Z-Domain Root Locus

M. Sami Fadali
Professor of Electrical Engineering
UNR

1
 Outline

• Root locus plots in z-plane.

• Pole locations and time response.
• Z-plane contours.
• Proportional control.

2
 Closed-loop Characteristic
Equation
1 + 𝐶 𝑧 𝐺𝑍𝐴𝑆 𝑧 = 0
1 + 𝐾𝐿 𝑧 = 0
𝐶 𝑧 = controller transfer function
𝐺𝑍𝐴𝑆 𝑧 = transfer function of DAC, analog
subsystem, and sampler
𝐿 𝑧 = loop gain
𝐾= gain

3
 Observations
• Identical equation to s-domain equation
with 𝑠 replaced by 𝑧.
• All the rules derived for s-domain are
applicable and can be used to obtain z-
domain root locus plots.
• The plots can also be obtained using the
root locus plots of most CAD programs
(MATLAB = rlocus)
4
 Example 6.1
• First order type 1 system with loop gain
1
𝐿 𝑧 =
𝑧−1
• Obtain the root locus plot.
• Obtain the critical gain.

5
 Solution
• Root locus rules give plot (MATLAB
rlocus).
• Root locus: real axis locus between pole
and zero.
• For a stable discrete system, real axis loci
must lie between 1,0 and −1,0 in the z-
plane.
• Critical gain 𝐾𝑐𝑟 is at point −1,0 .

6
 Root Locus for Example 6.1

• Closed-loop characteristic equation 𝑧 − 1 + 𝐾 = 0

• Substituting 𝑧 = −1 gives 𝐾𝑐𝑟 = 2 7
 Example 6.2

• Second order type 1 system with loop gain

1
𝐿 𝑧 =
𝑧 − 1 𝑧 − 0.5
• Obtain the root locus plot
• Obtain the critical gain.

8
 Solution
•Use root locus rules.
•RL: like Example 5.1(i) but in RHP.
•Breakaway point: 𝑧𝑏 = 1 + 0.5 /2 = 0.75
• 𝐾𝑐𝑟 (critical) intersection of RL & unit circle.
•Closed loop characteristic equation
𝑧 − 1 𝑧 − 0.5 + 𝐾 = 𝑧 2 − 1.5𝑧 + 𝐾 + 0.5 = 0
On the unit circle, 𝑧 = 1 (complex conjugate)

9
 Root Locus of Example 6.2

𝑧1,2 = 𝐾𝑐𝑟 + 0.5 = 1

⇒ 𝐾𝑐𝑟 = 0.5, 𝑧1,2 = 0.75 ± 𝑗0.661 10
 z-Domain Pole Locations &
Associated Temporal Sequences
Im{z}

. .. . .
. . . .
. . . . .
.
. . x
. . x
. Re{z}
x x x x x x
. . . . .
. . x .
. . . ..
..
. .

11
Time Functions & Real Poles

Continuous Laplace Sampled z-Transform

Transform

𝑓 𝑡 𝐹 𝑠 𝑓 𝑘𝑇 𝐹 𝑧
1 𝑧
𝑒 −𝛼𝑡 , 𝑡 ≥ 0 = 𝑒 −𝛼𝑘𝑇 , 𝑘 ≥ 0 =
=ቊ =ቊ 𝑧 − 𝑒 −𝛼𝑇
0, 𝑡 < 0 𝑠+𝛼 0, 𝑘 < 0

12
 Time Functions & Complex
Conjugate Poles
Continuous 𝑒 −𝜁𝜔𝑛 𝑡 sin 𝜔𝑑 𝑡 , 𝑡 ≥ 0
𝑓 𝑡 =ቊ
0, 𝑡 < 0
Laplace 𝜔𝑑
𝐹 𝑠 =
Transform 𝑠 + 𝜁𝜔𝑛 2 + 𝜔𝑑2
Sampled 𝑒 −𝜁𝜔𝑛 𝑘𝑇 sin 𝜔𝑑 𝑘𝑇 , 𝑘 ≥ 0
𝑓 𝑘𝑇 = ቊ
0, 𝑘 < 0
z-Transform 𝐹 𝑧
sin 𝜔𝑑 𝑇 𝑒 −𝜁𝜔𝑛 𝑇 𝑧
= 2
𝑧 − 2 cos 𝜔𝑑 𝑇 𝑒 −𝜁𝜔𝑛𝑇 + 𝑒 −2𝜁𝜔𝑛𝑇
13
2 nd Order Underdamped System
• z-domain characteristic polynomial
𝑧 − 𝑒 −𝜁𝜔𝑛+𝑗𝜔𝑑 𝑇 𝑧 − 𝑒 −𝜁𝜔𝑛 −𝑗𝜔𝑑 𝑇
= 𝑧 2 − 2 cos 𝜔𝑑 𝑇 𝑒 −𝜁𝜔𝑛 𝑇 𝑧 + 𝑒 −2𝜁𝜔𝑛 𝑇
• Poles
−𝜁𝜔𝑛 ±𝑗𝜔𝑑 𝑇
𝑧1,2 = 𝑒 = 𝑒 −𝜁𝜔𝑛 𝑇 ∠ ± 𝜔𝑑 𝑇

