Specifying the optimal
redundant power system

BY EVAN VOGEL
and RON DELUCA
Todd Products
Brentwood, NY

Redundant power systems continue to proliferate in response to the need for full-time availability of data. While no magic formula exists for defining the best redundant power solution, paying attention to a few essential elements will simplify the task. The key to specifying a redundant power supply system is to not overspecify. Instead, the designer, together with the power supply vendor, should follow several steps. Redundant system specifications that fail to consider these items often impose unnecessary costs for power that compromise the competitiveness of the system as a whole.

Redundant power systems often use hot-plug power supplies,
such as the Todd Rack System.

1. Determine both the maximum and typical power of the system. The decision that most affects the cost of power is the rating of the incremental basic power building block. For instance, a 1,500-W system with redundancy can be built from two 1,500-W or four 500-W supplies. The first system has 1,500 W for redundancy, while the second has 500 W. This does not necessarily mean that the latter costs less. The table compares three options for configuring a 1,500-W redundant system.

Dollars per watt are inversely proportional to power and each power supply manufacturer has a different price structure for a given power level. The best way to take advantage of cost-saving opportunities from multiple vendors is to specify the overall system requirement and not the incremental power level. Then, let the power supply manufacturer provide their best solution for the system.

2. Determine how the power can best be scaled. Many redundant systems can be tailored to specific customer requirements, with power as a function of the feature mix. In such instances, the minimum system configuration often cannot handle maximum system power without becoming prohibitively costly. At the same time, a maximum system configuration becomes too expensive if it comprises merely a large number of minimal system power supplies.

3. Agree to let the power supply manufacturer determine the best power increments and the ideal scalability for the system. The designer should provide a forecast of the most common three or four power configurations. Once again, the designer does not need to specify the incremental power level–instead letting the power supply manufacturer weigh the tradeoffs of volume and power.

This may be accomplished by providing a standard power supply solution, or a slightly modified standard product. In some cases where the package requirement is non-typical, a repackaging of the standard product is desirable. Not all power supply manufacturers can provide a non-standard solution.

4. Reach an understanding of how the power increments and the choice of the power supply will affect system reliability. When creating a fault-tolerant design, reliability must be considered. One might determine the failure rate of each power supply by using MIL-STD-217 or by the Bellcore method using the traditional approach. Both MIL-STD-217 and Bellcore are used to predict the frequency of failures over time. This is accomplished by computing the sum of each device failure-rate prediction multiplied by the environmental factor.

Bellcore and MIL-STD-217 differ by the failure rates predicted per grade of component. MIL-STD-217 is more suited for MIL-quality components. When commercial-grade components are used, the Bellcore method will predict a longer product life cycle.

All other things being equal, as component count goes up, the MTBF goes down. Therefore, more power supplies translates into more components and a lower MTBF. Remember that MTBF is a statistical failure rate prediction that does not account for factors related to failures caused by vendor processes such as design, component selection, assembly, and test and burn-in.

5. Determine the power system's remote monitoring and reporting requirements. Power supplies often require on-board status and monitoring of overvoltage, undervoltage, remote enable, current sharing, current limit, inverter good, shutdown, and overtemperature. To make it easy for service technicians to rapidly identify a failed power supply, redundant power systems often incorporate LED indicators and alarms.

There has been an increased need for power supplies that transmit serial numbers and actual output parameters over digital buses to provide real-time power supply status measurements at remote locations. This allows adequate time for repair, service, and preventive measures to maintain constant system operation. Often a single summary fault is sufficient to trigger a field service action. Because monitoring adds cost to a system, consider this feature carefully when defining the redundancy needs.

6. Select the type of connectors that will be used. Many redundant power systems are now being designed with hot-plug power supplies, such as the Todd Rack System (see photo). In these systems, a wide range of connectors can be used from companies such as Positronics, AMP, Elcon, Erni, and Amphenol. Many of these connectors come in pc-board-mount and wire-clamp styles for cable mounting. This feature allows for greater flexibility for the power supply to-system interface.

Key features for connectors include keying, current capability, engagement force, size, and cost. Another is the surface plating on the contacts, as this determines the contact resistance and the number of engagements specified. Most of these connectors come with options for pin length.

7. Agree how current sharing will be accomplished. The three current-sharing methods used in most redundant power systems are brute force paralleling, third-wire current sharing, and output slope program current share (sometimes referred to as droop).

/ Brute force paralleling is the simplest form of current sharing. With brute force, two similar units are wired in parallel and their output voltages are adjusted equally. The differential output voltage, as well as drops in the interconnections, time, and temperature, determine the degree to which the current is shared.

/ Third-wire current sharing utilizes active circuitry within each power supply to monitor its output current and develop a voltage proportional to current. The third-wire connection distributes this voltage to each power supply. The circuitry will modify the output voltages of each supply, thus balancing the output currents across all paralleled supplies. Current sharing of 3% to 5% can be achieved. Care must be taken to ensure that the third-wire connection does not become a single-point failure in the system. This would bring the whole power section down and negate redundancy.

/ Output slope program is perhaps the most elegant of all current-sharing approaches. With it, each power supply follows a preprogrammed output voltage change as a function of its output current. When similar power supplies are paralleled, the output current will change proportionally to its output voltage. Since each power supply is preprogrammed to follow a voltage slope that decreases with increases in current, current sharing is obtained. There are no additional connections between power supplies. With the output slope program technique, current sharing of 3% to 5% can be obtained with load regulation around 2% to 3%.