Torque Estimation Algorithms

for Stepper Motor

Master’s Thesis in Electrical Engineering

JULIA JANSSON
CHARLOTTE RÅDAHL

Department of Electrical Engineering

C HALMERS U NIVERSITY OF T ECHNOLOGY
Gothenburg, Sweden 2018
 Master’s thesis 2018: EENX30

Torque Estimation Algorithms for Stepper Motor

Julia Jansson
Charlotte Rådahl

Department of Electrical Engineering

Chalmers University of Technology
Gothenburg, Sweden
Torque Estimation Algorithms for Stepper Motor
JULIA JANSSON
CHARLOTTE RÅDAHL

© JULIA JANSSON, 2018.

© CHARLOTTE RÅDAHL, 2018.

Supervisor: Stefan Lundberg, Electric Power Engineering

Richard Kahl, Sigma Embedded Engineering
Examiner: Stefan Lundberg, Electric Power Engineering

Master’s Thesis 2018:EENX30

Department of Electrical Engineering
Chalmers University of Technology
SE-412 96 Gothenburg
Telephone +46 31 772 1000

Cover: Valve control system consisting of a butterfly valve, gearbox and 2-phase
bipolar hybrid stepper motor.

Gothenburg, Sweden 2018

iv
Torque Estimation Algorithms for Stepper Motor
JULIA JANSSON
CHARLOTTE RÅDAHL
Department of Electrical Engineering
Chalmers University of Technology

Abstract
Preventive health monitoring solutions of valve control systems for offshore applica-
tions are profitable due to possibility to plan maintenance work. Load torque is a
good indicator of system health as it provides information to the user whether the
controlling motor can generate enough torque to move the valve stem or if it is close
to its limits. To do this it is desirable to avoid expensive and large torque sensors,
opening up for the opportunity to investigate the possibility of estimating the torque
on the load based on motor parameters. The motor under investigation is a 2-phase
bipolar hybrid stepper motor with torque components from both the variable reluc-
tance and the permanent magnet in the rotor. Measurements confirm neglecting the
reluctance torque component to be acceptable, simplifying the derived motor torque
expressions. Based on a Simulink model of the system developed from mathemat-
ical expressions regarding motor and load together with measured motor specific
parameters, two motor torque estimation methods were developed and compared.
Simulations of the model were executed to estimate uncertainties in the load
torque algorithm. The load torque estimate was in turn depending on motor torque
algorithm simplifications in order to limit the number of sensors in a implementation
while upholding a approved tolerable accuracy. An estimation of the load torque
was presented based on the motor torque calculations derived from output power
and the assumption of a constant speed. The error of this estimate during a worst
case scenario is smaller than 13% having a total load torque above 2 Nm, a value
continuing to decrease with an increased load torque. In an implementation of this
algorithm, current sensors are the only necessary sensors while remaining parameters
are extracted from the reference values. The current sensors are used in systems
having a closed loop current control, making the suggested load torque estimation
not requiring any added sensors in an implementation.

Keywords: Load Torque Estimation, 2-phase Bipolar Hybrid Stepper Motor, But-
terfly Valve, Motor Torque, Simulink Model.

v
Acknowledgements
First and foremost we want to show our gratitude to Stefan Lundberg, our supervisor
at Chalmers University of Technology, for a tremendous patience when guiding us
through the snowy forest of motors. Without your knowledge and advise we would
not have learned half as much. We would also want to thank Richard Kahl and
Sigma Embedded Engineering for advice, feedback and the opportunity to write
this thesis. Without your support we would have never had this much fun while
learning at the same time. An additional thank to LK Valves for always taking the
time to answering our questions and providing this thesis with pictures of the valve.
We hope our work will be in your favour in order to prevent valve failures.

Julia Jansson and Charlotte Rådahl, Gothenburg, June 2018

vii
 Contents

1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Valve Control System 3

2.1 The Stepper Motor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.1.1 Hybrid Stepper Motor . . . . . . . . . . . . . . . . . . . . . . 8
2.1.2 Electrical Model Of The Hybrid Stepper Motor . . . . . . . . 8
2.1.3 Mechanical Model Of The Motor And Load . . . . . . . . . . 10
2.2 Controller and Drive Circuit . . . . . . . . . . . . . . . . . . . . . . . 11
2.3 The Valve And Gearbox . . . . . . . . . . . . . . . . . . . . . . . . . 13

3 Load Torque Estimation 15

3.1 2-Phase Stepper Motor Torque Expression . . . . . . . . . . . . . . . 15
3.2 Motor Torque Estimation Using Power . . . . . . . . . . . . . . . . . 17

4 Motor and Measurement Setup 19

4.1 Stepper Motor System In Thesis . . . . . . . . . . . . . . . . . . . . . 19
4.2 Motor Parameter Measurement Setup . . . . . . . . . . . . . . . . . . 21
4.2.1 Winding Resistance . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2.2 Maximum Flux Linkage . . . . . . . . . . . . . . . . . . . . . 22
4.2.3 Position Dependent Inductance . . . . . . . . . . . . . . . . . 22
4.3 Detent Torque Measurement Setup . . . . . . . . . . . . . . . . . . . 23

5 Extracted Stepper Motor Parameters 25

5.1 Motor Torque Measurements . . . . . . . . . . . . . . . . . . . . . . . 25
5.1.1 Winding Resistance . . . . . . . . . . . . . . . . . . . . . . . . 25
5.1.2 Maximum Flux Linkage . . . . . . . . . . . . . . . . . . . . . 25
5.1.3 Position Dependent Inductance . . . . . . . . . . . . . . . . . 25
5.2 Detent Torque . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

6 Simulink Model 31
6.1 Valve Control System model implementation . . . . . . . . . . . . . . 31
6.1.1 Motor Block . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
6.1.2 Mechanical System Block . . . . . . . . . . . . . . . . . . . . 35
6.1.3 Detent Torque Block . . . . . . . . . . . . . . . . . . . . . . . 36

ix
Contents

6.1.4 Load Torque Block . . . . . . . . . . . . . . . . . . . . . . . . 37

7 Simulation Results 39
7.1 Verification Of Model Motor Torque Estimation . . . . . . . . . . . . 39
7.2 Comparison Of Motor Torque Equations . . . . . . . . . . . . . . . . 41
7.2.1 Supply Voltage And Speed Variation . . . . . . . . . . . . . . 41
7.2.2 Resistance Variation . . . . . . . . . . . . . . . . . . . . . . . 45
7.2.3 Rotor Speed And Position Sensor Resolution . . . . . . . . . . 48
7.2.4 Reference Versus Simulated Rotational Speed . . . . . . . . . 51
7.3 Load Torque Assuming Constant Speed . . . . . . . . . . . . . . . . . 52
7.4 Load Torque Estimation During Worst Case Operation . . . . . . . . 54
7.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

8 Conclusion 59

Bibliography 61

x
 1
Introduction

1.1 Background
The stepper motor is a device generating a stepwise rotational motion, which has
an advantage in applications where a precise position control is of importance [1].
A microcontroller determines the sequence to produce winding currents and conse-
quently a magnetic field in different phases, creating a rotational movement of the
motor shaft by defined steps [2]. A result of the microcontroller is the stepper mo-
tors ability to be well functioning as an open loop system. In a closed loop system
the number of steps achieved by the motor could be counted and made sure to be
exact, an option of accuracy however is not necessary in all applications [3].
Due to its capability of creating sequential steps, the stepper motor has found
application in for example computerised numerical control machines (CNC) and
disk drives [1, 4]. Another load controlled by a stepper motor could be a valve. In
applications requiring a lightweight system the butterfly valve is of interest. The flow
of fluids in a butterfly valve is regulated through a disk opening and closing through
a quarter turn, with a rubber lining ensuring a minimisation of leakages. Every
opening and closing will tear the valve and as the number of openings and closings
depends on the application the valve lifetime is difficult to estimate. One option
to determine the mechanical health of the valve is to measure the torque required
to alter the position of the valve. A system with a preventive health monitoring
method could prevent failures and limit subsequent material and economical damage.
Furthermore, a product informing the user when reparation is required give rise to
a new business model. The supplier can sell the product as a subscription instead
of an one time cost, opening an opportunity to secure the relationship between the
supplier and buyer by signing more extensive contracts. In addition, unnecessary
substitution of valves can be prevented due to the increased knowledge of each
products individual lifetime. Replacements could be planned and performed at a
time most suitable for the company, an interesting solution concerning industries
directed towards offshore applications such as oil rigs and ships due to the limited
maintenance availability.
The torque produced by the motor to close and open the valve can be mea-
sured using a torque sensor. However, adding parts to the system also increase the
number of damageable components. Together with increased cost and space it is
therefore important to limit the number of system components. To minimise these
factors, the torque could instead be estimated from the stepper motor. This is a
favourable method as it would utilise the existing hardware and the estimator could

1
1. Introduction

be implemented in the controller program, compared to the bulky torque sensor.

1.2 Aim
In this master’s thesis a Simulink model of a specific stepper motor and load system
is developed with motor parameters extracted through measurements. Based on
the model at least two algorithms estimating the load torque without using torque
sensors will be constructed. An evaluation of the load torque estimation will be
performed, with respective algorithms compared to each other to establish the most
suitable choice for the given valve control system.

1.3 Limitations
The design of the stepper motor, its controller and the valve will not be altered or
optimised with respect to the scope of this thesis. Neither will considerations regard-
ing different materials going through the valve be taken into account, and thereby
their effects on the system behaviour. This due to that all system components are
not, in the beginning of the project, decided by the company.
The system will be modulated to consider a specific stepper motor. However,
by changing the motor specific parameters the model can be applied to other stepper
motors.
The remote diagnosis application of the system where the load torque esti-
mation algorithms are developed in an embedded application, will not be designed
in this thesis. The conclusions will be based on results provided by a simulation
environment and verification in a real system will not be performed.

2
 2
Valve Control System

This master’s thesis regard a valve control system consisting of a butterfly valve
controlled by a stepper motor. A schematic of the valve control system is shown in
Figure 2.1. The figure shows a power supply that feed a DC voltage to a drive circuit
regulated by a control circuit. Depending on the controller signals the windings of
the 2-phase stepper motor are energised in a specific order. The energised phases
A (A+, A−) and B (B+, B−) are seen in Figure 2.1 as a link between the drive
circuit and the 2-phase stepper motor. The current in the windings generate a
magnetic field which produce a torque that rotates the rotor. Through the gearbox
the mechanical torque is transformed to a preferred rotational force. This up scaled
force is used to open or close the valve and in turn control the flow. In the following
sections the components of the valve control system are described more in detail.

Figure 2.1: System components together with significant signals and their direc-
tion.

2.1 The Stepper Motor

The stepper motor is defined as a polyphase synchronous motor with both stator and
rotor having teeth consisting of a magnetic permeable material, making the stepper
motor salient as the source of flux depend on the rotor angular position [3, 5].
Shown in Figure 2.2 is the construction of the rotor, where the teeth, motor shaft
and the polarities are defined. With an advantage of low cost and high reliability
the stepper motor is additionally well functioning in open loop systems [2, 5]. As

3
2. Valve Control System

the name implies the stepper motor generates a stepwise rotational movement of the
motor shaft. The motor drive is regulated by a controller designed to transmit logic
signals to create the most suitable stepwise rotation for the given application [6].

Figure 2.2: Construction of the hybrid stepper motor rotor. The opposite polarity
of the permanent magnet rotor can be noted as well as the rotor shaft and the grey
rotor teeth.

The variable reluctance, permanent magnet and a mixture of the two, called
hybrid, are the three most common types of stepper motors. A hybrid stepper
motor practice both the source of flux principle from the variable reluctance mo-
tor together with the principle of the permanent magnet. Applying the principle
from the permanent magnet, the flux components considers both the stator and the
permanent magnets in the rotor [2]. With a current through the stator windings
the stator teeth are polarised. As a result of their polarisation and the permanent
magnets in the rotor, a field between the windings and rotor teeth occur moving
the rotor. Consequently the torque depend on the field strength resulting from the
flux components, and the angle between the fluxes [2]. Observed in Figure 2.3 is the
distribution of the rotor teeth of a hybrid stepper motor. Energising phase B so that
B+ becomes a north pole, and B- a south pole, will attract the rotor teeth and lock
the position shown in Figure 2.3. Based on that position, energising phase A so that
A+ has a north polarity and A- becomes a south pole will create a counterclockwise
rotational force, aligning the closest south tooth with the positive A-phase and the
closest north tooth with the negative A-phase.
A variable reluctance motor instead use the principle of reluctance to generate
torque. Ohm’s law for magnetic circuits is expressed as
mmf = Ψ<, (2.1)
and describes the relation between the magnetic field force mmf, the magnetic flux
Ψ and reluctance < [2]. Reluctance is a magnetic property equivalent to resistance
in electrical circuits, i.e. the magnetic flux will flow through the path with least
magnetic reluctance [7]. The working principle of the contribution from the reluc-
tance motor can be described by neglecting the magnetic poles in Figure 2.3, i.e.

