To summarize several comments:

Electrolytic capacitors reek.

Unfortunately I'm trying to build an extremely low cost device, and
I can tolerate a lot of variation from part to part.  I looked into a
PIC design, and it effectively increases the price of the whole
assembly by some 30% (do the math, this thing is really cheap.)

So ya get what ya pay for.

I Spent the last 2 days building an oven that can test components at
elevated temperatures and graph them vs. temperature.  My
conclusions are that electrolytics have a slightly positive
temperature coefficient (this was verified by a mfr)  a 100 uF
nominal cap at 25C was more like 105 uF at 85C.  That was just before
the can split open, however (sheesh) at 120C.

I've been afraid to use tantalums because the probably have excessive
leakage compared to a low leakage electrolytic.  But I'm not so sure.
 I may look into tantalums again soon.

My first experiment which seemed to indicate capacitance dropping is
still a mystery.  There are several components that could be
exibiting troublesome leakage or tolerance change at elevated
temperatures.  I'm testing them in my oven one by one.  I think my
original conclusion was erroneous.

To sum up, at high temperatures diodes are not diodes, resistors
become thermistors, capacitors become bombs, SCR's become
transistors, black becomes white,  politicians become honest, and
pigs can fly   etc. etc.   This is worse than designing for RF
circuits, it's like black magic.

Everyone who suggested a Timer or a PIC is right.

-- Lawrence Lile


Download AutoCad blocks for electrical drafting at:
http://home1.gte.net/llile/index.htm