An OVSF code tree partition policy for WCDMA systems based on the multi-code approach

Abstract

Orthogonal variable spreading factor (OVSF) codes for wide-band CDMA (WCDMA) systems have been proposed in the third generation (3G) mobile communications' standard to support services of variable rates. To assign OVSF codes, there are single-code and multi-code approaches. In this paper, the multi-code placement and replacement issue is addressed. We propose a tree partition policy for managing the OVSF code tree to reduce code fragments and the number of code reassignments. Besides, we adjust the multi-code rate according to a unit-based method, which is capable of decreasing the number of code fragments because fewer low-rate codes are utilized. Through extensive simulations, it turns out that the tree partition policy and the unit-based multi-code method jointly perform better than the left-most and crowded-first schemes. I. INTRODUCTION The universal mobile telecommunication system (UMTS) employs WCDMA to enable high-rate data transmission and variable-bit-rate services for different users using OVSF codes as channelization codes for connections. Since a different spreading factor (SF) can provide a different rate service, variety of requested rates can be flexibly fulfilled. OVSF codes may be represented as a binary code tree (1), in which the spreading factor of the downstream level is a power of two of the current level, while the relationship of code rates for these two levels is reverse. Any two codes are mutually orthogonal if and only if one code is not the ancestor or descendant code of the other. Avoiding simultaneous use of ancestor and descendant codes, the orthogonality can be maintained. In the literature, there are two assignment approaches for OVSF codes: single-code (2), (3), (4) and multi-code (5), (6), (7). Single-code assignment allocates only one code to a user. It may cause rate wastage and a code fragmenta- tion problem. Multi-code assignment can support multi-rate services using multiple codes with multiple rake combiners. It is able to reduce the code blocking probability because a small-SF OVSF code (a high-rate code) can be used to replace some large-SF OVSF codes (low-rate codes) to avoid code fragmentation. In (5), a multi-code assignment method is proposed. Among possible candidates, this method selects the best multi-code with less codes but more small-SF codes. Its main drawback is that it relies on the complicated computation. Although it indeed avoids using too many large-SF codes, it perhaps blocks small-SF ones. Moreover, the code placement and replacement are not taken into consideration. To prevent redundant computation, the scheme proposed in (6) uses the technique of dynamic programming scheme to build up a multi-code table in advance so that less time in selecting the best multi-code is required. In (7), the internal and external fragmentation problems are defined and solved by using multi- codes. To deal with the code placement and replacement, crowded-first and left-most schemes are proposed in (4). The disadvantages of these two schemes are extra storage and code reassignments required. What we have learned from the aforementioned work is that multi-codes have many merits but may cause side effects. When multi-codes are released, they may split resources, resulting in more fragments which may subsequently result in code blocking of high-rate codes. Moreover, the more code fragments we have, the more code reassignments are required. To eliminate the above-mentioned side effects, we propose an OVSF code tree partition policy called high rate right most (HRRM) and the unit-based multi-code (UBMC) method to solve the problem of code fragmentation and to reduce a large amount of code reassignments in this paper. Compared with the existing placement and replacement schemes, i.e., crowded-first and left-most, the advantages of the proposed schemes are demonstrated in the later section. The rest of the paper is organized as follows. The system model of multi-code OVSF is introduced in Section II. In Section III, the proposed schemes, i.e., HRRM and UBMC, are described in detail. After that, the simulation model and numerical results are presented in Section IV. Finally, Section V concludes the paper.