LAPPEENRANTA UNIVERSITY OF TECHNOLOGY

FACULTY OF TECHNOLOGY

Master’s Degree Programme in Technomathematics and Technical Physics

Tatiana Pavlenko

Tunable lumped-element bandpass filters

for Cognitive Radio application

Examiners: Professor Erkki Lähderanta

D. Sc. (Tech.) Tauno Vähä-Heikkilä
Supervisors: D. Sc. (Tech.) Tauno Vähä-Heikkilä
Professor Erkki Lähderanta
 ii

Abstract

Lappeenranta University of Technology

Faculty of Technology
Master’s Degree Programme in Technomathematics and Technical Physics
Tatiana Pavlenko

Tunable lumped-element bandpass filters for

Cognitive Radio application

Master’s thesis
2013
56 pages, 48 figures, 2 tables

Examiners: Professor Erkki Lähderanta

D. Sc. (Tech.) Tauno Vähä-Heikkilä
Keywords: filter design, bandpass filter, tunable filter, lumped-element filter, Cognitive
Radio, UHF

In this thesis the design of bandpass filters tunable at 400 MHz – 800 MHz was
under research.
Microwave filters are vital components which provide frequency selectivity in
wide variety of electronic systems operating at high frequencies. Due to the
occurrence of multi-frequency bands communication and diverse applications of
wireless devices, requirement of tunable filters exists. The one of potential
implementation of frequency-agile filters is frontends and spectrum sensors in
Cognitive Radio (CR). The principle of CR is to detect and operate at a particular
available spectrum without interfering with the primary user’s signals. This new
method allows improving the efficiency of utilizing allocated spectrum such as
TV band (400 MHz – 800 MHz).
The focus of this work is development of sufficiently compact, low cost tunable
filters with quite narrow bandwidth using currently available lumped-element
components and PCB board technology.
Filter design, different topologies and methods of tuning of bandpass filters are
considered in this work. As a result, three types of topologies of bandpass filter
were simulated and realised. They use digitally tunable capacitors (DTCs) for
adjusting central frequency at TV "white space" spectrum. Measurements
revealed that schematics presented in this work have proper output response and
filters are successfully tuned by DTCs.
 iii

Acknowledgments

I would like to express my sincere thanks and appreciation to my supervisor

Tauno Vähä-Heikkilä for possibility to work at VTT, guidance and full support.
Despite of high busyness, he always found time to listen and give advice. I would
also like to thank the members of the Adaptive Radios group of VTT which
helped me a lot. Thanks to Jan Holmberg for sharing experience and valuable
advices on RF design; to Mikko Kaunisto for his great help at layout design; to
Manu Lahdes for arranging control program so quickly; to Alpo Ahonen for
assembling circuits professionally and Hannu Hakojärvi for help in
measurements.
In addition, special thanks to Professor Erkki Lähderanta, who provide
opportunity to study in Finland, and to my department in St.-Petersburg
Electrotechnical University that gave deep background knowledge and introduced
me to the world of RF/Microwave engineering.
My warm wishes and thanks go as well to my dear friends at St.-Petersburg and
Lappeenranta. And the deepest appreciation belongs to family and Mikhail, whom
I love dearly. They provided priceless support and continuous encouragement.

Tatiana Pavlenko, May 2013

 iv

Table of Contents

Abstract ..............................................................................................................ii
Acknowledgments .............................................................................................ii
Table of Contents ............................................................................................. iv
List of Symbols and Abbreviations ................................................................. v
1. Introduction ................................................................................................... 1
1.1. Filters ............................................................................................................ 1
1.2. Tunable filters ............................................................................................... 2
1.3. Application of tunable filters for Cognitive Radio ....................................... 3
1.4. Research Goals and Approach ...................................................................... 4
2. Filter Design .................................................................................................. 6
Fundamentals and Design of LC passive bandpass filters .................................. 6
3. Tunable filters ............................................................................................. 14
3.1. Tuning methods ......................................................................................... 14
3.2. Examples of tunable filters ........................................................................ 15
4. Tunable filters simulations ......................................................................... 22
4.1. Introduction ................................................................................................ 22
4.2. Topologies of filter with ideal components ............................................... 22
4.3. Simulations with accounting influence of losses ....................................... 26
4.4. Filter simulations with real components .................................................... 29
4.5. Filter simulations with tuning elements ........................................... 32
4.6. Realisation of bandpass tunable filter prototypes .......................... 36
5. Measurement results ................................................................................... 41
6. Conclusions .................................................................................................. 45
6.1. Summary .................................................................................................. 45
6.2. Future work ................................................................................................ 46
References ........................................................................................................ 47
 v

List of Symbols and Abbreviations

bandwidth scaling factor

, , k, , auxiliary variables

ε passband ripple

angular frequency

BW bandwidth

C capacitance / capacitor

f frequency

g component value

IL passband insertion losses, losses in the components due to their

inherent resistive losses

stopband insertion losses

K inverter

L inductance / inductor

LC inductor and capacitor

N number of components

n order of filter

Q quality factor, ratio of insertion loss at central frequency to its 3dB

bandwidth

Qr resonator quality factor

RL passband return losses-

R resistance / resistor

S ratio of stopband to passband frequencies

S-parameters scattering parameters

input port reflection coefficient

forward transmission coefficient

 vi

polynomial of order N

A/D analog-to-digital converter

BAW bulk acoustic wave

BST barium strontium titanate Ba1-xSrxTiO3 ferroelectric material

CAD computer-aided design

CMOS FET Complementary Metal Oxide Silicon Field Effect Transistor

CPW coplanar waveguide

CR Cognitive Radio

DC direct current

DTC digitally tunable capacitor

ESD Electrostatic Discharge

FR4 composite thermo-laminated glass epoxy material

GND ground

LNA low-noise amplifier

LO local oscillator

MEMS MicroElectroMechanical Systems

MIM metal-insulator-metal capacitor

PCB printed circuit board

RF radio frequency

SAW surface acoustic wave

SCL serial clock data input

SDA serial data input

SEN serial enable line

SMA SubMiniature version A connector

SPI serial peripheral interface

UHF ultrahigh frequencies, frequency range 300 MHz – 3 GHz

 vii

VDD voltage source

YIG yttrium iron garnet

 1

1. Introduction

1.1. Filters

Microwave filters are vital components which provide frequency selectivity in

wide variety of electronic systems, including mobile and satellite
communications, radar, electronic warfare, and remote sensing systems operating
at microwave frequencies. Type of the filter is defined with frequency, which it
allows to pass or reject. Lowpass, highpass, bandpass and bandstop filters with
different topologies exist to perform frequency selectivity. An example of
response of ideal bandpass filter is shown in figure 1.1.

Fig. 1.1. Response of ideal bandpass filter.

