You are on page 1of 102

1

AP
I

IS

AP

AP
IS

AP
IS

AP

IS

LTE is in many respects similar to High Speed Packet Access (HSPA) for UMTS, which was originally
introduced already in Release 5 (finalised in 2002). Both HSPA and LTE makes use of the concept of
shared channels for signalling and data. Both use dynamic, channel quality dependent, packet
scheduling, adaptive modulation (QPSK/16QAM/64QAM) and a Hybrid-ARQ (HARQ) protocol for
retransmissions. Both may also be deployed with multiple-antenna techniques (MIMO). Finally, both
systems have clever base stations, where scheduling decisions may be taken very quickly to
compensate for changes in radio channel quality. (The LTE base station is a bit more clever than the
NodeB though, since it also performs tasks that, in HSPA, is performed by the Radio Network Controller,
RNC.)
The selected radio access technology, however, is very different between the two. HSPA uses Wideband
Code Division Multiple Access (WCDMA) with a fixed system bandwidth of 5 MHz while LTE uses
Orthogonal Frequency Division Multiplexing (OFDM) with variable system bandwidth (1.4, 3, 5, 10, 15 or
20 MHz). With OFDM, the data block to be transmitted to/from the UE is split and distributed over
many narrowband (15 kHz) so-called subcarriers, as opposed to over one wideband (5 MHz) carrier as in
WCDMA.
The figure lists the theoretical maximum bit rates for downlink and uplink in HSPA and LTE. Just by
looking at the numbers one gets the impression that LTE is clearly superior in this respect. This is not
true. When comparing the two, one should assume the same system bandwidth, say 5 MHz, and the
same number of antennas for MIMO operation, say 2. Otherwise the comparison becomes highly unfair
(with HSPA being the loser). Actually, the two have about the same performance in terms of maximum
bit rates when they are deployed with the same system bandwidth and the same number of transmit
antennas.

AP
I

The extreme bit rates for LTE Release 10 reflect the drastic extensions to the baseline LTE specification
that was introduced as part of that release. These R10 LTE enhancements are known under the name
LTE-Advanced.

AP
IS

The table lists all supported frequency bands as of Release 10. Frequency bands marked in blue were
added in Release 9 and those marked in red in Release 10. Additional frequency bands, as well as
extensions of some existing bands, are typically added in each release.

AP
I

AP

IS

The LTE physical layer specifications are written in such a way that it does not really matter on what
physical carrier frequency the system is deployed. Also, the Frequency Division Duplex (FDD) and Time
Division Duplex (TDD) options are harmonised to a very large extent (as opposed to the FDD and TDD
options for UMTS which are, more or less, completely different radio access technologies).

AP
IS

IS

The work on LTE started in Dec 2004, when a feasibility study on the UTRA & UTRAN Long Term
Evolution was started. The study phase for LTE was part of Release 7, with the actual standardisation
being part of Release 8. Similarly, the feasibility study phase for LTE-Advanced was part of Release 9,
with the actual standardisation being part of Release 10.

AP
I

AP

A substantial effort was put into aligning the LTE-Advanced work with the various deadlines set up by
the ITU for IMT-Advanced (as can be seen above).

AP
IS

AP
I

AP

IS

Some of the performance requirements for IMT-Advanced are listed above, together with the
performance of LTE Release 8/9 and Release 10 (LTE-Advanced). As can be seen, LTE R8/R9 almost
fulfills the ITU requirements. Furthermore, the performance numbers for LTE-Advanced will eventually
(R11+) exceed those given in the table. The LTE layer 1 specifications are, already in R10, prepared for
peak bit rates of 3000 Mbit/s in the downlink and 1500 Mbit/s in the uplink.

AP
IS

The table above is a summary of all 3PP releases from Release 5, introduced in 2002, to Release 13,
scheduled for completion in 2016, focusing on HSPA and LTE radio related features.

AP

IS

LTE and EPC were first introduced in Release 8 and this is the basic release still being deployed around
the World (Dec 2013). The evolution of the 3GPP standard in general and LTE/EPC in particular does
not, of course, end with Release 8. Release 9 was (officially) completed in December 2009 (with some
features postponed to the March/June standardisation meetings). R9 contains some left-overs that
where not finished in time for the R8 freeze (such as Multimedia Broadcast Multicast Service, MBMS) as
well as new features (such as emergency calls over LTE).
One of the major work items in R9 was the preparatory work for further development of E-UTRA and EUTRAN. This work is referred to as LTE-Advanced (LTE-A) and aims at fulfilling the requirements set up
by the ITU for future 4th generation mobile systems (IMT-Advanced). The actual standardization work
for LTE-A begun as part of Release 10 (frozen in June 2011) and continued in R11 (frozen in June 2013).
The work on R12 is still (Jan 2014) ongoing with an expected completion date of September 2014. Only
a few features have yet been announced for R13- please consult the 3GPP website for an up to date list
of Work Items.

AP
I

Please note that the features listed above are just examples of features introduced in the corresponding
release. Note also that a given 3GPP release includes new features for all radio technologies defined at
the time, i.e. UMTS and GSM and LTE. (A quite common misunderstanding is that R8 and onwards is
only relevant for LTE access.)

AP
IS

AP

IS

There are three different channel levels in the E-UTRAN channel architecture: Logical channels,
Transport channels and Physical channels. The logical channels describe the type of information to be
transmitted, the Transport channels describe in what format the information is to be transmitted and
the physical channels provide the transmission media through which the information is actually
transmitted. Flows from several logical channels can, in the same time instant, be mapped to the same
transport channel. Thus, there is not necessarily a one-to-one relationship between logical and
transport channels.
Please note also the location of Signalling Radio Bearers (SRB) and Data Radio Bearers (DRB) above the
Packet Data Convergence protocol in the figure above. The SRBs are used for RRC control signalling and
the DRBs are used for any form of data traffic.
Logical Channels
A logical channel is an information stream dedicated to the transfer of a specific type of information
over the radio interface. Logical channels are provided between the Radio Link Control (RLC) and
Medium Access Control (MAC) protocol layers in UE and eNB. The logical channels are shortly described
in the following:
Broadcast Control Channel (BCCH). Downlink channel for broadcasting system information. BCCH is
mapped onto the BCH and DL-SCH transport channels.
Paging Control Channel (PCCH). Downlink channel that carries paging information. Always mapped
onto the PCH transport channel.
Common Control Channel (CCCH). This is a bi-directional channel for transmitting initial RRC control
signalling between the UE and eNB. The CCCH logical channel is always mapped onto the UL/DL-SCH
transport channels.
Dedicated Control Channel (DCCH). Point-to-point bi-directional channel for sending dedicated RRC
control signalling between the UE and the eNB. This channel is always mapped onto the UL/DL-SCH
transport channels

AP
I

Dedicated Traffic Channel (DTCH). Point-to-point channel dedicated to one UE (uplink or downlink or
both) for transmission of user data. Always mapped onto the UL/DL-SCH transport channels.

AP
IS

Transport channels are offered from PHY to MAC for signalling or data transport. Different transport
channels are defined by how and with what characteristics the information is transmitted on the
physical layer. Information on transport channels is delivered to/from the physical layer in the form of
Transport Blocks (TB). One or two TBs are delivered per Transmission Time Interval (TTI). The TTI
length for E-UTRA is 1ms for most transport channels.
Broadcast Channel (BCH). Carries part of the System Information (SI) in a cell (the Master Information
Block, MIB). The BCH is mapped onto the PBCH physical channel. Additional SI blocks (SIB1, SIB2 etc)
are mapped onto DL-SCH.
Paging Channel (PCH). The PCH carries paging messages from eNB to the UE (or group of UEs). The
PCH is mapped onto the same physical resource as the DL-SCH.
Downlink Shared Channel (DL-SCH). This is the main downlink resource in E-UTRA. It carries data
(DTCH) and signalling (BCCH, CCCH and DCCH). The DL-SCH uses hybrid-ARQ (HARQ), dynamic packet
scheduling and adaptive modulation and coding. The DL-SCH is mapped onto the PDSCH physical
channel.
Random Access Channel (RACH). Uplink channel used to carry control information from the UE to the
eNB. The RACH is used for initial access, when the UE is not known in the eNB, and for handover
access. It is also used when the UE wishes to transmit on the PUSCH and the last assigned timing
advance value has expired. The corresponding physical channel is the PRACH.

Uplink Shared Channel (UL-SCH). This is the main uplink resource in E-UTRA. It carries data (DTCH)
and signalling (CCCH and DCCH). The UL-SCH uses hybrid-ARQ (HARQ), dynamic packet scheduling
and adaptive modulation and coding. The UL-SCH is mapped onto the PUSCH physical channel.
Physical Channels

IS

A physical channel corresponds to a well-defined radio resource where information pertaining to a


particular transport channel is transmitted. Different levels of reliability are achieved by applying
different layer 1 processing (channel coding, interleaving etc) to different physical channels.

AP

Physical Broadcast Channel (PBCH). This channel carries the MIB from the BCH transport channel.
Physical Downlink Shared Channel (PDSCH). This is the main downlink radio resource in a cell,
carrying data and/or higher layer signalling. The PDSCH is allocated to different UEs on a per-TTI basis
(i.e. every 1ms). The channel coding, modulation and subcarrier allocation for each transmission is
selected by the MAC-layer scheduler. Since the PDSCH is a shared resource, and since the Transport
Format used is dynamic, all downlink transmissions must be explicitly addressed to the receiving UEs.
This is done on the PDCCH.
Physical Downlink Control Channel (PDCCH). The downlink control channel is used for indications of
downlink transmission on PDSCH as well as for allocation of uplink resources on PUSCH/PUCCH. The
PDCCH signalling is located in the first 1-3 OFDM symbols in each 1ms long sub-frame. Multiple
physical downlink control channels are supported and a UE monitors a set of control channels.
Physical Control Format Indicator Channel (PCFICH). The sole purpose of this physical channel is to
indicate the size, in OFDM-symbols, of the control area used for the PDCCH transmission in the same
sub-frame.

Physical HARQ Indicator Channel (PHICH). The purpose of this channel is to transmit ACK/NACKs
related to uplink transmissions on the PUSCH. Thus, each PHICH is addressed to a single UE at a time.
Physical Uplink Shared Channel (PUSCH). This is the main uplink radio resource in a cell, carrying data
and/or higher layer signalling. The PUSCH is allocated to different UEs on a per-TTI basis. The channel
coding, modulation and subcarrier allocation for each transmission is selected by the MAC-layer
scheduler. As for the PDSCH, all transmissions must be explicitly allocated to a given UE. This is done
with the PDCCH as mentioned above.
Physical Uplink Control Channel (PUCCH). The PUCCH conveys uplink control information in the form
of channel quality indicator (CQI), uplink scheduling requests and ACK/NACKs for data received on
the PDSCH.

AP
I

Physical Random Access Channel (PRACH). The PRACH carries the random access preambles during
the random access procedure.

Transport Channels

AP
IS

AP

IS

Orthogonal Frequency Division Multiplexing (OFDM) can be seen as a combination of Frequency


Division Multiplexing (FDM) and Time Division Multiplexing (TDM) in that it allows users to share a
resource in both the frequency domain and the time domain. OFDM is a multi-carrier modulation
scheme that uses a large number of closely spaced orthogonal subcarriers. Each subcarrier is modulated
with a conventional modulation scheme (such as 16QAM) at a low symbol rate, maintaining data rates
similar to conventional single-carrier modulation schemes in the same bandwidth. The primary
advantage of OFDM over single-carrier schemes is its ability to cope with severe channel conditions
without complex equalization filters. Low symbol rate makes the use of a guard interval between
symbols affordable, making it possible to handle time-spreading and inter-symbol interference (ISI).
With hundreds or thousands of subcarriers available it becomes quite straightforward how to multiplex
users on the radio interface. Simply allocate different sets of subcarriers to different users (this is the
FDM in OFDM). More complex multiplexing schemes can be implemented by allowing users to share
the available subcarriers both in the frequency domain (FDM) and the time domain (TDM). This is
incicated with different colours in the figure.
The OFDM technology as such is about 50 years old (it was first used around 1957 by the US Navy).
During the 70s and 80s several important theoretical contributions from various sources made it
possible to implement more efficient and robust OFDM-based systems. Today, OFDM has proved itself
as the preferred radio access technology in a wide variety of communication systems. Some examples
of OFDM use: IEEE 802.11a/g (WLAN/WiFi), IEEE 802.16 (WiMAX), Digital Audio Broadcasting (DAB),
Digital Video Broadcasting (DVB-T and DVB-H) and Asynchronous Digital Subscriber Line (ADSL). Some
advantages of OFDM:
Allows adaptation to severe channel conditions without very complex equalization methods
Robust against Inter-Symbol Interference (ISI) and fading caused by multipath propagation
High spectral efficiency
Efficient implementation using FFT

Low sensitivity to time synchronization errors

AP
I

Some disadvantages of OFDM are its sensitivity to Doppler spreading and frequency synchronization
errors. OFDM also suffers from a high peak-to-average power ratio (PAPR), requiring more expensive
transmitters and lowering power efficiency.

AP
IS

AP

IS

There are many possible ways of realising an OFDM modulator, none of which is covered by the LTE
specifications. In practice though, the OFDM modulator is typically implemented by means of an Inverse
Fast Fourier Transform algorithm (IFFT) on the transmitting side and a Fast Fourier Transform algorithm
(FFT) on the receiving side. (There are a number of widely known algorithms available to realise the
IFFT/FFT logic itself)
The IFFT takes as input a block of N modulation symbols in parallel, where N equals the number of
assigned subcarriers. Each modulation symbol is then used for modulating one subcarrier, in effect
acting as a complex-valued weight setting the amplitude and phase of the subcarrier. The output
samples from the IFFT are then converted back to serial form. The OFDM symbol can thus be seen as a
train of samples.
The last few samples of the OFDM symbol are then copied and appended at the front of the symbol.
This is the Cyclic Prefix (CP) part of the OFDM symbol. It is common to refer to the original train of
samples as the useful part of the OFDM symbol (this does not by any means make the CP part unuseful!)

The CP is inserted to eliminate inter-symbol interference due to multipath propagation effects. The LTE
specifications define a normal length and an extended length of the CP, to cater for the different
requirements of small versus large cells. There are also different CP lengths defined for MBMS
transmission, when multiple synchronised eNBs act as a Single Frequency Network (SFN).
The OFDM symbols are then used for modulation of the actual physical carrier frequency. In addition,
various pulse shaping or filtering techniques may be applied at this stage (D/A box in the figure).

10

AP
I

The receiving side runs the process in reverse. The IFFT process must be inverted in order to retrieve
the information content of the individual subcarriers. This is done with the Fast Fourier Transform (FFT).
(The inverse of the Inverse-FFT is, of course, the FFT.)

AP
IS

The LTE radio frames are 10 ms in duration, divided into 10 subframes of 1ms duration. Thus, the
subframe length coincides with the TTI. Each subframe is further divided into two slots.

AP

IS

In the time domain, each slot consists of Nsymb OFDM symbols, where Nsymb is the number of symbols
during a slot. Nsymb=7 with the normal CP length and Nsymb=6 with the extended CP. In the frequency
domain, each slot consists of a system bandwidth dependent number of subcarriers. The system
bandwith is defined as a multiple of 12 subcarriers. Such a set of 12 consecutive subcarriers is called a
Physical Resource Block (PRB). The system bandwidth is labelled NRB, which is simply the system
bandwidth given in units of PRBs.
In Release 8, the smallest possible system bandwidth is NRB = 6 (corresponding to 72 subcarriers and a
1.4 MHz system) and the largest is NRB = 100 (correspondig to 1200 subcarriers and a 20 MHz system).
The number of resource blocks that may be assigned to the UE in the frequency domain can range from
NRB-min = 1 to NRB-max = 100. Please note that the LTE TTI-length is 1ms (2 slots). Hence the minimum
time-frequency resource that can be assigned to one UE is one Resource Block pair (one RB is each slot).

11

AP
I

The 2-dimensional time-frequency radio resource can visually be represented as a resource grid as
depicted in the figure. Each little square within the grid represents a single subcarrier for one symbol
period and is referred to as a Resource Element (RE), resulting in each PRB containing 84 REs. Each RE
corresponds, in terms of channel coded information content, to one QPSK/xQAM modulation symbol
(i.e. 2, 4 or 6 channel coded bits). Note that for multi-antenna operation there is a resource grid present
for each transmitting antenna.

AP
IS

AP

IS

In the downlink, one subframe consists of a Control Region and a Data Region. The Control Region
contains the Resource Elements (RE) that are assigned for the downlink control channels: PCFICH,
PHICH (s) and PDCCH(s). The time domain size of the Control Region may vary between each subframe
(primarily depending on the amount of PDCCH information that needs to be transmitted during a given
TTI). The size varies between 1-3 OFDM symbols for system bandwidths larger than 10 PRB and
between 2-4 OFDM symbols for system bandwidths equal to or smaller than 10 PRB.
The Data Region covers all OFDM symbols within a subframe not used by the Control Region and
contains the REs assigned for PDSCH transmission to one or more UE. As already mentioned, the
downlink resources assigned to UEs for PDSCH transmission are referred to as Physical Resource Blocks
(PRB). However, the PRB concept is not used for transmission of the downlink control channels.
The PDCCH(s) are formed by aggregation of Control Channel Elements (CCE). Each CCE is, in turn, an
aggregation of 9 RE Groups (REG), distributed over the Control Region. Each REG consists of 4
consecutive REs that are "not used for other purposes". The last sentence means that the 4 REs in a REG
need not be strictly consecutive, since some REs are 'consumed' by antenna reference signals (see
below). The PHICH and PCFICH channels are constructed from aggregations of 3 and 4 REGs respectively
(the CCE concept is only used for the PDCCH).

12

AP
I

In LTE downlink the UE performs channel estimation and data demodulation based on Cell-specific
Reference Signals (CRS) transmitted from each antenna port. The CRSs are transmitted on every 3rd
subcarrier in a repeated, symmetric, time-frequency pattern. The CRSs are also antenna port specific,
allowing channel estimation on a per-port basis. While one transmitter antenna is sending its CRS, all
other antenna ports are idle (the figure shows the REs needed for 2 antenna ports).

AP
IS

IS

The figure above (taken from TS 36.11) shows the REs needed for Cell-specific Reference Signals in the 4
antenna port case. As mentioned before, while one transmitter antenna is sending its CRS, the other
antennae must be idle. In order to reduce the CRS overhead in the 4-antenna case, the time-domain
CRS density has been reduced for ports 2 and 3 (the CRS only occurs once per slot instead of twice).

13

AP
I

AP

The reduced CRS density means that the UE will not be able to track fast channel variations as well as in
the case with 2 antennae. However, transmission to the same UE with 4 antennae is really only relevant
in low-mobility cases (and hence slower time varations) and should not pose any significant problem in
practice.

AP
IS

AP

IS

The LTE uplink subframe can, like the downlink, be described in terms of a Control Region and a Data
Region, but with a very important difference: uplink transmission from a given UE is either a PUSCH
transmission (data) or a PUCCH transmission (control). The concept of Data and Control regions
therefore makes sense only if one takes the full system bandwidth into account (i.e. all uplink
transmissions from all UEs during a subframe).
With this in mind, the uplink Control Region is mapped onto the very edges of the system bandwidth.
The control region is defined in multiples of 12 subcarriers- the Physical Resource Blocks introduced
earlier. A PUCCH transmission from one UE always occupies 2 PRBs, with frequency hopping at the slot
boundary. Thus, the 1st PRB is mapped onto one edge of the uplink system bandwidth and the 2nd PRB
to the opposite edge.
The Data Region is mapped onto the remaining PRBs between the band-edge Control Regions and
carries PUSCH transmissions from all scheduled UEs. In Release 8 there are never any multi-antenna
transmissions taking place in the uplink. Hence there is no need for any antenna-specific Reference
Signal pattern distributed over the system bandwidth. However, the eNB needs some form of pilot
signal in order to perform coherent demodulation of the various uplink transmissions.
This some form of pilot signal is called the Demodulation Reference Signal (DM-RS). Even though the
DM-RS serves a similar purpose to the downlink CRS (i.e. channel estimation) it differs in a number of
aspects. First, the DM-RS is UE-specific- each UE transmits its own DM-RS. Second, it is not distributed in
the time and frequency domains. The DM-RS for a PUSCH transmission occupies all REs in one SC-FDMA
symbol. As a matter of fact, the DM-RS is often referred to as the Demodulation Reference Symbol.
Third, the DM-RS only covers the actually allocated bandwidth, as opposed to the full system bandwidth
covered by the downlink RS.

14

AP
I

The DM-RS is present in both PUSCH and PUCCH tranmsissions. The PUCCH DM-RS mapping differs
depending on the transmitted PUCCH format. For PUCCH format 1 (and normal CP length) it occupies 3
consecutive SC-FDMA symbols in the middle of the slot. For PUCCH format 2 (with normal CP) it
occupies the second and the next-but-last SC-FDMA symbol. PUCCH format 1 is used for HARQ
ACK/NACK signalling and PUCCH format 2 for channel quality reporting. Both formats are shown in the
figure above.

AP
IS

AP

IS

For any reception of any scheduled downlink (or uplink) transmission the UE needs to decode the
Physical Downlink Control Channel (PDCCH). As mentioned earlier, the PDCCH is mapped onto a persubframe variable-size Control Region at the beginning of each subframe. The size (in units of OFDM
symbols) is indicated on a per-subframe basis with the Physical Control Format Indicator Channel
(PCFICH). The PCFICH itself is always mapped onto 4 Resource Element Groups (REGs) mapped onto the
first OFDM symbol in each subframe. The 4 REGs are distributed over the whole system bandwidth, with
the exact frequency-domain location dependent on the Physical Cell ID.
The PDCCH(s) are formed by aggregation of Control Channel Elements (CCE). Each CCE is, in turn, an
aggregation of 9 REGs interleaved over the Control Region. The UE tries to decode all possible PDCCH
candidates, based on its configured Transmission Mode (TM) and its monitored Search Spaces (more
about this later).
A downlink assignment on the PDCCH consists of an RNTI, a set of Physical Resource Blocks (PRB), a
Modulation and Coding Scheme (MCS) and HARQ-related parameters such as process id and
redundancy version. The set of PRBs are assigned using one of three different Resource Allocation
types. Depending on the Resource Allocation type and the system bandwidth, the set of PRBs may be
contiguous or non-contigous and may be assigned with single- or multiple-PRB granularity. For some
transmission modes, the PDCCH also includes MIMO related information for the downlink transmission.
After receiving a downlink data transmission on the Physical Downlink Shared Channel (PDSCH) in
subframe n the UE then responds with a HARQ ACK or NACK indication in uplink subframe n+4. In case a
NACK is received, the eNB may schedule a retransmission earliest in subframe n+8. Hence, the HARQ
Round-trip Time (RTT) in E-UTRA FDD is minimum 8 subframes. This is the reason why the HARQ entity
makes use of 8 parallel stop-and-wait processes, thus allowing continuous downlink transmission to the
UE.

15

AP
I

The downlink scheduler in the eNB needs Channel State Information (CSI) from each UE in order to take
efficient scheduling decisions. The contents of the CSI depends on the given UEs configured TM. For
single-antenna transmission and transmit diversity the CSI consists of a Channel Quality Indication (CQI),
specifying the estimated downlink channel quality at the UE. For spatial multiplexing the UE should, in
addition to CQI, also report a Rank Indication (RI) specifying the preferred number of transmission layers
for multi-antenna transmission. Finally, in case of closed-loop spatial multiplexing, the UE should also
report the preferred Precoding Matrix Indication (PMI). CSI can be sent periodically (2-128ms for FDD)
or aperiodically (on explicit request from eNB).

16

AP
I

IS

AP

AP
IS

17

AP
I

IS

AP

AP
IS

AP
IS

AP

IS

Any wireless communications system with one transmit (Tx) antenna and one receive (Rx) antenna is
referred to as operating in Single Input Single Output (SISO) mode. Systems with multiple Tx/Rx
antennas are divided into Single Input Multiple Output (SIMO), Multiple Input Single Output (MISO) and
Multiple Input Multiple Output (MIMO) systems. Simple multi-antenna systems have been around, in
one form or another, for over 50 years (Guglielmo Marconi used multiple antenna transmission to
transmit a Morse signal across the Atlantic Ocean, from England to Newfoundland, in 1901). But until
quite recently the amount of signal processing needed has been too expensive to be practical for largescale deployment and implementation in small mobile devices. Important factors driving MIMO today is
the advent of in-expensive high-speed Digital Signal Processors (DSPs) and significant research
breakthroughs in information theory over the last decade. MIMO is an integral part of the 3GPP R8
standards pertaining to eHSPA and LTE.
In a SIMO system the transmitter has one antenna and the receiver has two, or more, physically
separated antennas (the physical separation distance has a direct relationship with the wavelength of
the carrier). This allows for receive diversity (Rx diversity). With Rx diversity the receiver picks up two
(or more) versions of the same transmitted signal. The receiver may then either select the best input
(from one of the antennas) or combine the input from all antennas.
In a MISO system the transmitter has two or more physically separated antennas and the receiver has
one antenna. This allows for transmit diversity (Tx diversity). With Tx diversity, the transmitter sends
redundant copies of a signal to the receiver. Tx diversity is based on the hope that at least one of the
copies is received in a good enough state to allow reliable decoding.

18

AP
I

A MIMO system uses multiple antennas at transmitter and receiver. It can be shown that the diversity
schemes mentioned above increase the system capacity logarithmically with the number of antennas.
Symmetric MIMO configurations (same number of Tx and Rx antennas) on the other hand, increase the
system capacity linearly with the number of antennas. The magic of MIMO lies in its ability to take
multipath reception, which used to be an unavoidable and undesired by-product of radio propagation,
and convert it into an advantage that actually multiplies transmission speed and improves throughput.
With MIMO, mathematical algorithms are used in order to spread the user data across multiple
transmitting antennas. The signals transmitted are defined in 3 dimensions: time, frequency and space.
At the receiver, the different signals from each antenna must be identified and separately decoded.
Hence, the (mathematical) technique of separating out different paths on the radio link is what allows a
MIMO system to transmit multiple signals at the same time on the same radio resource.

AP
IS

AP

IS

One way of realizing Tx diversity is to make use of Space-Time Coding (STC). This means that the signal is
transmitted through multiple antennas (space) at different points in time (the ST in STC). In the simple
case of two antennas, a super-position of 2 modulation symbols (e.g. QPSK symbols) is transmitted on
both antennas simultaneously. The very same modulation symbols are then transmitted again with a
(very) slight delay. In addition, the modulation symbols will be coded, or weighted, differently for the
second, slightly delayed, transmission (the C in STC). The coding and antenna mapping of the input
modulation symbols is realized by multiplying a set of modulation symbols (a vector of symbols) with a
precoding matrix.
The Space-Time Coder acts on a block (or vector) of modulation symbols at a time. The figure shows a
case where 2 symbols at a time are Space-Time precoded prior to transmission. The particular
precoding matrix shown is the so-called (un-weighted) Alamouti matrix, named after its inventor
Siavash Alamouti. In general, such precoding schemes are often referred to as Space-Time Block Coding
(STBC). It is (at least mathematically) quite straight-forward to extend the basic 2-antenna STBC scheme
to more than 2 antennas. It is also quite straight-forward to extend the STBC scheme in order to realise
true MIMO (i.e. spatial multiplexing as opposed to spatial diversity).

19

AP
I

In the case of transmit diversity in LTE, the precoding is better described as Space-Frequency Coding
(SFC) rather than STC. A block of modulation symbols are mapped to different antennas and to different
subcarriers, as opposed to different time-instants. This process we can refer to as Space-Frequency
Block Coding, or SFBC, which is used for 2-antenna transmit diversity in LTE. In the 4-antenna transmit
diverity case there is an additional frequency shift applied, where pairs of modulation symbols are
transmitted by means of SFBC and with the transmission alternating between pairs of antenna ports
and pairs of subcarriers. This is referred to as combined SFBC and Frequency Shift Transmit Diversity,
abbreviated SFBC/FSTD (not shown in the figure).

AP
IS

AP

IS

The multiple streams of a MIMO transmission must be orthogonal to each other- catastrophic
interference would follow otherwise. The orthogonality is achieved by multiplying the transmitted
streams with a precoding matrix, as discussed on the previous page. At the receiver, the stream from
each antenna must be (mathematically) identified and separately decoded. This requires knowledge in
the receiver of the precoding applied. The receiver must also be aware of the current channel state. The
LTE reference signals discussed in chapter 1 are used for exactly this purpose, to estimate the state of
the channel in order to separate multiple antenna streams from each other.
Multi-antenna transmission schemes can be divided into closed loop and open loop schemes. A closed
loop MIMO scheme requires the transmitter to provide the receiver with explicit precoding information
for each transmission, specifying what precoding matrix that was applied at the transmitter. In a closed
loop scheme the receiver must (periodically) send feedback to the transmitter. The feedback provided
by the reciever is an indication of a preferred precoding matrix.
In an open loop scheme there is no need for the receiver to send feedback to the transmitter. An
example of an open loop scheme is Alamouti diversity, discussed on the previous page. An open loop
scheme also usually do not require the transmitter to provide the receiver with explicit precoding
information.

20

AP
I

MIMO transmission can further be divided into multi-user and single-user MIMO (MU-MIMO and SUMIMO). The difference between the two is that in SU-MIMO all the streams carry data to/from the
same user while in the case of MU-MIMO the data for different users is multiplexed onto a single timefrequency resource. Hence, SU-MIMO is used either to increase the reliability of the channel (i.e.
diversity) or to increase the throughput to a single user in a multiplicative manner while MU-MIMO can
be seen as yet another way of multiplexing data to/from different users. In the case of, say, a 2x2
antenna MU-MIMO configuration, two mobile terminals can transmit/receive their data streams
simultaneously using the same physical radio resource.

AP
IS

TM1: single antenna transmission


TM2: transmit diversity

AP

IS

The R8 specification allows for 7 different Transmission Modes (TM) in the downlink. The TM for a given
UE is configured through RRC signalling. Hence, RRC message exchange is required in order to change
the TM for the UE. TM3-6 also include a transmit diversity option. This makes it possible to fallback to a
basic spatial diversity scheme if the (temporary) radio conditions do not allow for spatial multiplexing.
The R8 TMs are as follows:

TM3: Open loop spatial multiplexing


TM4: Closed loop spatial multiplexing with multiple layers
TM5: Multi-user MIMO (MU-MIMO)
TM6: beamforming, single layer, codebook based
TM7: beamforming, single layer, non-codebook based.
The open loop mode (TM3) means that spatial multiplexing (MIMO) is in use, but there is no need for
any feedback from the UE, and no need for the eNB to signal the selected precoding matrix to the UE. In
the 2-antenna case a fixed precoding matrix is used all the time. In the 4-antenna case the eNB will cycle
through a sequence of 4 known precoding matrixes.
Closed loop spatial multiplexing (TM4 and TM6) means that spatial multiplexing (MIMO) is in use and
that the eNB can, on a per-TTI basis, select precoding matrix/vector from a standardised codebook. This
requires that the eNB indicates the applied precoding to the UE for each downlink transmission. The UE
may also be configured to send uplink feedback information in the form of Precoding Matrix
Information (PMI) and Rank Indicator (RI). The PMI specifies what precoding matrix/vector the UE
would prefer for the next downlink transmission and the RI specifies how many transmission layers the
UE would prefer for the next transmission. The eNB may, or may not, use the reported values. TM6 is in
effect a subset of TM4, where less signalling overhead is needed in downlink and uplink (e.g. for low
SINR UEs, where multi-layer transmission would not apply anyway).

Beamforming (TM7) is a special case of spatial multiplexing, referred to as non-codebook-based


beamforming, where the eNB need not signal the particular beamforming pattern applied to the UE.
Instead, UE-specific reference signals are used in order for the UE to estimate the channel experienced
by the beamforming transmission.

21

AP
I

Additional transmission modes, marked blue in the figure, have been added after Release 8. These will
be described in more detail in this chapter (TM8 and TM9) and in the CoMP chapter (TM10).

AP
IS

IS

As mentioned before, the Physical Downlink Control Channel (PDCCH) carries Downlink Control
Information (DCI) messages for assignment of uplink and downlink resources. The exact parameter
content within the DCI message is different for different transmission modes, corresponding to different
DCI formats. The DCI formats defined for Release 8 are listed below.

AP

Format 0 is used for assigning uplink resources. Only contiguous PRBs can be allocated in UL.
Format 1 is used for DL assignments, allocating contiguous or non-contiguous sets of PRBs to the UE.
Format 1A serves the same purpose as format 1, but only supports contiguous PRBs .
Format 1B carries the same information as format 1A, with precoding info for beamforming added.
Format 1C is used for paging messages, random access responses and system information broadcast.
Format 1D is used for Multi-User MIMO (MU-MIMO or SDMA).
Format 2 is used like format 1, with the addition that it supports also spatial multiplexing (e.g. MIMO).
Format 2A is used for spatial multiplexing assignments (open-loop MIMO without feedback).
Format 3 is used for power control commands. It carries 1 bit of information (per addressed UE).
Format 3A is also used for power control, with the difference that it carries 2 bits per addressed UE.
DCI formats 2 and 2A are used for R8 MIMO operation. Besides the normal downlink assignment
parameters (PRB assignment, MCS etc) they also include an indication of what precoding matrix/vector
the eNB has applied to the data transmission. Additional DCI formats, marked in blue in the figure, have
been added for TM8-10. DCI format 2B was introduced in R9 to support TM8. It is very similar to format
2/2A but with the following exceptions:
it carries no Precoding Control Information (since non-codebook precoding is used)
it indicates the scrambling sequence applied to the downlink DM-RS associated with the data (more
about this later).
DCI format 2C was introduced in R10 to support TM9. It can be seen as a generalisation of format 2B
(just as TM9 is a generalisation of TM8). It is, just like format 2B, very similar to format 2/2A with the
following exceptions:
it carries no Precoding Control Information (since non-codebook precoding is used)
it indicates the number of transmission layers used for the downlink.

DCI format 2C also differs to format 2B in that it does not indicate the DM-RS scrambling sequence
applied, as format 2B does. The reason is simply that such scrambling sequences are not used in
Transmission Mode 9.

22

AP
I

DCI format 2D was introduced in R11 to support TM10. The difference between DCI 2D and DCI 2/2A is
described in the CoMP chapter.

AP
IS

Transmission Mode 8 (TM8) introduced in Release 9 allows for combining beamforming with spatial
multiplexing. As of R9 the number of spatial multiplexing layers is restricted to two per UE (dual-layer
transmission).

AP

IS

TM8 also allows for Multi-User MIMO operation (MU-MIMO) by assigning different layers to different
UEs, thus allowing 2 or more UEs to share the same time-frequency resource as seen in the figure
above. There are different ways of realising MU-MIMO, for example by making all involved UEs
demodulate all transmitted layers and then simply discard data on any layer(s) not indended for the
individual UE. Such a MU-MIMO method implies that each UE is aware of the transmission to the other
involved UEs, that the same resource is being assigned to all involved UEs and that the involved UEs
must use the set of receive antennas as required for the full multi-layer transmission.

23

AP
I

The MU-MIMO solution offered by TM8 is a bit more flexible than that. With TM8 a specific UE need not
be aware of transmission(s) to other UEs on the same resource and the resource assignment to
different UE may fully or only partly overlap each other. Further, it allows for multiplexing UEs with
different numbers of receive antennas. The TM8 MU-MIMO scheme can use single-layer transmission to
up to 4 UEs or dual-layer transmission to up to two UEs.

AP
IS

IS

Transmission Mode 8 is a so-called non-codebook based scheme (as opposed to codebook based). In a
codebook based scheme (such as TM 3-6 in LTE) the transmission layers are mapped onto antenna ports
by means of predefined precoding vectors/matrices that are known to both transmitter and receiver.
The set of predefined vectors/matrices are referred to as the codebook.

AP

In LTE, the codebook based transmission modes all rely on the Cell-specific Reference Signals (CRS) for
proper channel estimation at the receiver (i.e. the UE). As can be seen in the figure, the CRS are applied
after the precoding. The CRS will therefore reflect the channel state for each antenna port, as seen by
the UE, but excluding the effect of precoding. The eNB therefore needs to signal to the UE which
vector/matrix from the codebook it has selected for the transmission (e.g. I have applied the precoding
matrix with index 3 in the codebook. Similarly, the feedback sent back from the UE is restricted to the
entries in the specific codebook used (e.g. please use precoding matrix number 4 from now on).
In a non-codebook based scheme, such as TM8, there is no need to indicate the precoding index to the
UE. Or, more properly, it is not even possible to do so, since there is no codebook defined! The
precoding then becomes an implementation option (the inner workings of the precoder is not visible in
the LTE specifications). For this to work, the UE-specific Demodulation Reference Signals (DM-RS) have
been defined. The DM-RSs are applied before the precoder, as can be seen in the figure. Therefore, they
will reflect the channel from each antenna port including the effect of the precoder. The UE can then
demodulate the data on each transmission layer without explicitly knowing what precoding that was
applied by the eNB (the UE needs to know the number of layers used though).

24

AP
I

It should be noted that the standardised codebooks can be used also in TM8- for UE feedback signalling.
That is, the UE reports a preferred vector/matrix from one of the predefined codebooks. This does not
restrict the precoding that can be applied by the eNB, but may help it to make a good choice.

AP
IS

The UE-specific Demodulation Reference Signals (DM-RS) defined in Release 9 for TM8 are seen to the
right in figure above. For comparison, the R8 version of the DM-RS, defined for TM7, is seen to the left.

AP

IS

It was clear already in R9 that multi-antenna transmission in LTE would eventually be extended beyond
4 layers. It was also foreseen that extending the R8 DM-RS scheme to support 8 layers (or more) would
be difficult. Therefore it was decided to specify a completely new DM-RS scheme for TM8. The same
scheme has since then been extended in Release 10 to support up to 8 layers.
In R8 the Resource Elements (RE) used for Cell-specific Reference Signal (CRS) transmission on a given
antenna port corresponds to unused REs on all the other ports (as described earlier). Also, the CRSs are
always transmitted over the full system bandwidth in the downlink (ports 0-4).
The DM-RSs are, in case of dual layer transmission, transmitted from both antenna ports using the same
REs simultaneously (ports 7-8). This means that they will interfer with each other. To combat this
interference, mutually orthogonal Orthogonal Cover Codes (OCC) are applied to the two DM-RSs.

25

AP
I

Also, the DM-RSs are only transmitted within those Resource Blocks (RB) that are actually used for the
PDSCH transmission to UEs configured in Transmission Mode 8. This means that a Release 8 UE will not
be confused by the new RS-pattern, since such a UE will never be scheduled to receive RBs in TM8.

AP
IS

As of Release 9 there are two different Orthogonal Cover Codes (OCC) defined for TM8. The purpose of
the OCC is to suppress interference, so that the UE can properly demodulate the DM-RSs (DM-RS 0 and
DM-RS 1) on two transmission layers. This is seen in case a) in the figure.

AP

IS

In MU-MIMO scenarios for TM8 there may be 4 layers present. Case b) above shows 2 UEs, each with
dual layer transmission, and case c) shows 4 UEs, each with single layer transmission. In both cases, this
means 4 different DM-RSs that are colliding on the same REs. (The different colours, and numbering, of
the DM-RSs above are intended to show that they are UE-specific.)

26

AP
I

To cope with such scenarios a pseudo-random sequence (Pseudo) is applied , in addition to the OCC, to
the REs carrying DM-RS. One and the same sequence is applied to 2 layers and there are 2 sequences
defined. In total this gives us 4 different combinations of OCC/Pseudo, which is what is needed for cases
b) and c).

AP
IS

Transmission Mode 9 (TM9) introduced in Release 10 allows for combining beamforming with spatial
multiplexing and is thus an extension, or generalisation, of TM8 described earlier. The TM9 scheme
allows up to 8 transmission layers to a single UE.

27

AP
I

AP

IS

The MU-MIMO capabilities of TM9 are the same as for TM8. The reason that no more than 4 layers can
be used for MU-MIMO lies in the way the demodulation reference signals have been defined for TM9.

AP
IS

28

AP
I

AP

IS

The UE-specific Demodulation Reference Signals (DM-RS) for TM9 are a natural extension of the ones
introduced for TM8 in Release 9. To support up to 8 transmission layers, the same RE location for the
DM-RS is now used on 4 different antenna ports, each with different Orthogonal Cover Codes (OCC)
applied. The same concept is then repeated for the remaining 4 antenna ports by means of a shift in
the frequency domain. Thus, two frequency shifts combined with 4 different OCCs allow for 8 different
DM-RSs. (We can here see the reason why MU-MIMO with TM9 can only support up to 4 layers: there
are only four different OCCs available and no scrambling sequences are used.)

AP
IS

IS

In LTE R8/R9, the UE uses the Cell-specific Reference Signals (CRS) for two quite different purposes: as a
demodulation reference for the control and data region and as input to the channel estimation process.
The CRS pattern is therefore designed to cope with worst case scenarios in terms of radio channel
conditions. This requires a quite dense pattern in both time and frequency domain.

AP

In LTE R10, where up to 8 transmission layers are supported, the R8 CRS structure is not enough since it
only allows for maximum 4 antenna ports. Furthermore, in the multi-layer transmission modes (TM8
and TM9), the UE will use its DM-RSs for data and control region demodulation and not the CRS.
For these, and other, reasons a new set of reference signals have been introduced in R10: the Channel
State Information Reference Signals (CSI-RS). While DM-RS are UE-specific and carried in the resource
blocks allocated to that UE only, the CSI-RS are cell-specific and transmitted in the whole DL bandwidth.
The CSI-RS in LTE R10 has a very sparse reference signal density in both time and frequency domain. The
motivation for this is that LTE-Advanced mainly targets low mobility scenarios with typically high spatial
correlation, in which case a sparse grid is enough.

29

AP
I

A number of different time/frequency configurations for CSI-RS transmissions are available in each cell.
The table above shows some of the specified CSI-RS configuration index values. The index specifies
where the resource elements carrying CSI-RS are located in a given subframe. The CSI-RS itself is
mapped onto 2 resource elements consecutive in time (such as those marked with red colour in the
figure). In total, there are 20 possible locations for the DM-RS, as shown above. In the time domain, the
reference signals may be repeated with different periods, ranging from 5ms up to 80ms. Multiple CSI-RS
patterns may be used in a single cell.

AP
IS

The CSI-RS is mapped onto antenna ports 15-22 (or a subset thereof, depending on the number of CSIRS transmitted in the cell). The figure, taken from TS 36.211, shows the case when CSI-RS is transmitted
on 8 antenna ports, assuming CSI-RS configuration index 0 (see table on the previous page).

AP

IS

3GPP defines CSI-RS on two, four and eight ports. An interesting question is then how many LTE R8 CRS
that should simultaneously be supported in a real LTE R10 network deployment. Backwards compatible
peak bit rates for R8/R9 terminals can be preserved by going for four LTE R8 CRS, but that leads to a
very high total overhead of reference signals. Reducing the peak rates of LTE R8/R9 UEs by using only
one LTE R8 CRS, in order to support eight layer transmisson using CSI-RS, is not a very good option
either. The middle-ground would then be to support two LTE R8 CRS in conjunction with eight LTE R10
CSI-RS (since all R8 terminals currently on the market do no not support more than two ports anyway).

30

AP
I

Another question altogether is what combinations of LTE R8 CRS and LTE R10 CSI-RS to support when
less than eight CSI-RS antenna ports are configured. Such choices are not part of the standardisation
process but must be considered anyway, if we would like to see R10 multi-antenna techniques deployed
in an efficient manner.

AP
IS

The Release 8 version of the LTE specification only supports single-antenna transmission in the uplink,
regardless of the actual number of antennas available at the UE. With LTE R10, support for uplink MIMO
for PUSCH transmission and transmit diversity for PUCCH transmission is introduced.

AP

IS

The UE can transmit data on the PUSCH with up to four transmission layers (2 transport blocks) per
component carrier. No MIMO is defined for the PUCCH. However, the UE may be configured to use socalled Spatial Orthogonal-Resource Transmit Diversity (SORTD) for its PUCCH transmissions. SORTD
simply means that the UE transmits PUCCH from two antennas, each using different
time/frequency/code resources (i.e. each modulation symbol is transmitted from two antenna ports, on
two separate orthogonal resources).
The R10 uplink multi-antenna scheme defined for the PUSCH is based on codebook-based precoding.
The UE applies its configured Demodulation Reference Signals (DM-RS) on a per-antenna basis prior to
the precoding step. This would, in principle, allow for non-codebook based precoding just as for the
downlink (TM8 and TM9). However, the current decision in 3GPP is to let the eNB have full control of
the precoding matrix applied by the UE. This simplification may, perhaps, be removed in future releases.

31

AP
I

The transmission of Sounding Reference Signal (SRS) from the UE, for uplink channel state estimation by
the eNB, is slightly changed in R10 in order to support uplink multi-antenna transmission. First, the UE is
configured with one SRS for each of its antenna ports. Second, the eNB can request a single aperiodic
SRS transmission from the UE (as opposed to the R8 periodic SRS) by means of an indication on PDCCH.
These PDCCH-triggered SRS transmissions are referred to as SRS type 1 (periodic SRS are, from R10,
called SRS type 0).

AP
IS

AP

IS

A new Downlink Control Information format has been introduced to support uplink multi-antenna
transmission: DCI format 4. It specifies the precoding matrix to be applied by the UE and, if needed, the
transport format for the second transport block. DCI format 4 is only used when the UE is configured for
uplink Transmission Mode 2. (Note that the concept of uplink transmission mode does not exist in the
R8/R9 version of the LTE specification.)
DCI format 4 may also assign a multi-cluster allocation to the UE (see next page) and trigger aperiodic
Sounding RS (SRS) from the UE. The SRS is located in the last OFDM symbol in the second slot in the
subframe. The SRS occupies every second subcarrier over the configured SRS transmission bandwidth
(wideband or sub-band). No PUSCH transmission, from any UE, is allowed in the OFDM symbol where
SRS is transmitted. The subframes where SRS transmissions may occur, from any UE, is indicated in
system information.

32

AP
I

Up to three different aperiodic SRS configurations may be configured in the UE (the 2-bit PDCCH
request for SRS specifies which one the UE should transmit). For example: the first configuration may
trigger SRS in the upper half of the system bandwidth, the second configuration may trigger SRS in and
the lower half and the third may trigger a wideband SRS transmission.

AP
IS

There are some additional changes, besides MIMO, to LTE uplink operation introduced in Release 10:

IS

In R8/R9 the UE can only be assigned a contiguous set of Physical Resource Blocks (PRB) in the uplink. In
R10, the UE may be assigned non-contiguous PRBs in the uplink. This is called multi-cluster allocation
(multi being restricted to two in R10).

AP

In R8/R9 the UE transmits either PUSCH or PUCCH in an uplink subframe, never both. In R10 the UE can
be configured for simultaneous PUSCH & PUCCH transmission. This will be discussed more in the Carrier
Aggregation chapter.

33

AP
I

The figure also shows the simultaneous transmission of two orthogonal PUCCH resources, one per
antenna port, associated with the SORTD transmission of PUCCH discussed earlier.

AP
IS

34

AP
I

AP

IS

The table above summarises the antenna ports, and their related reference signals, defined in the LTE
specification as of Release 11. It may be prudent to, again, remind the reader that the concept of an
antenna port in the LTE specifications has nothing to do with any particular implementation of physical
antennas used by the base station and/or terminal. In the LTE specification, an antenna port is simply
defined as the transmission of a specific reference signal.

35

AP
I

IS

AP

AP
IS

36

AP
I

IS

AP

AP
IS

AP
IS

AP

IS

The first release of LTE provided support for deployment in spectrum allocations of various
characteristics, with transmission bandwidths ranging from 1.4 MHz up to 20 MHz for both FDD and
TDD. The peak data rate targets for LTE Advanced can only be fulfilled in a reasonable way with a
further increase of the transmission bandwidth. Furthermore, one of the IMT Advanced requirements
specifies that system bandwidths up to (at least) 40 MHz must be supported.
Therefore, the concept of Carrier Aggregation (CA) is central to the R10 update of the LTE standard.
With CA, multiple Component Carriers (CC) are aggregated and jointly used for transmission to/from a
single UE, as illustrated in the figure above. Up to five CCs can potentially be aggregated, allowing for
transmission bandwidths up to 100MHz (assuming aggregation of 20 MHz carriers, as in the example).
Backward compatibility is ensured as each component carrier conforms to the R8 carrier structure.
Consequently, to a R8/R9 UE, each component carrier will appear as an LTE R8 carrier, while a carrier
aggregation-capable R10+ terminal can exploit the total aggregated bandwidth.

37

AP
I

It should be noted that the migration from single to multi-carrier systems has previously been
introduced for HSPA already in R8 (Dual Carrier HSDPA).

AP
IS

In principle, any combination of uplink and downlink component carriers can be realised. There are
three cases in terms of the frequency location of the different component carriers, as shown above:
intra-band aggregation with contiguous carriers (upper left box)

AP

inter-band aggregation (lower right).

IS

intra-band aggregation with non-contiguous carriers (upper left)


Aggregating non-adjacent component carriers means that fragmented spectrum can be utilized, which
in turn allows operators to provide high data rate services even without a single wideband spectrum
allocation. We can also define different cases when it comes to the bandwidth of the component
carriers:
aggregation of component carriers with the same bandwidth
aggregation of component carriers with different bandwidth (lower left box).
It is reasonable to expect that a terminal will have different aggregation capabilities in the uplink and
downlink directions. Therefore we can define different cases when it comes to the symmetry of uplink
and downlink component carriers:
symmetric uplink and downlink aggregation, with same number of carriers and same bandwidth
asymmetric uplink and downlink, with different number of carriers and/or bandwidth (upper right).

38

AP
I

Note that carrier aggregation may be useful also for a total bandwidth below 20 MHz, for example as a
means to coordinate interference in a so-called heterogeneous network (discussed in a later chapter).

AP
IS

AP

IS

Carrier aggregation is, naturally, designed to operate in the frequency bands supported by LTE (see
table in chapter 1). Given the large number of supported bands, and the CA possibilities mentioned on
the previous page, it is easy to realise that to define all possible CA scenarios involving all defined
frequency bands would be a herculean task. It would also be a quite meaningless task, since the vast
majority of such CA scenarios would never be used in reality!

39

AP
I

It makes a lot more sense to have a market-driven approach when defining CA scenarios and, at the
same time, to be respectful of terminal implementation constraints. Therefore, only a few CA scenarios
have been selected for standardisation in Release 10, as listed above. The R10 CA scenarios are all
limited to 2 component carriers.

AP
IS

AP

IS

A specific Aggregated Transmission Bandwidth in combination with a specific (maximum) number of CCs
corresponds to a specific Carrier Bandwidth Class (A, B, C, etc) as listed above. The classes are defined
with the explicit intention of limiting the number of ways a given CA bandwidth can be constructed
from different numbers of CCs. It would be inefficient from a testing perspective (and overly complex
specification work) to allow total freedom in this matter. For example, it makes sense to exclude the
possibility to construct 50 MHz bandwidth with five 10 MHz CCs- instead some combination of three
CCs should be used. The notation in the lower table is as follows:
CA_X

Carrier Aggregation for band X, where X is the applicable LTE operating band

CA_X_Y

Carrier Aggregation for bands X and Y, where X and Y are the LTE operating bands.

The letter (A, B, etc) after the band indicator specifies the Carrier Bandwidth Class. For example,
CA_1A_5A above means carrier aggregation between LTE operating bands 1 and 5, each with 1 CC and a
combined maximum bandwidth of 100 resource block. Notation which do not have Channel Bandwidth
Class indicator letter means that all classes are included. For example CA_1 includes CA_1A, CA_1B,
CA_1C, CA_1D, CA_1E and CA_1F.

In the general case, the aggregated channel bandwidth is defined as the bandwidth in which a UE
transmits/receives multiple CCs simultaneously. It includes guard bands at the edges of the outermost
CCs as well as reserved subcarriers between CCs. The reserved subcarriers are not used for
transmission. In principle, those subcarriers could be used for data transmission to/from a CA-capable
LTE terminal, but the net gain in spectrum efficiency from defining those (fractional) resource blocks
was considered small in comparison to the additional specification and implementation complexity.
Aggregated Transmission Bandwidth Configuration is defined as the number of aggregated resource
blocks on which a UE can transmit/receive simultaneously. This corresponds to the sum of the number
of actually used resource blocks on each CC and the unused (fractional) resource blocks between CCs.

40

AP
I

The spacing between centre frequencies (DC subcarriers) of contiguously aggregated component
carriers is always a multiple of 300 kHz. This is to be compatible with the 100 kHz frequency raster of
LTE R8/R9 and at the same time to maintain the orthogonality of the OFDM subcarriers with 15 kHz
spacing. It is to fulfil the 300 kHz requirement that some reserved subcarriers need to be inserted
between two component carriers.

AP
IS

AP

IS

Each downlink Component Carrier (CC) transmitted from the eNB is seen as a separate cell from the UE.
Therefore, all the downlink synchronisation signals and control channels (system information, paging,
scheduling etc) must be present on each CC. System information is always cell-specific, meaning that
the broadcasted system information on a given CC is specific to that CC. The geographical coverage of a
given CC is not necessarily the same as the coverage of another CC transmitted from the same base
station (irrespective if they are on the same frequency band or not).
An idle mode UE always camps on one CC/cell at a time, regardless if it is CA-capable or not. Thus, the
CA concept is transparent to idle mode terminals. Upon transition from idle to connected mode, the CAcapable UE will be configured with one or more Serving Cell. There will always be a Primary Serving Cell
(PCell) configured. There may also be zero, one or more Secondary Serving Cell (SCell) configured. A
specific Serving Cell for the UE is thus either a PCell or an SCell. Different UEs may have different PCells.
The PCell always has both a downlink component and an uplink component. These are called the DL
Primary Component Carrier (DL PCC) and the UL Primary Component Carrier (UL PCC) respectively. An
SCell will always have a downlink component (DL SCC) and may have an uplink component (UL SCC).

41

AP
I

Each configured Serving Cell, for a specific UE, is assigned a Serving Cell Index, where the PCell always
have the index 0 and an SCell has index 1 or higher.

AP
IS

AP

IS

The upper half of the figure shows the standard (R8 or later) format of a MAC Protocol Data Unit (or
transport block). In general, a MAC PDU consists of a header part and a payload part. The payload part
carries higher layer PDUs from one or more logical channel and, when needed, MAC layer signalling in
the form of so-called Control Elements (CE). The header part is divided into sub-headers, with one subheader for each segment carried in the payload part. Each sub-header indicates the content of the
corresponding payload segment by means of the Logical Channel ID field (LCID). In the example, subheader 1 indicates the presence of the new (R10) activation/deactivation CE.
To reduce terminal power consumption, a CA-capable terminal as baseline receives on one component
carrier, the PCell. Reception on additional SCells can be rapidly turned on or off in the UE by the base
station through MAC signaling. Similarly, in the uplink, feedback signaling is transmitted on the PCell
and SCells are enabled when necessary for data transmission.
The set of serving cells (PCell and SCells) is configured by means of RRC signalling. Any configured SCells
always start in the deactivated state. The UE performs no DL or UL activity on/for a deactivated SCell.
One or more SCell can then be activated by means of transmitting the activation/deactivation CE, as
part of a downlink transport block, to the UE. The CE consists of a bitmap, where the value 1 means
activation of the SCell with Serving Cell Index corresponding to the bit-position. Subsequent
deactivation of SCell(s) can be done with the same MAC CE, by setting the bit(s) in question to 0.
Deactivation may also be timer-based. The UE is configured with an SCell Deactivation Timer that runs
per configured SCell (range 20ms-1280ms). The same value applies for each SCell although the
associated functionality is performed independently for each SCell. The timer is reset to zero whenever
a downlink assignment or uplink grant associated with the SCell is received by the UE. The SCell is
deactivated when the the timer expires (i.e. due to inactivity of the SCell).

42

AP
I

The Release 8/9 Power Headroom Reporting (PHR) procedure is used to provide the eNB with
information about the difference between the UE's current transmit power and the maximum possible
transmit power. The difference is reported as a value in the PHR uplink MAC control element. The PHR
is sent periodically and is also triggered when the calculated pathloss has changed more than
'DL_PathlossChange' dB since last report. The eNB uses the PHR to select a combination of uplink
modulation & coding scheme (MCS) and resource size that does not lead to the UE becoming power
limited. The same mechanism is used for carrier aggregation, with the addition that power control is
operated on a per-carrier basis and hence the PHR should be for each component carrier. A new uplink
Extended Power Headroom Report (EPHR) control element was introduced in R10 for this purpose, with
which the UE reports the PCell and each active SCell separately in one and the same EPHR message.
(The EPHR control element is not shown in the figure.)

AP
IS

AP

IS

Carrier aggregation/de-aggregation takes place within the Medium Access Control (MAC) layer. Thus,
the higher layers are not aware of carrier aggregation. Hybrid-ARQ operation is done independently for
each component carrier. Once constructed, a transport block is associated with one, and only one,
component carrier. For example, a transport block initially transmitted on CC 2 cannot be retransmitted
on CC 1. All physical layer processing is also done independently for each component carrier. As a result,
all active CCs for a given UE may use different transport block sizes, coding rates, modulation and multiantenna transmission schemes.
As a baseline, scheduling assignments are transmitted on the same component carrier as the
corresponding downlink/uplink data (PDCCH option 1 in the figure). However, as an alternative, it is
possible to use cross-carrier scheduling where the scheduling information is transmitted to the UE on a
different component carrier to the corresponding data transmission (PDCCH option 2 in the figure).
Downlink ACK/NACK in response to uplink data transmission is always transmitted on the downlink
carrier where the uplink grant was given (and nowhere else). This means that the UE receives PHICH on
the same carrier as the one used for uplink transmission in case of PDCCH option 1. In case of PDCCH
option 2 this could potentionally be a different carrier than the one used for uplink transmission (crosscarrier scheduling).

43

AP
I

Uplink PUCCH transmissions (ACK/NACK and other feedback) is only allowed on the carrier
corresponding to the PCell. A new PUCCH format has been introduced to accommodate up to 10
parallel ACK/NACKs in response to downlink data transmission.

AP
IS

AP

IS

For any reception of any scheduled downlink (or uplink) transmission the UE needs to decode the
Physical Downlink Control Channel (PDCCH). As mentioned in chapter 1, the PDCCH is mapped onto a
per-subframe variable-size Control Region at the beginning of each subframe. The size (in units of OFDM
symbols) is indicated on a per-subframe basis with the Physical Control Format Indicator Channel
(PCFICH). The PCFICH itself is always mapped onto the first OFDM symbol in each subframe.
Two scheduling approaches can be used for CA: per-carrier scheduling and cross-carrier scheduling. It is
also possible to use a mix of the two, i.e. some component carriers are cross-carrier scheduled and
some are not. The scheduling mechanism to use is configured by RRC on a per-UE and per-carrier basis.

44

AP
I

Per-carrier scheduling (shown above) means that the R8/R9 mechanisms are reused, but now for each
active component carrier. The UE decodes PCFICH and the PDCCH(s) in the control region on all
currently active CCs (the control regions on different CCs can have different sizes in the same
subframe). A downlink assignment is only valid on the CC where it was received. Similarly, uplink grants
are only valid on the uplink carrier associated with the downlink carrier where the grant was received.
The only CA-specific rules introduced is that system information and paging messages can only be
scheduled on the PCell. If semi-persistent scheduling (SPS) is used, such assignment are also restricted
to the PCell. A downlink or uplink assignment on the PDCCH with per-carrier scheduling will consists of
the same information as in R8/R9 (UE identity, PRB allocation, MCS, HARQ-info etc).

AP
IS

AP

IS

Cross-carrier scheduling (shown above) implies that the PDCCH on one component carrier can be used
to allocate resources on another component carrier for downlink/uplink data transmission. This requires
an additional parameter on the PDCCH: the Carrier Indicator Field (CIF). The CIF is a 3-bit parameter
where the values correspond one-to-one to the Serving Cell Index (0-4) configured for the UE. Hence,
the meaning of a specific CIF-value is UE-specific since different UEs can have different PCells (index 0).
A given CC can only be scheduled from one CC, either itself or another CC. The who-schedules-who
mapping is configured on a per-UE and per-CC basis through RRC signalling. In the example in the figure,
Serving Cell 0 (the UEs PCell) schedules itself and SCells 1 and 2, while Serving Cell 3 schedules itself and
SCell 4. Thus, cells 0 and 3 are not cross-carrier scheduled while cells 1, 2 and 4 are. (This is what was
meant by ...mix of the two... on the previous page.)

45

AP
I

The UE does not need to decode the control region on any cross-carrier scheduled CC. This means that
the UE will not decode PCFICH and will not know where the data region in a given downlink subframe
begins on such CCs. For cross-scheduled CCs, the start of the data region is therefore configured, for the
UE, to be at a fixed position (PDSCH start configured by RRC). Assignments for system information,
paging and SPS resources are restricted to the PCell, as in the per-carrier scheduling case.

AP
IS

There are several PUCCH formats defined for carrying Uplink Control Information (UCI) in Release 8:

AP

IS

Format 1 is used for the Scheduling Request.


Format 1a: ACK/NACKs in response to a DL transmission of one transport block (may also carry SR).
Format 1b: ACK/NACKs in response to a DL transmission of two transport blocks (may also carry SR).
Format 2: carries channel state information (CQI and, possibly PMI and RI).
Format 2a: channel state information and ACK/NACK for one transport block.
Format 2b: channel state information and ACK/NACK for two transport blocks.
A new PUCCH format has been introduced especially for carrier aggregation: PUCCH format 3, which
supports HARQ-ACK/NACK for up to 10 transport blocks (TB). This is needed when 5 Component
Carriers (CC) are used, all with MIMO transmission in the downlink (2 TB per CC).
However, LTE R10/R11 is restricted to max 2 CCs, where each can be configured for MIMO operation.
This means the UE can potentially receive up to 4 transport blocks in the same TTI, resulting in 4
ACK/NACKs needed in the uplink. The R8 PUCCH format 1b supports only 2 HARQ-ACKs but can, by
introducing a R10 trick, actually be used to transmit 4 HARQ-ACKs. This is convenient since PUCCH 1b
consumes less network resources than PUCCH 3 and can also be used for simultaneous CSI reporting.
The trick lies in the fact that the eNB can configure up to 4 allowed PUCCH resources for the UE (in a
given subframe the UE will only use one PUCCH resource). When 4-bit HARQ-ACK is needed, the PUCCH
itself will carry only 2 bits (as in R8). The remaining, not transmitted, 2 HARQ-ACK bits are then used to
select one of the four configured PUCCH resources. The resource used to transmit the 2 actual bits will
thus implicitly carry 2 bits of information. This mechanism is referred to as PUCCH format 1b with
resource selection.

46

AP
I

Whether format 3 or format 1b with resource selection shall be used by the UE is configured by means
of RRC signalling.

AP
IS

0 = No CSI report requested

AP

IS

Some CA-related changes have been introduced for reporting of Channel State Information (CSI).
Aperiodic CSI reports (requested by the eNB when giving uplink grant) are, as in R8, always sent on the
PUSCH by using some of the assigned resource elements for UCI instead of for data. The CSI request
field in the uplink grant has in R10 been extended from one bit to two bits. The values of the new 2-bit
format has the following meaning:
1 = CSI report for this carrier requested (this being the carrier where the PDCCH was sent)
2 = CSI report for configured carrier-set 1 requested
3 = CSI report for configured carrier-set 2 requested
For example: this carrier may be the PCell, carrier-set 1 may be SCell 1 and carrier-set 2 may be both
PCell and SCell 1. The carrier-sets are configured by means of RRC signalling.
Some additional CA-specific rules also apply for simultaneous periodic CSI report and HARQ-ACK (sent
on PUCCH 2a or 2b in R8/R9). The additional rules for transmission of Uplink Control Information (UCI)
on PUCCH/PUSCH are summarised below:
PUCCH only: if PUCCH 3 is configured, there is no room for the CSI bits and the report is dropped
PUSCH: UCI shall be sent on the PUSCH with lowest Serving Cell Index (normally the PCell)

47

AP
I

Simultaneous PUSCH/PUCCH: CSI bits sent on lowest index PUSCH, HARQ-ACK on PUCCH.

AP
IS

AP

IS

In the R8/R9 specification of LTE there are 5 different UE categories defined. These have been extended
with 3 additional UE categories in R10. The table gives, for each category, the maximum number of
transport channel bits the UE can receive or transmit within a Transmission Time Interval (TTI). Given
the 1ms TTI length selected for LTE it becomes straight-forward to calculate the maximum possible
(layer 2) data rates from the table.
For a category 8 UE the max downlink bitrate is (roughly) 3000 Mb/s and the maximum uplink bitrate is
1500 Mb/s. These peak bitrates require CA operation with 5x20 MHz combined with 8-layer or 4-layer
MIMO (DL and UL, respectively). It should be noted that category 8 was defined as an extreme case to
show the full potential of LTE-Advanced (as of R11 there can never be more than 2 component carriers).
A category 6 or 7 UE shall also, as part of its UE Radio Access Capabilities, indicate support for R8/R9
category 4. In the same way, a category 8 terminal shall indicate also support for R8/R9 category 5. This
is needed since the network may not be upgraded to R10 and will thus only recognise categories 1-5.
Although not explicitly stated, category 6 and 7 were defined with 2-carrier CA capable UEs in mind.
Work is underway, within the scope of Release 12, to define CA with 3 component carriers. The final
R12 standard will therefore, almost certainly, see the introduction of yet another UE category.
Besides the UE category, the R10+ UE will indicate its supported band combinations (capabilities for
carrier aggregation, including the Carrier Aggregation Bandwidth Class) and UL/DL MIMO capabilites per
band and band combination. UE capability information from R10 also indicate support, or not, for the
following features:
TM 9 with 8 CSI-RS ports
PUCCH transmit diversity
Cross-carrier scheduling
Simultaneous PUSCH/PUCCH

48

AP
I

Multi-cluster allocation on PUSCH.

AP
IS

IS

As mentioned earlier, the CA combinations for LTE will be standardised based on the actual need for
them in various markets. A number of additional inter- and intra-band CA scenarios for both FDD and
TDD where therefore added as part of R11. More CA combinations will be added in each release- please
consult the references given at the end of this chapter for the latest developments.

49

AP
I

AP

In general, spectrum allocation is dependent on decisions taken by ITU-R as well as by regional/national


entities. The agreed approach in 3GPP is therefore to await the outcome of spectrum auctions and
regional/national spectrum allocation before deciding which specific CA scenarios that are needed in a
given release.

50

AP
I

IS

AP

AP
IS

51

AP
I

IS

AP

AP
IS

AP
IS

IS

The concept of relaying has been used in various mobile networks for a long time, typically through
some form of repeater. A repeater can be called an amplify-and-forward type of relay since it will
amplify and forward everything it receives- including noise and interference. Thus, it can be implied that
repeaters are best used under good signal to noise conditions.

AP

The relay architecture defined for LTE R10 makes use of so-called decode-and-forward relays. That is,
the relay will decode the received signal, including processing for active bit error correction, and then
re-encode the signal prior to forwarding. Thus, any noise and interference present will not be amplified
and forwarded, making this type of relay suitable also for low signal to noise radio conditions- such as at
the cell edge.
The LTE relays are self-backhauling in the sense that they communicate wirelessly with a Donor eNB
(DeNB) by means of the standard LTE radio interface and do not require any wireline connection. The
link between the DeNB and the Relay Node (RN) is called the backhaul-link (Un-interface), as opposed to
the access link (Uu-interface) between the RN and the UE. The RN controls its own cell, that in the
downlink carries normal R8 synchronisation signals, reference signals etcetera. The LTE relay concept is
therefore completely transparent to the UE and fully backwards compatible to R8/R9 terminals.
The S1- and X2-interfaces are terminated in the RN, with the DeNB acting as a proxy between the RN
and the core network (or eNB).

52

AP
I

The RN can be an outdoor relay, very similar to a regular macro eNB, offering improved coverage in
difficult spots such where the macro eNB is shadowed by buildings. It can also be deployed as an indoor
relay, offering better indoor coverage. An indoor relay may have one set of antennas inside, for
communicating with the UEs, and another set outside the building, for communicating with the DeNB.
Such a relay can be referred to as a through-wall relay.

AP
IS

AP

IS

The Relay Node (RN) acts as both eNB and UE. It is seen as an eNB from the UEs perspective and as a UE
from the DeNBs perspective. Upon power on, the RN will perform the Attach procedure just as a regular
UE does. Thus, the RN will be registered in an MME and will have user plane bearers established. The
RN then uses the user plane to download configuration parameter from an O&M system (the IP-address
of the O&M server may be preconfigured in the RN). Part of the configuration parameters is a list of
available Donor eNBs.
The RN selects a proper DeNB and connects to it. The RN then needs to perform the S1 Setup (and X2
Setup) procedure towards the DeNB. This is done by means of standard S1AP signalling, except that the
signalling messages are now sent wirerlessly over an LTE Data Radio Bearer.

53

AP
I

The Donor eNB takes the role of Serving Gateway (SGW) and Packet Data Network Gateway (PGW)
towards the RN. Thus, the DeNB needs to support certain SGW/PGW functionality needed for relay
operation. All downlink IP-packets that shall be forwarded to the RN will enter the DeNB through its
PGW function. The packets then need to be mapped onto the proper radio interface user plane bearer.

AP
IS

With the relay architecture, the S1-interface and X2-interface protocol stacks are stretched out over
the LTE radio interface and terminated in the Relay Node. The figure shows only the S1-interface stacks
for the control plane, but the same concept applies also to the X2-interface.

AP

IS

The IP packets for the S1 control plane (S1AP) are sent over the Un-interface encapsulated by the
normal radio interface user plane protocol stack (PDCP etc), that is over normal LTE Data Radio Bearers.
The difference is that the PDPC protocol will now provide integrity protection also for the Data Radio
Bearers (which it does not do for normal LTE operation). A new security key has been introduced for
this purpose: the Integrity Key for the User Plane (IKUP).

54

AP
I

The S1 and X2 proxy functionality of the DeNB includes modifying IP-addresses, Tunnel Endpoint
Identifiers (TEID) and S1AP/X2AP UE identities. No other S1/X2 message contents is changed.

AP
IS

55

AP
I

AP

IS

S1-interface and X2-interface proxying also applies to the user plane (again, only the S1-interface stacks
are shown). In this case, the GTP-U/UDP/IP packets received from the core network are encapsulated by
the normal radio interface user plane protocol stack and sent over normal LTE Data Radio Bearers.
Integrity protection for these Data Radio Bearers may be used as an option.

AP
IS

AP

IS

For each UE served by the RN there will be one S1 bearer (GTP tunnel) for each active EPS bearer, just
as in Release 8. The DeNB will then multiplex one or more UE-specific S1 bearer onto a radio bearer
towards the RN. UE bearers with similar QoS will be mapped onto the same radio bearer. The exact
meaning of similar QoS is not standardised, but it would make sense to use the QoS Class Identifier
value (QCI) for this purpose. The eNB may also base the mapping on IP-level DiffServ Codepoints as
received from the SGW within each packet. The RN then de-multiplexes bearers within a given radio
bearer based on the received TEID.

56

AP
I

In the uplink direction the RN will do reverse mapping, from QCI associated with a UE Data Radio Bearer
to DiffServ Code Points for the S1-U interface.

AP
IS

AP

IS

The backhaul link between Relay Node and DeNB may be outband or inband as compared to the
access link. In the outband case, the backhaul link and the access link operate in different frequency
bands and will not interfere with each other. For inband relaying, where the two links operate on the
same frequency, the DeNB-to-RN downlink will interfere with the RN-to-UE downlink (the same applies
to the uplink). Please note that downlink for the RN means reception on the backhaul link and
transmission on the access link (and vice versa for the uplink). Thus, the RN will interfere with its own
reception when it transmits on the downlink/uplink.
To avoid collisions between UE uplink and RN uplink, the RN can simply make sure not to schedule any
UEs in the uplink when it needs to transmit to the DeNB. This can be done since it is perfectly allowed to
have zero transmission in the LTE uplink. The access downlink is a different matter though. The UEs in
the cell of the RN will expect, at least, CRS and sync signal transmission in each downlink subframe. It is
therefore not allowed to have completely blank subframes in the downlink. This means there will
always be collisions between the DeNB-to-RN downlink transmission and the RN-to-UE downlink.
The downlink problem for inband relaying is handled by reusing the Release 8 MBSFN subframes. These
subframes only contain a control region occupying one or two OFDM symbols. The rest of the subframe
is completely empty, with not even CRS being transmitted. A repeating pattern of MBSFN subframes are
configured in the cell of the RN. The UEs will be aware of this from system information.
The MBSFN subframes will be repeated every 8 subframes. This matches the HARQ round-trip time
defined for LTE. The RN will then receive its downlink from the DeNB during the MBSFN subframes,
when there is (almost) no transmission on the access downlink. This naturally leads to a 4ms period
between RN reception and RN transmission from/to the DeNB, since there will be no uplink grants sent
to the UEs during the MBSFN subframes (an LTE UE uses the uplink 4ms after receiving its UL grant).
The only complication is that the 8ms pattern is not aligned with the LTE 10ms radio frames. Some
subframes in the 10ms radio frame are not allowed to be MBSFN subframes (such as those carrying sync
signals). This means that some of the per-8ms MBSFN subframes configired cannot be used. This is
shown in the figure. As a result, the HARQ round-trip time between RN and DeNB will sometimes be
16ms instead of the normal 8ms.

57

AP
I

Please note that the UE decodes the control region (1 or 2 symbols) even in MBSFN subframes. This
allows the RN to provide uplink grants to the UEs in its cell (but not downlink assignments, as that would
mean downlink data in the MBSFN region in the subframe).

AP
IS

AP

IS

The figure shows the contents on the backhaul downlink subframe coinciding with an access link MBSFN
subframe. The backhaul link cannot make use of the full subframe, since one or two OFDM-symbols are
actually used by the MBSFN frame. Furthermore, it may take some time for the RN to switch from
transmission to reception and vice-versa. The RN is therefore explicitly configured with the first and last
usable OFDM-symbol in the backhaul downlink. This is shown in the two tables to the right in the figure.
The PDCCH mapping for the RN cannot follow the same structure as a normal subframe, again because
the first fw OFDM-symbols are actually used by the MBSFN subframe. The PDCCH is instead mapped
onto the symbols where the MBSFN subframe is empty (possibly with a guard symbols used as defined
by the tables above).
This new mapping is referred to as the Relay PDCCH, R-PDCCH. Contentwise there is no difference
between the R-PDCCH and regular PDCCHs. The difference is that the R-PDCCH is mapped onto the
OFDM-symbols in the normal data region instead of onto the normal control region. The standard
downlink transmission modes also apply, meaning that R-PDCCH and/or PDSCH transmission to the RN
may be performed using normal CRS or UE-specific DM-RS. The resource blocks used by the R-PDCCH is
configured in the RN by means of RRC signalling.

58

AP
I

When there are R-PDCCH transmissions to more than one RN present, the different channels may be
mapped independently to their own resource blocks (red region in the figure) or they may be crossinterleaved with each other for better frequency domain diversity (as shown for relay node 2 and 3).

AP
IS

The DeNB to RN subframe differs in a number of aspects from the normal eNB to UE subframe:
The control region for R-PDCCH is mapped differently to the PDCCH, as already described.

IS

The PDSCH is mapped to the same OFDM symbols as the R-PDCCH (not the same subcarriers though).

AP

If no uplink grant is detected on the R-PDCCH (no control in the second slot) the PDSCH may occupy
the resource blocks in the second slot that would otherwise have been used by the R-PDCCH.
There is no normal control region in the first few OFDM symbols. This means that there is no PCFICH
present. Hence the DeNB needs to configure the PDSCHstart parameter in the RN (in the same way as
for cross-carrier scheduling for carrier aggregation). The absence of the control region also means that
there is no PHICH present. The RN will always assume positive ACK for each PUSCH transmission, but
will not flush the HARQ buffer. If the DeNB detects the need for a retransmission, this will be triggered
by an R-PDCCH grant (i.e. uplink retransmission are always scheduled).

59

AP
I

The R-PDCCH (and hence also PDSCH) may be configured not to include the last OFDM symbol in the
second slot (configuration 1 for slot 1 in the table as shown in the figure). In such a case it becomes
impossible to transmit all the UE-specific (i.e. RN-specific) demodulation reference signals (DM-RS) for
transmission modes 8 & 9. The seleted solution is to simply skip the DM-RS in the second slot and only
transmit them in the first.

AP
IS

The figure summarizes the paramaters that are configured in the RN by means of RRC signalling from
the DeNB. All procedures are applicable during the second phase of RN start-up (as described on the
next two pages).

AP

IS

Configuration of Signalling Radio Bearers (SRB) and Data Radio Bearers (DRB) is done in the same way,
using the same messages and parameters, as for a normal UE. The SRB is configured during the initial
RRC Connection Establishment procedure and one or more DRB is configured during the RRC Connection
Reconfiguration procedure (after activation of radio interface security).
The third procedure, RN Reconfiguration, is RN specific (as its name suggests...). The main purpose of
this procedure is to configure the MBSFN subframe pattern to be used between the DeNB and the RN,
and to configure the R-PDCCH. The RN indicates, to the DeNB, in the RRC Connection Establishment
procedure if it needs a subframe configuration or not (it may be pre-configured already).

60

AP
I

For further details, please consult TS 36.331 (E-UTRA RRC protocol).

AP
IS

Start-up of the Relay Node (RN) is done in the phases. Phase 1 is described here and Phase 2 on the next
page. Please note that the signalling flows are highly simplified, many messages and most parameters
have been completely left out.

AP

IS

In Phase 1 the RN performs a normal Attach just like any UE would do. The main difference is that the
HSS provides a RN-specific APN, that allows the RN to contact the Operation and Maintenance server
(OAM) but nothing else. The MME will select SGW and PGW (jointly named GW in the figure) based on
the RN-APN received from the HSS.
Normal Attach procedure signalling then follows, including activation of NAS security and creation of a
default bearer towards to gateways (Create Session in the figure). The PGW will allocate the UE IPaddress and pass it back to the MME. (This IP-address will really be used by the RN and not by any UE.)
After successful attach and activation of a default bearer, the RN will contact the OAM server and
download necessary configuration parameters. At the very least, the RN will download a list of available
DeNB cells in the area where it is located. The signalling between the RN and the OAM server takes
place through the EPS User Plane, on the default bearer.

After sucessfully downloading the configuration parameters, the RN initates the Detach procedure and
becomes de-registered in the network.

61

AP
I

This phase does not require any new functionality in any node (except the RN itself of course).

AP
IS

IS

In Phase 2 the RN starts by selecting one DeNB cell from the list retrieved in phase 1. The RN then
includes a specific indication to the DeNB in the RRC Connection Setup Complete message, making the
DeNB aware that it is dealing with a RN. Please note that the DeNB is not necessarily the same base
station as the eNB involved in phase 1.

AP

The RN then starts the Attach procedure in the normal manner. The DeNB will select an MME that it
knows supports the relay feature and adds the RN indication and the GW address to the uplink Attach
Request message. The GW is the co-located SGW/PGW functionality in the DeNB.
The HSS provides an indication that this specific IMSI number corresponds to a valid RN. After activation
of NAs security the MME then selects SGW/PGW based on the received GW address. The signalling to
activate the default bearer is then initiated towards the GW (i.e. the DeNB). The PGW functionality in
the DeNB allocates the RN IP-address.
After successful completion of the Attach procedure, the RN contacts the Operation and Maintenance
server (OAM) again, if needed. This step may, for example, include downloading a list of cell identities
for the RNs own cells.

Next follows establishment of the S1-interface between RN and DeNB. The DeNB updates the MME of
the new cells that now exist. The RN then initiates establishment of the X2-interface between RN and
DeNB. If applicable, the DeNB updates neighbouring eNBs of the existence of new cells.

62

AP
I

The last step is the configuration of Un-interface parameters such as the MBSFN subframe configuration
and the configuration of the R-PDCCH (as described earlier).

63

AP
I

IS

AP

AP
IS

64

AP
I

IS

AP

AP
IS

AP
IS

IS

Even though LTE, in theory, can be deployed as a network with GSM-like frequency reuse between
cells to eliminate inter-cell interference, the reality is that the same carrier frequency is used in all cells
(no network operator will have enough spectrum for the reuse solution). This obviously causes inter-cell
interference close to cell edges.

AP

A simple solution to this is to coordinate the resources scheduled for cell-edge users between cells. This
can be done in a semi-static way, by configuring a given cell to use only a subset of the total system
bandwidth when scheduling such UEs. Different cell-edge subsets are then configured for neighbouring
cells, thus eliminating inter-cell interference. This is what is shown in the figure.
Inter-Cell Interference Coordination (ICIC) may also be performed in a more dynamic way by making use
of the X2-interface. One eNB can inform another eNB about which uplink and downlink resource blocks
it is about to schedule for its cell-edge UEs. Similarly, an eNB may complain to another eNB about the
uplink interference situation. The result is really the same as the one shown in the figure, except that
the exact set of cell-edge resources will vary dynamically over time. Signalling support for this is part of
the specification already from Release 8. (see the X2AP slide a few pages hence for more information.)

A drawback of the R8 ICIC solution is that it only affects data transmission or, more precisely, the data
region within each subframe. The downlink control channels or, more precisely, the control region in
each subframe must be decodable by all UEs- including those at the cells edge. A large effort was put
into the LTE standardisation process to avoid collision of control channel transmissions at the cell edge.
For example, the exact OFDM resource elements a given control channel, such as PDCCH, is mapped
onto in a given subframe is dependent on the physical cell identity (which always differ between any
two neighbouring cells). This control channel resource randomization is not enough to completely
eliminate interference between cells though.
A second drawback is that the R8 ICIC is really only relevant at the cell-edge. This implies that
neighbouring cells are exactly that- neighbours. With hierarchical cell structures (cells on top of each
other) the R8 solutions is less useful or even useless.

65

AP
I

A third drawback is the fact that cell-edge UEs will never be able to use the full system bandwidth.

