09 Signals and Systems 4up

Signals, Systems, and Control (Part #1)
Review: Digital Systems

1 System Diagram
Signals, Systems, and Control
“system”
input output
Dr. Edmund Lam
with help from Dr. Hayden So and Prof. Y.S. Hung 2 Digital Logic Primitives
Department of Electrical and Electronic Engineering AND OR NOT D Flip-flop
The University of Hong Kong
D Q
ENGG1015: Introduction to Electrical and Electronic Engineering clk
(First Semester, 2012–13)
http://www.eee.hku.hk/˜engg1015 3 Schematics ⇐⇒ Boolean expressions
y = a · b̄ + ā · b
E. Lam (University of Hong Kong) ENGG1015 November, 2012 1 / 86 E. Lam (University of Hong Kong) ENGG1015 November, 2012 2 / 86
Signals, Systems, and Control (Part #1) Signals, Systems, and Control (Part #1)
Signals and Systems Examples
Communication system Imaging system

The system diagram applies not only to digital logic (input/output not signal: light intensity
signal: digitized voice (or image or
necessarily binary). video or text) system: consists of lenses,
system: consists of antenna, base photodetectors
stations, etc output = input? e.g. lens distortion
“system” output = input? e.g. background
input “signal” output “signal”
noise filtering
Financial system Biomedical system

Also called a block diagram
signal: $$ e.g. Electroencephalography, or EEG
A system maps an input signal to an output signal
system: consists of lots of signal: electrical (voltage)
A “signals and systems” abstraction computers system: consists of electrodes and
Examples: Combinational logic system, computer system, . . . output = input? A complicated other components
function! output = input?
A “Signals and Systems” Viewpoint Discrete-time Systems
Very diverse applications “system”

input “signal” output “signal”
A similar viewpoint: flow of information
Abstraction: same “language” (mathematics) to deal with signals
A system can be very complicated
and systems
Goal is to describe the system with simple “behaviors” or “rules”
Good news: what you learn here is very useful
We restrict ourselves to discrete-time systems, i.e., input and
(or, why many electrical and electronic engineers are in Wall
output signals are a sequence of numbers
Street, Medicine, Business, Music, Psychology, Linguistics, . . . )
Conceptually, such discrete systems can be built by representing the
numbers in binary and using lots and lots of combinational and
sequential logic gates
Four Different Representations #1: Signal flow graph
We restrict ourselves to three “primitives”:

Objective: Develop different ways to represent and understand a
discrete-time system a k·a
1 Multiplication (gain): k
(k can be integer, fraction, negative number. . . )
“Time-domain” methods:
Signal flow graph a a b b+c
+
Linear constant-coefficient difference equation (LCCDE) 2 Split/add (adder):
a c
“Transform” methods: (A value becomes two identical copies)
Operator notation and arithmetics (Two values added together)
z-transform and transfer function
{a, b, . . .} {0, 0, . . . , 0, a, b, . . .}
3 Delay: delay (d)
They are equivalent representations, but differ in their use. d of them
(The sequence is delayed by d integer units)
#1: Signal Flow Graph Example #1: Signal Flow Graph Example
Example (A):
Example (A): (cont.)
input = {0, 3, −1, 5, 2, 0} output =?
1
2 +
5
4 input
delay (1) 1 output
2 3
2
Assume: The system has no signal before the input.
1
1
0→ 2 · (0) = 0
1 1 3
sequence order
3→ 2 · (3) + 2 · (0) = 2 0 1 2 3 4 5
−1 → 1
· (−1) + 1
· (3) = 1 -1
2 2
1 1
5→ 2 · (5) + 2 · (−1) = 2
2→ 1
2 · (2) + 1
2 · (5) = 7
2 Observation: This output is a smoothed version of this input
1 1
0→ 2 · (0) + 2 · (2) = 1
Deduction: This discrete-time system achieves smoothing
Hence, output is {0, 32 , 1, 2, 72 , 1}.
#1: Signal Flow Graph Example #1: Signal Flow Graph Example
Example (B):
Example (B): (cont.)
input = {1, 0, 0, 0, . . .} output =?
+ 3 3 3
2 2 2
1 1 1
p delay (1)
0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5
Assume: The system has no signal before the input.
p = 0.8 p = 1.0 p = 1.2
1→1=1
0 → 0 + p · (1) = p
Observation: This output “decays” or “stays unchanged” or “grows
0 → 0 + p · (p) = p2 without bound” for a unit input
0 → 0 + p · (p2 ) = p3
0 → 0 + p · (p3 ) = p4 Deduction: The behavior of this discrete-time system depends a lot on
the value of p. This is extremely important for us later on!
Hence, output is {1, p, p2 , p3 , . . .}.
#1: Signal Flow Graph Mini-conclusions #2: Difference Equations

Look at the input sequence as a discrete-time signal
Some conclusions about flow graphs:
1 Given an input, we can “follow the flow” to deduce the output 5
2 Hardware implementation by putting in the appropriate 4
components (assume we have them) according to the flow graph 3
3 It is not always obvious what the system achieves, but we can guess 2
4 Intuitively, some changes to the flow graphs are permitted: 1
k delay (d) ⇔ delay (d) k 0

n
1 2 3 4 5 6 7 8 9 10
k k Data present at regular time intervals

⇔ Often due to sampling of a continuous signal
Some data are naturally discrete, e.g. daily stock price
k
For our purpose here, we’ll just assume the input and output signals are
discrete without worrying how they come to be
#2: Difference Equations: The Mathematics #2: Difference Equations: The Mathematics
Exercise: Try plotting the following signals
Conventions:
1 x[n] = Cαn , where n = 0, . . . , 99.
Signal: x[n] (square bracket) 1 α = 0.99, C = 10
Often n = 0, 1, . . . N − 1 for a length-N signal. 2 α = 1.00, C = 1
Assume x[n] = 0 outside this range. 3 α = 1.01, C = 0.1
Use x[n] for an input signal, y[n] for an output signal 2 x[n] = A cos(ωn + φ), where n = 0, . . . , 99.
1 A = 1, φ = 0, ω = 0.01π (low frequency)
2 A = 1, φ = 0, ω = 0.1π (middle frequency)
Alternatives:
3 A = 1, φ = 0, ω = π (high frequency)
1 Can have negative n, e.g., n = −N, −N + 1, . . . N − 1, N for a length 4 A = 1, φ = π/2, ω = π
2N + 1 signal
1 at n = 0
Can have infinite length signal (conceptually): n = all integers
3 x[n] =
2
0 otherwise.
This is the most important signal of all! It is called a delta function or a
unit impulse, denoted by δ[n].
#2: Difference Equations and Flow Graphs #2: Difference Equations Example
Example (A):
The flow graphs now operate on the entire signal (vector vs scalar)
1
x[n] k · x[n] x[n] 2 x[n] y[n] = 12 x[n] + 12 x[n − 1]
1 Multiplication (gain): k 1
2 +
1
x[n − 1] 2 x[n − 1]
delay (1) 1
x[n] x[n] x1 [n] x1 [n] + x2 [n] 2
+
2 Split/add (adder):
x[n] x2 [n]
Example (B):
(A signal becomes two identical copies)
(Two signals added together) x[n] y[n] = p · y[n − 1] + x[n]
+
x[n] x[n − d]
3 Delay: delay (d)
p · y[n − 1] y[n − 1] y[n]
p delay (1)
(A signal is delayed by d integer units)
#2: Difference Equations Example #2: Difference Equations Implementation

Example (C): The smoothing system in cascade
Software implementation: e.g. y[n] = p · y[n − 1] + x[n]
x[n] w[n] y[n]
1
+ 1
+
2 2 “Pseudo-code” in the computer
delay (1) 1
delay (1) 1 outside outside
2 2
p = 0.5;
Tedious to go through the flow graph for each point. Make use of y(-1) = 0;
A/D D/A
difference equations: n = 0;
1 1
LOAD x;
w[n] = x[n] + x[n − 1] REPEAT {
2 2
1 1 y(n) = p * y(n-1) + x(n);
y[n] = w[n] + w[n − 1]
2 2
1 1 1 1
n = n + 1;
= x[n] + x[n − 1] + x[n − 1] + x[n − 2] };
4 4 4 4
1
= x[n] + 2x[n − 1] + x[n − 2] .
4
Effect: further smoothing
#2: Difference Equations Implementation #2: Difference Equations Mini-conclusions
Real-time software implementation: e.g. y[n] = p · y[n − 1] + x[n] Some conclusions about difference equations:
“Pseudo-code” in the computer 1 Flow graphs and difference equations are equivalent
Can go from flow graphs to difference equations
outside outside Can go from difference equations to flow graphs (Do you know how?)
p = 0.5; 2 They correspond to different ways of implementation
y(old) = 0;
A/D D/A Flow graph is more “hardware”
REPEAT { Difference equation is more “software”
x(new) = INPUT;
y(new) = p * y(old) + x(new); 3 A general form of the difference equation:
OUTPUT = y(new);
y(old) = y(new); y[n] = a1 y[n − 1] + a2 y[n − 2] + . . . + b0 x[n] + b1 x[n − 1] + . . .
};
Hence, the name “constant-coefficient”
#1– #2: Time-domain Conclusions #3: Operator Notation and Arithmetics

Motivation: What does y[n] = 12 x[n] + 12 x[n − 1] mean?
Both flow graphs and difference equations are “time-domain” methods 5

y[3] = 12 (x[2]) + 12 (x[3])
4
1 Iterate for different n, e.g., for n = 3:
3
We compute the results as time passes 2
Generally speaking, these methods allow for easy computation of 1
the output 0 1 2 3 4 5
-1
Not so good with understanding the system behavior
2 Scale and add the entire signal together, i.e.,
As a result, if we want to analyze (understand) and even design a
5 5 5
discrete-time system, we need more advanced tools. 4 4 4
3 × 12 + 3 × 12 = 3
Next, we turn to “transform” methods. 2 2 2
1 1 1
0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5
-1 -1 -1
#3: Operator Notations #3: Operator Notations
A very important operation is called a delay operation, written as

