LED lighting teardowns: Five lighting designs that illuminate the future of lighting
Margery Conner -October 07, 2010
| View as PDF |
LED-based lighting is still far from a mainstream technology, and its designs are in flux. Consumers have not signaled the price-to-performance ratio for which they will open their wallets and homes, and businesses are reluctant to spend money in the current economic climate. Nevertheless, early SSL (solid-state-lighting) products are making their way onto store shelves and into inventory. These initial designs can indicate what direction SSL design will take, at least in its early stages.
This article describes the tear-down of five LED-lighting products to see how they perform and what components and design topologies they use. It’s OK to design in the abstract or to speculate about the most effective ways to use a brand-new technology. The designers and manufacturers of these products, however, have made many assumptions about component pricing and availability, manufacturing and distribution pricing, features that prospective customers will want, and the prices that the market will bear. This level of uncertainty is common when manufacturers are introducing technologies. Five years from now, you’ll know what the market wants and is willing to pay for in efficient lighting, but no one now has a clue, partially because of the existence of so many as-yet-undetermined variables. What will the price of energy do? Will the government set an energy policy and stick with it? Will global-energy needs affect investment in energy-efficient hardware?
Considering all the unknowns that face the introduction of SSL products, it’s amazing that companies and investors have the courage to invest. It falls to the engineers’ lot to make the best design they can with available components for the price point set by the marketers. It’s thus interesting and even exciting to peek inside these products and see the mind of the engineer and the mind of the marketer.
This tour of tear-downs begins with a 48-in. LED T8-sized tube light. You can’t call it a “replacement” T8 light because it doesn’t go into a fixture for fluorescent tubes. Fluorescent tube lighting requires a fixture with a ballast, the lighting industry’s term for a fixture-enclosed power supply for a light source. This arrangement works for technologies in which the light source, on average, wears out before the power-control circuitry. Fluorescent-lighting fixtures apply a voltage to a glass tube containing vaporized mercury. The excited mercury emits photons in the ultraviolet wavelength; these photons strike the phosphor coating on the inside of the tube, in turn emitting light in the visible wavelength. High-quality T8 fluorescent lights have efficacies of 100 lumens per watt and greater.
It is impractical to directly replace a fluorescent tube with an LED tube because the two lights have different power requirements. Most currently available LED tube lights contain their own ac/dc power supplies. In contrast, fluorescent-light fixtures contain the power-converting ballast.
Figure 1 shows an LED T8 tube light from Alpine Electronics. Alpine also provided a modified fluorescent-light fixture with no ballast. Each 18W tube emits 820 lumens, which works out to almost 46 lumens per watt, or just about half of what a high-quality fluorescent tube light emits. Figure 2 shows that each tube contains three rows of 96 LEDs. When the tube lights, the center row of LEDs are a warmer, yellowish white (Figure 3). The end cap routes ac power down to the tubes’ internal power supply (Figure 4). The aluminum back is a thin, rounded cover that touches the LED PCB (printed-circuit board) only at the edges and doesn’t provide much heat sinking. The PCB has no metal core; it looks like a garden-variety fiberglass board. The board itself is thus not a heat sink. Figure 5 shows the power supply. The part number on the three-terminal power regulator is missing, so it yields no part information. However, the part includes a lot of electrolytic capacitors (Figure 6). Two stapled-together PCBs make up the 4-foot-long light. Figure 7 shows the staples that connect the two boards, and Figure 8 shows the top view of the staples and the jumpers that route the power bus.
The specs for the light claim that the tube has a 50,000-hour lifetime; with all those electrolytic capacitors, though, this figure seems dubious. It’s possible to get a 50,000-hour lifetime with electrolytic capacitors, but I think the manufacturer may have just picked the general-lifetime number for LED components and used that figure for the whole light. The tube’s innards exemplify excellent manufacturing quality—much better than many other LED lights and CFLs (compact fluorescent lights).
The LEDs are in a matrix of 288 diodes in 18 parallel strings with 16 diodes in each string. Each LED has a drop of approximately 3.2V, totaling approximately 50V across the array. The specifications state that the light is 18W, so each string consumes about 1W, meaning that each diode uses approximately 0.0625W. This figure is a far cry from the HB (high-brightness) LEDs that you usually encounter in designing for LED lighting, which use approximately 0.5 to 1W.
The power supply apparently outputs 51V dc—that is, although it measures 51V dc at the load, it could be a constant current rather than a regulated-voltage power supply. Regardless, all of the diode strings are paralleled across the power-supply output—not an ideal load for an LED matrix because, as LEDs age, their current profiles change. In an array like the one in Figure 9, the string with the lowest resistance pulls the most current, heating the diodes and yielding differences in LED output. One of the most important characteristics of a light source is an even, consistent intensity and color; a matrix such as the one in this figure is asking for hot spots. A better choice would be a constant-current driver for each string (Figure 10).
