WIRELESS LANS ABSTRACT INTRODUCTION

Transcription

1 WIREESS AS CO-CHAE ITERFERECE CACEATIO BASED O MIMO SYSTEMS UCA GIAGASPERO AD UIGI AGAROSSI, PHIIPS RESEARCH MOZA GIOVAI PATEGHI, CEFRIE SHUTAI OKAMURA, OSAKA UIVERSITY MIORU OKADA, ISTITUTE OF SCIECE AD TECHOOGY SHOZO KOMAKI, OSAKA UIVERSITY RX antenna RX antenna RX antenna m RX antenna M OF demod OFD demod MIMO systems are currenty stimuating considerabe interest across the wireess industry as they appear to be a key technoogy for future wireess generations. ABSTRACT This artice deas with the expoitation of mutipe input mutipe output (MIMO) systems for broadband wireess indoor appications. Two systems, the Wind-Fex and Ubiquitous Antenna, are considered. Both aim to increase the system capacity by different approaches, to increase the singe ink data rate and the number of users in the whoe system, respectivey. Computer simuation resuts show the effectiveness of both MIMO systems. ITRODUCTIO The deveopment of wireess communication systems for high-bit-rate data transmission and high-quaity information exchange between terminas is becoming one of the new chaenging targets in teecommunications research. The market demand for broadband mutimedia services, ubiquitous networking, and Internet access via portabe devices is expected to grow enormousy, pushing the deveopment of modem and system architectures for high-bit-rate transmission. Mutipe input mutipe output (MIMO) systems are currenty stimuating considerabe interest across the wireess industry because they appear to be a key technoogy for future wireess generations. An (,M)-MIMO wireess system can be generay defined as a MIMO system in which signas are transmitted by antennas at the same time using the same bandwidth and, thanks to effective processing at the receiver side based on the M received signas by M different antennas, is abe to distinguish the different transmitted signas. The processing at the receiver is essentiay efficient co-channe interference canceation on the basis of the coected mutipe information. This permits improving system capacity whether the interest is to increase the singe ink data rate or increase the number of users in the whoe system. The Wind-Fex MIMO and the Ubiquitous Antenna System are exampes of these different ways to expoit MIMO properties. For the Wind-Fex MIMO system [], the target is to increase the capacity of a singe ink. The singe user transmitter (Tx) and receiver (Rx) are equipped with and M antennas, respectivey. The basic idea is to usefuy expoit the mutipath rather than mitigate it, considering the mutipath itsef as a source of diversity that aows the parae transmission of independent substreams from the same user. This is quite different from a time-division mutipexing (TDM), frequency-eivision mutipexing (FDM), or code-division mutipexing (CDM) technique: there is no expicit orthogonaization of the substreams. Instead, it is the rich mutipath environment that makes the substreams independent of each other. Since the user data is sent in parae over antennas, the effective bit rate is increased by a factor. To separate the simutaneousy transmitted signas at the receiver, the space time coding and space-division mutipexing techniques are used. Space time coding introduces a spatiotempora correation among transmitted signas to improve the information protection, whie the goa of space-division mutipexing is to increase the data rate. Severa different space time coding techniques have been considered: space time treis codes [], space time bock codes [3], and D-BAST [4]. A of them appear to be too compex for Wind-Fex-specific appication, so space-division mutipexing [5] has been chosen. Among space-division mutipexing techniques, V-BAST [6, 7] seems to exhibit the best tradeoff between performance and compexity. The V-BAST agorithm [6, 7] impements a noninear detection technique based on a spatia nuing process combined with symbo canceation to improve performance. The spatia nuing process, based on the zero forcing (ZF) approach (a minimum mean square error approach is aso possibe), is used to separate the individua substreams. Conceptuay, each substream in turn is considered to be the desired /0/$ IEEE IEEE Wireess Communications December 00

2 signa, and the remainders are considered as interferers to be canceed (nuing). To improve performance, symbo canceation is used in conjunction with spatia nuing. Symbo canceation means that after the strongest substream has been detected, the contribution of this substream is subtracted from the received signas. Thus, the remaining weaker substreams are easier to recover since the strongest substream has been removed as a source of interference. This process is reiterated unti a the substreams have been detected. Different from Wind-Fex, the target of the Ubiquitous Antenna System [8, 9] is to increase system capacity as indicated by the tota number of users in the system. The Ubiquitous Antenna System is composed of mutipe microceuar radio base stations (RBSs) depoyed over the service area and radio-on-fiber (RoF) ink [0] which connects RBSs to the centra contro station (CCS). In this system, the