This thesis proposes a novel approach for systematic design of reconfigurable continuous-time ΔΣ modulators. The approach follows a model-based, top-down design flow. The flow consists of two main parts: a technique for building linear and nonlinear circuit macromodels, and a method for finding performance-optimized ΔΣ modulator topologies. The flow was demonstrated for the design of three new topologies for re-configurable ΔΣ modulators, and the development of structural macromodels for four state-of-the-art analog circuits.
Top-down design has recently emerged as a promising methodology for designing analog circuits and systems. It is more efficient and less costly than traditional, flat design due to its capability to conduct hierarchical optimization at the system, circuit and device level. Hierarchical optimization focuses only on the design variables of a certain level and sets constraints for the next level. For high performance applications, however, it was observed that top-down design has serious limitations due to isolating the consecutive design levels. In our experience, even for "mass production" type of applications, like a high precision ΔΣ modulator, system design had to contemplate circuit level details, like circuit noise, maximum input swing and harmonic distortion. Thus, while promising, top-down design cannot be effective unless it considers design attributes of the lower levels too. Circuit macromodels are the abstract circuit models which are capable of predicting device-level nonidealities and fast to be simulated.
The proposed model-based top-down design flow uses linear and nonlinear circuit macromodels to generate optimized topologies for reconfigurable ΔΣ modulators. The flow includes two main components: a technique for building structural macromodels for analog circuits and a method for generating optimized reconfigurable ΔΣ topologies.
ΔΣ modulators have been widely used for wireless communication applications due to their good power efficiency, superior linearity at low bandwidths, and inherent bandwidth-resolution tradeoff in the noise-shaping characteristic. This thesis proposes a systematic methodology for design of reconfigurable continuous-time ΔΣ modulator topologies with circuit-level nonidealities and constrains, which can be used for multiple operating modes of wireless communication standards. Topologies are customized for a given set of operation modes. The methodology addresses the three main aspects of reconfigurable modulator design, e.g., cost, design overhead, and power consumption, by producing topologies with (i) minimum complexity, (ii) maximum robustness regarding performance degradation due to circuit nonidealities, and (iii) minimum estimated power consumption. Moreover, the design feasibility at circuit-level is also captured in the topology design by the formulation of the constrains abstracted from the circuit macromodels.
The developing of structural linear and nonlinear macromodels for analog circuits includes two steps: first, building block behavioral model is used to describe the basic building block of a circuit, and a decoupling technique to generate uncoupled building block behavioral structural circuit model. Then, a nonlinear system is represented as a system with nonlinear inputs and linearly coupled blocks, and the linear couplings are removed. Experiments are offered for single-stage OpAmp, three-stage OpAmp, two-stage folded cascode Miller OpAmp, and operational transconductor amplifier (OTA) circuits. The produced linear and nonlinear models are scalable, tunable for the required accuracy, offer insight into the circuit operation, and require low modeling effort. (1) To the best of our knowledge, the presented methodology is the first systematic methodology for designing reconfigurable ΔΣ modulator topologies with circuit-level constrain optimized for a set of performance requirements. Produced topologies (involving both selection of signal paths and topology parameters) are customized to an application rather than extensions of traditional, single-mode architectures. As a result, produced topologies have less reconfigurable cells, and are more robust to circuit nonidealities. (2) The novel contributions of the macromodeling technique include definition of the building block behavioral models, two original algorithms to generate structural models, and a novel description of circuit nonlinearities by successive composition of three operators. The produced models are stinctural, thus each of their composing elements has a physical interpretation in terms of the original circuit. In addition, macromodels are scalable, and their accuracy is controllable. The method requires a low modeling effort, as there is little designer input or data sampling required.