PARAMETER OPTIMIZATION OF THE ADAPTIVE MVDR QR-BASED BEAMFORMER FOR JAMMING AND MULTIPATH SUPRESSION IN GPS/GLONASS RECEIVERS

Transcription

1 PARAMETER OPTIMIZATION OF THE ADAPTIVE MVDR QR-BASED BEAMFORMER FOR JAMMING AND MULTIPATH SUPRESSION IN GPS/GLONASS RECEIVERS V. Behar 1, Ch. Kabakchiev 2, G. Gaydadjiev 3, G.Kuzanov 4, P. Ganchosov 5 1 Institute for Parallel Processing, Bulgarian Acadey of Sciences Acad. G. Bonchev St. 25-A, 1113 Sofia, Bulgaria E-ail: behar@bas.bg 2 Departent of Software Technologies, Sofia University St. Cleent of Ohrid J. Boucher Blvd. 5, 1164 Sofia, Bulgaria E-ail: ckabakchiev@fi.uni-sofia.bg 3, 4, 5 Departent of Coputer Engineering, Delft University of Technology Mekelweg 4, 2628 CD Delft, the Netherlands E-ail: g.n.gaydadjiev@ewi.tudelft.nl, G.Kuzanov@ewi.tudelft.nl, P.N.Ganchosov@tudelft.nl Abstract Key words: Antenna array, GNSS, interference itigation, STAP, Acquisition This paper analyzes the influence of the space-tie signal processing technique on the perforance of GPS signal acquisition in conditions of strong broadband interference (jaing) and ultipath. The space-tie processing ethod used for effective itigation of GPS interference before processing by a conventional acquisition algorith is the Miniu Variance Distortionless Response beaforing ethod with QR factorization (MVDR QR). The nuerical results obtained by siulations deonstrate that any factors such as array configuration, nuber of array eleents, and sapling rate of the incoing data have a considerable effect on the effectiveness of both beaforing and acquisition algoriths. Introduction The Global Positioning Syste (GPS) and the Global Navigation Satellite Syste (GLONASS) have been designed to provide precision location estiates for various ilitary and civilian applications. Each of the satellites transits digitally coded data, and GPS/GLONASS receivers deodulate and process these signals fro four or ore satellites siultaneously in order to generate three tie-difference-of arrival estiates, allowing the user to easure the range to three satellites, and, as a result, to deterine own position. Since a direct sequence spread spectru (DSSS) signal is used in transission, relatively low powers can be transitted by the satellites and still to have adequate signal-to-noise-ratio (SNR) for accurate position estiation. In fact, these signals have the signal-to-noise ratio (SNR) of between -20 and -30 db. Therefore, the key to achieve the precise position estiation perforance is the processing of very weak DSSS signals fro satellites that contain coarse acquisition (C/A) and precision (P) digitally coded data. Whatever, GPS/GLONASS signals have soe degree of jaing protection built into the signal structure itself the weak signal strength of the received signal akes it easy for an intentional or unintentional radio frequency interference (RFI) to overcoe jaing protection of the DSSS signal [1]. If a strong broadband jaing source is nearby, the receiver noise ay rise to the level where the post correlation SNR of the satellite signals is below the threshold value required for tracking. Multipath is the other liiting factor in any GPS/GLONASS applications that affects both the pseudo range and carrier phase estiates. Signal ultipath is the phenoenon where a satellite signal arrives at the receiver antenna after being reflected fro different surfaces or buildings. Various approaches can be used to itigate GPS interference before signal processing in a GPS/GLONASS receiver [2]. One of the is to use different beaforing techniques for broadband nulling having in ind that the satellite signals and interfering signals usually originate fro different spatial locations. The conventional (delay-and-su) beaforer is the siplest, with all its weights of equal agnitudes and the phases that are selected to steer the array in particular direction [3]. This beaforer has unity response in each look direction, that is, the ean output power of the beaforer in the look direction is the sae as the source power. In conditions of no directional interferences, this beaforer provides axiu SNR but it is not effective in the presence of directional jaing signals, intentional or unintentional. The others beaforers such as a Miniu Variance Distortionless Response (MVDR) beaforer can overcoe this proble by suppressing interfering signals fro off-axis directions [4]. To suppress jaing signals, this beaforer does not require the a priori inforation about the. It requires only the inforation for the direction-of-arrival of GPS signals. In this paper, the capability of the two beaforers, conventional and MVDR, to itigate GPS interference in order to iprove the acquisition of GPS signals in conditions of jaing and ultipath is studied and copared for different array configurations. The effectiveness of these beaforers is evaluated in ters of the following quality paraeters the iproveent in signal-to-noise-plus-interference ratio (SINR) - at the 325

