The deployment delays for EUVL, forces IC design to continue using 193nm wavelength lithography with innovative and costly techniques in order to faithfully print sub-wavelength features and combat lithography induced process variations. The effect of the lithography gap in current and upcoming technologies is to cause severe distortions due to optical diffraction in the printed patterns and thus degrading manufacturing yield. Therefore, a paradigm shift in layout design is mandatory towards more regular litho-friendly cell designs in order to improve line pattern resolution. However, it is still unclear the amount of layout regularity that can be introduced and how to measure the benefits and weaknesses of regular layouts. This dissertation is focused on searching the degree of layout regularity necessary to combat lithography variability and outperform the layout quality of a design. The main contributions that have been addressed to accomplish this objective are: (1) the definition of several layout design guidelines to mitigate lithography variability; (2) the proposal of a parametric yield estimation model to evaluate the lithography impact on layout design; (3) the development of a global Layout Quality Metric (LQM) including a Regularity Metric (RM) to capture the degree of layout regularity of a layout implementation and; (4) the creation of different layout architectures exploiting the benefits of layout regularity to outperform line-pattern resolution, referred as Adaptive Lithography Aware Regular Cell Designs (ALARCs). The first part of this thesis provides several regular layout design guidelines derived from lithography simulations so that several important lithography related variation sources are minimized. Moreover, a design level methodology, referred as gate biasing, is proposed to overcome systematic layout dependent variations, across-field variations and the non-rectilinear gate effect (NRG) applied to regular fabrics by properly configuring the drawn transistor channel length. The second part of this dissertation proposes a lithography yield estimation model to predict the amount of lithography distortion expected in a printed layout due to lithography hotspots with a reduced set of lithography simulations. An efficient lithography hotspot framework to identify the different layout pattern configurations, simplify them to ease the pattern analysis and classify them according to the lithography degradation predicted using lithography simulations is presented. The yield model is calibrated with delay measurements of a reduced set of identical test circuits implemented in a CMOS 40nm technology and thus actual silicon data is utilized to obtain a more realistic yield estimation. The third part of this thesis presents a configurable Layout Quality Metric (LQM) that considering several layout aspects provides a global evaluation of a layout design with a single score. The LQM can be leveraged by assigning different weights to each evaluation metric or by modifying the parameters under analysis. The LQM is here configured following two different set of partial metrics. Note that the LQM provides a regularity metric (RM) in order to capture the degree of layout regularity applied in a layout design. Lastly, this thesis presents different ALARC designs for a 40nm technology using different degrees of layout regularity and different area overheads. The quality of the gridded regular templates is demonstrated by automatically creating a library containing 266 cells including combinational and sequential cells and synthesizing several ITC'99 benchmark circuits. Note that the regular cell libraries only presents a 9\% area penalty compared to the 2D standard cell designs used for comparison and thus providing area competitive designs. The layout evaluation of benchmark circuits considering the LQM shows that regular layouts can outperform other 2D standard cell designs depending on the layout implementation.
© 2001-2026 Fundación Dialnet · Todos los derechos reservados