Savvy power sales reps listen to what designers are trying to achieve while they’re selling their products. Savvy system architects listen for emerging solutions, even in their power supply, to gain a competitive edge while continuing to buy products that are known to work for them. Both are trying to be open to better ways of doing things while following the tried and true.
But many power designers forgo the former and focus on the latter because they feel it’s their system that counts, not the power part. Yet ready or not, the marketplace forces power designers to increase the efficiency, reduce the noise, make things smaller, and provide lower voltages. And to those demands, the power-supply chain listens. It must.
THE MARKET CHANGES
The market, in the form of ever-lower dc-dc converter output voltages required by ever-smaller ICs, caused a seismic shift for power designers. In response, many changes were induced including higher output currents, smaller bricks, and new architectures, such as the intermediate bus and factorized power, greater power density, higher efficiency, and faster transient response.
But it works the other way, too. Innovative suppliers may see needs or trends that aren’t clear enough to convince some system architects to adopt a new concept. It may be better, but it may not be obvious. If the new concept requires retooling, rethinking the system design, or reworking cost and design tradeoffs—almost certainly to be the case—it’s easier to put off a decision. Wait and see.
Today’s brick was such a new concept at its introduction. What insight did it take or what degree of painful change was involved for the designer who bought the first brick? It addressed the hard part of power-supply design—the power conversion. Once the power-conversion function was encased in a brick, the designer no longer had to mess with discrete designs, close a control loop, and ensure stability over line and load conditions. That was already figured out.
The heart of the power supply was built and put into a compact brick. Discerning system architects gravitated to that because it would shorten their time-to-market. They no doubt paid a price, but they gained advantages over their competition. “We’re not power-supply engineers anyway,” some thought. “Why do we even need to be messing with this?”
Yet some designers still weren’t satisfied. They wanted the power supply to be more plug and play. The brick simplified the design of the power supply, but it was still necessary to put them on a printed-circuit board (PCB), make the interconnections, and add a few external components. Pre-engineered configurable power supplies provided the answer and some advantages: short time-to-market, low risk, and cost-effective power.
Other forces than those exerted by buyer and seller influence changes in the market as well. Initially, for instance, the military was fearful and resisted the commercial off-the-shelf (COTS) purchasing guidelines. They wanted their electronic systems and components exactly the way they wanted them, with high reliability and long lifetime. More or less, they’re now willing to accept the change for many of their mainstream applications.
THE ROLE OF ARCHITECTURE
New architectures often are accompanied by innovative power converters. Bricks helped distributed power become a successful system architecture. Intermediate bus converters and non-isolated point-of-load (POL) regulators did the same for the Intermediate Bus Architecture (IBA). The Factorized Power Architecture (FPA), using a miniaturized high-power-density pre-regulator followed by a point-of-load transformer in small packages, continues to gain traction.
In the high-end server market, overall system efficiency, dynamic performance, and lower POL voltage requirements have strained the IBA, which incorporates a multiphase buck converter for the lower load voltages. In addition to improving the thermal aspect of the power system while achieving high reliability, the FPA locates a current-multiplier-type device at the POL that isn’t hamstrung by an external control loop and provides a significant advantage for the demands of the latest processors.
While optimizing a distribution system for a move to a higher-voltage bus, it eliminates the need to use lower bus vol-tages to compensate for the duty-cycle limitations of the buck converter when it needs to supply a low POL voltage. It’s a natural next step in the evolution of high-end server architecture development.
That’s the challenge: to find the right solution, using new technology that really clicks with a certain application and a certain market. How do you match that up? Maybe the market speaks and power listens. Power speaks, and the market listens. It has to be a two-way street.