This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts

for publication in the ICC 2008 workshop proceedings.

Capacity and Coverage Analysis of a 3GPP-LTE

Multihop Deployment Scenario
Rainer Schoenen, Wolfgang Zirwas, and Bernhard H. Walke

Abstract— Broadband wireless access will be deployed in a consumes three resource blocks R0 , R1 , R2 , one on each hop
cellular way with 3GPP-LTE [1]. For the first rollout the main 0, 1, 2. The size of Ri in bits shall be the same, but the size in
demand is a huge area coverage. With only few available base terms of time and frequency bandwidth (T ×F ) depends on the
station sites that are connected to an access fiber, multihop
(relaying) techniques can be used well to fill the coverage gaps. modulation&coding scheme (P hyM ode) on this subchannel.
Later with increasing offered traffic, the demand shifts to higher Close to the sender, the higher received SIN R value allows
capacity over the area. Even for this purpose relays are beneficial. the highest P hyM ode, i.e. the highest data rate. At the cell
There is an area around relays where they provide better overall border the offered data rate is one order of magnitude lower
capacity to the user terminal, taking into account all resources (QP SK −1/3 compared to QAM 64−5/6 for LTE [1]). What
used for the first and second hop (the relaying overhead).
Relaying or Multihop operation therefore massively improves is bad in this situation is that a terminal operating at the lowest
the coverage as well as the capacity goals at low cost, without PhyMode occupies a ten times higher part of the spectrum than
the need of a cable or fiber access. a terminal operating at the highest PhyMode. That means the
This paper analyzes a realistic urban scenario on the island of average cell capacity is overproportionally determined by the
Jersey. We study the coverage and capacity over the area in three maximum possible rate at the outer regions.
cases. One base station (BS) only, one BS with four relay nodes
(RNs), and the latter plus another ring of nine RNs. The BS has Related work in this area mostly analyzes regular cellular
fiber access for rates beyond 100 Mbit/s, while the first hop of geometries without considering realistic pathloss due to ob-
Relays (H1) is fed over the air from BS using shared resources structions [4] [5] [6].
in the same LTE band. The second hop H2 is fed by the relays The paper is organized as follows. The first section defines
of group H1. In this paper we provide the results from numeric the scenario. Next, the used layer 1 (PHY) and layer 2 (DLC)
analysis based on models we explain here. It is shown that huge
gains in coverage and capacity are obtained by relaying. models are explained. Numerical results and 2D graphs are
presented in the next section. The paper ends with a summary.
Index Terms— Multihop, Relaying, LTE, Coverage, Capacity

II. U RBAN S CENARIO ( CITY OF J ERSEY )

I. I NTRODUCTION
The urban scenario of Jersey has been chosen because the

M ULTIHOP techniques are an elegant method to im-

prove coverage and capacity in relay enhanced radio
cells [2]. This is possible at much lower cost than providing
area of approx. 4.439km2 is a typical cell size. A radio
network planning has been done to place the base station and
all relay nodes at best possible locations. See figure 1 for
more base stations (BS) with their need for high-rate fixed the geographic positions. The topology (building placement)
network access at every location. Due to the limited power was known in advance and raytracing tools have been used
of transmitters and the high path loss in non-line-of-sight to obtain the received signal power PR,i at each location
conditions the received signal strength is not sufficient in many from each possible transmitter site. The physical parameters
urban areas. To improve the coverage we therefore need more and calculation steps are given in section III. In the end we
sending stations, which can be full BS or relay nodes (RN s obtain a coverage map which shows for each point on the map
of store-and-forward type). In this paper we study the realistic which serving station a terminal is best associated to (best
scenario of the city of Jersey. Compared to abstract and regular server). This can be the base station, if this is the rate optimal
cellular scenarios, where the benefit of relaying can be shown association, but it can also be one of the relays RN ∈ H1 or
easily [3], this provides a good proof-of-concept. RN ∈ H2. The maximum achievable rate at a certain point is
The OFDMA-based transmission scheme of 3GPP-LTE [1] then determined. From these matrix (2D) results we calculate
allows to coordinate the radio resources used in time and scalar performance measures to show that the overall coverage
frequency domain. In any relay enhanced cell we assume a and capacity is increased compared to using only one BS.
full coordination, so that there is no intra-cell interference. Therefore we compare the scenario with 1) one BS only, 2)
Therefore, a resource block can only be used by one of the the BS plus a ring of four RN ∈ H1 and 3) the BS plus a ring
actors, either BS or one of the four RN1 ∈ H1 or one of of four RN ∈ H1 plus a second ring of the nine RN2 ∈ H2.
the nine RN2 ∈ H2. For this reason, any traffic P DU that
goes from BS to a user terminal (U T ) associated to RN2
III. L AYER 1 AND 2 MODELS
– Wolfgang Zirwas is with Siemens, Munich, Germany From link level to MAC throughput, the performance of
– The other authors are with the Chair of Communication Networks at
RWTH Aachen University, Faculty 6, Germany the example system is evaluated by calculating the following
– This work is funded by the BMBF ScaleNet project in Germany steps.