transmitted signas from mobie terminas (MTs) are propagated through mutipath channes and received at the pura of the RBSs. The received radio frequency (RF) signa at each RBS is moduated into the intensity of the optica carrier and the moduated optica signa is sent to the CCS via RoF ink. At the CCS, a the optica signas from the RBSs are converted to RF signas again. Then the CCS performs a the demoduation sequences and signa processing. Since a the received signas at the RBSs are coected to the CCS, we can empoy more sophisticated signa processing such as co-channe interference canceation [] and joint detection at the CCS. Hence, the Uubiquitous Antenna System can achieve space-division mutipe access (SDMA), which aows mutipe MTs to operate in the same time sot and frequency channe, and then improve the tota number of users and frequency utiization efficiency of broadband wireess access systems. Both systems adopt the orthogona FDM () moduation scheme even if they have different features concerning the number of subcarriers, subcarrier moduation, and so on. The choice of the is basicay due to its interesting properties as a wideband moduation scheme and is generay accepted in the context of wireess indoor communications (e.g., IEEE 80.a). The artice is organized as foows. In the next section the Wind-Fex system mode and simuation resuts are presented. The third section is dedicated to the Ubiquitous Antenna System description and performance resuts. Concuding remarks are given in the fina section. WID-FEX SYSTEM WID-FEX OVERVIEW The aim of the Wind-Fex project [] is to design and deveop a wireess high-bit-rate fexibe and configurabe modem architecture, which works in singe-hop ad hoc networks and provides wireess access to the Internet for sowy moving users (about m/s) in an indoor environment. A basic requirement for the Wind-Fex modem is the capabiity to manage variabe bit rates, ranging adaptivey from 64 kb/s to more than 00 Mb/s of payoad, to bring new services and trigger new interesting and appeaing appications with different quaity of service (QoS) requirements. The 7 GHz unicensed frequency band has been chosen since the frequency bands around.4 and 5 GHz have aready been aocated and are cose to saturation. The avaiabe band of 00 MHz ( GHz) is divided into four 50 MHz wide channes not simutaneousy seectabe. Due to the adopted carrier frequency, the coverage ranges from 5 m for non-ine of sight (os), to 0 m for ine of sight (os) using omnidirectiona antennas. Subcarrier moduation schemes are binary phase shift keying (BPSK), quaternary PSK (QPSK), 6-quadrature ampitude moduation (QAM), and 64-QAM. The consteation of each subcarrier is adaptivey chosen among the various schemes according to the subcarrier signa-tonoise ratio (SR), the target bit error rate (BER), and medium access contro (MAC) requests. With respect to the channe coding, the coding scheme of the Wind-Fex modem is a parae convoutiona turbo code. The code rates estabished are /, /3, and 3/4. At the MAC ayer of the Wind-Fex system, the bit transmission is organized in a way time-division mutipe access/time-division dupex (TDMA/ TDD). The upper interface of DC ayer wi serve Internet Protoco version 4 (IPv4) and version 6 (IPv6). This seection gives the widest possibe range of potentia appications to the Wind-Fex modem. SYSTEM MODE MIMO techniques are based on the assumption of a fat fading channe. This requirement is obviousy not verified in a 50 MHz wide wireess indoor channe, such as the Wind-Fex one. However, the use of moduation makes the fat fading hypothesis true for each subband, aowing expoitation of the MIMO approach for broadband wireess appications as we. The transmitter is equipped with antennas (Fig. ). A traditiona channe encoder encodes the source information bits. The coded bits are then mapped on the symbos of the consteation adopted for each subcarrier. The vectoria nature of the transmission is introduced by demutipexing ( ) these symbos. In the considered architecture this demutipexer represents the space encoder. It maps symbos on the space channes, which are substreams of the same user. A seria to parae converter for each space channe takes of these symbos to form the input for the moduator. is the number of frequency channes, that is, the number of subcarriers (in WID-FEX = 8). To avoid any intersymbo interference (ISI) due to the deay spread of the channe, a cycic prefix is appended to each symbo. The corresponding antenna transmits the output from each moduator. This structure permits the simutaneous transmission of M-QAM symbos in the same bandwidth and with the same tota transmitted power required by an symbo to carry ony of them. The receiver is equipped with M antennas (note that V-BAST requires M ). Each The use of moduation makes the fat fading hypothesis true for each sub-band, aowing the expoitation of the MIMO approach aso for broadband wireess appications. IEEE Wireess Communications December 00 9

