Chapter Three

Digital Control Analysis

System discretization example
 Consider the Tank control system.
 H- steady state (stable) fluid height in the tank
 h- height perturbation from the nominal value.
 Q-steady state (stable) flow through the tank
 𝑞𝑖 − inflow perturbation from the nominal value.
 𝑞0 −outflow perturbation from the nominal
value.
 Solution:
h = R𝑞0
Assuming an incompressible fluid
𝑑𝐶 ℎ + 𝐻
= 𝑞𝑖 + 𝑄 − 𝑞0 + 𝑄
𝑑𝑡
Cont.
C- area of the tank
Substituting 𝑞0 ,
𝑑ℎ ℎ 𝑞𝑖 𝑑𝐻
+ = , 𝜏 = 𝑅𝐶, = 0 H is constant
𝑑𝑡 𝜏 𝐶 𝑑𝑡
The solution of this differential equation is:
𝑡
1
ℎ 𝑡 = 𝑒 − 𝑡−𝑡0 𝜏 ℎ 𝑡0 + 𝑒 𝑡−𝜆 𝜏 𝑞𝑖 𝜆 𝑑𝜆
𝐶 𝑡0
Let 𝑞𝑖 be constant over each sampling period T,
(from 𝑘𝑇 to (𝑘 + 1)𝑇) then we can solve the analog
equation over any sampling period to obtain:
ℎ 𝑘 + 1 = 𝑒 −𝑇 𝜏 ℎ 𝑘 + 𝑅 1 − 𝑒 −𝑇 𝜏 𝑞𝑖 𝑘
Where the variables at time KT denoted by k
Linear Difference Equation
 𝑒0 , 𝑒1 , 𝑒2 , … 𝑒𝑘 → 𝐼𝑁𝑃𝑈𝑇𝑆
 𝑢0 , 𝑢1 , 𝑢2 , … 𝑢𝑘−1 → 𝑂𝑈𝑇𝑃𝑈𝑇𝑆
 To get the next output the computer
needs to compute some function, can be
expressed as:
 𝑢𝑘 = 𝑓 𝑒0 , 𝑒1 , … 𝑒𝑘 ; 𝑢0 , 𝑢1 , … 𝑢𝑘−1
 𝑢𝑘 = −𝑎1 𝑢𝑘−1 − 𝑎2 𝑢𝑘−2 − ⋯ 𝑎1 𝑢𝑘−𝑛 +
𝑏0 𝑒𝑘 + 𝑏1 𝑒𝑘−1 + ⋯ 𝑏𝑚 𝑒𝑘−𝑚
 Cont.
 The z-transform is an important tool in the analysis
and design of discrete-time systems. It simplifies the
solution of discrete time problems by converting LTI
difference equations to algebraic equations and
convolution to multiplication.
Z-Transform
Definition 1.
given the causal sequence:
𝑢0 , 𝑢1 , 𝑢2 , … , 𝑢𝑘 , … its Z-transform is
 𝑈 𝑘 − 𝑖 = 𝑈𝑖 𝑧 −𝑖 , time delay operator
 𝑈 𝑧 = 𝑢0 + 𝑢1 𝑧 −1 + 𝑢2 𝑧 −2 + ⋯ +𝑢𝑘 𝑧 −𝑘
∞ −𝒌
 𝑼 𝒛 = 𝒌=𝟎 𝒖𝒌 𝒛
Cont.
 Definition 2.
 Given the impulse train representation of a discrete time signal
𝑢∗ 𝑡 = 𝑢0 𝛿 𝑡 + 𝑢1 𝛿 𝑡 − 𝑇 + 𝑢2 𝛿 𝑡 − 2𝑇 + ⋯ + 𝑢𝑘 𝛿 𝑡 − 𝑘𝑇
∞

= 𝑢𝑘 𝛿 𝑡 − 𝑘𝑇
𝑘=0
The Laplace transform of the above equation is:
𝑈 ∗ 𝑠 = 𝑈0 + 𝑈1 𝑒 −𝑠𝑇 + 𝑈2 𝑒 −2𝑠𝑇 + ⋯ + 𝑈𝑘 𝑒 −𝑘𝑠𝑇
∞

= 𝑈𝑘 𝑒 𝑠𝑇 −𝑘
𝑘=0
𝑠𝑇
Let z = 𝑒
∞

= 𝑈𝑘 𝑧 −𝑘
𝑘=0
Examples
1. Obtain the z-transform of the sequence
∞
𝑢𝑘 = 1, 3, 2, 0, 4, 0, 0, 0, …
0
𝑈 𝑧 = 1 + 3𝑧 −1 + 2𝑧 −2 + 4𝑧 −4
2. Unit impulse: 𝑢𝑘 = 1 𝑓𝑜𝑟 𝑘 = 0
3. Sampled Unit step: 𝑢𝑘 = 1 𝑓𝑜𝑟 𝑘 ≥ 0
4. Sampled Exponential: 𝑢𝑘 = 𝑎𝑘 𝑓𝑜𝑟 𝑘 ≥ 0
Sampled Step
1, 𝑘 ≥ 0
 𝑢 𝑘 =
0, 𝑘 < 0
𝑈 𝑧 = 1 + 𝑧 −1 + 𝑧 −2 + ⋯ + 𝑧 −𝑘
For 𝑧 < 1
∞
−𝑘
1 𝑧
𝑈 𝑧 = 𝑧 = −1
=
1−𝑧 𝑧−1
𝑘=0
 If the value z is greater than one the difference equation is
unstable .
 Meaning-for some finite initial conditions the solution will
grow with out bound as time goes to infinity.
 If all the roots of the characteristic equation are inside the
unit circle, the corresponding difference equation is stable.
Cont.
 E.g. is the characteristic equation stable?
𝑢 𝑘 = 0.9𝑢 𝑘 − 1 − 0.2𝑢 𝑘 − 2
 𝑧 𝑘 𝑈 𝑧 = 0.9𝑧 𝑘−1 𝑈 𝑧 − 0.2𝑧 𝑘−2 𝑈 𝑧
 Multiplying by 𝑧 −𝑘
 1 = 0.9𝑧 −1 − 0.2𝑧 −2
 Multiplying by 𝑧 2
 𝑧 2 − 0.9z + 0.2 = 0
 Solutions: z = 0.5 and z = 0.4
 The roots are inside the unit circle, the
equation is stable.
Properties of Z-Transform
1. Linearity:
𝓏 𝛼𝑓1 𝐾𝑇 + 𝛽𝑓1 𝐾𝑇 = 𝛼𝐹1 𝑧 + 𝛽𝐹1 𝑧
2. Convolution of time Sequences:
∞

𝓏 𝑓1 𝑙 𝑓2 𝑘 − 𝑙 = 𝐹1 𝑧 𝐹2 𝑧
𝑙=−∞
3. Time Delay:
∞

𝓏 𝑓 𝑘−𝑛 = 𝑓 𝑘 − 𝑛 𝑧 −𝑘 = 𝑧 −𝑛 𝐹 𝑧
𝑘=−∞
Example: Find the z-transform of the causal sequence
4, 𝑘≥2
𝑓 𝑥 =
0, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
Cont.