Historically, first were filters with discrete inductors (L) and capacitors (C) for
radio frequency (RF) application. Nowadays, lumped element filters are staying in
use at frequency range from 50 MHz till 2 GHz. The simplicity of design,
production and low cost are main advantages of using such filters. In addition,
they can often be very compact compared to filters based on half-wave resonators
structures (Fig.1.2) and they do not have "natural" eigenmodes that appear in
other filter technologies such as planar resonator structures. However, output
characteristics of lumped element filters suffer from low quality factor (Q) of
components. Despite of increase Q-values of modern inductors and capacitors up
to hundreds, they remain much below Q of thousands that cavity filters have.
 2

Fig. 1.2. Cavity filter (black) and small lumped-element (grey) filters
for 500 MHz [1].

The knowledge how to design discrete L and C passive filters is helpful for
synthesis of more complicated microwave filters [2].

1.2. Tunable filters

Due to the occurrence of multi-frequency bands in different regions and diverse

applications, requirement of tunable filters exists. It is significant for receivers and
transceivers operating in wide frequency spectrum to have maximal tuning range
and save filtering characteristics, while frequency is tuned.
One possibility is a switch with an array of individual filters (filter bank) which
have different frequency bands. This is quite simple approach, but needs a lot of
components and large space to have enough quantity of filters with fixed
frequency in an array. So, filter bank should consist of not less than 10 filters with
50 MHz bandwidth for smooth tuning in whole frequency range from 500 MHz to
1000 MHz. However, the quality factor of filters realised with fixed value
inductors and capacitors is better than Q of tunable elements. It allows filter banks
to have good performance.
 3

Fig. 1.3. Approaches of tuning center frequency in filter.

An alternative offer is to make one filter with tuning of central frequency

(Fig.1.3.). Mechanical frequency adjusting is possible with applying step-motors
in bulk resonator filters. It causes change the size of cavity that gives good output
characteristics, but such filters are too large, heavy and slow [3]. The best option
for mobile communication is electrically or magnetically tunable filters with
smaller size. One feasible variant of tuning is using of ferromagnetic resonance
based on yttrium iron garnet (YIG) [3], applying piezoelectric effect in surface
acoustic wave (SAW) and bulk acoustic wave (BAW) resonators, tuning of
capacitance [4]. First two of the approaches require special materials and
technology. Cost, size and complexity obstruct their implementation in RF
frontend board of handy device. The changing of capacitance has been very
explored and different types of variable capacitors are used now. They are rather
cheap, small and perform wide range of tuning.

1.3. Application of tunable filters for Cognitive Radio

The increasing number and capacity requirements of radio systems cause an

increasing demand for frequency spectrum. The Cognitive Radio (CR) offers a
tempting solution to this problem by proposing opportunistic usage of frequency
bands that are not occupied by their licensed users. The operation principle of a
CR circuit is to detect a particular spectrum that is currently in use and
immediately switch to an unused spectrum without interfering with the authorized
users. So, filters are needed for frontends and spectrum sensors in Cognitive
Radio devices. There are two frequency bands where a CR might operate in a near
 4

future: 400 - 800 MHz (UHF TV bands) and 3 - 10 GHz. In the lower UHF bands
almost every geographical area has several unused 6 MHz wide TV channels in
America or 8MHz wide in Europe, Asia and Africa. This frequency band is called
TV "white space" spectrum range and it is particularly appealed due to good
propagation properties for long-range communications [5].
The frontends for a Cognitive Radio could have architectures as depicted in figure
1.4 [6]. The first receiver circuit (Fig.1.4.a) has a RF wide bandpass filter to select
needed frequency range. Then the signal is reinforced by the low-noise amplifier
(LNA) and it goes into a notch tunable filter, which gives requiring bandgap,
slopes and losses in response for the following signal processing.

(a) (b)
Fig. 1.4. Simplified receiver architectures [6].

Also second structure of frontend (Fig.1.4.b) exists, where tunable filter is placed
before the amplifier to protect analog-to-digital (A/D) converter, LNA and mixer
with Local Oscillator (LO) from being over-driven by extraneous signals.
It is important to use widely tunable high-quality bandpass and low-loss bandstop
filters for RF receiver circuits.

1.4. Research Goals and Approach

Overall, the research on design of lumped element tunable filters is a currently

needed technical task that can be used in Cognitive Radio frontends and spectrum
sensors. Moreover, it performs quite useful goal for learning. Therefore, the
purpose of this work is to explore different types of topologies of bandpass filters
with quite narrow bandwidth, which is agile at whole 400 MHz – 800 MHz
frequency range. Due to considerably low operating frequency, it is expedient to
 5

design filter with lumped elements. The use of current capacitors and inductors
allows to make filter sufficiently compact, easy realisable and cheap.

Fig. 1.5. Ideal bandpass response for a tunable filter

used as a goal in this work.

The structure of this work can be represented as follows.

First, filter design and different methods of tuning is explored. This knowledge
became a basis for a new design.
Second, bandpass filters for required frequency with different topologies are
simulated to choose filter schematics that can be realised with current capacitors
and inductors. The compactness and possibility to tune future filter are under
consideration already. After simulation with real components and evaluating of
influence of losses, few bandpass filter topologies are chosen for realisation.
At the next step, tuning components are added to design and simulated to obtain
an agile central frequency with a constant bandwidth at 400 MHz - 800 MHz
frequency range. Then, layouts on the printed circuit board (PCB) of simulated
topologies with fixed and tuning components are prepared for the manufacturing
of filter prototypes. The attention is paid to minimise size and to provide the
control circuit for tunable components.
Finally, measurements of output response allow to compare simulation results
with reality and to show if presented filter designs are correct.
 6

2. Filter Design

Fundamentals and Design of LC passive bandpass

filters

In general, electrical performances of a filter are described in terms of the

insertion loss (IL), the return loss (RL), a frequency-selectivity (or attenuation at
rejection band), the group-delay variation in the passband, and so on. Filters are
required to have: small losses in the transmitted power (IL); a large reduction of
the reflected energy (RL) in a passband for proper impedance matching with
interconnecting components and a high frequency-selectivity to prevent any
interference. In the mechanical performance aspect, filters are needed to have
small volume, mass and good temperature stability. In cost and space parameters,
the filter order (quantity of LC resonators) has a considerable influence.
There are two methods of filter design. One was originated by Zobel and is known
as the image parameter method. The second was developed by Norton and
Bennett and is known as the polynomial method or insertion loss method [7]. In
the insertion loss method, a filter response is defined by a transfer function which
is the ratio of the output voltage to the input voltage of a filter and the goal is to
minimize the mismatch between the load and source resistors.
A second way to describe control of filter parameters is by the scattering matrix, if
a filter performs as the two-port network. In figure 2.1 incident currents and
voltages in both ports transformed to vectors a and b which have the transfer
matrix S. Components of this matrix relate to a measured transmission and a
reflection through a network and are called scattered parameters.
 7

Fig. 2.1. Two-port network and the scattering matrix [S].

Therefore, a filter response closed to ideal (Fig. 1.1) can be obtained by setting the
input reflection coefficient (S11) to the maximum and the forward transmission
coefficient (S21) to the minimum in a passband.
An insertion loss method deals directly with frequency responses and provides an
elegant solution to the approximation problem. Today, most of microwave filter
designs are done with the computer-aided design (CAD) based on an insertion
loss method.
A design by insertion loss starts from a mathematical description of the overall
desired response using approximation because ideal filter cannot be realized. An
acceptable approximation of the response has a magnitude function with
sufficiently small variation in a passband to be negligible and an attenuation that
is sufficiently large in a stopband. Four classical solutions to the approximation
are the Butterworth (maximally flat), Chebyshev (equal ripple in the passband),
inverse Chebyshev (equal ripple in the stopband) and elliptic (equal ripple in the
passband and stopband). The mathematically simplest and, consequently, the most
common approximation is Butterworth (Fig. 2.2.a), but it does not use the allowed
passband tolerance efficiently. In contrast, a Chebyshev filter (Fig. 2.2.b) has a
smaller transition band than Butterworth filter of the same order due to existing of
ripples. An elliptical is often a preferred approximation because it requires
significantly lower order than in the other types and does not provide more
attenuation than necessary in a stopband like Chebyshev and Butterworth filters
[8].
 8

(a) Butterworth passband response (b) Chebyshev passband response

Fig. 2.2. Butterworth and the Chebyshev approximations for response of a
bandpass filter with a different filter order n= 3, 4, 9 [8].

After choosing the approximation, a lowpass prototype should be synthesized

(Fig.2.3). The prototype means filter whose element values are normalized so that
the source resistance or conductance equals to one and is defined as g0 = 1. In
figure 2.3 are shown two identical forms of N-elements lowpass filter prototypes,
where the first element g1 is either a shunt capacitor C1 (Fig. 8.a) or a series
inductor L2 (Fig. 8.b).

(a) (b)

Fig. 2.3. Lowpass prototypes: (a) Pi-section and (b) T-section types.

Using of frequency transformations of lowpass response, highpass, bandpass, and

bandstop filters can be further designed.
Simple lowpass filter topologies (Fig. 2.3) can be converted to inverter coupled
architectures. They have identical properties but only one type of coupling
elements: a series L or a shunt C (Fig. 2.4). Inverters are two port networks which
transform the inductive impedance or the admittance into conductive ones and
vice versa with the invariance of frequency. In figure 2.4 they are depicted as
 9

boxes with the K label. In practice, inverters can be performed as LC circuits with
the negative value of L or C that can decrease a value of nearest components [9].

(a)

(b)
Fig. 2.4. Lowpass prototypes with inverters [9].

Because of a steeper filter response (Fig. 2.2) and better tolerance of components,
the Chebyshev passband and stopband filters are shown later as examples of a
design.
Extra control in a design of Chebyshev filter is the choice of the maximum loss in
a passband, referred to as the ripple. This parameter is usually specified in
decibels. The factor needed for transfer response is usually labelled as ε [8].
The insertion loss (IL) at ripple level can be expressed as
. (1)
The response depends on ripples and Chebyshev polynomial , which must
be obtained with the maximum value of number =1 at the maximum
number of points inside a passband . Therefore, a response and a
polynomial function should proceed as:
, (2)

. (3)
After adding a condition to the equation (1) the insertion loss equation will follow
expression
. (4)
 10

The result is that a transfer function appears as a polynomial of order N, where N

is a filter degree or the number of elements of a lowpass filter prototype. It can be
defined by use of required filter parameters.

, (5)

where is a stopband insertion loss, is a passband return loss and S is a ratio

of stopband to passband frequencies.
The solution gives g values of the normalised components which allow to obtain
needed response.
The filter order, g and finally the inductance and capacitance values can be
calculated by using filter parameters: a center frequency, a bandwidth, an insertion
loss, rejection requirements, and value of ripples. Second approach is to use the
tabular data.
Formulas to define g-values:
, (6)
, (7)

, , (8)

. (9)

Auxiliary equations:

, (10)

, (11)

, (12)

. (13)

To find elements values with using tables, it is required to know the degree of
filter (N) and a passband ripple (ε). g-values for filters with 0.01 dB and 0.1 dB of
passband ripple are provided as examples in table 1. It is clear from the tabular
data, that Chebyshev lowpass prototypes are symmetrical for an even filter
degree.
 11

Table. 1. Element values for Chebyshev lowpass filter prototypes [10].

Next step is frequency transformation because prototype filter has normalised

cutoff frequency equal to one.
For a bandpass filter it is as following:
, (14)

. (15)

where is a center frequency, is a minimum and is a maximum frequency

points of passband range of bandpass filter. BW is bandwidth and is a
bandwidth scaling factor.
Hence, reactive elements L and C will be transformed from lowpass values using
simple equations. L elements transform to series connected L’ and C’ (series LC-
resonator) by:
, (16)

. (17)
 12

C elements transform to parallel connected L’ and C’ (parallel or shunt LC-

resonator) by:
, (18)

. (19)

The bandpass filter, which was derived from the lowpass prototype filter by
introducing complementary elements and by producing shunt and series
resonators, is shown in figure 2.5.a. The alternating topology can be presented by
using only one type of a resonator between inverters (Fig. 2.5.b). These are basic
lowpass to bandpass transformations, and sometimes they lead to component
values, which are not readily available or have excessive losses.

(a) (b)
Fig. 2.5. Lowpass to bandpass transformation:
(a) ladder (b) inverter-coupled [9].

Inverters can be represented as a Pi-type network with capacitors (Fig. 2.6) and as
a capacitively-coupled bandpass filter (Fig. 2.7). However, it should be used in a
narrowband bandpass filter due to the frequency dependence of such inverter
transformation.
 13

Fig. 2.6. Schematic of inverter for a narrowband bandpass filter [9].

Fig. 2.7. Capacitively-coupled narrowband bandpass filter schematic [9].

The topology of bandpass filter may vary by choosing different types of lowpass
prototypes, but it is preferably to minimise numbers of inductors. It connects with
high losses in inductors because quality factor is lower than Q of capacitors.
Moreover, inductors, especially aircore ones, increase the size of filter in PCB.
 14

3. Tunable filters
Due to demand of tunable filters for broadband frontends and other applications,
several methods exist to accomplish this purpose. In this chapter are analysed
scientific papers for better understanding a current state of the issue.

3.1. Tuning methods

Tuning in frequency range of the simplest single resonator of lumped-element

filter demands a change of component values to shift a resonance frequency. For
filters with an order higher than one all resonator's elements should be changed. In
addition, coupled elements should be also tuned to keep the bandwidth of a filter
the same.
The main approach to tune frequency electrically is to change the capacitance of a
filter. The reason is that tuning of coils is more complicated. Nowadays, different
types of tunable capacitors are developed and can be successfully applied in agile
filters. There are varactor diodes, ferroelectric capacitors, digitally switchable
capacitor banks, and RF MEMS (MicroElectroMechanical Systems).
Varactor diodes have a widely spread application. The concept used in such type
of tunable capacitors is based on changes in the depletion region width of the
semiconductor diode and resultant changes in a capacitance. The reverse-bias DC
voltage utilizes to effect this change. Although, varactor tunable filters have a
small power consumption and a relatively fast tuning speed, they suffer from a
moderate Q value of resonators. Other varactor’s drawback is power handling.
Since varactor diodes are originally non-linear devices, the large input signal
generates unwanted nonlinear distortions [11].
Ferroelectric BST material has a high Q and a big tuning range; thus, it is a good
candidate for tunable varactors. Used in the paraelectric phase, BST exhibit large
dielectric constant which can be changed by an external bias voltage.
Ferroelectrics can be fabricated on a variety of substrates using standard
semiconductor manufacturing processes. So, planar tunable BST capacitors can be
 15

integrated directly into a filter PCB circuit [12]. Second advantage is quite
compact size due to a high value of the material permittivity.
Varactors are nonlinear and they need the voltage control circuit which makes the
filter design more complicated. Capacitor banks handle with such problems and
tune a filter digitally. Their harmonic performance and capacitance tuning range
are better than those of varactors [13]. The DTCs suffer from a low quality factor
of switches despite of a high Q of the MIM fixed capacitors.
Among these variants, the RF MEMS components can be employed in tunable
filters. In these devices an applied DC voltage can result in a change of the
capacitance by employing micrometer level movements. MEMS capacitors have
small size and zero power consumption. Other advantages are that they provide a
low insertion loss and show a wide tuning range. However, a RF MEMS suffer
from a considerably shortened life cycle in contrast to components steered purely
electronically. In addition, they are still not easily available on the market.
Depending on the technology, controlling a RF MEMS component requires an
analog voltage and, consequently, a proper control circuit [14].
To summarize, each option has own advantages and limitations. Therefore, in
further are presented examples of filters which use these types of tunable
capacitors in a desirable 400 - 800 MHz frequency range.

3.2. Examples of tunable filters

All mentioned tuning technologies have examples of the implementation in a filter

design at UHF band. They are found in scientific papers and presented below.
Successful example in UHF was obtained with 6 pairs of varactors by researcher
of Motorola Labs [15]. The tuning range is 476MHz - 698 MHz and the insertion
losses vary between 4.7 dB to 6.9 dB (Fig. 3.1).
 16

(a)

(b)

Fig. 3.1. Varactor-based tunable UHF filter (a) Schematic

(b) Measured insertion loss response [15].

Second excellent design of a filter with varactors is performed by researchers

from Kwangon University’s Micro/Nano Devices and Packaging Lab [16]. They
offer the highly compact bandpass filter with components embedded into the
eight-layered package substrate (Fig. 3.2).
 17

(a) (b)

Fig. 3.2. Circuit schematic (a) and view of structure (b) for the compact multi-
layered tunable filter design [16].

This multilayered bandpass filter operates in 500 - 900 MHz range if a bias
voltage of varactor diodes is changing from 1 V to 25 V (Fig. 3.3).

(a) (b)
Fig. 3.3. Measured S-parameters of compact multi-layered tunable filter [16].

Paper [17] shows compact 3-order inductively coupled bandpass filter with BST
thin film tunable capacitors (Fig. 3.4.a,b). It provides tunability at 490 - 790 MHz
range with acceptable output parameters (Fig. 3.4.c,d). Losses differ from -1.94
dB at the high edge of tuning range to -2.85 dB at the low edge.
 18

(a) (b)

(c) (d)

Fig. 3.4. Topology (a) and view on board (b) of the 3-order ferroelectric varactor-
based tunable bandpass filter with its measured
insertion losses (c) and return losses (d) [17].

The up-to-date offer of tuning with capacitor banks is a use of three 5-bit digitally
tunable capacitors (DTCs) to cover frequency range from 470 MHz to 860 MHz
[18]. The topology consists of three series LC resonator coupled by shunt
inductors (Fig. 3.5).

(a)
 19

(b)

Fig. 3.5. Schematic (a) and board view (b) of 3-order tunable bandpass filter with
DTCs [18].

Capacitor values vary from 0.6 pF to 4.6 pF and a center frequency is tuned from
470 MHz till 860 MHz. However, the insertion loss in a passband is rather high
and differs from 7.5 dB to 7.7 dB (Fig. 3.6). The main reason of such
characteristic is the influence of a quality factor of filter elements.

Fig. 3.6. Measured forward transmission response of tunable bandpass filter with
DTCs [18].

Electrostatically tuned membrane MEMS capacitor (Fig. 3.8.a,b) has a wide

tuning range 5:1 and Q greater than 100 in frequencies up to 1 GHz. It allows to
have the insertion loss at rate of 3 - 3.6 dB and a steep slope with 87 - 139 MHz
bandwidth (Fig. 3.8.c) [19].
 20

Fig. 3.7. Inductively coupling bandpass filter topology [19].

(a) (b)

(c)

Fig. 3.8. Picture of MEMS elements (a) and layout (b) of MEMS-based tunable
bandpass filter and (c) measured filter response [19].

All previous designs of filters successfully tunable in UHF band are presented in
table 2.
 21

Table. 2. Comparison of different tunable bandpass filter designs

operating at 400 - 800 MHz range.

3rd order
inductively
filter 3rd order inductively 3rd order 3rd order
coupling
configuration filter coupling filter filter
curcuits
filter
Bandwidth,
30 105-181 100-200 40 81-139
MHz
IL, dB 4.7-6.9 1.83-2.45 1.7-2.9 7.9±0.15 3.0-3.6
RL, dB - 10-25 >25 8-12.5 >15
thinfilm BST capacitor
tuning type varactors varactors MEMS
capacitors banks
tuning range,
476-698 510-910 490-790 450-940 600-1000
MHz
multilayered coupled
features - - -
PCB inductors
3.4x4.4x0.5
size - - 20x20 mm 15x11 mm
mm

It is clear from table that tunability causes low values of an insertion loss due to
additional losses in tuning components. Also it increases complexity of the design
to manage tuning elements. So, designs with a multilayered PCB and MEMS
structures need a high quality, expensive technology that can be laborious in a
fabrication and an implementing in RF frontend.
After weighing all advantages and disadvantages, digitally tunable capacitor
banks were chosen as tunable elements for the design. Commercially available
DTCs are small in size, easy-implemented in PCB design, have a rather large
tuning range, excellent linearity and power handling, low cost and they do not
need any complex technology.
 22

4. Tunable filters simulations

4.1. Introduction
In this work are investigated topologies of bandpass tunable filters for TV “white
space” spectrum tuning range.
In 400-800 MHz frequency range, lumped elements can be successfully used for
filters. Current capacitors and inductors have quite good parameters (high Q, wide
range of values, size) for a design of filters which are suitable for these
frequencies. Different topologies allow to choose values of components and sizes
to obtain a compact filter.
Filter schematics, which are appropriate for tuning, is the goal of this research.
First, filters with ideal components were under consideration to find topology with
values of capacitors, which can be changed by using digitally tunable capacitor
banks. Then influence of Q-factors of components was taken into the account.
This becomes the key factor when choosing a bandwidth of a filter. Finally, few
filter topologies were simulated with tuning by capacitor banks and performed in
a layout schematic for future realization.

4.2. Topologies of filter with ideal components

Many specified computer programs enable to design filters from their parameters.
In this work Agilent Genesys and AWR Microwave Office software were used.
Chebyshev 30 MHz bandwidth bandpass filter topologies were simulated for 400,
600 and 800 MHz to compare response and needed values of capacitance and
inductance. Second and third-order filters were considered to limit numbers of
elements to prevent complexity and big size of filter.
In figure 4.1 is shown 3rd order filters with 30 MHz BW for 600 MHz. Element
values are calculated and optimized to satisfy the requirements with using
insertion loss method in Agilent Genesys 2010.5 program.
 23

Schematics with a series LC-resonator (Fig. 4.1.a) and with a shunt LC-resonator
(Fig. 4.1.b) at first left places are bandpass topologies which are obtained with the
transformation from a T-type (Fig. 1.8.a) and a Pi-type (Fig. 1.8.b) lowpass
prototypes. They have a quite big inductance and small capacitance in series
resonators but in shunt resonators the situation is vice versa. Inverter-coupled
lowpass filter prototypes (Fig. 1.9) give C-coupled (Fig. 4.1.c) and L-coupled
(Fig. 4.1.d) shunt resonators filter topologies. Also, the capacitively coupled series
resonance filter (Fig. 4.1.e) can be constructed with the same way. They are
narrowband approximations of a bandpass filter and have desirable component
values in this frequency. The last topology (Fig. 4.1.f) is called tubular and
consists of alternating series of inductors and capacitors which is connected
capacitively grounded.

(a) T-type 3rd order filter (b) Pi-type 3rd order filter

(c) Top C-coupled 3rd order filter (d) Top L-coupled 3rd order filter

(e) Shunt C-coupled 3rd order filter

(f) Tubular 3rd order filter

Fig. 4.1. Topologies of the bandpass 3rd order 30 MHz BW filter at 600 MHz.
 24

Fig. 4.2. Simulated responses of 3rd order bandpass filters with different
topologies using ideal components.

All presented topologies have good output response (Fig. 4.2) but values of
components are sufficiently differ. So, in first two topologies capacitance values
are too small to be realizable and inductance values are high meaning that the
overall size of a filter will be large. Moreover, it is better to choose topologies
with grounded inductors to diminish the influence of a parasitic parallel
capacitance appearing with an inductance (Fig. 4.3).

Fig. 4.3. Equivalent circuit model of real soldered inductor.

To sum up, after investigation of topologies with attention to mentioned factors

the top C and the shunt C-coupled resonator filter schematics are the most suitable
 25

for the design of a lumped-element bandpass tunable filter at 400 MHz - 800 MHz
frequencies. Moreover, the change of capacitance values in these topologies are
less than 30 pF in a whole tuning range which can be covered using up-to-date
capacitor banks [13].
Also, a completely different topology to obtain small bandwidth was considered.
The peculiarity of this was very steep slope on one side of a response. The
approach is a transformation of a LC-resonator in such way to locate of the
transmition zero of a filter closely to the main resonance. It can be obtained by
replacing a typical shunt LC-resonator by a structure with one additional shunt
resonator in series [20]. The schematic of such notch filter with 10 MHz
bandwidth are shown in figure. 4.4.

Fig. 4.4. Schematic of notch bandpass filter.

Fig. 4.5. Simulation of a response for a notch bandpass filter.

 26

The S-parameters (Fig. 4.5) are simulated for ideal components with values
optimized for this frequency range. Big advantage of this topology is that the
bandwidth remains constant during tuning of the main resonance by change of
only one capacitor.
However, components with high Q are needed to realize approach that is
impossible with a current state of lumped elements losses at 400 MHz - 800 MHz
frequency range.

4.3. Simulations with accounting influence of losses

After the previous step, topologies (Fig. 4.1.c and e) appropriate for tuning with
capacitor banks should be considered with the losses impact. Up-to-date inductors
and capacitors have Q factors much less than the ideal ones. Taking this into an
account, the response of chosen topologies needs to be checked. To make it, the
second- (Fig. 4.6) and the third-order (Fig. 4.7) bandpass filters were simulated
with Q = 100, 200, 1000 and 10000 for four types of a bandwidth: 10 MHz, 30
MHz, 50 MHz and 100 MHz at AWR Microwave Office.

Fig. 4.6. Simulated S-parameters of 2nd order bandpass filter with different Q-
values of components.
 27

Fig. 4.7. Simulated S-parameters of a 3rd order bandpass filter with different Q-
values of components.

As shown in figures 4.6 - 4.8 simulations show that a design of a filter is the
balance between the Q factor, the bandwidth and order of the filter. So, with
current components (the average Q = 100 - 200 in UHF spectrum) only filters
with BW more than 50 MHz are reasonable to realize (Fig. 4.8). These
conclusions correlate with the equation which connects all mentioned parameters
and represents, that insertion losses are proportional to an order of a filter and
inversely proportional to Q and a passband bandwidth

, (20)

where IL is a midband insertion loss, f0 is a central frequency, is a bandwidth,

Qr is a resonator quality factor, and depends on an order of filter [9].
28
 29

To conclude, 2nd and 3rd order capacitively coupled and 3rd order shunt coupled
bandpass filter topologies with 50 MHz bandwidth are reasonable variants for
further work (Fig. 4.9).

Fig. 4.9. Chosen topologies of a bandpass filter: (a) 2nd order and
(b) 3rd order C-coupled filter, (c) 3rd order shunt-coupled filter topology.

4.4. Filter simulations with real components

Nowadays, special series of capacitors and inductors with advanced parameters

for RF application are commercially available. The highest values of quality
factors for lumped elements at 400 - 800 MHz belong to aircore inductors (Q up
to 230) [21] and chip capacitors (Q up to 400) [22].
 30

(a)

(b)

Fig. 4.10. Dependence of Q-factor with frequency for (a) Johanson Technology
capacitors [22] (b) Coilcraft minispring inductors [21].

However, capacitor banks have sufficiently more losses due to switches, average
value of Q is 40 but depends on the state and frequency, as shown in figure 4.11
[13].

Fig. 4.11. Measured shunt Q dependence on frequency for different states of

digitally tunable capacitor [13].
 31

To make a simulation closer to reality, components in chosen topologies were

replaced with equivalent models. They account losses, the dependence of a
frequency and correspond to real parameters provided by component
manufacturers (Coilcraft, Johanson Technology).
As an example, simulated S-parameters of a 3rd order bandpass topology are
shown in figures 4.11. Due to discrete range of available capacitance values, the
output response became less smooth and asymmetrical. It proves advantages of a
continuously tunable filter which does not require any high tolerance of
components.

Fig.4.12. S-parameters of 2nd order bandpass filter with real components.

 32

4.5. Filter simulations with tunable elements

In this work Peregrine digitally tunable capacitors were chosen as tunable

elements. They are based on the Peregrine’s UltraCMOS® technology, they have
linear capacitance change with switching of 32 tuning state (5 bits) and an
excellent harmonic performance compared to devices made of varactors. The
equivalent circuit model is shown in figure 4.13. It is visible, that the capacitance
is different for a series and a shunt configuration. For instance, in the DTC with
part number PE 64904 in a series connection, a capacitance can change from 0.6
pF till 4.6 pF. However, in a shunt configuration, a capacitance is calculated as the
sum of CP1 and CP2 and in this case the minimum value is 1.1 and the maximum is
5.1 [23].

Fig. 4.13. Equivalent circuit model schematic of a DTC [23].

Certain items of Peregrine’s DTCs appropriate to cover UHF frequency band

were defined with checking capacitor values for simulated filters at 400 MHz and
800 MHz. It appeared that the PE 64102 model [13] with tuning range between
1.88 pF - 14 pF in a shunt configuration and the PE 64904 model [23] with tuning
range between 0.6 pF - 4.6 pF in a series configuration are satisfied for the design.
To add capacitor banks, topologies should transform in the following way shown
at figure 4.14. Top coupled capacitors are replaced by PE 64904 DTCs. For
parallel capacitors, due to large capacitance values, they are performed as a
parallel connection of PE 64102 and capacitors with a fixed value. Five capacitor
banks are used for a 2nd order and a 3rd order shunt coupled filters and seven for a
3rd order filter topology. Use of agile coupled components allows keeping a
bandwidth of a filter at 50 MHz rate during tuning.
 33

(a)

(b)

(c)

Fig. 4.14. Topologies of tunable bandpass filters with capacitor banks:

(a) 2nd order and (b) 3rd order C-coupled filter,
(c) 3rd order shunt-coupled filter topology.

The distinctive feature of the last topology (Fig. 4.14.c) is that it requires less
capacitor banks as other 3rd order filter designs but needs three inductors with
rather high values of 35.5 nH and 43 nH. These are bigger in size and have less Q,
compared with 5 nH inductors for the rest chosen topologies. However, using of
the same capacitor banks PE 64904 for series and PE 64102 for shunt connections
allows to tune the center frequency by small changes in resonator capacitors. This
is an advantage of this scheme (Fig. 4.17).
Simulation results are presented below and they confirm a possibility of cover TV
"white space" spectrum range with using of chosen Peregrine digitally tunable
capacitors.
 34

Fig. 4.15. S11 and S21 for 2nd order tunable filter with 5 capacitor banks.

Fig. 4.16. S11 and S21 for 3rd order tunable filter with 7 capacitor banks.
 35

Fig. 4.17. S11 and S21 for 3rd order shunt coupled tunable filter
with 5 capacitor banks.

These graphs show that chosen topologies of a bandpass filter tuned with
capacitor banks provide insertion losses less than -5.7 dB and return losses better
than -10 dB in 50 MHz bandwidth over a whole tuning range. In schematics a
minimum of insertion loss is achieved at a low frequency edge. It agrees with the
theoretical prediction that IL should rise with an increasing of a central frequency
in a case of a constant bandwidth (Fig. 4.11).
The tuning range of these topologies is limited by a capacitance change of PE
64102 capacitor banks. The simple decision to have wider spectrum range is
replace it with a parallel connection of such DTCs or use a model with larger
tuning range.
The response of a shunt-coupled filter topology proved that it can be tuned by a
small change of a resonator capacitance. In addition, it has a small forward
transmission coefficient in a quite narrow bandwidth, meanwhile, insertion losses
of capacitively-coupled schematics increase, if coupling is too small.
 36

4.6. Realisation of bandpass tunable filter prototypes

To realise bandpass filter prototypes their PCB layouts were drawn in the PCB-
CAD software which is called Easy-PC by Number One System program.
The main factor of the design was the necessity to control digitally the tunable
capacitors by three wire SPI compatible interface. The SEN (serial enable), the
SCL (serial clock data input) and the SDA (serial data input) lines serve to adjust
a capacitance by controlling CMOS FETs that connect and disconnect a fixed
high-Q MIM capacitors inside a capacitor bank.

(a) (b)
Fig. 4.18. Peregrine’s digitally tunable capacitor pin configuration (topview)
(a) PE 64102 [13] and (b) PE 64904 [23].

Chosen Peregrine capacitor banks PE 64904 and PE 64102 have different pin
configuration but the same driving interface that allows to connect common lines
as SCL, SDA, and VDD into buses. However, each of DTCs need their own line
for a serial data input (SEN). The ChipKit max 32 by Digilent (Fig. 4.19) was
chosen as a core of a controlling system due to an easy programming and an
existing of the SPI interface. This microcontroller board is suitable for the
stacking filter PCB directly on the top of a ChipKit and to use metal rods for
connection and maintenance of both boards. Second advantage of a ChipKit max
32, in case of such type of a stacking, is that it has a voltage source among the
digital ports that allows the using close pins and make a filter board compact.
 37

Fig. 4.19. Top view of ChipKit max 32 board [24].

Due to considerably large amount of wires to control capacitor banks, the 2 layers
FR4 PCB board with thickness of 1.6 mm is used in the design. The inner layer
serves for the routing driving signals.
The whole filter PCBs of all topologies occupy the right part of a microcontroller
board and have dimensions as 77.5 mm x 35.9 mm, that also include places for the
SMA connectors.

Fig. 4.20. Top view of the layout of a 2nd order tunable bandpass filter
with capacitor banks.
 38

Fig. 4.21. Top view of the layout of a 3rd order tunable bandpass filter
with capacitor banks.

Fig. 4.22. Top view of the layout of a C-shunt 3rd order tunable bandpass filter
with capacitor banks.

Layouts represented in figures 4.20 - 4.22 have the same pins as the control
ChipKit max 32 board (Fig. 4.19) and coplanar grounded waveguide lines in input
and output. These lines are made with equal length in all topologies to make them
more identical to compare.
Moreover, the calibration line with two SMA connectors linked by coplanar
grounded waveguide lines are used to exclude losses of signal from the filter
response.
 39

Fig. 4.23. Top view of the layout of a calibration line.

