By Brett Etter, Enpirion

Integrated power devices open up more board real estate for application circuitry.

Developers of sophisticated embedded systems have classically cast power in a supporting role--a necessary slab of electrically noisy hardware that occupies an inordinate slice of printed circuit board (PCB) real estate. But new, fully integrated power-supply devices let developers distribute tiny power chips around their designs, physically opening up vast amounts of board space on behalf of the application circuitry for which the system design was originally conceived. Distributed power architectures have been limited by available technology. But now, fully-integrated silicon-based power supply devices let engineers convert power directly at the various points-of-load throughout a system, and this, in turn, lets developers think in entirely new ways about system design, elevating power to a peer level with other enabling technologies. This article will discuss how engineers can take advantage of the dramatic changes taking place in power-supply devices, leveraging these new technologies to increase the functional density of their system designs.

Decentralizing Power
Moore's Law has imposed demands for power supplies that support the very-low-voltage, high-current evolution. This trend has also expedited the migration from centralized power-supply bricks to distributed power architectures, giving rise to dc/dc converter devices specifically developed to meet the resulting need for higher efficiencies and current densities. Compounding the issue is the fact that the number of voltage rails in each system continues to increase. As the functionality of a system increases, it drives the need for smaller and smaller process geometries in order to achieve a cost-effective solution. In the wake of years of process geometry reductions, the number of different chips in a typical system has dramatically grown. Today, ten or more discrete voltage levels might be required throughout a system.

A distributed power architecture is a decentralized system in which power conversion is tuned for, and takes place physically near the circuit requiring power, that ie, at the point-of-load (POL). In conventional board-level power architectures, circuit-appropriate voltages are generated at a centralized location, typically by a power module brick, and then run across the board to various points-of-load. Running these low voltages is inherently detrimental to power integrity since trace impedance levels have to be exceedingly low to avoid losses, and in some cases with multiple chips as a load, it may be impossible to achieve the regulation requirements for both. Conversely, distributed power architectures bus a higher, raw dc voltage that is then converted by localized dc/dc converters positioned near, and tuned specifically for, the different circuit loads they serve. Each localized POL converter is matched to the load requirement, improving dynamic response and eliminating the problems associated with distributing low voltages around the system. With a distributed architecture, POL voltages of 1.2V or lower for many system-on-chip (SoC) devices, do not have to be routed any further than necessary, thereby eradicating voltage drops and noise interference while increasing the voltage regulation accuracy.

Another benefit of distributing power to the point of load instead of generating it centrally is improved management of the inevitable thermal issues associated with power supplies. By dispersing total heat dissipation across a board or system, the need for heat sinks or high velocity airflow mechanisms is greatly reduced. With temperatures more evenly maintained across the entire system, reliability and noise specifications are easier to meet for many demanding applications such as mission-critical military or telecom infrastructure equipment.

The 80% Monster
Despite the proliferation of distributed power architectures in physically larger systems like telecommunications and networking equipment, embedded systems developers continue to face power-design challenges. Board level power supplies have historically been accurately perceived as big and bulky. Centralized power supplies can occupy as much as 80% of a board, leaving little room for the multitude of components required for today's feature-rich electronic products. Typically, a power solution is implemented using a number of discrete components including a pulse width modulator (PWM) controller, at least two discrete power MOSFETs, a large power inductor, large input and output capacitors, and a number of feedback and compensation components. Total part count is typically between 15 and 20 components, requiring four to six square inches of board area. These space constraints force engineers to design around the power circuitry, optimizing the placement of components in order to minimize the power supply's size. Further compounding problems, today's embedded systems employ a variety of chips, requiring six or more supply voltages, with lower voltages becoming increasingly common. Of course, each supply voltage variant further cramps the space available on a board for the actual functionality for which the board was designed.

Compared with distributed power, centralized power architectures are inherently costly and complex, as a single power supply attempts to be all things to all circuits. As the number of voltages required at the board level increases, the cost of duplicating the full converter functionality escalates. Since isolation, regulation, transformation, electromagnetic interference (EMI) filtering and input protection are repeated at every load, the costs and real estate requirements for power conversion can become daunting. Sequencing each of the different power levels on a board also poses significant power design challenges, and with analog power circuitry often being mysteriously unfamiliar territory for system designers, the entire task of powering a design breaks down into an unpleasant mess.

