8.

Dynamic Behavior of Closed-loop Systems

In Chapter 7, we learned about the different elements of a typical feedback control

system. We also understand how the feedback control operates and simulated its
response to a setpoint and disturbance change. A process control employing a feedback
control loop is called a closed-loop system.

In this chapter, we will revisit the block diagram of a simple feedback control
presented in Chapter 7 to discuss some useful descriptions of closed-loop systems and
use the block diagrams to obtain a closed-loop system's overall transfer function model.
We then discuss the dynamic behavior and stability of closed-loop systems. The specific
learning outcomes at the end of this chapter are to:

1. perform block diagram reduction;

2. analyze the dynamic behavior of the closed-loop system;
3. discuss the general stability criterion of for closed-loop system; and,
4. perform stability analysis using the root locus, Routh stability test, and direct
substitution method.

8.1. Closed-loop transfer function

Consider the block diagram of a standard closed-loop system in Figure 8–1.

Unless otherwise defined, we will refer to this standard block diagram for most of the
discussion in this Chapter. It should be understood that other blocks and input variables
can be added to the diagram (such as in Figure 7–1) for a wide variety of control problems.
In a closed-loop system, the path between the comparator and the controlled variable is
referred to as a forward path. The feedback path is the path between the controlled
variable and the comparator. The open loop transfer function refers to the product of all
transfer functions in the loop; it relates the measured variable (U) to the set point (Ysp) if
the feedback loop is disconnected (i.e., opened) from the comparator. In Figure 8–1, the
forward path includes the blocks Gc, Gv, and Gp, the feedback path includes the block Gm,
and the open loop transfer function is the product of Gc, Gv, Gp, and Gm.

%
!%

!%
!"# " $ # !$ + !
+- !! !" !# +

!!
!$

Figure 8–1. Block diagram of a standard closed-loop system

165
 The overall transfer function of a closed-loop system refers to the transfer functions
that relate the response Y to changes in the inputs Ysp (setpoint) or D (disturbance).
These overall transfer functions provide useful information about the control system; its
usefulness will be apparent in succeeding chapters.

The closed-loop transfer function for setpoint changes is also referred to as the
servo (or servomechanism) problem or servo transfer function. The response of Y to a
change in setpoint Ysp is obtained by assuming no disturbance change occurs, that is, 𝐷
= 0. On the other hand, the closed-loop transfer function for disturbance changes is also
referred to as the regulator problem or regulator transfer function. The response of Y to a
change in disturbance D is obtained by assuming no setpoint change occurs, that is, Ysp
= 0. The desired transfer function is obtained by performing transfer function and block
diagram math operations.

8.1.1. Servo transfer function

S
To get the closed-loop transfer function for set-point changes, S*8
, first set 𝐷 = 0
(see Figure 8–2a). Then, the blocks in the forward path, Gc, Gv and Gp, which are
connected in series can be simplified by multiplicative property.
(Eq. 8-1)
𝐺 = 𝐺( 𝐺e 𝐺%

Accordingly, the following equations are written from transfer function and indicated math
operation in the diagram:
(Eq. 8-2)
𝑌 = 𝐺𝐸
𝑌P = 𝐺P 𝑌 (Eq. 8-3)

𝐸 = 𝑌$% − 𝑌P (Eq. 8-4)

Since there are four variables and three equations, we can solve the equations
simultaneously for Y in terms of Ysp, as shown
(Eq. 8-5)
𝑌 = 𝐺U𝑌$% − 𝑌P W
𝑌 = 𝐺U𝑌$% − 𝐺P 𝑌 W (Eq. 8-6)

𝑌 = 𝐺𝑌$% − 𝐺𝐺P 𝑌 (Eq. 8-7)

Finally, the following expression is obtained:

166
 𝑌 𝐺
= (Eq. 8-8)
𝑌$% 1 + 𝐺𝐺P

(Eq. 8-8) reduces the closed-loop block diagram to a single block diagram shown in Figure
8–2b.
%=0
!%

!%
!"# " $ # !$ + ! !"# ( !
+- !! !" !# +
1 + ((!

