TRANSMITTER FOR FBMC SYSTEM WITH BLOCK-ALAMOUTI TYPE SPACE-TIME CODING

Abstract

A transmission method and an FBMC transmitter to transmit at least a first and a second block of symbols (X.sub.0, X.sub.1), each symbols block including a temporal sequence of L vectors with predetermined size N. It uses a first and a second FBMC modulation channel, each FBMC modulation channel being associated with an antenna. During a first use of the channel, the vectors of the first block and the vectors of the second block are input to the first and to the second FBMC modulation channels respectively, in the order of the temporal sequence. During a second use of the channel, the vectors of the first and second blocks are multiplied by a factor j.sup.L 1 respectively and −(j.sup.L 1) input to the second and to the first FBMC modulation channels respectively, in the inverse order of the temporal sequence.

Claims

1. An FBMC transmission method using at least a first symbols block and a second symbols block (X.sub.0, X.sub.1), each symbols block comprising a temporal sequence of L real vectors with a predetermined size N, comprising a first and a second FBMC modulation channel, each FBMC modulation channel being associated with an antenna and wherein, during a first use of the transmission channel, the vectors of the first block and the vectors of the second block are input to the first FBMC modulation channel and to the second FBMC modulation channel respectively, in the order of said temporal sequence; the first block is transformed by multiplying the vectors of this block by a factor j.sup.L−1 in which L is an even number and by inverting the temporal order of the sequence of vectors thus obtained, and the second block is transformed by multiplying the vectors of this block by a factor −(j.sup.L−1) and by inverting the temporal order of the sequence of vectors thus obtained; during a second use of the transmission channel, the vectors of the first and second blocks thus transformed are supplied to the second modulation channel and the first FBMC modulation channel respectively.

2. The FBMC transmission method according to claim 1, in which each FBMC modulation channel comprises an OQAM preprocessing step alternately supplying real data and imaginary data, each data thus obtained being spread over a plurality 2K−1 of adjacent sub-carriers and filtered in the spectral domain by a prototype filter to obtain a vector of KN components, the vector of KN components being subjected to an IFFT with size KN to generate an FBMC symbol of KN samples, consecutive FBMC symbols being offset by N/2 samples, each FBMC symbol being combined with the K−1 preceding FBMC symbols and the K−1 following FBMC symbols to provide an antenna signal transmitted by an antenna associated with said channel, after translation in the RF band.

3. The FBMC transmission method according to claim 1, in which each FBMC modulation channel comprises an OQAM preprocessing step supplying a vector of N alternatively real and imaginary components, the vector of N components being subjected to an IFFT with size N to generate a plurality of sub-channels, each sub-channel being filtered by a polyphase filter, the polyphase filters being versions translated in frequency by 2k/T of a prototype filter for which the pulse response has a duration of KT in which T is the sampling period, the outputs from the polyphase filters being oversampled by a factor N/2 and delayed by 0 to N−1 sampling periods before being summated to provide an antenna signal transmitted by an antenna associated with said channel, after translation in the RF band.

4. The FBMC transmission method according to claim 1, wherein: when the channel is used for the first time, a guard block is provided consisting of a predetermined number of null vectors to the first and second modulation channels before providing the vectors of the first block and the vectors of the second block respectively to them, and in that when the channel is used for the second time, a guard block is provided consisting of said predetermined number of null vectors to the first and second modulation channels before providing the vectors of the second transformed block and the vectors of the first transformed block respectively to them.

5. The FBMC transmission method according claim 4, wherein the predetermined number of null vectors is equal to K+E in which E is the temporal spread of the transmission channel expressed as a number of samples.

6. The FBMC transmission method according to claim 1, wherein: when the channel is used for the first time, first and second preambles are provided consisting of a predetermined number of vectors known to the receiver, to the first and second modulation channels before providing the first block vectors and the second block vectors respectively to them, and in that when the channel is used for the second time, a guard block is provided consisting of said predetermined number of null vectors to the first and second modulation channels before providing the vectors of the second transformed block and the vectors of the first transformed block respectively to them.

7. The FBMC transmission method according claim 6, wherein said predetermined number is equal to K+E in which E is the temporal spread of the transmission channel expressed as a number of samples.

