You are on page 1of 5

Relaying in CDMA Networks:

Pathloss Reduction and Transmit Power Savings


Patrick Herhold, Wolfgang Rave, Gerhard Fettweis
Technische Universität Dresden, Vodafone Chair Mobile Communications Systems, D-01062 Dresden, Germany
Email: herhold@ifn.et.tu-dresden.de, Phone: +49 351 463 32739, Fax: +49 351 463 37255

Abstract— Relaying has recently emerged as a field of growing is taken in [5], where an additional air interface is used for
interest for wireless systems. The use of intermediate nodes for the relaying operation, thus inherently enhancing the overall
relaying information from a source to its destination promises bandwidth used.
improvements on various levels, ranging from increased connec-
tivity and reduced transmit powers to diversity gains. We examine For 3G CDMA networks, the idea of Opportunity Driven
various propagation models and network parameters and show Multiple Access (ODMA) to enhance TDD system capac-
to which extent the pathloss in cellular wireless systems can be ity and coverage was investigated [6]. Zadeh and Jabbari
reduced by the use of relay nodes in a two-hop scenario. Having considered a FDD system in which digital repeaters relay
highlighted these potentials, we discuss by means of numerical
data packets in a time-division manner, suggesting that power
analysis and system-level simulation under which conditions these
savings can be turned into a transmit power reduction for CDMA savings are feasible [7].
FDD systems. It becomes evident that the overall performance In this paper, we further discuss the possibility of relaying in
of the relay system depends on the node density and the relative CDMA systems. To this end, we enhance the work presented
load. in [8]. Following a presentation of our CDMA FDD system
I. I NTRODUCTION model in section II, we demonstrate the achievable pathloss
savings that provide the potential for a reduction of transmit
Compared to conventional wireless cellular systems, in power savings in Section III. Section IV first analytically
which all terminals are directly connected to the backbone analyzes the achievable transmit power reduction, showing
infrastructure via a single hop, the use of intermediate nodes that transmit power savings and interference levels depend on
to help transmit information from one node to another facili- system load. Using different carriers for the relay reception and
tates numerous improvements. For example, the connectivity transmission, respectively, we then investigate more complex
of nodes can be improved, a network-level advantage often scenarios by means of system-level simulation. The paper
referred to as enhancing coverage in cellular systems. With concludes in section V.
respect to the physical layer, the inherent diversity of the
relaying channel enables to benefit from these diversity gains II. S YSTEM M ODEL AND A SSUMPTIONS
[1].
Moreover, relaying splits longer paths into shorter segments, We investigate a cellular CDMA system, in which all users
thus reducing the resulting total pathloss by exploiting the non- simultaneously share the same radio resource. Uplink (UL) and
linear relation of pathloss vs. distance. At system level (MAC), downlink (DL) are separated from each other in the frequency
this potentially allows for a reduction of transmit powers, and, domain. Our system assumptions are as follows.
consequently, lower electromagnetic immission. It is this very a) Node types: Four types of nodes are present: base
topic – decreasing the exposure to electromagnetic radiation – stations (BS), relay stations (RS, these are user terminals that
that increasingly becomes relevant for system design, partially serve other nodes while simultaneously performing their own
for reasons of interference reduction, partially due to pressure communication with the BS), target stations (TS, served by
from public opinion. the RS), and direct stations (DS, conventionally connected to
The idea of relaying in wireless networks has long been the BS); see Fig. 1.
attracting attention [2], yet it was only until recently that b) Parameters: The total number of mobile nodes is K.
relaying is considered for practical systems. The probability of a mobile station having relay capability is
Various works addressed a variety of different relaying ap- p(R), so that in average there are p(R) · K relays to serve
proaches for existing systems. For example, a detailed concept the potential TSs. Note that p(R) = 0 corresponds to the
for incorporating relay functionality into contemporary GSM conventional case in which all mobiles directly communicate
networks is presented in [3]; another contribution for relaying with the BS, while for p(R) = 1 all mobiles can potentially act
in F/TDMA networks is provided in [4]. A different approach as relays for other terminals. The number of hops towards a BS
is limited to two (”single-relay”, or ”two-hop” operation). At
This work has been supported by the German Federal Ministry for Educa-
tion and Research under grant 01 BU 053. The authors take on responsibility most Mmax target stations can be served by a RS. Various
for the contents. propagation models are examined, each superpositioned by
BS 1
DS Pathloss model

