Path: utzoo!attcan!uunet!snorkelwacker!apple!vsi1!daver!bungi.com!news
From: Steven.D.Ligett@mac.dartmouth.edu
Newsgroups: comp.sys.nsc.32k
Subject: capacitor reliability
Message-ID: <1633025@mac.Dartmouth.EDU>
Date: 25 May 90 19:03:04 GMT
Sender: news@daver.bungi.com
Lines: 34
Approved: news@daver.bungi.com

There was recently a question about the reliability of the tantalum
capacitors.  I just got a new tantalum cap catalog from AVX.  Here's a summary
of the reliability info.

Reliability is dependent on three things
1. Vapplied/Vrated
2. Temerature
3. Series resistance.

The caps I'm ordering are rated at 1% failures/1000 hours,
 @85 (degrees)C, with 0.1ohm series resistance.

Factor 1:  All the caps will be rated at 20 volts or higher.  For the 12 volt
supply, this gives a failure factor of 0.03. For the 5 volt supply, this gives
a failure factor of 0.002.

Factor 2:  If you cool your system to 50 C, this gives a failure factor of
0.09.  If you cool your system to 40C, the factor is 0.05.

Factor 3: For a series resistance of 0.2 ohms/volt, the factor is 0.6; for 3
ohms/volt, it's 0.07.

-------------
So, calculating the failure rate of a 20 volt cap on the 12 volt line, in a
system at 50C, with a series resistance of 0.2 ohms/volt, gives 

0.03 x 0.09 x 0.6 x 1%/1000 hours = 0.00162%/1000 hours

That sounds pretty good to me.  Perhaps someone with a flair for statistics
can tell us when we'll likely see the first failure of that cap in the 100
systems that are being built.

Final note - I can buy 25 volt caps and halve that failure rate.  If you want
it, I'll do it.