Power Supplies Aim for Efficiency and Economy
“The fundamental question comes down to who will pay for efficiency?” said Kulbir Dhillon, director of product development at C&D Technologies. “Engineers might think, ‘Vendor A has an 80 percent efficient 1,000W supply for $150 but Vendor B sells a 90 percent efficient 1,000W supply for $200, so why should I buy the more expensive supply?’ They want efficiency, but in the short term they may not want to pay for it. They might not realize they could recoup that cost quickly.”
Manufacturers offer a variety of products and technologies that help engineers design economical and efficient supplies and get high efficiency from off-the-shelf supplies and components.
Many devices remain powered in a standby mode at all times. So designers must provide efficient standby supplies that do not waste energy. “We see requirements driven primarily by European initiatives for a no-load power consumption of 1/2Watt or less,” said Laszlo Balogh an executive technical fellow at Fairchild Semiconductor. “And the requirement may drop to 300 mW for power supplies rated less than 60W.”
Many designers connect a main power supply and a standby power supply in parallel so each can run efficiently at its own power level. Standby supplies typically use low-cost fly-back converters optimized for low power. Balogh points to the Fairchild FSCQ-Series devices optimized for use in quasi-resonant converters. “The resonant action for the switching intervals reduces switching losses and maximizes efficiency for low-power levels,” said Balogh. “And we have optimized every piece of silicon in the ‘power chain;’ FETs, diodes and controllers.” For instance, Fairchild also has introduced the Stealth-II Series of silicon diodes for power-factor correction (PFC) applications. Those devices approach the performance of silicon-carbide diodes, but with a lower forward voltage drop and at lower cost.
Technology improvements filter down to lower voltage applications, and integration of components such as MOSFETs and drivers in one package can offer a significant advantage in some low-voltage higher-speed applications. You gain a faster circuit, which means you reduce switching losses.
The management of the power systems in telecom applications has received a lot of attention recently. “When I designed telecom equipment in the early ‘90’s, we had a digital supervisor for the whole system in the central office even though analog controllers predominated for power-supply control,” said Balogh of Fairchild. “Now, operators of data centers, server farms and even smaller systems accept — and in some cases require — digital power management. In addition, digital loop control has made its debut in power supplies. Digital power management provides us with tools that optimize system efficiency, while digital control loops let us implement features that address the efficiency of individual power supplies. It would be cumbersome to provide those control functions using traditional analog-control techniques.”
International Rectifier has identified the resonant half-bridge configuration as a way to help engineers create an efficient power supply that minimizes radiated and conducted EMI. “Engineers know about this topology — related to a resonant tank circuit — but it is difficult to control,” noted Mario Battello, marketing manager for SMPS applications at International Rectifier. “Usually the secondary stage must rectify a square-wave-like signal that comes from the transformer. But in synchronous rectification you have a resonant signal much closer to a sinusoidal waveform, so the circuit produces less EMI.”
IR will soon have a new device in its synchronous-rectification IC family that will help engineers design the secondary stage. “Instead of using diodes, we switch MOSFETs in a resonant half-bridge configuration independent of the primary stage,” explained Battello. The company’s IR1166 and IR1167 devices use a MOSFET’s Rds(on) resistance to serve as a shunt resistor so the IC can sense the voltage across the MOSFET.
The IC detects when the current gets close to zero and turns the MOSFET off, which makes it operate like a diode. But because the IC turns the MOSFET off before the current changes direction, charge on the output capacitor does not flow back into the primary stage. “Our synchronous rectification approach can boost efficiency by about 2 percent. So a circuit that delivers 90 percent efficiency can now deliver 92 percent efficiency without requiring a design change,” explained Batello. “And you can eliminate the heat sink in the secondary stage, which reduces the space a supply needs and simplifies its mechanical design.”
But what happens when a supply must go into a standby mode? “All our products provide an enable pin that lets you put them in a ‘sleep’ mode,” said Battello. “In that mode, the IC stops all switching actions and draws only a small quiescent current — <1µA 100.”
“Many engineers want to design a power supply that can adapt itself to operating conditions,” said David Figoli, an applications engineer for the C2000 digital signal controller group at Texas Instruments. “Digital control lets us change a supply’s parameters so it operates as efficiently as possible under many load conditions. A circuit will always have switching, IR, magnetic and other losses. If you change operating frequencies, the boost levels, and the switching dead-band or gap times, you can operate a supply at high efficiency under almost any load condition.
“You can use digital or analog techniques to reduce a supply’s switching frequency as the load increases, or increase the frequency as the load decreases,” explained Larry Spaziani, product line manager for power-supply control products at Texas Instruments. “Changing the frequency as the load varies keeps efficiency high. The UCC28600, one of the green-mode power supply chips we introduced last year, provides frequency ‘fold back’ and several other operating modes.”
Designers also can substitute MOSFETs for diodes to create synchronous rectifiers. “A diode will typically drop 1V across it, so at higher currents that drop can create a lot of heat,” said Spaziani. “Using a MOSFET in place of a diode reduces that loss ten fold, but you need digital and analog ICs to control a MOSFET. So, synchronous rectifiers increase efficiency, but a design becomes a bit more complex.”
