Вы находитесь на странице: 1из 4

IEEE TRANSACTIONS ON MAGNETICS, VOL. 52, NO.

7, JULY 2016

8400804

Planar PCB Transformer Model for Circuit Simulation


Lew Andrew R. Tria1,2 , Daming Zhang1, and John E. Fletcher1
1 School

of Electrical Engineering and Telecommunications, University of New South Wales, Sydney, NSW 2052, Australia
and Electronics Engineering Institute, University of the Philippines Diliman, Quezon City 1101, Philippines

2 Electrical

An equivalent circuit model for planar printed circuit board (PCB) transformers is presented. The model utilizes the 1-D analysis
of Maxwells equations to develop a frequency-dependent representation of a multilayer, planar PCB transformer that can be
implemented in the circuit simulation software. In this transformer model, each conductor layer is implemented as a complex
impedance network, while each insulator layer is implemented as an air-cored inductor. Each magnetic core layer is modeled
as a non-linear inductance whose magnetic characteristic is based on a temperature-dependent JilesAtherton hysteresis model.
These impedances and inductances are then arranged side by side as they are arranged in the layer stack of the actual planar
PCB transformer. Through this model, the skin and proximity effect in the conductors and current distribution across windings
can be simulated. The developed model also enables the modeling of temperature-dependent hysteresis and saturation effects in
the magnetic material. The model provides a simpler method to derive the core and winding loss of the transformer than using a
finite-element analysis software. It also enables direct integration to circuit simulation tools. A prototype planar PCB transformer
was used to obtain experimental data for model validation. Comparisons made show good agreement between the performance of
the model and experimental results.
Index Terms JilesAtherton (JA) hysteresis model, lumped element model, printed circuit board (PCB) transformer.

I. I NTRODUCTION

LANAR printed circuit board (PCB) transformers are


widely used in high-frequency power electronic converters in order to achieve high-power density supplies. With
larger surface areas, planar transformers offer better thermal
performance than conventional magnetics. Using windings that
are pre-formed on, for example, PCBs, planar transformers
can be mass-manufactured and integrated with circuitry with
high repeatability. In addition, thinner and wider conductors
assist in reducing high-frequency losses due to skin and
proximity effects. Due to these advantages, numerous studies
have modeled planar transformers and inductors to predict and
characterize their performance and to improve their design.
One of the models developed particularly for multilayer planar magnetic structures is a lumped element model
based directly on the 1-D analysis of Maxwells equations.
These include lumped element models based on transmission
lines [1] and optical systems analogies [2]. A more intuitive
and systematic approach is presented in [3]. The model presented in [3] is most appropriate in circuit simulators, since
all the parameters are in the circuit domain (i.e., magnetic
parameters are converted into the electric domain). However,
these models are limited to the linear magnetic characteristics
of the transformer. In these lumped element models, the core
is represented as a lossless and linear component. In the linear
models, magnetic hysteresis and core loss are not considered.
In some cases, as in [4], a loss component is added that represents any loss unaccounted for in the total component loss,
but non-linear characteristics are still not considered. To model
the dynamic and transient response of a magnetic component,
such as inrush current, sub-harmonic, and chaotic responses,
the inclusion of the hysteresis of the magnetic core is essential.

Manuscript received November 6, 2015; revised January 2, 2016; accepted


January 4, 2016. Date of publication January 12, 2016; date of current
version June 22, 2016. Corresponding author: L. A. R. Tria (e-mail:
lew.tria@student.unsw.edu.au).
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TMAG.2016.2516995

