The 18th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC’07)
MULTI-ANTENNA SDMA IN OFDM RADIO NETWORK SYSTEMS: MODELING AND EVALUATIONS Afif Osseiran, Per Skillermark and Magnus Olsson Ericsson Research, Stockholm, Sweden {Afif.Osseiran, Per.Skillermark, Magnus.A.Olsson}@ericsson.com A BSTRACT This paper presents and evaluates different multi-antenna deployments in an OFDM wireless system. Two downlink implementations, higher order sectorization (HOS) and downlink SDMA using fixed beams (FB), are considered. HOS with 3-, 6-, and 12-sector sites are compared with downlink FB SDMA in a 3-sector site where each sector is using up to four antennas. In the uplink we evaluate SDMA using multi-antenna MMSE receiver at base station. The simulation results indicate that in the downlink FB SDMA with four antennas per sector provides a cell throughput improvement of more than 100% compared to a 3-sector site equipped with a single antenna each. The gain can be further improved by using HOS at the expense of additional complexity. The uplink results show that SDMA allows to increase the throughput by up to a factor of two. Key Words: Fixed beams, Higher Order Sectorization, MMSE receiver, OFDM, SDMA, SIC, System Performance.
I.
I NTRODUCTION
Conventional cellular systems typically make use of base station (BS) antennas with omni-directional or three-sectorized radiation patterns. A drawback of such BS antenna systems is the limited directivity. In the downlink the electro-magnetic energy, destined for a user located in a certain direction, is unnecessarily radiated in the entire sector, thus causing interference to other users in the network. Similarly, in the uplink the BS receives interfering signals from all directions even though it has only interest to collect the desired signal from a specific direction. One way to enhance the performance of such networks is to improve the directivity of the BS transmission and reception, which may be achieved by increasing the number of antennas. Higher order sectorization (HOS) is a straightforward solution to enhance the directivity by increasing the number of sectors per site, and reducing the width of each sector. An alternative solution is to keep the width of the sectors but increase the number of antennas per sector and to apply beam forming (BF). Beam forming comes in many different flavors that differ in performance and complexity. Transmit BF techniques include fixed BF and adaptive BF. With fixed BF the beams and the associated transmit weights are predefined and at each transmission attempt the best suited beam is used. The fixed BF method is also referred to as fixed beams (FB). With adaptive transmit BF the transmit weights are adaptively chosen with respect to a given criteria. An example is eigen BF for which the transmit weights are chosen to the principal eigenvector of c 1-4244-1144-0/07/$25.002007 IEEE
the channel covariance matrix [1]. Depending on the adaptation rate, BF schemes may further be classified as short-term or long-term beam forming techniques. Essentially, short-term BF implies that the adaptation is based on the instantaneous channel state while for long-term BF the adaptation is based on long-term statistics. Similar techniques may be used to enhance the directivity at the receiver side where it is often possible to estimate the channel state and the characteristics of the interference, which simplifies the selection of the receiver weights. The receiver antenna weights may e.g. be adapted to maximize the (postreceiver) signal-to-interference-and-noise ratio (SINR). Examples of receiver combining schemes include maximum ratio combining (MRC) and minimum mean square error (MMSE) combining also called optimum combining [2]. In cellular networks using BF, one way of taking advantage of increased directivity is to simultaneously serve multiple users on the same physical resource. This is referred to as spatial division multiple access (SDMA). In the downlink, SDMA may e.g. be implemented using the FB method. In this case users in the same sector but using different beams may be scheduled for simultaneous transmission on the same resource. In the uplink, multiple spatially separated users may similarly be scheduled for transmission on the same resource. The different signals are then restored by means of signal processing at the multi antenna BS receiver. In this paper we compare HOS with FB SDMA in the downlink of an OFDM radio network. FB method is used since it is simple but yet offers a performance almost as good as that of more advanced long-term BF schemes. In the uplink we analyze the benefits of SDMA with and without successive interference cancelation (SIC) at the BS receiver. II.
A NTENNA C ONFIGURATION
Assuming a plane wave incident in the horizontal plan on a Uniform Linear Array (ULA) from an angle θ relative to the axis of the array, it can be shown that the array factor (or antenna gain) of a ULA consisting of N Antenna Element (AE) is given by [3]: g(θ) = ωi H a(θ)
(1)
a(θ) = [1 e−jβx1 cosθ . . . e−jβxN −1 cosθ ]H
(2)
where
is the spatial signature, and ωi = [ωi,0 . . . ωi,N −1 ]H )
(3)
The 18th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC’07)
is the weight vector of the ith beam. β = 2π/λ is the phase propagation factor, and λ is the wavelength. In case of a FB system, the number of the weight vectors is fixed and usually equal to the number of the antenna elements in the ULA. The FB weight vectors are generated by a Beam Forming Network (BFN) that produces N orthogonal beams using N antenna elements. Let us define the N × N BFN Matrix (BFNM) as T = [ω0 ω1 ... ωN −2 ωN −1 ]. T comprises of the weight vectors of all the narrow fixed beams. The most commonly used techniques to implement a BFN is to use cascaded hybrid couplers known as the Buttler matrix, which is simply a realization of the Fast Fourier Transform (FFT), so the array weight are given by [4]: ωi,m = e−j2πmi , i ∈ {0, . . . , N − 1}
Figure 2: Antenna configuration of 12-sector site.
(4)
The default configuration is the 3-sector site where each sector is equipped with a single AE. For the SDMA case it is assumed that the 3-sector site is equipped with an ULA with 2 or 4 AEs. Fig. 1 illustrates as an example a 3-sector site each equipped with an ULA of several AEs. The central AE of each ULA form a uniform circular array with radius r. The orientation of the ULAs is on the tangent of the circle. The FFT method is used to generate the beam pattern. It is further assumed that the number of beams per sector is equal to the number of AEs of the ULA. The HOS configuration consisted of placing 3, 6 or 12 antennas equally spaced on a circle of radius r. An example 12sector sites is illustrated in Fig. 2. In this paper, we assumed that the 3-, 6- and 12- sector sites are equipped with an ULA of 1, 2 and 4 AEs, respectively. The sector antenna pattern is obtained by simply applying a unity weight vector to all AEs. Fig. 3 shows the antenna patterns of 3-, 6- and 12-sector sites.
20
Antenna Gain [dB]
10
wk,1,0 wk,1,−1
wk,1,1 wk,1,N2
wk,1,N2
xk
wk,0,1
xk
wk,0,0 wk,0,−1
r wk,2,N1
d
wk,0,N1
xk
wk,2,−1 wk,2,0 wk,2,1
5
0
−5
−10
−15 −200
−150
−100
−50
0 θ (deg)
50
100
150
200
Figure 3: Gains for 3, 6, and 12 sector antennas.
A. wk,1,N1
3Sec 6Sec 12Sec
15
Received Signal:
Let NR and NT denote the number of receive and the number of transmit antennas, respectively. The signal on the f th subcarrier at block (OFDM symbol) index k of the bth BS, is denoted by xb (f, k), a vector of size NX ×1, where 1 ≤ NX ≤ NT . NX is the number of transmitted streams. Here, without loss of generality, we assume the signal of interest is transmitted on the 0th BS. Let W b (f, k) and H b (f, k) denote that complex antenna weights (of size NT × NX ) and the channel coefficients (of size NR × NT ) on the f th subcarrier and time index k of the bth BS at UT 0, respectively. The received signal on the f th subcarrier at k, is given by y(f, k) = H 0 (f, k)W 0 (f, k)x0 (f, k) + ξ(f, k)
(5)
wk,1,N2
Figure 1: Antenna configuration of a 3-sector site.
III.
S IGNAL TO I NTERFERENCE AND NOISE D ERIVATION
In the following the expression of the received signal and the SINR will be derived.
where ξ(f, k) contains the received inter-cell interference and thermal noise, and is equal to B−1 H (f, k)W (f, k)x (f, k) + n(f, k). B denotes the b b b b=1 total number of BSs, and the term H b (f, k)W b (f, k)xb (f, k) is the interference signal from BS b. Finally, n(f, k), denotes the zero mean Additive White Gaussian Noise (AWGN). Since equalization and symbol detection operate on a subcarrier basis and block-by-block basis, we will henceforth omit the subcarrier index f and the time index k, respectively.
The 18th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC’07)
B. SINR Calculation In following the SINR for MMSE receiver is presented. The 0 , and we assume estimate of the signal x0 is denoted by x that the instantaneous channel matrix H 0 , and the second order statistics of the noise and the interference are known at the receiver. The MMSE estimate per subcarrier is simply obtained using the Wiener Filter [5] as follows: 0 = Λy 0 x
(6)
where the filter weight Λ is defined as −1 Λ = E x0 y H E yy H
(7)
Assuming the transmitted signals xb have unit power then using Eq. 5, Λ reduces to H −1 Λ = WH 0 H 0 S0
(8)
Where S0 is given by H S0 = H 0W 0W H 0 H 0 + R0 ,
(9)
sector, i.e., a deployment using twelve AEs per site. BS antennas are placed above rooftop. The network is assumed to operate at a carrier frequency of 3.5 GHz and OFDM with 128 sub-carriers is used within the 5 MHz transmission bandwidth. Table 1 provides a summary of the assumed system parameters. Parameter Number of sites Cell Radius [m] Number of sectors per site Number of BS antennas per sector Site output power BS receiver noise figure UT output power UT receiver noise figure Carrier frequency Transmission bandwidth Sub-carrier bandwidth Number of sub-carriers Cyclic prefix length
R0 is the interference plus noise covariance matrix and is given by R0 =
B−1
H H bW bW H b Hb + N 0
(10)
b=1
Furthermore, N 0 , the covariance matrices of the noise is given by N 0 = E nnH = σn2 I NR , (11) where σn is the thermal noise variance. Let el denote the lth column of the NX × NX identity matrix I NX , then using the above Equations the SINR of the l element of x0 is given by (l)
Γ0 = IV.
2 |eH l ΛH 0 W 0 el | H eH l ΛR0 Λ el
(12)
M ODELS AND A SSUMPTIONS
This section describes the models and assumptions used in the evaluations and the employed performance measures. A.
User behavior models
A metropolitan area with outdoor users on street level is considered. Users are assumed to be uniformly but randomly distributed in the area and move with an average speed of 3 km/h. Full-buffer traffic models are used in both uplink and downlink. B.
Network deployment model
A network deployment with seven sites where each site comprises three, six, or twelve sectors is considered. In downlink the number of BS antennas varies between three and twelve depending on the number of sectors per site (HOS) or the number of FBs (SDMA) that are employed. The uplink evaluations only considers a 3-sectorized deployment with four AEs per
Value 7 500 m, 1000 m, or 2000 m 3, 6, or 12 1, 2, or 4 120 W 5 dB 0.25 W 7 dB 3.5 GHz 5 MHz 39.0 kHz 128 3.2 µs
Table 1: System and Simulations Parameters.
C.
Radio Channel Model
The C2 metropolitan area pathloss and channel model from [6] is used in the evaluations. The model is applicable to a scenario with macro BS installation above rooftops and UTs located outdoors on street level. NLOS propagation is assumed between the BS antennas and the UTs. Shadow fading is modeled as a log-normally distributed random variable with a standard deviation of 8 dB. The ray-based channel model is an extension to the 3GPP spatial channel model (SCM) [7] with correlated shadow fading, delay spread and angular spread. D.
Transmitter and Receiver Structure
Both the BSs and the UTs are equipped with multiple antenna elements separated half a wavelength. For downlink SDMA the BS antennas are used for transmit beam forming. Furthermore, in the downlink a dual antenna MMSE receiver is employed at the UTs. In the uplink UTs transmits using a single antenna while the BS is equipped with multiple antenna MMSE receiver that allows to suppress multiple access interference. Moreover, for uplink SDMA SIC at the BS receiver is considered as a performance enhancing technique. With SIC the signals associated with already decoded packets are regenerated and subtracted from the received signal. In this study an ideal SIC model is used, i.e., it is assumed that the signals of already decoded packets can be perfectly removed. E.
Radio Network Algorithms
UTs connect to the sector with the lowest path-loss, shadowing included, and the downlink BF gain is considered in the cell selection procedure. In both uplink and downlink signals are
The 18th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC’07)
transmitted using a fixed output power and the modulation order and channel code rate is selected to maximize the data rate. Turbo coding with rates from 1/10 to 8/9 are used in combination with QPSK, 16QAM, or 64QAM to find an appropriate transmission format. Round-robin transmission scheduling is employed in both uplink and downlink. For downlink HOS one user per sector is scheduled for transmission while one user per beam is scheduled for downlink FB SDMA. In uplink, a predefined number of users per sector is scheduled for simultaneous transmission. For simplicity the user selection algorithm does not take into account the spatial separation between users. F.
Link-to-System Interface
To estimate the packet decoding error probability of a channel coded block transmitted over a multi-state channel, a mutual information (MI) based link-to-system interface is used [8]. The model uses the post-receiver SINRs of the symbols in the channel coded block to calculate the average MI for bit-interleaved coded modulation. The average MI is then used to estimate the packet error probability. All imperfection like e.g. channel estimation errors are neglected. G.
Performance Measures
to a 3-sector site. The gain of the 2FB is roughly 1.4 times compared to the 3-sector site case. In the same figure, SDMA is compared to HOS in case of 6- and 12-sector sites. The 12sector site offers up to 3.5 times relative gain to the 3-sector site while the 6-sector site offers 1.8 times relative gain. It is interesting to note that the relative gain of SDMA and HOS increases for a large cell size. In fact for a large cell size the capacity is limited by the noise and not the interference as it is the case for small cell sizes. Hence increasing the antenna gain will increase the coverage and allow larger number of users with a better SINR to be served. Doubling the number of beams from one to two yields slightly more than 40% relative cell throughput gain for SDMA while going from a 3- to 6- sector site provided 80% relative gain . On the other hand, doubling the number beams from two to four yields substantially higher relative cell throughput gain in the order of the 70%. The same observation was noted for WCDMA in [9]. In fact the low gain of a 2FB system is due to the method used to generate the beam patterns. The 6-sector antenna pattern has higher maximum antenna gain and lower side lobes compared a 3-sector site each covered by 2-FB. Finally the cross over beam region is wider for the 2FB case. Consequently, the intra-beam interference increases and yields a lower cell throughput.
The post-receiver SINR and the average throughput per site are used as performance measures, where the average throughput per site is defined as the number of correctly received bits relative the number of sites and the simulation time.
A.
R ESULTS
Downlink
In this section we compare the performance of HOS to FB SDMA. Fig. 4 shows the cdf of the SINR of all simulated schemes for a cell radius of 1km. The SINR decreases when the number of sector per site increases at the capacity limit. Similarly the SINR decreases when the number of FB per sector increases. 100
90
3Sec 3Sec−2FB 3Sec−4FB 6Sec 12Sec
3
Normalized Site throughput
V.
3.5
2.5
2
1.5
1 500
1000
1500
2000
Cell Raduis [m]
Figure 5: The normalized site throughput versus the cell radius for SDMA based on 3-sector sites and HOS for 6- and 12-sector sites.
80
70
cdf
60
50
40
30 3−Sec ,R=1000m 6−Sec ,R=1000m 12−Sec ,R=1000m 3Sec−2FB ,R=1000m 3Sec−4FB ,R=1000m
20
10
0 −10
−5
0
5 SINR [dB]
10
15
20
Figure 4: The cdf of the SINR for cell radius of 1km.
The relative system throughput gain compared to 3-sector sites equipped with single antenna, for HOS, FB are summarized in Table 2. The second column of the table indicates the number of FB used per sector. It is interesting to note that a large number of antenna configuration parameters can be optimized in order to increase further the downlink system capacity. For instance the antenna pattern e.g. side-lobes level, beam-width, antenna tilting and antenna spacing have been shown to impact the signal quality [10, 11]. Moreover the number of the beams can be also optimized for a FB concept [12] in order to further increase the capacity gain for a FB system. B.
The normalized site throughput versus the cell radius for the 3-sector sites where each is equipped with 2 or 4 AEs is shown in Fig. 5. The 4FB case offered a gain up to 2.4 times compared
Uplink
In the uplink SDMA can be realized by scheduling multiple users for transmission on the same physical resource. In order to separate the signals from various users at the base station a
The 18th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC’07) Technique
Sectors
FB/sector
Relative Gain
HOS FB SDMA HOS FB SDMA HOS
3 3 6 3 12
1 2 1 4 1
1 1.4 1.8 2.4 3.5
Table 2: Relative system gain to a single antenna. multi-antenna receiver is used. In an OFDM system, however, in which transmissions within a sector are orthogonal, SDMA introduces intra-cell multiple access interference in addition of increasing the inter-cell interference. In the studied deployment, MMSE receiver is employed at the BS to suppress interference and SIC after channel decoding is used as a performance enhancing technique to further limit the impact of the intra-cell multiple access interference introduced by SDMA. In each sector one, two or four randomly selected users are scheduled for transmission on the same physical resource and every scheduled user is assigned the entire transmission bandwidth. Fig. 6 depicts the average sector throughput as a function of the number simultaneously scheduled users. The plot includes results with and without SIC at the BS receiver. Without SIC, the average sector throughput is 33.0 Mbps, 46.5 Mbps and 5.5 Mbps, respectively, with one, two and four SDMA users. With SIC performance is further increased. The results indicate that with SIC and four users scheduled for simultaneous transmission the throughput is more than doubled in comparison to the case when a single user per sector is scheduled for transmission. 100 90
average throughput [Mbps/site]
80 70 60 50 40 30 20 10 0 0
MMSE−SIC receiver MMSE receiver 1
2 3 number of scheduled SDMA users
4
5
Figure 6: Uplink average throughput per site for a MMSE BS receiver and a MMSE-SIC BS receiver.
VI.
C ONCLUSIONS
The introduction of multiple antennas in OFDM-based radio network systems leads to a substantial system capacity increase. In this paper several multi-antenna configurations have been analyzed and assessed: Higher order sectorization, downlink Fixed Beams (FB) SDMA and uplink MMSE receiver based SDMA. The evaluations have been performed by means
of multi-cell computer simulations with accurate interference modeling. The results indicate that the downlink HOS configuration provides a higher throughput than the FB SDMA configuration at the expense of additional complexity. The HOS throughput increases almost in proportion to the number of sectors per site, e.g., a 3.5 times throughput gain when increasing the number of sectors per site from three to twelve. On the other hand, FB SDMA yielded 2.4 times gain compared to 3-sector sites where each is equipped with a single antenna. Furthermore, the investigations showed that the uplink MMSE SDMA enhances the system throughput. In fact, in a 3-sector deployment where each sector is equipped by four antenna elements, SDMA increased the throughput by more than a factor of two when SIC after channel decoding in conjunction of MMSE receiver is employed at the base station. ACKNOWLEDGEMENT Part of this work has been performed in the framework of the IST project IST-4-027756 WINNER II, which is partly funded by the European Union. The authors would like to acknowledge the contributions of their colleagues in WINNER II, although the views expressed are those of the authors and do not necessarily represent the project. R EFERENCES [1] F. R. Farrokhi et al., “Link-Optimal BLAST Processing With MultipleAccess Interference,” in Proceedings IEEE Vehicular Technology Conference, vol. 1, Sept. 2000, pp. 87-91. [2] J. H. Winters, ”Optimum Combining in Digital Mobile Radio with Cochannel Interference” in IEEE Journal on Selected Areas in Communications, Vol. 2, No 4, July 1984. [3] J. C. Liberti and T. S. Rappaport, Smart Antennas for Wireless Communications: IS-95 and Third Generation CDMA Applications. Prentice Hall, 1999. [4] S. J. Orfanidis, Electromagnetic Waves and Antennas Applications. Rutgers University, USA, 2004, www.ece.rutgers.edu/ orfanidi/ewa. [5] S. Haykin, Adaptive Filter Theory, 4th ed. Prentice Hall, 2002. [6] J. Meinil¨a, Ed., IST-2003-507581 WINNER I, D5.4, Final report on link level and system level channel models, 2005, no. v1, https://www.istwinner.org/Documents/Deliverables/D5-4-V1.pdf. [7] 3GPP, “Spatial channel model for multiple input multiple output (mimo) simulations, Tech. Rep. 3GPP TR 25.996 V6.1.0, Sept. 2003, http://www.3gpp.org/ftp/Specs/html-info/25996.htm. [8] K. Brueninghaus et al., “Link Performance Models for System Level Simulations of Broadband Radio Access Systems,” in IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), Berlin, Germany, September 2005. [9] A. Osseiran and A. Logothetis, “Smart Antennas in a WCDMA Radio Network System: Modeling and Evaluations,” IEEE Transactions on Antennas and Propagation, vol. 54, no. 11, pp. 3302–3316, Nov. 2006. [10] J. Niemela and J. Lempiainen, “Impact of the Base Station Antenna Beamwidth on Capacity in WCDMA Cellular Networks,” in Proceedings IEEE Vehicular Technology Conference, vol. 1, 2003, pp. 80–84. [11] A. Osseiran and A. Logothetis, “A Method for Designing Fixed MultiBeam Antenna Arrays in WCDMA Systems,” IEEE Antennas and Wireless Propagation Letters, vol. 5, pp. 41– 44, 2006. [12] J. R. Moreno, K. I. Pedersen, and P. E. Mogensen, “Capacity Gain of Beamforming Techniques in a WCDMA System Under Channelization Code Constraints,” IEEE Transactions on Wireless Communications, vol. 3, no. 4, pp. 1199–1208, July 2004.
