MODULE–IV

SAMPLING
Introduction
The "sampling theorem in signals and systems" is one of the most fundamental concepts in
digital signal processing that enables the conversion of continuous analog signals to discrete-
time digital signals. It establishes the theoretical foundation for the analog-to-digital
conversion process which is very important for applications like sound recording, video
processing, wireless communication, and more.

Need of Sampling

Sampling is an essential element in the world of electronics, which is why sampling is needed
in various steps. Sampling is needed for the conversion of continuous-time signals in areas
where the input signals are discrete-time signals. Most of the Digital Signal Processing
methods use sampling for the same. Also, the continuous-time signals are more challenging
in terms of transmission. Hence sampling comes into play to convert them into discrete time
signals.
Sampling is also needed in noise reduction techniques in [Link] addition to that, the
conversion of analog signals ad digital signals increases the capability between systems of
that kind. It also helps in interfacing as well.

Applications of Sampling Theorem

The sampling theorem finds numerous applications in digital signal processing and
communication systems.
Some key areas where it is utilized include:
o Sound recording and playback: Microphones sample soundwaves at frequencies
greater than 40 kHz as per the Nyquist criterion.
o Digital images: Cameras sample visual light waves into pixels at frequencies
satisfying the sampling rates as per dimensions and resolution of images.
o Wireless communication: Downconversion of RF signals to baseband for digital
processing and vice versa is based on sampling theory principles.
A signal may be distinctly determined by a succession of equidistant samples. In the provided
example (Figure 1 below), we provide three distinct continuous-time signals. Notably, these
signals exhibit equivalent values at integer multiples of T.

It is evident that a limitless quantity of signals has the capability to produce a specific
collection of samples. In the subsequent analysis, it will be demonstrated that when a signal is
band limited, and samples are sufficiently proximate with maximum frequency component of
the signal, then it enables a flawless reconstruction. The sampling theorem, sometimes
referred to as this outcome, has significant significance in the actual implementation of signal
analysis techniques.

Figure 1 illustrates three CT signals that possess same samples at integer multiples of the
period T.

One effective approach to achieve sampling involves using a periodic impulse train is
multiplied by x(t) that is intended to be sampled. The displayed process, may be seen in
Figure. Additionally, the fundamental frequency of p(t), denoted as Ws = 27π/T, is sometimes
denoted to as the sampling frequency. When considering the temporal dimension,

where,
It is well-established that the multiplication of x(t) with impulse of magnitude unity allows for
the sampling. Utilising this concept in the context of the equation. In equation (7.1), the
function Xp(t) represents an impulse train characterised by its amplitudes.

Figure 2: Impulse-train sampling

Based on the multiplication property, it is known that

Given that convolution with an impulse results in a signal shift, it may be inferred that
The above figure illustrates the impact of time domain sampling on the frequency domain.
Subfigure (a) & (b) displays the signal & pulse train in frequency domain
Figure 3 (c) & (d) displays the sampled signal in frequency domain with a sampling frequency
(Ws) more than twice the maximum frequency component (wM) and lower than twice the
maximum frequency component (wM).

It is a fundamental concept in signal processing that defines the number of samples to

represent continuous-time signal.

The sampling theorem, often elucidated via impulse-train sampling, asserts the fundamental
principle that a signal with finite bandwidth may be accurately represented by its discrete
samples. In practical applications, the generation and transmission of narrow, large-amplitude
pulses that closely resemble impulses may be challenging. So, the sampled signal is
constructed using a technique known as a ‘zero-order hold’. The system in question performs
as seen in the accompanying Figure.

This necessitates that

Figure 4: Sampling using a zero-order hold technique.

Once again, the frequency response in question is seen in practical applications. The value of
7.8 cannot be precisely achieved, necessitating the development of a suitable approximation.
Indeed, in several scenarios, the output of the zero-order hold is regarded as a satisfactory
approximation of the original signal in isolation, without the need for additional lowpass
filtering. Essentially, it serves as a potential, if rudimentary, interpolation between the sampled
values. In some instances, it may be desirable to execute a more seamless interpolation
between sample values.
 Figure 5 : The figure illustrates the concept of zero-order hold as a two-step process

Interpolation is a widely used technique in which a continuous signal is fitted to a collection of

sample values. This approach is often utilised to rebuild a function, either with an
approximation or an exact representation, based on the available samples. A commonly used
interpolation technique is the zero-order hold method. Linear interpolation is a valuable
method of interpolation in which consecutive sample locations are linked by a straight line.

In the case of a band-limited signal, it is possible to achieve precise signal reconstruction. This
entails using a lowpass filter to provide accurate interpolation between the sampled points.
The process of interpolation becomes apparent when examining the impact of the lowpass
filter seen in Figure 6 on the reconstruction of x(t). Specifically, the resulting output is,

The equation shown above illustrates the methodology for constructing a smooth
curve that passes over the discrete sample points x(nT). As a result, it may be regarded as a
formula for interpolation. Here is the revised version: Regarding the ideal lowpass filter,
denoted as H(jw),

In order to
Subsequently, it denotes the primary band-limited signal x(t), alongside x p(t), which refers to
the impulse train composed of samples. The illustration also demonstrates the superposition
of the separate phrases. The process of interpolation, which involves using the impulse
response of an ideal lowpass filter, is usually known as band-limited interpolation. This method
achieves precise reconstruction. In several instances, there is a preference for using a filter or
interpolating function that is less precise but simpler in nature.
A deeper understanding of the correlation between x(t) and Xr(t) under the condition Ws <
2wM may be obtained by examining the specific scenario of a sinusoidal signal. Hence,
Figure 9: The Impact of Aliasing on a Sinusoidal Signal. For each of the four values of wo. The
figure depicts the sinusoidal signal in its original form (represented by a solid curve), together
with its corresponding samples and the reconstructed signal (represented by a dashed curve).
The four scenarios shown are as follows: (a) when the frequency w0 is equal to one-sixth of
the sampling frequency ws, (b) when w0 is equal to two-sixths of ws, (c) when w0 is equal to
four-sixths of ws, and (d) when w0 is equal to five-sixths of ws. Aliasing does not occur in
scenarios (a) and (b), however it is present in scenarios (c) and (d).In this particular scenario,
the Fourier
Given that the Fourier transform of δ (t- nT) equals e-jwnT, it can be shown that.

Now let us investigate the discrete-time Fourier transform of the sequence xd[n].

Figure 12 Overall framework for the filtration of a continuous-time signal via the use of a
discrete-time filter.