1

CHAPTER-1
INTRODUCTION
1.1 INTRODUCTION

Digital signal processing (DSP) is one of the fast-growing technologies that will shape
Science and Engineering in the twenty-first century. Signal processing refers to the
science of creating, interpreting, modifying, manipulating, analyzing, and displaying
signal information. DSP made a revolutionary change in a broad range of fields like
medical imaging, wireless communications, radar and sonar, multimedia indexing
storage, hearing aid applications, etc. The marvelous performance effect is one of the
key reasons that DSP has become more famous. In signal processing, filters are
widely used and considered to be the first hands-on tool in any electronic circuit. The
primary aim of the filter is to allow the desired signal and remove the unwanted signal
based on a particular frequency. Filters are useful in performing various functions like
frequency selective digital filtering [1], anti-aliasing, signal preconditioning,
decimation/interpolation, band selection, Low-Pass Filtering (LPF), matched filtering,
spectral shaping, channel equalization, video convolution functions, and interference
cancellation [2]. There are two types of filters namely analog filters and digital filters.
Digital filters are more vantage than analog filters because of their stability in nature.

The exceptional performance of digital filters, which play a large role, is one
of the primary factors contributing to DSP's growing popularity. Digital filters
combine the benefits of high-speed digital circuitry with mathematical algorithms to
process a series of digital data collected over time. Signal separation and restoration
are the two main purposes of digital filters. When a signal is impacted by noise,
interference, or other signals, it has to be separated, and restoration is needed to
recreate the original signal. The same function is served by analog filters, which
operate on straight analog inputs. While analog filters are quicker, less expensive, and
have a wider range of amplitude and frequency, the performance of digital filters is
higher even though they are created using analog components like resistors, inductors,
and capacitors.

The benefit of a digital filter is simple to filter order calibration and filter
coefficient change to make them adaptable or user adjustable. The processing is more
 2

difficult even if similar approaches are present in analog filters. Due to analog to
digital and digital-to-analog conversions, digital filters have greater costs and longer
delays. Yet, compared to analog filters, many useful designs may be simply realized
using digital filters. Digital filters are used in many kinds of devices and are now a
required component of common electronics like radios, mobile phones, and AV
receivers.

1.2 FILTER

A Filter is a circuit that is capable of selectively filtering one frequency or a range of

frequencies out of a mix of different frequencies. Filters are mainly classified into two
categories.

 Analog filter
 Digital filter

Analog filter consists of resistors, capacitors, and inductors and operates on a

continuous time-varying signal. There are several analog filters like RC filter, LC
filter and RLC filter, and so on. These filters since they operate on continuous time
signals they don't need ADC (Analog-to-Digital Converter) and DAC (Digital-to-
Analog Converters). But analog filters are less flexible; also the components present
in the filters are more sensitive to environmental changes.

The digital filter is a system that can perform mathematical operations on

sampled, discrete-time signals to reduce or enhance certain aspects of that signal. It is
most widely used in signal processing [3]. Digital filters may be more expensive than
an equivalent analog filter due to their increased complexity, but the practical designs
can be implemented in digital filters that are not possible in analog filters. Digital
filters can be of a very high order. Due to these advantages, digital filters are still
widely used.

Digital filters are mainly classified into two types based on impulse responses.
They are

 Infinite Impulse Response (IIR) Filter

 Finite Impulse Response (FIR) Filter
 3

1.3 INFINITE IMPULSE RESPONSE (IIR) FILTER

IIR filters are digital filter that tends to give infinite response to the impulse input
signal which is high at a particular instant of time and low for the remaining time. The
construction of the IIR filter includes a feedback loop.

The IIR filter output is determined by present inputs and past input and output
values. IIR filters have better frequency responses. The phase characteristics of the
IIR filter are not linear which can cause a problem to the system which needs phase
linearity. When phase is essence IIR filters are not preferred.

The transfer function of the IIR Filter is given as

( )= (1.1)
ℎ( )

Where H(z) is the Transfer function and h(n) is the impulse response which tends to
infinity.

1.4 FINITE IMPULSE RESPONSE (FIR) FILTER

An FIR filter is a filter whose impulse response (or response to any finite length
input) is of finite duration because it settles to zero in finite time. One of the
fundamental components of digital signal processing is the FIR filter. In contrast, IIR
filters have the potential to have internal feedback and have an unlimited response
time.