4
 2. Valve Control System

Figure 2.3: Phase distribution of the rotor of a hybrid stepper motor. The figure
shows the rotor edge coinside with the shaft.

assuming the rotor not to be magnetised. The rotor is constructed of a permeable

material, therefor if phase B is energised, the rotor would be locked to the position
shown in the figure as the position provides the smallest air gap between the stator
and rotor teeth, resulting in a lower reluctance compared to the reluctance in a
larger air gap [2, 8]. If instead a current is applied to phase A the rotor would turn
either clockwise or counterclockwise in order to align the rotor teeth with phase
A. However, with the simple motor structure provided by Figure 2.3 the rotational
direction can not be determined since the distance between the closest rotor teeth
and the A phase is equal.
A hybrid stepper motor use both the characteristic of the permanent magnet
motor together with the variable reluctance stepper motor. It therefor use the
aspiration from the permanent magnet rotor teeth to attract or repel the excited
windings as well as creating a rotation in order to minimise the air gap between stator
and rotor, and thus obtain minimum reluctance [9, 10]. For applications requiring
high motion resolution a small step angle of the rotor is desired, as it denotes
the smallest individual rotational movement. Concerning systems employing larger
angular steps the variance reluctance stepper motor is profitable and for applications
requiring smaller steps the hybrid stepper motor is advantageous [3].
The corresponding static torque curve to Figure 2.3, used to describe the be-
haviour of the stepper motor, is shown in Figure 2.4, where the individual static
torque of each phase is illustrated together with the sum of the phase torque com-
ponents. The curve is obtained by altering the rotor position with mechanical angle
θm and measure the torque generated by each phase at the given angle [10]. Both
current direction and timing in the winding determine the performance of the step-
per motor and the total system, making the maximum torque limited with less
winding current [2]. Respective phase torque components shown in the figure are

5
2. Valve Control System

described as
τA = τkA cos(θm ) = kIA cos(N1 θm ),
(2.2)
τB = −τkB sin(θm ) = −kIB sin(N1 θm ),
where IA = IB = constant, k a scaling factor, θm is the mechanical rotor position
at 0° in Figure 2.4, and N1 is a constant depending on the motor design. For the
example in Figure 2.4 N1 is 3, for one mechanical turn the rotor position will be
repeated three times. Each torque component contributes to the total torque as

τtot = τA + τB . (2.3)

Assuming a full alignment of phase B at 0° and consequently a 90° misalignment of

phase A in the same position, phase B and A are positioned in equilibrium at 0°
respectively 30° as shown in Figure 2.4. Equilibrium defines the position where the
rotor and stator teeth are aligned [2], i.e. when the energised winding is aligned with
the opposite polarity of the rotor as shown in Figure 2.3. At angle θk the highest
torque component τk is generated. The component τk , also known as the motor pull
out torque, is dependent on the motor speed [10]. Using a τk -speed curve the relation
between maximum torque compared to the applied load torque at a given speed is
expressed. Additionally presented in a τk -speed curve are the motor resonance and
instability problems occurring at specific velocities [2]. Consequently, using a τk -
speed curve additional motor behaviour and dependencies can be observed.

H A
B
A+B
( k, k
)
1
Static Torque [Nm]

-1

-30 0 30 60 90
Mechanical Angle [deg]

Figure 2.4: Static torque curve of 2-phase motor.

As torque is a result of the motor winding energisation the sequence and ampli-
tude of the current fed to the windings are essential. The easiest stepping sequence

6
 2. Valve Control System

is single stepping where the rotor move in a number of distinct position during one
electrical period [2]. The static torque curves shown in Figure 2.4 describe the torque
components of the motor structure provided in Figure 2.3, whereas the position il-
lustrated in Figure 2.3 refers to the static torque curves at 0 mechanical degrees.
In that position winding B is fed a positive current, indicating phase B+ having
a north polarity. At position 30°, a positive current is instead fed to winding A,
moving the rotor another step counterclockwise. The two remaining movements of
the electrical turn are performed by in turn energising winding B with a negative
current, followed by an applied negative current in winding A at 60° respectively
90°. The described sequence is denoted as B+, A+, B−, A− [10]. For every stepwise
rotation of the rotor, the teeth will oscillate around the new position before settling.
These oscillations can be viewed as a damped oscillating system and every new step
will result in the same damped oscillations [10]. The step length is described by
motor geometry according to

360°
step length = , (2.4)
Nr · Ns

where Nr is the number of rotor teeth and Ns implies the number of phases in the
stator [2]. The two states and six rotor teeth shown in Figure 2.3 result in a step
length of 30°. Increasing the step resolution results in a smoother shaft rotation.
After single stepping half stepping is the next level to an increased step resolution.
Half stepping is introduced by repeating having one phase on and afterwards two
phases on simultaneously [2].
To further increase step resolution the excitation mode microstepping is ap-
plied. As stepper motors with a high number of stator and rotor teeth are compli-
cated to manufacture it is advantageous to divide the rotating steps into subsets.
Such an action is achieved by applying winding currents unequal in amplitude while
energising two phases simultaneously. This technique requires a closer surveillance
as feedback to control varying motor parameters such as temperature [2]. With a
position dependent current expressed as

ˆ
IA = Isin(θe ),
(2.5)
ˆ
IB = Icos(θe ),

the current together with (2.3) is altered as

ˆ
τtot = kIA sin(θm ) + kIB cos(θm ) = k Icos(θm − θe ), (2.6)

creating a motor torque dependent on both electrical position, θe , and mechanical

rotor position, θm , resulting in an increased stepping resolution.
Adding a load torque result in an angular displacement of the equilibrium
position and misalignment of the rotor and stator, a disadvantage associated with
the stepper motor [2]. This step position error is inevitable, but by designing a
steep curve around the step position the effects are decreased. Choosing a motor
with higher τk than necessary is another option to increase the steepness of the
curve, separating the actual curve from the sinusoidal approximation [2].

7
2. Valve Control System

The holding torque τH is the maximum achievable torque produced by the

motor before it starts an uncontrolled continuous rotation. This parameter is im-
portant as it provides an overall estimation of the system capability [2]. As shown
in Figure 2.4 the total torque produced by the motor at a given rotor angle is a sum
of the individual phase contributions. For a two phase motor τH is derived from τkA
and τkB as [10]
1 1 √
τH = √ τkA + √ τkB = 2τk , (2.7)
2 2
hence τk is defined as the effective value of τH .

2.1.1 Hybrid Stepper Motor

Since a variance reluctance stepper motor require at least three phases while the
investigated valve control system is designed to use a two phase stepper motor, a
hybrid stepper motor is consequently studied in this project [5]. The hybrid stepper
has the combined ability of both permanent magnet and variable reluctance stepper
motor. The stator is divided into poles, each with a surrounding coil [10]. As
the rotor of the hybrid stepper is a cylindrical permanent magnet with opposite
polarised edges it react to the produced magnetic field which results in a rotational
movement. An overview of the rotor is shown in Figure 2.2, where the structure
and polarisation distribution is defined. In order to provide a controllable stepwise
rotation of the rotor, the rotor and stator poles are divided into small sections i.e.
teeth, which are noticeable in Figure 2.2. Shown in the figure, the misalignment of
the north and south rotor teeth is 90°.
A motor with four stages of phase excitation as shown in Figure 2.4 the relation
of step length described in (2.4) is reduced to [3]

90◦
step length = , (2.8)
Nr

an equation true for all 2-phase bipolar hybrid stepper motors. With Nr = 50,
one mechanical turn of a 2-phase stepper motor is the result of 50 teeth alignment
periods. With one teeth alignment period being equal to one electrical period
θe = 360°
50
= 7.2° making one electrical period consisting of 4 steps making one step
length 1.8°. The mechanical angle in relation to the electrical angle is thereby derived
as θe = 50 · θm , indicating that one mechanical turn is translated into 50 electrical
periods. For a hybrid stepper motor the typical step length is 1.8° which provide a
higher angular position resolution compared to a variable reluctance motor [3].

2.1.2 Electrical Model Of The Hybrid Stepper Motor

The electrical model of the hybrid stepper motor refer to the winding circuit illus-
trated in Figure 2.5. The figure shows a simplified schematic of a basic winding
circuitry where the electric dynamics are mathematically described as

vj = Rij + ej , (2.9)

8
 2. Valve Control System

with R being the winding resistance, vj the applied phase voltage, i the system
current and ej the EMF for a specific phase j. The induced voltage depend on the
flux linkage as
dΨj
ej = , (2.10)
dt
consisting of both the winding inductance voltage and the induced back-EMF, ebj ,
seen in Figure 2.5.

ij R L

+ +
ωm θm

vj ej ebj +
−

− −

Figure 2.5: Circuitry of the stator windings, where resistance, inductance, EMF
and back-EMF are present, in addition to the winding applied input voltage and
the resulting winding current.

For a 2-phase stepper motor the total flux linkage in phase A and B are
described as
ΨA = LA (θe , i)iA + ΨM cos(θe ),
(2.11)
ΨB = LB (θe , i)iB + ΨM sin(θe )
where LA and LB are the position and current dependent inductance in winding A
respectively B. The currents iA and iB flows in respective phase while ΨM is the
maximum flux linkage from the permanent magnet through the winding. The angle
of the electrical period is denoted θe . By combining (2.10) and (2.11) the phase A
electrical dynamics are defined as
dΨA
vA = RiA +
dt
d
= RiA + (LA (θe , i)iA + ΨM cos(θe ))
dt
diA dLA (θe , i) d
= RiA + LA (θe , i) + iA + (ΨM cos(θe )) (2.12)
dt dt dt
diA
= RiA + LA (θe , i)
dt !
∂LA (θe , i) ∂θe ∂LA (θe , i) ∂i
+ iA + − ΨM ωe sin(θe ).
∂θe ∂t ∂i ∂t
The electrical dynamics for phase B is determined in the same manner and expressed
as
!
diB ∂LB (θe , i) ∂θe ∂LB (θe , i) ∂i
vB = RiB + LB (θe , i) + iB + + ΨM ωe cos(θe ).
dt ∂θe ∂t ∂i ∂t
(2.13)

9
2. Valve Control System

As a result of the permanent magnet in the hybrid stepper rotor there exists
a magnetic force unaffected of the winding energisation called detent torque τd [11].
Shown in Figure 2.6, τd exists between the stator and rotor adjacent teeth where the
latter attracts the former due to reluctance. With one electrical period being 7.2°
of one mechanical period, as described in Section 2.1.1, together with a step length
of 1.8°, there exist four steps within a period. One electrical period start with rotor
and stator teeth being aligned as shown in Figure 2.6 and ends when the rotor teeth
has moved four quarters.

Figure 2.6: Full alignment of stator and rotor teeth.

One electrical turn will therefor result in four positions where the detent torque
is maximal as shown in Figure 2.7. An assumption of a periodical detent torque is
due to the expectation of the teeth being equally separated. The detent torque is
defined as [12]
τd (θe ) = Ad sin(4θe ), (2.14)
with Ad being the amplitude of the position dependent detent torque.

2.1.3 Mechanical Model Of The Motor And Load

The mechanical model of a motor is derived from its mechanical dynamics. Con-
cerning a hybrid stepper motor the mechanical dynamic structure is set as
dωm 1
= (τe − τd (θm ) − τl ),
dt J (2.15)
dθm
= ωm ,
dt
with J being the motor inertia, τe , τd (θm ) and ωm the electrical torque, detent torque
depending on the mechanical rotor angle and the mechanical speed respectively [12].
A mathematical development of τe is provided in Chapter 3. Torque component τl
is the load torque of the system, generated by the valve, the gear box and other
mechanics in between [12].

