EX018/2005
M Shakeel Baig
ii
Abstract
802.16e provides specifications for non line of sight, mobile wireless communications in the
frequency range of 2-6 GHz. It is well implemented by using OFDMA as its physical layer
scheme. The OFDM symbol time ( Ts ) is to be selected depending on the channel conditions,
available bandwidth and, simulations provide a means of selecting right values of Ts in
different channel conditions. Additionally it has been shown that certain values of Ts
outperform others in all conditions, thus invalidating their use. Moreover, a solution proposed
by INTEL is also analyzed.
One of the major requirements of OFDM is high synchronization. Detecting the timing offset
of a new mobile user, entering the network, which is not time aligned using cross-correlation
and auto-correlation in time domain and cross-correlation in frequency domain at the base
station has been simulated. Results point that the processing load can be significantly reduced
by using frequency domain correlation of the received data or by using auto-correlation
followed by cross-correlation on localized data.
The use of adaptive antenna system in 802.16e improves the system performance, where
beamforming is implemented in the direction of desired user. Capons method and MUSIC
method have been simulated to compute the direction of arrival for OFDMA uplink. A new
user, while in the ranging process, transmits data with unknown time offset and unknown
direction. The thesis describes the procedure to find the two unknown one after another.
iv
Acknowledgements
This thesis was possible because of the will and wish of ALMIGHTY and I am grateful to
Him. Next, I express thanks to my Parents for their support during all these years of my
education.
I am grateful to Analog Devices Inc, Sweden for allowing me to work at their office for the
thesis, and my thanks to its employees for providing an encouraging environment, during my
stay at the office.
I express my sincere thanks to my supervisors, Yusuf Jamal, Klas Brink and Rickard
Fahlqvist, for their guidance and invaluable time spent on those numerous discussions we had
at the office. It is in those fruitful discussions that I learned many more things apart from
thesis itself and I am sure those will be helpful in future. Thanks for everything.
Whenever we did not find any solution, we had one man to call to, Michel Lopez, USA. My
special thanks to him for helping me during the thesis and for the careful reviewing of my
thesis report.
My thanks to Professor Mats Viberg, my thesis examiner, for his valuable ideas regarding
beamforming, the channel model and comments on the thesis report.
Lastly, I thank all my friends who helped me during my stay in Stockholm.
vi
Contents
ABSTRACT ............................................................................................................................IV
ACKNOWLEDGEMENTS...................................................................................................VI
CONTENTS.........................................................................................................................VIII
LIST OF FIGURES ................................................................................................................ X
PART 0: INTRODUCTION.................................................................................................... 1
0.1 GOAL OF THE THESIS ......................................................................................................... 3
0.2 REPORT OUTLINE ............................................................................................................... 3
0.3 INTRODUCTION TO WIMAX .............................................................................................. 3
0.3.1 802.16c .................................................................................................................................. 3
0.3.2 802.16a .................................................................................................................................. 4
0.3.3 802.16d .................................................................................................................................. 4
0.3.4 802.16e .................................................................................................................................. 4
viii
ix
List of figures
Figure 0.1: Three orthogonal subcarriers ................................................................................. 1
Figure 0.2: Comparison between FDMA and OFDM [5] ......................................................... 6
Figure 0.3: 3 subcarriers and multipath component ................................................................. 7
Figure 0.4: Perfect synchronization, no ICI .............................................................................. 7
Figure 1.1: Time plan from [1] ................................................................................................ 10
Figure 1.2: Tile Structure Uplink............................................................................................. 10
Figure 1.3: Simulated Doppler spectrum................................................................................. 13
Figure 1.4: A typical Rayleigh fading channel ........................................................................ 14
Figure: 1.5. The Block Diagram .............................................................................................. 16
Figure 1.6: SNR v/s BER for AWGN channel with QPSK modulation .................................... 19
Figure 1.7: Response of tail biting convolutional code in AWGN channel ............................. 19
Figure 1.8: Performance of 16 QAM modulation in AWGN channel...................................... 20
Figure 1.9: Performance in fading channel at 3GHz (without multipath) .............................. 21
Figure 1.10: Typical channel at different Doppler frequency (350, 680, 70, 135 Hz) ............ 21
Figure 1.11: Performance in fading channel at 5.9GHz (without multipath) ......................... 22
Figure 1.12: Performance of OFDM at 2 carrier frequencies ................................................ 23
Figure 1.13: Performance of 128 S for different speed of mobile ......................................... 24
Figure 1.14: Performance of 256 S for different speed of mobile ......................................... 24
Figure 1.15: Channel model .................................................................................................... 25
Figure 1.16: Figure showing subcarriers and multipath......................................................... 26
Figure 1.17: Performance at 3GHz in fading channel with multi path (guard time Ts/8)....... 26
Figure 1.18: Performance at 5.9GHz in fading channel with multi path (guard time Ts/8).... 26
Figure 1.19: Performance for different amplitudes of multipath............................................. 27
Figure 1.20: 16QAM in fading channel ................................................................................... 28
Figure 1.21: Performance when multipath delay is limited with in 16 S (guard time Ts/4) .. 29
Figure 1.22: Performance when multipath delay is limited with in 100 S (guard time Ts/4) 29
Figure 1.23: Convolutional encoder [1].................................................................................. 30
Figure 1.24: Convolutional coding performance .................................................................... 31
Figure 2.1: Initial ranging transmission symbol structure ...................................................... 36
Figure 2.2: Process of finding peak corresponding to time offset .......................................... 37
Figure 2.3 frequency domain correlations............................................................................... 38
Figure 2.4: Frequency domain correlation using IFFT .......................................................... 39
Figure 2.5: Comparison of various correlation techniques..................................................... 40
Figure 2.6: Cross-correlation with double precision code and with 2 bit quantized code...... 41
Figure 2.7: Effect of ranging signal amplitude on system performance.................................. 42
Figure 2.8: cross-correlation................................................................................................... 43
Figure 3.1: 2 element linear array........................................................................................... 47
Figure 3.2: 2 element linear array far field geometry ............................................................. 47
Figure 3.3: Plot of Array factor (AF) ...................................................................................... 48
Figure 3.4: Example (non-real) of switched beamforming...................................................... 49
Figure 3.5: Switched beamforming and Adaptive beamforming ............................................. 49
Figure 3.6: Downlink part of AAS frame structure.................................................................. 51
xi
PART 0: Introduction
Wireless communication systems have been in use for quite a long time. Many standards (to
name 802.11, Bluetooth) are available based on which these devices communicate, but the
present standards fail to provide sufficient data rate, when the user is moving at high speed. In
view of this requirement for future mobile wireless communication systems, the present
standard, 802.16e has been proposed by Institute of Electrical and Electronic Engineers
(IEEE). 802.16 (WiMAX) provides specifications for both fixed Line of sight (LOS)
communication in the range of 10-66GHz (802.16c), and fixed, portable, Non-LOS
communication in the range of 2-11GHz (802.16a, 802.16d). Also it defines wireless
communication for mobiles, moving at speed of 125 KMPH, in the range of 2-6 GHz
(802.16e). 802.16e is well implemented with OFDMA as its physical layer scheme, hence
OFDMA is discussed here.
One of the limiting factors in the performance of mobile wireless communication systems is
the Inter symbol interference (ISI), caused by the multipath. In single carrier systems the
symbol duration (for large system capacity) is very small and spans a wide bandwidth in
frequency domain and the multipath arriving at different time instants is spread over multiple
symbols leading to ISI. The complex solution is to implement an equalizer at the receiver to
mitigate the effect of the channel. A much simpler solution is to opt for multicarrier systems,
like OFDM, which transmit low rate data (large symbol time) on several overlapping
orthogonal subcarriers. In addition a guard time is provided (Figure 0.1) at the start of each
symbol. By doing so, the symbol time is made large enough so that the system becomes less
sensitive to multipath.
Figure 0.1: Three orthogonal subcarriers shown separately (in practice a sum of 3 is
transmitted)
0.3.1 802.16c
Wireless metropolitan area network- single carrier physical layer (WirelessMAN-SC PHY)
specification is targeted for operation in the 1066 GHz frequency band. The BS is essentially
an isotropic radiator, which transmits data (downlink) to all the users designated by their
connection identifier (CID). The subscriber station (SS) shall use highly directional antennas
directed towards the BS. The signal chain for this physical layer, at the transmitter is defined
as, randomization, forward error correction (FEC) encoder, symbol mapping followed by
0.3.2 802.16a
This part of the WiMAX standard uses single carrier (WirelessMAN-SCa PHY) physical
layer specification, similar to that of 802.16c, except that it is targeted for the frequency
below 11 GHz and at NLOS. The SS can be personal computers with an external box connect
to an outdoor isotropic antenna [14]. Hence this is fixed NLOS wireless communication.
Support for both TDD and FDD is provided, similar to 802.16c. Since single carrier in
multipath environment is used, a receiver needs to perform efficient channel estimation and
equalization techniques to overcome the multipath effects. Another difference is the
concatenated FEC using RS and pragmatic trellis coded modulation (TCM) (rate
convolutional coding (CC)) with optional interleaving. Optionally to improve the
performance support is provided for BTC, CTC, Adaptive antenna systems (AAS) and space
time coding (STC) are provided.
802.16a devices can be used to provide with T1/E1 level services to enterprises, thus
eliminating wire lines and saving the implementation cost and time. Additionally it can be
used to provide backhaul for hotspots being served by 802.11. Also it can be used in
residential locations to provide broadband internet connections.
0.3.3 802.16d
This is targeted to provide a broadband internet connection to indoor users. The SS operating
on this standard use indoor antenna and a limited mobility (portable devices) is allowed [14].
802.16d uses orthogonal frequency division multiplexing (OFDM) as its physical layer
specification to enable NLOS communication below 11 GHz. Since OFDM is used, the
receiver is made simple by elimination of bulky equalizer. The other features have nearly
been kept similar in all the physical profiles of the standards. FEC includes concatenated RSCC followed by interleaving. Similar to 802.16a, AAS, STC schemes are provided but are
kept optional. Variable FFT size and symbol time is specified, which could be fixed
depending on type of environment and allocated bandwidth.
0.3.4 802.16e
Specifications are provided such that mobility of the SS at 125 KMPH is allowed. Orthogonal
frequency division multiple access (OFDMA) is used as the physical layer scheme. Channel
Saving of bandwidth
-- (eq. 0.1)
Ideally, to maintain orthogonality we need that the symbol duration be exactly inverse of the
subcarrier spacing and the FFT be considered over symbol duration such that it covers integer
number of cycles. Moreover, the consecutive subcarriers differ by 1 full cycle only (Figure
0.1). If the system is to operate in a multipath environment, then each subcarrier should
experience a flat fading, hence the subcarrier spacing should be less than the coherence
bandwidth and each symbol should experience a time-invariant channel, hence the symbol
time should be less than the coherence time else the complexity of receiver increases when
overcoming the fading effect.
Reduction of inter symbol interference, which would require bulky equalizer to be constructed
at the receiver in a single carrier system, is overcome by the use of guard time in an OFDM
system. A guard time is added in time domain between two OFDM symbols and the FFT is
considered over duration such that there is no component from the previous or next symbol,
(Figure 0.3) which nulls the ISI and thus avoiding the bulky equalizer. ISI is completely
eliminated when the multipath signal delay is within the guard time. When designing an
OFDM system proper values are selected depending on the environment so as to satisfy the
above condition. Multi carrier systems have the problem of inter carrier interference (ICI),
which results from loss of orthogonality between the subcarriers. This happens when the FFT
is considered over duration where the subcarrier is not present (non-integer number of cycles),
which would be the case when multipath is present and the guard time has amplitude zero.
This is reduced by use of cyclic prefix [5], where we transmit a copy the last part of the
symbol followed by the symbol itself. This ensures orthogonality over the FFT period in case
of delayed multipath (Figure 0.3).
Figure 0.3: 3 subcarriers and multipath component shown separately, in practice the signal is
a sum of all subcarriers [5].
1.1 Introduction
A general communication system consists of two blocks, a transmitter and receiver, connected
by a channel. The information transmitted by the transmitter passes through the channel and
then reaches the receiver. If the channel does not distort the transmitted signal, then the
receiver can retrieve the transmitted information successfully, but in practice the channel
alters the transmitted information making the task difficult for the receiver. The main aim of
the designer is to reduce the number of errors made at the receiver. To achieve this,
information is required at the receiver, as to how the channel alters the information, so that the
channel impairments can be mitigated.
When the user is mobile, the channel characteristics do not remain constant for a very long
time. Hence the channel parameters need to be tracked, so that the effect can be mitigated and
reconstruct the transmitted data. This part of the thesis deals with the requirements of Channel
estimation at the Base station (BS) for an 802.16e uplink. Symbol time has an effect on
system performance depending on the channel conditions. Different symbol times are
proposed in [1] and each one has been simulated and compared for various channel condition.
In addition a solution proposed by Intel coop. has also been analyzed. It is concluded that the
performance of the system, for few proposed symbol times, is relatively good in all
conditions.
Symbol 1
Symbol 2
Symbol 3
Data subcarrier
Pilot subcarrier
10
1.3.1 Transmitter
Data Generation:
The data is generated from a random source, consists of a series of ones and zeros. Since the
transmission is done block wise, when forward error correction (FEC) is used, the size of the
data generated depends on the block size used, modulation scheme used to map the bits to
symbols (QPSK, 16QAM), and whether FEC is used or not [1]. The generated data is passed
on to the next stage, either to the FEC block or directly to the symbol mapping if FEC is not
used.
Forward error correction:
In case error correcting codes are used, the data generated is randomized so as to avoid long
run of zeros or ones, the result is ease in carrier recovery at the receiver. The randomized data
is encoded using tail biting convolutional codes (CC) with a coding rate of (puncturing of
codes is provided in the standard, but not simulated here). Finally interleaving is done by two
stage permutation, first to avoid mapping of adjacent coded bits on adjacent subcarriers and
the second permutation insures that adjacent coded bits are mapped alternately onto less or
more significant bits of the constellation, thus avoiding long runs of lowly reliable bits.
Symbol mapping:
The coded bits (uncoded, if FEC not used) are then mapped to form symbols. Modulation
scheme used is QPSK or 16QAM (QPSK unless otherwise specified) with gray coding in the
constellation map. In any case the symbol is normalized so that the average power is unity,
irrespective of the modulation scheme used [1].
Subcarrier allocation:
The subcarrier allocation is mentioned in the section 1.2 (Uplink transmission). This separates
data into set of 4 subcarriers for 3 time symbols, named as the tile structure. Symbols are
allocated indices representing the subcarriers and OFDM time symbol, and then passed onto
the next stage, the IFFT, to convert into time domain.
IFFT and cyclic prefix:
An N point inverse discrete fourier transform (IDFT) of X(k)is defined as
1
x(n ) =
N
N 1
X (k )e
n =0
2kn
N
(eq. 1.1)
From the equation we can infer that this is equivalent to generation of OFDM symbol. An
efficient way of implementing IDFT is by inverse fast fourier transform (IFFT). Hence IFFT
is used in generation of OFDM symbol. The addition of cyclic prefix is done on the time
domain symbol obtained after IFFT. The IFFT size (N value) is considered as 2048 in
11
1.3.2 Channel
In NLOS wireless communication, the received signal is a combination of many multipath
signals, which are result of reflections from surrounding objects. These multipaths have
different amplitude and phase and may add either constructively or destructively leading to a
complex envelope, i.e. fading. Fading characteristics depend on the channel parameters (rms
delay spread and Doppler spread) and signal parameters (symbol period and bandwidth).
Multipath delay spread leads to time dispersion and frequency selective fading and Doppler
spread leads to frequency dispersion and time selective fading. Any mobile channel is one of
the four mentioned below [2] based on
Based on multipath time delay spread
Flat fading
BW of Signal < BW of channel [ Bs << Bc ]
[ Ts << ]
Slow fading
low Doppler speed
[ Tc > Ts ]
12
-- (eq. 1.2)
13
r (t ) = (t )e j (t ) s (t )
-- (eq. 1.3)
According to the standard the maximum supported speed of mobile is 125 KMPH and the
operating frequency range is between 2 6 GHz. The system has been simulated for speeds
30, 80, 125 KMPH and frequency band of 3 GHz and 5.9GHz. Three multipaths were
simulated with uniformly distributed phase. For multipath the amplitude and delay has been
chosen as a random parameter, the first path does not have any excess delay and the amplitude
is scaled by a uniformly distributed number in the range of 0 to 1. The other 2 paths have their
amplitude scaled by uniformly distributed number between 0 to 0.9 and 0 to 0.7. The excess
delay is selected as a uniformly distributed random parameter. Finally additive white
Gaussian noise (AWGN) is added as a last component in the channel.
Doppler _ frequency( f d ) =
velocy(m / s ) * frequency( Hz )
speed _ of _ light (m / s )
14
1.3.3 Receiver
The first thing done at receiver (in simulation) is removal of cyclic prefix, thus eliminating the
inter symbol interference (ISI). Data is then passed through the serial to parallel converter of
size 2048 and then fed to the FFT for frequency domain transformation. The signal was
distorted by the channel, to reconstruct the original signal we need information as to how the
channel acted on the transmitted signal so that we can mitigate its effect. This is called
equalization. In an OFDM system, this is done by channel estimation and interpolation. As
we need at least one tile structure (3 OFDM symbols) to detect the data, storage of 3 OFDM
symbols is provided followed by the subcarrier de mapping. The pilot subcarriers are used for
channel estimation and synchronization at the receiver. In the simulation least squares (LS)
estimate has been used for channel estimation at the pilot subcarriers. If D (t ) is the
transmitted data (known if pilot), Y (t ) is the received data, and C (t ) is the unknown channel
response, then
Y (t ) = D (t ) * C (t ) + N (t )
-- (eq. 1.6)
where N (t ) represents the AWGN noise.
The channel can be estimated for known data symbols, i.e. pilot subcarriers as,
C (t ) =
Y (t )
D(t )
-- (eq. 1.7)
The estimate is simple but is highly affected by SNR or the noise power, as the assumption
made is absence of noise from the receiver power.
This information about channel at pilot subcarriers is interpolated over the whole tile
structure, to recover the data on each data subcarrier (Figure 1.1). Separate one dimensional
linear interpolation has been done for values between two subcarriers (the result: straight
line), hence the performance is not effected much for various one dimensional interpolation
algorithms.
Since we do linear interpolation the channel is assumed to be changing linearly with in the
tile, this assumption might not be true depending on the symbol time. This generates a noise
floor at the receiver (Errors are generated due to addition of AWGN noise and due to this
approximation of fading channel as a linearly varying channel. Beyond a certain value of
SNR, the BER is nearly constant for any further increase in SNR.). For larger symbol time, as
will be seen in simulations, this noise floor is reached at lower SNR, hence results in poor
performance.
15
Randomization
Convolutional
coding
Modulation scheme
(QPSK, 16QAM)
Subcarrier
allocation
S/
P
IFFT
Data bits
P/
S
Cyclic prefix
Interleaving
AWGN
fading
delay
fading
delay
fading
Remove cyclic
prefix
Deinterleaving
FFT
S/
P
P/
S
Subcarrier
demapping
Channel
Estimation
Interpolation
Symbol
Demapping
Convolutional
decoding
Derandomization
BER
calculation
-- (eq. 1.8)
With FD max = 700 Hz, Ts = 100 S , we get min pilot spacing in time domain should be 5.7,
which is well satisfied for 89.6 S of symbol time [3] but not all the values mentioned in the
standard (64 S ,128 S ,256 S ,512 S ,1024 S - ETSI)[1].
Even though the proposal from [3] works well theoretically, the performance is not as
expected, this along with all proposals from [1] are investigated by means of simulation.
17
Vary
INTEL recommendation
RMS delay
Amplitude
QPSK
16QAM
Speed
RMS delay
Frequency Amplitude
QPSK
16QAM
Simulation settings:
Symbol time: 64 S , 128 S , 256 S , 512 S , 1024 S - ETSI
Guard time: 1/4, 1/8, 1/16, 1/32 of symbol time.
Frequency: 3, 5.9 GHz (2-11 GHz specified in [1])
Speed of mobile: 125 KMPH (peak)
Modulation: QPSK, 16-QAM.
FFT size: 2048
Assumptions
Power in guard time is not considered.
18
-- (eq. 1.9)
Figure 1.6: SNR v/s BER for AWGN channel with QPSK modulation
Channel coding improves the performance significantly. The next simulation was done for
AWGN channel with QPSK modulation scheme with rate tail biting convolutional code
(G1 = 171; G2 = 133).
19
20
Figure 1.10: Typical channel at different Doppler frequency (350, 680, 70, 135 Hz)
21
Keeping all the parameters constant, we run the simulation for a carrier frequency of 5.9GHz.
Since the carrier is much higher, the Doppler increases, fading becomes fast and we get
degradation in the performance, which is shown in figure 1.11. Additionally, the response of
convolutional coding on the system with symbol time of 128 S and 256 S has been shown.
It indicates that the response of channel coding is also dependent on the symbol time and may
differ significantly.
The results at two frequencies become much clear when plotted together, in figure 1.12. We
see that the performance degradation is more significant for symbol time 256, 512 and
1024 S . The reason is that for these values the tile structure is not within the coherence time
of the channel.
22
Figure 1.12: Performance of OFDM at 2 carrier frequencies (dotted line for 3GHz and solid
line represents 5.9 GHz)
23
24
a1 U (0,1)
To receiver
Rayleigh
fading
simulator
a2 U (0,0.9)
AWGN
Rayleigh
fading
simulator
a3 U (0,0.7)
25
Figure 1.17: Performance at 3GHz in fading channel with multi path (guard time Ts/8)
Figure 1.18: Performance at 5.9GHz in fading channel with multi path (guard time Ts/8)
26
Figure 1.19: Performance for different amplitudes of multipath. Amplitudes are attenuated
relative to the main path
At the receiver we get an added version of pure sine waves (delayed) on each subcarrier. The
addition does not destroy the orthogonality, because we do not consider the cyclic prefix for
samples in FFT period (figure 1.16), but the addition results in different phase shifts on each
subcarrier [5]. Different phase shift on pilot and data subcarriers and due to the interpolation
scheme being used to equalize the symbol, the phase shift results in performance degradation.
This is seen in figure 1.19 for a time period of 256 S .
Figure 1.20 gives the performance of 16QAM in a Rayleigh fading channel with multi path.
The Doppler is ~680 Hz, and multi path is within the guard time, limited to 8 S . It should be
noted that the plot shows bit error rate versus SNR per bit.
27
28
Figure 1.21: Performance when multipath delay is limited with in 16 S (guard time Ts/4)
Figure 1.22: Performance when multipath delay is limited with in 100 S (guard time Ts/4)
In figure 1.21 we find that, when the delay is limited to 16 S , the system with a symbol time
of 128 S performs the best. Hence it should be the obvious choice in environment where the
maximum excess delay profile is around 16 S . Similarly Figure 1.22 shown simulation
results at 100 S and here also we find system with symbol time of 128 S performs best.
29
30
31
1.6 Discussion
In [1] it is proposed that for a FFT size of 2048, to get different bandwidths, one should use
symbols with different time durations. But, as we see, this does not result in good
performance in all cases. In [3] the use of a fixed time period is proposed and to get variable
bandwidth the size of FFT is to be varied i.e. scalable FFT. The solution seems to be good,
but the value of symbol time used might be worth discussing.
The coherence time and the coherence bandwidth of the channel, form the major parameters
influencing the value of the symbol time. We need to have a flat channel over the subcarriers
bandwidth in frequency domain and 3 subcarriers in time domain. Also from the calculations
we find that the coherence time for the maximum speed and the highest carrier frequency is
600 S . Hence in time domain the OFDM symbol including guard band should have a span
of not more than 200 S (600/3). The recommendation of 89.6 S in [3] well satisfies this
limit. This is also satisfied for a symbol time of 128 S in some cases.
As mentioned in [2] the worst case rms delay spread ( ) in urban environment can be
25 S , which translates to a coherence bandwidth ( Bc 50% correlation) of 8 KHz.
1
Bc =
-- (eq. 1.12)
5
In an OFDM system to achieve orthogonality we need to have symbol time (excluding guard
time) as multiple of 1/subcarrier spacing. Hence for 89.6 S we need to have a subcarrier
spacing of 11.16 KHz and for 128 S we require 7.81 KHz. This shows that 128 S is
theoretically a better option than 89.6 S . Also to remember is that 128 S is more effected
by fading than 89.6 S and in practice fading is more dominant and the maximum rms delay
of high value is seen in very few environments (urban). By observing figure 1.18, 1.21 and
1.22, we can fix the symbol time as either 64 or 89.6 or 128 S depending on the rms delay
profile of the environment [2].
32
33
PART 2: Ranging
2.1 Introduction
Total synchronization of an OFDM system is a very important criterion which should be
fulfilled to avoid any interference (ISI and ICI) leading to performance degradation. In
OFDMA it is required that all transmissions from various mobile subscriber station (MSS)
should arrive at the base station (BS) at the same time. Imagine a cell size of 20 Km. we, then
have a maximum round trip delay (RTD) of around 133.3 S . This means that, instead of
arriving at expected time at BS, data may arrive anytime within 0 to 133.3 S of delay. If the
symbol duration is 64 S , the amount of error in detection of data at BS is very high! The
whole network should be synchronized to one reference and in WiMAX this reference is the
BS clock. All data is expected to arrive at the same time at the BS receiver and all data
addressed to the MSS is transmitted at same time. The MSSs, which are distributed all over
the cell, receive data at different instant of time and similarly transmit data at different instant
of time depending on their distance from the BS. A MSS at the cell boundary receives data
quite late and transmits data very early when compared with MSS close to the BS.
When a new MSS is seeking entry into the network, its distance, with reference to the BS, is
not known hence the RTD is not known. The MSS does not have any idea as to what time or
power should be used for transmitting the initial signal. This is the BSs job to detect this new
MSS, find the misalignment between the new MSS and the network, and then send response,
to correct it.
Reducing system complexity without compromising with performance has been the major
focus of the testing various methods in this part of the thesis. It has been shown that a better
approach is to use frequency domain correlation by using IFFT, which is very simple and
efficient in implementation.
The below section starts with a brief description of this problem and how 802.16e system
handles this situation. Next the report explains the calculation of timing, frequency and power
offset. Major focus is to reduce the complexity of the system, and still maintain the system
performance at an acceptable level. Methods to estimate the timing offset, using both time and
frequency domain correlation have been explained and later corroborated with simulation
results. It is shown that, in time domain, the complexity of implementing a full crosscorrelation is very high and can be significantly reduced if the CDMA code at the receiver is
quantized and represented using just 2 bits.
34
35
Data is represented by using double precision in MATLAB, this is correlated with double
precision representation of the CDMA code (in time domain) to find the correlation peak
corresponding to the time offset (eq. 2.1).
n
-- (eq. 2.1)
Figure 2.1: Initial ranging transmission symbol structure (figure 240 from [1])
The process is made possible only because of the repetitive structure in which the data is
transmitted. Note that the second symbol does not have a cyclic prefix but a post fix! This fact
36
R ( m) =
m + N 1
-- (eq. 2.2)
k =m
Window a
Time
The above procedure can be simply represented as one time correlation over the total window
and the next step is nothing but one subtraction (of the first value) and one addition (of the
last value) to the previously obtained value. Thus the calculation is significantly reduced.
After locating the time offset plateau, cross-correlation is performed between the localized
data with different codes. One single code will have a good cross-correlation value and we get
two correlation peaks because of repetition of the code twice as in Figure 2.1. This is the
bases for calculation of frequency offset.
37
-- (eq. 2.3)
Delay
Rx data
CDMA code
CDMA code
FFT window
FFT window
FFT window
The detection of timing offset constitutes the following steps; the ranging subcarriers are
separated from data subcarriers by taking the FFT over the received signal. The data on the
ranging subcarriers are provided with different phase shifts (0 to 2 ) and correlated every
time with the CDMA code (in frequency domain), to check for the peak. The number of steps
in between 0 to 2 provides us with the resolution in detection of timing offset. For detecting
integer sample delay, the number of steps should be equal to size of the FFT. As shown in the
figure 2.3, the FFT is considered over the received data and since there is a delay involved,
only one of the three FFTs would correspond to the circularly shifted code. It is this FFT
output, that gives us the timing offset value (second FFT in figure 2.3). Then we multiply it
with phase component to get the required phase shift corresponding to integer number of
samples in time domain, finally we correlate it with the CDMA code. Simulation results prove
that this system works very well because of reduced interference from the data subcarriers,
but the complexity is very high because of the high number of multiplications required to
achieve the phase shifts.
Instead of providing phase shifts by multiplication of phase component, a very simple method
is to take an IFFT after multiplication of code in frequency domain. IFFT is equivalent to
providing phase shift (figure 2.4), hence reduces a lot of complexity.
X (k ) =
N 1
x ( n )e
1
x( n) =
N
j 2kn / N
n =0
Signal in
x ( n n 0 )
FFT
N 1
X ( k )e
k =0
IFFT
j 2kn0 / N
X (k )
j 2kn0 / N
CDMA code X (k )
38
j 2kn / N
X (k ) X (k )
For n = n0
we get a peak.
Of all the methods investigated here, the frequency domain correlation method mentioned
above, turns out to be the simplest in terms of complexity. Moreover the performance, as seen
from the simulation results is better. Hence, the frequency domain correlation method is a
better choice for detecting the time offset of the ranging MSS.
-- (eq. 2.4)
When the MSS achieves coarse synchronization, it decodes the UCD message which contains
information as to the maximum power that the BS can receive and the power which was
transmitted by the BS. The MSS calculates the Received signal strength and computes the
losses in the channel and calculates the maximum power that it can use for transmitting the
ranging request (CDMA code). After acquiring such information it will transmit at a power
level below the maximum level and start the ranging process, if it does not get a response
back form the BS, then it transmits the CDMA code at a higher power level. If in case the
MSS has achieved the coarse synchronization and yet unable to decode the UCD message, it
will start transmitting at the lowest possible power level and increasing it a level higher until
it receives a response. Power offset at the BS is simply calculated as (required power at BS received power).
39
Figure 2.5: Comparison of various correlation techniques when MSS and ranging MSS
experience different channel (ranging MSS on 6 subchannels and data transmitting MSS on
64 subchannels)
Also plotted are the results for correlation performed in frequency domain, over a window
size of 2048 samples. Since there is less interference the detection probability of the new user
is more, compared with correlation in time domain. As the CDMA code, in frequency domain
is represented by 2 bits, there is no quantized correlation in frequency domain.
Figure 2.6 shows the simulation results for cross-correlation performed with the same settings
as above (64 S symbol time), except that the data is transmitted on 32 subchannels instead
40
Figure 2.6: Cross-correlation with double precision code and with 2 bit quantized code
(ranging MSS on 6 subchannels and data transmitting MSS on 32 subchannels)
41
Figure 2.7: Effect of ranging signal amplitude on system performance (ranging MSS on 6
subchannels and data transmitting MSS on 64 subchannels)
42
4256
CDMA code
Figure 2.8: cross-correlation: multiplication between parts of received data and CDMA code
Computations required
(2048 multiplies + 2048 additions) * 4256 * n
6144 additions * 4256 * n
(6464 multiplies + 6464 additions) + 6144 additions * m * n
19392 + 6144 * m * n additions
(144 multiplies + 144 additions) * 6144 * n
Table 2.1: Table showing scheme used and computations required to execute it
43
44
45
3.1 Introduction
The present demand in the field of wireless communication is not only to provide data
communication when the user is mobile but also to provide high data rate by consuming less
bandwidth (achieve good spectral efficiency). Moreover, the system complexity and its
implementation are of major concern and sometimes limit the implementation of efficient
techniques. Efficient channel coding schemes and diversity schemes are used to achieve high
system capacity at less power. WiMAX, the IEEE standard provides specification for efficient
forward error correction techniques and optional schemes like adaptive antenna system
(AAS), space time coding (STC) and multi input multi output (MIMO) systems. Of these
AAS achieves high system capacity with implementation cost mainly concentrated at the base
station (BS), which can be easily tolerated. Hence is a good solution for increasing system
capacity with least cost.
