Method and transmitter for broadcast transmission with spatial spreading in a multi-antenna communication system

<figref>image 1</figref>

<figref>image2</figref>

<figref>image3</figref>

<figref>image4</figref>

5

10

15

20

25

30

35

40

E08153985

08-08-2014

For a single carrier MIMO system, a MIMO channel formed by the Nap antennas at the access point and the Nut antennas at a given user terminal or can be characterized by a channel response Hu matrix, of dimensions Nut x Nap, which can be expressed as:

<figref>image5</figref>

Ec. (1)

where the input hij, for i = 1 ... Nut and j = 1 ... Nap, indicates the complex coupling or gain between the antenna j of the access point and the antenna i of the user terminal. As shown in FIG. 1, the user terminals may be distributed throughout the coverage area of ​​the access point. A different MIMO channel is formed by the Nap antennas at the access point and the Nut antennas at each user terminal.

Data can be transmitted in various ways in the single carrier MIMO system. In a simple transmission scheme, a flow of data symbols is transmitted from each antenna of the access point, and NS data symbol flows are transmitted simultaneously from NS of the Nap antennas of the access point, where NS is the number of space channels and NS &lt;min Nap, Nut}. The symbols received at the user terminal for this transmission scheme can be expressed as:

ru = Hus + nu, Ec. (2)

where s is a vector of dimensions Nap x 1 with NS non-zero inputs for NS data symbols to be transmitted simultaneously by the access point;

ru is a vector of dimensions Nut x 1 with inputs for Nut received symbols, obtained by means of the Nut antennas in the terminal or of user; and

nu is a noise vector observed in the user terminal u.

For simplicity, the noise is assumed to be additive white Gaussian noise (AWGN) with an average vector

zero and a covariance matrix of <figref>image6</figref>, where <figref>image6</figref>it is the variance of the noise observed by the user u terminal and I is the identity matrix.

The NS flows of data symbols transmitted from the Nap antennas of the access point interfere with each other in the user terminal. A given flow of data symbols, transmitted from an antenna of the access point, can be received by all Nut antennas of the user terminal in different amplitudes and phases. Each received stream of symbols includes a component of each of the NS streams of transmitted data symbols. The Nut symbol flows received would collectively include all NS data symbol flows. However, these NS data symbol flows are scattered among the Nut symbol flows received. The user terminal or performs the spatial processing of the receiver on the Nut flows of symbols received, to retrieve the NS flows of data symbols transmitted by the access point.

The performance that can be achieved for the terminal or user depends (to a large extent) on its channel response matrix Hu. If there is a high degree of correlation within Hu, then each flow of data symbols would observe a large amount of interference from the other flows, which cannot be eliminated by the spatial processing of the receiver in the user terminal. The high level of interference degrades the SNR of each affected data symbol flow, possibly to a point where the data symbol flow cannot be decoded correctly by the user terminal.

For a data transmission to a specific user terminal, through a given MIMO channel, the system capacity can be achieved if the access point is provided with sufficient channel status information with respect to the MIMO channel. The access point can then use this information to process data in a way that maximizes performance for the user terminal (eg, select the appropriate speed for each data flow). Since different user terminals observe different MIMO channels, the access point would usually need to process data differently for each user terminal, to maximize the

6 5

10

15

20

25

30

35

40

45

E08153985

08-08-2014

performance for that user terminal.

For a broadcast transmission, the access point transmits the same data to a certain number of user terminals within a broadcast coverage area. For broadcasting, the access point usually has no channel status information for user terminals. In addition, it is usually not practical to process data designed for multiple user terminals based on channel status information for a specific user terminal.

The broadcast broadcast from the access point observes a set of MIMO channels for different user terminals in the broadcast coverage area. A certain percentage of MIMO channels can be considered as "bad." For example, a bad channel may appear when the channel response matrix H exhibits a high degree of correlation, or when there is insufficient dispersion, multipath (high coherence bandwidth) or temporary fading (high coherence time) in the channel . The appearance of "bad" channels is random, and it is desirable to minimize the percentage of time in which this can occur for each user terminal.

