This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
◆ LTE and HSPA�: Revolutionary andEvolutionary Solutions for Global MobileBroadbandAnil M. Rao, Andreas Weber, Sridhar Gollamudi, and Robert Soni
2G—Second generation3G—Third generation3GPP—3rd Generation Partnership Project4G—Fourth generationACK—AcknowledgementAMR—Adaptive multi rateBPSK—Binary phase shift keyingCDF—Cumulative distribution functionCDMA—Code division multiple accessCLTD—Closed-loop transmit diversityCP—Cyclic prefixCPC—Continuous packet connectivityCQI—Channel quality indicatorCS-RS—Channel sounding reference signaldB—DecibelDCH—Dedicated channelDFT—Discrete Fourier transformDL—DownlinkDM-RS—Demodulation reference signalDPCCH—Dedicated physical control channelDSCH—Downlink shared channelDRX—Discontinuous receiveEDGE—Enhanced data rates for GSM evolutionE-DPCCH—Enhanced DPCCHeNB—Enhanced node BEPC—Evolved packet coreEPS—Evolved packet systemE-UTRAN—Evolved UTRANEV-DO—Evolution data optimizedFACH—Forward access channelFDD—Frequency division duplexFDMA—Frequency division multiple accessFFR—Fractional frequency reuseFTP—File Transfer ProtocolGERAN—GSM/EDGE radio access networkGGSN—Gateway GPRS support nodeGPRS—General packet radio serviceGSM—Global System for Mobile CommunicationsGW—GatewayHARQ—Hybrid automatic repeat requesths—High speedHSDPA—High speed downlink packet accessHSPA—High speed packet accessHSPA� —HSPA evolutionHSUPA—High speed uplink packet accessID—IdentificationIDFT—Inverse discrete Fourier transformIFFT—Inverse fast Fourier transformIoT—Interference over thermalIP—Internet ProtocolIPv4—IP version 4ISI—Inter-symbol interferenceIST—Information Society Technologieskm/h—Kilometers per hourLMMSE—Linear minimum mean square errorLTE—Long Term EvolutionMAC—Medium access controlMC—Multi-carrierMCS—Modulation and coding scheme
MIMO—Multiple input-multiple outputMME—Mobility management entityMMSE—Minimum mean square errorMRC—Maximum ratio combiningms—MillisecondsNACK—Negative acknowledgementNGMN—Next-generation mobile networkOFDM—Orthogonal frequency division multiplexOFDMA—Orthogonal frequency division multiple
accessPAPR—Peak to average power ratioPARC—Per antenna rate controlPCH—Paging channelPDN—Packet data networkPMI—Precoding matrix indicatorPS—Packet switchedPSD—Power spectral densityPSTN—Public switched telephone networkPUCCH—Physical uplink control channelQAM—Quadrature amplitude modulationQoS—Quality of serviceQPSK—Quadrature phase shift keyingRACH—Random access channelRAN—Radio access networkRLC—Release completeRNC—Radio network controllerRoHC—Robust header compressionRoT—Rise over thermalRRC—Radio resource controlRTP—Real Time Transport ProtocolRx—ReceiveSA—System architectureSAE—System architecture evolutionSC—Single carrierSCCH—Shared control channelS-CCPCH—Secondary common control physical
channelSDMA—Spatial division multiple accessSFBC—Space frequency block codingSGSN—Serving GPRS support nodeSIC—Successive interference cancellationSINR—Signal-to-interference-plus-noise ratioSMS—Short message serviceSRS—Sounding reference signalTA—Timing advanceTS—Technical specificationTTI—Transport time intervalTU—Typical urbanTx—TransmitUDP—User Datagram ProtocolUE—User equipmentUL—UplinkUMTS—Universal Mobile Telecommunications
SystemUTRAN—UMTS terrestrial radio access networkURA—UTRAN registration areaVoIP—Voice over Internet ProtocolWINNER—Wireless World Initiative New Radio
DOI: 10.1002/bltj Bell Labs Technical Journal 9
high speed uplink packet access (HSUPA) technology
was introduced in 3GPP Release 6 in March 2005. The
combination of HSDPA and HSUPA is called simply
high speed packet access (HSPA), and is strongly posi-
tioned to become the dominant high speed wireless
data technology for many years.
Even before the standardization of HSUPA had
been completed, in December 2004, 3GPP initiated a
feasibility study regarding the long term evolution
of UMTS, in order to ensure that the GSM family of
technologies maintained a competitive position in the
world market. The introduction of LTE was seen as a
way to provide a smooth migration to the yet-to-be-
defined fourth generation (4G), and to take advantage
of new spectrum allocations with wider bandwidths
that would become available (e.g., in the 2.6 GHz 3G
extension band). During the LTE feasibly study, an
aggressive set of performance targets and require-
ments were agreed upon to form the basis for LTE
standardization work. To justify the introduction of a
new technology, LTE would be required to provide
very large performance gains compared to HSPA in
3GPP Release 6 and to fully take advantage of new
spectrum allocations as wide as 20 MHz. In order to
satisfy these requirements, it was clear that LTE would
have to be built from the ground up, and could not
offer backwards compatibility with UMTS/HSPA.
The fact that LTE would not be backwards com-
patible with HSPA was not necessarily received well
by the large number of operators who had already
made significant investments in UMTS/HSPA and had
not yet begun to realize the benefits of those invest-
ments. This spurned the introduction of the HSPA
evolution (HSPA�) effort in 3GPP in March 2006.
While 3GPP had already started working on Release 7
enhancements as early as 2005, there was no general
framework to these enhancements. HSPA� formally
defined a broad framework and a set of requirements
for the evolution of HSPA; the primary goal being to
provide performance similar to LTE in a 5 MHz carrier,
while offering backwards compatibility with Release
99 through Release 6. HSPA� would then provide a
compelling alternative to LTE for operators who were
already deploying UMTS/HSPA, allowing them the
flexibility to introduce LTE in new spectrum while
enjoying enhancements which would protect their
existing investments in UMTS/HSPA.
In this paper, we will describe the requirements set
forth in standardization for both LTE and HSPA�, and
then give an overview of the key features of each tech-
nology which allow them to achieve their performance
requirements. We will see that many of the features
introduced in HSPA� closely parallel the innovations
developed in LTE. Where applicable, we provide
detailed system performance studies which illustrate
how close the technologies come to meeting the desired
performance targets. The remainder of the paper is
organized as follows: we start with the performance
requirements set forth in standards, give an overview of
the system architecture enhancements, describe the key
features in the downlink, describe the key features in
the uplink, describe the features in LTE and HSPA� that
enable efficient transmission of Voice over Internet
Protocol (VoIP), and offer our conclusions.
Requirements and Performance TargetsDuring the initial study item phase for both LTE
and HSPA�, 3GPP agreed upon a set of requirements
and performance targets to form the basis of the stan-
dardization work, and to determine what key features
or enhancements should be included as part of the new
technology. In this section, we discuss the requirements
and performance targets for both LTE and HSPA�.
LTE Requirements and Performance TargetsWhile 3GPP understood in early 2005 that the
HSPA technology—the uplink component of which
had just been standardized—would provide a highly
competitive mobile broadband solution for several
years, potential threats from other technologies cre-
ated a desire to ensure competitiveness in an even
longer time frame (i.e., for the next 10 years and
beyond). This formed the justification for opening the
LTE study item in 3GPP very quickly.
Important considerations for the long term evo-
lution of 3GPP included reduced latency, higher user
data rates, improved system capacity and coverage,
and reduced cost for operators. In order to achieve
this, it was seen that both an evolution of the air
interface as well as the network architecture would
need to be taken into account. Looking to the future,
10 Bell Labs Technical Journal DOI: 10.1002/bltj
the desire for even higher data rates also needed to
– Control plane supporting transition time of less
than 100 milliseconds (ms) from a camped state
(i.e., idle) to an active state (i.e., CELL_DCH),
and a transition time of less than 50ms from a
dormant state (i.e., URA/CELL_PCH) to an
active state.
– User plane supporting latency of less than
5 ms should be possible on the user plane in
an unloaded condition. The user plane
latency is defined as the one way transit time
between the user equipment (UE) and the
radio access network (RAN) edge node.
• Co-existence and inter-working with UMTS and GSM.
Given that LTE would co-exist with both
UMTS/HSPA terrestrial radio access network
(UTRAN) and GSM/EDGE radio access network
(GERAN), requirements were placed on inter-
working with these legacy systems. An interrup-
tion time of less than 500 ms is targeted for a
handover of a non-real time service between LTE
and either UTRAN or GERAN, while an interrup-
tion time of less than 300 ms is targeted for real-
time services.
• Given the scope of these requirements for evolu-
tion work, 3GPP agreed upon a work split—the
evolution of the radio access network would take
place in the 3GPP RAN working groups, and in
parallel, work on an evolved packet core (EPC)
DOI: 10.1002/bltj Bell Labs Technical Journal 11
would take place in the system architecture (SA)
working groups. At this point, it is useful to clarify
some terminology: the radio access network
enhancements are referred to as either evolved
UMTS terrestrial radio access network (E-UTRAN)
or LTE; the names are used interchangeably. The
evolution work for the EPC is referred to as sys-
tem architecture evolution (SAE). For some time,
the combination of these enhancements was
referred to as LTE/SAE, but more recently it has
become known as the evolved packet system
(EPS).
HSPA� Requirements and Performance TargetsIn order to protect operator investments in HSPA
and provide a smooth evolution path towards LTE,
which would not offer backwards compatibility with
earlier releases of UMTS/HSPA, a study item on HSPA
evolution was opened in 3GPP in March 2006. While
development of HSPA Release 7 enhancements was
already underway in 2005 with open work items
regarding HSDPA multiple input-multiple output
(MIMO), continuous packet connectivity (CPC), and
the “one tunnel” solution for optimization of packet
data traffic, there was no general framework in place
to guide the evolution of HSPA. The HSPA� effort
provided a broad framework for HSPA evolution with
a clear set of requirements and performance targets,
with the intent of identifying what performance bene-
fits could be achieved with the existing Release 7
work items and what gaps still existed.
The goal of HSPA� is not to replace LTE, but
rather to enhance HSPA by providing an incremental
evolution path for both the RAN and core network
which will enhance performance while leveraging
existing infrastructure. In addition, HSPA� aims to
enable co-existence with the EPS since it will be part
of future 3G systems. As described in [3, 8], the guid-
ing principles behind HSPA� are as follows:
• HSPA spectrum efficiency, peak data rates, and
latency should be comparable to LTE in a 5 MHz
bandwidth.
• The inter-working between HSPA� and LTE
should be as smooth as possible and facilitate joint
technology operation; the possibility of reusing
the evolved packet core defined as part of the sys-
tem architecture evolution should be analyzed.
