Go

Digital Ebook | A Design World Resource

High-Performance Brushless Servo Systems

Inside
Comparison of DC servomotors and brushless servomotors 2
Sinusoidal and trapezoidal brushless servos 4
Sinusoidal EMF and sinusoidal current 6
Sinusoidal control for brushless and induction motors 7
Closed-loop control for high performance 9
This updated handbook provides a technical overview
of the components, theory and interaction of brushless
motion control systems.

AND MORE!
Sponsored by

Page 2
Figure 1:

Comparison of DC servomotors
and brushless servomotors

Block diagram of a typical motion control

system

Courtesy ElectroCraft
The traditional permanent magnet DC servomotor has

been the industry workhorse for many decades in high-performance

servo drive applications. The primary reason for this is that the DC
servomotor is very easy to control using adjustable DC voltage. A brief
review of the operating principle of the DC motor illustrates this point:
See Figure 2.
A fixed magnetic field created by permanent magnets in the stator
interacts with the armature current flowing in the rotor winding.
Interaction of the current-carrying conductors in the magnetic field
produces a rotor torque. This torque is at its maximum value when the
magnetic field vector is perpendicular to the current vector, but is zero if
the angle between the vectors is zero. The magnitude of the torque is:

A simple representation of a typical motion control system is shown in

Figure 1. Each of the primary components will be reviewed.

Figure 2:

Principle and
construction
of a DC
Servomotor

 Torque = K B I sin
 Where K is a constant, B is the magnetic flux density, I is the armature current, and
two vectors.

is the angle between the

Therefore, as the rotor motion reduces the torque angle to a zero value,
no further motion results. To eliminate this condition, the DC motor
incorporates a commutator on the rotor, which routes the current flow in
the armature windings as the rotor rotates. In other words, the current is
progressively reversed as the windings connected to the commutator bars
pass beneath the brushes. In a servomotor, the physical location of the
brushes is such that the current vector is maintained perpendicular to the
fixed magnetic field for any direction of rotor rotation or for any rotor
speed. This results in torque generation proportional to armature current
and motor speed proportional to armature voltage. The classic equations
that describe the DC servomotor are then:
Torque = Kt I
Eg = Back EMF Voltage = Ke n
Where Kt is the torque constant, Ke is the voltage constant,
and n is the motor speed.

While the control of a DC servomotor is straightforward, the primary

limitations are due to the mechanical commutator. Some of these
limitations include brush replacement, brush run-in after replacement,
brush radio frequency interference (RFI) and voltage/current limitations.
The construction of the DC servomotor also requires the commutator
to rotate. This means the armature windings must rotate as well which
results in high rotor inertia and a poor thermal situation because the heat
from losses is primarily generated in the rotor. Of course, for any type
of motor, the heat-producing losses must be minimized and effectively
dissipated so that the internal temperature remains below maximum
limits. Replacing the mechanical commutator with an electronic one can
eliminate all of these DC motor limitations.
The electronically commutated motor or brushless servomotor consists
of a permanent magnet rotor, a stator with usually three phases and a
rotor position sensor. (See Figure 3.)
Due to the construction technique of a brushless servomotor (insideout compared to the DC servomotor) the motor losses are almost entirely

