With “green-power” and energy -efficiency issues more important than ever, advancing power supply technologies are the brightest spot on the economic horizon. Energy sources continue to improve as a result of more attractive government stimulus plans, more exacting efficiency standardds and regulations, in the last reckoning, more ingenious power architectures.
Improved power supply technologies will be crucial to national energy strategies. The International Energy Agency (IEA), an intergovernmental energy policy advisory organization, gives one clue why: It figures electricity consumption from residential home computers, iPods, and other consumer electronics could be cut by more than half through the use of state-of-the-art power saving methods. This will ideally slow the growth in energy consumption to less than 1% per annum through 2030. Such savings represent a cut in consumer energy bills by over USD 130 billion in 2030 and the avoidance of 260 GW in additional power generation capacity — more than the current electrical generating capacity of Japan.
These realities are forcing supply designers to figure out ways of maintaining high system efficiency from full load down to the milliwatt level. Heat dissipation and thermal costs are the enemy, often estimated at $1 per year for every wasted watt.
Circuit-wise, system designs are migrating to more quasi-resonant and resonant systems, more power-factor correction, lower-loss synchronous rectification components and topologies on the secondary side, and greater use of digital power techniques. And IC manufacturers are coming to more finely match their controllers to the capabilities of their power MOSFETs on a single chip. Power management issues and smart power extend into the lighting arena, no surprise because lighting reportedly consumes more than 20% of the nation's power draw, with 11% of that in the home. Moreover, light-emitting diodes are being called on to match up against cold-cathode fluorescent lamps (CCFL) for backlighting in portables and the like; large-panel LEDs are fighting for recognition against the compact fluorescent lamp (CFL) in the home and commercial sectors, and high-brightness LEDs (HB-LEDs) compete against sodium-based technologies in industrial settings. With it all, the Power Sources Manufacturers Association (www.psma.com) has developed an extensive online Energy Efficiency Database to update designers on trends and developments in major areas.
Fig 1 - One example of an MCU-based power supply comes from Texas Instruments Inc., wherein a TI digital signal controller handles six interleaved synchronous buck converters. Interleaving individual converters this way allows some of them to be turned off if loads are light, thus boosting efficiency. The approach also reduces the size of required magnetic components to keep costs down.
There are generally three separate operating modes in which designers define power consumption efficiency. These can be broadly defined as shutdown, standby, and full-load operation. There's a separate category for power-factor correction (PFC), though this is more closely tied to cutting the reactive power the electric grid must supply than with improving supply efficiency. Typical techniques now employed to improve the light-load efficiency of today's switching supplies are to shut off power-factor correction circuitry, and use frequency scaling and burst-mode operation.
A rise in power system efficiency of just 1% brings a significant trickle-down effect by cutting the thermal load and reducing chip and board costs and size. “But it's not all about improving the ICs,” says Laurent Jenck, director of system applications engineering and marketing at ON Semiconductor (Phoenix, Az.). “We try to improve the parameters of the power supplies — a combination of factors. You can certainly improve components —use MOSFETs with lower RDS-on, or Schottky rectifiers with shorter recovery times. You can also secure changes on the architectural side. Power supplies traditionally were (80% of them) using flyback. However, that topology has a limitation regarding efficiency. We've moved to soft-switching topologies such as half-bridge resonant (see for example the company's new NCP1901, which contains power factor correction stage and a half-bridge resonant controller). In the past that was reserved for products from Tyco or Lucent, which were in the 400-500 W range. This topology is now used at 90 W and even lower. We've got products at 20 W. No one would have followed that five years ago.”
On a larger scale — as for server systems employing so-called virtualization — power management involves multiphase converter shedding techniques. Shedding schemes dynamically shut down various servers and let the IT specialist move a Web site between different computers as the demand (e.g., Web hits) changes on a given server.
