Detecting early signs of failure in switch mode power supplies

[angust_uk]

I have just experienced a few power supply failures (we use a lot of enclosed AC/DC supplies in the 100W - 500W range) and this got me wondering whether it is possible to spot the early signs of failure to be able to proactively swap a nearly-failed supply out before it goes bang (or in this case, simply stops).

My first throught is to measure the ripple on the output and also look at any changes in its AC current consumption (e.g. high frequency components which may be a sign of input capacitor failure?).

I appreciate there are many failure modes when it comes to PSUs, but most seem to fall into just a few categories (failed electrolytics, rectifier diodes and MOSFETs/drivers for example), so it would be great to be able to spot any degradation from outside the supply if at all possible (i.e. I don't really want to have to modify any supplies).

Any thoughts on whether this is a) feasible, or b) worth doing at all?

[tom66]

It depends on the failure, of course.  However, I would say it's mostly not feasible.

If the semiconductors are going bang the answer is pretty much no.  There is a very slight reduction in efficiency as the main transistors get too hot, but the difference is minimal until the moment of catastrophic failure.

For the output capacitors, you can likely measure ripple at the main switching frequency.  An increase in ripple implies increasing ESR and/or decreasing capacitance.  For the input capacitors, you'll see double your mains frequency (100/120Hz), though these caps going bad can also cause the PFC circuit to self-destruct.   The 100/120Hz primary side ripple will likely only get really bad under high load, whereas the output ripple will likely be present under light load. 

It's worth noting that while bad caps were a frequent issue 10 years ago, nowadays it seems like they are rarely the cause of a failure.  Most power supply failures tend to be "bang and gone", or rare issues you couldn't spot beforehand (control IC suddenly breaking down, for instance).

[coromonadalix]

And the brands of your "failures"  .... would be ?

In most cases not feasible, you would need intelligence inside of them to monitors lots of factors and have some warnings issued

Load(s) applied to them
Thermals
Over Current(s)
Over Voltage(s)

Many thing(s) to considerate

[shakalnokturn]

Comparing the in / out voltages, currents, startup times against a new model could tell you quite a bit about the state of capacitors in general.

Thermal camera and spectrum analysis may tell you couple more things without opening the box, the difficulty being having similar test conditions from one measurement to another (room temperature, surrounding EMI).

Start-up supply resistors going open, aging solders and I.C's failing internally are going to be tough to predict unless you have very well known PSU's with a history of the MTBF and failure modes.

I wouldn't say it's worth the bother... Seems easier to have a backup and repair the defective PSU once it has failed.

[mikeselectricstuff]

IME the most common failure is teh startup supply, and you only find that out when you power down and it won't power up again.
The best thing you can do as prevention is ensure the supply runs as cool as possible, e.g. by under-runnning and ensuring ample ventilation

[elecdonia]

In my experience the most common faults are caused by small electrolytics drying out and developing high ESR. The first observable symptom is often “failure to power up” when the ambient temperature is 10 degrees (or more) below normal room temperature.

Electrolytic capacitor ESR increases at low temperatures and drops considerably as the capacitor warms up. I do not know why this occurs but it is consistent. This temperature sensitivity becomes more extreme when the capacitor already has “higher than normal” ESR.

So, when evaluating the condition of small “wall-wart” SMPS power supplies I cool them in the refrigerator and then check whether they can still power up when cold.

At the moment I suspect one of my original Apple iPad power supplies is failing because it won’t start charging my iPad when the ambient temperature is <65 degrees F. However it always works perfectly at room temperatures >75 degrees F.

[elecdonia]

The best thing you can do as prevention is ensure the supply runs as cool as possible, e.g. by under-runnning and ensuring ample ventilation

One “rule of thumb” is that reducing the ambient temperature by ~10 degrees C will double the life expectancy of temperature-sensitive components.

[f4eru]

How to detect failures very early -> simply look for this mark, it indicates high potential of early failure:)
 
(Joking, of course)

[coromonadalix]

     

[Kim Christensen]

The easiest thing to do would be to just swap them out after so many hours of run time.
For more mission critical systems, redundant supplies could be a solution for increased reliability.

[MathWizard]

Sadly just to take apart a PSU, and desolder parts to check, takes so long, that only us hobbyists bother to do it. With the price of replacements and labour, it's usually cheaper to get a new 1.

I have a few old PSU I fixed, and it was fun learning stuff from doing that. I have a few more to try and fix. But IDK what to do with them after that, they are getting too old to sell.

[Mike_2012]

I agree with elecdonia,
however I had recently reviewed lots of HP PSUs "HSTNS-PDxx" (12V) ranging from 460W, 800W, 1200W and 1600W and it seems that HP is trying to fit as many components as possible in the same frame (and even increasing from one fan to two fans - placed side by side and acting 'serially' to increase air flow) - leaving almost no space for all that air to pass through the case
So I guess cooling is all but not very successful design - maybe the idea is that they are designed to work limited time only ...

[coppercone2]