| The need for analog and mixed-signal design is predicted to dramatically increase over the next years. Compared to digital designs, which can be efficiently designed with low effort using modern high-level, logic level, and physical-level design automation tools, analog design continues to seize a considerable portion of the total design time. Up to now, most of the analog designs have been done by analog designers, who mostly rely on their knowledge and experience. It is extremely difficult for a novice designer to do high-performance analog and mixed-signal design. Analog design automation is also motivated by the design productivity gap considering the trend of integrating hundreds of millions of transistors on a chip, which is infeasible to rely on full-custom designed blocks. To improve the productivity and time-to-market, design automation tools are highly desirable in this area.; Similar to digital design, analog design can also be done in a hierarchical manner. For example, for an analog continuous-time filter, high-level design includes selection of the filter topology and optimization of block specifications, such as gain and bandwidth in the case of an operational amplifier. Then, circuit-level design tools can be used to further synthesize the blocks, which are represented at the transistor-level. Finally, transistor layout is performed. The performance of an analog system depends heavily on the system topology, block specifications and transistor sizes in different levels of abstraction. It is possible that fixed system topology, fixed block specifications and transistor sizes do not leave enough performance margin to accommodate various performance-degrading factors, such as circuit nonidealities and layout-induced parasitics. In this case, the final design might be incorrect and costly re-iterations are needed to produce a new design.; To tackle this issue, the thesis presents a new top-down design methodology for synthesis and optimization of analog systems. Given the design specifications, the methodology starts with template-based optimal topology exploration for the analog system. Delta-Sigma modulators has been used as a case study in Chapter 2. We defined a generic modulator architecture that incorporates all possible feedback and feedforward signal paths, and we derived the symbolic Noise Transfer Function (NTF) and Signal Transfer Function (STF) for the generic topology. We then used the symbolic functions to formulate the topology exploration problem as a MINLP (Mixed-Integer Nonlinearly Constrained Programming) problem that simultaneously generates and selects the optimal modulator topology with respect to the cost function. (Abstract shortened by UMI.)... |