• HSPA� should be able to operate as a packet-only
network, based on the utilization of shared chan-
nels only (i.e., HSDPA and HSUPA).
• HSPA� shall be backwards compatible in the sense
that legacy terminals compatible with Release 99
through Release 6 are able to share the same
carrier with terminals implementing the latest
HSPA� features, without any performance degra-
dation.
• Ideally, existing infrastructure should only need a
simple upgrade to support the features defined as
part of HSPA�.
As we will see in later sections, the framework
provided by the HSPA� effort initiated the develop-
ment of several new HSPA enhancements beyond
what was already being considered in early 2006, in
Release 7.
Network Architecture ImprovementsGiven the requirements and performance targets
described in the previous section, it was clear that not
only were enhancements to the radio interface
required, but in addition, the network architecture itself
needed enhancements. In the next section, we describe
the network architecture enhancements for both LTE
and HSPA�.
Evolved Packet System Architecture for LTEThe goal of the system architecture evolution
effort in 3GPP is not just to define an efficient packet
core network and RAN architecture for LTE to meet
the requirements described in [2], but rather to
develop a framework for an evolution and migration
of current systems to a high data rate, low latency,
packet-optimized system that supports mobility and
service continuity across heterogeneous access net-
works, since it is envisioned that Internet Protocol
(IP)-based services would be provided through vari-
ous access technologies.
In its simplest form, the EPS architecture consists
of two basic nodes in the user plane: a single node
called the enhanced node B (eNB) comprises all radio
access functions and a single node called the EPS
gateway comprises the entire bearer plane (i.e., user
12 Bell Labs Technical Journal DOI: 10.1002/bltj
plane) in the core network. In the control plane, the
mobility management entity (MME) node is logically
separated from the user plane EPS gateway with an
open interface between them. Figure 1 provides a
comparison of the EPS network architecture and the
UMTS network architecture. EPS offers a flatter net-
work architecture than UMTS, especially as far as the
user plane is concerned, which reduces latency.
The clean separation of the user plane and control plane
is a key feature of the EPS architecture, as it allows for
independent scaling of control plane functionality and
user plane functionality. This is very important from
a technical viewpoint because the scaling of the two
depends on different factors: the capacity of the con-
trol plane functionality typically depends on the num-
ber of mobile devices and their mobility patterns,
whereas the capacity of the user plane depends on
aggregate data throughput required to be supported.
Drawing a parallel between the EPS architecture and
UTRAN, the enhanced node B absorbs all radio access
functions that were contained in the node B and radio
network controller (RNC) elements in UTRAN. Note
that the eNBs are directly connected to each other via
an interface called X2; this facilitates seamless mobility
and interference management.
Figure 2 presents a more detailed view of the EPS
architecture, with the interfaces that exist to support
mobility between 3GPP and non-3GPP networks. The
EPS gateway may be split into two separate logical
nodes with the optional S5 interface: the serving gate-
way (GW), and the packet data network (PDN)
gateway. The serving GW terminates the core network
interface towards 3GPP radio access networks and
serves as the local mobility anchor point for inter-eNB
handover within the EPS, as well as mobility anchor-
ing for inter-3GPP mobility (i.e., between EPS and
UTRAN/GERAN). Note the direct control plane inter-
face (S3) and user plane interface (S4) between the
EPS network and the serving GPRS support node
(SGSN) in the UMTS/GSM networks; such an inter-
face allows for a packet session to be maintained in a
way that is seamless to the user of a multimode ter-
minal that migrates across LTE, UMTS/HSPA, and
GSM/EDGE coverage areas. This meets the requirements
Node BNode B Node BNode B
Packet corenetwork
Radio accessnetwork
eNode B eNode B
Internet Internet
UMTS EPSU-planeC-planeGGSN
SGSN
RNCRNC
MME
EPSgateway
EPS—Evolved packet systemGGSN—Gateway GPRS support nodeGPRS—General packet radio serviceHSPA—High speed packet access
MME—Mobility management entityRNC—Radio network controllerSGSN—Serving GPRS support nodeUMTS—Universal Mobile Telecommunications System
Figure 1.Comparison of UMTS/HSPA network architecture and EPS network architecture.
DOI: 10.1002/bltj Bell Labs Technical Journal 13
for co-existence/inter-working and gives the operator
the flexibility to roll out LTE gradually, starting with
the areas of highest demand first. The PDN gateway
provides access to the packet data network through
the control of IP data services and allocation of IP
addresses; it also serves as an anchor for mobility
between 3GPP and non-3GPP access systems, which is
sometimes referred to as the SAE anchor function.
Note that the specification of logical nodes does not
mandate a mapping to physical entities. For example,
the serving GW, PDN GW, and MME may be imple-
mented in the same physical entity, or the MME may be
integrated into the eNB. The mapping of logical nodes
to physical entities may follow a highly integrated
approach or a more distributed approach, based on ven-
dor implementations and deployment scenarios.
HSPA� Network ArchitectureFor the HSPA� network architecture, we begin
with a description of the one tunnel enhancement
(also referred to as “direct tunnel”) that was already
being discussed as part of Release 7 prior to the
HSPA� initiative. 3GPP recognized that the amount of
user plane data would significantly increase in the
near future because of the introduction of HSPA. With
the existing system illustrated in Figure 1, packet data
traffic must traverse both the gateway GPRS support
node (GGSN) and serving GPRS support node in the
UMTS core network (through the use of two tunnels),
independently of how the data traverses the UTRAN.
A more scalable architecture is possible with the one
tunnel solution, which permits direct tunneling of the
user plane data between the GGSN and the RNC, as
illustrated in Figure 3. In this way, a cleaner separa-
tion between the control plane and the user plane is
achieved, the advantages of which were discussed
previously for LTE. The new SGSN controller performs
all the control functions of the SGSN, and the
enhanced GGSN takes over all data transport func-
tionality that resided in the previous GGSN and SGSN.
To further flatten the network architecture,
HSPA� introduced the option of integrating the RNC
serving GW PDN GW
EPS gateway
E-UTRAN
UTRAN
GERAN
Non-3GPP† access
S1-MME
S1-U
S11
S4S3
S5Internet
SGSN
MME
3GPP—3rd Generation Partnership Project EDGE—Enhanced data rates for GSM†† EvolutionEPS—Evolved packet systemE-UTRAN—Evolved UTRANGERAN—GSM/EDGE radio access networkGPRS—General packet radio serviceGSM—Global System for Mobile Communications††
†Trademark of the European Telecommunications Standards Institute. ††Registered trademarks of the GSM Association.
GW—GatewayMME—Mobility management entityPDN—Packet data networkSGSN—Serving GPRS support nodeUMTS—Universal Mobile Telecommunications SystemUTRAN—UMTS terrestrial radio access network
Figure 2.Detailed view of the EPS network architecture, with interfaces to support mobility across 3GPP and non-3GPP access.
14 Bell Labs Technical Journal DOI: 10.1002/bltj
and node B functionality into a single node for packet
switched services (denoted here as node B� ); this is
also shown combined with the one tunnel solution
in Figure 3. The flatter architecture reduces latency
and could be useful in deployments in which a high
level of integration is desirable (i.e., HSPA femtocells).
From Figure 3, it quickly becomes apparent how simi-
lar the HSPA� network architecture is to the EPS net-
work architecture shown in Figure 1, which allows
easy integration of HSPA� and EPS networks.
Key Features in the Downlink of LTE and HSPA�
The downlink (DL) of a RAN bears a higher
amount of data traffic compared to the uplink (UL)
due to the increasing demand for unbalanced data
services like, for example, FTP download or video
streaming. A number of features have been intro-
duced in order to support increasing data rates.
LTE Downlink Key FeaturesIn the following subsections, we will discuss key
features of the LTE DL such as OFDM transmission,
MIMO, the possibility of using higher order modulation
schemes, time and frequency selective scheduling,
and fractional frequency reuse. Details about the
LTE DL channel structure can be found in [7]
and [12].
Orthogonal Frequency Division MultiplexThe LTE DL air interface is based on orthogonal
frequency division multiplexing (OFDM) which is a
technique that avoids inter-symbol interference,
exploits the scarce frequency resource nearly opti-
mally, combines the advantages of broadband and
narrowband transmission, and, at the same time,
avoids their disadvantages. OFDM is further described
below.
In conventional high bit rate air interfaces, the
data symbols are transmitted sequentially over the air
interface. According to the Nyquist theorem, the mini-
mum required bandwidth B is related to the symbol
duration Tsym with B � 1/Tsym. In real systems,
guard bands are required at both ends of the used
spectrum due to the application of non-ideal filters. In
multipath environments, broadband channels show
Node-B
Packet corenetwork
Radio accessnetwork
Internet
HSPA�one tunnel
HSPA�one tunnel
withintegrated
RNC/node B
GGSN
SGSN
RNC
Internet
GGSN
SGSN
Node-B�
GGSN—Gateway GPRS support nodeGPRS—General packet radio serviceHSPA�—High speed packet access evolution
U-planeC-plane
RNC—Radio network controllerSGSN—Serving GPRS support node
Figure 3.HSPA� network architecture.
DOI: 10.1002/bltj Bell Labs Technical Journal 15
a frequency selective behavior with several deep fades
in the frequency domain. In the time domain, this
behavior corresponds to an overlapping of symbols,
which causes the so called inter-symbol interference
(ISI), illustrated in Figure 4. The smaller the symbol
duration, i.e., the higher the symbol rate, the more
symbols experience ISI. In broadband transmissions,
an inversion of the channel transmission function is
required which corresponds to an equalization of the
received signal in order to cope with inter-symbol
interference caused by the relatively short symbol
duration (compared to the delay spread of the chan-
nel echoes).
One means to reduce inter-symbol interference
is to extend the symbol duration so that it is longer
than the difference between the delays of the earliest
and latest channel echo. A further improvement is to
extend the symbol duration by a guard time, during
which transmission of the new symbol has already
started but which is discarded by the receiver. This
feature is called cyclic prefix (CP). The total symbol
duration in this case is the sum of the original sym-
bol plus the CP duration, which should be longer than
the difference between the delays of earliest and lat-
est echo in the multipath channel.
The reason for the low ISI in the time domain is
because of a flat channel in the frequency domain. In
a multipath environment, this corresponds to a nar-
row bandwidth used for the transmission of the sym-
bol. Many parallel narrowband transmissions are
required to obtain a high bit rate channel. OFDM
avoids the guard band between the so-called subcar-
riers by a modulation of these subcarriers with rec-
tangular pulses, using the rect(t) function. In the
frequency domain, the spectrum of a pulse with dura-
tion Tsym corresponds to the sinc(x) � sin(x)/x func-
tion with zero crossings at k/Tsym, k � . . . � 2, � 1,
1, 2, . . . . Consequently, if these pulses modulate a
number of subcarriers, the inter-subcarrier interfer-
ence is zero with a subcarrier spacing of 1/Tsym which
is, according to the Nyquist theorem, the optimal
value, as shown in Figure 5. Furthermore, this opti-
mal value is reached without any filter.
The modulation of the equally-spaced subcarri-
ers with rect pulses corresponds to an inverse discrete
Fourier transform (IDFT) in the time domain. At the
receiver, the original symbols are reconstructed using
the opposite function, namely a discrete Fourier trans-
form (DFT). Figure 6 shows a schematic view of the
OFDM transmission chain.
Transmitted signal
Echo 1
Echo 2
Echo 3
Received signal
Symbol n Symbol n�1ISI
ISI—Inter-symbol interference
Figure 4.Inter-symbol interference caused by channel echoes.
16 Bell Labs Technical Journal DOI: 10.1002/bltj
In OFDM the bandwidth can be easily adapted to
the needs of the network operator. LTE FDD, for
example, offers bandwidths of 1.4 MHz, 3 MHz ,
5 MHz , 10 MHz , 15 MHz , and 20 MHz with a sub-
carrier spacing of 15 kHz. The total bandwidth
includes guard bands at both ends of the spectrum,
so that 72, 180, 300, 600, 900, and 1200 subcarriers
are conveyed in the respective bandwidth, as outlined
in [10].
The inverse fast Fourier transform (IFFT) and
corresponding FFT enable a very efficient calcula-
tion of the transmitted signal and the correspon-
ding reconstruction of the symbols. FFT and IFFT
require that the number of subcarriers N is N � 2n
with an integer value of n, although not all of these
subcarriers have to be transmitted. The granularity
in which the transmitter has to calculate the IFFT
and in which the receiver has to sample the
1
0.8
0.4
0.6
0
0.2
�0.2
Am
plit
ud
e/m
axim
um
am
plit
ud
e
�0.4�10 �5 0 5 10
Frequency (1/Tsym)
OFDM—Orthogonal frequency division multiplexing
Figure 5.Frequency domain of seven subcarriers of an OFDM signal.
Mapper IDFT DFT DemapperChannel
Bits Symbols s(t) s’(t) Symbols Bits
DFT—Discrete Fourier transform IDFT—Inverse discrete Fourier transformOFDM—Orthogonal frequency division multiplexing
Figure 6.OFDM transmission chain.
DOI: 10.1002/bltj Bell Labs Technical Journal 17
channel is Tsym/N and, consequently, depends on
the FFT size.
Multiple Antenna AlgorithmsMIMO, a synonym for a technique that uses at
least two antennas at the transmitter and at least two
antennas at the receiver for the transmission of signals
over the air interface, is depicted in Figure 7. The
antennas can be used to obtain an array gain, i.e., a
diversity gain, to reduce co-channel interference, or to
enable multiplexing of several data streams to the
same or to different receivers. Consequently, MIMO is
able to increase quality of service (QoS), coverage,
spectral efficiency, and peak data rate.
One or several data streams are, after channel
coding and modulation, multiplied with a precoding
vector and mapped on the different transmission
antennas. The precoding vector describes the phase
shifts of the data symbols and the mapping of these
data symbols on the antenna ports. The transmit (Tx)
and receive (Rx) antennas, respectively, are either
widely-spaced (or alternatively cross polarized) or
closely spaced in a linear array. In the first case, the
channel state at the antennas is uncorrelated. For
widely-spaced antennas, the antenna pattern gener-
ated is frequency dependent and has an irregular
shape. In the case of a linear array, the channel state
at the antennas is correlated; the generated antenna
pattern is frequency independent over the used band-
width and shows a regular shape with main and side
lobes. A special case for the uplink is virtual MIMO,
shown in Figure 8. In this case, two or more widely-
spaced mobile terminals are multiplexed on the same
resources. In contrast to the base stations, these
devices do not require multiple antennas. In general,
the maximum number of data streams corresponds
to the minimum of Tx and Rx antennas. In the case of
virtual uplink MIMO, the number of antennas in the
base station is the limiting factor.
In a closed loop transmission procedure, the pre-
coding vector is chosen so that a constructive super-
position of the signals is obtained at the receiver, or
that different data streams conveyed over the same
resources can be easily separated. In the LTE closed
loop mode, the receiver chooses the precoding vector
out of a limited set of possible precoding vectors in
order to reduce feedback signaling load. In this case,
the precoding feedback is reduced to an index in a
table of predefined precoding vectors, known as the
precoding matrix indicator (PMI). In the open loop
mode of LTE DL, space frequency block coding (SFBC)
is used under bad to medium channel conditions,
while per antenna rate control (PARC) is used under
good channel conditions and enables two stream
transmissions. Feedback from the receiver is still
required in order to signal the supportable rank, i.e.,
the number of supportable streams and the channel
quality.
The LTE DL works with orthogonal pilots for the
different transmission antennas in order to enable
the receiver to calculate the channel transmission
function, i.e., the transmission function from every
CodingModulationWeightingMapping
Channel
Data stream(s)
WeightingDemappingDemodulationDecoding
Data stream(s)
MIMO—Multiple input-multiple output
Figure 7.MIMO transmission and reception.
