Size Reduction of a DC link Choke Using Saturation-gap

and Biasing with Permanent Magnets

Andrés, Revilla Aguilar, Aalborg University, Denmark, [email protected]
Stig, Munk-Nielsen, Aalborg University, Denmark, [email protected]
Marco, Zuccherato, Danfoss Power Electronics, Denmark, [email protected]
Hans-Jørgen, Thougaard, Sintex, Denmark, [email protected]

Abstract
This document describes the design procedure of permanent magnet biased DC inductors using the
Saturation-gap technique [1]. This biasing configuration can provide a 50% reduction in either the core
volume or the number of turns, while meeting its current and inductance requirements. A design
example on a 20A 1mH iron laminations DC choke is presented. MEC simulations and empirical
measurements on a physical unit were performed. Experimental results show a 50% reduction of the
core volume is achieved. The total volume reduction of the inductor is 37%.

1. Introduction
The use of permanent magnets for introducing bias magnetization in DC inductors is a known
technique for extending the saturation current limit and accordingly reducing the required inductor´s
size. The first designs of hybrid inductors presented the magnet inserted in the air-gap [2][3]. This
configuration can introduce a certain amount of bias flux; on the other hand the magnet size is limited
by the air-gap size, presenting limitations on the inductor design. More recent designs uses
nonstandard core shapes with magnets in the vicinity of the air-gaps [4][5]. These configurations can
achieve a higher level of bias flux; on the other hand the requirement of specially design core shapes
may be an important disadvantage.
The saturation-gap [1] configuration can be implemented with standard non-gaped UI iron laminations
and can achieve 100% bias flux. Fig.1. shows the basic core and magnets arrangement of the
saturation-gap configuration. The red vectors represents the flux from the coil, green vectors
represents flux from the magnets. The function of the permanent magnets in this configuration is a
twofold: to produce saturation in a localized segment of the core which will behave as a variable virtual
air gap and to create a biasing flux in the rest of the core.

Fig. 1. . Saturation-gap biasing configuration. Red vectors represent flux produced by the coils. Green vectors
represent flux produced by the magnets. Vectors in opposite direction represents biasing. Vectors in the same
direction represents saturation-gap.
The following section II will present an inductor design strategy based on the saturation-gap
configuration. A design example of 50% core reduction on a standard choke will also be presented.
Section III will present Magnetic Equivalent Circuit (MEC) simulation of the proposed solution. Section
IV will present the experimental analysis measured on the physically build hybrid and the standard
chokes. Finally section V will present the conclusions and further work.

2. Design Strategy
Any given standard EI (or EE) inductor design can be improved by using two UI (or UU) cores in
saturation-gap configuration. This section describes the generic procedure for reducing the size and/or
the copper losses of a given DC inductor, using the saturation-gap biasing configuration. A specific
design example targeting 50% core size reduction of an EI66 1mH 20A DC link choke is also
presented.

2.1. Introduction to Saturation-gap Inductor Design

The basic formula used for inductor size calculations is:

L I = N Bsat Ac
Where L is inductance, I is the current, N is the number of turns, Ac is the area cross section of the
core and Bsat is the saturation flux density of the material. In the following, when no additional sub-
indexes are attached to these variables, they will refer to the standardised EI core parameters. When
referring to the variables of either the hybrid inductor or the individual UI cores, the h and UI sub-
indexes will be used.
In the left hand of the equation we have the variables, directly related to the target inductor
specifications. In the right hand of the equation we have the parameters related to the choke geometry
and material characteristics. The product of these parameters will represent a constant. This design
constant will define the maximum LI product achievable of each inductor design. The specific nominal
inductance LN is adjusted by the length and shape of the air-gap. The blue line in Fig. 2. represents
the LI product region of a generic unity value design constant. Constant length air-gaps produce
constant LN as a function of current until crossing the LI product region. At this point it is defined the
saturation current ISat of the inductor. The use of stepped or non-constant length air-gaps will produce
nonlinear inductance profiles. This can maximize the working region of an inductor within a given LI
design constant.