A non-linear model appropriate for PCB transformers


is presented in this paper. The conductors and PCB are
represented as complex impedances and air-cored inductors, respectively. The magnetic core layers are represented as
non-linear inductances whose magnetic characteristic is based
on a temperature-dependent JilesAtherton (JA) model. With
this model, winding loss, current distribution among windings,
and core hysteresis loss can be calculated using the circuit
simulation software. The model is implemented in Simulink.
The performance of the model is validated by experimental
results using a prototype planar PCB transformer.
This paper is presented as follows. Section II presents
the proposed model. The experimental setup is described in
Section III. Simulation results and comparison with experimental results are presented in Section IV. Section V concludes
this paper.
II. E QUIVALENT C IRCUIT M ODEL
The equivalent circuit model is derived directly from the
1-D analysis of Maxwells equations. In this analysis, it is
assumed that the electromagnetic fields are parallel inside
and between the layers of the transformer. This assumption
is valid as long as the core materials permeability is high
and the conductors occupy the full breadth of the winding
window [3]. The transformer winding and insulator models
were adopted from [3]. A temperature-dependent non-linear
model for magnetic core layers is introduced in this model.
A. Transformer Winding
Fig. 1(a) shows a single-turn conductor of length d and
width w. A voltage V is applied across the terminals of the
conductor layer and current I flows through the conductor.
E T and E B are the electric field running along the length
on the top and bottom faces of the conductor, respectively,
while HT and H B are the magnetic fields along the width
of the conductor and on the top and bottom faces of the
conductor, respectively. Solving the Helmholtz equation and
using Amperes law yield a set of equations that captures the

0018-9464 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

8400804

IEEE TRANSACTIONS ON MAGNETICS, VOL. 52, NO. 7, JULY 2016

Fig. 1.
(a) One-turn conducting layer showing dimensional details and
E and H fields. (b) Equivalent three-port impedance network of the singleturn conductor layer [3].

relationship between the E and H fields in (1) and (2), which


can be represented by the three-port impedance network, as
shown in Fig. 1(b). The values of the impedances in Fig. 1(b)
depend on the layer geometry and material properties given
in (3)
(1)
E T = Z a HT + Z b (HT H B )
(2)
E B = Z b (HT H B ) Z a H B
(1 e h i )
2 e h i
Z ai =
; Z bi =
(3)
(1 + e h i )
(1 e2 h i )
where = (1 + j)/ is the propagation constant,
= (2/(r 0 ))(1/2) is the skin depth of the material at
the angular frequency , 0 is the permeability of free space,
r is the materials relative permeability, is the materials
conductivity, and h i is the layer thickness.
B. Insulators and Air Gaps
Spacings, insulators, and PCBs are modeled by air-cored
inductances whose impedance is given by (4), where ai is
the thickness of the space or insulator between layer i and
layer i + 1
Z Si = j 0 ai .

(4)

C. Magnetic Core
The magnetic core layer is represented by a nonlinear inductance whose characteristic is modeled using a
temperature-dependent JA hysteresis model [5], [6]. The
JA model uses the magnetic field, H , as the independent
variable and magnetization, M, as the dependent variable.
According to the JA model, the total magnetization, M, is
composed of an irreversible component, Mirr , and a reversible
component, Mrev
M = Mirr + Mrev .

(5)

A differential equation describing the dependence of magnetization on the magnetic field is provided in [8]
Man M
d Man
dM
= (1 c)
+c
dH
dH
sign( H )k(1 c) (Man M)
(6)

Fig. 2.
Prototype planar transformer. (a) Actual transformer prototype.
(b) Transformer geometry and winding arrangement. The magnetic path length
lm = 41.7 mm and the cross-sectional area of the center leg AC = 126 mm2 .

Fig. 3. Setup for acquiring the experimental performance of the transformer.

where


=

0,
1,

if sign( H ) (Man M) 0
otherwise.

Man is the anhysteretic magnetization and He is the effective


magnetic field given by (7) and (8), respectively
He = H + M


a
He

.
Man = Ms coth
a
He

(7)
(8)

In (6)(8), a, , c, k, and Ms are the JA parameters that can


be derived by curve-fitting experimental hysteresis curves. The
JA model has already been extended to include the temperature
dependence on the magnetic hysteresis in [7]. Equations to
reflect the temperature dependence on the JA parameters are
presented in
Ms (T ) = Ms0 (1 T /TC )
k(T ) =
a(T ) =
(T ) =
c(T ) =



2 TTc
k0 e


2 TTc
a0 e
2 T
0 e T c (1 T /TC )
2 T
c0 e T c (1 T /TC )

(9)
(10)
(11)
(12)
(13)

TRIA et al.: PLANAR PCB TRANSFORMER MODEL FOR CIRCUIT SIMULATION

8400804

Fig. 4. Equivalent model of the prototype PCB transformer. (a) Implementation of the model in Simulink. (b) Implementation of the non-linear impedance
for the core layer. Vpri is a sinusoidal voltage source with variable voltage and frequency. Each impedance and impedance network represents a material layer
in the actual transformer layer stack, as shown in Fig. 2.

