Practical Aspects of LTE Network Design and

Deployment
Ayman Elnashar
Abstract of fingers in rake receiver, and thus its complexity [8], [9].
Adding multiple-input multiple-output (MIMO) to the
This paper covers the practical aspects of commercial HSPA system on top of the above complexity in receiver
long term evolution (LTE) network design and design limits the anticipated gain from it [8], [9]. 3GPP
deployment. The end-to-end architecture of the LTE adopted other ways of extending HSPA system to
network and different deployment scenarios are presented. broadband systems, based on multi-carrier HSPA with
Moreover, the LTE coverage and link budget aspects are added complexity as well in terms of power amplifier (PA)
discussed in details. Theoretical and practical design, network cost, and network optimization [8], [9].
throughputs of LTE system are analyzed. In addition, OFDM can also be viewed as a multi-carrier system but
capacity dimensioning of LTE system is explained in each subcarrier is usually narrow enough that multipath
details and compared with the evolved high-speed packet channel response is flat over the individual subcarrier
access (HSPA+) system. Additionally, the quality of frequency range, i.e. frequency non-selective channel (flat
service (QoS) of the LTE system and the end-to-end fading) and hence receiver design is very simple. More
implementation scenarios along with testing results are specifically, OFDM symbol time is much larger than the
presented. Finally, the latency of the LTE system is typical channel dispersion [5]. Hence, OFDM is inherently
analyzed and compared with the HSPA+ system. This susceptible to channel dispersion due to multipath
paper can be used as a reference for best practices in LTE propagation. OFDM symbol detection requires that the
network design and deployment. entire symbol duration be free of interference from its
Keywords: LTE, OFDM, OFDMA, SC-FDMA, previous symbols i.e., Inter-Symbol-Interference (ISI) spill-
HSPA+, WCDMA, latency, QoS. over. However, ISI spill-over at the beginning of each
symbol can be tackled by adding a cyclic prefix (CP) to
I. Introduction each transmit symbol [5].
The LTE is the next major evolution step in mobile The aim of this paper is to provide the key practical
radio communication, introduced by 3GPP in Release 8 and aspects of design and deployment of a commercial LTE
provides initially downlink (DL) peak data rates of network. The analysis in this paper is presented in a
100 Mbps, an uplink (UL) data rate of 50 Mbps compared comparative manner with reference to the HSPA+ network
to 42Mbps/11Mbps (DL/UL) peak throughputs of HSPA+ to benchmark and evaluate the LTE network performance.
R8 3GPP standard [1]-[9]. The LTE system brings flat The remaining of the paper is organized as follows: Section
network architecture with improved data rates and less II provides LTE network architecture and typical
latency that provide best mobile data browsing experience implementation scenarios. LTE coverage and link budget
[2]-[6]. The LTE system brings two new multiple access (LB) analysis are presented in Section III. Network
techniques: orthogonal frequency division multiple access dimensioning and design exercise are introduced in Section
(OFDMA) on the DL and the single carrier frequency IV. LTE QoS and practical implementation exercises are
division multiple access (SC-FDMA) on the UL [2]-[6], introduced in Section V. The LTE network latency and a
[10]. Both of these access schemes are based on orthogonal comparison with the HSPA+ network are presented in
frequency division multiplexing (OFDM) transmission section VI. Finally, conclusions and key findings are
technique. The main advantage of OFDM, as is for SC- summarized in section VII.
FDMA, is its robustness against multipath signal
propagation, which makes it suitable for broadband II. LTE Network Architecture
systems, compared to wideband code division multiple
access (WCDMA) technique [5], [6]. The SC-FDMA
LTE system brings flat all IP architecture [2], [3], [11],
brings additional benefit of low peak-to-average power
[12]. This flat architecture offers saving in CAPEX and
ratio (PAPR) compared to the OFDMA technique making it
OPEX thanks to eliminating the radio network controller
suitable for uplink transmission of user equipment (UE) to
(RNC) and the circuit switch (CS) core while introducing
extend the battery backup time [7].
IP multimedia subsystem (IMS) [11], [12]. Moreover, the
Even though, the WCDMA system can be extended to LTE system offers higher network performance and
a broadband system, its complexity increases where it increased efficiency. This is achieved by reducing the
requires more number of rake receiver fingers since its latency since the eNB is directly connected via S1 interface
channel is a frequency selective fading channel [8], [9]. to the EPC and also faster handover thanks to direct
Therefore, extension of WCDMA system with high speed connectivity between eNBs via X2 interface. The EPC
packet access (HSPA) evolution to a 20 MHz broadband consists of the serving gateway (S-GW, SGW, or SAE
system requires extension of similar factor on the number GW), the mobility management entity (MME), and the
packed data network (PDN) gateway (PGW) [11], [12]. The noted that, the EPC nodes can have local redundancy
SGW is responsible for handovers with neighboring eNB, and/or geo-redundancy for high reliability and high
data transfer in terms of all packets across user plane, availability. Also, the home subscriber server (HSS) is a
mobility anchor to other 3GPP systems (i.e., 2G and 3G). new functionality that can be added as a separate entity
The MME is the centralized control unit for key operations or via upgrade to the existing home location register
on access network and core network. The PGW is (HLR). The service profile of each LTE subscriber is
responsible to act as an anchor of mobility between 3GPP defined in the HSS [11], [12].
and non-3GPP technologies [11], [12]. The PGW provides 3- Security domain: in addition to the well known firewall
connectivity from the UE to external packet data networks nodes, two new network elements are introduced as part
by being the point of entry/exit of traffic for the UE. In of the LTE system:
addition, LTE brings the always-on connected concept with i. IPSec Gateway (IPSec GW): it is responsible for
less than 100ms transition from idle state to connected encryption and for terminating the IPSec tunnels
state. This concept tackles the impact of the signaling storm with eNBs. All traffic and signals are encapsulated
generated from smartphones due to the transition from idle inside IPSec tunnel for security purpose and to
state to connected state. Finally, LTE system offers prevent any attack to the EPC nodes. The
increased throughput with 100Mbps/50Mbps DL/UL peak deployment of IPSec GW can be in cluster fashion
throughputs with category 3 (i.e., CAT3) modem and with for reliability and simplicity. Therefore, the network
seamless evolution to 150Mbps/75Mbps peak DL/UL can be divided into several clusters (i.e., regions)
throughputs with the introduction of category 4 (CAT 4) and eNBs traffic within certain cluster is
modem [2], [7]. encapsulated towards an IPSec GW pair with 1+1
hot standby. Furthermore, a feature such as X1-flex
A typical commercial LTE network high-level topology can allow the eNB to create two IPSec tunnels with
is depicted in Figure 2. The network consists of four two different IPSec GW nodes in two clusters (i.e.,
domains as follows: two different regions or locations) for reliability and
1- Access Network Domain: consists of eNBs that in this case the traffic can be distributed based on
provides the evolved universal terrestrial radio access load sharing or hot standby fashions.
(E-UTRA) user plane and control plane protocol ii. Certificate Authority (CA): responsible for allowing
terminations towards the UE and IP access layer that eNB to access the network by issuing, managing,
carry the traffic of eNB towards the core network and validating the digital authentication certificates.
domain. Each eNB may have one S1 connectivity with 4- Operation and management (O&M) domain: this
one EPC node and it can have two S1 connectivities domain includes O&M functionalities and NMS for all
with two EPC nodes if geo-redundancy is adopted network elements.
similar to the scenario in Figure 2. The S1 consists of
user plane and control plane. The S1 user plane (S1-U) III. LTE Link Budget and Coverage Analysis
is routed to the S-GW and the control plane (S1-MME)
is routed to the MME. The control plane capacity is ~ The aim of the link budget (LB) is to identify the
maximum allowable path loss (MAPL) between the
1%-3% of the user plane capacity. Moreover, X2
transmitter and receiver for UL and DL and therefore the
connects neighbor eNBs to support UE handover. The
cell radius can be calculated for different terrain
X2 capacity is estimated to be around 3-5% of S1
morphologies (i.e., dense urban, urban, suburban, and rural)
capacity. As per [3], 10ms delay is recommended for S1
based on the propagation model. A tuned version of
and 20ms for X2. Therefore, if the latency requirement
COST231-Hata model is used to estimate the pathloss. LTE
is met for S1 then, X2 shall be met accordingly. It is
Release 8 is a data centric technology; therefore the critical
important to mention that, X2 connectivity can be made
coverage constraint when designing an LTE network would
in the IP security gateway (IPSec GW), which is part of
the security domain, or in the IP access routers or even be the expected data rate at cell edge rather than the
in the IP core cloud. The selection of X2 cross connect received signal level. The outcome of the LB calculations
enables the network designer to determine the expected
node depends on the latency requirement, network
coverage calculated in theory and compare it with the
topology, and security requirement. The recommended
measured values in the field. Table 1 provides the
node to terminate the X2 is the IPSec GW to meet
theoretical link budgets for typical LTE system at
security requirement, reduce the latency, and maintain
1800MHz band (i.e., 3GPP band 3) with 20MHz channel
consistent network topology. The S1 latency is divided
bandwidth. The link budgets are calculated at different
into two parts; access network delay and transport
morphologies and for UL and DL. The cell radius of each
network delay where the 10ms can be divided 5ms for
terrain is determined based on the smallest cell radius from
each network. The maximum latency of the transport
UL and DL. The theoretical link budgets for all
network limits the number of cascaded transmission
links between the eNB and EPC. morphologies demonstrate that the LTE system is uplink
2- Core network domain: This is the main domain that limited (i.e., MAPL of UL < MAPL of DL) and there is ~3-
4 dB between the MAPLs for UL and DL at typical cell
includes IP core network cloud that carries the access
edge throughputs (i.e., DL at 512Kbps and UL at 128Kbps).
network traffic to core network elements. It is to be
 Figure 1 Typical LTE network end-to-end topology

