Powder Core Applications in High Performance EMI Filters

Tim Slattery,
Applications Engineer
The Arnold Engineering Company
Marengo, Illinois
May, 99

The use of molybdenum permalloy, 50% nickel-iron alloy and sendust-type powder cores in
power filter inductors is presented. Power line filtering to reduce differential-mode conducted
EMI is the featured application. The article begins with a simple explanation of filtering and an
example that demonstrates the advantages of using more than one inductor in a filter design.

In addition, power loss and inductance stability with respect to operating current frequency and
magnitude is compared between inductors made with the three different core types. Graphs of
equivalent series inductance, equivalent series resistance and impedance versus frequency
are used to show the effects of winding distributed capacitance and core material eddy cur-
rents.

This paper is a companion to “Powder Core Applications in Switching Amplifier and High
Performance EMI Filters,” an article written by Donald E. Pauly and sponsored by The Arnold
Engineering Company. It also complements “Power Supply Magnetics” (a three-part article) by
Mr. Pauly and published in the January, February and March, 1996 issues of PCIM Magazine.
Copies are available from The Arnold Engineering Company.

Multiple Pole Filters – Advantages
Filters as they apply to electrical and electronic
power conversion systems are circuits with
inductors and capacitors as elements. The
arrangement and sizes of these elements are
chosen so that only relatively low frequen-
cies of electrical energy are allowed to
pass. This creates a “low-pass filter.”
The nature of the circuits which are attached to
the filter input and output (source and load
impedances) have a profound effect on the
frequency response. The filter poles are not
seen as infinite responses because of damping
L = 13 µH (15 Turns)

Filter design is quite complicated, requiring

considerable knowledge of mathematics Power C = 15 µF Noise
and computer-aided engineering as well as
practical experience. The term “pole”
refers to a theoretically infinite output Two-Pole or “Single-Stage” Filter (15 Amperes)

response to input at a particular frequency.

For the two filters that will be considered L4 = 8 µH (10 Turns) L2 = 8 µH (10 Turns)

here, it is sufficient to identify the single

capacitor and single inductor arrangement
Power C3 = 15 µF C1 = 15 µF Noise
as a two-pole or single-stage filter and the
combination of two capacitors and two
inductors as a four-pole or two-stage filter. Four-Pole or “Two-Stage” Filter (15 Amperes)

See Figure 1. The number of poles corre-

sponds to the number of elements.
Figure 1. Two-Pole and Four-Pole Filters.

Page 1
by the load and source. In practice, the filter
elements include resistance associated with
inductor winding and core losses as well as
capacitor lead, electrode and dielectric losses.
Higher frequency loss provides additional
damping that is desirable for stability.1 Also,
parasitic elements such as capacitor lead
inductance and inductor winding distributed
capacitance influence filter performance.
These parasitic elements are shown in Figure 2
along with the equivalent series resistance of
the inductor, Rs, and capacitor, Rc.
Stand-alone constructions of the two types of
filters for testing are shown in Figure 3 and
Figure 4. The single-stage filter uses a 15 µF
Ls
Rs
Cl
50 Ω
50 S Test
(Output)
HP 4194A Gain-Phase Analyzer
Figure 2. Single-Stage Filter Model Showing Parasitic Elements and Test and Source Load.
Figure 3. Single-Stage Filter Construction and
Test Leads.
1
See Mitchell, Daniel M., DC-DC Switching Regulator Analysis,
McGraw-Hill, New York, 1988, ISBN 0-07-042597-3, Chapter 7,
“Effects of EMI Input Filtering.”
Page 2
polypropylene capacitor and a 13.2 µH inductor.
The inductor core is Arnold Engineering part
number MS-130060-2, a sendust-type core with
a permeability of 60. (Arnold manufactures and
sells this type of core under the trade name
Super-MSS.) The conductor is made from
three strands of 18 AWG magnet wire and
results in a DC resistance of 4.5 mΩ.
The two-stage filter uses two of the same 15 µF
capacitors as in the single-stage design. Each
of the two inductors has a value of 7.95 µH and
is based on a smaller core of the same material
and permeability, Arnold part number
MS-106060-2. The conductor is the same size
as for the larger inductor but only ten turns are
used. The total resistance for the inductors
Lc 50ΩS
50
Reference
Rc
(Input)
0.0 dBm
Cs
Figure 4. Two-Stage Filter Construction and Test
Leads.
connected in series is 5.4 mΩ. Because of the A similar graph for the capacitor is shown in
fewer turns, the magnetizing force on the Figure 6. In this case, parallel capacitance and
smaller cores is 14.5% less for the same value resistance are plotted against frequency. Note
of current. that the capacitor resonates with the inductance
of its leads at about 250 kHz.
Figure 5 shows how the equivalent series
inductance and resistance varies with fre- All of the frequency response graphs in this
quency for each type of inductor. The larger paper are based on data from a
one self resonates at about 26 MHz whereas Hewlett-Packard 4194A Impedance/
the smaller one is still inductive beyond Gain-Phase Analyzer. For each inductance
40 MHz. A wider frequency range is typical of and capacitance measurement, the test signal
smaller inductors. There will be more on voltage used was nominally 0.5 V rms.
inductor characteristics at high frequency in the
next section.
20 6000
16.67
Ls Single Stage
5500

13.33 5000
Rs Single Stage
10 4500

6.67 4000
Ls 2 Stage
3.33
Ls (µH)

3500

Rs (Ω)
0 3000

-3.33 2500

-6.67 2000

-10 1500

-13.33 Rs 2 Stage 1000

-16.67 500

-20 0
2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 2 3 4

10 100 1000 10000

Frequency (kHz)

Figure 5. Equivalent Series Inductance and Resistance versus Frequency for Single and Two-Stage
Inductors.

40 400

30 350
Cp
20 300

10 250
Cp (µF)

Rp (Ω)

0 200
Rp
-10 150

-20 100

-30 50

-40 0
2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 2 3 4

10 100 1000 10000

Frequency (kHz)

Figure 6. Equivalent Parallel Capacitance and Resistance versus Frequency for the 15 µF Capacitor.

Page 3
To show the advantage of multiple-pole filtering, For example, the maximum attenuation (mini-
the frequency response for the single-stage mum gain) for each filter is the result of series
(two-pole) and two-stage (four-pole) examples resonance of the capacitor and its lead induc-
are presented in Figure 7 and Figure 8. The tance. It is especially noticeable at about
most significant departures from a power circuit 175 kHz for the single-stage filter. The impor-
application are the 50 Ω source impedance and tance of minimizing lead length is apparent.
50 Ω load, which are provided by the Attenuation decreases at higher frequency
“gain-phase” part of the Hewlett-Packard 4194A because lead inductance is impeding the flow of
Analyzer. In a typical switching-type power return current through each capacitor.
supply, the impedances are variable, not
Another important observation with regard to
matched and usually much lower in value at low
filter behavior is the lower damping of the
frequencies. The test does provide useful
two-stage configuration around 20 kHz. Theo-
information for comparison even though the
retically, the four poles are two double poles at
application conditions differ.

0 240

-10 180

-20 120

Phase

Phase (deg)
-30 Single - Stage 60
Gain (dB)

-40 0

-50 -60
Single - Stage Gain
-60 -120

-70 -180

-80 -240
2 4 6 8 2 4 6 8 2 4 6 8 2 4 6 8 2 4 6 8 2 4 6 8 2 4 6 8

1e-005 0.0001 0.001 0.01 0.1 1 10 100

Frequency (MHz)
Figure 7. Gain and Phase versus Frequency for the Single-Stage Filter.

0 240

-10 180
Two - Stage Gain
-20 120

60
Phase (deg)

-30
Gain (dB)

Two - Stage Phase

-40 0

-50 -60

-60 -120

-70 -180

-80 -240
2 4 6 8 2 4 6 8 2 4 6 8 2 4 6 8 2 4 6 8 2 4 6 8 2 4 6 8

1e-005 0.0001 0.001 0.01 0.1 1 10 100

Frequency (MHz)
Figure 8. Gain and Phase versus Frequency for the Two-Stage Filter.

Page 4
 0 240
Single - Stage Gain
-10 180
Two - Stage Gain
-20 120

Single - Stage Phase

Phase (deg)
-30 60
Gain (dB)

Two - Stage Phase

-40 0

-50 -60

-60 -120

-70 -180

-80 -240
2 4 6 8 2 4 6 8 2 4 6 8 2 4 6 8 2 4 6 8 2 4 6 8 2 4 6 8

1e-005 0.0001 0.001 0.01 0.1 1 10 100

Frequency (MHz)
F ig u re 9. C o m p o site G ain an d P h ase versu s F req u en cy fo r O n e an d T w o -S tag e F ilters.

14.6 kHz and their effect is apparent because of Power Line Filters
the lack of damping between C1, L2 and C3. The term filter usually applies to an electrical
About 30 dB of attenuation is lost at 20 kHz. circuit or portion of a circuit that prevents
Above that frequency, the filter recovers and electromagnetic interference (EMI). Without
outperforms the single-stage design by a filtering, unwanted electrical signals could travel
remarkable 20 dB at 60 kHz. See Figure 9. from one device to another along the power line
Finally, note that the two-stage filter maintains or bus bar that they share. The conducted
a 10 dB advantage from 300 kHz to 1 MHz. interference can also be subsequently radiated
The additional inductor and capacitor reduce since a power line is an antenna for higher
the capacitor lead inductance effect.2 In the frequencies. The purpose of the filter is to
United States, because the AM broadcast band prevent this electrical “noise” from being con-
is from 540 kHz to 1.6 MHz, improved filter ducted to the line from the device while allowing
performance in this range of frequencies is of the desired electrical power to pass.
particular benefit. For equipment connected to a public utility,
The filters were tested with the input or refer- government agencies regulate the maximum
ence channel on the capacitor side and test conducted noise voltage over specific fre-
channel on the inductor. It was observed that quency ranges. For example, in the United
the frequency response was very similar with States, the Federal Communications Commis-
the inductor side connected to the reference sion specifies a conducted radio frequency
channel and the capacitor side tied to the test interference (RFI) limit of 48 dBµV from
channel. Therefore, the observations above 450 kHz to 30 MHz.3 The purpose is to prevent
also apply to the filters treated as the interference with radio, television and telephone
inductor-input type. services used by the general public.
Noise limits for the outputs of power supplies
2
See Don Pauly’s article “Power Supply Magnetics,” page 16 of
the reprint available from Arnold Engineering or “Part III” in the within an electronic system are determined by
March, 1996 issue of PCIM on battery line filters. the requirements of the attached loads. In most
3
FCC Part 15 Class B, devices for home or consumer use. cases, sufficient noise filtering is accomplished
“dBµV” stands for decibel microvolt; 48 dBµV is 0.00025 volt; with the same power filter elements (energy
“kHz” is kilohertz or thousand cycles per second and “MHz”
refers to megahertz or million cycles per second.
storage inductor and output capacitor) that

Page 5
control the output voltage ripple. However, The test data in Figure 12 and the data that
some designs include a second stage of filtering follows came from the same three cores. Each
to control EMI on the output of the supply.4 has a permeability of 60 and is the same size as
the core used in the single-stage filter. For
Equipment that contains continuously switching
reference, the Arnold part numbers are
components, such as the power transistors and
A-291061-2, HF-130060-2 and MS-130060-2,
diodes of a switched-mode power supply,
representing the molybdenum permalloy (MPP),
require conducted EMI filtering on the input
50% nickel-iron and sendust-type powders,
side. The abrupt changes of current in the
respectively.
circuit cause brief voltage rises or “spikes”
either across the input conductors or on both An example of a “fully” wound core is shown in
conductors to ground. (“Ground” includes Figure 10. Fully wound means that one-half of
ground return wires, ground planes and the core inside diameter remains. Usually, at
grounded enclosures.) least this much room must be provided for a
hook or shuttle to place the last turn. In this
EMI voltage between input conductors is called
case, the inductance value is 1.9 mH and is
differential-mode noise. Noise from both
typical for power line applications. Inductance
conductors to ground is termed common-mode.
requirements generally range from a few
An inductor for common-mode noise utilizes the
microhenries to several millihenries.
opposing currents in the input conductors (two
windings on one core) and a high permeability Low core loss at power line frequency is neces-
core material. sary to take advantage of the core material’s
high saturation flux density. Measurements are
In contrast, a differential-mode inductor requires
presented for Hi-Flux in Figure 13. Hi-Flux has
a core material that can maintain permeability
the most loss so it can be used as a worst-case
with a bias field. Refer to the graphs in Fig-
reference. Even at 400 Hz and 9000 gausses,
ure 11 showing inductor current, voltage and
the core loss density is low at 200 mW/cm3. For
core magnetic fields. Note that the source in
50 and 60 Hz applications, the flux density limit
Figure 11 is either a power line (via a wall outlet,
is determined by the change in permeability as
for example) or a battery such as the 48 volt
shown in Figure 14.
battery for telephone central office equipment. In
a battery system, the magnetizing current
is a constant DC. For an AC system with
high power factor, the magnetizing current
is nearly sinusoidal. With a low power
factor AC system, the current is a series
of alternating pulses.
Powder cores are appropriate for differen-
tial-mode, sometimes referred to as
“series-mode,” inductors or “chokes,”
because of their extraordinary capability
to maintain inductance with bias. The
50% nickel-iron alloy powder is particu-
larly useful at high flux densities. (Arnold
Engineering’s trade name for this material
is Hi-Flux.) For comparison, permeabil-
ity versus DC bias curves for the three
types of powder cores are shown in Figure 10. Example of “Fully” Wound Core
Figure 12. Figure 10. Example of “Fully” Wound Core.

