Performance Criteria

(and Controllers)
Introduction
Performance criteria are the requirements that a control system is required to meet. Generally, the criteria are derived from the
system’s response. A set of parameters are used to define the system and specifications of the system will require them to be at
certain levels. If the system does not meet certain required criteria, the response can be altered by either introducing a controller
or other compensating elements into the control system.

Transient Performance Specifications

Transient performance parameters are defined based on the second order step response of an under damped system as illustrated
in figure 4.1. The mathematical relationships between these parameters and the undamped natural frequency and damping ratio
can be obtained by applying mathematical principles to the time response expression.

ZMOOD FIGURE 6.1 AND BURNS FIGURE 3.21

Figure 4.1

Peak Overshoot
Peak overshoot or maximum overshoot is the largest deviation of the input over the final output value for a step input during the
transient period. The value of the overshoot can be obtained by determining the turning points of the response waveform. The
peak overshoot occurs at the first turning point. From calculus manipulations, the time at which the peak overshoot occurs is
𝜋
given by 𝑡𝑝𝑒𝑎𝑘 = 2
.
𝜔𝑛 √1−𝜁
The value of the peak overshoot is obtained by substituting this time value into the expression of the output. The overshoot can
be expressed in two ways:
2
1. Absolute : 𝑐𝑝 = 1 + 𝑒 −𝜋𝜁/√1−𝜁
2
2. As a fraction of the steady state output: 𝑀𝑝 = 𝑒 −𝜋𝜁/√1−𝜁

It can be observed that the peak overshoot depends only on the damping ratio. Typically peak overshoot is required to be below
a certain value.

Rise Time
Rise time, Tr, is the time required for the step response to rise from 10% to 90% of its final steady state value. The rise time is
1
can be approximated by 𝑇𝑟 ≈ (1 + 1.4𝜁) for 0 ≤ 𝜁 ≤ 0.7. Typically rise time is required to be below a certain value.
𝜔𝑛

41
Delay Time
Delay time, Td, is the time required for the step response to reach some value (usually 50%) of the final steady state value. The
1
value can be approximated by 𝑇𝑑 ≈ (1.1 + 1.2𝜁). Typically delay time is required to be below a certain value.
𝜔𝑛

Settling Time
Settling time, Ts, is the required for the system output to settle down to within a tolerance band of the final value, normally
between ±2% or ±5%.

𝑒 𝜁𝜔𝑛𝑡
The expression of the output is given by 𝑐(𝑡) = 1 − sin(𝜔𝑑 𝑡 + ∅). The peaks of the decaying oscillations touch the
√1−𝜁 2
𝑒 𝜁𝜔𝑛𝑡 𝑒 𝜁𝜔𝑛𝑡
envelope given by the expressions 1 + and 1 − . Figure 4.2 illustrates the envelope. The settling time can be
√1−𝜁 2 √1−𝜁 2
approximated by finding the point at which the envelope is within the tolerance band X. This gives the general formula for
𝑋 ln 𝑋
settling time as 𝑇𝑠 ≈ = .
𝜁𝜔𝑛 −𝛼
When the general formula is applied to the settling times of 2% and 5% it gives the following formulae:
4 4
2% settling time 𝑇𝑠 ≈ =
𝜁𝜔𝑛 −𝛼

3 3
5% settling time 𝑇𝑠 ≈ =
𝜁𝜔𝑛 −𝛼

FIGURE 3.21 BURNS

Figure 4.2

Exercise 4.1
𝐾𝜔𝑛 2
The step response of a control system for a step input is given by 𝐶(𝑠) = . If 𝜔𝑛 = 8𝑟𝑎𝑑/𝑠 and 𝜁 = 0.5,
𝑠(𝑠 2 +2𝜁𝜔𝑛 𝑠+𝜔𝑛 2 )
determine 𝑐(𝑡), 𝜔𝑑 , settling timer, the peak overshoot, the time at which peak overshoot occurs, rise time and delay time.

Disturbance Rejection
There are situations where there is need to specify the system in terms of how much it can reject disturbances. Disturbances can
be modeled as shown in figure 4.3. A disturbance signal in theory can add to any point on the block diagram but can be moved
to any appropriate point for ease of analysis using block diagram analysis.

42
 D(s)

Controller Process
+
R(s) + U(s) UT(s) C(s)
E(s) +
Gc(s) Gp(s)

Transducer
B(s)
H(s)

Figure 4.3

In this case peak overshoot, rise time and settling time are measured with respect to the disturbance input. Same definitions will
still however apply, only that the input would be D(s) instead of R(s).

With reference to figure 4.3, the closed loop system containing a controller can be analyzed as follows to show effects of both
the reference and the disturbance inputs:

By block analysis, 𝐶(𝑠) = 𝐺𝑐 (𝑠){𝐺𝑐 (𝑠)(𝑅(𝑠) − 𝐻(𝑠)𝐶(𝑠)) + 𝐷(𝑠)}

𝐺𝑐 (𝑠)𝐺𝑝 (𝑠)𝑅(𝑠) 𝐺𝑝 (𝑠)𝐷(𝑠)

By mathematical manipulations: 𝐶(𝑠) = +
1+𝐺𝑐 (𝑠)𝐺𝑝 (𝑠)𝐻(𝑠) 1+𝐺𝑐 (𝑠)𝐺𝑝 (𝑠)𝐻(𝑠)
= 𝐶𝑅 (𝑠) + 𝐶𝐷 (𝑠)
= 𝐺𝐶𝐿 (𝑠)𝑅(𝑠) + 𝑆(𝑠)𝐷(𝑠)

CR(s) is output due to the reference input. CD(s) is output to the disturbance input. 𝐺𝐶𝐿 (𝑠) is the output-to-reference TF while
𝑆(𝑠) is the output-to-disturbance. The two components of the output can be separately determined and added into total system
response using the principle of superposition.

Transient analysis measures can be applied to the output due to either the reference signal or the disturbance signal.

Obtaining a Second Order TF from the Step Response waveform

The TF of a system can be determined from the step response waveform of a system (response to a step reference input and
zero disturbance input) by performing the following steps:
𝑐
1. Determine the steady state output. This will give the steady state gain from the formula 𝐾 = 𝑠𝑠, where css is steady
𝐴
state output value and A is the magnitude of the step input.
2. Determine the peak overshoot from the waveform. The peak overshoot will give the damping ratio of the system from
2
the formula 𝑀𝑝 = 𝑒 −𝜋𝜁/√1−𝜁 .
3. Determine the period of the damped oscillations. This period will give the damped frequency from the formula 𝜔𝑑 =
2𝜋
where T is the damped period. Undamped natural frequency is then obtained from 𝜔𝑑 = √1 − 𝜁 2 .
𝑇
4. Substitute values of K, ωn and ζ into the general TF of a second order system.

Steady State Performance

In the steady state, performance is measured in terms of the steady state error. The steady state error tells how accurate the
system is. Parameters used to specify steady state performance are the error constants (also known as error coefficients) and the
system type.

Error Constants
𝑅(𝑠)
It is noted that the steady state error of a system is given by 𝑒𝑠𝑠 = lim ( ).
𝑠→0 1+𝐺(𝑠)𝐻(𝑠)

43
Position Error Constant
It is also known as the step error constant. Consider the steady state error when the input is a step input of magnitude A. The
𝐴
𝑠( )
𝑠
steady state error would be 𝑒𝑠𝑠 = lim ( )
𝑠→0 1+𝐺(𝑠)𝐻(𝑠)
𝐴
=
1+lim 𝐺(𝑠)𝐻(𝑠)
𝑠→0

The position error constant Kp is defined by the formula: 𝐾𝑝 = lim 𝐺(𝑠)𝐻(𝑠).

𝑠→0

𝐴
The steady state error is then given by 𝑒𝑠𝑠 = .
1+𝐾𝑝

Velocity Error Constant

It is also known as the ramp error constant. Consider the steady state error when the input is a ramp input of slope R. The steady
𝑅
𝑠( 2 )
𝑠
state error would be 𝑒𝑠𝑠 = lim ( )
𝑠→0 1+𝐺(𝑠)𝐻(𝑠)
𝑅
=
lim 𝑠𝐺(𝑠)𝐻(𝑠)
𝑠→0

The position error constant Kv or Kr is defined by the formula: 𝐾𝑣 = lim 𝑠𝐺(𝑠)𝐻(𝑠).

𝑠→0

𝑅 𝑅
The steady state error is then given by 𝑒𝑠𝑠 = or 𝑒𝑠𝑠 = .
𝐾𝑣 𝐾𝑟

Acceleration Error Constant

Consider the steady state error when the input is a parabolic input of magnitude R.
𝑅
𝑠( 3 )
𝑠
The steady state error would be 𝑒𝑠𝑠 = lim ( )
𝑠→0 1+𝐺(𝑠)𝐻(𝑠)
𝑅
=
lim 𝑠 2 𝐺(𝑠)𝐻(𝑠)
𝑠→0

The position error constant Ka is defined by the formula: 𝐾𝑎 = lim 𝑠 2 𝐺(𝑠)𝐻(𝑠).

𝑠→0

𝑅
The steady state error is then given by 𝑒𝑠𝑠 = .
𝐾𝑎

Interpretation of Error Constants

In all the three cases, the steady state error is seen to depend on the error constants. The error constants are, therefore, measures
of the system error in following the step input, ramp input and parabolic input. The actual value of the steady state error
depends on the error constant as well as on the magnitude of the input.

Exercise 4.2
𝐾
The open loop TF of a unity feedback system is given by 𝐺(𝑠) = .
𝑠(2𝑠+5)
a) If 𝐾 = 10, determine the steady state error and error constants for a step, ramp and parabolic input.
b) What value of 𝐾 is required for the steady state error for a step input to be at most 5%?

System Type
Type of a system is the order of the pole at 𝑠 = 0 in the open loop TF, 𝐺(𝑠)𝐻(𝑠), of the system. The open loop TF can be
𝐾(1+𝑇1 𝑠)(+𝑇2 𝑠)⋯(+𝑇𝑚 𝑠)
expressed in the form: 𝐺(𝑠)𝐻(𝑠) = where K and all T are constants. The type is defined as being l.
𝑠 𝑙 (1+𝑇𝑎 𝑠)(+𝑇𝑏 𝑠)⋯(+𝑇𝑛𝑠)

The form of the expression given above is known as the gain form. It is obtained by factorizing the denominator and numerator
of the TF and then expressing the factors in the standard form as given. In this form, the error depends entirely on the system
type, l and the gain, K of the system.

Examples to illustrate type of a system are:

3
1. 𝐺(𝑠)𝐻(𝑠) = 2 is of type 0.
𝑠 +3𝑠+2

44
 25
2. 𝐺(𝑠)𝐻(𝑠) = is of type 3.
𝑠 3 (𝑠+4)(𝑠+5)(𝑠+1)

𝐾(1+0.5𝑠)
3. 𝐺(𝑠)𝐻(𝑠) = is of type 1.
𝑠(1+𝑠)(1+2𝑠)

Effect of System Type on Steady State Error

Conditions for zero steady state error for various inputs
It is noted that for any of the three standard inputs, the steady state error is zero when the respective error constant goes to
infinity. In general specification of a system, the error due to a particular type of input is required to be below a certain value.
This will correspond to the corresponding error constant being above a particular value (as determined by the error formula).

Steady State Error due to a Step Input

For the steady state error to be zero, the system must be of at least type 1.
For type 0 systems, the error would be a constant (while for type 1 and higher, the error will be zero.

Steady State Error due to a Ramp Input

For the steady state error to be zero, the system must be of at least type 2.
For type 0 systems, the error would be a constant (while for type 1 and higher, the error will be zero).

Steady State Error due to a Parabolic Input

For the steady state error to be zero, the system must be of at least type 3.
For type 0 and type 1 systems, the error would be infinity. For type 2 systems, the error would be a constant (while for type 3
and higher, the error will be zero).

Table 4.1 summarizes effects of system type and input type on the steady state error.

Table 4.1
Input function Steady State Error
Type 0 Type 1 Type 2
Step Constant Zero Zero
Ramp Increasing Constant Zero
Parabolic Increasing Increasing Constant

Disturbance Rejection
For steady state analysis of disturbance rejection, the same approach is used, however for the output with reference to the
disturbance input.

Other Performance Measures

The other performance measures of a control system in addition to the ones described in the previous sections are listed next.
Some of the measures will be described in later topics.
1. Bandwidth
2. Damping ratio
3. Natural frequency
4. Performance Indices

Controllers
When the operation of a control system fails to satisfy specified transient or steady state performance, either or both of these
characteristics may be modified by introducing a controller, specified by the transfer function Gc(s) into the system. The
controller will alter the response of the system but it may also bring about additional changes that may not be desirable. The
choice of the controller is thus a compromise between a number of desired specifications.

45
The most common configuration is to introduce the controller in the forward path as illustrated in figure 4.4 (a) a controller may
however, also be introduced in the feedback path as illustrated in figure 4.4 (b). The configuration in figure 4.4 (a) will be used
as the basis for discussions on controllers.