18 Bell Labs Technical Journal DOI: 10.1002/bltj
transmission to every receive antenna. The following
antenna algorithms can be applied with multiple
antenna systems:
• Transmit and receive diversity (with widely-
spaced or cross polarized antennas),
• Beamforming, or beamswitching (with closely-
spaced antennas),
• Spatial multiplexing (with widely-spaced or cross
polarized antennas), and
• Combinations of the previous algorithms.
Transmit diversity generates a rich number of
channel echoes at the receiver, which arrive from vari-
ous directions. By leveraging maximum ratio com-
bining (MRC), the receiver may use this receive
diversity in order to combine the signal of the differ-
ent receive antennas so that the received signal level
is optimized.
For beamforming and beamswitching, the anten-
nas have to be closely spaced, with a spacing of half
the wavelength of the carrier frequency. In an open
loop algorithm, the antenna spacing has to be cali-
brated. For a closed loop algorithm with channel feed-
back from the receiver, calibration is not mandatory
since the receiver chooses the optimal beam. Several
data streams can be conveyed over the same resources
to different users, if the beams have sufficient angular
separation, possible via spatial division multiple access
(SDMA).
Spatial multiplexing is an antenna algorithm
based on widely-spaced or cross polarized antennas.
Several data streams are mapped on the transmit
antennas. The receiver, e.g., a minimum mean square
error (MMSE) receiver, repeats to combine the sig-
nals of the different receive antennas so that the
signal-to-noise-ratio of one data stream is optimized
while the other data streams are suppressed, until all
data streams are reconstructed. Successive interfer-
ence cancellation (SIC) is an enhanced receiver algo-
rithm that subtracts the interference of successfully
received data streams from other data streams with
low signal-to-interference-plus-noise ratio (SINR).
A combination of spatial multiplexing and beam
forming or beam switching is enabled by a number
of closely spaced, cross polarized antennas at the
transmitter. This antenna allows the transmission of
up to two data streams and, at the same time, it allows
a beam to form in order to optimize the signal level at
the receiver.
CodingModulationWeightingMapping
Channel
Data stream(s)WeightingDemappingDemodulationDecoding
Data stream(s)
Data stream(s)
WeightingDemappingDemodulationDecoding
MIMO—Multiple input-multiple output
Figure 8.Virtual MIMO in the uplink.
DOI: 10.1002/bltj Bell Labs Technical Journal 19
Other Performance Enhancing TechniquesLink adaptation is a state of the art technique that
allows the adjustment of channel protection to chan-
nel quality by choosing the best suited modulation
and coding scheme (MCS). LTE allows coding rates,
i.e., the ratio of data bits and transmitted bits, close to
1 for excellent channel quality. Quadrature phase shift
keying (QPSK), 16 quadrature amplitude modulation
(QAM), and 64 QAM are possible modulation
schemes and can be combined with any code rate.
QPSK conveys 2 bits in every data symbol (resource
element), 4 bits in 16 QAM, and 6 bits in 64 QAM.
Consequently, in good channel conditions, three times
more bits can be conveyed using 64 QAM than under
bad channel conditions when QPSK is applied.
However, only those MCS that have the best throughput
performance for a given channel quality will be
applied, i.e., those which are part of the hull curve of
all MCS as shown in Figure 9.
Frequency selective scheduling improves spectral
efficiency and cell border throughput for OFDM by
choosing only the best set of physical resource blocks
for a transmission. In a frequency division duplex
(FDD) system, for frequency selective scheduling, the
mobile terminal has to feed back the channel quality
for either all resources or for a subset of resources
with the best channel qualities, i.e., channel quality
indicator (CQI). The scheduler evaluates the individ-
ual sets of CQI values in combination with the indi-
vidual throughput of the mobile devices and
SISO AWGN, 630 resource elements per transport block
HSDPA in 3GPP Release 5 and Release 6 already pro-
vided an efficient high speed downlink air interface
through the use of a short subframe length (2 ms),
hybrid automatic repeat request (HARQ), and fast,
channel-sensitive scheduling on a shared channel,
facilitated by the use of channel quality feedback and
the addition of a new advanced scheduling entity,
MAC high speed (MAC-hs) located in the base sta-
tion. HSDPA, in Release 5 and Release 6, supports
QPSK and 16 QAM modulation, and offers a peak of
14. 4 Mbps. Several enhancements have been intro-
duced for HSDPA in Release 7 as part of HSPA� in
order to improve spectral efficiency and cell border
throughput [9].
Higher Order ModulationHSPA� allows up to 64 QAM modulation in the
downlink, which conveys 6 bits per symbol instead of 4
bits in the case of 16 QAM and consequently increases
the peak data rate by 50 percent to 21.6Mbps. 64 QAM
can be applied under good channel conditions. Due to
this fact, the possibility of using 64 QAM will enhance
spectral efficiency but will not have a high impact on
cell border throughput. For backward compatibility,
new terminal types have been defined that support
64 QAM.
MIMOHSPA� allows closed loop 2x2 MIMO with two
transmit antennas and two receive antennas. Under
good channel conditions, dual stream transmissions
are possible that can double the peak bit rate to
28.8 Mbps. As already described for LTE MIMO, the
mobile terminal chooses the best precoding vector out
of a set of predefined precoding vectors together with
CQI values for one or two streams. In order to enable
the device to measure the signal quality separately for
both antennas, the antennas carry orthogonal pilot sig-
nals. In case of dual stream transmission, both streams
can have different modulation and coding schemes
according to their channel quality. In case of low chan-
nel quality, the scheduler can decide to switch back to
single stream transmission, which then takes place on
the two antennas via closed-loop transmit diversity
(CLTD). The MIMO scheme, precoding vector, and
MCS signal the mobile device via the high speed
shared control channel (HS-SCCH). Dual stream
MIMO in HSPA� supports improved system capacity
rather than improved cell border throughput.
