The author is active in this field of interest as industrial designer for electrical drive systems for more
than 30 years and these lecture notes reflect a part of the knowledge gained with the electrical drives
in the central position.
Extending the knowledge on electrical drives is mainly driven by the high demanding projects done
by Philips Applied Technologies for its customers.
The electrical drives are the linking pin in these notes, directed to electrical drives for precision
engineering. In preparing these notes the decision is made to give preference to a system approach.
Many times a reference was made to thermal aspects, dynamics, control, electronics and the
consequences of tolerances. The electrical drives are considered here as a component to be selected,
based on its characteristics. The consequence is that typical motor design issues are described, but not
analyzed by means of mathematical formulations, as can be found in the literature directed to the
design of motors and actuators.
J.C. Compter
Eindhoven 22-06-2007
In controlled systems the information obtained from sensors or commanded signals is used to
influence the state of the system. The signals are used to exert forces or to control other types of
“actuators” (fans, valves, heaters, etc.). In electro-mechanical systems it is generally necessary to con-
vert electronic signals to mechanical forces or torques. The specific characteristics of the actuator may
be of noticeable influence on the performance of controlled systems.
Several types of electromechanical actuators exist. Examples are rotating or translating (linear) DC-
motors, step motors, AC-synchronous motors and AC-asynchronous motors. All these motors have
different characteristics and are specially suited for a typical application.
In this section some basic theory on electro-mechanical actuators will be presented. Tools used to
select motors for a certain application will be introduced. Finally a first impression of the influences
on the control system behaviour will be presented.
If we only need a constant speed, for instance for a transport system, we could take an AC-
synchronous motor or an AC-asynchronous motor. Looking to the dynamical behaviour one will see
however for these motors a considerable time-constant or oscillating response. In a servo system,
motors with a well predictable response, as holds for the DC-motor, are preferred. This can also be
obtained for the two mentioned AC-motors with the vector control technology, but this requires an
advanced and expensive motor controller and the related costs can only be accepted for powers above
1 kW. The majority of servo-systems within Philips are in the range 1 ... 400 watt, so we will
concentrate on the DC-motors.
Examples of DC-motors are: disc armature (Fig. 1.1), hollow rotor (Fig. 1.2), iron armature (Fig. 1.3)
and brushless (Fig. 1.4).
The toothed structure of the rotor gives rise to cogging, which are reduced by skewing the rotor slots
in the more expensive motors.
Strengths
• The power density, defined as Pmech/volume, is by far the highest, because the magnets only
have a small air gap to overcome. The rest of the flux path consists of the magnetically highly
conductive iron. The end result is less weight, less volume and low price.
• The iron guarantees mechanical robustness; the windings are anchored in it by means of a
moulding resin. Considerable attention has generally been given to secure fixing of the
winding head and to the connections to the collector. A high resistance to centrifugal forces
and accelerations is achieved in this way.
• The thermal capacity is great because of the presence of the rotor iron; short-time peak
powers are consequently readily absorbed. The small dimensions resulting from the high
power density do generally result in the thermal resistance Rth1 being slightly higher.
• A low mechanical time constant can be achieved through the slim structural shape of the rotor
in combination with the high power density; < 5 ms is feasible.
• The large number of producers means that “second sourcing” is possible at good prices.
Weaknesses
• The rotating iron produces cogging in addition to eddy current and hysteresis losses.
• The fact that the conductors are located in the slots of the iron rotor gives rise to a substantial
self-inductance of the rotor windings. This makes the commutation of the rotor coils difficult.
It also gives rise to a time constant in the control loop; with fast torque changes the self-
inductance means that the amplifier must have a safety margin on the voltage if it is to
overcome L.di/dt.
• The drawbacks of the iron armature motor will not be found in the hollow rotor motor. In this
motor the windings are located in the air gap and together form a cage that has been made
into a solid entity with resins. This cage has then been fixed to the shaft of the motor.
Because of this structure the reluctance for the magnets is factors greater than in the iron armature
motor; the magnetic field must cross four air gaps and pass the cage twice. Achieving acceptable field
strength requires large magnets in these motors.
Strengths
• No iron losses and no cogging.
• A small moment of inertia J, because in this case the iron is not turning in the same direction.
• The lower self-inductance moves the commutation limit to higher values. The self-inductance
is lower because there is an air gap on both sides of the windings. In addition, the magnets
located on one of the two sides also behave like air. (The permeability µr of magnets is very
similar to air.)
Weaknesses
• The power density is a factor 3 lower than that of comparable iron armature motors, because
appreciably larger magnets must be used.
• Greater vulnerability to peak torques and high speeds. The cage construction has lower limits
than the iron armature rotor.
• In the absence of the rotor iron the thermal capacity is low. At the same time the thermal
resistance Rth1 is moderate because of the large air gaps and the impeded heat dissipation via
the shaft. This motor is therefore better suited to short-time loads. An increase of 3* in the
thermal load can be achieved by the introduction of a forced air current via a separate blower.
Magnets are placed on both sides of the disc to obtain a sufficiently strong field.
Strengths
• No cogging, no iron losses.
• A small moment of inertia and a high value of S, possible through the use of a lot of magnetic
material, lead to a low mechanical time constant.
• A very low self-inductance and electric time constant, because of the existence of an air gap
and magnetic material respectively on either side of the copper. The commutation forms
hardly any limitation in the motor performance.
• Because of the large rotor area the thermal resistance is low. The low thermal capacity of the
PCB version is partly offset by the high permissible temperature of 155ºC. The epoxy version
has a temperature limit of 110ºC, but does have greater thermal capacity.
• The better controllability of the dimensioning leads to a higher permissible shaft speed,
because imbalance and distortion come less into play; 8000 - 10,000 rpm occur. Extremely
robust.
• The flat structural shape makes combination with an encoder, brake or tacho for example
easier.
Weaknesses
• The low power density and low values for S/volume and S/mass are a consequence of the
voluminous stators. Partly because of this prices are high.
• Usually AlNiCo is used as magnetic material. It is a material with a high sensitivity to
demagnetisation. Peak currents must therefore be controlled, it being important to remember
that these motors have hardly any damping effect by way of a high self-inductance.
• To make re-magnetisation possible by the user, the manufacturer generally has the
magnetisation windings located around the magnets.
• Depending on the type of PWM amplifiers, the low self-inductance can prevent the amplifier
working properly. The low self-inductance means that there are sharp variations in current.
Even with a torque T = 0, this leads to unnecessary ohmic losses; it may be necessary to take
the step of using (voluminous) smoothing coils.
• The good control engineering properties mean that the disc motors are frequently used for
robots and in servo-controlled machines. But the latest developments in the field of brushless
motors (AC synchronous and induction motors) mean that this choice is no longer as self-
evident as it was.
NOMINAL QUANTITIES
Nominal values of torque and speed (or current and voltage) are generally given for an ambient
temperature. The main usefulness this has is that it makes comparison with other motors possible. At
the same time a life of the motor can be given under these conditions. For the user of a motor this life
It is based on the theory of special relativity. Although very interesting this will not analysed here (see
Interpretation of Classical Electromagnetism, W.G.V.Rosser, ISBN 0-7923-4197-2).
The speed of the charge dq is given by:
v a = dl . 2.2
dt
The current [ampere] in the conductor is, according the definition of Ampere:
2.3
I = dq .
dt
The force [newton] on the conductor is obtained by substitution and integration over the length of the
wire. The force equals:
f t = B. I .l. 2.4
The field B is assumed to be generated by a permanent magnet. For a motor with N windings on a
radius r, the torque [newton.meter] equals:
So, the torque is proportional to the current, when using a permanent magnet. For this reason, motors
with a permanent magnet are used in servo systems. This type of motors we will call servomotors.
Now the induced voltage, the back EMF will be analysed. If the conductor is moving in a field B (in
the direction given in figure 2.1b), the force on the charge dq is given by: df a = dq.B.vt . The force
l l
a 2.6
EMF = ∫ Edl = ∫ dl = B.vt .l.
0 0
dq
For a motor with N windings on a radius r rotating with the speed ω [rad/sec] we have the next
relation:
vt 2.7
EMF = N .B.l.r. = K e .ω .
r
So the EMF is proportional to the motor speed if the field B is constant. The back-EMF constant is
equal to the torque constant and we will use from now on the symbol K with as units [Nm/A]! Be
aware of the units applied; many times one will meet Kt with the units Nm/A and Ke with as units
V/1000 rpm. From now on the symbol E will be used for the back EMF, because this is the usual
symbol.
The motor windings have a resistance R. There fore we must supply a voltage U to the motor
according:
U = E + I .R. 3.1
The incoming electrical power Pel is partly converted to mechanical power and dissipation:
Pel = U . I = E . I + I 2 R 3.2
The loss in the motor Pdiss results in to a temperature rise of the windings. The maximum temperature
of the windings limits the allowable continuous current (and torque) of the motor. An important
conclusion is that for the mechanical power holds:
Pmech = E. I = K .ω . I . 3.6
T 4.1
U = I .R + E = R + K .ω
K
U 4.2
ω0 =
K
K2 4.3
S=
R
T = S .(ω 0 − ω ) 4.4
S U 4.6
Ts = =K
ω0 R
i = U/R , T s =K.U/R
Ts
U=constant
T
S=Ts / 7 0 7 0 = U/K
0,0 7 70
Figure 4.1, Torque speed curve
The torque-speed curve is a straight line (figure 4.1). The derivative of this line is the motor steepness,
what can be proven by dividing the stall torque and the no-load speed. The higher the steepness S, the
better the motor. The background is that at a fixed torque T the loss Pdiss decreases at increasing
steepness, because the next formula holds:
K 2 K 2I 2 T 2 4.7
S= = =
R RI 2 Pdiss
Ts 4.8
S= .
ωo
In data sheets one will find the damping factor Km, which is equal to √S.
THERMAL LIMITS
The dissipation of the motor is limited by the maximum allowed temperature θmax of the winding
isolation. With a thermal resistance Rth between the windings and the ambient, and the temperature
θamb one obtains the maximum allowable dissipation Pmax,diss as:
T2 4.10
Pdis = I 2 R = .
S
Combination of the two preceding formulas leads to a maximum allowable continuous torque (T100):
Servomotors do not operate in continuous duty. The peak torque, necessary during acceleration can be
much higher, because the motor accelerates only during a limited period of time. Suppose a motor
operates in a cycle of Tcycl; during δ.100% of this time the torque T is produced. The dissipated power
is given now as:
Tδ2 4.12
Pδ = .
S
2
T100 T2 4.13
Pdiss ,max = =δ δ .
S S
Tδ = δ .T100 4.14
This torque level can be drawn in the torque speed curve (figure 3) as a function of δ. One has to be
aware that Pdiss,max, and so also Tδ, depends on the ambient temperature!
10 %
U nom
T 25 %
50 %
100 %
0,0 7 70
Figure 4.2, Tδ by thermal limits
When the torque changes rapidly in time an other definition is used many times; the effective torque
Trms, which is defined as:
1 4.15
Trms = ∫
t c tc
T (t ) 2 dt
The time interval tc is the cycle time of the process involved. Then holds:
2
Trms 4.16
P=
S
T 4.17
Pmech = ω .T = T .(ω 0 − )=
S
dPmech 2T 4.18
= ω0 −
dT S
ω0S Ts 4.19
T= =
2 2
ω 02 S U2 4.20
Pmech ,max = =
4 4R
Ts
P mech P
Torque max
T
Ts
2 P mech, max
0,0 70 /2 Speed/power
MAXIMUM EFFICIENCY
The efficiency of the motor depends also on the operating point (given by Τ fr and T) of the motor:
with Tfr as the friction torque of the motor and d as the viscous damping coefficient (the speed-
dependent friction). The torque T for maximum efficiency is determined by:
torque
P out
T
ω0 /2 speed/power
Figure 4.4, Efficiency
In the next figure all the available curves are drawn in the torque speed plane; the line Pmax represents
the points were the maximum mechanical output power can be found at changing voltage; the curve
belonging to the maximum efficiency is represented by 0max.
