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.
The manual you are reading is protected by copyright law. This means that Wray Castle Limited could take you and
your employer to court and claim heavy legal damages.
Apart from fair dealing for the purposes of research or private study, as permitted under the Copyright, Designs andPatents Act 1988, this manual may only be reproduced or transmitted in any form or by any means with the prior
permission in writing of Wray Castle Limited.
All of our paper is sourced from FSC (Forest Stewardship Council) approved suppliers.
LTE (Long Term Evolution) represents the next developmental step for the 3GPP (3rd Generation
Partnership Project) standards group. It provides for a continued evolutionary path from 2G GSM/GPRS
(General Packet Radio Services), beyond 3G UMTS/HSPA and ultimately towards a 4G solution.
UMTS (Universal Mobile Telecommunications System) has continued to build on the success of GSM
(Global System for Mobile Communications) and momentum is gathering behind its significantlyincreased capability with the introduction of HSPA (High Speed Packet Access). The classic fixed and
mobile telecommunications business models are undergoing enormous change with the move towards
all-IP (Internet Protocol) switching and a total-communications service profile. Meanwhile, the last
decade has seen the Internet develop into a serious business tool and fixed broadband access is fast
becoming a basic commodity.
This market landscape is ready for a technology that combines broadband capabilities with an efficient,
scalable switching infrastructure and a flexible service delivery mechanism. LTE provides just such a
solution and is designed to address growing global demand for anywhere, anytime broadband access
while maintaining efficient provision of more traditional telecommunications services and maximizing
compatibility and synergies with other communications systems.
Although LTE most obviously represents an evolutionary path for UMTS networks it has also been
designed to allow cost-effective upgrade paths from other technology starting points. For example, GSM
operators now have the possibility to access 3G-like performance through EDGE (Enhanced Data rates
for Global Evolution), and this in turn can be used as a direct pathway to LTE. Similarly, the interworking
capabilities of the EPC (Evolved Packet Core) make it possible for CDMA (Code Division Multiple
Access) to migrate radio access from 1x or 1xEV-DO (1x Evolution – Data Only) to LTE.
Evolution beyond LTE has been mapped out by 3GPP with the specification of LTE-Advanced, which
offers the possibility of downlink data rates (to stationary or low mobility users) of 1GBit/s or more.
Further Reading: 3GPP TS36.300 (LTE Radio Access), 23.401 (LTE Core Network)
Wide-area LTE radio access combined with the EPC represents a complete adoption of an all-IP
architecture, offering broadband delivery capability with the potential for bit rates of several hundred
megabits per second and QoS (Quality of Service) management suitable for real-time operation of high-
quality voice and video telephony.
LTE has a very important role in the overall telecommunications service convergence concept. LTE couldprovide a key to unlocking a truly converged fixed/mobile network for the delivery of quadruple play
services. Its potential bandwidth capabilities are sufficient for the support of services ranging from
managed QoS real-time voice or video telephony to high-quality streamed TV. Its flat all-IP architecture
means that it can act as a universal access network for a wide range of core network types.
Since the publication of the first GSM specifications in the late 1980s, the technologies and techniques
employed by GSM networks have continually evolved and developed. GSM itself underwent a series of
changes, from Phase 1 to Phase 2 and eventually to Phase 2+. Phase 2+ progressed with a series of
yearly releases, starting with Release 96.
The UMTS was introduced as part of Release 99 and from then onwards the 3GPP 3G networktechnology has also been undergoing a process of evolution. The evolutions that particularly affect the air
interface are mainly contained in Releases 5, 6, 7 and 8. Releases 5 and 6 introduced HSPA – HSDPA
(High Speed Downlink Packet Access) in R5 and HSUPA (High Speed Uplink Packet Access), or
Enhanced Uplink, in R6. Release 7 outlines the changes necessary to deliver HSPA+ and Release 8
specifications begin to describe LTE – the Long Term Evolution of UMTS.
Specification of LTE, generally described as 3.9G, was completed in Release 9. Specification of LTE-
Advanced, a full 4G solution, is detailed in Release 10.
Further Reading: 3GPP Release Descriptions – www.3gpp.org/ftp/Information/WORK_PLAN/Description_Releases/
Tests and evaluations carried out during 2007 led to the publication of the Release 8 36-series of
specifications, which began to detail the technological basis for LTE.
Of the original four candidate air interface technologies, two were chosen for the final version: OFDMA
(Orthogonal Frequency Division Multiple Access) and SC-FDMA (Single Carrier FDMA).
OFDMA is employed on the LTE downlink and is expected eventually to provide peak data rates
approaching 360 Mbit/s in a 20 MHz channel. SC-FDMA is employed on the LTE uplink and may deliver
up to 86 Mbit/s. SC-FDMA is also sometimes known as DFT-FDMA.
In addition to the air interface technologies, LTE simplifies the range of technologies employed in other
parts of the network.
LTE is an ‘all-IP’ environment, meaning that all air interface, backhaul and core network interfaces will
carry only IP-based traffic. The need to support different protocols for different traffic types, as was the
case with R99, is therefore avoided.
In this all-IP environment, layer 4 transport layer functions for signalling connections are performed usingan alternative to the traditional choices, TCP (Transmission Control Protocol) or UDP (User Datagram
Protocol).
SCTP (Stream Control Transmission Protocol) was developed with the needs of IP-based signalling in
mind and is used to manage and protect all LTE signalling services.
The basic building blocks of the E-UTRA access network are the eNB (Evolved Node B) plus backhaul –
and nothing else. All layers of the air interface protocol stack, including the elements that previously
resided in the RNC (Radio Network Controller) – RRC (Radio Resource Control), RLC (Radio Link
Control) and MAC (Medium Access Control) – have been moved out to the base station. As the eNB now
anchors the main backhaul link to the core network, it has also assumed responsibility for managing the
PDCP (Packet Data Convergence Protocol) service, which provides header compression and cipheringfacilities over the air interface.
HSDPA began the process of moving RRM (Radio Resource Management) functions, such as packet
scheduling, from the RNC to the Node B. In LTE, all remaining RRC functions are devolved to the eNB,
meaning that there is no longer a role for a device such as the RNC.
Among the RRM functions now devolved to the eNB are radio bearer control, radio admission control,
connection mobility control and the dynamic allocation (via scheduling) of resources to UEs (User
Equipments) in both uplink and downlink directions.
Following on from innovations in R4 and R5 networks, LTE also supports the concept of flexible
associations between access and core network elements, meaning that each eNB has a choice of MME(Mobility Management Entity) nodes to which to pass control of each UE. Dynamic selection of an MME
for each UE as it attaches is therefore also an eNB responsibility. An eNB may be associated with MMEs
belonging to different PLMNs (Public Land Mobile Networks), allowing for the easy creation of multi-
operator networks.
The eNB also receives, schedules and transmits control channel information in its cells, including paging
messages and broadcast system information, both of which are received from the MMEs. It retains many
of the traditional roles associated with base stations, such as bearer management. It is responsible for
routing U-plane traffic between each UE and its S-GW (Serving Gateway). The complexity of the eNB
and of the decisions it is required to make are therefore much greater than for an R99 Node B
The broadening of the range of services offered by the LTE EPS over time has led to the development of
several specialised sub-types of eNB. Femtocell services, for example, are provided via HeNBs (HomeeNBs), whilst LTE Relay facilities are offered by Relay Nodes and controlled by DeNBs (Donor eNBs).
If supported, logical X2 interfaces can be physically transported along either direct or indirect
connections.
A direct connection would require a point-to-point broadband connection to exist between the two related
eNB sites. This option offers advantages in terms of resilience, in the sense that if multiple physical
connections are supported the loss of one transmission link would not be catastrophic, but hasdisadvantages in terms of cost. If each eNB was expected to host connections to five or six neighbouring
sites, for example, the costs associated with the additional transmission requirements could be
unsustainably high.
Another disadvantage of using direct connections to support X2 interfaces is lack of flexibility. The LTE
E-UTRAN is designed to take advantage of a concept known as the SON (Self-Organizing Network). The
optional SON functionality supported by the eNB allows it to attempt to establish an X2 interface
connection automatically to any previously unknown local cells reported in UE measurements. Such
automatic discovery and connection is only possible when all local eNBs are connected to the same
common routing environment.
Most X2 connections can be expected to share the eNBs core transmission link with the S1 interface. X2traffic would then simply be routed back out towards the target eNB after arriving at a suitable E-UTRAN or
EPC IP router. The benefit of this approach, which was the preferred method of carrying Iu-CS and Iu-PS
connections between remote RNCs and the core network site in 3G networks, is that only one
transmission link per eNB is required. The disadvantage is that only one transmission link per eNB is
available, which introduces the potential for a lack of access network resilience.
Operators may decide to deploy a combination of direct and indirect X2 routing, with some heavily used
links between eNBs being provided with their own direct connections whilst other, less heavily used,
The reduced complexity in the RAN is mirrored by a similar reduction in the core network, where the EPC
(Evolved Packet Core) structure consists of five main nodes, although others may be required for
backwards-compatibility purposes.
The MME handles control plane functions related to mobility management (authentication and security)
and idle mode handling (location updates and paging), in which sense it is broadly analogous to the VLR(Visitor Location Register) or GMM (GPRS Mobility Management) functions found in legacy networks.
The MME is also responsible for EPC bearer control, and so handles connection control signalling.
The S-GW and PDN-GW (Packet Data Network Gateway) are broadly analogous to the SGSN (Serving
GPRS Support Node) and GGSN (Gateway GPRS Support Node) found in R99 networks and perform
user plane handling, switching/routing and interfacing functions. Unlike legacy systems, however, bearer
control has been removed from these devices and resides with the MME.
The PCRF (Policy and Charging Rules Function) is an optional network element. If deployed, it handles
QoS (Quality of Service) and bearer policy enforcement and also provides charging and rating facilities. If
a PCRF is not deployed then some of its functions can instead be performed by the PDN-GW.
Subscriber management and security functions are handled by the HSS (Home Subscriber Server),
which incorporates the functions of the legacy HLR (Home Location Register) and which is already
familiar from R5 elements such as the IMS (IP Multimedia Subsystem).
For backwards-compatibility purposes, SGSNs deployed to legacy parts of an operator’s network can be
interfaced to both the MME (for mobility management) and the S-GW (for user plane flows).
The MME then provides legacy systems with an interface to the HSS, and the S-GW and PDN-GW
assume the role previously performed by the GGSN.
The packet data services of legacy (GSM/GPRS, R99 and HSPA) networks and LTE/SAE systems can
therefore interwork via a unified set of core network elements if required.
The EPS is designed to provide IP connectivity between a UE and a PDN (Packet Data Network).
The connection provided to a UE is referred to as a PCS (PDN Connectivity Service), sometimes known
as just a PDN Connection.
This consists of one or more EPS bearer that connect(s) the UE to an Access Point in a PDN-GW andtraverses both the E-UTRAN and the EPC. The PDN-GW routes traffic between the EPS bearer and the
external PDN.
The EPS bearers, in turn, carry a Traffic Flow Aggregate that consists of one or more SDF (Service Data
Flow) connections between the UE and external data services.
If a UE requires additional connectivity that is only available via a different PDN-GW Access Point, then
additional PDN Connectivity Services may be established in parallel.
The MME is primarily a signalling node and each MME has to be accessible to and exchange control
data with MMEs and other devices within its own network and in other networks elsewhere in the world.
For this reason, each MME is assigned a unique and globally significant identifier known as a GUMMEI.
The GUMMEI consists of the network’s MCC and MNC followed by a MMEI (MME Identifier), which in
turn consists of the MMEGI and the MMEC. The MMEGI identifies the pool to which the MME belongsand the MMEC is its index within that pool.
The addressing of S-GW and PDN-GW nodes follows the model for addressing legacy PS (Packet
Switched) core network nodes – ultimately, each node will be identified by an IP address, which may or
may not be backed up with a DNS-resolvable device name. The termination and anchor point for an EPS
Bearer is an access point in a PDN-GW, which is analogous to a PDP Context terminating on GGSN
APN in 2G/3G networks. Each PDN-GW AP is assigned an IP address associated with a DNS-resolvable
name – the APN.
The EPS ECGI is globally unique and allows individual cells to be separately identified. The ECGI is a
28-bit identifier which consists of the PLMN ID (MCC + MNC), a 20-bit eNB ID (which will be unique
within a PLMN) and an 8-bit Cell ID (which will be unique within one eNB). This gives each PLMN scopeto identify up to 1 million eNBs and for each eNB to control up to 256 cells.
Consider a radio carrier being modulated by a 10 kbit/s bit steam using QPSK (Quadrature Phase Shift
Keying). It could be expected to see a spectral envelope following a (sin x)/ x function, as shown in the
diagram, with the first null located 5 kHz from the centre frequency.
In a classic FDM (Frequency Division Multiplexing) system, other radio carriers would be allocated and
spaced far enough away from the first to ensure minimal adjacent channel interference. The size of theguard band required would depend on the transmitter and receiver characteristics as well as the relative
powers.
However, in such a system it is assumed that there is no synchronization between the potential
interferers. It is this that leads to the need for large frequency spacing between adjacent carriers. In fact,
if there was synchronization between adjacent channels, a much smaller frequency spacing could be
used. The key is to be able to make use of the complex nature of the instantaneously transmitted
spectrum. The modulation envelope is only an artificial way of indicating all possibilities over time; a
snapshot at an instant in time would look different.
Consider a second radio carrier allocated such that its centre frequency coincides exactly with the null
in the first carrier’s envelope. It is using the same modulation scheme and carrying the same data rate.The result is as shown. Note that the carrier spacing of 5 kHz is the same magnitude as the symbol rate
of 5 ksps. The spectra of the two carriers now overlaps, but as long as the carrier frequencies and the
baseband data remain accurately synchronized, both can be demodulated successfully. The reason is
that this relationship between centre frequency offset and symbol rate maintains a high level of
Considering again the two overlapping QPSK radio carriers, it can be seen that there is a relatively large
spectral efficiency gain. If the effective bandwidth of the transmitted signal is considered to be the
frequency separation of the first nulls then a single QPSK carrier modulated with 10 kbit/s would have a
null-to-null bandwidth of 10 kHz.
However, here there are two subcarriers, each of which is carrying 10 kbit/s using QPSK. Their respective null-to-null spectra overlap by 5 kHz. This gives a collective null-to-null bandwidth for the pair
of subcarriers of 15 kHz. Thus QPSK is being used to carry 20 kbit/s in a radio bandwidth of 15 kHz.
Note that a single QPSK modulated carrier carrying 20 kbit/s would result in a null-to-null bandwidth of
20 kHz.
The principle of independent reception of orthogonal radio carriers with overlapping spectrum can be
extended by using a large number of narrowband radio carriers within one wideband channel allocation.
This results in a very spectrally efficient channel that can carry high bit rates.
For example, if 1000 orthogonal radio carriers were modulated using QPSK, each carrying 10 kbit/s, the
net throughput for the channel would be 10 Mbit/s. This would require a total channel bandwidth of
slightly more than 5 MHz. Carrying the same bit rate with QPSK modulation onto a single radio carrier would require a null-to-null bandwidth of 10 MHz. Thus OFDM (Orthogonal Frequency Division
Multiplexing) almost doubles the spectral efficiency. Moreover, the resulting OFDM transmission is more
Spectral efficiency is not the only benefit associated with using OFDM. It also exhibits good tolerance to
the effects of multipath propagation in the channel; both fading and time dispersion.
