Issue link: https://iconnect007.uberflip.com/i/1284035
126 SMT007 MAGAZINE I SEPTEMBER 2020 If we were to take a specimen after micro- etch and everything was fine (Figure 7), we would focus on ion-beam milling and look inside it. You can find a crack; in some cases, there are failures. You find these cracks in the board, and now everybody's in a panic about whether these boards are going to work. I have seen many customers who have had 1,400– 1,600 I/O BGAs using four-stacked microvias. I told them we were not going to build the boards. They said, "You build all our hardest boards." But I knew that this board was not going to get through their assembly. Six months later, I was in their quality lab with two Ph.D.s. They were looking at a board, and I said, "Let me guess. When you put your finger down, they work." Their response was, "How would you know?" I was not going to build that board because they have this type of failure in them due to the four-stack migrations. They are only detectable by using very expensive, very sophisticated focused ion beam milling. Another requirement these board fabricators have is monthly quality conformance testing for Class 3 suppliers. All we have to submit is the most complex board we built in that month to testing. The test checks rework simulation, bond strength, peel strength, dielectric with standing voltage, and moisture installation. But there's no way to test whether the board shop can actually produce microvias reliably. When we've finished building a PCB, we test the electrical circuit boards at ambient room temperature. But a weak microvia still has enough connection and will not fail an elec- trical test, especially when the electrical test threshold is typically at a 10 ohms resistance value and anything under 10 ohms in a net is considered a connection. Therefore, the weak microvia is not going to be detected. A funny story about testing at ambient room temperatures over the last five or six years is we've been working with reliability reflow test- ing of the coupons since the very beginning. In that time, some customers would not do the coupon testing and, therefore, had assem- bly problems. Typically, there were three- or four-stacked microvias involved. We shipped the boards. They called us and explained how they wanted to retest the boards after putting Figure 7: The older methods fail at detecting weak microvia defects.