Po w e r El e c t r o n i c s Te c h n o l o g y Oc t o b e r 2 0 0 3 w w w. p o w e r e l e c t r o n i c s .

c o m 14
The well-known problem of defining the
minimum induct ance value for a t arget
roll-off and dc current is addressed in the
frame of a hysteresis model.
By M MM MM a u r a u r a u r a u r a u r i ci o E i ci o E i ci o E i ci o E i ci o Esg u er sg u er sg u er sg u er sg u err rr rr a aa aa , Ferri t e Di vi si on, EPCOS AG, Muni ch, Ger many
T
he widespread use of gapped ferrite cores for
inductive components subject to a superposed
dc cur rent poses a challenge to the way the
required properties are specified. The impor-
tance of attaining a precise definition of tar-
get values is dr iven by the need of miniaturization, on one
hand, and the slow progress of material performance, on
the other. This is because incrementing key material prop-
erties requires advancement in both material composition
and processing in to push the limits.
The conventional way to formulate a dc-bias specifica-
tion for a gapped core based on measurements shown in
Fig. 1.
1) A target roll-off in the range of 10% to 30% is linked
to a dc flux density as estimated from the dc current and the
gapping factor b, defined by the reciprocal difference be-
tween the effective and the initial permeabilities: (Eq. 1, 2)
B
NI
l
e

1 1
e i e
g
l
N stands for the number of turns, I for the current, le
for the effective magnetic length of the core and g for the
air gap. Note that the validit y of Equation 1 improves with
increasing air gaps.
2) It is t hen ver ified if t he result ing value is dist ant
enough, e.g. 20%, from the material saturation figure as
judged by empirical data. The roll-off referred to the lower
tolerance limit of the inductance minus a safet y margin is
then assigned to this cur rent.
The above described semi-quantitative procedure can
be improved by taking into account the material behavior
under dc-bias. To this end, a model of the nonlinear mag-
net ic response descr ibing t he permeabilit y roll-off as a
Po w e r El e c t r o n i c s Te c h n o l o g y Oc t o b e r 2 0 0 3 w w w. p o w e r e l e c t r o n i c s . c o m 16
DC-BIAS SPECIFICATIONS
function of flux densit y is needed. One possibility is the
classical model by Gans [1], which describes a universal
function that depends on initial permeabilit y and satura-
tion. However, only a certain type of materials follows the
Gans cur ve because features such as the squareness of the
hysteresis loop (which is reflected in the inductance vs.
current cur ve) are missing. Therefore, a closer representa-
tion has to star t from hysteresis modeling.
Hysteresis Model and Reversible Permeability
The model proposed by the author [2] consists of solv-
ing the Hodgdon [3] hysteresis differential equation in the
formulation H(B) with the major loop, as a particular solu-
tion. For the branches of the major loop a heuristic descrip-
tion has been found, which fits the lower H
L
(B) and upper
H
U
(B) branches for soft magnetic materials: (Eq. 3a, 3b)
H B
B
B
B
H
L
c
s
a c
( )

|
.

`
,

+

0
1
1

H B
B
B
B
H
U
c
s
b c
( )

|
.

`
,

0
1
1

The fit parameters are besides the saturation B

S
, coer-
cive field H
c
and permeability at the coercive field
c
, the
two squareness exponents a and b. The reversible perme-
ability can be derived from the equations for asymmetric
minor loops in the limit
0

B
and for a ~ b, valid for
most power ferrites as follows: (Eq. 4)

rev dc
dc
s
a
dc
s
a
c
B
a
B
B
B
B
( )
+
|
.

`
,

|
.

`
,

]
]
]
]
+
1 1
1
1
2
( )

11
1 2 1
1 1

|
.

`
,

|
.

`
,

]
]
]
]

|
.

`
,

B
B
B
B
dc
s
dc
s
a B
i c
o s

||
.

`
,

1
To compare the model with measured cur ves, a rela-
tionship between the non-directly measurable dc flux den-
sit y B
dc
and the ap-
plied dc field H
dc
is
necessar y: (Eq. 5)
H
B
B
B
dc
c
dc
dc
s
a

|
.

