See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/221638974

On the Impact of Repeaters Deployment on WCDMA, Networks Planning

Conference Paper · May 2006

DOI: 10.1109/VETECS.2006.1682858 · Source: DBLP

CITATIONS READS
2 50

5 authors, including:

Mario Garcia-Lozano Luis Alonso

Universitat Politècnica de Catalunya Universitat Politècnica de Catalunya
88 PUBLICATIONS   412 CITATIONS    244 PUBLICATIONS   2,881 CITATIONS   

SEE PROFILE SEE PROFILE

Fernando Jose Casadevall Palacio Silvia Ruiz Boqué

Universitat Politècnica de Catalunya Universitat Politècnica de Catalunya
140 PUBLICATIONS   1,587 CITATIONS    92 PUBLICATIONS   437 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

4WARD (ICT) View project

Greenet project View project

All content following this page was uploaded by Mario Garcia-Lozano on 19 May 2014.

The user has requested enhancement of the downloaded file.

 On the Impact of Repeaters Deployment on WCDMA
Networks Planning

M. Garcia-Lozano, L. Alonso, F. Casadevall, S. Ruiz L.M. Correia

Department of Signal Theory and Communications Instituto de Telecomunicacoes
Technical University of Catalonia (UPC) Instituto Superior Tecnico (IST). Tech. University of Lisbon
Barcelona, Spain Lisbon, Portugal
mariogarcia gtsc.upc.edu

Abstract- This paper addresses the analysis of WCDMA systems example fiber optic or copper links. Moreover, repeaters have a
with repeaters deployment. The real different path delays, taking typical internal delay of 5-10 gs. On the other hand, in
into account the repeaters presence and the finite nature of the WCDMA systems, the RX usually is a Rake one, which
time window of Rake receivers are considered. This allows an performs a delay profile characterization and just those
enhanced analysis with respect to classical approaches from a components within a certain time window are constructively
system level viewpoint. Results show that not taking into account combined. Thus, with the introduction of repeaters, the system
these effects imply erroneous metrics. Indeed, some of the most
relevant parameters in network performance evaluation are
analysis is significantly modified and must be revised.
clearly affected. Although much previous research efforts have been focused
on the analysis of CDMA-based systems ([3]-[5], only to cite a
Keywords- WCDMA, network planning, repeaters. few), not many studies in the literature analyze the effect of
repeaters on the capacity and feasibility conditions for a
I. INTRODUCTION CDMA mobile communications system. In [6], the authors
Optimizing the design of radio access networks in mobile study repeaters as coverage extenders and how they modify
communications systems is a technical and economical interference in highways environments. Similarly, the works in
challenge for operators. In this sense, active repeaters (non- [2] and [7] deal with coverage, but also investigate variations in
regenerative) are of special interest when considering the capacity due to the noise rise in the donor BS caused by
coverage of a mobile communications system in a certain set of repeaters. In order to minimize this effect, [8] proposes an
situations, which include filling in coverage holes in valleys, automatic on-off switching repeater. Finally an analytic scheme
tunnels, buildings or extending the service area beyond cells for estimating the required power at the BS in an environment
boundaries. Despite their use in second generation networks, with fiber-optic repeaters is found in [9]. Published studies
repeaters are expected to play even a major role in planning always consider one of following assumptions:
WCDMA (Wideband Code Division Multiple Access) systems. * RXs receive power just through the radiolink with the
For example, since adjacent cells share the same power, BS or through one of the repeaters. Other contributions
repeaters are a cost-effective option to reduce inter-cell are not considered in any way [8],[9].
interference, particularly in environments with hotspots [1]. On
the other hand, the use of higher frequencies implies higher * All the contributions are always perfectly combined in
propagation losses and repeaters can be key devices to the mobile terminal (MT). Ideal maximum ratio
guarantee indoor coverage. Finally radio over fiber, optical combining (MRC) is applied [1],[2],[6].
wireless, as well as O&M capabilities have improved However, in comparison to classical approaches, in real
significantly during the last years and will benefit a dense systems the Rake RX has to perform a delay profile
deployment of repeaters, especially in urban environments. measurement to designate the time offsets of the strongest
Unfortunately, repeaters are not noiseless devices and, signal components. Only those paths within a certain time
therefore, they modify the interference and thermal noise window are constructively combined, the others not being
patterns of the donor receiver (RX). This implies an effect in considered and causing a certain level of 'self-interference'.
both the coverage and capacity of the cell [2]. The UMTS (Universal Mobile Telecommunications System)
Signal contributions from the base station (BS) and from BSs and user equipments can usually handle a 20 gs time delay
repeaters arrive through different paths and at different times to between two paths [10], and this value can be easily exceeded
the RX. Moreover, the paths passing through a repeater will be if repeaters are deployed, even for small values of the delay
usually 'longer' (both in spatial and time domains), because of spread of the radio channel.
the delay introduced by the link between the repeater and the The novelty of this paper is a generic analysis of a
donor BS, and the internal delay of the repeater itself. In fact, WCDMA system with repeaters deployment from a system
the link can be established via various transmission media level viewpoint. Considering the different path delays and the
which can imply lower propagation velocities than radio, as for