14
 Observation

If the Laplace transform 𝐹 𝑠 of a

continuous-time function 𝑓 𝑡 has a pole
𝑝𝑠 , then the z-transform 𝐹 𝑧 of its
sampled counterpart 𝑓 𝑘𝑇 , with sampling
period 𝑇 has a pole at
𝑝𝑧 = 𝑒 𝑝𝑠𝑇

15
 Primary Strip


s 
Primary 
Strip

s 

𝑝𝑧 = 𝑒 𝑝𝑠 𝑇 = 𝑒 𝜎𝑇 𝑒 𝑗𝜔𝑑 𝑇
= 𝑒 𝜎𝑇 𝑒 𝑗 𝜔𝑑 𝑇+𝑘2𝜋 , 𝑘 = 0,1,2, …
16
Pole Contours in the s-Domain
and the z-Domain
Poles 𝑠1,2 = 𝜎 ± 𝑗𝜔𝑑 , 𝑧1,2 = 𝑒 𝜎𝑇 ∠ ± 𝜔𝑑 𝑇
Contour s-Domain Contour z-Domain
Poles Poles
𝜎 vertical 𝑧 = 𝑒 𝜎𝑡 circle
constant line constant

𝜔𝑑 horizontal ∠𝑧 radial line

constant line constant

17
 Constant  Contours
j Im(z)
5 3
4
10
3 2
6
2 2
1
-2
1
-6
0 0 -10

-1
-1
-2

-3
-2
-4

-5  -3 Re(z)
-10 -5 0 5 10 -3 -2 -1 0 1 2 3

Constant  contours in s-plane. Constant  contours in z-plane.

18
 Constant 𝜔𝑑 Contours
Im(z)
3

j 2
    

3/4
1
/2

/4 0  0

0
-1
/4

/2
-2
    
3/4

-3 -2 -1 0 1 2 3
 -3 Re(z)
-3 -2 -1 0 1 2 3

Constant 𝜔𝑑 contours in the s-plane. Constant 𝜔𝑑 contours in the z-plane.

19
 Constant 𝜁 Contours


Poles 𝑧1,2 = 𝑒 −𝜁𝜔𝑛 𝑇 ∠ ± 𝜔𝑛 𝑇 1 − 𝜁 2

• Logarithmic spirals that get smaller for
larger values of 𝜁.
• The spirals are defined by the equation
𝜋𝜃°
1−𝜁 2 −𝜁 / 1−𝜁 2
𝑧 = 𝑒 −𝜁𝜃/ =𝑒 180°

𝑧 = magnitude of pole
𝜃= angle of pole
20
 Constant 𝜔𝑛 Contours

− 𝜔𝑛 𝑇 2 −𝜃2
𝑧 =𝑒
𝑧 = magnitude of pole
𝜃= angle of pole
To obtain the expression, eliminate 𝜁 from
𝑧1,2 = 𝑒 −𝜁𝜔𝑛 𝑇 ∠ ± 𝜔𝑛 𝑇 1 − 𝜁 2

21
 MATLAB
>> g=tf(num, den, T) % sampling period T
>> rlocus(g) % Root locus plot
>> zgrid(zeta, wn) % Plot contours
% zeta = vector of damping ratios
% wn = vector of undamped natural
% frequencies

22
Z-Domain Grid

23
 Characteristics of Log Spirals
1. Two spiral for each 𝜁 value, corresponding to ±𝜃:
−𝜃 spiral below real axis=mirror image of +𝜃 spiral
2. For any spiral, 𝑧 drops logarithmically with 𝜃
increase.
3. At the same angle , increasing 𝜁 gives smaller 𝑧 ,
i.e. spirals are smaller for larger 𝜁 values.
4. All spirals start at 𝜃 = 0, 𝑧 = 1 but end at
different points.
5. When given 𝜁 and 𝑧 , obtain  by substituting in
the equation
1 − 𝜁2
𝜃= ln 𝑧
𝜁 24
z-Domain Design Specifications

• Similar to those for s-domain design.

