Issue link: https://iconnect007.uberflip.com/i/1428512
34 PCB007 MAGAZINE I NOVEMBER 2021 showing up. And when defects show up, some can be reworked, some are just false alarms, but if you go from one or none at AOI (iden- tifying defects on a core) to 40 or 50, because things are finer, there are more false alarms and there are more actual things that can be reworked, things will get missed. en the final product is just a scrap board and we run more. Johnson: e people who need to be paying particular attention to this really seem to be the design teams, working with their fabrica- tors to make sure that they're getting analysis information, long term feedback to ensure a more reliable, higher yield, a more profitable board. Maybe we're starting to change how we do business a little bit. What advice do you have for the designers? Partida: It's collaboration. e complexity of current designs will benefit from collabora- tion between the fabricator and design. e fabricator can give guidance, improve yields. I worked with one customer who had a dou- ble-blind [via] that ended up into a buried that had five laser vias working its way out with a laser via that passed the buried via structure. Johnson: Is there a catch? Partida: No. e overall area was smaller than what it had to do and how they made the board. We went through the stackup for nine months. I said, "You cannot stack six laser vias. It's not going to work." I explained why. ey stag- gered them. ey did what they were supposed to. We went through the design three times. We went through part numbers three times com- pletely, and we found little minor things that could help. ey had a ground plane where they dog-boned the laser via. ey said, "Okay, the impedance is a three-and-a-half mil line on the ground plane." I said, "ere are things to understand about a board shop. Etchers over- etch, platers over-plate. On this layer when that's plated and over-etched, those three-and- a-half mil lines may disappear, but the track length is less than the two pad sizes combined. So, there really is no impedance. Could you change it from a three-and-a-half to an eight- mil wide trace? en there's no way we're going to etch it out." e EE said, "You're right, absolutely. Go ahead and change these layers and this layer and that layer." We went through and changed everything we could, and where we needed to when they were outside the components, we went to a larger via and land sizes so there wouldn't be any failures there. We only used a small feature when we had to. When we built these boards, it's complex, there's two double blinds that go to a buried. ere are five lasers on the way out, six total lasers down. We built Image of microvia separation 400X and with SEMs.