3 We assume that the channe remains fixed during a frame, but it randomy changes from frame to frame. This is a totay reasonabe assumption in wireess appications where there is imited mobiity, such as the Wind-Fex one. Info bits Information source Channe encoder Code bits C k,n, Symbos Mapping bit symbo k: discrete time index n: Space channe index : Frequency channe index Demux n S/P S/P S/P moduator moduator moduator TX antenna TX antenna TX antenna n TX antenna Figure. Transmitter architecture. antenna receives a different noisy superposition of the faded versions of the transmitted signas (Fig. a). If the antennas are sufficienty spatiay separated (more than λ/) and there is sufficienty rich scattering, the transmitted signas arriving at different receive antennas undergo independent fading. Moreover, if the channe response is known at the receiver, V-BAST agorithm is abe to detect the transmitted signas. Channe response can be estimated at the receiver using a training sequence embedded in each TDMA frame. In so doing, we assume that the channe remains fixed during a frame, but it randomy changes from frame to frame. This is a totay reasonabe assumption in wireess appications where there is imited mobiity, such as Wind-Fex. At each symbo time, the output of the demoduator corresponding to the receive antenna m is a set of signas, r m,, one for each frequency channe rm, = Hmn,, Cn, + ηm, with =,, n= where H m,n, is the compex coefficient representing the frequency response of the channe from the transmit antenna n to the receive antenna m at muticarrier frequency, and η m, are independent sampes of a compex Gaussian random variabe with zero mean and variance 0, representing noise (note that 0 is the variance of the noise at the receiver input [6]). The M outputs reated to the muticarrier frequency are the inputs to a V-BAST signa processor that detects the different M-QAM symbos transmitted in this frequency channe. There are V-BAST signa processors, each operating in a specific frequency channe (Fig. b). The M-QAM symbos, obtained at the output of the V-BAST signa processors, are then seriaized by a parae to seria converter in order to compete the traditiona receiver processing. Concerning the channe mode, the singe input singe output (SISO) mode adopted to cacuate the H m,n, coefficients is the one presented in [3]. The os channe is modeed with 8 taps. The taps ampitudes have been shaped foowing different probabiity density functions (pdfs): a combination of exponentia and Weibu pdfs for the first bin and exponentia pdfs for the others. In the considered context, however, the SISO mode is not sufficient to competey characterize the channe, since the MIMO approach aso entais consideration of the spatia correation characteristics of the transmission system. Severa studies have recenty demonstrated that if channe path gains of a (,M) MIMO system are independent, the channe capacity scaes ineary with n, where n = min(,m) [4]. The probem is that in rea propagation conditions, these channe coefficients coud be partiay correated. Generay, the correation of the coefficients depends on many factors such as the physica parameters of transmit and receive antennas (e.g., antenna spacing), and the characteristics and distribution of the scatterers. It has been observed that when paths are correated, the channe capacity can be significanty smaer than when they are not. This suggests that further research is required into the spatia correation probem [5]. In this work, the correation has been taken into account by means of a synthetic parameter (i.e., the mean correation coefficient of the MIMO channe). Some correation matrices K (the generic eement k i,j of the correation matrix 0 IEEE Wireess Communications December 00

4 RX antenna RX antenna RX antenna m RX antenna M demoduator demoduator demoduator, C, C V-BAST signa processor for frequency C n m, channe M, C, C +, C + V-BAST signa processor for frequency C +n m, channe M, C n r, = Σ H,n, 'C + n, η, n = n r, = Σ H,n, 'C + n, η, n = n r m, = Σ H n = m,n, 'C + n, η m, n r M, = Σ H n = M,n, 'C + n, η M, P/S (a) Symbos Demapping symbo bit V-BAST signa processor for frequency channe Code bits n Channe decoder Info bits C (-)+ C (-)+ C (-)+n C (-)+ Destination In our simuations, we have assumed that the received signas are corrupted by additive white Gaussian noise. We have aso assumed idea symbo and sampe-cock synchronization at the receiver. We have considered the non-coded system performance.,, m, M, V-BAST signa processor for frequency channe C (-)+ C (-)+ C (-)+n C (b) Figure. Receiver architecture: a) demoduation and space channe detection; b) V-BAST signa processors detection and traditiona receiver processing. is the correation coefficient between the ith and jth path gains) have been defined with an increasing vaue of the mean correation coefficient ranging from 0 (totay uncorreated path gains) to 0.8 (amost competey correated path gains) with step granuarity of 0.. SIMUATIO RESUTS In order to verify the performance of the considered architecture in idea and nonidea propagation conditions (i.e., the presence of spatia correation between path gains), a simuator, based on Wind-Fex specifications, has been impemented. In our simuations, we have assumed that the received signas are corrupted by additive white Gaussian noise. We have aso assumed idea symbo and sampe-cock synchronization at the receiver. We have considered noncoded system performance. The BER, as a function of the SR at each receive antenna, has been evauated. The first set of performance curves refers to idea propagation conditions (i.e., the path gains can be considered independent). The impact on the performance of and M has been anayzed. First, a system having the same number of transmit and receive antennas has been simuated. In Fig. 3 (eft) the performance of this system, with 6-QAM on each subcarrier, for different is reported. IEEE Wireess Communications December 00

5 0 0 (,) system 6-QAM = M = = M = 3 = M = (,M) system 6-QAM 0 0 BER BER SR at each RX antenna = M = = M = 3 = M = 3 = 3 M = 4 = M = 4 = 3 M = 5 SR at each RX antenna Figure 3. A system performance comparison for different vaues of = M (eft) and for different vaues of M (right). This figure proves that the V-BAST technique permits increasing the bit rate without significanty worsening the BER. Remember, in fact, that in idea propagation conditions the bit rate of the SDM systems scaes ineary with the number of transmit antennas. It is important to note that the tota transmitted power and used bandwidth are constant. The performance curves of a MIMO system with different numbers of receive antennas, M, using 6-QAM on each subcarrier, have aso been evauated (Fig. 3, right). As expected, increasing M improves performance. This is due to the avaiabiity of a greater number of received signas, which can be combined in a more efficient way to obtain a more accurate estimate of the transmitted signas. Moreover, some other simuations indicate that the performance improvement is a function of the difference M, resuting in being substantiay independent from or M separatey. Concerning the nonidea propagation condition, the anaysis has been imited to the (,) and (3,3) MIMO systems. Figure 4 shows the simuation resuts for a (3,3) system with different vaues of the mean correation coefficient among paths. The mean correation coefficient ranges from 0 to 0.8 with steps of 0.. The anaysis of these curves shows an interesting noninear reation between performance degradation and the mean correation coefficient. Moreover, the negative effect of the spatia correation becomes more and more evident as SR increases. Under a threshod SR, BER performance is imited mainy by the presence of noise. The curves show that the SR 0 0 (3,3) system 6-QAM 0 0 (3,3) system 64-QAM 0 0 BER BER Uncorreated paths Mean corr. coeff. = 0. Mean corr. coeff. = 0.4 Mean corr. coeff. = 0.6 Mean corr. coeff. = SR at each RX antenna Uncorreated paths Mean corr. coeff. = 0. Mean corr. coeff. = 0.4 Mean corr. coeff. = 0.6 Mean corr. coeff. = SR at each RX antenna 5 30 Figure 4. Performance of the (3,3) MIMO system for different mean correation coefficients and moduations. IEEE Wireess Communications December 00

6 (a) E/O,O/E RBS oss of the (3,3) system due to spatia correation does not exceed 5 db up to a mean correation coefficient equa to 0.6 in the considered SR range. It must be pointed out that 0.6 is a high vaue of correation. Beyond this vaue the use of a MIMO system woud be inefficient for the specific propagation environment. The resuts for the (,) system exhibit simiar properties. UBIQUITOUS ATEA SYSTEM Radio-on-fiber ink Figure 5. Genera description of (a) the ubiquitous antenna system and (b) the centraized antenna system. CCS E/O,O/E, modem, signa processing (b) RBS modem, signa processing UBIQUITOUS ATEA SYSTEM OVERVIEW A genera description of the Ubiquitous Antenna System is shown in Fig. 5a. The system is composed of mutipe microceuar RBSs depoyed over the service area, a CCS, and the RoF ink. Each RBS has ony eectrica-to-optica (E/O) and optica-to-eectrica (O/E) converters, and requires no RF moduation or demoduation functions. The radio signas received at the RBSs are appied to the E/O converter to moduate the intensity of the optica carrier, and they are sent to the CCS via the RoF ink, maintaining the radio signa waveform. The CCS converts the optica signas from RBSs to eectrica signas again by corresponding O/E converters, and then performs a the signa processing and demoduation sequences. Since a the signas received at the RBSs are obtained at the CCS, co-channe interference canceation and joint detection can be performed at the CCS simiar to the adaptive array antenna systems []. The ubiquitous antenna system enabes the mutipe MTs to operate at the same frequency channe simutaneousy by making effective use of joint detection. Thus, it does not require frequency aocation for each RBS. That is, a the RBSs can use a whoe band assigned to the system. Hence, the ubiquitous antenna system achieves higher frequency utiization efficiency whatever the assigned bandwidth is stricty imited. Furthermore, since the CCS has expensive components such as RF components, moduator, demoduator, and signa processing components such as a joint detector, and RBSs requires O/E and E/O converters, the ubiquitous antenna system aows us to reduce the tota cost to depoy and maintain the system. A ubiquitous antenna system is aso known as