2 beaforer output and the probability of detection - at the output of the acquisition algorith. The influence of different factors, naely array configuration, nuber of array eleents and sapling rate of the incoing data, on the overall beaforing + acquisition algorith is evaluated using the Monte Carlo approach. 2. Software-based GPS receiver The structure of a software-based GPS receiver is shown in Fig.1. It processes signals received fro the GPS satellites, which are in view, and then uses the inforation extracted to deterine and display the user position, velocity, and tie [1]. The GPS receiver is a ulti-channel device, where one channel processes the incoing signal fro one satellite. The first blocks of a receiver are an antenna and RF front-end devices, which are usually ipleented in hardware. Antenna (L1) RF Front-end hardware Acquisition & Tracking Software-based Processing Navigation Algorith Position & Velocity Coputation Fig.1: Architecture of a software-based GPS receiver In the RF front-end section, the GPS signal is down converted fro the RF frequency to an interediate frequency (IF) by one of ultiple stages. The down converters reduce the GPS carrier frequency fro GHz to a couple MHz. The last stage of the RF front-end section is an Analog-to Digital Converter, where the IF signals are sapled at a suitable sapling frequency and digitized. In the next blocks of a GPS receiver, the satellite signal is processed in ultiple (until 8 to 12) parallel channels, where each channel acquires and dynaically tracks one visible satellite. A iniu four channels tracking the signals of four satellites would be required to deterine three position coordinates and the receiver clock offset. In a conventional hardware-based receiver, the last three blocks in Fig.1 are ipleented in an IC chip, and the algoriths built inside the chip are not available for the user. In a software-based GPS receiver, however, these blocks are ipleented in software and hence the user has a free access to the algoriths and can exercise control over the. This is the difference between the software-based GPS receiver and a conventional hardware-based GPS receiver. 3. Signal odel The GPS eploys spread-spectru (SS) techniques for transission at two frequency bands L1 ( MHz) and L2 ( MHz). The signal transitted at L1 contains coarse acquisition (C/A) code, precision (P) code and navigation data while the signal transitted at L2 has P-code only. In this paper we liit discussion within the L1 band and C/A code only because the frequency band L2 and P-code serves the ilitary purposes only and the civilian counity does not have a free access to the encryption codes of P-code. The C/A code has a rate of Mchips/s with a period of 1023 chips (1s). The C/A code based on Gold codes is a sequence of zeros and ones, and it is unique for every satellite. The ain property of the C/A code is that it has the best cross-correlation characteristic. The cross-correlation between two codes is very uch lower than the auto-correlation of each of the codes. In GPS applications, when the receiver is on the ground, the input SNR value depends on the RF-bandwidth of the receiver front-end and is typically around -20dB (2.046 MHz C/A code bandwidth). The ideal received signal is coposed of the GPS signal, theral noise and a variety of interference. The coplex saples of the received signal at tie instant k can be atheatically described as: L x( = ac s( + bl jl ( + q p p ( + n( l = 1 where x( is the M x 1 data vector, s( is the desired GPS signal, j l ( is the lth broadband interference, p ( is the pth ultipath, a c, b l and q p are M x 1 antenna array response vectors of the GPS signal, the lth broadband interference and the pth ultipath, respectively, n( is the noise vector and L and P is the nuber of broadband interference sources and ultipath signals, respectively. The GPS signal is given by: s( = P c( d( (2) s where P s is the signal power, c( is the C/A coded signal of length (20 x 1023), separate for each satellite, d( is the GPS data bit which reains constant over the length of one cycle of the C/A code. The jaing signal j( occupies the entire RF bandwidth of the GPS receiver and usually has a flat power spectru over the passband of the GPS receiver and thus it can be represented atheatically as bandliited white additive Gaussian noise (AWGN). The ability of the GPS receiver to perfor critical functions is called as its anti-jaing capability, which is defined as the ratio of interference power to GPS signal power, above which the receiver function P p= 1 (1) 326

3 cannot be perfored. Multipath is another significant source of errors for high accuracy positioning in GPS applications. Multipath is the phenoenon whereby a signal is reflected fro ultiple objects in the environent and arrives at the receiver via ultiple paths. According to [5], the ultipath signal p ( in (2) can be odeled as: = α s( k δ )exp( i f k + iφ ) (3) p ( p p p p where α p is the ultipath aplitude, δ p is the delay, f p is the difference between Doppler frequencies of the GPS signal and ultipath. This frequency is very sall since the distance and the relative speed between the obile and reflecting surface are very sall with respect to the satellite-obile ones. According to [5], the worst situation for position estiation is when the phase Φ p is 0 or 180. In that case the GPS receiver can not distinguish between a direct and reflected signal, and as a result, the receiver tracking loops align the locally generated code and carrier to the coposite signal instead of the direct signal causing the ultipath error. 4. Antenna array geoetry Antenna arrays are coposed of any antenna eleents working jointly to establish a unique radiation pattern in the desire direction. The antenna eleents are put together in a known geoetry, which is usually unifor - Unifor Linear Arrays (ULA), Unifor Rectangular Arrays (URA) or Unifor Circular Arrays (UCA) [6, 7]. Since the ULA bea pattern can be controlled only in one diension (aziuth), so in GPS applications, a URA configuration or a UCA configuration with the eleents extended in two diensions ust be used in order to control the bea pattern in two diensions (aziuth and elevation). URA configuration In a URA, all eleents are extended in the x-y plane. There are M X eleents in the x-direction and M Y eleents in the y-direction creating an array of (M X x M Y ) eleents. All eleents are uniforly spaced d apart in both directions. Such a rectangular array can be viewed as M Y unifor linear arrays of M X eleents or M X unifor linear arrays of M Y eleents. Usually, the first antenna eleent is considered as the origin of Cartesian coordinates as shown in Fig.2. The direction of a signal arriving fro aziuth φ and elevation θ can be described with a unit vector e in Cartesian coordinates as: e ( ϕ, θ ) ( e, e, e ) = (cos θ sin ϕ, cos θ cos ϕ,sin θ ) (4) = x y z The vector in the direction of the (i, eleent can be described in Cartesian coordinates as: r ( i, k ) = ( d( i 1), d( k 1),0) (5) In (5), i and k denote the eleent position along the y- and the x-axis, respectively. The sequential eleent nuber (i, is defined as: ( i, = ( i 1) M X + k, i =1 MY, k = 1 M X (6) If the first eleent in the rectangular array is a reference eleent, the path-length difference d (i, for a signal incident at eleent (i, can be defined as the projection of the vector r (i, on the signal direction vector e: T d = e r = cosθ d [sin ϕ( i 1) + cosϕ( k 1)] (7) ( i, k ) ( i, Therefore, the URA response vector a c in (1) takes the for: ac ( ϕ, θ) = [1, exp( j d 2 ), K,exp( j d ( i, k ) ), K,exp( j d M )], where M=M X x M Y (8) UCA configuration In a UCA, all eleents are arranged along the ring of radius r (Fig.3). The ring contains M array eleents. Since these eleents are uniforly spaced within the ring, so they have an intereleent angular spacing φ=/m and a linear intereleent spacing d=2rπ/m. It is usually assued that the first antenna eleent is located at the y-axis, and the ring center is the origin of Cartesian coordinates. The vector in the direction of the th array eleent can be written in Cartesian coordinates as: r = ( r sinϕ, r cosϕ,0), where ϕ = ( 1) / M (9) The unit vector e(φ,θ) in the direction of a signal source is given by (4). If the ring center serves as a reference point, the propagation path-length difference d for a signal incident at eleent can be defined as the projection of the vector r on the direction vector e: T d = e r = d cosθ (sin ϕ sin ϕ + cosϕ cosϕ ) = d cosθ cos( ϕ ϕ ) (10) Therefore, the UCA response vector a c in (1) takes the for: ac ( ϕ, θ ) = [exp( j d1),exp( j d 2 ), K,exp( j d ), K,exp( j d M )] (11) where d is calculated by (10) for =1,2, M. 327