4
See Billings, Keith H., Switchmode Power Supply Handbook,
McGraw-Hill, New York, 1989. ISBN 0-07-005330-8. pp. 1.151
and 1.152.
Page 6
 Differential-Mode
Inductor, L
I
+ V - Switching
Power Circuit
Capacitor, C and Load

Two-Pole or “Single-Stage” Common-Mode

Filter for Differential-Mode Inductor
Conducted EMI.

50/60 Hz Sinusoidal Current

B Battery I (High Power Factor)
Current
BD C
0 t
Battery DC
Operating
Point
H 50/60 Hz Current Pulses (Low Power Factor)
HDC V
50/60 Hz
Hysteresis
Loop
0 t

Switching Noise Voltage

Figure 11. Typical EMI Filter Configuration and Differential-Mode Inductor

Voltage,
Figure 11. Typical Current
EMI Filter and Magnetic
Configuration Waveforms.
and Differential-Mode Inductor Voltage, Current
and Magnetic Field Waveforms.

Page 7
 100 100

95 95
MPP
90 100 % = 60µ 90

85 85

Percent permeability
Percent Permeability Hi-Flux
80 80

75 75

70 70
SMSS
65 65

60 60

55 55

50 50

45 45

40 40
2 3 4 5 6 7 8 9 20 30 40 50 60 70 80 90
1 10 100
D. C. Magnetizing Force (Oersteds)
Figure 12. Permeability versus DC Bias.

Another important consideration is the

300

variation of inductance with frequency. The 200

frequency responses of single layer 60 µH 60µ Hi - Flux

inductors made with each type of core are 100

90
80

shown in Figure 15 through 17. 70

60
Core Loss Density (mW / cm3)

50

The absence of the series inductance peak 40

z
for the Hi-Flux inductor indicates that its
30
0H
40
core permeability is dropping with fre- 20

z
quency. Higher eddy current loss in the 0H
20
50% nickel-iron powder is responsible. As 10
9
8 z
mentioned before, loss at higher frequency
7
0H
6 12
can be an advantage in filters because of
5

the additional stability the damping pro- 3

vides. Greater detail regarding series 2

inductance and resistance in the frequency

range of 100 kHz to 1 MHz is given in Figure 1

18. The relatively low eddy current loss in 1000

2000 3000 4000 5000 6000 7000 8000 9000
10000

Super-MSS is apparent. Flux Density (Gauss)

Figure 13. Core Loss versus Flux Density.

10 10

7.5 7.5

5 60µ Hi - Flux 5
% Change of Permeability

2.5 2.5

0 0

-2.5 -2.5

-5 -5

-7.5 -7.5

-10 -10

-12.5 -12.5

-15 -15

-17.5 -17.5

-20 -20
2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9

10 100 1000 10000

Flux Density (Gauss)

Figure 14. Modulation of Permeability with Flux Density.

Page 8
 150
MPP
125

100

75

50
Hi-Flux
25

Ls (µH)
SMSS
0

-25

-50

-75

-100

-125

-150
2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 2 3 4

10 100 1000 10000

Frequency (kHz)
Figure 15. Equivalent Series Inductance versus Frequency, Single-Layer Winding.

20000

18000
MPP
16000

14000

12000
Rs (Ω)

10000
SMSS
8000

6000
Hi-Flux
4000

2000

0
2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 2 3 4

10 100 1000 10000

Frequency (kHz)

Figure 16. Equivalent Series Resistance versus Frequency, Single Layer Winding.

100000 8
6
4

2 MPP
10000 8 SMSS
6
4

Hi-Flux
Impedance (Ω)

1000 8
6
4

100 8
6
4

10 8
6
4

1
2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 2 3 4

10 100 1000 10000

Frequency (kHz)

Figure 17. Impedance versus Frequency, Single Layer Winding.

