Full Frequency Reuse in OFDMA-Based Wireless

Networks with Sectored Cells

Serdar Sezginer(1) and Hikmet Sari(1),(2)
(1)
Sequans Communications, Bat. Citicenter, 19 Le Parvis, 92073 Paris La Défense, France
(2)
Supélec, Plateau de Moulon, 1-3 Rue Joliot Curie, 91190 Gif sur Yvette, France

Abstract – Due to overlap of antenna radiation patterns and interference-free regions, where there is no overlap between
the resulting interference between sectors, different frequency the radiation diagrams of adjacent sector antennas. Users
bands are typically used in different sectors of a cell in cellular located in those regions can be served by the base station (BS)
systems. In this paper, a full frequency reuse concept, which is antennas of Sectors A, B and C using the same frequency
based on multiple-input multiple-output (MIMO) techniques, is
introduced and analyzed for the uplink and downlink. The re-
band (or group of subcarriers). More specifically, suppose that
sults indicate a significant increase of the cell capacity. Sector A is serving a user in its blank region using a particular
Keywords—Frequency reuse, cell sectoring, multiple-input subcarrier group. The same subcarrier group can be used by
multiple-output (MIMO). the other sectors to serve users in their own blank regions.
Therefore, 3 different users can be served simultaneously
I. INTRODUCTION using the same subcarriers.
In cellular communications systems, cell sectoring is com-
monly used to increase cell and network capacity [1]. This
consists of replacing the omnidirectional base station (BS) f2 f1
antenna by several directional antennas each radiating within
a specified sector. With hexagonal cell patterns, the most
common sectoring schemes employ 120° or 60° sectors. The f2 f1 f3 f2 f1
conventional approach is to allocate a different channel to
each sector. Fig. 1 shows a hexagonal cell pattern with 120°
cell sectoring and the corresponding frequency allocation to f3 f2 f1 f3
different sectors. Splitting each cell into three sectors multi-
plies the cell capacity by three, but since a separate channel is
allocated to each sector, three channels are needed to serve the f2 f1 f3 f2 f1
entire cell and the spectral efficiency remains unchanged.
As is well known, frequency is a rare and expensive re-
source, which needs to be utilized efficiently. Therefore, there f3 f2 f1 f3
is a significant interest to use the same channel in different
sectors of a cell. This would not be a problem if the radiation
diagrams of adjacent sector antennas did not overlap. How- f3
ever, in practice, the radiation diagrams of the sector antennas
used exceed the sector boundaries and there is a significant
overlap region between adjacent sectors. This prohibits full Fig. 1: Hexagonal cell pattern with 120° cell sectoring.
frequency reuse (FFR).
One way to handle this problem in cellular systems based Interference
on orthogonal frequency-division multiple access (OFDMA) region
[2] is to reserve a group of subcarriers to serve users located
in overlap regions. With 120° sectors, this group is divided in
3 subgroups, and one subgroup is allocated to each overlap
region. The remaining subcarriers can be reused in all sectors.
This scheme ensures partial frequency reuse (PFR).
In this paper, we are interested in full frequency reuse. We Sector A Sector B
introduce a new FFR concept based on multiuser multiple-
input multiple-output (multiuser MIMO) techniques. The BS
proposed technique is applicable to both uplink and downlink.
The paper is organized as follows: In the next section, we
give a formulation of the problem and describe the proposed
technique for the downlink. Next, in Section 3, we describe
the case of the uplink. Finally, we give some performance Sector C
results in Section 4 and our conclusions in Section 5.

II. FULL FREQUENCY REUSE ON THE DOWNLINK

The problem of full frequency reuse in sectored cells is
illustrated in Fig. 2. The blank regions in this figure are the Fig. 2: Interference regions in a cell with three sectors.
 The shaded regions are the interference regions between First, from the received signal samples, the following
adjacent sectors. A simple way to handle this interference in signals are computed.
OFDMA is to allocate different carrier groups to different
interference regions (static separation), or to avoid using w1 = r1 − b(h1 s3 + h2 s 4 )
simultaneously in several overlap regions the same group of = a (h1 s1 + h2 s 2 ) + n1 (3.a)
carriers (dynamic separation). Both schemes reduce the
frequency reuse factor and are particular cases of partial w 2 = r2 − d (h2 s 3* − h1 s 4* ) .
frequency reuse. (In this paper, the frequency reuse factor is
defined as the number of times a given channel is used in the = c (h2 s1* − h1 s 2* ) + n 2 . (3.b)
same cell.)
The proposed solution to make full frequency reuse in From these signals, the receiver computes
sectored cells is based on using the BS antennas and
transceivers of two adjacent sectors to handle the interference ( 2
h1* w1 = a h1 s1 + h1* h2 s 2 + h1* n1 ) (4.a)
in the overlap region of their antenna radiation diagrams. It
applies to both downlink and uplink and uses existing ( 2
h2 w2* = c * h2 s1 − h1* h2 s 2 + h2 n 2* ) (4.b)
multiuser MIMO concepts.
and then,
Focusing on the downlink, suppose that two users u1 and u2
are located in the overlap region of sectors A and B and that y1 = h1* w1 / a + h2 w2* / c *
u1 is to be served by sector A and u2 is to be served by sector
B during the same time slot and using the same carrier group.
( 2
= h1 + h2 s1 + η1
2
) (5)
Suppose further that during two consecutive OFDM symbol
periods and on a given subcarrier, sector A wants to transmit where η1 = h1* n1 /a + h2 n*2 /c* . The resulting signal can be sent
symbols (s1, s2) to user u1 and sector B wants to transmit to a threshold detector to make an optimum estimate of
symbols (s3, s4) to user u2. Rather than transmitting their symbol s1 conditional on (s3, s4).
symbols independently, the two sectors cooperate using a full- To make an estimate of symbol s2, the receiver further
rate 2x2 space-time code (STC). Note that the well-known makes the following computations:
Alamouti code matrix [3] is not full rate (under the definition
of the rate as the number of transmitted symbols per antenna (
h2* w1 = a h1h2* s1 + h2 s 2 + h2* n1
2
) (6.a)
use), as it transmits only 2 symbols using 2 antennas and 2
time slots. Two possible full-rate code matrices are given in (
h1 w 2* = c * h1 h2* s1 − h1 s 2 + h1 n 2*
2
) (6.b)
[4] and [5]. From those, we use here the code matrix of [5], and then,
which leads to simple maximum-likelihood (ML) decoding.
y 2 = h2* w1 / a − h1 w2* / c *
This matrix is given as follows:
( 2
= h1 + h2 s2 + η2
2
) (7)
⎡ as + bs3 − cs 2* − ds 4* ⎤
X=⎢ 1 ⎥, (1)
⎣as 2 + bs 4 cs1* + ds3* ⎦ where η2 = h1* n1 /a − h2 n*2 /c* . As previously, an optimum
estimate of symbol s2 conditional on (s3, s4) can be obtained
where a, b, c, and d are complex-valued design parameters by sending the signal y2 to a threshold detector. Thus, the
and the star designates complex conjugate. The first row of receiver gets an optimum (maximum likelihood) estimate of
this matrix represents the signals transmitted from sector A, symbols s1 and s2 conditional on (s3, s4), which we denote by
and the second row represents the signals transmitted from ( ML ML
)
sˆ1 s , s , sˆ2 s , s . Note that user u1 in fact does not need to
sector B. Next, the first column of this matrix represents the 3 4 3 4

signals transmitted during the first symbol interval, and the
second column represents the signals transmitted during the
second symbol interval.
Denoting the channel response (at the subcarrier frequency
at hand) from the transmitter (Tx) of sector A to the receiver
(Rx) of user u1 by h1 and the channel response from the Tx of
sector B to the Rx of that user by h2, the received signal
samples corresponding to the symbol periods T1 and T2 can be
written as:
r1 = h1 (as1 + bs3 ) + h2 (as 2 + bs4 ) + n1 .
r2 = − h1 ( cs2* + ds4* ) + h2 (cs1* + ds3* )+ n2 .
where n1 and n2 are additive noise terms.
It is assumed that the channel responses h1 and h2 are
known to the receiver. It was shown in [5] that optimum
decoding of the transmitted symbols can be performed with a
complexity on the order of M2, where M is the constellation
size. The receiver works as follows:
decode (s3, s4), but to obtain the optimum estimate of (s1, s2),
it needs to make the above computations for all values of the
symbol pair (s3, s4). For each symbol pair (s3, s4), it thus gets
an optimum estimate of (s1, s2) and the associated metric, and
it makes a final decision corresponding to the smallest metric.
The final decision on (s1, s2) is denoted by sˆ1 , sˆ2
ML ML
. ( )
Equations (5) and (7) clearly indicate that second-order
diversity is achieved.
Similarly, denoting the channel response from the Tx of
sector A to the Rx of user u2 by h3 and the channel response
(2.a) from the Tx of sector B to the Rx of that user by h4, the signal
(2.b) samples received by this user corresponding to the symbol
periods T1 and T2 can be written as:
r3 = h3 (as1 + bs 3 ) + h4 (as 2 + bs 4 ) + n 3 . (8.a)
r4 = − h3 ( cs 2* + ds 4* ) + h4 (cs1* + ds 3* )+ n 4 . (8.b)
The receiver of user u2 makes the same type of computa-
tions as that of user u1 to detect the (s3, s4) symbols. Equations
(2.a) through (7) remain valid with h1 replaced by h3 and h2
replaced by h4.
