Beruflich Dokumente
Kultur Dokumente
Title:
Running Head:
Abstract
Filter inductors are probably one of the heaviest and more voluminous components found in
power supplies of most electronic devices. A known technique to reduce the inductor size in DC
applications is the use of permanent magnet inductors (PMIs). One of the latest developed
biasing topologies, termed the Saturation-gap, uses two standard UU cores simultaneously
biased by PMs. This paper presents a size reduction design strategy, implemented on a 10A
5mH output filter inductor pair of a DC-DC converter, using the Saturation-gap topology. Two
different PMIs designs are presented. The standard filter and the two versions using PMIs were
physically implemented and tested in the power electronics lab. The two different implemented
PMIs filters present 50% core volume reduction and 25% and 50% copper volume reduction
respectively. The measurements on the three filters are compared with simulation results using
Inductors are probably one of the heaviest and more voluminous components found in most
power electronic converters. Permanent magnet inductors (PMIs) can be used in place of
standard inductors operating in DC applications. PMIs present a higher energy density than
standard non-biased inductors and can minimize the weight and size requirements [1] - [13].
These special kinds of inductors utilize permanent magnets (PMs) in order to introduce an
opposing bias magnetic flux in the core material, effectively extending the saturation current
limit. The different core and PM topologies used on PMIs, has been evolving from its early
beginnings [3]. The main limitations of the first basic topologies have been minimized at
One of the recently develop PMI topologies, the Saturation-gap, can be implemented using
standard U cores, avoiding the need of special shaped cores and the related manufacturing
process. This topology uses one pair of magnets simultaneously biasing two non-coupled
inductors [1] - [3]. This is especially suitable for applications using a symmetric pair of inductors,
one at the positive and other at the negative terminal, like: the output filter in power converters,
Fig. 1. Block Diagram of output filters inductors topologies. a) Standard Output Filter inductor pair. b) Output filter
with the Saturation-gap biasing configuration, it uses two non-coupled Inductors, simultaneously biased with a
single pair of permanent magnets (PMs).
Fig. 2. Typical Inductance vs Current profiles of three inductor types. The colored areas are representative of
storable energy and are proportional to inductor size. The typical operation current range of a DC output filter is
defined between the red lines (Imax, Imin). F1 and F2 indicate the bias magnetization force introduced in PMI1
and PMI2 respectively.
In this paper, the Saturation-gap biasing topology has been used to reduce the size of the output
filter pair of a DC to DC converter. The block diagram of the conventional and the PMI filter
versions are shown in fig.1. The filter, including the two inductors in series, is required to provide
an inductance of 5mH and operate at a DC current level of 9A, with a ripple current from 8A
Fig.2. presents the typical inductance vs current (LvsI) profile of a standard inductor and two
PMIs with different size reduction strategies. The area enclosed by the LvsI profiles is
representative of storable energy and is proportional to the physical size of the inductor. Non-
biased inductors present a symmetric LvsI profile, with a linear range extending from –𝐼𝑚𝑎𝑥 to
𝐼𝑚𝑎𝑥 . This paper presents two different PMI design approaches. The first PMI design shifts the
LvsI profile, concentrating its linear region from 0A to 𝐼𝑚𝑎𝑥 . This approach can be described as
full linear biasing, and is commonly found in the related PMI literature [1] - [15]. In order to
further reduce the inductor size requirements, a second PMI design is been implemented, where
the linear region of the LvsI profile have been further concentrated to the operating current
region, from 𝐼𝑚𝑖𝑛 to 𝐼𝑚𝑎𝑥 . This approach can be described as oversaturated biasing or non-linear
biasing. The PMI filter implemented with the design approach of full linear biasing (PMI1),
achieve 50% core reduction, and 25% copper volume reduction. The PMI filter implemented with
the oversaturated biasing approach (PMI2), achieves 50% reduction in both core and copper
volume. The following section II covers a basic summary of the relevant theory background,
including the basics on inductor sizing, permanent magnet characterization and a review of the
evolution of PMIs topologies. Section III presents the basic design procedure with saturation-gap
where the two size reduction approaches are calculated. Section IV covers the implementation
and empirical measurements perform on the physical units. Section V presents the 2D FEA
simulations implemented with FEMM software and the measured and simulated results are
compared. Finally section VI summarizes the conclusions and future research on the subject.
