0% found this document useful (0 votes)
101 views10 pages

IET Electric Power Appl - 2021 - Tokgöz - Comparison of PCB Winding Topologies For Axial Flux Permanent Magnet Synchronous

Uploaded by

alex.evseenko
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
101 views10 pages

IET Electric Power Appl - 2021 - Tokgöz - Comparison of PCB Winding Topologies For Axial Flux Permanent Magnet Synchronous

Uploaded by

alex.evseenko
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 10

IET Electric Power Applications

Research Article

Comparison of PCB winding topologies for ISSN 1751-8660


Received on 17th July 2020

axial-flux permanent magnet synchronous


Revised 27th October 2020
Accepted on 28th October 2020
E-First on 8th February 2021
machines doi: 10.1049/iet-epa.2020.0622
www.ietdl.org

Furkan Tokgöz1, Gökhan Çakal1, Ozan Keysan1


1Electrical-Electronic
Engineering Department, Middle East Technical University, Ankara, Turkey
E-mail: [email protected]

Abstract: Although axial-flux permanent magnet machines have high torque densities, challenges regarding mass production of
stators make them a less appealing choice. Printed circuit board (PCB) axial-flux machine is a type of machine with a stator that
is made of layers of PCB. Given the precise, fast, and cheap mass production capabilities of PCB manufacturers, PCB axial-flux
machines stand as a viable alternative for conventional round-wire winding machines. In this study, five different winding
topologies are compared. Their induced phase voltages and torque are calculated using the developed magnetic scalar
potential method and finite element analysis (FEA). Proposed windings are tested on a 16-pole, 2000-RPM, double rotor-single
stator axial-flux permanent magnet synchronous machine. Results showed that the parallel winding had the smallest resistance
and loss. Moreover, radial and concentric winding had the highest induced voltage and torque while the radial winding had 20%
less phase resistance than concentric. Also, the induced voltage of radial winding had the smallest total harmonic distortion in
comparison with other winding types. A novel unequal width parallel winding is proposed and it is compared with parallel
winding separately. It is found that by simply increasing the cross-section area of wave windings, it is possible to decrease
copper loss by 17%.

1 Introduction induced phase voltages. Different axial-flux PCB motors are


reviewed in [13].
Axial-flux machines are gaining popularity thanks to their short Detailed comparative analysis and the analytical calculation
axial-length, higher torque and power densities. In literature, there method of induced phase voltages and output torque of various
are cases in which axial-flux machines are used in wind turbines, winding topologies lacks in literature. In this paper, five different
electric and hybrid vehicles, flywheels [1–5]. However, axial-flux stator winding topologies are compared with a double rotor-single
machines are not very suitable for mass-production, due to their stator axial-flux permanent magnet motor. Other machine
planar stator geometry which makes the manufacturing of the stator parameters are kept constant while comparing different windings.
challenging. Printed circuit board (PCB) motor offers a solution to Rotors, air-gap distance, PCB inner and outer diameters, PCB
this problem by use of the precise and cheap manufacturing options thickness, number of layers, clearance, width and thickness of the
of PCB technology that enables axial-flux permanent magnet copper tracks are identical in order to have a fair comparison.
synchronous machines (AFPMSM) with PCB winding to be used Unequal width parallel winding is proposed and compared with
in small wind-turbines, water current turbines, hard-disk drives, parallel winding separately in order to observe the effect of unequal
video disk drives and spindle motors [6–10]. width trace technique on wave windings. Basic principles that
Fast and cheap PCB manufacturing process ensures that all govern the axial-flux motor design, calculation of the air-gap flux
PCBs can be produced with precision in micrometres range, which density using magnetic scalar potential, analytical calculation of
enables the construction of small electric motors. Control induced phase voltages and torque values of different winding
algorithms such as model predictive control or direct torque control topologies and validation of the calculations by FEA and tests
require the motor parameters to be known. Since each winding results are presented in this paper.
PCB will be identical to each other, these control methods can be
accurately implemented. Moreover, cogging torque can be
eliminated if air-cored stator is used. Lastly, air-cored PCB motors 2 PCB winding design
have a fewer number of turns and large effective air gaps which 2.1 Winding types
make the phase inductances low. The lower the inductance, the
smaller the electrical time constant, and the better the transient In this section, five different winding types are investigated. Their
performance is. advantages and disadvantages are listed. The designed concentric
In literature, there are different winding designs for PCB winding, radial winding, parallel winding, arc winding and unequal
motors. The concentric winding, which is a commonly used width parallel winding can be seen in Figs. 1 and 2.
winding type, has a high induced voltage and torque output
because of the high flux linking area and long radial winding 2.1.1 Concentric winding: The concentric winding has a high
length. However, end winding connections of the concentric induced voltage and torque values thanks to long radial copper
winding are long. As a result, the resistance of the current path paths and high flux linking area. Concentric winding design is
increases, so the efficiency of the motor is reduced and thermal presented in Fig. 1a. There are several studies in the literature that
stress increases. Alternatively, wave winding topologies are concentric winding is used [10, 14–16]. However, only the radial
proposed in [11]. The main purpose of design is to increase the part of the copper traces is active in the torque generation of the
efficiency of the winding without losing the dynamic performance machine. So, long inactive end winding connections of concentric
of the motor. A comparison of different windings for PCB motors winding make it an inefficient solution. Moreover, harmonic levels
is presented in [12]. However, PCB inner and outer diameters of of the induced voltage are high because of the winding shape.
various windings are chosen differently and fractional pitched
stator coils are used which changes the harmonic levels and

IET Electr. Power Appl., 2020, Vol. 14 Iss. 13, pp. 2577-2586 2577
© The Institution of Engineering and Technology 2020
17518679, 2020, 13, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-epa.2020.0622 by Readcube (Labtiva Inc.), Wiley Online Library on [07/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Fig. 1 Five different winding designs that are covered in the paper. Red colour represents the top layer and blue colour represents the bottom layer
(a) Concentric winding, (b) Parallel winding, (c) Radial winding, (d) Arc winding, (e) Unequal width parallel winding

2.1.3 Radial winding: Alternatively, the radial winding is


proposed in [12] and it is found that radial winding has higher
induced voltage with a high third harmonic level compared to
parallel winding. Long radial copper traces, which give the name to
the radial winding, increase the output torque. Also, it covers most
of the pole area that creates the induced voltage. Since the copper
trace length of the radial winding is longer than parallel winding,
losses will be higher.

2.1.4 Arc winding: A winding topology, which is called as arc


winding in this paper, is proposed in [17, 18] that induces a better
sinusoidal voltage at the expense of increasing total copper trace
length. However, total harmonic distortion (THD) levels are not
presented in [17, 18]. So, it is considered one of the design options
that is covered in this paper.

2.1.5 Unequal width parallel winding: Unequal width winding


design is proposed in [19] for concentric windings. In order to
investigate the effect of the unequal width traces, a winding
Fig. 2 One loop of each of the designed windings compared with the size topology that is called unequal width parallel winding is proposed
of the magnets. (1) Concentric winding, (2) parallel winding, (3) radial in this paper. The difference between parallel winding and unequal
winding, (4) arc winding, (5) unequal width parallel winding width parallel winding is that the trace width of the unequal
parallel winding increases as the radius increases as can be seen in
2.1.2 Parallel winding: Parallel winding is proposed in [11]. The Fig. 2. Thanks to that, the resistance of the copper trace decreases,
advantage of parallel winding is that it follows the shortest path to and the PCB surface area is utilised better. The total copper surface
create a loop for the stator winding. As a result, the resistance of area is increased that improves the cooling performance of the
the copper trace is smaller than the resistance of the concentric unequal width parallel winding. Unequal width traces can also be
winding. Due to the low flux linking area, the induced voltage applied to other wave winding types such as radial or arc winding.
magnitude will be lower than the concentric winding. Moreover, Since the width of the copper traces of unequal width parallel
the output torque of the parallel winding is low because the copper winding is greater than the other four winding types, it is separately
paths are not aligned in the radial direction. As a result, the created investigated and compared with parallel winding to observe the
torque vector should be multiplied by a cos(α) term, where α is the effect of unequal width trace.
angle between the copper trace and the radial direction, in order to
find the circumferential output torque component.

2578 IET Electr. Power Appl., 2020, Vol. 14 Iss. 13, pp. 2577-2586
© The Institution of Engineering and Technology 2020
17518679, 2020, 13, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-epa.2020.0622 by Readcube (Labtiva Inc.), Wiley Online Library on [07/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Table 1 Comparison of physical and electrical properties of considered, as the fractional pitched winding changes the harmonic
the designed windings levels of the induced voltage waveforms which disturb fair
Type Trace length, Covered area, Phase resistance, comparison between different winding designs. Since the machine
mm mm2 Ω is a 16-pole 3-phase machine, slot per pole per phase of the motor
Concentric 3244 171 1.04
can be found as 3. The tracks start from the top and end in bottom
as presented in Fig. 2. The standard phase connection of A, -C, B, -
Radial 2740 340 0.87
A, C, -B is used. The resultant trace length of one phase, the area
Arc 2645 364 0.85 covered by one loop of the winding and the calculated phase
Parallel 2470 255 0.79 resistances are presented in Table 1. Note that, the covered area of
UEW 2470 255 0.59 the concentric winding is the area covered by the loop in the
middle of the number of turns. Concentric winding has the highest
phase resistance compared to other winding types because of the
high trace length. Moreover, it can be seen that increasing trace
width with unequal width technique reduces parallel winding's
phase resistance by 25%.
There a few parameters that affect the cost of a PCB: thickness
of the board, the copper thickness, the number of layers and the
clearance between the copper traces. Default values of these PCB
parameters are 1.6 mm, 35 μm, 2 layers and 0.2 mm, respectively.
If the maximum power density is aimed at design stage, the
maximum amount of copper should be used while keeping the PCB
thickness minimum. So, the highest number of layers with
maximum copper thickness and minimum board thickness is
required. However, custom PCB parameters increase the
manufacturing cost. Several PCB options are compared in this
Fig. 3 Analysis of manufacturing cost per pieces of 1000 PCBs with
paper.
different options
In order to investigate the effects of changing the copper
thickness and number of layers, the thickness of the board and the
clearance between copper traces are chosen as 1.6 and 0.2 mm,
respectively. The production cost of 1000 boards is obtained from a
PCB manufacturer. The production cost versus the number of
layers and the copper thickness are presented in Fig. 3. The cost of
the PCB with more than six layers increases immensely. Moreover,
it is found that there is a significant cost difference when copper
thickness is increased from 35 to 70 μm, but increasing the copper
thickness from 70 to 105 μm costs almost the same. Stacking the
boards in parallel is a valid option to reduce the cost of the stator.
Another factor that affects the cost per board is the number of
boards that is going to be produced. Production cost per pieces of
1000 PCBs is just 13% of the production cost per pieces for 10
PCBs. An optimum point of cost per power should be investigated
in detail for every PCB motor depending on the application.
While benchmarking different windings, stators are
implemented with the same rotor design. The motor designed for
Fig. 4 Exploded view of double rotor-single stator axial-flux permanent this paper is a 16-pole, 2000-RPM axial-flux permanent magnet
magnet PCB motor. (1) Rotor cover, (2) the first rotor, (3) magnets, (4) synchronous machine as can be seen in Fig. 4. The motor is
PCB, (5) the second rotor designed and optimised using a genetic algorithm to maximise the
torque per kilogram value of the motor. The details of algorithm
Table 2 Specifications of the designed motor are omitted since it is not in the scope of the paper. The machine is
Rated speed 2000 RPM Rated frequency 266.67 Hz in a double rotor-single stator configuration. Eddy current losses in
this motor configuration are minimal since both of the rotor yokes
Number of poles 16 Air-gap 1 mm
rotates at the same speed. Having thin copper traces on PCBs, eddy
Magnet thickness 2 mm Residual flux density 1.33 T loss on stator winding will be negligibly small in the range of 10–
Outer diameter 92 mm Inner diameter 50 mm 20 mW at 2000 RPM. Moreover, having two rotors rotating at high
PCB thickness 1 mm Core thickness 5 mm speed at both sides of the PCB helps the cooling of the stator
Trace clearance 0.2 mm Trace width 0.8 mm windings since rotors act as fans that create airflow. The detailed
Number of layers 2 Copper thickness 70 µm parameters of the motor are given in Table 2.

