Prof. M. M.

Solanki
Electronics Engineering Dept.
B.V.M. Engineering College
 Basics of control design -
Proportional, Integral and Derivative actions.

2
 Consider the closed-loop control system -
𝐷(𝑠)

𝑅𝑠 𝐸 𝑠 + 𝑌𝑠
+- 𝐺𝑐(𝑠) + 𝐺(𝑠)

++
𝑁(𝑠)

We have seen that feedback control is capable of 𝑅𝑠 is the reference input

 Improving the stability 𝑌𝑠 is the controlled output
 Meeting the performance specifications 𝐷𝑠 is the disturbance input
 Decreasing the sensitivity to parametric variations 𝑁𝑠 is the measurement noise
 Improving disturbance rejection
 Attenuating the measurement noise We especially care about 𝐸 𝑠 , the difference
between the reference input and the measured
output

Control Engineering 3
 We studied the responses of typical first order and second order closed-loop systems to different inputs.

 First order system :

1
𝐺1 𝑠 =
1 + 𝑇𝑠
 Second order system :

𝜔𝑛2
𝐺2 𝑠 =
𝑠2 + 2𝜁𝜔𝑛𝑠 + 𝜔𝑛2

 We identified the features that captured the nature of these responses. For the second order system -

𝜋−𝜃 𝜋−cos −1 𝜁 1+0.7𝜁

 Rise time 𝑡 𝑟 = =  Delay time 𝑡𝑑 =
𝜔𝑑 𝜔𝑛 1−𝜁 2 𝜔𝑛
𝜁𝜋
−
1−𝜁 2 4
 Percentage Overshoot 𝑀 𝑝 = 100𝑒 %  Settling time 𝑡𝑠 = 4𝑟 = (2% tolerance)
𝜁𝜔 𝑛

4
 We typically specify the performance of a control system in terms of these metrics.

 Suppose that the plant has order that is relatively high and the closed-loop system has the
following pole zero distribution in the s-plane.

𝜔
What is the contribution of these poles and zeros to the
response of the system?

 The poles closest to the imaginary axis have the most

dominant contribution and are called the dominant poles.
𝜎
 Let us begin by analysing the system when there is an extra
pole or zero along with the dominant poles.

5
Consider a typical second order closed-loop system

𝜔𝑛2
𝐶 𝑠 =
𝑠2 + 2𝜁𝜔𝑛𝑠 + 𝜔𝑛2

For simplicity, let 𝜁 = 0.5 and 𝜔𝑛 = 1. Let us start by adding a zero at s= −𝑎.

𝑠
To maintain the DC gain at one we add + 1 , to obtain
𝑎

𝑠 +1
𝑎 1 1 𝑠
𝐶1 𝑠 = = +
𝑠2 + 𝑠 + 1 𝑠2 + 𝑠 + 1 𝑎 𝑠2 + 𝑠 + 1

Let 𝑓 −1 𝐶 𝑠 = 𝑐 𝑡 and 𝑓 −1 𝐶1 𝑠 = 𝑐1 𝑡 . Then

1
𝑐1 𝑡 = 𝑐(𝑡) + 𝑐 𝑡
𝑎

6
 1
𝑌 𝑠 = 𝐶1 𝑠
𝑠

= 𝐶 (𝑠) + 𝑠 𝐶 (𝑠)
𝑠 𝑎 𝑠

𝑠
= 𝑌1 𝑠 + 𝑌1 𝑠
𝑎

Let 𝑓 −1 𝑌 𝑠 = 𝑦 𝑡 and 𝑓 −1 𝑌1 𝑠 = 𝑦1 𝑡 .

1
⇒ 𝑦 𝑡 = 𝑦1 𝑡 + 𝑦1 𝑡
𝑎

With 𝑦2 𝑡 = 𝑦1 𝑡 , we have
1
𝑦 𝑡 = 𝑦1 𝑡 + 𝑦2 𝑡
𝑎

Let us look at the response of the system for different values of 𝑎.

