LED illumination for building interiors and outdoor public spaces continues to gain serious traction, opening up new opportunities for electronic designers. For example, National Semiconductor’s LM3445 and NXP’s SSL2101 monolithic controllers for high-brightness (HB) white LED (WLED) building lighting accommodate legacy triac dimmers while providing wide-range dimming and power factor correction (PFC).
Meanwhile, Texas Instruments and Microchip both provide online resources for designers who want to develop custom microcontroller dimming applications. Every company that makes ICs for driving LEDs now offers chips that provide for pulse-width-modulation (PWM) control among their products.
Outside the building, things are changing fast. Mark McClear, director of business development, solid-state lighting, at Cree, observes that two years ago, it simply wouldn’t have been practical to use LEDs for street lighting. A year ago, the idea was barely into the trial stage. “A few cities would replace the fixtures on one block of one street,” he says. This year, it’s full steam ahead around the world (Fig. 1).
Last summer, the city council of Anchorage, Alaska authorized the replacement of 16,000 streetlight fixtures—a quarter of all the streetlight fixtures in the city—with LED luminaires. The city is investing $2.2 million in the plan. The new streetlights will consume half the energy used by the current fixtures, leading to a potential savings of $360,000 each year. According to Cree, the LED fixtures, based on Cree XLamp LEDs, typically last up to seven times longer than high-pressure sodium fixtures, allowing the city to better utilize maintenance resources.
One interesting aspect of Anchorage’s situation is that 85 days a year have fewer than eight hours of sunlight, which suggests that the payback in power costs increases closer to the Earth’s poles. However, it ignores the corollary that Anchorage must also have 85 days in a year with fewer than eight hours of darkness.
Down south, around latitude 47N, the city of San Jose, Calif., just announced plans to replace its 62,000 streetlights with new LED versions that will “use state-ofthe- art technology to vary their intensity and timing.” San Jose plans to convert 100 lights this spring and is seeking $20 million from the stimulus package to install 20,000 new lights. The intent is to have all of the city’s streetlights changed by 2022.
According to the San Jose Mercury News, the city’s street lighting bill is nearly $4 million a year. Last year, costs rose so high that the city turned off 900 streetlights to save money. Although the new lights will have a high acquisition cost, the city estimates it could recoup that within five years. That’s because, says the newspaper, the current lights need to be changed every few years, and often the city doesn’t know a light is out until it gets a call from a resident.
“Under the new system, the lights will last 10 to 15 years, and they will alert the city automatically when they are out,” the Merc reports.
GET THE YELLOW OUT
The LED lighting will replace sodiumvapor lights used in cities at the southern end of San Francisco Bay for the sake of the optical telescopes at Lick Observatory on nearby Mount Hamilton. (With just two spectral lines, the loom of sodium lighting can be dealt with using optical filters.) While the accommodation had the support of the astronomers, the sodium lights have been unpopular with the general citizenry. The complaint is that they’re too easily confused with traffic signals and distort the colors of cars and painted curbs.
That’s where the San Jose technology goes beyond merely swapping light fixtures. The plan may even give the astronomers control over the lighting when they want to take a picture. Well, maybe not initially, but all of the lights will be radio-controlled, and the first concession to the astronomers’ need for “dark skies” will be selective dimming, cutting them back, say, by 30% from 3 to 5 a.m., when Lick’s telescopes are most active (see “LED Lighting And Light Pollution").
In another Silicon Valley wrinkle, to help power the streetlights, the city plans to set up solar panels along light poles, on top of buildings, and on canopies over sidewalks. In time, the goal is to power these lights entirely from renewable sources.
More than doubling San Jose’s volume of replacements, the city of Los Angeles, Calif., announced in February plans to replace 140,000 existing streetlight fixtures in the city with LED units over the next five years. The city expects electricity savings of at least $48 million over seven years and a reduction in carbon emissions by approximately 40,500 tons a year.
McClear says this fresh activity is possible partly due to tighter binning and improvements in the Color Rendering Index—getting a broader range of photon energy levels out of the phosphors so that colors appear more natural. Yet even more important may be what he calls “managing photons,” making sure light generated inside the lamp exits the lamp and can be directed so that it illuminates what it was intended to illuminate, instead of radiating off into space. Managing photons translates into more lumens/watt and better energy efficiency.
