Issue link: https://iconnect007.uberflip.com/i/1276973
44 DESIGN007 MAGAZINE I AUGUST 2020 Fugitt: I have examples. Things like—I'll throw the term out there, and you can put yourself to sleep Googling it—"self-intersecting polygons." Almeida: It's just a conversion of the intelligent design into unintelligent Gerber files, and what that entails to drive a machine, as opposed to screen graphics. Fugitt: We're doing a translation, and just to carry that one just a little bit further is it's ille- gal, per the Gerber spec, to get all the CAD system's output and know how you can go any further with that. Almeida: It always goes back to that common issue between design systems and manufactur- ing systems. This is at the end of the process, and it's where the smallest amount of R&D is spent on the manufacturing side of the design tools. And that's a large reason why people are still gravitating toward Gerber files, as opposed to an ODB++ or an IPC-2581 file; they just don't have that experience. Fugitt: I'll go a little bit further with this. I have a major CAD system right now that outputs Gerber data so badly that most users output IPC-274D instead of 274X. If you know the dif- ferences between 274D and 274X, it's almost inconceivable. I love that word. Right now, I have a rep in the Czech Republic, trying to write a translator for 274D—remember that there was no spec 274D—and then I have a user in Germany who sent us an example of 274X from this same CAD system. It's unus- able, bad data. There's this thought in the industry that Gerber is good, and the simple fact is it's not, in a lot of cases. Shaughnessy: But 98% of designs are still out- put in Gerber, right? Almeida: Something like that. It's still very high. I think the other problem you have is that the ownership of the design over the last 10–20 years has been shifting from a special- ist PCB designer, who had a lot of manufac- turing expertise, to the EE, who is often more of a generalist. There's nothing wrong with that, but the EEs now have more ownership of the process, and they're not as well-versed in things like PCB fabrication or assembly. How- ever, they can do everything from chip design right through to the board design and simula- tion, signal integrity, and so forth. The elec- tronic circuit is getting faster and better and more complex but with the trade-off of less manufacturing information. Shaughnessy: Now, the data gets to the fabricator's CAM depart- ment. What typically happens at that stage if the data package comes in and has incomplete or bad data? Almeida: We discussed this in the earlier days of DownStream. We talked to fabricators, and some of them told us that up to 50%—or even more—of the designs they received were put on hold because there was an issue. It either didn't align with their require- ments, or there was ambiguous or missing data. But typically, the design goes on hold, and the production process comes to a stop until what- ever issue they found can be resolved. Some- times, that percentage can even be closer to 100% because it's just ambiguous when they get the information. Fugitt: We spend so much time on database integrity, and now you're putting out data that somebody has to fix before you can manufac- ture this design. And of course, if you've sent it to me, I'm going to fix it one way. But if you send it to Rick, he may fix it another way. Now, you've lost your database integrity. Shaughnessy: What is the impact on the job? What happens when fabricators have to deal with this bad data? Rick Almeida