Novel motor design for rotating anode x-ray tubes operating in the fringe

field of a magnetic resonance imaging system

Prasheel Lillaneya)
Department of Radiology, Stanford University, Stanford, California 94305 and Department of Bioengineering,
Stanford University, Stanford, California 94305
Mihye Shin
Department of Radiology, Stanford University, Stanford, California 94305 and Department of Mechanical
Engineering, Stanford University, Stanford, California 94305

Waldo Hinshaw
Department of Radiology, Stanford University, Stanford, California 94305
N. Robert Bennett
Department of Radiology, Stanford University, Stanford, California 94305 and Qualcomm MEMS
Technologies, San Jose, California 95134

Norbert Pelc
Department of Radiology, Stanford University, Stanford, California 94305 and Department of Bioengineering,
Stanford University, Stanford, California 94305
Rebecca Fahrig
Department of Radiology, Stanford University, Stanford, California 94305

(Received 3 May 2012; revised 20 November 2012; accepted for publication 7 December 2012;
published 16 January 2013)
Purpose: Using hybrid x-ray/MR (XMR) systems for image guidance during interventional proce-
dures could enhance the diagnosis and treatment of neurologic, oncologic, cardiovascular, and other
disorders. The authors propose a close proximity hybrid system design in which a C-arm fluoroscopy
unit is placed immediately adjacent to the solenoid magnet of a MR system with a minimum dis-
tance of 1.2 m between the x-ray and MR imaging fields of view. Existing rotating anode x-ray tube
designs fail within MR fringe field environments because the magnetic fields alter the electron trajec-
tories in the x-ray tube and act as a brake on the induction motor, reducing the rotation speed of the
anode. In this study the authors propose a novel motor design that avoids the anode rotation speed
reduction.
Methods: The proposed design replaces the permanent magnet stator found in brushed dc motors
with the radial component of the MR fringe field. The x-ray tube is oriented such that the radial
component of the MR fringe field is orthogonal to the cathode-anode axis. Using a feedback po-
sition sensor and the support bearings as electrical slip rings, the authors use electrical commuta-
tion to eliminate the need for mechanical brushes and commutators. A vacuum compatible proto-
type of the proposed motor design was assembled, and its performance was evaluated at various
operating conditions. The prototype consisted of a 3.1 in. diameter anode rated at 300 kHU with
a ceramic rotor that was 5.6 in. in length and had a 2.9 in. diameter. The material chosen for all
ceramic components was MACOR, a machineable glass ceramic developed by Corning Inc. The
approximate weight of the entire assembly was 1750 g. The maximum rotation speed, angular accel-
eration, and acceleration time of the motor design were investigated, as well as the dependence of
these parameters on rotor angular offset, magnetic field strength, and field orientation. The resonance
properties of the authors’ assembly were also evaluated to determine its stability during accelera-
tion, and a pulse width modulation algorithm was implemented to control the rotation speed of the
motor.
Results: At a magnetic flux density of 41 mT orthogonal to the axis of rotation (on the lower
end of the expected flux density in the MR suite) the maximum speed of the motor was found
to be 5150 revolutions per minute (rpm). The acceleration time necessary to reach 3000 rpm was
found to be approximately 10 s at 59 mT. The resonance frequency of the assembly with the
anode attached was 1310 rpm (21.8 Hz) which is far below the desired operating speeds. Pulse
width modulation provides an effective method to control the speed of the motor with a resolution
of 100 rpm.
Conclusions: The proposed design can serve as a direct replacement to the conventional induction
motor used in rotating anode x-ray tubes. It does not suffer from a reduced rotation speed when
operating in a MR environment. The presence of chromic steel bearings in the prototype prevented

022302-1 Med. Phys. 40 (2), February 2013 0094-2405/2013/40(2)/022302/14/$30.00 © 2013 Am. Assoc. Phys. Med. 022302-1
022302-2 Lillaney et al.: A MR compatible motor for x-ray tubes 022302-2

testing at the higher field strengths, and future iterations of the design could eliminate this shortcom-
ing. The prototype assembly demonstrates proof of concept of the authors’ design and overcomes one
of the major obstacles for a MR compatible rotating anode x-ray tube. © 2013 American Association
of Physicists in Medicine. [http://dx.doi.org/10.1118/1.4773313]

Key words: XMR, motor design, x-ray production, magnetic fields, MR compatibility

I. INTRODUCTION
Using hybrid X-ray/MR (XMR) systems for image guid- fringe field environment. Existing rotating anode x-ray tube
ance during interventional procedures could enhance the di- designs fail within MR fringe field environments because the
agnosis and treatment of neurologic, oncologic, cardiovascu- magnetic fields alter the electron trajectories in the x-ray tube
lar, and other disorders. XMR suites typically have become and act as a brake on the induction motor, reducing the rota-
more available, with various vendors offering dual-modality tion speed of the anode. Solutions to correct for the altered
solutions.1–13 These systems combine the three-dimensional electron trajectories have been discussed previously,20–24 and
imaging capabilities and excellent soft tissue contrast pro- are focused on designing magnetostatic or electrostatic mech-
vided by MR with the high spatial/temporal resolution and ac- anisms to correct for the changes in the trajectories such that
curate device tracking provided by x ray. To eliminate system the focal spot size, shape, and position are preserved. We de-
compatibility concerns, the suites typically have long travel scribe here in detail a novel motor design that eliminates the
distances between MR and x-ray components. As a result, reduced rotation speed of the anode. The proposed design is
switching between modalities requires shuttling the patient analogous to a modified three pole brushed dc motor, with the
several meters from one system to the other. Because patients radial component of the MR fringe field replacing the perma-
typically have critically placed monitoring systems and in- nent magnet stator field used in conventional brushed dc mo-
travenous lines for drug delivery and anesthesia, the cumber- tors. The motor support bearings provide rotating electrical
some shuttling process impedes repeated switching between contacts, and feedback signals from a position sensor control
the modalities and introduces image registration errors. electrical commutation, eliminating the need for mechanical
An XMR system geometry in which the modalities were brushes and commutators.
fully colocalized into a single imaging unit has also been suc- In our proposed system design the x-ray tube is positioned
cessfully demonstrated.14 In this approach a static anode x-ray below the common patient table (Fig. 1) such that the ra-
tube and digital flat-panel detector were installed into the gap dial component of the fringe field is orthogonal to the axis
of a double-donut 0.5 T interventional MR system (Signa SP, of rotation of the motor, while the longitudinal component of
GE Milwaukee, WI). This system was used successfully for the fringe field is oriented along the axis of the rotor. The
clinical procedures such as brain biopsy, transjugular intra-
hepatic portosystemic shunts (TIPS), chemoembolization of
hepatic tumors, hysterosalpingograms, and loopograms.15–18
The advantage of this system geometry is that it does not
require shuttling the patient to switch between the modali-
ties. However, the major limitations are the low field strength
of the double-donut MR system, and the limited x-ray flux
output of the static anode tube. A static anode design limits
the instantaneous heat loading capabilities of the target area
on the anode, which limits the maximum x-ray flux. Typical
maximum power ratings for static anode tubes are less than
1 kW, whereas rotating anode tubes can support power rat-
ings well beyond 10 kW. Hence, using static anode tubes is
problematic for several high flux angiographic procedures and
abdominal interventions in larger patients.
To overcome the limitations of the colocalized system ge-
ometry, and to circumvent the hurdles associated with alter-
nating between modalities that are separated by several me-
ters we proposed a close proximity hybrid system design in
which a C-arm fluoroscopy unit is placed immediately adja-
cent to a closed bore MR system with a minimum distance
of 1.2 m between the x-ray and MR imaging fields of view.19
The C-arm unit consists of a rotating anode x-ray tube and F IG . 1. Geometry of the proposed hybrid system with common table to shut-
tle the patient between the x-ray and MR imaging fields of view. The x-ray
digital flat panel x-ray detector mounted opposite from each tube and detector are mounted on opposite ends of a wooden C-arm, and the
other. Placing the C-arm so close to the MR bore requires rectangular prism is meant to depict the x-ray flux that is collimated onto the
an x-ray tube capable of operating in a relatively strong MR area of the flat panel detector.

Medical Physics, Vol. 40, No. 2, February 2013

022302-3 Lillaney et al.: A MR compatible motor for x-ray tubes 022302-3

F IG . 3. Forces on a wire loop in a constant magnetic field according to the

Lorentz force relationship.

F IG . 2. Depiction of the local field map of one potential location where the
x-ray tube motor could be positioned. The rectangular box outline gives a
the segment. The cross product in Eq. (1) is maximized when
rough idea of the dimensions of the cross section of the motor. the vectors B and L  are orthogonal to each other. Hence, the
shorter wire segments in Fig. 3 have no force exerted on them
approximate magnitude of the radial component (BR ) ranges because their direction of current flow is parallel to the mag-
from 20 to 120 mT, while the longitudinal component (BZ ) netic field direction. The longer segments have their current
ranges from 5 to 30 mT depending on the exact position and direction perpendicular to the magnetic field vector resulting
orientation of the C-arm.24 Figure 2 illustrates the local mag- in a force on each segment with magnitude equal to ILB. Note
netic field map for one possible position of the motor in the that the forces on the longer segments are equal in magnitude
MR environment to give the reader an understanding of the but point in opposite directions. If the axis of rotation is cho-
field orientation. This geometry would be used in neurological sen as shown in Fig. 3, then the magnitude of the total torque
applications where the center of the motor is positioned at Z can be derived according to Eq. (2):
= 116 cm, R = 66 cm relative to the MR imaging field of τ | = |r × F | = 2(W/2ILB ) = IAB .
| (2)
view.
Here τ is the torque vector, r is the moment arm vector de-
In this study a vacuum compatible prototype of the pro-
fined from the axis of rotation to the wire segment, and W
posed motor design was assembled, and its performance was
is the length of the shorter wire loop segments. The product
evaluated at various operating conditions. The rotation speed
of the wire segment lengths can be simplified as the area (A)
(ω), angular acceleration (α), and acceleration time to achieve
of the wire loop yielding the expression on the right hand
minimum operation speed of the motor design were investi-
side of Eq. (2). It is helpful to rewrite the torque in a more in-
gated (tAccel ), as well as the dependence of these parameters
tuitive and generalized form, which is shown in Eq. (3). This
on rotor angular offset (γ ), magnetic field strength (|B|), and
equation holds true for arbitrary wire loop shapes, and can be
field orientation (|Bz | relative to |BR |).
derived by integrating the torque contributions from infinites-
imally small segments of wire:
II. METHODS τ = μ 
 × B. (3)
II.A. Theory of operation  is referred to as the magnetic moment of
In this expression μ
II.A.1. Fundamental torque relationship the wire loop and is defined below:
The fundamental principle behind generating rotation is  = N I An̂.
μ (4)
based on the interaction of a current carrying wire loop in a N refers to the number of turns in the wire loop and n̂ is
magnetic field. Figure 3 illustrates an example of a wire loop a unit vector normal to the plane that contains the wire loop.
with current flow in the counter-clockwise direction, with a From Eq. (3) it is evident that maximum torque is created
magnetic field oriented perpendicular to the long axis of the when μ 
 is perpendicular to B.
loop. Using the Lorentz force law [Eq. (1)] the force exerted
on each segment of wire due to the magnetic field can be de-
II.A.2. Efficient winding commutation
rived:
and rotor orientation
F = q v × B = I L
 × B.
 (1)
The motor design has three separate coil windings attached
In the above expression F is the force vector, q is the to the rotor body. The windings are arranged as shown in the
amount of charge, v is the velocity of the charge, B is the schematic diagram depicted in Fig. 4. The direction of the nor-
 is a vector
magnetic field vector, I is the current value, and L mal vectors (n̂) for each coil is spaced 120◦ apart from each
with magnitude equal to the length of the wire segment and other. Consider the case where only coil winding A is ener-
with direction oriented parallel to the direction of current in gized with current flowing in the counter clockwise direction.

Medical Physics, Vol. 40, No. 2, February 2013

022302-4 Lillaney et al.: A MR compatible motor for x-ray tubes 022302-4

F IG . 5. Diagram of one mechanical cycle of the rotor broken into 60◦ sec-
tors. Each sector is color coded to show which direction current should flow
when a coil winding has its magnetic moment within that sector. With the
magnetic field direction shown this particular current polarity configuration
will produce a clockwise rotation.

enters sectors 4 and 5 a negative polarity must be applied. If a

negative polarity is applied for sectors 1 and 2 and a positive
F IG . 4. Schematic representation of the wiring geometry relative to the rotor
polarity is applied for sectors 4 and 5, then the coil will cause
body (gray cylinder). There are three coil windings (represented by the black the rotor to rotate in the counter-clockwise direction. Finally,
rectangular loops) that are separated by 120◦ . When current flows through a when a coil enters sectors 3 or 6 it should have no current de-
particular winding it will align its magnetic moment with the direction of the livered to it from the power supply. As the angular spans of
external magnetic field causing the rotor to turn. each sector are equal to each other, this implies that over the
course of one full mechanical cycle a given coil winding will
spend an equal portion of the cycle in all three polarity states.
Current flow in this direction dictates that μ  A is parallel to n̂A .
Furthermore, the 120◦ offsets between the coil positions dic-
If B is perpendicular to μ A , then the coil will exert torque in
tate that at any point within a mechanical cycle each of the
a clockwise direction causing rotation until μ  A is parallel to
 At this position if nothing is changed the motor will stall, coil windings will occupy different polarity states.
B.
To achieve efficient acceleration during startup the angu-
and if further clockwise rotation is desired the current to coil
lar locations of the transition boundaries of the commutation
winding A must be switched off and the current to coil wind-
states must be determined. For convenience during this analy-
ing C must be turned on. Thus, as the rotor turns the windings
sis it is assumed that the motor is always positioned such that
can be turned on and off according to a commutation sched-
the magnetic moment for coil winding A is aligned with the
ule such that there is always one coil winding that will create
 transition boundary between sector 6 and sector 1, and that
clockwise rotation by trying to align its μ  with B.
clockwise rotation is desired. The objective of the following
Using a commutation schedule where only one coil wind-
analysis is to determine the angular location of the transition
ing is energized at a time is not the most efficient approach. A
boundary between sector 6 and sector 1 such that the motor
more efficient method requires energizing two coil windings
has the largest angular acceleration once started. Since each
at a time, one with current flowing in the counter-clockwise
sector spans 60◦ , once the position of this transition boundary
direction and the other with current flowing in the clockwise
is known, all the other boundaries will also be determined.
direction. By controlling the polarity of the current in the two
By maximizing the work performed on the motor as the mag-
coils it is possible to generate magnetic moments that add
netic moment for coil winding A travels through sectors 1 and
constructively instead of destructively. To determine the po-
2, it is possible to determine the optimal angular position for
larity schedule for the coils (i.e., the direction of current in
the transition boundary. The work can be defined as the inte-
each coil during one mechanical revolution of the rotor) the
gral of the torque performed on the motor as it rotates; analo-
360◦ mechanical cycle of the rotor is divided into six sepa-
gous to how integrating force over a distance yields the work
rate 60◦ sectors, as shown in Fig. 5. The arrows labeled A,
for an object that is being subjected to translational motion
B, and C represent the magnetic moments (n̂) for the three
[Eq. (5)]:
coil windings. Each 60◦ sector can be assigned one of three
different polarities: positive, neutral, or negative. The polarity   θ+120
indicates in which direction the current will flow in the coil W = τ (φ)dφ = |μ| |B| sin(φ)dφ. (5)
θ
when it is positioned within the given sector, or if the coil
should be turned off. A positive label indicates that current Equation (5) is derived by integrating Eq. (3) over the ap-
should flow in the counter-clockwise direction, a negative la- propriate angular span, where θ denotes the boundary be-
bel indicates clockwise current flow, and a zero indicates that tween sectors 6 and 1 and θ + 120 denotes the boundary be-
the coil should not be energized at all. If clockwise rotation tween sectors 2 and 3. Maximizing this integral maximizes the
of the rotor is desired, then when a coil enters sectors 1 and torque performed on the motor, which effectively maximizes
2 a positive polarity must be applied to it, and when the coil the angular acceleration.