Because the data rate on individual subcarriers with the channel is very low, the symbol period is
correspondingly long. The resulting symbol period is typically significantly longer than the time dispersion
that occurs in the channel. This means that relatively simple equalization can be used to counteractmultipath even though the net rate in the whole channel is very high.
Furthermore, a guard period can be inserted in every symbol period that covers the expected time
dispersion for the channel. This removes most of the time dispersion distortion from the useful symbol
period.
This guard period is usually created by repeating a copy of the last part of the symbol at the start. In this
Tolerance to multipath fading effects comes from the overall wideband characteristic in the channel. A
narrowband channel tends to exhibit flat fading characteristics; that is to say, the fading characteristics
are coherent across the whole channel bandwidth. The effects of this can be seen in the diagram.
OFDM channels, on the other hand, are usually used to carry very high data rates and therefore require
many subcarriers occupying a relatively large bandwidth. In most cases the bandwidth will exceed thecoherence bandwidth by a large factor, so differing fading characteristics will be seen in different parts of
the channel. In effect, the wide channel provides a degree of frequency diversity with a resulting
improvement in performance.
However, it would be wrong to assume that this benefit for OFDM results solely because the channel
bandwidth is wide. A single carrier system with the same bit rate would also result in a wide radio
channel. Therefore, a single carrier system also benefits from this form of frequency diversity to some
extent.
In the single channel system, energy from each symbol will be spread across the whole radio channel
and each symbol will therefore suffer some distortion from any fading that may occur in any one part of
the channel. In an OFDM system only those symbols transmitted on subcarriers in the part of the channelaffected by a fade will be distorted. Symbols transmitted on other subcarriers will remain unaffected. It is
then possible to adapt the subcarriers in use according to the varying fading characteristics. This means
The diagram shows a block representation of the transmitter that brings together the elements of symbol
mapping for QAM (Quaderature Amplitude Modulation) and the application of the IFFT (Inverse Fast
Fourier Transform) in order to produce an OFDM signal.
The serial data to be carried on the radio link is first passed through a serial-to-parallel conversion
process. The number of parallel streams will be equivalent to the number of data-carrying subcarriers inthe system. This number will usually be a power of two since this makes best use of the efficiencies
offered by the IFFT.
Bits on the parallel data streams will also be grouped as appropriate for the symbol constellation of M-ary
QAM scheme in use. For example, for QPSK bits are grouped in pairs; for 16QAM they are grouped in
fours and for 64QAM they are grouped in sixes.
The next process is symbol point mapping for the bit groups on each parallel data stream. The resulting
complex number symbols then form the input to an N-point IFFT where N will be a power of two
equivalent to the number of subcarriers in use.
The output of the IFFT will be a series of complex number digital samples representing the OFDM signalduring each symbol period. At this point the cyclic prefix is added by copying the last samples onto the
beginning of the symbol period. These complex real and imaginary sample values are used to form the I
and Q symbol streams. Next, the I and Q branches are subsequently multiplied onto sine and cosine
representations of the radio carrier. This generates a digital representation of the required multicarrier
M-ary QAM modulated transmit signal.
After digital-to-analogue conversion the resulting signal can be up-converted to the required channel
centre frequency before amplification and transmission.
The simplest option for multiple access in an OFDM system is to use a form of time multiplexing on the
OFDM radio bearer. This is illustrated in the top part of the diagram. Each user is allocated the full
channel bandwidth and all data subcarriers exclusively for a defined number of symbol periods.
The greatest efficiency can be achieved if dynamic time allocation is applied so that users with higher bit
rate requirements are allocated a greater proportion of time. However, in such a system the minimumresource allocation is one OFDM symbol. Even with dynamic time allocation, such an arrangement can
still become very inefficient when there is strong demand for multiple lower bit rate connections, for
example when multiple voice circuits are active. Consider an OFDM system operating in a 10 MHz
bandwidth, with a 512-point FFT and using 16QAM. Allowing for null and reference subcarriers, such a
system could transfer in the order of 1,600 bits in a single OFDM symbol period. This may seem a
modest resource unit, but delay requirements must also be accounted for. For a real-time service such
as voice it is essential to avoid excessive round-trip delay. To meet the delay requirement for a voice
service, resources may need to be allocated, for example once every 20 ms. This would mean in a
minimum bandwidth allocation to one user of 80 kbit/s (or 120 kbit/s if 64QAM is in use). Even allowing
for the error protection overhead this minimum resource will significantly reduce system efficiency and its
ability to benefit from optimal techniques such as discontinuous transmission and channel adaptation.
Greater efficiency in resource allocation can be gained from the use of subchannelization. This involves
division of resource by time and by frequency. Thus a user may be allocated a subset of the subcarriers
available in the system, as illustrated in the lower part of the diagram. This approach allows much finer
granulation in resource allocation and therefore greater efficiency. OFDM systems that support this are
The quality of the radio link is affected by many factors including fading, interference and time dispersion.
Terrestrial mobile radio channels, which are usually assumed to be non-line of site, can be very poor.
Therefore most terrestrial cellular radio systems are designed with robust modulation schemes and large
error protection overheads.
However, close examination of real channel conditions shows them to be very variable in short timeframes, and much of the time any given channel will show good performance. Thus the standard
approach engineers the channel to deal with the worst case, which only occurs for a small amount of
time.
It is clear that if the channel could be adapted at a rate fast enough to track changing channel conditions
then the average performance of a channel could be significantly improved. This is the principle of
channel adaptation. Channel adaptation is a common approach in many broadband radio systems and in
most cases involves the adaptation of the modulation scheme and the error protection overhead applied.
Adaptive scheduling can also be very effective, enabling the cell to make the best use of the pool of
channels allocated to different mobiles, each of which will be varying independently.
MIMO (Multiple Input Multiple Output) antenna arrays offer significant performance improvements over
conventional single antenna configurations.
The technique involves placing several uncorrelated antennas at both the receiving and transmitting ends
of the communication link. If there are four uncorrelated antennas at the transmitter and a further four
uncorrelated antennas at the receiver, then there will be 16 possible direct radio paths between thetransmitter and the receiver. Each of these is open to multipath effects, creating even more radio paths
between the transmitter and the receiver. These radio paths can then be constructively combined, thus
producing micro diversity gain at the receiver.
Since the receiver can distinguish between the various uncorrelated antennas, it is possible to transmit
different data streams in different paths. The stream applied to each antenna can be referred to as a
‘layer’ and the number of antennas available at the transmitter and receiver can be referred to as ‘rank’.
For example, a system operating with a 4x4 MIMO antenna array can be described as having four layers
and being of rank four. The way in which data streams are mapped to layers will change the specific
benefits offered by a particular MIMO implementation, and the specification of this is an important part of
system design. Pre-coding may also be used to improve the MIMO system performance. Pre-coding may
be adaptive and as such would be based on some source of channel estimation. This could be derived atthe transmission or the reception end of the link.
It is relatively easy to mount antennas on the base station in an uncorrelated manner. For a 2x2 MIMO
array a single cross-polar panel could be used. A 4x4 MIMO array would require two cross-polar polar
panels with suitable space separation. This is harder to achieve in a mobile. However, as for the base
station, 2x2 MIMO could be achieved with cross polarization, but this could result in some undesirable
MIMO is potentially a complex technology but it can provide very significant benefits in system capability.
There are three key ways in which MIMO improves system performance. Any given MIMO
implementation may make use of all these benefits or may be configured to take particular advantage of
one of them. Ideally, a system should be designed with sufficient flexibility in MIMO implementation to
allow a system operator to choose the most suitable implementation for different environments or system
goals.
Diversity gain arises out of the provision of multiple antennas at the transmitting and/or receiving end of
the radio link. This creates multiple transmission paths with decorrelated fading characteristics. The
result is an overall improvement in channel signal-to-noise ratio leading to increased channel throughput
and reliability.
Array gain refers to the beamforming capability of a multiple antenna array. With suitable signalling of
feedback from the receiver, or with measurements made on a return link, it is possible to direct radiated
energy toward the receiver in a steered beam. The result is improved channel performance and
increased throughput.
Spatial multiplexing gain arises out of the orthogonality between the multiple transmission paths createdby the multiple antenna array. Since the receiver can resolve independent transmission paths it is
possible to map different information streams into the transmission paths, identifiable by their spatial
signature. This results in a direct increase in the channel throughput in proportion to the number of
The basic implementation of MIMO is generally referred to as SU-MIMO (Single-User MIMO).
The SU-MIMO concept can be developed into MU-MIMO (Multi-User MIMO). In this case the spatial
multiplexing capability of MIMO is used to multiplex a link to more than one mobile using the same
time/frequency resource. The order of multiplexing available depends on the number of antennas (or
rank) available at the transmitter and receiver ends of the link. For example, the diagram shows a 2x2MIMO arrangement being used for MU-MIMO with two mobiles. In this case, the rate available to each
mobile would be lower than that potentially available to a single mobile with an SU-MIMO configuration,
but both mobiles are allocated the same time/frequency resource and still have the potential for diversity
and array gain. Thus cell capacity is increased, but the resource can be shared between a larger number
of users. The use of more than one transmitting or receiving station in this way is sometimes called
virtual MIMO.
It is also possible to implement MU-MIMO in one direction only with just single antennas on each of the
mobiles. In this case, array and diversity gain would be reduced, but time/frequency resources can still
be reused in the cell.
MU-MIMO can be further developed into multi-cell MU-MIMO. In this case the data streams are mappedto the combined antenna resources of two or more base stations that provide a combined connection to
multiple mobiles in multiple cells. The scenario in the diagram is in effect 4x4 MIMO but shared between
two connections. Note that spatial diversity will be significant in such a scenario because of the
geographical separation of the base station and of the mobiles.
E-UTRA supports services in a variety of channel bandwidths. In fact, the specification explicitly labels
E-UTRA as ‘bandwidth agnostic’, meaning that it has no rigidly defined or preferred channel bandwidth
and can be scaled to channels of almost any size. Both FDD and TDD modes are supported, as is a
‘half duplex’ mode.
E-UTRA has also been designed to work as the bearer for Multicast and Broadcast Multimedia Services(MBMS) and as such includes support for SFN (Single Frequency Network) operation.
Support for advanced antenna configurations has also been designed into the specification with MIMO
and beam-forming adaptive antennas both being referenced.
E-UTRA/LTE is designed to work in a variety of bandwidths ranging initially from 1.4 MHz to 20 MHz. As
E-UTRA is described as being ‘bandwidth agnostic’, other bandwidths, ones that allow E-UTRA to be
backwards compatible with channel allocations from legacy network types, for example, could be
incorporated in the future.
The version of OFDMA employed by E-UTRA is similar to the versions employed by WiMAX or DVB, butwith a few key differences. In systems such as WiMAX, OFDMA schemes occupying different channel
bandwidths employ different subcarrier spacing, meaning that there is a different set of physical layer
parameters for each version of the system.
The E-UTRA scheme allows for two fixed subcarrier spacing options, 15 kHz in most cases, with an
optional 7.5 kHz spacing scheme, only applicable for TDD (Time Division Duplex) operation and intended
for in very large cells in an SFN. Fixing the subcarrier spacing reduces the complexity of a system that
can support multiple channel bandwidths.
Further Reading: 3GPP TS 36.211, 36.101:5.5, 36.104:5.5
There is considerable regional variation in the availability of spectrum for LTE operation and this is
reflected in the standards. Along with flexibility in bandwidth there is considerable flexibility for spectrum
allocation. There are no requirements for minimum band support nor for band combinations. It is
assumed that this is determined by regional requirements.
The standards identify a range of bands for FDD operation, ranging from frequencies of approximately700 MHz through to frequencies in the range 2.7 GHz+. There also eight bands identified for TDD
operation ranging from approximately 1900 MHz to 2.6 GHz. Considerable scope has been left in the
standards to add more frequency bands as global requirements evolve.
Further Reading: 3GPP TS 36.101; 5.5, TS 36.104; 5.5
For both uplink and downlink operation subcarriers are bundled together into groups of 12. This grouping
is referred to as an RB (Resource Block). The RB also has a dimension in time and when this is
combined with the frequency definition it forms the basic unit of resource allocation.
The number of resource blocks available in the system is dependent on channel bandwidth, varying
between 100 for 20 MHz bandwidth to just six for 1.4 MHz channel bandwidth. The nominal spectralbandwidth of an RB is 180 kHz for the standard 15 kHz subcarrier spacing. Note that this means there is
a difference between the stated channel bandwidth and the transmission bandwidth configuration, which
is expressed as n x RB. For example, in a 5 MHz channel bandwidth the transmission bandwidth would
be approximately 4.5 MHz. This difference acts as a guard band.
OFDMA channels are allocated within an operator’s licensed spectrum allocation. The centre frequency
is identified by an EARFCN (E-UTRA Absolute Radio Frequency Channel Number). The precise location
of the EARFCN is an operator decision, but it must be placed on a 100 kHz raster and the transmission
bandwidth must not exceed the operator’s licensed spectrum.
Further Reading: 3GPP TS 36.101:5.6, 5.7; 36.104:5.6, 5.7
The range of modulation schemes used in E-UTRA comprises BPSK, QPSK, 16QAM (16-state
Quadrature Amplitude Modulation) and 64QAM (64-state Quadrature Amplitude Modulation). BPSK is
only employed for a limited set of signalling and reference functions, while 64QAM is optional on the
uplink.
The range of error coding options used in E-UTRA devices is far more limited than those available to, for example, a UMTS device. For most channels the only option is one-third rate turbo coding based on
convolutional coding.
Broadcast traffic channels are only permitted to use 1/3 Tail Biting convolutional coding. Various control
channels have been assigned either convolutional coding, block coding or simple repetition as their error
coding options.
In addition to error coding, transport blocks containing user and control traffic may also optionally have a
CRC (Cyclic Redundancy Check) block attached. Transport blocks on connections that have CRC
selected have a 24-bit CRC block appended to the end of the data container.
The familiar UMTS error monitoring levels of Bit Error Rate (BER), derived from the error coding service,and BLER (Block Error Rate), derived from CRC, continue to be available in E-UTRA.
The physical layer involves the transmission and reception of a series of physical channels and physical
signals. The physical signals relate to the transmission of reference signals, the PSS (Primary
Synchronization Signal) and the SSS (Secondary Synchronization Signal).
The PBCH (Physical Broadcast Channel) carries the periodic downlink broadcast of the RRC
MasterInformationBlock message. Note that system information from BCCH (Broadcast Control Channel)is scheduled for transmission in the PDSCH (Physical Downlink Shared Channel).
The PDCCH (Physical Downlink Control Channel) carries no higher layer information and is used for
scheduling uplink and downlink resources. Scheduling decisions, however, are the responsibility of the
MAC layer, therefore the scheduling information carried in the PDCCH is provided by MAC. Similarly the
PUCCH (Physical Uplink Control Channel) is used to carry resource requests from UEs that will need to
be processed by MAC.
The PHICH (Physical Hybrid ARQ Indicator Channel) is used for downlink ACK/NACK of uplink
transmissions from UEs in the PUSCH (Physical Uplink Shared Channel). It is a shared channel and
uses a form of code multiplexing to provide multiple ACK/NACK responses.
The PCFICH (Physical Control Format Indicator Channel) is used to indicate how much resource in a
subframe is reserved for the downlink control channels. It may be either one, two or three of the first
symbols in the first slot in the subframe.
The PRACH (Physical Random Access Channel) is used for the uplink transmission of preambles as part
of the random access procedure.
The PDSCH and the PUSCH are the main scheduled resource on the cell. They are used for the
transport of all higher-layer information including RRC signalling, service-related signalling and user
traffic. The only exception is the system information in PBCH.
There are two basic frame types employed in E-UTRA, which are common to both uplink and downlink.
Type 1 frames are employed for FDD full- and half-duplex systems, while Type 2 frames are reserved for
TDD operation only.
The Type 1 frame duration is 10 ms and it is divided into 20 slots, each of 0.5 ms duration. More
significantly, however, for most information transmission, two slots are combined to form a subframe.Thus subframe duration is 1 ms, which corresponds to the TTI (Transmission Time Interval) for
E-UTRA.
Type 1 slots contain either 7 or 6 symbols, depending upon which CP (cyclic prefix) type is in use.
Additionally, the length of the CP prefixed applied in a particular symbol within a slot varies, also
dependent on which CP length is in use. With the normal CP, symbol 0 in each slot has a CP equal to
160 x Ts or 5.2 µsec, while the remaining symbols in the slot have slightly shorter CPs of just 144 x Ts or
4.7 µsec. When using the extended CP, all symbols are prefixed with a CP of 512 x Ts or 16.7 µsec.
Scheduling occurs across a subframe period. Up to the first three symbols in the first slot of each
subframe can be defined as a ‘control region’ carrying control and scheduling messages. The remaining
symbols of the first and all symbols in the second slot within the subframe are then available for user traffic.
The diagram shows an example of a populated downlink FDD frame using the normal CP, 2x2 MIMO
and implemented in a 5 MHz bandwidth channel.
The PBCH is transmitted during subframe 0 of each 10 ms frame and occupies the centremost six
resource blocks. Alongside this and also in the sixth subframe in the frame are the primary and
secondary synchronization signals. Reference signal position for two resource blocks within a singlesubframe are shown for both antenna ports in the 2x2 MIMO system.
The diagram also shows the space allocated for downlink control channels, which includes PDCCH,
PCFICH and PHICH resources. A UE will be required to monitor some proportion of this dependent on
the connectivity state and the cell configuration.
The remainder of the allocation space will be used for scheduled downlink transmission in the PDSCH.
This includes common control signalling (system information and paging), dedicated control signalling
There are three sublayers within the E-UTRA layer 2, PDCP, RLC and MAC (Medium Access Control).
All the sublayers, including PDCP, span both the control and user planes of the protocol stack, although
in most cases the functions provided in each plane differ.
PDCP provides SAP (Service Access Point) access to protocol functionality through SRB (Signalling
Radio Bearer) provision in the control plane and DRB (Data Radio Bearer) provision in the user plane. Atthe eNB end a set of SRBs and DRBs is created on a per-UE basis as required. For system information
and paging, PDCP has a null function. PDCP provides sequencing of higher-layer PDUs and implements
the integrity and ciphering security functions as required.
RLC provides three levels of service through three SAP types, TM (Transparent Mode), UM
(Unacknowledged Mode) and AM (Acknowledged Mode). TM is only applicable to system information
broadcasting, paging and RRC connection establishment in SRB 0. AM is used for all dedicated
signalling functions and packet traffic transfer, providing retransmission and sequencing. For real-time
traffic, when AM would not be desirable in achieving the delay requirements UM can be used for
sequencing only.
MAC SAPs are known as logical channels. The MAC layer is responsible for mapping and multiplexinglogical channels to transport channels at the physical layer. MAC also controls scheduling for resource
allocation at the physical layer as well and control for a number of physical layer processes.
The MAC layer is defined as part of layer 2. However, many of its functions are closely related to physical
layer behaviour, so the architecture indicated in the standards should be treated as informative.
Manufacturers are left to determine an efficient implementation for the realization of MAC and physical
layer interaction.
The MAC layer is accessed through logical channels as well as a control SAP. It maps information flowsinto the physical layer through transport channels. The mapping of logical channels to transport channels
is a key function of the MAC layer.
In addition to channel mapping, the MAC layer has important control functionality including management
of multiple HARQ processes for each information flow and the random access process.
Most significantly, the MAC layer is responsible for channel prioritization and scheduling of resources on
the physical layer.
The MAC layer has a null function for paging and for system information that will be transmitted in the
The main function of the MAC is to manage the shared access to a common transmission medium by
multiple devices. This is achieved through the eNB’s scheduling function. Resource allocation will be
performed on the basis of a scheduling algorithm, the specifics of which are not defined by the standards.
However, channel performance, data buffer fill, UE power capability and traffic priority are likely to be
considered.
When a UE establishes an RRC relationship with an eNB it is assigned a C-RNTI (Cell Radio Network
Temporary Identifier), which will uniquely identify that UE in that cell. The C-RNTI will be used to address
any control and scheduling messages to or from the UE. Each UE is capable of establishing multiple EPS
bearers, which are the NAS traffic and signalling connections that travel from the UE to the core network.
Resource allocations are defined in terms of one or more PRB (Physical Resource Block), which will be
populated using a specified MCS (Modulation and Coding Scheme). The allocations can be made for one
or more TTI periods. LTE offers three scheduling modes. The first, known as dynamic scheduling ,
involves the use of MAC downlink assignment messages and uplink grant messages in the PDCCH to
allocate resources as required. Dynamic scheduling is intended for typical bursty packet data traffic.
For VoIP (Voice over IP) traffic where regular and reliable allocation of resources is required to meetmore demanding QoS requirements, LTE offers persistent scheduling . This is achieved through a
combination of RRC signalling in the DL-SCH (Downlink Shared Channel), for the initial specification of
the resource allocation interval, and MAC signalling in the PDCCH for more specific PRB and MCS
information. The result is a lower overhead in the PDCCH for these regular resource allocations. The
third scheduling option, known as semi-persistent scheduling , is used specifically for the purpose of
resource allocation for the establishment or reconfiguration of a persistent scheduled resource, i.e. for the
transport of RRC messages relating to the persistent scheduled resource. In this case an SPS-C-RNTI
(Semi Persistent Scheduling) will be used to address the UE, which is different from the UE’s C-RNTI.
The random access procedure is handled by the MAC and the physical layer and operates using a
combination of the PRACH on the uplink and the PDCCH on the downlink. UEs are informed of the range
of random access preambles available in system information, as are the contention management
parameters. When a random access event is required, the UE will perform the following functions:
review and randomly select a preamble
check the BCCH for the current PRACH configuration; this will indicate the location and periodicity
of PRACH resources in uplink subframes
calculate open loop power control parameters – initial transmit power, maximum transmit power
and power step
discover contention management parameters
Once the UE transmits an initial preamble it will wait a specified period of time for a response before
backing off and retrying. Open loop power control ensures that each successive retry will be at a higher
power level.
Upon receipt of a successful uplink PRACH preamble, the eNB will calculate power adjustment andtiming advance parameters for the UE based on the strength and delay of the received signal and
schedule an uplink capacity grant to enable the UE to send further details of its request. This will take the
form of the initial layer 3 message. If necessary, the eNB will also assign a Temporary C-RNTI for the UE
to use for ongoing communication.
Once received, the eNB reflects the initial layer 3 message back to the UE in a subsequent downlink
scheduled resource to enable unambiguous contention resolution. After this point further resource
allocations may be required for signalling or traffic exchange and these will be addressed to the
The table summarizes the RNTI types defined for E-UTRA. In all cases they have a length of 2 octets
and for some RNTI types there is a limited number range or specific reserved values. Outside of these
reserved values there is no structure to the RNTI.
A SPS-C-RNTI is allocated to a UE when Semi-Persistent scheduling is used and indicates resources for
higher-layer signalling that relates the UE's current persistently scheduled resource. The range of potential values will therefore be dependent on the PRACH configuration used in a cell. Any number in
this range cannot be allocated for use as any other RNTI type.
An Temporary C-RNTI is allocated to a UE on initial access as part of the random access procedure. On
successful completion of the random access procedure the Temporary C-RNTI becomes the
C-RNTI. This is cell specific and is the main identity for the UE within the cell.
A SPS-C-RNTI is allocated to a UE when persistent scheduling is used and indicates resources for
higher-layer signalling that relates the UE’s current persistently scheduled resource.
The fixed SI-RNTI (System Information RNTI) and P-RNTI (Paging RNTI) are used to indicate the
allocation of resources in the PDSCH containing system information or paging respectively.
TPC-PUCCH-RNTI (Transmit Power Control PUCCH RNTI) and TPC-PUSCH-RNTI are used to indicate
power control information for the PUCCH and PUSCH respectively.
RLC provides three levels of service: acknowledged mode, unacknowledged mode and transparent
mode. Radio bearers are mapped through RLC to logical channels and an RLC entity is created for each
active radio bearer.
For the transparent mode and the unacknowledged mode RLC entities are configured as either
transmitting or receiving entities. For acknowledged mode a single entity provides both transmit andreceive functionality for one side of the link. This configuration facilitates retransmission of failed RLC
Unacknowledged mode entities are accessed through a UM-SAP. Unacknowledged mode reorganizes
RLC SDUs into a size requested by the MAC layer. Unacknowledged mode also provides sequence
numbering for in-order delivery to higher layers at the receiving end. Reordering in the RLC layer is used
in support of the HARQ functions provided by the MAC layer.
Reorganization of RLC SDUs is provided by the segmentation and concatenation function. As shown inthe diagram, higher-layer SDUs can be fragmented and reassembled into the RLC PDU payload area to
produce a packet size suitable for scheduling by the MAC layer for transmission over the air interface.
The RLC header enables the receiving entity to reassemble the higher-layer SDU in the correct order.
The application of unacknowledged mode is limited to the user plane, where it would be utilized for
packet traffic flows with low tolerance to delay. The most common example would be VoIP connections.
The acknowledged mode of RLC is applicable in the control plane for RRC signalling messages carried
in DCCH and for user plane traffic carried in DTCH. Acknowledged mode entities are accessed through
an AM-SAP.
General transmission and reception functionality in terms of segmentation, concatenation, buffering and
HARQ reordering for AM mode are similar to those for UM mode. However, AM mode also providesretransmission of failed RLC PDUs. In this respect a number of enhancements in functional architecture
are provided. Firstly, a single entity for transmission and reception is required for interaction between the
transmitting and receiving side. Secondly a retransmission buffer is required in the transmit side. All
transmitted RLC PDUs are retained in the transmission buffer until acknowledgement is received.
Additionally, control (status) PDUs are required in addition to data PDUs in order to manage the
retransmission process. These must be multiplexed with data PDUs at the transmission end and
demultiplexed (routed) from data PDUs at the reception end.
A PDCP ent ity is created for each SRB and/or DRB on a per-UE bas is. Al l PDCP ent it ies are
bidirectional, thus when the AM mode of RLC is being used there is a one-to-one mapping between a
PDCP entity and AM SAP in RLC. However, for the UM mode of RLC one PDCP entity will be associated
with two UM SAPs, one configured for transmit functions and the other configured for receive functions.
Within a PDCP entity sequence numbering is applied for higher layer PDUs. This ensures in-order delivery at the receiving end. In the user plane PDCP control PDUs can be used to indicate missing
PDUs.
In the user plane, only IETF-defined ROHC (Robust Header Compression) is provided. Support for this is
only mandatory for UEs that have VoIP (Voice over IP) capability.
In the control plane, integrity protection is provided for RRC signalling messages.
Ciphering is then applied in both control and user planes, although separate cipher keys are applied for a
RRC exists only in the control plane of the air interface AS protocol stack. RRC receives information from
functional entities in the NAS (Non Access Stratum) in the form of complete messages for direct transfer,
and also in the form of requests, information elements and parameters that will trigger RRC activity and
be used in RRC messages.
For broadcast functions over the air interface RRC messages are mapped directly to logical channels.This includes paging and system information broadcasting using the PCCH and BCCH logical channels
respectively.
For dedicated signalling functions between a UE and an eNB signalling flows are mapped into an SRB.
When a UE transitions to the RRC connected state a set of SRB instances is created. SRB 0 is used only
for the initial establishment of the RRC connection and is mapped to the CCCH. Once the RRC
connection is established the UE will be issued with a C-RNTI and SRB 1 and optionally SRB 2 will be
created. SRB 1 is used for all RRC specific signalling functions. SRB 2 is used for RRC direct transfer of
NAS signalling messages. However, NAS messages may also be piggybacked with RRC signalling in
SRB 1. Both SRB 1 and SRB 2 are mapped to DCCH logical channels.
If required, one or more DRB may be created during or subsequent to an RRC connection establishment.These exist in the user plane and carry traffic. However, ‘traffic’ in this context includes service-related
signalling between service applications in higher layers, for example VoIP connection establishment
using the IMS. DRBs are mapped to DTCH logical channels.
The overall function of RRC is to create, maintain and clear DRBs as required to provide the radio link
segment of one or more EPS bearer relating to one or more EPS connectivity service. RRC receives
instructions on what EPS bearers are required from the NAS. The NAS activity in turn is driven by
instructions from service applications (via the PCRF on the EPC side).
In order to manage DRBs, RRC must exchange signalling with its peer entity and provide direct transfer for NAS signalling exchange. Connectivity for this comes from SRBs. However, signalling relating to
service applications, which are always external to the LTE/EPS, are treated as traffic flows and as such
are carried in DRBs within an EPS bearer. Note that an EPS bearer has only one set of associated QoS
characteristics, so if application signalling were to require different QoS treatment to the traffic that it
facilitates then a second EPS bearer would have to be defined. Multiple EPS bearers may or may not be
part of the same EPS connectivity service dependent on their respective connectivity requirements.
The RRC connection establishment procedure is always initiated from the UE. It begins with the
transmission of the RRCConnectionRequest message containing an identity and a cause value. If the UE
has already registered with the network then it will use the S-TMSI (SAE-TMSI) as its identity. If this is a
new mobile needing to perform an initial registration then it will generate and use a 40-bit random value.
The message is carried in the CCCH/UL-SCH channel combination. This requires a scheduled resource
allocation, which is secured using the lower-layer random access procedure and the RACH. The lower-layer random access procedure also facilitates the allocation of a C-RNTI at this stage.
The eNB responds with an RRCConnectionSetup message containing a transaction identifier, used to
relate future messages as part of this signalling sequence, and the radio resource configuration for SRB
1. Note that the exchange of the two messages to this point has involved the use of the implicitly
configured SRB 0.
The final part of this three-way handshake is the confirmation from the UE in the form of the
RRCConnectionSetupComplete message now using the defined SRB 1 and DCCH/UL-SCH
combination. For registered UEs this message contains identities of the PLMN and MME with which it is
registered. In any case the message will also piggyback the initial NAS message that triggered the RRC
establishment procedure, for example, a service request or registration message.
Further Reading: 3GPP TS 36.331:5.3.3, TS36.321:5.1
IP is the only packet transport mechanism employed by the EPS transport network. It does not support
the layer 2 transmission protocols employed in legacy systems such as TDM (Time Division Multiplexing)
and ATM (Asynchronous Transfer Mode).
An IP 'cloud' provides logical and physical interconnections between EPS network nodes. The design of
the cloud is intended to ensure that redundant paths exist between all nodes to allow the network tooperate in a resilient and fault-tolerant manner.
Equipment vendors and network operators have the option of deploying systems that support IPv4 (IP
version 4) or IPv6 (IP version 6) or a combination of both (functionality which is referred to by EPS nodes
as 'IPv4v6').
Compared to legacy IPv4, which has been in use since the early 1980s, IPv6 has a flatter protocol
structure – with many functions that required additional protocols in IPv4 being performed within the IP
layer itself in IPv6.
