A common element among new developments in discrete semiconductor power devices involves PFC, or power factor correction (see "Power Factor Basics," below). About 30% of the world's markets—including Europe, China, Japan, and several states in India—now require PFC in switching supplies for computing gear. Though PFC hasn't been mandated yet in the U.S., the IEEE is at work drafting standards.
Stephen Oliver, who manages International Rectifier's marketing to OEMs that make ac-dc front ends, believes that PFC has broader benefits than regulatory compliance. "If you put PFC on your front end, you automatically have a universal-input supply," he says.
"One design can run from mains voltages from 80 to 264 V," he explains. "Moreover, PFC up front means you have a very cleanly controlled bus voltage, usually at 380 or 400 V. Downstream from that, you can use lower-voltage MOSFETs or diodes, which means that with the same size chips, you can trade off performance versus cost. For example, instead of using maybe an 800-V MOSFET, you can use a 500- or 600-V device."
Without PFC, switching supplies feed harmonics back onto the powerlines. The problem arises when systems draw current from the energy storage capacitors in the ac-dc front-end supply in narrow, high-amplitude pulses.
These pulses contain harmonics that interfere with other equipment on the line and reduce the maximum power that can be drawn. In addition, the distorted line voltage causes capacitor overheating, dielectric stress, and overvoltages in insulation.
The problem with power factors that are significantly less than unity involves the rms current content of the harmonics created by the switching supply. The rms current is:
where n is the order of the respective harmonic currents. Eliminating the harmonics by making the power-supply input current approach a sinusoid decreases the rms input current and brings the circuit power factor closer to unity. Without PFC, a typical switched-mode power supply (SMPS) exhibits a power factor between 0.6 and 0.7.
DISCRETES' ROLE IN PFC
Historically, switched-mode power supplies with PFC have used a boost converter topology that contains two semiconductor components: a power switch (MOSFET or IGBT) and a boost diode (Fig. 1). This is where semiconductor manufacturers are concentrating their product differentiation efforts.
Diodes used in boost converters are considered "soft" or "snappy" depending on their reverse-recovery characteristics. Consider what happens in a boost converter operated in continuous current mode (see "PFC Operational Modes," p. 60). The boost diode and the switching device operate in the hard-switched mode, which involves a short recovery period in which the diode conducts briefly in the reverse direction (see "Reverse Recovery," p. 67).
The diode's reverse-recovery characteristics increase the switching device's turn-on losses and generate EMI. If the reverse current characteristic is snappy—that is, if it shuts off too abruptly—there are voltage spikes and ringing. To deal with this source of EMI, designers have either slowed down the switch turn-on di/dt with a softer-switching diode and/or added snubber circuits. But a slower switch turn-on rate increases turn-on loss, and efficient snubbers are tough to design.
SOFT AND SNAPPY DIODES
To begin to get a handle on reverse-recovery characteristics, take a look at a cross section of a power diode (Fig. 2). The diode incorporates an n doped region between the n+ cathode and the p+ anode. This n region determines the diode's blocking-voltage characteristics and absorbs the depletion layer that's created when the PN junction is reverse-biased. It's made as thin as possible to minimize its effect on forward voltage drop (VF).
The diode can't switch instantaneously from conduction to blocking or vice versa. Follow the negative current, starting with turn-on. At that time, the n region is more resistive than the n+. But as soon as the diode starts conducting, injection of minority carriers into the n region lowers its resistance.
Still, just at turn-on, the current may increase much more quickly than the minority carriers injected from the junction diffuse into it. This causes a temporarily higher voltage across the diode than that which will exist in the steady state. Such overshoot can damage a downstream switching device if it exceeds that device's breakdown rating.
Then at turn-off, the n region has a different effect on the transition from the conduction state. This is referred to as reverse recovery time (tRR in the figure in "Reverse Recovery"). When the diode tries to turn off in response to a reversal of the voltage applied across it, the excess carriers stored in this region must first be removed. The process involves the flow of a large reverse current, followed by recombination and sweep-out.
Afterward, the depletion layer acquires a substantial amount of space charge from the reverse-bias voltage and expands into the drift region, enabling the diode to block the reverse voltage. This part of the reverse-recovery time is designated tA.
Once the excess carriers are dealt with, the junction becomes reverse-biased and tB starts. The diode voltage now decreases, and if the slope of dIREC/dt is large, the peak reverse voltage also is high and VR may ring, or the diode may fail due to excess voltage. The induced overvoltage may damage the diode or the switching element if the breakdown rating is also exceeded.