4. Multiplication by Exponential:
𝓏 𝑟 −𝑘 𝑓 𝑘 = 𝐹 𝑟𝑧
5. Inversion:
 Long Division
 Partial Fraction Expansion
Cont.
 Find the z-transform of the exponential
sequence
𝑓 𝑥 = 𝑒 −𝑎𝑘𝑇 , 𝑘 = 0,1,2, …
Cont.
6.Final value theorem
 𝑓 𝑘 =𝑓 ∞ +𝑔 𝑘 , 𝑘 = 0,1,2, …
 𝑔 𝑘 − a sequence that decays to zero as k tends to infinity.
lim 𝑓 𝑘 = 𝑓 ∞
𝑘→∞
The z- transform of 𝑓 𝑘 = 𝑓 ∞ + 𝑔 𝑘 ,
𝑓 ∞ 𝑧
𝐹 𝑧 = +𝐺 𝑧
𝑧−1
𝑧−1 𝑓 ∞ 𝑧
lim +𝐺 𝑧 =𝑓 ∞
𝑧→1 𝑧 𝑧−1
Final-Value Theorem:
Poles of 𝐹 𝑧 are inside the unit circle if
lim 𝑓 𝑘 = lim 𝑧 − 1 𝐹 𝑧 = 𝑓 ∞
𝑘→∞ 𝑧→1
Discrete Integration
𝑡
 𝑢 𝑡 = 0
𝑒 𝑡 𝑑𝑡
Forward rectangular rule
𝐴 = 𝑡𝑘 − 𝑡𝑘−1 𝑒𝑘−1
𝑢 𝑘 = 𝑢 𝑘 − 1 + 𝑇𝑒𝑘−1 ,𝑇 = 𝑡𝑘 − 𝑡𝑘−1
Backward rectangular rule
𝐴 = 𝑡𝑘 − 𝑡𝑘−1 𝑒𝑘
𝑢 𝑘 = 𝑢 𝑘 − 1 + 𝑇𝑒𝑘
Cont.
Trapezoid rule
𝑒𝑘 + 𝑒𝑘−1
𝐴 = 𝑡𝑘 − 𝑡𝑘−1
2
𝑇
𝑢 𝑘 = 𝑢 𝑘 − 1 + 𝑒𝑘 + 𝑒𝑘−1
2
Transfer Function
𝑏0 + 𝑏1 𝑧 −1 + ⋯ + 𝑏𝑚 𝑧 −𝑚
𝐻 𝑧 =
1 + 𝑎1 𝑧 −1 + 𝑎2 𝑧 −2 + ⋯ +𝑎𝑛 𝑧 −𝑛
If 𝑛 ≥ 𝑚
𝑏0 𝑧 𝑛 + 𝑏1 𝑧 𝑛−1 + ⋯ + 𝑏𝑚 𝑧 𝑛−𝑚 𝑏 𝑧
𝐻 𝑧 = 𝑛 𝑛−1 𝑛−2
=
𝑧 + 𝑎1 𝑧 + 𝑎2 𝑧 + ⋯ +𝑎𝑛 𝑎 𝑧
𝑈 𝑧
𝐻 𝑧 =
𝐸 𝑧
 Zeros - 𝑏 𝑧 = 0
 Poles - 𝑎 𝑧 = 0
Linearity and Stationarity
 Linearity: A system with input e and
output u is linear if superposition applies.
𝑢1 𝑘 𝑖𝑠 𝑡ℎ𝑒 𝑟𝑒𝑠𝑝𝑜𝑛𝑠𝑒 𝑜𝑓 𝑒1 𝑘
𝑢2 𝑘 𝑖𝑠 𝑡ℎ𝑒 𝑟𝑒𝑠𝑝𝑜𝑛𝑠𝑒 𝑜𝑓 𝑒2 𝑘
 Then for any scalar 𝛼 𝑎𝑛𝑑 𝛽
The response to 𝛼𝑒1 𝑘 + 𝛽𝑒2 𝑘 is
𝛼𝑢1 𝑘 + 𝛽𝑢2 𝑘 .
Cont.
 Stationarity: A system is stationary or
time invariant, if a time shift in the input
results in only a time shift in the output.
Impulse response
 the response of a discrete time system to
a unit impulse is known as the impulse
response sequence.
 𝑢 𝑘 = 𝑢 0 ,𝑢 1 ,…,𝑢 𝑖 ,…
Cont.
 To derive this relationship, we first represent
the input sequence in terms of discrete
impulses as follows:
 𝑢 𝑘 =𝑢 0 𝛿 𝑘 +𝑢 1 𝛿 𝑘−1 +
𝑢 2 𝛿 𝑘 − 2 + ⋯+ 𝑢 𝑖 𝛿 𝑘 − 𝑖 + ⋯
∞
 = 𝑖=0 𝑢 𝑖 𝛿 𝑘 − 𝑖
 For a linear system, the principle of
superposition applies and the system output
due to the input is the following sum of
impulse response sequence.
𝑦 𝑙 =ℎ 𝑙 ∗𝑢 𝑙
Cont.
 𝑦 𝑙 = ℎ 𝑙 𝑢 0 + ℎ 𝑙−1 𝑢 1 +
ℎ 𝑙 − 2 𝑢 2 + ⋯+ ℎ 𝑙 − 𝑖 𝑢 𝑖
 The output at time k is given by
∞

