OUTLOOK

Visions and research directions for the Wireless World

2014, No 15

LTE Small Cell

Enhancement by
Dual Connectivity
 Working Group C
“Communication Architectures and Technologies”

White Paper

LTE Small Cell Enhancement

by Dual Connectivity
November 2014

Version 1.1
 Contributors

Full Name Affiliation Contact

Jian Zhang Huawei Technologies, China [email protected]
Qinghai Zeng Huawei Technologies, China [email protected]
Mahmoodi, Toktam King’s College London, UK [email protected]
Andreas Georgakopoulos University of Piraeus, Greece [email protected]
Panagiotis Demestichas University of Piraeus, Greece [email protected]
 WGC – Communication Architectures and Technologies

Table of Contents

0. Abbreviation ..................................................................................................................... 5
1. Purpose ............................................................................................................................. 6
2. Scope ................................................................................................................................. 6
3. Motivation.......................................................................................................................... 6
3.1 Challenges.................................................................................................................... 6
3.2 Motivations ................................................................................................................... 7
3.3 Scenarios...................................................................................................................... 7
4. Architecture ...................................................................................................................... 8
4.1 Control plane architecture ............................................................................................ 8
4.2 User plane architecture ................................................................................................ 8
4.3 Backhaul aspects ....................................................................................................... 10
5. Network interface ........................................................................................................... 11
5.1 S1 interface ................................................................................................................ 11
5.2 X2 interface ................................................................................................................ 11
5.3 Flow control ................................................................................................................ 12
6. Control plane handling .................................................................................................. 12
6.1 eNB Sychronization .................................................................................................... 12
6.2 Security handling ........................................................................................................ 12
6.3 System information acquisition .................................................................................. 12
6.4 SeNB radio resource management ............................................................................ 13
6.5 Selection of connection at the UE .............................................................................. 14
6.5.1 Use of Analytics .......................................................................................................... 14
6.6 UE capability handling ................................................................................................ 14
6.7 SeNB radio link failure handling ................................................................................. 14
7. User plane handling ....................................................................................................... 15
7.1 UL-DL split .................................................................................................................. 15
7.2 UL bearer split ............................................................................................................ 15
7.3 UCI transmission of SeNB.......................................................................................... 15
7.4 PDCP handling ........................................................................................................... 16
7.5 RLC handling .............................................................................................................. 16
7.6 UE MAC entity modeling ............................................................................................ 16
7.7 Random access .......................................................................................................... 16
7.8 BSR procedure ........................................................................................................... 17
7.9 LCP procedure ........................................................................................................... 17
7.10 PHR procedure ........................................................................................................... 18
7.11 DRX ............................................................................................................................ 19
7.12 Multi-Path TCP support .............................................................................................. 19
7.13 Activation/deactivation................................................................................................ 19
8. Potential enhancements in future releases................................................................. 19
9. Summary ......................................................................................................................... 19
10. References ...................................................................................................................... 20

Page 3 , Total20
 WGC – Communication Architectures and Technologies

Executive Summary

The WWRF and WGC

The White Paper on “LTE Small Cell Enhancement by Dual Connectivity” is the outcome of discussions and
suggestions in the context of the Wireless World Research Forum (WWRF) meetings and the Working Group C
(WGC) in particular, which handles issues related to “Communication Architectures and Technologies”.

LTE small cell enhancement by dual

connectivity towards 5G evolution
Framed in this context, the paper describes the common vision on small cell enhancement by dual
connectivity. Small cells are becoming a promising technology to meet ever increasing traffic capacity and data
rate demand. Small cells are typically deployed as hotspots within macro cell coverage. By enhancing small
cells by dual connectivity certain benefits are anticipated such as:
 Increased UE throughput especially for cell edge UEs;
 Mobility robustness enhancement;
 Reducing signaling overhead towards the core due to frequent handover.

Architecture and technical

considerations
In this respect, the White Paper elaborates on the architecture of dual connectivity where the UE can
receive/transmit data from/to multiple eNBs simultaneously. There is a Master eNB (MeNB) and one or more
Secondary eNBs (SeNB). In LTE release 12 specifications, only the case of one MeNB and one SeNB is
considered. For the support of dual cell connectivity, some modifications to existing S1 interface and X2
interface are needed as well and these aspects are elaborated on Chapter 5. Finally technical considerations
regarding the control and user planes are presented in sections 6 and 7 respectively.

Proposed enhancements for LTE

R-12 and beyond
LTE small cell enhancement by dual connectivity is a technology which extends carrier aggregation (CA)
and coordinated multi-point (CoMP) to inter-eNB with non-ideal backhaul while LTE release 12 focuses on
inter-eNB CA case for per UE throughput improvement and mobility robustness. Specifically, Chapter 8 of this
White Paper discusses ‘Potential enhancements in future releases’ where ultra-dense infrastructure
deployments, utilization of mmWave bands, multiple antennas, integrated use of M2M communications with
ultra-low latency and very high reliability will set the requirements of future releases towards 5G.

Way Forward
More White Papers related to innovations in the field of “Communication Architectures and Technologies”
will follow in order to address sufficiently several trending topics on: Management of evolved RATs;
Ultra-Dense Networks; Dynamic/Flexible Spectrum Management and Spectrum Sharing; Virtualization of
Infrastructure by exploiting the concepts of Software-Defined Networking (SDN) and Network Function
Virtualization (NFV); Cloud-RANs.