A. Inductor sizing
The first step of an inductor design is to determine the basic sizing limitations of the inductor
operation requirements. The product of the maximum operating current, 𝐼𝑚𝑎𝑥 and nominal
(1)
Where, 𝐵𝑆𝑎𝑡 is the saturation flux density of the core material, 𝐴𝑐 is the area cross-section of the
core and 𝑁 is the number of turns. The 𝐿𝐼 product is a constant, particular of each specific
design requirement, and it defines the maximum achievable area for the LvsI profile of a given
inductor design. A generic 𝐿𝐼 product area of unity value is depicted by the blue lines in Fig.3.
The requirements of the 𝐿𝐼 product can be achieved by different compromises in the choice of
the core material (𝐵𝑆𝑎𝑡 ), number of turns and area cross-section. Different strategies exist to
determine these compromises, as function of the optimization target. Most common targets are:
Minimal losses, Minimal thermal resistance, Minimal size or Minimal cost [4] - [7].
Fig. 3. LvsI profile of an inductor as function of air-gap length. Blue curve is a generic unity LI product region.
Orange line represents the LvsI profile for each air-gap length. Left: short air-gap; Center: long air-gap; Right: non-
constant length air-gap.
After the parameters in (1) have been chosen, according to the desired target, the air-gap length
must be adjusted in order to define the nominal inductance range. Assuming the total reluctance
𝑁2
𝐿𝑁 =
𝑅𝑔
(2)
𝑙𝑔
𝑅𝑔 =
𝜇0 𝐴𝑔
(3)
Where 𝑙𝑔 is the length of the air-gap along the flux axis, 𝜇0 is the permeability of free air and 𝐴𝑔
is the area cross-section of the air-gap. For a given cubic air-gap with dimensions: [𝑙𝑥 ; 𝑙𝑦 ; 𝑙𝑔 ], in
order to account for the fringing flux outside the gap, a commonly found approach is to calculate
𝐴𝑔 as:
𝐴𝑔 = (𝑙𝑥 + 𝑙𝑔 )(𝑙𝑦 + 𝑙𝑔 )
(4)
Fig. 4. Second Quadrant of the Hysteresis Loops of a Lineal Permanent Magnet. In purple is the Intrinsic (or
polarization) curve. In blue is the Extrinsic (or normal) curve. HK, BK indicates the threshold of irreversible
demagnetization (knee-point). Green, blue and grey areas represents: stored energy, effective energy product and
co-energy, respectively.
In Fig. 3. It is shown the approximate LvsI profiles of three hypothetical air-gaps. The longer the
air-gap, the bigger the reluctance and lower the nominal inductance, 𝐿𝑁 . Saturation occurs at the
current value at which 𝐿𝑁 intersects with the 𝐿𝐼 area limit (blue curves). The use of non-constant
length air-gaps produces non-linear LvsI profiles. Some inductor applications and can benefit
from these LvsI profiles presenting higher 𝐿𝑁 values at low currents and therefore optimize the
B. Permanent Magnets
Permanent magnets PMs, are used to provide the desired biasing flux in PMIs. A careful choice
of the PM material and dimensions is required for the design of PMIs. The basic characterization
parameters can be analyzed by the PMs demagnetization curves, as depicted in Fig. 4. The
intrinsic curve (purple) represents the polarization, J achieved in the PM material as function of
the applied magnetizing field, 𝐻. This measurement is typically done in a magnetic short circuit
condition. The extrinsic curve (blue) represents the flux density 𝐵 produced by the PM as
function of the magnetic field, 𝐻 after being magnetized. The operation point of the PM can be
(5)
Where ɸ is the actual PM flux, 𝐴𝑃𝑀 is the area cross-section of the PM, 𝐵 is the magnetic flux
density, 𝐵𝑟 is the remanent flux density and 𝜇𝑃𝑀 is the permeability of the PM. In cases where no
𝐻𝑐 𝑙𝑃𝑀
ɸ=
𝑅𝑃𝑀 + 𝑅𝑙
(6)
Where, 𝐻𝑐 is the coercive force of the PM, and 𝑙𝑃𝑀 is the PM length along its axes of
magnetization. 𝑅𝑃𝑀 is the reluctance of the PM and 𝑅𝑙 is the reluctance of the media
surrounding the PM or load reluctance. The blue area in Fig.4. represents the effective magnetic
energy product sourced by the PM, at a given operation point. Applying an external
demagnetizing field, 𝐻 or increasing the load reluctance, 𝑅𝑙 , decreases the PM’s operation point,
increasing the stored energy in the PM, represented by the green area in Fig. 4.
