CONVENTIONAL AND EVOLUTIONARY DIGITAL FILTER DESIGN

Mr. Rajesh Uday1, Mr. Gaurav Phulwari2, Ms. Nidhi Verma

1
[Link] Scholar, Electronics & Comm., Bhagwant University, Ajmer, India, [Link]@[Link]
2
[Link] Scholar, Electronics & Comm., Bhagwant University, Ajmer, India, [Link]@[Link]
3
Assistant Professor, Electronics & Comm., Bhagwant University, Ajmer, India, verma.nidhi17@[Link]

Abstract— Digital Filter style is rising space of analysis in current technological era. Varied techniques are
obtainable and explored for style functions. A previous study shows that digital filter style is achieved mistreatment
neural network approach by considering Continuous Hopfield Network model. Another approach mistreatment ANN
considers coaching of multilayer perceptron with varied sets of sinusoids. Genetic formula approach and applied math
approach have conjointly been thought-about for digital filter style. This paper covers the fundamental needs of
digital filter style and a study of standard and organic process styles like genetic formula, etc. is additionally
conferred. It conjointly includes the boundaries of standard approach for planning. A style example is additionally
studied for elliptical filter and its magnitude and section response is studied.

Keywords— Evolutionary approach (EA), elliptic filter

I. INTRODUCTION
Filters are wide utilized in signal process and communication circuits systems in applications corresponding to channel
deed, noise reduction, radar, audio process, video process, medical specialty signal process. In an exceedingly wireless,
band pass filter or tuner is employed to extracts the signals in an exceedingly radio channel. The Dolby system [7] is
another example of the appliance of filters. During this case pre-filtering and post-filtering are wont to minimize the
impact of noise. In hi-fi audio a compensating filter could also be enclosed in an exceedingly preamplifier to complete
the non-ideal frequency response characteristics of the speakers.

Depending on style of the filter equation and structure of implementation filters could broadly speak be classified into the
subsequent classes:

 Linear filters versus non-linear filters.

 Adaptive time variable filters versus non adaptive time invariant filters.

 Algorithmic versus non algorithmic filters.

 Direct type, cascade type, parallel and lattice structures.

There are two main forms of filter, analog and digital. Analog filter uses analog electronic circuits created up from parts
corresponding to resistors, capacitors and op-amps to provide the desired filtering impact. Such filter circuits are wide
utilized in such applications as noise reduction, video signal sweetening, graphic equalizers in hi-fi system, and plenty of
different areas. There are well established customary techniques for coming up with associate in analog filter circuit for a
given demand. In an exceedingly digital filter, the signal is diagrammatic by a sequence of numbers, instead of a voltage
or current. The most benefits of digital filters over the analog filters [7], [8] are:

1. A digital filter is programmable, i.e. its operation is set by a program hold on within the processor's memory. This
suggests the digital filter will simply be modified while not moving the electronic equipment. Analog filter will solely be
modified by redesigning the filter circuit.

2. Digital filters are simply designed, tested and enforced on an all-purpose pc or digital computer.

3. In contrast to their analog counterparts, digital filters will handle low frequency signals accurately. Because the speed
of DSP technology continues to extend, digital filters are being applied to high frequency signals within the frequency
domain additionally.

4. The characteristics of analog filter circuits are subject to drift and are captivated with temperature. Digital filters don't
suffer from these issues, and then are extraordinarily stable with respect each to time and temperature.

Digital filters are classified either as Finite period unit pulse response (FIR) filters or Infinite period unit pulse response
(IIR) filters [8], looking on the shape of unit pulse response of the system. Within the FIR system, the impulse response
sequence is of finite period, i.e. it's a finite variety of non-zero terms. Digital filters are classified as algorithmic and Non-
Recursive filters. The response of algorithmic or FIR filters rely solely upon
present and former input of signal. FIR filters have the subsequent advantages:
  They will have a definite linear section.

 They are perpetually stable.

 The style strategies are usually linear.

 They are complete expeditiously in hardware.

 The filter set out transients has finite period.

II. FILTER DESIGN

Filter style is that the method of coming up with a filter, usually a linear shift-invariant filter, that satisfies a collection of
necessities, a number of that square measure contradictory. The matter is to search out a realization of the filter that
meets every of the wants to an adequate degree to form it helpful. The filter style method [8] may be delineated as
associate degree improvement downside wherever every demand contributes with a term to a blunder perform that ought
to be decreased. Sure components of the planning method may be machine-driven.

