An LTE-Femtocell Dynamic System Level Simulator - IEEE Xplore

2010 International ITG Workshop on Smart Antennas (WSA 2010)

An LTE-Femtocell Dynamic System Level Simulator Meryem Simsek, Tarik Akbudak, Bo Zhao, Andreas Czylwik University of Duisburg-Essen Bismarckstrasse 81, 47048 Duisburg Email: {simsek,akbudak,bo.zhao,czylwik}@nts.uni-due.de Abstract—A dynamic system level simulator for LTE networks was developed for investigating the interference behaviour of femtocells placed within macrocells. We simulate a multicell, multi-user and multi-carrier system in the downlink for Single-Input, Single-Output (SISO) and Multiple-Input, MultipleOutput (MIMO) antenna configurations. Index Terms—System level simulations; LTE; Femtocell

Initialization Snapshot number

k=0

Mobility: position update

Traffic update

I. I NTRODUCTION Channel calculation

978-1-4244-6072-4/10/$26.00 ©2010 IEEE

k=k+1

The demand for high data rates in wireless communications is increasing tremendously. Both the coverage area and the capacity of existing cellular network systems are not sufficient to meet the expected demand of multimedia traffics. Since a small distance between transmitter and receiver in a wireless system increases the capacity of this link and creates dual benefits of higher quality and more spatial reuse, the femtocell approach seems to be the most beneficial one. Femtocells use low-power, low-cost cellular base stations. They are typically deployed indoors to improve coverage and provide high data rates. There are a number of economic factors that argue strongly in favour of femtocells, including the capacity improvements that are enabled by allowing mobile stations (MS) to transmit with very low power to a femtocell. Therefore there is high interest in femtocells among network providers and equipment manufacturers. One of the most interesting open technical questions concerns how to optimize the performance of networks including femtocells in light of the interactions between femtocells and MSs that may be near, but not communicating with the femtocell [1]. In order to deploy and operate femtocell networks properly, several issues like, e.g., system architecture, RF-related issues and interference control should be considered [2]. In this paper, we focus on the interference problem between macrocells and femtocells. We discuss the effect of interference for the OFDMA/FDD based mobile LTE system by dynamic system level simulations (SLS). Several works about LTE system level simulations like [3] have been published. In contrast to published system level simulators our simulator is totally dynamic. It provides the opportunity to choose different antenna configurations for both macro and femto stations independently. Papers concerning femtocells have also been published [4] [5] [6]. Most of these publications deal with WIMAX femtocell system level simulations that are, different to our presented simulator, in TDD mode or totally static.

SINR calculation

Link level quality estimation

Scheduling and power control

Evaluation

Fig. 1.

Simulation flow.

The rest of this paper is organized as follows. The components of the system model and the performance metrics are described in section II. The simulation parameters, their results and interpretation are presented in section III. Finally, the paper ends with some concluding remarks and consideration for future works. II. S IMULATION C ONCEPT A. Simulation Flow Figure 1 shows the simulation flow. Our simulator continuously simulates the temporal development of cells by calculating densely spaced snapshots k. Within each snapshot, the position of the macro mobile stations (MS) changes and the traffic situation is updated. Next, the channels between each femto/macro mobile station and each femto/macro base station are newly calculated and scheduling is performed for each snapshot. UMTS-LTE allows different timing granularities [7]. In this simulation we use the time interval T = 1 ms. B. Initialization In the first part of the initialization procedure all the parameters that remain constant during the whole simulation

66

are set. The following parameters can be set: • number of macro base stations (BS) / number of macro sites, • inter-site distance, • number of macro MSs per macrocell (changes due to mobility, but total number of macro MSs stays constant), • velocity of macro MSs, • number of macro transmit and macro receive antennas, • femtocell scenario, • number of femto blocks per cell, • number of femto transmit and femto receive antennas, • bandwidth, • carrier frequency, • scheduling algorithm. A sectorized system with 3 sectors per Bs is considered. A cellular configuration consisting of a main area and six surrounding areas is created for simulating the effects of intraand intercell interference in the 3 sectors that are located at the center as well as at the boundary of the main area. For the surrounding areas a wrap-around technique is used. The macro mobile stations are distributed uniformly in the whole cellular environment. Although the SINR evaluation is done only for mobile stations in the main area, because of the wrap-around technique the effect of equivalent mobile and base stations of all other areas has to be considered. Thus, a base station of the main area and its equivalent base stations in all the other areas same transmission powers. The femto blocks are randomly inserted within each macrocell. Femto base stations and the femto mobiles within these femto blocks are also randomly located, taking into account a minimum distance of mobile stations and femto base stations. The following femtocell scenarios are available [8]: 1) Femtocell scenario 1 - Suburban scenario: Femto blocks are dropped within the macro coverage area with a random uniform distribution, subject to minimum distance to the macro sites. A suburban type deployment is considered where each femto block is modelled as a (2-dimensional, 12 m × 12 m) rectangular femtocell/house. Within each house the femto BS and femto MSs are randomly dropped. The number of femto MSs varies between four and eight randomly. With a probability p a femto MS might be outdoors. This scenario enables to investigate the effect of femtocells within macrocells by altering the number of femtocells per macrocell. 2) Femtocell scenario 2 - Dense urban dual stripe model: In a dense urban femtocell modelling, each femto block represents two stripes of apartments, each stripe has 2 by N apartments (see Figure 2). Each apartment is of size 10 m × 10 m. There is a street between the two stripes of apartments, with width of 10 m. In each macrocell one or several femto blocks are randomly dropped. It is assumed that femto blocks are not overlapping with each other. To simulate the realistic case that an apartment may not have a femtocell, we use a ’deployment ratio’ to determine whether an apartment is deployed with a femtocell or not. For femtocell apartments, the femto BS and femto MSs are dropped randomly and uniformly within the apartment with a minimum separation

femto MS femto BS

apartment

10m 10m

... 10m

femto cell

STREET

... 10m 10m

Fig. 2.

Dual stripe model.

of 20 cm. This scenario enables the investigation of the interference of neighbouring femtocells. 3) Femtocell scenario 3 - Dense urban 5×5-grid model: The 5×5-grid model is an alternative simple femtocell scenario. We consider a single floor building with 25 apartments (see Figure 3). The apartments are of size 10 m × 10 m and are placed next to each other on a 5×5 grid. In addition we assume that with probability p, there is a femtocell in each apartment. This probability represents the density of femto deployment. For the apartments that are a femtocell, the femto BS and femto MSs are dropped randomly and uniformly in the apartment with a minimum separation of 20 cm. femto MS femto BS

10m 10m

femto cell

apartment

Fig. 3.

5x5-grid model.

Since macro MSs are dropped uniformly and randomly throughout the macrocell environment, it is possible that some macro MSs will be dropped into a femtocell area. C. Mobility Model The locations li (t), direction of movement di (t) of macro MSs and their velocities vi (t) are updated every snapshot according to: li (kT + T ) = li (kT ) · vi (T ) · T

(1)

di (kT + T ) = di (kT ) + A · ∆di (kT )

(2)

The direction of movement di (t) of each macro MS at the (k + 1)-th snapshot is obtained by updating its direction at the k-th snapshot by multiplying a uniformly distributed random 1 i variable ∆di (with f∆di (∆di ) = ∆d rect( ∆d ∆d )) with another random variable A generated from a discrete probability density function fA (A) = pdc δ(A − 1) + (1 − pdc )δ(A). Thus

67

the maximum change in direction, also called maximum swing angle is limited to ∆dmax . The probability of direction change pdc is used to make a decision whether or not a particular macro MS changes its direction. Obviously, the random variable A can take values of either zero or one. If the mobile station does not change its direction (A = 0), the direction calculated at the previous snapshot remains unchanged [9]. Femto MSs are assumed to be at a fixed position during the whole simulation time.

considering user mobility, different mixed user channel models were chosen for macrocell and femtocell users respectively. In case of a macrocell, a typical mixed user channel model [15] was used (Ped-B for macro mobile stations at the velocity of 3 km/h). In case of a femtocell, the channel mixed model for an indoor small office scenario [13] is used (PedB for fixed femto mobile stations, 0 km/h). The multipath channels between the mobile stations and the base stations were generated once per snapshot for each subcarrier. The resulting channel frequency response is calculated afterwards and averaged over each set of 12 subcarriers to achieve an averaged gain per resource block.

D. Traffic Model The communication traffic in the network is also updated on a snapshot. Traffic models we consider can be grouped in the service types ’best effort packet transmission’ and ’real time transmission’. The best effort packet service type includes FTP and HTTP, while real time services include streaming, VoIP and gaming. The detailed statistical traffic model and associated parameters are defined in [10].

G. SINR Calculation The link performance can be evaluated using block error ratio (BLER) or throughput. In this work we focus on the throughput performance. A mapping method ’attenuated and truncated Shannon bound’ could be used for link to system mapping (3GPP TR 36.942 V8.1.0 A.1). The principle of this method is to map the obtained SINR at each snapshot to the throughput. First of all, the post-processing SINR is calculated for each user and each subcarrier n. For different downlink transmission modes different post-processing SINR calculation methods are implemented. The post-processing SINR of the desired user i with the power Pi,n on the n-th subcarrier is given by: • Transmission mode 1: Single-input Single-output (SISO)

E. Propagation Model 1) Path Loss Model: The path loss model implemented in this simulator is given by [8]. The path loss between a macro BS and a MS is characterized can be calculated as follows: • P L = 15.3 dB + 37.6 log10 (R/m), for outdoor MS and • P L = 15.3 dB+37.6 log10 (R/m)+Low , for indoor MS, where R is the distance from the transmitter (Tx) to the receiver (Rx) in meters and Low the penetration loss of an outdoor wall, which is 10 dB or 20 dB. For the pathloss between a femto BS and a MS different cases for each femto scenario are considered. It is distinguished between the case whether an MS is inside the same femto house or inside a different femto house as the femto BS, or whether a MS is outside for the suburban scenario. For the dense urban dual stripe model the cases whether a MS is inside the same apartment stripe or not are also considered. An alternative simplified model based on the LTE-A evaluation methodology which avoids modelling any walls is used for the 5×5-grid model. Here the pathloss is calculated as follows: P L = 127 dB + 30 log10 (R/1000 m).

Pi,n (|hi,k(i),n |2 ) SIN Ri,n = PN 2 2 ( l=1, Pl,n |hH i,k(l),n | ) + σn

(3)

l6=i

• •

F. Multipath Model The LTE system exploits instantaneous channel conditions for the performance enhancement such as channel dependent scheduling, adaptive modulation and coding (AMC) and different transmission modes for uplink and downlink [11]. For a realistic system performance evaluation, short-term timevarying channel characteristics are considered during simulation as well as geometric path loss and long-term shadow fading. Time-varying short-term fading effects are described by power delay profiles. Conventional six-tap models [12] were developed for 5 MHz bandwidth channels, but they are not sufficient to describe a system which has 10 MHz bandwidth. As the bandwidth increases, the time resolution in the delay domain increases so that more taps are required [13]. Modified power delay profiles with 24 taps proposed in [14] were used. By

•

•

•

68

hi,k(i),n : channel frequency response on the n-th subcarrier between the i-th user and k-th BS k: BS index N : total number of user σn2 : variance of the white Gaussian noise on the n-th subcarrier Transmission mode 2: Transmit diversity Transmission mode 3: Open-loop spatial multiplexing (SM) LTE uses open-loop SM, where independent data streams are transmitted from different antennas. A minimum mean square error (MMSE) receiver is assumed. Transmission mode 4: Closed-loop spatial multiplexing In closed-loop spatial multiplexing, precoding is applied at the base station before transmission. Precoding is carrier out corresponding to TS 36.211, 6.3.4.2.3. Transmission mode 5: Multiuser multiple-input multipleoutput (MU-MIMO) This case corresponds to transmission mode 4, i.e. the single layer transmission of the closed-loop spatial multiplexing. Transmission mode 6: Closed-loop rank 1 precoding This mode amounts to beamforming and it can be regarded as a special case of transmission mode 4 (single layer transmission)

•

slowly varying channels. Consequently, retransmissions are likely to occur soon after the initial transmission and the channel state remains constant during retransmissions. Once a BS selects an MCS level of a given resource block, the reselection procedure of MCS level for that resource block will not occur unit all HARQ retransmissions are completed for that resource block.

Transmission mode 7: If the number of the physical broadcast channel antenna ports is one, Single-antenna port, port 0; other wise transmit diversity See transmission mode 1 and 2.

H. Link Level Quality Estimation Link adaptation techniques significantly increase user throughput by providing efficient ways to maximize spectral efficiency [16] [17] [18]. In case of multi-carrier transmission as in LTE, the set of subcarrier SINRs are mapped with the help of an exponential effecitve SINR mapping (EESM) into a scalar instantaneous effective SINR value. Next from linklevel performance curves an estimate of the BLER value is then obtained, using the effective SINR value. EESM is used when all the resource blocks allocated to a MS are modulated using the same modulation and coding scheme (MCS). The basic idea of EESM is to find a function that maps the set of SINRs to a single value that is a good predictor of the actual BLER. In Figure 4 the link level to system level interface used in our simulator is depicted. Several input parameters e.g., channel frequency response, powers, MIMO modes and CQI values, are given from the system level to the link level. In the link level abstraction function the post-processing SINR on each subcarrier is calculated in order to determine the effective SINR. The output of the link level abstraction is BLER and throughput.

Link Level

The scheduler of a femtocell has an essential functionality at the BS. It manages and allocates network resources considering boundary conditions and optimization criteria, but with a strongly modified design in comparison to macro BSs. Our simulator will be able to compare scheduling algorithms which do and do not use cooperation between femtocells. As a baseline we implemented two known scheduling algorithms namely, ’Round-Robin’ and ’Proportional Fair’. In the further developed versions of our simulator these scheduling algorithms will be compared with more complex newly designed scheduling algorithms. III. S IMULATION PARAMETERS AND R ESULTS A cellular configuration consisting of a main macro site (with 3 hexagonal cells/sectors) and six surrounding sites is created in order to simulate the effects of intra- and intercell interference in the site that is located at the center as well as at the boundary of the main macro site. The simulated femto scenario is the ’suburban model’. One femto block and 2 macro MSs are distributed uniformly within each macrocell. The number of femto MSs varies between four and eight and is randomly chosen for each femtocell. All stations have two transmit and two receive antennas. This scenario is randomly created 800 times and each scenario is simulated for 5 milliseconds (5 snapshots). The same simulation is done for a scenario with two femto blocks per macrocell, too. Detailed simulation parameters are listed in Table 1. A Round-Robin scheduling algorithm are used. Since Round-Robin scheduling does not require any power control, we do not activate one MS after the other. We assume that all MSs are active at the begin of the simulation. For statistical analysis we create 800 random scenarios all with the same simulation parameters. After resource allocation, SINR calculation for transmission mode 4 is done only for the allocated resources for each MS in the main macro site. In order to model the HARQ retransmission improvements of BLER, link level simulations have been used [22]. The LTE link level simulations including H-ARQ processes were performed for a single-user scenario corresponding to the simulation parameters shown in Table 2. All of the 15 MCS defined by the channel quality indicator (CQI) values in the LTE standard [23] have been used. 1000 subframe-long simulations were sufficient to obtain reliable data at BLERs of 10−1 . For physical layer abstraction we use look-up tables (BLER vs. SINR and Throughput vs. SINR curves) obtained from these link level simulations. The MCS selection for each

System Level

+ Compute post processing SINR on each subcarrier

+ Channel frequency response H(f) + Transmit power + Noise power + Interference power + MIMO modes + MCS

Look up beta

+ Compute effective SINR using EESM

+ PHY abstraction mapping

Fig. 4.

I. Scheduling

BLER, Throughput

Link level to system level interface.

In cellular systems, it is state-of-the-art to use hybrid automatic repeat request (HARQ) protocols accompanied by retransmissions to achieve a high system capacity. The use of HARQ is also an undisputed assumption to LTE. HARQ techniques are also used to improve the throughput performance of the link adaptation techniques by compensating link adaptation errors caused by inaccurate channel estimation and the channel quality feedback delay [19] [20]. With our simulator a retracement of each resource block over the whole simulation time and the variation of their delay times are available. An optimal and totally adaptive modulation and coding scheme selection criterion for maximizing system capacity and user throughput in cellular networks will be used. As introduced in [21] a synchronous HARQ scheme will be employed for

69

TABLE I S YSTEM LEVEL SIMULATION PARAMETERS

1

macrocells femtocells

0.9 0.8

0.7 Value 0.6 Hexagonal grid, 0.5 3 cells per site, frequency reuse 1 0.4 5 1 ms 0.3 7 0.2 2 0.1 1 or 2 0 0 2 4 6 8 10 12 14 16 18 500 m Mbps 2000 MHz 8 dB Fig. 5. Throughput of a scenario with 2 macro MSs per macrocell, 1 femtocell Between cells: 0.5 fixed per macrocell, 2 × 2 MIMO at 1.4MHz Between sites: 1.0 fixed For 3-sector cell sites A(Θ) = −min[12( ΘΘ )2 dB, Am ] 3dB 1 where Θ3dB = 70◦ , Am = 20dB 0.9 macrocells 9 dB femtocells 0.8 3 km/h

cdf

Parameter Cellular Layout

Antenna pattern (fixed)

macro/femto MS noise figure macro MS velocity Number of Tx, Rx antennas for macro and femto Total macro BS TX power Total femto BS TX power Minimum distance between MS and macro BS Minimum distance between MS and femto BS

0.7

2x2 46 dBm 20 dBm

0.6

cdf

Number of snapshots Snapshot duration Number of macro BS Number of macro MS per cell Number of femto BS per macrocell Inter-site distance Carrier Frequency Shadowing standard deviation Shadowing correlation

0.5 0.4 0.3

35 m

0.2 0.1

20 cm

0

0

2

4

6

8

10

12

14

16

18

Mbps

TABLE II L INK LEVEL SIMULATION PARAMETERS Parameter System bandwidth Subcarrier spacing Channel profile Max. number of retransmissions CQI values MIMO transmission mode Simulation length

Fig. 6. Throughput of a scenario with 2 macro MSs per macrocell, 2 femtocells per macrocell, 2 × 2 MIMO at 1.4MHz

Value 1.4 MHz 15 kHz AWGN 3 1-15 4 (MIMO), 1 (SISO) 1000 subframes

The ideas presented in many publications rely on a centralized network architecture, where a centralized entity should collect the data, generate the plan and distribute the information. However, a distributed/decentralized architecture where each femtocell is able to select the sub-channels itself would be more suitable. Further improvements currently under research in our group, include the study of distributed algorithms a truly dynamic frequency allocation for OFDMA networks. Regarding this purpose we will develop new scheduling algorithms for femtocells and compare them with our simulator. An optimal MCS and adaptive transmission mode selection rule for maximizing user throughput taking HARQ operation into account will also be developed for both uplink and downlink transmissions. In this paper, we present a preliminary version of a dynamic system level simulator for LTE networks. The interference behaviour of femtocells placed within macrocells is shown in a multi-cell and multi-user multi-carrier MIMO system.

allocated resource block of a MS is done for a target BLER of 10−1 . The achievable rate per MS is determined after the lowest MCS of all allocated resource blocks for that user is selected as the users MCS level. The cumulative distribution function (cdf) of the system throughput for the macrocell and femtocell is shown in Figure 5 and 6, respectively. Figure 5 shows the case that only one femtocells is within each macrocell and in Figure 6 the case that two femtocells are distributed within each macrocell is depicted. The total simulation time for one snapshot (1ms in real time) with MATLAB (Version 7.8.0 2009a) in a Intel 2.8 GHz, 3.25 GB RAM computer is for the scenario with one femtocell per macrocell 7 seconds, for the case of two femtocells per macrocell the simulation time is 17 seconds.

V. ACKNOWLEDGEMENT The described system level simulator is developed within the project Optifemto in cooperation with our partners mimoOn, Duisburg and Heinrich-Hertz-Institut, Berlin. The project is funded by the Federal Ministry of Economics and Technology.

IV. F UTHER W ORK AND C ONCLUSION In future works we will also focus on the uplink transmission and investigate the effect of uplink transmission upon the overall system capacity.

70

R EFERENCES [1] Jie Zhang, Guillaume de la Roche , Femtocells : Technologies and Deployment, Wiley, 2009. [2] V. Chandrasekhar, J. Andrews, and A. Gatherer, “Femtocell networks: A survey,” IEEE Communications Magazine, vol. 46, pp. 59–67, Sept. 2008. [3] E. Haro, S. Ruiz, D. Gonzalez, M. Garcia-Lozano, and J. Olmos, “Comparison of different distributed scheduling strategies for static/dynamic lte scenarios,” in COST 2100 TDs 9th MCM - Wien. [4] D. Lopez-Perez, A. Valcarce, G. De La Roche, E. Liu and J. Zhang, “Access methods to WiMAX femtocells: A downlink system-level case study,” in Communication Systems, 2008. ICCS 2008. 11th IEEE Singapore International Conference on, Nov. 2008. [5] Y. Sung, N. Jeon, B. Yoon, J. Lee, S. Lee, and S. Kim, “Femtocell/macrocell interference analysis for mobile WiMAX system,” in The 5th IEEE VTS Asia Pacific Wireless Communications Symposium (APWCS2008), Aug. 2008. [6] D. Lopez-Perez, A. Valcarce, G. De La Roche, A. Juttner and J. Zhang, “Interference avoidance and dynamic frequency planning for WiMAX femtocells networks,” in Communication Systems, 2008. ICCS 2008. 11th IEEE Singapore International Conference on, Nov. 2008, pp. 1579– 1584. [7] 3GPP TR 25.814, “Physical layer aspects for evolved universal terrestrial radio access (UTRA),” 3GPP, Sept. 2006. [8] Femto Forum, “OFDMA interference study: Evaluation methodology document,” Femto Forum, Apr. 2009. [9] S. A. Ghorashi, E. Homayounvala, F. Said, and A.H. Aghvami, “Dynamic simulator for studying WCDMA based hierarchical cell structures,” in Personal, Indoor and Mobile Radio Communications, 2001 12th IEEE International Symposium on, Sept. 2001. [10] “Ngmn radio access performance evaluation methodology,” 2008. [11] S. Sesia, I. Toufik, and M. Baker, LTE, The UMTS Long Term Evolution: From Theory to Practice, Wiley, 2009. [12] REC. ITU-R M.1225, “Guidelines for evaluation of radio transmission technologies for IMT-2000,” ITU-R M.1225. [13] IEEE, “Draft IEEE 802.16m evaluation methodology,” IEEE, Dec. 1997. [14] IEEE 802.16 Broadband Wireless Access Working Group, “Wideband extension of the ITU profiles with desired spaced-frequency correlation,” 2007. [15] “3GPP TS 36.101: Evolved Universal Terrestrial Radio Access (EUTRA); User Equipment (UE) radio transmission and reception,” . [16] A. J. Goldsmith, and S.-G Chua, “Adaptive coded modulation for fading channels,” IEEE Transactions on Communications, vol. 46, pp. 595–602, May 1998. [17] S. R. Kadaba K. Balachandran and S. Nanda, “Channel quality estimation and rate adaptation for cellular mobile radios,” IEEE Journal on Selected Areas in Communications, vol. 17, pp. 1244–1256, 1999. [18] K. Balachandran S. Nanda and S. Kumar, “Adaptation techniques in wireless packet data services,” IEEE Communications Magazine, vol. 38, pp. 54–64, Jan. 2000. [19] G. Caire and D. Tuninetti, “The throughput of hybrid-ARQ protocols for the gaussian collision channel,” IEEE Transactions on Information Theory, vol. 47, pp. 1971–1988, 2001. [20] J. Cheng, “Coding performance of hybrid ARQ schemes,” IEEE Transactions on Communications, vol. 54, pp. 1017–1029, 2006. [21] H. Lee D. K. Sung D. Kim, B. C. Jung and H. Yoon, “Optimal modulation and coding scheme selection in cellular networks with hybrid-ARQ error control,” IEEE Transactions on Wireless Communications, vol. 7, pp. 5195–5201, Dec. 2008. [22] C. Mehlfuhrer, M. Wrulich, J. Colom Ikuno, D. Bosanska, and M. Rupp, “Simulating the Long Term Evolution Physical Layer,” in The 17th European Signal Processing Conference (EUSIPCO 2009), Aug. 2009. [23] “3GPP TS 36.213: Evolved Universal Terrestrial Radio Access (EUTRA); Physical layer procedures,” .

71