Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

PExprt From UPM PDF

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

9005 1

Teaching Magnetic Design using CAD tools


Roberto Prieto, Jesús A. Oliver and Pedro Alou
Centro de Electrónica Industrial
Universidad Politécnica de Madrid

Abstract. Design of magnetic components is a multivariable is needed, so the Steinmetz model is considered the
problem. There are many different combinations of shapes, sizes best.
and materials for the core with many diameters for the wires. So
it is difficult to find the optimum design without a great number Ÿ Wire losses have two components Direct Current (DC)
of iterations. Analytically only a few combinations are usually losses and Alternating Current (AC) losses. The AC wire
studied but it is very easy to take into account all the losses are produced by the mechanisms of skin effect and
combinations using a CAD tool [1]. In this work the CAD tool proximity effect. Analytically, wire losses are usually
used is PExprt (ANSYS [2]) which is being developed at UPM- evaluated with the following equation:
CEI.
2
P ≅ R DC ·Ieff
l
R DC = ρ ·
I. ANALYTICAL DESIGN s (3)
Design of magnetic components for power converter is Where only DC resistance (RDC) is considered but not AC
based on the specifications of the power converter (input resistance. This is not very recommendable if the ripple of the
voltage and current waveforms). current waveform is important compared with the DC level
(then AC resistance is not negligible).
A. Inductors analytical design
If the losses value is right the process is finished, otherwise
To design the inductor it is necessary to select both the core a new design must be recalculated (iterative process). It is not
and the conductors. The core is characterized by the size, the effective to do a lot of iterations and so it is difficult to obtain
shape and the material. The size is commonly selected in order the optimum component.
to manage the specified power. The shape depends on the
specific application of the component. The material is selected B. Transformers analytical design
in terms of the switching frequency of the converter. The type In a transformer the combinations are multiplied by the
of conductor (solid wire or Litz wire) is function of this windings number. The initial specifications of the converter
frequency too. The conductor area implies a limit on the losses define the value of turns ratio. In a transformer the rules to
and is commonly selected considering the skin effect. The select the core and to select the conductor type are the same as
number of turns, N, is selected in order to avoid core in the inductor. But the conductor area is selected to optimize
saturation (1). the losses. Usually the different wires of the windings have the
same current density. The following step is to obtain the
L·i ma x
N= number of turns of primary side which maintain the core
A e ·B ma x (1) losses bellow a fixed value.
With L and N it is possible to calculate the total reluctance Losses calculation is equal than for the inductor. If the core
of the component, which determines the gap dimensions, if it losses and the wire losses are right the design is finished,
is necessary. otherwise it will necessary to redesign the component.
Once the design is obtained the next step is the losses However, not always the optimum design is to reach the same
calculation: core losses and wire losses, it depends on the kind of core
Ÿ Core losses are well evaluated with vendor databooks or material.
with the classical Steinmetz model. This model appears in
equation (2). II. CAD DESIGN
β The use of the software tool makes possible a better
⎛ ΔB ⎞
P = Ve ·k ·f α ·⎜ ⎟ optimization of the components by analyzing a lot of
⎝ 2 ⎠ (2) combinations of wires, bobbins and cores possibilities. This
Where k, α and β are properties that depends on the would take a lot of time in an analytical way.
core material and the frequency and Ve is the effective A. Inductor and Transformer Losses
volume of the core. The goal of the design is to find the
magnetic component with least losses, so it is not Ÿ Core losses. Steinmetz model continue being the best in
necessary a very accurate result because the the trade-off between efficiency and accuracy.
comparative analysis of different designs is more Ÿ Wire losses. Both DC losses and AC losses are
important. A trade-off between precision and efficiency considered. The expression for the losses is:
9005 2