Medical Physics, Vol. 40, No. 2, February 2013

022302-5 Lillaney et al.: A MR compatible motor for x-ray tubes 022302-5

Performing the optimization on the work integral yields operating speed of slightly less than 3000 revolutions
the optimum value of θ to be equal to 30◦ , and by symmetry per minute (rpm) and have delays on the order of
arguments this implies that the angular location of the bound- 1–2 s before reaching this speed.26 To generate suf-
ary between sectors 3 and 4 should be located at 180◦ + 30◦ . ficient torque the rotational moment of inertia of the
Intuitively, it makes sense that the location of the transition proposed design is larger than an induction motor, and
boundary (i.e., the value of θ ) that maximizes the work inte- hence acceleration of the assembly will take longer.
gral is bounded by 0◦ and 60◦ . Values for θ that are outside Future iterations of the design can use a larger startup
of this domain cause portions of sector 1 or 2 to be in the boost current (as done in the induction motor design)
lower half plane (180◦ –360◦ ) of the mechanical cycle. Apply- in combination with a lower rotational moment of in-
ing counter-clockwise current to the windings in this scenario ertia to reduce this acceleration time.
leads to counter-clockwise rotation of the rotor. Using the (iv) Fabrication. Custom components should be fabri-
convention defined above where positive torque is assumed cated using conventional metal working tools to ex-
to create clockwise rotation (the desired rotation direction), pedite the prototyping process.
any counter-clockwise rotation would reduce the work per-
formed on the rotor. Furthermore, the work performed on the Exploded views of the motor assembly along with an iso-
rotor could be increased if the angular span of sectors 3 and 6 lated view of the rotor body are shown in Fig. 6. The rotor
was smaller than 60◦ . Ideally, the span for these sectors would body is machined out of ceramic to prevent the creation of
be just large enough to allow the coil windings to switch be- eddy currents that could lead to a reduction in motor speed.
tween positive and negative polarity. However, the nature of The approximate height and outer diameter of the rotor are
the control electronics, which are explained in more detail in 15.2 and 7.4 cm, respectively. The anode used in this design
Sec. II.C., dictates that all three polarity states have equal an- has a 7.9 cm outer diameter, a 14◦ target angle, a tungsten-
gular spans limiting each sector to 60◦ . rhenium molybdenum target, weighs 0.64 kg, and is rated
at 300 kHU. It is from a rotating anode x-ray tube (Varian
II.B. Prototype construction Rad-4),27 which is intended to be used for general purpose ra-
diographic and fluoroscopic procedures. The anode plate was
A prototype of the motor design depicted in Fig. 4 was machined to mate with the stem of the anode in order to hold
machined and assembled. The geometry and materials used it securely in place. There are four steel bearings used in the
for the prototype were chosen such that they satisfied several prototype, with each bearing providing mechanical support to
design constraints, which are enumerated below along with the rotor body while also serving as an electrical slip ring.
their rationale. The bearing positioned closest to the anode serves as a re-
(i) Materials. Whenever possible the materials used in turn path to ground for the tube beam current. The three bear-
the assembly were chosen to be nonmagnetic and ings farthest from the anode are used to provide current to the
nonconductive. Any magnetized components increase coil windings. The stationary connections to the bearings are
the apparent friction in the system as the magnetized made by soldering vacuum compatible polyimide insulated
domains tend to align with the direction of the fringe magnet wire to the inner races. The free ends of these wires
field making it more difficult for the assembly to ro- are fed along slots along the inner spacers until they reach the
tate. Furthermore, conducting structures create eddy base plate where they are connected to a vacuum feedthrough.
currents when they are rotated in the presence of a The rotating connections to the bearings are made by solder-
magnetic field. The direction of these eddy currents ing wires to the outer bearing races, which are then fed out
is such that they will reduce the rotation speed of the through holes in the rotor body, or for the case of the ground
rotor as well. return wire out through an opening in the anode plate. The
(ii) Vacuum compatibility. All materials must be vacuum coils for the different motor phases are wound around the
compatible and capable of surviving vacuum process- ceramic rotor along slots at the top, bottom, and side of the
ing steps such as the bake out procedure that is used in rotor body. The coil windings are wired together in the con-
the manufacturing of x-ray tubes. To be vacuum com- ventional wye configuration, which is discussed in more de-
patible the material must have acceptable surface and tail in Sec. II.C. The wye configuration requires that there is
bulk desorption rates, low vapor pressure at the limits a common connection point between all three coil windings,
of the operating temperatures, and not contain cracks, and in our assembly this connection point is located in the
crevices, or trapped volumes that can be the source bottom slot of the rotor body. Wave springs are placed in be-
of virtual leaks later on.25 During bake out, the entire tween each inner spacer and bearing to create a slight dis-
assembly will be subjected to 300 ◦ C for a period of placement of the inner races relative to the outer races. The
24–48 h while being connected to an ion pump for displacement leads to a better contact between the balls and
outgassing. races of each bearing, which improves the conduction of cur-
(iii) Acceleration speed. Accelerate to minimum operating rent through the bearing. Ceramic sleeves are positioned be-
speed on a timescale of 10 s over the range of oper- tween the outer races to ensure the bearings are in the cor-
ating field strengths stated earlier (20–120 mT per- rect positions along the length of the rotor body, and to elec-
pendicular to the axis of rotation). For comparison, trically insulate the bearings from each other. All ceramic
typical x-ray tube induction motors have a minimum components in the assembly are made out of MACOR, a

Medical Physics, Vol. 40, No. 2, February 2013

022302-6 Lillaney et al.: A MR compatible motor for x-ray tubes 022302-6

TABLE I. Components and material list.

Component name Material composition Manufacturer

Anode Tungsten, Rhenium, TZM Alloy Varian (Palo Alto, Ca)

Anode plate 316 Stainless steel Custom Machined
Rotor body MACOR (machineable glass-ceramic, Corning Inc.) Custom Machined
Center shaft 316 Stainless steel Custom Machined
Shaft key 316 Stainless steel McMaster (Elmhurst, IL)
End cap MACOR Custom Machined
Base plate 316 Stainless steel Custom Machined
Vacuum feedthrough 304 Stainless steel Solid Sealing Technology (Watervilet, NY)
Washer 316 Stainless steel McMaster
Spring clip Type PH 15-7 MO Stainless steel McMaster
Wave spring Carbon steel Smalley (Lake Zurich, IL)
Preload spring Carbon steel Smalley
Sleeve MACOR Custom Machined
Bearing Chromic steel VXB Ball Bearings (Anaheim, CA)
Inner spacer MACOR Custom Machined
Coil windings Kapton, Gold, Copper Accu-Glass (Valencia, CA)

machineable glass ceramic developed by Corning Inc. A com- rotor via a support bearing. Each coil winding is represented
plete list of the components used in the prototype is shown in by lumped resistance and inductance (Fig. 8), and they are
Table I. connected to each other in a wye configuration. The inverter
bridge is part of an integrated circuit inside a commercially
available three-phase low voltage power module. The control
II.C. Circuit implementation signals for the inverter bridge are generated using the MC1
motor control development board which has an embedded
II.C.1. Power control electronics microcontroller (dsPICDEM6010A). Both the power module
The control electronics are identical to those used in three- and the development board are manufactured by Microchip
phase brushless dc (BLDC) motor designs.28 The circuit con- Technology (Chandler, AZ). The development board samples
sists of a three-phase inverter where each leg of the inverter is the output signals from the motor feedback position sensor
connected to one of the coil windings that are attached to the (the details of the sensor are discussed in more detail below),

F IG . 6. (a) Exploded view of the motor assembly depicting how the exterior components such as the rotor body, anode, plate, end cap, and base plate mate
with each other. (b) Exploded view of the assembly with the rotor and the center shaft hidden from view in order to demonstrate how the sleeves, inner spacers,
wave springs, and bearings are positioned with respect to each other. (c) Isometric view of the isolated rotor body illustrating how the coil windings are arranged
within the wire slots. (d) The complete rotor assembly with mounting interface, anode, and coil windings attached to the rotor.

Medical Physics, Vol. 40, No. 2, February 2013

022302-7 Lillaney et al.: A MR compatible motor for x-ray tubes 022302-7

TABLE II. Electrical commutation schedule for coil windings. II.C.2. Position feedback and speed
monitoring system
State No. A B C

1 + − 0
The feedback position system consists of a Hall effect
2 + 0 − quadrature sensor, and a ferrous nylon magnet wheel con-
3 0 + − sisting of 120 alternating north and south poles, mounted at
4 − + 0 the base of the rotor (both parts purchased from ElectroSen-
5 − 0 + sors, Minnetonka, MN). This positioning system was chosen
6 0 − + as it was one of the few commercial options available that was
capable of mating with the nonstandard rotor size. However,
this sensor does not meet two of the critical design criteria
specified for the motor design. It is not vacuum compatible
and these signals are interpreted by the microcontroller to de- because it is contains nylon which has a high outgassing rate,
termine the position of the rotor so that the correct control sig- and it cannot survive the temperatures necessary for the bake
nals can be sent to the power module. If the magnetic field di- out procedure. For operation in vacuum the magnet wheel
rection is assumed to be oriented relative to the rotor as shown needs to be custom designed with a high Curie temperature
in Fig. 5, then the optimal commutation schedule for the coil magnetic material such as samarium-cobalt. For initial test-
windings will be as listed in Table II. This sequence also as- ing purposes the nylon magnet wheel was used to perform
sumes the rotor is oriented such that n̂A is aligned at 30◦ (the electrical commutation. The Hall effect sensor is positioned
start of sector 1) before any coil windings are energized. Each at a distance no greater than 0.102 cm (0.040 in.) relative
state in the sequence has one coil winding with positive po- to the magnet wheel, as this is the maximum operating dis-
larity, another winding with negative polarity, and the third tance specified by the manufacturer. To function correctly in
winding inactive. As there are six states in the sequence, each the MR fringe field environment the axis of the sensor is po-
state will be active through 60◦ of the mechanical cycle. sitioned orthogonal to the fringe field direction (i.e., for the
Each leg of the inverter consists of two power metal oxide geometry shown in Fig. 5 the axis of the sensor would be po-
semiconductor field-effect transistors (MOSFETs) connected sitioned at the transition boundary between sectors 1 and 2
in series. These transistors are referred to as the high side and or at the transition boundary between sectors 4 and 5). The
low side transistor depending on which side of the dc volt- situation may arise where the exact fringe field direction is
age bus they are connected. Each transistor is operated in en- not known, but correct operation can be achieved if the sen-
hancement mode, meaning that when a positive voltage is ap- sor can be placed orthogonal to the plane that contains the
plied to the gate, a conduction path opens between the source magnetic fields. As the ferrous magnet wheel spins with the
and the drain, effectively making each transistor a switch. For rotor, the sensor generates two square wave outputs (chan-
example, to operate the coil windings the first state shown in nel A and B) with 60 pulses/revolution in quadrature (i.e.,
Fig. 5 (A positive, B negative, C zero) voltage is applied to the 90◦ phase shift between waveforms). The direction of rotor
gates of transistors 1 and 4, with all other gates receiving no travel (clockwise or counter-clockwise) determines whether
voltage. This configuration is the first row shown in Table III, channel A leads channel B or vice versa. These signals are
and the following rows are the FET gate voltages for the sub- sampled by the development board and by a data acquisition
sequent states that were listed in Table II. Both high and low card (National Instruments, Austin, TX). A LabVIEW vir-
side transistors of the same leg should never be conducting at tual instrument is used to calculate the instantaneous speed
the same time, as this would create a short circuit across the of the motor by calculating the time period in between ris-
dc bus and damage the control circuitry. The diodes shown in ing and falling edges of the square wave signals on channels
the schematic represent the intrinsic parasitic body diodes of A and B.
the MOSFETs. Their purpose is to protect the transistor from To start the commutation sequence the rotor is carefully
any reverse current that will arise from abruptly switching the aligned such that n̂A is aligned at 30◦ . A short duration 5 V
currents on and off to the coil windings. pulse is sent to the index port on the development board which
triggers the microcontroller to proceed into the first commu-
tation state (state #1 in Table II). The microcontroller incre-
TABLE III. FET gate voltage schedule. ments or decrements the position count register depending on
the direction of travel by the rotor, which can be clockwise or
FET No. counter-clockwise. The numerical value from this register is
State No. 1 2 3 4 5 6 used to compute the angular position of the rotor, which then
determines the next appropriate commutation state. The max-
1 +V 0 0 +V 0 0 imum value of the position count register is predetermined
2 +V 0 0 0 0 +V
in the microcontroller software routine that controls the com-
3 0 0 +V 0 0 +V
mutation of the motor. This value is computed based on the
4 0 +V +V 0 0 0
5 0 +V 0 0 +V 0
number of poles of the specific magnetic encoder wheel used
6 0 0 0 +V +V 0 in the assembly. The microcontroller recognizes the comple-
tion of one mechanical revolution when the maximum value is

Medical Physics, Vol. 40, No. 2, February 2013

022302-8 Lillaney et al.: A MR compatible motor for x-ray tubes 022302-8

reached in the register. At this point the position count register II.D.2. Duty cycle calculation using SVM
is reset to its starting state which causes the microcontroller
As mentioned previously the motor is a three-phase sys-
to set the voltages to the coil windings back to the first com-
tem, with the windings connected to each other in a wye con-
mutation state (state #1).
figuration. The Park transformation30 allows for analysis of
a balanced three-phase system as a two-phase system. Intu-
II.D. Angular velocity control itively, the system can be reduced since the winding voltages
are dependent because they are wired together. The transfor-
II.D.1. Introduction to speed control using pulse width mation to convert the voltages from the three-phase inverter
modulation (PWM) system to an orthogonal two-phase system in the complex
Speed control for the motor is accomplished using PWM, a plane is shown in Eq. (6):
common technique for controlling the power delivered to the aS = 2/3[aX (t) + a · aY (t) + a 2 · aZ (t)]. (6)
motor. In this approach, a switch placed between the power
supply driving the motor and the load (i.e., the coil windings) Here we define the constant vectors a = e j 2π/3
and a 2
alternately turns the current to the load on and off in order to =e j 4π/3
. The variables aX (t), aY (t), and aZ (t) represent the
control the average power delivered to the load. Controlling voltages for each of the three windings, respectively. The vec-
motor speed using PWM is advantageous because the method tor aS is the transformed two-dimensional representation of
allows for minimal power losses in the additional switching the system in the complex plane. The purpose of the Park
devices. When the switches are off there is virtually no cur- transformation is to simplify the analysis of the system in the
rent, and when they are on there is virtually no voltage drop subsequent SVM calculations.
across the switch. In both cases the product of the current and In the SVM approach each leg of the inverter shown in
voltage in the switch (i.e., the power loss associated with the Fig. 8 must have either the high side or low side MOSFET
switch) is relatively small. By varying the duty cycle of the switch closed. When the high side switch is closed and the
switching, it is possible to control exactly how much power is low side switch is open the leg is represented by a 1, and when
supplied to the coil windings which, in turn, determines how the high side switch is open and the low side switch is closed
fast the motor will turn. the leg is represented by a 0. Since there are three legs with
As mentioned previously in the description of the circuit two possible configurations for each leg, the inverter bridge
implementation (Sec. II.C.1) the motor control circuitry con- is capable of producing eight (23 ) discrete voltage states. The
sists of an inverter bridge with six MOSFETs that function voltage states where all of the high side switches are closed
as switches. The state of each MOSFET switch (on or off) (1,1,1) and where all of the low side switches are closed
depends on its respective gate voltage. We can control the (0,0,0) are effectively the same state. In both of these degen-
percentage of time that each MOSFET is active by driving erate cases there is no current in any of the coil windings be-
each gate using square waves with varying duty cycles, thus cause each segment of the wye configuration is held at the
controlling the percentage of time each coil winding is ac- same voltage potential. The seven unique voltage states can
tive. The polarity of current in the winding can also be con- be mapped to discrete vectors in the complex plane using
trolled depending on if the high or low side MOSFET for the Eq. (6). These vectors are illustrated in Fig. 9 where V1 –V6
winding is modulated. In the software control routine for the are the active voltage vectors, while V0 and V7 are the degen-
motor, the task of determining the appropriate duty cycles is erate zero vectors.
accomplished using space vector modulation (SVM).29 The Ideally to implement perfect speed control the inverter
SVM routine takes the current orientation of the rotor and the bridge would be able to output a voltage state that could map
desired speed of rotation, and computes the duty cycle and to a voltage vector V with general orientation and magni-
polarity for each coil winding (Fig. 7). The orientation and tude in the complex plane. However, it is restricted to the
desired speed are updated every sampling period (TS ), which eight discrete voltage states mentioned previously. To solve
is a set time interval predetermined in software. The details of this challenge the SVM algorithm decomposes the general
the SVM calculations are given in Sec. II.D.2. voltage vector V into a linear combination of the neighbor-
ing active voltage vectors. The mathematical representation
of the decomposition is shown in Eq. (7), where V is the gen-
eral voltage vector being decomposed, while VA and VB are
the neighboring vectors used in the decomposition, and V0 is
the degenerate zero vector. The coefficients or weights in the
decomposition, denoted as ta , tb , and t0 , correspond to time
intervals and TS is a predetermined sampling interval:
V · TS = VA · ta + VB · tb + V0 · t0 . (7)
Intuitively, Eq. (7) is stating that if we require the inverter
bridge to output a general voltage vector V for a time interval
F IG . 7. Schematic representation of the SVM feedback routine. The routine
samples the angular position and angular velocity of the motor every sam- TS , then it is possible to do so by outputting a weighted av-
pling period (TS ). From this input data, the routine then updates the duty erage of the neighboring voltage vectors (VA and VB ). In this
cycles and polarities for each coil winding. instance, the averaging is done over time with ta and tb being

Medical Physics, Vol. 40, No. 2, February 2013

022302-9 Lillaney et al.: A MR compatible motor for x-ray tubes 022302-9

F IG . 8. Circuit implementation of the motor demonstrating how each coil winding is connected to the inverter bridge. Each of the outputs from the inverter
bridge is wired to the inner race of one of the support bearings. The outer race of the bearing is then wired to one of the coil windings through one of the bearing
connection holes in the rotor. The other connection points of the coil windings are all wired together forming a “wye” configuration.

a fraction of the sampling interval TS . To be able to control and can generate a maximum flux density of approximately
the magnitude of V , we also average in the degenerate zero 44 mT at its center. The motor assembly was mounted in be-
vector (V0 ) for a time interval t0 . The magnitude of V0 will tween the coil pair at the center point such that the magnetic
decrease as the length of t0 increases because the motor will field was orthogonal to the rotation axis. This experimental
spend a greater proportion of time in the zero voltage vector setup was used to measure the time necessary to accelerate
state. to 3000 rpm as a function of the magnetic field strength gen-
The definitions for the pulse width modulation variables ta , erated by the coil pair. The time required to reach 3000 rpm
tb , and t0 are given below in Eqs. (8)–(10): benchmark is significant because the x-ray tube current is typ-
√ ically not energized until the anode achieves this minimum
3V
ta = TS sin(π/3 − φ), (8) speed. The speed of the motor was monitored continuously
Vd throughout the acceleration process, and after crossing the
√ 3000 rpm threshold, the motor was allowed to decelerate to
3V rest.
tb = TS sin(φ), (9)
Vd The Helmholtz coil pair setup was also used to determine
the effect of the rotor angular starting position on the aver-
age acceleration and maximum rpm. The magnetic wheel en-
t0 = TS − ta − tb . (10)
coder could be used to deliberately offset the starting position
of the rotor in 6◦ increments (the equivalent of 1 pole pair on
In this formulation the angle φ is normalized to the sector the magnetic wheel). An angular offset causes the work in-
in which V resides, restricting the minimum and maximum tegral in Eq. (5) to be not optimized, which results in a drop
value of φ to 0◦ and 60◦ , respectively. The variable V is the off in acceleration and maximum rpm. The Helmholtz coil
magnitude of V , and Vd is the magnitude of the voltage dif- pair was also used to test the PWM speed control algorithm.
ference between the high and low side of the inverter bridge. The pulse width modulation timing parameters (ta , tb , t0 ) were
These variables are computed and updated once every sam- controlled by adjusting the above-mentioned potentiometer,
pling period in the PWM algorithm that runs on the micro- which in turn modulated the angular velocity of the motor.
controller. Thus, by altering the potentiometer setting on the
development board it is possible to determine how long the
II.E.2. MR fringe field
motor spends in one of the zero voltage states (V0 or V7 ) in
each sampling period, which directly affects the instantaneous After these initial tests in the Helmholtz coil field, the mo-
rotation speed of the rotor. A demonstration of the angular ve- tor assembly was also tested in the fringe field of a 0.5 T GE
locity control algorithm for the motor is shown in Sec. III. Signa SP MR system (Fig. 10). The purpose of these tests was
twofold:
II.E. Motor operation in magnetic fields (i) To evaluate the performance of the motor at higher
magnetic field strengths than the Helmholtz coil pair
II.E.1. Helmholtz coil field
could generate.
Initial evaluation of the motor prototype performance was (ii) To determine the effect of misalignment between axis
carried out using a Helmholtz coil pair (manufactured by of rotation and the magnetic field, i.e., the magnetic
Stangenes Industries, Palo Alto, CA). The coil pair has a field is not orthogonal to the axis of rotation of the
21 cm inner radius, a 15 cm gap between the inner coil faces, motor.