a distributed antenna system, which has been studied in recent years [6, 7]. In a distributed antenna system, antenna eements are distributed over the service area as a microceuar RBS and their outputs are brought, in anaog form, to a centra server. Simiar to the adaptive array, interference canceation is performed at the centra server. Since the distance between antenna eements is much onger than in a conventiona centraized adaptive array antenna system (Fig. 5b), we can further obtain macrodiversity. In [0], Cark et a. evauated and compared centraized and distributed antenna systems. From their remarks, the distributed antenna system aows us to obtain improvement in system capacity. SYSTEM MODE In order to evauate the performance of the ubiquitous antenna system, a system mode is considered in this artice. The system mode is iustrated in Fig. 6. In the foowing, we concentrate on the upink connection. et us assume that MTs in the service area transmit C signas simutaneousy at the same frequency channe. At the nth MT, the binary streams are error correction encoded and intereaved over each C subcarrier. The encoded binary streams are mapped onto the QAM symbos at the moduator. In order to estimate the channe impuse response (CIR) between each MT and each RBS, piot symbos are attached before data symbos. The symbos are then inverse discrete Fourier transformed (IDFT). The guard interva, aso known as the cycic extension, is inserted in order to remove ISI due to deay spread. After that, the signa is transmitted to the RBSs through the radio propagation channe. In a radio propagation path, the transmitted signa from each MT is affected by fading, cochanne interference, path oss, and propagation deay. The signas are received at the M RBSs depoyed over the service area. The received signa at each RBS is corrupted by additive white The ubiquitous antenna system enabes the mutipe MTs to operate at the same frequency channe simutaneousy by making effective use of joint detection. Thereby, it does not require the frequency aocation for each RBS. That is, a the RBSs can use a whoe band assigned to the system. IEEE Wireess Communications December 00 3

7 Transmitter RBS Conv. corder Bit intereaver S/P k QAM k mod. QAM mod. IFFT GI insertion E/O Piot insertion Fading Hm Propagation deay Co-channe interference E/O Radio-on-fiber ink Receiver O/E MMSE-DC Propagation deay O/E FFT FFT Σ MSE normaization MSE normaization MSE k QAM demod. k QAM demod. P/S Bit deintereaver Soft-decision Viterbi decoder H opt Channe estimator Piot generator Figure 6. The system mode. Gaussian noise (AWG) as we as co-channe interference from other MTs. The received signa is converted to an optica signa by an E/O converter, and sent to the CCS via the RoF ink. Since the spacing among RBSs is hundreds of meters, about severa hundred nanoseconds of deay difference arises in the RoF ink. In genera, this deay difference coud be a probem in wideband transmission. Fortunatey, since the C signa has a guard interva inserted at the head of each C symbo in order to avoid ISI and interchanne interference due to deay spread, the deay difference can be ignored. At the CCS, the O/E converters convert the optica signas from RBSs to eectrica signas again. For the sake of notationa convenience, we define the M-dimensiona received signas vector as y k k k k = [y, y,, ym] T, where k denotes the subcarrier index. The received vector is then given by y k = H k x k +z k, where x k k k k = [x, x,, x] T is the -dimensiona transmitted signa vector, z k k k = [z, z,, k z M]T is the -dimensiona AWG vector, where the eements z k m are independent and identicay distributed Gaussian random variabes with zero mean and variance σ n. x T denotes the transpose of x. The frequency response matrix H k = [h k mn ] is an M matrix whose eement h k mn denotes the response between the nth MT and the mth RBS. In this artice, we assume that the frequency responses for different MTs and RBSs are statisticay independent, stationary, and compex Gaussian random variabes. Joint Detection As a joint detector, we empoy the minimum mean square error diversity combiner (MMSE-DC). The MMSE-DC performs the joint detection by using M signas form RBSs. Before the MMSE-DC, the received signa from each RBS is divided into each subcarrier by the corresponding FFT processor. Then the MMSE- DC is performed subcarrier by subcarrier. The received signa from each RBS is weighted by the optimum weight matrix given by H k opt = (H k )H(Rk yy ), where Rk yy is M M correation matrix of received signas. The weighted signas are then combined. The output of the MMSE-DC is given by x k = H k opt y k, where x is the estimated transmitted signas vector. Channe Estimation ot ony can the MMSE- DC-based joint detection described previousy mitigate the performance degradation due to fading, it can aso remove the co-channe interference transmitted from other MTs. However, in order to estabish joint detection, estimation of CIR is required at the receiver. In the proposed system, each MT inserts the piot symbo satisfying the optimum MSE condition discussed in [8]. The receiver estimates the CIR by cacuating the correation between the received signa at each RBS and reference piot signa and the estimated CIR. Then the receiver estimates H k n =[h k n, h k n,, h k mn ] T by transforming CIR into the frequency domain using FFT, where H k n is an M-dimensiona channe response vector. 4 IEEE Wireess Communications December 00

