You are on page 1of 6

Cross-tier Interference Mitigation for Self

Organized-Self Optimized Femtocells
Massa Ndong, Takeo Fujii

Advanced Wireless Communications Research Center
The University of Electro-Communications
Email: {massa.fujii}

Abstract-Maximizing frequency reuse is of high interest to
multi-tier wireless cellular networks. The coexistence of macrocell
and femtocell requires advanced cross-tier interference mitigation
to satisfy the spectral efficiency sought for the evolution of
cellular networks.

In this paper,

our proposal combines an

opportunistic cancellation of the interference at the base station
of the femtocell(HeNB) which mitigates dead zone occurrences
at the femtocell premises in closed access regime, and a transmit
power limitation on the femtocell mobile user for interference
avoidance at the base station of the macrocell. Our proposed com­
bination profits from the network infrastructure where successive
interference cancellation does not apply, and neighbor discovery
enabled at the HeNBs. Through simulation results and analysis,
our proposal improves the underlay network performance with
implementation of a self organized-self optimized network.

I1Idex Terms-Macrocell, femtocell, two-tier networks' MMSE ,
Cross-tier interference mitigation, power control, SON



Contemporary wireless operators are in the process of
deploying short range cellular systems called femtocells. A
femtocell tier is made of the base station of the femtocell and
its connected mobile terminals denoted as home users(HUEs).
The base station of the macrocell(eNB) and its related mobiles
denoted as MUEs constitute the macrocell tier. The communi­
cation between the HeNB and the eNB goes through a wired
backhaul. Femtocells are deployed to satisfy the increasing
demand of indoor data rate under the condition of mitigating
the cross-tier interference.
As an underlaid system to the macrocell, the femtocell
needs to handle the cross-tier interference from the surround­
ing MUEs in addition to avoiding the interference to the
eNB [1] [2]. To address the issue, sensing and adaptive power
allocation are additional considerations urged by cross-tier
interference management. In [ 3], a rate optimization has been
considered with a power control and a soft sensing by a
secondary system (SS) terminal to avoid any impairments at a
primary system (PS) base station. Optimizing the subchannel
allocation in a decentralized frequency planning macrocell em­
bedded with femtocells is considered in [4]. Self optimizing of
SONs is considered [5]. The proposed coverage coordination
scheme based on the measurement results of received signals
simultaneously improves an indoor coverage while avoiding
impeding outdoor leakage. However, the derived HUE transmit
power is not based explicitly on its positioning angle W.r.t
978-1-4577-0580-9/12/$26.00 @2012 IEEE

to the HeNB and the distances from the HUE and the base
stations. Such parameters can better determine the mobile
terminal as HUE or MUE. Furthermore, signal powers and
channel gain estimations are required at HUE in [5] while
our simple interference avoidance from each HUE profits
more of the idea of a mobile user confined in the femtocell
vicinity, and is not loaded with power estimation and extra
neighbor discovery which is already a feature of the HeNB
active at its initialization. In [6], the signal of a strongly
interfering terminal is canceled out by feedback using the
infrastructure connecting base stations from different cells.
The delay effect which might be induced by the feedback is
considered in contrast with our work. However, our simulation
uses the Recursive Least Square (RLS) for the interference
cancellation differing from the scheme in [6]. Indeed, in the
signal recovery process, the channel of the user whose data
have to be recovered is solely estimated at the receiver. In this
paper, a general analysis is given with the conventional MMSE
using the user's channel gains for theoretical performance
evaluation. Furthermore, [6] resorts to joint multi-antenna
processing while our receiver considers a single antenna for
the wireless interface. Additional considerations on the effect
of potential backhaul overhead can be found in [7]. Therein,
the traffics from both the transport and the radio layer are
jointly used to optimize the overall network throughput. The
proposed interference cancellation in this paper extends the
interference cancellation using infrastructure network in [8]
by stating the transmit power and position of each terminal.
These parameters are used to derive further analysis and to
propose the mitigation of the interference coming from both
This paper considers uplink spectrum sharing usage between
a PS constituted by the macrocell, and an SS represented by
the femtocell. We assume that the femtocell operates in a
closed access, i.e. only an HUE is allowed to communicate
with the considered HeNB. We propose an interference can­
cellation and a power control schemes which applied together
implement a SON scheme for the femtocell:

the proposed interference cancellation at the HeNB uses
the macrocell information feedback through the two-tier
network infrastructure. The interference signal from MUE
is fed back from the eNB to the HeNB to mitigate the
interference from the macrocell tier to the femtocell tier.

(I. Macrocell embedded with femtocell in Uplink interference. Proposed HeNB receiver structure. The illustration is given in Fig. V indicates vector transpose. YI is the combination of the HUE and MUEs signals while the Yi for i = 2. PROPOSED INTER FERENCE CANCELL ATION AT THE HENB In this part.. We propose to combine an interference cancel­ lation at the SS which cancels out the signals from the MUE to the HeNB. hi is the channel gain between the HeNB and the ith MUE modeled as an i. The PS is represented by the Macro-tier while the Femto-tier illustrates the SS.. and a power control based on the parameters D.MandY= [ YI Y2 Y3 YM ] Tis the received vector at the HeNB. We propose a general analysis of our proposed interference cancellation with multiple interfering MUEs and one HUE transmitting the symbol XI on the channel with fading coef­ ficient hI. section IV emphasizes on the proposed power control scheme at the SS and the paper ends with the conclusion in section VI. hp between MUE and eNB and hs between HeNB and HUE. M are the fed back signals from the eNB. We suppose that the data of the MUEs can be fed back to the HeNB. III. . and the femtocell operates in a closed regime i. the femtocell owner does not allow a non registered mobile to be connected to the HeNB. where [.i. The computer simulations confirm that the throughput and BER can be improved in our proposed cross-tier interference mitigation. For the purpose of the analysis. d and the transmit power of the MUE. HUE o o (I. section III is dedicated to the proposed interference cancel­ lation at the HeNB. we provide an extension to [8] considering the variation of the HUE's position and a general analysis. -----.l by the crossing of the line representing the transmission from the HUE to the eNB. is the distance between HeNB and HUE.d Rayleigh fading coefficient with zero mean and unit variance and estimated from pilot symbols where i=2.1.+h"S. . .<::-?��/// Fig. .. The reminder of this paper is organized as follows: Section II describes our underlaid two-tier network model.e. A view of such coexistence is given in Fig. D is the distance between HeNB and MUE. The interference from the MUE to the HeNB is canceled out while the HUE interference to the eNB is avoided. �o Co �. Sp and Ss respectively stand for the transmitted symbols from the MUE and the HUE.� . it derives a suitable HUE's transmit power under a given threshold based on the Fig. distances and power parameters. the cross-tier interference from the macrocell to the femtocell is illustrated by the transmission of the MUE intended to the eNB which can be close enough to the HeNB to create a dead zone at the femtocell. We consider Uplink transmission and the interference is defined as follows: the transmission from the HUE intended to the HeNB interferes at the eNB if the HUE's transmit power is not regulated pertaining to its position.. If we assume the knowledge of geographical position of the eNB and the estimation of angles positioning the HUE. The channel gains are denoted as hps between MUE and HeNB. h. we consider the following notations: M-1 stands for the number of MUEs.-. We consider spectrum sharing between the two tiers. the HeNB derives the suitable transmit power for the HUE as a function of distances separating the terminals and the MUE's transmit power.. 2. This part of our proposal can be seen as complementary to the neighbor discovery performed by the HeNB at its initialization.t. and d.S. . MUE HeNB eNB • II.. /� o -.1/ /..The symbols sent by the HUE can be recovered from the interfered signals. The power control can be seen as an interference avoidance scheme mitigating the interference from the HUE to the eNB. MACROCELL EMBEDDED WITH FEM TOCELL UNDERLAID NETWORK MODEL Our network model is an underlay system. !. The proposed method enhances the femtocell throughput by canceling out the interference from the MUE. the power control scheme is proposed as follows: the HeNB uses derivations from its received power mea­ surements to adapt its operating parameters concordant with the interactions with the PS.

R= S+ � · The diagonal elements of H being non zero. H is non singular. The resulting digital signal Tj and the network infrastructure signal Tm are the inputs to the detector module encompassed in the dashed line box in Fig.Using matrix notation. we propose to use Wiener combining techniques in the spectral domain to provide the linear output estimating Ss with . The noise added interfered received signal hpsSp+hsSs+n is converted in the analog to digital converter (AI D).l and using the RLS algorithm. Using matrix notation.d with zero mean and unit variance. the particular case of one HUE under one MUE.:. single-symbol detection is equivalent to nulling out the interference. We consider in the following analysis. (8) is Rayleigh faded such as in a 1 x 1 channel. the feedback makes them separable at the HeNB. Each column of H is the channel matrix of one user. By using the notation Y=HX+N.:.i.:.N = (�) .- ---. The decoupling of the two symbols is given by: and N = [n 0 0 o r n ( ) hs 0 ' I H.j = Ss+ 1fhJ' (7) where z has zero mean and its variance is No. (2) and (3). the second term of Tj has the variance E { 1 ::.:. we obtain: n T.2. Since the columns are linearly independent. we consider Eqs. (1) (3) Referring to Fig. Therefore. (8) We can notice that the scalar channel in Eq.12} : (6) Tj can be rewritten as: z T. thereby averting the estimation of interfering users' channel gains.1 0 0 o -h3 -hM -. where o o o H= 0 0 0 (�s ) hs f . Scaling Tj by the factor multiplying z gives the following: Tj = Ilhsll Ss+ z. X I can be extracted from the first component of H-IY. resulting into Xl + ::. . S= (�.:.2: + XM 1 (2) 0 where Xi is the transmitted symbol from the ith user and n is an AWGN. we can write the received signals as a received vector R: where H= h I h2 h3 0 1 0 0 0 1 0 0 0 1 hM 0 0 0 which can be rewritten as: R=HS+N. To avoid noise enhancement while mitigating the interfer­ ence. s (5 ) Denoting by No the variance of n. single-symbol detection can be performed. The utilization of the network infrastructure feedback offers signaling space design.j = Ss+ h. the HeNB is able to invert H in order to recover the desired symbol which is Xl in this case. The signal space is then M and the receiver can differentiate between the signals. Besides. H being non singular and triangular. it is upper triangular. I H. and the fading coefficients hs and hps are i. The proposed HeNB receiver where If we denote by Tj the first component of H-IR. it admits an inverse matrix H-I which is upper triangular as well: I h. we have at the HeNB: YI Y2 Y3 YM h I h2 h3 0 1 0 0 0 1 0 0 0 1 0 0 0 hM 0 0 0 0 0 0 n Xl X2 0 0 X3 structure is illustrated in Fig. Considering the two trans­ mitters as mutually interfering users.:.) (4) .= 0 0 0 o -h2 -. While the signals from the wireless channel interfere with each other.- 0 0 1 0 0 0 o 0 0 0 1 1 The directions of arrival constituted by the channel of each transmitted signal in the columns of H are linearly independent vectors.

Pl. = (19) where Pl is the indoor penetration loss. The demodulation of S8 gives the data symbols of the HUE.r. we have: CD )£2 . Since hCd and hCD are not correlated. Given that the fading channels can be modeled each as a Single Input Single Output channel.2LDcos(e) + D2. (17) Following the same derivation method with a at the eNB and e at the HeNB in the triangle formed by eNB. Jensen's inequality provides a lower bound of the average SIR: E[SIR] > [ ] ] PhE lhCD12 CDclKm . MUE). The SIR in this section is defined as the ratio of the symbol strengths of the MUE over the HUE. (20) at the eNB: we denote by If the interference from the HUE located at distance Cd from the eNB. Denoting by S8 the estimated value of S8 and d as the desired chosen signal correlated with the inputs of the equalizer. • 1m Pm -lOclloglO(D) . we propose an interference avoidance from one HUE to the eNB. d. This leads us to: (dsin(e))2 = C/ .) indicates ensemble average.cland flare respectively the macrocell and the femtocell path loss components. At each base station. MUE and HeNB. The SIR is represented as: SIR = Ph IhcDI2 Ci/Km f pm IhCd12 Cd lKf ' (12) where 1. and by Rf the received power from the HUE located at distance d from the HeNB: Rf • = Ph -lOfllog10(d) . HUE and HeNB. We approximate the SIR by its average to derive the proposed power control scheme: = = SiR = PhCDclKm . l .l . (16) Cd is therefore obtained as: Cd = )£2 .1 indicates the amplitude. MUE) and Cd (resp. we have an interfering and a desired signals following [ 4] : at the HeNB: we denote by 1m the interference from the MUE located at distance D from the HeNB. ) H denotes the complex conjugate transpose vector and I is the unit matrix having the same dimension as H. PROPOSED POWER CONTROL FOR SON Considering the proposed spectrum sharing two-tier network system. we can derive Cd and CD: dsin(e) = Cdsin(a). Denoting by W the weights matrix derived from the equalization.) denotes the complex conjugate vector. considering the triangle formed by 0 at the eNB. the values of D and d can be computed at the HeNB through energy detection of the neighbors [ 1 0] and the interference patterns defined in this section. the . Land e are available at the HeNB.Pl. The estimated signal X for the general analysis or S8 in the application involving the HUE and the MUE is obtained from the MMSE equalization as follows: (11) The weights can be generated through the recursive least square technique which avoids direct matrix inversion. we can derive the suitable transmit power for the HUE to avoid its interference at the eNB.37. MUE) and the eNB. We derive the SIR at the eNB using the signal strengths of both MUE and HUE.37. lKf (14) Given the parameters d. With the macrocell embedded femtocell parameters given in Table II. (22) Replacing the constant loss terms kf and km from the inter­ ference patterns. f PmE Ihcdl2 C:. hCD) between the HUE (resp. (18) = Given an SIR constraint on HUE's transmit power. from [ 9] we have: (10) where ( . lKf [ ( 1 3) where E(. CD) is the distance between HUE (resp. The proposed power control implemented at the HeNB regulates the HUE transmit power pertaining to D. MUE) and the eNB.37. If = Ph .Pl. we obtain the approximate SIR: SiR = PhCDcl -30loglO(je) + 71 f PmC:.37 (23) Our proposed decentralized power control at the femtocell is based on: (24) where the constraint is restrictive in the sense that we consider only two users mutual interference. Ph (resp. Km) is the constant power loss for the HUE (resp.2Ldcos(e) + d2.lOfllog lO(Cd) . the random variables are such that E(lhcJ 12 E(lhcD)12 l. L and the angles giving the positions of the HUE and MUE w. From this model. and Wf and Wm are the weights minimizing the error: E{ S8 d 2}. (21) and by Rm the received power from the MUE located at distance CD from the eNB: Rm = Pm-lOclloglO(CD)-30logl0(je)+71. f PmC:. * ( _ ) The MMSE equalization optimally trades off mitigating the noise and the interference. (15) where a is the angle at the vertex 0.Minimum Mean Square Error (MMSE) which is more tolerant to noise. The channel gains are represented by hCd (resp. we can obtain: (9) where (. IV.t to the HeNB in Fig. Pm) is the transmit power of the HUE (resp. Kf (resp.(L-dcos(e))2.

packet) error rate is obtained by the ratio of the correctly receives bits (packets) over the overall transmitted ones. in addition to increasing femtocell capacity [ 1 2]. Nsym is the number of data symbols. As for obtaining the angle e several methods can be advocated from the literature. In the femtocells realm. Corrections of Pth can be done by requesting received power at the eNB from the HUE which can be forwarded to the HeNB through infrastructure network. This cooperation is similar to [11] where the macrocell base station assigns additional resources to the femtocell base station through the backhaul.1 for the numerical results obtained from our proposal and its related conventional scheme. the performance of the MUE improves as D increases.(Pth . from Eq.dDconv ---e--- Performance evaluation with the proposed protection of the eNB from the HUE T (1 . B. V. we derive the maximum transmit power of the HUE denoted as Ptfmax: Ptfmax Pm .25E-6s OFDM QPSK Rayleigh flat fading RLS AWGN 112 7 10' r---�---�--�----' PERFORMANCE ASSESSMENT T HROU G H SIMULATIONS We consider Fig. dDproposed 10' �--�---�--�-----' 5 10 15 5 20 SIR (dB) A. (26) where Pe is the packet error rate. it can be utilized to estimate the angles used for the purpose of our interference avoidance to the eNB. 10' r---�--�---�-��====�==� ---e--- dDproposed ----A. Its half power-bandwidth can span within � radians and the beam can be switched to the user location. After characterizing such interference as a fixed loss varying at a large scale.3 through each terminal's average symbol energy decay based on D and d replacing respectively Cd and CD in Eq. Therefore. 3. Both terminals have the same data modulation. Interference cancellation evaluation considering the pro­ posed power control The throughput in Fig. the E­ plane horns based reconfigurable antenna presents low power side lobe and dynamically switchable main beam.dDConv 10 ��L O -_�1 5=-- -- _1�0 -- _5 �'0 ---�--�--�-- SIR (dB) Fig.TABLE I SIMULATION CONDITIONS interference level of neighboring base station can be known to the HeNB. Bit error rate of the MUE with variable d and D considering the power control. the random cross-tier interference is left to be dealt with.37 . nmod is the number of bits per data symbol. dDconv illustrates the throughput of the MUE at the eNB under an HUE interference without the proposed power control while dDproposed represents the performance with the interference avoidance.4E-6s 1.Pi). The simulation parameters related to the macrocell-femtocell two­ tier network are represented in Table I.5 illustrates the interference can­ cellation performance and the power reduction effect due to . = (25 ) If Pth decreases. The analysis is similar for dDconv and dDproposed considering the bit error rate in 4.Pe)rNsubNsymnmod/Ts. The MUE's throughput steadiness illustrates the power control effect.(14). Our simulation uses the packet error rate to estimate the throughput denoted by T as derived in [8] : Parameters Bandwidth Number of subcarriers Useful symbol time Guard interval Data modulation small scale channel model Weight estimation algorithm Noise Convolutional code rate Convolutional code constraint length Fig.10logI0(CD) .10logI0(Cd) . The varying SIR between the HUE and MUE terminals is obtained in Fig. r is the channel coding rate and Ts is the symbol time. 4. The simulation results illustrate the power control performance with varying D and d which are normalized distances chosen randomly between 0 and 1.30log10(fc) + 71 . The bit (resp. Throughput of the MUE with variable d and D considering the power control. Therefore. = Value 5MHz 512 6. -----A. Nsub is the number of subcarriers.(24). Alternative methods for the estimation of the angles can resort to indoor global positioning system.

the random positions of mobile terminals connected to either tier generate a cross-tier interference issue which has to be handled for a successful implementation of femtocells." EURASIP J -15 -10 -5 SIR (dB) 10 Fig. 0 10' VI. Feb 2011. M. pp. 2011. Li. M. G. vol.r..70. 1. 1-6.. L. 2010. Oct. "Joint load balancing of radio and transport networks in LTE system. [4] V. Wang and R. and Netw. Negative and low SIRs represent the presence of the MUE in the femtocell minimal area or an MUE transmitting in the vicinity of the HUE while being far apart from the eNB. Balachandran. 2010. [3] V. Our paper has addressed the design of SON by handling the cross-tier interference from both tiers mainly by the means of the femtocell. 10' � E-< => '" :c '" 10' => 0 '" :c E-< 10° base station assignment.. the proposed interference cancellation effectively mitigates the interference from the MUE to the HeNB.. 1. Zhao. 65 . Nevertheless. volA7. Yavuz et aI.. This is similar to the MUE's. [7] L.1155/2010/285714. and Comput." High Capacity Fem­ networks. Oct 2010. Timm-Giel and C. tocells with Directional Antennas. M. Jul. The next generation cellular system gains in area spectral efficiency with the deployment of femtocells into macrocell in a spectrum shared usage. Mun. Hwang.9.8 3 1O. no. Hwang.lO. It defines the maximum transmit power of the HUE by means of the measurements of its angular position and distance towards the HeNB. pp.In addition to comparIng the the interference avoidance. TABLE II HUE AND MUE CELLULAR PARAMETERS. Huang. A."IEEE Trans. The simulation results confirm the resilience of the cross-tier overall operative mitigation. Asghari and S. Fujii. From the femtocell side. doi:10. [10] V. 2009. Chandrasekhar and 1. "Intelej rence cancellation for spectrum shared femtocell networks with macrocell information feedback. "Implementation of the Levinson algorithm for MMSE equalizer" in SoC Design con!. [12] A. Jo. Yook. Lee and Y. Wireless Commun. M. Oct. At high SIRs." in Int. [9] M. "Self­ EURASIP J Wireless Commun. 2008. Aissa. Kim. pp. Nov. Wireless Commun. con! on Ubiquitous and Future Netw. G. "Interference management and performance analysis of UMTSIHSPA+ femtocells. Further analysis and results van be found in [8]." IEEE Trans. Weerawardane.. Ndong and T.t to the power reduction due to the proposed power control affecting the HUE. Moon and J.95 2500 MHz 5dB 23dBm 3. 5314-5327. vol. Chandrasekhar. Gorg. Rege. pp. Value Parameters(Variable) Macrocell radius (Re) femtocell radius (Rf) normalized distance from HeNB to eNB (L) Carrier Frequency (fc) Wall penetration loss (Pz) Mobile maximum transmit power (Pt'max) Macrocell path loss (cz) femtocell path loss (fz) SIR threshold(Pth) MUE_angle HUE_angle 1000m 30m 0.115512010/240745.IO. Kountouris and J. 2010. "Self-optimized coverage coor­ dination in femtocell networks. Syst. 2011.8." in Int.. The effect of the power control is illustrated by HUE. J."in con!(WCNC). HUE's throughput from the power control and the interference cancellation. 5.9. Torregoca. It uses the backhaul to feedback the MUE's symbols to the HeNB which can then recover its signal of interest from the received aggregated signals of the HUE and MUE." JIEEE Commun. Andrews.. "Adaptive rate and power transmission in spectrum-sharing systems. 8 pages. Tsai. 2008. 1583 . From the macrocell. . [6] K..229. Kang.. Oct. . Enkhbat and W. It also provides some insights for the self organizing-self optimizing feature of SONs. [II] P. we proposed a power control which limits the hindrance of the HUE to the eNB receptions. Jun. con! on Ubiquitous and Future Netw. Mag. Karakayali and K. Wireless Commun. Kim. "Spectrum allocation in two-tier networks. 3272-3280 .(25). K. lO. 102-109.2. pp. no. T. No. pp. the combined proposals leads to a similar maximum throughput for both HUE and MUE singly w. [8] M. "On uplink inteljerence sce­ narios in two-tiers macro and femto co-existing UMTS networks. HU Epro illustrates the HUE performance with the interference cancellation without the effect of the power control. J." in 42nd Asilomar con! on Signals. vol. pp.III-16. Jun." IEEE Trans. [2] Z. 2009. H. Netw. D D D D o 0 0 0 Wireless Commun. C. and channel assignment in cognitive femtocell Wireless Commul1. 224 . [5] H. Reed and M. Shi. doi:IO. lO. "Joint power control.. the HUE performance is given by HU Econv. Zhao. 10' CONCLUSION REFERENCES [I] M. C. Netw. vol. 14 pages. III-15 . R. optimized coverage coordination in femtocell networks."IEEE Trans. Andrews. "Coverage in multi-antenna two-tier networks. 0 G- '" ---'-' .16dB 7r 4 O"'i HUE and the MUE respectively.. Wireless Commun. J. Compared to the conventional method. X..1587. Without our proposal. Article ID 285714. No. the interference is canceled out by the proposed interference mitigation leading to a throughput curve remaining horizontal for the proposed method at a level permitted by the noise power and the protection of the eNB. 2010." Article ID 240745. we applied our proposed interference cancellation in Section III to the HUE controlled through Eq. pp.