10
 2. Valve Control System

1
0.8

0.6

Ideal Normalised Detent Torque

0.4 0.5
Normalised Total Voltage

0.2

0 0

-0.2

-0.4 -0.5

-0.6

-0.8
-1

-1
0 1 2 3 4 5 6 7
Electric Angle [deg]

Figure 2.7: Wave form of the normalised detent torque and normalised total phase
voltage over one electrical period.

2.2 Controller and Drive Circuit

The stepper motor is digitally regulated by a controller unit sending pulses to the
motor driver circuit [5]. Due to digital logic control problems appearing at moments
such as start, stop, acceleration and deceleration, the design of the driver circuit is
important when developing a stepper motor system. The controller can be applied
as an application specific integrated circuit (ASIC), on a microprocessor or other
logic integrated circuit (IC) [10]. The advantage of a microprocessor is the re-
programmability, a quality that makes it slower in comparison to an ASIC, as this
logic construction is optimised for the given application.
In the elaborated system the drive circuit act as a bridge between the controller
and the actual motor. Within the driver circuit several modules are integrated,
which are all resulting in a control of the winding current. Depending on the stepper
motor, two different driver types are used. The drives are divided as bipolar and
unipolar, and differs in the number of used windings and energisation sequence. The
unipolar drive use four windings, each attached to a positive power supply which in
turn is connected to a centre tap. Each winding is connected to ground by a switch,
controlling the current flow from the centre tap through the windings in a given
order making each winding being energised in turn. An example of the sequence of
energising is illustrated in a timing diagram shown in Figure 2.8 where A, B, C and
D are the separate windings i.e. the separate phases. The unipolar control ensures
the rotor to move one position at a time and is advantageous in applications with
a variance reluctance stepper motor due to the direction of the current not being
relevant when creating the torque [3, 5].

11
2. Valve Control System

Figure 2.8: Timing diagram of a unipolar motor control sequence.

A bipolar two phase control is utilised by turning on and off switches in two H-
bridges controlling the direction of the current flow in the motor windings. Figure 2.9
illustrates the A phase drive configuration where switch S1 and S4 as well as switch
S2 and S3 work in pairs. When S1 and S4 are turned on the voltage and current
through the winding inductance will be positive. With S2 and S3 turned on the
voltage and current flow in the opposite direction. The switch pairs cannot be
turned on at the same time, but by being off simultaneously the output current will
reach zero. This is necessary as the inductance current can not change direction
instantaneously and the stored energy in the inductance is dissipated through the
diodes during off-time. Current is controlled by altering the voltage on the windings
though Pulse Width Modulation, PWM.

Figure 2.9: A phase generated from H-bridge in the driver circuit.

The stepwise energisation of the two windings establish a timing diagram

shown in Figure 2.10 where a high value indicates a positive current in the wind-
ing and a low value represents a negative phase current with full stepping driver

12
 2. Valve Control System

mode[10]. Due to the direction of the current the windings will be energised in a
specific manner providing a four step rotation of the rotor. The bipolar drive is
suitable for the hybrid stepper motor [5].

Figure 2.10: Timing diagram of a bipolar motor control sequence.

2.3 The Valve And Gearbox

The load of the valve control control system consist of a gearbox and valve and the
torque from both components contributes to the load torque τl . The valve in this
valve control system is a butterfly valve which open and closes by 90°. It has a
compact design requiring a small installation space making it advantageous in space
limited applications [13]. The gearbox is connected between the motor shaft and
the load. It converts the motor torque and speed output to the high torque and
low speed required by the valve, enabling the possibility of using smaller motor.
Depending on the torque τe produced by the motor and the torque required by the
valve τv the ratio between the two torques, the gear ratio nG , is obtained as
τv = nG τe (2.16)
assuming a positive and constant gear rate [14].
The valve is used to control liquid flow. Its main components are the body,
stem, and disc mounted as shown in Figure 2.11, provided by Johan Deger at LK
Valves. Leakages are minimised as the disc closing tightly around the lining. Two
components creates the total valve torque as
τv = τc + τa , (2.17)
where τc is the torque originating from construction and τa is the aerodynamic torque
coefficient component. Construction torque τc depends on construction parameters
such as friction and will increase with age [15, 16]. This is why having a motor
able to produce more torque than nominally necessary is important, to count for
increased construction torque and thereby extend the life time of the motor and valve
system. The fluid flowing in the pipe applies pressure on the disk causing the second
torque component. This additional aerodynamic valve torque depend on the disk
angle and geometry of the nearby pipe. It is described by a normalised aerodynamic
torque coefficient component τa . When the disk is closed and positioned at 0°, τa
is generally small due to equal forces on both sides of the disk. Opening the disk
increase the contribution from the fluid, changing τa . At 80° the aerodynamic valve
torque reach its maximum value, corresponding to the position where the fluids in
the pipe has the largest impact on the disk. In the end of the opening process, τa
decrease to arrive at a small and balanced value at a fully opened disk at 90° [16, 17].

13
2. Valve Control System

Figure 2.11: CAD model of the butterfly valve provided by Johan Deger at LK
Valves.

14
 3
Load Torque Estimation

Torque is the rotational analogue to force, measured in newton meter and translating
Newtons second law F = ma into
dω
τ = J θ̈ = J , (3.1)
dt

where J is the body inertia equivalent the mass m and θ̈ the angular accelera-
tion equivalent the linear acceleration a. Superposition is utilised to measure a
net torque, hence the total torque of a motor is described as (2.15), relating the
produced electrical torque τe to the detent torque τd (θe ) and the load torque τl .
Rewriting (2.15) results in

dωm
τl = τe − τd (θe ) − J − Bωm , (3.2)
dt
which additionally includes the viscous damping constant B [9]. This constant
denotes for the friction and second order effects due to eddy currents and hysteresis
in the motor [9]. Noticeably, the produced electromagnetic torque is equal to the
sum of the applied load torque and the detent torque when the rotor speed is zero.
With an assumption of a constant rotational speed, i.e. the stepping of the motor
is smooth enough to be seen as uninterrupted reduce (3.2) to

τl = τe − τd (θe ) − Bωm . (3.3)

Estimating τl by evaluating the detent torque, loss component Bωm and motor
torque for the specific system enables the possibility to monitoring the valve health
without the usage of a torque sensor. Both B and the detent torque are small
components compared to the motor torque, why the latter is the main focus when
trying to estimate τl .

3.1 2-Phase Stepper Motor Torque Expression

An expression regarding the torque produced by a 2-phase hybrid stepper motor
is initiated using the rule of energy conservation in the coupling field. The same
reasoning is applicable to a motor with more phases. For the 2-phased motor the
differential form of the comprise

dWe = dWf + dWm , (3.4)

15
3. Load Torque Estimation

where dWe is the electrical energy applied to the coupling field, dWf the energy
from electrical system reserved in the coupling field and the mechanical energy dWm
moving the rotor [5].
Dividing the electrical input energy dWe in its fundamental components, it
can be expressed as
dWe = (eA iA + eB iB )dt, (3.5)
where eA and eB are the induced EMF from a change in flux concerning each phase
as shown in (2.10) and iA and iB are the respective phase currents [5].
The relation between mechanical energy and torque τe is described as as [5]

dWm = τe dθm . (3.6)

Fixing the rotor in a set position making dθm = 0 and causing dWm to be zero,
entails (3.4) and (3.5) to merge in [5]

dWf = dWe = (eA iA + eB iB )dt. (3.7)

Using (2.10) in (3.7) and integrating over the rotor angles, the total amount of
energy stored in the coupling field is described as
Z ΨA Z ΨB
Wf = iA dΨA + iB dΨB . (3.8)
0 0

With the rotor rotating a small distance dθe together with an assumption of constant
currents during this period, (3.4), (3.6) and (3.8) form

iA dΨA + iB dΨB = dWe = dWf + τe dθm . (3.9)

With dθm approaching zero, (3.9) is reformulated, expressing the motor torque as

∂Wf ∂ΨA ∂ΨB ∂Wf0

τe = − + iA + iB = , (3.10)
∂θm ∂θm ∂θm ∂θm

where the latter term Wf0 is the coenergy of the system. Coenergy is a concept not
having a physical aspect, however being used to express electromagnetic force [18].
In a motor assumed not to reach saturation the energy, as a function of angular
position and current, is linear in the magnetisation curve [5]. Coenergy resembles
the energy stored in the coupled field and is expressed as
Z iA Z iB
Wf0 = ΨA diA + ΨB diB . (3.11)
0 0

The flux linkage will then also be linear to the current as

ΨA = LA (θe )iA + ΨM cos(θe )

(3.12)
ΨB = LB (θe )iB + ΨM sin(θe ),

where LA (θe ) respectively LB (θe ) are the inductance of each phase depending on
electrical angle θe [5]. Inserting (3.12) in (3.11) together with θe = Nr · θm , then

16
 3. Load Torque Estimation

(3.10) results in
1 ∂LA (Nr θm ) 1 2 ∂LB (Nr θm )
τe = i2A + iB
2 ∂θm 2 ∂θm
| {z }
Reluctance Torque (3.13)
− Nr ΨM iA sin(Nr θm ) + Nr ΨM iA cos(Nr θm ),
| {z }
Permanent Magnet Torque

where the former segment of the equation is the contribution from the reluctance
torque and the latter is the contribution from the permanent magnet.
As shown in (3.13) the contribution from the reluctance torque is depending
on a rotor position dependent inductance. With aligned stator and rotor teeth
the reluctance is low introducing a high flux at a given current, enabling a high
inductance. At a misaligned position, the reluctance at a given current is higher
resulting in a lower inductance. Due to the construction of the stator and rotor
teeth, the inductance is approximately periodic with an average phase inductance
L0 and a rotor position dependent inductance variation L1 [3]. Concerning the 2-
phase stepper motor the A and B position dependent inductance are 180° misaligned
from each other and expressed as

LA (θm ) = L0,A + L1,A sin(2Nr θm )

(3.14)
LB (θm ) = L0,B − L1,B sin(2Nr θm ).

Using (3.14) in (3.13) the torque expression for a 2-phase stepper motor is
derived as  
τe = i2A Nr L1,A cos(2Nr θm ) − i2B Nr L1,B cos(2Nr θm )
(3.15)
+Nr ΨM (−iA sin(Nr θm ) + iB cos(Nr θm )) .

3.2 Motor Torque Estimation Using Power

Another estimate of τe can be derived from output power and rotational speed as
Pout
τe = . (3.16)
ωm
In a realistic model of the motor the system losses need to be accounted for, as they
are present during operation. Thereby, the input power can be related to the output
power as
Pout = Pin − Ploss . (3.17)
Calculating the output power for a 2-phase stepper motor is executed using (2.12)
and (2.13). Ignoring the saturation effect, leaves only a position variance [8]. The
input power to a system is then used to extract the electrical torque of the stepper
motor. Concerning phase A such an operation results in

Pin,A = vA iA
!
diA ∂LA (θe ) (3.18)
= Ri2A + LA (θe ) iA + ωe i2A − iA ΨM sin(θe ) ,
dt ∂θe

17
3. Load Torque Estimation

while phase B resulting in

Pin,B = vB iB
!
diB ∂LB (θe ) (3.19)
= Ri2B + LB (θe ) iB + ωe i2B + iB ΨM cos(θe ) .
dt ∂θe

Combined, (3.18) and (3.19) is described as

Pin,2-phase = PinA + Pin,B

diA diB
= R(i2A + i2B ) + LA (θe )iA + LB (θe ) iB
dt dt !
2 ∂LA (θe ) 2 ∂LB (θe )
+ ωm Nr iA − iA ΨM sin(θe ) + iB + iB ΨM cos(θe ) ,
∂θe ∂θe
(3.20)
with the the last term being Pout according to (3.17) using term identification.
Identifying the motor torque component from (3.13) and rewriting (3.20) as (3.16)
generates
Pin,2-phase = PinA + Pin,B
diA diB
= R(i2A + i2B ) + LA (θe ) iA + LB (θe ) iB
dt dt
!
(3.21)
1 2 ∂LA (θe ) 1 2 ∂LB (θe )
+ ωm τe + iA + iB .
2 ∂θe 2 ∂θe
To describe the generated motor torque in terms of input power and losses,
(3.21) result in
!
1 diA diB
τe = Pin,2-phase − R(i2A + i2B ) − LA (θe ) iA − LB (θe ) iB
ωm dt dt
! (3.22)
1 2 ∂LA (θe ) 1 2 ∂LB (θe )
− i + iB ,
2 A ∂θe 2 ∂θe

as an additional motor torque expression in comparison to (3.13), having both posi-

tion and rotational speed dependencies. Due to (3.14), (3.23) the torque expression
for a 2-phase stepper motor is written as
!
1 diA diB
τe = Pin,2-phase − R(i2A + i2B ) − LA (θe ) iA − LB (θe ) iB
ωm dt dt (3.23)
 
− Nr i2A L1,A cos(2Nr θm ) − Nr i2B L1,B cos(2Nr θm ) .