Tmax
SOA Umax
Unom ωmax
0.5 Unom
T25
T100 η opt
speed
Figure 4.5, SOA
The last four lines enclose the so-called “Safe Operation Area”, the SOA.
As general advice for motor selection one has to operate always within the Safe Operation Area,
preferably operate between the Popt and ηopt line to combine an acceptable efficiency and a high output
power. Additionally one has to reconsider the motor chosen when the worst case operation point can
be found under the T-ω line belonging to 0.5 Unom (motor under loading). Within Philips one also
prevents to exceed momentary the T25-line; this is based on a questionable carefulness; the more one
knows concerning the application and motor one can shift this limitation upwards. Finally remains the
area with a dashed contour as preferred operation area.
di 5.1
u=E+L + i.R
dt
dω T 5.2-5.5
T = J. , E = K .ω , i = , and u = K .ω 0
dt K
ω 1 5.6
=
ω 0 τ eτ m s + τ m s + 1
2
with:
J .R J 5.7
τ e = L / R, τ m = =
K2 S
The electrical time constant τe is in general much smaller than τm. The conclusion is that one deals
with a second order system, what has to be considered as a complication for the design of the
controller.
current source
u i T 1 ω
K + _ sJ T load
voltage source
u 1 i T 1 ω
+ _ R(1+s) τe K + _ sJ T load
The solution to prevent that additional time constants are introduced is to apply a current source
amplifier, characterized by forcing a current, pre-described by its input, through the load, which is
here of course a motor.
Taking the effects mentioned above one has to take more affects into account to obtain the required
supply voltage for the motor-amplifier. One has to consider also the voltage drops over:
connectors
mains
+
E
_
• the cable and connectors between the amplifier and the motor
• the commutation system (brush resistance and voltage drop over the contact layer between
brushes and commutator)
• the voltage drop over the end-stage of the amplifier
• the voltage drop over the supply at high currents.
These additions also hold for the voltage of a current amplifier. In chapter 14 a more extended list of
amplifier choice related items will be given.
R
th1
rotor stator
P R
th2
Loss
C R C
th1 th0 th2
ambient
Figure 6.1.1, Thermal model
If one has to consider the time dependent thermal behaviour depends on the ratio between the smallest
thermal time constant and the cycle time of the load. The value for the thermal constant is between 10
seconds (<< 10 watt) and 1000 seconds (1 kW motors).
The rotor loss consists of the losses i2R, eddy current loss and friction loss in the bearings and
brushes. The next figure gives the interactions involved.
U + I Tem
1 + + - 1 ω
pL+R(θ r ) s - - - - pJ
-
E d Thyst Tfric TCog(Θ)
Θ
1
sign(ω)
p
θr Rth 1(ω) θs
2
I(θ ) .R(θ r)+ω2.d+
s Rth 2
|ω|.{T +T } Cth Cth 2
θambient
2
Trms 2
Trms Ramb .(1 + α.∆θ r ) Trms
2
Ramb .(1 + α.P.(Rth1 + Rth2 )) 6.2.1
P= R(θ ) = =
K amb .(1 + k t .∆θ s )
r
K(θ s ) 2 2 2 2
K amb .(1 + k t .Rth2 .P) 2
with P as the copper losses. Be aware of Ramb and Kamb, numbers belonging to the actual ambient
temperature, which might differ from the values Rref and Kref at the reference temperature used in data
sheets from the motor supplier. Rearrangement leads to:
Rth 2 + P.{K amb − αTrms Ramb .( Rth1 + Rth 2 )}− Trms 6.2.2
2
P 3 .K amb
2
k t Rth 2 + 2 P 2 k t K amb
2 2 2 2
Ramb = 0
The roots of this third order equation in the power loss P can be obtained in analytical form either by
numerical methods, with as result that a direct link exists between the torque Trms and the losses P.
The temperature rise of the rotor and stator follow by the multiplication of the solution P with Rth1+
Rth2 and Rth2 respectively.
The link between the ambient temperature θamb, the maximum allowed rotor temperature θr,max and the
maximum allowed continuous torque T100 can be obtained with the following approach, based on
θr,max=100 ˚C.
θref = 30 θrmax = 100 Rth1 = 0.5 Rth2 = 0.3 K := 1 R=6 kt := −0.002
Rθmax := R⋅⎡⎣ 1 + α ⋅(θrmax − θref) ⎤⎦ Rotor resistance at max rotor temperature
(θrmax − θamb)
P (θamb) := Allowed losses now
Rth1 + Rth2
θrmax − θamb
θs (θamb) := ⋅Rth2 + θamb Resulting stator temperature
( Rth1 + Rth2)
K (θamb) := K ⋅⎡⎣ 1 + kt ⋅(θs (θamb) − θref) ⎤⎦ Resulting motor constant
P (θamb)
T100(θamb) := K (θamb) ⋅ Resulting allowed torque
Rθmax
(
T100 θambi ) (
P θambi )
Ti := Pi :=
T100(θref) P (θref)
1
Ti Pi
1
0.5
0 0
0 50 100 0 50 100
θambi θambi
Figure 6.3.1, De-rating
Substitution of the function K(θamb) and P(θamb) in the function T100 in Fig. 6.3.1 leads to the analytical
expression as an alternative for the program given.
De-rating is not only a question of going to tropical area’s, mounting a motor in a warm housing has
the same consequences!
dθ s θ − θ s θ s − θ amb 6.4.2
= r −
dt Rth1 .C th 2 Rth 2 .C th 2
For the rotor and stator are used θr0 and θs0 respectively as initial values. Let us assume that the
motor load, a constant torque, is removed after 450 seconds. The problem is solved with Mathcad as
follows in fig. 6.4.1.
⎡⎢ ⎡ 1 + α ⋅( θ 0 − θref) ⎤ 1 ( θ0 − θ1) ⎥⎤
P ( t) ⋅⎢ ⎥⋅ −
⎢ ⎢ ⎡ 1 + kt⋅( θ 1 − θref) ⎤ 2 ⎥ Cth1 Cth1⋅Rth1 ⎥
D( t , θ ) := ⎢ ⎣⎣ ⎦ ⎦ ⎥
⎢ θ0 − θ1 ( θ1 − θamb) ⎥
⎢ − ⎥
⎣ Cth2⋅Rth1 Cth2⋅Rth2 ⎦
100
Θri
50
Θsi
0
0 200 400 600 800 1000
Ti
Θri 100
Θsi
50
20 0
0 200 400 600 800 1000
0 Ti 999
The mechanical commutation with brushes and a commutator is the main cause of performance
limitations for this type of motors. The maximum brush current limits the maximum torque, the
maximum speed is determined by the lift of the brushes, the maximum voltage by the limited lamellae
voltage and the mechanical output by the commutation power handling capacity of the commutation
system. Furthermore it is a fact that the limited reliability of motors with brushes is highly related to
mechanical troubles with the brushes.
The principal geometry of a servomotor with brushes is given by permanent magnets on the stator and
(heat generating) coils on the rotor. So the thermal resistance from the heat source towards the cool
ambient is high, because the heat has to be transferred over the air gap. Electronically commutation
removes the limitations from the brushes and the other geometry, as given in the next figure, leads to
a better heat transfer to the ambient.
Coil a
Stator
Magnet
Coil c Coil b
On the market ECM’s are offered and sold from mW’s up to MW’s. The higher costs of electronically
commutated motors (ECM) are many times accepted as price for an increased reliability.
The principal geometry of an ECM is given in the figure. A permanent magnet is mounted on the
rotor and the stator is provided with at least 3 coils.
The value of the individual coil currents is determined by the (measured) momentary rotor position
and by the required torque production.
Two types of brushless motors are available. At first there is the DC-brushless type, based on a six-
step controller. The second type is AC-synchronous, based on sinusoidally changing currents. In the
following paragraphs more details will be given.
7.1 DC-BRUSHLESS
The following figure represents the six-step controller in combination with an ECM. Some logical cir-
cuits control by means of six power transistors the current entering the coils.
In the simplest form the motor is provided with 3 coils, which are mounted in the stator under and
spatial angle of 120 degrees. To obtain a trapezoidal induced voltage as function of the rotor position
one applies a careful distribution of the winding sides of each coil over the stator slots in combination
with a fitted magnet shape and magnetization.
Z
S3 S2
-
Position Hall
Current PWM-logic Decoder
I ref
Figure 7.1.1, DC-brushless
The spatial distribution over 120 degrees leads also to an electrical phase shift of 120 degrees between
the 3 induced voltages in the coils (EMF), as indicated by the next figure.
EMK
0
Coil 3
0
Coil 2
0
Coil 1
=0 2%
Figure 7.1.2, Induced voltage
This figure reveals the 3 induced voltages as a function of the rotor position; at constant speed a
constant relation exists between position and the time.
The control circuit has as target to realize that each coil carries current during the intervals with a
constant induced voltage. The polarity of the current has to be related to the polarity of the induced
voltage and the amplitude to the value of the required torque.
This procedure leads to a constant torque production. To prove this we first consider a single coil,
assuming a constant speed ω. The K-factor of each coil is constant during each interval with a
constant induced voltage, because the following relation holds here:
E (θ ) 7.1
K=
ω
In the next figure the relations between the rotor position and the current, the induced voltage and the
transferred power Pmech are indicated. The currents are in phase with the induced voltages, block
shaped and have an alternating polarity. Adding the contributions of the 3 coils leads to a constant
torque:
Coil 3
Coil 2
EMK
Coil 1 Pmech
=0 2% Current
Figure 7.1.3, Induced voltage
So the value of the torque is determined by the value of the current. Furthermore a characteristic of
this type of control is that two coils are carrying current at the same instant of time. The figure also
reveals that at 6 positions within one revolution the current distribution over the coils has to be
changed. This is the background of the term six-step control.
1 1 1
1 2 3
3 2 3 2 3 2
1 1 1
4 5 6
3 2 3 2 3 2
Figure 7.1.4, Six step
In practice one finds a number of causes that leads to a less ideal behaviour:
• the position of the Hall sensors is not at the correct position. This leads to commutation at a
wrong position
• deviations between the three coils
• a non-trapezoidal shape of the EMF
• the amplitude of the currents are not equal, caused by inequalities in the amplifier end-stages
(offset, gain errors, drift).
The causes mentioned above lead to torque fluctuations near the commutation; this occurs 6 times per
revolution. Especially in position servo-systems one has to be aware of this phenomenon, because an
Suppose the motor is running at a constant speed. The mechanical power produced by each coil j is
given by:
2π 7.2.1
Pmech, j = E , (t ).i (t ) = Eˆ .I . sin 2 (ω .t + j )
3
Eˆ (:)ω . 7.2.3
So we may write:
T (:)Iˆ. 7.2.4
Fulfilling the condition that the same phase angle holds for the current and EMF of a coil leads to a
similar situation as for the DC-brushless motor; the torque production is determined by the current
amplitude.
