PCB007 Magazine

PCB007-Oct2018

Issue link: https://iconnect007.uberflip.com/i/1038326

Contents of this Issue

Navigation

Page 83 of 125

84 PCB007 MAGAZINE I OCTOBER 2018 it should never get more than two or three steps further in the process before a "flag" pops up. It is possible but unlikely an error could get all the way through to electrical test before find- ing the problem. Ultimately, that's our first goal, no escapes. Goldman: Instead of a first article, you run a first piece to test the recipe. Brown: Yes, but it's in every operation, that's the cool part. Because when we run cores, they'll get over to automated optical inspec- tion (AOI). And our first piece is checked at AOI. So if we have to make any adjustments, we make the adjustment right there and re - start the cores. We don't have to wait until we do a cross-section. If we do a first article inspection, we're going to look at it as more of a validation than an inspection process because of all the in-situ measurements we're doing. We're going to predict what the board is going to look like and then we're simply going to validate it. Goldman: And, especially early on, you'll learn with every board if things need to be tweaked. Brown: Correct. All that data is corrected and fed back. We can make "on-the-fly" adjust- ments for impedance, as an example, and not wait until it gets to E-test. Goldman: A lot of people try to inspect their quality in, but what they should be doing is using all those steps to fine-tune the process. Brown: We're not going to inspect it in; we're going to build it in. It's counter to how it's be- ing done because everybody does that. To your point, Patty, with first article inspection, you're kind of sitting there with your fingers crossed. I hope it's good, I hope it's good, I hope it's good. We'll be able to say, we know it's good. Then we just validate it because all of the data that we collected along the way tells us the of the layer and just by doing the math, we know what the dielectric is. If you know how far apart the copper is, you know what the di- electric is. And of course, we're using that to map the copper layer location in the holes. So, we'll rip a backdrill file from the Schmoll ma- chines because we've identified where every copper layer is in every hole. When we know what (holes) we have to backdrill, we'll just enter it in our backdrill file right from there and it's tuned to the actual board. With all of these process controls we've put in place, we're confident we are going to minimize escapes. With that aspect of servicing the cus- tomer in mind, we're evolving what customer service looks like. Because we're still new, we've made the decision that we're go- ing to delay adding sales peo- ple until we understand what we need to do to service the cus- tomers. Maybe we simply need more people in the front-end be- cause once we tool the job, we're good. We're already learning that front- end engineering is going to be our single largest headcount. The way our process is set up, once you start a board, there's no queue time. There's no need for queue time because it's part of the recipe. As that board gets to the next step in the process, the recipe is al- ready downloaded, and the piece of equipment knows what to do, and it does it. And the next panel that comes right behind it can have a dif- ferent recipe. Because all those 2D codes and radio-frequency identification (RFID) tracking that we put on up front tell the machine what's there. Goldman: So you totally depend on what hap- pens up front to get that recipe right. Brown: The recipe is key and again that's where the human factor comes in, the front-end en- gineering. We're going to have to do checks and double checks because we're digital; if we make an error there, that error is now part of the recipe. But with our in-situ data collection, Jim Brown

Articles in this issue

Archives of this issue

view archives of PCB007 Magazine - PCB007-Oct2018