18
 4
Motor and Measurement Setup

In this section the employed stepper motor is described together with measurement
setups to determine different motor specific parameters.

4.1 Stepper Motor System In Thesis

The stepper motor under investigation is module PD60-4-1160 from Trinamic, com-
prising a controller, driver and motor. Specifications for the module are collected
from datasheets with an summary of important parameters stated in Table 4.1 and
a simplification of the module implementation with specific processors and motor
shown in Figure 4.1 [19, 20, 21]. The implementation consists of the 2-phase bipolar
hybrid stepper motor QSH6018-86-28-310 from Trinamic together with a controller
unit, driver stage and the sensOstep™ magnetic encoder to detect step [22]. The con-
troller unit is implemented with microcontroller ARM Cortex-M3, motion controller
circuit TMC429, pre-driver circuit TMC262 and a TMCL-memory [21]. Dividing
the controller design in subdivisions establish a possibility for the microcontroller to
control three different stepper motors simultaneously [23]. A MOSFET drive circuit
implemented with H-bridges transfer the required currents to the motor windings.
Typically the applied phase voltage has a value of 12, 24 or 48 V although the motor
can withstand a maximum value of 51 V [20]. With the number of rotor teeth being
1
50, each full step length is equal to 7.2°, corresponding to 50 of one mechanical turn.
The motor controller unit enables each full step to be divided into 256 mi-
crosteps. It also has an implemented current control, spreadCycle™, which be-
havioural characteristics are based on the chopper principle [22]. The chopper regu-
late the current to match a reference current by adjusting the applied phase voltage,
determining the current slope. Trinamics own coolStep™ function is additionally
implemented, adapting the winding current level depending on the load to minimise
the power consumption and place the module in energy saving mode when steady
state operation is reached [22]. By controlling the power transmission the efficiency
is maximised and heating problem decreases. Benefits arising from higher efficiency
is more perceptible with an increased rotational speed. As the coolStep™ function
require knowledge about the applied load torque the feature stallGuard2™ is inte-
grated in the controller unit [22]. This function measure the back-EMF level and
translate the result to a corresponding load torque value for the coolStep™ function.
The feature is used to locate the load torque value both for the coolStep™ function
and to monitor the load torque and thereby prevent motor failure due to overload.
Another feature implemented a microPlayer™ which increase the smoothness in ro-

19
4. Motor and Measurement Setup

Figure 4.1: Simplified circuit description of motor module PD60-4-1160 with spec-
ifications stated in [21] and [20].

tation and enable microstepping [22]. Motor systems from Trinamic are operated
using the combined protocol and programming language Trinamic Motion Control
Language TMCL™. TMCL™ for motor applications is developed using a PC-based
integrated development environment from Trinamic TMCL-IDE, connected to the
motor through USB or field buses CAN, RS232 or RS485 [21]. Commands can be
stored in the application and performed as stand-alone operation or be operated
directly from a host. The TMCL-IDE furthermore offers a possibility to graphically
monitor motor behaviour and parameters in real time. A large range of parameters
can be altered to achieve an optimal application setting. However, in the project
the standard values are considered and no fundamental characteristics have been
changed.

The 2-phase bipolar hybrid stepper motor controls a butterfly valve from LK
Valves with a holding torque of 250 Nm. According to Johan Deger at LK Valves,
a gearbox with gear ratio 84:1 is used in the system to accomplish this torque. The
gear box is supposed to be ideal, why the load torque of the system τl equals the
torque acquired by the valve τv . Furthermore the gearbox should transfer a rotor
speed of 21 rpm – 97 rpm to a valve speed of 0.3 rpm – 1.2 rpm with an actuation
time of 13 – 60s. Due to the torque of the valve increasing with age, the motor
should be able to deliver a torque to move a load requiring 300 Nm, having a safety
margin of 20%.

20
 4. Motor and Measurement Setup

Table 4.1: Specifications for Trinamic motor module PD60-4-1160 [19, 20].

Dimension Nema24
#Rotor Teeth 50
Rated Current, irms 2.8 A
Peak Current, imaxpeak 4A
Phase Voltage, vmax 51 V
Phase Voltage, vtyp 12, 24, 48 V
Holding Torque, τH 3.1 Nm
Phase Resistance at 20°, R 1.5W
Phase Inductance (typ.), L 6.8 mH
Rotor Inertia, J 840 g·cm2

4.2 Motor Parameter Measurement Setup

Some parameters required for the theoretical model and torque estimators were not
provided by the manufacturer and datasheet, whereas actual measurements on the
real motor were necessary. Experiments and measurements were also conducted to
verify the information and graphs provided by the datasheet.

4.2.1 Winding Resistance

In order to obtain an accurate model of the motor the winding resistance, seen in
Figure 2.5, needs to be determined. This was done using a multimeter. The resis-
tance measurement required the windings to be open circuited, as the measurement
probes were connected according to Figure 4.2.

ii R L

Multimeter ebi +
−

Figure 4.2: Measurement setup for winding resistance determination. The multi-
meter was connected to the open circuit winding.

21
4. Motor and Measurement Setup

4.2.2 Maximum Flux Linkage

To estimate the electrical torque given as (3.13) one of the motor specific parame-
ters to determine is the maximum flux linkage ΨM . As e = dΨ dt
, knowing the flux
linkage from each phase from (3.12), the expressions for each phase induced voltage
contribution are described as
dΨA diA ∂L(θe ) ∂θe
eA = = LA (θe ) + iA − ΨM ωm sin(θe ),
dt dt ∂θe ∂t
(4.1)
dΨB diB ∂L(θe ) ∂θe
eB = = LB (θe ) + iB + ΨM ωm cos(θe ).
dt dt ∂θe ∂t

Open circuit measurements result in zero current through the windings, establishing
the induced voltage being equal to the input voltage, arriving in

dΨA
eA = = −ΨM ωm sin(θe ), (4.2)
dt
dΨB
eB = = ΨM ωm cos(θe ). (4.3)
dt
The induced voltage was created by rotating the motor shaft in a known and con-
stant speed. By extracting the voltage curves over time using a digital oscilloscope,
the mechanical speed was derived. Knowing the induced voltage over time the exact
speed of the shaft rotation was established using the assumption of the induced volt-
age being two phase and offset by 90°. Such an estimation creates the opportunity
of constructing the two phases as a complex voltage vector V = VA + jVB . The
absolute of the created complex vector is the length of the vector and the ampli-
tude of the induced voltage. The mean value of the vector during a set number of
times will create a linear model whose slope is the frequency of the system. From
this frequency the correct speed of the rotation was calculated and the value of ΨM
determined.

4.2.3 Position Dependent Inductance

The rotor position dependent inductance was estimated applying the configurations
shown in Figure 4.3. When conducting the experiment an external arm was attached
perpendicular to the rotor shaft. As shown Figure 4.3 the arm had an equal length
x on both sides of the shaft and an evenly distributed weight to minimise the impact
on the rotor. Manual adjustment of the rotor angle was obtained by altering the
position by y length units. With a measuring stick at a given length from the
shaft, the rotational angle θ of the motor was with these lengths calculated using
trigonometric identities.
The inductance values were extracted by connecting a LCR meter to the wind-
ings and measure the inductance at distinct rotor angles. The used equipment was
an Atlas LCR model LCR40, with an accuracy of ±1.5% concerning inductance
measurements [24]. In order to obtain conclusive results the experiment was done
for both sides of the rotor, i.e. when the rotational movement was both clockwise
and counter clockwise.

22
 4. Motor and Measurement Setup

Figure 4.3: Setup of the position dependent inductance measurement.

4.3 Detent Torque Measurement Setup

In order to measure the detent torque the rotor was rotated by an external force. The
motor was therefor disconnected from its digital control to ensure the measurements
occurred without impact from other sources of force. To create a controlled rotation
of the rotor an arm was placed perpendicular to the shaft. The procedure to extract
the rotor angle θm were the same as to the measurements of the position dependent
inductance in Section 4.2.3. At a given angular position the weight this rotational
movement caused on the arm on the opposite side of the shaft was measured. The
setup of the detent torque measurement is shown in Figure 4.4.

Figure 4.4: Setup for detent torque measurement.

Using the length l from the rotor centre to the point where the weight m
was measured the torque on the shaft at a specific angular position is calculated as
τ = m · g · l with g being the gravitational acceleration of 9.81 m/s2 . The sampling

23
4. Motor and Measurement Setup

frequency of the data, i.e. the data points analysed during the experiment, needs
to fulfil the requirement of the specific motor being bipolar and having 2 phases
together with one electrical turn being 7.2°. This translates into that for one electric
period there must be at least two measurement points.

24
 5
Extracted Stepper Motor
Parameters

The following chapter contain results of performed motor measurements required to

obtain a correct model of the system.

5.1 Motor Torque Measurements

5.1.1 Winding Resistance
The individual resistance of each winding was measured using a multimeter as illus-
trated in Figure 4.2. As a result, measurements concluded that both windings had
a resistance value of 1.4 W.

5.1.2 Maximum Flux Linkage

Exposing the shaft of the investigated motor to a mechanical speed cause an induced
voltage of each phase as shown in Figure 5.1. As assumed in Section 4.2.2 the two
voltages are offset by 90° why the introduction of the complex voltage vector is
accurate. Calculation of the complex voltage angle from such an estimation creates
the model shown in Figure 5.2. Linearising this model results in the linear equation
of y = 1735x+42.53. From the slope of this straight line the mechanical speed of the
system is determined as 324.67 rpm, a value close to the, by a tachometer, measured
331.2 rpm. Rearranging (4.1) to describe a cosine function with amplitude Ψm the
flux linkages is additionally plotted in Figure 5.1. The maximum flux linkage Ψm is
determined to be 0.8247 Wb.

5.1.3 Position Dependent Inductance

Measurements of the position dependent inductance resulted in the graph shown in
Figure 5.3. Although some measured values are misplaced, the winding inductance
can be mathematically described as a sine. Due to one electrical period being
7.2° and the bipolar drive of the investigated motor, two periods of the inductance
variation are seen in the figure. The 180° shift of the measured phase inductance is
explained by the 90° phase misalignment and the inductance position dependency
provided in (3.14).
Presented in Table 5.1 are the key values extracted from the measurements
such as the maximum, minimum and mean inductance values in addition to the

25
5. Extracted Stepper Motor Parameters

30
1.5

20
1

10
0.5
Voltage [V]

[mWb]
0 0

-10 -0.5

Voltage phase A -1
-20
Voltage phase B
|Complex voltage|
Mean complex voltage
-1.5
-30
-4 -3 -2 -1 0 1 2 3 4
Time [ms]

Figure 5.1: Induced voltages and their corresponding flux linkage values.

50

45
Voltage Angle [rad]

40

35
-4 -3 -2 -1 0 1 2 3 4
Time [ms]

Figure 5.2: Model of approximate linear model resulting in a speed of 324.67 rpm.

largest inductance value deviation for both phases. With the individual phase in-
ductance mean value differing by 15 µH, the total mean inductance for both phases
is calculated as 6.4 mH. As the maximum deviation is 0.23 mH and 0.265 mH for
phase A and B respectively, which is approximately 4% of the mean value, the induc-
tance deviates with ±2% from the mean inductance value. Thereby, the inductance

26
 5. Extracted Stepper Motor Parameters

6.55
Phase A
Phase B

6.5

6.45
Inductance [mH]

6.4

6.35

6.3

6.25
0 1 2 3 4 5 6 7 8 9
Angle [deg]

Figure 5.3: The figure shows the position dependent inductance for phase A and
phase B.

is assumed to be constant and not position dependent.

Table 5.1: Result from position dependent inductance measurement.

Lmax Lmin Lmean Lmax deviation

Phase A [mH] 6.518 6.288 6.408 0.23
Phase B [mH] 6.529 6.264 6.393 0.265

An explanation to the outlying data could be a result from the difficulty to

establish a stable connection between the LCR meter probes and the winding circuit.
The used equipment to measure the inductance was the Atlas LCR model LCR40 [24]
with an accuracy of ±1.5% ±1.6µH, additionally explaining the edgy sine shape of
the plot.