Rare-earth PM materials, like NdFeB or SmCo present a linear extrinsic curve in the second
quadrant, as depicted in Fig. 4. As long as the operation point of the PM is kept within the linear
region, no irreversible demagnetization would take place. The knee-point at the extrinsic curve
temperature reduces the linear range, bringing the knee-point towards the 2nd quadrant. The
minimum operation point for the PM at the maximum working temperature of the PMI must be
The specific core shapes and PMs topologies used in PMIs have been evolving from its early
beginnings. The basic characteristics of the main topologies, found in the literature, are sketch
on Fig. 5. The red vectors represent flux induced by the coils and green vectors represent flux
induced by the PMs. In the following, the main advantages and limitations of the different
The first designs of PMIs date back to the early 1960’s [4] - [7], with the development of barium
ferrite PMs. These first designs consisted of standard gaped inductors with a permanent magnet
inserted in the air-gap. Fig. 4. (left) shows the basic configuration. The use of standardized core
shapes allows for the use of common inductor design techniques, by introducing correction
factors for the effective 𝐵𝑆𝑎𝑡 in presence of the PMs flux. This arrangement can effectively
introduce a certain amount of bias flux. There is however several limitations intrinsic to this
configuration:
The flux from the coil is passing directly through the PM. Strong demagnetizing field and
These limitations greatly reduced the applicability of this basic topology and more sophisticated
configurations proceed through the years. On the other hand, the relatively recent development
of rare-earth PM materials have improve the design tradeoffs and extending the range of
In order to overcome the limitations of placing the PM inside the air-gap, a logical approach has
been to place the PMs in the vicinity of the air-gaps. This strategy can effectively lower the
demagnetizing fields at the PM and minimize eddy currents. On the other hand, the use of non-
standard core shapes implies a higher production cost and requires the development of specific
inductor design techniques. Relevant documentation on this topology is mostly limited to patent
documents. A long list of particular designs with different core shapes and PMs, based on this
topology can be found in patent documents. Only four scientific publications have been found
regarding the design of PMIs with this topology [11] - [14]. In Fig. 4. (center), an example design
The Saturation-Gap
A new biasing concept is been recently develop [1] – [3]. This topology, as depicted in Fig. 4.
(right), can be implemented using standard UU or UI core shapes with no air-gaps. PMs in this
topology are perpendicular and outside the flux path of the coils, avoiding demagnetizing fields
at the PMs and minimizing eddy currents. One pair of PMs is simultaneously biasing two
inductors. The flux linkage among these two inductors is zero and therefore they are mutually
decoupled. This can be appreciated observing the coil’s flux paths (red vectors) in Fig. 4. On the
other hand, both inductors can be connected in series and operate as a single inductor. The
To introduce the desired biasing flux in the longest section of the cores.
To induce certain level of magnetic saturation in a localized section of the cores. This
section presents higher reluctance and performs the equivalent purpose of the typical air-
gap.
