Three trends are emerging in dc-dc converters. First, the news in bricks is all about incredible power densities and efficiencies. Second, differentiation has become a matter of control and distribution architecture for point-of-load (POL) converter makers. And third, while some readers cringe when they see yet another white-LED power story, white LEDs define something of a paradigm for power conversion in products from handhelds to vehicular lighting.
The essence of distributed power, bricks constitute a leapfrog market with continuous one-upmanship. Artesyn Technologies recently introduced 200-W quarter-bricks with up to 92% efficiency. In that form factor, the 5-V model provides 200 W/in.3 In the half-brick size, TDK Innoveta's Veta line includes a 28-V model with 450-W capability.
The intermediate bus architecture (IBA) was the next step after distributed power and bricks. IBA steps down the system's nominal 48 V and distributes isolated, loosely regulated 12-V (occasionally 8 or 5 V) power around each board. There, power is stepped down to the tightly regulated level needed by the load for non-isolated buck-architecture POL regulators (or niPOLs).
But it isn't quite that simple. Some IC makers have specs for the rate at which power ramps up. Plus, chips with separate power rings for core, memory, and I/O may be fussy about which ring is energized first. As a result, POLs acquired pins that allowed them to be brought up at a controlled dV/dt and in sequence. Simple sequencing involves daisy-chaining the POLs on a board so each is inhibited until the POL immediately ahead of it in the chain appears.
For slew-rate control, some POLs have a "track" pin. By connecting the track pins of multiple POLs, engineers can force them all to follow a common RC ramp waveform, applied through a single MOSFET. An alternative is an analog power-supply tracking manager IC that permits independent control of turn-on sequencing and slew rate.
Multiple-POL control doesn't stop there. Unless the frequencies and phases of the switchers in multiple POLs are staggered, they can interact with each other. This can lead to electromagnetic interference on the intermediate bus, and it becomes even more challenging in multiphase POLs.
Last year, it started to become clear to POL suppliers that adding sequencing, slew-rate, and synchronizing features ad hoc was akin to building the Tower of Babel. This meant a return to the drawing board. In the process, the suppliers also began to see that a consistent, system-oriented approach would enable system operators to collect data from the POLs and make it available for monitoring and diagnostics.
The I2C bus is one candidate for control. Last fall, Artesyn Technologies and Emerson's Astec Power announced a coalition of power-supply and semiconductor manufacturers. Its aim is to create a new communications standard that defines an open architecture for power systems using I2C, namely the "Power Management Bus," or PMBus.
In addition to the power-supply manufacturers, semiconductor participants include Intersil, Microchip Technology, Summit Microelectronics, Texas Instruments, Volterra Semi, and Zilker Labs. By using PMBus-compliant products, OEMs should be able to control all converters using the same set of commands.
The PMBus consortium was no doubt encouraged by an announcement last summer from Power-One concerning the availability of the POL component of its Digital IBA scheme. The company's Z-One Digital IBA approach relies on proprietary Digital Power Manager (DPM) chips, smart POL converters, and a Windows GUI-based programming interface to set output voltages, sequencing, tracking, and protection limits and to program for monitoring and diagnostic feedback.
The conceptual differentiator between Power-One's Digital IBA and PMBus is where the I2C connection stops. In Digital IBA, I2C commands go to the DPM, which talks to the POLs via a custom single-wire bus. With PMBus, the POLs themselves will each have an I2C port.
For a system designer, deciding which to use probably comes down to the number of loads that need to be controlled and monitored. According to Power-One, the greater the number of loads, the more system-board real estate is saved by the approach. But even with relatively few POLs, Digital IBA has advantages in the graphical elements of its design interface and its range of control.
A contrary point of view comes from Vicor, articulated strenuously by company president Patrizio Vinciarelli. Inventor of the zero-current switching and zero-voltage switching technologies, he founded Vicor in 1981 and has served as chairman of the board, president, and CEO since then. Vinciarelli has been a Fellow at both the Institute for Advanced Studies and CERN, so when he says we're marching over a cliff, it's wise to listen.
He first notes that the greater the I/O voltage ratios are in conventional niPOLs, the more their efficiency falls off. The obvious response is to reduce the bus voltage. But when the bus is carrying tens of amps of current, Joule's law losses make that strategy counterproductive. IBA also places a lot of bulk capacitance on the output of each niPOL to store the energy needed when there's a sudden increase in load demand.
So, Vinciarelli and Vicor created the factorized power architecture (FPA). It uses a higher bus voltage, with concomitant low I2R losses and voltage drop. Bulk capacitance moves from the POL's output to the input, reducing the amount of capacitance needed by the square of the step-down ratio. The regulation function can reside anywhere in the system, instead of at the point of load. FPA also makes it easy to regulate the load voltage through the isolation barrier without long, noise-sensitive feedback lines or opto- or magnetic couplers.
There are two FPA modules. Taking power from the 48-V system bus upstream, the pre-regulator module (PRM) is a high-VOUT buck-boost regulator that provides a controlled, non-isolated FPA bus voltage to the second module, called the voltage-transformation module (VTM). Load regulation is performed using feedback to the upstream PRM. The PRM adjusts its output to maintain the load voltage in regulation.
The VTM's task is to multiply the current (or divide the voltage) by some factor, K. This takes place with essentially a 100% transformation duty cycle, so there's no loss of efficiency at high values of K. Consequently, the bus voltage is limited only by safety standards.
VTMs can respond virtually instantaneously to changing loads, making it possible to place bulk capacitance at the VTM's input instead of its output. The benefit of this is that the actual capacitance can be reduced by the square of the VTM current gain.
PRMs can operate efficiently in both buck and boost modes even when VOUT is very close to VIN. VTMs use a zero-voltage switching and zero-current switching topology called a sine amplitude converter. The power train is a low-Q, high-frequency, controlled oscillator.
It's easy to think a system must be all IBA or all FPA. Not so, Vinciarelli says. He envisions applications in which IBA and factorized power coexist peacefully on the same board, factorized power doing the heavy lifting and IBA effectively handling the lighter tasks. Clearly, some companies have seen the light. In fact, Sony licensed factorized power from Vicor last July.
Many LEDs populate battery-powered consumer products, and just as many companies make ICs to light them. LEDs also are taking over in cars and even planes (see the figure). Beyond LEDs, other converters manage charging, supply voltages that bias the displays, run separate cellular telephone and wireless (Bluetooth and WiFi) transceivers, and keep the system going. If the success of the iPod is any indicator, your next cell phone also will need power for its own disk drive.