P = ∑ R w ·I 2w Ÿ It also allows an automatic model generation of the


w magnetic component. The model can be 1D analytical,
(4) based in transmission lines for PSpice and Simplorer
Where Rw and Iw are the resistance and current for the w-th
simulators. It is also possible to obtain a 2D/3D model,
harmonic. So, it is necessary an harmonic decomposition of based in Finite Element Analysis Tools (FEA). It is very
the current waveform and to know the dependence of the useful to have a model to study the component not only as
resistance with frequency. an isolated component but also as a part of a power
Another way to calculate wire losses is with Dowel model, converter.
which allows splitting up the losses due to skin effect and the
losses due to proximity effect.
It also takes into account the fringing flux, caused by the
effect of the gap [3-4], which modifies the window energy.
This means a different gap length too.
Especially significant is the losses optimization. In an
analytical way the optimization can be done refereed to the
flux density (if the current ripple is big) or to the number of
turns (if the current is mainly DC). The frontier between two
limits is not clear. As a solution the CAD tool optimize both of
Figure 2. PExprt Text Files for documentation.
them, starting with the design with the minimum number of
turns and increasing it discreetly, it finds a minimum in the
C. Design Steps
losses.
Next, the design steps of magnetic components with PExprt
B. CAD Design versus Analytical Design are described (grouped by sections).
Ÿ CAD design is more flexible than analytical design and it
is not necessary to be an expert designer. Specifications
Ÿ Design of magnetic components is a multivariable problem There are two ways to input the initial specifications of the
and the optimization is not easy. However, using a CAD power converter, which are shown on Figure 3. The first one
tool it is very easy to take into account all the is using the current and voltage waveforms and the second one
combinations. with the specific topology of the power converter.
Ÿ Analytical equations of magnetic components are difficult
to solve, but it is easy to implement these equations in a
software tool.
Ÿ The main drawback is that it is necessary to buy and to
learn how to use the different CAD tools.

PExprt specific advantages


Ÿ It has a friendly interface.
Ÿ It prevents construction problems, by selecting only the Figure 3. Diagram of PExprt Input Specifications
stock components.
If the input is based on the waveforms, there is not any
Ÿ The output of the analysis is a list of components (Figure restriction in the type of inductor or transformer that can be
1. ), so the designer can compare different designs and
designed.
choose the better using his/her own guidelines (less power
If the input is based on the converter specification, the
losses, volume, height, footprint, temperature...).
available are:
Ÿ Inductor of a Buck.
Ÿ Inductor of a Boost.
Ÿ Inductor of a Buck-Boost.
Ÿ Transformer of a Classic Forward.
Ÿ Transformer of a Half Bridge.
Ÿ Transformer of a Full Bridge.
Ÿ Transformer of a Half-Bridge.
Ÿ Transformer of a Push Pull.
Figure 1. List of results of PExprt.
Ÿ Transformer of a Flyback.
Ÿ It allows a quick documentation, because of the easy
generation of text files and screen captures.
9005 3

Component Selection Then, the equation to evaluate the losses is:


