To Spread or Not to Spread in a Wireless Data Network John Leung and Stephen Hanly ARC Special Research Centre for Ultra-Broadband Information Networks The University of Melbourne Australia Abstract – Performance comparisons are provided for different receivers in a time-slotted mobile network designed to carry delay- insensitive traffic. We show that even with one user per timeslot, capacity gains are achieved from some degree of spreading and multi- user detection against no-spreading and also against spreading plus matched filtering. Our results suggest that frequency partitioning between cells may be superior still, for the scenario considered in the present paper. However, our results also suggest that with optimal scheduling of users (not included in our model) full frequency reuse and no spreading may be optimal, but this issue requires further study. I. INTRODUCTION In CDMA networks, all the connected users transmit and receive at the same time, with the signature code differentiating between the signals. This approach is good for delay-sensitive traffic such as voice, but it may not be optimal for elastic traffic because useful power may be assigned to transmit to users with bad channel conditions or for users who are far away from the base-station. It may be advantageous to delay transmissions for users with bad channel or large attenuations, and assign valuable power resources to users with good channel and transmit data at higher rate. This type of scheduling is known to be information theoretically optimal [1][2]. We believe that future high data rate mobile systems, designed to carry delay- insensitive traffic, will be based on this approach. We note that Qualcomm’s CDMA-HDR [3] technology is of this type. In the present work, we consider the forward link of a time-division multiple access mobile system with frequency reuse factor of one as in Qualcomm’s HDR. For simplicity, we do not consider scheduling based on channel quality in the present paper, but rather take a simple round robin (TDMA) approach. In such systems, the throughput performance is limited by the outer-cell interference. As a mobile moves away from its base-station (and closer to the sources of interference), its channel capacity degrades rapidly. We find in our simulations that by spreading the transmitted signal, we are able to reduce the effect of interference, and that the capacity degradation rate across distance is lower. Three types of receivers for the spread spectrum signal are considered in the present paper, they are the single-user matched filter, the multi-user MMSE receiver, and the ideal system running orthogonal spreading codes in adjacent cells. Of the receivers we encountered, we observe that the MMSE receiver (that is, using multi-user detection) works very efficiently and results in higher capacity over the system with no spreading for mobiles far away from the base- station. The exception is that when the mobile is near the base-station, spreading is sub-optimal. However, we find that the optimal amount of spreading for the MMSE is enough to allow orthogonality between neighbouring cells. Indeed, we find that the capacity with orthogonal spreading codes is the highest of all. Although spreading-code orthogonality cannot be achieved in practice, this result suggests that full frequency reuse is suboptimal and bandwidth should be partitioned between cells (see also [4] for information theoretic results which support this), but this issue requires further study and more detailed models. Moreover, this conclusion may be false when scheduling is included in the model, and we note that scheduling is included in Qualcomm HDR. In the next section, we describe our network and simulation model in detail. In Section III, we focus on the receivers models and we discuss the simulation results and findings in Section IV. II. NETWORK AND SIMULATION MODEL Let us consider the forward link of a synchronous DS- CDMA connection between a base-station and a mobile. The information symbols are spread by a signature sequence of length N at the transmitter, that is, the spreading factor is N. These signature codes are chosen randomly. The received signal power at the mobile is the transmitted power attenuated with a path-loss of d α , where d is the distance between the transmitter and the receiver, and α is the path- loss exponent. We assume the fluctuations in the received power can be modelled by log-normal shadowing with a standard deviation of S dB. There are six cells surrounding the cell in consideration in the typical hexagonal structure. We assume that the base- station in a cell transmits, in a round-robin fashion to mobiles in its cell. There is no power control and the transmission powers of the base-stations are assumed to be the same. The signal power originating from these surrounding base-stations becomes interference to the user in the centre cell. We wish to compare the performance of different receiver structures over a range of carrier-to-interference ratios, therefore we place the centre cell user at various distances 599 0-7803-7206-9/01/$17.00 © 2001 IEEE