where M0 , a0 , 0 , c0 , and k0 are the JA parameters at T = 0 K.


is a material-dependent critical component and TC is the
Curie temperature.
Therefore, using (6) and Faradays law, a magnetic core
layer is modeled as an inductance with voltage, d/dt, which
is a non-linear function of H


dBA
d(H + M)
dH
dM
d
=
= d0 h F
= d0 h F
1+
dt
dt
dt
dt
dH
(14)
where h F is the magnetic core layer thickness. This was
implemented as a non-linear inductor, as shown in Fig. 4(b).
D. Model Implementation
The impedance networks and inductors representing each
material layer are connected side by side as they are arranged
in the physical layer stack of the transformer. Conducting
layers connected by cross-layer structures are represented by
connecting their input ports according to how they are connected in the actual transformer. A layer port is represented,
as in [3], by an ideal transformer with turns ratio, m i :1, where
m i is the number of series-connected turns in layer i .
Thus, the prototype planar PCB transformer, shown in
Fig. 2, is modeled as in Fig. 4(a). Here, the impedances are
multiplied by a factor of d/w to convert the electromagnetic
system into an electric circuit domain for implementation in
the circuit simulation software. It should be noted that the
top and bottom parts of the core are considered as separate
layers in the multilayer stack.
III. E XPERIMENTAL S ETUP
The prototype planar PCB transformer is shown in Fig. 2(a).
The transformer has a turns ratio of 1:6. The winding
was made of three double-sided 1.6 mm FR4 PCB with
35 m-thick copper traces. PCBs were separated by 0.5 mm

thick insulators. The primary winding is composed of


two sets of parallel-connected windings, each with three turns
of 2.4 mm wide copper traces. The secondary winding is composed of 18 turns of 0.4 mm wide copper traces in two layers.
The windings are interleaved in a P-S-S-P configuration. The
magnetic core is a planar E-E core of the 3F3 material and
with dimensions shown in Fig. 2(b).
Fig. 3 shows the setup to obtain the experimental data.
A sinusoidal input voltage is supplied by a signal generator driving a power amplifier. The temperature-dependent
characteristics of the prototype transformer were obtained
using a controlled-temperature chamber. Signal amplitude was
varied at 1 Vpk increments until the transformer exhibited
saturation characteristics. This was conducted for frequencies
from 20 to 100 kHz at 20 kHz interval. The load, placed at the
secondary side, is a 50 resistor. The primary and secondary
voltages and currents were obtained using an oscilloscope with
voltage and current probes and data processed in MATLAB.
Tests were conducted at a temperature range of 20 C80 C
at 10 C intervals.
To acquire the hysteresis curve of the magnetic core, an
open-circuit test was conducted on the transformer. The magnetic field was derived directly from the input current, i p , as
the secondary side is left open circuited, using
H j = N p i p, j /lm

(15)

where N p is the number of turns in the primary and lm is the


magnetic path length of the core. The flux density is derived
from Faradays law, using the secondary voltage, vs , referred
to the primary side, to discount the effect of the winding series
impedance

(Vs, j /n)
t
(16)
B j = 1/(N Ac )
where Ac is the cross-sectional area of the center leg of the
core, N is the number of turns in the primary, n is the turns
ratio, and
t is the sampling time.

8400804

IEEE TRANSACTIONS ON MAGNETICS, VOL. 52, NO. 7, JULY 2016

TABLE I
JA PARAMETERS

Fig. 5.
Comparison of experimental and simulated waveforms. Primary
voltage V p = 5 Vpk and frequency f = 100 kHz in loaded condition
R L = 50 .

Fig. 6.
Comparison of experimental and simulated hysteresis curves at
different values of temperature, T , with f = 10 kHz. (a) T = 20 C.
(b) T = 40 C. (c) T = 60 C. (d) T = 80 C.

