TD-CDM-OFDM: Evolution of TD-SCDMA Toward 4G

WIRELESS COMMUNICATIONS IN CHINA: TECHNOLOGY VS. MARKETS TD-CDM-OFDM: Evolution of TD-SCDMA Toward 4G Kan Zheng, Lin Huang, and Wenbo Wang, Beijing University of Posts & Telecomms Guiliang Yang, Datang Mobile Co ABSTRACT TD-SCDMA, which is the homemade 3G standard in China, has received considerable attention and is believed to play a critical role in the development of China s mobile communication. Meanwhile, advanced MIMO and OFDM techniques shed light on the feasibility of highperformance 4G broadband systems. We discuss an evolutionary path of TD-SCDMA toward 4G systems in this article, which combines the existing advanced traits in TD-SCDMA with new features for broadband wireless communication systems. INTRODUCTION China has become the largest market for mobile phone services in the world. There is no sign of this market stopping its growth thus far. It is predicted that the total number of mobile service subscribers will reach much more than 300 million by 2005, as shown in Fig.1. Global System for Mobile Communications (GSM) and narrowband IS-95 code-division multiple access (CDMA), which are the second-generation (2G) mobile systems, have been growing steadily and successfully in the last decade. Their success promoted the evolution from 2G to third-generation (3G), while a variety of wireless systems, including General Packet Radio Service(GPRS), cdma2000 1X and wireless local area network (WLAN), have also been developed all over the country. However, these systems were designed independently, targeting different service types, data rates, and users. None of them is capable of replacing the other systems to meet all the services. Meanwhile, 3G mobile systems including the homemade time-division synchronous CDMA (TD-SCDMA) standard [1] are expected to land in China soon. Although the peak data rates of the current 2.5G and evolving 3G systems are promised to be up to 384 kb/s and 2 Mb/s, respectively, the average throughput per user is not guaranteed to be more than 171 kb/s during busy hours. Thus, the 2.5G and 3G systems are qualified to provide fundamental services such as voice, basic data communications, and low-speed wireless Internet access, but not to provide new interactive multimedia services such as multiparty videoconference and video on demand, wherein data rates up to 100 Mb/s are required frequently. In order to provide high-data-rate services, a new 4G system, possibly based on new radio access technology, is currently envisioned in several research programs such as China s FuTURE and the European MATRICE. A number of technologies have been proposed for 4G systems including orthogonal frequency-division multiplexing (OFDM) and multi-input multi-output antennas (MIMO) technologies. OFDM is widely recognized as the modulation scheme for broadband wireless communication systems due to its ability to eliminate the intersymbol interference (ISI) caused by multipath channels. In order to achieve high spectrum efficiency and enlarge system coverage, multiple antennas must be employed in 4G systems as well to support high data rates. Besides, there are several advanced characteristics in 3G TD-SCDMA that can be directly applied in 4G systems. CDMA as one of the promising multiple access methods can be combined with OFDM. Time-division duplex (TDD) allows the flexible allocation of the ratio between uplink (UL) and downlink (DL), which makes it the top candidate to provide both symmetrical services such as voice and asymmetrical services such as mobile Internet access. The air interface of 4G will be a mixture of mature technologies used in 3G and newly developed ones targeting higher data rates. Our vision of a 4G system, referred to as TD- CDM-OFDM (time-division code-division multiplexing OFDM), is that it should be a system with the wide coverage to spread over both hot spot and cellular areas using different modulation schemes. Flexible DL and UL capacity allocation, the advanced detection mechanism, the combination of CDMA with OFDM, and multiple antenna technologies will make TD-CDM- OFDM a strong candidate for the evolving 4G system. IEEE Communications Magazine January 2005 0163-6804/05/$20.00 2005 IEEE 45

Number of subscribers (Millions) 500 400 300 200 100 0 258 2003 339 2005 Year n Figure 1. The mobile market in China and its growth. 405 2007 The article is organized as follows. First, we briefly review the promising characteristics and limitations of TD-SCDMA. Then we discuss several key enabling technologies for TD-CDM- OFDM systems. Furthermore, the architecture of TD-CDM-OFDM systems in the physical layer is provided. Finally, conclusions are given. PROMISING CHARACTERISTICS AND LIMITATIONS OF TD-SCDMA The Chinese Academy of Telecommunications Technology (CATT) proposes a system running TD-SCDMA to meet the challenges of highdata-rate transmission. TD-SCDMA is a combination of three principal multiple access technologies: frequency-division multiple access (FDMA), time-division multiple access (TDMA), and synchronous CDMA. TDD is incorporated as well inside the TD-SCDMA system to provide asymmetric and flexible data service between its DL and UL. Besides, TD-SCDMA makes use of a variety of existing techniques to optimize its spectrum efficiency. These techniques, such as joint detection (JD), smart antenna (SA), uplink synchronization, and dynamic channel allocation (DCA), are implemented to eliminate intracell interference and significantly reduce intercell interference, thus leading to a considerable improvement in spectrum efficiency. To be more specific, JD allows the receiver to estimate the radio channel and process all users signals simultaneously, thus eliminating multiple access interference (MAI), minimizing intracell interference, and increasing transmission capacity. For the SA technique, it can both efficiently mitigate multi-user interference and improve reception sensitivity by assigning a different narrow beam to each user in a smart way. As a consequence, SA offers increased system capacity and cell coverage. Moreover, by periodically adjusting the transmission timing of each individual mobile terminal with closedloop control, TD-SCDMA improves the accuracy of UL synchronization, thus reducing the calculation time for position location and handover search. Since the air interface of TD- SCDMA takes advantage of three available multiple access techniques, it can make full use of these degrees of freedom by DCA, minimizing intercell interference. However, TD-SCDMA has its own limitations. First, the position of midamble in each slot reduces its capability to track the fast fading characteristic and the Doppler effect in the mobile environment, thus limiting terminal mobility. Second, in order to support high data rates up to 2 Mb/s, the spreading factor has to be decreased to one, resulting in poor interference resistance. This may cause some other problems when receiving signals under serious multipath fading. Third, TD-SCDMA uses exactly the same orthogonal variable spreading factor (OVSF) code as proposed in wideband CDMA (WCDMA), which cannot solve ratematching problems well for multimedia traffic due to its code generation tree structure. It does provide flexible control of the data rates in the sense of multiples of twice the basic rate, but do nothing more to support services with other rates in future systems. Therefore, a new wireless communication system compatible with current TD-SCDMA must be developed to meet the demand for higher mobile speed, higher data rates up to more than 2 Mb/s, and more flexibility for multirate services. On the way toward 4G networking one of the major challenges is the (r)evolution in the physical layer, discussed later in this article. ENABLING TECHNOLOGIES TDD is very likely the right choice for 4G system duplex mode. TDD provides higher flexibility in assigning the capacity ratio between UL and DL than frequency-division duplex (FDD), which leads to better efficiency in bandwidth utilization. Also, appropriate selection of modulation and multiple access schemes is crucial to satisfy the high data rate requirement, support multimedia services efficiently, and provide greater immunity against the severe effect of frequency selective fading over broader signal bandwidth. Moreover, multiple antenna technology is promising and indispensable in achieving high spectral efficiency. Finally, link adaptation technologies, where signal transmission parameters such as modulation/coding levels are dynamically adapted and retransmission strategies are properly selected according to changing channel conditions, have emerged as powerful approaches not only to increase the data rate but also to improve the spectral efficiency of wireless communication systems. TDD FDD and TDD are two commonly used duplex modes in wireless communication systems. Different frequency bands are used for DL and UL transmission in an FDD system, while the same frequency band is used alternatively during its specified time slot in a TDD system. Therefore, flexible spectrum allocation may be easier to implement in TDD than in FDD when spectrum is a valuable resource. In a TDD system, the 46 IEEE Communications Magazine January 2005

length of slots can be unequal, and the number of DL and UL slots per frame can be different as well. The advantage of using TDD is the capability to accommodate asymmetric high-bitrate services for the DL and UL, which will be one of the prominent features in 4G systems. In addition, the channel reciprocity between DL and UL in TDD can be utilized for link adaptation, such as adaptive beamforming, transmission diversity, and adaptive modulation. The accomplishment of link adaptation can enlarge system throughput and simplify the receiver structure. Finally, the cost of the radio frequency transceiver is lower in TDD than in FDD since no strict spectral isolation between the transmitting and receiving multiplexing is required. MODULATION AND MULTIPLE ACCESS SCHEMES CDMA has gained the most attention in 3G mobile communication systems primarily because of its higher capacity over other multiple access techniques. However, the performance of wideband CDMA suffers from multiple access interference (MAI) and ISI due to the severe multipath fading effect when the data rate is high up to 100 Mb/s. To better utilize the spectrum resources for multimedia services with satisfactory bit error rate (BER), OFDM has been proposed to aid CDMA in mitigating the severe effect of multipath fading while maintaining bandwidth efficiency. OFDM adopts longer symbol duration with a guard period to avoid frequency selectivity. It also minimizes the separation between adjacent carriers to increase frequency efficiency. Incorporated with OFDM, CDMA can bring in better performance, and is becoming one of the most promising multiple access candidates for future broadband mobile communication systems. Various OFDM-CDMA schemes [2] have been proposed; they can be categorized into several groups according to code spreading direction. The first is to spread the original data stream in the frequency domain; the second is to spread it in the time domain, similar to the DS- CDMA scheme. Consequently, a frequency or time Rake receiver will be used, respectively. The former scheme, which is usually referred to as multicarrier CDMA (MC-CDMA), can obtain a good frequency Rake diversity effect through the despreading operation since the fading of each subcarrier is different. However, such a scheme cannot achieve the time diversity gain by itself. The latter scheme is a good scheme in which to introduce OFDM technology into DS- CDMA systems, especially for the quasi-synchronous mobile communication environment. However, the frequency diversity gain, the main advantage of using such technique, cannot be achieved without a good combination of channel coding and interleaving in the frequency domain. Therefore, besides the two groups mentioned above, spreading in two dimensions, which exploits both time and frequency diversity [3], is a good alternative to the conventional approach that spreads in only the frequency or time direction. With two-dimensional spreading, the maximum diversity gain in time/frequency domain can be achieved by using a sufficiently long onedimensional spreading code to spread data all over two dimensions. The spreading pattern has to be appropriately designed in both domains in order to ensure that all chips with spread data fade independently. MIMO It has been demonstrated that MIMO technology has the potential to significantly improve the capacity and performance of a wireless system. Several space-time processing techniques have also been developed in recent years. Therefore, it is natural to combine two powerful technologies, MIMO and OFDM, in the physical layer design. An attractive approach for the transmit diversity technique is space-time block code (STBC) [4] based on orthogonal design, which achieves full diversity with a simple linear maximum likelihood(ml) decoder. It utilizes orthogonal design to separate signals from different transmit antennas, and its decoding algorithm is very simple linear combining because of the orthogonality. On the other hand, spatial multiplexing techniques, such as Bell Laboratories layered space-time (BLAST) [5], that can dramatically increase the frequency efficiency have drawn considerable attention recently because they can provide high-data-rate communication without increasing transmission power and bandwidth. In BLAST, a high-rate data signal is divided into a set of lower-rate streams, each of which is encoded, modulated, and transmitted at a different antenna. The receiver separates the different signals using a spatial equalizer and an interference cancellation scheme. Furthermore, the number of antennas at a mobile terminal is often not greater than that at the base station because of limitations on hardware implementation at the terminal side in practice. It is fairly easy to apply BLAST on the uplink in MIMO systems since the number of receiver antennas a BLAST detector requires is greater than or equal to the number of independent transmit antennas. In the DL, combining spatial multiplexing with transmit diversity (i.e., combining BLAST with STBC) can reduce the number of receive antennas to half or less and achieve both increased data rate and more diversity gain, which is one of the best solutions for the DL. LINK ADAPTATION The basic idea of link adaptation is to adjust transmission parameters and schemes to take advantage of variations in channel conditions. Fundamental quantities to be adapted include modulation/coding levels and retransmission strategies. Others can also be adjusted for the benefit of systems, such as transmission power levels, spreading factors, and antenna weights. Among all those link adaptation techniques, adaptive modulation and coding (AMC) and hybrid-automatic request repeat (H-ARQ) are two of the most efficient algorithms that have already been successfully implemented in highspeed downlink packet access (HSDPA) [6]. The principle of AMC is to change the modulation and coding format in accordance with variations in channel conditions, subject to system restrictions. The main benefits of AMC are: In the downlink, combining spatial multiplexing with transmit diversity, i.e., combining BLAST with STBC, can reduce the number of receive antennas to half or less and achieve both increased data-rate and more diversity gain, which is one of the best solutions for the downlink. IEEE Communications Magazine January 2005 47

Carrier Bandwidth Duplex mode 5.8 GHz 65 MHz TDD Total number of subcarriers 1024 OFDM symbol duration Subcarrier separation Number of subcarriers used 832 FRAME STRUCTURE As shown in Fig. 3, a radio frame with duration of 5 ms is subdivided into 10 main time slots (TSs) of 473.6 µs duration each and three special time slots: DL synchronization (DLSync), switch point, and UL synchronization (ULSync). Time slot TS0 is always for DL and TS1 is for UL, whereas the other time slots can be either UL or DL, depending on flexible switching point configuration. The burst structure of the main time slots consists of data blocks, pilot signals, and a guard period of 9.4 µs. In the time domain, the period of one OFDM block is 14.8 µs including an effective block period of 12.8 µs and a cyclic prefix of 2.0 µs. A TS of 473.6 µs is equivalent to 32 OFDM symbols. The fast Fourier transform (FFT) size is 1024 and the total bandn Table 1. TD-CDM-OFDM system parameters. 12.8+2.0 µs (cyclic prefix) (1024 + 160 samples) 78.1 khz Number of virtual subcarriers, left/right 95 + 96 Number of DC subcarriers 1 Duration of guard period (GP) per TS 9.4 µs Duration of TS Number of fixed pilot OFDM 2 symbols per TS Duration of DLSync 44.4 µs Duration of ULSync 59.2 µs Duration of switch point 66.4 µs 473.6 µs (32 OFDM symbols + GP) Channel coding/decoding Turbo coding(1/2, 2/3, 3/4 )/ Max_Log_MAP (iteration = 8) Modulation Configuration of multiple antennas Application environment QPSK, 16-QAM, 64-QAM BS (4 antennas)/ MT (2 antennas) Microcell (mobile speed 3 km/h) Macrocell (mobile speed 30 120 km/h) Higher data rates available for users in favorable positions, which in turn increases the average throughput of the cell Reduced interference variation due to link adaptation based on variations in the modulation/coding scheme rather than in transmit power In AMC, explicit signal-to-noise ratio (SNR) measurements or similar measurements are used to select the modulation and coding format, whereas in H-ARQ link layer acknowledgments are used for retransmission decisions. AMC by itself does provide some flexibility to choose an appropriate modulation and coding scheme for