Issue link: https://iconnect007.uberflip.com/i/873992
September 2017 • The PCB Design Magazine 19 ing the schedule or the price quoted. There's no performance or quality measures when they re- spond. It's simply that schedule is the only met- ric we're trying to meet. I said, "Well, we defi- nitely have a problem in that, in design, there's no measure of quality." Design takes absolutely no ownership that this thing is not manufactur- able or producible, low-quality manufacturing. We invented metrics, which I think I elabo- rated, but we also looked at methods of con- stantly improving the design process and how we can dig back to first batch yield in electrical test because we discovered that if we physically analyze first test yield in electrical test, the so- called random defects are not random. If you do an F test or T test of significance physically, log after log, you'll find that certain nets always have shorts in them. When you go back and look at that you'll find some minimum spacing and things like that which moves them, then those things no longer randomly show up as electrical shorts. One of the first papers I ever wrote, in 1971, an- alyzed electrical tests; even today, nobody ana- lyzes electrical test. They just have a pile of good boards and a pile of bad boards. They don't real- ize the test data is telling them something about design or something about manufacturing. You have to understand what a map test is in order to understand that what they're calling random really isn't random. It's designed in. It's a very complex topic. I sympathize with you, Andy, in terms of how to get your arms around it. How do you make this an interesting topic and not just somebody standing up on a soap- box complaining about lead-free soldering? Matties: Happy, can you share a story you once told me about taking over the design area at HP? Holden: I complained to the vice president of our process group that Six Sigma and total qual- ity manufacturing were having a profound ef- fect on our manufacturing. We were eliminat- ing final inspection, empowering employees to own quality of the product, and everything else. Everything was going fine except for the front- end and tooling. We were accumulating all of our resources to inspect the incoming design file, because they had so many problems and wouldn't make these design mistakes. Dieter thought you have to offer tours of board shops so designers can see how PC boards are made. After managing print circuit design, I realized that this was a pipe dream. It's difficult to learn how to design. You can spend your entire life trying to become an expert on design. How can a one-hour tour of a board shop, where you see all these complex processes and machines, provide any insight on how that board should have been designed? Because it takes an entire lifetime to become an expert in printed circuit board technology. How is a de- signer going to learn that just by observing a step? We went back and said that we have to provide designers, number one, with the best design process that we can in terms of simula- tion and tools available to them. The second thing is that in the design pro- cess, there are all these critical decision points that a designer has to choose in order to contin- ue the design. At those critical decision points, they need to know the pros and cons of the al- ternate path that they can take. We put out a design for manufacturing manual that created what we called the relative cost index. When they would hit one of these decision points, in the design manufacturing manual, they could find out what the relative costs and a couple other boundary conditions would be if they pulled some particular path on this design. That at least gave them some insight into the decisions they were making. Because, when the design was sent out for a prototype or quoting, they'd say, "Well, why is this quote so high?" They said, "Well, if you had simply not done this or you've not done that…" The designers said, "That's interesting, but you should have told me this four weeks ago, because this is an early decision, and I can't go back and change that now. I'd have to start all over." Because we were a design organization owned by manufacturing, not by R&D, our focus was to design the most manufacturable board, which meant we could meet delivery times at the highest calling and the lowest cost. That was our goal with design, not just to meet the schedule. One thing that's interesting, if you ask designers, is "What's the measure of perfor- mance?" The measure of performance is meet- PREDICTIVE ENGINEERING: HAPPY HOLDEN DISCUSSES TRUE DFM