On-Chip Single-Inductor Multiple-Output Power Converter Design with Adaptive Cross Regulation and Supply Variation Control for Power-Efficient VLSI Systems

Autor: Dongsheng Ma, Rajdeep Bondade, Yi Zhang
Rok vydání: 2011
Předmět:
Zdroj: Journal of Low Power Electronics. 7:71-86
ISSN: 1546-1998
DOI: 10.1166/jolpe.2011.1118
Popis: Dynamic voltage and frequency scaling (DVFS) algorithms are widely used power management techniques which enable the optimization of system-level power dissipation, energy consumption and operation performance. One cost-effective hardware-enabling platform to implement the DVFS is through single-inductor multiple-output (SIMO) DC-DC converters. However, the major drawback with such converters is cross regulation, which leads to undesired supply voltage variations. These variations have significant impact on modern VLSI applications, from the circuit to the system level. Based on these challenges, this paper presents a cross-layered design of two SIMO DC-DC converters with an adaptive freewheel switching technique. The proposed converters are able to minimize cross regulation by adaptively adjusting the freewheel switching durations and currents. This also increases the efficiency of the SIMO converters, by reducing the conduction losses due to the lossy freewheel switch. To verify the cross regulation performance of the SIMO converters, its influence on clock frequencies generated by a ring oscillator is investigated. By simulating various load transient conditions, it is shown that the SIMO converters are able to significantly minimize frequency variations. The operation of the converters is verified through transistor-based HSPICE simulation results, with a 130-nm CMOS process. The SIMO converter with the digital-based adaptive freewheel switching controller achieves a maximum efficiency of 93.2%, while the analog counterpart achieves a maximum efficiency of 93.4%. For a load change from 200 mA to 40 mA, the proposed technique reduces the conduction losses during the freewheel switching duration by 99.96%, compared with traditional methods.
Databáze: OpenAIRE