While sectors A and B cooperate as described above to
serve two users in their overlap region, sector C can serve a
third user located in its blank region which is not subjected to
any interference from sectors A and B. Hence a cell sectored
into 3 sectors with overlapping antenna radiation diagrams
can serve 3 users per channel use. This process being true for
all carriers, the scheme corresponds to full frequency reuse.
III. FULL FREQUENCY REUSE ON THE UPLINK
Next, let us consider the uplink case. Since the two users do
not cooperate and exchange any information, no space-time
coding between user terminals is supposed in that direction.
Instead, we consider spatial multiplexing of two user signals
(virtual MIMO). Suppose that user u1 wants to transmit
symbol a1 and that user u2 wants to transmit symbol a2 to the
BS and that both users are located in the overlap region of
sectors A and B. Denoting by g1 the channel response from
user u1 to the Rx of sector A, by g3 the channel response from
that user to the Rx of sector B, by g2 the channel response
from user u2 to the Rx of sector A, and finally by g4 the
channel response from user u2 to the Rx of sector B, the signal
samples received by the receivers of the two sectors are
respectively given by:
x1 = g1a1 + g 2a2 + w1
x2 = g3a1 + g 4a2 + w2 .
where w1 and w2 are additive noise terms. The receivers of
sectors A and B detect jointly the transmitted symbols (a1, a2).
Assuming ML detection, they compute the following metric
for all combinations of (a1, a2) and select the one which
corresponds to the smallest metric value.
{
D ( p1 , p2 ) = x1 − g1a1 − g 2 a2 + x2 − g 3a1 − g 4 a2
2 2
} (10)
An exhaustive search clearly involves the computation of M2
metrics and M2-1 comparisons, where M is the size of the
signal constellation. The receiver complexity in exhaustive
search is similar to that of the downlink receiver described in
the previous section, but it is known that using a simple
receiver architecture based on conditional estimation similar
to the one explained above the complexity can be reduced to
the order of M.
As for the downlink, we have thus shown that in a sectored
cell, two sectors with overlapping antenna radiation diagrams
can simultaneously serve two users on the uplink using
existing MIMO concepts. (Full-rate full-diversity space-time
coding is used on the downlink and full-rate spatial
multiplexing is used on the uplink.) Also as on the downlink,
while sectors A and B serve two users in their overlap region
using spatial multiplexing and joint detection, sector C can
serve a user in its blank region, which is free of interference.
In other words, we have shown that full frequency reuse is
possible despite the presence of the overlap regions between
adjacent sector antenna radiation diagrams. Of course,
implementation of the proposed scheme requires knowledge
of user positions by the scheduler. However, this does not
mean full knowledge of their positions, but only knowledge of
the region in which the users are located (blank regions of A,
B, C, overlap region of A and B, overlap region of B and C,
overlap region of A and C).
V. PERFORMANCE ANALYSIS
We now discuss the performance of the proposed technique
making some simplifying assumptions. To do so, we consider
120° cell sectoring and a simplified sector antenna radiation
diagram illustrated in Fig. 3. The diagram is given from the
center labeled 0 and extends to its edge labeled π / 3 + θ .
Ideally, the antenna radiation diagram should have a brickwall
shape, i.e., it should not exceed π/3 while having constant
amplitude over the entire angular span, but this is not realistic.
In practice, the attenuation at the sector edges ±π/3 with
respect to the center must be kept small, typically less than 1
dB, and this is only achievable with an excess angle θ.
Magnitude
(9.a)
(9.b)
0 π/3 π/3+θ Angle
Fig. 3: The sector antenna radiation pattern.
Given the antenna radiation diagram above, a conventional
radio resource management scheme would assign specific
carrier groups to the sector overlap regions of angular width
2θ. The group assigned to one of these overlap regions cannot
be used in other regions. In contrast, full frequency reuse can
be achieved in non-overlap regions of angular width
2(π/ 3 − θ ) . Then, using a proper normalization, the frequency
reuse factor of such a scheme can be expressed as
2θ + 6(π/ 3 − θ )
FRF =
2π/ 3
6θ
= 3− (11)
π
The FRF hence reduces linearly with the excess beamwidth
parameter θ. For θ = π/20 (or 9 degrees), we have an FRF of
2.7, which corresponds to a loss of 10% of the cell capacity
compared to a scheme which allows full frequency reuse (i.e.,
FRF = 3). For θ = π/10, the FRF is 2.4 and the cell capacity
loss is 20%. By allowing full frequency reuse in the sector
overlap regions, the proposed technique simply avoids this
capacity loss.
In addition to avoiding the capacity loss in the sector
overlap regions, the proposed technique also provides some
performance improvement in terms of bit error rate (BER),
particularly on the uplink. The spatial multiplexing with two
receive antennas used on the uplink provides second-order-
diversity when detected using an ML detector. On the
downlink too, equations (5) and (7) explicitly show that a
diversity order of 2 is achieved in the sector overlap regions
using the presented full-rate full-diversity space-time code.
The reference performance against which the proposed
scheme should be benchmarked is the single-input single-
output (SISO) transmission scheme using the same
modulation in a non-overlap region. Indeed, since we are
assuming that user terminals have a single antenna, non-
overlap regions of the three sectors can only use SISO
transmission, and this is our reference scheme.
Figure 4 shows the BER performance results using uncoded
QPSK and uncorrelated Rayleigh fading. The results show
that on the uplink the proposed multiuser spatial multiplexing
technique leads to a tremendous performance improvement
compared to that of the non-overlap regions. The SNR
improvement is as high as 12 dB at the BER of 10-3, and it
increases at lower BER values. On the downlink, the
multiuser full-rate full-diversity space-time code used gives a
performance curve with the same slope as the SM used on the
uplink, but there is a significant performance difference
between the two schemes. The reason is that two receive
antennas are used on the uplink and only one is used on the
downlink. Looking at the equations describing the encoding
and reception of the two schemes, the SM technique used on
the uplink transmits two symbols and we have two equations
to determine them, but the STC used on the downlink has
only two equations to determine the 4 symbols transmitted per
matrix. But the main result here is that at BER values lower
than 5.10-3, the STC used on the downlink has better
performance than the SISO scheme used in the interference-
free regions.
Fig. 4: Performance comparison between the SM used on the
uplink, the STC used on the downlink, and the SISO used in
interference-free regions using QPSK.
Similarly, Figure 5 shows the same type of performance
results with 16-QAM modulation. Here too, the multiuser SM
technique used on the uplink leads to a tremendous
improvement over SISO transmission, and there is a cross-
over point between the respective performances of the STC
used on the downlink and the SISO technique used in non-
overlap regions. A comparison of Fig. 4 and Fig. 5 reveals
that at the BER of 10-3, SM with 16-QAM achieves over 6 dB
performance improvement compared to SISO with QPSK.
Since 16-QAM doubles the data rate with respect to QPSK,
this result means that the proposed technique significantly
improves performance while doubling the uplink capacity.
Fig. 5: Performance comparison between the SM used on the
uplink, the STC used on the downlink, and the SISO used in
interference-free regions using 16QAM.
VI. CONCLUSIONS
In this paper, we have introduced a new concept for full
frequency reuse in sectored cells. It is based on using multi-
user MIMO techniques to handle multiple users in the overlap
regions of adjacent sectors. On the downlink, a full-rate full-
diversity 2×2 space-time code is used for transmitting sym-
bols to 2 users from the 2 adjacent sectors, while the third
sector transmits to a third user located in its interference-free
region. On the uplink, simple spatial multiplexing is used
between 2 users located in the overlap region of 2 sectors, and
a third user is served in parallel from the third sector. The
analysis indicates that the proposed technique leads to a sig-
nificant increase of the cell capacity with an additional per-
formance improvement on the uplink.
ACKNOWLEDGMENT
The authors would like to acknowledge the support of the
European Commission through the FP7 project WiMAGIC
(www.wimagic.eu).
REFERENCES
[1] T. S. Rappaport, Wireless Communications – Principles and Practice,
Prentice Hall, 1996.
[2] IEEE 802.16-2005: IEEE Standard for Local and Metropolitan Area
Networks – Part 16: Air Interface for Fixed and Mobile Broadband
Wireless Access Systems – Amendment 2: Physical Layer and Medium
Access Control Layers for Combined Fixed and Mobile Operation in
Licensed Bands, Feb. 2006.
[3] S. M. Alamouti, “A simple transmit diversity technique for wireless
communications,” IEEE J. Sel. Areas Commun., vol. 16, pp. 1451–
1458, Oct. 1998.
[4] J.-C. Belfiore, G. Rekaya, and E. Viterbo, “The Golden code: A 2×2
full-rate space-time code with non-vanishing determinants,” IEEE
Trans. Inf. Theory, vol. 51, pp. 1432–1436, Apr. 2005.
[5] S. Sezginer and H. Sari, “Full-rate full-diversity 2×2 space-time codes
of reduced decoder complexity,” IEEE Commun. Letters, vol. 11, no.
12, pp. 973-975, Dec. 2007.