D(s)

Controller Process
+ C(s)
R(s) + E(s) U(s) UT(s)
Gc(s) + Gp(s)
-

Transducer
B(s)
H(s)

Figure 4.4 (a)

D(s)

Process

Ua(s) +
R(s) + E(s) UT(s) C(s)
+ + Gp(s)
- -

U(s) Controller

Gc(s)

Transducer
B(s)
H(s)

Figure 4.4 (b)

𝑈(𝑠)
The actuating signal, U(s), is directly dependent upon the controller transfer function, G(s). Mathematically, 𝐺𝑐 (𝑠) = .
𝐸(𝑠)

Controller Assessment Framework

In a control system, the output is required track the input signal r(t), and attenuate or reject the disturbance signals, d(t). Hence
a control system is assessed in terms of:
1. Reference tracking performance
2. Disturbance rejection performance.
In addition, it is important to ensure that after implementation of the controller, the system is stable.

46
 𝐺𝑐 (𝑠)𝐺𝑝 (𝑠)𝑅(𝑠)
With reference to figure 4.4(a) and as already shown in an earlier section, tracking response is given by 𝐶𝑅 (𝑠) =
1+𝐺𝑐 (𝑠)𝐺𝑝 (𝑠)𝐻(𝑠)
𝐺𝑝 (𝑠)𝐷(𝑠) 𝐶(𝑠) 𝐺𝑐 (𝑠)𝐺𝑝 (𝑠)
and disturbance rejection response is given by 𝐶𝐷 (𝑠) = . The respective TFs are = and
1+𝐺𝑐 (𝑠)𝐺𝑝 (𝑠)𝐻(𝑠) 𝑅(𝑠) 1+𝐺𝑐 (𝑠)𝐺𝑝 (𝑠)𝐻(𝑠)
𝐶(𝑠) 𝐺𝑝 (𝑠)
= .
𝐷(𝑠) 1+𝐺𝑐 (𝑠)𝐺𝑝 (𝑠)𝐻(𝑠)

To analyze the performance of a system for a given controller, the controller is incorporated into the system, the resulting
closed loop responses determined after which transient and steady state performance measures are applied separately to the
reference tracking response and the disturbance rejection response. The results are compared to those obtained when the
controller is not implemented. The comparison results will directly imply the attributes of the controller in terms of effect on
parameters like response speed, peak overshoot, steady state error, etc.

Stability
The reference tracking and disturbance rejection TFs share the same denominator, hence, they have the same poles. Stability of
a system is determined from the position of its poles, hence the analysis of the stability of tracking and disturbance rejection
responses can be performed once. Methods of checking stability will be covered in a later topic.

Reference Tracking Performance

Analysis of reference tracking involves looking at the shape of the time response in the output with respect to the reference
input. The transient part concerns speed of response and size of the overshoot. The steady state part concerns the steady state
error.

Disturbance Rejection Performance

Analysis of disturbance rejection involves looking at the shape of the time response in the output with respect to a disturbance
input. The transient part concerns the size of the overshoot. The steady state part concerns the steady state error.

Servo and Regulator Design Problems

1. When a controller is designed to make a system follow a frequently changing reference signal, this is referred to as a
servo-control design problem.
2. When a controller is designed for cases where the reference signal is held at a steady value for long periods and the
system is required to reject disturbance signals, this is referred to as a regulator control design problem.

In general, it is noted that always 𝐺𝐶𝐿 (𝑠) + 𝑆(𝑠) + 1. Hence reference tracking cannot be changed without changing
disturbance rejection properties. This makes the control design difficult.

The Basic Forms of Common Type of Controllers

The most commonly used control Laws are the proportional, integral and derivative (PID) control Laws. They are based on
three basic forms of controllers which are proportional, integral and derivative control. The name, thus takes after the first
letters of the respective controller terms. PID control laws are commonly used because:
1. They are easy to understand
2. One can design without much understanding of the Physics.
3. A model of the system is not required (especially important in Chemical Industry).

The underlying principle is to act according to:

1. Current output (proportional)
2. Past outputs (integral)
3. Prediction of future outputs (derivative)

Proportional Control
The control action is given by 𝑢(𝑡) = 𝐾𝑃 𝑒(𝑡)
In Laplace domain 𝑈(𝑠) = 𝐾𝑃 𝐸(𝑠). Hence the transfer function of the controller is given 𝐺𝑐 (𝑠) = 𝐾𝑃 .

KP is called the proportional gain constant. Increasing the proportional action involves increasing the gain KP. Increasing the
proportional action has the following effects:
1. It speeds up the transient response of both the reference tracking and disturbance rejection responses.
2. It reduces but does not entirely remove steady state errors for both reference tracking and disturbance rejection
response.
3. It generally increases the size of the control signal which achieves good reference tracking and disturbance rejection.

47
 4. It may cause the controller output signal to be too large which may lead to saturation or limiting problems of actuators.
Actuators have a limit on the physical range of inputs they can respond to.
5. It may cause instability.

