Root Locus in Design of Discrete Control

Systems
Introduction
The discrete-time root locus, drawn in the z-plane, is essentially the same as the continuous-time one drawn in the s-plane,
except that the stability boundary is the unit circle instead of the imaginary axis. It is especially useful for preliminary design
work, e.g. for stability investigations or locating dominant poles at specific positions in the z-plane in order to meet a
performance requirement.

The method consists of expressing the characteristic equation of the discrete-time system in the form:
1
1 + 𝛼𝐹 (𝑧) = 0, i.e., 𝐹 (𝑧) = − 𝛼 

where 𝛼 is some system parameter (e.g. gain) which can be varied and 𝐹 (𝑧) is the actual or dummy open-loop transfer function.

A point will lie on this root locus if it satisfies two conditions:

1. phase condition: ∠𝐹(𝑧) = 𝑜𝑑𝑑 𝑚𝑢𝑙𝑡𝑖𝑝𝑙𝑒 𝑜𝑓 180° = ±(2𝑛 + 1)180° for 𝑛 = 0, 1, 2, ⋯ (2)
1
2. magnitude condition: |𝐹(𝑧)| = taking 𝛼 > 0 (3)
𝛼

The root locus is a plot of all points satisfying the phase condition. The magnitude condition gives the value of 𝛼 corresponding
to a point on the locus.

General features of the Root Locus

The root locus has the following general features:
1. It is symmetrical about the real axis.
2. It has n branches, each a continuous curve where n is the number of poles.
3. It has each branch starting (𝛼 = 0) on an open-loop pole and finishing (𝛼 = ∞) either on an open-loop zero or at infinity
along an asymptote.
4. It has 𝑛 − 𝑚 asymptotes where m is the number of zeros.

Construction
A root locus can be constructed using computer packages such as MATLAB or manually as follows.

Pole-zero map
𝑁(𝑧)
If the characteristic equation is 1 + 𝛼𝐹 (𝑧) = 0 where 𝐹 (𝑧) = 𝑄(𝑧), then
open loop zeros are the roots of 𝑁(𝑧) = 0, i.e. 𝑧1 , 𝑧2 , ⋯ , 𝑧𝑚 say
open loop poles are the roots of 𝑄(𝑧) = 0, i.e. 𝑝1 , 𝑝2 , ⋯ , 𝑝𝑛 say

Draw the corresponding z-plane open-loop pole-zero map (𝑛 ≥ 𝑚).

Real axis segments

A point on the real axis in the z-plane lies on the root locus iff it lies to the left of the odd total number of open-loop poles and
zeros.

Asymptotes
The 𝑛 − 𝑚 asymptotes are defined as follows:

A point of intersection (𝜎, 0) on the real axis where 𝜎 is given by:

(𝑠𝑢𝑚 𝑜𝑓 𝑜𝑝𝑒𝑛 𝑙𝑜𝑜𝑝 𝑝𝑜𝑙𝑒𝑠)−(𝑠𝑢𝑚 𝑜𝑓 𝑜𝑝𝑒𝑛 𝑙𝑜𝑜𝑝 𝑧𝑒𝑟𝑜𝑠)
𝜎= (𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑝𝑜𝑙𝑒𝑠)−(𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑧𝑒𝑟𝑜𝑠)

67
Angles of intersection 𝜃 made with the real axis are given by:
(2𝑞+1)𝜋
𝜗= where 𝑞 = 0, 1, 2, ⋯ , (𝑛 − 𝑚 − 1)
𝑛−𝑚

Points of intersection with the imaginary axis are given by (0, |𝜎| tan 𝜃 ).

Breakaway and break-in points

If the characteristic equation is 1 + 𝛼𝐹 (𝑧) = 0, then these points are the roots of:
𝑑𝛼 𝑑 1
= 0, i.e. [ ]=0
𝑑𝑧 𝑑𝑧 𝐹(𝑧)

𝑑2 𝛼
If <0 → breakaway
𝑑𝑧 2

𝑑2 𝛼
If >0 → break-in
𝑑𝑧 2

Departure or arrival angles

The angle of departure 𝜙𝑑 from a complex open-loop pole is given by:
𝜙𝑑 = 180° − ∑(𝑎𝑛𝑔𝑙𝑒𝑠 𝑓𝑟𝑜𝑚 𝑜𝑡ℎ𝑒𝑟 𝑝𝑜𝑙𝑒𝑠) − (𝑎𝑛𝑔𝑙𝑒𝑠 𝑓𝑟𝑜𝑚 𝑧𝑒𝑟𝑜𝑠)

