198 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 15, NO.

1, JANUARY 2000

Magnetic Flux Distributions in Transformer Core

Joints
Günther F. Mechler and Ramsis S. Girgis, Fellow, IEEE

Abstract—This paper presents results of a rigorous analytical the performance of real cores. No explanation could be found.
study performed in order to understand the role which joints play The majority of the remaining work in this area concentrated on
in determining the performance of cores of power and distribu- the effect of the T-joint on the flux distribution in the plane of
tion transformers. The paper presents the details of the magnetic
flux distribution and redistribution in the core joint region along the joint and the additional losses caused by this joint in 3-phase
the core stack height direction as a function of the operating flux 3-limb cores [5]–[7].
density, core joint material, joint configuration, and joint design This paper presents the magnetic flux distribution in the -di-
parameters. The results are presented for both conventional and rection (core stack direction). Having available all the detailed
step-lap joints. The results illustrate the benefits of the step-lap information about flux distribution in the core joints over the
joint in reducing core losses, noise and excitation current. The re-
sults also explain the impact of overlap length, the number of lam- whole induction range, we are able to: 1) develop an accurate
inations per step, and the contribution of the joint region to the model for the whole magnetic circuit of the transformer; 2) de-
whole core performance. The calculated flux density wave shapes rive the precise contribution of localized joint losses to the total
in the joint region were found to be in good agreement with mea- core losses, taking into account the harmonic content of the flux
sured wave shapes. waves especially in conventional nonstep-lap joints; and 3) se-
The results presented in this paper have been used in developing
an accurate calculation of core losses for both power and distribu- lect the optimum core joint configuration for a family of trans-
tion transformers. They have also been used to arrive at the op- former core designs.
timum core joint configuration for a transformer core. First the calculation method and modeling technique used are
Index Terms—Core joints, finite difference method, magnetic described briefly. Results of the calculations are then presented
flux, no-load-loss, transformer core. starting with the flux distribution in a nonstep-lap joint, then in a
step-lap joint. The distributions are then presented for a number
of operating flux densities, grades of steel, joint configuration
I. INTRODUCTION
and joint geometry parameters. Finally calculated and measured

I N THE recent years we have made efforts to get a better un-

derstanding of the magnetic flux behavior in stacked trans-
former cores [1]. The final goal is to find a reliable method to
flux wave shapes are compared.

II. CALCULATION METHOD AND MODELING

calculate core losses and acoustic excitations in the design stage.
We use the finite difference (FD) method for magnetic field
For the magnetic field calculation we use a finite difference code
calculation [1] to calculate the flux distribution in a 2-dimen-
which takes into account the nonlinear behavior of the core steel
sional cut of steel laminations normal to the sheet plane and
permeability as well as its anisotropy.
parallel to the flux direction.
A detailed analysis of the magnetic flux distribution in the
The nonlinear magnetization behavior is modeled as well as
transformer core joint region is very critical to the accurate de-
the magnetic anisotropy between the rolling direction of the
termination of the magnitude and distribution of flux in the core
steel sheets and that perpendicular to the sheets.
and hence, for the required accurate calculation of core joint re-
As no measured material properties in normal direction
luctance. The joint reluctance influences the global distribution
was available, the assumption was made that magnetization
of the core flux and the flux concentrations in the neighborhood
in normal direction behaves similar to that in cross rolling
of the joint gaps cause additional localized joint losses. Both of
direction, which can be justified by the crystalline structure of
these effects can only be calculated using a detailed flux distri-
the steel. Also this assumption is not critical for the accuracy of
bution in the joint region.
the calculations since the thickness of the steel is a very small
Much of the past work dealing with the effect of core joints
fraction of the dimension in the direction of the main flux.
on transformer core performance has been experimental [2]–[7].
Flux direction inside the steel sheets is almost identical with
Z. Valkovic and Nakata used experimental models to determine
the rolling direction of the steel sheets on both sides of the joints.
the overall effect of the step-lap on core losses, excitation cur-
The bending to cross rolling direction near the joints is only a
rent [2], [3] and noise [4]. Because the models did not replicate
few degrees [1]. In the interlaminar gaps the flux lines are almost
the geometry of the actual transformer cores properly, the re-
normal to the sheet surfaces.
sults were somewhat contradictory and did not coincide with
As only the normal component of B is continuous at material
borders, the flux directions inside neighboring steel sheets are
Manuscript received January 20, 1999. almost independent of each other. So the overlap length in the
G. F. Mechler is with ABB Corporate Research, Heidelberg, Germany.
R. S. Girgis is with ABB Power T&D Company Inc., St. Louis, MO. model has to be the overlap length taken in rolling direction. On
Publisher Item Identifier S 0885-8977(00)00613-0. the other hand, the width of the joint gap, separating limb and
0885–8977/00$10.00 © 2000 IEEE
MECHLER AND GIRGIS: MAGNETIC FLUX DISTRIBUTIONS IN TRANSFORMER CORE JOINTS 199

Fig. 1. Magnetic flux lines in a conventional lap core joint at 1.7 T induction.
Fig. 2. Sketch to identify lines for line plots (not to scale).
yoke sheets, has to be interpreted as the shortest distance of both
in the real joint.
We always consider a finite number of lamination groups
which define entirely the overlap scheme. Because of a peri-
odical repetition of the sheet pattern, it is not necessary to study
the whole lamination package, which can consist of several hun-
dred laminations.
The average induction is defined by the total magnetic flux
penetrating the considered lamination groups. This average in- Fig. 3. Magnetic flux density in the steel lamination along line 1 of Fig. 2.
duction corresponds to the operating flux density of the core.
The joint gaps cause local disturbances of flux density. Without
gaps, the flux density inside the sheets would be uniform.
Every core joint type is characterized by the gap pattern in a
group of laminations which repeats periodically throughout the
whole steel package; e.g., in Fig. 1 we have a model with 10
steps. Two adjacent steps with two different gap positions form
a group (conventional overlapping or nonstep-lap = NSL). Each
step usually (not always) consists of 2 laminations. Significant
Fig. 4. Normal component of flux density along line 2 of Fig. 2.
parameters are the average air gap width, the overlap length be-
tween 2 steps of the same group, the number of steps per group
and the number of laminations per step. The thickness and ma-
terial properties of the considered core steel grade plays also an
important role.

III. RESULTS
A. Magnetic Flux Distribution in a Conventional Lap Joint
Fig. 1 shows a representative example for a conventional
Fig. 5. Normal component of flux density along line 4 of Fig. 2.
joint in power transformers with a gap width of 3 mm, while
the overall induction level is 1.7 T.
Far from the gaps the flux density level is identical in all In Fig. 5 the flux transfer between 2 sheets of the same step
sheets and equal to the value of the core induction. As the mag- (line 4 of Fig. 2) is shown. Its magnitude is much smaller and
netic flux approaches a gap it turns into the sheet of the next distributes on a wider range, because the influence of the gap
step. As the region between two gaps becomes saturated, some edges is only indirect. The normal flux along this line can only
of the flux passes through the air gap. be explained by even number of steps in the FD model. So the
Fig. 3 shows that, at the joint region, the flux density in the top-bottom symmetry is broken. The negative net flux in the
core sheet area shunting the air-gap rises up to 2.7 T while that overlap range is due to the fact that in the lamination at the
in the gap rises to a value of 0.7 T. These values are a function bottom of the package the gap is on the left side, at the top it
of the core induction as explained in Section II. is on the right side and the positive flux direction is from left to
In Fig. 4, values of the normal component of flux traveling right.
between neighboring sheets of adjacent steps are plotted. The Flux density along a line crossing the steel sheets at the gap
figure shows that this traveling between the sheets occurs in a position in normal direction is presented in Fig. 6. We observe a
very narrow strip (about 1 mm) before and after the gap. It also homogeneous distribution across the steel thickness and inside
shows that the magnitude of the flux traveling perpendicular to the gaps. We see again the flux density level in the gaps and
the sheets approaches 0.7 T. beside the gaps, where the steel is fully saturated.
200 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 15, NO. 1, JANUARY 2000

Fig. 6. Magnetic flux density along line 3 of Fig. 2.

