Performance Evaluation of Approximately MAI-Free Multiaccess OFDM Transceiver Shang-Ho Tsai 1 , Yuan-Pei Lin 2 and C.-C. Jay Kuo 1 Department of Electrical Engineering, University of Southern California, CA, U.S.A. 1 Department of Electrical and Control Engineering, National Chiao Tung University, Hsinchu, Taiwan 2 Abstract — The performance of an approximately MAI-free OFDM transceiver, called the repetitively coded multicar- rier CDMA (RCMC-CDMA), is evaluated and compared with MC-CDMA in this work. In particular, its advan- tages on the MAI reduction capability and the low im- plementational complexity are presented. We first demon- strate that RCMC-CDMA has the performance comparable to MC-CDMA using a linear MUD (multiuser detection) MMSE (minimum mean squared errors) detector, while it outperforms MC-CDMA with a MUD decorrelating detec- tor. Then, the complexity of RCMC-CDMA is shown to be lower than that of MUD-based MC-CDMA. Keywords — Multiaccess OFDM, MAI-free, RCMC- CDMA, MC-CDMA, multiuser detection. I. Introduction Multicarrier modulation code division multiple access (MC-CDMA) systems have been proposed for multiaccess transmission recently [1]. As compared with CDMA sys- tems, MC-CDMA can effectively combat the inter sym- bol interference (ISI) caused by frequency-selective fading. However, the capacity of MC-CDMA system is limited by multiaccess interference (MAI). To suppress the MAI ef- fect, multiuser detection (MUD) is usually used at the re- ceiver end [2]. To perform MUD, the receiver has to know the channel statistic information (CSI) and channel esti- mation is needed. As the number of active users grows, ac- curate channel estimation for MC-CDMA systems becomes more difficult due to MAI [3]. Moreover, the complexity of the MC-CDMA system increases as a result of the use of MAI suppression techniques. An approximately MAI-free multiaccess OFDM transceiver, called the repetitively coded multicarrier CDMA (RCMC-CDMA), was proposed in [4]. The approximately MAI-free property holds when the num- ber of subchannels, N , is sufficiently large. Unlike the conventional MC-CDMA system, there is no need to suppress MAI in the RCMC-CDMA system. Conse- quently, the complexity is lower and channel estimation can be done more easily. Moreover, the RCMC-CDMA system promises the capacity increase as the transmit power increases. This result stands in contrary to the conventional MC-CDMA system, in which increasing the transmit power for one user will also increase the MAI for other users. Under the assumption that only the receiver knows the CSI, the performance of RCMC-CDMA is evaluated and † Author for all correspondence: shanghot@usc.edu (Shang-Ho Tsai) compared with that of conventional MC-CDMA with MUD in this work. In particular, the following two advantages of the proposed RCMC-CDMA scheme will be demonstrated. First, the RCMC-CDMA outperforms the MC-CDMA with a decorrelating detector and has the performance compa- rable to MC-CDMA with a linear MMSE (LMMSE) de- tector. Second, when the number of active users grows or multiple antennas are used, the complexity superiority of RCMC-CDMA over MUD-based MC-CDMA becomes more pronounced. II. RCMC-CDMA System Description and Properties A. System Description Fig. 1 shows the block diagram of the RCMC-CDMA system with T users. The transmit side (TX) contains four stages. At the first stage, the input of the ith user is an N × 1 vector x i , consisting of N modulation symbols, e.g. PSK or QAM. Each symbol in x i is repeated M times and then scaled to form an NM × 1 vector y i . Let x i [k] denote the kth symbol of x i ,0 ≤ k ≤ N − 1, and y i [l] denote the lth symbol of y i ,0 ≤ l ≤ NM − 1. The relation between x i [k] and y i [l] is given by y i [m + kM ]= 1 √ M x i [k], 0 ≤ m ≤ M − 1, (1) where 1 √ M is included to preserve the power before and af- ter the symbol repetition. At the second stage, each vector y i is passed through an NM × NM diagonal matrix W i with its diagonal elements drawn by the M × M unitary matrix D, i.e. D † D = M I, where D † is the conjugate- transpose of D and I is the M × M identity matrix. Let d i denote the ith column of D. Then, W i is obtained by repeating d i by N times along the diagonal, i.e. W i = diag(d t i d t i ··· d t i ), where d t i denotes the transpose of d i and diag(d t ) is the function which puts the elements of d t on the diagonal. When M is equal to the power of 2, an example for D is the Hadamard matrix, whose columns form the M Hadamard- Walsh codes [2]. For instance, letting M = 2, the second column of the 2×2 Hadamard matrix is (+1 −1) t . If N = 2, we have W 2 = diag(+1 − 1 +1 − 1). Since D † D = M I, the diagonal elements of W i satisfy the following property: M−1  m=0 w i [m + kM ]w ∗ j [m + kM ]=  M, i = j 0, i = j , (2) 0-7803-8521-7/04/$20.00 (C) 2004 IEEE