Issue link: https://iconnect007.uberflip.com/i/622045
January 2016 • SMT Magazine 57 Goldman: in your paper do you discuss the cours- es? Borkes: No, that can be found at our website. This paper is very specific in its scope because one of the things you find out, especially if you're a contract manufacturer or an EMS com- pany for example, is that you'll get a job to quote, and the product may not be designed for automation, so you're at a big disadvantage because you want to be competitive and win the job, but yet you know that the way this is designed you're not going to be able to be suc- cessful at automating it. The paper I'm present- ing here at SMTAI specifically talks about de- signing for automation and an important tool that should be part of every high tech electronic product assembly's infrastructure. Goldman: not just for manufacturability, but au- tomated manufacturability. Borkes: Absolutely right. In the paper we do a case study that shows how, in very dramatic terms, we can do that. As long as you do it in the design phase you'll be successful in manu- facturing. If you design it without that thought process in mind, you put the manufacturer in a very difficult position where no matter what they do they're not going to be able to success- fully achieve those high yield rates. Goldman: ideally the designer should be working with the manufacturer from the beginning. Borkes: That's always been the DFMA mentality. If you're an OEM that actually designs your own products you have a better shot, but even then it can be difficult. I do a lot of consulting work and my clients sometimes seem like they are three or four different companies because the depart- ments are so self-contained, with the proverbial silo mentality. A lot of the time it seems they're competing against one another within the same company. That's one thing we do in the paper: describe a case study on automation design and coordination, and the dramatic effect that it can have on the overall labor cost. The other part of the paper addresses au- tomated process control. Years ago we started to talk about the need for what we call proac- tive process control. It's one thing to design the product for automation, but when you get into a production run, there's going to be variation in your process. The robustness of the process (your process window) will determine how well that process will respond to the variation of the different independent variables as they change over time. In the paper we suggest that to achieve the yields needed to compete we have to go beyond what I have called proactive process control, where you don't wait for a pro- cess to go out of control before you stop the line and fix whatever is causing the problem. Pro- active process control suggests that you want to constantly monitor the process in real time, and when the process parameters start to vary in a non-random way, you get an alert. A flag goes up and someone realizes, "I don't have any defects yet, but we're headed that way." There is what we call in statistical process control an "assignable cause" that has entered the process. That's proactive process control. In the paper, what we're suggesting is that to achieve those 99.5% and above yield rates you have to go one step further. I call it "meta- FEATurE inTErViEw THE JEFFErSOn PrOJECT Figure 1: passive SMT component family (with english designations) including their patriarch: the axial-leaded resistor (all on a Jefferson nickel).