- (Absolute) encoder
or resolver
PWM-logic
PI PI PI
Logic ω
- θ
- + Prom Generate 3 sinuses
-+
sin (θ ) + X
X
sin( θ + 12 0 )
X
sin(θ + 2 40 )
I Ref θ Re f
controller
Figure 7.2.3, AC-synchronous circuit layout
So the required bandwidth for the amplifier of an AC-synchronous motor is not determined by
the specified control loop bandwidth only but also by the required torque capability at high
speed.
Also the selection of the supply voltage of the amplifier has to be done carefully; one has to consider
the required rise time of the torque, the level of the torque, the speed and the commutation process. It
can be proven that one needs as amplifier specification at least:
2 7.2.6
2 K ⎛ dT ⎞
Uˆ ph − ph = + S .ω ⎟ + ( p.ω .τ .T )
2
⎜T + τ
3 S ⎝ dt ⎠
T K2 L ph − ph 7.2.8-
K= ,S = , τ=
I rms 1.5.R ph − ph R ph − ph 7.2.10
The lifetime of electronically commutated motors is determined by the bearings only. The torque
performance is given by the thermal limitations of the electrical isolation. In rare cases one also meets
a torque limitation to prevent demagnetisation of the magnets.
Torque ECM-limits
thermal limit brushes or
demagnetisation
T
T 25
0,0
n (speed)
Figure 7.3.1, Limitations for motors with and without brushes
The figure shows the limitation of a motor with brushes; added to this are the limitation for a
brushless motor, indicated by the dashed lines. The speed for this last type is limited by the bearings
and/or the heat production by eddy-currents in the rotor. The horizontal line is the torque limit related
with demagnetisation.
The hyperbola in the figure represents the maximum mechanical output of a motor with brushes.
(Going over this line holds Pmech=ω.T is constant). For the ECM represents the working point at the
maximum speed and torque the point of maximum mechanical output power.
In the electronically commutated motors one can find the heat producing windings at the stator. In
comparison with the mechanically commutated motors does this mean a considerably reduced thermal
resistance from the windings towards the cool environment.
The differences given leads to the conclusion that the absence of mechanical commutation leads
to a considerable higher mechanical output for the same motor volume.
For reasons of costs one finds in motors with brushes the classical magnetic materials as ferrites and
AlNiCo, whereas SmCo and NdFeB magnets, with at least four times higher ratio between magnetic
field production capability and volume, are applied in ECM’s. The cost of NdFeB drops dramatically
and so also motors with brushes and NdFeB magnets are offered nowadays on the market.
In general high power magnets as NdFeB or SmCo are in applied in the rotor of an ECM. Those
magnets generate high field densities in the stator iron. This is very attractive when one prefers a high
ratio between performance and volume. However this implies also iron-losses in the stator when the
motor rotates. In an experiment the stator temperature of an ECM is considered at 5000 rpm; a second
motor realizes this speed. The ECM was disconnected from its supply. Under these conditions only
eddy-current losses and hysteresis losses can be found in the stator. These losses resulted into a
temperature rise of 50 degrees!
This phenomenon can also be approach from another way, namely look to the required current to
obtain a certain value of the torque at increasing speed. The following figure reveals an increase of 50
%. The stator copper loss rises to more then 200 % and one has to deal also with the heating by iron-
losses. This coupled phenomenon limits the performance of an ECM!
1
0 500 1000 1500 2000 2500 3000 3500
Speed (rpm)
As indication the T10, T30, T50, T80, and the T100 curves are given for an ECM. The background is that
the allowable copper loss is reduced at increasing speed by the fact that the iron loss claims an
increasing part of the allowed stator-dissipation.
T10
T30
T50
T80
T100
Figure 7.4.2, T10, T30, T50, T80 and T100 for a motor
We can regard eddy current losses as a viscous friction; a good approach is to say that: Peddy = d.ω2,
so that the total losses are:
δ ( ω .d)
2 7.4.1
Pdiss + Peddy = (Tδ + ω .d )2 + (1 - δ ). + ω 2 .d ≤ Q max
S S
Qmax is the maximum allowed losses, based on the maximum allowable stator temperature. If we
employ a duty cycle δ, then Tδ is related to the duty cycle δ and the rotational speed ω as follows:
S 1 7.4.2
Tδ ( ω ,δ ) = .( Q max - ω 2 .d) - ω 2 .d 2 .(1 - ) - ω .d
δ δ
It is possible to combine an amplifier, intended for an EC-DC motor, with an AC-synchronous motor.
The background can be a required cost reduction or the availability of a motor with a certain
performance. This combination will suffer with a torque fluctuation of at least 13 %! This also holds
for the combination of an AC-synchronous amplifier and an EC-DC motor. The number of 13 % can
be proven by analysing the torque production of an EC-DC motor at a three-phase sinusoidal supply
system.
The toothed structure of the rotor gives rise to cogging, which are reduced by skewing the rotor slots
in the more expensive motors.
Strengths
• The power density, defined as Pmech/volume, is by far the highest, because the magnets only
have a small air gap to overcome. The rest of the flux path consists of the magnetically highly
conductive iron. The end result is less weight, less volume and low price.
• The iron guarantees mechanical robustness; the windings are anchored in it by means of a
moulding resin. Considerable attention has generally been given to secure fixing of the
winding head and to the connections to the collector. A high resistance to centrifugal forces
and accelerations is achieved in this way.
• The thermal capacity is great because of the presence of the rotor iron; short-time peak
powers are consequently readily absorbed. The small dimensions resulting from the high
power density do generally result in the thermal resistance Rth1 being slightly higher.
• A low mechanical time constant can be achieved through the slim structural shape of the rotor
in combination with the high power density; < 5 ms is feasible.
• The large number of producers means that “second sourcing” is possible at good prices.
Weaknesses
• The rotating iron produces cogging in addition to eddy current and hysteresis losses. A
reduction of cogging can be obtained by applying skewing of the slots; a full removal of the
cogging is rarely possible by tolerances on magnet dimensions, positioning, strength and
homogeneity.
• The fact that the conductors are located in the slots of the iron rotor gives rise to a substantial
self-inductance of the rotor windings. This makes the commutation of the rotor coils difficult.
Inside the cage there is a stationary magnet. The tolerances of the cage force suppliers to select a large
air gap.
Because of this structure the reluctance for the magnets is factors greater than in the iron armature
motor; the magnetic field must cross four air gaps and pass the cage twice. Achieving acceptable field
strength requires large magnets in these motors.
Strengths
• No iron losses and no cogging.
• A small moment of inertia J, because in this case the iron is not turning in the same direction.
• The lower self-inductance moves the commutation limit to higher values. The self-inductance
is lower because there is an air gap on both sides of the windings. In addition, the magnets
located on one of the two sides also behave like air. (The permeability µr of magnets is very
similar to air.)
Weaknesses
• The power density is a factor 3 lower than that of comparable iron armature motors, because
appreciably larger magnets must be used.
• Greater vulnerability to peak torques and high speeds. The cage construction has lower limits
than the iron armature rotor.
• In the absence of the rotor iron the thermal capacity is low. At the same time the thermal
resistance Rth1 is moderate because of the large air gaps and the impeded heat dissipation via
the shaft. This motor is therefore better suited to short-time loads. An increase of 3* in the
thermal load can be achieved by the introduction of a forced air current via a separate blower.
Strengths
• No cogging, no iron losses.
• A small moment of inertia and a high value of S, possible through the use of a lot of magnetic
material, lead to a low mechanical time constant.
• A very low self-inductance and electric time constant, because of the existence of an air gap
and magnetic material respectively on either side of the copper. The commutation forms
hardly any limitation in the motor performance.
• Because of the large rotor area the thermal resistance is low. The low thermal capacity of the
PCB version is partly offset by the high permissible temperature of 155ºC. The epoxy version
has a temperature limit of 110ºC, but does have greater thermal capacity.
• The better controllability of the dimensioning leads to a higher permissible shaft speed,
because imbalance and distortion come less into play; 8000 - 10,000 rpm occur. Extremely
robust.
• The flat structural shape makes combination with an encoder, brake or tacho for example
easier.
Weaknesses
• The low power density and low values for S/volume and S/mass are a consequence of the
voluminous stators. Partly because of this prices are high.
• Usually AlNiCo is used as magnetic material. It is a material with a high sensitivity to
demagnetisation. Peak currents must therefore be controlled, it being important to remember
that these motors have hardly any damping effect by way of a high self-inductance.
• To make re-magnetisation possible by the user, the manufacturer generally has the
magnetisation windings located around the magnets.
• Depending on the type of PWM amplifiers, the low self-inductance can prevent the amplifier
working properly. The low self-inductance means that there are sharp variations in current.
Even with a torque T = 0, this leads to unnecessary ohmic losses; it may be necessary to take
the step of using (voluminous) smoothing coils.
• The good control engineering properties mean that the disc motors are frequently used for
robots and in servo-controlled machines. But the latest developments in the field of brushless
motors (AC synchronous and induction motors) mean that this choice is no longer as self-
evident as it was.
Static load
If the servo loop has the task of driving a load at an almost constant torque and speed (a static load), it
is easy to show in Fig. 7.7.1, where the working point of the motor is. Verification of the maximum
permissible value of the torque and the speed is followed by a check on Trms, which must be less than
T100. And if the point Trms/nominal speed is also between the lines of Popt and ηopt, then we have a
suitable motor.
If this is not the case, a transmission can provide a solution. Let’s say that we have the working point
1 in Fig. 7.7.1. For the mechanical power we have Pout = ωload.Tload. A transmission (without losses)
with a transmission ratio i leads to Tmotor = i.Tload and ωmotor = ωload/i. The result is that, depending on
the value of i, a different point on the curve in Fig. 7.7.1 can be used.
P max
torque P opt
Tmax
SOA 1 Umax
ωmax
i
Pout =
constant
η opt
speed
With static loads we have complete freedom to choose such a transmission ratio that the efficiency for
example is at a maximum or it is possible for example to suffice with a(n) (available) supply voltage.
ω&load 7.7.1
T = J motor . ω&motor + i.J load .ω&load = J motor . + i.J load .ω&load
i
The transmission ratio i largely determines what torque the motor must deliver. Since the cost price of
a motor and the maximum torque are closely linked, a low torque is preferable.
The question now is what transmission ratio i must be chosen in order that a minimum torque T
suffices. Depending on the load, the curve of the motor torque as a function of the transmission ratio
can be a flat or a strong minimum (see Fig. 7.7.2). Differentiation of the above equation to ωmotor leads
to the condition that is known as “INERTIAL MATCH".
dT J .ω& 7.7.2
= - motor 2 load + J load .ω&load = 0
di i
J motor 7.7.3
i=
J load
The fact that there is a minimum can be explained as follows: if an extremely large transmission ratio
i is chosen, the term in equation 7.7.2 with the moment of inertia of the load will predominate and the
torque will grow proportionately to i. This leads to an initial conclusion: reduce i. If this were to go
too far, then the term in the formula with the motor moment of inertia would predominate. The reason
is that with a low i the angular velocity of the motor is high, so that a lot of energy has to be expended
on the acceleration/deceleration of the motor.
torque
SOA T(i)
i
T(i)
speed
Fig.7.7.2 The curve of the motor torque and speed as
a function of the transmission ratio i for two cases
The curve of the motor torque as a function of i is an hyperbola (: 1/i) at low values of i, which leads
to the conclusion: increase i. Result: somewhere there is an optimum.