The following section presents a design procedure with the saturation-gap topology.
saturation-gap topology, the simplified magnetic equivalent circuit (MEC) presented in Fig. 6.
can be considered. The biased section of each core is defined as the bias reluctance, 𝑅𝑏𝑖𝑎𝑠 and
the saturated section of each core is defined as the saturation-gap reluctance, 𝑅𝑠𝑔 . Since the
relative permeability, 𝜇𝑟 in any core material, is a function of the instantaneous flux density, 𝐵,
these reluctances are depicted as non-lineal resistors. The basic current source model is used
for the PMs. The internal reluctance of the PMs, 𝑅𝑃𝑀 can be calculated as:
𝑙𝑃𝑀
𝑅𝑃𝑀 =
𝜇0 𝜇𝑟 𝐴𝑃𝑀
(7)
The same approximation taken in (4) can be used in order to account for the fringing flux at the
ɸ𝑟 = 𝐴𝑃𝑀 𝐵𝑟
(8)
The actual PM’s flux, ɸ𝑃𝑀 , introduced in the core reluctances (𝑅𝑏𝑖𝑎𝑠 and 𝑅𝑆𝑔 ) is calculated as:
(9)
Fig. 6. The Saturation-gap topology. Simplified MEC model (left) and Operation Mode Analysis (right). Three
operation modes as function of PMs flux over core area cross-section (ɸPM/AC). Green areas delimit the linear
region of the BH curves of the core material; Red areas delimit the non-linear region. Orange and blue markers
represents the operation point of RSg and Rbias at initial flux condition (I = 0A). QSE is the Quasi Saturated
Equilibrium flux density.
Where ɸ𝑑 is the self-demagnetization flux through 𝑅𝑃𝑀 . The value of ɸ𝑃𝑀 can also be calculated
𝑅𝑏𝑖𝑎𝑠 𝑅𝑆𝑔
𝑅𝑙 =
𝑅𝑏𝑖𝑎𝑠 + 𝑅𝑆𝑔
(10)
Taking the PM current source as reference, the core reluctances (𝑅𝑏𝑖𝑎𝑠 and 𝑅𝑆𝑔 ) appear in
parallel configuration. Taking the coil voltage source as reference, the core reluctances (𝑅𝑏𝑖𝑎𝑠
and 𝑅𝑆𝑔 ) appear in series configuration. This can be observed looking at the flux vectors
The basic flux and reluctance behavior in a saturation-gap topology and the concept of the quasi
saturated equilibrium QSE, have been introduced in [2][3]. The QSE can be defined as: the limit
flux density achievable at any point in the length of the core by a given PM. When the flux
density at the saturation-gap reluctance, 𝑅𝑆𝑔 reaches the QSE value, the topology presents a
linear inductor behavior. Increments in flux at 𝑅𝑆𝑔 , due to increments in the coil’s current, are
compensated by equivalent decrements of PM’s operation point, lowering the induced PM’s flux
ɸ𝑃𝑀 and maintaining the QSE at 𝑅𝑆𝑔 . In the MEC model presented in Fig. 6., the current source
ɸ𝑟 is in parallel with both: 𝑅𝑆𝑔 and 𝑅𝑃𝑀 . The reluctance value of 𝑅𝑆𝑔 at the QSE flux density level
is equal to the PM’s reluctance 𝑅𝑃𝑀 . This can be used with (2) in order to define the nominal
𝑁2 𝑁2
𝐿𝑁 = =
𝑅𝑆𝑔 𝑅𝑃𝑀
(11)
C. Operation Modes
The initial flux conditions can be defined as the total flux density, in 𝑅𝑏𝑖𝑎𝑠 and 𝑅𝑆𝑔 , at zero current
(𝐵[0𝐴]). Three different operation modes has been identified, as function of the initial flux
conditions 𝐵[0𝐴].