5.2 Detent Torque

Shown in Figure 5.4 are the measurement results of the investigated motor detent
torque with the rotor rotating in both directions. As shown in the figure, the assump-
tion of a periodical detent torque is correct and can additionally be mathematically
described as a sine with an average amplitude of 0.09 Nm. The figure further present
the four maxima within one electrical period occurring at 1.8°, 3.6°, 5.4° and 7.2°, a
result of the four excitation steps of the bipolar 2-phase stepper motor to move the
rotor teeth one period. As the detent torque should be zero during an alignment
between the rotor and stator teeth, the illustration in Figure 5.4 is inaccurate. An

27
5. Extracted Stepper Motor Parameters

explanation to this result shown in Figure 5.4 could be a static torque component
due to motor construction which needs to be applied to rotate the rotor. Conse-
quently the result shows a combined static and detent torque. Figure 5.5 shows the
detent torque without the static torque offset, resulting in amplitude of the position
dependent detent torque determined to be 0.05 Nm.

0.14
Clockwise
Counterclockwise

0.12

0.1
Detent Torque [Nm]

0.08

0.06

0.04

0.02
0 1 2 3 4 5 6 7 8 9
Mechanical Angle [deg]

Figure 5.4: Position dependent result with static torque from detent torque mea-
surement with rotor direction specified. The stars represent the outlines of one
mechanical step, i.e. 7.2°.

28
 5. Extracted Stepper Motor Parameters

0.06
Clockwise
Counterclockwise

0.04

0.02
Detent Torque [Nm]

-0.02

-0.04

-0.06
0 1 2 3 4 5 6 7 8 9
Mechanical Angle [deg]

Figure 5.5: Position dependent without static torque result from detent torque
measurement with rotor direction specified. The stars represent the outlines of one
mechanical step, i.e. 7.2°.

29
5. Extracted Stepper Motor Parameters

30
 6
Simulink Model

A model of the motor system was constructed using MATLAB Simulink. This
section will elaborate on the Simulink model constellation of module PD-4-1160
from Trinamic.

6.1 Valve Control System model implementation

Using the extracted module PD-4-1160 parameters a valve control system model
was created using Simulink. In the Simulink model, all torque components earlier
denoted τi are changed to T_i, i being the system torque terms. An overview of
the model setup is shown in Figure 6.1 where the calculation of motor torque τe ,
load torque τl , detent torque τd (θe ) and the valve systems mechanical dependency
are represented with individual blocks and expressed by

dω
J = τe − τl − τd (θe ) − Bωm , (6.1)
dt
which is a rearrangement of the relation presented in (3.2). As shown in Figure 6.1
the Load Torque block is a stand-alone component of the valve control system and
is thus not dependent on the remaining blocks. Mechanical speed and position are
calculated in the Mechanical System block and are required in order to generate τe
in the Motor block as well as the torque contribution from the Detent Torque block.

6.1.1 Motor Block

A close-up of the Motor block is shown in Figure 6.2. A reference current is applied
to the system, followed by a Driver and Current Controller block where a phase
voltage is constructed by comparing the real current to the reference current. By
applying the winding characteristics and the constructed phase voltage, the block
representing Electrical Motor Component determine the real phase currents. The
real phase currents are used in a closed loop as both an input to the Driver block
as well as input to the electric torque calculation within the Mechanical Motor
Component block. As seen in Figure 6.2, both the mechanical speed and electrical
position are required input variables.
Shown in Figure 6.3 is the close-up of the Current Reference block which pro-
duces reference currents i_A_ref and i_B_ref. The currents are designed as sine
shaped signals with 90° phase shift. By applying an electric speed reference and
a current amplitude i_amp_ref, the system is controlled. To minimise oscillations

31
6. Simulink Model

Figure 6.1: An overview of the Valve Control System model implemented in

Simulink.

the speed reference omega_e_ref is created using a ramping step function with a
reference frequency. Based on the desired electrical speed the reference frequency is
chosen. The created reference speed omega_e_ref is integrated to access the electri-
cal position theta_e_ref. Euler’s formula is used to compose its exponential complex
function. After multiplication with a reference current amplitude i_amp_ref, the
exponential complex function real and imaginary components are extracted to ex-
press the reference currents i_A_ref and i_B_ref as sine and cosine wave forms.

Signals i_A_ref and i_B_ref are the input variables to respective phase
Driver and Current Controller block illustrated in Figure 6.4. The figure only shows
the build up for the Driver and Current Controller block for phase A as the block
for phase B is identical. Within this block each current reference is compared to its
corresponding real phase winding current. The current difference act as an input to
the relay block in which the input is compared to the value 0.05 and -0.05, represent-
ing the width of the hysteresis band. If the reference current is 0.05 A larger than
the real phase current, indicating that the real phase current should increase, the
relay output is 1. For a reference current 0.05 A smaller than the real phase current,
the relay output is set to 0. As the function of the Driver and Current Controller
block is to create a phase voltage, the output from the relay, either 1 or 0, act as
an input to the switch block. In the switch block the relay output is compared to
the value 0.1. If the relay output is larger than 0.1, i.e. 1, a positive voltage V_pos
will be applied as winding voltage and vice versa, creating a PWM voltage with a
magnitude of either V_pos or V_neg.
The resulting voltage from Figure 6.4 act as an input to the block Electrical

32
 6. Simulink Model

Figure 6.2: Motor block calculating the produced motor torque with individual
blocks determining the reference current, motor driver and current controller, the
motor electrical part and the mechanical motor characteristics.

Figure 6.3: Current Reference block producing reference currents.

Motor Component. Other input variables to the block Electrical Motor Component
are the mechanical speed omega_m and electrical position theta_e. A deconstruc-
tion of the block is shown in Figure 6.5 where the design of the block is derived
from (2.12) and (2.13). Measurements on the stepper motor conducted the position
dependent inductance to be negligible, reducing (2.12) and (2.13) to
diA
vA = RiA + LA − ΨM ωe sin(θe ),
dt (6.2)
diB
vB = RiB + LB + ΨM ωe cos(θe ).
dt
Noticeable in Figure 6.5 is the usage of omega_m instead of omega_e as stated in
the equation. The reason is that the flux component Psi_m was measured based

33
6. Simulink Model

Figure 6.4: Hysteresis Regulator block, exemplified with phase A, creating a phase
voltage depending on the difference between the reference and real phase current.

on the mechanical speed, hence a factor of Nr is already implemented in that term.

The position dependent terms for each phase in (6.2) is the induced back-EMF
earlier denoted as ebi and seen in Figure 2.5. Applying Laplace transformation to
the equation result in
vA = RiA + sLA iA − ΨM ωe sin(θe ) → vA + Ψm ωe sin(θe ) = iA (R + sLA ),
(6.3)
vB = RiB + sLB iB + ΨM ωe cos(θe ) → vB − Ψm ωe cos(θe ) = iB (R + sLA ),
as presented in Figure 6.5. The output consist of the real winding current i_A and
i_B.

Figure 6.5: Electrical Motor block resulting in phase currents i_A and i_B.

Shown in Figure 6.6 is the close-up of the Mechanical Motor Component block
constructing motor torque τe . Due to the negligible position dependent inductance
(3.15) is reduced to
τe = Nr ΨM (−iA sin(Nr θm ) + iB cos(Nr θm )) . (6.4)
By applying the same reasoning as for the Electrical Motor Component equation,
Nr in (6.4) is neglected when constructing the model in Simulink. This because the

34
 6. Simulink Model

measured flux component is based on the mechanical speed and not the electrical
speed.

Figure 6.6: Mechanical Motor Component block which calculates the electrical
torque produced by the motor using the real phase current, the electrical position
of the rotor and the maximum flux linkage using (6.4).

6.1.2 Mechanical System Block

In the Mechanical System block seen in Figure 6.1 the angular speed and position of
the rotor is derived as described by (6.1), whereas the block components are shown
in Figure 6.7. By dividing the input variable with the rotor inertia J, the derivative
of the mechanical speed is obtained. An integration block is applied, creating the
mechanical speed. Thereafter the mechanical rotor position is created using an
additional integration block. Multiplying the mechanical speed by the number of
rotor teeth determine the electrical speed and thereby the electrical position can be
obtained using an integration block. Shown in Figure 6.7 is a modification block
and a 2*pi constant, which result in a electrical position value ranging from 0 to
2π. Both the mechanical speed and electrical position are used as input variables in
the Motor block. In addition, the mechanical speed is multiplied with the viscous
damping constant B creating the term Bωm described in (6.1) and seen in Figure 6.1.

35
6. Simulink Model

Figure 6.7: Mechanical System block where the system input J dω dt

is used to de-
termine the mechanical speed, mechanical position, electrical speed and electrical
position.

6.1.3 Detent Torque Block

The Simulink model of the detent torque is based on (2.14) with parameter values
extracted from motor measurements to produce a sinusoidal detent torque with four
maximum during one electrical period. The implemented model of the detent torque
is seen in Figure 6.8 where the electrical position act as an input variable. In the
figure, an initial detent torque value T_d0 is defined and set to the static torque
seen in Figure 5.4. Thereafter, the sum of the components are multiplied by 4 due
to the relation in (2.14). A sine curve is formed with an amplitude of detent_amp,
resulting in the output variable T_d.

Figure 6.8: Detent Torque Block creating a sinusoidal torque component with four
maximas during one electrical period.

36
 6. Simulink Model

6.1.4 Load Torque Block

The total torque τv of the butterfly valve consist of the aerodynamic torque com-
ponent τa and a construction torque component τc as described in Section 2.3. As-
suming an ideal gearbox the load torque of the system equals the valve torque as
τl = τv . The construction of the load torque is shown in Figure 6.9 where τa is
simplified as a triangular wave. As the load torque shape and magnitude depend
on the valve position, the mechanical position act as an input variable, shaping the
component T_a. By applying a constant T_c an offset value can be implemented
in the Simulink model.

Figure 6.9: Load Torque Block creating the aerodynamic torque component as
a triangular wave whose vertical offset is determined by the contruction torque
component.

37
6. Simulink Model

38
 7
Simulation Results

Estimation mehods of the load torque τl in the valve control system are compared
based on the system employed in the Simulink model described in Chapter 6. Deter-
mining motor torque τe is a necessity when estimating τl and in the Simulink model
τe is generated by electrically modelling the motor as described by (3.15). Mo-
tor measurements determined the position dependent inductance to be negligable,
simplifying (3.15) to (6.4) and reproduced as τe1 according to

τe1 = Nr ΨM (−iA sin(Nr θm ) + iB cos(Nr θm )) . (7.1)

Another approach of calculating τe is derived from (3.16) together with (3.17)

using the output power and mechanical speed to describe the motor torque. For
a 2-phase hybrid stepper motor (3.16) is expressed as (3.23). Due to the position
dependent inductance being neglected, (3.23) is re-formulated and denoted τe2 as

VA iA + VB iB − (RA i2A + RB i2B )

τe2 = . (7.2)
ωm
In each section the used parameters for the specific simulation are specified.
These parameters act as input values to the Simulink model and denotes the ref-
erence values to the controller in an implementation. Simulation sequences corre-
sponding to signals measured by sensors in an implementation such as VA , VB , iA ,
iB and ωm , have been derived based on the two last simulated periods. The average
of these values are used when calculating τe2 . Concerning τe1 , its value is derived
calculating the mean of (7.2) during the two last simulated periods, rather than
calculating (7.2) using the mean of the individual parameters. The value of B is
not known in this thesis. Thereby an assumption of it having a 5% impact during
operation with a mechanical speed of 400 rpm was made, i.e. B = 0.0024 Nms/rad.