0-7803-9392-9/06/$20.00 (c) 2006 IEEE

466
finite duration of the time window in Rake RXs allows an
enhanced prediction of the network performance. A
comparison with the results obtained using classical
approaches has also been done. Different relevant parameters
that characterize networks performance have been evaluated I u3
and compared. The differences with respect to simplified
classical analysis are shown to be remarkable, stressing the
improvement provided by the proposed analysis. The paper is
focused just on the UL because of space reasons, however a
deeper analysis and considering the DL has been done [11].
Figure 1. Different reception situations in an environment with repeaters
II. MODEL
Let us consider a mobile communications systems Y, . <(k)P
PTX (k)(1 )
consisting in any number NB of BSs, each one of them /JL [k,s(k)]- [k, s(k)]
icTt{k,s(k) LIL
connected to any number of repeaters. The system layout can
be completely generic, with no restriction on the spatial 11,=l i(=-tl( S(()l I'. [m, s(k1 )] ictl( S(l(k)}- ST,s(k)]
distribution of the different devices. It is assumed that there are M#k ii T[k,s(k)]
NM MTs in the network and there are no restrictions on the
service the MTs can use, either. A power control algorithm is From (1), it can be seen that the useful received power at
present, and it is composed of the so called inner and outer s(k) from k (numerator) can be expressed as the product
loops. The former aims at adjusting the transmitted powers, so between the transmitted power PTX (k) and a certain term
that a certain signal to interference plus noise ratio (SINR) which only depends on propagation conditions. Note that it is
target before de-spreading for a user k is reached; this ratio is possible to see the inverse of this factor as an effective
denoted as f)k,s(k)], s(k) being the BS that serves k. The latter equivalent loss whose value is calculated as the parallel (using
intends to keep the quality of communications at a desired level an electrical equivalency) of the loss of the paths inside the
in terms of block error rate depending on higher layers' time window of the Rake. We denote this as LeA[kj]:
requirements. The expression for f2k,s(k)] is straightforward in
generic WCDMA systems without repeaters deployment [13].
However, with the introduction of repeaters, the analysis has to (2)
be modified. The different cases to be considered are shown by L iT[kk,](k[ )] Li[k, j)] )
means of an example in Fig. 1. Indeed, based on the position of
an MT with respect to the network layout, the signal could be Regarding the denominator, the first term is the UL multi-
received in different ways (numbered in the figure): user interference received by the link between k and s(k). This
can be expressed as the transmitted power divided by another
1. Directly from the donor BS. effective loss, named Lefj[k,j], which is the parallel of all the
2. Simultaneously from the BS and one repeater.
3. Just from one repeater. paths between k and s(k).
4. From the BS and several repeaters.
5. From several repeaters. I

6. All the previous ones in a soft-handover situation with Y, I (3)

other BSs or repeaters in the system.
4fF [k, j] =

ic F[k,s(k)] -4 [k, j) ]
For example, type 2 areas are defined as the zones in which The second term is the self-interference power, which is
the difference of propagation times of both received paths is generated by signal replicas from k that arrive with such a
smaller than the Rake time window. Otherwise, one of the delay that fall out of the Rake window at s(k). Finally, n[s(k)] is
paths would generate a level of self-interference, which is the total thermal noise after considering the noise at the BS and
nonexistent in a deployment without repeaters. In this sense, the received noise produced by the repeaters connected to s(k).
the position of the MT with respect to each BS or repeater, the
internal delay at the repeater, and the transmission media in the If we define the total power received at BS j, including all
link between donors and repeaters are of key importance. the signal and thermal noise terms:
Let define F(kj) {LL(kj), L4kj),... } as the set of mean
propagation losses of the different paths between MT k and BS PfLW(J)
j. Let also define the subset T(k,j) c F(kj) as that containing
X uY
M=1
(m)/Lf, (m, j) + n(j) (4)
the paths constructively combined by the Rake RX, that is,
those within the Rake observing time window limits. And OUL is the SSIR (Signal to Signal-plus-Interference-
Taking these definitions into account, the SINR for a MT k plus-Noise Ratio) measured in the UL:
in UL before de-spreading yuL[k,s(k)] can be written as:
(5)
1 +/JL [k, s(k)] PRuxL$j ,t [s(k)ILjq, [k,s(k)]

467
 Then, it is obtainned a compact and general expression that these users into account, the different BSs to which the MT is
allows calculating the power that an MT must transmit to reach connected (active set) should be evaluated independently and, a
a certain SIR target: posteriori, the connection that requires less power from the user
should be selected. That would mean solving the equations
system twice. Firstly, a coarse adjust is done, with all virtual
PTuxL (k) = 0uL [k, s(k)] PRuxL,,,, [s(k)] Lfp [k, s(k)] (6) connections. Next, a refined one should be done, once the
optimum connection is chosen and the others are eliminated.
Notice that the required transmission power depends on the
total amount of power received at s(k) and the effective path III. NUMERICAL RESULTS
loss (parallel) of those paths within the Rake time window. In order to evaluate several performance indicators of a
Obviously, it also depends on the SIR target. Observe as well WCDMA network, and to quantify the differences obtained
that the total effective loss, i.e. considering all transmission when using the proposed analysis with respect to classical
paths, is also considered within the total received power term. approaches (which include some approximations), the
From (6) it is clear that the transmitted powers can be numerical analysis has been embedded in a static system level
calculated by solving the NMi-order linear equations system simulator. The number of MTs in the first set of simulations
formed by the application of this expression for each one of the has been set so that the number of users not reaching the E,INO
MT in the system. In general, NM is high and, therefore, this required by the power control is 5 % when using the ideal full
type of microscopic approach requires a lot of computational MRC approach. This implies 93 users inside the common pilot
power, especially if it has to be solved on a frame-by-frame channel coverage area (some more details are given later).
basis. Nevertheless, a second relationship between PRx,0,ULu) The study case scenario consists of a long road or railway
and the individual UL transmitted powers was established in like scenario, with one BS and one repeater as a coverage
(4), so that the the problem can be turned into a macroscopic extender. The distance between the donor BS and the repeater
approach, while remarkably reducing the computational cost: is 4 km. The link between them is considered to be an optical
fiber with refraction index of 1.48 in the core and a length of
4.5 km. The gains of the transmitter and receiver in the link and
pUL (j ) ={E [ks, ) uL [k, s(k)] PRux,to [s(k)]I+n (j) the internal gain of the repeater are adjusted so that there is no
amplifier saturation at the repeater. For radio propagation
evaluation, COST231-Hata propagation model for suburban
As a consequence, the dimension of the equations system, areas has been used. A summary of other important parameters
can be reduced to NB, in the same way as it is usually done in is shown in Table 1.
WCDMA systems without repeaters [12],[13]. Since NB is far
smaller than NM, the analysis is considerably reduced in terms The first parameter to evaluate is the MT's transmission
of computational cost. The final linear equations system can be power as a function of the distance to the BS. Fig. 2 shows
expressed in matrix notation as follows: different curves with the behavior of MTs for different internal
delays at the repeater (from 5 to 11 s). The curve with legend

