PID Controller: DC motor

Krutika Gite
VJTI Mumbai, India

Abstract— The analysis of a DC machine (DC motor) is

I. INTRODUCTION
Direct current (DC) motors have been widely used in
adjustable speed drives or variable torque controls because
their torque is easy to control and their range of speed control Fig. 1.
is wide. However, DC motors have a major drawback in
which they need mechanical devices such as commutators
and brushes for a continuous rotation. These mechanical
parts require regular maintenance and hinder high-speed op-
erations. Recently due to a major development in alternating
current (AC) motor control technology, power electronics
technology, and microprocessors, AC motors have been used
instead of DC motors, but DC motors are still used in
lower power ratings below several kilowatts.The DC motors
can be mainly classified into two different categories: a
separately excited DC motor and a self-excited DC motor.
In a separately excited DC motor, the armature winding and Fig. 2.
field winding each have separate DC power supplies, while
in a self-excited DC motor, the two windings share one DC
power supply follows the reference input as accurately as possible, is the
An electric motor is an electromechanical device that most important factor. The control system is also needed to
converts electrical energy into mechanical torque. As a have the disturbance rejection capability while tracking the
torque-producing device, the electric motor needs to have reference input. This means that no disturbance can affect
its torque controlled as a priority. When the torque produced the system tracking the reference input as much. There are
by the motor is applied to a mechanical load, the torque will two basic types of control: open-loop control and closed-
change the speed m of the driven load and, in turn, change its loop control, also known as nonfeedback and feedback,
position m.Therefore, in motor drives, the most effective way respectively.
for controlling the position or speed of the load connected
to the motor is to control the torque of the motor directly. II. DC M ACHINE M ODELING
However, the torque control is expensive due to the high cost
Figure 3 shows the equivalent electric circuit of a DC
of the torque sensor for motor torque feedback.We can see
motor. From this circuit we obtain the system of equations
that the torque is directly proportional to the current under
defined later by using the Kirchhoff law of mesh currents
a constant field flux. This tells us that we can control the
for the armature and field circuits and Newton’s law of
torque by using the current instead of the torque itself. Thus
summation of torques for the rotor of the machine
the current control is important for the speed or position
control of a motor. d
V (t) = Ra i(t) + L i(t) + w(t)i(t)k0 (1)
The aim of control is to adjust the control input so that dt
the state or output of the system reaches the required goal.
d
In the motor drive systems, the state or output may be the i2 (t)k0 = TL (t) + bw(t) + J w(t) (2)
current, flux, speed, position, or torque, while the control dt
input is only the input voltage. For control performance, d
the ability of reference tracking, i.e., the system output V (t) = (Ra + R f ) ∗ i(t) + L f i f (t) + E(a) (3)
dt
 Fig. 5.
Fig. 3. Caption

Fig. 4.

Fig. 6.
d
Te (t) = TL (t) + b ∗ w(t) + j w(t) (4)
dt
V. P ROPORTIONAL CONTROLLER (P CONTROLLER )
III. O PEN - LOOP CONTROL ( OR NONFEEDBACK CONTROL )
In this type of controller, which is illustrated in Fig.5, the
Open-loop control is a type of control that determines the controller output is proportional to the present error
control input Us of a system by using the system model Gss e(t) = r(t) − y(t) (5)
as shown in Fig 3. A drawback of the open-loop control
is that it requires perfect knowledge of the system. Since as given by
the open-loop control does not use the feedback of the U(t) = K pe(t) (6)
system output Ys, the error between the reference Rs and
An advantage of the proportional control is its immediate
the output Ys cannot be reflected in determining its control
response to errors. The larger the proportional gain is, the
input Us. Therefore its output may be different from the
faster the response is. Thus it should be noted that the pro-
desired goal due to the change in system parameters or
portional gain determines the control bandwidth. However,
disturbances. Because of its simplicity and low cost, the
there is a limit to increasing the proportional gain because
open-loop control is often used in systems where it has
a large gain can lead to system instability or may require
the well-defined relationship between input and output and
unrealizable control input. A drawback of the proportional
are not influenced by disturbances. However, this open-loop
control is that the error can never be zero. In other words,
control is inadequate for the high-performance motor drive
there exists a steady-state error in the P control. This is
systems, which are easily affected by parameter variations
because when the error comes to zero, the output of the
and disturbances such as the back-EMF and load torque.
P controller becomes zero. Thus the system will have an
error again. We can evaluate the steady-state error of the P
IV. C LOSED - LOOP CONTROL ( OR FEEDBACK CONTROL ) controller from its close-loop transfer function of the block
diagram in Fig.5 as,
Closed-loop control is a type of control that adjusts the
R(s) G0 b(s) K pGp(s)
control input U(s) by the feedback of the output Y(s), as = =
0
Y (s) 1 + G b(s) 1 + K pGp(s)
shown in Fig.4. The closed-loop control is generally used
to control the position, speed, current, or flux in the motor (7)
drive systems. The controller Gc(s) adjusts its control input
U(s) to reduce the error between the output and the desired
goal. Thus the role of the controller Gc(s)in the closedloop
control is very critical to the system’s performance. There are VI. I NTEGRAL CONTROLLER (I CONTROLLER )
several types of widely used types of the feedback controller In this type of controller, the controller output is the
in the motor drive systems as follows. integral of the error as shown in Fig.6, so the past errors
 Fig. 2.29 shows the frequency response of the PI controller.
The PI controller has an infinite DC gain (i.e., gain when 5
0) due to an integrator. The transfer function of Eq. (2.25)
can be rewritten as Gpis 5 Kp 1 1 1 Tpis 5 Kp s 1 1
Tpi s where Tpi 5 Kp=Ki is the integral time constant of
the PI controller and 1=Tpi is called the corner frequency
of the PI controller. From Eq. (2.26), we can see that the
PI controller can add one pole at s 5 0 and one zero at
s 5 2 Ki=Kp to the transfer function of the system. The
location of this zero has a strong influence on the system
Fig. 7. performance, which will be described in detail later. If the
zero is correctly placed, then the damping of the system
can be improved. As mentioned earlier, the pole at s 5 0 of
the PI controller causes the system type to be increased by
adding an integrator 1=s to the transfer function, improving
the steady-state error. However, the increased system type
may make the system unstable when gains are large. In
general, the PI controller can improve the steady-state error
but at the expense of stability. In addition, the PI controller
can reduce the maximum overshoot and improve the gain and
phase margins as well as the resonant peak, but it reduces the
bandwidth and extends the rise time. The PI controller also
Fig. 8. has a characteristic of a low-pass filter that reduces the noise.
Now, let us evaluate the performance of the PI controller for
current control of a DC motor. The transfer function of a DC
as well as the present error have an effect on the control motor system that includes a PI current controller as shown in
input as given by Fig. 2.30 is given as Is 5 Kpcs 1 Kic Las2 1 Ra 1 Kpcs 1 Kic
Z
Kp
Z I s 2 s Las2 1 Ra 1 Kpcs 1 Kic Es (2.27) where Kpc and Kic
U(t) = Ki e(t)dt = e(t)dt (8) represent the proportional and integral gains of the current
K pi
controller, respectively. We can see from Eq. (2.27) that the
The integral controller can remove the steady-state error back-EMF Es of the DC motor will act as a disturbance to
completely for the non-time varying reference R(s). This the current control system. If the current reference of theDC
is because it continuously produces nonzero control input motor remains a constant value (i.e., s 5 0), then Es will have
from the accumulated past errors even though the present no influence on the current control performance. However,
error is zero. However, due to the accumulated errors in the when the current reference I s is changed, Es may have an
integral term, it responds slowly to change in errors. For this influence on the dynamic performance of the actual current Is
reason, the integral controller is not normally used alone but following its reference. For a large system inertia, the effect
is combined with the P controller. of Es on the current control can be ignored since the variation
of Es due to the speed variation is very small. However,
VII. P ROPORTIONALINTEGRAL CONTROLLER for servo motors designed with a small inertia for a fast
A widely used controller in motor drive systems is the speed response, the effect of Es cannot be ignored. To obtain
proportionalintegral controller (PI controller), which is a a good performance of the current control by eliminating
combination of a proportional controller and an integral this undesirable effect of the back-EMF, we can adopt the
controller to achieve both a fast response and a zero steady- feedforward control as described in the Section 2.5.2.3. Since
state error. The block diagram of a proportionalintegral the back-EMF of a DC motor is ea 5 kefm, we can easily
controller is shown in Fig.7. estimate it from the speed information. When the disturbance
Es is compensated by the feedforward control as in Fig.
VIII. CURRENT CONTROLLER DESIGN 2.31A, the DC motor will be simplified as a R-L circuit
The PI controller is the most commonly used in the DC shown in Fig. 2.31B. For the case of Fig. 2.31B, the current
motor drive systems for current control. In this section we control will be easier and its performance will be improved.
will introduce how to design a current controller, i.e., how In this case, Eq. (2.27) becomes Is I s 5 Kpcs 1 Kic Las2 1
to select the P and I gains of a PI current controller for Ra 1 Kpcs 1 Kic (2.28) From Eq. (2.28), it can be seen that
the desired performance. The PI controller consists of a the current control characteristic depends on the performance
proportional term, which depends on the present error, and of the current controller, which will be determined by the P
an integral term, which depends on the accumulation of past and I gains selected for the PI controller. Now, we will look
and present errors as follows: Gpis 5 Kp 1 Ki s (2.25) where at how the proper gains are selected to achieve the desired
Kp and Ki are the proportional and integral gain, respectively. performance of the current control. Here, we will introduce
a gain selection method based on the frequency response of gain Kic can be obtained from Eq. (2.31) as Proportional
the system. gain: Kpc 5 La cc (2.37) Integral gain: Kic 5 Ra La Kpc
5 Ra cc (2.38) It is important to note that the gains of
IX. PROPORTIONALINTEGRAL CURRENT the controller depend on the motor parameters La and Ra.
CONTROLLER Therefore the gain values to achieve the same current control
A current control system containing a PI controller is performance may differ from motor to motor. For example,
shown in Fig. 2.32. Here, assume that the feedforward con- a motor with a large value of winding inductance, whose
trol compensates the back-EMF of the DC motor perfectly. current is difficult to change, needs a larger proportional gain
In this PI current controller, the relationship between the to obtain an equal bandwidth of the current control than that
current error and its output voltage is given as Va 5 Kpc 1 1 with a small value of winding inductance, whose current is
1 Tpis I a 2 Ia (2.29) where Tpi 5 Kpc=Kic is the integral easy to change. Therefore, accurate information about the
time constant of the PI current controller and Kic is the motor parameters is essential to achieve the required current
integral gain. There are several PI gain tuning rules such as control performance.
the famous ZieglerNichols method. Here, we will adopt the
technique called the pole-zero cancellation method. By using X. S ELECTION OF THE BANDWIDTH FOR CURRENT
CONTROL
the pole-zero cancellation, we can remove the current control
characteristic of the DC motor itself so that the PI controller The value of the proportional gain Kpc determines how
may determine the performance of the current control. The fast the system responds, whereas the value of the integral
open-loop transfer function Go c s of this system is given gain Kic determines how fast the steady-stateerror is elim-
by Go c s 5 Kpc s 1 1 Tpi s 0 BB@ 1 CCA 1 Las 1 Ra inated. When the value of these gains is larger, the control
5 Kpc s 1 Kic Kpc s 1 La s 1 Ra La (2.30) If the zero performance is better. However, large gains may lead to
2 Kic=Kpc of the PI controller is designed to cancel the an oscillatory response and result in an unstable system.
pole 2 Ra=La of the DC motor by the pole-zero cancellation As can be seen in Eq. (2.37) and (2.38), since the gains
method, i.e. 1 Tpi 5 Kic Kpc 5 Ra La (2.31) then, Eq. (2.30) of a current controller are determined by its bandwidth cc,
becomes Go c s 5 1 La Kpc s (2.32) Fig. 2.33 shows the we first need to select an appropriate bandwidth cc for the
bode plot of the open-loop frequency response from Go c s. target current control system. A wider bandwidth will lead
The phase is 290 at the gain cross-over frequency cc where to a faster response but an unstable system. The bandwidth
the magnitude Go c j 5 0 dB. Thus the gain margin is cc of a current controller is limited by the following two
positive and the system will bestable. Also, in this case, the factors: the switching frequency of the power electronic
transfer function of Eq. (2.32) becomes the system of type 5 converter realizing the output of the current controller and the
1 and, thus, we can expect the steady-state error to be zero sampling period for detecting an actual current for feedback
as eN 5 lim s-0 1 1 1 Go c s 5 1 1 1N 5 0 (2.33) From Eq. in the digital controller. Since a motor current cannot change
(2.32), the gain cross-over frequency cc can be determined faster than the switching frequency of the power electronic
by Go c jcc 5 1 La Kpc jcc 5 1 - cc 5 Kpc La (2.34) converter, the switching frequency limits the bandwidth of
The gain cross-over frequency cc of the open-loop frequency the current control. If the current is sampled twice every
response is equal to the gain cut-off frequency of the close- switching period, as a rule of thumb, the maximum available
loop frequency response, which indicates the bandwidth of bandwidth can be up to 1/10 of the switching frequency. On
the current control system. Let us check the bandwidth of the other hand, if it is sampled once every switching period,
this current control system. The closed-loop transfer function the maximum available bandwidth can be up to 1/20 of the
of this system is given by Is I s 5 Gc cs 5 Go c s 1 1 Go c switching frequency. For this case, it is desirable to restrict
s 5 1 La Kpc s 1 1 5 cc s 1 cc (2.35) Its frequency response the bandwidth to 1/25 of the sampling frequency [1,2]. As an
is shown in Fig. 2.34. We can see that the bandwidth of this example, suppose that the switching frequency of a chopper,
system is given as cc by letting Gc cj equal to 1= ffiffiffi which provides the voltage applied to a DC motor, is 5
2 p (523 dB) as Gc cj 5 cc j 1 cc 5 1 ffiffiffi 2 kHz. In this case, the maximum available bandwidth of the
p (2.36) Thus the bandwidth is equal to the gain cross-over current controller can be up to at most 1/10 of the switching
frequency of the open-loop frequency response.Like this, if frequency, i.e., 5 kHz/10 5 500 Hz( 3100 rad/s). However,
the zero of the PI controller is designed to cancel the pole of for a stable current control or if the current is sampled once
the system, the current control system is expressed simply every switching period, it is better to restrict to 1/20 of
as a first-order system with a cut-off frequency cc, which is the switching frequency, i.e., 5 kHz/20 5 250 Hz ( 1550
stable. In this case, there will be no overshoot in the response rad/s). Fig. 2.35 compares the current control performances
and the time required for the response to reach the final value according to the gains, which are determined by the gain
will be about four times the time constant of the system. selection method as described earlier. We can identify the
Based on the explanation given above, the proportional and different responses according to the control bandwidth.
integral gains of the current controller needed to achieve
the required bandwidth can be determined as follows. If XI. ANTI-WINDUP CONTROLLER
the required control bandwidth is cc, then the proportional As stated before, we can see that the integral controller can
gain Kpc can be obtained from Eq. (2.34) and the integral effectively eliminate the steady-state error. This is because of
the nature of the integral controller producing its output from Ra cc 4. Select the gain of the anti-windup controller as Ka
the accumulated past errors. However, this nature is often 5 1=Kpc. The procedures 1 and 2 are interchangeable with
the cause of control performance degradation in cases where each other, i.e., the switching frequency can be determined
the output of the controller is limited. The output of the by the required bandwidth cc for current control.
PI current controller, which indicates the voltage reference
applied to a DC motor, should be limited to a feasible value ]SPEED CONTROLLER DESIGN
by the following reasons. First, a voltage exceeding its rated When controlling the speed of a motor, the speed con-
value should not be applied to a motor. Furthermore, power troller must be placed in the outer loop of the current
electronic converters, which produce the voltage applied to controller as shown in Fig. 2.38. In this case, if the bandwidth
the motor, normally have a restricted output voltage due to of current control is chosen to be sufficiently wider than
the limited availability of the input voltage and the voltage that of speed control, then the current control will not have
rating of switching devices. Once the output of the PI any influence on the speed control performance, and the
controller exceeds its limit due to sustained error signals for stability of the speed control will be improved. Like the
a significant period of time, the output will be saturated, but current controller, a PI controller is widely used for speed
the integrator in the controller may have a large value by control. Now, we will explain how the PI speed controller is
its continual integral action. This phenomenon is known as designed, i.e., how to select the proper gains to achieve the
integral windup. When the windup occurs, the controller is desired speed control performance. The transfer function of
unable to immediately respond to the changes in the error a PI speed controller is given by Gpis 5 Kps 1 Kis s (2.39)
due to a large accumulated value inside the integrator. It is where Kps and Kis are the proportional and integral gains of
required that the error have an opposite sign for a long time a PI speed controller, respectively. From the block diagram in
until the controller returns to the normal state. This system Fig. 2.38, the open-loop transfer function of the speed control
becomes an open-loop system because the output remains at system is given by Go s s 5 GpisGc csU KT Js 5 Kps 1 Kis
its limit irrespective of the error. As a result, the system will s U cc s 1 cc U KT Js (2.40) This open-loop characteristic
exhibit a large overshoot and a long setting time. To avoid is the sum of the frequency responses of three parts: a PI
such integrator windup, the magnitude of the integral term speed controller, a PI current controller, and a mechanical
should be kept at a proper value when the saturation occurs, system of a motor and a load. Here, we assume that the
so that the controller can resume the action as soon as the current controller is a first-order system with a bandwidth of
control error changes. There are several methods to avoid cc. Also, the load torque TL is neglected and KT 5 kTf . The
the integrator windup such as back calculation, conditional bode plot of the open-loop frequency responses for a speed
integration, and limited integration [3,4]. The anti-windup control system is shown in Fig. 2.39. Here, cs is the gain
scheme of back calculation as shown in Fig. 2.36 is widely cross-over frequency, which is equal to the bandwidth of the
used because of its satisfactory dynamic characteristic.In speed controller, and pi5Kis=Kps is the PI corner frequency.
