A Novel Transmitter-Based Selective-Precoding Technique for DS/CDMA systems C. Masouros and E. Alsusa School of Electrical & Electronic Engineering The University of Manchester Manchester, UK email: Chris.Masouros@postgrad.manchester.ac.uk, E.Alsusa@manchester.ac.uk Abstract— In this paper a new transmitter precoding technique is presented that outperforms conventional precoding by making use of a portion of the interference between the users in a CDMA system downlink. The proposed technique selectively pre- decorrelates users that are experiencing destructive interference while allowing interference to other users when it is expected to contribute to their signal. The existence and exploitation of constructive interference effectively spreads the signal constellation and enhances the SNR at the receiver. The SNR improvement happens by making use of energy that is already in the system so the performance improvement is attained with no additional power-per-user investment. This however comes with the trade-off of some extra processing at the transmitter for the measurement of the expected interference. The proposed technique applies to the downlink of cellular CDMA systems employing PSK modulation. Theoretical analysis supported by comparative simulations of this and other precoding methods are presented and discussed. Keywords—Adaptive signal processing, code division multi- access, interference multiuser channels, suppression I. INTRODUCTION The capacity of a code-division multiple-access (CDMA) system is limited by multiple access interference (MAI) from other users as well as intersymbol interference (ISI) between the symbols of the user of interest due to the frequency selectivity of the transmission medium. Multiuser Detection (MUD) techniques are traditionally used to mitigate these effects and improve the performance and capacity of CDMA systems. In order for complexity reduction at the Mobile Units (MUs) of a CDMA communication system, the current trend is towards transferring the computational burden to the Base Station (BS) by use of precoding techniques for the downlink transmission. Various methods have been proposed towards this end following an initial idea introduced in [1] applicable to general pulse amplitude modulation (PAM) systems. In [2] the authors propose transferring the Rake processing to the BS which yields the Pre-Rake technique. This technique’s main advantage is that matched filtering (MF) is applied for detection which alleviates the need for channel estimation and removes a significant burden from the MU. However, this is primarily a single user detection (SUD) technique so performance is poor in a multiuser scenario. The authors in [3] propose a system similar to the conventional receiver-based decorellator-detector where the decorrelation procedure happens at the BS prior to transmission. The orthogonalization of the users comes with an increase in transmitted energy which calls for either scaling of the signal to be transmitted or applying constrained optimization in order for the power limitation to be maintained. Both these techniques are investigated in [3]. An improvement is attained by applying the decorrelating procedure in [4]. This optimization leads to the use of a decorrelation scheme that also employs Pre-Rake processing. This method offers both the benefits of pre- decorrelation as well as the advantages of Pre-Rake over the Rake technique as explained in [2]. The decorrelating methods introduced in [3,4] are blockwise, which results in a high computational complexity. In aim of mitigation of this defect, the authors in [5] propose a zero forcing bitwise decorrelating technique that has comparable performance when the ISI is limited by use of guard intervals. When severe multipath is introduced, however, the performance rapidly deteriorates. An improved bitwise technique using minimum mean square error (MMSE) optimization is presented in [6] that achieves performance comparable to [3,4] while maintaining reasonable computational complexity. In this paper an improvement of the Transmitter Precoding (TP) and Joint Transmission (JT) techniques in [3,4] respectively is suggested which takes advantage of the constructive interference concept applicable in PSK modulation that will be analyzed below. The system analysis follows the one presented in [3]. In contrast to the TP and JT techniques where the users are fully orthogonalized, we propose a partial orthogonalization by means of pre-decorrelation, orientated to the users that are expected to suffer from destructive MAI. This reduces complexity and yields an increased signal to noise ratio (SNR) at the receiver. It should be noted that this technique is applicable to both the TP and JT methods irrespective of whether scaling or constrained optimization are used, assuming phase shift keying (PSK) modulation where the constructive MAI concept stands. In this paper unconstrained optimization followed by scaling is investigated. II. CONSTRUCTIVE MAI DERIVATION FOR PSK MODULATION A. Downlink Signal Model and Constructive MAI Definition Consider the downlink transmission in a discrete-time synchronous frequency selective CDMA system of K users, where the channels’ path delays are assumed to be an integer 1-4244-0353-7/07/$25.00 ©2007 IEEE This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the ICC 2007 proceedings.