In his first article, Paul discussed silicon manufacturing quality. In the second blog, Paul shares his insight into cost management in the manufacturing process and the factors determining chip units costs. In his third instalment below, Paul discusses how much control you should have during the manufacturing process and what elements matter most.
Where is the Motivation For Success
Let’s say you’ve got a great idea for a new musical Christmas card solution, and you go to a Big Foundry and a Big OSAT and convince them to supply you. Great. But let’s say there are several other musical Christmas card solution providers, and they’re also working with the same group of suppliers. The Big companies can see that the musical Christmas card market is going to grow, and there’s a lot of innovation there, so supplying this market is a strategic objective for them: they don’t care who wins, as long as they’re in the game and are likely to be serving the winner. They don’t care who wins! So why would they work hard to make sure that you win, as long as their other customers include your competitors – someone’s got to win, and they will try and bet on all the potential winners, irrespective of their potential technical superiority. You have to work hard to make sure that you win: nobody else cares as much as you do about your success. So whatever you do in manufacture, you have to make sure that motivation for success is in the right place: the people who control the detail need to care a lot about its outcome.
So how do you decide what you need to be in control of?
The Boundary of Awareness and The Boundary of Control
When you’re having chips made for you, somewhere there is a boundary beyond which you know nothing about how things are done and what the issues and risks might be. At one end of the scale might be the choice of which make and model of photo stepper is used for your chips. At the other end might be the content of the test program that sorts good chips from bad. Somewhere in the middle might be the suppliers of the solder balls that end up on your BGA. How would these differ from each other in their significance? Well, the photo stepper really is someone else’s problem: your time on it is a very small proportion of its total capacity, and if it fails then your foundry has got a Really Big Issue to deal with. The test program, however, is for your product only, and it determines the DPPM (Defective Parts Per Million) rate that your customer will suffer; clearly you care more than anyone else in the world that this is done correctly. And the solder balls supplier you technically don’t really care about, but when a top tier customer asks you about your compliance with the US Dodd-Frank legislation, you’ll start to take an urgent interest in the solder ball suppliers, and the smelters that supply them.
So for all aspects of your product’s manufacture, there is a boundary of control, and a boundary of awareness. You might not need to choose the solder ball supplier, but it might be good to know who it is, or at least know how to find out who it is. That’s awareness, and it’s pretty easy to get, if you know you need it. Control is more difficult, mainly because of the expertise and time required.
The ATE test program is a good example: ATE programming is a specialism that is not essential to product development, so test engineering is frequently considered to be an optional skill in a company’s workforce, and test program writing is easily subcontracted to a specialist supplier. But the test program will contribute massively to a product’s cost and quality characteristics, so you actually care an awful lot about how well it is done. More than anyone else in the world, in fact. So how important is it really that you have this skill in-house? And if you subcontract it, how do you make sure that it is done well enough?
A similar analysis can be done across all things that contribute to a product’s manufacture. This analysis should be done, with the aim of identifying where everything lies relative to the boundaries of awareness and control. Those that need to lie within the boundary of control need to be addressed with high priority: a clear strategy should be defined for their implementation and management. If this isn’t done properly, your company could have some seriously annoying issues to deal with.
Paul will be speaking at the forthcoming Silicon for Systems event at Rutherford Appleton Laboratory on September 3rd. Take me to the Silicon for Systems event page