You are on page 1of 8

4374 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 29, NO.

8, AUGUST 2014

New Core Loss Measurement Method for


High-Frequency Magnetic Materials
Mingkai Mu, Qiang Li, Member, IEEE, David Joel Gilham, Fred C. Lee, Fellow, IEEE,
and Khai D. T. Ngo, Senior Member, IEEE

Abstract—Magnetic core loss is an important concern for power


converters. As the switching frequency increases and converter
size reduced, the core loss will have significant impact to the con-
verter efficiency and temperature. Accurate evaluation is impor-
tant for magnetic design and converter loss estimation. The classic
two-winding method is limited to low frequencies (usually below
1 MHz) because it is sensitive to phase discrepancy. In this paper,
a new method is proposed for high-frequency core loss measure-
ment that utilizes capacitive cancellation, which is suitable for HF
and VHF core loss measurement. The new method greatly reduces
the sensitivity to phase discrepancy, which is the dominating error
source in the conventional two-winding method. An experimental
demonstration is performed at 10 MHz, and the possible errors are
analyzed in detail. With the proposed method, the high-frequency
core loss can be accurately measured.
Index Terms—Core loss measurement, ferrite, high frequency,
magnetic integration, magnetic loss.

I. INTRODUCTION
ORE loss is always an important concern for power elec- Fig. 1. Two-winding measurement method and equivalent model. (a) Two-
C tronics applications. The converter is less efficient without
proper magnetic cores. With the advance of semiconductors like
winding method. (b) Equivalent model.

gallium nitride devices, the power converters have been pushed


to higher and higher switching frequencies for the ease of mag- permeability material test (or gapped core), the winding loss can
netic integration. As the magnetic components get smaller, the be a significant part of the total loss. Because the presence of the
core loss will become a limiting factor to converter efficiency core will change the eddy current in the winding, the winding
and power density. An accurate measurement of the core loss loss is difficult to test, estimate or compensate. In addition, if
is important to magnetic and thermal design, as well as the ex- the sample does not generate enough heat, the measurement
ploration of new high-frequency materials. There are several accuracy can be questionable.
existing methods to test the core loss [1]–[7]. The widely used two-winding method [3]–[5] is shown in
One method is the thermal approach [1], [2]. The basic idea Fig. 1(a). The core-under-test (CUT) carries two windings: the
is to put the wound core in the thermally isolated chamber and excitation winding and the sensing winding. The current flowing
measure the temperature difference between the inlet and outlet through the excitation winding and the voltage on the sensing
coolant. With the temperature difference and coolant flow rate, winding are measured by the probes. Integrating the product
the heat generated by the excited inductor can be calculated. This of the sensing winding voltage and excitation winding current
method is universal, but it is time consuming. Also, it needs extra waveforms will give the loss consumed by the CUT. This method
effort to exclude the winding loss. For high-frequency and low does not have significant drawbacks in principle, and it excludes
the ac winding loss from the measured core loss, which is its
biggest advantage. However, in practice it is sensitive to phase
discrepancy when the core’s quality factor is high. Usually the
Manuscript received July 22, 2013; revised September 18, 2013; accepted phase discrepancy comes from the current-sensing resistor’s par-
October 3, 2013. Date of current version March 26, 2014. This work was
supported by the Power Management Consortium (PMC) of the Center for Power asitic inductance, mismatch between probes and oscilloscope’s
Electronics Systems (CPES) at Virginia Tech. Recommended for publication sampling resolution limit. Thus it is not accurate when testing
by Associate Editor C. R. Sullivan. high-frequency loss, especially when the material is low perme-
The authors are with the Center for Power Electronic Systems, Virginia
Polytechnic and State University (Virginia Tech), Blacksburg, VA 24060 USA ability and low loss.
(e-mail: mmk@vt.edu; lqvt@vt.edu; dgilham@vt.edu; fclee@vt.edu; kdtn@vt. Another method for high-frequency measurement connects a
edu capacitor in series or in parallel with the inductor of the core
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. under test. [6]–[8] In [6], the series capacitor is finely tuned to
Digital Object Identifier 10.1109/TPEL.2013.2286830 resonate with the inductor at the test frequency. Because the
0885-8993 © 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
Authorized licensed use limited to: J.R.D. Tata
See Memorial Library Indian Institute of standards/publications/rights/index.html
http://www.ieee.org/publications Science Bengaluru. Downloaded on July 09,2023
for moreatinformation.
11:57:05 UTC from IEEE Xplore. Restrictions apply.
MU et al.: NEW CORE LOSS MEASUREMENT METHOD FOR HIGH-FREQUENCY MAGNETIC MATERIALS 4375

voltages on the inductor and capacitor cancel each other, the


quality factor or resistance in this RLC network can be easily
tested by the ratio of the amplitudes of the measured voltages
on capacitor and the resonant RLC branch. After compensat-
ing the parasitic resistances, such as the winding resistance and
the resonant capacitor equivalent series resistor (ESR), one can
calculate the core loss. This method is not sensitive to phase
discrepancy because it only uses the voltage amplitude ratio
to calculate loss. However, it is difficult to compensate the ac
winding losses, which cannot be estimated very accurately at
high frequencies with the presence of cores. The resonant ca-
pacitance’s value is critical; it should be the exact value that
resonates with the inductor at the given frequency. What is
more, when the nonlinearity of the core sample is strong, the Fig. 2. Loss error induced by phase discrepancy at different phase angles
between voltage and current.
waveform distortion will invalidate this method.
There are other methods like the impedance analyzer based
method [9] and H-bridge method [10]. But they have similar
problems as mentioned earlier. To summarize, although there
are methods that can be effective, they all have one or more
significant drawbacks:
1) need extra efforts to separate the winding loss from the
core loss, careful compensation is required for such situ-
ations, especially for high frequency;
2) sensitive to phase discrepancy when the core’s quality
factor is high;
3) time consuming.

II. PROPOSED METHOD


A. Sensitivity to Phase Discrepancy
Before introducing the proposed method, it is necessary to
discuss the sensitivity to phase discrepancy of the two-winding
method in Fig. 1(a). Its equivalent model is in Fig. 1(b). This
method uses the integration between sensing winding voltage
v2 and current sensor voltage vR to get the core loss, as shown
in (1) Fig. 3. Proposed measurement circuit and equivalent model. (a) Improved
 method. (b) Equivalent model.
1
Pcore = v2 vR dt. (1)
T Rref T
the current sensing resistor, mismatches of two probes, and the
In (1), Rref is the current sensing resistance, and T is the
oscilloscope sample rate limitation, begin to emerge. For exam-
period of the excitation frequency. Because the impedance of
ple, a 2 Ω sensing resistor with 1 nH will produce 0.9◦ phase
magnetizing inductor Lm is usually much larger than the equiv-
discrepancy at 5 MHz. For a 5 GS/s digital oscilloscope, the
alent core loss resistance Rcore , the phase difference between v2
minimum sampling period is 200 ps, which is 0.36◦ at 5 MHz.
and vR is very close to 90◦ . This means a small phase discrep-
Fig. 2 shows the percentage of error that 0.1◦ , 1◦ , and 10◦ phase
ancy will lead to a significant loss error after the integration.
discrepancy will cause at different phase angles between v2 and
Assuming the excitation is sinusoidal, we can calculate the sen-
iR . We can see that at near 90◦ , it will produce over 100% error
sitivity to phase discrepancy as [8]
for 1◦ phase discrepancy. This type of error is more severe if the
Δ = tan(ϕv −i ) × Δϕ (2) measuring frequency is high, and the target material is high-Q.
A high frequency will increase the phase discrepancy, a low loss
where ϕv −i is the phase angle between the two terms for in- will make the first term of (2) very large.
tegration: voltage and current. When the voltage and current
have nearly 90◦ phase difference, the small phase discrepancy
B. Proposed Method
Δϕ will be amplified by tan(ϕv −i ), and leads to a significant
loss error percentage Δ. For low-frequency measurements, the To alleviate the sensitivity of the two-winding method, a new
parasitic effect in the test circuit is small, so the phase dis- method for sinusoidal excitation is proposed. The improved
crepancy is small and won’t cause a significant error. As the method is shown in Fig. 3(a). The equivalent circuit model is
frequency increases, the imperfections, such as the parasitic of shown in Fig. 3(b). The core loss is represented as a resistor in

Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on July 09,2023 at 11:57:05 UTC from IEEE Xplore. Restrictions apply.
4376 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 29, NO. 8, AUGUST 2014

series with the ideal magnetizing inductor. The CUT still carries where Ae is the equivalent core cross-section area. If the core
an excitation winding and a sensing winding. One terminal of the under test is a toroid core with a large ID/OD ratio, we can
sensing winding is connected to the same-polarity terminal of assume the flux density in the core is almost uniform. Dividing
the excitation winding. A resonant capacitor is added in series the core loss by the equivalent volume Ve , we can obtain the
with the excitation winding at this connected terminal. Now core loss density at the specific flux density. Hence, the core
three voltages should be monitored: sensing winding voltage loss density is
v2 , the current sensor’s voltage vR , and the new voltage v3 .
 T
Voltage v3 is actually the total voltage on the sensing winding N1
and the resonant capacitor. When the turns ratio is 1:1, it is Pv = v3 vR dt. (7)
N2 T Rref Ve 0
simply the total voltage on Rcore , Lm , and Cr .
The purpose of adding the series resonant capacitor Cr is to Although in principle any turns ratio could be used, we recom-
cancel the measured voltage on the magnetizing inductor Lm . If mend a turns ratio of 1:1 for simplicity and to facilitate the use
the transformer turns ratio is 1:1, and Cr resonates with Lm at of bifilar windings to minimize leakage inductance. The bifilar
the input frequency, their voltages are the same in magnitude but windings are recommended due to stronger coupling between
180◦ out of phase. As a result, v3 is equal to the voltage on the excitation winding and sensing winding.
equivalent core loss resistance Rcore , and it is in phase with the For the current-sensing resistor, to reduce the phase discrep-
current through the excitation winding iR , as well as its voltage ancy caused by its ESL, we can parallel multiple larger value
vR . We know that the integration of v3 ×vR will produce the loss resistors [3]. However, too many resistors in parallel could in-
consumed on Rcore , because Lm and Cr do not consume real duce additional parasitic inductor from the layout. To simplify
power. As v3 and vR are in phase, the integration of their product the layout, we chose to use only a few current-sensing resistors
is not sensitive to phase discrepancy, which is proved in Fig. 2. in parallel. All the probes used in this test circuit are the same
Another major advantage of this method is that it automatically model so that their time delays are identical. To further verify
excludes the winding loss of the excitation winding from the that the probes match well, we can use them to test a square
measurement. In [7], although it uses a capacitor in parallel wave simultaneously, and check if their rising edges are close to
with the transformer excitation winding to correct the phase each other in the oscilloscope. In addition, v3 and vR ’s probes
angle between the measured voltage and current, the measured should be exchanged during the test. If the two groups of data
loss is not only the loss on the core. With careful analysis, one are close enough, we can assume that their time delay difference
can find out that the measured loss will include the excitation won’t produce significant error. High quality factor is preferred
winding’s loss. for the resonant capacitor. The loss on the resonant capacitor is
Though Cr is used to cancel the voltage on the inductor, it included in the measurement, so high-Q capacitors like silver
does not have to completely cancel Lm ’s voltage. Fig. 2 shows mica or RF porcelain capacitors are recommended, and their
that when the phase angle is 30◦ , the power error is only 1% loss should be subtracted from the test result. During the test,
for 1◦ phase discrepancy and 10% for 10◦ phase discrepancy, the magnetizing inductance tends to change due to environmen-
which is sufficiently small. That means the resonant capacitor tal changes such as temperature and ac flux amplitude, so the
Cr has a tolerance range, and it does not have to resonate with resonant capacitor may be tuned a little bit to keep the phase
Lm at the input frequency. It only needs to keep the phase angle angle between v3 and vR close to 0◦ . Sometimes, when the core
between v3 and vR far from 90◦ . When the turns ratio is N1 :N2 , loss is fairly small or the core is tested under biased conditions,
the Cr value should be close to the voltage waveform of v3 may not be a good sinusoidal shape.
As a result, it is difficult to judge if v3 is in phase or not with
N1
Cr = . (3) vR by observation. A recommended method is to add a small
N2 (2πf )2 Lm delay perturbation (e.g., 1◦ ) on the channel of vR or v3 with
If the turns ratio is 1:1, Cr is simply the resonant capacitor the deskew function of the scope, and compare the difference
for Lm at the measuring frequency. between the loss measurement result before and after the de-
The phase angle between vR and v3 is calculated as follows: lay. If the difference is sufficiently small, we can consider that
  the waveforms are close to in-phase condition and the result is
−1 ωLm − (N1 /N2 ωCr ) accurate.
φv 3 −i R = tan . (4)
Rcore
By selecting the correct resonant capacitor Cr , we can move C. Experiment Example
the phase angle between v3 and vR far from 90◦ . To demonstrate the method, the core loss density of the com-
The formulas for the flux density and core loss calculation mercial high-frequency power magnetic material 4F1 (NiZn
are ferriteR
) from Ferroxcube is tested. The core loss curve from
 T its datasheet is also displayed. The authors do not have enough
1
B= v2 dt (5) information about the test process behind the datasheet value,
N2 Ae 0
so it is displayed just as a comparison here.
 T
N1 The test setup is shown in Fig. 4. Fig. 4(a) shows a 4F1 toroid
Pcore = v3 vR dt (6) sample wound as a transformer. The core is customized from
N2 T Rref 0

Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on July 09,2023 at 11:57:05 UTC from IEEE Xplore. Restrictions apply.
MU et al.: NEW CORE LOSS MEASUREMENT METHOD FOR HIGH-FREQUENCY MAGNETIC MATERIALS 4377

Fig. 5. Working waveform in the improved method (10 MHz).

Fig. 4. Improved core loss measurement setup. (a) Core under test (4F1 NiZn
ferrite). (b) Test board layout. (c) Test setup. Fig. 6. Result of conventional and proposed methods compared with datasheet.

bulk and the dimensions are: OD = 25 mm, ID = 21 mm,


and h = 3 mm. The 12-turn bifilar winding is made of 105
strand AWG 48 Litz wire. The core under test is suspended
underneath the test board and immersed in the oil bath on a
hot plate, and its temperature is supervised by the thermocou-
ple and controlled to be 100 ◦ C. The current sensing resistor
is a 4.7 Ω SMD thick film resistor. Its resistance at 10 MHz is
4.55 Ω, tested with impedance analyzer (Agilent R
4294 A). The
transformer’s excitation winding is excited by the voltage from a
power amplifier (Amplifier Research R
25A250 A) driven by the
Fig. 7. Probe used to test the sensing resistor’s parasitic inductance.
10 MHz sinusoidal signal from the function generator. Three dif-
ferential probes (Tektronix R
P6251) measure the voltage v2 , v3 ,
and vR , respectively. The waveform on the digital oscilloscope A. Phase Discrepancy
(Tektronix TDS R
7054) is shown in Fig. 5. We can see that v3
◦ The ESL of the current sensor not only depends on the re-
is 4 leading vR , which means their phase angle difference is
sistor, but also on the test board layout. The sensed voltage
close to 0◦ . So an integration of v3 and vR will have less phase-
will be interfered by the coupling between the sensing loop and
induced error. The measured core loss is compared with the
the excitation loop. The effective ESL of the current sensor is
datasheet result in Fig. 6. The results of measurement with the
tested using the following procedure. Firstly, the parasitic induc-
conventional two-winding method using the same equipment
tance of the sensing resistor is tested using the probe (Agilent R
are also displayed; these results are several times higher.
42941 A) as shown in Fig. 7. Then, replace the resistor with an
identical sized copper foil, and test its inductance with the same
III. ERROR ANALYSIS probe (the calibration setting and probe tip position should be the
The error sources in the earlier measurement example in- same as the first step). There could be some inductance differ-
clude phase discrepancy, resonant capacitor ESR, parasitic ca- ence because the resistor’s current routing may be not as simple
pacitances, eddy current loss, and etc. as a straight conductor. After that, inject a known-amplitude

Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on July 09,2023 at 11:57:05 UTC from IEEE Xplore. Restrictions apply.
4378 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 29, NO. 8, AUGUST 2014

Fig. 9. Simplified circuit of parallel parasitic capacitance.

C. Error Caused by Parasitic Capacitances


The parasitic capacitors in the test circuit will introduce extra
error. For example, the probe’s impedance is not infinite and the
transformer has intrawinding capacitances. So there will be a
small current flowing through the sensing winding, which will
Fig. 8. Resonant capacitor ESR compared with R c o re . invalidate the assumption that the sensing winding is open when
the parasitic capacitance is big. Though v3 and vR ’s probes also
have input capacitances, their loading effect is much smaller
10 MHz ac current into the copper foil, and test the sensing than v2 ’s probe. This is because the magnitude of v3 and vR
voltage. This voltage is the inductive voltage inducted by the are much smaller than v2 , so the currents flowing through their
parasitic inductance and coupled flux from the excitation cir- probes’ input capacitors are much smaller than v2 ’s probe. Thus,
cuit. Compensating the inductance difference between resistor when we analyze the loading effect of v2 ’s probe, we can safely
and copper foil, we can get an estimate for the effective ESL assume the other two probes are open.
of the current sensor. Following this procedure, we found out The simplified equivalent circuit is shown in Fig. 9. Cp is
that there is about no inductance difference between the 4.7 Ω the total parasitic capacitance in parallel, including the probe’s
resistor and copper foil, and the effective ESL of the current input capacitor and the intrawinding capacitances from both
sensor is about 0.97 nH. As stated earlier, the sensitivity to windings. Due to this capacitor, a current will flow through the
phase discrepancy is greatly reduced if the v3 –vR phase angle sensing winding and Cp . This current will then induce voltage
is within 30◦ . If we assume current sensing resistor ESL, probe drop on the sensing winding’s leakage inductor L2 and winding
mismatch and scope resolution limit contribute 1.6◦ , 0.7◦ , and resistor R2 , which means v2 will be different from the voltage on
0.7◦ phase discrepancy respectively, the total phase discrepancy Lm and Rcore , which is vm . Furthermore, the current measured
is 3◦ . Even though the actual phase discrepancy could be higher by current sensor iR is not equal to the current flowing through
or lower than this estimation, we can still feel safe to have <1% magnetizing inductor im . When the measurement frequency is
loss error according to Fig. 2. Thus, this part of the error is not high, these differences cannot be ignored.
a significant contributor to overall error. To quantify the error caused by the parallel parasitic capaci-
tance, we assume
Rcore  ωLm (8)
B. Error Caused by ESR of the Resonant Capacitor 1
ωLm  (9)
The capacitor will introduce ESR into the measurement. To ωCp
estimate and compensate this error, we can calculate the equiv- 1
alent core loss resistance Rcore using the tested loss and current R1 , R2 , ωL1 , ωL2  . (10)
ωCp
amplitude, as shown in Fig. 8. We can see that Rcore varies from
2.7 to 10 Ω depending on the flux density. Under these assumptions, the error percentage can be approx-
In the measurement, a 120 pF silver mica capacitor is imated as

used to resonate with the magnetizing inductor. The ESR is Pm easured − Pactual (v3 iR − vm im )dt
Δ= = 
about 160 mΩ, measured by the impedance analyzer (Agilent R
Pactual vm im dt
4294 A). Multiple tests are performed and the oscillatory voltage
R2 ω 2 Lm Cp
level in the impedance analyzer is changed from 10 mV to 1 V, ≈ + ω 2 L2 Cp . (11)
and the measured ESR does not vary much from 160 mΩ (the Rcore
reading varies from 152 to 185 mΩ and 160 mΩ is the averaged The derivation of (11) is complicated, so the final approxi-
value). Since the rated voltage (600 V) is much higher than the mated equation is given directly here. Simulation proves that
voltage level (<40 V) in the test circuit (see Fig. 5), we assume this approximate equation can fairly accurately estimate the er-
the capacitor ESR is constant 160 mΩ during the test. Under ror caused by Cp if the assumptions in (8), (9), and (10) are
this assumption, the capacitor ESR is compared with Rcore in true.
Fig. 8. The loss error caused by ESR ranges from 1.6% to 5.9%, For the measurement setup in Section II-C, Cp ≈ 10 pF, Lm ≈
and this part of error can be compensated accordingly. 2.2 μH, L2 ≈ 100 nH, R2 ≈ 0.233 Ω, and ω = 2π × 10 MHz.

Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on July 09,2023 at 11:57:05 UTC from IEEE Xplore. Restrictions apply.
MU et al.: NEW CORE LOSS MEASUREMENT METHOD FOR HIGH-FREQUENCY MAGNETIC MATERIALS 4379

Fig. 12. Simplified circuit model with transformer interwinding capacitor.

