IEEE TRANSACTIONS ON MAGNETICS, VOL. 44, NO.

11, NOVEMBER 2008

4127

A New Technique for Measuring Ferrite

Core Loss Under DC Bias Conditions
C. A. Baguley1 , U. K. Madawala1 , and B. Carsten2
Department of Electronics and Computer Engineering,University of Auckland, Auckland 48 New Zealand
Bruce Carsten Associates Inc., Corvallis, OR 97330-9243 USA
It is well known that the measurement of ferrite core losses is difficult, particularly under low power factor conditions for which the
excitation VA drawn is large relative to the power dissipated in the ferrite core. This paper proposes a new technique, called mutual
inductance neutralization, which overcomes this difficulty by compensating for the reactive power drawn by the circuit through the use
of an air cored mutual inductor. Based on the technique, circuits are presented which are appropriate for core loss measurement in the
absence, and the presence of a dc bias. Measured results are presented which verify the accuracy of the technique, and results are also
presented showing the effect of dc bias conditions on the losses of the ferrite core under test.
Index TermsCore losses, ferrite, switched-mode power supply (SMPS).

I. INTRODUCTION

II. MUTUAL INDUCTANCE NEUTRALIZATION

UE TO the large quantities of switched-mode power supplies (SMPSs) that are produced, strong economic and
regulatory pressures exist to minimize SMPS size and losses
[1][5]. Within SMPS circuits, ferrite-cored magnetic components operating under dc bias conditions are commonly used,
and are typically among the largest components. They therefore have a strong impact on SMPS size and losses, particularly in view of the knowledge that ferrite core losses increase
with dc bias in a highly nonlinear manner [6][12]. However,
investigation into the phenomenon is hampered by the difficulty
of accurately measuring ferrite core losses under dc bias [6].
This paper proposes a low-cost, accurate, expedient technique
for measuring ferrite core losses under dc bias conditions that,
to the knowledge of the authors, has not previously been presented.
Average core losses can be measured by exciting a winding on
a ferrite core under test (CUT) then integrating the product of the
over one cycle.
winding emf, and the excitation current
However, if the CUT is a low-permeability and low-loss type,
this procedure is prone to large core loss measurement errors. A
primary source of these errors is the unequal time delays intro, and winding emf
duced by parasitic impedances in the
sensing channels [13], [14]. An example illustrating the high
degree of time/phase matching necessary between the sensing
channels to avoid significant measurement errors is given in [6].
The susceptibility to this form of measurement error occurs because the phase shift between the
and winding emf waveforms is very close to 90 . The measurement technique presented in this paper, which determines core loss through integrating the product of sampled winding emf and
waveforms, overcomes this difficulty by phase-shifting these waveforms using an air-cored mutual inductor.

Digital Object Identifier 10.1109/TMAG.2008.2002862

The improvement to the power factor (pf) between the

waveforms is achieved using the
winding emf and
mutual inductance neutralization technique illustrated by
Fig. 1(a). The technique is implemented using an adjustable
air-cored mutual inductor with windings of a relative phasing
in Fig. 1(a) is measured, the reactive volt
such that when
cancels that across
.
drop across the mutual inductor
As is apparent from the equivalent circuit shown in Fig. 1(b),
this leaves only an in-phase resistive core loss component
in the
voltage, which therefore means
is in-phase
with
. Through integration of the product of the
and
waveforms over one cycle, the CUT core losses
can be accurately determined. From Fig. 1(b), it is also apparent that the self inductances and winding resistances of the
, as well as those of the mutual inductor,
CUT,
, and
have no effect on the phase relationship
and
if the voltmeter used to measure
between
has a sufficiently high impedance.
Based on Fig. 1(b), the pf of the measurement circuit can be
calculated using (1), and Fig. 1(c) can be obtained

(1)
For Fig. 1(c), the sinusoidal excitation frequency is 100 kHz,
is 5 H and assumed to be constant,
is varied from
0 to 10 H, and three different values of
are used to give
three ratios of
, where
is the mutual reactance of the CUT. It is apparent that the smaller
is, then the better is the pf between
and
. It is also
apparent that the greater is the ratio of
, the more
significant is the improvement to the pf in the uncompensated
case, for which
H, compared to that in the
fully compensated case when the pf is equal to one. This shows
that the potential benefit of the mutual inductance neutralization

0018-9464/$25.00 2008 IEEE

4128

IEEE TRANSACTIONS ON MAGNETICS, VOL. 44, NO. 11, NOVEMBER 2008

Fig. 2. (a) Excitation waveforms in an uncompensated circuit. (b) Excitation

waveforms in a compensated circuit.