!!
!$

(a) (b)
Figure 8–2. Block diagram showing a setpoint change, that is D = 0 (a) and equivalent
block diagram of servo transfer function (b)

8.1.2. Regulator transfer function

S
To get the closed-loop transfer function for disturbance changes, h, set 𝑌$% = 0 (see
Figure 8–3a) and carry out a similar block diagram math operation as previously. From
the diagram the following equations can be written:

(Eq. 8-9)
𝑌 = 𝑌[ + 𝑌h
𝑌` = 𝐸𝐺( 𝐺e 𝐺% (Eq. 8-10)

𝑌' = 𝐷𝐺' (Eq. 8-11)

𝐸 = −𝑌P (Eq. 8-12)

𝑌P = Y𝐺P (Eq. 8-13)

Solve the equations simultaneously for Y in terms of D, as shown

(Eq. 8-14)
𝑌 = −Y𝐺P 𝐺( 𝐺e 𝐺% + 𝐷𝐺'
𝑌 + Y𝐺P 𝐺( 𝐺e 𝐺% = 𝐷𝐺' (Eq. 8-15)

167
 𝑌U1 + 𝐺P 𝐺( 𝐺e 𝐺% W = 𝐷𝐺' (Eq. 8-16)

Substitute (Eq. 8-1) into (Eq. 8-16)

𝑌(1 + 𝐺𝐺P ) = 𝐷𝐺' (Eq. 8-17)

Finally, the following expression is obtained:

𝑌 𝐺'
= (Eq. 8-18)
𝐷 1 + 𝐺𝐺P

The equivalent block diagram for (Eq. 8-18) is shown in Figure 8–3b. Note that the closed-
loop transfer functions for setpoint and disturbance change, (Eq. 8-8) and (Eq. 8-18),
respectively, have identical denominator 1 + 𝐺𝐺P .

'
!%

!%
!"# = 0 " & # !$ + ! ' (% !
+- !! !" !# +
1 + ((!

!!
!$

(a) (b)
Figure 8–3. Block diagram showing a disturbance change, that is 𝑌$% = 0 (a) and
equivalent block diagram of the regulator transfer function (b)

For a single loop feedback control system, the overall transfer function can be
conveniently derived by following a generalized rule for any pair of input (X) and output
(Y) variables.

𝑌 𝜋+
Negative feedback*: = (Eq. 8-19)
𝑋 1 + 𝜋no

𝑌 𝜋+ (Eq. 8-20)
Positive feedback*:
=
𝑋 1 − 𝜋no

168
Where:
𝜋+ = product of transfer function in the path between the locations of the signal X and Y
𝜋no = is the open loop transfer function or product of all transfer functions in the loop
*Negative and positive feedback is indicated by the operator for Ym as it enters the
comparator (see in Figure 8–1).

8.2. Block diagram reduction

A control system may contain multiple loops and multiple input variables. To obtain
the overall transfer function of such a system, it is often convenient to combine several
blocks into a single block. This section will explore the method to reduce the block
diagram of a multiloop system to a single-loop block diagram to obtain an overall transfer
function expression for the process through an example.

Example 8.1. Overall transfer function of multiloop system

Given the block diagram below, derive the closed-loop transfer function for
setpoint, D1, D2 changes.

#1 !%"
#2

!
" + +
+- !! 1 +- !! 2 !" !# + + !&

!$#

!$"

Figure 8–4. Block diagram for Example 8.1

Solution:
Reduce the inner loop involving Gc2, G1, G2, and Gm2 to a single block using the
generalized rule for negative feedback in (Eq. 8-19).

𝐺(# 𝐺" 𝐺#
𝐺O = (Eq. 8-21)
1 + 𝐺(# 𝐺" 𝐺# 𝐺P#

The block diagram is then reduced to a single loop in Figure 8–5.