Fig. 10. Error caused by the parallel parasitic capacitances. the current-sensing resistor. This common ground is required if
single-ended probes are used.
Another parasitic capacitor is the transformer interwinding
capacitance. Fig. 12 shows a simplified transformer model with
interwinding capacitor Cw w . If there is a voltage difference
between the two windings, there will be a small current flowing
through Cw w and the sensing winding.
Assume that
Rcore  ωLm (12)
1
Fig. 11. Improved version of proposed method. ωLm  (13)
ωCw w
1
With the value of Rcore in Fig. 8, we can calculate the error R1 , R2 , ωL1 , ωL2  . (14)
ωCw w
percentage for the example in Section II-C, as shown in Fig. 10.
From (11), we know that this error increases very fast as the The error caused by the interwinding capacitance can be sim-
frequency goes up. To reduce this error, the sensing winding’s plified as shown (15), at the bottom of the page.
resistance R2 and leakage inductance L2 should be small, the If the windings are wound in the bifilar way, N1 = N2 , the
ratio of Lm to Rcore should be small (which means the magne- equation is simplified to
tizing inductor’s quality factor is low, but this is predetermined R1 ωL2 + R2 ωL1
by the core material), and the parasitic capacitance Cp need to Δ≈− ωCw w . (16)
Rcore
be small. Thus the intrawinding capacitance of the transformer
For the measurement in the example, R1 ≈ R2 ≈ 0.233 Ω,
should be well-controlled, and a low-input-capacitance probe is
L1 ≈ L2 ≈ 100 nH, Cw w ≈ 20 pF, ω = 2π × 10 MHz, the turns
preferred. In addition, to reduce the loading effect of the probe,
ratio is 1:1, the equivalent core loss resistance is as in Fig. 8, and
an improved version of the proposed method is shown in Fig. 11.
the measurement error is plotted in Fig. 13. In summary, the error
This improved circuit does not measure the voltage on the
caused by the parallel capacitors and interwinding capacitor is
sensing winding. Instead, it tests the voltage on the resonant
generally below 1%.
capacitor v2 . By measuring this way, we can reduce the current
through the sensing winding, and merge Cprob e into the resonant
D. Eddy Current Loss Due to Mutual Resistance
capacitance Cr . The voltage on the sensing winding can be
calculated by simply subtracting v2 from v3 . Another benefit of The transformer configuration allows this method to exclude
this improved version is that the three probes can find a common the ac winding loss on the excitation winding. However, a small
ground at the connection between the resonant capacitor and portion of eddy current loss can be sensed by the sensing


Pm easured − Pactual (v3 iR − vm im )dt
Δ= = 
Pactual vm im dt

−N1 R1 ωL2 − N1 R2 ωL1 − N12 R2 ωLm − N12 Rcore ωL2 + N1 N2 R1 ωLm + N1 N2 R2 ωLm
2 2

N1 N2 Rcore

−N1 N2 Rcore ωL1 + N1 N2 Rcore ωL2 − N22 R1 ωLm + N22 Rcore ωL1
+ ωCw w (15)
N1 N2 Rcore

Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on July 09,2023 at 11:57:05 UTC from IEEE Xplore. Restrictions apply.
4380 IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 29, NO. 8, AUGUST 2014

one would prefer certain distance between the two windings and
to keep the core sample away from other conductive parts.

E. Other Error Sources


The interconnect resistance between the sensing winding and
the resonant capacitor could also add some error. To reduce
this part of loss, the solder points of the sensing winding and
the resonant capacitor should be as close as possible. On the
testing board shown in Fig. 4(b), the interconnect resistance is
measured to be around 1.3 mΩ at 10 MHz. So this part of error
is negligible.
There are some other error sources, such as the oscilloscope’s
magnitude error, temperature variation, etc. Since these errors
Fig. 13. Error caused by parasitic interwinding capacitance of the transformer.
are small and do not increase much with frequency, the error
from these sources is under control.

IV. CONCLUSION
This paper proposes a new method of core-loss measurement
for high-frequency magnetic materials. The new method offers
a significant improvement to the previous methods at high fre-
quencies. It excludes the ac winding loss from the measured re-
sult, and reduces the error induced by phase discrepancy, which
is the major error source in the classic two-winding method at
high frequency. The main concept is the addition of a capacitor
to cancel the reactive voltage on the magnetizing inductor of the
transformer, which in turn causes the voltage and current to be
roughly in phase. In other words, it makes the calculated total
power have less reactive power. This will make the measurement
Fig. 14. Simulation model with AWG 20 wires in ANSYS Maxwell. less sensitive to phase error, and thus increase the accuracy.
A 4F1 core sample is used to demonstrate the proposed
winding. This impact can be interpreted as the mutual resis- method. The possible error sources are analyzed in detail and
tance [11]. Because the existence of mutual resistance, the mea- the estimated total error is below 8%. Though only a 10 MHz
sured voltage v3 will include a resistive voltage vm r example is discussed in this paper, this method can measure the
core loss at much higher frequencies.
vm r = Rm r iR . (17)
As a result, this portion of voltage will produce a loss when ACKNOWLEDGMENT
multiplying the current on excitation winding’s current. Its in- The author would like to thank Charles Sullivan for helpful
duced loss density error is expressed as comments.