Due to finite response, FIR filters are more stable. These filters have a wide
range of applications, including equalizers, Telecommunications, image processing,
and several related fields. The FIR filter output depends only on the present input and
there is no feedback as in the IIR filter.

The transfer function of the FIR filter is given as

(1.2)
( )= ℎ( )

Where H(z) is the Transfer function and h(n) is impulse response and N is the
number of taps of the FIR filter. The block diagram of the 4-tap FIR filter is shown in
 4

Figure 1.1.The coefficients of the FIR filter are d[0], d[1], d[2] and d[3]. X[n] is the
input sequence. The sequence gets delayed for each clock cycle and enters the
multiplier and gets multiplied with the appropriate coefficients and the output of each
multiplier is added resulting in the output.

The output for 4-tap FIR Filter with coefficients d[0], d[1], d[2] and d[3] is given as

Y[n]=d[0]*X[n]+d[1]*X[n-1]+d[2]*X[n-2]+d[3]*x[n-3] (1.3)

X [n ]  ri
Delay Delay Delay

d[0] d[1] d[2] d[3]

Y [n]  ri'

Figure 1.1 Basic Block diagram of FIR Filter

Digital FIR Filter is a widely used digital signal processing technique. It is a

type of digital filter whose impulse response has a finite duration. An FIR filter can be
used for various applications such as smoothing, signal separation, noise reduction,
and digital signal processing in general. An FIR filter is a type of digital filter whose
output depends only on the current and past input samples. This means that the output
of an FIR filter at a given time depends on a finite number of input samples. The
impulse response of an FIR filter is a finite sequence of values. The length of the
impulse response is determined by the filter order. The filter order is the number of
delay elements used in the filter structure [4].

For N-tap filter the output is given as

(1.4)
[ ]= [ ] [ − ]

where Y[n] =output sequence, N= order of FIR Filter, [0], [1], [2] … … …
[N]are the FIR filter coefficients and X[n]= Input sequence.
 5

1.5 DESCRIPTION OF FIR FILTER

The basic structure of an FIR filter consists of a delay line and a set of coefficients,
which are multiplied with the input signal samples and then summed to produce the
output samples. The number of coefficients used in the filter determines the filter
order, and higher order filters can achieve more complex frequency response
characteristics

The frequency response of an FIR filter can be designed using various

methods such as windowing, frequency sampling, or optimal design techniques.
Windowing involves multiplying the impulse response of an ideal filter with a
window function to obtain the desired frequency response. Frequency sampling
involves specifying the frequency response at a finite number of frequency points and
using the inverse discrete Fourier transform to obtain the impulse response. Optimal
design techniques involve minimizing the error between the desired and actual
frequency response using mathematical optimization algorithms

Overall, the basic FIR filter is a fundamental building block in digital signal
processing, widely used in a variety of applications such as audio processing, image
processing, and communication systems.

1.6 DESIGN OF FIR FILTER

The design of a digital FIR filter involves determining the filter coefficients that
define the desired frequency response. FIR filters are widely used in signal processing
and have several desirable properties, such as linear phase response and stability. An
overview of the design process for a digital FIR filter involves

 Filter Specification: Beginning with specifications of the desired filter

characteristics, such as the filter type (low-pass, high-pass, band pass, etc.), cutoff
frequency, transition bandwidth, stop band attenuation, and any other relevant
parameters. These specifications will guide the design process.
 Filter Length: Determining the required length of the filter, which influences the
frequency resolution and stop band attenuation. Longer filters provide better
frequency resolution but require more computational resources.
 6

 Filter Windowing: Choosing a windowing function to shape the impulse response

of the filter. Common window functions include Hamming, Hanning, Blackman,
and Kaiser. The choice of the window function affects the trade-off between main
lobe width and side lobe attenuation.
 Filter Design Method: Selecting an appropriate design method to compute the
filter coefficients. Some popular methods include:
a) Frequency Sampling: In this method, you directly specify the desired
frequency response at specific frequencies and interpolate between them. It's
simple and intuitive but may result in an inefficient filter length.
b) Window Method: Here, you design an ideal filter (typically, a rectangular
frequency response) and then apply a window function to achieve the desired
characteristics. This method allows more control over the trade-off between
main lobe width and side lobe attenuation.
c) Optimal Methods: Optimal design methods, such as the Parks-McClellan
algorithm (Remez algorithm) or least squares approximation, aim to minimize
the error between the desired and actual frequency responses. These methods
provide better control over filter characteristics but are more computationally
intensive.
 Filter Coefficient Calculation: Once the design method was chosen, the filter
coefficients based on the chosen method was computed. The specific algorithm or
equations used depend on the method selected. This step yields a set of
coefficients that determine the filter's frequency response.
 Filter Implementation: By using the computed coefficients, the FIR filter was
implemented. This was done by using MATLAB, and the filter was simulated
using Verilog code in Vivado tool for signal processing.
 Filter Evaluation: The performance of the designed filter by evaluated by
analyzing its frequency response, phase response, group delay, and other relevant
parameters.

