International Conference on Renewable Energies and Power Quality (ICREPQ’17)

Malaga (Spain), 4th to 6th April, 2017

Renewable Energy and Power Quality Journal (RE&PQJ)
ISSN 2172-038 X, No.15 April 2017

Non-linear inductor modelling for a DC/DC Buck converter

G. Lullo1, D. Scirè1 and G. Vitale2.

1
Dipartimento di Energia, ingegneria dell’Informazione e modelli Matematici, Università di Palermo,
Viale delle Scienze, Edif.9, 90128 Palermo (Italy), email: giuseppe.lullo(at)unipa.it, daniele.scire91(at)gmail.com
2
Istituto di Studi sui Sistemi Intelligenti per l’Automazione (ISSIA), Consiglio Nazionale delle Ricerche (CNR),
Via Dante Alighieri, n. 12 - 90141 Palermo (Italy), corresponding author, email: gianpaolo.vitale(at)cnr.it

Abstract. The paper is focused on the modelling of an during the switching period. For this reason a suitable
inductor to exploit its non-linear behaviour in the roll-off region model is required.
when it is employed in a DC/DC converter. The model is set up Some models have been proposed in literature. In [12] a
on the basis of experimental data measured in operating first approximation is obtained using Finite Element
conditions, which are fitted to a polynomial curve describing the Analysis (FEA), then the model is verified by comparing
inductance variations. The analysis of the buck converter, the results obtained from the simulation with
performed by including the proposed model, is validated by measurements. Reference [13] proposes a model of
experimental tests.
magnetic components which integrates a nonlinear model
of hysteresis, electro-magnetic windings and thermal
behavior in a single model for use in circuit simulation of
Key words power electronics systems.
Inductor, modelling, moderate saturation, roll-off region.
The operating efficiency of the full bridge converters via
one and two output inductors was evaluated and analysed
1. Introduction in [14], where to take into account the magnetic saturated
effects of the core material, a nonlinear inductance model
Switch-mode power supplies (SMPS) in step-down and is used. A dynamic model including the effects of
step-up configuration are widely used in many portable saturation, hysteresis and eddy currents for lossy
applications especially where energy is obtained by inductors is presented in [15]. In [7] a behavioural
renewable sources [1-3]. The optimal design implies that a analytical model of inductance versus current saturation
high power density value is achieved. These circuits use curve, obtained starting from the data provided by
the inductor as an energy reservoir, which is energized by inductors manufacturers, is proposed and the inductor is
the voltage source and transfers energy to the load kept at analytically described by means of an arctangent
constant voltage by a capacitor. The inductor remains a function. In [16] three different models are described: an
bulky component of the converter since its core is average model, a piecewise linearized model and a model
dimensioned to prevent the onset of saturation. This based on the Volterra-series expansion.
approach raises the size and cost of the inductor[4]. In this paper a new model of the inductor, based on a
In most of the theoretical studies on the static and dynamic polynomial curve fitting of the inductance variation, is
performances of converters, the inductor is considered as proposed. The model is obtained by an experimental set
a linear device, i.e. a dipole in which the magnetic flux is up in which the inductor is DC biased and a flux ripple is
proportional to the current.In particular the inductor is superimposed. This approach allows the model to be
operated with a constant DC current value and a moderate identified when data provided by manufacturer are
ripple allowing the hysteresis curve to be considered as lacking or not sufficient. The suitability of the model
linear and the inductance as constant [5, 6]. Anyway, this has been assessed by comparing simulation and
jeopardizes the power density since during operation the experimental data obtained by a buck converter whose
inductor’s capabilities are not fully exploited [7, 8]. inductor has been previously characterized by following
Recently the use of inductors working in partial saturation the proposed method.
has been considered in literature where it has been This paper is organized as follows: in section two the
assessed that the advantages in using a more compact behaviour of a DC/DC Buck converter with inductor in
inductor imply an acceptable amount of losses [9-11]. moderate saturation is described. Section three is devoted
From the theoretical point of view, if the hypothesis of to the inductor characterization and modelling. A DC/DC
linearity for the inductor is removed, the value of the buck converter implementation is presented in section
inductor varies with the current. During its operation in a four and the experimental rig to validate the model is
DC/DC converter, when it is biased with a DC value of the described in section five. Simulation and experimental
flux with a superimposed ripple, the inductance will results are finally given in section six.
experience a variation from a minimum to a maximum

