WWRF21-WG4-07 C Hoymann, A Rácz, N Johansson, J Lundsjö - WWRF 21-WG4 Stockholm, Sweden, 2008 - 1

wireless-world-research.org

A Self-backhauling Solution for LTE-Advanced

Christian Hoymann, András Rácz, Niklas Johansson, Johan Lundsjö
Ericsson Research

support combinations of aggregated carriers with various

Abstract— After years of research, multi-hop technology bandwidths.
seems to be ready for introduction in cellular technology Spectrum aggregation: Spectrum aggregation is seen as a
standards. In 3GPP various multi-hop solutions are being
discussed for potential inclusion in Long-Term Evolution (LTE)
generalization of carrier aggregation such that non-adjacent
Advanced, a coming release of the LTE standard to meet the carriers are aggregated there. This technique allows
requirements of IMT-Advanced as defined by the International leveraging multiple carriers, such as carriers already
Telecommunications Union (ITU). In this paper we discuss allocated, carriers newly allocated and even re-farming of
various options for inclusion of multi-hop capabilities in LTE- spectrum used for other technologies. These bands are, in
Advanced with particular focus on self-backhauling, a layer 3 general, scattered in the frequency domain.
multi-hop solution for cost efficient backhauling of base
stations.
Multi antennas: Multi antenna techniques are a key
component of LTE and they are further enhanced in LTE
I. INTRODUCTION Advanced. Spatial multiplexing increases the peak spectral
efficiency. In uplink, spatial multiplexing with up to four
Multi-hop communication claims to offer cost efficient
simultaneous data streams is a strong candidate for LTE
coverage extension and capacity increase. Hence it has been
Advanced. In downlink, higher-order spatial multiplexing is
a hot research topic over decades and multi-hop
considered as well. Since beamforming can increase cell
functionality has been integrated in all kinds of wireless
edge user throughput, a combined beamforming / spatial
systems ranging from system concepts and prototypes
multiplexing approach is recommended to be introduced,
(WINNER) [12][13], over standards for local area networks
too.
(IEEE 802.11s)[8][10] and metropolitan area networks
Coordinated multipoint transmission: Coordinated
(IEEE 802.16j) [7][9].
multipoint transmission is a technology where transmission
Apart from simple repeaters, multi-hop communication
and/or reception is coordinated across several
has not yet been part of commercial cellular networks from
geographically separated points. In uplink the support for
the 3GPP or 3GPP2 technology families. However, 3GPP is
joint processing of signals received at multiple
discussing multi-hop as one technical component of a
geographically separated points is relatively straight forward
coming release of Long-Term Evolution (LTE) [1][4][5]
and the impact on the LTE radio interface specification is
called LTE-Advanced [3] the 3GPP proposal for IMT-
expected to be marginal. In downlink, coordinated
Advanced.
multipoint transmission can have different flavors. It ranges
This paper presents how a multi-hop enabled LTE-
from dynamically coordinated scheduling across separate
Advanced system might look like, leveraging on the large
points to joint transmissions from multiple geographically
amount of research on multi-hop concepts that has been
separate points. The impact on the radio interface
conducted. First, it provides a brief overview of relevant
specification depends on the selected approach, but is
technology components that have been proposed for LTE-
expected to be higher than in uplink.
Advanced, in particular different options for how to extend
Multihop functionality: Multihop functionality is claimed
the presented LTE functionality with multi-hop capabilities.
to offer cost efficient coverage extension and/or capacity
Second, it details on one of these options, namely layer 3
increase [11]. Different approaches, e.g., Layer 1 repeater,
self-backhauling.
Layer 2 relay, and Layer 3 self-backhauling will be outlined
in section III. As main topic of this paper, Layer 3 self-
II. LTE-ADVANCED
backhauling is presented in detail in sections IV and V.
In March 2008 3GPP started to discuss the evolution of
LTE, named LTE-Advanced. LTE-Advanced targets to meet III. CLASSIFICATION OF MULTI-HOP SCHEMES
the requirements of the International Telecommunications
Numerous multi-hop schemes have been proposed in the
Union (ITU) for next generation mobile systems, named
literature. One way of classifying them is according to the
IMT-Advanced. Several technology components have been
protocol layer at which user plane data forwarding takes
discussed [2][6], including:
place. We then arrive at three main categories of multi-hop
Wider bandwidth: The widest carrier bandwidth specified
schemes of relevance for LTE-Advanced, namely Layer 1
for LTE is 20 MHz. The extension to wider bandwidths is
repeater, Layer 2 relay, and L3 wireless router. The self-
part of the LTE evolution towards LTE Advanced since
backhauling solution described in the following chapter
spectrum allocated might have carriers larger than 20 MHz.
belongs to the third category, L3 wireless router.