3 Analytical calculation
2.2 Winding design The analytical model for the proposed PCB motor is derived in this
In order to have a fair comparison between different winding section. The analytical modelling of PCB motor starts with open-
topologies, the number of layers, thickness and the width of the circuit magnetic field modelling. Then, the induced voltage
traces, clearance between traces, the outer and inner diameters of characteristics will be derived for different winding topologies
the PCBs are kept the same in concentric, radial, parallel and arc using the derived field model with the help of FEA results.
winding. Since the unequal width winding has a higher width of
traces of other types, it should be separately investigated with 3.1 Magnetic field modelling
parallel winding. The area of the PCB is fully utilised by putting
the maximum number of vias that can be fit into the inner diameter There are several methods proposed in the literature for magnetic
of the PCB. Radial, parallel, arc and unequal width parallel field analysis. In [20–23], the authors used magnetic scalar
windings have 144 vias or ‘slots’. A full-pitched winding potential to derive the magnetic field. There are studies that use
configuration is used for the design of the wave windings to have a magnetic vector potential for magnetic field modelling as in [24,
maximum winding factor. Fractional pitched winding is not 25]. In this study, the magnetic scalar potential is used to derive the

IET Electr. Power Appl., 2020, Vol. 14 Iss. 13, pp. 2577-2586 2579
© The Institution of Engineering and Technology 2020
17518679, 2020, 13, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-epa.2020.0622 by Readcube (Labtiva Inc.), Wiley Online Library on [07/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Table 3 Specifications of the selected motor for magnetic [20–22], and the details of the derivation can be found in these
field analysis studies. To summarise, first, the magnetisation of the permanent
Rotor outer diameter 92 mm magnets is represented using Fourier series. Note that permanent
Number of poles 16 magnets are magnetised only in the axial direction. Then, the
Remanence flux density, Br 1.33 T
magnetic scalar potential is written in terms of magnetisation
vector. The solution for magnetic scalar potential is governed by
Distance between rotor cores, L 7 mm Laplace equation with some unknown coefficients. Using
Magnet to pole pitch ratio, αp 0.8 constitutive relations, the axial component of the air gap flux
Magnet thickness, hm 2 mm density can be written as

πnμ0 πnx
BIy = − ∑ C eπny / τp − C2e− πny / τp cos
τ p n = 1, 2, 3… 1 τp
(1)

where n stands for nth order harmonic, μ0 is the permeability of


free space, τ p is pole pitch, x is the circumferential position and y is
the axial position. The unknown coefficients, C1 and C2 are
determined using boundary conditions.
In order to measure the accuracy of the analysis, the magnetic
field of a selected motor is obtained using both analytically and
FEA. The specifications of the selected motor are given in Table 3.
It is a double-sided axial flux motor employing NdFeB permanent
magnets. The motor is constructed for a gimbal camera
stabilisation system to be used in an aerospace application. FEA is
Fig. 5 Comparison of the air gap flux density distribution obtained with
carried out using ANSYS Maxwell 3D solver. The air-gap flux
the analytical model and FEA
density results are compared at the mean radius as shown in Fig. 5.
According to results shown in Fig. 5, there is <1% difference at
the fundamental components of the air gap flux density. The
obtained results show that the flux density obtained using the
proposed analytical model is in good agreement with finite-element
results, verifying the accuracy of the analytical field model.
Note that pole pitch increases with radius in the motor. The
presented analytical model is capable of finding the field solution
for all points by adjusting pole pitch. For simplicity, the magnetic
field only in the middle radius is considered for the induced voltage
and torque calculation in the following subsection.
Additionally, although the magnetic scalar potential method
successfully takes the leakage flux on the circumferential direction
into the account, it does not calculate the leakage flux at the inner
and outer edges of the magnets. Therefore, the variation of the
Fig. 6 Distribution of the magnetic flux density in z-direction on ANSYS magnetic flux density with respect to the circumferential direction
Maxwell can be calculated accurately but the variation with respect to radius
cannot be calculated using this method. As can be seen in Figs. 6
and 7, flux density indeed varies with the radius. Thus, in order to
increase the accuracy of the analytical model, magnetic scalar
potential and FEA results are used together in the following
subsection, where the induced phase voltage and torque
characteristics will be obtained.

