An account (not mine) of the total failure of 150 high 
quality highly engineered systems under test for reasons 
that were retrospectively understandable but which would 
have been unbelievable at the start.

ie a high quality mil spec opto coupler that was AC rated to 
5000 VAC RMS caused utterly fatal long term problems when 
run at 500 VDC.

An excellent lesson in what can go very very badly wrong.


        Russell

____________________________




Lots of subtle issues here....
Strange things happen with large DC voltages.

I once had a vacation canceled because all 150 of the high 
tech
switching power supplies that
had been 100% qualified at 50C for 24 hours were shipped to 
the
integrator where all
150 failed spectacularly near simultaneously after 72 hours 
at 40C.

The fault was an opto isolator in the current feedback path.
It was a an expensive full mil spec part specified for 5000V 
RMS AC,
but carried no DC rating. At 500VDC with at elevated 
temperature charge
migrated across the barrier depositing on the receiver side 
until the
charge built up to the level
where it jumped into the receiver logic and killed it.
Causing the large multi kilowatt power supply to loose its 
primary
current feedback
path causing the output transformer to saturate and take out 
the
power transistors
that then failed shorted and took out the battery backup 
transfer
switch shorting
the large lead acid batteries they were connected 
to....blowing the last line
of  defense large DC fuse. (DC fuses are physically much 
larger than
an equivalent AC fuses
as AC extinguishes the arc 60 times a second.)
Batteries were separate form the main electronics and 
separately
cooled for thermal reasons.

This was a supply design that had been rigorously reviewed 
and
subjected to extreme
testing both mechanical and thermal. The first test supply 
with lots of buried
temp sensors in the potted magnetics, etc...reached thermal 
equilibrium
in 30 minutes after running at  65C. So we felt safe with a 
24hour test at 65C.
The whole job was a rush to power some battle field 
computers for the original
dessert shield/storm.  In additional testing after the 
failure we found that
at 65C the opto coupler failed at 26 hours +/-1 hour at 
500V.

Via FEMA we had identified the opto coupler as a potential 
bad failure point,
and had upgraded the otherwise COTS equipment in that area 
to a high
reliability
opto.

This job and failure has always given me nightmares, we did 
everything right
and it still blew up in our face. What if it had been a 
spacecraft with
a 10 year travel mission and the DC charge migration took 
9years to go bang?

Complex systems fail in complex unknown ways...

Paul




-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist