Neural Networks for Improving Delay Lock Loops Performances in DS-CDMA Systems

In this work a novel approach for asynchronous CDMA receivers, based on artificial neural networks, is presented. Starting from code tracking systems which can be found in literature, the aim of this study is directed at finding an alternative solution to non-coherent delay lock loop, with the purpose to improve performances. Introduction Direct-sequence code-division multiple access (DS-CDMA) is a promising technology for both wireless personal communication systems (PCS) and navigation systems (GPS, Galileo). The performances of these systems are mostly influenced by the accuracy of the synchronization modules. In fact, loss of fine tracking in PCS case means loss of information for a while and a less precise position estimation in the case of navigation systems. Furthermore, the complete loss of synchronization implies the connection loss in both cases. Thus, to achieve a good service, the synchronization (coarse synchronization and tracking) modules must be designed in order to have the highest performances. The synchronization phase between the incoming signal and the locally generated replica is divided in two steps: coarse alignment (first acquisition) and fine synchronization (tracking) [1]. The first task is achieved when there is a small delay offset between signature of the incoming signal and the code replica at the receiver. The second task performs and continuously maintain the best possible alignment between incoming coded signal and the local code replica. The literature proposes several methodologies to have a tracking module achieving good performances, i.e. narrow correlators [2]. In this article we propose to use classifiers based on artificial neural networks. In particular, three different types of classifiers have been developed and have been tested in a receiver model. The most interesting of these systems uses a synaptic weights variation technique which makes it adaptable to the channel in real-time. For the evaluation of the performances, many signal synchronization simulations have been executed and, by the collected data, the behavior of the implemented systems has been compared to the classical DLL. The tests have pointed out a clear improvement of the DLL performances concerning synchronization error frequency and estimation precision. The classifier which has revealed the most effective behavior is the adaptive one, confirming the starting intuition. The tests have been performed simulating a navigation system but the solution is general and can be used in other contexts dealing with CDMA, for example: personal communication systems. The choice of a navigation system has been done because these systems are very sensible to tracking errors. The paper is organized as follows: first we describe the signal model; then, after introducing the NN, we present the receiver structure. Finally we show some simulation results and draw some conclusions. System The Galileo positioning system, like GPS, will be composed by a constellation of MEO satellites to broadcast the information using the CDMA technique as a multi-user transmission technique. This system will offer many kinds of services, some of those will be public (Open Access Service, Controlled Access Service) and some private (Governmental Access Service). At this time, the system definition phase is not yet finished, therefore, the chosen Galileo system data is one of the mainly possible scenario proposed by project industrial partners. This scenario is named “European standalone scenario (termed “reference scenario” before WRC2000) that comprises the frequency bands E5, E6, E2, E1 and C. In this scenario, limited in this paper to band E6, the Galileo signal is composed by one carrier frequency modulated by spread spectrum codes and a QPSK modulated data signal. The pseudonoise codes are gold sequences, like the C/A code in the GPS system, with chip rate of 1.023 Mchip/s and length of 1023. The data message has a bit-rate of 125 bps. The modulation is a QPSK, where the in-phase signal carries the data message and the quadrature signal is used as pilot signal without actual signal message. From the receiver point of view, the signal sensed by the antenna can be modeled as equation (1), see ref. [3]