4 Z to a signal source Z to a signal source e r e θ φ Y r1 φ φ 1 θ 1 Y X X Fig.2: URA configuration Fig.3 UCA configuration 5. GPS signal processing The signal processing in a conventional GPS receiver includes acquisition and tracking of the GPS signal. In this paper we liit our discussion with the acquisition stage only. In our case, the GPS signal processing includes two stages (Fig.4). At the first stage the digital beaforing is perfored in order to itigate broadband interference and ultipath. At the second stage the standard acquisition algorith is perfored. Antenna array (URA or UCA) RF Front-end (IF signal) Digital Acquisition Beaforing (BF) Analysis Analysis Fig.4: The flow-chart of the evaluation process Beaforing stage The digital beaforer increases the gain in the direction of arrival of the desired signal, and decreases the gain in all other directions (interference). The output of an antenna array of M eleents is fored as: H y( = W x(, where k=1 N (12) where (.) H denotes conjugate transpose. Conventional Beaforer: In a conventional beaforer, the coplex vector of weights W is equal to the array response vector a c, which is defined by array configuration [8]: W = (13) conv a c MVDR Beaforer: The objective of the adaptive beaforing is to preserve the gain in the direction of arrival of the desired signal and itigate broadband interference incoing fro the other directions. The weight vector W is chosen to axiize the signal-to-interference-plus-noise ratio at the antenna output [9, 10]: 2 H σ S W ac SINR = H W K W where K j+n is the interference + noise covariance atrix of size (M x M), and σ 2 S is the signal power. The easy solution can be found by aintaining the distortionless response toward signal and iniizing the power at the filter output. This criterion of optiization is forulated as: H in W K W to subject W H a = 1 (15) W j+ + n The solution of (15) is known as the iniu variance distortionless response beaforer (MVDR): 1 W K j+ nac MVDR = (16) H a K 1 a In practical applications, K j+n, is unavailable. For that reason the saple covariance atrix K ) is used instead of it. The saple covariance atrix is estiated as: ) N 1 K = x( n) x H ( n) (17) N n= 1 Many practical applications of MVDR-beaforers require online calculation of the weights according to (16), and it eans that the covariance atrix (17) should be estiated and inverted online. However, this operation is c j+ n j+ n 2 c c (14) 328

5 very coputationally expensive and it ay be difficult to estiate the saple covariance atrix in real tie if the nuber of saples MN is large. Furtherore, the nuerical calculation of the weights W MVDR using the expression (16) ay be very unstable if the saple covariance atrix is ill-conditioned. A nuerical stable and coputationally efficient algorith can be obtained by using QR decoposition of the incoing signal atrix. This atrix is decoposed as X=QR, where Q is the unitary atrix and R is the upper triangular atrix. Hence the QR-based algorith for calculation of beaforer weights includes the following three stages: H * H 1 The linear equation syste R z1 = ac is solved for z 1, and the solution is z1 = ( R ) ac * * 1 * The linear equation syste Rz 2 = z 1 is solved for z 2, and the solution is z2 = R z1 The weight vector W ) ) * H * is obtained as W = z /( a ) 2 c z2 Acquisition stage The ain purpose of the acquisition stage is to identify the visible satellites in the incoing data and then find the beginning point of the C/A code and estiate the rough Doppler shift by correlating the incoing signal with the local signal replica. Since the C/A code is 1s long, the acquisition ust be perfored on at least 1s of the incoing data. According to [1] the acquisition algorith consists of the following steps: Perfor the FFT of the input data x( converting the into frequency doain as X(. Take the coplex conjugate X( obtaining the outputs X*(. Generate 21 local codes l si ( as l i (=c(exp(jf i ),, where f i =f c +i khz, f c - is the interediate frequency and i=-10,-9,.9, 10. The local code is the product of the C/A code satellite and a coplex IF signal. The frequencies f i of the local codes are separated by 1 khz. Perfor the FFT on l si ( to transfor the to the frequency doain as L i (. Multiply X*( and L i ( point by point obtaining the result R i (.. ( r i Take the IFFT of R i ( to transfor the result into tie doain as r i ( and find its absolute value The axiu of r i ( k ) in the kth location and ith frequency bin gives the beginning point of the C/A code in (1/f s ) resolution in the input data and the carrier frequency in 1kHz resolution (f s is the sapling frequency). When both paraeters, the beginning point of the C/A code and the carrier frequency, are found, this inforation is passed on to the tracking algorith. 