EKSTRÖM LAYOUT 2/14/06 11:03 AM Page 38
`
`TOPICS IN
`RADIO COMMUNICATIONS
`
`Technical Solutions for the
`3G Long-Term Evolution
`
`Hannes Ekström, Anders Furuskär, Jonas Karlsson, Michael Meyer, Stefan Parkvall,
`Johan Torsner, and Mattias Wahlqvist, Ericsson
`
`ABSTRACT
`Work has started in the 3GPP to define a
`long-term evolution for 3G, sometimes referred
`to as Super-3G, which will stretch the perfor-
`mance of 3G technology, thereby meeting user
`expectations in a 10-year perspective and
`beyond. The fundamental targets of this evolu-
`tion — to further reduce user and operator costs
`and to improve service provisioning — will be
`met through improved coverage and system
`capacity as well as increased data rates and
`reduced latency. This article presents promising
`technologies to fulfill these targets, including
`OFDM, multi-antenna solutions, evolved QoS
`and link layer concepts, and an evolved architec-
`ture. Furthermore, the results of a performance
`evaluation are presented, indicating that the
`requirements can indeed be reached using the
`proposed technologies.
`
`BACKGROUND AND
`TARGETS FOR 3G EVOLUTION
`Third-generation (3G) wireless systems, based
`on wideband code-division multiple access
`(WCDMA) radio access technology, are now
`being deployed on a broad scale all over the
`world. The first step in the evolution of
`WCDMA has also been taken by the Third Gen-
`eration Partnership Project (3GPP) through the
`introduction of high-speed downlink packet
`access (HSDPA) [1] and enhanced uplink [2].
`These technologies provide 3GPP with a radio
`access technology that will be highly competitive
`in the mid-term future.
`However, user and operator requirements
`and expectations are continuously evolving, and
`competing radio access technologies are emerg-
`ing. Thus, it is important for 3GPP to start con-
`sidering the next steps in 3G evolution, in order
`to ensure 3G competitiveness in a 10-year per-
`spective and beyond. As a consequence, 3GPP
`has launched the study item Evolved UTRA and
`UTRAN, the aim of which is to study means to
`achieve further substantial leaps in terms of ser-
`vice provisioning and cost reduction. The overall
`
`target of this long-term evolution (LTE) of 3G,
`sometimes also referred to as Super-3G, is to
`arrive at an evolved radio access technology that
`can provide service performance on a par with
`or even exceeding that of current fixed line
`accesses, at substantially reduced cost compared
`to current radio access technologies. As it is gen-
`erally assumed that there will be a convergence
`toward the use of Internet Protocol (IP)-based
`protocols (i.e., all services in the future will be
`carried on top of IP), the focus of this evolution
`should be on enhancements for packet-based
`services. 3GPP aims to conclude on the evolved
`3G radio access technology in 2007, with subse-
`quent initial deployment in the 2009–2010 time-
`frame. At this point, it is important to emphasize
`that this evolved radio access network (RAN) is
`an evolution of current 3G networks, building on
`already made investments.
`Among others, the targets of long-term 3G
`evolution are [3]:
`• The possibility to provide significantly high-
`er data rates than do the current steps of
`3G evolution (HSDPA and enhanced
`uplink), with target peak data rates up to
`100 Mb/s for the downlink and up to 50
`Mb/s for the uplink.
`• The capability to provide three to four times
`higher average throughput and two to three
`times higher cell-edge throughput (mea-
`sured at the 5th percentile) when compared
`to 3GPP Release 6 (Rel-6) systems (i.e.,
`systems based on HSDPA and enhanced
`uplink).
`• Improved spectrum efficiency, targeting an
`improvement on the order of a factor of 3
`compared to current standards.
`• Significantly reduced control and user plane
`latency, with a target of less than 10 ms
`user plane RAN round-trip time (RTT)
`and less than 100 ms channel setup delay
`• Reduced cost for operator and end user.
`• Spectrum flexibility, enabling deployment in
`different spectrum allocations. This involves
`a smooth migration into other frequency
`bands, including those currently used for
`second-generation (2G) cellular technolo-
`gies such as GSM and IS-95.
