It is no coincidence that this week's "Focus" on power management runs concurrently with the Penton Wireless Symposium/Portable by Design Conference in San Jose, Calif. (Feb. 22-26). Power management is one of the fastest-growing markets for analog and mixed-signal ICs. And portable appliances-especially computers and cell phones-are among the biggest consumers of power-management devices.
Len Sherman, an expert on cell-phone power-management
issues at Maxim, recommends using active 'ringing clamps'
across inductive filters to control high-current loads in the
RF power amplifiers used in portable cell phones.
Power-management devices-essentially voltage regulators-must transform the voltages provided by computer or cell-phone batteries into the 5, 3.3 or 2.5 V needed to power the internal logic of the portable appliance. In some cases, the voltage regulator is a charge pump or dc/dc converter that uses a pulsewidth modulator (PWM) and MOSFET switches connected to the battery terminal to step up ("pump up") or step down the battery voltage.
In other cases, the regulator will be a linear type in which an IC error amplifier is used to open or close the MOSFET switch linearly on the battery terminal. In all cases, the goal is to make the regulator highly efficient, to avoid tapping into the battery life of the portable-or the wattage drawn from the ac line.
There are profound differences between linear and switching regulators, in the regulators used for portable vs. desktop computers, and in those used for wireless transmission devices. With cell phones, the trend is to replace multicell nickel-metal-hydride rechargeable batteries with single-cell lithium-ion, whose voltage can vary from 4.2 to 2.5 V as it discharges. That means the voltage regulator must step down the battery voltage at the beginning of the cell-phone usage cycle to provide, say, 3.3 V for baseband processors, but then step it up again as the battery nears the end of its discharge cycle.
Portable computers have a different set of problems and contingencies. Laptops have bigger batteries and higher voltages-10.8 to 15 V-than cell phones. However, mobile Pentium cores, buses and I/O ports, disk drives, floppies, CD-ROM players and LCD screens each have different voltage and current requirements, and those take their toll on battery life. Portable computers have their own arsenal of switching and linear regulators, but each must be extraordinarily efficient to avoid draining battery life or dissipating heat.
For desktop power management, many systems builders are implementing the Power Supply '98 (PS98) specification, an outgrowth of the Advanced Computer Power Initiative begun by Intel, Microsoft and portable-computer maker Toshiba. Basically, PS98 was meant to bring the same kind of power-conserving "sleep modes" to desktop computers that are seen in laptops.
It's been a major problem keeping certain devices in the PC-especially those on the PCI bus-awake while the rest of the unit sleeps. The solution seems to revolve around Pentium-controlled voltage regulators, some with multiple output voltages, distributed across the motherboard, along with MOSFET switching for the PCI power lines.
Switching regulators (dc/dc converters) with "buck boost" capabilities are increasingly used. Many experts, such as Jim Schuessler, strategic marketing manager at National Semiconductor Corp. (Santa Clara, Calif.), believe that the Li-ion battery's discharge from 4.2 to 2.5 V will require a buck-boost topology for systems using 3.3-V logic.
But that technique is not without controversy. Anthony Armstrong, MOSFET marketing manager at Vishay-Siliconix (Santa Clara), believes that step-down regulators in cell phones will be the norm. According to Armstrong, it is very likely that cell-phone makers will simply tune their current 3.3-V systems to slip into a standby mode and demand a recharge from the user as the single-cell Li-ion battery drops from 4.2 to 3.6 V.
Moreover, the trend toward lower-voltage logic-2.8, 2.5 and eventually 1.8 V-means that RF transmitters, receivers, ARM-based microcontrollers and DSP baseband processors will all be powered by low-dropout linear regulators (LDOs) connected directly to the battery terminals. That would all but eliminate the requirement for boost converters-except for cell-phone power amps, which have specialized power requirements in any case.
Where dc/dc converters are used, according to Len Sherman, staff scientist at Maxim Integrated Products (Sunnyvale, Calif.), care must be taken to keep the switching noise out of the sensitive RF receiver circuitry. LDOs are widely used behind the dc/dc converter to provide the distribution voltages for the many cell-phone devices. In his Focus article, Sherman discusses a primary concern: suppressing the noise from dc/dc converters. He recommends active "ringing clamps" across the inductive filters of the converter and capacitor reservoirs for dealing with high-current pulsed loads of power amplifiers. He also addresses the problem of generating the negative bias voltages required by gallium-arsenide RF transmitters.
The RF power amp, in fact, is the single largest consumer of cell-phone battery power. "Users today enjoy more than 200 hours of standby time on cellular telephones because the power amp is inactive," writes National's Schuessler in his piece, which serves as a review of the battery-load trends for cellular handsets. But because of the pulsed-power requirement of power amps, actual talk time is less than a tenth of the standby time.
