Issue link: https://iconnect007.uberflip.com/i/1514628
JANUARY 2024 I PCB007 MAGAZINE 13 the board. Generating a lot of TQs to under- stand the design intent is mostly frowned upon because it increases the production tooling release time, leaving less time on the manufac- turing floor to meet shipping commitments. Initial qualification orders may also have short lead times. e manufacturer qualifica- tion team (if there is one) will hand-walk the design through the process and make real-time judgment calls when issues arise. ere isn't much time to synchronize with the designer. is is especially true when the designer is one level removed from the PCB fabricator, as when the board is ordered from an EMS sup- plier. Multiple part numbers should be ordered from the same supplier/plant to allow the designer and manufacturer to get in sync. e front-end engineer will see common documen- tation errors and solutions, along with com- mon answers to TQs that are generated. is is the start of the designer knowledge transfer to manufacturing, resulting in reduced cycle times and improved PCB quality and perfor- mance. When the customer continues to shop for different suppliers to get the best lead time, lowest cost, etc., that makes it nearly impossi- ble to establish a bi-directional technical rela- tionship between the designer and PCB fabri- cator. DFM Software to the Rescue? Millions of dollars have been spent on designer DFM and fabricator engineering/ CAM soware, yet it hasn't fully solved the problem of unmanufacturable data packages being released for production. Why is this? First, fabricators don't provide enough process capability and rules to cus- tomers or DFM soware developers. ere are multi-variable conditions and logic that can't be provided in a simple data table format. Pro- cess engineering oen doesn't provide suffi- cient process rules to their own front-end engi- neering teams, resulting in conducting meet- ings to resolve questions. It can take a year to train a new product engineer to become pro- ficient with the internal rules and many cus- tomer requirements. eCAD/mCAD DFM soware also doesn't have all the manufacturer's rules available. Material suppliers don't provide broadband TDR-based dielectric constants to map mate- rials to the intended impedance values. TDR measurements are the most specified imped- ance test method. Fabricators must perform their own testing to generate these values. So, the analysis soware both parties are using with supplied material values can't be relied on to yield a perfect stackup. Designers generally don't have sufficient allocated time to review multiple fabricators' rules, which allows them to develop an aver- age set of rules that can be successfully used by all the approved fabricators. Many large com- panies do require a nearly common set of rules to be used by all suppliers because they have technology teams that are separate from the layout engineer to develop this. Generative AI and Digital Twin to the Rescue? Alas, generative AI, along with all the manu- facturing and processing data in the PCB digi- tal twin will not solve the problem. e theory states that the design intent digital twin can be combined with the production lot-by-lot pro- cess and quality data using AI to yield the per- fect data package for future designs. ere is a minor problem with this solution: It doesn't exist. is has the most potential to bond the designer to the front-end engineer. e multi- variable DFM requirements, along with the information garnered from prior production of the down revision or similar part number, could be answered by computers. is would reduce the human communication interchange to resolve the most complex issues. Unfor- tunately, PCB manufacturing is a low-mar- gin business, with most companies maintain-