The next step is to select the desired components from the P ≅ R DC ·I ef2 f
Design Libraries of the vendor (Figure 4. ). These components
l
are the shape, size and material of the core, the bobbin, and the R DC = ρ ·
wire types. s (6)
Analytically, this is the way to evaluate the losses,
without taking into account AC losses. This s not very
accurate in the designs with an important current ripple
compared with DC level, because AC losses are not
negligible.
Ÿ DC losses and AC losses:
Skin effect only or Skin and Proximity effects (Dowell
model [5])
In these cases, the expression for the losses is:
Figure 4. Component Libraries P = ∑ Rw ·I w2
w
Designing with a CAD tool allows a better design
(7)
optimization, by analyzing a lot of combinations of wires, Where R w and Iw are the resistance and current for the
bobbins and cores. This analysis will cost a lot of time in an w-th harmonic. It is necessary a harmonic
analytical way. decomposition of the current waveform and to know the
dependence of the resistance with frequency.
Design Inputs
Therefore, AC losses calculation is not effective
The main options are (Figure 5. ): analytically. However, it is very easy to compute AC
Ÿ The gap can be central, can be divided in both legs or the losses with a CAD tool like PExprt.
magnetic component can have no gap.
Ÿ Losses optimization.
Ÿ The design can be with or without bobbin.
Ÿ The wire geometry can be planar, concentric or toroidal. There are two optimization modes. In Mode 1 the
Ÿ The separation between the wires of a layer or between optimum number of turns is calculated to minimize the
different layers must be specified. total losses. In Mode 2 the optimum number of turns
Ÿ Winding distribution can be 2D, 1D partially full or 1D with the optimum parallel wires are calculated to
completely full. minimize the total losses.
Ÿ The user can limit some values, for example the
Ÿ Options of the List of Results.
temperature rise, the gap size, the number of parallel wires.
Ÿ The fringing gap energy can be considered. The list of results can be made with all the solutions or
with a limited number of solutions that keep some
requirements (minimum losses, minimum volume,
minimum temperature rise...)

Figure 5. Design Inputs

Modeling options
Figure 6. Modeling Options
Modeling options are divided in three groups.
Ÿ Wire losses calculation.
List of Results
Wire losses have two components, the Direct Current
The output of the analysis is a list of components (Figure
(DC) losses and the Alternating Current (AC) losses.
The AC wire losses are produced by the high frequency 1. ). PExprt does not calculate only one design, but a list of
mechanisms of skin effect and proximity effect. possibilities for the specifications. The designer selects the
most suitable with his/her own design guidelines.
PExprt has three options to calculate the losses: If a design is selected out of the list of results, it is
Ÿ Only DC losses possible to see the constructive results and the performance
results.
9005 4

Constructive Results Input Voltage 48 V


Output Voltage 50 V
For example: Power output 100 W
Ÿ Wire type. Switching Frequency 100 kHz
Ÿ Number of turns. Maximum Current 0.458 A
Ÿ Gap length. Current Ripple 1.66 A
Ÿ Core size and material. Inductance 150 mH
Duty Cycle 50%
Table 1. Converter Specifications

Figure 9. Flyback topology

B. Analytical Design
Figure 7. Constructive Results The Analytical design of the inductor has been obtained in
the way described in previous sections:
Performance Results Design A.1 Design A.2
Core Size RM12/I RM12/I
For example: Core Material 3F3 3F3
Ÿ Losses. Primary
Irms 2.83 A 2.83 A
Ÿ Window occupancy. Wire Diameter 0.4 mm 0.4 mm
Ÿ Flux density. Number of Turns 27 27
Number of Parallel Turns 3 5
Secundario
Irms 2.95 A 2.95 A
Wire Diameter 0.4 mm 0.4 mm
Number of Turns 27 27
Number of Parallel Turns 3 5
Gap 855 mm 855 mm
Windings Losses 1.24 W 744 mW
Core Losses 37 mW 37 mW
Total Losses 1.51 mW 781 mW
Table 2. Results of the analytical designs
Designing is an iterative process. As the winding losses
Figure 8. Performance Result
were very high in Design A.1, a new design of less losses
(Design A.2) is obtained with more parallel conductors in both
primary and secondary side of the transformer.
Modeling
C. PExprt Design
There is a link with the Maxwell FEA tool, for a 2D/3D
PExprt data input is based on the converter specifications.
modeling of magnetic components.
The input voltage and current waveforms on the transformer
There is also another one for building a 1D analytical
extracted from the tool are shown in Figure 10. and Figure 11.
model based in transmission lines to study the global
converter. The model can be implemented in an electrical
simulator, like PSpice or Simplorer.

III. EXAMPLE. TRANSFORMER OF A FLYBACK.

A. Specifications Figure 10. Voltage and Current waveforms on the primary side
The main converter specifications are shown in the
following table. The circuit is shown in Figure 9. The
transformer is working in Continuous Conduction Mode.

