Electrical Design for Manufacturability Solutions: Fast Systematic Variation Analysis and Design Enhancement Techniques

Abstract

The primary objectives in this research are to develop computer-aided design (CAD) tools for Design for Manufacturability (DFM) solutions that enable designers to conduct more rapid and more accurate systematic variation analysis, with different design enhancement techniques. Four main CAD tools are developed throughout my thesis. The first CAD tool facilitates a quantitative study of the impact of systematic variations for different circuits' electrical and geometrical behavior. This is accomplished by automatically performing an extensive analysis of different process variations (lithography and stress) and their dependency on the design context. Such a tool helps to explore and evaluate the systematic variation impact on any type of design. Secondly, solutions in the industry focus on the "design and then fix philosophy", or "fix during design philosophy", whereas the next CAD tool involves the "fix before design philosophy". Here, the standard cell library is characterized in different design contexts, different resolution enhancement techniques, and different process conditions, generating a fully DFM-aware standard cell library using a newly developed methodology that dramatically reduce the required number of silicon simulations. Several experiments are conducted on 65nm and 45nm designs, and demonstrate more robust and manufacturable designs that can be implemented by using the DFM-aware standard cell library. Thirdly, a novel electrical-aware hotspot detection solution is developed by using a device parameter-based matching technique since the state-of-the-art hotspot detection solutions are all geometrical based. This CAD tool proposes a new philosophy by detecting yield limiters, also known as hotspots, through the model parameters of the device, presented in the SPICE netlist. This novel hotspot detection methodology is tested and delivers extraordinary fast and accurate results. Finally, the existing DFM solutions, mainly address the digital designs. Process variations play an increasingly important role in the success of analog circuits. Knowledge of the parameter variances and their contribution patterns is crucial for a successful design process. This information is valuable to find solutions for many problems in design, design automation, testing, and fault tolerance. The fourth CAD solution, proposed in this thesis, introduces a variability-aware DFM solution that detects, analyze, and automatically correct hotspots for analog circuits.