This application is based on and hereby claims priority to PCT Application No. PCT/DE01/03923 filed on 12 Oct. 2001 and German Application No. 100 51 133.3 filed on 16 Oct. 2000, the contents of which are hereby incorporated by reference.
The invention relates to a beam shaping method.
In the embodiments of the beam shaping method described in this application, covariance matrices are calculated singly for all sufficiently strong taps of a radio signal, eigenvectors and eigenvalues of the covariance matrices thus obtained are calculated and the eigenvectors corresponding to the strongest eigenvalues are used for the beam shaping.
The number of eigenvalues which can be used in practice in the beam shaping is limited. There are different types of reasons for this.
In a method as described with reference to FIG. 4 of patent application DE 10032426.6, a problem lies in the limited bandwidth which is available for transmitting information describing the receiving situation back from the subscriber station to the base station and which must be divided into the transmission of components of the measured eigenvectors, which are valid or usable over relatively long periods, on the one hand, and on the other hand, a desired short-term weighting of these eigenvectors in the downlink signal. This makes it necessary to restrict oneself to the transmission of only the most important eigenvectors in the interest of rapid updating of the weighting.
In the method described with reference to FIG. 5 of patent application DE 10032426.6, the base station uses a linear combination of eigenvectors as weighting vectors for radiating. A greater number of eigenvectors used the greater the number of propagation paths covered with the radio signal, both to the subscriber station for which the signal is intended and to other stations for which the signal represents interference. To keep this interference within limits, the number of eigenvectors included in the linear combination must be limited.
It is one possible object of the invention to specify a beam shaping method in which the greatest possible protection against fading drop-outs is achieved even with a limited number of eigenvectors simultaneously taken into consideration.
One aspect of the invention is based on the finding that, under certain conditions of propagation, limiting the number of eigenvectors can lead to problems (Canyon effect) as is shown in
The base station BS has an adaptive antenna with M elements and it is assumed that the number of eigenvectors taken into consideration in the beam shaping by the base station BS is limited to two.
When the base station BS carries out the method described with reference to FIG. 5 of patent application 10032426.6, there is the possibility that the strongest eigenvalues correspond to the transmission paths M1, M2 in each case and that, in consequence, the base station uses a linear combination of the eigenvectors corresponding to these two propagation paths in the beam shaping. In such a case, it generates two beams which extend in the same direction from the base station.
The partial coincidence of the propagation paths inevitably results in a correlation between them, i.e. if one of them experiences extinction, there is an increase in the probability of extinction of the second one, and thus the probability that no further communication is possible between base station and subscriber station.
A similar problem also exists in the method described with reference to FIG. 4 of patent application 10032426.6. In this case, the eigenvectors corresponding to the propagation paths M1, M2 are not used simultaneously but following one another. Here, too, however, the risk that both propagation paths M1, M2 are simultaneously affected by extinction is greater than in the case of two non-overlapping propagation paths such as, for instance, M1 and M3.
To eliminate this problem, a method for beam shaping in a radio communication system is proposed, the radio communication system comprising subscriber stations and a base station which has an antenna apparatus with a plurality of antenna elements which radiate a downlink signal in each case weighted with coefficients wi, i=1, . . . , M of a current weighting vector w, in which a plurality of first weighting vectors w(j) are formed in an initialization phase, and the current weighting vector w, used for radiating a time slot of the downlink signal intended for the subscriber station (MSk), is cyclically redetermined in an operating phase by the first weighting vectors formed, the first weighting vectors w(j) being determined with the stipulation that they are all orthogonal to one another.
A particularly simple possibility of ensuring the orthogonality of the first weighting vectors formed is to set up a single first spatial covariance matrix which is composed of contributions of short-term covariance matrices determined for the individual taps of the uplink or downlink signal, and to select the first weighting vectors among their eigenvectors. In particular, this first covariance matrix can be obtained by forming a mean value of the short-term covariance matrices obtained for the individual taps. Since this first spatial covariance varies only slowly, it can also be called a “long-term” covariance matrix in contrast to the short-term covariance matrices. Since the eigenvectors of a covariance matrix are orthogonal, the desired orthogonality thus occurs by itself taking into consideration only a single first spatial covariance matrix.
The short-term covariance matrices are also suitably averaged in time for determining the first spatial covariance matrix. Thus, the short-term covariance matrices can be used at the same time as second spatial covariance matrices in the sense of patent application 10032426.6.
These and other objects and advantages of the present invention will become more apparent and more readily appreciated from the following description of the preferred embodiments, taken in conjunction with the accompanying drawings of which:
Reference will now be made in detail to the preferred embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to like elements throughout.
The configuration of a base station and of a subscriber station which are suitable for carrying out the method proposed by the inventor have been described in U.S. patent application Ser. No. 10/312,769 filed on Dec. 30, 2002, which is hereby incorporated by reference.
The first embodiment of the method is described with reference to
In an initialization phase, the signal processor SP of the subscriber station MS determines in step 11 by training sequences, transmitted in an i-th time slot of the downlink signal, individually for each tap of the downlink signal the impulse responses of each antenna element AE1, . . . , AEM. (Naturally, “each” tap does not mean the totality of all contributions in the received signal which may be detectable but only all those which can actually be detected and isolated by the receiver of the subscriber station or only those which have a sufficient signal strength so that their evaluation can improve the reliability of the symbol estimation.) These impulse responses are assembled by the signal processor SP in the manner known e.g. from the quoted DE 198 03 188 to form spatial covariance matrices Rxx,a,i where a is an index of a tap. These spatial covariance matrices Rxx,a,i are supplied to a computing unit RE.
In step 12, the computing unit RE forms a mean value Rxx,i of the covariance matrices Rxx,a,i supplied by the signal processor SP for the i-th time slot, in that it adds them and divides the result by the number of taps. (Since only the proportions of the coefficients of this averaged matrix are of importance for the purposes of the method, the division step can also be omitted.) The matrix thus obtained is called short-term or second spatial covariance matrix.
In step 13, the computing unit RE forms a sequence of temporal mean values
(
(
where (
Steps 11 to 13 are repeated a predetermined number z of times in order to obtain a temporal averaging over an informative period in which fluctuations of the channel estimation due to rapid fading essentially average out. This period can be of the order of magnitude of minutes. ρ is selected in such a manner that the contribution of Rxx,0 to the mean value (
This is followed by an analysis of the eigenvalues and eigenvectors of the first covariance matrix (
The analysis can extend to all eigenvectors and eigenvalues of the covariance matrix (
The coefficients of the eigenvectors w(k,1), w(k,2), . . . are stored at the subscriber station MS (step 15) and combined with the user data stream coming from the voice processing unit SSV and transmitted via the antenna to the base station (step 16) where they are used as first weighting vectors for the beam shaping. They are received by the base station and stored in its storage unit SE for use as coefficients for the multiplier M of the beam shaping network.
Since these eigenvectors w(k,1), w(k,2), . . . have in each case been obtained from a single second covariance matrix, they are orthogonal; the transmission paths corresponding to them have different directions of arrival at the base station BS and are uncorrelated.
The computing unit RE now changes to an operating phase in which it continues to form short-term covariance matrices (step 17) as described above from the covariance matrices supplied for each tap and each time slot by the signal processor SP and multiplies these by each one of the eigenvectors stored in the storage unit and transmitted to the base station, in order to find the eigenvalues of these vectors for the relevant short-term covariance matrix Rxx,i (step 18). The number of the eigenvector having the greater eigenvalue is transmitted to the base station via the control unit KE in step 19. This eigenvector is called the dominant eigenvector because it supplies the strongest and, as a rule, best contribution to the received signal. If only two eigenvectors are stored in the storage element SE and have been transmitted to the base station, one bit is sufficient for designating the eigenvector having the in each case greater eigenvalue. In consequence, if one bit is available per time slot for signaling the characteristics of reception back to the base station, the vector used for the beam shaping by the base station can be updated in each time slot and used for the beam shaping in the subsequent time slot.
If four eigenvalues have been transmitted to the base station, two bits are required for designating the in each case dominant eigenvector. If one bit per time slot is available for transmitting back the characteristics of reception, two time slots are, therefore, required for transmitting the complete designation of the dominant vector. As a consequence, it is used for the beam shaping for the two time slots following its transmission; in the course of these two time slots, the designation of the eigenvector to be used subsequently is transmitted.
The steps of the operating phase can be cyclically repeated many times before the initialization phase has to be carried out again for updating the coefficients of the eigenvectors.
The steps of the two phases are preferably carried out interleaved in time. This means: a short-term covariance matrix is used both for calculating a first spatial covariance matrix which will be transmitted from the subscriber station MS to the base station at a later time, and for calculating the eigenvalues by which the subscriber station selects the dominant one among the eigenvectors formed from an earlier first spatial covariance matrix and transmitted to the base station; in other words: steps 12 and 17 are a single step, the results of which are evaluated differently in steps 13 ff., on the one hand, and, on the other hand, 18 ff.