Considering spectrum compatibility with legacy LTE, carrier
aggregation is the natural choice. Such approach should
WWRF21-WG4-07
A. Layer 1 Repeater
A repeater receives a certain signal, amplifies it and
transmits it again (aka amplify and forward). User plane
forwarding to and from the User Equipments (UEs) is
performed on Layer 1, i.e., the physical layer. However,
control signaling may possibly occur up to Layer 3, so the
term Layer 1 repeater really only refer to the user plane.
A repeater typically introduces very little delay compared
to other multi-hop solutions operating on higher layers. At
the same time, a repeater can not differentiate between
received desired signals and received noise and interference
since no decoding operation is performed in the repeater.
Both noise and desired signal are amplified and forwarded
and therefore the repeater can not improve the SINR from
input to output.
A repeater can either be frequency translating, where the
repeated signal is transmitted on a different carrier frequency
relative to the received signal, or on-frequency operating,
where received and transmitted signal are on the same
carrier. In case of an on-frequency repeater, the repeated
signal and any direct signal will add like channel multipath
in the receiver. On-frequency repeaters typically need some
form of self-interference cancellation functionality.
Conceptually a simple repeater can be thought of as an
analog Power Amplier (PA). However, a repeater could
also be more advanced and e.g. contain a controllable bank
of band-pass filters, perform measurements, transmit
reference signals etc. An even more advanced repeater could
consist of several receive and transmit antennas enabling
multi-stream signal repetition.
B. Layer 2 Relay
A layer 2 relay forwards user plane traffic on Layer 2. As
for the Layer 1 repeater, control signaling may still occur on
higher layers. As the relay node decodes, re-encodes, and
forwards received data blocks a delay is introduced.
However, no noise is forwarded by the relay node and rate
adaptation may be performed individually for each link. The
direct and the relayed signal interfere so that relay and
eNodeB transmissions have to be separated, e.g., by time or
frequency multiplexing.
Many different technology options exist for a layer 2 relay
solution. Layer 2 protocols, such as MAC, RLC, PDCP can
operate either end-to-end or on a per-hop basis. Control
functionality, such as resource allocation and system
broadcast, can be performed centrally by the eNodeB or it
can be distributed amongst relays. Control signaling may be
placed on layer 2 and/or layer 3.
Common to all layer 2 technology options is that they will
require rather large standard changes. For example, not
terminating PDCP (the “highest” L2 sub-layer of LTE) in
the relay node would mean that the in-order delivery service
of PDCP function would have to be replaced by something
new, including the associated control signaling between
relay nodes and eNodeBs at handover.
The European research project Wireless World Initiative
New Radio (WINNER) investigated Layer 2 relaying as a
key technology that can be applied to enhance coverage and
capacity of a base station [12]. Performance evaluation
showed the potential to improve coverage, especially to
2
overcome uplink power limitations. However, capacity
improvement in terms of spectral efficiency could only be
realized by intelligent and dynamic resource partitioning and
reuse schemes. An additional cost evaluation showed that
the introduction of relays has a positive impact on capital
and operational expenditures provided that the cost of a
relay is less than approximately 1/3 that of a base station
[13].
C. Layer 3 Wireless Router
A wireless router (a.k.a. Layer 3 relay) forwards IP
packets on Layer 3, i.e., the network layer. A wireless router
has similar capabilities and characteristics as a Layer 2 relay,
such that it does not amplify noise and interference but that
it introduces processing delays. In contrast to a Layer 2
relay, radio protocols (layer 1 and 2) are not affected. In
fact, since a layer 3 wireless router will include full eNodeB
functionality it can be seen as an “ordinary” eNodeB being
backhauled via the same RAT that is used over the access
link towards the UE.
The wireless router provides advantages of decode &
forward (Layer 2) relaying without requiring new network
nodes or modified radio protocols. The characteristics of a
wireless router motivate its usage for self-backhauled
eNodeBs as it will be discussed in the following sections.
IV. SELF-BACKHAULED ENODEB
The network infrastructure that is used to connect
eNodeBs to the core network is an IP-based transport
network, which can comprise of different Layer 1 / Layer 2
technologies, e.g., leased telephone lines, bre optic cables,
Ethernet or microwave links. The type of transport network
and Layer 2 technologies employed is a deployment issue,
depending on the availability, cost, ownership, operator
preferences, etc., of such networks in the particular
deployment scenario. However, in some deployment
scenarios it may be technically difficult, or not cost efficient,
