This action might not be possible to undo. Are you sure you want to continue?
Institute of Communications and RadioFrequency Engineering
Master Thesis
CPICH Power Optimization for MIMO HSDPA
Andreu Mateu Torrelló
executed for the purpose of obtaining the Master degree MASTEAM
Vienna March 2009
under the direction of
Projektass. Dipl.Ing. Martin Wrulich
Univ.Prof. Dipl.Ing. Dr.techn. Markus Rupp
Abstract
The increasing demand of bandwidth from new network services has reached mobile
technologies. Multipleinput multipleoutput (MIMO) Highspeed downlink packet
access (HSDPA) can be the answer to this demand, being able to double the data
rate of its predecessor singleinput singleoutput (SISO) HSDPA. The optimization of
mobile network is seen as an important issue from network operators, which see the
opportunity to reduce costs by optimizing their networks through simulations. This
thesis presents an overview of the main characteristics of enhanced MIMO HSDPA,
describes the basic MIMO HSDPA linklevel simulator we built and consequently
focuses on the common pilot channel (CPICH) and its optimization. Furthermore,
we derive a computationally eﬃcient improvement, of the a linkmeasurement model
utilized in a MIMO HSDPA systemlevel simulator, to take the eﬀects of CPICH power
variation into account.
i
Acknowledgements
I would like to thank Martin Wrulich, who has guided me through this path called
thesis, for his constant support and his optimistic approach, but especially for his
patience.
I would also like to thank Professor Markus Rupp for giving me the opportunity to
work in such a friendly environment, I felt like at home. My thank also goes to
the whole institute of communications and radiofrequency engineering and my thesis
students colleagues for helping me to clarify my ideas at any time during the day or
night.
A la meva família, per donarme suport en els moments més difícils, perquè tenirvos
al costat tot i que no sigueu a prop fa que em senti amb conﬁança per afrontar nous
reptes.
Finally, I would to thank you, who helped me somehow at some point over the last 5
years, because without you I never would not have made it.
Palma de Mallorca Andreu Mateu Torrelló
April 4, 2009
iii
for my mother
v
Contents
1. Introduction 1
1.1. MIMO HSDPA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2. How important is the common pilot channel? . . . . . . . . . . . . . . 1
1.3. Topics covered in this Master Thesis . . . . . . . . . . . . . . . . . . . 2
2. HSDPA Basics 3
2.1. Standardization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2. WCDMA principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2.1. DirectSequence Code Division Multiple Access . . . . . . . . . 4
2.2.2. Spreading and despreading . . . . . . . . . . . . . . . . . . . . 6
2.2.3. Radio propagation . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3. HSDPA principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3.1. HSDPA vs Release 99 DCH . . . . . . . . . . . . . . . . . . . . 11
2.3.2. Highspeed downlink shared channel . . . . . . . . . . . . . . . 12
2.3.3. Multiple Input Multiple Output HSDPA . . . . . . . . . . . . . 14
2.3.4. HSDPA physical layer operation procedure . . . . . . . . . . . 16
3. Basic MIMO HSDPA LinkLevel Simulator 17
3.1. General structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2. Block Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2.1. Common Pilot Channel & Highspeed Downlink Shared Channel 18
3.2.2. Transmission modes . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2.3. Channel model . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2.4. Equalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.2.5. Channel Estimation . . . . . . . . . . . . . . . . . . . . . . . . 29
3.3. Implementation issues . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4. Mobile Network Simulations 35
4.1. SystemLevel Simulations vs. LinkLevel Simulations . . . . . . . . . . 35
4.2. MIMO HSDPA System Level Simulator . . . . . . . . . . . . . . . . . 36
4.2.1. General Structure . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2.2. LinkMeasurement Model . . . . . . . . . . . . . . . . . . . . . 37
4.2.3. LinkPerformance Model . . . . . . . . . . . . . . . . . . . . . . 41
vii
4.3. Linkmeasurement model enhancement . . . . . . . . . . . . . . . . . . 42
4.3.1. Validation of the current model . . . . . . . . . . . . . . . . . . 42
4.3.2. Inﬂuence of CPICH in the current model . . . . . . . . . . . . 42
4.3.3. Modeling the eﬀects of CPICH . . . . . . . . . . . . . . . . . . 44
4.3.4. Validation of the enhanced model . . . . . . . . . . . . . . . . . 46
5. CPICH Power Optimization 47
5.1. Importance of CPICH power optimization . . . . . . . . . . . . . . . . 47
5.2. Simulation methodology . . . . . . . . . . . . . . . . . . . . . . . . . . 49
5.2.1. Preequalization SINR . . . . . . . . . . . . . . . . . . . . . . . 49
5.2.2. HSDSCH SINR optimization . . . . . . . . . . . . . . . . . . . 51
5.3. Simulation results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.4. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
A. Appendix 57
A.1. Channelisation Codes . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
A.2. Scrambling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
List of Figures
2.1. 3GPP release timeline [1] . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2. Spread Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3. Spreading and Despreading . . . . . . . . . . . . . . . . . . . . . . . . 7
2.4. Modulation constellations . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.5. Generic downlink transmitter structure MIMO DTxAA [2] . . . . . . 16
3.1. Basic linklevel simulator block diagram . . . . . . . . . . . . . . . . . 18
3.2. Relationship between modulation, spreading and scrambling [3] . . . . 19
3.3. Frame structure for Common Pilot Channel . . . . . . . . . . . . . . . 20
3.4. Transmission schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.5. Transmission scheme comparison . . . . . . . . . . . . . . . . . . . . . 23
3.6. Rayleigh fading for diﬀerent speed . . . . . . . . . . . . . . . . . . . . 25
3.7. PDP Pedestrian B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.8. MMSE equalizer structure . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.9. Overall impulse response . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.10. Comparison of channel estimators . . . . . . . . . . . . . . . . . . . . . 32
3.11. Basic MIMO HSDPA linklevel simulator ﬁle structure . . . . . . . . . 33
4.1. Schematic block diagram of system level simulations [4] . . . . . . . . 36
4.2. MIMO HSDPA systemlevel simulator main structure . . . . . . . . . 38
4.3. System Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.4. Model validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.5. MSE channel coeﬃcients . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.6. Eﬀect of CPICH in the current model and LS estimator . . . . . . . . 45
4.7. Model validation and LS estimator . . . . . . . . . . . . . . . . . . . . 46
5.1. Cell resizing eﬀect due to CPICH power variation . . . . . . . . . . . . 48
5.2. Path loss ﬁgures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
5.3. I
all
[dBW] of the target sector . . . . . . . . . . . . . . . . . . . . . . . 51
5.4. I
all
Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.5. SISO . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.6. MISO 2x1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.7. MIMO 2x2 TxAA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.8. MIMO 2x2 DTxAA . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
ix
A.1. Codetree for generation of OVSF codes . . . . . . . . . . . . . . . . . 57
Glossary
3GPP 3rd generation parnership project
AWGN Additive white gaussian noise
BLER Block error ratio
CPICH Common pilot channel
CQI Channel quality indicator
DCH Dedicated channel
DSCDMA Directsequence code division multiple access
DSCH Downlink shared channel
EDCH Enhanced DCH
FACH Forward access channel
GSM Global system for mobile
HARQ HybridARQ
HSDSCH Highspeed downlink shared channel
HSPDSCH Highspeed physical downlink shared channel
HSSCCH Highspeed shared control channel
HSDPA Highspeed downlink packet access
ISI Intersymbol interference
ITU International telecommunication union
LOS Lineofsight
LS Least squares
MAI Multiple access interference
MIMO MultipleInput MultipleOutput
MISO Multipleinput and singleoutput
xi
MMSE Minimum mean squared error
MNO Mobile network operators
MRC Maximum ratio combining
MSE Mean square error
NLOS Nonlineofsight
OVSF Orthogonal variable spreading factor
PCPICH Primary CPICH
PDP Power delay proﬁle
RNC Radio network controllers
RRM Radio resource management
Rx receiver
SCCPCH Secondary common control physical channel
SCPICH Secondary CPICH
SF Spreading factor
SINR Signaltonoiseandinterference ratio
SISO Singleinput singleoutput
SNR Signaltonoise ratio
STMMSE Spacetime minimum mean square error
TFC Transport format combination
TTI Transmission time interval
Tx Transmitter
UE User equipment
UMTS Universal mobile telecommunications system
WCDMA Wideband code division multiple access
1. Introduction
1.1. MIMO HSDPA
Mobile radio communication represents one of the most persistent growing technol
ogy markets since the introduction of the Global System for Mobile communications
(GSM). The Universal Mobile Telecommunications System (UMTS) was the evolu
tion introducing Wideband Code Division Multiple Access (WCDMA) technology in
mobile networks. Since then, the increasing availability of a broad range of new high
speed data services is fuelling demand for more bandwidth in order to improve the
user experience.
HighSpeed Downlink Packet Access (HSDPA) was introduced in Release 5 by the
third Generation Parnership Project (3GPP) as the evolution of UMTS to give re
sponse to the new bandwidth demand. The natural evolution is HSPA+, which in
cludes an enhanced version of HSDPA, doubles the data capacity and increases voice
capacity by three times enabling operators to oﬀer mobile broadband at even lower
cost.
MultipleInput MultipleOutput (MIMO), was introduced in Release 7, is one of the
main techniques introduced to allow this increase of bandwidth. The optimization
of MIMO HSDPA is on the mind of any operator wishing to launch this new sys
tem.
1.2. How important is the common pilot channel?
In HSDPA, as well as in UMTS, channel estimation is accomplished through the use
of a signaling channel. Channel estimation is an essential part of a mobile system,
because is the element that provides the knowledge of the channel coeﬃcients which
are crucial for the equalizer performance.
The Common Pilot CHannel (CPICH) is the signaling channel which aids the chan
nel estimation. The variation of the CPICH power produces two main eﬀects. On
the one hand, the CPICH is used in the measurements for the handover and cell se
lection/reselection. The use of CPICH reception level at the terminal for handover
1
1. Introduction
measurements has the consequence that, by adjusting the CPICH power level, the cell
load can be balanced between diﬀerent cells. Reducing the CPICH power causes part
of the terminals to hand over to other cells, while increasing it invites more terminals
to hand over to the cell, as well as to make their initial access to the network in that
cell. On the other hand, the increase of the CPICH power derives better performance
of the channel estimator, since the better the channel coeﬃcients are estimated the
higher the equalizer performance.
In this study we will focus on:
1. Enhancing the linkmeasurement model of the MIMO HSDPA systemlevel sim
ulator in [5] to take the eﬀects of CPICH power variation into consideration.
2. Determining an optimal CPICH power value for MIMO HSDPA networks.
1.3. Topics covered in this Master Thesis
To give a short overview, the thesis starts in Chapter 2 by presenting the principles of
WCDMA and HSDPA, making a comparison on how the data is handled in HSDPA
and UMTS. Chapter 3 covers in detail the design of a basic MIMO HSDPA linklevel
simulator. In Chapter 4 we discuss the importance of the combination of linklevel
and systemlevel simulators to provide a comprehensive study of the mobile network
technologies and present the systemlevel model improvement. The CPICH power
optimization is treated in Chapter 5. Additional material regarding some speciﬁc
parts may be found in the Appendix.
2
2. HSDPA Basics
This chapter introduces the standardization of HSDPA in Section 2.1. In addition,
the principles of WCDMA and HSDPA are presented in Sections 2.2 and 2.3 respec
tively.
2.1. Standardization
3GPP is the forum created at the ends of 1998 by US, Europe, Korea and Japan as the
result of the desire to introduce a new single global standard for mobile communication
based on WCDMA technology.
At the end of 1999 the ﬁrst release was published, termed Release 99, containing the
ﬁrst full series of WCDMA speciﬁcations. Two years later, Release 4 speciﬁcations
were issued. In the meantime it became obvious that some improvement for packet
access would be needed [1].
A feasibility study for HSDPA was started in March 2000. The study was initially
supported by Motorola and Nokia from the vendor side and BT/Cellnet, TMobil
and NTT DoCoMo from the operator side. The study comprehended a set of im
provements to be done over Release 99 speciﬁcation. The main topics included phys
ical layer retransmissions, BTSbased scheduling, adaptive coding and modulation,
multiantenna transmission and reception technology, called MIMO, as well as fast
cell selection.
The feasibility study showed that substantial improvement could be reached by the
introduction of some of the studied techniques. Accordingly, HSDPA speciﬁcations
were published in Release 5 in March 2002, as shown in Figure 2.1. MIMO was not
included in Release 5 speciﬁcations but later in Release 7 termed HSPA+ which is an
enhanced version of HSPA [6]. Fast cell selection was discarded since it was concluded
that the complexity introduced would not justify the beneﬁts [1].
3
2. HSDPA Basics
2.2. WCDMA principles
WCDMA is a Wideband DirectSequence Code Division Multiple Access (DSCDMA),
also known as directsequence spread spectrum. The following sections describe the
technology of the 3rd generation of mobile communications, especially emphasizing
the technology advantages and disadvantages.
2.2.1. DirectSequence Code Division Multiple Access
Code Division Multiple Access is a multichannel access method. The main idea
of this technology is to allow for sending independent information simultaneously
over a single communication channel meaning that several users share common band
width.
Spread spectrum uses wide band, noiselike signals, see Figure 2.2. The user data
signal is spread over a wide bandwidth by multiplying the user data with a code
sequence of quasirandom symbols (called chips). Because spread spectrum signals
are noiselike, they are hard to detect. Spread spectrum signals are also hard to
intercept or demodulate. Further, spread spectrum signals are harder to jam than
narrowband signals. These low probability of intercept and antijam features are why
the military has used spread spectrum techniques for so many years. Spread signals
are intentionally made to occupy a much larger bandwidth than the information they
are carrying to make them look more noiselike.
Spread spectrum transmitters use similar transmit power levels than narrowband
transmitters. However, they transmit at a much lower spectral power density than
narrowband transmitters. Spread and narrow band signals can occupy the same band,
with little or no interference. This capability is the main reason for all the interest in
spread spectrum today [7].
2000 2001 2002 2003 2004 2005 2006 2007
2000 2001 2002 2003 2004 2005 2006 2007
Release 99
12/1999
Release 4
03/2001
Release 5
03/2002
Release 6
12/2004
Release 7
06/2007
Figure 2.1.: 3GPP release timeline [1]
4
2.2. WCDMA principles
Spread
Signal Bandwidth
Original
Signal Bandwidth
Noise
frequency
Figure 2.2.: Spread Spectrum
DSCDMA advantages and disadvantages
The use of DSCDMA in cellular communications introduces several improvements and
has certain advantages with respect to other multiple access schemes such as TDMA
or FDMA. Let us brieﬂy introduce some of them [7, 8]:
1. DSCDMA systems imply a universal frequency reuse in each cell as all the cells
use the entire available bandwidth. The frequency channel concept disappears
making frequency planing much easier.
2. Narrowband interferences are now practically harmless. These interferences only
aﬀect certain parts of the spectrum. Because of the signal spreading, this nar
rowband interference will only have some impact to a small set of the signal
(and thus, it aﬀects only to a relatively small part of the overall power, as signal
spreading implies also the spread of power along all the spectrum).
3. The lack of frequency channels enables User Equipments (UE) to be connected
to more than one radio base station. Due to this, soft handovers can be done
as well as macrodiversity techniques, which allows Radio Network Controllers
(RNC) to combine diﬀerent base station signals in order to make reception more
robust.
4. Communication privacy is increased because only the transmitter and the re
ceiver know the pseudonoise sequence to despread the signal, and thus can
decode the signal.
Nevertheless, CDMA systems have a certain number of disadvantages which are enu
merated as follows:
1. Power requirements should be strictly controlled as DSCDMA schemes are par
ticularly sensitive to the NearFar problem
1
. Despreading is more diﬃcult when
1
UEs may use the same carrier frequency and are distinguished only by the use of diﬀerent spreading
codes. In this case, the position of the users becomes relevant i.e. a UE closer to the base station
5
2. HSDPA Basics
another station emits at a higher power. Due to this a power control system
must be implemented [9].
2. Self jamming might appear. This eﬀect happens when the supposed orthogonal
coded sequences do not cancel each other perfectly due to multipath propagation.
3. A rigid chipleveled synchronization is needed between transmitter and receiver.
2.2.2. Spreading and despreading
The spreading operation is the multiplication of each symbol of a data sequence of
rate R with a code sequence of SF symbols, called chips. The result is a spreaded
sequence of the data sequence at a rate of SF × R. Spreading codes are chosen to
be orthogonal among each other, i.e. the inner product between two codes is zero,
see Appendix A.1. This wideband signal would then be transmitted across a wireless
channel to the receiving end. The chip rate used in WCDMA is 3.84 Mcps leading to
a carrier bandwidth of approximately 5 MHz.
At the receiver, the despreading consists of the multiplication of the spread data/chip
sequence with the very same SF code chip sequence that was used during the spreading
of these symbols. Then the receiver integrates the resulting products for each user
symbol. The increase of the data rate by SF corresponds also to an increase of the
spectrum occupied by a same factor. Despreading restores a bandwidth proportional
to R for the signal.
The Figure 2.3a depicts a spreading example when the chip sequence has length 5,
and the transmitted user data sequence is modulated in BPSK and has a rate R.
We see that the resulting sequence acquires the same appearance as the spreading
code. During the despreading in the receiver, see Figure 2.3b, the original transmitted
sequence can be recovered perfectly thanks to the orthogonality of the codes which
have a crosscorrelation equal to zero; in other words, they do not interfere with each
other. Note that perfect synchronization is presumed [9].
2.2.3. Radio propagation
Air is the access medium for mobile communications. Also known as radio channel, this
medium is highly hostile compared to the cabled transmission mediums.
may block a large part of the other users farther in the cell.
6
2.2. WCDMA principles
Sent Data
1
1
1
1
1
1
Spreading code sequence
Spread signal
symbol
chip
(a) Spreading
Spreading code sequence
Received Data
1
1
1
1
(b) A large subﬁgure
Figure 2.3.: Spreading and Despreading
Many propagations mechanisms are involved in radio channel transmissions that have
eﬀects on the signal. This section will deal with these mechanisms and its conse
quences. Besides, we will also focus on the propagation models that have been used
in the present study.
Propagation mechanisms in space with objects
As covered in the introduction of this section, radio channel transmissions suﬀer from
a series of mechanisms that aﬀect the signal quality. The main ones are reﬂection,
diﬀraction and scattering [10].
• Reﬂection occurs when a propagating electromagnetic wave impinges on a smooth
surface with very large dimensions compared to the signal wavelength (λ).
• Diﬀraction occurs when the radio path between the transmitter and receiver
is obstructed by a dense body with large dimensions compared to (λ), causing
secondary waves to be formed behind the obstructing body. Diﬀraction is a phe
nomenon that accounts for energy traveling from transmitter to receiver without
a lineofsight path between the two. It is often termed shadowing because the
7
2. HSDPA Basics
diﬀracted ﬁeld can reach the receiver even when shadowed by an impenetrable
obstruction.
• Scattering occurs when a radio wave impinges on either a large rough surface or
any surface whose dimensions are on the order of (λ) or less, causing the reﬂected
energy t spread out (scatter) in all directions. In an urban environment, typical
signal obstructions that yield scattering are lampposts, street signs and foliage.
These propagation mechanisms inﬂuence the signal propagation and cause diﬀerent
eﬀects such as path loss, shadowing or multipath loss, which will be explained fur
ther on. These eﬀects can be grouped as smallscale fading and largescale fading
[10].
Smallscale fading
Smallscale fading aﬀects the instantaneous signal power and therefore aﬀects the link
quality. There are fast and sudden changes on the signal power, which can vary 30 or
40 dBs in only a few seconds or within a few wavelength variations (λ). Fading can also
create signal phase shifts (that is, changes in the signal space) and signal dispersion
in time, known as echos. Fading is mainly caused by multipath propagation and by
the Doppler eﬀect which is caused by changing channel conditions (i.e. movement)
[8, 7].
• Multipath propagation. Due to the reﬂection, diﬀraction and dispersion phe
nomenons, several signal versions are formed and arrive at the receiver with
diﬀerent phases, delays and amplitudes and as a result the signal suﬀers from
large and fast power shifts. On the other hand, delays can cause symbol overlaps.
This eﬀect is known as Intersymbol Interference (ISI). The resulting signal can be
represented as the sum of all the signal versions. So, for an emitted x(t) signal,
the received signal, formed from N versions, will follow the next expression:
y(t) =
N
¸
i
a
i
(t, τ)x(t −τ
i
(t))e
jφ
i
(t,τ)
(2.1)
where a(t, τ) is the signal attenuation, τ
i
(t) represents the delay and φ
i
(t, τ) the
phase diﬀerence.
• Doppler eﬀect. This eﬀect is produced when the wavetransmitter and its re
ceiver are in relative movement. A change in phase is experienced with the
consequent frequency shift, which is known as the Doppler frequency (f
d
) and
can be approximated to the following expression:
f
d
=
v
λ
(2.2)
8
2.2. WCDMA principles
where λ is the wavelength and v is the relative speed between transmitter and
receiver. The degradation due to this eﬀect can be split in two: fast fading and
slow fading. When the coherence time of the channel (which is related with
the Doppler frequency within a multiplicative constant) is short compared to
the symbol duration, we say that the signal suﬀers from fast fading. In a fast
fading situation, the channel is expected to change several times while a symbol
is propagating causing a distorsion of the baseband pulse, resulting in a loss of
signaltonoise ratio (SNR) which may lead to an irreducible error rate. Slow
fading occurs when the coherence time of the channel is long enough compared to
the symbol duration. Then the channel is expected to remain unchanged during
the time in which a symbol is transmitted [10].
Largescale fading
Largescale fading aﬀects the average signal power, mainly caused by freespace path
loss and shadowing. It is well known that the power of an airtransmitted electro
magnetic wave proportionally decreases with the squared distance to the transmitter.
The received power expressed in terms of transmitted power is attenuated by a fac
tor L
s
(d), where this factor is called free space loss. When the receiving antenna is
isotropic this factor is expressed as:
L
s
(d) =
4πd
λ
2
(2.3)
This eﬀect is used in cellular systems because the rapid attenuation with distance
makes it feasible to reuse channels and thus, increase system capacity. Not only free
space propagation is aﬀected by path loss, other parameters such as the base station
or mobile antenna height or terrain characteristics are inﬂuencing too. The shadowing
concept involves all the unique characteristics of the scenario which can hamper the
communications and include for instance buildings and mountains. These environ
mental peculiarities increase the complexity of building up theoretical models [7, 8].
The following section presents the model used in the present study termed COST231
WalﬁshIkegami model which does not cover shadowing modeling.
COST231 WalﬁshIkegami model
This empirical model is a combination of the models from J. Walﬁsch and F. Ikegami.
It was further developed by the COST 231 project. It is now called Empirical COST
WalﬁschIkegami Model.
The model considers the buildings in the vertical plane between the transmitter and
the receiver. The accuracy of this empirical model is quite high because in urban
9
2. HSDPA Basics
environments especially the propagation over the rooftops (multiple diﬀractions) is
the most dominant part.
The main parameters of the model are listed in Table 2.1. Only the main equations
are explain below, please refer to [11] for further details.
The model distinguishes between lineofsight (LOS) and nonlineofsight (NLOS)
situations. In the LOS case, between base and mobile antennas within a street canyon,
a simple propagation loss formula diﬀerent from free space loss is applied. The loss is
based on measurements performed in the city of Stockholm
L
b
(dB) = 42.6 + 25 log(d km) + 20 log(f MHz) for d ≥ 20 m. (2.4)
In the NLOScase the basic transmission loss is composed of the terms free space loss
L
0
, multiple screen diﬀraction loss L
msd
, and rooftoptostreet diﬀraction and scatter
loss L
rts
.
L
b
=
L
0
+ L
rts
+ L
msd
for L
rts
+ L
msd
> 0
L
0
for L
rts
+ L
msd
≤ 0
(2.5)
The freespace loss is given by
L
0
(dB) = 32.4 + 20 log(d km) + 20 log(f MHz). (2.6)
The term L
rts
describes the coupling of the wave propagating along the multiplescreen
path into the street where the mobile station is located. The determination of L
rts
is
mainly based on Ikegami’s model. It takes into account the width of the street and its
orientation. COST 231, however, has applied another streetorientation function than
Ikegami.
L
rts
= −16.9 −10 log(wm) + 10 log(f MHz) + 20 log(∆h
Mobile
m) +L
Ori
. (2.7)
Scalar electromagnetic formulation of multiscreen diﬀraction results in an integral for
which Walﬁsch and Bertoni published an approximate solution in the case of base
station antenna located above the rooftops. This model is extended by COST 231
for base station antenna heights below the rooftop levels using an empirical func
tion based on measurements. The heights of buildings and their spatial separations
along the direct radio path are modelled by absorbing screens for the determination
of L
msd
.
L
msd
= L
bsh
+k
a
+k
d
log(d km) +k
f
log(f MHz) −9 log(b m) (2.8)
10
2.3. HSDPA principles
Parameters Restrictions
Frequency f 800  2000 MHz
Height of the transmitter h
TX
4  50 m
Height of the receiver h
RX
1  3 m
Distance d between transmitter and receiver 0.02  5 km
Table 2.1.: Parameters COST231 WalﬁshIkegami model
2.3. HSDPA principles
This section covers HSDPA principles for WCDMA  the key new features included
in Release 5, 6 and 7 speciﬁcations that are relevant for this study. HSDPA has
been designed to increase downlink packet data throughput of release 99 by means of
fast physical layer retransmission and transmission combining as fast link adaptation
controlled by the base station. First a comparison between Release 99 and HSDPA is
performed and then HSDPA key aspects are presented.
2.3.1. HSDPA vs Release 99 DCH
Three diﬀerent methods for data packet transmission are speciﬁed in Release 99:
Forward Access CHannel (FACH), Dedicated CHannel (DCH) and Downlink Shared
CHannel (DSCH). The last one has been replaced in Release 5 for the new HighSpeed
Downlink Shared CHannel (HSDSCH) and therefore will not be analyzed in this sec
tion. Table 2.2 shows the main diﬀerences between DCH and HSDSCH channels, and
it is explained in more detail below[1].
• FACH. This channel is used to transport small data volumes or connection set
ups during state transfers. In HSDPA, it is used to carry signalling information
when a terminal has changed its state. FACH does not support fast power
control or soft handover. The Secondary Common Control Physical CHannel
(SCCPCH) is the responsible to carry its content and the used spreading code
is ﬁxed. If there is a need to carry mixed services, FACH cannot be used.
• DCH. The key part of Release 99 and Release 5 is always operated together with
HSDPA. When circuitswitched service is demanded it runs always on DCH. In
Release 5 the uplink user data always goes through the DCH, but in Release 6
there is an alternative with the use of an enhanced version of DCH (EDCH).
DCH can carry any kind of service using a ﬁxed spreading code and ﬁxed allo
cation time, although these parameters can be changed from upper layers. The
theoretical maximum peak rate is 2 Mbps, in practice only 384 Kbps, and re
transmissions are handled in the RNC. It supports the use of soft handover,
11
2. HSDPA Basics
Feature DCH HSDSCH
Variable spreading factor Yes No
Fast power control Yes No
Adaptive modulation and coding No Yes
Multicode operation Yes Yes, extended
Physical layer retransmissions No Yes
BTSbased scheduling and link adaptation No Yes
Table 2.2.: Comparison DCH and HSDSCH [1]
meaning that an UE can be connected with more than one station at a time and
receiving information from all of them. The DCH also allows the use of the fast
power control feature.
2.3.2. Highspeed downlink shared channel
The logical transport channel, which carries the actual user data in HSDPA, is the
HSDSCH and is mapped physical channels named HSPDSCH. The key diﬀerences
between Release 99 DCHbased packet data operation, which are described in [1], are
as follows:
• Lack of closedloop power control or so called fast power control. In HSDPA,
link adaptation selects the suitable combination of codes, coding rates, and mod
ulation.
• Support of higher order modulation than DCH, the last release standardizes
64QAM, 16QAM and 4QAM for the downlink. In release 99 only 4QAM was
available.
• The scheduling is done on a 2ms basis, with the addition of fast physical layer
signaling. With DCH the Transmission Time Interval (TTI) could be as long
as 80ms and not shorter than 10ms. The use of the new short allocation period
implicitly implies a more dynamic nature of the system.
• Lack of soft handover. Data are sent from only one serving HSDSCH cell.
• Lack of physical layer control information on the HSPDSCH. This is carried on
the Highspeed shared control channel (HSSCCH) for HSDPA use and on the
associated DCH.
• Only spreading factor 16 is used, therefore the UEs will be able to support up
to 15 codes because common channels and associated with DCHs need one of
them. The support of multiple channelisation codes is called codemultiplex.
12
2.3. HSDPA principles
• Only turbocoding is used, with DCH also convolutional code could be used.
• Implements HybridARQ (HARQ) which can operate in two modes: ’Chase
combining’ and ’incremental redundancy’.
HSDSCH coding
The use of turbocoding outperforms convolutional codes, therefore HSDPA presents
a restriction on the use of convolutional codes and just turbocoding will be used from
now on.
There are some changes introduced in the channel coding chain due to the use of
new modulation schemes. Also a bit scrambling functionality is introduced to avoid
having long sequences of ’1s’ or ’0s’, to ensure good signal properties for demodula
tion.
The HARQ functionality consists of twostage rate matching functionality which allows
tuning the redundancy version of diﬀerent retransmissions when using nonidentical
retransmissions. HARQ can operate in two modes, ’Chase combining’ or ’incremental
redundancy’ [1].
• In Chase combining, the rate matching is identical between transmissions so
the same bit sequence is sent. The receiver stores the received samples as soft
values, and therefore the memory consumption is higher than if it was storing
hard values.
• Incremental redundancy uses a diﬀerent rate matching between retransmissions.
The relative number of parity bits to systematic bits varies between retransmis
sions. This solution requires more memory in the receiver. The rate matching
function is varied between diﬀerent retransmissions and in the actual implemen
tation channel encoding can be done for each transmission or the data can then
be kept in the virtual buﬀer.
When the physical retransmissions fail, these are handled by the RNC like in Release
99.
HSDSCH link adaptation
As covered in the section 2.3.2 the use of 2ms TTIs allows the system a great dynamic.
Apart from the scheduling decisions, the base station will also decide every 2 ms which
coding and modulation combination to transmit.
13
2. HSDPA Basics
4QAM 16QAM 64QAM
Figure 2.4.: Modulation constellations
Link adaptation is based on physical Channel Quality Indicator (CQI), which is the
feedback provided by the UE. To avoid the nearfar problem, link adaptation takes the
extra power that results from being too close to the base station and uses it to select
the transmission parameters in such way that the required symbol energy corresponds
more accurately to the available symbol power. The dynamic range obtained using
this technique can reach 30dB according to [1].
HSDSCH modulation
The transport channel associated with R99, DCH, uses only 4QAM modulation. Re
lease 5 and 6 oﬀer additional support for higher modulation order on the downlink:
16QAM. Moreover Release 7, the so called HSPA+, introduces 64QAM on the down
link which increases the data rates by 50% and 16QAM on the uplink. The constel
lations are shown in Figure 2.4. The higher order modulation the higher the num
ber of bits that can be carried per symbol. But higher order modulation introduces
more complicated decision boundaries, and therefore the signal quality needs to be
better when for example 16QAM is used instead of 4QAM. A goodquality CPICH
allows the estimation of the optimum channel without userspeciﬁc pilot overhead
[1, 12].
2.3.3. Multiple Input Multiple Output HSDPA
MIMO systems enable an increase of the throughput without having to increase the
bandwidth nor the transmitted power. The same carrier frequency is used for all
the transmitted antennas. These as well as the receiver antennas are usually uni
formly separated by distances close to the size of a wavelength. MIMO HSDPA was
introduced in Release 7 within what is called HSPA+ or HSPA evolved [6]. HSPA+
14
2.3. HSDPA principles
supports 2x2 downlink MIMO with up to two antennas at the transmitter and the
receiver.
The standardized MIMO transmission scheme in Release 7 is termed Adaptative An
tenna Array (see Figure 2.5). There are two modes deﬁned: TxAA, in this case one
stream is transmitted over both antennas, and DTxAA or dualstream TxAA, for this
case, two separate data streams are transmitted on two orthogonal weight sets simulat
neously. The weights w
(k)
t
are selected to maximize the signaltonoiseandinterference
ratio (SINR) at the UE. The weight selection is signaled by the UE to the base sta
tion. Note that when DTxAA is used both data streams are transmitted on the same
orthogonal spreading code(s). Thus, this achieves code reuse.
DTxAA transmits the data streams using orthonormal array weight vectors drawn
from the closed loop transmit diversity codebook. Because they are orthogonal, knowl
edge of one of the two array weight vectors will completely determine the other. In ad
dition, it should be noted that in the MIMO case the second data stream is only turned
on at high SINR conditions. In other words, either one data stream or two data streams
are transmitted depending on the terminal SINR conditions.
MIMO performs most eﬀectively when the SNR at the UE is high, ensuring a successful
decoding of the signal in spite of distributing the power among the transmit antennas.
MIMO also needs a rich scattering environment to keep the two data stream orthogonal
when they reach the UE, so that there are enough uncorrelated paths, i.e. the channel
matrix has to be of suﬃcient rank. Therefore MIMO beneﬁts can be seen in dense
urban areas where the size of the cells is typically small. Higher order modulation
complements MIMO by providing signiﬁcant gains in lineofsight scenarios, where
MIMO gains are limited.
The key beneﬁts of MIMO systems compared to SISO systems summarized in [13]
are:
• Antenna grouping gain. The processing in the transmitter and the receiver in
creases the mean SNR received.
• Diversity gain. The signal power in a wireless communication suﬀers from ran
dom fading. If the fading in each of the MIMO channels are uncorrelated, then
better results will be obtained in detection. Thus, the more uncorrelated paths,
the better the system will work. A diversity gain is obtained compared to SISO
systems when there is just one path.
• Spatial multiplexing gain. MIMO channels provide an increase of the system
capacity without the need to increase the bandwidth nor the transmitted power.
This gain can also be enhanced by the capacity to transmit independent data
signals in each of the transmission antennas.
15
2. HSDPA Basics
w
4
w
1
w
2
TrCH
processing
HSDSCH
primary
transport block
+
+
CPICH
1
spread/scramble
w
3
TrCH
processing
HSDSCH
secondary
transport block
+
+
CPICH
2
Figure 2.5.: Generic downlink transmitter structure MIMO DTxAA [2]
2.3.4. HSDPA physical layer operation procedure
This section presents the HSDPA physical layer basic steps once at least one user have
been conﬁgured to use HSDSCH and the data is already present at the buﬀer of the
base station [1].
• Every 2ms the base station evaluates each user in order to schedule which users
will be served in the future. The selection criteria is not set in the standard, so
schedulers are an ongoing research topic at the moment.
• When a UE is selected as served in a determined TTI, the base station identiﬁes
the HSDSCH parameters needed for the transmission, including the number of
codes, the modulation order and the UE limitations.
• HSSCCH is transmitted two slots before the corresponding HSDSCH TTI. This
is because it has two diﬀerent parts and Part 1 carries information needed to
decode the frame HSDSCH properly.
• The terminal monitors the HSSCCHs (there are up to four). Once the Part 1
of HSSCCH is decoded it will start to decode the Part 2 and will buﬀer the
necessary codes from HSDSCH.
• When decoding Part 2 the UE discovers the ARQ process, and can then deter
mine whether there is the need of combining or not.
• After decoding, the UE sends in the uplink direction an ACK/NACK indicator
after the combination of the data (if applied), depending on the result of the
CRC over the HSDSCH data.
16
3. Basic MIMO HSDPA LinkLevel
Simulator
This chapter presents the description of the implemented simulator used in this study
to model a basic MIMO HSDPA link. The outline is as follows, Section 3.1 contains
the general description of the simulation model. This is followed by a more detailed
study of each component in Section 3.2. Finally, implementation issues are discussed
in Section 3.3.
3.1. General structure
The general structure of our basic link level simulator is depicted in Figure 3.1. Basi
cally, like in any communication system, there are three diﬀerent parts: the transmit
ter, which in our case is the base station due to the study of the downlink (HSDPA);
the channel, a MIMO frequencyselective channel plus Gaussian white noise; and ﬁ
nally the receiver or user terminal. All these elements are explained in detail in Section
3.2.
The Basic MIMO HSDPA linklevel simulator emulates the physical transmission of
the data channel (HSDSCH) and the pilot channel (CPICH), the last one is used to
estimate the channel for the data channel.
The bit sequences in the simulator ﬂow encapsulated in packets of 2 ms, as cited
before in Section 2.3.2 this is the TTI in HSDPA systems. For sake of simplicity and
convenience the channel is constant during the duration of a TTI, i.e. block fading is
utilized. The length of a radio frame is 10 ms and is composed of 15 slots, thus a TTI
is formed by 3 slots.
3.2. Block Description
The dotted blocks in Figure 3.1 are presented in detail in Section 3.2.1, covering the
generation of the CPICH and the HSDSCH. In Section 3.2.2 we discuss the transmitter
schemes supported in the simulator and introduce the concept of beamforming. The
17
3. Basic MIMO HSDPA LinkLevel Simulator
Figure 3.1.: Basic linklevel simulator block diagram
channel model is covered in Section 3.2.3. This is followed by the receiver analysis in
3.2.4, emphasizing the new MMSE receiver preferred in HSDPA, as commented in [14].
Finally the schemes of channel estimators implemented in the simulator are discussed
in Section 3.2.5.
3.2.1. Common Pilot Channel & Highspeed Downlink Shared Channel
Before describing how the channels are created it is needed to understand the re
lation between the modulation, the spreading and the scrambling of the sequences.
Figure 3.2 shows the relation between each other for a HSDSCH stream taking into
account that more than one bit stream is transmitted per stream. As can be seen,
encoded ratematched bits, b
(k)
n
[m], go through the modulation mapper in groups
of the length according to the modulation selected. There are three modulation
schemes supported in enhanced HSDPA [12]. Our simulation has support for all of
them.
Once bits come out of the modulation mapper in shape of symbols, these are multi
plied by the corresponding channelisation code, C
SF,c
, spreading the signal, and there
fore increasing the symbol rate by the length of the channelisation code or spreading
factor. In addition to spreading, part of the process in the transmitter is the scram
bling operation. This is needed to separate UEs or base stations from each other.
Scrambling is used on top of spreading, so it does not change the signal bandwidth
but only makes the signals from diﬀerent sources separable from each other [9]. The
functionality and characteristics of the scrambling and channelisation codes for the
HSDSCH are summarised in Table 3.1. To understand how the channelisation and
scrambling codes, C
scr
, are generated, refer to Appendix A.1 and Appendix A.2, re
spectively.
18
3.2. Block Description
Channelisation code Scrambling code
Usage Separation of connections to
diﬀerent users within one cell
Separation of sectors (cells)
Length 16 chips 2 ms = 7680 chips
Code family Orthogonal Variable Spreading
Factor
Long 10 ms code: Gold code
Short code: Extended S(2) code
family
Spreading Yes, increases transmission
bandwidth
No, does not aﬀect transmission
bandwidth
Table 3.1.: Functionality of the channelisation and scrambling codes [9]
4QAM, 16QAM, 64QAM
b
1
[m]
(k)
4QAM, 16QAM, 64QAM
4QAM, 16QAM, 64QAM
C
16,2
C
16,N1
C
scr
b
2
[m]
(k)
b
SF1
[m]
(k)
s
sp
[i]
(k)
s
[i]
(k)
[i]
C
16,1
s
y
[n]
(k)
1
s
y
[n]
(k)
2
s
y
[n]
(k)
SF1
Figure 3.2.: Relationship between modulation, spreading and scrambling [3]
CPICH
Figure 3.3 shows the structure of the CPICH as well as the bit sequence assigned
to this channel. In case of MIMO transmission, the CPICH has to be transmitted
from both antennas using the same spreading factor and scrambling code even though
the predeﬁned bit sequence of the CPICH is diﬀerent for Antenna 1 and Antenna
2. When no diversity is used, Antenna 1 transmits just the sequence assigned to
it.
Although there are two types of CPICH, only the primary (PCPICH) is implemented
in the simulator, the implementation of the secondary (SCPICH)has no interest in this
study. Normally each cell has only the PCPICH, which is broadcasted to the entire
cell. Occasionally the SCPICH can be found implemented for serving dedicated hot
spot areas. This study focus on the optimization of the PCPICH power, and therefore
the term CPICH will be used for PCPICH in the following.
19
3. Basic MIMO HSDPA LinkLevel Simulator
0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000
0011 1100 0011 1100 0011 1100 0011 1100 0011 1100 0011 1100 0011 1100 0011
Slot 2 Slot 1 Slot 15
Transmitter Antenna 2
Transmitter Antenna 1
T = 2560 chips, 20 bits slot
1 radio frame: T = 10 ms
f
Figure 3.3.: Frame structure for Common Pilot Channel
The main features of the CPICH are the following [3]:
• The same channelisation code is always used and has length 256. Particularly the
Orthogonal Variable Spreading Factor (OVSF)used is the C
256,0
. (See Appendix
A.1)
• The CPICH is scrambled by the primary scrambling code.
• The modulation scheme used is always 4QAM.
• There is only one CPICH per cell.
• CPICH is broadcast over the whole cell.
• The bit rate is 30 kbps, as a result of the use of spreading factor of 256, and
4QAM modulation it implies 20 bits per slot, and 2560 chips per slot. See Figure
3.3.
HSDSCH
For the data channel, the bit sequence is generated randomly with a length according to
the spreading factor and modulation selected so that the number of chips transmitted
during a TTI keeps constant. The equation (3.1) shows this relation
N
r
=
2560 ×M
SF
(3.1)
where N
r
denotes the number of bits to be generated, M is the modulation order, SF
is the spreading factor used. Using the chip rate (3.84 Mcps) and the TTI (2 ms), the
number of transmitted chips per TTI can be easily derived and the result is 7680 chips.
20
3.2. Block Description
s
[i]
(1)
p [i]
(1)
v [i]
(1)
h
(1,1)
r
[i]
(1)
(a) SISO
s
[i]
(1)
w
1
(1)
w
2
(1)
p [i]
(1)
p [i]
(2)
v [i]
(1)
h
(1,1)
h
(1,2)
r
[i]
(1)
(b) MISO
s
[i]
(1)
w
1
(1)
w
2
(1)
p [i]
(1)
s
[i]
(2)
w
1
(2)
w
2
(2)
p [i]
(2)
v [i]
(1)
v [i]
(2)
h
(1,1)
h
(2,2)
h
(2,1)
h
(1,2)
dual streammode
r
[i]
(1)
r
[i]
(1)
(c) MIMO, TxAA and DTxAA
Figure 3.4.: Transmission schemes
A packet contains 3 slots, then the number of chips transmitted per TTI is equal to
2560 chips, here the origin of the constant.
The main characteristics of the HSDSCH channel as implemented in the simulator
are:
• The channelisation code has always length 16. There are 15 channelisation codes
available from up to 16, one of them is reserved for the CPICH.
• The supported modulation schemes are 4QAM, 16QAM and 64QAM. The last
one is standardized for the non MIMO case only.
• The bit rate is variable depending on the modulation.
3.2.2. Transmission modes
The simulator can cope with up to four diﬀerent transmissions schemes (Figure 3.4):
SISO, MISO 2x1, MIMO 2x2 TxAA, and MIMO 2x2 DTxAA. Singleinput single
output SISO(Figure 3.4a) and multipleinput and singleoutput MISO (Figure 3.4b)
are degenerate cases of MIMO (Figure 3.4c), when respectively either only one trans
mitter and one receiver antenna or two transmitter and one receiver antenna are
used. Release 7 includes the speciﬁcation for the MIMO case limited to the two an
tenna scheme. Detailed information concerning MIMO HSDPA is presented in Section
2.3.3.
The standardized precoding vectors, which are used in multiple antenna cases, are
deﬁned as follows:
21
3. Basic MIMO HSDPA LinkLevel Simulator
w
(1)
1
= w
(2)
1
= 1/
√
2
w
(1)
2
= −w
(2)
2
w
(1)
2
∈
1 +j
2
,
1 −j
2
,
−1 +j
2
,
−1 −j
2
Figure 3.4 also reﬂects the fact that no precoding is applied to the CPICH, denoted
by p
(t)
.
The spreading factor (SF) used for the data channel is SF = 16, one of them is
reserved for CPICH. For this reason, each data stream has a capacity of 15 bit streams.
Each bit stream is denoted in the Figure 3.2 as b
(k)
n
with n ∈ 1...SF −1 and k is the
stream index. Hence, as many as 30 diﬀerent bit streams can be transmitted for a
particular user, when DTxAA is the selected mode.
The number of active streams in the case of our simulator is handled manually. Even
though in general, a variable amount of power could be allocated to each stream in
order to maximize capacity, this is not implemented in our simulator, because the
extra complexity did not have substantial beneﬁts for the purpose of this study, and
therefore the power allocated to the data channel will be split equally among the
number of streams.
Selection of the best beamforming
The selection of the best beamforming is done according to [15]. The precoding matrix
is selected in the transmitter in concordance with a past channel determined by a delay,
d
bf
. This is done because in a real system the best beamforming is selected in the UE
and signaled to the base station which implies a certain delay.
The frequencyselective channel H
n
is modeled as an array of length L×N
T
associated
to the receiver antenna n. In order to maximize the received power, and therefore
choose the best beamforming we deﬁne R :=
¸
N
R
n=1
H
H
n
H
n
. Then the problem of
choosing the weights W, which is the matrix containing the precoding vectors, so that
maximize the received power becomes
argmax
W:W
2
=1
W
H
RW (3.2)
The optimal weight vector is the dominant eigenvector of R. We always optimize over
the ﬁrst stream in the DTxAA case.
22
3.2. Block Description
−10 −5 0 5 10 15 20 25 30
−20
−15
−10
−5
0
5
10
15
20
25
30
noise power [dBW]
S
I
N
R
H
S
−
D
S
C
H
[
d
B
]
SISO
MISO 2x1
MIMO 2x2 TxAA
stream 1 MIMO 2x2 D−TxAA
stream 2 MIMO 2x2 D−TxAA
Figure 3.5.: Transmission scheme comparison
Transmission schemes comparison
The HSDSCH SINR is increased by the use of MIMO just as expected. This is
depicted in Figure 3.5. The SINR in the case of MIMO DTxAA is evaluated per
stream. Note that in these simulations full knowledge of the channel is assumed in the
equalizer.
The HSDSCH SINR is the SINR observed at the demapper input, so after despreading
and is obtained as follows: Consider the transmitted data symbol vector s
(k)
y
n
of the
kth stream and n spreading code (see Figure 3.2) and the corresponding received
symbol vector at the demapper input ˆs
(k)
y
n
. The post equalization SINR then is given
by
SINR
k,n
=
s
(k)
y
n
γ
(k)
2
2
s
(k)
y
n
γ
(k)
−ˆs
(k)
y
n
2
2
(3.3)
where γ
(k)
represents the attenuation of the overall impulse response at delay τ and
is given by
γ
(k)
=
f
(k)
ˆ
H
ˆ
H
(τ+(k−1)(L
f
+L
h
−1))
w
2
2
. (3.4)
where f
(k)
are the equalization ﬁlters for each kth stream and
ˆ
H
j
w
denotes the column
j of the matrix
ˆ
H
w
.
23
3. Basic MIMO HSDPA LinkLevel Simulator
3.2.3. Channel model
The channel presents two types of fading eﬀects: largescale and smallscale fading,
these eﬀects have been presented in Section 2.2.3. In our simulator we deal with both
of them. The interference power in the cell is modeled as a white Gausian noise and
the channel is normalized to 1.
Smallscale fading
Smallscale fading are produced by changing channel conditions. The received signal is
made up of multiple reﬂective rays, therefore the envelope amplitude due to smallscale
fading has a Rayleigh probability density function [10] given by,
p(r) =
r
σ
2
exp
¸
−
r
2
2σ
2
r ≥ 0
0 otherwise
(3.5)
where r is the envelope amplitude of the received signal, and 2σ
2
is the mean power
of the multipath signal. Because of this sometimes the smallscale fading is also called
Rayleigh fading. These interferences are generated as described in [16].
The channel will be fast fading or slow fading depending on the doppler frequency,
f
d
. As we have discussed before our simulator uses block fading, i.e. the channel is
kept constant during the transmission of a packet (7680 chips or 2 ms). This limits
our simulator so that the coherence time of the channel, T
o
, must be larger than
2 ms. If the relative speed between the base station and the UE is v = 3km/h then
T
o
≈ 32 ms but when the speed turns to be v = 50 km/h then T
o
≈ 2 ms. This
means that our simulator is limited to deal with relative speeds lower than 50 km/h.
The attenuation introduced by the rayleigh fading for diﬀerent velocities can be seen
in 3.6.
In order to deﬁne the diﬀerent number of propagation paths, we use the ITU channel
model Pedestrian B [17], as speciﬁed in the standard [18]. Figure 3.7 shows the
Power Delay Proﬁle (PDP) of Pedestrian B model, which determines the length of
the channel (L
h
). The maximum excess delay time, T
m
, is also determined by the
PDP. In this case, T
m
= 3700ns. We can determine the duration of a chip, T
c
,
from the chip rate R = 3.84Mcps, which leads to a T
c
= 260.41ns. Therefore the
channel is frequency selective (T
m
> T
c
) and such multipath dispersion yields ISI. The
equalization described in the next section mitigates these eﬀects.
24
3.2. Block Description
0 200 400 600 800 1000 1200 1400 1600 1800 2000
10
−3
10
−2
10
−1
10
0
10
1
Rayleigh fading v=3km/h
Samples
S
ig
n
a
l
L
e
v
e
l
A
t
t
e
n
u
a
t
io
n
(
d
B
)
0 200 400 600 800 1000 1200 1400 1600 1800 2000
10
−3
10
−2
10
−1
10
0
10
1
Samples
S
ig
n
a
l
L
e
v
e
l
A
t
t
e
n
u
a
t
io
n
(
d
B
)
Rayleigh fading v=10km/h
0 200 400 600 800 1000 1200 1400 1600 1800 2000
10
−3
10
−2
10
−1
10
0
10
1
Samples
S
ig
n
a
l
L
e
v
e
l
A
t
t
e
n
u
a
t
io
n
(
d
B
)
Rayleigh fading v=120km/h
0 200 400 600 800 1000 1200 1400 1600 1800 2000
10
−3
10
−2
10
−1
10
0
10
1
Rayleigh fading v=30km/h
Samples
S
ig
n
a
l
L
e
v
e
l
A
t
t
e
n
u
a
t
io
n
(
d
B
)
Figure 3.6.: Rayleigh fading for diﬀerent speed
3.2.4. Equalization
RAKE is the conventional receiver for WCDMA which approximately implements a
matched ﬁlter for the channel impulse response. The receive signal r
i
is ﬁrst de
layed and then descrambled, despread and combined to Maximum Ratio Combining
(MRC), and the coeﬃcients of the tap delay line are found through channel estima
tion.
HSDPA uses WCDMA for multiuser communication and thus orthogonal spreading
codes are used to separate diﬀerent users in the downlink. However, the orthogonality
of these codes is destroyed by the multipath characteristics of the channel resulting
in Multiple Access Interference (MAI). The performance of the MMSE equalizer is
compared to the RAKE receiver in [14] proving that signiﬁcant performance gain is
obtained with the use of equalizers and that when these are used the overall system
is no longer interference limited.
The MMSE equalizer implemented in our simulator is based on the papers [19, 20].
25
3. Basic MIMO HSDPA LinkLevel Simulator
0 500 1000 1500 2000 2500 3000 3500 4000
−25
−20
−15
−10
−5
0
ns
d
B
Power Delay Profile − Pedestrian B
Figure 3.7.: PDP Pedestrian B
System model
In order to derive an MMSE equalizer we present the system model as derived in [19].
The kth spread and scrambled chip stream at time instant i is deﬁned as
s
(k)
i
=
¸
s
(k)
[i], . . . , s
(k)
[i −L
h
−L
f
+ 2]
T
, (3.6)
where L
h
and L
f
are the length of the channel impulse response and the equalizer
length, respectively. The chip streams are weighted by the complex precoding coeﬃ
cients w
(k)
1
and w
(k)
2
where the subindex indicates the transmit antenna. Afterwards
the pilot sequences p
(1)
i
and p
(2)
i
are added at the resulting sequences in each an
tenna.
By stacking the transmitted stream vectors which can be up to K = 2, in case of
DTxAA and the CPICH vectors which depend on the number of transmit anten
nas,
s
i
=
¸
s
(1)
i
T
, . . . ,
s
(K)
i
T
T
, (3.7)
p
i
=
¸
p
(1)
i
T
, . . . ,
p
(N
T
)
i
T
T
. (3.8)
The frequency selective channel between the n
t
th transmit and the n
r
th receive an
tenna is modeled by the L
f
×(L
h
+L
f
−1) dimensional band matrix,
26
3.2. Block Description
H
(n
r
,n
t
)
=
h
(n
r
,n
t
)
0
. . . h
(n
r
,n
t
)
L
h
−1
0
.
.
.
.
.
.
0 h
(n
r
,n
t
)
0
. . . h
(n
r
,n
t
)
L
h
−1
¸
¸
¸
¸
, (3.9)
where h
(n
r
,n
t
)
i
represent the channel impulse response of the n
t
th transmit antenna to
the n
r
th receive antenna. Then, the full frequency selective MIMO channel is modeled
by a block matrix Hconsisting of N
R
×N
T
band matrices deﬁned in (3.9)
H =
H
(1,1)
H
(1,N
T
)
.
.
.
.
.
.
H
(N
R
,1)
H
(N
R
,N
T
)
¸
¸
¸
¸
. (3.10)
By stacking the received signal vectors of all N
R
receive antennas we have
r
i
=
¸
r
(1)
i
T
, . . . ,
r
(N
R
)
i
T
. (3.11)
In addition to the channel eﬀects a noise v
i
term is added representing the ther
mal noise in the receiver and the interference power received from other base sta
tions. This is modeled as an additive white gaussian noise (AWGN) with variance
σ
2
v
We can obtain a compact system description of received symbols
r
i
= H(W⊗I
L
h
+L
f
−1
)s
i
+Hp
i
+v
i
= H
w
s
i
+Hp
i
+v
i
. (3.12)
Here, ⊗ denotes the Kronecker product. The matrix W is deﬁned in (3.13) containing
the precoding values of each stream (up to K=2). How the selection of this precoding
matrix is done is explained in Section 3.2.2.
W =
¸
w
(1)
1
. . . w
(K)
1
w
(1)
2
. . . w
(K)
2
¸
(3.13)
27
3. Basic MIMO HSDPA LinkLevel Simulator
v [i]
(1)
r
[i]
(1)
f
r
(n ,1)H
f
r
(n ,2)H
dual stream mode
ŝ
[i]
(1)
ŝ
[i]
(2)
Figure 3.8.: MMSE equalizer structure
STMMSE equalizer
The Spacetime minimum mean square error (STMMSE) equalizer coeﬃcients can be
calculated by minimizing the distance between the equalizer chip stream and the trans
mitted chip stream through the following quadratic cost function [20]
J
f
(k)
= E
f
(k)
H
ˆr
i
−s
(1)
i−τ
2
¸
, (3.14)
where
ˆr
i
= H
w
s
i
+v
i
, (3.15)
and
f
(k)
=
¸
f
(1,k)
T
, . . . ,
f
(N
R
,k)
T
T
, (3.16)
that deﬁnes N
R
equalization ﬁlters for each kth stream, see Figure 3.8. Each ﬁlter is
deﬁned by
f
(n
r
,k)
=
¸
f
(n
r
,k)
0
, . . . , f
(n
r
,k)
L
f−1
T
. (3.17)
The minimization of the cost function is performed by deriving (3.14) with respect to
f
(k)
∗
obtaining [20]:
f
(k)
= σ
2
s
H
w
R
ss
H
H
w
+R
vv
−1
H
w
e
τ
k
,2(L
h
+L
f
−1)
. (3.18)
Here, the vector e
τ
k
,2(L
h
+L
f
−1)
is a zero vector of length 2(L
h
+ L
f
−1) with a single
one at position
τ
k
= τ + (k −1)(L
h
+L
f
−1) k = 1 . . . K, (3.19)
28
3.2. Block Description
0 5 10 15 20 25 30 35 40 45
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Figure 3.9.: Overall impulse response
The variable τ speciﬁes the delay of the equalized signal and due to causality must fulﬁll
the condition: τ ≥ L
h
. The matrices R
ss
and R
vv
are the signal and noise correlation
matrices, respectively. Note that because we assume that the noise vector v
i
is AWGN
with variance σ
2
v
and mean µ = 0, the covariance matrix and the correlation matrix
are the same and we can write R
vv
= σ
2
v
I. We also assume that the data signals of
the users are uncorrelated and therefore R
ss
= σ
2
s
I where σ
2
s
is the energy allocated
to kth stream.
Figure 3.9 shows an exemplary overall impulse response of the equalizer and channel,
which is described in Section 3.2.3. It represents the ideal case when the noise term
is assumed to be zero. The length of the impulse response is the sum of the length of
the channel in this case is 18 chips because of the use of Pedestrian B, and the span
of the equalizer which is 30 chips minus 1 chip, thus 47 chips.
3.2.5. Channel Estimation
Our simulator utilizes two channel estimators: Least Squares Estimator and the
Correlationbased Estimator. Both of them are based on [21, 22].
System Model
In this case the spread pilot chip sequence p
(n
t
)
i
and the spread data sequence s
(n
k
)
i
are deﬁned as follows
p
(n
t
)
sp
i
=
¸
p
(n
t
)
sp
[i] . . . , p
(n
t
)
sp
[i +N
c
−1]
T
, (3.20)
29
3. Basic MIMO HSDPA LinkLevel Simulator
s
(n
k
)
sp
i
=
¸
s
(n
k
)
sp
[i] . . . , s
(n
k
)
sp
[i +N
c
−1]
T
, (3.21)
where i is the time index in chips and N
c
is the number of chips considered for the
channel estimation. In our case N
c
will have the length of a packet, and therefore 7680
chips. The precoding of the data sequence is omitted by sake of simplicity and without
lost of generality, note that the precoding could be done before the scrambling without
alteration of the ﬁnal transmitted sequence. Consequently, the scrambled transmit
signal at antenna n
t
is given by
x
(n
t
)
i
= C
scr
i
p
(n
t
)
sp
i
+s
(n
t
)
sp
i
n
t
= 1 . . . N
T
(3.22)
with the matrix C
scr
i
= diag[c
scr
i
, . . . , c
scr
i+N
c
−1
] comprising the N
c
chips of the base
station’s scrambling sequence at the main diagonal. The combined transmit signal for
all N
T
is modeled like this,
P
sp
i
=
¸
p
(1)
sp
i
, . . . , p
(N
T
)
sp
i
, (3.23)
S
sp
i
=
¸
s
(1)
sp
i
, . . . , s
(N
T
)
sp
i
, (3.24)
X
i
=
¸
x
(1)
i
, . . . , x
(N
T
)
i
= C
scr
i
(P
sp
i
+S
sp
i
). (3.25)
The deﬁnition of the channel in this section has been adapted according to [21]. The
MIMO channel at the delay k (k = 0, . . . , L
h
−1) is deﬁned as
H
k
=
h
(1,1)
k
. . . h
(1,N
R
)
k
.
.
.
.
.
.
.
.
.
h
(N
T
,1)
k
. . . h
(N
R
,N
T
)
k
¸
¸
¸
¸
(3.26)
we can deﬁne a stack matrix of the channel matrices like this
H =
¸
H
T
0
. . . , H
T
L
h
−1
T
. (3.27)
Once we have deﬁned our channel and the signal transmit in each antenna, we can
express the receive signal of all receive antennas as
30
3.2. Block Description
R
i
=
¸
r
(1)
i
, . . . , r
(N
R
)
i
=
L
h
−1
¸
k=0
C
scr
i−k
(P
sp
i−k
+S
sp
i−k
)H
k
+N
i
(3.28)
The matrix N
i
includes the interference of other base stations as well as the thermal
noise at the receiver.
Least Squares Estimator
Despite of being a low complexity channel estimator the Least Squares (LS) channel
estimator has a reasonable good performance. From the model presented we can
rewrite the equation (3.28) as
R
i
=
˜
P
i
H+
˜
N
i
(3.29)
where
˜
P
i
= [C
scr
i
P
sp
i
, C
scr
i−1
P
sp
i−1
, . . . , C
scr
i−L
h
+1
P
sp
i−L
h
+1
] (3.30)
and
˜
N
i
=
L
h
−1
¸
k=0
C
scr
i−k
S
sp
i−k
H
k
+N
i
(3.31)
Then, the LS estimator for the system description (3.29) is given by
ˆ
H
(LS)
=
˜
P
H
i
˜
P
i
−1
˜
P
H
i
R
i
(3.32)
Correlationbased Estimator
The correlator estimator is a simpliﬁed case of the LS estimator. It tries to avoid
the calculation of the matrix
˜
P
H
i
˜
P
i
−1
. The approximation and therefore the
correlationbased estimator is given by
ˆ
H
(cor)
=
1
p
2
2
˜
P
H
i
R
i
(3.33)
Where the term p
2
2
denotes the energy of the CPICH sequence transmitted at one
transmit antenna.
31
3. Basic MIMO HSDPA LinkLevel Simulator
0 2 4 6 8 10 12 14 16 18 20
1
2
3
4
5
6
7
8
CPICH E
c
/I
or
[dB]
M
S
E
[
d
B
]
MSE H (LS)
MSE H (cor)
Figure 3.10.: Comparison of channel estimators
Comparison of channel estimators
Correlationbased estimator is a simpliﬁed version of the LS estimator, therefore a
similar result is expected. The mean square error (MSE) of both channel estimators
for a SISO case and a noise power of 4dBW is depicted in ﬁgure 3.10, they look very
much the same over a large range of values, but it can also be observed that the larger
the power allocated to the common pilot channel the larger the diﬀerence between the
channel estimator MSE.
3.3. Implementation issues
The Basic MIMO HSDPA linklevel simulator is implemented in Matlab. The tools
provided by Matlab are very powerful and useful for the purpose of this study,
because some of the functions such as convolution and ﬁltering are already imple
mented.
The main ﬁle structure of the simulator is depicted in Figure 3.11 . The structure of
the simulator is contained in hsdpa_link_body ﬁle. E_HSDPA_link level_simulator
ﬁle contains the diﬀerent simulations which has been used in this study, it oﬀers the
possibility of being upgraded with more simulations. Each simulation ﬁle simu_xxx
has associated a load_parameters_xxx ﬁle, which has the parameters that can be set
by the user for the chosen simulation.
32
3.3. Implementation issues
E_HSDPA_link_level_simulator.m
hsdpa_link_body
load_parameters_xxx.m
load_parameters_yyy.m
load_parameters_zzz.m
simu_xxx.m
simu_yyy.m
simu_zzz.m
Figure 3.11.: Basic MIMO HSDPA linklevel simulator ﬁle structure
33
4. Mobile Network Simulations
This chapter presents the importance of systemlevel simulators and computational
eﬃcient modeling. First, the diﬀerences between linklevel and systemlevel simula
tors are covered in Section 4.1. In Section 4.2 we present a computationally eﬃcient
MIMO HSDPA systemlevel simulator showing its main structure and focusing in the
linkmeasurement model. Finally in Section 4.3, we propose a model improvement
to take into consideration the eﬀects of channel estimation in the linkmeasurement
model.
4.1. SystemLevel Simulations vs. LinkLevel Simulations
Linklevel simulators are commonly used to study the behavior of transmission and
reception schemes. As the name indicates, this kind of simulators just study one link, in
this case we are talking about the link between one UE and one base station. Linklevel
simulators aim to study the physical characteristics of the link and sometimes some
MAC functionalities such as development of receiver algorithms, feedback strategies,
coding design, and so on.
Mobile Network Operators (MNO) are aware that an optimization of their networks
increases the performance and reduces costs. For this reason, it is important to identify
whether, and to which amount, predicted link level performance gains can be obtained
in an entire network. Hence, a comprehensive study of a mobile network technology
can not stick just to the way how one UE performs with the base station. The goal
of systemlevel simulators is to evaluate the performance of a whole network (or part
of it), where multiple users, multiple cells and therefore base stations are taken into
consideration. Cell planning, scheduling and multiuser and multi base station inter
ference are some of the investigations that systemlevel simulators try to cover. This
type of simulators have to rely on simpliﬁed link models that still must be accurate
enough to capture the essential behavior due to complexity reasons because the com
putational cost of evaluating a whole network with the use of linklevel simulators is
prohibitive[23].
In this study the implemented basic MIMO HSDPA linklevel simulator (see Chapter
3) is used to enhance the characteristics of the MIMO HSDPA systemlevel simulator
in [23, 5].
35
4. Mobile Network Simulations
channel characteristics
(pathloss, shadowing,
fast fading)
intercellinterference
characteristics
link
measurement
model
feedback
strategy
link
performance
model
pilot signaling
PHY processing/
power allocation
resource
allocation
link
adaption
strategy
resource
scheduling
strategy
QoS
requirements
PER
figures of merit,
e.g. cell throughput
channel quality
information
traffic model
Figure 4.1.: Schematic block diagram of system level simulations [4]
4.2. MIMO HSDPA System Level Simulator
Figure 4.1 depicts a schematic diagram of a basic dynamic systemlevel simulator.
Generally two kind of models are required: a link measurement model, which models
the measurement used for link adaptation and resource allocation, and a link per
formance model, which determines the BLock Error Ratio (BLER) given a certain
resource and power allocation as well as signal processing. Both models are related
in the sense that they provide ﬁgures for performance prediction and can be referred
to as system level interface [23]. The following sections present the main structure
of the MIMO HSDPA Systemlevel simulator and how these two models are imple
mented.
4.2.1. General Structure
In [5] a computationally eﬃcient linktosystem level model is proposed and its em
bedding in a Matlabbased systemlevel simulator which includes as features MIMO
with DTxAA, and a Minimum Mean Squared Error (MMSE) equalizer. This system
level simulator presented shows a structure that identiﬁes the relevant interference
terms and allows for the generation of scalar fading parameters prior to systemlevel
simulation. Utilizing this special structure nearly all linkdedicated procedures can
be included in these fading parameters, thus during the runtime of the system level
simulation only scalar multiplications are needed to compute the SINR, thus reducing
signiﬁcantly the computational eﬀort.
36
4.2. MIMO HSDPA System Level Simulator
The structure of the simulator is depicted in Figure 4.2. It can be decomposed in the
main network elements [23]:
• Node B. It represents all network related procedures, in particular it carries
out the scheduling trying to balance the throughput and fairness as well as the
Transport Format Combination (TFC) decision based the UE feedback.
• SL model. The channel model is represented using an abstract model, so that
it needs lower computational complexity, even though it is capable of support
diﬀerent transmission schemes and receivers.
• UE. User speciﬁc algorithms and feedback decisions are done here. In addition,
the evaluation of the transmission success is performed.
• delay. A signaling delay is imposed.
In following sections we will focus on the linkmeasurement model since the improve
ments in this study will be done on this model. Only a short reference to the link
performance model is done, but more information can be found in [23].
4.2.2. LinkMeasurement Model
This model reﬂects base station and terminal measurements, such as estimated SINR
used for channel dependent scheduling and link adaptation. Measurement results not
only depend on channel and intercell interference, in fact, they also depend on the
measurement phase as for example transmission power and beamforming weights. This
model is needed to provide appropriate estimates of the channel quality.
In this section we present the system model utilized in [23] to derive the computation
ally eﬃcient linkmeasurement model, we explain shortly the main parts and let the
reader refer to [23] for ﬁnding more detailed information.
System model
Figure 4.3 depicts the model where the transmitter (Tx) and the receiver (Rx) are
equipped with n
T
and n
R
antennas, respectively. Accordingly the input data stream s
is demultiplexed into N parallel streams, s
0
, . . . , s
N−1
with the individual data streams
s
n
, 0 ≤ n ≤ N −1, being spread by a number of spreading sequences, ϕ
n
(multicode
usage) and scrambling sequences.
These spread and scrambled sequences are then mapped to the n
T
transmit anten
nas using a preﬁltering matrix D ∈ C
n
T
×N
, which contains the precoding weights,
w
1
, . . . , w
n
T
N
. In this work the maximal amount of streams and transmitter antennas
is ﬁxed to 2, and therefore there will be 4 precoding weights. At the receiver, the
37
4. Mobile Network Simulations
stream decision
BLER evaluation
HARQ model
SINRCQI mapping
UE position update
mapping generation
BLER data
Node B
MAChs
SL model UE
delay
SINR (slot)
get ff parameters
shadow fading
antenna gain
macro pathloss
SINR averaging
ff generation
MI mapping
Node B
output
avg. SINR per stream
stream decision
BLER evaluation
HARQ model
SINRCQI mapping
UE position update
mapping generation
BLER data
avg. SINR
Node B output
UE output
power management
QCI/TFC decision
stream decision
scheduler
buffer
transmission settings
delayed UE output
Structure
MIMO HSDPA systemlevel simulator
Node B
SLmodel UE
Figure 4.2.: MIMO HSDPA systemlevel simulator main structure
signals are gathered with n
R
antennas and chip spaced sampled before they enter the
discrete time STMMSE equalizer. The MIMO channel H ∈ C
n
R
×n
T
L
is modeled as
timediscrete, frequencyselective channel,
H =
h
1,1
(0) . . . h
1,n
T
(0) . . . h
1,n
T
(L −1)
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
h
n
R
,1
(0) . . . h
n
R
,n
T
(0) . . . h
n
R
,n
T
(L −1)
¸
¸
¸ (4.1)
where the entry h
r,t
(l) denotes the lth sampled chip of the channel impulse response
from transmit antenna t to receive antenna r, with a total length of L chip intervals.
Note that the pulse shaping the transmit and receive ﬁltering, as well as the sampling
38
4.2. MIMO HSDPA System Level Simulator
ϕ
0
ϕ
N1
ϕ
0
*
*
ϕ
N1
D
E
M
U
X
S
T

M
M
S
E
D
s
0
s
x
0
s
N1
x
N1
H
y
1
y
n
R
n
T
x
0
x
N1
22
27
∧
∧
1
s
0
s
N1
∧
∧
Figure 4.3.: System Model
operation can be incorporated in the MIMO channel matrix. For sake of notational
simplicity, an equivalent discrete channel Γ ∈ C
n
R
×n
T
L
is deﬁned that includes the
preﬁltering matrix D and the MIMO channel H, i.e. Γ = H · (I
L
⊗ D), with I
L
denoting the identity matrix of size L, and ⊗ being the Kronocker product. With
this, the input output relation at time instant k, formulated by means of the equivalent
channel matrix, is given by
y(k) = Γx(k) +n(k), (4.2)
where we introduced the receive vector y(k) = [y
1
(k), . . . , y
n
R
(k)]
T
, the transmit vector
x(k) = [x
0
(k), . . . , x
N−1
(k), . . . , x
N−1
(k −L+1)]
T
and the receive noise vector n(k) =
[n
1
(k), . . . , n
n
R
(k)]
T
.
The STMMSE solution can be computed again according to Equation (3.14).
Equivalent Fading Parameteres Description
In this section we just present the conclusions obtained in [23]. The receive power is
decomposed into diﬀerent interference terms to derive the system level model. With
this decomposition it is possible to describe the characteristics of the individual terms
by means of fadingparameters that are real valued scalar processes. These parameters
can be computed oﬄine and loaded for the runtime of a systemlevel simulation, thus
signiﬁcantly reducing the computational burden.
First of all we decomposed the equalizer coeﬃcients W
d
= [w
1
, . . . , w
n
, . . . , w
N
]
T
and the channel Γ
u,b
= [γ
0
u,b
, . . . , γ
m
u,b
, . . . , γ
N(E+L−1)−1
u,b
] where n denotes the stream
index and m is the index of the Tx chips for all streams entering the equalizer span.
Without losing generality, the user u = 0 and the basestation b = 0 are the one
deﬁned of interest.
1. Desired Signal: The desired power signal is given by
P
s,n
=
w
T
n
γ
dN+n
00
2
· P
n,ζ
= G
s,n
· P
n,ζ
(4.3)
39
4. Mobile Network Simulations
where P
n,ζ
denotes the power on stream n and spreading code ζ spent for user
u = 0 by basestation b = 0. G
s,n
describes the equivalent fading of the useful
signal power.
2. Intracel l Interference: The intracell interference is composed by a number of
terms, i.e. the remaining ISI after equalization, the intercode interference when
the same scrambling but a diﬀerent spreading code is used, the intracell inter
ference from users that are not served in the same instant as the user of interest
but with the same instant as the user of interest but with the same scrambling
and spreading code, and the intracell interference from users with the same
scrambling but diﬀerent spreading code. The following equation represents a
simpliﬁcation of all this interferences.
P
intra,n
=
¸
P
0
+o
BF,n
·
U
0
¸
u=1
P
u
¸
· o
intra,n
· G
s,n
, (4.4)
where the intracell orthogonality, o
intra,n
, is deﬁned as
o
intra,n
1
N
·
¸
N(E+L−1)−1
m=0
m/ ∈[dN,dN+N−1]
w
T
n
γ
m
00
2
w
T
n
γ
dN+n
00
2
, (4.5)
and the precoding orthogonality, o
BF,n
, is given by
o
BF,n
1
G
s,n
1
Ω
¸
ω∈Ω
N(E+L−1)−1
¸
m=0
m/ ∈[dN,dN+N−1]
w
T
n
γ
m
ω0
2
(4.6)
3. Intrastream Interference: The interference generated by the parallel transmission
of a second stream is given by
P
INT,n
=
N−1
¸
m=0
m=n
w
T
n
γ
dN+m
00
2
· P
m,ζ
= G
s,n
·
N−1
¸
m=0
m=n
o
INT,m
· P
m,ζ
, (4.7)
where o
INT,n
w
T
n
γ
dN+m
00
2
denotes the intrastream orthogonality factor.
4. Intercel l Interference: This interference term assumes that all users in a neigh
bouring cell will apply the same precoding coeﬃcient, and that all substreams
designated for one user are equally powered. Having considered this the intercell
interference is given by
P
inter,n
=
B
¸
b=1
P
tot,b
1
N
w
T
n
¸
N(E+L−1)−1
¸
m=0
γ
m
0b
(γ
m
0b
)
H
¸
w
∗
n
=
B
¸
b=1
P
tot,b
G
inter,b,n
, (4.8)
40
4.2. MIMO HSDPA System Level Simulator
5. Thermal Noise: The thermal noise is modeled as i.i.d. white Gaussian noise,
with identical power on all antenna and over all chips that enter the equalizer.
The power (on symbol level) of the thermal noise can denoted as
W = E
w
T
n
n
E
(k)
2
2
= σ
2
n
w
T
n
2
2
. (4.9)
Inﬂuence of NonData Channels
The intereference caused by nonspread channels is given by
P
ns,n
=
P
non−spread
N
G
s,n
¸
1 +
N−1
¸
m=0
m=n
o
INT,m
¸
+P
non−spread
o
intra
G
s,n
(4.10)
and the interference caused by other spread channels than HSDSCH assuming the
total power of other spread channels in the cell to be P
other−spread
, can be evaluated
to be
P
s,n
= P
other−spread
o
intra,n
G
s,n
(4.11)
Resulting SINR Description
The SINR on substream n and spreading code ζ, as observed after equalization
and despreading, taking into consideration all the exposed above, can be expressed
by
SINR
n,ζ
=
SF · P
s,n
SF · P
INT,n
+P
intra,n
+P
ns,n
+P
s,n
+P
inter,n
+W
, (4.12)
where SF denotes the spreading factor and P
INT,n
simpliﬁes to P
INT
= o
INT
· G
s,n
· P
m,ζ
for the case DTxAA, since there is only one interfering parallel substream.
4.2.3. LinkPerformance Model
The link performance model provides an estimate of the link performance when the
decisions on the radio resource management (RRM), scheduling and link adaptation
are already known. This model generates either a bit or a block error probabilities and
is utilized for quality estimation at runtime, which in turn can generate, for example
retransmissions and aﬀect also slow link adaption. Such a model can be viewed as
a conditional probability; the probability that the transmitted code word is decoded
erroneously given the channel, Tx power, beamforming weights and interference plus
noise during the interleaving period [24].
41
4. Mobile Network Simulations
Parameter Value
power allocated at HSDSCH 20 W
channel model ITU PedB
equalizer length L
f
30 chips
equalizer delay τ 15 chips
number of spreading factors per stream 15
channel estimator full knowledge of the channel
Table 4.1.: SINR validation simulation settings
4.3. Linkmeasurement model enhancement
This section presents a validation of the current linkmeasurement model, and shows
the fact that there is the eﬀect of the channel estimation missing, and therefore its link
with the power assigned to the common pilot channel. An improvement of the current
model is presented, which extends the capabilities of the MIMO HSDPA systemlevel
simulator.
4.3.1. Validation of the current model
In order to validate the linkmeasurement model presented in [23], an adaptation
of the model has been implemented in the basic MIMO HSDPA linklevel simulator
presented in Chapter 3. The intercell interference and the interference due to the
inﬂuence of nondata channels has been omitted since they are not implemented in
the basic linklevel simulator.
Figure 4.4 depicts the performed simulations to validate the proposed SINR descrip
tion, for each transmission scheme. The genie SINR is the SINR observed at the
demapper input, so after de despreading and is obtained as deﬁned in (3.3).
