Roger Held wrote:

>     The problem is that the evaluation board had a shear crack in the 
>     barrel of a 12mil small via hole after 100 cycles of thermal shock

	There is a publication, IPC-TR-579, dated Sept 1988, "Round Robin
	Reliability Evaluation of Small Diameter Plated Through Holes in
	Printed Wiring Boards" (I typed it right! I typed it right!), that
	presents a large amount of analysis data from a cross-section of
	vendors on via failures vs size and board type for several thermal
	excursion tests.

        I'm not sure the results presented are all that current, but there
        were themes that probably persist.

	I'll apologize for being unable to do justice summarizing the results,
	but it doesn't lead me to think that your soldermask is the problem.
	In fact this doc (along with some TechNet discussions) led me to
	conclude "It is better to allow tangency or breakout (with fillets)
	then to reduce via hole size", at least for cases with excursions
	beyond 0-100C/more than 100 cycles, or a where a broad pool of ven-
	dors is desired.

	Good luck,

        Jeff Seeger                             Applied CAD Knowledge Inc
        Chief Technical Officer                      Tyngsboro, MA  01879
        [log in to unmask]                               508 649 9800