Method and device for generating a multicarrier OFDM signal, method and device for attenuating extrema of such a signal, corresponding computer program products

Abstract

A method is proposed for generating an OFDM type multicarrier signal including OFDM blocks constituted by M carriers modulated by source symbols. The method includes the following steps: interleaving the M symbols of a block of source symbols into R sub-blocks of N interleaved symbols; obtaining a block of M time domain samples corresponding to the block of M source symbols; forming a peak vector containing N maximum amplitudes determined from among the M samples; attenuating the extrema of each sub-block of N time domain samples corresponding to the R sub-blocks of N interleaved symbols by a correction of symbols taking account of the peak vector and delivering R sub-blocks of N corrected interleaved symbols; de-interleaving the R sub-blocks of N corrected interleaved symbols delivering a block of M corrected source symbols; generating an OFDM block of the signals corresponding to the block of M corrected source symbols.

Claims

1. A method for generating an OFDM type multicarrier signal comprising OFDM blocks constituted by M carriers modulated by constellation symbols, called source symbols, a block of M source symbols corresponding to an OFDM block of M carriers, wherein the method comprises the following acts performed by a signal generating device: interleaving the M symbols of a block of source symbols into R sub-blocks of N interleaved symbols, where R, M and N are natural integers such that M=RN; obtaining a block of M time domain samples by discrete Fourier transform of said block of M source symbols; determining N maximum amplitudes among said samples of said block of M time domain samples and forming a vector containing said N maximum amplitudes, called a peak vector; attenuation, by correction of constellation symbols, of extrema of a sub-block of N time domain samples corresponding to the discrete Fourier transform of one of said R sub-blocks of N interleaved symbols, said correction taking account of at least said peak vector and delivering a sub-block of N corrected interleaved symbols, said step of attenuation, repeated for each sub-block among said R sub-blocks of N interleaved symbols, delivering R sub-blocks of N corrected interleaved symbols; de-interleaving said R sub-blocks of N corrected interleaved symbols delivering a block of M corrected source symbols; and generating an OFDM block of said multicarrier signal by discrete Fourier transform of said block of M corrected source symbols.

2. The method for generating according to claim 1, characterized in that, during said step for interleaving the M symbols of a block of source symbols into R sub-blocks of N interleaved symbols, an i-th sub-block of N interleaved symbols, where i is an integer ranging from 1 to R, comprises symbols indexed (i1)+k*R of said block of source symbols, k being an integer ranging from 0 to N1.

3. The method for generating according to claim 1, wherein an element indexed j of said peak vector, where j is an integer ranging from 0 to N1, is equal to said maximum amplitude of the samples indexed j+r*N of said block of M time domain samples with r an integer ranging from 0 to R1.

4. The method for generating according to claim 1, wherein R>1 and wherein said attenuating extrema by correction of constellation symbols is done successively for each of said R sub-blocks of N interleaved symbols, and wherein: said attenuating, performed for the sub-blocks 1 to v1 of N interleaved symbols, where v is an integer ranging from 2 to R, delivers v1 sub-blocks of N corrected interleaved symbols; when said attenuating is performed for a v-th sub-block of N interleaved symbols, the peak vector taken into account by the correction is updated as a function of an amplitude of at least one time domain sample among v1 sub-blocks of N corrected interleaved time domain samples corresponding to the discrete Fourier transform of said v1 sub-blocks of N preliminarily obtained corrected interleaved symbols.

5. The method for generating according to claim 4, said attenuating is performed for said v-th sub-block of N interleaved symbols, said updating of said peak vector taken into account by said correction corresponds, for an element indexed s of said peak vector, s being an integer ranging from 0 to N1, to: an incrementing of said element indexed s by a value corresponding to a difference, weighted by first corrective factor Gup, between an amplitude of a sample indexed s of a sub-block of N corrected interleaved time domain samples corresponding to the discrete Fourier transform of said sub-block ranked v1 of N preliminarily obtained, corrected interleaved symbols and an amplitude of a sample indexed s of a sub-block of N interleaved time domain samples corresponding to the discrete Fourier transform of said sub-block ranked v1 of N interleaved symbols when said difference is positive or null; a decrementing of the element indexed s by a value corresponding to said difference weighted by a second corrective factor Gdown when said difference is negative.

6. The method for generating according to claim 5, wherein said first corrective factor Gup is equal to 1 and said second corrective factor Gdown is equal 0.707.

7. The method for generating according to claim 1, wherein, at the end of said obtaining a block of M time domain samples by discrete Fourier transform of said block of M source symbols, the device implements an act of over-sampling of said block of M time domain samples obtained.

8. The method for generating according to claim 1 wherein said attenuating the extrema by correction of the constellation symbols implements a correction belonging to the group consisting of: an ACE (Active Constellation Extension) correction; a TI-CES (Tone Injection-Constellation Extension Scheme) correction; a CD (Constellation Distortion) correction; a TR (Tone Reservation) correction; and a PCTS (Pre-Constructed Temporal Signal) correction.

9. The method for generating according to claim 1, wherein the method further comprises a preliminary act of pre-correction of said block of M source symbols, a source symbol being pre-corrected by a pre-defined intermediate correction value between a zero correction value and an extreme correction value.

10. A method for attenuating extrema of an OFDM type multicarrier signal, the method comprising the following acts performed by an attenuating device: attenuating, by correction of constellation symbols, of extrema of a block of M time domain samples corresponding to a discrete Fourier transform of a block of N constellation symbols, said correction taking account of a peak vector representing extrema of said multicarrier signal and delivering a block of N corrected symbols; wherein said OFDM type multicarrier signal comprises OFDM blocks constituted by M carriers modulated by constellation symbols, called source symbols, a block of M source symbols corresponding to an OFDM block of M carriers: said block of N constellation symbols is a sub-block among R sub-blocks of N interleaved symbols obtained by interleaving of M symbols of a block of source symbols, where R, M and N are natural integers such that M=RN; said peak vector is obtained by determining N maximum amplitudes among time domain samples obtained by discrete Fourier transform of said block of M source symbols; and de-interleaving R sub-blocks of N corrected interleaved symbols, obtained by repetition of said act of attenuating, for each block of N constellation symbols among said R sub-blocks of N interleaved symbols.

11. A non-transitory computer-readable medium comprising a computer program product stored thereon, comprising program code instructions for implementing a method for generating an OFDM type multicarrier signal, when said instructions are executed by a processor of a signal generating device, wherein the OFDM type multicarrier signal comprises OFDM blocks constituted by M carriers modulated by constellation symbols, called source symbols, a block of M source symbols corresponding to an OFDM block of M carriers, and wherein the method comprises the following acts performed by the signal generating device: a step for interleaving the M symbols of a block of source symbols into R sub-blocks of N interleaved symbols, where R, M and N are natural integers such that M=RN; obtaining a block of M time domain samples by discrete Fourier transform of said block of M source symbols; determining N maximum amplitudes among said samples of said block of M time domain samples and forming a vector containing said N maximum amplitudes, called a peak vector; attenuation, by correction of constellation symbols, of extrema of a sub-block of N time domain samples corresponding to the discrete Fourier transform of one of said R sub-blocks of N interleaved symbols, said correction taking account of at least said peak vector and delivering a sub-block of N corrected interleaved symbols, said step of attenuation, repeated for each sub-block among said R sub-blocks of N interleaved symbols, delivering R sub-blocks of N corrected interleaved symbols; de-interleaving said R sub-blocks of N corrected interleaved symbols delivering a block of M corrected source symbols; and generating an OFDM block of said multicarrier signal by discrete Fourier transform of said block of M corrected source symbols.

12. A non-transitory computer-readable medium comprising a computer program product stored thereon, comprising program code instructions for implementing a method for attenuating extrema of an OFDM type multicarrier signal, when the instructions are executed by a processor of an attenuating device, wherein the method comprises: attenuating, by correction of constellation symbols, of extrema of a block of M time domain samples corresponding to a discrete Fourier transform of a block of N constellation symbols, said correction taking account of a peak vector representing extrema of said multicarrier signal and delivering a block of N corrected symbols; wherein said OFDM type multicarrier signal comprises OFDM blocks constituted by M carriers modulated by constellation symbols, called source symbols, a block of M source symbols corresponding to an OFDM block of M carriers: said block of N constellation symbols is a sub-block among R sub-blocks of N interleaved symbols obtained by interleaving of M symbols of a block of source symbols, where R, M and N are natural integers such that M=RN; said peak vector is obtained by determining N maximum amplitudes among time domain samples obtained by discrete Fourier transform of said block of M source symbols; and de-interleaving R sub-blocks of N corrected interleaved symbols, obtained by repetition of said act of attenuating, for each block of N constellation symbols among said R sub-blocks of N interleaved symbols.

13. A device for generating an OFDM type multicarrier signal comprising OFDM blocks constituted by M carriers modulated by constellation symbols, called source symbols, a block of M source symbols corresponding to an OFDM block of M carriers, wherein the device comprises: an interleaver of M symbols of a block of source symbols into R sub-blocks of N interleaved symbols where, R, M and N are natural integers such that M=RN; a first generator of a block of M temporary samples by discrete Fourier transform of said block of M source symbols; a computer of N maximum amplitudes among the samples of said block of M time domain samples and a second generator of a vector containing said N maximum amplitudes, called a peak vector; an attenuator by correction of constellation symbols, of extrema of a sub-block of N time domain samples corresponding to the discrete Fourier transform of a sub-block of N interleaved symbols, said correction taking account of at least said peak vector and delivering a block of N corrected interleaved symbols; a de-interleaver of R sub-blocks of N corrected interleaved symbols delivering a block of M corrected source symbols; and a third generator of an OFDM block of said multicarrier signal by discrete Fourier transform of the block of M corrected source symbols.

14. A device for attenuation of extrema of an OFDM type multicarrier signal comprising: an attenuator, by correction of constellation symbols, of the extrema of a block of N time domain samples corresponding to a discrete Fourier transform of a block of N constellation symbols, said correction taking account of a peak vector representing extrema of said multicarrier signal and delivering a block of N corrected symbols, said signal OFDM type multicarrier signal comprising OFDM blocks constituted by M carriers modulated by constellation symbols, called source symbols, a block of M source symbols corresponding to an OFDM block of M carriers; an interleaver of M symbols of a block of source symbols into R sub-blocks of N interleaved symbols, where R, M and N are natural integers such that M=RN, the block of N constellation symbols being a sub-block among said R sub-blocks of N interleaved symbols; a computer of N maximum amplitudes among time domain samples obtained by discrete Fourier transform of said block of M source symbols and a generator of a vector containing said N maximum amplitudes, called a peak vector; a de-interleaver of R sub-blocks of N corrected interleaved symbols delivering a block of M corrected source symbols.

Description

LIST OF FIGURES

(1) Other features and advantages of the invention shall appear from the reading of the following description, given by way of an indicative and non-exhaustive example, and from the appended drawings, of which:

(2) FIG. 1 illustrates the steps of methods for generating and attenuating extrema of an OFDM type multicarrier signal according to different embodiments of the invention;

(3) FIG. 2 illustrates the interleaving of the carriers and of the corresponding modulating symbols according to one embodiment of the invention;

(4) FIG. 3 illustrates a device for generating and attenuating extrema of an OFDM type multicarrier signal according to different embodiments of the invention;

(5) FIG. 4 illustrates a device for attenuating extrema of an OFDM type multicarrier signal according to one particular embodiment of the invention;

(6) FIGS. 5a and 5b illustrate variants of implementation of the methods for generating and attenuating extrema according to different embodiments of invention;

(7) FIG. 6 illustrates a device for attenuating extrema of an OFDM type multicarrier signal according to another particular embodiment of the invention;

(8) FIGS. 7a and 7b illustrate the gains obtained in terms of the number of logic gates and the memory size needed to implement a PCTS type method for reducing extrema in using the technique described; and

(9) FIGS. 8a, 8b and 8c present examples of structures of a device for generating an OFDM type multicarrier signal as well as for attenuating the extrema of an OFDM type multicarrier signal according to one particular embodiment of the invention.

