DE10058336B4 - Method for beam shaping in systems with array antennas at transmitting and receiving stations - Google Patents

Abstract

Method for beamforming in systems with group antennas at transmitting and receiving stations (base and / or mobile stations), wherein 1 ≤ k ≤ K subscriber stations comprise a group antenna with M _k individual antennas, characterized in that
1.) the transmitter has knowledge of the long-term properties of the channel in the form of channel covariance matrices
2.) Beamform vectors w TS / q, k and w BS / q, k are selected from the dominant eigenvectors of the channel covariance matrix for the respective TS, each beamform vector allowing operation of a channel and are the number of available channels Q _k of the number gives the dominant eigenvalues of the channel covariance matrix;
3.) by dividing the total signal current into Q _k channels, the demultiplexed signals S BS / q, k (t) in the downlink and S TS / q, k (t) for the uplink result;
4) the transmission power is equally divided between the demultiplexed signals S _{q, k} (t);
5.) the demultiplexed signals S _{q, k} (t) are correlated with the same spreading code and modulated with the same modulation format;
6.) Q BS / k channels for the kth TS through ...

Description

The The invention is based on a method for beam shaping in systems with group antennas at transmitting and receiving stations of a mobile communication system considering the correlations of the mobile radio channel in space and time according to the generic term of claim 1.

The Method pursues the goal under realistic assumptions an increase to achieve the spectral efficiency of the communication system. The spectral efficiency is a measure of the ratio between the transmission rate achieved and bandwidth.

1.1 Technical environment

With Growing growth of communications markets with large customer growth and increasingly higher rate Data services increase the need for cost-effective and easy-to-implement communication systems with high spectral efficiency. On the one hand, WCDMA / UMTS systems apply as the dominant solution for communication systems the third generation, i.a. because high licensing costs for bandwidth for UMTS / WCDMA in many European countries have been paid. On the other hand, just because of the high License costs the spectral efficiency in WCDMA be high, affordable Offer high-data-rate services that have been promised to be able to. Increasingly in the field of vision for future and efficient communication systems therefore come from systems that have one increased Number of antennas (group antennas) first at the base station (BS) and then also at the subscriber station (TS).

The maximum achievable data rate, with the digital cordless communication systems can operate is determined by the so-called channel capacity described by Shannon [1]. The Channel capacity formulated the theoretical upper limit of the amount of data transmitted over a channel can be. Central parameters for describing a channel and thus the maximum channel capacity Among other things, this is the signal-to-noise power available relationship (SNR) at the receiver, the number of transmit antennas, the number of receive antennas, the Bandwidth of the mobile radio channel and the statistics of both the transmission coefficients from the transmitter to the receiver, as well as the transmission signal.

The increase in the number of antennas for the purpose of capacity increase can be done in two stages. In the first stage you equip the BS with group antennas. In the second stage, the TS, especially cars, laptops, PDAs (Personal Digital Assistants), and finally cell phones will also be equipped with group antennas. Each step is associated with an increase in capacity. In the following we focus on algorithms for M _BS > 1 antenna elements at the BS and M _TS ≥ 1 elements at the TS.

• Ein gegenwärtiger Vorschlag zur Erzielung von hohen Datenraten in Funkkanälen basiert auf einer gesteigerten Kanalkapazität durch Einrichtung von virtuellen, unabhängigen Sub-Kanälen in Mehrantennensystemen basierend auf einer Schätzung der Kanalkoeffizienten und einer anschließenden Singulärwertzerlegung (SVD) dieser Kanalmatrix. Zur Einrichtung dieser virtuellen Subkanäle werden die realen Daten mit den unitären Matrizen aus der SVD transformiert. Die Zahl der dominanten Singulärwerte bestimmen die Anzahl der möglichen virtuellen Sub-Kanäle. Desweiteren kann die Leistungsverteilung innerhalb der Subkanäle anhand der Singulärwerte bestimmt werden [2]. Eine alternative Vorgehensweise dieser Idee besteht darin, die existierenden Sub-Kanäle in der Empfangsrichtung in ihren Richtungen ("Direction of Arrival" (DOA)) zu schätzen und diese dann in Abstrahlrichtung wieder zu bedienen [3]. Diese Vorgehensweise beruht auf einer geometrischen Vorstellung und Interpretation der Wellenausbreitung im Kanal.
• Ein weiterer Vorschlag zur Steigerung der Datenrate und der spektralen Effizienz wurde im Rahmen der WCDMA/3GPP Standardisierung eingebracht. Weil in einem Frequency-Division Duplex (FDD) System wie WCDMA durch die eingeschränkte Reziprozität des Kanals Kanalinformationen in Abstrahlrichtung nur eingeschränkt vorliegen, werden unterschiedliche Datenströme ungewichtet auf die Antennenelemente gegeben [4]. (Aufgrund der eingeschränkten Reziprozität lassen sich nur langfristige, d.h. über den schnellen Schwund gemittelte Kanaleigenschaften von Aufwärts- auf Abwärtsverbindung (und umgekehrt) anwenden.)