ɸ𝑃𝑀
𝐵[0𝐴] =
𝐴𝑐
(12)
In Fig. 6. (right) are represented the characteristic LvsI profiles of each operation mode. The
operation point, at initial flux conditions of 𝑅𝑏𝑖𝑎𝑠 and 𝑅𝑆𝑔 are depicted in the BH diagram as the
blue and orange markers respectively. The basic characteristics of each operation mode are
Sub-Saturated
The flux density 𝐵[0𝐴] at 𝑅𝑏𝑖𝑎𝑠 is within the linear region at the 3rd quadrant.
This operation mode produces LvsI profiles similar to those of a non-constant length air-gap.
The reluctance value of 𝑅𝑆𝑔 (at 𝐵[0𝐴]) is relatively low and the instantaneous inductance is quite
high. As current increases, the flux density 𝐵, at 𝑅𝑆𝑔 approaches QSE and the inductance
decreases reaching its nominal value (11). Further increase of the current at this state is
The flux density 𝐵[0𝐴] at 𝑅𝑏𝑖𝑎𝑠 is at the limit of the linear region (−𝐵𝑆𝑎𝑡 ) at the 3rd
quadrant.
As current increases, the flux density, 𝐵 at 𝑅𝑏𝑖𝑎𝑠 moves towards the 1st quadrant and the
operation point of the PM decreases, maintaining the QSE at 𝑅𝑆𝑔 . The value (11), of the nominal
inductance, 𝐿𝑁 remains constant while 𝑅𝑏𝑖𝑎𝑠 is operating within the linear region of the 1st and 3rd
quadrants. The saturation current, 𝐼𝑠𝑎𝑡 , is determined when 𝑅𝑏𝑖𝑎𝑠 reaches the non-linear region
Over-Saturated
The flux density 𝐵[0𝐴] at 𝑅𝑏𝑖𝑎𝑠 is at the non-linear region at the 3rd quadrant.
The totality of the core is saturated, the reluctance value of both 𝑅𝑏𝑖𝑎𝑠 and 𝑅𝑆𝑔 is relatively high
and the instantaneous inductance is very low. As current increases, the flux density 𝐵, at 𝑅𝑏𝑖𝑎𝑠
approaches the linear region at the 3rd quadrant and the inductance increases approaching its
nominal value (11). Further increase of the current at this state is equivalent to the operation
D. Design Procedure
PMI Sizing
The first step is to define the required LI constant of the design. For standard inductors, the LI
constant is defined by (1). The limit of the LI area, as depicted by the blue curve in Fig. 3, is
asymptotic with both axes (𝐼 = 0𝐴 and 𝐿 = 0𝐻). In the case of non-biased inductors, the axes
𝐼 = 0𝐴 is equivalent to the zero flux axes ( 𝐵 = 0𝑇 ). In the case of PMIs the 𝐵 = 0𝑇 axes is
equivalent to the 𝐼 = 𝐹 axes. Where 𝐹 is the biasing magnetization force, as depicted in Fig. 2.
(13)
The factor 𝐹⁄𝑁 is equal to the current at which 𝐵 = 0𝑇. For designs in Full Linear Bias operating
mode:
𝐹⁄ = 𝐼𝑚𝑎𝑥
𝑁 2
(14)
𝐹⁄ = 𝐼𝑚𝑎𝑥 − 𝐼𝑚𝑖𝑛
𝑁 2
(15)
Once the required LI constant is defined, the different compromises in the choice of the core
material (𝐵𝑆𝑎𝑡 ), 𝑁 and 𝐴𝑐 , can be made based on the desired design priority (minimal losses,
minimal size, etc… ). On the other hand, some limitations in the choice of core size and number
of turns must be taken into account in order to provide a sufficient window area, 𝑊𝑎 for the coil
Defining PM flux
The second step is to define the required ɸ𝑃𝑀 at zero current. This parameter defines the active
operation mode as indicated in Fig. 6. The choice of the operation mode stabilizes the
requirements of the PM material (𝐵𝑟 ) and its area cross-section, 𝐴𝑃𝑀 and can be calculated as:
(16)
For Full Linear Bias mode of operation, 𝐵𝑆𝑔 is equal to QSE and 𝐵𝑏𝑖𝑎𝑠 is equal to 𝐵𝑠𝑎𝑡 .