Half-bridge quasi-resonant and resonant DC/DC converters, generally useful to a few hundred watts, are often applied to flat-panel TVs and computer supplies. “These converters use zero-voltage switching for high efficiency and very low EMI,” says Van Niemela, technical marketing manager for industrial power conversion at Fairchild Semiconductor (San Jose, Calif.). “It's more complex than the standard forward converter or flyback, but these days with the push towards higher efficiency, power supply makers are adopting more complex topologies (and usually a few more components) where it makes sense to drive efficiency higher,” he says. He cites the company's LLC resonant converter as a substantial improvement over fixed-frequency PWM and traditional half-bridge converters. One example is the company's just-released FAN7621. Niemela cites a FAN9612 controller running a PFC stage followed by the FAN7621 for an LLC resonant converter as a superior off-line power supply design.
Beyond these issues, there's topological innovation across the system. “In the past, designers concentrated on primary-side power switches to cut losses,” said Eric Persson, executive director at International Rectifier (El Segundo, Calif.), noting primary-side controller issues can involve expensive tradeoffs of isolation, speed, and cost.
In many applications today, though, the issue hinges on maximizing efficiency at output operating voltages down in the 1.x-V range, and getting away from traditional diode-based designs. To that end, the company's IR1167 and IR1168 SmartRectifier drivers, for advanced (secondary-side) synchronous rectification in resonant half-bridges, are designed with attention to turn-on and turn-off times of zero-current switching transitions in MOSFET power stages. The combination can improve overall system efficiency by 1.5% in AC/DC systems for LCD TVs, reduce operating temperatures by up to 25°C, and cut board space requirements by two thirds.
Whither digital power?
In addition, IR's second generation of compact SupIRBuck regulators, for point-of-load applications, provides an ultra-optimized synchronous buck PWM controller, control FET, and sync FET in a single package. The regulator boasts 96% efficiency for server, storage, and netcom applications. Integrated chips that include better matched controller-driver-output stages often extend to all-in-one reference designs. These products are exemplified in such expanding lines as Fairchild's Green Power Switches and ON Semiconductor's GreenPoint reference designs.
Viewed by most practitioners as a potentially valuable tool for enhancing power supply performance, digital power techniques are slowly making their way into the designer's arsenal.
“People are moving to digital power to meet efficiency standards in a low-cost flexible way,” says Bill Hutchings, power product marketing manager for high performance microcontrollers at Microchip Technology (Chandler, Az.). Microchip defines “digital power” in four segments that extend from on-off functions in power management to full control using a digital feedback loop for DC/DC conversion.
“We're in the third phase of the evolution,” adds Jeff Shepard, director of the Darnell Group (Corona, Calif.). “The first phase was focused on components. The second focused on optimized devices as substitutes for analog components. The present phase centers on adaptive control and nonlinear functions. When you can adjust poles and zeroes on the fly, you can operate in regions that would not be possible in the analog world,” he says. “And digital is hitting price points that analog won't be able to match.”
But how are new chips directly contributing to power supply efficiency? “Say you want to make a 94% efficiency supply,” says Bill Hutchings. “To do that, I need to implement a zero-voltage transition, zero-current switching scheme. To get that switching scheme, I need a controller that has both a high-resolution PWM for duty cycle and high resolution in frequency to implement full-bridge technology. In the end, it goes back to the requirements of the controller and relates to the efficiency of the supply in a cost-competitive way. We're speaking of an adaptive control technique to get half-a-percent more efficiency here, half-a-percent more there. And getting efficiency at a certain practical cost level.” In that context, Microchip's second-generation dsPIC33 Digital Signal Controller (i.e., providing certain functions of a DSP and MCU) doubles the speed of its predecessor, provides more options for its PWM and Flash memory, and consumes half the power.
The range of controllers encompasses all four segments of digital-power. Given the pros and cons, Texas Instruments (Dallas) has made what's perhaps the biggest statement on the future of digital power. Just a year ago, TI was content to play at levels 2 and 3 with non-DSP architectures. But its newly released Piccolo line of microcontrollers can work in level-4 applications suitable for, say, resonant-mode converters.
Another new arrival suited to level-4 is Analog Devices' (Norwood, Mass.) new ADP1043 controller IC. It boasts a full digital control loop for both fixed-frequency and resonant-mode applications. Also, National Semiconductor Inc. (Santa Clara, Calif.) just released its LM5553 energy management unit, an adaptive-voltage scaling (AVS) product, which complies with the System Power Management Interface, SPMI.