8 MSE ormaization At the output of the MMSE- DC, a subcarriers have the same desired signa power, whie the noise power at each subcarrier is different. On the other hand, the foowing Viterbi decoder uses the Eucidean distance as the path metrics. That is, the decoder is optimum in terms of minimizing the BER when a the noises are independent and identicay distributed (i.i.d.) Gaussian random variabes. In order to perform the Viterbi decoder in the optimum condition, the output signa of the MMSE-DC is normaized by the MSE or the noise variance of the corresponding signa. The MSE for the nth MT is given by MSE k n = σ d (H k n ) H (Rk yy ) H k n, where σ d is the desired signa power. This normaization is performed subcarrier by subcarrier as we as for the MMSE-DC. The noise at the output of the MSE normaization is i.i.d., and the Viterbi decoder performs as the optimum decoder. Then the normaized signas are demoduated and deintereaved. The deintereaved signa is then appied to the soft-decision Viterbi decoder. Finay, we can obtain the desired user s binary streams. SIMUATIO RESUTS In this section we anayze the performance of the proposed ubiquitous-antenna-based wireess A system by computer simuations. In the foowing, we evauate the BER and frequency utiization efficiency of upink connection in comparison with other systems. The C parameters are based on the IEEE 80.a standard. The entire channe bandwidth is divided into 64 subchannes. The 48 subchannes are used to transmit data. Differentia quadrature phase shift keying (DQPSK) is empoyed as a subchanne symbo moduation format. The symbo duration is 4. µs incuding.0 µs of guard interva, which is just a itte bit onger than that of the WA standards in order to mitigate performance degradation due to the deay difference among RoF inks. For forward error correction (FEC), haf-rate convoutiona error correction coding with constraint ength of 7 (64-state) is empoyed. In the system, each MT can transmit a signa with a data rate of.5 Mb/s over a 5 MHz channe, and its transmission efficiency is approximatey 0.77 b/s/hz. As a propagation channe, two-sampe spaced two-ray Rayeigh fading channe with Dopper shift, f d = 40 Hz, is assumed. The interva between two rays is 50 ns. The path oss exponent is 4.0. We aso assume the symbo timing of the C signas is synchronized at every MT. We ignore the frequency offset between the MTs and CCS oca osciators, and noniner distortion due to the RoF ink. First, we anayze the BER performance of the ubiquitous antenna system in order to demonstrate the basic characteristics of the proposed system and the effect of the MSE normaization scheme. In this simuation, we assume that two RBSs are depoyed in the horizonta axis, and each RBS is connected to the CCS by the RoF ink. The distance between RBSs is 00 BER Joint detection with MSE normaization Joint detection without MSE normaization Without joint detection 5 E b / 0 (db) Figure 7. BER performance of the ubiquitous antenna system. m, and it causes 500 ns of deay difference due to the RoF ink. Two MTs are ocated at the center of two RBSs. The BER performance of the proposed system against E b / 0 is shown in Fig. 7. The BER performance without joint detection is intoerabe for transmitting data. However, the ubiquitous antenna system with joint detection can drasticay improve the BER performance. Furthermore, in contrast to the performance without joint detection, the proposed system without MSE normaization drasticay improves BER performance. Furthermore, MSE normaization gives a gain of 0 db at BER = 0 3 by making effective use of the impicit subcarrier diversity effect. ext, we anayze the frequency utiization efficiency of the ubiquitous antenna system. The aocation of the RBSs is shown in Fig. 5a. In this simuation, we assume 6 ces, and the ubiquitous antenna system is composed of four RBSs corresponding to the four ces in the center of the area. Each RBS is ocated at the center of the ce and connected to the CCS with the RoF ink. Each ce size is 00 m. A the MTs transmit C signa simutaneousy at the same frequency channe. The ubiquitous antenna system ony demoduates the signas from MTs in the centra four ces, and signas from the other MTs around the four ces are considered co-channe interference. The ubiquitous antenna system performs the MMSE-DCbased joint detection by using the signas received at four RBSs. For comparison purposes, we evauate the performance of the ubiquitous antenna system without joint detection. In this case, a the RBSs have their own demoduators and each RBS demoduates the received signas independent of the other RBSs. A the MTs are sti IEEE Wireess Communications December 00 5