Table 1: LTE Link Budgets

LTE Link Budget
Morphology Dense Urban Urban Suburb Rural Formulas
Data Channel Type PUSCH PDSCH PUSCH PDSCH PUSCH PDSCH PUSCH PDSCH Physical UL/DL Shared Channel
Duplex Mode FDD FDD FDD FDD FDD: Frequency Division Duplexing
User Environment Indoor Indoor Indoor Indoor
System Bandwidth (MHz) 20.0 20.0 20.0 20.0
Channel Model ETU 3 ETU 3 ETU 120 EVA 120
MIMO Scheme 1×2 2×2 SFBC 1×2 2×2 SFBC 1×2 2×2 SFBC 1×2 2×2 SFBC SFBC: Space-Frequency Block coding
Cell Edge Rate (kbps) 128.00 512.00 128.00 512.00 128.00 512.00 128.00 512.00
MCS QPSK 0.20 QPSK 0.12 QPSK 0.20 QPSK 0.12 QPSK 0.20 QPSK 0.12 QPSK 0.20 QPSK 0.12 QPSK: Quadrature Phase Shift Keying
Tx
Max Total Tx Power (dBm) 23.00 46.00 23.00 46.00 23.00 46.00 23.00 46.00 A
Allocated RB 3 19 3 19 3 19 3 19 B
RB to Distribute Power 3 100 3 100 3 100 3 100 C
Subcarriers to Distribute Power 36 1200 36 1200 36 1200 36 1200 D = 12*C
Subcarrier Power (dBm) * 7.44 15.21 7.44 15.21 7.44 15.21 7.44 15.21 E = A-10*Log10(D)
Beamforming Gain 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 F
Tx Antenna Gain (dBi) 0.00 17.00 0.00 17.00 0.00 17.00 0.00 17.00 G
Tx Cable Loss (dB) 0.00 0.50 0.00 0.50 0.00 0.50 0.00 0.50 H
Tx Body loss (dB) 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 I
EIRP per Subcarrier (dBm) 7.44 31.71 7.44 31.71 7.44 31.71 7.44 31.71 J = E+F+G-H-I
Rx
SINR (dB) -4.19 -5.37 -4.19 -5.37 -2.33 -4.94 -2.20 -4.43 K
Rx Noise Figure (dB) 2.30 7.00 2.30 7.00 2.30 7.00 2.30 7.00 L
Receiver Sensitivity (dBm) -134.13 -130.61 -134.13 -130.61 -132.26 -130.18 -132.14 -129.67 M = K+L-174+10*Log10(15000)
Rx Antenna Gain (dBi) 17.00 0.00 17.00 0.00 17.00 0.00 17.00 0.00 N
Rx Cable Loss (dB) 0.50 0.00 0.50 0.00 0.50 0.00 0.50 0.00 O
Rx Body loss (dB) 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 P
Target Load 75.00% 90.00% 75.00% 90.00% 75.00% 90.00% 75.00% 90.00%
Interference Margin (dB) 0.89 2.72 0.89 2.72 1.46 3.13 2.71 3.74 Q
Min. Signal Reception Strength (dBm) -149.74 -127.89 -149.74 -127.89 -147.31 -127.05 -145.93 -125.93 R = M-N+O+P+Q
Path Loss & Cell Radius
Indoor Penetration Loss (dB) 19.00 19.00 15.00 15.00 11.00 11.00 8.00 8.00 S
Std. Dev. of Shadow Fading (dB) 11.70 11.70 9.40 9.40 7.20 7.20 6.20 6.20
Area Coverage Probability 95.00% 95.00% 95.00% 95.00% 95.00% 95.00% 90.00% 90.00%
Shadow Fading Margin (dB) 9.43 9.43 8.04 8.04 5.99 5.99 1.87 1.87 T
Maximum Allowable Path Loss (dB) 128.74 131.16 134.13 136.56 137.76 141.77 143.50 147.77 U = J-R-S-T
Propagation Model Modified Cost231-Hata Modified Cost231-Hata Modified Cost231-Hata Modified Cost231-Hata
eNodeB/UE Antenna Height (m) 25.00 1.50 30.00 1.50 40.00 1.50 50.00 1.50
Frequency (MHz) 1800 1800 1800 1800 1800 1800 1800 1800
Cell Radius (km) 0.47 0.55 0.87 1.02 2.13 2.78 5.64 7.54