The facts can be verified by analyzing the various performance parameters (Tr, Ts, ess, ζ, ωn, etc.) as the value of KP is varied.
VERIFY THE SAID EFFECTS OF PROPORTIONAL CONTROL.

As an example, consider the effect of KP on the position error constant. Position error constant 𝐾𝑝 = lim {𝐾𝑃 𝐺𝑝 (𝑠)𝐻(𝑠)}.
𝑠→∞
Hence, as KP increases, the position error constant increases thereby reducing the steady state error.

The extent to which KP can be increased to achieve improvement of particular performance measures is limited by desired peak
overshoot and the need for the system to be stable. In principle, we can, thus, choose a value of KP to give a required steady
state error as long that value will not lead to other undesirable effects.

Integral Control
𝑡
The control action is given by 𝑢(𝑡) = 𝐾𝐼 ∫0 𝑒(𝑡)𝑑𝑡
𝐾𝐼 𝐾𝐼
In Laplace domain 𝑈(𝑠) = 𝐸(𝑠). Hence the transfer function of the controller is given 𝐺𝑐 (𝑠) = .
𝑠 𝑠

KI is the gain of the controller. Increasing integral action means increasing the gain KI. Integral control has the following
effects:
1. It leads to wider range of closed loop system responses and sometimes even unstable ones.
2. The value of KI shapes both the dynamics of the reference tracking and disturbance rejection responses.
3. It eliminates the steady state error for both reference tracking and disturbance rejection performances. Also of note is
that integral control adds an open loop at 𝑠 = 0, to the system. Hence the system type is increased thus improving the
steady state performance.
4. Higher values of KI increase the speed of the response but also increases the peak overshoot. This is desirable for
reference tracking but may be bad for disturbance rejection.

VERIFY THE SAID EFFECTS OF INTEGRAL CONTROL.

Derivative Control
𝑑{𝑒(𝑡)}
The control action is given by 𝑢(𝑡) = 𝐾𝐷
𝑑𝑡
In Laplace domain 𝑈(𝑠) = 𝐾𝐷 𝑠𝐸(𝑠). Hence the transfer function of the controller is given 𝐺𝑐 (𝑠) = 𝐾𝐷 𝑠.

KD is the d gain of the controller. Increasing the derivative action is achieved by increasing the gain KD. Derivative control is
usually associated with the controller anticipating the future direction of the error signal. It cannot be used alone because it does
not respond when the error is constant. Derivative control has the following effects:
1. It introduces an open loop zero to the system.
2. Affect the shape of both the reference tracking and disturbance rejection responses.
3. Has no effect on the steady state error for both reference tracking and disturbance rejection.
4. Derivative control increases damping ratio (directly related to damping ratio) of the system and hence damping.
Damping ratio can, thus be tuned by changing the amount of derivative control gain of the controller. Derivative
control can be used to improve system stability.
5. It can amplify noise and so impairing the system robustness and causing saturation.

VERIFY THE SAID EFFECTS OF DERIVATIVE CONTROL.

Composite Control Laws

It is possible to select any combination of any of the basic control forms to give a composite control law. The aim of combining
them is to exploit advantages from the individual forms.

Proportional plus Integral Control (PI controller)

𝐾
It is a combination of proportional and integral control. The transfer function of the controller would be 𝐺𝑐 (𝑠) = 𝐾𝑃 + 𝐼. The
𝑠
block diagram in figure 4.5 gives a model of the PI controller.

𝐾𝐼
In standard form the TF of the controller is expressed as : 𝐺𝑐 (𝑠) = 𝐾𝑃 (1 + )
𝐾𝑃 𝑠
48
 1 𝐾𝑃
= 𝐾𝑃 (1 + ), where 𝑇𝑖 =
𝑇𝑖 𝑠 𝐾𝐼
Ti is known as the integral action time formally defined as the time interval in which the part of the control signal due to
integral action increases by an amount equal to the part of the control signal due to the proportional action when the error is
unchanging.

Controller

E(s) U(s) C(s)

R(s) + +
Gp(s)
𝐾𝑃

- +

𝐾𝐼
𝑠

B(s)
H(s)

Figure 4.5

Proportional plus Derivative Control (PD controller)

It is a combination of proportional and derivative control. The transfer function of the controller would be 𝐺𝑐 (𝑠) = 𝐾𝑃 + 𝐾𝐷 𝑠.
𝐾 𝑠
In standard form the TF of the controller is expressed as : 𝐺𝑐 (𝑠) = 𝐾𝑃 (1 + 𝐷 )
𝐾𝑃
𝐾𝐷
= 𝐾𝑃 (1 + 𝑇𝑑 𝑠), where 𝑇𝑑 =
𝐾𝑃
Td is known as the derivative action time formally defined as the time interval in which the part of the control signal due to
proportional action increases by an amount equal to the part of the control signal due to the derivative action when the error is
changing at a constant rate.