These additional features include functions such as dynamic IP address allocation, which required an
additional protocol such as DHCP (the Dynamic Host Configuration Protocol) in IPv4, but is managedautomatically (by means of router prefixes and host link local addresses) in IPv6. Support for security
mechanisms such as IPsec (IP security) are also incorporated into the IP layer in IPv6.
IPv6 is a backwards-compatible system, however, so network operators have the opportunity interface a
new IPv6-based EPS with existing IPv4-based legacy packet core networks.
GTP was originally designed as part of the 2.5G GPRS packet core network and was employed to route
encapsulated traffic packets between GPRS Support Nodes (SGSNs and GGSNs).
The initial 2.5G version of GTP became known as GTPv0.
As it matured, a number of basic problems were discovered. Chief amongst these was the fact that GTPv0placed tunnel control and administrative information in fields in the headers of packets that also
encapsulated user data. This meant packets that carried user data but no control information had a greater
amount of header overhead than necessary, leading to a potentially inefficient service.
GTPv1 was developed to offer an evolved service to 3G packet core networks. The most obvious
difference with GTPv0 was the extension of the service beyond the SGSN and down to the RNC.
Another major difference was the separation of the protocol into parts that dealt with control plane (GTP-C)
and user plane (GTP-U) traffic. GTP-U packet headers could therefore be smaller and offer a more
efficient service, as all control data was carried in its own logical stream.
discuss options for interworking the EPC with legacy packet core networks
describe the main points of interest related to EPC topics such as pooling
list the set of S interfaces described for the EPC and outline their basic functions and protocols
discuss options for User Plane connectivity between a UE and a PDN-GW
outline how combinations of redundant S interfaces can provide for EPC resilience
list the basic set of identifiers used to describe EPC areas
outline the set of node identifiers that have been defined for the EPC
discuss the impact of the evolved device/subscriber identifiers employed by the EPC
outline the fundamental properties of an EPS Bearer and describe the structure of an EPS Bearer ID
describe the relationship that exists between an EPS Bearer and an E-RAB
outline the role of the APN (Access Point Name) in the handling of a PCS
describe the interaction between the EPC and the GTP
outline the interaction between the EPC, GTP and IP
discuss the concept of the PCS and its relevance within the EPC
outline the functions of the default EPS Bearer
describe the differences between the default and dedicated bearer types and outline their
relationship with the Service Data Flow
describe the EPC connection hierarchy and list the set of parameter types that define them
outline the QoS concepts employed by the EPC and define the roles of the QCI and the ARP
outline the methods that are available for providing CS services to EPS attached UEs, includingGeneric Access Network functions, CS Fallback and Voice Call Continuity
outline the security functions employed by the EPC
The MME assumes many of the functions that would previously have been performed by the VLR or
SGSN and which in the evolved network are termed EMM functions.
The MME’s main responsibility is to terminate the Control Plane NAS signalling flows from individual UEs
and to manage the authentication and security functions for each attached UE. Unlike the legacy VLR,
however, the MME is also responsible for bearer establishment. It receives Service Requests from UEsand issues appropriate instructions to the S-GW that will handle each user plane connection.
The EMM functions also include responsibility tracking UEs that are in idle mode and the MME ensures
‘UE Reachability’ by receiving TAU messages, maintaining the tracking area lists and performing paging
of individual UEs when required.
To assist with service resilience, MMEs can be grouped into ‘pools’. eNBs are able to contact any MME
within the pool(s) with which they are associated when passing on UE Attach requests. The MME then
has flexibility as to the S-GW chosen to establish the user plane connection for each UE.
The MME is also in charge of roaming and handover functions to 2G/3G SGSNs.
The S-GW handles user plane connectivity between UEs and the EPC and acts as the EPC mobility
anchor for UEs roaming within part of a PLMN. This entails performing IP packet routing and buffering
functions and also managing QoS by inserting DSCP (DiffServ Code Point) data into IP packet headers.
The S-GW also provides mobility anchoring for connections that roam onto legacy 3GPP GERAN (GSM
EDGE Radio Access Network) (2G) and UTRAN (UMTS Terrestrial Radio Access Network) (3G) accessnetworks. As all EPS user traffic must pass through an S-GW it is a logical node within which, in concert
with the PDN-GW, to base the EPS Lawful Interception interface and also the charging functions.
The standard S5 and S8 interfaces that link the S-GW and PDN-GW are based on the 3GPP GTP; many
non-3GPP systems obtain similar IP mobility functionality by employing the MIPv4 (Mobile IPv4) or
PMIPv6 (Proxy Mobile IPv6) protocols developed by the IETF (Internet Engineering Task Force).
Adapted versions of the S5 and S8 interfaces are available that support the PMIP protocol for IP mobility.
In such cases, the S-GW will also act as the FA (Foreign Agent) to anchor mobile IP tunnels.
To provide some legacy perspective, taken together the MME and S-GW provide the EPC with
functionality similar to that previously provided by the SGSN, with the MME handling the signalling and
session control aspects and the S-GW dealing with the user traffic.
Early in its development, the S-GW was also known as the UPE (User Plane Entity), although this
If the functionality of the MME/S-GW can be thought of as analogous to that of the legacy SGSN, then
the PDN-GW can be thought of as similar in function to the legacy GGSN. The PDN-GW (also known in
some versions of the specifications as the P-GW) routes traffic between EPS Bearers and the SGi
interface, which leads to external data networks such as the IMS and the Internet.
As all inbound and outbound EPS traffic must pass through a PDN-GW it is the logical node in which thenetwork’s packet filtering and classification functions are based. These include the ‘deep packet
inspection’ techniques that are used to classify packets into particular SDFs before routing them over an
EPS Bearer or the SGi interface, which in turn allows the PDN-GW to act as the network’s PCEF Under
direction from the PCRF (Policy and Charging Rules Function) the PDN-GW will apply ‘per SDF’
charging, service level and rate enforcement and QoS-related traffic shaping functions that control the
‘gating’ of user traffic flows.
Each PDN-GW contains a number of logical access points (each identified by an APN which act as the
GTP tunnel endpoints and mobility anchors of the EPS Bearers that extend service out to mobile UEs. As
in the legacy GGSN, the APNs are responsible for the allocation of IP addresses to UEs during the
establishment of each EPS Bearer and for routing traffic between the Bearers and particular external
The PCRF is responsible for propagating the network’s connection policies and charging rules to the
PDN-GW via the S7/Gx interface and to traffic gateway elements within the IMS via the Rx interface. It is
the element that decides if new connections are to be allowed and, if so, whether they will be carried by
an existing EPS Bearer or whether a new one is required.
The PCRF is responsible for providing service data flow detection, gating, QoS and flow-based charginginformation to traffic handling entities within the network. This includes rules that allow the PDN-GW to
provide the correct level of service to user data flows once the type of traffic being carried has been
determined. For example, if the PDN-GW determines that the SDF to a user is carrying real-time traffic it
may ‘gate’ up to the data rate and QoS level indicated by the PCRF and the user’s subscription profile.
The PCRF’s charging rules allow the operator to apply the appropriate rating to CDRs (Call Data
Records) generated for each SDF so that, for instance, real-time connections can be differentiated from
an Internet browsing session.
In the case of EPS roaming, when users use their terminals abroad, 3GPP has developed an extended
PCRF architecture, based on the S9 interface, that defines Home Policy and Charging Rules Function
(H-PCRF) and Visited Policy and Charging Rules Function (V-PCRF) logical functions.
In order to offer effective service to UEs, the EPS needs to be able to define and keep track of the
availability and reachability of each terminal. It achieves this by maintaining two sets of ‘contexts’ for
each UE – an EMM (EPS Mobility Management) context and an ECM (EPS Connection Management)
context – each of which is handled by ‘state machines’ located in the UE and the MME.
A further state machine operates in the UE and serving eNB to track the terminal’s RRC state, which canbe either RRC-IDLE (which relates to a UE in idle mode) or RRC-CONNECTED (which relates to a UE
EMM is analogous to the MM processes undertaken in legacy networks and seeks to ensure that the
MME maintains enough location data to be able to offer service to each UE when required.
The two EMM states maintained by the MME are EMM-DEREGISTERED and EMM-REGISTERED.
A UE in the EMM-DEREGISTERED state has no valid context stored in an MME, so its current locationis unknown and paging and traffic routing cannot take place. This is generally consistent with a UE that
is either powered off or is out of EPC-connected network coverage.
The EMM-REGISTERED state relates to UEs that have performed either an attach or a TAU (Tracking
Area Update) and for which the MME maintains a valid context. In this state the UE will have been
assigned an M-TMSI and will be performing TAU functions when necessary. This means that the MME
knows the UE’s location (at least to the current TA level) and can page and route traffic for it.
A UE in the EMM-REGISTERED state will have at least one active EPS bearer (the ‘always-on’ initial or
default bearer).
In order for a UE in ECM-Idle state to perform an Explicit Detach and move from EMM-Registered toEMM-Deregistered, it must first move to ECM-Connected state to ensure that a signalling bearer is
The ECM states describe a UE’s current connectivity status with the EPC, e.g. whether an S1
connection exists between the UE and EPC or not.
There are two ECM states, ECM-IDLE and ECM-CONNECTED.
A UE in ECM-IDLE has no S1 active relationship with an MME, although UE and Bearer Contexts will bestored in the serving MME, and no NAS signalling is passing between those elements.
A UE in this state will perform network and cell selection/reselection and will send TAU messages, but
has no RRC or S1 traffic bearers established. In ECM-IDLE the location of the UE is known by the MME
only to the level of the current TA or TA List.
In the ECM-CONNECTED state a UE has established a signalling relationship with an MME, which will
know the UE’s location to the eNB level, not the current cell level. The UE’s Bearer Contexts will be
activated and RRC and S1 transport resources will have been assigned to it.
A UE will move to the ECM-CONNECTED state during functions such as Attach, TAU and Detach and
when an EPS bearer is active for traffic transfer.
A UE moves from ECM-IDLE to ECM-CONNECTED by sending a Service Request to the MME.
Each eNB is able to establish associations with multiple MME and S-GW devices following a principle
known as S1-Flex.
The main benefit of Flex capability is that it provides redundant core network services for each eNB and
the set of UEs they serve. When a new UE requests service in a cell, whether due to an initial Attach or
following a Handover, the eNB is able to select the MME to which it forwards the NAS connectivityrequest from one or more MME Groups. If an individual MME fails or becomes overloaded, the Flex
concept ensures that only a subset of each cell’s users will be affected. Similar redundancies are
provided for EPS Bearer connections through S-GWs.
Associated with the benefi ts of service redundancy are those of load balancing. If eNBs tailor the
numbers of UEs they introduce to each MME to the advertised ‘relative capacity’ of each of those devices
then the chances of individual MMEs becoming overloaded is minimized.
A further, but less obvious, benefit of the S1-Flex process is the ability it offers to allow each eNB to
connect to and offer services for more than one PLMN. In theory, a physical network of base stations can
advertise the services of, and connect traffic to, multiple PLMNs; this is a key enabler of LTE’s ability to
support shared or Multi-Operator network environments. In this model, user connection requests areforwarded by the eNB to an MME belonging to one or other of the core networks that are sharing the
The EPC was designed as an ‘all IP’ environment and as such carries all traffic, even voice, in IP
streams but interfaces have been developed that allow for backwards compatibility with and handover of
CS (Circuit Switched) traffic to legacy networks, if required.
The SGs interface is based on the GERAN/UTRAN Gs interface and carries mobility management and
handover signalling between an MME and a legacy MSC (Mobile-services Switching Centre) or MSCServer. It was created to serve the interfacing requirements of the CS Fallback service, which allows
EPC-Attached UEs to drop back to 2G/3G networks to handle CS calls.
The SGsAP (SGs Application Part) message format employed on the interface is an adaptation of the
BSSAP+ (Base Station System Application Part +) protocol employed on the legacy Gs interface, and
provides much the same set of services.
Other interfaces have been developed to support other forms of EPC-CS Core interaction; the SGs
interface, for example, carries MME-MSC/MSC-S signalling to support the SRVCC (Single Radio VCC),
which allows IMS-anchored real time sessions to be seamlessly handed over between EPS Bearers and
GERAN/UTRAN CS Bearers.
Further Reading: 3GPP TS 23.216 (SRVCC), 23.272 (CS Fallback)
The EPS Bearer ID (EBI) is assigned by the MME upon bearer establishment.
The EBI consists of 4 bits which in theory allows a maximum of 16 EPS bearers to be created for each
UE. However, the relevant specification indicates that 5 values are reserved which limits the number of
EPS Bearers per UE to 11. EBI values are always assigned by the MME which sets the EBI value for the
default bearer and sends it to the S-GW. In the same way, the MME also assigns the EBI value todedicated bearers. In UMTS networks the equivalent of an EBI is the NSAPI (Network Layer Service
Access Point Identifier) which is used to identify a PDP context. When the UE moves from LTE to UMTS,
the EBI is mapped to an NSAPI – this mapping is not complex as both NSAPI and EBI are 4 bit values.
The EPS Bearer travels between the UE and the PDN-GW; during handovers it may also extend over the
X2 interface between source and target eNBs.
When travelling over the S1 and X2 interfaces, there is a one-to-one mapping between the EPS Bearer
and the E-RAB and between the identities assigned to each of those entities.
Each UE will establish an initial or default EPS Bearer as part of the attach process. This will provide the
required ‘always on’ IP connectivity to the UE and may be to a default APN (Access Point Name), if one
is stored in the user’s subscriber profile, or to an APN selected by the network.
In networks that interconnect to an IMS, the default bearer allows the UE to perform SIP registration and
thereafter to provide a path for session initiation messaging. In these circumstances, the data rate andQoS assigned initially to the default bearer is commensurate with the expected low level of SIP-based
traffic flow, but these parameters could be modified to accommodate the requirements of application
traffic flows when a connection is established. In reality, however, a separate EPS Bearer is established
to carry LTE call media traffic, so the Default EPS Bearer would only need to be adjusted if the level of
SIP traffic exceeded the capacity provided by the original QoS settings.
If a UE has a requirement to establish an application connection whose QoS or data rate demands are
incompatible with those currently assigned to the default bearer (but which can still be routed through the
current APN), the PDN-GW or PCRF may initiate the establishment of an additional EPS Bearer to carry
the new traffic flow. Any additional bearers assigned to a UE in addition to the Default EPS Bearer are
termed Dedicated EPS Bearers and will be identified by different EPS Bearer/E-RAB and radio bearer IDs.
A UE may have more than one PDN Connectivity Service running if it has connections established
through more than one APN/PDN-GW. In that case, there will be one Default Bearer and an optional
number of Dedicated Bearers created for each PCS. The 4-bit EPS Bearer ID and the set of reserved IDs
limit the total number of bearers that can be established for one UE to eleven (numbered 5 to 16).
QoS in the EPC is currently defined by three levels: GBR, MBR and AMBR (Aggregate Maximum Bit
Rate).
GBR connections are assigned a guaranteed data rate and are therefore useful for carrying certain types
of real-time and delay-sensitive traffic. MBR connections are non-guaranteed, variable-bit-rate services
with a defined maximum data rate. If a connection’s data rate goes beyond the set maximum the networkmay decide to begin discarding the excess traffic.
GBR and MBR parameters are applied on a ‘per bearer’ basis, whereas AMBR is applied to a group of
bearers; specifically, a group of non-GBR bearers that terminate on the same UE. AMBR allows the EPS
to set a maximum aggregate bit rate for the whole group of bearers that can then be shared between
them.
The APN-AMBR parameter sets the shared bit rate available to a group of non-GBR bearers that
terminate on the same APN and can therefore be seen to be applied on a ‘per PCS’ basis; the UE-AMBR
parameter aggregates all non-GBR bearers associated with one UE.
Dedicated bearers can be established as GBR or non-GBR (i.e. MBR) as required. Default bearers, dueto the probable need to adjust their bandwidth after the initial Attach has taken place, must be non-GBR.
VCC (Voice Call Continuity) is designed to make use of the combined resources of the IMS and legacy
CS core network by allowing IMS-anchored real-time or CS calls to be handed over from the E-UTRAN
and the GERAN/UTRAN.
The specific variant of this concept outlined in the diagram is SRVCC (Single Radio VCC), which
supports UEs that only contain one radio and can therefore only connect to one air interface method at atime; in this scenario, the UE is capable of connecting to E-UTRAN, UTRAN or GERAN cells but only
one at a time.
Call- and handover-related signalling is passed between the MME and MSC-MSC Server via the Sv
interface. Handover or hand back of calls from UTRAN/GERAN to E-UTRAN is not supported; once a
call drops down to 2G/3G it stays there.
Any active PS sessions will be split from the CS sessions and handed over to a 2G/3G SGSN at the
same time as the CS sessions are transferred.
The SRVCC specification also provides options for handing over IMS-anchored real-time sessions from
UTRAN (HSPA) and 3GPP2 1xRTT CDMA2000 access networks to GERAN/UTRAN resources.
EPS employs the same AKA (Authentication and Key Agreement) mechanism as is used by 3G UMTS
networks.
The EPS AKA mechanism aims to ensure that the network can authenticate users and vice versa, and
that once authenticated, users and network can agree on a set of encryption mechanisms to employ to
protect user and control traffic. EPS AKA operates between the UE and the MME and is facilitated bysubscription data stored in the USIM (Universal Subscriber Identity Module) and the HSS.
As in 3G UMTS, when a user is required to authenticate, the HSS will generate a quintet of AVs
(Authentication Vectors): a random 128-bit number (RAND), an XRES (Expected Response), a CK
(Cipher Key), an IK (Integrity Key) and an AUTN (Authentication Token) – which are passed to the
serving MME.
RAND is used as a challenge and is transmitted to the UE. The USIM processes RAND through its copy
of the ‘shared secret’ K authentication key and generates a response, which is transmitted back to the
MME. If the USIM response matches XRES then the USIM is deemed to be genuine and the UE is
allowed to access network services.
The CK is passed to the serving eNB to allow user plane encryption to and from the UE to take place,
while the IK is employed between the UE and the MME to protect the integrity of signalling messages.
Finally, the AUTN is passed to the UE to allow it to authenticate the network.
As with legacy 3GPP systems, the EPS uses the IMSI to absolutely and uniquely identify each user. The
user confidentiality mechanism provides subscriber anonymity by ensuring that the IMSI is transmitted
across the network as little as possible.
A UE accessing a network for the first time or after a long period of inactivity has no option but to transmit
its user’s IMSI to the network to allow identification and authentication to take place. Once the user hasbeen authenticated, however, the MME generates an ‘alias’ that may then be used in place of the IMSI to
identify the subscriber.
Generically in 3GPP networks this alias is known as a TMSI. The specific variety employed in the EPS is
the M-TMSI. The correspondence between M-TMSI and a user’s true IMSI is known only to the MME and
user’s UE. An M-TMSI will be unique within the MME that issued it. When combined with an MMEC to
make an S-TMSI it becomes unique within an MME pool. When the M-TMSI is combined with a GUMMEI
to form a GUTI it becomes unique within all EPS networks.
The MME may elect to request UEs to reauthenticate periodically and will issue a new M-TMSI at this
time. A UE may be issued a new M-TMSI when it moves to the control of a new MME.
The EPS user confidentiality mechanism is essentially the same as that employed in the GERAN and
UTRAN, although the identities of the relevant network elements have changed.
Idle mode represents a state of operation for the UE where it has successfully performed the following:
PLMN selection, cell selection and location registration (by tracking area).
Once in idle mode, the UE will continue to reassess the suitability of its serving cell and, in some
circumstances, its serving network. In order to do this it will implement cell and PLMN reselection
procedures. A UE in idle mode will be monitoring its current serving cell in terms of radio performanceand signalling information. The radio performance measurements are done on the basis of a quality
measure. This is an assessment of radio signal strength and interference level, and it can be made for
both the serving cell and its neighbours. The aim will be to ensure that the UE is always served by the
cell most likely to give the most reliable service should information transfer of any kind be required.
The UE will also be monitoring two key types of signalling from the serving cell system information
messages and paging or notification messages. System information messages convey all the cell and
system parameters. The UE will record changes in these parameters that may affect the service level
provided by the cell, or access rights to the cell. Changes in these parameters could provoke a cell
reselection, or a PLMN reselection. Paging or notification messages will result in connection
establishment.
All of these procedures are performed through communication between the AS and the NAS. In general,
instructions are sent from the NAS to the AS; the AS then performs the requested procedure and returns
a result to the NAS.
If CSG (Closed Subscriber Group) is supported then these procedures are modified such that a cell’s
broadcast CSG ID forms another level of differentiation between cells. CSG is intended for use with
Cell reselection in LTE both reuses many principles that were are well established in legacy technologies
and introduces new strategies. A key addition for LTE is the use of RAT/frequency prioritization. Each
frequency layer that the UE may be required to measure, either E-UTRA or any other RAT, is assigned a
priority. The cell-specific priority information is conveyed to UEs via system information messages.
Additionally, UE-specific values can be supplied in dedicated signalling, in which case they take priority
over the system information values. Any indicated frequency layers that do not have a priority will not beconsidered by the UE for reselection.
In general, the measurement rules are used to reduce unnecessary neighbour cell measurements. The
UE always measures cells on a higher priority E-UTRA inter-frequency or I-RAT frequency. The UE will
only measure E-UTRA intra-frequency cells if the Srexlev value for the current selected cell falls below
an indicated threshold (Sintersearch). Similarly, the UE only measures E-UTRA inter-frequency or I-RAT
frequency cells on equal or lower priority layers if the Srexlev value for the current selected cell falls
below an indicated threshold (Snonintrasearch).
Measurements are then evaluated for potential reselection. Again, the frequency/RAT priority level is
used along with system-defined threshold for this assessment. A UE will always reselect a cell on a
higher priority frequency if its value of Srxlex exceeds Threshx,high for longer than TreselectionRAT. It willonly select a cell on a lower priority frequency when the Srxlev of the serving cell falls below
Threshserving,low and Srxlev of the neighbour is above Threshx,low for TreselectionRAT and there is no other
alternative. For neighbour cells on intra-frequencies or on equal priority E-UTRA inter-frequencies, the
UE uses a ranking criterion ‘Rs’ for the serving cell and ‘Rn’ for the neighbour cell. Ranking is based on a
comparison of the respective Srxlev values with a hysteresis added to the serving cell value and an offset
added to the neighbour cell value. The UE will select the highest ranked cell if the condition is maintained
for TresectionRAT.
In addition to all of this, the UE will apply scaling to Treselection, hysteresis values and offset values
dependent on an assessment of its mobility state, which may be high, medium or low. This is based on
When the UE becomes RRC connected, the measurement and reporting process as well as mobility
decisions becomes the responsibility of the eNB. The required measurement and reporting settings are
signalled to the UE in the RRCConnectionReconfiguration message.
The measurement object defines what the UE is to measure. This is defined as a frequency and
measurement bandwidth; optionally it may also contain a list of cells. If it does contain a list of cells thenthey will be indicated as either white list or black list. The UE will measure any cells it detects but will not
report black list cells. Frequency- or cell-specific offsets will also be included in this field.
The reporting configuration sets what quantities the UE is to measure, what quantities the UE is to report
and under what circumstances a measurement report is to be set. Reporting may be set as either trigger-
based, periodic or triggered periodic. This field also defines the other contents of the measurement report
message.
Measurement identities provides a reference number such that some part of this identified measurement
can be modified or removed in future.
The Quantity configuration sets the filtering to be used on the measurements that are taken.
The gap configuration defines periods when the UE can take measurements of neighbour cells.
In order to maintain orthogonality between uplink transmissions from multiples UEs in a cell, timing
adjustment must be applied to compensate for variations in propagation delay.
Initial timing advance is calculated at the eNB from a UE’s preamble transmission on the PRACH. The
timing advance correction is given as an 11-bit value although the range is limited to 0–1282 timing
advance steps. Granularity is in steps of 16Ts (0.52 µs) so timing advance can be varied between 0 and0.67 ms. One timing advance step corresponds to a distance change of c.78 m and is significantly
smaller than the normal CP. The maximum timing advance value corresponds to a range of c.100 km.
The maximum specified speed for a UE relative to an eNB is 500 km/h (139 m/s), which would require
slightly more than one timing advance change every two seconds. Consideration also needs to be given
to the possibility of more extreme changes in the multipath characteristics of a channel, for example the
sudden appearance or disappearance of a strong reflected path from a distant object or delay through a
repeater. However, these are extreme examples and, in any case, timing advance update commands
can indicate up to +/– 16 µs in a single step. Thus the rate at which timing advance commands need to
be sent in practice is typically much less than one every two seconds.
Timing update commands are transmitted to UEs as MAC control messages and as such are included inMAC PDUs carrying data for the UE on the PDSCH. The command itself is a six-bit value giving a
number range from 0–63. Values less than 31 will reduce timing advance and values greater than 31 will
In its first release, LTE is specified with several options for SU-MIMO implementation and a more limited
option for MU-MIMO operation. The specification include descriptions of operation up to rank 4 (4x4
MIMO).
The simplest option is not MIMO, as such, but uses the multi antenna array at an eNB to provide transmit
diversity. The standards allow configuration with up to four antennas at the base station. It is likely thatcross-polar antennas would be used as part of the antenna array, so a two-antenna array could be
implemented using a single cross-polar panel, with a four-antenna array requiring two cross-polar panels.
Transmit diversity involves the transmission of a single data stream to a single UE, but makes use of the
spatial diversity offered by the antenna array. This can increase channel throughput or increase cell
range.
There are also two beamforming options available. These are based on the use of a single layer with rank
one pre-coding but make use of a multi antenna array for beamforming to a single UE. The two options for
this are a closed loop mode, which involves feedback of PMI (Pre-coding Matrix Indicators) from the UE,
and an open loop mode, which involves the transmission of UE-specific reference signals and the eNB
basing the pre-coding for beamforming on uplink measurements.
Full SU-MIMO configurations are available in LTE in the downlink direction with ranks up to four. However,
a maximum of two data streams is used, even when four antenna ports are available. In SU-MIMO the UE
can be configure to provide PMI feedback as well as RI (Rank Indicators), which indicates the rank that
the UE calculates will give the best performance.
In the first release of the LTE specification there is only a limited implementation of MU-MIMO specified. It
is applicable in the uplink direction and allows two UEs to use the same time frequency resource within
one cell.
Further Reading: 3GPP TS 36.211:6.3.3, 6.3.4, 36.213:7.1
The UE’s objective when performing an attach is to register the subscriber’s identity and location with the
network to enable services to be accessed. During the attach procedure the UE will be assigned a default
EPS bearer to enable always-on connectivity with a PDN. The UE may be provided with details of a local
P-CSCF to enable it to register with the IMS.
A simplified view of the attach process – assuming that it is an initial attach with stored details from arecent previous context for a UE using its H-PLMN (Home PLMN) and accessing via the Home E-UTRAN
– is shown, and the stages of the process are described below.
Once a suitable cell has been selected the UE employs the Random Access procedure to request an
RRC connection with the chosen eNB. With that in place an Attach Request message (1) can be
transmitted. If the UE has previously been registered with the PLMN, it may include a previously assigned
GUTI in the message, otherwise the Attach Request message contains the subscriber’s IMSI and some
other parameters.
On receipt of the Attach Request the eNB either derives the identity of the previously used MME from the
supplied GUTI or selects an MME from the pool available and forwards the message (2).
The MME contacts the HSS indicated by the subscriber’s IMSI and in response receives the relevant
elements of the ‘quintuplet’ that allows the EPS-AKA process to take place (3).
Optionally, at this point the MME may be required to check the identity and status of the UE via the EIR
(4) using the ME Identity Check process. Ciphering may then be invoked over the air interface.
Once the AKA procedures have successfully concluded the MME transmits an Update Location message
to the HSS and receives the Insert Subscriber Data message in response containing the user’s service
profile (5). An Insert Subscriber Data Ack from the MME is followed by an Update Location Ack from the
ISR is designed, as the name suggests, to reduce the amount of UE-network and MME-SGSN signalling
required to manage idle mode terminals. ISR is a feature of the S3 and S16 interfaces and is not
available to legacy SGSNs that do not support them.
When an Idle UE activates (or is instructed to activate) ISR, copies of UE Context and Bearer Contexts
are stored in both an MME (for E-UTRAN access) and SGSN (for GERAN/UTRAN access). The UE isable to reselect freely between registered RATs without transmitting location updates, unless a change in
RAI or TAI is detected. Any location updates that are sent need only be transmitted via the RAT currently
in use; the receiving core network element will forward the update to its peer over the S3 interface.
The MME and SGSN both store copies of the UE’s bearer contexts and will both page for the UE. When
the UE needs to move to connected mode, whether in response to a page or to a user-initiated event, it
can do so by sending a Service Request via whichever RAT it is currently camped on. The receiving
device will then instruct the S-GW to re-establish the parked bearers.
A UE with ISR activated maintains details of the RAT and therefore the RAT-specific temporary identifier
that is in use using the TIN (Temporary Identity used in Next update) parameter.
The TIN can be set to P-TMSI (for GERAN/UTRAN access), GUTI (for E-UTRAN access) or RAT-related
TMSI. This last option means that the UE will use the P-TMSI or GUTI depending upon which RAT is
currently in use.
A UE will deactivate ISR if it loses contact with one of the registered access networks. For example, a UE
might be within the coverage of both an E-UTRAN and a GERAN cell when ISR is activated but may
roam out of coverage of the E-UTRAN cell; in such circumstances it would revert to being attached to just
User connectivity in a combined EPS/IMS network requires two levels of connection to be established:
firstly, the radio and EPS bearers that will carry traffic through the E-UTRAN and EPC, and secondly the
IMS SIP and media connections that will carry call-related signalling and end-to-end user traffic.
A UE’s default bearer may be an operator’s first choice for carrying application traffic, but if the QoS
demanded by a new service data flow is incompatible with that of the default bearer, then the PDN-GW/PCRF may decide that an additional dedicated bearer is established.
When a UE enters idle mode the physical S1 and radio resources assigned to the default EPS bearer will
be released and the bearer context details will be stored. Any existing dedicated bearers may be
released or stored also.
When the UE moves from ECM-IDLE to ECM-CONNECTED the stored bearer contexts will be
A UE will trigger a Service Request to reactivate its parked bearer contexts in response to a command
from an application client, the terminal management software or the user interface. A response to a
network initiated paging message will also trigger a Service Request.
The process begins with the transmission of a NAS: Service Request either following the random access
procedure or carried in scheduled uplink capacity. The NAS: Service Request contains the UE’s currentS-TMSI and the service type (data or paging response). The request is initially forwarded to the eNB
encapsulated in an RRC message (1).