169
 #1 !"#
#2

"
+ +
! +- !& 1 !! + + !$

!%#

Figure 8–5. Block diagram reduction of Figure 8–4

To obtain the transfer function Y/R, set D1 = 0 and D2 = 0, apply the generalized
rule for negative feedback to Figure 8–5.

𝑌 𝐺(" 𝐺O 𝐺B
= (Eq. 8-22)
𝑅 1 + 𝐺(" 𝐺O 𝐺B 𝐺P"

The overall transfer functions Y/D1 and Y/D2 can also be obtained in the same manner.

𝑌 𝐺'" 𝐺B
= (Eq. 8-23)
𝐷" 1 + 𝐺(" 𝐺O 𝐺B 𝐺P"

𝑌 𝐺B (Eq. 8-24)
=
𝐷# 1 + 𝐺(" 𝐺O 𝐺B 𝐺P"

Aside from the generalized rule for feedback systems, there are other rules that
can be applied for block diagram manipulation to reduce a complicated block diagram
into a single loop system. These rules are summarized in Table 8-1.

170
Table 8-1. Block diagram equivalent transformation

171
 Table 8-1. Continued

Source: https://www.benardmakaa.com/wp-content/uploads/2018/05/Block-Diagram-
Transformation-Theorems.pdf

8.3. Stability of Closed-loop Systems

By definition, the system response is stable if the output response is bounded for
all bounded inputs; otherwise, it is unstable. A bounded input is an input that stays within
upper and lower limits (or bounds) for all time values. For example, a step and sinusoidal
function are bounded, while a ramp function is not bounded since its value continually
increases or decreases with time. We also perform stability analysis of a feedback control

The block diagram of a simple proportional feedback control system and its
corresponding unit step response to a setpoint change at different values of controller
gain (Kc) are shown in Figure 8–6. With the unit-step change (a bounded input), notice
that as Kc increases, the system's response becomes more oscillatory but remains
bounded at moderate values of Kc. However, at a high enough value of Kc, successive
amplitudes of the response increasingly become unbounded -hence, the system
becomes unstable.

172
 #

! + 1 "
+- !% = (% + !! =
$!% + 1 $"% + 1
1 1
!! = 1, !" = , &'( !# =
2 4

1
!# =
$$ % + 1

(a)

(b)
Figure 8–6. Proportional feedback control diagram (a) and responses for a unit step
change in setpoint (b)

As demonstrated, feedback control can cause an oscillatory response. The

response may become unstable depending on controller parameters, such as Kc.
Typically, if the oscillation has a small amplitude and damps out quickly, the control
system performance is generally considered satisfactory. In general, for satisfactory
control, the control system must be stable.

173
8.4. General stability criterion

The stability of a closed-loop system is determined by its characteristic equation,

which is the denominator of the closed-loop transfer function equated to zero. In general,
the characteristic equation for the closed-loop system is:

1 + 𝜋no = 0 (Eq. 8-25)

Where 𝜋no (as defined earlier) is the open loop transfer function or product of all transfer
functions in the loop. A closed-loop system is stable if and only if all roots of its
characteristic equation are negative or have negative real parts (or at the left-hand side
of the complex plane). Otherwise, the system is unstable. Recall from Section 4.2.5 that
the roots of characteristic equation are also called the poles. The implication of the
location of the poles in control stability is revisited in Figure 8–7.

Stability of typical roots in characteristic equation

Limit of stability!!!
Is when the pole is purely
imaginary (no real part)

At RHP:
Variation exponentially
At LHP: grow; he farther from the
Variation exponentially origin, the faster is the
decay; the farther from growth
the origin, the faster is
the decay

Figure 8–7. Implication of the location of the roots of the characteristic equation in
control stability

Root/s on the right-hand side of the complex plane give an unstable response.
Roots at the imaginary axis indicate the limit of stability; the value of Kc corresponding to
these roots of the characteristic equation of the controller is called the ultimate controller
gain or Kc,u, which is the maximum value of Kc that can be used to achieve a stable control.

