October 4, 2004
by Will Robertson
October 4, 2004
This document has been typeset with the LATEX document preparation
system using Peter Wilson’s Memoir class.
Page margins have been chosen as a trade-off between achieving
the optimal number of characters per line for ease of reading† , and
trying to fit the typeblock onto the poorly-sized (for books), although
convenient, a4 stock.‡ pdfTEX’s margin kerning via character protru-
sion is used to ensure optical straightness of these margins.
pdf hyperlinks have been automatically inserted in all appropriate
positions: at footnote, page number, bibliographical, figure, and table
references. Furthermore, bibliography entries are appended with page
numbers for convenience.
The fonts used in the document are Prof. Hermann Zapf’s Aldus at
11 pt for body text, Palatino for the titles, and Euler for maths. Figure
labels are set in Bogusław Jackowski’s Latin Modern Sans at various
design sizes (based of course on Prof. Don Knuth’s Computer Modern
Sans). This same font is used on the title page, and the embellished
author name is in Hermann Zapf’s Zapfino.
†
Approximately 66.
‡
For reference, see The Elements of Typographic Style by Robert Bringhurst.
Abstract
This document is the eighteen month progress report required for the continu-
ation of the author’s Ph. D. candidature in the School of Mechanical Engineering,
University of Adelaide. The research aim is to design and build a vibration isol-
ation table using non-contact magnetic springs; the progress of the project is de-
tailed within.
Vibration isolation tables are using in situations that require very precise con-
trol of the environment, in which outside disturbances can have a significant det-
rimental effect (for example, photo-lithography in the semiconductor industry).
Commercial isolation tables use pneumatic springs that still transmit vibration at
very low frequencies; with non-contact support, it is believed that greater perform-
ance may be obtained.
It has been found that non-contact vibration isolation tables, in general, have
not yet been built to support large loads. The literature demonstrates using Hal-
bach arrays to increase the flux of magnetic structures; in this work, an analysis
of using such arrays in opposition as springs is undertaken, with preliminary re-
search providing some surprising results.
The basic geometry of the proposed magnetic spring design is presented, and
shown how it can be adapted to accommodate Halbach arrays in order to increase
the load bearing ability of the isolator.
Finally, the details of the equipment to be used for the forthcoming prototype
experiments is covered and the testing plan explained.
iii
Table of Contents
1 Introduction 1
1.1 Project exposition 1
1.2 Document structure 2
3 Understanding magnets 13
3.1 Sources of the magnetic field 13
3.2 Properties of magnetic flux 15
3.3 Analytic equations 16
3.4 Magnetic materials 18
iv
4.4.5 Arrays for the prototype 29
4.5 Other considerations 29
5 Prototype considerations 32
5.1 Magnets 32
5.2 Forces between magnets 32
5.3 Actuators 34
5.3.1 Prototype electromagnets 34
5.3.2 Actuator placing 35
5.4 Sensors 35
5.4.1 Prototype sensor 36
5.5 Prototypes and testing 36
6 Conclusions 39
6.1 Future work 39
List of Tables
v
List of Figures
vi
1 Introduction
This chapter introduces the project and describes its aims and its relevance. It
concludes with a description of the document structure.
†
http://www.newport.com/
1
2 Section 1.2: Document structure
plication. An efficient and scalable magnetic design has thus far been overlooked:
this project attempts to rectify the situation.
Boerdijk (1956a)
This chapter is a review of the literature in the field of magnetic levitation. The
chapter concludes with the selection of papers that relate specifically to tables that
float on magnets.
3
4 Section 2.1: Levitation techniques
The proof is conceptually quite simple. We start with the equation for the mag-
netic field; when there are no external current terms, it can be shown to be ex-
pressed as Laplace’s equation:
∇ · B = 0 =⇒ ∇2 B = 0. (2.1)
d2 U d2 U d2 U
∇2 U = + + 2 = 0. (2.2)
dx2 dy2 dz
The double differentiations of the energy are the stiffnesses in each direction. But
for stable equilibrium, these three terms must be greater than zero. This cannot
satisfy Equation (2.2) and thus levitation cannot occur.
that causes the levitation. The technique uses a large amount of power, and is not
especially suitable for the purposes of this research for this reason.
For an actively stabilised levitation system, the levitation forces are created by
permanent magnets (which have dc magnetic fields) and the necessary stabilisa-
tion applied with variable dc electromagnets with a feedback control system. For
this reason the magnetic fields are known as quasi-static. This system was first
implemented by Holmes who levitated a magnetic needle, as cited by Boerdijk
(1956b). Some more practical examples of these types of system are covered in
the next section.
This summary is fairly brief; Bleuler (1992) wrote a more detailed overview.
His paper introduces the ‘self-sensing amb’ (Vischer and Bleuler, 1993) which uses
back–electromotive force from the controlling electromagnet to sense the position
of the floating element. This eliminates the need for a more classical position
sensor (for example, optical or capacitive), but the control system is necessarily
more complex and the behaviour not as precise.
Stator Rotor
magnet magnet
Shaft
N S S N N S S N
Figure 2.2: The cross-section of two radially magnetised ring magnets in a radial
bearing.
Rotor Stator
magnet magnet
Repulsion
Rotor Stator
magnet magnet
Repulsion
Figure 2.3: Two equivalent radial bearings (with equal forces of repulsion), despite
their different directions of magnetisation.
8 Section 2.2: Non-contact applications of magnetics
His later paper (Yonnet, 1981) continues this work, describing how magnetic
bearings may re-arranged to suit different applications, showing every possible
simple equivalent axial and radial bearing.
Figure 2.6: The magnetic flux lines of a Halbach array made from magnets with
45° magnetisation increments (indicated by the arrowheads). The magnetic field
can be seen to be much stronger above the array than below it.
S S
N S N
N
N
S S
(a) Planar array created by the superposition of two (b) The planar array by Cho et al. (2001)
orthogonal linear Halbach arrays. Non-solid arrow- (top view). Flux travels out of the page from
heads correspond to diagonal directions of magnetisa- the north- to south-faced magnets, and back
tion. through the array in the triangular magnets.
control system used was quite advanced, utilising a combination of two independ-
ent control systems for stable levitation and robust vibration isolation. The mag-
netic actuator design is not described, however. It is believed that high-powered
electromagnets were required for this design; with the current availability of cheap
rare-earth magnets, a more efficient design is now possible.
Chang (2001) applied non-linear control to the problem of magnetic levitation,
using coupled hybrid magnets (that is, electromagnets biased with permanent mag-
net cores) that create a magnetic circuit with the levitated table of 20 kg. The paper
looks at non-linear analysis and neither passive nor active vibration isolation res-
ults are shown.
More recently, Choi et al. (2003) have designed a levitation table capable of
supporting 15 kg, that is unstable in only one horizontal degree of freedom. How-
ever, the magnetic arrangement used for the spring, despite their claims, appears
quite unstable and is not suited for scaling up to bear greater loads. Their experi-
ments prove that only a single axis requires control for stability, but the position
resolution they achieve is fairly coarse.
A novel design for achieving a theoretically infinite stiffness spring is shown
by Mizuno et al. (2003a,b). By using physical springs in series with electromag-
netic suspension springs, three degrees of freedom of an isolation platform are
actively controlled, with the total weight supported around 30 kg. The unstable de-
gree of freedom is in the vertical direction, however, which may prove problematic
when increasing the load on the table.