ULUL fliULRX,tot
N

UL
"MRC" stands for the case in which all paths are constructively
combined, while "SEL" stands for the classical approach in
which only one of the paths is selected, that is, the
where HRx,totUL is a vector containing the unknowns, that is, communication of the MT is established just through the BS or
the total received power at each BS, NUL is a vector with the the repeater. It can be seen that all curves show an upwards
total thermal noise power, and Q1UL represents a NB x NB matrix trend in the required transmission power as the MT gets farther
with individual elements calculated as follows, being 5i the away from the BS and enters the repeater's area of influence.
Kronecker delta: Comparing the cases, differences with respect to the full MRC
approach reach up to 3 dB when the repeater has a 9 or 11 s
delay, this is because the path from the BS is out of the Rake
time window (20 gs). This implies two facts: first, it puts in
~ULQUL((I,)
, i) N,i L=1fF(k,
-, L
j)_* (k, i))9
4fp (k i) 0 (kI
evidence the presence of a term of self-interference that causes
kci an increase of the transmission power; second, in this situation,
fewer paths can be combined, and therefore the attenuation
Note that the number of equations does not depend on the suffered by the MT is greater than the one measured in an ideal
number of installed repeaters. Therefore, the computational full MRC case. Since, the percentage of area in which all the
cost to analyze the network with repeaters deployment is paths cannot be constructively combined is directly
independent of their number, and we can afford using as much proportional to the repeater internal delay, the degradation
repeaters as needed without adding significant complexity to observed with respect to the optimistic classical consideration
the network planning. Once the total received power has been will also increase. Notice that with an internal repeater delay of
calculated, the individual MT transmitted power can be found 5 gs, there are very few areas in which all paths cannot be
by using (6). An additional usefulness of this formulation is combined, hence, results are very similar to the full MRC case.
that it allows considering also those MTs that have initiated a For distances over 7 km, the error mantains a constant value of
soft handover process with minor changes. In order to take

