View Single Post
  #68   Report Post  
Posted to uk.d-i-y
Commander Kinsey Commander Kinsey is offline
external usenet poster
 
Posts: 4,540
Default Why doesn't everything use solid aluminium capacitors?

On Fri, 01 May 2020 21:05:34 +0100, Paul wrote:

Commander Kinsey wrote:
On Fri, 01 May 2020 01:07:53 +0100, whisky-dave
wrote:

On Thursday, 30 April 2020 20:35:27 UTC+1, Commander Kinsey wrote:
Cost is only 20% higher. Lifetime is many times greater.

Not compared to the types of capacitors check out the data sheets
typically up to 2000 hours.


Anything with only 2000 hours is not fit for sale. That's going back to
the days of the filament lightbulbs.


At the temperature of the surface of the sun,
the lifetime of the capacitor is 0.1 milliseconds.


They're not used there.

While you're sitting on the surface of the Sun,
a spherical cow flies by.


WTF?

At 105C, the capacitor lasts 2,000 hours (as communicated in
some military reliability handbook or similar). The numbers
selected, are intended by the manufacturer to resemble
"classes of applications". Military, industrial, consumer.


Rubbish. I've got graphics cards approaching that temperature. 2000 hours is only 83 days, that would fall foul of the warranty.

At 45C, the capacitor lasts 60,000 hours.
This is the temperature inside your PC, on a particularly
hot summers day. A day so hot, you've stepped out of the
computer room, to go for a swim down at the lake.


Really? Capacitors tend to be next to power circuits, they're much hotter than that.

The rubber seal on the cheap capacitorss,


Which is why I said we should be using the solid ones. I see no seal on those.

is rated for a 17 year
lifespan. Which is an issue that "runs in parallel" with
the accelerated life testing curve being communicated above.
If the room is filled with concentrated ozone, the cap
lasts a day or so, before the seal is gone and the cap
now has evaporating electrolyte.


But you'd suffocate first.

The behavior of the reliability number, follows an Arrhenius
style equation, with suitably extracted empirical constants.
This requires the engineer in the lab doing the measurements,
to own some "curve fit software", because nobody writes
this software for themselves any more. (Although we were taught
how when I went to school. Not that this matters.)