Escolar Documentos
Profissional Documentos
Cultura Documentos
2005 8
OFDMA-IOTA OFDM/CDMA-IOTA
Performance Comparison between OFDMA-IOTA and
OFDM/CDMA-IOTA Systems
2005 8
2005 8
: ________________________
: ________________________
: ________________________
Abstract
Although conventional OFDM modulation scheme using guard interval is robust
to channel induced ISI (Inter-Symbol Interference) in time-domain, it is very sensitive
to ICI (Inter-Carrier Interference) due to doppler effect in frequency domain.
OFDM/OQAM-IOTA modulation scheme utilizes IOTA (Isotropic Orthogonal
Transform Algorithm) filter that has orthogonality in time and frequency domain so
that it is robust to delay spread and doppler effect. OFDMA-IOTA system can increase
bandwidth efficiency since it does not use guard interval.
In this paper, we show the design of IOTA filter and the structure of
OFDM/OQAM-IOTA system and perfect channel estimation for it. The advantage and
performances of OFDMA-IOTA are investigated by the comparison with the
conventional OFDM modulation.
Simulation results show that OFDMA-IOTA system outperforms OFDM system.
And finally, OFDMA-IOTA and OFDM/CDMA-IOTA are compared on their BER
performance. Due to the inherent orthogonality in OFDMA-IOTA, OFDMA-IOTA
performs better than OFDM/CDMA-IOTA under fair comparison condition of the
same transmission data rate.
Contents
Abstract
......................................................................................................................
Contents .........................................................................................................................ii
Lists of Figures ................................................................................................................iv
List of Tables ..................................................................................................................vi
Chapter 1 Introduction...................................................................................................1
Chapter 2 Wavelet Basics................................................................................................3
2.1 Wave & Wavelet .................................................................................................3
2.2 Scaling Function & Wavelet Function ..............................................................3
2.2.1 Scaling Function ........................................................................................ 3
2.2.2 Wavelet Function ....................................................................................... 5
2.3 QMF Filter Bank ................................................................................................7
2.4 Wavelet Transform & Wavelet Packets...........................................................10
2.4.1 Fast Wavelet Transform........................................................................... 10
2.4.2 Wavelet Packet ......................................................................................... 12
2.5 Polyphase for Filter Banks...............................................................................17
Chapter 3 IOTA Filter Design......................................................................................20
3.1 IOTA Filter ........................................................................................................20
3.2 Polyphase IOTA Filter Design..........................................................................24
Chapter 4 Design of OFDM/OQAM-IOTA System ....................................................28
4.1 OFDM/QAM System ........................................................................................28
4.2 OFDM/OQAM-IOTA System...........................................................................29
ii
iii
Lists of Figures
Figure 2.1. The nested function spaces spanned by a scaling function ............................5
Figure 2.2. The relationship between scaling and wavelet function spaces .....................5
Figure 2.3. (a) A two-band filter bank for one-dimentional subband coding and
decoding, and (b) its spectrum splitting properties .........................................8
Figure 2.4. An FWT analysis bank ................................................................................. 11
Figure 2.5. The inverse FWT synthesis filter bank ........................................................12
Figure 2.6. A two-stage or two-scale FWT analysis bank ..............................................13
Figure 2.7. A coefficient (a) and analysis (b) tree for the two-scale FWT analysis
bank of Fig. 2.6 .............................................................................................14
Figure 2.8. A three-scale FWT filter bank: (a) block diagram; (b) decomposition
space tree; and (c) spectrum splitting characteristics....................................14
Figure 2.9. A three-scale wavelet packet analysis tree ...................................................15
Figure 2.10. The (a) filter bank and (b) spectrum splitting characteristics of a
three-scale full wavelet packet analysis tree.................................................17
Figure 2.11. (a) Direct form of the analysis bank and (b) polyphase form of the
analysis bank.................................................................................................18
Figure 3.1. IOTA waveform and its Fourier transform...................................................24
Figure 3.2. Fourier transforms of IOTA function ...........................................................24
Figure 4.1. OFDM/QAM system....................................................................................29
Figure 4.2. OFDM/OQAM time and frequency lattices (compared to OFDM w/o
guard interval)...............................................................................................31
iv
List of Tables
vi
Chapter 1
Introduction
Wavelet has been a very hot topic in recent years. Its application ranges from the
function approximation, signal multiresolution representation, image compression to
signal processing and other fields. The popularity of wavelets is primarily due to the
interesting structure they provide based on dilation and location. A few investigators
have begun to exploit those features of wavelets that suggest their applications in
communications [1].
This thesis, is the result of a research work that of a design of wavelet-based
OFDM systems, classical OFDM modulation uses a guard interval (or cyclic prefix) to
efficiently combat the multi-path effect, at the price of a loss of spectral efficiency. It is
robust to channel induced ISI (Inter-Symbol Interference) in time-domain, but sensitive
to ICI (Inter-Carrier Interference) due to doppler effect in frequency domain. So we
design a novel wavelet function and utilize it to the OFDM system called OFDMAIOTA system. Since this prototype functionIOTA(Isotropic Orthogonal Transform
Algorithm) filter is very well localized both in the time and frequency domain to limit
the inter-symbol and inter-carrier interferences, and can also guarantee orthogonality
between sub-carriers, so OFDMA-IOTA is very robust to delay spread and doppler
effect. It can increase bandwidth efficiency since it does not use guard interval.
OFDMA-IOTA is an interesting alternative to classical OFDM, it has very good
characteristic, but more complexity may be induced at the reception. Classical channel
estimation can be based on scattered pilots or on a preamble, when it is applied straight
Chapter 2
Wavelet Basics
2.1 Wave & Wavelet
Wave is an oscillating function of time or space (eg. sinusoid). Fourier transform
uses the wave or sinusoid to represent a signal, the basis of fourier transform is
completely localized at the frequency domain, but not localized in time domain.
Wave analysis is useful for periodic, time-invariant, stationary signal.
Wavelet is a small wave (a pulse), which has its energy concentrated in particular
time. Wavelet transform uses the wavelets to represent a signal. Unlike the case of
Fourier transforms, there exists a large selection of wavelet families depending on the
choice of the mother wavelet. However, desirable properties such as orthogonality,
compactness of support, rapid decay, and smoothness impose a variety of restrictions.
Wavelet is a good tool to analyze transient, non-stationary, time-varying signal.
f ( x) = k k ( x)
k
(2.1)
where k is an integer index of the finite or infinite sum, the k are realvaluedexpansion coefficients, and the k ( x ) are real-valued expansion functions.
Consider the set of expansion functions composed of integer translations and
binary scalings of the real, square-integrable function k ( x ) ; that is, the set k ( x )
where [2]
(2.2)
Here k determines the position of j,k ( x ) along the x-axis, j determines j,k ( x ) s
widthhow broad or narrow it is along the x-axisand 2 j 2 controls its height or
amplitude. Because the shape of j ,k ( x ) changes with j , ( x ) is called a scaling
function. By choosing ( x ) wisely, { j ,k ( x )} can be made to span L2 (R ) , the set of
all measurable, square-integrable functions.
The scaling function usually obeys the four fundamental requirements of multiresolution analysis [2]:
z The scaling function is orthogonal to its integer translates.
z The subspaces spanned by the scaling function at low scales are nested within
those spanned at higher scales.
z The only function that is common to all V j is f ( x ) = 0 .
z Any function can be represented with arbitrary precision.
V V1 V0 V1 V2 V
(2.3)
V2
V1
V0
V j = Span{ j , k ( x )}
V0 V1 V2 ... V
Given a scaling function that meets the MRA requirements, we can define a
wavelet function ( x ) that, together with its integer translates and binary scalings,
spans the difference between any two adjacent scaling subspaces, V j and V j +1 . The
situation is illustrated graphically in Fig. 2.2. We define the set { j ,k ( x )} of wavelets
j ,k ( x) = 2 j / 2 (2 j x k )
(2.4)
V2
V1
W1
W0
V0
W j = Span{ j ,k ( x)}
V j +1 = V j W j
Figure 2.2. The relationship between scaling and wavelet function spaces.
for all k Z that spans the W j space in the figure. As with scaling function, we write
W j = Spank { j ,k ( x )}
(2.5)
f ( x ) = k j ,k (x )
(2.6)
The scaling and wavelet function subspaces in Fig. 2.2 are related by
V j +1 = V j W j
(2.7)
where denotes the union of spaces (like the union of sets). The orthogonal
complements of V j in V j +1 is W j , and all members of V j are orthogonal to the
members of W j . Thus,
j ,k ( x ), j ,l ( x ) = 0
(2.8)
weighted sum of shifted, double-resolution scaling functions. That is, we can write
( x) = h (n) 2 (2 x n)
n
(2.9)
where the h (n ) are called the wavelet function coefficients and h is the wavelet
vector. Using the condition that wavelet spans the orthogonal complement spaces in Fig.
2.2, and that integer wavelet translates are orthogonal, it can be shown that h (n ) is
related to h (n ) by
h (n ) = ( 1) h (1 n )
n
(2.10)
In the frequency domain, the space spanned by the scaling function is a lowpass
channel, while the space spanned by the wavelet is a bandpass channel.
Fig. 2.3 (a) shows the principal components of a two-band quadrature mirror filter
(QMF) bank. The input of the system is a one-dimensional, band-limited discrete-time
signal x(n ) for n = 0,1,2,...; the output sequence, x (n ) , is formed through the
decomposition of x(n ) into y0 (n ) and y1 (n ) via analysis filters h0 (n ) and h1 (n ) ,
and subsequent recombination via synthesis filters g 0 (n ) and g1 (n ) . Note that filters
h0 (n ) and h1 (n ) are half-band digital filters whose idealized transfer characteristics,
H 0 and H 1 , are shown in Fig. 2.3(b). Filter H 0 is a low-pass filter whose output is an
g 0 (n )
y 0 (n )
x(n )
x (n )
y1 (n )
(a)
h1 (n )
g1 (n )
H 1 ( )
H 0 ( )
(b)
Figure 2.3. (a) A two-band filter bank for one-dimensional subband coding and
decoding, and (b) its spectrum splitting properties.
The Z-transform, a generalization of the discrete Fourier transform, is the ideal tool
for studying discrete-time, sampled-data systems like the one in Fig. 2.3(a). The Ztransform of sequence x(n ) for n = 0,1,2,... is
X ( z ) = x(n )z n
(2.11)
where z is a complex variable. Our interest in the Z-transform stems from the ease with
[ ( ) (
1
X z1 2 + X z1 2
2
)]
(2.12)
where the double arrow indicates that the expressions on the left and right forma Ztransform pair. In a similar manner, upsamplingagain by a factor of 2is defined by
the transform pair
x(n 2 ) n = 0,2,4,...
x up (n ) =
otherwise
0
( )
X up ( z ) = X z 2
(2.13)
)
1
X ( z ) = [X ( z ) + X ( z )]
2
)
where x (n ) = Z 1 X ( z )
(2.14)
Z 1 [ X ( z )] = ( 1) x(n )
n
(2.15)
With this brief introduction to the Z-transform, consider again the subband coding
and decoding system of Fig. 2.3(a). In accordance with Eq. (2.14), we can express the
systems output as
)
1
1
X ( z ) = G0 ( z )[H 0 ( z ) X ( z ) + H 0 ( z ) X ( z )] + G1 ( z )[H 1 (z ) X ( z ) + H 1 ( z ) X ( z )] (2.16)
2
2
H 0 ( z )G0 ( z ) + H 1 ( z )G1 ( z ) = 0
(2.17)
H 0 ( z )G0 ( z ) + H 1 ( z )G1 ( z ) = 2
(2.18)
and, the following equations satisfy the requirement of quadrature mirror filters:
H1 ( z ) = H 0 ( z )
(2.19)
G0 ( z ) = H 0 ( z )
(2.20)
G1 ( z ) = H 0 ( z )
(2.21)
W ( j, k ) = h ( n ) W ( j + 1, n ) n= 2 k ,k 0
W ( j , k ) = h ( n ) W ( j + 1, n ) n= 2 k ,k 0
(2.22)
(2.23)
W ( j + 1, n )
h ( n )
h ( n )
W ( j , n )
W ( j , n )
system in Fig. 2.3(a). The synthesis and analysis filters must be time-reversed versions
of one another. Since the FWT analysis filters (see Fig. 2.5) are h0 (n ) = h ( n ) and
h1 (n ) = h ( n ) , the required FWT
h1 (n ) = h (n ) .
W ( j + 1, k ) = h (k ) WUP ( j , k ) + h (k ) WUP ( j , k )
k 0
(2.24)
W ( j , n )
h (n )
+
W ( j, n )
W ( j + 1, n )
h (n )
the partitioning of the time-frequency plane (e.g., smaller bands at the higher
frequencies), the FWT must be generalized to yield a more flexible decomposition
called a wavelet packet. The cost of this generalization is an increase in computational
complexity from O(M ) for the FWT to O(M log M ) [5].
f (n ) = W ( J , n )
h ( n )
W ( J 1, n )
h ( n )
h ( n )
W ( J 2, n )
W ( J 2, n )
W ( J 1, n )
h ( n )
Figure 2.6.
Consider the two-scale filter bank of Fig. 2.6, Figure 2.7(a) details the structure of
the tree and links the appropriate FWT scaling and wavelet coefficients [from Fig. 2.6]
to its nodes. The root node is assigned the highest-scale approximation coefficients,
which are samples of the function itself, while the leaves inherit the transforms
approximation and detail coefficients outputs. The lone intermediate node, W ( J 1, n ) ,
is a filter bank approximation that is ultimately filtered to become two leaf nodes. Note
that the coefficients of each node are the weights of a linear expansion that produces a
band-limited piece of root node f (n ) . Since any such piece is an element of a known
scaling or wavelet subspace, we can replace the generating coefficients in Fig. 2.7(a) by
the corresponding subspace. The result is the subspace analysis tree of Fig. 2.7(b).
Although the variable W is used to denote both coefficients and subspaces, the two
quantities are distinguishable by the format of their subscripts.
W ( J , n ) = f (n )
W (J 1, n )
VJ
W (J 1, n )
W (J 2, n ) W (J 2, n )
Figure 2.7.
WJ 1
VJ 1
WJ 2
VJ 2
A coefficient (a) and analysis (b) tree for the two-scale FWT analysis bank
of Fig. 2.6.
h ( n )
f (x ) VJ
WJ 1
h ( n )
h ( n )
WJ 2
V J 1
h ( n )
h ( n )
W J 3
h ( n )
V J 3
VJ 2
H ( )
VJ
VJ
WJ 1
VJ 1
WJ 2
VJ 2
V J 3
Figure 2.8.
W J 3
VJ 2
V J 3
VJ 1
W J 3
WJ 2
WJ 1
These concepts are further illustrated in Fig. 2.8, where a three-scale FWT analysis
bank, analysis tree, and corresponding frequency spectrum are depicted. The block
diagram of Fig. 2.8(a) is labeled to resemble the analysis tree in Fig. 2.8(b)as well as
the spectrum in Fig. 2.8(c). Thus, while the output of the upper-left filter and
VJ = VJ 1 WJ 1
(2.25)
VJ = VJ 2 WJ 2 WJ 1
(2.26)
VJ = VJ 3 WJ 3 WJ 2 WJ 1
(2.27)
They correspond to the one-, two-, and three-scale FWT decompositions. In general, a
P-scale analysis tree supports P unique decompositions.
Vj
V j 1
V j 2
W j 1
W j2
W J 1, A
W J 1, D
V j 3 W j 3 W j 2 , A W j 2, DW j 1, AAW j 1, AD W j 1, DA W j 1, DD
Figure 2.9.
Analysis trees are also an efficient mechanism for representing wavelet packets,
which are nothing more than conventional wavelet transforms in which the details are
iteratively filtered. Thus, the three-scale FWT analysis tree of Fig. 2.8(b) becomes the
three-scale wavelet packet tree of Fig. 2.9. The additional subscripting is introduced.
The first subscript of a double-subscripted node identifies the scale of the FWT parent
node from which it descended. The seconda variable length string of As and Ds
encodes the path from the parent to the node. An A designates approximation filtering,
while a D indicates detail filtering. Subspaces WJ 1, DA , for example, is obtained by
filtering the scale J 1 FWT coefficients (i.e., parent WJ 1 in Fig. 2.9) through an
additional detail filter (yielding WJ 1, D ), followed by an approximation filter (giving
WJ 1, DA ). Figures 2.10 (a) and (b) are the filter bank and spectrum splitting
characteristics of the analysis tree in Fig. 2.9. The evenly spaced frequency bands that
are characteristics of full packet decompositions.
The three-scale packet tree in Fig. 2.9 almost triples the number of decompositions
(and associated time-frequency tilings) that are available from the three-scale FWT tree.
The wavelet packet tree of Fig. 2.9 supports 26 different decompositions. For instance,
V j can be expanded as
VJ = VJ 3 WJ 3 WJ 2, A WJ 2, D WJ 1, AA WJ 1, AD WJ 1, DA WJ 1,DD
(2.28)
h ( n )
h ( n )
h ( n )
W J 1, DD
h ( n )
W J 1, DA
h ( n )
W J 1, AD
h ( n )
WJ 1, AA
h ( n )
WJ 2 ,D
h ( n )
W J 2, A
h ( n )
VJ 2
W J 3
h ( n )
V J 3
W J 1, D
W J 1
h ( n )
WJ 1, A
f ( x ) VJ
h ( n )
h ( n )
WJ 2
VJ 1
h ( n )
H ( )
VJ
VJ 1
VJ 2
VJ 3
0
W J 3
WJ 1
WJ 1, A
WJ 2
W J 2, A
WJ 1,D
WJ 2, D WJ 1, AA WJ 1, AD W J 1, DA W J 1, DD
Figure 2.10. The (a) filter bank and (b) spectrum splitting characteristics of a threescale full wavelet packet analysis tree.
The polyphase idea extends from one filter to a bank of filters. The polyphase form
does the downsampling first. In the block diagram of the filter bank, the decimators
move outside the filters. We can write C and D or H 0 and H 1 for the lowpass and
highpass filters:
(a) Direct
C
x(n )
D
(b) Polyphase
x(n )
y 0 (n )
y1 (n )
0 (n )
1 (n )
x0 (n )
with
Delay
x1 (n 1)
0 (n )
H p (z )
1 (n )
Figure 2.11. (a) Direct form of the analysis bank and (b) polyphase form of the
analysis bank.
V0 ( z ) C0 ( z )
V ( z ) = D ( z )
1 0
C1 ( z )
D1 ( z )
X 0 (z )
X 0 (z )
1
= H p ( z ) 1
z X 1 ( z )
z X 1 ( z )
(2.29)
This defines and displays H p (z ) . For FIR causal filters, the kind we expect to use, the
polyphase components are polynomials in z 1 . When the input x is also causal, the
outputs are causal.
The indices in X 0 and X 1 refer to even and odd. The indices in V0 and V1
refer to the two channels. This is normal for matrix multiplication, when H ij
multiplies X j and contributes to Vi . Rows of H p ( z ) go with channels, and columns
of H p ( z ) go with phases.
In an M-channel bank, i is the channel index and j is the phase index in
Chapter 3
IOTA Filter Design
3.1 IOTA Filter
Ot Gauss ( f ) =
Gauss ( f )
0 Gauss ( f k 0 )
(3.1)
O f x (t ) =
x(t )
0 x(t k 0 )
(3.2)
= O f Ot Gauss
(3.3)
(3.4)
(3.5)
0 = 0 = 1 2
(3.6)
(t ) = O f Ot .2 4 ef
2
with
Ot X ( f ) =
X( f )
1 2 X f k
k
(3.7)
and
x(t )
O f x(t ) =
1
2 x t k
(3.8)
1
4
here Gauss ( f ) = 2 e f .
In practice, we cannot numerically compute perfectly orthogonal IOTA function as
given by (3.6), and we have to use an approximate expression given as a finite series
expansion. The approximate IOTA function, denoted 1
Ks
(t ) = 1 d k ,1
2
2 k =0
k
k K s
d l ,1
G
t
G
t
+
+
auss
auss
2
1 2
1 2 l =0
t
(3.9)
cos 2l
1 2
jk
d k ,0 = bk , j e (2 j + k ) ,
0k K
(3.10)
j =0
5
8
35
64
63
128
to
429
1024
6435
16384
14
)
12155
32768
46189
131072
88179
262144
676039
2097152
123
128
763
1024
219
64
675
256
2289
1024
13875
8192
1395
2048
202281
131072
20691
32768
38753
65536
1400487
1048576
9765
2048
34871
8192
790815
262144
1434705
524288
457107
65536
161925
16384
596277
65536
3679941
262144
969375
131075
12097169
1048576
70545315
4194304
2067909 26060847
131072
1048576
394159701
16777216
51182445
4194304
13861065 87185895
2097152 8388608
56163
16384
7797
4096
76233
16384
6055
1024
1545
512
213
256
231
512
15
8
19
16
3
4
105
64
3
4
23600537
4194304
85037895
16777216
374325 5297445
2097152
262144
146289
262144
277797
524288
20050485
16777216
9895893
4194304
2641197
2097152
2120495
4194304
4063017
8388608
1300075
4194304
5014575
16777216
Figures 3.1 and 3.2 show the simulation results of various representations of the
IOTA function.
1.2
0.8
0.8
0.6
0.6
Am plitude
A m plitude
IOTA
1.2
0.4
0.4
0.2
0.2
-0.2
-4
-3
-2
-1
0
1
Symbol duration
Figure 3.1.
-0.2
-4
-3
-2
-1
0
1
Inter-carrier spacing
20
0
-20
Amplitude(dB)
-40
-60
-80
-100
-120
-140
-160
-4
Figure 3.2.
-3
-2
-1
0
1
Normalized intercarrier spacing
N
2 L = 2 ML coefficients of the
2
function 1 2 (t ) , equally spaced. Here N is the FFT length and M=N/2. The ideal IOTA
filter would be infinite but at it vanishes rapidly; it can be truncated to a reduced number
of coefficients. They are denoted k with k = 0 2ML 1 . We have for each k:
k =
2 0
1 2 (
k
L
)
2
M 2
(3.11)
1
M
M 1
2M
( 2 L 1) M
M +1
2 M +1
( 2 L 1) M +1
2 M 1
3M 1
2 LM 1
For example, with L=2 and M=4 (that is with an FFT size of N=2M=8), we would
have the buffer represented in Table 3.3.
Table 3.3. The buffer with L=2 and M=4.
0
12
13
10
14
11
15
If we denote C m,n the outputs of the IFFT where m is the sub-carrier index in
[02M1], an n the time index, we will have the following table after the first outputs
of the IFFT:
C4,04
C0,08
C4,012
C1, 0 1
C5,05
C1,09
C5,013
C 2, 0 2
C 3, 0 3
At each 0 the M outputs of the buffer corresponds to its first column, then the
columns of the buffer are shifted to the left.
C0,08 + C4,14
C4,012 + C0,18
C 4,1 12
C5,05 + C1,11
C1,09 + C5,15
C5,013 + C1,19
C 5,1 13
C6,06 + C2,12
C2,010 + C6,16
C6,014 + C2,110
C 6 ,1 14
C7,04 + C3,13
C3,011 + C7,17
C7,015 + C3,111
C 7 ,1 15
L 1
s kM +i = Ci ,k 2 q .2 qM +i + Ci + M ,k ( 2 q +1) .( 2 q +1) M +i
(3.12)
q =0
L 1
Ri , j = s i + jM + 2 qM i + 2 qM
(3.13)
q =0
N
, which illustrates that the
2
Chapter 4
Design of OFDM/OQAM-IOTA System
4.1 OFDM/QAM System
s (t ) =
n = + m = M 1
n =
m =0
d mn e ( 2im f t ) g (t nTu )
(4.1)
where d m , n is the transmitted OFDM symbol (Complex value) on the mth sub-carrier at
Classical OFDM modulation using guard interval is well known for its robustness to
multi-path time varying propagation channels. OFDM/OQAM-IOTA modulation is an
alternative to it, which has the advantage not to require the use of a guard interval, this
leads to a gain in spectral efficiency.
For this purpose, the prototype function modulating each sub-carrier must be very
well localized in the time domain, to limit the inter-symbol interference. Moreover, it
can be chosen very well localized in the frequency domain, to limit the inter-carrier
interferences (doppler effects, phase noise). This function must also guarantee
orthogonality among sub-carriers and among multi-carrier symbols. Functions having
these characteristics exist, which guarantee the orthogonality only in real domain.
Consequently, the complex QAM data stream (cmn) must be separated into its two real
components: real part (amn) and imaginary part (bmn.) (see Fig. 4.1), the imaginary part
being modulated with a half-symbol-duration (Tu/2) shifted version of the modulation
filter (thus the connotation Offset). It is important to notice that the density of the timefrequency frame related to OFDM/OQAM equals 2, i.e. 0 0 = 1 . In other words, this
2
modulation is as spectral efficient as OFDM/QAM without guard interval. Indeed, c.f.
Fig 4.1, for a given inter-carrier spacing 0 , on each sub-carrier, OFDM/OQAM carries
one real value each 0 while OFDM/QAM without guard interval carries one complex
value each 2 0 [12].
The classical OFDM signal (without cyclic prefix) can be expressed as (4.1), where
g(t) is a rectangular filter. By separating the two parts of (cmn), the corresponding
OFDM/OQAM modulated signal can be written as:
s(t ) =
n=+m= M 1
mn
n= m=0
T
i me( 2imft ) g (t nTu ) + ibmni me( 2imft ) g t + u nTu
2
(4.2)
M 1
d m,n = am ,n or bm,n
(4.3)
m ,n m ' ,n '
(4.4)
0
2 0 = Tu
Figure 4.2 illustrates the OFDM/IOTA system transmission /reception chain. The
demodulated data is obtained by taking the real part of the projection of the received
signal on the corresponding matched poly-phase IOTA filter. And at the reception,
d m,n
i m+n
d m ,n
s (t )
i m+ n
Complexity of the receive side is almost the same to the transmit side.
z FFT (Complex multiplication) :
The use of polyphase IOTA filter brings more complexity to both of the
transmitter and receiver of this novel scheme, the main parts including symbol
dephase part, IFFT, polyphase IOTA transmitter filter in the transmitter, and the
corresponding parts in the receive side.
In the transmitter, there are three parts, complexity increase mainly due to
the IFFT and polyphase IOTA filter.
mth
sub-carrier, n is the
time index.
The value of i m + n has 4 cases: i , i ,1,-1, hardware has already saved
these 4 values, so there is no need multiplication in symbol dephase part.
Hardware space retain 2 spaces for saving real & imaginary value, for
example, 2 complex value: a + bj , c + dj , we will save in the hardware space in
the follows:
Imag
Multiplicand
Multiplier
Product
ac-bd
ad+bc
z IFFT :
x[0]
x[1]
x[2]
x[3]
x[4]
x[5]
x[6]
x[7]
W
W
0
N
2
N
W
W
0
N
2
N
x[256]
x[257]
x[258]
WN248
WN249
WN250
251
N
x[259]
W
W
x[260]
x[261]
x[262]
x[263]
252
N
253
N
WN254
W
255
N
W N0
W N0
W N0
W N0
W N0
W N0
WN2
W N0
W N0
W N0
WN2
W N0
256 2 = 512 ;
Complex addition/symbol:
N (log 2 (N ) 1) = (512) (log 2 ( N ) 1) = 4096 ;
Real addition/symbol:
512 ;
x[256]
x[2]
x[258]
x[4]
x[260]
x[6]
x[262]
x[0]
x[1]
x[257]
x[3]
x[259]
x[5]
x[261]
x[7]
x[263]
s kM + i =
where
Ci , k :
L 1
C
q=0
i,k 2q
. 2 qM + i + C i + M , k ( 2 q + 1) . ( 2 q + 1) M + i
(4.5)
2 N 8 = 2 256 8 = 4096 ;
Complex Addition/symbol:
N 7 = 256 7 = 1792 ;
z Memory for poly-phase coefficients :
2048 C bits : where C : # of bits for coefficients;
In receiver side, the algorithm of IOTA receive filter is a little different, but
the complexity of FFT and polyphase receive filter is not much different
compared to that of the transmit side.
z FFT :
At the receiver side, the same filter is applied to the signal before the
FFT operation, but the algorithm is slightly different.
Ri, j =
L 1
s
q=0
i + jM + 2 qM
i + 2 qM
Complex Addition/symbol:
N 8 = 256 8 = 2048 ;
(4.6)
4.3.3 Summary
(N 2) log 2 (N ) 4 2 = 18432 ;
z Total real addition/symbol of OFDM system:
N log 2 (N ) 2 2 = 18432 ;
Complexity comparisons between OFDM and OFDM/IOTA are given in Table 4.2
and Table 4.3.
Table 4.2. Complexity comparison between OFDM and OFDM/IOTA (1).
OFDM
Tx
OFDM/IOTA
Rx
Tx
Rx
Polyp
IFFT
FFT
IFFT
hase fi
lter
Polyphase
filter
N 8
N 8
Real
FFT
N=51
Multiplication
=409
6
Compl
ex
(N 2) log2
=2304
(N 2) log
=2304
( N 2) (log2 N
=409
6
(N 2) log2 (N
=2048
multiplication
=2304
N=512
Real add
ition
Compl
N log 2 ( N
N log 2 ( N )
N 27
N (log2 N
N log2 (N )
N 28
ex
additio
=4608
=4608
=4096
=1792
=2048
=4608
OFDM/IOTA
Ratio
18,432
26,112
1.42
18,432
25,600
1.39
Total # of equivalent
real multiplication
Total # of equivalen
t
real addition
Chapter 5
OFDMA-IOTA System vs. OFDM/CDMAIOTA System
5.1 Ideal Channel Estimation Method
M 1
r (t ) = H m ,n a m ,n m ,n (t ) + b(t )
(5.1)
n m =0
where a m,n denotes the real information value (Offset QAM) sent on the m th subcarrier at the n th symbol, is the IOTA function. If (m0 ,n0 ) denotes a reference
sub-carrier, we must look at the complex projection of the received signal r (t ) on
m0 ,n0 to estimate H m0 ,n0 :
H m0 ,n0 =
r (t )g
m0 , n0
(t )dt
(5.2)
am0 ,n0
s(t )
m0 ,n0
(t )dt = am ,n
0
a )
) (
m ,n
( m ,n m0 ,n0
m ,n
m0 ,n0
(5.3)
The second item I m0 ,n0 represents the intrinsic interference, whose real part is null.
As the prototype function is chosen well localized, the main interfering term is due to
the sub-carriers adjacent to the (m0 , n0 ) sub-carrier.
C1m0 ,n0 =
( m,n )1m0 , n0
(5.4)
where 1m ,n represents the indexes of these 8 adjacent sub-carriers (1st order neighbor
0
1
m {1,..., M 2}, C m , n0 = 0
(5.5)
This structure depends on the parity of n0. Therefore, symbol sequence for the case of
even centre symbol is illustrated on Fig. 5.2(a), and for the case of odd symbol, symbol
sequence to reduce intrinsic ISI can be similarly formulated. Symbol sequence maybe
composed of BPSK symbols { e ;+ e }.
It can be demonstrated that these structures set every Nth order intrinsic ISI
0
n0 1
n0
t
n0 + 1
m0 1 m0
m0 + 1
f
Figure 5.1.
n0 3
n0 2
n0 1
n0
n0 + 1
n0 + 2
n0 + 3
n0 3
n0 2
n0 1
n0
n0 + 1
n0 + 2
n0 + 3
+ e
+ e
Based on the specific symbol sequence, the synoptic of this new ideal channel
estimation is illustrated on Figure 5.3.
H 2 n ,k
H n,k
H 2 n+1,k
In the following, we verify through simulations that this method provides reliable
bounds of performances. We test this symbol-sequence based channel estimation
method under AWGN and 1-path Rayleigh fading channel environment. It can obtain
good results, but under multipath channels, error appears. In this case, we can try to
apply similar method to realistic channel estimation with scattered pilot symbols [16].
We will go on this research work in the future.
The OFDM-CDMA transmitter spreads the original signal using a given spreading
code in the frequency domain. The basic transmitter structure of OFDM-CDMA scheme
is similar to that of a normal OFDM scheme. The main difference is that the OFDMCDMA scheme transmits the same symbol in parallel through many sub-carriers,
whereas the OFDM scheme transmits different symbols.
Similar to OFDMA-IOTA system, OFDM/CDMA-IOTA uses IOTA filter as the
pulse waveform, and IOTA filter is implemented by the poly-phase technique as used in
j
MC
GMC
(t ) = d mj ,n i m+ n cmj e j 2mft (t n 0 )
(5.6)
n m =1
where dm,nj denotes real information value (Offset QAM) sent on the mth sub-carrier at
the nth symbol of user j, cmj is Walsh code for the jth user, f is the sub-carrier
separation, and is the IOTA filter.
In receiver side of OFDM/CDMA, after the serial-to-parallel conversion, the mth
sub-carrier is multiplied by the gain qm to combine the received signal energy scattered
in the frequency domain. We believe this is the main advantage of OFDM/CDMA
scheme over other schemes. This is also used in our OFDM/CDMA-IOTA scheme. The
decision variable is given by [17]
GMC
(5.7)
m =1
(5.8)
'
5.3
Values
Carrier frequency
2GHz
Signal bandwidth
4.8MHz
Sampling frequency
7.68MHz
Vehicle speed
50 Km/h
Jakes spectrum
Modulation
QPSK/OQPSK
Convolutional Encoder
Channel coding
/ Viterbi Decorder
(K=7,[133 171])
Code rate
1/2
(QPSK/OQPSK)
Block Interleaver
Interleaver
Channel estimation
Perfect Compensation
Values
16.67 us
15 kHz
FFT size
512
Modulated sub-carrier
320
Code Rate
QPSK
1/2
OFDM/QPSK Information
Increase of Spectral
Bit Rate[Mbps]
Efficiency
3.84
4.8
25%
OFDM/QPSK vs OFDM-OQPSK/IOTA
10
OFDM-OQPSK/IOTA
OFDM/QPSK
-1
10
-2
10
BER
-3
10
-4
10
-5
10
-6
10
10
12
14
16
18
20
Eb/No
(a) AWGN
OFDM/QPSK vs OFDM/OQPSK-IOTA in 1 path Rayleigh fading
10
OFDM/OQPSK-IOTA
OFDM/QPSK
-1
10
-2
10
BER
-3
10
-4
10
-5
10
-6
10
10
15
20
25
Eb/N0
5.4
Performance
Comparison
between
OFDMA-IOTA
and
OFDM/CDMA-IOTA Systems
In Fig. 5.6, comparison scenario of OFDMA-IOTA system and OFDM/CDMAIOTA system are given.
Values
Carrier Frequency
2 GHz
Signal Bandwidth
4.8 MHz
Sampling Frequency
7.68 MHz
Vehicle Speed
60 Km/h
Modulation
QPSK
Jakes Method
Channel Estimation
Ideal Channel
Estimation
Values
FFT Size
512
320
Inter-carrier Spacing
15 KHz
66.66 us
Values
FFT Size
512
Inter-carrier Spacing
15 KHz
Number of Users
10
OFDMA-IOTA
OFDM/CDMA-IOTA (1 user)
OFDM/CDMA-IOTA (4 users)
-1
10
-2
10
BER
-3
10
-4
10
-5
10
-6
10
10
15
20
25
30
Eb/N0
10
OFDMA-IOTA
OFDM/CDMA-IOTA (1 user)
OFDM/CDMA-IOTA (4 users)
-1
10
-2
10
BER
-3
10
-4
10
10
15
Eb/N0
20
25
30
The performance of OFDM/CDMA- IOTA (1 user) is similar to that of OFDMAIOTA under both AWGN and 1-path Rayleigh fading channel. But in this case, due to
Chapter 6
Conclusions
This paper provides two novel systems: OFDMA-IOTA and OFDM/CDMA-IOTA
that use IOTA function as their prototype filter. IOTA filter is well known for its good
characteristic that can guarantee a quasi-optimal localization in Time/Frequency domain.
But IOTA filter only guarantee orthogonality on real values, that is why we use offset
QAM modulation(In this paper, we do the simulation by QPSK modulation).
OFDMA-IOTA and OFDM/CDMA-IOTA are very spectral efficient, but require
specific treatment at the reception, to limit what we called the intrinsic ISI. We adapt a
new channel estimation method that based on a specific symbol sequence. Finally
OFDMA-IOTA and OFDM/CDMA- IOTA are compared on their BER performance.
Through comparison, OFDMA-IOTA system outperforms OFDM system in both
AWGN and 1-path Rayleigh flat fading channel.
Due to the inherent orthogonality in OFDMA-IOTA, OFDMA-IOTA performs
better than OFDM/CDMA-IOTA under fair comparison condition of the same
transmission data rate.
It is an interesting challenge to utilize OFDM-IOTA as the novel wavelet platform
of 4G environment.
References
[1] KyungHi Chang, XuDuan Lin and HuiJuan Li, Wavelet-based multi-carrier CDMA
for personal communications systems, in Proc. IEEE Int. Conf. on Acoustics,
Speech, and Signal Processing, May. 1996, vol. 3, pp. 1443-1446.
[2] Rafael C.Gonzalez and Richard E. Woods, Digital image processing, Prentice Hall,
Apr. 1995.
[3] P.P Vaidyanathan, Multirate systems and filter banks, Prentice Hall, Sept. 1992.
[4] Gilbert Strang, Truong Nuyen and T. Ngugen, Wavelets and filter banks,
Wellesley-Cambridge Press, Oct. 1996.
[5] Alan V.Oppenheim and Ronald W.Schafer, Discrete-Time signal processing,
Prentice Hall, Feb. 1995.
[6] 3GPP TSG-RAN-1 Meeting #31, Technical description of the OFDM/IOTA
modulation, Feb. 2003.
[7] 3GPP TR 25.892 v 0.4.1, Oct. 2003.
[8] 3GPP TSG-RAN-1 Meeting #30, IOTA, an advanced OFDM modulation for
UTRAN enhancement: presentation and potentials, Jan. 2003.
[9] Pierre Siohan and Christian Roche, Derivation of extended Gaussian functions
based on the Zak transform, IEEE signal processing letters, vol. 11, pp. 401-403,
Mar. 2004.
[10] Christian Roche and Pierre Siohan, A family of extended gaussian functions with
a nearly optimal localization property, in Proc.1st Int. Workshop on Multi-Carrier
Spread-Spectrum, Apr. 1997, pp. 179-186.
[11] P.Siohan and C.Roche, Cosine modulated filterbanks based on extended Gaussian
functions, IEEE Trans. On Signal Processing, vol. 48, pp. 3052-3061, Nov. 2000.
[12] D.Lacroix, N.Goudard and M.Alard, OFDM with guard interval versus
OFDM/OffsetQAM for high data rate UMTS downlink transmission, in Proc.
IEEE Veh. Tech. Conf.-Fall, Oct. 2001, vol. 4, pp. 2682-2686.
[13] Xiao Zhou, Joo Heo and KyungHi Chang, OFDM/OQAM-IOTA System, Journal
of KICS, vol. 8, No. 2, pp. 120-128, Dec. 2004.
[14] Joo Heo, Xiao Zhou and KyungHi Chang, OFDM/OQAM-IOTA System, in Proc.
KONI Conf, Oct. 2004, vol. 8, No. 2, pp. 179-183.
[15] D.Lacroix, N.Goudard and M.Alard, A new channel estimation method for
OFDM/OQAM, in Proc. of 7th International OFDM workshop, Sept. 2002.
[16] Jean-Philippe Javaudin, D.Lacroix and Alexandre Rouxel, Pilot-Aided channel
estimation for OFDM/OQAM, in Proc. IEEE Veh. Tech. Conf. - Spring, Apr. 2003,
vol. 3, pp. 1581-1585.
[17] Ramjee Prasad and Shinsuke Hara, An overview of multi-carrier CDMA, in Proc.
IEEE 4th International Symposium on Spread Spectrum Techniques and
Applications, Sept. 1996, vol. 1, pp. 107-114.
[18] Xiao Zhou, Joo Heo and KyungHi Chang, 'OFDM/IOTA vs. MC/CDMA-IOTA
using a New Channel Estimation Method, in Proc. KICS Fall Conf, Nov. 2004,
vol. 30, 4A-4, p. 66.
Acknowledgements
This thesis is completed under my professor KyungHi Changs guidance and
encouragement. First I give my deepest regards and thanks to him, my dear professor,
who teach me to study in the field of mobile telecommunication for Masters degree in
Korea. During these two years, he encourages me to study hard, directs good life road
for me, gives me great help on guiding my studying and research, and supports my life.
During the process of writing this thesis, he is concerned all along and gives me many
valuable comments and suggestions. And I also want to say thanks to professors of
Graduate School of Information and Telecommunications, Inha University for their
directions and help.
My lab mates also give me a lot of good suggestions on my study and research.
When I meet difficulties in study and research, they always help me patiently. Here I
want to express my sincere thanks to them. They are Mr. Joo Heo, Mr. InSuk Cha, Mr.
SungHo Park, Mr. JungJu Gim, Mr. JungHo Noh, Mr. YuPeng Wang and Mr. YiXian
Chen. Thanks to my Chinese friends Mr. GuangYun Jin, Mr. WeiHua Zhang, Mr.
QingHai Yang, Mr. HuaMin Zhu and Mrs. ShaoYi Xu for their kind help on my study
and my 2 years study abroad life.
I am extremely grateful to my dear parents, whose love, endurance and sacrifice
has made my research possible. Finally, I dedicate the thesis to them for their hard
fostering, great direction and eager hope.