Page 4 , Total20
 WGC – Communication Architectures and Technologies

0. Abbreviation

AP Access Point
CA Carrier Aggregation
CCE Control Channel Element
CN Core Network
CoMP Coordinated Multi Point
CRE Cell Range Extension
DC Dual Connectivity
DRX Discontinuous Reception
eNB eNode B
HO Handover
LTE Long Term Evolution
MAC Medium Access Control
MCG Master Cell Group
MME Mobility Management Entity
MeNB Master eNB
NAS Non-Access Stratum
NGMN Next Generation Mobile Networks
OPEX Operational Expenditures
PCRF Policy and Charging Rules Function
PDCCH Physical Downlink Control Channel
PSCell Primary SCell
QoS Quality of Service
RAN Radio Access Network
RAT Radio Access Technology
RBG Resource Block Group
RNC Radio Network Controller
Rx Receiver
SCE Small cell enhancement
SCG Secondary Cell Group
SeNB Secondary eNB
SI Study Item
SON Self-organizing network
TCH Traffic Channel
TDD Time Division Duplexing
Tx Transmitter
UE User Equipment
UMTS Universal Mobile Telecommunications System
WI Work item

Page 5 , Total20
 WGC – Communication Architectures and Technologies

1. Purpose
This paper introduces the 3GPP progress of LTE small cell enhancement (SCE) – higher layer aspects study
item (SI) [1] and its corresponding work item (WI): dual connectivity (DC) [2]. The SCE SI is completed in
December 2013, and the corresponding WI was created. Multiple Rx/Tx capable UE in inter-frequency scenario
is prioritized in the WI.

2. Scope
This paper focuses on the architecture and key technologies introduced in LTE DC to improve radio
resource efficiency and mobility robustness. It mainly consists of 5 parts
Chapter 3: motivation
Chapter 4: architecture
Chapter 5: network interface
Chapter 6: control plane handling
Chapter 7: user plane handling

3. Motivation
3.1 Challenges
Small cells are becoming a promising technology to meet ever increasing traffic capacity and data rate demand.
Small cells are typically deployed as hotspots within macro cell coverage. Backhaul between low power nodes
providing small cells and macro nodes may be ideal and non-ideal. Existing intra-eNB CA and CoMP
architectures assume ideal backhaul in which centralized scheduling can be implemented for efficient radio
resource utilization. LTE release 12 SCE SI concerns inter-eNB CA and CoMP operations with non-ideal
backhaul. Inter-eNB CoMP scenario is mainly studied in a separate study item [3]. For inter-eNB CA and
CoMP with non-ideal backhaul, distributed resource allocation and coordination have to be relied upon and new
challenges emerge, e.g.:
 Efficient radio resource utilization across eNBs
Difficult to improve per-user throughput by utilizing radio resources in more than one eNBs with non-ideal
backhaul while taking QoS requirements into account;

 Mobility robustness
Mobility robustness in inter-frequency scenario is not as good as in a macro only network, but less of a problem
than in co-channel scenario if no DRX is used;

 Increased signalling load

Signalling over X2 interface as well as signalling towards the MME and the S-GW is increased with increasing
UE speed due to frequent handover;

 UL/DL imbalance between macro and small cells

UE’s best uplink cell and best downlink cell may be different. It is less of an issue for inter-frequency scenario
than for co-channel case. Cell Range Extension (CRE) may be used for the latter case.

Page 6 , Total20
 WGC – Communication Architectures and Technologies

3.2 Motivations
Considering these new challenges, the SCE SI is mainly based on the follow motivations:
 Increase UE throughput especially for cell edge UEs
LTE DC can significantly increase the UE throughput especially for cell edge UEs by transmitting/receiving
multiple streams and dynamically adapting to the best radio conditions of multiple cells. Small cells provide
additional capacity for UEs having multiple radio connections.

 Mobility robustness enhancement

In heterogeneous network deployment, moving UEs suffers frequent handover failure, inefficient offload and
service interruption due to short Time-of-Stay. The consequences are more severe if UE’s velocity is higher and
cell coverage is smaller. LTE DC can greatly reduce the handover failure rate by maintaining the macro cell
connection as the coverage layer.

 Reducing signaling overhead towards the CN due to frequent handover

This can be achieved by not issuing handover operations as long as the UE is within macro coverage. Signaling
generated towards the CN should be reduced as much as possible if the mobility only happens between small
cells. Appropriate LTE DC architecture is the key for this purpose, e.g. only one S1-MME for the UE with the
MME, as will be discussed later.

3.3 Scenarios
LTE DC can be deployed in the following 3 typical scenarios as depicted in figure 1; the aggregated serving
cells of the involved eNBs can be intra-frequency or inter-frequencies. LTE DC is usually configured for low to
medium mobility speed cases, and support indoor and outdoor, ideal and non-ideal backhaul scenarios. The
focus is on non-ideal backhaul case.
 Scenario 1: co-channel scenario
 Scenario 2: inter-frequency scenario
 Scenario 3: small cell out of coverage of macro cell scenario

f1 f1
f1 f2

Scenario 1: co-channel scenario Scenario 2: inter-frequency scenario

f1

F1 or f2

Scenario 3: small cell out of coverage of macro cell scenario

Figure 1: LTE DC typical scenarios

Page 7 , Total20
 WGC – Communication Architectures and Technologies

4. Architecture
In LTE DC, the UE can receive/transmit data from/to multiple eNBs. There is a Master eNB (MeNB) and one or
more Secondary eNBs (SeNB). In LTE release 12 specifications, only the case of one MeNB and one SeNB is
considered.

4.1 Control plane architecture

There is only one S1-MME connection per UE and it is terminated at the MeNB. The main purpose is to reduce
signaling overhead towards the CN in case of SeNB change. The UE RRC connection is terminated only at the
MeNB to reduce RRM and signaling complexity. There is no RRC entity in the SeNB. SeNB related RRC
configurations are transmitted to the MeNB in the form of RRC container. MeNB makes final decision and
constructs the eventual RRC configuration message and transmits it to the UE. The UE always maintains RRC
connection with the MeNB as long it is under the macro cell coverage. This kind of control plane and user plane
split architecture can avoid frequent handover and reduce handover failure rate. RRC diversity, i.e. both the
MeNB and the SeNB may transmit RRC messages to the UE to improve signaling robustness, is discussed in
3GPP however not adopted in LTE release 12 due to its potential complexity. The control plane architecture is
depicted in Figure 2.

MME

S1-MME

MeNB X2
SeNB
RRC

Uu Uu
UE
RRC

Figure 2: control plane architecture

4.2 User plane architecture

There are many possibilities of user plane architecture. User traffic may split at the S-GW or at the MeNB. Only
bearer level split is possible in the S-GW where different E-RABs for the UE are routed by the MeNB and the
SeNB respectively. Both the MeNB and the SeNB have S1-U connection with the S-GW for the UE. Both
Bearer level and packet level split can be supportable if split takes place at the MeNB, and only one S1-U
interface via the MeNB is needed. For the bearer level split, all user data of a radio bearer are routed between
the SeNB and the UE. For the packet level split, user data may be routed between MeNB or SeNB and the UE.
User data may be split as IP packets i.e. PDCP SDUs, PDCP PDUs, RLC PDUs etc. It is difficult to be split as
MAC PDUs due to non-ideal backhaul and that MAC PDUs are generated in real time according to radio
conditions at that TTI. 3GPP Rel-12 adopts the user plane architecture of bearer level split at the S-GW and
PDCP PDUs level split at the MeNB, i.e. UP architecture 1A and 3C [4]. The architectures are depicted in figure
3 and figure 4 respectively. In Figure 4, it is also possible that all PDCP PDUs of bearer 2 are routed by the
SeNB and in Rel-12, only RLC-AM type bearer split is considered.

Page 8 , Total20
 WGC – Communication Architectures and Technologies

Radio bearer 1

UE MeNB S1-bearer 1
1 2 3 4 1 2 3 4

1 2 3 4

1 S-GW
X2
2

3 SeNB
Radio bearer 2
4

S1-bearer 2

Figure 3: UP 1A - Bearer level split at S-GW

Radio bearer 1
UE MeNB S1-bearer 1
1 2 3 4 1 2 3 4
S-GW
1 3 S1-bearer 2
1 2 3 4

Radio bearer 2
4

X2
2

SeNB
Radio bearer 2 4

Figure 4: UP 3C - Packet level split as PDCP PDUs at MeNB

Table 1 shows pros and cons for the above user plane architectures.
Table 1 Comparison of user plane architectures

Alt1: UP 1A Alt2: UP 3C
Pros  MeNB needs not buffer traffic for  SeNB mobility hidden to CN
an EPS bearer of SeNB  no security impacts with ciphering
 limited impact on PDCP/RLC and being required in MeNB only
GTP-U/UDP/IP  no data forwarding between SeNBs
 low requirements on the backhaul required at SeNB change
link between MeNB and SeNB
Cons  SeNB mobility visible to CN  need to route, process and buffer all
 security impacts due to ciphering dual connectivity traffic in MeNB
being required in both MeNB and  PDCP reordering needs to be always on
SeNB  flow control required between MeNB
and SeNB

Page 9 , Total20
 WGC – Communication Architectures and Technologies

4.3 Backhaul aspects

A typical backhaul deployment scenario is that user traffic of the MeNB and the SeNB are aggregated and split
at a higher level site, e.g. at a router between EPC and eNBs as depicted in figure 5. Backhaul capacity
requirement of SeNB does not impact that of MeNB for UP 1A, but imposes additional requirement on that of
MeNB for UP 3C. For example, the aggregation site needs to route forth and back from the MeNB for the user
traffic destined to the SeNB.

SeNB SeNB

④
MeNB ③

②
MeNB routing
Aggregation
site
SGW Routing

EPC

Figure 5: Higher level aggregation site than MeNB

Another typical backhaul deployment scenario is that user traffic of the MeNB and the SeNB are aggregated and
split at the MeNB, as depicted in figure 6. In this scenario, the impact on MeNB backhaul by backhaul capacity
requirement of SeNB is the same for both the UP 1A and UP 3C.

SeNB SeNB
SeNB

MeNB routing

MeNB
SGW Routing

EPC

Figure 6: Aggregation site is the MeNB

Page 10 , Total20
 WGC – Communication Architectures and Technologies

5. Network interface
Some modifications to exiting S1 interface and X2 interface are needed to support DC.

5.1 S1 interface
In LTE DC architectures, there is only one S1-MME interface between the MeNB and the MME. Depending on
UP 1A or UP 3C, two S1-U interfaces or one S1-U interface are needed. For both UP 1A and UP 3C, S1-MME
is not visible to the SeNB and the MME is not aware of the presence of SeNB. For E-RAB management
function for UP 1A, the MeNB needs to initiate path switch procedure when E-RAB establishment or
modification signaling is received from the MME. The MeNB decides which E-RABs are to be established in
the SeNB and requests the MME to issue path switch operation. If legacy path switch procedure is used, an
indication should be included to inform the MME that only some of the bearers need to be switched. In order
not to impact existing specifications too much, a new procedure for partial path switch is considered, as depicted
in figure 7. For UP 3C the MeNB provides its own GTP endpoint addresses for the bearers split into the SeNB
so that the SeNB can transfer uplink PDCP PDUs according to the address. In the initial context setup procedure
and the E-RAB establishment procedure, the MeNB also indicates the MME the downlink TNL information of
the MeNB or both the MeNB and the SeNB depending on whether it is UP 3C or UP 1A.

MME MeNB
E-RAB MODIFICATION INDICATION

E-RAB MODIFICATION CONFIRM

E-RAB MODIFICATION FAILURE

Figure 7: E-RAB modification for partial path switch

5.2 X2 interface
Existing X2 interface needs to be extended to support DC. X2 user plane new functions are responsible for
PDCP PDU transmissions for UP 3C and data forwarding for UP 1A upon SeNB change or MeNB handover.
SeNB provides feedback on PDCP PDUs delivery status e.g. the highest successfully transmitted PDCP PDU
SN (sequence number) to the MeNB. Packet loss over X2 also needs to be aware by the MeNB, X2-U SN is
introduced so that SeNB can report to MeNB packet loss over the air and X2 separately. The new X2-AP
functions for DC are used to establish, modify and release UE context at the SeNB:
SeNB (SCG) Addition: used to add radio resources of a SeNB, initiated by the MeNB.
SeNB (SCG) Modification: used to modify radio resource configuration of SeNB, may be triggered by the
MeNB or the SeNB.
SeNB (SCG) Release: used to release UE specific context at the SeNB, may be triggered by the MeNB or the
SeNB.
SeNB (SCG) Change: This procedure can be realized by SeNB Addition/Modification procedure in which the
source SeNB is released and the target SeNB is added.
Flow control function over X2 is beneficial for UP 3C to have more efficient rate control across eNBs and avoid
congestion in the SeNB.

X2 signalings to support mobility operations in LTE DC are needed. SeNB change while keeping the MeNB
can be implemented by the combination of SeNB addition and SeNB release. Downlink data forwarding from
source SeNB to MeNB is needed for UP 1A, but not necessary for UP 3C because the MeNB can maintain
copies of all packets transmitted to SeNB. Uplink data forwarding is needed for both architectures. Path switch

Page 11 , Total20
 WGC – Communication Architectures and Technologies

related signalings are needed for both UP 1A and UP 3C.

MeNB change can reuse existing handover procedure with slight updates and resources of SeNBs are released
in the MeNB change procedure. SN status transfer and data forwarding are performed from the SeNB to the
source MeNB, and then from the source MeNB to the target MeNB.

5.3 Flow control

For UP 3C, flow control between MeNB and SeNB is needed to avoid buffer overflow and packet losses at the
SeNB. Mechanisms similar to existing flow control in UMTS Iub interface can be considered. SeNB can grant
certain amount of capacity/buffer size to the MeNB periodically or based on the radio resource monitoring
status. SeNB can indicate the buffer size on per UE basis or per bearer basis. MeNB may also be able to request
SeNB for capacity allocation. New GTP-U extension header is introduced for the inter-eNB RRC container to
transfer flow control information..

6. Control plane handling

6.1 eNB Sychronization
It is not required for the MeNB and SeNB to be SFN synchronized, not like the carrier aggregation (CA) case in
which the aggregated serving cells from the same eNB are strictly SFN schronized. The UE may only need to
obtain SFN of one special cell in the SeNB, i.e. the always activated cell with configured PUCCH. To obtain the
SFN information, the UE may read MIB of this special cell, or the MeNB informs SFN offset of the special cell
to the UE by dedicated signaling.

6.2 Security handling

In LTE DC, the data transmitted via X2 interface and the new Uu interface between the SeNB and the UE
should encryted. For UP 1A, the user plane encryption key in the MeNB and the encryption key in the SeNB
will be cryptographically separated. MeNB derives a base-key (S-KeNB) from the currently active KeNB and
sends it to SeNB over X2. The SeNB derives its user plane encryption key from the received S-KeNB.

6.3 System information acquisition

In Rel-10 and Rel-11 CA technology, system information of all SCells is configured to the UE by dedicated
RRC signaling. This applies to initial system acquisition and system information update. In LTE DC, similar
mechanism is reused. System information of small cells on SeNBs is firstly transferred to the MeNB in the RRC
container via X2 interface and then configured to the UE by RRC signaling, as depicted in figure 8. This may
contains SFN offset between the MeNB and the SeNB if they are not synchronized. Ambiguous time due to
backhaul latency can be avoided by an indication of configuration complete information from the MeNB to the
SeNB. Other means are also possible, e.g. UE acquires MIB and potentially also SIs of a special cell with
configured PUCCH on the SeNB. However, additional complexity or requirements will be introduced and it
may make these other means unnecessary.

Page 12 , Total20
 WGC – Communication Architectures and Technologies

UE MeNB SeNB
X2-AP: RRC Container with SIs of
RRCConnectionReconfiguration cells in SeNB
(including SIs of cells in SeNB)

RRCConnectionReconfigurationComplete
X2-AP:
RRCConnectionReconfigurationComplete

Figure 8: SeNB system information acquisition

6.4 SeNB radio resource management

The SeNB is primarily responsible for allocating radio resources of its own cells. At least one special cell in the
SCG has configured uplink with PUCCH resources. Inter-eNB coordination is needed for UE capabilities,
cell/bearer management, QoS requirements, etc. Information is exchanged through RRC containers carried in
X2 messages and both eNBs can understand each other’s radio resource configurations. The MeNB does not
modify the contents in the RRC container of the SeNB.
Generally, the following procedures can be used:

SeNB (SCG) addition/modification procedure:

Figure 9 shows this procedure.
- The MeNB sends the SCG Addition/Modification Indication message including the MeNB (MCG)
configuration.
- The SeNB responds with SCG Addition/Modification Request messages including its radio resources
configurations in the RRC container. SeNB also allocates an independent C-RNTI for SCG.
- The MeNB sends the RRCConnectionReconfiguration message to the UE including configurations of both
the MCG and the SCG.
- The UE applies the new configuration and reply the RRCConnectionReconfigurationComplete message to
the MeNB.
- The MeNB replies the SeNB SCG Addition/Modification Confirm message to the SeNB forwarding
RRCConnectionReconfigurationComplete message.
- The UE may perform random access procedure for uplink synchronization with the SeNB.
SeNB can also independently initiate the SCG Modification Request message to perform configuration changes
of the SCG.

UE MeNB SeNB
SCG ADDITION/MODIFICATION INDICATION

SCG ADDITION/MODIFICATION REQUEST

RRCConnectionReconfiguration

RRCConnectionReconfigurationComplete
SCG ADDITION/MODIFICATION CONFIRM

Random access procedure

Figure 9: SCG Addition/Modification procedure

SeNB (SCG) Release procedure:

Page 13 , Total20
 WGC – Communication Architectures and Technologies

The SCG release procedure can be realized by the SCG Modification procedure. The
RRCConnectionReconfiguration message comprises the release of the SCG. This part of message may be
generated by the MeNB or the SeNB.

SeNB (SCG) Change procedure:

The SCG change procedure can be realized by the SCG Addition/Modification procedure. The
RRCConnectionReconfiguration message comprises the release of the source SCG and the addition of the target
SCG.

6.5 Selection of connection at the UE

In release 12 LTE DC, it is the MeNB who decide which SeNB as additional connection for the UE based on
UE’s measurement report. In future releases, it is a potential candidate or migration path to have the UE to
decide which SeNB to connect with. In choosing which eNB to conect to in the small-cell environment, UE
stands in a unique position to make a decision. Given the heterogity of the small-cell both in terms of size
(macro/pico/femto), and access technology, such decision could depend on complexset of data. The quality of
wireless signal, can be best measured at the UE, the battery uaseg of UE is an important point to consider, and
the running application and the device spec have significant effect on the perception of quality of the connection
at the UE. In this line, UE can employ range of Multi Attribute Decision Making processes to select the eNB(s)
to connect to, not only based on signal strenght, but number of rather more important criteria.

6.5.1 Use of Analytics

Having UE in control of the connectivity in the LTE DC, the network congestion information could
significantly affect the connection experience since performance of good quality wireless channel could be
degraded by a congested backhaul link, or the access. To this end, cell-level congestion based on historical
analysis of weekley or daily trends could help the UE in choosing the right eNB(s) for its conection. Such
analytics can be delivered to the UE through standrad S2c interface.

6.6 UE capability handling

In LTE DC, the MeNB and the SeNB schedule the same UE independently. UE radio access capabilities may be
exceeded if both eNBs exert their radio resources to the UE simultaneously without any coordination. These UE
capabilities mainly concerns UE specific RF parameters defined by UE category and they may need to be split
between the MeNB and the SeNB, e.g., Maximum number of DL-SCH transport block bits received within a
TTI, Total number of DL-SCH soft channel bits, Maximum number of UL-SCH transport block bits transmitted
within a TTI, etc. In addition to the UE capability parameters, the QoS parameter UE-AMBR should also be
considered. It is the maximum aggregated bit rate of all non-GBR bearers of a UE. If a non-GBR bearer is split,
UE-AMBR will be enforced across eNBs. It is difficult to coordinate UE capability distribution in real time due
to latency of non-ideal backhaul. For coordination and split of the UE capability parameters, the MeNB is an
appropriate node to decide the split ratio. MeNB can grant a part of the UE capability i.e., provide UE capability
restrictions to the SeNB. MeNB can obtain SeNB capability information, cell configuration, radio condition,
load conditions etc via X2 interface, it can estimate the SeNB acceptable UE capability constraints. MeNB and
SeNB can also coordinate UE capability split parameters, e.g. SeNB may suggest its acceptable constraints and
informs the MeNB to modify the parameters.

6.7 SeNB radio link failure handling

When UE detects radio link failure (RLF) in the SeNB, e.g. maximum number of RACH attempts failure,
maximum number of RLC PDU retransmissions failure, physical layer link problems etc, it should report the
failure to the MeNB and stops all uplink transmissions. The MeNB may initiate SeNB release or SeNB change
procedure or temporarily disable the SeNB.

Page 14 , Total20
 WGC – Communication Architectures and Technologies

7. User plane handling

7.1 UL-DL split
UL-DL imbalance refers to the situation that the UE’s best uplink cell and best downlink cell are different due
to significant transmission power difference between the macro eNB and the small eNB, as depicted in figure 10.
UL-DL imbalance issue exists in LTE DC and it is less severe in inter-frequency case than in intra-frequency
case because cell edge interference may be significant in the latter case. UL-DL split scheme can reduce cell
edge interference for intra-frequency case and save UE power for both cases.
In this scheme, UE selects the node as uplink to which the UL path loss is lower and selects the node as
downlink from which the downlink reference signal receiving power (RSRP) is higher. The eNB determines the
downlink cell edges and uplink cell edges of the involved cells separately for the UE. Due to backhaul latency,
the physical layer control signals and UCIs may still need to be bound with the same eNB, e.g. small eNB
allocates UL resources for the UE and transmits PHICH in response to the UL transmissions if UE uplink
traffics are targeting the small eNB.
It is observed that good performance can already be achieved with appropriate CRE and ABS configurations in
the co-channel Hetnet deployment. Therefore UL-DL split is deprioritized in SCE SI and may be further studied
in the future.

f1 f1/f2

Figure 10: UL-DL split

7.2 UL bearer split

In UP 3C, UL bearer split introduces some complexity to the handling of buffer status report (BSR), logical
channel prioritization (LCP), power headroom report (PHR), etc. It is possible that the UL bearer is not split
even though the corresponding DL bearer is split. By this means, UL user traffic are only targeted to one
specific eNB and it can be configured by the network. Consequently, BSR except for BSR for RLC status report,
LCP enforcement are only addressed to the configured eNB. However, RLC status reports in response to DL
RLC PDU transmissions are still transmitted to the concerned eNB. DL bearer split is supported in release 12
LTE DC and UL bearer split is postponed to future releases.

7.3 UCI transmission of SeNB

In Rel-10 and Rel-11 CA, PUCCH is configured only on the PCell. Uplink control information (UCI) e.g.
HARQ ACK/NACK, CSI information of SCells are transmitted on PUCCH of the PCell if not piggybacked on
SCell’s own PUSCH. In LTE DC, it is not suitable any more to carry UCIs of SeNB in PUCCH of PCell in
MeNB due to backhaul latency. Therefore, PUCCH is configured on a special SCell of SeNB, i.e. the PSCell.
This PSCell is never deactivated and RACH procedure needs to be initiated upon its initial configuration.

Page 15 , Total20
 WGC – Communication Architectures and Technologies

7.4 PDCP handling

Several challenges exist for PDCP operation in UP 3C: Out-of-order arrival of PDCP PDUs, X2 loss and PDCP
PDU SN gap due to PDCP discard functionality etc. To address these issues, PDCP reordering needs to be
always on, unlike that in legacy LTE system in which PDCP reordering is only needed upon handover and RRC
re-establishment. Timer based reordering can be used to avoid too much latency. This means that PDCP
receiving side is allowed to move on window regardless of the gap upon reordering timer expiry. The loss may
be caused by X2 congestion or PDCP discard at the transmitting side. Based on this scheme, the RLC-UM like
window operation is utilized by PDCP, this can avoid complexity of legacy MRW (move receiving window)
like mechanisms in UMTS if RLC-AM like window is used.

7.5 RLC handling

For UP 1A, a radio bearer is either located on the MeNB or the SeNB. RLC transmissions/receptions will not
occur across eNBs. No special handling is needed. It is also the case for UP 3C. Two logical channels
corresponding to the split bearer are located on the MeNB and the SeNB respectively. A RLC entity only
belongs to the MeNB or the SeNB and it will not operate across eNBs. If UL bearer is not split, RLC status
reports in the uplink are still sent to the corresponding eNB.

7.6 UE MAC entity modeling

In UP 1A, dual MAC entity modeling is feasible because the eNB specific bearers are not multiplexed together
in UE MAC. In UP 3C, both single and dual MAC entity modeling are applicable from the implementation
perspective. The dual MAC entity modeling is adopted not only because a unified modeling is beneficial from
the specification perspective, but also because it is clearer and simpler for the operations of the split bearers, e.g.
logical channel group mapping to the MCG and the SCG, BSR and LCP procedures etc. The serving cells
associated with the MeNB are configured as a MCG (master cell groups), and the serving cells associated with
the SeNB are configured as a SCG (secondary cell group). The dual MAC entities of the UE are corresponding
to the MCG and the SCG respectively.

