Method for improving the performance of multiple input multiple output transmit diversity using feedback
Field of the Technology
The present invention is related to the field of wireless communication, and more specifically to an improved method for reducing interference without degrading the performance in a multiple input multiple output (MIMO) wireless communication system utilizing transmit diversity.
Background of the Invention
In wireless communication systems, such as for example IEEE 802.16 broadband wireless access (BWA) system or the 3rd generation wireless wideband code division multiple access (W-CDMA), the use of multiple (or smart) antennas is much preferred in order to increase the performance of the system. The basic idea is to improve the performance of the system by using multiple antennas in a transmitting end (e.g. node B or base station), and focusing the transmitted energy from the transmitter by appropriate phasing and power distribution over the transmitter antennas, and by phasing and weighting the receiver antenna signals. The antenna coverage patterns may thus be customized for specific traffic conditions, and the use of smart antennas thereby provide flexibility by enabling network operators to change antenna patterns to adjust to changing traffic and prevailing radio frequency conditions.
In the above referred-to kind of wireless communication systems, transmit diversity is adopted to mitigate the fading channel. This transmit diversity is commonly performed by using the so called Alamouti code, where every group of two symbols of data to be transmitted are jointly space time encoded and transmitted from two transmitter antennas over two symbol time intervals. The Alamouti code is one realization of the so called space time transmit diversity (STTD) encoding.
In order to increase the data throughput over a wireless channel, the number of antennas at the transmitter can be increased to four, thereby enabling the transmission of two parallel Alamouti encoded data streams. This is called double space time transmit diversity (D-STTD). While increasing the data throughput, the D-STTD scheme also increases the complexity of the receiver, since the two STTD encoded data streams will interfere with each other. In order to reduce the mutual interference, a linear interference suppression receiver is employed which is adapted according to the minimum mean square error (MMSE) criterion.
In view of increasing the performance of D-STTD further, a feedback channel can be utilized for having the receiver give some information to the transmitter about the current channel state. The feedback information is used by the transmitter to adapt the transmission and thereby increasing the signal to noise ratio in the receiver. This gives a closed loop D-STTD scheme.
Ideally, in a frequency division duplex (FDD) system, where the channel between a transmitter and a receiver is unknown to the transmitter, the receiver should inform the transmitter about the complete channels between all the transmitter-receiver antenna pairs. However, to convey such detailed knowledge of the channels entails a large amount of information to be transmitted over the feedback channel, and is thus very bandwidth consuming. Further, in mobile communication, one of the communicating parties is most often moving at some speed, and since the feedback information sent in for example a 3r generation system is conveyed one bit at a time, so at the time when a transmission of large amounts of information from such party reaches the other communicating party, the information may already be obsolete. It is therefore difficult to provide a transmitter with full channel knowledge. Consequently, in order to assure a high spectral efficiency of the system, it is of high importance to keep the feedback data rate low, while still giving the highest possible benefits of the feedback information. Hence, what channel information to select for transmission on the feedback channel is of great importance.
Instead of feeding back the complete channel information, the feedback information is used to set complex valued weights wj, Vt2, W3, W4 on the antenna signals that are to be transmitted, see Figure 1. The optimal weights will depend on the channels between the transmitter and receiver antennas and must be conveyed to the transmitter by using the feedback channel.
In a document entitled "Double-ASTTD with Sub-group Rate Control", by Huawei, authors Mattias Wennstrδm and Branislav M. Popovic, Montreal, Canada, May 10-14, 2004, an improvement of the performances of D-STTD is described. In accordance with the discussed solution, it is proposed that instead of adapting the phase of the transmitted signals, the signal amplitudes are adapted, that is, the weights in Figure 1 are real valued. This scheme increases the throughput considerably compared to an open loop D-STTD. However, this method emanates from a solution for STTD, in which there is only one antenna branch and thus no mutual interference between different data streams to consider, and although this prior art solution may be used also for D-SSTD, it is in no way optimised for such use.
It would thus be desirable to provide a method giving an improved trade-off between the channel information and the system resource utilization, and which method also considers the mutual interference between data streams of different antenna branches in a transmitting end.
Summary of the Invention
It is an object of the present invention to provide a method that reduces the amount of feedback conveyed, without adversely affecting the performance of the system, while simultaneously taking into account the mutual interference between different transmission data streams.
This object is achieved, according to a first aspect of the invention, by a method as defined in claim 1.
