A colleague once told me, “You can’t cheat physics.” Of course, this wasn’t a novel concept to either of us, but it was in the context of our conversation that made it a striking statement. We were discussing a recurring issue of fractured solder on an assembly. As Process/Manufacturing/Quality Engineers, when it comes to loss of solder integrity, the lengths taken to find fault in the manufacturing process is often the sole focus and exclusionary of other fault potentials. It seemed next to impossible to say that the failure had no manufacturing solution and attempts to find and implement one was to cheat physics. Likewise, until every button was pressed and every knob was turned to fix the problem, could such a statement be made. In manufacturing, there are so many knobs, buttons, and combinations of each. This makes the design of experiment (DOE) our most powerful tool. You can find archived webinars on DOE, presented by Indium Corporation’s Dr. Ron Lasky, Ph.D., PE, Senior Technologist, on our website. Simply scan the QR code below:
While you are there, be sure to register for future webinars that pique your curiosity.
Before pursuing a DOE, let data drive your investigation. First, examine the assembly pedigree. Pull all manufacturing records for the assembly serial number. Does the data indicate any yield loss or anomalies? Hopefully, your process has a robust serialization, traceability, and data collection system that makes data mining simple and straightforward. Data mining should take no more than a few minutes upon receiving the assembly serial number. With that single serial number, you should be able to drill backward to the incoming material certificates if necessary. All test data—not simply that the assembly passed or failed—should be available as to what was tested and the quantitative results for each test. Depending on the assembly complexity, all inspection steps, solder paste inspection, reflow oven settings, automated optical inspection, automated X-ray inspection, visual inspection, assembly screw torques, etc., comprise the assembly pedigree for the failed part. If no anomalies are found, what is the chance that a process fault occurred? In my next blog, I will provide tips on many more failure analysis techniques.