Showing papers by "Dragan Maksimovic published in 2021"

Reduction of AC Winding Losses Due to Fringing-Field Effects in High-Frequency Inductors With Orthogonal Air Gaps

Abstract: In high-frequency inductors, ac winding losses are affected by skin and proximity effects, including uneven current distribution due to fringing magnetic fields around air gaps. It is well known that fringing effects can be mitigated using distributed air gaps. This article is focused on an orthogonal air-gap approach, which is a distributed air-gap technique where gaps are placed in core segments parallel with the windings and in core segments perpendicular to the windings. The orthogonal air-gap approach is developed using a one-dimensional analytical framework and validated by two-dimensional and three-dimensional finite-element simulations. Analytical guidelines are presented to optimize the air-gap distribution to achieve minimum ac resistance. As a case study, a planar inductor is designed for an 8-kW SiC-based buck converter operating at 250 kHz. It is shown how the orthogonal air gaps result in approximately 50% reduction in ac resistance and substantially reduced inductor losses compared to the design using standard air gaps. Experimental validation includes measurements of losses on the converter prototype as well as quality factor measurements in a resonant-circuit test setup.

A Two-Stage Automotive LED Driver With Multiple Outputs

Abstract: This article presents a two-stage automotive LED driver architecture delivering independently regulated output currents to multiple LED strings. The system consists of a multiphase noninverting buck–boost front-end stage, which allows for a wide battery voltage range, followed by high-frequency immittance-network-based LCL-T resonant converters, which operate as current sources over wide output voltage ranges. The two-stage buck–boost + resonant (BB+resonant) architecture takes advantage of the buck and boost capability of both stages, and the flexibility in setting the intermediate bus voltage to minimize losses. Advantages of the BB+resonant architecture include the use of lower voltage rated devices and soft switching in the resonant stage, leading to reduced losses and size. Experimental results are provided for a prototype consisting of a 250-kHz two-phase front-end stage and 2-MHz LCL-T resonant stages delivering independently regulated 1 A currents to four LED strings with $N = 1$ $-$ 18 LEDs. The measured system efficiency is greater than 88% over wide input (8 $-$ 18 V) and output (3–50 V) voltage ranges, with a peak efficiency of 93%.

Transformerless Stacked Active Bridge Converters: Analysis, Properties, and Synthesis

Abstract: This article presents analysis, properties, and systematic synthesis of a new class of hybrid dc–dc converters named transformerless stacked active bridge (TSAB) converters. The TSAB converters, which are obtained from switched-capacitor (SC) converters by insertion of small ac inductors, inherit advantages of parent SC converters together with conversion and control characteristics similar to transformer-isolated dual-active-bridge (DAB) dc–dc converters. Features of TSAB converters include soft charging and discharging of capacitors, zero voltage switching, low peak and rms current stresses, low energy storage requirement of magnetic components, and regulation capabilities using simple phase-shift control. Based on a network-theoretic approach, an algebraic representation of TSAB converters yields general results for dc characteristics and component stresses. Furthermore, a systematic synthesis approach is formulated allowing construction of TSAB converter topologies starting from various two-phase SC converters. Synthesis results are presented for TSAB converters with two inductors obtained from well-known SC parent topologies, including Dickson, ladder, stacked-ladder, doubler, and Fibonacci. Experimental results are summarized for 48-to-12 V doubler, 48-to-12 V Dickson, and 36-to-12 V ladder TSAB prototypes. It is shown how the TSAB prototypes use small ac inductors (tens to hundreds of nanohenries) while operating at relatively low switching frequencies (150–200 kHz) and have measured efficiency above 98% over wide load range.