DETAILED DESCRIPTION OF THE INVENTION

(10) In all the figures of the present document, the identical elements and steps are designated by a same reference.

(11) The general principle of the technique described consists in obtaining R sub-blocks of N symbols by interleaving a block of M source symbols that are meant to modulate the M carriers of an OFDM block, M being equal to RN. A peak vector is also obtained on the basis of N maximum amplitudes among M time signals corresponding to the discrete Fourier transform of the block of M source symbols. As a result, a method for attenuating extrema can be implemented with reasonable complexity for each of the R sub-blocks of N interleaved symbols in taking account of the peak vector. The de-interleaving of R sub-blocks of N corrected interleaved symbols thus obtained then gives a block of M corrected source symbols corresponding to an OFDM block having extrema that are limited as compared with the original waveform corresponding to the block of M source symbols.

(12) Referring now to FIGS. 1, 2 and 3, a description is provided of a technique for generating an OFDM type multicarrier signal comprising OFDM blocks constituted by M carriers modulated by constellation symbols according to different embodiments of the invention. Such a technique for generating an OFDM signal, implemented in a device for generating a multicarrier OFDM type signal 310, implements an attenuation of the extrema of this signal.

(13) In a step INTERL. 100, a block of M constellation symbols, called source symbols, intended for modulating the M carriers of the corresponding OFDM block, is interleaved by an interleaver 300 so as to obtain the R sub-blocks of N interleaved symbols such that M=RN, with R, M and N being natural integers.

(14) To this end, it is proposed in one embodiment to observe the expression of the signal obtained at the instant t=k.sub.M.Math.T.sub.e, with T.sub.e being the sampling period, k.sub.M being an integer such that 0k.sub.M<M, and M.Math.T.sub.e=T.sub.M the period of an OFDM symbol, at output of the inverse discrete Fourier transform of a block of M source symbols X.sub.m, with m being an integer ranging from 0 to M1:

(15) $x\left(k_M.Math.\mathrm{Te}\right)=\underset{m=0}{\overset{M-1}{.Math.}}{X}_m.Math.e^{2.Math.j.Math..Math.\frac{m}{M}.Math.k_M}$

(16) In choosing N as an integer sub-multiple of M, and in taking R=M/N, it is seen that this equation can then be re-written as:

(17) $x\left(k_M.Math.\mathrm{Te}\right)=\underset{n=0}{\overset{N-1}{.Math.}}\left[X_{R.Math.n}.Math.e^{2.Math.j.Math..Math.\frac{R.Math.n}{M}.Math.k_M}+X_{R.Math.n+1}.Math.e^{2.Math.j.Math..Math.\frac{R.Math.n+1}{M}.Math.k_M}+\mathrm{.Math.}+X_{R.Math.n+R-1}.Math.e^{2.Math.j.Math..Math.\frac{R.Math.n+R-1}{M}.Math.k_M}\right]$

(18) or again:

(19) $x\left(k_M.Math.\mathrm{Te}\right)=\underset{n=0}{\overset{N-1}{.Math.}}{e}^{2.Math.j.Math..Math.\frac{n}{N}.Math.k_M}.Math.\left[\begin{array}{c}{X}_{R.Math.n}+X_{R.Math.n+1}.Math.e^{2.Math.j.Math..Math.\frac{1}{M}.Math.k_M}+\mathrm{.Math.}+\\ X_{R.Math.n+R-1}.Math.e^{2.Math.j.Math..Math.\frac{R-1}{M}.Math.k_M}\end{array}\right]$

(20) In taking

(21) $\left(y,k_M\right)=e^{2.Math.j.Math..Math.\frac{y}{M}.Math.k_M},$
and with (k.sub.M mod N)=k.sub.N, k.sub.N being an integer ranging from 0 to N1, we then obtain the equation (Eq1):

(22) $x\left(k_M.Math.\mathrm{Te}\right)=\underset{n=0}{\overset{N-1}{.Math.}}{X}_{R.Math.n}.Math.e^{2.Math.j.Math..Math.\frac{n}{N}.Math.k_N}+\left(1,k_M\right).Math.\underset{n=0}{\overset{N-1}{.Math.}}{X}_{R.Math.n+1}.Math.e^{2.Math.j.Math.\frac{n}{N}.Math.k_N}+\mathrm{.Math.}\mathrm{.Math.}\mathrm{.Math.}\mathrm{.Math.}\mathrm{.Math.}\mathrm{.Math.}\mathrm{.Math.}\mathrm{.Math.}\mathrm{.Math.}+\left(R-1,k_M\right).Math.\underset{n=0}{\overset{N-1}{.Math.}}{X}_{R.Math.n+R-1}.Math.e^{2.Math.j.Math..Math.\frac{}{N}.Math.k_N}$

(23) According to this equation, the signal x(k.sub.M. Te) can therefore be expressed as the sum of R periodic signals of a period N, weighted by a phase term (y, k.sub.M); these R signals being derived from an equivalent number of N-sized inverse discrete Fourier transforms carried out on sub-blocks 201 of N source symbols obtained by interleaving of the M source symbols of the original block 200.

(24) More specifically, in this embodiment, the symbol indexed k, k being an integer ranging from 0 to N1 of the i-th sub-block 201 of N interleaved symbols, i being an integer ranging from 1 to R, corresponds to the symbol indexed (i1)+k*R of the block of M source symbols 200.

(25) In one variant, the interleaver INTERL. 300 is implemented in the form of a complex matrix interleaver with N lines and M/N=R columns, the latency of which is M clock periods at the frequency F.sub.e=1/T.sub.e, the principle of matrix interleaving being that of writing the data to a memory in rows and reading the data in columns.

(26) The interleaver INTERL. 300 then delivers the R sub-blocks of N interleaved symbols with a view to their correction by an attenuator ATT. EXTREME 303.

(27) To this end, in a step IFFT 101, a block of M time domain samples is obtained by a first generator TF M 301 via the inverse discrete Fourier transform of the block of M source symbols.

(28) In one variant, M is expressed as a power of 2, enabling this discrete Fourier transform to be implemented as a fast Fourier transform.

(29) At the step VEC. 102, a peak vector containing N values of maximum amplitude is determined by a peak vector determining module MAX 302 on the basis of the block of M time domain samples delivered by the first generator TF M 301. More particularly, the peak vector determining module MAX 302 comprises a computer of N maximum amplitude values among the M time domain samples delivered by the first generator TF M 301 and a second generator of a vector containing the N maximum amplitudes.

(30) In the embodiment corresponding to the interleaving of the symbols described here above with reference to the equation (Eq1), it can be seen that the amplitudes of the R samples indexed k.sub.M0, k.sub.M1, . . . , k.sub.MR-1 such that (k.sub.Mr mod N)=k.sub.N, r being an integer ranging from 0 to R1, of the OFDM block obtained by inverse discrete Fourier transform of the block of M source symbols, are all a function of the amplitudes of the samples indexed k.sub.N of the sub-blocks obtained by inverse discrete Fourier transform of the R sub-blocks of N interleaved symbols delivered by the interleaver INTERL. 300 during the step 100.

(31) Thus, if an extremum exists for one of the R samples indexed k.sub.M0, k.sub.M1, . . . , k.sub.MR-1, such that (k.sub.Mr mod N)=k.sub.N, of the OFDM block obtained by inverse discrete Fourier transform of the block of M source symbols, then it can be reduced by action on the value of the samples indexed k.sub.N of the R sub-blocks obtained by inverse discrete Fourier transform of the sub-blocks of N corresponding interleaved symbols.

(32) In this embodiment, the element indexed k.sub.N of the peak vector, where k.sub.N is an integer ranging from 0 to N1, is equal to the maximum amplitude of the samples indexed k.sub.N+r*N of the block of M time domain samples, with R being an integer ranging from 0 to R1.

(33) It can then be seen that, if the element indexed k.sub.N of the peak vector thus defined is above a given threshold, i.e. if the amplitude of one of the samples indexed k.sub.M0, k.sub.M1, . . . , k.sub.MR-1 such that (k.sub.Mr mod N)=k.sub.N of the OFDM block obtained by inverse discrete Fourier transform of the block of M source symbols is above this given threshold, then this information can be used to deduce a constraint of reduction of amplitude of the samples indexed k.sub.N of the sub-blocks obtained by inverse discrete Fourier transform of the sub-blocks of N corresponding interleaved symbols.

(34) The peak vector thus defined can therefore serve as a constraint vector for a method for reducing extrema by correction of symbols applied to each of the R sub-blocks of N interleaved symbols delivered by the interleaver INTERL. 300 at the step 100.

(35) Thus, at a step ATT. EXTREME 103, the attenuator 303 takes the peak vector determined by the peak vector determining module 302 at the step VEC. 102 as the basis for applying a method for reducing extrema by correction of symbols to each sub-block of N interleaved symbols delivered by the interleaver INTERL. 300 at the step 100. More specifically, in the embodiment described here above with reference to the equation (Eq1), if a maximum amplitude indexed k.sub.N in the peak vector is above a given threshold, the method for reducing extrema will tend to reduce the amplitude of the samples of a same index of the R sub-blocks obtained by inverse discrete Fourier transform of the sub-blocks of N interleaved symbols delivered by the interleaver 300 at the step 100 and will do so in a given ratio relative to this given threshold.

(36) In variants, the method for reducing extrema by correction of symbols applied to each sub-block of N interleaved symbols is a known method, for example belonging to the group: an ACE (Active Constellation Extension) correction; a TI-CES (Tone Injection-Constellation Extension Scheme) correction; a CD (Constellation Distortion) correction; a TR (Tone Reservation) correction; and a PCTS (Pre-Constructed Temporal Signal) correction.

(37) The application of the method for reducing extrema by correcting symbols at the R sub-blocks of N interleaved symbols thus leads to the obtaining of R sub-blocks of N corrected interleaved symbols delivered by the attenuator ATT. EXTREMA 303 to a de-interleaver DE-INTERL. 304.

(38) Thus, the attenuator 303 processing the sub-blocks sized N, N being an integer sub-multiple of the size M of the OFDM blocks composing the multicarrier signal considered, the complexity of implementation of the attenuator 303 is reduced relative to an implementation, as encountered in the prior art, that processes he M-sized blocks in their totality.

(39) In a step DE-INTERL. 104, the de-interleaver 304 then carries out an operation that is the reverse of that carried out by the interleaver 300 at the step INTERL. 100 with the aim of delivering a block of M corrected source symbols corresponding to an OFDM block having extrema reduced in comparison with the OFDM block corresponding to the block of M source symbols.