𝑦 𝑘 =ℎ 𝑘 ∗𝑢 𝑘 = ℎ 𝑘−𝑖 𝑢 𝑖
𝑖=0
 For a causal system: the response due to an
impulse at time i is an impulse response
starting at time i and the delayed response
ℎ 𝑘 − 𝑖 satisfies
 ℎ 𝑘 − 𝑖 = 0 ,𝑖 > 𝑘
Cont.
In other words a causal system is one
whose impulse response is a causal time
sequence .
 𝑦 𝑘 = 𝑢 0 ℎ 𝑘 + 𝑢 1 ℎ 𝑘−
Cont.
Given the discrete time system
𝑦 𝑘 + 1 − 0.5𝑦 𝑘 = 𝑢 𝑘 , 𝑦 0 = 0
Find the impulse response of the system
ℎ 𝑘 :
1. From the difference equation
2. Using z-transform
Cont.
Stability of Digital Control Systems
1) Asymptotic Stability: a system is
Asymptotically stable if the response to any
initial conditions decays to zero
asymptotically.
lim 𝑦 𝑘 = 0
𝑘→∞
Ifthe response due to the initial
conditions remains bounded but does not
decay to zero, the system is said to be
marginally stable.
Cont.
2) Bounded-Input-Bounded-Output
Stability:
A system is said to be bounded-input-
bounded-output(BIBO) stable if its
response to any bounded input remains
bounded.
𝑢 𝑘 < 𝑏𝑢 , 𝑘 = 0,1,2, … }𝐼𝑁𝑃𝑈𝑇 0 < 𝑏𝑢 < ∞
𝑦 𝑘 < 𝑏𝑦 , 𝑘 = 0,1,2, … }𝑂𝑈𝑇𝑃𝑈𝑇 0 < 𝑏𝑦 < ∞
Stable Z-Domain Pole Location
Asymptotic Stability: in the absence of
pole-zero cancellation, an LTI digital
system is asymptotically stable if its
transfer function poles are inside the unit
circle and marginally stable if the poles
are in the unit circle with no repeated
poles on the unit circle.
Cont.
 Example1:
4 𝑧−2
𝐻 𝑧 = , not stable
𝑧−2 𝑧−0.1
4 𝑧−0.2
𝐻 𝑧 = , Asymptotically stable
𝑧−0.2 𝑧−0.1
5 𝑧−0.3
𝐻 𝑧 = , Asymptotically stable
𝑧−0.2 𝑧−0.1
8 𝑧−0.2
𝐻 𝑧 = , Marginally stable
𝑧−0.1 𝑧−1
BIBO Stability
 A discrete time linear system is BIBO
stable if and only if its impulse response
sequence is absolutely summable.
∞

ℎ 𝑖 <∞
𝑖=0
 A discrete time linear system is BIBO
stable if and only if the poles of its
transfer function lie inside the unit circle.
Frequency Response of Discrete
Time Systems
 Discrete time Frequency response is a
response of a system to a sampled
sinusoidal input.
Cont.
 It is now possible to define a transfer function for sampled
inputs as:
∗
∗
𝑌 𝑠
𝐻 𝑠 = ∗ 𝑎𝑡 𝑧𝑒𝑟𝑜 𝑖𝑛𝑖𝑡𝑖𝑎𝑙 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛
𝑈 𝑠
𝐻 ∗ 𝑗𝜔 = 𝐻∗ 𝑠 |𝑠=𝑗𝜔
In terms of the complex variable 𝑧 = 𝑒 𝑠𝑇
𝐻 𝑧 = 𝐻∗ 𝑠 | 1
𝑠=𝑇𝑙𝑛 𝑧
Thus, the frequency response is given by
𝐻 ∗ 𝑗𝜔 = 𝐻 𝑧 |𝑧=𝑒 𝑗𝜔𝑇
= 𝐻 𝑒 𝑗𝜔𝑇
 𝐻 𝑒 𝑗𝜔0 𝑇 ∠𝐻 𝑒 𝑗𝜔0 𝑇