Designed layouts were realized at a 2 layered FR4 board by the PCB

manufacturer, and all passive components were soldered in the laboratory. Next
step was a stacking filter circuits with a microcontroller board by pins. In figure
4. 24 is shown filter board which was ready for measurements.

Fig. 4.24. Top view of a realised 2nd order bandpass filter circuit stacked with a
microcontroller board to drive DTCs.

The microcontroller program to control tunable capacitors provides a comfortable

interface to manually change DTC states from 0 to 31 with the control panel.
Since designed topologies are symmetric, it is not required to tune all the DTCs
separately. So, resonator capacitors at the edge of filters can tune simultaneously
and form one tuning group. Also, coupling elements and edge capacitors can be
tuned in such way. Thus, a 2nd order filter required only three driving groups, a 3rd
order needed 5 groups. For a shunt-coupled topology, DTCs in a first and a
 40

second resonator should be managed separately and the control board is arranged
for 4 tuning groups.

Finally, after preparing circuits and controlling facilities, proposed tunable filters
become ready for a work and measurements.
 41

5. Measurement results
The actual goal of this work was to develop functional lumped-element tunable
filters which can be used in Cognitive Radio receivers or spectrum sensors. To
design filters that will be further implemented in different devices, it is important
to check the correctness of the simulations. Therefore, the performance
measurement of filter prototypes is significant part of filter development.
Measurements of three realised filter schematics were made with the HP 2-ports
network analyser, which allows to see S-parameters at a wide range of
frequencies. In figures 5.1 - 5.3 are presented measured responses of designed
topologies.

Fig. 5.1. Measured S11 and S21 for a 2nd order tunable filter with
5 capacitor banks
 42

Fig. 5.2. Measured S11 and S21 for a 3rd order tunable filter with
7 capacitor banks

Fig. 5.3. Measured S11 and S21 for a 3rd order shunt-coupled
tunable filter with 5 capacitor banks

The capacitance of DTCs was changed manually with a control panel to choose
required 50 MHz bandwidth.
Measurements revealed that in the 2nd order bandpass tunable filter topology
insertion losses change from -3.6 dB to -4.6 dB in 370 MHz - 570 MHz tuning
range while DTCs capacitance alter from a minimum to a maximum values. The
3rd order capacitively coupled bandpass filter is agiled at the same spectrum band
 43

and has -4 dB at 383 MHz and -4.8 dB at 576 MHz. Their out-of-band
suppression is better than the previous topology has. At the last circuit schematic
(Fig. 5.3) S21-parameters differ from -3.6 dB to -4.5 dB. In terms of a shape of
response and a tuning range, this topology cede to the rest.
It is evident from figures 5.1 – 5.3 that frequency range downshifts in comparison
with simulated (Fig. 4.14 - 4.16) for all topologies. So, the high edge of the
measured response is less than 600 MHz. The left side of a tuning range also
shifts from 400 MHz to lower frequencies. The reason can be that real aircore
inductors have a value higher than 6 nH. This decreases central frequencies of
resonators. However, great contribution in downshifting belongs to the parasitic
capacitance. The proof of this suggestion is that the high side of the frequency
range moved more than the low edge. The manufactured boards have a gap
between an edge and a source line that appears as a gap capacitor at higher
frequencies. In addition, parasitic capacitances to the ground at component pads
exist. These things can be prevented by a proper electromagnetic simulation of the
board which will take into account the possible appearance of such parasitic
elements as well as losses in a transmission line. Measurements of the calibration
of connectors (Fig. 4.23) showed that they do not add significant losses to the
output response.
Measured values of insertion losses at the edge of tuning range are considerably
close to simulations or they differ less than 1 dB. The increase of losses in realised
filters can be caused by a parasitic resistance at element interconnects and wiring
because the soldering is imperfect.
The benefit of proposed tunable filters is the flexibility that gives wider variety of
applications. It is possible also to simply tune the bandwidth of filters by the
change of a coupling capacitance. In figure 5.4 is shown measurements of 50
MHz and 100 MHz bandwidth for 3rd order tunable filter. The increase of
bandwidth causes the enhance of insertion losses. So, the filter with 100 MHz
bandwidth has S21 = -2.7 dB which is less in twice than losses in 50 MHz
bandwidth filter. This result totally corresponds with the equation 20.
 44

Fig. 5.4. Measured response of 50 MHz and 100 MHz bandwidth 3rd order tunable
bandpass filter.

To sum up, these measurements show that bandpass filters have an appropriate
rate of losses, a quite stable 50 MHz bandwidth and steep descents during tuning.
These allow to imply designs of the offered filters for a Cognitive Radio
application. In addition, bandwidth of filters can be tuned. This is a useful tool in
filtering devices.
 45

6. Conclusions

6.1. Summary

In this work a design of tunable bandpass filters at UHF band was under a
research. One promising application of such devices is in frontends and sensors
for the Cognitive Radio. The main goal was to development sufficiently compact,
low cost tunable filters with a quite narrow bandwidth using currently available
lumped-element components and a PCB board technology. Filter design, different
topologies and methods to tune bandpass filters were explored to choose the best
suitable variant to comply with the required purpose.
It appeared, that the 2nd and the 3rd order of capacitively-coupled and the 3rd order
of shunt-coupled topologies of a bandpass filter have realisable values of
capacitors and inductors and appropriate output characteristics. They were
simulated with taking into account the influence of losses in real components.
Digitally tunable capacitors (DTCs) are used to adjust a central frequency at
whole TV "white space" spectrum.
Measurements revealed that the presented schematics have proper output response
and filters are successfully tuned by DTCs. But the frequency range sufficiently
downshifts in comparison with simulation for all topologies. This can be
prevented by a proper electromagnetic simulation of a filter board which will
calculate a possible appearing of parasitic elements as well as losses in a
transmission line.
Measured values of insertion losses have good agreement with simulated ones.
Difference between them is less than 1 dB. The increase of losses in real filters
can be caused by parasitic resistance at element interconnects and wiring because
the soldering is imperfect.
In comparison with designs of filters presented in scientific papers, tunable filters
of this work have several benefits. The first is a constant and relatively small
bandwidth during tuning. Second advantage is that the range of insertion losses is
3.6 - 4.8 dB. It is less than 7.9 dB, which other filter that also used Peregrine
digitally tunable capacitors has [18]. Moreover, if bandwidth is wider than 50
MHz, losses in proposed filters can be smaller than 3 dB.
 46

To summarize, considering the acceptable level of losses, the constant bandwidth

during tuning, small dimensions and cost of realised filters, all topologies are suit
for a Cognitive Radio application. The digital tuning type perfectly corresponds
with the CR agenda of the spectrum-defined radio.