1.7 OPERATION OF FIR FILTER

Consider a FIR Filter with coefficients as

d[0]= 0010, d[1]= 0100, d[3]=0110, d[4]=1000 (in binary)

 7

The coefficients are given as input to the Multipliers and the output of the
multipliers are added and output as Y[n].

Consider the input sequence or word to be X[n] = 1010 (in binary) and the
output of the FIR Filter is

Y[n] = 0010* X[n] +0100 * X[n-1] + 0110 * X[n-2] + 1000 * x[n-3]

Where X[n-1], X[n-2],X[n-3] are delayed input sequence

First clock cycle:

X[n]= 1010, X[n-1] = 0000, X[n-2] = 0000, X[n-3] =0000

Output: Y[n] = 1010*0010 +0000*0100 + 0000*0110 + 0000*1000

= 10100

Second clock cycle:

X[n]= 1010, X[n-1] = 1010, X[n-2] = 0000, X[n-3] =0000

Output: Y[n] = 1010*0010 +1010*0100 + 0000*0110 + 0000*1000

= 10100 + 101000 = 111100

Third clock cycle:

X[n] = 1010, X[n-1] = 1010, X[n-2] = 1010, X[n-3] =0000

Output: Y[n] = 1010*0010 +1010*0100 + 1010*0110 + 0000*1000

= 10100 + 101000 + 111100 = 1111000

Fourth clock cycle:

X[n] = 1010, X[n-1] = 1010, X[n-2] = 1010, X[n-3] = 1010

Output: Y[n] = 1010*0010 + 1010*0100 + 1010*0110 + 1010*1000

= 10100 + 101000 + 111100 + 1010000 = 11001000

1.8 CHARACTERISTICS OF FIR FILTER

A Finite Impulse Response (FIR) filter is a type of digital filter that has several key
characteristics.
 8

Finite impulse response: An FIR filter has a finite impulse response, which means
that its output depends only on a finite number of past input samples. The filter
response decays to zero after a finite number of samples.

Linear phase response: One significant characteristic of FIR filters is that they can
have a linear phase response. This means that the phase shift introduced by the filter is
directly proportional to the frequency of the input signal. Linear phase is desirable in
applications where preserving the relative timing of different frequency components is
important, such as in audio and video processing.

Stability: FIR filters are inherently stable. Unlike IIR filters, FIR filters do not have
feedback loops, which eliminates the possibility of unstable behavior or oscillations.

Design flexibility: FIR filters offer a high degree of design flexibility. Design
parameters such as filter order, cutoff frequency, and desired frequency response can
be adjusted to meet specific filtering requirements. Various design methods, such as
windowing, frequency sampling, and least squares, can be used to design FIR filters
with different characteristics.

Arbitrary frequency response: FIR filters can achieve any desired frequency
response within their operating range. By carefully designing the filter coefficients, it
is possible to shape the frequency response to meet specific filtering requirements,
such as low-pass, high-pass, band-pass, or notch filtering.

Impulse response length: The length of the filter's impulse response directly affects
its frequency response. Longer filter lengths allow for more precise frequency shaping
but require more computational resources. The choice of filter length involves a trade-
off between frequency response accuracy and computational complexity.

1.9 ADVANTAGES OF DIGITAL FILTER

Digital filters offer several advantages compared to their analog counterparts. Here
are some of the key advantages of digital filters:

Flexibility: Digital filters provide greater flexibility in terms of design and

implementation. They can be easily modified and adjusted by changing the filter
 9

coefficients or parameters, allowing for fine-tuning and optimization. In contrast,

analog filters often require physical modifications or adjustments.

Accuracy: Digital filters can achieve high levels of accuracy and precision in signal
processing. They can provide precise control over filter characteristics such as cutoff
frequencies, stop-band attenuation, and pass-band ripple. With sufficient
computational power, digital filters can achieve better accuracy than analog filters.