Many power-management-IC vendors have developed their own LED-driver chips, such as Texas Instruments’ C2000 DSP-based IC driver, which lends itself well to applications with several strings. National Semiconductor, International Rectifier, Marvell, NXP, NEC, On Semiconductor, and several others also offer LED-driver chips, but the C2000 uses a DSP core with multiple PWM (pulse-width-modulated) outputs; one chip can provide a constant-current source for as many as seven LED strings.
You may be thinking that 18 strings would require 18 control loops. This requirement would be a problem for a cost-constrained tube light. Why not dispense with those wimpy 0.0625W LEDs and use some HB LEDs that will each put out 0.5W? Then you would need to use only 36 HB LEDs. This approach brings up a couple of other constraints, though. For example, 0.5W HB LEDs provide distinct, intense-point sources of light, and lighting designers and consumers alike don’t want that type of illumination. In addition, HB LEDs of this power have heat-dissipation issues: The 288 0.0625W LEDs more uniformly disperse heat and can use an inexpensive PCB. Using fewer high-power LEDs, however, requires a heat-dissipating substrate and may require the use of heat sinks, increasing the price of the tube light.
This design uses fewer expensive LEDs, power-management devices, and intense-point sources of light, but it has uneven current sourcing because LEDs age unevenly, which affects light quality and reliability. The challenge for an LED-based T8 LED-replacement light is to cost-effectively replace today’s $2 fluorescent light and maintain the quality of light. Prices for the Alpine T8 tube light range from $65 (1000) to $95 (one) per tube.
Alpine can find customers, even though its tube is competing against $2 fluorescents, because LEDs’ longer life can justify their higher costs in some difficult-to-reach applications when you consider replacement costs, including labor, downtime, and difficulty of access. Early adopters who value the color quality of LEDs and who simply like having the latest in technology also may be willing to pay the premium.
You can’t consider the tube light as a true replacement of a fluorescent light because of the modifications you must make to a fluorescent-light fixture. An example of a true replacement light for a 40W incandescent bulb is Home Depot’s recently introduced EcoSmart dimmable LED bulb (Reference 1). The 8.6W light sells for $20 and comes with a five-year warranty. The light provides warm, diffuse light; dims nicely; and produces no noticeable audio noise. It has a glass, dome-shaped outer shell that covers the LEDs and that doesn’t easily come apart, as you can see in Figure 11. The LEDs are not the usual intense light sources you see in other LED lights, such as the nondimmable, 7.5W bulb from TESS (Topco Energy Saving System) Corp that I disassembled in March (Reference 2 and Figure 12). That light uses seven LEDs that output 560 lumens, according to the specifications on the packaging. These large-surface-area LEDs provide a pleasant, diffuse light source, and only two of them output 429 lumens at 8.6W.
Figure 13 shows a close-up of the EcoSmart LEDs: I removed one to look for a manufacturer’s label or mark because officials at LSG (Lighting Science Group), the bulb’s designer, don’t want to divulge the company’s suppliers. I couldn’t find a manufacturer’s label, but there is an apparent part number, AM6L1, and the part looks like an LED array, meaning that the LED packages several tiny LED chips in one package and covers them with a single phosphor. It’s a good choice to use such a diffused light source because there is no pixilation.
To determine whose LEDs the light uses, I perused an LED catalog from Japanese LED manufacturer Citizen (Reference 3). It looks as though AM6L1 is similar to Citizen’s 6W CCL-L251 LED. In other words, the LSG derates the bulb’s two LEDs and runs them at less than 6W each—a smart, conservative design choice.
A rubbery compound encapsulates the electronics—a good choice for lighting technology because encapsulation cushions the electronics from all the vibrations inherent in a small, easily accessed light bulb (Figure 14). It’s not so nice for a tear-down, however. Nevertheless, the rubbery encapsulation material comes off fairly easily, exposing all of the drive electronics. The most promising IC—that is, the one with the most leads—is a 10-pin MSOP with a cryptic “SULB” marking on the top (Figure 15). A quick Google search reveals SULB as the “top mark” for National Semiconductor’s LM3445 TRIAC (triode-alternating-current)-switch-dimmable LED driver (Reference 4). I could see only one 50-μF Nichicon electrolytic capacitor, which operates at 105°C. The black capacitor-like components in the figure are inductors.
The electrolytic capacitor, which is partially visible in the right side of the figure, is a potential weak link, and this design uses a high-quality part to mitigate the risk of failure. The solder joint is the Achilles’ heel of LED-lighting reliability (Reference 5); using a highly integrated LM3445 LED driver decreases the number of solder joints. The metal baseplate of the LEDs mounts directly on the finned metal heat sink using a dab of thermal grease (figures 16 and 17).Compare this approach with the seven-LED light from TESS, in which the LEDs sit on a metal-core substrate and then on a flexible thermal interface before mounting on the heat sink. EcoSmart uses a simple approach that quickly removes the heat from the LEDs. Its overall design philosophy increases reliability by reducing the parts count and, thus, the associated solder points.
