| Performance estimation techniques that effectively evaluate design alternatives at early stages of the design process can effectively increase the quality of the final implementations. In order to fully benefit from performance estimates, they must be integrated into the synthesis environment such that estimated topologies can be directly refined into a physical layout. In this dissertation, we present a set of methodologies for estimating the performance of analog and mixed-signal systems at different levels of abstraction to assist computer-automated synthesis tools. We discuss the estimation of various Parameters of analog components, analog networks (with and without feedback), mixed-signal interface elements and high frequency electronics in detail.;We start by presenting a hierarchical knowledge-based approach to estimate analog circuit components. This methodology determines the performance parameters (UGF, slew rate, gate area, DC gain, etc.) of analog circuit components along with anticipated sizes of all the circuit elements. The specification inputs are the circuit topology and design parameters (bias current, capacitance load, etc.). The estimation is hierarchically performed using symbolic performance models of analog circuits at various levels of abstractions. These levels include basic circuit elements (MOS transistors, resistors, capacitors, etc.), simple analog circuits (current mirrors, differentiators, etc.), operational amplifiers in various configurations and analog library cells (integrators, filters, amplifiers, etc.). Also, we present a methodology for estimating the distortion effects in RF circuits produced by the nonideal behavior of circuit devices operating at high frequencies.;At the system level, we present two different techniques to estimate the performance of analog networks. The two techniques are founded on the component level performance estimation methodology. The first approach uses block diagram simplification rules (ignoring loading effects) to produce the system level performance estimates. This approach decomposes the specified system level topology into basic block diagram configurations, and recursively evaluates the performance estimation equations for the new simplified block configurations. The second methodology uses a symbolic analysis technique (considering loading effects) to evaluate the performance at the system level. In this approach, the system net-list is represented by a signal-flow graph using analog component models instead of transistor level descriptions.;Using a linear programming approach, we automate the estimation, selection and constraint transformation processes of analog-digital interface elements of mixed-signal systems. Sharing techniques are used to minimize silicon area while meeting system constraints. Also, we describe how the estimation methodologies can be effectively used during the analog sizing, constraint allocation, constraint transformation and behavioral synthesis processes. |
