Emb DSP1

Fundamentals
2/1/2013
Amit Kulkarni
What Is DSP?
Analog Computer
a bit loud
Digital Computer
DSP
ADC 1010 1001 DAC
OUTPUT
2/1/2013
Amit Kulkarni
A Typical DSP System

MEMORY
ADC DSP DAC
DSP Chip Memory Converters (Optional)

Analog to Digital Digital to Analog
Communication Ports
Serial Parallel
PORTS
Multiply and Add

Add
+ 1+2 = 3 0001 0010 0011
Multiply
0 1 0 1 5 x 8 x x 4 x x 2 x xShifted 1 x and 0011 0011 0011 0011 3
5*3 = 15 0000 0011 0000 0011 =
added multiple times
Most Common Operation in DSP A = B*C + D E = F*G + A
MAC Operation Typically 70 Clock Cycles With Ordinary Processors Typically 1 Clock Cycle With Digital Signal Processors
. . .
Multiply, Add, and Accumulate MAC Instruction
2/1/2013
Amit Kulkarni
Drop in Multiplication Times

TIME (ns) 600 500 400 300 200 100 0 1971
5 ns
1976
1998 YEARS
2/1/2013
Amit Kulkarni
Digital Computers
von Neuman Machine
A
STORED PROGRA M AND DATA
INPUT/ OUTPU T
ARITHMETIC A = ADDRESS LOGIC D = DATA UNIT
Harvard Architecture
A A
ARITHMETIC LOGIC UNIT INPUT/ OUTPUT
STORED PROGRAM
D
STORED DATA
2/1/2013
Amit Kulkarni
DSP Development
ADD A, B 11100010010100001001
ASSEMBLER HIGH-LEVEL LANGUAGE
CODE
TEST
EMULATOR
S/W DESIGN
N DSP
OK?
Y
Tools of the Trade
PRODUCT
2/1/2013
Amit Kulkarni
Why Digital Processing?

ADC PROCESS DAC
Advantages to Digital Processing

Programmability Stability Repeatability Special Applications
2/1/2013
Amit Kulkarni
Programmability
One Hardware = Many Tasks
SOFTWARE 1 SOFTWARE 2 SOFTWARE N LOW-PASS FILTER MUSIC SYNTHESIZER
. .
SAME HARDWARE
. . MOTOR CONTROL
Upgradability and Flexibility

Develop New Code Upgrade Analog Solder New Component
2/1/2013
Amit Kulkarni
Analog Variability
Analog Circuits are affected by
Temperature Aging
Tolerance of Components
Two Analog Systems using the same design and components may differ in performance
1k + 10 years =
1.1k
2/1/2013
Amit Kulkarni
10
Digital Repeatability
Perfect Reproducibility
Nearly identical performance from unit to unit Performance not affected by tolerance No drift in performance due to temperature or aging Guaranteed accuracy
A CD player always plays the same music quality
2/1/2013
Amit Kulkarni
11
Performance
Some special functions are best implemented digitally
Lossless Compression
Adaptive Filters gain f
Linear Phase Filters phase
frequency f1 f2
frequency
2/1/2013
Amit Kulkarni
12
Practical DSP Systems

Hi-Fi Equipment Toys Videophones Modems Phone Systems 3D Graphics Image Processing And More ...
2/1/2013
Amit Kulkarni
13
Analog Advantages
Low cost and simplicity in some applications
Attenuators/amplifiers Simple filters
Wide bandwidth (GHz) Low signal levels Infinite effective sampling rate
Infinite resolution in frequency No aliasing/reconstruction issues
Infinite resolution in amplitude

No quantization noise
2/1/2013
Amit Kulkarni
14
Digital Signal Processing (DSP) Advantages

Repeatability
Low sensitivity to component tolerances Low sensitivity to temperature changes Low sensitivity to aging effects

Nearly identical performance from unit to unit Matched circuits cost less
High noise immunity In many applications DSP offers higher performance and lower cost
CD players versus phonographic turntable
2/1/2013
Amit Kulkarni
15
Analogs Place in DSP

Most transducers are analog by nature
Microphones, speakers, etc.
Analog circuits are required to pre-process low level signals before ADC Analog filters may be required to limit the bandwidth of signals
Anti-alias (before ADC) and reconstruction filters (after DAC)
Analog circuits may be required to drive output transducers

A power amplifier is required to enable a DAC to drive a speaker
2/1/2013
Amit Kulkarni
16
Number Systems
Represent numbers digitally
Decimal 2 Digit Number Digit Number 128 27 7 64 26 6 32 25 5 16 24 4 8 23 3 4 22 2 2 21 1 1 20 0
Any number can be represented as a series of 1s and 0s Decimal 3 in binary

Decimal 2
Digit Number
0 2
7
0 2
6
0 2
5
0 2
4
0 2
3
0 2
2
2+ 2
1
1= 2
0
Digit Number Binary
7 0
6 0
5 0
4 0
3 0
2 0
1 1
0 1 0000 0011
Decimal 26 in binary
Decimal 2
Digit Number
0 2
7
0 2
6
0 2
5
16+ 2
4
8+ 2
3
0 2
2
2= 2
1
0 2
0
26
Digit Number Binary
7 0
6 0
5 0
4 1
3 1
2 0
1 1
0 0 0001 1010
2/1/2013
Amit Kulkarni
17
Binary and Hex

Decimal 0,1,2,.,9 Binary 0,1 Hex 0,1,2,..,A,B,C,D,E,F 16 Decimal 20 Decimal 0x10 Hex 0x14 Hex
4 bits of binary system is represented by a single hex digit

Decimal 2
Digit Num ber
8+ 2
3
4+ 2
2
2+ 2
1
1= 2
0
15
Binary Hex
1 F
1111 F
Decimal 26 in binary and hex

Decimal 2
Digit Number
0 2
7
0 2
6
0 2
5
16+ 2
4
8+ 2
3
0 2
2
2= 2
1
0 2
0
26
Digit Number Binary Hex
7 0
6 0 1
5 0
4 1
3 1
2 0 A
1 1
0 0 0001 1010 1A
2/1/2013
Amit Kulkarni
18
Signed Integers
Signed magnitude integers
Signed Decimal 2 3 -2 -3 Hex Sign 00 00 00 02 00 00 00 03 80 00 00 02 80 00 00 03 0 0 1 1 Binary Number 000 0000 0000 0000 0000 0000 0000 0010 000 0000 0000 0000 0000 0000 0000 0011 000 0000 0000 0000 0000 0000 0000 0010 000 0000 0000 0000 0000 0000 0000 0011
2/1/2013
Amit Kulkarni
19
Twos Complement Notation

2 Digit number Decimal Binary twos complement Decimal calculation Binary twos complement Decimal calculation -2 3 -2 7 -128 0 0 1 -128 26 64 0 0 1 +64 25 32 0 0 1 +32 24 16 0 0 1 +16 23 8 0 0 1 +8 22 4 0 0 1 +4 21 2 1 +2 1 +2 20 1 1 +1= 0 +0= -2 3
2/1/2013
Amit Kulkarni
20
Twos Complement
Conversion to twos complement notation
Action
Signed integer Strip sign bit Invert Add one Twos complement Hex
Decimal
Sign -2 1 0 1 1 1 F
Binary
Number 000 0000 0000 0000 0000 0000 0000 0010 000 0000 0000 0000 0000 0000 0000 0010 111 1111 1111 1111 1111 1111 1111 1101 1 111 1111 1111 1111 1111 1111 1111 1110 F F F F F F E
Twos complement addition

Decimal
-2 3 -2 + 3 = 1
Hex
Sign FF FF FF FE 00 00 00 03 00 00 00 01 1 0 0
Twos Complement Binary

Number 111 1111 1111 1111 1111 1111 1111 1110 000 0000 0000 0000 0000 0000 0000 0011 000 0000 0000 0000 0000 0000 0000 0001
2/1/2013
Amit Kulkarni
21
What a DSP system is?

Digital Signal Processing: application of mathematical operations to digitally represented signals Signals represented digitally as sequences of samples Digital signals obtained from physical signals via transducers (e.g., microphones) and analog todigital converters (ADC) Digital signals converted back to physical signals via digital-to-analog converters (DAC) Digital Signal Processor (DSP): electronic system that processes digital signals
2/1/2013 Amit Kulkarni 22
Common DSP algorithms and applications

Applications Instrumentation and measurement Communications Audio and video processing Graphics, image enhancement, 3- D rendering Navigation, radar, GPS Control - robotics, machine vision, guidance Algorithms Frequency domain filtering - FIR and IIR Frequency- time transformations - FFT Correlation
What Do DSPs Need to Do Well?

Most DSP tasks require: Repetitive numeric computations Attention to numeric fidelity High memory bandwidth, mostly via array accesses Real-time processing DSPs must perform these tasks efficiently while minimizing: Cost Power Memory use Development time
Who Cares?
DSP is a key enabling technology for many types of electronic products DSP-intensive tasks are the performance bottleneck in many computer applications today Computational demands of DSP-intensive tasks are increasing very rapidly In many embedded applications, general purpose microprocessors are not competitive with DSPoriented processors today 2012 market for DSP processors: $6 billion
Todays DSP Killer Apps

In terms of dollar volume, the biggest markets for DSP processors today include: Digital cellular telephony Pagers and other wireless systems Modems Disk drive servo control Most demand good performance All demand low cost Many demand high energy efficiency Trends are towards better support for these (and similar) major applications.
Decoding DSP Lingo
2/1/2013
Amit Kulkarni
27
Signals and their representation

Recall that the Digital Signal Processing is application of mathematical operations to digitally represented signals. Signals represented digitally as sequences of samples. Signal representation is done in following two ways:
Time-domain and Frequency domain
2/1/2013
Amit Kulkarni
28
The Time domain

The traditional way of observing signals. The time domain is a record of what happened to a parameter of the system versus time.
Figure shows a simple spring-mass system where we have attached a pen to the mass and pulled a piece of paper past the pen at a constant rate. The resulting graph is a record of the displacement of the mass versus time, a time domain view of Displacement.
2/1/2013
Amit Kulkarni
29
The Frequency Domain

It was shown over one hundred years ago by Baron Jean Baptiste Fourier that any waveform that exists in the real world can be generated by adding up sine waves. (The Fourier series)
By picking the amplitudes, frequencies and phases of these sine waves correctly, we can generate a waveform identical to our desired signal.
2/1/2013
Amit Kulkarni
30
Contd..
Conversely, we can break down our real world signal into these same sine waves. This combination of sine waves is unique; any real world signal can be represented by only one combination of sine waves.
2/1/2013
Amit Kulkarni
31
Have we lost some information?

