AU8735398A - Method for estimating the angular dispersion of a signal transmitted by a transmitter to a receiver and corresponding receiver - Google Patents

Description

1 A METHOD OF ESTIMATING THE ANGULAR SPREAD OF A SIGNAL TRANSMITTED BY A TRANSMITTER TO A RECEIVER, AND A CORRESPONDING RECEIVER The field of the invention is that of radio 5 transmission systems, in particular cellular systems. To be more precise the present invention concerns a method consisting in estimating the angular spread of a signal transmitted by a transmitter to a receiver. The invention can be applied with advantage in an 10 SDMA (Space Division Multiple Access) transmission network. The invention also concerns a receiver implementing the above method. The following description is given in the context of determining the angular spread of a signal transmitted 15 between a transmitter and a receiver. As shown in Figure 1, when a transmitter 10 transmits a signal for the attention of a receiver 11 multipath propagation due to the presence of obstacles in the environment of the transmitter 10 generates angular spread of the 20 transmitted signal. This angular spread is characterized by an angle a that to a first approximation is inversely proportional to the distance between the transmitter 10 and the receiver 11. As a general rule, for any receiver receiving a 25 signal transmitted by a transmitter, whether it is part of a cellular network or not, angular spread is an important parameter that has to be allowed for. For example, it is desirable to adapt the algorithm for equalizing the signal received at the receiver 11 to the 30 value of angular spread a. One aim of such adaptation is to use a more powerful equalization algorithm when there is a high angular spread and a less powerful equalization algorithm when there is a low angular spread. Also, when there is a high angular spread, for 35 example in the case of SDMA transmission in a cell of a cellular network - in which case the receiver 11 is a base station and the transmitter 10 is conventionally a 2 mobile terminal - it is not possible to associate a precise direction of arrival (DOA) of the signal and therefore to form a lobe in the direction of the transmitter 10. The re-use factor is then high. To be 5 more precise, in the case of SDMA transmission, lobes in the direction of the transmitter cannot be formed when the transmitter is near the base station, i.e. when a has a high value. For the above reasons in particular, it is 10 therefore desirable to know the value of the angular spread a at the receiver 11. There are various techniques for estimating the value of a. The receiver usually includes an array of antenna elements each receiving the transmitted signal, 15 this array being connected to means for calculating angular spread. The main drawback of the prior art techniques is that they all require considerable processing of the received signals. For example, MUSIC or ESPRIT type algorithms can determine the angular 20 spread of a received signal but require computation of covariance matrices, calling for considerable computing power. Another disadvantage is that the received signal must be observed over a long time period. Also, the above algorithms are lacking in robustness in terms of 25 convergence. An object of the present invention is to remedy the above drawbacks. To be more precise an object of the invention is to provide a method of estimating the angular spread of a 30 signal transmitted by a transmitter to a receiver, which method is easy to use and requires little processing power. Another object of the invention is to provide a receiver comprising means for estimating angular spread. 35 The above objects, together with others that become apparent below, are achieved by a method of estimating the angular spread of a signal transmitted by a 3 transmitter and received by a receiver, the method consisting in estimating the angular spread exclusively by observing data representative over at least a given time period of a variable related only to the amplitudes 5 or powers of the signals received by each antenna element of a receiver antenna array. In a first embodiment the data is an average of the variances of the amplitude differences of signals received by the antenna elements, the differences being 10 calculated taking the antenna elements in pairs. In a second embodiment the data is an average of the variances of the amplitudes of signals received by each antenna element. In a preferred application the transmitter is a 15 mobile terminal and the receiver a base station of a cellular transmission network. The cellular transmission network advantageously uses SDMA transmission. The estimated angular spread is preferably compared 20 to a reference value to enable the transmitter to perform SDMA transmission selectively. In a variant, the data is compared to a reference value selectively to enable the transmitter to perform SDMA transmission. 25 The invention also provides a receiver comprising an antenna array adapted to enable the angular dispersion of a signal received at the antenna array to be estimated, the receiver including means supplying data representative of a variable related only to the 30 amplitudes or powers of the signals received by each antenna element of the array. The receiver advantageously further includes converter means for supplying the value of the angular spread. 35 Other features and advantages of the invention become apparent on reading the following description of one preferred embodiment of the invention given by way of 4 non-limiting example and from the accompanying drawings, in which: - Figure 1 is a diagram showing the angular spread of a signal transmitted from a transmitter to a receiver; 5 - Figure 2 is a plan view of an antenna array; - Figure 3 shows the angular spread of the power of a component received at an antenna of the antenna array; - Figure 4 shows the average of the variances of the power differences obtained for pairs of antenna 10 elements of an antenna array as a function of the angular spread for various average DOA; - Figure 5 shows a base station of an SDMA network and an area in which SDMA transmission is not possible; - Figure 6 shows the results of simulating the 15 probability that the angular spread value calculated using a first calculation method in accordance with the invention is higher than a reference value, this probability being a function of angular spread; - Figure 7 shows the result of another simulation 20 of the probability that the angular spread value calculated using the first calculation method is higher than a reference value, this probability also being a function of angular spread; - Figure 8 shows the average of the variances of 25 the powers obtained for the antenna elements of an antenna array as a function of angular spread for various average DOAs; - Figure 9 shows the result of simulating the probability that the angular spread value calculated 30 using a second calculation method in accordance with the invention is higher than a reference value, this probability being a function of angular spread; and - Figure 10 shows a receiver employing the method of the invention. 35 Figure 1 is described above with reference to the prior art. The invention proposes to exploit statistical -4o 5 properties of ,spatial and temporal fluctuations in the amplitude or power of signals received by a receiver antenna array. The antenna array comprises a number M of antennas where M is at least equal to two, for example 5 M = 12. The array can be a three-dimensional array but for simplicity the following description refers to a plane array (two dimensions) of equally spaced antennas. The signal s(t) transmitted by the transmitter has a wavelength k. Each antenna of the array receives a set 10 of signals confined within the angle a, as shown in Figure 2, which shows antennas 200 through 2 0 m_1. Figure 2 shows the distance don between the antenna 200 and the antenna 20n of the array (n lies in the range 0 to M-1) . The parameter Oen is the angle between the 15 average DOA of the signal and the normal to the straight line segment connecting the antenna 200 and 20n. Particular values are denoted as follows: Doo = 0 and 000 = 0 The angular distribution P (a) of the average power 20 arriving in an average direction 0 is characterized by a parameter arms (rms signifying "Root Mean Square") shown in Figure 3. Because each variable a has a (Gaussian) average value of zero: 25 arms - fa 2 p 2 (a)da arm In practice a discrete number N of separate components received corresponding to N discrete angles ai 30 can be considered. Consequently, the general expression for the signal received at the N antenna elements of the antenna array can be determined: N 2 j1-don sin(9on +a) xn,(t) = i g(t)s(t - r je + bn (t) i=1 6 for n from 0 through M-1, and where gi(t) is the complex gain of the component I, -r is its relative delay and bn(t) is the noise component. 5 It is known that: E[Igi (t)12] = P(a 1 ) The amplitude of the signal received at each antenna 10 element is therefore considered hereinafter: T Xn = xnW(t)|2dt 0 for n from 0 through M-1, or more simply the power 15 Xn2 of the signal received at each antenna element. To be more precise, the invention proposes to estimate the angular spread of a signal transmitted by a transmitter and received by a receiver from the single observation of data representative over at least a given 20 time period of a variable related only to the amplitudes or powers of the signals received by each antenna element of the receiver antenna array. In practice a number B of successive measurements are performed on the M antenna elements, as described below. 25 The statistical properties of the powers of the components received by the antenna elements are directly related to the value of the angular spread arms. Any combination of the measured powers is characterized by a statistical model that can be identified and thereafter 30 used to estimate the angular spread. Depending on the intended application of that estimate, it is possible to determine a combination of the powers yielding an appropriate statistical model. Below, Gaussian type noise is considered having 35 variance an that is not necessarily equal. In particular 7 cases (for example transmitting known learning sequences), the contribution of noise can be eliminated more or less completely. In other cases the solutions proposed below offer significant performance even if 5 noise power cannot be eliminated. For example, the first method proposed eliminates the variance of the noise if that variance is the same for all of the antenna elements. Various embodiments of the method of the invention 10 are described below. In all these embodiments it is assumed, for the purposes of the simulations, that the noise is Gaussian noise, that the transmission channel is a Rayleigh channel, that the distribution angular power profile is uniform (P(a) is constant), that N (the number 15 of multipaths) is equal to 10 and that the antenna array is a plane array comprising M = 12 antenna elements spaced apart by a distance d/X = 0.5. In all the simulations the average DOA 0 is uniformly distributed in a sector extending from -60' to +60' relative to the 20 normal to the antenna array. First method The first method consists in determining the angular spread of the signal received at an array comprising a plurality of antenna elements. The method consists in 25 determining an average of the variances of the amplitude differences of the signals received by the antenna elements, the differences being calculated taking the antenna elements in pairs. To be more precise, a set Zn of variables is 30 considered that is defined as follows: Z = Xn - Xn+ 1 for n varying from 0 through M-2 This yields a set of variables Zn each representative of the power (or amplitude) difference between the signals received at two antennas spaced by 35 d/X. The average VARz of the variances of the above differences as a function of the angular spread is shown in Figure 4 for discrete angles 0 equal to O 7 8 (corresponding to a transmitter facing the antenna array), ±30' and ±600. For each value of 0 it is possible to determine a single value a from VARz. A one-to-one relation is 5 established here. Note also a low spread of the values of a as a function of VARZ for angles 0 in the range ±30', which corresponds to a coverage angle of 60'. The following relations are used here: iB b VARz = XVAR B b=1 10 with b 1 M-2 VARZ= M1 (Zn,b - Zb) M-1 n=O 15 and 1M-2 Zb -I Zn,b M-1 n=0 In other words, VARZ is the average of the B 20 variances, each variance being evaluated from M-1 differences Za,b calculated for an analysis b. B preferably has a high value, as explained below in connection with method 2. This yields a matrix with M-1 rows and B columns of variables Zf,b (n denotes the signal 25 and b denotes the occurrence) with b from 1 through B. Application of method 1 Method 1 is advantageously applied to indicating the probability that the calculated angular spread value (or the calculated value of VARz) is higher than a reference 30 value aref (or VARref) - This probability is directly related to the accuracy of the estimate obtained. It is ideally equal to 0 if the estimate is below the threshold value or equal to 1 if the estimate is above the threshold value. This probability is referred to below 9 as the correct detection probability. To be more precise, in the context of an application of the invention to an SDMA transmission network, it is necessary to know the value of the angular spread of the 5 signals transmitted by terminals for the attention of a base station, as shown in Figure 5. Figure shows a base station of an SDMA network (for example of the GSM or CDMA type) and an area within which SDMA transmission is not possible. 10 The base station 11 covers a cell 50 which is ideally a circular cell. SDMA transmission with terminals 101 through 105, for example mobile terminals, presupposes that the base station 11 is capable of forming lobes in the directions of terminals 10, through 15 105. Here this formation of lobes forms four sectors 51 through 54 and the transmission resources used in those sectors can be the same (the same carrier frequency and the same spreading code for CDMA transmission or allocation of the same time interval for TDMA 20 transmission). For example, terminal 10, can use the same transmission channel as terminal 104 because they are in different sectors. In an area 55 around the base station 11 SDMA transmission is not possible because the angular spread is too high in this area: because of secondary 25 lobes of the transmit antennas of the base station 11, it is not possible to send data to a terminal within the area 55 without another terminal sharing the same resource receiving data which is not intended for it. In this situation mutual interference between calls occurs. 30 It is therefore important to detect whether a terminal is inside or outside the area 55. For this purpose, the invention proposes, for example, to use the value of the angular spread obtained by method 1 as the criterion for whether a terminal is present or absent in 35 the area 11. If the value of a obtained is below a reference value aref, the terminal concerned is deemed to be outside the area 11, and if the value of a is above 10 the reference value arer, the terminal concerned is deemed to be inside the area 11. The value of cIref therefore corresponds to the angular spread of signals transmitted by terminals in the immediate vicinity of the boundary of 5 the area 11. The above-mentioned concept of the probability of correct detection of the angular spread value corresponds precisely to this criterion because it indicates the pertinence of the method employed to determine angular 10 spread. For example, Figure 6 shows a simulation of the correct detection probability Pbd as a function of angular spread (obtained using method 1) for a reference value (ref equal to 7, which corresponds approximately to a VARz 15 equal to 0.03 (see Figure 4). The three characteristics 60, 61, 62 shown respectively correspond to B = 3, B = 10 and B = 30. Note that the slope and therefore performance increases with the number of measured samples B processed. 20 Variant application of method 1 A variant application of method 1 consists in taking a majority vote using the B different decisions based on VARZb (variance of the power differences obtained for two adjacent antenna elements), rather than comparing the 25 average of the B variances to a reference value. Each of the values VARZ is compared to the same reference value, and after each of the b comparisons, a majority vote is taken to determine whether or not the number of comparisons for which the response was "yes" is greater 30 than the number of comparisons for which the response was "no". The characteristics of Figure 7 are constructed in this way, with Pbd being the correct detection probability. The characteristics 70, 71 and 72 are respectively 35 obtained for B = 3, B = 10 and B = 30. Characteristics 71 and 72 coincide which indicates that there is no further improvement in accuracy for B greater than 10.

11 Apart from this, performance is comparable to that of the first application of method 1. Method 2 The second method proposed is also based on 5 observing data representative of a variable related to the amplitudes of the signals received by each antenna element. However, the data is not the value of VARz, as in method 1, but an average VAR, of the variances of the amplitudes (or powers) of the signals received by each 10 antenna element. Thus in method 2 there is no subtraction of the amplitudes of signals received by two antenna elements. To be more precise: M-1 15 VAR =- I (Xn,b - Xb) M n=O and 1 M-1 Xb =- Xn,b M n=O 20 The value of b is in the range 1 through B (B successive measurements). VAR is the average of the variances VAR . Figure 8 shows the average VAR, of the variances of 25 the powers obtained at the antenna elements of an antenna array as a function of the angular spread ax for various average DOA 0. Here there is also a one-to-one relation between VAR and cx and a low spread of the estimate for angles 0 30 in the range ±300. Application of method 2 As in the first application of method 1 it is beneficial, for example in the context of transmission in an SDMA network, to compare the average of the calculated 35 variances with a reference value VARref or aref to determine a correct detection probability Pbd. R4 T 12 Figure 9 shows the probability Pbd as a function of the angular spread a calculated for B analyses, B being equal to 3, 10 or 30. The reference value used is aref = 7 which corresponds to VARref = 0.75. Note that 5 performance improves as B increases. Practical implementation The invention also provides a receiver employing the above-described method. Figure 10 shows one such receiver. 10 An antenna array 100 comprises a plurality of antenna elements 200 through 2 0 m_1. Each antenna element is connected to processing means 101 supplying data (for example VARz or VARx depending on the method employed) representative of a variable related only to the 15 amplitudes or to the powers of the signals received by each antenna element of the array 100. This data is then fed to converter means 102 supplying the value of the angular spread a. The processing means 101 separate the channels in the case of CDMA transmission. 20 At a subsidiary level, and as previously indicated, the value of VARZ (VARy) or that of a can be compared to a threshold value VARref or aref in comparator means 103 supplying a comparison signal COMP indicating whether the condition VARZ (or VARX) < VARref or a < aref is achieved or 25 not. Such comparator means are particularly beneficial in an application to an SDMA receiver. The invention applies not only to estimating the angular spread of a signal received by a receiver but in 30 particular to base stations of cellular networks, whether they are of the GSM, CDMA or other type, to enable re-use of transmission resources (SDMA).

