a novel pilot pattern for minimizing …ijesr.org/admin/upload_journal/journal_m.archana...
TRANSCRIPT
ISSN 2277-2685 IJESR/July 2015/ Vol-5/Issue-7/887-895
M. Archana et. al./ International Journal of Engineering & Science Research
*Corresponding Author www.ijesr.org 887
A NOVEL PILOT PATTERN FOR MINIMIZING SYNCHRONIZATION ERRORS
IN CARRIER AGGREGATION SYSTEMS
M. Archana*1, Ch.Venkat Ramaiah
2
1Dept of Electronics and communications, SRM University, Chennai, India.
2Asst. Prof, Dept. of ECE, R.M.D. Engineering College, Kavaraipettai, Tamil Nadu, India.
ABSTRACT
Orthogonal Frequency Division Multiplexing (OFDM) is a multi-carrier modulation scheme, which is used as
the air interface technique for LTE (Long Term Evolution) standard. Furthermore Carrier Aggregation (CA) is
defined by 3GPP (3rd Generation Partnership Project) to support wide-bandwidth transmission. However OFDM
is sensitive to synchronization errors due to multipath fading, Doppler shift and oscillator instability. In this
paper, a block type pilot pattern error suppression algorithm is proposed to reduce the synchronization errors in
carrier aggregation systems. A block type pilot pattern used here is weighted CAZAC (Constant Amplitude Zero
Auto-Correlation) sequence which is well exploited in LTE Advanced standard. Simulation results show that, it
can significantly improve the system performance and has a smooth performance when synchronization errors
vary.
Keywords: Carrier Aggregation (CA), Carrier Frequency Offset (CFO), Orthogonal frequency Division
Multiplexing (OFDM), Sampling Time Offset (STO),weighted Constant Amplitude Zero Auto-Correlation
(CAZAC).
1. INTRODUCTION
A multicarrier communication system with orthogonal subcarriers is called Orthogonal Frequency Division
Multiplex (OFDM) system. The word “orthogonal” indicates that there is a precise mathematical relationship
between the frequencies of the carrier in the system. The basic principle of OFDM is to split a high data rate
sequence into a number of low rate sequences that are transmitted simultaneously over a number of subcarrier.
In order to support more number of users than OFDM, carrier aggregation technique is used. Carrier
Aggregation (CA) is a technique that aggregates multiple component bands into an overall wider bandwidth, is
proposed to support wide bandwidth transmission in LTE- Advanced standard.CA also provides flexibility in
spectrum assignment. It is of three types, Inter-band non-continuous, intra-band continuous and intra-band non
continuous.
However, OFDM system is sensitive to synchronization errors, such as Carrier Frequency Offset (CFO) and
Sampling Time Offset (STO). Inaccurate synchronization will break the orthogonality among subcarriers and
introduce ICI (Inter Carrier Interference) into received signal, which will lead to serious performance
degradation. Furthermore, in non-continuous CA scenario, signals transmitted on different component bands
always experience different synchronization errors, which make synchronization much more difficult. There
have been many previous works on synchronization problem. In [4], a timing synchronization algorithm was
discussed but CFO compensation method wasn’t considered. CFO compensation, methods are proposed in [5-
14]. These frequency synchronization methods can be classified as two groups. The first is called feedback
method, which will increase the transmission overhead and possibly cause outdated estimation in time-varying
scenario. An alternative is to achieve synchronization via signal processing at the receiver without the help of a
control channel. In [8-10], SIC (Successive Interference Cancellation) as well as PIC (Parallel Interference
Cancellation) methods were raised. In these methods, the received signals are classified as reliable group and
unreliable group. The reliable signals are directly detected while the unreliable signals are detected after the
cancellation of the MAI (Multi Access Interference) effects due to the reliable signals. Inverse interference
matrix method was discussed in [11-13]. Unfortunately, the methods in [8-11] need perfect multiple CFO’s
M. Archana et. al./ International Journal of Engineering & Science Research
Copyright © 2015 Published by IJESR. All rights reserved 888
estimation, which is impossible in practical system. And our previous algorithms proposed in [12-13] didn’t
consider timing offset. Furthermore, all the methods mentioned above are applied non continuous –CA system,
and to our best knowledge, there are few papers about the synchronization problems in CA system. In this
project work, mainly focus on the synchronization problems in non-continuous CA system and proposed a block
type pilot based synchronization errors suppression algorithm. First use the correlation of the pilot block to
estimate the STO. Interference estimation is followed after STO estimation. In this stage, pilot block are
exploited to estimate the ICI components directly. Then suppress the components by inverse matrix method.
Since block type pilot is a common pilot pattern in wireless communications, this method can be easily extended
to the other systems.
2. SYSTEM MODEL
Fig.1 shows the simplified block diagram of the orthogonal frequency division multiplexing (OFDM) based
non-continuous carrier aggregation (non-continuous CA) system. Consider the case that is without loss of
generality, we suppose K component bands are exploited in this communication system. The carrier frequency
of the kth component band is denoted as fck and the subcarrier amount of the kth component band is Nk.
Here we assume the vector dk=[dk
0,dk
1,..............dkNk-1]
T (k=1, 2, 3 …, K) denotes the transmitted data on the k
th
component band. Then the output of the Nk-IFFT is
∑−
=
=1
0
/2/1)(
k
k
N
i
Ninji
kk
k edNnsπ (1)
(a) Transmitter
(b) Receiver
Fig. 1: Block diagram of OFDM based non-continuous CA system (a) Transmitter (b) Receiver
The base band signal will be modulated to the transmission band with the carrier frequency fck which is
generated by the transmitter frequency synchronizer at the up-converter. Here we consider each signal
M. Archana et. al./ International Journal of Engineering & Science Research
Copyright © 2015 Published by IJESR. All rights reserved 889
experiences a frequency selective fading channel with the time domain impulse response hk(n) (k=1, 2, 3 …K).
Then the received signal is the superposition of signals from all active component bands and can be written as
∑=
+⊗=K
k
kktfjk tvthetstrkc
1
2)()()).(()(
π (2)
Where ⊗ denotes convolution and Vk (t) is the AWGN on the k
th component band. To extract the transmitted
signal rk(t) on the k
th component band, the generated signal will pass through a low-pass filter and cancel the
signals from the other component bands. Through Nk-FFT processing, the output of the lth subcarrier on the kth
component band is
Ylk=dl
kHl
k (3)
Where Hlk
denotes the frequency domain channel response of the kth
component band.
3. INFLUENCE OF SYNCHRONIZATION ERRORS
To start the influence of inaccurate synchronization, multiple CFOs and STOs will be introduced into the
received signals. CFO and STO of the kth
component band are denoted as ∆fek
and ∆tek
respectively. After
transmitted through fading channels, the CFOs and channel fading corrupted time domain signal can be written
as
ke
ke
kc
tnt
kK
k
ktffjk tvthetstr∆+
=
∆+ +⊗=∑ )()()).(()(1
)(2π (4)
For non-continuous CA, different component bands are usually separated by sufficient bandwidth, so the
interference between them can be negligible. The STO corrupted received signal can be written as
kes
ke
tnTt
kktfjkk tvthetsnr∆+=
∆ +⊗= |)()(]).([)(2π
)(/1/)(2/2
1
0
k
kNenijNnijk
i
N
i
k
ik nveeHdN kkkkkk
k
δδεδππ ++= ++−
−∑
∑−
=
++ ++=1
0
/)(2)(/1
k
kkkk
N
i
k
kNninjk
i
k
i
k
ik nvepHdN δδεεπ
(5)
Where Ts=T/Nk is the sampling period, T denotes the symbol period, εk and δk denote the normalized CFO and
STO respectively as, after FFT,
∑∑−
=
−
=
−++ +=1
0
1
0
/2/)(2/1
k k
kkkkk
N
n
N
i
k
l
NnljNnnijk
i
k
i
k
ik
k
l VeepHdNYπδεεπ
∑ ∑−
=
−
=
+− +=1
0
1
0
/2])[(2/1
k k
kkkk
N
n
N
i
k
l
Njlinjk
i
k
i
k
ik VeepHdNδπεεπ
∑−
≠=
− ++=1
0
0
kN
lii
k
l
k
li
k
i
k
i
k
i
kk
l
k
l
k
l VQpHdQpHd
(6)
Where ��� = ���∑ �� ���������������� �� ������ denotes the interference factor and Vkl denotes AWGN. Obviously, we
can obtain the matrix expressed frequency domain signal vector according to the above equation
k
k
v
K
M
K
M
k
v RPQ VY += (7)
Where ��� = [��� , � �……… . �#�� � ]T (Rl
k=dl
kHl
k) denotes the received signal vector distorted by channel
fading, PMk=diag(p0
k,p1
k,......pNk-1
k) denotes the phase rotation matrix caused by STO and
M. Archana et. al./ International Journal of Engineering & Science Research
Copyright © 2015 Published by IJESR. All rights reserved 890
=
−−−−
−
kk
N
k
N
k
Nkk
k
M
QQQ
QQQ
Q
kk
k
0)2()1(
110
........