And, MOSFET-based designs require attention to details. “If you give a MOSFET a 12V gate-drive signal but it needs only 5V, you waste power,” said Spaziani. “Manufacturers have improved MOSFETs so they now work with a gate drive of only 5V to 8V.” TI now offers a high-frequency synchronous driver — the TPS28225 — that matches the drive levels required by the newest MOSFETs.
“Designers also can use an interleaving technique,” noted Spaziani. “If you need 1,000W, you connect two 500W supplies in parallel. The interleaving technique operates these two power stages 180° out of phase, which offers cost and size advantages. If you do not need both power stages, you turn one off. By turning it off at the right operating point, you can make design tradeoffs between switching losses and conduction losses. At full-power output, conduction losses produce the most heat, but at mid-range the power-output switching losses predominate.” TI has several analog ICs under development that will support the interleaving technique.
In addition to using AC/DC power supplies and DC/DC converters, engineers also may use an intermediate bus architecture that distributes a high DC voltage, say 48V, throughout a rack of equipment. Then, an unregulated intermediate bus converter converts the 48V power to an intermediate output or, say 12V DC. Finally, a voltage-regulator module (VRM) provides an even lower voltage directly to a microprocessor chip.
“Problems arise, however,” said Stephen Oliver, vice president of marketing and sales for the V.I Chip at Vicor. “First, you have the cost of a bus converter. Second, distributing 12V at high current can lead to power losses in bus bars, connectors and PCB traces.” So, Vicor replaces the intermediate bus with V.I chips that allow a direct step from 48V to the voltage a load requires. The chips multiply currents and divide voltages, and preserve the V-times-I power relationship.
Engineers can place a regulator on a board or backplane and put a voltage “transformer” circuit next to the high-current load (Figure 1); and the V.I chips are the key to building those efficient regulator and transformer modules. Engineers can use either analog or digital circuits to control V.I chips, which can adapt to new power-supply designs and architectures.
If you use a factorized regulator with a factorized voltage transformer, then you can increase the efficiency of that portion of your power system by about 7 to 8 percent. “We found that replacing traditional 48V load converters with factorized power modules saves about $30 per year in power and air-conditioning costs for each per microprocessor,” said Oliver. For companies that have thousands of processors in a data center, savings quickly add up.
John Wanes, director of technology development at C&D Technologies sees a growing interest in “digital power” or the use of digital circuits and communication to optimize power-supply operations. “Digital control lets you optimize the power levels throughout a system,” said Wanes. “Suppose
C&D Technologies has provided digital interfaces in its power supplies for some time and now licenses the Z-One technology from Power One and serves as a second for Power One’s Z-One DC/DC point-of-load products. “That license gives us access to the Z-One digital-power technology, and we have started to use it in some products,” said C&D’s Kulbir Dhillon, director of business solutions. “Given a specific voltage, current or load condition, we can optimize a supply’s performance. And engineers can incorporate the digital control in high-level system-management software.” (Figure 2.)
“Power One incorporates a digital power-manager chip in the DC/DC point-of-load modules,” said Dhillon. “You can connect modules in parallel and control voltage rise, tracking and sequencing, and you can monitor temperature, current, voltage and other parameters.”
John Wanes explained that C&D will use digital control in its AC/DC power supplies to selectively control operating conditions and optimize efficiency. “Digital control has another benefit: You can reduce the number of components in a design. That means circuitry does not waste power on housekeeping tasks under light-load conditions.”
No matter what technology engineers choose, they should remember components, packaging techniques and circuit designs must work hand in hand. You want economy and reliability; but a power supply must meet safety and EMC requirements, too.
For Further Reading
• “Code of Conduct on Energy Efficiency of External Power Supplies,” Version 2, 24 November 2005. European Commission, Institute for Environment and Sustainability. http://re.jrc.ec.europa.eu/energyefficiency/html/standby_initiative.htm
• “Converting Analog Controllers to Smart Controllers with the TMS320C2000 DSPs,” Application Report SPRA995. Texas Instruments. www.ti.com/litv/pdf/spra995
• “Design Guidelines for Quasi-Resonant Converters Using FSCQ-Series Fairchild Power Switch (FPS),” Application Note AN4146. www.fairchildsemi.com
• “Design of Secondary Side Rectification Using IR1167 SmartRectifier Control IC,” Application Note AN-1087. International Rectifier. www.irf.com.
• “Designing a TMS320F280x Based Digitally Controlled DC/DC Switching Power Supply,” Application Report SPRAAB3. Texas Instruments. focus.ti.com/lit/an/spraab3/spraab3.pdf
• “Factorized Power Architecture and V.I Chips,” Vicor. www.vicorpower.com/products/vichip/. Click on “FPA Overview.”
• “Using Virtualization and Digital Control Technologies to Increase Data Center Operating Efficiency,” Liebert Corp. www.liebert.com.
• For information about the Z-One technology, see: www.z-alliance.org.
• For information about the 80 PLUS program that aims to improve efficiency of PC and server power supplies, see: www.80plus.org.