Designers parallel their power supplies to increase system power output or to provide fault tolerance. Paralleling also improves system performance. When both power supplies are load sharing, for example, the two supplies share the dissipated heat, improving reliability. Furthermore, if one power supply goes offline, the other one is coming from an already loaded state, so the perturbation on the output is not significant.
The first requirement for fault tolerance is redundancy, i.e., at least one extra converter in the system. This is commonly called an N+M array, where N converters are required to satisfy the power requirements, and M additional modules provide redundancy. All modules in the array must be able to supply undisturbed power on failure of one module, in spite of the sudden increased load demanded of each. The same is true when an additional module is placed on line, suddenly reducing the current demanded of each module.
To satisfy these criteria, it’s essential for the individual converters to share the load current to minimize the dynamic response or recovery required of each. Current sharing is required to parallel converters for increasing power. Although current sharing is generally not required for 2N redundant converters, it does provide better system performance. Current sharing is required, though, for N+1 redundancy.
How To Do It
Different techniques are used to achieve both active and passive load sharing. In one passive load-sharing scheme, the droop-share method, the output voltage is allowed to droop or decrease as load is increased, forcing two power supplies to load share. That’s acceptable for a fault-tolerant system, because no single point of failure can compromise the system.
Having the units load share without an interconnect minimizes the potential points of failure. But load sharing and load regulation with a passive scheme can be compromised more readily than with an active communication scheme, where a single wire interconnects control signals.
The active single-wire connection (load-share port) between power converters is useful because it allows the power converters to communicate, improving the accuracy of load sharing, and accurate load sharing can maximize the power output of the converter array. If you’re simply paralleling converters to scale the power supply to a higher level, it’s desirable to get tight load sharing between the converters, so they can reach the maximum capability.
Some converters use a primary side communication bus. The most popular ones use an ac signal. The technique first brought to market was a daisy chain technique. The signal was passed from a master, or driver, to a slave, or booster. Then it was regenerated to another power booster, and next to another booster. It didn’t allow for fault tolerance, but it was an effective technique, and the fact that the power trains were matched allowed equal energy to be delivered through each power train. As a result, they would load share within 5%.
More development on the primary side communication technique made it more of a parallel bus structure that could be made redundant. Two buses were needed in case there was a fault on one of the buses. However, the advantage was that it used an ac signal, and the converters could be capacitively or transformer coupled to the parallel control bus.
This affords a level of fault tolerance relative to each converter. Since it is isolated through the capacitor (or transformer), it prevents a fault from taking down the common communication bus and affecting all the converters in the array. Thus, the other converters still communicate and load share. In addition, it is a high-frequency ac signal, so it allows those components to be small and easily implemented.
Many pulse-width modulation (PWM) bricks have a secondary-side load-share dc control for load-share purposes, basically another control loop that’s slower than the main control loop, so there could be issues from a stability standpoint. With recent advances in employing digital communication techniques, paralleling schemes will take advantage of the centralized control architecture (CCA), where things can be managed from a digital standpoint.
Through continuous monitoring of the power supply, that approach can provide even tighter load sharing. It also can determine the status of other operational parameters. That all plays into the ultimate in fault tolerance, i.e., being able to monitor the power supply and then letting the system know that a power supply is compromised and service is required.
What You Need
Load sharing and scalability are important elements in the implementation of the power system. They lend themselves not only to fault-tolerant approaches in designing electronic systems, but also in an OEM’s ability to meet market needs and maintain cost initiatives.
Scalability is a useful capability. Many manufacturers today use a plug-and-play approach to address the power-supply requirement, leading off with an entry-level system, sold at an attractive price. By using scalability, the system can be upgraded in terms of outputs or functionality. If the entry-level system is a telecom system, for example, it may be able to handle the basic requirements for a small business in many cases.
As the business grows, it may want to add more capability in its data lines or phone lines. So, it doesn’t make sense to install the maximum power supply and have the customer pay for it up front. It’s an incremental cost with each upgrade. Brick-style power converters are especially well suited to this approach, having always used scalability as a key feature, not an option.
However, it all comes down to the application. Everybody has different goals and different needs, e.g., simplicity, greater functionality, high power density, small size, light weight, fault tolerance, tight regulation, low noise, and on and on.
Related Articles
DC-DC Converter Architectures: What Are People Talking About?