The angle of arrival 𝜙𝑎 at a complex open-loop pole is given by:

𝜙𝑎 = 180° − ∑(𝑎𝑛𝑔𝑙𝑒𝑠 𝑓𝑟𝑜𝑚 𝑜𝑡ℎ𝑒𝑟 𝑧𝑒𝑟𝑜𝑠) − (𝑎𝑛𝑔𝑙𝑒𝑠 𝑓𝑟𝑜𝑚 𝑝𝑜𝑙𝑒𝑠)

Imaginary axis crossings

To find where the root locus crosses the imaginary axis in the z-plane, put 𝑧 = 𝑗𝑦 in the characteristic equation, equate the real
and imaginary parts to zero and solve for 𝛼 and y.

Unit circle crossings

Use Jury’s stability criterion or the Routh-Hurwitz method to find the values of z and the corresponding values of 𝛼for which
the system becomes just unstable, i.e. where the root locus crosses the unit circle.

Dominant pole locations

Dominant poles at 𝑠 = −𝜁𝜔𝑛 ± 𝑗𝜔𝑛 √1 − 𝜁 2 in the s-plane map into points
𝑇(−𝜁𝜔𝑛 ±𝑗𝜔𝑛 √1−𝜁 2 )
𝑧=𝑒 = 𝑒 −𝑇𝜁𝜔𝑛 𝑒 ±𝑗𝑇𝜔𝑑 in the z-plane, i.e.

2𝜋𝜁 𝜔
−( × 𝑑)
𝜔𝑠
−𝑇𝜁𝜔𝑛 √1−𝜁2 2𝜋
|𝑧 | = 𝑒 =𝑒 where 𝜔𝑑 = 𝜔𝑛 √1 − 𝜁 2 and 𝑇 = 𝜔
𝑠

𝜔
∠𝑧 = ±𝑇𝜔𝑑 = ±2𝜋 𝜔𝑑
𝑠

𝜔𝑑
where is the number of oscillations per damped oscillation period.
𝜔𝑠

Charts of the type shown in topic 6 are also helpful for indicating the regions of the z-plane in which the dominant poles must
lie to meet the system performance requirements.

Application to controller design

The root locus method can be used as an aid to the design of digital controllers with the following general approach:
1. Determine the position of the dominant closed-loop poles from the performance specification requirements for
the system.
2. Draw the open-loop pole-zero map in the z-plane and calculate the root-locus angle at a dominant pole position.
For example, for point P in figure 8.1, where 𝑍1 is the open-loop zero and 𝑃1 , 𝑃2 are open-loop poles,
𝑟𝑜𝑜𝑡 𝑙𝑜𝑐𝑢𝑠 𝑎𝑛𝑔𝑙𝑒 = 𝜙 − (𝜃1 + 𝜃2 )

68
 Im
P

𝜃2
𝜃1
𝜙
×𝑃 0
𝑍1
Օ ×
𝑃
Re
1 2

Figure 8.1

3. Calculate the extra phase angle needed to make this – 180°, i.e. to put point P on the root locus.
4. Decide on a suitable form of the controller 𝐷(𝑧) and calculate the controller poles and zeros required to meet the
phase condition.
5. Calculate the gain of the controller such that P also meets the root locus magnitude condition.

Standard controller forms

PID controllers are often used in discrete-time control. Some of the following simpler forms can be used in addition to the full
PID controller discussed in an earlier topic:
1. Proportional: 𝐷 (𝑧 ) = 𝑘 𝑃 a constant.
𝑘𝐼 𝑘𝐼 𝑧
2. Integral: 𝐷(𝑧) = (1−𝑧 −1) = (𝑧−1) form
𝑘 𝑧−𝑎
3. P + I: 𝐷(𝑧) = 𝑘𝑃 + (1−𝑧𝐼−1) = 𝑘 (𝑧−1) form
𝑧−𝑎
4. P + D: 𝐷(𝑧) = 𝑘𝑃 + 𝑘𝐷 (1 − 𝑧 −1 ) = 𝑘 𝑧
form
𝑧−𝑎
5. General: 𝐷(𝑧) == 𝑘 𝑧−𝑏 for pole cancellation

Examples
The following examples show the application of the above procedures to some typical example problems.