7.1 Verification Of Model Motor Torque Estima-

tion
The Simulink model is built to mimic Trinamic module PD60-4-1160. In the model,
τe1 is utilised to describe the mechanical motor behaviour. Verification of the Simulink
model is executed by comparing maximum motor specific torque versus velocity
curves generated by Trinamic, accessible in the hardware datasheet [20] for module
PD60-4-1160. A specification of the used parameters in the simulation are defined
in Table 7.1. The stepper motor controlling the valve is supposed to operate in a

39
7. Simulation Results

range of 21 rpm – 97 rpm, resulting in the full speed range provided by Trinamic not
being evaluated. Instead the tests were performed in the mechanical speed range of
20 rpm – 200 rpm to demonstrate the curve shape trends.
The verification was executed by varying the mechanical speed and observing
the maximum applied load torque before reaching instability. In order to detect
system instabilities a curve defining the difference between the electrical reference
angle and the real electrical rotor angle was established. As the motor is described as
a microstepping synchronous motor, the applied winding current lead the real rotor
angle and thereby cause a rotation. System instability indicates an angle difference of
90°. Instability was noted when the rotor speed became high and negative, indicating
a loss of rotor and stator synchronism. By successively increasing τl the system was
forced in to instability, enabling observation of at which torque the instability occurs.
Observations were thereafter compared to the torque curve provided by the PD60-4-
1160 datasheet [20], where an increase in torque for low speeds, reaching a top value
of approximately 2.7 Nm at 50 rpm can be noted. The top is followed by a decrease
in torque where the value reach 1.5 Nm at 200 rpm and 0.8 Nm at 400 rpm.

Table 7.1: Parameter settings for simulation verifying the Simulink model motor
torque estimation being equal to the torque curve provided by Trinamic.

Vin 24 V
Iin,rms 2.8 A
R 1.4W
ωm 20 rpm – 200 rpm
τl,app 10 Nm
B 0.0024 Nms/rad

The results from the Simulink simulations are shown in Figure 7.1. It satisfac-
torily indicates the motor to manage a load over 2.5 Nm up to 100 rpm comparable
to the result in the datasheet for module PD60-4-1160 [20]. Comparing the torques
during the period 100 rpm – 200 rpm, the simulation results in a different maximum
torque than defined by Trinamic. However, due to the low resolution in the figure in
the datasheet for module PD60-4-1160 [20], the supposed torque at a given speed is
difficult to extract and the result rather demonstrates a satisfying trend verification.

40
 7. Simulation Results

2.8

2.6

2.4
Max Load Torque [Nm]

2.2

1.8

1.6

1.4

1.2

1
20 40 60 80 100 120 140 160 180 200
Mechanical Speed [rpm]

Figure 7.1: Results from Simulink simulations with different rotational speeds and
their corresponding maximum allowed torque values.

7.2 Comparison Of Motor Torque Equations

As the two methods of calculating τe require different sensors when implemented,
a comparison of the two is important in order to draw the conclusion of the most
suitable approach when estimating τl . Different settings and simulated sensor noise
has been generated in order to provide a comparison of necessary sensor resolution
between τe1 and τe2 . Measurement uncertainties affect all necessary extracted param-
eters e.g. currents and voltages. However, in the following simulations uncertainties
concerning resistance and speed when calculating τe2 together with position and ΨM
when estimating τe1 are the only considered.

7.2.1 Supply Voltage And Speed Variation

One aspect of evaluation when comparing the torque algorithms are the affects of
different input voltages. With parameter settings defined in Table 7.2, two results
were generated varying the mechanical speed for a supply voltage of 24 V respective
48 V.
Figure 7.2 shows the estimated motor torques τe1 and τe2 and the total load
torque τl,app + Bωm during operation with a supply voltage of 24 V. The relation
between the estimates and the total load torque during 24 V input voltage operation
is shown in Figure 7.3. Increasing the supply voltage to 48 V results in the τe1 , τe2
and a total load torque as shown in Figure 7.4. The relation between the estimates
and the total load torque as for the 24 V case was done for 48 V operation and the
results are presented in Figure 7.5.

41
7. Simulation Results

Table 7.2: Parameter setting for simulation measurement comparing motor torque
algorithms for two separate supply voltages and speed variations.

Vin 24 V & 48 V
Iin,rms 2.8 A
R 1.4 W
ωm 20 rpm – 200 rpm
τl,app 1 Nm
B 0.0024 Nms/rad

1.08
1
e
1.07 2
e

l,app
+B m

1.06

1.05
Estimated Torque [Nm]

1.04

1.03

1.02

1.01

0.99
20 40 60 80 100 120 140 160 180 200
Mechanical Speed [rpm]

Figure 7.2: Comparison of τe1 , τe2 and the total load torque with a supply voltage
of 24 V during different operational speeds.

42
 7. Simulation Results

105
1
e
/ l,app +B m
2
104 e
/ l,app +B m

103
Estimated Torque Difference [%]

102

101

100

99

98

97
20 40 60 80 100 120 140 160 180 200
Mechanical Speed [rpm]

Figure 7.3: Relation in percentage between calculations of τe1 and τe2 with a supply
voltage of 24 V during different operational speeds.

Shown in Figure 7.2 and Figure 7.4 are the increase in generated motor torque
for respective supply voltage when increasing the mechanical speed. This is expected
due to the speed dependent term comprised as Bω. The curve representing the total
load torque is linearly increasing in both cases, whereas there is a slight variation in
τe1 and τe2 increment. As seen in Figure 7.2 and Figure 7.4, spikes occur in the same
applied speeds for both estimates, which are 125 rpm for 24 V input voltage and
100 rpm at 48 V voltage operation. These values are a result of system resonance,
a problem occurring in the real motor during certain speeds due to characteristics
of motor system and the load [2, 3]. The following simulations requiring a constant
speed will avoid these resonances. By considering the operational range provided
by LK Valves, a mechanical speed of 90 rpm is chosen for these simulations.
Further, the difference between the total load torque and the estimates are
depicted in Figure 7.3 and Figure 7.5. By comparing the figures concludes that
the differences are low. Including the spikes the maximum deviations are during
24 V 4.2% and 2.8% for τe1 respective τe2 and 0.9% and 2.8% for τe1 respective τe2
during 48 V. Neglecting the spikes, the maximum deviation shown in Figure 7.3
are 0.9% respective 1.7% for τe1 and τe2 . For operation with 48 V supply voltage
the maximum difference in percentage is 0.9% for τe1 and 1.1% for τe2 . Ideally
τe1 and τl,app + Bωm would be equal but due to the low motor inertia the curves
deviates from this assumption. Applying a low inertia result in the mechanical
speed oscillating in steady state, making the average detent torque not zero, causing
a torque difference between the estimated τe1 and the correct torque τl,app + Bωm .

43
7. Simulation Results

1.08
1
e
1.07 2
e

l,app
+B m

1.06

1.05
Estimated Torque [Nm]

1.04

1.03

1.02

1.01

0.99
20 40 60 80 100 120 140 160 180 200
Mechanical Speed [rpm]

Figure 7.4: Comparison of τe1 , τe2 and the total load torque with a supply voltage
of 48 V during different operational speeds.

Even so, the maximum difference is, neglecting the resonance frequencies, under 1%
and thereby the assumption of equality is acceptable.
Conclusions can be drawn that an increase in supply voltage will lower the
difference in percentage between the generated motor torque and the estimates,
making the estimates more accurate. This is however not unexpected as the system
is controlled using PWM and a higher voltage level would result in a better current
control and thereby a more adjustable system.

44
 7. Simulation Results

105
1
e
/ l,app +B m
2
104 e
/ l,app +B m

103
Estimated Torque Difference [%]

102

101

100

99

98

97
20 40 60 80 100 120 140 160 180 200
Mechanical Speed [rpm]

Figure 7.5: Relation in percentage between calculations of τe with a supply voltage

of 48 V during different operational speeds.

7.2.2 Resistance Variation

A comparison between the two motor torque algorithms was performed with respect
to variations in resistance value in the model and in the estimator τe2 . An increase
in temperature in the motor windings will increase the resistance value, which will
have direct impact when calculating power losses required in τe2 according to (7.2)
and indirect affecting the estimate τe1 . Two tests were executed where the first one
regarded a variation of winding resistance of the Simulink model and the calculation
of τe2 . The second test estimated the motor torque deviation of τe2 using a resistance
value unequal to the winding resistance in the Simulink model. The resulting mean
values of the individual estimates from the first test are shown in Figure 7.6. During
the simulation the parameters described in Table 7.3 were used.

Table 7.3: Parameter setting for simulation measurement comparing motor torque
algorithms for resistance variation.

Vin 48 V
Iin,rms 2.8 A
R 1W–2W
ωm 90 rpm
τl,app 1 Nm
B 0.0024 Nms/rad

45
7. Simulation Results

1.028
1
e
2
e
1.026 +B m
l,app

1.024
Motor Torque [Nm]

1.022

1.02

1.018

1.016

1.014
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
Resistance [ ]

Figure 7.6: Comparison of τe1 and τe2 as a function of a varying resistance with a
mechanical speed of 90 rpm.

Shown in Figure 7.6 the estimate τe2 is more affected by the Simulink model
parameter variation than τe2 , an expected result. However, even if there occur a
spike at a resistance of 1.7 W, τe2 has a fairly constant amplitude, making the over
all difference between τl,app + Bω and τe2 constant due to the simulated system
being current controlled. With an increase in winding resistance, the PWM voltage
compensate and gain a higher average value in order to provide the required current.
This behaviour can also be concluded as the maximum deviation in Figure 7.6 is
0.7%, which is small in comparison to the resistance variation. Thereby, a variation
in all resistive components will not contribute to a large variation in the estimates.
During the second test the resistance value affecting the power loss was varied
to compare a calculation of τe2 with a resistance deviating from the value used in
the Simulink model. The Simulink model winding resistance was held constant at
1.4 W while changing the resistance in (7.2) from 1 W to 2 W. Except for R = 1.4W,
the second test had the same parameter settings as described in Table 7.3. The
variation in τe2 is shown in Figure 7.7 together with the total load τl,app + Bω.
In Figure 7.7 a linear decrease in the estimated motor torque is depicted. This
is expected as the average current is assumed to be constant over the two evaluated
periods and that the power loss term in (7.2) is linearly dependent on resistance,
resulting in the curve slope. A similar reason can be applied to τe1 which is directly
dependent on the measured motor value ΨM . This value act as a pure gain in the
system and a parameter variation of that component will have a linear affect of τe1
as well.

46
 7. Simulation Results

1.8
2
e
1.6 l,app
+B m

1.4

1.2
Motor Torque [Nm]

0.8

0.6

0.4

0.2

0
0.8 0.9 1 1.1 1.2 1.3 1.4
Resistance [p.u]

Figure 7.7: Resistance variation in the power loss calculation affecting the esti-
mated motor torque calculated by τe2 and compared to the total applied load torque.
The resistance is defined in p.u. based on the measured winding resistance of 1.4 W

47
7. Simulation Results

7.2.3 Rotor Speed And Position Sensor Resolution

Within a simulation environment the presence of external noise and sampling fre-
quencies of sensors are neglected. In the real stepper motor system components
such as speed and position sensors depend on sampling time and possible resolu-
tion consequently making their determination imprecise. By introducing a random
noise in the simulation system the algorithm’s noise tolerance can be observed, from
which required sensor resolution can be determined. As the estimate τe1 is position
dependent and τe2 speed dependent, noise on both individual components need to be
added and their corresponding impact on the evaluated estimates. When conducting
the simulation test the model parameters described in Table 7.4 were applied in the
system. The varying parameters were the noise levels and the mechanical speed.

Table 7.4: Parameter setting for simulation measurement comparing motor torque
algorithms with position and speed noise.