12 Section 2.4: Non-contact magnetic support
The magnetisation of a permanent magnet creates the magnetic fields that are
of such great interest. Inside the magnet (with no other external fields present),
the magnetic field, B, is given by the simple relation:
B = µ0 M, (3.2)
Jm = ∇ × M. (3.3)
This is a good beginning for describing the effects of an external current density
(J) acting on the magnet. To separate the effects of induced magnetisation and that
13
14 Section 3.1: Sources of the magnetic field
J = ∇ × H. (3.4)
Now the earlier Equation (3.2) can be adjusted to allow for both internal and ex-
ternal forms of magnetisation (that is, magnetic field caused by permanent mag-
nets or by current carrying conductors). This is the fundamental equation relating
the three important terms in magnetics,
B = µ0 (M + H), (3.5)
Equation (3.5), may now be used to describe the situation at all points in space.
Use Figure 3.1 to note that while inside the magnet, the magnetic field is the vector
sum of two components, whereas outside the magnet, the magnetisation is zero
and the magnetic field is related to the magnetic field strength by a constant. This
results in B being continuous everywhere, and both M and H being discontinuous.
Magnet Air
µ0 M
µ0 H
µ0 H
B B
Figure 3.1: The magnetic field, B, both inside and outside a magnet.
This equivalence in air is essentially the reason that there is often confusion
between B and H. It can be seen that within a magnet, however, their relationship
is more complex and important. The performance of a magnet is shown by its B-H
curve, which is shown for an ideal magnet in Figure 3.2. This curve demonstrates
†
The names of these terms are not always consistent in the literature. M is also known as
polarisation, and B and H are both sometimes known as the magnetic field.
Chapter 3: Understanding magnets 15
the non-linear and hysteresic effects of the magnetic flux density of a magnetic
material as external magnetic field is applied to it.
Two important features are shown in the B-H curve. First, the remanence
of the magnet, Br . This value is equal to µ0 Msat and occurs when there is no
external magnetic field. The other is the coercive force, Hc , which is the amount
of magnetic field strength required to reduce the flux density of the magnet to
zero.
B
Br
-Hc
H
Figure 3.2: The characteristic B vs. H curves for an ideal rare-earth magnet.
ous ether’. Nowadays, scientists tend toward more modern interpretations using
electromagnetic fields involving quantum theory. Nonetheless, the name sticks.
Magnetic flux, Φ, is therefore defined as the amount of ‘fluid’ passing through an
area:
Φ=B·A (3.8)
This flux is almost analogous to electric current; the only difference being that
electric current is constrained by the conductor it is flowing through, whereas
while magnetic flux is known to prefer areas of greater permeability, it occasionally
can deviate from these simple paths
It is more instructive for a basic understanding of how magnets behave to look
at the ways their flux lines interact. The following ‘magnet design axioms’ are
adapted from Moskowitz (1995), whose book covers permanent magnet design for
a wide range of uses.
1. Flux lines follow the path of least resistance. This means that they will travel
through the shortest path possible, through the material with the greatest
permeability—so they will travel more readily through magnetic or ferrous
material than air, and more readily through air (although only slightly) than
diamagnetic material.
2. Flux lines travelling in the same direction repel each other. This means flux
lines will never cross.
4. Permeability of ferrous material is ‘used up’ by flowing flux; when the ma-
terial reaches saturation, flux lines travel as easy though air as through the
saturated material.
From these axioms, one can generate incorrect, yet applicable, theories how
and why magnets attract and repel each other. For example, two magnets in repul-
sion have flux lines opposing each other. It can be imagined that the reason forces
occur between them is due to a ‘squashing’ of the flux lines which the magnets try
to oppose—but theories like this only help visualising magnetic behaviour, not for
explaining the reasons behind it.
still very much a work-in-progress, for the vector calculus can be fairly involved
and for preliminary results, it is quicker to use a finite element analysis. This con-
trasts with the opinions of the authors from years past, whose computers were not
powerful enough to solve complex feaproblems within a reasonable time frame.
Akoun and Yonnet (1984) derived an analytic solution to the problem of two
cube magnets with parallel magnetisation. Note that this geometry allows for at-
traction when the magnetisation direction is in the direction of the offset between
the magnets, and repulsion when perpendicular—essentially any configuration is
possible.
It is useful to include their equation in the text so that its complexity may be
somewhat appreciated.
