<The story on cosmic rays affecting memory is well-known. It was proposed
<as the reason why one of the early 4Kbit (I think) chips had a very high
<error rate. It was later found (I think) that the real reason was alpha
<particles from the radioactive decay of something in the (ceramic?)
<package of the chip.
it originated with the transiston from 16k Drams to 64kdrams as the feature
size was small enough that an Alpha particle hitting a sense line or other
die level feature could offset the charge on the capacitor(storage element).
Also different refresh schemes, better logic, faster parts
and fewer interconnects all contributed to this soft error rate going
down rather than up. The source of the particale was the gold eutectic
bond used to hold the die to the lead frame or the lide for the brazed
lid cases. The solution was a overcoat of particle absorbing material on
the die. Since there there has been a many fold(or greater) geometry
shrink to get the current megabyte parts and we do not have the forcasted
problem, though the potential exists. It's not to say the problem wasn't
real but win95 crashes(on its own) more often than the probability
of a actual soft error in practice.
Allison
Received on Tue Jan 05 1999 - 21:45:34 GMT