Typical Design Requirements

To design a filter suggests that to pick out the coefficients specified the system has specific characteristics. The specified
characteristics area unit explicit in filter specifications [9]. Most of the time filter specifications talk to the frequency
response of the filter. Typical style needs that area unit thought-about within the style methods are:

1. The filter ought to have a selected frequency response.

2. The filter ought to have a selected impulse response.

3. The filter ought to be causative.

4. The filter ought to be stable.

5. The filter ought to be localized.

6. The machine complexness of the filter ought to be low.

7. The filter ought to be enforced especially hardware or software package.

Computational Complexity

A general need in any style is that the quantity of operations (additions and multiplications) required figuring the filter
response is as low as attainable. In sure applications, this need could be a strict demand, for instance thanks to restricted
computational resources, restricted power resources, or restricted time. The last limitation is typical in time period
applications. There are many ways that during which a filter will have completely different process complexness [10].
For instance, the order of a filter is additional or less proportional to the quantity of operations. This suggests that by
selecting an occasional order filter, the computation time may be reduced.
For discrete filters the machine complexness is additional or less proportional to the quantity of filter coefficients. If the
filter has several coefficients, for instance within the case of third-dimensional signals reminiscent of picturing
information, it's going to be relevant to scale back the quantity of coefficients by removing those that area unit
sufficiently on the point of zero. Another issue regarding machine complexness is disconnectedness, that is, if and the
way a filter are often written as a convolution of two or additional less complicated filters. Specially, this issue is of
importance for third-dimensional filters, e.g. second filter that area unit employed in image process. during this case, a
big reduction in machine complexness are often obtained if the filter are often separated because the convolution of one
1D filter within the horizontal direction and one 1D filter within the vertical direction.

III. CONVENTIONAL AND EVOLUTIONARY DIGITAL FILTER DESIGN FEATURES

Evolvable hardware refers to one specific kind of hardware whose architecture/structure and functions
amendment dynamically and autonomously so as to enhance its performance in playacting bound tasks [10].
The emergence of this new field in recent years has been influenced deeply by the progresses in
reconfigurable hardware and organic process computation.
 1. Digital filters play a very important role in communications systems, typically at the interface between
digital and analog signal process systems. Examples square measure mobile communications, speech
process, modems, etc.

2. Whereas the science of digital filter style is extremely well established and researched, there are not
any ‘conventional’ style procedures that result in optimum styles with acceptable effort within the general
case. A survey of relevant journals reveals that digital filter style is an energetic space of analysis [2][8].

3. The look house for digital filters is well outlined however giant and complicated. A well outlined house
facilitates comparison of various results. An outsized and complicated style house can challenge our organic
process system and can be smart at evaluating our system’s quality in handling robust style issues.

4. A quantitative live of filter performance is usually accessible, providing a fitness live for EAs that's
comparatively straightforward to cypher. It conjointly provides an easy metric in examination completely
different styles. Since no general style methodology is offered, organic process styles are going to be a decent
and automatic various to manual style.

IV. ADVANTAGES OF EVOLUTIONARY DESIGN

An evolutionary process style approach offers variety of benefits over the traditional one utilized by human
designers though there are some vital problems that stay open.

First, the evolutionary process style approach will explore a way wider vary of style alternatives than those
can be thought-about by people. This has been shown by several experiments in alternative style tasks, like
evolutio-nary style of neural networks [2], [3], or of building architectures [4].These experiments
incontestable however evolutionary process techniques can be applied to evolving novel styles that were
troublesome to get by people.

Second, the evolutionary process style approach doesn't assume a priori data of any specific style domain. It
will be applied by users while not resorting to domain consultants. It will be utilized in domains wherever
little prior data is on the market or wherever such data is incredibly expensive to get.

Third, the evolutionary process style approach is incredibly versatile. It will touch upon non -differential or
maybe discontinuous objective functions. It will touch upon varied linear and nonlinear constraints further as
objectives.

V. LIMITS OF THE CONVENTIONAL PROCESS

Designing digital filters, particularly algorithmic filters, isn't an easy drawback. Instead, varieties of approximation
strategies are developed, sometimes applicable just for a selected category of style issues. There square measure two
issues with this style methodology.