However, the fallback mode of CLTD for single stream
transmission will increase cell border throughput com-
pared to the case of transmitting with a single antenna
only. The combination of MIMO with 64 QAM is not
foreseen for Release 7, but will be part of Release 8 of
HSPA�, increasing the peak rate to 43. 2 Mbps.
Enhanced Receiver TypesOne common means to increase downlink sys-
tem capacity and cell border throughput is to enhance
the requirements for the mobile receiver. For Release 5,
requirements are based on a single antenna rake
receiver. Release 6 defined requirements based on a
DOI: 10.1002/bltj Bell Labs Technical Journal 21
rake receiver with dual antenna receive diversity
(enhanced receiver type 1) and on a single antenna
receiver with equalization, e.g., an MMSE receiver
(enhanced receiver type 2). Release 7 defines require-
ments for the combination of dual antenna receive
diversity and equalization (enhanced receiver type 3).
The introduction of so-called interference aware
receivers further improves performance. Using this
feature, the receiver reduces the interference from
neighbor cells, which works to enhance cell border
throughput. This feature is used in conjunction with
equalization for single antenna (enhanced receiver
type 2i) or dual antenna Rx diversity (enhanced
receiver type 3i).
Downlink Performance ComparisonTable I and Table II show the performance com-
parison of HSDPA Release 6 with 5 MHz bandwidth
and LTE DL with 5 MHz and 10 MHz bandwidth,
respectively. For the basic assumptions we used the
HSDPA Improvement Release 6 LTE LTE compared(5 UEs/cell (5 UEs/cell (10 UEs/cell to HSDPAin 5 MHz) in 5 MHz) in 10 MHz) 1 � 2
Case 1 (1 � 2) 0.47 1.33 1.52 2.8 – 3.2x
Case 1 (2 � 2) NA 1.47 1.60 3.1 – 3.4x
Case 1 (4 � 2) NA 1.73 1.85 3.7 – 3.9x
Case 3 (1 � 2) 0.44 1.24 1.40 2.8 – 3.2x
Case 3 (2 � 2) NA 1.37 1.50 3.1 – 3.5x
Case 3 (4 � 2) NA 1.60 1.70 3.6 – 3.9x
Table I. Average downlink spectral efficiency (bps/Hz/cell) with NGMN assumptions.
HSDPA—High speed downlink packet accessLTE—Long term evolutionNA—Not applicableNGMN—Next-generation mobile networkUE—User equipment
HSDPARelease 6 LTE LTE Improvement
(5 UEs/cell in (5 UEs/cell in (10 UEs/cell in compared to5 MHz) 5 MHz) 10 MHz) HSDPA 1 � 2
Case 1 (1 � 2) 195 223 321 1.1 – 1.6x
Case 1 (2 � 2) NA 257 345 1.3 – 1.8x
Case 1 (4 � 2) NA 337 462 1.7 – 2.4x
Case 3 (1 � 2) 170 140 209 0.8 – 1.2x
Case 3 (2 � 2) NA 186 262 1.1 – 1.5x
Case 3 (4 � 2) NA 257 323 1.5 – 1.9x
Table II. Five percent CDF downlink user throughput (kbps) with NGMN assumptions.
CDF—Cumulative distribution functionHSDPA—High speed downlink packet accessLTE—Long term evolutionNA—Not applicableNGMN—Next-generation mobile networkUE—User equipment
22 Bell Labs Technical Journal DOI: 10.1002/bltj
3GPP performance verification framework [5], which
is based on TS 25.814 [4]. Contrary to this frame-
work, we scaled the average number of users per sec-
tor according to the bandwidth in order to have a fair
comparison across the different systems. For HSDPA
mobile terminals we used enhanced receiver type 1,
i.e., a rake receiver with two antenna receive diver-
sity. For LTE, we used a maximum ratio combining
receiver in the case of single stream transmission and
a linear minimum mean square error (LMMSE)
receiver in the case of dual stream transmission.
