EDITOR'S PAGE

TI's crown jewel

Just off the LBJ Expressway in Dallas lies what will soon be the crown jewel of Texas Instruments' production facilities—DMOS 6. Originally built in 1996 as a 200-mm fab but mothballed when demand turned south, the once-empty shell is quickly filling up with equipment and support systems as teams of engineers and technicians ramp the facility for 300-mm production. During a late-March trip to Texas, I toured the fab and spoke with some key personnel bringing DMOS 6 on-line.

The largest fab in the company's family, DMOS 6's main cleanroom area sprawls 118,000 sq ft over the freestanding waffle slab, with another 32,000 sq ft of adjacent clean manufacturing area for CMP, wafer recycling, pod services, and the like. A familiar golden-orange rectangular glow delineates the lithography area in the center of the huge production floor, which is about as large as two football fields. Not surprisingly, the proportions of the equipment match the scale of the room itself—from miniwarehouse-like implanters, to windowed wet benches reminiscent of streetcars, to a prodigious cluster tool called by one engineer "an Endura on steroids."

DMOS 6 will be a fully automated, computer-integrated manufacturing (CIM) facility. Every wafer will be tracked on the CIM system, with the overhead transport system carrying color-coded wafer pods (green for nonmetal, red for metal, amber for copper) to and from the prodigious stockers and the various process and metrology cells. Because minienvironment technology is used, the main cleanroom operates at Class 100 and features a layout strikingly different from traditional ballroom fabs—there are no bays and chases. Besides opening up the workspace, the scheme has other obvious benefits: the tools' subsystems and components are readily accessible for fine-tuning or maintenance.

Since the fab was in the engineering phase during my visit, most of the wafers were still transferred manually, but the one operational litho cluster—a rambling, 36-ft-long scanner/track combo—was benefiting from the first deliveries of the automated materials handling system. When the fab reaches production levels, the peripatetic robots will carry a hell of a lot of silicon: the final capacity could reach 35,000+ wafer starts per week (a whopping 78,000+ in 200-mm equivalent wafers, if you use the 2.25x ratio).

Greg Winterton, DMOS 6's engineering manager, was part of the original 200-mm project group and has worked on the 300-mm project core team for more than 18 months. "We started receiving tools last September, and we continue to receive and install tools. [More than 180 tools had been received by late March.] We started silicon in mid-February, and we got transistors out and they look pretty good. The transistors match nearly to our mother fab [the Kilby Fab, or K-Fab] and to the spec. We'll have yielding devices out of here in the next month or so."

TI's plan is to qualify the 0.18-µm aluminum interconnect module this September, quickly followed by the 0.13-µm copper process in December, according to Winterton. "We're starting up on our proven high-yielding aluminum interconnect process. The process is running in four other factories and the technology is well understood. So if I bring it in here, I can see what squeaks and rattles in the process flow and I can be guaranteed that it's not some marginality with new technology—it's an issue with our integration of the tool set.

"There's been a couple of minor integration issues going to 300 mm. We do have a full tool set in place for processing but many are one-of-a-kind tools, so we're suffering some when a tool goes down for a couple of days. But the results I've gotten so far on 300 mm closely match what I've seen on 200 mm.... You plan for the worst, you hope for the best, and it's actually been more toward the better: all the things I feared really haven't happened. Things like CMP on 300 mm—it's actually better than 200 mm."

A veteran engineer who has seen his share of fab follies, Winterton nonetheless seems cautiously optimistic about the DMOS 6 ramp. "It's been easier than I thought it would be thus far. That's not to say there isn't some land mine out there, but we haven't stepped on it yet."

Tom Cheyney
Editor

tom.cheyney@cancom.com

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit