The Islamic University of Gaza
Electrical Engineering Department
Op t i m a l PLL l o o p Fi l t e r
De s i g n f o r Mo b i l e Wi Ma x
Vi a LMI
By
Ayman Akram AlQuqa
Supervisors
Dr. Hatem Elaydi
Dr. Hala J. El-Khozondar
Submitted in Partial Fulfillment for the Master Degree in Electrical
Engineering
March, 2009
Dedicated to
My Father,Mother, Wife and sweet sons: Mohamed, Abd Allah and Ahmad.
ii
Abstract
Achieving optimal design of phase-locked loop (PLL) is a major challenge in
WiMax technology in order to improve system behavior against noise and to enhance
Quality of Service (QOS). A new loop filter design method for phase locked loop
(PLLs) is introduced taking into consideration various design objectives: small settling
time, small overshoot and meeting Mobile WiMax requirements. Optimizing conflicting
objectives is accomplished via linear programming and semidefinite programming
(especially Linear Matrix Inequality (LMI)) in conjunction with appropriate adjustment
of certain design parameters. Digital filters, Infinite Impulse Response (IIR) and Finite
Impulse Response (FIR) are designed using linear programming and convex
programming.
Simulations show that IIR digital lowpass filter with narrow transition band could
not work properly with mobile WiMax system. Simulations show that FIR digital
lowpass filter utilizing linear programming managed to improve the transient behavior.
The FIR digital lowpass filter utilizing semidefinite programming (LMI) will much
improve the transient behavior; therefore it is recommended for mobile WiMax
systems.
iii
Acknowledgements
Praise be to Allah, the Almighty for having guided me at every stage of my life.
I have been fortunate to have the opportunity to work under the supervision of
Dr. Hatem Elaydi and Dr. Hala J. El-Khozondar. Their quidance, valuable suggestions
and support throughout all stages of my thesis are highly appreciated. I would also like
to place on record my appreciation for the cooperation and guidance by my committee
members, Dr. Basil Hamad and Dr. Ammar Abu Hadrous.
I recall with deep gratitude and respect the debt I owe to my mother, father and
brothers for their unfathomable love, prays, support and encouragement throughout my
studies. I would also like to thank my wife for the support and patience she extended
during my thesis. May Allah bless them all.
iv
Contents
1.
Introduction ................................................................................................................. 1
1.1
Motivation .......................................................................................................... 1
1.2
Linear Matrix Inequalities .................................................................................. 2
1.2.1
Linear Programming (LP) ..................................................................................... 3
1.2.2
Semi-definite Programming (SDP) ....................................................................... 5
1.3
Phase Locked Loop Fundamentals..................................................................... 7
1.3.1
Phase Detector (PD) .............................................................................................. 7
1.3.2
Loop Filter (LF) .................................................................................................... 8
1.3.3
Voltage Controlled Oscillator (VCO) ................................................................... 8
1.4
PLL Application: Frequency Synthesizer .......................................................... 8
1.5
The Problem ..................................................................................................... 10
1.6
Thesis Organization ......................................................................................... 11
2.
Literature Review ...................................................................................................... 12
3.
Mobile WiMax .......................................................................................................... 22
3.1
Fixed WiMAX vs. Mobile WiMAX ................................................................ 23
3.2
WiMAX Working ............................................................................................ 24
3.3
WiMAX: Building Blocks ............................................................................... 25
3.3.1
WiMAX Base Station ......................................................................................... 25
3.3.2
WiMAX Receiver (CPE) .................................................................................... 26
3.3.3
Backhaul.............................................................................................................. 27
3.4
WiMAX Application ........................................................................................ 28
3.4.1
Metropolitan Area Network (MAN) ................................................................... 28
3.4.2
Last-Mile: High-Speed Internet Access or Wireless DSL .................................. 29
3.4.3
Broadband on Demand ........................................................................................ 30
3.4.4
Cellular Backhaul ................................................................................................ 30
3.4.5
Residential Broadband: filling the gaps in cable & DSL coverage ..................... 31
v
3.4.6
Wireless VoIP ..................................................................................................... 31
3.4.7
Mobility ............................................................................................................... 32
3.5
4.
5.
3.5.1
Scalability............................................................................................................ 32
3.5.2
Relative Performance .......................................................................................... 32
3.5.3
QoS...................................................................................................................... 33
3.5.4
Range .................................................................................................................. 33
3.5.5
Security ............................................................................................................... 33
Design of PLL Filter .................................................................................................. 34
4.1
Fractional-N PLL block diagram ..................................................................... 35
4.2
IIR Low-Pass Filter Design .............................................................................. 37
4.3
FIR Low-Pass Filter Design ............................................................................. 40
4.3.1
LP formulation .................................................................................................... 41
4.3.2
SDP formulation.................................................................................................. 41
Results Analysis ........................................................................................................ 44
5.1
IIR low pass Filter ............................................................................................ 45
5.2
FIR low-pass Filter ........................................................................................... 48
5.2.1
Linear Programming ........................................................................................... 49
5.2.2
SDP Programming (LMI) ................................................................................... 51
5.3
6.
WiMAX versus WiFi ....................................................................................... 32
Discussion ........................................................................................................ 54
Conclusion and Future Work..................................................................................... 58
6.1
Conclusion........................................................................................................ 58
6.2
Future work ...................................................................................................... 58
7.
References ................................................................................................................. 59
8.
Appendix ................................................................................................................... 64
8.1
IIR1 MatLab Code ........................................................................................... 64
8.2
FIR1 MatLab Code .......................................................................................... 67
8.3
FIR2 MatLab Code .......................................................................................... 70
8.4
List of Acronyms.............................................................................................. 73
vi
List of Tables
Table 3-1 Fixed Wimax Vs. Mobile Wimax .................................................................. 24
Table 3-2 Scalability Comparison .................................................................................. 32
Table 3-3 Relative Performance Comparison ................................................................. 32
Table 3-4 Qos Comparison ............................................................................................. 33
Table 3-5 Range Comparison ......................................................................................... 33
Table 3-6 Security Comparison ...................................................................................... 33
Table 4-1 Design Specifications .................................................................................... 34
Table 5-1 Comparison Between IIR Digital Filter Design Using LP And Other Designs
................................................................................................................................ 55
Table 5-2 Comparison Between FIR Digital Filter Design Using LP And Other Designs
................................................................................................................................ 56
Table 5-3 Comparison Between FIR Digital Filter Design Using SDP And Other
Designs.................................................................................................................... 56
vii
List of Figures
Figure 1-1 Linear Programming Example ........................................................................ 4
Figure 1-2 A Simple Semidefinite Program With
And
....................... 6
Figure 1-3 A Basic PLL Block ......................................................................................... 7
Figure 1-4 Basic Frequency Synthesizer .......................................................................... 9
Figure 3-1 Wimax Working ............................................................................................ 25
Figure 3-2 Wimax Base Station ...................................................................................... 26
Figure 3-3 Wimax Receivers .......................................................................................... 27
Figure 3-4 Wimax Technology ....................................................................................... 27
Figure 3-5 Metropolitan Area Network .......................................................................... 29
Figure 3-6 Last Mile ....................................................................................................... 30
Figure 3-7 Cellular Backhaul .......................................................................................... 31
Figure 4-1 Basic Configuration Of A Frequency Synthesizer ........................................ 34
Figure 4-2 Basic Fractional-N PLL Block Diagram ....................................................... 35
Figure 4-3 Designed Fractional-N Synthesizer Block ................................................... 36
Figure 5-1 PLL Frequency Synthesizer Simulation Model ............................................ 44
Figure 5-2 3rd Order IIR Filter Magnitude/Phase Response .......................................... 46
Figure 5-3 3rd Order IIR Filter Magnitude/Phase Response With Much Higher
Stopband Frequency ............................................................................................... 47
Figure 5-4 The Control Signal Of VCO Input Using Designed IIR Filter .................... 48
Figure 5-5 FIR Impulse Response (LP) ......................................................................... 49
Figure 5-6 FIR Filter Magnitude/Phase Response (LP) ................................................. 50
Figure 5-7 The Control Signal Of VCO Input Using Designed FIR Filter ................... 51
Figure 5-8 FIR Impulse Response (LMI) ...................................................................... 52
Figure 5-9 FIR Filter Magnitude/Phase Response (LMI) .............................................. 52
Figure 5-10 The Control Signal Of VCO Input Using Designed FIR Filter (LMI) ...... 53
viii
1. Introduction
1.1 Motivation
The number of telecommunications innovations grew rapidly during the last half
of the 20th century [1]. Currently, there is widespread and growing use of cellular
phones, cordless phones, digital satellite systems, and personal mobile radio networks.
Over the past decade, there has been a tremendous growth in the popularity of wireless
networking technologies [2]. Schools and universities are providing wireless access to
students and faculty members. Malls are providing customers with wireless connectivity
to allow them to search products, virtually navigate shops, and interact with services.
Tourist sites are providing wireless devices to aid tourists in navigating, exploring and
learning about attractions. Moreover, the emergence of new computing paradigms such
as pervasiveness, ubiquity, and mobility has necessitated the rapid deployment of
wireless networks as an infrastructure underneath such technologies.
One of the main technology that can lay the foundation for the next generation
,fourth generation (4G), of mobile broadband networks is popularly known as
“WiMAX.” WiMAX, Worldwide Interoperability for Microwave Access, is designed to
deliver wireless broadband bitrates, with Quality of Service (QoS) guarantees for
different traffic classes, robust security, and mobility.
WiMAX is a set of specifications created by the WiMAX Forum [3] and is
based on standards developed by the IEEE 802.16 Working Group (WG) [4]. There are
two standards of note here. The first is IEEE 802.16-2004, sometimes called 802.16d,
which specifies a common air interface for fixed (both ends stationary) microwave
equipment. But since the high-growth market opportunity for wireless of any form
today is in mobile systems, the IEEE 802.16 WG subsequently issued IEEE 802.16e2005, which specifies a mobile broadband technology. The 802.16e, as it is commonly
known, is now seeing significant product development and production deployments on
a global basis. Farpoint Group expects that effective per-user throughput of 2-4 Mbps
will become common on carrier WiMAX networks over the next few years, with
monthly pricing perhaps below that currently charged for mobile broadband services
1
with far less throughput. WiMAX will also become a platform for application
deployment and could even be catalytic in the broad availability of Web services and
software as a service (SaaS), which Farpoint Group believes will become the dominant
model for IT in the future – mobile or not. A significant part in WiMax system is the
phase-locked loop.
Phase Locked Loop has become one of the most versatile building blocks in
electronics [5]. They are at the heart of circuits and systems ranging from clock
recovery blocks in data communications to the local oscillators that power the
ubiquitous cellular phones. The property of making its output frequency an exact
multiple of the reference frequency makes the Phase Locked Loop (PLL) the circuit of
choice for frequency synthesizers. PLL is also used for aligning various clocks in
synchronous systems and for a myriad of applications ranging from tracking satellite
Doppler shift to sensing minute reactance changes in industrial proximity sensors.
However, phase-locked loop (PLL), is an electronic circuit that controls an
oscillator so that it maintains a constant phase angle relative to a reference signal [6]. In
communications, the oscillator is usually at the receiver, and the reference signal is
extracted from the signal received from the remote transmitter.
1.2 Linear Matrix Inequalities
A wide variety of problems arising in system and control theory can be reduced
to a few standard convex or quasi-convex optimization problems involving linear matrix
inequalities (LMIs), that is constraints of the form
,
.
Where x ∈ R m is the variable and Fi = FiT ∈ R n×n , i = 0,..., m, are given.
Although the LMI form appears very specialized, it is widely encountered in system and
control theory. Lists of many comprehensive examples are found in Boyd et al [7].
2
Since these resulting optimization problems can be solved numerically very
efficiently as showed in [7], there are special cases with few analytical solutions to LMI
optimization problems. Indeed, the recent popularity of LMI optimization for control
can be directly traced to the recent breakthroughs in interior point methods for LMI
optimization [8]. The growing popularity of LMI methods for control is also evidenced
by the large number of publications in recent control conferences. Much of the research
effort in the application of LMI optimization has been directed towards problem from
control theory while, many of the underlying techniques extend to problems from other
areas of engineering as well, for instant, truss topology design and VLSI design.
1.2.1 Linear Programming (LP)
A linear programming problem may be defined as the problem of maximizing or
minimizing a linear function subject to linear constraints. The constraints may be
equalities or inequalities. Here is a simple example[9].
Find numbers x1 and x2 that maximize the sum x1 + x2 subject to the constraints
x1 ≥ 0, x2 ≥ 0, and
x1 + 2x2 ≤ 4
4x1 + 2x2 ≤ 12
(1.2)
−x1 + x2 ≤ 1
In this problem there are two unknowns, and five constraints. All the constraints are
inequalities and they are all linear in the sense that each involves an inequality in some
linear function of the variables. The first two constraints, x1 ≥ 0 and x2 ≥ 0, are special.
These are called nonnegativity constraints and are often found in linear programming
problems. The other constraints are then called the main constraints. The function to be
maximized (or minimized) is called the objective function. Here, the objective function
is x1 + x2 . Since there are only two variables, we can solve this problem by graphing the
set of points in the plane that satisfies all the constraints (called the constraint set) and
then finding which point of this set maximizes the value of the objective function. Each
inequality constraint is satisfied by a half-plane of points, and the constraint set is the
intersection of all the half-planes. In the present example, the constraint set is the five
sided figure shaded in Figure 1-1.
3
Figure 1-1 Linear Programming Example
We seek the point (x1, x2), that achieves the maximum of x1 + x2 as (x1, x2) ranges over
this constraint set. The function x1 + x2 is constant on lines with slope −1, for example
the line x1 + x2 = 1, and as we move this line further from the origin up and to the right,
the value of x1 + x2 increases. Therefore, we seek the line of slope −1 that is farthest
from the origin and still touches the constraint set. This occurs at the intersection of the
lines x1 + 2x2 = 4 and 4x1 + 2x2 = 12, namely, (x1, x2) = (8/3, 2/3). The value of the
objective function there is (8/3) + (2/3) = 10/3.
It is easy to see in general that the objective function, being linear, always takes on its
maximum (or minimum) value at a corner point of the constraint set, provided the
constraint set is bounded. Occasionally, the maximum occurs along an entire edge or
face of the constraint set, but then the maximum occurs at a corner point as well. Not all
linear programming problems are so easily solved. There may be many variables and
many constraints. Some variables may be constrained to be nonnegative and others
unconstrained. Some of the main constraints may be equalities and others inequalities.
However, two classes of problems, called here the standard maximum problem and the
standard minimum problem, play a special role. In these problems, all variables are
constrained to be nonnegative, and all main constraints are inequalities.
Standard form is the usual and most intuitive form of describing a linear programming
problem. It consists of the following three parts:
•
•
A linear function to be maximized or minimized, e.g. maximize
Problem constraints of the following form, e.g.
4
(1.3)
•
,
Non-negative variables, e.g.
.
The problem is usually expressed in matrix form, and then becomes:
(1.4)
maximize
subject to
,
Other forms, such as minimization problems, problems with constraints on alternative
forms, as well as problems involving negative variables can always be rewritten into an
equivalent problem in standard form.
1.2.2 Semi-definite Programming (SDP)
The (linear) semidefinite programming problem (SDP) is essentially an ordinary
linear program where the nonnegativity constraint is replaced by a semidefinite
constraint on matrix variables.
SDP has many applications, ranging from control theory to structural design. In
particular, many hard optimization problems (with integer constraints) can be relaxed to
a problem with convex quadratic constraints which, in turn, can be formulated as an
SDP. This SDP provides a polynomial time approximation to the original, hard
problem. Usually, approximations from SDP relaxations are better than those from
linear programming.
A semidefinite program is an optimization problem of the following form[10]:
(1.5)
minimize
subject to
Where
The problem data are the vector
,
,
and
. The inequality sign in
5
symmetric matrices
means that
is positive
semidefinite, i.e.,
for all
. We call the inequality
a linear
matrix inequality and the problem (1.5) a semidefinite program.
A semidefinite program is a convex optimization problem since its objective and
constraint are convex: if
, then, for all ,
and
,
. Figure 1-2 depicts a simple example
with
and
. Our goal here is to give the reader a generic picture that
shows some of the features of semidefinite programs, so the specific values of the data
are not relevant. The boundary of the feasible region is shown as the dark curve.
Figure 1-2 A simple semidefinite program with
The feasible region, i.e., |
and
consists of this boundary curve along with the
region it encloses. Very roughly speaking, the semidefinite programming problem is to
move as far as possible in the direction -c, while staying in the feasible region. For this
semidefinite program there is one optimal point, Xopt.
This simple example demonstrates several general features of semidefinite programs.
We have already mentioned that the feasible set is convex. Note that the optimal
solution Xopt is on the boundary of the feasible set, i.e.,
is singular; in the
general case there is always an optimal point on the boundary (provided the problem is
feasible). In this example, the boundary of the feasible set is not smooth. It is piecewise
smooth: it consists of two line segments and two smooth curved segments. In the
general case the boundary consists of piecewise algebraic surfaces. Skipping some
technicalities, the idea is as follows. At a point where the boundary is smooth, it is
defined locally by some specific minors of the matrix
boundary is locally the zero set of some polynomials in
surface.
6
vanishing. Thus the
,
,
, i.e., an algebraic
1.3 Phase Locked Loop Fundamentals
A Phase Locked Loop or a PLL is a feedback control circuit. As the name
suggests, the phase locked loop operates by trying to lock to the phase of a very
accurate input signal through the use of its negative feedback path. A basic form of a
PLL consists of three fundamental functional blocks namely [11]
•
•
•
A Phase Detector (PD)
A Loop Filter (LF)
A voltage controlled oscillator (VCO)
with the circuit configuration shown in Figure 1-3
Fin
Phase
Detector
Loop
Filter
VCO
Fout
Figure 1-3 A basic PLL Block
1.3.1 Phase Detector (PD)
The phase detector (PD) compares the phase of the output signal to the phase of
the reference signal. If there is a phase difference between the two signals, it generates
an output voltage, which is proportional to the phase error of the two signals. This
output voltage passes through the loop filter and then as an input to the voltage
controlled oscillator (VCO) controls the output frequency. Due to this self correcting
technique, the output signal will be in phase with the reference signal. When both
signals are synchronized the PLL is said to be in lock condition. The phase error
between the two signals is zero or almost zero. As long as the initial difference between
the input signal and the VCO is not too big, the PLL eventually locks onto the input
signal. This period of frequency acquisition, is referred as pull-in time, this can be very
long or very short, depending on the bandwidth of the PLL. The bandwidth of a PLL
depends on the characteristics of the phase detector (PD), voltage controlled oscillator
and on the loop filter.
7
1.3.2 Loop Filter (LF)
The filtering operation of the error voltage (coming out from the Phase Detector)
is performed by the loop filter. The output of PD consists of a dc component
superimposed with an ac component. The ac part is undesired as an input to the VCO,
hence a low pass filter is used to filter out the ac component. Loop filter is one of the
most important functional blocks in determining the performance of the loop. A loop
filter introduces poles to the PLL transfer function, which in turn is a parameter in
determining the bandwidth of the PLL. Since higher order loop filters offer better noise
cancelation, a loop filter of order 2 or more are used in most of the critical application
PLL circuits.
1.3.3 Voltage Controlled Oscillator (VCO)
VCO is an electronic oscillator (nonlinear device) designed to be controlled in
oscillation frequency by a voltage input. The frequency of oscillation is varied by the
applied DC voltage, while modulating signals may also be fed into the VCO to cause
frequency modulation (FM) or phase modulation (PM); a VCO with digital pulse output
may similarly have its repetition rate (FSK, PSK) or pulse width modulated (PWM).
1.4 PLL Application: Frequency Synthesizer
One of the most common uses of a PLL is in Frequency synthesizers of Wireless
systems. A frequency synthesizer generates a range of output frequencies from a single
stable reference frequency of a crystal oscillator [11]. Many applications in
communication require a range of frequencies or a multiplication of a periodic signal.
For example, in most of the FM radios, a phase-locked loop frequency synthesizer
technique is used to generate 101 different frequencies. Also most of the wireless
transceiver designs employ a frequency synthesizer to generate highly accurate
frequencies, varying in precise steps, such as from 600 MHz to 800 MHz in steps of 200
KHz. Frequency Synthesizers are also widely used in signal generators and in
instrumentation systems, such as spectrum analyzers and modulation analyzers.
A basic configuration of a frequency synthesizer is shown in Figure 1-4 [11]. Besides a
PLL, it also includes a very stable crystal oscillator with a divide by N-programmable
divider in the feedback loop. The programmable divider divides the output of the VCO
by N and locks to the reference frequency generated by a crystal oscillator.
8
fr
Phase
Detector
/
Loop
Filter
fo
VCO
Programmable
Counter (
)
Figure 1-4 basic Frequency Synthesizer
The output frequency of VCO is a function of the control voltage generated by
the PD. The Output of the phase comparator, which is proportional to the phase
difference between the signals applied at its two inputs, control the frequency of the
VCO. So the phase comparator input from the VCO through the programmable divider
remains in phase with the reference input of crystal oscillator. The VCO frequency is
thus maintained at
. This relation can be expressed as
.
This implies that the output frequency is equal to
.
Using this technique one can produce a number of frequencies separated by
and a
multiple of N. For example if the input frequency is 24KHz and the N is selected to be
32 ( a single integer ) then the output frequency will be 0.768 MHz In the same way, if
N is a range of numbers, the output frequencies will be in the proportional range. This
basic technique can be used to develop a frequency synthesizer from a single reference
frequency. This is the most basic form of a frequency synthesizer using phase locked
loop technique.
Return to Figure 1-4, note that
and
is the phase input,
the phase error,
output phase. Phase error (phase detector output) can be calculated from
9
.
and the VCO output can be calculated from
.
where
is the product of the individual feed forward transfer functions, and
is
the product of the individual feedback transfer functions
1.5 The Problem
My goal is to design PLL loop filter that accomplish all requirements to work
properly and efficiently with Mobile WiMax system. The design is based on
optimization particularly Linear Programming and Linear Matrix Inequalities (LMI)
techniques. Much of the research effort in the application of LMI optimization has been
directed towards problems from control theory while, many of the underlying
techniques extend to problems from other areas of engineering as well, for instant,
wireless applications. The problem to design and optimize the loop filter using LMI is
new in that it is used for Mobile WiMax system.
Mobile WiMax systems use digital low-pass filter (IIR or FIR) as main
component of frequency systhesizer. In other world, the problem is to design IIR Filter
or FIR Filter and replace it instead of loop filter in PLL block.
The designed loop filter will be stable and compatible with:
1. Frequency range used for Mobile WiMax systems (2.3 – 2.7) GHz.
2. 15 Mbps capacity up to 3km per channel.
3. Channel resolution (125 KHz).
4. Non line-of-site Requirement.
5. Low settling time, to lower the lock-in range.
6. Low overshoot.
10
In this thesis, we will utilize the Chou’s approach [12] which implement simple
PLL architecture to design loop filter by transforming the problem to convex
optimization particularly LMI. The implemented VCO was replaced with an
integrator multiplied by gain and their designed filter was used specially for GPS
system. Trying to use their technique to design loop filter for mobile WiMax is not
good idea because in wireless systems, we do not use simple PLL architecture,
instead we can use integer-N frequency synthesizer and for more modern systems
we use N-Fractional synthesizer.
In our design, the VCO noise is assumed to be white Gaussian noise and it is
neglected. N-Fractional Synthesizer is used instead of N-Integer Synthesizer to
reduce noise resulted from the factor N.
Traditional frequency synthesizers use low-pass analog filter to eliminate the
high frequency components, but in the new design we will use IIR or FIR digital
low-pass filter to eliminate the high frequency components resulting in more
immunity to noise.
1.6 Thesis Organization
Chapter 2 will cover literature review of PLL loop filter designs and optimizations.
Theoretical background on mobile WiMax standard will be presented and outlined in
chapter 3. The design of fractional-N frequency synthesizer using IIR and FIR digital
filters will be discussed in chapter 4. Chapter 5 will display the solution of the problem
using lmitool (cvx) and linear programming with comparison with other techniques.
Finally the conclusion and suggestions for future work will be given in chapter 6.
11
2. Literature Review
A wide variety of problems arising in system and control theory can be reduced to a
handful of standard convex and quasi-convex optimization problems that involve matrix
inequalities [7]. For a few special cases there are “analytic solutions” to these problems,
but their main point is that they can be solved numerically in all cases. These standard
problems can be solved in polynomial-time, and so are tractable, at least in a theoretical
sense. Recently developed interior-point methods for these standard problems had been
found to be extremely efficient in practice. Therefore, they considered the original
problems from system and control theory as solved.
Abbas-Turki, et al. [13], presented an LMI formulations for designing controllers
according to time response and stability margin constants. Convex mathematical
translations of both kinds of objectives (time and frequency-domain) were proposed
using Linear Matrix Inequalities (LMI). The application of Youla parameterization
allows restoring the linearity in the compensator parameters, but a huge state space
representation of the system was induced. Showing that the Cutting Plane Algorithm
(CPA) was efficiently used to overcome the problem of having a huge number of added
variables, which often occurs in Semi-Definite Programming (SDP) particularly when
used in conjunction with Youla Parameterization. The application of the CPA leads to
prevent the introduction of additional decision variables, which implies that a high order
of the Youla parameter can be considered without numerical difficulties. So the
feasibility of the problem can be easily checked by increasing gradually the order of
Youla parameter to be determined. The simplicity of using the CPA made it attractive,
although some numerical improvements can be a subject of forthcoming works. The
stability margins constrained were considered for MISO (Multi Input Single Output) or
SIMO (Single Input Multi Output) plants, the extension to the MIMO (Multi Input
Multi Output) case was being under investigation ( up to the authors [13]).
Focusing our attention to the new filter design methods, we start with the Henrion,
et al. [14] simultaneous Optimization over the numerator and denominator polynomials
in the Youla-Kurcera (YK) parameterization. They showed that optimizing
simultaneously over the numerator and denominator polynomials of the rational YK
parameter provides the designer with a great flexibility. Stability of the denominator
12
polynomial was ensured with the sufficient condition, meaning that the key ingredient
in the design procedure was the choice of the so-called central polynomial around which
the closed-loop dynamics must be optimized. With the help of numerical examples, it
was shown that the approach is suitable for fixed-order and
∞
controller design.
Stability of the denominator polynomial, as well as fixed-order controller design with
∞
performance were ensured via notation of a central polynomial and LMI conditions
for polynomial positivity. It was therefore possible to control at will the growth in the
controller order, hence overcoming an often mentioned difficulty of the YK
parameterization.
Since the theoretical description of phase-lock loop (PLL) was well established,
there have been several approaches to design a loop filter; In 1988, Abramovitch [15]
proved that Lyapunov stability techniques were adequate for analyzing the stability of a
third order phase-lock loop and was not substantially more difficult than that of a
second order loop. He treated the third order PLL as a nonlinear control system: first
examine the small signal (linear) operation and then extending the analysis to the
nonlinear region. To accomplish his work he used the second method of Lyapunouv and
LaSall’s Theorem. In 1990 [16], Abramovitch solved the problem of stability and
tracking analysis for the nonlinear model of analog phase-locked loops using
Lyapunouv redesign [17]. However, there was a difficulty in applying the proposed
method to high order loops.
After that a concise review of the PLL technique, which is applicable to
communication and servo control system is illustrated in Phase-locked loop Techniques
Survey presented by Hisch and Hung, 1996[18]. As a result, it is expected that PLL will
contribute to improvement in performance and reliability for future communication
systems. It will also contribute to the development of higher accuracy and higher
reliability servo control systems, such as those involved in machine tools. The status of
the PLL technology and its applications has been discussed, and a summary of the PLL
technology and its development trends are also included. It is pointed out that the
development of better PLL technology and the associated modular IC’s is continuing.
The PLL-based servo control system has become important and popular in the
development of mechatronics.
13
After that, Suplin and Shaked [19] introduced a simple systematic procedure to
design PLLs that entail parameter uncertainty and time delay. The procedure was based
on translating the resulting output-feedback control problem into a problem of designing
a state-feedback on an augmented non minimal system. A mixed H2/H∞, was used to
obtain the required stability margin for the PLL while minimizing the effect of the
measurement and the phase noise inputs. One of the main drawbacks of the above
method of translating the output-feedback problem into a state-feedback one was that it
was tuned to the exact model one assumes for the phase noise spectrum.
Yaniv and Raphael [20] presented a design method for near-optimal PLL taking
into consideration the phase noise, the thermal noise, the undesired but unavoidable
loop delay caused by delayed decisions, and margins for protection from gain
uncertainty and insuring good step response. Unlike the conventional designs, their
design incorporated a possible large decision delay and S-curve slope uncertainty. Large
decision delays frequently existed in modern receivers due to, for example,
a
convolutional decoder or an equalizer. The new design also applied to coherent optical
communications where delays in the loop limit the laser linewidth. They provided an
easy-to-use complete design procedure for second-order loops, and also introduced a
design procedure for higher order loops for near-optimal performance. They also
showed that using the traditional second-order loop was suboptimal when there was a
delay in the loop, and also showed large improvements, either in the amount of allowed
delay, or the phase error variance in the presence of delay.
To show the stability problem, an analog phase-lock loop design using Popov
Criterion [21] was presented by N. Eva Wu, 2002. The phase-lock criterion, when
combined with some straightforward numerical search, can be used to design the lowpass filter in a phase-lock loop with a guaranteed lock range.
A PLL with digitally-controllable loop parameters was designed to optimize jitter
performance, by Mansuri and Ken Yang, 2002 [22]. They developed an intuition for
designing low-jitter PLLs both by deriving a closed-form solution for a second-order
loop and by plotting the sensitivity to various loop parameters for higher order loops.
Furthermore, the loop served as a test bench to verify their analysis. The analysis
showed a simple expression for long term jitter due to VCO and buffering noise to the
14
damping factor and natural frequency. They derived an expression that relates the jitter
contribution of clock buffering (in the feedback) and VCO to the same parameters. They
also validated the common design practice of using high loop bandwidth to reduce
VCO-induced jitter. However to minimize jitter, they found that accounting for the loop
delay in the phase margin is critical. Interestingly, this minimum is very insensitive to
Pressure Volume Temperature (PVT) and parameter variations making such a design
robust. For applications that require small short-term jitter (i.e., short distance links and
block to block interconnect), an underdamped loop can result in much higher short-term
rms jitter. For applications that filters input jitter, their modeling showed that very low
bandwidths (0.002%
) were necessary to reduce noise by a factor of 10 while a
damping factor greater than 2 was sufficient.
However, Fahim and Elmansury [23] introduced fast lock digital phase-locked loop
architecture for wireless applications. The main advantages of this architecture include
small area and digitally selectable frequency resolution. Also, a fully digital solution to
reduce the phase-lock time was introduced. In their study, techniques for fast lock PLL
design were reviewed. A rigorous time-domain based approach was adopted in order to
better understand the frequency locking phenomenon in PLLs. Bounds on expressions
of frequency lock time were developed. Phase-locking phenomena were also
investigated. It was determined that there seems to be an optimal trajectory in which the
phase lock time of the PLL is minimized. Based on this study, an effective method of
reducing the lock time was demonstrated. It was shown that the lock time and frequency
resolution trade-off can be performed in the digital domain, as opposed to the analog
domain. The advantage of this was that in the digital domain, the tradeoff deals with the
area and power penalties incurred for extra performance gains. The digital PLL
architecture was shown to be effective for low frequency resolution standards such as
cordless and wireless LAN standards. For high frequency resolution standards, such as
GSM, the architecture could only be used as a frequency aid circuit. The cost of fine
resolution became excessively high in terms of area penalty as well as lock time. A
simple, yet effective method of reducing the phase lock time was also introduced.
An efficient, systematic, and robust method for the optimization of PLL circuits
using geometric programming was illustrated in [24], where PLL circuits with VCO
frequencies ranging from 814MHz to 1.9GHz were automatically generated. The
15
optimization returns a clean design rule checking (DRC) and layout versus schematic
(LVS) of PLL. Power consumption predictions down to 3mW and accumulated jitter
predictions below 6ps agreed with silicon measurements. Using this technique, the PLL
design cycle was reduced from a matter of weeks to a matter of hours. To the authors’
knowledge [24], this was the first example of fully automated PLL design.
In 2004, A Magnitude/Phase-Locked Loop System Based on Estimation of
Frequency and In-Phase/Quadrature-Phase Amplitudes was presented by M. KarimiGhartemani, et al. [25]. A new PLL system that provides the dominant frequency
component of the input signal and estimates its frequency was presented. The QPLL
(Quadrature PLL) is advantageous for communication system applications such as
QPSK, QAM, and DSB-SC which deals with quadrature modulation techniques. The
proposed system provided the dominant frequency component of the input signal and
estimated its frequency. The mechanism of the proposed PLL was based on estimating
in-phase and quadrature-phase amplitudes of the desired signal and, hence, had
application advantages for communication systems which employed quadrature
modulation techniques. The studies demonstrated that the proposed PLL also provided a
superior performance for power system applications. Derivation of the mathematical
model and theoretical stability analysis of the proposed PLL were carried out using
dynamical systems theory. Advantages of the proposed PLL over the conventional
PLLs were its capability of providing the fundamental component of the input signal
which is not only locked in phase but also in amplitude to the actual signal while
providing an estimate of its frequency. Computer simulation was used to evaluate its
performance. Evaluations confirmed structural robustness of the proposed PLL with
respect to noise and distortions.
All-pole phase-locked loop: calculating lock-in range by using Evan’s root-locus
was presented by Piqueira, and Monteiro [26] in 2006. By studying the problem of
synchronizing the PLLs with all-pole filter by using an equivalent linear feedback
control system they calculated the lock-in range, for any order n+1 of the loop. This
contribution had, mainly, a theoretical point of view showing the conditions for the
existence of the lock-in range in a higher-order PLL with all-pole filter but giving
practical conditions to calculate it when designing circuits. Analysis was performed by
detecting a Hopf bifurcation on the synchronous state by using the root-locus method
16
combined with the dynamical system theory. The lock-in range was calculated by
applying the classical control tools defining an equivalent feedback control system.
In 2007, PLL Equivalent Augmented System Incorporated with State Feedback
Designed by LQR was presented by Wanchana, Benjanarasuth et al [27]. They derived
the optimal value of filter time constant of loop filter (LF) in the phase-locked loop
control system and the optimal state feedback gain designed by using linear quadratic
regulator approach. In designing, the structure of phase-locked loop control system
rearranged to be a phase-locked loop equivalent augmented system by including the
structure of loop filter into the process and by considering the voltage control oscillator
as an additional integrator. The designed controller consisting of state-feedback gain
matrix K and integral gain KI is an optimal controller. The integral gain KI related to
weighing matrices q and R was an optimal value for assigning the filter time constant of
loop filter. The experimental results in controlling the second-order lag pressure process
using two types of loop filters showed that the system response was fast without steadystate error, the output disturbance effect rejection was fast and the tracking to step
changes was good. It can be concluded that the proposed technique allowed the designer
to easily assign the filter time constant of the loop filter (LF) from the gain KdKo and the
weighting matrices q and R.
Now, we want to focus on PLL frequency Synthesizer; one of the drawbacks of a
traditional frequency synthesizer [28], also known as an Integer-N frequency
synthesizer, is that the output frequency is constrained to be N times the reference
frequency. If the output frequency is to be adjusted by changing N, which is constrained
by the divider to be an integer, then the output frequency resolution is equal to the
reference frequency. If fine frequency resolution is desired, then the reference frequency
must be small. This in turn limits the loop bandwidth as set by the loop filter, which
must be at least 10 times smaller than the reference frequency to prevent signal
components at the reference frequency from reaching the input of the VCO and
modulating the output frequency, creating spurs or sidebands at an offset equal to the
reference frequency and its harmonics. A low loop bandwidth is undesirable because it
limits the response time of the synthesizer to changes in N. In addition, the loop acts to
suppress the phase noise in the VCO at offset frequencies within its bandwidth, so
reducing the loop bandwidth acts to increase the total phase noise at the output of the
17
VCO. The constraint on the loop bandwidth imposed by the required frequency
resolution is eliminated if the divide ratio N is not limited to be an integer. This is the
idea behind fractional-N synthesis.
‘SWRA029’ is technical brief published by Texas Instruments and edited by Curtis
Barrett [29] described -in depth- Fractional/Integer-N PLL basics. This document
detailed basic loop transfer functions, loop dynamics, noise sources and their effect on
signal noise profile, phase noise theory, loop components (VCO, crystal oscillators,
dividers and phase detectors) and principles of integer-N and fractional-N technology.
The approach was mainly heuristic, with many design examples.
A new methodology for designing fractional-N frequency synthesizers and other
phase locked loop (PLL) circuits is presented by Lau, and Perrott, “Fractional-N
frequency synthesizer design at the transfer function level using a direct closed loop
realization algorithm”[30]. The approach achieved direct realization of the desired
closed loop PLL transfer function given a set of user-specified parameters and
automatically calculated the corresponding open loop PLL parameters. The algorithm
also accommodated nonidealities such as parasitic poles and zeros. The entire
methodology was implemented in a GUI-based software package which was used to
verify the approach through comparison of the calculated and simulated dynamic and
noise performance of a third-order fractional-N frequency synthesizer.
Furthermore, Design and Simulation of Fractional-N PLL Frequency Synthesizers
is presented by Kozak and Friedman [31] where a behavioral level simulation
environment was developed for Fractional-N PLL frequency synthesizers on a mixed
MATLAB™ and CMEX™ platform. A uniform simulation time step was allowed by
appropriately modeling the continuous-time average current-to-voltage loop filter
transfer function as a discrete-time charge difference-to-voltage transfer function. The
simulator enabled the exhaustive behavioral level simulation of Fractional-N PLL
frequency synthesizers in a fast and accurate manner. The simulation results
demonstrated the effectiveness of the “1” LSB initial condition imposed on the first
integrator of the ∆∑ modulator in rejecting fractional spurs.
18
The Nonlinear Phase-Locked Loop Design using Semidefinite Programming is
illustrated by Wang, et al. [12]. This design approach was based on the polynomial
nonlinear model of the PLL system. This approach started with the linear design of the
controller and then estimated the domain-of-attraction of the linear designed system to
get the suitable local Lyapunov function for the system. The Lyapunov function was
then used as the performance constraints to further refine the performance of the system
outside the linear region. In otherworld, a Lyapunov function was searched as a
certificate of the lock-in region of the PLL system. Moreover, the polynomial design
technique was used to further refine the controller parameters for system response away
from the equilibrium point. Various simulation results were provided to show the
effectiveness of the approach.
Chou, et al. 2006 [32], presented a new filter design method for PLL taking into
consideration the various design objectives such as small noise bandwidth, good
transient response (small settling time, small overshoot), and larger gain and phase
margins. This method is simple and applicable to PLL of any order. Particularly, it
allows one to specify the filter poles to desired locations in advance (including the
special case of PI form filters which have all the poles at the origin). Numerical
simulation of a GPS (Global Positioning System) application was performed using
nonlinear PLL model. It was observed that this method yields much better performance,
when compared with the traditional GPS PLL design. Trade-off among the conflicting
objectives was made via recently developed convex optimization skill in conjunction
with appropriate adjustment of certain design parameters.
An LMI approach to
∞
Optimal Filter Design for GPS Receiver Tracking Loop
was presented by Phi Long, et al. [33]. They investigated a new approach for GPS
receiver tracking loop, using
∞
theory, where a closed-loop system was designed with
objective to minimize the worst case amplification from the input noise power to the
output. The main drawback of this approach was the much higher complexity compared
with the
∞
controller design.
Wasim Al-Baroudi in his thesis titled “Digital Filter Design using LMI Based
Techniques” in 1997 [34] formulated the FIR filter design problem as a Linear
Objective Optimization problem with LMI constraints, and developed a necessary
19
software for solving it. He also formulated the IIR filter design problem as a Linear
Objective Optimization problem with LMI constraint and with iterative scheme to
overcome the nonlinear term, and developed the necessary software for solving it.
Another contribution was the introduction of the Frequency Selection Algorithm that
reduced the number of LMI’s solved to reach the optimal solution. He also introduced
the Formulation of the FIR/IIR Optimal Power Spectrum Approximation Problem as
Linear Objective Optimization problem with LMI constraint, and developed the
necessary software for solving it. Finally he introduced an LMI-based Model Reduction
Technique.
FIR Filter Design via Semidefinite Programming and Spectral Factorization was
presented by Wu, et al. [35]. They presented a new semidefinite programming approach
to FIR filter design with arbitrary upper and lower bounds on the frequency response
magnitude. It was shown that the constraints can be expressed as linear matrix
inequalities (LMIs), and hence they could be easily handled by recent interior-point
methods. Using this LMI formulation, we can cast several interesting filter design
problems as convex or quasi-convex optimization problems, e.g., minimizing the length
of the FIR filter and computing the Chebychev approximation of a desired power
spectrum or a desired frequency response magnitude on a logarithmic scale. Many
other extensions that were not discussed in the paper can be handled in the same
framework, such as, maximum stopband attenuation or minimum transition-band width
FIR design given magnitude bounds, or even linear array beam-forming. Recent
interior-point methods for semidefinite programming can solve each of these problems
very efficiently.
Linear programming design of IIR digital filters with arbitrary magnitude function
was presented by Rabiner, et al. [36].
This paper discussed the use of linear
programming techniques for the design of Infinite Impulse Response (IIR) digital
filters. In particular, it was shown that, in theory, a weighted equiripple approximation
to an arbitrary magnitude function can be obtained in a predictable number of
applications of the simplex algorithm of linear programming. When one implements the
design algorithm, certain practical difficulties (e.g., coefficient sensitivity) limit the
range of filters which can be designed using this technique. However, a fairly large
20
number of IIR filters were successfully designed and several examples were presented
to illustrate the range of problems for which they found this technique to be useful.
Previous studies mentioned above have made important contributions to this
challenging problem; however, none of the published studies appear to have completely
accounted of the optimization of the Phase Locked Loop (PLL) using the Linear Matrix
Inequality (LMI) method for Mobile WiMax application.
21
3. Mobile WiMax
The term ‘WiMAX’ has been used generically to describe wireless systems that are
based on the WiMAX certification profiles of the IEEE 802.16-2004 Air Interface
standard. With additional profiles pending that are based on the IEEE 802.16e-2005
Mobile Amendment, it is necessary to differentiate between the two WiMAX systems.
‘Fixed’ WiMAX is used to describe 802.16-2004-based systems while ‘Mobile’
WiMAX is used to describe 802.16e-2005-based systems [37].
Mobile WiMAX [38] is a broadband wireless solution that enables convergence of
mobile and fixed broadband networks through a common wide area broadband radio
access technology and flexible network architecture. The Mobile WiMAX Air Interface
adopts Orthogonal Frequency Division Multiple Access (OFDMA) for improved multipath performance in non-line-of-sight environments.
Mobile WiMAX systems offer scalability in both radio access technology and
network architecture, thus providing a great deal of flexibility in network deployment
options and service offerings. Some of the salient features supported by Mobile
WiMAX are:
•
High Data Rates : The inclusion of MIMO antenna techniques along with
flexible sub-channelization schemes, Advanced Coding and Modulation all
enable the Mobile WiMAX technology to support peak DL (Down Link) data
rates up to 63 Mbps per sector and peak UL (Up Link) data rates up to 28 Mbps
per sector in a 10 MHz channel.
•
Quality of Service (QoS) : The fundamental premise of the IEEE 802.16 MAC
architecture is QoS. It defines Service Flows which can map to DiffServ code
points or MPLS flow labels that enable end-to-end IP based QoS. Additionally,
sub-channelization and MAP-based signaling schemes provide a flexible
mechanism for optimal scheduling of space, frequency and time resources over
the air interface on a frame-by-frame basis.
22
•
Scalability : Despite an increasingly globalized economy, spectrum resources
for wireless broadband worldwide are still quite disparate in its allocations.
Mobile WiMAX technology therefore, is designed to be able to scale to work in
different channelizations from 1.25 to 20 MHz to comply with varied worldwide
requirements as efforts proceed to achieve spectrum harmonization in the longer
term. This also allows diverse economies to realize the multifaceted benefits of
the Mobile WiMAX technology for their specific geographic needs such as
providing affordable internet access in rural settings versus enhancing the
capacity of mobile broadband access in metro and suburban areas.
•
Security : The features provided for Mobile WiMAX security aspects are best in
class with AP based authentication, AES-CCM-based authenticated encryption,
and CMAC and HMAC based control message protection schemes. Support for
a diverse set of user credentials exists including; SIM/USIM cards, Smart Cards,
Digital Certificates, and Username/Password schemes based on the relevant
EAP methods for the credential type.
•
Mobility : Mobile WiMAX supports optimized handover schemes with latencies
less than 50 milliseconds to ensure real-time applications such as VoIP perform
without service degradation. Flexible key management schemes assure that
security is maintained during handover.
802.16m is the next generation standard beyond 802.16e-2005 and will be adopted
by the WiMAX Forum once the standard is completed in the 2009 timeframe. 802.16m
is considered to be a strong candidate for a 4G technology. The IEEE has defined its
expected parameters for 802.16m, which can be found on their Web site [37].
3.1 Fixed WiMAX vs. Mobile WiMAX
The most useful resource along this section is taken from Mobile WiMAX
Handbook [38]. WiMAX is also called Mobile WiMAX as it can serve all usage models
from fixed to mobile with the same infrastructure. Table 3-1 shows a comparison
23
between fixed and mobile WiMax technologies. Based on the IEEE 802.16e-2005
standard, Mobile WiMAX offers fixed, nomadic, portable, mobile capabilities and:
•
It does not rely on line-of-sight transmissions in lower frequency bands (2 to 11
•
GHz).
•
It is currently uses Time Division Duplexing (TDD).
•
It provides enhanced performance, even in fixed and nomadic environments.
It’s system bandwidth is scalable to adapt to capacity and coverage needs.
Table 3-1 Fixed WiMAX vs. Mobile WiMAX
Fixed WiMAX
Mobile WiMAX
Frequency(GHz)
3.5, 5.8
2.3, 2.5, 3.5, etc
Channel (MHz)
3.5, 7, 10, 14
3.5, 7, 8.75, 10, 14,
etc
TDD/FDD
TDD/FDD
TDMA
OFDMA
Duplexing
Multiple Access
3.2 WiMAX Working
The most useful resource along this section is taken from ‘The Business of
WiMAX [39]’. WiMAX has been designed to address challenges associated with
traditional wired and wireless access deployments [39]. A WiMAX network has a
number of base stations and associated antennas communicating by wireless to a much
larger number of client devices (or subscriber stations).
The WiMax MAN is schematically similar to the point-to-multipoint layout of a
cellular network. The original 802.16 specification paved the way for fixed wirelessaccess coverage, which requires a mounted outdoor antenna at the customer’s access
point. This fixed wireless-access coverage enables clients to communicate with their
respective base station, but the 802.16e ‘mobility’ extension enables seamless
communication from station to station (roaming property).
A WiMax base station is connected to public networks using optical fiber, cable,
microwave link or any other high-speed point-to-point connectivity, referred to as a
backhaul. In a few cases, like mesh networks, a point-to-multipoint WiMAX link to
24
other base stations is used as a backhaul. Ideally WiMAX should use point-to-point
antennas as a backhaul to connect aggregate subscriber sites to each other and to base
stations across long distances.
The base station serves subscriber stations (also called customer premise
equipment) using non-line-of-sight or line-of-sight point-to multipoint connectivity
referred to as ‘last mile’. Ideally, WiMAX should use non-line-of-sight point-tomultipoint antennas to connect residential or business subscribers to the base station
(See Figure 3-1).
Figure 3-1 WiMAX Working
3.3 WiMAX: Building Blocks
The most useful resource along this section is taken from ‘The Business of
WiMAX [39]’. Typically, a WiMAX system consists of two parts: a WiMAX base
station and a WiMAX receiver also referred as customer premise equipment (CPE).
While the backhaul connects the system to the core network, it is not the integrated part
of WiMAX system as such.
3.3.1 WiMAX Base Station
A WiMAX base station consists of indoor electronics and a WiMAX tower.
Typically, a base station can cover up to a radius of 30 miles or 50 km -theoretically;
however, any wireless node within the coverage area would be able to access the
Internet (Figure 3-2).
25
Figure 3-2 WiMAX Base Station
The WiMAX base stations would use the MAC layer defined in the standard – a
common interface that makes the networks interoperable – and would allocate uplink
and downlink bandwidth to subscribers according to their needs, on an essentially realtime basis. Each base station provides wireless coverage over an area called a cell. The
maximum radius of a cell is theoretically 50 km (depending on the frequency band
chosen). As with conventional cellular mobile networks, the base-station antennas can
be Omni-directional, giving a circular cell shape, or directional to give a range of linear
or sectoral shapes for point-to-point use or for increasing the network’s capacity by
effectively dividing large cells into several smaller sectoral areas.
3.3.2 WiMAX Receiver (CPE)
A WiMAX receiver (Figure 3-3) may have a separate antenna (i.e. receiver
electronics and antenna are separate modules) or could be a stand-alone box or a
PCMCIA card that sits in your laptop or computer or Impeded CPE used inside Laptop.
Access to a WiMAX base station is similar to accessing a wireless access point in a
WiFi network, but the coverage is greater. So far one of the biggest deterrents to the
widespread acceptance of BWA has been the cost of CPE. This is not only the cost of
the CPE itself, but also the installation cost.
26
Figure 3-3 WiMAX Receivers
3.3.3 Backhaul
Backhaul refers both to the connection from the access point back to the
provider and to the connection from the provider to the core network. A backhaul can
deploy any technology and media provided it connects the system to the backbone. In
most of the WiMAX deployment scenarios, it is also possible to connect several base
stations to one another using high-speed backhaul microwave links. This would also
allow for roaming by a WiMAX subscriber from one base station coverage area to
another, similar to the roaming enabled by cell phones (Figure 3-4).
Figure 3-4 WiMAX Technology
27
3.4 WiMAX Application
The most useful resource along this section is taken from ‘The Business of
WiMAX [39]’. The 802.16 standard will help the industry provide solutions across
multiple broadband segments. WiMAX was developed to become a last mile access
technology comparable to DSL, cable and T1 technologies. It is a rapidly growing
technology that is most viable for backhauling the rapidly increasing volumes of traffic
being generated by Wi-Fi hotspots.
WiMAX is a Metropolitan Area Network (MAN) technology that fits between wireless
LANs, such as 802.11, and wireless wide-area networks (WANs), such as the cellular
networks. Bandwidth generally diminishes as range increases across these classes of
networks. Proponents believe that WiMAX can serve in applications such as cellular
backhaul systems, in which microwave technologies dominate, backhaul systems for
Wi-Fi hotspots and most prominently as residential and business broadband services.
WiMAX was developed to provide high-quality, flexible, reduce cost, BWA using
certified, compatible and interoperable equipments from multiple vendors. There are
many application of WiMAX that discussed in followed subsections.
3.4.1 Metropolitan Area Network (MAN)
What makes WiMAX so attractive is its potential to provide broadband wireless
access to entire sections of metropolitan areas as well as small and remote locales
throughout the world. People who could not afford broadband will now be able to get it,
and in places where it may not previously have been available. WiMAX enables
coverage of a large area very quickly (Figure 3-5).
28
Figure 3-5 Metropolitan Area Network
Today, MANs are being implemented by a wide variety of innovative
techniques such as running fiber cables through subway tunnels or using broadband
over power lines (BPL). In response to these new techniques, there has been a growing
interest in the development of wireless technologies that achieve the same results as
traditional MANs without the difficulty of supplying the actual physical medium for
transmission, such as copper or fiber lines.
Undeniably, wireless MANs (WMANs) are emerging as a viable solution for
broadband access. MANs are intended to serve an area approximately the size of a large
city; MANs serve as the intermediary network between LANs and WANs. WMANs
consist of a fixed wireless installation that interconnects locations within a large
geographic region.
3.4.2 Last-Mile: High-Speed Internet Access or Wireless DSL
DSL operators, who initially focused their deployments in densely populated
urban and metropolitan areas, are now faced with the challenge to provide broadband
services in suburban and rural areas where new markets are quickly taking root.
Governments are prioritizing broadband as a key political objective for all citizens to
overcome the ‘broadband gap’ also known as the ‘digital divide’ (Figure 3-6).
WiMAX is also a natural choice for underserved rural and outlying areas with
low population density.
29
Figure 3-6 Last Mile
3.4.3 Broadband on Demand
One aspect of the existing IEEE 802.16a standard that will make it attractive to
service providers and end customers alike is its provision for multiple service levels.
Thus, for example, the shared data rate of up to 75 Mbps that is provided by a single
base station can support the ‘committed information rate’ to business customers of a
guaranteed 2 Mbps (equivalent to a E1), as well as ‘best-effort’ non-guaranteed 128
kbps service to residential customers.
The key parameters of WiMAX receiving attention are concerned with its capability to
provide differential services. Quality of service enables NLOS operation without severe
distortion of the signal from buildings, weather and vehicles. It also supports intelligent
prioritization of different forms of traffic according to its urgency.
3.4.4 Cellular Backhaul
The robust bandwidth of IEEE 802.16 makes it an excellent choice for backhaul
for commercial enterprises such as hotspots as well as point-to-point backhaul
applications (Figure 3-7). Also, with the WiMAX technology, cellular operators will
have the opportunity to lessen their independence on backhaul facilities leased from
their competitors. Here the use of point-to-point microwave is more prevalent for
mobile backhaul, but WiMAX can still play a role in enabling mobile operators to costeffectively increase backhaul capacity using WiMAX as an overlay network.