.
.
.
.
........
.
.
(8)
denotes the ICI matrix caused by CFO. The diagonal components in k QkM are the CPE (Common Phase Error)
components and the other components denote the ICI from the corresponding subcarriers.
4. SYNCHRONIZATION ERRORS SUPPRESSION ALGORITHM
In this section, we will discuss the proposed block type pilot based synchronization errors suppression
algorithm. The pilot pattern used here is the weighted CAZAC sequence, which is one of the strongest
candidates as pilot pattern in OFDM system and has been exploited by LTE-Advanced standard. Let L to be any
positive integer larger than one and M to be any number, which is relatively prime with L. Then an example of
weighted CAZAC sequence is given as
−==
−==
+
++
LisevenLnec
LisoddLnenc
nnLMj
nM
nnnLMj
M
,1................1,0,
,1.,.........1,0,)(
)2/(/2
)(
)2/)1((/2
2π
π (9)
and the circular auto-correlation of weighted CAZAC sequence is
∑−
=
≠
==+
1
0
mod0,0
0,)()(
L
n
LMM
Lncnc
τ
ττ (10)
(a) STO Estimation Algorithm
In this part, we discuss the proposed frequency domain STO estimation algorithm to obtain the STO between
the practical sampling time and optimum sampling time. To implement this algorithm, special weighted
CAZAC sequence should be exploited, in which pilots spaced by half subcarriers amount should satisfy the
following requirement
k
Nl
k
l kPP 2/+±= (11)
Where Plk denotes the pilot transmitted on the lth subcarrier of the kth component band. In OFDM system, since
the amount of subcarrier is always an even number, we can obtain the expression of %&�as
1..............1,0,
)2/(/2 2
−== +k
llNMjk
l NleP kπ
(12)
Together with equation (11), we can obtain
int2/1)2/2/2/1(
])2/()2/[(/)2/(/ 22
])2/()2/[(/2)2/(/2 22
=++
+++=+
±= ++++
lNM
NlNlNMllNM
ee
k
kkkK
NlNlNMjllNMj kkkk ππ
(13)
Obviously, to satisfy the above requirement, we just need to set M as an odd and Nk as a multiple of 4 (in this
scenario, M is naturally relatively prime with Nk). In OFDM system, to achieve high speed FFT/IFFT,
subcarrier amount is usually set as exponential times of 2. Thus the above requirement can be achieved by just
set M as an odd number. Therefore
−=
=
+
+
iflisevenPP
iflisoddPP
k
Nl
k
l
k
Nl
k
l
k
k
;
,
2/
2/ (14)
As illustrated in previous section, ICI has the appearance of Gaussian noise. Furthermore, according to the
interference self-cancellation algorithm, the pilot pattern used here could alleviate the interference. Therefore,
we could treat the interferences as noise and don’t consider it. In this situation, the lth
received pilots can be
expressed as
M. Archana et. al./ International Journal of Engineering & Science Research
Copyright © 2015 Published by IJESR. All rights reserved 891
kk
l
k
l
k
l
k QpHPlP 0)( = (15)
We suppose the frequency domain channel response has already been estimated. Then the correlation can be
calculated as follows
' = ∑ (��)*�+,�-&.��� /0*����� %�(1��� ��&�� )
=∑ |%&�|3|���|3[4&.��/3���� ��&�� . (4&�)∗] =89:� ∑ |%�(1)|3|���|3��� ��&�� (16)
Therefore, the STO of the kth component band can be obtained from the phase of η as
;� = <=>?<@-AB(C)DE(C)/9 (17)
Since the variation range of arc tan(x) is (-π/2, π/2), the estimation range is (-0.5, 0.5), which fully covered the
available value of STO.