(40) For example, in the embodiment described with reference to the equation (Eq1), it can be seen that the operation that is the reverse of the interleaving operation described with reference to FIG. 2 corresponds to having: the corrected symbol indexed 0 of the block of M corrected source symbols which corresponds to the corrected symbol indexed 0 of the first sub-block of N corrected symbols; the corrected symbol indexed 1 of the block of M corrected source symbols that corresponds to the corrected symbol indexed 0 of the second sub-block of N corrected symbols; . . . the corrected symbol indexed R1 of the block of M corrected source symbols that corresponds to the corrected symbol indexed 0 of the R-th sub-block of N corrected symbols; the corrected symbol indexed R of the block of M corrected source symbols that corresponds to the corrected symbol indexed 1 of the first sub-block of N corrected symbols; the corrected symbol indexed R+1 of the block of M corrected source symbols that corresponds to the corrected symbol indexed 1 of the second sub-block of N corrected symbols; . . . the corrected symbol indexed 2*R1 of the block of M corrected source symbols that corresponds to the corrected symbol indexed 1 of the R-th sub-block of N corrected symbols; . . . the corrected symbol indexed M1 of the block of M corrected source symbols that corresponds to the corrected symbol indexed N1 of the R-th sub-block of N corrected symbols.

(41) In other words, if we consider a block of M corrected symbols, called a concatenated block, obtained by the ordered concatenation of the first to the R-th of the R sub-blocks of N corrected symbols, the modified source symbol indexed m, where m is an integer ranging from 0 to M1, in the block of M modified source symbols corresponds to the corrected symbol indexed (m*N) mod M+(m*N)/M, where . designates the integer part in the concatenated block.

(42) The block of M modified source symbols is then delivered to a third generator TF M 305 which, in a step GENER. 105, generates an OFDM block constituted by M carriers modulated by the M modified source symbols through the implementing of an inverse discrete Fourier transform sized M of the block of M modified source symbols. The OFDM block thus generated from the block of M modified source symbols has extrema with amplitude that is reduced in comparison with the OFDM block generated from the block of M original source symbols by the first generator TF M 301.

(43) In one variant, the technique described is applied not directly to a block of M source symbols but to a block M pre-corrected source symbols. Thus, in a preliminary pre-correction step, the pre-correction module PRE-CORR. 307 delivers a block of M pre-corrected source symbols in an intermediate state between the original source symbols and an extreme correction position, this intermediate state corresponding to a correction of the source symbols as permitted by the method for reducing extrema by correction of symbols, implemented in the attenuator 303. Thus, the implementation by the attenuator 303 of the method considered for reducing extrema by correction of symbols is facilitated, thus reducing its convergence time.

(44) Referring to FIG. 4, we shall now describe a device 410 for attenuating extrema of an OFDM type multicarrier signal according to one particular embodiment of the invention.

(45) In this embodiment, the method for reducing extrema by correction of symbols applied to each of the sub-blocks of N interleaved symbols is a PCTS correction method corresponding to the method described extensively in the French patent application published under number FR 3 003 107.

(46) Here the PCTS method implemented in the attenuator 303 successively processes the R sub-blocks of N interleaved symbols delivered by the interleaver INTERL. 300 on the basis of a block of M source symbols. To this end, the attenuator 303 also takes the peak vector delivered by the peak vector determining module 302 as input data. This module 302, as indicated earlier, comprises a computer of N maximum amplitudes among the M time domain samples delivered by the first generator TF M 301 and a second generator of a vector containing these N maximum amplitudes.

(47) In this embodiment, the block of time domain samples delivered by the first generator TF M 301 is over-sampled by a factor of two by a first over-sampling device 400. Indeed, as described in detail in the French patent application FR 3 003 107, the over-sampling of an OFDM signal at a sampling frequency higher than the natural sampling frequency of the discrete Fourier transform used for its generation enables a better representation of the extrema of the corresponding analog signal. The implementing of a method for reducing extrema on the basis of such an over-sampled digital signal in the time domain thus makes it possible to take account of the extrema more efficiently and therefore leads to better results. In this context, the greater the over-sampling, the more accurate is the time domain representation of the extrema and the greater is the extent to which they are taken into account for their reduction as well. However, the greater this over-sampling, the more complex will be the implementation of the processing operation. In practice, it appears that an over-sampling by a factor of two already gives satisfactory results for most applications.