978-1-4244-2052-0/08/$25.00 ©2008 IEEE 31

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the ICC 2008 workshop proceedings.

PhyMode modulation code rate SIN Rmin in [dB]

1 QPSK 1/3 0.9
2 QPSK 1/2 2.1
3 QPSK 2/3 3.8
4 QAM16 1/2 7.7
5 QAM16 2/3 9.8
6 QAM16 5/6 12.6
7 QAM64 2/3 15.0
8 QAM64 5/6 18.2
TABLE I
LTE P hyM odes AND THEIR REQUIRED SIN Rmin

1
M I(SIN R, m) =
([s · M Ishannon (SIN R)]−w + m−w )1/w
(2)
s = s(m) = 0.95 − 0.08 · (m mod 2) (3)

w = w(m) = 2 · m + 1 (4)
m is the modulation index, i.e. the number of bits per symbol
(1=QPSK,...8=QAM256). Figure 2(a) shows the result graph.
Fig. 1. The Scenario map of Jersey showing the BS (middle) and RN The net PHY throughput is obtained by multiplying with the
placement
coding rate. For LTE, coders have rates 1/3, 1/2, 2/3 and
5/6 [8]. So the results in Figure 2(b) were obtained. The
P hyM odes in this figure are given in Table I. Within this
• Transmit Power: 37dBm at the BS, 34dBm at the RN,
cell, all RN s are coordinated by the BS, so there is no intra-
• Bandwidth: b = 18MHz net (20MHz system),
cell interference.
• Frequency: 2.5GHz apropriate for LTE or WiMAX,
• Scenario Geometry: Cell Radius R = 1600m,
• Pathloss I: 3D model of the city scenario (walls of B. DLC layer
buildings), On DLC (MAC) layer, there is an overhead due to framing,
• Pathloss II: ray tracing to capture multi-path propagation, signaling and ARQ retransmissions. The latter depends on
• Noise: Thermal noise power is N = −96.4dBm, P ER, which can be taken into account when assuming
• SINR: the first performance measure below PHY layer, selective repeat ARQ by equation 5.
• MI: mutual information determined from SIN R and
modulation (Eq. 2), raboveARQ = rbelowARQ · (1 − P ER) (5)
• BER: bit error ratio, the PHY performance result,
In total we obtain a MAC overhead of M AC/P HY =
• PER: packet error ratio, the result after channel decoding,
132.3%. The PHY overhead of P HY /RAW = 107.1%
• Throughput: determined by bandwidth, PhyMode (mod-
comes from OFDM cyclic prefix duration. Under multihop
ulation and code rate), ARQ overhead (Eq. 5),
operation there are individual resources needed on every
• Second Hop Throughput: reduced by resources required
hop. The constant packet length requires a different resource
on first hop (Eq. 6).
share depending on the used P hyM ode which determines the
• Third Hop Throughput: reduced by resources required on
maximum rate ri,max usable on each hop. Therefore we can
first and second hop (Eq. 7).
get the maximum rate on the second hop to be
−1 −1
r2 = (r1,max + r2,max )−1 (6)
A. PHY layer
and on the third hop
The received power PR,i on every location is the output
−1 −1 −1
of software tools for raytracing. The next steps were ana- r3 = (r1,max + r2,max + r3,max )−1 (7)
lyzed analytically-numerically using Matlab. With SIN R =
PR,i /(N +I) the signal to noise ratio is easily determined. For For every location (x, y) we can now determine the best rate
each SIN R level between around 0 and 20 there in another out of r1 , r2 , r3 which gives us the result in Figure 7. One of
P hyM ode chosen, depending on the estimated performance the three rates is maximum and the index i of the maximum ri
of this P hyM ode in terms of bis/s/Hz. For determining determines the “best server”, i.e. it shows which station the UT
the required link level results we build upon the mutual at that location should be associated with. Note that we can
information (MI) method [7]. We apply the steps SIN R → also determine the best server by choosing the highest SIN R.
M I, M I → BER and BER → P ER to get the packet error But in the case of relaying, this would neglect the overhead due
probability. For the SIN R → M I we use [3]: to the resources used in the hops before. Figure 7(f) shows the
best server determined by rate (optimum) and for comparison
M Ishannon (SIN R) = log2 (1 + 10SIN R/10dB ) (1) Figure 5 shows the best server determined by SIN R.