Vin 48 V
Iin,rms 2.8 A
R 1.4 W
ωm 90 rpm
τl,app 1 Nm
B 0.0024 Nms/rad

In order to obtain the desired information the noise level was implemented in
the Simulink model. Simulink block Uniform Random Number was applied in the
model to return a random number within a defined range. As the noise only should
impact the output and not affect the mechanical behaviour of the motor, the Uniform
Random Number block was added to the output signal before extraction. Because
τe1 is position dependent and estimate τe2 speed dependent, disturbances were at-
tached to respective components. Noise levels ranging ±0.5 · 10−4 to ±0.5 · 103 were
applied to the system and the resulting torque estimates were evaluated using the
disturbed sensor signals. The different torque estimates with noise implemented
were compared to corresponding ideal conditions with no implemented disturbance.
Figure 7.8 show the results obtained from noise simulations with a mechanical speed
of 90 rpm. The figure show the estimated torque deviation in percentage compared
the individual torque values when no noise was applied to the system. According to
Figure 7.8, a noise with a value higher than ±0.5 · 10−2 will affect the τe1 estimation.
A noise level of ±0.5 · 101 can be applied to the system before τe2 deviates. How-
ever, both algorithms have to be evaluated individually as the signals with applied
disturbance has different nominal values.
A summary of the noise dependencies is shown in Table 7.5, where the me-
chanical speed, its corresponding electrical speed, the maximum allowed noise level
before deviation and the required mechanical resolution are presented. Position de-
pendent estimator τe1 has a maximum noise level determined to ±0.5 · 10−2 . As the
electrical position varies between 0 and 2π rad, a noise of ±0.5 · 10−2 is calculated

48
 7. Simulation Results

150
1
e
2
e

100
Motor Torque W Noise / Ideal Motor Torque [%]

50

-50

-100

-150
-4 -3 -2 -1 0 1 2 3
Noise ±0.5*10 x

Figure 7.8: Relation in percentage between τe1 and τe2 when noise is and is not
applied with a mechanical speed of 90 rpm. The applied noise levels reach ±0.5·10−4
to ±0.5 · 103 .

to its corresponding mechanical resolution in degrees as

0.5 · 10−2 1
Mechanical Resolution = ± 360° · . (7.3)
2π 50
1
The term 50 in (7.3) represent the transformation from electrical degrees, considering
the number of rotor teeth. The required mechanical resolution for 90 rpm is 0.006°
as seen in Table 7.5. Angular resolution denotes the allowed difference between the
sampled current and rotor angle in order to obtain a correct motor torque estimate.
If instead exposing the system to a noise level of ±0.5 · 10−1 , i.e. a position
sensor with a resolution of 0.06°, the error increase to 6.64%. Further increasing the
noise to ±0.5 · 100 result in an error of 70% during 90 rpm operation, a larger error
than the 20% error margin defined by LK Valves. Table 7.5 present the resolution
of sensors with results from Figure 7.8. Figure 7.8 also show a maximum allowed
disturbance level of ±0.5 · 102 for τe2 , which translated from electrical rad/s to an
error in mechanical speed of ±477.5 rpm, an interesting result as the reference speed
was 90 rpm.
A further investigation of the speed observed by the simulated sensor result in
the values described in Table 7.6. According to Figure 7.8 and Table 7.6, a differ-
ence in reference and observed mean speed can be 90.39 rpm without disturbing the
motor torque estimation. This means even if a mean sensor speed is twice as high
as the reference during 90 rpm operation, according to Figure 7.8, the difference will

49
7. Simulation Results

only be 1.8%. The unexpected result presented in Table 7.6 is due to the mean value
of approximation of the extracted parameters being used when calculating τe2 , i.e.
by averaging the noise the impact on the calculation will be smaller. If instead using
the extracted values directly in (7.2) and thereafter calculating the mean of the two
extracted periods, another result in required resolution would probably be obtained.
However, the result provided in Table. 7.6 show that an implementation using av-
eraged values of speed will depend less on the speed resolution up to 477.5 rpm or
a noise level of ±0.5 · 102 . Even so, the most correct and probable resolution in
implementation would be ±4.77 rpm, which in noise levels refer to ±0.5 · 100 .

Table 7.5: Result from maximum position and speed sensor noise and correspond-
ing resolution when ωm = 90 rpm.

Mechanical Speed Electrical Speed Noise Mechanical Resolution

τe1 90 rpm 471.2 rad/s ±0.5 · 10−2 0.006°
τe2 90 rpm 471.2 rad/s ±0.5 · 102 ± 477.5 rpm

Table 7.6: The resulting reference speed and mean sensor speed from tests where
noise is applied to the system.

Applied Noise ±0.5 · 100 ±0.5 · 101 ±0.5 · 102 ±0.5 · 103
Reference Speed [rpm] 90 90 90 90
Mean Sensor Speed [rpm] 90.91 99.03 180.39 993.13
Speed Difference [rpm] 0.91 9.03 90.39 903.13

50
 7. Simulation Results

7.2.4 Reference Versus Simulated Rotational Speed

To further simplify an implementation, the controller’s reference rotational speed
could be used to calculate τe2 rather than using a speed sensor. Simulating the model
with the same parameter values as in Table 7.2 give the results shown in Figure 7.9
and Figure 7.10, the difference between the simulations being supply voltage. Shown
in the two figures are the difference between calculating τe2 using the reference speed
and calculating τe2 with the simulated speed. A lower supply voltage result in an
outlying deviation of 1.5 Nmm due to resonance at approximately 120 rpm, not
included in the plot to instead clearer view the smaller deviations. Except this
outlying value, the methods of calculating τe2 does not differ more than 0.08%.
Increasing the supply voltage to 48 V results in an even smaller difference as shown
in Figure 7.10, except for an increased, again not included, outlying differential spike
due to resonance at 100 rpm measured to 19 Nmm. Neglecting the outlying value,
the largest deviation occur at 110 rpm where the difference between the reference
and measured speed was determined to 0.018%.

1.5

1
Difference [Nmm]

0.5

0
2
e

-0.5

-1
20 40 60 80 100 120 140 160 180 200
Mechanical Speed [rpm]

Figure 7.9: Difference in calculating τe2 using the reference speed or calculating τe2
using the simulated speed when the supply voltage is 24 V. The outlying spike at
100 rpm has a value of 1.5 Nmm.

51
7. Simulation Results

0.05

0
Difference [Nmm]

-0.05

-0.1
2
e

-0.15

-0.2

-0.25
20 40 60 80 100 120 140 160 180 200
Mechanical Speed [rpm]

Figure 7.10: Difference in calculating τe2 using the reference speed or calculating
τe2 using the simulated speed when the supply voltage is 48 V. The outlying spike
at 120 rpm has a value of 19 Nmm.

7.3 Load Torque Assuming Constant Speed

To simplify the load torque estimation algorithm a comparison between the load
torque when based on τe1 and τe2 were done while assuming a constant mechanical
speed. Thereby, the mechanical dynamics described in (3.2), reproduced as
dωm
τl = τe − τd (θe ) − J − Bωm , (7.4)
dt
is reduced to (3.3), re-formulated as
τl = τe − τd (θe ) − Bωm , (7.5)
simplifying the expression as no derivative need to be taken into consideration.
An investigation of this relation is of importance as a simplified expression would
facilitate the execution in hardware.
During simulations, the Simulink configurations presented in Table 7.7 were
applied showing only the varying factor during the tests being τl,app . The simulations
provided both τe1 and τe2 which were inserted in (7.5) where the corresponding τl,app
for each motor torque algorithm were estimated and compared to the real applied
load torque for evaluation of their accuracy.
Differences between the applied load torque τl,app and the estimated load
torques τl (τe1 ) and τl (τe2 ) are displayed in Figure 7.11, showing the largest differ-
ence of 5.3% held by τl (τe2 ) when the applied load is 0.2 Nm, decreasing with an

52
 7. Simulation Results

Table 7.7: Parameter setting for simulation measurements assuming a constant

speed.

Vin 48 V
Iin,rms 2.8 A
R 1.4 W
ωm 90 rpm
τl,app 0 – 3.3 Nm
B 0.0024 Nms/rad

increased load torque to after 0.6 Nm not exceed a deviation of 0.7%. The error of
τl (τe1 ) is much smaller and overestimate the load torque with a deviation of 0.1% at
1 Nm.

102

l
( 1e ) / l,app

l
( 2e ) / l,app
101
Estimated Load Torque / Applied Load Torque [%]

100

99

98

97

96

95

94
0 0.5 1 1.5 2 2.5 3
Load Torque [Nm]

Figure 7.11: Relation in percentage between estimated and applied load torque
assuming a constant speed of 90 rpm, for the both τl (τe1 ) and τl (τe2 ).

53
7. Simulation Results

7.4 Load Torque Estimation During Worst Case

Operation
Evaluations of the load torque estimates during non ideal conditions are essential in
order to establish a range in which the estimates are reliable and well functioning. It
is also important to investigate the system during a set of worst case scenarios as a
contemporary parameter deviation may affect the motor model behaviour differently
than when only considering individual deviations.
Two different simulations were done in order to provide information regarding
the estimates behaviour. The first simulation conclude a variety in load torque
having a constant speed while the second regard a constant load torque and speed
parameter variation. In both tests extraction of the necessary estimate parameters
were done during steady state operation and the assumption of constant mechanical
speed, why the term J dωdtm become zero. In addition, the detent torque component
was also neglected as the mean value over the last two simulation periods is close
to zero. A detailed description of the simulation parameter settings for respective
test is shown in Table 7.8. In order to provide reasonable results for τl (τe2 ), the
speed disturbance was set to ±0.5 · 100 , indicating a required sensor resolution of
±4.77 rpm. During the simulations τe1 and τe2 were estimated with a 10% error in
the ΨM and resistance value as noted in Table 7.8.

Table 7.8: Parameter setting for worst case simulations.

Speed Variation Load Torque Variation

Vin 24 V 24 V
Iin,rms 2.8 A 2.8 A
Rmodel 1.4 W 1.4 W
Rτe2 1.54 W 1.54 W
ΨM,model 0.8247 Wb 0.8247 Wb
ΨM,τe1 0.907 Wb 0.907 Wb
ωm 0-200 rpm 90 rpm
τl,app 1 Nm 0 – 3 Nm
B 0.0024 Nms/rad 0.0024 Nms/rad
Noise θe ±0.5 · 10−2 ±0.5 · 10−2
Noise ωm ±0.5 · 100 ±0.5 · 100

Results from the first test are shown in Figure 7.12 presenting the difference in
percentage between the estimated load torques τl (τe1 ) and τl (τe2 ) and the applied load
torque τl,app , where the estimated load torques both are comprised by the estimated
motor torque and the speed dependent torque component Bω. The figure shows a
constant τl (τe1 ), independent on the applied torque but with an overestimation of
approximately 10% which is explained by the +10% error of ΨM . In comparison,
the estimation of τl (τe2 ) is clearly dependent on the applied load. This is explained
by, due to the constant current amplitude and resistance, the loss component in the

54
 7. Simulation Results

120

100

80

60
Motor Torque Difference [%]

40

20
l
( 1e ) / l,app

l
( 2e ) / l,app
0

-20

-40

-60

-80

-100
0.5 1 1.5 2 2.5
Applied Load Torque [Nm]

Figure 7.12: Relation in percentage between τl (τe1 ) and τl (τe2 ) varying the load
torque during worst case parameter settings.

estimation being constant and overestimated, making its influence on τl (τe2 ) high,
especially during low torque levels. Estimate τl (τe2 ) become better when increasing
the applied load torque, reaching a difference of 11% compared to the total applied
load torque.
The results from the second test is shown in Figure 7.13. The figure illustrates
the difference in percentage between the estimated load torques and the applied
load torque τl,app during different speed settings. Similar to the result in the first
test, estimation method τl (τe1 ) is more accurate and stable than τl (τe2 ), who show a
large deviation for lower speed settings due to the speed error being more prominent
during lower speeds. Due to it overestimating during the higher speeds τl (τe1 ) would
in an implementation for health monitoring provide a more secure system. Even
though the estimation of τl (τe2 ) approach the applied load torque in both Figure 7.12
and Figure 7.13, the estimate provide a minimum deviation of 11% and 3.2% for the
first respectively the second test. During the second test, and shown in Figure 7.13,
τl (τe1 ) is behaving similarly as in the first test, with a constant overestimation of
approximately 10%.

55
7. Simulation Results

120

100

80
Motor Torque Difference [%]

60
l
( 1e ) / l,app

l
( 2e ) / l,app
40

20

-20
20 40 60 80 100 120 140 160 180 200
Mechanical Speed [rpm]

Figure 7.13: Relation in percentage between τl (τe1 ) and τl (τe2 ) varying the speed
during worst case parameter settings.

7.5 Discussion
Finding a suitable load torque estimator depends both on accuracy and convenience
in implementation. Limiting the implemented components will both reduce the price
together with restraining the number of possible failing component. The possible
options for estimating the load torque, e.g. using motor torque estimator τe1 or
τe2 , are therefor compared regarding specifications of the algorithms and necessary
sensors.
To evaluate the Simulink model results τe1 was compared with a curve in
datasheet [20] regarding motor module PD60-4-1160 from Trinamic. The results
suggest the τe1 estimator to be similar to Trinamic’s own curve. It is however im-
possible to conclude if the Simulink model is being correct in terms of maximum
torque due to the limited information existing around Trinamic’s data extraction.
We do for example not know if the data from Trinamic is extracted from simulations
or measurements on the real motor and if there exist individual changes between
motors. It is moreover difficult to confirm the exact moment of error occurring in
the simulation environment making the extracted torque value slightly higher than
they would in a real case setup. Therefor the shape of the simulated curve and the
curve provided by Trinamic were compared by pattern rather than exact output
torques. Optimally, a comparison between the physical motor and its simulated
version should be executed with furthermore extensive tests. This was however not
possible to obtain due to the physical motor not being able to produce the desired

56
 7. Simulation Results

torque at given speeds.

