Common concerns about migration to nanometer process technologies too often suggest that successful nanometer design demands a total break away from familiar methods. For many designers, growing emphasis on design for manufacturing (DFM) as part of emerging solutions leaves the impression that nanometer design will require a major revolution in their design approach. In fact, nanometer design is really just more of the same. Its solution is less about revolution and more about evolution toward greater accuracy, more efficient data management, and earlier anticipation of manufacturing effects.
Designers labor at achieving closure in nanometer designs with techniques that succeeded marvelously at 130 nm and up, but cannot successfully address timing problems arising from nanometer effects. To succeed at 90 nm and below, designers will find the solution lies in better-characterized models, more accurate extraction, and greater attention to a new generation of process design rules—typically bundled under the DFM rubric.
In fact, the current emphasis on DFM in EDA is simply the realization that nanometer manufacturing brings its own constraints to design. Just as designers recognize that they need to deal with timing, power, or area constraints in their circuits, designers who have worked through a nanometer design now understand that much more efficient results are possible when they account for those manufacturing constraints early in design.
Dealing with the additional critical details of nanometer design will continue to depend on design analysis and verification, which are just more types of analysis and more comprehensive verification of complete designs. Of course, those analysis and verification methods can provide results only as accurate as the models and data used to drive them.
Not surprisingly, the more successful design organizations make sure their designs are analyzed with libraries that are re-characterized for each specific process and at all critical corners. Similarly, they make sure parasitic extraction is as accurate and complete as possible, and that it reflects lithography and CMP process distortions. Why? Because they understand that for nanometer designs, relying on over-simplified parasitics is a certain path to a silicon re-spin. As mask sets top $1.5 million for advanced processes, details like model re-characterization and extraction accuracy will factor more prominently into successful design methodologies.
In fact, the types and sources of data will continue to expand in line with expanding awareness that many factors ultimately influence the performance and yield of advanced designs. For example, foundries augment design-rule sets with optional rules that are essentially signposts for improving performance and yield. Designers can ignore these optional rules and still achieve some measure of silicon success, but designs that accommodate these expanded rule sets will see correspondingly better results in production. With the expanded influence of design, lithography and manufacturing on silicon, each participant in the design chain—including IP vendor, EDA tool provider, and foundry—will play an increasingly collaborative role in better preparing designers to accommodate the diverse factors that influence results.
As a necessary enabler of these solutions, however, nanometer technology does demand some significant enhancements to current environments. Designers will gain better results by adopting new types of design analysis able to anticipate significant downstream effects, such as those associated with sub-wavelength lithography. At a more fundamental level, data management must become more efficient to deal with the explosion in data volume. New data architectures are needed to supply a consistent view of a design to the expanding list of participants in the development process, most notably manufacturing. At the same time, design flows must provide incremental capabilities that give designers a practical solution for coping with the explosion of data, particularly after tapeout and extraction.
Nanometer technologies certainly do present a challenge. However, when you look deeper at the issues, it’s apparent that nanometer design is fundamentally just more of the same—with much more data, a greater attention to accuracy, and a better anticipation of downstream effects.