1 vR
Pm r = Rm r iR dt. (18) REFERENCES
T Ve T Rref
The value of the winding mutual resistance of a cored trans- [1] C. Xiao, G. Chen, and W. G. H. Odendaal, “Overview of power loss
measurement techniques in power electronics systems,” IEEE Trans. Ind.
former is difficult to measure and separate experimentally. Appl., vol. 43, no. 3, pp. 657–664, May/Jun. 2007.
Thus simulation is used to estimate this part of the error. The [2] R. Linkous, A. W. Kelley, and K. C. Armstrong, “An improved calorimeter
ANSYS R
Maxwell and SFD method introduced in [12] are for measuring the core loss of magnetic materials,” in Proc. Appl. Power
Electron. Conf., Feb. 6–10, 2000, vol. 2, pp. 633–639.
used to calculate the mutual resistance of the sample shown [3] V. J. Thottuvelil, T. G. Wilson, and H. A. Owen Jr., “High-frequency mea-
in Fig. 4(a). The calculated mutual resistance of the sample is surement techniques for magnetic cores,” IEEE Trans. Power Electron.,
2.1 mΩ, which is below 1% of the self-resistance and negligible vol. 5, no. 1, pp. 41–53, Jan. 1990.
[4] V. Cekov Valchev and A. Van denBossche, Inductors and Transformers
compared with the core loss resistor. Its induced error is well for Power Electronics. Boca Raton, FL: CRC Press, 2005, ch. 11.
below 1%. Even when the wires are changed to AWG 20 and [5] A. Brockmeyer, “Experimental evaluation of the influence of DC premag-
closely placed as shown in Fig. 14, the simulated mutual resis- netization on the properties of power electronic ferrites,” in Proc. Appl.
Power Electron. Conf., Mar. 3–7, 1996, vol. 1, pp. 454–460.
tance is 1.12 mΩ, still much lower than the core loss resistor. [6] Y. Han, G. Cheung, A. Li, C. R. Sullivan, and D. J. Perreault, “Evaluation
This analysis indicates that the mutual resistance caused by the of magnetic materials for very high frequency power applications,” IEEE
eddy current is not a significant error in the proposed method, Trans. Power Electron., vol. 27, no. 1, pp. 425–435, Jan. 2012.
[7] F. Dong Tan, J. L. Vollin, and S. M. Cuk, “A practical approach for mag-
and using solid wire is as fine as the Litz wire for this test method. netic core-loss characterization,” IEEE Trans. Power Electron., vol. 10,
Even though it is not significant, to keep it as small as possible, no. 2, pp. 124–130, Mar. 1995.

Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on July 09,2023 at 11:57:05 UTC from IEEE Xplore. Restrictions apply.
MU et al.: NEW CORE LOSS MEASUREMENT METHOD FOR HIGH-FREQUENCY MAGNETIC MATERIALS 4381