FIXTURES, NOT BULBS
Surprisingly, McClear thinks we shouldn’t expect to see much direct replacement of old, Edison-based bulbs or fluorescent tubes with direct LED equivalents. The implication for dimming designs is that even though chip companies are now focusing on dimming room lighting using legacy triac dimmers, this doesn’t mean just changing out old bulbs and tubes. Instead, it means replacing existing can and overhead fixtures at the junction box in the ceiling while retaining the existing wiring and controls upstream from that.
The caveat reflects Cree’s roadmap for 2008 through 2012. In part, it looks at three types of LED replacements for screw-in and plug-in form-factor incandescent bulbs of the type normally used in recessed can fixtures and flood/spotlight track lighting: the A19 Edison-base “light bulb,” the PAR38 Edison-base “flood,” and the MR17 pin-socket compact flood (Fig. 2).
Continue to page 2
There’s a lot to be learned by comparing the performance of today’s LED replacements to their off-the-shelf incandescent counterparts (see the table). LEDs provide better efficiency in terms of lumens/watt, across the board, but with some falloff in actual light output as the rating goes up. Also, as the actual wattage rises, efficiency drops.
Higher Kelvin color temperatures would be more efficient, but the lower temperature is more like incandescent bulbs. Also, LED lifetimes slip below Energy Star’s L70 lifetime requirements at higher color temperatures (see “\\[\\[LED-Life-Standards-In-And-Out-Of-Luminaires20824|LED Life Standards In And Out Of Luminaries\\]\\]”).
This is only a problem for the direct replacement of certain common incandescent bulb types, though. The LEDs themselves keep improving. Yet design engineers should be thinking in terms of fixtures and heat and photon management, as well as light sources. It’s a new ball game.
MULTIPLE LED-DRIVING BASICS
Simple arrays of LEDs are driven in series because parallel connection can cause non-uniform sharing of current between LEDs, even when these LEDs are forward-voltage-matched, but it isn’t really that simple. In a big illumination array, a separate driver for each string isn’t desirable. The end result, then, is larger current-capacity drivers and multiple parallel strings of LEDs.
Still, you don’t want any string to go dark if one of its series LEDs fails open. Preventing that requires parallel Zener diodes or silicon-controlled rectifiers (SCRs) across each LED. Zeners should be sized so that VZ(min) is greater than VF(max), and each should be rated to dissipate IF times VZ. SCRs have a power advantage over Zeners in that each SCR need only dissipate IF times its threshold voltage.
Paralleling series strings brings us back to the current-sharing issue, though it gets easier if you assume the variation in VF across n series LEDs in the string averages out. In that case, it is possible to calculate a common value for ballast resistors for each string. Supertex applications engineering manager Alexander Mednik explains: If the nominal output voltage of the driver is VO, he says, and assuming a ±10% variation in VF across any LED string, and that the currents in each parallel string are maintained within ±20%, then it’s clear that n × VF × 1.1 + VBALLAST × 0.8 must equal VO. From that, VBALLAST must equal n × VF × 0.5, and RBALLAST equals VBALLAST/ IO, where IO is the constant current output of the driver.
That’s what is needed if using individual LEDs. At a higher cost and with less flexibility in terms of form factor, but less nonrecurring engineering expense, the LED manufacturer can match its devices and connect them in series/parallel cross-connect grids, as in Philips’ Luxeon Flood products.
PFC requirements come from IEC 61000-3-2, the European standard that covers electrical safety and electromagnetic interference. (For more on international standards, see “Conforming With Worldwide Safety And EMC/EMI Standards”)
TRIAC DIMMING BASICS
While dc LED dimming is simply a matter of controlling the PWM switching of a constant current, the task gets more challenging when dealing with an existing building that already has old-fashioned triac dimmers. When controlling ordinary light bulbs, those dimmers automatically shut off the current for every zero crossing of the line voltage and turned the light circuit back on when the line voltage climbed back up to a certain level. This is set by the dimmer pot, which is in series with a capacitor. The node between the pot and the capacitor connects to the gate of the triac.
The RC time constant of the pair determines how long it takes from the zerovoltage crossing to the time the voltage on the capacitor reaches a level that will fire the triac. Once the triac starts to conduct, it continues to conduct through the rest of the ac half-cycle until the ac voltage crosses through zero. Theoretically, the range of hold-off times for any given half-cycle ranges from 0° to 90° (or 180° to 270°) of the ac waveform.
Those old dimmers generally also include some inductance and capacitance to slow the rate of turn-on and turn-off during each half-cycle. This reduces RF noise and is gentler on the incandescent filament, rather than abruptly turning it on and off 120 times per second. That’s the basic idea, though there are some complications.