X
1
−J1 J2 φx
Fx = · (−1)i+j+k+l+p+q
4πµ0
i,j,k,l,p,q=0
X
1
J1 J2 φy
Fy = · (−1)i+j+k+l+p+q
4πµ0
i,j,k,l,p,q=0
X
1
J1 J2 φz
Fz = · (−1)i+j+k+l+p+q (3.9)
4πµ0
i,j,k,l,p,q=0
where
uv
1
v2 −w2 ln (r−u) + uv ln (r−v) + vw arctan + 21 ru
φx = 2 rw
uv
φy = 12 u2 −w2 ln (r−v) + uv ln (r−u) + uw arctan + 21 rv
uv rw
φz = −uw ln (r−u) − vw ln (r−v) + uv arctan − rw (3.10)
rw
and
For these equations, 2a, 2b and 2c are the dimensions of the first magnet; 2A, 2B,
2C dimensions of the second; α, β, and γ are the distances between their centres.
Furlani (1993) derived similarly complex equations for the exact axial forces
between two ring magnets. A different technique that has been used for model-
ling magnetic forces was shown by Bancel (1999), who introduces the concept of
magnetic nodes, which may be used to significantly reduce the complexity of the
equations of forces between magnets. This technique looks very promising for cre-
ating parametrically optimisable equations of force for arbitrary geometries, and
it is planned to investigate this method for this project.
18 Section 3.4: Magnetic materials
Magnet type
Property Ferrite Alnico Neodymium
Max. temperature (°C) 400–500 800–900 80–200
Remanence (T) 0.2–0.4 0.5–1.3 1–1.3
Coercivity (kA/m) 100–200 50–160 800–900
Max energy product (kJ/m3 ) 6–33 10–80 200–300
Table 3.1: Typical values for various permanent magnets. Adapted from informa-
tion from http://www.magtech.com.hk/.
4 Non-contact magnetic spring design
This chapter deals with the geometry of the proposed magnetic spring design.
19
20 Section 4.2: Simple magnet arrangement
radial and thrust bearings shown by Yonnet (1981), an improved design is shown
in Figure 4.2. Now the spring forces are caused by horizontal magnets in attrac-
tion, in contrast to the earlier arrangement shown in Figure 4.1 which uses repuls-
ive force.) Two fixed outer magnets both attract a centred floating magnet, so the
unstable degree of freedom is in that horizontal direction. Perturbations in the
other horizontal direction are restored by the aforementioned attraction, which is
similarly the cause of the positive vertical spring stiffness. Note that the diagram
works as both a top and side view, demonstrating how it is unstable in only one
direction.
Displacement force
Equilibrium forces
Spring forces
Figure 4.2: A ‘horizontal’ spring with attracting magnets to create vertical stiff-
ness, as shown by the reaction forces after displacement.
60
V Spring Vertical Spring:
H Spring (Gap = 0.1)
50
H Spring (Gap = 0.5)
H Spring (Gap = 1)
40
Vertical force/(N)
30
20
10 Horizontal Spring:
Gap
0
Figure 4.3: Forces from fea comparing the simple vertical and horizontal mag-
netic springs. The initial gap between the magnets in the vertical spring has been
arbitrarily set to 3 magnet widths.
nets: one below, which provides the majority of the stiffness of the spring, and
two aside, for stabilising one degree of freedom in the ‘into the page’ direction.
This simple design gives a basic framework from which to build up more complex
arrangements.
Restoring forces
Stabilising
magnets
Support magnet
Figure 4.4: Combination spring, with the strength of the vertical spring and the
stability of the horizontal spring.
22 Section 4.3: Halbach arrays
flux linkage for reluctance actuators, not for generating force; a notable separation
exists here.
The ‘patchwork’ array shown in Figure 4.6(a) is the most simple. See Hinds
cited by Kim (1997), or Chitayat cited by Cho et al. (2001) for its use in other
circumstances. The array used by Kim is shown in Figure 4.6(b), despite the fact
that he demonstrated the Halbach superposition as shown in §2.3. → page 9
N S N
S N S
N S N
(a) ‘Patchwork’ array. (b) Extended Halbach array.