30
Figure 3-7 Cellular Backhaul
Some salient points about WiMAX use as cellular backhaul are:
•
•
High-capacity backhaul.
•
There is capacity to expand for future mobile services.
•
Multiple cell sites are served.
It is a lower cost solution than traditional landline backhaul.
3.4.5 Residential Broadband: filling the gaps in cable & DSL coverage
The range, absence of a LOS requirement, high BW, flexibility, and low cost
help to overcome the limitations of traditional wired and proprietary wireless
technologies.
3.4.6 Wireless VoIP
Wireless VoIP is a simple and cost-effective service which allows a subscriber
to use VoIP services while on the move. This is possible because of WiMAX which can
provide carrier-grade connectivity while being wireless. WiMAX using a dynamic
resource allocation protocol (DRAP) that provides admission & congestion control
within the wireless domain of the network.
31
3.4.7 Mobility
IEEE 802.16e allows users to connect to a WISP even when they roam outside
their home or business, or go to another city that also has a WISP.
3.5 WiMAX versus WiFi
The main and most useful resource for these comparisons is taken from [40].
3.5.1 Scalability
Table 3-2 shows a comparison between the IEEE 802.11 & 802.16a
Table 3-2 Scalability Comparison
802.11
•
•
Wide (20MHz) frequency
MHz to 20 MHz width channels.
Channel bandwidths can be
channels.
•
802.16a
•
MAC designed to support 10's
of users
chosen by operator.
MAC designed to support
thousands of users.
3.5.2 Relative Performance
Table 3-3 shows a relative performance comparison between IEEE 802.11 &
802.16a.
Table 3-3 Relative Performance Comparison
Standard
802.11
802.16a
Channel Bandwidth
20 MHz
Selectable channel
MAX. Data Rate
Mb/s
32
bandwidths between .
and
M(z
Mb/s
3.5.3 QoS
Table 3-4 shows a Quality of Service comparison between IEEE 802.11 &
802.16a.
Table 3-4 QoS Comparison
•
•
•
•
802.11
Contention‐based MAC CSMA/CA =>
•
no guaranteed QoS
Standard cannot currently guarantee
latency for Voice, Video.
TDD only – asymmetric
.
e proposed QoS is
prioritization only.
•
•
•
802.16a
Grant‐request MAC
-
TDM (for DL)
-
TDMA (for UL)
Designed to support Voice and Video.
TDD/FDD – symmetric or asymmetric
Centrally‐enforced QoS
3.5.4 Range
Table 3-5 shows a range comparison between IEEE 802.11 & 802.16a.
Table 3-5 Range Comparison
•
•
802.11
Up to
•
meters
Optimized for indoor performance.
•
802.16a
Up to
Km
Optimized for outdoor NLOS
performance.
3.5.5 Security
Table 3-6 shows security comparison between IEEE 802.11 & 802.16a.
Table 3-6 Security Comparison
•
•
802.11
Standard WEP + WPA
.
•
Triple‐DES
‐bit .
i in process of addressing
security.
802.16a
33
‐bit and RSA
4. Design of PLL Filter
One of the most common uses of a PLL is in Frequency synthesizers. A frequency
synthesizer generates a range of output frequencies from a single stable reference
frequency of a crystal oscillator.
Basic configuration of a frequency synthesizer is shown in Figure 4-1. Besides a
PLL it also includes a very stable crystal oscillator with a divide by N-programmable
fr
Crystal
Oscillator
Phase
Detector
Loop
Filter
VCO
fo
Programmable
Divider (
)
Figure 4-1 Basic configuration of a frequency synthesizer
divider in the feedback loop. The programmable divider divides the output of the VCO
by N and locks to the reference frequency generated by a crystal oscillator.
To gain some design experience and some more insight, consider a real design
problem of a frequency synthesizer loop filter with the following specifications as
shown in Table 4-1.
Table 4-1 Design Specifications
DESIGN SPECIFICATIONS [41]
PARAMETER
SPECIFICATION
Frequency Range
2.3 GHz – 2.7 GHz
Resolution
125 KHz
Overshoot
Less than 20%
Settling time
Less than
Directivity
Non Line-of-Sight
34
The design process is divided into several stages. We first present the overall block
of frequency synthesizer, then select the integer value of N according to reference
frequency and resolution. The next step is the design of digital IIR low-pass loop filter.
IIR low-pass loop filter design is carried out using linear programming (LP) technique.
After that, FIR low-pass loop filter will be designed. FIR low-pass loop Filter
design is carried out using linear programming (LP) and then using Semi-Definite
programming (SDP) utilizing Linear Matrix Inequalities formulation (LMI).
In the next chapter, we will implement the algorithm and simulate the designed filters
using our generated MATLAB codes.
4.1 Fractional-N PLL block diagram
The fractional-N PLL block diagram is showed in Figure 4-2,
Figure 4-2 Basic Fractional-N PLL Block Diagram
The fractional-N PLL consists of:
1. Phase/Frequency detector which is assumed XOR type.
2. Loop Filter which is the objective of our designe, is a low-pass filter (LPF).
3. Voltage Control Oscillator (VCO).
We first begin the design with Integer-N PLL: 125 kHz Reference pushes N
from 18400 to 21600 (2700/.125). As a result the loop filter cutoff (<12.5 KHz)
produces long settling time and VCO phase noise increased by 20*log10(N) ≈ 87dB.
35
To overcome the previous drawback, we use the Multi-Modulus Fractional PLL with
these properties:
•
•
•
•
Fractional value between N and 2N-1 (64-127).
Sigma Delta Modulator (Programmable resolution).
Large Reference (20MHz) for good tradeoff with settling time.
Reduced N impact on phase noise by 45dB over Integer N.
Example 4.1:
To produce 2300MHz, we produce 1533MHz (from VCO) and then upconvert it to
2300MHz (1533MHz * 1.5 ≈ 2300MHz). The 1533MHz can be produced with N = 76
and a fraction = 0.65 (means that 20MHz * (76 + 0.65) = 1533MHz).
As a result, for N = (76 ~ 90), it can produce frequency range (1533MHz ~
1800MHz), which can be upconverted to (2300MHz ~ 2700MHz).
1533-1800MHz
Ref
20MHz
Loop
Filter
PFD
VCO
76 - 90
Div-N
Fout
2300-2700MHz
ΣΔ
Div-2
Figure 4-3 Designed Fractional-N Synthesizer Block
Figure 4-3 shows the Fractional-N PLL design block diagram, with N range
from 76 to 90 and
is the fraction.
Our goal now is to design the low pass loop filter in order to meet the previously
mentioned requirements.
36
4.2 IIR Low-Pass Filter Design
We consider the design of infinite impulse response (IIR) filters subject to upper
and lower bounds on the frequency response magnitude. The associated optimization
problems, with the filter coefficients as the variables and the frequency response bounds
as constraints, are in general non-convex. Using a change of variables and spectral
factorization, we can pose such problems as linear or nonlinear convex optimization
problems. As a result we can solve them efficiently (and globally) by recently
developed interior-point methods. The design procedure follows [36].
Let H(z) be the transfer function of an IIR digital filter. Assume H(z) has the form
∑
.
∑
(4.1)
where the numerator polynomial N(z) is of mth degree, and the denominator polynomial
D(z) is of nth degree. The a term in (4.1) can be set to 1.0 without any loss in
generality. The magnitude response of the filter is obtained by evaluating (4.1) on the
unit circle (i.e., for
|
exp
) , and taking its magnitude, thus giving
∑
|
exp
∑
exp
.
(4.2)
In many frequency domain filter design problems, the magnitude required of the
resulting filter can be approximated to a given magnitude function
,
is a monotically increasing function of
and bi’s) to minimize the quantity
consistent with that constraint inequality
a tolerance
,
, where
, with
for fixed
. Thus, the resultant approximation problem is selecting the filter coefficients (the ai’s
exp
exp
exp
,
37
.
(4.3)
Inequality (4.3) is generally evaluated over a union of disjoint subintervals of the band
.
The above approximation problem is a nonlinear one in that the filter
coefficients enter into the constraint equation nonlinearity. There are several methods to
solve the previous problem where a linear approximation problem can be defined by
considering the magnitude squared function of the filter. The derivations are presented
in reference [36].
The resultant magnitude squared function of the filter is
|
|
exp
|
∑
cos
∑
cos
.
.
Equation (4.4) shows that the magnitude squared function of the filter is a ratio
of trigonometric polynomials. It is also seen that both
and
, the numerator polynomial,
, the denominator polynomial, are linear in the unknown filter coefficients
{ci} and {di}. Now, linear programming technique can be used to determine ci’s and di’s
such that |
exp
| approximates a given magnitude squared characteristic
where the peak weighted error of approximation is minimized.
If we let
be the desired magnitude squared characteristic, then the
approximation problem consists of finding the filter coefficients such that
,
where
(4.5)
is a tolerance function on the error which allows for unequal weighting of
errors as a function of frequency.
Equation (4.5) can be expressed as a set of linear inequalities in the ci’s and di’s
as follows
38
,
,
Or
,
The additional linear inequalities
(4.6)
(4.7)
.
(4.8)
,
(4.9)
.
(4.10)
completely define the approximation problem.
Thus, the question of whether or not there exists a digital filter with magnitude
squared characteristic F(w) and tolerance function
is equivalent to the question of
whether or not there exists a set of filter coefficients satisfying the system of constraints
defined by (4.6)-( 4.10). The question can be answered by using linear programming
techniques. First, an auxiliary variable v is subtracted from the left side of each
constraint, forming the new set of constraints
,
,
.
.
,
.
.
.
The objective function
is chosen to be minimized under the constraints of
(4.11) –(4.14). Clearly a solution to constraints (4.6) - (4.10) exists if and only if the
minimum value of z under constraints (4.11) - (4.14) is zero. If the minimum value of
is 0, then a solution exists to the approximation problem and the filter coefficients may
be obtained directly as the output of the linear programming routine. If
solution to the approximation problem exists, and either
be modified in order to obtain a solution.
39
, or
, then no
, or both must
4.3 FIR Low-Pass Filter Design
We consider the problem of designing a finite impulse response (FIR) filter with
upper and lower bounds on its frequency response magnitude [35]: given filter length
,
N, find filter tap coefficients
frequency response
∑
,….,
, such that the
satifies the magnitude bounds
|
|
,
Ω
.
,π .
over the frequency range Ω of interest.
One conventional approach to FIR filter design is Chebychev approximation of a
desired filter response
, i.e., one minimizes the maximum approximation error
over Ω.
We present a new way of solving the proposed class of FIR filter design
problems, based on magnitude design i.e., instead of designing the frequency response
of the filter directly, we design its power spectrum |( ω | to satisfy the
magnitude bounds [35].
Let autocorrelation function
denote
,
where we take
around
transform of
,
.
. The sequence
,
,
Is the power spectrum of
is symmetric
. Note that the Fourier
|
| ,
. If we use r as our design variables, we can reformulate
the FIR design problem in RN as
40
,…..,
find
subject to
,
,
,
The non-negativity constraint
for the existence of
Once a solution
.
Ω
is a necessary and sufficient condition
satisfying (4.14) by the Fejér-Riesz theorem (see § 4 in [35]).
of (4.15) is found, an FIR filter can be obtained via spectral
factorization. An efficient method of minimum-phase spectral factorization is given in
Section 4 in [35].
4.3.1 LP formulation
A common practice of relaxing the semi-definite program (4.15) is to solve a
discretized version of it, i.e., impose the constraints only on a finite subset of the [0, ]
interval and the problem becomes
,…..,
find
subject to
where
,
,
,…..,
. Since
.
Ω
,
is a linear function in r for each i,
(4.16) is in fact a linear program and can be efficiently solved. When M is sufficiently
large, the LP formulation gives very good approximations of (4.15) in practice. A rule
of thumb of choosing M,
, is recommended in [42].
4.3.2 SDP formulation
We will show that the non-negativity of
LMI constraint and imposed exactly at the cost of
will use the following theorem.
41
for all
,
can be cast as an
⁄ auxiliary variables. We
Theorem 1 Given a discrete-time linear system
minimal and
. The transfer function
, , ,
, A stable,
, ,
satisfies
,
if and only if there exists real symmetric matrix P such that the matrix inequality
.
.
is satisfied.
For more information, proof of this theorem can be found in [35].
, , ,
In order to apply Theorem 1, we would like to define
in terms of
such that
.
An obvious choice is the controllability canonical form:
,
,
(4.19)
,
It can be easily checked that
, , ,
given by (4.19) satisfies (4.18) and all the
hypotheses of Theorem 1. Therefore the existence of
and symmetric
that satisfy the
matrix inequality (4.17) is the necessary and sufficient condition for
,
by Theorem 1.
42
,
Note that (4.17) depends affinely on
and . Thus we can formulate the SDP
feasibility problem:
Find
,
Subject to
with
, , ,
Ω
(4.20)
given by (4.19). The SDP feasibility problem (4.20) can be cast as an
ordinary SDP and solved efficiently.
43
5. Results Analysis
The design process is divided into three setps presented as follows: The first step is
to design IIR digital low pass filter using LP MATLAB-based algorithm. Second, FIR
digital low-pass filter design using LP and SDP algorithms. The final step is to simulate
the designed filters, discuss the results, and compare it with others. To simulate the
designed filters we constructed simulation module shown in Figure 5-1.
Filter Bank
Figure 5-1 PLL Frequency Synthesizer Simulation Model
The simulation module consists of:
1. Reference Frequency: Pulse generator is chosen to produce 20MHz reference
frequency.
2. Filter Bank: three filters are designed and separated by two manual switches as
shown in Figure 5-1.
3. Voltage Controlled Oscillator (VCO) with output signal amplitude equal to 1V,
quiescent frequency equal to 1.511 GHz, and input sensitivity equal to
10MHz/V.
4. Phase Detector: XOR type selected.
5. Frequency Divider which produces (synN + synM) values used to divide the
output of VCO. Where synN is an integer and synM is the fraction.
6. Sigma/Delta Modulator: to produce the required fraction synM.
44
7. The Gain formula
.
, where
output of IIR filter, FIR filter respectively.
,
. after the
Note that the output synthesized frequency from the previous simulation module
(1.533GHz – 1.8GHz) must be multiplied with 1.5 to obtain the required range (2.3GHz
– 2.7GHz), see Figure 4-3.
5.1 IIR low pass Filter
We begin the design process by using CVX (convex optimization tool) software,
which is a powerful tool developed by Michael Grant and Stephen Boyd to work under
MatLab environment. Such as any language, CVX has its own syntax, for more
information refer to CVX webpage [43]. IIR low pass Filter can be designed by
implementing the procedure described in the previous chapter.
We begin IIR filter design process using the following values:
M = 4; % nominator
N = 4; % denominator, order = N-1
wpass = .006*pi; % end of the passband
wstop = .2*pi; % start of the stopband
delta = 1;
% maximum passband ripple in dB (+/- around 0 dB)
where, wpass value complies with mobile WiMax system specifications as follows:
Ω
.
.
,
.
/ .
Ω
/ .
The maximum passband ripple was chosen very low to lower the noise in the overall
system. Note that, the stopband attenuation is not constrained here. Using IIR1 code
listed in the appendix, the resultant IIR filter Transfer Function (transformed to sdomain using zero-order hold):
.
.
.
.
note, 3rd order filter obtained.
45
.
.
.
The filter magnitude/Phase response is shown in Figure 5-2:
10
mag H(w) in dB
0
-10
-20
-30
-40
X: 0.6283
Y: -59.53
-50
-60
0
0.5
1
1.5
w
2
2.5
3
0
0.5
1
1.5
w
2
2.5
3
3
phase H(w)
2
1
0
-1
-2
-3
Figure 5-2 3rd order IIR Filter Magnitude/Phase Response
It is clear from Figure 5-2 that we can summarize the designed IIR filter
specifications as follow:
•
•
The maximum pass band ripple = 1dB with wpass = .
Stop band attenuation below -59.53 dB with wstop = .
/ .
/ .
This means that our obtained IIR filter is in complement with passband stage and
stopband stage. However, we are encouraged to simulate the obtained IIR filter with
mobile WiMax simulation block diagram shown in Figure 5-1.
For our bad luck, the simulation is very slow and does not work properly, as a result we
cannot obtain our expected output frequency. Yes the designed IIR filter is ok but not
for mobile WiMax system. Anyway, let us try to design another IIR filter with much
more stop band frequency (wstop = .
/ , chosen after several iterations).
46
Applying IIR1 code again with new stopband frequency, this will diverge from our
required optimal filter design specifications. The designed IIR filter transfer function
(transformed to s-domain using zero-order hold) is:
.
.
.
.
.
.
.
The filter magnitude/Phase response is shown in Figure 5-3:
10
mag H(w) in dB
0
-10
-20
-30
-40
X: 2.199
Y: -59.53
-50
-60
0
0.5
1
1.5
w
2
2.5
3
0
0.5
1
1.5
w
2
2.5
3
3
phase H(w)
2
1
0
-1
-2
-3
Figure 5-3 3rd order IIR Filter Magnitude/Phase Response with much higher
stopband frequency
By looking at Figure 5-3 we conclude that increasing wstop will result in a filter with
much wider transition band. Anyway, let us try to simulate it with Mobile WiMax
Simulation block diagram shown in Figure 5-1.
The simulation run properly and the correct output frequency obtained. Figure 5-4
shows the control signal of VCO input using current designed IIR filter.
47
2.5
X: 3.999e-006
Y: 2.204
Control Signal (v)
2
1.5
1
0.5
0
0
0.2
0.4
0.6
Time (s)
0.8
1
1.2
-5
x 10
Figure 5-4 The Control Signal of VCO input using Designed IIR Filter
Figure 5-4 shows that
•
The settling time is about
•
The rise time is very low ( .
•
.
).
The overshoot is eliminated (zero) which agrees with our design specifications
listed in Table 4-1.
As a result we conclude that designing IIR digital filter with narrow transition band
using LP is not a good idea because it does not work properly with mobile WiMax
simulation. One reason for this draw back is that IIR phase does not taken into
consideration (IIR phase consideration is not with in our thesis goal). Another way to
overcome this draw back is to use FIR filter instead of IIR filter.
5.2 FIR low-pass Filter
The design of FIR filter is implemented using two different methods described in
the previous chapter. Linear programming is used first in order to check if it is the best
way for optimal FIR filter design or we need to use Semi-definite programming
technique.
48
5.2.1 Linear Programming
In this section, we design FIR low-pass filter using Linear Programming (LP) to
replace IIR lowpass filter. After several iterations, the best outcome is obtained with
these parameters:
max_order = 22; % the proposed filter order (2n + 1)
wpass = 0.006*pi;
% passband cutoff freq (in radians)
wstop = 0.2*pi;
% stopband start freq (in radians)
delta = 0.4;
% max (+/-) passband ripple in dB
atten_level = -44.5;
% stopband attenuation level in dB
Note that, we added constraints on the passband and stopband regions and not between
them (transition region). By applying FIR1 code listed in the appendix, the resultant FIR
filter Impulse Response is shown in Figure 5-5:
0.1
0.09
0.08
0.07
h(t)
0.06
0.05
0.04
0.03
0.02
0.01
0
0
2
4
6
8
10
t
12
14
16
18
Figure 5-5 FIR Impulse Response (LP)
And the FIR filter magnitude/Phase response is shown in Figure 5-6:
49
20
10
mag H(w) in dB
0
X: 0.01885
Y: -0.4
-10
-20
-30
X: 0.6283
Y: -44.5
-40
-50
0
0.5
1
1.5
w
2
2.5
3
0
0.5
1
1.5
w
2
2.5
3
3
phase H(w)
2
1
0
-1
-2
-3
Figure 5-6 FIR Filter Magnitude/Phase Response (LP)
It is clear from Figure 5-6 that the designed FIR filter length is 21 taps where,
filter order equal to
•
•
, and n = 10. Figure 5-6 shows that:
.
The maximum passband ripple does not exceed 0.4 dB with
.
/ .
Stopband attenuation below -44.5 dB with
.
.
/ .
Let us try to simulate this FIR filter with Mobile WiMax Simulation block
diagram shown in Figure 5-1. The simulation work properly and correct output
frequency obtained. The simulation result of the control signal using this FIR filter is
shown in Figure 5-7.
50
2.5
X: 5.498e-007
Y: 2.223
Control Signal (v)
2
1.5
1
0.5
0
0
0.2
0.4
0.6
Time (s)
0.8
1
1.2
-5
x 10
Figure 5-7 The Control Signal of VCO input using Designed FIR Filter
We see from Figure 5-7 that:
•
The overshoot equal zero (eliminated).
•
The settling time is about .
•
The rise time equal .
.
.
It is clear from the results that using LP to design FIR filter is a good idea that results
FIR filter in complement with all mobile WiMax system requirements presented in
Table 4-1. In other words, we conclude that using LP to design FIR digital filter will
improve settling time, rise time and eliminate overshoot. In the next section, we will use
SDP programming technique to answer the next question. Is SDP better than LP or not?
5.2.2 SDP Programming (LMI)
In this section, we use SDP method to design FIR low-pass filter. We implement
the SDP algorithm with these parameters:
n = 19; % proposed filter length (order)
wpass = 0.006*pi; % end of the passband
51
wstop = 0.2*pi; % start of the stopband
delta = 0.4;
% maximum passband ripple in dB (+/- around 0 dB)
Using FIR2 code listed in the appendix, the resultant FIR filter Impulse Response
shown in Figure 5-8:
0.1
0.09
0.08
0.07
h(t)
0.06
0.05
0.04
0.03
0.02
0.01
0
0
2
4
6
8
10
12
14
16
18
t
Figure 5-8 FIR Impulse Response (LMI)
And the FIR filter magnitude/Phase response is shown in Figure 5-8:
mag H(w) in dB
0
X: 0.01885
Y: -0.4
-10
-20
-30
X: 0.6283
Y: -44.5
-40
-50
0
0.5
1
1.5
w
2
2.5
3
0
0.5
1
1.5
w
2
2.5
3
3
phase H(w)
2
1
0
-1
-2
-3
Figure 5-9 FIR Filter Magnitude/Phase Response (LMI)
52
It is clear from Figure 5-9 that we obtained 19 taps FIR filter order (fewer than LP FIR
langth) with:
•
•
.
Maximum passband ripple does not exceed 0.4 dB with
.
/ .
.
Stopband attenuation below -44.5 dB with
.
/ .
Let us try to simulate it with Mobile WiMax Simulation block diagram shown in
figure 5-1. The simulation work properly and we got the correct output frequency. The
control signal using this FIR filter is shown in Figure 5-10.
2.5
X: 4.998e-007
Y: 2.194
Control Signal (v)
2
1.5
1
0.5
0
0
0.2
0.4
0.6
Time (s)
0.8
1
1.2
-5
x 10
Figure 5-10 The Control Signal of VCO input using Designed FIR Filter (LMI)
Figure 5-10 shows that:
•
•
•
Overshoot is also eliminated.
The rise time =0.25
The settling time = .
(agree with FIR filter designed by LP).
(better than value obtained using LP).
It is clear that using SDP technique to design FIR filter is better than using LP technique
as result of the following:
1. Filter order is reduced from 21 taps to 19 taps.
53
2. Settling time is enhanced (lowered) from .
SDP.
3. Rise time is enhanced (lowered) from .
with LP to .
with LP to .
with
with SDP.
The next section discusses results in more detail and compares it with others.
5.3 Discussion
Simple PLL contains passive low pass loop filter to eliminate high frequency
components and pass low frequency components into VCO in order to generate much
higher frequency based on the input dc voltage. To eliminate the VCO noise and
external noise we must take care of the loop filter design. Modern communication need
to corporate much faster PLL as frequency synthesizer to produce many high
frequencies. Designers of the first GPS system incorporated simple PLL architecture.
The main challenge was how to improve the performance and reduce noise. Low pass
loop filter was the major key for their designs in order to reduce noise and improve the
behavior of the overall system. Chou et al [32] introduced the design procedure for the
simple GPS PLL, the loop filter was designed using LMI method, and the design is a
good choice for reducing bandwidth noise. We tried to use their procedure for Integer-N
frequency synthesizer however, the design introduced very high bandwidth noise up to
thousands resulted from the N divider. The noise was increased by a value equal
log
which can be reduced (about 45dB) by replacing the Integer-N with N-
fractional frequency synthesizer. In other words, the Chou’s procedure is not a good
choice for N-fractional frequency synthesizer design. More immunity to noise is the
main advantage of Digital over analog filter. For this reason we replaced loop filter with
digital low pass IIR or FIR filter. IIR digital lowpass filter was designed using Linear
Programming (LP) technique. From the design results, we concluded that increasing the
filter order have somehow negligible effect on the overall filter performance, instead, it
will increase the complexity of the filter. To this end, the most good low pass behavior
can be obtained with 3rd order IIR filter. The designed IIR filter with narrow transition
band does not work properly with mobile WiMax systems. IIR filter with much wider
transition band works properly with mobile WiMax system with much degradation in
performance compared to FIR filter. Linear programming was used for FIR filter design
with filter length (order) increased to about 21 taps. Much high FIR taps provide the
54
system with large delay and more complexity to implement. Semi-Definite
Programming SDP (especially LMI) is the best way required to obtain optimal FIR filter
design with lower filter length. The design problem of loop filter for mobile WiMax
system using LP and SDP did not discussed previously. To our best knowledge this
study is the first focused on the usage of LP and SDP (LMI) to design loop filter that
incorporated into frequency synthesizer. As a result, we will present our design
significance and benefits in comparison to other relative methods such as for GPS and
other systems.
Beginning with IIR digital lowpass filter, we conclude, from Figure 5-3, that the
settling time is about
which is very low compared with Chou’s [32] result (
).
Chou designed loop filter for GPS using LMI technique. On the other hand, the rise
time obtained from our design ( .
) is higher than Chou’s result ( .
).
Moreover, the maximum overshoot obtained by IIR filter design (eliminated) is good
compared to Chou’s value (about 21%). This elmination of overshoot is good for GPS
as well as for mobile WiMax requirements. The standard 3rd order loop filter for mobile
WiMax system designed by Staggs [41] had lower overshoot (about 25 %) than our
design. On the other hand, the settling time is about (
obtained value (about 4
) which is not better than our
).
Design and simulation of fractional-N PLL frequency synthesizers presented by
Kozak’s [31] had larger overshoot (about 29.6 %) than our design. On the other hand,
the rise time is better than our obtained values. In contrast, our obtained settling time is
better than Kozak’s value. Table 5-1 present the comparison between IIR digital low
pass filter design and others.
Table 5-1 Comparison between IIR digital filter design using LP and other designs
IIR filter design
Chou’s
Staggs
Kozak
Settling time
Rise time
Maximum
.
.
overshoot
55
21%
.