Reproducibility: Digital filters offer reproducibility, meaning that the filter response
can be precisely replicated across different systems and platforms. Once the filter
coefficients are determined, they can be easily shared and implemented on various
devices or software applications without any loss of performance.

Stability: Digital filters are inherently stable and not susceptible to variations caused
by temperature, aging, or component tolerances, which can affect analog filters. This
stability ensures consistent and reliable performance over time.

Signal Processing Techniques: Digital filters allow for the use of advanced signal
processing techniques, such as adaptive filtering, nonlinear filtering, and multirate
filtering. These techniques can enhance the filtering capabilities and enable more
sophisticated processing of signals, such as noise reduction, echo cancellation, and
signal enhancement.

Implementation Efficiency: Digital filters can be implemented using efficient

algorithms and computational techniques, such as fast Fourier transforms (FFT) or IIR
structures. These algorithms can provide efficient real-time processing and reduce
computational complexity, making digital filters suitable for various applications,
including real-time audio, image, and video processing.

Easy Integration: Digital filters can be easily integrated into digital systems and
embedded devices. They can be implemented using digital signal processors,
microcontrollers, or software algorithms running on general-purpose processors. This
ease of integration allows for seamless incorporation of digital filters into existing
digital systems and enables their use in a wide range of applications.

Filter Design Tools: Digital filters benefit from the availability of sophisticated
design tools and software packages that facilitate the design, analysis, and
 10

optimization of filters. These tools provide designers with graphical interfaces,

simulation capabilities, and optimization algorithms, simplifying the filter design
process and reducing development time.

1.10 COMPARISON OF FIR AND IIR FILTERS

FIR and IIR filters are two commonly used digital filter designs with different
characteristics. FIR filters are based on a feed forward structure, where the current
output is a weighted sum of past input samples.IIR filters, on the other hand,
incorporate feedback, where the current output depends on both past input and output
samples.FIR filters have a finite impulse response, which means their impulse
response decays to zero within a finite time.IIR filters have an infinite impulse
response, which means their impulse response can extend indefinitely.
FIR filters are always stable, as their impulse response decays to zero. They do
not exhibit stability issues such as oscillations or unbounded growth.IIR filters can be
unstable if their pole locations are not properly controlled. Unstable IIR filters can
result in erratic behavior or numerical instabilities.FIR filters typically have linear
phase response, meaning all frequency components experience the same delay. This
characteristic makes them suitable for applications requiring phase linearity, such as
audio signal processing.IIR filters generally have non-linear phase response, causing
different frequency components to experience different delays. This characteristic
may introduce phase distortion in the filtered signal.
FIR filters can provide a sharp cutoff with high stop-band attenuation, making
them suitable for applications where a precise frequency response is required.IIR
filters can achieve a similar frequency response with fewer filter taps (coefficients),
making them computationally more efficient in certain cases.FIR filters offer more
design flexibility and control over the frequency response. You can design FIR filters
with arbitrary magnitude and phase characteristics using various algorithms.IIR filters
are more constrained in design due to the presence of feedback. Their frequency
response is generally shaped by pole and zero locations, limiting the flexibility to
achieve certain response types.
FIR filters typically exhibit a zero-phase response, meaning they introduce no
phase shift in the filtered signal. This characteristic makes them suitable for
 11

applications where preserving the signal's timing is crucial.IIR filters introduce phase
shift in the filtered signal, which can affect the timing of the output relative to the
input.FIR filters are generally preferred when linear phase response, stability, and
precise frequency response control are important.

1.11 RNS FIR FILTER

The RNS is a method of representing integers using a set of residue classes. In RNS,
an integer is represented by its remainder modulo a set of pairwise co-prime integers.
The RNS provides a way to perform modular arithmetic on the residues in parallel,
which can be advantageous for certain, types of digital signal processing operations.

RNS (Residue Number System) FIR filter is a type of digital filter that
operates on residue numbers instead of traditional binary numbers. The Residue
Number System is a number representation technique that is based on the Chinese
Remainder Theorem. In an RNS FIR filter, the input signal is first converted into
residue numbers, and then the filter coefficients are also represented in residue form.
The filter performs multiplication and addition operations on the residue numbers to
produce the output signal.