7
We see the following major effects

 Increased overshoot
 No significant change in the settling time
 Faster response
 As 𝑎 → ∞, the contribution of the zero to the system response decreases and we get back
the original behaviour.

8
Let us now place the zero at s = 𝑎. To maintain the DC gain at one we add 1 − 𝑠 , to obtain
𝑎

1 − 𝑠𝑎 1 1 1
𝐶1 𝑠 = 2 = −
𝑠 + 𝑠 + 1 𝑠2 + 𝑠 + 1 𝑎 𝑠2 + 𝑠 + 1

Let 𝑓 −1 𝐶 𝑠 = 𝑐 𝑡 and 𝑓 −1 𝐶1 𝑠 = 𝑐1 𝑡 . Then

1
𝑐1 𝑡 = c t − 𝑐(𝑡)
𝑎
The step response of the system is
1
𝑌 𝑠 =𝐶 𝑠
𝑠

= 𝐶1(𝑠) − 𝑠 𝐶1(𝑠)
𝑠 𝑎 𝑠

𝑠
= 𝑌1 𝑠 − 𝑌1 𝑠
𝑎

1
⇒ 𝑦 𝑡 = 𝑦1 𝑡 − 𝑦1 (𝑡)
𝑎

9
 The interesting feature is the undershoot in the transient
response for small 𝑎.

 The undershoot leads to sluggish response with the

increase in rise time and delay time.

 Poles and zeros in the RHP are called non-minimum

phase poles and zeros

 The term ‘non-minimum phase’ will be dealt with in more

detail when we take up design in the frequency domain.

10
 Effects of adding a pole in the Left Half s-Plane.
To the second order system considered so far, let us add a pole at 𝑠 = −𝑎.

1 1
𝐶 𝑠 = 𝑠
(𝑎 + 1) 𝑠2 + 𝑠 + 1

Step responses of the system for different 𝑎.

 As 𝑎 → ∞, the contribution of the pole to the system response decreases and we get back the original behaviour.

 In the design process, as a thumb rule, we roughly take 𝑎 ≫ 5 × the real part of the dominant poles.

11
  Dominant Poles :- The closed-loop poles that have
dominant contribution on the transient response of the
system.

 It is often the case that these dominant poles occur in the

form complex conjugate poles.

 Higher order systems are generally adjusted such that that

there exists a pair of dominant complex conjugate poles.

 It is desirable to have the real parts of the other poles and

zeros at least five times farther away than the real parts of
the dominant poles.

Dominant poles

12
 The addition of a pole to an open-loop transfer function has the effect of pulling the Root
Locus to the right.

 Some of the closed-loop poles are nearer to the imaginary axis, which implies the relative
stability is reduced.

 We begin with a single pole system and progressively add poles to illustrate these points.

1 1 1
𝐺 𝑠 = 𝐺 𝑠 = 𝐺 𝑠 =
𝑠+2 (𝑠 + 2)(𝑠 + 4) (𝑠 + 2)(𝑠 + 4)(𝑠 + 8)
13
  The addition of a zero to an open-loop transfer function has the effect of pulling the
Root Locus to the left.

 The effect of the zero is prominent when it is close to the imaginary axis.

 We begin with a third order system and add a zero at different locations on the real
axis to illustrate these points.

1 𝑠 +7 𝑠 +3 𝑠 +1
𝐺 𝑠 = 𝐺 𝑠 = 𝐺 𝑠 = 𝐺 𝑠 =
𝑠(𝑠 + 2)(𝑠 + 5) 𝑠(𝑠 + 2)(𝑠 + 5) 𝑠(𝑠 + 2)(𝑠 + 5) 𝑠(𝑠 + 2)(𝑠 + 5)
14
  The controller acts on the error between the reference input and the measured output.

𝑅 𝑠 𝐸 𝑠 𝑌𝑠
+- 𝐾 𝐺(𝑠)

 We have already performed significant analysis with this

controller through the Root Locus, Nyquist and Bode plots. ×

 With the freedom to vary only the gain 𝐾 we are restricted to

move on the Root Locus.
×
 Works when the poles on the Root Locus meet the
specifications.

 If not, more sophisticated controllers are necessary. 1

𝐺 𝑠 =
𝑠(𝑠 + 2)(𝑠 + 5)
15
  The controller acts on the error between the reference input and the measured output.

𝑅 𝑠 𝐸 𝑠 𝑌𝑠
+- 𝐾 𝐺(𝑠)

 We have already performed significant analysis with this

controller through the Root Locus, Nyquist and Bode plots.

 In Bode plots, the gain 𝐾 provided the freedom to shift the

magnitude plot up and down.

 In general cases, this may not be sufficient.

 If not, more sophisticated controllers are necessary.

200
𝐺 𝑠 =
𝑠(𝑠 + 2)(𝑠 + 5)
16
17
 Let us take a closer look at the proportional control action. Consider a first order plant
𝐾
𝐺 𝑠 =
𝑇𝑠 + 1
Then the closed-loop transfer function is
𝐾
𝐶𝑠 =
𝑇𝑠 + 1 + 𝐾

𝑅 𝑠 𝐸 𝑠 1 𝑌𝑠
+- 𝐾
𝑇𝑠 + 1

1 1 1 𝑇𝑠 + 1 1 𝑇𝑠 + 1 1
𝐸 𝑠 = 𝑅 𝑠 = = 𝑒𝑠𝑠 = lim 𝑒 𝑡 = lim 𝑠𝐸 𝑠 = lim =
1+𝐺 𝑠 1+ 𝐾 𝑠 𝑇𝑠 + 1 + 𝐾 𝑠 𝑡 →∞ 𝑠→0 𝑠→0 𝑇𝑠 + 1 + 𝐾 𝐾+1
𝑇𝑠 + 1
𝑇
 Proportional controller has improved the time constant from 𝑇 to . However, there is steady state error.
(1+𝑘)
 The steady state error can be reduced by choosing a large 𝐾, but high gain has the tendency to destabilize the
higher order plants.

18
 Let us choose a controller that acts not on the current error but on the accumulated error.

𝑅 𝑠 𝐸 𝑠 𝐾 1 𝑌𝑠
+-
𝑠 𝑇𝑠 + 1

The output of the controller

𝐾 𝑡
𝑌𝑐 𝑠 = 𝐸 𝑠 ⟹ 𝑦𝑐 𝑡 = 𝐾 ∫ 𝑒 𝑡 𝑑𝑡
𝑠 −∞

1 𝑠(𝑇𝑠 + 1) 1 𝑠2(𝑇𝑠 + 1) 1
𝐸 𝑠 = 𝑅 𝑠 = 𝑒𝑠𝑠 = lim 𝑒 𝑡 = lim 𝑠𝐸 𝑠 = lim =0
1 + 𝐺 𝑠 𝐻(𝑠) 𝑠 𝑇𝑠 + 1 + 𝐾 𝑠 𝑡 →∞ 𝑠→0 𝑠→0 𝑠(𝑇𝑠 + 1) + 𝐾 𝑠

 With the integrator as the controller, the steady state error is now zero.

 Note that the order of the system has increased from one to two.

 As we have seen for higher order systems, addition of an extra pole may lead to instability.
19
• Comment on the stability of the system given below.

𝑅𝑠 𝐸 𝑠 𝐾 1 𝑌𝑠
+-
𝑠 𝑠(𝐽𝑠 + 𝑏)

20
Consider the second order plant
1
𝐺 𝑠 =
𝑠(𝐽𝑠 + 𝑏)

Models a rotational element that consists of moment of inertia 𝐽 and viscous friction with coefficient 𝑏.

𝐷 𝑠

𝑅𝑠 𝐸 𝑠 + 1 𝑌𝑠
+- 𝐾 +
𝑇 𝑠 𝑠(𝐽𝑠 + 𝑏)

The controller has access to the input torque 𝑇 𝑠 . An external disturbance 𝐷 𝑠 acts on the system.

