Issue link: https://iconnect007.uberflip.com/i/1284035
12 SMT007 MAGAZINE I SEPTEMBER 2020 understand what's going on in the field, so they'll add more heat, cold, cycles, or vibra- tion to try and be tough on the product. They'll do some sort of acceleration that may or may not correlate to what's going on in the field. I get these all the time. People say, "I want 500– 1,000 cycles of that." It reminds me of the American Tourister and Samsonite commercials from the '70s and '80s, where they used to take the suitcase and put it in the cage with the gorilla. The gorilla would beat this around for a while. If it didn't break open, it was reliable—a true robustness test. Sadly, that gorilla type of robustness approach is a lot of the testing that's done today. They just say, "If it holds up to the gorilla, then it's going to hold up for my customer." That's an attitude, at various stages, that's usually cheap and easy to do. People have some confidence in that. That's what most "reliability testing" has become these days—a robustness test— because there's a disconnect between what's being done in the testing and what's going on in the field. It's easy, cheap, or they're copy- ing what a competitor does or using an indus- try standard. But they're not going through the process of understanding what the correlation is between the testing that they're doing and what's happening to their product in the field. Matties: Is part of the issue what they may do the reliability testing on, such as a prototype run, but the mass production run yields differ- ent results? How do you overcome that? Neves: The assumption is that, whatever you're going to run your reliability testing on, the variation in the product that you're manufac- turing isn't going to change much from when you run the test. That is rarely the case with PCBs as batch to batch, run to run, and lot to lot, introduce many possibilities for variation in the hundreds of processes that are needed to create a PCB. shows that it's gone through this component attachment process, getting you all the way up to the point of where it would be going out the door. That's one feature that not everyone does before testing. A lot of people will take a com- ponent and then do testing on that component as it comes in the door, or they'll electrically test a PCB before component attachment and say, "Every- thing's great." But any type of testing that you do to see how long your product might last needs to have some sort of simulation of the soldering process before testing. That simula- tion also needs to include the rework and repair part of the component attachment process. Reliability is expensive to understand because you first have to understand where your prod- uct is being used, how it's being used, and the environment that it's going to be used in. Once you understand these things, you have to find a way to accelerate them in a reason- able amount of time. You don't want to wait 10 years to see if your product has a 10-year life. You have to find some way to acceler- ate product life, which keeps the connection between what you're doing to it in an accel- erated way and what your customer is going to do to it out in the field. That takes a lot of studying and isolating all the individual attri- butes that go into what your customers will do to your product and how to accelerate that in a controlled manner without adding or remov- ing factors that contribute to failure. Whether it's an environmental acceleration, a mechan- ical acceleration, or all the other weird things that are going to happen to your product, it's about how to accelerate those in a way that you can use the results from that acceleration to directly understand "in the field" influences on your product's life. Most people don't do that. Most people are happy to do more of a robustness type of test. They want to do something that's worse than the customer is going to do to it. They don't really Bob Neves