The advantage of using the RNS FIR filter is that it can achieve high-speed
processing of digital signals with low power consumption. This is because the RNS
arithmetic can be implemented using simple and fast operations such as modular
additions and subtractions, which require fewer logic gates and consume less power
compared to traditional binary arithmetic.

The advantage of using RNS is that it allows for parallel and fast arithmetic
operations on the residues. This is because modular arithmetic operations such as
addition, subtraction, and multiplication can be performed independently on each
residue, without the need for carry propagation.

RNS offers carry-free, parallel processes, and since it works with tiny
numbers, it is quicker than other traditional approaches. The forward converter,
arithmetic operation, and reverse conversion make up the RNS system. The forward
converter can convert binary to residue using modulo operation, while the reverse
 12

converter can transfer back from residue to binary number using the Chinese
remainder theorem.The block diagram of RNS-based FIR filter is shown below.

FIR Filter
mod m1

RNS FIR Filter RNS

Input ENCODER mod m2 DECODER Output

FIR Filter
mod m3

Figure 1.2 Block diagram of RNS-based FIR filter

RNS has gained significant importance in the field of High speed digital signal
processors in recent years because to its carry-free, parallel nature for addition,
subtraction, and multiplication operations, which leads to faster, computing.An
integer number is converted into a collection of residues in RNS, which are discretely
processed small binary integers. Therefore, it's crucial to have quick fundamental
building blocks for multiplication, subtraction, and addition operations.

1.12OPERATION OF RNS-BASED FIR FILTER

In an RNS-based FIR filter, the input word is converted into a set of residues
( , , ) which are derived from modular arithmetic. The moduli values are fixed
and are in{2 ,2 ,2 } format. If n=4 then the modulo set will be {7,8,9} format.
The filter coefficients and input signals are converted into residue using the moduli
operation shown in Figure 1.2. The output signal from forward converters is given to
various FIR filters and the output is given to the RNS decoder.

Let us consider the input sequence to be x[n]=1010 and moduli set be

m={ , , } = {7,8,9}.

1.13 RNS ENCODER WORKING

The RNS encoder contains three moduli operators for three moduli sets and performs
parallel modulus operation on the input sequence and produces a residue set which is
 13

applied to the FIR filter shown in Figure 1.3 and the output of FIR filters is connected
to the RNS Decoder.

%7 r1

%8 r2
Input signal

%9 r3

Figure 1.3 RNS Encoder implementation

For input x[n]=1010 and moduli set = {7,8,9} the residues are given as

= x[n]%7 = (1010) %7 = 0011

= x[n]%8 = (1010) %8 = 0010

r = x[n] % 9 = (1010) % 9 = 0001

1.14 RNS DECODER WORKING

The Chinese Remainder Theorem in mathematics states that, provided the divisors are
pair wise coprimes, one can uniquely determine the remainder of the division of an
integer n by the product of these integers if one knows the remainders of the
Euclidean division of that integer by a number of other integers.

1.14.1 Chinese Remainder Theorem

Let us consider the residual number ( , , … . . ) with modulo ( , ,….. )

where all , ,…. are mutually prime.

Let M = ( × × … . .× )

= (1.5)
 14

Let be the result that ( ∗ )% = 1, then we have the corresponding number

(1.6)
= ( ) %

1.15 CHARACTERISTICS OF RNS ALGORITHM

The RNS algorithm is a technique used in digital signal processing for high-speed and
low-power applications. The RNS algorithm has several characteristics that make it
useful in these applications:

Parallelism: The RNS algorithm is inherently parallel, allowing for multiple

operations to be performed simultaneously. This results in faster computation times
and increased throughput.

Modularity: The RNS algorithm is modular in nature, allowing for easy scaling and
customization of the system. This makes it useful in a wide range of applications,
from low-power embedded systems to high-performance computing.

Fault tolerance: The RNS algorithm is inherently fault-tolerant, as errors in one

residue do not affect the accuracy of other residues. This makes it useful in
applications where reliability is critical, such as aerospace and defense systems.

Reduced hardware complexity: The RNS algorithm requires less hardware

complexity than other digital signal processing algorithms, such as the Fast Fourier
Transform algorithm. This result in reduced power consumption and hardware costs.

Efficient computation: The RNS algorithm is optimized for integer arithmetic,

resulting in efficient computation of arithmetic operations. This makes it useful in
applications where high-speed arithmetic is required, such as digital signal processing
and cryptography.