Assume we have the following: Y = D{X}. It means simultaneously the following:
1 X represents the entire input signal (x[n] for all n). y[0] = 0
2 Y represents the entire output signal (y[n] for all n). y[1] = x[0]
y[2] = x[1]
3 Y = S{X} represents an operation:
..
.
X Y = S{X}
S{·}
y[n] = x[n − 1]
From X, operate on it (called S), and output Y.
Often write this way instead of y[n + 1] = x[n].
e.g. Can think about S in qualitative terms: take X, halving every
value, and then add to a delayed version of X, also halving every value. In this specific case, we can omit the bracket and write Y = DX. It
behaves like ordinary multiplication: e.g. D(X1 + X2 ) = DX1 + DX2 .
#3: Operator Arithmetics #3: Operator Example

The flow graphs now acts as operators
X k·X Example (A):
1 Multiplication (gain): k
(k can be integer, fraction, negative number. . . ) 1
Y = 12 X + 12 DX
X 1 2X
2 +
X X X1 X1 + X2
+
2 Split/add (adder): 1
X X2 DX 1 2 DX
delay (1) 2
(Two identical copies / signals added together)
X Dd X We are allowed to write Y = 1

(1 + D) X. How to interpret this?
3 Delay: delay (d) 2

There is a direct correspondence with y[n] = 12 x[n] + x[n − 1] .
(We write Dd X to mean D{. . . D{ X } . . .})

d of them
#3: Operator Example #3: Operator Example
We can tackle composite flow graphs using operator arithemtics.

We can generalize the cascade system to the following:
Example (B): The smoothing system in cascade
X W Y X W Y
1
+ 1
+ S1 S2
2 2
delay (1) 1 delay (1) 1

2 2
W = S1 X
Y = S2 S1 X
W = 12 (1 + D)X
This gives us a “hierarchical” way of analyzing a system: look at the
Y = 12 (1 + D)W = 12 (1 + D) · 12 (1 + D)X = 14 (1 + 2D + D2 )X “big picture”, then more details of each sub-system, and further details
of each sub-system, etc.
Each of the sub-system is S = 12 (1 + D) and Y = S2 X.
#3: Operator Example #3: Operator Example

These expressions are algebraically equivalent:
Exercise: What are the respective operator equations? Y = (1 + D)2 (1 − D)X
1 X Y
+ + + Y = (1 + D)(1 − D2 )X
delay (1) delay (1) −1 delay (1) Y = (1 + 2D + D2 )(1 − D)X
Y = (1 + D − D2 − D3 )X
2 X Y
+ +
Hence the earlier three flow graphs are equivalent, as is the following:
delay (1) −1 delay (2)
X Y
3 X Y +
+ +
delay (1) +
−1 delay (1) 2 delay (1) +
−1 delay (2) +
delay (2)
delay (3)
#3: Operator: Output #3: Operator Arithmetics for Feedback
Operation notation provides a powerful tool to analyze feedback,

Once we understand the system, we can compute the specific output when we
where output is “looped back” to the input.
are given a particular input.
Flow graph:
Often, we are concerned with an impulse input: x[n] is zero
everywhere, except x[0] = 1. x[n]/X y[n]/Y
+
Example: What’s the output of the previous system with x[n] = δ[n]?
Y = X + DX − D2 X − D3 X ;
p delay (1)
Dk X represents δ[n − k] ; therefore,
Output is y[n] = δ[n] + δ[n − 1] − δ[n − 2] − δ[n − 3].
Difference equation: y[n] = py[n − 1] + x[n]
You can try tracing the flow graph or the difference equation to see if you can
arrive at the same result. Operator equation: Y = pDY + X, or, (1 − pD)Y = X
#3: Operator Arithmetics for Feedback #3: Operator Arithmetics for Feedback
If we can work on D algebraically, we can express Y in terms of X: The corresponding flow graph:
X Y

+
1
Y= X
1 − pD p delay (1) +
p2 delay (2) +
What does it mean by “performing the delay operation in the
denominator”?
.. .. ..
Ans: Use the relationship X Y
Alternatively: +
(1 − pD)(1 + pD + p2 D2 + . . .) = 1. (why?)
p delay (1) +
Therefore, p delay (1) +

Y = (1 + pD + p2 D2 + . . .)X.
.. ..
#3: Operator Arithmetics for Feedback #3: Operator Arithmetics for Feedback
“Stable” only if |p| < 1:
3 3 3
If x[n] = δ[n], what is y[n]? 2 2 2
Ans: Y = X + pDX + p2 D2 X + . . . implies 1 1 1
0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5
y[n] = δ[n] + pδ[n − 1] + p2 δ[n − 2] + . . .
p = 0.8 p = 1.0 p = 1.2
We had the same result when we studied the flow graph before! For 3 3 3
this feedback system, {1, 0, 0, . . .} −→ {1, p, p2 , . . .}. 2 2 2
1 1 1
Value of p determines whether the output in response to an impulse input is

0 2 3 5 0 2 3 5 0 2 3 5
stable or not! In particular, three cases: |p| < 1, |p| = 1, |p| > 1.
1 4 1 4 1 4
-1 -1 -1
-2 -2 -2
-3 -3 -3
p = −0.8 p = −1.0 p = −1.2
#3: Operator Mini-conclusions #4: z-transform and Transfer Function
We now have a good tool to analyze feedback systems. Two (related) questions:
The key to understanding the behavior of a feedback system: think
1 How to mathematically represent X that incorporates “the whole
about what happens to the signal when it goes through a loop or a cycle signal”?
2 How to mathematically represent the operations “multiplication”,
|p| < 1: signal weakens after the loop, so output decays
|p| = 1: signal magnitude remains the same after the loop, so output “addition”, and “delay”?
maintains
|p| > 1: signal is amplified after the loop, so output grows A brilliant way: X is a polynomial where the coefficients are the
various values of x[n].
We are most concerned with an input that is a unit impulse: it’s
called the impulse response convention to use negative power
Example: x[n] = {3, 4, 1, −1}
Feedback gives rise to a persistent response with only a transient input
The system has a “similar” behavior as long as the input is of a X(z) = (3) + (4)z−1 + (1)z−2 + (−1)z−3
finite duration. (Why?)
X is now a polynomial in terms of z
#4: z-transform Notations #4: z-transform Notations
From x[n] to X(z), we call it the “z-transform”: The z-transform notation makes it very convenient to represent system
operations.
X(z) = x[0] + x[1]z−1 + x[2]z−2 + x[3]z−3 + . . .
A delay of one unit is equivalent to multiplication with z−1 , since
Using the summation symbol, we can write
z−1 X(z) = x[0]z−1 +x[1]z−2 + x[2]z−3 + x[3]z−4 + . . .

∞
value at time n = 1
X(z) = x[n] z−n
n=0
A delay of d units is multiplication with z−d .
We “build” X(z) by putting putting x[n] as its coefficients; conversely, This is very similar to the operator notation D! You can think of z-transform
we can recover x[n] by reading off the coefficients in X(z). as a “practical way” of realizing the operations.
#4: z-transform Arithmetics #4: z-transform Example
X, X1 , and X2 are now z-transforms of the time-domain signals. We Example (A):

omit the “(z)” when there is no ambiguity.
X k·X X
1
2X Y = 12 X + 12 z−1 X
1
1 Multiplication (gain): k 2 +
1 −1
z−1 X 1 2z X
X1 X1 + X2 delay (1) 2
X X
+
2 Split/add (adder):
X X2

(Two identical copies / signals added together) We can write Y = 1
2 1 + z−1 X.
There are direct correspondences with:
−d
X z X Y= 1
(1 + D) X
3 Delay: delay (d) 2
y[n] = 12 x[n] + x[n − 1]
#4: z-transform Example #4: Transfer Function
Example (B): Define the transfer function H(z) =

Y(z)
X(z) . (We’ll also omit “(z)” later on)
X Y = pz−1 Y + X
+ input X output Y = HX
H
pz−1 Y z−1 Y Y
p delay (1)

Example (A): H = 1
2 1 + z−1 .
1
1 − pz−1 Y = X Example (B): H = .
1 − pz−1
1
Y= X The transfer function “blows up” when z = p. Hence, we call p the “pole” of
1 − pz−1 the system.
#4: z-transform Example #4: z-transform Example

Example (C): A second-order feedback system, first with flow graph. Using z-transform (g1 = 34 , g2 = − 18 ):
Assume g1 = 34 and g2 = − 18 :
X Y = g1 z−1 Y + g2 z−2 Y + X
x[n] y[n] +
+
g1 z−1 Y z−1 Y Y
+ g1 delay (1) + g1 delay (1)
g2 delay (2) g2 z−2 Y

g2 z−2 Y
delay (2)
1→1
0 → 0 + 34 (1) = 3
4
Algebra with polynomials provides a powerful tool for analysis!
0 → 0 + 34 ( 34 ) + (− 18 )(1) = 7
16

1 − 34 z−1 + 18 z−2 Y = X
3 7 1 3 15
0→0+ 4 ( 16 ) + (− 8 )( 4 ) = 64
0→0+ 3 15 1 7 31
4 ( 64 ) + (− 8 )( 16 ) = 256

1 − 12 z−1 1 − 14 z−1 Y = X
Do you see a pattern?