Medical Physics, Vol. 40, No. 2, February 2013

022302-10 Lillaney et al.: A MR compatible motor for x-ray tubes 022302-10

F IG . 9. Possible voltage configurations of the inverter bridge mapped

to the complex plane using the transformation in Eq. (6). The space F IG . 11. Graph of the time required by the motor to reach 3000 rpm versus
vector modulation approach decomposes the requested voltage vector V the transverse field strength.
into a linear combination of the two neighboring vectors and a zero
vector (V0 or V7 ).

erated from rest to 3000 rpm. In addition, each test run was
performed with and without the anode attached to the rotor
The prototype was tested at five different positions within
assembly in order to determine the effect of the anode on the
the MR fringe field. For three of these positions the fringe
acceleration times.
field was virtually orthogonal to the axis of rotation of the
motor [Fig. 10(b)], with the only change between the posi-
tions being the magnitude of the magnetic flux density. The
approximate values of the flux density at the center of the III. RESULTS
rotor for these three positions were 38, 50, and 59 mT. The
III.A. Acceleration times under different
other two positions placed the motor at an angle relative to
field conditions
the face of the magnet such that there was a component of
the magnetic field parallel to the axis of rotation of the motor. The effect of the |B| on the time required to accelerate to
Figure 10(c) illustrates one such orientation. In both of these 3000 rpm (tAccel ) with the anode attached is demonstrated in
angled positions the motor is still oriented along the central Fig. 11. As the magnitude of the flux density orthogonal to
axis of the magnet. For the angled motor orientations the mag- the axis of rotation increased, tAccel decreased dramatically.
nitudes of the parallel components of the flux density at the The value of tAccel dropped to 10.56 ± 0.06 s at the highest
center of the rotor were B|| = 13 and 31 mT. The magnitude magnitude field condition (59 mT). The rise time dynamics of
of the orthogonal flux density (B⊥ ) was held constant (35 mT) the motor at 38 mT and how they are affected by the presence
for both orientations by altering the distance between the mo- of the anode are illustrated in Fig. 12. Since determining the
tor and the magnet. At each position the motor was accel- maximum rpm was not the purpose of these experiments the

F IG . 10. (a) Diagram illustrating the orientation of the magnetic field lines of the 0.5 T open bore MR system with respect to the rotor when the fringe field is
orthogonal to the axis of rotation. (b) Experimental setup with MR fringe field orthogonal to the rotor axis of rotation as seen from the front of the MR system.
(c) Experimental setup with the rotor axis of rotation misaligned with the MR fringe field such that there is a component of the fringe field parallel to the axis of
rotation.

Medical Physics, Vol. 40, No. 2, February 2013

022302-11 Lillaney et al.: A MR compatible motor for x-ray tubes 022302-11

F IG . 14. Graph of the average angular acceleration of the motor and how it
F IG . 12. The rpm versus time graphs of the motor in a MR environment depends on angular offset.
[see Fig. 9(b)] with a 35 mT orthogonal field. Curves are shown both with
and without the anode attached to give the reader an idea of the rise time
dynamics. III.B. Effect of rotor angular offset
In these performance tests the starting angular position of
the rotor was offset from 0◦ in 6◦ increments up to a maxi-
current was switched off after the motor passed the 3000 rpm mum offset angle (γ ) of 72◦ . The motor was allowed to ac-
benchmark. The presence of the anode caused a significant celerate until it reached a steady state maximum rpm, after
increase in tAccel (Fig. 13), ranging from 54% at 38 mT to which it was brought back to rest. For each value of γ the
75% at 59 mT. The motor could not be tested at higher field average angular acceleration (α) of the rotor over the first
strengths because the bearings, which are unfortunately made 10 s of operation was calculated, and the maximum obtainable
out of chromic steel, become strongly magnetized at fields ω was recorded. The results of these calculations are shown
larger than 65 mT and cause the motor to stall. Last, tAccel in Figs. 14 and 15, respectively. During each test run B⊥ was
shows no strong dependence on the presence of a magnetic held constant at 20.5 mT and the anode was not attached to
field component parallel to the axis of rotation (Fig. 13). The the rotor assembly. The average α started at a maximum of
values of tAccel for B⊥ = 35 mT, B|| = 13 mT and B⊥ = 35 mT, 21.5 rad/s2 at 0◦ and fell to 4.1 rad/s2 at γ equal to 72◦ , while
B|| = 31 mT were nearly identical (14.35 s versus 14.34 s with the maximum ω started at 4740 rpm and fell to 1370 rpm.
anode, and 9.35 s versus 9.28 s without anode). These times The angular offset experimental setup allows for precise
were also comparable to (but higher than) those recorded for control over the value of γ for the motor. We used this fact
B⊥ = 38 mT, B|| = 0 mT (13.03 s with anode and 7.93 s to our advantage and tuned α via γ to find the resonant fre-
without anode). quency of our motor assembly. At γ equal to 60◦ the input
α was just large enough so that the motor spent a 10-s pe-
riod fighting through mechanical resonance between 30 and
40 s. This is illustrated by the boxed region on the solid line in

F IG . 13. Acceleration times of the motor with and without the anode at- F IG . 15. Graph of the maximum obtainable rpm of the motor and how it
tached for various MR fringe field strengths and orientations. depends on angular offset.

Medical Physics, Vol. 40, No. 2, February 2013

022302-12 Lillaney et al.: A MR compatible motor for x-ray tubes 022302-12

F IG . 17. Acceleration curve of the rotor when the PWM speed control
F IG . 16. Acceleration curves of the motor highlighting the location of the
method is employed.
resonance frequency of the assembly with and without the anode attached.