Fig. 1. (a) Mutual inductance neutralization circuit, (b) its equivalent circuit,
and (c) variation of power factor with the difference between air and CUT mutual inductances for different R to X
ratios.

technique is greatest when low permeability, low loss cores are

tested, for which the excitation VA drawn is high relative to the
core loss.
The improvement to the pf that can be achieved is apparent
in the measured waveforms shown in Fig. 2 for which the CUT
is a MMG T25 15 10 toroidal in F49 material. A sinusoidal
excitation frequency of 50 kHz is used, and the ac flux density is
monitored and controlled to be 10 mT. Such a low flux density is
chosen in order to illustrate the effects of the proposed technique
under the core loss measurement extreme of exceptionally poor
pf conditions. In Fig. 2(a), the measured
and
waveforms using an uncompensated circuit are shown, and it is evident that the pf is very poor with a phase shift close to 90 . For
the same CUT, the
and
waveforms measured using
the mutual inductance neutralization circuit given in Fig. 1(a),

are shown in Fig. 2(b). It is clearly evident that the

and
waveforms are approximately in phase, and that the magnitude of the
waveform is dramatically reduced, indicating the reactive component of
has been largely cancelled. The third harmonic apparent in
is caused by the
nonlinear nature of the ferrite BH loop. This is because any third
harmonic component introduced by the nonlinear BH loop into
will, according to
, produce a harmonic
voltage in the mutual inductor output that, by the action of the
differential operator, is greatly exaggerated and is 90 out of
phase with the current harmonic. Therefore, when the output
voltage of the mutual inductor is summed with the secondary
voltage of the CUT to give
, the third harmonic will be
apparent, even if sinusoidal excitation is maintained across the
CUT. Harmonics do not affect the accuracy of the core loss measurement. This is because if the harmonic voltage components
at the output of the mutual inductor are exactly 90 out of phase
with the corresponding harmonics contained in
, then the
average of the product of the two is zero for each harmonic.
However, this is contingent on minimizing the impurity of the
mutual inductor, as explained in Section III.
III. MUTUAL INDUCTOR
Mutual inductance standards were extensively used in highaccuracy ac bridge measurement circuits during last century.

BAGULEY et al.: NEW TECHNIQUE FOR MEASURING FERRITE CORE LOSS UNDER DC BIAS CONDITIONS

4129

Fig. 3. Comparison of core loss measurements with no dc bias.

Fig. 4. Winding connection diagram of the core loss test circuit for dc bias
conditions.

Because of their importance as a measurement device, the errors associated with mutual inductance standards have been investigated, and the term impurity [15], [16] has been associated with a particular form of error. Impurity causes the phase
angle between the open circuit secondary winding emf, and the
primary excitation current waveforms of the mutual inductor to
deviate from quadrature. Such a deviation would add to, or subtract from, the core losses measured using the mutual inductance neutralization technique; therefore, it must be minimized
to ensure core losses are measured accurately. The causes of
impurity have been identified [15][19] as the self and interwinding capacities of mutual inductor windings (in conjunction with winding resistances and/or dielectric losses), as well
as the winding eddy current losses. Therefore, the mutual inductor used for the results presented in this paper is constructed
so as to minimize in particular the inter-winding capacitance.
To minimize inter-winding capacitance, the primary and secondary windings are placed end to end on a single layer. To further reduce inter-winding capacitance, as well as the associated
dielectric losses, the ends of the windings that are adjacent to
each other are both brought to the circuit ground [20]. To minimize primary winding eddy current losses, the primary is wound
using Litz wire. The secondary is wound using 0.25-mm diameter magnet wire, and the primary to secondary turns ratio is
2:21. The relatively low number of primary winding turns minimizes the voltage drop across the primary winding impedance,
so allowing most of the supply voltage to be used to excite the
CUT. The high number of secondary winding turns is necessary
to
in order for a sufficiently large reactive volt drop across
and improve the pf. The
be developed to cancel that across
primary and secondary windings are wound atop separate pieces
of mylar sheet which are slid along a glass former to adjust the
gap between the windings to allow the mutual inductance to be
varied to the optimal position. Such a variation is necessary in
order to allow the measurement technique to be used with CUTs
of different permeabilities and under different temperature and
excitation conditions.
The benefit of using the mutual inductor is shown in the results presented in Fig. 3, which gives the measured core losses as
of an MMG T25 15 10
a function of the ac flux density
toroidal in F49 material that is cut into two halves which are
taped together with no spacer between the halves. The core loss
measurements are made in an ambient temperature of 24 C, at
a sinusoidal excitation frequency of 50 kHz, and the
levels