Transmitting data from a single antenna to cover the entire cell (isotropic) is difficult and has
many disadvantages hence sectorized antennas are widely used in wireless communications,
where the cell is divided into sectors; 3 or 6 generally. This gives many advantages as reduced
interference, increased range and SNR, at the cost of some complexity at the transmitter (BS)
which is usually acceptable. Beamforming is the next step used to further improve the
performance of the system, where the number of sectors are many more! Apart form the
advantages gained by using sectorization, Space division multiple access (SDMA) can be
implemented, thus increasing the system capacity.
In simple terms, beamforming can be explained as Instead of using incandescent lamp
throwing light all over the room, use a torch, with same power, and point it in the direction
where you want to see. Certainly you will see more clearly; in terms of Antennas this is
beamforming. Beamforming is achieved by making the antenna array radiation pattern, point
in one particular direction; this is obtained by using an array of antenna, fed with same signal
at different time instant or provided with phase shifts.
A Ranging MSS trying to enter an AAS system poses two problems to the base station: one is
the unknown time offset and another is the unknown direction of arrival of signals. A simple
solution to this problem is to use a single antenna output to detect the time and frequency
offset information and uses this information to find data on which we can run the direction of
arrival (DOA) algorithm. The present thesis simulates two DOA algorithms and compares the
performance for 802.16e, OFDMA system in uplink. Rayleigh fading channel has been
assumed during the simulation.
46
1
Et = w
f ( , ) cos (kd cos + )
--- (eq. 3.1)
r
N
axis
r
N = number of antenna elements
d
2
= phase shift between antenna elements
= elevation angle
f ( , ) = field pattern of single element
d = distance between two antenna elements
Figure 3.2: 2 element linear array
far field geometry
This can be written as E (total) = [E (single element at reference point)] X [(array factor)]
Where the normalized array factor is
1
47
-- (eq. 3.3)
The Half power beam width (HPBW) of any antenna radiation pattern [12] is defined as In
the plane containing the direction of the maximum of a beam, the angle between two
directions in which the radiation intensity is half the maximum value of the beam. The
HPBW varies (figure 3.3) for different angles (minimum for 90 and maximum for 0). To
differentiate between users in the same cell we would require the HPBW to be small, hence in
practice, sectorized antenna array is used [13], each array operating for 30 to 150 degrees.
Figure 3.3: Plot of Array factor (AF) for 6 linearly arranged / 2 dipole antenna elements
with spacing of lamda/2
48
Direction of
arrival
computation
The number of beams which can be formed is such that we can either allow N-1 maxima
(users) or block N-1 minima (interferers) or a combination of two, where N is the number of
antenna elements in the array. This flexibility of an N element array to fix the pattern at N-1
places is termed as degree of freedom [9]. It is cost effective to implement the antenna array
at the BS instead at the MSS.
Signal out
Direction of
arrival
computation
Complex
weight
selection
Signal out
Complex
weight
computation
49
50
51
AAS-DLFP
AAS-DLFP
AAS-DLFP
AAS-DLFP
52
u (t ) = a ( l )sl (t ) + n(t )
-- (eq. 3.4)
l =0
Where
Ruu = uu H
-- (eq. 3.5)
1
R uu =
K
K 1
u u
K =0
H
k
-- (eq. 3.6)
PCapon ( ) =
1
a H ( )R uu1a ( )
-- (eq. 3.7)
The input data is divided into set of blocks and the covariance matrix is estimated over each
block consisting of K input samples, followed by the spectrum estimation. The same process
of estimating spectrum is repeated for many blocks and an average is considered over them to
obtain the estimate of Capons spectrum.
The peaks in the spectrum determine the transmitting user location. The method requires
estimation of matrix inverse, which could be highly complex in case of large arrays.
53
R uuV = V
Where
uu
-- (eq. 3.8)
V = [q 0 , q 1 ...... q M
uu
If there are D input signals from different directions (such that M>D), then the noise
subspace is given by Vn = [q D , q D +1 ......q M 1 ] . The noise subspace and signal subspace are
orthogonal to each other. The steering vectors corresponding to the direction of arrival of the
signals lie in the signal subspace and are orthogonal to the noise subspace.
Hence, a H ( )VnVnH a( ) = 0 when corresponds to the direction of arrival of the incident
signal.
a H ( )a( )
PMU ( ) = H
a ( )VnVnH a( )
-- (eq. 3.9)
The DOA corresponds to a peak in the spectrum, resulting due to denominator approaching
zero
The numerator results in a constant value equal to the number of antennas in the array hence
can be removed. Another significant improvement can be gained by avoiding the reciprocal in
the equation and by doing so the DOA is represented by a dip in the spectrum, instead of a
peak.
54
Figure 3.8: Capons method for DOA computation (DOA of signals from two MSS is not
distinguished).
Figure 3.9: MUSIC method for DOA computation, DOA of signals from two MSS is seen
55
By observing the MUSIC spectrum, we find that the values in and around the DOA of the
signal have the amplitude higher than 1, which would be obtained only if the denominator is
less than 1. Hence while searching for dips; we can rearrange the y axis, so as to get the
range as 0 to 1. This results in very fine resolution as shown in figure 3.10. In general, to
avoid non detection of DOA, the range of y axis can be set from 0 to 50.
56
In all cases MUSIC method for DOA computation outperforms Capons method.
