New methodologies and their implementations for integrated circuit design for quality and manufacturability /
This dissertation discusses the developments and implementations of methodologies to calculate Propagation of Variance (POV) in the area of Design for Quality (DFQ) of integrated circuit design. First, using the concept of a-cut representation of fuzzy numbers a methodology is developed to handle un...
| Main Author: | |
|---|---|
| Format: | Thesis Book |
| Language: | English |
| Published: |
[Place of publication not identified] :
[publisher not identified] ;
1999.
|
| Subjects: | |
| Online Access: | http://proxy.library.tamu.edu/login?url=http://proquest.umi.com/pqdweb?did=731681661&sid=1&Fmt=2&clientId=2945&RQT=309&VName=PQD |
| Summary: | This dissertation discusses the developments and implementations of methodologies to calculate Propagation of Variance (POV) in the area of Design for Quality (DFQ) of integrated circuit design. First, using the concept of a-cut representation of fuzzy numbers a methodology is developed to handle uncertain statistical information and its propagation. Secondly, utilizing quadratic approximation to the performance functions, an approximation is used to predict integrated circuit (IC) performances. Knowing the statistics of IC performance indices, finding manufacturing yield, maximizing yield and reducing performance variability are important in statistical design. But these processes are also difficult. Very often, the available statistical information is not precise and only vague estimates of statistical parameters are possible, for instance standard deviation, because the manufacturing process and an IC are developed simultaneously. The method of the Propagation of Fuzzy Variance was used to express uncertain statistical information and how to propagate it through a circuit. Also, being limited by linear approximation, the standard POV method does not provide sufficiently accurate predictions of the performance of circuits. Building quadratic models for performance functions for a circuit, the approximation was used to increase the accuracy of the POV for DFQ applications (variability minimization and tuning). Some examples are chosen to illustrate the applications of these methods. These methods are important in practice, since the first method makes it possible to investigate the influence of the vague statistical information (often encountered in practice) on the overall IC design process while the second method is cheaper than the Monte Carlo analysis and more accurate than linear POV commonly used in statistical circuit design. This dissertation also focuses on device mismatch in statistical circuit design. In order to obtain the highest yield for integrated circuits, influence of fabrication process must be considered in initial design. The device mismatch in circuits, both digital and analog, is one of the most important problems resulting from processing. This dissertation discusses building mismatch statistical models and incorporating the effect of mismatch to design flow for a set of transistors. The proposed methodology is used to study the Power Supply Rejection Ratios (PSRR) in circuits. Theoretically developing new design methodologies is the big progress in the field of statistical circuit design, but applying these methodologies to industries and getting admitted by circuit designers are important as well. This dissertation emphasizes how to solve real problems that exist in the current situation of industries. Examples are given for each of the methodologies proposed and the tools developed from these methodologies. |
|---|---|
| Item Description: | Vita. "Major Subject: Electrical Engineering". |
| Physical Description: | xi, 107 leaves : illustrations ; 28 cm. Issued also on microfiche from University Microfilm Inc. |
| Bibliography: | Includes bibliographical references (leaves 66-72). |