Issue link: https://iconnect007.uberflip.com/i/1479191
SEPTEMBER 2022 I PCB007 MAGAZINE 59 these smaller packages of 0.4 mm, 0.5 mm, it's beneficial because it's a tight density. But don't take that same three-stack configuration out- side of those type packages. You must stagger. You have room. If you have a discreet compo- nent or something that's big, there's room to stagger, and those can be reliably staggered by themselves. I want to share with the design community that there's a rule that is dependent on density. What's illuminating is there's a deeper under- standing of the reliability of the interconnect of microvias and where they fail. What breaks the microvia is material expan- sion. So, when you stack them, it's the mate- rial expansion of the entire stack from top to bottom and how much strain is created during reflow. Do you have enough surface area to hold on for the expansion expected? at's what the simulation soware does. It looks at the pre- preg, how much is retained copper, total dis- tance, and says, "ere's this much force. You have this much surface area. It's not enough to hold on. It will fail at this reflow." at's what I hope to present at IPC APEX EXPO so that it's helpful to the designers to understand, "Okay, I can do this safely in a very tight pitch device because the microvias sup- port each other in an area. But in my design, I make sure I had different rules outside of those packages." For the most part, they use the same rule everywhere. Shaughnessy: What was your process to get here? Was it just trial and error? Partida: Luckily, we've had a D coupon tester in-house for years, so I had all the cross-section data for all the jobs that went through D cou- pon testing. With the ones that were weaker, I could go back and cross-section them. I could see how thick it was, and I could check the laser. I had access to the CAM data, so I knew what the laser drill size was. I knew what we were projecting for the dielectric. I could look at the cross-sections to see if we got there or not. In early 2021, when I started using the simu- lation soware, I had all this data. I wanted to plug it in backward from what's already tested and see how closely they match. For the most part, they matched. I had this one weird out- lier, it wasn't matching, and I was looking for something else. But then we worked with our suppliers, and they said, "Look at your laser via. It's too smooth at the bottom. You didn't prepare this correctly." And then we said, "You know what, that's right." Once we did that, then the simulation soware was in line with what was going on. It helps us make sure our processes are set right. Make sure that we are going with the right starting point of building a board. But the guidance is to have the custom- ers design and use stacking where they can for their benefit, but not to put it where it's going to make the reflow in danger by stacking too much in the middle of nowhere. And that's really my mantra for this year in sharing with the customers or the design community. Shaughnessy: Don't stack out in the middle of nowhere. at's good. Partida: About six months aer playing around with the simulation soware, I changed the pitch from a real tight pitch, and I got a lot of reflows. I said, "Oh, wait a minute. is customer uses a 1-millimeter pitch," and I switched it. It was like four reflow cycles to fail- ure. I thought, "Something is wrong with this soware. e bigger, the better. It should have been better." So, I called the soware guy, and It helps us make sure our processes are set right. Make sure that we are going with the right starting point of building a board.