Abstract-Novel low-density signature (LDS) structure is proposed for transmission and detection of symbol-synchronous communication over memoryless Gaussian channel. Given as the processing gain, under this new arrangement, users' symbols are spread over chips but virtually only chips that contain nonzero-values. The spread symbol is then so uniquely interleaved as the sampled, at chip rate, received signal contains the contribution from only number of users, where denotes the total number of users in the system. Furthermore, a near-optimum chip-level iterative soft-in-soft-out (SISO) multiuser decoding (MUD), which is based on message passing algorithm (MPA) technique, is proposed to approximate optimum detection by efficiently exploiting the LDS structure. Given = as the system loading, our simulation suggested that the proposed system alongside the proposed detection technique, in AWGN channel, can achieve an overall performance that is close to single-user performance, even when the system has 200% loading, i.e., when = 2. Its robustness against near-far effect and its performance behavior that is very similar to optimum detection are demonstrated in this paper. In addition, the complexity required for detection is now exponential to instead of as in conventional code division multiple access (CDMA) structure employing optimum multiuser detector. Index Terms-Chip-level iterative SISO MUD, Gaussian channel, low-density signature, message passing algorithm, overloaded condition.

Novel Low-Density Signature (LDS) structure is proposed for synchronous Code Division Multiple Access (CDMA) systems for an uplink communication over AWGN channel. It arranges the system such that the interference pattern being seen by each user at each received sampled chip is different. Furthermore, new near-optimum chip-level iterative multiuser decoder is suggested to exploit the proposed structure. It is shown via computer simulations that, without forward error correction (FEC) coding, the proposed LDS structure could achieve near single-user performance with up to 200% loading condition. As the proposed iterative decoding converges relatively fast, the complexity is kept much more affordable than that of optimum multiuser detection (MUD) with conventional structure.

Abstract-The performance of various iterated soft-input softoutput (SISO) multiuser detection (MUD) techniques is evaluated and compared for the case of overloaded condition for symbolsynchronous uncoded Code-Division Multiple Access (CDMA) systems over Additive White Gaussian Noise (AWGN) channel. We show that, via computer simulation, while the Bit-Error Rate (BER) performance of the serial version of the Symbol-Level iterated (SLi) SISO MUD technique is inferior to its parallel counterpart, its convergence rate is the faster between the two. Yet, the Chip-Level iterated (CLi) SISO MUD performs the best. Moreover, the Max-Log-MAP version of the CLi SISO MUD complementing the recently proposed Low-Density Signature (LDS) structure is introduced and compared to its Log-MAP version. The sensitivity of the various CLi SISO MUD techniques to the existence of the short-cycles in the underlying factor graph of the LDS structure is also analyzed. Our simulations confirm that the Max-Log-MAP version performs better compared to its more complex Log-MAP version in higher system loading condition. We also show that a performance close to single-user bound can be achieved at higher Signal-to-Noise Ratio (SNR) regime for system loading of up to 250%.

Abstract-The brute-force MAP-based Chip-Level Iterated (CLi) Multiuser Detection techniques exploiting the newly proposed Low-Density Signature (LDS) structure have been shown to approach the performance of Single-User even at overloading condition of up to 200%. However, their complexity is exponential to the number of interfering users in a chip that can be impractical. In this letter, by using Gaussian-Forcing (GF) technique, we trade-off performance and complexity and it results in a dynamic Grouped GF (G-GF) CLi MUD technique. We show via computer simulation, at the system with 200% condition, the loss of approximately 0.3 dB compared to its brute-force counterpart is observed while reducing the complexity to more than half.Index Terms-Low-density signature, synchronous overloaded CDMA, chip-level iterated MUD, dynamic user grouping, Gaussian-forcing technique.

No abstract