6. Perforance easures The perforance easures presented here serve to evaluate the perforance of the two beaforing techniques and select the best paraeters of the space-tie processing in order to provide a high degree of the acquisition efficiency. These quality easures are defined as follows: SINR iproveent factor For a single eleent array the input SINR is defined as: PS SINR = INPUT PN + Pi + P (18) where P S is the average power of the desired GPS signal, P N is the receiver noise power, P i is the total broadband jaing power, and P is the total ultipath power. Using the superposition principle, the SINR at the beaforer output can be evaluated as H H H ( W s)( W s) SINROUT = (19) H H H ( W x0 )( W x0 ) where x 0 is the total noise + interference + ultipath signal that arrives at an antenna array eleent. Therefore the iproveent in SINR provided by the beaforer can be found as: K = SINR / SINR (20) SINR OUT This quality easure evaluates the capability of the beaforer (conventional or MVDR) to cancel the interference power in the incoing signal and at the sae tie to preserve the desired signal power. Probability of detection (P D ) This quality easure evaluates the capability of the overall beaforing + acquisition algorith to detect the beginning point of the C/A code and find correctly the carrier frequency of the incoing IF signal. This perforance easure can be evaluated using the Monte Carlo approach: P = K / N (21) D success where K success is the nuber of events, in which both paraeters being estiated, the beginning point and the carrier frequency, are found correctly; N total is the total nuber of Monte Carlo runs. total INPUT 329

6 7. Siulation results In this section, 500 coputer siulations of the overall beaforing + acquisition algorith are perfored in order to evaluate the influence of such factors as array configuration, nuber of array eleents and sapling rate on the capability of the algorith to operate effectively in conditions of strong broadband interference and ultipath. The sallest intereleent spacing in antenna arrays is usually equal to or slightly less than half a wavelength of the satellite carrier frequency (/2) in order to avoid the proble of spatial undersapling. However, a saller intereleent spacing than /2 increases the risk of utual coupling between antenna eleents. In GPS/GLONASS applications, the intereleent spacing is approxiately 0.09 for the L1- band, and this technical deand puts a physical liitation on how sall the array can be used in a GPS/GLONASS receiver and how any eleents is appropriate to be used in such an antenna array. For that reason we consider a liited nuber of array configurations, which have a sall nuber of eleents. Four exaples of such arrays are presented in Table 1. URA-4 UCA-4 UCA-7 URA-9 Table 1: Array configurations The first two arrays, rectangular (URA-4) and circular (UCA-4), contain four eleents with half-wavelength intereleent spacing. The third of arrays is a unifor circular array with seven eleents (UCA-7) where the first eleent is located at the array center. The radius of this array is equal to one half-wavelength. The last of arrays is a rectangular array with 9 eleents. It is well known that the nuber of antenna eleents M is related to the nuber of broadband jaers that can be nulled by the space-tie beaforing algorith. Typically, the nuber of broadband jaers that can be nulled by the space-tie filtering corresponds to (M-1). With this in ind five scenarios with the ain paraeters described in Table 2 are siulated. Scenario Jaing Multipath GPS signal Scenario 1 Receiver noise only Scenario 2 URA-4 and UCA-4 Three jaing sources: Elevation: θ=40 Aziuth: φ 1 =-70 ;φ 2 =-60 ;φ 3 =60 ISR: 10dB. 100dB - Elevation: θ=40 Aziuth: φ=0 Doppler shift:5 khz SNR: -20dB Scenario 3 URA-9 and UCA-7 Four jaing sources: Aziuth: φ 4 =70 ISR: 10dB 100dB - Scenario 4 - Scenario 5 - URA-4 and UCA-4 One or three echoes: Elevation: θ=10 Aziuth: φ=rav[-90,-40 ]& Rav[90,40 ] Attenuation: 2dB Doppler shift: 0; Phase: 0 Excess delay: 0.5 PRN chip URA-4 and UCA-4 One, three or five echoes: Elevation: θ=10 Aziuth: φ=rav[-90,-40 ]& Rav[90,40 ] Attenuation: 2dB Doppler shift: 0; Phase: 0 Excess delay: 0.5 PRN chip Table 2: Jaing and ultipath scenarios Variant 1 Carrier: MHz Sapling: MHz Variant 2 Carrier: MHz Sapling: MHz Duration: 1s C/A code: satellite

7 In the first scenario, the received signal consists of a single GPS signal and additive white Gaussian noise with unity variance. For this scenario, the estiates of the iproveent in SNR evaluated by 500 siulation runs are tabulated in the first row of Table3. It can be seen that in the interference-free environent both beaforers have the sae iproveent in SNR. The second scenario, in which where three broadband jaing signals are added to the GPS signal and noise, is siulated with four-eleent arrays: URA-4 and UCA-4. Analogically, the third scenario, in which four jaing signals are cobined with the GPS signal and noise, is siulated with arrays URA-9 and UCA-7. For all these arrays, the estiates of the iproveent in SINR are evaluated as a function of the interference-to-signal ratio (ISR) (Table 3). Nuerical results given in Table 3 are plotted in Fig. 5 for URA-4 and UCA-4 and in Fig.6 for URA-9 and UCA-7. SINR Iproveent (db) ISR Conventional Beaforer MVDR Beaforer (db) URA-4 UCA-4 URA-9 UCA-7 URA-4 UCA-4 URA-9 UCA Table 3: SINR Iproveent evaluated for the two beaforers Fig. 5: SINR iproveent for URA-4 and UCA-4 Fig. 6: SNR iproveent for URA-9 and UCA-7 The antenna pattern of the UCA-7 calculated for the case of four strong jaing signals (ISR =60 db) is shown in Fig.7. The inial gain is created in the direction of arrival of jaing signals (-70,-60, 60, 70 ) while the axial gain is created in the direction of the GPS signal (0 ). The correlator outputs are plotted in Fig. 8. Fig. 7: Antenna pattern (4 jaers, ISR =60 db) Fig. 8 Correlator output (UCA-7, 4 jaers, ISR =60dB) 331

8 As is shown above, the capability of each of the two beaforers to suppress broadband interference depends not only on the array geoetry - URA and UCA but the nuber of array eleents as well - 4, 7, 9. However, unlike the conventional beaforer the MVDR-beaforer very successfully itigates broadband interference even if the interference intensity becoes 100 db over the desired GPS signal. The MVDR-beaforer achieves the ost effect for URA-9. As whole, the rectangular array configuration is ore effective for broadband interference suppression than circular one. The effectiveness of the overall beaforing + acquisition algorith can be estiated in ters of the detection probability, which is defined as the probability of an event in which both paraeters of the incoing GPS signal, the beginning point and the carrier frequency, are correctly estiated. The detection probability is calculated using 500 siulation runs. The probability of detection is estiated for two values of the sapling frequency of the incoing data. The estiates obtained are presented in Table 4 for URA-4 and UCA-4 and in Table 5 for URA-9 and UCA-7. The analogical graphical results are plotted in Fig. 9, 10, 11 and 12. ISR (db) Probability of detection (P D ) Sapling frequency ( MHz) Sapling frequency ( MHz) Conventional Beaforer MVDR QR Beaforer Conventional Beaforer MVDR QR Beaforer URA-4 UCA-4 URA-4 UCA-4 URA-4 UCA-4 URA-4 UCA Table 4: Probability of detection evaluated for URA-4 and UCA-4 In Table 4 and Table 5, the ISR values that correspond to the probability of detection above 0.9 (noted in bold) can characterize the anti-jaing capability of the overall beaforing + acquisition algorith. The estiates in Table 4 show that the MVDR-beaforer increases the antijaing capability of the acquisition algorith by 10dB copared to the conventional beaforer. ISR (db) Probability of detection (P D ) Sapling frequency ( MHz) Sapling frequency ( MHz) Conventional Beaforer MVDR QR Beaforer Conventional Beaforer MVDR QR Beaforer URA-9 UCA-7 URA-9 UCA-7 URA-9 UCA-7 URA-9 UCA Table 5: Probability of detection evaluated for URA-9 and UCA-7 According to the results fro Table 4 and Table 5, the increase of the sapling frequency iproves the antijaing protection of the acquisition algorith only in case of URA-4 (Fig.9 and Fig.10) while for the other arrays (UCA-4, URA-9 and UCA-7) the effect of the sapling frequency on the anti-jaing capability of the acquisition algorith is insignificant. Taken as a whole, the anti-jaing capability of the acquisition algorith entirely depends on the effectiveness of the beaforing algorith. The graphics in Fig.11 and Fig.12 clearly illustrate that the acquisition algorith ensures save operation in conditions of very strong jaing signals if the MVDR-beaforing is a preliinary to acquisition. 332