6.2. Future work

Because the frequency range in prototypes considerably shifts, an electromagnetic

simulation of filters on the board and more careful choice of equivalent models of
elements should be made to account all possible parasitics. It will allow to correct
component values to obtain more agreement with measured and simulated results.
A simple decision can be used to provide tuning in whole TV "white space"
spectrum. In capacitively-coupled bandpass topologies, the resonator capacitor
should be replaced by parallel connections of several PE64102 DTCs, the edge
and the coupling PE64904 capacitors should be changed by DTCs with bigger
capacitance range. It will not increase the complexity and the cost of filters in a
large rate.
Filter schematics are independent of frequency and can be recalculated to higher
spectrum for other applications. Digitally tunable capacitors allow operation from
100 MHz to 3000 MHz. However, if frequency will be more than 1 GHz, filter
should be constructed with other technology, for example with the use of
microstrip lines or bulk resonators. The reason is that the quality factor Q and
parasitics do not allow to implement lumped-elements at GHz range.
The actual filter size is 35x12 mm for a 3rd order capacitively-coupled topology
and 30x27 mm for a shunt-coupled schematic, but they can be even reduced by
the decreasing of spaces between components. The implementing of proposed
filters at the receiver circuit can require changes in layouts.
 47

References

[1] Anatech Electronics Inc., 2009. Guidelines for choosing RF and Microwave
products, [pdf document]. [Accessed January 2013]. Available:
http://www.anatechelectronics.com/products/large/CMS/Files/Anatech%20Filter
%20Design%20Guide3.pdf

[2] G.L. Matthaei, L. Young, E.M.T. Jones, 1980. Microwave Filters, Impedance-
Matching Networks, and Coupling Structures. Dedham, Massachusetts: Artech
House Inc.

[3] J. Uher and W. J. R. Hoefer, 1991."Tunable microwave and millimeter-wave

band-pass filters", Microwave Theory and Techniques, IEEE Transactions on, vol.
39, pp. 643-653.

[4] K. Hashimoto, S. Tanaka and M. Esashi, 2011. "Tunable RF SAW/BAW

filters: Dream or reality?" in Frequency Control and the European Frequency and
Time Forum (FCS), 2011 Joint Conference of the IEEE International, pp. 1-8.

[5] M. R. Dzulkifli, M. R. Kamarudin and T. A. Rahman, 2011. "Spectrum

occupancy at UHF TV band for cognitive radio applications," in RF and
Microwave Conference (RFM), 2011 IEEE International, pp. 111-114.

[6] D. Cabric, S. M. Mishra and R. W. Brodersen, 2004. "Implementation issues

in spectrum sensing for cognitive radios”, Signals, Systems and Computers, 2004.
Conference Record of the Thirty-Eighth Asilomar Conference on, 7-10 Nov.,
Vol.1, pp. 772-776.

[7] A. I. Zverev, 1967. Handbook of Filter Synthesis. New York: John Wiley and
Sons.

[8] G. D. Vendelin, A. M. Pavio, Rohde and Ulrich L., Eds., 2005. Microwave
Circuit Design using Linear and Nonlinear Techniques. John Wiley & Sons,
pp.273-289

[9] I. C. Hunter, 2001. Theory and Design of Microwave Filters. England: The
Institution of Electrical Engineers, pp. 49-99, 107-116, 125-131.

[10] J. -S. Hong and M. J. Lancaster, 2001. Microstrip Filters for RF/microwave
Applications. New York: Wiley, p.43.

[11] K. E. Mortenson, 1974. Variable Capacitance Diodes: The Operation and

Characterization of Varactor, Charge Storage and PIN Diodes for RF and
Microwave Applications. Dedham, Massachusetts: Artech House Inc.
 48

[12] Manh-Tai Nguyen, W. D. Yan and E. P. W. Horne, 2008. "Broadband

tunable filters using high Q passive tunable ICs", Microwave Symposium
Digest,2008 IEEE MTT-S International, 15-20 June, pp. 951-954.

[13] Peregrine Semiconductor Corp., 2013-last update, 64102 UltraCMOS®

Digitally Tunable Capacitor [Datasheet]. [Accessed February 2013]. Available:
http://www.psemi.com/pdf/datasheets/PE64102_70-0428-01.pdf

[14] R. L. Borwick III, P. A. Stupar, J. F. DeNatale, R. Anderson and R.

Erlandson, 2003. "Variable MEMS capacitors implemented into RF filter
systems," Microwave Theory and Techniques, IEEE Transactions on, vol. 51, pp.
315-319.

[15] N. Tanzi, 2006. "Varactor-tuned coupled resonator front-end bandpass filters

for cognitive radio applications", Radio and Wireless Symposium, 2006 IEEE,17-
19 January, pp. 155-158.

[16] T. C. Lee and J. Y. Park, 2009. "Compact PCB embedded tunable filter for
UHF TV broadcasting", Microwave Symposium Digest, 2009. MTT '09. IEEE
MTT-S International, 7-12 June, pp. 505-508.

[17] M. K. Roy and J. Richter, 2006. "Tunable ferroelectric filters for software
defined tactical radios", Applications of Ferroelectrics, 2006. Isaf '06. 15th Ieee
International Symposium on, 30 July – 6 Aug., pp. 348-351.

[18] A. Jaschke, M. Tessema, M. Schuhler and R. Wansch, 2012. "Digitally

tunable bandpass filter for cognitive radio applications", Computer Aided
Modeling and Design of Communication Links and Networks (CAMAD), 2012
IEEE 17th International Workshop on, 17-19 Sept., pp. 338-342.

[19] Yonghyun Shim, Zhengzheng Wu and M. Rais-Zadeh, 2012 "A High-

Performance Continuously Tunable MEMS Bandpass Filter at 1 GHz" Microwave
Theory and Techniques, IEEE Transactions on, vol. 60, pp. 2439-2447.

[20] W. Pöhlmann, 1966. "Variable Bandpass Filters with Tracking Poles and
Attenuation", Rohde & Schwarz Mitt. no. 22, pp.224-232.

[21] Coilcraft Inc., 2011-last update, Mini SpringTM Air Core Inductors
[Datasheet]. [Accessed March 2013]. Available:
http://www.coilcraft.com/pdf_viewer/showpdf.cfm?f=pdf_store:mini.pdf

[22] Johanson Technology, 2013-last update, Multi-Layer High-Q Capacitors

[Datasheet]. [Accessed March 2013]. Available:
http://www.johansontechnology.com/images/stories/catalog/JTI_CAT_2012_ML
CC_HighQ.pdf

[23] Peregrine Semicoductor Corp., 2013-last update, 64904 UltraCMOS®

Digitally Tunable Capacitor [Datasheet]. [Accessed March 2013]. Available:
http://www.psemi.com/pdf/datasheets/PE64904-70-0325-06.pdf
 49

[24] Digilent Inc., 2011-last update, ChipKIT™ Max32™ Board Reference

Manual, [online document]. [Accessed March 2013]. Available:
http://www.digilentinc.com/Products/Detail.cfm?NavPath=2,892,894&Prod=CHI
PKIT-MAX32