`
`38
`
`0163-6804/06/$20.00 © 2006 IEEE
`
`IEEE Communications Magazine • March 2006
`
`Ericsson v. IV II LLC
`Ex. 1014 / Page 1 of 8
`
`

`

`EKSTRÖM LAYOUT 2/14/06 11:03 AM Page 39
`
`Based on the
`requirements of
`reduced latency and
`cost, it is natural to
`consider system
`architectures that
`contain a reduced
`number of network
`nodes along the data
`path. This would
`reduce both
`the overall
`protocol-related
`processing as well as
`the number of
`interfaces, which in
`turn reduces the cost
`of interoperability
`testing.
`
`GGSN
`
`SGSN
`
`RNC
`
`RNC
`
`SGSN
`ACGW
`
`Node B
`
`Node B
`
`Node B
`
`Node B
`
`Node B
`
`Node B
`
`Node B
`
`Node B
`
`UE
`
`UE
`
`nnnnFigure 1. The current 3GPP Release 6 architecture (left) and one possible evolved 3G architecture reduc-
`ing the number of nodes along the data path from 4 to 2 (right).
`
`One additional requirement is the possibility
`for smooth introduction of technical solutions
`that fulfill these targets. Thus, any new or
`evolved radio access technology must be able to
`coexist with current 3G radio access technologies
`and radio network architectures and vice versa.
`To achieve the above-mentioned targets,
`3GPP needs to consider new radio transmis-
`sion technologies as well as updates and mod-
`ifications to the existing radio network
`architecture. Many such technologies have
`been proposed in the context of new fourth-
`generation (4G) mobile systems research
`[4–7]. However, in order to protect operator
`and vendor investments, the performance gain
`of any proposed update to or evolution of the
`3G radio access or RAN must always be trad-
`ed off against its impact on already made
`investments.
`In this article candidate building blocks of
`a possible long-term 3G evolution are
`described. These building blocks are: an
`evolved system architecture, evolved quality of
`service (QoS) and link layer concepts, the use
`of orthogonal frequency-division multiplexing
`(OFDM) as a new access technology enabling
`frequency domain adaptation, and finally, the
`possibility of employing multi-antenna solu-
`tions. It should be noted that the standardiza-
`tion of 3G long-term evolution is currently
`ongoing. It is therefore uncertain to what
`extent these technical solutions will be includ-
`ed in the standard. An initial performance
`evaluation of some of these building blocks is
`also provided, and finally, conclusions are
`drawn and presented.
`
`TECHNICAL SOLUTIONS
`This section presents candidate technical solu-
`tions for the evolved radio access and RAN. A
`top-down approach is followed, beginning with
`architecture aspects and ending with physical
`layer issues.
`
`ARCHITECTURE EVOLUTION
`Based on the requirements of reduced latency
`and cost, it is natural to consider system archi-
`tectures that contain a reduced number of net-
`work nodes along the data path. This would
`reduce both the overall protocol-related process-
`ing as well as the number of interfaces, which in
`turn reduces the cost of interoperability testing.
`A reduction of the number of nodes also makes
`it possible to reduce call setup times, as fewer
`nodes will be involved in the call setup proce-
`dure. Such a reduction also gives greater possi-
`bilities to merge control plane protocols, thereby
`potentially further reducing call setup times.
`Figure 1 illustrates the current Rel-6 architec-
`ture and a possible path for an architecture evo-
`lution.
`In Rel-6, the Node B handles the lower lay-
`ers of the wireless access, as this is the node
`with the antenna. The radio network con-
`troller (RNC) handles radio resource manage-
`ment, mobility management (locally), call
`control, and transport network optimization. It
`further acts as a termination point for the
`radio protocols. The gateway General Packet
`Radio Service (GPRS) support node (GGSN)
`acts as an anchor node in the home network.
`The serving GPRS support node (SGSN) acts
`as an anchor node in the visiting network and
`handles both mobility management and ses-
`sion management. Typically all traffic is rout-
`ed back to the home network so that a
`consistent service environment can be main-
`tained while also allowing the operator to fil-
`ter traffic and provide security to the end user
`(e.g., by means of firewalls).
