Analysis and design of switched-capacitor DC-DC converters with discrete event models

Abstract

Switched-capacitor DC-DC converters (SCDDCs) play a critical role in low power integrated systems. The analysis and design processes of an SCDDC impact the performance and power efficiency of the whole system. Conventionally, researchers carry out the analysis and design processes by viewing SCDDCs as analogue circuits. Analogue attributes of an SCDDC, such as the charge flow current or the equivalent output impedance, have been studied in considerable detail for performance enhancement. However, in most existing work, less attention is paid to the analysis of discrete events (e.g. digital signal transitions) and the relationships between discrete events in SCDDCs. These discrete events and the relationships between discrete events also affect the performance of SCDDCs. Certain negative effects of SCDDCs such as leakage current are introduced by unhealthy discrete states. For example, MOS devices in an SCDDC could conduct undesirably under certain combinations of signals, resulting in reversion losses (a type of leakage in SCDDCs). However, existing work only use verbal reasoning and waveform descriptions when studying these discrete events, which may cause confusion and result in an informal design process consisting of intuitive design and backed up merely by validation based on natural language discussions and simulations. There is therefore a need for formalised methods to describe and analyse these discrete events which may facilitate systematic design techniques. This thesis presents a new method of analysing and designing SCDDCs using discrete event models. Discrete event models such as Petri nets and Signal Transition Graphs (STGs) are commonly used in asynchronous circuits to formally describe and analyse the relationships between discrete transitions. Modelling SCDDCs with discrete event models provides a formal way to describe the relations between discrete transitions in SCDDCs. These discrete event models can be used for analysis, verification and even design guidance for SCDDC design. The rich set of existing analysis methods and tools for discrete event models could be applied to SCDDCs, potentially improving the analysis and design flow for them. Moreover, since Petri nets and STGs are generally used to analyse and design asynchronous circuits, modelling and designing SCDDCs with STG models may additionally facilitate the incorporation of positive features of asynchronous circuits in SCDDCs (e.g. no clock skew). In this thesis, the relations between discrete events in SCDDCs are formally described with SC-STG (an extended STG targeting multi-voltage systems, to which SCDDCs belong), which avoids the potential confusion due to natural language and waveform descriptions. Then the concurrency and causality relations described in SC-STG model are extended to Petri nets, with which the presence of reversion losses can be formally determined and verified. Finally, based on the STG and Petri net models, a new design method for reversion-loss-free SCDDCs is proposed. In SCDDCs designed with the new method, reversion losses are entirely removed by introducing asynchronous controls, synthesised with the help of a software synthesis toolkit “Workcraft”. To demonstrate the analysis capabilities of the method, several cross-coupled voltage doublers (a type of SCDDC) are analysed and studied with discrete event models as examples in this thesis. To demonstrate the design capabilities of the method, a new reversion-loss-free cross-coupled voltage doubler is designed. The cross-coupled voltage doubler is widely used in low power integrated systems such as flash memories, LCD drivers and wireless energy harvesting systems. The proposed modelling method is potentially used in both research and industrial area of those applications for a formal and high-efficiency design process