Example 9.1
For the sampled data system in the following diagram, 𝑇 = 0.5 seconds and 𝐷(𝑧) is an integral controller with variable gain,
𝑘 𝑘𝑧
i.e. 𝐷(𝑧) = (1−𝑧𝐼−1) = (𝑧−1) .

𝐺 (𝑧 )

𝑇
1
+
𝐷 (𝑧 ) ZOH 𝑠+1
−

Figure 8.2.

1 𝑧−1 1 1 𝑧−1 1 𝑧
𝐺 (𝑧) = (1 − 𝑧 −1 )𝑍 [𝑠(𝑠+1)] = 𝑍 [ 𝑠 − 𝑠+1] = [ − 𝑧−𝑒 −𝑇 ]
𝑧 𝑧 𝑧−1

𝑧−1 𝑧(1−𝑒 −𝑇 ) 0.39

𝐺 (𝑧 ) = [(𝑧−1)(𝑧−𝑒 −𝑇 )] = 𝑧−0.61 for 𝑇 = 0.5
𝑧

Therefore, the characteristic equation of the system is 1 + 𝐷(𝑧)𝐺(𝑧) = 0, i.e.

𝑘𝑧 0.39
1 + 𝑧−1 × 𝑧−0.61 = 0

69
 0.39𝑧
1 + 𝑘 (𝑧−1)(𝑧−0.61) = 0

Proceeding through the stages of plotting a root locus:

1. Open-loop zeros: 𝑧=0
2. Open-loop poles: 𝑧 = 0.61 , 𝑧 = 1
3. Branches: two branches, one of which ends at −∞.
4. Real axis segments: [0.61 → 1] , [0 → −∞].
5. Asymptote is the negative real axis
6. Breakaway/break-in given by:
(𝑧−1)(𝑧−0.61)
−𝑘 = 0.39𝑧

𝑑𝑘 0.39𝑧(2𝑧−1.61)−0.39(𝑧 2 −1.61𝑧+0.61) 0.39𝑧 2 −0.39×0.61 𝑧 2 −0.61

That is: = (0.39𝑧)2
= (0.39𝑧)2
=
𝑑𝑧 0.39𝑧 2
𝑑𝑘
7. Breakaway/break-in points at 𝑑𝑧 = 0, i.e. 𝑧 = ±0.78
8. That is, breakaway at 𝑧 = 0.78 and break in at 𝑧 = −0.78
9. From this information, the root locus is as shown in figure 8.3.

Im

Unit circle

Re
0 0.61 1.0

Figure 8.3.

The critical value of k (i.e. where the system becomes unstable) occurs when the root locus crosses the unit circle at 𝑧 = −1,
that is:
(𝑧−1)(𝑧−0.61) (−2)(1.61)
𝑘𝑐𝑟𝑖𝑡𝑖𝑐𝑎𝑙 = | 0.39𝑧 | = | 0.39(−1) | = 8.26
𝑧=−1

Therefore, the system is stable for 0 < 𝐾 < 8.26.

Example 9.2
𝑧−𝜆
For the sampled data system in figure 8.4, 𝑇 = 0.5 seconds and 𝐷(𝑧) is a 𝑃 + 𝐼 controller, i.e. 𝐷 (𝑧) = 𝑘 (𝑧−1) with k and  to
be determined to give 8 samples per damped oscillation and dominant poles with a damping ratio 𝜁 = 0.5.

𝐺 (𝑧 )

𝑇
𝑈 (𝑧 ) 𝑒 −𝑠𝑇
𝑅 (𝑧 ) + 𝐸 (𝑧 )
𝐷 (𝑧 ) ZOH 𝑠+1
−

Figure 8.4.