`In the proposed LTE architecture, the Rel-6
`nodes GGSN, SGSN, and RNC are merged into
`a single central node, the access core gateway
`(ACGW) as shown in Fig. 1. The ACGW termi-
`nates the control and user planes for the user
`equipment (UE), and handles the core network
`functions provided by the GGSN and SGSN in
`Rel-6. The control plane protocol for the UE
`
`IEEE Communications Magazine • March 2006
`
`39
`
`Ex. 1014 / Page 2 of 8
`
`

`

`EKSTRÖM LAYOUT 2/14/06 11:03 AM Page 40
`
`ACGW
`
`Node B
`
`UE
`
`IP packet
`
`IP packet
`
`IP packet
`
`IP packet
`
`RLC
`
`RLC payload
`
`RLC payload
`
`RLC payload
`
`RLC payload
`
`RLC payload
`
`RLC payload
`
`RLC header
`
`FEC block
`
`FEC block
`
`FEC block
`
`FEC block
`
`MAC
`FEC fragment
`
`FEC fragment
`
`FEC fragment
`
`FEC fragment
`
`FEC fragment
`
`FEC fragment
`
`nnnnFigure 2. Schematic data flow through the RLC and MAC layer for downlink traffic.
`
`TTI1
`
`TTI2
`
`TTI1
`
`TTI2
`
`will be similar to radio resource control (RRC)
`in Rel-6, for example. handling control of mobil-
`ity and radio bearer configuration. In the user
`plane the ACGW will handle functions like
`header compression, ciphering, integrity protec-
`tion, and automatic repeat request (ARQ).
`The proposed architecture has the following
`merits:
`• User-plane latency is reduced, as there are
`fewer nodes, and less protocol packing/
`unpacking.
`• Call/bearer setup time is reduced, as there
`are fewer nodes involved in the setup pro-
`cedure.
`• Complexity is reduced, as there are fewer
`interfaces to implement and test. The
`amount of interoperability testing required
`will therefore also be reduced.
`• Placing an ARQ protocol in the ACGW will
`provide both robustness against lower-layer
`losses and a simple way to provide lossless
`mobility.
`• Performing ciphering and integrity protec-
`tion of control and user plane data in the
`ACGW allows for a security solution at
`least as strong as in Rel-6.
`• Support for macrodiversity can be provided
`with centralized radio control handling.
`This has been shown to give significant cov-
`erage and capacity gains.
`• There is no need for a direct Node B–Node
`B interface for mobility. Such an interface
`would increase the operational burden for
`the operator (through additional configura-
`tion and planning) and also impose a new
`security threat to the network.
`• A new function in the proposed architecture
`compared to Rel-6 is support for ACGW
`pooling. This allows for network redundan-
`cy solutions that increase the reliability of
`the network.
`QUALITY OF SERVICE
`The key driver behind the QoS concept
`described in this section is to provide operators
`with effective and simple means to provide ser-
`vice differentiation over networks that employ
`high-speed shared channels. Two components of
`the QoS concept are described: service differen-
`tiation and simplified bearer realizations. It
`should be noted that the concept presented
`
`below should be seen as one possible evolution
`of the current 3GPP QoS concept.
`Service differentiation is enabled by classifi-
`cation and marking of each packet at the net-
`work edge (i.e., ACGW for downlink traffic and
`UE for uplink traffic). The edge node classifies
`each incoming packet into different predefined
`service classes, such as Internet access and voice
`over IP (VoIP). This classification could, for
`example, be done on the basis of information
`contained in the protocol headers. Following
`classification, the packet is marked. An explicit
`form of packet marking is the use of IP layer dif-
`ferentiated services (DiffServ) code points, while
`an implicit form of packet marking is the map-
`ping of packets to “marked bearers,” such as a
`“marked” packet data protocol (PDP) context or
`radio access bearer. This marking is then used
`by each subsequent node to identify the service
`class to which the packet belongs. The edge
`node further performs rate policing and/or
`admission control to ensure that flows do not
`exceed a specified maximum bit rate. For some
`service classes (e.g., Internet access), this maxi-
`mum bit rate may be specified on a subscription
`basis, whereas for others (e.g., VoIP) it may be
`specified on a session basis during the session
`setup phase.
`Once all incoming packets have been marked
`and policed, each node in the data path uses the
`markings to carry out appropriate queuing and
`policy-based scheduling. The queuing in the
`nodes may be service class dependent; that is,
`the size and dropping strategies of the queue
`may differ depending on the characteristics of
`the traffic belonging to the service class. Policy-
`based scheduling denotes the process of schedul-
`ing according to predefined policies. Such
`policies can, for example, govern the distribution
`of bandwidth between different service classes. It
`is foreseen that such policies can be modified
`dynamically depending on the expected usage of
`particular services. It should be possible for the
`operator to push new policies to the relevant
`nodes through the network management system.
`For a more detailed discussion of service differ-
`entiation and scheduling, see [8].
`Another key component of the QoS concept
`is the use of a reduced number of radio bearer
`realizations. For example, it is believed that due
`to the significant latency reduction in the evolved
`
`40
`
`IEEE Communications Magazine • March 2006
`
`Ex. 1014 / Page 3 of 8
`
`

`

`EKSTRÖM LAYOUT 2/14/06 11:03 AM Page 41
`
`FDD only
`
`Combined FDD/TDD
`
`TDD only
`
`fDL
`fUL
`Highest data rates for given
`bandwidth and peak power
`
`fDL
`fUL
`
`nnnnFigure 3. Duplex schemes.
`
`fDL/fUL
`
`Reduced UE complexity
`
`Unpaired spectrum
`
`RAN, even real-time services like VoIP could be
`supported over a reliable link layer (acknowl-
`edged mode). The bearers will, however, differ
`in the scheduling policies that are assigned to
`them. It is believed that this reduction in radio
`bearer realizations will help to reduce the time
`to market when introducing new services, since
`no service-specific bearer will need to be defined
`and tested prior to introduction of a new service.
`LINK LAYER SOLUTIONS
`While the Rel-6 link layer protocols support the
`peak data rates of HSDPA and enhanced uplink
`effectively, the requirements on the evolved
`RAN demand enhanced link layer concepts. A
`fixed radio link control (RLC) protocol data unit
`(PDU) size is regarded as too inflexible to oper-
`ate over a wide range of data rates. Small PDUs
`lead to too large header overhead, while large
`PDUs would introduce too much padding over-
`head for small packets like VoIP frames or TCP
`acknowledgments. Therefore, another solution,
`called the packet-centric link layer, is outlined
`here.
`The concept foresees two layer 2 ARQ proto-
`cols as in Rel-6. The RLC protocol, which con-
`tains ARQ functionality, operates between
`ACGW and UE, while the hybrid ARQ (HARQ)
`protocol is embedded in the medium access con-
`trol (MAC) layer, and operates between Node B
`and UE. The RLC protocol is needed to provide
`a reliable mobility and ciphering anchor point,
`and cope with congestion losses on the Iub inter-
`face, while radio interface transmission errors
`are typically not handled by the RLC, but by
`HARQ.
`The key characteristic of the packet-centric
`link layer is to map packets (i.e., either IP pack-
`ets or RRC messages) one-to-one to RLC PDUs,
`thereby making the size of these PDUs variable,
`as depicted in Fig. 2. This concept deems seg-
`mentation and concatenation at the RLC layer
`obsolete, thereby eliminating padding overhead.
`An additional field, specifying the PDU size, is
`required in the protocol header. However,
`despite this added overhead, an overall gain in
`terms of overhead is typically achieved since
`padding is avoided.
`In addition, the concept has the advantage
`that IP packets become implicitly visible in the
`Node B, because each RLC PDU corresponds to
`exactly one IP packet. This fact can be exploited
`by the scheduler in the MAC layer, which now
`sees complete IP packets as opposed to seg-
`ments thereof. This is expected to allow for
`more efficient scheduling decisions.
`A potential problem of the packet-centric
`concept is that one RLC PDU may be too large
`to be transmitted in one frame (e.g., when the
`
`IEEE Communications Magazine • March 2006
`
`receiver is experiencing bad signal quality). In
`this case, segmentation is required in the Node
`B. However, instead of segmenting an RLC
`PDU into multiple pieces,1 it is proposed to first
`encode the RLC PDU into forward error correc-
`tion (FEC) blocks and then use rate matching to
`form FEC fragments, which fit into the available
`radio resources. If the RLC PDU is large, this
`may result in a very high initial code rate, in
`some cases even higher than one, making it
`highly unlikely that such a transmission can be
`decoded correctly. Therefore, in combination
`with incremental redundancy HARQ, so-called
`autonomous retransmission is performed, where-
`by more data from this PDU is transmitted with-
`out waiting for a negative acknowledgment. This
`is repeated until the probability of successful
`reception has exceeded a certain threshold. Sub-
`sequently, conventional HARQ feedback is used
`to request further retransmissions if needed.
`The RAN transport is expected to remain an
`expensive part of the network, and overdimen-
`sioning of these links cannot generally be
`assumed. Therefore, packet losses due to con-
`gestion in the transport network will occur
`despite deployment of enhanced flow control
`mechanisms. A further enhancement deals with
`this problem. In such scenarios the Node B can
`act as an RLC relay node and send negative
`acknowledgments back to the ACGW to request
`local retransmissions. This avoids time-consum-
`ing ARQ operations over the radio interface.
`Furthermore, since the sequence number is visi-
`ble in the Node B, head-of-line blocking for such
`retransmissions can be avoided by ordering the
`PDUs according to their sequence number in
`the Node B transmission queue.
`
`PHYSICAL LAYER AND
`RADIO RESOURCE MANAGEMENT
`
`OFDM [9] is an attractive choice to meet
`requirements for high data rates, with corre-
`spondingly large transmission bandwidths, and
`flexible spectrum allocation. OFDM also allows
`for a smooth migration from earlier radio access
`technologies and is known for high performance
`in frequency-selective channels. It further
`enables frequency domain adaptation, provides
`benefits in broadcast scenarios, and is well suited
`for multiple-input multiple-output (MIMO) pro-
`cessing.
`The possibility to operate in vastly different
`spectrum allocations is essential. Different band-
`widths are realized by varying the number of
`subcarriers used for transmission, while the sub-
`carrier spacing remains unchanged. In this way
`operation in spectrum allocations of 1.25, 2.5, 5,
`10, 15, and 20 MHz can be supported. Due to
`
`The RAN transport is
`expected to remain
`an expensive part of
`the network and
`over-dimensioning of
`these links cannot
`generally be
`assumed. Therefore,
`packet losses due to
`congestion in the
`transport network
`will occur despite
`deployment of
`enhanced flow
`control mechanisms.
`
`1 The segmentation of
`RLC PDUs at the MAC
`layer is another viable
`alternative. Since it is the
`traditional approach, fur-
`ther discussion is omitted
`here.
`
`41
`
`Ex. 1014 / Page 4 of 8
`
`

`

`EKSTRÖM LAYOUT 2/14/06 11:03 AM Page 42
`
`≈200 kHz
`
`0.5 ms
`
`User A
`User B
`User C
`User D
`
`Time
`
`Time
`
`Frequency
`
`Downlink
`
`Frequency
`
`Uplink
`
`nnnnFigure 4. Time-frequency structure for downlink (left) and uplink (right).
`
`the fine frequency granularity offered by OFDM,
`a smooth migration of, for example, 2G spec-
`trum is possible. A 2G GSM operator can in
`principle migrate on a 200 kHz GSM carrier-by-
`carrier basis by using only a fraction of the avail-
`able OFDM subcarriers. Frequency-division
`duplex (FDD), time-division duplex (TDD), and
`combined FDD/TDD, as illustrated in Fig. 3, are
`supported to allow for operation in paired as
`well as unpaired spectrum.
`
`Downlink: OFDM with Frequency Domain
`Adaptation — The basic time-frequency struc-
`ture of the OFDM downlink is illustrated on the
`left of Fig. 4. A subcarrier spacing of 15 kHz is
`adopted, allowing for simple implementation of
`dual mode Rel-6/LTE terminals as the same
`clock frequencies can be used. To minimize
`delays, the subframe duration is selected as short
`as 0.5 ms, corresponding to seven OFDM sym-
`bols. The cyclic prefix length of 4.7 µs is suffi-
`cient for handling the delay spread for most
`unicast scenarios, while only adding modest
`overhead. Very large cells, up to and exceeding
`120 km cell radius, with large amounts of time
`dispersion are handled by reducing the number
`of OFDM symbols in a subframe by one in order
`to extend the cyclic prefix to 16.7 µs. Broadcast
`services are supported by transmitting the same
`information from multiple (synchronized) base
`stations. To the terminal, the received signal
`from all base stations will appear as multipath
`propagation and thus implicitly be exploited by
`the OFDM receiver.
`Exploiting channel variations in the time
`domain through link adaptation and channel-
`dependent scheduling, as is done in current 3G
`systems such as WCDMA and HSDPA, has been
`shown to provide a substantial increase in spec-
`tral efficiency. With the evolved radio access,
`this is taken one step further by adapting the
`transmission parameters not only in the time
`domain, but also in the frequency domain. Fre-
`quency domain adaptation is made possible
`through the use of OFDM and can achieve large
`performance gains in cases where the channel
`varies significantly over the system bandwidth.
`Thus, frequency domain adaptation becomes
`increasingly important with an increasing system
`bandwidth. Information about the downlink
`channel quality, obtained through feedback from
`the terminals, is provided to the scheduler. The
`scheduler determines which downlink chunks to
`
`allocate to which user and dynamically selects an
`appropriate data rate for each chunk by varying
`the output power level, the channel coding rate,
`and/or the modulation scheme. Quadrature
`phase shift keying (QPSK), 16-quadrature ampli-
`tude modulation (16-QAM), and 64-QAM mod-
`ulation schemes are supported in the downlink.
`
`Uplink: Single-Carrier FDMA with Dynamic
`Bandwidth — For uplink transmission, an
`important requirement is to allow for power-effi-
`cient user-terminal transmission to maximize
`coverage. Single-carrier frequency-domain multi-
`ple access (FDMA) with dynamic bandwidth,
`illustrated on the right of Fig. 4, is therefore pre-
`ferred. For each time interval, the base station
`scheduler assigns a unique time-frequency inter-
`val to a terminal for the transmission of user
`data, thereby ensuring intracell orthogonality.
`Primarily time domain scheduling is used to sep-
`arate users, but for terminals with limitations in
`either transmission power or the amount of data
`awaiting transmission, frequency domain
`scheduling is also used. Note that a terminal is
`only assigned chunks contiguous in the frequen-
`cy domain to maintain the single-carrier proper-
`ties and thereby ensure power-efficient
`transmission. Frequency domain adaptation is
`typically not used in the uplink due to lack of
`channel knowledge, as each terminal cannot con-
`tinuously transmit a pilot signal covering the
`whole frequency domain. Slow power control,
`compensating for path loss and shadow fading, is
`sufficient as no near-far problem is present due
`to the orthogonal uplink transmissions.
`Multipath propagation is handled by frequen-
`cy domain equalization at the base station, aided
`by the insertion of a cyclic prefix in the transmit-
`ted signal. Transmission parameters, coding, and
`modulation are similar to the downlink transmis-
`sion.
`
`MULTI-ANTENNA SOLUTIONS
`In order to fulfill the requirements on coverage,
`capacity, and high data rates, various multi-
`antenna schemes need to be supported as part
`of the long-term 3G evolution. For example,
`beamforming can be used to increase coverage
`and/or capacity, and spatial multiplexing, some-
`times referred to as MIMO, can be used to
`increase data rates by transmitting multiple par-
`allel streams to a single user. However, conven-
`tional multi-antenna diversity techniques, at both
`
`42
`
`IEEE Communications Magazine • March 2006
`
`Ex. 1014 / Page 5 of 8
`
`

`

`EKSTRÖM LAYOUT 2/14/06 11:03 AM Page 43
`
`Multipath
`propagation is
`handled by
`frequency domain
`equalization at the
`base station, aided
`by the insertion of a
`cyclic prefix in the
`transmitted signal.
`Transmission
`parameters, coding,
`and modulation are
`similar to the
`downlink
`transmission.
