Issue link: https://iconnect007.uberflip.com/i/942744
FEBRUARY 2018 I DESIGN007 MAGAZINE 29 Gorajia: We want to tie these together. First, you want to simulate as much up front, first- time-right design. The second is feeding back; that's the big thing that's missing in today's world. Shaughnessy: Because the manufacturer will fix the crap data and just update the board and never even tell the designer. Gorajia: The designers don't even know what's happening. Matties: So many people say that designers really need to understand the manufacturing process. To a degree, yes, but there's the other side that says the designers just need feed- back. The lag time between design and completed board, they're on to other designs by then and it's not a natural feedback loop. How do you improve that feedback loop so it's more automated? Gorajia: First of all, in order to solve the problem, it needs to be a systematic or systems- based feedback, not a people- based feedback. People obviously will be the authors of feedback. But the system should drive the communication. As you said, it might not be what either side is working on right then and there…they've moved on to some- thing else. For example, if a design organiza- tion is going to have an EMS partner run design reviews before production, is there a platform in which they can collaborate and capture that feedback such that both organizations can then monitor those recommendations that were done or not done, accepted, or waived? Is it a repeated violation? Does it show up two revisions later? Does the same design group have a consistent DFM violation compared to others? We need tools, process, people. We need a business process that says, "The last five designs that we've done with these guys, they've recommended these changes. Can we then take it and add a process that updates our design characteris tics, DRCs, the design rules, the DFM rules, the signal integrity rules, in order to compensate for all of these recurring problems in manufacturing?" That's step two. Then step three, how can we pull real quality data back i nto the design so that DFM is based on actuals? I'll give you an example of that. Most of the time, the bill of materials on the design side are chosen by an engineer who's familiar with the product and has worked with various components for some functional ben - efit. Usually, in every bill of materials there are three or four options of which part to use in manufacturing, many times called alternate parts. Which one is the best? That is usually determined in manufacturing by what best cost they get. If they have a choice of three, and one is already in inven - tory, that's usually the first one they choose. If it's not in inventory, then which one has the lead time with the target cost? In manufacturing, that makes sense. In design, when they're picking parts based on form, fit, and function, where's cost? What ends up happening is there isn't any system- atic way of understanding what the risks are when you select a part. A lot of designers now are moving towards integrating their upfront library management, component management process, with various suppliers. They're doing that part great. What they're missing, though, is with my manufacturing partner, whether it's an EMS A, B, C, or whether it's their inte- grated manufacturing line, what if we can feed back defect rates for a specific component or component vendor or distributor? That data is still missing. If you're integrating with IHS or Silicon Expert or any 3rd party component aggregator, and you can get all the parametric data, cool. You can get cost data, cool. But for the EMS partner, if a component has a higher defect rate compared to another, it can really