This work has been founded by the Spanish Research Council

CICYT+FEDER through the projects TEC2005-07326-C02-O1/TCM and
TIC2003-08609

468
 TABLE I. SIMULATION PARAMETERS that there will be also important differences in the prevision of
Min. required common pilot channel EJo -12 dB
the percentage of users not reaching their EbINo target between
the classical approaches and the proposed one. Some results are
BS Max. transmission power 43 dBm shown in Fig. 3. According to the full MRC or SEL
Noise Figure 5 dB
approximations, around a 95 % of users could be correctly
MT Max. transmission power 21 dBm
Noise Figure 8 dB served, which is a typical design constraint. However, this
Repeater
Max. transmission power 43 dBm value is reduced to 86 % when the repeater has an internal
~~ Noise Figure 5 dB delay of 11 gs. Given this, the optimistic result may lead to
UL Eb No target 2.9 dB accept as good an inappropriate system design. The differences
DL EblNo target 4.4 dB
Service Bit Rate 12.2 kbps are always higher with respect to the MRC approach, which is
the most optimistic case. Even with a repeater with very low
internal delay (5 gs) a difference of 1.4% of users arises.
17.5
Coverage is subsequently evaluated in Fig. 4, which shows
the probability of coverage for each pixel in the scenario. A
1 16.5 pixel is considered to have coverage as long as it accomplishes
four conditions. Firstly, the EJIo measured on the common
blo
15.5 pilot channel must be higher than -12 dB. Moreover, once the
MT has selected a cell, it should have enough power to reach
`
14.5 the UL EINo target. Similarly, the MT will not need more
power than the maximum allowed level per connection to reach
13.5 the DL E,No target. Finally, the BS will not be transmitting at
3 5 its maximum power.
Distance to the BS [kin]
It can be seen that, when the full MRC approach is used,
Figure 2. UL transmission power as a function of the distance to the BS (an orthogonality factor of 0.4 is assumed, meaning 1 a full loss
of orthogonality and, thus, maximum intra-cell interference)
95 the coverage is supposed to be guaranteed in every point
"I between the BS and the repeater. However, when considering
t 93
the real delays of the paths, a clear loss is obtained, especially
91
in the repeaters area of influence. With the proposed
e
X
evaluation, certain zones turn out to demand more power than
, 89 the available one and as a consequence connections are
degraded or dropped. Since interference is also increased, the
\ 87 evaluated levels of EJIo on the pilot signals are affected too.
These differences would be much sharper when a higher
85 orthogonality factor is present. Thus, the analysis reveals that
MRC SEL 5 7
Repeater internal delay [,us]
9 11
the proposed layout would be inappropriate in a real situation
but this conclusion would not have been reached if a classical
Figure 3. Correctly served users for classical approaches and different
internal delays at the repeater evaluation had been performed.
1.5 dB. Under some circumstances, inaccuracies of this order
Finally, admission regions for the network have been
studied. Admission Control (AC) is a key Radio Resource
may not be tolerable. Management strategy in WCDMA systems [15]. Since
If just the best path is considered (SEL case) and the coverage and capacity are tightly coupled, a method that
repeater had an internal delay of 5 gs, an error around 2 dB handles all new incoming traffic is mandatory. In our case, the
would be committed by the approximated analysis along 1 km. AC strategy decides whether a new radio access bearer can be
The SEL curve will tend to the MRC case when the MT is admitted or not according to a certain estimation of the current
close to the BS. This is coherent, since the propagation path load, the so-called load factor [14]. If the load stays below a
through the repeater is more than 10 dB below than that certain threshold the new RAB will be allowed. In our
through the BS, thus this signal contribution can be considered simulations this maximum has been adjusted to 0.85 because it
negligible. Conversely, when the MT reaches the repeater and implies a typical value of 95 % of users correctly served when
continues moving away from the BS, the curve tends to the evaluating the scenario with classical approaches.
9 gs case. The curve does not tend to the 11 s case because In this sense, Fig. 5 shows the number of users that could
the "SEL" approximation only considers one path, the others be admitted in the study-case scenario. It can be seen that those
not being taken into account, and then not generating any kind results obtained when considering traditional approximations
of interference or self-interference. Between these two clear are optimistic. Indeed, when differences in delays are
areas, the results will be more or less accurate depending on the considered, the resulting admission region becomes smaller
internal delay of the repeater. and is inversely proportional to the internal delay of the
Considering that all MTs have a maximum available repeater. This degradation in capacity reaches the 8 % when
transmission power, the previous reasoning induces to think the repeater has an internal delay of 11 s. Moreover, it can be

469
 WCDMA networks have been studied and results show that
classical approximations lead to optimistic results. This could
lead to accept bad planned networks as good designs. UL
transmitted powers have shown differences up to 3 dB. The
previsions on the number of correctly served users would be
also quite optimistic when classical approaches are used; in our
example this value evolved from 95 % to 86 % when using the
proposed method. Reductions of coverage have been also
1000_
observed when the realistic behavior of the paths is considered.
Finally, admission regions have been also compared, one more
_