`
`Traffic models
`
`User distribution
`
`Terminal speed
`
`Data generation
`
`Radio network models
`
`Uniform, on average10 users/sector
`
`0 km/h
`
`On-off with activity factor 20, 40, 60, 80, 100 percent
`
`Distance attenuation
`
`L = 27.5 + 37.1 * log(d), d = distance in meters
`
`Shadow fading
`
`Multipath fading
`
`Cell layout
`
`Log-normal, 8 dB standard deviation
`
`3GPP typical urban
`
`Hexagonal grid, 3-sector sites, 21 sectors in total
`
`Cell radius (intersite distance = 3 × radius)
`
`500–2250 m
`
`System models
`
`Spectrum allocation
`
`20 MHz (4 x 5 MHz for WCDMA Rel-6)
`
`Base station output power
`
`40 W into antenna (10 W/5 MHz carrier for WCDMA)
`
`Maximum antenna gain
`
`18 dBi
`
`Modulation and coding schemes
`
`QPSK and 16-QAM, turbo coding according to WCDMA Rel-6
`
`OFDM parameters
`
`According to an earlier section
`
`WCDMA receiver
`
`Two-branch antenna diversity with rake receiver, maximum
`ratio combining of all channel taps; 7 dB noise figure
`
`Evolved RAN MIMO scheme
`
`Two streams with per-antenna rate control (PARC)
`
`Evolved RAN receiver
`
`Two-branch MMSE, 7 dB noise figure
`
`Scheduling
`
`Round-robin in time domain
`
`nnnnTable 1. Simulation models and assumptions (MMSE: minimum mean square error).
`
`the receiver and transmitter, will also play an
`important role in fulfilling the requirements.
`It is necessary to consider multi-antenna tech-
`nologies as a well-integrated part of the evolved
`radio access, and not as an extension added to
`the specification at a later stage. The potential
`of using the spatial domain is large, and the
`development of new and even more efficient
`multi-antenna algorithms is expected to continue
`for a long time into the future. Hence, to make
`the evolved radio access future-proof, it should
`be able to support new and improved multi-
`antenna algorithms in an efficient manner. In
`addition to an initial negotiation between the
`transmitter and receiver about the transmission
`scheme, this can be achieved by using the follow-
`ing key components:
`• Multipurpose measurement signals
`• Adjustable preprocessing rules for the mea-
`surements
`• A few well-defined measurement result for-
`mats
`It should be possible for the transmitter to
`send several multipurpose measurement signals.
`
`These measurement signals should be orthogo-
`nal to each other, and the receiver needs no
`knowledge of the spatial properties of the signal
`(i.e., the antenna pattern or the beam pattern
`used for transmission). The receiver only needs
`to be informed of which signals to measure.
`Different multi-antenna algorithms require
`measurements with different resolutions in the
`time, frequency, space, and stream domains. The
`Doppler spread of the radio channel and the
`velocity of the UE will also affect which resolu-
`tions are appropriate. By using a set of
`adjustable preprocessing rules for the receiver, it
`is possible to adapt the measurement resolution
`to the current conditions. For example, the aver-
`aging of measurements in time could be
`selectable between 0.5 ms and 100 ms, the aver-
`aging in frequency could be selectable between
`300 kHz and 5 MHz, the averaging in space
`could be selectable between one antenna and all
`antennas, and the averaging in stream domain
`could be selectable between one stream and all
`streams.
`For a large group of multi-antenna schemes,
`
`IEEE Communications Magazine • March 2006
`
`43
`
`Ex. 1014 / Page 6 of 8
`
`

`

`Evolved RAN
`WCDMA R6
`
`EKSTRÖM LAYOUT 2/14/06 11:03 AM Page 44
`
`2
`
`1.8
`
`1.6
`
`1.4
`
`1.2
`
`1
`
`0.8
`
`0.6
`
`0.4
`
`0.2
`
`0
`
`Max normalized system throughput (b/s/Hz/sector)
`
`Evolved RAN mean
`Evolved RAN 5th perc
`WCDMA R6 mean
`WCDMA R6 5th perc
`
`3.5
`
`3
`
`2.5
`
`2
`
`1.5
`
`1
`
`0.5
`
`0
`
`Normalized active radio link bit rate (b/s/Hz)
`
`0
`
`0.2
`
`0.4
`0.6
`0.8
`1
`1.2
`1.4
`1.6
`Normalized served traffic (b/s/Hz/sector)
`
`1.8
`
`2
`
`400
`
`600
`
`800 1000 1200 1400 1600 1800 2000 2200
`Cell radius (m)
`
`2400
`
`nnnnFigure 5. Mean and 5th percentile normalized active radio link bit rate vs. traffic load for a cell radius of 500 m (left); normalized cell
`throughput vs. cell radius in fully loaded systems (right).