celerate once again and continue with the space vector mod-
Fig. 16 where there is a noticeable “dip” in the ω versus time
ulation control scheme. The ability to reaccelerate is demon-
graph at 2050 rpm. At smaller values of γ (<60◦ ) α was too
strated past the 220 s mark.
large and this effect was not visible because the motor did
not spend as much time at the resonant frequency. At larger
values of γ (>60◦ ) α was too small and the maximum ω of
IV. DISCUSSION
the motor was below the resonant frequency. A similar exper-
imental setup was performed with the anode attached to the IV.A. Dependence of acceleration times on field
assembly. Using a γ value of 0◦ and setting B⊥ to 6.2 mT, the strength and orientation
resonance of the assembly was visualized at a lower speed of For B⊥ lower than 6 mT the input torque to the system is
1310 rpm, which is illustrated between the 70 and 80 s mark not large enough to overcome the static friction in the system
on the dashed line in Fig. 16. The γ value and B⊥ were altered and the motor fails to rotate, while at B⊥ lower than 8 mT the
so that with the anode attached α was just large enough to vi- maximum ω of the assembly is lower than 3000 rpm. As B⊥
sualize the mechanical resonance of the assembly. For both is increased past the 8 mT threshold the acceleration times de-
configurations the entire assembly could be seen to visibly vi- crease drastically until about 30 mT after which the effect of
brate as it accelerated through resonance. It should be noted the increased input torque becomes marginal. At higher val-
that the reader should not directly compare the two curves ues of B⊥ the magnetization of the bearings and anode is sig-
in Fig. 16 against each other and assume the only difference nificantly larger. Furthermore, the magnitudes of the induced
between the curves was the presence of the anode. These ex- back EMF current in the anode, anode plate, and bearings
periments were run at different sets of nonoptimal operating are larger as well. Both of these effects cause increased re-
conditions for the motor simply to visualize mechanical res- sistance to rotation thus increasing the apparent friction in the
onance, which also explains why these curves do not cross system, which counteracts the effect of the increased input
3000 rpm, as opposed to those curves run at optimal condi- torque. Replacing the chrome steel bearings with austenitic
tions which do cross the 3000 rpm benchmark (Fig. 12). stainless steel, a material that is virtually nonmagnetic, can
mitigate the increase in apparent friction. Chromic steel bear-
ings were chosen for this version of the prototype because
III.C. Speed control using pulse width modulation
they were readily available in a wide variety of sizes with the
The instantaneous ω of the motor was recorded while run- intention that they would be replaced before the motor assem-
ning the space vector modulation control scheme discussed bly was vacuum sealed. However, from the motor data gath-
above with B⊥ equal to 41 mT. The speed curve for the mo- ered at higher fields it is evident that the high field data is sig-
tor using this algorithm is shown in Fig. 17. The motor was nificantly affected by the presence of the magnetic bearings,
run for 10 to 20 s intervals for ω ranging from 5150 rpm to and that this data needs to be collected again without mag-
1800 rpm. The algorithm can also be used to bring the motor netic bearings. We are in the process of replacing the chromic
to rest by requesting a 0 magnitude voltage vector, which is steel bearings with their austenitic stainless steel counter-
shown at approximately 220 s. In this scenario the stall detec- parts from an in-state vendor (KMS Bearings, Inc. Anaheim,
tion in the control electronics cuts off the current to the coil CA), which we predict will improve the performance of the
windings to prevent any damage by overheating the power motor at higher field strengths. The austenitic stainless steel
MOSFETs or the coil windings themselves. However, since does have the disadvantage of being a “softer” material than
the position of the rotor is still known once a nonzero mag- chromic steel, so there will be quicker long term wear on the
nitude voltage vector is requested the motor will begin to ac- bearings that warrants further investigation.

Medical Physics, Vol. 40, No. 2, February 2013

022302-13 Lillaney et al.: A MR compatible motor for x-ray tubes 022302-13

The value of tAccel of the motor increases when the anode current.32 If the “normal” mode is deemed insufficient for a
is attached to the assembly due to the additional rotational particular exposure level, the tube must be run at “high speed”
moment of inertia added to the system. Intuitively, attaching mode because it is the only other option. However, running
the anode adds more mass to the assembly and more energy the tube at “high speed” mode puts significant wear and tear
is required to accelerate and decelerate the additional mass on the bearings, which are usually the first component to fail
to a given ω. Since the input torque remains the same it in the assembly. An advantage of our design is that it allows
must be applied for a longer period of time to achieve the for a continuum of operating speeds so that the anode rotation
desired ω. This fact leads to the question of how much slower speed can be precisely set depending on the exposure level.
would the motor accelerate with a heavier anode attached In the specific PWM implementation used for this motor
to the assembly. There are specific applications such as prototype a lookup table method is used for calculation of the
cineradiography that are much higher dose than conventional sine function. The discrete approximation to the sine function
radiography or fluoroscopy which use an anode such as the leads to slight errors in the computation of the weight vari-
Varian G-1593BI (5.25 in. diameter, 1500 kHU) (Ref. 31) ables ta , tb , and t0 , which translate into small variations in
that is much heavier than the Rad-4 anode attached to our the motor speed. These variations can be visualized as small
assembly (1.75 versus 0.64 kg). By simply replacing the spikes in the speed curve shown in Fig. 17. These variations
Rad-4 anode with G-1593BI anode the moment of inertia of are small enough to be ignored, but a lookup table that is more
our assembly is increased by a factor of three, which means finely sampled can be used if higher precision is required. The
that to a first order approximation it would take three times as ability to bring the motor to rest and restart it without requir-
long to accelerate to the desired ω. The geometry of our motor ing a new index pulse sent to the microcontroller is important
prototype would need to be altered in order to accommodate for a hybrid system implementation. If the motor is rotating
an anode as heavy as the G-1593BI, and methods to increase during MR imaging it can potentially cause artifacts in the im-
the acceleration of the assembly would need to be explored. ages. However, once the MR imaging is completed additional
Last, the value of tAccel is not affected by the presence of a x-ray imaging might be required so the motor will need to
magnetic field parallel to the axis of rotation of the motor. Any be brought back up to speed quickly. Using the PWM control
magnetic field in this direction produces forces that do not method the motor can be immediately restarted from its pre-
have large moment arms implying that any torque created by vious state and brought up to speed (3000 rpm within 10.56
these forces is small. Furthermore, the direction of this torque ± 0.06 at 59 mT) without having to run any start-up or ini-
causes it to be absorbed by the shaft and mounting assembly. tialization sequence, which will help to keep any delay time
The value for tAccel for B|| = 0 mT was slightly lower than the as short as possible.
values for B|| = 13 mT and B|| = 31 mT because the value of By mounting the motor assembly and keeping the position
B⊥ was slightly higher for the former setup (38 vs 35 mT). of one end fixed the assembly becomes analogous to a can-
tilever beam. During acceleration the assembly will vibrate,
IV.B. Effects of rotor angular offset and at a specific speed these vibrations will excite the natu-
ral frequency of the structure causing the magnitude of the
The effect of offsetting the rotor is moderate up to a 30◦
vibrations to increase dramatically. The motor should not be
offset after which α and maximum ω fall off more dramati-
operated at its resonance frequency for extended periods of
cally. After 30◦ any further offset generates negative torque
time as the vibrations could potentially damage components
for a fraction of the commutation period. The sensitivity of
as well as affect x-ray imaging performance. Attaching the
the performance of the motor to the value of γ is important
anode to the assembly changes the speed at which resonance
for this prototype because the feedback position sensor does
occurs because it changes the mass distribution of the system.
not generate an error checking index pulse once per rotation.
If the resonance is deemed to be at an unsafe location it is
The purpose of the index pulse is to indicate that a full ro-
possible to alter the location by changing various geometri-
tation has been completed and that the rotor position counter
cal parameters of the assembly such as the diameter of the
can be reset back to its initial state. In our implementation a
support shaft or the diameter of the bearings. Ideally, the res-
full rotation is determined after a predetermined number of
onance frequency would be higher than the desired value of
pulses from the quadrature sensor are counted by the micro-
ω. If not, the motor should be accelerated past resonance as
controller. Once the required number of pulses is reached the
quickly as possible. After passing through resonance, ω can
position counter is set back to 0. This method is less robust
be adjusted using a control method such as PWM to bring the
than using an index pulse because if there is any error during
anode to the desired rotation speed.
this process (i.e., a pulse is not recognized by the microcon-
troller for any reason), then the rotor becomes slightly offset.
V. CONCLUSIONS
IV.C. Speed control and resonance
We have developed a novel motor design for rotating an-
Pulse width modulation is well suited for electrically com- ode x-ray tubes that can be operated in the fringe field of a
mutated motors as it provides a stable method to control the MR system. The motor design is analogous to a three-phase
ω of the device. Typical induction motors have two operating dc motor with the radial component of the fringe field serving
speeds, a “normal” mode and a “high speed” mode. The op- as the stator field. A vacuum compatible prototype of this de-
erating speed is determined by the frequency of the applied sign was assembled using machineable ceramic or austenitic