Direct Transfer NAS messages were transparent to the UMTS Node B and were only accessible to the
RNC. In the E-UTRAN, NAS messages are switched from the RRC bearer used on the air interface to an
S1AP bearer for forwarding to the MME (2) and in some cases are interpreted by the eNB.
Depending upon configuration, the MME may initiate a reauthentication of the UE/USIM before
processing the Service Request (3).
The MME sends the eNB an S1AP: Initial Context Setup Request, which issues the commands that re-
establish physical resources for the stored bearer contexts on the S1 interface between the UE and theS-GW (4). The eNB allocates radio resources (5) on the air interface and informs the UE. Uplink traffic is
then able to flow (6). The eNB confirms these actions with an S1AP: Initial Context Setup Complete
message (7).
The MME instructs the S-GW to establish its end of the S1-U tunnels using the Update Bearer Request
message (8). If the PDN-GW has requested updates regarding the UE’s location, the S-GW will pass this
on in an Update Bearer Request (9). After the PDN-GW and S-GW return Update Bearer Responses,
data can begin to flow on the downlink (9, 10 and 11).
If a UE determines that there is a requirement to establish a traffic flow aggregate (which may contain
one or more SDFs) to a new AF (Application Function) destination – in response to a user interface
request, for example – it will transmit a Request Bearer Resource Modification to the MME. If the UE had
been in Idle Mode when it made this determination it will first send a Service Request to reactivate the
existing bearers.
The MME forwards the request to the S-GW currently dealing with the UE’s EPS Bearer(s), which in turn
forwards it to the appropriate PDN-GW. If dynamic PCC is in use, the PDN-GW interacts with the PCRF
to determine how best to deal with the request: if static PCC is in use then the PDN-GW makes the
determination itself.
The Modification request includes the required QoS, the EPS Bearer ID and a TAD (Traffic Aggregate
Descriptor), which describes the modification function to be performed (add, modify or delete) and the
SDF 5-tuple details that enable the PCRF to build a packet filter for the flow. The PCC function will
evaluate the request and either accept or reject it. Accepted requests result in new or updated packet
filters.
In the case of a new traffic flow that is to be added to an existing bearer, the PCC function will add anadditional packet filter to the TFT (Traffic Flow Template) related to the bearer over which the flow will
travel. If the addition of the new flow alters the bearers QoS requirements the adjustment will be
communicated to other elements using the Update Bearer Request process.
In addition to UE-initiated Bearer Modification the EPC also supports PDN-GW-initiated Bearer
Modification; HSS-initiated Bearer QoS Modification and MME and PDN-GW initiated Bearer
IMS connection establishment is the responsibility of SIP. The EPS default bearer is established to a
home network PDN-GW and maintained mainly to provide a path for SIP messaging between a UE and
its serving I-CSCF.
Consider an example SIP flow between a roaming UE and its home S-CSCF during which a media
session to a distant IMS-connected UE is initiated. Not all network elements involved in the processhave been shown.
In response to an instruction received via the user interface, the originating UE initiates the session by
transmitting a Service Request to reactivate its bearers followed by a SIP Invite message to the current
I-CSCF (1). The Invite message contains an SDP payload, which describes the type of connection the
originating UE wishes to establish with the destination UE.
The I-CSCF passes the message on to the assigned S-CSCF for authorization (2). The S-CSCF
discovers the called party’s home network and passes the Invite to an I-CSCF in that network for
forwarding to the S-CSCF and the destination UE (3).
Once discovered, the destination UE inspects the SDP payload and determines if it can support thetype of service and QoS parameters specified. A Session Progress message is returned to the
originating UE containing the IP address of the distant terminal and a response to the SDP parameters
(4).
Each CSCF in the return path is able to approve or edit the SDP response so that the eventual media
session’s parameters match the capabilities of the busiest link in the chain.
The originating UE returns a PRACK (Provisional Acknowledgement), which confirms the parameters of
the media session (5). This triggers the reservation of resources for the distant UE, which it confirms
In this scenario a UE with active EPS bearers initiates the inter-eNB cell handover procedure when both
source and target eNBs are associated with the serving MME and S-GW and an X2 path can be
established between them.
At the start of the process the UE is connected to the E-UTRAN and is taking neighbour cell
measurements (1). Traffic may or may not be flowing over the connected EPS Bearers. A neighbour cellachieves the criteria necessary for the UE to initiate handover, which is effected by the source and target
eNBs without core network involvement beyond the establishment of S1 resources towards the target
eNB (2).
Once the handover is complete the source eNB forwards any further downlink traffic received for the UE
to the target eNB either directly via the X2 interface (3); this is termed Direct Forwarding. Uplink traffic
travels from the target eNB to the S-GW via the newly established tunnel. The target eNB sends a Path
Switch Request to the MME informing it of the handover (4).
The MME determines that the existing S-GW is still capable of serving the UE and instructs it to switch
the Downlink user plane over to the tunnel created towards the new cell using a User Plane Update
Request message (5). If the PDN-GW has indicated that needs to be kept informed of the UE’s location(for variable charging purposes), the S-GW informs it using an Update Bearer Request (6).
The S-GW realigns the tunnel carrying the EPS bearer(s) to point to the target eNB and sends
confirmation to the MME that the path has been switched successfully (7); both uplink and downlink
traffic now travels over the new S1 tunnels (8). The S-GW sends ‘end marker’ packets to the source eNB
to confirm that the path has been switched, which are forwarded to the target eNB to indicate that X2
forwarding will cease (9).
The MME confirms the procedure to the new eNB with the Path Switch Request Ack message (10) and it
in turn confirms the handover to the old eNB using the Release Resource message (11).
The UE remains connected to the source E-UTRAN cell during the preparation phase, but once
alternative resources are in place the source MME issues a Handover Command to the source eNB (9),
which in turn sends a HO from E-UTRAN Command to the UE (10a). This message encapsulates a
‘transparent container’ that travels between the target RNC and the UE, which contains details of the
resources that have been reserved for the UE in the target cell.
The UE releases its E-UTRAN resources and performs the access activities required to establish
connectivity in the target UTRAN cell and sends the Handover to UTRAN Complete message (10b) in the
new cell to confirm the connection.
As the tunnel from the PDN-GW has not yet been realigned, Downlink packet traffic destined for the UE
is still being sent to the source eNB and must be forwarded to the target RNC. Direct forwarding between
the source eNB and target RNC (11a) uses an unnamed forwarding interface. Indirect Forwarding travels
between source eNB, source S-GW, target SGSN and target UTRAN (11b). Once the handover is
complete, the UE can send traffic on the uplink via the PDP Context that has been created towards the
SGSN, from where it will be forwarded to the S-GW and on to the PDN-GW (11c).
Once the UE has successfully connected to the UTRAN cell the target RNC sends a RelocationComplete message (12) to the target SGSN, which in turn informs the source MME using the Forward
This KPI describes the ratio of the number of successfully performed incoming handover procedures to
the number of attempted incoming handover procedures to evaluate inter RAT incoming handover
performance from a GSM network.
Inter-RAT Incoming Handover Success Rate (UMTS EPS)This KPI describes the ratio of the number of successfully performed incoming handover procedures to
the number of attempted incoming handover procedures to evaluate inter RAT incoming handover
The ongoing design and implementation of cellular systems is pushed forward by equipment vendors,
network operators and, in most cases, international specifications bodies such as 3GPP. A degree of
central guidance to this process is provided by the ITU, a branch of the United Nations responsible for
attempting to harmonize the standards and techniques employed in worldwide telecoms systems.
The ITU set out its vision for 3G cellular systems at the end of the 1990s by devising a framework knownas IMT-2000, although the 3G market ended up consisting of several competing technological platforms
all of them ultimately adhered to the guidelines espoused by the ITU.
The ITU created a similar framework of expectations for emerging 4G technologies in 2007 with the initial
publication of the IMT-Advanced guidance.
Instead of specifying a particular technology or mandating the functionality of a 4G system, the IMT-
Advanced guidelines proposed the basic set of capabilities that a system should be able to support for it
to be regarded as being a true ‘4G’ system.
In addition to the aspirations outlined in the diagram, the ITU (in ITU-R report M.2134) also provided
some more definite technical requirements for 4G systems, including: that an IMT-Advanced systemshould be able to accommodate radio channels of at least 40MHz in bandwidth; that their peak spectral
efficiency should meet or exceed 15bits/Hz in the downlink and 6.75bits/Hz in the uplink; that it should
support latencies of less than 100ms for control-plane connections and less than 10ms for the user-
plane.
It should be noted that the original (Release 8) version of LTE did not meet the full set of IMT-Advanced
requirements, mainly due to its inability to support 1 Gbit/s data connections. To date only Release 10
LTE-Advanced and the updated version of Mobile WiMax (IEEE 802.16m) have been accepted as fully
compliant with IMT-Advanced.
Further Reading: www.IMT-2000.org (ITU website), ITU-Report M.2134
Basic LTE-Advanced functionality was first described in 3GPP Release 10 specifications, with the more
detailed and enhanced functionality held over to Release 11.
Release 10 functions include: Carrier Aggregation, which allows a UE to be scheduled with data-carrying
capacity on up to five parallel cells simultaneously; enhanced MIMO techniques that allow a UE to make
use of up to 8 MIMO layers on a downlink and up to 4 layers on an uplink; additional types of MIMO for uplink connections; the extension of LTE into a variety of new radio bands; enhancements to the SON
concept; and a range of data-handling improvements such as LIPA (Local IP Access) and SIPTO
(Selective IP Traffic Offload).
Release 11 introduced only one significant new feature, known as CoMP, and mainly concentrated on
adding more depth and complexity to existing R10 services.
Detailed information about the work and study items scheduled for each release are contained in the
relevant 3GPP Release Description documents, which provide a regularly-updated view of the work
being undertaken, or that has been completed, for each release.
Further Reading: http://www.3gpp.org/ftp/Information/WORK_PLAN/Description_Releases/
CA (Carrier Aggregation) is the most prominent feature of Release 10 LTE-Advanced. It offers an inverse
multiplexing facility that allows a UE to substantially increase the overall data rate it can achieve by
allowing an eNB to schedule capacity for it on multiple cells (or 'carriers') simultaneously.
Each carrier (either downlink or uplink) assigned for use by a UE is known as a CC (Component Carrier)
and the set of CCs allocated to a UE at any one time forms a Carrier Aggregate. R10 CA permits up tofive CCs to be bound into a Carrier Aggregate, potentially providing a suitably-equipped UE with up to
100MHz of bandwidth and an aggregate downlink data rate of over 3Gbit/s.
The lowest level of carrier aggregation allows a UE to connect via just one cell. The radio connectivity of
this cell is described as the PCC (Primary Carrier Component) and the cellular service it offers is known
as the PCell (Primary Serving Cell). The PCell carries NAS and RRC services for a UE and is also the
carrier measured by the UE to support functions like quality feedback and handover measurements.
A Release 8/9 UE (or an R10 UE that didn’t require CA services) would just connect via the PCell and
would not be assigned any additional carriers. An R10, CA-capable UE that did require a CA service
would be scheduled with capacity on between one and four SCells (Secondary Serving Cells), each of
which would be carried by an SCC (Secondary Component Carrier).
The PCC and any SCCs aggregated to provide a CA service for a UE must all be under the control of the
same eNB, but the terms ‘primary’ and ‘secondary’ used in relation to CA carriers are determined from
the point of view of each UE – different UEs in the same area may have selected different cells to be
their PCell and may therefore regard an assigned cell as an SCC which may be employed by another UE
as a PCC.
A PCell is always used in a bidirectional manner, as befits the cell that carries NAS and RRC traffic, but
an SCell may be used in either a bidirectional or unidirectional manner depending upon local
configuration and current requirements. If an SCell is used unidirectionally then it is only able to operate
in 'downlink-only' mode, there is no provision for cells to operate in 'uplink-only' mode.
Carrier Aggregation operates by allowing an eNB to schedule uplink and downlink capacity for a UE
across a set of component carriers that have been combined to form an Aggregated Channel.
The total bandwidth occupied by the Aggregated Channel is termed the Aggregated Channel Bandwidth
and includes all used and unused subcarriers; unused subcarriers are generally those set aside to act as
guard bands at the upper and lower end of each CC. The net usable bandwidth of the AggregatedChannel (which is total bandwidth less all guard subcarriers) is termed the Aggregated Transmission
Bandwidth.
Due to the requirements imposed by OFDMA, each CC will still need to have its centre-frequency left
unused to act as a DC Subcarrier. To ensure that orthogonality is maintained between subcarriers across
an entire aggregated channel, the DC Subcarriers of all CCs within the aggregate must be spaced at
multiples of 300kHz from each other. The number of guard subcarriers employed at the top and bottom
ends of each CC may be adjusted to accommodate this spacing requirement if necessary.
In a contiguous aggregated channel (i.e., one in which all CCs are adjacent to each other) the DC
subcarrier of the central CC will be used as the baseband transmission DC Subcarrier for the entire
aggregate. For this reason, it is recommended that aggregate channels are created in a ‘balanced’manner, as shown in the diagram. The total 25MHz aggregated channel bandwidth in the diagram has
been created by aggregating 2x10Mhz and 1x5MHz CCs, but to balance this aggregate the 5MHz CC
has been placed in the centre and its DC Subcarrier will be used for the entire aggregated channel.
Un-balanced configurations are also possible – a 25MHz channel aggregate could be created by
combining CCs in a 10+10+5 pattern, for example – but in these cases the aggregate DC Subcarrier
must be created by using whichever ‘transmission’ subcarrier happens to be at the centre frequency of
the aggregate.
In non-contiguous aggregates (like those employed in inter-band configurations), each contiguous
section of the aggregate will require its own aggregate DC subcarrier.
Further Reading: 3GPP TS36.807:5.6A & 5.7, 36.104:5.7.1A
The cost of LTE UEs is partly determined by the sophistication of the set of functions that each can
perform. Only a limited subset of UEs will be designed to take advantage of the maximum possible
capabilities of an LTE network. It is therefore necessary to allow each UE to signal its capabilities to the
network to allow the services offered to be tailored to its abilities.
UE capabilities are signalled in RRC message IEs and are classified into UE Categories.
Release 8/9 UEs have capabilities – those that relate to ‘standard’ LTE access – that place them in UE
Categories 1-5, these are signalled in the UE-EUTRA_Capability IE.
Release 10 UEs may have capabilities that relate to LTE-Advanced access which place them in UE
Categories 6-8, these are signalled in the UE-EUTRA-Capability-v10 IE.
A Release 8/9 UE with UE Category 5 can, in theory, achieve data rates of up to 300MBit/s in the
downlink and 150MBit/s in the uplink using 64QAM modulation in both directions and up to 4 layers of
downlink MIMO.
A Release 10 UE with a UE Category of 8 can, in theory, achieve data rates of 3000MBit/s downlink and1500MBit/s uplink using 64QAM in both directions, up to 8 MIMO layers in the downlink and up to 4 in the
uplink. The use of CA with 4 SCells to achieve these data rates is implicitly indicated as they wouldn’t be
possible otherwise.
Both UE-EUTRA-Capabilities signal a large number of other parameters in addition to UE Category and
other IEs are also employed to carry UE configuration and capability information.
Further Reading: 3GPP TS36.306:4.1 (UE-Category IE), 36.331:6.3.6 (UE-EUTRA-Capability)
The level of Carrier Aggregation that is supported by a UE is determined by its Bandwidth Class.
