DFM needs new collaborative mindset
06/01/2004
In recent years, companies in the semiconductor industry have been collaborating much more than in the past. Major collaborative efforts, in various forms, have sped the development of leading-edge technologies and processes while simultaneously keeping semiconductor prices at a level compatible with end-market development.
Collaboration becomes a compelling need for different segments of manufacturing industries at different stages of that particular industry's evolution. The degree and pace of collaboration often tend to increase as an industry matures. Cost pressures mount, consolidation takes place, and collaboration often becomes necessary simply for survival. While some in our industry don't like to hear it, the global semiconductor industry is maturing, and the increased focus on collaboration is one of several signs of the maturation process at work.
Sometimes, key segments of maturing industries have a lot of trouble collaborating. These segments often tend to be the ones where collaboration is most needed because they are where a lot of value resides, and organizations do not freely share critical information. One such segment in our industry is DFM (design for manufacturing).
There is no debate as to whether DFM has the potential to deliver value to device design and manufacturing. As the industry moves to the 90nm node and beyond, fixing problems that were introduced by design becomes more difficult and a great deal more costly. DFM techniques — and perhaps more importantly, a DFM mindset — help designers avoid creating problems for manufacturing. In other, more mature manufacturing industries such as automotive, DFM has drastically reduced product development costs and time to market.
But DFM is far less advanced in semiconductor manufacturing. Inherent problems place limits on DFM. Most notably, manufacturing processes are not established until at least a couple of years after designs are created. And the rise of foundries has further solidified the natural wall that separates design and manufacturing.
Still, progress has been made. As Frank Schellenberg of Mentor Graphics said earlier this year at the SPIE conference and exposition, designers and process engineers are at least talking to each other more these days. Last month, Synopsys announced an expansion of its "lithography verification" software, which helps chipmakers deal with RET by predicting results before silicon is processed. And Hans Stork, senior VP, CTO, and director of silicon technology development for Texas Instruments, pointed out in a recent interview that TI designers now are taking into account device performance factors that change during the lifetime of the device. He used threshold voltage drift, which degrades over time, as an example. This is new thinking on the design side.
But a lot more progress is needed for DFM to even approach its potential value. Linard Karklin, VP of technical marketing at BindKey Technologies Inc., a DuPont Photomask company that supplies DFM software tools, points out that basic technological changes must be implemented in DFM tools for them to work at 90nm and beyond. Extending what has worked in the past is not a viable approach. New materials will be introduced at each new process node, and DFM tools must adapt. As processes evolve, DFM tools must follow.
"We have to rebuild our tool set," stresses Karklin. Areas of DFM that need technological advancement, he says, include the transfer of process information upstream to designers, automation of massive data gathering and data sharing, and making metrology information easily accessible to designers.
Karklin believes that joint development efforts by innovative companies, especially startups, would greatly speed DFM tools to market. Right now, these companies are duplicating each other's development work, but "it's very hard to talk to each other because we're competitors," says Karklin. He advocates the creation of "technology clusters" in which companies would merge at an early stage and combine technology and resources to move DFM technology forward. But he is not optimistic this will occur soon.
 |
The semiconductor industry has shown its resiliency many times, and it is likely to solve its problems in this area and move badly needed DFM approaches forward. But companies must be willing to share more information, and device designers must continue to develop a mindset that includes manufacturing.
Kevin Fitzgerald
Editor-in-Chief