MUTHAYAMMAL ENGINEERING COLLEGE
(An Autonomous Institution)
(Approved by AICTE, New Delhi, Accredited by NAAC & Affiliated to Anna University)
Rasipuram - 637 408, Namakkal Dist., Tamil Nadu
Course Code &
21ECC09
Course Name
& DIGITAL SIGNAL PROCESSING
:
Year/Sem/Sec
II/IV/A,B,C
: MUST KNOW CONCEPTS MKC
S.No
ECE Term
Notation Concept/Definition/Meaning/Units/ 2022-23
Units
( Symbol) Equation/Expression
UNIT I FOURIER ANALYSIS OF DISCRETE TIME SIGNALS
A Signal a function that Conveys information
1 Signals
about a phenomenon
2 Systems It produces an output for a given input signal.
X(t)=Asinωt
3 Sinusoidal signals
X(t)=A cos ωt
A continuous signal that contains time-varying
4 Analog signal
quantities.
It is a signal that is being used to represent data as
5 Digital signal
a sequence of discrete values.
A discrete time signal x (n) is a function of an
6 Discrete time signal
independent variable that is an integer.
A discrete or an algorithm that performs some
7 Discrete time system
prescribed operation on a discrete time signal.
Elementary discrete Unit Step signal, Unit Ramp, Unit Impulse and
8
time signals Exponential Signal.
9 Exponential signal x (n)=an where a is real x(n)-Real signal
1. Energy and power signals
Classification of 2. Periodic and A periodic signals
10
discrete time signals 3. Symmetric(even) and Ant symmetric (odd)
signals
11 Energy Signal E =∑ │x (n)│2
P = Lt (1/2N+1)
12 Power Signal
N-->∞
N-1
13 DFT Equation X(k)= ∑ x(n) e-j2πnk/N , 0<=k<=N-1
n=0
1
N-1
14 IDFT Equation x(n)= 1/N ∑ X(k) e j2πnk/N , 0<=n<=N-1
n=0
Periodicity, Linearity and symmetry,
Multiplication of two DFTs, Circular convolution,
15 Properties of DFT
Time reversal, Circular time shift and frequency
shift, Complex conjugate, Circular correlation
Two methods of 1. Concentric Circle Method
16
Circular convolution 2. Matrix multiplication Method
To find N point DFT, have to add (N-L) zeros at
17 zero padding
the sequence x(n).
Methods used for the 1. Overlap-add method and
18
sectional convolution 2. Overlap-save method
To each data block we append M-1 zeros and
19 Overlap-add method
perform N point circular convolution.
To each data block we add M-1 zeros at the initial
20 Overlap-save method
point of input signal.
The Fast Fourier Transform is an algorithm used
21 FFT
to compute the DFT.
If the number of output points N can be expressed
as a power of 2 that is N=2M, where M is an
22 Radix-2 FFT
integer, the algorithm is known as radix-2
algorithm.
Decimation-In-Time algorithm is used to calculate
23 DIT algorithm the DFT of a N point sequence. The sequence x(n)
is often splitted into smaller sub-sequences.
The output sequence X(k) is divided into smaller
24 DIF algorithm and smaller sub-sequences , that is why the name
Decimation In Frequency.
Applications of FFT 1) Linear filtering 2)Correlation
25
algorithm 2) Spectrum analysis
UNIT II DESIGN OF IIR FILTER
A filter is a circuit capable of passing certain
26 Filter
frequencies while attenuating other frequencies.
1. IIR filter
Types of filters based
27 2. FIR filter
on impulse response
IIR filter IIR filters are easily realized recursively
28
The round off noise in IIR filters is more.
The Magnitude response of Butterworth
filter decreases monotonically as the
Butterworth filter
29 frequency increases.
The Poles of the Butterworth filter lies
along the circle.
A baseband signal is centered around DC (zero)
30 Low pass signal
frequency.
A high-pass filter (HPF) is an electronic filter that
31 High pass signal passes signals with a frequency higher than a
certain cutoff frequency.
A band of frequencies ranging from some non
32 Band pass signal
zero value to another non zero value.
2
The design of IIR filter is realizable and stable.
33 Structure of IIR filter The impulse response h(n) for a realizable filter is
h(n)=0 for n≤0
Advantage of direct In direct form ΙΙ structure, the number of memory
34 form ΙΙ structure over locations required is less than that of direct form Ι
direct form Ι structure structure.
1. Map the desired digital filter specifications into
Design digital filters those for an equivalent analog filter.
35
from analog filters 2. Derive the analog transfer function for the
analog prototype.
Procedures for 1. Impulse invariance method.
36 digitizing the TF of 2. Bilinear transformation method.
analog filter
The impulse response of resulting digital filter is a
Impulse invariant
37 sampled version of the impulse response of the
method
analog filter.
The Magnitude response of Chebyshev
filter will not decrease monotonically with
frequency because it exhibits ripples in pass
Chebyshev Filter
38 band or stop band.
The Transition width is very small
The poles of chebyshev filter lies along the
ellipse.
The mapping from the S-plane to the Z-plane is in
Bilinear bilinear transformation is
39
transformation -1 -1
S=2 (1- z )/ T (1+ z )
Response of analog
40
filter
sampled signal of
41
analog filter output
When the desired magnitude response is piece-
wise constant over frequency, this compression
42 Pre-warping can be compensated by introducing a suitable pre-
scaling, or pre-warping the critical frequencies by
using the formula
1. The bilinear transformation provides one-to-
Advantages of one mapping.
43 bilinear 2. Stable continuous systems can be mapped into
transformation realizable, stable digital systems.
3. There is no aliasing.
1. The mapping is highly non-linear producing
Disadvantages of frequency, compression at high frequencies.
44 bilinear 2. Neither the impulse response nor the phase
transformation response of the analog filter is preserved in a
digital filter obtained by bilinear transformation.
Advantage of cascade Quantization errors can be minimized if we
45
realization realize an LTI system in cascade form.
It is a graphical representation of the relationships
46 Signal flow graph between the variables of a set of linear difference
equations.
47 Transposition If we reverse the directions of all branch
3
transmittance and interchange the input and output
theorem in the flowgraph, the system function remains
unchanged.
1. Reverse the directions of all branches in the
signal flow graph
48 Transposed structure
2. Interchange the input and outputs.
3. Reverse the roles of all nodes in the flow graph.
1.center frequency
Important parameters
2. upper critical frequency
49 of band pass
3. low critical frequency
50 Sampling frequency T=1/Fs
UNIT III DESIGN OF FIR FILTER
A filter is a circuit capable of passing certain
51 Filter
frequencies while attenuating other frequencies.
Types of filters based 1. IIR filter
52
on impulse response 2. FIR filter
1. Low pass filter
Types of filters based
2. High pass filter
53 on frequency
3. Band pass filter
response
4. Band reject filter
A band of frequencies ranging from some non
54 Band pass signal
zero value to another non zero value.
A baseband signal is centered around DC (zero)
55 Low pass signal
frequency.
A high-pass filter (HPF) is an electronic filter that
56 High pass signal passes signals with a frequency higher than a
certain cutoff frequency.
FIR filters can be realized recursively and non-
recursively.
57 FIR filter
Errors due to round off noise are less severe in
FIR filter
IIR filter IIR filters are easily realized recursively
58
The round off noise in IIR filters is more.
Linear phase FIR Phase delay, α = (N-1)/2 (i.e., phase delay is
59
filter constant)Impulse response, h(n) = h(N-1-n)
Impulses occur at the mirror image in the first
Anti Symmetric FIR
60 quadrant and third quadrant or second quadrant
Filters
and Fourth quadrant or both.
The Optimum Equi ripple design Criterion is used
Optimum equi ripple
61 for designing FIR Filters with Equal level
design criterion
filteration throughout the Design.
Design techniques of (1) Window method, (2) Frequency sampling
62
FIR filters method, (3) Optimal or minimax design
Reason that FIR FIR filter is always stable because all its poles are
63
filter is always stable at the origin.
Gibb’s One possible way of finding an FIR filter that
64 Phenonmenon approximates H(ω) would be to truncate the
infinite Fourier series at n=±(N-1/2).