Proportional plus Integral plus Derivative Control (PID controller)

𝐾
It is a combination of proportional, integral and derivative control. The transfer function would be 𝐺𝑐 (𝑠) = 𝐾𝑃 + 𝐼 + 𝐾𝐷 𝑠.
𝑠
𝐾𝐼 𝐾𝐷 𝑠
In standard form the TF of the controller is expressed as : 𝐺𝑐 (𝑠) = 𝐾𝑃 (1 + + )
𝐾𝑃 𝑠 𝐾𝑃
1
0 = 𝐾𝑃 (1 + + 𝑇𝑑 𝑠)
𝑇𝑖 𝑠

Illustration
𝐾𝐼 1
Suppose a PID controller of the form 𝐺𝑐 (𝑠) = 𝐾𝑃 + + 𝐾𝐷 𝑠 is used to control a process with a TF of 𝐺𝑝 (𝑠) = in a
𝑠 𝑠(𝐴𝑠+𝐵)
unity feedback system as shown in figure 4.6.

49
 Controller Process

E(s) 1 C(s)
R(s) + 𝐾𝐼
𝐾𝑃 + + 𝐾𝐷 𝑠
𝑠 𝑠(𝐴𝑠 + 𝐵)
-

Figure 4.6
The expressions for the output and the error can be worked out. They will come out as:
𝐾𝐷 𝑠 2 +𝐾𝑃 𝑠+𝐾𝐼
Output: 𝐶(𝑠) = 𝑅(𝑠)
𝐴𝑠 3 +𝑠 2 (𝐵+𝐾𝐷 )+𝐾𝑃 𝑠+𝐾𝐼

𝑠 2 (𝐴𝑠+𝐵)
Error: 𝐸(𝑠) = 𝑅(𝑠)
𝐴𝑠 3 +𝑠 2 (𝐵+𝐾𝐷 )+𝐾𝑃 𝑠+𝐾𝐼

A proper selection on the values of KP, KI and KD can be made to give a desired system performance. For example values can
be selected to make the system exhibit an oscillatory behavior with an acceptable damping ratio. The selection process needs to
be carefully done to avoid instability of the system.

Another observation we may make is on the steady state performance. When there is no controller {𝑢(𝑡) = 𝑒(𝑡)}, the
expressions of the output and error are:
1
Output: 𝐶(𝑠) = 2 𝑅(𝑠)
𝐴𝑠 +𝐵𝑠+1

𝐴𝑠 2 +𝐵
Error: 𝐸(𝑠) = 𝑅(𝑠)
𝐴𝑠 2 +𝐵𝑠+1

The steady error without control is zero for a step input, B for a ramp input and infinity for a parabolic inputs. When the
controller is introduced, the steady state error for step and ramp inputs becomes zero, that for a ramp input becomes equal to B.

Similar type of analyses can be applied to any other type of controller.

Example 4.1
1
A PD controller is used in cascade configuration to control a process with a TF 𝐺𝑝 (𝑠) = 2 in a unity feedback
𝑠 +𝑏𝑠+𝑎
configuration. Obtain expressions for the damping ratio and natural frequency after the controller is introduced.

SOLUTION
The TF of the controller can be written as 𝐺𝑐 (𝑠) = 𝐾𝑃 + 𝐾𝐷 𝑠
1
(𝐾𝑃 +𝐾𝐷 𝑠)( 2 )
𝑠 +𝑏𝑠+𝑎
The closed loop TF of the system after introduction of the controller is given by 𝐺 = 1
1+(𝐾𝑃 +𝐾𝐷 𝑠)( 2 )
𝑠 +𝑏𝑠+𝑎

(𝐾𝑃 +𝐾𝐷 𝑠) (𝐾𝑃 +𝐾𝐷 𝑠)

The TF simplifies to 𝐺 = (𝑠2 =
+𝑏𝑠+𝑎)+(𝐾𝑃 +𝐾𝐷 𝑠) 𝑠 2 +(𝑏+𝐾𝐷 )𝑠+(𝑎+𝐾𝑃 )

Denominator shows a second order system response would be obtained. Hence natural frequency and damping ratio are
respectively given by:
𝜔𝑛 2 = 𝑎 + 𝐾𝑃

2𝜁𝜔𝑛 = 𝑏 + 𝐾𝐷
𝑏+𝐾𝐷
Hence 𝜔𝑛 = √𝑎 + 𝐾𝑃 and 𝜁 =
2√𝑎+𝐾𝑃

50
It can be observed that values of controller gain terms cam=n be selected to meet particular natural frequency and damping
ratio.