Relative total pathloss β


RS
TS 0.8
WI LOS
0.6
HATA
0.4 BS height=20m
Roof top height=25m
0.2 K=20 WI NLOS
σ =10 dB
0
0 0.2 0.4 0.6 0.8 1
Fraction of # relays w.r.t. total # terminals p(R)

Fig. 1. An example system snapshot. The plotted antenna characteristics


of the BS reflect the employed three-sectorization. While this plot shows a Fig. 2. Relative pathloss reduction β as a function of the relay density p(R).
single BS only, multiple BS were simulated in order to reduce border effects. Parameter is the pathloss model (WI=COST Walfish Ikegami, HATA=COST-
HATA 259, LOS=line of sight, NLOS=non line of sight). BS antennas are
installed below roof top. As the relative number of available relays increases,
the pathloss in the relaying case reduces with respect to the direct case
a log-normally distributed random shadowing with standard (p(R) = 0.0). Antenna gains at the BS have not been taken into account.
deviation σ.
c) Routing: We assume the pathlosses between all
nodes to be known. The routing scheme is a sub-optimum IV. T RANSMIT P OWERS
min-pathloss algorithm [8]: based on the knowledge of all We now focus on the actual transmit power savings of
pathlosses between the network elements, the scheme routes a power-controlled CDMA systems. Transmit powers in such
potential target station via a relay if this results in a reduction a system are primarily determined by two measures. First, the
of the pathloss with respect to the direct link to the “nearest” received power should match the receivers sensitivity, which
BS. requires that transmit powers be set such that the propagation
d) Orthogonality Constraint: Relaying requires a relay loss is overcome. We denote this fraction of the transmit power
node to receive information from a source and to forward as pathloss-determined. Second, the interference level due to
this information to the intended destination. With respect to multiple access interference (MAI) may require an additional
capacity considerations, it would be beneficial to retransmit increase of the power to ensure a sufficient SINR at the
the relayed signal at the same resource, i.e. at the same time detector. As interference in a CDMA system is influenced by
and frequency, at which it has been received. For the purpose the network load, we describe this part of the transmit power
of a simple analysis, we initially assume the relay nodes to as the load-determined fraction. Power control autonomously
be capable of performing this operation. However, as this full- adapts the transmission power to the interference conditions.
duplex mode is technically not feasible in a small-size radio, The previous section has demonstrated the potential for re-
it is necessary to allocate orthogonal resources to the relays’ ducing the pathloss-determined fraction of the transmit power,
receive- and transmit paths. This is later considered in the and it remains to discuss the relation of relaying and network
simulations and will then be explained in more detail. load and its influence on the load-determined power fraction.
In the following section, we investigate to which extent the
pathloss can be lowered by introducing relay stations. A. Analysis
III. PATHLOSS R EDUCTION 1) General Analysis of CDMA Transmit Powers: The sys-
tem consists of a set of point-to-point links; each link i is
Our aim is to estimate the achievable pathloss reduction
characterized by its required SINR γi at the intended receiver.
for typical wide-area cellular configurations. To this end, we
In a perfectly power controlled system, the transmit power Pi
denote the total pathloss in the relay case, normalized to the
is set such that this SINR is achieved exactly. Assuming that
total pathloss in the direct case, by β. Total pathloss refers
interfering signals can be regarded as white noise, this SINR
to the sum of all individual link’s pathlosses in the system.
relation is expressed as
For the employed min-pathloss routing algorithm, we always
achieve pathloss savings, i.e. 0 ≤ β ≤ 1. αi,i gi Pi
Figure 2 shows this relative pathloss reduction β as a γi = X . (1)
αi,k Pk + Ni
function of the relaying probability p(R) for various propa-
k6=i
gation models. Assuming that each mobile in a network has
the ability to act as a relay (p(R) = 1.0), the total relay In this equation, αi,k is the path gain that relates the transmit
pathloss amounts to approximately 30% of the pathloss in the power of link k to the power level at the receiver of link i by
direct case. This strong potential for reduced transmit powers capturing the effects of path loss and shadowing (0 < αi,k ≤
motivates to investigate the actual effects that relaying has on 1). Moreover, gi is the processing gain of link i, and Ni is the
the transmit powers in a power-controlled system. thermal noise power at the receiver of link i. We now aim at
computing the transmit powers Pi . Reordering (1) yields given K users with a SINR requirement γ0 . Using (7) in (6)
γ i Ni X γi αi,k and solving for P0 yields
= Pi − Pk . (2)
αi,i gi αi,i gi γ0 N 1
| {z } k6=i | {z } P0 = · , (8)
ni Ψi,k α g0 − gpole (K)
| {z0 } | {z }
Defining ni = γi Ni /(αi,i gi ) and pathloss-determined load-determined