* Subcarrier power is estimated assuming that the TX power is equally distrusted across the total bandwidth.
 In order to validate the theoretical LB, a field For the sake of comparative analysis, the same exercise
measurement is obligatory. Figure 2illustrates the UL and is conducted with HSPA+ system at 2.1GHz band (i.e., the
DL throughputs versus the pathloss from field results for most popular band for UMTS system) [8], [9]. Figure 4
LTE system at dense urban terrain with similar parameters illustrates the loading impact on the cell radius for UL and
as in Table 1 [13]. It is evident that the LTE system is UL DL of HSPA+ system. The figure indicates that the DL is
link limited and the difference in path losses is almost the limiting link in HSPA+ as long as the UL loading is less
similar to the theoretical one (3-4 dB). After validating the than 90%. The DL cell radius get decreased by 45% at
LB, the cell dimensioning process can be conducted as 100% load while the UL cell radius goes close to 0 at 100%
illustrated in the next section. UL load [14]. This illustrates the well know cell breathing
drawback of the UMTS system. To further analyze the cell
One of the important factors that may impact the cell breathing, the cell radius is estimated versus the allocated
radius is the loading. Unlike the legacy universal mobile power of the PA per user to guarantee 512Kbps DL
telecommunications system (UMTS) system, it is not throughput at cell edge. Figure 5 demonstrates the cell
expected that the LTE system will be severely degraded radius and number of users with 512Kbps at cell edge
with the increase of the loading thanks to OFDM technique. versus power allocation per user. The figure indicates that
Since the LB has been validated, we can estimate the cell the cell radius shrinks as a function of number of users at
radius based on different loading scenarios to analyze the cell edge with 512 Kbps throughput and only max of 6
impact of the loading on the cell radius. Figure 4, provides users can achieve 512 Kbps throughput at cell edge while
the theoretical DL/UL cell radii versus loading for an LTE the cell radius shrunk to 200m. Therefore, one native
system at urban terrain with same parameters of Table 1. HSDPA carrier (i.e., only HSDPA data) with 40 watt power
The figure reconfirms that the LTE is UL limited. The UL amplifier (PA) can serve only six users with 512kbps at cell
cell radius reduction is about 10% at 100% load and it is edge, which is only 200 meter away from the cell center
5% at 50% load (i.e., typical practical UL load), which is a [8], [9]. This limitation is a major differentiator for the LTE
very graceful degradation compared to HSPA+ system as system thanks to the OFDM technique.
illustrated later.
HSPA Cell Radius as a function of Loading

0.9
0.8
0.7
Cell Radius(km)

0.6
0.5 UL
0.4 DL
0.3
0.2
0.1
0
10 20 30 40 50 60 70 80 90 100
Cell Loading(%)

Figure 2 Average layer DL/UL throughputs for LTE system Figure 4 Impact of cell loading on cell radius for HSPA+ at
versus pathloss 2.1GHz in urban indoor scenario at 128kbps/512kbps cell edge
throughputs for UL/DL, respectively
Cell#Radius#versus#Cell#Loading

1.30##

1.20##

1.10##

1.00##

0.90##
Cell#Radius(Km)

