Control Systems ELEC3114

Root Locus Technique

Dr. Arash Khatamianfar

School of Electrical Engineering and Telecommunications

T2 2021
Content
• Root locus concept Mainly from
• Root locus procedure
Nise Textbook:
• Compensator design using Root Locus technique Chapter 8 and 9
• Lead-lag controller
And Dorf textbook:
Chapter 7

And some other

resources

Page 1
Root locus concept
Reference
• A simple LTI closed-loop system is characterized by Input Output
the transfer functions in the forward path and 𝑅(𝑠) 𝑌(𝑠)
feedback path with some adjustable parameters.

• The roots of the characteristic equation of the closed-loop

system determine the behaviour of the system response 𝑌(𝑠).
• As you can see, in the closed-loop system, the characteristic
equation depends on the variable parameter 𝐾 for 𝐾 ≥ 0.
𝑅(𝑠) 𝑌(𝑠)
𝐾𝐺(𝑠)
𝑇(𝑠) = ⟹ 1 + 𝐾𝐺 𝑠 𝐻 𝑠 = 0
1 + 𝐾𝐺 𝑠 𝐻(𝑠)
Closed-loop transfer function
characteristic equation

• The closed-loop poles must satisfy the characteristic equation,

and their location on 𝑠-plane changes as 𝐾 varies.
• The open-loop poles (which are the same as loop transfer
function poles) are the roots of the denominator of 𝐾𝐺 𝑠 𝐻 𝑠 .

Page 2
Root locus concept (2)
Reference
• Since the Laplace variable 𝑠 = 𝜎 + 𝑗𝜔 is a Input Output
complex variable, the closed-loop characteristic 𝑅(𝑠) 𝑌(𝑠)
equation is a complex function which can be
described in polar form:
1 + 𝐾𝐺 𝑠 𝐻 𝑠 = 0 ⟹ 𝐾𝐺 𝑠 𝐻 𝑠 = −1 + 𝑗0

⟹ 𝐾𝐺 𝑠 𝐻 𝑠 ∠𝐾𝐺 𝑠 𝐻 𝑠 = −1

𝑅(𝑠) 𝑌(𝑠)
Therefore: 𝐾𝐺 𝑠 𝐻 𝑠 =1
Closed-loop transfer function

∠𝐾𝐺 𝑠 𝐻 𝑠 = 2𝑘 + 1 180°

• Any value of 𝑠 satisfying the characteristic equation is a closed-

loop pole which causes the magnitude of the open loop transfer
function to be unity 𝐾𝐺 𝑠 𝐻 𝑠 = 1
• Any closed-loop pole causes the angle of the open loop transfer
function to be an odd multiple of 180°. Page 3
Root locus concept (3)
• The location of the closed-loop poles as the parameter 𝐾 varies is
called locus of the roots of the characteristic equation, or in short
the root locus.
• Let’s consider a simple overdamped 2nd-order
system with unity feedback: R(s) + E(s)
𝐾
U(s)
𝐺(𝑠)
Y(s)
− Uc (s)
1 𝑌 𝑠 𝐾𝐺 𝑠
𝐺 𝑠 = ⟹ 𝑇𝑠 = =
𝑠(𝑠 + 10) 𝑅 𝑠 1 + 𝐾𝐺 𝑠
𝐾 Open-loop
𝑇(𝑠) = 2 ⟹ 𝑠1,2 = −5 ± 25 − 𝐾
𝑠 + 10𝑠 + 𝐾 poles

The root locus is the

path of the closed-
loop poles traced
out in the 𝑠-plane as
a system parameter
varies from zero to
infinity.

Page 4
Vector Representation of Complex Numbers
• Given the complex numbers 𝑧 = 𝑥 + 𝑗𝑦 = 𝑟∠𝜙 = 𝑟𝑒 𝑗𝜙 , 𝑧1 = 𝑥1 +
𝑗𝑦1 = 𝑟1 ∠𝜙1 = 𝑟1 𝑒 𝑗𝜙1 , and 𝑧2 = 𝑥2 + 𝑗𝑦2 = 𝑟2 ∠𝜙2 = 𝑟2 𝑒 𝑗𝜙2 , some
basic properties to be used for vector analysis are as follows:

• Addition: • Reciprocal:
𝑧1 + 𝑧2 = 𝑥1 + 𝑥2 + 𝑗 𝑦1 + 𝑦2 1 1 1 −𝑗𝜙
= ∠ −𝜙 = 𝑒
𝑧 𝑟 𝑟
• Subtraction: • Square Root:
𝑧1 − 𝑧2 = 𝑥1 − 𝑥2 + 𝑗 𝑦1 − 𝑦2 𝜙 𝜙
𝑗2
𝑧 = 𝑟∠ = 𝑟𝑒
2
• Multiplication: • Complex Conjugate:
𝑧1 𝑧2 = 𝑟1 𝑟2 ∠ 𝜙1 + 𝜙2 𝑧 ∗ = 𝑥 − 𝑗𝑦 = 𝑟∠ − 𝜙 = 𝑟𝑒 −𝑗𝜙
= 𝑟1 𝑟2 𝑒 𝑗(𝜙1+𝜙2) 1
• Division: = −𝑗 and 𝑧𝑧 ∗ = 𝑟 2 = 𝑧 2
𝑗
𝑧1 𝑟1 𝑟1 𝑗(𝜙 −𝜙 )
= ∠ 𝜙1 − 𝜙2 = 𝑒 1 2 • Also: 𝑒 ±𝑗90° = ±𝑗,
𝑧2 𝑟2 𝑟2 𝑒 ±𝑗180° = −1

