Issue link: https://iconnect007.uberflip.com/i/1401769
16 PCB007 MAGAZINE I AUGUST 2021 Thompson: Good point. Shaughnessy: at's interesting. If I give you data, I'm assuming that you're going to check it to make sure it's right. Korf: ey're assuming you're going to fix it. Why would I check it if it's right? Having worked in the volume side and proto side, it's oen you get a job in Asia, like Happy said, from a pro- totype shop in North America. e prototype shop is being paid to get in and out quickly and make some money at it. ey're not paid to necessarily make it low cost because they don't know where it's going to be purchased in volume anyway. So customers say, "My last fabricator didn't have that problem." I'll say, "You probably didn't need to check it. ey didn't have to op- timize the yields like we have to do when running volume, or their rules and capability were dif- ferent." Prototype shops typically will have a wider process because they must run many designs through in very low volumes and get it out quickly. Volume shops would tend to have a narrower process win- dow, so you can tailor it to get a high yield out of it. Kind of competing standards there. Holden: I knew Steve Jobs, and I worked with Wozniak, and Apple actually had two differ- ent design organizations. One group worked with R&D to get the prototype in the first it- eration out as quickly as possible. en they had the second group whose job was to make it manufacturable and lower-cost. But two to four months later, competitors would reverse- engineer and start to copy them. en, as they were looking at ramping up, Apple would in- troduce the second, more cost-effective de- sign. Apple knew that getting to market first is not the same thing as cost in manufacturing. Korf: Different goals. en you overlay the fact that there are many companies purchasing from multiple vendors to get the best price. e design is done, so now you go pick your vendor. Well, it's kind of too late for that and do anything of real value to affect the cost or performance, and that's just real life. at will always be a problem. Thompson: Good point. A good fabricator has both a test and validation department that is separate from manufacturing. An example of that would be that many manufacturers can cheat on stuff. Let's say you over- etch something, and you're running high in imped- ance at the TDR tester. Do you hand it back to the sol- der mask guy and he throws another coat on it, and you bring it down to 50 ohms? No. If you did some- thing like that, then the board gets to the cus- tomer and they've got two boards in their hand, one that's built the right way the first time, and the second one that has that extra coat of solder mask. e other thing that happens all the time is customers call and ask, "What's the dielectric constant for P370?" It's as simple as that. I'd say, "Well, what speed? What dielectric?" It's an open- ended question. Korf: Mark, I figured out the other day that teams I worked with have probably seen over a million different part numbers over my career. Impedance vs. operating and test frequen- cy were always conflicting. Many years ago, a company I worked at said, "Timeout. We