21
 𝑇
𝐷 𝑠 =
𝑠
𝑅𝑠 𝐸 𝑠 + 1 𝑌𝑠
+- 𝐾 +
𝑇 𝑠 𝑠(𝐽𝑠 + 𝑏)

Setting 𝑅 𝑠 = 0, let us obtain the transfer function from 𝐷 𝑠 to 𝑌 𝑠 .

𝑌(𝑠) 1
=
𝐷(𝑠) 𝐽𝑠2 + 𝑏𝑠 + 𝐾
and
𝐸(𝑠) 𝑌𝑠 1
=− =−
𝐷(𝑠) 𝐷𝑠 𝐽𝑠2 + 𝑏𝑠 + 𝐾
The steady state error
−𝑠 𝑇 −𝑇
𝑒𝑠𝑠 = lim 𝑒 𝑡 = lim 𝑠𝐸 𝑠 = lim =
𝑡 →∞ 𝑠→0 𝑠→0 𝐽𝑠 2 + 𝑏𝑠 + 𝐾 𝑠 𝐾
The steady state error can be reduced by increasing K. However, the system becomes more oscillatory. For
higher order plant can lead to instability.
22
Let us look at how Proportional + Integral control fares in this situation.
𝑇
𝐷 𝑠 =
𝑠

𝑅 𝑠 𝐸 𝑠 1 + 1 𝑌𝑠
+- 𝐾(1 + ) +
𝑇𝑖 𝑠 𝑠(𝐽𝑠 + 𝑏)

The transfer function from the disturbance signal to the output is

𝑌(𝑠) 𝑠
=
𝐷(𝑠) 𝐽𝑠 3 + 𝑏𝑠2 + 𝐾𝑠 + 𝐾
𝑇 𝑖

With 𝑅 𝑠 = 0,
𝐸(𝑠) 𝑌𝑠 𝑠
=− =−
𝐷(𝑠) 𝐷 𝑠 𝐾
𝐽𝑠 3 + 𝑏𝑠2 + 𝐾𝑠 +
𝑇𝑖

23
 𝐾
𝐽𝑠3 + 𝑏𝑠 2 + 𝐾𝑠 + =0
𝑇𝑖
have negative real parts.

−𝑠2 𝑇
𝑒𝑠𝑠 = lim 𝑒 𝑡 = lim 𝑠𝐸 𝑠 = lim =0
𝑡 →∞ 𝑠→0 𝑠→0 3 𝐾 𝑠
𝐽𝑠 + 𝑏𝑠 2 + 𝐾𝑠 +
𝑇𝑖

The proportional + integral control action eliminated the steady state error.

Exercise II
Reason why integral control alone cannot be used in this case. The answer may be found in the previous exercise.

In this case, both the proportional and integral terms are necessary. The proportional term ensures stability
while the Integral terms eliminates steady state error.
24
Consider the plant
1
𝐺 𝑠 = ,
𝐽𝑠2
a rotational element with moment of inertial 𝐽.

Consider the proportional control of this plant

𝑅𝑠 𝐸 𝑠 𝑌𝑠
1
+- 𝐾
𝐽𝑠2

The closed – loop transfer function is

𝐶(𝑠) 𝐾
=
𝑅(𝑠) 𝐽 𝑠 2 +𝐾

The poles of the characteristic equation are on the imaginary axis. Hence the system
has pure oscillations to a step input.
These oscillations can be damped if we incorporate a derivative term in the controller.
25
The controller has the following structure

𝐺𝑐 𝑠 = 𝐾 1 + 𝑇𝑑 𝑠

The output from the controller𝑌𝑐 (𝑠) is

𝑌𝑐 𝑠 = 𝐺𝑐 𝑠 𝐸 𝑠

= 𝐾𝐸 𝑠 + 𝐾𝑇 𝑑 𝑠𝐸 𝑠

⇒ 𝑦𝑐 𝑡 = 𝐾𝑒 𝑡 + 𝐾𝑇 𝑑 𝑒 𝑡

 The derivative term acts on the rate of change of the error signal. The output is proportional
to the rate of change of the error signal.

