Digital systems that in the past would have used a single, centralized-power supply now use distributed-power supplies. A distributed-power architecture is particularly attractive to several classes of digital systems. The first class is those systems that need multiple voltage levels. Also, highly configurable systems, such as workstations, can benefit from distributed power. Distributed power is not limited to just small- and medium-sized digital systems; by using distributed power, large digital systems, such as private-branch exchanges and digital telecommunications systems, can eliminate both high-current, low-voltage bus bars and the single-point failure mode of a centralized-power supply.

The term "distributed power" means that each pc board or module in a digital system has its own local dc/dc converter situated physically close to the point of load (Ref 1). One key factor of the attractiveness of distributed power is purely mechanical: It lets you handle distributed-power dc/dc converters as components, assembling them on your pc boards, just as you would any other component. Bulk power supplies, on the other hand, are typically large mechanical assemblies that you must install and connect separately.

Distributed power leads to a significant advantage for racked systems. Rather than distributing low voltages, such as 3 or 5V, at high current, a distributed-power system's ac/dc "bulk" converter dispenses higher voltages, such as 48 or 300V dc to its pc boards (Fig 1). Board-level distributed-power converters are available that accept standard telecomm, military, and industrial input-voltage levels. These higher intermediate voltages obviously result in proportionally lower currents. Consequently, conducting the lower currents requires much less copper and fewer backplane-connector pins.

Distributed-power systems have many physically smaller power assemblies than do centralized-power systems. Centralized-power systems tend to have a small number of heavy assemblies. The weight of the power assemblies is important not only in manufacturing but also in the impact on a system's resistance to vibration and shock.

As the inexpensive supplies in PCs show, the hardware cost of a custom, bulk supply can be very low, but other custom-supply costs are not. A custom supply often takes a significant amount of time to design. The power supply's designer must foretell the maximum load currents the supply will encounter over the life of a product, even if the owner later installs options. While standard ICs' data sheets provide the means to estimate power consumption, such estimations for custom devices can require advanced tools, such as Systems Science's $18,500 PowerSim for VHDL ICs. Consequently, custom supplies are often overdesigned.

Any changes in requirements entail design changes. After each redesign, you must requalify the custom supply with safety agencies. If the custom power supply has a fan, the fan affects reliability because fans are a limited-lifetime component with a relatively high failure rate.

Rather than concentrate power converters and their resulting power dissipation, distributed-power systems diffuse heat throughout a system. Using distributed power, onboard converters in the 5 to 50W range can supply most loads. In these cases, natural convection can often cool the system, eliminating the use of fans. Higher loads often require forced convection. The price for poor cooling is a 50% reduction in MTBF for every 10°C temperature rise. Or, as Calex's Steve Hageman says, "If you cannot touch your design because it runs too hot, it probably isn't reliable."

Distributed power is not a new concept. Engineers have long been using DIP-sized dc/dc converters to develop tiny amounts ±12 or ±15V for RS-232C ports or small analog circuits from local 5V digital-circuit power. Even today, a less efficient—but much less costly—onboard linear regulator is often the best choice for deriving small amounts of power from a higher intermediate voltage. The telecomm industry is also using small, board-mounted, dc/dc converters to develop electronics voltages from standard telecomm-equipment voltages, such as 48V dc. However, these dc/dc converters are limited to specific applications.

The concept of distributed power really took off in the mid-1980s when Vicor Inc fielded high-power, compact converters in component form, and, simultaneously, the workstation industry needed to develop highly configurable products. Vicor's early lead led to an industrywide "Vicor-standard" footprint—but, alas, not a standard pinout or any compatibility between products from different suppliers (Fig 2).

Benefits of distributed power

Distributed power can reduce development cycles. You can select a converter for each pc board as you develop the board. The power supplies can thus be integral to your system—not an afterthought. You do not have to wait until the end of your development cycle to determine a system's power requirements all at once. A distributed-power system design is very predictable and can lead to reduced NRE costs. These savings can outweigh the distributed-power modules' higher cost. Power Micro expects 1500W distributed-power systems to cost less than $0.75/W within the next two to three years.

Using the 1-converter/board approach eliminates low-voltage dc distribution—except for distributing power on individual pc boards themselves, of course. In other words, you can eliminate hefty wiring harnesses and bus bars. Because a distributed-power system minimizes parasitics, it can have better transient response than that of a centralized-power system.

Upgrades are often easier, too. Consider that when you want to upgrade a system, you may want to do more than just increase the power. You may need to add a new voltage for some advanced ICs that operate from lower voltages. As part of an upgrade, you can sometimes simply swap out the local power converter rather than the whole power system. Upgrading a system having a centralized-power supply this way may not be physically possible. The pc traces and backplane connector may not have enough pins or enough power-handling capacity.