(b) Interference Suppression Algorithm
After STO estimation, we perform interference estimation and then suppress the inaccurate synchronization.
Instead of estimating CFOs, which is fairly difficult to get accurate results, we directly estimate the interference
components under the aid of the received pilot block. After that, the ICI matrix can be reconstructed by very
simple mapping method. Then we can suppress interferences by inverse matrix method and finally improve
system performance.
From the expression of QLk
we can obtain that
���� = 1G� H �� ��I������� �� �������������
= ���∑ 839(����.J�)/��������� �� ������
= QkNk-1 (18)
Therefore, the ICI matrix Q Mk can be rewritten as
�K� = L������ ⋯ ������⋮ ⋱ ⋮����3� ⋯ ��� P (19) We can see that the matrix QM
k is a circulant matrix with only Nk different components. Therefore, equation (7)
can be written as
RS� = L������ ⋯ ������⋮ ⋱ ⋮����3� ⋯ ��� PTK� L ���⋮������ P + V�
=L ���4�����4�� ⋯ ������ 4�����⋮ ⋱ ⋮������ 4����� ���4�� ⋯ ����3� 4���3� P L ���⋮������ P + V�
=�K� WS� + V� (20)
Where WS� = [������ …������ ]Xand RMk=L ���4�����4�� ⋯ ������ 4�����⋮ ⋱ ⋮������ 4����� ���4�� ⋯ ����3� 4���3� P (21)
M. Archana
Copyright © 2015 Published by IJESR. All rights reserved
We can treat equation (20) as an equations set and the
estimate these Nk values by solving this equations set, the
of CFO can be suppressed by multiply the inverse
been estimated and the pilots are known to the receive,
rank matrix. Thus, (RMk)
-1 can be easily constructed
easily estimated by using the following equation is as shown below:
WY�Z = WS� + (�[� )��V�After that, the interference matrix QM
k’
CFO are highly correlated during the pilot block and the foll
errors can be easily suppressed by LS (Least Square) method as
R\]^^_`\\`a� = �V� + (T[� )0Since the interferences from other component bands are ignorable, the synchronization errors of each band
could be independent and have few influences on the suppression performan
process to suppress the inaccurate synchronization on each band.
Fig. 2: Block diagram of synchronization errors suppression algorithm
Fig. 2 shows the block diagram of the suppression algorithm. The whole processing
follows:
(1)When a symbol arrives at the receiver, detect whether this symbol is a pilot symbol:
Yes, go to step (2); No, go to step (5);
(2) Do channel estimation;
(3) Do STO estimation under the aid of channel estimation;
(4) Do Q matrix estimation under the aids of channel estimation results and STO estimation results.
(5) Do inverse-matrix based suppression algorithm.
5. SIMULATION RESULTS
Fig. 3 shows the MSE (Mean Square Error) performance of the proposed STO estimation algorithm
are evaluated in AWGN and ITU Ped
reduced to 10-4
level which is a fairly small value. So this achieves smooth performance when synchronization
errors vary.
M. Archana et. al./ International Journal of Engineering & Science Resear
Copyright © 2015 Published by IJESR. All rights reserved
We can treat equation (20) as an equations set and the Nk components in QMk
as the unknown values. If we can
values by solving this equations set, the QMk
matrix can be reconstructed and the interferences
of CFO can be suppressed by multiply the inverse QMk
matrix with the received signal. Since STO has already
been estimated and the pilots are known to the receive, RMkcan be easily calculated. Obviously,
can be easily constructed Qvk
and QMk
which contains the Nk components in, can be
easily estimated by using the following equation is as shown below:
(22)
k’ can be reconstructed from the estimatedWY�Z .We assume the STO and
CFO are highly correlated during the pilot block and the following data blocks. Therefore, the synchronization
errors can be easily suppressed by LS (Least Square) method as
0[(WK�Z )0WK� ]Z ��(WY�Z )0V�Since the interferences from other component bands are ignorable, the synchronization errors of each band
could be independent and have few influences on the suppression performance. Therefore, we can use parallel
process to suppress the inaccurate synchronization on each band.
