New lamps are more efficient, but what about the electronics? Designers need to learn a few new tricks about driving LEDs and CFLs.
It isn’t just more-efficient lighting that is driving down energy bills. Compact fluorescent lamps (CFLs) and LEDs wouldn’t be found in many light sockets today if it weren’t for electronics able to economically drive these bulbs. And there are still lessons being learned about how to get the cost out of illumination systems.
Take fluorescent lighting, for example. The old-time ballasts powering fluorescent tubes were little more than transformers that energized the tube gas by applying a high voltage to heat the filaments. The ballast also serves as a current limiter when the lamp is on. The problem with old-style ballasts was one of both bulk and inefficiency.
CFLs only took off with the advent of electronic-ballast circuits that were both economical and compact enough to fit in the base of a lamp holder. Today’s CFL drivers are basically switch-mode power-supply circuits that include power-factor correction and protection against such conditions as shorts and openbulb filaments. These use switching circuitry instead of transformers to generate the high voltages (about 500 V) that initially energize fluorescent lights and the lower voltages (about 200 V) that sustain lamp operation. Fluorescent bulbs are most efficient when operating at the 20 kHz and higher frequencies that electronic switchers generate. Operation at higher frequencies also lets ballast components be physically smaller and makes for a more-compact package.
It isn’t just CFLs that have electronic ballasts. Linear fluorescents have gone electronic as well. As of 2006, DOE regulations dictated what are called ballast-efficacy ratings — basically a measure of energy efficiency. The ratings are such that transformer-style ballasts aren’t efficient enough for many of the most common fluorescents used in shop and factory lighting. In the same year, the EU banned all magnetic ballasts, forcing a move to electronic ballasts for fluorescent bulbs sold there.
Ballasts may be going electronic but not all of them have the same level of integration. Some manufacturers still design their own. “Cost has been a barrier to the use of singlechip ballasts,” says Fairchild Semiconductor Director of Marketing Claudia Innes. But there are subtleties to driving a fluorescent bulb that can be a learning process for some manufacturers. “Compared to powering an incandescent bulb, you have to account for more conditions and provide safety features for different kinds of failures,” she says. “A lot of designers don’t know how to do this. So the electronic-ballast chips build in a lot of failure protection to make sure a problem doesn’t damage the entire ballast.”
For example, lamp impedance changes with age. This can move the oscillation frequency away from its most-efficient operation point. To check for faults, ballast circuits must watch the crest factor (the ratio of peak to rms current). A crest factor exceeding four generally indicates the lamp is at its end of life.
Dimming is another issue. Ballast circuits usually adjust a voltage-controlled oscillator to dim CFLs, but “If you put a dimmable CFL next to a dimmed incandescent, you’ll notice they don’t dim to the same extent and they don’t dim the same way. From a design point of view, there are several more things you have to account for,” says Innes.
A typical electronic ballast first rectifies ac, then converts the resulting dc to a signal in the range of 50 kHz through a MOSFET or IGBT switch. This switching action can generate harmonics in the current and voltage. These distortions cause radiated interference and put a damper on efficiency. So electronic ballasts generally incorporate power-factor-correction (PFC) circuits to compensate. PFC chips basically keep the switch on time at a fixed relationship to the input line voltage so the load appears resistive to the ac line.
A ballast-control chip then handles preheating and ignition, watches for conditions that indicate an open filament, and implements zero-voltage switching of the final high-voltage stage. The high-voltage stage that actually connects to the lamp is usually a half-bridge powering either MOSFETs or IGBTs.
Whether to use one or several chips to implement these functions often depends on how manufacturers view trade-offs between the cost of components versus the total system. “Every connection is a point of failure and every component picked-and-placed has a cost. Still, some people design their own,” says Innes.
Spotlight on LEDs
Two years ago, there was no such thing as an LED streetlight. That all changed in 2006 with the advent of superbright LEDs. “Now it takes under 100 LEDs to generate the equivalent to a high-pressure sodium light,” says Cree Inc. Director of Business Development Mark McClear.
Key to this turn of events was CREE’s development of its EZBright LED power chip. Since then, other manufacturers have brought out versions of high-output LEDs. But CREE has come up with a new LED topology that it says is more efficient than earlier chips by a factor of two and figures it is perhaps a year ahead of its closest competitors.
Current research by LED makers focuses on bettering power efficiency and lumens/ dollar spent. Today these figures are about 100 lumens/W and 40 lumens/dollar. Expectations are that the year 2010 will see 150 lumens/W with costs down substantially. “Every time we improve efficiency, it makes possible another wave of new applications,” says McClear.
It turns out that the benefits of LEDs are not limited to efficiency. “It costs a municipality as much to change a bulb as to buy a new lamp. Because LEDs last two to five times longer than incumbent bulbs, they avoid a lot of maintenance costs,” says McClear. And there is a sleeper benefit to using them for outside lighting: “When you replace a yellow sodium light with LEDs, people think you have cleaned up the place,” McClear says. “That’s because the eye has more visual acuity in the LED’s light range. Surveillance cameras work better with LED light and people actually feel safer in parking decks illuminated with LEDs.”
Several lamp makers now make outdoor fixtures incorporating LEDs. One of these, Beta Lighting in Sturtevant, Wis., employs CREE LEDs configured as light bars, each containing 20 LEDs. Beta adds light bars to get fixtures of a specific output. The firm says its design is protected by over 20 patents.
“Our biggest issue was thermal management. Once we solved that, we optimized the optical design to get the most out of the LED,” says Beta Sales Director Kevin Orth.
Though LED-powered streetlamps are more expensive than the conventional lights they replace, they cost less to own, Orth says.
How to do Drivers
LEDs may be the wave of the future, but there doesn’t seem to be a consensus about how best to configure their source of power. “So far, there is no set topology for driving LEDs,” says National Semiconductor Corp. Senior Application Engineer Chris Richardson. “If you want to drive 100 LEDs to get the maximum amount of light, there are many ways to do it — so many, in fact, that a lot of people get intimidated by the task.”
There are three general approaches to driving banks of LEDs today, Richardson says. The first, and most efficient, is to simply drive the LEDs in series from a dc supply. The problem with this approach is that it can involve voltages high enough to be classified as hazardous by UL. The high-voltage components involved can be expensive. “It is okay if you really understand all the safety codes and are willing to double insulate and isolate. But it gets ugly in terms of safety testing and I don’t recommend it,” says Richardson.
A second slightly different approach also uses a singlestage power supply but incorporates galvanic isolation, usually in the form of a transformer. This gets around some of the safety issues and has the advantage of availability as commercial off-the-shelf units. The problem is that this approach is only practical for driving strings of about eight LEDs at most, says Richardson. “You might produce at most 1 A this way,” he explains. “It is expensive because you pay a premium for the engineering that goes into the supply.”
The third way is the most widely used. It employs a commercial ac/dc converter that produces an output below 60 V, thus staying below hazardous voltages. The output goes to multiple dc/dc converters, each driving an LED string. Besides avoiding dangerous voltage levels, the approach guarantees some of the LEDs stay lit in the event one fails open.
“You need more engineering time to design this sort of circuit, but the result is the most flexible and reliable of the three possibilities,” says Richardson. By eliminating the need to work at high voltages, it may also be the least difficult to realize for most engineering staffs. “I haven’t met many power-supply engineers well versed in both high-voltage ac and low-voltage dc,” says Richardson.