I met Peter Borgesen back in the mid 1980s when he was a research scientist at Cornell working with Professor Che-Yu Li. Later we worked together at Universal Instruments. Currently Peter is a professor a Binghamton University. All during this time, Peter has been working on materials science-related topics in electronics packaging and assembly, most notably reliability. In addition to his many technical skills, he is a gifted linguist, speaking multiple European languages. Etched in my mind is Peter talking to several European graduate students in their native European languages in the space of five minutes, switching from one to the other effortlessly.
Few people know more about lead-free solder reliability than Peter. So I thought I would get his perspective on my recent post on lead-free field reliability data. His comments follow.
I agree that the sky is not falling. Also, we should be talking much more (only?) about life in service. I realize that we don’t know enough about this (and our predictions based on test results are much more off than people want to recognize). The vast majority of practitioners focusing on ‘engineering tests’ are doing worse than wasting time and effort if comparisons of test results do not translate to relative performances in service. There is a lot of ‘sticking heads in the sand’ here.
I am not concerned about the long term life of cell phones, and not very worried about in which respect they do better or worse in service than with SnPb. Intermetallic bonds have generally gotten weaker and more prone to sporadic defects, and cratering is greatly enhanced for the devices Vahid Goudarzi mentions when discussing Motorola field data. I agree those are limited concerns, no sky falling indeed.
What still scares me (in the case of critical applications) or concerns me (in the case of expensive applications) is the naivete with which many seem to think we can learn much about sporadic disasters or long-term reliability of those from consumer electronics experiences.
I am not often interested in comparisons to actual life of SnPb either (any more). We face ever more applications (designs and service conditions) for which we don’t have sufficiently accurate critical experience with SnPb either. The first challenge becomes not to be surprised by effects of long-term aging, combinations of loading, minor differences in pad finish, joint configuration, latent damage, process, …. and their interactions for the specific solder alloy used (!).
While I can’t yet extrapolate test results to life in long-term service (I think we are close, but I really need an extra $1M to prove my hypothesis and turn it into a quantitative model) I can show how current models may easily be off by 2-3 orders of magnitude or more (worse, how they may screw up comparisons of alternatives). It obviously depends on the application whether this really matters (I side with companies who have cut drastically back on testing for many applications).
Keep up the good work.
I will keep in touch with Peter in the future for updates on his perspective.