In essence, the prior art is characterized by two proposals, which are briefly described here:

• A current proposal for achieving high data rates in radio channels is based on increased channel capacity by establishing virtual, independent subchannels in multiple antenna systems based on an estimate of the channel coefficients and a subsequent singular value decomposition (SVD) of this channel matrix. To set up these virtual subchannels, the real data is transformed with the unitary matrices from the SVD. The number of dominant singular values determines the number of possible virtual sub-channels. Furthermore, the power distribution within the subchannels can be determined from the singular values [2]. An alternative approach of this idea is to estimate the existing sub-channels in the direction of arrival in their directions ("Direction of Arrival" (DOA)) and then to operate them again in the emission direction [3]. This approach is based on a geometric conception and interpretation of wave propagation in the channel.
• Another proposal for increasing data rate and spectral efficiency was introduced as part of WCDMA / 3GPP standardization. Because in a frequency-division duplex (FDD) system such as WCDMA channel information in the radiation direction are limited due to the limited reciprocity of the channel, different data streams are given unweighted on the antenna elements [4]. (Due to the limited reciprocity, only long-term, ie fast-averaged channel characteristics from uplink to downlink (and vice versa) can be used.)

1.3 Defects of previously known designs

Of the first proposal [2] has the disadvantage that extensive channel information not only at the receiver, but also on the transmitter must be known. Due to the limited reciprocity in one FDD system requires the replacement of part of the channel information. This is problematic because the feedback rate is limited in WCDMA is. The needed Amount of feedback would be especially with variable Huge channels, which is illustrated by the following numerical example: 170 antennas per page, 30kHz bandwidth and 20dB SNR results after Shannon [1] a data rate of about 20Mb / s [2]. However, this is one Data amount of at least 170 x 170 x 2 x 8 bits = 0.46Mb per retransmission the channel information needed to the transmitter. Is the channel not constant? and will be a retransmission Required per slot at 1500 timeslot / s, the feedback is 693.6 Mb / s.

Of the Disadvantage of the DOA-based method is that this based on a geometrically interpreted channel model, the the reality does not exactly replicate or reproduce is.

Of the second proposal [4] has the disadvantage that by non-directional blasting Interference at the other TS inside and outside the cell is generated, because in particular in WCDMA the spreading codes used for frequency-selective channels in the downlink are not orthogonal. This is especially critical here because CDMA Systems are generally interference limited. Thus, both the channel estimation as well as the data detection critical.

1.4 Provided technical Problem, motivation

• • FDD System, dadurch eingeschränkte Reziprozität. Zusätzlich geringe, beziehungsweise keine Feedbackrate. D.h., der Kanal ist in Senderichtung nur eingeschränkt bekannt.
• • Der Kanal ist in der Regel frequenzselektiv.
• •In der Abwärtsverbindung stören die Signale für einen Teilnehmer die jeweils anderen Teilnehmerstationen und erzeugen dort Interferenz. Falls die Teilnehmer in der gleichen Zelle sind, ist das Intrazellinterferenz, sonst Interzellinterferenz (aufgrund der Korrelationseigenschaften der verwendeten Codes tritt bei flachen Kanälen keine Intrazellinterferenz auf). In der Aufwärtsverbindung tritt auch bei flachen Kanälen Intrazellinterferenz auf.