`
,

1
1
0

These equations
yield cur ves in ex-
cellent agr eement
with measured val-
ues for a var iet y of
soft ferrite materi-
als [4].
Roll-off at
Saturation
With the help of
Equ at i on 4, t h e
permeabilit y vs. dc
field for cores with
air gap can be cal-
culated. For the in-
ductance, the effec-
tive reversible per-
meabilit y follows
fr om Equat ions 2
and 4: (Eq. 6)

rev
rev
e
+
1
and for the effective dc field: (Eq. 7)
H H B
dc dc dc
e
+

0
The above equations can then be applied to answer the
following question: If the shearing line described by Equa-
tion 1 is prolonged to cross the saturation cur ve, which roll-
off corresponds to the resulting value for the effective field
(dubbed H
s
for saturation field, Fig. 2)? The result depicted
in Fig. 3 clearly shows that for sufficient large gaps, such that
roughly [4] (Eq. 8)

e
i
<
20
the roll-off is 50%. This can be proofed both theoreti-
cally [4] and by measurements on a gapped toroids (Fig. 4:
R9.5 in material T38, i=10000, e=270) for which the dc
current has been rescaled by a factor (Eq. 9)
I
N
l
dc
e

0
following Equation 2 and 7, yielding values in Tesla. The
saturat ion value can then be read at the 50% saturat ion
yielding 410 mT in excellent agreement with the directly
measured value. Of course, values higher than the satura-
tion do not have any physical meaning.
Idc [A]
L

[
m
H
]
L
nom
Is Iset
L
min
F FF FFi g i g i g i g i g. .. .. 1. 1. 1. 1. 1. Convent i onal defi ni t i on of a dc-bi as speci fi cat i on: t he set t i ng
current I
set
i s defi ned wi t h an empi ri cal di st ance from t he current
correspondi ng t o t he mat eri al sat urat i on I
s
accordi ng t o Eq. 1. The
speci fi ed i nduct ance l i mi t L
min
corresponds t o t he rol l -off for t he l ower
t ol erance wi t h a saf et y margi n.
0 1000 2000 3000 4000 5000 6000
0
0.5
Hdc [A/m]
B

[
m
T
]
Bs
H
s
0
0
120

r
e
v
H
s
1000 2000 3000 4000 5000 6000
Hdc [A/m]
0.4
0.3
0.2
0.1
100
80
60
40
20
?
F FF FFi g i g i g i g i g. .. .. 2. 2. 2. 2. 2. The sat urat i on fi el d st rengt h H
s
,
defi ned i n t he upper graph as t he cross poi nt
of t he sheari ng l i ne (bl ack) and t he mat eri al
sat urat i on, i s shown i n t he l ower graph t o
correspond t o a rol l -off val ue t o be det ermi ned.
~
Po w e r El e c t r o n i c s Te c h n o l o g y Oc t o b e r 2 0 0 3 w w w. p o w e r e l e c t r o n i c s . c o m 18
When applying the above proce-
dure to core shapes other than rings,
a distribution of cross-sectional areas
has to be t aken into account . This
means t hat t he cor responding flux
densit y distribution of the core has to
be calculated section by section. This
is shown for a highly inhomogeneous
shape, such as EP13 in Fig. 5, along
with a rescaling of the current by the
0.00 0.5 1.0 2.0
0
300

e
I
dc
[A]
0.0 0.410 0.6
1.0
250
200
150
100
50
1.5
0.5
0.0
I
dc
*m
o
*N/I
e
(1/m
e
-1/m
i
) [T]
m
r
e
v
/
m
e
-
1
F FF FFi g i g i g i g i g. .. .. 4. 4. 4. 4. 4. Permeabi l i t y vs. dc-bi as measurement for gapped R9.5 cores i n t he hi ghperm
mat eri al T38 at T=21C (upper graph). The rescal i ng of t he current axi s (Eq. 8) yi el ds a
mat eri al sat urat i on val ue at t he 50% rol l -off of 410 mT.
DC-BIAS SPECIFICATIONS
2000 4000 6000 8000 110
4
-100
-90
-80
-70
-60
-50
-40
-30
-20
-10
0
i
R
o
l
l
-
o
f
f

@

H
=
H
s
e=200
e=50
F FF FFi g i g i g i g i g. .. .. 3. 3. 3. 3. 3. Cal cul at ed rol l -off val ue for t he
sat urat i on fi el d H
s
as a funct i on of i ni t i al
permeabi l i t y for gi ven effect i ve permeabi l i t y
val ues
e
= 200 and 50.
minimum cross-sectional area, which
represents the bottleneck in terms of
saturation. (Eq. 10)
I
N
l
A
A
dc
e
e