The architecture of a power system includes not only the power buses, but also power-supply control, fault diagnosis, and status reporting. Distributed-power systems obviate remote sensing along with associated reliability and diagnostic problems in most systems.

You could monitor or control a wide range of power-system elements: output voltages, airflows, temperatures, and energy saving during battery operation, among others. The most basic and useful control is turning the converter on and off with an external signal. Using such signals, you can easily accomplish power sequencing. Some ac-control panels, such as those from Pulizzi Engineering, can help you sequence your bulk supplies.

Telecomm systems often require the converter to sense its own input voltage and to turn itself off if the input voltage goes below a certain value to safeguard a battery. Some newer converters allow you to program the voltage levels at which the converter turns on or off. You could also adjust the dc/dc converters' output-voltage margins with fixed resistors and analog switches or with D/A converters.

Fault isolation

You can isolate faults and contain damage more easily in a distributed-power system than in centralized-power system. You may need no more than simple board-level diagnostics. Distributed-power systems usually combine the converter with the "field-replaceable unit" (FRU) it powers. Standard engineering techniques can make pc boards and modules "hot-swappable." Thus, a service technician can simply replace the entire function and its power supply at the same time in the event of a failure (Fig 3). If a converter's output in a distributed-power system goes high, it damages only one pc board. If a centralized-power supply sustains an overvoltage condition, it can fry every component in the entire system.

Acceptable reliability differs, depending on whether you want a fault-tolerant or a high-availability system. By definition, no single failure ever brings down a fault-tolerant system. "Fault tolerant" implies full-blown duplication of hardware and exhaustive self-diagnostics. "High availability" means that only the rarest and most unlikely failures can bring down the system. High availability trades off availability for cost.

The most obvious potential culprit for a catastrophic, single-point failure in a distributed-power system is the ac/dc bulk converter. The probability of an output short in an ac/dc converter is very small, but not zero. Techniques used in high-availability systems to make the ac/dc conversion less failure-prone include N+1 redundant ac/dc converters (or N+2... N+M). A fault-tolerant system would have 2N-redundant ac/dc converters.

In some cases, a pc board may demand more current than a single board-level dc/dc converter can supply while still meeting the component-height restrictions of your card cage. In such cases, consider paralleling on-board converters. You can also parallel on-board converters for N+1 redundancy.

Whether you are paralleling ac/dc converters or dc/dc converters, paralleling adds complexity to the system and typically entails accepting some performance or cost compromises. When paralleling converters, mount all the converters in a common thermal environment so that they experience as close to the same temperature as possible.

Paralleling supplies with blocking ("ORing") diodes is more reliable than simply paralleling the supplies' outputs. Run such diodes hot, and use very low forward-drop devices. After all, reverse-leakage current is an issue only on failure.

Distributed power makes hot swapping easier. Because hot swapping a module of a distributed-power system affects only a small portion of the total power, "glitch-free" swaps are easy to ensure. Hot swapping can be a big advantage for large systems that must remain continuously on-line. Blocking diodes also simplify hot swapping.

If you step back and take a systemwide view, you will see that a distributed-power architecture duplicates many power-supply circuit elements. In a distributed-power system, each converter has its own control and fault-handling circuitry. In a bulk-supply system, the bulk-supply has only one of each of these elements.

Given that increasing the number of components decreases reliability, distributed-power makers have had to increase the reliability of their dc/dc converters. For example, Vicor has demonstrated MTBF of greater than 20 million hours. However, not all converter makers have taken the time to characterize their products over such long periods. Consequently, you often have no choice other than to rely on calculated MTBF.

Although telecomm standards exist for calculating MTBF, most power-supply vendors use MIL-HDBK-217 instead. Even though this practice is widespread, MIL-HDBK-217 has its problems; it depends on a database of component types and their field-failure rates. This database focuses on military components and takes time to accumulate. As a consequence, most newer commercial technologies are not available in the database.

MIL-HDBK-217 imposes a harsh—and possibly unjustified—penalty on nonmilitary components. Further, some of its component-failure rates are not consistent with those components' actual performances. For example, transformers and magnetic devices have a very low actual failure rate, but MIL-HDBK-217's predicted rate for these components is very high. ICs fare even worse than do magnetic components.

At least two converter companies have compared MIL-HDBK-217's predictions to the actual field performance of their dc/dc converters. Ericsson finds that its converters run 3 to 10 times longer than MIL-HDBK-217 predicts, while Vicor sees two to three times longer performance.

To select your intermediate-bus voltage, first consider the ease of safety approval vs cost. A lower voltage entails more expense to handle the higher currents, but the lower voltage may be more acceptable to regulatory agencies.