65 Desired frequency
4
π
response Hd(ω) hd(n)=1/2π∫ Hd(ω) ejωn dω
-π
(N-1)/2
Transfer function of
66 H(z)=z-(N-1)/2 [h(0)+∑ h(n)(zn +z-n)]
the realizable filter
n=0
1. Rectangular window
2. Hamming window
Types of windowing
67 3. Hanning window
techniques
4. Bartlett window
5. Kaiser window
Equation for WR(n) = 1 for 0 ≤ n ≤ M-1
68
Rectangular window = 0 otherwise
Equation for WHam(n)= 0.54-0.46 cos(2πn/M-1) for 0 ≤ n ≤ M-1
69
Hamming window =0 otherwise
WHan(n)= 0.5- 0.5 cos (2 πn /M-1 ) for 0 ≤ n ≤ M-
Equation for
70 1
Hanning window
=0 otherwise
Necessary and h(n) = h(N-1-n) for Symmetric
sufficient condition h(n) = - h(N-1-n) for Asymmetric
71
for linear phase FIR
Filter
What type of filter
72 frequency sampling Narrow Band Frequency Selective Filters
method is suitable
1. FIR Filter is always stable.
73 Merits of FIR filters 2. FIR Filter with exactly linear phase can easily
be designed.
Demerits of FIR 1. High Cost.
74
filters 2.Require more Memory
1. The central lobe frequency level should be
narrow.
Desirable
2. The highest side lobe frequency level should be
75 Characteristics of a
small.
window
3. The side lobe frequency magnitude decreases
with increasing as ω tends to π.
UNIT IV FINITE WORD LENGTH EFFECTS
Types of arithmetic Fixed point arithmetic, floating point, block
76
in digital systems floating point arithmetic
In fixed point number the position of a binary
Fixed point number
77 point is fixed. The bit to the right represent the
representation
fractional part and those to the left is integer part.
Types of fixed point sign magnitude, 1’s complement, 2’s complement
78
arithmetic
The leading binary digit is used to represent the
Sign magnitude
79 sign. If it is equal to 1 the number is negative,
representation
otherwise it is positive.
1’s Complement all the bits of the positive number
80
complement form
2’s Complement all the bits of the positive number
81
complement form and add 1 to the LSB
5
Advantages of 1. Large dynamic range
82 floating pint 2. Overflow is unlikely
representation
1. Input quantization errors 2. Coefficient
83 Quantization errors quantization errors
3.Product quantization errors
Input quantization The filter coefficients are computed to infinite
84
error precision in theory
Product quantization The product quantization errors arise at the output
85
error of the multiplier
Input quantization The input quantization errors arise due to A/D
86
error conversion
Quantization Truncation and Rounding
87
methods
A process of discarding all bits less significant
88 Truncation
than LSB that is retained
Rounding a number to b bits is accomplished by
89 Rounding choosing a rounded result as the b bit number
closest number being unrounded
In recursive system these nonlinearities often
cause periodic oscillation to occur in the output,
90 Limit cycles
even when input sequence is zero or some
nonzero value
The reduction of a continuous-time signal to
91 Sampling
a discrete-time signal
The average number of samples obtained in one
92 sampling rate
thus Fs = 1/T
A bandpass signal is sampled slower than
Under its Nyquist rate, the samples are indistinguishable
93
sampling from samples of a low-frequency alias of the high-
frequency signal.
Oversampling is used in most modern analog-to-
Over digital converters to reduce the distortion
94
sampling introduced by practical digital-to-analog
converters
95 Sampling Theorem Fs ≥2Fm
96 Nyquist rate Fs=2Fm
1. Zero limit cycle behavior
Types of limit cycle
97 2. Over flow limit cycle behavior
behavior of DSP
A high-level oscillation that can exist in an
otherwise stable filter due to the nonlinearity
98 Overflow limit cycle
associated with the overflow of internal filter
calculations
1. Saturation arithmetic
Methods to prevent
99 2. Scaling
overflow
100 Safe Scaling v ( n) = f( n) * x ( n )
6
UNIT V MULTIRATE AND DIGITAL SIGNAL PROCESSORS
Data communication require more than one
Multirate signal
101 sampling rate for processing data in such a cases
processing
increase and/or decrease the sampling rate.
Examples of multi- Decimator and interpolator
102
rate digital systems
Input- output Fy = Fx/D
103 relationship for a
decimator
Input- output Fy = IFx
104 relationship for an
interpolator
The original shape of the signal is lost due to
105 Aliasing
under sampling. This is called aliasing
106 Avoid Aliasing Placing a LPF before down sampling
How sampling rate Cascade connection of interpolator and decimator
107 be converted by a
factor I/D
It is an efficient coding technique by allocating
108 Sub-band coding lesser bits for high frequency signals and more
bits for low frequency signals.
