Automated system identification and controller tuning for digitally controlled dc–dc converters

Abstract

Digital controllers are often used in interleaved multi-phase dc-dc converters. System identification is desirable for precise frequency-domain characterisations and subsequent stability assessment and/or controller design. The digital nature of the feedback signal facilitates the communication between the converter and a processing unit, which is used here to estimate the system parameters by means of system identification in situ . The phase loss, zero-order-hold effect and asynchronised sampling of the digital controller are discussed. Design limitations and desired dynamic behaviours complicate the controller design process and necessitate automated tuning of the controller. Therefore a computer-generated auto-tuning framework is set forth that provides the required controller coefficients to meet desired time and frequency performance criteria. This is an offline optimisation-based process that minimises a cost function formulated using appropriate error terms. The identification results for single- and multi-phase dc-dc converters closely match those obtained from conventional network analysers. Also, the auto-tuning performance is verified in time and frequency domains using hardware measurements and numerical simulations.