Regarding single sheet prepregs, I feel comfortable down to about 3 
mil average pressed thickness.  I have issues with single layers of
1080 or 106, the 106 being worse. 

I have run some quick tests on single sheets of 1080/106.  These were
biased humidity tests on very dense computer boards.  Shorts would first
form in the area of high density arrays, between power/ground planes.
The reason is the large number of plane keepouts in these areas would
result in a thin web of metal between the keepouts.  During lamination,
the epoxy flows into the keepout, while the metal webs on facing planes
could get as close as 1 mil apart.

In all fairness, I did not thoroughly characterize this failure
mechanism.  To correlate to real world conditions would involve a large
factorial experiment in which voltage, temperature, and humidity were
varied.  Results will also be very design specific.  Under the
conditions I used, the single layer prepreg failed much earlier than a
standard construction.  Whether this accellerated test was
capable of representing field conditions, has not been determined. 

AT&T has presented excellent papers on accelerated life testing that
would be applicable to this situation.  Search for author M.J. Luvalle.

Thin prepregs are already here in PCMCIA cards.  We are hesitant to
introduce them on larger, more complex boards because of the greatly
increased number of potential failure sites.  However, we are going to
have to face this issue in the very near future.

-Jeff Deeney-
[log in to unmask]
Hewlett Packard Company
Fort Collins, Colorado