AP
IS

AP

IS

Current mobile networks are typically deployed as homogeneous networks, with a macro cell centric
planning process. In general terms, a homogeneous system is a network of base stations in a planned
layout in which all the base stations have similar transmit power and capacity (they serve roughly the
same number of terminals). The locations of the macro base stations are carefully chosen through
network planning, and the cell settings are properly configured to maximize the coverage and control
the interference between cells. As the traffic grows the network relies on densifying the network by
adding more cell sites (and/or additional carriers). This is a complex and expensive process. A pure cellsplitting approach will also surely lead to some cells being under-utilized. Yet another problem is the
higher number of handovers that will need to be performed for non-stationary users (causing an
increased overall signalling load in the network).

66

AP
I

A potentially more attractive approach is to complement a macro-cell layer, providing basic coverage,
with additional low output-power pico cells where needed. The result of such a strategy is a
heterogeneous deployment with two or more cell layers. The idea of multiple cell layers is in itself not
new- it has been used in GSM/UMTS (and other) networks for well over a decade. Thus, heterogeneous
networks is a deployment strategy and not a technology. As such, heterogeneous deployments are
very much possible with LTE R8. However, R10 provides explicit features that support an efficient usage
of this type of network deployment.

AP
IS

IS

In a heterogeneous network environment, low-power base stations are deployed to eliminate coverage
holes in the macro system and to improve capacity in hot spots. There are many possibilities when it
comes to the exact type of low-power node to deploy. The figure shows an example where pico cells,
femto cells (home base stations), remote radio heads and relay nodes are all deployed.

AP

While the placement of macro base stations in a cellular network is generally based on network
planning, the placement of pico-type base stations may be more or less ad hoc, based on just a rough
knowledge of coverage and traffic density. Due to their lower transmit power and smaller physical size,
pico/femto/relay/etc base stations can offer flexible (read: cheap and hazzle-free) site acquisitions.
Because of the (more or less) unplanned deployment, most networks with low power base stations can
be expected to have areas with low signal-to-interference conditions, resulting in a challenging radio
environment for control channel transmissions to UEs on the cell edge of low-power nodes.

67

AP
I

The R10 features relating to heterogeneous network deployments are primarily aimed at combating
these new interference situations that may/will arise in the network.

AP
IS

AP

IS

Due to the large difference in output power between macro and pico base stations, there is always a
risk that the pico cell will be flooded by the macro cell. Any UE that uses downlink signal strength as
the major criterion for cell selection will most often lock on to the macro cell. In this scenario, the
larger coverage of macro base station will attract most UEs and the pico station will be under-utilized
(while the macro station may become overloaded). If cell selection is based on downlink signal strength,
as it is in LTE R8, the usefulness of pico base stations will thus be greatly diminished.
We must also consider the uplink transmissions. The strength of the signal received by a base station
depends on the UEs transmit power, which does not change with the type of base station serving it.
Hence, the uplink coverage of all base station types is similar. As an example, the UE in the pico cell in
the upper right of the figure may select the macro cell as soon as the downlink signal from the macro
cell is better than that from the pico cell. However, from that location its output power may not be
enough for proper communication with the macro base station. For this UE, the macro cell offers the
best downlink at the same time as the pico cell offers the best uplink!
Thus, mixing macro and pico cells can create mismatches between downlink and uplink coverage and
makes the base station-to-UE association less clear, as compared to homogenous networks where
downlink and uplink coverage is more closely matched.
A possible remedy for this is to use the uplink pathloss, as opposed to downlink signal strength, as the
main criterion for base station-to-UE association. This will cause the UE to stay with the pico cell longer,
thus balancing the load between macro and pico base stations. In effect, the coverage of the pico base
station has been extended in such a scenario. This concept is, logically, called range extension.
However, then we are back with the problem of interference to downlink control signalling at the celledge. Any UE in the extended range area of a pico cell will be the victim of interference from the macro
cells high-power downlink transmissions. Note that failure to decode the downlink control information
in LTE will affect both downlink and uplink data transmission.

68

AP
I

The problem is even worse when it comes to femto cells. Any macro cell UE that is close to the femto
cell (but not allowed to use it) risks suffering from downlink control channel outage due to the downlink
transmission from the femto base station. At the same time, the femto base station will be severly
interfered by the UEs uplink transmission. Thus, the so-called aggressor-victim relation goes both-ways
in this case. From the macro base stations point of view, femto cells can be seen as punching holes in
its coverage area.

AP
IS

The R10 LTE specifications offer two mechanisms for Inter-Cell Interference Coordination (ICIC) in
heterogeneous networks. The first one operates in the frequency domain and the second one in the
time domain.

AP

IS

The frequency domain ICIC makes use of carrier aggregation and is also called CA-based ICIC. The basis
for the mechanism is quite straight-forward: both the macro layer and the pico layer make use of two
(or more) component carriers (CC) by means of cross-carrier scheduling.
The macro cell performs cross-carrier scheduling from one CC and the pico from another CC, as shown
in the figure. Furthermore, the macro cell reduces its transmit power in the control region of its second
CC. This will offer protection for pico UEs located in the high interference extended range region. As an
option, protection for the data region in the extended pico cell area can be achieved with R8 ICIC.

69

AP
I

Obviously, this solution requires more than one LTE carrier. One possibility to deploy this option for
spectrum-limited network operators is to split their existing LTE spectrum into two (or more) carriers.
This will be, more or less, transparent to R10 terminals since they can still make use of the aggregated
spectrum. CA-based ICIC is fully backwards compatible to previous LTE releases, with the exception that
pre-R10 terminals cannot make full use of the system bandwidth- they will be limited to single-carrier
operation.

AP
IS

The time domain ICIC solution makes use of aso-called Almost Blank Subframes (ABS), introduced in LTE
Release 10. This solution is also called non-CA-based ICIC.

AP

IS

An Almost Blank Subframe is exactly that- almost blank. It only carries Cell-specific Reference Signals
(CRS), the Primary and Secondary Synchronisation Signals (PSS & SSS) and the Physical Broadcast
Channel (PBCH, carrying the crucial Master Information Block). No PDCCH or data region transmissions
take place within an ABS, with the exception of Paging messages and System Information Block 1 (both
of which are scheduled by PDCCH). The red and green patches in the ABS shown in the figure are
supposed to represent transmission of a PDCCH (red) carrying scheduling for a paging message and/or
SIB 1 (green).
The macro base station informs the pico (or other) base station about its currently used ABS-pattern via
X2-interface signalling. The pico cell can then make use of these subframes for scheduling of UEs in the
high interference area. The ABS patterns are defined over 40ms to be compatible with both the 10ms
radio frame timing and the 8ms HARQ-cycle timing used in LTE.
The usage of ABS subframes is, in principle, transparent to the UE. It will, however, affect the
measurements performed by the UE. In some subframes (non-ABS) the inter-cell interference will be
severe and in others hardly noticable (ABS). This will be true for measurements on the serving cell for
CQI reporting as well as measurements on neighbour cells for regular (handover) reporting.

70

AP
I

It does not make sense to report wildly fluctuating values, nor to report some kind of average between
ABS and non-ABS subframes. The pico cell will therefore configure the UE with measurement resource
restriction patterns. The UE will then know in what (ABS and non-ABS) subframes it should perform
measurements in order to get decent values. It is assumed that three different patterns will be needed,
as listed in the figure.

AP
IS

AP

IS

The purpose of the Load Indication procedure is to transfer interference information between eNBs for
interference coordination purposes. The initiating eNB (eNB 1) sends an UL Interference Overload
Indication' to another eNB (eNB 2) indicating in which parts of the spectrum it experiences low,
medium or high interference. It also specifies in which parts of the spectrum it is especially sensitive to
uplink interference. The interference is specified per Physical Resource Block (PRB) where one PRB
corresponds to 12 consecutive OFDM subcarriers (180 kHz bandwidth). The intention is that eNB 2
should use this information when setting its uplink scheduling policies.
The message may also include an indication, per PRB, of whether the transmitted DL output power from
eNB 1 will, in the immediate future, be above or below a signalled threshold. This is the DL Relative
Narrowband Tx Power parameter listed in the figure. The intention is that eNB 2 should use this
information when setting its downlink scheduling policies. Similarly, the parameter UL High Interference
Information indicates if, in the immediate future, eNB 2 should expect UL interference above or below
a signalled threshold (also set per PRB).
The Load Indication procedure thus provides important input for the ICIC function and may also be used
for node self-optimisation purposes. The parameters mentioned so-far were introduced in R8. New for
R10 is the ABS Information parameter. It contains a bitmap (40 bits) that specifies which subframes
that are configured as ABS subframes in the sending cell. It also specifies which subframes (out of 40)
that should be used in measurement resource restriction patterns (see previous page).
The Resource Status Reporting procedure can be used by a given eNB to request (periodic) reporting of
load measurements from a neighbouring eNB. The reporting may relate to one or more of the
measured objects indicated in the figure above. The eNB that receives the report may use this
information when setting mobility policies (e.g. when taking handover decisions).

71

AP
I

New for R10 are the last two parameters listed. The Relative Capacity specifies the available UL/DL
resources in eNB 2 given as a percentage relative other cells and/or relative the total E-UTRAN capacity.
The ABS Status reports the percentage of PRBs within ABS subframes allocated for UEs protected (by
ABS) from inter-cell interference. The parameter also specifies which ABS (as previously configured by
eNB 1) that have been used for protecting UEs.

72

AP
I

IS

AP

AP
IS

73

AP
I

IS

AP

AP
IS

AP
IS

AP

IS

Coordinated Multi Point (CoMP) transmission/reception refers to a system where transmission and/or
reception at multiple, geographically separated, antenna sites is dynamically coordinated in order to
improve system performance. The basic concept is shown in the upper part of the figure. Some sources
refer to CoMP as network MIMO. This term is somewhat misleading since the signals transmitted to
the UE from the distributed antennas are identical in terms of data content. It would make more sense
to refer to CoMP as network transmit diversity (at least as of R11).
CoMP has been standardised for LTE-Advanced in R11 as a potential tool to improve the coverage area
for high data rate availability and/or to improve the throughput for cell-edge UEs. CoMP can also be
seen as yet another mechanism to reduce inter-cell interference, i.e. spatial domain ICIC. CoMP is
applicable to both homo- and heterogeneous network deployments and can be realised in many
different ways, some of which are briefly described below.
The simplest form of CoMP would be intra-site with a single eNB. That is, the eNB can schedule
transmission to/from a single UE via two or more of its sectors/cells. This case naturally does not
require any coordination between base stations. (It can be debated if this case really is CoMP since the
antennas at the eNB are not geographically separated.)
Intra-site CoMP can also involve a real eNB and one or more, geographically separated, Remote Radio
Heads (RRH). In this case, the eNB performs all the processing and transmits a baseband signal to one or
more RRH. The Macro eNB and one or more of the RRHs transmit identical signals to the UE. In such a
scenario, the UE may not even be aware that multiple points are involved in the transmission (as in
MBMS SFN transmissions).
Inter-site CoMP means that two or more base stations are, to some extent, involved in the transmission
to the UE (or reception). This requires substantially more coordination between the base stations. With
some CoMP solutions it may be enough to exchange only scheduling decisions, while in others the base
stations may also need to exchange user data and/or CSI feedback in addition to the scheduling
information.
The backhaul (sidehaul?) between transmission points is, in R11, assumed to be fibre optic cables with
zero latency, even though solutions based on the legacy X2-interface are not precluded.

74

AP
I

Most of the details when it comes to radio interface signalling (PDCCH assignments, CSI reporting etc)
and the use of reference signals for CoMP has been defined in R11, but much work still remains to be
done in R12. The following pages provide an overview of the CoMP feature as of R11 (Dec-2013).

AP
IS

For downlink CoMP operation, the following 3GPP definitions apply:

IS

CoMP cooperating set: the set of (geographically separated) points directly or indirectly participating
in PDSCH transmission to the UE. The cooperating set may controlled by a single eNB or multiple eNB.
The cooperating set may be transparent to the UE.

AP

CoMP transmission point: a set of geographically co-located antennas. The points, within the CoMP
cooperating set, that are actively transmitting PDSCH to the UE are called directly participating points.
Those that do not actively transmit PDSCH to the UE are called indirectly participating points.
Indirectly participating transmission points contribute to the scheduling/beamforming decisions for a
given UE, even though they do not (currently) transmit data. The active transmission points for a UE
can change dynamically (i.e. an indirectly participating point can become a directly participating point,
and vice versa).
CoMP measurement set: the set of points about which the UE is measureing and reporting Channel
State Information (CSI). The reported results from the CoMP measurement set may cause points to be
added or removed from the UEs CoMP cooperating set.

RRM measurement set: the set of cells the UE performs normal neighbour cell measurements on for
handover preparation. Note: the cells outside the CoMP measurement set are not called transmission
points. RRM reports may trigger an update of the CoMP measurement set.

75

AP
I

It should be noted that the number of TPs in the figure above are chosen just to visualise the concept.
As of R11 the maximum number of TPs in the CoMP cooperating/measurement set is 4.

AP
IS

AP

IS

In the study phase for CoMP, a few scenarios were selected for simulations. This does not preclude
more scenarios being added as the standardisation work progresses. One should also not forget that
CoMP is not really a technology as such, but rather a class of techniques that can be realised in a large
number of ways. The intention of 3GPP is to minimise the impact on existing SU-MIMO, MU-MIMO
schemes as well as on existing reference signal structures.
The investigation of CoMP and its performance was divided into two phases in 3GPP. Two different
scenarios are investigated for each phase, as summarised below:
Scenario 1, phase 1: homogeneous network with intra-site CoMP
Scenario 2, phase 1: homogeneous network with high power Remote Radio Heads (RRH)
Scenario 3, phase 2: heterogeneous network with low power RRH within macro cell, different cell ID
Scenario 4, phase 2: heterogeneous network with low power RRH within macro cell, same cell ID.
The aim of the study was to evaluate the performance benefits, as opposed to single cell operation, and
the standardisation impact of CoMP for LTE. Reference network set-ups have been defined for each of
the 4 scenarios. A reference setup specifies, among other things, UE transmisison mode, system
bandwidth, channel model, UE feedback scheme and site distances to use for the simulations. The
reference setup and simulation results are gathered in TR 36.819.

76

AP
I

The figure above shows the setup for scenarios 1 & 2 for phase 1 simulation. The phase 1 simulation
results were completed by May 2011 and were based on contributions from a total of 20 companies.
The outcome of simulation phase 1 was somewhat inconclusive since the different contributions
showed highly varying gains (some contributions even showing negative gains).

AP
IS

77

AP
I

AP

IS

The figure above shows the setup for scenarios 3 & 4 for phase 2 simulation. The phase 2 simulation
results were completed by October 2011 and were based on contributions from 25 companies. The
outcome of the phase 2 simulations showed (almost) unanimousely that significant gains can be
achieved by using CoMP for cell-edge UEs. As a result, the R11 standardisation work for CoMP focused
mostly on heterogeneous network deployment (i.e. the scenarios shown above).

AP
IS

AP

IS

The figure above summarises the CoMP categories/schemes considered by 3GPP during the pre-study
phase for the CoMP feature. It should be noted that there is no universally agreed upon nomenclature
when it comes to different CoMP schemes. The following terminology only reflects the 3GPP technical
report for CoMP (TR 36.819). Downlink CoMP can be divided into Coordinated Link Adaptation (CLA),
Coordinated Scheduling (CS) and Joint Processing (JP).
Coordinated Link Adaptation is the simplest CoMP category, where a given TP makes use of scheduling
decisions received from neighbouring TPs when selecting MCS to use when transmitting to its UE(s).
With CLA there is no actual coordination of transmission decisions between the TPs. A given TP simply
informs neighbour TPs about its scheduling decisions. It is then up to those neighbour TPs how to use
that information in the link adaptation process.
Coordinated Scheduling can be seen as an extension of CLA where the involved TPs not only share
information but also make joint scheduling decisions. As a result of sharing scheduling information a TP
may not transmit at all in a given set of PRBs in a given subframe in order to decrease interference to
UEs served by a neighbouring TP on the same time-frequency resource. This scheme is referred to as
Dynamic Point Blanking. For the same reason a TP may, for a given time-frequency resource, adjust its
output power or transmission direction. These schemes are referred to as Coordinated Power Control
and Coordinated Beamforming respectively. In all schemes described so far the scheduling of a given UE
is, more or less, coordinated between multiple TPs. However, the actual data to be transmitted to that
UE is available only at a single TP.
Joint Processing (JP) allows true multi-point transmission where data to be transmitted to a given UE is
available at multiple TPs. The JP scheme is further divided into Joint Transmission (JT) and Dynamic Point
Selection (DPS). With JT, two or more TPs cooperate by sending data to a single UE. The multiple
transmissions are either coherent or non-coherent. Coherent transmission means, in effect, coordinated
beamforming towards a single UE where the different antenna ports belong to different TPs. Noncoherent transmission means that the same signal is sent to the UE from multiple TPs (compare soft
handover for UTRAN) thus achieving a power gain. It should be noted that none of the two schemes are
explicitly supported by the R11 standard due to restrictions in CSI reporting and L3/DCI signalling.
Dynamic Point Selection means that only one TP actively transmits data to the UE, but the active TP can
be switched on a per-subframe basis. The non-active, but cooperating, TPs may, or may not, perform
muting on the REs used for transmission by the active TP.