Verify with flow graph using {1, 0, 0, . . .} as input:
We can convert the above second-order feedback system to the
following equivalent form: cascade of two first-order feedback system x[n] w[n] y[n]
+ +
X W Y
+ +
1 delay (1) 1 delay (1)
2 4
1 delay (1) 1 delay (1)
2 4
1→ 1 →1

1 − 12 z−1 W = X 0→ 0 + 12 (1) = 1
2 → 12 + 14 (1) = 3
4
1 1 3 7
1 − 14 z−1 Y = W 0→ 0 + 12 ( 12 ) = 1
4 → 4 + 4 ( 4 ) = 16
0→ 0 + 12 ( 14 ) = → 1 1 7 15
8 + 4 ( 16 ) = 64
1
=⇒ 1 − 12 z−1 1 − 14 z−1 Y = X 8
1 1 15 31
0→ 0 + 12 ( 18 ) = 1
16 → 16 + 4 ( 64 ) = 256
A corollary: Can also interchange the order of the two systems

Another equivalent form: two first-order feedback system in parallel
Further algebra reveals yet another form! X Y
+ 2 +
1
Y= X
1 − 12 z−1 1 − 14 z−1 1
2
delay (1)
⎛ ⎞
⎜⎜ 2 −1 ⎟⎟ + −1
= ⎜⎜⎝ + ⎟⎟⎠ X (why?)
1 − 12 z−1 1 − 14 z−1
⎛ ⎞ ⎛ ⎞ 1
⎜⎜ 1 ⎟⎟ ⎜⎜ 1 ⎟⎟ 4
delay (1)
= ⎜⎝⎜ ⎟⎟ · (2) · X + ⎜⎜
⎠ ⎝
⎟⎟ · (−1) · X
⎠
1 − 12 z−1 1 − 41 z−1 1 → 2(1) − 1(1) = 1
0 → 2[0 + 12 (1)] − 1[0 + 14 (1)] = 3
4
1
Two first-order feedback system in parallel. Poles are at 2 and 14 . 0 → 2[0 + 1 1 1 1 7
2 ( 2 )] − 1[0 + (
4 4 )] = 16
1 1 1 1 15
0 → 2[0 + 2 ( 4 )] − 1[0 + 4 ( 16 )] = 64
The mathematics is called partial fraction.
1 1 1 1 31
0 → 2[0 + 2 ( 8 )] − 1[0 + 4 ( 64 )] = 256

#4: z-transform Example #4: z-transform Example Mini-conclusions
z-transform allows us to use polynomials to analyze discrete-time

Remember that for a first-order feedback system with gain p, an
systems, particularly feedback systems
impulse input (x[n] = δ[n]) gives an output
The poles are the roots of the polynomial in the denominator

y[n] = pn
The poles (in particular, the one with the largest magnitude)
determine the stability of the feedback system
Hence for our second-order feedback system, we can analytically
represent the output as Partial fraction, as a technique in polynomial manipulation, is
very useful to give insight into the system behavior
y[n] = (2)( 12 )n + (−1)( 14 )n = ( 12 )n−1 − ( 14 )n
(Almost all) high-order feedback systems can be turned to a sum
This is not obvious at all from the original flow graph analysis! of first-order feedback systems
#3 – #4: Transform Conclusions Problems that Need Control

Transform method is less straightforward compared with time-domain A design problem: Fix the angle θ[n] of an antenna. You have
methods, but is more powerful, particularly in design. 1 A voltage source v[n] (assumed discrete-time)
2 A motor that produces an angular movement proportional to the
Control the rate of decay by designing the poles. voltage input
Can break into sub-systems and treat the system hierarchically.
A possible configuration:
For implementation, can always go back to time-domain.
1 − 32 z−1
Example: H = . Thus,
1 − 56 z−1 + 16 z−2

1 − 65 z−1 + 16 z−2 Y = 1 − 32 z−1 X
+
Y − 56 z−1 Y + 16 z−2 Y = X − 32 z−1 X v − θ
y[n] − 56 y[n − 1] + 16 y[n − 2] = x[n] − 32 x[n − 1]
Problems that Need Control Problems that Need Control

Modification by adding feedback control:
The corresponding discrete-time model:

move by K2 K1 (θD − θ)
v[n] motor θ[n]
K1 θD K1 (θD − θ)
θD potentiometer comparator amplifier
Problems:
θ
Need to know initial angle
Need exact electrical and mechanical characteristics of the motor K1 θ
potentiometer
Need to give very precise instruction to the voltage source to first
accelerate and then decelerate the motor
A desired angle θD
This is called an open-loop system, which is generally not desirable. Potentiometer maps angle to voltage (K1 )
Comparator takes the difference
Amplifier magnifies the input signal (K2 )
Problems that Need Control Problems that Need Control

The corresponding discrete-time model: Many problems benefit from feedback control. Examples:
θD [n] + gain motor θ[n] 1 Proper driving: controlling the steering wheel to stay in lane
desired actual
driver car
−1 position position
What about drunk driving?

gain = K1 K2 . This is our design parameter
Can disturb the antenna position — the system will correct itself 2 Air-conditioning: controlling the room temperature
No need to know initial angle, model the exact electrical and desired actual
thermostat heating/cooling
mechanical characteristics of the motor, or give very precise temperature temperature
instruction to the voltage source to first accelerate and then
decelerate the motor What if you don’t want to turn on and off the air-conditioning too
This is called a closed-loop system, which is generally desirable! frequently?
Mathematics of Feedback Control Feedback Control Example

A general feedback pattern: Example (A): Reliable amplification with a constant K

X − GY K X − GY X K Y
+ K is huge (K 1) but unreliable. How to build a reliable gain system?

