An Improved Linear Model for High Frequency Class-DE Resonant Converter using the Generalized Averaging Modeling Technique

Research output: Contribution to journalJournal articleResearchpeer-review

48 Downloads (Pure)

Abstract

As the operating frequency of power converters increases, the passive component values likewise decrease. This results in the effect of the parasitic components becoming
more prominent, leading to significant modeling errors if not considered. For resonant converters this especially becomes a problem at high frequencies. This paper presents a reduced model for a class-DE series resonant converter based on generalized averaging that incorporates the relevant parasitics and uses multiple harmonics to obtain an accurate linear model. Comparison between the proposed model, prior art, and a prototype converter running at 1 MHz is conducted, and a PI-controller is designed based on each model and tested. The results show that the parasitics have a significant impact on the DC-gain and dynamics of the converter, and that the proposed model improves on the prior art by reducing the DC-gain error by more than 7 dB, and the error in the low frequency pole from 168 % to 16.9 %. Furthermore, the PI-controller designed on the prior art was found to have more than 40 times larger overshoot in the control signal when measured compared to the model prediction, while the controller based on the proposed model showed correct performance when simulated and measured.
Original languageEnglish
JournalI E E E Journal of Emerging and Selected Topics in Power Electronics
Number of pages12
ISSN2168-6777
Publication statusAccepted/In press - 2019

Keywords

  • Resonant converters
  • Modeling
  • PI control
  • State-space methods
  • DC-DC power converters

Cite this

@article{5ba2c7395c9f4c48a78cc7eb51e548c5,
title = "An Improved Linear Model for High Frequency Class-DE Resonant Converter using the Generalized Averaging Modeling Technique",
abstract = "As the operating frequency of power converters increases, the passive component values likewise decrease. This results in the effect of the parasitic components becomingmore prominent, leading to significant modeling errors if not considered. For resonant converters this especially becomes a problem at high frequencies. This paper presents a reduced model for a class-DE series resonant converter based on generalized averaging that incorporates the relevant parasitics and uses multiple harmonics to obtain an accurate linear model. Comparison between the proposed model, prior art, and a prototype converter running at 1 MHz is conducted, and a PI-controller is designed based on each model and tested. The results show that the parasitics have a significant impact on the DC-gain and dynamics of the converter, and that the proposed model improves on the prior art by reducing the DC-gain error by more than 7 dB, and the error in the low frequency pole from 168 {\%} to 16.9 {\%}. Furthermore, the PI-controller designed on the prior art was found to have more than 40 times larger overshoot in the control signal when measured compared to the model prediction, while the controller based on the proposed model showed correct performance when simulated and measured.",
keywords = "Resonant converters, Modeling, PI control, State-space methods, DC-DC power converters",
author = "Dahl, {Nicolai Jerram} and Ammar, {Ahmed Morsi} and Arnold Knott and Andersen, {Michael A. E.}",
year = "2019",
language = "English",
journal = "I E E E Journal of Emerging and Selected Topics in Power Electronics",
issn = "2168-6777",
publisher = "Institute of Electrical and Electronics Engineers",

}

TY - JOUR

T1 - An Improved Linear Model for High Frequency Class-DE Resonant Converter using the Generalized Averaging Modeling Technique

AU - Dahl, Nicolai Jerram

AU - Ammar, Ahmed Morsi

AU - Knott, Arnold

AU - Andersen, Michael A. E.

PY - 2019

Y1 - 2019

N2 - As the operating frequency of power converters increases, the passive component values likewise decrease. This results in the effect of the parasitic components becomingmore prominent, leading to significant modeling errors if not considered. For resonant converters this especially becomes a problem at high frequencies. This paper presents a reduced model for a class-DE series resonant converter based on generalized averaging that incorporates the relevant parasitics and uses multiple harmonics to obtain an accurate linear model. Comparison between the proposed model, prior art, and a prototype converter running at 1 MHz is conducted, and a PI-controller is designed based on each model and tested. The results show that the parasitics have a significant impact on the DC-gain and dynamics of the converter, and that the proposed model improves on the prior art by reducing the DC-gain error by more than 7 dB, and the error in the low frequency pole from 168 % to 16.9 %. Furthermore, the PI-controller designed on the prior art was found to have more than 40 times larger overshoot in the control signal when measured compared to the model prediction, while the controller based on the proposed model showed correct performance when simulated and measured.

AB - As the operating frequency of power converters increases, the passive component values likewise decrease. This results in the effect of the parasitic components becomingmore prominent, leading to significant modeling errors if not considered. For resonant converters this especially becomes a problem at high frequencies. This paper presents a reduced model for a class-DE series resonant converter based on generalized averaging that incorporates the relevant parasitics and uses multiple harmonics to obtain an accurate linear model. Comparison between the proposed model, prior art, and a prototype converter running at 1 MHz is conducted, and a PI-controller is designed based on each model and tested. The results show that the parasitics have a significant impact on the DC-gain and dynamics of the converter, and that the proposed model improves on the prior art by reducing the DC-gain error by more than 7 dB, and the error in the low frequency pole from 168 % to 16.9 %. Furthermore, the PI-controller designed on the prior art was found to have more than 40 times larger overshoot in the control signal when measured compared to the model prediction, while the controller based on the proposed model showed correct performance when simulated and measured.

KW - Resonant converters

KW - Modeling

KW - PI control

KW - State-space methods

KW - DC-DC power converters

M3 - Journal article

JO - I E E E Journal of Emerging and Selected Topics in Power Electronics

JF - I E E E Journal of Emerging and Selected Topics in Power Electronics

SN - 2168-6777

ER -