Overall, the RNS algorithm is a powerful tool in digital signal processing,

offering arrange of characteristics that make it useful in a wide range of applications.
Its parallelism, modularity, reduced hardware complexity, and efficient computation
make it an attractive option for high-speed and low-power applications.
 15

1.16 PROBLEMS RELATED TO RNS FIR FILTER DESIGN

Many existing hardware-efficient FIR filters usually deal with its filter tap
computation as a sequence of binary numbers which is handled by some LUT blocks
and carries out summation as a fully concurrent process in a hierarchical manner. In
most cases, finite state machines are used for holding the intrusion database
formulated based on the rule set given and matching the payload received according
to rules which cannot achieve impressive performance for many valid reasons such as
sequential state transition and comparator units.

This section briefly categorizes the performance gap that needs to be filled and
the demands over next-generation FIR filter design.

 The path propagation delay and hardware complexity level of any FIR filter
design are increased linearly concerning the filter. Only through appropriate
arithmetic models and associated parallel tasks, these problems can be solved.

 Though DA-based arithmetic is widely preferred for area-efficient FIR filter

design still it has major path delay constraints.

 The power consumption level is very high in any FIR design since convolution
comes summation of each tap results computed and its subsequent bit
transitions.

 Most of the existing FIR designs are incompatible with attainable frequency
response while supporting high-speed and energy-limited applications.

1.17 APPLICATION OF RNS FIR FILTER

ECG (Electrocardiogram) signal is a type of bioelectrical signal that measures the

electrical activity of the heart. The ECG signal is recorded by placing electrodes on
the skin, which detect the small electrical impulses generated by the heart as it
contracts and relaxes. The ECG signal is typically measured in millivolts (mV) and
plotted over time. The signal consists of a series of waves and intervals that
correspond to the different phases of the cardiac cycle.

The different components of the ECG signal include:

P wave: The P wave represents the depolarization (contraction) of the atria.

 16

QRS complex: The QRS complex represents the depolarization (contraction) of the
ventricles. It is composed of the Q wave, R wave, and S wave.

T wave: The T wave represents the repolarization (relaxation) of the ventricles.

PR interval: The PR interval is the time between the onset of the P wave and the
onset of the QRS complex. It represents the time it takes for the electrical impulse to
travel from the atria to the ventricles.

QT interval: The QT interval is the time between the onset of the QRS complex and
the end of the T wave. It represents the time it takes for the ventricles to depolarize
and repolarize.

Figure 1.4 ECG signal

The ECG signal provides important information about the function and
condition of the heart. Abnormalities in the ECG signal can indicate various cardiac
conditions, such as arrhythmias, heart block, and myocardial infarction (heart attack).
The ECG signal can be analyzed using various techniques, such as time-domain
analysis, frequency-domain analysis, and wavelet analysis. These techniques can
provide information about the amplitude, frequency, and timing characteristics of the
signal, which can be used to detect and diagnose cardiac abnormalities.

ECG signals can be affected by external sources of interference, such as power

lines, electromagnetic fields, and other electrical equipment. Incorrect placement of
ECG electrodes can result in poor signal quality and noise. Patient movement can
cause artifacts in the ECG signal, which can be mistaken for noise. Muscle activity,
such as muscle tremors or contractions, can generate noise in the ECG signal.
 17

High-frequency noise in ECG signals can be caused by a variety of sources,

including external interference from power lines or electrical equipment, poor
electrode contacts or skin preparation, and physiological sources such as muscle
activity or tremors. One of the most common techniques for removing high-
frequency noise from ECG signals is low-pass filtering. A low-pass filter allows low-
frequency signals to pass through while attenuating high-frequency signals. The
cutoff frequency of the low-pass filter should be set high enough to allow the ECG
signal to pass through but low enough to remove high-frequency noise.

It is important to note that removing high-frequency noise from ECG signals

can be challenging, as high-frequency noise may contain important information.
Careful consideration should be given to the choice of filtering technique to ensure
that the important features of the ECG signal are not lost in the filtering process.

