New methods and software tools for integrated circuit design for quality and manufacturability

This dissertation addresses some of the problems encountered in Statistical Design of Analog Integrated Circuit (IC) blocks. The objective is to develop new methodologies and software tools for more efficient and practical circuit optimization using a combination of the existing and newly developed IC optimization software, in the specific area of IC design for Quality and Manufacturability.; Design for Quality and Manufacturability (DFQM) is normally formulated as an optimization problem, where the goal is to minimize the variabilities of the circuit performances and to tune the statistical mean values of these performances to their desired (target) nominal values. New methodologies are developed in this dissertation to estimate these performances statistics, and apply specific performance measures.; The major contributions of this dissertation in this area of ongoing research are: (1) A new methodology based on the Singular Values Decomposition of the scaled system Jacobian matrix of the circuit performances with respect to the noise parameters to estimate a measure of overall performance variability suitable for optimization, using a reduced space characterization of performance variabilities. (2) A new methodology to create performance linear models dynamically using data from the previous simulations and only a small number of additional circuit simulations to estimate the performance statistics. (3) A new methodology to identify the trade-offs between different circuit performances, and to apply visualization facilities for making design decisions. (4) Development of the above and other related software into a design framework for IC optimization. These contributions are supported by suitable IC optimization and design examples.; The new DFQM methodologies presented in this dissertation were implemented as the following CAD tools: (1) svdfm: a program supporting first two methodologies; (2) simvis: a utility program for visualization; and (3) other associated software tools providing suitable interfaces of an optimizer (the GOSSIP system, developed at Electrical Engineering Department, Texas A&M University, in this case) with different circuit simulators.; Examples are given for each of the methodologies proposed and the tools developed. Practical application and comparison of the new tools and methodologies is demonstrated through the design and optimization of an industrial circuit.