When you weigh the pros and cons of the various power solutions for low-voltage loads, it seems clear that the best-case scenario for a designer would be a miniature, high-performance POL converter that can be easily programmed for voltage scaling. Fortunately, recent power management breakthroughs have helped deliver turnkey power supply solutions with drastically smaller footprints to market. In particular, advances including higher-frequency power transistors and the use of new and novel magnetics technologies have resulted in the development of a dc/dc converter with an integrated inductor in a single IC package, dramatically reducing PCB real-estate, lowering cost, and increasing performance.

New Power Solutions
An example of this new breed of power solutions is a unique line of switch-mode synchronous dc/dc converters that Enpirion has developed for POL applications. The devices integrate the power transistors with gate drive circuits, control and compensation networks, and magnetics into one standard-footprint IC package (see Figure 1).

The devices are the result two significant engineering breakthroughs: 1) the development of proprietary CMOS technologies that enable high-speed, high-power transistors for low-cost CMOS silicon processes; 2) pairing proprietary magnetic materials with a MEMS process to fabricate the inductor on chip. The high-speed CMOS process allows the power transistors to switch at five to ten times the speed of comparable devices, but with identical or better efficiency. This power transistor topology, in turn, enables the use of far smaller and lower-cost associated passive components. The high switching speed also enables a very wide bandwidth control loop to achieve better transient performance.

Using a MEMs process to integrate magnetics on chip eliminates large, external power inductors, as well as the poor performance characteristics some exhibit at high frequencies. The Enpirion POL converters' magnetic alloy enables high inductance and improved high-frequency performance, and they can be manufactured in a standard CMOS process. This approach enables the complete integration of the converter, which results in PCB area and parts-count reduction, ease of design, and significantly enhanced transient performance.

The end result is a fully integrated converter that takes up to 70% less space and has 60% fewer parts than conventional solutions. By shrinking the power supply so dramatically, designers get back the space and budget to increase value-added application capabilities and circuitry. This improvement in functional density is particularly important in applications where constrained space is a key design consideration. The alternative to increasing features, of course, is to reduce the overall product size and bill-of-material (BOM) costs, which is an appealing alternative for many designers.

As one can imagine, these advances in power management have turned the design paradigm on its head. No longer does an engineer have to design around the power supply, but rather can view it as a key enabler of a smaller, faster, more efficient system. With more space for functional circuitry rather than support circuitry, designers can focus on the features that will differentiate their product and get it to market faster (see Figure 2).

Power to the People
How can engineers take advantage of the space savings and ease of design afforded by ultra-integrated, miniaturized power converters? For starters, design teams won't necessarily require a full-time power supply expert, since a turnkey solution takes the guesswork out of power system configuration. Instead, design expertise and additional board space can be directly applied to the development of value-added features. Take telecommunications and data communications applications, for example, where power-supply circuitry overwhelmingly dominates board space (as much as 80%) due to the odd assortment of required voltages. By distributing dedicated power chips at each POL, designers can easily double, or even triple, communication-channel density by eliminating the old space-hungry power supply bricks.

Mass storage is another application that can realize a benefit from a miniaturized power supply in the form of increased storage density and higher retrieval speeds. And consider the improvements possible in regard to mission-critical systems, such as specialized medical equipment or aerospace systems. Many mission-critical applications are adversely affected by weight, shock, and vibration. With the additional board space afforded by a miniaturized power supply, designers can focus on features that mitigate these affects and deliver optimum reliability.

As system requirements continue to become more complex, power systems must also evolve. The recent breakthroughs in power management technology, which yielded miniaturized, fully-integrated dc/dc converters, is a boon to designers looking to maximize their PCB area, simplify design, and reduce costs. Engineers can now explore a full range of system design considerations that were once hampered by the power supply and are now enabled by the power system architecture and design.

Brett Etter is the Director of Product Marketing for Enpirion. He can be reached by phone at (908) 575-7550 and by e-mail at brett.etter@enpirion.com.