0 i=k where we have separated the parameters into two factors that
Ψi,k = γi αi,k , (3) reflect the pathloss-dependency and the load-dependency of
αi,i gi i 6= k
the transmit powers.
equation (2) can be rewritten as To see how data rate affects the transmit powers, consider
X a fixed number K of users. For low data rates, i.e., for
ni = P i − Ψi,k Pk ∀i . (4) large processing gains (g0  gpole (K)), the transmit powers
i6=k
in (8) are primarily pathloss-determined. As the rates and
This set of linear equations can be expressed in matrix form hence the load increases, i.e. as g0 → gpole (K), the mutual
with the introduction of a noise vector nT = (n1 , .., ni , .., nL ) interferences cause the transmit powers to become dominantly
and a transmit power vector PT = (P1 , .., Pi , .., PL ), so that load-determined. The powers grow unboundedly as the data
we have n = P − Ψ · P. The matrix Ψ consists of the mutual rate per user approaches the limit; that is, P0 → ∞ as
interferences Ψi,k . Thus, we obtain g0 → gpole (K).
For the purpose of further exposition, consider the ratio of
P = (I − Ψ)−1 · n . (5) the pole processing gain to the actually used processing gain,
Note that equation (5) defines a unique set of transmit powers 0 ≤ gpole (K)/g0 < 1. This ratio represents the normalized
given the links’ path gains, required SINRs, noise powers, load with respect to the pole capacity, and is for a fixed number
and processing gains. A valid set of transmit powers is of users solely determined by the employed processing gain
characterized by Pi > 0 ∀ i. g0 . In the following example we will see how this system load
2) Pole Capacity, Processing Gain and Data Rate: Our affects the transmit powers of direct and relaying systems.
baseline system for comparison of relaying performance is the
conventional system in which all mobiles directly transmit to B. Relay Case and Numerical Example
a BS. For such a system, the concept of pole capacity can be Consider the simple single-cell configuration in Figure 3.
readily applied [9]. To this end, and in order to further simplify An inner tier of mobiles serves as relay stations (RSs) for
our analysis, we assume in the sequel that the system-level distantly located terminals. The pathlosses αi,k reduce with
parameters of all mobile stations are equal, i.e., γi = γ0 , Ni = the introduction of relay hops. This reduction of pathlosses,
N and gi = g0 for all i. Note that choosing equal processing however, comes at the cost of an increased total data rate
gains (gi = g0 ) implies that all mobile stations operate at the as relay stations retransmit information that has already been
same data rate. For the purpose of a pole capacity analysis, emitted by the base station or target stations, respectively.
we assume furthermore that the path gains between all mobiles Recalling that RSs transmit their own data in addition to the
and the central BS are equal: αi,i = αi,k = α0 . This yields relayed information, it is obvious that the links between BS
ni = γ0 N/α0 g0 and Ψi,k = γ0 /g0 . We study an isolated cell, and RSs carry the total data rate of the cell. Parts of this data
i.e. intercell interference is neglected. For K users, equation is then additionally relayed to/from the targets.
(4) then reads In order for the links between BS and RSs to be able to
γ0 N γ0 transport this increased rate, the spreading factor of these links
= P0 − (K − 1) P0 . (6) needs to be reduced appropriately. That is, the processing gains
α0 g 0 g0
gi need to be lowered for the inner links.
Solving for K and allowing unlimited transmit powers For the example placement of nodes depicted in Fig. 3, we
(P0 → ∞) yields the well-known pole capacity for an isolated can easily determine the transmit powers using equation (5).
cell of the conventional, direct system Kpole = (g0 /γ0 ) + 1 Towards this end, we assume a log-distance pathloss model
for P0 → ∞. Equivalently, we can solve for the minimum in which αi,k = d−a , where di,k is the distance between
i,k
required processing gain gpole (K) necessary for serving K transmitter and receiver, and a is the pathloss exponent. For
users: the numerical examples in this section a pathloss exponent
gpole (K) = γ0 · (K − 1) for P0 → ∞ . (7) of a = 3 is used. The nominal processing gain is g0 ; the
links between BS and RS operate with a processing gain of
Hence, gpole (K) is the minimum processing gain that is g0 /2 as double data rates are required for these connections.
required to serve K users with SINR requirement γ0 if an For K = 20 mobiles and γ0 = 1 the pole processing gain
unlimited transmit power budget was available. In terms of becomes gpole (K)P = 19. The resulting average UL transmit
1
data rate, gpole (K) is the spreading gain that corresponds to power P̄ = K i Pi is plotted in Fig. 4 as a function of the
the maximum data rate per user that is theoretically achievable relative load gpole (K)/g0 of the cell.
node densities. An analytical analysis for the case of K = 20
1.5
users shows that the transmit powers diverge for a relative load
1 of 0.98 (see also Fig. 4). In other words, the pole capacity of
0.5 the relay system is smaller than that of the direct system.
y 0 We see that there are two major system-level drawbacks of
BS the relaying system. First, the total data rate to be transported
−0.5 RS in the relay case increases with the introduction of (additional)
−1 TS
relay hops. Second, the power control must ensure a sufficient
−1.5 SINR at the relays, thus increasing the number of points at
which signals need to be detected from one (BS in the direct
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2
x case) to many (BS and relays in the relay case).
To summarize, two converse trends affect the actual transmit
Fig. 3. Example analysis system. Many of the effects of complex relay powers: on one hand, relaying reduces the average pathloss, on
networks can be qualitatively described using this simple configuration.
the other hand the network load increases due to the immission
of additional signal copies, thus requiring stronger powers to
10
overcome enhanced interferences. It hence depends on the load
Avg. UL transmit power [dBm]

