Issue link: https://iconnect007.uberflip.com/i/1245291
MAY 2020 I DESIGN007 MAGAZINE 15 started. We got into a culture of, "Get the de- sign close, throw it over to manufacturing, and they'll find all of our capability issues and fix it for free. We don't have to worry about it." That's kind of the standard process today. Get it close, and throw it to them because they'll do it for free since it's so competitive. I expect to have a design to come in that I don't have to modify. I hated having front-end engineer- ing/CAM people modify a customer's design as they might mess it up accidentally. And I know that for board layout—especial- ly for a poor layout person who is laying that board out—it's not like they have to just wor- ry about DFM. They have schedule, SI con- straints, schematics, timing, thermal, cost, and mechanical constraints. There are lots of constraints involved when you lay the board out, and because I've had quite a few design- ers work for/with me over the years, I appreci- ate how hard their job is and how the software tools aren't up to speed enough to allow them to do all this automatically. Shaughnessy: Right. And it's funny because the designers will say, "Don't mess with my design," but almost every CAM person I talk to says that, unless it's a long-time customer, they're going to have to do some tweaking. It's just accepted. Korf: It is accepted to a certain point, and there are various levels of that; it's interesting. You look at the broad industry overall, and this has been true in North America 20+ years ago and very true in Asia; someone will get a design in, edit it, and not tell the customer they changed something. They build it, send it back, and the customer goes through all the testing—including development and software testing—and it works. Then, they send the design out to a higher volume shop that has more financial risk for it if it doesn't work, and they come back with a long list of questions. A lot of times, the designer responds, "My last supplier didn't have that issue." One of my favorite responses was, "This is a law of phys- ics problem. They couldn't have built the board if they didn't fix this. It wouldn't have worked." I guarantee you they modified X, Y, and Z to make it work. Another example: A customer with a via fill board wanted a 100% via fill guarantee. I told him, "That is physically not possible." Num- ber one, it's not possible because there are al- ways air bubbles when you print any kind of ink, cure it, and the bubble gets trapped in the material. Number two, how do you even inspect it? You can't inspect the inside of ev- ery hole. They said, "My vendors never had the issue." And I said, "Go ahead, give me a board. Let me cross-section one via on a board randomly from one of your current vendors." About three months later, it was an automo- tive application, and he came back and said, "What should the specification be?" He went and tested some boards and found that they weren't 100% either. But they were working just fine. Sometimes, there are laws of physics that are real. One big problem is, and I wrote a column about this a few months ago, that board shops consider design rules proprietary. They don't want to tell the designer what the rules are be- cause they will know all my processes, and it's secret. At one company, they wouldn't let me give the designers our design rules. I said, "How do you expect us to spend less time in quick-turn if you don't tell them what we want?" They finally let me do it, and the de- signs got better. We have to break this mentali- ty of design rules being kept secret and not giv- ing them to layout people. We should tell them everything that we can do—not how we do it, but what we can do. Shaughnessy: I'm wondering how this thing has gotten so askew because most of these designers have been doing it for 30–40 years. Korf: People have various forms of running DFM during design, post-placement, and whatever various stages of the layout process, but the problem is if you don't know what the rules are, you're always going to be wrong. You're always going to have at least one issue, prob- ably. This is why it's a two-ended problem. It's a CAD tool problem; it's a sourcing prob-