Issue link: https://iconnect007.uberflip.com/i/1307491
40 DESIGN007 MAGAZINE I NOVEMBER 2020 The industry has been using microvias for decades, and most of us shared a basic misconcep- tion about microvias and high- reliability applications. The microvia was believed to be the most robust type of intercon- nect. In many ways, it still is. This is a low-probability prob- lem, but its latency and inabil- ity to be detected by traditional testing methods is unfortunate. Seeking certain density designs, we incorporated two, three, four, or more stacked micro- vias, possibly without sufficient forethought as to what that might do to uncover new failure mechanisms. Because of our preconceptions, we miss diagnosed problems quite often and early on. In some cases, companies drifted along for some years, thinking they uniquely had the problem—until they started comparing notes [1] . We are in a containment position. We have received guidance from the IPC, and we have internal companies with their own guidance. You've seen the Motorola red, yellow, green structure [2] . That involves a kind of contain- ment by limiting the complexity and being very careful to qualify sources because not every- body is equally good at this. Using intensive screening and perhaps over-conservatively defined lot rejection, this problem is corralled but limits advancement. Many OEMs felt confident about what they were getting and fielding. With limited capabil- ity, especially as you get to the more complex structures, that limited capacity in the industry in North America. Some of those failures and rejections have reportedly led to late deliver- ies, etc. In today's world, the biggest impact is scheduling. We have made a lot of progress in screening technology. Screening helps prevent escapes to the field. We are much further along in containment than we are in understanding the interaction with the multiple contributing mechanisms [3] . There are some papers listed. This project is intended to add to the publicly available knowledge of those mechanisms. Project History It was a little before mid-2019 that we started having conver- sations. From some of the can- didate organizations—mostly centered around bemoaning the absence of definitive con- ferences of industry accessible data—we know that there is very good testing that has been done by some of the larger OEMs and large printed circuit networks. However, due to the nature of the commercial world we live in, a lot of that is proprietary IP and has been difficult sharing. That is perhaps starting to open a little bit. But when we started this project, it was a real prob- lem, and the VTSL data sorting microvia group had hoped to do a lot of mining of existing data and was thwarted by that inability to get it past the IP police of various organizations. We began the process of saying, "What are we going to do about this?" We began discus- sions among a select group of manufacturers, test organizations, simulation labs, and univer- sities. The market segments were served, and similar processes were especially important. One of the things that took the longest, if you are going to have two or more fabricators build- ing a test vehicle, is getting an exact consensus on the details of layup and construction, which became an exceptionally long, hard, arduous process. In September of last year, the fabrica- tion companies began the consensus material selection; we were in a fairly quick and early agreement. We wanted a material and stackup that was widely used and available, but per- haps not the most used, as the ultimate mate- rial. And that was also covered by those who would supply the material to support this test in return for access to the information. In October of last year and running through at least January, the two fabricators collabo- rated in an exceptionally open manner. Those of us in the industry are used to helping each other out, but this was remarkable and resulted Marc Carter