Self-Reference Spatial Diversity Processing for Spread Spectrum Communications∗ Daniel Perez-Palomar,´ Montse Najar´ and Miguel Angel Lagunas
Total Page:16
File Type:pdf, Size:1020Kb
International Journal © Urban & Fischer Verlag of Electronics http://www.urbanfischer.de/journals/aeue and Communications Self-reference Spatial Diversity Processing for Spread Spectrum Communications∗ Daniel Perez-Palomar,´ Montse Najar´ and Miguel Angel Lagunas Abstract In this paper, the application of three blind (or self- interfering signals by linearly combining the signals re- reference) spatial diversity signal processing methods to Spread ceived at the antenna elements. The design of the beam- Spectrum (SS) communications is described. These methods do former requires some a priori knowledge about the de- not require any kind of side information beyond knowledge of sired signal in order to discriminate it from the inter- the signal structure, in contraposition to methods that depend ference. Depending on the character of this information, on training sequences. Each self-reference method is specifically designed for a par- classical beamforming techniques can be classified into ticular SS transmission scheme and uses a particular signal’s two types. The first one is Time Reference Beamforming structural information. All three methods, however, are derived (TRB), which needs a training sequence that in turn im- under the common goal of finding the optimum beamformer, in plies a loss of spectral efficiency and channel bandwidth the sense of maximum signal to interference-plus-noise ratio, in utilization but yields robust methods. The second type a blind manner. is Spatial Reference Beamforming (SRB), which requires The Code Reference Beamformer for Frequency Hopping sys- the knowledge of the direction of arrival (DOA) of the tems is an approach based on the knowledge of the hopping desired signal, the knowledge of the array geometry, and sequence or code structure of the signal. This method is the strict array calibration. best alternative for array signal processing on this particu- lar SS scheme. Two recently proposed beamforming methods, An alternative to requiring either a spatial or a time which are based on the knowledge of the redundancy struc- reference is Code Reference Beamforming, which can be ture of the desired signal, are described under the common included in the more general category of blind beamform- framework of efficiently using the inherent diversity – either ing. These blind (or self-reference) algorithms exhibit in frequency or time – of two signalling formats, namely, Fre- good behavior in terms of robustness and spectral/channel quency Diversity (FD) SS and DS-CDMA. The former uses efficiency. They are based on the minimization of an ob- frequency diversity, while the latter uses time diversity. The jective cost function which is defined based on certain in- diversity approach presented for FDSS seems to be the best herent structural properties of the desired signal; hence, no choice at the moment. Regarding DS-CDMA, for which many side information, such as a training sequence or the DOA beamforming algorithms have been developed in the literature, the self-reference beamforming is simply an alternative blind of the desired signal, is needed. This implies that self- method that shows very good performance. These diversity ap- reference spatial processing techniques can be applied to proaches were compared via simulations to the standard Time existing systems without the need of any modification of Reference Beamformer (TRB) and showed a performance simi- the emitter (e.g. the uplink channel of a cellular communi- lar to it (with 10% of the bits as a training sequence) even cation system [1]). An important feature of self-reference without the need of any side information whatsoever. beamforming algorithms is that they are not dependent on Keywords Spatial signal processing, Arrays, Beamforming, channel properties or array calibration, which makes them Blind methods, Self-reference, Diversity, Redundancy, Fre- robust methods. It is important to note, on the other hand, quency hopping, CDMA, FDSS, Spread Spectrum that each particular transmission scheme presents a dif- ferent signal structure and, therefore, requires a specific particular blind method. 1. Introduction Many blind algorithms have been devised exploiting different properties of the signals. They can be divided Array signal processing allows interference rejection ac- into two main groups: deterministic blind beamforming cording to the bearing angles or spatial signatures of the (using the known structure of the signal) and statisti- cal blind beamforming (using statistical properties of the source). All these techniques are also referred to as prop- Received August 2, 1999. Revised May 8, 2000. erty restoral algorithms, because they force the signal estimates to exhibit certain properties that the actual sig- D. Perez-Palomar,´ M. Najar,´ M. A. Lagunas, Universitat Politècnica nals are known to possess. For man-made signals, such de Catalunya (UPC), Department of Signal Theory and Communi- as those encountered in wireless communications, these cations, c/ Jordi Girona, 