Issue link: https://iconnect007.uberflip.com/i/1327102
22 DESIGN007 MAGAZINE I JANUARY 2021 the same because even some of our major laminate manufacturers manufacture some of the material here for North America, and they manufacture material in China for Asia. The process equipment we use is going to affect that material a little bit differently, and have different tolerances. We have to optimize our simulators based on the output of the actual products from our fabrication sites. I've dubbed the term "fabrication equivalent stackups." Some of my customers understand that we're providing two stackups, and even though they are slightly different, like the pre- preg thicknesses here, for the three plies of prepreg it might be 9.2 in China, where it's 9.08 here (for San Jose), for example. Fabri- cation equivalent stackups are going to meet all the IPC and customer requirements. Even though there are minor variances, those vari- ances are all within the fabrication and process tolerances, and they're all within the customer specification tolerances. Holden: When you do that, you've just eliminated one respin, since making that mistake, between local low-volume quick-turn and high-volume Asia, will result in enough differences to require a further change. Ellis: Richard, do you want to bring up the subject of best practices that you mentioned this morning, with regard to the prepregs? Dang: Yes. One of the things we consider when we're making a stackup is that we want to keep the prepregs as similar as possi- ble across the stackups. We have a process called lay-up before we go to a lamination, and the operator has to basically get all the materials in line and lay it up according to the stackup. If your design is using four or five different prepregs, once you take the prepreg out of their bag, you can't tell what it is. You really can't tell a 1080 from a 106 sometimes, just by looking at it. We try not to do too many, like adding 106 and 1080 in the same prepreg opening. If we can do it in two plies (sheets) of 1080 (68% resin content), that's what we'll try to do, and we'll try to do it across the entire stackup. That helps us eliminate potential operator error, especially in mass production, when they're working on hundreds of panels. Another best practice is to consider the glass weave used for different applications, such as low layer count, high-voltage, or high-speed applications. For 1.6 mm (.063") thick four- layer boards, we use thicker glass prepregs to achieve the dielectric thicknesses without using too many sheets, or plies, between layers of copper. However, when it comes to high volt- age applications, years of testing has proven that thicker glass styles (i.e., 7628) don't hold up as well against conductive anodic filament Figure 8: Typical glass weave constructions.