Issue link: https://iconnect007.uberflip.com/i/1508761
OCTOBER 2023 I SMT007 MAGAZINE 55 e newer designs have access to all the data. e key is disseminating that data to the test service. We still see many long-life boards that are five, 10, 15, 20 years old. Just the bare min- imum of the data is there. Avionics is a good example of that, where you'll see a system that was designed 10 or 15 years ago, and they will still be building it for the next 10 or 15 years. Data and documentation are the challenges. Maybe we don't have an updated schematic, or it's been scanned three times from paper and it's hard to read. Has current test equipment made the older legacy work better or easier for you? We can reverse engineer a board today. I can take a raw PCB and create Gerber files. We can create ASCII CAD. One must ask, "Do you want your test guys to re-engineer your board to come out with the design files, or is it cheaper to go back to a designer and get updated data?" We pro- cess Gerbers all the time here, but at the same time, it's a manual or semi-manual step. You're suggesting that for some long-life designs, it's time to bring them into the 21st century? Yes, we still see stuff that is 1980s technol- ogy that needs to be built and tested. ere are other times you must break free and say, "Look, we need to update that design data." Let's talk about the user interface improve- ments you mentioned. What's going on there, and how does user interface development help? We don't have to use the native tester so- ware to do everything. We use the soware to process certain BOMs and CAD files. With these interchangeable solutions, they have all made it more automated, reducing the chance of human error, even from a newer developer of tests. Now the challenge becomes ensuring that the data supplied is current and correct. Can you quantify that? In 2018, for example, what it was like to set up and test a brand- new job? Compare the effort involved then to what you're doing now. How has that changed? e tools are more automated in handling certain known factors and setting up for- mat of data is more flexible. ey're building more intel- ligence into the parsing tools. Back then, we had to describe things a little bit more in the parsing algorithm. Now it's built in with its own level of intelligence. When we pull in a new BOM, it's intuitive: It will see certain measurements and description files, and it recog- nizes the data type. Would you characterize it as predictive? I would say it is predictive to a certain level. If it pre- dicts wrong, it's a lot easier to clean up one or two errors than to describe the whole board to the tester. at process is a lot smoother than to go through the development cycle by brute force. at's what test engineers had to do back then. How long would a setup process take? In 2016 to 2018, a 300-node test job took a full day just bringing in the data. Now, it's probably 20–25% more efficient, and it seems to be get- ting better each time we talk to our partners; we're pretty active in the dialogue with all the ATP in the tester manufacturers. ey'll say, "By the way, it came out last week." e big guys are very much leading the way. ey're listening, and they keep my tester relevant. We can reverse engineer a board today. I can take a raw PCB and create Gerber files.