Each Bandwidth Class describes the level of CA capabilities supported by a UE, it includes descriptions
of the maximum number of Resource Blocks that a UE can use simultaneously and the maximum
number of Component Carriers (with each CC classed as an uplink/downlink pair whether the uplink is
actually used or not) the UE can access simultaneously. The largest channel bandwidth available in R10is 20MHz, so the maximum number of CCs parameter implicitly identifies the maximum operational
bandwidth a UE could be assigned.
Bandwidth Class A devices do not support CA, so this classification would be used to describe all
Release 8/9 UEs and any Release 10 devices that do not require CA capabilities. Class A devices are
able to access up to 100 Resource Blocks on a 20MHz channel in a PCell only and will not be able to
access any SCells. Examples of Bandwidth Class A devices might include voice-only LTE terminals and
devices designed to support low-bandwidth data applications such as smart utilities meters.
Bandwidth Classes B and upwards support increasing levels of aggregation capability. Class B devices
are also limited to using up to 100 RBs but may be assigned one SCell in addition to the PCell. The 100
RB limit means that although a Class B device is able to access two CCs it would not be able to accesstwo 20MHz CCs, as that would entail processing 200 RBs.
Bandwidth Class C devices are able to handle up to 200 RBs and two CCs and would therefore be able
to be assigned to, for example, a 20MHz PCell and a 20MHz SCell, although other channel
configurations are also possible.
As of v10.2.1 of TS36.101, the 3GPP specification that describes Bandwidth Classes, classes D-F had
parameters assigned to them but were described as being FFS (For Further Study), so the parameters
shown in the diagram are only provisional and may be subject to change by later versions of the
document. However, the maximum capability for a Class F device is likely to be up to five CCs covering a
A PCell, or Primary Serving Cell, is essentially the same as a standard R8/9 LTE cell.
A UE, whether it supports R8/9 or R10, will undertake the same Idle Mode and cell selection procedures
when active and will camp on a serving cell from the set locally-available based on signal quality and
strength indicators. Once a cell has been selected and the UE has camped on it, it becomes that UE’s
PCell. The current PCell is always assigned Serving Cell Index 0 by the UE.
A UE will only monitor the BCCH of a PCell, any system information required to access an SCell will be
provided to the UE using dedicated RRC signalling during SCell Addition or Modification.
An Idle Mode UE will monitor the PCell’s paging channel and will use measurements taken of the PCC
as the basis for idle mode reselection actions. If necessary, the UE will use PCell’s RACH to transmit
Service Requests to initiate a move to Connected Mode (moving from ECM-Idle to ECM-Connected).
A UE in Connected Mode will initially use the resources of the PCell; all RRC and NAS transactions take
place via the PCell and any ciphering applied to established traffic and control channels will be based on
the ciphering calculated for the PCell. The UE will send Uplink Capacity requests via the PCell PUCCH
(Physical Uplink Control Channel) and will monitor the PCell PDCCH (Physical Downlink ControlChannel) for downlink capacity notifications and uplink scheduling grants.
Scheduling for all forms of LTE is managed via the PDCCH.
For R8/9 UEs and for R10 UEs that either don't support CA or that do not currently have CA activated,
only the PCell's PDCCH will carry relevant scheduling information.
For R10 UE’s with CA activated and SCells assigned, there are two scheduling information options,depending upon whether cross-carrier scheduling has been enabled.
Cross-carrier scheduling allows scheduling notifications for one or more SCells to be carried by the
PDCCH of another active cell, typically the PCell. In such cases the capacity notification will indicate
details of the capacity granted plus the index of the intended CC. The typical benefit of cross-carrier
scheduling is that it allows a UE to monitor just one PDCCH to get access to all of its capacity
notifications. Cross-carrier scheduling is not an option for the PCell, whose capacity notifications are
always carried on its own PDCCH.
If cross-carrier scheduling is not enabled for an SCell then capacity notifications for that cell will be
carried on its own PDCCH.
Cross-carrier scheduling can be enabled or disabled on an SCell by SCell basis, based on an option in
the RRC SCell Addition and Modification messages, meaning that a UE could in theory have it enabled
for some cells in its CA set but not for others.
Whatever its complexities, LTE-Advanced Carrier Aggregation can ultimately be seen as an exercise in
multi-carrier scheduling and its success or otherwise in an individual vendor's equipment will be
determined by the efficiency of the scheduling algorithm they have developed.
R10 LTE-Advanced shares the same air interface protocol stack as standard R8/9 LTE, namely PDCP
over RLC over MAC over the OFDMA-based PHY (Physical Layer).
Carrier Aggregation is mainly accommodated by making changes to the MAC layer. Each UE-specific
MAC instance in an eNB and the corresponding MAC layer in each UE perform a set of functions that are
common for all active serving cells (such as MAC PDU creation and scheduling), but there is also a set of cell-specific functions performed separately for the PCell and each Scell. These functions mainly relate to
the establishment of separate HARQ sessions for each UL or DL SCH (Shared Channel) that is active.
At the Physical layer, the eNB and UE will establish separate instances of UL-SCH and DL-SCH for each
CC and for each MIMO layer on each CC that is active for the UE connection. Traffic travelling over each
of these separate physical layer connections will be scheduled capacity by the MAC layer as required to
meet the QoS agreed for each E-RAB.
The upper reaches of LTE-Advanced capability, as evidenced by the parameters associated with UE
Category 8, can in theory allow a UE to achieve 3000MBit/s on the downlink and up to 1500MBit/s on the
uplink. To enable this to take place, the UE is likely to be operating with a full set of serving cells (one
PCell and four SCells), to be using the highest available modulation scheme (64QAM on both the uplinkand the downlink) and to be using the full set of available MIMO streams. This would equate to 8 streams
per DL CC and 4 streams per UL CC – or a combined 40 separate logical downlink physical layer
connections and 20 in the uplink.
The possibility of a UE (or a network) achieving commercial support for such extreme data rates and
operating parameters in the near term is exceedingly remote and far more modest data rates of a few
hundred megabits per second is a much more achievable target for most LTE-Advanced operators.
3GPP Release 10 and 11 specifications also introduce a number of techniques designed to enhance theways in which mobile data is handled, especially when users have access to femtocells or wifi hotspots.
LIPA is designed to allow a UE to make use of the resources of a home or office LAN (to which the user has access) via LTE; the intention is that a UE could access a network printer, or send content to an
Internet-enabled TV, or stream music from a home content server, all via a LAN-connected femtocell.Managed Remote Access aims to provide the same kind of home/office LAN access service but workswhile the user is roaming away from their home femtocell; remote access to local resources is providedfrom anywhere on the LTE network via the secure connection created to serve the user’s homefemtocell.
SIPTO is designed to detect and make use of geographically local EPS Gateways (S-GW and PDN-GW)for certain PDN connections. A roaming UE may, for example, have one PDN connection (connected toan application server that only exists in the home network) directed back to an APN in a home networkPDN-GW, whilst a PDN connection to the general Internet could be directed to a local, visited network APN. SIPTO works in both roaming and non-roaming cases and an APN’s applicability for SIPTO issignalled in the user’s HSS profile.
IP Flow Mobility aims to enable better interworking between LTE and WiFi, in this case by providing themeans to seamlessly hand active connections over between LTE and Wi-Fi access devices, but only if the Wi-Fi access is controlled by the LTE network operator. A UE that, for example, has an active voicecall and is also in the process of downloading a large file from an Internet-based server, could use IPFlow Mobility to divert the data download to an active (operator-controlled) Wi-Fi connection, thusoffloading traffic from the LTE network and preventing the user from consuming unnecessary amounts of their data bundle.
Multi-Access PDN Connectivity allows an operator to make efficient use of their portfolio of cellular access systems by allowing UEs to setup simultaneous connections via more than one of them at a time.Traditionally, a multi-RAN capable UE (e.g. one that could connect to 2G, 3G and 4G access networks)would only ever be connected to one RAT at a time, and all connections would be established via theactive RAT. Multi-access PDN Connectivity allows a UE to maintain, for example, a low-bit rate VoLTE
(Voice over LTE) connection via an enhanced 2G GERAN connection whilst also maintaining a high-bitrate data streaming connection via LTE-Advanced. ANDSF (Access Network Discovery & SelectionFunction) is also designed to improve multi-RAT operations by providing a means for a UE to availableinterrogate non-3GPP access types (e.g. WiFi, WiMax, CDMA2000 cells) and negotiate with their homecore network about which of the available options is best, thus ensuring optimal roaming connectivity.
CoMP transmission/reception was proposed as part of Release 10 LTE-Advanced as a means of
improving cell-edge coverage and of therefore boosting overall connection quality and data throughput
levels. It provides, in effect, a macro diversity facility for LTE in a manner that has also been described as
‘network MIMO’. Detailed specification of CoMP was held over the R11 and R12 work schedules.
The CoMP concept is simple; UEs located at cell edges or in other poor coverage areas may be offereda more consistent connection by either using multiple ‘points’ (eNBs, Relays, Remote Radio Heads, etc.)
to serve them or, by collating feedback from several neighbouring nodes, adjusting the scheduling
applied to them to minimize interference and improve reception quality.
CoMP operation is based on intra-site and inter-site co-ordination between eNBs; intra-site co-ordination
operates between different cells, relays and remote head units controlled by the same eNB and is
relatively simple to achieve, inter-site co-ordination operates between neighbouring eNBs and is more
complex to implement. The group of ‘points’ currently serving a UE are classed as its CoMP Co-
operating Set. Points in a set may be either ‘directly participating’, in which case they will be actively
handling traffic for the UE, or ‘indirectly participating’ in which case they will supplying measurements and
other information to assist with co-operative scheduling. CoMP can operate in both downlink (DL-CoMP)
and uplink (UL-CoMP) directions.
The two main forms of DL-CoMP are JP (Joint Processing) and DPS (Dynamic Point Selection). JP can
operate on the basis of JT (Joint Transmission), where multiple points transmit copies of the same data
in the same subframe simultaneously to improve reception of that data at the UE, or CS/CB (Co-
ordinated Scheduling/Co-ordinated Beamforming), where only the serving cell transmits data to the UE.
During CS/CB however, the serving cell and its co-operating neighbours manage scheduling to ensure
that interference to and by all UEs is minimized. In DPS only one point out of a Co-operating Set will
transmit to a UE at any one time, but the choice of which point to transmit from can vary on a subframe,
or even a Resource Block, basis.
UL-CoMP consists of JR (Joint Reception), where transmissions from a UE are received by multiple
points and are passed to the serving cell for processing, and uplink CS/CB scheduling management.
The ongoing design and implementation of cellular systems is pushed forward by equipment vendors,
network operators and, in most cases, international specifications bodies such as 3GPP. A degree of
central guidance to this process is provided by the ITU, a branch of the United Nations responsible for
attempting to harmonize the standards and techniques employed in worldwide telecoms systems.
The ITU set out its vision for 3G cellular systems at the end of the 1990s by devising a framework knownas IMT-2000, although the 3G market ended up consisting of several competing technological platforms
all of them ultimately adhered to the guidelines espoused by the ITU.
The ITU created a similar framework of expectations for emerging 4G technologies in 2007 with the initial
publication of the IMT-Advanced guidance.
Instead of specifying a particular technology or mandating the functionality of a 4G system, the IMT-
Advanced guidelines proposed the basic set of capabilities that a system should be able to support for it
to be regarded as being a true ‘4G’ system.
In addition to the aspirations outlined in the diagram, the ITU (in ITU-R report M.2134) also provided
some more definite technical requirements for 4G systems, including: that an IMT-Advanced systemshould be able to accommodate radio channels of at least 40MHz in bandwidth; that their peak spectral
efficiency should meet or exceed 15bits/Hz in the downlink and 6.75bits/Hz in the uplink; that it should
support latencies of less than 100ms for control-plane connections and less than 10ms for the user-
plane.
It should be noted that the original (Release 8) version of LTE did not meet the full set of IMT-Advanced
requirements, mainly due to its inability to support 1 Gbit/s data connections. To date only Release 10
LTE-Advanced and the updated version of Mobile WiMax (IEEE 802.16m) have been accepted as fully
compliant with IMT-Advanced.
Further Reading: www.IMT-2000.org (ITU website), ITU-Report M.2134
Basic LTE-Advanced functionality was first described in 3GPP Release 10 specifications, with the more
detailed and enhanced functionality held over to Release 11.
Release 10 functions include: Carrier Aggregation, which allows a UE to be scheduled with data-carrying
capacity on up to five parallel cells simultaneously; enhanced MIMO techniques that allow a UE to make
use of up to 8 MIMO layers on a downlink and up to 4 layers on an uplink; additional types of MIMO for uplink connections; the extension of LTE into a variety of new radio bands; enhancements to the SON
concept; and a range of data-handling improvements such as LIPA (Local IP Access) and SIPTO
(Selective IP Traffic Offload).
Release 11 introduced only one significant new feature, known as CoMP, and mainly concentrated on
adding more depth and complexity to existing R10 services.
Detailed information about the work and study items scheduled for each release are contained in the
relevant 3GPP Release Description documents, which provide a regularly-updated view of the work
being undertaken, or that has been completed, for each release.
Further Reading: http://www.3gpp.org/ftp/Information/WORK_PLAN/Description_Releases/
CA (Carrier Aggregation) is the most prominent feature of Release 10 LTE-Advanced. It offers an inverse
multiplexing facility that allows a UE to substantially increase the overall data rate it can achieve by
allowing an eNB to schedule capacity for it on multiple cells (or 'carriers') simultaneously.
Each carrier (either downlink or uplink) assigned for use by a UE is known as a CC (Component Carrier)
and the set of CCs allocated to a UE at any one time forms a Carrier Aggregate. R10 CA permits up tofive CCs to be bound into a Carrier Aggregate, potentially providing a suitably-equipped UE with up to
100MHz of bandwidth and an aggregate downlink data rate of over 3Gbit/s.
The lowest level of carrier aggregation allows a UE to connect via just one cell. The radio connectivity of
this cell is described as the PCC (Primary Carrier Component) and the cellular service it offers is known
as the PCell (Primary Serving Cell). The PCell carries NAS and RRC services for a UE and is also the
carrier measured by the UE to support functions like quality feedback and handover measurements.
A Release 8/9 UE (or an R10 UE that didn’t require CA services) would just connect via the PCell and
would not be assigned any additional carriers. An R10, CA-capable UE that did require a CA service
would be scheduled with capacity on between one and four SCells (Secondary Serving Cells), each of
which would be carried by an SCC (Secondary Component Carrier).
The PCC and any SCCs aggregated to provide a CA service for a UE must all be under the control of the
same eNB, but the terms ‘primary’ and ‘secondary’ used in relation to CA carriers are determined from
the point of view of each UE – different UEs in the same area may have selected different cells to be
their PCell and may therefore regard an assigned cell as an SCC which may be employed by another UE
as a PCC.
A PCell is always used in a bidirectional manner, as befits the cell that carries NAS and RRC traffic, but
an SCell may be used in either a bidirectional or unidirectional manner depending upon local
configuration and current requirements. If an SCell is used unidirectionally then it is only able to operate
in 'downlink-only' mode, there is no provision for cells to operate in 'uplink-only' mode.
Carrier Aggregation operates by allowing an eNB to schedule uplink and downlink capacity for a UE
across a set of component carriers that have been combined to form an Aggregated Channel.
The total bandwidth occupied by the Aggregated Channel is termed the Aggregated Channel Bandwidth
and includes all used and unused subcarriers; unused subcarriers are generally those set aside to act as
guard bands at the upper and lower end of each CC. The net usable bandwidth of the AggregatedChannel (which is total bandwidth less all guard subcarriers) is termed the Aggregated Transmission
Bandwidth.
