Issue link: https://iconnect007.uberflip.com/i/1284035
16 SMT007 MAGAZINE I SEPTEMBER 2020 where these reflow monitoring tests are starting to make their way into the PCB requirements, where people want to do this type of "electrically moni- tored" reflow simulation. There's elec- trical testing during reflow simulation, and then some type of thermal shock regimen afterward to make sure that you get some sort of cycling to show an acceleration of life in the field. Johnson: Could you also revisit another phrase you used earlier when you said, "Reliability is expensive to under- stand?" Neves: Take a vehicle, for instance. We do a lot of automotive testing at the lab. Most automo- tive customers have five or six different envi- ronments that they deal with and that their products need to survive in, such as a single car. You have the interior, on engine, in engine compartment, transmission sensors, brake sensors, etc., and even though they might be in the same car, they're going to be subjected to wildly different environmental and mechanical stresses. What's happening with your transmis- sion sensor versus what's happening in your air conditioning electronics or your stereo sys- tem in your car is going to be very different. It would be easier to model life for the worst case and use that model everywhere (e.g., robust- ness), but that means you will spend a lot of money for parts that don't need that level of reliability to operate for the expected lifetime under less demanding environmental stresses found elsewhere in the vehicle. If I buy parts that are certified for my braking system at 200°C and use them for my stereo, that stereo is going to be a very expensive stereo. It doesn't make economic sense to do that for every environment that you're in. It makes a lot of cost sense to understand exactly where your reliability level needs to be. That will limit your costs for your product as it goes into the field. Automotive companies have spent a lot of time and money modeling the different environments and understanding what it takes for decades worth of life in a vehicle, which is what they have to guarantee. They've taken that model and pulled it back to squeeze it into 30–45 days' worth of testing to say, "I know that if I use this model, I do a Weibull correlation on the data, and I take it out to a decade or two, my product is going to last for that time in the field. If it can survive this accelerated test, in this modeled environment, I know that product will also do that." They've done this for each of the different environmental operating zones that they have in a car. When you qualify a product for a zone in your car, there is a set of specific reliability requirements associated with that zone, depending on the end-use environment. There are a certain number of cycles and specific temperatures or type of environment to subject the parts to. With all the different operating environments that you would expect to see throughout a vehicle, they work to understand and model what it means to accelerate that environment. It has taken them many years to come up with those acceleration models. They're very protective of those models because it cost them millions of dollars and years to create. With companies like Tesla that are relatively new to the game, their approach didn't evolve directly from another manufacturer or Tier 1 supplier. They're not using the procurement model that most established car manufacturers are using, where components are purchased from Tier 1 suppliers (HELLA, Continental, Bosch, Aptiv, etc.). These Tier 1 suppliers