The main simulation parameters can be seen in Table 4.1. Notice that full knowledge
of the channel is assumed. It can be observed that the linkmeasurement model shows
a very good agreement with the genie or true SINR.
4.3.2. Inﬂuence of CPICH in the current model
The CPICH is a very important channel since the channel estimation in the UE is
performed according to this. Figure 4.5a depicts the MSE of the channel coeﬃcients,
deﬁned as
42
4.3. Linkmeasurement model enhancement
−10 −5 0 5 10 15 20 25 30
−20
−15
−10
−5
0
5
10
15
20
noise power [dBW]
S
I
N
R
H
S
−
D
S
C
H
[
d
B
]
genie
approximation
(d) SISO
−10 −5 0 5 10 15 20 25 30
−20
−15
−10
−5
0
5
10
15
20
noise power [dBW]
S
I
N
R
H
S
−
D
S
C
H
[
d
B
]
genie
approximation
(e) MISO 2x1
−10 −5 0 5 10 15 20 25 30
−15
−10
−5
0
5
10
15
20
25
30
noise power [dBW]
S
I
N
R
H
S
−
D
S
C
H
[
d
B
]
genie
approximation
(f) MIMO 2x2 TxAA
−10 −5 0 5 10 15 20 25 30
−20
−15
−10
−5
0
5
10
15
20
noise power [dBW]
S
I
N
R
H
S
−
D
S
C
H
[
d
B
]
stream 1 genie
stream 1 approximation
stream 2 genie
stream 2 approximation
(g) MIMO 2x2 DTxAA
Figure 4.4.: Model validation
MSE(
ˆ
h) = E
ˆ
h −h
2
(4.13)
where
ˆ
h are the estimated channel coeﬃcients using the LS estimator or the correlation
based estimator. The result is obvious, a smaller MSE is obtained when the power
allocated to the CPICH is higher. It is also important to notice that the curves
are unique for a certain CPICH SINR or CPICH
E
c
/I
0
. Figures 4.5b4.5d, show the
MSE for the rest of schemes, note that the MSE increases with the channel complex
ity.
The linkmeasurement model, described in Section 4.2.2, does not take into account
this eﬀect and supposes a perfect knowledge of the channel when performs the equal
ization, which is a situation far from the reality. The consequence is a poor per
formance of the approximation when the CPICH is considered. Figure 4.6 depicts
the eﬀect, where in each case, the lower curve denotes a dependence with the MSE
43
4. Mobile Network Simulations
0 10 20 30 40 50 60 70 80 90 100
1
2
3
4
5
6
7
8
CPICH EcIor [%]
M
S
E
[
d
B
]
MSE H (LS)
MSE H (cor)
(a) SISO
0 10 20 30 40 50 60 70 80 90 100
4
5
6
7
8
9
10
11
12
13
14
CPICH EcIor [%]
M
S
E
[
d
B
]
MSE H (LS)
MSE H (cor)
(b) MISO 2x1
0 10 20 30 40 50 60 70 80 90 100
6
7
8
9
10
11
12
13
14
15
16
CPICH EcIor [%]
M
S
E
[
d
B
]
MSE H (LS)
MSE H (cor)
(c) MIMO 2x2 TxAA
0 10 20 30 40 50 60 70 80 90 100
6
7
8
9
10
11
12
13
14
15
CPICH EcIor [%]
M
S
E
[
d
B
]
MSE H (LS)
MSE H (cor)
(d) MIMO 2x2 DTxAA
Figure 4.5.: MSE channel coeﬃcients
of the channel coeﬃcients, again for a particular CPICH
E
c
/I
0
and channel estimator
algorithm.
4.3.3. Modeling the eﬀects of CPICH
In principle, in a real system, the UE would accomplish the equalization using the
channel estimated coeﬃcients provided by the channel estimator. In the case of our
linklevel simulator, the equalizer shows the following structure
f
(k)
= σ
2
s
ˆ
H
w
R
ss
ˆ
H
H
w
+R
vv
−1
ˆ
H
w
e
τ
k
,2(L
h
+L
f
−1)
. (4.14)
where the vector e
τ
k
,2(L
h
+L
f
−1)
is a zero vector of length 2(L
h
+L
f
−1) with a single
one at position
τ
k
= τ + (k −1)(L
h
+L
f
−1) k = 1 . . . K, (4.15)
and
ˆ
H
w
ˆ
H(W⊗I
L
h
+L
f
−1
).
44
4.3. Linkmeasurement model enhancement
(a) SISO (b) MISO 2x1
(c) MIMO 2x2 TxAA (d) MIMO 2x2 DTxAA
Figure 4.6.: Eﬀect of CPICH in the current model and LS estimator
The solution of including a channel estimator in the calculation of the fading parame
ters is not desired, for the reason that it would increase the computational complexity
of the systemlevel simulator.
Since we do not want to increase the computational complexity, we have to ﬁnd out
how to incorporate the estimated channel without directly utilizing channel estimators.
Our proposal is based on considering the estimated channel matrix
ˆ
H as the true
channel matrix H and a matrix H
∆
representing the channel estimation error. This
can be express as
ˆ
H = H+H
∆
. (4.16)
The matrix H
∆
is constructed like the channel matrix H (in 3.10, or in 4.1 if we
use the structure of the linkmeasurement model). The nonzero elements of H
∆
are
assumed to be a i.i.d. Gaussian with 0 mean and a variance equal to the MSE of the
channel estimator, that as it has been shown before depends on the noise term and
the estimator itself. Thus, we can say that
Var{H
∆
} = f(estimator algorithm, CPICH
E
c
/I
0
) (4.17)
45
4. Mobile Network Simulations
(a) SISO
0 10 20 30 40 50 60 70 80 90 100
−12
−10
−8
−6
−4
−2
0
2
CPICH E
c
/I
or
[%]
S
I
N
R
H
S
−
D
S
C
H
[
d
B
]
genie
new approximation
(b) MISO 2x1
(c) MIMO 2x2 TxAA (d) MIMO 2x2 DTxAA
Figure 4.7.: Model validation and LS estimator
Hence we can say that the STMMSE equalizer in (4.14), becomes
f
(k)
= σ
2
s
(H
w
+H
w∆
)R
ss
(H
w
+H
w∆
)
H
+R
vv
−1
(H
w
+H
w∆
)e
τ
k
,2(L
h
+L
f
−1)
. (4.18)
This solution enables the possibility to have a set of precalculated MSE curves which
can be used to model the eﬀects of the CPICH channel in a systemlevel simulator
without increasing the computational complexity in excess because only a scalar prod
uct and a sum of matrix have to be performed during the generation of the fading
parameters.
4.3.4. Validation of the enhanced model
We have conducted a set of validations of the new model, for diﬀerent CPICH
E
c
/I
0
, the
results can be seen in Figure 4.7. Although the matrix H
w∆
depends on the variance
of a random variable, it can be seen that the ﬁtting is very good.
46
5. CPICH Power Optimization
This chapter covers the analysis of the CPICH, and the optimization of the power
allocated to this channel. The outline is as follows: Section 5.1 contains the anal
ysis of the CPICH showing its main eﬀects due to power variation. The simulation
methodology utilized which leads to obtain the optimization of the CPICH power
is explained in Section 5.2. Finally the simulation results are presented in Section
5.3.
5.1. Importance of CPICH power optimization
In HSDPA, as in UMTS, channel estimation is accomplished through the use of a
signaling channel. The CPICH is this signaling channel. It is a ﬁxed rate downlink
physical channel that carries a predeﬁned bit sequence. Apart from being utilized
for the channel estimation in the HSDSCH, the CPICH also provides the channel
estimation reference for common channels.
The measurements of the CPICH are used in the cell selection, cell reselection and
handover procedures. The UEs scan for the CPICH signals continuously and measure
the received level of chip energy to interference plus noise density ratio (E
c
/I
0
) of all
pilot signals they can detect. I
0
denotes the total received power density, including
signal and interference, as measured at the UE antenna connector. In order to keep a
mobile referenced to a cell, the E
c
/I
0
at the mobile terminal must exceed a minimum
threshold at all times. The E
c
/I
0
ratio can be described as
CPICH
E
c
/I
0
=
P
CPICH
L
p
¸
numBS
i=1
P
TX,i
L
p
i
+I
intra
+N
0
, (5.1)
where P
CPICH
is the CPICH power of the cell we are measuring, L
p
is the overall path
loss to the best server, P
TX,i
is the total transmit power by the base station i, L
pi
is
the overall path loss to base station i, I
intra
is the intracell interference, due to loss
of orthogonality between channelisation codes, N
0
is the thermal noise of the UE and
numBS is the number of base stations in the network [25].
47
5. CPICH Power Optimization
Data Channels
(HSDPA, R99)
CPICH
Other Channels
Data Channels
(HSDPA, R99)
CPICH
Other Channels
Figure 5.1.: Cell resizing eﬀect due to CPICH power variation
The cell with the highest CPICH level at the UE is selected as the serving cel l. As
a consequence by adjusting the CPICH power level, the cell load can be balanced
between diﬀerent cells. Reducing the CPICH power causes part of the terminals to
hand over to other cells, while increasing it invites more terminals to hand over to the
cell, as well as to make their initial access to the network in that cell. This eﬀect is
depicted in Figure 5.1.
If a mobile terminal is unable to clearly receive one dominant CPICH signal, due
to interference or coverage problems and therefore the terminal can not perform an
eﬀective estimation of the channel, the result is likely to be dropped calls, failed
initiations, poor voice quality and/or data throughput. Figure 4.5d shows the MSE
of the channel coeﬃcients for a MIMO 2x2 TxAA link. It can be observed that
the MSE increases when the amount of power allocated to the CPICH decreases, in
other words, the less power assigned to the pilot channel the worse estimation of the
channel.
Nevertheless, the more power is spent for pilot signalling, the less power is left to
serve the user traﬃc. Excessive pilot power can easily take too large proportion of
the total available transmit power, so that not enough power is left for traﬃc chan
nels. Even considering the hypothetical case of having inﬁnite power in both chan
nels, the increase of the pilot channel would be beneﬁcial just before reaching an
upper limit. One of the limiting factors is the loss of orthogonality between spreading
codes.
Another limiting factor, which we do not take into account for our analysis, that is
produced due to too high values of CPICH power is called pilot pol lution. Hearing
equal powered CPICH signals or multiple strong CPICH with their multipath com
ponents causes pilot pollution. One of the possible consequences of pilot pollution is
48
5.2. Simulation methodology
Parameter Value
total power available at Base Station 20 W
distance between Base Stations 1000 m
antenna pattern 3gpp 25.996
path loss model urban micro, COST231 WalﬁshIkegami
number of cells 19
number of Sectors/cell 3
transmitter frequency 2 GHz
receiver noise ﬁgure 9 dB
Table 5.1.: Simulation scenario
that a UE may change the best serving cell continuously due to slow faded CPICH sig
nals, increasing the signaling load in the network by causing additional cell reselection
messages.
5.2. Simulation methodology
To ascertain the optimal CPICH power in a network we have used a basic network
simulator (Section 5.2.1) that provides the overall intercell interference plus thermal
noise in the UE (I
all
) for a particular scenario. This I
all
depends basically on the
scenario parameters and the position of the UE inside the cell. Consider one UE
with a particular I
all
associated, this overall interference term is then introduced in
the basic MIMO HSDPA linklevel simulator (Section 3) and with it we derive the
CPICH power value that maximizes the HSDSCH SINR. In any case, this value
will be only optimal for this particular user, so the process is repeated for a uniform
grid inside one sector for the purpose of obtaining a uniform number of maximums
inside a sector which allows us to calculate the optimal CPICH power of the cell by
averaging.
5.2.1. Preequalization SINR
In fact, the simulator, which we use to obtain the overall interference term (I
all
), is
designed to calculate a preequalization SINR of set of points in a deﬁned area for a
given base station transmission power of the sector i, P
TX,i
, and a set of parameters
that deﬁne a uniform HSDPA network. In our case this SINR in particular spot inside
the deﬁned area corresponds to Equation (5.1) when I
intra
is set to 0, from now on
we will refer to this preequalization SINR as
CPICH
E
c
/I
0
,s
. Table 5.1 shows the
parameters that have been used in the simulations.
49
5. CPICH Power Optimization
distance in m
d
is
ta
n
c
e
in
m
−2000 −1500 −1000 −500 0 500 1000 1500 2000
−1500
−1000
−500
0
500
1000
1500
−6
−4
−2
0
2
4
6
8
10
12
(a) Antenna Pattern [dB]
distance in m
d
is
ta
n
c
e
in
m
−2000 −1500 −1000 −500 0 500 1000 1500 2000
−1500
−1000
−500
0
500
1000
1500
90
100
110
120
130
140
150
160
(b) Path loss [dB]
(c) Overall path loss [dB]
distance in m
d
is
ta
n
c
e
in
m
−2000 −1500 −1000 −500 0 500 1000 1500 2000
−1500
−1000
−500
0
500
1000
1500
−120
−110
−100
−90
−80
−70
−60
−50
−40
(d) Radiated power [dBm]
−2000 −1500 −1000 −500 0 500 1000 1500 2000
−1500
−1000
−500
0
500
1000
1500
−80
−70
−60
−50
−40
−30
(e) Overall interference [dB] (f) preequalized SNR [dBm]
Figure 5.2.: Path loss ﬁgures
The
CPICH
E
c
/I
0
,s
, as it is calculated in the simulator, for a particular spot is described
as
CPICH
E
c
/I
0
,s
=
P
TX,i
L
p
¸
numBS
k=1
P
TX,k
L
kp
+N
0
(5.2)
where s is the spot index, N
0
is the thermal noise at the receiver, L
ip
is the overall
link loss between the base station at cell i and the user at the position p, and it is
calculated as follows
L
ip
= l
ip
/G
ip
(5.3)
where l
ip
is the path loss (Figure 5.2b) calculated using COST231 WalﬁshIkegami
model for urban micro environments (see Section 2.2.3) and G
ip
is the antenna gain
(Figure 5.2b) which depends on the antenna pattern deﬁned in [26]. The resulting
overall path loss for the deﬁned area can be seen in Figure 5.2c.
The radiated power of the target sector,
P
TX,i
L
p
, can be seen in Figure 5.2d.
50
5.2. Simulation methodology
distance in m
d
i
s
t
a
n
c
e
i
n
m
0 100 200 300 400 500 600
−250
−200
−150
−100
−50
0
50
100
150
200
250
−2
0
2
4
6
8
10
12
14
Figure 5.3.: I
all
[dBW] of the target sector
The sum of the intercell interference and thermal noise can be seen in Figure 5.2e
where each dot represents one base station and the target base station is situated at
the position (0,0). The resulting
CPICH
E
c
/I
0
in the target sector for each spot is
depicted in Figure 5.2f.
The channel in the basic MIMO HSDPA linklevel simulator is normalized, it means
that the path loss term must be included in the interference term. The noise term
in this case includes the intercell interference and the thermal noise in the receiver.
Once this have been considered the I
all
for a speciﬁc spot that will be inserted in the
linklevel simulator can be easily derived from (5.2) and becomes
I
all
= L
p
· (I
inter
+N
0
) (5.4)
where I
inter
=
¸
numBS
k=1
P
TX,k
L
kp
. Figure 5.3 shows the total number of spots that have
been considered inside the target cell. Every cross has a diﬀerent I
all
that will be
entered in the linklevel simulator.
5.2.2. HSDSCH SINR optimization
Our approach considers the optimization of the power allocated to the CPICH when
the SINR of the data channel (HSDSCH) is maximal. Thus the cost function is
E
c
I
or
CPICH,opt
= argmax
E
c
I
or
CPICH
SINR
HS−DSCH,s
(5.5)
51
5. CPICH Power Optimization
Figure 5.4.: I
all
Comparison
where SINR
HS−DSCH,s
is the SINR in the data channel and the subindex s indicates
the spot position and therefore the I
all
associated.
E
c
I
or
CPICH
is the power allocated
to the CPICH as a fraction of the total transmission power I
or
.
The procedure that leads to obtain the optimal point is the following:
1. Insert the (I
all
) of a particular spot (see Figure 5.3), which has previously been
acquired with the use of the preequalization SINR simulator, into the basic
MIMO HSDPA linklevel simulator.
2. Deﬁne SINR
HS−DSCH,s
as a function of
E
c
I
or
CPICH
. Figure 5.4 shows an example
with two functions in SISO transmission, each one is deﬁned for a diﬀerent I
all
.
It is important to emphasize that the base station transmitted power remains
constant over the curve, in other words, an increase of the CPICH transmitted
power infers a decrease of the HSDSCH transmitted power, and vice versa.
3. Find the maximal SINR
HS−DSCH,s
of the function, and its
E
c
I
or
CPICH
associated.
Note that the maximal of the function also changes with the change of the overall
interference term.
4. Repeat the process for every spot inside the target sector.
Once the set of maximal SINR
HS−DSCH,s
, thus the optimal
E
c
I
or
CPICH
values associ
ated to them are found, we proceed to average the optimal
E
c
I
or
CPICH
for the target
52
5.3. Simulation results
sector as follows. Suppose we have S spots in the target sector, in the single stream
case the optimal value is originated from
Optimal
E
c
I
or
CPICH
=
1
S
S
¸
s=0
argmax
E
c
I
or
CPICH
SINR
HS−DSCH,s
(5.6)
In the case of double stream, the optimization is done independently per stream.
Nevertheless, we see that both results are quite similar and decide to obtain the optimal
by averaging both streams optimal values.
5.3. Simulation results
In this section we present the results of the optimal CPICH power that maximizes
SINR
HS−DSCH
, for each transmission scheme supported by our basic MIMO HSDPA
linklevel simulator. The conﬁguration of the linklevel simulator is done according to
Table 5.2, and the parameters concerning the network are deﬁned according to Table
5.1.
Figures 5.5  5.8 show the optimal CPICH transmission power for each spot according
to the scenario speciﬁed for the SISO, MISO 2x1, MIMO TxAA and MIMO DTxAA,
respectively, for a Pedestrian B channel model and considering in the case of double
stream the averaging of the results. Note that an increase of the CPICH power is
required when the number of antennas increases, this eﬀect is due to the increase in
complexity of the channel. This extra complexity of the channel has a negative eﬀect
on the channel estimation which needs more help in terms of power to perform better.
Table 5.3 sums up the results.
We have performed the simulation for Pedestrian A and Pedestrian B channel model.
Pedestrian B (18 chips) is a very large channel when compared to Pedestrian A (3
chips). The optimal CPICH power when Pedestrian B is utilized is higher than when
Pedestrian A is used, again, this is due to the diﬀerence in complexity of the channel.
In the case of Pedestrian B the equalizer perform worse, which leads to an increase of
the CPICH power to obtain the maximal data SINR.
However, the maximal data SINR is similar over a large range of values (See Figure
5.4. Thus the selection of the optimal CPICH power is actually only critical in a
particular range, which is the range closer to the minimal CPICH power, where the
slope is more pronounced.
53
5. CPICH Power Optimization
Parameter Value
total power available at Base Station I
or
20 W
supported Transmission schemes SISO, MISO, MIMO TxAA & DTxAA
antenna utilization 2x1, 2x2
number of transmitted packets 500
equalizer span L
f
30 chips
equalizer delay τ 15 chips
transmitter frequency 2 GHz
channel estimator LS Estimator
channel model ITU PedA, PedB
user mobility 3 km/h
UE receiver type STMMSE
beamforming delay d
bf
4 packets
utilized spreading codes for HSDSCH 15
HSDSCH modulation 4QAM
Table 5.2.: linklevel simulation parameters
Transmission scheme PedA Optimal CPICH PedB Optimal CPICH
SISO 4.16 W 7.00 W
MISO 2x1 4.40 W 7.93 W
MIMO 2x2 TxAA 4.86 W 8.53 W
MIMO 2x2 DTxAA 5.78 W 8.81 W
Table 5.3.: CPICH power optimization results
5.4. Conclusion
The optimal CPICH power obtain in this chapter is far from the one used in a real
network, usually set to 510% of the total transmission power of the base station. The
main reason is that pilot pollution is not considered, in any case the cost function
utilized shows the eﬀect of the CPICH power variation in each transmission scheme
and sets the basis to develop new studies able to ﬁt the real networks more accu
rately.
54
5.4. Conclusion
Figure 5.5.: SISO
Figure 5.6.: MISO 2x1
55
5. CPICH Power Optimization
Figure 5.7.: MIMO 2x2 TxAA
Figure 5.8.: MIMO 2x2 DTxAA
56
A. Appendix
A.1. Channelisation Codes
The utilized channelisation codes are OVSF codes that preserve the orthogonality
between a user’s diﬀerent physical channels. The OVSF codes can be deﬁned using
the code tree of Figure A.1 [27].
C
ch,1,0
C
ch,2,0
C
ch,2,1
C
ch,4,2
C
ch,4,3
C
ch,4,0
C
ch,4,1
= (1)
= (1,1)
= (1,1)
= (1,1,1,1)
= (1,1,1,1)
= (1,1,1,1)
= (1,1,1,1)
Figure A.1.: Codetree for generation of OVSF codes
In Figure A.1, the channelisation codes are uniquely described as C
ch,SF,k
, where
SF is the spreading factor of the code and k is the code number, 0 ≤ k ≤ SF −
1.
Each level in the code tree deﬁnes channelisation codes of length SF, corresponding
to a spreading factor of SF in Figure A.1.
The generation method for the channelisation code is deﬁned as:
C
ch,1,0
= 1,
¸
C
ch,2,0
C
ch,2,1
¸
=
¸
C
ch,1,0
C
ch,1,0
C
ch,1,0
−C
ch,1,0
¸
=
¸
1 1
1 −1
¸
57
A. Appendix
C
ch,2
(n+1)
,0
C
ch,2
(n+1)
,1
C
ch,2
(n+1)
,2
C
ch,2
(n+1)
,3
.
.
.
C
ch,,2
(n+1)
2
(n+1)
−2
C
ch,,2
(n+1)
2
(n+1)
−1
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
=
C
ch,2
n
,0
C
ch,2
n
,0
C
ch,2
n
,0
−C
ch,2
n
,0
C
ch,2
n
,1
C
ch,2
n
,1
C
ch,2
n
,1
−C
ch,2
n
,1
.
.
.
.
.
.
C
ch,2
n
,2
n
−1
C
ch,2
n
,2
n
−1
C
ch,2
n
,2
n
−1
−C
ch,2
n
,2
n
−1
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
¸
The leftmost value in each channelisation code word corresponds to the chip transmit
ted ﬁrst in time.
A.2. Scrambling
The scrambling code sequences are constructed by combining two real sequences into a
complex sequence [27]. Each of the two real sequences are constructed as the position
wise modulo 2 sum of 38400 chip segments of two binary msequences generated by
means of two generator polynomials of degree 18. The resulting sequences thus consti
tute segments of a set of Gold sequences. The scrambling codes are repeated for every
10 ms radio frame. Let x and y be the two sequences respectively. The x sequence is
constructed using the primitive (over GF(2)) polynomial 1+X
7
+X
18
. The y sequence
is constructed using the polynomial 1 +X
5
+X
7
+X
1
0 +X
18
.
The sequence depending on the chosen scrambling code number n is denoted z
n
, in the
sequel. Furthermore, let x(i), y(i) and z
n
(i) denote the i:th symbol of the sequence x,
y, and z
n
, respectively. The msequences x and y are constructed as:
Initial conditions:
• x is constructed with x(0) = 1, x(1) = x(2) = . . . = x(16) = x(17) = 0.
• y(0) = y(1) = . . . = y(16) = y(17) = 1
Recursive deﬁnition of subsequent symbols:
• x(i + 18) = x(i + 7) +x(i) mod 2, i = 0, . . . , 2
18
−20.
• y(i + 18) = y(i + 10) +y(i + 7) +y(i + 5) +y(i) mod 2, i = 0, . . . , 2
18
−20.
The n:th Gold code sequence z
n
, n = 0, 1, 2, . . . , 2
18
−2, is then deﬁned as:
• z
n
(i) = x((i +n) mod (2
18
−1)) +y(i) mod 2, i = 0, . . . , 2
18
−2.
58
A.2. Scrambling
These binary sequences are converted to real valued sequences Z
n
by the following
transformation:
Z
n
(i) =
+1 ifZ
n
(i) = 0
−1 ifZ
n
(i) = 1
for i = 0, 1, . . . , 2
18
−2
Finally, the n:th complex scrambling code sequence S
dl,n
is deﬁned as:
• S
dl,n
(i) = Z
n
(i) +jZ
n
((i + 131072) mod (2
18
−1)), i = 0, 1, . . . , 38399.
Note that the pattern from phase 0 up to the phase of 38399 is repeated.
59
Bibliography
[1] H. Holma and A. Toskala, HSDPA/HSUPA for UMTS. The Atrium, Southern
Gate, Chichester, West Sussex PO19 8SQ, England: John Wiley & Sons, Ltd,
2006.
[2] 3GPP, Technical Speciﬁcation Group Radio Access Network; Multiple Input Mul
tiple Output in UTRA (Tech. Spec. 25.876 V7.0.0), release 7 ed., March 2007.
[3] 3GPP, Technical Speciﬁcation Group Radio Access Network; Physical channels
and mapping of transport channels onto physical channels (FDD) (Tech. Spec.
25.211 V8.2.0), release 7 ed., September 2008.
[4] M. of WINNER, “Assessment of advanced beamforming and MIMO technologies,”
Tech. Rep. Tech. Rep IST2003507581 D.2.7 v1.0, WINNER, 2003.
[5] M. Wrulich, S. Eder, I. Viering, and M. Rupp, “Eﬃcient linktosystem level
model for MIMO HSDPA,” GLOBECOM Workshops IEEE, 2008.
[6] “www.3gpp.org.”
[7] J. Ramis, “Mobile communication networks lecture notes.” Lecture slides, avail
able online: http://dmi.uib.es/ jramis/, University of Balearic Island, 2005.
[8] G. L. Stüber, Principles of Mobile Communication. Kluwer Academic Publishers,
2nd ed., 2001.
[9] H. Holma and A. Toskala, WCDMA for UMTS. The Atrium, Southern Gate,
Chichester, West Sussex PO19 8SQ, England: John Wiley & Sons, Ltd, 2004.
[10] B. Sklar, “Rayleigh fading channel in mobile digital communication systems part
I: Characterization,” IEEE Communications Magazine, 1997.
[11] COST 231, Digital mobile radio towards future generation systems.
[12] Qualcomm, “Release 7 hspa+ for mobile broadband evolution.” available online:
http://hspa.gsmworld.com/upload/news/ﬁles/10122007082416.pdf, 2007.
[13] P. Soler, “MIMO en sistemes hiperlan/2 i ieee 802.11a.,” Master’s thesis, Univer
sity of Balearic Island, 2007.
61
Bibliography
[14] S. Geirhofer, C. Mehlführer, and M. Rupp, “Design and realtime measurement
of HSDPA equalizers,” SPAWC, 2005.
[15] E. Zacarías, “Adaptive transmit beamforming with closed loop feedback in MIMO
systems,” Postgraduate course in Radio Communications, 2004.
[16] Y. Rosa and C. Xiao, “Simulation models with correct statistical properties for
rayleigh fading channels,” IEEE TRANSACTIONS ON COMMUNICATIONS,
vol. 51, June 2003.
[17] ITUR, “Recommendation itur m.1225: Guidelines for evaluation of radio trans
mission technologies for imt2000,” tech. rep., Tech.Rep., 1997.
[18] 3GPP, Technical Speciﬁcation Group Radio Access Network; User Equipment
(UE) radio transmission and reception (FDD) (Tech. Spec. 25.101 V7.14.0), re
lease 7 ed., December 2008.
[19] C. Mehlführer, M. Wrulich, and M. Rupp, “Intracell interference aware equaliza
tion for TxAA HSDPA,” Wireless Pervasive Computing, 2008.
[20] C. Mehlführer, S. Caban, and M. Rupp, “Measurement based evaluation of low
complexity receivers for DTxAA HSDPA,” EUPISCO, 2008.
[21] C. Mehlführer and M. Rupp, “Novel tapwise LMMSE channel estimation for
MIMO WCDMA,” Global Telecommunications Conference, 2008.
[22] S. M. Kay, Fundamentals of statistical Signal Processing: Estimation Theory,
vol. 2. Prentice Hall, 1993.
[23] M. Wrulich and M. Rupp, “Computationally eﬃcient MIMO HSDPA systemlevel
evaluation,” EURASIP, WIRELESS COMMUNICATIONS AND NETWORK
ING, August 2009.
[24] M. Wrulich, C. Mecklenbräuker, and J. Wehinger, “Upgrading methodology for
MIMO systemlevel simulations,” in Wireless Evolution Beyond 3G, ftw. C10,
2007.
[25] W. Karner, “Optimum default base station parameter settings for UMTS net
works,” Master’s thesis, Technischen Universit¨at Wien, 2003.
[26] 3GPP, Technical Speciﬁcation Group Radio Access Network; Spatial channel
model for Multiple Input Multiple Output (MIMO) simulations (Tech. Spec.
25.996 V6.1.0), release 6 ed., September 2003.
[27] 3GPP, Technical Speciﬁcation Group Radio Access Network; Spreading and mod
ulation (FDD) (Tech. Spec. 25.213 V7.6.0), release 7 ed., September 2008.
62
Abstract
The increasing demand of bandwidth from new network services has reached mobile technologies. Multipleinput multipleoutput (MIMO) Highspeed downlink packet access (HSDPA) can be the answer to this demand, being able to double the data rate of its predecessor singleinput singleoutput (SISO) HSDPA. The optimization of mobile network is seen as an important issue from network operators, which see the opportunity to reduce costs by optimizing their networks through simulations. This thesis presents an overview of the main characteristics of enhanced MIMO HSDPA, describes the basic MIMO HSDPA linklevel simulator we built and consequently focuses on the common pilot channel (CPICH) and its optimization. Furthermore, we derive a computationally eﬃcient improvement, of the a linkmeasurement model utilized in a MIMO HSDPA systemlevel simulator, to take the eﬀects of CPICH power variation into account.
i
.
but especially for his patience. I felt like at home. 2009 Andreu Mateu Torrelló iii . Palma de Mallorca April 4.Acknowledgements I would like to thank Martin Wrulich. A la meva família. who helped me somehow at some point over the last 5 years. Finally. for his constant support and his optimistic approach. I would also like to thank Professor Markus Rupp for giving me the opportunity to work in such a friendly environment. per donarme suport en els moments més difícils. My thank also goes to the whole institute of communications and radiofrequency engineering and my thesis students colleagues for helping me to clarify my ideas at any time during the day or night. perquè tenirvos al costat tot i que no sigueu a prop fa que em senti amb conﬁança per afrontar nous reptes. I would to thank you. who has guided me through this path called thesis. because without you I never would not have made it.
.
for my mother v .
.
3. . . . . .2. . . . . . . . . . . MIMO HSDPA System Level Simulator 4. . . . . 2.Contents 1. . . . . 1 1 1 2 3 3 4 4 6 6 11 11 12 14 16 17 17 17 18 21 24 25 29 32 35 35 36 36 37 41 . . . . How important is the common pilot channel? . . Multiple Input Multiple Output HSDPA . . . 1. . . . . . . LinkLevel 4. . . . . . . . . . . . . . . . . . . LinkMeasurement Model . DirectSequence Code Division Multiple Access 2. . . . . .2. .3. . . . . .1. . .2. . . . . . . . . . . . . . . . . . . Topics covered in this Master Thesis . . Highspeed downlink shared channel . . . .2. . . . . WCDMA principles . . .2. . . . . . 2. . . .1. .1. . . Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . 1. . .3. . . . . . . 3. . . . . . . . Block Description . . . . . . . . . . . . Basic MIMO HSDPA LinkLevel Simulator 3. . . .2. . . . . MIMO HSDPA . . . . . . . . . . . . . . . . . . . . 3. Implementation issues . . . HSDPA Basics 2. . . . 3. . . Channel model . . .2. . . . . General Structure . .3.1. . . . . . .3. Mobile Network Simulations 4. . . . . . .3. . . . . .2. . . . Channel Estimation . . . . 3. . Spreading and despreading . . . . . HSDPA principles . .2. . . . . . . . . . . . . . . . . . . .1. . . . . . . 4. . . . .2. . 3. . 4. . . . . . . . . . .1. . . . . .2. . . . . . . .3. . . . . . . Introduction 1. . . . . . . HSDPA physical layer operation procedure . . . Standardization .1. . LinkPerformance Model . . . . . . . . . .5. . . . . . Radio propagation . . . . . . . . . . . . . . HSDPA vs Release 99 DCH . . . . . . . . . Equalization . 3. . . . . .4. . . . . . . . . . . . . . . . . . . . . . . .4. .2. . . . . . . Common Pilot Channel & Highspeed Downlink Shared Channel 3. . . 2. . . . . . . . . . .2. . .3. . . . . 2. . . . . . . . . . . . . . . . . SystemLevel Simulations vs. . . . Transmission modes . . . . . . . . .3. . . . . . . .2. 2. . . .2. 2. . .3. . . . . . . . .2. . . . . .2. . . . . 2. . . . . . . . . . . .1. .3. 4. . . . . 2. . . . . . . General structure . . . . . . .2. 3. . . . . . . . . . . . . . . . . . .2. . . . vii . . 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. CPICH Power Optimization 5. . . . . . . . . . . . . . . . . . Scrambling . . .4. . . .3. . Simulation methodology . . . . . . . . . . . . . . . . . . . . . .2. . Importance of CPICH power optimization 5. . . . . . Validation of the current model . . . . . . . . 5. . .1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1. . . . .3. . . . . . . . . .3. . 4. . . 5. . . . . . . . . . . . . . . . . . . . . . . . . . Appendix A. . 5. . . .3. . . . . . . . . . . . . . . . . .3. .2. . . . . . . . . . . . . . . . . Inﬂuence of CPICH in the current model 4. .2. . . . . . A. . . . .1. . . 4. . . . . . . . . . . . . . . . . 5. . . . . . Modeling the eﬀects of CPICH . . . . . . . 4. .3. . . 5.4. . . Preequalization SINR . . . Linkmeasurement model enhancement . . . . . . 42 42 42 44 46 47 47 49 49 51 53 54 57 57 58 . .4. . Channelisation Codes . .2. . . . . .2. . . . . . . . . . .3. . . . . . . . . . . . HSDSCH SINR optimization . . . .2. . A. . Validation of the enhanced model . . . . . . . . . . . . Simulation results . . . . . . . . .1. . Conclusion . . .
. . . . . . . . Overall impulse response .7. . . . . . . . . . MIMO 2x2 DTxAA . . . . . . MSE channel coeﬃcients .2. . . . . . . . . . . .3. 2. . . . . . . . . . . . . . . . . . . . Rayleigh fading for diﬀerent speed . . . 2. . Generic downlink transmitter structure MIMO DTxAA [2] . . . . . . . .1.4.1. . . . 3. . .5. 3. . . .8. . . . . . . Cell resizing eﬀect due to CPICH Path loss ﬁgures . . . . . . . . . . . . . . . 3. . . . . . . . . . . . . . . . SISO .9. . .10. 3.1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 5 7 14 16 18 19 20 21 23 25 26 28 29 32 33 36 38 39 43 44 45 46 48 50 51 52 55 55 56 56 3. . . .11. . . . . . Transmission scheme comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. . . . . . . . 5. . . . . . . . . . . . Relationship between modulation. . . . . . .7. . . . . . .6. . . . .7. . . . . . . . 4.2. . . . . .2. . . . . . . . . . . . . . .4. . . . . . . . .1. . . . . . . . . 5. . 5. . . . . . . . . . . . . . . 5. . .5. . . . . . . . MISO 2x1 . . . . . . . . 4. System Model . . . . . Iall [dBW] of the target sector . . . . Basic MIMO HSDPA linklevel simulator ﬁle structure . .5. . . . . . . . . . . . . . .List of Figures 2. . .5. . . . . . . . . . . .4. . . . . . . . . . . Schematic block diagram of system level simulations [4] MIMO HSDPA systemlevel simulator main structure . . . .8. . . 5. . . . . . . . . . . . . . 4. . . . . . . . . . . . 4. . . . . . . . . . . . . . . . . . . 4. . . . . . .6. . . . . . . . . . . . . . . . . 3. . . . 3. 5. . Comparison of channel estimators . . . . . . . . . .3. . . . . . . . . . . . . . . . . . . . . . . . . . . MMSE equalizer structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3. . . . . . . . . . .2. . . . . . . . . . . . . . . . . Model validation . 5. . Transmission schemes . . . . . . . . . . spreading and scrambling 3. . .4. Frame structure for Common Pilot Channel . 3. . . . Iall Comparison . . . . . . . . . . . . MIMO 2x2 TxAA . . . . Basic linklevel simulator block diagram . [3] . . . . . . . . PDP Pedestrian B . . . . . . .6. . . . . . . . . .3. . . . . 5. Eﬀect of CPICH in the current model and LS estimator Model validation and LS estimator . . . . . . . . . . Spread Spectrum . . . . . . . . . . . . . . 4. . . . 2. . . . . . . . . . . . . . . . . 3. . . . . . . . . power variation . ix . . . . . Modulation constellations . . 3GPP release timeline [1] . . . . . 3. . . . . . . . . . . . . 4. . . . . . . . . . . . Spreading and Despreading . . . . . . . .
. . . . . .1. . Codetree for generation of OVSF codes .A. . 57 . . . . . . . . .
Glossary 3GPP AWGN BLER CPICH CQI DCH DSCDMA DSCH EDCH FACH GSM HARQ HSDSCH HSPDSCH HSSCCH HSDPA ISI ITU LOS LS MAI MIMO MISO 3rd generation parnership project Additive white gaussian noise Block error ratio Common pilot channel Channel quality indicator Dedicated channel Directsequence code division multiple access Downlink shared channel Enhanced DCH Forward access channel Global system for mobile HybridARQ Highspeed downlink shared channel Highspeed physical downlink shared channel Highspeed shared control channel Highspeed downlink packet access Intersymbol interference International telecommunication union Lineofsight Least squares Multiple access interference MultipleInput MultipleOutput Multipleinput and singleoutput xi .
MMSE MNO MRC MSE NLOS OVSF PCPICH PDP RNC RRM Rx SCCPCH SCPICH SF SINR SISO SNR STMMSE TFC TTI Tx UE UMTS WCDMA Minimum mean squared error Mobile network operators Maximum ratio combining Mean square error Nonlineofsight Orthogonal variable spreading factor Primary CPICH Power delay proﬁle Radio network controllers Radio resource management receiver Secondary common control physical channel Secondary CPICH Spreading factor Signaltonoiseandinterference ratio Singleinput singleoutput Signaltonoise ratio Spacetime minimum mean square error Transport format combination Transmission time interval Transmitter User equipment Universal mobile telecommunications system Wideband code division multiple access .
The optimization of MIMO HSDPA is on the mind of any operator wishing to launch this new system.1.2. the CPICH is used in the measurements for the handover and cell selection/reselection. doubles the data capacity and increases voice capacity by three times enabling operators to oﬀer mobile broadband at even lower cost. On the one hand. The variation of the CPICH power produces two main eﬀects. the increasing availability of a broad range of new highspeed data services is fuelling demand for more bandwidth in order to improve the user experience. is one of the main techniques introduced to allow this increase of bandwidth. HighSpeed Downlink Packet Access (HSDPA) was introduced in Release 5 by the third Generation Parnership Project (3GPP) as the evolution of UMTS to give response to the new bandwidth demand. The natural evolution is HSPA+. MultipleInput MultipleOutput (MIMO).1. channel estimation is accomplished through the use of a signaling channel. as well as in UMTS. Introduction 1. How important is the common pilot channel? In HSDPA. The use of CPICH reception level at the terminal for handover 1 . MIMO HSDPA Mobile radio communication represents one of the most persistent growing technology markets since the introduction of the Global System for Mobile communications (GSM). 1. Since then. was introduced in Release 7. because is the element that provides the knowledge of the channel coeﬃcients which are crucial for the equalizer performance. which includes an enhanced version of HSDPA. The Common Pilot CHannel (CPICH) is the signaling channel which aids the channel estimation. The Universal Mobile Telecommunications System (UMTS) was the evolution introducing Wideband Code Division Multiple Access (WCDMA) technology in mobile networks. Channel estimation is an essential part of a mobile system.
since the better the channel coeﬃcients are estimated the higher the equalizer performance. making a comparison on how the data is handled in HSDPA and UMTS. as well as to make their initial access to the network in that cell. 2 . Determining an optimal CPICH power value for MIMO HSDPA networks. by adjusting the CPICH power level.1. the thesis starts in Chapter 2 by presenting the principles of WCDMA and HSDPA. the cell load can be balanced between diﬀerent cells.3. Additional material regarding some speciﬁc parts may be found in the Appendix. In this study we will focus on: 1. 2. Enhancing the linkmeasurement model of the MIMO HSDPA systemlevel simulator in [5] to take the eﬀects of CPICH power variation into consideration. Introduction measurements has the consequence that. Chapter 3 covers in detail the design of a basic MIMO HSDPA linklevel simulator. In Chapter 4 we discuss the importance of the combination of linklevel and systemlevel simulators to provide a comprehensive study of the mobile network technologies and present the systemlevel model improvement. 1. while increasing it invites more terminals to hand over to the cell. the increase of the CPICH power derives better performance of the channel estimator. Reducing the CPICH power causes part of the terminals to hand over to other cells. The CPICH power optimization is treated in Chapter 5. Topics covered in this Master Thesis To give a short overview. On the other hand.
2. called MIMO. 3 . containing the ﬁrst full series of WCDMA speciﬁcations. 2. termed Release 99.2 and 2. Accordingly. HSDPA Basics This chapter introduces the standardization of HSDPA in Section 2. multiantenna transmission and reception technology. as well as fast cell selection. adaptive coding and modulation. In the meantime it became obvious that some improvement for packet access would be needed [1]. MIMO was not included in Release 5 speciﬁcations but later in Release 7 termed HSPA+ which is an enhanced version of HSPA [6]. A feasibility study for HSDPA was started in March 2000. The main topics included physical layer retransmissions. Europe.1. In addition. Release 4 speciﬁcations were issued. The study was initially supported by Motorola and Nokia from the vendor side and BT/Cellnet. TMobil and NTT DoCoMo from the operator side. HSDPA speciﬁcations were published in Release 5 in March 2002. Standardization 3GPP is the forum created at the ends of 1998 by US.1. as shown in Figure 2. BTSbased scheduling.1.3 respectively. At the end of 1999 the ﬁrst release was published. The feasibility study showed that substantial improvement could be reached by the introduction of some of the studied techniques. Korea and Japan as the result of the desire to introduce a new single global standard for mobile communication based on WCDMA technology. The study comprehended a set of improvements to be done over Release 99 speciﬁcation. Two years later. the principles of WCDMA and HSDPA are presented in Sections 2. Fast cell selection was discarded since it was concluded that the complexity introduced would not justify the beneﬁts [1].
especially emphasizing the technology advantages and disadvantages. The following sections describe the technology of the 3rd generation of mobile communications. they transmit at a much lower spectral power density than narrowband transmitters. Spread spectrum transmitters use similar transmit power levels than narrowband transmitters. Spread signals are intentionally made to occupy a much larger bandwidth than the information they are carrying to make them look more noiselike. Because spread spectrum signals are noiselike. However.2.2. These low probability of intercept and antijam features are why the military has used spread spectrum techniques for so many years.2. see Figure 2. Spread spectrum signals are also hard to intercept or demodulate. spread spectrum signals are harder to jam than narrowband signals. with little or no interference. The main idea of this technology is to allow for sending independent information simultaneously over a single communication channel meaning that several users share common bandwidth. WCDMA principles WCDMA is a Wideband DirectSequence Code Division Multiple Access (DSCDMA).2. 2. Further. The user data signal is spread over a wide bandwidth by multiplying the user data with a code sequence of quasirandom symbols (called chips). also known as directsequence spread spectrum. HSDPA Basics 2. DirectSequence Code Division Multiple Access Code Division Multiple Access is a multichannel access method.: 3GPP release timeline [1] 4 . noiselike signals.1. Release 99 12/1999 Release 4 03/2001 Release 5 03/2002 Release 6 12/2004 Release 7 06/2007 2000 2000 2001 2001 2002 2002 2003 2003 2004 2004 2005 2005 2006 2006 2007 2007 Figure 2. Spread spectrum uses wide band. Spread and narrow band signals can occupy the same band.1. they are hard to detect. This capability is the main reason for all the interest in spread spectrum today [7].
The frequency channel concept disappears making frequency planing much easier.2.: Spread Spectrum DSCDMA advantages and disadvantages The use of DSCDMA in cellular communications introduces several improvements and has certain advantages with respect to other multiple access schemes such as TDMA or FDMA. In this case. Power requirements should be strictly controlled as DSCDMA schemes are particularly sensitive to the NearFar problem1 . 4. Let us brieﬂy introduce some of them [7. These interferences only aﬀect certain parts of the spectrum. 8]: 1.2. 2. 3. Despreading is more diﬃcult when 1 UEs may use the same carrier frequency and are distinguished only by the use of diﬀerent spreading codes. DSCDMA systems imply a universal frequency reuse in each cell as all the cells use the entire available bandwidth. as signal spreading implies also the spread of power along all the spectrum). and thus can decode the signal. this narrowband interference will only have some impact to a small set of the signal (and thus. which allows Radio Network Controllers (RNC) to combine diﬀerent base station signals in order to make reception more robust. Nevertheless. Narrowband interferences are now practically harmless. The lack of frequency channels enables User Equipments (UE) to be connected to more than one radio base station. the position of the users becomes relevant i. Due to this. WCDMA principles Original Signal Bandwidth Spread Signal Bandwidth Noise frequency Figure 2. soft handovers can be done as well as macrodiversity techniques. a UE closer to the base station 5 .e.2. it aﬀects only to a relatively small part of the overall power. Because of the signal spreading. Communication privacy is increased because only the transmitter and the receiver know the pseudonoise sequence to despread the signal. CDMA systems have a certain number of disadvantages which are enumerated as follows: 1.
We see that the resulting sequence acquires the same appearance as the spreading code. Also known as radio channel. in other words. The result is a spreaded sequence of the data sequence at a rate of SF × R. may block a large part of the other users farther in the cell. i.e. the original transmitted sequence can be recovered perfectly thanks to the orthogonality of the codes which have a crosscorrelation equal to zero.2. 2. called chips. Spreading codes are chosen to be orthogonal among each other. and the transmitted user data sequence is modulated in BPSK and has a rate R. see Figure 2. At the receiver. HSDPA Basics another station emits at a higher power.3a depicts a spreading example when the chip sequence has length 5. they do not interfere with each other.2. 2. Spreading and despreading The spreading operation is the multiplication of each symbol of a data sequence of rate R with a code sequence of SF symbols. Note that perfect synchronization is presumed [9].1. Despreading restores a bandwidth proportional to R for the signal. The chip rate used in WCDMA is 3. The increase of the data rate by SF corresponds also to an increase of the spectrum occupied by a same factor. 6 .3b. see Appendix A. The Figure 2. the despreading consists of the multiplication of the spread data/chip sequence with the very same SF code chip sequence that was used during the spreading of these symbols. During the despreading in the receiver. This eﬀect happens when the supposed orthogonal coded sequences do not cancel each other perfectly due to multipath propagation. Self jamming might appear. A rigid chipleveled synchronization is needed between transmitter and receiver. this medium is highly hostile compared to the cabled transmission mediums. 2. Due to this a power control system must be implemented [9]. Then the receiver integrates the resulting products for each user symbol. 3. This wideband signal would then be transmitted across a wireless channel to the receiving end.2.2.84 Mcps leading to a carrier bandwidth of approximately 5 MHz.3. the inner product between two codes is zero. Radio propagation Air is the access medium for mobile communications.
This section will deal with these mechanisms and its consequences. we will also focus on the propagation models that have been used in the present study. Besides. Propagation mechanisms in space with objects As covered in the introduction of this section. Diﬀraction is a phenomenon that accounts for energy traveling from transmitter to receiver without a lineofsight path between the two. • Diﬀraction occurs when the radio path between the transmitter and receiver is obstructed by a dense body with large dimensions compared to (λ).2.: Spreading and Despreading Many propagations mechanisms are involved in radio channel transmissions that have eﬀects on the signal. It is often termed shadowing because the 7 .2. • Reﬂection occurs when a propagating electromagnetic wave impinges on a smooth surface with very large dimensions compared to the signal wavelength (λ). The main ones are reﬂection. causing secondary waves to be formed behind the obstructing body.3. diﬀraction and scattering [10]. WCDMA principles symbol chip Sent Data 1 1 1 1 1 1 Spreading code sequence Spread signal (a) Spreading 1 1 1 1 Spreading code sequence Received Data (b) A large subﬁgure Figure 2. radio channel transmissions suﬀer from a series of mechanisms that aﬀect the signal quality.
delays and amplitudes and as a result the signal suﬀers from large and fast power shifts. On the other hand. A change in phase is experienced with the consequent frequency shift. This eﬀect is produced when the wavetransmitter and its receiver are in relative movement. HSDPA Basics diﬀracted ﬁeld can reach the receiver even when shadowed by an impenetrable obstruction. street signs and foliage. causing the reﬂected energy t spread out (scatter) in all directions. 7]. formed from N versions. There are fast and sudden changes on the signal power. the received signal. for an emitted x(t) signal. • Doppler eﬀect. These eﬀects can be grouped as smallscale fading and largescale fading [10]. So. These propagation mechanisms inﬂuence the signal propagation and cause diﬀerent eﬀects such as path loss. • Multipath propagation. will follow the next expression: N y(t) = i ai (t.2) 8 . Fading can also create signal phase shifts (that is.e. delays can cause symbol overlaps.1) where a(t. diﬀraction and dispersion phenomenons. which will be explained further on. This eﬀect is known as Intersymbol Interference (ISI). τi (t) represents the delay and φi (t. τ )x(t − τi (t))ejφi (t. • Scattering occurs when a radio wave impinges on either a large rough surface or any surface whose dimensions are on the order of (λ) or less. Fading is mainly caused by multipath propagation and by the Doppler eﬀect which is caused by changing channel conditions (i. In an urban environment. movement) [8. several signal versions are formed and arrive at the receiver with diﬀerent phases. τ ) is the signal attenuation. typical signal obstructions that yield scattering are lampposts. The resulting signal can be represented as the sum of all the signal versions. changes in the signal space) and signal dispersion in time. which can vary 30 or 40 dBs in only a few seconds or within a few wavelength variations (λ). Smallscale fading Smallscale fading aﬀects the instantaneous signal power and therefore aﬀects the link quality. which is known as the Doppler frequency (fd ) and can be approximated to the following expression: fd = v λ (2. known as echos. shadowing or multipath loss. τ ) the phase diﬀerence.2. Due to the reﬂection.τ ) (2.