1-3, Mòdul D5, Campus Nord UPC, 08034 Barcelona, Spain. signal properties are often known with great accuracy, E-mail: [email protected] leading to robust algorithms. ∗ This work was partially supported by the European Comission Some of the most representative deterministic prop- under ACTS Project AC347 SUNBEAM; the Spanish Government erties, Constant Modulus (CM) [2, 3] and Finite Alpha- (CYCIT) TIC98-0412, TIC98-0703; and the Catalan Government bet (FA) (exploited by decision-directed methods [4]), are (CIRIT) 1998SGR 00081, 1999FI 00588. properties exhibited by many modulation schemes. Other Int. J. Electron. Commun. (AEU)¨ 54 (2000) No. 5, 267−276 1434-8411/2000/54/5-267 $15.00/0 268 D. Perez-Palomar,´ M. Najar,´ M. A. Lagunas: Self-reference Spatial Diversity Processing for Spread Spectrum Communications signals, known as gated-like signals, have the property of gorithm, a specific method for adaptive arrays using FH not being present on the channel during a certain time signals (assuming knowledge of the FH sequence) based (or at least being present with low power); this allows on the spectral characteristics of the received signals. Fi- anticipative processing to estimate the parameters of the nally, Eken [22] devised a modified sidelobe canceller for scenario free of the desired signal [5–7]. A particular ex- FH signals that needed a priori knowledge of the DOA ample of this type of signal is Frequency Hopping, where of the desired signal. The main drawback of the previous knowledge of the frequency hopping sequence can be approaches is the drop in SINR at hop instants. The rea- used for the anticipative processing [8–10] (see Sect. 2). son for this behavior, as was pointed out in [18, 20], is Another property, the property of diversity, which refers to that changes in signal frequency due to FH modulation the existence of signal replicas along the time, frequency, are seen by the algorithm as changes in the DOA’s, result- code or space axis (usually to combat fading or jamming), ing in discontinuities of the adaptive algorithms’ perform- provides the signal with a redundancy that allows opti- ance. mum beamforming using the replicas as embedded self- The anticipative processing algorithm proposed in [8– references [11–13] (see Sect. 3). 10] and denoted Code Reference Beamformer takes ad- Regarding statistical properties, the most important vantage of the knowledge of the FH sequence at the re- ones are probability density function fitting and statistical ceiver, as in [20], and requires neither temporal nor spa- independence of the sources [14], higher order statistics tial references. The method relies on the possibility of (HOS) [15, 16], and cyclostationary properties [17]. It is the measurement and estimation of the interference-plus- important to note that blind statistical methods usually noise covariance matrix. This property has already been present bad convergence properties, yielding algorithms used by several authors in non FH contexts [5–7]. that are not very reliable. The proposed system, depicted in Fig. 1, is composed Blind beamforming techniques lend themselves to of two parallel processors; both processors begin by de- block or direct implementation (usually based on a gen- hopping the received signal. The first one, called the antic- eralized eigenvector decomposition) and to adaptive al- ipative processor, is used to obtain the interference-plus- gorithms (with a reduced computational load). Clearly, noise covariance matrix, Rin, before the desired signal is adaptive methods are especially suitable for tracking in present. The second one, the on-line processor, performs non-stationary environments. the actual beamforming. In this paper, three blind deterministic beamforming The beamforming part of the on-line processor is made methods, based on the knowledge of the structure of the up of two different stages, corresponding to the two terms = κ −1 κ signal, are described for Spread Spectrum (SS) commu- of the optimum beamvector wopt Rin ad,where is nication systems. Each one is specifically designed for a scalar factor, and ad is the steering vector of the desired a particular SS signalling scheme, but derived from a com- signal. The first stage preprocesses the dehopped received mon framework of diversity aimed at the maximization of snapshot, xol (t), using the inverse of the interference-plus- ( ) = −1 ( ) the Signal to Interference-plus-Noise Ratio (SINR). The noise covariance matrix to obtain x t Rin xol t .The performance of each method is verified via simulations. second stage is devoted to properly weighting the prepro- In Sect. 2, we describe the Code Reference Beamformer cessed signal vector x(t) with the desired steering vec- based on anticipative processing for slow Frequency Hop- tor ad.Sincead is assumed to be unknown, a blind es- ping SS (FHSS). Sect. 3 introduces beamforming tech- niques based on a general framework of diversity prop- erties. Then, Sects. 3.1 and 3.2 present specific self- reference algorithms based on frequency diversity, for both fast FHSS and Frequency Diversity SS (FDSS), and ANTICIPATIVE PROCESSOR based on time diversity, for Direct Sequence SS (DSSS), Q IF 5 -1 respectively.