0.80##

0.70##
UL#
0.60##
DL#
0.50##

0.40##

0.30##

0.20##

0.10##

0.00##
10# 20# 30# 40# 50# 60# 70# 80# 90# 100#

Cell#Loading#(%)
Figure 5 Cell radius and number of users with 512Kbps DL
Figure 3 Impact of cell loading on cell radius for LTE system at thrghouput at cell edge versus power allocation per user
1800MHz band with urban indoor scenario at 128kbps/512kbps
cell edge throughputs for UL/DL, respectively.
 IV. LTE Throughput Analysis
For a TCP/IP application and using the generalized
protocol parameters in [2], [13], the expected peak
theoretical DL and UL throughputs of LTE are provided in
Figure 6 for MIMO 2x2 and for single-input single output
(SISO) [2]. The SISO throughputs are the expected peak
throughputs to be seen in the indoor distributed antenna
solution (DAS) that deployed inside tall buildings or
underground stations/tunnels to enhance the coverage.
Unfortunately, the deployed legacy DAS systems do not
support MIMO configuration as only one antenna is
deployed and therefore, it is difficult and very costly to
upgrade the legacy DAS systems to support MIMO as a
complete parallel set of RF feeders and antenna system
need to be installed. Considering the throughput gain of
MIMO, an alternative solution to deploy MIMO with DAS
system based on hotspot approach could be an efficient Figure 7 LTE network DL throughput using CAT 3 Modem
interim strategy. Another alternative solution is by
alternatively feeding the two RF ports of the MIMO 2x2 to V. LTE Network Dimensioning
feed neighbor DAS antennas, which would provide less Following the average sector throughput estimation as
degraded MIMO performance. However, DAS design can provided in the previous section, the capacity of a single
be revisited to improve MIMO performance by maintaining LTE site can be calculated. Figure 8 provides LTE
multiple of half wavelength separation distance between dimensioning exercise and a comparison with DC-HSPA+
MIMO antennas. system. The key input to the dimensioning exercise is the
average sector throughput and the outcome is the number of
the subscriber based on the provided traffic profile. The
average sector throughput of LTE is 33Mbps as estimated
in the previous section. The average sector throughput of
the HSPA+ system is estimated using a similar manner to
be 12.3Mbps based on collocated LTE/HSPA+ sites and
testing same route depicted in Figure 7. A complete
comparative comparison between the collocated LTE and
HSPA+ systems for the route in Figure 7 is summarized in
Table 2 [13]. A typical DL loading of 70% is used in the
dimensioning exercise in Figure 8 and the traffic profile of
the user is 50kbps during busy hour (i.e., typical user
throughput if all users accessed the system at the same
time). For sake of comparative analysis, the HSPA+ system
is considered with two native HSPA+ carriers (i.e., no R99
Figure 6 Theoretical LTE UL and DL throughputs for MIMO traffic) and using the DC-HSPA+ feature where two
2x2 and SISO carriers are aggregated to provide 42Mbps peak throughput.
A peak to average margin of 20% is considered to
The most important factor during network design is accommodate burst traffic. Figure 8 indicates that the LTE
the average sector throughput. In this paper we have system offers capacity improvement of 34%, which is
estimated the average throughput for LTE system based directly reflected by the average sector throughput gain
on a commercially deployed LTE network with CAT 3 depicted in Table 2 and thus the spectrum efficiency gain.
UE. Figure 7 provides drive test throughput The complete network dimension exercise is summarized in
performance of LTE system from a commercially Figure 9. The full coverage and capacity dimensioning
deployed LTE network with same simulation exercise of LTE system is demonstrated in Figure 9 which
parameters used to estimate the peak throughputs in provides the total number of the required eNBs to meet the
Figure 6 using MIMO 2x2 configuration. The average coverage and capacity requirements.
throughput is estimated to be around 33Mbps over the
entire route with mobility at 80Km/hour. Without losing
the generality, we assume the average sector throughput
equals the average user throughput (i.e., 33Mbps).
 Figure 8 LTE and HSPA+ networks dimensioning

Table 2: Comparison between LTE and HSPA+ based on commercially deployed networks
Criteria DC-HSPA+ (2.1GHz) LTE (1800MHz)
Mobility average throughput 9 Mbps with DC (2x5MHz) 33 Mbps with 20MHz channel BW
Average scheduling rate 73%* 100%
Normalized mobility average
12.3 Mbps with DC (2x5MHz) Same as above
Throughput
Mobility spectrum efficiency 1.23 (i.e., 12.3Mbps/10MHz) 1.65 (i.e., 33 Mbps/20Mhz)
Throughput % 2.1% of the route > 21Mbps 50% of the route > 28Mbps
Number of serving cells 100 67
Estimated cell radius (meter) 390 500 (28% improvement)
64QAM utilization % 8% of the route 40% of the route
MIMO usage % MIMO +DC is not available yet 62% of the route
* Estimated from the number of successful high speed-shared control channel (HS-SCCH) decoded by the UE

Start
LTE Link Budget

MAPL of MAPL of Cell Average Configuration

Downlink Uplink Throughput Calculation Analysis
(MAPLDL) (MAPLUL)

MAPL = Subscribers Supported Traffic Model

Min (MAPLDL , MAPLUL) per Cell Analysis

Cell Radius per Cell Area per Cell Count per eNodeB Number
Morphology Morphology Morphology (initialized by Coverage Dimensioning)