the channel conditions, determined through either terminal measurement or network measurement. However, accurate measurement is required, and there is a delay effect. Compared to AMC, H-ARQ is an implicit link adaptation technique. It autonomously adapts to instantaneous channel conditions and is insensitive to measurement error and delay. Combining AMC with H-ARQ leads to the best use of both approaches: AMC provides gross data rate selection, while H-ARQ is for fine data rate adjustment based on channel conditions. Future 4G systems with OFDM and MIMO technologies will have much more flexibility for link adaptation by providing many subchannels in the frequency and/or space domain. For example, the system can perform: Adaptive subcarrier, bit, and power allocation that fully exploits the characteristics of the OFDM link Adaptive transmission power according to the MIMO links TD-CDM-OFDM ARCHITECTURE ON THE PHYSICAL LAYER Following the introduction of the new technologies to cover the needs for high data rates and new services, the integration of existing TD-SCDMA into a common platform is an important objective for the evolution to 4G systems. Therefore, the design of a generic multiple access scheme for new wireless systems is essential. A generic architecture allows capacity optimization with seamless transition from TD-SCDMA to the future TD-CDM- OFDM system. TD-CDM-OFDM, which absorbs several good characteristics of TD- SCDMA, can provide much higher-data-rate service where new OFDM and MIMO technologies are well combined. Multiple antennas are employed at both base station (BS) and mobile terminal (MT) in order to achieve high spectrum efficiency and enhance cell range. The main system parameters of TD-CDM- OFDM are shown in Table 1. SPECTRUM ALLOCATION The air frequency spectrum is a limited resource. How to select the frequency carrier depends not only on the requirements of the system, but also on minimizing the impact of interference to other existing systems. A large part of the spectrum has been allocated for mobile communication in the 3 8 GHz bands. In 3G systems, extra spectrum was required because of the increased number of users and rapid growth of mobile data services. Accordingly, China s Radio Management Bureau has allocated new spectra of paired 2*90 MHz and unpaired 155 MHz for future 3G deployments of FDD and TDD systems, respectively, including the 2 GHz core band and the nearby supplemental band as shown in Fig. 2. Therefore, not much band is left at 2 GHz. Since 4G systems require more bandwidth, the unlicensed band in the 5.8 GHz range, where larger bandwidth is available, becomes very attractive. 48 IEEE Communications Magazine January 2005

MHz 1700 1800 1900 2000 2100 2200 2300 2400 MS TX BS TX MS TX BS TX China 1755-1785 1850-1880 1920-1980 2110-2170 1880-1920 2010-2025 2300-2400 FDD core band TDD core band FDD supplemental band TDD supplemental band n Figure 2. 3G spectrum allocations in China. For signal detection at the terminal station side, the single-user detection with MMSE principle is applied before demodulation, which is a good tradeoff between the implementation complexity and the receiver performance. width is divided into 78.1 khz per subchannel. Spectrum null subcarriers (i.e., virtual subcarriers) are put on both sides of the spectrum for the sake of implementation. Furthermore, in order to avoid the direct current (DC) problem, a null subcarrier is also put in the middle of the spectrum (i.e., the DC subcarrier is not used). The remaining 832 subchannels are used for data transmission, thus resulting in an efficient 65 MHz bandwidth. DOWNLINK With its flexible frequency reuse, OFDM- CDMA is chosen in order to achieve high link capacity in the DL. Each user will be distinguished by its orthogonal spreading sequence. Different spreading types with variable spreading factors will be applied depending on the channel conditions such as delay spread, Doppler spread, intercell interference, and so on. In a slow fading environment with high SNR, spreading in the time domain (MC-DS- CDMA) is preferred because it is easier to maintain orthogonality among different users spreading signals by spreading in the time domain than in the frequency domain under such an environment. However, if frequency selective fading is very serious with low SNR, spreading in the frequency domain (MC- CDMA) will lead to better performance because of its ability to achieve frequency diversity gain. Moreover, with increased mobile speed (i.e., in a fast fading channel), two-dimensional spreading in both time and frequency Frame (5 ms) Time slot (473.6 µs) GP Switch point 9.4 µs (66.4 µs) TSO TS1 TS2 TS3 TS9 Pilot Data 32 OFDM symbols (473.6 µs) DLsync (44.4 µs) ULsync 59.2 µs 32 OFDM symbols (473.6 µs) Code Code User K User 2 User 1 OFDM-CDMA (DL) Frequency User 1 User 2 User K OFDM-CDMA (UL) Frequency n Figure 3. The frame structure in a TD-CDM-OFDM system. IEEE Communications Magazine January 2005 49