The procedure to follow to see whether a particular motor is suitable for this dynamic load:
1) Determine the transmission ratio that goes with the "Inertial Match".
2) Process the moment of inertia of a suitable transmission (seen from the motor side) in the total
moment of inertia and over one or more strokes see whether the "Inertial Match" can be achieved
in combination with available transmissions.
3) After 2) are all the working points of the motor inside the SOA?
(maximum torque, maximum speed; this point is also called "Worst-case operation point").
4) Is Trms less than T100? Since T100 falls at elevated speeds with electronically commutated motors, a
safe design is obtained if the value of T100 at the maximum speed occurring in the design is
compared with Trms.
It is recommended that the curve of the motor torque be drawn as a function of the transmission ratio,
because depending on motor moment of inertia, mechanical time constant and required acceleration
for example there can be a flat area around the optimum.
Analysis of the open loop transfer function of a servo system shows that an advantage can be obtained
by the selection of a certain ratio between the moments of inertia of the motor and the load plus
transmission. Choosing equal ratio or simply deliberately a relatively light motor also depends on the
position of the sensors. The dissertation by Mr Groenhuis (TU Eindhoven) goes into this more deeply.
INTRODUCTION
For each servo system containing a brushless servo motor questions arise concerning the amplifier
voltage needed. This appendix describes the link between DC-based servo systems and AC
synchronous motors.
8.1 DC SYSTEMS
For DC voltage systems based on a classical servo motor with brushes there is a well-known formula,
d i(t) 8.1.1
u(t) = i(t).R + L + K.ω (t)
dt
with:
u terminal voltage [V]
i motor current [A]
R terminal resistance [Ω]
L terminal inductance [H]
K motor constant [N/A]
ω rotational speed [rad/s]
t time [s]
Taking into account the relation T(t) = K.i(t) for the torque, one can rearrange this expression as:
K d T(t) 8.1.2
u(t) = {T(t) +τ + ω (t).S}
S dt
with
S = K2/R (known as the steepness of the motor)
τ = L/R, the electrical time constant.
The momentary power needed is obtained by multiplying the previous equation by the current i:
The AC synchronous motor has at least three coils. There is a phase shift of 120 degrees in general
between the induced voltages of those coils. First, a single coil is considered for a motor with pole
pair p. The coupled flux, generated by the magnet, is given by:
dφ ph 8.2.2
T = i. = i.φˆph .p.cos(p.θ )
dθ
K ph = p.φˆph 8.2.3
The principle of AC synchronous motors is based on maintaining a coil current according to the
following expression:
Substitution of the current in the voltage equation of a single coil, on the assumption that the
inductance does not depend on the angle θ:
d L ph .i ph dφ ph 8.2.5
u ph = i ph .R ph + +
dt dt
diˆph 8.2.7
u ph ( θ , t) = {iˆph (t).R ph + L ph . + K ph .ω } . cos(pθ ) - L ph .iˆph .p.ω .sin(pθ )
dt
To obtain the amplitude of this voltage one differentiates the previous equation to pθ ; then one finds
the angle of pθ at which the maximum can be found. Subsequent substitution leads to:
8.2.8
d iˆph
Uˆ ph = (iˆph .R ph + L ph . + K ph .ω ) + (L ph .iˆph .p.ω )2
2
dt
Before we continue with this expression the definition of the steepness and the motor constant for AC
synchronous has to be given. The torque production of three coils together equals:
2
2.π 3 8.2.9
T = ∑ K ph .iˆph .cos(p.θ + j. ) 2 = K ph .iˆph
j=0 3 2
3 8.2.10
P = iˆph2 .R ph
2
2
3 K ph 8.2.11
S=
2 R ph
T 8.2.12
K=
i rms
The background of this definition is that one can get the motor constant simply by measuring the
torque and dividing the result by the rms value of a phase current. With the relation
1 8.2.13
irms = iˆph . 2
2
we arrive at:
2 2 8.2.12
S= K = K
3 R ph 1.5 R phph
with Rphph as the phase-phase resistance. The definition of the steepness S with this K-factor becomes:
3 8.2.14
K= 2 .K ph
2
2 8.2.15
2 K ⎛ dT ⎞
Uˆ phph = 3 Uˆ ph = ⎜ T + τ . +S.ω ⎟ + ( p.ω .τ .T )2
3 S ⎝ dt ⎠
Where a linear motor has to be analysed, this equation can be derived in the same way; after the
introduction of:
F
2 8.2.16
S=
P
and τm as pole pitch (distance N-N poles) and v as the speed, one arrives at:
2 2 8.2.17
2 K ⎛ dF ⎞ ⎛ 2.π .v.τ ⎞
Uˆ phph = ⎜ F + τ . +S.v ⎟ + ⎜⎜ F ⎟⎟
3 S ⎝ dt ⎠ ⎝ τm ⎠
To specify the amplifier one also needs to know the peak output power. Then the power of the three
phases has to be included. Let j indicate the phases with j = 0, 1 or 2. For the phase currents we now
have:
2.π .j 8.2.18
i ph , j ( θ ,t) = iˆj (t).cos(p.θ (t) + ))
3
d L ph i ph , j dφ ph , j 8.2.19
u ph , j = i ph , j .R ph + +
dt dt
2.π .j 8.2.20
φ ph , j ( θ , t) = φˆph , j .sin(p.θ (t) + )
3
8.2.21
2
T (t ) 2 τ .T(t ) dT (t )
P(t) = ∑ u ph , j (x, t).i ph , j (x, t) = + + ω (t).T(t)
j=0 S S dt
It is nice to recognise in this equation the similar terms as found for a DC motor given at the
beginning of this chapter. When for a linear movement in the x-direction a linear motor is applied the
equation becomes:
8.2.22
2
F (t ) 2 τ .F(t ) dF (t )
P(t) = ∑ u ph , j (x,t).i f , j (x,t) = + + v(t).F(t)
j=0 S S dt
vi Fi
1 500
0 0
0 0.05 0.1 0 0.05 0.1
ti ti
Ui Ii
200 10
0 0
0 0.05 0.1 0 0.05 0.1
ti ti
Pi
2000
0
0 0.05 0.1
ti
Figure 8.2.2, Example of the link between motion profile and amplifier requirements.
CONCLUSIONS
The result of this exercise is that the determination of the supply voltage and power needed for an AC
synchronous servo drive can be done by hand and the only input needed is:
• the motor constants K, S, τ and τm or p
• the motion profile related force or torque in the form of F(t) , dF(t)/dt or T(t) and dT(t)/dt
• and the speed (v(t) or ω(t))
An example is now given that demonstrates the use of the voltage equations for an AC synchronous
machine in combination with a third order motion profile.
For systems with a high demand on the accuracy one prefers the third order motion profile, which are
characterized with a finite value of the jerk. The background of this jerk limitation is that a jerk
energizes the mechanical vibrations in a motion system.
For the 3rd order speed profile holds that the acceleration pattern can be split in 9 equal time intervals
with a specific acceleration function; for the trapezoidal profile holds 3 equal time intervals.
When the stroke to be made equals S and one has Ts as available time then holds:
triangular 1 1 1 1
trapezoidal 1.13 0.75 0.92 0.85
3rd order 1.7 0.75 1.03 0.95
The practical meaning of the preceding table is, that a lower dissipation in a motor can be obtained by
applying a trapezoidal speed profile instead of a triangular profile, however then one has to take into
account that:
• high jerk values are present again
• the maximum acceleration in increased 13 %; this holds also for the amplifier current
The 3rd order speed profile prevents the high jerk, however then the acceleration, and so the motor
current, is increased with 70 % and the dissipation with 6 % in comparison with the first speed profile.
Despite the higher claims on the amplifier and motor this 3rd order motion profile is the nowadays
standard for advanced production equipment.
Remark:
• the relation between dissipation and the effective acceleration is quadratic
• a parabolic speed profile requires leads to a minimum motor dissipation. However this profile
is hardly used based on its high initial acceleration level of 12.S/Ts2 in combination with a
very high jerk at the start of the motion.
To translate the preceding table into voltages and power required from the motor amplifier one should
apply the related equations given earlier.
The voltage required is obtained with the worst-case operation point, given by the highest acceleration
in combination with a high speed. For a triangular and second order profile this is a unique instance of
time at the end of the acceleration interval. For a 3rd order profile one has to scan the time interval
with a decreasing acceleration and still rising speed.
The torque or force in the case of a triangular and second order profile is in general constant
between time intervals as indicated in the following figure.
Worst case OP
T1
Speed Torque
T2 Time
0,0
t1 t2 T3 t4 t5 t6 t7
tc
T3
1 9.1
t c ∫tc
Trms = T (t ) 2 dt ,
1 8 9.2
Trms =
tc
∑T
i −1
i
2
.ti
It prevents integration, however it can only be applied in the case of constant torque- (or force-
) levels during the time intervals.
The conclusions are that the motion profile applied influences significantly:
• peak current, voltage and power of the amplifier and
• the dissipation in the motor.
It is sometimes sufficient to use a simple circuit to control a motor, as shown in Fig. 10.1.1. At the
input of the operational amplifier (OpAmp) a reference for the motor current is compared with the
actual current. If there is any deviation in the current, the transistor is made more or less conductive
by changing the base voltage.
The diode parallel to the motor provides protection for the transistor, activating if the transistor blocks
so fast that the EMF and the voltage term L.di/dt together reach a value that is too high for the
transistor in question. A limitation of this circuit is that only a positive current i can be controlled; this
corresponds to a controlled torque in one direction. A torque in the other direction is not possible,
because a transistor only allows the current to flow in one direction.
+
+
I ref - U ce
I
-
Figure 10.1.1, Motor control
The solution is to make use of the self-inductance of the motor. If the transistor is driven via the base-
emitter voltage alternately within a return time T, fully non-conductive during Toff and fully
conductive during Ton, then little transistor dissipation will remain (see Fig.10.2.1).
The current im through the motor can be determined via simple first-order differential equations,
which show that the ratio Ton/T determines the mean value of the current. There are two approaches.
First, to keep Toff constant and to influence Ton via the controller; the cycle time is consequently a
function of the desired current; in this case we speak of frequency modulation. The second approach
is to keep the cycle time T constant, in which case the ratio Ton/Toff is varied. In this case we have
pulse width modulation.
Ub
U ce I
P=I * Uce
U ce
I
-
T Ub
Toff Ton
Figure 10.2.1, Pulse Width Modulation
Notes:
• There are clear similarities to a heating boiler.
• The cycle time T is generally much less than 1 ms.
The use of a motor in the four quadrants of the torque-speed plane is possible with the standard circuit
shown in Fig. 10.3.1. This circuit is called the “full H-bridge”.
i i
I ref Control L L L
-
i A B
Figure 10.3.1, H-bridge circuit Figure 10.3.2, Path of current with positive and
negative torque
Fig. 10.3.2 A gives the path for the current with a torque in one direction, Fig. 10.3.2 B for the other
direction. Here again one of the functions of the diodes is over voltage protection. The control of the
transistors must be such that two transistors can never be conductive in the same branch
simultaneously. Pulse width modulation and frequency modulation are the methods of control in this
circuit.
There are applications in which the motor returns energy to the power supply. In these circumstances
the motor is working as a generator and will return current to the power supply via two recovery
diodes. If no measures are taken, the voltage across the power supply capacitor C will rise sharply, the
power supply eventually suffering damage. This can happen in a crane for example.