Fig. 7. FEMM Model of the Implemented Inductors. The STD model presents only one inductor from the filter pair.
The biased models PMI1 and PMI2 include both inductors from the filter pair. Depth into the page is 28 mm. Core
material: Ferrite MF198A; Core shape: U 46x40x28; STD:[g=2.4 mm; N=95]. PMI1:[a=30mm; b=1mm; N=95].
PMI2:[a=4mm; b=0.2m N=62].
For Over-Saturation mode of operation, 𝐵𝑆𝑔 is equal to QSE and 𝐵𝑏𝑖𝑎𝑠 is equal to the desired
displacement of the zero flux axes. The QSE value is specific of each core material and PM
reluctance, 𝑅𝑃𝑀 , and it is independent of the PM’s strength or energy density product [3]. A
The last step is to define the nominal inductance, 𝐿𝑁 according to (11). This determines the
required PM reluctance, 𝑅𝑃𝑀 and accordingly its length, 𝑙𝑃𝑀 . If the resultant 𝑙𝑃𝑀 is too small for
practical uses, a smaller value of 𝑁, needs to be choose in the compromises taken after (13).
Once the basic material properties and dimensions, of both core and PMs, are defined, it is
possible to use small adjustments in the relative positions of PMs and cores, in order to fine tune
the desired LvsI profile. These adjustments are indicated as: distances a and b in Fig. 7.
The output filter, of a DC to DC converter, has been implemented using the Saturation-gap
biasing topology, with the design priority of minimal size. The complete filter, including two non-
30kHz.
The standard filter, STD and two different PMIs designs have been implemented, following the
procedure from section III. The first design, PMI1 uses a saturation-gap topology operating in
Full Linear Bias mode. The second design, PMI2 uses a saturation-gap topology operating in
Over-Saturation mode, concentrating the linear region, of the LvsI profile, to the operative
current range.
The basic profile of the three different filter inductors is depicted in Fig. 7. The specific design
The electric performance of the three implemented filters has been characterized using a Wayne
Kerr 3260B Magnetic Analyzer with WK3265B bias unit. The first test, performed on each filter
(including two inductors in series), measures the incremental Inductance, 𝐿 and the Equivalent
with a frequency of 30kHz. The test is repeated with different levels of DC current, from 0 to 20A.
The obtained LvsI and ESRvsI profiles of the three filters are presented in Fig. 8. (Top Left) and
(Top Right), respectively. A second test is been perform in order to see the losses behavior of
the filters as function of frequency. In Fig. 8. (Bottom Left) and (Bottom Right) it is presented the
Quality Factor, Q and the ESR as a function of frequency. The used stimulus is 500mV AC with
10A DC bias. The measurement is repeated for a set of stimuli frequencies ranging from 10kHz
to 50kHz.
Fig. 8. Inductor characterization measurements. Perform with WK3260B + WK3265 magnetic analyzer.
Measurements performed on the three complete filters (including the two inductors in series). Top: Inductance (L)
and Equivalent Series Resistance (ESR) as function of DC current (I); Stimuli: 500mV AC 300 kHz. Bottom:
Quality Factor (Q) and ESR as function of frequency; Stimuli: 500mV AC 10A DC.
Fig. 9. Inductance vs Current (LvsI) profile of three filter inductor pairs. Left: Standard Inductors (STD); Center:
Biased Inductors (PMI 1); Right: Over-saturated Biased Inductors (PMI 2). Black curves represent measured data.
Green curves represent simulated data with FEMM.
V. FEA SIMULATIONS
The three filter inductors have been simulated using FEA with the software FEMM. The used
models are shown in Fig. 7. The PMs and cores specifications in Table 1. have been used for
the FEMM models. The non-hysteretic BH curve of the core material is also required. The BH
loop, of the used ferrite cores, has been measured following the procedure presented in [20].