3.2 Induced voltage and torque calculation of PCB windings


In axial-flux machines, the cross-product of the current density of
the winding and magnetic flux density, which is perpendicular to
the current, creates the force density vector. Force density is
integrated over the volume of the copper paths. So, increasing the
Fig. 7 Variation of the magnitude of the magnetic flux density at the number of layers, trace width or thickness increases the copper
middle of the magnet with respect to radius in FEA volume as it increases the torque. One must keep in mind that
output torque is the circumferential component of the created
open-circuit magnetic field model. Then, the accuracy of the torque vector. So, the longer the radial part of the copper trace, the
analysis will be compared with FEA results. Induced phase voltage greater the output torque. The output torque of a machine can be
and output torque calculation methodology of wave windings will calculated as
be presented in the next section.
The reason why no-load magnetic field analysis is used in this f = J×B (2)
paper is that the magnetic field in the air gap is created by
permanent magnets on the rotor and this field can be distorted by
the magnetic field created by current-carrying armature coils on the
F= ∫ ∫ ∫ f ⋅ dυ = ∫ ∫ ∫ J × Bdυ (3)
stator, called armature reaction. However, the proposed topology is υ υ

air-cored and has a large magnetic air-gap compared to iron-cored


machines. Therefore, armature reaction is low and negligible [26,
27]. Accordingly, only no-load magnetic field by permanent
T = Fr = r ∫ ∫ ∫ J × Bdυ (4)
υ
magnets is considered in the study.
The magnetic field derivation of air-cored axial flux permanent where J is the current density, B is the flux density, f is the force
magnet machines using magnetic scalar potential is presented in density, F is the force, r is the radius and T is the torque.

2580 IET Electr. Power Appl., 2020, Vol. 14 Iss. 13, pp. 2577-2586
© The Institution of Engineering and Technology 2020
17518679, 2020, 13, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-epa.2020.0622 by Readcube (Labtiva Inc.), Wiley Online Library on [07/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

B (r, θ, t) = a^ zBr(r)Bmaxsin(wt + pθ) (7)

in which Bmax is the peak of fundamental magnetic flux density, w is


the angular frequency, p is the number of pole pairs and Br(r) is the
normalised function that is found via curve fitting the flux density
variation graph which was represented in Fig. 7. Magnetic flux
density is rotating with an angular speed of w. Also, flux density
varies with the position in angle, θ, with a multiple of pole pairs, p.
The magnetic model of the axial-flux machine with parallel
winding can be seen in Fig. 8. The grey area represents the region
where the flux density is present, i.e. the section which overlaps
with magnets. The red line is the copper path on the top layer and
the blue line is the copper path on the bottom layer. A single loop
of the winding is enough to calculate induced phase voltage
magnitude.
The integration is divided into two parts for the parallel
winding. The first part calculates the flux linkage in the region
between the lines of L1 and L4 of Fig. 8. Alternatively, the second
part calculates the flux linkage in the region between the lines of L2
and L3. In order to find the induced voltage of different windings,
the limitations of the integral should be set properly. θ is defined as
the angle difference between two points in lines with the same
Fig. 8 Magnetic model of the motor with the parallel winding for induced radius. For example, θ can be found as 45° between L1 and L4 when
phase voltage calculation (red: top layer, blue: bottom layer)
radius is equal to r1. After finding the angle difference for different
radii, θ can easily be written as a function in terms of radius. Then,
these functions are set as the limits of the integration. The flux
linked by the single loop of the coil is then multiplied by the
number of turns in order to find the total flux linkage of one phase.
The analytical calculation can be done as follows:
r2 θ1 /2
ϕ1(t) = ∫ ∫
r1 − θ1 /2
a^ z ⋅ kwBr(r)Bmaxsin(wt + pθ) ⋅ rdθdr (8)

r3 θ2 /2

Fig. 9 Linear approximation of the wave windings. From left to right: arc
ϕ2(t) = ∫ ∫
r2 − θ2 /2
a^ z ⋅ kwBr(r)Bmaxsin(wt + pθ) ⋅ rdθdr (9)

winding, parallel winding, radial winding


ϕtotal(t) = ϕ1(t) + ϕ2(t) (10)