Page 5
Vector Representation of Complex Numbers (2)
• Any complex number in the s-plane can be represented
in Cartesian (rectangular) form or polar form (which is a
vector form): 𝑠
𝑗𝜔1
𝑀 = 𝜎12 + 𝜔12
𝑠 = 𝜎1 + 𝑗𝜔1 = 𝑀∠𝜃 ⟹
𝜔1 𝜎1
𝜃 = tan−1
𝜎1
where 𝑀 is the magnitude and 𝜃 is the phase angle (or just angle).
• Evaluating a simple complex function like 𝐹(𝑠) = (𝑠 + 𝑎) at any particular
point like 𝑠 = 𝜎 + 𝑗𝜔 can also be represented in vector form:

𝐹 𝑠 𝜎1 + 𝑎 2 + 𝜔12
=
𝐹 𝑠 ቚ = 𝜎1 + 𝑎 + 𝑗𝜔1 ⟹ 𝜔1
𝑠=𝜎1 +𝑗𝜔1 −1
∠𝐹(𝑠) = tan
𝜎1 + 𝑎
𝐹(𝑠) 𝑠
𝑗𝜔1
𝑗𝜔1 Shift the vector 𝑎 units to the left
∠(𝑠 + 𝑎)
𝜎1
𝜎1 + 𝑎
The magnitude and the angle
of 𝐹 𝑠 remain the same.
Page 6
Vector Representation of Complex Numbers (3)
• As you notice, 𝐹(𝑠) = (𝑠 + 𝑎) has a root at −𝑎. So we can Positive
describe the value of the function 𝐹(𝑠) at any point 𝑠 = 𝜎1 + 𝑠
angle
(+)
𝑗𝜔1 by finding the magnitude and angle of the vector drawn 𝑗𝜔1
from the root of the function to the point 𝑠. ∠(𝑠 + 𝑎)
𝜎1
• Consider a loop transfer function given in zero-pole-gain form
(first-order complex factors)
𝐾 𝑠 + 𝑧1 𝑠 + 𝑧2 … Negative
𝐾𝐺 𝑠 𝐻 𝑠 = angle
𝑠 + 𝑝1 𝑠 + 𝑝2 … (−)

• We can evaluate the loop gain 𝐾𝐺 𝑠 𝐻 𝑠 at any complex value 𝑠 = 𝜎1 + 𝑗𝜔1 by

• Multiplying the magnitude of the zero vectors divided by the magnitude of the pole
vectors for 𝐾𝐺 𝑠 𝐻(𝑠) .
• Adding the angles of the zero vectors (zero angles) and subtracting them from the
sum of the pole angles for ∠𝐾𝐺 𝑠 𝐻 𝑠
• 𝐾 is a constant gain and does not affect the angle.

𝑚 𝑛
ς𝑚
𝑖=1 𝑠 + 𝑧𝑖
𝐾𝐺 𝑠 𝐻 𝑠 =𝐾 𝑛 ∠𝐺 𝑠 𝐻 𝑠 = ෍ ∠(𝑠 + 𝑧𝑖 ) − ෍ ∠(𝑠 + 𝑝𝑖 )
ς𝑖=1 𝑠 + 𝑝𝑖 𝑖=1 𝑖=1
Page 7
Example 1
• Evaluate 𝐺(𝑠) at the point 𝑠 = −3 + 𝑗4.
• (Worked solutions of examples will be given either in lecture or uploaded later)

𝑠+1
𝐺 𝑠 =
𝑠(𝑠 + 2)

Answer:
𝐺 𝑠 ቚ = 0.217∠ − 114.3°
𝑠=−3+𝑗4

Page 8
Practice Problem 1
• Find 𝐹 𝑠 at the point 𝑠 = −7 + 𝑗9 by directly substituting the point into 𝐹 𝑠 and also by
using vector analysis.
𝑠 + 2 (𝑠 + 4)
𝐹 𝑠 =
𝑠(𝑠 + 3)(𝑠 + 6)

Answer:
𝐹 𝑠 ቚ = 0.034 + 𝑗0.09 = 0.096∠ − 110.7°
𝑠=−7+𝑗9

Page 9
Example 2
• Evaluate 𝐺(𝑠) at the location of the closed-loop poles in a unity feedback system when
𝐾 = 50.
• (Worked solutions of examples will be given either in lecture or uploaded later)

𝐾
𝐺 𝑠 =
𝑠(𝑠 + 10)

Answer:
𝐺 𝑠 ቚ = 1∠ − 180°
𝑠=−5+𝑗5

Page 10
Properties of the Root Locus
𝑅(𝑠) + 𝑌(𝑠)
• Recall that for any point in 𝑠-plane to be a closed- 𝐾 𝐺(𝑠)
−
loop pole, it should be a root of the characteristic
equation of the closed-loop transfer function.
• For a non-unity feedback system with a variable gain 𝐻(𝑠)
𝐾 as the controller, a value of 𝑠 could be a closed-loop
pole if: 𝐾𝐺(𝑠)
𝑇(𝑠) =
1 + 𝐾𝐺 𝑠 𝐻(𝑠)
1 + 𝐾𝐺 𝑠 𝐻 𝑠 = 0 ⟹ 𝐾𝐺 𝑠 𝐻 𝑠 ∠𝐾𝐺 𝑠 𝐻 𝑠 = −1

1
Therefore: 𝐾𝐺 𝑠 𝐻 𝑠 =1 ⇒ 𝐾= ቚ Gain
𝐺 𝑠 𝐻 𝑠 𝑠=𝐶.𝐿 𝑝𝑜𝑙𝑒 criterion

∠𝐾𝐺 𝑠 𝐻 𝑠 = 2𝑘 + 1 180° ⇒ ∠𝐺 𝑠 𝐻 𝑠 ቚ = 2𝑘 + 1 180° Angle

𝑠=𝐶.𝐿 𝑝𝑜𝑙𝑒 criterion

If a particular point causes the angle of the loop gain 𝐺 𝑠 𝐻 𝑠 to

be 2𝑘 + 1 180°, there exists a gain 𝐾 such that the point becomes
a closed-loop pole, i.e., the point is located on the root locus.
Page 11
An example
• Consider the unity feedback system below (𝐻 𝑠 = 1):

𝐾(𝑠 + 3)(𝑠 + 4)
𝐾𝐺 𝑠 𝐻 𝑠 =
(𝑠 + 1)(𝑠 + 2)
𝑌 𝑠 𝐾(𝑠 + 3)(𝑠 + 4)
𝑇 𝑠 = =
𝑅 𝑠 1 + 𝐾 𝑠 2 + 3 + 7𝐾 𝑠 + (2 + 12𝐾)

To find a gain 𝐾 such that a point like 𝑠 = −2 + 𝑗3 is one of the

closed-loop poles (lies on the root locus), the angle criteria
must be first satisfied:

∠𝐺 𝑠 𝐻 𝑠 ቚ = 𝜃𝑧1 + 𝜃𝑧2 − 𝜃𝑝1 − 𝜃𝑝2 Angle contribution

𝑠=−2+𝑗3
𝑠 = −2 + 𝑗3

3 3 3
= tan−1 + tan−1 − 90° − (180 − tan−1 )
4−2 3−2 2−1
= 56.31° + 71.57° − 90° − 108.43° = −77.55° ≠ −180°

Therefore 𝑠 = −2 + 𝑗3 cannot be a closed-loop pole 𝐿𝑧1 𝐿𝑝1

𝐿𝑧2 𝐿𝑝2
for any gain (i.e., it is not a point on the root locus). 𝜃𝑧1 𝜃𝑧2 𝜃𝑝1 𝜃𝑝2

Page 12
An example (2)
𝑗𝜔
• If we check for 𝑠 = −2 + 𝑗 2/2 in the same system, the total
angle contribution at point 𝑠 is: 𝑠 = −2 + 𝑗 2/2
𝑗 2/2
𝐿𝑧1 𝐿𝑝1 𝐿𝑝2
𝐿
∠𝐺 𝑠 𝐻 𝑠 ቚ = 𝜃𝑧1 + 𝜃𝑧2 − 𝜃𝑝1 − 𝜃𝑝2 𝜃𝑧1 𝑧2 𝜃𝑧2 𝜃𝑝1 𝜃𝑝2
𝜎
𝑠=−2+𝑗 2/2
−4 −3 −2 −1
0.707 0.707 0.707
= tan−1 + tan−1 − 90° − (180 − tan−1 )
2 1 1 𝐾(𝑠 + 3)(𝑠 + 4)
= 19.46° + 35.26° − 90° − 180° + 35.26° = −180° 𝐾𝐺 𝑠 𝐻 𝑠 =
(𝑠 + 1)(𝑠 + 2)

𝐾(𝑠 + 3)(𝑠 + 4)
Therefore 𝑠 = −2 + 𝑗 2/2 is a closed-loop pole for 𝑇 𝑠 =
some value of gain 𝐾. 1+𝐾 𝑠2 + 3 + 7𝐾 𝑠 + (2 + 12𝐾)

To find the value of gain 𝐾 the gain criterion must be

satisfied:

1 𝑠+1 𝑠+2 𝐿𝑝1 × 𝐿𝑝2

𝐾= ቚ = ቚ = 0.333 or 𝐾 =
𝐺 𝑠 𝐻 𝑠 𝑠=−2+𝑗 2/2 𝑠 + 3 𝑠 + 4 𝑠=−2+𝑗 2/2 𝐿𝑧1 × 𝐿𝑧2

• Check the roots of the closed-loop

polynomial when 𝐾 = 0.333:
1.33𝑠 2 + 5.331𝑠 + 5.996 = 0 ⟹ 𝑠1,2 = −1.9996 ± 𝑗0.7068 = −2 ± 𝑗 2/2
Page 13
Properties of Root Locus (2)
In summary, given the poles and zeros of the open-loop transfer
function, 𝐾𝐺(𝑠)𝐻(𝑠), a point in the 𝑠-plane is on the root locus
for a particular value of gain 𝐾 if:

• The angles of the zeros minus the angles of the poles, all drawn to the
selected point on the s-plane, add up to 2𝑘 + 1 180°.
𝑚 𝑛

∠𝐺 𝑠 𝐻 𝑠 ቚ = ෍ ∠(𝑠 + 𝑧𝑖 ) − ෍ ∠ 𝑠 + 𝑝𝑖 ቚ = 2𝑘 + 1 180°
𝑠=𝐶.𝐿.𝑝𝑜𝑙𝑒 𝑠=𝐶.𝐿.𝑝𝑜𝑙𝑒
𝑖=1 𝑖=1

• The gain 𝐾 at that point for which the angles add up to 2𝑘 + 1 180° is found by dividing
the product of the pole lengths by the product of the zero lengths (it is the reciprocal
of the magnitude of 𝐺(𝑠)𝐻(𝑠) at that point).