First, for brand new drawback categories, approximation approach should be developed initial. Let’s say, in [5], a method
for style of stable IIR filters with equiripple pass bands and peak-constrained least-squares stop bands is developed. This
is often totally different from biological process approaches, which may typically be written for abundant larger
categories of issues, permitting it to be used for abundant totally different filter style issues.

The second drawback with typical approach is that, reckoning on the precise approach taken, the ensuing style is
probably going to be suboptimal. Let’s say, the approach taken in [6] is predicated on repetitious quadratic programming
methodology with linearized constraints with a least-square objective.

VI. DESIGN EXAMPLE

Design a low pass and high pass FIR filter of order five with pass band edge frequency 2.4 kHz, pass band ripple =1
decibel, stop band attenuation = 40dB, stop band edge frequency = 2kHz and normalized frequency of ten kilo hertz. An
elliptic filter may be a signal process filters with equal ripple behaviour in each the pass band and also the stop band. The
number of ripple in every band is severally adjustable, and no different filter of equal order will have a quicker transition
in gain between the pass band and also the stop band for the given values of ripple.
 ∈C
N ( ΩΩ )
2

√
│H (jΩ) │=
2
1+∈ C N
2
( ΩΩ )
2

Where C- is pass band ripple parameter.

CN is Jacobian elliptic function of order N.

W is pass band edge frequency.

Fig. 1: Magnitude and Phase Response of Low Pass Elliptical Filter of Order 5

Fig. 2: Magnitude and Phase Response of High Pass Elliptical Filter of Order 5

VII. CONCLUSIONS
In this paper filter style needs are thought of and study of typical and biological process approaches show that for crucial
applications evolutio-nary approach is best. One style example for elliptical filter shows that the part response of elliptic
filters is a lot of nonlinear in pass band than a comparable Butter-worth filter or a Chebyshev filter, particularly close to
the band edge. In future optimisation techniques is applied for coming up with of filters and might be compared with
conven-tional techniques.

REFERENCES
[1] Rabiner L.R., J.H. McClellan, and T.W. Parks, “FIR Digital Filter Design Techniques using Weighted Chebyshev
Approximations”, Proc. IEEE, 63, (1975).
[2] X. Yao and Y. Liu, “A New Evolutionary System for Evolving Artificial Neural Networks”, IEEE Transactions on
Neural Networks, 8 (3), pp. 694 – 713, 1997.
[3] X. Yao and Y. Liu, “Towards Designing Artificial Neural Networks by Evolution”, Applied Mathematics and
Computation, 91 (1), pp. 83 – 90, 1998.
[4] X. Yao and Y. Liu, “Evolving Artificial Neural Networks through Evolutionary Programming”, in Evolutionary
Programming V: Proc. of the Fifth Annual Conference on Evolutionary Programming (L.J. Fogel, P.J. Angeline, and T.
Back, eds.), (Cambridge, MA), pp. 257 – 266, The MIT Press, 1996.
[5] X. Yao and Y. Liu, “Evolutionary Artificial Neural Networks that Learn and Generalize Well”, in 1996 IEEE
international Conference on Neural Networks, Washington, DC, USA, Volume on Plenary, Panel and special Sessions,
pp. 159 – 164, IEEE Press, New York, NY, 3-6 June 1996.
[6] W.S. Lu. “Design of Stable IIR Digital Filters with Equiripple Pass Bands and Peak-constrained Least-squares Stop
Band”. IEEE Transactions on Circuits and Systems II: Analog and Digital Signal Processing, 46 (11), 1421 – 1426,
1999.
[7] D.E. Johnson, “Introduction to Filter Theory”, Prentice Hall Inc... 1976.
[8] A. Antoniou, “Digital Filters: Analysis, Design and Applications”, 2nd ed."McGraw Hill, 1993.
[9] H.D. Helms, “Digital Filters with Equiripple and Min-max Responses”, IEEE [Link] Electroacoust, AU-19, pp.
87 – 93, Mar 1971.
[10] O. Hermann, and H.W. Schüssler, “Design of Nonrecursive Digital Filters with Minimum Phase”, Electron. Lett., 6,
pp. 329 – 330, 1970.
[11] X. Yao, “Following the Path of Evolvable Hardware”, Communications of the ACM, 42 (4), pp. 47 – 49, 1999.