EDA Tech Forum

The real objective of design for manufacturability (DFM) is to improve a product’s profitability and manufacturing predictability for its market window and unit volume by optimizing tradeoffs between design costs and manufacturing improvements according to a holistic, lifetime view of the product. Current DFM practice often falls far short of that goal. For instance, the most common design mistake made during the DFM stage is the pursuit of small, incremental improvements in yield that are out of proportion with attendant design costs or schedule delays.

Examples are all too familiar. A design team might blindly follow an arbitrary deck of ‘DFM rules’ that lack the confirmation of quantified and up-to-date models and process data. Resulting errors cost millions of dollars attributable to slipped schedules, yield problems, and/or poor performance. Alternatively, a ‘recommended rule’ might demand redundant contacts in every cell and redundant vias on every route, even though these can actually decrease overall yield because they increase the chip’s critical area and lead to higher leakage surrounding every gate.