ς𝑚
𝑖=1 𝑠 + 𝑧𝑖 1 ς𝑛𝑖=1 𝑠 + 𝑝𝑖
𝐾𝐺 𝑠 𝐻 𝑠 =𝐾 𝑛 ⟹ 𝐾= ቚ = ቚ
ς𝑖=1 𝑠 + 𝑝𝑖 𝐺 𝑠 𝐻 𝑠 𝑠=𝐶.𝐿.𝑝𝑜𝑙𝑒 ς𝑝𝑖=1 𝑠 + 𝑧𝑖 𝑠=𝐶.𝐿 𝑝𝑜𝑙𝑒

Page 14
Practice Problem 2
• For a unity feedback system which has the forward transfer function below, do the
following
𝐾 𝑠+2
𝐺 𝑠 =
𝑠 2 + 4𝑠 + 13
• Calculate the angle of 𝐺 𝑠 at the point 𝑠 = −3 by finding the algebraic sum of
angles of the vectors drawn from the zeros and poles of 𝐺(𝑠) to the given point.
• Determine if the point 𝑠 = −3 is on the root locus.
• If the point 𝑠 = −3 is on the root locus, find the gain 𝐾 so that 𝑠 = −3 is a closed-
loop pole.

Answer: s=-3+0j;
• ∠𝐺 𝑠 𝐻 𝑠 ȁ𝑠=−3 = 180° G=(s+2)/(s^2+4*s+13);
Theta=(180/pi)*angle(G)
• Point 𝑠 = −3 is on the root locus M=abs(G);
K=1/M
• 𝐾 = 10

Page 15
Sketching the Root Locus
• At the end of the day, we are interested to see how we can adjust the behaviour of the
closed-loop system by varying some parameters in the system (mostly controller
gains).
• We have learned that the behaviour of the closed-loop system relies on the location of
the closed-loop poles. So we need to see if we can bring the closed-loop poles to a
desired region in 𝑠-plane by varying those parameters.
• The Root locus is a graphical technique which allows us to see how the closed-loop
pole locations change in 𝑠-plane as a parameter/gain varies.
• Sketching the graph provides us with lots of useful information about the closed-loop
system from the open-loop transfer function including:
• The range of gains for which the system remains stable,
• the exact value of the gain that brings the closed-loop system to the edge of instability (critical
gain or ultimate gain in Ziegler-Nichols method).
• The frequency/period of the pure oscillation at the critical gain (used for PID design using
Ziegler-Nichols method).
• Whether or not we can find a value for the gain such that the closed-loop poles can be located
inside the desired region based on the design specification, or we need to add some zeros and
poles to the open-loop system to manipulate the shape of the root locus such that it passes
through the desired region and then find the gain (lead-lag compensator design). Page 16
Sketching the Root Locus – Basic rules 𝑅(𝑠) + 𝑌(𝑠)
𝐾 𝐺(𝑠)
• Number of branches −
• The number of branches of the root locus equals the number of
closed-loop poles.
𝐻(𝑠)
• Symmetry
𝐾𝐺(𝑠)
• The root locus is symmetrical about the real axis. 𝑇(𝑠) =
1 + 𝐾𝐺 𝑠 𝐻(𝑠)
• Real-axis segments The root locus uses the open-loop transfer
• On the real axis, for 𝐾 > 0 the root locus exists to the left of an function (loop gain) 𝐾𝐺 𝑠 𝐻(𝑠) to plot the
odd number of real-value and finite open-loop poles and/or location of the closed-loop poles for all
zeros. positive values of gain 𝐾: 0 → ∞.

• Starting and ending points • 𝑛 is the number of finite open-loop poles

(the system’s order)
• The root locus begins at the finite and infinite poles of 𝐺(𝑠)𝐻(𝑠)
and ends at the finite and infinite zeros of 𝐺(𝑠)𝐻(𝑠). • 𝑚 is the number of finite zeros (order of
the numerator of 𝐺(𝑠)𝐻(𝑠)).
• Behaviour at infinity • 𝑘 = 0, ±1, ±2, … is the multiplicity or the
number of asymptotes which is equal to
• The root locus approaches straight lines, known as asymptotes, 𝑘 = 𝑛 − 𝑚.
as the locus approaches infinity. Further, the equations of the
• Physical systems have strictly proper
asymptotes are given by the real-axis intercept 𝜎𝑎 and angle transfer functions (𝑛 > 𝑚) so infinite zeros
𝜃𝑎 in radians as follows: are assumed to match the number of finite
σ finite O. L. poles − σ finite O. L. zeros 2𝑘 + 1 𝜋 poles.
𝜎𝑎 = 𝜃𝑎 = Page 17
𝑛−𝑚 𝑛−𝑚
Example
𝑅(𝑠)+ 𝐾 𝑠+3 𝑌(𝑠)
• Consider the following system.
− 𝑠(𝑠 + 1)(𝑠 + 2)(𝑠 + 4)
• We expect to have four branches as there will be four
closed-loop poles
• Real-axis segments of the root locus are on the left side
of the odd number of finite open-loop poles and zeros:
𝑗𝜔
• −1 ≤ 𝜎 ≤ 0, −3 ≤ 𝜎 ≤ −2, 𝜎 ≤ −4
−𝑗3
• There are four finite poles, one finite zero, which
means we need three infinite zeros to match the
number of finite poles (𝑛 − 𝑚 = 4 − 1 = 3). asymptotes −𝑗2
• Three poles will approach three infinite zeros.
• The asymptotes will show how the three closed-loop −𝑗1
poles diverge to infinity. 𝜋
3
σ finite O. L. poles − σ finite O. L. zeros 0
𝜎
𝜎𝑎 = asymptotes −4 −3 −2 − 4 −1
𝑛−𝑚 3
− −𝑗1
(0 − 1 − 2 − 4) − (−3) 4
𝜎𝑎 = =−
4−1 3 asymptotes
− −𝑗2
𝜋
for 𝑘 = 0
2𝑘 + 1 𝜋 2𝑘 + 1 𝜋 3
𝜃𝑎 = = = 𝜋 for 𝑘 = 1 Number of asymptotes: 𝑛 − 𝑚 = 3 − −𝑗3
𝑛−𝑚 3 5𝜋
for 𝑘 = 2
3 Page 18
Example (2)
𝑅(𝑠)+ 𝐾 𝑠+3 𝑌(𝑠)
• Some important information from the root
− 𝑠(𝑠 + 1)(𝑠 + 2)(𝑠 + 4)
locus sketch.
• The open-loop zeros are the same as the
closed-loop zeros.
• As the gain increases, the closed-loop poles move toward the
system zeros. 𝑗𝜔
• Right-hand side zeros pull the closed-loop poles to the RHS of 𝑠-plane, The zeroes are like −𝑗3
making the closed-loop system unstable at some value of the gain 𝐾. magnets for the
• As the gain increases, the closed-loop poles tend to diverge (go locus; they pull the
−𝑗2
away from each other). closed-loop poles
closer to them.
• Two closed-loop poles might initially move toward each other, but −𝑗1
they always break away at some point on the real axis (how to find
that point?).
𝜎
• When 𝑛 − 𝑚 ≥ 3 with all the zeros on the left-hand side, as the gain −4 −3 −2 −1
0