109 Up sampling Increasing the sampling rate
110 Down sampling Decreasing the sampling rate
111 Decimator own sampling and a anti-aliasing filter
112 Interpolator An anti-imaging filters and Up sampling
Sampling rate Changing one sampling rate to other sampling rate
113
conversion is called sampling rate conversion
Quadrature Mirror Filters-Analysis section and
114 Sections of QMF
synthesis section
Cutoff frequency of Pi/M where M is the down sampling factor
115
Decimator
Cutoff frequency of π/L where L is the UP sampling factor.
116
Interpolator
Difference in Number of delayed multiplications are reduced
117 efficient transversal
structure
Shape of the white Flat frequency spectrum.
118
noise spectrum
transform coding that breaks a signal into a
number of different frequency bands, typically by
119 Sub-band coding
using a fast Fourier transform, and encodes each
one independently
Channel Vo-coder A bank of filters that breaks two incoming sound
120
sources into compatible frequency regions.
121 Random Processes A collection of Random Variables usually indexed
7
by time.
Stationary and Non-Stationary Random Processes,
Classification of Ergodic and Non-Ergodic Random Processes,
122
Random Processes Deterministic and In-deterministic Random
Processes.
A random process is called weak-sense stationary
Wide Sense
or wide-sense stationary (WSS) if its mean
123 Stationary Random
function and its correlation function do not change
Processes
by shifts in time.
Power Spectral It is a measure of how the average power of the
124
Density process is distributed with respect to frequency.
Examples of Random Gaussian Random Processes, Markov Random
125
Processes models Processes
PLACEMENT QUESTIONS
Digital signal processing improves the sensitivity
126 Define DSP
of a receiving unit.
A signal is a function that
127 Signals
conveys information about a phenomenon.
128 Systems It produces an output for a given input signal.
A signal of continuous amplitude and time is
continuous-time
129 known as a continuous-time signal or an analog
signal
signal
continuous-time The signals at input and output are continuous-
130
system time signal
Speech processing, Communication, Biomedical
131 Applications of DSP signal processing, Image processing, Radar signal
processing, Sonar signal processing etc.
1. The programs can be modified easily for
better performance.
132 Advances of DSP 2. Better accuracy can be achieved.
3. The digital signal can be easily stored and
transported.
A continuous signal that contains time-varying
133 Analog signal
quantities.
It is a signal that is being used to represent data as
134 Digital signal
a sequence of discrete values.
A discrete time signal x (n) is a function of an
135 Discrete time signal
independent variable that is an integer.
A discrete or an algorithm that performs some
136 Discrete time system
prescribed operation on a discrete time signal.
Methods to prevent 1. Saturation arithmetic
137
overflow 2. Scaling
The reduction of a continuous-time signal to
138 Sampling
a discrete-time signal
139 Up sampling Increasing the sampling rate
140 Down sampling Decreasing the sampling rate
141 Quantization Transforming a continuously valued input into a
representation that assumes one out of a finite set
8
of values
1. Zero limit cycle behavior
Types of limit cycle
142 2. Over flow limit cycle behavior
behavior of DSP
Truncation is a process of discarding all bits less
143 Truncation significant than LSB that is retained
Rounding a number to b bits is accomplished by
144 Rounding choosing a rounded result as the b bit number
closest number being unrounded.
To find N point DFT, have to add (N-L) zeros at
145 Zero padding
the sequence x(n).
The Fast Fourier Transform is an algorithm used
146 FFT
to compute the DFT.
When the desired magnitude response is piece-
wise constant over frequency, this compression
147 Pre-warping
can be compensated by introducing a suitable pre-
scaling, or pre-warping the critical frequencies
It is a graphical representation of the relationships
148 Signal flow graph between the variables of a set of linear difference
equations.
The region of convergence (ROC) of X(Z) the set
Region Of
149 of all values of Z for which X(Z) attain final
Convergence
value.
Linear filtering, Correlation
Applications of FFT
150 Spectrum analysis
algorithm
Faculty Team Prepared Signatures
Dr.T.R. Ganeshbabu,
1.
Prof/ECE
2. Dr.C.Selvi,ASP/ECE
Mr.M.Hariharan,AP/ECE
3.
HOD