Figure 4.6: Two simple planar magnetic arrays: (a) shows the planar analogue to
Backers’ arrangement; (b) is a Halbach array extended into a planar structure.
M(i, j) = Mx x̂ + My ŷ + Mz ẑ (4.1)
where i and j are the magnet numbers in the x̂ and ŷ directions, respectively. This
equation is redefined to remove the constant magnetisation magnitude, M, from
24 Section 4.4: Forces between arrays
Figure 4.7: Opposing Halbach arrays in the strong and weak fields.
y x
2 2
1 1
j=0 i=0
Figure 4.8: Arbitrary planar array with cuboid magnets. i and j are the magnet
numbers in the x and y directions.
Chapter 4: Non-contact magnetic spring design 25
each term:
Note that this is the magnetisation of only the bottom array; the magnetisation of
the opposite, repulsive, array can be generated by inverting the magnetisation in
the facing direction: M̂z2 = −M̂z .
For simple ‘patchwork’ magnet arrangement as in Figure 4.6(a), the magnetisation → page 23
directions are:
M̂x = 0,
M̂y = 0,
M̂z = cos (iπ) cos (jπ). (4.5)
For the case when the linear Halbach array (with 90° rotations) is extended into
the third dimension as in Figure 4.6(b), the magnetisation of the array is: → page 23
The forces between facing pairs of these three arrays is shown in Figure 4.9. For
this graph, and subsequent ones, 5 × 5 arrays of half-inch rare-earth cube magnets
have been used . Surprisingly, the patchwork array did not perform better than
the homogeneous array except in close proximity. The finite element analysis
in ANSYSwas capable of generating solutions for distances equal to 0.3 magnet
widths between the arrays; closer than this and the number of nodes required for
the solution surpassed the capacity of the department’s license for the software.
1200
Single magnet
Patchwork
1000
Extended 2D
800
Vertical Force/(N)
600
400
200
0 0.5 1 1.5 2
Vertical displacement/magnet width
given in two ways. A strict superposition, as performed by Kim, would simply add
the magnetisations of two orthogonal Halbach arrays, yielding:
However, an inspection of the magnetisations shows that the magnets are now sig-
nificantly stronger than for the other arrays that have been looked at. This is not
particularly convenient for comparing array performance. To make a comparison
worthwhile, some care must be taken to ensure the normalised magnetisations of
each block are equal and within the same range (0 6 M̂x,y,z 6 1) as the other
arrays already shown:
The first array will be referred to as a ‘Halbach addition’, whereas the second will
be called a ‘Halbach superposition’. Figure 4.10 demonstrates the significant differ-
ences between the two, showing the forces with the extended Halbach array (refer
to Equation (4.6)) for comparison. The simple addition of two orthogonal arrays
Chapter 4: Non-contact magnetic spring design 27
2400 Extended 2D
Halbach superposition
2100
Halbach addition
1800
Vertical Force/(N)
1500
1200
900
600
300
0
0 0.5 1 1.5 2
Vertical displacement/magnet width
Figure 4.10: Forces between planar arrays: Halbach addition and Halbach super-
position.
.
M̂i,j = 41 (1 − (−1)j ) M̂V (1 − (−1)i ) − M̂H (1 − (−1)i+1 )
+ 41 (1 − (−1)j+1 ) M̂H (1 − (−1)i ) − M̂V (1 − (−1)i+1 ) (4.9)
28 Section 4.4: Forces between arrays
where
→ page 11 It may be helpful to refer to Figure 2.7(b) which shows the full magnetisation dir-
ections for the triangular blocks. Force analysis is yet to be performed comparing
this array to the others.
2 0
-MH -MV -MH -MV 1
i=0,1,2,3,...
Figure 4.11: Cho’s array, decomposed into repeating elements.
