Aat 2

Subject : Digital signal processing

Name : G Shivaprasad
Roll no : 21951a04j9
Ece d

Question :

1. Discuss the condition for causality and stability ?

Causality:

time ttt depends only on the current and past inputs, and not on any future
Causality refers to the property of a system where the output at any given

inputs. Mathematically, for a system to be causal, the output y(t)y(t)y(t) at

time ttt should be determined by the input x(t′)x(t')x(t′) for t′≤tt' \leq tt′≤t.

y(t)=f[x(t)]y(t) = f[x(t)]y(t)=f[x(t)] is causal if and only if the function fff

Condition for Causality: A system described by a mathematical model

satisfies: y(t)=f[x(t′)]for all t′≤ty(t) = f[x(t')] \quad \text{for all } t' \leq
ty(t)=f[x(t′)]for all t′≤t

Implication:

 Causality ensures that the system responds to inputs in a physically

meaningful way. It aligns with the notion that a system cannot
anticipate future inputs to determine its current output.
Example:

 A physical system with a delay element (such as a simple resistor-

capacitor network) is typically causal because the output at any time
depends only on the past and current inputs.

Stability:

Stability refers to the property of a system where bounded inputs result in

bounded outputs. In other words, if the input to a stable system is bounded
(finite), then the output will also be bounded (finite), ensuring that the
system response does not grow uncontrollably over time.

Condition for Stability:

stable if, for any bounded input x(t)x(t)x(t), the output y(t)y(t)y(t)
 Bounded Input, Bounded Output (BIBO) Stability: A system is

remains bounded.

∣x(t)∣<Mx⇒∣y(t)∣<Myfor some Mx,My<∞|x(t)| < M_x \Rightarrow |y(t)| <

M_y \quad \text{for some } M_x, M_y < \infty∣x(t)∣<Mx ⇒∣y(t)∣<My for some
Mx ,My <∞

Implication:

 Stability ensures that the system behaves predictably and does not
exhibit unbounded oscillations or exponential growth in response to
input signals.

Example:

 An electric circuit with resistors, capacitors, and inductors, when

properly designed, tends to be stable. It dissipates energy over time
rather than accumulating it, which prevents the outputs from growing
indefinitely.

2 ) Define convolution sum and properties of convolution ?

Convolution is a fundamental operation in signal processing and systems analysis. It

involves combining two signals to produce a third signal that represents the
magnitude and duration of overlap between the original two signals. Here’s an
explanation of the convolution sum and its properties:

### Convolution Sum:

The **convolution sum** is defined for discrete-time signals \( x[n] \) and \( h[n] \) as
follows:

\[ y[n] = (x * h)[n] = \sum_{k=-\infty}^{\infty} x[k] \cdot h[n-k] \]

Where:
- \( y[n] \) is the output signal (convolution result) at time index \( n \).
- \( x[k] \) is the input signal at time index \( k \).
- \( h[n-k] \) is the impulse response of the system at time index \( n-k \).

### Properties of Convolution:

1. **Commutativity**: Convolution is commutative, which means the order of the

signals can be reversed without affecting the result:
\[ x[n] * h[n] = h[n] * x[n] \]

2. **Associativity**: Convolution is associative, meaning the grouping of convolutions

doesn't affect the final result:
\[ (x[n] * h[n]) * g[n] = x[n] * (h[n] * g[n]) \]

3. **Distributivity**: Convolution distributes over addition of signals:

\[ x[n] * (h[n] + g[n]) = x[n] * h[n] + x[n] * g[n] \]

4. **Shifting**: Convolution with a shifted signal is equivalent to shifting the

convolution result:
\[ x[n] * h[n-m] = (x[n] * h[n]) \text{ shifted by } m \]

5. **Linearity**: Convolution is linear, meaning it satisfies the principle of

superposition:
\[ a \cdot (x[n] * h[n]) + b \cdot (g[n] * h[n]) = a \cdot y_1[n] + b \cdot y_2[n] \]
where \( y_1[n] = x[n] * h[n] \) and \( y_2[n] = g[n] * h[n] \).
 6. **Zero Input Response**: Convolution of a signal \( x[n] \) with a delta function \( \
delta[n] \) gives the signal itself:
\[ x[n] * \delta[n] = x[n] \]

7. **Identity Element**: The delta function \( \delta[n] \) acts as the identity element
under convolution:
\[ x[n] * \delta[n] = \delta[n] * x[n] = x[n] \]

3) explain circular convolution using DFT ?