Block diagram of synchronization errors suppression algorithm
2 shows the block diagram of the suppression algorithm. The whole processing can be described as
(1)When a symbol arrives at the receiver, detect whether this symbol is a pilot symbol:
(3) Do STO estimation under the aid of channel estimation;
rix estimation under the aids of channel estimation results and STO estimation results.
matrix based suppression algorithm.
Fig. 3 shows the MSE (Mean Square Error) performance of the proposed STO estimation algorithm
are evaluated in AWGN and ITU Ped-B scenarios. This shows that MSE of the proposed algorithm can be
level which is a fairly small value. So this achieves smooth performance when synchronization
International Journal of Engineering & Science Research
Copyright © 2015 Published by IJESR. All rights reserved 892
as the unknown values. If we can
matrix can be reconstructed and the interferences
matrix with the received signal. Since STO has already
can be easily calculated. Obviously, RMk is a full
components in, can be
.We assume the STO and
owing data blocks. Therefore, the synchronization
(23)
Since the interferences from other component bands are ignorable, the synchronization errors of each band
ce. Therefore, we can use parallel
can be described as
rix estimation under the aids of channel estimation results and STO estimation results.
Fig. 3 shows the MSE (Mean Square Error) performance of the proposed STO estimation algorithm. The results
B scenarios. This shows that MSE of the proposed algorithm can be
level which is a fairly small value. So this achieves smooth performance when synchronization
M. Archana et. al./ International Journal of Engineering & Science Research
Copyright © 2015 Published by IJESR. All rights reserved 893
Table 1: Simulation Parameters
Baud Rate 9600 symbols/second
Modulation Type 16 QAM
Component Carriers 4
FFT size 256
Cyclic prefix length 16
Number of Sub Carriers 512
Channels AWGN, Pedestrian-B(Ped-B)
Fig. 3: MSE performance of the proposed STO estimation algorithm
Now selecting the algorithm proposed in [11] as comparison and choosed 5, 10, 50 as the bandwidth of the
banded-matrix. Fig. 4 shows the BER performance in AWGN channel and Fig. 5 shows the BER performance
in Ped-B channel. Comparing the simulation results, we can see that due to the co-effects of CFO and STO, the
BER performance will be degrade seriously. If we used banded matrix suppression method, when the
bandwidth is small a serious error flow phenomenon occurs in high SNR region. Thus the Bit Error Rate (BER)
is reducing for the proposed system; hence it improves the performance of the system.
Fig. 4: BER performance in AWGN channel
0 2 4 6 8 10 12 14 16 18 2010
-5
10-4
10-3
10-2
10-1
100
SNR in dB
MS
E
MSE performance of proposed STO estimation algorithm
MSE of Proposed STO Estimation in Ped B Scenario
MSE of Proposed STO Estimation in AWGN Scenario
0 2 4 6 8 10 12 14 16 18 2010
-5
10-4
10-3
10-2
10-1
100
SNR in dB
BIT
ER
RO
R R
AT
E
BER performance in AWGN channel
No Sync Error
No Suppression
Proposed System
Banded Matrix Suppression Bandwidth=50
Banded Matrix Suppression Bandwidth=10
Banded Matrix Suppression Bandwidth=5
M. Archana et. al./ International Journal of Engineering & Science Research
Copyright © 2015 Published by IJESR. All rights reserved 894
Now comparing the simulation results in Ped-B scenario in fig. 5,at SNR= 6dB in Pedestrian-B channel, Bit
Error Rate=15x10-3
for Banded Matrix suppression method when sub band bandwidth is 50KHz whereas
proposed system gives Bit Error Rate of 5x10-3. Thus the Bit Error Rate (BER) is reducing for the proposed
system; hence it improves the performance of the system
Fig. 5: BER Performance in Ped-B Channel
6. CONCLUSION
In this project work, focus on the synchronization problems in non-continuous CA OFDM systems and propose
a novel block type based synchronization errors suppression algorithm. Unlike other STO estimation
algorithms, the proposed method exploits a special kind of weighted CAZAC sequence which abides the rule of
interference self cancellation and can improve the estimation accuracy. Furthermore, the proposed algorithm
could directly estimate the interference components without CFO estimation. Thus, system complexity can be
reduced. Most of all, the synchronization errors of each band could be independent, which is more close to
practical situation. According to the simulation results the proposed algorithm could significantly improve
system performance and achieve smooth performance when synchronization errors vary. Since the weighted
CAZAC sequence pilot pattern is defined in LTE- Advanced standard this algorithm can be treat as a candidate
method in LTE-Advanced systems and other block type pilot based systems.
