11
\$\begingroup\$

There are many articles on the internet about life expectancy of switching power supplies, but most of them have discussed about the situations that the PSU is at near full load. I'm designing an electronic device that needs a standby state, I'm using a COTS PSU with an output of 24V-5A, something like this:

PSU

My device power consumption is nearly like this:

5% of times at near full load.

55% of time at 20% of full load.

40% of time at standby (about 1% of full load).

I know that everything is depend on the quality of PSU and it's design specs, but Chinese suppliers doesn't provide enough and reliable information about their products. Although we are buying them from a sort of responsible supplier.

Can I generally expect of 10 years life of my PSU?

Someone says that the inputs capacitors (at high ac voltage) or the optocouplers regardless of the load consumption will destroy after some years. Is it true? If yes some other electronic systems like TVs that are on ac power for many years how had solved this problem?

\$\endgroup\$
12
  • \$\begingroup\$ Yes, your input capacitors will degrade over time regardless of load. No, that's usually not a problem in the first 10 years. \$\endgroup\$
    – Mast
    Commented May 1, 2017 at 6:41
  • \$\begingroup\$ What's your ambient temperature? \$\endgroup\$
    – winny
    Commented May 1, 2017 at 6:53
  • \$\begingroup\$ +1, This is a very good question! I've seen Murata PSU quoted 575k hrs of MTBF, which is about 65 years!. TDK RWS50 power supply lists the MTBF at 4,170,949 hours !!!, which is about 500 years. Yet they contain mechanical fans that usually have at most 40,000 hrs, and best electrolytic caps are rated at 100,000 hours at 60C. Although the life expectancy doubles with every 10c drop in operating temperature, I still can't figure out the 4M hrs rating. I am all ears... \$\endgroup\$ Commented May 1, 2017 at 6:55
  • 1
    \$\begingroup\$ @MarcusMüller: Calculating the true value of MTBF based on a references like MIL-HDBK-217F needs exact information on all electronic parts and some extra information like their supplier and environment conditions. I know that many electronic parts that are used in these devices doesn't have any reliable documents. For example look at this link: link , they said 316.2K hrs for MTBF (= 36 years), you believe this? \$\endgroup\$
    – electro
    Commented May 1, 2017 at 9:00
  • 1
    \$\begingroup\$ @AliChen MTBF != life expectancy. A battery may have one hour lifetime before depleated but 10 million hours MTBF or more is common since batteries are very reliable. \$\endgroup\$
    – winny
    Commented May 1, 2017 at 17:35

3 Answers 3

7
\$\begingroup\$

You can't tell the expected life time of any device from its looks.

In your specific case, there's a few things that I'd expect would be the most likely to fail:

  1. defects due to surges on the power source
  2. aging of capacitors
  3. semiconductors due to heat due to dust on the cooler

The first one is totally out of your hands, usually. The question how bad a grid malfunction must be to fry your SMPS depends on the design of your SMPS, and, simple luck. Still, from a pure stochastic point of view, you must expect anything to go wrong once in a while (lightning strikes, car crashing into power line mast) – but that while might be half a century, or half a month, depending, again, on odds totally out of the influence of the SMPS.

Of course, you can limit the effect of things going wrong by sufficient input protection – but that's something that will only be spec'ed for larger, much more expensive supplies than yours.

Then, we have aging and failing probabilities. In an industrial environment, you'll usually find a measure called "MTBF", mean time between failures, that captures exactly that. That number should be higher than what you need it to be, under a heavy-load usage scenario. It would not make that much sense to estimate MTBF under lighter load – a) could you have used a smaller SMPS then, and b) it's simply pretty hard to guess whether things would last longer; I don't really see why a fixed-frequency SMPS would have anything be less stressed in a 1% load vs a 90% load scenario, iff cooling is adequate.

So, if you need that reliability, buy something that comes with an MTBF. That number is pretty hard to honestly and qualifiedly estimate, and only large suppliers with significant experience should be trusted on that.