Understanding Circular Convolution:

o Let's denote two sequences x[n]x[n]x[n] and h[n]h[n]h[n], each

2. Finite-Length Sequences:

of length NNN. The circular convolution y[n]=x[n]∗h[n]y[n] =

x[n] * h[n]y[n]=x[n]∗h[n] of x[n]x[n]x[n] and h[n]h[n]h[n] can
be computed efficiently using the DFT.

o To apply the DFT for convolution, both sequences x[n]x[n]x[n]

3. Zero-Padding:

and h[n]h[n]h[n] are zero-padded to a length L=N+M−1L = N +

M - 1L=N+M−1, where MMM is the length of h[n]h[n]h[n]. This
ensures that the circular convolution result matches the linear
convolution result.

o Compute the DFTs X[k]X[k]X[k] and H[k]H[k]H[k] of the

4. DFT of Sequences:

sequences x[n]x[n]x[n] and h[n]h[n]h[n], respectively:

X[k]=∑n=0L−1x[n]⋅e−j2πLkn,k=0,1,…,L−1X[k] = \
sum_{n=0}^{L-1} x[n] \cdot e^{-j \frac{2 \pi}{L} kn}, \quad k =
0, 1, \ldots, L-1X[k]=∑n=0L−1 x[n]⋅e−jL2π kn,k=0,1,…,L−1
H[k]=∑n=0L−1h[n]⋅e−j2πLkn,k=0,1,…,L−1H[k] = \
sum_{n=0}^{L-1} h[n] \cdot e^{-j \frac{2 \pi}{L} kn}, \quad k =
0, 1, \ldots, L-1H[k]=∑n=0L−1 h[n]⋅e−jL2π kn,k=0,1,…,L−1

o Multiply the DFTs X[k]X[k]X[k] and H[k]H[k]H[k] element-wise:

5. Element-wise Multiplication:

Y[k]=X[k]⋅H[k],k=0,1,…,L−1Y[k] = X[k] \cdot H[k], \quad k = 0,

1, \ldots, L-1Y[k]=X[k]⋅H[k],k=0,1,…,L−1

o Compute the inverse DFT of Y[k]Y[k]Y[k] to obtain the circular

6. Inverse DFT:

convolution y[n]y[n]y[n]: y[n]=1L∑k=0L−1Y[k]⋅ej2πLkn,n=0,1,

 …,L−1y[n] = \frac{1}{L} \sum_{k=0}^{L-1} Y[k] \cdot e^{j \
frac{2 \pi}{L} kn}, \quad n = 0, 1, \ldots, L-1y[n]=L1 ∑k=0L−1
Y[k]⋅ejL2π kn,n=0,1,…,L−1

o The result y[n]y[n]y[n] obtained from the inverse DFT is

7. Circular Nature:

correct result y[n]=x[n]∗h[n]y[n] = x[n] * h[n]y[n]=x[n]∗h[n]

circularly shifted due to the circular nature of DFT. To get the

(linear convolution), you may need to discard M−1M-1M−1

samples from the beginning of y[n]y[n]y[n], where MMM is the
length of h[n]h[n]h[n]
o
4) Write the properties of twiddle factor ?

The twiddle factor is a complex exponential term used in the context of

Discrete Fourier Transform (DFT) and Fast Fourier Transform (FFT)
algorithms. It plays a crucial role in defining the frequency components and
phase shifts involved in transforming signals between the time domain and
frequency domain. Here are the key properties of the twiddle factor:

8. Definition: The twiddle factor WNknW_N^{kn}WNkn is defined as:

WNkn=e−j2πNknW_N^{kn} = e^{-j \frac{2 \pi}{N} kn}WNkn =e−jN2π
kn where NNN is the size of the DFT/FFT, kkk is the frequency index,
and nnn is the time index.

WNkn=WN(k+mN)nW_N^{kn} = W_N^{(k + mN)n}WNkn

9. Periodicity: The twiddle factor exhibits periodicity:

=WN(k+mN)n for any integer mmm. This periodicity arises from the
periodic nature of the complex exponential e−j2πNkne^{-j \frac{2 \pi}
{N} kn}e−jN2π kn.
Unity Property: When k=0k = 0k=0, the twiddle factor
simplifies to: WN0n=e0=1W_N^{0n} = e^0 = 1WN0n =e0=1 This
10.

property signifies that WN0nW_N^{0n}WN0n is always equal to 1 for

any nnn, implying no phase shift.