The aim of the method is beamforming with array antennas at the BS and / or TS to increase the spectral efficiency. The disadvantages mentioned in section 1.3 should be avoided. The assumed framework conditions in WCDMA are the following:

• • FDD system, thus limited reciprocity. Additionally low or no feedback rate. That is, the channel is limited known in the transmission direction.
• • The channel is usually frequency selective.
• • In the downlink, the signals for one subscriber interfere with the other subscriber stations and generate interference there. If the participants are in the same cell, this is intracellular interference, otherwise intercell interference (due to the correlation properties of the codes used, there is no intracellular interference in shallow channels). In the uplink, even in shallow channels, intracellular interference occurs.

1.5 Solution proposal

Reached The object is achieved by the in the characterizing part of claim 1 specified characteristics. under claims FIGS. 2 to 8 relate to embodiments of the subject matter of the claim 1.

in the The following is assumed to be the channel estimation the TS (in the downlink) based on common (com) pilot sequences (training sequences) carried out which is radiated omnidirectionally from each BS antenna element be used by all TS in the cell. Furthermore it is assumed that information of the downstream channel sent to the BS via feedback become. The TS sends to the BS (in the uplink) with them assigned (dedicated) pilot into the data channels and thus enables a channel estimation the BS.

alternative it is possible, to give up the feedback. In addition, there is a possibility, too from the BS with assigned pilots to the data channels. This will be discussed in section 1.5.2.

1.5.1 Concept and Embodiment

Based on 1 - 3 an embodiment will be explained.

downlink

• 1. The BS radiates different (common) Pilot sequences omnidirectionally from each BS antenna element from [5, 6].
• 2. The TS estimates the channel in the downlink in space and time by correlation with the known (common) pilot sequencers. It describes
  
  the current, short-term (= instantaneous) channel estimation on the TS antenna element for the tap n at the time slot i as a result of a correlation based on the pilots emitted at the BS and known at the TS. M _BS is the number of BS antennas. • A conventional rake searcher finds the long-term dominant samples / taps (rake finger). Since the antenna elements are close to the TS, the narrowband approximation [7] applies. The temporally dominant samples are therefore the same for all antenna elements on the TS. • For channels with constant macroscopic properties, the TS becomes the long-term spatial covariance matrices
  
  for all TS antenna elements, 1 ≤ m ≤ M _{TS formed} by expected time slots i and averaging over the dominant time taps n. There are two alternative approaches: the first one is the macroscopic properties at the TS in the form of the long term spatial Covariance matrices per tap n to be determined as R BS / S, n, m. The second approach is to capture the entire temporal structure in a covariance matrix (space-time covariance matrix) R BS / ST, m. The further procedure with the respective covariance matrices is adjusted accordingly compared to the case of the averaging over the temporal taps n (see text). • Just as the channel's long-term temporal characteristics are equal at the TS antennas, so too are the long-term spatial characteristics between the BS array antenna and the TS antennas due to the narrow band assumption. Therefore:
  
  It can therefore be averaged over the covariance matrices of the TS antenna elements to increase the estimation accuracy and speed up the convergence. In the following, the index m is therefore omitted for long-term variables. Since the macroscopic / long-term channel properties change slowly over time, a forgetting factor p is typically used to track the long-term covariance matrix R SB / S over the time slots i:
  
  • From the long-term covariance matrices of (1) and (3), the Q ^BS dominant eigenvectors w BS / q are determined. These are used at the BS in the form of beamforming vectors to service the spatial multipath structure of the channel from the BS to the TS with multiplexed data streams s _q (t) (see point 4). • The current channel coefficients of the data channels for each TS antenna element correspond
  