increases, there are not enough finite zeros to pull the poles and
− −𝑗1
keep them in the LHP.
• So they diverge and at some value of the gain 𝐾, they enter the
− −𝑗2
right-hand side, making the closed-loop system unstable (how to
find that critical gain and the crossing point on the 𝑗𝜔 axis?).
− −𝑗3
• Adding zeros to the LHS of the 𝑠-plane is stabilizing at high
gains.
Page 19
More examples
𝑅(𝑠) + 𝑌(𝑠)
• 2nd-order system without zero: 𝐾 𝐺(𝑠)
−
1
𝐺 𝑠 =
(𝑠 + 1)(𝑠 + 3)
Underdamped system at
𝐾 > 1, complex conjugate
C.L. poles

Critically damped
system at 𝐾 = 1,
break away point is
s = −2

Overdamped
system at 𝐾 < 1,
real C.L. poles

s=tf('s');
G = 1/((s+1)*(s+3));
rlocus(G)
Page 20
More examples (2)
𝑅(𝑠) + 𝑌(𝑠)
• 2nd-order system with a LHP zero: 𝐾 𝐺(𝑠)
−
(𝑠 + 4)
𝐺 𝑠 =
(𝑠 + 1)(𝑠 + 3)

A break-away point
and a break-in point

Page 21
More examples (3)
𝑅(𝑠) + 𝑌(𝑠)
• 2nd-order
system with a RHP zero (nonminimum 𝐾 𝐺(𝑠)
−
phase system):
(𝑠 − 2)
𝐺 𝑠 =
(𝑠 + 1)(𝑠 + 3)

Critical gain (≈ 1.48) where

a closed-loop pole enters the
RHS of the 𝑠-plane

Page 22
More examples (4)
𝑅(𝑠) + 𝑌(𝑠)
• 2nd-order
system with two RHP zeros (nonminimum 𝐾 𝐺(𝑠)
−
phase system):

(𝑠 − 1)(𝑠 − 3)
𝐺 𝑠 = Critical gain (≈ 0.969)
(𝑠 + 1)(𝑠 + 3) with frequency of
oscillation of ≈ 1.73 rad/s

Page 23
More examples (5)
𝑅(𝑠) + 𝑌(𝑠)
• 3rd-order system without zero: 𝐾 𝐺(𝑠)
−

1
𝐺 𝑠 =
(𝑠 + 1)(𝑠 + 2)(𝑠 + 4)

Page 24
More examples (6)
𝑅(𝑠) + 𝑌(𝑠)
• 3rd-order system with a LHP zero: 𝐾 𝐺(𝑠)
−

(𝑠 + 3)
𝐺 𝑠 =
(𝑠 + 1)(𝑠 + 2)(𝑠 + 4)

Page 25
More examples (7)
𝑅(𝑠) + 𝑌(𝑠)
• 3rd-order system with complex conjugate zeros on the 𝐾 𝐺(𝑠)
−
LHP:
(𝑠 2 + 6𝑠 + 10)
𝐺 𝑠 = 2
(𝑠 + 4𝑠 + 8)(𝑠 + 4)

Departure
angle
Arrival
angle

Page 26
Transient response design via root locus
• When we sketch a root locus, it shows us whether or not there exists a
gain 𝐾 such that the closed-loop poles can be located inside a desired
region given by design specifications (transient response
specifications %𝑂𝑆, 𝑇𝑝 , and steady state response specifications 𝑇𝑠 , 𝑒𝑠𝑠 ).
• However, remember that the formulas describing %𝑂𝑆, 𝑇𝑝 and 𝑇𝑠 were
derived only for a system with two closed-loop complex poles and no
closed-loop zeros.
• So, if such a gain exists that can satisfy the specifications on the root
locus, we need to check that the closed-loop system would behave like
an ideal second-order system. Otherwise, further tweak the gains in
simulation by trial and error.
• If we cannot find any gain that can bring the root locus to the desired
region in the 𝑠-plane, we need to compensate for system response
deficiency by adding zeros and poles to the system (using PID or
Lead/Lag compensators).

