August 13, 2004 -- Proverbs and design rules provide useful, and often conflicting, guidance. For example, it has often been said, "good things come in small packages." This advice is certainly followed in the design rule world, where minimum width and spacing rules are set as small as possible to reduce area.

On the other hand, people also say, "the bigger, the better." This is true in the design rule world as well, because design for manufacturability (DFM) and resolution enhancement technologies (RET) often require geometries to be larger than a strictly minimum-area focus would allow.

Much of the work in creating a good set of design rules is balancing out these kinds of valid but conflicting concerns. The final values for the rules are chosen such that the likelihood of a defect is low, rather than to prevent defects altogether.

Calculating the impact of a global design-rule change to the average area of a standard cell isn't difficult. A more useful analysis, performed on a cell-by-cell basis, can determine how close each geometry can get to its preferred, highest yield configuration. What if the vast majority of geometries can be created without penalty using the high yield rules? And unless a single rule's yield effects would be catastrophic, why globally increase the area by unnecessarily increasing a design rule?

In some cases, the penalty for not using DFM- and RET-enhancing methods is significant yield reduction. In those cases, the DFM and RET requirements become mandatory design-rules. However, the values are chosen for those design-rules such that yield degradation is an acceptable value.

By Paul de Dood. (de Dood is president and founder of Prolific, Inc.)