Issue link: https://iconnect007.uberflip.com/i/1245291
18 DESIGN007 MAGAZINE I MAY 2020 Korf: It is in bad times. There are two approach- es that have worked in the past. One is, I've presented to small companies, and the VP of engineering is sitting in there, and I say, "You pay me more money to do a three-day turn, but the cycle time starts after the approval of the TQs. We both agree what the final layouts are, then the clock starts. If we take three days to go back and forth to get all the answers right, you are paying me three days of mon- ey for a six-day turn. They say, "Oh, that's in- teresting. It's actually a six-day turn from our standpoint." The other way is we've worked with companies that don't want to have issues because they understand it's a cycle time im- pact. Their management wrote it into their re- views saying, "Part of your KPI is to reduce TQs by X% every year." This way, they had some skin in the game also. That has worked very well. I also worked with a consumer company. We went from 20-some TQs per part number—and these were little tiny thumbnail-sized boards— down to about two or three. It was a tremen- dous improvement. We were getting near zero, and then, all of sudden, we're back up to 15 or 20 again. I flew to the U.S., sat down with them, and said, "What happened?" They looked to each other and said, "Oh, we went from our in-house design to subcontracting de- sign. We forgot to tell them what the rules are." They made an unintentional mistake, then cor- rected it. Matties: You keyed in on an important aspect of this: time. We did that issue on design eco- nomics, looking at the cost of design. The cost of re-spins is incredibly expensive, and it seems to me that people would be highly moti- vated to avoid that at almost any cost—the old adage, "There's no time to do it right but plen- ty of time to do it over." Korf: I've done this—and what I'm going to say is not a negative opinion—but there are shops that all they do is quick turn, meaning 1–3 days, for instance; get it in and get it out. By default, some of them don't have time to do the back and forth with you. The order must be on the floor in an hour. They will go ahead and look at it and say, "I'll yield enough to get four pieces out. I'll put four panels in the process, and if we get one panel to work, we're okay. It's factored into our pricing." The other ap- proach is, "We'll change it to make it work and not tell them. We don't have time." The cus- tomer gets a board back, and it's different than what they sent, but they weren't notified that it was modified. Then, when they go to the vol- ume shop, and it is thoroughly reviewed, the same issues are brought up this time for reso- lution. But the customer says, "Wait a second. Our prototype shop didn't have that problem." Actually, they did, or they changed it, and you didn't know and couldn't tell during testing. Sometimes, it's just a capability mismatch be- tween the proto and volume plants. Matties: What's built in prototype will often not work in production. Korf: The yields may be too low and won't meet the production cost goals, etc. I would re-state it that they may not work. It depends on how close the design is to the fabricator capability. Matties: Or material supply is perhaps differ- ent as well. Korf: Yes, I had a customer that went through full FDA approval based on a prototype shop they used in the U.S. They transferred it to Asia, and we said, "We can't get that material in Asia because they don't ship that here." We got it for them. They paid an arm and a leg for that board because we had to ship to Asia be- cause no one uses it there. They didn't know that. It wasn't their fault. Shaughnessy: Designers say that sometimes they don't know where it's going to be proto- typed, much less where it's going to be built in volume. Korf: Sometimes, purchasing waits to the last second to pick a supplier based on delivery, cost, etc. They may have three, four, or even five vendors that they send that technology to