Meet the demand for high-availability power systems at lower power levels

Redundancy in a power system means that an alternative power source is automatically accessed if all, or part, of the main system fails. Redundant power architectures have been employed in practically all large telecom, datacom, networking and IT infrastructure for many years. In fact, they’re found anywhere where the system up-time is a determinant of revenue. Now, there is increasing demand for power supply redundancy in many smaller systems, even where applications are not mission-critical. In some cases it might be over-engineering with respect to the true needs of the systems but marketers often see the provision of redundancy in power systems as a way of differentiating their products from those of competitors. This trend has made its way into all types of relatively low-end communications/enterprise equipment: small business network security gear, LAN and wireless LAN switches, session border controllers, SAN servers, and other applications that may previously have relied upon a single power source.

Using diodes to connect power supplies in parallel
Diodes are used in the output lines of power supplies in redundant systems so that if one power supply fails the other will continue to operate without the ‘dead’ power supply pulling down the output rail. Diodes should always be rated higher than the power supply output current limit.

Adding diodes in the output lines of a power supply causes degradation of the output regulation, due to the voltage drop across the diode at different current levels, and reduced system efficiency. This needs to be considered when using a redundant system as a solution, as the load must be able to accept the poorer regulation. To get around this problem it is possible to use the remote sense function and connect it after the diode. When doing this, better transient response can be achieved if a current share connection is available and utilized. This will allow the power supply to compensate for the diode voltage drop.

Ideally, all units within a redundant power system should share the load equally when they are operating to minimize stress and extend their lives. In most applications, droop-sharing (also known as the “programmed slope” method) is sufficient, provided that the initial output voltage set-points of the redundant units are very precisely set (roughly to within 0.5%). Active or “forced” current sharing, which involves a dedicated current share control connection between redundant supplies, is often requested because it is often perceived as ensuring better current sharing accuracy. However, droop sharing can be shown to be as effective as active sharing with the right set-point control and it does not introduce a possible single point of failure.