It is very important to understand that we have neither gained nor lost information, we are just representing it differently. We are looking at the same three-dimensional graph from different angles. Why the Frequency Domain?
to resolve each frequency component contained in a signal.
2/1/2013
Amit Kulkarni
32
The Frequency Domain: A Natural Domain

At first the frequency domain may seem strange and unfamiliar, yet it is an important part of everyday life. Your ear-brain combination is an excellent frequency domain analyzer. The ear-brain splits the audio spectrum into many narrow bands and determines the power present in each band. It can easily pick small sounds out of loud background noise.
Spectrum Examples
2/1/2013
Amit Kulkarni
34
Review
Signal classification Signal operations Fourier Series representation of periodic signals Continuous time Fourier transform and signal spectra
2/1/2013
Amit Kulkarni
35
Classification of Signals
Continuous and discrete-time signals Continuous and discrete-valued signals Even and odd signals Periodic signals, non-periodic signals Deterministic signals, random signals Causal and anticausal signals Right-handed and left-handed signals Finite and infinite length
Continuous and discretetime signals

Continuous signal
- It is defined for all time t : x(t)
Discrete-time signal
- It is defined only at discrete instants of time : x[n]=x(nT)
2/1/2013
Amit Kulkarni
37
Continuous and Discrete valued singals

CV corresponds to a continuous y-axis DV corresponds to a discrete y-axis
Digital signal
Even and odd signals

Even signals : x(-t)=x(t) Odd signals : x(-t)=-x(t) Even and odd signal decomposition xe(t)= 1/2(x(t)+x(-t)) xo(t)= 1/2(x(t)x(-t))
2/1/2013
Amit Kulkarni
39
2/1/2013
Amit Kulkarni
40
Periodic signals, nonperiodic signals

Periodic signals
- A function that satisfies the condition x(t)=x(t+T) for all t - Fundamental frequency : f=1/T - Angular frequency : = 2/T
Non-periodic signals
Deterministic signals, random signals

Deterministic signals
-There is no uncertainty with respect to its value at any time. (ex) sin(3t)
Random signals
- There is uncertainty before its actual occurrence.
2/1/2013
Amit Kulkarni
42
Causal and anticausal Signals

Causal signals : zero for all negative time Anticausal signals : zero for all positive
time
Noncausal : nozero values in both

positive and negative time
causal signal
2/1/2013 Amit Kulkarni
anticausal signal
noncausal signal
43
Right-handed and left-handed Signals

Right-handed and left handed-signal : zero between a given variable and positive or negative infinity
2/1/2013
Amit Kulkarni
44
Finite and infinite length

Finite-length signal : nonzero over a finite interval tmin< t< tmax Infinite-length singal : nonzero over all real numbers
2/1/2013
Amit Kulkarni
45
Basic Operations on Signals

Operations performed on dependent signals Operations performed on the independent signals
2/1/2013
Amit Kulkarni
46
Operations performed on dependent signals

Amplitude scalingy (t ) = cx (t ) Addition Multiplication Differentiation Integration
y (t ) = x1 (t ) + x2 (t )
y (t ) = x1 (t ) x2 (t )
d y (t ) = x (t ) dx
y (t ) =
x( )d
47
2/1/2013
Amit Kulkarni
Operations performed on the independent signals

Time scaling y (t ) = x ( at ) a>1 : compressed 0<a<1 : expanded
2/1/2013
Amit Kulkarni
48

Reflection y (t ) = x(t )
2/1/2013
Amit Kulkarni
49

Time shifting y (t ) = x(t t0 )
- Precedence Rule for time shifting & time scaling b
y (t ) = x(at b) = x(a (t )) a
2/1/2013
Amit Kulkarni
50
The incorrect way of applying the precedence rule. (a) Signal x(t). (b) Time-scaled signal v(t) = x(2t). (c) Signal y(t) obtained by shifting v(t) = x(2t) by 3 time units, which yields y(t) = x(2(t + 3)).
The proper order in which the operations of time scaling and time shifting (a) Rectangular pulse x(t) of amplitude 1.0 and duration 2.0, symmetric about the origin. (b) Intermediate pulse v(t), representing a time-shifted version of x(t). (c) Desired signal y(t), resulting from the compression of v(t) by a factor of 2.
2/1/2013
Amit Kulkarni
51
Elementary Signals
Exponential signals x(t ) = Beat Sinusoidal signals x (t ) = A cos( t + ) Exponentially damped sinusoidal signals x(t ) = Ae at cos( t + )
2/1/2013
Amit Kulkarni
52
Elementary Signals
Step function
x (t ) = u (t )
2/1/2013
Amit Kulkarni
53
(a) Rectangular pulse x(t) of amplitude A and duration of 1 s, symmetric about the origin. (b) Representation of x(t) as the difference of two step functions of amplitude A, with one step function shifted to the left by and the other shifted to the right by ; the two shifted signals are denoted by x1(t) and x2(t), respectively. Note that x(t) = x1(t) x2(t).
2/1/2013
Amit Kulkarni
54
Elementary Signals
Impulse function
x (t ) = (t )
(a) Evolution of a rectangular pulse of unit area into an impulse of unit strength (i.e., unit impulse). (b) Graphical symbol for unit impulse. (c) Representation of an impulse of strength a that results from allowing the duration of a rectangular pulse of area a to approach zero.
2/1/2013
Amit Kulkarni
55
Elementary Signals
Ramp function
x(t ) = r (t )
2/1/2013
Amit Kulkarni
56
Systems Viewed as Interconnection of Operations

input signal
system
output signal
2/1/2013
Amit Kulkarni
57
Properties of Systems
Stability Memory Invertibility Time Invariance Linearity
2/1/2013
Amit Kulkarni
58
Stability(1)
BIBO stable : A system is said to be bounded-input bounded-output stable iff every bounded input results in a bounded output. Its Importance : the collapse of Tacoma Narrows t |bridge, x(t ) | M t | y (t ) | M y < suspension pp .45 x <
2/1/2013
Amit Kulkarni
59
Dramatic photographs showing the collapse of the Tacoma Narrows suspension bridge on November 7, 1940. (a) Photograph showing the twisting motion of the bridges center span just before failure. (b) A few minutes after the first piece of concrete fell, this second photograph shows a 600-ft section of the bridge breaking out of the suspension span and turning upside down as it crashed in Puget Sound, Washington. Note the car in the top righthand corner of the photograph.
Stability(2)
Example pp.46 - y[n]=1/3(x[n]+x[n-1]+x[n-2]) - y[n]=rnx[n], where r>1
1 y[n] = x[n] + x[n 1] + x[n 2] 3 1 (| x[n] | + | x[n 1] | + | x[n 2] |) 3 1 (M x + M x + M x ) = M x 3

Memory
Memory system : A system is said to possess memory if its output signal depends on past values of the input signal Memoryless system (example) i (t ) = 1 v(t ) R
i (t ) = 1 L
v( )d
62
y[n] = x[n] + x[n 1]

Memory or memoryless?
2/1/2013
Amit Kulkarni
63
Causality
Causal system : A system is said to be causal if the present value of the output signal depends only on the present and/or past values of the input signal. Non-causal system (example) y[n]=x[n]+1/2x[n-1] y[n]=x[n+1]+1/2x[n-1]
2/1/2013
Amit Kulkarni
64
Invertiblity(1)
Invertible system : A system is said to be invertible if the input of the system can be recovered from the system output. H:xy, H-1:yx H-1{y(t)}= H-1{H{x(t)}}, H-1H=I
H x(t)
H-1 y( t ) x(t)
65
Invertiblity(2)
(Example) y (t ) = 1 L
x ( )d x(t ) = L
d y (t ) dt
y (t ) = x 2 (t )
2/1/2013
Amit Kulkarni
66
Time Invariance (1)

Time invariant system : A system is said to be time invariant if a time delay or time advance of the input signal leads to a identical time shift in the output signal.
yi (t ) = H {x(t t0 )} = H {S t 0 {x(t )}} = HS t 0 {x(t )}
y0 (t ) = S t 0 { y (t )} = S t 0 {H {x(t )}} = S t 0 H {x(t )}
Time Invariance (2)

Are following two systems equivalent?
St0 x(t) H x(t) x(t-t0) St0 y0 ( t ) H y i( t )
2/1/2013
Amit Kulkarni
68
Time Invariance (3)

Examples
y (t ) = 1
x( )d L
x(t ) y (t ) = R(t )
2/1/2013
Amit Kulkarni
69
Linearity(1)
Linear system : A system is said to be linear if it satisfies the principle of superposition.
x(t ) = ai xi (t )
i =1
y (t ) = H {x(t )} = H { ai xi (t )}
i =1
= ai H {xi (t )} = ai yi (t )
i =1 i =1
Linearity(2)
x1(t) x2(t) . . . xN(t) a1 x1(t) a2 . . . aN . . . H y(t) xN(t) x2(t) H . . . H H a1 a2 . . . aN . . . y(t)
2/1/2013
Amit Kulkarni
71
Linearity(3)
Examples y[ n] = nx[n] -
y (t ) = x (t ) x (t 1)
Check superposition with simple two inputs.
x(t ) = a1 x1 (t ) + a2 x2 (t )
Theme Examples
Example of multiple propagation paths in a wireless communication environment.
2/1/2013
Amit Kulkarni
73
Tapped-delay-line model of a linear communication channel, assumed to be time-invariant
2/1/2013
Amit Kulkarni
74
Stock Price : filtering

(a) Fluctuations in the closing stock price of Intel over a three-year period. (b) Output of a four-point moving-average system.
2/1/2013
Amit Kulkarni
75
n The frequencies are w n = T
Fourier transform
and
w =
Therefore (1) and (2) are represented as
1 ~ f ( x) = 2
iwx T iwt e e f ( t )dt w n = T
(3)
Since, on one hand the function with period T has also the periods kT for any integer k, and on the other hand any non-periodic function can be considered as a function with infinite period, we can run the T to infinity, and obtain the Riemann sum with w, converging to the integral:
T 1 ~ iwx iwt e e f ( t )dt dw f ( x) = 2 T
(4)
76
Fourier transform definition

