Systematic modeling and fast simulation of mixed-signal integrated systems in deep submicron technologies

Alex Doboli,Hui Zhang
2005-01-01
Abstract:In spite of many remarkable advancements in CAD methodologies and tools, high-performance analog and mixed-signal circuit design still remains an art. A main reason is that each design reflects complex and customized trade-offs performed across successive abstraction levels. Traditionally, trade-offs are manually contemplated based on extensive design experience and knowledge. In complex design, however, manual analysis is not only tedious, but also impractical due to the many nonidealities that ought to be simultaneously considered. On the other hand, there are currently no mature analog circuit analysis and synthesis methods that can efficiently tackle all nonidealities and nonlinearities in a circuit, including layout parasitic and process parameter variations. This thesis presents a bottom-up modeling and simulation methodology and related techniques for constructing structural circuit models that incorporate many circuit nonlinearities as well as knowledge about process and model variations. For fast simulation, the thesis proposes compiled-code simulators for nonlinear analog and mixed-signal systems. The simulation method is very useful in synthesis, where tens of thousands of design points need to be analyzed in reasonably long time. Process and device model variations are modeled in three refinement steps performed bottom-up: process/model parameter variation analysis, electrical parameter variation analysis and circuit level variation behavior analysis. For process/model parameter variation analysis, an improved method is proposed. As compared to other process parameter variation modeling methods, such as σSA method, the proposed methodology is capable of correctly modeling not only differences of process/model parameters (mismatch), but also the process parameter variations for individual devices. Using the structural behavioral models, a compiled-code simulator is produced for an entire system through symbolic analysis and piecewise linear modeling. The customized simulation code is optimized for simulation speed. Our experimental results show that the proposed methodology generates more accurate circuit models, and that the customized simulators perform significantly faster than numerical simulators.
What problem does this paper attempt to address?