X + K(z) Y
−
Ans: Assume we can build an attenuator G reliably.
GY
G(z)
+
X + K Y
−
Negative gain is embedded in the “+” and “−” symbols G

Y = K X − GY
If KG 1: independent of the actual value of K,
Y K Y K 1
H= = = ≈
X 1 + KG X 1 + KG G
Feedback Control Example Feedback Control Example

Example (A): (cont.) Reliable amplification using circuits Example (A): (cont.) Reliable amplification using circuits

Use an operational amplifier (op amp) where Vo = K V+ − V− +
X + K Y
−
V− −
R2
Vo G=
V+ + R1 + R2
Build this circuit (V− is the feedback): Assume KG 1: (precise value of K does not matter)
R2 R1 K 1 R1 + R2
H= ≈ =
1 + KG G R1

R2
−
y V− = Y
R1 + R2 Resistor values are much more accurate than op amp gain → this
+

x
+
− Y = K X − V− new circuit has much more reliable gain
Tradeoff: the gain is much smaller, since we need K 1/G
Called non-inverting amplifier
Feedback Control Example Feedback Control Example

Example (B): Inverse system design
Example (C): Stabilization of unstable systems (e.g., microphone echo)
y[n] = 12 (x[n] + x[n − 1]) has a smoothing effect; can we undo it?
y[n] = 2y[n − 1] + x[n] is an unstable system; can we stabilize it?
Ans: Let G = 1
2 (1 + z−1 ), so Y = GX. Pass y[n] through:
Ans: Let K = 1
1−2z−1
and G = ρz−1 .
+
Y + K W
− + 1
X + K= 1−2z−1 Y
−
G(z) = 12 (1 + z−1 )
G = ρz−1
Assume KG(z) 1: (we’ll ignore what that really means for a polynomial in
1
z for now) K −1 1 1
= 1−2z = −1 =
W K 1 1 + KG 1 + 1
ρz−1
−1
(1 − 2z ) + ρz 1 − (2 − ρ)z−1
H= = ≈ 1−2z−1
Y 1 + KG G
ρ is our design parameter. Pick ρ so that |2 − ρ| < 1.
1
Hence, W = HY ≈ G X = X.
G
Feedback Design Feedback Control Design Example

A more generic setup of feedback loop
Example (A):
+ error input
reference + controller system output
− + ρ 1
x[n] + 1−2z−1
y[n]
−
measurement
sensor
ρ ρ
controller = C(z), system/plant = P(z), sensor = G(z) Y −1 ρ 1+ρ
= 1−2z ρ = −1
=
P(z) may be unstable X 1+ (1 + ρ) − 2z 2
1 − 1+ρ z−1
1−2z−1
Design C and G such that the closed-loop system
Bigger ρ means a smaller pole
CP
H= is stable e.g. x[n] = δ[n], then a smaller pole means y[n] → 0 faster
1 + CPG
Feedback Control Design Example Feedback Control Design Example

Example (B): ρ
Example (B): (cont.) H =
(1 + ρ) − (ρσ)z−1 − 4z−2
+ ρ 1
x[n] + 1−4z−2
y[n]
−
We can make different choices of ρ and σ.
55
1 ρ= 9 , σ = 0:
1 − σz−1
55 55 55 55
9 64 128 128
Original system is unstable because H= = 9 −2
= +
64
9 − 4z−2 1 − 16 z 1 − 34 z−1 1 + 34 z−1
1 1
1 1
= = + 2 2
Therefore, if x[n] = δ[n], then
1 − 4z−2 (1 + 2z−1 )(1 − 2z−1 ) 1 − 2z−1 1 + 2z−1
n ⎧ n
Overall feedback system: 55 3 n 3 ⎪
⎨ 55 3
if n is even.
y[n] = + − =⎪
⎩
64 4
ρ 128 4 4 0 if n is odd.
1−4z−2 ρ ρ
ρ = =
1 + 1−4z−2 (1 − σz ) (1 − 4z ) + ρ(1 − σz ) (1 + ρ) − (ρσ)z−1 − 4z−2
−1 −2 −1
Feedback Control Design Example Feedback Control Design Example

