Documente Academic
Documente Profesional
Documente Cultură
AbstractــــIn OFDMA multiple access wireless communication of active users and, their corresponding subchannels, which
systems, carrier frequency offsets between the transmitter and limits its application. Throughout this paper, we will refer to
the receiver tend to destroy the orthogonality among subcarriers, the algorithm in Ref. [6] as the Cao-Tureli-Yao (CTY)
and hence, introduces intercarrier interference. A two-stage scheme. In the present paper we focus on the CFO estimation
frequency offset estimation algorithm based on subspace
in the uplink of interleaved-OFDMA systems. We propose a
processing is proposed. The main advantage of the proposed
method is that it can obtain the CFOs of all users simultaneously two-stage CFO-estimation algorithm based on subspace
using only one OFDMA block. Compared with previously known processing which maintains the advantage of estimating the
methods, our system is seen not to have a relatively low CFOs of all users simultaneously using only one OFDMA
implementation complexity, but is also suitable for random block without the need for any a priori information.
subchannel assignment.
II. SYSTEM MODELING
Key words
Orthogonal Frequency-Division Multiple Access (OFDMA); In this section, we introduce the signal model for the
Carrier Frequency Offset (CFO); Normalized Root Mean interleaved-OFDMA uplink following [6]. Consider an
Square Error (NRMSE). OFDMA system with N subcarriers and M users, where each
SS communicates with the BS through an independent
I. INTRODUCTION multipath channel. Suppose that the N subcarriers are divided
into Q subchannels, and each subchannel has P = N/Q
Orthogonal Frequency Division Multiple-Access (OFDMA) is subcarriers. For the interleaved subcarrier assignment, the q-th
selected as the physical-layer technology for the IEEE subchannel is composed of subcarriers with indexes
802.16/WiMAX family of standards[1,2], and is pursued as {q, Q + q,…… .,(P-1 )Q + q}, q = 0,1,… ,Q-1 .
the potential candidate for future mobile communication
systems.In OFDMA, subcarriers are assigned to different users Let superscript (.) ( m ) denotes the m-th user. Fig.1 depicts the
for parallel data transmission, and frequency guard- bands transmitter structure for the m-th user, with
between adjacent users are achieved. Moreover, s (m) = [ S 0(m) S1(m) " S P(m)−1 ] denoting the information block that the
in an Additive White Gaussian Noise (AWGN) channel, sub
carrier not employed, there by a high spectral efficiency is m-th user will transmit during one OFDMA block. This block
being achieved. Moreover, in an Additive White Gaussian is first mapped into a set of N modulation symbols: {X i(m) }iN=0−1 ,
Noise (AWGN) channel, subcarrier orthogonalities prevent where,
Inter Carrier Interferences (ICI) as well as Multi User ⎧S (m) , i = pQ + q ( m )
Interferences (MUI). But, OFDMA inherits the Weakness ⎪
X (m) = ⎨ g,p (1)
of OFDM, i.e., being sensitive to Carrier Frequency Offset g,i
⎪⎩0, otherwise
(CFO) induced by Doppler effects and/or poor oscillator
Alignments this, would, in turn, destroys the orthogonality And p = 0 ,1,…… P-1 , and q (m ) denotes the subchannel index
among subcarriers, resulting in ICI and MUI. In order for of user m. For a data bearing subcarrier, the modulation could
OFDMA to achieve a desirable performance, it has to have be a phase shift keying (PSK) or quadrature amplitude
strict frequency synchronization between the Subscriber modulation (QAM). For virtual subcarriers, the modulation
Station (SS), and the Base Station (BS) over both the uplink as symbols are effectively zero padded in the IFFT. For pilot
well as the downlink. As a result, CFO estimation schemes subcarriers, the modulation symbols are pilot symbols or
have become a major task in the OFDMA system. In Ref.[6], training symbols used for channel estimation purpose. The
Cao, etal. studied the CFO estimation for the uplink of the CP(Cyclic prefix) is, then, added to avoid inter-block
interleaved-OFDMA system, and proposed a CFO-estimation interference (IBI) caused by multipath fading. As a result, the
algorithm based on the signal structure. The main advantage of Baseband signal transmitted from the m-th user can be
this estimation algorithm is that, it can estimates the CFOs of represented as follow
all users simultaneously using only one OFDMA block.
However, it requires a priori information such as the number
m = 0 p =0
M −1 P −1
(7)
= ∑∑H (m)
p
S e(m)
p
j ( 2π / N ) ni
,
m =0 p =0
l = 1
l Υ ( m ) ( n ) = ∑ H p( m ) S p( m ) e N
p=0
(m )
(10)
where L is the total number of paths, α l(m ) and τ l(m ) , are j
2π
( q +ε ( m ) ) n P −1 j
2π
pn
the complex gain and time delay of the l-th path respectively.
=e N
∑H
p =0
(m)
p =0
(m)
S e
p
N
At the (uplink) receiver, the signal of one OFDMA block is Where n=0, 1… N-1. It is important to note that,
the superposition of signals from all M active users. Assume (m)
Υ ( m ) ( n + P ) = e j 2 π ( q +ε ) / Q Υ ( m ) ( n ) , (11)
all m-users are synchronized in time, the received (sampled)
Moreover,
signal in the absence of noise can be written as (m)
M L( m ) Υ ( m ) ( n + νP ) = e j 2πν ( q+ε ) / Q Υ ( m ) ( n ), (12)
Υ ( nTs ) = ∑ ∑ α l
(m)
( nTs )x
( m) ( m)
( nTs − τ l ) (5) Where ν is an integer. Equations (11), and (12), indicate that
m =1 l =1
Where Ts = T / N is the sampling interval. Consider the the received N-signal samples of one OFDMA block from the
m-th user, {Υ ( m ) ( n )}nN=−01 , have a special periodical structure
signal of one OFDMA block, and let H p( m ) denotes the
every P samples. Let R denotes the number of periods within
frequency response of the i-th subcarrier for the m-th user’s the N signal samples. We have R = N/P = Q .The number of
channel during one OFDMA block,
periods is, the number of sub-channels. {Υ ( m ) ( n )}nN=−01 Can thus at the receiver. {Υ ( m ) (n)}nN=−01 can also be arranged into
be arranged into a Q × P matrix a Q × P matrix, Y, in the same fashion as (13). From (14) and
⎡ Y( 0 ) ... Y(P − 1 ) ⎤ (16), the following relationship holds,
⎢ Y(p) ... Y( 2 P − 1 )⎥
M
Υ = ∑ Y ( m ) = VS + W = V {UΘ( BFp )} + W (17)
Υ=⎢ ⎥ m=1
⎢ e " e ⎥
⎢# # % # ⎥
⎢ 2 π ( P −1)
j P j
2 π ( P −1)( P −1)
⎥ ⎢ # # % # ⎥
⎣1 e " e P
⎦ ⎢ j 2π (Q−1)θ ⎥
(1)
j 2π (Q−1)θ ( 2 ) j 2π (Q−1)θ ( m)
effective CFO’s fall in a non-overlapping ranges. Hermitian operation, Φ = E[ sl slH ] is the covariance of sl ,and I
is a Q×Q identity matrix. From Eq.(18), it is known that Ψ is
C. Multiple user signal, and its effective CFO a Hermitian matrix, and its eigenvalues are all positive, real
numbers. Let λ1 ≥ λ2 ≥ " ≥ λQ denote the Q eigenvalues of Ψ
From (8), at the uplink receiver, the remaining N superposed
signal samples of one OFDMA block after the removal of CP in descending order. Based on the subspace processing
are given by, technique [9], we find that {θ ( m ) }mM=1 corresponds to the M
of PCFO (θ ) ,
M P −1
Υ (n) = ∑ ∑ H p( m ) S p( m ) e j ( 2π / N )( pQ + q
(m)
+ε ( m ) ) n
(16) largest values which is given by
m =1 p =0
1
where the m-th user is assigned to sub-channel q. In (16), we PCFO (θ ) = 2
(19)
H H
assume signals from different users are synchronized in time a (θ )U zU z a (θ )
where a (θ ) = [1 e j 2πθ " e j 2π ( Q−1)θ ] , U z is Q × (Q − M )
matrix composed of the Q − M eigenvectors corresponding to
λM +1 , ", λQ and the subspace spanned by U z is the noise
subspace of Ψ . As PCFO (θ ) is a function of the effective CFO,
we define PCFO (θ ) as the CFO spectrum according to the
definition of Direction-Of-Arrival (DOA) spectrum in the
MUSIC algorithm. With a large number of columns in Y, the
covariance matrix Ψ can be estimated by
ˆ = 1 ΥΥ H = 1 ∑ y y H
P
Ψ l l
(20)
P P l =1
Therefore, Û z ,the estimation of U z ,can be derived by
introducing eigen decomposition to Ψ̂ .