8. The FBMC transmission method according to claim 1, wherein L is a power of 2.

9. The FBMC transmitter to transmit at least a first and a second symbols block (X.sub.0, X.sub.1), each symbols block comprising a temporal sequence of L real vectors with predetermined size N, comprising first and second FBMC modulation means associated with a first and a second transmission antenna respectively, wherein: during a first use of the transmission channel, the vectors of the first block and the vectors of the second block are provided to the first FBMC modulation channel and to the second FBMC modulation channel respectively, in the order of said temporal sequence, and in that said transmitter comprises: first transformation means adapted to transform the first block by multiplying the vectors of this block by a factor j.sup.L−1, in which L is an even number, and by inverting the temporal order of the sequence of vectors thus obtained, and second transformation means adapted to transform the second block by multiplying the vectors of this block by a factor −(j.sup.L−1) and by inverting the temporal order of the sequence of vectors thus obtained; and in that when the transmission channel is used for the second time, the first and second transformation means supply vectors of the first and second blocks thus transformed to the second modulation channel and to the first FBMC modulation channel.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Other characteristics and advantages of the invention will become clear after reading preferred embodiments of the invention, with reference to the appended figures among which:

[0068] FIG. 1 diagrammatically shows a known FS-FBMC telecommunication system according to the state of the art;

[0069] FIG. 2A illustrates the spectral spreading done upstream from the IFFT module in FIG. 1;

[0070] FIG. 2B illustrates the spectral despreading done downstream from the IFFT module in FIG. 1;

[0071] FIG. 3 illustrates the combination of FBMC symbols in FIG. 1;

[0072] FIG. 4 diagrammatically represents the transmission of two sequences of symbol blocks transmitted by an FBMC transmitter using block-Alamouti coding known in the state of the art;

[0073] FIG. 5 diagrammatically represents the architecture of an FBMC receiver capable of receiving sequences of symbol blocks transmitted by the transmitter in FIG. 4;

[0074] FIG. 6 diagrammatically represents the architecture of an FS-FBMC receiver capable of receiving sequences of symbol blocks coded by a block-Alamouti coding;

[0075] FIG. 7A diagrammatically represents the transmission of two sequences of symbol blocks by an FBMC transmitter using a first block-Alamouti coding according to a first example embodiment of the invention;

[0076] FIG. 7B diagrammatically represents the transmission of two sequences of symbol blocks by an FBMC transmitter using a second block-Alamouti coding according to a second example embodiment of the invention;

[0077] FIG. 8 diagrammatically represents the architecture of an FS-FBMC transmitter capable of transmitting sequences of symbol blocks according to FIGS. 7A and 7B.

[0078] FIG. 9 diagrammatically represents the architecture of an FBMC transmitter according to an example embodiment of the invention different from that of FIG. 8.

DETAILED PRESENTATION OF PARTICULAR EMBODIMENTS

[0079] We will start by considering an FS-FBMC transmitter like that described with reference to FIG. 1, to facilitate understanding the notation.

[0080] Unlike the previous notations, the column vectors X.sup.m, m=0, . . . , L−1, with size N, will represent vectors of input data in the following, in other words data at the input to the OQAM modulator. Therefore the values of elements of these vectors are real.

[0081] The signal transmitted by the transmitter at time m can be represented by a column vector Z.sup.m with size KN for which the elements are samples at frequency Nf. The vector Z.sup.m can be expressed as a function of input data vectors X.sup.m−(K−1), . . . , X.sup.m, . . . , X.sup.m+(K−1), namely:

[00006]$\begin{array}{cc}{Z}^m=F^H.Math.G\left(X^m.Math..Math..Math.M^m\right)+\underset{p=1}{\overset{K-1}{.Math.}}.Math..Math.Q_{\frac{\mathrm{pN}}{2}}.Math.F^H.Math.G\left(X^{m-p}.Math..Math.M^{m-p}\right)+Q_{\mathrm{KN}-\frac{\mathrm{pN}}{2}}.Math.F^H.Math.G\left(X^{m+p}.Math..Math.M^{m+p}\right)& \left(8\right)\end{array}$

in which □ is the Hadamard product, F is the matrix of the discrete Fourier transform with size KN×KN, G is a matrix with size KN×N representing spectral spreading and the transfer function of the prototype filter in the frequency domain, namely:

[00007]$\begin{array}{cc}G=\left(\begin{array}{cccc}{G}_{K-1}& 0& \mathrm{.Math.}& 0\\ \mathrm{.Math.}& G_{K-1}& \ddots & \mathrm{.Math.}\\ G_0& \mathrm{.Math.}& \ddots & 0\\ \mathrm{.Math.}& G_0& \ddots & G_{K-1}\\ G_{-K+1}& \mathrm{.Math.}& \ddots & \mathrm{.Math.}\\ 0& G_{-K+1}& \ddots & G_0\\ \mathrm{.Math.}& \ddots & \ddots & \mathrm{.Math.}\\ 0& \mathrm{.Math.}& 0& G_{-K+1}\end{array}\right)& \left(9\right)\end{array}$

M.sup.m is a column vector with size N that translates the OQAM modulation, namely a vector for which the elements are given by:

M.sup.m[k]=j.sup.m+k(−1).sup.km  (10)

and Q.sub.l is an offset matrix of l samples with size KN×KN defined by:

[00008]$\begin{array}{cc}{Q}_{}=\left(\begin{array}{cc}{0}_{\times \left(\mathrm{KN}-\right)}& 0_{\times }\\ I_{\mathrm{KN}-}& 0_{\left(\mathrm{KN}-\right)\times }\end{array}\right)& \left(11\right)\end{array}$

in which I.sub.KN−l is the size identity matrix (KN−l)×(KN−l)

[0082] It will be understood that the terms under the sum sign in expression (8) represent the 2K−1 FBMC symbols that are combined in FIG. 3.

[0083] The signal received by the FBMC receiver at time m can be expressed similarly in the form of a data vector at the output from the OQAM demodulator, in this case denoted Y.sup.m, with size KN. The vector Y.sup.m can be expressed as a function of the vector Z.sup.m representing the transmitted signal, that is, if we ignore the noise term:

Y.sup.m=(G.sup.HFH.sub.0Z.sup.m)□M.sup.m*  (12)

or, allowing for the fact that G.sup.HFF.sup.HG=I.sub.N and that (X.sup.m□M.sup.m)□M.sup.m*=X.sup.m:

[00009]$\begin{array}{cc}{Y}^m=H_0\left(X^m+\underset{p=1}{\overset{K-1}{.Math.}}.Math..Math.U^p\left(X^{m-p}.Math..Math.M^{m-p}\right).Math..Math.M^{m*}+\underset{p=1}{\overset{K-1}{.Math.}}.Math..Math.V^p\left(X^{m+p}.Math..Math.M^{m+p}\right).Math..Math..Math.M^{m*}\right)& \left(13\right)\end{array}$

in which:

[00010]$\begin{array}{cc}{U}^p=G^H.Math.{\mathrm{FQ}}^{\frac{\mathrm{pN}}{2}}.Math.F^H.Math.G.Math..Math.\mathrm{and}.Math..Math.V^p=G^H.Math.{\mathrm{FQ}}^{\mathrm{KN}-\frac{\mathrm{pN}}{2}}.Math.F^H.Math.G& \left(14\right)\end{array}$

[0084] It will be noted that G.sup.H=G.sup.T, considering that the coefficients of the filter transfer matrix are real.

[0085] We will now assume that a block-Alamouti coding is made, with a coding matrix defined by:

[00011]$\begin{array}{cc}\stackrel{\_}{C}=\left(\begin{array}{cc}{\stackrel{\_}{X}}_0& {\stackrel{\_}{X}}_1\\ -{\stackrel{\_}{X}}_1.Math.T& {\stackrel{\_}{X}}_0.Math.T\end{array}\right)& \left(15\right)\end{array}$

[0086] As described below, it is possible to use a receiver implemented in the frequency domain (FS-FBMC receiver) and the two blocks at the output from the FFT module (module 170 in FIG. 1) can be combined, during the first and second use of the channel respectively.

[0087] We will use X.sub.0.sup.m to denote the m.sup.th input data vector to the first block X.sub.0 and X.sub.1.sup.m to denote the m.sup.th input data vector to the second block X.sub.1. We will also use W.sub.0.sup.m to denote the m.sup.th samples vector at the output from the FFT module, before despreading and filtering, during the first use of the channel. Similarly, we will use W.sub.1.sup.m to denote the m.sup.th samples vector at the output from the FFT module, before despreading and filtering, during the second use of the channel.

