New roadmap targets cycle time, waste
06/01/2008
Assume that a semiconductor manufacturing factory is completely empty, with all resources ready to run. You have just one wafer to process–the hottest hot lot of all time–and you have to do so ASAP. That, said Eric Englhardt, senior director, systems engineering, Applied Materials, is the theoretical, minimum cycle time, or SingleWaferCTmin.
Now imagine that it’s a real fab, where cycle time is far from this ideal, slowed by the challenges of maximizing production and the inefficiencies of the real world. Wafers are slowed by unscheduled equipment downtime and by simply waiting to be processed. Even after processing has begun, individual wafers in the carrier wait for other wafers from the carrier to finish processing. At inspection or metrology tools, all wafers in the carrier must wait while only a few of the wafers are sampled.
Fab productivity and cycle time can also be hampered by inefficient use of factory resources, said Englhardt. Production equipment often remains idle, even when work–in–process (WIP) is available for processing. This is sometimes due to tool setup required when process recipes are changed. While most efforts have been aimed at improving tool and factory productivity, that can be a trap, claims Taichii Ohno, the “father” of the Toyota Production System. He says to instead focus on waste: Productivity will come. Englhardt believes the semiconductor industry has done an excellent job reducing silicon waste through scaling, but today, more and more of that waste has been redistributed in the form of increased cycle time and lost equipment productivity. Englhardt has suggested a new metric to capture the difference between minimum and actual cycle time called WaitTimeWaste, defined here:
 |
“The WaitTimeWaste equation can be applied to a tool, a module, or the whole factory,” Englhardt said. “All of that is waste that we put on the factory to try to maximize factors other than cycle time.”
A second similar metric is used to define equipment waste, or the fraction of potential output that is not realized.
 |
A third metric would be energy and resource waste.
Englhardt described a hypo–thetical, yet typical, tool that can do 100 wafers/hr and has 90% availability. “You should be able to get 90 wafers/hr out of that tool. That’s how we’ve designed it,” he said. “But customers [do not get close to that] as an average once they put it into the factory; in fact, we don’t really know what they get because we don’t get all the information.”
Englhardt said most equipment companies have programs to continue to expand throughput, say, to go from 100 wafers/hr to 120 wafers/hr in capability. “That’s something equipment companies can easily measure internally to their operations in their own labs. But, the opportunity exists to figure out how to improve other parts of that efficiency. Perhaps we can invest more effectively to go from 60 wafers/hr of real production to 70 or 80 wafers/hr, rather than just increasing the top–level spec.”
What Englhardt is proposing is a new roadmap that incorporates the three waste metrics described above. “In order to systematically and aggressively eliminate these wastes, the baseline level of waste in today’s factories must first be quantified; since waste is not explicitly measured in the semiconductor industry today, many fabs may not even know how much waste exists,” says Englhardt. “Next, a roadmap must be established to reduce this waste by 30% every 2 to 2.5 years and 50% every 4 to 5 years.” (Eric’s boss, Applied’s president and CEO Mike Splinter, has challenged the industy [during his ISSM keynote address last year] to aim for 50% reduction in cycle time and energy usage.)
Englhardt’s Roadmap suggests moving WaitTimeWaste from 14 in 2008 to 11.8 in 2009, 9.9 in 2020, 8.3 in 2011, 7.0 in 2012, etc. “Commitment will be required from factories and suppliers to agree on the right values for these metrics, publish the roadmap, collaborate to develop and implement solutions, and relentlessly follow the roadmap to reduce factory waste.”
Englhardt believes (and I concur) that by concurrently attacking the scaling and waste reduction roadmaps, we can eradicate waste instead of redistributing it. The result will be real productivity gains for the entire industry. The industry’s present struggle to understand the economics of moving to 450mm wafer makes it even more of an imperative.
 |
Pete Singer
Editor–in–Chief