Fig. 7. Flux lines in step-lap core joint at B (overall) = 1.7 T.

Fig. 8. Sketch for identification of plot lines in a SL joint.

Fig. 9. Magnetic flux density in all 6 different laminations in a step-lap core
B. Magnetic Flux Distribution in a Step-Lap Joint joint along lines 1–6 of Fig. 8.

A different and more common overlapping scheme at core

joints is the step-lap (SL) which is presented in Fig. 7. The loss and the noise level of a transformer. This also explains the
step-lap has 6 steps per group, which is typical for modern large difference between magnitudes of the excitation current of
power transformers. The calculation model comprises 3 groups. transformers with step-lap cores versus those with nonstep-lap
The flux lines describe the flux path in the joint environment at cores.
a core induction level of 1.7 T. Although the overlap length be- Fig. 10 shows the normal component of flux density between
tween neighboring steps of a group is chosen smaller than in two adjacent sheets of neighboring steps. We can see, how the
a NSL joint, the joint volume which is affected by the gaps is flux turns from one sheet to the next in the regions before and
bigger than with a NSL joint. after the gaps.
The basic property of a SL joint with 6 steps is that the flux in The upper plot along line 7 of Fig. 8 is taken in the inter-
one lamination step travels, upon meeting a gap, on to the next laminar gap between two groups, below the leftmost gap of the
5 steps. upper group and above the rightmost gap of the lower group.
In Fig. 9, the flux density in the mid sheet plane of the lower The plot along line 8 of Fig. 8 shows the behavior between the
lamination of each step of a group is plotted (along lines 1–6 of first and the second gap of a group.
Fig. 8). These plots demonstrate that, in the case of this type of It can be observed from Fig. 10, that normal flux density
step-lap, the flux almost totally avoids the gap (0.04 T) and re- before and after the gaps reaches a peak magnitude of 0.25 T,
distributes almost equally in the laminations of the other 5 steps, which is much less than the 0.7 T observed in a NSL joint (refer
bringing the flux density in these close to the level of total satu- to Fig. 4).
ration in the steel regions above and below this gap region. This In Fig. 11, the flux density along a line crossing the sheets
repeats in all other laminations. Note that this is quite different of a step-lap in normal direction at one of the 6 gap positions
from what occurs in a nonstep-lap joint, shown in Fig. 3. This (line 9, see Fig. 8). We find a very small variation of induction
explains the significant effect the step-lap has on improving the (about 0.005 T) within those ten sheets of one group, where no
flux distribution in the joint region and hence on the total core gap is passed. In Fig. 11, one can see that the regions in the ten
MECHLER AND GIRGIS: MAGNETIC FLUX DISTRIBUTIONS IN TRANSFORMER CORE JOINTS 201

Fig. 11. Flux density distribution along line 9.

Fig. 10. Normal magnetic flux density between sheets of different lamination
steps in a step-lap core joint (along lines 7, 8 of Fig. 8).

laminations above and below a gap exhibit flux densities that

are very much equal. The variation is only between 2.028–2.035
T, indicating again a very effective flux loading in the step-lap
joint. Conversely, the flux density in these regions rises to as
much as 2.7 T (far beyond the saturation point) in a nonstep-lap
joint, see Fig. 6.

C. Change of Local Flux Distribution with Overall Flux

Density Fig. 12. Flux in a step-lap joint at different operating flux densities (along line
2 of Fig. 8).
In Sections III-A and III-B we have described the flux density
distribution at 1.7 T operating induction.
Figs. 12 and 13 illustrate the flux redistribution in the joint
region at lower and higher operating flux densities.
Fig. 12 shows the flux density in a SL sheet at 3 flux density
levels (along line 2, see Fig. 8). It illustrates that the step-lap
joint results in a very effective redistribution of the flux in the
joint region at all inductions. At lower inductions, the maximum
flux density in the regions above and below the gap region pro-
portionally increases to 1.8–1.9 T. Correspondingly, at 1.8 T, the
flux density in these regions only increases to a level of 2.15 T.
This explains the more effective role the step-lap plays at lower
operating flux densities compared to that at higher flux densi-
ties. This is again illustrated in Fig. 13, where the magnitude of
the flux density of the flux traveling between neighboring steps
of the step-lap is shown at different levels of the operating flux
density of the core.
Fig. 13. Normal flux between lamination of a step-lap joint at different
operating flux densities (along line 8 of Fig. 8).
D. Flux Distribution versus Material Grades

