COMPARISON OF DIFFERENT MIMO STRATEGIES FOR THE DOWNLINK TDD MODE OF UTRA Roger Gaspa, Javier R. Fonollosa Department of Signal Theory and Communications Universitat Polit`ecnica de Catalunya Jordi Girona,1-3, M`odul D-5, Campus Nord UPC 08034 Barcelona, SPAIN e-mail: frgaspa,
[email protected] ABSTRACT Employment of multi-element antenna arrays at both tranmit and receive sites is capable of enormous theoretical capacity over wireless communication systems. In this paper different decoding strategies for space-time coded systems and layered architectures are analysed from a performance and complexity point of view. 1. INTRODUCTION New wireless communication services require higher data rates and lower Bit Error Rates to provide acceptable Quality of Service. Turbo Codes represented a true milestone in channel coding over additive white Gaussian channels, allowing very low BER using a sub-optimum iterative decoding scheme. Turbo codes, as well as Turbo Trellis Coded Modulations, showed significant coding gains over Rayleigh fading channels provided that temporal diversity is available (i.e. fast fading channel, full interleaving). Under highly correlated fading channels their performance decrease significantly. It is well known that employing multiple transmit and receive antennas considerably increases the capacity of the wireless channels. Transmit diversity schemes represent a powerful technique to combat and mitigate the destructive effects of multipath fading, but their decoding complexity increases with the number of transmit and receive antennas. Several papers appeared in the literature that derive code construction criteria to maximize both transmit diversity and coding gain. The simplest code was first proposed by Alamouti [1] for two transmit antennas and lately extended to higher number of antennas in [2]. Space-time codes based on a trellis structure were first proposed in[3], * This work was partially supported by the European Commission under Project IST-1999-11729 METRA; the Spanish Government (CICYT) TIC98-0703, TIC99-0849, TIC2000-1025, FIT-070000-2000-649; and the Catalan Government (CIRIT) 2000SGR 00083.
and space-time trellis codes with higher coding gain were later proposed in [4][5]. Turbo space-time codes were introduced in [6] as a natural extension of turbo codes to multiple transmit antennas, and design rules to guarantee full antenna diversity were developed in [7]. In [8] analytical tools for designing space-time codes with PSK modulation were presented and in [9] the authors derived the rank criteria to construct powerful space-time codes for QAM modulations based on classical coding theory. A layered architecture, named BLAST (Bell Labs Layered Space-Time Architecture)[10], provides high spectral efficiencies at reasonable decoding complexity based in an interference nulling, interference canceling and compensation procedure. This layered architecture basically demultiplexes the bitstream (coded or uncoded data from previous blocks) into different substreams (one per each transmit antenna), modulates them and finally symbols are fed to its respective transmit antenna. On one hand, inter-substream coding, such that some redundancy between substreams is incorporated, leads to D-BLAST. On the other hand VBLAST does not encode the signal (or, at least, only individual substream coding is applied). The results presented in this paper are part of the METRA project. The METRA project is carried out under the IST (Information Society Technologies) Programme, which is a single integrated research program managed by the Information Society Directorate-General of the European Commission. The METRA consortium is formed by Universitat Polit`ecnica de Catalunya, which acts as project coordinator, the Center for PersonKommunikation of Aalborg University, Nokia Networks, Nokia Mobile Phones and Vodafone Ltd. Updated information about the METRA project including public deliverables can be found at http://www.istmetra.org/. The METRA project builds on the experience of previous European Commission funded projects on smart antennas for mobile communications: TSUNAMI and SUNBEAM. TSUNAMI (Technology in Smart Anten-
nas for Universal Advanced Mobile Infrastructure project) was a RACE (Research in Advanced Communications in Europe)-II project, running for two years from January 1994 to December 1995. Together with its extension, the ACTS (Advanced Communications Technologies and Services) project TSUNAMI-II, running almost for 3 years, clearly demonstrated that adaptive base station antennas are a feasible technology for capacity enhancement and range extension in 2nd and 3rd generation mobile communication systems. TSUNAMI II successfully influenced the standardization process of UMTS (Universal Mobile Telecommunications System) to support adaptive base station antennas by means of optional dedicated pilot channels in downlink (in contrast to a common pilot channel as in IS95). The ACTS project SUNBEAM (Smart Universal Beamforming) followed and concentrated into implementation of flexible multi-mode GSM/WCDMA air interface basestation architectures supporting adaptive antenna arrays. However, the scope of both projects was limited to adaptive antenna array implementations at the base station exclusively. In this paper we compare the performance of spaceTime Block Codes for two transmit antenna (as standardized for UMTS), BLAST architecture and beamforming for the TDD mode of UTRA.
ANT1 FIR
Midamble
We use a stochastic MIMO radio channel model that accounts for both time- and space-domain characteristics based on real environment measurements developed by Aalborg University, a METRA partner [11][12]. The wideband radio channel ( ) can be modelled as:
H
H( ) =
L X
A Æ(
)
l
=1
l
Data
ENC
where l is a nT nR complex matrix whose elements i;j describe the relation between transmit antenna i and receive antenna j at time delay l and takes into account the spatial correlation between antenna elements. With this channel model the received signal can be expressed as:
r(t) = n
L X
A (t)d(t l
=1
) + n(t) l
INT
SPR+SCR
ANT2
FIR
RF
w2
Uplink channel estimate
Figure 1: Transmit Diversity for DPCH
2
R
MS
=
6 6 4 2
R
=
6 6 4
1
: : 0:1615
0 3394
0 0856 1
: : 0:1216
0 2628
0 2550
:
1
: 0:2947
0 2947
1
0 3394
: 0:1379 0:2628
0 0856
1
: 0:2550 0:2417
0 2417
1
: 0:2143
0 2490
:
0 2896
3
: 0:1379 7 7 0:2490 5
0 1615
1
3
: 0:2143 7 7 0:2896 5
0 1216
(3)
(4)
1
which have been obtained through experimental measurements in real indoor scenario. The signal constellation (assumed QPSK) is scaled such that the total transmitted energy in normalized to unity. We assume that the energy is equally distributed along all transmit antennas, hence the average energy out of each antenna is 1=nT .
(1)
l
A
w1 MUX
BS
2. SYSTEM MODEL
RF
(2)
l
(t) is a nR 1 additive white Gaussian noise vector modeled as independent samples of a zero mean complex Gaussian random variable with variance N 0 =2 per dimension. We consider a 4x4 MIMO channel with spatial power correlation matrices for the mobile station (MS) and base station (BS), RBS and RM S respectively,
3. TRANSMIT DIVERSITY SCHEMES The TDD mode of UTRA does not envision using transmit diversity codes for Dedicated Physical Channels (DPCH) and only considers using block space-time codes for the P-CCPCH (Primary Common Control Physical Channel). Figure1 shows the transmit structure as specified by 3GPP. Note that only selective transmit diversity (antenna selector) or beamforming are considered for downlink dedicated channels [13]. In this paper we investigated the performance of space time block codes and BLAST architectures for the dedicated channels. The STBC matrix G is defined as:
G=
s1 s2 s2 s1
(5)
where rows indicated those symbols transmitted from each transmit antenna and stands for complex conjugate.
Beamforming
Pilot Pilot
SPRD
c 1 ,...,c K*m
s 1 ,...,s K
ANT 1
SPRD
W ANT 2
G
Pilot SPRD
2 m -PSK, 2 m QAM Modulator
M U X
Space - Time Block Encoder
Pilot
M U X
ANT 3
SPRD
W ANT 4
Convolutional Code rate 1/4
Beamforming
Pilot SPRD
M U X
ANT 1
M U X
ANT 2
M U X
ANT 3
M U X
ANT 4
Figure 2: STBC and beamforming Pilot SPRD
Space-time block codes for 4 transmit antennas and complex constellations have rate 3=4 or 1=2 and hence they should be combined with proper data puncturing to maintain the same spectral efficiency than the beamforming approach. Instead, we commbined into 2 groups of 2 antennas the available 4 antennas and applied a beamforming to each group of two antennas. The two groups of antennas are shown in Figure 2 for the case when space-time blocks codes are applied. 2.
Figure 3: BLAST transmit architecture with code re-use Downlink INDOOR 3km/h
0
10
Beamforming STBC+beamforming BLAST mem:3 BLAST mem:4
−1
10
−2
BER
10
The STBC decoder is based on a multiuser MMSE detector and a space-time decoder.
−3
10
The Blast architecture is depicted in figure 3.Data is demultiplexed into 4 different subchannels, each one transmitted from each transmit antenna. Since these scheme increases the data rate by a factor of 4, we first applied a convolutional code rate 1/4 to maintain the overall spectral efficiency equal for all architectures under consideration. As described in [10] the detection process for layered space-time coding of all substreams is divided into three key aspects: interference nulling, interference canceling and compensation. Interference nulling projects out interference from those substreams not yet detected, interference canceling substracts out interference of those substreams already detected, finally stronger substreams compensates weaker ones.
4x1
−5
10 −25
−20
−15
−10
−5
0 CIR (dB)
5
10
15
20
25
20
25
Figure 4: Downlink 4 users Downlink PEDESTRIAN 3km/h
0
10
−1
10
BER
Note that with BLAST all parallel subchannels (i.e. data out of each transmit antenna) should be estimated. Since channel codes are limited, users employs the same channel code for each transmit antenna (code re-use), but different scrambling codes and training sequences are used for each antenna for each user. In this way, all parallel channels of all users can be jointly detected and finally those subchannels associated to the same user are properly combined. If no coding is applied the spectral efficiency increases with the number of antennas, being possible to achieve very high data rates with a QPSK modulation per antenna. For this reason this architecture has been proposed as a possible implementation for HSDPA.
4x4 −4
10
−2
10
4x4 4x1 −3
10
Beamforming STBC+beamforming BLAST mem:3 BLAST mem:4 −4
10 −25
−20
−15
−10
−5
0 CIR (dB)
5
10
Figure 5: Downlink 4 users
15
Downlink INDOOR 3km/h
0
will be analyzed in the IST I-METRA project, as a natural extension of the METRA project.
10
−1
10
6. REFERENCES −2
BER
10
4x4
−3
10
[1] S.M. Alamouti, “A simple transmit diversity technique for wireless communications,” Selected Areas in Communications, IEEE Journal on, vol. 16, no. 8, pp. 1451–1458, Oct. 1998.
4x1
−4
10
Beamforming STBC+beamforming BLAST mem:3
−5
10
−6
10 −25
−20
−15
−10
−5
0 CIR (dB)
5
10
15
20
25
Figure 6: Downlink 6 user
4. SIMULATION RESULTS Simulation results for the downlink UTRA-TDD indoor channel are presented in this section. Simulations are made in a slot-basis (i.e. one slot simulated per frame). SF is set to 16 and 4 and 6 intracell interferences are considered. Pilot sequences for each antenna are generated from different Basic Midamble Codes [14] and different scrambling sequences are used at each transmit antenna, hence a single user uses 1 channel code but 4 different pilot sequences. For the BLAST architecture a multiuser detector is used to detect all sub-streams transmitted from each antenna. After multiplexing those data symbols a Viterbi decoder is used to decode the sequence and recover the original transmitted data. Figure 4 and figure 5 shown the BER curves for an scenario with 4 intracell users. We observe that BLAST does not perform better than the beamforming approach, nor STBC. The reason could be the high spatial correlation between antennas, that makes beamforming much more useful rather than trying to exploit parallel subchannels, which indeed do not exist. In figure 5 the performance with 6 intracell users is drawn. Note that BLAST performs much worst than with 4 users. The reason is that channel estimation matrix becomes ill conditioned. 5. CONCLUSIONS The results presented in this paper are part of the METRA project. Further analysis will be performed until the end of the project, focusing on combining different transmit architectures with inner channel coding. New powerful spacetime code and a new promising concept denominated HSDPA, which combines hybrid-ARQ with space-time codes,
[2] H.; Calderbank A.R. Tarokh, V.; Jafarkhani, “Spacetime block coding for wireless communications: performance results,” Selected Areas in Communications, IEEE Journal on, vol. 17, no. 3, pp. 451–460, March 1999. [3] N.; Calderbank A.R. Tarokh, V.; Seshadri, “Spacetime codes for high data rate wireless communication: performance criterion and code construction,” Information Theory, IEEE Transactions on, vol. 44, no. 2, pp. 744–765, March 1998. [4] R.S. Qing Yan; Blum, “Optimum space-time convolutional codes,” Wireless Communications and Networking Confernce, 2000. WCNC. 2000 IEEE, vol. 3, pp. 1351–1355, 2000. [5] J; Vucetic B Firmanto, W; Yuan, “Space-time trellis coded modulation for fast fading channels,” International Symposium on Information Theory and Its Applications, Honolulu, Hawaii, Nov. 2000. [6] T.M. Stefanov, A.; Duman, “Turbo coded modulation for systems with transmit and receive antenna diversity,” Global Telecommunications Conference, GLOBECOM ’99, vol. 5, no. 5, pp. 2336–2340, 1999. [7] E. Hsuan-Jung Su; Geraniotis, “Space-time turbo codes with full antenna diversity,” Communications, IEEE Transactions on, vol. 49, no. 1, pp. 47–57, January 2001. [8] Jr.; El Gamal H. Hammons, A.R., “On the theory of space-time codes for psk modulation,” Information Theory, IEEE Transactions on, vol. 46, no. 2, pp. 524– 542, March 2000. [9] O.Y.; Zhongxin Han Youjian Liu; Fitz, M.P.; Takeshita, “A rank criterion for qam space-time codes with application to turbo coding,” Sensor Array and Multichannel Signal Processing Workshop. Proceedings of the 2000 IEEE, pp. 193–197, 2000. [10] G.D.; Valenzuela R.A.; Wolniansky P.W. Foschini, G.J.; Golden, “Simplified processing for high spectral efficiency wireless communication employing multielement arrays,” Selected Areas in Communications,
IEEE Journal on, vol. 17, no. 11, pp. 1841–1852, Nov. 1999. [11] “Mimo channel characterisation,” METRA Project Deliverable AAU-WP2-D2-V1.1.pdf, , no. D2, pp. available at http://www.ist–metra.org, Dec. 2000. [12] S.H. Jensen J.B. Andersen F. Frederiksen T. B. Srensen J.P. Kermoal, P.E. Mogensen and K.I. Pedersen, “Experimental investigation of multipath richness for multi-element transmit and receive antenna arrays,” IEEE Vehicular Technology Conference VTC 2000 Spring, Tokyo, Japan, vol. 3, pp. 2004–2008, May 2000. [13] “Pysical layer procedures (tdd). ts 25.224 v3.1.0.,” 3rd Generation Partnership Project, Technical Specification Group, Radio Acces Network, Working Group 1 (RAN-WG1)., Dec. 1999. [14] “Pysical channel and mapping of transport channels onto pysical channels (tdd). ts 25.221 v3.1.0.,” 3rd Generation Partnership Project, Technical Specification Group, Radio Acces Network, Working Group 1 (RAN-WG1)., Dec. 1999.