In accordance with the present invention, a method is provided, providing closed- loop transmit diversity in a wireless communication system, whereby the method comprises the step of receiving feedback information about a channel state in a transmitting end in the wireless communication system from at least one receiving end. The transmitting end comprises at least a number of antennas, e.g. three or four antennas, providing for at least one Alamouti encoded data substream. The method comprises the step of conveying, at the receiving end, feedback information containing the complex valued weights of a subset of said antennas, thereby reducing the bandwidth used for feeding back information from said receiving end. The method in accordance with the present invention does in fact show a comparable system performance as a method using only complex valued weights. Thereby the present invention provides a method overcoming the disadvantages of the prior art, accomplishing corresponding, or in one embodiment the same, system performance at a lesser feedback rate.
In accordance with one embodiment of the present invention, the feedback information contains the complex valued weight of only one antenna. The feedback information may thereby be chosen so as to reduce the bandwidth used for feeding back information even further.
In accordance with one embodiment of the present invention, the complex valued weight that maximizes the signal to noise ratio SNR in the receiving end is chosen. The method in accordance with the present invention is thus flexible in that it is not necessary, but preferred, to optimise the SNR. The phase weight chosen need not maximise the SNR.
In accordance with another embodiment of the present invention, a linear minimum mean square error (MMSE) receiver is used, and wherein the calculation is performed by maximizing the signal to noise ratio, SNR, in the expressions:
SNRm +K|2|A21|2 +KI2IAnI2 +H>22|2)-2£l Re{Λ} (∑)
SNR
B2
+KI
2IA
14I
2 +K|
2|A
24|
2)-2*
2 Re{Λ}
(π)
where
where
and where
is the phase of
and
This embodiment thus gives general expressions for calculating complex weights that minimizes the mutual interference between different Alamouti data streams, maximizing the signal to noise ratio for the respective substreams, while the bandwidth used for feedback information is reduced.
In accordance with another embodiment of the present invention, the complex antenna weight of one antenna and the real-valued amplitude weights of the remaining antennas are fed to the transmitting end. This embodiment provides a means for accounting for all the different amplitudes involved, thereby giving a means for balancing the antennas (data streams) in any desired way. It may be shown that this embodiment of the present invention presents exactly the same system performance as a method using the only complex valued antenna weights.
In accordance with another embodiment of the present invention, the phase weight of one of the antennas and the amplitude weights of the remaining antennas are fed to the transmitting end. This embodiment gives a slightly improved bandwidth reduction as
compared to the previous embodiment, still without any considerable loss of system performance.
In accordance with another embodiment of the present invention, the amplitude weights are set to be constant, whereby the feedback information contains only the phase weight of any one of the antennas. This embodiment gives further yet reduction of the utilized bandwidth, without essentially degrading the system performance, but without a means for adjusting for the amplitudes. This embodiment thus provides a flexible solution, where systems or environments where the adjustment of amplitudes is not essential are able to further reduce the bandwidth for feedback information without much degradation of system performance.
In accordance with yet another embodiment of the invention, a linear minimum mean square error (MMSE) receiver is used, n = 2 receivers is assumed, the amplitudes of the antennas are set to constant values, and the calculation is performed by maximizing the signal to noise ratio, SNR, in the expressions now being reduced to:
SNR
2{θ) = gξ(β
ι g? + g
2g« + σ
2l)
'lg
3 (1)
where H denotes Hermitian transpose and the vectors g, are defined as the columns in the following matrix
whereby the phase θ maximizing 57Vi?; and SNR
∑ streams simultaneously for n = 2 receive antennas is calculated as
,
<.-.
where
and where the phases
are defined as
where
denotes the phase of the channel
.This embodiment of the present invention gives explicitly the optimal phase weights of the antennas in phase weighted D- STTD in a flat fading channel for n = 2 antennas. Thereby an easily implemented method is provided for reducing the bandwidth being used for feedback, while maintaining essentially the same system performance.
In accordance with yet another embodiment of the invention, a four antenna system is used in which the feedback information contains the complex valued weight of only one antenna. In this embodiment, the amplitude weights { cl ,c2,ci ,cii ) of the four antennas correspond substantially to one or more from the values in the group consisting
of: 1, . Further, the phase in the phase weight e
jθ correspond
substantially to any from the group:
This has the advantage that a codebook which shows a significant performance gain as compared to the existing codebook in the IEEE standard 802.16 can be achieved when using Space Time Code matrix A or B plus antenna grouping in a 4 BS antenna system.