IV. R ESULTS
The JA parameters were derived at each temperature value
by curve-fitting experimental BH curves using a differential
evolution-based method as presented in [8]. TC , Curie temperature, was derived from the data sheet value for the
material used. was then derived by curve fitting M S (T ).
With and TC , the initial conditions (at T = 0 K) used in
the model, M S0 , a0 , 0 , c0 , and k0 were derived by curvefitting experimental results for (9)(13). The values of the
J-A parameters derived and used in the model are presented
in Table I.
The model was implemented in Simulink, as shown
in Fig. 4, using the same excitation voltage waveforms, Vpri ,
used in the experiments. Good agreement between the experimental and simulated signals has been derived, as shown
in Fig. 5, for a sinusoidal voltage input, Vpri = 5 Vpk ,
frequency 100 kHz, temperature of 20 C, and a resistor load,
R L = 50 . Possible reasons for model deviation from
experimental results include 2-D fields introduced by vias,

terminals, and air gaps, and non-inclusion of terminal and


wiring losses in the model.
An open-circuit test was also conducted on the model to
acquire a simulated hysteresis curve at various temperature
conditions. The models hysteresis curve showed good agreement with the experimental results as shown in Fig. 6. Small
deviation from experiment may be attributed to the accuracy
in the estimation of the JA parameters.
V. C ONCLUSION
This paper presents a model for the planar PCB transformer
for the implementation in the circuit simulation software. The
model simulated the performance of a prototype transformer,
including its non-linear characteristic and saturation effects, at
various temperature conditions. The model presents a simpler
and cheaper alternative to the finite-element analysis (FEA)
simulation. With this model, winding loss can be derived
with only the geometry and material properties. Since the
model captures the hysteresis curve of the core, a temperaturedependent core loss can also be derived. The accuracy of the
model is limited by the assumption that the electromagnetic
fields are uniform along the breadth of the winding window
and that the field is changing only in the vertical dimension.
For structures that does not satisfy this assumption, a 2-D or
3-D FEA may be a more appropriate model. The accuracy
of the hysteresis is tied to the JA parameters. Thus, the
accurate determination of these parameters is essential. Only
the static JA model was used in this paper, and for better
accuracy and operation over a wider range of frequencies, the
dynamic JA should be implemented. Circuit losses, such as
those contributed by terminals, source, and connecting wires,
can be incorporated into the model as external impedances for
a more accurate model. The model assumes sinusoidal excitations. For non-sinusoidal waveforms, such as those commonly
used in switched-mode power electronics, the effect of each
harmonic content shall be considered and recommended for
further study.
R EFERENCES
[1] R. Prieto, J. A. Oliver, J. A. Cobos, and M. Christini, Magnetic
component model for planar structures based on transmission lines,
IEEE Trans. Ind. Electron., vol. 57, no. 5, pp. 16631669, May 2010.
[2] J.-P. Keradec, B. Cogitore, and F. Blache, Power transfer in a twowinding transformer: From 1-D propagation to an equivalent circuit,
IEEE Trans. Magn., vol. 32, no. 1, pp. 274280, Jan. 1996.
[3] M. Chen, M. Araghchini, K. K. Afridi, J. H. Lang, C. R. Sullivan, and
D. J. Perreault, A systematic approach to modeling impedances and
current distribution in planar magnetics, IEEE Trans. Power Electron.,
vol. 31, no. 1, pp. 560580, Jan. 2016.
[4] A. Schellmanns, P. Fouassier, J.-P. Keradec, and J.-L. Schanen, Equivalent circuits for transformers based on one-dimensional propagation:
Accounting for multilayer structure of windings and ferrite losses, IEEE
Trans. Magn., vol. 36, no. 5, pp. 37783784, Sep. 2000.
[5] D. C. Jiles and D. L. Atherton, Theory of ferromagnetic hysteresis
(invited), J. Appl. Phys., vol. 55, no. 6, pp. 21152120, Mar. 1984.
[6] D. C. Jiles and J. B. Thoelke, Theory of ferromagnetic hysteresis:
Determination of model parameters from experimental hysteresis loops,
IEEE Trans. Magn., vol. 25, no. 5, pp. 39283930, Sep. 1989.
[7] A. Raghunathan, Y. Melikhov, J. E. Snyder, and D. C. Jiles, Theoretical
model of temperature dependence of hysteresis based on mean field
theory, IEEE Trans. Magn., vol. 46, no. 6, pp. 15071510, Jun. 2010.
[8] D. Zhang, Y. Liu, and S. Huang, Differential evolution based parameter
identification of static and dynamic J-A models and its application to
inrush current study in power converters, IEEE Trans. Magn., vol. 48,
no. 11, pp. 34823485, Nov. 2012.

Вам также может понравиться