Type
Text
Type
Dissertation
Advisor
Doboli, Alex | Hong, Sangjin | Stanacevic, Milutin | Salman, Emre | McConaghy, Trent.
Date
2013-12-01
Keywords
Electrical engineering
Department
Department of Electrical Engineering.
Language
en_US
Source
This work is sponsored by the Stony Brook University Graduate School in compliance with the requirements for completion of degree.
Identifier
http://hdl.handle.net/11401/77460
Publisher
The Graduate School, Stony Brook University: Stony Brook, NY.
Format
application/pdf
Abstract
Analog circuit design activities mainly depend on designers' expertise and their ability to produce new designs by combining basic devices, sub-circuits, and ideas from similar solutions as the source for innovation. There are very few systematic methods that can characterize similarities and differences between analog circuits while maintaining the correlation between design variables, performance, and trade-offs. Moreover, many computer-aided design tools are focused only on routine tasks, like transistor sizing and layout design. This dissertation presents novel techniques to automatically characterize the analog design space based on feature uniqueness and variety and to perform systematic circuit comparisons using symbolic models. Initially, we evaluate, for different analog circuit families, design science metrics aimed at capturing feature uniqueness and variety. The studies are useful in producing an overall characterization of analog circuit features. The insight obtained can help in enhancing the circuit design process and developing new automated synthesis techniques that can explore solution space regions that are likely to include novel design features. A symbolic technique is proposed to automatically create ordered feature clustering schemes that express the main structural similarities and differences among analog circuits. Four separation scores, based on entropy, item characteristics, category characteristics, and Bayesian classifiers are investigated for large sets of state-of-the-art amplifier circuits. The generated representations offer understanding about the uniqueness and importance of specific design features and can be used in topology refinement and automated synthesis. For detailed analysis, an automated mechanism for systematically producing comparison data between two analog circuits is developed. The similar and distinguishing performance characteristics of circuits with respect to gain, bandwidth, common-mode gain, noise, and sensitivity are captured. The technique utilizes matching of both topologies and symbolic expressions of the compared circuits to find the nodes with similar behavior. The impact on performance of the unmatched nodes is used to express the differentiating characteristics of the circuits. The produced comparison data is important for getting insight into unique benefits and limitations of a circuit, selecting fitting circuit topologies for system design, and optimizing circuit topologies. Systematic comparison is the basic operator of a prototype framework for modeling the analog circuit design feature variety. The proposed concept structure model expresses symbolically the design features as well as their advantages and limitations at different levels of abstraction and includes systematic mechanisms that can create new conceptual solutions. Case study examples illustrate application of the proposed methods in a reasoning-based analog circuit synthesis technique. The procedures incorporate cause-effect understanding of the performance limitations generated by circuit structures and precisely addresses them by finding alternatives that relax performance trade-offs. | 270 pages
Recommended Citation
Ferent, Cristian, "Systematic Modeling and Characterization of Analog Circuits using Symbolic and Data Mining Techniques" (2013). Stony Brook Theses and Dissertations Collection, 2006-2020 (closed to submissions). 3272.
https://commons.library.stonybrook.edu/stony-brook-theses-and-dissertations-collection/3272