  where the channel matrix
  
  describes the transmission coefficients from the BS antenna elements to the TS antenna element m at the taps 1 ≦ n ≦ N. In the case of [4], the channel estimation in the TS is simpler since both the orthogonal (common) pilots and the multiplexed data streams from each antenna element are radiated at the BS. • Normally spatially uncorrelated antenna elements at the TS can be assumed [8]. If this is not the case, the correlations of these current channel coefficients (4) can be exploited in order to obtain an improved channel estimation and lower computation effort according to the principle as described in [9] for receiver structures (eigenrake concept).
• 3. The determined beamforming vectors w BS / q are determined by feedback in the uplink sent to the BS. Feedback can basically consist of long-term and short-term feedback [10]. Because the feedback rate in WCDMA limited is on 1bit / time slot, short-term feedback seems only at a lot low speeds, if any, makes sense. Last Minute Feedback is usually not necessary because of several Antennas on the receiver for sufficient spatial diversity especially when the channel is frequency-selective, or more receive antennas exist as beamforming vectors. Thus, only long-term feedback should be used only from the slowly changing one macroscopic arrangement in the channel is affected
• 4. By the feedback of the beamforming vectors w BS / q, 1 ≤ q ≤ Q ^BS , the BS knows the long-term dominant characteristics of the channel in the downlink. Different signals s _q (t), which consist of different symbol streams which are modulated onto the same spreading code, are emitted by the BS via the Q ^BS different beam-shaping vectors (eigenbeams). Alternatively, the same symbols may be radiated over the channels opened by the beamforming vectors. These should then be made separable at the receiver so that they can be advantageously combined to compensate for the rapid fading (diversity gain) by using different spreading codes or space-time block codes with the same spreading code [4]. The radiation power P _{q of} the beamforming vectors is a function of the mean attenuation of the eigenbeam by the channel, which is characterized by the corresponding eigenvalue. One possibility here is, for example, the so-called water filling [11], or a uniform distribution: the transmission power. In a frequency-flat scenario, the number of beamforming vectors on which to radiate should not exceed the number of TS antennas to ensure unambiguous detection. Therefore, the number of beamforming vectors depends on both the multipath structure of the channel and the number of TS antennas.
• 5. The detector on the TS has two options. On the one hand, the joint detection of all Q ^BS data streams s _q (t), which are intended for a TS, or on the other an iterative approach according to the BLAST principle [4]. • Joint detection will produce better results if the data streams of similar power reach the receiver. The disadvantage is the increased complexity. • The iterative procedure according to the BLAST principle is less computationally intensive and still powerful if the data streams at the receiver are damped to different degrees. In the sense of error propagation it is expedient to start with the strongest self-crew. The strongest data stream in the ith time slot results
  
  the channel estimate for the data channel is from (4) and
  
  the Frobenius norm is calculated. In the last step, the channel estimate H ded / q (i) according to (6) is applied to the despread data signal of the q th eigenbeam in the space and time domain at the TS about the decision variable, for example according to the Maximum Ratio Combining (MRC) principle to obtain. This means that the data stream of the qth eigenbeam is detected with an MRC space-time rake.

uplink

• 1. If the antenna elements on the TS are spatially uncorrelated, only processing of the signals for the uplink remains [4]. However, if the TS antenna elements are spatially correlated, weight vectors can be radiated in the upward direction from the TS to the BS: • It becomes the long-term spatial covariance matrix
  
  and h com, BS / m, n (i) describe the channel estimate for the pilots from the BS (see above). • This covariance matrix
  
  will be added to the uplink in the next step
  
  converted to take into account the frequency offset [12]. This must be based on calibrated antenna elements on the TS [12]. The Q ^TS dominant eigenvectors w TS / q of the converted spatial covariance matrix R TS / S form the beamforming vectors at the TS for the uplink. In each dominant eigenbeam w TS / q, multiplexed data streams are broadcast with different pilots assigned to the TS.
• 2. By the different assigned pilots in each Eigenbeam of the TS are the different Eigenbeams inclusive contained therein data stream at the BS separable. Through a correlation with For the pilots assigned to the OS, the BS estimates the data channel each Eigenbeams in the uplink from the TS.
• 3. The channel estimates of the Q ^TS eigenbeams now detect the individual data channels from the uplink. A short illustration of this concept gives 1

1.5.2 Extensions

In Section 1.5.1 was a concept for data transfer between BS and TS described on the assumption that common pilot sequences and feedback are available.

• 1. Abwärtsverbindung mit gemeinsamen Piloten ohne Feedback: Damit die TS die Datenkanäle mit Hilfe des gemeinsamen Piloten schätzen kann nach (4), müssen TS und BS von den gleichen Strahlformungsvektoren an der BS ausgehen. Um an der BS zu den gleichen Strahlformungsvektoren zu kommen, muß an der BS die langfristige Kanalschätzung für die Abwärtsverbindung aus der Aufwärtsverbindung gewonnen werden. Hierzu muß das System kalibriert sein und der vorhandene Frequenzversatz korrigiert werden [12]. In diesem Fall mittelt die BS über die Q^TS räumlichen Kovarianzmatrizen, die nach der Korrelation mit den zugewiesenen Piloten von der TS entstehen. Eine Zusammenfassung dieses Konzeptes gibt 2.
• 2. Abwärtsverbindung mit zugewiesenen Piloten ohne Feedback: Die BS erhält eine Schätzung der Abwärtsverbindung, indem sie mit den bekannten, zugewiesenen Piloten von der TS korreliert. Durch Korrektur des Frequenzversatzes nach [12] erhält die BS eine Schätzung für die Abwärtsverbindung. Die Q^BS dominanten Eigenvektoren dieser langfristigen Kanalschätzung bilden die Strahlformungsvektoren. Der gesamte Datenstrom wird auf alle dominanten Eigenbeam gemultiplext. Zusätzlich wird über jeden Eigenbeam ein anderer zugewiesener Pilot gesendet. Auf der Basis der zugewiesenen Piloten erhält die TS sofort eine Schätzung des Datenkanals in der Abwärtsverbindung H ded / q(i) entsprechend (6). Anhand dieser Schätzung des Datenkanals kann die TS die Detektion der einzelnen, gemultiplexten Datenströme durchführen. Nachdem an mindestens einer Seite der Datenverbindung der gesamte Kanal vermessen werden muß, um die Bildung neuer Strahlformungsvektoren zu ermöglichen, ist es notwendig, in der Aufwärtsverbindung omnidirektional abzustrahlen. Alternativ besteht auch die Möglichkeit an beiden Seiten gerichtet abzustrahlen und erst beim Eintreten bestimmter Ereignisse von der TS oder von der BS für eine Zeitdauer ungerichtet abzustrahlen, um so den Kanal an der Gegenseite neu zu vermessen. Typische Beispiele für diese Ereignisse wären z.B. das Verstreichen eine vorgegebenen Zeitspanne, oder die Verschlechterung der Übertragungsstrecke, so daß die Datenübertragung zu der gewünschten Datenrate nicht mehr möglich ist. Eine Zusammenfassung dieses Konzeptes gibt 3.

Two alternatives are conceivable:

• 1. Common Pilot Downlink with No Feedback: In order for the TS to estimate the data channels using the common pilot (4), TS and BS must start from the same beamforming vectors at the BS. In order to arrive at the same beamforming vectors at the BS, the long-term downlink channel estimation must be obtained at the BS. For this, the system must be calibrated and the existing frequency offset corrected [12]. In this case, the BS averages the Q ^TS spatial covariance matrices that arise after correlation with the assigned pilots from the TS. A summary of this concept gives 2 ,
• 2. Downlink with Assigned Pilots without Feedback: The BS receives an estimate of the downlink by correlating with the known, assigned pilots from the TS. By correcting the frequency offset according to [12], the BS gets an estimate for the downlink. The Q ^BS dominant eigenvectors of this long-term channel estimation form the beamforming vectors. The entire data stream is multiplexed to all dominant eigenbeams. In addition, a different assigned pilot is sent over each self-crew. On the basis of the assigned pilots, the TS immediately gets an estimate of the data channel in the downlink H ded / q (i) according to (6). Based on this estimation of the data channel, the TS can perform the detection of the individual, multiplexed data streams. Since the entire channel must be measured on at least one side of the data link to allow the formation of new beamforming vectors, it is necessary to omnidirectionally radiate in the uplink. Alternatively, it is also possible to radiate directed on both sides and only be on entering be to broadcast non-directional events from the TS or BS for a period of time so as to re-measure the channel at the far end. Typical examples of these events would be, for example, the elapse of a predetermined period of time or the degradation of the transmission link so that data transmission at the desired data rate is no longer possible. A summary of this concept gives 3 ,

