Many and probably most power-supply failures are easily preventable. They are most frequently the result of overstressing the supply with heat (either ambient or self-generated), transients or overloading. If you're a power-supply designer, many of these causes may be obvious to you. But don't assume they are obvious to your customers.
Or, if you're specifying power supplies for use in your systems, be aware of the not-so-obvious pitfalls such as misleading reliability specs or low-cost power supplies with specifications that may be too good to be true. By following the guidelines listed here, most power-supply failures can be prevented.
Be sure that air can circulate freely around and through the supply (if necessary, use a fan), and don't block the vented surfaces of enclosed supplies. If the input power is likely to contain large transients — for example, from motors and high currents being switched — use transient suppressors on the troublemakers when possible and put an input filter on the supply.
Continuously drawing more current than that for which the supply is rated will likely result in an early failure. This is particularly true when the environment is hot and the manufacturer's derated specification for that temperature is not heeded.
Frequently, supposed “failures” are actually misapplication problems. Be certain the power supply can handle not only the steady-state operating current, but also the maximum surge current the load may draw. Most power supplies incorporate electronic current limiting on their outputs and may simply “lock up” when the load tries to draw a substantial surge. This is especially true with high capacitance loads and nonlinear loads such as incandescent lights.
Another simple consideration is your input source. Are you certain the actual ac input voltage is always within the specified range? Keep in mind that a supply with a 105-Vac to 125-Vac input range most likely won't work properly (or at all) in Japan, where the nominal is 100 Vac.
As with most other aspects of life, you get what you pay for. You won't get the highest-reliability supply for the lowest price. Relatively low-priced power supplies usually have components stressed to much higher levels than higher-priced supplies, so they have less operating margin and are likely to fail more quickly.
Heatsinking also may be skimpy, resulting in semiconductors running hot. Note that many low-priced units have a warranty period of less than a year, which says something about their manufacturers' confidence in their product.
And be realistic when it comes to your expectations of reliability. Mean-time-between-failure (MTBF) calculations are often based solely on parts counts, without taking into account the effects of temperature and other stresses. So don't place a great deal of confidence in those ratings.
For example, how can the MTBF of a power supply be 500,000 hours when the electrolytic capacitors in it typically degrade in less than 10 years, which is only 88,000 hours? If you must have a power-supply function for many years without risk of failure, consider redundancy. The outputs of two power supplies, each capable of individually supporting the load, can be connected in parallel (through diodes, to prevent interaction) so that if one drops out, the other will continue to power the load without interruption while the failed supply is being repaired or replaced.
Even when the individual power supplies each have a modest MTBF rating, redundancy can raise the effective MTBF of the power system to astronomical levels. Of course it's important to continuously monitor the outputs of both, so that a failure doesn't go undetected.
Tom Skopal has been at Acopian Technical Company for almost 40 years and has written numerous articles for electronics publications. He received a BSE degree from Drexel University and has done extensive graduate work at both Drexel University and Seton Hall. He was the recipient of an Idea of the Year award from Electronic Design magazine.