These two tear-downs have been of lights that comply with a lighting form factor. For my next project, I’ll take a look inside a new lighting engine from Cree, a manufacturer of LED components. You can think of the Cree LMR4 module as a light engine that can serve as a building block for a lighting luminaire (Figure 18 and Reference 6). You can quickly disassemble this light by removing several screws. A large, white-metal hood encloses the entire device (Figure 19). The back of the LED unit lies to the right of the penny, and the light cover is above the unit. The cover has a simple paper cone that serves as a reflector, and a diffuser sits between it and a clear-plastic light cover.
Cree’s TrueWhite color-mixing technology combines discrete white and red LEDs. Other approaches to creating a consistent warm white from LEDs rely on combining multiple colored emitters in one LED package or by tweaking the phosphor. The LMR4’s TrueWhite implementation has five white LEDs and three red LEDs. When you turn on the light and gradually turn up the power, the four primary white LEDs and the two primary red LEDs come on somewhat uniformly (Figure 20). As you continue to crank up the power, the secondary white one and then the secondary red one turn on. If you crank it all the way up, the secondary white LED comes fully on and rivals the primary white LEDs in brightness, whereas the secondary red LED never appears to turn on much at all.
Plus, if you leave the light on at a power level in which the fifth white LED initially is off, it comes on after a couple of minutes, which perhaps means that the LED lights’ color changes with temperature and that the other two are balancing LED-color changes for both temperature and power. Cree’s marketing videos refer to “active color management” when describing TrueWhite, so the power and thermal response must be the active part.
An eight-pin TI 9C L2903 dual differential comparator sits next to the LEDs. This chip perhaps compares the current through the main LEDs and turns on the secondary color-balancing LEDs when the current exceeds a maximum threshold (Reference 5). Figure 21 shows the substrate after it has popped off the baseplate, displaying the metal core. The marking on the front, “Berg MP A2,” looks like a Bergquist Thermal Clad metal-core substrate, which comprises a circuit layer over a dielectric layer over a base-metal layer. The substrate clamps onto an adhesive, thermally and electrically conductive layer to the baseplate of the power supply (Figure 22). The National Semiconductor LM3445 TRIAC-dimming SULB is probably the power-management IC, and the capacitors are 22-μF, 200V, 100°C Nichicon devices. The power and ground wires loop through a large ferrite bead to filter noise.
Cree specifies the power factor of the LMR4 module at greater than 0.80 for 120V ac/60 Hz, or more than 0.90 for 230V ac/50 Hz. Measuring with my trusty $20 Kill A Watt power meter from P3 International yields a power factor of 0.56. Granted, the Kill A Watt is not the most sophisticated power meter going, but 0.56 is a far cry from 0.8. Removing the Lutron dimmer from the circuit causes the module to operate directly off ac-line voltage, increasing the power factor to 0.91, so the TRIAC dimmer is evidently not fully on even when the switch indicates 100%.
The previous tear-downs were all production-LED lights currently for sale. The next example is a Helieon demo unit from Bridgelux (Figure 23). Bridgelux and Molex teamed up to design a socket-and-module combination for new installations (Figure 24).
The Helieon module includes a Bridgelux LED array mounted on an aluminum spreader, a lens, and a socket. The LED array can deliver 500 to 1500 lumens in 3000K warm white or 4100K neutral white, and the module’s optics shape the light path to deliver narrow, medium, or wide flood-beam angles. You can change the white-light unit’s color temperature and beam focus by swapping out the LED module. The socket attaches the LED module to the ceiling or the wall and delivers power to the fixture. The Helieon lacks the heat sinking necessary for a fully functional light; I suspect that this omission is the reason that it has a momentary on switch: to prevent evaluators from turning on and leaving on the evaluation kit, resulting in overheating. The Helieon design also lacks power-management circuitry, but it serves as another example of LED emitters for LED lighting. Bridgelux LEDs package a matrix of LED emitters into one LED device, an approach that’s similar to—but on a larger scale than—the one that Citizen LED takes in the EcoSmart bulb. The Bridgelux device provides as many as 1500 lumens in this package (Figure 25).
Bridgelux intends the demo unit for designers who want to evaluate the Helieon LED module-and-socket combination; the power-management circuitry is there only to enable the demonstration of the Helieon module. Nevertheless, it illustrates that the power management for LEDs is not trivial. The unit audibly ticks whenever you plug in the brick and hums when you hold down the momentary power switch. The dimmer circuit doesn’t use a TRIAC and dims the light by only approximately 50%, rather than virtually off. “Power management is the bane of my existence,” says Jason Posselt, vice president of sales for Bridgelux, commenting on these undesirable characteristics and likely voicing the thoughts of many other LED manufacturers.