[8] J. Zhang, G. Skutt, and F. C. Lee, “Some practical issues related to core Fred C. Lee (S’72–M’74–SM’87–F’90) received the
loss measurement using impedance analyzer approach,” in Proc. Appl. B.S. degree in electrical engineering from the Na-
Power Electron. Conf., Mar. 5–9, 1995, vol. 2, no. 2, pp. 547–553. tional Cheng Kung University, Tainan City, Taiwan,
[9] P. M. Gradzki, “Core loss characterization and design optimization of in 1968, and the M.S. and Ph.D. degrees in electri-
high-frequency power ferrite devices in power electronics applications,” cal engineering from Duke University, Durham, NC,
Virginia Tech Dissertation, 1992. USA, in 1972 and 1974, respectively.
[10] B. Carsten, “Why a magnetics designer should measure core loss; with He is a University Distinguished Professor at Vir-
a survey of loss measurement techniques and a low cost, high accuracy ginia Polytechnic Institute and State University (Vir-
alternative,” in Proc. High Freq. Power Convers., San Jose, CA, May ginia Tech) in Blacksburg, VA, and the Founder and
1995, pp. 103–119. Director of the Center for Power Electronics Systems,
[11] J. H. Spreen, “Electrical terminal representation of conductor loss in trans- an engineering research center consisting of 80 cor-
formers,” IEEE Trans. Power Electron., vol. 5, no. 4, pp. 424–429, Oct. porations. The mission of the center is “to provide leadership through global
1990. collaboration to create electric power processing systems of the highest value
[12] C. R. Sullivan, “Computationally efficient winding loss calculation with to society.” Dr. Lee’s research interests include high-frequency power conver-
multiple windings, arbitrary waveforms, and two-dimensional or three- sion, magnetics and EMI, distributed power systems, renewable energy, power
dimensional field geometry,” IEEE Trans. Power Electron., vol. 16, no. 1, quality, high-density electronics packaging and integration, and modeling and
pp. 142–150, Jan. 2001. control. Dr. Lee holds 72 U.S. patents, and has published over 250 journal arti-
cles and over 640 refereed technical papers. During his tenure at Virginia Tech,
Dr. Lee has supervised to completion 75 Ph.D. and 83 Master’s students.
Mingkai Mu received the B.S.E.E. and M.S.E.E. de- Dr. Lee served as the President of the IEEE Power Electronics Society
grees from Zhejiang University, Hangzhou, China, in (1993–1994) and is a recipient of the William E. Newell Power Electronics
2004 and 2007, respectively, and the Ph.D. degree Award in 1989; Fellow of IEEE in 1990; PCIM Award for Leadership in Power
from Virginia Tech, Blacksburg, VA, USA, in 2013. Electronics Education presented at HFPC in 1990; the Arthur E. Fury Award for
He is currently a Research Scientist at the Cen- Leadership and Innovation in1998; the Honorary Sun Yuen Chuan Chair Pro-
ter for Power Electronics Systems, Virginia Tech. fessor from National Tsing Hua University, Taiwan in 2001; the Ernst–Blickle
His current research interests include high-frequency Award sponsored by SEW-EURODRIVE FOUNDATION in 2005; the Distin-
magnetics and high power density integration. guished Alumni Award from National Cheng Kung University in 2006; the
Honorary Li Kwoh-Ting Chair Professor, National Cheng Kung University in
2011; and the inaugutal member of the Virginia Tech Entrepreneur Hall of Fame
in 2012. Dr. Lee is a member of the US National Academy Of Engineering and
an Academician of the Academia Sinica.

Qiang Li (M’12) received the B.S. and M.S. de-


grees in power electronics from Zhejiang University,
Hangzhou, China, in 2003 and 2006, respectively, and
the Ph.D. degree from Virginia Tech, Blacksburg, VA,
USA, in 2011.
He is currently an Assistant Professor at the Cen-
ter for Power Electronics Systems, Virginia Tech. His
Khai D. T. Ngo (S’82–M’84–SM’02) received the
current research interests include high-density elec- B.S. degree from California State Polytechnic Uni-
tronics packaging and integration, high-frequency
versity, Pomona, CA, USA, in 1979, and the M.S.
magnetic components, and high-frequency power
and Ph.D. degrees from the California Institute
conversion.
of Technology, Pasadena, CA, USA, in 1980 and
1984, respectively, all in electrical and electronics
engineering.
David Joel Gilham was born in San Diego, CA, He is currently a Professor of Electrical and
USA, in 1979. He received the B.S and M.S. degrees Computer Engineering at Virginia Tech, Blacksburg,
in electrical engineering from Virginia Tech, Blacks- USA, where he is currently involved in technologies
burg, VA, USA, in 2007 and 2013, respectively. for integration and packaging of power passive and
Since 2007, he has been under the direction of Fred active components to realize building blocks for power electronic systems. He
C. Lee at the Center for Power Electronic Systems, was a Member of Technical Staff at General Electric Corporate Research and
Virginia Polytechnic and State University, Blacks- Development Center, Schenectady, NY, from 1984 to 1988. Between 1988 and
burg, where he is currently the Electrical Lab Man- 2006, he was with the University of Florida, Gainesville, FL, USA. These tech-
ager and Packaging Engineer. His current research nologies lead to power conversion systems with higher efficiency and higher
interests include high-frequency converters, mag- power density. His current research interests include topologies, control, emis-
netic materials, and hybrid packaging using ceramic sion, integration issues for RF power converters, magnetic materials and com-
substrates. ponents, energy reclamation, and power integrated circuits.

Authorized licensed use limited to: J.R.D. Tata Memorial Library Indian Institute of Science Bengaluru. Downloaded on July 09,2023 at 11:57:05 UTC from IEEE Xplore. Restrictions apply.

You might also like