Fig. 2. . Inductor’s working range as a function of gap dimensions. Blue curve is a unity LI product region. Orange
curve represents the inductance versus current profile for each air-gap. Left: Short air-gap; Middle: Long air-gap;
Right: Nonlinear gap.
The saturation-gap configuration, presented in Fig. 1, does not use physical air-gaps. The equivalent
reluctance of an air-gap is induced in a segment of each core, by means of partial saturation. The
length and shape of the induced equivalent air-gap is a complex function involving: the core saturation
profile; the instantaneous flux levels produced by coils and magnets; and the magnets and cores
relative geometry.
Permanent magnets behaviour can be simulated as a constant mmf source with a series internal
reluctance. The instantaneous flux produced by the magnet is inversely proportional to the reluctance
of its magnetic path. The minimum and maximum flux levels produced by a magnet are defined by its
length ML and area cross-section AcM respectively [6]. The flux level produced by the permanent
magnets in the present configuration, is continuously adapting to the variations induced by the coil’s
flux, maintaining a Quasi-Saturated Equilibrium (QSE) in the saturation-gap region.
For the biasing of laminated iron cores a high flux density permanent magnet material needs to be
used. The chosen material is Neodymium Iron Boron, NdFeB with Br = 1.2 T. As a rough
approximation we can suppose the maximum flux density Br of the NdFeB magnets, to be equivalent
to the saturation level of the laminated iron. In order to fully bias the inductor and to bring the
saturation-gap region to an almost-saturated state, the permanent magnet needs to provide enough
flux for the two flux paths (the saturation-gap path and the biasing path). Accordingly, the area cross-
section of the magnets AcM, needs to be at least two times the area cross-section of the core Ac.
AcM / Ac > 2
The AcM / Ac ratio of standard UI laminations is equal to 2.5. By using UU laminations the ratio can be
increased up to 4. Since the hybrid inductor is composed of two UI cores connected in series, the
inductance requirement for each UI core is half of the standard inductor. That provides and extra
reduction factor of 2 in the design of the UI cores. Both UI cores will contribute equally to the total
volume of the inductor. Accordingly this extra reduction factor is applied to the required AcUI in order to
maintain the volume proportion of the total hybrid core.
AcUI = Ach / 2
By reducing AcUI, the required winding turn length is also reduced. If the applied reduction factor is
perform with equal rectangular proportion RAc, the wire length is been reduced by the same proportion.
The perfect square proportion (RAc=1) will provide the shortest wire length. Assuming full bias is
achieved, we can extend the max flux density of the hybrid inductor to double the B sat value of the
material. Accordingly, we have the possibility of reducing Nh or Ach by a factor of 2, while meeting the
same current and inductance requirements.
L I = Lh Ih = Nh Ach 2Bsat
Nh Ach = N Ac / 2
The focus of the specific design strategy should be used in order to decide the distribution of the
reduction factor. Applications using a high DC current could benefit from a reduction in copper
resistance and increase the efficiency. On the other hand, many designs require smaller and lighter
cores.
Using the entire reduction factor for the area cross-section Ac will allow for the use of the smallest core
and decrease the component weight. On the other hand, the required window area, Wa will remain the
same. In this situation is possible to use the equivalent Wa size of UI laminations (2xUI = EI), and
reduce the laminations stack length, lstack by half. This procedure is illustrated in Fig. 3. Another
possible procedure could be to replace one bigger EI core with two smaller UU cores.

Fig. 3. Half core volume reduction strategy with saturation-gap hybrid configuration.
Using the entire reduction factor for the number of turns N will decrease the wire resistance and the
required window area by a factor of two, allowing for core size reduction or bigger wire cross-section
for additional reduction of copper resistance. Since the required window area becomes smaller while
the area cross-section Ac must remain the same, this strategy can be used to replace one bigger EE
core with two smaller UI cores.
A compromised distribution of the reduction factor, between Ac and N, is also possible and needed to
a certain extent, in order to accommodate to the finite possible standardised lamination sizes .

