INDUSTRY NEWS

Special Report

Texas Instruments' Kilby Center links R&D to production floor

The Kilby Center is tucked away in Texas Instruments' sprawling Expressway campus in Dallas. Hailed as one of the leading semiconductor R&D facilities in the world, "the house that Jack (Kilby, that is) built" began operations in late 1997, after 20 months of design, construction, and tool installation phases. The 500 employees there share a four-shift, around-the-clock schedule. The heart of the center is what TI'ers call the K-Fab, a 50,000-sq-ft cleanroom where the company not only carries out all research and development on current and future technologies but also pilots processes and brings them to low-level production.

The K-Fab may not be a huge facility, but it is a veritable showroom of manufacturing technology, including late models of many different pieces of semiconductor process, analytical, and metrology equipment. As fab manufacturing supervisor and tour guide Ron Weingartner puts it, comparing his facility with production sites, "we're going to have one to two to three [tools], they're going to have 10, 15, 20; they have rows, we have a row." Its distinctive 16-ft ceiling (versus the normal 12-ft clearance) was designed to accommodate the increasing vertical footprint of some of the newest machines, he notes, and also makes the work environment less claustrophobic. The yellow-orange safe lighting familiar to anyone who's ever worked in a lithography area is used throughout the relatively open K-Fab cleanroom floor. "One of the things we found out at other places," explains Weingartner, is that "when you're transferring wafers—not just in the photolithography area—you want to make sure that the light is the same no matter where you are on the transfer."

TI developed the strategic concept of integrating research, development, and early production at the K-Fab to achieve better time to market and cost-effectiveness as well as to improve technology learning curves and margins before delivering the new processes to the manufacturing sites, according to Faa-Ching Wang, the fab's manager. "We plan to actually do a small volume, to check out the volume-related variations, check out the good margins. . . and when the market demand increases, then hand it off to the production fab." With the recent sell-off of TI's memory chip business to Micron and increased emphasis on digital signal processors (DSPs) and logic product lines at the company, Wang sees K-Fab's mission as "more focused and easier right now. . . . It used to be our resources were more diluted, stretched much thinner. With a better focus, we are applying the full sets of methodologies only on the technology required for DSPs, and the methodology itself is pretty much a good consolidation of learning accumulated through the years. . . . We have spent a very significant amount of effort to help make the transition, to really let our people understand that this is a state-of-the-art facility with the best people, best-focused product now, and that there's a lot of leadership effort going on to help with this."

This new focus has had a positive impact on defect reduction and yield optimization activities at the K-Fab and throughout the rest of the company. Venu Menon, yield manager at the facility, still deals with TI's global yield strategies portfolio, but his days of living out of a suitcase have greatly decreased, a fact he admits makes his family very happy. Menon says the "yield manager's role is to support yield-enabling technology in the research phase to make sure that processes are ready in that phase. When you move into the latter stages of development, you make sure you have demonstrated yield and margin in small volume. With the new charter, we have to demonstrate yields and benchmark yields, [and do so] with a reasonable amount of wafer volume. We expect to ramp the yields, higher yields, [develop] a more mature process, and hopefully transfer [to the production fabs] at equal or even higher yields."

MICRO editor Tom Cheyney visited the Kilby Center in late October and discussed fab operations and yield management issues as well as TI's new personality with Wang and Menon. The following excerpts from those conversations exude an optimism in the future of TI and offer a realistic assessment of some of the critical challenges on the ever-evolving technological roadmap.

MICRO: Venu, who's working for you, what are the various departments under your tutelage?

Menon: The main folks in my organization are focused around in-line yield enhancement. There are three or four pieces we look at. The first piece is modeling—what are the goals, what are the defect budgets, and so on. The second part is defect inspection methodology; a group is working on implementing that at all these stages—research, development, early production. The third part, which is the core of what we do, is parsing the problem set into the systematic issues and the random particle issues, and having teams that drive solutions to those problems.

MICRO: Do you have much interface with your DMOS 5 [production fab] counterparts?

Menon: Daily. There are joint teams for equipment selection that involve DMOS 5 membership as well as my team. We look at future technology requirements together, decide what equipment or software we need to evaluate and put in place. Even on many of the problem-solving teams, we have membership from both sides.

MICRO: How many people are on your team right now?

Menon: 24 including technicians.

MICRO: Are there also part-timers from the process modules who come to the meetings when it pertains to their part of the line?

Menon: Many of our cross-functional teams have module involvement as well as involvement from product integration groups.

Wang: Venu is one of the key leaders for the TI-wide yield efforts, and the TI teamwork in various areas is very good, either within K-Fab or among the yield community. There are many yield-solving teams led by the yield team that embrace all of the equipment engineering and process engineering [teams] to physically solve problems.

MICRO: In terms of your supplier base, how closely are they involved with you here?

Menon: We work very closely with them. This is very important, especially when you're doing research and development because you're at a point where equipment maturity is not there, so you need additional support. Oftentimes you're trying to codevelop things, when it's a new process. We need a higher level of partnership and so far it's been quite good.

MICRO: Your facility is also one that's testing the vendors' ability to deliver what they promise.

Wang: We're talking about very-high-purity facilities, with all key process critical chemicals at sub-ppb levels.

