Paralleling voltage sources (also called power supplies, but there are also current-source power supplies) is done for a variety of reasons:

  • you just found out that your design needs may need more current than originally planned for, and paralleling a supply seems like a good way to have some just-in-case "insurance;"
  • it gives you the option of implementing a redundant or N+1 topology;
  • it allows you to use the same supply as a single unit or in multiples across various products, thus simplifying the BOM, inventory, and other practical issues;
  • it lets you work with one make/model of supply and get to know its inherent characteristics and idiosyncrasies (and they all have some!).

Unfortunately, paralleling supplies is not as simple as the term suggests. Strange and often unpleasant things can happen if you just connect the outputs "as is." Generally speaking, you can’t just connect the outputs of two supplies together unless the supplies have been specifically designed for such operation. Problems include output voltage inaccuracy, poor regulation, output droop, all the way to possible damage if the "stronger" supply tries to force current into a weaker one (and no two supplies–even the same model–are identical).

Even using remote sensing to maintain the output voltage won’t help in many cases. While there are ways to allow many supplies which are not designed for paralleling to be used that way, such as by using a ballast resistor to equalize outputs, doing so brings new issues in regulation, supply efficiency, and resistor dissipation.

The good news is that there is help. A recent application note from supply-vendor CUI is very enlightening on this subject. "Current Sharing with Power Supplies" is an overview of the issues related to using supplies in parallel. It looks at what can go right, what can go wrong and to what extent, and what the designer can do to ensure the "going right" outcome is the one that happens. (It is not a thinly disguised pitch for their products; other vendors have published similar application notes, but this one is fresh and well written.)

Ironically, the issue of paralleling is more difficult for switching supplies, which are generally, but not always, more effective than linear supplies. There are many LDOs that can easily be used in parallel, which can be a real plus when the localized IC or subsection of a circuit that the LDO is supplying needs just a little – or a lot – more current.

Have you ever paralleled supplies simplistically, without checking if they were designed for it or needed special considerations? How did the experience go?