With millions of cell phones, PDAs and portable computers to power, voltage regulators are the hottest segment of the standard linear business. Yet Steve Ohr sees a couple of market missteps on the power-management scene.
You've undoubtedly encountered this situation on a long plane trip, to Asia, say, or a cross-country flight: Your laptop computer says you have 80 percent of your battery life available; the next moment it starts beeping at you frantically to save your work and shut down before a dead battery causes you to lose data. "Fuel gauging" - accurate fuel gauging - is one task of portable power management.
I had promised to give the 10,000-foot view in voltage regulators. Well, "power management" is the fashionable term for what Dataquest says is the fastest growing part of the standard linear IC business. This market was $2.8 billion in 1999, and was dominated by companies like National Semiconductor (with $379 million in 1999 shipments), Linear Technology ($243 million) and ON Semiconductor ($203 million). Texas Instruments' 1999 acquisition of Unitrode gave it the No. 1 slot (with $480 million). Fairchild, Maxim Integrated Products, and STMicroelectronics were, respectively, Nos. 5, 6 and 7 in this market.
There are reasons for that growth. An increasing number of portable devices - cell phones, PDAs, and palmtop computers, in addition to laptops - depend on analog skills to regulate voltages without taking any additional juice from the battery. Everything is couched in terms of efficiency: 100 percent, a one-to-one power transfer from input to output, is the unobtainable ideal. The difference between, say, a 100 percent transfer and a 90 percent transfer is dissipated as heat.
Yet each portable device will require a different voltage regulator topology. Portable computers require multiple voltages from a 10- or 15-V battery source, one voltage to power disk drives and I/O devices, another to light up an LCD screen, still a different one to spark the Pentium core. Switching regulators (dc-to-dc converters), in which an oscillator and pulse-width modulator (PWM) are used to tap out a dc voltage from a dc source, generally have the highest efficiency under high-load (high current) conditions. They can be less efficient under light load conditions. So manufacturers like Linear Technology have devised "pulse-skipping" techniques for switchers in portable computers. The switcher operates at a lower frequency as the computer slips into a standby mode and actually resembles a linear regulator when the computer goes into a "sleep" condition.
A big requirement for switching regulator ICs these days for both portables and desktops is multiple outputs, often with differing voltage levels. Linear Technology and others have made inroads on a phasing technique that ensures that the duty cycles of multiple switchers on the same controller are effectively out of phase. This neutralizes current surges in the system and ensures an overall higher efficiency.
Another efficiency tweak to pay attention to is the matching between the power MOSFET switches and PWM controllers in a dc-to-dc converter. As MOSFET makers like International Rectifier and Vishay Siliconix like to point out, small increments in power supply efficiency can be obtained when the rise-and-fall curves for power switches and controllers are carefully matched.
Linear regulators, especially the low-dropout (LDO) variety, are used in cell phones and PDAs. The linear regulator uses a series pass transistor (a power device) as something of a current valve, regulated by an op amp in the feedback loop. It is much smaller and easier to use than a switching regulator, and it generates far less noise, which makes them popular in cellular handsets. The strength of this market (430 million cell phones likely will be shipped this year) makes a little company with strength in LDOs, Micrel, No. 8 in the overall voltage regulator market.
But the efficiency of these devices is contingent on relatively small differences between the input and the stepdown voltage. Even if it uses no power of its own, a linear regulator that, say, steps 5 V down to 3 V can only be 60 percent efficient. The best efficiency is obtained where an LDO must drop, say, a 3.6-V cell phone battery voltage to 3.3 V for its internal logic. Despite the potential for induced noise, some technologists argue that a miniature switcher - with a topology called "buck-boost" - would be most useful for 3.6-V cell phone batteries. The regulator would step down the phone voltage when the battery has all its power available and step up the drooping voltage as the battery runs down. National Semiconductor, I believe, has been instrumental in driving this topology for cell phones.
Yet with tons of wasted promotional effort out there, I see a couple of things that look like nonstarters. One is the smart battery concept. This is essentially an embedded data-acquisition system for computer batteries. An A/D converter reads the voltage level on the battery, and a microcontroller compares it with the level in a lookup table and reports back to the computer host. Although the main advantage is accurate fuel gauging for laptop PC batteries, the smart battery technology offers close matching for varying battery chemistries and charging control for one-time hazardous lithium-based batteries.
I think this is tremendous technology. The people who helped drive this, like Dave Timm of Maxim or Dale Stolitzka when he was at National (he's now at Analog Devices), deserve some kind of a medal. But it is a nonstarter, even when promoted by companies like BenchMarq (a Unitrode company now under the TI umbrella), a nonstarter. When the Smart Battery Specification was initially promoted as part of a partnership between Intel and Duracell, its backers believed that smart batteries would be ubiquitous as alkalines. But computer batteries reflect a wide variety of form factors; they are all custom made and no one manufacturer at this point has enough volume to sustain someone's smart battery IC product line. The result is that this technology may whither and die. Though my company-issue Dell computer has a smart battery fuel gauge, if I fly a 777 cross country (and upgrade myself to business class), I will plug my computer into the airliner's power line. To heck fuel-gauging.
The other nonstarter, I think, is the desktop version of ACPI, the Advanced Configuration and Power Interface. Developed by Compaq, Intel, Microsoft, Phoenix and Toshiba and sometimes called "instant on," ACPI puts computer management under control of the operating system. For example, Windows 98 will control the power-saving shutdown sequences for portable computers, turning off the disk drives, then the computer screen and eventually putting the entire machine to sleep.
Although power-saving sleep modes conserve battery life in portables, they have generally not been utilized in desktop machines. To be sure, Intel's Platform Architecture Lab has developed a desktop specification that allows certain parts of the desktop machine to go to sleep and come awake instantly with a tap of the keyboard. (No more 20-minute boot-up routines.) But unlike a notebook computer, which can go to sleep until its owner wakes it up, the desktop machine must retain enough consciousness to respond to a network or a modem call.
Unfortunately, most 56K modems and network interface cards (NIC) live on the PCI bus. Supporters of the desktop ACPI effort had to wrestle with the notion of putting the PC to sleep while keeping certain devices on the PCI bus awake. The answer was a partial redesign of the PCI bus: A previously unused line on the bus becomes Vaux, a 3.3-V standby voltage for modems and NICs. In operation, a network call wakes up the NIC, which in turn wakes up the machine and switches the PCI bus back to 5-V operation.
Once again, bravo for the engineers who thought this up (they've made great presentations at the annual Analog and Mixed-Signal Applications Conference). There seem to be some clever implementations of Vaux. But once again, the market seems to be indifferent to the advantages of desktop ACPI. If I'm typical of the average business user, nobody will use this. I'll leave my desktop NT machine on for weeks at a time (even when I'm traveling), indifferent to the impact on my electric bill.