2.2. Practical Design Example with Saturation-gap

The target inductor rating parameters of our standard choke are: I=20A, L=1mH, lstack = 34mm,
2
accordingly the area cross-section is: Ac = 34 x 22 = 748 mm . The number of turns is N = 36. The
2
wire diameter is ø = 2.24 mm. Gap length lgap = 1.2 mm. The area used by the coil is Cwa = 140 mm
2
and the total window area Wa = 363 mm . The packing factor in the window area is k = 0.4.
Since the present design strategy is focused on core size reduction, it is desired to use the entire
reduction factor for the cross-sectional area of the UI cores. The following formulas describe the
design requirements for each UI core:
NUI = N
AcUI = Ac / 4
Since the required number of turns, and accordingly the window area did not change, it is
recommendable choosing the UI core equivalent of half the target standard EI core (2 x UI33 = EI66).
By using a stack length, lstack equal to half of that in the standard, we achieve the required ¼ reduction
of AcUI, while keeping the same RAc and accordingly the total wire length. This design will present 50%
core volume while keeping equal wire length and resistance. Unfortunately, UI33 laminations are not a
standardized size. The closer standard UI core size found is UI30. The Wa of UI30 is slightly smaller
than the standard EI66 and the same coil could fit by a small increase in the packing factor. In order to
achieve the required AcUI with UI30 laminations the minimum required stack thickness is calculated.
Accordingly to the previous calculation, the final UI cores specifications are: NUI = 36 turns; Ø = 2.24
mm; lstackUI = 20mm. The magnet specifications: ML = 3mm; AcM = 20x25mm covering all the area
between UI cores. The total volume of the hybrid core (including the two UI cores) has decreased
50%. The achieved rectangular proportion of the windings is higher than desired, RAcUI = 2, and a
small increase in wire resistance is expected. By comparing the Ac perimeter length of the standard
and the hybrid designs, the expected increase in wire resistance is approximately 18%.

3. Simulation Analysis
In order to verify the proposed design and to further investigate the behavior of the saturation-gap
configuration a Magnetic Equivalent Circuit (MEC) model of both inductors was created using LTspice.
Fig. 4. shows the MEC models of both the hybrid and the standard inductors. Each core is divided into
segments. The UI segments connected to the magnets are subdivided in 9 segments in series, in
order to provide a higher resolution in the saturation-gap area. The magnets are modeled as an Fmag
(voltage) source with series reluctance Rmag. A parallel reluctance Rfrg is inserted to account for
fringing flux [4]. The total surface of each magnet, Ma is simulated with four smaller magnets with ¼ of

Fig. 4. Magnetic Equivalent Circuit (MEC) models and simulation results. Inductance versus current.
the original Ma, in order to match the present segment resolution. The reluctance of each core
segment is simulated by a non-linear resistor. The values of these non-linear resistor’s are introduced
as V vs I look-up tables. The actual values were calculated using the segment dimensions and the B
vs H curve found on the core specifications. Each coil is represented as a magnetic flux (current)
source which is related to the electrical winding of the models as in [8]. The values of the different
MEC elements were calculated using the design’s geometrical specifications presented in section II,
and applying the common simulation formulas as in [1][8].
The models were tested using a 20V DC pulse stimulus of 3 ms duration. The produced current and
voltage waveforms were computed and the inductance value as a function of the current was
calculated. The obtained results from the hybrid and standard inductors are shown next to their
respective models in Fig. 4.

Fig. 5. MEC simulation: Voltage pulse test instantaneous signals. Top: Current and voltage at electric terminals;
Middle: Magnetic flux present at coil segment (red), saturation-gap segment (purple) and at the total magnet area
(green); Bottom: Magnetic flux present at each magnet segment ( ɸm1 is the magnet segment closest to center).
The standard choke model presents a characteristic inductance profile, having a nominal inductance,
Ln = 1 mH. The saturation knee point appears at Isat = 30 A, approximately. The hybrid choke model
presents a nonlinear inductance profile. The saturation point and the inductance value above 20A, is
perfectly matching the standard choke profile. The inductance at zero current is two times the nominal
value and decreases exponentially until reaching Ln = 1 mH at 20A. This nonlinear inductor behavior is
due to the variations of the equivalent virtual air-gap produced in the saturation-gap region.
In order to further investigate the behavior of the saturation-gap and to identify the possible tuning
mechanisms, the different flux signals in the model were analyzed. Fig. 5. shows the voltage pulse
stimulus and the obtained current ramp at the electrical terminals of the hybrid model. Additionally, the
instantaneous flux, ɸ[Wb] at the key points of the MEC are presented. At the top plot of Fig. 5.
different current behavior zones (Z1 to Z4) are identified and delimited along the voltage pulse
duration. The different zones are representative of different flux behavior within the saturation-gap and
they are delimited by different current threshold values:
 Z1. Nonlinear inductor behavior: The flux level, present at the saturation-gap, at 0A current,
has not reached the Quasi-Saturated Equilibrium (QSE) point. The initial equivalent air-gap is
very short and exponentially increases as the flux from the coil is adding to the saturation-gap
flux level. Accordingly, an exponential decrease in inductance is achieved.
 Z2. Linear inductor behaviour: The flux level, present at the saturation-gap has reached the
QSE point. A linear increase of the coil’s flux is equilibrated by a linear decrease of the
magnet’s flux, maintaining the QSE point at the saturation-gap center. The starting current
threshold of this zone is controllable by the amount of magnet’s flux. A bigger magnet could
introduce the necessary flux to reach the QSE from the 0A current point.
 Z3. Core saturation: The saturation current, Isat delimits the onset of this zone. The saturation-
gap flux is constant at QSE and the behavior of the coil and magnet fluxes is identical as in Z2.
On the other hand, the core material has reached saturation and the inductance value
decreases rapidly.
 Z4. Demagnetization threshold: The flux produced by the magnet segment closer to the edge
(ɸm4) has reached zero value. No flux is produced by this section of the magnet. As current
increases, flux from the coil crosses through the magnet ending up in demagnetization effects.
The desired working point in most applications would be Z2. Nonlinear behavior zones like Z1 are in
principle not desired. On the other hand, several applications using standard nonlinear-gap inductors
could benefit from this configuration. The Core saturation threshold of the present model appears at
30 A. The Demagnetization threshold is not observed until reaching 60 A current. Accordingly, the
magnets will not suffer demagnetization in the working current range of the hybrid inductor (Z1 to Z2).

