Roundtrip design strategy of high-speed delta-sigma A/D converters

With the advance of process technologies, analog circuits are integrated with digital circuits to form system-on-a-chip (SoC). Compared with digital circuits, the sizes of analog circuits are small. However, their development time usually dominates time-to-market of products, since analog circuits are sensitive to noises and circuit nonidealities. Redesign is necessary when process technologies migrate, but design reuse is not easy for analog circuits. Therefore, very limited synthesis exists in analog circuit design and makes development time longer than that in digital circuits.; In order to accelerate analog circuit design, design strategies are proposed to reduce redundant efforts and improve designers' productivity. The strategies also offer some required design procedures that are important in circuits. Traditional top-down design strategy is not sufficient for analog circuits. Analog behavioral modeling and some backtrackings are required to ensure desired performance in different design stages. Therefore, roundtrip design strategy (top-down design and bottom-up verification) with enhanced behavioral models is proposed to minimize the unnecessary tasks in designing analog circuits. There are three accomplishments in this dissertation: (1) A novel behavioral model for Delta-Sigma modulators is invented to combine time domain analysis with noise effects. With updated simulation results (transient analysis and AC analysis) from circuit simulators, virtual tests from behavioral simulations can help designers find potential problems in early design stages. Design margins for parameter variation from statistical information are also maintained in this model. (2) A high-resolution (14-bit) high-speed (2-Msamples/s) Delta-Sigma modulator is designed, modeled and fabricated in AMI 0.5um CMOS technology to demonstrate the roundtrip design flow. (3) This behavioral model is implemented in different popular computer tools/languages: MATLAB, VHDL-AMS, Java and C/C++. By following brief comparisons at the end of this dissertation, designers can choose suitable tool/language to build their own behavioral models based on their needs.