This shortcoming can be solved by dissipating the energy released in a braking resistor Rbrake, as
shown in Fig. 10.3.3. This resistor is used as the occasion arises via transistor Q. Of course it would
be even better if the surplus energy went back to the mains. But the cost price of the additional
electronics required often seems to be a barrier to such a step.
+ T T T
Rbrake 7 7
i
I ref Control
C L 7
Q
T T
7 7
- i
Figure 10.3.3, H-bridge with energy dissipation Figure 10.3.4 The four
quadrants
The presence of self-inductance in the circuit is highly desirable, because it is a means of limiting the
ripple in the current. This is important because the dissipation in the copper of the rotor is
proportional to irms and the output torque to imean! See for example Fig. 10.3.5, where the torque T is
equal to zero. It is not uncommon to place an extra coil in series with the motor to reduce the current
fluctuations due to PWM.
0 0
time time
I rms = 1/3 I top
T T 10.3.1
1 2 1
T ∫0
irms = i dt imean = ∫ i dt
T0
A second reason is that the fluctuations in the current also produce flux variations in the iron and the
magnets of the motor. These variations induce eddy current losses, which can mean considerable
dissipation if the self-inductance is too low. (Even at standstill and with Tshaft = 0!). Fig. 10.3.6 shows
the measured current as a function of time for a moving coil motor in combination with a PWM
amplifier without an extra coil. It also shows what we can expect if the motor details given in the
product catalogue are used for calculating the current. The extreme difference is due to neglected
eddy currents in the magnets and the yoke!
Current with PWM
1
0.8
0.4
theoretical via L/R
0.2
0
1 74 147 220 293 366 439 512
-0.2
time in microsec.
Figure 10.3.6, Current in a moving coil motor without external inductance
(Measurements made available by M. van der Steen, MSD.)
Apart from an increase in the dissipation, we must also expect a decrease in the life of the
commutation system. A supplier of moving coil motors mentioned that a PWM frequency of more
than 100 kHz has to be used to prevent excessive wear, whereas 18 kHz as a usual PWM frequency.
The full PWM-amplifier supply voltage is connected to the motor terminal, even at no-load conditions
(then 50 % of the time in positive sense and 50 % in the negative sense). It is noticed once that this
continuous high voltage resulted into silver migration from a silver plated commutator towards the
shaft, leading to a short circuit to earth and finally a serious damage.
The ideal behaviour of an amplifier is got when the output voltage or output current follows the input
voltage perfectly. There are two effects that we must always consider carefully, that are the bandwidth
and the linearity around the zero-axis crossing.
I
+
R
I R C L
+
L 10.4.1
τ amp = Ramp .C amp , τe = , γ = R / Ramp
R
4 3
τ e = 0.013 τ amp = 6.283 10 Ramp = 1 10 γ = 0.01
I/Iamp I/Iamp
1.4 0
1.2 10
Amplitude
Amp Phase
Phase
i i
1 20
0.8 30
0 100 200 300 400 0 100 200 300 400
f f
i i
Frequency Frequency
To be concluded with the given numbers: more than 10 degrees lost below 200 Hz. Now look to I/E.
0.002
10
Amplitude
Amp Phase
Phase
i i
0.0015
20
0.001
4
5 10 30
0 100 200 300 400 0 100 200 300 400
f f
i i
Frequency Frequency
Conclusion is that the induced voltage causes a current via the output impedance; a rising frequency
of the induced voltage leads to a rising torque or force in the case of a coupled linear motor. The
translation to precision engineering design is that the torque (or force) level is influenced by vibration
of the stator. In other words, frame vibrations are transferred to the load of the motor via the finite
output impedance of the amplifier. (Sven Hol, ASML, thesis to be published)
Another consideration is that with electronically commutated motors the phase shift occurring makes
the generation of the torque less efficient at a rising speed (see Chapter 7).
The last remark related to the output impedance of the current source amplifier is, that its presence
leads to a rising sensitivity for the load impedance on the bandwidth of the transfer function motor
current over the amplifier set point.
Linearity is the second effect. Fig. 10.4.4 and Fig. 10.4.5 show the transfer of a 15 Ampere amplifier
near the zero-axis crossing for a sine of 10 Hz and an output current amplitude of 0.6 A. For a control
this phenomenon means that the transfer function around the zero-axis crossing is not well defined; as
frequency increases, this distortion becomes more pronounced. The cause of the distortion lies in the
imperfect behaviour of the power transistors of the H-bridge, which cannot be fully controlled by the
internal feedback within the amplifier. So, the linearity around the zero-crossing should be a
specification point.
PM3380A PM3380A
ch2: dT=-----s V2=- 148mV ch2: dY= 584mV
ch1
ch2
2
2
Figure 10.4.2, Input voltage and the current as Figure 10.4.3, Input voltage versus
function of time current
The gain of a motor amplifier is one of the terms in the open loop gain, so variations caused by e.g.
temperature dependency should be investigated in relation to the control stability. Offset in the case of
a single phase amplifier for a motor with brushes or an actuator are in general no problem, because the
I(ntegrator) action of the control will compensate this.
A serious attention has to be given to gain errors and offsets in the case of an electronically
commutating motor. In the case of a DC-brushless motor one will notice stepwise torque changes near
the commutating positions when those amplifier errors exists. For an AC-synchronous motor the more
complex consequences has to be described with equations.
dφ ph 10.5.1
T ph = i ph . = i ph .φˆph .p.cos(p.θ )
dθ
K ph = p.φˆph 10.5.2
Let us assume deviating relative gains, ∆Kamp(ph) and offsets Ioffset(ph) per phase.
2π . ph
i ph ( θ ,t , ph) = iˆph (t){1 + ∆K amp (ph)}.cos(p.θ (t) +
10.5.3
) + I offset (ph)
3
The torque production of three coils together equals:
2
2π . ph 10.5.4
T= ∑
K ph cos(p.θ + . 3
).i ph ( θ ,t , ph) =
ph=0
2
2π . ph 2π . ph
∑ K .cos(p.θ + 3 ).⎡⎢⎣iˆ (t){1 + ∆K amp (ph)}.cos(p.θ (t) +
⎤
ph ph ) + I offset (ph )⎥
ph = 0 3 ⎦
3 2
2π . ph 2
= K ph .iˆph + ∑ K ph ∆K amp (ph)iˆph (t).cos(p.θ + ) +
2 ph = 0 3
.cos(p.θ + 2π3.ph )
2
∑K
ph =0
ph I offset (ph )
The conclusion is that gain errors and offsets are introducing position dependency in the torque. The
frequency of the gain error related disturbance is twice the frequency of the offset related disturbance.
So one has to analyse carefully the consequences of these amplifier errors when precision engineering
drives are designed. As an initial value one can use 2 % gain error and 2 % of the maximum current
(or voltage) as offset errors.
In graphical form one gets now including the output impedance Z of the current source
amplifiers the following figure for a three phase system. By assuming a star point connection
Offset1(θ)
K1
T
ref Θ
K2
Offset2(θ)
Figure 10.5.1, The amplifier errors combined with a thermal model of a rotating
brushless motor.
A general representation of the amplifier, taking also into account the bandwidth of the
current source is as follows.
Amplifier Load
K(∆,θ)
Rm (∆,θr )
Lm
Offset(θ,t) K(∆,θ )ω
The frequency dependent behaviour of the current source and offset can be measured when
the current amplifier output is short-circuited. The measurement of the output impedance
should be done by adding a controllable voltage source in series with the amplifier output and
load. Fix the current amplifier input on zero, make a frequency scan with the controllable
voltage source and measure the current running through the source.
J.C. Compter, Electrical drives for precision engineering designs, 2007 56
10.6 CABLES
When designing a system with a PWM power supply the designer must design to a short distance
between the PWM power supply and the motor. The reason is that the wires to the motor are
connected alternately to the + and the – voltages of the power supply respectively with a high
frequency and steep edges. This causes electro-magnetic interference in the surrounding area, which
can for example affect sensor lines (the incremental encoder!) or prevent the installation from meeting
inspection standards (= approbation). The subsequent installation of shielding involves additional
expense through time lost as a result of the fault finding and the costs of re-cabling.
The minimisation of the electro-magnetic interference is just one reason; another is the electrical
resistance of the cable. For a 20-metre cable (out and back), with 30 wires of 0.2 mm section as
conductors, this amounts to approximately 0.7 Ohm (without taking in to account a certain
temperature rise). A servomotor with currents of 10 Ampere is no exception; this means a 7 Volt
voltage loss over the cable and a 70-Watt loss. For the power supply, usually a servo amplifier, this
means a higher specification in terms of voltage and power to be delivered! If we also think that a
power supply voltage below 50 V does not usually mean any special requirements for approbation,
the selection of a sufficiently thick cable can avoid many arguments.
With voltage loss in mind attention must also be paid to the resistance of thermal fuses and the
number of connectors in the cable between the amplifier and the motor. This latter point can involve a
conflict with the division of a drive system into separate modules for logistical or service reasons.
The cost of a servomotor is of course linked to its size. Also the available space in equipment is
forcing designers to apply the smallest motor possible in many cases. The application of a
transmission is one way as far as the backlash, additional friction, the stiffness reduction and the
higher motor speed can be accepted. The result is that the motor torque required is reduced and then
one can go for a smaller motor, because the motor volume is linked with the torque capability. Also in
Chapter 5 it is concluded that a motor should be used between 0.5 ω0 and ω0 to combine a high
efficiency and mechanical output power.
But one should also include the cost of the amplifier; the smaller the motor the more power is needed
to get the same mechanical output. So profit at the motor side is lost at the amplifier side.
Additionally holds that a smaller motor leads to a higher operating temperature, so more heat is
entering the equipment. The related drawbacks are potentially:
• shorter lifetime of the brushes and bearings
• shorter lifetime of the electrical insulation
• the rising heat leads to thermal expansion in the mechanics being a risk in high accuracy
systems
• the positive effects of a well tuned feed-forward is partly lost by the temperature dependent
motor constant K.
So a transmission should be investigated always, but additionally using a motor at its limits leads to a
rising development effort.
11.1 INTRODUCTION
Linear motors and actuators are the most suited drives for advanced machines with a high accuracy
and fast reaction time. An example is the lens actuator of a CD-player. This actuator type can be
considered as an extreme example of a mass product with a short reaction time in combination with a
high accuracy. The control loop bandwidth of this system goes up to 2 kHz.
Looking to data sheets of linear motors and actuators one gets the impression that the right choice for
an application can be made on the base of a small set of considerations. Target of the following
chapters is to teach potential users the meaning of the keywords in the linear motor and actuator
technology and to link their application aspects to the characteristics of linear motors and actuators
and to find a fitted amplifier.
The linear drive has got a clear position in the servo-technology for direct driven systems in the last
decade. The main characteristic of a direct driven system is the absence of a mechanical transmission
between the drive and the load. A mechanical transmission is for example a set of gear wheels or a
gearwheel in combination with a toothed bar. This last combination can be used to transfer a rotation
into a linear movement. Wear, play, a limited stiffness, hysteresis and friction are often performance
limiting factors related to mechanical transmissions and the direct drive is the answer.
Pneumatic and hydraulic solutions are preferred when force levels of several thousands of Newtons
are required and when the costs and volume of the compressor and pump respectively can be
accepted. The accuracy and reaction time of these systems can be brought to the same level as an
electric drive by the selection of fitted sensors and control valves. However this is outside the scope of
this treatise.