[0088] During the first use of the channel, the vector W.sub.0.sup.m can be expressed as follows:

[00012]$\begin{array}{cc}{W}_0^m=H_0\left(G\left(X_0^m.Math..Math.M^m\right)+\underset{p=1}{\overset{K-1}{.Math.}}.Math.A^p\left(X_0^{m-p}.Math..Math.M^{m-p}\right)+\underset{p=1}{\overset{K-1}{.Math.}}.Math.B^p\left(X_0^{m+p}.Math..Math.M^{m+p}\right)\right)+H_1(.Math.G\left(X_1^m.Math..Math.M^m\right)+\underset{p=1}{\overset{K-1}{.Math.}}.Math.A^p\left(X_1^{m-p}.Math..Math.M^{m-p}\right)+\underset{p=1}{\overset{K-1}{.Math.}}.Math.B^p\left(X_1^{m+p}.Math..Math.M^{m+p}\right))& \left(16\right)\\ .Math.\mathrm{where}.Math..Math.A^p={\mathrm{FQ}}^{\frac{\mathrm{pN}}{2}}.Math.F^H.Math..Math.\mathrm{and}.Math..Math.B^p={\mathrm{FQ}}^{\mathrm{KN}-\frac{\mathrm{pN}}{2}}.Math.F^H& \left(17\right)\end{array}$

[0089] Similarly, during the second use of the channel, the vector W.sub.1.sup.m can be expressed as follows:

[00013]$\begin{array}{cc}{W}_1^m=-H_0\left(G\left(X_1^{L-1-m}.Math..Math.M^m\right)+\underset{p=1}{\overset{K-1}{.Math.}}.Math.A^p\left(X_1^{L-1-m+p}.Math..Math.M^{m-p}\right)+\underset{p=1}{\overset{K-1}{.Math.}}.Math.B^p\left(X_1^{L-1-m-p}.Math..Math.M^{m+p}\right)\right)+H_1\left(G.Math.\left(X_0^{L-1-m}.Math..Math.M^m\right)+\underset{p=1}{\overset{K-1}{.Math.}}.Math.A^p\left(X_0^{L-1-m+p}.Math..Math.M^{m-p}\right)+\underset{p=1}{\overset{K-1}{.Math.}}.Math.B^p\left(X_0^{L-1-m-p}.Math..Math.M^{m+p}\right)\right)& \left(18\right)\end{array}$

expression in which advantage is taken of the fact that input data vectors are real values. It will be noted that in this case the size of the elementary channel transfer matrices H.sub.0 and H.sub.1 is KN×KN, due to spectral spreading.

[0090] If the vector block W.sub.1.sup.m, m=0, . . . , L−1 is transformed by temporal inversion and complex conjugation of the block, the m.sup.th vector of the block thus transformed can be written as follows, from (18):

[00014]$\begin{array}{cc}{W}_1^{L-m-1^{*}}=-H_0^{*}\left(G\left(X_1^m.Math..Math.M^{L-1-m^{*}}\right)+\underset{p=1}{\overset{K-1}{.Math.}}.Math..Math.A^{p*}\left(X_1^{m+p}.Math..Math.M^{L-1-m-p^{*}}\right)+\underset{p=1}{\overset{K-1}{.Math.}}.Math..Math.B^{p*}\left(X_1^{m-p}.Math..Math.M^{L-1-m+p^{*}}\right)\right)+H_1^{*}\left(G\left(X_0^m.Math..Math.M^{L-1-m^{*}}\right)+\underset{p=1}{\overset{K-1}{.Math.}}.Math..Math.A^{p*}\left(X_0^{m+p}.Math..Math.M^{L-1-m-p^{*}}\right)+\underset{p=1}{\overset{K-1}{.Math.}}.Math..Math.B^{p*}\left(X_0^{m-p}.Math..Math.M^{L-1-m+p^{*}}\right)\right)& \left(19\right)\end{array}$

Namely since:

M.sup.L−1−m*=−M.sup.mj.sup.L−1;M.sup.L−1−m−p*=−M.sup.m+pj.sup.L−1;M.sup.L−1−m+p*=−M.sup.m−pj.sup.L−1

in which it was assumed that the size L of the block was an even number, and that:

A.sup.p*=B.sup.p;B.sup.p*=A.sup.p

the vector W.sub.1.sup.L−m−1* of the inverted block can finally be written as follows:

[00015]$\begin{array}{cc}{W}_1^{L-m-1^{*}}=H_0^{*}.Math.j^{L-1}.Math.\left(G\left(X_1^m.Math..Math..Math.M^m\right)+\underset{p=1}{\overset{K-1}{.Math.}}.Math..Math.B^p\left(X_1^{m+p}.Math..Math..Math.M^{m+p}\right)+\underset{p=1}{\overset{K-1}{.Math.}}.Math..Math.A^p\left(X_1^{m-p}.Math..Math..Math.M^{m-p}\right)\right)-H_1^{*}.Math.j^{L-1}\left(G\left(X_0^m.Math..Math.M^m\right)+\underset{p=1}{\overset{K-1}{.Math.}}.Math..Math.B^p\left(X_0^{m+p}.Math..Math.M^{m+p}\right)+\underset{p=1}{\overset{K-1}{.Math.}}.Math..Math.B^{p*}\left(X_0^{m-p}.Math..Math.M^{m-p}\right)\right)& \left(20\right)\end{array}$

[0091] We can then estimate vectors of transmitted data X.sub.0.sup.m, X.sub.1.sup.m by combining vectors W.sub.0.sup.m W.sub.1.sup.L−m−1*:

{hacek over (X)}.sub.0.sup.m=μ(H.sub.0*W.sub.0.sup.m+j.sup.L−1H.sub.1W.sub.1.sup.L−1−m*)  (21-1)

{hacek over (X)}.sub.1.sup.m=μ(H.sub.1*W.sub.0.sup.m−j.sup.L−1H.sub.0W.sub.1.sup.L−1−m*)  (21-2)

in which

[00016]$\mu =\frac{1}{\mathrm{Tr}\left(H_0^H.Math.H_0+H_1^H.Math.H_1\right)},$

then by filtering and spectral despreading and finally an OQAM demodulation:

{circumflex over (X)}.sub.0.sup.m=μG(H.sub.0*W.sub.0.sup.m+j.sup.L−1H.sub.1W.sub.1.sup.L−1−m*)□M.sup.m*  (22-1)

{circumflex over (X)}.sub.1.sup.m=μG(H.sub.1*W.sub.0.sup.m−j.sup.L−1H.sub.0W.sub.1.sup.L−1−m*)□M.sup.m*  (22-2)

[0092] FIG. 6 diagrammatically represents the architecture of an FS-FBMC receiver capable of receiving sequences of symbol blocks coded by block-Alamouti coding;

[0093] The receiver comprises a sampling module 610 to sample the signal received in base band at rate Nf in which N is the number of sub-carriers and f is the frequency of FBMC symbols. Samples are grouped together in the form of blocks with size KN by a series-parallel converter 620.

[0094] The receiver is assumed to be synchronised on the FBMC symbols, in other words the beginning of an FFT window coincides with the first sample of an FBMC symbol (transmitted by one of the transmission antennas). Moreover, the receiver is assumed to be synchronised on channel use times such that it knows reception times of the first and second blocks.

[0095] The sample blocs are subjected to an FFT with size KN in the FFT module 630.

[0096] A demultiplexer 640 provides FFT output vectors on a first output 641 during the first use of the channel and on a second output 642 during the second use of the channel. The L vectors (size KN) generated sequentially on the first output are stored in a first buffer memory 651 configured in the form of a FIFO. The L vectors generated sequentially on the second output are also stored in a second buffer memory 652 configured in the form of a LIFO. The module 660 thus reads the L vectors in the inverse order (LIFO) to the order in which they are stored to achieve a temporal inversion, and also makes a complex conjugation of each of these vectors. A multiplier 670 multiplies the elements of the vectors at the output from the module 660 by (j).sup.L−1, in other words by j if L is an even number. In particular we could choose L equal to a power of 2:L=2.sup.l where l is an integer greater than 1.