High-Frequency Wide-Range Resonant Converter Operating as an Automotive LED Driver

Abstract: This article introduces an immittance network-based wide-range LCL-T resonant dc–dc converter, where two control variables, a phase shift between two inverter half-bridges, and a phase shift between inverter and rectifier half-bridges, are utilized to achieve output current regulation while minimizing losses over wide ranges of input and output voltages. A practical control strategy is developed to adjust the phase shifts so that zero voltage switching of all devices is achieved while minimizing conduction losses. Efficiency improvements over a standard LCL-T converter are demonstrated by loss modeling and by experiments. The features of the proposed WR-LCL-T converter are well suited for automotive LED drivers. Experimental results demonstrating operation with the loss minimizing control strategy are shown for a 2-MHz converter prototype operating from an input voltage ranging from 8 to 18 V, and delivering 0.5-A output current to a string of 1–14 LEDs, which corresponds to an output voltage range of 3–45 V. Using silicon MOSFETs, the prototype achieves a peak power stage efficiency of 92.4% and maintains greater than 86% power stage efficiency across the wide input and output voltage ranges.

Efficiency-Optimized Current-Source Resonant Converter for USB-C Power Delivery

Abstract: This paper presents a resonant dc-dc converter capable of providing load-independent output current without the need for active current control. This converter comprises an LC3L resonant tank, with inductance and capacitance values selected based on a new design methodology that achieves two objectives: 1) the converter’s output current is independent of its output voltage, and 2) the resonant tank losses are minimized. The design methodology is based on minimizing inductive energy storage to improve the power conversion efficiency. This design methodology makes the converter suitable for applications that require constant current over a wide range of output voltages, such as USB-C power delivery (PD) based battery charging. An efficiency-optimized current-source LC3L converter prototype operating from 20-V input and delivering 6 A current with USB-C PD output voltages in the range from 5 V to 15 V is designed, built and tested. The prototype achieves a peak efficiency of 92%, maintains >87% efficiency over the entire output voltage range, and is also shown to have >10% lower losses compared to a conventionally designed prototype. A loss breakdown demonstrates how reduced inductive energy storage results in lower system losses.

Voltage Sharing With Series Output Connected Battery Modules in a Plug-and-Play DC Microgrid

Abstract: This article presents output voltage sharing among series output connected battery power modules (BPMs) in plug-and-play (PnP) dc microgrids with a wide bus voltage range. The system provides active cell balancing and interfaces cells to the bus voltage through series output connected BPMs. A cell current sensor is utilized to achieve both state-of-charge regulation and output voltage sharing. To improve microgrid modularity, the battery pack behaves as a current source. A current-dependent droop between the individual output voltages and input currents is used to facilitate series output connected voltage sharing with bidirectional power flow, simple control, and smooth transitions between charge and discharge regions. Experimental results are reported for a 3 kWh 1C-rate prototype consisting of two parallel 24 V subsystems, with five series output connected BPMs in each subsystem. Each BPM consists of three battery cells and three 100 W dc–dc converters.

Grid-connected Self-synchronizing Cascaded H-Bridge Inverters with Autonomous Power Sharing

Abstract: Cascaded inverters are widely applied in applications where elevated ac voltages are required while using semiconductor devices with lower voltage ratings. Here, we focus on structures that require localized power transfer between low-voltage sources/loads dispersed across inverter dc links and the inverter ac-sides are series-connected across a three-phase medium-voltage ac grid. To date, decentralized controllers that allow for bi-directional power transfer in such systems are limited. To fill this gap, we propose a virtual oscillator controller which modulates the power processed by each inverter in a purely decentralized manner. The proposed controller uses only locally measured current, provides communication-free synchronization of the inverter modules, and enables control of power transfer in both directions. Moreover, it is implemented purely in time domain as opposed to phasor-domain based droop controllers. Stability is analyzed and a design procedure for the oscillator parameters is provided alongside relevant simulation and experimental results for a system of five series-connected inverters.