Claims (9)

1. A method of estimating the angular spread (ca) of a signal transmitted by a transmitter and received by a receiver, said method consisting in estimating said 5 angular spread (a) exclusively by observing data (VARz, VARx) representative over at least a given time period of a variable related only to the amplitudes or powers of signals received by each antenna element (200 through 20 m-1) of a receiver antenna array (100). 10

2. A method according to claim 1, characterized in that said data is an average of the variances of the amplitude differences of signals received by said antenna elements, said differences being calculated taking said antenna 15 elements in pairs.

3. A method according to claim 1, characterized in that said data is an average of the variances of the amplitudes of signals received by each antenna element. 20

4. A method according to claim 1 or claim 2, characterized in that said transmitter is a mobile terminal and said receiver a base station of a cellular transmission network. 25

5. A method according to claim 4, characterized in that said cellular transmission network uses SDMA transmission. 30

6. A method according to any one of claims 1 to 5, characterized in that said estimated angular spread (c) is compared to a reference value (aref) selectively to enable SDMA transmission by said transmitter. 35

7. A method according to any one of claims 1 to 5, characterized in that said data (VARZ, VAR,) is compared to a reference value (VARref) selectively to enable SDMA 14 transmission by said transmitter.

8. A receiver comprising an antenna array (100) adapted to enable estimation of the angular dispersion of a 5 signal received at said antenna array, characterized in that it includes means (101) supplying data (VARz, VAR,) representative of a variable related only to the amplitudes or powers of signals received by each antenna element of said array (100). 10

9. A receiver according to claim 8, characterized in that it further includes converter means (102) for supplying said value of said angular spread (a).