5 of the network whether or not relaying results in transmit


Relay case (Mmax=1) power savings, with the load being determined by both the
0
number of nodes and their individual data rates to be carried.
−5 Note that up to this point we have not made any constraint
Direct case
on the technical capabilities of the relay nodes. In particu-
−10 lar, we disobeyed the orthogonality constraint by assuming
Break−even that a relay station is capable of transmitting and receiving
−15
load
simultaneously at the same resource. This justified a direct
−20 comparison of the direct and the ideal relay system as both
0.7 0.75 0.8 0.85 0.9 0.95 1
gpole/ g0 (Relative load) techniques require a single carrier for continuous operation. It
was shown that even under these idealized circumstances re-
Fig. 4. UL transmit power (P̄ ) for the example system depicted in Fig. 3 as laying may exhibit stronger transmit powers for high network
a function of the relative load, expressed in terms of the ratio of processing
gains. As the data rate per link increases, i.e. as gpole (K)/g0 → 1, the power loads.
advantage of relaying reduces. Eventually, the transmit powers in the relay In order to study these general trends for more realistic
case exceed those of the direct case for loads greater than 0.96. assumptions, the next subsection discusses a realistic relaying
strategy for a CDMA FDD system and details the simulation
model that is used in this more complex study.
Clearly, it depends on the load of the system whether
relaying yields transmit power savings with respect to the C. Simulation Model
direct case. For the considered example, the break-even load 1) Frequency Assignment: As discussed previously, one
is 0.96: for relative loads stronger than this break-even load, needs to assign orthogonal resources for the receive and
relaying requires stronger powers than conventional direct transmit path of a relay node. One approach is to use different
communication. Furthermore, the results in Table I suggest time slots for the two operations. This store- and forward
that the break-even load depends on the node density, a fact option is especially suitable for data packet communication
that becomes intuitively clear considering the stronger mutual and other services that exhibit low delay sensitivity, and is
interferences that arise from the smaller distances between frequently considered in the literature; see, for example, [7].
receiving and interfering nodes. Another possibility is to assign different frequencies for the
Moreover, it becomes obvious that the capacity of the relay receive- and transmit paths, an option desirable for the inves-
system is smaller than the that of the direct system for high tigated CDMA FDD system for it allows to retain continuous
transmission. However, it requires that different frequencies
TABLE I be used for reception and transmission at the relay. These
T HE BREAK - EVEN LOAD , I . E . THE SYSTEM LOAD FOR WHICH DIRECT resources can be made available through the use of a second
SYSTEM AND RELAY SYSTEM REQUIRE THE SAME TOTAL TRANSMIT carrier. Since each carrier provides two frequencies, a total of
POWER ( FOR THE STAR - CONFIGURATION OF F IG . 4). F OR LOADS four frequencies is then available.
SMALLER THAN THE BREAK - EVEN LOAD , RELAYING YIELDS TRANSMIT In the framework of a German national project, an algorithm
POWER SAVINGS . was developed that assigns carriers such that (i) relay stations
receive and transmit at different frequencies, (ii) mutual inter-
K 10 20 30 40 ferences are avoided to the best possible extent, and (iii) both
Break-even load >1 0.96 0.83 0.69 carriers are loaded equally in order to avoid load unbalances.
Similar to the routing scheme, the algorithm takes as input
15

[dBm]
the pathlosses between the links, and then iteratively assigns Powers
frequencies subject to the above mentioned conditions. A 10 pathloss− load−

UL
detailed description of this procedure is beyond the scope of 5 determined determined

Avg. UL transmit power P


this paper. To permit a fair evaluation, this two-carrier relaying 0
needs to be compared to a conventional (non-relaying) system −5
Direct communication
that likewise utilizes two carriers.
−10
2) Simulation Method: A static simulation tool was used N=10 (per cell)
to investigate the effects of relaying on the average transmit −15
p(R) =1.0
powers for more realistic scenarios. Relevant simulation pa- −20 Relay case σ (dB)=8.0
rameters are summarized in Table II. A perfect power control −25
2 4 6 8 10 12 14
(PC) algorithm was implemented [10]. Data rate per mobile [104 kbit/s]

