Issue link: https://iconnect007.uberflip.com/i/871322
22 The PCB Magazine • September 2017 in—instead of inspecting it in—we had to al- locate all our resources to the front-end tooling because we were getting such variations in de- sign. I complained to HP's vice president that manufacturing is building quality into the prod- uct but how come design can't design quality into the product. The designer says, "We don't need standardization. You guys just build what we design." Well, the vice president didn't like that reaction and fortunately he promoted me to be head of PC board design at HP. I've always been aware of design rule check- ers and things like that since at one time HP ac- tually sold PC board design tools. One of HP's key successes in the IC business was having its own IC design tools that were highly automat- ed. And the IC design tools were interesting; before you start designing the IC, you have to thoroughly characterize manufacturing or the IC design tool won't even let you start. A printed circuit board design tool had the heritage that you could put in any numbers you want, just because it comes out of a computer and a CAD system, it's very easy for it to make things that are totally impossible to manufac- ture. So, the concept of quality/producability has been one of my soapboxes over the years talking about EDA design tools for PC board. There is no measure of producability or quality. Instead, everybody has to run something like the DRC. And then where do you get the num- bers to put into the DRC is an interesting thing. Essentially, you're looking for errors. But errors are not really design for manufacturing as de- fined by the industry and two university pro- fessors that developed the technique. As Figure 5 shows humorously, just because a CAD sys- tem can describe a product does not necessarily mean it can be built. When HP took a DFM/A license and put us all in school, we learned that DFM is "doing it right the first time." It's not designing it, trick- ing it out, building it and then going back and redesigning it, looping it around and around on re-spins. How do you figure out how it can be designed right the first time so it only takes a lit- tle tweaking and you're off and running a high- volume product with high yields and excellent quality? Because DFM and DFX have been so mis- used I had to use the term "predictive engineer- ing" in order to get people to understand what I'm talking about. But I'm talking about what we need is software systems that help us do it right the first time. I have a chapter on DFM/ DFX, talking about the predictive engineering point of view [3] , to explain the methodology and how that applies to PC board, IC, and sur- face mount assembly. Predictive engineering works fundamentally by feeding back manufacturing data. So instead of being "over the wall," this is a true feedback loop where you look at how things actually were manufactured in their class, and their queue, their problems, and their quality. You feed that back into your predictive algorithms so the next time, you design it better. It's not just finding the mistakes from a design rule checker type of thing. The design rule checker helps, but what you'd potentially like to see is that these things are manufacturable and the design rule check- er doesn't find anything. You can see in Figure 6 that the statistical process control of processes can lead to enlarged process latitudes and to en- hanced design rule specifications. And the high volume is the same as the pro- totype you made. All that time and wasted ef- fort isn't there anymore. The EDA and the CAD venders are not particularly interested because they don't consider a PC shop a customer. They consider the person that buys the tool the cus- tomer. They have absolutely no visibility over what Six Sigma and total quality management PCB PROCESS ENGINEERING: DETAILS FROM ONE OF THE ORIGINALS Figure 5: Even though your CAD system can describe a product, there is no guarantee that it can be manufactured.