For the proportional feedback control in Figure 8–6a, the characteristic equation is

1 1 (Eq. 8-26)
1 + 𝐾( Ÿ f g=0
(𝑠 + 1)(0.5𝑠 + 1) 0.25𝑠 + 1

174
which can be simplified to

(𝑠 + 1)(0.5𝑠 + 1)(0.25𝑠 + 1) + 𝐾( = 0 (Eq. 8-27)

Note that assigning different values to Kc in (Eq. 8-27) will yield a different set of
roots, as presented in Table 8-2.

Table 8-2. Roots of characteristic equation (Eq. 8-27) at different values of Kc

Kc s1 s2 s3
2 -4.31 -1.19 + 1.51j -1.19 - 1.51j
6 -5.93 -0.38 + 3.05j -0.38 - 3.05j
10 -6.6 -0.05 + 3.65j -0.05 - 3.65j
12 -6.87 0.08 + 3.89j 0.08 - 3.89j

At values of Kc = 2, 6, and 10, the roots of the character equation are all located at
the left-hand side of the complex plane (or have negative real parts), indicating stability.
For Kc = 12, two of the roots are on the right-hand complex plane, thus, indicating
instability (see Figure 8–7). Note that these results are consistent with the response
simulation of the control system shown in Figure 8–6b.

The stability of a control system ensures that its output is under control and thus,
is the most important characteristic of control systems. Controller stability analysis
determines the range of controller gains that lead to a stabilizing controller. Several
methods can be used to determine the controller gains between a lower limit and the
upper limit. For our purpose, we will only consider the Root-locus method, Routh test, and
the direct substitution method.

8.5. Root-locus diagram

The root locus diagram is the plot of the roots of the characteristic equation of the
closed-loop system on a complex plane as the system parameters, such as 𝐾𝑐, is varied.
It is a convenient graphical display in analyzing the system stability, the nature of closed-
loop responses as the value of the 𝐾𝑐 is changed, and in determining the range of 𝐾𝑐
values for which the controller will continue to be stable. In this section, we will explore
the controller stability analysis by root locus method using the “sisotool” function and
Control System Designer in MATLAB. It should be noted that for this application to work,
the Control Systems Toolbox is required to be installed in MATLAB.

Example 8.3. Stability analysis by root locus method

Perform stability analysis by root locus method for the proportional feedback
control system in Figure 8–6a.

175
Solution:
Let GH = the open-loop transfer function without the controller gain 𝐾𝑐

1 1 (Eq. 8-28)
𝐺𝐻 = Ÿ f g
(𝑠 + 1)(0.5𝑠 + 1) 0.25𝑠 + 1

Build the transfer function GH in Matlab, then call the Sisotool built-in application
as shown in Figure 8–8. Note that the transfer function GH created is in polynomial form
and the Control System Designer interface of the Sisotool application opens as shown in
Figure 8–9. By default, the interface shows the Bode editor and Root locus editor and
the response windows, for our purpose the Bode editor window may be ignored or
minimized.

Figure 8–8. Building transfer function and calling “sisotool” application in MATLAB to
generate the root locus

176
 Figure 8–9. Control System Designer interface of the Sisotool application

In the root locus window, drag a conjugate pole (pink marker), notice that the
response window correspondingly changes, as well as the value for C (which is also the
value of Kc) as shown in Figure 8–10 to Figure 8–13. Drag the pole until it crosses the
imaginary axis, and the response become unstable. Through the root locus diagram, we
can simulate the control system's response by experimenting with different values
of Kc to determine the range of values of Kc that will give a satisfactory control. In this
example, it appears that a controller gain Kc > 11.25 will give an unstable response.