It is now called Empirical COSTWalﬁschIkegami Model. Then the channel is expected to remain unchanged during the time in which a symbol is transmitted [10]. we say that the signal suﬀers from fast fading. The model considers the buildings in the vertical plane between the transmitter and the receiver. Not only free space propagation is aﬀected by path loss. where this factor is called free space loss. COST231 WalﬁshIkegami model This empirical model is a combination of the models from J. The shadowing concept involves all the unique characteristics of the scenario which can hamper the communications and include for instance buildings and mountains. Walﬁsch and F. It was further developed by the COST 231 project. The degradation due to this eﬀect can be split in two: fast fading and slow fading. 8]. increase system capacity. Slow fading occurs when the coherence time of the channel is long enough compared to the symbol duration. Largescale fading Largescale fading aﬀects the average signal power. When the coherence time of the channel (which is related with the Doppler frequency within a multiplicative constant) is short compared to the symbol duration.2. The received power expressed in terms of transmitted power is attenuated by a factor Ls (d). WCDMA principles where λ is the wavelength and v is the relative speed between transmitter and receiver. In a fast fading situation. other parameters such as the base station or mobile antenna height or terrain characteristics are inﬂuencing too. These environmental peculiarities increase the complexity of building up theoretical models [7. The following section presents the model used in the present study termed COST231 WalﬁshIkegami model which does not cover shadowing modeling. the channel is expected to change several times while a symbol is propagating causing a distorsion of the baseband pulse. The accuracy of this empirical model is quite high because in urban 9 . mainly caused by freespace path loss and shadowing. When the receiving antenna is isotropic this factor is expressed as: Ls (d) = 4πd λ 2 (2.3) This eﬀect is used in cellular systems because the rapid attenuation with distance makes it feasible to reuse channels and thus.2. Ikegami. resulting in a loss of signaltonoise ratio (SNR) which may lead to an irreducible error rate. It is well known that the power of an airtransmitted electromagnetic wave proportionally decreases with the squared distance to the transmitter.
has applied another streetorientation function than Ikegami.4 + 20 log(d km) + 20 log(f MHz).8) 10 .6 + 25 log(d km) + 20 log(f MHz) for d ≥ 20 m. (2.2. It takes into account the width of the street and its orientation. The heights of buildings and their spatial separations along the direct radio path are modelled by absorbing screens for the determination of Lmsd . please refer to [11] for further details. a simple propagation loss formula diﬀerent from free space loss is applied. (2. The loss is based on measurements performed in the city of Stockholm Lb (dB) = 42.1. The determination of Lrts is mainly based on Ikegami’s model.7) Scalar electromagnetic formulation of multiscreen diﬀraction results in an integral for which Walﬁsch and Bertoni published an approximate solution in the case of base station antenna located above the rooftops. multiple screen diﬀraction loss Lmsd . Only the main equations are explain below. and rooftoptostreet diﬀraction and scatter loss Lrts .5) L0 for Lrts + Lmsd ≤ 0 The freespace loss is given by L0 (dB) = 32. L0 + Lrts + Lmsd for Lrts + Lmsd > 0 Lb = (2.4) In the NLOScase the basic transmission loss is composed of the terms free space loss L0 . Lrts = −16. The main parameters of the model are listed in Table 2. however. The model distinguishes between lineofsight (LOS) and nonlineofsight (NLOS) situations. COST 231. HSDPA Basics environments especially the propagation over the rooftops (multiple diﬀractions) is the most dominant part.9 − 10 log(w m) + 10 log(f MHz) + 20 log(∆hMobile m) + LOri . This model is extended by COST 231 for base station antenna heights below the rooftop levels using an empirical function based on measurements. Lmsd = Lbsh + ka + kd log(d km) + kf log(f MHz) − 9 log(b m) (2.6) The term Lrts describes the coupling of the wave propagating along the multiplescreen path into the street where the mobile station is located. In the LOS case. between base and mobile antennas within a street canyon. (2.
Table 2.3. and retransmissions are handled in the RNC. In HSDPA. in practice only 384 Kbps. Dedicated CHannel (DCH) and Downlink Shared CHannel (DSCH). First a comparison between Release 99 and HSDPA is performed and then HSDPA key aspects are presented. • FACH. This channel is used to transport small data volumes or connection set ups during state transfers. The last one has been replaced in Release 5 for the new HighSpeed Downlink Shared CHannel (HSDSCH) and therefore will not be analyzed in this section. FACH cannot be used. and it is explained in more detail below[1]. but in Release 6 there is an alternative with the use of an enhanced version of DCH (EDCH). 2. FACH does not support fast power control or soft handover. The Secondary Common Control Physical CHannel (SCCPCH) is the responsible to carry its content and the used spreading code is ﬁxed.1.2000 MHz 4 . it is used to carry signalling information when a terminal has changed its state.3.2 shows the main diﬀerences between DCH and HSDSCH channels. DCH can carry any kind of service using a ﬁxed spreading code and ﬁxed allocation time. HSDPA principles Parameters Frequency f Height of the transmitter hT X Height of the receiver hRX Distance d between transmitter and receiver Restrictions 800 . 11 . • DCH. When circuitswitched service is demanded it runs always on DCH.3. The key part of Release 99 and Release 5 is always operated together with HSDPA.the key new features included in Release 5.: Parameters COST231 WalﬁshIkegami model 2.5 km Table 2.2. HSDPA principles This section covers HSDPA principles for WCDMA .1. It supports the use of soft handover.50 m 13m 0. although these parameters can be changed from upper layers. If there is a need to carry mixed services. The theoretical maximum peak rate is 2 Mbps. HSDPA vs Release 99 DCH Three diﬀerent methods for data packet transmission are speciﬁed in Release 99: Forward Access CHannel (FACH). HSDPA has been designed to increase downlink packet data throughput of release 99 by means of fast physical layer retransmission and transmission combining as fast link adaptation controlled by the base station.02 . In Release 5 the uplink user data always goes through the DCH. 6 and 7 speciﬁcations that are relevant for this study.
• Lack of physical layer control information on the HSPDSCH.2. In release 99 only 4QAM was available. the last release standardizes 64QAM. The use of the new short allocation period implicitly implies a more dynamic nature of the system. The support of multiple channelisation codes is called codemultiplex. therefore the UEs will be able to support up to 15 codes because common channels and associated with DCHs need one of them. • Lack of soft handover. The DCH also allows the use of the fast power control feature. 12 . Data are sent from only one serving HSDSCH cell.2. which are described in [1]. extended Yes Yes Table 2. • Only spreading factor 16 is used. with the addition of fast physical layer signaling. are as follows: • Lack of closedloop power control or so called fast power control. 2. • The scheduling is done on a 2ms basis. The key diﬀerences between Release 99 DCHbased packet data operation. Highspeed downlink shared channel The logical transport channel. With DCH the Transmission Time Interval (TTI) could be as long as 80ms and not shorter than 10ms. HSDPA Basics Feature Variable spreading factor Fast power control Adaptive modulation and coding Multicode operation Physical layer retransmissions BTSbased scheduling and link adaptation DCH Yes Yes No Yes No No HSDSCH No No Yes Yes. 16QAM and 4QAM for the downlink. is the HSDSCH and is mapped physical channels named HSPDSCH. and modulation. which carries the actual user data in HSDPA.: Comparison DCH and HSDSCH [1] meaning that an UE can be connected with more than one station at a time and receiving information from all of them.3. link adaptation selects the suitable combination of codes. • Support of higher order modulation than DCH. This is carried on the Highspeed shared control channel (HSSCCH) for HSDPA use and on the associated DCH.2. coding rates. In HSDPA.
13 . When the physical retransmissions fail. ’Chase combining’ or ’incremental redundancy’ [1]. There are some changes introduced in the channel coding chain due to the use of new modulation schemes. HSDSCH link adaptation As covered in the section 2. HSDSCH coding The use of turbocoding outperforms convolutional codes. The relative number of parity bits to systematic bits varies between retransmissions. these are handled by the RNC like in Release 99. Also a bit scrambling functionality is introduced to avoid having long sequences of ’1s’ or ’0s’. • In Chase combining. The HARQ functionality consists of twostage rate matching functionality which allows tuning the redundancy version of diﬀerent retransmissions when using nonidentical retransmissions. The receiver stores the received samples as soft values. the base station will also decide every 2 ms which coding and modulation combination to transmit.2 the use of 2ms TTIs allows the system a great dynamic. HARQ can operate in two modes. HSDPA principles • Only turbocoding is used. to ensure good signal properties for demodulation. the rate matching is identical between transmissions so the same bit sequence is sent. therefore HSDPA presents a restriction on the use of convolutional codes and just turbocoding will be used from now on.3.3. • Implements HybridARQ (HARQ) which can operate in two modes: ’Chase combining’ and ’incremental redundancy’. The rate matching function is varied between diﬀerent retransmissions and in the actual implementation channel encoding can be done for each transmission or the data can then be kept in the virtual buﬀer. This solution requires more memory in the receiver. and therefore the memory consumption is higher than if it was storing hard values. Apart from the scheduling decisions. • Incremental redundancy uses a diﬀerent rate matching between retransmissions.2. with DCH also convolutional code could be used.
4. the so called HSPA+.3. To avoid the nearfar problem. link adaptation takes the extra power that results from being too close to the base station and uses it to select the transmission parameters in such way that the required symbol energy corresponds more accurately to the available symbol power.: Modulation constellations 64QAM Link adaptation is based on physical Channel Quality Indicator (CQI). uses only 4QAM modulation. DCH. HSPA+ 14 . and therefore the signal quality needs to be better when for example 16QAM is used instead of 4QAM. The dynamic range obtained using this technique can reach 30dB according to [1]. introduces 64QAM on the downlink which increases the data rates by 50% and 16QAM on the uplink. Release 5 and 6 oﬀer additional support for higher modulation order on the downlink: 16QAM. A goodquality CPICH allows the estimation of the optimum channel without userspeciﬁc pilot overhead [1.3. The higher order modulation the higher the number of bits that can be carried per symbol. These as well as the receiver antennas are usually uniformly separated by distances close to the size of a wavelength.4. HSDPA Basics 4QAM 16QAM Figure 2. HSDSCH modulation The transport channel associated with R99.2. which is the feedback provided by the UE. Multiple Input Multiple Output HSDPA MIMO systems enable an increase of the throughput without having to increase the bandwidth nor the transmitted power. The constellations are shown in Figure 2. But higher order modulation introduces more complicated decision boundaries. 12]. 2. The same carrier frequency is used for all the transmitted antennas. MIMO HSDPA was introduced in Release 7 within what is called HSPA+ or HSPA evolved [6]. Moreover Release 7.
The key beneﬁts of MIMO systems compared to SISO systems summarized in [13] are: • Antenna grouping gain. In addition. In other words. • Spatial multiplexing gain. The weights wt are selected to maximize the signaltonoiseandinterference ratio (SINR) at the UE. There are two modes deﬁned: TxAA. If the fading in each of the MIMO channels are uncorrelated. the channel matrix has to be of suﬃcient rank. The processing in the transmitter and the receiver increases the mean SNR received. This gain can also be enhanced by the capacity to transmit independent data signals in each of the transmission antennas. Thus. The standardized MIMO transmission scheme in Release 7 is termed Adaptative Antenna Array (see Figure 2. for this case. where MIMO gains are limited. in this case one stream is transmitted over both antennas. so that there are enough uncorrelated paths. knowledge of one of the two array weight vectors will completely determine the other. Note that when DTxAA is used both data streams are transmitted on the same orthogonal spreading code(s). either one data stream or two data streams are transmitted depending on the terminal SINR conditions. then better results will be obtained in detection. MIMO also needs a rich scattering environment to keep the two data stream orthogonal when they reach the UE.3. Therefore MIMO beneﬁts can be seen in dense urban areas where the size of the cells is typically small. Thus. The weight selection is signaled by the UE to the base station.e. Higher order modulation complements MIMO by providing signiﬁcant gains in lineofsight scenarios. the better the system will work. • Diversity gain. ensuring a successful decoding of the signal in spite of distributing the power among the transmit antennas. it should be noted that in the MIMO case the second data stream is only turned on at high SINR conditions. DTxAA transmits the data streams using orthonormal array weight vectors drawn from the closed loop transmit diversity codebook. this achieves code reuse.2. the more uncorrelated paths. The signal power in a wireless communication suﬀers from random fading. two separate data streams are transmitted on two orthogonal weight sets simulat(k) neously.5). MIMO performs most eﬀectively when the SNR at the UE is high. Because they are orthogonal. HSDPA principles supports 2x2 downlink MIMO with up to two antennas at the transmitter and the receiver. i. and DTxAA or dualstream TxAA. MIMO channels provide an increase of the system capacity without the need to increase the bandwidth nor the transmitted power. A diversity gain is obtained compared to SISO systems when there is just one path. 15 .
• HSSCCH is transmitted two slots before the corresponding HSDSCH TTI. HSDPA Basics CPICH 1 w1 primary transport block + + HSDSCH TrCH processing w2 spread/scramble w3 secondary transport block HSDSCH TrCH processing + w4 CPICH 2 + Figure 2. HSDPA physical layer operation procedure This section presents the HSDPA physical layer basic steps once at least one user have been conﬁgured to use HSDSCH and the data is already present at the buﬀer of the base station [1]. Once the Part 1 of HSSCCH is decoded it will start to decode the Part 2 and will buﬀer the necessary codes from HSDSCH. so schedulers are an ongoing research topic at the moment. 16 . • Every 2ms the base station evaluates each user in order to schedule which users will be served in the future. • When decoding Part 2 the UE discovers the ARQ process. • After decoding. • When a UE is selected as served in a determined TTI. the base station identiﬁes the HSDSCH parameters needed for the transmission.2. depending on the result of the CRC over the HSDSCH data.: Generic downlink transmitter structure MIMO DTxAA [2] 2. the UE sends in the uplink direction an ACK/NACK indicator after the combination of the data (if applied).3. and can then determine whether there is the need of combining or not.5. the modulation order and the UE limitations. The selection criteria is not set in the standard. This is because it has two diﬀerent parts and Part 1 carries information needed to decode the frame HSDSCH properly.4. including the number of codes. • The terminal monitors the HSSCCHs (there are up to four).
covering the generation of the CPICH and the HSDSCH.1 are presented in detail in Section 3. The length of a radio frame is 10 ms and is composed of 15 slots. which in our case is the base station due to the study of the downlink (HSDPA).3.2. Finally.3. the last one is used to estimate the channel for the data channel. the channel.1 contains the general description of the simulation model. there are three diﬀerent parts: the transmitter. 3. as cited before in Section 2.2. implementation issues are discussed in Section 3. All these elements are explained in detail in Section 3. The 17 . i. a MIMO frequencyselective channel plus Gaussian white noise. This is followed by a more detailed study of each component in Section 3.3. Block Description The dotted blocks in Figure 3. The bit sequences in the simulator ﬂow encapsulated in packets of 2 ms.1.e.2. thus a TTI is formed by 3 slots.2 this is the TTI in HSDPA systems.2 we discuss the transmitter schemes supported in the simulator and introduce the concept of beamforming.2.2.1. 3. block fading is utilized. For sake of simplicity and convenience the channel is constant during the duration of a TTI. Basically. The outline is as follows. Section 3.1. General structure The general structure of our basic link level simulator is depicted in Figure 3. and ﬁnally the receiver or user terminal. like in any communication system. Basic MIMO HSDPA LinkLevel Simulator This chapter presents the description of the implemented simulator used in this study to model a basic MIMO HSDPA link. The Basic MIMO HSDPA linklevel simulator emulates the physical transmission of the data channel (HSDSCH) and the pilot channel (CPICH). In Section 3.
: Basic linklevel simulator block diagram channel model is covered in Section 3. so it does not change the signal bandwidth but only makes the signals from diﬀerent sources separable from each other [9]. This is followed by the receiver analysis in 3. and therefore increasing the symbol rate by the length of the channelisation code or spreading factor. Once bits come out of the modulation mapper in shape of symbols. Scrambling is used on top of spreading.2.1.2.2.4. respectively.3. As can be seen. Figure 3. CSF.3.5.1 and Appendix A. This is needed to separate UEs or base stations from each other.c .2. the spreading and the scrambling of the sequences. Basic MIMO HSDPA LinkLevel Simulator Figure 3. To understand how the channelisation and scrambling codes. 3. spreading the signal.2 shows the relation between each other for a HSDSCH stream taking into account that more than one bit stream is transmitted per stream. The functionality and characteristics of the scrambling and channelisation codes for the HSDSCH are summarised in Table 3. (k) encoded ratematched bits. There are three modulation schemes supported in enhanced HSDPA [12]. go through the modulation mapper in groups of the length according to the modulation selected. are generated. bn [m]. Finally the schemes of channel estimators implemented in the simulator are discussed in Section 3. In addition to spreading.2. 18 . as commented in [14]. part of the process in the transmitter is the scrambling operation. these are multiplied by the corresponding channelisation code. Our simulation has support for all of them. Cscr .1. emphasizing the new MMSE receiver preferred in HSDPA. Common Pilot Channel & Highspeed Downlink Shared Channel Before describing how the channels are created it is needed to understand the relation between the modulation. refer to Appendix A.1.
In case of MIMO transmission. and therefore the term CPICH will be used for PCPICH in the following. increases bandwidth transmission Scrambling code Separation of sectors (cells) 2 ms = 7680 chips Long 10 ms code: Gold code Short code: Extended S(2) code family No.1. 19 . 16QAM. When no diversity is used.3 shows the structure of the CPICH as well as the bit sequence assigned to this channel. spreading and scrambling [3] CPICH Figure 3. Although there are two types of CPICH.2. which is broadcasted to the entire cell.: Relationship between modulation. only the primary (PCPICH) is implemented in the simulator. Occasionally the SCPICH can be found implemented for serving dedicated hotspot areas. 16QAM.1 (k) b1 [m] 4QAM.N1 bSF1 [m] (k) 4QAM. 64QAM (k) sy [n] SF1 Figure 3. Block Description Channelisation code Separation of connections to diﬀerent users within one cell 16 chips Orthogonal Variable Spreading Factor Yes.2 b2 [m] (k) 4QAM. does not aﬀect transmission bandwidth Usage Length Code family Spreading Table 3. 16QAM. the CPICH has to be transmitted from both antennas using the same spreading factor and scrambling code even though the predeﬁned bit sequence of the CPICH is diﬀerent for Antenna 1 and Antenna 2. This study focus on the optimization of the PCPICH power. 64QAM (k) sy 1 [n] C16. 64QAM (k) sy 2 [n] (k) ssp [i] Cscr [i] (k) s [i] C16. Antenna 1 transmits just the sequence assigned to it.2. Normally each cell has only the PCPICH. the implementation of the secondary (SCPICH)has no interest in this study.: Functionality of the channelisation and scrambling codes [9] C16.3.
Particularly the Orthogonal Variable Spreading Factor (OVSF)used is the C256. (See Appendix A. and 4QAM modulation it implies 20 bits per slot. 20 bits 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 0000 1 radio frame: Tf = 10 ms Slot 1 Slot 2 Slot 15 0011 1100 0011 1100 0011 1100 0011 1100 0011 1100 0011 1100 0011 1100 0011 Transmitter Antenna 2 Figure 3.3.3. Basic MIMO HSDPA LinkLevel Simulator Transmitter Antenna 1 Tslot = 2560 chips. SF is the spreading factor used. • CPICH is broadcast over the whole cell. M is the modulation order.3. and 2560 chips per slot.0 .1) shows this relation 2560 × M SF Nr = (3. • The modulation scheme used is always 4QAM.: Frame structure for Common Pilot Channel The main features of the CPICH are the following [3]: • The same channelisation code is always used and has length 256. Using the chip rate (3. HSDSCH For the data channel.1) where Nr denotes the number of bits to be generated.1) • The CPICH is scrambled by the primary scrambling code. the bit sequence is generated randomly with a length according to the spreading factor and modulation selected so that the number of chips transmitted during a TTI keeps constant. 20 . The equation (3. • The bit rate is 30 kbps. as a result of the use of spreading factor of 256. • There is only one CPICH per cell. See Figure 3. the number of transmitted chips per TTI can be easily derived and the result is 7680 chips.84 Mcps) and the TTI (2 ms).
here the origin of the constant. There are 15 channelisation codes available from up to 16.3. when respectively either only one transmitter and one receiver antenna or two transmitter and one receiver antenna are used. • The supported modulation schemes are 4QAM.3. one of them is reserved for the CPICH. and MIMO 2x2 DTxAA.: Transmission schemes A packet contains 3 slots.2) (2) v [i] (1) s [i] (1) p [i] h (1) (1. The last one is standardized for the non MIMO case only. MISO 2x1.1) v [i] r [i] (1) (1) p [i] s [i] (1) (1) h (1. which are used in multiple antenna cases.2. Detailed information concerning MIMO HSDPA is presented in Section 2.3. • The bit rate is variable depending on the modulation.4): SISO.2. Block Description (1) w1 p [i] s [i] (1) (1) h (1. Singleinput singleoutput SISO(Figure 3.1) v [i] (1) r [i] (1) h (1) (2. The standardized precoding vectors. Transmission modes The simulator can cope with up to four diﬀerent transmissions schemes (Figure 3. then the number of chips transmitted per TTI is equal to 2560 chips.4b) are degenerate cases of MIMO (Figure 3. Release 7 includes the speciﬁcation for the MIMO case limited to the two antenna scheme. 16QAM and 64QAM.1) (1) w1 (2) h p [i] (2) (1. TxAA and DTxAA Figure 3. 3.4a) and multipleinput and singleoutput MISO (Figure 3.1) (a) SISO w1 (1) w2 p [i] h (1.4c). MIMO 2x2 TxAA.2.2) v [i] (1) r [i] w2 w2 dual stream mode (2) (b) MISO (c) MIMO.4.2) (2) r [i] (1) (2) s [i] h (2. are deﬁned as follows: 21 . The main characteristics of the HSDSCH channel as implemented in the simulator are: • The channelisation code has always length 16.
2 as bn with n ∈ 1. each data stream has a capacity of 15 bit streams. which is the matrix containing the precoding vectors.4 also reﬂects the fact that no precoding is applied to the CPICH. . (k) Each bit stream is denoted in the Figure 3. Hence. when DTxAA is the selected mode. Selection of the best beamforming The selection of the best beamforming is done according to [15]. so that maximize the received power becomes argmax WH RW W: W 2 (3.. and therefore the power allocated to the data channel will be split equally among the number of streams.2) =1 The optimal weight vector is the dominant eigenvector of R. We always optimize over the ﬁrst stream in the DTxAA case. The frequencyselective channel Hn is modeled as an array of length L×NT associated to the receiver antenna n. dbf .3. 2 2 2 2 Figure 3. and therefore choose the best beamforming we deﬁne R := NR HH Hn . a variable amount of power could be allocated to each stream in order to maximize capacity. The precoding matrix is selected in the transmitter in concordance with a past channel determined by a delay. The number of active streams in the case of our simulator is handled manually. Then the problem of n n=1 choosing the weights W. this is not implemented in our simulator. as many as 30 diﬀerent bit streams can be transmitted for a particular user. denoted by p(t) . Even though in general. . one of them is reserved for CPICH. This is done because in a real system the best beamforming is selected in the UE and signaled to the base station which implies a certain delay. For this reason.SF − 1 and k is the stream index. because the extra complexity did not have substantial beneﬁts for the purpose of this study. The spreading factor (SF ) used for the data channel is SF = 16.. Basic MIMO HSDPA LinkLevel Simulator √ (1) (2) w1 = w1 = 1/ 2 w2 = −w2 w2 ∈ (1) (1) (2) 1 + j 1 − j −1 + j −1 − j . In order to maximize the received power. 22 .
3.2. Block Description
30 25 20 SINR HS−DSCH [dB] 15 10 5 0 −5 −10 −15 −20 30 25 20 15 10 5 noise power [dBW] 0 −5 −10 SISO MISO 2x1 MIMO 2x2 TxAA stream 1 MIMO 2x2 D−TxAA stream 2 MIMO 2x2 D−TxAA
Figure 3.5.: Transmission scheme comparison
Transmission schemes comparison The HSDSCH SINR is increased by the use of MIMO just as expected. This is depicted in Figure 3.5. The SINR in the case of MIMO DTxAA is evaluated per stream. Note that in these simulations full knowledge of the channel is assumed in the equalizer. The HSDSCH SINR is the SINR observed at the demapper input, so after despreading (k) and is obtained as follows: Consider the transmitted data symbol vector syn of the kth stream and n spreading code (see Figure 3.2) and the corresponding received (k) symbol vector at the demapper input ˆyn . The post equalization SINR then is given s by SINRk,n = syn γ (k)
(k) syn γ (k) (k) 2 2 (k) 2 − ˆyn s 2
(3.3)
where γ (k) represents the attenuation of the overall impulse response at delay τ and is given by
2
γ
(k)
=
f
(k)
ˆ H
ˆ (τ +(k−1)(Lf +Lh −1)) Hw
2
.
(3.4)
ˆw where f (k) are the equalization ﬁlters for each kth stream and Hj denotes the column ˆ w. j of the matrix H
23
3. Basic MIMO HSDPA LinkLevel Simulator
3.2.3. Channel model
The channel presents two types of fading eﬀects: largescale and smallscale fading, these eﬀects have been presented in Section 2.2.3. In our simulator we deal with both of them. The interference power in the cell is modeled as a white Gausian noise and the channel is normalized to 1.
Smallscale fading Smallscale fading are produced by changing channel conditions. The received signal is made up of multiple reﬂective rays, therefore the envelope amplitude due to smallscale fading has a Rayleigh probability density function [10] given by,
r exp − r2 2 2σ 2 p(r) = σ
r≥0 otherwise
(3.5)
0
where r is the envelope amplitude of the received signal, and 2σ 2 is the mean power of the multipath signal. Because of this sometimes the smallscale fading is also called Rayleigh fading. These interferences are generated as described in [16]. The channel will be fast fading or slow fading depending on the doppler frequency, fd . As we have discussed before our simulator uses block fading, i.e. the channel is kept constant during the transmission of a packet (7680 chips or 2 ms). This limits our simulator so that the coherence time of the channel, To , must be larger than 2 ms. If the relative speed between the base station and the UE is v = 3km/h then To ≈ 32 ms but when the speed turns to be v = 50 km/h then To ≈ 2 ms. This means that our simulator is limited to deal with relative speeds lower than 50 km/h. The attenuation introduced by the rayleigh fading for diﬀerent velocities can be seen in 3.6. In order to deﬁne the diﬀerent number of propagation paths, we use the ITU channel model Pedestrian B [17], as speciﬁed in the standard [18]. Figure 3.7 shows the Power Delay Proﬁle (PDP) of Pedestrian B model, which determines the length of the channel (Lh ). The maximum excess delay time, Tm , is also determined by the PDP. In this case, Tm = 3700ns. We can determine the duration of a chip, Tc , from the chip rate R = 3.84Mcps, which leads to a Tc = 260.41ns. Therefore the channel is frequency selective (Tm > Tc ) and such multipath dispersion yields ISI. The equalization described in the next section mitigates these eﬀects.