1.6 Achieved benefits

• • in einer verbesserten Störunterdrückung durch direktionales Abstrahlen der Daten. Andere TS erhalten weniger Interferenz. Dies führt zu einer verbesserten Kanalschätzung und Datendetektion. Die Datendetektion profitiert dabei doppelt; zum einen durch die genauere Kanalschätzung und zum anderen durch die geringere Interferenz.
• • in einer verbesserten Kanalschätzung in der TS durch eine akkurate Kenntnis der langfristigen Strahlformungsvektoren. Weil für die langfristige Kovarianzmatrix in (1), bzw. (3) oft gemittelt wird, sind die entsprechenden Strahlformungsvektoren exakt bekannt.
• • in einem einheitlichen Konzept für unterschiedliche Kanalsituationen zwischen BS und TS. Beispielsweise wird bei starker räumlicher Korrelation nur ein Eigenbeam ausgeprägt, wohingegen bei schwacher räumlicher Korrelation mehrere gebildet werden. Bei der Berechnung der Eigenbeams wird die Frequenzselektivität des Kanals ebenfalls berücksichtigt.
• • dadurch, daß in CDMA Systemen mit nicht-orthogonaler Teilnehmertrennung durch Störunterdrückung noch große Gewinne für M_BS > M_TS erzielt werden, wenn der Kanal räumlich korreliert ist. Typischerweise sind CDMA Systeme interferenzlimitiert. In Systemen mit orthogonaler Teilnehmertrennung, z.B. in OFDM, wäre die Kapazität im wesentlichen durch min(M_TS, M_BS) beschränkt. Hintergrund ist, daß es wesentlich wirtschaftlicher ist, die Anzahl der Antennenelemente an der BS als an den TS zu steigern.

The advantages of the above concept compared to the prior art are the utilization of the long-term channel characteristics, in particular

• • in an improved interference suppression by directional radiation of the data. Other TS receive less interference. This leads to improved channel estimation and data detection. Data detection benefits twice; on the one hand by the more accurate channel estimation and on the other hand by the lower interference.
• In an improved channel estimation in the TS by an accurate knowledge of the long-term beamforming vectors. Because the long-term covariance matrix in (1) or (3) is often averaged, the corresponding beamforming vectors are known exactly.
• • in a uniform concept for different channel situations between BS and TS. For example, only one eigenbeam is pronounced in the case of strong spatial correlation, whereas several are formed if the spatial correlation is weak. When calculating the eigenbeams, the frequency selectivity of the channel is also taken into account.
• • in that in CDMA systems with non-orthogonal subscriber separation due to interference suppression, large gains are still achieved for M _BS > M _TS , when the channel is spatially correlated. Typically, CDMA systems are interference limited. In systems with orthogonal subscriber separation, eg in OFDM, the capacitance would be essentially limited by min (M _TS , M _BS ). The background is that it is much more economical to increase the number of antenna elements on the BS than on the TS.