The non-hysteretic curve is calculated from the BH loop data, following the method presented in
effects, like eddy currents, are not predicted by the simulations. The static calculation of the
Fig. 9. shows the LvsI profiles, obtained from the simulations data, compared with the empirical
measurements of the three filters. The LvsI profiles, predicted with FEMM, present some
deviations from the measured profiles. These deviations are more sever in the PMI models. The
limitations of the 2D magneto-static analysis, makes the tool not suitable for accurate design. On
the other hand, the simulations provide a good tool for the analysis and understanding of the
Fig. 10. presents a sequence of simulation plots, from the two PMI models, at different currents.
The flux density, 𝐵 is represented with the color scale. The black lines are equipotential contours
of magnetic vector potential (or A-field) and must not to be mistaken with magnetic flux, ɸ. The
A-field contours represent the path of net magnetic flux, taking into account the magnetic
Fig. 10. Finite Element Analysis. Left: PMI1; Center: PMI2; Right: Representation of Generic Saturation-gap flux
behavior. FEMM: Color represent: Magnetic Flux Density; Black lines: Equipotential A-field contours (net magnetic
flux contours). Schematic representation (right): Green lines represent flux from PMs; Red lines represent flux
from the windings.
cancelations of the biasing effect. In the right column at Fig. 10. it is represented the generic
behavior of the magnetic flux interactions between the coils and the PMs. The flux, ɸ generated
by the PMs is represented in green. The flux, ɸ generated by the coils is represented in red.
Magnetic cancellations are represented by shading the individual flux lines. The behavior of the
At 0A, all the flux is introduced by the PMs. The flux distribution between 𝑅𝑏𝑖𝑎𝑠 and
𝑅𝑆𝑔 defines the operation mode. The PMs are at their maximum operation point,
close to 𝐵𝑟 . The portion of the core, delimiting 𝑅𝑆𝑔 , presents its shortest length.
As current increases, new magnetic flux is introduced in the core. The flux density at
𝑅𝑏𝑖𝑎𝑠 decreases, due to magnetic cancelations. The resulting net magnetic flux
paths, delimiting 𝑅𝑏𝑖𝑎𝑠 and 𝑅𝑆𝑔 is increasing the effective length of 𝑅𝑆𝑔 . Energy is
being stored in the PMs, lowering their operation point. The 𝑅𝑆𝑔 portion of the core is
At a current level of 6A in PMI1 and 9A in PMI2, the 𝑅𝑏𝑖𝑎𝑠 portion of the core
presents zero net magnetic flux. All the A-field contours are contained in 𝑅𝑆𝑔 .
As current increases further, more flux is introduced without cancellation effects. The
path of magnetic flux introduced by the coils can easily be recognized now. The flux
density in 𝑅𝑏𝑖𝑎𝑠 increases. The flux density in 𝑅𝑆𝑔 is, at some extent, equilibrated by
a decrease in the operation point of the PMs. The inductor reach its saturation
current, when the flux density at 𝑅𝑏𝑖𝑎𝑠 reaches the saturation region.
In order to avoid irreversible losses in the PMs, their operation point, at the maximum operating
current, must be above the 𝐵𝐾 value. Fig. 11. shows the flux density of the PMs at different
current levels. The maximum operating current is 10A. The PMs stay above 𝐵𝐾 even for currents
VI. CONCLUSIONS
The inductors of an output filter for a DC to DC converter have been implemented using the
Saturation-gap biasing topology. Two different PMI designs have been implemented targeting
maximum size reduction. The relevant analysis and design procedures have been presented.
The designs have been physically implemented and tested. The measurement results show a
significant size reduction while presenting the same inductance within the operating current
range. The PMI designs achieve 50% core volume reduction and 25% and 50% copper volume
reduction. The implemented PMI filters present higher losses than the standard filter. The use of
PMs with higher resistivity can reduce the total losses. On the other hand, the required PM area
cross-section will increase and over-saturated states like in PMI2, may not be possible.
Simulations based on 2D magneto-static FEA can provide a good analysis tool, but they do not
provide the required precision for design purposes. The development of more sophisticated 3D
models, and the design optimization, targeting minimum losses, will also be the focus of future
work.