Example 4.2
A liquid level process control system is shown in figure 4.7(a). The system parameters are:
𝐴 = 2𝑚2 𝑅𝑓 = 15𝑠/𝑚2 𝐻1 = 1𝑉/𝑚
3
𝐾𝑣 = 0.1𝑚 /𝑠𝑉 𝐾1 = 1

𝐻𝑎 (𝑠) 𝑅𝑓
The TF of the tank is = . PROVE THIS AS EXERCISE.
𝑉1 (𝑠) 1+𝐴𝑅𝑓 𝑠

a) What are the values of 𝑇𝑖 and 𝜁 when 𝜔𝑛 = 0.1𝑟𝑎𝑑/𝑠?

b) Find the expression for the time response of the system when there is a step change in ℎ𝑑 (𝑡) from 0m to 4m.

BURNS FIGURE 4.26

Figure 4.7(a)

BURNS FIGURE 4.27

Figure 4.7(b)

SOLUTION
a) The block diagram for the system is shown in figure 4.7 (b).
𝐻𝑎 (𝑠) 𝐾1 𝐾𝑣 𝑅𝑓 (1+𝑇𝑖 𝑠)
The closed loop TF of the system works out to be =
𝐻𝑑 (𝑠) (𝐴𝑅𝑓 𝑇𝑖 )𝑠 2 +(1+𝐾1 𝐾𝑣 𝑅𝑓 𝐻1 )𝑠+𝐾1 𝐾𝑣 𝑅𝑓 𝐻1

(1+𝑇𝑖 𝑠)
𝐻𝑎 (𝑠) ( )
𝐻1
In standard form = 𝐴𝑇𝑖
𝐻𝑑 (𝑠) 1
( )𝑠 2 +( +1)𝑠+1
𝐾 1 𝐾𝑣 𝐻 1 𝐾1 𝐾𝑣 𝑅 𝑓 𝐻 1

1 𝐴𝑇𝑖
From the TF, = . Putting necessary substitutions gives 𝑇𝑖 = 5 𝑠𝑒𝑐𝑜𝑛𝑑𝑠.
𝜔𝑛 2 𝐾1 𝐾𝑣 𝐻1

2𝜁 1
For damping ratio, = + 1. Putting necessary substitutions gives 𝜁 = 0.417.
𝜔𝑛 𝐾1 𝐾𝑣 𝑅𝑓 𝐻1

𝐻𝑎 (𝑠) (1+5𝑠) 0.01(1+5𝑠)

b) Inserting values into the closed loop TF gives the TF as = = .
𝐻𝑑 (𝑠) 100𝑠 2 +8.34𝑠+1 𝑠 2 +0.0834𝑠+0.01

51
 (1+5𝑠) 4 4 (−4𝑠−0.1336)
For step input of height 4m, the output is 𝐻𝑎 (𝑠) = [ ] = + (𝑠+0.0417)2
100𝑠 2 +8.34𝑠+1 𝑠 𝑠 +0.09092

In time domain, the response is ℎ𝑎 (𝑡) = 4[1 − 𝑒 −0.0417𝑡 (𝑐𝑜𝑠 0.0909𝑡 − 𝑠𝑖𝑛 0.0909𝑡)]

Physical Interpretation of PID Control Laws

1. Proportional control pushes on the system with a “force” proportional to the error, i.e, if the output is a long way from
the reference, the push is hard and if it is close, the push is soft. The problem with P control is that if there is no
damping it can create oscillations.
2. Derivative control effectively adds or subtracts damping. The force is proportional to the rate at which the gap between
the output and the reference is being closed.
3. Integral control is mainly used to eliminate steady state errors. The force is proportional to the area between the output
curve and the reference curve. If there is a steady state error, the area adds up to create a force to reduce it.

Figure 4.8 shows illustrations.

Figure 4.8 (OLD NOTES BOOK 3)

Background to PID Tuning Methods

Design of a controller involves selecting 𝐺𝑐 (𝑠)to make the system follow the reference set point as well as to reject disturbances
whilst ensuring the system is stable. Certain performance criteria (response speed, peak overshoot, steady state error, etc.) need
to be satisfied when the controller is introduced.

