On 19 January, Texas Instruments Inc., in Dallas, named Johannes ”Hans” Stork as its new chief technology officer. Born in the Netherlands and educated at Delft University of Technology and then at Stanford University in Palo Alto, Calif., Stork was made a Fellow of the IEEE in 1994 for his work on silicon germanium heterojunction bipolar technology. Before Texas Instruments tapped him, he managed areas of research at IBM Corp. and Hewlett-Packard Co., Palo Alto, Calif. Stork told IEEE Spectrum senior technical editor Linda Geppert his plans for the company in a 23 January interview.
Q: What is the focus of R&D at Texas Instruments today?
The company spends about 80 percent of its total R&D budget on product development and 20 percent on process development. Last year that budget was projected to be US $1.7 billion.
On the processing side, we are moving 90-nanometer CMOS [a semiconductor manufacturing technology capable of producing chip features with dimensions of 90 nm] into volume production, and developing the technologies needed for 65 and 45 nm. Of course, this is a gradual process. We have done work on all these technologies already, but the emphasis shifts from one generation to the next over time.
In connection with that, in June 2003 we announced the building of a new fabrication facility in the Dallas area that we plan to have completed sometime in 2005.
We expect it will start operation using 65-nm technology for production, along with the development of the materials and tools needed for 45 nm. And that raises one of the big issues in the industry: the economics of doing process development and process manufacturing in the same semiconductor fab.
Manufacturing has economies of scale. So as long as the company grows, we can make the kind of investments required to produce the wafers. But on a development level, we don't have those synergies. So I encourage my team to think a lot about doing development in a systematic way--to be clever and create new methods to improve the development process.
Will you be overseeing circuit development as well?
The primary focus of the CTO [chief technology officer] role is for process development. I am also expected to raise the broad technology issues within the company. So, yes, through the interaction with the various business groups, I am expected to provide leadership consistent with the interests of those business units.
For a company like TI, where we have so much product design as well as internal process development, the coordination of the process and the product is critical. It's critical in the sense that we can do better as a company than we would if these two things were decoupled. So, for example, fine-tuning our system-on-chip capability is--I don't want to say only possible--but certainly enabled by having both the circuit and process skills inside the company.
Looking ahead, what are the most important areas for R&D at TI?
A close second to developing advanced digital semiconductor processes should be good optimization of our mixed-signal and analog process. TI competes in the signal-processing world and signal processing has a digital part, which is enabled by the continual scaling of CMOS, and also a mixed-signal and analog part. It's not as big an investment, but it is clearly an important element.
What specific things do you look at when optimizing the mixed-signal and analog processes?
What is happening there is just like what is happening in advanced CMOS. More and more, advanced CMOS includes some analog and RF [radio frequency] elements. And, vice versa, a lot of the analog world now includes digital logic and memory. What separates the two worlds is the difference in voltage. Advanced CMOS is operating internally at close to 1 volt, and externally at 3 V. In the mixed-signal world, you are talking about 20 V and up. It's a different set of constraints.
Wireless communications is an area where we see value in single-chip implementations-- systems-on-chip--that include both digital and analog components. We believe a good system-on-chip technology is a pretty competitive weapon. And the same idea applies to broadband communications. There's also a high level of digital-analog integration on the base station side.
Across many of the markets where we compete, it's important to carefully determine where a system-on-chip makes the most sense and where it makes no sense.
And then on the analog side, we have recently challenged the industry by proclaiming that we can build high-speed A/D [analog to digital] converters at least as well as, or better than, anyone in the world. That's a market that never stands still and it's also a market that values technology very highly.
One area that's a little bit outside of the digital-analog IC space is digital light processing (DLP). We build micro-mirror arrays--hundreds of thousands of tiny mirrors on top of an IC whose angles are controlled electrostatically by circuits on the silicon surface. Light is reflected off these mirrors onto a screen. Today, projectors are the main application, but they are growing quite nicely in digital television as well. Digital TV also requires a good deal of signal processing, an area that TI excels in. So we need to coordinate the right portfolio of products and technologies for those markets.
Digital large-screen TV has been talked about for a long time. What's different now?
