Power Adaptation in Space-Time Block Code Myeongsook Seo and Sang Wu Kim Department of Electrical Engineering Korea Advanced Institute of Science and Technology 373-1 Kusong-dong, Yusong-gu, Taejon 305-701, Korea Abstract- Transmission power adaptations are considered in a space-time block code (STBC) system utilizing feedback channel gain informations. We investigate the optimal power adaptation scheme that minimizes the average BER subject a fixed average transmission power constraint. We present the power gain provided by the optimal adaptation scheme over the conventional STBC without power adaptation. The power gain provided is found to be 13.8dB at BER of in a Rayleigh fading channel. Also, we analyze the performance of STBC system with optimal truncated power adaptation which compensates for fading above a certain fade depth: below the cutoff level data transmission is suspended (i.e. outage condition is declared). Finally, we propose a simple adaptation scheme for STBC systems based on two-bits feedback information that provides a power gain of 4.3dB at . BER of I. I NTRODUCTION Recently, transmit diversity has drawn a lot of interest because of its simplicity of implementation and feasibility of having multiple antennas at the base station. One attractive approach is space-time block code (STBC), proposed by Alamouti [1], which can achieve full diversity without bandwidth expansion. STBC system can achieve the diversity order of for transmit antennas and one receiver antenna. The reason is that STBC maintains orthogonality between the antennas and it can avoid selfinterference in flat fading channels. Furthermore, since the receiver can process the received signal using simple linear processing, the complexity is low. This transmit diversity is currently considered in the WCDMA system [2]. In this paper, we consider the transmission power adaptation in space-time block code (STBC) systems, where the transmitter is equipped with two antennas and the receiver is equipped with one antenna. We assume that channel gain informations at each antenna are obtained at the receiver and then transported over a dedicated feedback link to the transmitter for power adaptations. We investigate the optimal power adaptation scheme that minimizes the average BER under a fixed average transmission power constraint. We show that the optimal adaptation provides a power gain of 13.8dB at BER of over conventional STBC without power adaptation. Then we analyze the STBC system employing optimal trun- cated power adaptation scheme that minimizes the average BER. We show that the optimal truncated power adaptation provides a power gain of 15dB at BER of and outage probability of . The optimal adaptation requires a complete channel gain information, and thus presents the theoretical performance limit. In practice, the feedback channel information may be limited. So, we propose a simple adaptation scheme based on two-bits feedback information (one bit for each antenna). We show that the proposed scheme provides a power gain of 4.3dB at BER of over the conventional STBC without power adaptation. An attempt was made in [3] to adapt STBC according to the available channel information, where a predetermined orthogonal STBC was adapted by means of a linear transformation under a fixed instantaneous transmission power constraint. The remainder of this paper is organized as follows. Section II describes the system model. In Section III, we investigate the optimal adaptation scheme that minimizes the average BER. In Section IV, we propose a simple adaptation scheme based on two-bits feedback information. In Section V, numerical results are presented. Finally, conclusions are made in Section VI. II. S YSTEM MODEL We consider the downlink of a mobile communication system where the base station is equipped with two transmit antennas, and the mobile is equipped with one received antenna. The channel gain between the transmit antenna one and the receiver is denoted by and be tween the transmit antenna two and the receiver is denoted by . We assume that and are constant across two consecutive bits, and they are independent. We assume a single user and BPSK signaling in a frequency nonselective Rayleigh fading channel. Fig. 1 shows the model of proposed STBC system, wherein the gains (weights) , at each antenna is adapted based on the feedback channel information and . The traditional STBC system without power control corresponds to a spe cial case of . The received signals and at time and !#" , respectively, where " is the symbol duration, are given 0-7803-7206-9/01/$17.00 © 2001 IEEE 3188 This indicates that the optimal strategy is to use only one antenna providing the larger gain. Then, it follows from (6) that the average BER is given by by " ! " " ! ! ! 7 8 ;:=? 9 # < >A@9B ! # " DCFEG9H IKJ C L C (8) (1) where the pdf of - is given by $ + where is the average transmission power and and G9H IKJ C MN# OQP SRTVUXW P SRTVUXW are information bits at time and ! " , and and (9) are complex random variables representing noise and interference. The combiner is designed to keep the orthogonality and MN# of STBC so as to detect separately without self (10) %'& $ ( interference. The combiner generates outputs and , MY# where In what follows, we will assume . 7 8 that 9 Now, we want to minimize in (8) by adapting ac ! cording to C , subject to (2) %'& ( Z:=# < [ G H IKJ C L C (11) Then, it follows from (1) and (2) that " ! ! " ! ! ! Using calculus of variations [4], we can show that the optimal is the solution of "! $ ! $ (4) $ Adfe d T aW bC "Oc # (12) b gC "Oc # g h # gC "Oc (13) i is the Lambert W function [5], defined as where d where (3) which yields Thus, the received SNR is given by " # P ]R \_^4` KjFPk + l - j (14) (5) Thus, the average BER with the optimal power adaptation can be obtained from (8) and (13) as III. STBC WITH C OMPLETE C HANNEL I NFORMATION 7 8 9 m:n< u rt #?>poqs td @ b wv gC " #yx C i| D QP SR P SR L Ez{ In this section, we assume that the transmitter is provided with complete feedback channel informations, and , where b is the solution of and adapts the power based on them. We consider adapt# g C " and , subject to %'& ! ( , such that ing the :=# < b #yx E ' % & ( @ v " C d the average BER is minimized. If we let ! , then i} OQP SR P SR L C "!*) " # " # ! - (15) (16) Fig. 2 is a plot of the optimum weighting factor ver- . We find that the optisus the channel gain C of (6) mum is decreasing with $ C ,+ indicating that deep fades $,+ should be compensated by transmitting at a higher power. 0/2143 This may require a large average transmit power and may where -. $ + $ $ and the equality is satisfied cause a large interference to other users. Note that the opwhen timal adaptation requires channel gain informations and , since - ~/2143 if 6 5 $ + . In contrast, the $ $ + (7) required feedback informations in [3] are and . otherwise. $,+ 3189 where b Optimal Truncated Power Adaptation In this subsection, we investigate the optimal truncated power adaptation in a space-time block code system. The transmitter sends data only when the channel gain - is # $ + greater than a given threshold level . Otherwise, data transmission is suspended (i.e. outage is declared). Thus, # the data transmission suffers an outage probability of is the solution of %'& ( # gC " :=< b " C d @ v #yx W i} OQP SR P SR L C E (24) Fig. 3 is a plot of the optimum weighting factor ver# C sus the . # : # W G H IKJ C L C # channel gain of $ + - for As increases, the optimal weighting factor increases, i.e. received SNR in (18) increases. As consequence, we (17) DQP W can expect a reduction of BER at the expense of an inThis type of power control has been analyzed in [6] [7]. creased outage probability. For comparison purpose, we also include the weighting factor for channel inversion The power adaptation # without # truncation corresponds to a ( c C ), where the factor 0.72 has been chosen to special case of X 4 . ! satisfy (11). In order to maintain a constant data rate c " cP independent of the outage, the bit duraIV. STBC WITH L IMITED C HANNEL # # tion during $ + - should be set to " . I NFORMATION Also, to make an average transmission# power of , the - should be set to transmission# power during c . As a result,$ + the received SNR, during # $ + - , is given by In this section, we consider the situation where the chan # "! nel information at the transmitter is limited. We assume " $ + # # i D that the transmitter knows only whether the channel gain D " $ for antenna is above a threshold . The required feed# $ + - (18) back information in this case is two bits (one bit for each antenna). The weighting factor is adapted as Then, it follows from (18) that the average BER is given by 6 5 (25) $ , 7 8 ; 9 : < > @ B A W " DCFEn # G9H IKJ G9H IKJ C L C (19) W where the conditional pdf of $ + - given $ + - given by # is $ for some constants and , subject to %'& ! ( . This is illustrated in Fig. 5. Then, it follows from (4) and (25) the average BER is given by # - C 7 8 9 $ + , $,+ W # DQ SRP PSR C (20) # D 7 Now, we want to minimize 98 in (19) by adapting according to C , subject to G H IKJ G H IKJ C :=< [ G H IKJ C %'& ( ; L C W Using calculus of variations, the optimal given by dfe for C b gC "Oc # h # gC "Oc :=< > oqs u rt # p td D W C i| DQP SR P SR L @ b v # gC # " x :# : 6 # >A@ B # I R P S R L C L C i?P S m: ! < : # " # : < : < >A@ B ! " >A@ B I R P S R L C L C i?P S (21) # " C ! C C ! ! C E C E C E I R P SR L C L C (26) i?P S ) P ! T , we can show that Using the inequality > 3 (22) ) #" I 97 8 #" (27) ! is It follows from (19), (20) and (22) that the average BER with the optimal truncated power adaptation is given by 7 8 9 where #" I Ez{ #" (23) 3190 . . :# P : < P $ &% '` $ &% W '` ` " I R S )(R C L W ` " RS )(R C L (28) 7 We want to minimize the upper bound on 98 ject to g , sub- the total power on one antenna providing a higher channel gain and controlling the transmission power on that antenna according to the Lambert W function of the chan%'& ! ( (29) nel gain. The optimal adaptation is found to provide a power gain of 13.8dB at BER of over the convenIt follows from (25) that this constraint can be restated as tional STBC without power adaptation. The optimal strategy requires a large transmit power to ! (30) D compensate for deep fades. In order to avoid the large transmit power (thereby avoiding a strong interference to where other users in multi-user systems), we considered the opZ: < . P SR L C (31) timal truncated power adaptation, wherein the transmit ter sends data only when # channel gain $ + - is greater . Otherwise, data transmission than a threshold level Since the upper bound is minimized by minimizing " I ! is suspended and an outage is declared. The power gain #" , the power adaptation problem is characterized by over the conventional STBC without# power adaptation in creases as the outage probability increases. The opti# " # " Min " I " I ! (32) mal truncated power control is found to provide a power gain of 15dB over the conventional STBC without power subject to D ! . adaptation at BER of and outage probability of . These optimal strategies require a complete channel gain information, and thus presents the theoretical perV. N UMERICAL R ESULTS formance limit. # In practice, the feedback channel information may be Fig. 4 is a plot of average BER versus 9 "Oc with limited. So, we proposed a simple adaptation scheme optimal power adaptation. We find that the STBC with opbased on two-bits feedback information (one bit for each timal power adaptation provides a power gain of 13.8dB antenna): The transmitter needs to know only whether the over the conventional STBC without power adaptation at BER of . When the optimal truncated power adapta- channel gain $ for antenna is above a threshold. This tion is applied to STBC, the power gain over the conven- adaptation scheme is found to provide a power gain of tional STBC without power adaptation increases at the ex- 4.3dB over the conventional STBC without power adap# pense of increased outage probability . Figure 4 shows tation at BER of . that the optimal truncated power adaptation provides a power gain of 15dB over the conventional STBC without power adaptation at BER of and outage probability References # of . We also find that the power gain provided [1] S. M. Alamouti, “A simple transmit diversity techby the optimal power adaptation in (13) over the channel nique for wireless communications,” IEEE Jounal inversion is 0.3dB at BER of . on Selected Areas in Communications, Vol. 16, No. Fig. 6 is a plot of the optimal weighting # factors 8, pp. 1451-1458, October 1998. 9 and and the threshold versus . We find "Oc that the optimal threshold decreases with the transmis[2] 3GPP, “Technical Specification 25.214, Physical sion power . Also, the optimal increases with 9 , layer procedures(FDD),” June 2000. whereas the optimal is almost invariant to 9 .# . The [3] G. Jongren and B. Ottersten, “Combining transmit Fig. 7 is a plot of average BER versus 9 "Oc antenna weights and orthogonal space-time block adaptive STBC scheme with limited feedback information (two-bits feedback) provides a power gain of 4.3dB over codes by utilizing side information,” in Proc. 33th the conventional STBC without power adaptation at BER Asilomar Conference on Signals, Systems and Comof . puters, Oct. 1999. VI. C ONCLUSIONS [4] R. Courant and D. Hilbert, Methods of Mathematical Physics, Vol. I, New York:Interscience, 1953. We considered transmission power adaptations in a space-time block code (STBC) system. We investigated the optimal adaptation scheme that minimizes the average BER under a fixed average transmission power constraint. We presented the power gain provided by the optimal adaptation scheme over the conventional STBC without power adaptation. The optimal strategy is to allocate [5] R. M. Corless et al., “On the Lambert W function,” Advances in Computational Mathematics, Vol. 5, pp.329-359, 1996. [6] Sang Wu Kim and Andrea J. Goldsmith, “Truncated power control in code-division multiple access communications,” IEEE Trans. Veh. Technol., Vol. 49, No. 3, pp. 965-972, May 2000. 3191 [7] A. J. Goldsmith and S. G. Chua, “Variable-data variable-power MQAM for fading channels,” IEEE Trans. on Commun., Vol. 45, pp. 1218-1230, Oct. 1997. h1 = α1e jθ1 b2 b1 STBC − b2 b1 Encoder b1 b2 PTTw1 PTTw2 r2 , r1 h2 = α2e Power adaptation jθ 2 Combiner h1 h2 Channel estimator b̂1 b̂2 h1 , h2 (b) Receiver (a) Transmitter Figure 1: Block diagram of the proposed system. 7 PTT/N0= 2dB PTT/N0= 4dB PTT/N0= 8dB PTT/N0= 10dB PTT/N0= 12dB 6 5 w 4 3 2 1 0 0 1 2 3 4 g Figure 2: The optimal weighting factor C versus channel gain C for complete feedback information. 3192 10 PTT/N0=10dB P0=0 8 -2 P0=10 -1 P0=10 Channel inversion (w=0.72/g) w 6 4 5.0 5.0 4.5 4.5 4.0 2 4.0 q1 q2 3.5 3.5 q1 , q2 0 0.0 0.5 1.0 1.5 2.0 g Figure 3: The optimal weighting factor C versus channel gain C for complete feedback information when truncated power adaptation is applied. 10 -1 10 -2 2.5 Threshold η 2.0 2.0 1.5 1.5 1.0 1.0 0.5 0.5 0.0 0.0 0 5 10 15 20 PTT/N0[dB] Figure 6: The optimal weighting factors , and the # . threshold versus 9 "Oc 0 10 3.0 2.5 STBC without power adaptation STBC with channel inversion (w=0.72/g) STBC with optimal power adaptation when P0=0 -2 STBC with optimal power adaptation when P0=10 -1 Average BER STBC with optimal power adaptation when P0=10 10 -3 1x10 -4 1x10 -5 10 -6 10 -7 10 -8 10 -9 5 10 15 20 25 PTT/N0[dB] # Figure 4: Average BER versus 9 "Oc with complete channel information. 10 1 10 0 No diversity STBC without power adaptation STBC with two-bits feedback adaptation STBC with channel inversion STBC with optimal power adaptation when P0=0 Average BER -2 10 -1 10 -2 10 -3 1x10 -4 1x10 -5 10 -6 STBC with optimal power adaptation when P0=10 -1 STBC with optimal power adaptation when P0=10 wi q1 5 10 15 20 25 PTT/N0[dB] q2 Figure 7: Average BER versus 9 "Oc η Threshold η 3.0 g Figure 5: Weighting factors , versus channel gain C with two-bits feedback information. 3193 # .