Issue link: https://iconnect007.uberflip.com/i/1245291
16 DESIGN007 MAGAZINE I MAY 2020 lem. You see a lot of designs where a person lays the board out, a CM will do the assembly, and they'll pick the fabricator based on what- ever their criteria are. The designer has no clue who's going to build it, let alone what the rules are. A lot of times, the CM will impose con- straints on top of the design that the designer didn't have and add another layer of complex- ity to the problem. To your point, over the years, my fabrica- tion pre-engineers have "DFM-ed" hundreds of thousands of part numbers—probably more than a million part numbers over my career. Some of these would be for a company they've been building for for over 10 years, and they're still coming back bad. Why? "It takes too long to update the libraries." "It takes too long to update the rules and get them all approved." "We have three vendors, and we can't get them all to agree," which is false. I've been through many successful standardization exercises on that one. There are lots of excuses why they can't do it. Another generic issue which is kind of inter- esting: Say I get revision A of a board. We go back and forth and resolve all the issues dur- ing the DFM process. Then, we all agree on the design changes we're going to have to do as the fabricator. When revision B comes in, the issues are still there. A lot of companies don't generally go back and update their de- signs with the issues so that revision B comes out clean. It's a standard practice of board fab- ricators to run the DFM review again on re- vision B, compare that against the down rev, and you'll see all the same issues—maybe a few more, maybe a few less. One of the TQs that's sent back on revision B is, "Can I do all the edits I did on revision A that you approved? Here's what you approved." It's kind of silly. Barry Matties: What's the incentive for the de- signer to send complete and accurate docu- mentation to a fabricator? Korf: I'll give you another good example of why your question is valid. I had a large custom- er request a pre-laid out, pre-released DFM re- quest. The Gerber data had a round hole in it, and the fab print had a slot. My DFM en- gineer sent back a TQ, questioning, "Which one do you want? Do you want the slot, or do you want the hole?" Unfortunately, the cus- tomer replied in the same email with two at- tachments—one reply was the hole, the other reply was the slot. Two different answers in the same email. We filed the response away because we weren't going to build the board for a while. An order comes in a few months later with the same design documentation er- ror still there. Unfortunately, the engineer se- lected the hole; they didn't notice there were two answers. We built the board, and it came back saying that there was supposed to be a slot, not a hole. They said, "How could you possibly do that?" I said, "We asked you which one it was, and you said both of them were okay." We responded that we should have caught it and got back to you, but it's your design mis- take; your documentation was wrong. I wrote a corrective action back to the customers saying the root cause was their design documentation. Again, it was a large company, and I wrote to them that the root cause in the report that their documentation was wrong. Oh boy, everyone freaked out on that one because, in Asia, you don't tell your customers they're wrong—espe- cially big ones. We argued over that. We paid for the mistake, even though they were the root cause. As I mentioned earlier, in the '60s, some fabricator decided to rebuild a board for free because they didn't catch the customer's error, and that became standard practice. Matties: How do you incentivize it? A fabrica- tor could say, "We're going to give you X% dis- count if you send us the complete and accurate data package, and we're going to conversely try to do X% more if you don't." Korf: I've tried that in a couple of ways. One, we'll charge you X% if you don't send me a good one. They went, "What? The other sup- pliers don't add money, sorry, we're gone." They won't come back. Matties: Is that necessarily a bad thing?