1 𝑧−1 𝑧(1−𝑒 −𝑇 ) 0.39

𝐺 (𝑧) = (1 − 𝑧 −1 ) × 𝑧 −1 × 𝑍 [𝑠(𝑠+1)] = [(𝑧−1)(𝑧−𝑒 −𝑇 )] = 𝑧(𝑧−0.61)
𝑧2

70
Now for the dominant pole of the system:
2𝜋𝜁 𝜔
−( × 𝑑)
𝜔𝑠
√1−𝜁2 𝜔
|𝑧 | = 𝑒 and ∠𝑧 = ±2𝜋 𝜔𝑑
𝑠

±2𝜋
If 𝜔𝑠 = 8𝜔𝑑 as given, then ∠𝑧 = = ±45°, i.e. dominant poles lie at ±45° to the real axis and
8
2𝜋𝜁 𝜔
−( × 𝑑) 2𝜋(0.5) 1
𝜔𝑠 −( × )
√1−𝜁2 √1−(0.5)2 8
|𝑧 | = 𝑒 =𝑒 = 𝑒 −0.4535 = 0.635

Therefore, the dominant poles lie at 𝑧 = 0.635(cos 45° ± 𝑗 sin 45° ) = 0.45 ± 𝑗0.45

Consider the dominant pole in the upper half of the z-plane (call it P). For the root locus we have the following characteristic
equation:
0.39(𝑧−𝜆)
1 + 𝑘 𝑧(𝑧−1)(𝑧−0.61) = 0

That is,
Open loop zeros: 𝑧=𝜆
Open loop poles: 𝑧 = 0 , 𝑧 = 0.61 , 𝑧 = 1

Since P lies on the root locus, ∑(𝑧𝑒𝑟𝑜 𝑎𝑛𝑔𝑙𝑒𝑠) − ∑ 𝑝𝑜𝑙𝑒 𝑎𝑛𝑔𝑙𝑒𝑠 = ±(2𝑞 + 1)180°

Therefore, if any angle from the unknown zero at 𝑧 = 𝜆 is 𝜙, we have:

𝜙 − (45° + 𝜃1 + 𝜃2 ) = −180° i.e. 𝜙 = −180° + 45° + 109.6° + 140.7° = −115.3° (𝜃1 and 𝜃2 obtained as shown
in figure 8.5).

Im 0.45
𝜃1 = 180° − tan−1 ൬ ൰ = 109.6°
P (0.45 + 𝑗4.5) 0.16

0.45
𝜃2 = 180° − tan−1 ൬ ൰ = 140.7°
0.55

𝜃1 𝜃2
𝜙
45° Re
0× ×
0.61
Օ ×
1.0

Figure 8.5.

0.45
Therefore 𝜆 lies between 0.61 and 1 and tan 64.7° = 𝜆−0.45 = 2.115 → 𝜆 = 0.66

Also
0.39𝑧(𝑧−0.66)
𝑘 |𝑧(𝑧−1)(𝑧−0.61)| = 1 by the magnitude relation for P on the root locus.
𝑧=0.45+𝑗0.45

(0.45+𝑗0.45)(0.45+𝑗0.45−1)(0.45+𝑗0.45−0.61)
𝑘=| | = 1.11
0.39(0.45+𝑗0.45)(0.45+𝑗0.45−0.66)

1.11(𝑧−0.66)
Thus, the required controller is 𝐷(𝑧) = 𝑧−1

To put this into the form of digital algorithm, we have

𝑈(𝑧) 1.11(𝑧−0.66)
𝐷(𝑧) = 𝐸(𝑧) = i.e. (1 − 𝑧 −1 )𝑈 (𝑧) = (1.11 − 0.73𝑧 −1 )𝐸 (𝑧)
𝑧−1

The algorithm is: 𝑢(𝑘) = 𝑢(𝑘 − 1) + 1.11𝑒(𝑘 ) − 0.73𝑒(𝑘 − 1)

71
Chapter 9 Tutorial Questions
1. In the system shown below, D(z) is an integral controller of the following form:
αz
D(z) 
z 1

+ 1
D(z) ZOH
- s 1

Draw the root loci for T = 0.5, 1 and 2 seconds and find the corresponding values of  for which the system becomes
unstable. [ 8.17, 4.33, 2.63 ]

e 2s
2. For the system shown in Q3, but with G p (s)  , design D(z) to be a PI algorithm such that the dominant poles of
s 1
the closed loop system have a damping ratio of 0.5 and the error is sampled 12 times each damped oscillation period.
 0.513(z  0.49) 
D(z)  z 1 
 
1 za
3. For the system shown in Q3, but with G p (s)  , T = 200ms. Design a controller D(z) of the form D(z)  k
s(s  2) zb
to give a damping ratio z = 0.6 and a sampling rate of 8 samples per damped period. What is the velocity constant of the
resulting system?
 15.85(z  0.67) 
D(z)  , 2.79 sec 1 
 z  0.062 

72