A critical concept that many production engineers fail to grasp is that process tweaks have unintended consequences, and at 20 years into a process, it's highly unlikely that you will stumble upon a beneficial process tweak. In fact, it's highly likely that such a process tweak will produce grossly suboptimal results.
I was involved in a 1986 tigerteam investigation into why an EPROM process was failing miserably, to the point that writing data into one location would erase data from other locations. Turns out that that production engineer came up with the brilliant idea to increase the longevity of the diffusion furnace tube by decreasing the EPROM oxidation process temperature from 1100°C to 1050°C. Increasing the process time accordingly, and the oxide appeared on the test patterns to match perfectly with the previous runs, there appeared to be no apparent loss in yield, and the furnace tube life increased by a factor of 4. Obviously, a win-win.
2 yrs later, we ran an EPROM memory device on the same process, and instead of getting 40% yield we got 0.3%. This was traced, EVENTUALLY, to the the aformentioned oxidation process. While others concocted various "fixes," I concentrated on looking at related process literature, which I eventually found. A full 15 years prior to our ingenious production engineer's wonderful tweak, the engineers at IBM published an article fully explaining the deleterious effects of reduced temperatures for that oxidation step, fully describing the exact effects that we were seeing on our devices. Changing the temperature back to 1100°C immediately improved our yields to over 30% on the very first run, and subsequent runs had even better yields.
So, why did the other product not show the same effect? It turns out that it did, but it has so many other yield loss mechanisms that the EPROM yield hit just wasn't obvious enough. Had the production engineer done a thorough process yield analysis, he probably would have seen the effects.
The moral of the story is that a production engineer MUST do a massively thorough due diligence search before making any process changes. There will have been hundreds of prior production engineers looking for the same sort of changes, and they will have likely published why they were bad ideas.
TTFN
FAQ731-376
Chinese prisoner wins Nobel Peace Prize