Other strategies, widely applied across industry, selectively apply digital circuitry for control. One example is Fairchild's new FAN9612 interleaved boundary conduction PFC controller. The control topology boosts a supply's robustness (start-up/shut-down, fault recovery, multiple recovery conditions) using internal (digital) multipliers.
Beyond power ICs for basic DC/DC conversion, designers are heavily focused on carefully conserving power in the subsystems most users can visually identify with — displays and lighting. Most notable are newer devices that self-monitor ambient light to save battery power. “We look at the usage model, human-condition cases,” said Joy Wrigley, marketing manager for portable power solutions at Analog Devices. The company's ambient light products, for example, let systems sense the light environment and throttle back on backlighting for keypads. ADP5520 and ADP5501 backlight drivers, launched within the last six months, detect ambient light levels for up to six white LEDs and accordingly adjust LED current (up to 128 levels) for the right amount of backlighting. The ADP8860 is an LED charge pump with automatic phototransistor control stage (i.e., processor-free monitoring of the control stage).
At the other end of the usage spectrum, various controllers for compact fluorescent lamps (CFLs), and especially LED lighting (though largely not yet economical) for the home are well along on the learning curve. Says Rick Zarr, National Semiconductor's Technologist for PowerWise Products, “CFLs were a step forward, but they have their own problems. It's mainly the environmental impact (mercury). You can't throw them in the trash.”
There are also wavelength issues for a product that many see as an interim solution until LEDs mature. But nobody thinks the interim period will end any time soon.
“In Asia they prefer white light, a cultural thing maybe; the western world prefers amber,” says Zarr. “The CFL manufacturers have tried to adjust to that, tried to make their bulbs warm so they don't look washed out.” And CFLs can smoke during multiple transients on loss of mains power. Beyond that, “The CFL isn't easily dimmable,” says Zarr. There are two problems from moving from CFLs to LEDs, or incandescents to LEDs, he notes. The heat from LEDs must be conducted away, creating a thermal issue. Second is the issue of dimming and implementing LEDs using existing infrastructure (e.g., backward compatibility with existing wall dimmers).
Triacs are the typical means of dimming incandescents, and they can adjust light levels smoothly and gradually. Dimmable CFL circuitry, on the other hand, generally provides only coarse control of light output. When it comes to LEDs, National's LM3445 triac dim decoder driver chip, a constant-current controller that the company says is the first dimmable off-line chip for zero-to-100% dimming, reads the triac signature and automatically adjusts LED currents up to 1 A.
How long will it take LEDs to mature? Perhaps the better part of a decade for the numerous applications designers envision. But what about CFLs? “The CFL has a substantial future,” says Persson. To that end, IR's IRS2530D is a linear ballast control IC for CFLs with the capability to dim to below 10%. This 600-V part is the world's first 8-pin dimming IC, and as such addresses one big issue for dimmer ICs: the higher costs associated with the wire bonding of larger chips.
Things to come
Most observers agree though that, with LEDs, it's just a matter of time. One needs only go to Europe to see immediate opportunities for high-brightness LEDs (HB-LEDs). Linear Technology's HB-LED drivers are being used in a European oil refinery right now, says Tony Armstrong, because they make sense. “Take a look at the oil refinery — they run 24/7. They use (high pressure) sodium lighting, off yellow/orange. The lamps are inefficient and costly and not very nice to the environment. The cost of (HB-LED) conversion for the oil refinery is $500k. But you can save $160,000/yr using LED lighting. So after three and a half years, you have a payback. And no maintenance costs for 11 years. And you have the ability to detect differences between smoke from a fire, and steam, something you get with HB-LEDs that the human eye cannot detect with sodium.”
Looking at power management a year or two out, “you're going to see a lot more products in production,” said Bill Hutchings. “A broadening of applications. We won't see changes in peak efficiency in servers and computer supplies, but what we will see is somewhat better efficiency over all load conditions. The entire power chain is not particularly efficient in a server farm. There are UPSes hanging off the system, a lot of point-of-loads too. In the whole scheme of things, your 98% supply is buried in a morass of a whole bunch of power conversions so the power to the load is not that much. Things are going to get a lot smarter, efficiency of the entire system will increase.”
