6
\$\begingroup\$

I am doing aging of my product by placing to a chamber for 2hrs 50degrees 95% humidity and 2 hours 0degrees 24/7, I have finished 1 week and the product is still okay. I am wondering what does this 1 week of under a cycle of hot and cold temperature translate to in terms of aging or normal product cycle? Presume that the consumer use it for 8hrs a day under room temp.

Thanks,

Jasp

\$\endgroup\$
  • 3
    \$\begingroup\$ I suspect you could get the answer from someone in your company (the boss who asked you to do this test, maybe?). If it's the case, you'd better ask him. But then, come back to us and answer your own question, because it's interesting. \$\endgroup\$ – dim Jun 24 '16 at 13:03
  • 3
    \$\begingroup\$ This is sort of like asking "if I have light on my garden for 3 hour on and three hours off, will it grow?" Well first of all, are there plants in your garden? what type of plants? Do you have water as well ? ... Determining failures rates is very difficult, it seems from your thermal cycling that this is a very ad hoc program. Why those values? what is your thermal ramp times? What does your FIT analysis tell you? ... What does this program tell you? that you can determine if you roughly have infant mortality issues and probably nothing else. \$\endgroup\$ – placeholder Jun 24 '16 at 16:08
  • \$\begingroup\$ This is usually to simulate the aging of the product this is some kind of stressing the system by placing to this cycle. I am wondering how the institution simulate the life span of a product. There should be a translation of this hours compare to normal operation. This is to accelerate the life of the product. \$\endgroup\$ – jasp Jun 27 '16 at 2:13
2
\$\begingroup\$

I see this as a statistical modelling issue. You need to determine what the real-world requirement is before attempting any kind of accelerated ageing testing. Then you need to determine the risk you are willing to take in relying on your testing. Testing one example of a component (sample size, N = 1) has absolutely no predictive value whatever. There are plenty of reference texts available to help you with sampling and accelerated testing. Tighten up your specification. Do the reading. Then come back here.

\$\endgroup\$
  • \$\begingroup\$ This is a topic I wish there were better Q&As about on this site. It's a very rabbit-hole-esque topic! \$\endgroup\$ – Daniel Jun 29 '16 at 2:33
1
\$\begingroup\$

First off no, you are not really doing anything to do with reliability. The best that these test can even come close to is "infant mortality" test in which you are winnowing out the device that might have failed early. Either because of marginal devices, damage during assembly and poor mechanical qualities of the soldering (from the thermal cycling).

You haven't given any details as to what the temperature ramp rates are, how many devices you are testing and other operational parameters.

You can't do reliability testing without some serious work a head of time, simply because of the expanse (in equipment, in time [ it may take months] and in plain old "figuring" it out).

You start with a FIT analysis (if you can get data even) this will apply to simple devices. Complex devices ... likely never.

You then have to study many devices to the point of failure at different temperatures. What is meant by failure? It very much depends upon what the device is, it could be a single hiccup in a pipeline load of a processor so occasionally it doesn't "compute right" to a leakage current measurement exceeding specification.

Once you've run the statistically significant number of devices at each temperature you run them until you get a statistically significant drop out rate. Then you turn to use that data to estimate the activation energy of that failure mode. But then if another failure mode comes along ...

Anyways, assume you have the activation energy. You then can use the Arrhenius equation to estimate the reliability of the device.

Here is a link to a FIT discussion here on EE.SE

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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