The integral (4) suggests the formal definition: The funciotn F(w) is called a Fourier Transform of function f(x) if:
F ( w ) := F { f ( t )} :=
The function
iwt e f ( t )dt
(5)
1 F { F ( w )} := 2
1
iwx e F ( w )dw
(6)
Is called an inverse Fourier transform of F(w).
2/1/2013
Amit Kulkarni
77
Example 1
The Fourier transform of
1 | t | 1 f ( t ) := 0 | t |> 1
1
sin w 2 iwt F { f ( t )} = e dt = w 1 2 The inverse Fourier transform is

2 2
iwx e
is
w0 w=0
sin w 2 sin w dw = cos wx dw = 0 w w 1 | x |< 1 1 sin( w ( x + 1)) 1 sin( w ( x 1)) = dw dw = 1 / 2 x = 1 0 w 0 w 0 x>1

2/1/2013
Amit Kulkarni
78
Fourier Integral
If f(x) and f(x) are piecewise continuous in every finite interval, and f(x) is absolutely integrable on R, i.e.
converges, then
1 1 [ f ( x ) + f ( x + )] = 2 2
iwx iwt e e f ( t )dt dw
Remark: the above conditions are sufficient, but not necessary.
2/1/2013
Amit Kulkarni
79
Properties of Fourier transform

1 Linearity: For any constants a, b the following equality holds:
F {af ( t ) + bg ( t )} = aF { f ( t )} + bF { g ( t )} = aF ( w ) + bG ( w )
Proof is by substitution into (5).
Scaling:
For any constant c, the following equality holds:
1 w F { f (ct )} = F( ) |c| c
2/1/2013
Amit Kulkarni
80
Properties of Fourier transform 2

3
Time shifting:
F { f ( t t 0 )} = e iwt 0 F ( w )
Proof:
F { f ( t t 0 )} =
4.
f ( t t 0 )e iwt dt = e iwt 0
f ( u)e iwu du
Frequency shifting:
F {e iwt 0 f ( t )} = F ( w w 0 )
Proof:
F {e iw0 t f ( t )} =
2/1/2013
iwt 0 iwt ( ) e f t e dt = F ( w w 0 )
Amit Kulkarni
81

5.
Symmetry:
F { F ( t )} = 2f ( w )
Proof: The inverse Fourier transform is
therefore
1 1 f ( t ) = F { f ( w )} = 2 1 2f ( w ) = 2
iwt F ( w ) e dw
itw F ( t ) e dt = F { F ( t )}
2/1/2013
Amit Kulkarni
82

6.
Modulation:
1 F { f ( t ) cos( w 0 t )} = [ F ( w + w 0 ) + F ( w w 0 )] 2 1 F { f ( t ) sin( w 0 t )} = [ F ( w + w 0 ) F ( w w 0 )] 2
Proof: Using Euler formula, properties 1 (linearity) and 4 (frequency shifting):
1 F { f ( t ) cos( w0 t )} = [ F {e iw0 t f ( t )} + F {e iw0 t f ( t )}] 2 1 = [ F ( w + w 0 ) + F ( w w0 )] 2

Differentiation in time
7.
Transform of derivatives
Suppose that f(n) is piecewise continuous, and absolutely integrable on R. Then
F { f ( n ) ( t )} = ( iw ) n F ( w )
In particular
F { f ' ( t )} = iwF ( w )
Proof:
and
F { f '' ( t )} = w 2 F ( w )
From the definition of F{ f(n)(t)} via integrating by parts.
2/1/2013
Amit Kulkarni
84
Example 2
The property of Fourier transform of derivatives can be used for solution of differential equations:
y 4 y = H ( t )e 4 t
0 t < 0 H (t ) = 1 t 0
F { y} 4 F { y } = F { H ( t )e
Setting F{y(t)}=Y(w), we have
4t
1 }= 4 + iw
1 iwY ( w ) 4Y ( w ) = 4 + iw
Example 2
Then
1 1 Y (w) = = (4 + iw )( 4 + iw ) 16 + w 2
1 4|t | y( w ) = F {Y ( w )} = e 8
1
Therefore
2/1/2013
Amit Kulkarni
86
Frequency Differentiation
F {t n f ( t )} = i n F ( n ) ( w )
In particular
F {tf ( t )} = iF ( w )
and
F {t 2 f ( t )} = F ( w )
Which can be proved from the definition of F{ f(t)}.
2/1/2013
Amit Kulkarni
87
Convolution
The convolution of two functions f(t) and g(t) is defined as:
f *g
Theorem:
f (u) g(t u)du
F { f * g } F { f }F { g }
Proof:
1 F { f ( t ) g( t )} [ F * G ]( w ) 2
iwt F { f * g} = e f ( u) g ( t u)du dt = iwu iw ( t u ) e f ( u) g ( t u)d ( t u) e du = F { f }F { g }
2/1/2013
Amit Kulkarni
88
Fundamentals
2/1/2013
Amit Kulkarni
89
CHAPTER
TimeTime -Domain Representations of LTI Systems
2. Interconnection of LTI Systems

2.1 Parallel Connection of LTI Systems 1. Two LTI systems: Fig. 2.1(a 2.1(a). ). Figure 2.1 Interconnection of two LTI systems. (a) Parallel connection of two systems. (b) Equivalent system.
2. Output: y (t ) = y (t ) + y (t ) 1 2 = x(t ) h1 (t ) + x (t ) h2 (t ) y (t ) = x( )h1 (t )d + x( )h2 (t )d

y (t ) = x( ){h1 (t ) + h2 (t )} d

where h(t) = h1(t) + h2(t) Fig. 2.1(b 2.1(b) )

90
= x ( )h(t )d = x (t ) h(t )
2/1/2013
Amit Kulkarni
CHAPTER

(2.1)
Distributive property for Continuous-time case:
x(t) h1(t) + x(t) h2 (t) = x(t) {h1(t) + h2 (t)}
Distributive property for Discrete-time case:
x[n] h1[n] + x[n] h2 [n] = x[n] {h1[n] + h2 [n]} (2.2)

2.2 Cascade Connection of LTI Systems 1. Two LTI systems: Fig. 2.2(a 2.2(a). ).
Figure 2.2 Interconnection of two LTI systems. (a) Cascade connection of two systems. (b) Equivalent system. (c) Equivalent system: Interchange system order.
CHAPTER
2. The output is expressed in terms of z(t) as y(t) = z(t) h2 (t) (2.17)
y(t) = z( )h2 (t )d
-
(2.3)
Since z(t) is the output of the first system, so it can be expressed as
z( ) = x( ) h1( ) = x( )h1( )d
-
(2.4)
Substituting Eq. (2.19) for z(t) into Eq. (2.18) gives y (t ) =

x( v )h1 ( v )h2 (t )dvd
Change of variable = (2.5)
y(t) = x( ) h1( )h2 (t )d d - Define h(t) = h1(t) h2(t), then
h(t v) = h1 ( )h2 (t v )d
y(t) = x( )h(t )d = x(t) h(t)

-
(2.6)
Fig. 2.2(b 2.2(b). ).
3. Associative property for continuous-time case:

CHAPTER
{x(t) h1(t)} h2 (t) = x(t) {h1(t) h2 (t)} (2.7) 4. Commutative property: Write h(t) = h1(t) h2(t) as the integral
h(t ) = h1 ( )h2 (t )d
Change of variable =t (2.8) Fig. 2.2(c 2.2(c). ).
h(t) = h1(t )h2 ( )d = h2 (t) h1(t)

-
Interchanging the order of the LTI systems in the cascade without affecting the result:
x(t ) {h1 (t ) h2 (t )} = x(t ) {h2 (t ) h1 (t )} ,

Commutative property for continuous-time case:
h1(t) h2 (t) = h2 (t) h1(t)
(2.9)
5. Associative property for discrete-time case:
{x[n] h1[n]} h2 [n] = x[n] {h1[n] h2 [n]} (2.10) Commutative property for discrete-time case: h1[n] h2 [n] = h2 [n] h1[n] (2.11)
CHAPTER
Example 2.1 Equivalent System to Four Interconnected Systems Consider the interconnection of four LTI systems, as depicted in Fig. 2.20. 2.20 The impulse responses of the systems are h1[n] = u[n], h2 [n] = u[n + 2] u[n], h3[n] = [n 2], and h4 [n] = nu[n]. Find the impulse response h[n] of the overall system.
Figure 2.3 Interconnection of systems for Example 2.11. <Sol.> 1. Parallel combination of h1[n] and h2[n]: h12[n] = h1[n] + h2[n] Fig. 2.4(a 2.4(a). ).
CHAPTER
Figure 2.4 (a) Reduction of parallel combination of LTI systems in upper branch of Fig. 2.20. (b) Reduction of cascade of systems in upper branch of Fig. 2.21(a). (c) Reduction of parallel combination of systems in Fig. 2.21(b) to obtain an equivalent system for Fig. 2.20.
2/1/2013
Amit Kulkarni
95
CHAPTER
2. h12[n] is in series with h3[n]: h123[n] = h12[n] h3[n] h123[n] = (h1[n] + h2[n]) h3[n] 3. h123[n] is in parallel with h4[n]: h[n] = h123[n] h4[n] Fig. 2.4(b).
h[n] = (h1[n] + h2 [n]) h3 [n] h4 [n], h12 [n] = u[n] + u[n + 2] u[n] = u[n + 2]
Convolving h12[n] with h3[n] gives
Fig. 2.4 (c).
Thus, substitute the specified forms of h1[n] and h2[n] to obtain
h123[n] = u[n + 2] [n 2] = u[n]

h[n] = 1 n u[ n].
Table 2.1 summarizes the interconnection properties presented in this section.
CHAPTER
2.3 Relation Between LTI System Properties and the Impulse Response
2.3.1 Memoryless LTI Systems 1. The output of a discrete-time LTI system:
y[n ] = h[n] x[n] =
k =
h[k ]x[n k ]
(2.3)
y[n] = L + h[ 2]x[n + 2] + h[ 1]x[n + 1] + h[0]x[n] + h[1]x[n 1] + h[2]x[n 2] + L

CHAPTER
2. To be memoryless, y[n] must depend only on x[n] and therefore cannot depend on x[n k] for k 0. A discrete-time LTI system is memoryless if and only if
h[k ] = c [k ]
Continuous-time system: 1. Output:
c is an arbitrary constant
y (t ) = h( )x(t )d ,
2. A continuous-time LTI system is memoryless if and only if
h ( ) = c ( )
c is an arbitrary constant
2.3.2 Causal LTI Systems The output of a causal LTI system depends only on past or present values of the input. Discrete-time system: 1. Convolution sum: y[ n] = L + h[ 2] x[n + 2] + h[1]x[ n + 1] + h[0]x[n]
+ h[1]x[n 1] + h[2]x[ n 2] + L.
CHAPTER
2. For a discrete-time causal LTI system,
h[k ] = 0 for k < 0

3. Convolution sum in new form:
y[n] = h[k ]x[n k ].

k =0
Continuous-time system: 1. Convolution integral:
3. Convolution integral in new form:
y (t ) = h( ) x(t )d .
y(t) = h( )x(t )d .
0
2. For a continuous-time causal LTI system,
h( ) = 0 for < 0
2.3.3 Stable LTI Systems A system is BIBO stable if the output is guaranteed to be bounded for every bounded input. y[n] M y Output: Discrete-time case: Input x[n ] M x
CHAPTER
1. The magnitude of output:
y[n ] = h[n] x[n] =

y[ n ] y[n]
x[n] M x
and it follows that
k =
h[k ]x[n k ]
a+b a + b
ab = a b
k =
h[k ]x[n k ] h[ k ] x[ n k ]
k =
2. Assume that the input is bounded, i.e.,
x[n k ] M x
y[n] Mx h[k]
k =
(2.28)
Hence, the output is bounded, or |y[n]| for all n, provided that the impulse response of the system is absolutely summable. 3. Condition for impulse response of a stable discrete-time LTI system:
CHAPTER
k =
h[k ] < .
Continuous-time case: Condition for impulse response of a stable continuous-time LTI system:

h ( ) d <
Example 2.2 Properties of the First-Order Recursive System The first-order system is described by the difference equation
y[n] = y[n 1] + x[n]

and has the impulse response
h[n] = nu[n]
Is this system causal, memoryless, and BIBO stable? <Sol.> 1. The system is causal, since h[n] = 0 for n < 0. 2. The system is not memoryless, since h[n] 0 for n > 0. 3. Stability: Checking whether the impulse response is absolutely summable?
CHAPTER

k k
k =
h[ k ] = = <
k =0 k =0
if and only if || < 1
Note: A system can be unstable even though the impulse response has a finite value. 1. Ideal integrator:
y(t) = x( )d
(2.29)
Input: x() = (), then the output is y(t) = h(t) = u(t). h(t) is not absolutely integrable Ideal integrator is not stable! 2. Ideal accumulator:
y[n] =
k =
x[k ]
Impulse response: h[n] = u[n] h[n] is not absolutely summable Ideal accumulator is not stable!
CHAPTER
2.3.4 Invertible Systems and Deconvolution A system is invertible if the input to the system can be recovered from the output except for a constant scale factor. 1. h(t) = impulse response of LTI system, Fig. 2.5 inv 2. h (t) = impulse response of LTI inverse system
Figure 2.5 Cascade of LTI system with impulse response h(t) and inverse system with impulse response h-1(t). 3. The process of recovering x(t) from h(t) x(t) is termed deconvolution. 4. An inverse system performs deconvolution.
x(t ) (h(t ) hin v (t )) = x(t ). h (t ) h inv (t ) = (t )

5. Discrete-time case:
2/1/2013
(2.12)
Continuous-time case (2.13)

103
h[n] hinv [n] = [n]

Amit Kulkarni
CHAPTER
Example 2.3 Multipath Communication Channels: Compensation by means of an Inverse System Consider designing a discrete-time inverse system to eliminate the distortion associated with multipath propagation in a data transmission problem. Assume that a discrete-time model for a two-path communication channel is
y[n] = x[n] + ax[n 1].

Find a causal inverse system that recovers x[n] from y[n]. Check whether this inverse system is stable. <Sol.> 1. Impulse response:
n=0 1, h[n] = a, n =1 0, otherwise

2. The inverse system hinv[n] must satisfy h[n] hinv[n] = [n].
hinv [n] + ahinv [n 1] = [n].
(2.32)
1) For n < 0, we must have hinv[n] = 0 in order to obtain a causal inverse system
CHAPTER
2) For n = 0, [n] = 1, and eq. (2.14) implies that
hinv [n] + ahinv [n 1] = 0,

hinv [n] = ahinv [n 1]
(2.15) 3. Since hinv[0] = 1, Eq. (2.33) implies that hinv[1] = a, hinv[2] = a2, hinv[3] = a3, and so on. The inverse system has the impulse response
hinv [n] = (a) n u[n]

4. To check for stability, we determine whether hinv[n] is absolutely summable, which will be the case if
k =
h [k ] =
inv
k =
is finite.
For |a| < 1, the system is stable. Table 2.2 summarizes the relation between LTI system properties and impulse response characteristics.
2/1/2013
Amit Kulkarni
105
CHAPTER
2.4 Step Response

1. The step response is defined as the output due to a unit step input signal. 2. Discrete-time LTI system: Let h[n] = impulse response and s[n] = step response.
s[ n] = h[ n]* u[ n] =
k =
h[k ]u[n k ].
3. Since u[n k] = 0 for k > n and u[n k] = 1 for k n, we have

CHAPTER
s[n] =
k =
h[k ].
The step response is the running sum of the impulse response. Continuous-time LTI system:
s(t) = h( )d
(2.16)
The step response s(t) is the running integral of the impulse response h(t). Express the impulse response in terms of the step response as
h[n] = s[n] s[n 1]
and
h(t ) =
d s (t ) dt
Example 2.14 RC Circuit: Step Response The impulse response of the RC circuit depicted in Fig. 2.6 is
t 1 RC h(t ) = e u (t ) RC
Find the step response of the circuit. <Sol.>

Figure 2.6 RC circuit system with the voltage source x(t) as input and the voltage measured across the capacitor y(t), as output.
107
CHAPTER
1. Step response:
1 RC s(t ) = e u ( )d . RC
t
0, s(t ) = 1 t RC e u ( )d RC
0, t<0 s(t ) = 1 t RC e d t 0 0 RC 0, t<0 = t RC 1 e , t 0
Fig. 2.7
t<0 t0
Figure 2.7 RC circuit step response for RC = 1 s.
2/1/2013
Amit Kulkarni
108
CHAPTER
2.5 Differential and Difference Equation Representations of LTI Systems

1. Linear constant-coefficient differential equation:
M Input = x(t), output = y(t) dk dk a y(t) = b x(t) (2.17) k k k k dt dt k =0 k =0 2. Linear constant-coefficient difference equation: N
ak y[n k] = bk x[n k]
k =0 k =0
(2.18)
Input = x[n], output = y[n]
The order of the differential or difference equation is (N, M), representing the number of energy storage devices in the system. Ex. RLC circuit depicted in Fig. 2.8 1. Input = voltage source x(t), output = loop current 2. KVL Eq.: Often, N M, and the order is described using only N.
Ry ( t ) + L
2/1/2013
d 1 t y ( t ) + y ( ) d = x ( t ) dt C
Amit Kulkarni 109
CHAPTER
Figure 2.8
Example of an RLC circuit described by a differential equation.
1 d d2 d y (t ) + R y (t ) + L 2 y (t ) = x (t ) C dt dt dt
Ex. Accelerator modeled in Section 1.10:
N=2
d2 y (t ) + 2 y (t ) = x (t ) y (t ) + Q dt dt
2 n
n d
N=2
where y(t) = the position of the proof mass, x(t) = external acceleration.
2/1/2013
Amit Kulkarni
110
CHAPTER
Ex. Second-order difference equation:
1 y[n 2] = x[n] + 2x[n 1] (2.19) N=2 4 Difference equations are easily rearranged to obtain recursive formulas for computing the current output of the system from the input signal and the past outputs. y[n] + y[n 1] +
Ex. Eq. (2.36) can be rewritten as
M N
1 y [ n] = a0
1 b x n k ] k [ a0 k =0
a y [n k ]
k =1 k
Ex. Consider computing y[n] for n 0 from x[n] for the second-order difference equation (2.37). <Sol.> 1. Eq. (2.37) can be rewritten as
y[n] = x[n] + 2x[n 1] y[n 1]

2. Computing y[n] for n 0:
1 y[n 2] 4
(2.20)
111
CHAPTER

1 y[ 2] 4
(2.21) (2.22)
y[0] = x[0] + 2x[ 1] y[ 1]

y[1] = x[1] + 2x[0] y[0]
1 y[ 1] 4
1 y [ 0] 4 1 y [3] = x [3] + 2 x [ 2] y [ 2] y [1] 4 y [ 2] = x [ 2] + 2 x [1] y [1]

Initial conditions: y[ 1] and y[ 2]. The initial conditions for Nth-order difference equation are the N values
y [ N ] , y [ N + 1] ,..., y [ 1] ,
The initial conditions for Nth-order differential equation are the N values
y (t )
t =0 ,
d y (t ) dt
t = 0
d2 , 2 y (t ) dt
, ...,
t =0
d N 1 y (t ) N 1 dt
t =0
2/1/2013
Amit Kulkarni
112
CHAPTER
Example 2.4 Recursive Evaluation of a Difference Equation Find the first two output values y[0] and y[1] for the system described by Eq. (2.38), assuming that the input is x[n] = (1/2)nu[n] and the initial conditions are y[ 1] = 1 and y[ 2] = 2. <Sol.> 1. Substitute the appropriate values into Eq. (2.20) to obtain
1 1 y [ 0] = 1 + 2 0 1 ( 2 ) = 4 2
2. Substitute for y[0] in Eq. (2.40) to find
y [1] =
1 1 1 3 + 2 1 (1) = 1 2 2 4 4
Example 2.5 Evaluation of a Difference Equation by means of a Computer A system is described y the difference equation
y [ n ] 1.143 y [ n 1] + 0.4128 y [ n 2] = 0.0675 x [ n ] + 0.1349 x [ n 1] + 0.675 x [ n 2]