this scheme, when the output is saturated, the difference As a rule of thumb, pi is selected as one-fifth of cs.steady-
between the controller output and the actual output is fed state performance will be degraded, but if it is too large,
back to the input of the integrator with a gain of Ka so the system will exhibit an overshoot. As stated before, in
that the accumulated value of the integrator can be kept order for current control to not exert any influence on speed
at a proper value. The gain of an anti-windup controller is control, the current control bandwidth cc should be at least
normally selected as Ka 5 1=Kp to eliminate the dynamics five times wider than the speed control bandwidth cs. Under
of the limited voltage. Fig. 2.37 shows the phenomenon this condition, the gain of the current controller around the
of integrator windup for a PI current controller, which is cross-over frequency cs becomes almost one, i.e., Gc cs 5 cc
generated by a large change in the reference value. Fig. s 1 cc 1 (2.41) Also, if the PI corner frequency pi is much
2.37A shows the performance of a current controller without smaller than the speed control bandwidth cs, then the transfer
an anti-windup control. Due to its saturated output voltage, function of the PI current controller around cs can be reduced
the actual current exhibits a large overshoot and a long as Gpis 5 Kps 1 Kis s Kps (2.42) Consequently the open-
setting time. On the other hand, Fig. 2.37B shows a current loop transfer function of Eq. (2.40) around the crossover
controller with an anti-windup control. When the output is frequency cs can be simplified as Go s s Kps KT Js (2.43)
saturated, the accumulated value of the integrator can be kept The gain of this open-loop transfer function at frequency cs
at a proper value, resulting in an improved performance. is 0 dB, i.e., Go s jcs 5 0 dB (2.44) From this, we can
obtain the proportional gain Kps and the integral gain Kis
XII. G AINS SELECTION PROCEDURE OF THE as Proportional gain: Kps 5 Jcs KT (2.45) Integral gain: Kis
PROPORTIONALINTEGRAL CURRENT CONTROLLER
5 Kps pi 5 Kps cs 5 5 J2 cs 5KTIt is important to note that
1. Find the switching frequency fsw of the system. 2. the gains of a speed controller depend on the system inertia
Select the control bandwidth cc of the current controller to J and the torque constant KT of the target motor. Since J
be within 1/101/ 20 of the switching frequency fsw and and KT can be different from each system, the gains of a
below 1/25 of the sampling frequency. 3. Calculate the speed controller may vary with the applied system. From Eq.
proportional and integral gains from the selected bandwidth (2.45), it can be seen that, to achieve an equal bandwidth of
cc as Proportional gain: Kpc 5 La cc Integral gain: Kic 5 speed control, the value of the proportional gain needs to
be large for a large J and small for a large KT . Like the
current controller, the PI speed controller needs to include
an anti-windup controller to prevent integrator saturation
XIII. CONCLUSIONS
The basic term is the proportional term, P, which generates
a corrective control action proportional to the error. The
integral term, I, generates a correction proportional to the
integral of the error. This assures us that, if we apply a
sufficient control effort, the tracking error is reduced to
zero. The derivative term, D, generates a control action
proportional to the error range change. This tends to have a
stabilizing effect, but usually generates large control actions.
It is extremely important to analyze in the first instance the
type of system with which you are working, the degree of it
will determine how narrow the values of the tuning gains can
reach. It is also important to determine from the beginning
the real stability of the system, since it does not make sense
to implement a control process to a system that is naturally
unstable, in which case it will be necessary to modify the
plant to maintain the poles in the negative real half-axis,
guaranteeing so its stability.
R EFERENCES
[Electric Motor Control DC, AC, and BLDC Mo-
tors] [Contemporary Engineering Sciences, Vol. 11, 2018,
no. 99, 4913 - 4920 HIKARI Ltd, www.m-hikari.com
https://doi.org/10.12988/ces.2018.810539]