to employ existing transport network technologies. The costs
of the transport network often play a significant part of the
overall operation costs of the network. The need for
alternative solutions might be accelerated when moving
towards extremely dense deployments of eNBs, a target
scenario for LTE-Advanced. For instance, microwave
solutions require line-of-sight deployment, and it may not be
economically viable to deploy optical fibers everywhere.
This is the reason why it could be interesting to use the
LTE radio interface as a backhaul link to connect an
eNodeB (named Self-backhauled eNodeB (sNB)) via
another eNodeB (named Anchor eNodeB (aNB)) to the core
network. We call this method self-backhauling.
UL
core
UL network
Self-backhauled eNB Anchor eNB
Fig. 1. Transmissions on access and backhaul link in UL band
WWRF21-WG4-07
Fig. 2. Self-interference at a self-backhauled eNodeB
The purpose of self-backhauling is to reduce the cost for
the transport network and in turn to provide cost-efcient
backhauling as a complement to other backhaul technology
options.
LTE uses different transmission schemes for Downlink
(DL) and Uplink (UL). In general, an eNodeB transmits DL
signals using OFDMA and it receives UL signals using SC-
FDMA. An sNB whose backhaul trafc is served like
"regular UE trafc" by the aNB requires additional
functionality: such an sNB additionally transmits SC-FDMA
UL signals, see Fig. 1 and it receives OFDMA DL signals,
see Fig. 2. Consequently, an sNB requires UE-like
transceiver capabilities.
A. In-band self-backhauling
In the concept of in-band self-backhauling physical
resources are (dynamically) shared between self-backhauling
and UE trafc, i.e. backhauling is performed inside the
regular spectrum band. Since an sNB transmits and receives
in the same band its transmitted signal interferes with the
received signal, see Fig. 2. So called self-interference occurs
at the sNB in DL as well as in UL. Note that the aNB does
not generate self-interference.
At the sNB, the receive power of its own transit signal is
orders of magnitude higher than the receive power of the
desired UE signal so that the SINR is reduced drastically.
Without additional effort the desired signal cannot be
decoded successfully.
frequency
1 2 3 4 time [ms]
aNB DL schedule
frequency
1 2 3 4 time [ms]
Tx Rx Tx Rx
sNB DL schedule
Fig 4. Example resource allocation of aNB (top) and sNB (bottom)
using coordinated Tx and Rx phases
3
Fig 3. Desired DL transmissions (solid line) and interfering signals
(dotted lines) of access and backhaul link
There are at least two potential technical means to
mitigate self-interference. The first one, coordinated
transmission and reception, means that self-interference at
the sNB is mitigated by time multiplexing the backhaul into
the sNB’s access link, where scheduling of the sNB and aNB
is coordinated. Figure 3 shows the involved DL
transmissions on the access links (aNB-UE and sNB-UE) as
well as the DL transmissions on the backhaul link (aNB-
sNB). Figure 4 shows 4 TTIs of the aNB’s DL schedule
(top) and the corresponding DL schedule of the sNB
(bottom). In the aNB, resource allocations for the backhaul
link (marked with horizontal lines) are embedded into
allocations for regular UEs (marked with vertical lines). All
transmissions are scheduled on orthogonal resources by the
aNB scheduler.
In order to mitigate self-interference, the sNB’s DL
schedule is subdivided into subsequent transmit (Tx) and
receive (Rx) phases. During Tx phases, e.g., 1st and 3rd
TTI, the sNB allocates resources to its own UEs, i.e., the
sNB transmits user data. In the 2nd and 4th TTI the sNB
switches to Rx so that it can receive transmissions on the
backhaul link. During TTIs dedicated to Rx, no UE can be
served by the sNB. An analog scheme needs to be applied in
the UL.
Rx phases need to be negotiated on a long-term basis so
that the aNB can allocate resources and the sNB can switch
to Rx mode. They should occur periodically. The length of
Rx (and Tx) phases can be one or more TTIs. Within the
negotiated Rx phase, the actual resource can be allocated
(channel dependent) anywhere in the time-frequency grid.
The mechanism to negotiate Rx phases can be quite similar
to the Discontinuous Reception (DRX) operation of LTE, a
UE power saving feature.
The second method to reduce self-interference is to
employ a dedicated antenna for self-backhauling. Such an
antenna could be highly directive, e.g. a parabolic antenna. It
could be pointed directly towards the peer eNodeB without
sacricing the downtilt of the eNodeB antenna used for the
access link.
At the aNB a dedicated antenna allows to allocate
backhaul resources independently from UE resources. The
dedicated antenna could be seen as a separate sector that
serves only one UE, which is the sNB. Since physical
resources are no longer shared between backhaul and user
traffic, UE performance is not affected by self-backhauling.
If the separation of backhaul and access antenna at the
sNB is sufcient to suppress self-interference, access and
backhaul are independent at the sNB as well. No Tx and Rx
phases are required and the sNB can continuously serve UEs
WWRF21-WG4-07 4