For broadcasting, the access point needs to transmit each flow of data symbols at a sufficiently low rate, so that the flow can be recovered by the user terminals even in the worst case channel condition. Broadcasting benefits are then dictated by the expected channel condition of the worst case, for all user terminals in the coverage area.

A. Broadcast of MIMO

Spatial widening can be used to randomize an effective MIMO channel observed by each user terminal, so that broadcasting performance is not dictated by a single channel embodiment on a code block. With spatial widening, the access point performs spatial processing on each code block, with multiple guide matrices to effectively randomize the MIMO channel for each user terminal. Consequently, each user terminal observes a set of channels in the extension of each code block and does not get stuck in a single channel for an extended part of the code block.

The spatial processing at the access point for spatial widening in the MIMO system can be expressed as:

xmimo (m) = V (m) · s (m), Eq. (3)

where

s (m) is a vector of dimensions NS x 1 with NS data symbols to be sent in the transmission section m;

V (m) is a guide matrix of Nap x NS dimensions for the transmission section m; and

xmimo (m) is a vector of dimensions Nap x 1 with Nap transmission symbols to be sent from the Nap antennas of the access point in the transmission section m.

A transmission section may cover time and / or frequency dimensions. For example, in a single carrier MIMO system, a transmission segment may correspond to a symbol period, which is the time duration for transmitting a data symbol. As another example, in a multi-carrier MIMO system, such as a MIMO system that uses the OFDM, a transmission segment may correspond to a sub-band in a period of OFDM symbols. A transmission segment can also cover multiple periods of symbols and / or multiple subbands. Thus, m can be an index for time and / or frequency. A transmission segment may also be called a transmission interval, a signaling interval, a slot, etc.

A set of L guide matrices can be generated as described below, and used for spatial widening. This set of guide matrices is indicated as {V}, or V (i), for i = 1 ... L, where L can be any integer greater than one. A guide matrix in the assembly can be selected for each transmission section m, and used for spatial processing by the access point, for that transmission section. The results of the spatial processing are Nap streams of transmission symbols for dissemination from the Nap antennas of the access point.

The symbols received in each user terminal with spatial widening can be expressed as:

r (m) = H (m) · V (m) · s (m) + n (m) = Heff (m) · s (m) + n (m), Ec. (4)

where

r (m) is a vector of dimensions Nut x 1 with Nut symbols received for the transmission section m;

7 5

10

15

20

25

30

35

40

45

50

55

E08153985

08-08-2014

H (m) is a channel response matrix of Nut x Nap dimensions for the transmission section m;

Heff (m) is an effective channel response matrix, of Nut x NS dimensions, for the transmission section m, which is Heff (m) = H (m) · V (m); and

n (m) is a noise vector for the transmission section m.

For simplicity, the channel response H (m) is assumed to be constant over each transmission segment. The quantities H (m), Heff (m), r<figref>image7</figref>(m) and n (m) are different for different user terminals, while the quantities V (m) and s (m) are the same for all user terminals. To simplify the notation, the subscript “u” is omitted for the terminal or user, in the user specific amounts in equation (4) and in the following description.

As shown in equation (4), due to the spatial widening performed by the access point, the NS data symbol flows observe the effective response of channel Heff (m) instead of the actual response of channel H (m) for each user terminal. If multiple guide matrices are used for broadcast transmission, then each flow of data symbols effectively observes a set of spatial channels of H (m). In addition, if multiple guide matrices are used in the extension of a code block, then the data symbols in the code block would observe different channels in the extension of the code block.

In general, the access point can broadcast any number of (ND) data streams simultaneously to the user terminals, where NS> ND> 1. For example, if ND = NS, then the access point can broadcast a stream of data for each spatial channel of Heff (m). The maximum number of data streams that can be broadcast simultaneously is determined by the number of space channels for all user terminals, which, in turn, is determined by (1) the number of antennas at the access point and (2) the minimum number of antennas in all user terminals. If all user terminals are equipped with the same number of antennas, then min {Nap, Nut}> NS> ND. If ND = 1, then the access point can broadcast a data stream from its Nap antennas.