 𝑒 𝑗𝜔0 𝑇 = cos 𝜔0 𝑇 + 𝑗 sin 𝜔0 𝑇

Frequency response of first order
system
𝑧
 𝐻 𝑧 =
𝑧−𝑒 −𝑎𝑇
 𝑧 = 𝑒 𝑗𝜔𝑇
𝑒 𝑗𝜔𝑇 cos 𝜔0 𝑇+𝑗 sin 𝜔0 𝑇
 𝐻 𝑧 = =
𝑒 𝑗𝜔𝑇 −𝑒 −𝑎𝑇 cos 𝜔0 𝑇+𝑗 sin 𝜔0 𝑇−𝑒 −𝑎𝑇
cos 𝜔0 𝑇+𝑗 sin 𝜔0 𝑇 cos𝜔0 𝑇−𝑒 −𝑎𝑇 −𝑗 sin 𝜔0 𝑇

cos𝜔0 𝑇−𝑒 −𝑎𝑇 +𝑗 sin 𝜔0 𝑇 cos𝜔0 𝑇−𝑒 −𝑎𝑇 −𝑗 sin 𝜔0 𝑇
cos 𝜔0 𝑇 cos𝜔0 𝑇−𝑒 −𝑎𝑇 + sin 𝜔0 𝑇 2
 +
cos𝜔0 𝑇−𝑒 −𝑎𝑇 2 + sin 𝜔0 𝑇 2
cos𝜔0 𝑇−𝑒 −𝑎𝑇 sin 𝜔0 𝑇−cos𝜔0 𝑇 sin 𝜔0 𝑇
𝑗
cos𝜔0 𝑇−𝑒 −𝑎𝑇 2 + sin 𝜔0 𝑇 2
Cont.
 Frequency response design approaches,
especially design based on bode plots, are
very popular in the design of control
systems.
 The fact that the bode plots of rational loop
gains can be approximated by straight lines, if
the transfer function is minimum phase, the
phase can be determined from the plot of
the magnitude, and this allows as to simplify
the design of a compensator that provides
specified stability properties.
Cont.
 Unfortunately, discrete transfer functions are
not rational functions in 𝑗𝜔 because the
frequency is introduced through the
substitution 𝑧 = 𝑒 𝑗𝜔𝑇 in the z-transfer
function. Hence, the simplification provided
by methods based on Bode plots for analog
system is lost. A solution to this problem is
to bilinearlly transform the z-plane into a
new plane, called the W-plane, where the
corresponding transfer function is rational
and where the Bode approximation is valid.
Cont.
𝑧−1 2
 𝑤 =𝑐 , 𝑤ℎ𝑒𝑟𝑒 𝑐=
𝑧+1 𝑇
𝑤𝑇
1+
2
 𝑧= 𝑤𝑇
1−
2
2 𝜔𝑇
 𝑤 = 𝑗𝑣 = j tan
𝑇 2
 This transforms the transfer function of a
system from the z-plane to the w-plane.
 Then we can plot the frequency response in
the someway we plot continuous time
frequency response.
Example 1
1
𝐺 𝑠 =
𝑠+1
At T=0.1
0.09516
𝐺 𝑧 =
𝑧 − 0.9048
−0.05𝑤 + 1
𝐺1 𝑤 =
𝑤+1
At T=0.001
0.00995
𝐺 𝑧 =
𝑧 − 0.99
−0.005𝑤 + 1
𝐺2 𝑤 =
𝑤+1
Cont.
 By calculating values of w for
corresponding values of 𝜔 at sample
points and using the w-plane transfer
function it is possible to plot the
frequency response Bode plot.
Cont.
 Discrete State-Space
Continuous time State-Space
𝑋 𝑡 = 𝐴𝑋 𝑡 + 𝐵𝑈 𝑡
𝑌 𝑡 = 𝐶𝑋 𝑡 + 𝐷𝑈 𝑡
The general solution of continuous time difference equation is
𝑡𝑓
𝑋 𝑡𝑓 = 𝑒 𝐴 𝑡𝑓 −𝑡0
𝑋 𝑡0 + 𝑒𝐴 𝑡𝑓 −𝜏
𝐵𝑈 𝜏 𝑑𝜏
𝑡0
Let the initial time 𝑡0 = 𝑘𝑇, and the final time
𝑡𝑓 = 𝑘 + 1 𝑇, the discrete time solution becomes;
𝑘+1 𝑇
𝑋 𝑘 + 1 = 𝑒 𝐴𝑇 𝑋 𝑘 + 𝑒𝐴 𝑘+1 𝑇−𝜏
𝐵𝑈 𝑘 𝑑𝜏
𝑘𝑇
Let, λ = 𝑘 + 1 𝑇 − 𝜏, so dλ = −𝑑𝜏
0 𝑇
𝑒 𝐴λ 𝐵𝑈 𝑘 −𝑑λ = 𝑒 𝐴λ 𝐵𝑈 𝑘 𝑑λ
𝑇 0
Cont.
𝑋 𝑘 + 1 = 𝐴𝑑 𝑋 𝑘 + 𝐵𝑑 𝑈 𝑘
𝑌 𝑘 = 𝐶𝑑 𝑋 𝑘 + 𝐷𝑑 𝑈 𝑘
𝐴𝑇 𝑇 𝐴λ
𝐴𝑑 = 𝑒 , 𝐵𝑑 = 0
𝑒 𝐵𝑑λ
𝐴𝑑 -is the discrete time state matrix
𝐵𝑑 -is discrete input matrix
The discrete time state matrix is the state
transition matrix for the analog system
evaluated at the sampling period T.
Discretization method 1
 The most important function is the
exponential function 𝑒 𝐴𝑡 .
 Based on Taylor series,
2 2 𝑛 𝑛
𝜆𝑡
𝜆 𝑡 𝜆 𝑡
𝑒 = 1 + 𝜆𝑡 + +⋯+ …
2! 𝑛!
Converges for all finite 𝜆 and t, therefor
2 2 𝑛 𝑛
𝐴𝑇
𝐴 𝑇 𝐴 𝑇
𝑒 = 𝐼 + 𝐴𝑇 + + ⋯+ …
2! 𝑛!
∞
1 𝑘 𝑘
= 𝑡 𝐴
𝑘!
𝑘=0
Cont.
𝑇 2 3 4
𝑇 𝑇 𝑇
𝑒 𝐴λ 𝑑λ = 𝑇𝐼 + 𝐴+ 𝐴2 + 𝐴3 …
0 2! 3! 4!
𝑇
𝐴λ
𝑇2 𝑇3 2 𝑇4 3
𝐵𝑑 = 𝑒 𝐵𝑑λ = 𝑇𝐼 + 𝐴+ 𝐴 + 𝐴 … 𝐵
0 2! 3! 4!
Discretization method 2
 For invertible state matrix A, the integral of the
matrix exponential given by;
𝑒 𝐴𝑇 𝑑𝑡 = 𝐴−1 𝑒 𝐴𝑡 − 𝐼𝑛 = 𝑒 𝐴𝑡 − 𝐼𝑛 𝐴−1