indices: WN−kn=WNkn‾=ej2πNknW_N^{-kn} = \overline{W_N^{kn}}

11. Symmetry: The twiddle factor exhibits symmetry for conjugate

= e^{j \frac{2 \pi}{N} kn}WN−kn =WNkn =ejN2π kn This symmetry is

crucial in the computation of inverse DFT and handling complex
conjugate terms in DFT/FFT operations.
 WNkn⋅WNln=WN(k+l)nW_N^{kn} \cdot W_N^{ln} =
12. Multiplicative Property: When twiddle factors are multiplied:

W_N^{(k+l)n}WNkn ⋅WNln =WN(k+l)n This property simplifies the

multiplication of twiddle factors and directly relates to how frequency
components are combined in DFT/FFT computations.

WNknW_N^{kn}WNkn corresponds to the kkk-th root of unity scaled

13. Relation to Roots of Unity: The twiddle factor

by nnn. Roots of unity are complex numbers that satisfy zN=1z^N =

1zN=1, and they are uniformly spaced around the unit circle in the
complex plane.
14. Implementation in FFT: In FFT algorithms, the twiddle factors
are precomputed and stored, which significantly speeds up the
computation of the DFT by reducing redundant complex exponential
calculations.

5) What is decimation by factor D ?

Decimation by a factor DDD refers to a downsampling operation in digital

of DDD. This operation is commonly used in various applications such as

signal processing where the sampling rate of a signal is reduced by a factor

signal filtering, data compression, and reducing computational complexity.

Here's how decimation by a factor DDD works:

Decimation Process:

Input Signal: Start with a discrete-time signal x[n]x[n]x[n],

where nnn represents the discrete time index.
15.

Downsampling: Decimate x[n]x[n]x[n] by a factor DDD to

obtain the decimated signal y[n]y[n]y[n]: y[n]=x[Dn]y[n] =
16.

x[Dn]y[n]=x[Dn]
o y[n]y[n]y[n] is obtained by taking every DDD-th sample from
x[n]x[n]x[n].
o This operation reduces the sampling rate of x[n]x[n]x[n] by DDD.

x[n]x[n]x[n] is fsf_sfs samples per second (or per unit time). After
17. Reduction in Sampling Rate: The original sampling rate of

decimation by a factor DDD, the sampling rate of y[n]y[n]y[n]

becomes fsD\frac{f_s}{D}Dfs .
Example:

 Suppose x[n]x[n]x[n] is sampled at fs=1000f_s = 1000fs =1000 Hz

Decimation by a factor D=2D = 2D=2 would mean taking every

(1000 samples per second).

second sample: y[n]=x[2n]y[n] = x[2n]y[n]=x[2n]



After decimation, y[n]y[n]y[n] would have a sampling rate of

10002=500\frac{1000}{2} = 50021000 =500 Hz.


Application:

 Signal Compression: Decimation reduces the number of samples in a

signal, reducing storage or transmission requirements.
 Anti-Aliasing Filtering: Often used in conjunction with a low-pass
filter to remove high-frequency components before downsampling to
prevent aliasing.
 Computational Efficiency: Reduces the number of computations
required for further processing, such as in digital signal processing
algorithms.

Considerations:

 Anti-Aliasing Filtering: Before decimation, it's essential to apply an

anti-aliasing filter to remove high-frequency components that could
alias into the signal after downsampling.
 Nyquist-Shannon Sampling Theorem: Decimation should respect
the Nyquist criterion to avoid aliasing. The new sampling rate after
decimation should be sufficient to capture the desired signal
bandwidth.

6)Draw the frequency response of N point Bartlett window ?

To draw the frequency response of an NNN-point Bartlett (triangular)

window, we first need to understand its mathematical form. The NNN-point
Bartlett window w[n]w[n]w[n] is defined as:
w[n]={2nN−1for 0≤n≤N−122−2nN−1for N−12<n≤N−1w[n] = \
begin{cases} \frac{2n}{N-1} & \text{for } 0 \leq n \leq \frac{N-1}{2} \\ 2 - \
frac{2n}{N-1} & \text{for } \frac{N-1}{2} < n \leq N-1 \
end{cases}w[n]={N−12n 2−N−12n for 0≤n≤2N−1 for 2N−1 <n≤N−1

This window function is symmetric and tapers from 0 at n=0n = 0n=0 and
n=N−1n = N-1n=N−1 to 1 at n=N−12n = \frac{N-1}{2}n=2N−1 .