Linear electrical drives and actuators are an alternative. Force levels exceeding 1000 N, speeds
beyond 2 m/s and accuracies better than 1 micrometer are nearly on stock products. The combination
of all these numbers is not so easy, because then the full system behaviour starts to be important. The
behaviour of the bearing, sensors, mechanical dynamics, the control, power electronics starts to be
important for short time intervals and thermal effects might disturb the absolute accuracy.
Attention will be given here to linear motor and actuators for servo systems. The principal difference
between these two is that the stroke of a linear motor can be extended with limited consequences. E.g.
the extension of the magnet strip of a linear motor is all needed to get a longer stroke.
The definition of a servo system is not given until now. Within the context of electrical drives we
consider a drive as a servo system when the object to be moved has to follow accurately a changing
(electronic or software) reference. This reference might be linked to the acceleration, speed or
position. Control engineering learns that accuracy is gained when the control behaviour of the loop
components is predictable. Electric motors and actuators with permanent magnet are the ideal
components within this respect and the rising strength and falling costs of modern rare earth magnets
(Samarium-Cobalt and especially Neodymium-Iron-Boron) have given these drive components a very
competitive position.
The lack of a transmission means that the force and speed have to be the same for driving motor or
actuator and the load. The required force level determines highly the costs of the motor/actuator with
as consequences that one should not expect reduced motor costs in comparison to a solution with a
rotating motor and a rotation to linear transmission.
But direct drive means also one looses the acceleration torque needed for the transmission and an
example is known where the total costs of a spindle driven solution was comparable with a linear
direct drive.
Chapter 12 will give attention to actuators; chapter 13 deals with linear motors and chapter 14 gives a
critical view on the motor constants in general.
The most famous electrical actuator is the electro magnet of a doorbell. However one will meet this
rarely in servo systems because its main characteristic is a force that depends on the ratio of the
current and the air gap to the power of two. Control engineering rules on stability are valid only for
linear systems and this is the reason why actuators with a linear transfer function are preferred.
Examples of linear transfer functions are:
• the force linearly related to the current,
• a displacement linearly related to the voltage applied
• the speed linearly related to the voltage applied.
An additional preference is that the actuator should not have preferred positions as can be seen when a
supply voltage is removed. An example is the cogging torque present in fan motors and in toy motors.
Those torques are external disturbances in the eyes of a control engineer, which has to be counteracted
by an additional control action with at least a transient error during set point changes. Exceptions in
this respect are the electronic throttle of fuel engines and hoisting equipment, where safety has to be
guaranteed when the electric supply fails.
First attention will be given to the electro dynamic actuator, which can be found in loudspeakers, CD-
players and many servo systems.
The electro dynamic actuator is characterized by a current carrying coil in the field of a permanent
magnet where holds that the coil moves relatively with respect to the magnet. The design with a
moving coil is widely spread, because then one will find:
• the highest ratio possible between the force and the moving mass
• a good linear relation between force and the current and
• no preferred positions.
This description allows many forms, but only the rotational symmetric geometry will be discussed
here.
X
Figure 12.1.3, A coil in the field of a permanent magnet system
The time derivative of the coupled flux N .φ satisfies:
dφ ⎛ ∂φ d x ∂φ d i ⎞ ∂φ dx di 12.1.3
N. = N.⎜⎜ + ⎟⎟ = N. + L.
dt ⎝ ∂x dt ∂ i dt ⎠ ∂x dt dt
One should be aware that this last results only holds in the case that the inductance L does not depend
on the position x. The next step to come to the force is the multiplication of equation 12.1.1 with the
current i to get the power balance, with as result:
dφ ∂φ d x di 12.1.4
P = u.i = i 2 .R + i.N. = i 2 .R + i.N. + i.L.
dt ∂x dt dt
Rearrangement gives:
U U
current in the coil rises according i (t ) = t . At t=T holds I T = T .
L L
T
U2 2 1
The stored energy equals: E (T ) = ∫ U .i (t )dt = T = 2 L. I T2
0
2L
Mechanical engineering learns:
dx 12.1.6
Pmech = F .v = F.
dt
The combination of the equation 12.1.5 and 12.1.6 leads to the force:
∂φ 12.1.7
F = i.N.
∂x
The ratio between the force in [Newton] and the current in [Ampere] is called the force constant Kf ;
this ratio becomes:
F ∂φ 12.1.8
Kf = = N.
i ∂x
The voltage equation 12.1.1 can be written also as:
di ∂φ dx 12.1.9
u = i.R + L. + N. .
dt ∂x dt
This last term is well known as the EMF (the electro motoric “force”). The word motoric force is
based on the fact that a movement is required to get the voltage. It should be noted that the second
term is based on the change of the current level!
The EMF can be rewritten as:
∂φ dx ∂φ 12.1.10
EMF = N. = N.v.
∂x dt ∂x
The ratio between the EMF in [Volts] and the speed in [m/sec] is usually indicated as Ke with the
equation:
∂φ 12.1.11
K e = N.
∂x
Ke [V sec m-1] and Kf [N Amp–1] are clearly equal to each other with the given S.I. units and from now
on we will use for both the motor constant K.
Question: Prove with the preceding equations that the mechanical power is given by: Pmech=EMF.i .
Equation 12.1.11 gives implicitly a way to determine the motor constant K. Suppose one connects the
coil with an electronic integrator and starts moving the coil while measuring the position x, starting
from x0. The output of the integrator will be:
The derivative of Uint to x has to be obtained by post processing the data and now holds:
dU int (t ) dφ ( x ) 12.1.13
=N
dx dx
This means, that the motor constant can be determined without having the usual troubles with friction
in the case one uses a force sensor.
The temperature of the coil is always the performance-limiting factor in electro mechanics. The cause
of the temperature rise is the dissipation in the current carrying coil in relation to a finite thermal
conductivity for the heat flow to the ambient. This paragraph will spend attention to this performance
limitation.
The time dependent force F [N] in servo systems is clearly linked to the dissipation. For the current
holds by using the motor constant K [N/A]:
F (t )
i (t ) = . 12.2.1
K
With R [Ω] as the electrical resistance of the coil one gets as time averaged dissipation:
Tc Tc
1 1 R
∫
i (t ) 2 .Rdt = ∫ F (t ) .Rdt
2
P= 12.2.2
Tc 0 Tc K 2 0
The time Tc is the cycle time of the process to be controlled. To link the requirements of the
mechanics to the dissipation P we introduce the effective force Frms as:
T
1 c 2
Tc ∫0
Frms = F ( t ) dt , 12.2.3
It is also very helpful to introduce the steepness S=K2/R. The value of Frms is determined by the load
characteristics (mass, damping and a spring constant) and the movement to be realized in time
(position, speed and acceleration). The steepness S is a figure of merit for each actuator. The
combination of the introduced variables gives:
2
Frms
P= . 12.2.4
S
It should be clear by the preceding text that one always needs the peak force level and the effective
force Frms to determine if an actuator is fitted for a job.
Remark: in English data sheets one will meet many times the constant Km, which is equal to S and
called the specific damping. The link between damping and this constant is that a short-circuited
actuator gives S [Ns/m] as damping. The induced voltage EMF leads to a current I, which gives
finally with the motor constant K an opposing force according:
This relation between force and speed can be found also for viscous dampers. So a short-circuited
permanent magnet based actuator behaves like a viscous damper.
t t t
The consequence of displacements is of course a relative moment of the coil with respect to the yoke.
The motor constant K is in general a function of the relative position and this position dependency can
be reduced by extending the length of the coil or by extending the length of the gap in the axial
direction. The first solution leads to more dissipating copper without an increase of the force and the
second leads to a bigger yoke and magnet.
A position dependency of K of 10 to 20 % is hardly a risk for the control loop stability. However
when one applies a feed forward for e.g. the acceleration force one will over- or under compensate the
acceleration force leading to a rise of the servo errors.
The conclusion is that one has to specify the stroke and the allowable position dependency of the
motor constant K before ordering or designing an actuator.
The dissipation in the coil leads to a temperature rise, which is limited by:
• the maximum allowable temperature of the wire insulation class (90 tot 220 ºC),
• the temperature dependent mechanical stability of the synthetic materials used
• the decreasing strength of permanent magnets at rising temperature (0.1 .. 0.2 %/K for
modern rare earth magnets.
Additionally holds that the heat flow to the ambient might be restricted and that the copper resistivity
rises with 0.4%/K (so 60 degrees temperature rise means a rise of the coil resistance with 24 %). This
kind of considerations leads to a certain allowable temperature rise θmax for the coil.
The following figure represents the thermal model of an electro dynamic actuator, including the heat
capacity of the coil and the yoke. For the sake of simplicity we will only analyse the steady state
behaviour allowing us to forget the heat capacities. The assumption is also made that Re can
neglected, what is allowed when the coil is not fixed to a heat conducting body or frame.
2
i .R(θCu ) Re R th,2
C th,1 C th,2
θamb
Figure 12.5.1, Thermal model
The ambient temperature is indicated by θamb and the total thermal resistance between copper and
ambient by Rth [W/K]. The temperature of the copper follows with:
Rth,1 is the thermal resistance between the coil and the yoke, Rth,2 is the thermal resistance from the
yoke to the ambient. The first one, Rth,1, is determined by:
• the width of the air gap at both side of the coil
• the material and surface of the coil carrier (paper or aluminium)
• the winding technology applied (“wild” or orthocyclic)
• the speed of the actuator with respect to the yoke
• the thermal contact between the wires and the coil carrier
• the thickness of the insulation layer on the wires
• the altitude
The inner and outer surface of the coil and coil carrier are transferring the heat under the assumption
the Re can be neglected. When the total surface is given as A, we can introduce the specific heat
conductivity λ according:
1
λ= . 12.5.2
A. Rth
For a vertical plate in free air holds approximately λ=10 W/m2K; other shapes like a cylinder can go
up to λ=14 W/m2K. The presence of cool surface on a short distance will reduce the thermal
resistance significantly.
The position dependency of the motor constant K is discussed in the preceding paragraph. In this
paragraph is introduced the temperature dependency of the resistance. Both are influencing the
voltage and power consumption, as described in Chapter 6.
12.6 MECHANICS
Electrical aspects are linked now to the thermal aspects. Mechanical aspects are not requiring special
attention, because the material stress remains in general far below a dangerous level. The most critical
part is the coil, which can be made very robust by applying the orthocyclic winding technology in
combination with melting the wires to each other by means of a thin nylon layer on top of the wire
insulation. The only exception is that the shear stress on the glue layer between the coil and the coil
carrier might become a danger at a maximum coil temperature.
The electro dynamic actuators described are intended to linear movements. The application of a linear
air bearing system is an expensive solution (investment in air equipment, maintenance and energy
consumption) with as additional penalty the weight of the moving member of the air bearing. An
alternative is a leaf spring construction, but this starts to be bulky when the stroke exceeds some
millimetre.
Many times a solution is found by applying a pivot point, with the attractive option that the position of
the pivot point with respect to the actuator and load can be used to maximize the system performance.
Roller bearings are nice when they are rotating over big angles, but they are not fitted for small
repetitive rotations. A set of leaf spring acting as a pivot point is here the solution. Those components
are commercially available and this paragraph is directed to the bearing system
A weak spot of those leaf spring designs is a limited stiffness for forces, which gives in combination
with the mass eigen-frequencies and position errors. So, for a design as given in the next figure, one
should minimize the forces in the pivot point.