Satisfy Capacity No Adjust eNodeB

Total Subscribers
Requirement? Number

Yes

eNodeB Number

End

Figure 9 LTE network dimensioning flowchart

 VI. LTE QoS Aspects 1. QoS Class Identifier (QCI): (for inter/intra-user QoS) is
used to control packet-forwarding treatment (e.g.
The quality of service (QoS) indicates the expected scheduling weights, admission thresholds, queue
service class in terms of packet delay tolerance, acceptable management thresholds, link layer protocol
packet loss rates, and required minimum bit rates during configuration, etc.), and typically pre-configured by the
network communication. QoS ensures that request and operator.
response of a user (i.e., inter-user QoS) or application (i.e., 2. Allocation and Retention Priority (ARP) (for Inter-user
intra-user QoS) correspond to a certain predictable service QoS): the ARP is stored in the subscriber profile in HSS
class. The QoS is a general term that used on various on a per access point name (APN) basis (at least one
conditions with service supplies and demands to assess the APN must be defined per subscriber) and it can take
capability of meeting customer service requirements. The value between 1 – 15 based on the user priority (i.e.,
assessment is not based on accurate scoring, but on analysis gold, silver, and bronze). The primary purpose of the
of service quality in different conditions instead. Then, ARP is to decide if a bearer establishment/modification
specific measures can be taken to improve service quality. request can be accepted or rejected (i.e., admission
The most common problems of the IP-based transmission control) in case of resource limitation.
networks are packet delay, jitter, and packet loss. 3. Guaranteed Bit Rate (GBR) and Maximum Bit Rate (MBR)
Additionally, different applications require different – this parameter is defined for GBR bearer only.
bandwidths. Therefore, the problem becomes more severe 4. Aggregate Maximum Bit Rate (AMBR) sums all non-GBR
and a robust QoS mechanism is mandatory. The QoS bearers per terminal/APN.
provides a comprehensive solution in such situation. Figure
10 illustrates the end-to-end bearer service architecture. The eNB guarantees the downlink GBR associated with
a GBR bearer and enforces the downlink AMBR associated
An evolved packet system (EPS) bearer uniquely with a group of Non-GBR bearers [2]. In order to maintain
identifies packet flows that receive a common QoS the same priority of the QCI over end-to-end
treatment between the UE and the PGW (i.e., same implementation, the QCI should be mapped to the IP
scheduling, queue, management/rate, shaping, and transport network. A key QoS parameter that used for this
policy). As shown in Figure 10, the EPS bearer consists of purpose is the DiffServ Code Point (DSCP) that
the radio bearer between the UE and the eNB, the S1-bearer encapsulated with each IP packet. A typical mapping
between the eNB and the SGW, and the S5/S8 bearer between QCI and DSCP is also shown in Table 3. The
between the SGW and PGW. An EPS bearer can be a signaling is mapped to the highest priority (i.e., DSCP =
guaranteed bit rate (GBR) or non-guaranteed bit rate (Non- 46) while other applications with different QCIs are
GBR) [2]. Table 3 provides the 3GPP QoS class identifiers mapped accordingly. Furthermore, the QCI/DSCP need to
(QCI) for different applications with corresponding QoS be mapped to the transmission network such as the
requirements. The QCI is further used within the LTE microwave (MW) links air interface. The mapping to the
access network to define the control packet-forwarding MW air interface is based on queuing technique and the
treatment from an end-to-end perspective [11], [12]. It also number of queues depends on the MW manufacturer. If we
ensures a minimum standard level of QoS to ease the assumed MW with 8 queues, then the mapping is illustrated
interworking between the LTE networks mainly in roaming also in Table 3. It is recommended to reduce the number of
cases and in multi-vendor environments. The packet delay queues and number of DSCP values to reduce the
budget (PDB) defines an upper bound delay that a packet is complexity during the implementation and later during
allowed to experience between the UE and the PGW. network expansion or upgrade.

The key QoS parameters attached to a bearer are

outlined as follows:

Figure 10 End-to-end SAE bearer service architecture

 Table 3. QCI mapping to DSCP and MW queuing
Traffic GBR/ QCI Priority Packet Packet DSCP MW Service Sample
type Non- Delay loss Queuing
GBR Budget rate
Signaling N/A 46 7 Stream Transmission Control Protocol
(SCTP)
Sync 46 7 Synchronization Signal
(1588V2)
O&M 46 7 Operation and maintenance
User Data GBR 1 2 100 10-2 46 7 Conversational voice
2 4 150 10-3 26 4 Conversational video (Live Streaming)
3 3 50 10-3 34 5 Real time gaming
4 5 300 10-6 26 4 Non-conversational video (Buffer streaming)
Non- 5 1 100 10-6 46 7 IMS signaling
GBR 6 6 300 10-6 18 2 Video (buffer steaming) TCP based (www. e-
mail, chat, ftp)
7 7 100 10-3 18 2 Voice, Video (live streaming) interactive
streaming.
8, 9 8, 9 300 10-6 0 0 Video (buffer steaming) TCP based (www. e-
mail, chat, ftp)