177
Figure 8–10. Control System Designer interface showing only the Root locus editor and
the unit step response window

Figure 8–11. Moving poles towards but before the imaginary axis

178
Figure 8–12. Moving poles on to the imaginary axis

Figure 8–13. Moving poles beyond the imaginary axis

8.6. Routh stability test

The roots of the 1st- or 2nd-order system can be found easily by the analytical
method. However, for the higher-order transfer function, which often occurs in a closed-
loop system, finding the roots of the characteristic equation may be a challenge; hence,

179
numerical root-finding techniques are normally employed, often with the aid of a
computing software like MATLAB. The Routh test provides an alternative technique in
evaluating the stability without requiring calculation of the roots of the characteristic
equation; the test is based on the analysis of the coefficients of the characteristic
equation. It is only applicable to systems whose characteristic equations are polynomials
in s, hence, it cannot be directly used in systems containing a time delay (transportation
lag). However, an approximation of time delay by Taylor and Padé technique can be
employed.

The Routh stability test (also called Routh Stability Criterion) states that a
necessary and sufficient condition for all roots of the characteristic equation (in standard
polynomial) to have negative real parts is that all of the elements in the left column of the
Routh array are positive.

The Routh test can be used to check the system's stability at specific values of the
controller gain (Kc) or determine the possible range of values of the Kc that gives a
stabilizing control. Such cases are illustrated through examples. The steps for Routh
stability test, are summarized as follows:

1. Check the applicability to the control system under study

a. Write the characteristics equations in standard polynomial form:

𝑎6 𝑠 . + 𝑎" 𝑠 .7" + 𝑎# 𝑠 .7# + ⋯ + 𝑎. = 0 (Eq. 8-29)

Where ao is positive. (If ao is originally negative, both sides are multiplied by -1);
ao , a1 , a2 , . . . , an-1 , an are coefficients of the polynomial equation.

b. If any coefficient is negative, the system is definitely unstable, hence, there is

no need to apply Routh test
c. If all the coefficients are positive, the system may be stable or unstable, then,
apply the Routh test by proceeding to step 2.

2. Arrange the coefficients of the characteristic equation (in standard polynomial form)
into the first two rows of the Routh array as shown in Table 8-3.

Table 8-3. Routh array

Row
1 𝑎6 𝑎# 𝑎A 𝑎p
2 𝑎" 𝑎B 𝑎b 𝑎d
3 𝑏" 𝑏# 𝑏B
4 𝑐" 𝑐# 𝑐B
5 𝑑" 𝑑#
6 𝑒" 𝑒#
7 𝑓"
n+1 𝑔"

180
3. Solve values in the remaining rows in Routh array from the following:
𝑎" 𝑎# − 𝑎6 𝑎B (Eq. 8-30)
𝑏" =
𝑎"
𝑎" 𝑎A − 𝑎6 𝑎b (Eq. 8-31)
𝑏# =
𝑎"

𝑏" 𝑎B − 𝑎" 𝑏# (Eq. 8-32)

𝑐" =
𝑏"

𝑏" 𝑎b − 𝑎" 𝑏B (Eq. 8-33)

𝑐# =
𝑏"

The elements for the other rows can be solved by following the same equation pattern
of (Eq. 8-30) to (Eq. 8-33).

4. Apply Theorems of Routh test criterion to determine stability

a. Theorem 1: All elements of the first column of the Routh array are positive and
nonzero
b. Theorem 2: If some elements in the first column of the Routh array is negative, the
number of sign changes is equal to the number of roots with positive real parts.
c. Theorem 3: If one pair of roots is on the imaginary axis, equidistant from the origin,
and all other roots are in the left half plane, the location of the pair of imaginary
roots can be found by solving the equation:

𝐶𝑠 # + 𝐷 = 0 (Eq. 8-34)

where the coefficients C and D are the elements of the array in the (n - 1)th row as
read from left to right, respectively.

Example 8.4. Routh stability test

Perform stability analysis by root locus method for the proportional feedback
control system in Figure 8–6a.