https://doi.org/10.24084/repqj15.433 686 RE&PQJ, Vol.1, No.15, April 2017

2. DC/DC buck converter with inductor in 3. Inductor characterization.
moderate saturation
A. Linear behaviour
The scheme of a Buck converter is shown in Figure 1,
where the equivalent series resistance (ESR) of the As it is known, an inductor is characterized by the
inductor and of the capacitor are neglected. following relationship between the magnetic flux (or
simply flux) and the current:
φ = iL (1)

where φ is the flux concatenated with the wires and L the

self-inductance (or simply inductance). The voltage vs.
current relationship for a linear inductor is obtained by
applying the Faraday’s law considering the inductance as
constant:

dφ di
Fig. 1. Buck converter scheme vL = =L L (2)
dt dt
In Continuous Conduction Mode (CCM), turning on the where the voltage and the current are defined according
switch increases the inductor current. When the switch is to Figure 3.
in “off” state , the inductor current “freewheels” through
the diode. During the “on” time the inductor experiences a
constant voltage and, if the inductor is linear, the current
raises with a constant slope given by (vin-vout)/L, whereas
during the “off” time the slope is given by (-vout)/L as
shown in Figure 2a. If the average current in the inductor
is increased, the weak saturation region is reached at the
end of the “on” time. As a consequence, the maximum Fig. 3. Inductor dipole representation
current will be higher, if compared with the linear case,
and the shape will be different according to the variation of
the inductance (Figure 2b). B. Commercial power inductors characteristic

The inductors used in DC/DC converter applications

often exhibit an inductance whose value is variable with
the current. Figure 4 shows, as an example, the variation
of inductance of a set of commercial inductors vs.
current, where LR and IR denote the rated values while Lop
and Iop are the operating values [17]. As example the
inductor B2302M001 corresponds to LR=2.5 mH and
IR= 3A, it can be noted that the maximum value of the
inductance is exhibited for low currents. The greater the
current, the smaller is the inductance, which assumes the
smallest value (about 30% of the maximum) when the
current reaches 1.4 times of the rated value. In a DC/DC
converter the inductor is operated around a DC point. If
the ripple of the current around this point is small, the
inductor will show a small variation of the flux and
equation (2) is still valid. However in the design phase, it
implies an oversizing of the inductor as a great part of
magnetic material remains unexploited. On the other
hand, for larger current ripple around the DC point, an
increase of the maximum current will be expected and
the variation of the inductance vs. current has to be
considered. As a consequence the model described by
(2) is no longer valid.
In the curve of the inductance vs. current three regions of
operation can be identified: a) the weak saturation region,
Fig. 2. Inductor current vs. time in a Buck converter in case of:
in which the inductance slightly decreases with respect to
a) inductor operated in linear region, b) inductor operated in
moderate saturation.
LR , until about 70% of LR ; b) the roll-off region, where
the inductance swings from values close to LR to values
close to the value that represents the saturation Ldeepsat; c)
the deep saturation region, where the inductance is much

https://doi.org/10.24084/repqj15.433 687 RE&PQJ, Vol.1, No.15, April 2017

lower than about 30% of LR and approaches the value Comparing simulated and experimental results (see
corresponding to Ldeepsat. Figure 5), however it has been noted that the
The extension of the operating points to the roll-off region corresponding waveforms are different. It can be
implies a significantly higher current, flowing through the explained by a sensitivity analysis of (4). In particular,
inductor, compared to the exploitation of the weak the sensitivity respect to the coefficients Li is given by
saturation region only. the corresponding term of the current (6).

 L0 = 3.13 ⋅ 10 −3
 −4
 L1 = −3.21 ⋅ 10
−4
 L2 = −1.94 ⋅ 10 (5)
 L = 5.66 ⋅ 10 −5
 3
 L4 = −4.74 ⋅ 10 −6

∂L(i )
∆L = ∑ ∆Lk = ∑ i k ⋅ ∆Lk
k ∂Lk k =0 ,1,..,n
k
k =0 ,1,..,n

(6)

Eq. (6) shows that the error on the coefficients with lower
index is more relevant at lower current, on the contrary
the error on the coefficients with higher index is more
relevant at higher current.
A tuning algorithm has been used to adjust the
Fig. 4. Inductance vs. current of a commercial inductor coefficients. This algorithm provides the matching of the
coefficients starting from L0 obtained by a DC current of
C. Experimental characterization of the inductor 0.5A up to L4 which is optimized for a DC current of 4A.
The flow chart of the algorithm is shown in Figure 6. On
In order to identify the function L=L(i) in principle the this basis a new set of coefficients (7) has been obtained.
curves given by manufacturers could be used; These coefficients allow a good fitting of simulated and
unfortunately due to the fabrication tolerances this experimental data as shown in Figure 7.
introduces an appreciable error in the model. For this
reason we decided to characterize the inductor  L0 = 3.50 ⋅ 10 −3
experimentally. The measurement circuit is made by a  −4
simple Buck converter similar to Figure 1. The DC current  L1 = −1.20 ⋅ 10
if imposed by varying the load, while switching frequency, −4
input/output voltages and duty cycle are optimized in order  L2 = −1.30 ⋅ 10 (7)
to limit the current ripple, so to avoid the variation of the  L = 3.00 ⋅ 10 −5
inductance. On the basis of the voltage applied to the  3
inductor terminals and of the slope of the current, the  L4 = −1.90 ⋅ 10 −6
inductance is given by:

vL
L(i ) = (3)
diL
dt i = I DC

The measurement has been performed on a ring core

chokes with iron powder core, with rated inductance of
2.5 mH and rated current of 3 A, setting the DC current
equal to: 0.5, 1, 2, 3, 4, 5 and 6A. Then these points have
been interpolated by a polynomial curve. The best results
has been obtained by a 4th order curve:

L = L0 + L1i + L2i 2 + L3i 3 + L4i 4 (4)

Fig. 5. Simulated and experimental value of the current through
the inductor for a DC current of 4A before the tuning of the
with the coefficients given by (5). coefficients.

https://doi.org/10.24084/repqj15.433 688 RE&PQJ, Vol.1, No.15, April 2017

 The inductance curve obtained on the basis of the new set
of coefficient (7) of polynomial fitting is plotted in
Figure 8. The corresponding characteristics values are:
LR=3.5mH, Ldeepsat=2.3mH, L70%=3.14mH, L30%=2.66mH.

Fig. 8. Inductance vs current curve

D. Proposed model of the inductor

The inductor, as described by (4) with coefficients (7),

has been modelled to be used in the simulation analysis.
The magnetic flux is given by:

φ = i ⋅ L(i ) = L0i + L1i 2 + L2i 3 + L3i 4 + L4i 5 (8)

On the basis of the magnetic flux, eq. (8) is solved in

recursive form:

φ L1 2 L2 3 L3 4 L4 5
i= − i − i − i − i (9)
L0 L0 L0 L0 L0
with the constraint:
φ = ∫ vdt < max{i ⋅ L(i )} (10)

Fig. 6. Flow chart of the tuning algorithm This last condition imposes that the horizontal line of the
flux must intersect the curve sketched in Figure 9 so that
eq. (9) admits a solution. The solution used in simulation
lies on the left of the maximum.

Fig. 7. Simulated and experimental value of the current through

the inductor for a DC current of 4A after the tuning of the
coefficients. Fig. 9. Flux curve, given by i · L(i).

https://doi.org/10.24084/repqj15.433 689 RE&PQJ, Vol.1, No.15, April 2017

The implementation of (9) in Simulink® environment is
shown in Figure 10. TABLE II. COMPOSITION OF THE EXPERIMENTAL RIG
supplier function Model
TDK- TDK GEN600V-
Power Supply
Lambda 5.5A
Signal
Agilent 33521A
generator
Resistor RS 136-238 200W
passive Load 3R3 J 01.03
Electronic
Sorensen SLH 500V-6A
Load
Digital DMM4050 61∕2-
Tektronix
Multimeter Digit
Agilent Oscilloscope MSO6104A
Fig. 10. Implementation of eq. (9) in Simulink® environment

6. Results
4. DC/DC buck converter implementation.
A. Model validation
The DC/DC buck converter has been implemented in
Simulink® environment following the approach described The model has been validated comparing simulation and
in [18]. This approach does not require a specific circuit experimental data of the inductor current and of the
simulator and it is faster than using conventional circuit output voltage of the converter. Figures 11a and 11b
simulators, in addition it takes into account the parasitic show the shape of the current flowing through the
effects of inductors and capacitors. Instead of a linear inductor in case of small and large ripple respectively.
inductor, the model of Figure 10 has been adopted. Figure 11a has been obtained with a supply voltage of
In addition, to validate the simulation results, the circuit 30V, a switching frequency of 20 kHz, duty cycle equal
has been built in hardware form. The list of the to 0.5 and a DC current of 2 A; it can be noted that with a
components used for the experimental circuit is ripple of about 200 mA the shape of the current is the
summarized in table I. well-known triangular waveform.
TABLE I. : COMPONENTS OF THE DC/DC BUCK
CONVERTER

electrical Rated Supplier

code
symbol value
IR=3A, Epcos
L B82615-B2302M
LR=2.5mH
C 100 µF

VDSS = 100V International

HEXFET® Power
RDS(on) = Rectifier
T MOSFET
0.036Ω
IRFP150N
ID = 42A
IF(AV) =12 A ST Turbo 2 ultrafast
VRRM= 600 V Microelectronics, high voltage
D
trr (max) = Inc. rectifier
25 ns STTH12R06DI