In Figs. 14 and 15, the flux densities in the joint region are E. Influence of Joint Parameters
shown for two vastly different core materials; namely, an ex- Fig. 16 shows that the flux density level in the sheets between
tremely thin 0.23 mm (9 mil) laser scribed highly grain oriented the gaps is independent of the overlap length (distance of gaps
material and a thick 0.3 mm (12 mil) regular grain oriented ma- of neighboring lamination steps in the same group). Although
terial. Fig. 14 shows the distribution for a nonstep-lap joint we have different gap widths, the flux density in the gap is ne-
while Fig. 15 shows the distribution for a step-lap joint. glegible in both cases. Thus the gap width does not have a major
The figures illustrate that the higher anisotropy of the laser influence on flux distribution.
scribed steel deters some of the flux from traveling around the Fig. 17(b) and (c) show, that the flux density of the flux trav-
gap to the adjacent laminations. This effect is more evident in eling from one step to the next in a step-lap joint is higher for a
the nonstep-lap joint and is of less impact for a step-lap joint. In shorter overlap length.
the nonstep-lap, the flux density in the adjacent lamination rises Fig. 18 shows that this situation is different in a nonstep-lap
to a magnitude of 2.9 T at the gap region. This is compared to joint, where the overlap length does not have an appreciable
about 2.7 T for the laser scribed highly oriented material. effect. Comparing Fig. 17(a) and (b) we can see that in a step-lap
202 IEEE TRANSACTIONS ON POWER DELIVERY, VOL. 15, NO. 1, JANUARY 2000

Fig. 14. Flux density along line 1 (see Fig. 2) in a conventional joint using
different steel grades at a B-level of 1.7 T: (a) laser scribed super oriented steel
and (b) regular grain oriented steel.

Fig. 17. Flux change between steps with different overlap lengths and different
numbers of laminations per step in a step-lap joint along line 8 of Fig. 8: (a)
9 mm overlap length and 3 mm gap width, 1 sheet per step, (b) 9 mm overlap
length and 3 mm gap width, 2 sheets per step, and (c) 3 mm overlap length and
1 mm gap width, 2 sheets per step.

Fig. 15. Flux density along line 2 (see Fig. 8) in a step-lap joint using different
steel grades at a B-level of 1.7 T: (a) laser scribed super oriented steel and (b)
regular grain oriented steel.

Fig. 18. Flux distribution with different overlap lengths in a nonstep-lap joint
along line 2 of Fig. 2: (a) 9 mm overlap length, 3 mm gap and (b) 20 mm overlap
length, 3 mm gap.

bored step. Below saturation the flux density rises proportional

with overall induction. As soon as saturation is reached, part of
the additional flux doesn’t stay in the steel but passes through
the joint gap. So we have a flattened sine wave in the steel sheet
beside the gap.
The calculated flux density wave shapes in the joint region
Fig. 16. Flux distribution with different overlap lengths in a step-lap joint were found to be in good agreement with wave shapes mea-
along line 2 of Fig. 8: (a) 9 mm overlap length and 3 mm gap width and (b) 3 sured at the university of technology in Vienna by A. Ilo and
mm overlap length and 1 mm gap width.
H. Pfützner. This allows to determine analytically the harmonic
content of the flux density waves in the core joint.
joint with 2 sheets per step, the flux traveling to the overlapping
More comparison with measured results will be presented in
step is about twice the flux of the case with one sheet per step,
a future publication.
even though the overlap length is the same.