1.18 LOW-PASS FIR FILTER DESIGN

A low-pass FIR filter is a type of digital filter that is designed to allow low-frequency
components of a signal to pass through while attenuating or blocking high-frequency
components. The filter is implemented using a finite number of coefficients, and it is
said to have a finite impulse response because its output response will decay to zero
after a finite number of samples. The design of a low-pass FIR filter involves
following steps. The first step in designing an FIR filter is to specify the desired
characteristics of the filter. This includes the pass-band frequency range, the stop-
band frequency range, and the desired attenuation levels in the stop-band and pass-
band. The window function determines the shape of the filter's frequency response.
There are several types of window functions, including the rectangular window, the
Hamming window, the Blackman window, and the Kaiser window. Each window
function has its own advantages and disadvantages, depending on the specific
requirements of the filter. Choosing the best windowing technique for a particular
application depends on the specific requirements of that application.

Windowing techniques are used to design digital filters with finite impulse
response by reducing the effect of Gibbs phenomenon. The Gibbs phenomenon is a
type of overshoot or ripple that occurs in the frequency response of a filter due to
abrupt truncation of an IIR filter. The windowing technique involves multiplying the
 18

desired impulse response of the filter by a window function. The window function
essentially smooth’s out the sharp transitions in the filter's impulse response and
reduces the effect of Gibbs phenomenon. Therefore, windowing techniques help to
improve the frequency response of FIR filters, by reducing the magnitude of the side
lobes and minimizing the ripple in the pass band and stop-band. Different windowing
techniques result in different frequency responses.

Some window functions, like the rectangular window, have a wide main lobe
and high side lobes, which can cause significant distortion and poor frequency
resolution. Other window functions, like the Hamming or Blackman windows, have
narrower main lobes and lower side lobes, which can provide better frequency
resolution and less distortion. Therefore, it is important to choose a window function
that provides the desired frequency response.In general, narrower main lobes result in
higher side lobe levels, while wider main lobes result in lower side lobe levels. This
trade-off must be considered when selecting a window function. For example, if low
side lobe levels are critical, we may choose a window function with a wider main
lobe, such as the Blackman window.Window functions can affect the time-domain
properties of the filter, such as the filter's settling time, transient response, and pass-
band ripple. For example, the Kaiser window can provide good frequency resolution
and low sidelobes, but it can also result in a longer settling time compared to other
window functions.

Some window functions are more computationally complex than others. For
example, the Kaiser window requires more computation time than the Hamming
window. Therefore, it is important to consider the computational complexity of the
window function when choosing a suitable technique. The filter order determines the
number of filter coefficients required to achieve the desired filter characteristics. The
higher the filter order, the better the filter's frequency response, but the more
computational resources required to implement the filter. The filter order in a digital
filter determines its frequency response, accuracy, and computational
complexity.Increasing the number of orders in a digital filter generally results in a
more accurate frequency response, with a sharper cutoff and less ripple in the pass-
band and stop-band. However, this also increases the computational complexity of the
 19

filter and requires more processing power.Conversely, reducing the order in a filter
reduces its computational complexity but also results in a less accurate frequency
response with a wider transition band and more ripple in the pass-band and stop-band.

Choosing the appropriate filter order depends on the desired filter

specifications and the available computational resources. A balance must be choosen
between the filter's performance and complexity. Once the filter order and window
function have been determined, the filter coefficients can be computed using a
mathematical formula.

There are various windowing techniques that are used in digital signal
processing for designing FIR filters. Here is a brief explanation of some commonly
used windowing techniques along with their advantages and disadvantages:

Rectangular window

The rectangular window is the simplest windowing technique where all the samples in
the window are equal to one. The advantage of using the rectangular window is that it
provides a very sharp cutoff but it also results in significant sidelobes which degrade
the filter performance.

( )=1 (− ≤ ≤ ) (1.7)

=0 ℎ

N=25

-12 -8 n

Figure 1.5 Rectangular window sequence

Hamming window

The Hamming window is designed to reduce the sidelobes of the rectangular window.
It is defined as:
 20

( ) = 0.54 + 0.46 cos (− ≤ ≤ ) (1.8)

=0 ℎ

WH(n) N=25

-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 n

Figure 1.6 Hamming window sequence

where N is the filter length and n is the sample index. The Hamming window has a
wider main lobe and smaller side lobes compared to the rectangular window.
However, it still suffers from significant side lobe levels which can be problematic for
some applications.