Page 27
An example: Gain adjustment via root locus
• Consider a unity negative feedback system 𝑅(𝑠)+ 𝐾 𝑠 + 1.5 𝑌(𝑠)
with a 3rd-order transfer function model having 𝑠(𝑠 + 1)(𝑠 + 10)
a zero. −
• We wish to design the value of gain 𝐾 to
achieve %𝑂𝑆 ≤ 1.52%. 𝜁 = 0.8
• The root locus is already inside the region of
interest (damping ratio of 𝜁 = 0.8), but what exact
gain would result in %𝑂𝑆 = 1.52% for the closed-
loop step response?
• As you can see, at three points the root locus
crosses the 𝜁 = 0.8 line, suggesting that there
are three values for the gain 𝐾 that could result
in %𝑂𝑆 = 1.52%.
• The main question here is that which gain would
make the closed-loop system behave more like a
second-order system, and therefore resulting in
%𝑂𝑆 being closer to 1.52%.

𝜁 = 0.8

Page 28
An example: Gain adjustment via root locus (2) 𝐺 𝑠 =
𝐾 𝑠 + 1.5
𝑠(𝑠 + 1)(𝑠 + 10)
• Using MATLAB or manual calculation, we can
find the location of the complex conjugate
poles (and their corresponding gains) on the
root locus that crosses the 𝜁 = 0.8 line. They
are shown in the table:
• If we wish to have %𝑂𝑆 ≈ 1.52% we need to
find a gain that not only causes root locus to
cross the 𝜁 = 0.8 line, but also to make the
closed-loop system behave like a 2nd-order
system. The conditions are:

• The third pole of the closed-loop system remains at Best

least 5 to 10 times farther to the left of the real-part Gain
of the dominant closed-loop pole(s).
Gain 𝐾 7.36 12.79 39.65
• The zero remains at least 5 to 10 times farther to the
left of the real-part of the dominant closed-loop pole. Real part of
• If neither of the above conditions are satisfied, then complex −0.87 −1.19 −4.6
the third pole should be close to the zero to cancel conjugate poles
its effect. Third pole −9.25 −8.6 −1.8
Zero −1.5 −1.5 −1.5
Page 29
An example: Gain adjustment via root locus (3)
• Plotting the step response of the ideal 2nd-order
closed-loop system and comparing it with the step 𝑅(𝑠) + 𝐾 𝑠 + 1.5 𝑌(𝑠)
response of the actual 3rd-order system with a zero − 𝑠(𝑠 + 1)(𝑠 + 10)
using the same gain values of 12.79 and 39.65
shows that the third gain provides much better
approximation to a 2nd-order system response.
𝑅(𝑠) + 𝐾 𝑌(𝑠)
• However, if we cannot achieve the desired − 𝑠(𝑠 + 1)
specification with just a simple gain, we need to
add more flexibility to the system by adding zeros
and poles.

Page 30
PID controller design using root locus
• We learned that with PI, PD, and PID controllers, we can place
the closed-loop poles at any desired location on the 𝑠-plane
using the pole-placement technique.
• But this requires solving an algebraic equation, and for higher
order systems, we may not have enough equations to find
unique solutions.
• We can use the root locus to identify where the additional
zero(s) of the PI, PD and PID controllers should be located to
make sure that the desired closed-loop poles lie on the root 𝐺𝑃𝐷 𝑠 = 𝐾𝑃 + 𝐾𝐷 𝑠 = 𝐾(𝑠 + 𝑧𝑐 )
locus.
𝐾𝐼 𝐾 𝑠 + 𝑧𝑐
• To do this, we can write the transfer functions of the PI, PD and 𝐺𝑃𝐼 𝑠 = 𝐾𝑃 + =
PID controllers in zero-pole-gain form. 𝑠 𝑠
• As you can see, we still have the same number of variables to 𝐾𝐼
adjust the system’s behaviour as the number of gains, but 𝐺𝑃𝐼𝐷 𝑠 = 𝐾𝑃 + + 𝐾𝐷 𝑠
𝑠
instead of the 𝐾𝑃 , 𝐾𝐷 and 𝐾𝐼 gains, we can adjust the location 𝐾 𝑠 + 𝑧𝑐1 𝑠 + 𝑧𝑐2
of the zero(s) and find a corresponding gain 𝐾 to alter the =
root locus to cover the desired dominant closed-loop poles. 𝑠

Page 31
PI/PD controller design using root locus
• Here are the steps to accomplish the design: 𝑅(𝑠) + 𝑌(𝑠)
𝐺𝑐 (𝑠) 𝐺(𝑠)
• Find the desired closed-loop poles from the given −
specification (find 𝜁 and 𝜔𝑛 using %𝑂𝑆, 𝑇𝑝 and/or 𝑇𝑠 ).
𝐺𝑐 (𝑠) = 𝐺𝑃𝐷 𝑠 = 𝐾(𝑠 + 𝑧𝑐 )
𝑠𝐶𝐿 = −𝜁𝜔𝑛 ± 𝑗𝜔𝑛 1 − 𝜁 2
• For PI, place the zero somewhere near the origin. 𝐾 𝑠 + 𝑧𝑐
𝐺𝑐 𝑠 = 𝐺𝑃𝐼 𝑠 =
• For PD, place the zero somewhere to the left of the first two 𝑠
real open-loop poles.
• These are just a guide and the location is assumed to be unknown! Note: If we are only given
• Use the angle criterion of the root locus assuming an percent overshoot %𝑂𝑆 or
unknown zero angle 𝜃𝑧𝑐 and find what should be the angle of the damping ratio 𝜁 with
the zero to have an angle contribution of 2𝑘 + 1 180°. some reduction ratio on 𝑇𝑝 or
𝑇𝑠 , we need to sketch the
• Use the obtained zero angle 𝜃𝑧𝑐 to find the location of the zero uncompensated root locus
𝑧𝑐 geometrically. (only for the process transfer
function 𝐺(𝑠)) and find the
∠𝐿𝑜𝑜𝑝 𝐺𝑎𝑖𝑛 ቚ = 2𝑘 + 1 180° intersection of the root locus
𝑠=𝐶.𝐿. 𝑝𝑜𝑙𝑒
and the 𝜁 line. Then, we use
• Finally, use the gain criterion to find the corresponding gain 𝐾.
that point to find the
1 uncompensated 𝑇𝑝 or 𝑇𝑠 and
𝐾= ቚ finally find the desired closed-
𝐿𝑜𝑜𝑝 𝐺𝑎𝑖𝑛 𝑠=𝐶.𝐿. 𝑝𝑜𝑙𝑒 loop poles. Page 32
Example 3
• Design a PD controller for 𝐺(𝑠) to have damping ratio of 0.8 and peak time of 1.047 s.
• (Worked solutions of examples will be given either in lecture or uploaded)