An Arbitrary Waveform Generator based on an Eight-Level Flying-Capacitor Multilevel Converter

Abstract: Arbitrary waveform generators (AWG) are essential in various industrial and medical applications. These are typically large-signal amplitude and high frequency waveform generators, which require switching power converters in order to meet stringent transient performance requirements with high bandwidth and high efficiency. This paper presents an AWG design using an eight-level flying-capacitor multilevel (FCML) converter. The power stage design and a voltage controller are designed to achieve fast responses reference waveform transients, and an active control scheme based on duty cycle compensation is used to ensure that the flying capacitor voltages remain balanced under steady-state and transient conditions. The approach is verified on a 500 W eight-level digitally controlled GaN-based FCML prototype operating at a switching frequency of 400 kHz from 200 V input voltage, with output voltage varying from 25 V to 180 V, and with maximum output current of 2 A. Experimental results show that the prototype is capable of tracking various reference waveshapes including step, ramp, and sinusoidal, while at the same time regulating the flying capacitor voltages. For step transients between 25 V and 180 V the rise and the fall times are shorter than 3 µs.

Thermal Design, Optimization, and Packaging of Planar Magnetic Components

Abstract: This article is focused on the thermal design and three-dimensional (3-D) package optimization of planar magnetic components (PMCs), including transformers and inductors for application in an electric vehicle composite boost dc–dc converter. Each PMC comprises electrical windings in printed circuit board (PCB) form in combination with a ferrite core. Multiple features of each PMC package are thermally optimized for the proposed device configurations with given core size, core loss distribution, number of turns in the PCB winding, winding copper thickness, and winding loss distribution. These heuristically optimized features include a lower level cold plate structure with a conformal base for enhanced convective heat transfer, an upper level PMC cap structure for doubled-sided cooling through conductive heat flow to the cold plate, the implementation of functionally distributed copper thermal and electrothermal vias in the PCB winding for improved cross-plane thermal conductance, and judicious implementation of select materials at various locations and interfaces within the package. Detailed numerical modeling reveals the combined effect of this 3-D packaging strategy with a 79.3 °C and 48.5 °C maximum temperature reduction in the core and PCB winding, respectively, relative to a baseline device configuration. Select PMC experimental validation confirms the expected thermal performance of an optimized PCB design. The thermal design approach is relevant for a range of high-power-density electronics PMC packaging applications.

Systematic Optimization of Multiple-Output DC Distribution Architectures

Abstract: Complex electronic systems often require a power distribution architecture that provides multiple, separate voltage domains for various subsystem loads, such as microprocessor cores, interface, memory, analog, and radio frequency components. The multiple point-of-load regulated voltages are typically generated using multiple dc–dc converters operating from a single input dc voltage. This article introduces the permutation-graph (PG) method to systematically optimize multiple-voltage-domain dc distribution architectures using commercially available or custom single-input single-output dc–dc converter building blocks. The PG method enumerates possible converter arrangements and selects the converter blocks to achieve the best system figure of merit (FOM) in terms of system efficiency, size, cost, power density, or a combination of these metrics. In representative case study examples, considering a system operating from 48-V input dc voltage and requiring seven regulated point-of-load dc voltages {1, 1.3, 3.3, 5, 12, 24, and 40} V, it is shown how the proposed approach can improve the system FOM by more than 50% compared with solutions offered by a commercially available tool using standard dc–dc blocks. It is also shown how the inclusion of custom dc–dc blocks results in a different optimal arrangement, which yields further system FOM improvements. The approach is validated by experimental results on two 120-W system prototypes, providing five regulated dc voltages {1, 3.3, 5, 12, and 24} V from a 48-V dc bus.

A High-Frequency Planar Transformer with Medium-Voltage Isolation

Abstract: This paper presents the design of a low-loss, high-frequency planar transformer having medium-voltage (10’s of kV) isolation capability while transformer primary and secondary windings are interleaved to reduce losses. Medium-voltage isolation between adjacent printed circuit board (PCB) layers is extremely challenging using traditional PCB dielectrics. The isolation requirement is met using PCB with 7 kV/mil polyimide (Panasonic Felios RF775) as the dielectric, and by an appropriate layout of the windings and the inter-winding vias. The transformer is used to implement a dual active bridge (DAB) converter in a stackable dc-ac architecture where the dc port is connected to a photovoltaic (PV) string and ac outputs are connected in series to achieve direct PV string-to-medium voltage conversion without the need for low-voltage collection or a bulky line frequency transformer. Since each DAB transformer processes time-varying power, a design methodology is developed to minimize line-cycle-averaged losses. Experimental results are presented for a 1:1 planar transformer in a 7.5 kW SiC-based dc-to-ac module operating at 200 kHz. Isolation of 26 kV between the primary and secondary layers and between the windings and the core is verified using a hipot tester.

