Wouter van Ooijen wrote: > I wrote: > > Think of it as a specific instance of the more general > > process of margin testing your overall design. > > > > You DO margin test your designs, right? > > My *design* yes, but I don't margin-test each produced item against all > margins (there are many more than the few you mention :( ). There are two kinds of margin testing, the kind you do in the lab (all margins) on prototypes before going into production, and then there's a second kind (external or environmental margins) that you do on a sample of production machines. The latter helps catch the statistical tail ends that are too easy to miss, overlook or even disregard in the lab. At Apollo, we called the latter "DVT" (design verification testing). We would take a number of early production machines, typically 10 to 12 of them, and cycle all of them through all of the corner cases that I listed previously. Each machine would run 24 to 48 hours at each corner, and this really helped to shake out any weak points in either the components or the design. Later, we would take a larger sample of machines (20 to 30) and run them continuously under more-or-less nominal conditions in a process we called "DMT" (design maturity testing). This helped to catch any reliability (infant mortality) issues. DVT and DMT were run by personnel other than the design team, which added a lot of formality to the process, and every issue that cropped up had to be addressed one way or another. And yes, the process was expensive, in terms of the amount of equipment tied up in the tests, the staff to run the tests, and the time delay in getting to full production. However, it made a huge difference in terms of customer satisfaction with our reliability, and in long-term support costs (warranty repairs/returns). -- Dave Tweed -- http://www.piclist.com PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist