398 IEEE COMMUNICATIONS LETTERS, VOL. 12, NO.

5, MAY 2008

Load Dependent Interference Margin for

Link Budget Calculations of OFDMA Networks
Andreas Fernekeß, Graduate Student Member, IEEE, Anja Klein, Member, IEEE,
Bernhard Wegmann, Karl Dietrich, and Matthias Litzka

Abstract— Link budget calculations are essential in the plan- reduced. This effect should also be considered when planning
ning of wireless networks to get a reliable estimate of maximum wireless networks.
cell radius and guaranteed signal to interference plus noise In networks using code division multiple access (CDMA),
ratios at the cell border. In future networks using orthogonal
frequency division multiple access, the interference introduced it is state of the art to consider a load dependant margin for
by co-channel cells depends on the amount of subcarriers the intra- and inter-cell interference [7]. There is no intra-cell
utilized in the interfering as well as in the considered cell. If interference when using OFDMA due to orthogonality among
distributed subcarrier allocation is applied, frequency diversity the subcarriers. Nevertheless, there is inter-cell interference
and interference averaging effects can be obtained. A method which depends upon the load, i.e. the amount of utilized sub-
is derived in this paper to consider load dependent margins for
interference from co-channel cells in the link budget calculation. carriers, in the co-channel cells. In this paper a load dependant
This margin makes the estimation of the maximum cell radius interference margin (IM) for wireless networks using OFDMA
more accurate. with distributed subcarrier allocation is derived. Compared to
Index Terms— Interference margin, link budget calculation, a noise limited scenario the maximum pathloss is reduced by
orthogonal frequency division multiple access. the IM so that still a given required signal to interference plus
noise ratio (SINR) is guaranteed to the user at the cell border.
This paper is organized as follows. In section II, the system
I. I NTRODUCTION
model is described that is used throughout this investigation.

L INK budget calculations are the first step in the network

planning process. By considering gain and loss during
transmission and assuming a required signal to noise ratio
The IM due to partial load is derived in section III. Section IV
provides results obtained from link level simulations. Finally,
conclusions are drawn in section V.
(SNRreq ) at the receiver and a maximum transmit power at
the sender a maximum pathloss can be determined. This leads II. S YSTEM M ODEL
to a maximum cell radius for that the assumed SNR can still A wireless network using OFDMA is considered. The
be guaranteed at the cell border [1, 2]. To compensate for subcarrier allocation to users is performed in a distributed
slow or fast fading, margins can be added in the link budget fashion. In that case, subcarriers allocated to a single user
calculation. If the pathloss is known, the achievable SNR at the are distributed over the whole available bandwidth. Users in
cell border can be adjusted by changing the transmit power co-channel cells utilize different sets of subcarriers. The al-
or the maximum cell radius taking only thermal noise into location procedure is cell dependant. Coding and interleaving
account. In scenarios where the inter-cell interference is larger is performed over a set of subcarriers. Convolutional codes
than the thermal noise changing the cell radius will have no are used. Therefore, interference averaging and frequency
influence on the signal to interference ratio (SIR). Improved diversity can be obtained [6].
SIR conditions at the cell border can only be achieved by A regular hexagonal cell grid with omnidirectional antennas
increasing the distance to co-channel interfering cells with a is assumed. The transmit power per subcarrier is constant. The
higher frequency reuse [3]. SIR at the cell border depends on the reuse distance D and
A promising multiple access schemes for future mobile can be approximated by
communication systems is Orthogonal Frequency Division


S Rα
Multiple Access (OFDMA) [4]. OFDMA is considered for SIR = 10 log10 = 10 log10 (1)
I (D − R)α
4th Generation wireless networks and is currently used for
broadband wireless access networks according to the IEEE with S the received power density, I the interference power
802.16e standard [5]. OFDMA provides frequency diversity density at the receiver, α the pathloss coefficient and R the
and interference averaging if distributed subcarrier allocation cell radius [3]. The reuse distance D is a linear function of R
is considered [6]. If subcarriers are not utilized in neighboring so that the SIR is independent of R.
cells, the effective interference experienced by a user is A set of subcarriers A is utilized in the considered cell. The
co-channel interfering cells utilize a set of subcarriers B with
Manuscript received January 8, 2008. The associate editor coordinating the B ⊆ A. This is achieved if co-channel interfering cells are not
review of this letter and approving it for publication was A. Banchs. fully loaded and therefore not all subcarriers are needed for
A. Fernekeß and A. Klein are with Technische Universitaet Darm-
stadt, Merckstr. 25, 64283 Darmstadt, Germany (e-mail: [email protected] transmission. A measure for the load l is given by the ratio
darmstadt.de). of the cardinalities of B and A given by
B. Wegmann, K. Dietrich, and M. Litzka are with Nokia Siemens Networks
|B|
GmbH & Co KG, Munich, Germany. l= (2)
Digital Object Identifier 10.1109/LCOMM.2008.080032. |A|
1089-7798/08$25.00

c 2008 IEEE
FERNEKESS et al.: LOAD DEPENDENT INTERFERENCE MARGIN FOR LINK BUDGET CALCULATIONS OF OFDMA NETWORKS 399

with 0 ≤ l < 1. For l = 0, the system is noise limited and there 4.5
is no co-channel interference. For l = 1, the system would be QPSK
16QAM
4
interference limited and all subcarriers are interfered by co-
channel interfering cells. Due to assumption of a constant SIR, 3.5

channel capacity in bits/(s Hz)

no optimization can be performed in that case.
3
It is assumed that the interference power density, is larger
than the spectral noise power density I  N . Therefore 2.5

subcarriers with index k ∈ A \ B experience an SNRnl which 2

depends on the cell radius, subcarriers with index i ∈ B
experience an SIR given by (1). Interference power as well 1.5

as noise power is assumed to be Gaussian distributed [6]. 1

Corresponding to a normal link budget calculation, a noise
limited scenario with l = 0 is assumed. An SNRreq can be 0.5
0 2 4 6 8 10 12 14 16 18
SNR in dB
achieved at the cell border by adjusting, e.g., the cell radius.
In a second scenario, interference from neighbouring cells
Fig. 1. channel capacity as function of the SNR for QPSK and 16QAM
is considered on part of the subcarriers thus 0 < l < 1. modulation.
The performance at the cell border of the scenario including
interference should match the performance at the cell border TABLE I
of the noise limited scenario. Therefore, it is obvious that S IMULATION PARAMETER
the SNRnl at the cell border has to be readjusted if SIR =
Parameter Value
SNRreq . The difference between SNRnl and SNRreq gives the System bandwidth 10 MHz
interference margin that has to be considered in the link budget Subcarrier spacing 11.16 kHz
calculation to get reliable maximum cell radius estimations for symbol duration 95.2 µs
Channel Model Additive White Gaussian Noise
the network planning. SIR 0 dB (reuse 1); 8 dB (reuse 3)
Modulation and coding QPSK 1/2, 16 QAM 1/2, 16 QAM 3/4
III. I NTERFERENCE M ARGIN
To determine the performance at the cell border, the chan- It should be noted that for l = 1, Creq is limited by Cil
nel capacity [8] is used throughout this paper. The channel which is assumed to be independent of the cell radius in this
capacity may be expressed in the form work. The minimum SNR at the cell border in a scenario with
I(X; Y ) = h(Y ) − h(Y |X) (3) interfering co-channel cells is given by the inverse function
of f which is valid for channel capacities smaller than the
with X and Y the input and output signal, respectively, maximum achievable channel capacity with that modulation:
using a certain modulation and h(·) the entropy function [8].
The channel capacity for PSK and QAM modulation can be SNRnl (l) = f −1 (Cnl (l)). (8)
obtained using the Monte Carlo method [9]. It is not possible The IM depending on SNRreq and l is given by
to give a closed form solution for the relationship between
SNR and channel capacity for QPSK and 16QAM modulation, IM(l) = SNRnl (l) − SNRreq . (9)
so that we will use the expression
IV. S IMULATION R ESULTS
C = f (SNR) (4)
Link level simulations according to the IEEE 802.16e
throughout this paper. The function can be found in Fig. 1. standard [5] are performed. In this section, the results will
In a noise limited scenario with l = 0 an average capacity be compared to the IM model developed in section III. The
per subcarrier convolutional code of [5] is considered. Decoding is performed
Creq = f (SNRreq ) (5) using the Viterbi algorithm [10]. 288 bits are coded and
can be guaranteed at the cell border by adjusting the transmit interleaved together in one code block according to [5]. In
power and the cell radius. The SNR achieved at the cell border each code block a fraction l of the subcarriers experiences a
in this configuration is termed SNRreq . predefined SIR. The SNR experienced on the remaining sub-
The average capacity per subcarrier achieved in a scenario carriers is variable. The IM is the difference in SNR between
where l · |A| subcarriers are interfered by co-channel cells is a scenario with l = 0 and l = 0 when the same performance
given by for the user can be achieved. The performance is measured as
C(l) = Cnl · (1 − l) + Cil · l (6) transmitted bits per code block based on the achieved bit error
performance. The main simulation parameters can be found in
where Cnl is the capacity achieved on subcarriers interfered by Table I.
noise and Cil the capacity achieved on subcarriers interfered Fig. 2, 3 and 4 show the results of the IM as function
by co-channel cells. To get the required average capacity Creq , of the required SNR at the cell border for transmission with
Cnl has to be adjusted so that QPSK 1/2, 16QAM 1/2 and 16QAM 3/4, respectively. It can
Creq − l · Cil be seen that the analytical model from section III fits well to
Cnl (l) = . (7) the results gained from simulation if the required SNR is not
1−l
400 IEEE COMMUNICATIONS LETTERS, VOL. 12, NO. 5, MAY 2008

4 4
l=0.1 analyt. l=0.1 analyt.
l=0.2 analyt. l=0.2 analyt.
3.5
3.5 l=0.5 analyt. l=0.5 analyt.
l=0.1 sim. l=0.1 sim.
l=0.2 sim. 3 l=0.2 sim.
3

interference margin in dB
l=0.5 sim. l=0.5 sim.
interference margin in dB

2.5
2.5

2
2

1.5
1.5
1
1
0.5
0.5
0
0 2 4 6 8 10 12 14
0 required SNR in dB
0 1 2 3 4 5 6
required SNR in dB

Fig. 4. Interference Margin for 16QAM rate 3/4, reuse 3 (SIR = 8 dB).
Fig. 2. Interference Margin for QPSK rate 1/2, reuse 1 (SIR = 0 dB).

at the cell border. It was shown that a realistic margin can

4
be given to achieve the same performance at the cell border
l=0.1 analyt.
l=0.2 analyt.
in terms of channel capacity, if some subcarriers experience
3.5
l=0.5 analyt. high interference from co-channel cells, as in a noise limited
l=0.1 sim.
3 l=0.2 sim. scenario. Of course, this performance cannot be achieved if the
interference margin in dB

l=0.5 sim.
2.5
required SNR is high compared to the SIR, the load is high
in co-channel cells and higher order modulation and coding
2
schemes shall be used. This is also reflected in the proposed
1.5 model. Due to the assumption of a constant SIR the model
1
leads to lower bound results. If the interference power density
gets close to the spectral noise power density an SINR > SIR
0.5
has to be considered in (7) which is no longer independent
0 of the cell radius. Nevertheless, the proposed model helps
0 2 4 6 8 10 12 14
required SNR in dB to improve the link budget calculation due to taking load
dependant inter-cell interference for OFDMA networks into
Fig. 3. Interference Margin for 16QAM rate 1/2, reuse 3 (SIR = 8 dB). account. The estimation of the maximum possible pathloss in
the link budget calculation gets more accurate.
high compared with the SIR and l is small. Due to interleaving R EFERENCES
and coding over a certain number of bits, a high SNR on
[1] A. R. Mishra, Fundamentals of Cellular Network Planning and Opti-
some symbols can compensate a low SIR on the other. But misation. John Wiley & Sons, Ltd., 2004.
the interference of part of the subcarriers leads to an error floor [2] J. Laiho, A. Wacker, and T. Novosad, Radio Network Planning and
in the bit error curves [11]. For instance, QPSK 1/2 requires Optimisation for UMTS. John Wiley & Sons, Ltd., 2002.
[3] J. Zander and K. S.-L., Radio Resource Management for Wireless
an SNR of 5 dB to achieve a bit error rate less than 10−6 Networks. Artech House, 2001.
according to [5]. Therefore, QPSK 1/2 can be guaranteed at [4] G. Parsaee and A. Yarali, “OFDMA for the 4th generation cellular
the cell border for a scenario with SIR equal to 8 dB. On the networks,” in Proc. Canadian Conference on Electrical and Computer
Engineering, Niagara Falls, Canada, May 2004, pp. 2325–2330.
other hand QPSK 1/2 is not feasible in an interference limited [5] IEEE standard 802.16e-2005, “Air interface for fixed broadband wireless
scenario with SIR equal to 0 dB and l > 0.2 as seen in fig. access systems,” Feb. 2006.
2. This is also covered by the developed model. [6] S. N. Moiseev, S. A. Filin, M. S. Kondakov, A. V. Garmonov, D. H. Yim,
J. Lee, S. Chang, and Y. S. Park, “Analysis of the statistical properties of
the interference in the IEEE 802.16 OFDMA network,” in Proc. IEEE
V. C ONCLUSION Wireless Communications and Networking Conference, Las Vegas, USA,
Apr. 2006, pp. 1830–1835.
An analytical model to consider interference in the link bud- [7] K. Hiltunen and N. Binucci, “WCDMA downlink coverage: interference
get calculation for OFDMA based wireless networks is pro- margin for users located at the cell coverage border,” in Proc. IEEE 55th
posed in this paper. In the considered scenario, it is assumed Vehicular Technology Conference (VTC Spring), Birmingham, USA,
May 2002, pp. 270–274.
that some subcarriers are interfered by co-channel interfering [8] T. M. Cover and J. A. Thomas, Elements of Information Theory. John
cells while the remaining subcarriers experience only thermal Wiley & Sons Inc., 1991.
noise. The same performance in terms of channel capacity can [9] B. M. Hochwald and S. ten Brink, “Achieving near-capacity on a
multiple-antenna channel,” IEEE Trans. Commun., vol. 51, pp. 389–399,
be guaranteed at the cell border as in a noise limited system Mar. 2003.
if the SNR is improved by a reduction of the cell radius. [10] A. J. Viterbi, “Error bounds for convolutional codes and an asymp-
Link level simulations for IEEE 802.16e OFDMA show that totically optimum decoding algorithm,” IEEE Trans. Inform. Theory,
vol. 13, pp. 260–269, Apr. 1967.
the results from the proposed model are accurate so that the [11] S. Plass, A. Dammann, and S. Kaiser, “Analysis of coded OFDMA in a
influence of load in co-channel cells can be considered in the downlink multi-cell scenario,” in Proc. International OFDM-Workshop,
link budget calculation to get an estimate for the required SNR Dresden, Germany, Sept. 2004.