4. Experimental Analysis
The standard and the hybrid design, presented in the previous sections, were physically implemented
and compared using two different experimental tests. The first test consisted in generic inductor
characterization measurements, using a Wayne Kerr magnetics analyzer. In the second test, the units
were analyzed at higher power using a Spitzenberger Grid Simulator. The units were driven at nominal
AC and DC Bias level, simulating the working conditions of DC link in a frequency converter .

Fig. 6. Inductance versus DC bias current (top) and Equivalent Series Resistance (ESR) versus DC bias current
(bottom). Measured with WK3260B magnetic analyser and WK3265B bias units. Stimulus: 1Vrms AC 300Hz.

4.1. Inductor Characterization Measurements

A Wayne Kerr 3260B Magnetics Analyser with two WK3265B bias units was used to characterise the
two inductor designs. The analyser uses a low power AC voltage stimulus for measurement
calculations. The two WK3265B bias units can provide adjustable DC bias current levels up to 50 A.
The inductance and the ESR (Equivalent Series Resistance) as a function of the DC bias current were
measured for both inductors. Fig. 6. shows the obtained measurement results.
The Inductance versus DC Current Bias graph shows the comparison between the inductance value
of the standard and hybrid solutions as a function of the applied DC current. The standard choke
exhibits the typical inductance vs current profile of a gapped laminated steel inductor. The inductance
value is maximal at the lower bias current and slowly decreases along the working current range of the
inductor. At current levels higher than 28A, the inductance value is rapidly decreasing due to
saturation of the core material. The hybrid choke replicates quite well the standard choke behaviour
over all the DC current bias range, but shows a slightly lower inductance value for almost all the
working points. It is also noticeable how the inductance value of the hybrid choke is initially increasing
in the lower bias current range, reaching its maximum value at 4 A. This is contrary to the expectations
of the MEC analysis. This effect is an indication of the permanent magnet biasing flux is reaching the
negative saturation level of the core (Z0).
The ESR versus DC Current Bias graph shows the comparison between the ESR value of the
standard and hybrid solutions as a function of the applied DC current. The measured ESR value of the
choke under test represents the total power losses in the component, seen as the sum of copper
losses and core losses. The measured DC winding resistance of the standard and hybrid chokes are
27 mΩ and 35 mΩ respectively. The higher copper resistance of the hybrid inductor is expected
according to the design calculations provided in section II. The standard choke exhibits a flat ESR
curve up to 28 A DC bias current. This is the working region of the core and losses are almost
constant. At 28 A DC bias current, the laminated core material starts saturating and gradually reduces
the contribution of the core losses in the choke, leading to decreasing values of the measured ESR.
The hybrid choke presents an unexpected loss profile. The ESR value is maximal at low DC current
levels, and decreases constantly as the DC current increases. At the nominal DC current level (20 A)
the ESR value of the hybrid inductor is 20% higher than that of the standard choke and at the
saturation point (28 A) the ESR value is only a 10% higher .

4.2. Nominal Power AC +DC testing

In order to verify the actual behavior at nominal power rating, the hybrid and the standard chokes were
tested using a Spitzenberger Grid Simulator EMV D 45000/PAS. A 20A DC bias current with a
sinusoidal component of 7.5Apeak was applied to each inductor in order to simulate the approximate
conditions of a DC link choke operating in a frequency converter. Voltage and current waveform were
measured and the losses were calculated. The following Table. 1. summarizes the measured data.

Nominal Power Testing

Device Total Power Losses Copper Losses Core Losses

Standard 17.24 W 13.82 W 3.42 W

Hybrid 22.04 W 18.51 W 3.53 W

Hybrid losses increment 28% 34% 3%

Table. 1. Power losses at nominal power: 20A DC bias + 7.5A peak AC. Tested with Spitzenberger Grid Simulator
EMV D 45000/PAS.
The total power losses at the nominal ratings are clearly dominated by the copper losses in both
inductors. The hybrid presents 28% more total losses than the standard choke. This increment in the
total losses of the hybrid is also dominated by the increase in copper losses due to the increased
length and resistance of the windings.

5. Conclusions and Further Work

The saturation-gap design strategy is been presented. 50% reduction factor of the core or the winding
turns is achievable with this hybrid configuration. The proposed hybrid design was simulated and
physically implemented. Both, hybrid and standard inductors were empirically tested. Measurement
results show a size reduction of 50% of core volume is achieved with a small increase in losses. This
increase in losses is mostly dominated by the winding resistance and it is a particularity of the design
example.
MEC simulations can provide a useful view for the understanding of the saturation-gap behaviour and
possibilities. On the other hand, a Finite Element Modelling, (FEM) will provide higher accuracy and
the magnet tuning could be tested at the design stage.
The specific core loss mechanisms acting on the hybrid inductor are not fully identified. Core losses in
standard inductors can be classified in two different types: eddy currents and hysteresis [9]. Eddy
currents are produced in the magnetic core material, induced by the variations in flux. Additional eddy
currents induced in the magnet are not initially expected, since the magnets are outside the coil’s flux
path. On the other hand, it is known that the flux level produced by a permanent magnet is dependent
on the internal reluctance of the magnet and the reluctance of its magnetic path [6]. Accordingly, the
permanent magnet is expected to constantly vary its flux level, following the reluctance variations in
the core, inducing eddy currents within the magnet material.
Hysteresis losses are produced by friction, magnetostriction and delays in the rotation of dipoles and
domains within the crystalline structure of the core [9]. The presence of the permanent magnet flux will
introduce an additional tension in the domain alignment reaching an equilibrium point different from
that of the core material alone. This mechanism could increase or even decrease the hysteresis loses
of a given core.
A deeper analysis and experimentation dedicated to the understanding and minimization of the loss
mechanism in the hybrid configuration will be the focus of further work .

6. References
[1] Andres Revilla Aguilar and Stig Munk-Nielsen. ”Method for introducing bias magnetization
in ungapped cores: The Saturation-gap,” APEC 2014. Unpublished.
[2] 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. 23 October 2003. IEEE, pp. 416-420.
[3] 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, 2008. PESC 2008. IEEE, pp 3171 – 3177.
[4] G.M. Shane and S.D. Sudhoff. “Permanent Magnet Inductor Design,”. Electric Ship
Technologies Symposium, 2011 IEEE, pp. 330-333, April 10-13, 2011.
[5] G.M. Shane and S.D. Sudhoff. “Design and Optimization of Permanent Magnet
Inductors,”. Applied Power Electronics Conference and Exposition (APEC), 27th Annual
IEEE, pp 1770 – 1777.
[6] C.M. Andrews. “Understanding Permanent Magnets,”. TECH Notes Group Arnold.
[7] W.L. Soong. “BH Curve and Iron Loss Measurements for Magnetic Materials,”. Power
Engineering Briefing Note Series. PEBN 5. 12 May 2008
[8] Stig Munk-Nielsen. “Experience with Spice teaching power electronics,”. Power
Electronics and Applications, 2009. EPE ´09. 13th European Conference. IEEE, pp 1 – 8.
[9] John B. Goodenough. “Summary of Losses in Magnetic Materials”. IEEE Transactions on
Magnetics. Vol. 38 NO 5 September 2002.