TABLE II
Fig. 5. Simulation results. Transmit powers of the conventional system and
S IMULATION PARAMETERS . the relay system vs. the data rate per mobile. As the data rate per mobile
increases, the load-determined power fraction becomes dominant over the
pathloss-determined fraction, and the rate increase then causes the relaying
Parameter Value system to exhibit stronger transmit powers than the conventional one.
Cell radius (hexagon) 800 m
Log-normal shadowing (σ) 8.0 dB
Target SINR Eb /N0 (γ0 ) var.
Noise power (N ) -106.7 dBm
node density. Even for optimistic assumptions (ideal node
Receiver noise figures {BS,DS,RS,TS} {5.0,8.0,5.0,8.0} dB placement in a star-scenario, relays have the capability of
Max. tx powers {BS,DS,RS,TS} {43,24,35,24} dBm transmitting and receiving simultaneously at the same fre-
Maximum DCH powers {DL,UL} {38,24} dBm quency), it was demonstrated that for high system loads the
DL orthogonality factor 0.4 transmit powers of the relay system exceed those of the direct
BS antenna height 30 m
BS antenna pattern Realistic sector
system.
DS, RS, TS antenna gain 18 dBi For interference-limited power-controlled systems this sug-
DS, RS, TS antenna (gain) Omni (0 dBi) gests that direct transmission eventually becomes favorable
Pilot & common power fraction 10% with respect to capacity considerations.
R EFERENCES
A simulation consists of numerous snapshots, the results of [1] Jon Boyer, David D. Falconer, and Halim Yanikomeroglu, “A Compari-
which are averaged to obtain reliable statistics. Each snapshot son of Relaying Models for Multihop Channels,” Mar. 2002, Submitted
represents a single realization of a random distribution of users to IEEE Trans. Wireless Comm.
[2] T.M. Cover and A.A. El Gamal, “Capacity Theorems for the Relay
and log-normal shadowing. At the beginning of a snapshot, the Channel,” IEEE Trans. Inform. Theory, vol. 25, no. 5, pp. 572–584,
specified number of mobiles are uniformly distributed over Sept. 1979.
the area. The subsequent routing procedure then establishes [3] George Neonakis Aggélou and Rahim Tafazolli, “On the Relaying
Capability of Next-Generation GSM Cellular Networks,” IEEE Pers.
the connection between the nodes, and finally frequencies are Commun., vol. 8, no. 1, pp. 40–47, Feb. 2001.
assigned. [4] V. Sreng, H. Yanikomeroglu, and D. Falconer, “Coverage Enhancement
through Two-Hop Relaying in Cellular Radio Systems,” in Proc. IEEE
D. Simulation Results Wireless Commun. and Netw. Conf. (WCNC) 2002, Mar. 2002, vol. 2,
pp. 881–885.
Fig. 5 compares the transmit powers achieved in the direct [5] Hongyi Wu, Chunming Qiao, Swades De, and Ozan Tonguz, “Integrated
case and in the relaying case as a function of the data rate per Cellular and Ad Hoc Relaying Systems: iCAR,” IEEE J. Select. Areas
mobile station. Clearly, with increasing rate per mobile, i.e. Commun., vol. 19, no. 10, pp. 2105–2215, Oct. 2001.
[6] T. Rouse, S. McLaughlin, and H. Haas, “Coverage–Capacity of
with higher network loads, relaying becomes less attractive Opportunity Driven Multiple Access (ODMA) in UTRA TDD,” in 3G
as predicted by analysis. In this example, relaying does not 2001 Mobile Communications Technologies, IEE Conf. Publication 477,
provide any power savings for rates greater than eighty kbit/s Mar. 2001, pp. 252–256.
[7] Ali N. Zadeh and Bijan Jabbari, “A High Capacity Multihop Packet
per mobile. While a complete discussion of the simulation CDMA Wireless Network,” in First ACM Workshop on Wireless Mobile
results is beyond the scope of this paper, it can be summarized Internet 2001, 2001.
that power savings ranging from 1 to 8 dB are feasible for low [8] M. Bronzel, W. Rave, P. Herhold, and G. Fettweis, “Interference
Reduction in Single-Hop Relay Networks,” in Proc. 11th Virginia
to medium network loads. Tech/MPRG Symposium on Wireless Personal Communications, June
2001, pp. 49–60.
V. S UMMARY AND C ONCLUSIONS [9] K.S. Gilhousen, I.M. Jacobs, R. Padovani, A.J. Viterbi, and L.A. Weaver,
“On the Capacity of a Cellular CDMA System,” IEEE Trans. Vehic.
We quantified the significant pathloss reductions that are Techn., vol. 40, pp. 303–312, May 1991.
achievable by relaying in wireless networks. However, due to [10] G.J. Foschini and Z. Miljanic, “A Simple Distributed Autonomous
the load increase caused by repeated emissions of essentially Power Control Algorithm and its Convergence,” IEEE Trans. Veh.
Technol., vol. 42, pp. 641–646, Nov. 1993.
the same signal, it was shown that the extent to which transmit
powers can be reduced strongly depends on network load and

You might also like