The amplitude weights (c
1 ,c
2,c
3 ,c
ii ) of the four antennas may be selected such that C
1 2 +
= 4 . In this way, the resulting total radiated power is invariant irrespective of the selection.
The invention also relates to a receiver and transmitter, respectively, including means for performing the improved feedback information method. The invention further relates to a wireless communication system utilizing said improved method and comprising at least on of said receivers. The corresponding advantages, described above are achieved by a receiver as claimed in claim 14, a transmitter as claimed in claim 26 and a system as claimed in claim 39.
In yet another embodiment of the present invention, in which the system comprises at least three transmitter antennas, the feedback information contains the amplitude weights of said antennas, and the amplitude weights correspond substantially to one or
more of the values in the group consisting of: 1, J — and J — . This embodiment
enables a IEEE 802.16 codebook which considerably increases the performance for 3 BS antenna systems when Space Time Code matrix A or B plus antenna grouping is used. As above, the amplitude weights ( cl 5 c2,c3 ) of the antennas may be chosen such that + cl + cl = 3 in order to achieve radiated power invariance.
The invention also relates to a system comprising at least three transmitter antennas, wherein the feedback information contains the amplitude weights of said antennas. The advantages of such a system are achieved by a method as claimed in claim 39, a receiver as claimed in claim 41, and transmitter as claimed in claim 43.
Brief Description of the Drawings
Figure 1 shows schematically antenna weights W1, W2, W3 and w4 multiplied with antenna signals to be transmitted.
Figure 2 shows schematically a system architecture for two Alamouti encoded substreams and two receiver antennas, in which system the present invention may be utilized.
Figure 3 shows a diagram comparing the present method to two prior art methods.
Embodiments of the Invention
The Alamouti code is one realization of the so called space time transmit diversity (STTD) encoding, and the present application is concerned with such Alamouti coding. In order to facilitate the understanding of the present invention, a brief description of this encoding is given in the following. In short, Alamouti coding provides a way to accomplish a two-branch transmit diversity with only one receiver. Two signals, S1 and s2; are transmitted simultaneously from the two antennas ai and a2, respectively, during a first symbol period. During the following symbol period, signal (- S1 * ) is sent from antenna a2, and signal s2 * is sent from antenna ai, where * denotes the complex conjugates. The use of Alamouti encoding provides enhanced performance in terms of bandwidth and diversity.
With a view to increasing the performance of a D-STTD even further, a feedback channel can be utilized. Since the prior art method described in the introductory part emanates from a solution for STTD, in which there is only one antenna branch sending one data stream, and thus no mutual interference between different data streams to consider, it is not crucially important to convey phase information from the receiving end to the transmitting end, and was therefore not considered in this prior art solution. The inventors of the present invention have identified this short-coming of the prior art, and an inventive method will be described presenting a way to optimally choose what channel information to convey to the transmitting end from the receiving end.
Figure 1 shows schematically antenna weights W1, W2, W3 and W4 multiplied with four antenna signals to be transmitted. In short, to improve the performance of closed loop D-STTD, i.e. a system with feedback, the method of the present invention allows, in
one embodiment, the use a complex valued weight of any one (which could be arbitrarily chosen) of the four antennas in Figure 1, and to use real valued weights of the three remaining ones. This gives a significant reduction of the amount of information that has to be fed back from the receiving end to the transmitting end, thus saving bandwidth used. Further, it will be shown that this reduction may be done without any substantial loss of signal to noise ratio, SNR, in the receiving end, if a minimum mean square error (MMSE) receiver is adopted. In the description, a subset of complex valued weights of the transmitter antennas used in the system, is intended to mean comprising one or more, but not all, of the respective complex valued antenna weights.
With reference still to figure 1, the inventors of the present invention have realized, as will be shown in the following, that all these antenna weights W1, W2, W3 and W4 need not be complex in order to accomplish a performance just as good as if all antenna weights in fact were complex valued. The SNR on the output from the linear MMSE receiver will, if n = 2 receiver antennas is assumed, for the two substreams be
where
and where
is the phase of
and
Hence, it can be seen that the SNR for both substreams are affected by the same phase angle Δ . Thereby, it is possible to control the phase angle Δ by using the phase of a single antenna, e.g. ^1 , and set «3 - α4 = α5 - ° without loss of performance. Hence, only a single complex weight, the antenna weight of one of the antennas, is necessary to be fed back to the transmitting end, and only real antenna weights of the remaining antennas. With one complex weight and three real valued weights, the SNR for the two substreams are exactly the same as in the case where complex weights are fed back for all four antennas, a scheme which requires a much higher feedback data rate. In accordance with the present invention, the performance is thus maintained at the same time as the bandwidth used for the feedback channel is decreased.
The use of a linear MMSE receiver might seem as a restriction, but is in fact a rather general assumption, since many advanced receiver algorithms, such as iterative receivers, also use a MMSE receiver to obtain a first bit estimate for initialization of the iterative receivers. Their performance depends on the reliability of the first estimate.
Figure 2 shows a system setup with two Alamouti encoded substreams and n = 2 receivers. It is to be noted that the present invention is not limited to the case with two Alamouti encoded substreams, but it is conceivable to use a different number of encoded substreams, e.g. even three (i.e. three sets of STTD encoders) or more (correspondingly increased number of encoders) encoded substreams. The transmitter of Figure 1 may be used for the disclosed D-STTD encoded scheme, with feedback of a single phase value θ, according to the invention. The two streams of data symbols are space time transmit
diversity encoded using the Alamouti code, and the output signal from the Alamouti encoder connected to one of the four transmit antennas is phase shifted by an angle θ. The information of the values of these parameters are obtained from the feedback channel, either directly or based on a combination of the most recent and previous received feedback information.
The choice of antenna for which the phase weight is applied is entirely arbitrary and gives the same performance. Hence the description given here considers antenna 1 weighting but can straightforwardly be derived for any other antenna weight. In the case of choosing the amplitude weight, choosing a lower amplitude may give a lower power, in which case the performance of the system may depend on the choice of antenna. But in general, the choice of which antenna weight phase to use is totally arbitrary.
The feedback phase weight θ that maximizes the signal to noise ratio (SNR) for the two substreams is the optimal choice of θ, and is utilized in the preferred embodiment of the present invention.
Assume without loss of generality that the transmitter is using unit power and the noise power at the two receiver antennas are independent and have variance σ . The signal to noise ratio of the two substreams, SNRi and SNR∑, after linear MMSE filtering, are then given by
where
H denotes Hermitian transpose and the vectors gv are defined as the columns in the following matrix
where * denotes complex conjugate and where
H1 =[K h2 Kf (3)
is the vector containing the received channel coefficients from transmit antenna i to all n receive antennas.
In accordance with the present invention the signal to noise ratio (1) for the two substreams is maximized by finding the optimal phase angle θ.
As an exemplary illustration of the present invention, a special embodiment is described below, but it is to be realised that the present invention is not limited to only two receive antennas as in the described example.
In this embodiment, the amount of feedback can be reduced even further by using only a feedback value for the phase of the antenna weight of any one of the antennas, and no weighting at all (constant weight) on the remaining three antennas.
Hence, the weights in Figure 1 is in this case set to
The phase is optimized by maximizing the signal to noise ratio in the receiver. For the case of n = 2 receive antennas, we then have:
Applying weights on more antennas than antenna one does not give more degrees of freedom to improve the SNRs in (1). The phase θ that maximizes SNRj and SNR2 streams simultaneously for n = 2 receive antennas is calculated as
and where the phases a' P are defined as
a = ZA12 + ZA21 - Z(A14A23 - A13A24) β = Zh1 j + ZA22 - Z(A13A24 - A14A23 ) ^
where rt denotes the phase of the channel « .
With reference now to Figure 3, the performance of the present method for n = 2 receiver antennas is shown, where the bit error probability for one of two Alamouti encoded substreams as a function of bit energy per noise power density (EJ/NQ) is plotted, obtained from a Monte Carlo simulation. This is shown in the lowest graph. Binary PSK modulation was assumed and the channel coefficients were assumed to be independent and identically distributed (i.i.d) and subject to Rayleigh fading. For comparison, the bit error probabilities are also shown for an open loop method presented by Texas Instruments, "Double-STTD scheme for HSDPA systems with four transmit antennas: link level simulation results", Rl-01-0458, Busan, Korea, May 21-25, 2001, and a method presented in Huawei (A.Cao, B. Popovic, M.Wennstrδm), "A method and system for adaptive space-time closed-loop transmit diversity", Patent application WO 2004/062132. In the comparison ideal feedback is utilized. Clearly, the method according to the present invention improves the average bit error probability over the case with no feedback and for higher Et/No. The proposed method also has improved performance compared the method in the international patent application WO 2004/062132.