The deployment scenario in Table 3 is for application-

based QoS (i.e., provide different treatment for different
applications). This scenario consumes almost all MW
capability in terms of queues and therefore, there is no
room to accommodate other technologies on the same MW
network. Another interesting scenario to deploy QoS is the
inter-user QoS by providing differentiated treatment for
users based on their importance, for example, gold, silver,
and bronze users. This approach will allow the operator to
offer differentiated services to their customers while
maximizing the utilization efficiency of the scarce
resources in air interface and transport network. Table 4
provides a practical deployment scenario for the inter-user
QoS in commercial LTE network. In this exercise, QCI 6,
8, and 9 are used to represent gold, silver, and bronze users,
respectively. The user priority (i.e., ARP) is stored in the
HSS profile of the user and the PGW can map the QCI
values to DSCP values as per table 4. The PGW cannot
map the ARP to DSCP. The scheduling weight for each
user category is defined in the eNB and it is associated with Figure 11 FTP download throughput for three users with different
each ARP/QCI. The throughput testing results of this QoS as per Table 4.
exercise is shown in Figure 11. In this exercise, one LTE
cell with 20MHz channel bandwidth is used and three users As shown in Table 5, the actual average users throughputs
(gold, silver, and bronze) are accessing the LTE cell at the are well presented by the allocation/priority weight. The
same time. The demonstrated throughput is obtained using allocation/priority weight is operator adjustable value and it
FTP download. depends on operator strategy and pricing scheme for each
user category or the data bundle.
Table 4: Inter-user QoS deployment scenario in commercial LTE
network Table 5: Throughput Analysis for the scenario in Table 4
DSCP Values in User Allocation/ Total Cell Throughput (Mbps) 54.7

User Transport Network Priority Priority User Priority Gold Silver Copper
QCI Weight in
Category Data (ARP) in
Allocation/ Priority Weight in eNB 100 50 20

Signaling Voice eNB

Traffic HSS Number of Online Users 1 1 1
Expected Average User Throughput (Mbps) 32.2 16.1 6.4