32
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the ICC 2008 workshop proceedings.

9
7
BPSK
QPSK−1/3
8 QPSK QPSK−1/2
6 QPSK−2/3
QAM8
7 QAM16−1/2
QAM16 QAM16−2/3
QAM32 5 QAM16−5/6
6 QAM64−2/3
QAM64
MI [bit/s/Hz]

QAM64−5/6
QAM128

MI [bit/s/Hz]
shannon
5 4
QAM256
4 Shannon
3

3
2

1
1

0 0
−20 −10 0 10 20 30 40 −20 −15 −10 −5 0 5 10 15 20 25 30
SINR [dB]
SINR [dB]

(a) Mutual information (MI) depending on SIN R and PhyMode (b) resulting net rate taking P ER and ARQ into account

Fig. 2. Link level results for different modulation&coding schemes (P hyM ode).

(a) coverage using relays (b) capacity increase using relays

Fig. 3. Coverage and capacity compared for three scenarios: BS (one BS only), BS+H01 (with one tier of relays), BS+H01+H02 (BS with two relay hops
H1+H2). F 0..F 3 are 2D FIR data filters.

IV. N UMERICAL R ESULTS requires more share of the resources than a UT close to the
BS. The following equation [9] for the capacity C considers
The analysis using our own numerical code in Matlab this:

1 1
has been carried out to generate the two-dimensional data = dxdy (8)
in fig. 6 and 7. Here we show the downlink only that also C cellarea Capacity(x, y)

applies to the uplink if the uplink pathloss is the same (FDD). Figure 3(b) shows the capacity of each scenario. Two-
But the benefit in terms of capacity reveals if we derive dimensional FIR filters have been applied to upsample the
scalar performance measures from it. The coverage (in % original quantized data from raytracing engine. Figure 4 shows
of the area) of the scenarios differing by the number of that in a multihop scenario, more and more of the coverage
relays involved is determined by counting all locations with area of BS is taken over by RN s. The capacity C in bit/s can
SIN R > SIN Rmin . For LTE SIN Rmin = 0.9dB holds. be used to calculate the spectral efficiency e = C/b using the
Figure 3(a) shows the coverage of each scenario. The system used bandwidth b. As result we get the performance metrics
capacity is determined by assuming equal traffic load for each in Table II. According to this, two tiers of relays, compared
user terminal and a homogeneous user density over the area. to the BS only scenario, increase the coverage by a factor of
This means that a UT far outside, having a low P hyM ode, 2.08, and both the capacity and spectral efficiency by a factor

33
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the ICC 2008 workshop proceedings.