1
𝐺 𝑠 =
(𝑠 + 1)(𝑠 + 4)

Answer:
𝐺𝑐 𝑠 = 3 𝑠 + 7 = 21 + 3𝑠

Page 33
Practice Problem 3
• Design a PI controller for 𝐺(𝑠) to have percent overshoot of 25.4% and a settling time
of 2.94 s.
1
𝐺 𝑠 =
(𝑠 + 1)(𝑠 + 3)(𝑠 + 10)

Answer:
102 𝑠 + 0.1 10.2
𝐺𝑐 𝑠 = = 102 +
𝑠 𝑠

Page 34
Lead and lag compensators
• To implement a PID controller we might need to use active circuits 𝑅(𝑠) + 𝐺𝑐 (𝑠) 𝐺(𝑠)
𝑌(𝑠)
(based on Op. Amps) which require a secondary power source. −
• A similar effect can be achieved using passive circuits, however
neither pure integration nor differentiation are feasible in this case.
𝐾(𝑠 + 𝑧𝑐 )
• Also, sometimes, only adding a zero (PD) or a zero with a fixed 𝐺𝑐 𝑠 =
pole at the origin (PI) or a full scale PID might not satisfy the 𝑠 + 𝑝𝑐
given design specifications which would make alteration of the real
plant necessary.
• The Cascade compensator of the following form can be used to
fix some of the system deficiencies.
• Cascade compensators are also known as phase lead, phase lag or a
combination of both (phase lead-lag).
• Compensators are mostly used in the forward path of the feedback
system.
• The name comes from their frequency response characteristic
as they can control the total phase and magnitude of the transfer
function at different frequencies (phase margin and gain margin).

Page 35
Lag compensator
𝑅(𝑠) + 𝑌(𝑠)
• Similar to PI and PD, cascade compensators add a zero and a 𝐺𝑐 (𝑠) 𝐺(𝑠)
pole to the open-loop system allowing us to adjust both the −
transient and steady-state characteristics of the system response.
• The lag compensator is designed to act similarly to the PI
controller in the time domain. 𝐾(𝑠 + 𝑧𝑐 )
𝐺𝑐 𝑠 =
𝑠 + 𝑝𝑐
• The main objective of using a PI controller is to increase the
system Type to eliminate steady-state error for a step reference
input. Lag compensator
zero and pole map
• So for a lag compensator to reduce the steady-state error, the
pole 𝑝𝑐 should be placed close to the origin.
• However, adding a pole close to the origin can easily destabilize
the system and dramatically deteriorate the transient response, so
the zero 𝑧𝑐 should be placed close to the pole on its left. −1

Note: A rule of thumb

Choose the location of the lag zero and pole such that 𝑧𝑐 = 10 𝑝𝑐 < 1. This will
guarantee a significant increase in the static error constant by a factor of 𝑧𝑐 /𝑝𝑐 (if 𝑒𝑠𝑠
cannot be eliminated) as well as having a negligible impact on the transient response.
Page 36
Lag compensator and static error constant
• Consider a Type 1 system. The static error
constant of the uncompensated system is
given by:
𝐾𝑧1 𝑧2 …
𝐾𝑣,uncomp = lim 𝑠𝐾𝐺 𝑠 =
𝑠→0 𝑝1 𝑝2 …

• The static error constant of the

compensated system is given by:

𝐾 𝑠 + 𝑧𝑐 𝐾(𝑧𝑐 )𝑧1 𝑧2 …
𝐾𝑣,comp = lim 𝑠 𝐺 𝑠 =
𝑠→0 𝑠 + 𝑝𝑐 (𝑝𝑐 )𝑝1 𝑝2 …

• The ratio of the new 𝐾𝑣 to the

original 𝐾𝑣 is:
𝐾𝑣,comp 𝑧𝑐
=
𝐾𝑣,uncomp 𝑝𝑐

• So if 𝑝𝑐 < 𝑧𝑐 < 1, the static error constant will improve by a

factor of 𝑧𝑐 /𝑝𝑐 resulting in significant reduction in steady-state
error 𝑒𝑠𝑠 for a ramp input (for instance).
Page 37
Design Lag compensator using root locus
• Similar to PI and PD, we can use the root locus to find the 𝑅(𝑠) + 𝑌(𝑠)
location of either the zero or the pole using the angle 𝐺𝑐 (𝑠) 𝐺(𝑠)
−
criterion, and then find the corresponding gain to bring the
dominant closed-loop poles to the desired location.
𝐾(𝑠 + 𝑧𝑐 )
• Here are the design steps: 𝐺𝑐 𝑠 =
𝑠 + 𝑝𝑐
• Case 1: Use the steady-state error requirement to find the new static error
constant and divide it by the uncompensated static error constant to find the
ratio of 𝑧𝑐 /𝑝𝑐 (if this requirement is given).
• Choose 𝑝𝑐 = 0.01 or 0.001 and then use the ratio of static error constants to
find 𝑧𝑐 . Then find the gain that satisfies the %𝑂𝑆 specification (sketch the root
locus and find the crossing point with the 𝜁 line, or use the MATLAB Control
Systems Designer App).