Furthermore, to reduce the amount of feedback, it might be necessary to quantize the feedback weights so that the feedback transmission can be performed with a limited number of bits. The method can be combined with any quantization method for reducing the number of bits required to feed back the antenna weights, such as for example Gray
encoding, or any method for combining and using already sent feedback bits and newly sent feedback bits. A method for regenerating θ requiring as few bits as possible is naturally preferred.
Speaking in terms of bits for exemplifying the bandwidth savings accomplished by the present invention: in prior art four bits are used to convey a phase information and one bit per amplitude of one antenna weight, thus 5 bits for each antenna, which gives 20 bits for a D-STTD using all complex valued antenna weights. In accordance with the present invention, it is possible to use one complex valued antenna weight, i.e. 5 bits and the real valued amplitude weights for the remaining antennas, i.e. an additional 3 bits, which gives a total of 8 bits. This is a considerable bandwidth saving, and the performance is unchanged in the two cases, whereby the present invention shows a remarkable improvement. In another embodiment of the invention the number of bits sent may be reduced even further, described below; sending only the phase of one of the antennas and constant amplitude weights, results in feedback information consisting of only 4 bits.
An exemplary embodiment of the present invention, which is suitable for use with broadband wireless access according to IEEE 802.16, will now be described. In particular, the present invention may advantageously be used in 3 or 4 BS antenna systems where Space Time Code (STC) matrix A and B plus antenna grouping is used. STC matrix A and B are described in detail in IEEE P802.16e/D6 "Air interface for Fixed and Mobile Broadband Wireless Access Systems- Amendment for Physical and Medium Access Control Layers for Combined Fixed and Mobile Operations in Licensed Bands".
The 3 and 4 BS antenna closed loop MIMO precoding codebooks disclosed in 802.16e/D6 are optimized for the use of STC matrix C, and the STC encoders in these cases correspond to spatial multiplexing. However, closed loop MIMO precoding can also be used together with Alamouti encoding, corresponding to the STC matrices A and B, if they are used together with antenna grouping, see M.Wennstrόm, B. Popovic, "Closed Loop Precoding for STC", IEEE C802.16e-04/451rl.
An alternative codebook for 4 BS antenna systems, which shows significant performance gain compared to using the existing codebook in 802.16e will now be described.
For a 4 BS antenna system using the STC matrix B, the spatial rate for the encoder two and hence the two groups of Alamouti space time encoded signals will interfere with each other in the receiver. The output from the antenna grouping operation is 4 streams and there are 4 transmit antennas, so all the elements in the proposed codebook set are 4 by 4 matrices. According to M. Wennstrom, B. Popovic, "Closed Loop Precoding for STC", IEEE C802.16e-04/451rl, sufficient to consider the following precoding matrix structure.
where c,. i - 1,2,3,4 are real valued and θ is a phase which is selected to make the interference between the two subspaces corresponding to the two Alamouti encoded signal groups as small as possible. As is disclosed in the description above, it is sufficient if one of the four diagonal elements of the precoding matrix is complex valued, as nothing is gained by adding more complex valued coefficients.
The elements c, / = 1,2,3,4 and θ in (9) are selected using a 6 bit CQICH. The variables c, i = 1,2,3,4 can take the values 1,
0.343 or
1.372 and the phase angle θ is selected from the set 0°, ±10° and ±60°. Note that these phases do not sample the whole 360° angle space. The reason is due to the preceding antenna grouping, which has effectively reduced the useful angle space (in other words, angles |ø| > 90° are obsolete since they would reduce the SNR in the receiver and can thus be removed, thereby increasing the resolution in the useful range of angles).
Not all possible combinations of c,. / = 1,2,3,4 can be used in the codebook, since the total radiated power should be invariant to the selection, hence C
1 2 +
= 4 must hold for all matrices in the codebook. The proposed codebook can be seen in detail in Table 1 below.
TABLE 1
The codebook elements has the following structure
where the elements are selected by the 6-bit CQICH feedback according to the table below and wherein
At the Mobile Subscriber Station (MSS), the selection of a codebook matrix can be determined based on any desired criteria such as minimum BER, minimum mean square error or minimized cross interference between the two Alamouti encoded signal groups. The selection can be made jointly with the antenna grouping selection or separately, that is, antenna grouping selection is performed first and then the best possible precoding matrix is selected given the antenna grouping selection. This allows a use of different CQICH reporting periods for the precoder and the antenna grouping, if desired. Furthermore, more advanced selection methods (filtering, prediction ...) can also be used to reduce the effects of feedback delay.