Every country has some kind of safety standard or requirement that limits the maximum voltage to which you can expose equipment operators and service personnel. The common term for this limit is "safety extra-low voltage" (SELV), but not all agencies set SELV at the same level. The most commonly accepted value for SELV is slightly more than 60V. Consequently, if your intermediate bus voltage is less than 60V, your product more easily complies with safety shielding and regulations.

However, your nominal intermediate-bus voltage has high and low limits for conditions such as battery charging and load switching. For the 48V-dc telecomm standard, for example, the maximum voltage is 60V—very close to the most generally accepted SELV limit. Therefore, a nominal 48V is currently the highest SELV for a distributed-power system's intermediate voltage.

But you could follow the example of mainframe-computer makers, rectifying and filtering the ac line to yield 300V-dc intermediate voltage. This scheme reduces the cost of both the ac/dc converter and the intermediate-voltage distribution. Your genuine safety concerns for a high-voltage bus are creepage and clearance, preventing access to shock hazards, and large amounts of stored energy available to short circuits.

If the load on the intermediate-voltage bus switches rapidly, such as when a fuse opens, the bus's inductance can generate a voltage pulse having as much as 70 Wsec of energy.

For a 300V intermediate bus, overload and short-circuit protection require large devices to handle inrush and arcing. However, there is a dearth of standard connectors and fewer standard converters for 300V. And backing up a 300V bus with a battery obviously requires more cells than does backing up a 48V bus.

You need to carefully consider overcurrent protection for the intermediate bus. Two common problems are that start-up requires large currents to charge the bus's capacitance. This large charging current means that the overcurrent-protection circuits can trip at start-up. The result is that the system never actually gets started. In this case, you must sequentially enable load converters only after bus voltage is stable.

You must also carefully choose your board-level converters' overcurrent protection. For example, in a battery-backed system, the constant-power nature of the load could trap a brick-wall-limiting converter at a point beyond the knee of its overcurrent characteristic.

Opinions differ about the relative prevalence of low- and high-voltage intermediate-bus distributed-power systems. According to Ericsson, most distributed-power systems have bus voltages below the SELV limit. Vicor, on the other hand, sees a 50:50 distribution between 48 and 300V systems.

You can opt for isolated and nonisolated dc/dc converters. Isolated converters are more expensive but are also safer, and they reduce problems with system noise, ground loops, and interaction between outputs. In addition to operating from either polarity of input voltage, isolated converters permit flexible system grounding.

In a centralized-power system or a distributed-power system using nonisolated converters, the common of the power-distribution system is also signal common. The common of the power-distribution bus is isolated from signal commons in distributed-power systems using isolated dc/dc converters.

No engineer runs dc/dc converters continuously at their rated full load. Ericsson reports that most designers allow margins of 15 to 40%. The penalty for under-margining is obviously more extreme than that for over-margining. The power-supply margins are easier to determine in a distributed-power system.

Constant vs variable frequency

Each converter manufacturer has its own circuit topology. Some employ constant-frequency converters that use PWM for voltage control. Vicor uses a variable-frequency resonant scheme. According to Vicor, the efficiency of PWM converters is usually lower than that of similar-capacity resonant converters. Vicor also notes that a PWM converter's efficiency drops rapidly with load, culminating with high dissipation under output short circuit. The company also states that PWM converters emit difficult-to-filter conducted and radiated common-mode (Denka-plate), normal-mode, and radiated noise and that their output ripple increases with load.

PWM-converter makers, however, have been busy enhancing their designs. So your best guides are spec sheets and your own tests. However, make sure your quality-assurance staff is not using outmoded tests designed for linear supplies (see Ref 2 for proper setups). Test your candidate converters in a realistic circuit.

A converter's operating frequency is important, however, because it determines the time required to sense and respond to a change in load current. The converter's topology and circuit design set a limit for the amount of energy delivered to the load per converter operating cycle. A converter may take several operating cycles to meet a demand for dynamic current.

You can synchronize many converters, but AT&T questions this practice. The fear is that two units operating at nearly the same frequency will "beat" and produce extraneous emissions. But synchronizing makes emissions worse because it causes all the converters' emissions to add arithmetically. Without synchronization, the reflected currents add in rms fashion.

You must look very closely at efficiency. Small size combined with low efficiency spells disaster. The higher the efficiency, the higher the MTBF for both the converter and the system. Also, high efficiency extends backup-battery holdup time. Efficient converters permit the use of smaller heat sinks and quieter fans.

Converter efficiency is a family of curves, not a single figure. So, look at efficiency across both line and load variations. For safety's sake, also look at dissipation under short-circuit conditions.

Efficiency for dc/dc converters currently ranges from about 75 to 83%. At first glance, this small range may appear to be meaningless, but it is actually a very significant difference because converting differences in percentages into percentage differences is not intuitive. A 75%-efficient converter dissipates 60% more power at full load than does an 83%-efficient converter. Politicians take advantage of this weakness in human intuition when they call an increase in taxes from 5 to 6% a "1% increase" (when it's really a 20% increase).