The second design, Figure 4.12(b), was developed by extending two Halbach
arrays in orthogonal directions starting from an initial magnet and filling in the
Chapter 4: Non-contact magnetic spring design 29
N N
S S S S S
N N N N N N N N
S S S S S
N N
(a) A proposed design with rotated lin- (b) Other proposed design
ear Halbach arrays.
Figure 4.12: Two proposed planar arrays with 90° angles of magnetisation.
Magnetic spring
Stabilising
magnet arrays
Supporting
magnet array
Delamare et al. (1994) look at this problem and manage to remove the rota-
tional instability by adding a weaker, axial bearing to the system—that is, oppos-
ite in effect to the main aims of the system. The coupling of the axial and radial
bearings eliminates the rotational instability, but the strength of the radial bearing
bearing overpowers the axial bearing for normal operation.
This is more easily explained with a diagram. Figure 4.14 shows how this
idea can be applied to a simplified model. Attracting magnets close to the centre
→ page 20 of rotation provide the forces for the spring (refer to Figure 4.2). Further away,
repelling magnets are placed such that the translational forces they apply are less
than the attractive forces of the inner magnets (see Figure 4.14(a)). However. their
strengths and distances away have been carefully chosen so that the rotational
forces (moments) they apply are greater than the moments applied by the inner
magnets (see Figure 4.14(b)). The system my be represented by the following
equations for each equal and opposite magnet pair. For the forces:
D N
S d
(Stabilising) (Unstable)
forces forces
N N
Smaller forces
Finner Fouter
due to increased
S distance
N
(a) The added magnets are spaced farther away, so they do
not affect the translational stability.
N
S
Unstable Stabilising
moments moments
N N
Minner Greater moments
due to increased
Mouter
S radius
N
r
R
(b) The added magnets do affect the radial stability
because their distance from the centre of rotation is
much greater.
Figure 4.14: Secondary (outer) magnets may be added to add stability in the rota-
tional direction.
5 Prototype considerations
This chapter covers the implementation details of the prototype spring, looking at
how the forces in the spring will be estimated and secondly at the actuators and
sensors that will be used. Finally, the testing plan is briefly covered.
5.1 Magnets
Magnets are required in order to build a magnetic spring! K & J Magnetics† is
an American supplier of surplus rare earth magnets. They offer a large range of
sizes, for example, from one-eighth inch cubes (100 for us$8.20) all the way up
to one inch cubes, which sell for us$14.25 each. (These prices are very good.) For
the initial stages of the project, 25 half-inch magnets were purchased from them
for about us$45. They offer bulk discounts for when many more magnets will be
required in order to build up the planar arrays for the supports.
†
http://www.kjmagnetics.com/
32
Chapter 5: Prototype considerations 33
20
15
Vertical force/(N)
10
−5
2 2
1.5 1.5
1 1
0.5 0.5
0.1 0
Norm. H. distance Norm. V. distance
(a) Analytic forces between two half-inch cube magnets.
20
15
Vertical Force/(N)
10
−5
2 2
1.5 1.5
1 1
0.5 0.5
0.1 0
Normalised H. dist. Normalised V. dist.
(b) fea forces between two half-inch cube magnets.
5.3 Actuators
The next aspect of the prototype design involves the arrangement of the electro-
magnets that will be used to apply variable non-contact forces on the spring. Two
separate functions are needed: stability of levitation, and rejection of vertical dis-
turbances. These two forces are orthogonal in direction and functionality.
64 64
Force/(N) − log scale
32 32
16 16
8 8
4 4
2 2
1 1
0.5 0.5
‡
One of the disadvantages of a strict budget
Chapter 5: Prototype considerations 35
5.4 Sensors
To keep the magnetic spring centred, it is necessary to use sensors to detect where
the floating component is spatially located so that deviations may be corrected.