are monitored and controlled. Three sets of measurements are

made: first using a ClarkeHess 2335 wattmeter [21], second
using an Agilent DSO6034 oscilloscope [22] as a wattmeter,
and third using the same oscilloscope and the mutual inductance neutralization technique. The readings taken using the
ClarkeHess 2335 wattmeter are expected to be the most accurate, as this instrument is optimized for accurate high-frequency power measurement at any pf through the high degree
of phase/time matching between the current and voltage sensing
channels. By comparison, the core loss data measured using the
oscilloscope is strongly subject to errors resulting from poor
phase/time matching between the current and voltage sensing
channels. The poor matching is due primarily to the parasitic
impedances introduced into the current sensing channel by the
0.22- thick film resistive current shunt that is used, and the
1-m-long probe leads connecting it to the oscilloscope, as well
as digitization errors. It is apparent that large differences exist
between the wattmeter and the oscilloscope core loss measurements. These differences occur because the phase angle in the
core loss measurement circuit is high, which means very small
phase/time matching errors cause large core loss measurement
errors [13]; the phase angle as measured by the wattmeter varies
from 87.97 at 10 mT, to 81.2 at 80 mT. It is also apparent
that, through the use of the mutual inductance neutralization
technique, measurements made using the oscilloscope are dramatically improved in accuracy. By implication, this shows that
the impurity of the mutual inductor does not significantly affect
measurement accuracy.
IV. CORE LOSSES MEASURED UNDER DC BIAS
To measure core losses in the presence of a dc bias, the measurement circuit is modified to that shown in Fig. 4. In this cirand
are measured and multiplied together to
cuit
calculate the total losses of two gapped ferrite CUTs at the same
time. Two CUTs are used in order to allow dc bias windings to
be placed on each with polarities such that the dc bias windings
are effectively decoupled from the ac windings in the circuit.
Decoupling is necessary in order for the dc bias to be exact [6],
[23]. The primary windings of the CUTs are connected in parallel to ensure that each core is run at the same ac flux density.
The sense windings of the CUTs are connected in series rather

4130

IEEE TRANSACTIONS ON MAGNETICS, VOL. 44, NO. 11, NOVEMBER 2008

Fig. 5. Core loss measurements made under dc bias conditions.

than parallel to avoid the generation of circulating currents in

these windings that would generate an error in the core loss measurement.
Fig. 5 shows core loss results measured as a function of con, and dc magnetic field strengths,
, using the cirtrolled
cuit given in Fig. 4. The measurements are made at a sinusoidal
excitation frequency of 100 kHz in an ambient temperature of
24 C, and the CUTs are gapped MMG T25 15 10 toroidals
in F49 material. The total core gap used is 0.152 mm, which is
found to be sufficient to reduce the tolerance on the relative permeability between the two gapped CUTs to within 3%. Such
a gap is necessary to ensure the CUTs operate at same dc flux
. A nonlinear
density, within a tolerance, for a given level of
is clearly apincrease in the core loss characteristic with
parent.

V. CONCLUSION
The mutual inductance cancellation of inductive sense voltages can be used to greatly enhance the accuracy of core loss
measurements. This technique will be of greater value in measuring the losses of low permeability/distributed gap cores such
as powdered iron and MPP. Further research into the quantification and reduction of the mutual inductance phase angle impurity at high frequencies will be required for these more demanding applications.
ACKNOWLEDGMENT
This work was supported by the University of Auckland
under UARC Grant 9273-3607841 and MMG, Ltd., (U.K.)
for supplying the cores used for the test results. The concept
of using mutual inductance to neutralize inductive voltages in
powdered iron cores was developed by Carsten in (unpublished)
work for Micrometals.