Medical Physics, Vol. 40, No. 2, February 2013

022302-14 Lillaney et al.: A MR compatible motor for x-ray tubes 022302-14

stainless steel for the majority of the components. The assem- 8 P. Acher et al., “An analysis of intraoperative versus post-operative dosime-

bly was tested using a Helmholtz coil pair and the fringe field try with CT, CT-MRI fusion and XMR for the evaluation of perma-
nent prostate brachytherapy implants,” Radiother. Oncol. 96(2), 166–171
from a 0.5 T split bore MRI system to evaluate acceleration (2010).
times and determine the location of the first mode resonance 9 M. Carlsson et al., “Magnetic resonance imaging quantification of left ven-

frequency. The motor was able to accelerate to the minimum tricular dysfunction following coronary microembolization,” Magn. Reson.
operating speed of 3000 rpm in nearly 10 s at 59 mT, and Med. 61(3), 595–602 (2009).
10 P. Acher et al., “Comparison of combined x-ray radiography and magnetic
the resonance mode of the assembly was found to be below resonance (XMR) imaging-versus computed tomography-based dosimetry
the desired value of ω. Minor alterations to the current design for the evaluation of permanent prostate brachytherapy implants,” Int. Ra-
need to be carried out before it can be tested in vacuum us- diat. Oncol., Biol., Phys. 71(5), 1518–1525 (2008).
11 P. P. Chinchapatnam et al., “Anisotropic wave propagation and apparent
ing an electron gun. The position encoder must be replaced
conductivity estimation in a fast electrophysiological model: Application
with a vacuum compatible counter-part, and the chromic steel to XMR interventional imaging,” Med. Image Comput. Comput. Assist.
bearings must be replaced with austenitic stainless steel bear- Interv. 10(Pt 1), 575–583 (2007).
12 M. Matsumae et al., “World’s first magnetic resonance imaging/x-
ings that have a vacuum compatible lubricant. For future iter-
ray/operating room suite: A significant milestone in the improvement of
ations of the motor design further work is necessary to model
neurosurgical diagnosis and treatment,” J. Neurosurg. 107(2), 266–273
how acceleration and the location of the resonance frequency (2007).
change with the geometry and materials used for the assem- 13 M. Sermesant et al., “A fast-marching approach to cardiac electrophysi-

bly. This specific motor design was motivated by the need ology simulation for XMR interventional imaging,” Med. Image Comput.
Comput. Assist. Interv. 8(Pt 2), 607–615 (2005).
for a MR compatible rotating x-ray tube. The motor concept 14 R. Fahrig et al., “A truly hybrid interventional MR/X-ray system: Feasibil-
can also be adapted for other applications that require rotation ity demonstration,” J. Magn. Reson. Imaging 13(2), 294–300 (2001).
within a MR environment. 15 S. T. Kee et al., “MR-guided transjugular intrahepatic portosystemic shunt

creation with use of a hybrid radiography/MR system,” J. Vasc. Interv. Ra-

diol. 16(2 Pt 1), 227–234 (2005).
16 R. Fahrig et al., “Truly hybrid interventional MR/X-ray system: In-
ACKNOWLEDGMENTS
vestigation of in vivo applications,” Acad. Radiol. 8(12), 1200–1207
This work is supported by NIH Grant No. EB007626, (2001).
17 R. Fahrig et al., “First use of a truly-hybrid x-ray/MR imaging system
Stanford Bio-X Fellowship, and the Richard M. Lucas Foun- for guidance of brain biopsy,” Acta Neurochir. (Wien) 145(11), 995–997
dation. The authors would like to acknowledge Gary Virshup (2003).
18 A. Ganguly et al., “Truly hybrid x-ray/MR imaging: Toward a streamlined
and the x-ray tube team at Varian Medical Systems for ad-
vice related to the motor and overall x-ray tube design. The clinical system,” Acad. Radiol. 12(9), 1167–1177 (2005).
19 L. Brzozowski et al., “Compatibility of interventional x-ray and magnetic
authors would also like to thank Karlheinz Merkle and John resonance imaging: Feasibility of a closed bore XMR (CBXMR) system,”
Kirk of the Stanford Physics Department machine shop for Med. Phys. 33(8), 3033–3045 (2006).
20 Z. Wen et al., “Study of increased radiation when an x-ray tube is placed in
their invaluable advice related to machining details for the
motor prototype designs. a strong magnetic field,” Med. Phys. 34(2), 408–418 (2007).
21 J. A. Bracken et al., “Closed bore XMR (CBXMR) systems for aortic valve

replacement: Active magnetic shielding of x-ray tubes,” Med. Phys. 36(5),

1717–1726 (2009).
a) Author to whom correspondence should be addressed. Electronic mail: 22 Z. Wen, R. Fahrig, and N. J. Pelc, “Robust x-ray tubes for use within mag-

[email protected] netic fields of MR scanners,” Med. Phys. 32(7), 2327–2336 (2005).

1 T. J. Vogl et al., “Hybrid MR interventional imaging system: Combined 23 P. V. Lillaney et al., “Electrostatic focal spot correction for x-ray tubes op-

MR and angiography suites with single interactive table. Feasibility study erating in strong magnetic fields,” in American Association of Physicists in
in vascular liver tumor procedures,” Eur. Radiol. 12(6), 1394–1400 (2002). Medicine (Vancouver, Canada, 2011).
2 M. W. Wilson et al., “Experimental renal artery embolization in a combined 24 P. Lillaney et al., “Magnetostatic focal spot correction for x-ray tubes op-

MR imaging/angiographic unit,” J. Vasc. Interv. Radiol. 14(9 Pt 1), 1169– erating in strong magnetic fields using iterative optimization,” Med. Phys.
1175 (2003). 39(9), 5567–5583 (2012).
3 K. S. Rhode et al., “Registration and tracking to integrate x-ray and MR 25 B. S. Halliday, “An introduction to materials for use in vacuum,” Vacuum

images in an XMR facility,” IEEE Trans. Med. Imaging 22(11), 1369–1378 37, 583–585 (1987).
(2003). 26 J. T. Bushberg, The Essential Physics of Medical Imaging, 2nd ed. (Lippin-
4 J. Kucharczyk et al., “Cost-efficacy of MR-guided neurointerventions,” cott/Williams & Wilkins, Philadelphia, 2002), Vol. xvi, p. 933.
Neuroimaging Clin. N. Am. 11(4), 767–772 (2001). 27 Varian, RAD-4 Rotating Anode X-Ray Tube, 2004.
5 R. M. Chu et al., “Minimally invasive procedures. Interventional MR 28 P. Yedamale, “Brushless DC motor fundamentals,” Microchip Technical

image-guided functional neurosurgery,” Neuroimaging Clin. N. Am. 11(4), Note AN885, 2003.
715–725 (2001). 29 D. O. Neacsu, “Space vector modulation - An introduction,” in IEEE In-
6 K. S. Rhode et al., XMR Guided Cardiac Electrophysiology Study and Ra- dustrial Electronics Society (IEEE, 2001), pp. 1583–1592.
dio Frequency Ablation (SPIE, 2004). 30 R. H. Park, “Two reaction theory of synchronous machines,” AIEE Trans.
7 M. R. Ginks et al., “A simultaneous x-ray/MRI and noncontact mapping 48, 716–730 (1929).
study of the acute hemodynamic effect of left ventricular endocardial and 31 Varian, G-1593BI Rotating Anode X-Ray Tube, 2007.

epicardial cardiac resynchronization therapy in humans,” Circ. Heart Fail. 32 P. Sprawls, The Physical Principles of Diagnostic Radiology (University

4(2), 170–179 (2011). Park Press, Baltimore, 1977), Vol. ix, p. 365.

Medical Physics, Vol. 40, No. 2, February 2013