Page 9
Finally, the effect of greater distributed capaci-
tance with overlapping conductor turns (multiple
layers of magnet wire) is shown graphically in
Figure 20 and Figure 21. Beyond 1.6 MHz, this
stray capacitance actually causes the
high-value, multiple-layer inductor to have a
lower impedance than the low-value
single-layer inductor.5
In conclusion, each of the powder core types is
applicable to power line filtering. The 50%
nickel-iron material performs the best because
65
Ls Hi-Flux
64.5
of its ability to sustain inductance with higher
magnetizing current. It also provides some
desirable damping at higher frequencies.
Another important consideration is acoustic
noise caused by magnetostriction of the mag-
netic metal alloy. A 50% nickel-iron alloy core
can make a humming sound at high 50 or
60 Hz flux levels. Of course, DC magnetizing
current does not cause audible noise so the
50% nickel-iron is usually the best material for
battery power line filters.
40
35

64 30

x
lu
63.5 25

i-F
H
Rs
63 20
Ls (µH)

Rs (Ω)
Ls MPP
62.5 15

62 10
PP
Rs M
61.5 5
SS
Rs SM
61 0
Ls SMSS
60.5 -5

60 -10
200 300 400 500 600 700 800 900
100 1000
Frequency (kHz)
Figure 18. Equivalent Series Inductance and Resistance versus Frequency, 100 kHz to 1 MHz.

1000900
800
700
600
500

400

300

Hi-Flux
HI-Flux
200

MPP
Impedance

SMSS
100 90
80
70
60
50

40

30

20

10
200 300 400 500 600 700 800 900
100 1000
kHz
Figure 19. Impedance versus Frequency, 100 kHz to 1 MHz.

5
The multiple-layer winding is 170 turns of 18 AWG magnet wire.
Page 10
Both MPP and Super-MSS have exceptionally The sendust-type core has the lowest cost per
low magnetostriction and either could be used unit volume and is a good choice where the
to minimize audible noise. MPP has the advan- highest possible performance is not required.
tage of better incremental permeability with DC The difference in cost of molybdenum permalloy
bias. The 50 or 60 Hz current is essentially DC powder and the 50% nickel-iron alloy powder is
compared to the frequency of electrical noise. negligible.
Incremental permeability versus DC bias
curves can be used to predict the inductance at
any point of the 50/60 Hz current waveform.

80 16000
70 14000
60 12000
Single - Layer
50 10000
40 8000
Single Layer Ls (µH)

Full Wound Ls (µH)

30 6000
20 4000
10 2000
0 0
-10 Full Wound -2000
-20 -4000
-30 -6000
-40 -8000
-50 -10000
-60 -12000
-70 -14000
-80 -16000
2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 2 3 4

10 100 1000 10000

Frequency (kHz)
Figure 20. Equivalent Series Inductance versus Frequency for Single-Layer and Fully Wound Designs.

100000 8
6
4

2
Full Wound Single - Layer
10000 8
6
4
Impedance (Ω)

1000 8
6
4

100 8
6
4

10 8
6
4

1
2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 2 3 4

10 100 1000 10000

Frequency (kHz)
Figure 21. Impedance versus Frequency for Single-Layer and Fully Wound Designs.

Page 11
Acknowledgements
The author thanks Jim Motley for his careful and
skillful work in preparing all of the data curves in
the paper and Jean Wickman for assisting him.
Don Pauly was a very helpful and encouraging
proofreader and his insights regarding filter
design were essential.
References
Billings, Keith H., Switchmode Power Supply Hand-
book, McGraw-Hill, New York, 1989.
ISBN 0-07-005330-8.
Mitchell, Daniel M., DC-DC Switching Regulator
Analysis, McGraw-Hill, New York, 1988,
ISBN 0-07-042597-3
Reference Data for Engineers: Radio, Electronics,
Computer, & Communications, Eighth Edition, Mac
E. Van Valkenburg, Editor-in-Chief, Sams Publishing,
Carmel, Indiana, USA, 1995, ISBN 0-672-22753-3
Pauly, Donald E., “Power Supply Magnetics, Part III,”
PCIM Magazine, March, 1996. (Available from The
Arnold Engineering Company.)
Pauly, Donald E., “High Fidelity Switching Audio
Amplifiers Using TMOS Power MOSFETs,” AN1042/
D, 1989, Motorola Technical Information Center,
Phoenix, AZ.
Brown, Paul, “Not All Millihenries Are Equal,” pre-
sented at an IEE Colloquium on Capacitors and
Inductors for Power Electronics, Savoy Place,
London, March, 1996. Paper available as various
abstracts through Almag Ltd, Braintree, Essex,
England (e-mail: enquiry@almag.co.uk)

Page 12