25%
29.6%
Because IIR digital lowpass filter designed with narrow transition band does not
work properly with mobile WiMax system, we designed FIR digital low pass filter by
using LP technique in order to work properly with mobile Wimax system. FIR digital
filter results showed that the maximum overshoot (eliminated) and the settling time
( .
) are much better than IIR digital filter. In addition, they were better than
Chou’s, Staggs’s, and Kozak’s values. Unfortunately, rise time was degraded from that
obtained by Chou’s, Staggs’s values. On the other hand, it is better than Kozak’s and
our IIR result. Table 5-2 shows the comparison between FIR digital filter design using
LP and other designs.
Table 5-2 Comparison between FIR digital filter design using LP and other designs
FIR filter
Chou’s
design (LP)
Settling time
Rise time
.
Maximum overshoot
.
.
Zero
Staggs
.
21%
25%
Kozak
29.6%
To obtain much better filter performance, we designed another FIR digital lowpass
filter using SDP (LMI). The maximum overshoot (eliminated) is better than previously
mentioned designs and is the same compared with IIR/FIR digital filter designed using
LP. The settling time ( .
) and the rise time ( .
) are better than FIR filter
designed by LP. To this end, the FIR filter designed by SDP is the best for mobile
WiMax system. Table 5-3 shows the comparison between FIR digital filter design using
SDP and other designs.
Table 5-3 Comparison between FIR digital filter design using SDP and other
designs
FIR filter
design (SDP)
Settling time
Rise time
Maximum overshoot
.
Chou’s
.
.
56
21%
Staggs
.
25%
Kozak
29.6%
From the previous discussion, we conclude that using FIR digital lowpass filter
designed with LP will improve the transient behavior of the overall system. Much better
transient performance with can be achieved with FIR lowpass filter designed using
SDP.
57
6. Conclusion and Future Work
Phase locked loop remained an interesting topic for the research, as it covered
many discipline of electrical engineering such as communication theory, control theory,
signal analysis, design with transistors and op amps, digital circuit design and nonlinear analysis.
6.1 Conclusion
A new loop filter design method for frequency synthesizer was introduced taking
into consideration various design objectives: small settling time, small overshoot and
meeting mobile WiMax requirements. IIR and FIR digital low pass filters were
designed using linear programming and semi-definite programming.
Simulations showed that IIR digital lowpass filter with narrow transition band is
not good choice for mobile WiMax system. On the other hand, simulations showed that
FIR digital lowpass filter utilizing linear programming managed to improve the
transient behavior A FIR digital lowpass filter utilizing semi-definite programming
(LMI) much improve transient performance and fit best mobile WiMax systems.
6.2 Future work
We can summarize our suggestions for future works as follows:
It is recommended for IIR low pass digital filter design to extend the constrained
region to include the transition band between pass band and stop band, and to take into
consideration the filter phase. Although, IIR digital filter was designed using traditional
Linear Programming, we can expand the design and add LMI-based constraints.
The designed low-pass digital filters were compatible with Mobile WiMax systems
working for (2.300 GHz – 2.700 GHz); however, it can be extended to include much
higher frequency bands. On other hand, the proposed procedure can be examined with
Fixed WiMax system.
58
7. References
Bystrom, M. (2001). Maja Byston. Retrieved April 21, 2008, Boston University
[1]
: http://iss.bu.edu/bystrom/WirelessCommunications.doc
Al-Bin-Ali, F. (2004). Inside Wireless Evolution. Retrieved Octobar 31, 2008,
[2]
from The Association for Computing Machinery:
http://www.acm.org/crossroads/xrds9-4/wirelessevolution.html
Wiax Forum. (n.d.). Retrieved Octobar 31, 2008, from WiMax Forum:
[3]
http://www.Wimaxforum.org/home/
IEEE groups. (n.d.). Retrieved Octobar 31, 2008, from IEEE Group:
[4]
http://grouper.ieee.org/groups/802/16/
Tektronix. (n.d.). Characterizing Phase Locked Loops Using Tektronix Real-
[5]
Time
Spectrum
Analyzers.
Retrieved
April
21,
2008,
tektronix:
http://www.tek.com/Measurement/App_Notes/37_18170/eng/37W_18170_1.pdf
Digital PLL. (n.d.). Retrieved April 21, 2008, University of Aberdeen :
[6]
http://www.erg.abdn.ac.uk/users/gorry/course/phy-pages/dpll.html
Boyd, S., El Ghaoui, L., Feron, E., & Balakrishnan, V. (1994). Linear Matrix
[7]
Inequalities in System and Control Theory. Philadelphia: SIAM.
Vandenberghe, L., & Balakrishnan, V. (1998). Linear matrix inequalities for
[8]
signal processing : an overview. 32nd Annual Conference on Information Sciences
and Systems. Princeton: CISS.
Ferguson, T. (n.d.). Ferguson homepage. Retrieved MAY 01, 2008, UCLA
[9]
Department of Mathematics: http://www.math.ucla.edu/~tom/LP.pdf
[10]
Vandenberghe, L., & Boyd, S. (1996). Semidefinite programming. Society for
Industrial and Applied Mathematics , 38, pp. 49-95.
59
[11]
Best, R. E. (2007). Phase-Locked Loops: Design, Simulation, and Applications,
Sixth Edition. New York: McGraw Hill International.
[12]
T. Wang, T. Chiou, and S. Lall. (2008). Nonlinear Phase-Locked Loop Design
using Semidefinite Programming. Proceedings of the Mediterranean Conference on
Control and Automation , 1640 - 1645.
[13]
Abbas-Turki, M., Duc, G., & Clement, B. (2005). LMI formulation for
designing controllers according to time response and stability margin constraints.
44th IEEE Conference on Decision and Control, 2005 and 2005 European Control
Conference, (pp. 5740 - 5745). Seville, Splain ,France.
[14]
Henrion, D., Kucera, V., & Molina-Cristobal, A. (2004). Optimizing
simultaneously over the numerator and denominator polynomials in the YoulaKučera paramertrization. 43rd IEEE Conference on Decision and Control, 2, pp.
2177 - 2181. Nassau.
[15]
Abramovitch, D. (1988). Analysis and design of a third order phase-lock loop.
Military Communications Conference. 2, pp. 455-459. San Diego, CA, USA: Ford
Aerosp. Corp., Palo Alto, CA.
[16]
Abramovitch, D. (1990). Lyapunov redesign of analog phase-locked loops.
IEEE Trans. on Communication, 38, pp. 2197-2202.
[17]
Parks, P. (1966). Lyapunov redesign of model reference adaptive control
systems. IEEE Transaction Automatic Control, 11, pp. 362- 367.
[18]
Hsieh, G.-C., & Hung, J. (1996). Phase-locked loop techniques : a survey. IEEE
Transaction On Industrial Electronics, 43, pp. 609-615.
60
[19]
Suplin, V., & Shaked, U. (2000). Mixed H∞ /H2 design of digital phase locked
loops with polytopic type uncertainties. Proceedings of the 39th IEEE Conference
on Decision and Control, 2, pp. 1930 - 1935.
[20]
Raphaeli, D., & Yaniv, O. (1999). Near-optimal PLL design for decision
feedback carrier and timing recovery. IEEE International Conference on
Communications, 3, pp. 1515 - 1520.
[21]
Wu, N. (2002). Analog phaselock loop design using Popov criterion.
Proceedings of the American Control Conference, 1, pp. 16 - 18.
[22]
Mansuri, M., & Ken, C.-K. (2002). Jitter optimization based on phase-locked
loop design parameters. IEEE Journal of Solid-State Circuits, 37, pp. 1375 - 1382.
[23]
Fahim, A., & Elmasry, M. (2003). A fast lock digital phase-locked-loop
architecture for wireless applications. IEEE Trans. on circuits and systems-II:
Analog and Digital Signal Processing, 50, pp. 63 - 72.
[24]
Colleran, D., Portmann, C., Hassibi, A., Crusius, C., Mohan, S., Boyd, S., et al.
(2003). Optimization of phase-locked loop circuits via geometric programming.
Proceedings of the Custom Integrated Circuits Conference (CICC), (pp. 377-380).
[25]
Karimi-Ghartemani, M., Karimi, H., & Iravani, M. (2004). A magnitude/phase-
locked loop system based on estimation of frequency and in-phase/quadrature-phase
amplitudes. IEEE Trans. On Industrial Electronics , 51, pp. 511- 517.
[26]
Piqueira, J. R., & Monteiro, L. H. (2006). All-pole phase-locked loops:
calculating lock-in range by using Evan's root-locus. IEEE International Journal of
Control, 79, pp. 822 - 829.
[27]
Wanchana, S., Benjanarasuth, T., Komine, N., & Ngamwiwit, J. (2007). PLL
Equivalent augmented system incorporated with state feedback designed by LQR.
International Journal of Control, Automation, and Systems, 5, pp. 161-169.
61
[28]
Kundert, K. (2006, August). Predicting the phase noise and jitter of PLL-based
frequency synthesizers. Retrieved Octobar 01, 2008, The Designer’s Guide
Community: http://www.designers-guide.org/
[29]
Barrett, C. (1999, August). Fractional/Integer-N PLL Basics. Retrieved June 03,
2008,
Texas
Instruments,
Wireless
Communication
Business
Unit:
http://focus.ti.com/lit/an/swra029/swra029.pdf
[30]
Lau, C. Y., & Perrott, M. H. (2003). Fractional-N frequency synthesizer design
at the transfer function level using a direct closed loop realization algorithm.
Proceedings Design Automation Conference (pp. 526 - 531). New York: ACM.
[31]
Kozak, M., & Friedman, E. (2004). Design and simulation of fractional-N PLL
frequency synthesizers. Proceedings of the 2004 International Symposium on
Circuits and Systems, 4, pp. IV- 780-3.
[32]
Chou, Y., Mao, W., Chen, Y., & Chang, F. (2006). A novel loop filter design for
phase-locked loops. IEEE International Conference on Systems, 4, pp. 2932 - 2938.
[33]
Long, N. P., Hyunsoo, K., Yang, E., & Hansil, K. (2007). LMI approach to H∞
optimal filter design for GPS receiver tracking loop. IEEE International Symposium
on Electrical & Electronics Engineering. HCM City, Vietnam.
[34]
Al-Baroudi, W. (1997). Digital Filter Design using LMI Based Techniques. A
thesis. Faculty of The Graduate Studies, King Fahd University of Petroleum &
Minerals.
[35]
Wu, S.-P., Boyd, S., & Vandenberghe, L. (1996). FIR filter design via
semidefinite programming and spectral factorization. Proceedings of the 35th IEEE
Decision and Control, 1, pp. 271 - 276.
62
[36]
Rabiner, L., Graham, N., & Helms, H. (1974). Linear programming design of
IIR digital filters with arbitrary magnitude function. IEEE Transactions on
Acoustics, Speech and Signal Processing, 22, pp. 117 - 123.
[37]
WiMax Forum. Retrieved May 1, 2008. WiMax Forum :
http://www.wimaxforum.org/home.
[38]
Telecoms of Korea. (2008). Mobile WiMAX Handbook. Mobile Monday Global
Summit. Kuala Lumpur, Malaysia.
[39]
Pareek, D. (2006). The Business of WiMAX. John Wiley & Sons, Ltd.
[40]
Finneran, M. F. (2004). WiMax versus WiFi: A comparison of technologies,
markets, and business plans. Retrieved June 03, 2008. dBrn Associates :
http://searchmobilecomputing.techtarget.com/searchMobileComputing/downloads/F
inneran.pdf
[41]
Staggs, E. (n.d.). Mobile WiMax radio phase locked loop design. Retrieved
March 2008, Ansoft: http://www.ansoft.com/converge/staggs_ansoft.pdf
[42]
Alkhairy, A., Christian, K., & Lim, J. (1991). Design of FIR filters by complex
Chebyshev approximation. Proceedings of the Acoustics, Speech, and Signal
Processing (pp. 1985 - 1988). Washington, DC, USA: IEEE Computer Society.
[43]
CVX: matlab software for disciplined convex programming. Retrieved May 1,
2008. Stanford university: http://www.stanford.edu/~boyd/cvx/.
63
8. Appendix
8.1 IIR1 MatLab Code
% Maximize stopband attenuation of a lowpass IIR filter
% "Linear Programming Design of IIR Digital Filters with Arbitrary
% Magnitude Functions" by Ayman AlQouqa
%
% Designs a lowpass IIR filter using spectral factorization method
where we:
% - minimize maximum stopband attenuation
% - have a constraint on the maximum passband ripple
%
%
minimize
max |H(w)|
for w in the stopband
%
s.t.
1/delta <= |H(w)| <= delta
for w in the passband
%
% where we now have a rational frequency response function:
%
%
H(w) = sum_{m=0}^{M-1} b_m exp{-jwm} / sum_{n=1}^{N-1} a_n exp{jwn}
%
% We change variables via spectral factorization method and get:
%
%
minimize
max R(w)
for w in the stopband
%
s.t.
(1/delta)^2 <= R(w) <= delta^2 for w in the passband
%
R(w) >= 0
for all w
%
% where R(w) is the squared magnited of the frequency response
% (and the Fourier transform of the autocorrelation coefficients r).
% We represent R(w) = N_hat(w)/D_hat(w), where now R(w) is a rational
% function since we deal with IIR filter (see the reference paper).
%
% Variables are coeffients of the numerator, denoted as c, and
% denominator, denoted as d. delta is the allowed passband ripple.
% This is a quasiconvex problem and can be solved using bisection.