Switching between single and dual stream in the case
of 2�2 and 4�2 transmission was performed accord-
ing to the precoding matrices presented in 3GPP TS
36.211 [7]. For all DL cases, we used the spatial channel
model WiM C2 “macro urban” from the Information
Society Technologies Wireless World Initiative New
Radio (IST WINNER) project. We performed the
simulations for case 1 with an inter site distance of
500 meters (m) and for case 3 with an inter site dis-
tance of 1732 m; both cases use a penetration loss
of 20 decibels (dB) and are at a carrier frequency of
2 GHz. The 3GPP performance verification framework
requests an improvement factor between 3 and 4 for
DL spectral efficiency and between 2 and 3 for DL cell
border throughput, which is defined as the fifth per-
centile of the mobile terminal’s cumulative through-
put distribution function. However, for the 3GPP
performance verification framework, HSDPA with
5 MHz bandwidth is compared to LTE with 10 MHz
bandwidth, and with 10 users in an average per sec-
tor for both systems, which penalizes the average
mobile device and cell border throughput of HSPA by
a factor of two. Consequently, in our comparison, as
we scale the number of users with the bandwidth,
the required improvement factor for the cell border
throughput of 3GPP has to be divided by a factor of 2
and, hence, shall be between 1 and 1.5. The required
improvement factor for spectral efficiency remains the
same. For all simulations, a standard proportional fair
scheduler has been used which, in the long term,
assigns approximately the same number of resources
to the mobile devices. For HSDPA, the proportional
fair algorithm has been performed in time, for LTE DL
in time and frequency. It is interesting that for case 3,
i.e., for an inter-site distance of 1732 m, the most
comparable case, namely 1�2 with 5MHz bandwidth,
the LTE cell border throughput is 18 percent smaller
but the spectral efficiency is 180 percent higher
compared to HSDPA. However, the LTE DL propor-
tional fair scheduler could be easily tuned towards
higher cell border throughput on the cost of spectral
efficiency. For case 1, a 500 meter inter-site distance,
cell border throughput is superior to HSDPA Release
6 while we still obtain a 180 percent gain in spectral
efficiency. Due to a higher channel diversity, spectral
efficiency and cell border throughput increases with
increasing bandwidth. With increasing numbers of
transmit antennas, we also obtain a gain in both spec-
tral efficiency and cell border throughput due to better
exploitation of channel diversity. As a consequence, if
we combine both effects, the spectral efficiency gains
for increasing bandwidth decreases with increasing
numbers of transmit antennas due to the fact that
both exploit channel diversity.
Further improvements are possible for HSDPA and
LTE. HSDPA, according to Release 7, introduces 64
QAM and MIMO in the DL as well as new enhanced
receiver types with equalizers and with intra-cell inter-
ference cancellation. 64 QAM will lead to an improve-
ment of spectral efficiency. Dual stream MIMO will
not lead to an improvement of cell border through-
put. However, transmit diversity, the fallback mode
for 2�2, may enhance cell border throughput as well
as the new enhanced receiver types. All these features
will reduce the gap between HSDPA and LTE DL per-
formance. However, all results presented for LTE are
without interference rejection combining, which will
[10] 3rd Generation Partnership Project, “EvolvedUniversal Terrestrial Radio Access (E-UTRA),Base Station (BS) Radio Transmission andReception (Release 8),” 3GPP TS 36.104, v8.0.0,Dec. 2007, �http://www.3gpp.org/ftp/Specs/html-info/36104.htm�.
[11] K. Balachandran, Q. Bi, A. Rudrapatna, J. Seymour, R. Soni, and A. Weber, “PerformanceAssessment of Next-Generation Wireless MobileSystems,” Bell Labs Tech. J., 13:4 (2009), 35–58.
[12] Informa Telecoms & Media, WCIS, and 3GAmericas, Global UMTS and HSPA OperatorStatus, Feb. 27, 2008.
(Manuscript approved October 2008)
ANIL M. RAO is a member of technical staff in Alcatel-Lucent’s wireless research anddevelopment (R&D) organization inNaperville, Illinois. He received a B.S. inapplied mathematics from the University ofAlaska, Fairbanks, and M.S. and Ph.D.
degrees in electrical engineering from the Universityof Illinois at Urbana Champaign where he held aNational Science Foundation graduate researchfellowship. Dr. Rao joined Alcatel-Lucent afterassignments with NASA’s Jet Propulsion Laboratoryand TRW. His work at Alcatel-Lucent has involvedvarious aspects of system design, performance analysis,and algorithm development for UMTS, HSPA/HSPA�,and LTE. He has actively contributed to both thestandardization and product realization of thesetechnologies. His interests include intelligentantennas, scheduling and resource allocationalgorithms, and optimizing the end-to-endperformance of mobile broadband wireless systems.
ANDREAS WEBER is team leader of the mobile systemperformance evaluation group in Bell Labs’Radio Access domain in Stuttgart, Germany.He received Dipl.-Ing. and Dr.-Ing. degreesin electrical engineering from the Universityof Stuttgart, Germany. Prior to joining
Alcatel-Lucent, Dr. Weber worked in the field ofsatellite communications as a member of scientific staffat the Institute of Communications Switching and DataTechnics, University of Stuttgart. During his tenure atAlcatel Research & Innovation and later at Bell Labs, heworked on the performance evaluation andoptimization of 2G, 3G, and beyond 3G mobilecommunication systems. Currently, he and his teamwork on LTE Advanced and WiMAX.
SRIDHAR GOLLAMUDI is a member of technical staffwith Alcatel-Lucent’s Wireless research anddevelopment (R&D) organization inWhippany, New Jersey. He received his Ph.D.in electrical engineering from the Universityof Notre Dame, Indiana.
Dr. Gollamudi worked at Motorola Inc. beforebeginning his career at Alcatel-Lucent. His researchinterests include statistical signal processing, resourceallocation in wireless systems, physical and MAC layeralgorithm design, and performance analysis ofcommunications systems.
34 Bell Labs Technical Journal DOI: 10.1002/bltj
ROBERT SONI is a technical manager in Alcatel-Lucent’sWireless business group in Whippany, NewJersey. He supervises a group which isinvestigating and developing new advancedantenna, physical layer and MAC layertechnologies for 3G/4G cellular systems. He
received a Ph.D. and MSEE in electrical engineeringfrom the University of Illinois at Urbana-Champaign,and received his BSEE, summa cum laude, from theUniversity of Cincinnati in Ohio. Dr. Soni began hiscareer as a member of technical staff at Alcatel-Lucentten years ago. He also teaches part-time at ColumbiaUniversity in New York City, and the New JerseyInstitute of Technology in Newark, New Jersey. ◆