DFM Progress Will Require a Leap of Faith

If there's one thing this industry is good at, it's fixing problems. After all, that's what engineers do. And sometimes, it's just a matter of identifying the problem itself that leads the way to rapid problem solving. This is exactly what's happening in the design-for-manufacturing (DFM) space today, and it couldn't be more exciting or, some would say, more challenging, than feats the industry has undertaken in the past. The difference is an air of urgency, intent and commitment that is truly remarkable. And the drivers are the ever-present motivators — yields and profits.

The pressures of time-to-market and time-to-yield were greater than ever, and when 130 and 90 nm designs were not yielding as expected, big red warning lights were set off, indicating something was drastically wrong. When fabless companies brought their designs to a wafer foundry and the foundry said they could manufacture the device given their design rules, and later when the device was manufactured with zero functional yield, it was unclear who was to blame. Was it the fault of the design (as the foundry contended) or the fault of manufacturing (as the fabless company believed)? So began the pressing need for design-for-manufacturing.

IDMs allegedly didn't run into these problems because of the close link between the manufacturing process windows and the design tolerances. Or at least that is how many people perceive the DFM picture. However, in actuality, all manufacturers and designers must address a broad range of DFM issues affecting many aspects of manufacturing, from initial mask design to burn-in efficiency. Those who believe it is strictly a lithography problem are sorely underestimating the reach of DFM (see "DFM: Worlds Collide, Then Cooperate," p. 40).

But regardless of the situation at the IDMs, the foundries and fabless companies found themselves in a most difficult situation. After all, the fabless company's most closely guarded information is its intellectual property — its design. Some of the foundry's most guarded information lies with its lithography process. The DFM bridge that will deliver high-yielding wafers requires sharing of this extremely valuable information.

A few companies are stepping up as suitable mediators for the DFM challenges that lie between the foundry and fabless companies. However, will either party trust a third party with their most valuable information? It seems they may not have a choice, because there is expertise required in so many different areas and all the pieces of the puzzle must fit together. But such changes and relationships will be forged slowly and must be based on a great deal of trust. We are talking about a level of partnership that is perhaps unprecedented in the industry.

Another challenge is finding engineers with the right mix of skills. People say it's hard to find a good analog design engineer. But try to find a litho engineer with some design expertise, or someone on the design side who has knowledge of rework. The level of specialization in engineering is such that finding people with any combination of design and manufacturing expertise is difficult. These worlds have literally collided, and restructuring has been required on both sides of the fence. Neither side is necessarily happy about the added responsibilities associated with making lithography more design-friendly and designs more manufacturable. But this is the reality of the day, and the only path that will lead to acceptable yields and profits.

Many tools for DFM exist, and many startup companies are addressing everything from lithography simulation to resistance distribution modeling. The harder part of DFM will be making the organization changes necessary to communicate the information needed in a timely manner and modifying the fabless/foundry model to include DFM solutions without having either side give up its IP. Mediators may be needed on the outside, but they will definitely be needed within organizations to understand DFM's ongoing complexities and address new DFM issues as they arise.