3.8 Conclusion
As we see beamforming can be implemented very easily at the BS, and the problem of two
unknowns is solved one after another. First we find the unknown time offset for the ranging
MSS by considering only a single antenna element output, followed by DOA computation on
the data following the time offset. Though, while calculating the time offset, we loose, by not
considering all the antenna elements, but it is compensated by the gain from the CDMA code
used during ranging.
The DOA computation can be accurately computed by using the spectrum based method like
MUSIC algorithm. Reduction in complexity is obtained by searching for dips instead of
peaks.
57
58
59
A. References
1. IEEE Standard for Local and metropolitan area networks, Part 16: Air Interface for
Fixed Broadband Wireless Access Systems, IEEE Std 802.16-2004, 1 Oct 2004.
2. Theodore S. Rappaport. Wireless communication: Principles and practice, second
edition, Pearson Education, Inc.
3. Hassan Yaghoobi, Scalable OFDMA Physical Layer in IEEE 802.16 WirelessMAN,
Intel Communications Group, Intel Corporation. 2004.
4. A.A. Hutter, R. Hasholzner, J.S. Hammerschmidt. Channel estimation of mobile
OFDM system - Institute for Integrated Circuits, BRIDGELAB Digital Signal
Processing, Technical University of Munich, D - 80290 Mnchen, Germany.
5. Richard Van Nee, Ramjee Prasad. OFDM for wireless multimedia communications,
Artech house publisher, 2000
6. Niklas Parnesten. Phase and frequency offset correction in a combined OFDMCDMA system, Master thesis, radio communication systems laboratory, Royal
institute of technology, Sweden. February 1997.
7. Draft IEEE Standard for Local and metropolitan area networks Part 16: Air Interface
for Fixed and Mobile Broadband Wireless Access Systems, Amendment for Physical
and Medium Access Control Layers for Combined Fixed and Mobile Operation in
Licensed Bands, IEEE P802.16e/D5, September 2004
8. Saeed, M.A.; Ali, B.M.; Habaebi, M.H. Performance evaluation of OFDM schemes
over multipath fading channels. The 9th Asia-Pacific conference on
Communications, 2003, APCC 2003. Volume: 1, Pages:415 419, 21-24 September
2003
9. Lal Chand Godhra, Smart antennas CRC press, 2004.
10. George Jngren, Utilizing channel state information in space time coding:
performance limits and transmission techniques, PhD thesis, Department of signals,
sensors and systems, Royal institute of technology, Sweden. 2003.
11. Shetty, Kiran Kumar, A novel algorithm for uplink interference suppression using
smart antennas in mobile communications, Master thesis, Department of Electrical
and Computer Engineering, Florida state university. February 2004.
12. Constantine A. Balanis, Antenna theory, second edition, John Wiley & sons, Inc.1997.
13. Joseph C. Liberti, JR., Theodore S. Rappaport, Smart antennas for wireless
communications: IS 95 and third generation CDMA applications, Prentice hall
publications, 1999.
60
14. Laine Philippe, Boscher Christophe, Boettle Dietrich, Feijt Laurence, WiMAX,
making ubiquitous high-speed data services a reality, Whitepaper, Alcatel. June 28,
2004.
15. Tony Ottosson, Arne Svensson, Pl Frenger, Bartosz Mielczarek, Maxime Flament,
Hongxia Zhao, Simulation of a mobile radio system, laboratory manual and project
2, 2005.
16. Monson H. Hayes, Statistical digital signal processing and modeling, John Wiley &
sons, Inc. 2002.
17. Menam K. Ragheb., S. H. Elramly and Sami Mostafa, The Effect Of Varying The
Input Parameters On The Performance Capabilities Of Three Different DOA
Estimation Algorithms Using Smart Antennas, Twentieth national radio science
conference, March 18-20, 2003, Cairo, Egypt.
18. Maxime Flament, Arne Svensson, John M. Cioffi, On the design of an OFDM system
for 60 GHz wireless LAN, Chalmers university of technology & Stanford university
19. ADSP-TS201 TigerSHARC processor programming reference, Revision 1.0, August
2004
20. Richard V. Cox, Carl-Erik W. Sundberg, An efficient adaptive circular viterbi
algorithm for decoding generalized tailbiting convolutional codes, IEEE transactions
on vehicular technology, Volume 43 No.1, February 1994.
21. Draft IEEE Standard for Local and metropolitan area networks Part 16: Air Interface
for Fixed and Mobile Broadband Wireless Access Systems, Amendment for Physical
and Medium Access Control Layers for Combined Fixed and Mobile Operation in
Licensed Bands, IEEE P802.16e/D6, February 2005.
22. John G. Proakis, Masoud Salehi, Communication systems engineering, second edition,
Pearson Education Inc. 2003.
61
B. Acronyms
AAS
AF
AWGN
BCC
BER
BPSK
BR
BS
BTC
BW
BWA
CDMA
CID
CP
DAMA
DCD
DSP
DIUC
DL
EIRP
FDD
FEC
FFT
GS
HPBW
ICI
IEEE
IFFT
ISI
LDPC
LMS
LOS
MAC
MAN
Mb/s
MIMO
MISO
MSS
NLOS
OFDM
OFDMA
PHY
PMP
PPP
PRBS
QAM
QoS
QPSK
62
RADAR
RNG
RS
RTD
Rx
SDMA
SIMO
SNR
SS
STC
TCM
TDD
TDM
TDMA
Tx
UCD
UIUC
UL
WiMAX
WirelessMAN
63