%
clear;clc;
%*********************************************************************
% user's filter specs (for a low-pass filter example)
%*********************************************************************
% number of coefficients for the IIR filter (including the zeroth one)
% (also without loss of generality we can assume that d_0 = 1, which
% is the zeroth coefficient of the autocorrelation denominator)
M = 4; % nominator
N = 4; % denominator
wpass = .006*pi;
% end of the passband
wstop = .7*pi;
% start of the stopband
delta = 1;
% maximum passband ripple in dB (+/- around 0 dB)
%*********************************************************************
% create optimization parameters
%*********************************************************************
% rule-of-thumb discretization (from Cheney's Approx. Theory book)
64
sample_order =30;
m = 15*(sample_order);
w = linspace(0,pi,m)'; % omega
% A's are maiir1trices used to compute the power spectrum
Anum = [ones(m,1) 2*cos(kron(w,[1:M-1]))];
Aden = [ones(m,1) 2*cos(kron(w,[1:N-1]))];
% passband 0 <= w <= w_pass
ind = find((0 <= w) & (w <= wpass));
Ap_num = Anum(ind,:);
Ap_den = Aden(ind,:);
% passband
% transition band is not constrained (w_pass <= w <= w_stop)
% stopband (w_stop <= w)
ind = find((wstop <= w) & (w <= pi));
As_num = Anum(ind,:);
As_den = Aden(ind,:);
% stopband
%********************************************************************
% optimization
%********************************************************************
cvx_quiet(true);
% use bisection (on the log of vars) to solve for the min stopband
atten
Us_top = 1e-0; % 0 dB
Us_bot = 1e-6; % -60 dB (in original variables)
while( 20*log10(Us_top/Us_bot) > 1)
% try to find a feasible design for given specs
Us_cur = sqrt(Us_top*Us_bot);
% formulate and solve the magnitude design problem
cvx_begin
variable c(M,1)
variable d(N-1,1)
% feasibility problem
% passband constraints
(Ap_num*c) <= (10^(+delta/20))^2*(Ap_den*[1;d]); % upper constr
(Ap_num*c) >= (10^(-delta/20))^2*(Ap_den*[1;d]); % lower constr
% stopband constraint
(As_num*c) <= (Us_cur)*(As_den*[1;d]); % upper constr
% nonnegative-real constraint
Anum*c >= 0;
Aden*[1;d] >= 0;
cvx_end
% bisection
if ~any(isnan(c)) % feasible
fprintf(1,'Problem is feasible for stopband atten = %3.2f dB\n',
...
10*log10(Us_cur));
Us_top = Us_cur;
b = spectral_fact(c);
a = spectral_fact([1;d]);
else % not feasible
65
fprintf(1,'Problem not feasible for stopband atten = %3.2f dB\n',
...
10*log10(Us_cur));
Us_bot = Us_cur;
end
end
% display the max attenuation in the stopband (convert to original
vars)
fprintf(1,'\nOptimum min stopband atten is between %3.2f and %3.2f
dB.\n',...
10*log10(Us_bot),10*log10(Us_top));
disp('Optimal IIR filter coefficients are: ')
disp('Numerator: '), b
disp('Denominator: '), a
cvx_quiet(false);
%*********************************************************************
% plotting routines
%*********************************************************************
% frequency response of the designed filter, where j = sqrt(-1)
H = ([exp(-j*kron(w,[0:M-1]))]*b)./([exp(-j*kron(w,[0:N-1]))]*a);
% magnitude plot
figure(1)
subplot(2,1,1)
plot(w,20*log10(abs(H)), ...
[0 wpass],[delta delta],'r--', ...
[0 wpass],[-delta -delta],'r--', ...
[wstop pi],[10*log10(Us_top) 10*log10(Us_top)],'r--')
xlabel('w')
ylabel('mag H(w) in dB')
axis([0 pi -65 10]);
grid
% phase plot
subplot(2,1,2)
plot(w,angle(H))
axis([0,pi,-pi,pi])
xlabel('w'), ylabel('phase H(w)')
grid
%convert to analog
Ts=1e-008;
F=tf(b',a',Ts);
Fc=d2c(F);
66
8.2 FIR1 MatLab Code
% Minimize order of a linear phase lowpass FIR filter
% "Filter design" by Ayman Alqouqa
%
%
%
%
%
%
%
%
%
%
%
%
%
%
%
%
%
Designs a linear phase FIR lowpass filter such that it:
- minimizes the filter order
- has a constraint on the maximum passband ripple
- has a constraint on the maximum stopband attenuation
This is a quasiconvex problem and can be solved using a bisection.
minimize
s.t.
filter order n
1/delta <= H(w) <= delta
|H(w)| <= atten_level
for w in the passband
for w in the stopband
where H is the frequency response function and variable is
the filter impulse response h (and its order/length).
Data is delta (max passband ripple) and atten_level (max stopband
attenuation level).
%********************************************************************
% user's filter specifications
%********************************************************************
% filter order that is used to start the bisection (has to be
feasible)
max_order = 20;
wpass = 0.006*pi;
wstop = 0.2*pi;
delta = 0.4;
atten_level = -44.5;
% passband cutoff freq (in radians)
% stopband start freq (in radians)
% max (+/-) passband ripple in dB
% stopband attenuation level in dB
%********************************************************************
% create optimization parameters
%********************************************************************
m = 30*max_order; % freq samples (rule-of-thumb)
w = linspace(0,pi,m);
%*********************************************************************
% use bisection algorithm to solve the problem
%*********************************************************************
cvx_quiet(true);
n_bot = 1;
n_top = max_order;
disp('Rememeber that we are only considering filters with linear
phase, i.e.,')
disp('filters that are symmetric around their midpoint and have order
2*n+1.')
disp(' ')
while( n_top - n_bot > 1)
% try to find a feasible design for given specs
67
n_cur = ceil( (n_top + n_bot)/2 );
% create optimization matrices (this is cosine matrix)
A = [ones(m,1) 2*cos(kron(w',[1:n_cur]))];
% passband 0 <= w <= w_pass
ind = find((0 <= w) & (w <= wpass));
Ap = A(ind,:);
% passband
% transition band is not constrained (w_pass <= w <= w_stop)
% stopband (w_stop <= w)
ind = find((wstop <= w) & (w <= pi));
As = A(ind,:);
% stopband
% formulate and solve the feasibility linear-phase lp filter design
cvx_begin
variable h_cur(n_cur+1,1);
% feasibility problem
% passband bounds
Ap*h_cur <= 10^(delta/20);
Ap*h_cur >= 10^(-delta/20);
% stopband bounds
abs( As*h_cur ) <= 10^(atten_level/20);
cvx_end
% bisection
if strfind(cvx_status,'Solved') % feasible
fprintf(1,'Problem is feasible for n = %d taps\n',n_cur);
n_top = n_cur;
% construct the full impulse response
h = [flipud(h_cur(2:end)); h_cur];
else % not feasible
fprintf(1,'Problem not feasible for n = %d taps\n',n_cur);
n_bot = n_cur;
end
end
n = n_top;
fprintf(1,'\nOptimum number of filter taps for given specs is 2n+1 =
%d.\n',...
2*n+1);
cvx_quiet(false);
%********************************************************************
% plots
%********************************************************************
figure(1)
% FIR impulse response
plot([0:2*n],h','o',[0:2*n],h','b:')
xlabel('t'), ylabel('h(t)')
grid
figure(2)
% frequency response
H = exp(-j*kron(w',[0:2*n]))*h;
% magnitude
subplot(2,1,1)
plot(w,20*log10(abs(H)),...
68
[wstop pi],[atten_level atten_level],'r--',...
[0 wpass],[delta delta],'r--',...
[0 wpass],[-delta -delta],'r--');
axis([0,pi,-50,10])
xlabel('w'), ylabel('mag H(w) in dB')
grid
% phase
subplot(2,1,2)
plot(w,angle(H))
axis([0,pi,-pi,pi])
xlabel('w'), ylabel('phase H(w)')
grid
69
8.3 FIR2 MatLab Code
% Maximize stopband attenuation of a lowpass FIR filter (magnitude
design)
% "FIR Filter Design via Spectral Factorization and Convex
Optimization"
% by Ayman AlQouqa
%
% Designs an FIR lowpass filter using spectral factorization method
where we:
% - minimize maximum stopband attenuation
% - have a constraint on the maximum passband ripple
%
%
minimize
max |H(w)|
for w in the stopband
%
s.t.
1/delta <= |H(w)| <= delta
for w in the passband
%
% We change variables via spectral factorization method and get:
%
%
minimize
max R(w)
for w in the stopband
%
s.t.
(1/delta)^2 <= R(w) <= delta^2 for w in the passband
%
R(w) >= 0
for all w
%
% where R(w) is the squared magnited of the frequency response
% (and the Fourier transform of the autocorrelation coefficients r).
% Variables are coeffients r. delta is the allowed passband ripple.
% This is a convex problem (can be formulated as an LP after
sampling).
%
%*********************************************************************
% user's filter specs (for a low-pass filter example)
%*********************************************************************
% number of FIR coefficients (including the zeroth one)
n = 19;
wpass = 0.006*pi;
% end of the passband
wstop = 0.2*pi;
% start of the stopband
delta = 0.4;
% maximum passband ripple in dB (+/- around 0 dB)
%*********************************************************************
% create optimization parameters
%*********************************************************************
% rule-of-thumb discretization (from Cheney's Approx. Theory book)
m = 15*n;
w = linspace(0,pi,m)'; % omega
% A is the matrix used to compute the power spectrum
% A(w,:) = [1 2*cos(w) 2*cos(2*w) ... 2*cos(n*w)]
A = [ones(m,1) 2*cos(kron(w,[1:n-1]))];
% passband 0 <= w <= w_pass
ind = find((0 <= w) & (w <= wpass));
% passband
Lp = 10^(-delta/20)*ones(length(ind),1);
Up = 10^(+delta/20)*ones(length(ind),1);
Ap = A(ind,:);
% transition band is not constrained (w_pass <= w <= w_stop)
70
% stopband (w_stop <= w)
ind = find((wstop <= w) & (w <= pi));
% stopband
As = A(ind,:);
AA = [zeros(1,n-1);eye(n-2) zeros(n-2,1)];
BB = [1 ;zeros(n-2,1)];
%********************************************************************
% optimization
%********************************************************************
% formulate and solve the magnitude design problem
cvx_begin sdp
variable r(n,1)
CC = r(2:n)'
DD = 0.5*r(1)
variable P(n-1,n-1) symmetric;
% this is a feasibility problem
minimize( max( abs( As*r ) ) )
subject to
% passband constraints
Ap*r >= (Lp.^2);
Ap*r <= (Up.^2);
% nonnegative-real constraint for all frequencies (a bit
redundant)
A*r >= 0;
[P-AA'*P*AA, CC'-AA'*P*BB;...
CC-BB'*P*AA, DD+DD'-BB'*P*BB]>=0;
cvx_end
% check if problem was successfully solved
disp(['Problem is ' cvx_status])
if ~strfind(cvx_status,'Solved')
return
end
% compute the spectral factorization
h_lmi = spectral_fact(r);
% compute the max attenuation in the stopband (convert to original
vars)
Ustop = 10*log10(cvx_optval);
fprintf(1,'The max attenuation in the stopband is %3.2f
dB.\n\n',Ustop);
%*********************************************************************
% plotting routines
%*********************************************************************
% frequency response of the designed filter, where j = sqrt(-1)
H = [exp(-j*kron(w,[0:n-1]))]*h_lmi;
figure(1)
% FIR impulse response
plot([0:n-1],h_lmi','o',[0:n-1],h_lmi','b:')
xlabel('t'), ylabel('h(t)')
grid
figure(2)
% magnitude
subplot(2,1,1)
plot(w,20*log10(abs(H)), ...
[0 wpass],[delta delta],'r--', ...
71
[0 wpass],[-delta -delta],'r--', ...
[wstop pi],[Ustop Ustop],'r--')
xlabel('w')
ylabel('mag H(w) in dB')
axis([0 pi -50 5])
grid
% phase
subplot(2,1,2)
plot(w,angle(H))
axis([0,pi,-pi,pi])
xlabel('w'), ylabel('phase H(w)')
grid
72
8.4 List of Acronyms
Acronyms
Meaning
BPL
Broadband over Power Line
CPE
Customer Premises Equipment
CSMA
Carrier Sense Multiple Access
CVX
Convex software
DES
Data Encryption Standard
DL
Down Link
DSCP
Differentiated Services Code Point
DSL
Digital Subscriber Line
EAP
Extensible Authentication Protocol
FDD
Frequency Division Duplexing
FIR
Finite Impulse Response
GPS
Global Positioning System
IIR
Infinite Impulse Response
IP
Internet Protocol
LAN
Local Area Network
LF
Loop Filter
LMI
Linear Matrix Inequality
LOS
Line of Sight
LP
Linear Programming
MAC
Media Access Control
MAN
Metropolitan Area Network
MIMO
Multi-Input Multi-Output
MPLS
Multi Protocol Label Switching
NLOS
Non Line of Sight
OFDMA
Orthogonal Frequency Division Multiple Access
73
PFD
Phase/Frequency Detector
PLL
Phase Locked Loop
QOS
Quality of Service
RAS
Remote Access Service
SDP
Semi-Definite Programming
SIM
Subscriber Identity Module
TDD
Time Division Duplexing
TDMA
Time Division Multiple Access
UP
Up Link
USIM
Universal Subscriber Identity Module
VCO
Voltage Controlled Oscillator
VoIP
Voice Over IP
WEP
Wired Equivalent Privacy
WiFi
Wireless Fidelity
WiMax
Worldwide Interoperability for Microwave Access
WMAN
Wireless Metropolitan Area Network
WPA
Windows Privacy Activation
XOR
Exclusive OR
74