24
3.2. Block Description
10
1
Rayleigh fading v=3km/h
10
1
Rayleigh fading v=10km/h
Signal Level Attenuation (dB)
10
−1
Signal Level Attenuation (dB)
10
0
10
0
10
−1
10
−2
10
−2
10
−3
0
200
400
600
800
1000 Samples
1200
1400
1600
1800
2000
10
−3
0
200
400
600
800
1000 Samples
1200
1400
1600
1800
2000
10
1
Rayleigh fading v=30km/h
10
1
Rayleigh fading v=120km/h
10
−1
Signal Level Attenuation (dB)
Signal Level Attenuation (dB)
10
0
10
0
10
−1
10
−2
10
−2
10
−3
0
200
400
600
800
1000 Samples
1200
1400
1600
1800
2000
10
−3
0
200
400
600
800
1000 Samples
1200
1400
1600
1800
2000
Figure 3.6.: Rayleigh fading for diﬀerent speed
3.2.4. Equalization
RAKE is the conventional receiver for WCDMA which approximately implements a matched ﬁlter for the channel impulse response. The receive signal ri is ﬁrst delayed and then descrambled, despread and combined to Maximum Ratio Combining (MRC), and the coeﬃcients of the tap delay line are found through channel estimation. HSDPA uses WCDMA for multiuser communication and thus orthogonal spreading codes are used to separate diﬀerent users in the downlink. However, the orthogonality of these codes is destroyed by the multipath characteristics of the channel resulting in Multiple Access Interference (MAI). The performance of the MMSE equalizer is compared to the RAKE receiver in [14] proving that signiﬁcant performance gain is obtained with the use of equalizers and that when these are used the overall system is no longer interference limited. The MMSE equalizer implemented in our simulator is based on the papers [19, 20].
25
(1) T T (K) T si = si .6) where Lh and Lf are the length of the channel impulse response and the equalizer length. .: PDP Pedestrian B System model In order to derive an MMSE equalizer we present the system model as derived in [19]. (3. (3. Basic MIMO HSDPA LinkLevel Simulator Power Delay Profile − Pedestrian B 0 −5 −10 dB −15 −20 −25 0 500 1000 1500 2000 ns 2500 3000 3500 4000 Figure 3.7) pi = pi (1) T .7. pi T (NT ) T . The kth spread and scrambled chip stream at time instant i is deﬁned as (k) si T = s (k) [i]. 26 . .8) The frequency selective channel between the nt th transmit and the nr th receive antenna is modeled by the Lf ×(Lh +Lf −1) dimensional band matrix. The chip streams are weighted by the complex precoding coeﬃ(k) (k) cients w1 and w2 where the subindex indicates the transmit antenna. in case of DTxAA and the CPICH vectors which depend on the number of transmit antennas. . . Afterwards (1) (2) the pilot sequences pi and pi are added at the resulting sequences in each antenna. . . si . . (3. . respectively. s (k) [i − Lh − Lf + 2] . By stacking the transmitted stream vectors which can be up to K = 2. . . .3. .
. . How the selection of this precoding matrix is done is explained in Section 3. . (3.9) 0 where hi r t represent the channel impulse response of the nt th transmit antenna to the nr th receive antenna..2.9) (n . W= w1 ..13) 27 . .10) By stacking the received signal vectors of all NR receive antennas we have ri = ri (1) T .1) .NT ) .2. . (3. . .nt ) H = . (3. . w2 (1) (K) (3.2. w1 (1) (K) w2 . Block Description (nr . . H= .. . (n . ..n ) hLhr−1t . the full frequency selective MIMO channel is modeled by a block matrix H consisting of NR ×NT band matrices deﬁned in (3. (3.nt ) h0 .1) H(NR . Then.NT ) H(1.13) containing the precoding values of each stream (up to K=2). This is modeled as an additive white gaussian noise (AWGN) with variance 2 σv We can obtain a compact system description of received symbols ri = H(W ⊗ ILh +Lf −1 )si + Hpi + vi = Hw si + Hpi + vi .n ) 0 (n .11) In addition to the channel eﬀects a noise vi term is added representing the thermal noise in the receiver and the interference power received from other base stations. The matrix W is deﬁned in (3.. (n .n ) H(1. ⊗ denotes the Kronecker product. . (nr .3.12) Here. .n ) h0 r t hLhr−1t . . .. H(NR . ri (NR ) T .
15) T T . (3.14) where ˆi = Hw si + vi . . .8. . .1)H (1) ŝ [i] f (nr .19) 28 .k) r .16) that deﬁnes NR equalization ﬁlters for each kth stream.2)H (2) ŝ [i] dual stream mode Figure 3.17) The minimization of the cost function is performed by deriving (3. fLf−1 (n . Basic MIMO HSDPA LinkLevel Simulator (1) v [i] (1) r [i] f (nr .8. . K.2(Lh +Lf −1) . see Figure 3.k) . the vector eτk . (3.k) T (3.3.: MMSE equalizer structure STMMSE equalizer The Spacetime minimum mean square error (STMMSE) equalizer coeﬃcients can be calculated by minimizing the distance between the equalizer chip stream and the transmitted chip stream through the following quadratic cost function [20] J f (k) = E f (k) H ˆi − si−τ r (1) 2 .2(Lh +Lf −1) is a zero vector of length 2(Lh + Lf − 1) with a single one at position τk = τ + (k − 1)(Lh + Lf − 1) k = 1 .18) Here.14) with respect to f (k) ∗ obtaining [20]: −1 2 f (k) = σs Hw Rss HH + Rvv w Hw eτk . (3. (3. r and f (k) = f (1. . (3. . Each ﬁlter is deﬁned by f (nr . f (NR . .k) T . .k) = f0 (nr . .
3.9 0.9. Note that because we assume that the noise vector vi is AWGN 2 with variance σv and mean µ = 0. the covariance matrix and the correlation matrix 2 are the same and we can write Rvv = σv I.: Overall impulse response The variable τ speciﬁes the delay of the equalized signal and due to causality must fulﬁll the condition: τ ≥ Lh .2.3. p(nt ) [i + Nc − 1] sp sp sp . The matrices Rss and Rvv are the signal and noise correlation matrices.5. 22].4 0. It represents the ideal case when the noise term is assumed to be zero.7 0. System Model In this case the spread pilot chip sequence pi are deﬁned as follows (nt ) and the spread data sequence si (nk ) T t p(ni ) = p(nt ) [i] . . We also assume that the data signals of 2 2 the users are uncorrelated and therefore Rss = σs I where σs is the energy allocated to kth stream. respectively. Figure 3.6 0. . which is described in Section 3. Both of them are based on [21. thus 47 chips. The length of the impulse response is the sum of the length of the channel in this case is 18 chips because of the use of Pedestrian B. . 3.3 0.20) 29 .2 0. Channel Estimation Our simulator utilizes two channel estimators: Least Squares Estimator and the Correlationbased Estimator. Block Description 1 0.2. and the span of the equalizer which is 30 chips minus 1 chip.9 shows an exemplary overall impulse response of the equalizer and channel.2.5 0.1 0 0 5 10 15 20 25 30 35 40 45 Figure 3. (3.8 0.
. . . p(NiT ) . . . . spi sp Sspi = s(1) . Consequently. . The combined transmit signal for all NT is modeled like this..27) Once we have deﬁned our channel and the signal transmit in each antenna.NR ) .22) with the matrix Cscri = diag[cscri . . . . xi = Cscri (Pspi + Sspi ). The precoding of the data sequence is omitted by sake of simplicity and without lost of generality.N ) hk R T we can deﬁne a stack matrix of the channel matrices like this T H = HT . HTh −1 0 L . hk (1.1) hk T (N .26) (N . Hk = . . .. . (3. note that the precoding could be done before the scrambling without alteration of the ﬁnal transmitted sequence.21) where i is the time index in chips and Nc is the number of chips considered for the channel estimation..24) Xi = xi .. . .23) (3. . Pspi = p(1) . . The MIMO channel at the delay k (k = 0. and therefore 7680 chips. (3. . . .3. . NT (3. . we can express the receive signal of all receive antennas as 30 . In our case Nc will have the length of a packet. cscri+Nc −1 ] comprising the Nc chips of the base station’s scrambling sequence at the main diagonal.25) The deﬁnition of the channel in this section has been adapted according to [21]. . . . . . Lh − 1) is deﬁned as h k. spi sp (1) (NT ) (3. . (1.. s(NiT ) .1) . Basic MIMO HSDPA LinkLevel Simulator T k s(ni ) sp = s(nk ) [i] . . . . . (3. (3. s(nk ) [i sp sp + Nc − 1] . the scrambled transmit signal at antenna nt is given by xi (nt ) t t = Cscri p(ni ) + s(ni ) sp sp nt = 1 . . .
. . Cscri−1 Pspi−1 . the LS estimator for the system description (3.30) ˜ Ni = k=0 Cscri−k Sspi−k Hk + Ni (3. .32) Correlationbased Estimator The correlator estimator is a simpliﬁed case of the LS estimator. Cscri−Lh +1 Pspi−L and Lh −1 h +1 ] (3.28) The matrix Ni includes the interference of other base stations as well as the thermal noise at the receiver.33) Where the term p transmit antenna. Block Description Ri = ri . .28) as ˜ ˜ Ri = Pi H + Ni (3. . The approximation and therefore the correlationbased estimator is given by ˆ H(cor) = 1 ˜H Pi Ri p 2 2 (3. 2 2 denotes the energy of the CPICH sequence transmitted at one 31 . From the model presented we can rewrite the equation (3.29) is given by −1 ˆ ˜i ˜ H(LS) = PH Pi ˜i PH R i (3. . .3. It tries to avoid −1 the calculation of the matrix ˜ ˜ PH Pi i .2. .31) Then. Least Squares Estimator Despite of being a low complexity channel estimator the Least Squares (LS) channel estimator has a reasonable good performance.29) where ˜ Pi = [Cscri Pspi . ri (1) (NR ) Lh −1 = k=0 Cscri−k (Pspi−k + Sspi−k )Hk + Ni (3.
The structure of the simulator is contained in hsdpa_link_body ﬁle. The mean square error (MSE) of both channel estimators for a SISO case and a noise power of 4dBW is depicted in ﬁgure 3. 32 . The main ﬁle structure of the simulator is depicted in Figure 3. which has the parameters that can be set by the user for the chosen simulation.10.11 .3. The tools provided by Matlab are very powerful and useful for the purpose of this study. 3. it oﬀers the possibility of being upgraded with more simulations. because some of the functions such as convolution and ﬁltering are already implemented. Implementation issues The Basic MIMO HSDPA linklevel simulator is implemented in Matlab.10. they look very much the same over a large range of values. Each simulation ﬁle simu_xxx has associated a load_parameters_xxx ﬁle. E_HSDPA_link level_simulator ﬁle contains the diﬀerent simulations which has been used in this study. but it can also be observed that the larger the power allocated to the common pilot channel the larger the diﬀerence between the channel estimator MSE. Basic MIMO HSDPA LinkLevel Simulator 8 7 6 MSE [dB] 5 4 3 2 MSE H (LS) MSE H (cor) 2 4 6 8 10 12 CPICH Ec/Ior [dB] 14 16 18 20 1 0 Figure 3.3.: Comparison of channel estimators Comparison of channel estimators Correlationbased estimator is a simpliﬁed version of the LS estimator. therefore a similar result is expected.
m hsdpa_link_body load_parameters_zzz.11.m load_parameters_yyy.m simu_yyy.3.m simu_zzz.m simu_xxx. Implementation issues E_HSDPA_link_level_simulator.: Basic MIMO HSDPA linklevel simulator ﬁle structure 33 .m load_parameters_xxx.3.m Figure 3.
.
For this reason. Mobile Network Simulations This chapter presents the importance of systemlevel simulators and computational eﬃcient modeling.2 we present a computationally eﬃcient MIMO HSDPA systemlevel simulator showing its main structure and focusing in the linkmeasurement model. Hence. coding design. predicted link level performance gains can be obtained in an entire network. Finally in Section 4. we propose a model improvement to take into consideration the eﬀects of channel estimation in the linkmeasurement model. multiple cells and therefore base stations are taken into consideration. Mobile Network Operators (MNO) are aware that an optimization of their networks increases the performance and reduces costs. Linklevel simulators aim to study the physical characteristics of the link and sometimes some MAC functionalities such as development of receiver algorithms. the diﬀerences between linklevel and systemlevel simulators are covered in Section 4.3. Cell planning.1. it is important to identify whether. a comprehensive study of a mobile network technology can not stick just to the way how one UE performs with the base station. LinkLevel Simulations Linklevel simulators are commonly used to study the behavior of transmission and reception schemes. this kind of simulators just study one link. 5]. SystemLevel Simulations vs. 4. In Section 4. As the name indicates. scheduling and multiuser and multi base station interference are some of the investigations that systemlevel simulators try to cover. and so on.4. In this study the implemented basic MIMO HSDPA linklevel simulator (see Chapter 3) is used to enhance the characteristics of the MIMO HSDPA systemlevel simulator in [23. 35 . First. where multiple users. This type of simulators have to rely on simpliﬁed link models that still must be accurate enough to capture the essential behavior due to complexity reasons because the computational cost of evaluating a whole network with the use of linklevel simulators is prohibitive[23]. feedback strategies.1. and to which amount. The goal of systemlevel simulators is to evaluate the performance of a whole network (or part of it). in this case we are talking about the link between one UE and one base station.
shadowing. and a link performance model. which models the measurement used for link adaptation and resource allocation.1. The following sections present the main structure of the MIMO HSDPA Systemlevel simulator and how these two models are implemented. 4. thus reducing signiﬁcantly the computational eﬀort. 36 . This systemlevel simulator presented shows a structure that identiﬁes the relevant interference terms and allows for the generation of scalar fading parameters prior to systemlevel simulation. and a Minimum Mean Squared Error (MMSE) equalizer. Both models are related in the sense that they provide ﬁgures for performance prediction and can be referred to as system level interface [23]. e. General Structure In [5] a computationally eﬃcient linktosystem level model is proposed and its embedding in a Matlabbased systemlevel simulator which includes as features MIMO with DTxAA.1 depicts a schematic diagram of a basic dynamic systemlevel simulator.4.: Schematic block diagram of system level simulations [4] 4.1. thus during the runtime of the system level simulation only scalar multiplications are needed to compute the SINR.2. Generally two kind of models are required: a link measurement model. which determines the BLock Error Ratio (BLER) given a certain resource and power allocation as well as signal processing. cell throughput PER traffic model QoS requirements Figure 4. Utilizing this special structure nearly all linkdedicated procedures can be included in these fading parameters.2. MIMO HSDPA System Level Simulator Figure 4. fast fading) pilot signaling feedback strategy intercellinterference characteristics link performance model PHY processing/ power allocation link adaption strategy resource scheduling strategy resource allocation figures of merit. Mobile Network Simulations channel quality information link measurement model channel characteristics (pathloss.g.
the 37 . Only a short reference to the linkperformance model is done.2. 0 ≤ n ≤ N − 1.2. but more information can be found in [23]. The channel model is represented using an abstract model. A signaling delay is imposed. and therefore there will be 4 precoding weights. System model Figure 4. in fact. This model is needed to provide appropriate estimates of the channel quality. . we explain shortly the main parts and let the reader refer to [23] for ﬁnding more detailed information. . ϕn (multicode usage) and scrambling sequences. in particular it carries out the scheduling trying to balance the throughput and fairness as well as the Transport Format Combination (TFC) decision based the UE feedback. At the receiver. User speciﬁc algorithms and feedback decisions are done here. It can be decomposed in the main network elements [23]: • Node B. s0 . respectively. . • SL model. wnT N . Measurement results not only depend on channel and intercell interference. . Accordingly the input data stream s is demultiplexed into N parallel streams. • delay. . . w1 . In this section we present the system model utilized in [23] to derive the computationally eﬃcient linkmeasurement model. sN −1 with the individual data streams sn . . MIMO HSDPA System Level Simulator The structure of the simulator is depicted in Figure 4. being spread by a number of spreading sequences. In addition. These spread and scrambled sequences are then mapped to the nT transmit antennas using a preﬁltering matrix D ∈ CnT ×N .4. which contains the precoding weights. even though it is capable of support diﬀerent transmission schemes and receivers. It represents all network related procedures.2. . such as estimated SINR used for channel dependent scheduling and link adaptation. • UE. so that it needs lower computational complexity. the evaluation of the transmission success is performed. 4.2. In this work the maximal amount of streams and transmitter antennas is ﬁxed to 2. LinkMeasurement Model This model reﬂects base station and terminal measurements. In following sections we will focus on the linkmeasurement model since the improvements in this study will be done on this model. they also depend on the measurement phase as for example transmission power and beamforming weights.3 depicts the model where the transmitter (Tx) and the receiver (Rx) are equipped with nT and nR antennas.
4.1 (0) . . . . with a total length of L chip intervals.: MIMO HSDPA systemlevel simulator main structure signals are gathered with nR antennas and chip spaced sampled before they enter the discrete time STMMSE equalizer.nT (0) . . . . . .nT (L − 1) (4. . SINR Node B output macro pathloss UE position update UE position update antenna gain SINRCQI mapping SINRCQI mapping mapping generation mapping generation shadow fading HARQ model HARQ model get ff parameters SINR (slot) stream decision stream decision SINR averaging MI mapping UE output ff generation BLER evaluation BLER evaluation BLER data BLER data avg.. h1. . hnR .nT (L − 1) . . . Note that the pulse shaping the transmit and receive ﬁltering.2. . . h1. SINR per stream Figure 4. . as well as the sampling 38 . . Mobile Network Simulations Node B Structure MIMO HSDPA systemlevel simulator delayed UE output buffer scheduler stream decision QCI/TFC decision power management transmission settings Node B MAChs SL model UE delay SLmodel Node B output UE avg. hnR .t (l) denotes the lth sampled chip of the channel impulse response from transmit antenna t to receive antenna r.1 (0) . h1. The MIMO channel H ∈ CnR ×nT L is modeled as timediscrete. frequencyselective channel. .1) where the entry hr. . . H= . hnR .nT (0) . .
. These parameters can be computed oﬄine and loaded for the runtime of a systemlevel simulation. . . . . an equivalent discrete channel Γ ∈ CnR ×nT L is deﬁned that includes the preﬁltering matrix D and the MIMO channel H. Equivalent Fading Parameteres Description In this section we just present the conclusions obtained in [23]. i. the user u = 0 and the basestation b = 0 are the one deﬁned of interest. . . ynR (k)]T . The receive power is decomposed into diﬀerent interference terms to derive the system level model. . thus signiﬁcantly reducing the computational burden. .: System Model operation can be incorporated in the MIMO channel matrix. . formulated by means of the equivalent channel matrix. With this. . and ⊗ being the Kronocker product.b = [γu. First of all we decomposed the equalizer coeﬃcients Wd = [w1 . . wN ]T N (E+L−1)−1 0 m and the channel Γu.3.b ] where n denotes the stream index and m is the index of the Tx chips for all streams entering the equalizer span. . Desired Signal: The desired power signal is given by T dN Ps. 1. . . Γ = H · (IL ⊗ D). . . . .n = wn γ00 +n 2 · Pn. wn .e. nnR (k)]T . . With this decomposition it is possible to describe the characteristics of the individual terms by means of fadingparameters that are real valued scalar processes. . is given by y(k) = Γx(k) + n(k). .2) where we introduced the receive vector y(k) = [y1 (k).b . . . γu. (4.ζ = Gs. γu.2.14). . . the transmit vector x(k) = [x0 (k).ζ (4. . For sake of notational simplicity. with IL denoting the identity matrix of size L. xN −1 (k − L + 1)]T and the receive noise vector n(k) = [n1 (k). MIMO HSDPA System Level Simulator x0 1 22 27 ϕ0 y1 x0 ∧ * ϕ0 D s H STMMSE DEMUX s0 s0 ∧ sN1 ϕN1 xN1 nT sN1 xN1 ∧ ∧ yn * ϕN1 R Figure 4. . Without losing generality.n · Pn. .b . .4. .3) 39 . xN −1 (k). the input output relation at time instant k. . The STMMSE solution can be computed again according to Equation (3.
ζ denotes the power on stream n and spreading code ζ spent for user u = 0 by basestation b = 0.n T dN wn γ00 +m 2 denotes the intrastream orthogonality factor.dN +N −1] / 2 (4. 4. The following equation represents a simpliﬁcation of all this interferences.4) where the intracell orthogonality.n .m · Pm. U0 Pintra.n · m=0 m=n oINT. Intracell Interference: The intracell interference is composed by a number of terms. Gs.n · Gs. Intrastream Interference: The interference generated by the parallel transmission of a second stream is given by N −1 PINT. the intracell interference from users that are not served in the same instant as the user of interest but with the same instant as the user of interest but with the same scrambling and spreading code. ointra.n = m=0 m=n 2 T dN wn γ00 +m N −1 · Pm.ζ .n . is deﬁned as 1 · N N (E+L−1)−1 m=0 m∈[dN.5) and the precoding orthogonality. (4.b 1 T w N n N (E+L−1)−1 m m ∗ γ0b (γ0b )H wn = m=0 B Ptot.b. the remaining ISI after equalization.n 1 1 Gs.e. (4. (4.n . (4. and that all substreams designated for one user are equally powered.b Ginter.n .7) where oINT. Intercell Interference: This interference term assumes that all users in a neighbouring cell will apply the same precoding coeﬃcient.4. the intercode interference when the same scrambling but a diﬀerent spreading code is used.n = b=1 Ptot.n Ω ω∈Ω N (E+L−1)−1 T m wn γω0 m=0 m∈[dN. oBF.dN +N −1] / T dN wn γ00 +n T m wn γ00 2 2 ointra.8) b=1 40 . is given by oBF.n = P0 + oBF.n · u=1 Pu · ointra.6) 3. and the intracell interference from users with the same scrambling but diﬀerent spreading code. Having considered this the intercell interference is given by B Pinter. 2.ζ = Gs. Mobile Network Simulations where Pn.n . i.n describes the equivalent fading of the useful signal power.
n + Pintra.n Gs.2. Such a model can be viewed as a conditional probability.n = N −1 Pnon−spread oINT.2. 41 . The power (on symbol level) of the thermal noise can denoted as W =E T wn nE (k) 2 2 2 T = σn wn 2 2 . beamforming weights and interference plus noise during the interleaving period [24].ζ = .9) Inﬂuence of NonData Channels The intereference caused by nonspread channels is given by Pns. MIMO HSDPA System Level Simulator 5.n (4.n ·Pm.n = Pother−spread ointra.ζ for the case DTxAA.n Gs.n + Pinter. LinkPerformance Model The link performance model provides an estimate of the link performance when the decisions on the radio resource management (RRM). (4.d. can be evaluated to be Ps.i.n + Ps. can be expressed by SF · Ps.n + W where SF denotes the spreading factor and PINT.n SINRn. since there is only one interfering parallel substream. Tx power.m + Pnon−spread ointra Gs. for example retransmissions and aﬀect also slow link adaption.3.n 1 + N m=0 m=n (4. the probability that the transmitted code word is decoded erroneously given the channel.n + Pns. Thermal Noise: The thermal noise is modeled as i. white Gaussian noise.n simpliﬁes to PINT = oINT ·Gs. (4. scheduling and link adaptation are already known.10) and the interference caused by other spread channels than HSDSCH assuming the total power of other spread channels in the cell to be Pother−spread . as observed after equalization and despreading. with identical power on all antenna and over all chips that enter the equalizer. which in turn can generate.12) SF · PINT.4.11) Resulting SINR Description The SINR on substream n and spreading code ζ. taking into consideration all the exposed above. 4. This model generates either a bit or a block error probabilities and is utilized for quality estimation at runtime.
4.3. 4.3). The main simulation parameters can be seen in Table 4.3. for each transmission scheme.4.1. Figure 4.: SINR validation simulation settings 4.1. The intercell interference and the interference due to the inﬂuence of nondata channels has been omitted since they are not implemented in the basic linklevel simulator. and therefore its link with the power assigned to the common pilot channel. deﬁned as 42 . An improvement of the current model is presented. Validation of the current model In order to validate the linkmeasurement model presented in [23]. and shows the fact that there is the eﬀect of the channel estimation missing.5a depicts the MSE of the channel coeﬃcients. Mobile Network Simulations Parameter power allocated at HSDSCH channel model equalizer length Lf equalizer delay τ number of spreading factors per stream channel estimator Value 20 W ITU PedB 30 chips 15 chips 15 full knowledge of the channel Table 4. The genie SINR is the SINR observed at the demapper input.1. Figure 4.4 depicts the performed simulations to validate the proposed SINR description. Notice that full knowledge of the channel is assumed.3. so after de despreading and is obtained as deﬁned in (3. It can be observed that the linkmeasurement model shows a very good agreement with the genie or true SINR.2. which extends the capabilities of the MIMO HSDPA systemlevel simulator. Linkmeasurement model enhancement This section presents a validation of the current linkmeasurement model. Inﬂuence of CPICH in the current model The CPICH is a very important channel since the channel estimation in the UE is performed according to this. an adaptation of the model has been implemented in the basic MIMO HSDPA linklevel simulator presented in Chapter 3.