F. Calculated versus Measured Core Joint Flux Wave-Shape IV. CONCLUSIONS

Fig. 19 shows the measured and calculated flux density wave Having available all the detailed information about flux dis-
inside a steel sheet at a position directly beside a gap of a neigh- tribution in the core joints over the whole induction range, we
MECHLER AND GIRGIS: MAGNETIC FLUX DISTRIBUTIONS IN TRANSFORMER CORE JOINTS 203

[7] A. J. Moses and B. Thomas, “The spatial variation of localized power

loss in two practical transformer T-joints,” IEEE Trans. Mag., vol.
MAG-9, pp. 655–659, 1973.

Günther F. Mechler was born in Gundelsheim,

Germany, on August 13, 1952. He studied physics at
the University of Heidelberg and obtained the degree
of a Dipl.Phys. in 1979 in the field of experimental
plasma physics. Between 1979 and 1981, he was
with Endress & Hauser Corp., Maulburg, developing
humidity sensors.
Since 1981, Günther has been with ABB
Fig. 19. Flux density wave in a sheet beside the gap of the neighbor step in a Corporate Research Heidelberg (formerly BBC
nonstep-lap joint at 1.7 T operating induction. Corp., Mannheim) in the department for modeling
and simulation in the field of electromagnetism
and multiphysics. He participates in the development and improvement of
superconducting magnets, switchgear systems, electromagnetic bearings, and
are able to: 1) develop an accurate model for the whole magnetic transformers. He also participates in analyzing EMC problems in big industrial
circuit of the transformer; 2) derive the precise contribution of plants.
localized joint losses to the total core losses, taking into account
the harmonic content of the flux waves especially in conven-
tional nonstep-lap joints; and 3) select the optimum core joint
Ramsis S. Girgis is presently Manager of the Devel-
configuration for a family of transformer core designs. opment Engineering Department in the Power Trans-
former Division of ABB Power T&D Company lo-
REFERENCES cated in St. Louis, MO.
Dr. Ramsis was born in Egypt in 1946. He
[1] G. F. Mechler and R. S. Girgis, “Calculation of spatial loss distribution received the Ph.D. degree from the University
in stacked power and distribution transformer core,” IEEE Transactions of Saskatchewan, Canada, in electrical power
on Power Delivery, vol. 13, no. 2, pp. 532–537, Apr. 1998. engineering in 1978.
[2] Z. Valkovic, “Some aspects of additional losses in step-lap joints of Dr. Girgis has about 30 years of research and devel-
transformer core,” The Int. J. for Computation and Mathematics, vol. opment experience in the area of power, distribution
11, no. 1, pp. 137–140, 1986. and high frequency transformers, rotating machines
[3] T. Nakata, N. Takahashi, and Y. Kawase, “Magnetic performance of and pulse power components. His main area of interest is electromagnetics and
step-lap joints in distribution transformer cores,” IEEE Trans. Mag., vol. noise of electric power equipment. He has published and presented more than 50
18, no. 6, pp. 1055–1057, Nov. 1982. scientific papers in IEEE, IEE, CIGRE and other international journals. Ramsis
[4] Z. Valkovic, “Effects of transformer core design on noise-level,”, is presently the chairman of the IEEE Working Group on no load and load
Grenoble, France, paper #Tpm/6A-11, SMM13, 4 pp., 1997. loss testing of power and distribution transformers and is a member of several
[5] A. Ilo, B. Weiser, T. Booth, and H. Pfuetzner, “Influence of geometric working groups and subcommittees in the IEEE Transformers Standards Com-
parameters on the magnetic properties of model transformer cores,” J. mittee. He coauthored chapters in two recent electrical engineering handbooks
Magn. Magn. Mat., vol. 160, pp. 38–40, 1996. on transformer design and transformer noise. He also is the past Technical Ad-
[6] Z. Valkovic, “Additional losses in 3-phase transformer cores,” J. Magn. visor representing the US National Committee in the IEC Power Transformer
Magn. Mat., vol. 41, pp. 424–426, 1984. Technical Committee (14).