9 Frequency utiization efficiency per ce (bits/s/hz) Ce size: 00 m fd = 40 Hz 0 ormina E b / 0 (db) Figure 8. Frequency utiization efficiency. Ubiquitous with joint detection Centraized with joint detection Ubiquitous w/o joint detection Conventiona TDMA Centraized w/o joint detection 0 30 operating at the same frequency. Furthermore, we evauate the macroce system, or one RBS ocated at the center of the service area as shown in Fig. 5b. In the foowing, we ca it the centra antenna system. We assume the centra antenna system with and without joint detection. With joint detection, the RBS has a four-eement array antenna, and MMSE-DC-based joint detection is performed using the array antenna. In the centra antenna system without joint detection, however, the RBS has ony one antenna eement, and no joint detection is performed. As in the ubiquitous antenna system, a the MTs operate at the same frequency simutaneousy. Moreover, we aso evauate the conventiona digita communication system in which a the MTs access the RBS, ocated at the center of the four ces and equipped with ony one antenna eement, in TDMA mode. In a cases, a the MTs are uniformy distributed over the service area. In this simuation, we assume the transmission is successfu when there is no bit error in a packet composed of a piot symbo and 0 information symbos. Figure 8 shows the frequency utiization efficiency against nomina E b / 0. The transmission power of an MT is determined so that E b / 0 per branch equas nomina E b / 0 at one RBS antenna that is 70 m apart from the corresponding MT. In this simuation, we assume four MTs are in the four ces. Furthermore, eight MTs are on the outside of these four ces as co-channe interce interferers. In the conventiona system, the efficiency is about 0. b/s/hz since it aows ony one MT to operate at the same frequency channe at one time. In the centraized system, it improves the efficiency to 0.5 b/s/hz by using joint detection. On the other hand, in the ubiquitous antenna system, the efficiency is improved to 0.65 b/s/hz by making effective use of joint detection. Furthermore, even the ubiquitous antenna system with seection diversity can achieve higher efficiency than a centraized one with joint detection in the ower E b / 0 region. COCUSIOS This artice focuses on two MIMO -based systems, Wind-Fex and the ubiquitous antenna system. Their capabiity to increase overa system capacity has been investigated and the simuation resuts on their performance are reported. The Wind-Fex system, based on the V- BAST MIMO and moduation scheme, proves capabe of greaty improving bit rate without increasing tota transmitted power or required bandwidth. The simuation resuts confirm this feature for both idea and nonidea propagation conditions. The ubiquitous antenna system, which is composed of mutipe RBSs depoyed over the service area, the CCS, and an RoF ink, aows a the MTs to operate at the same frequency simutaneousy. The computer simuation resuts show the ubiquitous antenna system is capabe of improving overa system capacity. This property is very important in designing wireess networks, especiay when the radio resource is stricty imited. REFERECES []. Giangaspero, G. Patenghi and. Agarossi, A MIMO architecture for Wireess Indoor Appications, Proc. IEEE Int. Conf. Wireess As and Home ets., Singapore, Dec. 00, pp [] V. Tarokh,. Sehadri, and A. R. Caderbank, Space-Time Codes for High Data Rate Wireess Communications: Performance Criterion and Code Construction, IEEE Trans. Info. Theory, vo. 44, Mar. 998, pp [3] S. M. Aamouti, A Simpe Transmit Diversity Technique for Wireess Communications, IEEE JSAC, vo. 6, no. 8, Oct. 998, pp [4] G. J. Foschini, ayered Space-Time Architecture for Wireess Communication in a Fading Environment When Using Muti-Eement Antennas, Be abs Tech. J., Autumn 996. [5] A. van Zest, Space Division Mutipexing Agorithms, 0th Mediterranean Eectrotech. Conf., vo. 3, pp. 8. [6] P. W. Woniansky et a., V-BAST: An Architecture for Reaizing Very High Data Rates Over the Rich-Scattering Wireess Channe, invited paper, Proc. ISSSE 98, Pisa, Itay, Sept [7] G. J. Foschini et a., Simpified Processing for High Spectra Efficiency Wireess Communication Empoying Muti-Eement Arrays, IEEE JSAC, vo. 7, no., ov. 999, pp [8] M. Toyama, M. Okada, and S. Komaki, Maxima Ratio Combining Macro Diversity for Micro-Ceuar Sotted AOHA, IEICE B-I, vo. J79-B-I, 5, May 996, pp [9] S. Okamura, M. Okada, and S. Komaki, On the Performance of the Ubiquitous Antennas for the Reception of C Signas, Proc. IEEE Int. Conf. W As and Home ets., Singapore, Dec. 00, pp [0] S. Komaki et a., Proposa of Radio Highway etworks for Future Mutimedia-Persona Wireess Communications, ICPWC 94, Bangaore, India, Aug. 994, pp [] Y. i and R. Soenberger, Adaptive Antenna Arrays for Systems With Cochanne Interference, IEEE Trans. Commun., vo. 47, no., Feb.999, pp [] [3] M. obeira et a., Parameter Estimation and Indoor Channe Modeing at 7 GHz for -Based Broadband WA, IST Mobie Commun. Summit 000, Gaway, Ireand, Oct. 000, pp [4] G. J. Foschini and M. J. Gans, On imits of Wireess Communications in a Fading Environment when Using Mutipe Antennas, W Pers. Commun., vo. 6, no.3, March 998, pp IEEE Wireess Communications December 00

10 [5] D.-S. Shiu et a., Fading Correation and Its Effects on the Capacity of Mutieement Antenna Systems, IEEE Trans. Commun., vo. 48, no. 3, Mar. 000, pp [6] M. V. Cark et a., Distributed Versus Centraized Arrays in Broadband Wireess etworks, Proc IEEE. VTC, Rhodes, Greece, MA-, May 00. [7] A. Skavos et a., Joint Channe Estimation in Muti- User System, Proc. 6th Int. Wks., Sept. 00. [8] Y. i,. Seshadri, and S. Ariyavisitaku, Channe Estimation for Systems with Transmitter Diversity in Mobie Wireess Channes, IEEE JSAC, vo. 7, Mar. 999, pp BIOGRAPHIES UCA GIAGASPERO graduated in eectronic engineering, teecommunications speciaization, from the Poytechnic of Bari in 00. In the same year he received a Master s degree in information technoogy, in the fied of advanced transmission systems, from CEFRIE, Poytechnic of Mian, where he worked on MIMO wireess transmission systems. He is now a research scientist at Phiips Research Monza in the wireess systems and terminas area. He is currenty invoved in an IST European research project aimed at defining a fexibe high-rate wireess modem for indoor appications. His main research interests are adaptivity and reconfigurabiity for future-generation wireess systems. UIGI AGAROSSI (uigi.agarossi@phiips.com) received a degree in eectronic engineering from the University of Boogna in 98. He started working as a research scientist in the GTE Teecommunications Radio ink R&D Department, within the area of baseband and IF signa processing. In 987 he joined Phiips Research Centra aboratories (ATAB), Eindhoven, The etherands. After one year he moved to Phiips Research Monza aboratory, Itay, where he started working on storage systems and architecture. He has experience in optica storage, optica channe modeing, and detection for highdensity optica recording. He is the author of internationa patents and papers, and has taken part in many European projects. Since 997 he has been working on wireess systems and terminas, and is currenty invoved with the Wireess Indoor Fexibe High Bitrate Modem Architectures (Wind-Fex) European project as a project eader. Current interests are digita baseband processing and moduation for high-capacity systems, software-defined radio (SDR), and adaptive and reconfigurabe radio. GIOVAI PATEGHI (pateng@cefrie.it) received a degree in eectronic engineering (summa cum aude) from the University of Brescia in 999. He has been with CEFRIE- Poitecnico di Miano since 999 and currenty is a researcher in the Advanced Mobie and Wireine Transmission Systems unit. His main research interests are in the fieds of residentia/soho wireess systems and fiber optic communication systems. In particuar, he is carrying out research on moduation, MIMO transmission systems, optica transport networks, and optica devices for inteigent WDM networks (optica add/drop mutipexers and optica crossconnects). He is a member of WID-FEX, an IST European project devoted to the design of a highbit-rate fexibe and configurabe wireess architecture for indoor environments. SHUTAI OKAMURA (okamura@roms.comm.eng.osaka-u.ac.jp) received a B.E. degree in eectrica and eectronic engineering from Shizuoka University, Japan, in 000, and an M.E. degree from Osaka University, Japan, in 00. He is currenty pursuing a Ph.D. degree at Osaka University and engaged in research on radio communication systems. He is a member of the Institute of Eectronics and Information Communication Engineers (IEICE) of Japan. MIORU OKADA (mokada@is.aist-nara.ac.jp) received a B.E. degree in communications engineering from the University of Eectro-Communications, Tokyo, Japan, in 990, and M.E. and Ph.D. degrees, both in communications engineering, from Osaka University, Japan, in 99 and 998, respectivey. In 998 he joined the Department of Communications Engineering, Osaka University, as a research associate. From 999 to 000, he was with the University of Southampton, United Kingdom, as a visiting research feow. In 000 he joined the Graduate Schoo of Information Science, ara Institute of Science and Technoogy, Japan, where he is currenty an associate professor. He is a member of the IEICE, IEEE, and Institute of Image Information and Teevision Engineers of Japan (ITE). He received the Young Engineer Award from IEICE in 999. SHOZO KOMAKI [SM] (komaki@comm.eng.osaka-u.ac.jp) received B.E., M.E., and Ph.D. degrees in eectrica communication engineering from Osaka University, in 970, 97, and 983, respectivey. In 97, he joined TT Radio Communication aboratories, where he was engaged in repeater deveopment for a 0 GHz digita radio system, and 6-QAM and 56-QAM systems. From 990 he moved to Osaka University, Facuty of Engineering, and engaging in the research on radio and optica communication systems. He is currenty a professor at Osaka University. He is a member of the IEICE and (ITE). He was awarded the Paper Award and the Achievement Award of IEICE in 977 and 994, respectivey. The Wind-Fex system, based on the V-BAST MIMO and moduation scheme, shows the capabiity to greaty improve the bit-rate without increasing the tota transmitted power or the required bandwidth. The simuation resuts confirm this feature for both idea and non-idea propagation conditions. IEEE Wireess Communications December 00 7