[0097] Each element of a vector generated on the first output is multiplied in 681 by the complex conjugate of the coefficient of the first elementary channel between the first transmission antenna and the reception antenna, at the frequency of the sub-carrier carrying the element concerned (in this case the operation is symbolised by a multiplication of the vector at the buffer memory output by the matrix H.sub.0*) and in 683 by the complex conjugate of the coefficient of the second elementary channel between the second transmission antenna and the reception antenna, at the same sub-carrier frequency (in this case the operation is symbolised by a multiplication of the samples vector at the FFT output by the matrix H.sub.1*). It is understood that the matrices H.sub.0 and H.sub.1 in this case have a size of KN×KN and represent the coefficients of the elementary channels for the KN spectrally spread sub-carriers. We could choose an identical channel coefficient for the K frequencies derived from the same carrier. It is assumed that the matrices H.sub.0 and H.sub.1 are constant over the duration of the sequence (assumption of flat fading in time).

[0098] Similarly, each element of a vector generated on the second output is multiplied in 682 by the coefficient of the channel between the first transmission antenna and the reception antenna at the frequency of the sub-carrier carrying the element concerned (operation symbolised by a multiplication of the vector at the FFT output by the matrix H.sub.0) and in 684 by the coefficient of the channel between the second transmission antenna and the reception antenna at the same sub-carrier frequency (operation symbolised by a multiplication of the vector at the FFT output by the matrix H.sub.1).

[0099] The vectors at the output from the multiplier 681 are summated element by element with the vectors at the output from the multiplier 684, in the summator 691. Successive vectors, with size N output from the summator 691 are then supplied to a first spectral spreading and filtering module 695.

[0100] Similarly, the vectors at the output from the multiplier 682 are subtracted element by element from the vectors at the output from the multiplier 683, in the summator, 692. Successive vectors, with size N output from the summator 692 are then supplied to a second spectral spreading and filtering module 696.

[0101] An OQAM demodulation (not shown) is then performed on the vectors obtained by the first and second modules 695 and 696, to obtain estimated data vectors X.sub.0.sup.m and X.sub.1.sup.m, m=0, . . . , L−1.

[0102] This invention is based on the observation that the structure of the receiver in FIG. 6 can be simplified when the FBMC transmitter uses the block-Alamouti coding defined as follows, instead of the coding given by (15):

[00017]$\begin{array}{cc}{\stackrel{\_}{C}}^{\prime }=\left(\begin{array}{cc}{\stackrel{\_}{X}}_0& {\stackrel{\_}{X}}_1\\ -\left(j^{L-1}\right).Math.{\stackrel{\_}{X}}_1.Math.T& \left(j^{L-1}\right).Math.{\stackrel{\_}{X}}_0.Math.T\end{array}\right)& \left(23\right)\end{array}$

[0103] In this case, multiplication by the factor (j.sup.L−1) can be removed at reception and consequently the multiplier 670 can be omitted.

[0104] FIG. 7A diagrammatically represents the transmission of two sequences of symbol blocks by an FBMC transmitter using a first block-Alamouti coding according to a first example embodiment of the invention.

[0105] The data blocks to be transmitted in this case are considered upstream from the OQAM modulation.

[0106] A first sequence of blocks, 701, is formed by a first guard block 711, a first block of L input data vectors, X.sub.0, 721, a second guard block, 731, followed by a first transformed block, −(j.sup.L−1)X.sub.1T, 741, obtained by temporal inversion and multiplication by the factor −(j.sup.L−1) of the first input data block.

[0107] A second sequence of blocks, 702, is formed by a first guard block 712, a second block of L input data vectors, X.sub.1, 722, a second guard block, 732, followed by a second transformed block, {circumflex over (X)}.sub.0T, 742, obtained by temporal inversion and multiplication by the factor (j.sup.L−1) of the second input data block.

[0108] The size L of the data blocks is assumed to be even, in other words (j.sup.L−1)=j or (j.sup.L−1)=−j.

[0109] The guard blocks are composed of null vectors to prevent interference between data blocks and transformed blocks. The number of null vectors in the guard blocks is advantageously equal to K+E where K is the length of the prototype filter and E is the temporal spreading of the channel expressed as a number of samples at the sampling frequency (Nf).

[0110] The first and second sequences are transmitted by the first and second antennas, 791 and 792 respectively, after FBMC modulation.

[0111] FIG. 7B diagrammatically represents the transmission of two sequences of symbol blocks by an FBMC transmitter using a second block-Alamouti coding according to a second example embodiment of the invention.