Though there is a tendency to conserve power consumption with low-voltage CMOS, Schuessler believes that certain devices will continue to demand 3 V or more. He advocates buck-boost voltage regulators but suggests that a switched-capacitor topology may offer higher efficiency than inductive PWMs.
In his follow-up piece, Milt Wilcox, an engineering design manager for Linear Technology Corp. (Milpitas, Calif.), argues that inserting additional regulators into the power-conversion chain from the battery may not compromise efficiency. A two-step conversion process, which first transforms a 15-V battery source into a 5-V line and then steps that 5 V down to 1.5 V, may be just as efficient as the conversion process that takes the 15 V directly down to 1.5 V. What's more, the two-step process may offer additional advantages in terms of noise and spike reduction and power distribution.
However, power management in desktops creates some interesting challenges. In Intel's conception, first disclosed in 1997 at the Intel Developers' Forum, an unused PC would waltz from fully awake to standby and deep-sleep modes-yet pop awake at the touch of a keyboard. In its S3 (or "Suspend to RAM") mode, the PC would consume less than 5 W. The screen would be dark-even the fans would be silent.
More than a year after the release of Intel's desktop power-management spec, PS98, manufacturers of analog switches and regulators are finally devising ways for controlling power distribution on desktop motherboards. In addition to voltage-regular modules for Pentium core voltages, manufacturers such as Analog Devices, Harris Semiconductor and Semtech are developing voltage regulators with multiple outputs-1.5 V for GTL+ bus logic and 2.5 V for the clock system, in addition to the CPU core voltage, which itself could vary from 3.6 to 1.8 V in 100-mV increments. Parts such as the ADI ADP3155 or the Semtec SC1185 include a synchronous buck regulator with a DAC-controlled variable output voltage for the CPU core and two linear regulators. The most sophisticated devices have enable or standby pins to handle power-up and power-down states.
Implementing much of the PS98 power spec is reasonably straightforward, said Chuck Stancil, a systems-design manager with Compaq Computer Corp. (Houston). The major issue is responding to what the spec calls a "power management event" (PME)-a modem call or Ethernet network header that would require devices on the PCI bus to awaken the entire system.
"A modem card could require as much as 375 mA to stay awake on the PCI, while sleeping devices require only 10 or 20 mA," said Stancil, whose responsibilities include power-supply design. "But how do you direct 375 mA to just one PCI slot-the one holding the modem or network card?"
The solution, derived by Intel, actually creates a 3.3-V standby power line (Vaux). In operation, a PME will force the power-system controller not to pump into the standby line but to switch power lines entirely, explained Stancil. The switching of bus lines must be coordinated with the maintenance of SDRAM (or Rambus) memory banks in S3 states. An important question that came up for his group was how much switching control to put into the "silver box" and how much to embed onto the motherboard.
Fortunately, Stancil reminds us, control signals for power switching and sequencing are available from the CPU and/or its south-bridge controller. He also credits Analog Devices Inc. (Norwood,Mass.) for providing a number of power-line and system-monitoring options.
Meanwhile, Texas Instruments Inc.'s Jonathan Bearsfield discusses the means of implementing sleep modes in desktops and server PCs in his power-management piece. The Vaux line in the PS98 specification (Rev 1.0) and the newest PCI spec (2.1) call for a 3.3-V standby line on the PCI bus. That keeps modems and network PHYs awake while the rest of the PC sleeps, enabling wake-on-LAN and wake-on-modem calls. But rather than pull 3 or 5 A or more from the 3.3-V standby line, the spec asks the system to switch over entirely from standby to a more powerful line.
"Until now there have not been many elegant ways to support both low-power modes and instant-on capabilities in desktop computers and servers," writes Bearsfield, who serves as a systems engineer. The key word here is "elegant." Bearsfield shows some Vaux line-switching techniques using diodes and MOSFET switches, as well as self-contained devices that implement Vaux on the Cardbus.
Later in this section, David Heacock, director of portable products at Unitrode (Merrimack, N.H.), offers a tutorial on smart-battery ICs. These miniature data-acquisition systems adjust to battery chemistries and monitor their charging and discharging states, reporting battery conditions to the host processor and often adjusting to error conditions. Their major advantage, Heacock reports, is "fuel gauging"-more accurate reporting of battery depletion and availability.
Additional updated information on specifications related to power-management issues can be found at Intel's Web site (developer. intel.com/technology/iapc/tech.htm). The site includes information on the Power Supply 98 specification (Rev 1.0), PCI specification 2.1, PCI Bus Power Management Interface and 3.3- Vaux ECR, as well as a PC Power Management Design Guide.