• Often approximate values based on
continuous time definitions.
• Allow selection of pole locations for z-
domain design.

25
 Time Constant
• Time constant of exponential decay for
the continuous envelope of sampled
waveform.
1
𝜏=
𝜁𝜔𝑛
• Not necessarily equal to a specified
percentage of the final value after one
time constant

26
 Settling Time
• Period after which the envelope of the
sampled waveform stays within a
specified percentage (usually 2%) of the
final value.
• Multiple of time constant depending on
the specified percentage.
• Settling time for a 2% specification
4
𝑇𝑠 =
𝜁𝜔𝑛
27
 Other Specification
• Frequency of Oscillations 𝜔𝑑 : angle of the
dominant complex conjugate poles divided by
the sampling period.
• Other design criteria such as the percentage
overshoot, 𝜁, 𝜔𝑛 , defined analogously to the
continuous case.
• As in analog design, select a dominant closed-
loop pair in the complex plane to obtain a
satisfactory time response.
• Analytical design: possible for low order
systems but more difficult than analog design.
28
 Example 6.3
Design a proportional controller for the
digital system with sampling period
T=0.1s to obtain
a) 𝜔𝑑 = 5 rad/s
b) A time constant of 0.5 s
c) 𝜁 = 0.7
1
𝐿 𝑧 =
𝑧 − 1 𝑧 − 0.5
29
 Solution (a), (b)
• Obtain the results with calculator or
MATLAB.
• Use MATLAB: rlocus
(a) 𝜔𝑑 = 5 rad/s:
angle of the pole = 𝜔𝑑 𝑇
= 5 × 0.1 = 0.5 rad
= 28.65°
(b) 𝜏 = 0.5 s:
1/( time constant)= 𝜁𝜔𝑛 = 1/0.5 = 2 rad /s
pole magnitude = exp −𝜁𝜔𝑛 𝑇 = 0.82

30
 Solution (c)
• Obtain the results with calculator or
MATLAB.
• Use MATLAB: rlocus
(c) 𝜁 = 0.7 :
• Use 𝜁 directly to get the results the gain and
the undamped natura frequency 𝜔𝑛 .
• Sampled step response: MATLAB command
step
• Higher gain designs have a low 𝜁
(oscillatory response).

31
 P-Control Design Results
Design Gain 𝜁 𝜔𝑛 rad/s

𝜔𝑑 = 5 0.23 0.3 5.24

rad/s
𝜏 = 0.5 s 0.17 0.4 4.60

𝜁 = 0.7 0.10 0.7 3.63

32
 Time Response Plots

Designs of Table 6.1 (a) =  , (b) = , (c) = +.