FIG. 2 shows a process 200 for spreading data with spatial widening. Initially, the access point processes data for ND data flows, to obtain a set of ND blocks of data symbols, one block for each data flow (block 212). Each data symbol block contains data symbols generated from a block of encoded data, which can be called a code block or a packet of encoded data. Data processing can be performed as described below. The access point then divides the ND blocks of data symbols into NM sub-blocks of data symbols to be disseminated in NM transmission segments, a subblock in each transmission segment (block 214). NM is also called the block length and is greater than one,

or NM> 1. Each sub-block may contain one or more data symbols of each of the ND blocks. For example, if ND = NS, then each sub-block may contain NS data symbols from NS blocks for NS data flows. As another example, if ND = 1, then each sub-block may contain NS data symbols from a block for a data flow. The index m used to indicate the transmission segment for the current set of data symbol blocks is fixed at 1 (block 216).

For each transmission section m, the access point selects a guide matrix, which is indicated as V (m), p. eg, from the set of L guide matrices (block 218). The access point then performs the spatial processing on the sub-block m of data symbols with the guide matrix V (m), to obtain transmission symbols (block 220). If the transmission section m encompasses a vector of data symbols, then the access point forms a vector s (m) with up to NS data symbols, from the sub-block m of data symbols, and spatially processes the vector s (m) with the matrix V (m), to obtain the vector xmimo (m), as shown in equation (3). If the transmission section m encompasses multiple (NV) data symbol vectors, then the access point forms NV vectors sl (m), for l = 1 ... NV, from the sub-block m of data symbols, and spatially processes each vector sl (m) with the same guide matrix V (m), to obtain a corresponding minimum vector, l (m) of transmission symbols. In any case, the access point uses the same guide matrix V (m) for spatial processing, for all the data symbol vectors in the transmission section m. The access point also processes and disseminates the vectors resulting from transmission symbols by means of the Nap transmission antennas in the transmission section m (block 222).

A determination is then made as to whether or not all NM sub-blocks of data symbols have been processed and transmitted (ie, if m = NM) (block 224). If the answer is "No", then the index m is increased for the next sub-block, or transmission segment (block 226), and the process returns to block 218. If the answer is "Yes" for block 224, then a determination is made as to whether or not there is more data to disseminate (block 228). If the answer is "Yes", then the process returns to block 212 to start processing for the next set of symbol blocks. Otherwise, the process ends.

Each set of data symbol blocks is therefore spatially processed with NM guide matrices, to obtain Nap transmission symbol sequences. Each sequence of transmission symbols is broadcast from

8

<figref>image8</figref>

<figref>image9</figref>

<figref>image10</figref>

<figref>image11</figref>

<figref>image12</figref>

<figref>image13</figref>

<figref>image14</figref>

<figref>image15</figref>

<figref>image16</figref>

<figref>image17</figref>

<figref>image18</figref>

<figref>image19</figref>

E08153985

08-08-2014

and

<figref>image20</figref>

where is an estimate of heff, i (m), for i = 1, 2; and

<figref>image21</figref>they are post-processed noise, respectively, for the detected symbols

<figref>image22</figref>and<figref>image6</figref>.

<figref>image6</figref>

If the user terminal is equipped with multiple antennas, then the received symbols can be expressed as:

<figref>image23</figref>

where ri (m) is a vector of dimensions Nut x 1 with Nut symbols received for the period i of symbols of the transmission section m; 10 H (m) is a channel response matrix, of Nut x Nap dimensions, for the transmission section m;

Heff, sttd (m) is an effective channel response matrix, of dimensions Nut x 2, for the transmission section m, which is Heff, sttd (m) = H (m) · Vsttd (m) = [heff, 1 (m) heff, 2 (m)]; and ni (m) is a noise vector for the period i of symbols of the transmission section m. It is assumed that the channel response H (m) is constant over the two symbol periods of the section m of

fifteen transmission. The multi-antenna user terminal can obtain estimates of the two data symbols, sa (m) and sb (m), as follows:

<figref>image24</figref>

and

<figref>image25</figref>

twenty where

<figref>image6</figref>

it is an estimate of heff, i (m), for i = 1, 2; and

<figref>image6</figref>and

<figref>image6</figref>they are post-processed noise, respectively, for the detected symbols

<figref>image6</figref>and<figref>image6</figref>.

