Tim Watts wrote:
On 06/01/15 06:10, John Robertson wrote:

As usual, heat is the enemy. Reading the technical specs for caps is
enlightening. They are rated usually at something like 2,000 to 5,000
hours at their rated temperature. So an 80C cap will die after something
like 2000 hrs at 80C or 4000 hours at 50C and 10000 hours at 40C (not
looking it up!), whereas a 105C cap will last 10000 hours at 85C, etc.
So, the better the grade of cap the longer it will last in warm to hot
environments. And there is the equivalent resistance and inductance to
consider as well. Some caps are much more tolerant of 50/60hz and others
are better at 20,000hz. Selecting those takes time and the cost
accountants slip in at some point...

That's very interesting.

Based on this random driver circuit:

http://www.ecnmag.com/sites/ecnmag.c...206-web(1).jpg


and looking at RS for 680uF around 50V electrolytics:

the 105C are around 50-70 pence
125C are around 130 pence
150C are 252 pence

There seems to be 1 big cap in that circuit - I don't have time now to
cost all of the electrolytics but this looks like a case of adding maybe
3 pounds would make the difference between a short life and a very long
life.

I suspect Philips use the good components - their LED bulbs seem to last
a long time (my tests are still pending) but they cost rather more than
3 pounds over the cheaper LEDs.



The things that kill electrolytic capacitors are ripple current and
ESR. If you double the capacitor size, the life dramatically increases
as the running temperature decreases. I once had a very nice DOS program
which forecast capacitor life extremely well based predominantly on
ripple current.

I see that Ford have had to add cooling fans to their infinite life LED
headlamps. I've never yet seen a fan with an infinite life!