Fig. 11a. Experimental waveforms measured in the DC/DC

converter with small ripple: output voltage (top), inductor
5. The test rig current (middle), duty cycle (bottom)

The test rig is composed of a power supply, the DC/DC The waveforms of Figure 11b have been obtained with
converter under test, the resistor load, a digital multimeter, the same DC current but, the switching frequency has
an oscilloscope equipped with voltage and current sensors been lowered to 3 kHz and the supply voltage has been
to acquire the duty cycle, the output voltage and the increased to 70V, maintaining the duty cycle of 0.5. This
current in the inductor of the DC/DC converter. The main causes the large ripple and the peaks on the current.
components of the experimental rig are summarized in These test conditions have been adopted to validate the
table II. model. The results are shown in the following Figures.

https://doi.org/10.24084/repqj15.433 690 RE&PQJ, Vol.1, No.15, April 2017

 The experimental and simulated output voltage are
shown in Figure 14. In this case a good agreement can be
noted as well.

Fig. 11b. Experimental waveforms measured in the DC/DC

converter with high ripple: output voltage (top), inductor current
(middle), duty cycle (bottom).

The experimental and simulated current are shown in Fig. 14. Comparison of the simulation and experimental data
Figure 12 and the relative error is shown in Figure 13. It for the output voltage
can be noted that in the worst case the error in a switching
period is lower than 13%.
B. Static Gain

The static gain, defined as the ratio of the output voltage

V0 and the input voltage Vs has been evaluated in
simulation considering the same DC point with a linear
inductor whose value is equal to Lo (Figure 15a) and with
the inductor modelled by eq. (8) (Figure 15b). The
switching frequency is equal to 6 kHz and the supply
voltage is equal to 70V. In an ideal case the Buck
converter should have a gain equal to the duty cycle D. It
can be noted that in both cases the intersection of the
surface with a plane at constant current gives a straight
line for low currents whereas it is influenced by a drop
voltage on the parasitic parameters increasing the current.
In any case the gain does not differ significantly in case
of linear inductor and non-linear one, as a matter of fact
the maximum error is equal to 2.5·10-3 %.
Fig. 12. Simulation and experimental data of the current flowing
through the inductor.

Fig. 15a. Static gain of the Buck converter adopting a linear

Fig. 13. Percentage relative error in the inductor current inductor
comparing simulation and experimental data

https://doi.org/10.24084/repqj15.433 691 RE&PQJ, Vol.1, No.15, April 2017

Fig. 15b. Static gain of the Buck converter adopting an inductor
operated in moderate saturation
Fig. 17. Output voltage step response with a small current
ripple on the inductor.
C. Step response

The simulated step response has been analysed in both

cases, i.e. when the DC/DC converter is operated with a
small ripple (obtained increasing the switching frequency
to 20 kHz) and with a large ripple (the inductor is
operated in roll-off region). The step has been obtained by
varying the load resistance from 70 Ω to 11 Ω. The current
flowing through the inductor and the output voltage when
the ripple is small are respectively plotted in Figure 16
and in Figure 17. In this last case the settling time
(evaluated in a range of 5%) is equal to 3.47 ms. The
initial voltage is equal to 34.3 V whereas the final voltage
is equal to 32.8V. The initial ripple on the voltage is equal
to 0.017 V and the final value is equal to 0.031 V.

Fig. 18. Inductor current step response with large ripple

Fig. 16. Inductor current step response with a small ripple

Figure 18 and 19 show the current through the inductor

and the output voltage with large ripple obtained lowering
the switching frequency to 6 kHz. The settling time of the
Fig. 19. Output voltage step response with large current ripple
output voltage (evaluated in a range of 5%) is equal to 2.62 on the inductor.
ms. The initial voltage is equal to 34.3 V whereas the final
voltage is equal to 32.8V. The initial ripple on the voltage
7. Conclusion
is equal to 0.19 V and the final value is equal to 0.34 V.
Comparing Figures 16 and 18 it can be noted that the
The exploitation of roll-off region of the inductor in a
increased ripple implies a higher current peak in the
DC/DC Buck converter has been analysed by a suitable
inductor as expected.
model. The non-linear model of the inductor in a DC/DC