• Case 2: If the desired closed-loop poles from the given specification can be
obtained (i.e., find 𝜁 and 𝜔𝑛 using %𝑂𝑆, 𝑇𝑝 and/or 𝑇𝑠 ) and either 𝑧𝑐 or 𝑝𝑐 is already
given, simply find the location of the other one using the angle criterion and then
geometry (like PD/PI design).

• Finally, use the gain criterion to find the corresponding gain 𝐾.

Page 38
Example 4
• Compensate the unity feedback system with feedforward transfer function 𝐺(𝑠) to
improve the steady-state error by a factor of 10 if the system is operating with a
damping ratio of 0.174.
• (Worked solutions of examples will be given either in lecture or uploaded later)
1
𝐺 𝑠 =
(𝑠 + 1)(𝑠 + 2)(𝑠 + 10)

Answer:
158.1 𝑠 + 0.111
𝐺𝑐 𝑠 =
𝑠 + 0.01

Page 39
Lead compensator
• Lead compensators are designed to act similarly to PD 𝑅(𝑠) + 𝑌(𝑠)
controller in time domain. 𝐺𝑐 (𝑠) 𝐺(𝑠)
−
• The main objective of using a PD controller is to improve 𝑗𝜔
the transient response (%𝑂𝑆, 𝑇𝑝 and to a good extent 𝑇𝑠 ).
• The transient response improvement is possible by adding 𝐾(𝑠 + 𝑧𝑐 )
𝐺𝑐 𝑠 =
a zero 𝑧𝑐 to pull the root locus further to the left and bend it 𝑠 + 𝑝𝑐
toward the real zero (without interfering with the desired
dominant closed-loop poles). Lead compensator
zero and pole map
• But the pole of lead compensator 𝑝𝑐 should be placed far 𝑗𝜔
to the left, so that it can approximate the behaviour of
𝑠-plane
lead compensator like a normal PD controller (also being
less destabilizing). 0
𝜎
−𝑝𝑐 −𝑧𝑐
Note: A rule of thumb
Choose the location of the lead zero and pole, if no particular specification is given, to:
• place 𝑧𝑐 directly below the desired dominant closed-loop poles,
• or place 𝑧𝑐 to the left of the first two real open-loop poles,
• and then placing the pole between 5 to 10 times farther to the left of the least
dominant open-loop pole to have minimum impact on the transient response.
Page 40
Design Lead compensator using root locus
• Here are the design steps:
𝑅(𝑠) + 𝑌(𝑠)
• Find the desired closed-loop poles from the given 𝐺𝑐 (𝑠) 𝐺(𝑠)
specification if possible (find 𝜁 and 𝜔𝑛 using %𝑂𝑆, 𝑇𝑝 and/or 𝑇𝑠 ). −
• If you can directly place the zero 𝑧𝑐 under the desired dominant
closed-loop poles, then place the pole 𝑝𝑐 somewhere far to the 𝐾(𝑠 + 𝑧𝑐 )
left of the all open-loop poles. 𝐺𝑐 𝑠 =
𝑠 + 𝑝𝑐
• Using the angle criterion of the root locus, find the angle of the
pole 𝑝𝑐 such that it has an angle contribution of 2𝑘 + 1 180°.
• Otherwise, place the zero 𝑧𝑐 somewhere to the left of the first Note: The lead pole 𝑝𝑐 is
two real open-loop poles. Then use the angle criterion to find mostly the unknown variable
the angle of the pole 𝑝𝑐 such that it has an angle contribution of to be located in lead
2𝑘 + 1 180°. compensator design, but
• Use the obtained angle to find the location of the pole 𝑝𝑐 sometime the location of the
geometrically. lead pole is already given, so
you need to only find the
• Finally, use the gain criterion to find the corresponding gain 𝐾. location of the zero using the
given specifications and the
angle and gain criteria.

Page 41
Example 5
• Design a lead compensator for the unity feedback system with feedforward transfer
function 𝐺(𝑠) that will reduce the settling time by a factor of 2 while maintaining 30%
overshoot.
• (Worked solutions of examples will be given either in lecture or uploaded)

1
𝐺 𝑠 =
𝑠(𝑠 + 4)(𝑠 + 6)

Page 42
Lead-lag compensator
• Similar to PID, if we wish to improve both steady state error and transient
response, we can combine these two compensators known as the lead-lag
compensator.
• To avoid complication of the design procedure, we design each of them
separately, mostly starting with the lead compensator to improve the
transient response.
• If the transient response requirements are met after the simulation and
verification, then we design the lag compensator to improve and satisfy the
steady-state error requirement.
• Final note: With any control system design, you need to always verify your
final control system by simulation and if needed, make any proper
adjustments by trial-and-error using your understanding of the behaviour of
each control parameter to achieve the desired response.
• MATLAB Control System Designer is a great tool that provides a visual and
graphical display of the root locus as well as allowing you to move the added
zeros and poles to see the result in the response instantaneously (with lots of
other features).

Page 43
Summary
• From Nise Textbook

Page 44
Realization of PID and Compensators
• Active circuit realization

Page 45
Realization of PID and Compensators (2)
• Active circuit realization

Page 46
Realization of PID and Compensators (3)
• Active circuit realization

Page 47
Lead-Lag Compensator with Op Amp

Page 48
Passive Circuit Realization of Lead/Lag Compensators

Page 49
Questions?

Page 50