Machine Learning Estimators for Power Electronics Design and Optimization

Abstract: In the design and optimization of power electronics systems consisting of multiple converters, there is a need to characterize performance of a large number of available converter modules in order to determine the system performance characteristics such as efficiency or size. This characterization process can be computationally expensive, requiring multiple module-level design and optimization steps. Machine learning (ML) based estimators can be used to reduce the computational expense associated with dc-dc converter characterization. This paper compares ML techniques for generating dc-dc converter performance models and shows that random forest and gradient boosting techniques are capable of accurately predicting the performance of commercially available dc-dc converters. Test cases show that system optimization approaches using ML estimators can generate power electronics systems with Figure of Merit (FOM) expressed in terms of efficiency and size within 15% of systems designed using high fidelity but computationally expensive models.

Systematic Design and Optimization of Large DC Distribution Architectures Using Simulated Annealing

Abstract: Electronics applications often require DC distribution architectures capable of producing multiple DC outputs from a single input using a collection of single-input, single-output dc-dc converters. Prior work has described exhaustive-search based optimization and design techniques that can produce high quality distribution architectures when the number of required outputs is low. However, the design space becomes intractable when the number of outputs increases beyond single digits. Presented in this work is a technique that can efficiently optimize large systems without the use of design space heuristics. Test cases show that the optimization algorithm, which is based on simulated annealing, can increase system figure of merit by more than 18 % compared to a heuristic for a system with 22 outputs and reduce solution time by 63 % compared to an exhaustive-search approach for a system with eight outputs.

Proximate Time-Optimal Control of Flying-Capacitor Multi-Level Converters Using a Fixed Frequency PID Framework

Abstract: The paper proposes a near time-optimal control method for flying-capacitor multi-level (FCML) converters to improve the transient performance under large-signal output voltage reference or step-load transients. The proposed control scheme is based on a simple voltage-mode proportional-integral-derivative (PID) framework, in which the proportional gain is tuned using a state-plane diagram to achieve proximate time-optimal transient recovery. A direct output voltage derivative replaces the capacitor current, while a small integral gain is enabled only during the steady-state operation. Importantly, the proposed control scheme uses a single unified controller structure and a fixed-frequency modulator at all times, which ensures seamless transitions between transients and steady-state operation. The approach is verified on an eight-level digitally controlled GaN-based FCML prototype. Experimental results obtained using the proposed approach demonstrate a five-fold reduction in the rise time and a two-fold reduction in the fall time compared to a conventional small-signal based tuning methods, for a step-reference transient from 25 V to 180 V and back respectively, at $v_{in}=200 {\mathrm V}$ and load resistance $R=100 \Omega$.

A Novel Decentralized PWM Interleaving Technique for Ripple Minimization in Series-stacked DC-DC Converters

Abstract: Cascaded dc-dc converters are commonly used in applications where distributed energy sources or loads are connected to elevated voltage levels for power transfer. In such systems, it is advantageous to minimize the ripple on the bus current and voltage by proper phase shifting of the pulse-width modulation (PWM) pulses among the converters via a method known as interleaving. Existing approaches use either a centralized controller or separate communication lines among the stacked converters to control their relative PWM switch transitions. The key drawbacks are that these methods entail significant wiring, the central controller acts as a single point of failure, and implementation on very large numbers of units is impractical. In this paper, we introduce a decentralized interleaving control (DIC) strategy that acts on local current measurements at every converter and achieves communication-free PWM interleaving among the series-stacked converters. The proposed controller is simple in structure and is shown to converge asymptotically to the interleaved state irrespective of clock drifts among the digital signal processors. Experimental results are provided for a system of five series-connected converters showing a 10× reduction in the current ripple compared to normal operation.