Write a recursive formula that computes the present output from the past outputs and the current inputs. Use a computer to determine the step response
CHAPTER
of the system, the system output when the input is zero and the initial conditions are y[ 1] = 1 and y[ 2] = 2, and the output in response to the sinusoidal inputs x1[n] = cos(n/10), x2[n] = cos(n/5), and x3[n] = cos(7n/10), assuming zero initial conditions. Last, find the output of the system if the input is the weekly closing price of Intel stock depicted in Fig. 2.9, 2.9 assuming zero initial conditions. <Sol.> 1. Recursive formula for y[n]:
y [ n ] = 1.143 y [ n 1] 0.4128 y [ n 2] + 0.0675 x [ n ] + 0.1349 x [ n 1] + 0.675 x [ n 2]
2. Step response: Fig. 2.9(a 2.9(a). ). 3. Zero input response: Fig. 2.9(b 2.9(b). ). 4. The outputs due to the sinusoidal inputs x1[n], x2[n], and x3[n]: Fig. 2.9(c 2.9(c), ), (d), and (e). 5. Fig. 2.9(f 2.9(f) ) shows the system output for the Intel stock price unit. A comparison of peaks in Figs. 2.8 2.8and 2.9(f 2.9(f) ) Slightly delay!
2/1/2013
Amit Kulkarni
114
CHAPTER
Figure 2.8 Weekly closing price of Intel stock.
2/1/2013
Amit Kulkarni
115
CHAPTER
Fig. 2.9 (a).
2/1/2013
Amit Kulkarni
116
CHAPTER
Fig. 2.9 (b).
2/1/2013
Amit Kulkarni
117
CHAPTER
Fig. 2.9 (c).
Fig. 2.9(d 2.9(d). ).

CHAPTER
Fig. 2.9(e 2.9(e). ). Figure 2.9 Illustration of the solution to Example 2.16. (a) Step response of system. (b) Output due to nonzero initial conditions with zero input. (c) Output due to x1[n] = cos (1/10n). (d) Output due to x2[n] = cos (1/5n). (e) Output due to x3[n] = cos(7/10n).
2/1/2013
Amit Kulkarni
119
CHAPTER
Figure 2.9f Output associated with the weekly closing price of Intel stock.
2/1/2013
Amit Kulkarni
120
CHAPTER
2.6 Solving Differential and Difference Equations

Complete solution: y = y (h) + y (p) y (h) = homogeneous solution, y (p) = particular solution 2.6.1 The Homogeneous Solution Continuous-time case: N d k (h) 1. Homogeneous differential equation: ak k y ( t ) = 0 dt k =0 2. Homogeneous solution:
y (t) = c ierit
(h) i= 0
(2.41)
Coefficients ci is determined by I.C.

k k
3. Characteristic eq.:
a r
k =0
=0
(2.23)
Discrete-time case: 1. Homogeneous differential equation: 2. Homogeneous solution:
a y( ) [n k ] = 0
h k =0 k
y [n] = c irin
(h) i=1
(2.43)
Coefficients ci is determined by I.C.

Nk k
2/1/2013
a r
k =0
=0
(2.24)
121
Amit Kulkarni
CHAPTER
If a root rj is repeated p times in characteristic eqs., the corresponding solutions are r r r Continuous-time case: e j t , te j t , ..., t p 1e j t Discrete-time case:
rjn , nrjn , ..., n p 1rjn
Example 2.6 RC Circuit: Homogeneous Solution The RC circuit depicted in Fig. 2.10 is described by the differential equation
y ( t ) + RC
d y (t ) = x (t ) dt
Determine the homogeneous solution of this equation. <Sol.> d 1. Homogeneous Eq.: y ( t ) + RC y ( t ) = 0
dt
2. Homo. Sol.: y ( h ) t = c e r1t V () 1 3. Characteristic eq.:
Figure 2.10 RC circuit. r1 = 1/RC

t RC
1 + RCr1 = 0
y
(h)
4. Homogeneous solution:
( t ) = c1e
V
122
2/1/2013
Amit Kulkarni
CHAPTER
Example 2.7First-Order Recursive System: Homogeneous Solution Find the homogeneous solution for the first-order recursive system described by the difference equation
y [ n ] y [ n 1] = x [ n ]
<Sol.> 1. Homogeneous Eq.: 2. Homo. Sol.:
y [ n ] y [ n 1] = 0
y ( h ) [ n ] = c1r1n
r1 = 0
r1 =
4. Homogeneous solution:
y ( h ) [ n ] = c1 n
2.6.2 The Particular Solution A particular solution is usually obtained by assuming an output of the same general form as the input. Table 2.3
CHAPTER
Example 2.8 First-Order Recursive System (Continued): Particular Solution Find a particular solution for the first-order recursive system described by the difference equation
y [ n ] y [ n 1] = x [ n ]
if the input is x[n] = (1/2)n. <Sol.> 1. Particular solution form:
y [n] = c p
( p)
()
1 2
2. Substituting y(p)[n] and x[n] into the given difference Eq.:

CHAPTER
cp
()
1 2
cp
()
1 2
n 1
()
1 2
Both sides of above eq. are multiplied by (1/2) n
c p (1 2 ) = 1
3. Particular solution:
(2.25)
n
( p)
1 1 [ n] = 1 2 2
Figure 2.11 RC circuit.
Example 2.9 RC Circuit (continued): Particular Solution Consider the RC circuit of Example 2.6 and depicted in Fig. 2.10 2.10. Find a particular solution for this system with an input x(t) = cos(0t). <Sol.> d 1. Differential equation: y ( t ) + RC y (t ) = x (t )
dt
2. Particular solution form:
y ( p ) (t ) = c1 cos( t ) + c2 sin( t )
3. Substituting y(p)(t) and x(t) = cos(0t) into the given differential Eq.:
CHAPTER
c1 cos( t ) + c2 sin( t ) RC0c1 sin(0t ) + RC0c2 cos(0t ) = cos(0t ) c1 + RC 0c2 = 1
RC0 c1 + c2 = 0
4. Coefficients c1 and c2:
c1 =
1 1 + ( RC 0 )
2
and
c2 =
1 + ( RC 0 ) RC0
RC0
y(
p)
(t ) =
1 1 + ( RC0 )
cos ( 0t ) + 2
1 + ( RC0 )
sin (0t )
2.6.3 The Complete Solution Complete solution: y = y (h) + y (p) y (h) = homogeneous solution, y (p) = particular solution The procedure for finding complete solution of differential or difference equations is summarized as follows:
CHAPTER
Procedure 2.3: Solving a Differential or Difference equation 1. Find the form of the homogeneous solution y(h) from the roots of the characteristic equation. 2. Find a particular solution y(p) by assuming that it is of the same form as the input, yet is independent of all terms in the homogeneous solution. 3. Determine the coefficients in the homogeneous solution so that the complete solution y = y(h) + y(p) satisfies the initial conditions. Note that the initial translation is needed in some cases. Example 2.21 First-Order Recursive System (Continued): Complete Solution Find the complete solution for the first-order recursive system described by the difference equation
1 y[n 1] = x[n] (2.46) 4 if the input is x[n] = (1/2)n u[n] and the initial condition is y[ 1] = 8. <Sol.> n ( h) 1 1. Homogeneous sol.: y [ n] = c y[n]
1 4
()
2/1/2013
Amit Kulkarni
127
CHAPTER

3. Complete solution:
n
1 y [ n] = 2 2
( p)
1 1 y[n] = 2( )n + c1( )n 2 4
(2.47)
4. Coefficient c1 determined by I.C.: I.C.:
y [ 0] = x [ 0] + 1 4 y [ 1]
0 0
y [0] = x [0] + (1 4) 8 = 3
We substitute y[0] = 3 into Eq. (2.47), yielding
1 1 3 = 2 + c1 2 4
5. Final solution:
c1 = 1
n n
1 1 y [ n] = 2 + 2 4
for n 0
Example 2.10 RC Circuit (continued): Complete Response Find the complete response of the RC circuit depicted in Fig. 2.10 to an input x(t) = cos(t)u(t) V, assuming normalized values R = 1 and C = 1 F and assuming that the initial voltage across the capacitor is y(0) = 2 V.
CHAPTER

y
(h)
<Sol.> 1. Homogeneous sol.: 2. Particular solution:
( t ) = ce
t RC
V
sin ( t ) V
0 = 1
y(
p)
(t ) =
1 1 + ( RC )
cos ( t ) + 2
RC 1 + ( RC )
2
1 1 y ( t ) = ce t + cos t + sin t V 2 2
4. Coefficient c1 determined by I.C.:
R = 1 , C = 1 F
y(0) = y(0+) c = 3/2
2 = ce
0+
1 1 1 + cos 0+ + sin 0+ = c + 2 2 2
5. Final solution:
3 1 1 y ( t ) = e t + cos t + sin t V 2 2 2
2/1/2013
Amit Kulkarni
129
CHAPTER
Example 2.11 Financial Computations: Loan Repayment The following difference equation describes the balance of a loan if x[n] < 0 represents the principal and interest payment made at the beginning of each period and y[n] is the balance after the principal and interest payment is credited. As before, if r % is the interest rate per period, then = 1 + r/100.
y [ n ] = y [ n 1] + x [ n ]
Use the complete response of the first-order difference equation to find the payment required to pay off a $20,000 loan in 10 periods. Assume equal payments and a 10% interest rate. <Sol.> 1. We have = 1.1 and y[ 1] = 20,000, and we assume that x[n] = b is the payment each period. 2. The first payment is made when n = 0. The loan balance is to be zero after 10 payments, thus we seek the payment b for which y[9] = 0. 3. Homogeneous sol.:
y(
h)
[ n] = ch (1.1)n
Amit Kulkarni 130
2/1/2013
CHAPTER
y(
p)
[ n] = c p
Substituting y(p)[n] = cp and x[n] = b into the difference equation y[n] 1.1y[n 1] = x[n], we obtain
c p = 10b
y[n] = ch (1.1)n 10b, n 0
(2.48)
4. Coefficient ch determined by I.C.:

0
I.C.: y [ 0] = 1.1 y [ 1] + x [ 0] = 22, 000 + b
22,000 + b = ch (1.1) 10b

ch = 22,000 + 11b y [ n ] = ( 22, 000 + 11b )(1.1) 10b
n
Fig. 2.12.
5. Payment b: By setting y[9] = 0, we have
0 = ( 22, 000 + 11b )(1.1) 10b

9
b=
22, 000 (1.1)

9
11(1.1) 10
= 3, 254.91
131
2/1/2013
Amit Kulkarni
CHAPTER
Figure 2.12 Balance on a $20,000 loan for Example 2.23. Assuming 10% interest per period, the loan is paid off with 10 payments of $3,254.91.
Tools to analyze systems
2/1/2013
Amit Kulkarni
133
Outline
Transforms in context of problem solving Convolution Dirac delta function (t) Fourier Transform Sampling Laplace Transform
2/1/2013
Amit Kulkarni
134
Why use Transforms?