Example (B): (cont.)
ρ
Example (B): (cont.) H =
(1 + ρ) − (ρσ)z−1 − 4z−2
0.9 1
0.8
0.7 0.8
0.6
8
2 ρ = 31, σ = 31 : 0.5
0.6
0.4
31 31 2 31 1 0.4
31 32 32 · 3 32 · 3 0.3
H= = = +
32 − 8z−1 − 4z−2 1 − 14 z−1 − 18 z−2 1 − 12 z−1 1 + 14 z−1
0.2 0.2
0.1
0 0
Therefore, if x[n] = δ[n], then
0 2 4 6 8 10 0 2 4 6 8 10
55 8
ρ= 9, σ=0 ρ = 31, σ =
n n 31
31 1 1
y[n] = 2 + − poles at 34 , − 34 poles at 12 , − 14
96 2 4
Decay is determined by the larger of the two poles (magnitude < 1)
Feedback Design: The Antenna Orientation Problem Feedback Design: The Antenna Orientation Problem
Set the input angle, and expect the output to follow
Input is θD [n]. Output is θ[n]
Sensor senses current angle to feed back to the input
move by K2 K1 (θD − θ)
θs [n] = θ[n]
K1 θ D K1 (θD − θ)
θD potentiometer comparator amplifier Controller produces an angular velocity proportional to the difference
between desired angle and sensed data
θ
K1 θ
v[n] = K(θD [n] − θs [n])
potentiometer
System turns the angle of the antenna from the “previous angle” to the
“current angle” (based on the previous angular velocity!)
Connect the “specific problem” to the “general model”
θ[n] = θ[n − 1] + Tv[n − 1]
+ error input
reference + controller system output
− where T is the time between the discrete samples (fixed beforehand)
measurement
sensor Design question: How to choose K?
system
Equation: controller
θ[n] = θ[n − 1] + KT(θD [n − 1] − θ[n − 1]) + v[n]
θD [n] + K T + delay (1) θ[n]
−
Flow graph:
controller system θs [n]

sensor
Within the system block, we have a feedback with positive loop! Hence,
+ v[n]
θD [n] + K T + delay (1) θ[n] transfer function within this block has minus in the denominator:
−
z−1
T·
1−z−1
θs [n] Substituting this into the system block, we can apply the feedback
sensor equation again to get

z−1
Make use of the mathematics of feedback control to find the transfer Θ K T · 1−z−1 (KT)z−1
H= = =
function of the entire process ΘD 1 + K T · z −1
−1
1 − (1 − KT)z−1
1−z
θD [n] + v[n]

θD [n] + + K T + delay (1) θ[n]
We can use this transfer function to understand the behavior of −
delay (1)
the output with an inpulse input.
θs [n]
But what we want is a bit different: we want to understand the
behavior of the output with an input staying at a certain angle.
Let p = 1 − KT. Overall transfer function:
Change to the following system! (1 − p)z−1
= 1 1 1
H = −
θD [n] + v[n] 1 − z−1 1 − pz−1 1 − z−1 1 − pz−1

θD [n] + + K T + delay (1) θ[n]
−
delay (1)
Thus, if θD [n] is an impulse input, output is a difference of two
geometrically decaying sequences, i.e.
θs [n]
θ[n] = (1)n − (p)n = 1 − pn
What if sensor feeds back to the input with a delay, i.e.
θs [n] = θ[n − 1]
θ[n] = 1 − pn
Adding delay tends to destabilize the system
Interpretations: Difficult to have an intuitive way to set K
Let’s say a value of 1 means 30◦ . We give an impulse input, or
controller system
equivalently keep θD [n] = 1, leads to θ[n] = 1 − pn , which
approaches 1 when |p| < 1 and n → ∞.
+ v[n]
Best value of p: p = 0, i.e. KT = 1. θD [n] + K T + delay (1) θ[n]
−
Note that lim p0 = 1. So, output is 1 only when n ≥ 1.
p→0
Let’s say we get a discrete sample every one second, i.e. T = 1s. If
we want to turn 30◦ (a value of 1 in our discrete model), then we delay (1)
should set K = 30◦ per second (to multiply the input). θs [n]
This makes perfect sense! We will be done in one step. sensor
1 1 Develop the transfer function

0.8 0.8 z−1
Systems block unchanged: T ·
0.6 0.6
1 − z−1
0.4 0.4
Transfer function includes the delay in sensor
0.2 0.2

z−1
0 0
Θ K T · 1−z−1 (KT)z−1

0 5 10 15 20 0 5 10 15 20
H= = =
KT = 0.1 KT = 0.25 ΘD 1 + K T · z−1−1 (z−1 ) 1 − z−1 + (KT)z−2
2 2 5 1−z
4
1.5 1.5 3 Full transfer function including the “step” input:

2
(KT)z−1
1 1
= Θ = 1
1
H
D 1 − z−1 1 − z−1 + (KT)z−2
0
Θ
0.5 0.5
−1
z z
= −
0 0 −2
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
1−z −1 1 − z + (KT)z−2
−1
KT = 0.75 KT = 1.0 KT = 1.1
Poles are the values where the denominator is zero, i.e.