Frequency Response:

The frequency response H(ejω)H(e^{j\omega})H(ejω) of the Bartlett window

is given by the Discrete-Time Fourier Transform (DTFT) of w[n]w[n]w[n]:

H(ejω)=∑n=0N−1w[n]e−jωnH(e^{j\omega}) = \sum_{n=0}^{N-1} w[n]

e^{-j\omega n}H(ejω)=∑n=0N−1 w[n]e−jωn

frequency response ∣H(ejω)∣|H(e^{j\omega})|∣H(ejω)∣.

However, for visualization purposes, we often plot the magnitude of the

Steps to Plot the Frequency Response:

Compute the Window Function: Calculate w[n]w[n]w[n] for

NNN points using the given formula.
18.

Evaluate H(ejω)H(e^{j\omega})H(ejω) for various frequencies ω\

19. Compute the Discrete-Time Fourier Transform (DTFT):

omegaω.
Plot the Magnitude Response: Plot ∣H(ejω)∣|H(e^{j\omega})|
∣H(ejω)∣ versus ω\omegaω (normalized frequency).
20.

7) What are the categories of multirate ?

multirate signal processing involves manipulating signals at different

sampling rates. There are several categories or techniques within multirate
signal processing that are commonly used:

integer factor DDD, which reduces the sampling rate of the signal.
21. Decimation: Decimation involves downsampling a signal by an
 Decimation is typically followed by filtering to remove the aliasing
introduced by the downsampling process.

rate of a signal by an integer factor III. It inserts I−1I-1I−1 zeros

22. Interpolation: Interpolation involves increasing the sampling

between each sample of the original signal and then applies a low-pass
filter to reconstruct the signal at the higher sampling rate.

of a signal by a non-integer factor RRR. It combines decimation and

23. Resampling: Resampling refers to changing the sampling rate

interpolation techniques to achieve the desired sampling rate

conversion.
24. Polyphase Decomposition: Polyphase decomposition is a
technique used to implement multirate filters efficiently. It involves
decomposing a filter into multiple phases (branches), each operating at
a different rate, to reduce computational complexity.
25. Filter Banks: Filter banks are systems of filters used to split a
signal into multiple subbands or channels, each containing a different
range of frequencies. They are often used in applications such as audio
and image compression, where different frequency components are
processed separately.
26. Wavelet Transform: The wavelet transform is a powerful tool in
multirate signal processing that decomposes a signal into components
at different scales. It allows for both time and frequency localization of
signal

8) Why decimate ?

Decimation, in the context of digital signal processing, is the process of reducing the
sampling rate of a signal. This operation serves several important purposes and is applied in
various practical scenarios:

Reasons for Decimation:

27. Reducing Data Rate: Decimation reduces the number of samples in a signal, thereby
reducing the data rate required for storage, transmission, or further processing. This is
particularly useful in applications where bandwidth or storage capacity is limited.
 28. Improving Efficiency: By reducing the sampling rate, decimation can make
subsequent signal processing operations more efficient. Algorithms operating on
fewer samples can be computationally faster and require less memory.
29. Anti-Aliasing Filtering: Decimation is often used in conjunction with anti-aliasing
filtering. Before decimating a signal, a low-pass filter is applied to remove high-
frequency components that could alias into the signal after downsampling. This
ensures that the downsampling operation does not introduce unwanted artifacts or
distortion due to aliasing.
30. Signal Compression: Decimation can be part of signal compression techniques. By
reducing the sampling rate, the signal's information content is reduced, leading to
potential savings in storage space or transmission bandwidth.
31. Matching Sampling Rates: In systems where signals of different sampling rates need
to be processed or combined, decimation can be used to bring all signals to a
common sampling rate. This simplifies subsequent processing and avoids
synchronization issues.
32. Noise Reduction: Decimation can improve the signal-to-noise ratio (SNR) of a signal in
certain cases. If the noise power is spread across a wider bandwidth than the signal of
interest, reducing the bandwidth through decimation can effectively reduce the noise
power within the signal band.

Practical Applications:

 Telecommunications: Decimation is commonly used in digital communication

systems to reduce the bandwidth requirements and to match the sampling rates of
different parts of the system (transmitters, receivers, etc.).
 Audio and Video Processing: In digital audio and video processing, decimation is
used for downsampling signals before storage or transmission to reduce file sizes
without significant loss of quality.
 Medical Signal Processing: Decimation can be used in biomedical signal processing
to reduce the amount of data that needs to be analyzed or transmitted, while still
retaining critical information.
 Sensor Networks: In sensor networks, where multiple sensors may produce data at
different rates, decimation can be used to align data streams for synchronization and
processing.
In summary, decimation is a fundamental operation in digital signal processing that enables
efficient data handling, reduces computational complexity, and ensures effective signal
representation for various applications across different fields.

