Issue link: https://iconnect007.uberflip.com/i/1284035
SEPTEMBER 2020 I SMT007 MAGAZINE 17 have dozens of years' worth of experience in the automotive business, and since they are ultimately financially responsible for the reliability, they have done their homework on modeling. Newer players in the automotive market decided they were going to start from scratch and build their own supply chain. It has taken them a little bit more time to step into understanding long term reliability because they've had to start from scratch, they're the "new" kids on the block. They're bigger companies, so they're able to spend that kind of money and effort to do that. But if you're a small manufacturer, you may not have the time or money necessary to do an appropriate job of modeling for reliability. In today's changing market, it's difficult going out there to say, "I need to model this," and spend all the money and time to understand it. You end up copying something from somebody else or saying, "This works for my competitor," or, "Let's just do the gorilla test. If it survives, that's good enough." A lot of that goes around, where people have just said, "I'm not going to spend the time. I'm not going to do a reliability test. I'm going to do a robustness test, and I'm going to call it a reliability test." "If it's robust enough to survive this particular set of tortures, it's surely enough to last in my…" which may be true. That very well could be the case. But people tend to error on the side of conservancy. If you're doing the robustness test, you may end up spending more money than you really need to in your supply chain. Your product may be more expensive because of that because you've had to over-engineer it to be sure, and it may make you a lot less competitive in the marketplace. Johnson: How does one approach this as a smaller manufacturer? Is one of those ways to use robustness instead of reliability? Neves: Robustness is what most people end up doing. They call it reliability testing. People mislabel it as an ongoing long-term reliability test. But in reality, they're doing a robustness test. They truly don't understand the correla- tion between what they're testing and what's going on in their product in the field. They're setting a bar. What's happening right now, at least on the PCB side, is people are saying, "We have to do this reflow preconditioning for soldering. Why don't we just expand on that? Instead of just doing five reflows, let's do 10. If it survives 10 reflows, it's going to be okay in our environment." You start to lose any sort of correlation between what you're doing and what's going on in the field because you have so many additional acceleration factors. It doesn't make a lot of sense anymore. If you fail this robustness test, it doesn't necessarily mean that you're not reliable in a given envi- ronment. It could mean that, but it might not mean that. A lot of products are thrown away today because people aren't passing these robustness tests, where the product may be fully reliable in the intended use environment. We have a long history of correlating ugly attributes as being unreliable. We do a lot of visual inspections, such as evaluating cross-sections or looking at solder joints. If it was ugly, we threw it away. We considered it not reliable if it's ugly, and that's just not necessarily the case. Ugly can be reliable. Johnson: We have the ugly test, which doesn't necessarily show reliability but is often used. Then, we have robustness, which doesn't show reliability but has also been used. What other tests have masqueraded for reliability? Neves: Those tests for "ugly" are easy and rela- tively inexpensive to do. That's why they were used. It's easy to hire an inspector to look for physical defects. It's easy to expose a board to more cycles of soldering. It's hard and expen- sive to do longer-term tests where you're look- ing at 30, 40, 60, or 90 days' worth of testing. The hardest thing is to really know where your product's being used and how that correlates to an accelerated test. That's the hardest thing. If people took the time to really know that, and really understand that, working backward toward true reliability would be a lot easier; but few focus on that. Everybody's focusing