Transforms are not simply math curiosity sketched at the corner of a woodstove.
Way to reframe a problem in a way that makes it easier to understand, analyze and solve.
2/1/2013
Amit Kulkarni
135
General Scheme using Transforms

Problem Equation of the problem Solution of the equation Result
Transformation Inverse transformation
= EASY
Transformed equation
Solution of the transformed equation
= HARD
2/1/2013
Amit Kulkarni
136
Which Transform to Use?

Application Signal Processing Control Theory Continuous Domain Fourier T. Laplace T. Discrete Domain Discrete F.T. (DFT/FFT) z-Transform
2/1/2013
Amit Kulkarni
137
Typical Problem
Given an input signal x(t), what is the output signal y(t) after going through the system?
To solve it in the time domain (t) is cumbersome!

Integrating Differential Equation?

Lets have a simple first order low-pass filter with resistor R and capacitor C:
The system is described by diff. eq.:
RCy ' (t ) + y (t ) = x (t )
To find a solution, we can integrate. Ugh!
Convolution
Math operator (symbol *) that takes two input functions (x(t) and h(t)) and produces a third (y(t))
y (t ) = x (t ) h(t ) = x (t )h ( )d
Expresses the amount of overlap of one function x(t) as it is shifted over another function h(t). Way of blending one function with another.
2/1/2013
Amit Kulkarni
140
Frame-by-frame convolution
Can be visualized as flipping one function (x), sliding it, and doing the dot product.
2/1/2013
Amit Kulkarni
141
More convolution examples

* =
Smoothing
2/1/2013
Amit Kulkarni
142
(Moving Average of GOOG)
2/1/2013
Amit Kulkarni
143
(Gaussian Blur in 2D) Convolution is heavily used in image processing
2/1/2013
Amit Kulkarni
144
Dirac Delta function (t)
2/1/2013
Amit Kulkarni
145
Convolution with Dirac delta (t)

Convolving a signal with Dirac delta (t) simply yields the same signal.
2/1/2013
Amit Kulkarni
146
Delay operator as convolution with (t-)

Convolving with a shifted delta (t-) shifts the original signal: delay.
2/1/2013
Amit Kulkarni
147
Convolution Properties
Convolution is a linear operation and therefore has the typical linear properties:
Commutativity Associativity Distributivity Scalar multiplication
2/1/2013
Amit Kulkarni
148
Using Convolution to Solve

Again same first order low-pass filter:
The system is described by its impulse response:

t
Solution is convolution impulse resp. with x(t)
t
2/1/2013
*
Amit Kulkarni
t
149
Use a Convolution to Solve

Convolution is expensive to compute. Little intuition about output signal y(t).
2/1/2013
Amit Kulkarni
150
Fourier Transform
Jean-Baptiste Fourier had crazy idea (1807): Any periodic function can be rewritten as a weighted sum of sines and cosines of different frequencies. Called Fourier Series
2/1/2013
Amit Kulkarni
151
Square-Wave Deconstruction
2/1/2013
Amit Kulkarni
152
Other examples
2/1/2013
Amit Kulkarni
153
FT expands this idea

Take any signal (periodic and non-periodic) in time domain and decompose it in sines + cosines to have a representation in the frequency domain.
Time Domain
FT
Frequency Domain
Real: Cosine Coefficients
f t
FT
Imaginary: Sine Coefficients
f
FT: Formal Definition
Convention: Upper-case to describe transformed variables: Transform: F{ x(t) } = X() or X(f) Inverse:
2/1/2013
(=2f)
F-1{Y() or Y(f) }= y(t)

Amit Kulkarni 155
FT gives complex numbers

You get complex numbers
Cosine coefficients are real Sine coefficients are imaginary
2/1/2013
Amit Kulkarni
156
Complex plane
Complex number can be represented:
Combination of real + imaginary value:
x +iy
Amplitude + Phase A and
2/1/2013
Amit Kulkarni
157
Alternative representation of FT
Complex numbers can be represented also as amplitude + phase.
Real Amplitude
f t
FT
+
Imaginary
OR
+
Phase
2/1/2013
Amit Kulkarni
158
Example Fourier Transform

Fast moving vs slow moving signals
Amplitude Spectrum
FT
Amplitude Spectrum
t
FT
f
159

Time Domain t Frequency Domain Real
Real
Real

Time Domain t Frequency Domain
Real
Real
DC component
2/1/2013
Amit Kulkarni
161
2/1/2013
Amit Kulkarni
162
2/1/2013
Amit Kulkarni
163
2/1/2013
Amit Kulkarni
164
Note: FT is imaginary for sine

FT of Delay (t)
Amplitude + phase is easier to understand: Amplitude:
Gives you information about frequencies/tones in a signal.
Phase:
More about when it happens in time.
2/1/2013
Amit Kulkarni
166
Important FT Properties
Addition Scalar Multiplication Convolution in time t
F {a (t ) + b(t )} = A( ) + B( )
F {ka (t )} = kA( )
F {x (t ) h(t )} = F {x (t )}F {h(t )} = X ( ) H ( )

Convolution in frequency
F { X (t ) H ( )} = 2x (t )h(t )
1
FT time
frequency duality
Frequency Domain
wide narrow Convolution Multiplication Sinc Box Gauss Real+Even (just cosine) Im + Odd (just sine) Etc..
168
Time Domain
narrow wide Multiplication Convolution Box Sinc Gauss Real + Even Real + Odd Etc..
FT: Reframing the problem in Frequency Domain

Problem x(t),h(t)
Solution of the equation
Result
Fourier Transform Inverse Fourier Transform
= EASY
X(), H()
X()H()
= HARD
Completely sidesteps the convolution!

Amit Kulkarni 169
2/1/2013
FT: Another Example

What is the amplitude spectrum |Y(f)| of a voice signal (bandlimited to 5 kHz) when multiplied by a cosine f=15 kHz?
(Note: this is Amplitude Modulation

- 5 kHz
5 kHz
AM radio)
170
FT: Solution
x(t)=?
t
-15 kHz
|X(f)|
|Y(f)|
15 kHz
*
f
-15 kHz 15 kHz
=
f
-20 kHz -10 kHz 10 kHz 20 kHz
Remember! Convolving with (f-f) == Shifting signal
(Look Ma! No Algebra!)

FT Gaussian Blur
Space
Frequency
x
f f
=
f
2/1/2013
Amit Kulkarni
172
Sampling Theorem
In order to be used within a digital system, a continuous signal must be converted into a stream of values. Done by sampling the continuous signal at regular intervals. But at which interval?
2/1/2013
Amit Kulkarni
173
Sampling Theorem
Sampling can be thought of multiplying a signal by a pulse train:
...
x(t)
t
x
... t
=
... ... t
2/1/2013
Amit Kulkarni
174
Aliasing
If sampling rate is too small compared with frequency of signal, aliasing WILL occur:
2/1/2013
Amit Kulkarni
175
Fourier Analysis of Sampling

The FT of a pulse train with frequency fs is another pulse train with interval 1/fs:
1/fs
... ... t
fs FT
... ... f
2/1/2013
Amit Kulkarni
176
Fourier Analysis of Sampling

Aliasing will happen if fs <2 fmax
Nyquist frequency = fs/2
Time D Domain Frequency Domain
-fmax
fmax
2f m f s<
ax
2/1/2013
Amit Kulkarni
177
A few sampling frequencies

Telephone systems: 8 kHz CD music: 44.1 kHz DVD-audio: 96 or 192 kHz Aqua robot: 1 kHz Digital Thermostat (HMTD84) : 0.2 Hz
Laplace transform
Introduced by Pierre Simon Marquis de Laplace (17491827). Practically used by Oliver Heaviside (1850-1925) Transform relationship between x(t) and X(s):
Laplace Transform
x(t)
X(s)
Inverse Laplace Transform
2/1/2013
Amit Kulkarni
179
Laplace Transform
Formal definition:
L[ f (t )] = F ( s ) = f (t )e dt
Compare this to FT:
0
st
F ( ) =
Small differences:
f (t )e it dt
Integral from 0 to to for Laplace

f(t) for t<0 is not taken into account
-s instead of -i
Laplace Transform Pairs

The Laplace transform originated from the Fourier transform.
Definition:
X ( s ) = x ( t ) e st dt
1 st x(t) = X ( s ) e ds 2j j
where s is a complex variable that is defined as s=+jw .
+ j
2/1/2013
Amit Kulkarni
181
Laplace transform
Definition:
X (s) = x (t ) =
st x ( t ) e dt
1 2 j
+ j
j
X ( s ) e + st ds
For a causal system, where t 0, the Laplace Transform will be:
X (s) =
st x ( t ) e dt 0
+ j
x (t ) =
X ( s ) e + st ds
182
The Laplace transform originated from the Fourier transform.
X ( s ) | s = + j 2 f =
0 0
( + j 2 f ) t x ( t ) e dt
X( + j 2 f) =
] [ x ( t ) e e
t
j 2 ft
dt
0 2/1/2013 Amit Kulkarni
-j 2ft dt y(t) e
183
Common Laplace Transfom

Name Impulse Step Ramp Exponential Sine f (t ) 1 f (t ) = 0 F(s)
t=0 t>0
1
1 s 1 s2 1 s+a
f (t ) = 1 f (t ) = t
f (t ) = e at
f (t ) = sin(t )
at
s2 + 2
Damped Sine f (t ) = e
2/1/2013
sin(t )
( s + a )2 + 2
184
Amit Kulkarni
Laplace Transform Properties