Due to the requirements imposed by OFDMA, each CC will still need to have its centre-frequency left
unused to act as a DC Subcarrier. To ensure that orthogonality is maintained between subcarriers across
an entire aggregated channel, the DC Subcarriers of all CCs within the aggregate must be spaced at
multiples of 300kHz from each other. The number of guard subcarriers employed at the top and bottom
ends of each CC may be adjusted to accommodate this spacing requirement if necessary.
In a contiguous aggregated channel (i.e., one in which all CCs are adjacent to each other) the DC
subcarrier of the central CC will be used as the baseband transmission DC Subcarrier for the entire
aggregate. For this reason, it is recommended that aggregate channels are created in a ‘balanced’manner, as shown in the diagram. The total 25MHz aggregated channel bandwidth in the diagram has
been created by aggregating 2x10Mhz and 1x5MHz CCs, but to balance this aggregate the 5MHz CC
has been placed in the centre and its DC Subcarrier will be used for the entire aggregated channel.
Un-balanced configurations are also possible – a 25MHz channel aggregate could be created by
combining CCs in a 10+10+5 pattern, for example – but in these cases the aggregate DC Subcarrier
must be created by using whichever ‘transmission’ subcarrier happens to be at the centre frequency of
the aggregate.
In non-contiguous aggregates (like those employed in inter-band configurations), each contiguous
section of the aggregate will require its own aggregate DC subcarrier.
Further Reading: 3GPP TS36.807:5.6A & 5.7, 36.104:5.7.1A
The cost of LTE UEs is partly determined by the sophistication of the set of functions that each can
perform. Only a limited subset of UEs will be designed to take advantage of the maximum possible
capabilities of an LTE network. It is therefore necessary to allow each UE to signal its capabilities to the
network to allow the services offered to be tailored to its abilities.
UE capabilities are signalled in RRC message IEs and are classified into UE Categories.
Release 8/9 UEs have capabilities – those that relate to ‘standard’ LTE access – that place them in UE
Categories 1-5, these are signalled in the UE-EUTRA_Capability IE.
Release 10 UEs may have capabilities that relate to LTE-Advanced access which place them in UE
Categories 6-8, these are signalled in the UE-EUTRA-Capability-v10 IE.
A Release 8/9 UE with UE Category 5 can, in theory, achieve data rates of up to 300MBit/s in the
downlink and 150MBit/s in the uplink using 64QAM modulation in both directions and up to 4 layers of
downlink MIMO.
A Release 10 UE with a UE Category of 8 can, in theory, achieve data rates of 3000MBit/s downlink and1500MBit/s uplink using 64QAM in both directions, up to 8 MIMO layers in the downlink and up to 4 in the
uplink. The use of CA with 4 SCells to achieve these data rates is implicitly indicated as they wouldn’t be
possible otherwise.
Both UE-EUTRA-Capabilities signal a large number of other parameters in addition to UE Category and
other IEs are also employed to carry UE configuration and capability information.
Further Reading: 3GPP TS36.306:4.1 (UE-Category IE), 36.331:6.3.6 (UE-EUTRA-Capability)
The level of Carrier Aggregation that is supported by a UE is determined by its Bandwidth Class.
Each Bandwidth Class describes the level of CA capabilities supported by a UE, it includes descriptions
of the maximum number of Resource Blocks that a UE can use simultaneously and the maximum
number of Component Carriers (with each CC classed as an uplink/downlink pair whether the uplink is
actually used or not) the UE can access simultaneously. The largest channel bandwidth available in R10is 20MHz, so the maximum number of CCs parameter implicitly identifies the maximum operational
bandwidth a UE could be assigned.
Bandwidth Class A devices do not support CA, so this classification would be used to describe all
Release 8/9 UEs and any Release 10 devices that do not require CA capabilities. Class A devices are
able to access up to 100 Resource Blocks on a 20MHz channel in a PCell only and will not be able to
access any SCells. Examples of Bandwidth Class A devices might include voice-only LTE terminals and
devices designed to support low-bandwidth data applications such as smart utilities meters.
Bandwidth Classes B and upwards support increasing levels of aggregation capability. Class B devices
are also limited to using up to 100 RBs but may be assigned one SCell in addition to the PCell. The 100
RB limit means that although a Class B device is able to access two CCs it would not be able to accesstwo 20MHz CCs, as that would entail processing 200 RBs.
Bandwidth Class C devices are able to handle up to 200 RBs and two CCs and would therefore be able
to be assigned to, for example, a 20MHz PCell and a 20MHz SCell, although other channel
configurations are also possible.
As of v10.2.1 of TS36.101, the 3GPP specification that describes Bandwidth Classes, classes D-F had
parameters assigned to them but were described as being FFS (For Further Study), so the parameters
shown in the diagram are only provisional and may be subject to change by later versions of the
document. However, the maximum capability for a Class F device is likely to be up to five CCs covering a
A PCell, or Primary Serving Cell, is essentially the same as a standard R8/9 LTE cell.
A UE, whether it supports R8/9 or R10, will undertake the same Idle Mode and cell selection procedures
when active and will camp on a serving cell from the set locally-available based on signal quality and
strength indicators. Once a cell has been selected and the UE has camped on it, it becomes that UE’s
PCell. The current PCell is always assigned Serving Cell Index 0 by the UE.
A UE will only monitor the BCCH of a PCell, any system information required to access an SCell will be
provided to the UE using dedicated RRC signalling during SCell Addition or Modification.
An Idle Mode UE will monitor the PCell’s paging channel and will use measurements taken of the PCC
as the basis for idle mode reselection actions. If necessary, the UE will use PCell’s RACH to transmit
Service Requests to initiate a move to Connected Mode (moving from ECM-Idle to ECM-Connected).
A UE in Connected Mode will initially use the resources of the PCell; all RRC and NAS transactions take
place via the PCell and any ciphering applied to established traffic and control channels will be based on
the ciphering calculated for the PCell. The UE will send Uplink Capacity requests via the PCell PUCCH
(Physical Uplink Control Channel) and will monitor the PCell PDCCH (Physical Downlink ControlChannel) for downlink capacity notifications and uplink scheduling grants.
If a UE in ECM-Connected mode supports CA (e.g. if it has a Bandwidth Class greater than Class A), if
the serving eNB supports CA, if the traffic flow demanded by the UE’s currently-active applications is
sufficiently high, and if there is enough capacity available to support it, then the UE may be assigned one
or more SCells and could have capacity scheduled on them.
SCells are assigned by a controlling eNB for UE traffic capacity expansion purposes only and are notused for signalling or administrative purposes. SCells will be assigned UE-specific SCell Index values of
between 1 and 7 by the eNB, the UE will use this index as the basis for the Serving Cell Index that it
stores for each SCell.
The assignment of an SCell by an eNB will be signalled to the UE using the RRC SCell Addition
message, which will also specify the SCell Index, its Physical Cell ID and EARFCN channel number, its
bandwidth, mode (bidirectional or downlink-only), whether cross-carrier scheduling is to be employed,
and other parameters. SCell Modification messages are used to change the configuration of an existing
SCell and the SCell Release message is used to deassign them.
Once an SCell has been added to a UE's CA set it can be activated and deactivated as traffic flow and
capacity availability dictate – on initial assignment all SCells are deactivated. SCell status is signalled bythe eNB using the SCell Activation/ Deactivation MAC control element, which can optionally be appended
to a downlink MAC PDU header to signal any changes to the status of the set of SCells. SCell
deactivation is usually triggered by the expiry of an SCell inactivity timer in the eNB and SCell activation
is usually triggered by an increase in uplink or downlink traffic flow. The PCell cannot be deactivated at
any time.
When an SCell is active, a UE will be required to monitor its PDCCH for scheduling notifications (unless
cross-carrier scheduling is enabled) and will be expected to provide quality feedback in relation to the
SCell CCs. When an SCell is deactivated a UE Is not required to perform any actions in relation to it.
Scheduling for all forms of LTE is managed via the PDCCH.
For R8/9 UEs and for R10 UEs that either don't support CA or that do not currently have CA activated,
only the PCell's PDCCH will carry relevant scheduling information.
For R10 UE’s with CA activated and SCells assigned, there are two scheduling information options,depending upon whether cross-carrier scheduling has been enabled.
Cross-carrier scheduling allows scheduling notifications for one or more SCells to be carried by the
PDCCH of another active cell, typically the PCell. In such cases the capacity notification will indicate
details of the capacity granted plus the index of the intended CC. The typical benefit of cross-carrier
scheduling is that it allows a UE to monitor just one PDCCH to get access to all of its capacity
notifications. Cross-carrier scheduling is not an option for the PCell, whose capacity notifications are
always carried on its own PDCCH.
If cross-carrier scheduling is not enabled for an SCell then capacity notifications for that cell will be
carried on its own PDCCH.
Cross-carrier scheduling can be enabled or disabled on an SCell by SCell basis, based on an option in
the RRC SCell Addition and Modification messages, meaning that a UE could in theory have it enabled
for some cells in its CA set but not for others.
Whatever its complexities, LTE-Advanced Carrier Aggregation can ultimately be seen as an exercise in
multi-carrier scheduling and its success or otherwise in an individual vendor's equipment will be
determined by the efficiency of the scheduling algorithm they have developed.
R10 LTE-Advanced shares the same air interface protocol stack as standard R8/9 LTE, namely PDCP
over RLC over MAC over the OFDMA-based PHY (Physical Layer).
Carrier Aggregation is mainly accommodated by making changes to the MAC layer. Each UE-specific
MAC instance in an eNB and the corresponding MAC layer in each UE perform a set of functions that are
common for all active serving cells (such as MAC PDU creation and scheduling), but there is also a set of cell-specific functions performed separately for the PCell and each Scell. These functions mainly relate to
the establishment of separate HARQ sessions for each UL or DL SCH (Shared Channel) that is active.
At the Physical layer, the eNB and UE will establish separate instances of UL-SCH and DL-SCH for each
CC and for each MIMO layer on each CC that is active for the UE connection. Traffic travelling over each
of these separate physical layer connections will be scheduled capacity by the MAC layer as required to
meet the QoS agreed for each E-RAB.
The upper reaches of LTE-Advanced capability, as evidenced by the parameters associated with UE
Category 8, can in theory allow a UE to achieve 3000MBit/s on the downlink and up to 1500MBit/s on the
uplink. To enable this to take place, the UE is likely to be operating with a full set of serving cells (one
PCell and four SCells), to be using the highest available modulation scheme (64QAM on both the uplinkand the downlink) and to be using the full set of available MIMO streams. This would equate to 8 streams
per DL CC and 4 streams per UL CC – or a combined 40 separate logical downlink physical layer
connections and 20 in the uplink.
The possibility of a UE (or a network) achieving commercial support for such extreme data rates and
operating parameters in the near term is exceedingly remote and far more modest data rates of a few
hundred megabits per second is a much more achievable target for most LTE-Advanced operators.
3GPP Release 10 and 11 specifications also introduce a number of techniques designed to enhance theways in which mobile data is handled, especially when users have access to femtocells or wifi hotspots.
LIPA is designed to allow a UE to make use of the resources of a home or office LAN (to which the user has access) via LTE; the intention is that a UE could access a network printer, or send content to an
Internet-enabled TV, or stream music from a home content server, all via a LAN-connected femtocell.Managed Remote Access aims to provide the same kind of home/office LAN access service but workswhile the user is roaming away from their home femtocell; remote access to local resources is providedfrom anywhere on the LTE network via the secure connection created to serve the user’s homefemtocell.
SIPTO is designed to detect and make use of geographically local EPS Gateways (S-GW and PDN-GW)for certain PDN connections. A roaming UE may, for example, have one PDN connection (connected toan application server that only exists in the home network) directed back to an APN in a home networkPDN-GW, whilst a PDN connection to the general Internet could be directed to a local, visited network APN. SIPTO works in both roaming and non-roaming cases and an APN’s applicability for SIPTO issignalled in the user’s HSS profile.
IP Flow Mobility aims to enable better interworking between LTE and WiFi, in this case by providing themeans to seamlessly hand active connections over between LTE and Wi-Fi access devices, but only if the Wi-Fi access is controlled by the LTE network operator. A UE that, for example, has an active voicecall and is also in the process of downloading a large file from an Internet-based server, could use IPFlow Mobility to divert the data download to an active (operator-controlled) Wi-Fi connection, thusoffloading traffic from the LTE network and preventing the user from consuming unnecessary amounts of their data bundle.
Multi-Access PDN Connectivity allows an operator to make efficient use of their portfolio of cellular access systems by allowing UEs to setup simultaneous connections via more than one of them at a time.Traditionally, a multi-RAN capable UE (e.g. one that could connect to 2G, 3G and 4G access networks)would only ever be connected to one RAT at a time, and all connections would be established via theactive RAT. Multi-access PDN Connectivity allows a UE to maintain, for example, a low-bit rate VoLTE
(Voice over LTE) connection via an enhanced 2G GERAN connection whilst also maintaining a high-bitrate data streaming connection via LTE-Advanced. ANDSF (Access Network Discovery & SelectionFunction) is also designed to improve multi-RAT operations by providing a means for a UE to availableinterrogate non-3GPP access types (e.g. WiFi, WiMax, CDMA2000 cells) and negotiate with their homecore network about which of the available options is best, thus ensuring optimal roaming connectivity.
CoMP transmission/reception was proposed as part of Release 10 LTE-Advanced as a means of
improving cell-edge coverage and of therefore boosting overall connection quality and data throughput
levels. It provides, in effect, a macro diversity facility for LTE in a manner that has also been described as
‘network MIMO’. Detailed specification of CoMP was held over the R11 and R12 work schedules.
The CoMP concept is simple; UEs located at cell edges or in other poor coverage areas may be offereda more consistent connection by either using multiple ‘points’ (eNBs, Relays, Remote Radio Heads, etc.)
to serve them or, by collating feedback from several neighbouring nodes, adjusting the scheduling
applied to them to minimize interference and improve reception quality.
CoMP operation is based on intra-site and inter-site co-ordination between eNBs; intra-site co-ordination
operates between different cells, relays and remote head units controlled by the same eNB and is
relatively simple to achieve, inter-site co-ordination operates between neighbouring eNBs and is more
complex to implement. The group of ‘points’ currently serving a UE are classed as its CoMP Co-
operating Set. Points in a set may be either ‘directly participating’, in which case they will be actively
handling traffic for the UE, or ‘indirectly participating’ in which case they will supplying measurements and
other information to assist with co-operative scheduling. CoMP can operate in both downlink (DL-CoMP)
and uplink (UL-CoMP) directions.
The two main forms of DL-CoMP are JP (Joint Processing) and DPS (Dynamic Point Selection). JP can
operate on the basis of JT (Joint Transmission), where multiple points transmit copies of the same data
in the same subframe simultaneously to improve reception of that data at the UE, or CS/CB (Co-
ordinated Scheduling/Co-ordinated Beamforming), where only the serving cell transmits data to the UE.
During CS/CB however, the serving cell and its co-operating neighbours manage scheduling to ensure
that interference to and by all UEs is minimized. In DPS only one point out of a Co-operating Set will
transmit to a UE at any one time, but the choice of which point to transmit from can vary on a subframe,
or even a Resource Block, basis.
UL-CoMP consists of JR (Joint Reception), where transmissions from a UE are received by multiple
points and are passed to the serving cell for processing, and uplink CS/CB scheduling management.