The price point is coming down rapidly: 50-inch screens will soon be about $2000 and the price will continue to drop.
DLP has extremely good contrast and light efficiency because the light reflects directly off the mirrors. So it can make possible slim-line televisions with very large dimensions. This is an area that is difficult for LCD [liquid-crystal display] panels to get into, and it probably will be so for a long time. Yet here in North America, the growth rate for large-screen TVs is very rapid. And DLPs have a good market share in that. In fact, we have been gaining market share.
What are your thoughts on LCOS[liquid crystal on silicon], which Intel Corp. has been promoting for displays?
It is an alternative implementation of the large-screen TV concept. But we are not developing any technology along that line. LCOS has not yet been able to demonstrate a level of maturity to be proven manufacturable. Many reputable manufacturers have tried to bring LCOS to market--Samsung, Thomson, and Toshiba have all abandoned their LCOS TVs, and both Samsung and Thomson soon after began utilizing DLP technology. DLP has already established a pretty sizable market share. We have products being built, we have the market infrastructure, and we have the confidence of the large consumer electronics companies that sell these things.
Is anybody else doing digital light projection with micromirror arrays?
We are the only ones. It's a unique position, and it's a technology that has taken a long time to mature. But clearly I can say without reservation that we can produce these arrays in volume with good yield. We make them on 200-mm wafers so we get good efficiency, even for very large chips.
How is R&D at TI affected by globalization?
With our global economy, you have to be world-class in what you are doing. No ifs, ands, or buts. People can get a component or a design or a piece of software anyplace and anytime. So, if you can't produce it on a competitive level, worldwide, then customers will find better ways to collect all the various elements they need for their products.
Another aspect of globalization is that information is available at record speed. It's almost exactly 10 years ago I left IBM. And these days, compared to a decade ago, the amount of time that goes by before knowledge spreads is counted in hours, certainly not in days and weeks.
We now have a 24-hour culture, and that's not just manufacturing. Development is also a 24-hour activity. People in another part of the world can do something overnight and tell you in the morning what they did. You can coordinate [the work of] different design teams around the world so that you get two or three times the productivity. And all the teams can use the same tool sets and communicate with each other across the globe. So global competitiveness, timeliness, and productivity are very big consequences of globalization.
Any differences in R&D cultures between IBM, H-P, and TI?
IBM and H-P are systems companies, whereas TI is essentially a component supplier. Now components have become systems in their own right. But there are more pieces needed to make an end product.
For TI, even though we have a large portfolio, the product space that we occupy is semiconductors--not as broad as the diversity of products and services that come out of a company like H-P. On the flip side, we cover everything from building the materials to developing the software for our products. So even though it's on a chip, it has the complexity of a whole system.
What advice would you give to young engineers who would like to become CTOs of large companies?
The first step is to be world-class in what you do. It doesn't really matter what you pick to specialize in. Be world-class in what you are doing and then, only after that, develop some breadth of interest.
It's also advantageous to understand the disciplines that are related to what you're involved in, and to consider the supply side and the customer side. And then, if you're really talking about the CTO perspective, it's important to have some sense of how a business runs. During my Ph.D. days, I took a variety of classes on industrial engineering and on how to run a company. And for the first 10 years of my career, I don't ever remember looking at those books.
But in the last few years, I have actually used them. So business acumen certainly complements technology skills. But I would say it's only a complement, not a replacement. There are a lot of good business people out there and I would never try to measure myself against them. It's just that I can speak their language and understand how they look at the world.
A general skill, and not just for engineers and higher-level professionals, is to understand how other people look at your work. It is vital in figuring out why your work is not successful and how it can be made successful.
Jacob Ziv (with glasses), who became chair of Technion's electrical engineering department in the 1970s, worked earlier on information theory with Moshe Zakai. The two collaborated on a paper describing what became known as the Ziv-Zakai bound.Photo: Jacob Ziv/Technion
Jacob Ziv (left) and Abraham Lempel published algorithms for lossless data compression in 1977 and 1978, both in the IEEE Transactions on Information Theory. The methods became known as LZ77 and LZ78 and are still in use today.Photo: Jacob Ziv/Technion