 The derivative term anticipates the large overshoot, when the rate of change of the error is
high, and takes corrective action.

 The effect can be noticed in the improved damping in the system.

26
 𝑅 𝑠 𝐸 𝑠 1 𝑌𝑠
+- 𝐾(1 + 𝑇𝑑 𝑠)
𝐽𝑠 2

The output of the system 𝑌 𝑠 is

𝐾 1 + 𝑇𝑑 𝑠
𝑌 𝑠 = 2 𝑅(𝑠)
𝐽𝑠 + 𝐾𝑇 𝑑 𝑠 +
𝐾
For positive values of 𝐾, 𝑇𝑑 and 𝐽, is always stable. Note that the damping term in the characteristic
equation is due to the derivative control.

27
Let us look at the effect of Proportional + Derivative + Integral control on a second order unstable plant.
Consider the second order plant
1
𝐺 𝑠 =
(𝑠2−𝑏)
with 𝑏 > 0. Clearly the system is unstable with poles at 𝑠 = ± 𝑏.

Let us begin with a proportional controller.

𝑅 𝑠 𝐸 𝑠 1 𝑌𝑠
+- 𝐾𝑝
(𝑠2−𝑏)

Note that with just a proportional controller 𝐾𝑝, the closed loop transfer function is
𝐾𝑝
𝑌 𝑠 =
𝑠2−𝑏+𝐾𝑝
This system cannot be stabilized using just proportional control. At best marginal stability can be achieved.

28
 𝑅 𝑠 𝐸 𝑠 1 𝑌𝑠
+- (𝐾𝑝 +𝐾𝑑 𝑠)
(𝑠2−𝑏)

The closed-loop transfer function is

𝑌(𝑠) (𝐾𝑝+𝐾𝑑𝑠)
=
𝑅(𝑠) 𝑠2 + 𝐾𝑑𝑠 + 𝐾𝑝 − 𝑏

For 𝐾𝑑 > 0 and 𝐾𝑝 > 𝑏, the system is stable and arbitrary pole-placement is possible. The
closed-loop poles can be now be placed in the s-plane to meet any given specifications.

29
 𝑠(𝐾𝑝+𝐾𝑑𝑠) 1 𝐾𝑝
lim 𝑦 𝑡 = lim 𝑠𝑌 𝑠 = lim = ≠1
𝑡→∞ 𝑠→0 𝑠→0 (𝑠2+𝐾𝑑𝑠+𝐾𝑝−𝑏) 𝑠 𝐾 𝑝 −𝑏

There is steady state error in the system.

Let us add an integral term to the PD controller.

𝑅 𝑠 𝐸 𝑠 𝐾𝐼 1 𝑌𝑠
+- 𝐾𝑝 + 𝐾𝑑 𝑠 +
𝑠 (𝑠2−𝑏)

The DC gain from 𝑅 𝑠 to 𝑌(𝑠) is

𝑠(𝐾𝑑 1
lim 𝑦 𝑡 = lim 𝑠𝑌 𝑠 = lim 𝑠 2 +𝐾 𝑠+𝐾 ) =1
𝑡→∞ 𝑠→0 𝑠→0 (𝑠 3 +𝐾𝑑 𝑠𝑝2 +(𝐾𝑝𝐼 −𝑏 )𝑠+𝐾𝐼 ) 𝑠

We have perfect tracking. Moreover we know that an integral controller rejects constant
disturbance inputs.
30
 In the previous example, using PID control, we could

 Stabilize an unstable second order system.

 Through the multiple degrees of freedom offered by PID

control, we can place the poles arbitrarily.

 The system has perfect tracking of step signals because of

integral control.

 The system rejects constant disturbance inputs.

31