78

AP
I

Some of the schemes described above can also be applied for Uplink CoMP and can be realised without
the UE being aware of it. UL CoMP has very little (if any) impact on specifications and is therefore seen
as a network implementation issue. Uplink CoMP is not covered further in this course.

AP
IS

AP

IS

Coordinated Link Adaptation (CLA) allows one TP (i.e the eNB) to make use of scheduling decisions taken
by neighbouring TPs in order to reduce interference for its served UEs. The process as such is not
standardised but can be broken down into the steps described below and shown in the figure. The CLA
process needs to be performed per subframe per UE and may involve an arbitrary number of
neighbouring TPs.
Step 1: The eNB analyses Channel State Information (CSI) reports as sent by the UE. The difference here,
as compared to non-CoMP, is that the UE has been configured with multiple so-called CSI processes.
Each such process corresponds to a hypothesis regarding transmission decisions in neighbouring TPs.
For example, neighbour TP X transmits on the same resource as serving TP is one such hypothesis and
neighbour TP X does not transmit on the same resource as serving TP is another.
Step 2: The serving TP makes a scheduling decision regarding the upcoming transmission to the UE. At
this point, the TP has selected a resource (a set of PRBs) to use for the transmission. It may also have
made an initial choice regarding link adaptation parameters such as modulation and coding scheme
(MCS), coding rate, modulation etc.

Step 3: Neighbouring TPs share scheduling decisions with each other. It should be noted that the
specifications are completely silent on the details on this information exchange.
Step 4: The serving TP checks the reported CSI hypotheses received from the UE agains the decisions
taken by neighbouringh TPs and then selects/changes values for relevant link adaptation parameters.

79

AP
I

Step 5: Finally, the serving TP transmits data to the UE on the scheduled resource.

AP
IS

IS

For the different Coordinated Scheduling (CS) categories data for a single UE is available in and
transmitted from a single transmission point at a time (just as with CLA described on the previous page).
The difference to CLA is that, with CS, the actual scheduling decisions are coordinated between TPs
within the CoMP cooperating set in order to control the interference within and between cells.

AP

The upper part of the figure shows Dynamic Point Blanking (DPB) while the lower part of the figure
shows Coordinated Beamforming (CB). For simplicity, only two TPs are involved in the examples.
With DPB the serving TP predicts the interference impact on the UE assuming that neighbouring TPs
would transmit on the same time-frequency resource. It also predict how much better channel quality
would be if the neighbours did not transmit on the same resource. If needed, the serving TP can then
restrict or prohibit neighbours to transmit on a given resource in a given subframe. Again, the
specifications are silent as to how this would be communicated to neighbouring TPs.
With CB, the two (or more) transmission points can use the same time/frequency resource for PDSCH
transmissions towards cell-edge UEs without causing any inter-cell interference by agreeing on what
beamforming precoders to use. This scheme can be seen as an extension of the single-cell MU-MIMO to
include multiple cells, thus realising inter-cell Multi-User Beamforming (MU-BF). One can also see CB as
an extension of the R10 eICIC feature to also make use of the spatial domain.

80

AP
I

The Coordinated Scheduling schemes rely on the same CSI reporting mechanism as DLA, where the UE
reports multiple interference hypotheses. The reference signal configurations needed by the UE to
make these multi-TP measurements are discussed later in this chapter.

AP
IS

IS

Joint Processing (JP) is a category of CoMP operation where downlink data to a single UE is available at
multiple transmission points, as opposed to DLA/CS where data is only available at a single point. Data
transmission may take place from multiple points (Joint Transmission, JT) or a single point (Dynamic
Point Selection, DPS) in a given subframe. Hybrid schemes mixing JP and CS/CB are potentially possible.

AP

With Dynamic Point Selection (DPS) the data is available at multiple points but only sent from one point
at a time. The transmitting point may change on a per-subframe basis based on UE CSI reports. The
involved TPs may or may not be part of the serving cell (i.e. they may be eNBs or RRHs).
In this scheme the UE is configured to report interference hypotheses, as before, but also to report CSI
measurements, based on CSI-RS, for all involved TPs. The serving cell/TP them decides whether to
transmit PDSCH in the serving cell or to transmit PDSCH from one of the other TPs. Please note that the
scheduling assignments (on E-PDCCH) are always transmitted from the serving cell/TP. This decision
may change on a per-subframe basis. How data for a given UE is made available in a non-serving TP is
not defined in the specifications. The Enhanced PDCCH (E-PDCCH) and the reference signal
configurations mentioned in the figure are described later in this chapter.

With Joint Transmission (JT) downlink data for a given UE is simultaneously transmitted, using the same
time/frequency resource, from multiple transmission points within the CoMP cooperating set. In its
simplest form the involved TPs transmit identically processed signals to the same UE, thereby
achieving a power gain at the UE. Care must be taken though, so that the extra signals do not cause
unwanted intererence to transmissions towards other UEs. Two or more TPs may also jointly select
precoders to use for transmission to a single UE, resulting in beamforming involving multiple cells (or at
least with geographically separated antennas).

81

AP
I

The JT category of CoMP puts much higher requirements on the coordination links and the backhaul
since the user data need to be made available at and synchronously transmitted from multiple
transmission points. The amount of control information to be exchanged over the coordination links is
also larger. For example, it may need to include ACK/NACKs and periodic/aperiodic CSI reports sent
from the UE. More elaborate CSI reporting mechanisms, compared to previously described CoMP
schemes, are also needed. As of R11 these signalling advances have not been standardised.

AP
IS

AP

IS

A new CoMP-specific transmission mode was, after much debate in the standardisation process,
introduced as part of R11. This new transmission mode (TM10) is very similar to TM9, but excludes
some functionality which is allowed for TM9 (such as the use of R8 cell-specific reference signals, CRS).
As we have seen before in this course, with a new TM also follows a new DCI format- in this case DCI
Format 2D. Format 2D carries the same information as Format 2C with the addition of a new 2-bit field:
the PDSCH RE mapping and quasi co-location indicator. This field refers to one of (up to) four
parameter sets configured by higher layers (RRC). This new 2-bit field in DCI Format 2D provides the UE
with the folowing:
It informs the UE where the data region starts in the subframe (pdsch-Start)
It identifies the TP used, by means of an index number associated with a CSI-RS resource
It specfifies the number of CRS antenna ports, and the subcarrier/frequency shift of their location
It specifies a zero power CSI-RS confoguration (ZP CSI-RS)
It tells the UE if the PDSCH transmission is associated with a CRS resource which is quasi co-located
(QCL) with the corresponding CSI-RS resource or not.
The pdsch-Start information is needed since the UE only monitors PCFICH in the serving cell/TP and thus
do not know where the subframe data region starts in another TP (the pdsch-Start value must be semistatically configured in the network on a per-TP level).
The CRS and ZP CSI-RS information is needed in the UE for proper (de)rate matching of the PDSCH. This
is because the PDSCH transmission will not be mapped onto REs used for CRS, CSI-RS and ZP CSI-RS. This
information is needed for time/frequency synchronisation purposes as well.
The QCL information is needed by the UE for proper channel estimation. The definition of QCL is as
follows:
Two antenna ports are said to be quasi co-located if the large-scale properties of the channel over
which a symbol on one antenna port is conveyed can be inferred from the channel over which a symbol
on the other antenna port is conveyed. The large-scale properties include one or more of delay spread,
Doppler spread, Doppler shift, average gain, and average delay (from TS 36.211).

82

AP
I

In practice, all antenna ports at the eNB site will be quasi co-located with each other while antenna
ports at, for example, a remote radio head will not be quasi co-located with the ports at the eNB.

AP
IS

AP

IS

Transmission Mode 10 does not allow data transmission with Cell-specific Reference Signals (CRS). For
PDSCH reception in TM10 the UE therefore must be configured with a set of UE-specific Demodulation
Reference Signals (DM-RS). This is the same concept as the R9/10 DM-RS usage for TM8-9 but with one
important difference. The difference is that the R11 DM-RS are generated from a virtual physical cell id,
thus decoupling the relationship between cell and reference signal. Such cell agnostic reference signals
are an absolute requirement for any CoMP scheme that involves more than one cell.
As mentioned earlier, the UE needs to be configured with one or more CSI process for CoMP operation.
A CSI process is a combination of one CSI Interference Measurement resource (CSI-IM) for interference
estimation and a set of CSI Reference Signals (CSI-RS, one per antenna port).
The CSI-RSs are the same as described earlier (in the MIMO chapter). The exakt time-frequency
locations of Resource Elements (RE) carrying the CSI-RS resources are flexible and will, from a CoMP
point of view, be different in neighbouring TPs. Depending on the CoMP scheme used, the UE will
measure CSI-RS only in the serving cell/TP (e.g. DLA) or in multiple cells/TPs (e.g. DPS).
In a given cell/TP, a CSI-IM resource is simply a set of resource elements where nothing is transmitted
(i.e. configured in the same way as a Zero Power CSI-RS). These empty REs are used by the UE to
measure interference from one or more neighbouring TPs.
The figure shows an example with two TPs and the Dynamic Link Adaptation Scheme. Two CSI processes
are configured, corresponding to the following interference hypotheses:
Hypothesis A: no transmission from TP Right when serving cell/TP transmits, and
Hypothesis B: transmission from TP Right, when serving cell/TP transmits.
For hypotheses A, the UE needs to measure CSI-RS in the serving cell. This corresponds to RE-group 1.
The UE also needs to measure the interference from TP Right when it is not transmitting. This
corresponds to RE-group 2 (these REs are configured as ZP CSI-RS in TP Right).
For hypotheses B, the UE needs to measure CSI-RS in the serving cell. This corresponds, again, to REgroup 1. The UE also needs to measure the interference from TP Right when it is transmitting. This
corresponds to RE-group 3 (these REs are configured as CSI-RS in TP Right).

83

AP
I

The same appoach can be extended to the more complex CoMP schemes. For example, in the case of
Dynamic Point Selection with two TPs the UE would be configured with two CSI-RS resources (one per
TP) and four CSI-IM resources (TP transmit or not for both TPs) resulting in four CSI processes. The
approach can of course also be extended to more than two TPs. (Just for fun, the reader may try to
figure out the number of CSI processes and CSI-RS/IM resources that are needed for the case three-TP
Dynamic Link Adaptation.)

AP
IS

IS

A new mapping of the PDCCH, called the Enhanced PDCCH (E-PDCCH), was introduced in R11 for both
CoMP and non-CoMP operation. The E-PDCCH is mapped in the data region similarly to the R-PDCCH for
relay operation. This mapping of the E-PDCCH allows for frequency domain eICIC (see HetNet chapter)
also for downlink control signalling.

AP

The E-PDCCH carries no Cell-specific Reference Signals (CRS) but instead make use of the UE-specific
Demodulation Reference Signal (DM-RS). The E-PDCCH can therefore be transmitted to the UE using
non-codebook based precoding (see MIMO chapter). This is useful for the CoMP feature in particular
but also for more advanced antenna solutions in general (where the network wants the freedom to
change precoding without signalling the UE). The E-PDCCH is transmitted with transmit diversity on one
or more of the antenna ports 107-110.
One E-PDCCH is mapped onto one PRB-pair (more for high aggregation levels). The UE is configured with
one or two sets of PRB-pairs where the E-PDCCH may be transmitted. One such set consists of 2, 4 or 8
PRB-pairs. Within a given PRB-pair set either localised or distributed mapping of a single E-PDCCH is
used (the figure shows localised mapping). Any PRB-pair not used for E-PDCCH transmission in a given
subframe can instead be used for PDSCH transmission.
There can be several reasons for configuring a single UE with two E-PDCCH sets. One could be to allow
for both localised and distributed mapping (and switch from one to the other depending on radio
conditions). Another reason may be if the UE is configured for Dynamic Point Switching with two TPs. It
is then possible to associate each TP with one of the E-PDCCH sets. The E-PDCCH used for assignment
will then implicitly identify from which TP the associated PDSCH transmission will take place.
Content-wise there are only small differences between the regular PDCCH and the E-PDCCH. The main
difference is that the E-PDCCH is never used for carrying multi-UE DCI assignments such as DCI format
3/3A (power control) and DCI format 1C (system information).

84

AP
I

Like the PDCCH, the E-PDCCH is constructed from smaller building blocks. One E-PDCCH is constructed
from Enhanced Control Channel Elements (ECCE). Depending on aggregation level, payload size and
radio conditions the E-PDCCH can be mapped onto 1, 2, 4, 8, 16 or 32 ECCEs. Each ECCE, in turn, is
constructed from 4 Enhanced Resource Element Groups (EREG). Finally, one EREG consists of 9 Resource
Elements (RE). For the finer details regarding mapping of E-PDCCH onto REs, EREGs and ECCEs the
reader should consult TS 36.211.

85

AP
I

IS

AP

AP
IS

86

AP
I

IS

AP

AP
IS

AP
IS

IS

The 3GPP R12 Study Item Small Cell Enhancements (SCE) is a natural evolution of R10/R11 HetNet and
CoMP. The small cells concept integrates a number of R10/R11 features such as HetNet, CoMP, CA etc
into a single framework. The studied scenarios and some of the basic assumptions for small cells are
summarised in the figure and briefly described in the following.

AP

Both indoor and outdoor deployment. Indoor deployment may not coincide with macro coverage.
The small cell layer and the macro layer will operate on different frequencies (except in scenario 1).
The 3.5 GHz band was extensively used for the small cell layer during the study phase.
Much denser deployment of cells as compared to R10/R11 HetNet/CoMP. Small cells will be grouped
into clusters, with one or more clusters per macro cell. (4 or 10 cell clusters where used in the study.)
Non-ideal backhaul is assumed. This could be an updated X2-interface or a completely new interface.
Co-existence of FDD and TDD modes. There are a few related R12 study/work items relating to joint
FDD/TDD operation. For example, there is a work item on dynamic TDD for small cells, where the
uplink/downlink subframe configuration can be dynamically adapted in order to match the real-time
traffic situation.
The small cell low power nodes are expected to be small (obviously), cheap, energy efficient and easy
to operate and maintain. The latter point goes hand-in-hand with the R12 work on next generation
Self-Organising Networks.
Mechanisms to increase spectrum efficiency has been studied, for example the introduction of
256QAM modulation. Introduction of 256QAM will have an impact on CQI/MCS/TBS tables and will
most likely need new DCI/UCI formats. It has also been discussed if the reference signal overhead
(CRS, DM-RS, CSI-RS) can be reduced without hurting legacy pre-R12 terminals.

87

AP
I

It is still (jan-2014) not certain what functions will end up in the final standard for SCE. The reader is
therefore cautioned not to take everything written in this chapter too literally.

AP
IS

AP

IS

In legacy LTE systems (up to R11) the eNB is always-on regardless if any users are in the cell or not. The
eNB always transmits Cell-specific Reference Signals (CRS), synchronisation signals, System Information
Blocks (SIB) and PDCCH for scheduling of the SIBs. This is needed for synchronisation and cell search
purposes by idle mode UEs in case they come within the coverage area of the cell. The CRS and
synchronisation signals are also used by UEs in other cells for Idle and Connected mode measurement
purposes.
A number of mechanisms for switching small cells on and off have been studied. The aim is, of course,
to make the network as energy efficient as possible while still not sacrificing network capacity. Having
some cells switched off will also benefit the interference situation in neighbouring cells.
The upper part of the figure shows existing mechanisms for energy saving. In the simplest case, the cell
can simply be switched off completely during certain time periods (e.g. a few hours in the middle of the
night). From R9 and onwards there is support in the X2-interface specification for signalling between
eNBs related to cell activation/deactivation. A cell with no active users can decide to switch itself off
and will signal this to its neighbours. One or more of those neighbours can be configured as so-called
compensation cells. A compensation cell will increase its coverage area for as long as the energy saving
cell is deactivated. The compensation cell(s) can also order the dormant cell to activate itself again, for
example if the traffic load in the compensation cell rises above some configured threshold. the time
scale for the R9 energy saving feature is in the order of seconds or minutes.
The need for a much faster on/off mechanism is foreseen for small cells, operating on time scales from
100s of milliseconds to seconds. The lower part of the figure shows some of the studied mechanisms. It
should be noted that there is no need for compensation cells in a small cell environment since macro
cell coverage is always present (with the exception of indoor scenario 3). Faster transitions for small cell
on/off (below 100ms) may be acheived by using so-called discovery signals and/or a new R12 feature
called Dual Connectivity. This is elaborated upon later in this chapter.