Some conclusions about feedback control:
1 − z−1 + (KT)z−2 = 0
z2 − z + KT = 0 1 Feedback allows us to change the system characteristics, e.g. from
unstable to stable
(−1)2 − 4(1)(KT) 1 1
1±
So poles are at = ± 1 − 4(KT). 2 Feedback allows design parameters
2(1) 2 2
3 The mathematics of z-transform and pole locations are powerful
For 0 < (KT) < 14 : the larger of the pole is 12 + 12 1 − 4(KT). It is
tools to analyze discrete-time systems and feedback
smallest when KT = 14 .
4 We now can analyze and design a system that would be impossible to do
Smaller values of KT also works, but not as good (due to a larger otherwise!
pole).
Sensor delay results in slower response.
Feedback Design: The Antenna Orientation Problem The Way Forward
First Year Second Year Third Year

A postscript: We had some observations earlier that
1 ELEC 3222:
For 4 < (KT) < 1, the system oscillates but still converges Robotics
For (KT) = 1, the system oscillates (Hint: what is the magnitude of ELEC 3206: Con-
trol Systems
the poles?) ELEC 2205: Con-
trol and Instrumen-
tation
For (KT) ≥ 1, the system is unstable! ENGG 1015
ELEC 2201: Sig-
nals and Linear
Systems
ELEC 2204: Digital
Why? You’ll find out if you continue with EEE! Signal Processing
ELEC 3225: Digital
Image Processing
ELEC 3224: Multi-
media Signals and
Applications

09 Signals and Systems 4up

Uploaded by

Copyright:

Available Formats

You might also like

09 Signals and Systems 4up

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

09 Signals and Systems 4up

Uploaded by

Copyright:

Available Formats

Signals, Systems, and Control (Part #1)

Review: Digital Systems

Signals and Systems Examples

Communication system Imaging system

Financial system Biomedical system

A “Signals and Systems” Viewpoint Discrete-time Systems

Very diverse applications “system”

Four Diﬀerent Representations #1: Signal ﬂow graph

We restrict ourselves to three “primitives”:

#1: Signal Flow Graph Mini-conclusions #2: Diﬀerence Equations

k delay (d) ⇔ delay (d) k 0

k k Data present at regular time intervals

#2: Diﬀerence Equations Example #2: Diﬀerence Equations Implementation

#2: Diﬀerence Equations Implementation #2: Diﬀerence Equations Mini-conclusions

Hence, the name “constant-coeﬃcient”

#1– #2: Time-domain Conclusions #3: Operator Notation and Arithmetics

Both ﬂow graphs and diﬀerence equations are “time-domain” methods 5

#3: Operator Notations #3: Operator Notations

A very important operation is called a delay operation, written as

#3: Operator Arithmetics #3: Operator Example

X Dd X We are allowed to write Y = 1

#3: Operator Example #3: Operator Example

We can tackle composite ﬂow graphs using operator arithemtics.

delay (1) 1 delay (1) 1

#3: Operator Example #3: Operator Example

#3: Operator: Output #3: Operator Arithmetics for Feedback

Operation notation provides a powerful tool to analyze feedback,

Therefore, p delay (1) +

Ans: Y = X + pDX + p2 D2 X + . . . implies 1 1 1

this feedback system, {1, 0, 0, . . .} −→ {1, p, p2 , . . .}. 2 2 2

Value of p determines whether the output in response to an impulse input is

#3: Operator Mini-conclusions #4: z-transform and Transfer Function

#4: z-transform Notations #4: z-transform Notations

#4: z-transform Arithmetics #4: z-transform Example

X, X1 , and X2 are now z-transforms of the time-domain signals. We Example (A):

#4: z-transform Example #4: Transfer Function

Example (B): Deﬁne the transfer function H(z) =

#4: z-transform Example #4: z-transform Example

g2 delay (2) g2 z−2 Y

#4: z-transform Example #4: z-transform Example

#4: z-transform Example #4: z-transform Example

Do you see a pattern?

#4: z-transform Example #4: z-transform Example Mini-conclusions

z-transform allows us to use polynomials to analyze discrete-time

The poles are the roots of the polynomial in the denominator

#3 – #4: Transform Conclusions Problems that Need Control

Problems that Need Control Problems that Need Control

The corresponding discrete-time model:

Problems that Need Control Problems that Need Control

What about drunk driving?

Mathematics of Feedback Control Feedback Control Example

+ K is huge (K  1) but unreliable. How to build a reliable gain system?

Feedback Control Example Feedback Control Example

Feedback Control Example Feedback Control Example

Feedback Design Feedback Control Design Example

Feedback Control Design Example Feedback Control Design Example

Feedback Control Design Example Feedback Control Design Example

controller system θs [n]

+ K is huge (K 1) but unreliable. How to build a reliable gain system?