Standard cell optimization and physical design in advanced technology nodes
MetadataShow full item record
Integrated circuits (ICs) are at the heart of modern electronics, which rely heavily on the state-of-the-art semiconductor manufacturing technology. The key to pushing forward semiconductor technology is IC feature-size miniaturization. However, this brings ever-increasing design complexities and manufacturing challenges to the $340 billion semiconductor industry. The manufacturing of two-dimensional layout on high-density metal layers depends on complex design-for-manufacturing techniques and sophisticated empirical optimizations, which introduces huge amounts of turnaround time and yield loss in advanced technology nodes. Our study reveals that unidirectional layout design can significantly reduce the manufacturing complexities and improve the yield, which is becoming increasingly adopted in semiconductor industry [61, 89]. The lithography printing of unidirectional layout can be tightly controlled using advanced patterning techniques, such as self-aligned double and quadruple patterning. Despite the manufacturing benefits, unidirectional layout leads to more restrictive solution space and brings significant impacts on the IC design automation ow for routing closure. Notably, unidirectional routing limits the standard cell pin accessibility, which further exacerbates the resource competitions during routing. Moreover, for post-routing optimization, traditional redundant-via insertion has become obsolete under unidirectional routing style, which makes the yield enhancement task extremely challenging. Regardless of complex multiple patterning and design-for-manufacturing approaches, mask optimization through resolution enhancement techniques remains as the key strategy to improve the yield of the semiconductor manufacturing processes. Among them, Sub-Resolution Assist Feature (SRAF) generation is a very important method to improve lithographic process windows. Model-based SRAF generation has been widely used to achieve high accuracy but it is time-consuming and hard to obtain consistent SRAFs. This dissertation proposes novel CAD algorithms and methodologies for standard cell optimization and physical design in advanced technology nodes, which ultimately reduces the design cycle and manufacturing cost of IC design. First, a standard cell pin access optimization engine is proposed to evaluate the pin accessibility of a given standard cell library. We further propose novel pin access planning techniques and concurrent pin access optimizations to efficiently resolve the routing resource competitions, which generates much better routing solutions than state-of-the-art, manufacturing-friendly routers. To systematically improve the manufacturing yield in the post-routing stage, a global optimization engine has been introduced for redundant local-loop insertion considering advanced manufacturing constraints. Finally, we propose the first machine learning-based framework for fast yet consistent SRAF generation with the high quality of results.