From Eq.(19), estimation of {θ ( m ) }mM=1 corresponds to the m-th Fig .2 Flowchart of the proposed two-stage CFO-estimation algorithm based
on subspace processing
largest peaks of PCFO (θ ) .Cao,etal. proposed a CFO-estimation
algorithm based on the signal structure[6].The main advantage
of the estimation algorithm in [6], is that it can estimate the
CFOs of all users simultaneously using only one OFDMA
block. However, it requires apriori information such as the
number of active users and, the occupied subchannels, which
limits its application. In the case of random subchannel
assignment, the number of users and the occupied subchannels
are unknown to the uplink receiver, thus an accurate peak- Fig.3 The relation between the effective CFOs and the subchannel indexes
search in the CFO spectrum over the range
θ ∈ [(0 − 0.5) / Q, (Q − 1 + 0.5) / Q ] is required in the CTY Where qˆ ( m ) is the estimate of q (m ) , and ⎣X ⎦ denotes the integer
method [6], which is computationally very expensive. To equal to or less than X. Because the number of active users
solve this problem, we propose a two-stage CFO-estimation and occupied subchannels can be estimated, the proposed
algorithm based on subspace processing in this paper. The algorithm is suitable for systems with random subchannel
flowchart of the algorithm is shown in Fig.2 assignment.
In summary, the algorithm is similar to the MUSIC algorithm,
1) Coarse Estimation and the steps of the coarse-estimation stages are described as
follows:
First, we introduce SVD (Eigen decomposition) to Ψ̂ . Then Step 1: arrange the received signal samples into matrix form
the number of users can be estimated by means of the AIC [8], Y;
or the MDL [9], criterion using the Eigen values of Ψ̂ .After ˆ = ΥΥ H / P ;
Step 2: calculate the covariance matrix Ψ
knowing the user number, the coarse estimation of {θ ( m ) }mM=1 ,
Step 3: introduce eigen-decomposition to Ψ̂ ;
{θˆ ( m ) }mM=1 can be derived by performing the coarse peak-search Step 4: estimate the number of users M̂ by using the AIC or
in the CFO spectrum, PCFO (θ ) with the MDL criterion and obtain Û z with the eigenvectors
θ ∈ [(0 − 0.5) / Q, (Q − 1 + 0.5) / Q ] , where the increment of θ is of Ψ̂ ;
θinc = 0.1/Q .From Eq.(15), we can obtain the relation formula Step 5: find the M̂ largest peaks of PCFO (θ ) over the range
between the effective CFO and the subchannel index of the m- θ ∈ [(0 − 0.5) / Q, (Q − 1 + 0.5) / Q ] in order to obtain the coarse
th user, which is given by,
θ ( m)Q = q ( m) + ε ( m) , m=1,2,…,M (21) estimations {θˆ ( m ) }mM=1 , where the increment of θ in each round is
Fig.3 illustrates the relation between {θ ( m ) }mM=1 where the arrows θinc = 0.1 / Q ;
Step 6: estimate users’ subchannels according to Eq.(22).
represent the occupied subchannels. Since ε ( m ) ∈ (−0.5,0.5) ,the
subchannel index of the m-th user can be estimated via, 2) Fine Estimation
qˆ ( m ) = [θˆL( m ) Q + 0.5] (22) In this stage, the fine peak-search is performed in the CFO
spectrum PCFO (θ ) with
TABLE 2
θ ∈ [(qˆ ( m ) − 0.5) / Q, (qˆ ( m ) + 0.5) / Q ] , m = 1,2,..., Mˆ , and, hence, SIMULATION PARAMETERS
the accurate estimations of all the CFOs can be obtained. The Uplink Band Width (BW) 20MHz
increment of θ in each round is θ inc = 0.001 / Q . Modulation QPSK
The steps for the fine-estimation stage are summarized as Number of subcarriers (N) 2048
follow:
Subcarrier spacing (Δƒ) 11.16kHz
Step 1: select user m from M̂ users;
Step2: find the largest peak of PCFO (θ ) in Useful symbol duration (T) 89.6 μs
θ ∈ [(qˆ ( m ) − 0.5) / Q, (qˆ ( m ) + 0.5) / Q ] , in order to obtain Cyclic prefix length 256
2
10
1
P e a k v a lu e
ENRMSE
10
0
10
-3
10
-1
10
-2
10
-4
-3 10
10 0 5 10 15 20 25 30
-0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 SNR (dB)
Effective CFO
Fig.6 Normalized RMSE versus SNR
Fig.5 Enlarged CFO spectrum of all users