What semiconductor makers call the softness of the diode—the ratio tB/tA—is determined by the quantity of charge left in the n region after the full spread of the depletion zone. And, that's largely where the product differentiation occurs.
For example, Fairchild's Stealth diodes use a new technology to combine the fast-recovery characteristics of Hyperfast technology with soft-recovery characteristics. They provide tB/tA values of greater than 1.2 and have a reverse-recovery time of less than 40 ns. Also, ON Semiconductor's soft 250-V Schottky (the first Schottky rated for more than 200 V) can handle 40-A and 2- and 4-A 200-V Schottkys, all with much softer characteristics than ultrafast rectifiers.
OTHER POWER-SEMI MARKET DRIVERS
Another evolving trend in ac-dc supplies is a shift from diodes to synchronous rectification on the secondary side. Today's typical 120-W laptop adapter parallels a trio of 20-A Schottky rectifiers to reduce conduction losses, but this approach is losing ground, according to some semi makers.
Oliver says that to evaluate the practicality of synchronous rectification, IR took one of those designs and replaced the Schottkys with a single MOSFET. The result was the same efficiency with a 60% reduction in bill-of-materials (BOM) cost. IR's customers responded by adopting synchronous, although they're using dual MOSFETs to achieve not the same but better efficiency than nonsynchronous, yet with only a 40% BOM cost reduction.
In response, IR introduced synchronous-rectification MOSFETs rated at 75 and 100 V. Both are designed for something around a 12-V output. The 100-V devices are used in applications like laptops that have a flyback topology on the front end. The 75-V FETs are employed in half-bridge or full-bridge topologies in server systems.
But not every company is hearing the same thing from its customers. "The reason you need different kinds of diodes has to do with switching performance," says Jeff Olsen, ON Semiconductor's director of power-supply market development. "In continuous mode, you're more concerned with switching losses than conduction losses. In the other modes, the current is actually at zero when you switch the device. So you're looking for low forward-voltage drop. In continuous mode, the average current in the conductor is still high, so switching losses affect efficiency."
He notes that for continuous-mode operations, ON Semiconductor recently released a series of 4- and 8-A ultrafasts that are rated at less than the traditional 600 V (they're rated at 450 and 500 V, the latter guaranteed to 520 V) and have a lower forward drop than 600-V ultrafasts. "Customers told us they didn't really need 600 V," he says. "They wanted the lower conduction losses."
Packaging innovations are providing another source of product differentiation. Until recently, designers were happy with through-hole semiconductor device packages. It made sense because many of the other components on boards, such as transformers and capacitors, had to be through-hole mounted.
However, through-hole starts to limit power density because through-hole packages aren't as efficient in heat transfer as newer designs. Only five years ago, a power density of 5 W/in.3 was state of the art for an ac-dc supply. Now, densities typically hover at the 20-W/in.3 level, with a handful of prototype supplies achieving 25 W/in.3
New semiconductor packages include IR's DirectFET, a proprietary SO-8-size surface-mount package in which a copper "can" forms the drain connection from the other side of the die from the board (Fig. 3). ON Semiconductor and Diodes Inc. announced products in a different SO-8 flat-lead package—it has the conventional tab on the bottom and a clip on top.
Supply power density is a critical element in products ranging from blade servers to laptops. Blade-server supplies, which are about the size of a carton of cigarettes and put out around 2 kW, often take much design effort.
Even more challenging are laptop adapters, along with external supplies for LCD monitors and PCs that incorporate a processor and peripherals into the display unit. Adapters are essentially small, closed plastic boxes with no provision for fans or external heatsinks. Yet some laptop adapters now put out as much as 150 W. Adapters for PCs with displays incorporating the electronics and drives may double that number.
It gets really interesting when the trend intersects newly mandated energy-efficiency standards such as Blue Angel, Energy Star, and One-Watt Standby. Blue Angel and Energy Star require certain efficiencies when a product is running. The One-Watt Standby rule adds that when the product goes to sleep, it can't consume more than a watt. The power supply may only dissipate half a watt in standby mode.
For cell phones, this is no problem. Their efficiency is great over the whole load range. But a 150-W laptop adapter, which generally achieves its greatest efficiency (circa 85% to 92%) at three-quarters of full load, must still operate at no less than 50% efficiency down at 1 W.
That's a tough challenge. It requires some kind of intelligence in the power supply to turn off everything unessential during standby. Since no PFC limitations exist for products operating in standby mode, it turns off the PFC stage. Then on the secondary side, it can start to pulse-skip so that the device only wakes up every few milliseconds to see if the load needs more power. Ultimately, the semi industry must come up with new ideas and technologies to become more efficient.