The same code book can independently be used if matrix A or matrix B encoders are used.
For a 4 BS antenna system using the STC matrix A the spatial rate is one and therefore, there is no cross-interference between the two groups of antenna signals to consider in the receiver. Therefore the value of the phase angle θ has no impact on the performance and need not be taken into account in the selection criteria. Also, the selection of the precoding for antenna (it is assumed that antenna 1 and 2 are grouped to one Alamouti encoded signal and 3 and 4 to the other. This is of course dependent on the selection in the preceding antenna grouping operation. Antennas 1 and 2 can be made independently of the precoding selection for antennas 3 and 4).
To make the scheme more robust to feedback delay, the following selection criteria can be adopted. At the MSS, the index of the transmission matrix can be determined based as follows; if = V2 /17
otherwise choose C
1 = 1, c
2 = 1 where T>1 is a threshold parameter and /-Z
15Zz
2 are the channels from BS antenna 1 and 2 respectively (where these two antennas transmit one Alamouti encoded signal group. Hence they may not correspond to the physical antenna 1 and 2 due to the preceding antenna grouping operation). A similar selection is made for antenna 3 and
4 independently of the selection for antenna 1 and 2. This scheme is robust to feedback delays since when
open loop (C
1 =l, c
2 =1 ) transmission is used. So, in effect, when there is a large uncertainty about the order relation between the two channels in the future (i.e.
), the open loop scheme is selected.
For a 3 BS antenna system using the STC matrix B the spatial rate is two, so there is cross-interference to consider in the receiver. However, it can be seen from the discussion in M. Wennstrδm, B. Popovic, "Closed Loop Precoding for STC", IEEE C802.16e-04/451rl that complex antenna weights has no impact on the performance and the following matrix structure is proposed:
where c, , i = 1,2,3 are real valued. The variables C
1 i = 1,2,3 can take the values or 1 under the constraint - 3 , hence 3 bits
are needed to select a precoding matrix from the codebook. At the MSS, the index of the transmission matrix can be determined based on for example minimizing BER, minimum square error or maximizing SNR.
In an exemplary codebook, the codebook elements have the following structure: frfi 0
0 d2 0
1° 0
where the elements are selected by the 6-bit CQICH feedback according to TABLE 2 below and P —
2
TABLE 2
As is apparent from the above table, this information could equally well be transmitted in a 3-bit field instead of a 6-bit field. A 3 BS antenna system using the STC matrix A may also use the above codebook.
In the above description a transmitting end could for example be a base station, or a node B, and the receiving end could be any mobile user equipment, including at least two receive antennas. The channel calculations may be performed in either end.
The present invention is not limited to the use of two antennas. The first described method using a subset of complex weights and also real valued weights may thus also be extended to an arbitrary number of receive antennas, larger than or equal to two. It is possible to use two complex valued weights, and two real valued weights, in the case of two receivers, and still gain bandwidth.
Further, any method or criterion can be used for calculating the phase angle θ and real valued antenna weights, not necessarily the one that maximizes the SNR.
The method according to the present invention may be applied in an OFDM (Orthogonal Frequency Division Multiplexing) system such as IEEE 802.16e where each sub carrier is flat fading.
The method can also be applied to time division duplex (TDD) systems where the channel coefficients from the reverse link directly, as opposed to using a separate feedback information channel, can be applied to calculate the phase angle in the transmitter.
In summary, the present invention gives a means for minimizing the mutual interference between different Alamouti encoded data streams or a single Alamouti encoded stream and a directly transmitted data stream from a third antenna. The invention further gives explicitly the optimal phase weights of the antennas in phase weighted D-STTD in a flat fading channel for n = 2 antennas. As a consequence of the space time coding structure and the linear MMSE receiver, if no amplitude weighting is applied, only one of the antennas (arbitrarily selected to antenna number one) needs to be phase adjusted to maximize the signal to noise ratio in the receiver for both the STTD encoded data streams simultaneously. The selected phase
minimizes the mutual interference between the two Alamouti encoded data substreams. The present invention thus provides a solution to the trade-off between the limited bandwidth available and the need to send as much feedback information as possible without degrading the performance of the system.