9

Its a common specification for household devices such as lightbulbs. However, I can't figure out how you'd truly evaluate/prove such a claim without running the device for the specified amount of time.

Consider a lightbulb that is said to have a lifetime of 9000 hours. If I were to test this, the only way I can think of truly measuring this is to let the lightbulb run for 9000 hours, which is approximately a year!

If a year is not long enough, consider certain LED bulbs rated at 50,000 hours!

Clearly its not feasible to run a test for this long. So I guess I'm asking; on what basis are these claims made?

Perhaps one way to test it is to stress the component at higher than normal operating conditions, so that it burns out faster, and then somehow create a prediction based on measurements. Or perhaps run the component for some (short) time and measure the deterioration/ageing and use that to create a prediction.

midnightBlue
  • 1,805
  • 4
  • 21
  • 35
  • 1
    And if THAT is not enough, consider the 100 years data retention claimed for [EEPROM](http://www.atmel.com/Images/2545s.pdf). – Vorac Dec 01 '15 at 11:53

3 Answers3

2

One of the methods is, probably as you've predicted, that of accelerated aging. This is used when the lifespan of a product is such that it would be impractical to run a normal lifespan test (such as LEDs, which have MTBFs of possibly 100,000 hours plus). Here, one stresses the test item past what it will ever receive "in the field" to achieve a shortened lifespan, from which the data can be extrapolated.

A major consideration in the use of accelerated aging is the non-linear effect of operating outside of a part's recommended operating range. This can be illustrated with mechanical systems, such as running a gearbox at 45,000 rpm instead of 15,000, and extrapolating the data by three. However, suppose you're testing your lightbulb. Common sense would say that running it at twice the current would make it run half as long; however due to those non-linear effects you may find that lifetime is only 1/4th of that at the proper current because of the additional stress of the overrating. An important consideration is that the test device/subject/item should be well-characterized in both its intended and unintended operating range behavior prior to performing the test.

Because of that there are any number of studies (a LOT of them) on the various refinements to the accelerated aging tests in various fields. LEDs come to mind; photovoltaics are another.

ecfedele
  • 607
  • 4
  • 11
1

As you intuited, lifetime rating usually involves testing at operating conditions more severe than the specifications allow. Mathematical models - which can be empirical or theoretically derived - are then used to map the tested time to failure to a practical set of conditions. In semiconductor devices, one well-known such 'law' is Black's equation and the general technique is called HTOL testing. As you might imagine, it can be difficult to establish the validity of accelerated tests, and some engineers would recommend taking the numbers that result with a grain of salt. In the semiconductor industry, many standards have been created and continue to develop.

user49628
  • 1,218
  • 6
  • 7
0

I think you basically answered your question. Depending on the technology the test cases and protocols to estimate the life-time vary of course, but overall they measure the degradation to a certain percentage and estimate the total life-time based on the protocols specific to that technology. I found this regarding the LED life-time estimations that you mentioned: http://apps1.eere.energy.gov/buildings/publications/pdfs/ssl/lifetime_white_leds.pdf Hope this helps.

Arash
  • 249
  • 2
  • 9