`
`such as various open-loop beamforming schemes,
`open-loop transmit diversity, and basic spatial
`multiplexing techniques, a requested data rate is
`a sufficient measurement result format. For
`other multi-antenna schemes, more feedback
`information about the radio channel is needed.
`By letting the transmitter specify how many bits
`should be used to represent the phase and ampli-
`tude, respectively, various multi-antenna schemes
`that require knowledge of the radio channel will
`be supported (e.g., closed loop transmit diversity
`and eigenvalue-based MIMO).
`
`PERFORMANCE EVALUATION
`In this section a simple radio network perfor-
`mance evaluation of a possible evolved RAN
`concept is presented. The intention is to indicate
`whether the performance requirements present-
`ed earlier can be met. To this end, assessments
`of downlink user quality, capacity, and coverage
`of a system employing a possible evolved RAN
`concept are made, and compared to a Rel-6 sys-
`tem based on WCDMA using the system config-
`urations mandated in [3].
`Simple models and assumptions are used.
`A summary grouped into traffic, radio net-
`work, and system models is provided in Table
`1. The ambition is to achieve relative assess-
`ments of the gains associated with OFDM and
`MIMO. Frequency domain adaptation and
`other higher-layer improvements of the
`evolved RAN are not included in the evalua-
`tion. It should also be noted that many con-
`trol plane and user plane protocol aspects
`above the physical layer are omitted, yielding
`optimistic absolute values.
`A simple static simulation-based evaluation
`methodology is used. In each iteration of the
`simulation, terminals are randomly positioned in
`the system area, and the radio channel between
`each base station and terminal antenna pair is
`calculated according to the propagation and fad-
`ing models. To study different system load lev-
`
`els, base stations are randomly selected to trans-
`mit with an activity factor f ranging from 20 to
`100 percent. In cells with active base stations, a
`single receiving user is selected independent of
`channel quality. This models channel-indepen-
`dent time domain scheduling (e.g., round-robin).
`The total number of active users for activity fac-
`tor f is denoted U(f). Based on the channel real-
`izations
`and
`active
`interferers,
`a
`signal-to-interference-plus-noise ratio (SINR) is
`calculated for each terminal receive antenna.
`Using results from link-level simulations, includ-
`ing HARQ, the SINR values are then mapped
`to active radio link bitrates Ru for each active
`user u. In the case of MIMO, Ru is modeled as
`the sum of the rates achieved per MIMO stream.
`Note that R u is the bit rate user u gets when
`scheduled. When the channel is shared between
`multiple users, a correspondingly lower bit rate
`than R u is experienced above the MAC layer.
`Active base stations and users differ between
`iterations, and statistics are collected over a
`large number of iterations.
`For each activity factor, the served traffic T(f)
`is calculated as the sum of the active radio link
`U(f) Ru),
`bit rates for the active users (i.e., T(f) = Σu=1
`and the mean and 5th percentile of the active
`radio link bitrate are used as measures of aver-
`age and cell-edge user quality, respectively. Note
`that as the activity factor increases, individual
`active radio link bit rates decrease because of
`increased interference and thereby decreased
`SINR. The served traffic, however, increases as
`the number of active users increases.
`Figure 5a shows the mean and 5th percentile
`(cell-edge) active radio link bit rate vs. served
`traffic for a cell radius of 500 m. The bit rates
`are normalized with the spectrum allocation to
`enable comparison between the evolved RAN
`and WCDMA Rel-6. It is seen that the evolved
`RAN concept yields significantly improved bit
`rates for both average and cell-edge users. Com-
`paring the bit rates at a served traffic of 0.75
`b/s/Hz/sector, gains in cell-edge and mean active
`
`44
`
`IEEE Communications Magazine • March 2006
`
`Ex. 1014 / Page 7 of 8
`
`

`

`The fun