Sliding Mode Control with Minimum-Deviation Transient Response for Non-Inverting Buck-Boost DC-DC Converters

Abstract: Recent work has demonstrated that utilization of all possible switching states in non-inverting buck-boost (NIBB) DCDC converters leads to substantial transient response improvements compared to responses in buck-only or boost-only modes of operation. This paper presents a hybrid digital controller for NIBB converters, which utilizes PID control and sliding mode control based the novel control algorithm. The sliding mode controller executes the transient response with an optimal sliding surface achieving the lowest possible output voltage deviation. The implemented digital controller enables seamless transition to PID regulation, which handles standard fixed-frequency steady-state regulation. The proposed solution is verified by circuit simulations and experimental results on a digitally controlled low-voltage NIBB converter suitable for mobile applications.

Composite Hybrid Energy Storage System utilizing Capacitive Coupling for Hybrid and Electric Vehicles

Abstract: An innovative architecture is presented that combines energy-dense and power-dense battery packs through a supercapacitor that provides capacitive coupling and a low-power DC-DC converter that provides energy balancing. A sizing algorithm is developed to optimize the design of such systems for plug-in hybrid and battery electric vehicles (PHEVs and BEVs). The proposed composite architecture extends vehicle range and battery lifetime by fully utilizing the capabilities of energy-dense and power-dense battery chemistries. A power-dense battery is coupled to an energy-dense battery using a small supercapacitor module that naturally distributes the system current between the two packs, requiring no additional contactors or full-power processing DC-DC converters. The proposed algorithm provides a tool for designing the composite architecture to achieve maximum weight reduction under given conditions for both ideal and practical scenarios. A design example is provided based on a PHEV with 68-mile range using the US06 drive cycle. The design achieved a 42% weight reduction when compared to a similar design with a conventional single chemistry battery system. Experimental results of a 1.5 kW, 0.2 kWh small-scale prototype with 25 Ah NMC and 2.9 Ah LTO battery cells and a 30 F supercapacitor verify the natural distribution of system current between the energy-dense and power-dense packs.

A Transformerless Composite Step-Down DC-DC Converter with Wide Input Voltage Range

Abstract: This paper is focused on a composite step-down dc-dc converter architecture capable of operating efficiently over wide input voltage range. The composite architecture consists of a transformerless fixed-ratio stage (DCX) and a non-inverting buck-boost stage, which provides regulation while processing a relatively small fraction of the load power. The DCX stage is based on a 4-to-l transformerless stacked active bridge (TSAB) converter and a capacitively isolated l-to-l dual active bridge (DAB) converter. A high overall system efficiency is obtained over a wide input voltage range as the bulk of the indirect power is processed through the highly efficient DCX stage. A decoupled control scheme is presented for the composite system, which allows for fast output voltage regulation as well as efficient operation of the DCX stage close to its nominal conversion ratio. Experimental results are provided for a composite converter prototype delivering a well regulated 12V output voltage with a maximum output power of 120W, while the input voltage varies from 48 V to 65 V. Using 30V and 40V rated Silicon devices and small planar coupled inductors, the measured power stage efficiency peaks at 97.8% and is greater than 96% over the 48- 65V input voltage range and the 30-100% output power range.

Efficient Design and Optimization of Large DC Distribution Architectures Using Descent Based Methods

Abstract: Complex dc distribution architectures are often required in electronic systems to generate multiple point-of-load regulated dc outputs from a single dc input. Typically these systems are realized using a set of single-input, single-output dc-dc modules. The system optimization involves choosing an arrangement of dc-dc modules to minimize the system losses, size or a user-defined figure of merit. Exhaustive-search based system optimization methods become intractable for large systems, i.e., when the number of required outputs, n o , exceeds eight. Presented in this work is an efficient descent-based design automation technique that optimizes the system without the use of ad-hoc heuristics or complexities associated with an alternative efficient method based on simulated annealing. Test cases show that the proposed algorithm can, for small and medium problems, decrease solve time by up to 60 % with no more than a 1 % decrease in figure of merit compared to the exhaustive search and, for large systems, finds a solution with a figure of merit 1 % better compared to simulated annealing with a 38 % increase in solve time.