These factors will ultimately descend on thermal issues. “The bulk of power used in server farms is all in cooling, so any 1° of temperature you can take things down will benefit,” says Hutchings. Thus we can expect to see a change in the metrics as they relate to the cost per transaction as a function of cooling units and so on, he says. Meanwhile, power management ICs will be coming into play for solar cells, smart metering, and energy harvesting applications.
Understanding digital power supplies
|Buck, boost, buck-boost||75 - 78%|
|Resonant, zero-voltage switched||85 - 95%|
Inside resonant converters -
Microcontrollers are increasingly integrated into power conversion applications. The competing “digital power” philosophies, generally categorized in four levels of complexity from 1 to 4, are widely perceived as the “state machine” analog method for low-end applications; the microcontroller-with-hardware accelerator approach (often called DSP-like) for midlevel tasks; and the digital signal processor technique at the high end for full digital control. But “state machine” doesn't necessarily equate to “analog circuitry.” Digital circuitry can be found at any of the four levels, and DSP versus DSP-like may often be far apart. Level 1 digital power supplies typically use an MCU to provide low-power standby modes, programmable soft start, and to sequence multiple power supply stages. Level 2 applications include output voltage margining, load sharing and balancing, and historical performance logging. Level 3 control applications use an analog control loop to regulate the power supply output. However, the MCU can change and optimize the control loop configuration, increasing power supply efficiency across multiple load levels. Level 3 applications provide a common, configurable platform that can be programmed to support multiple applications. And Level 4 supplies replace the analog control loop with an analog/digital converter to acquire power supply feedback and a digital pulse-width modulator for power stage control. Level 4 devices allow for dynamic adjustment of control loops and predictive control algorithms that optimize power supply efficiency.
Quasi-resonant and resonant topologies are replacing traditional flyback and two-switch forward topologies for higher efficiency and lower EMI. Both the resonant and quasi-resonant reduce the turn-on switching losses by introducing a sinusoidal waveform to the system's power stage and switching the stage at zero voltage (or zero-current). A resonant (tank) circuit is required for this type of operation. In a resonant converter, two MOSFETs or equivalent circuitry generate a square wave to excite the resonant circuit. The output from the tank is sinusoidal; higher-order components are attenuated. The current lags the voltage. During the time the current reverses across the power switch at the zero-crossing voltage, the switch consumes virtually no power.
A resonant converter, as diagramed here by STMicroelectronics, uses a resonant tank circuit to convert square waves into sinusoids for more efficient rectification. In particular, LLC resonant converters can provide a substantial improvement over fixed-frequency PWM and traditional half-bridge converters.
The drawback of these supplies is that at light loads, the frequency of operation goes up and so then does the system's turn-off losses. On the other hand, typical discontinuous/continuous-conduction mode (DCM/CCM) flyback systems (switching not synchronized) have the highest switching losses, and their efficiency is a function of the switching point. The quasi-resonant converter, used up to about 100 watts, can be viewed as an extension of a DCM flyback design but the switching is timed; thus, losses are reduced. Similarly, the resonant converter appears somewhat like the two-switch design, but it's very different. For instance, the LLC resonant converter (named for its two inductors and a capacitor in the supply's tank circuit, and used in applications up to about 500 watts) doesn't need an inductor in the output rectifier stage, as does a two-switch forward design. In addition, the circuit/ transformer topology is such that the resonant converter will usually be physically smaller for a given power output. On the other hand, its ability to handle widely varying voltages without appropriate front-end circuitry when used in an AC-input supply is not particularly good.
Texas Instruments Inc., www.ti.com
Darnell Group, www.darnell.com
National Semiconductor Corp., www.national.com
Fairchild Semiconductor International Inc., www.fairchildsemi.com
International Rectifier, www.irf.com
Microchip Technology Inc., www.microchip.com
ON Semiconductor, www.onsemi.com
Analog Devices Inc., www.analog.com
Linear Technology, www.linear.com
Power Sources Manufacturers Association (www.psma.com) online Energy Efficiency Database