VII. REFERENCES
[1] Andres Revilla Aguilar, and Stig Munk-Nielsen, “Method for introducing bias magnetization in ungaped cores:
The Saturation-Gap,” Applied Power Electronics Conference and Exposition (APEC) 2014, 29th Annual
[2] Andres Revilla Aguilar, Stig Munk-Nielsen, Marco Zuccherato and Hans-Jørgen Thougaard, “Size reduction
of a DC line choke using saturation gap and biasing with permanent magnets,” PCIM Europe 2014, pp 1667
[3] Andres Revilla Aguilar, and Stig Munk-Nielsen, “Design analysis and simulation of magnetic biased inductors
with Saturation-Gap,” Power Electronics and Applications (EPE'14-ECCE Europe), 2014 16th European
[4] J.T. Ludwig, “Design of optimum inductors using magnetically hard ferrites in combination with magnetically
soft materials,” Journal of Applied Physics. Volume 29, Issue 3, pp. 497-499. Mar 1958.
[5] J.T. Ludwig, “Inductors biased with permanent magnets. Part I: Theory and analysis,” American Institute of
Electrical Engineers, Part I: Communication and Electronics, Transactions of the. Volume 79, Issue 3, pp.
[6] J.T. Ludwig, “Inductors biased with permanent magnets. Part II: Design and synthesis,” American Institute of
Electrical Engineers, Part I: Communication and Electronics, Transactions of the. Volume 79, Issue 3, pp.
[7] J.T. Ludwig, “Inductors biased with permanent magnets,” Electrical Engineering, Volume 80, Issue 6, p 408.
June 1961.
[8] Akio Nakamura, and Junpei Ohta, “A new reverse-biased choke coil,” Proceedings of Powercon 9 C-5. TDK-
[9] Teruhiko Fujiwara, and Hatsuo Matsumoto, “A new downsized large current choke coil with magnet bias
method,” Telecommunications Energy Conference, 2003. INTELEC ´03. The 25th International. IEEE, pp.
[10] Rafal Wrobel, Neville McNeill, and Phil H. Mellor, “Design of a high-temperature pre-biased line choke for
power electronics applications,” Power Electronics Specialists Conference. PESC 2008. IEEE, pp 3171–
[11] Kuo Baoquan, Song Liwei, Zhang Qianfan, and Cheng Shukang, “The principle and design of the permanent
magnet bias DC reactor,” Proceedings of the Fifth International Conference on Electrical Machines and
[13] G.M. Shane, and S.D. Sudhoff, “Design and Optimization of Permanent Magnet Inductors,” Applied Power
Electronics Conference and Exposition (APEC) 2012, 27th Annual IEEE, pp 1770 – 1777. Feb 2012.
[14] G.M. Shane, and S.D. Sudhoff, “Design Paradigm for permanent-magnet-inductor-based power converters,”
IEEE Transactions on Energy Conversion, Vol 28, NO 4, pp 880 – 893. December 2013.
[15] Zhigang Dang and Jaber A. Abu Qahouq, “Permanent magnet toroid power inductor with increased
saturation current,” Applied Power Electronics Conference and Exposition (APEC) 2013, 28th Annual IEEE,
[16] ABB, “Reducing AFD-caused harmonics at partial load conditions.” TECH Notes ABB. 2005.
[17] C.M. Andrews, “Understanding permanent magnets,” TECH Notes Group Arnold.
[18] H. Lovatt and P. Watterson, “Energy stored in permanent magnets,” IEEE Transactions on Magnetics, 35(1)
[19] David Meeker, “Magnetic Circuit Derivation of Energy Stored in a Permanent Magnet,” FEMM online
documentation, http://www.femm.info/wiki/PMEnergy.
[20] W.L.Soong, “BH curve and iron loss measurements for magnetic materials,” Power Engineering Briefing
[21] Stig Munk-Nielsen and Hans Christiansen, “A simple and accurate non-linear transformer model including