The two motor torque equations τe1 and τe2 were compared to a total load
torque τl,app + Bωm to investigate their accuracy. Simulation results indicates that
the two methods both are accurate, however with a favour to τe1 being the most
correct estimator of the two. Not surprisingly, due to it being derived from the
electrical motor definition. The two motor torque equations do however not differ
more than 1.1% in test scenarios with optimal conditions having a supply voltage of
48 V. Implementing τe1 requires a current sensor and position sensor not deviating
in sampling more than 0.006° in order to provide an error smaller than 6.6%. Such
an implementation additionally requires a measurement of ΨM for each individual
motor. This parameter is important to accurately determine as it acts as a gain
in (7.1), i.e. in an implementation with a 10% incorrect ΨM τe1 will have an equally
large error.
An implementation of τe2 requires current and voltage sensors for each phase.
Due to the winding current being controlled by PWM it is easy to extract the voltage
level directly from the controller using the reference voltage. Thereby the need for
voltage sensors is eliminated. A correct measurement of the resistance is important
as this, similar to ΨM , act as a linear component in the estimate. An implementation
of τe2 additionally requires a mechanical speed sensor with an accuracy of ±4.77 rpm.
Results proved a worse resolution could be managed if the algorithm is implemented
using the mean of the speed during a larger time interval. According to Figure 7.9
and Figure 7.10, the difference of calculating τe2 using the reference speed versus
measured speed is only 0.08% and 0.018%, with the smaller value corresponding
to the larger supply voltage. An increased supply voltage in other words decrease
the difference in calculating τe2 with the reference speed compared to the measured
speed. Both cases result in a difference from which an assumption of using the
reference speed rather than using a speed sensor is reasonable in order to simplify
an implementation of τe2 . Due to the dependency of ωm when calculating τe2 , it is
not possible to estimate the load torque when the speed is 0, i.e. in the opening
and closing moment of the valve. As the estimation of load torque in this thesis
concern a valve where the interesting torque to estimate is the maximum value
during an opening and closing period of 13 – 60 seconds, an alternative of not to
measure τe during a specified first and last time interval is acceptable. A conclusion
of the individually required sensors for τe1 and τe2 are defined in Table 7.9. Maximal
flux linkage ΨM , the number of rotor teeth Nr and winding resistance R are motor
specific and defined as constants in the implementation.
Ultimately the load torque estimation τl which use either of the two alternatives
of calculating τe is evaluated. Simplifying the expression by assuming a constant
speed and thereby neglecting the equation derivative results in an inaccurate τl
estimation up to 0.1% for τl (τe1 ) and 0.7% for τl (τe2 ), having a load torque larger
than 0.6 Nm in the latter case. With a load torque below 0.6 Nm τl (τe2 ) has an
deviation up to 5.3%. However, with a gearbox with ratio 84:1, a produced motor
torque of 0.6 Nm result in a torque of 50 Nm at the load which is only one fifth
of the max holding torque required by the motor according to LK Valves. As the
interesting torque to estimate is much higher than this breaking point of 50 Nm,
this result is determined to be insignificant for estimations. The detent torque can

57
7. Simulation Results

Table 7.9: Comparisons of necessary parameters measurements and sensors needed

for the two methods to calculate τe

τe1 τe2
Current Sensor x x
Voltage sensor
Rotor position sensor x
Rotor speed sensor
ΨM x
Nr x
R x

additionally be neglected in (7.4) due to it being periodical within a teeth alignment

making it an average zero. Such an assumption further simplifies the load torque
estimation algorithm as it neglects another position dependent variable.
During the worst case simulations the estimator τl (τe1 ) provides a more stable
result than τl (τe2 ) due to it providing a constant 10% overestimation of the load
torque. Estimator τl (τe2 ) provide large errors during low mechanical speeds due to the
speed dependency in (7.2). With a low mechanical speed an error in speed accuracy
will have a more significant impact on the estimation. However, by increasing the
speed the deviation of τl (τe2 ) will decrease with the two estimates both providing an
error of 10% at 150 rpm. It is clear, that for an implementation the use of accurate
parameters are more important considering τl (τe2 ) compared to τl (τe1 ).
We suggest that in order to improve the accuracy of the models 48 V should
be applied as a supply voltage, the necessary sensors should have a satisfying reso-
lution. Further, parameters ΨM and R used in the estimates of motor torque should
be accurately measured for each motor. We also want to emphasise that every sys-
tem has unique parameters and behaviour that needs to be determined before an
implementation.

58
 8
Conclusion

By estimating the load torque of a valve controlled by a hybrid stepper motor, the
mechanical health of the system could be determined. A study has been made
to provide and evaluate algorithms for load torque estimation, using only motor
parameters. Two motor torque estimations have been evaluated, together with sim-
plifications of the load torque expression, in order to determine a suitable algorithm
to implement in the valve control system.
Due to the presented results of error deviations and the fewer required sensors
in an implementation compared to the other option as presented in Table 7.9, the
suggested method to estimate the load torque of the valve control system is expressed
as
VA iA + VB iB − (RA i2A + RB i2B )
τl = − Bωm , (8.1)
ωm
which is a combination of the motor torque estimate τe2 and the assumption of
constant mechanical speed while neglecting the detent torque. The detent torque
is negligible due to it being periodical, having an average of zero. Based on the
simulation results the assumption of constant speed resulted in an underestimation
for τe2 of 5.3% during an applied load of 0.2 Nm. However, for applied load torques
above 0.6 Nm the estimate did not differ more than 0.7%. During all load torque
levels applied to the system during the simulation the estimate τe1 had a maximum
deviation of 0.1%, making the assumption of constant speed reasonable for both
cases for applied load torques above 0.6 Nm. Noted in (8.1) the estimate depend on
the Bωm term, in turn depending on the mechanical system and mechanical speed.
For systems with a low B value, the term can be neglected, further reducing (8.1).
However, measurements and determination of B are necessary in order to conclude
the term to be neglected. It is additionally important to implement the estimator
using a high voltage, a well defined resistance and not estimating the load torque
during the closing and opening moment of the valve to provide a well functioning
algorithm.
During the last test the system was put in a worst case scenario, where the
estimates were evaluated during operation with low supply voltage, a 10% overesti-
mation of ΨM and R in calculations, as well as the assumption of having a mechanical
speed sensor with the resolution of 4.77 rpm and a mechanical position sensor with
an accuracy of 0.006°. According to the results, the load torque estimation τl (τe1 )
had a constant overestimation of 10% for variations in both speed and applied load.
This overestimation is due to the implemented error in ΨM , making it important
to accurately measure the component if that estimation is to be used. For τl (τe2 )
the load estimation was not acceptable for both low speed levels and applied load,

59
8. Conclusion

whereas the difference between the real and estimated load torque in percentage for
90 rpm and 1 Nm operation was 24%. However, as the reason for the large underesti-
mation of τl (τe2 ) is due to the direct impact on speed and resistance, the mechanical
speed and accurate measurements of winding resistance important parameters to
extract correctly when implementing the estimate. Due to the knowledge LK Valve
possess regarding the specific system and that the algorithm with this knowledge
can be altered for specific implementations, we assume this worst case scenario to
have a low probability of occurring.
Thereby, this study proves it theoretically possible to estimate the load torque
within reasonable error margins in a valve control system based only on parameters
from a 2-phase bipolar stepper motor. Before an algorithm implementation mea-
surements on the intended motor need to be conducted. Motor parameters such
as winding resistance, winding inductance and maximum flux linkage are motor
dependent and the conclusions drawn in this thesis only being valid for module
PD60-4-1160 from Trinamic. With a torque sensor on the load, the measured and
estimated τl can be compared to determine if the estimation is acceptable outside
the simulated environment. If the results are accurate, the estimation of τl is imple-
mented in hardware to be calculated in a real time system. Questions regarding a
hardware implementation require further investigation in order to obtain an optimal
mechanical health monitoring system.

60
 Bibliography

[1] K. V. Arzhanov and Y. A. Shurygin, “Energy saving control algorithms of step-

per motor drives for metal-working machines equipped with cnc unit,” in 2015
International Conference on Mechanical Engineering, Automation and Control
Systems (MEACS), pp. 1–4, 2015.
[2] A. H. et. al., Electric motors and drives: fundamentals, types, and applications.
Amsterdam: Elsevier/Newnes, fourth ed., 2013.
[3] P. Acarnley, Stepping motors: a guide to theory and practice, vol. 63. London:
Institution of Electrical Engineers, 4.;4; ed., 2011.
[4] S. E. Mirbagheri, M. Al-Bassyiouni, and A. Dasgupta, “Bearing wear model for
optical disk drive stepper motor,” pp. 1274–1280, IEEE, 2012.
[5] F. Khorrami, P. Krishnamurthy, and H. Melkote, “Modeling and adaptive non-
linear control of electric motors,” 2003.
[6] B. C. Kuo, Stepper motor. 2014.
[7] K. V. Manning, “Reluctance,” 2014.
[8] P. Zaskalicky and M. Zaskalicka, “Analytical method of calculation of the cur-
rent and torque of a reluctance stepper motor using fourier complex series,”
pp. 899–902, IEEE, 2008.
[9] T. Kenjo, Stepping motors and their microprocessor controls. Oxford University
Press, New York, 1984.
[10] H. K. et. al., Small Electric Motors. Institution of Engineering and Technology,
1998.
[11] M. Zribi and J. Chiasson, “Position control of a pm stepper motor by exact
linearization,” IEEE Transactions on Automatic Control, vol. 36, no. 5, pp. 620–
625, 1991.
[12] B. Henke, O. Sawodny, S. Schmidt, and R. Neumann, “Modeling of hybrid
stepper motors for closed loop operation,” IFAC Proceedings Volumes, vol. 46,
no. 5, pp. 177 – 183, 2013.
[13] K. e. a. Ogawa, “Hydrodynamic characteristics of a butterfly valve — prediction
of torque characteristics,” ISA Transactions, vol. 34, no. 4, pp. 327–333, 1995.
[14] H. Grotstollen, “Optimal design of motor and gear for drives with high acceler-
ation by consideration of torque-speed and torque-acceleration product,” IEEE
Transactions on Industry Applications, vol. 47, no. 1, pp. 144–152, 2011.
[15] C. Solliec and F. Danbon, “Aerodynamic torque acting on a butterfly valve.
comparison and choice of a torque coefficient,” Journal of Fluids Engineering,
vol. 121, no. 4, p. 914, 1999.

61
Bibliography

[16] M. J. Morris and J. C. Dutton, “Aerodynamic torque characteristics of butterfly

valves in compressible flow,” Journal of Fluids Engineering, vol. 111, no. 4,
p. 392, 1989.
[17] B. Nesbitt, S. (e-book collection), E. C. (e-book collection), and K. (e-book
collection), Valves manual international: handbook of valves and actuators.
Burlington, MA;Oxford;: Butterworth-Heinemann, 1st ed., 2007;2011;.
[18] P. C. Krause, O. Wasynczuk, S. D. Sudhoff, and S. Pekarek, Theory of Elec-
tromechanical Energy Conversion, vol. 75, pp. 1–52. Hoboken, NJ, USA: Wiley,
3 ed., 2013.
[19] Trinamic, “QMOT QSH6018 manual.” [Link]
fileadmin/assets/Products/Motors_Documents/QSH6018_manual.pdf,
2014. [Online; accessed 25-May-2018].
[20] Trinamic, “Hardware manual PD-1160.” [Link]
fileadmin/assets/Products/Drives_Documents/PD-1160_hardware_
[Link], 2013. [Online; accessed 25-May-2018].
[21] Trinamic, “Firmware manual PD-1160.” [Link]
fileadmin/assets/Products/Drives_Documents/PD-1160_TMCL-firmware_
manual_V104.pdf, 2017. [Online; accessed 25-May-2018].
[22] Trinamic, “TMC262 datasheet.” [Link]
assets/Products/ICs_Documents/TMC262_Datasheet.pdf, 2016. [Online;
accessed 25-May-2018].
[23] Trinamic, “TMC424 datasheet.” [Link]
assets/Products/ICs_Documents/TMC429_datasheet.pdf, 2015. [Online;
accessed 25-May-2018].
[24] PEAK Electronic Design ltd, “Atlas LCR Model: LCR40, Product Breef,”
2012.

62