dϕtotal
V(t) = − N (11)
dt

In this paper, the analytical calculation of induced voltage of the


parallel winding is shown. Other winding designs can be calculated
with the same procedure. In fact, the proposed analytical
calculation method can be applied to any other wave winding
topology regardless of the winding shape. The only difference
between various windings is the integration limits. Linear
approximation of other winding types, which is illustrated in Fig. 9,
is enough to determine integration limits. Parallel winding has been
Fig. 10 Magnetic model of the motor with the parallel winding for torque calculated using two integrals and relations of θ and r, whereas
calculation radial winding has been calculated using three regions. As the
complexity of the winding shape gets higher, the number of
Also, the pole area that is covered by the turns of the winding integration to find the induced voltage waveform increases. Arc
results in flux linkage which creates induced voltage by well- winding can be an example of this since it has five regions to
known Faraday's law of induction. So, induced phase voltages can integrate. For every region which is separated by the lines, the
be calculated as follows: relation between θ and r should be calculated and flux linkage in
these regions should be found. Torques of other winding types are
ϕ= ∫ ∫ B ⋅ dS (5)
calculated similarly to parallel winding. However, the torque
produced on every piece of lines in other windings should be
S
calculated. Similar to the parallel winding model in Fig. 10, the
dϕ output torque of radial and arc windings are calculated.
V induced = − N (6) In order to calculate the voltage for different harmonic levels,
dt
the same procedure is used. However, kw, winding factor, should be
in which ϕ is the magnetic flux, N is the number of turns and V is calculated at harmonic orders of the fundamental frequency.
the induced phase voltage. Winding factor calculation of PCB motors is similar to radial-flux
After obtaining the magnetic flux density distribution, induced machines. Copper traces of PCB can be defined as slots. Knowing
voltage and torque can be calculated. First, the magnetic flux the slot number, number of poles and phases, one can calculate the
density distribution across the air gap should be approximated as a winding factor for fundamental and harmonic frequencies. Also,
function that varies with radius, angle, speed and time. So, the sin(wt + pθ) term in (7) should be modified according to frequency
airgap flux density distribution is written as follows: by changing w term with appropriate frequency.

IET Electr. Power Appl., 2020, Vol. 14 Iss. 13, pp. 2577-2586 2581
© The Institution of Engineering and Technology 2020
17518679, 2020, 13, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-epa.2020.0622 by Readcube (Labtiva Inc.), Wiley Online Library on [07/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Table 4 Analytical results of induced phase voltage and torque calculations of the windings
Winding type Induced voltage, V Torque, mN·m
Concentric 6.21 182
Arc 5.20 140
Parallel 4.87 129
Radial 6.29 175
UEW 4.87 129

Table 5 Finite-element analysis results of the windings


Winding type Induced voltage, V Torque, mN·m
Concentric 6.20 178
Arc 5.14 146
Parallel 4.89 136
Radial 6.30 178
UEW 4.88 136

The magnetic model used for torque calculation can be seen in has higher efficiency. The phase inductance of concentric winding
Fig. 10. RMS current density of the copper traces is multiplied with is found greater than any other winding type since the number of
mean air gap flux density in order to find force density. The turns of the concentric winding is greater. This might be a reason
circumferential component of the created force vector is the part for choosing the concentric winding at the expense of greater
that is responsible for the output torque. So, it is multiplied with resistance in some cases.
cos(α), in which α is the angle between the circumferential The windings are tested in a 16-pole, 2000 RPM axial-flux
direction and the normal of the vector, to find the circumferential permanent magnet synchronous motor. Rotors are positioned on
part of the force density vector. Then, force density is multiplied both sides of the stator winding. The prototype is presented in
with the volume of the copper traces. As done earlier with the Fig. 12. Due to differences of shaft diameter and hole diameter of
induced voltage calculation, torque calculation is done in two parts the rotor cover, there was a slight eccentricity in the axial direction.
for parallel winding. The torque created on L1 and L2 is calculated Fine-tuning of the position of rotor blocks had to be done to
using minimise the effect of the eccentricity on the experimental results.
A DC motor is mechanically coupled to the shaft of the AFPMSM
f = JrmsBavgBr(r)cos(α) (12) and the machine is tested under the generator mode of operation at
2000 RPM. The test setup is shown in Fig. 13. Motor is tested
υcopper = Nltracewtracettrace (13) under no-load and full-load conditions. Load resistance is varied
such that 2 A of load current is drawn from every stator winding at
r2 r3 2000 RPM at full-load. Electrical output power is measured by the
T= ∫ r1
υcopper f 1 ⋅ dr + ∫
r2
υcopper f 2 ⋅ dr (14) three-phase wattmeter and mechanical input power is calculated
from speed and torque values.
The comparison of analytical, finite-element analysis, and test
where Bavg is the average flux density in the air gap, ltrace, wtrace and results of the Fourier transform of open-circuit induced phase
ttrace are the length, width and the thickness of the copper traces, voltages can be seen in Fig. 14. Analytical, finite-element analysis
respectively. N is the total number of copper traces on both layers. and test results showed similar results in all winding types. Induced
α is defined as the angle between the tangent line at the middle phase voltage waveforms of the winding for one period can be seen
point of L1 and L2 and the force density vector. in Fig. 15. The harmonics of the induced phase voltages are
The analytical calculation results of induced phase voltages and compared in Fig. 16. It can be seen that while the concentric
output torque for different windings are presented in Table 4. winding has a higher fundamental voltage than the radial winding,
Concentric and radial windings have the highest induced phase the magnitude of the third harmonic is four times higher than the
voltages and output torque of all windings. radial winding.
Analytical, FEA and experimental results are in good agreement
3.3 Finite-element analysis with each other. Radial and concentric winding showed the highest
fundamental voltage value. However, induced phase voltages of
Finite-element analysis of the windings is conducted on ANSYS concentric winding have 5.7% THD, whereas radial winding has
Maxwell. Table 5 shows the results of each winding topologies. 1.3%. Harmonic levels of the concentric winding are higher than
Analytical and finite-element results are in good agreement with any other winding type because of the winding shape. The induced
each other. Concentric winding and radial winding showed the voltage of a copper path can be represented as the summation of
highest induced voltage and torque value compared to other infinitesimal voltage sources connected in series. If these voltage
topologies. Also, eddy current losses are investigated since the sources are connected at the same angle as in concentric winding,
frequency is high. It is found that eddy current losses of windings the harmonic levels will be higher because the voltage will vary
are around 10–20 mW at 2000 RPM. Eddy losses are minimal greatly when the angle changes. However, the copper path, voltage
because of thin copper traces. Also, both rotors rotate at the same sources, of the radial winding has a smooth transition between
speed in double rotor-single stator configuration. As a result, there different angles that creates a more sinusoidal voltage waveform.
will be no eddy current which is induced on the back-cores. Hence, the harmonic levels of the radial winding are smaller
compared to those of the concentric winding. This phenomenon
4 Experimental verification can also be observed in comparison to arc winding and parallel
winding as well. Parallel winding's voltage waveform changes
The PCBs are manufactured as shown in Fig. 11. PCBs have two linearly as which results in a triangular waveform shape with
layers with 70 μm copper thickness. Measured resistances and higher THD levels as 5.2%, whereas arc winding has 2.9%.
inductances of the windings are presented in Table 6. As expected, Unequal width parallel winding and parallel winding showed
the concentric winding has the highest resistance among four similar results. It is expected since their winding shape is the same.
different topologies. Radial winding, which has similar torque The difference occurs at their thermal performance as the unequal
output with concentric winding, has 20% less resistance and hence width winding has 18% less copper loss than parallel winding.

