|
ISMI Symposium drilldown: Weather.com control knob now on an atmospheric oxidation furnace near you |
|
|
|
Oct 25, 2007 at 09:22 AM |
Advanced process control (APC) has come a long way in the diffusion furnaces at Spansion's Fab 25 in Austin.
In a presentation from Chris Lanford today at the ISMI Symposium, he described numerous process improvements they've seen since installing their first APC controller on a tool in 2000. ("APC controller"---isn't that redundant?)
Since then, the fab has transitioned from logic to flash (and from AMD to Spansion), which meant that the back end of line became less critical and tool/process intensive and the front end---where the furnaces reside---became much more complex, with many additional process recipes and the like in play. Chris said they went from about 35 furnaces in the logic days to about 80 furnaces today, with no real increase in engineers to watch over them (meaning each team member has more than twice as many tools to look after).
The need for solid APC approaches has been acute, so the successes seen by Chris and his colleagues in drastically improved film thickness variabilities/uniformities, CpKs, distribution times, productivity, etc. have allowed the team to spend more time working on dialing in baselines than trying to get the tools to run under control.
One of the more intriguing case studies had to do with how atmospheric pressure changes affect the oxidation rate, and the need for a sturdy tunnel-oxide APC scheme. Chris noted that when a storm blows in toward the fab, the atmospheric pressure changes can lead to increased oxidation rate variation and a higher number of charge-induced fails in the tool. So how does one measure the pressure changes and compensate for the effects of Mother Nature on the process?
First, they borrowed pressure traces from monitors in the fab's lithography cell, since as Chris put it, "the litho guys didn't care about pressure, they only cared about changes in pressure." But the litho area is on the other side of the fab from the furnaces, so there was some logistical inconvenience. To perfect a feedforward tunnel oxide APC model, they needed to get the normalized pressure and oxidation rate into line and add an atmospheric pressure knob to the controller.
Then someone had a bright idea: Instead of relying on the monitor in the litho cell, why not just check the pressure readings from the nearby Austin airport at weather.com? As the Guinness guys would say, "brilliant!" What's important is not the pressure inside the fab, as Chris noted, but the reading from the exhaust line, a number pretty much the same as what the airport's barometrics would show. (btw, the pressure-compensated oxidation time results were much better, and now they not only have the feedforward capabilities, but a feedback loop as well, with thickness and metrology results.)
Who knew that a very popular site on the Internet would provide a key control knob on a process furnace's APC box. What other knobs might be out there?
|
Blogs 
