Design of Microstrip Filters Using Neural Network

Vivek Singh Kushwah & G.S. Tomar; Member IEEE

Amity School of Engineering & Technology, New Delhi-india

Vikrant Institute of Technology & Management, Indore 452009

E-mail:vskushwah@aset.amity.edu, gstomar@ieee.org

Abstract for the desired cutoff frequency and the

This paper presents the general design
procedure for microstrip filters using
artificial neural networks and this is
demonstrated using Low pass filters. In this
design procedure, synthesis is defined as the
forward side and then analysis as the reverse
side of the problem. in this work, the neural
network is employed as a tool in design of
microstrip filters. MATLAB programming
language GUI (Graphical User Interface )
tools are used for implementing results.
Neural network Training
algorithms are used for trained the samples
so that error can be minimized. Therefore ,
one can obtain the geometric dimensions
with high accuracy.
1. Introduction:
The design of low pass filters
involves two main steps. The first one is to
select an appropriate low pass prototype. The
choice of the type of response, including Pass
band ripple and the number of reactive
elements will depend on the required
specifications. The element values of the low
pass prototype filters, which are usually
normalized to make a source impedance go =
1 and a cutoff frequency Ωc = 1.0, are then
transformed to the L-C elements
desired source impedance, which is
normally 50 ohms for microstrip filters.
The next main step in the design of
microstrip low pass filters [9] is to find an
appropriate microstrip realization that
approximates the lumped element filter.
The element values for the low pass
prototype with Chebyshev response at pass
band ripple m normalized values gi i.e. g1,
g2, g3, g4......, gn. The filter is assumed to be
fabricated on a substrate of dielectric
constant εr and of thickness t mm. for
Angular (normalized) cutoff frequency Ωc,
characteristic impedance source/load, Zo =
50 ohms, are taken , whose physical
lengths are smaller than a quarter of
guided wavelength (λg/4) at which they
operate, are the most common components
for approximate microwave realization of
lumped elements in microstrip filter
structures.

Figure 1: Lumped-Element Lowpass

Prototype Filter
The simplest form of low pass filter may
just consist of series inductors, which is
often found in applications for direct current
or dc block. In design and realization of
microstrip filters, short section of
transmission line or stub, whose length is Figure 2: Lumped-Element Low Pass
much smaller than a quarter of guided Filter
wavelength, are the most common
components . As a small open circuited loss 2.1 Design Parameters and Analysis
less microstrip line stub is equivalent to shunt For the proposed design shown in figure
capacitor and similarly small short-circuited 1, which is consisting of short-circuited
line is equivalent to shunt inductor. For a stub (grounded line) on a commercial
more selective low pass filter, more of such substrate; the following parameters are
elements are required [1]. The quasi lumped considered:
elements discussed for this design have their g1 = g3 = 1.0316
size smaller then the quarter of the guided g2 = 1.1474
wavelength at the cutoff frequency 0.75 Cut-off frequency, fc = 0.75 GHz.
GHz. This type of Relative Dielectric Constant, єr = 4.7
low pass filter may be designed based on a Height of substrate, h = 1.6 mm
lumped-element lowpass prototype filter, Characteristic Impedance, ZO = 50 Ω
using normalized elements values gi and by Corresponding width of the microstrip, Wc
terminating into impedance ZO, as shown in = 2.91mm
figure 1. ZOC = 20 Ω
ZOL = 100 Ω
2. Design for Microstrip Filters The step impedance open circuited step
In order to design a three-pole open stub low microstrip low pass filter has been design
pass filter while considering the normalized using the above equations and parameters
elements values; the values of series .the conventional microstrip low pass filter
inductors L1 , L3 and capacitor C2 (as shown is shown in figure 3.
in figure 2) are obtained by using design
equations (1) and (2). The components L1, L3
& C2 are realized by high and low impedance
microstrip open circuited stub respectively.
The values of various inductance Li and
capacitance Ci are obtained from the
equations, as
Li = (Zo/go) (Ωc/2πfc) g1 (1)
Ci = (go/Zo) (Ωc/2πfc) g2 (2)
Physical length (in mm) of the high and low
impedance lines (inductance and capacitance
respectively) is found out by the formulae
given below, Figure 3: Microstrip Low Pass Filter
IL = λgl /2π Sin-1 (ωc Li / ZOL) (3) Design
IC = λgc /2π Sin-1 (ωc Ci Zoc) (4)

