原创 Measuring power supply efficiency: Not that easy

Of course, talking about it is one thing, but measuring it and doing it properly is another. In the "old days" when efficiencies were in the range of 40 to 80%, a measurement error of a few percentage points might not be a serious problem. As efficiencies are, however, now in the 85% to 90%+ zone and every percentage point increase is judged as critical, any inadvertent or unintentional errors in the test set-up or execution can be a major concern.

A recent blog post by Josh Mandelcorn at Texas Instruments' Power House site clearly showed how easy it is to allow apparently inconsequential or overlooked issues to seriously affect the validity of the results. He points out that the potential errors in the final results start with a basic fact: there are four simultaneous readings (input and output current and voltage) from four meters, and they each have errors which can conveniently cancel out, or just as easily add up.

In addition, the effect of temperature on wire and connector resistivity—not an issue with very low-power supplies, of course—can easily cause other small errors as well. Furthermore, there are issues associated with forced cooling versus convection cooling, thermal time constants for the supply's circuitry, and other factors that can affect the legitimacy of the numbers, even if they are measured with high accuracy.

It's an old story: making a fairly good measurement is often easy; making a really good one is much harder. As most engineers know, the difference between the two can become a serious problem. Even if the results are tagged as "preliminary, subject to review," they get repeated in reports and often develop a life and credibility of their own, at least in management's mind.

Pretty soon, the numbers become accepted as solid fact rather than as tentative subject to verification; for supplies, the rough number will soon become the lower-bound threshold of presumed efficiency. It’s tough to go back to the boss and say "you know that 94% number I gave you last week? —after careful review, it's really only 90%."

Have you have to make power supply efficiency measurements? What problems did you encounter, or not foresee? Did they come back and "bite you" when you had to re-visit them?