Sponsored by

Page 3

continued

Comparison of DC servomotors and

brushless servomotors
in the stator, resulting in a short thermal path to the ambient allowing
more input power into the windings. Passing air over the motor frame
can further increase heat transfer to the ambient. By eliminating the DC
motors mechanical commutator and armature winding on the rotor,
the brushless motor design results in lower rotor inertias, higher rotor
speeds, and higher motor supply voltages compared to the conventional
DC servomotor. The polyphase stator winding for a brushless servomotor
lends itself to automatic winding processes.
The permanent-magnet materials used for typical rotor construction
are the rare earths (samarium cobalt and neodymium iron boron) or
ceramic (ferrite). Typical demagnetization plots of the second quadrant
B-H curves for these materials are shown in Figure 4. B is the flux density
and H is the magnetizing force. The ferrite magnet has inferior magnetic
properties, yet is low cost and readily available.
The samarium cobalt material (SmCo5) has excellent magnetic
properties, can operate at higher temperature, but is expensive.
The rare-earth material neodymium iron boron has many impressive
magnetic properties. At the same time, however, there are limitations.
Neodymium iron boron magnet materials have less-than-optimal
thermal properties and this must be taken into account in the motor
design. Two forms of Neodymium iron boron magnets are shown in
Figure 4. The highest performance is achieved with the fully dense
sintered material. The bonded neodymium can be formed into more
complex shapes, simplifying manufacturing processes and lowering costs
but at the expense of performance. Rare-earth permanent magnets are
used for high-performance brushless servomotors that have the lowest
rotor inertias and the smallest overall motor size/weight for a given
torque rating.
From an application perspective, the torque-speed curves of a
typical DC servo system versus a typical brushless servo system are
shown in Figure 5. Note the ability of brushless servo systems to
operate at higher speeds with higher peak torques than conventional
DC servo systems. These advantages of electronic commutation versus
mechanical commutation result in superior performance and reduced
maintenance.

Figure 3:

Principle and construction of a brushless servomotor

Figure 4:

B-H curves for permanentmagnet materials

Figure 5:

Torque-speed curves

Sponsored by

Page 4

Sinusoidal and trapezoidal

brushless servos
Courtesy ElectroCraft

Trapezoidal EMF and

sinusoidal EMF and square
wave current
Courtesy ElectroCraft

The basic structure of the DC servomotor and drive amplifier is

relatively standard throughout the world. However, the same cannot
be said about brushless servomotors and drive amplifiers. Some of the
common brushless technologies and terminology are shown in Figure 6.

Figure 6:

Brushless drive technology and terminology

Trapezoidal EMF and square wave current

This type of brushless servo technology was developed first because
of its analogy to traditional DC motors. The theory of operation is
illustrated in Figure 7. As the rotor turns, the current is electronically
commutated from one pair of windings to another. In this way, only the
flattop portions of the back EMF are active, and a composite DC voltage
is created proportional to motor speed. The current amplitude, which is
proportional to developed torque, is usually controlled by pulse width
modulating the active transistors.
In practice, achieving the trapezoidal back EMF waveform imposes
difficult motor design constraints. For this reason and because of the
evolution of brushless servo technology (which we will descibe in
following sections) this type of brushless servo drive is rarely used today.
ElectroCraft uses various neodymium iron boron magnets (rare earth magnets) in the ElectroCraft RP, RPP, and EXC
families of brushless servomotors. The newer RP and RPP Series motor families were designed to exhibit superior
velocity and torque performance, while maintaining a small footprint and competitive prices.

Sponsored by

Page 5

continued
Sinusoidal EMF and square wave current
Today most brushless servomotors actually have sinusoidal back
or frequency is increased. Phase advancing the commutation angle
EMF. Because of the simplicity and low cost of the square-wave current
can improve this situation at the expense of additional complexity.
control, it is common to have square-wave current servo amplifiers
In summary, excitation of a sinusoidal EMF brushless motor with
operating sinusoidal back EMF brushless motors. This is commonly
a square wave current drive is practical and can result in acceptable
referred to as a Brushless DC servo system. The theory of operation is
performance for many applications. This very basic control principle
shown in Figure 7.
is relatively simple to accomplish in a servo drive. This approach is
Notice that with ideal back EMF and current waveforms, the motor
still common and results in a lower cost solution, which requires
torque constant Kt has a peak-to-peak divided by peak ripple value
less hardware and, in many cases, is simpler to install than more
equal to 13.4%. The average Kt
sophisticated brushless servo
value is about 10% higher than
technology. The ElectroCraft
Figure 7:
the Kt for sinusoidal currents. In
CompletePower series of
Waveforms for two EMF styles and square wave
closedloop velocity servos the Kt
brushless servo drives and speed
brushless servo drives
ripple is not necessarily a problem
controls utilize this control
since a high velocity loop gain
methodology. The Completeproduces a very uniform rotation.
Power Series of drives can be
However, the torque constant ripple
a cost effective and simple
does increase motor heating, because
solution to many motion control
current is modulated by the ripple
applications.
function to produce the uniform
velocity. In fact, the continuous
torque rating is decreased by about
Figure 8:
5% as compared to sinusoidal
Inductance effect
with square wave current
current excitation.
Another subject that has not
been discussed yet is motor winding
inductance. The coils of wire that
make up the motor windings
have resistance and inductance.
From Faradays law, current in
an inductor cannot be changed
instantaneously. In practice, the
square wave currents suffer from
the inductance effect as shown
in Figure 8. This torque loss
progressively worsens as speed

Sponsored by

Page 6

Sinusoidal EMF and sinusoidal current

Courtesy ElectroCraft
Sinusoidal brushless servo drives provide the best rotational
uniformity at any speed or torque. Compared to previous technologies,
the sinusoidal servo drive uses a more complex control algorithm but the
motor, feedback and power electronics remain the same. In recent years,
high-performance microcontrollers and Digital Signal Processors (DSPs)
have been developed that can handle complex calculations, increasing

Figure 9:

Sinusoidal three-phase principle of operation

the capability of the sinusoidal brushless servo system. At the same time
component costs have decreased, making this the most popular servo
technology today.
The sinusoidal back EMF motor excited with three phase sinusoidal
currents (in the proper relationship to the back EMF at every rotor
position) produces a constant torque. An explanation of this phenomenon
at a steadystate speed and torque is illustrated in Figure 9.
The three-phase sinusoidal quantities which are displaced
spatially by 120 represent the magnetic field and current. The
three-phase quantities produce a resultant vector with constant
amplitude that rotates at the sinusoidal frequency. The rotor position
sensor tracks the back EMF position and allows the current vector
command to be generated perpendicular to the magneticfield vector
at any instand. A pulse width modulated (PWM) current amplifier
is necessary to ensure the ability to control the current amplitude,
frequency and phase with sufficient dynamic performance. At
this point, the fixed magnitude magnetic field vector, which is
perpendicular to the adjustable magnitude current vector, is analogous
to the operation of a DC motor.
There are many brushless servo drives and brushless positioning
drives that are designed to produce sinusoidal currents with
incremental encoder feedback. This type of servo system combines
the optimum motor design with a sinusoidal current PWM drive
to produce the best low-speed and high-speed performance. The
ElectroCraft CompletePower Plus digital servo drive takes this
control capability one step further. True sinusoidal current control
can be accomplished in these drives without the need for highresolution incremental encoder feedback. Only the lower resolution
commutation feedback is required. In some applications, this reduces
the need for and cost of the additional encoder.
Other benefits that arise from the microprocessor-based drive design
include the ability to operate induction motors using field orientated
control and the ability to share the motor mounted encoder with the
position controller. This maximizes flexibility.

Sponsored by

Page 7

Sinusoidal control for permanent magnet brushless

motors and induction motors
Courtesy ElectroCraft
The mechanical commutator of the DC servomotor ensures that

the armature current vector is kept perpendicular to the permanent magnet

field at all speeds and torques. This provides for control of torque by simply
adjusting the armature current level and makes using the DC servomotor
very straightforward. Using input commands analogous to those for
a DC servomotor, a universal control computes the torque producing
sinusoidal currents for permanent-magnet brushless motors. (Otherwise, for
induction motors, it computes the torque and field-producing sinusoidal
currents.) This universal control strategy, known as field-oriented or vector
control, is shown in Figure 10. Field-oriented control ensures that the
torque-producing current vector is perpendicular to the field vector at any
torque or speed. Some form of this control is used with all sinusoidal back
EMF and sinusoidal current brushless servo drives. From the servo users
perspective, the torque, velocity, and position control is then analogous to
the traditional DC servomotor.

Also shown in Figure 10 is an optional phase advance angle. The

phase advance angle can be used to optimize the amplifier/motor
performance characteristics. The most common use of the phase
advance angle is to compensate for the inductance effect that causes a
torque reduction as speed (frequency) increases. A less common use is to
allow permanent magnet brushless motors to operate to a higher speed
than would normally be possible without the phase advance angle. If the
phase advance angle is used, the magnitude of the angle as a function
of torque and/or speed is normally determined by the servo drive
manufacturer and is not user adjustable.

