I had asked the original question regarding methods used to determine the predicated reliability of an assembly, whether it be a PCB assembly, or the product containing the PCB assembly. There are a number of software packages out there that will do this, and there are differing methods ranging from simple - load the BOM into the package and it sums up all of the failure rates and spits out an MTBF prediction, to complex where you load in observed temperatures and other stresses which then are factored into the prediction (generally making the MTBF lower). 

Accompanying this are methods used to evaluate how to increase reliability and raise the MTBF. Increasing the wattage of parts that have higher stresses, heat sinking, etc. are common methods. 

I have typically been using field data to calculate reliability for our customers who request an MTBF number, however, on newer designs that number seems suspect, as the devices have not seen significant service life. 

A while back I worked for a company that produced electromechanical relays under Mil specs. We had banks of ovens and load testers that would test relays for weeks on end, and each week one group would come off and a new group would go on test. As you accumulated hours of test time with no failures DESC would allow you to claim a higher reliability level for the parts, as I remember L was lowest, then M, P and S. A failure would need to be factored into the calculation and could cause you to lose a rating. I think we attempted to get to 100,000 operations on each group being tested. 

Thanks to everyone who contributed to the discussion so far. I've appreciated all the responses. 

Blair


On Mon, 15 Apr 2013 17:15:48 -0400, Honsowetz, Eric <[log in to unmask]> wrote:

Was the question about the reliability of electronic assemblies or the reliability of PCBs?

The University of Maryland, http://www.calce.umd.edu/general/education/physics_of_failure_and_reliabili.htm

has been one place that research into electronic assembly reliability has been ongoing for a few years...

I seem to remember that someone there had done some work of the reliability of PCB materials and did a presentation Webinar via IPC but it has been a few years and I never have had a good memory...