0
min
Remar kably, t he cor rected EP13
cur ves and the r ing core coincide ex-
actly at the 50% roll-off, yielding the
same material saturat ion figure.
Distance to Saturation
The results presented so far show
that the dc bias can be precisely de-
fined relative to the saturation point.
With Equations 4 and 6, cur ves for the
roll- off vs. t he r elat ive dist ance t o
sat ur at ion have been calculated for
t ypical material parameters at differ-
ent temperatures, as shown in Fig. 6.
Wit h t he help of t hese cur ves, t he
choice of roll-off can be converted via
t he r elat ive dist an ce in t o a cor r e-
Po w e r El e c t r o n i c s Te c h n o l o g y Oc t o b e r 2 0 0 3 w w w. p o w e r e l e c t r o n i c s . c o m 20
the selection of A
L
value and of the
number of turns is a degree of free-
F FF FFi g i g i g i g i g. .. .. 6. 6. 6. 6. 6. Di st ance t o sat urat i on for N87 mat erial
as a funct i on of rol l -off (for an RM8 wi t h

e
=75.6, see exampl e) and as a funct i on of
e
for a rol l -off of 20% at T= 25C and T=100C.
DC-BIAS SPECIFICATIONS
sp on d i n g cu r r en t , based on t h e
kn owled ge of mat er i al an d cor e
par ameters.
A
L
-Tolerances
When considering the effect of A
L
-
tolerances, it is important to first note
that based on a stat ist ical toler ance
simulat ion analysis, the toler ance is
propor t ional to t he A
L
-value it self:
(Eq. 11)
A A Tol
L L
nom
t ( ) 1
(Eq. 12)
Tol
C
l
A sym l
A
pk
o
e
e i
g
e
L
i
nom

|
.

`
,

+
|
.

`
,

6
1
2
2 2

The constant involves the standard

deviation for the permeabilit y
i
and
the air gap
g
,

a factor sym for sym-
met r i c ( = 1) or asymmet r i c ( = 2)
grinding as well as the target Cpk-
value. The above equation gives a gen-
erally valid relationship to be consid-
ered for designs. However, concrete
values will change fr om vendor t o
vendor, even though an IEC-standard
for A
L
values and its tolerances is in
prepar at ion.
Dc-bias Specification
Since the target of an application
i s t o h ave a cer t ai n i n du ct an ce:
(Eq. 13)
L A N
A
l
N
L e
e
e

2
0
2

10 12 14 16 18 20 22 24 26 28 30
0
5
10
15
20
Roll-Off [%]
D
i
s
t
a
n
c
e

t
o

S
a
t
u
r
a
t
i
o
n

[
%
]
20 30 40 50 60 70 80 90 100
0
4
8
12
16
20
e
D
i
s
t
a
n
c
e

t
o

S
a
t
u
r
a
t
i
o
n

[
%
]
25C
100C
25C
100C
0 0.1 0.2 0.3 0.4 0.5
0
20
40
60
80
100
120
Ring core calculation
EP13 calculation
EP13 measurement
Idc*(N/le)*(o/)*(Ae/Amin) [T]
A
L

[
n
H
]
50
F FF FFi g i g i g i g i g. .. .. 5. 5. 5. 5. 5. A
L
-val ue vs. dc-bi as measurement for
gapped EP13 cores i n t he hi ghsat mat eri al
N45 at T=25C wi t h A
L
= 100 nH.
Po w e r El e c t r o n i c s Te c h n o l o g y Oc t o b e r 2 0 0 3 w w w. p o w e r e l e c t r o n i c s . c o m 22
DC-BIAS SPECIFICATIONS
Idc [A]
L

[
m
H
]
L
nom Is Iset
L
min
F FF FFi g i g i g i g i g. .. .. 7. 7. 7. 7. 7. Speci fi cat i on exampl e for an RM8 core i n N87 wi t h A
L
=160 nH
+/ -3% at T=25C. The red area corresponds t o t he set current and l i mi t
i nduct ance as per t he proposed procedure whi l e t he bl ack corresponds
t o a convent i onal speci fi cat i on.
dom. In the frame of the present discussion, the upper tol-
erance limits the maximum attainable dc current and needs
to be kept as small as possible. Therefore, the following
specification procedure is recommended (Fig. 7):
1) The target roll-off RO with respect to the nominal
in duct an ce is defin ed as t he lower in duct an ce limit :
(Eq. 14)
L L RO
nom min
( ) 1
2) The current corresponding to the target roll-off re-
ferred to the upper tolerance (of
e
) is the setting param-
eter: (Eq. 15)
I B
l
N
A
A
Tol
DTS
set s
e
e
e i

+

|
.