7.7 Random access

In legacy CA, only one random access procedure is ongoing at any point in time. UE has to monitor PDCCH on
PCell during RAR (random access response) window of PCell to receive RAR of random access procedure for a
SCell. Contention based RA is not support for SCell.
In LTE DC, existing rules may not be that reasonable any more due to backhaul latency. MeNB or SeNB may
not be aware that a RA procedure is being ongoing when it initiates a new RA procedure. Contention based RA
procedure may be ongoing when an eNB initiate a contention free RA procedure. If UE ignores one of the RA
attempts, unnecessary latency will be introduced. Therefore support of parallel RA procedure has some
qualifications. Contention based RA for SeNB is also beneficial to reduce latency, e.g. in case of high priority
uplink data arrival in SeNB specific bearers or maximum number of SR (scheduling request) transmissions
failure. Contention free based RA for SeNB can be triggered by DL data arrival and addition of the first special
SCell with PUCCH.
During the parallel RA procedure, RA preamble transmissions and/or msg3 transmissions may collide or not. In
case of colliding, power limitation issue may occur, and some priority rules may be needed, e.g. prioritize
PDCCH triggered RA or RA on MeNB etc. In release 12 LTE DC, parallel RA procedure is only supported in
case the preamble transmissions do not collide. For RA procedure in the SeNB, UE needs to monitor PDCCH of
SCell on the SeNB to receive RAR which requires support of PDCCH common search space for SeNB. This
will increase number of UE blind detections.

Page 16 , Total20
 WGC – Communication Architectures and Technologies

7.8 BSR procedure

It is natural that buffer status of eNB specific bearers is only reported to the corresponding eNB in UP 1A. For a
split bearer in UP 3C, two corresponding logical channels locate on the MeNB and the SeNB respectively. RLC
entities are logical channel specific so that buffer status in RLC is reported to the corresponding eNB for both
logical channels of the split bearer. Buffer in the PDCP entity is shared by both logical channels and in order to
avoid double counting, a split ratio needs to be configured. The UE calculates PDCP buffer size in the BSR for
the MeNB and the SeNB according to the configured ratio, as shown in figure 11. When overall data amount is
small, it may not be efficient to be scheduled by both eNBs. A threshold can be configured to the UE such that
UE only needs to report BSRs to both eNBs when the threshold is crossed. UE can report BSR to an configured
eNB when data amount is less than this threshold. In release 12 LTE DC, the UL bearer does not split and the
MeNB will configure the UE to which eNB the PDCP PDUs will be delivered, so that the BSR for data packets
is only reported to the configured eNB. For the other eNB, UE only transmits BSR containing buffer size of
reverse RLC status report for the split DL bearer.

UE
RB1 RB2

PDCP PDCP

RLC RLC RLC

MAC MAC
Correspond to Correspond to
MeNB SeNB

LCG1 LCG2
Portion for Portion for
LCG1 LCG2
Figure 11: buffer status report

7.9 LCP procedure

In legacy LTE releases, token bucket algorithm is used for logical channel prioritization (LCP) procedure. In UP
1A, it is natural that independent LCP procedures per eNB can be applied, as the bearers are eNB specific.
However, there is some correlation between LCP for the MeNB and for the SeNB for a split bearer in UP 3C.
Two kinds of token bucket modeling can be considered: common token bucket modeling and separate token
bucket modeling. For both models, the PBR (prioritized bit rate) of the original split bearer needs to be divided
by a configured ratio between the corresponding logical channels, otherwise the PBR of a split bearer may be
enforced twice which results in unfair resource allocation.
In the common token bucket modeling, two logical channels of the split bearer share the same token bucket and
the bucket size reuses that for the original bearer. Tokens are increased for each TTI at the rate of PBR of the
original bearer, and decreased the sizes of each MAC SDU generated for the MeNB and for the SeNB. A
sequential PBR enforcement mechanism may be needed because the common variable Bj for real time tokens
can only be decreased mutually exclusive by the MeNB and the SeNB in any TTIs. Figure 12 depicts such
common token bucket modeling.

Page 17 , Total20
 WGC – Communication Architectures and Technologies

Bj = Bj + (PBR X TTI)

PBR X BSD

LCH 1 of split bearer

Bj = Bj – MAC SDU size
MeNB

Factor1 X Bj

LCH 2 of split bearer

Bj = Bj – MAC SDU size
SeNB

Factor2 X Bj
Figure 12: common token bucket modeling

In the separate token bucket modeling, each logical channel maintains its own token bucket independently. The
increasing rate of tokens for each bucket is according to its split rate of PBR, tokens are decreased by the size of
MAC SDUs generated for the corresponding eNB independently. This model supports parallel PBR
enforcement operation.
For both token bucket modeling, ambiguous resource utilization still exists from the eNB perspective. For
example, MeNB is not aware of the amount of MAC SDUs UE generates using radio resources of SeNB, and
MeNB may still allocate radio resources for the served data until an updated BSR is received. Further
optimization may be considered by utilizing coordination between eNBs or indication from the UE.
In release 12 LTE DC, the UL bearer does not split and the separate token bucket modeling is adopted.
Starvation of RLC status report may occur if common token bucket model is utilized. The common tokens may
be exhausted for multiple consecutive subframes by the logical channels uplink towards the configured eNB for
which PDCP data are delivered.