Clearly, non-contact sensors are required for a non-contact spring. In this capacity,
there are four main choices:
Ultra-sonic These sensors work by sending out modulated ultra-sonic pulses to
be reflected off the target. The time spent in the round trip gives a linear
indicator of the distance. Ultra-sonic sensors are cheap, but slow and inac-
curate. They are unsuited for use in this project.
Inductive An inductive, or Hall effect, sensor works by exciting a coil with a high
frequency ac current which induces eddy currents in the target. These eddy
currents may be measured very accurately, but the whole effect is very de-
pendent on a lack of magnetic noise. This makes these type of sensors diffi-
cult to use in magnetic applications.
Capacitive A capacitive sensor measures the capacitance between a plate and the
target. It can be very accurate and quite fast. To measure large distances,
however, a large capacitive head is required (approx. 1 cm diameter for every
1 mm of range.) They are also very expensive due to the conditioning elec-
tronics required.
36 Section 5.5: Prototypes and testing
Laser A laser sensor uses interferometry to calculate position of the target. For
their price, they offer very good accuracy and speed; also, they do not suffer
from electrical noise.
See Boehm et al. (1993) for a complete overview. At this stage in the project,
monetary restrictions weigh most heavily on the choice of sensor. Several sensors
suitable for this project are sold by the German company Micro-Epsilon†† , whose
products are re-sold in Australia by Bestech‡‡ . A selection of commercial sensors
are shown in Table 5.1.
Sensor type
Laser Capacitive Inductive
Range (mm) 6.35 ≈4 4
Resolution (µm) 1.9 0.4 20
Price $2500 $4500 “low-cost”
16
Steel
Log curve fit
8 Aluminium
Log curve fit
Voltage/(V) − log scale
0.5
0.25
0 0.25 0.5 0.75 1 1.25 1.5 1.75 2
Distance/(mm)
simple and become more complex as each variation is shown to work. A conceptual
diagram of the prototype is shown in Figure 5.4. The stabilising magnet arrays are
set up so they can be offset both horizontally and vertically from the edge of the
floating spring in order to reduce the forces they apply.
The initial spring will be levitated and supported only with single cube mag-
nets arranged horizontally as in Figure 4.2. This model will test the performance → page 20
of both the sensors and the actuators. It is intended that once successful, the
individual magnets will then be linked up in linear Halbach arrays, still in the ho-
rizontal configuration. Finally, the lower support magnets will be put into place
and the whole system tested as a passive vibration isolator.
Once these tests are complete, a whole table can be manufactured with the final
spring design, which is expected to be iteratively improved during the prototype
stage. At this time vertical actuators can be included in the system and a fully
active control system developed to attenuate the ground vibrations.
38 Section 5.5: Prototypes and testing
FLOATING SPRING
Electro- Electro-
Steel surface
magnet for attraction forces
magnet
Aluminium
Sensor
surface for sensor
Sensor
Electromagnets apply
Electro- variable force to keep the Electro-
magnet spring centred. magnet
39
40 Section 6.1: Future work
Sept.
FEA Modelling Progress report
Basic geometry
2005
Horizontal linear arrays
Paper — spring design
Planar array integration
Construction
Thesis
Testing
2006
Akoun G. and Yonnet J.P. (1984): 3D analytical calculation of the forces exer-
ted between two cuboidal magnets. IEEE Transactions on Magnetics, MAG-
20(5):1962–1964. (Cited on page 17)
Berry M.V. and Geim A.K. (1997): Of flying frogs and levitrons. The European
Physical Society, 18:307–313. (Cited on page 5)
Boehm J., Gerber R. and Kiley N.R.C. (1993): Sensors for magnetic bearings. IEEE
Transactions on Magnetics, 29(6):2962–2964. (Cited on page 36)
Boerdijk A.H. (1956a): Levitation by static magnetic fields. Philips Technical Re-
view, 18:125–127. (Cited on pages 3 and 5)
Chang S.C. (2001): Dynamics of a magnetically levitated table with hybrid mag-
nets. The Japan Society of Applied Physics, 40(10):6163–6170. (Cited on page 11)
Cho H.S., Im C.H. and Jung H.K. (2001): Magnetic field analysis of 2-D per-
manent magnet array for planar motor. IEEE Transactions on Magnetics,
37(3):3762–3766. (Cited on pages 10, 11, 23, 27, and 28)
41
42
Choi K.B., Cho Y.G., Shinshi T. and Shimokohbe A. (2003): Stabilization of one
degree-of-freedom control type levitation table with permanent magnet repuls-
ive forces. Mechatronics, 13(6):587–603. (Cited on page 11)
Delamare J., Yonnet J.P. and Rullière E. (1994): A compact magnetic suspension
with only one axis control. IEEE Transactions on Magnetics, 30(6):4746–4748.