The eigenvectors can be transmitted at fixed predetermined time intervals from the subscriber station to the base station. However, a variant of the method described above in which the eigenvector analysis is repeated at the averaged covariance matrices (
If, at the beginning of a transmission link between subscriber station MS and base station BS, no averaged covariance matrix (
A second embodiment of the method is described with reference to
The spatial covariance matrices Rxx,a,i generated by the signal processor SP in step 11 are supplied to the computing unit RE which, as described above with reference to the first embodiment, forms from these short-term spatial covariance matrices Rxx,i in steps 12, 13, forms their moving mean value and performs in step 14 an eigenvalue and eigenvector analysis on the first long-term spatial covariance matrix (
These eigenvectors in each case correspond to transmission paths and contain the information about the relative phase angles of the corresponding contribution of the uplink signal at the individual antenna elements and thus about the direction from which the contribution is received. If the frequencies of uplink and downlink are the same as in a TDD radio communication system, the phase information contained in the eigenvector can be used directly for weighting the downlink signal. If the frequencies of uplink and downlink are different, it is necessary to convert the phase information contained in the eigenvector into a corresponding direction, using the uplink frequency as a basis, and to convert this direction back into phase information by the downlink frequency in order to obtain suitable eigenvectors for the beam shaping in the downlink.
The further steps of the method up to and including step 18 again correspond to those of the first embodiment, the difference being that the entire processing takes place at the base station BS and is based on measurements performed on the uplink signal, which is also why transmission step 16 is omitted.
In a first, simple variant of the second embodiment, the eigenvalue determination in step 18 is followed by a step 20 in which a current weighting vector w(k) is calculated by forming a linear combination of the stored eigenvectors w(k,1), w(k,2), . . . , each of the eigenvectors w(k,1), w(k,2), . . . being included in the linear combination with a weight dependent on its eigenvalue obtained in step 18 or its amount. Normalization of the linear combination is possible. This weighting in the formation of the linear combination ensures that those transmission paths having the best transmission characteristics in the short term dominate the downlink signal radiated by the base station. The other eigenvectors included in the current weighting vector w(k) are used for ensuring that a usable signal arrives at the subscriber station even in the case where the transmission path weighted most highly fails from one time slot to the next.
If one of the transmission paths between base station and subscriber station is a line-of-sight link, the eigenvector allocated to this transmission path can be directly used as current weighting vector w(k), in other words, all other eigenvectors are included in the formation of the linear combination with coefficients 0.
According to a further development of the second embodiment, space/time block coding can be performed on the downlink signal as a result of which, for example, two mutually orthogonal sequences with the same information content are obtained from an original transmitted symbol sequence. Each of these sequences is radiated by the base station weighted with one of the eigenvectors, the eigenvalues of which proved to be the largest in step 17.
The invention has been described in detail with particular reference to preferred embodiments thereof and examples, but it will be understood that variations and modifications can be effected within the spirit and scope of the invention.
Number | Date | Country | Kind |
---|---|---|---|
100 51 133 | Oct 2000 | DE | national |
Filing Document | Filing Date | Country | Kind | 371c Date |
---|---|---|---|---|
PCT/DE01/03923 | 10/12/2001 | WO | 00 | 4/16/2003 |
Publishing Document | Publishing Date | Country | Kind |
---|---|---|---|
WO02/33850 | 4/25/2002 | WO | A |
Number | Name | Date | Kind |
---|---|---|---|
5634199 | Gerlach et al. | May 1997 | A |
5886988 | Yun et al. | Mar 1999 | A |
5901075 | Offord et al. | May 1999 | A |
5999800 | Choi et al. | Dec 1999 | A |
6154661 | Goldburg | Nov 2000 | A |
6317612 | Farsakh | Nov 2001 | B1 |
20030144032 | Brunner et al. | Jul 2003 | A1 |
20050195915 | Raleigh et al. | Sep 2005 | A1 |
Number | Date | Country |
---|---|---|
19803188 | Jul 1999 | DE |
10032426 | Jan 2002 | DE |
0899896 | Mar 1999 | EP |
WO9843106 | Oct 1998 | WO |
Number | Date | Country | |
---|---|---|---|
20040014434 A1 | Jan 2004 | US |