Solution:
Write the simplified characteristic equation (Eq. 8-27) of Figure 8–6a in standard
polynomial form

0.125𝑠 B + 0.875𝑠 # + 1.75𝑠 + (1 + 𝐾( ) = 0 (Eq. 8-35)

181
Since coefficients are positive and the value of Kc still unknown, the Routh test may be
applied to the control system.

Row
1 0.125 1.75
2 0.875 1 + 𝐾(
3 1.61 − 0.143𝐾(
4 1 + 𝐾(

Apply Theorem 1, all elements of the first column of the Routh array are positive and
nonzero for stability, hence
1.61 − 0.143𝐾( > 0 (Eq. 8-36)

11.259 − 𝐾( > 0 (Eq. 8-37)

𝐾( < 11.259 (Eq. 8-38)

If the proportional sensitivity of the controller Kc is a positive quantity, the fourth

entry in the Routh array is positive. Therefore, the system will be stable only if Kc < 11.259.
Verify that this result is consistent with the result using the Root locus method in the
previous example. Take note that the Routh test tells us only that for Kc < 11.259 the
system is stable, it does not provide further information on the nature of closed-loop
responses within the stable value of Kc, unlike in the Root locus method.

8.7. Direct substitution method

The direct substitution method is used to find the stability limit of a control system
(see Figure 8–7), Kcu or ultimate controller gain. The stability limit of the control is
determined by directly substituting 𝑠 = 𝑗𝜔 in the polynomial form of the characteristic
equation. This method is related to Theorem 3 of the Routh test, discussed in the previous
section. This method will be illustrated in the preceding examples.

Example 8.4. Direct substitution method

Using the direct substitution method, determine Kc,u for the control system given in
Figure 8–6a.

Solution:
Using the characteristic equation already derived in Example 8.3

0.125𝑠 B + 0.875𝑠 # + 1.75𝑠 + (1 + 𝐾( ) = 0 ((Eq. 8-35)

182
Substitute 𝑠 = 𝑗𝜔 and 𝑗 # = −1 into (Eq. 8-35)

−0.125𝑗𝜔B − 0.875𝜔# + 1.75𝑗𝜔 + (1 + 𝐾( ) = 0 (Eq. 8-39)

Group the real and imaginary terms

(1 + 𝐾( − 0.875𝜔# ) + 𝑗(1.75𝜔 − 0.125𝜔B ) = 0 (Eq. 8-40)

If both the real and imaginary parts are identically zero, hence

1.75𝜔 − 0.125𝜔B = 0 (Eq. 8-41)

1 + 𝐾( − 0.875𝜔# = 0 (Eq. 8-42)

Solving 𝜔 from (Eq. 8-41) gives the crossover frequency 𝜔(6

𝜔# = 14 (Eq. 8-43)

𝜔(6 = 3.74 (Eq. 8-44)

Substitute (Eq. 8-43) into (Eq. 8-42), finally gives

𝐾(,` = 11.25 (Eq. 8-45)

Note that 𝜔 is also zero in (Eq. 8-41), however, using 𝜔 = 0 in (Eq. 8-42) gives
𝐾(,` = −1, but since the proportional sensitivity of the controller Kc is a positive quantity,
it cannot be the solution. Therefore, it can be concluded that the limit of stability of the
control system is at Kc,u = 11.25, which is consistent with the first two methods.

Summary

In this chapter, we further our discussion on feedback control, also called a closed-
loop system. Using the block diagram algebra, we derived the servo and regulator transfer
function of a closed-loop system or the overall transfer function of a closed-loop system
due to a setpoint change and disturbance change, respectively. We also demonstrated
using Simulink that some values of feedback controller setting, particularly proportional
gain, can give an unstable response. In general, for satisfactory control, the control
system must be stable. The stability of a closed-loop system can be determined by
examining the roots of its characteristic equation. A closed-loop system is stable if and
only if all roots of its characteristic equation are negative or have negative real parts (or
at the left-hand side of the complex plane). Otherwise, the system is unstable. Stability
analysis using the root-locus diagram, Routh stability test, and direct substation can be
used to determine the controller gain stability range.

183