88

AP
I

In the best of worlds, in some future release, the cell on/off mechanism would/will operate on a persubframe basis. This is currently not possible, even though simulations carried out during the small cell
study phase used such an ideal on/off mechanism to get an upper bound for the performance gain
achievable.

AP
IS

AP

IS

As for any LTE cell in any release, some form of cell search or cell discovery mecanism is needed for
small cells. In pre-R12 LTE this is supported by the continuous transmission of synchronisation signals,
reference signals and system information messages. In a small cell scenario things get a little more
complicated since a given small cell may or may not be on at a given point in time. Dense deployment
of small cells (on a different frequency than the macro) will also lead to substantial UE effort to perform
inter-frequency measurements. Yet another issue is that inter-cell interference on synchronisation and
reference signals can be severe in a dense small cell environment.
The figure shows some suggested signals to use for small cell discovery. This includes legacy signals,
modified legacy signals and new signals (all can be referred to as small cell discovery signals). Some level
of coordination between small cell eNBs regarding transmission of discovery signals may be beneficial in
order to avoid RS collisions (e.g. by using different frequency domain offsets). The UE may also be
equipped with a receiver capable of cancelling out interference to Synchronisation Signals (SS) and/or
Cell-specific Reference Signals (CRS) between cells.
As for new reference signals, one suggestion is the so-called Companion Discovery Signal (CDS). This
makes use of the currently unused resource elements adjacent to the legacy synchronisation signals
that occur twice per 10ms radio frame (in slot 0 and slot 10).

89

AP
I

The figure also shows how discovery signal transmission may be used in combination with the small cell
on/off mechanism. For example, all dormant cells in a cluster may periodically transmit discovery signals
that are easy to detect by the UE (the macro eNB can provide UE with the discovery signal transmission
period). Once detected, the UE reports the cell identities of the dormant cells to the macro eNB. The
macro eNB may then, if needed, send an activation order to the small cell(s).

AP
IS

IS

The Small Cell Enhancements study item (SCE) looks into both synchronised and unsynchronised
network scenarios. However, having small cells synchronised (at least within the same cluster) will
greatly benefit features such as eICIC, CoMP, carrier aggregation and Dual Connectivity. Synchronised
transmission will also benefit UEs equipped with receivers for reference signal interference cancellation.

AP

In legacy networks synchronisation is based on GPS (or other satellite system) and/or special backhaul
sync protocols. In a small cell environment this may not be possible (indoor cells and/or non-ideal
backhaul). The SCE study therefore looks into mechanisms for radio interface based synchronisation
between cells. Synchronisation is needed between the macro cell and each cluster, between clusters
and between small cells within a cluster. The synchronisation accuracy target was set to 3 s for the
study.
A new scheme for radio interface synchronisation, called network listening, will be introduced for small
cells. In this scheme, the cell providing sync information to another cell is referred to as the source and
the cell acquiring sync from another cell is referred to as the target. A target may use multiple sources
and a target may also act as source to another target cell (thus leading to multi-hop sync links). The
target will actively select the source e.g. based on channel conditions and the number of hops. In a
given cluster, at least one cell will use the macro cell and/or GPS as the source for synchronisation.
The target cell will periodically listen for bursts of synchronisation signals from all sources. The time
scales involved for listening will depend on, among other things, the stability of the internal clock in the
eNB. During the study phase a 1ms long listening window was used with a repetition period of 10
seconds. Legacy signals (CRS/PRS/CSI-RS) were used as synchronisation signal. The target will mute its
own transmission during the listening window (at least when source and target operate on the same
frequency). The corresponding downlink subframe in the target is, for the UEs, configured as an MBSFN
subframe.
In some case there may also be a need for inter-operator synchronisation. Some markets use contiguous
TDD spectrum allocation among different operators. This leads to significant mutual interference and
degrade the network performance of both sides. Such cases will require a mutual network listening
mechanism (i.e. a cell is both source and target). An inter-operator X2 interface may also be needed in
order for the involved eNBs to exchange synchronisation signal configuration parameters.

90

AP
I

An alternative/complementary mechanism to network listening is to use UE-assisted synchronisation.


The target will in this case order one or more UE to obtain synchronisation information from one or
more source and report back to the target. It is not yet defined what the UE will report. Nor is it defined
if this will lead to new performance requirements for the UE.

AP
IS

Inter-node radio resource aggregation is a potential tool for improving the per-UE throughput in the
network. This means that the UE receives data from more than one eNB at the same time. (Please note
that these data streams will not carry copies of the same data as in the CoMP solution.)

AP

IS

The UE will always be connected to a Master eNB (MeNB) and may in addition be connected on or more
Secondary eNBs (SeNB). The feature is referred to as Dual Connectivity (DC) regardless of the number of
SeNBs involved. Non-ideal backhaul is assumed between MeNB and SenBs. As of R12 the UE cannot be
connected to more than 5 cells simultaneously (one MeNB and four SeNBs).
Various scenarios for Dual Connectivity have been studied, for example DC for Small Cell Enhancements
scenario 1 (upper-right) and uplink/downlink split between MeNB and SenB (lower-left).
Particular attention was given to Small Cell Enhancements scenario 2 (macro and small on separate
frequencies) with no explicit uplink/downlink split. The remainder of this chapter will focus on Dual
Connectivity for this scenario.

91

AP
I

(The reader is encouraged to consult technical report 36.842 for the reasons why the other scenarios
were down-prioritised.)

AP
IS

Nine different options for User Plane (UP) protocol stacks between MeNB and SenB and between SeNB
and Serving Gateway (SGW) were studied (see TR 36.842 for all options and their pros and cons). Two
options were finally selected for further standardisation, as shown in the figure above.

AP

IS

On the left-hand side of the figure is shown the option where all UP connections over the S1-U interface
terminate in the MeNB. For inter-node resource aggregation the MeNB can then decide to perform
bearer splitting and allocate part of the data flow for one (or more) EPS bearer to SeNB(s). This SeNB
addition procedure will require signalling between the eNBs as well as signalling towards the UE. No
signalling towards the core network (i.e. MME) is foreseen for this option.
The bearer to split is divided at the Packet Data Convergence Protocol (PDCP) level, thus requiring new
functionality in the PDCP protocol specification. The interface between MeNB and SeNB is, currently,
labelled the Xn-interface. Most likely this is simply a R12 version of the already existing X2-interface.
(Even if a new Xn-interface is defined it will use the same GTP-based UP stack as on S1/X2-interface.)
On the right-hand side of the figure is shown the option where all S1-interface UP connections go to
both MeNB and SeNB(s). No bearer splitting is performed, each node manages one or more complete
EPS bearer. Signalling on the Xn-interface is still needed since it is the MeNB that decides whether a
bearer should be managed by SeNB or not. With this option also the MME must be involved in SeNB
addition, since the S1-interface UP connection goes directly from SGW to SeNB.

92

AP
I

There are still many open questions regarding these two options. For example, is it possible to have a
mix of the two active for a single UE? Another question, valid for both options, is what will happen with
uplink transmissions: will the UE have uplink only to MeNB or to all involved cells?

AP
IS

Several different options were investigated also for the Control Plane architecture for Dual Connectivity
(again, please refer to TR 36.842 for more details). The final conclusions were as follows:

IS

Only the MeNB terminates the CP connection (for a given UE) towards the MME.

AP

RRC messages are only sent from the MeNB. Since the radio resource managment (RRM) functionality is
independent in MeNB and SeNB this requires signalling between the two prior to signalling the final RRC
configuration to the UE (as shown in the lower-left of the figure). Logically, the two RRM entities
communicate by exchanging RRC containers. These containers are carried inside specific Xn Application
Protocol (XnAP) messages.

93

AP
I

The right side of the figure shows the signalling for addition of SeNB for a given UE. Please note that the
stage 3 protocol specifications will not be finalised until Sept-2014. The flow shown is included to
illustrate the concept of SeNB addition on a high level and should not be taken as an accurate
representation of (future) protocol functionality.

94

AP
I

IS

AP

AP
IS

95

AP
I

IS

AP

AP
IS

AP
IS

96

AP
I

AP

IS

The tables on the next few pages summarize the Release 12 feature list as of Jan-2014. The table above
shows Work Items defined for Release 12 (and 13). Some of these where originally part of Release 11
but where, at the Sept-2012 RAN meeting, pushed from Release 11 into Release 12.

AP
IS

The table above shows Study Items defined for Release 12 and 13 (including those that have been
pushed from Release 11 into Release 12).

97

AP
I

AP

IS

Please note that the lists on this and the previous page are not 100% complete. The author has taken
the liberty to include only those work/study items that, in one way or another, relate to the topics
discussed in this course. For a complete and updated list, please consult the link to the 3GPP Work Plan
referenced at the end of this chapter.

AP
IS

AP

IS

Over the last decade or so, cell site architectures have been evolving from the legacy architecture
where large radio units (BSC, RNC) are located remote from the antennas, to an architecture wherein a
separate RF portion of the radio can be located more closely to the antenna. This separation of the
digital radio, BBU (Base Band Unit), from the analog radio, RRH (Remote Radio Head), allows for a
reduction of the equipment foot print at the site and for a more efficient operation of the network. it is
common that a digital fiber optic link provides the connection between the BBU and the RRH (as in the
HetNet architectures).
The next logical step is to integrate the radio functions into the antenna and distribute the radio
functionality across the antenna elements present. This gives us an active anntenna system, AAS.
There are several benefits with an AAS configuration. The, perhaps, most obvious benefit with the
integration of the radio within the antenna elements themselves is the elimination of several
components (such as cables, connectors, mounting hardware etc) giving an overall reduction in the
physical tower space required. (Also, by integrating the RRH functionality into the antenna, the
aesthetics of the site can be improved.)

98

AP
I

Another benefit of an AAS configuration is the ability to electronically tilt transmission beams (as
opposed to mechanical tilt) by having independent control of the phase, amplitude, and delay of
individual carriers on each antenna element. This is good for multi-RAT systems where different
technologies (UMTS, LTE etc) coexist in the same frequency band, but may require different downtilt
angles. An AAS configuration also opens up for more advanced beamforming techniques within one and
the same RAT. The latter is discussed on the next page.

AP
IS

AP

IS

Since AAS antenna systems have electronic (or virtual) tilt capabilities it becomes possible to utilise the
vertical dimension for beamforming, as opposed to only the horizontal dimension in R8-11. The figure
to the upper left shows a simple case of vertical beamforming where the eNB can form two separate
sub-sectors, thus creating an inner cell and an outer cell (these may or may nor be defined as actual
cells with different Phy Cell-IDs).
A natural progression from vertical beamforming is shown in the upper right figure. Here, elevation
beamforming is used to provide coverage in high-rise buldings, thus creating upper cells and lower
cells.
With more advanced AAS arrays it becomes possible to fully utilise all three spatial dimensions and
create very narrow beams. This is shown in the lower part of the figure, where the beams are
dynamically directed towards individual UEs. This is sometimes called pencil beamforming.

99

AP
I

These more advanced MIMO schemes are addressed in Release 12 by means of two Study Items looking
into RF/EMC requirements for AAS base stations and 3D channel models. (In principle, 3D MIMO can be
realised with the non-codebook based precoding concept introduced in R10.)

AP
IS

IS

Release 10 increased the number of Tx/Rx antenna ports at the eNB from 4 to 8 in order to increase the
system capacity and/or single UE peak throughput. The natural progression is, of course, to go beyond 8
antenna ports. These so-called massive MIMO configurations are not be expected in the specifications
until, earliest, Release 13.

AP

With AAS configurations, as discussed earlier, it becomes easier to shrink the physical size of antenna
arrays. Further, the massive MIMO deployments are expected to be used only in high or very high
frequency bands, which allows for even more miniturization.
it is assumed that the 3D beamforming techniques discussed on the previous page are not really
feasible unless the number of antenna elements drastically increasy (at least up to 64 elements,
probably beyond).

100

AP
I

The lower part of the figure shows two examples, a 64 element eNB array for 2.5 GHz to the left and a
256x16 eNB/UE configuration for the 30 GHz band to the right. There are currently (Release 12) no LTE
frequency bands defined above 3.5 GHz. Release 13 will probably contain a Study Item on deployment
of LTE in unlicensed frequency bands, including bands above 3.5 GHz.

AP
IS

The Release 11 version of the CoMP standard assumes that the backhaul interface between two or
more TPs is based on fibre optics with negligible latency (in the s range). Due to this limitation, some
operators cannot enjoy the performance benefits made possible by CoMP operation.

AP

IS

A new so-called Study Item (SI) was therefore started as part of Release 12 with the aim of evaluating
the performance benefits and identifying potential standardization impacts for candidate CoMP
techniques involving multiple eNBs with non-ideal backhaul (quote from TR 36.874). The SI looks into
three different scenarios, described on the next page, and different assumed backhaul latencies. The
node-to-node backhaul latencies studied/simulated are 2ms, 5ms, 10ms, 30ms and 50ms.
After the SI is completed a decision will be taken whether the possible performance benefits are large
enough to justify further standardisation or not (i.e. whether a so-called Work Item, WI, shall be
initiated or not). Such a WI will need to focus on defining new signalling for proper inter-eNB CoMP
operation under the non-ideal backhaul assumption. It was also stated, when this CoMP-related SI was
created, that any standardised solutions should be aligned with the ongoing Small Cell Enhancements
Work Item (also part of R12 and described in other parts of this document).

The Technical Report associated with this SI (TR 36.874) lists some possible scheduling and UE-related
information parameters that may or may not need to be exchanged between the involved eNBs. The
information is roughly divided into two groups:
information which is considered valid for a longer time than the backhaul delay, and
information which is considered valid for a shorter time than the backhaul delay.

101

AP
I

In the figure these are referred to as LOCAL and SHARED information sets. The LOCAL information is
assumed to be based on, among other things, CSI reports received from the UE.

AP
IS

IS

The three deployment scenarios to investigate during the SI phase where selected partly from the
completed R11 CoMP work and partly from the ongoing R11/R12 Small Cell Enhancements (SCE) work.
The scenarios are numbered according to the CoMP Technical Report (TR 36.819) and the Small Cell
Enhancements Technical Report (TR 36.872).

AP

CoMP scenario 2: homogeneous network deployment with CoMP operation between macro eNBs. The
scenario assumes three sectors/cells per site and cooperation between either 9 cells or 21 cells. (It
should be noted that TR 36.819 uses the term high-power Remote Radio Head for all but the central
TP in this scenario.)
SCE scenario 1: heterogeneous network deployment with CoMP operation between macro eNB and
clusters of small cell eNBs. The scenario assumes three sectors/cells at the macro eNB and one or two
clusters of 4 or 10 small cell eNBs per sector. It is further assumed that the macro layer and the small
cell layer operate on the same carrier frequency (the simulations use 2GHz).
SCE scenario 2a: heterogeneous network deployment with CoMP operation between small cell eNBs.
The assumptions for this scenario are identical to SCE scenario 1 with the exception that the macro and
small layers operate at different carrier frequencies (the simulations use 2GHz for the macro layer and
3.5 GHz for the small cell layer).
The backhaul assumptions, as of TR 36.872 and TR 36.874, are the same in all three scenarios:
there exists backhaul links between macro eNB sites,
there exists backhaul links between a macro eNB and all small cell eNBs within its coverage,
there exists backhaul links between small cell eNBs within a cluster,
other backhaul topologies as defined by each involved company.
The final simulation results for all three scenarios can be found in TR 36.874. The results include
different combinations of backhaul delay, backhaul topology, cluster size, CoMP scheme etc and shows
highly varying results in performance. The reference scheme for performance comparison is referred to
as the best pre-release-12 scheme. According to TR 36.874 this mysterious scheme includes:

R11 intra-site CoMP between the 3 sectors of each macro,


R11 eICIC and other R10/R11 coordination signalling between cells and...
... other features/functions that each company believes to be the best pre-release-12 scheme.
Given all the above it is not very surprising that the conclusion in TR 36.874 states that:

102

AP
I

it was observed that CoMP-NIB gain varies as a factor of deployment scenario, backhaul delay,
coordination scheme, resource utilization factor, and coordination size. (NIB = Non-ideal Backhaul.)

You might also like