PHY mode (DL) Modulation Code rate Maximum rate (Mb/s) 1 QPSK 1/2 44.9 2 QPSK 3/4 67.4 3 16-QAM 1/2 89.9 4 16-QAM 2/3 119.8 5 16-QAM 3/4 134.8 6 64-QAM 2/3 179.7 7 64-QAM 3/4 202 n Table 2. TD-CDM-OFDM modulation and coding schemes for downlink. domains can fully exploit the diversity gain that can be achieved. Since the number of transmit antennas (i.e., four) is larger than that of receive antennas (i.e., two) in the DL, no STBC with a code rate of one exists. So the combination of STBC and vertical BLAST (V-BLAST) [7] is applied in TD-CDM-OFDM to achieve both good system performance and high spectrum efficiency. Channel estimation is performed based on the time-multiplexed pilot structure as shown in Fig. 3. Under a slow fading channel, two OFDM symbols with reference data at the beginning and end of each TS are transmitted periodically. For detection, the channel frequency response (CFR) during data OFDM symbols can be assumed to be the same as the estimated CFR from these two pilot symbols. With increased mobile speed, the channel state cannot be assumed to be constant within one TS. The CFRs for data detection can be interpolated from both sides of estimated CFRs by pilot symbols. If fast fading becomes more serious (e.g., mobile speed up to 160 km/h), more pilot OFDM symbols must be arranged in the middle of the TS in order to achieve more accurate channel estimation. For signal detection on the terminal station side, single-user detection with minimum mean square error (MMSE) is applied before demodulation, which is a good trade-off between implementation complexity and receiver performance. The frame synchronization and estimation of the timing/frequency offset are realized by the training sequence transmitted in the DLSync TS and pilot symbols in each TS. As shown in Table 2, the use of flexible channel coding and modulation in TD-CDM- OFDM provides up to seven physical (PHY) modes, a combination of forward error correction (FEC) and modulation. The data rates that can be supported are in the range of 10 kb/s 200 Mb/s, depending on the coverage and channel conditions. Under good channel environment, higher-order modulation such as 16- quadrature amplitude modulation (QAM) or 64-QAM with small spreading factor is used to achieve higher spectrum efficiency. In a dense cellular system with high interference and timefrequency selectivity, the lowest-order modulation, quaternary phase shift keying (QPSK), with the highest spreading factor in both directions is employed. UPLINK Different from the DL, the extension of orthogonal frequency-division multiple access (OFDMA) by code-division multiplexing (CDM) [8] will be adopted in the UL as a multiple access scheme. It applies OFDMA for user separation, which can cope with a certain amount of asynchronization. In addition, it uses CDM on data symbols belonging to the same user. The CDM component is introduced in order to achieve additional frequency diversity gain. Like MC-CDMA, this OFDMA exploits the advantages brought by the combination of spread spectrum and multicarrier modulation. The basic difference between OFDMA-CDM and MC-CDMA is that in the former, CDM is used for the simultaneous transmission of the data of one user on the same subcarriers, whereas it is for the transmission of the data of different users on the same subcarriers in the latter. OFDMA assigns each user its own subcarrier subset according to an FDMA scheme, whereas MC-CDMA is a CDMA scheme. Since one user exclusively uses each subset of subcarriers, the MAI caused only by asynchronism between different users is very small. And the self-interference of one user can easily be decreased by interference cancellation since all superimposed modulated spreading codes of its subcarrier subset are affected by the same channel fading. When considering a cellular system, a frequency reuse factor of one can be realized by using different