Thermal study exposes PSU improvements

Article By : Michael Dunn

See how a thermal camera helps evaluate updates to a cheap PSU’s componentry.

In Whipping a cheap LED PSU into usable shape, I looked at a very low cost 12 V, 18 W PSU from China that had issues, to say the least. A bit of re-engineering converted it into a more usable supply, and now, with EDN’s thermal camera, we can get a better look at how the new parts affect thermal behaviour. 

The 12 V, 18 W PSU. Note the tape over some shiny components to help balance their emmisivity.

For these images, I ran the supplies at 70% rated load – about 1.1 A – until they reached thermal equilibrium. Ambient was 15°C (being able to enter a temperature offset would be really handy for displaying rise over ambient).

I measured three PSUs:

  • From original batch, with poor output capacitors
  • From second batch (AliExpress), which came with low-ESR output caps
  • Modified version of second unit, with Rubycon output cap and Schottky output rectifier

Based on my initial investigation, the components most suited to and needful of replacement were the 12 V output cap and the output rectifier, a UF5404 3A fast-recovery (50 ns) part. I replaced the rectifier with an SR540 5A Schottky. The differences are apparent in the images (in addition to the tape over a few components, there’s also a strip on the case, from top-left to bottom-right; the effect is small but noticeable).

Component and case thermal images, original batch

Component and case thermal images, second batch

Component and case thermal images, modified unit

Unfortunately, colour amongst the images can’t be directly compared because the camera app doesn’t allow for freezing the temperature scale (this feature is enabled when using the more expensive camera). Nonetheless, the results are intriguing, and not exactly as expected.

Some surprising readings

Looking at the components, the rectifier actually runs 5° hotter in the 2nd unit, with the better output cap. The Schottky is 32° cooler than the 2nd. Or, more usefully, it’s 33° above ambient, versus 65° for the 2nd.

The switching transistor & sink are eerily cool on unit #1, but show a 35° rise on the others.

The output capacitor temperature is not as bad as I remember from my initial testing. It seems very nonlinear with current. Well, I2R, eh. The Rubycon is at least 10° cooler than the factory caps.

The case temperatures also show some surprises. The 1st, supposedly worst unit, has the lowest case temp! It would likely become the hottest at 1.5 A, but at 70% load, it’s not bad. The hot spot is over the output components.

The 2nd PSU, with the better factory cap, has the highest hot spot, at 45° over ambient, and it has moved to the transistor area.

The modified PSU’s hot spot is also over the transistor, but 39° above ambient. Above the output components, it’s about 10° cooler than the 2nd unit.

Even given the modified PSUs’ improved performance, I’m using them at only 50% of their rated output. And it appears from this investigation that using the original unmodified batch at this level would be okay too. But yeah…these really shouldn’t be available to the general public.


Related articles:


Michael Dunn is Editor in Chief at EDN with several decades of electronic design experience in various areas.

Leave a comment