33
 Analytical Design
1
𝐿 𝑧 =
𝑧 − 1 𝑧 − 0.5
• Closed-loop characteristic equation
𝑧 2 − 1.5𝑧 + 0.5 + 𝐾
= 𝑧 2 − 2 cos 𝜔𝑑 𝑇 𝑒 −𝜁𝜔𝑛 𝑇 𝑧 + 𝑒 −2𝜁𝜔𝑛 𝑇
• Equating coefficients
𝑧1 : 1.5 = 2 cos 𝜔𝑑 𝑇 𝑒 −𝜁𝜔𝑛 𝑇
𝑧0: 𝐾 + 0.5 = 𝑒 −2𝜁𝜔𝑛 𝑇 34
 Design (a) 𝜔𝑑 = 5, 𝑇 = 0.1
• 𝑧1 equation: 1.5 = 2 cos 𝜔𝑑 𝑇 𝑒 −𝜁𝜔𝑛𝑇
1 1.5 1.5
𝜁𝜔𝑛 = − ln = −10 ln
𝑇 2 cos 𝜔𝑑 𝑇 2 cos 0.5
≈ 1.571
𝜔𝑑2 = 𝜔𝑛2 1 − 𝜁 2 = 25
𝜔𝑑2 1 − 𝜁2 25 𝑟𝑎𝑑
2
= 2
= 2
⇒ 𝜁 = 0.3, 𝜔𝑛 = 5.24
𝜁𝜔𝑛 𝜁 1.571 𝑠
• 𝑧 0 equation: 𝐾 + 0.5 = 𝑒 −2𝜁𝜔𝑛𝑇
𝐾 = 𝑒 −2𝜁𝜔𝑛𝑇 − 0.5 = 𝑒 −2×1.571×0.1 − 0.5 ≈ 0.23 35
 Design (b) 𝜏 = 0.5 𝑠
𝜁𝜔𝑛 = 1/𝜏 = 1/0.5 = 2 rad /s
𝑧1 equation: 1.5 = 2 cos 𝜔𝑑 𝑇 𝑒 −𝜁𝜔𝑛 𝑇
𝜁𝜔𝑛 𝑇
1 1.5𝑒
𝜔𝑑 = cos−1
𝑇 2
= 10 cos −1 0.75𝑒 0.2 ≈ 4.127𝑟𝑎𝑑/𝑠
• Solve for 𝜁
𝜔𝑑2 1 − 𝜁2 4.127 2
2
= 2
=
𝜁𝜔𝑛 𝜁 22
⇒ 𝜁 = 0.436, 𝜔𝑛 = 4.586 𝑟𝑎𝑑/𝑠
𝐾 = 𝑒 −2𝜁𝜔𝑛 𝑇 − 0.5 = 𝑒 −2×2×0.1 − 0.5 ≈ 0.17 36
 Design (c) 𝜁 = 0.7
• 𝑧1 equation: 1.5 = 2 cos 𝜔𝑑 𝑇 𝑒 −𝜁𝜔𝑛𝑇
= 2 cos 0.0714𝜔𝑛 𝑒 −0.07𝜔𝑛
Solve numerically by trial and error with a calculator
𝜔𝑛 = 3.63 𝑟𝑎𝑑/𝑠
• 𝑧 0 equation: 𝐾 + 0.5 = 𝑒 −2𝜁𝜔𝑛𝑇
𝐾 = 𝑒 −2𝜁𝜔𝑛𝑇 − 0.5 = 𝑒 −0.14×3.63 − 0.5 ≈ 0.1
• Graphically: draw root locus and a segment of the
constant 𝜁 spiral and find their intersection.
• (Rough results and the solution is difficult without
MATLAB for all but a few simple root loci).
37
 Example 6.4
Design a proportional controller for the
unity feedback digital control system with a
sampling period T=0.1 s to obtain
a) 𝑒 ∞ due to ramp = 10%
b) 𝜁 = 0.7
1
𝐺 𝑠 =
𝑠 𝑠+5

38
 Solution
𝑧 + 0.847
𝐺𝑍𝐴𝑆 𝑧 = 4.261 × 10−3
𝑧 − 1 𝑧 − 0.606
• Closed-loop characteristic equation
1 + 𝐾𝐺𝑍𝐴𝑆 𝑧
= 𝑧 2 − 1.606 − 4.261 × 10−3 𝐾 𝑧 + 0.606 − 3.608 × 10−3 𝐾
= 𝑧 2 − 2 cos 𝜔𝑑 𝑇 𝑒 −𝜁𝜔𝑛 𝑇 𝑧 + 𝑒 −2𝜁𝜔𝑛 𝑇
• Equation involves three parameters 𝜁, 𝜔𝑛 & 𝐾
• Equating coefficients yields two equations
• Evaluate two unknowns and obtain the third from the
design specification.
39
 (a) Design for 𝑒 ∞ % = 10%
• Velocity error constant 𝐾𝑣 (Type 1)
1 𝑧−1
𝐾𝑣 = 𝐾𝐺𝑍𝐴𝑆 𝑧 ቉
𝑇 𝑧 𝑧=1
4.261 × 10−3 1 + 0.847 𝐾
= 10𝐾 =
1 − 0.606 5
• Same 𝐾𝑣 as for the analog proportional control
system: 10% steady-state error due to ramp.
𝐾 100 100
= 𝐾𝑣 = = = 10
5 𝑒 ∞ % 10
𝐾 = 50
40
Root locus for 𝐾 = 50

41
Time response for 𝐾 = 50

42
 b) Design for 𝜁 = 0.7
• Difficult analytical solution for constant 
• MATLAB: move cursor to 𝜁 = 0.7 ⇒ 𝐾 = 10
• 𝐾𝑐𝑟 = 109, stable system at 𝐾 = 50 and 𝐾 = 10.
• Design specs. are met for both (a) and (b).
• Must check other design criteria.
• For (a), 𝐾 = 50, 𝜁 = 0.18 (highly oscillatory)
• For (b), 𝐾 = 10, 𝑒 ∞ % = 50% due to unit ramp.
• P-control cannot provide good steady-state
error together with good transient response.

43
Root Locus for Constant 𝜁 Design

44