scramble codes in neighbor cells, and intercell interference can be avoided by selecting different subcarrier sets for users in neighboring cells if the system load is not heavy. Considering the antenna configuration in TD- CDM-OFDM, it is straightforward to adopt space multiplexing (i.e., V-BLAST) in the UL. FDMA in the uplink enables low-complexity channel estimation, which is quite difficult and requires more overhead in the MC-CDMA UL. Similar to the DL, the pilot OFDM symbols are put at both sides of each time slot for the uplink, where the reference signals for different users are transmitted at different users designated subcarrier subsets on the frequency domain but simultaneously on the time domain. Because of the reciprocity between the DL and UL in the TDD system, the estimation of the channel state and the timing/frequency offset in the downlink can be used to facilitate uplink synchronization. When accessing the system, an MT searches the training sequence in the DLSync TS from a nearby BS. The MT will roughly estimate its next transmission time and power level according to the detected arrival time and power level of the received sequence in DLSync. Meanwhile, the frequency offset estimated from the DL can be compensated for before UL transmission at MTs. Then another training sequence will be sent in the ULSync TS with the estimated timing and power level from the MT. Once the BS detects the output from one MT, it will determine the necessary timing offset and power level adjustment by finding the peak correlation in ULSync. This information can be fed back to 50 IEEE Communications Magazine January 2005

the MTs by the closed-loop control channel. According to the received messages, the MTs must adjust both their transmission time and transmission power levels and try to synchronize to the BS timing, achieving time alignment at the receiver side (i.e., the BBS). After UL timing is finished, UL frequency offset estimation will be performed using the well designed pilot sequences in each TS. Furthermore, the transmission of a wide range of multimedia services with different data rates can easily be supported by an OFDMA scheme with a variable number of subcarriers/ codes per user. PERFORMANCE EVALUATION For the purpose of software evaluation, the multipath channel is modeled as a finite tapped delay line with L = 18 Rayleigh fading paths of maximum delay τ max = 1.05µs. The multipath intensity profile of the channel is according to IST MATRICE CHANNEL C [9]. Each path of the channel is modeled by a classical Doppler spectrum in simulations where the Doppler frequency f d up to 96 Hz is taken into account for mobile speed up to 18 km/h. The ratio of the energy per bit (E b ) to the spectral noise density (N 0 ) per antenna is defined by Required E b /N 0 (db) 14 12 10 8 6 4 2 0 1 1 0.1 TD QPSK NTT QPSK TD 16-QAM NTT 16-QAM 2 4 C mux (a) 8 SISO downlink channel C 16 QPSK C mux = 1 QPSK C mux = 16 16QAM C mux = 1 16-QAM C mux = 16 32 Eb C SF =, N0 I Rc log2m where C/I is the carrier-to-interference ratio per antenna, M is the size of the data symbol alphabet, SF is the spreading factor, and R c denotes the code rate. For the sake of simplification, in BLAST detection only MMSE linear detection without interference cancellation is applied in both the DL and UL, where the performance can be furthermore improved if ML detection or iterative interference cancellation is adopted. In Fig. 4a we compare the required average E b /N 0 for achieving BER = 10 4 between DL TD-CDM-OFDM with SF = 32 and DL NTT OFCDM systems [10] employing 2D spreading with SF = SF Time SF Freq = 16 2 in a singleinput single-output (SISO) configuration. Under low system load (i.e., the number of multiplexed codes C mux is small), the performance of TD-CDM-OFDM is much better than that of NTT OFCDM because a higher frequency diversity effect is achieved in the former and the effect of intercode interference due to destroyed code orthogonality is relatively minor. Although the influence of intercode interference becomes more serious with increased C mux, the gain of TD-CDM-OFCDM still exists in high system load. Figures 4b and 4c provide an assessment of a MIMO TD- CDM-OFDM system in terms of BER in various modulation schemes and system loads, respectively. The performances are good enough because of the space diversity introduced by MIMO and the frequency diversity due to spreading in the frequency domain. Similar to SISO cases, system performances are also deteriorated with increased intercode interference. Bit error rate Bit error rate 0.01 1E-3 1E-4 1E-5 1E-6 1 1 0.1 0.01 1E-3 1E-4 1E-5 1E-6 6 0 1 2 3 4 5 6 7 8 9 E b /N 0 per antenna n Figure 4. BER performances of TD-CDM-OFDM systems: a) SISO downlink; b) MIMO downlink; and c) MIMO uplink. (b) MIMO downlink V-BLAST and STBC channel C 5 4 3 2 1 0 1 2 3 E b /N 0 per antenna (c) QPSK C mux = 1 QPSK C mux = 32 16-QAM C mux = 1 16-QAM C mux = 32 MIMO uplink V-BLAST channel C IEEE Communications Magazine January 2005 51