Figure 10:

Field-oriented control for induction motors

and permanent-magnet brushless motors

Sponsored by

Page 8

Incremental encoders and resolvers

Courtesy ElectroCraft
The most commonly used motor-mounted position feedback

devices for sinusoidal brushless servos are either incremental optical

encoders or brushless resolvers. The primary advantages of the resolver
are that the position information is absolute and it is robust because it
is similar in construction to the motor. However, other factors favor the
encoder over the resolver, including lower overall cost, digital feedback,
higher resolution and accuracy, and easy line count flexibility (binary or
decimal). The advancement of magnetic and capacitance-based incremental
encoder technology further enhances the position of encoder feedback
design. However, the use of both encoder and resolver feedback will likely
continue into the foreseeable future as both solutions service the need of
specific applications and operating environments.
The resolver used in brushless servo drives is illustrated in Figure 11.

Figure 11:

Resolver feedback for brushless motors

The high-frequency excitation signal is transferred to the rotor via a circular

transformer. The raw resolver feedback is a high-frequency AC signal
modulated by the sine and cosine of the rotor angle. The raw resolver
feedback is not very useful, so some form of external circuitry is required to
create usable information. In brushless servo drive applications, the resolver
feedback is usually processed by commercial resolver-to-digital-converters
that add significant cost. The output of the resolver to digital converter is an
absolute digital position word and analog velocity.
The incremental encoder used in brushless servo drives is illustrated in
Figure 12. The raw encoder feedback, already in digital format, is typically
processed with low-cost commercial circuitry to produce a digital position
word. The position information interfaces directly to the microcontroller of
digital drives. If the velocity loop is analog, then an additional circuit processes

Figure 12:

Incremental encoder feedback for brushless

motors

Sponsored by

Page 9

continued

Incremental encoders and resolvers

the encoder feedback to produce an analog tachometer signal. Notice
that encoder feedback signals are differential, for high noise immunity
and for locating the encoder at long distances from the drive.
Permanent-magnet servomotors also use a low-resolution
absolute signal in addition to the incremental encoder. This
is used to locate the magnetic field vector at startup. The lowresolution absolute signal may be built into the incremental
encoder (usually called commutation signals) or it may be
provided from a separate Hall-effect or commutation encoder.
In some cases, these commutation signals are incorporated into
the incremental encoder and are provided by a single device,
thus eliminating the separate commutation feedback device and
reducing the cost and complexity of the servomotor design.
Induction motors do not require an additional low-resolution
absolute encoder because they do not use permanent magnets to
establish the magnetic field vector.
Closed-loop control for high performance motion control,
the most common structure for high performance motion
controllers, is illustrated in Figure 13. This cascade control
structure has an innermost current loop, a velocity loop around
the current loop, and a position loop around the velocity loop.
The sequence of current (torque), velocity and position is natural
as it matches the structure of the process to be controlled.
This multi-loop control structure functions properly only if
the bandwidths of the loops have the proper relationship.
Bandwidth is the measure of how well the controlled
quantity tracks and responds to the command signal. Figure
13 shows a closed loop frequency response and definitions for
the two most common bandwidths -3 dB bandwidth and
-45 phase shift bandwidth. The current loop must have the
highest bandwidth, then the velocity loop; finally, the position
loop has the lowest bandwidth. Therefore, tuning control-loop
regulators is accomplished by starting with the innermost loop
and working outward.

Figure 13:

Closed-loop control for high-performance motion

Sponsored by

Page 10

Current regulation
Courtesy ElectroCraft
The current control for three-phase brushless servomotors is usually