Figure 11. Voltage and Current waveforms on the secondary side


9005 5

Searching for the Analytical Design with PExprt Core Size RM14/I
Core Material 3F3
Among the numerous solutions that are calculated with Primary
PExprt, one of them (Design B) is equal to the analytical Irms 2.9 A
Wire Diameter AWG25 (0.51 mm)
design (Design A.2):
Number of Turns 20
Core Size RM12/I
Number of Parallel Turns 5
Core Material 3F3
Secondary
Primary
Irms 2.8 A
Irms 2.9 A
Wire Diameter AWG24 (0.57 mm)
Wire Diameter AWG27 0.4 mm
Number of Turns 20
Number of Turns 27
Number of Parallel Turns 5
Number of Parallel Turns 5
GAP 633.4 mm
Secondary
Window Filling Factor 42.5 %
Irms 2.8 A
Windings DC Losses 405.9 mW
Wire Diameter AWG27 0.4 mm
Windings AC Losses 444.7 mW
Number of Turns 27
Core Losses 58.2 mW
Number of Parallel Turns 5
Total DC Losses 464.1 mW
GAP 873.8 mm
Total AC Losses 502.9 mW
Window Filling Factor 48.8 %
Temperature Rise 6.3 ºC
Windings DC Losses 836.523 mW
Windings AC Losses 881.1 mW
Table 4. Design C
Core Losses 35.1 mW It is interesting to increase a bit the size of the transformer
Total DC Losses 871.6 mW
Total AC Losses 916.2 mW
because the Window Filling Factor of Design B is very high
Temperature Rise 16 ºC and it can be difficult to build. With this new design, Design
Table 3. Design B C, winding losses fall and core losses are still low, so it is a
very good design.

Gap effect
In this section, the purpose is to emphasize the utility of use
2D/3D FEA models to design magnetic components. With
these models it is possible to study the gap effect in the losses
of the magnetic component.
The way to use a FEA model is with PExprt Analysis
Figure 12. Axial View of the Design B Mode. Previously, a PExprt FEA model must be extracted
The losses results are very similar. There are only little from the appropriate link. It is easier to implement these
differences with the winding losses because of the effective models in a CAD tool versus its use in the analytical
currents, which are not exactly the same for both designs. design.
Using the wire diameter as degree of freedom and keeping The design selected is Design B. It has not a very big gap
the core constant. but as it is a design with a lot of turns, the gap effect becomes
The core is kept constant and all wires of the stock library very important, so accurate winding losses can be calculated
are selected. PExprt design of fewer losses is the previous one. with this model:
Core Size RM12/I
The harmonic content of the current waveform is very high, so Core Material 3F3
the wire selected analytically is the best. Skin effect must be Primary
considered for the wire selection. Irms 2.9 A
Wire Diameter AWG27(0.4 mm)
Number of Turns 27
Number of Parallel Turns 5
Secondary
Irms 2.8 A
Wire Diameter AWG27 (0.4 mm)
Number of Turns 27
Number of Parallel Turns 5
GAP 873.8 mm
Window Filling Factor 48.9 %
Windings DC Losses 836.5 mW
Figure 13. Axial View of the Design C Windings AC Losses 25.3 W
Core Losses 43.1 mW
Total DC Losses 879.2 mW
Changing the core type Total AC Losses 25.4 W
Temperature Rise 442.2 ºC
If the core selected is smaller, there is not a design of fewer Table 5. Design C modified
losses than the Design B.
If the core selected is bigger (RM14/I), the losses decrease: AC losses have a great increase compared with the Dowell
9005 6

