make more sense there. I'm not certain that street lighting, for instance, is necessarily one of them because LEDs need to compete against high-intensity-discharge lamps. But for offices or parking lots where you need, say, 50 to 100 W, LED panels make a lot of sense.”
And added to cost, there's the business of aesthetics. “Some people in Europe don't like compact fluorescents too much because of the look,” says Green. And they don't like the light — it's white, too cold. The CFLs don't dim properly or at all. LEDs would be more attractive, but very expensive. Some people will stay with incandescents if they can.” n
Sign says: Growth for high-brightness LEDs
The markets for high-power LEDs and IC drivers can be confusing. The world market for LED drivers for signage, outdoor displays, and traffic lights was put at over 100 million in 2008 by IMS Research (Wellingborough, England). This is one-seventh of the whole market which is put at $777 million. IMS figures the total market will rise to $1.8 billion by 2016, with double-digit growth from 2010-2014.
Strategies Unlimited (Mountain View, Calif.) says the market for high-brightness LEDs (HB-LEDs) exceeded $5 billion in 2008, despite a downturn in the automotive lighting and outdoor video screen segments. They see a market worth $14.9 billion by 2013.
Darnell Group (Corona, Calif.) says signage will continue to dominate in high-power applications, probably using (low-cost) colored LEDs. Vehicles, which encompasses mainly low-power applications, will grow slowly; white-LED headlights won't be cost-competitive against existing technologies until after 2013. Finally, low-power LEDs have unseated neon bulbs for signage but backlighting will eventually recede in importance as a market mover.
How to power up bright LEDs
Most system designers expect wide-area lighting to run off the 120-277 Vac power grid, whether the lighting is LED or not. The prototypical ideal: “The basic input circuitry for the LED driver includes a bridge rectifier, some storage capacitors, and an active power-factor correction stage that converts the ac input to a regulated (internal) high-voltage dc bus (e.g., 450 V) at the driver's input section,” says Peter Green, Senior Systems Engineer for the Lighting Group at International Rectifier. “The driver's back-end includes a step-down transformer stage that provides isolation, which in turn has a fairly large impact on the choice of the driver design itself and the specific topology. Regulation circuitry delivers a constant-current for driving the LEDs. The typical driving voltage is in the range of 24 to 48 V.”
“What we're seeing is a modular two-stage topology,” concurs ON Semiconductor Director of Applications, Computing and Consumer Group Bernie Weir. “You have an off-line converter that handles the power-factor correction and voltage or current regulation, followed by multiple dc/dc converters driving independent strings of LEDs.”
There's also one other issue: the supply's electrolytic capacitors. “Most of the high-frequency designs that we do require some kind of capacitors. They're a necessary evil in high-frequency switching power supplies,” says Vipin Vothra, STMicro Power Supply Applications Manager, (Schaumburg, Ill.). “Now LEDs would give you perhaps 50,000 hours of life, but electrolytic caps don't have that much life, perhaps 25,000 hours. It's a challenge to come up with a design that doesn't use electrolytic caps.”
The remaining concern is isolation. “Office lighting doesn't require much in the way of isolation,” says Fairchild Semiconductor, Europe marketing director Alfred Hesener, noting that users are not likely to come into physical contact with these systems. “With street lights, the designer must deal with humidity issues, ground faults, and various cabling issues, and so there we require double isolation. And with medical lighting, we also need some sort of safety isolation.”
The Darnell Group calculated the percentage of total market accounted for by various lighting technologies. Their figures cover both low-power and high-power applications but do not break out these areas separately. The notation “LED BLUs” signifies LED backlighting units. The “other” sector includes projection systems. The green sector refers to flash units in handsets and digital still cameras. The automotive-use sector (red) relates to mostly low-power applications, billboards (yellow) are high-power. The implication: Big-market areas, in terms of units, are low-current applications; in terms of dollars, high-current uses will dominate. Darnell also says white-LEDs will probably be too expensive for use in billboards. But despite the expense, they will probably find their way into automotive headlights after 2014.
A Stronger Light Source
Luminus Devices Inc. has been getting a lot of attention recently for its approach to fabricating high-intensity LEDs. “If you want to increase the amount of light you get from a source, you cannot simply put more die next to each other because luminous intensity is light output divided by surface area,” says Rick Zarr, technologist for National Semiconductor's PowerWise brand. “Luminus had an idea to increase the luminous intensity in panels so it is intense enough for streetlights and headlights, where before it was just on the edge of being good enough.” That idea might be described as redirecting photons in the LED substrate that would otherwise be lost as heat.
“Photonic lattice LED technology scales up LEDs into larger and larger surface emitters, larger than anyone had been able to do before,” says Don McDaniel, director of technical lighting solutions at Luminus. “At the same time it maintains high light-extraction efficiency, it's no longer a physical limit to the size of the LED. We targeted the projection display market, especially the market for projection TVs.”
“We replaced the arc lamps with LEDs,” he continues. “The lamp's color wheel was not particularly efficient. However, a projection system has tight constraints on the beam angle and aperture size. The LED technology was there, but to do it with LEDs required making arrays of red, green, and blue and projecting it first on the DLP chip where it was modulated with a picture and then to the screen.” To get around the brightness lost by the dead space between the LEDs, the company made LEDs with an aspect ratio of 16-by-9, thus making it suitable for large-panel TV systems.
“It was a big challenge because you need a lot of light,” says McDaniel. ”These LED systems would be driven at 30 A. So we time-interleaved red and green and blue. They were combined with dichroic filters, the three beams overlapped, and then projected on the DLP system. These were 30 A at 3 to 4 V (100 W peak power). So in addition to being able to scale up these LEDs, there had to be a lot of work on thermals so these things didn't burn themselves up.”
At the same time, the company developed a white-LED platform for general lighting. “Right now, there's a 15,000-lumen variable-zoom Vari-Light from Philips, and that uses seven of our RGB and W modules. An RGB LED product can put out a lot of saturated light, because you're not putting out white light and filtering it down. But if you tune them to color-match white, it's not an efficacious way to get white. So we put a white chip in there and now those things don't give up any performance compared to other lamp products. They're bright and saturated colors.”
McDaniel cites two areas where a large chip conveys an important advantage. One is for systems requiring 750/1,000 lumens where customers look for a single-emitter product. At the other end are streetlighting, high-bay lighting, and architectural lighting. One of their products can deliver 4,000 lumens in a single part (it uses four 9-mm2 devices in series). A 20,000-lumen application or high-bay or large streetlighting applications thus needs just five or six modules.
As another example, 200 LEDs providing 100 lumens per LED deliver 20,000 lumens to provide the equivalent of a 250-W metal halide lamp. Equivalently, just 20 of the company's SST-90 devices (a 1,000 lumen, 10-W part) are required; and just five of their CSM-360 devices (a 4,000 lumen array) would handle the job.