For wire bonding an ultra sonic transducer has to land on a silicon dye to fix a wire on the surface of
the dye. It is preferred to minimize the virtual mass as seen by the landing area to prevent any damage
during landing. Now we have two design objectives. A third objective, minimizing the peak force
required, will be added in the next paragraph.
J beam ,m
v
b
r
c
Figure 12.6.1, A tool.
gold
ball
crystal
Suppose that the transducer is fixed at the end of a homogenous beam with l as length and m as mass.
The question is to determine the position of the pivot point giving a minimum impact force and force
acting on the bearing.
Jbeam is the inertia of the beam in its centre of mass. The impact mass at the landing zone is given by:
J l 2 . 121 + r 2
mimpact = = m . 12.6.2
(r + l. 21 ) 2 (r + l. 21 ) 2
Lets us assume as stiffness of the landing zone c and assume as landing speed vb. The impact force
follows as the result of energy conservation:
l 2 . 121 + r 2 12.6.3
F = vb . mimpact .c = vb . m.c .
( r + l. 12 ) 2
Differentiation to r gives at r = 1/6 l a minimum impact mass mimpact = ¼ m!
The force acting on the bearing is the next question. The force F de-accelerates the beam:
F . 23 .l = ( J beam + m.( 16 .l ) 2 ).ϕ&
&. 12.6.4
Substitution of 12.6.5 and 12.6.6 in 12.6.7 proves F=Fm, so the landing force is fully spend to the de-
acceleration of the centre of mass. No force is left on the bearing and excitation of vibrations on the
bearing is prevented.
a
x1
x2
J work
m mb mb m work
act 2 1
Figure 12.6.3, Actuator, beam, pivot point and the load.
The derivation of the analytical equations involved is not too complex. The following issues should be
taken into account:
• the cross section of the beam is kept constant,
• the mass of the moving coil is in linear relation to the peak force required and
• the mass and inertia of the transducer is fixed.
The enclosed Mathcad program calculates as a function of r1 and r2 the impact mass, the peak force of
the actuator and the force on the bearings. The following graphs are the result with along the
horizontal axis r1 and along the vertical axis r2.
Figure 12.6.4 indicates as optimum r2=3 cm and r1=8.5 cm, figure 12.6.5 r2=4 cm and r1=8 cm and
finally figure 12.6.6 r2=3 cm and r1=9. It is nice to notice that the three objectives are not in conflict
with each other.
0.07
0.04
0.02 0.015
0.035
0.06 0.03
0.015
0.025
0.04 0.015
0.03 0.01
Mimpact
Figure 12.6.4, The impact mass in [kg] versus r1 (horizontal) and r2 (vertical) in [m].
0.05
6
0.04
0.03 8 7 6 5 7 8
Fact
Figure 12.6.5, The actuator force in [N] versus r1 (horizontal) and r2 (vertical) in [m].
0.07
0.06
0.02 0.015
0.01
0.005
0.05
0.04
FLZ
Figure 12.6.6, The bearing force in [N] versus r1 (horizontal) and r2 (vertical) in [m].
Fimp. r1
α ( r1 , r2 ) Angular acceleration beam
2
mtot( r1 , r2 ) . ( r1 x( r1 , r2 ) ) J ( r1 , r2 )
dzz( r1 , r2 ) α ( r1 , r2 ) . ( r1 x( r1 , r2 ) ) Z-acceleration mass centre
FL( r1 , r2 ) dzz( r1 , r2 ) . mtot( r1 , r2 ) Fimp Z-force on bearing
FLZi, j FL r1i , r2j
Several types of the electro dynamical actuators will be treated in the following chapter and the most
dominant differences will be indicated.
The human ears are very sensitive sound; a range of more than 100 dBA is covered over a frequency
range from 20 to 16.000 Hz. Distortions as clipping, crossover and inter-modulation are easily
detected. The electro-dynamic actuator satisfies high specifications without problems and is
considered as the best performing electro-mechanical actuator. A typical loudspeaker geometry is
given in figure 15. A ring shaped, axially magnetized, ferrite magnet is clamped between two iron
plates and an iron cylinder is placed in the inner bore. The iron concentrates the magnetic flux in the
gap between the upper plate and the cylinder and 1 Tesla is a usual field strength here.
The turns on the coil are directed tangentially and the axially directed force is generated in the gap
consequently. Only a part of the coil generates indeed the force. The axial length of the coil is related
to the required stroke and this clarifies why a raising stroke leads to a decreasing efficiency.
Current is brought to the coil by two Litze-wires, bundles of very thin wires to reduce their stiffness
and to enhance their lifetime. The coil itself is wound with orthocyclic winding technology to increase
the number of turns and to improve the heat transfer.
The coil carrier is made of paper or aluminum foil. This last choice is mainly based on heat transfer
improvement. One will get eddy currents in this foil, opposing the currents in the coil with a reduced
force as consequence when no precautions are made. The foil thickness is chosen as less than 0.1 mm
and a split in the axial direction prevents circular eddy currents.
For servo applications is a general rule to minimize the moving weight as much as possible to get the
highest possible bandwidth. So the dimensions of an actuator are nearly always a point of discussion.
One should give attention to reluctance forces in those cases where the peak force is only required for
a small part of the time, allowing a heavily overloaded actuator for a short time.
The theoretical background of the reluctance force is given first, starting with the voltage equation
for the coil:
dφ di.L ∂φ dx 12.7.1.3
u = R.i + N. = i.R + + N. pm
dt dt ∂x dt
Let us assume that the current is fixed in time. Multiplication with the current gives the power
balance:
dL.i ∂φ dx 12.7.1.4
u .i = i 2 .R + i + N.
dt ∂x dt
The first and third term at the right hand side are treated earlier as the dissipation and the power
going to the mechanics. For the second term holds under the assumed time independent current:
Comparing the equations 12.7.1.2 and 12.7.1.7 shows a linearly and quadratically dependency on the
current respectively. This explains why reluctance forces start to be important at high current levels.
F − = − c1 ( x ). I + c2 ( x ). I 2
F+ − F− F+ + F−
c1 ( x ) = c2 ( x ) =
I I2
So, by current reversal one can find if a reluctance force exists and if it has to be considered as
relevant.
An extreme improvement is reached in the last two decade as far as magnets are concerned. The most
popular magnet type around 1970 was ferrite, with 0.37 T as strength. Nowadays NdFeB is the
favorite material with 1.2 .. 1.4 Tesla as strength. More powerful or compacter designs can be made
now.
Typical examples of NdFeB based magnet systems are given in Figure 12.7.1.3. An important issue is
that the magnet fields outside the yoke are highly reduced compared to a configuration as given in
Figure 12.7.1.1.
The sintering process needed during the production of the magnets does not allow the aspect ratio
found in the right hand figure, so those rings has to be assembled by e.g. 16 segments. The related
costs give a preference to the left solution. The advantage of the right solution at the other hand is that
the force constant depends less on radial displacements.
It is hard to get a long stroke with a loudspeaker actuator. A good option is to apply the geometry
given in Figure 12.7.2.1 The coil is moving here also. The magnets are extended in direction of
motion and here a force is generated over the whole length of the coil. At the other hand only 50 % of
the circumference of the coil is in a magnet field. A field strength of 0.5 Tesla can be expected in this
geometry when NdFeB-magnets are applied.
The linear guiding system requires attention, because an air bearing adds a significant mass and a
roller bearing will introduce lifetime questions and disturbances.
The self-inductance of this actuator is far higher than for a loudspeaker actuator and has to be taken
into account when the amplifier voltage has to be determined. A reluctance force exists when the self-
inductance depends on the position as mentioned in the preceding paragraph. This position
dependency is present here significantly. The designer of this actuator can reduce the position
dependency by the addition of small air gaps in the vertically drawn areas.
One of the weak points of this actuator is the assembly to the load. The mounting bracket can be fixed
with an epoxy compound to one coil side, but this is surely influencing the dynamical behavior in
negative sense.
Applications are e.g. driving the sledge of a CD-player and driving a sub-micron measurement
equipment.
Actuators with a limited building height are required in wafer steppers. Thermal conditions in wafer
steppers are very tight and volume/weight are interfering always with other features. On top of this
holds that the system performance is determined by the position accuracy and the throughput,
expressed as wafer per hour. For the design of these actuators one has to judge the dynamical,
thermal, and control behavior, the power consumption, weight and volume.
A typical configuration is given in the following figure.
The coils are mounted in a stainless steel housing provided with a water-cooling system. Stainless
steel plates are glued on top of the coils to reduce the thermal resistance from the coil to the water
channels.
The stroke of this actuator is limited to some millimeters. Accuracies better than 5 nanometer within
milliseconds after a force pulse of 500 N are the targets. This can be achieved only with a very well
predictable behavior of the system and as far as the actuator is concerned, one has to find a good
combination of magnet and coil dimensions.
Mounting this actuator well in the system is much easier even when force vector has to go through the
center of mass.
The mass of the magnet yoke is twice the mass of the coil part, so it seems the most appropriate
solution to fix the coil part to the load. However this interferes with the nanometer accuracy required,
because the noisy water flow and rather stiff conductors of the coils will act as a significant
disturbance for the position control.
A general point of attention to investigate the positioning of the two actuator parts with respect to
each other in relation to stacked mechanical tolerances in a drive system. Misalignment will introduce
force in another direction than intended as indicated in Figure 12.7.3.2.
An interesting actuator for a short stroke (<2 cm) at a force level of 100 is given in Figure 20.
Magnet
+ current
- current
Figure 12.7.4.1, Short stroke moving magnet actuator.
The two coils in the inner bore are electrically connected in series in opposite direction. The two iron
disks on the magnet concentrate all the flux of the magnet to guide it radially through the copper to
the stator iron. The enclosed flux of each coil is changing when the magnet is moving, so a force is
created when current is going through the copper.
The simplicity of the production and robustness of this actuator is the major attractive side.
The ideal behavior of an actuator is a position independent force in combination with a linear relation
between the generated force and the current. There are four issues here to be discussed, a cogging and
a reluctance force and damping.
Cogging is present when the actuator has preferred positions when no current is in the coils. Here is
the mid position indeed a preferred position for the magnet and this effect becomes more dominant
with a shorter iron tube. This is the reason why the iron tube is extended in comparison to the coils.
Reluctance is present when the inductance is position dependent and the moving iron disks give
indeed a position dependency. The motor constant K [N/A] depends on the position.
Damping is present when a force is opposing when the magnet is moving with a certain speed, even
when the coils are disconnected. The explanation that damping exists here is based on the position
dependent flux in the stator iron. The flux going through in a thin ring of the stator near the iron disk
is changing when the magnet moves. The derivation of the (estimated) damping goes as follows:
R 0.02 . m Radius of the magnet
Amagn π . R2 Surface of the magnet
Bmagn 0.5 . tesla Assumed flux density in the magnet
Φ Amagn. Bmagn Flux
4 . 10 . m
3
d Thickness of the disks
26 . 10 . m
3
Ri Inner radius cylinder
29 . 10
3.
Ro m Outer radius cylinder
Φ
dΦdx The change of the flux
d
EMF ( v ) dΦdx. v The induced voltage
ρ 88 . 10 9. ohm . m The specific resistivity of iron
ρ . π . ( Ro Ri) The resistance of a ring with the axial length d
R
d . ( Ro Ri)
EMF ( v ) 2 The dissipated power as a function of the spee
P ( v) of the translator
R
P ( v)
damping( v ) 2 . The damping found for both sides
v2
damping( 1 . m . sec ) = 38.945 newton . sec . m
1 1
Suppose that the shaft is made of iron. The left hand side is then a magnetic north pole and the other
side a magnetic south pole. Handling this actuator starts to be difficult and not every bearing type is
fitted for operation in a strong magnetic field. So stainless steel shafts are preferred.
