Transcript performance of frequency offset synchronization in a

PERFORMANCE OF FREQUENCY OFFSET
SYNCHRONIZATION IN A SINGLE
AND MULTI-ANTENNA IEEE 802.16-2004
SYSTEM
José A. Rivas Cantero
M. Julia Fernández-Getino García
Dpto. de Teoría de la Señal y Comunicaciones,
Universidad Carlos III de Madrid
3rd COST 289 Workshop
ENABLING TECHNOLOGIES FOR B3G SYSTEMS
July 12-13, 2006
Aveiro, Portugal
1
Outline
 1. Background
•Motivation
•OFDM
•MIMO-OFDM
•IEEE 802.16-2004
•STC
 2. Frequency offset estimation algorithms
•SISO systems
•MIMO systems
 3. Results
 4. Summary and Conclusions
2
1. Background-Motivation
Nowadays the combination of Multiple-Input Multiple-Output
(MIMO) systems and Orthogonal Frequency Division Multiplexing
(OFDM) technologies (MIMO-OFDM) is one of the most attractive
techniques to provide broadband communications
IEEE 802.16-2004, also known as IEEE 802.16d, is the standard that
describes the air interface for fixed broadband wireless
communications. Physical layer based on OFDM modulation.
This standard just proposes a typical SISO system, and leaves as
optional the development of a MISO 2x1 system.
This work extends the standard to a MIMO scheme. Several
scenarios (SISO, MISO, MIMO) are developed and compared.
3
1. Background-Motivation
A critical issue is frequency offset estimation and correction:
SISO systems
MIMO systems
In all these schemes several algorithms are compared:
Channel estimation algorithms:
Maximum Likelihood Time Frequency (ML-TF)
LS estimator (Time domain)
Space Time Coding (Alamouti configuration) is usually employed.
Influence in:
Bit Error Rate (BER)
Data transfer rate
4
1. Background - OFDM
OFDM: Multicarrier modulation which divides the bandwidth in
several ortogonal channels.
Suitable for data transmission in wireless channels due to its
robustness against multipath fading.
Easy implementation by FFT.
5
1. Background - OFDM
Time-Frequency scheme
Cyclic prefix: avoids ISI and ICI
OFDM block diagram
6
1. Background MIMO- OFDM
MIMO: Use of multiple antennas both in the transmitter and in the
receiver
Several channels among emitter and receiver.
High capacity system.
Diversity in a fading environment.
MIMO-OFDM system :
7
1. Background- IEEE 802.16-2004
Air interface for fixed broadband wireless communications standard
Revision of IEEE Std 802.16-2001.
IEEE802.16e. Approved December 2005. WMAN mobile.
NLOS propagation.
2-11 GHz
OFDM. FFT 256 points
Data subcarriers (QPSK, 16-QAM, 64-QAM-optional)
Pilot subcarriers: Estimation purposes (BPSK)
Null subcarriers: DC and guard band.
8
1. Background- IEEE 802.16-2004
Standard specifies preambles both for UL and DL:
UL: One OFDM symbol. Only even subcarriers are not null:
DL: Two OFDM symbols. In the second one only even
subcarriers are not null :
One symbol with only even subcarriers different from zero =>
Two equal halves in time domain.
9
1. Background- IEEE 802.16-2004
PMP => Point – Multipoint structure
In simulations TDD is employed:
DL
preamble
UL
preamble
10
1. Background- IEEE 802.16-2004
Wimax scenarios
11
1. Background - STC
Alamouti scheme
12
1. Background
Implementation of MISO system is optional.
2x1 system employing Space-Time Coding.
When using more than one transmitter preamble emitted in the
DL is not the long preamble (2 OFDM symbols). It is a OFDM
symbol where only odd subcarriers are not null.
Preambles emitted by both antennas are orthogonal.
Schemes studied in this work:
SISO
* The first one is the
MISO 2X1. STC
standard one.
* The second one is
MIMO 2X2. STC
optional.
* The rest ones are new,
MIMO 2X2. NO STC
and are not implemented
in the standard yet.
13
2. Frequency offset
In the simulations two channel estimation algorithms compatible with
IEEE 802.16-2004 standard are employed:
1) ML Algorithm:
Estimation in frequency domain (Subcarrier by subcarrier).
Interpolation is needed.
Frequency estimator:
2) LS Algorithm:
Estimation in Time domain.
Expression:
14
2. Frequency offset
CHANNEL ESTIMATION ALGORITHMS
Very similar performance in terms of BER.
NO STC
15
2. Frequency offset
Frequency synchronization must be performed in the receiver.
 No synchronization => orthogonality loss among symbols.
Why this offset appears?
 Channel effects.
 Synchronization loss among system elements, especially between
emitter and receiver oscillators.
ε represents the normalized frequency offset:
16
2. Frequency offset
Preambles composed of two equals halves in time domain =>
algorithms based on finding them (Correlation).
Offset is composed of an integer and a fractional part.
Received signal in
time domain
Correction:
¡¡ Frequency offset  Change in the phase of the received
signal (in time domain) !!
Target: Residual offset as small as possible.
17
2. Frequency offset
Using this fractional part estimation and LS channel estimation a
joint channel estimation and frequency estimation can be derived. It
takes into account the estimation of the integer part of the frequency
offset:
1) Estimation and correction of the fractional part of the frequency offset
2) Consider the integer frequency offset hypothesis from (-M,-M+2…,
-2,0,2,…,M) where M is the maximum possible even integer offset and obtain
the corresponding LS channel estimates by circularly shifting the FFT outputs
accordingly.
3) Calculate the corresponding LS error for the channel estimates obtained
in the previous step
4) Iterate over steps 2 and 3 till all frequency offset hypotheses are
considered and choose the one that minimizes the LS error.
18
2. Frequency offset
MIMO systems: Based on correlation between signals too!
Adapted from an algorithm proposed for WLAN systems.
First of all we estimate time-domain channel responses
between any pair of transmit and receive antenna assuming
that the frequency offset has been completely compensated.
We define two different signals:
Signals which really arrive to the antennas (yt ).
From yt first channel estimation is performed: (Hl) . We define the
signal which should arrive in case that this estimation were correct:
(yt’ ).
19
2. Frequency offset
Therefore yt (r,1)’ represents the signal which would arrive to the rth
receive antenna in the time instant 1, supposing that the first channel
estimation is correct. With 1 or 2 we distinguish between the two equal
halves which composes the OFDM symbol in time domain (T(p,1);T(p,2)),
t = 0,1,…,127.
To obtain the fractional offset we can measure the phase change
between yt (r,1) yt (r,1)’ * and yt (r,2) yt (r,2)’ * :
Once the fractional offset is found, the correction is performed as in
the SISO systems.
20
3. Results
BER OF DIFFERENT SCHEMES:
2X2 System
No Space Time Coding.
Spatial Multiplexing => Double data
transfer rate.
Highest BER.
21
3. Results
BER OF DIFFERENT SCHEMES:
In all of them ML-TF channel estimator has been employed.
Using a 2X2 scheme data transfer rate is doubled, in case no space time
coding is applied. It could be very useful for situations where a big amount
of data must be transferred, although BER is higher than in the typical
SISO scenario.
Last two curves in Figure show the benefits of employing Space Time
Coding (Alamouti configuration). When using ST Coding data transfer rate
is not doubled, keeping the data rate of the SISO case, although a second
antenna has been added in transmission.
 On the other hand BER of the system decreases significantly. The 2X1
system is leaved as optional in the standard. If a second antenna is added
in reception, it can be clearly appreciated how much BER decreases,
reaching 10-8 values just with a signal to noise ratio of 20 dB
22
3. Results
FREQUENCY OFFSET ESTIMATION EFFECTS:
SISO SYSTEMS
The maximun aceptable residual offset can be ε = 0.01
23
3. Results
FREQUENCY OFFSET ESTIMATION ALGORITHMS:
LS channel estimation (ε = 0.3).
εresidual_SISO ≈ 0.001 y εresidual_MIMO ≈ 0.01.
24
4.Summary and Conclusions
Extension to the IEEE 802.16-2004 standard.
Addition of a second antenna in the receiver.
Several scenarios, combining SISO, MISO and MIMO configurations. Use/
Not use of Space Time Coding.
Frequency offset must be taken into account. With presented algorithms
residual error is almost null. In MIMO systems this offset in perceptible in
terms of the MSE of the channel estimation, but can be considered as offset
free in terms of BER.
Depending on the requeriments of the systems in terms of BER, data
transfer rate, physical space to add more antennas to the system and cost,
one of the schemes studied in this paper may be chosen to implement a next
generation fixed broadband wireless access downlink system based on IEEE
802.16-2004 standard.
25