performed with a PWM power amplifier and closed-loop control of the

current in each phase. A block diagram of the current loop and power
amplifier is shown in Figure 14.
The power devices must be able to withstand high voltages, switch high
currents, and exhibit low conduction and switching losses. Traditionally,
the bipolar transistor and power field effect transistors (FETs) have been
the most common output devices for high performance servo systems.
However, these switches are being replaced with insolated-gate, biploar
transistors (IGBTs) and intelligent power modules as these devices have
lower losses and can operate at higher power levels. These devices combine
the rugged output of a bipolar transistor with the gate drive and fast turnoff times of a power FET. The PWM frequency of modern servo drives is
typically between 5 and 20 kHz. The high PWM frequency allows for a
high current loop gain and keeps the current ripple frequency and audible
noise to a low level.
The current feedback sensor is critical and must provide an exact
representation of the actual current. The current feedback signal is
compared to the current command to generate a current error signal.
The current regulator processes current error to create a motor voltage
command. The voltage command signal is compared to a triangle wave to
create the PWM signal that commands the power devices to turn on and
off at the proper time. There is additional circuitry that provides lockout to
ensure that the upper and lower devices are never on at the same time, even
during turn-off and turn-on transitions. Too much lockout time results in
excessive deadband in the current loop while too little lockout time results
in short-circuit or shoot-through current flowing through upper and
lower devices. The PWM technique results in the most efficient conversion
of DC to variable AC power.
Because the current controller tuning is very important for proper drive
performance, most drive manufacturers do not allow users to perform
this adjustment. In the past, to eliminate the need for current controller
adjustments by the drive user, specific amplifier model numbers were
matched with specific motor model numbers, or plug-in personality
modules matched an amplifier with a motor. The next generation of
ElectroCraft CompletePower Plus brushless servo motors and drives will

Figure 14:

Three-phase current loop for a PWM power amplifier

self-detect the motor and drive combination and automatically determine

the correct current control gain settings, eliminating the possibility of
incorrect setting. The end result of a properly tuned current controller is an
actual current that follows the commanded current with -3 dB bandwidth,
commonly in excess of 1 kHz. To the outer velocity loop, a properly tuned
current controller can be approximated by a fixed gain and first order lag
such that the low frequency characteristics (up to frequencies of concern to
the velocity loop) approximate the low frequency characteristics of the more
complicated actual transfer function.
One limitation of the current loops is gain, phase and offset errors
that occur due to imperfect sensors and other circuitry. These errors are
one source of torque ripple so it is important to keep them to an absolute
minimum. Another limitation of the properly designed current controller
is insufficient voltage to generate the necessary current. This situation
occurs for large value current commands at higher motor speeds when the
back EMF of the motor begins to approach the motor supply voltage. In
applications requiring high torque at high speeds, careful observation of the
motor/drive system peak torque envelope is required.

Sponsored by

Page 11

Velocity regulation and tuning guidelines

Courtesy ElectroCraft

The most common velocity controller structure is

the proportional plus integral (PI) regulator. A block diagram of

a simple velocity loop appears in Figure 15. The choices of the
P gain and I gain for the desired response are based upon the
application requirements. Proportional gain is always used with
higher bandwidths resulting from higher P gain values. The integral
gain provides stiffness to load torque disturbances and reduces the
steady-state velocity error to a zero value. However, integral gain
does add phase shift to the velocity controller closed-loop frequency
response and can lengthen settling time. Therefore, a low or zerovalue integral gain is sometimes used with very high bandwidth
position controllers in point-to-point positioning applications, while
a significant integral gain value is used with contouring applications
to provide high stiffness.

In practice, the velocity controller tuning is rarely determined solely by

calculation. Often tuning is done manually through trial-and-error, with
the motor connected to the actual load. This tuning is simplified though
the use of sophisticated set-up software and a PC connected to the servo
drive. The ElectroCraft CompletePower Set-Up Software Utility provides
the user a manual tuning mode that allows a small step velocity command
to be applied to the drive while the motor is attached to the actual load.
Within the software utility, real-time adjustment of the velocity loop gains
can then be made while observing oscilloscope waveforms on the PC to
optimize the velocity loop tuning for the application. A typical example of
velocity responses to the step changes in velocity command and load torque
for a poorly tuned velocity loop and a well-tuned velocity loop are shown
in Figure 16. Tuning should be performed with small signal responses
which means that the current stays away from the current limit at all times.

Figure 15:

Figure 16:

Velocity loop block diagram

Velocity responses to poorly tuned and welltuned velocity loops

Sponsored by

Page 12

Position regulator and tuning guidelines

Courtesy ElectroCraft
Position control applications typically fall into two basic categories:
contouring and point-to-point. Contouring applications require that
the actual position follow the commanded position in a very predictable
manner with high stiffness to reject external torque disturbances. Notice
that predictability is required, but this does not necessarily mean that
position error must be zero at all times.

