Capacitor lifetime

[Kelte]

The more I look at the capacitor, the more puzzle I get. First we could expect to find the life time of the component in the datasheet, but what we finally get? 1000, 2000, 3000, 5000 hours! That is from one to eight months!!! A colleague explains to me that this figure is calculated in the worse condition, of course working condition are not so bad so the expected life is much longer… Hum, I may miss a point but I don't really care that I can use a capacitor during six months in the worse condition, all I care about is how long it will last in nominal configuration. There I got the explanation that I must measure the component temperature, and apply a mathematical formula to discover the expected lifetime… Anything easier ?? It's like buying a car and getting the information that it can last in average 3000 Km in battlefield condition, yes, but I use mostly my car to go to work and to go shopping (Saturday morning)…
So actually the 3000 hours (probably cheaper) component was selected. And after two year it broke down… Do I have to go to see the capacitor manufacturer with my “mathematical” formula and ask for money back? Really this 3000 / 5000 doesn't help don't you think? Could someone explain why these components are defined this way? How do you select electrolytic caps, on lifetime point of view, for your circuit?…

[shadow]

Re: Capacitor lifetime

Capacitors are rated to a common known standard or way.

The problem with giving a value based on a nominal condition is ambiguity.

What is a nominal condition?
Unfortunately nominal conditions change with each application.

What 'nominal value' should be the standard?
Should we even have standards?

Therefore we have been given the life expectancy based on a common and standard set of parameters.
From there we can judge which capacitors to use based on relative comparison (a higher number is better, get the biggest number that you can afford).
Or we could calculate the expected life time using these mathematical formulas or possibly with rules of thumbs.

Even if we have capacitors life expectancy based on your nominal values (what you think they should be). Another application for them with different conditions will need to use mathematical formulas to convert your life expectancy based on nominal values to their own life expectancy based on their nominal values because the conditions are different.

[KeriJane]

Re: Capacitor lifetime

Hi Kelte!
Welcome to the forum.

Yes, that's right.
It's all relative and depends upon the application.
Your car might last a very short time during a war or a race, or maybe much longer if driven carefully under moderate conditions and Capacitors are exactly the same in that regard.

And worse, many of the cheaper manufacturers lie or exaggerate. A somewhat reliable indicator is can size. An impossibly high value in a small diameter can (like 4700uf/10v in a 10mm∅ can) usually indicates at least some exaggeration. You wouldn't want to believe the rating on useful life for this situation either.

The best thing to do is the same as buying a car:

Stick with a reputable brand and give it moderate conditions to operate in. For electrolytic capacitors that means adequate cooling and not being over-stressed electrically.
And hope that you didn't get a lemon.

Have Fun,
Keri

[Kelte]

Re: Capacitor lifetime

Thanks a lot for the clarification, KeriJane and Shadow.
“Get the biggest number that you can afford” and “Stick with a reputable brand and give it moderate conditions to operate in” are quite pragmatic rules. (Unfortunately in consumer design these are not always followed)
Kelte

[shadow]

Re: Capacitor lifetime

For one off projects, you will find that sticking to reputable brands and getting the best that you can afford do go ahead and are followed.

The problem is with mass production.

If you can save 50 cents on each board by using a slightly less reputable brand (with possibly exaggerated specs) and you produce 1,000,000 units. You will save $500,000 on production alone. This obviously increases profits.

For one off projects, spending an extra few dollars here and there typically doesn't matter as much since you are only producing one (or a small number of) end product.

[gdement]

Re: Capacitor lifetime

Another problem with giving nominal lifetime values is that it would just be based on estimates and guessing.

The endurance rating under worst case conditions is something the manufacturer can actually test and measure, so those values have real meaning. Testing at moderate conditions would take years, which is impractical.
I like that capacitor manufacturers instead publish what they can actually test.

This is something that bugs me about some devices like Flash memory. Typically they claim data retention of either 10 or 100 years. They don't elaborate on how they got that figure, but I admit I haven't really tried to find out either.

They also promise some number of rewrite cycles, that at least I can believe they were able to test.

I wish that datasheets would express endurance test results in statistical terms, rather than just promising that they'll meet some minimum.
Something like "120K rewrites with standard deviation of 30K" would be more honest and useful. Especially in a mass production scenario, where you want to estimate what % of your products will still be working after X cycles/years.