2582 IET Electr. Power Appl., 2020, Vol. 14 Iss. 13, pp. 2577-2586
© The Institution of Engineering and Technology 2020
17518679, 2020, 13, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-epa.2020.0622 by Readcube (Labtiva Inc.), Wiley Online Library on [07/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Fig. 11 Produced boards with different winding styles
(a) Concentric winding, (b) Parallel winding, (c) Radial winding, (d) Arc winding, (e) Unequal width parallel winding

Table 6 Electrical test results of the windings


Type Phase resistance, Ω Phase inductance, µH
Concentric 1.07 14.8
Arc 0.91 13.9
Parallel 0.79 11.8
Radial 0.86 12.2
UEW 0.66 9.3

Fig. 12 Constructed double rotor-single stator axial-flux synchronous machine with concentric winding stator

Then the machine is tested under full load, 2 A phase currents at change of the windings with respect to time is shown in Fig. 18. As
2000 RPM, torque values are found as in Table 7. In order to check expected, concentric winding has the highest temperature around
the thermal stress of the stator windings, all stators are tested under 44°C. Radial winding which had the same induced voltage and
full-load for 5 min. The temperatures of the stators are recorded via torque value as concentric winding has reached the final
a thermal camera as can be seen in Fig. 17. The temperature temperature of 35°C. Parallel winding has the lowest temperature

IET Electr. Power Appl., 2020, Vol. 14 Iss. 13, pp. 2577-2586 2583
© The Institution of Engineering and Technology 2020
17518679, 2020, 13, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-epa.2020.0622 by Readcube (Labtiva Inc.), Wiley Online Library on [07/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Fig. 13 Test setup of the PCB motor. (1) Human interface, (2) PCB motor, (3) torque transducer, (4) prime mover, (5) armature resistance, (6) oscilloscope,
(7) 3-phase wattmeter, (8) resistive load bank

Fig. 14 FFT comparison of no load induced phase voltages of analytical, FEA and test results at 2000 RPM
(a) Concentric winding, (b) Parallel winding, (c) Radial winding, (d) Arc winding, (e) Unequal width parallel winding

Fig. 15 Induced phase voltage waveforms of the windings at no load condition at 2000 RPM

Fig. 16 FFT comparison of no load induced phase voltages of the windings at 2000 RPM

Table 7 Test results of the windings under full load


Type Phase voltage, V Torque, mN·m Copper loss, W Output power, W THD, %
Concentric 6.22 175 12.8 24.5 5.7
Arc 5.27 146 10.8 20.8 2.8
Parallel 4.96 134 9.5 20.3 5.2
Radial 6.21 178 10.3 26.9 1.2
UEW 4.75 130 7.9 20.6 4.7

of 33.6°C in comparison with concentric, radial, arc and parallel One can observe that the unequal width parallel winding has a
winding. Unequal width parallel winding which has the lowest lower stator temperature compared to parallel winding although it
resistance value has also the lowest stator temperature of 32°C. has the same induced phase voltages and torque output values.