Figure 17:

Position controller block diagram

The other type of position control point-to-point positioning is

typically defined by move time, settling time and velocity profile, not paths.
Independent of the positioning application, the form of a simple
position controller is shown in Figure 17. The velocity controller has been
approximated by a unity gain and a first-order lag with a time constant
equal to the velocity loop -45 phase shift bandwidth.
A position controller with only proportional gain K is very common
(particularly for contouring applications), and the position loop response
can be easily calculated for a certain crossover frequency or gain. The openloop frequency response for the proportional position controller is shown
in Figure 18. The crossover frequency is related to the common method of
expressing gain:
2
 .65 Hz = 16.66 rad/sec =
1 inch/min./mil = velocity/position error

Figure 18:

Position controller response and Kp gain

Where 1 inch/min./mil also equals 1 meter/min./mm. In most

position controllers, the Kp gain is related to position loop gain as shown
in Figure 18. A general position loop regulator is more complicated
than the simple proportional gain-only type. The general position loop
regulator is Proportional-Integral-Derivative with feed-forward (velocity
and acceleration) and is illustrated in Figure 19. The following is a brief
explanation of the purpose for the five adjustment terms.
The proportional gain Kp is the most important term and generates a
velocity command proportional to position error. In other words, if just
Kp gain is present, motion is only possible if a position error exists. In fact,
higher velocities result in proportionally higher position following error.
Only increasing Kp gain can reduce the position following error. However,
because the velocity loop has a given -45 phase-shift bandwidth, Kp
gains that push position-loop bandwidth above about 1/3 of this velocity
loop -45 phase-shift bandwidth cause actual position to overshoot the
commanded position which is usually undesirable.
The feed-forward gain Kff generates a velocity command signal
proportional to the derivative of the position command. Therefore, if there
is no change in the position command, then the feed-forward command

Sponsored by

Page 13

continued

Position regulator and tuning guidelines

is always zero. Ideally, 100% feed-forward provides the exact velocity
command without the need for any position error. In practice, actual
systems including loads are not ideal, so a more conservative setting of feedforward gain is often taken (100% or less) because too much feed-forward
gain causes actual position to overshoot the commanded position. Because
the feed-forward command is generated open loop, there is no effect on
the position loop stability. The function of the feed-forward command is
to significantly reduce the constant velocity following error even though
the Kp gain is maintained at a proper level for stability. Figure 20 shows an
example of the feed-forward velocity command effect on position following
error when making a trapezoidal velocity profile move.
The derivative gain Kd as shown in Figure 19 creates a command signal
that is proportional to the derivative of actual position feedback. The
derivative term is used in two different situations. One situation occurs
when the normal velocity servo is replaced by a torque (current) servo,
which requires that the derivative term be used to provide the necessary
damping in the form of a velocity feedback signal. The other situation
occurs with a normal velocity servo to reduce overshooting of the position if
a higher-than-normal Kp gain is necessary.
The next gain term is the acceleration feed forward gain, Fgain, which
scales the second derivative of position command. This term allows position
following error to be reduced when the velocity is changing. The acceleration
feed-forward term is useful in combination with the velocity feed-forward
term when trying to maintain a low following error at all times which can
be useful in tracking applications or if a fast settling time is required.
The last term is the integral gain, Ki, which can provide a velocity
command signal to reduce static position errors to zero. Integral gain
provides stiffness against torque disturbances and friction torques and is
usually handled in the velocity controller. Therefore, integral gain in the
position controller is normally avoided except for special situations. Because
integral gain causes overshooting, some position controllers with integral
gain allow the integrator to only be active during certain conditions
such as when the position command is not changing and when the actual
position is very close to the commanded position.

Figure 19:

General structure of position loop

Figure 20:

Effect of velocity feedforward on position error profile

Sponsored by

Page 14

E-Mail:
E-Mail:

[email protected]

Web:
Web:

www.electrocraft.com

Sponsored by