Issue link: https://iconnect007.uberflip.com/i/1284035
SEPTEMBER 2020 I SMT007 MAGAZINE 13 Accepting this, reliability testing becomes a capability test. "I am capable of being reliable, but I don't know necessarily if my product, as I build it over time, is going to be reliable for the customer." I do my upfront reliability test- ing and say, "I'm capable of being reliable. I then have to monitor all the features that could vary, in the process of making my product, and make sure they don't vary in a way that would affect the reliability level I have established." I'm going to understand what variation does to my reliability. I may take lows and highs and run some reliability testing, so I can under- stand, "If this happens to the board, or that happens to the component, or this happens to the solder joint, this is how it's going to skew my reliability results." Matties: Often, it may be using different man- ufacturers for their pre-production run versus where they get their final quantity built. Neves: In today's environment, it's every com- ponent on the board. You have different man- ufacturers. You have different materials you are sourcing. You have all your trusted source issues associated with that as well and making sure you're getting what you really want—not seconds or copies. You're focusing on getting the product that you expected. Then, you have the normal variation in processing or switch- ing from one component supplier to another component supplier. There's a variety of issues that you have to monitor to ensure your reli- ability doesn't change from that first big long test you did to make sure your product is reli- able enough to be used in the field. Matties: Are there any repeating defects that you see, time and time again? Neves: When you start to look at the variables, the more complex the component, the more likely the component is to fail. When you look at a resistor or a capacitor, each of those typi- cally has <50 processes that it goes through to get to the finished product. Your opportunity for failure drops dramatically when compared to the many processes required to create a PCB because the component itself really doesn't have a lot of variables that can go wrong. It's a lot easier for the manufacturer to control that, and to have a good quality system in place that monitors those fewer processes very well and makes sure that the product continues to be repeatable as they manufacture it. Moving over to things like transformers, PCBs, or other sub-assemblies that have into the hundreds of processes to create them, now you start getting interactions that become very difficult for the manufacturer to fully account for. Variation inherently creeps into the process via dependencies, and that ultimately affects the reliability of the product in the field. Johnson: Earlier, you said, "Soldering simula- tion," in the context of preparing for testing. What is soldering simulation? Neves: When components were primarily placed in holes, we used to do the solder dip test. We would take a coupon representing the PCB and dip it into a vat of solder to trans- fer a similar amount of heat to the coupon, as you would see in a wave solder operation. These days, it's mostly convection air reflow assembly and hand soldering for repair or some other sort of feature attachment that you can't do with an air type of reflow. In the field, at least on the PCB side, we have a variety of tests that put the boards through a simulated reflow process. These simulations try to get the surface temperature of the board the same as you would with components in a worst- case scenario of as many times and for the longest exposure that you allow the finished product to endure. However, many times, you would reflow solder the board and then possi- bly rework or repair the board afterward. The simulation of the component attachment pro- cess usually comprises four to six times of a severe reflow exposure. That's pretty typical in the industry. Someone would run a board through a worst-case reflow process four to six times. Understanding the most extreme dam- age that could happen during the reflow pro- cess on production boards, you try to simu- late that same kind of damage and stress to