3. Implementation & Results

Suppose that we have to design the
microstrip low pass filter at 0.7 Ghz
resonant frequency using the following
parameters.
Dielectric constant = 4.7
Substrate thickness = 1.6 mm
Now we can calculate the length and width
of capacitive element of filters using general
relations. Length and width of capacitive
element of the filters for 0.7 Ghz frequency is
calculated , given by
In mm.
εr =4.7, h=1.6, W=7.684, L=15.14, f=0.7Ghz

Figure 4: Microstrip Low Pass Filter Design

When IE3D electromagnetic simulation is

performed and changes the dimensions of
filters then we obtain the different resultant
graphs between return loss and frequency: Figure 6: Response of microstrip low pass
filter for 0.5 Ghz

Now trained the neural network by using

training algorithms and transfer
function.Figure 7 shows the neural
network training graph results for the
Synthesis of microstrip Low pass Filter. it
is clear from this figure ,training performs
in 76 epochs . We pass the full set of input
samples through the neural network to
compute the least squared error function
we will use in the back propagation of the
errors step. Each such pass is called an
epoch. From this figure it is clear that error
minimizes from103 to nearly 100.

Figure 5: Response of microstrip low pass

filter for 0.7Ghz

If εr =4.7, h=1.6 , W=11.484, L=15.14


1 dielectric constant.
Figure 7 training graph results for
microstrip low pass filter
Figure 8 Neural network architecture for
microstrip low pass filters
As shown in the above neural network
architecture, it consists of three layers.
The three-layer network has one output layer
(layer 3), one hidden layer (layer 2) and one
input layer (layer 1 ).
Input and output layer consists of two
neurons .inputs f1 and f2 are applied at the
input neurons while outputs W & L are
obtained from output neurons. A constant
input 1 is fed to the biases for each neuron.
Note that the outputs of each intermediate
layer are the inputs to the following layer. A
layer that produces the network output is
called an output layer.
Conclusion
In this work, the neural network is
employed as a tool in design of the
microstrip filters[11] . In this design
procedure, synthesis is defined as the
forward side and then analysis as the
reverse side of the problem. Therefore, one
can obtain the geometric dimensions with
high accuracy, which are the length and
the width of the filters in our geometry, at
the output of the synthesis network by
inputting resonant frequency, height and
REFERENCES
.
[1] R.K. Mishra, A. Patnaik, “Neural network-
based CAD model for the design of square-
patch antennas”, Antennas and Propagation,
IEEE Transactions, Vol. 46, No. 12, pp. 1890
– 1891, December 1998.
[2] S. Devi, D.C. Panda, S.S. Pattnaik, “A
novel method of using artificial neural
networks to calculate input impedance of
circular microstrip antenna”, Antennas and
Propagation Society International Symposium,
Vol. 3, pp. 462 – 465, 16-21 June 2002.
[3] K. G¨uney, N. Sarıkaya, “Artificial neural
networks for calculating the input resistance of
circular Microstrip antennas”, Microwave and
Optical Technology Letters, Vol. 37, No. 2,
pp. 107-111, 20 April 2003.
[4] G. Angiulli, M. Versaci, “Resonant
frequency evaluation of microstrip antennas
using a neural-fuzzy approach”, Magnetics,
IEEE Transactions, Vol. 39, No. 3, pp. 1333 –
1336, May 2003.
[5] R.K. Mishra, A. Patnaik,
“Neurospectral computation for input
impedance of rectangular microstrip
antenna”, Electronics Letters, Vol. 35, No. 20,
pp. 1691 – 1693, 30 Sept. 1999.
[6] R.K. Mishra, A. Patnaik, “Designing
rectangular patch antenna using the
neurospectral method”, Antennas and
Propagation, IEEE Transactions, Vol. 51, No. 8,
pp. 1914 – 1921, Aug. 2003.
[7] Q. J. Zhang, K. C. Gupta, Neural Networks
for RF and Microwave Design, Artech House
Publishers, 2000.
[8] J. Park, W. I. Sandberg, “Universal
Approximation Using Radial Basis Function
Networks”, Neural Computation, Vol. 3, pp. 246-
257, 1991
[9] Jia-Shen G. Hong & M.J. Lancaster,
“Microstrip Filters for RF/ Microwave
Applications” John Wiley & Sons Inc., 2001.
[10] Jia-Sheng Hong; Lancaster M.J, “Recent
progress in planar microwave filters,” IEEE
Trans. Antennas Propagat., Vol. 2, pp. 1134 –
1137, August 1998.
[11] G. Mathaei, L.Young & E.M.T. Jones,
“Microwave Filter impedance matching networks
and coupling structures,” Artech House,
Norwood, MA, 1980.