For the case of PID controller, the design process is basically the selection of values for KP, KI and KD (or KP, Ti and Td) that
will make the controller bring up the required performance specifications in the response of the system being controlled. The
steps that are followed when designing a PID controller are:
1. Fully understanding the capability of the PID controller.
2. Discovering the strengths and weaknesses of the system’s existing response performance.
3. Deciding on the required type of closed loop system performance.
4. Choosing a suitable method for selecting the PID controller coefficients. Choosing of the controller coefficients is
referred to as tuning the controller.

Step 1: Capabilities of a PID Controller

This is knowledge of the effects of the P, I and D terms as has already been covered in earlier sections.

52
Step 2: Discovering Strengths and Weaknesses
In practice, this is achieved by conducting some simple time-domain experiments, often step response tests. Knowledge of first
and second order time responses can then be applied to the results to get performance measures (steady state error, settling time,
peak overshoot, rise time, etc.) for the system, after which the system can be identified as meeting the requirements or not.

Step 3: Type of required Closed Loop Performance

This step consists of two parts: general considerations which will dictate the structure of the appropriate controller and the
design specifications that the controller has to achieve.

General consideration is about type of steady state accuracy required in response to reference and to disturbance input signals.
If steady state accuracy is required, a PI or PID controller is chosen. If steady state accuracy is not important, a P or PD
controller may be chosen.

The typical list of design specifications is:

 Closed loop time constant
 Rise time
 Peak overshoot
 Damping ratio
 Natural undamped frequency
 Closed loop pole positions
 Settling time
 Steady state error
 Disturbance peak value
 Disturbance settling time

Some of the specifications can be given typical values, e.g., closed loop system damping is often specified in the range 0.6-.75.
Other control system parameters will depend on the system being controlled, and engineering experience. Sometimes one of the
specifications is especially important and is given a critical value which must be achieved. Table 4.2 shows examples of how
control strategies are chosen for the particular systems.

Table 4.2
System Specification Design Solutions
Ship autopilot control
𝑀𝑝 = 0% Design for critical damping 𝜁 = 0
No overshoot to step demand in heading, but good
𝑒𝑠𝑠 = 0 Use PI control
speed of response
Liquid tank level system
𝜏𝐶𝐿 ≪ 𝜏𝑂𝐿 Design for first order response type
Good speed of response, but steady state accuracy
𝑒𝑠𝑠 small Use P control
not required
Gas turbine temperature control
Disturbance rejection design Underdamped response design for
Load changes frequent.
𝑒𝐷𝑠𝑠 = 0 speed of response.
Steady state accuracy essential.
𝐷𝑡𝑠 < 5 𝑚𝑖𝑛𝑢𝑡𝑒𝑠 Use PI control
Fast disturbance rejection needed within 5 minutes

Another further skill needed on this step is being able to manipulate a given control specification into a form which can then be
used in the PID tuning method (relationship of control gains to, for example, parameters in standard first and second order
parameters).

Step 4: Choosing PID Controller Coefficients

This stage involves choosing the PID controller coefficients KP, KI and KD (or KP, Ti and Td). If a PID controller is expressed in
terms of KP, KI and KD, it is said to be in the text book or gain format. If it is expressed in terms of KP, Ti and Td, it is said to be
in the industrial or time constant fomat. The text book format is more appropriate to use as compared to the standard format.
The following methods can be used to tune the controller.
A. M ANUAL TUNING FOR P AND PI CONTROLLERS
This is a trial and error method where the controller (P or PI) is implemented and the values of the terms tuned as
follows:
i. The proportional gain, KP is increased until speed of response is acceptable.

53
 ii. If integral control is required to eliminate steady state errors, the integral gain, KI is increased until acceptable
settling time is achieved.
iii. KP and KI are then balanced to compromise overshoot and response speed.
The manual method disturbs the process for long periods without production.

B. KNOWING PROCESS M ODEL

The closed loop TF (for tracking or disturbance responses as necessary) is obtained. The TF will be in terms of KP, KI
and/or KD as necessary. The TF is expressed in a form that relates it to the required specifications. Suitable controller
coefficient values are then obtained by comparing the actual specification expressions to the required specification
values. It is important to note that fixing a controller term for a particular specification automatically fixes other
specifications which could result in some undesirable effects.

C. POLE PLACEMENT
The performance of a system depends on the position of its poles. Sometimes, the closed loop poles may be required to
be at particular positions in the s-plane to give a particular performance. In such cases, the required CLCE is
determined. Then the actual closed loop TF of the system (in terms of KP, KI and/or KD) is determined. The
coefficients of the required CLCE are compared to those of the actual CLCE so as to select suitable values of the
control terms.

54