REFERENCES
[1] California standards for external power supplies, California Energy
Commission, Mar. 3, 2005 [Online]. Available: http://www.energy.ca.
gov.
[2] A. N. Alderman, PSMA position on voluntary and mandatory standards that will regulate the minimum acceptable efficiency of power
supplies, Jan. 14, 2005 [Online]. Available: http://www.psma.com.
[3] A. Fanara, Energy star for external power supplies, in Proc. PPDC,
San Jose, CA, Apr. 1820, 2005.
[4] C. Calwell, A. Mansoor, and R. Keefe, Active mode power supply
efficiency: Key issues, measured data and the design competition opportunity, in Proc. IEEE APEC04, 2004, vol. 1, pp. 323328.
[5] P. Bertoldi, European code of conduct to improve energy efficiency of
power supply: The first policy action around the worldwide addressing
external power supplies, in Proc. IEEE APEC04, 2004, vol. 1, pp.
329331.
[6] C. A. Baguley, B. Carsten, and U. K. Madawala, An investigation into
the impact of DC bias conditions on ferrite core losses, IEEE Trans.
Magn., vol. 44, no. 2, pp. 246252, Feb. 2008.
[7] A. Brockmeyer, Experimental evaluation of the influence of DC-premagnetization on the properties of power electronic ferrites, in Proc.
IEEE APEC96, Mar. 37, 1996, vol. 1, pp. 454460.
[8] A. Brockmeyer and J. Paulus-Neues, Frequency dependence of the
ferrite-loss increase caused by premagnetization, in Proc. IEEE
APEC97, Feb. 2327, 1997, vol. 1, pp. 375380.
[9] J. Reinert, A. Brockmeyer, and R. W. A. A. De Doncker, Calculation of losses in ferro- and ferrimagnetic materials based on the modified steinmetz equation, IEEE Trans. Ind. Applicat., vol. 37, no. 4, pp.
10551061, Jul.Aug. 2001.
[10] P. M. Gradzki and F. C. Lee, Domain wall resonance and its effect
on losses in ferrites, in Proc. IEEE PESC91, Jun. 2427, 1991, pp.
627632.
[11] H. Y. Chung, F. N. K. Poon, C. P. Liu, and M. H. Pong, Analysis
of buck-boost converter inductor loss using a simple online B-H curve
tracer, in Proc. IEEE APEC00, Feb. 610, 2000, vol. 2, pp. 640646.
[12] V. C. Valchev, A. P. Van den Bossche, and D. M. Van de Sype, Ferrite
losses of cores with square wave voltage and DC bias, in Proc. IEEE
IECON05, Nov. 610, 2005, pp. 837841.
[13] J. Thottuvelil, T. G. Wilson, and H. A. Owen, High frequency measurement techniques for magnetic cores, IEEE Trans. Power Electron.,
vol. 5, no. 1, pp. 4153, Jan. 1990.
[14] B. Carsten, Why a magnetics designer should measure core loss; with
a survey of loss measurement techniques and a low cost, high accuracy
alternative, in Proc. HFPC, San Jose, CA, May 1995, pp. 103119.
[15] S. Butterworth, Capacity and eddy current effects in inductometers,
in Proc. Phys. Soc. London, 1921, vol. 33, pp. 312354.
[16] L. Hartshorn, The properties of mutual inductance standards at telephonic frequencies, in Proc. Phys. Soc. London, 1925, vol. 38, pp.
302320.
[17] F. A. Laws, Electrical Measurements. New York, London: McGrawHill, 1938.
[18] D. Karo, Electrical Measurements Part II. London, U.K.: MacDonald, 1953.
[19] Z. Fei, A new construction of the mutual inductor used as quadrature
standard, IEEE Trans. Instrum. Meas., vol. IM-29, no. 4, Dec. 1980.
[20] W. A. Edson, The single-layer solenoid as an RF transformer, in
Proc. IRE, Aug. 1955, pp. 932936.
[21] Clarke and Hess Communications Research Corp., [Online]. Available:
www.clarke-hess.com.
[22] Agilent Technologies, [Online]. Available: www.agilent.com
[23] J. Zhang, G. Skutt, and F. C. Lee, Some practical issues related to core
loss measurement using impedance analyser approach, in Proc. IEEE
APEC95, Mar. 59, 1995, vol. 2, pp. 547553.

Manuscript received March 03, 2008. Current version published December

17, 2008. Corresponding author: C. A. Baguley (e-mail: cbag002@ec.auckland.ac.nz).