Hanning window

The Hanning window is similar to the Hamming window but it has a smoother
transition from the main lobe to the side lobes. It is defined as:

( ) = 0.5 + 0.5 cos (− ≤ ≤ ) (1.9)

=0 ℎ

N=25

-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 n

Figure 1.7 Hanning window sequence

The Hanning window provides a good balance between the main lobe width and side
lobe levels but it still suffers from some level of side lobes.
 21

Blackman window

The Blackman window is designed to provide even lower side lobes compared to the
Hamming and Hanning windows. It is defined as:

( ) = 0.42 + 0.5 cos + 0.08 cos (− ≤ ≤ ) 1.10)

=0 ℎ

WB (n) N=25

-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 n

Figure 1.8 Blackman window sequence

The Blackman window provides the best side lobe suppression among the four
windowing techniques mentioned here. However, it has a wider main lobe compared
to the other windows which can result in higher pass band ripple.

Kaiser window

The Kaiser window is a parameterized windowing technique that allows for adjusting
the trade-off between the main lobe width and side lobe levels. The Kaiser window
provides good control over the side lobe levels and main lobe width and is widely
used in filter design applications.

0.8

0.6
WB (n)

0.4

0.2

0
0 5 10 15 20
n

Figure 1.9 Kaiser window sequence for α=3

 22

1.19 METHODOLOGY

In general, many DSP applications need to accommodate a large number of Multiply

and Accumulator (MAC) operations, and its hardware accumulation is also increased
accordingly. Moreover, many existing investigation works also included several
arithmetic optimization methods to reduce the design complexity overhead and path
delay propagation overhead with the condensed critical path.
To achieve both complexity reduction and performance efficiency residue
number system has been highly motivated in many DSP applications since input
operand bit size is down-converted into smaller units to carry out arithmetic
operations. Critical path delay optimization also can be easily optimized by reducing
carry propagation as bit size reduction which is also directly related to the design
complexity reduction. Though RNS arithmetic is widely used as a Multiply and
Accumulation (MAC) unit in many dominant applications such as Fast Fourier
transform, Digital signal processing, and wireless communications, it is observed to
have limitations in terms of hardware complexity during the conversion process.

1.20 THE MOTIVATION OF THE RESEARCH

Many portable and battery power electronic gadgets like mobile phones, laptops,
hearing aids, pacemaker, etc demand chip that consumes less power and area. Several
applications like biomedical imaging, wireless communication, Software-Defined
radio, high-performance computing, etc. are based on digital signal processing
algorithms. All these applications demand low-power implementation of DSP
architectures. In many of these DSP applications, filters are the most essential blocks.
FIR filters play a paramount role in DSP applications. The MAC operations involved
in FIR filters will occupy more area and power consumption.
A fundamental problem over FIR filter systems in the signal analysis is
hardware complexity overhead for improved frequency response. The efforts made by
the research community to design optimization model is always growing, but
hardware efficient design without compromising performance is difficult.
It is feasible to employ ROM-based LUTs since the standard DA method used
to create an FIR filter assumes that the impulse response coefficients are fixed.
Nevertheless, the memory need for DA-based FIR filter implementation exponentially
 23

grows with filter order. In many DSP applications, FIR filters are crucial.
Implementing a higher-performing FIR filter is now essential due to the development
of VLSI technology and the growing popularity of DSP.

1.21 ORGANIZATION OF THE THESIS

The organization of the thesis is as follows

Chapter 1 details the introduction to FIR filter, description and operation of FIR filter,
RNS system, Application of RNS FIR filter, Low-pass FIR Filter design,
motivation and scope of the research.

Chapter 2 summarizes various topics associated with this work, and the untried
observations on the literature survey conclude the chapter. The
investigation is carried over various topologies with the measures such as
area utilization rate, operating speed, throughput rate, and energy
consumption. This chapter also investigates the trade-off that exists
between FIR length and the achieved performance of previous techniques.

Chapter 3 deals with the architectural design including LUT decomposed RNS FIR
Filter to reduce area and power consumption.

Chapter 4 highlights with the Modified Binary Distributed Arithmetic RNS FIR Filter,
Partitioned DA RNS FIR filter, Memory Less DA-I, Memory Less DA-II,
RNS FIR Filter to increase the performance of the filter.

Chapter 5 describes with RNS Filter using Optimized Adders and Optimized
Multipliers to recognize the high performance sub modules.

Chapter 6 focuses onimplementation of RNS FIR filter using Pipelining, and

Retiming techniques to optimize the delay by balancing the FIR filter
design and evaluated to denoise the ECG signal.

Chapter 7 concludes the work on design area delay power optimized RNS FIR Filters
for ECG denoising applications, and suggests future research directions.