Linkmeasurement model enhancement 20 15 10 SINRHS−DSCH [dB] 5 0 −5 −10 −15 −20 30 SINRHS−DSCH [dB] genie approximation 20 15 10 5 0 −5 −10 −15 −20 30 genie approximation 25 20 15 10 5 noise power [dBW] 0 −5 −10 25 20 15 10 5 noise power [dBW] 0 −5 −10 (d) SISO 30 25 20 SINRHS−DSCH [dB] SINRHS−DSCH [dB] 15 10 5 0 −5 −10 −15 30 genie approximation 20 15 10 5 0 −5 −10 −15 −20 30 (e) MISO 2x1 stream 1 genie stream 1 approximation stream 2 genie stream 2 approximation 25 20 15 10 5 noise power [dBW] 0 −5 −10 25 20 15 10 5 noise power [dBW] 0 −5 −10 (f) MIMO 2x2 TxAA (g) MIMO 2x2 DTxAA Figure 4. where in each case. described in Section 4. show the MSE for the rest of schemes. It is also important to notice that the curves are unique for a certain CPICH SINR or CPICHEc /I0 .6 depicts the eﬀect.2. The consequence is a poor performance of the approximation when the CPICH is considered. note that the MSE increases with the channel complexity. the lower curve denotes a dependence with the MSE 43 . The result is obvious. The linkmeasurement model.4.2. Figures 4.5d. Figure 4. a smaller MSE is obtained when the power allocated to the CPICH is higher.5b4.3. does not take into account this eﬀect and supposes a perfect knowledge of the channel when performs the equalization.4.13) ˆ where h are the estimated channel coeﬃcients using the LS estimator or the correlationbased estimator.: Model validation ˆ MSE(h) = E ˆ h−h 2 (4. which is a situation far from the reality.
.3. 44 .4. . (4. the UE would accomplish the equalization using the channel estimated coeﬃcients provided by the channel estimator.14) where the vector eτk .2(Lh +Lf −1) is a zero vector of length 2(Lh + Lf − 1) with a single one at position τk = τ + (k − 1)(Lh + Lf − 1) k = 1 . again for a particular CPICHEc /I0 and channel estimator algorithm. (4. In the case of our linklevel simulator.: MSE channel coeﬃcients of the channel coeﬃcients.2(Lh +Lf −1) . in a real system.15) ˆ and Hw ˆ H(W ⊗ ILh +Lf −1 ). K. Modeling the eﬀects of CPICH In principle. 4.3.5. Mobile Network Simulations 8 MSE H (LS) MSE H (cor) 14 13 12 MSE H (LS) MSE H (cor) 7 6 11 10 9 8 7 6 MSE [dB] 4 3 2 5 1 0 MSE [dB] 5 10 20 30 40 50 60 CPICH EcIor [%] 70 80 90 100 4 0 10 20 30 40 50 60 CPICH EcIor [%] 70 80 90 100 (a) SISO 16 15 14 13 12 11 10 9 8 7 6 0 10 20 30 40 50 60 CPICH EcIor [%] 70 80 90 100 MSE H (LS) MSE H (cor) 15 14 13 12 MSE [dB] 11 10 9 8 7 6 (b) MISO 2x1 MSE H (LS) MSE H (cor) MSE [dB] 0 10 20 30 40 50 60 CPICH EcIor [%] 70 80 90 100 (c) MIMO 2x2 TxAA (d) MIMO 2x2 DTxAA Figure 4. the equalizer shows the following structure −1 2 ˆ ˆw f (k) = σs Hw Rss HH + Rvv ˆ Hw eτk .
3. Thus.6.: Eﬀect of CPICH in the current model and LS estimator The solution of including a channel estimator in the calculation of the fading parameters is not desired.16) The matrix H∆ is constructed like the channel matrix H (in 3.d. ˆ Our proposal is based on considering the estimated channel matrix H as the true channel matrix H and a matrix H∆ representing the channel estimation error. Linkmeasurement model enhancement (a) SISO (b) MISO 2x1 (c) MIMO 2x2 TxAA (d) MIMO 2x2 DTxAA Figure 4. This can be express as ˆ H = H + H∆ .1 if we use the structure of the linkmeasurement model). Gaussian with 0 mean and a variance equal to the MSE of the channel estimator. CPICHEc /I0 ) (4.i. that as it has been shown before depends on the noise term and the estimator itself. The nonzero elements of H∆ are assumed to be a i. we can say that Var{H∆ } = f (estimator algorithm. (4.10. or in 4. Since we do not want to increase the computational complexity.4. we have to ﬁnd out how to incorporate the estimated channel without directly utilizing channel estimators.17) 45 . for the reason that it would increase the computational complexity of the systemlevel simulator.
: Model validation and LS estimator Hence we can say that the STMMSE equalizer in (4.4.4.7. 46 . becomes −1 2 f (k) = σs (Hw +Hw∆ )Rss (Hw +Hw∆ )H +Rvv (Hw +Hw∆ )eτk .3. for diﬀerent CPICHEc /I0 . Validation of the enhanced model We have conducted a set of validations of the new model.18) This solution enables the possibility to have a set of precalculated MSE curves which can be used to model the eﬀects of the CPICH channel in a systemlevel simulator without increasing the computational complexity in excess because only a scalar product and a sum of matrix have to be performed during the generation of the fading parameters.7. (4. it can be seen that the ﬁtting is very good. the results can be seen in Figure 4.2(Lh +Lf −1) . Mobile Network Simulations 2 0 −2 SINRHS−DSCH [dB] −4 −6 −8 −10 −12 0 genie new approximation 10 20 30 40 50 60 CPICH Ec/Ior [%] 70 80 90 100 (a) SISO (b) MISO 2x1 (c) MIMO 2x2 TxAA (d) MIMO 2x2 DTxAA Figure 4.14). 4. Although the matrix Hw∆ depends on the variance of a random variable.
The simulation methodology utilized which leads to obtain the optimization of the CPICH power is explained in Section 5. It is a ﬁxed rate downlink physical channel that carries a predeﬁned bit sequence. the Ec /I0 at the mobile terminal must exceed a minimum threshold at all times. PTX. due to loss of orthogonality between channelisation codes. Apart from being utilized for the channel estimation in the HSDSCH.5.i i=1 Lp i CPICHEc /I0 = . The UEs scan for the CPICH signals continuously and measure the received level of chip energy to interference plus noise density ratio (Ec /I0 ) of all pilot signals they can detect. In order to keep a mobile referenced to a cell. cell reselection and handover procedures. The Ec /I0 ratio can be described as PCPICH Lp numBS PTX.i is the total transmit power by the base station i. N0 is the thermal noise of the UE and numBS is the number of base stations in the network [25].2.1) + Iintra + N0 where PCPICH is the CPICH power of the cell we are measuring. Importance of CPICH power optimization In HSDPA.1. as in UMTS. channel estimation is accomplished through the use of a signaling channel. the CPICH also provides the channel estimation reference for common channels. including signal and interference. I0 denotes the total received power density. 5. and the optimization of the power allocated to this channel.1 contains the analysis of the CPICH showing its main eﬀects due to power variation. Iintra is the intracell interference. Finally the simulation results are presented in Section 5.3. 47 . CPICH Power Optimization This chapter covers the analysis of the CPICH. as measured at the UE antenna connector. The outline is as follows: Section 5. The CPICH is this signaling channel. The measurements of the CPICH are used in the cell selection. Lpi is the overall path loss to base station i. (5. Lp is the overall path loss to the best server.
while increasing it invites more terminals to hand over to the cell. Hearing equal powered CPICH signals or multiple strong CPICH with their multipath components causes pilot pollution.1.5d shows the MSE of the channel coeﬃcients for a MIMO 2x2 TxAA link. poor voice quality and/or data throughput. R 9 9 ) Othe r C ha nne ls Othe r C ha nne ls C P IC H C P IC H Figure 5. If a mobile terminal is unable to clearly receive one dominant CPICH signal.5. in other words. Figure 4. failed initiations. Nevertheless. so that not enough power is left for traﬃc channels. Another limiting factor. the less power is left to serve the user traﬃc.: Cell resizing eﬀect due to CPICH power variation The cell with the highest CPICH level at the UE is selected as the serving cell. which we do not take into account for our analysis. One of the possible consequences of pilot pollution is 48 . Even considering the hypothetical case of having inﬁnite power in both channels. as well as to make their initial access to the network in that cell.1. It can be observed that the MSE increases when the amount of power allocated to the CPICH decreases. Excessive pilot power can easily take too large proportion of the total available transmit power. the increase of the pilot channel would be beneﬁcial just before reaching an upper limit. Reducing the CPICH power causes part of the terminals to hand over to other cells. the more power is spent for pilot signalling. the less power assigned to the pilot channel the worse estimation of the channel. As a consequence by adjusting the CPICH power level. the result is likely to be dropped calls. R 9 9 ) Da ta C ha nne ls (HS DP A. One of the limiting factors is the loss of orthogonality between spreading codes. that is produced due to too high values of CPICH power is called pilot pollution. CPICH Power Optimization Da ta C ha nne ls (HS DP A. due to interference or coverage problems and therefore the terminal can not perform an eﬀective estimation of the channel. This eﬀect is depicted in Figure 5. the cell load can be balanced between diﬀerent cells.
this value will be only optimal for this particular user. Simulation methodology To ascertain the optimal CPICH power in a network we have used a basic network simulator (Section 5. the simulator. increasing the signaling load in the network by causing additional cell reselection messages. Consider one UE with a particular Iall associated.2. is designed to calculate a preequalization SINR of set of points in a deﬁned area for a given base station transmission power of the sector i.996 urban micro. and a set of parameters that deﬁne a uniform HSDPA network. In any case. PTX.1) that provides the overall intercell interference plus thermal noise in the UE (Iall ) for a particular scenario. 5. Table 5.2.i . 5.s .1. 49 .1) when Iintra is set to 0.1 shows the parameters that have been used in the simulations. so the process is repeated for a uniform grid inside one sector for the purpose of obtaining a uniform number of maximums inside a sector which allows us to calculate the optimal CPICH power of the cell by averaging.: Simulation scenario that a UE may change the best serving cell continuously due to slow faded CPICH signals.2. Preequalization SINR In fact. which we use to obtain the overall interference term (Iall ). In our case this SINR in particular spot inside the deﬁned area corresponds to Equation (5. this overall interference term is then introduced in the basic MIMO HSDPA linklevel simulator (Section 3) and with it we derive the CPICH power value that maximizes the HSDSCH SINR.5.1. from now on we will refer to this preequalization SINR as CPICHEc /I0 . COST231 WalﬁshIkegami 19 3 2 GHz 9 dB Table 5. This Iall depends basically on the scenario parameters and the position of the UE inside the cell.2. Simulation methodology Parameter total power available at Base Station distance between Base Stations antenna pattern path loss model number of cells number of Sectors/cell transmitter frequency receiver noise ﬁgure Value 20 W 1000 m 3gpp 25.
2. N0 is the thermal noise at the receiver.2b) calculated using COST231 WalﬁshIkegami model for urban micro environments (see Section 2. PTX. The radiated power of the target sector. 50 .2d.s = PTX.s . Lip is the overall link loss between the base station at cell i and the user at the position p.3) where lip is the path loss (Figure 5. for a particular spot is described as CPICHEc /I0 . as it is calculated in the simulator.2b) which depends on the antenna pattern deﬁned in [26]. The resulting overall path loss for the deﬁned area can be seen in Figure 5.2) + N0 where s is the spot index.2c. can be seen in Figure 5.5.k k=1 Lkp (5.i Lp numBS PTX.: Path loss ﬁgures The CPICHEc /I0 .3) and Gip is the antenna gain (Figure 5.i Lp . and it is calculated as follows Lip = lip /Gip (5. CPICH Power Optimization 1500 1000 500 0 −500 −1000 −1500 −2000 −1500 −1000 −500 0 500 distance in m 1000 1500 2000 1500 1000 500 0 120 −500 −1000 −1500 −2000 −1500 −1000 −500 0 500 distance in m 1000 1500 2000 110 100 90 160 150 140 130 12 10 8 distance in m 4 2 0 −2 −4 −6 (a) Antenna Pattern [dB] 1500 1000 500 0 −500 −1000 −1500 distance in m 6 (b) Path loss [dB] −40 −50 −60 distance in m −70 −80 −90 −100 −110 −120 1000 1500 2000 −2000 −1500 −1000 −500 0 500 distance in m (c) Overall path loss [dB] 1500 1000 500 0 −500 −70 −1000 −80 −1500 −2000 −1500 −1000 −500 0 500 1000 1500 2000 −30 −40 −50 −60 (d) Radiated power [dBm] (e) Overall interference [dB] (f) preequalized SNR [dBm] Figure 5.2.
5.: Iall [dBW] of the target sector The sum of the intercell interference and thermal noise can be seen in Figure 5.4) TX.3.opt Ec Ior CPICH (5.2. HSDSCH SINR optimization Our approach considers the optimization of the power allocated to the CPICH when the SINR of the data channel (HSDSCH) is maximal. it means that the path loss term must be included in the interference term.2f. Simulation methodology 14 250 200 150 100 distance in m 50 0 −50 −100 −150 −200 −250 0 100 200 300 400 distance in m 500 600 −2 12 10 8 6 4 2 0 Figure 5.2e where each dot represents one base station and the target base station is situated at the position (0.5.2) and becomes Iall = Lp · (Iinter + N0 ) P (5.s CPICH.5) 51 .2.0). Figure 5. The resulting CPICHEc /I0 in the target sector for each spot is depicted in Figure 5. The noise term in this case includes the intercell interference and the thermal noise in the receiver. The channel in the basic MIMO HSDPA linklevel simulator is normalized.k where Iinter = numBS Lkp . Once this have been considered the Iall for a speciﬁc spot that will be inserted in the linklevel simulator can be easily derived from (5.3 shows the total number of spots that have k=1 been considered inside the target cell. Thus the cost function is Ec Ior = argmax SINRHS−DSCH. Every cross has a diﬀerent Iall that will be entered in the linklevel simulator.2.
Ec 3. Insert the (Iall ) of a particular spot (see Figure 5. each one is deﬁned for a diﬀerent Iall .: Iall Comparison where SINRHS−DSCH. 4. and its Ior associated.5. Repeat the process for every spot inside the target sector.4 shows an example CPICH with two functions in SISO transmission. we proceed to average the optimal Ec Ior CPICH Ec Ior CPICH values associfor the target 52 .3).s as a function of Ior . into the basic MIMO HSDPA linklevel simulator. Figure 5. CPICH Power Optimization Figure 5. in other words.s . Once the set of maximal SINRHS−DSCH. Ec 2.s of the function.s is the SINR in the data channel and the subindex s indicates Ec the spot position and therefore the Iall associated. Deﬁne SINRHS−DSCH. and vice versa. Ior is the power allocated CPICH to the CPICH as a fraction of the total transmission power Ior . thus the optimal ated to them are found. It is important to emphasize that the base station transmitted power remains constant over the curve. CPICH Note that the maximal of the function also changes with the change of the overall interference term. Find the maximal SINRHS−DSCH. which has previously been acquired with the use of the preequalization SINR simulator. The procedure that leads to obtain the optimal point is the following: 1. an increase of the CPICH transmitted power infers a decrease of the HSDSCH transmitted power.4.
5 . and the parameters concerning the network are deﬁned according to Table 5.s s=0 Ec Ior CPICH (5.4. for each transmission scheme supported by our basic MIMO HSDPA linklevel simulator. MISO 2x1. MIMO TxAA and MIMO DTxAA. where the slope is more pronounced. 5. This extra complexity of the channel has a negative eﬀect on the channel estimation which needs more help in terms of power to perform better.3. which is the range closer to the minimal CPICH power. again.1. Figures 5. 53 .6) In the case of double stream. We have performed the simulation for Pedestrian A and Pedestrian B channel model. we see that both results are quite similar and decide to obtain the optimal by averaging both streams optimal values. this is due to the diﬀerence in complexity of the channel. which leads to an increase of the CPICH power to obtain the maximal data SINR. this eﬀect is due to the increase in complexity of the channel. respectively. Nevertheless. The optimal CPICH power when Pedestrian B is utilized is higher than when Pedestrian A is used.8 show the optimal CPICH transmission power for each spot according to the scenario speciﬁed for the SISO. Pedestrian B (18 chips) is a very large channel when compared to Pedestrian A (3 chips). Thus the selection of the optimal CPICH power is actually only critical in a particular range. the optimization is done independently per stream. However. the maximal data SINR is similar over a large range of values (See Figure 5. Note that an increase of the CPICH power is required when the number of antennas increases. Simulation results In this section we present the results of the optimal CPICH power that maximizes SINRHS−DSCH .3. The conﬁguration of the linklevel simulator is done according to Table 5. In the case of Pedestrian B the equalizer perform worse.2. Suppose we have S spots in the target sector.3 sums up the results. Table 5. Simulation results sector as follows. in the single stream case the optimal value is originated from S Optimal Ec Ior = CPICH 1 S argmax SINRHS−DSCH. for a Pedestrian B channel model and considering in the case of double stream the averaging of the results.5.5.
40 W 4. MISO. 2x2 500 30 chips 15 chips 2 GHz LS Estimator ITU PedA.: linklevel simulation parameters Transmission scheme SISO MISO 2x1 MIMO 2x2 TxAA MIMO 2x2 DTxAA PedA Optimal CPICH 4.86 W 5.5. 54 .53 W 8.93 W 8. Conclusion The optimal CPICH power obtain in this chapter is far from the one used in a real network. The main reason is that pilot pollution is not considered.00 W 7.4. CPICH Power Optimization Parameter total power available at Base Station Ior supported Transmission schemes antenna utilization number of transmitted packets equalizer span Lf equalizer delay τ transmitter frequency channel estimator channel model user mobility UE receiver type beamforming delay dbf utilized spreading codes for HSDSCH HSDSCH modulation Value 20 W SISO.: CPICH power optimization results 5.2. in any case the cost function utilized shows the eﬀect of the CPICH power variation in each transmission scheme and sets the basis to develop new studies able to ﬁt the real networks more accurately. usually set to 510% of the total transmission power of the base station. PedB 3 km/h STMMSE 4 packets 15 4QAM Table 5.3.16 W 4.78 W PedB Optimal CPICH 7.81 W Table 5. MIMO TxAA & DTxAA 2x1.
6.: MISO 2x1 55 . Conclusion Figure 5.4.5.5.: SISO Figure 5.
8.: MIMO 2x2 TxAA Figure 5.5.7.: MIMO 2x2 DTxAA 56 . CPICH Power Optimization Figure 5.
1.1. The generation method for the channelisation code is deﬁned as: Cch.4.1) Cch.0 Cch.0 = (1) Cch.0 1 1 = = Cch.2 = (1.4.1.4. 0 ≤ k ≤ SF − 1.2.A. Each level in the code tree deﬁnes channelisation codes of length SF.1.4.0 = (1.1.0 = 1. Cch. The OVSF codes can be deﬁned using the code tree of Figure A.1.1) Cch.1 = (1.1) Cch.1.1) Cch.SF.1 [27].1.0 −Cch.1 Cch.3= (1.1 = (1. corresponding to a spreading factor of SF in Figure A.1.1) Cch.: Codetree for generation of OVSF codes In Figure A.1.2. where SF is the spreading factor of the code and k is the code number.1. Appendix A.2.1.1.1.1.1.0 Cch.2.1) Figure A. the channelisation codes are uniquely described as Cch. Cch.1.1.0 1 −1 57 . Channelisation Codes The utilized channelisation codes are OVSF codes that preserve the orthogonality between a user’s diﬀerent physical channels.0 = (1.k .
2(n+1) . . . .. . n .2n −1 −Cch. . The resulting sequences thus constitute segments of a set of Gold sequences.1 . and zn . 2. n = 0.3 . . 1. 218 − 20. let x(i). 218 − 20. . i = 0. . • y(0) = y(1) = .2(n+1) 2(n+1) −2 Cch. x(1) = x(2) = . . Cch.0 Cch.2n −1 The leftmost value in each channelisation code word corresponds to the chip transmitted ﬁrst in time.0 Cch.0 −Cch. . Furthermore. . 58 . . Each of the two real sequences are constructed as the position wise modulo 2 sum of 38400 chip segments of two binary msequences generated by means of two generator polynomials of degree 18. The scrambling codes are repeated for every 10 ms radio frame. in the sequel.2 Cch. A. y(i) and zn (i) denote the i:th symbol of the sequence x.2n .1 Cch. y.2n .2n .2(n+1) . The x sequence is constructed using the primitive (over GF(2)) polynomial 1+X 7 +X 18 . respectively. 218 − 2.2 Cch. . . • y(i + 18) = y(i + 10) + y(i + 7) + y(i + 5) + y(i) mod 2. The sequence depending on the chosen scrambling code number n is denoted zn .2n .1 −Cch. . The n:th Gold code sequence zn ..A. . 218 −2. Appendix Cch.2n . = x(16) = x(17) = 0. n . Scrambling The scrambling code sequences are constructed by combining two real sequences into a complex sequence [27].2(n+1) 2(n+1) −1 Cch. . . i = 0. Let x and y be the two sequences respectively.2n . . . . .1 Cch. = y(16) = y(17) = 1 Recursive deﬁnition of subsequent symbols: • x(i + 18) = x(i + 7) + x(i) mod 2. .2(n+1) .2 Cch.2n −1 Cch. .2n −1 Cch. .2n . The msequences x and y are constructed as: Initial conditions: • x is constructed with x(0) = 1. .2. is then deﬁned as: • zn (i) = x((i + n) mod (218 − 1)) + y(i) mod 2.0 Cch.1 = . i = 0.2(n+1) .2n .0 Cch. The y sequence is constructed using the polynomial 1 + X 5 + X 7 + X 1 0 + X 18 .2n .2n .
Scrambling These binary sequences are converted to real valued sequences Zn by the following transformation: Zn (i) = +1 ifZn (i) = 0 for i = 0. 218 − 2 −1 ifZn (i) = 1 Finally.n is deﬁned as: • Sdl. . the n:th complex scrambling code sequence Sdl. Note that the pattern from phase 0 up to the phase of 38399 is repeated. . . 1.2. . 1. . . .A. 38399.n (i) = Zn (i) + jZn ((i + 131072) mod (218 − 1)). i = 0. 59 . .
.
Principles of Mobile Communication. WCDMA for UMTS. I. [12] Qualcomm.2. [11] COST 231. Chichester. [5] M. [13] P. Rupp.” Lecture slides. 61 .” GLOBECOM Workshops IEEE. 2007. [2] 3GPP.7 v1. L. WINNER. 2008. [10] B. Southern Gate. “Eﬃcient linktosystem level model for MIMO HSDPA. Viering. “Rayleigh fading channel in mobile digital communication systems part I: Characterization. 25. 2004. University of Balearic Island. Wrulich.0.876 V7. The Atrium. 1997. 2005. 25. and M.” [7] J. Digital mobile radio towards future generation systems. Physical channels and mapping of transport channels onto physical channels (FDD) (Tech.com/upload/news/ﬁles/10122007082416. “Mobile communication networks lecture notes.” available online: http://hspa.. Multiple Input Multiple Output in UTRA (Tech. Ltd. Kluwer Academic Publishers.” Master’s thesis. Sklar.uib.org. Ltd. “MIMO en sistemes hiperlan/2 i ieee 802. Rep IST2003507581 D. West Sussex PO19 8SQ.” IEEE Communications Magazine. S. 2007.0). 2003.0). Stüber. [9] H.pdf. Holma and A. “Assessment of advanced beamforming and MIMO technologies. [8] G. 2001.Bibliography [1] H. Holma and A. Eder. Spec. West Sussex PO19 8SQ.gsmworld. Soler. Toskala. England: John Wiley & Sons. England: John Wiley & Sons. Technical Speciﬁcation Group Radio Access Network. release 7 ed.es/ jramis/. release 7 ed. available online: http://dmi.. March 2007.0. HSDPA/HSUPA for UMTS. Chichester. September 2008.11a.. 2nd ed. The Atrium. Rep. Ramis.211 V8. 2006. Southern Gate. Toskala. Tech. Technical Speciﬁcation Group Radio Access Network. [4] M.3gpp. [3] 3GPP. Spec.2. “Release 7 hspa+ for mobile broadband evolution.. of WINNER.” Tech. University of Balearic Island. [6] “www.
“Recommendation itur m. Spreading and modulation (FDD) (Tech. [19] C. Rupp. 2008. C. “Simulation models with correct statistical properties for rayleigh fading channels. “Computationally eﬃcient MIMO HSDPA systemlevel evaluation. User Equipment (UE) radio transmission and reception (FDD) (Tech. and M. Mehlführer. Rupp. 1997. Wehinger.” EUPISCO.14. August 2009. Mecklenbräuker. release 7 ed. [22] S. “Optimum default base station parameter settings for UMTS networks. WIRELESS COMMUNICATIONS AND NETWORKING. Karner. [21] C.0). 1993..Rep. release 6 ed. “Intracell interference aware equalization for TxAA HSDPA. vol. Spatial channel model for Multiple Input Multiple Output (MIMO) simulations (Tech. Spec. Prentice Hall. 2005. [17] ITUR.. “Adaptive transmit beamforming with closed loop feedback in MIMO systems.” Global Telecommunications Conference. 2. “Measurement based evaluation of low complexity receivers for DTxAA HSDPA. Rosa and C. Xiao. release 7 ed. June 2003. December 2008. [26] 3GPP. Technical Speciﬁcation Group Radio Access Network. Mehlführer. September 2008. Rupp. Technical Speciﬁcation Group Radio Access Network. Zacarías. [18] 3GPP. [23] M. 51. 2004. C10. and M. September 2003. [20] C. Mehlführer. 25.Bibliography [14] S.” Master’s thesis. [16] Y. Rupp.6. Fundamentals of statistical Signal Processing: Estimation Theory. Kay. vol. S.1225: Guidelines for evaluation of radio transmission technologies for imt2000. [15] E.” SPAWC. 25. Wrulich. Technical Speciﬁcation Group Radio Access Network.. Spec. [25] W. Caban. rep. Geirhofer. “Design and realtime measurement of HSDPA equalizers. Mehlführer and M. 25. and J.1.101 V7. 2003.213 V7. Rupp.” tech. [24] M.” Wireless Pervasive Computing. and M. “Upgrading methodology for MIMO systemlevel simulations..” EURASIP. M. 2007..0).0). Wrulich and M. ftw. “Novel tapwise LMMSE channel estimation for MIMO WCDMA.” IEEE TRANSACTIONS ON COMMUNICATIONS. 2008. 2008. Spec. Technischen Universit¨at Wien. C. Tech.” Postgraduate course in Radio Communications. M.996 V6. Wrulich. 62 . [27] 3GPP.” in Wireless Evolution Beyond 3G.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue listening from where you left off, or restart the preview.