Although we are buying them from a sort of responsible supplier.

Lol, OK, so, not that supplier, but a fundamentally responsible one. Yes, there will be a cost factor of > 2.

You usually don't buy such components for consumer electronics; the price/benefit trade-off simply doesn't allow that:

If yes some other electronic systems like TVs that are on ac power for many years how had solved this problem?

Not at all? Ok, of course, as engineer, you try to build the best SPMS you can with the things given to you. You correctly pick your components with some headroom in their specs so that of the thousands of devices that will be build, the outage probability is low. But if a friggin' 400€ TV fails in the field after a decade … no one cares. That's about the upper end of life expectancy of consumer electronics anyway. If it fails after 5 years, you don't even infer a bad reputation. If it fails after 2 years, well, look that these cases don't accumulate, and if they don't, you're fine as manufacturer; if it fails within warranty, still cheaper to exchange 1 in 1000 TVs than to build 1000 TVs that are 1% more expensive, but fail years later on average.

Luckily, economics might be on your side here: while any solidly made SMPS will be more expensive than your "sort of responsible" supplier, if you can get access to supplies for mass produced higher-end devices, you'll benefit from the fact that rigorous testing has been done on these, but the mass production distributes the cost of these among more devices, and they still are relatively cheap.

The comments already mention TDK and Murate as typical SMPS OEM, but I'd simply go and open a device that has a similar usage profile as you have – in your case, look for a broken high-end office printers/copiers, and then for the names on the power supplies.

\$\endgroup\$
4
  • \$\begingroup\$ "I don't really see why a fixed-frequency SMPS would have anything be less stressed in a 1% load vs a 90% load scenario, iff cooling is adequate" : Of course my SMPS cooling is not that enough, obviously it heated more at 90% load, so surely it will affect on electrolyte caps and other parts lifetime. About the failure rate, I'm really OK with that 1 per 1000 ones. But how could I sure about that rate? The SMPS supplier is a mass producer and they surely test their products, plus we will test them, but I think you could not estimate lifetime by these tests. \$\endgroup\$
    – electro
    Commented May 1, 2017 at 7:41
  • \$\begingroup\$ My question is more fundamentally about SMPS parts especially input caps, that are they degrade or destroy while they are in ac plug but the SMPS is not at full load? And do they age in average 10 years in this situation? \$\endgroup\$
    – electro
    Commented May 1, 2017 at 7:44
  • \$\begingroup\$ @electro as said, this is impossible to tell without knowing how the SMPS was designed and what caps were chosen. You can buy oversized or more rigidly tested caps for money – but whether you spend that money or not is up to the manufacturer of the SMPS. No general answer can be given. \$\endgroup\$ Commented May 1, 2017 at 8:04
  • \$\begingroup\$ Well, the power supplies' heating is usually much higher at 90% than at 80% and extremely high at 100%. So considering heating, it's usually good not to put more than 80% load on the supply for longer periods. \$\endgroup\$
    – yo'
    Commented May 1, 2017 at 17:47
2
\$\begingroup\$

Life expectancy is not really equal to MTBF!

MTBF is computed using a probability function distribution for electronic component (usually the exponential distribution function is used). Life expectancy is a parameter which guaranties that your assumption about probability distribution function still stands.

There is establised standard like FIDES that allows estimation of PSUs device (or COTS in general). What you need to define is your application profil (ambiant temperature, ON Time and OFF time ratio, etc.).

You can make a quick tour on FIDES website to know more.

\$\endgroup\$
0
\$\begingroup\$

Simple temperature cycling destroys electronic devices, because the expansion coefficients are not identical and the constant mechanical flexing and constant change in stresses..........eventually breaks the components.

Cure: do not turn off the circuits.

Oh...... keep them cool.

One reason for lead in solder.....its mechanical compliance (ductility).

EDIT Consider a SMPS with 100,000Hz clock........100,000 times a second the thermal expansion and contraction occur.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.