POWER SUPPLY ENDURANCE TEST

[PFF]

We have a small power supply (45 watts, 3.75A at 12VDC) want to know how long it will last. Any idea what would be the quickest way to determine this...not withstanding the manufactures word on it. Could load it down and run the heck out of it. Heat it up etc. Ideas?

Thanks

 

[MikeHalloran]

Sure; that's called accelerated life testing. You can get a pretty good argument going over how it actually correlates to real time expectations, but it's better than nothing. The first problem you run into is figuring out what to change, and how much, and that's when the first argument starts.

Probably the fastest way to verify or refute the manufacturer's assertions is to engage an expert to reverse engineer and analyze the power supply and estimate its service life based on known data about each component and the way they interact with each other.

... which does not mean that it's actually a fast process.

The military uses it, because their stuff has to work.
The automotive industry uses it, because their stuff has to work, too.
( ... and the demonstrated improvement has driven down the fleet replacement rate so far as to produce a global capacity glut, but that's another discussion. )


For more, search on:
{"power supply" "reliability prediction"}
with the quotes, without the braces.





Mike Halloran
Pembroke Pines, FL, USA

 

[IRstuff]

I suggest you hire a consultant. Without some serious physics of failure and electronics reliability experience, there will be no way for you determine the degree of acceleration you create. Additionally, you will most likely need to kill quite a few power supplies to gather the data.

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529

 

[VEBill]

MIL-HDBK-217. PDF copies online. Commercial software available.

 

[OperaHouse]

We built a product that failed all the time. A new customer asked us to do life expectancy calculations. One of the other engineers used a MIL calulation and came up with 120 years. Everyone was happy. I don't think they had a fill in number doe bad design.

 

[VEBill]

The nice thing (_{not really}) about the mil handbook reliability calculations is when the calculated MTBF is something like "106835873.4787 hours" (*) and they include this exact figure, all 13 (in)significant figures, in the final report. And then, nine months later, all the early deliveries fail almost immediately and have to be sent back for rework.

(* 12,000 years to the nearest one-third of a second. The precise number has been changed to protect the guilty.)

 

[IRstuff]

Note also, though, MIL-STD-217 predictions, when the hardware is working correctly, are very pessimistic, due to some issues in the modeling applied, and the conditions applied. For example, there are only gross categories, like ground mobile, which could be anything from a golf cart on asphalt to a cross-country tank trek.

TTFN
faq731-376

Need help writing a question or understanding a reply? forum1529

 

[DHambley]

VE1BLL,
Hah, I'm laughing while reading your post because I see that silly stuff quite frequently. Program managers who don't understand mathematics enough to know what 13 (in)signifigant figures means are running a lot of shows in aerospace. When a nice report is created because the PM thinks that is a more important delivery item than the product actually working, then you end up with the situation desicribed in Operahouse's post right before yours.