while doing self-backhauling. If the antenna separation at the frequency

sNB is not sufficient, the sNB need to coordinate
transmission and reception by negotiating Rx phases as
shown above.
The drawback of an extra antenna is the extra cost for
equipment and deployment, the advantage is a higher
capacity for both access and backhaul link.
B. Out-band self-backhauling
In contrast to in-band self-backhauling where access and
backhaul link share the same spectrum band, self-
backhauling could also be performed out-band, i.e., access
and backhaul link operate on separate spectrum bands.
At the aNB the out-band solution leads to independent
resource allocation of both links. With separate carriers, 1 2 3 4 time [ms]
access and backhaul link do not interfere each other at the aNB DL schedule
sNB. The sNB can continuously serve its UEs on one carrier Fig 5. Example resource allocation of an aNB aggregating three
carriers
while performing self-backhauling on the other carrier. Self-
interference does not occur.
The drawback of the out-band solution is the extra V. PROTOCOL ARCHITECTURE
transceiver costs. The advantage is a better performance of
access and backhaul link. Out-band self-backhauling could A. User plane
use spectrum bands which are high up in the radio spectrum Figure 6 shows the user plane protocol stack including the
(frequencies above 3 GHz) and which are therefore not very E-UTRAN and the S1 interface of a conventional, i.e., non-
useful for Non Line-of-Sight transmission. However, such self-backhauled system. The radio access uses the protocols
bands could be used for self-backhauling under Line-of- MAC, RLC and PDCP [5]. The user plane part of the S1
Sight propagation conditions. Self-backhauling could even interface is based on the GPRS Tunneling Protocol (GTP)
utilize unused TDD spectrum. To do so the self-backhauling protocol, which uses a tunneling mechanism ensuring that IP
would be based on LTE TDD while the access would remain packets destined to a given UE are delivered to the eNodeB
FDD. where the UE is currently located. GTP encapsulates the
Spectrum (and carrier) aggregation functionality is original IP packet into an outer IP packet which is addressed
proposed as potential technical component of LTE to the proper eNodeB. The S1 interface can be operated over
Advanced, refer to section II. Such functionality would various Layer 1 / Layer 2 technologies, e.g., fiber optic
allow for a more spectrally efficient performance of out- cables, leased (copper) lines, or microwave links.
band self-backhauling. Figure 5 shows an example DL
resource allocation of an aNB that aggregates three carriers.
In the lower two carriers the aNB allocates resources for UE App. App.
trafc only (vertical lines). On the third subcarrier the aNB TCP/UDP TCP/UDP
IP IP
allocates resources for backhaul trafc (horizontal lines) as GTP-u GTP-u
well. UDP UDP
The sNB uses the carriers differently. The lowest carrier IP IP

is used for the access link only, i.e., the sNB transmits user PDCP PDCP
RLC RLC L2 L2
trafc. The highest carrier is the backhaul carrier, there the MAC MAC

sNB receives backhaul data from the aNB. The middle PHY PHY L1 L1

carrier serves as guard band to avoid self-interference; it is UE eNodeB S-GW

not used by the sNB. Fig 6. User plane protocol stack (E-UTRAN and S1 interface)
With carrier aggregation functionality, the aNB can Transmissions on access and backhaul link in UL band

efficiently utilize resources from all three carriers. Beside

the restriction of allocating backhaul traffic to
certain carriers only, resources can be App. App.
dynamically shared between access and TCP/UDP TCP/UDP
IP IP
backhaul. GTP-u GTP-u GTP-u
conventional
UDP self- UDP UDP
IP packets
backhauling
IP IP IP
IP packets
PDCP PDCP PDCP PDCP
RLC RLC RLC RLC L2 L2
MAC MAC MAC MAC

PHY PHY PHY PHY L1 L1

self-backhauled anchor
UE eNodeB S-GW
eNodeB

Fig 7. Potential user plane protocol stack for self-backhauled eNodeB access and backhaul
link in UL band
WWRF21-WG4-07
S1-AP S1-AP
SCTP SCTP
IP IP
PDCP PDCP
RLC RLC L2
MAC MAC
PHY PHY L1
self-backhauled anchor
eNodeB eNodeB
Fig 8. Potential control plane protocol stack for self-backhauled eNodeBs
Figure 6 also shows an example TCP/IP based
application, such as web browsing. The corresponding peer
entities operate in the UE and at the server hosting the web
application. For simplicity, peer protocol entities of the
server are drawn in the Serving Gateway (S-GW), however,
in general they are located somewhere in the Internet.
One potential approach to integrate an sNB into the user
plane protocol architecture is shown in Fig. 7. There, the
aNB is seen as part of the transport network, acting like a
wireless IP router in-between the sNB and the core network.
In principle IP packets addressed to the sNB are routed via
the aNB based on the sNB IP address. Regular IP routing
mechanisms could be used.
That approach imposes that the aNB has to handle
(conventional) IP packets destined for a UE associated to
that aNB differently than (self-backhauling) IP packets
destined for the sNB (actually destined for UEs associated to
the sNB). The former IP packets are de-capsulated and plain
user IP packets are transmitted over the air. The mapping of
IP packets to radio bearers at the aNB is done based on the
GTP tunnel endpoint. The latter IP packets are directly
routed towards the sNB resulting in GTP-encapsulated IP
packets being transmitted over the air. Here the mapping to
radio bearers cannot rely on GTP tunnel endpoints but, e.g.,
on IP Quality of Service (QoS) mechanisms such as IP
Differentiated Services. As a result the sNB cannot be
controlled like a regular UE by the Mobility Management
Entity (MME). A modified QoS management would have to
be introduced.
B. Control plane
The control plane of the S1 interface connects the
eNodeB and the MME. The corresponding protocol is called
S1-AP and operates on top of the Stream Control
Transmission Protocol (SCTP). The MME uses the S1-AP
protocol to establish, configure and tear down bearers, to
authenticate UEs, to control ciphering of NAS signaling, and
to support UE mobility.
The approach to introduce self-backhauling in the control
plane could look similar to the one on the user plane: IP
packets carrying control plane data for an sNB are routed via
the aNB, which is acting like an IP router in the transport
network. IP packets carrying control plane data for the aNB