Scenario coverage [%] capacity [M bit/s] spec.eff.[bit/s/Hz] 0 3.5

BS only 47.257 16.223 0.901

BS+H1 79.891 18.217 1.012 3
BS+H1+H2 98.519 21.583 1.199 100

TABLE II 2.5

SCALAR RESULTS FOR THE RELAY SCENARIOS

200

300

y
1.5

400

0.5

500

600 −0.5
0 100 200 300 400 500 600
x

Fig. 5. Best server of BS+H1+H2 determined by SINR instead of net rate

[3] R. Schoenen and B. Walke, “On PHY and MAC performance of

3G-LTE in a multi-hop cellular environment,” in Proceedings of
the 3rd IEEE International Conference on Wireless Communications,
Networking and Mobile Computing (WiCOM), Shanghai, China, Sep
2007. [Online]. Available: http://www.comnets.rwth-aachen.de
Fig. 4. The fraction of area served by each group of servers [4] V. Sreng, H. Yanikomeroglu, and D. Falconer, “Coverage enhancement
through two-hop relaying in cellular radio systems,” in Proc. IEEE Wire-
less Communications and Networking Conference(WCNC’02), 2002.
[5] H. Yanikomeroglu, “Fixed and mobile relaying technologies for cellular
of 1.33. networks,” in 2nd Workshop on Applications and Services in Wireless
Networks (ASWN’02), 3-5 July 2002, pp. 75–81.
[6] R. Schoenen, J. Eichinger, and B. Walke, “On the OFDMA FDD mode
V. C ONCLUSION in 3G-LTE,” in Proceedings of the 12th International OFDM Workshop
This paper treats the numeric coverage and capacity analysis (InOWo’07), Hamburg, Germany, Aug 2007. [Online]. Available:
http://www.comnets.rwth-aachen.de
of a real city scenario. We analyzed the topologic structure [7] K. Brueninghaus and D. e. a. Astely, “Link performance models for
of the city to get the path loss, SIN R and data rate values system level simulations of broadband radio access systems,” in Pro-
for every location in that area. From that we derive the ceedings of the 17th Annual IEEE International Symposium on Personal,
Indoor and Mobile Radio Communications, Sep 2005, pp. 2306–2311.
coverage (in %) and capacity (in M bit/s) as well as the [8] K. J. e. a. Ekstrom H., Furuskar A., “Technical solutions for the 3G
spectral efficiency calculated for the whole system. Doing this long-term evolution,” IEEE Communications Magazine, pp. 38–45, Mar
analysis for three scenarios, one with a base station only, one 2006.
[9] C. Hoymann and S. Goebbels, “Dimensioning cellular wimax
with two hops (one tier of relays) and one with three hops part i: Singlehop networks,” in Proceedings of European Wireless
(two tiers of relays), we observed remarkable gains when 2007, Paris, France, Apr 2007, p. 7. [Online]. Available: http:
using relays. Since relays are served via radio by the BS, //www.comnets.rwth-aachen.de
[10] Bogdan Timus, “Deployment Cost Efficiency in Broadband Delivery
the number of interfaces to the fixed network remains the with Fixed Wireless Relays,” Ph.D. dissertation, KTH Stockholm, 2006.
same in all three scenarios. But the relays help to cover
areas behind obstructions or other locations where the path
loss from the BS is too large. Even in terms of capacity the
multihop techniques improve the spectrum efficiency, because
for all covered areas there is only one radio channel needed.
In contrast, having several BSs to serve the same area tends to
require more channels, since a reuse factor of one would result
in too high interference. It would be interesting to perform an
economic comparison of the CAPEX and OPEX related to the
scenarions considered, e.g. [10]. In addition to abstract and
regular cellular scenarios [3], this provides a realistic view on
the application of relays.

R EFERENCES
[1] http://www.3gpp.org/Highlights/LTE/LTE.htm.
[2] R. Pabst, B. Walke, D. C. Schultz, and et al, “Relay-Based Deployment
Concepts for Wireless and Mobile Broadband Radio,” IEEE Communi-
cations Magazine, pp. 80–89, Sep 2004.

34
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the ICC 2008 workshop proceedings.

0 8

0 7
60 100

50
100 6

200
40

200 5
30

300

y
20
300
y

10
400

400 3
0

−10
500
500
2

−20

600
600 1
0 100 200 300 400 500 600 0 100 200 300 400 500 600
x x

(a) SINR of BS only (in dB) (b) PhyMode of BS only

0 8

0 7
60 100

50
100 6

200
40

200 5
30

300
y

20
300
y

10
400

400 3
0

−10
500
500
2

−20

600
600 1
0 100 200 300 400 500 600 0 100 200 300 400 500 600
x x

(c) SINR of BS with H1 (in dB) (d) PhyMode of BS with H1

0 8

0 7
60 100

50
100 6

200
40

200 5
30

300
y

20
300
y

10
400

400 3
0

−10
500
500
2

−20

600
600 1
0 100 200 300 400 500 600 0 100 200 300 400 500 600
x x

(e) SINR of BS with H1 and H2 (in dB) (f) PhyMode of BS with H1 and H2

Fig. 6. On the area map of Jersey, these figures show the SINR [dB] and PhyMode [1..8] for a scenario with BS only, with one tier of relays H1, and with
two relay hops H1+H2

35
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the ICC 2008 workshop proceedings.

7
x 10
0 0 3.5

100 100

5
2.5

200 200
4 2

300 300
y

y
1.5
3

400 400
2

0.5

500 500
1
0

600 0 600 −0.5

0 100 200 300 400 500 600 0 100 200 300 400 500 600
x x

(a) Rate of BS only (b) Best server (coverage) of BS only

7
x 10
0 0 3.5

100 100

5
2.5

200 200
4 2

300 300
y

1.5
3

400 400
2

0.5

500 500
1
0

600 0 600 −0.5

0 100 200 300 400 500 600 0 100 200 300 400 500 600
x x

(c) Rate of BS with H1 (d) Best server (coverage) of BS with H1

7
x 10
0 0 3.5

100 100

5
2.5

200 200
4 2

300 300
y

1.5
3

400 400
2

0.5

500 500
1
0

600 0 600 −0.5

0 100 200 300 400 500 600 0 100 200 300 400 500 600
x x

(e) Rate of BS with H1 and H2 (f) Best server (coverage) BS with H1 and H2

Fig. 7. On the area map of Jersey, these figures show the available rate capacity [bit/s] and best server (middle=3=BS, 2=H1, 1=H2) for a scenario with BS
only, with one tier of relays H1, and with two relay hops H1+H2

36