The requirements of large coverage, high data rate, and high spectrum efficiency in TD-CDM-OFDM can be met by the combination of two powerful technologies in the physical layer: MIMO and OFDM. CONCLUSION In this article we propose an evolutionary path from TD-SCDMA to future 4G systems based on TDD. It shows that TDD mode is very promising for 4G systems since it offers a number of advantages: more flexible capacity allocation, better frequency utilization efficiency, better resource allocation for packet services, and inherent channel reciprocity to facilitate link adaptations. In the end, the requirements of large coverage, high data rate, and high spectrum efficiency in TD-CDM-OFDM can be met by the combination of two powerful technologies in the physical layer: MIMO and OFDM. ACKNOWLEDGMENTS This research has been funded in part by China National 863 Project (2003AA12331004 and 2004AA123160). REFERENCES [1] CWTS-WG1, Physical Layer General Description, TS C101, V3.1.1, Sept. 2000. [2] R. Prasad and S. Hara, An Overview of Multi-Carrier CDMA, IEEE ISSSTA, Mainz, Germany, Sept. 22 25, 1996, pp. 107 14. [3] K. Zheng, G. Zeng, and W. Wang; Performance Analysis for Synchronous OFDM-CDMA with Joint Frequency-Time Spreading, IEEE ICC, June 20 25, 2004, Paris, France. [4] M. Alamouti, A Simple Transmit Diversity Technique for Wireless Communications, IEEE JSAC, vol. 16, Oct.1998, pp. 1451 58. [5] G. J. Foschini, Layered Space-time Architecture for Wireless Communication, Bell Labs Tech. J., vol. 1, Autumn 1996, pp. 41 59. [6] 3GPP,TR25.848, Physical Layer Aspects of UTRA High Speed Downlink Packet Access. [7] P. W. Wolniansky et al., V-BLAST: An Architecture for Realizing Very High Data Rates over the Rich-Scattering Wireless Channel, Int l. Symp. Signals, Sys., and Elect.., 29 Sept. 2 Oct. 1998, pp. 295 300. [8] S. Kaiser and K. Fazel, A Flexible Spread-Spectrum Multicarrier Multiple Access System for Multimedia Applications, IEEE PIMRC,vol. 1, Sept. 1997, pp. 100 04. [9] IST-2001-32620 MATRICE D1.3 [10] N. Maeda et al., Variable Spreading Factor-OFCDM with Two Dimensional Spreading that Prioritizes Time Domain Spreading for Forward Link Broadband Wireless Access, IEEE VTC 2003-Spring, vol. 1, pp. 127 32. BIOGRAPHIES KAN ZHENG (zkan@buptnet.edu.cn) received B.S and M.S. degrees from Beijing University of Posts and Telecommunications (BUPT), China, in 1996 and 2000, respectively, where he is currently working toward a Ph.D. degree. From April 2000 to October 2001 he was a system development engineer at the TD-SCDMA R&D Center of Siemens (Ltd), Beijing. His current research interests lie in the field of signal processing for digital communications, with emphasis on multicarrier modulation and MIMO systems. LIN HUANG (huanglin_bupt@163.com) received a B.S. degree in electronics engineering from BUPT, China, in 2002. Now she is a Master s student in the wireless communication and signal processing laboratory of BUPT. Her current research interest lies in channel estimation and detection in multicarrier systems. WENBO WANG (wbwang@bupt.edu.cn) is currently a professor and dean of the School of Telecommunication Engineering at BUPT. He received his B.S., M.S., and Ph.D. from BUPT in 1986, 1989, and 1992, respectively. His research interests include signal processing, mobile communications, and wireless ad hoc networks. YANG GUILIANG (yangguiliang@datangmobile.cn) is currently a senior engineer at Datang Mobile Communication Company where he heads the department of system standards. He received a B.S. degree in mechanics engineering from Tsinghua University in 1995, and an M.S. degree in electrical engineering from Chinese Academy of Telecommunications Technology (CATT) in 1998. His research interests include signal processing for digital communications and system design with the TD-SCDMA standard. 52 IEEE Communications Magazine January 2005