7.10 PHR procedure

In CA, the power headroom report (PHR) contains PH information and cell Pcmax information of the PCell and
all the activated Scells. If simultaneous PUCCH and PUSCH transmission is configured, type II PH of the PCell
shall also be included. Type II PH concerns the power headroom considering power consumed by the PUCCH
and the PUSCH altogether.
In LTE DC, PHR to either eNB still includes PH and cell Pcmax information of the PCell and all the activated
SCells. Type II PH of PSCell of SeNB also needs to be included because there is an additional PUCCH in the
SeNB. To reduce probability of too aggressive or too pessimistic resource allocation situation, PH and cell
Pcmax information of cells in the other eNB is useful for the eNB who received the PHR to evaluate the
pathloss and resource utilization status of the other eNB. This requires PHR be conveyed to both eNBs if it is
triggered according to triggering conditions of either eNB. The radio conditions and pathloss variation
characteristics may be quite different for cells of the MeNB and cells of the SeNB, hence the different PHR
related parameters can be configured for the MeNB and the SeNB. PHR can be triggered independently for the
MeNB and the SeNB, and once triggered for either of the eNBs, it can be transmitted to both eNBs or only one
of the eNBs depending on whether UL resources are available. For the PHR of the activated cells belonging to
another CG/eNB, UE is configured to report always virtual PH or actual PH when there is a PUCCH/PUSCH
transmission for a cell in the other CG, otherwise virtual PH. For UE’s uplink transmission towards MeNB and
SeNB, MeNB decides the minimum guaranteed power allocation P_MeNB and P_SeNB, and informs SeNB
these parameters via inter-eNB RRC container.

Page 18 , Total20
 WGC – Communication Architectures and Technologies

7.11 DRX
It is difficult to inherit common DRX principle of CA for all serving cells due to backhaul latency. The MeNB
and the SeNB cannot obtain each other’s cell status in time. Per eNB common DRX can still be used in which
UE follows common DRX operation for all cells of the same eNB. It is still possible to align DRX on duration
time, i.e. the point in time when UE re-enters active time from the inactive period. This can be achieved by
exchange of the DRX parameters configuration between the MeNB and the SeNB.

7.12 Multi-Path TCP support

Since single application’s data can be delivered through multiple connection, and more specifically the UL and
DL may have been splited over multiple connections, there is a need to support multiple latency and out of order
packet delivery. One of the best solution that can deal with such cases is multipath TCP. While the support for
multipath TCP is available in today’s stack, linking such multipath flows to the dual wireless connectivity is
missing, and is important to be considered in the future releases of dual connectivity in LTE.

7.13 Activation/deactivation
SCells are allowed to be activated / deactivated by the corresponding eNB. There will be no activation /
deactivation operation across eNBs since the cell status cannot be known in time by the concerned eNB due to
backhaul latency if its cells are cross activated / deactivated by the other eNB. There is a special cell with
PUCCH in the SeNB which will always be activated upon configuration.

8. Potential enhancements in future releases

This white paper provides a brief introduction of the LTE DC WI in 3GPP. In future releases of LTE, some
potential enhancements may be further expected:
- Support UL bearer split
- Introduction of small cell controller
- Support LIPA/SIPTO in dual connectivity
- Support of CSG in dual connectivity
- Enhancement of Location Information Reporting for dual connectivity
- RRC diversity
- Small cell deployed on unlicensed spectrum or licensed shared access spectrum
- Support of more connections, e.g. three or more
- Combination with multi-path TCP
The forthcoming 5G study items may start in 3GPP from release 14. RAN architecture will be further evolved,
e.g. small cells will be ultra densely deployed. UE may have multiple connection paths towards the RAN, e.g.
connections with multiple antennas of a centralized RAN, or with multiple small eNBs. The backhauls are also
wireless e.g. in millimeter wave band, and may be shared with access links. The air interface will be more
flexible and scalable, e.g. support of multiple different air interfaces in a single cell with filtered OFDM
techniques. The air interfaces are adapted to different service requirements in a unified infrastructure, e.g.
different frame structure and multiple access schemes for H2H communications and industrial M2M
communications which needs ultra low latency and very high reliability.

9. Summary
LTE DC is a technology to extend CA and CoMP to inter-eNB with non-ideal backhaul, LTE release 12 focuses
on inter-eNB CA case for per UE throughput improvement and mobility robustness. The gain of throughput and
mobility comes from multiple streams and maintaining single RRC connection. The system architecture,
mobility, inter-eNB RRM coordination, user plane handlings etc are the key technical issues.

Page 19 , Total20
 WGC – Communication Architectures and Technologies

10. References

[1] RP-122033: "New Study Item Description: Small Cell enhancements for E-UTRA and E-UTRAN –
Higher layer aspects".
[2] RP-132069: "New Work Item Description: Dual Connectivity for LTE".
[3] RP-130847: "New Study Item Proposal: Study on CoMP for LTE with Non-Ideal Backhaul".
[4] TR 36.842: "Study on Small Cell Enhancements for E-UTRA and E-UTRAN – Higher layer aspects".
[5] R2-144660: “Introduction of Dual Connectivity”, LTE stage 2 CR for TS 36.300

Page 20 , Total20
Imprint
Wireless World Research Forum
c/o Format A AG
Pfingstweidstrasse 102b
CH-8005 Zürich

Secretariat:
Vinod Kumar
Alcatel-Lucent France
Centre de Villarceaux
Route de Villejuste
91 620, NOZAY
France
e-Mail: [email protected]
Phone : + 33 1 30 77 27 37
Fax : + 33 1 30 77 61 75

The WWRF is a non-profit organisation registered in Switzerland

Chairman of the Forum:

Dr. Nigel Jefferies

Editor-in-Chief: Mr Sudhir Dixit

The WWRF Outlook Visions and research directions for the Wireless World

ISSN 1662-615X is published non-periodically by the Wireless World Research Forum

http://www.wwrf.ch

Responsibility for the contents rests with the Steering Board of the Forum.