Today's tools in the design for manufacturing space are essentially set up to perform post layout processing on the physical design implementation. This results in a more robust preparation of a layout for manufacturing, but does not really include much in the way of design involvement. In terms of data, the flow is backwards from manufacturing to physical implementation and doesn't get as far as the designer. The DFM mostly involves modification of the design to encompass manufacturing and yield issues.

After Mentor released Calibre nm, I asked where's the D in DFM.

To get more of design involvement in this design for manufacturing flow, the data flows must be bi-directional, manufacturing must be aware of the design intent and design must be aware of manufacturing requirements and limitations. Just like the experiences with design for test, DFM adoption will need much more education and experience for the design engineers. The large digital designs cost more to develop than earlier designs because they're harder to do, require new skills and tools, and must be finished within ever-shorter design schedules.

Designs must learn and exchange philosophies, methodologies, and practices with manufacturing. Manufacturing has always been change oriented, the engineers expect to invest time and money to develop the next generation process. Even as the process is developed and put into production, the engineers continue to learn, experiment, and modify the process throughout its manufacturing life. In comparison, design flows change over time scales measured in decades.

Now, the DFT tools are bringing information to the designer about process and maufacturability. These characteristics were secondary or tertiary effects in the larger processes. It is becoming apparent that DFM is more like the process for timing closure than DRC (design rules check), because the DFM tools must incorporate probability and a cumulative level of quality as a part of their analyses. Requirements for more analysis are driven by the greater complexity of today's designs and by the designers themselves, who are hungry for the next generation process so they can continue to pursue Moore's law.

One important change in recent times is that foundries are now releasing DFM and process data in their design kits. This change in behavior from the foundries is partially due to the need for greater differentiation as semiconductor processing becomes more of a commodity across the manufacturing sector as the fabs create partnerships to spread out the costs of process development. Increases in yield and reliability are just as important to the buyers of silicon as performance and cost in evaluating suppliers.

So far, most of the data is flowing from manufacturing to design. DFM analysis tools can guide fab inspection requirements and help to coordinate diagnostics and failure analysis of new designs. The information generated by defect and critical area analysis may not be necessary as the rules or models become more robust.

Analog and mixed signal designers are already doing a lot of DFM work as a part of their design work. The digital designers may be clueless about the techniques and analyses that are becoming necessary for design success. To be successful, digital designers will have to change their design philosophy through a combination of increased education, practice and style modifications, and experience. The top fabless companies have invested in DFM to get an advantage in the marketplace.

These changes in education, training, and experience will be similar to the transition from schematics to HDLs, since significant portions of the design flow must be altered to address the statistical characteristics of the smaller processes. Companies must start to measure themselves in meaningful ways to force changes in the design community. By relating metrics of behavior and other categories of issues to design success, companies can motivate designers to change their design approaches and flows. If the companies wait for the designers to change on their own, the companies will fall behind.

Contributors to this opinion piece were Jean-Marie Brunet, product development manager for the Litho-Friendly Design products, and David Abercrombie, Design for Manufacturing program manager, both from Mentor Graphics.

To comment on this article send email to:gmoretti@gabeoneda.com