Advertised power levels for dc/dc converters are often very optimistic. Don't neglect the fine print, which says that the converters need heat sinks to achieve their advertised performance. Also note at what ambient temperature the converter needs derating. Converters from different manufacturers exhibit a wide range of ambient-temperature operation. "Ambient temperature" means different things to different manufacturers. See Table 1 for temperature definitions.

Ericsson finds that distributed converters have 3 to 11 times the power density of the pc board they occupy. That is, converters are a concentrated source of heat. Ericsson recommends, therefore, for free convection the converter can occupy no more than about 2% of the pc board's area and that for forced convection, the converter can occupy no more that about 7% of the pc board's area.

Only converters at both extremes of the power range use conduction for cooling (Table 2). Mainframe-computer converters that supply hundreds or even thousands of amps use a recirculating coolant. At the other extreme are low-power converters, 10W or less, that conduct heat out through their leads.

Convection cooling is more difficult to model and analyze than is conduction cooling. Free-convection cooling is very simple and reliable. Also, convection cooling does not entail the acoustic noise, maintenance, cost, and degraded reliability that fans introduce. However, many systems require fans because forced convection can cool about four times the power per board compared with free convection.

Opinions differ on filtering. Both Ericsson and Vicor say that you do not need to use a filter at the input of the dc/dc converter if you have properly designed and executed your dc distribution and decoupling. Datel, on the other hand, says most of its customers want such filters. Datel adds that engineers are specifying IEC noise limits for dc/dc converters (but not, of course, the IEC test setup because the spec actually applies to ac-line noise).

Table 3—Available power vs geographical area
Area	Voltage	Current	Max VA	Max 48V (W)
North America	120	15	1800	1183.3
		20	2400	1560.2
	208	15	3120	2066.2
		20	4160	2737.4
United Kingdom	240	10	2400	1597.7
		13	3120	2072.5
Continental Europe	220	10	2200	1464.0
		16	3520	2329.9
Japan	100	15	1500	982.8
		20	2000	1292.9
	200	15	3000	1985.9
		20	4000	2630.4
(Courtesy AT&T)

Selecting ac/dc converters for distributed-power systems is much like selecting any ac-input supply. You need to decide if you want manually strapped or autoranging inputs for single- or 3-phase mains voltages. Even with power-factor correction, many systems are already drawing the maximum amount of current allowable from a single-phase connection (Table 3). Power-factor-correcting ac/dc converters are also becoming more common as regulatory agencies tighten up on conducted noise. One more hint: Look for ac/dc supplies that require no preload.

Problems with distributed power

One major problem with distributed power is that the most frequently promoted spec for dc/dc converters, power density, is also the most useless. Fantastic power densities tend to wilt after you expose them to the harsh glare of your application's environment. You can compare power densities of different makers' dc/dc converters only after taking into account heat sinks, derating, and other design considerations. Fully configured, some high-density dc/dc converters are large and heavy enough to damage their host pc board during shipping and usage.

Next, the switch to a 3.3V digital standard is not as easy as just swapping out converter modules. If you are to take the JEDEC standard seriously, its ±0.1V tolerance means that a µP drawing 4A could have no more than 0.025V trace resistance between itself and its dc/dc converter. In other words, a fine trace or a connector can put you out of spec. Also, noise currents in the system ground can quickly eat up 3.3V noise margins. If you have a mixed-voltage board, carefully check your margins for the worst-case supply condition: 3.3V supplies at their high end (3.4 or 3.6V) and 5V supplies at their low end (4.75 or 4.5V). You might be in for a nasty surprise if you are interfacing 3.3 and 5V ICs.

Although JEDEC has promulgated a 3.3V standard, little actual conformance exists in the industry. Various manufacturers are going ahead with low-voltage standards other than 3.3V. Semtech notes that on mixed-voltage boards, you might have to be very careful how you sequence your supplies up and down to avoid failures. The company also notes that transient-voltage-protection devices for 3.3V circuits are rare.

At the new, lower voltages, large current surges can occur on the pc board itself. So-called "green" PCs (which switch large digital devices on or off as needed) and low-voltage disk drives are two possible sources of such surges. These surges may necessitate remote sensing for board-mounted converters, reintroducing a problem that distributed power supposedly eliminates (Fig 4). Further, the digital ICs themselves may be drawing pulsed currents at a high enough frequency that the skin effect may come into play in their power and ground lines.

1. "The Power Book," Ericsson Components AB, K3(93025) A-Ue, Stockholm, Sweden.

2. Hageman, Steven C, "DC/DC Converter Application Notes," Calex, Concord, CA.

3. "Applying DC/DC Converters," Conversion Devices Inc, Brockton, MA.