time, traditional approximations lead to unrealistic and

9200
4000 ~n| F optimistic results. In our layout, reductions in the number of
admitted users can reach 8 %. Therefore, we conclude that the
ubguE) andth
500 1250enane
2500 3750prpoa fo inera
50001 6250 7500 8750deay
10000Q 11 proposed analysis is a highly useful tool to plan and manage
x [m]
WCDMA networks that include the presence of repeaters.
Figure 5.30091Probability of coverage for classical full MRC (first
subfigure) and the enhanced proposal for internal delay I11 ts
500
REFERENCES
93 [1] M. Rahman, P. Emstrom, "Repeaters for Hotspot Capacity in DS-
92 CDMA Networks", IEEE Trans. on Vehicular Technology, Vol. 53, No.
3, May 2004, pp. 626-633.
901
[2] M. Garcia-Lozano, L. Alonso, F. Casadevall, S. Ruiz, "Capacity and
Coverage Tradeoff in WCDMA Environments with Repeaters
Deployment", Proc. of Wireless Personal Multimedia Communications
Conf. 2005 (WPMC 2005), Aalborg (Denmark), Sep. 18-22, 2005.
87 [3] J. Zander, "Performance of Optimum Transmitter Power Control in
86 Cellular Radio Systems", IEEE Trans. on Vehicular Technology, Vol.
41, No. 1, Feb. 1992, pp. 57-62.
[4] S.A. Jafar, A. Goldsmith, "Adaptive Multirate CDMA for Uplink
84 Throughput Maximization", IEEE Trans. on Wireless Communications,
MRC 5 7 9 11 Vol. 2, No. 2, Mar. 2003, pp. 218-228.
SR
Repeater internal delay [,usJ [5] A. Baier, U.C. Fiebig, W. Granzow, W. Koch, P. Teder, J. Thielecke,
"Design Study for a CDMA-based Third-generation Mobile Radio
Figure 6. Admission region evolution for classical approaches and System", IEEE Journal on Selected Areas in Communications, Vol. 12,
differenent interal delays at the repeater. No. 4, May 1994, pp. 733-743.
[6] W. Lee, D. Lee, "The Impact of Repeaters on CDMA System
seen that there is almost no difference between the MRC and Performance", Proc. of IEEE Vehicular Technology Conference 2000
SEL cases, this is because the only difference between these (VTC-Spring 2000), Tokyo (Japan), May 15-18, 2000.
two approaches is the effective attenuation that perceives the [7] M.N. Patwary, P.B. Rapajic, I. Oppermann, "Capacity and Coverage
MT. On the other hand, the enhanced proposal also introduces Increase with Repeaters in UMTS Urban Cellular Mobile
self-interference terms, which have a direct and evident impact Communication Environment", IEEE Trans. on Communications, Vol.
on the load factor (it is increased) and therefore on the 53, No. 10, Oct. 2005, pp. 1620-1624
maximum number of admitted users. [8] W. Choi, Y. Cho, T.W. Ban "Automatic On-Off Switching Repeater for
DS/CDMA Reverse Link Capacity Improvement", IEEE
Communications Letters, Vol. 5, No. 4, Apr. 2001, pp. 138-141.
IV. CONCLUSIONS [9] K.S. Jeong, J.M. Cheong, T.H. Park, T.G. Kim, S. Park, "Performance
Along this paper, a realistic analysis of WCDMA networks Analysis of DS-CDMA Reverse Link with Fiber-Optics Repeaters",
Proc. of IEEE Vehicular Technology Conference 2000 (VTC-Spring
with repeaters deployment has been presented. Generic and 2000), Tokyo (Japan), May 15-18, 2000.
compact expressions for the UL have been derived, omitting [10] 3GPP Specification. TR 25.956 (Release4) - "UTRA Repeaters
the DL for space reasons. Transmitted powers can be Planning Guidelines and System Analysis" http://www.3gpp.org/
calculated without simplifications and the effect of interference [11] M. Garcia-Lozano, L. Alonso, F. Casadevall, S. Ruiz, L.M. Correia,
generated by signal replicas that fall out of the Rake RX "Enhanced Analysis of WCDMA Networks with Repeaters
window has been considered realistically. A dimension Deployment", (unpublished, under review process).
reduction of the power control problem has been attained as in [12] S.V. Handy, "Congestion Measures in DS-CDMA Networks", IEEE
basic WCDMA systems. Moreover, solving the obtained Trans. on Communications, Vol. 47, No. 3, Mar. 1999, pp. 426-437.
equations systems is independent of the number of repeaters. [13] L. Mendo, J.M. Hernando, "On Dimension Reduction for the Power
Control Problem", IEEE Trans. on Communications, Vol. 49, No. 2,
The proposal considers realistically the paths delays and the Feb. 2001, pp. 243-248.
duration of the time window in the Rake RXs, which allows [14] H. Holma, A. Toskala, WCDMA for UMTS radio access for third
more accurate metrics from a system level viewpoint. generation mobile communications, Ed. Wiley & Sons, 2nd ed, 2002.
It has been simulated the major differences with respect to [15] 3GPP Specification. TS 25.133 (Release5)- "Requirement for Support
of Radio Resource Management (FDD)" http://www.3gpp.org/
classical approaches. Different performance indicators of

470

View publication stats