https://doi.org/10.24084/repqj15.433 692 RE&PQJ, Vol.1, No.15, April 2017

Buck converter has been implemented and verified [13] Wilson PR, Ross JN, Brown AD. Simulation of
experimentally. magnetic component models in electric circuits including
From the analysis of the performance it has been noted that dynamic thermal effects. IEEE Transactions on Power
the operation in roll-off region does not influence the static Electronics 2002; 17:55–65.
gain of the converter. The exploitation of the roll-off [14] Wang JB, Li R, Chen J. Efficiency comparison of
region makes the dynamic response faster but causes an full bridge converters in considered magnetic saturation.
higher ripple on the output voltage and a higher peak of IEEE Industrial Electronics Conference, November 2008;
the current in the inductor. 717–722.
Moreover, the exploitation of the roll-off region allows [15] Chapman PL, Sudhoff SD. Dynamic lossy inductor
either the switching frequency to be lowered obtaining a model for power converter simulation. Applied Power
reduction of the switching losses or the inductor to be Electronics Conference, March 2002; 137–143.
designed with a smaller value of the inductance increasing [16] R. A. Mastromauro, M. Liserre, A. Dell’Aquila, “Study of
the power-density of the converter. the Effects of Inductor Nonlinear Behavior on the Performance
On the other hand, a current control loop is recommended of Current Controllers for Single-Phase PV Grid Converters,
to avoid that a current peak in the inductor could damage IEEE trans on Ind. Electr. Vol. 55, n. 5, May 2008.
[17] https://en.tdk.eu/tdk-en/530144/products/product-
the power switch or the diode. catalog/emc-components/power-line-chokes--epcos-
[18] G. Vitale, “Analysis of DC/DC converters by mean node
Acknowledgement voltage method”, Renewable Energy and Power Quality Journal
(RE&PQJ) ISSN 2172-038 X, No.13, April 2015.
This paper has been funded by the research project: RITmare,
Ricerca ITaliana per il mare (Italian Research for the sea)
CUP:B91J11000740001.

References
[1] Palma L, Todorovic MH, Enjeti P. Design considerations for
a fuel cell powered DC-DC converter for portable applications.
21 st Annual IEEE Applied Power Electronics Conference and
Exposition, March 19–23, 2006; 1263–1268.
[2] Vazquez MJ, Marquez JMA, Manzano FS. A methodology
for optimizing stand-alone PV-system size using parallel-
connected DC/DC converters. IEEE Transactions on Industrial
Electronics 2008; 55(7): 2664–2673.
[3] Yuan L, Qu G. Analysis of energy reduction on dynamic
voltage scaling-enabled systems. IEEE Transactions on
Computer-Aided Design of Integrated Circuits and Systems
2005; 24(12):1827–1837.
[4] L. Milner, G. Rincon-Mora, “Small saturating inductors for
more compact switching power supplies”, IEEJ transactions on
electrical and electronics engineering, trans 2012; 7; 69-73.
[5] N. Mohan, Power Electronics A First Course, John Wiley &
Sons, Inc. United States of America, 2012.
[6] S. Ang, A. Oliva, Power Swiching Converters, CRC Press,
Taylor and Francis Group, Boca Raton, FL 2011.
[7] G. Di Capua, N. Femia, “A Novel Method to Predict the Real
Operation of Ferrite Inductors With Moderate Saturation in
Switching Power Supply Applications”, IEEE Transactions on
Power Electronics, year: 2016, Volume: 31, Issue: 3.
[8] G. Di Capua, N. Femia, K. Stoyka, “Power magnetics volume
and weight reduction in aerospace power supply units”, 2016
IEEE 17th Workshop on Control and Modeling for Power
Electronics (COMPEL), Trondheim, Norway, 27-30 June 2016.
[9] R. A. Salas and J. Pleite, “Simulation of the saturation and
air-gap effects in a pot Ferrite Core with a 2-D finite element
model,” IEEE Trans. Magnet., vol. 47, no. 10, pp. 4135–4138,
Oct. 2011.
[10] R. A. Salas and J. Pleite, “Equivalent electrical model of a
Ferrite Core inductor excited by a square waveform including
saturation and power losses for circuit simulation,” IEEE Trans.
Magn., vol. 49, no. 7, pp. 4257–4260, Jul. 2013.
[11] A. Stadler, T. Stolzke, and C. Gulden, “Nonlinear power
inductors for large current crest factors,” in Proc. Eur. Power
Convers. Intell. Motion Conf., May 2012, pp. 1548–1553.
[12] Salas R, Pleite J, Olias E, Barrado A. Nonlinear
saturation modelling of magnetic components for circuit
simulation. IEEE International Magnetics Conference,
May 2006; 993.

https://doi.org/10.24084/repqj15.433 693 RE&PQJ, Vol.1, No.15, April 2017