2584 IET Electr. Power Appl., 2020, Vol. 14 Iss. 13, pp. 2577-2586
© The Institution of Engineering and Technology 2020
17518679, 2020, 13, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-epa.2020.0622 by Readcube (Labtiva Inc.), Wiley Online Library on [07/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Fig. 17 Thermal view of the windings at full load after 5 min
(a) Arc winding, (b) Radial winding, (c) Parallel winding, (d) Concentric winding, (e) Unequal width parallel winding

However, it is not the most optimal winding type; if low THD


levels are desired, as radial winding has fewer harmonics compared
to arc winding. The parallel winding has the smallest copper loss
and in comparison with concentric, radial and arc windings. Its
efficiency is found as 68% which makes it suitable in applications
where the cooling of the stator is a challenge. Unequal width
parallel winding has even less copper loss than parallel winding
that increases the efficiency of the motor to 72%. Unequal width
winding design can be used in other wave winding topologies to
reduce the phase resistance and total copper loss by utilising the
PCB surface area better.

Fig. 18 Temperature change of the stators under full load


5 Conclusion
Thanks to the double rotor-single stator motor configuration, Five different PCB windings, namely concentric, radial, arc,
thermal stress on the stator windings is dissipated with the airflow parallel and unequal width parallel, are compared in this paper for
created by the rotors rotating at high speeds. If the motor is put into a 16-pole, 2000-RPM axial-flux permanent magnet synchronous
a housing such that it blocks the thermal convection, the machine. Analytical calculation methods of induced phase voltages
temperature of the stator is going to increase. A thermal conduction and output torque of wave windings are presented and validated
path can be constructed to cool the windings via extending the with finite-element analysis tools and experimental results. Radial
copper traces of the windings to outside in order to have a thermal winding and concentric winding showed the highest torque and
connection with housing. induced phase voltages. However, the concentric winding had the
The results showed that radial winding has a high induced highest loss and THD levels in comparison with other winding
phase voltage and output torque with 72% efficiency. Moreover, types. Parallel winding has the smallest resistance in comparison
THD is the lowest compared to other winding types which makes it with concentric, radial and arc windings. A novel unequal width
suitable for high precision applications. Concentric winding has parallel winding has been proposed and compared with parallel
similar induced phase voltage and output torque but the copper loss winding which resulted in 17% smaller phase resistance than
is 24% higher than radial winding which reduces the efficiency to parallel winding while holding the same induced phase voltage and
66%. Note that, concentric winding has the highest phase output torque.
inductance in comparison with other windings. Higher inductance
can be desirable in the motoring mode of operation since it 6 References
decreases the ripple in phase currents. So, concentric winding can
be used in applications that require lower current ripples in [1] Chalmers, B.J., Spooner, E.: ‘An axial-flux permanent-magnet generator for a
gearless wind energy system’, IEEE Trans. Energy Convers., 1999, 14, (2),
motoring mode. Arc winding has lower induced voltage and output pp. 251–257, ISSN 1558-0059, doi: 10.1109/60.766991
torque than radial and concentric windings. Moreover, its [2] Latoufis, K., Troullaki, K., Pazios, T., et al.: ‘Design of axial flux permanent
efficiency is the lowest of all with 66% which is same as concentric magnet generators using various magnetic materials in locally manufactured
winding's efficiency. THD of arc winding is low as stated in [17]. small wind turbines’. 2016 XXII Int. Conf. on Electrical Machines (ICEM),
Lausanne, Switzerland, 2016, pages pp. 1545–1551

