DFM Becomes More Than Hype

The flurry of startup activity in the design for manufacturing (DFM) space has led some people to joke that the acronym really stands for "design for merger," because some companies structure themselves as good targets for eventual acquisition. But really, DFM has become a legitimate sector of the industry, now that real tools are helping real companies prevent zero functional yield at first silicon — the event that began the DFM craze not too many years ago.

DFM came about because of the mismatch between the precisely constructed and followed design rules and layouts and the chip yield on printed silicon. Subwavelength lithography has largely been the cause of this mismatch, but not the sole contributor. Process variability, smaller process windows, power integrity issues, and changing defect mechanisms also play a role in driving a need for a better relationship between IC design and manufacturing. When process nominal yields from test chips do not match actual yields, designers suddenly needed to concern themselves with more than power and timing issues, and branch into areas involving leakage currents, electromigration and modeling at lower supply voltages. From the manufacturing side, systematic mechanism-limited yield loss, which has become the biggest source of yield loss in advanced chips, is tied to the design. DFM is needed to address the biggest source of yield loss.

Most of the early DFM products targeted optical proximity correction (OPC) — first rule-based, then model-based corrections that accommodate the complexity of the design. Tools have been developed to start with the best focus and exposure conditions, but to then test the output throughout the process window to see if the design will print correctly. Now, the methodologies and tools are branching into new areas — no longer just concerned with such first steps as adding redundant vias, checking process windows, and performing verification on verification.

Because the industry is only now beginning to realize the effects of process and yield variation, it is also only beginning to understand the role that DFM will play in the overall yield and productivity picture. Certainly, this is a new area of expertise for which engineers with talent on both sides of the aisle are sorely needed, and will be much appreciated in the years to come.