Menon: One of the unique advantages the K-Fab has, if you look across the industry at most of the research and pilot-line facilities, is that there aren't many that were built in the last two years. Usually what happens is, you have a lot of older fabs, which are research infrastructures for the companies. There you have an uphill battle because you have to constantly upgrade your facility, your chemicals. For us, starting off as a new research fab is an advantage. We don't tend to worry too much about chemical purity being a big yield issue, for example, because we have such good quality here. This allows us to focus on the other yield issues, such as the process equipment, the integration issues, and the design.

MICRO: There are some real show-stoppers ahead, though, unless the chipmakers and vendors figure out how to deal with them.

Menon: In the area of defect inspection and defect detection, for example, we have some serious concerns about the ability of the vendors or suppliers to meet our needs when we are down in the 0.1-µm range, to be able to detect defects and watch the edge contacts and things like that. Looking at high-aspect-ratio structures, how do you see what's at the bottom, on the sides; those are our challenges. . . . Photolithography is a challenge for us, being able to find smaller defects in the photo module. We are concerned about the ability of the industry as a whole to make sure we have the solutions in time.

Wang: As the technology moves to smaller sizes, the complexity is incredible. The requirements also become incredible to make every tool able to do everything we can, lowering costs, running fast; it's hard. It would be good to have some tools that were more versatile, but so far they don't seem able to do that.

MICRO: Because of the development mission of your fab, do you try out a variety of tools?

Menon: We do a bit of that, but at the same time we tend to watch the equipment diversity, since we don't want to get too diverse. It becomes a huge management challenge, in equipment engineering, to maintain the instruments.

MICRO: So you have this combination of being the brand-new tool evaluator but at the same time if you have too many options going on, how do you transfer that to the manufacturing fab that in the future will need those same tools.

Menon: We are very cognizant of the fact that we would like to minimize the changes that our production factories and our development line need to make node to node. That's cost, that's new learning, and it's always an immaturity curve that you have to climb. We only go after new equipment when there's a fairly clear gap that's identified. In almost all these areas we have gap analysis teams, or GATs. They go and look and make sure we have a clear gap identified and then we work with the suppliers in those areas to say, "OK, how do we fill those gaps?"

Wang: The focus we talked about earlier helps somewhat in these aspects. We work very closely with the production fabs, and we try to have the same equipment set and understand their problems, and they try and understand our problems, and so on. Once again, with DSP only, that is simplified.

MICRO: What are the other defect and yield challenges you're dealing with? Which ones are you devoting the most effort to?

Menon: Broadly, our objective, our vision, is that we want to detect the problems in-line, not wait for electrical test. We're spending a lot of time to make sure we have the right tools, methods, sampling and review strategies, autodefect classification, to catch the defect events very quickly in-line. Within that umbrella, photolithography is our number-one area. Because they are pushing the design margins, equipment capabilities, and photoresist capabilities so much, we watch that area very closely to make sure we catch any unique problems.

MICRO: There you have the added challenge of having to deal with the whole reticle issue, not just the wafer.

Menon: I was just going to say that. Getting the right reticle-monitoring tool, working with the reticle vendors, is a key part of that activity for us in K-Fab. For us, the other key area is design-to-process integration, because we are developing new processes from the ground up. We have to understand integration and design issues here. We don't want them ever to go into DMOS 5.

MICRO: In other words, what effect the previous process step is having on the step we're now going through.

Menon: Right: When you're doing metallization with six or seven levels of metal, you want to make sure you understand the integration issues in here. One point to make here is that particles are not the only thing. In fact, particles are secondary for us. We do worry about them, but the first order is to make sure we have good integrated processes. Otherwise, after lithography, I would say broadly, the interconnect loop, including CMP, plug technologies, damascene, and all that, is an area of concern.

MICRO: What's going on in terms of advanced materials work at K-Fab, such as copper, low-k dielectrics, and the like?

Wang: There is no question that copper and low-k provide at least a couple of major advantages, such as low resistance, lower cost, fewer process steps, so we are moving toward this for high-performance applications at the 0.15-µm [technology] node. So the timeline for [production of] this type of technology is probably 2000, 2001.

MICRO: Some companies are segregating their copper work from the rest of the fab. Are you taking this approach?

Wang: That's an interesting question. Copper raises a major concern with contamination, but our ultimate goal is to have the right protocol, the right discipline so that we don't have to do special segregation. In the early [copper] part of the facility we do have a few segregated tool sets. But even today the overall fab, the rest of the layout, is not segregated because we are implementing good protocol with lots of training and everything.

MICRO: Are you color coding the cassettes and tools involved with copper?

Menon: We need tags on all of the equipment so our people know which are for copper and which are not.

Wang: The cassette color code in our factory is on all of our cassettes; it's not only limited to copper. It's a general practice. We have different levels of contamination control, different color cassettes. Anyway, copper just adds more concern to this general consideration.

MICRO: What's the proposed life span of K-Fab, as TI's cutting-edge facility?

Wang: In this industry, I can't tell you anything beyond 5 to 10 years, we change so fast, but I would say at least for the next 5 years this will be it, for all the innovation that occurs at the 0.15-µm node and 0.13-µm node. At least the next 5 years.

Photos by Michael Blackwell/Ti

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