9 Fig. 9: Detection probability (URA-4 and UCA-4) Fig. 10: Detection probability (URA-4 and UCA-4) Fig. 11: Detection probability (URA-9and UCA-7) Fig. 12: Detection probability (URA-9and UCA-7) The last two scenarios with the paraeters given in Table 2 describe such an environent when ultipath signals arrive at the receiver input in cobination with the GPS signal. Scenario 4 is the case when one or three ultipath signals are present in the received signal while scenario 5 is the case when one, three or 5 ultipath signals are cobined with the desired GPS signal. In siulations, it is assued that the direction of arrival of each ultipath is a rando variable uniforly distributed in range [-90,-40 ] and [90, 40 ]. For these scenarios, the detection probability estiated by siulations is tabulated in Table 6. When five ultipath signals are present at the receiver input, the signals at the correlator outputs are depicted in Fig. 13 and Fig. 14. Nuber of echoes Probability of detection (P D ) Acquisition (Conventional Beaforer) Acquisition ( MVDR QR Beaforer) URA-4 UCA-4 UCA-7 URA-9 URA-4 UCA-4 UCA-7 URA Table 6: Probability of detection evaluated for two beaforers and two array configurations Analysis of the results in Table 6 shows that in both cases, with the conventional beaforer and with the MVDR-beaforer, the acquisition perforance degrades with increasing of the nuber of ultipath signals. One can see that the MVDR-beaforer has no discernible advantage over the conventional one in the ultipath environent. In such environent, the key advantage gives the usage of rectangular array configurations (URA-4 and URA-9) because these arrays allow aintaining a ore stable operation of the acquisition algorith in conditions of ultipath. However this is not true for circular array configurations. 333

10 Fig. 13: Correlator output (URA-9, 5ultipath) Fig. 14: Correlator output (UCA-7, 5ultipath) Conclusion The results obtained show that the stable acquisition of GPS signals depends on several iportant factors when strong broadband interference and ultipath signals are present at the receiver input. In the first place the effectiveness of the acquisition algorith depends on the capability of the beaforing odule to effectively suppress jaing signals and ultipath before processing by the acquisition algorith. In this context, the adaptive MVDR-algorith is the preferable algorith for itigation of jaing signals. It is also shown that the physical array configuration has a very significant effect on the overall acquisition perforance. In order to keep the size of arrays and beaforing coplexity sall, only array configurations with a sall nuber of eleents have been considered in the paper. Four different array configurations are copared in both interference-free and ulti-interference environents. The results show a very good anti-jaing capability of all arrays if the MVDR-algorith is used for beaforing. On that score the best results, however, are obtained for two rectangular arrays. As for the sapling rate of the incoing data, the influence of this factor is significant in case of arrays with a sall nuber of eleents (URA-4 and UCA-4). For arrays with a larger nuber of eleents (URA-9 and UCA-7) this influence is insignificant. Acknowledgent This work is partially supported by the Bulgarian Science Fund (the project MI-1506/05 and the project MU-FS-05/2007) Reference 1. Jaes Bao-Yen Tsui, Fundaentals of Global Positioning Syste Receivers: A Software Approach, Wiley Interscience, John Wiley&Sons, J. Sklar, Interference itigation approaches for the Global Positioning Syste, MIT Lincoln Laboratory Journal, vol.14, No 2, 2003, H. L. Van Trees, Optiu Array Processing. Part IV of Detection, Estiation, and Modulation Theory. New York, NY: JohnWiley and Sons, Inc., Z.Fu, A. Hobostel, J. Haesfahr, A. Konovaltsev, Suppression of ultipath and jaing signals by GPS/Galileo applications, GPS Solutions, No 6, 2003, J. Soubielle, I. Fijalkow, P. Duvau and A. Bibaut, GPS positioning in a ultipath environent, IEEE Trans. on Signal Processing, vol. 50, No 1, 2002, D.Moelker, E. van der Pol, Adaptive antenna arrays for interference cancellation in GPS and GLONASS receivers, Proc. IEEE 1996 Position Location and Navigation syp, Apr. 1996, J.Lee, L.Song, J. Joung, Unifor circular array in the paraeter estiation of coherently distributed sources, Proc. IEEE Military Counications Conf. MILCOM 2002.,Oct.,2002,vol.2, R. Monzingo and T. Miller, Introduction to Adaptive Arrays, New York: Wiley, L. Tuonery, I. Proudler, A. Farina, J. McWhirter, QRD-based MVDR algorith for adaptive ulti-pulse antenna array signal processing, in Proc. Radar, Sonar, Navigation, vol.141, No 2, 1994, P. Vouras, B. Freburger, Application of adaptive beaforing techniques to HF radar, in Proc. IEEE conf. RADAR 08, May, 2008,