Similar to Fourier transform:
Addition/Scaling
L[af1 (t ) bf 2 (t )] = aF1 ( s ) bF2 ( s )

Convolution
t
f 1(t )f 2( ) d = F1 ( s ) F2 ( s )
Derivation
2/1/2013
d L f (t ) = sF ( s ) f (0 ) dt
Amit Kulkarni 185
Transfer Function H(s)

Definition
H(s) = Y(s) / X(s) X(s) H(s) Y(s)
Relates the output of a linear system (or component) to its input. Describes how a linear system responds to an impulse. All linear operations allowed
Scaling, addition, multiplication.
2/1/2013
Amit Kulkarni
186
RC Circuit Revisited
Time e Domain
RC
dy +y=x dt
Laplace Domain
1 s
1 1 + RCs
1 1 1 = 1 s(1 + RCs ) s +s RC
2/1/2013
Amit Kulkarni
187
Poles and Zeros

A( s ) Given F ( s ) = B( s) A( s ) = an s + ... + a1s + a0
n
B( s ) = bm s m + ... + b1s + b0 Poles are the values of s for which B(s) = 0 Zeros are the values of s for which A(s) = 0 Poles and zeros are complex Order of system = # poles = m
Poles and Zeros

1 For example, L[e ] = s+a Pole is s = a No Zeros
at
For sine : L[sin(t )] = 2 s + 2 Poles are s = i, i

No Zeros
Poles and Zeros

Name Impulse Step Ramp Exponential Sine f (t ) 1 f (t ) = 0 F(s)
t=0 t>0
Poles n/a 0 0 (double) -a -i,i
1
1 s 1 s2 1 s+a
f (t ) = 1 f (t ) = t
f (t ) = e at
f (t ) = sin(t )
at
s2 + 2
Damped Sine f (t ) = e
2/1/2013
sin(t )
( s + a )2 + 2 -a-i,-a+i
190
Amit Kulkarni
Poles and Zeros

If pole has: Real negative: exponential decay Real positive: exponential growth If imaginary 0: oscillation of frequency
2/1/2013
Amit Kulkarni
191
Effect of Poles Location
2/1/2013
Increasing Frequency
Amit Kulkarni
192
Where do you dont want to be?
2/1/2013
Increasing Frequency
Amit Kulkarni
193
MATLAB functions
fourier laplace ifourier ilaplace Search for these functions and their syntax in MATLAB help
2/1/2013
Amit Kulkarni
194
Representation of discretetime systems

x [n]
T{.}
x[n]
y [n]
Example: Ideal delay system
y [n] = x [n nd]
- < n < y[n]
A memoryless system
y[n] = x[n]
The accumulator system

y[n] =
k =
x[k ]
Moving average
1 y[n] = ( x[n 1] + x[n] + x[n + 1] + x[n + 2]) 4
1 y[n] = M1 + M 2 + 1
k = M1
M2
x[n k ]
It is a low pass filter!!!
Memory Causality Stability Time invariance Linearity
Memory:
A system is memoryless if y[n] = f ( x[n] )
i.e. it sees only present values.
A system has memory if y [n] depends on previous values

it can also depend on present and future values!
Causality
A system is causal if the output y[n] depends only on present and/or past values.
On-line systems are causal by definition
Time-invariance
A system is time invariant if a shift in the input causes a corresponding shift of the output.
For all n0 x1 [n] = x [n-n0] gives y1[n] = y [n-n0]
Stability
A system is stable if every bounded input sequence produces a bounded output
i.e. it never diverges. If |x[n]| Bx < then |y[n]| By <
Linearity
Linear systems obey the principle of superposition.
1) Additive property.
T {x1[n] + x2[n]} = T {x1[n]} + T {x2[n]} = y1[n] + y2[n]
2) Scaling property
T {a x1[n]} = a T {x1[n]} = a y[n] Altogether: T {a x1[n] + b x2[n]} = a T{x1[n]} + b T{x2[n]}
More generally: If x[n] = k ak xk[n] then y[n] = k ak yk[n]
where yk[n] is the system response to the input xk[n]
Exercise:
Which properties (linearity, causality, time-invariance, stability and memory) posses the following systems:
a) y[n] = 3 x[n] 4 x[n-1] b) y[n] = 2 y[n-1] + x[n+2] c) y[n] = n x[n] d) y[n] = cos (x[n]) e) y[n] = log10 (x[n]) f) y[n] = x[n]4 g) The accumulator system h) The ideal delay system i) The moving average system
A sequence can be represented as a linear combination of delayed impulses:
x[n] =
k =
x[k ] [n k ]
Linear time-invariant systems (LTI)

Let hk[n] be the response to [n-k] (an impulse at n = k)
y[n] = T { x[n]} = T x[k ] [n k ] k =

y[n] = x[k ] T { [n k ]}

If the system is linear
= =
k =
x[k ] h [n]
k
k =
x[k ] h[n k ]
If the system is time-invariant
Convolution sum
y[n] =
k =
x[k ] h[n k ]
Convolution sum
y[n] = x[n] h[n]
Linear time-invariant systems can be described by the convolution sum!
Note that:
x[n] = x[k ] [n k ]

y[n] = x[k ] h[n k ]
A linear time-invariant system can be completely characterized by its input response h[n] [n] LTI h[n]
Properties of LTI systems

(or properties of convolution)
Convolution is conmutative
x[n] h[n] = h[n] x[n]
Convolution is distributive
x[n] (h1[n] + h2[n]) = x[n] h1[n] + x[n] h2[n]
Cascade connection:
y[n] = h1[n] [ h2[n] x[n] ] = [ h1[n] h2[n] ] x[n]
x[n]
h2
h1
y[n]
=
x[n] h1h2 y[n]
Parallel connection
y[n] = h1[n] x[n] + h2[n] x[n] ] = [ h1[n] + h2[n] ] x[n]
h1 x[n] h2 + y[n]
=
x[n] h1+h2 y[n]
LTI systems are stable iff
k =
h[k ] <
LTI systems are causal if
h[n] = 0
n<0
2/1/2013
Amit Kulkarni
212
Content
Introduction z-Transform Zeros and Poles Region of Convergence Important z-Transform Pairs Inverse z-Transform z-Transform Theorems and Properties System Function
Why z-Transform?
A generalization of Fourier transform Why generalize it?
FT does not converge on all sequence Notation good for analysis Bring the power of complex variable theory deal with the discrete-time signals and systems
2/1/2013
Amit Kulkarni
214
Definition
The z-transform of sequence x(n) is defined by
X ( z) =
Let z = ej.
n =
x ( n) z
n
Fourier Transform
X (e ) =
n =
x ( n)e
j n
215
The Laplace Connection

Consider the Laplace Transforms of x(t) and y(t):
Y (s ) = L{y (t )} = L{x(t T )} = e
1
Ts
X (s )
Ts
Equate the transform domain delay operators:
=e
Ts
z =e
Examine s-plane to z-plane mapping . . .
2/1/2013
Amit Kulkarni
216
z-Plane
X ( z) =
n =
x ( n) z
Im
z = e j
X (e ) =
n =
x ( n)e
j n
Re
Fourier Transform is to evaluate ztransform on a unit circle.

z-Plane
X(z) Im
z = e j
Re Im Re
2/1/2013
Amit Kulkarni
218
Periodic Property of FT
X(z) X(ej)
Im Re
Can you say why Fourier Transform is a periodic function with period 2?
2/1/2013
Amit Kulkarni
219
Definition
Give a sequence, the set of values of z for which the ztransform converges, i.e., |X(z)|<, is called the region of convergence.
| X ( z ) |=
n =
n x ( n ) z =
n =
n | x ( n ) || z | <
ROC is centered on origin and consists of a set of rings.

Example: Region of Convergence

| X ( z ) |=
n = n x ( n ) z = n = n | x ( n ) || z | <
Im
ROC is an annual ring centered on the origin.
r Re
Rx <| z |< Rx +
ROC = {z = re | Rx < r < Rx + }
j
2/1/2013
Amit Kulkarni
221
Stable Systems
A stable system requires that its Fourier transform is uniformly convergent.
Im Fact: Fourier transform is to evaluate ztransform on a unit circle. A stable system requires the ROC of ztransform to include the unit circle. Re
2/1/2013
Amit Kulkarni
222
Example: A right sided Sequence
x(n)
...
-8 -7 -6 -5 -4 -3 -2 -1 1 2 3 4 5 6 7 8 9 10
2/1/2013
Amit Kulkarni
223

For convergence of X(z), we require that
X ( z) =
n =
a u (n)z
n n =0
1 | az |<
n =0
| az 1 |< 1
| z |>| a |
= a n z n = (az 1 ) n
n =0
1 z X ( z ) = (az ) = = 1 1 az za n =0
1 n
| z |>| a |
224
Example: A right sided Sequence n ROC for x(n)=a u(n)

Which one is stable?
Im Im
a
Re
a
Re
2/1/2013
Amit Kulkarni
225
Example: A left sided Sequence
-8 -7 -6 -5 -4 -3 -2 -1
1 2 3 4 5 6 7 8 9 10
2/1/2013
...
x(n)
Amit Kulkarni
226

For convergence of X(z), we require that
X ( z ) = a u ( n 1)z
n n =
1
1 | a z|< n =0
| a 1 z |< 1
| z |<| a |
= a n z n
n =
= a n z n
n =1
= 1 a n z n
n =0
1 z X ( z ) = 1 (a z ) = 1 = 1 1 a z z a n=0
1 n
| z |<| a |
227
2/1/2013
Amit Kulkarni
Example: A left sided Sequence n ROC for x(n)=a u( n1)

Which one is stable?
Im Im
a
Re
a
Re
2/1/2013
Amit Kulkarni
228
Represent z-transform as a Rational Function

P( z ) X ( z) = Q( z )
where P(z) and Q(z) are polynomials in z.
Zeros: The values of zs such that X(z) = 0 Poles: The values of zs such that X(z) =

x( n) = a n u ( n)
Im
X ( z) =
z , za
| z |>| a |
a
Re
ROC is bounded by the pole and is the exterior of a circle.
2/1/2013
Amit Kulkarni
230

x(n) = a n u (n 1)
Im
X ( z) =
z , za
| z |<| a |
a
Re
ROC is bounded by the pole and is the interior of a circle.
2/1/2013
Amit Kulkarni
231
Example: Sum of Two Right Sided Sequences

n 1 n x ( n) = ( 1 ) u ( n ) + ( 2 3 ) u (n)
1 z z 2 z ( z 12 ) X ( z) = + = 1 1 1 z2 z+3 (z 1 2 )( z + 3 )
Im
1/12 1/3 1/2
Re
ROC is bounded by poles and is the exterior of a circle. ROC does not include any pole.
2/1/2013
Amit Kulkarni
232
Example: A Two Sided Sequence

n 1 n x ( n) = ( 1 ) u ( n ) ( 3 2 ) u ( n 1)
1 z z 2 z ( z 12 ) X ( z) = + = 1 1 1 z+3 z2 (z + 1 3 )( z 2 )
Im
1/12 1/3 1/2
ROC is bounded by poles and is a ring.