are de-capsulated at the aNB. Regular routing protocols can
be applied. The corresponding control plane protocol stack
is depicted in Fig. 8.
Although packet forwarding and routing for the control
plane would be similar to the user plane, the control plane
traffic should be mapped to bearers with higher QoS. The
MME acting as the serving MME for the sNB will have to
configure the radio bearers and set the packet classification
5
rules in the aNB such that control plane traffic is mapped on
S1-AP
SCTP high priority radio bearers. This needs to be configured
IP when the S1 interface between sNB and MME is
established, i.e., at sNB setup/configuration.
L2
L1 VI. CONCLUSIONS
MME Multi-hop capability is likely to become part of the 3GPP
LTE-Advanced standards. We have described different
options of multi-hop forwarding on Layer 1, 2 or 3, detailing
on Layer 3 self-backhauling, a cost-efficient solution for
backhauling of eNodeBs.
Self-backhauling provides major benefits of conventional
Layer 2 relaying such as signal regeneration due to decode
and forward operation, and per-link optimization of the radio
transmissions. Furthermore, self-backhauling avoids major
disadvantages of layer 2 relays: it does not introduce new
network nodes and it requires only little standardization
effort. These characteristics are very important when
considering smooth integration of a new feature into an
existing standard and fast adoption in the commercial
market. Thus we conclude that self-backhauling is a
promising technology component of LTE-Advanced, for
which further research on various optimizations is highly
motivated.
REFERENCES
[1] 3GPP (2006). 3GPP TS 36.300, Technical Specication Group Radio
Access Network; E-UTRA and E-UTRAN; Overall description.
[2] 3GPP (2008b). REV-080060, Report of 3GPP TSG RAN IMT-
Advanced Workshop.
[3] 3GPP TR 36. 814, Technical Specication Group Radio Access
Network; Further Advancements for E-UTRA, Physical Layer
Aspects
[4] Dahlman, Erik, Parkvall, Stefan, Skold, Johan, and Beming, Per (2nd
edition 2008). 3G Evolution: HSPA and LTE for Mobile Broadband.
Academic Press.
[5] Ericsson AB (2007). Long Term Evolution (LTE): an introduction.
White paper.
[6] Ericsson AB (2008). R1-082024, A discussion on some technology
components for LTE-Advanced. Contribution to TSG-RAN WG1
#53.
[7] Hoymann, C., Klagges, K., and Schinnenburg, M. (2006). Multihop
Communication in Relay Enhanced IEEE 802.16 Networks. In
Proceedings of the 17th IEEE PIMRC, page 4, Helsinki, Finland.
[8] IEEE (2004). IEEE Std 802.16-2004, IEEE Standard for Local and
metropolitan area networks, Part 16: Air Interface for Fixed
Broadband Wireless Access Systems.
[9] IEEE (2008a). IEEE 802.16j/D5, Draft Standard for Local and
Metropolitan Area Networks Part 16: Air Interface for Fixed and
Mobile Broadband Wireless Access Systems Multihop Relay
Specication.
[10] IEEE (2008b). IEEE Std 802.16Rev2-D5, IEEE Draft Standard for
Local and metropolitan area networks, Part 16: Air Interface for Fixed
Broadband Wireless Access Systems.
[11] Pabst, R., Walke, B., Schultz, D., and et al. (2004). Relay-Based
Deployment Concepts for Wireless and Mobile Broadband Radio.
IEEE Communications Magazine, pages 80–89.
[12] WINNER, IST-4-027756 (2006). Relaying concepts and supporting
actions in the context of CGs. WINNER deliverable D3.5.1.
[13] WINNER, IST-4-027756 (2007). Final assessment of relaying
concepts for all CGs scenarios under consideration of related
WINNER L1 and L2 protocol functions. WINNER deliverable
D3.5.3.