This problem can be prevented by changing the concept as given in Figure 12.7.4.2. The production
of the shaft is an easy job compared to the previous design but it is very likely that the magnets has to
be ordered as segments of e.g. 45 degrees.
Iron
Magnet
+ current
- current
Figure 12.7.4.3 gives a patented moving magnet actuator for a wire bonder. The operation principle is
proven by the position dependent permanent flux in the coil. Both stator parts should be made of
laminated SiFe to reduce eddy currents and of course both sides can be made of E-shaped SiFe
provided with a coil.
A general remark for all linear actuators with a moving magnet and iron at the fixed world: high
bearing forces when the alignment is not done well.
12.8 SUMMARY
Linear short stroke actuators are treated as components to be used as part of a drive system and the
link between the system requirements and specification of the drive is made. Several actuator types
are analyzed and attention is given to the strong and weak aspects.
Two principles can found nowadays. The first consists of a magnet strip and an armature consisting of
an iron core and coils and the second one has two magnet strips with and ironless coil block between
the magnets strips.
Magnets
Magnetstrip
Figure 13.1, the iron core motor
Magnet yoke
Colis
Magnets
Magnetstrip
Figure 13.1.1, The iron core linear motor
The layout of an electronically commutating linear motor with an iron core is given in Figure 13.1.1.
One part is provided with permanent magnets and the armature is provided with coils. The current
values in the coils, usually a 3-phase system, are link to the armature position with respect to the
magnets strip and the current amplitude is linked to the required force level.
13.2 DC-BRUSHLESS
Figure 13.2.1 gives the basic circuit for a DC-brushless linear motor. Some digital circuits are
controlling the transistors linked with the coils. The motor is provided with sets of three coils. Within
each set are the individual coils spatially shifted that the EMF’s are shifted 120 degrees wit respect to
each other. Very specific is that the EMF’s behave trapezoidally as a function of the position.
+ EMF
0
Coil block Coil 1
0 Coil 2
-
S1 S3
Coil 3
Current PWM-logic Position Hall
Decoder 0
θ =0 2π
I v, s
ref
The target of the logic is to put current in the motor coils only when the EMF has reached its constant
value. The direction of the current is linked to the polarity of the EMF. As indicated earlier holds for
the power transfer from the electrical to the mechanical part that this power is equal to EMF.i. Based
on the previous philosophy holds that two coils are carrying a current, leading to:
P mech = F .v = 2.EMF.i 13.2.1
Figure 13.2.2 gives the tranferred power per coil and summing the power leads to the same transferred
power level at each instant of time. Assuming a constant speed leads now to a constant force level.
θ =0 2π
Figure 13.2.2, Mechanical power for a DC-brushless motor
The force level is determined by the current amplitude and its direction is determined by the current
polarity.
One will notice that the ideal behaviour is disturbed by the following causes:
1. deviations related to the Hall sensors (offset, drift and positioning) lead to current switching at
a deviating position
2. the induced EMF is not flat over 120 degrees
3. unequal phase currents by offset and/or gain differences.
The result is that one will notice at each commutation position a disturbance in the force. Positioning
in the neighbourhood of the commutation positions might give rise to a limit cycling phenomenon.
The application of this DC-brushless principle is limited to ironless linear motors, because the
switching of currents leads to a disturbing noise level in iron armature motors.
The second group of brushless motors is based on a sinusoidally changing EMF. To get this sinusoidal
behaviour one adapts the magnet dimensions and/or coils shapes and the shape of the teeth in the case
of an iron armature.
Figure 13.3.1 shows how the current has to behave as a function of the position in relation to the EMF
values. The next figure shows the mechanical power produced. The underlying equations are the same
as valid for a rotating electronically commutated motor, which are described earlier.
1 1
The three The three
phase EMF's as a
I1i currents as a E1i function of
function of the the position
I2i 0 E2i
position in 0 in rad.
I3i rad.
E3i
1 1
0 2 4 6 0 2 4 6
θi θi
0
0 2 4 6
θi
Explanation:
Stroke: the costs of the magnet strips start to be dominant at a long stroke.
Volume: iron core motors are more compact by using the magnets more efficient. In the case of
water-cooling holds, that this can be done more efficient in iron core motors.
Cogging: the interaction between the teeth of the iron core and the magnets leads to preferred
positions which are detectable easily at i=0. A usual level is 2 to 5 % of the continuous allowed
force level. For a control loop is this noticeable as a disturbance, which can be counteracted by a
(learning) feed forward. Always one should specify the amplitude of the cogging and its spatial
frequency.
Load on the bearings: for an iron core motor holds as load on the bearing 2 to 5 times the peak
force of the motor, whereas the bearing load of an iron core can neglected usually. The
consequences are that the bearing moving mass will be significantly higher for an iron core motor.
In cases where a air bearing is used the presence of a high pretension is attractive.
Accuracy: at standstill the controller, the amplifier and the sensor determine the accuracy and
resolution for both motor types. At constant speed are cogging and the high bearing loading the
causes of a limited accuracy for the iron core motors.
Noise level: the teeth structure of the iron core motor moving along the magnet strip gives highly
changing attraction and shear forces with noise as consequence.
Frms/Fpeak: this ratio is low for ironless motors and the application (mechanics and the motion
profile to be realized) might give rise to a preferred motor type.
Ftop/moving mass: the iron of an iron core motor is the issue involved. The mass of the load
determines to what extent this has to be considered as a selection criterium.
The market moves clearly to ironless linear motors, because the costs of the widely used magnet
material, NdFeB, is dropped with a factor of ten in the last ten year. So, the efficient use of the magnet
material in an iron core motor became less important.
The previous linear motor concepts are based on moving copper and magnets fixed to the world.
Figure 13.5.1 gives a concept based on moving magnets and the coils fixed to the world. This motor
was the heart of a successful lithographic machine for many years with 5 nano-meter accuracy.
The two moving iron yoke are fixed to each other and provided with SmCo magnets. The coils are
wound around and fixed to a bar of SiFe sheets, glued well to a nearly solid body.
IJzer
Magneet (4*)
+ stroom
_ stroom
Figure 13.5.1, Linear motor with moving magnets and a long stroke
The direction of the currents in the stator coils is linked to the required force direction and their
strength determines the force value, as seen before in the AC-synchronous and DC-brushless linear
motor. The coils are connected in two series connections, consisting the coils 1,3, 5 and 2,4,6
respectively. Two amplifiers are needed to feed the motor. Of course one has to apply also here a
position sensor for the electronic commutation.
The bar is made of the SiFe sheets to reduce eddy currents damping and magnetic hystersis friction to
improve the servo behaviour. It is an advantage to use the double-sided construction to reduce the
bearing load.
The series connection means, that all coils are carrying current, so losses can be found over the total
length of the motor. This is the payback for preventing moving motor cables.
The relevant constants will be treated in the following paragraphs with having in mind linear motors.
The following issues are involved with respect to the motor constant K:
• Current level; the load requires a certain torque or force level and a changing motor constant
K means a changing current level.
• Voltage level; in the voltage equation of a motor one can find the EMF, linearly related to
the speed via K and the current, for a given force (or torque) linear related to 1/K. It depends
on the motor, load and motion profile whether one can find a rising or decreasing voltage in
combination with a decreasing motor constant K.
• Power losses; the losses in the motor are linearly related to square of the current, so P(:)1/K2
• Control loop behavior; the motor constant is one of the components of the open loop gain.
Another open loop gain changes always the phase and gain margin.
• Feed forward; a deviating motor gain leads to wrong feed forward value, so K-factor
deviations leads to a lower efficiency of the feed forward.
So consequences can be found in the motor, the amplifier and the control loop behaviour.
• Magnetic saturation
o The field of the coils in an iron core motor can magnetically saturate the iron of the
armature. The result is that the flux density caused by the magnets decreases with
finally a decreasing motor constant at a rising current level. So for iron core motors
one should ask a supplier the behaviour of the motor constant as a function of the
current level for a certain value of the air gap.
The items given above are based on the experience built up by designing linear drives for precision
equipment. It is overdone to check all these aspects for a linear transport system!
It should be noted that the constant K gives together with the current the force or torque produced in
the motor; the shaft torque is got from this number by subtracting the internal motor damping,
friction, cogging and that part of the force or torque needed for the acceleration. Additionally has to
be mentioned, that amplifier gain errors and offsets influences the force or torque of electronically
commutated motors, as described earlier.
The steepness S equals K2/R. The preceding paragraphs show the factors influencing K and R, so it
should be clear, that the steepness is subject to much variation.
Data sheets are mentioning always the thermal resistances, but the lack of standards for servomotors
allows many test conditions. Questions to be answered by the supplier are at least:
• Is the motor mounted to a frame; what material is used and what are the dimensions?
• What is the orientation? Is air moving freely around the motor?
• What is the relation between the speed and the thermal resistance?
• The thermal resistance is a function of the temperature of the heated body and the ambient
temperature by the changing ratio between convection and radiation. Consequently one
should know the temperatures involved.
• What is the reference altitude? A de-rating of the allowed loss of 5%/km altitude should be
applied.
Suppliers have two interests, getting the best position in comparison to competitors and to prevent
customer disappointments. So open the discussion with a supplier as soon as fitted to ensure that your
own conditions are in line with the test conditions.
14.5 LIFE-TIME
The lifetime of electric motors are dominantly determined by the bearings and, in the case of a motor
with brushes, by the latter. A lifetime indication can be listed in data-sheets; however, most likely this
number is based on the motor running continuously at a constant speed. That is rarely the case in
servo-systems and it remains always the responsibility of the designer to verify the lifetime.
As far as the bearings are concerned we have to consider:
• radial and axial load
• temperature
• acceleration (noticed: slipping roller cage!)
• ambient conditions (e.g. glass particles)
• repetitive point-to-point movements and grease concentration at the end of the stroke
• PWM supply induced capacitive currents running through the bearings.
The life of brushes are influenced by
• peak currents
• repetitive point to point movements
• current pulses by PWM
• peak voltage of the amplifier
• temperature
Based on the contents of the preceding chapters we can summarize all relevant items related to the
specification of an amplifier:
• the cable and connectors between the amplifier and the motor
• the temperature dependent K-factor and resistance R
• the commutation system (brush resistance and voltage drop over the contact layer between
brushes and commutator)
• the voltage drop over the end-stage of the amplifier
• the voltage drop over the supply at high currents.
• voltage fluctuations of the mains
• a margin on the peak current and voltage for control purposes
• tolerances on the motor constant and resistance given by the supplier
• uncertainties concerning the load data
• changing ambient conditions (ambient temperature or altitude)
• damping by eddy currents in the motor friction related to the bearings and brushes
• a position or current level depending motor “constant”
• bandwidth
• output impedance
• noise level
• PWM induced motor losses and life time reduction
• offset, drift and phase gain errors in the case of 3-phase amplifiers
• non linear behavior around zero current for PWM-amplifier.
This in addition to costs, volume, standardization, interfacing, safety, serviceability, remote sensing,
communication protocols, mains and radio interference, cooling, electro-magnetic compatibility,
approbation, life time, stability, supply voltage, supply voltage disturbance rejection, monitoring, etc.