literature

• [1] C. Shannon, "A mathematical theory of communication, "Bell Syst. Tech. Journal, vol. 27, pp. 379-423 and 623-656, July 1948.
• [2] B. Hochwald and T. Marzetta, "Multiple antenna communication system and method thereof, # EP 0905 920 A2 "Patent, 1998. Lucent.
• [3] H. Boche, M. Bronzel, and T. Kuhwald, "Method and Apparatus for Generating Predetermined Directional Characteristics," 1998. Patent, Serial No. DE 198 58 951 A1 ,
• [4] Lucent, "Preliminary link level results for HSPDA using multiple antennas. "TSG-RAN Working Group 1 meeting # 16, TSGRi # 16 (00) 1218, Korea, Oct. 2000th
• [5] "Third Generation Partnership Project (3GPP), 3G TS 25.211, "2000. www.3gpp.org.
• [6] "Third Generation Partnership Project (3GPP), 3G TS 25.214, "2000. www.3gpp.org.
• [7] D.H. Johnson and D.E. Dudgeon, Array Signal Processing: Concepts and Techniques. Englewood Cliffs, NJ: Prentice Hall, 1993.
• [8] J. Fuhl, J.P. Rossi, and E. Bonek, "High-resolution 3D direction-of-arrival determination for urban mobile radio, "IEEE Trans. Antennas and Propagation, vol. 45, pp. 672-682, Apr. 1997.
• [9] C. Brunner, J.S. Hammerschmidt, A. Seeger, and J.A. Nossek, "Space-time eigenrake and downlink eigenbeamformer: Exploiting long-term and short-term channel properties in WCDMA, "in Proc. IEEE GLOBECOM, (San Francisco, CA), Nov. 2000.
• [10] Siemens, "Advanced closed loop tx diversity concept (eigenbeamformer). "TSG-R WG1 document, TSGRi # 14 (00) 0853, Oulu, Finland, July 2000. www.3gpp.org.
• [11] R.G. Gallager, Information Theory and Reliable Communication. John Wiley & Sons, 1968th
• [12] T. Aste, P. Forster, and L. Fety, "Downlink beamforming avoiding DOA estimation for cellular mobile communications, "in Proc. IEEE International Conf. Acoust., Speech, Signal Processing, (Seattle (USA)), 1998.

Claims (8)

Method for beamforming in systems with group antennas at transmitting and receiving stations (base stations and / or mobile stations), where 1 ≤ k ≤ K subscriber stations having a group antenna with M _k individual antennas, characterized in that 1.) the transmitter is aware of the long-term properties of the Channels in the form of channel covariance matrices be 2.) beamform vectors w TS / q, k and w BS / q, k are selected from the dominant eigenvectors of the channel covariance matrix for the respective TS, each beamform vector allowing operation of a channel and are the number of available channels Q _k from the Number of dominant eigenvalues of the channel covariance matrix gives; 3.) by dividing the total signal current into Q _k channels, the demultiplexed signals S BS / q, k (t) in the downlink and S TS / q, k (t) for the uplink result; 4) the transmission power is equally divided between the demultiplexed signals S _{q, k} (t); 5.) the demultiplexed signals S _{q, k} (t) are correlated with the same spreading code and modulated with the same modulation format; 6.) Q BS / k channels for the k-th TS are established by beamforming at the BS with different beamform vectors w BS / q, k, 1 ≤ q ≤ Q BS / k at the transmission side in the downlink direction and these channels with (different ) Signals Q BS / q, k (t), and Q TS / k channels from the k th TS by beamforming with different beamform vectors w TS / q, k, 1 ≤ q ≤ Q TS / k at the transmission side in the uplink direction and these channels are operated with (different) signals Q TS / q, k (t). Method according to claim 1, characterized in that that the Beamform vectors w TS / q, k and w BS / q, k result from a base change in the subspace defined by the dominant eigenvectors of the channel covariance matrix for the respective TS. Method according to Claim 1, characterized in that the same total signal current is applied to Q _k signals S _{q, k} (t), the Q _k signals S _{q, k} (t) being separated by different space-time block codes, so that these can be separated and advantageously combined at the receiver (diversity gain). Method according to Claim 1, characterized in that the multiplexed signals s _{q, k} (t) are correlated with different spreading codes and / or modulated at different modulation stages. A method according to claim 1, characterized in that the beamforming vectors w _{q, k} , 1 ≤ q ≤ Q _{k are} determined on the basis of the measured long-term channel characteristics of the receiving link (at the BS and / or at the TS _k ). A method according to claim 1, characterized in that the beamforming vectors w _{q, k} , 1 ≤ q ≤ Q _{k are} determined on the basis of the measured and frequency offset adjusted long-term channel characteristics of the receiving link (at the BS and / or at the TS _k ). Method according to Claim 1, characterized in that the (long-term) beamforming vectors w BS / q, k, 1 ≤ q ≤ Q _{k are} determined by channel estimation of the common BS-radiated pilots at the k th TS and by means of a back channel of the BS be communicated. Method according to the above claims, characterized in that the demultiplexed signals s _{q, k} (t) are provided with a power which is a function of the dominant eigenvalues of the channel covariance matrix in terms of power control.