Applying the above matrix property,

𝑇
𝐵𝑑 = 𝑒 𝐴λ 𝐵𝑈 𝑘 𝑑λ = 𝑒 𝐴𝑡 − 𝐼𝑛 𝐴−1 B
0
Solution of Discrete State Space
𝑋 1 = 𝐴𝑑 𝑋 0 + 𝐵𝑑 𝑈 0
𝑋 2 = 𝐴𝑑 𝑋 1 + 𝐵𝑑 𝑈 1
:
𝑋 2 = 𝐴𝑑 𝐴𝑑 𝑋 0 + 𝐵𝑑 𝑈 0 + 𝐵𝑑 𝑈 1
= 𝐴𝑑 2 𝑋 0 + 𝐴𝑑 𝐵𝑑 𝑈 0 + 𝐵𝑑 𝑈 1
2−1

𝑋 2 = 𝐴𝑑 2 𝑋 0 + 𝐴𝑑 2−𝑖−1
𝐵𝑑 𝑈 𝑖
𝑖=0
 The equation generalized to;
𝑘−1

𝑋 𝑘 = 𝐴𝑑 2 𝑋 0 + 𝐴𝑑 𝑘−𝑖−1
𝐵𝑑 𝑈 𝑖
𝑖=0
 The Discrete time output becomes;
𝑘−1

𝑌 𝑘 = 𝐶𝑑 𝐴𝑑 2 𝑋 0 + 𝐴𝑑 𝑘−𝑖−1
𝐵𝑑 𝑈 𝑖 + 𝐷𝑑 𝑈 𝑘
𝑖=0