`
,

min
( )
( )
1
1
1
1
0

The above procedure provides a valid specificat ion for
all possible A
L
values within the limits, provided the toler-
ance range is smaller than the roll-off: (Eq. 16)
2Tol RO <
Otherwise, the inductance vs. dc-current cur ve corre-
sponding to the lower A
L
-tolerance limit would lie outside
the minimum inductance value defined by Equation 13.
Temperature Effects
The material par ameters of ferrite materials are known
to be strong temperature-dependent. To adapt the above
results from a given temperature, i.e. 25C, the following
parameters need to be corrected in their temperature de-
pendence:
B
s
: Look up value in material data sheet; the slope of the
temperature cur ve is in the range of 0.8 ... 1.5 mT/C for
power grades and 1.6...-2.3 mT/ C for highperm grades

e
: Since t he air gap and hence r emain const ant
t hr oughout t he t emper at ur e, t he value at a differ en t
Po w e r El e c t r o n i c s Te c h n o l o g y Oc t o b e r 2 0 0 3 w w w. p o w e r e l e c t r o n i c s . c o m 24
DC-BIAS SPECIFICATIONS
Core: RM8
A
e
=64 mm, l
e
=38 mm, A
min
=55 mm
Mat erial: N87

i
(25C)=2200,
i
(100C)=4000
3
)B
s
(25C)=465 mT,
3
)B
s
(100C)=370 mT
a(25C)=2.9, a(100C)=5.1

c
(25C)=5500,
c
(100C)=4300
H
c
(25C)=21 A/ m, H
c
(100C)=13 A/ m
Induct ance: L=1.3 mH
A
L
=160 nH +/ - 3% (
e
=75.6) and N=90
Roll-off : RO=20%
Result :
1) From Eq. 14 follows
L mH
min
. 1 04
2) From Fi g. 6:
DTS C I C A
set
( ) % ( ) . 25 12 25 1 47
3) For T
2
=100C wit h DTS (100C)=8%:
B C
B T
DTS C
DTS C
s
s
( )
( )
. ;
( )
( )
.
100
0 796
1 100
1 25
1 042
1

I
set
(1 000 1 27 C A ) .
Conclusion
A well-defined dc-bias specification for ferrite cores with
a sufficiently large gap star ts from the saturation proper-
ties of the core and the material via the 50% roll-off point.
Further consideration of tolerances and temperature ef-
fects makes it possible in a straightforward manner to have
consistent setting cur rents with one target minimum in-
ductance as required by most applications. The result is a
stable specification with higher values in both setting cur-
rent and minimum inductance as compared to the con-
ventional method.
The equations shown are not limited to the case of large
gaps, which represents the most common design case and
which can be handled in a simple way. As shown in [4] the
model can be also applied even for ungapped cores such as
r ings. PETe c h
References
1. R.M. Bozort h, Ferrom agnet ism , D.van Norst ran d
Co.Inc., New York: 1951.
2. M. Esguerra, Modelling Hysteresis Loops of Soft Fer-
rite Materials, Internat ional Conference on Ferrites ICF8,
Kyoto: 2000, 220- 222.
3. M. L. Hodgdon, IEEE Trans. Magn. 24 ( 1988) 3120.
4. M. Esguerra, to be submitted to IEEE Trans. Magn.
For more i nf or mat i on on t hi s ar t i cl e,
CIRCLE XXX on Reader Ser vi ce Card
temperature is given by (Eq. 17)
1 1 1 1
2 1 2 1

e e i i
T T T T ( )

( )
+
( )

( )
DTS: cf. Fig. 6
The adaptation of these parameters mainly affects the
setting current given by Equation 15. Because of the com-
parably large gaps considered here (Eq. 8), the minimum
inductance may be considered as temperat ure-indepen-
dent. This is desir able from an application point of view
and correct as long as T
2
>T
1
. In most cases,
i
( T
2
)>
i
( T
1
).
The specification procedure can be therefore comple-
mented with a final step: 3.) Setting current at the tem-
perature T
2.
(Eq. 18)
I T
B T
B T
DTS T
DTS T
set
s
s
2
2
1
2
1
1
1
( )
( )

( )
( )
( )
The advantage of this equation as opposed to a direct
calculation for ever y temperature is that a known specifi-
cation that has been defined at, say, room temperature can
easily be t r ansfer r ed t o a differ ent t emper at ur e. Even
though the r atio of saturation values is the dominating fac-
tor (see example below), the DTS term is necessar y to have
correct setting currents for all temperatures. The specifi-
cation procedure can be best illustrated by the following
example (Fig. 7):