Testing power supply

[PFF]

Hey Guys, Got a question about a switching supply I'm testing. The environment it will be used in is typically at 158 degrees F. So I'm testing it in an oven with a load on it to pull max power non stop. How long do I need to test it to determine if these will work for our needs? I could run it around the clock if need be. Is there any standard for this? Thanks

 

[IRstuff]

Well, what's the objective in doing this test? What are you trying to determine?

TTFN
faq731-376

 

[VEBill]

"The nice thing about standards is that there are so many from which to choose."

MIL-STD-810G, Method 501.5, Procedure II (Operational), 55ºC for 2 hours after temperature stabilization.

In other words, monitor the temperature of the Unut Under Test somewhere sensible (perhaps on the hotest heat sink?), and after it stabilizes then give it at least two more hours.

 

[IRstuff]

Again, it depends on the objective. We've experienced rather mindless operational temperature testing, and it turned out that while the system operated nicely at 55ºC, it didn't operate at 40°C. Test performed, but proved nothing.

TTFN
faq731-376

 

[PFF]

Interesting. objective is endurance. Have had several units fail last year. Wish I had one to look at but I don't. These power supplies run the control circuitry in a hot environment. The area where the PS units are mounted is a bit hot...around 160 F. Just want to test to make sure these PS don't fail. Would like them to last a very long time but I'm not at all sure how to determine this. I'd like to be able to give a good guess on how long we can expect them to perform as specified. How would you extrapolate out from a 2 hour test?

 

[VEBill]

Google: MIL-HDBK-217

It'll lead into the topic.

Temperature is typically an input to the calculations. Keep in mind the true meaning of words such as 'prediction' and 'estimate'. The real world MTBF can easily be in a different order of magnitude than the calculation (sometimes for good reasons not included in the model).

 

[OperaHouse]

We always think of semiconductors failing because of heat. A customer was complaining of one of our units that would fail almost every two years of continious 24 hour operation. This was a product designed way before my time and the part was an electrolytic. I duplicated the environment ad did temperature readings on the part. Plugged the numbers into the manufacturers life equasion and it came out almost exactly two years. Perty much all my home entertainment electronics was free due to poor selection of capacitors. A lot of designers don't realize that how limited the life is on standard electrolutics, it is not years and years like many think. Those that do are often playing a numbers game counting on the customer not having the device on that long or often so they can save a few cents. Do a like calculation. You will be shocked. You will buy a better cap.

 

[IRstuff]

"We always think of semiconductors failing because of heat"

That is still correct, however, which is why the Arhenius equation is so widely used for electronics. Also in answer to the OP's question, you need to perform an accelerated life test, assuming you want to spend the money. You'd need to run at least two batches of parts at different temperatures to failure, and the failure rates as a function of temperature can then be determined, as well as identifying which components are the weakest. There are a series of military standards that describe how life testing is performed.

160ºF = 71ºC, which is already at the upper limit of commercial component temperature ratings; when you factor in the thermal resistance of the components within the supply to that ambient temperature, you could be easily at 85ºC. which is the temperature limit for many industrial temperature range components. Given that, you definitely need to run a few supplies at a minimum of 85ºC ambient, with instrumentation on the components to determine what the induced case temperatures are. I suggest the higher ambient temperature because it's almost invariably true that a "measured" ambient temperature is taken where it's convenient, and may not reflect the true thermal environment. We once ran a life test in an oven with 125ºC ambient temperatures, but the test engineer didn't do any thermal calculations to predict that the 1.5 kW being dissipated by the components might raise the local ambient temperatures above 180ºC, resulting in junction temperatures in excess of 200ºC. Naturally, we got failures almost immediately.

As far as the duration is concerned, therefore, if the part is truly that susceptible, you might try three supplies at 70ºC, 85ºC, and 100ºC, and run them to failure, since that's the actual objective, to determine the weakest component(s) within the power supply. This may become and iterative process, as you may have multiple weak links. As a first step, though, you may need to revisit the parts list of the supply and determine whether all the components are rated for at least 85ºC ambient. Note, also, your supplies may not necessarily have permanent failures,; they may just be operating outside of the design conditions and may regain operability when cooled down.

TTFN
faq731-376

 

[PFF]

IRStuff, this supply is for an oven. In the test fixture(an oven) I placed the power supply with a thermocouple attached and a second thermocouple inside to measure the surrounding temperature. The PS does get quite a bit hotter than the surrounding area. So far it's run okay. Haven't been able to make it fail. Suppose I could keep increasing the load and/or temperature until it fails however the particular oven I'm using only goes as high as 180F. What's the best way to proceed? Suppose I get it to fail running at 120% of it's rated output power after 2 hours of use. That's only a sample of one. Would you use this data as parameters for a weibull distribution and then calculate the normal usage life?

 

[VEBill]

I'd examine the possibility of remoting the power supply to a cooler location. Or adding simple features (insulation, air ducts) to the present location to drop the temperature.

Perhaps these approaches are not possible.