(Cited on page 29)
Earnshaw S. (1842): On the Nature of the Molecular Forces which regulate the
Constitution of the Luminiferous Ether. Transactions of the Cambridge Philo-
sophical Society, 3(1):97–112. (Cited on page 3)
Furlani E.P. (1993): A formula for the levitation force between magnetic disks.
IEEE Transactions on Magnetics, 29(6):4165–4169. (Cited on page 17)
Halbach K. (1981): Physical and optical properties of rare earth cobalt magnets.
Nuclear Instruments and Methods, 187:109–117. (Cited on page 9)
Kim W.j. (1997): High-Precision Planar Magnetic Levitation. Ph.D. thesis, Mas-
sachusetts Institute of Technology. (Cited on pages 8, 10, 23, 25, and 28)
Ma K.B., Postrekhin Y.V. and Chu W.K. (2003): Superconductor and mag-
net levitation devices. Review of Scientific Instruments, 74(12):4989–5017.
(Cited on page 5)
Nagaya K., Atsumi M. and Endoh M. (1993): Simplified vibration isolation con-
trol using disturbance cancellation and velocity feedback controls for a magnetic
levitation table. International Journal of Applied Electromagnetics in Materials
4, 4(4):115–122. (Cited on page 10)
Puppin E. and Fratello V. (2002): Vibration isolation with magnet springs. Re-
view of Scientific Instruments, 73(11):4034–4036. (Cited on page 10)
Simon M.D. and Geim A.K. (2000): Diamagnetic levitation: Flying frogs and float-
ing magnets. Journal of Applied Physics, 87(9):6200–6204. (Cited on page 5)
Simon M.D., Heflinger L.O. and Geim A.K. (2001): Diamagnetically stabilized
magnet levitation. American Journal of Physics, 69(6):702–713. (Cited on page 5)
Trumper D.L., Kim W.j. and Williams M.E. (1996): Design and analysis frame-
work for linear permanent-magnet machines. IEEE Transactions on Industry
Applications, 32(2):371–379. (Cited on page 35)
Trumper D.L. and Queen M.A. (1992): Control and actuator design for a precision
magnetic suspension linear bearing. In Controls for Optical Systems, volume
1696 of Proceedings of SPIE - The International Society for Optical Engineering,
pages 2–15. Orlando, FL, USA. (Cited on page 8)
Verma S., Kim W.j. and Gu J. (2004): Six-axis nanopositioning device with preci-
sion magnetic levitation technology. IEEE/ASME Transactions on Mechatron-
ics, 9(2):384–391. (Cited on page 8)
Watanabe K., Hara S., Kanemitsu Y., Haga T., Yano K., Mizuno T. and
Katamure R. (1996): Combination of h∞ and pi control for an electromagnetic-
ally levitated vibration isolation system. In Proceedings of the 35th IEEE Con-
ference on Decision and Control, volume 2, pages 1223–1228. (Cited on page 10)
Yonnet J.P. (1978): Passive magnetic bearings with permanent magnets. IEEE
Transactions in Magnetics, MAG-14(5):803–805. (Cited on page 6)
Yonnet J.P. (1981): Permanent magnet bearings and couplings. IEEE Transactions
on Magnetics, MAG-17(1):1169–1173. (Cited on pages 8 and 20)