REFERENCES
[1] ITU-R Rec M. 1645. Framework and Overall Objectives of the Further Development of IMT-2000 and
Systems Beyond IMT-2000, 2003.
[2] ITU-R Rec.M.2134. Requirements related to Technical performance for IMT-Advanced Radio Interface(s),
2008.
[3] Yuan G, Zhang X, Wang W, Yang Y. Carrier aggregationfor LTE- advanced mobile communication
systems. IEEE Communications Magazine 2010; 48(2): 88– 93.
[4] Chen N, Tanaka M, Heaton R. OFDM timing synchronization under multi-path channels. IEEE 57th
Vehicular Technology Conference, April 2003.
[5] Morelli M. Timing and frequency synchronization for the uplink of an OFDMA system. IEEE Transactions
on Communications 2004; 52(2): 296-306.
[6] Zhang X, Ryu H, Kim J. Suppression of synchronization errors in OFDM basedcarrier aggregation system.
Proc. IEEE APCC., Nov. 2010; 106-111.
[7] Adakane DV, Vasudevan K. An Efficient Pilot Pattern Design for ChannelEstimation in OFDM Systems.
IEEE International Conference on Signal Proceesing, Computing and control, 2013.
0 5 10 15 20 25 30 3510
-5
10-4
10-3
10-2
10-1
100
SNR in dB
BIT
ER
RO
R R
ATE
BER performance in Ped-B channel
No Sync Error
No Suppression
Proposed System
Banded Matrix Suppression Bandwidth=50
Banded Matrix Suppression Bandwidth=10
Banded Matrix Suppression Bandwidth=5
M. Archana et. al./ International Journal of Engineering & Science Research
Copyright © 2015 Published by IJESR. All rights reserved 895
[8] Jiang Y, Zhu X, Lim E, Huang Y. Joint Semi-Blind Channel Equalization and ICI Mitigation for Carrier
Aggregation Based CoMP OFDMA Systems with Multiple CFOs. IEEE ICC 2013 Signal Processing for
Communications Symposium.
[9] Wang H, Zhu L, Shi Y, Xing T, Wang Y. A Novel Synchronization Algorithm for OFDM Systems with
Weighted CAZAC Sequence. Journal of Computational Information Systems, 2012; 2275–2283.
[10] Yucek T, Arslan H. Carrier Frequency Offset Compensation with Successive Cancellation in Uplink
OFDMA Systems. IEEE Transactions on Wireless Communications 2007; 6(10): 3546-3551.
[11] Cao Z, Tureli U, Yao Y-D, Honan P. Frequency synchronization for generalized OFDMA uplink. IEEE
Global Telecommunications Conference, Nov. 2004.
[12] Zhang X, Ryu H-G. Joint Estimation and Suppression of PhaseNoise and CFO in MIMO SC-FDMA with
SC- SFBC, IET Communications.
[13] Zhang X, Ryu H-G. Suppression of ICI and MAI in SC-FDMACommunication System with Carrier
Frequency Offsets. IEEE Transaction onConsumer Electronics 2010; 56(2): 359-365.
[14] Wu Y, Bergmans JWM, Attallah S. Carrier frequency offset estimation formultiuser MIMO OFDM uplink
using CAZAC sequences performance and sequence optimization. EURASIP Journ.Wireless Com- mun. and
Net., 570680-1/11, 2011.