One is holding current (see STMicroelectronics’ AN302 “Thyristors and TRIACs: Holding Current” applications note at www.st.com/stonline/products/literature/an/3563.pdf ). In an electromechanical relay, it’s necessary to have a minimum current circulating in its coil to keep it turned on. The same goes for a triac. This is called the hypostatic or holding current (IH). Whenever the triac current falls below IH, the triac is blocked and requires another gate pulse before it can turn on again.
The STMicro app note discusses IH in incandescent lamps and motor control, but its significance in LED control is this: That legacy triac dimmer was designed for a certain IH, drawn by a resistive incandescent light bulb. In other words, the LED dimmer must provide a way to emulate the resistance of the incandescent bulb. It must also provide complete 0% to 100% dimming and PFC.
To get a sense of all the considerations involved in dimming LEDs in general lighting fixtures with legacy triac dimmers, try analyzing a specific design. For instance, National’s LM3445 buck controller uses a proprietary constant off-time method to maintain constant current through the string of LEDs (Fig. 3). With constant offtime, the on-time becomes the only variable, and controlling on-time is achieved by varying the switching frequency. In the National chip, on-time is determined by the sensed inductor current through a resistor to a voltage reference at a comparator.
Continue to page 3
Figure 3, an application circuit adapted from the National datasheet, shows a Zener bridge (BR1) across the output of the old triac controller. A “valley-fill” circuit follows the bridge. The network of diodes and capacitors between D3 and C10 constitutes one possible configuration for such a circuit. Valley fill allows the buck regulator in the chip to draw power even while the triac is in cutoff. This allows a smaller capacitance for C10, adds passive PFC, and eliminates some 120-Hz flicker.
In operation, the chip senses the rectified line voltage at pin 10 (BLDR). The “bleeder” terminology means this is the node at which the circuit emulates the resistance of an incandescent bulb. That is, resistor R5 maintains IH.
Also at pin 10, the schematic shows an external series-pass regulator comprising R2, D1, and Q1. D1 is typically a 15-V Zener. It forces Q1 to “stand off ” most of the rectified line voltage. Note that there’s no capacitance on Q1’s source. This means the voltage on the sensing pin can rise and fall with the rectified line voltage as the line voltage drops below D1’s Zener voltage. To provide a steady voltage on the main power pin, a diode-capacitor network (D2, C5) keeps that node up while the voltage on the bleeder pin goes low.
FIRING ANGLE TO PWM
National’s design provides a 10% to 100% dimming range, based on triac-dimmer firing angles between 45° and 135° of the ac line waveform. Inside the chip, a ramp generator produces a 5.85-kHz sawtooth wave with a level between 1 and 3 V. The sawtooth is placed on Pin 1 (ASNS), where it’s filtered by R1 and C3 and applied to Pin 2 (FLTR1). Back inside the chip, the signal on FLTR1 is compared to the output of a ramp signal.
The ramp comparator’s output is a series of pulses whose on-time is inversely proportional to the average voltage level at FLTR1. Because the FLTR1 signal can vary between 0 and 4 V (the limits of the sawtooth on the ASNS pin), and the ramp only swings between 1 and 3 V, the ramp comparator’s output will be on whenever the FLTR1 voltage is below 1 V and off when it is above 3 V.
The ramp comparator’s output drives a common-source N-channel MOSFET through a Schmitt trigger. (The comparator output also appears on Pin 3, (DIM), which is used in a scheme to gang multiple LM3445 chips.) The MOSFET inverts the ramp comparator’s output, and its output voltage directly controls the peak current that will be delivered by Q2 during its on-time.
The MOSFET’s drain voltage is proportional to the duty cycle of the triac dimmer. The amplitude of the ramp causes this proportionality to “hard limit” for duty cycles above 75% and below 25%. To reduce ripple in this signal, the chip provides for a second, one-pole, 10-Hz low-pass filter stage that comprises C4 on pin 5 (FLTR2) and an internal 370-kO resistor.
Pin 4 (COFF) is used with C11 for setting PWM off time. Pin 5 (FLTR2) is the input to the second filter for the LED current control voltage. Capacitor C4 filters the PWM dimming signal to supply that dc voltage. Finally, maximum LED current is set via pin 7 (ISNS) by means of R3 at the source of the switching MOSFET Q2.
That describes one way to do the job monolithically. If you’re looking at a special case, you may want to tackle the job with a custom design. Companies that can help you include Texas Instruments (search www.ti.com for “dimming”) and Microchip.