Re
ROC does not include any pole.

Example: A Finite Sequence

x(n) = a n ,
X ( z) = a z
n n =0
N-1 zeros
0 n N 1
n
N 1
= ( az )
n =0
N 1
1 n
Im
1 (az 1 ) N = 1 az 1
1 zN aN = N 1 z za
ROC: 0 < z <

N-1 poles
ROC does not include any pole.

Re
Always Stable
Properties of ROC
A ring or disk in the z-plane centered at the origin. The Fourier Transform of x(n) is converge absolutely iff the ROC includes the unit circle. The ROC cannot include any poles Finite Duration Sequences: The ROC is the entire z-plane except possibly z=0 or z=. Right sided sequences: The ROC extends outward from the outermost finite pole in X(z) to z=. Left sided sequences: The ROC extends inward from the innermost nonzero pole in X(z) to z=0.
2/1/2013
Amit Kulkarni
235
More on Rational z-Transform

Consider the rational z-transform with the pole pattern:
Im
Find the possible ROCs
a b
c
Re
2/1/2013
Amit Kulkarni
236

Im Case 1: A right sided Sequence.
a b
c
Re
2/1/2013
Amit Kulkarni
237

Im Case 2: A left sided Sequence.
a b
c
Re
2/1/2013
Amit Kulkarni
238

Im Case 3: A two sided Sequence.
a b
c
Re
2/1/2013
Amit Kulkarni
239

Im Case 4: Another two sided Sequence.
a b
c
Re
2/1/2013
Amit Kulkarni
240
Z-Transform Pairs
Sequence
( n ) ( n m ) u ( n) u ( n 1)
z-Transform
1
ROC All z All z except 0 (if m>0) or (if m<0)

| z |> 1 | z |< 1
z m 1 1 z 1 1 1 z 1
1 1 az 1
1 1 az 1
a u ( n)
a u ( n 1)
n
| z |>| a |
| z |<| a |
241
2/1/2013
Amit Kulkarni
Z-Transform Pairs
Sequence
[cos 0 n]u (n) [sin 0 n]u ( n)
[r cos 0 n]u (n)
n
z-Transform
1 [cos 0 ]z 1 1 [2 cos 0 ]z 1 + z 2
[sin 0 ]z 1 1 [2 cos 0 ]z 1 + z 2 1 [r cos 0 ]z 1 1 [2r cos 0 ]z 1 + r 2 z 2 [r sin 0 ]z 1 1 [2r cos 0 ]z 1 + r 2 z 2
ROC
| z |> 1 | z |> 1 | z |> r | z |> r
[r sin 0 n]u (n)

n
a n 0
2/1/2013
0 n N 1 otherwise
Amit Kulkarni
1 a N zN 1 az 1
| z |> 0
242
Linearity
Z [ x(n)] = X ( z ),
Z [ y (n)] = Y ( z ),
z Rx
z Ry
Z [ax(n) + by (n)] = aX ( z ) + bY ( z ),
z Rx R y
Overlay of the above two ROCs
2/1/2013
Amit Kulkarni
243
Shift
Z [ x(n)] = X ( z ), z Rx
Z [ x(n + n0 )] = z X ( z )
n0
z Rx
244
Multiplication by an Exponential Sequence
Z [ x(n)] = X ( z ),
Rx- <| z |< Rx +
Z [a x(n)] = X (a z )
n
z | a | Rx
245
Differentiation of X(z)
Z [ x(n)] = X ( z ), z Rx
dX ( z ) Z [nx(n)] = z dz
z Rx
246
Conjugation
Z [ x(n)] = X ( z ), z Rx
Z [ x * (n)] = X * ( z*)
z Rx
247
Reversal
Z [ x(n)] = X ( z ), z Rx
Z [ x(n)] = X ( z )
z 1 / Rx
248
Real and Imaginary Parts

Z [ x(n)] = X ( z ), z Rx
Re[ x(n)] = 1 2 [ X ( z ) + X * ( z*)]

Im [ x(n)] =
2/1/2013
z Rx
z Rx
249
1 2j
[ X ( z ) X * ( z*)]
Amit Kulkarni
Initial Value Theorem

x(n) = 0, for n < 0
x(0) = lim X ( z )
z
Convolution of Sequences
Z [ x(n)] = X ( z ), Z [ y (n)] = Y ( z ), z Rx z Ry
Z [ x(n) * y (n)] = X ( z )Y ( z )
z Rx R y
251
Convolution of Sequences
x ( n) * y ( n) =
k =
x(k ) y (n k )
n Z [ x(n) * y (n)] = x(k ) y (n k ) z n = k =
k =
x(k ) y(n k )z
n =
Amit Kulkarni
k =
x(k ) z k
n =
n y ( n ) z
= X ( z )Y ( z )
2/1/2013 252
Shift-Invariant System
x(n) y(n)=x(n)*h(n)
h(n)
X(z)
H(z)
Y(z)=X(z)H(z)
253
Shift-Invariant System
X(z)
Y ( z)
H(z)
Y ( z) H ( z) = X ( z)
Nth-Order Difference Equation
k =0 N
r =0
Y ( z ) ak z
k =0
= X ( z ) br z
r =0
H ( z ) = br z
r =0
k =0
ak z
255
Representation in Factored Form

Contributes poles at 0 and zeros at cr
A (1 cr z )
1 1 ( 1 d z r ) k =1 r =1 N M
H ( z) =
Contributes zeros at 0 and poles at dr

Stable and Causal Systems

Causal Systems : ROC extends outward from the outermost pole.
Im
H ( z) =
A (1 cr z )
1 1 ( 1 d z r ) k =1 r =1 N
Re
2/1/2013
Amit Kulkarni
257
Stable and Causal Systems

Stable Systems : ROC includes the unit circle.
Im
H ( z) =
A (1 cr z )
1 1 ( 1 d z r ) k =1 r =1 N
1 Re
2/1/2013
Amit Kulkarni
258
Example
Consider the causal system characterized by
y (n) = ay (n 1) + x(n)
Im
1 H ( z) = 1 1 az
h( n) = a u ( n)
n
Re
259
Determination of Frequency Response from polezero pattern

A LTI system is completely characterized by its polezero pattern.
Im
Example:
z z1 H ( z) = ( z p1 )( z p2 )
j 0
p1 z1 p2
260
j 0
Re
H (e
2/1/2013
)=
(e j0
e j0 z1 p1 )(e j0 p2 )
Amit Kulkarni
|H(e )|=?
Example:
A LTI system by its polej is completely characterized j zero pattern.

Im
H(e )=?
z1
z z1 H ( z) = ( z p1 )( z p2 )
j 0
p1
e
j 0
Re
H (e
2/1/2013
)=
(e j0
e j0 z1 p1 )(e j0 p2 )
p2
261
Amit Kulkarni
|H(e )|=?
Example: |H(ej)| = | | ||
A LTI system by its polej is completely characterized j zero pattern.

Im 2
H(e )=?
z1
| |
p1
e
1 3
j 0
Re
H(ej) = 1(2+ 3 )
p2
262
Example
20 10
dB Im
0 -10 0 2 1 2 4 6 8
a
Re
0 -1 -2 0 2 4 6 8
263

Emb DSP1

Uploaded by

Copyright:

Available Formats

You might also like

Emb DSP1

Uploaded by

Document Information

Original Description:

Original Title

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

Emb DSP1

Uploaded by

Copyright:

Available Formats

Fundamentals

A Typical DSP System

ADC DSP DAC

DSP Chip Memory Converters (Optional)

Multiply and Add

5*3 = 15 0000 0011 0000 0011 =

added multiple times

Most Common Operation in DSP A = B*C + D E = F*G + A

Multiply, Add, and Accumulate MAC Instruction

Drop in Multiplication Times

STORED PROGRA M AND DATA

ARITHMETIC A = ADDRESS LOGIC D = DATA UNIT

Tools of the Trade

Why Digital Processing?

Advantages to Digital Processing

Upgradability and Flexibility

A CD player always plays the same music quality

Adaptive Filters gain f

Linear Phase Filters phase

Practical DSP Systems

Infinite resolution in amplitude

Digital Signal Processing (DSP) Advantages

Analogs Place in DSP

Analog circuits may be required to drive output transducers

Any number can be represented as a series of 1s and 0s Decimal 3 in binary

Digit Number Binary

Digit Number Binary

Binary and Hex

4 bits of binary system is represented by a single hex digit

Decimal 26 in binary and hex

Digit Number Binary Hex

Twos Complement Notation

Twos complement addition

Twos Complement Binary

What a DSP system is?

Common DSP algorithms and applications

What Do DSPs Need to Do Well?

Todays DSP Killer Apps

Decoding DSP Lingo

Signals and their representation

The Time domain

The Frequency Domain

Have we lost some information?

The Frequency Domain: A Natural Domain

Continuous and discretetime signals

Continuous and Discrete valued singals

Even and odd signals

Periodic signals, nonperiodic signals

Deterministic signals, random signals

Causal and anticausal Signals

Noncausal : nozero values in both

Right-handed and left-handed Signals

Finite and infinite length

Basic Operations on Signals

Operations performed on dependent signals

Operations performed on the independent signals

Operations performed on the independent signals

Most Common Operation in DSP A = BC + D E = FG + A