IET Electr. Power Appl., 2020, Vol. 14 Iss. 13, pp. 2577-2586 2585
© The Institution of Engineering and Technology 2020
17518679, 2020, 13, Downloaded from https://ietresearch.onlinelibrary.wiley.com/doi/10.1049/iet-epa.2020.0622 by Readcube (Labtiva Inc.), Wiley Online Library on [07/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
[3] Pop, A.A., Radulescu, M.M., Balan, H., et al.: ‘Electromagnetic design and [16] Wang, X.-y., Li, X., Li, C.-p., et al.: ‘Design of a PCB stator coreless axial
finite-element analysis of an axial-flux permanent-magnet machine’. 2014 Int. flux permanent magnet synchronous motor based on a novel topology halbach
Conf. on Applied and Theoretical Electricity (ICATE), Craiova, Romania, array’, Front. Inf. Technol. Electron. Eng., 2019, 20, (3), pp. 414–424, ISSN
2014, pages pp. 1–4 2095-9230, doi: 10.1631/FITEE.1700345
[4] Oh, S.C., Emadi, A.: ‘Test and simulation of axial flux-motor characteristics [17] Neethu, S., Nikam, S.P., Wankhede, A.K., et al.: ‘High speed coreless axial
for hybrid electric vehicles’, IEEE Trans. Veh. Technol., 2004, 53, (3), pp. flux permanent magnet motor with printed circuit board winding’. 2017 IEEE
912–919, ISSN 1939-9359, doi: 10.1109/TVT.2004.827165 Industry Applications Society Annual Meeting, Ohio, USA, 2017, pages pp.
[5] Nguyen, T.D., Tseng, K., Zhang, S., et al.: ‘A novel axial flux permanent- 1–6
magnet machine for flywheel energy storage system: design and analysis’, [18] Neethu, S., Nikam, S.P., Pal, S., et al.: ‘Performance comparison between
IEEE Trans. Ind. Electron., 2011, 58, (9), pp. 3784–3794, doi: 10.1109/ PCB-stator and laminated-core-stator-based designs of axial flux permanent
TIE.2010.2089939 magnet motors for high-speed low-power applications’, IEEE Trans. Ind.
[6] Wen, Z., Xiong, B., Gu, G.: ‘Optimization design of low speed axial flux Electron., 2020, 67, (7), pp. 5269–5277
halbach permanent-magnet generator with pcb winding’. 2019 22nd Int. Conf. [19] Wang, X., Pang, W., Gao, P., et al.: ‘Electromagnetic design and analysis of
on Electrical Machines and Systems (ICEMS), Harbin, People's Republic of axial flux permanent magnet generator with unequal-width PCB winding’,
China, 2019, pages pp. 1–4 IEEE Access, 2019, 7, pp. 164696–164707
[7] Moury, S., Iqbal, M.T.: ‘A permanent magnet generator with pcb stator for [20] Zhu, Z.Q., Howe, D., Bolte, E., et al.: ‘Instantaneous magnetic field
low speed marine current applications’. 2009 1st Int. Conf. on the distribution in brushless permanent magnet dc motors. i. Open-circuit field’,
Developements in Renewable Energy Technology (ICDRET), Dhaka, IEEE Trans. Magn., 1993, 29, (1), pp. 124–135, doi: 10.1109/20.195557
Bangladesh, 2009, pages pp. 1–4 [21] Furlani, E.P.: ‘Computing the field in permanent-magnet axial-field motors’,
[8] Tsai, M., Hsu, L.: ‘Design of a miniature axial-flux spindle motor with IEEE Trans. Magn., 1994, 30, (5), pp. 3660–3663, doi: 10.1109/20.312734
rhomboidal pcb winding’, IEEE Trans. Magn., 2006, 42, (10), pp. 3488–3490 [22] Chan, T.F., Lai, L.L., Xie, S.: ‘Field computation for an axial flux permanent-
[9] Jang, G.H., Chang, J.H.: ‘Development of an axial-gap spindle motor for magnet synchronous generator’, IEEE Trans. Energy Convers., 2009, 24, (1),
computer hard disk drives using pcb winding and dual air gaps’, IEEE Trans. pp. 1–11, doi: 10.1109/TEC.2008.2011830
Magn., 2002, 38, (5), pp. 3297–3299 [23] Cakal, G., Keysan, O.: ‘Design of double sided linear motor with easy to
[10] Chuo, Y.-C., Wang, C.-C., Liu, C.-S., et al.: ‘Development of a miniature manufacture hairpin plate winding’. 2019 12th Int. Symp. on Linear Drives
axial-field spindle motor’, IEEE Trans. Magn., 2005, 41, (2), pp. 974–976 for Industry Applications (LDIA), Neuchâtel, Switzerland, 2019, pages pp. 1–
[11] Jang, G.H., Chang, J.H.: ‘Development of dual air gap printed coil bldc 5
motor’, IEEE Trans. Magn., 1999, 35, (3), pp. 1789–1792 [24] Parviainen, A., Niemela, M., Pyrhonen, J.: ‘Modeling of axial flux
[12] Paul, S., Farshadnia, M., Pouramin, A., et al.: ‘Comparative analysis of wave permanent-magnet machines’, IEEE Trans. Ind. Appl., 2004, 40, (5), pp.
winding topologies and performance characteristics in ultra-thin printed 1333–1340, doi: 10.1109/TIA.2004.834086
circuit board axial-flux permanent magnet machine’, IET Electr. Power Appl., [25] Azzouzi, J., Barakat, G., Dakyo, B.: ‘Quasi-3-d analytical modeling of the
2019, 13, (5), pp. 694–701 magnetic field of an axial flux permanent-magnet synchronous machine’,
[13] Taqavi, O., Mirimani, S.M.: ‘Design aspects, winding arrangements and IEEE Trans. Energy Convers., 2005, 20, (4), pp. 746–752, doi: 10.1109/
applications of printed circuit board motors: a comprehensive review’, IET TEC.2005.845538
Electr. Power Appl., 2020, 14, (9), pp. 1505–1518 [26] Fei, W., Luk, P.C.K., Jinupun, K.: ‘Design and analysis of high-speed coreless
[14] Wang, X., Hu, C., Zhao, M., et al.: ‘Design of multi-layer PCB coreless axial axial flux permanent magnet generator with circular magnets and coils’, IET
permanent magnet synchronous motor’. 2019 22nd Int. Conf. on Electrical Electr. Power Appl., 2010, 4, (9), pp. 739–747
Machines and Systems (ICEMS), Harbin, People's Republic of China, 2019, [27] Fei, W., Luk, P.C.K., El-Hasan, T.S.: ‘Rotor integrity design for a high-speed
pages pp. 1–4 modular air-cored axial-flux permanent-magnet generator’, IEEE Trans. Ind.
[15] Cupertino, F., Ettorre, S.: ‘Torque production capabilities of electrical Electron., 2011, 58, (9), pp. 3848–3858
machines with planar windings’. IECON 2012 - 38th Annual Conf. on IEEE
Industrial Electronics Society, Montreal, Canada, 2012, pages pp. 2080–2085

2586 IET Electr. Power Appl., 2020, Vol. 14 Iss. 13, pp. 2577-2586
© The Institution of Engineering and Technology 2020

You might also like