model losses [5]. This can induce errors when less losses Core Size RM14/I
design is being selected, and this is illustrated with the Core Material 3F3
Primary
following example. Irms 2.937 A
In Figure 14. , there is a list of results calculated with Wire Diameter AWG25 (0.51 mm)
Dowell model sorted by losses. Number of Turns 20
Number of Parallel Turns 5
Secondary
Irms 2.878 A
Wire Diameter AWG24 (0.57 mm)
Number of Turns 20
Number of Parallel Turns 5
GAP 633.4 mm
Window Filling Factor 42.5 %
Windings DC Losses 405.9 mW
Windings AC Losses 14.3 W
Core Losses 67.8 mW
Figure 14. List of results. Total DC Losses 473.7 mW
Total AC Losses 14.4 W
Temperature Rise 180.8 ºC
Table 7. Final design
Design D presents fewer losses than Design C, because the
first one has less number of turns. The optimum design (of
less losses) has changes when a PExprt model is used.

IV. CONCLUSIONS
As the current ripple is important, skin effect must be
Figure 15. Axial View of the Design D considered for the wire selection. That is why the wire
Two designs are compared; the Design B (in blue) and the selected analytically is the PExprt optimum too.
following design in losses, named Design D (in red). Design D Other way to compute AC wire losses is the use of 2D/3D
has more losses than Design B, but the gap of Design D is FEA models and, so 2D/3D effects can be taken into account,
smaller. As the gap effect increases the losses, this influence for example, gap effect. If the gap effect is important, as in
will be less considerable in Design D. This can be observed in this example, these models are very useful: there are changes
the following table: in the design positions in the list of results ordered by losses. It
Dowell model PExprt model is easier to implement these models in a CAD tool versus its
Core Size RM14/I RM14/I use in the analytical.
Core Material 3F3 3F3 This example has illustrated the advantages of using
Primary magnetic design CAD tools in order to teach particular effects.
Irms 2.9 A 2.9 A
Wire Diameter AWG23 AWG23 (0.64mm)
The combination of slides, blackboard and CAD tools are an
(0.64mm) excellent receipt for explaining magnetic design.
Number of Turns 20 20
Number of Parallel Turns 2 2 V. REFERENCES
Secondary
Irms 2.8 A 2.8 A [1] Final Master degree work "Diseño de Componentes Magnéticos para
Wire Diameter AWG27 AWG27 Alta Frecuencia", R. Rubio, División de Ingeniería Electrónica
(0.41 mm) (0.41 mm) (E.T.S.I.I, UPM).
Number of Turns 20 20 [2] Ansoft Corporation. Web site, http://www.ansoft.com.
Number of Parallel Turns 4 4
GAP 633.4 mm 633.4 mm [3] "Air-gap reluctance and inductance calculations for magnetic circuits
Window Filling Factor 21.7 % 21.7 % using a Schwarz-Christoffel transformation", Balakrishnan, A.; Joines,
Windings DC Losses 785.2 mW 785.3 mW W.T.; Wilson, T.G. Power Electronics, IEEE Transactions on ,
Windings AC Losses 858.5 mW 12.9 W Volume: 12 Issue: 4 , July 1997 Page(s): 654 –663
Core Losses 58.2 mW 67.8 mW [4] "New modeling strategy for the fringing energy in magnetic
Total DC Losses 843.5 mW 853.1 mW components with air gap", Escribano, L.M.; Prieto, R.; Oliver, J.A.;
Total AC Losses 916.7 mW 13 W Cobos, J.A.; Uceda, J. Applied Power Electronics Conference and
Temperature Rise 10.8 ºC 154 ºC Exposition, 2002. Seventeenth Annual IEEE, 2002 Page(s): 144 -150
Table 6. Design D vol.1

PExprt losses of Design D are smaller than the ones of [5] "Effect of eddy currents in transformer windings", P. L. Dowell.
Proceedings of the IEE. Vol 113 no.8. August 1966.Page(s).1387-1394
Design B. If the effect of the gap in the losses is taken into
account, the position of both designs will exchange the order
in the list of results.
Finally, PExprt losses of Design C have been calculated in
order to study if this design is still the optimum design:

You might also like