GOLD 46 46 34 5 6 100 Expected Throughput Percentage (%) 59% 29% 12%

Achieved Average Throughput (Mbps) 32.9 14.7 7.1
Silver 46 46 18 7 8 50
Achieved Throughput Percentage (%) 60% 27% 13%
Bronze 46 46 0 9 9 20
 VII. LTE network Latency is analyzed and validated using field test results. It has
been validated that the LTE system is UL link limited
The latency is one of the most critical factors that can
and the difference between UL and DL path losses is
impact the performance of the LTE network. Therefore, a
around 3-4 dB. On the other side, the DL is the limiting
special attention should be given to the latency at designing
link in HSPA+ system as long as the UL loading is less
phase of the LTE network to meet the X1 and S2 latency
than 90%. Moreover, with LTE system, the UL cell
requirements and also to maintain the gain of the reduced
radius is reduced by 10% at 100% load, which is a very
latency that offered by the LTE system. A latency
graceful degradation compared to the HSPA+ system
comparison is conducted between the LTE simplified
where the DL cell radius is decreased by 45% at 100%
topology (termed as LTE trial) where the eNB is directly
load. We have conducted an end-to-end dimensioning
connected to the EPC without IPSec GW via one
exercise for LTE system including coverage and
transmission media (i.e., fiber network in this case) and a
capacity planning. In addition, we have presented the
full commercial LTE network similar to the topology in
key practical aspects of the QoS deployment and the
Figure 2. Also, for sake of comparative analysis, the latency
end-to-end deployment scenario for a commercial LTE
of the collocated HSPA+ network is presented. The latency
network for both application-based QoS and inter-user
is measured in terms of round trip time (RTT) which is
QoS. Also, a practical throughput testing results from
estimated by sending packets of different sizes (i.e., 10,
live LTE network is presented for inter-user QoS.
100, 1000, 1460 Bytes) using ping command to a local FTP
Finally, latency of the LTE network is compared with
server directly connected to the EPC in case of LTE
the HSPA+ network. It has been demonstrated that the
network and to the packet core of the HSPA+ network (i.e.,
full commercial LTE network offers average reduction
GGSN; gateway GPRS support node). As demonstrated in
of 40% in latency compared to HSPA+ network.
Figure 12, the simplified LTE network offers significant
latency reduction. Moreover, the LTE system can maintain
Acknowledgments
the same latency even with bigger packet size. The full
I wish to express my appreciation to the wireless
commercial LTE network introduces higher latency
broadband team for their cooperative support. Many thanks
compared to the simplified LTE network due to the
to Huawei and Qualcomm teams who supported this work
introduction of the IPSec GW, firewall, and other IP routers
with a lot of resources and practical results.
in access and core clouds. Despite this, the full commercial
LTE network yields a reduced latency compared to the References
collocated HSPA+ network that uses the same backhauling.
More specifically, the full commercial LTE network offers [1] J. Zyren and W. McCoy, “Overview of the 3GPP Long
Term Evolution Physical Layer,” Freescale Semiconductor,
average reduction of 40% in latency compared to HSPA+ Inc., white paper, July 2007
network. Furthermore, with HSPA+ network, the latency is [2] 3GPP TS 36.300 V8.5.0, 2009, “Evolved Universal
significantly increased with the packet size increase where Terrestrial Radio Access (E-UTRA) and Evolved Universal
the difference in latency between the smallest and biggest Terrestrial Radio Access Network (E-UTRAN); Overall
packets is 21ms versus 12ms with the full commercial LTE description”.
network and only 1ms for the simplified LTE network. [3] 3GPP TR R3.018, “Evolved UTRA and UTRAN radio
access architecture and interfaces”.
[4] 3GPP TS 36.201 V8.3.0, 2009 "LTE Physical Layer -
General Description”.
[5] 3GPP TS 36.211 version 8.7.0 Release 8, “LTE; Evolved
Universal Terrestrial Radio Access (E-UTRA); Physical
channels and modulation”.
[6] 3GPP TS 36.213: "Evolved Universal Terrestrial Radio
Access (E-UTRA); Physical layer procedures".
[7] 3GPP TS 36.101 V8.15.0, 2011 "Evolved Universal
Terrestrial Radio Access (E-UTRA); User Equipment (UE)
radio transmission and reception”.
[8] 3GPP TS 25.214 V8.12.0, 2011 "Physical layer procedures
(FDD)”.
[9] 3GPP TS 25.213 version 3.4.0 Release 1999, “Universal
Mobile Telecommunications System (UMTS); Spreading
and modulation (FDD).
[10] G. Berardinelli, et. all, “OFDMA VS. SC-FDMA:
Performance Comparison in Local Area IMT-A Scenarios,”
Figure 12 Latency comparison among LTE simplified network, IEEE Wireless Communications, pp. 1536-1284, Oct. 2008
full LTE network, and collocated HSPA+ network [11] 3GPP TS 23.002 V8.7.0 (2010-12) (Release 8) “Network
architecture”.
VIII. Conclusions [12] 3GPP TS 23.401 version 8.6.0 Release 8 "General Packet
Radio Service (GPRS) enhancements for Evolved Universal
In this paper, we have provided the key practical Terrestrial Radio Access Network (E-UTRAN) access".
[13] A. Elnashar and M. A. El-Saidny, “A practical performance
aspects and the best practices for design and
analysis of the Long Term Evolution (LTE) system based on
deployment of a commercial LTE network. The end-to- field test results,” submitted for IEEE Vehicular technology
end LTE network topology is presented along with key magazine.
network domains. The theoretical LB of the LTE system
[14] A. Elnashar, “Driving Broadband innovation in UAE; du June 2005. He has managed several large-scale networks and
LTE evolution,” GSA information papers, Dec. 2012; mega projects with approximate capital of one billion USD
http://www.gsacom.com/gsm_3g/info_papers.php4. including start-up, networks expansion, and swap projects. He
obtained his PhD degree in multiuser interference cancellation and
Ayman Elnashar was born in Egypt in 1972. He received the smart antennas for cellular systems. He published 18 papers in
B.S. degree in electrical engineering from Alexandria University, wireless communications arena in highly ranked journals such as
Alexandria, Egypt, in 1995 and the M.Sc. and Ph.D. degrees in IEEE Transactions on Antenna and Propagation, IEEE
electrical communications engineering from Mansoura University, Transactions Vehicular technology, and IEEE Transactions
Mansoura, Egypt, in 1999 and 2005, respectively. He has more Circuits and Systems and international conferences. His research
than 15 years of practical experience in telecoms industry interests include digital signal processing for wireless
including GSM, GPRS/EDGE, UMTS/HSPA+/LTE, WiMax, communications, performance analysis of cellular systems,
WiFi, and transport/backhauling technologies. Currently, he is Sr. CDMA, OFDM, mobile network planning and design, multiuser
Director of Wireless Broadband with the Emirates Integrated detection, smart antennas, beamforming, and robust adaptive
Telecommunications Co. “du“, UAE. He is in charge of mobile detection.
and fixed wireless broadband networks.
Prior to this, he was with Mobily, Saudi Arabia, from June 2005 to
Jan 2008 and with mobinil (orange), Egypt, from March 2000 to