9) what is trunctuation ?

(a) Truncation and Rounding:

Truncation and rounding are methods used in digital signal processing (DSP)
and numerical computation to convert real numbers (often obtained from
analog-to-digital conversion or mathematical operations) into a fixed-point
representation suitable for further processing or storage in digital systems.

Truncation:

 Definition: Truncation involves discarding the digits beyond a certain

point in a number's representation without rounding off.
 Usage: In DSP, truncation is often used to simplify computations or
reduce storage requirements by keeping only a certain number of
significant digits after the decimal point.
 Example: If a number 3.14159 is truncated to two decimal places, it
becomes 3.14.

Rounding:

 Definition: Rounding involves adjusting a number to the nearest value

that can be represented in a specified number of decimal places or
significant figures.
 Usage: Rounding is crucial in applications where precision is
important, such as financial calculations or scientific computations.
 Example: Rounding 3.14159 to two decimal places gives 3.14 (if
rounding down) or 3.15 (if rounding up).

Significance:

 Both truncation and rounding can introduce errors or inaccuracies in

computations, especially if not carefully managed.
 In DSP, these operations are critical when converting analog signals to
digital form (quantization) or when processing digital signals with
limited precision to avoid cumulative errors.
10) Find the linear convolution of the sequences x(n) and h(n) using DFT.
x(n) = 1, 0, 2, h(n) = 1, 1 ?

To find the linear convolution of sequences \( x(n) \) and \( h(n) \) using the Discrete
Fourier Transform (DFT), we can follow these steps:

Given sequences:
\[ x(n) = [1, 0, 2] \]
\[ h(n) = [1, 1] \]

### Step-by-Step Solution:

1. **Zero-padding**: Pad both sequences \( x(n) \) and \( h(n) \) to the length \( L = N +

M - 1 \), where \( N \) is the length of \( x(n) \) and \( M \) is the length of \( h(n) \).
- \( x(n) \) has length \( N = 3 \)
- \( h(n) \) has length \( M = 2 \)
- \( L = N + M - 1 = 3 + 2 - 1 = 4 \)

So, pad \( x(n) \) and \( h(n) \) with zeros to make them of length 4:
\[ x(n) = [1, 0, 2, 0] \]
\[ h(n) = [1, 1, 0, 0] \]

2. **Compute DFT**: Compute the Discrete Fourier Transform (DFT) of \( x(n) \) and \
( h(n) \).
\[ X(k) = \text{DFT}\{x(n)\} \]
\[ H(k) = \text{DFT}\{h(n)\} \]

Using the DFT formulas:

\[ X(k) = \sum_{n=0}^{3} x(n) e^{-j \frac{2 \pi}{4} kn} \]
\[ H(k) = \sum_{n=0}^{3} h(n) e^{-j \frac{2 \pi}{4} kn} \]

Calculate \( X(k) \) and \( H(k) \):

\[ X(k) = [3, 2+2j, -1, 2-2j] \]
\[ H(k) = [2, 2, 0, 0] \]
3. **Multiply in Frequency Domain**: Multiply the DFTs \( X(k) \) and \( H(k) \) element-
wise:
\[ Y(k) = X(k) \cdot H(k) \]
\[ Y(k) = [6, (2+2j) \cdot 2, 0, 0] \]
\[ Y(k) = [6, 4+4j, 0, 0] \]

4. **Inverse DFT**: Compute the inverse DFT of \( Y(k) \) to obtain the linear
convolution \( y(n) \):
\[ y(n) = \frac{1}{4} \sum_{k=0}^{3} Y(k) e^{j \frac{2 \pi}{4} kn} \]

Calculate \( y(n) \):

\[ y(n) = [1.5, 3+3j, 0, 0] \]

Here, \( y(n) \) represents the linear convolution of \( x(n) \) and \( h(n) \) calculated
using the DFT method.

### Summary:

The linear convolution \( y(n) \) of sequences \( x(n) = [1, 0, 2] \) and \( h(n) = [1, 1] \)
using the Discrete Fourier Transform (DFT) method results in:
\[ y(n) = [1.5, 3+3j, 0, 0] \]

Note: The result \( y(n) \) is typically truncated to the original length \( N + M - 1 = 4 \),
but in this case, the last two values are zero due to the length of \( h(n) \).