For the SFTD scheme, the two vectors s1 (m) and s2 (m) are broadcast by two different subbands in the same symbol period. The same transmission and reception processing described above can be performed

25 for the SFTD scheme, where index i now indicates sub-band instead of symbol period. Since the channel response is assumed to be approximately constant over the transmission segment, two adjacent sub-bands kyk + 1 can be used to diffuse the two vectors s1 (m) and s2 (m).

21

<figref>image26</figref>

<figref>image27</figref>

<figref>image28</figref>

5

10

15

20

25

30

35

E08153985

08-08-2014

each. This condition ensures that the NS data symbols diffused simultaneously, using the guide matrix V (i), have the same power and are orthogonal to each other before transmission.

Some of the guide matrices may also be uncorrelated, so that the correlation between two uncorrelated guide matrices is zero, or a low value. This condition can be expressed as:

<figref>image29</figref>for i = 1 ... L, j = 1 ... L ei ≠ j, Ec. (24)

where C (ij) is the correlation matrix for V (i) and V (j), and 0 is a matrix of all zeros. The condition in the equation

(24) may improve performance for some applications, but is not necessary for most applications.

A set of L guide matrices {V} can be generated using various schemes. In a first scheme, the L guide matrices are generated based on random variable matrices. A matrix G of dimensions NS x Nap, with elements that are Gaussian random variables complex, independent and identically distributed (IID), each having a mean zero and unit variance, is initially generated. A correlation matrix of G, of dimensions Nap x Nap, is calculated as R = GH .G and decomposed using decomposition into eigenvalues, such as:

R = E · D · EH, Ec. (25)

where

E is a unitary matrix, of Nap x NS dimensions, of R autovectors;

D is a diagonal matrix, of dimensions NS x NS, of eigenvalues ​​of R.

The diagonal matrix D contains eigenvalues ​​of R, which represent the power gains for the NS automodalities of G. The matrix E is used as a guide matrix V (i) and is added to the set. The guide matrix V (i) is a unitary matrix because the matrix E is obtained by decomposition into eigenvalues. The process is repeated until all the guide matrices are generated.

In a second scheme, the L guide matrices are generated based on a set of (log2L) +1 independent unit matrices, isotropically distributed. A random unit matrix is ​​isotropically distributed if its probability density remains unchanged by pre-multiplication by any deterministic unit matrix of

Nap x Nap dimensions. The index i for the guide matrices in the set can be indicated as<figref>image6</figref>i = <figref>image6</figref>where Q

= log2 L,

<figref>image6</figref>it is the first bit of index i, it is the last bit of index i, and each bit can take a value of 0 or 1. The guide matrices can then be generated as follows:

<figref>image30</figref>

where V0 is an isotropically distributed independent unit matrix, of dimensions Nap x NS; and

<figref>image6</figref>

, for j = 1 ... Q, is an isotropically distributed independent unit matrix, of dimensions Nap x Nap.

The matrix V0 can be defined, for example, as <figref>image6</figref>, where INS is an identity matrix of dimensions NS x NS. The second scheme is described by TL Marzetta in "Structured Constellations of Unitary Space-Time Self-Codification", IEEE Transaction on Information Theory, Vol. 48, No. 4, April 2002.

In a third scheme, the L guide matrices are generated by successively rotating an initial unit guide matrix V (1) in a complex Nap-dimensional space, according to the following:

<figref>image31</figref>for i = 1 ... L – 1, Ec. (27)

where

<figref>image6</figref>It is a diagonal unit matrix of Nap x Nap dimensions, which can be defined as:

25

<figref>image32</figref>

<figref>image33</figref>

<figref>image34</figref>