[0112] The second example is identical to the first except that the first guard block is replaced in the first sequence by a first preamble 711′ and in the second sequence by a second preamble 712′. The other blocks remain unchanged and are therefore not described again.

[0113] The first and second preambles generate an interference affecting the first symbols in blocks X.sub.0 and X.sub.1, interference that does not affect blocks −X.sub.1T and X.sub.0T symmetrically. This asymmetry does not eliminate the interference for input data vectors X.sub.0.sup.m, X.sub.1.sup.m at the beginning of the block. However, since the preamble symbols are known to the receiver, this interference can be eliminated provided that an estimate of the transmission channel is available.

[0114] FIG. 8 diagrammatically represents the architecture of an FS-FBMC transmitter according to a first embodiment of the invention; This transmitter transmits sequences of symbol blocks coded by Alamouti coding according to FIGS. 7A and 7B.

[0115] The symbol blocks to be transmitted are denoted X.sub.0 and X.sub.1, as above. The input data vectors X.sub.0.sup.m are stored in a FIFO buffer, 810 and a LIFO buffer, 811. They are supplied to the multiplexer 841 in their order of arrival and to the multiplexer 842 in the inverted order of their arrival after having been delayed by the delay 821 and after having been multiplied by the factor (j.sup.L−1) in 831. Similarly, the input data vectors X.sub.1.sup.m are stored in a FIFO buffer, 812 and a LIFO buffer, 813. They are supplied to the multiplexer 842 in their order of arrival and to the multiplexer 841 in the order of their arrival and in the inverted order of their arrival after having been delayed by the delay 822 and after having been multiplied by the factor −(j.sup.L−1) in 832. The first time that the channel is used, the multiplexers 841 and 842 switch the outputs of the FIFO buffers 810 and 811, onto the OQAM preprocessing modules 851 and 852 respectively. The second time that the channel is used, the multiplexers 841 and 842 switch the outputs of the LIFO buffers 812 and 813, after delay and multiplication by the above mentioned factors, onto the preprocessing modules 852 and 851 respectively. The man skilled in the art will realise that other implementations could equally well be envisaged. In particular, the arrangement composed of 810, 811, 821 can be replaced by a simple memory read in the forward direction the first time that the channel is used and in the inverse direction the second time that the channel is used, the vectors read during the second use of the channel having firstly been multiplied by the factor (j.sup.L−1) before being supplied to the OQAM preprocessing module 852. Similarly, the arrangement composed of 812, 813, 822 can be replaced by a simple memory read in the forward direction the first time that the channel is used and in the inverse direction the second time that the channel is used, the vectors read during the second use of the channel having firstly been multiplied by the factor (j.sup.L−1) before being supplied to the OQAM preprocessing module 851.

[0116] Modules 861-862, 871-872, 881-882 are identical to modules 820, 830, 840 respectively and therefore they will not be described again herein. Output signals from combination modules 881 and 882 are translated into the RF band before being transmitted by antennas 891 and 892 respectively.

[0117] FIG. 9 diagrammatically represents the architecture of an FBMC transmitter according to a second embodiment of the invention; This transmitter is different from that shown in FIG. 8 in that the latter is conventionally implemented in the temporal domain using a polyphase network as described in the above mentioned paper by Hirosaki.

[0118] Element references 910 to 952 are identical to elements 810 to 852 respectively.

[0119] More precisely, the transmitter comprises two FBMC modulation channels. For each of these channels, unlike the first embodiment, the data vector at the output from the OQAM module is supplied to a synthesis filter bank composed of an IFFT module with size N (961, 962), a plurality N of polyphase filters (971, 972) and a plurality N of over samplers (981, 982) by factor N/2, at the output from the different polyphase filters and finally a plurality of delays arranged in parallel and varying from 0 to N−1 sampling periods. The polyphase filters are versions translated in frequency by 2k/T from the prototype filter, the pulse response of which has duration KT.

[0120] Each of the sub-channels at the output from the IFFT are filtered by a polyphase filter. The outputs from the N oversampled and retarded polyphase filters are summated by an adder (981, 982). The output signal from the adder is translated in RF band to output an antenna signal that is then transmitted by the antenna associated with the channel (991, 992).