(48) Thus, if we consider an u-th block of M source symbols formed by the symbols Y.sub.(u.Math.M), Y.sub.(u.Math.M+1), . . . , Y.sub.(u.Math.M+M1), the u-th OFDM block of M corresponding carriers, obtained at output of the first generator TF M 301 at instants that are integer multiples of the sampling period T.sub.e, can be expressed as:

(49) $y\left(\left(u.Math.M+k_M\right).Math.\mathrm{Te}\right)=\underset{m=0}{\overset{M-1}{.Math.}}{Y}_{\left(u.Math.M+m\right)}.Math.e^{2.Math.j.Math..Math.\frac{m}{M}.Math.k_M}$

(50) After over-sampling by a factor of two in the time domain, the first over-sampling device 400 delivers 2*M samples corresponding to the u-th OFDM block taken at the instants that are integer and semi-integer multiples of the sampling period T.sub.e.

(51) In one variant, the first over-sampling module 400 is implemented in the form of a linear-phase finite impulse response (FIR) filter so as to obtain a simple implementation that does not cause any phase distortion of the over-sampled signal.

(52) The corresponding over-sampled digital signal is then delivered to the peak vector determining module MAX 302. If we reconsider the embodiment described with reference to FIGS. 1, 2 and 3 and with reference to the equation (Eq1) and if we take account of the present over-sampling, the maximum amplitude indexed l of the peak vector, l being an integer ranging from 0 to 2*N1, is equal to the maximum amplitude of the samples indexed l+r*2*N of the block of 2*M time domain samples, with R being an integer ranging from 0 to R1.

(53) Thus, the peak vector determining module MAX 302 delivers a peak vector MAXM.sub.u.Math.M[ ] associated with the u-th block of M source symbols which can be expressed as:

(54) $\mathrm{MAX}{M}_{u,M}\left[\right]=\left[\begin{array}{c}{\mathrm{MAX}}_{r=0}^{R-1}\left\{.Math.y\left(\left(u.Math.M+r.Math.N\right).Math.\mathrm{Te}\right).Math.\right\}\\ {\mathrm{MAX}}_{r=0}^{R-1}\left\{.Math.y\left(\left(u.Math.M+r.Math.N+\frac{1}{2}\right).Math.\mathrm{Te}\right).Math.\right\}\\ \mathrm{.Math.}\\ {\mathrm{MAX}}_{r=0}^{R-1}\left\{.Math.y\left(\left(u.Math.M+r.Math.N+l\right).Math.\mathrm{Te}\right).Math.\right\}\\ \mathrm{.Math.}\\ {\mathrm{MAX}}_{r=0}^{R-1}\left\{.Math.y\left(\left(u.Math.M+r.Math.N+\frac{2.Math.N-2}{2}\right).Math.\mathrm{Te}\right).Math.\right\}\\ {\mathrm{MAX}}_{r=0}^{R-1}\left\{.Math.y\left(\left(u.Math.M+r.Math.N+\frac{2.Math.N-1}{2}\right).Math.\mathrm{Te}\right).Math.\right\}\end{array}\right]$

(55) In one variant, the maximum amplitudes of the peak vector are determined by means of two delay lines of a depth N, implemented in the form of a succession of registers clocked at F.sub.e=1/T.sub.e, with N intermediate outputs. A first of the two delay lines is associated with the processing of the M samples taken at instants that are integer multiples of the sampling period T.sub.e among the 2*M values delivered by the first over-sampling module or over-sampler 400. The second delay line is then associated with the processing of the M samples taken at instants that are semi-integer multiples i of the sample period T.sub.e among the 2*M values delivered by the first over-sampling module 400. The input of each delay line is the result of the choice of the maximum amplitude made between that of the sample y((k.sub.N+(r+1).Math.N+u.Math.M).Math.Te) or y((k.sub.N+(r+1).Math.N+u.Math.M+).Math.Te), derived from the first over-sampling module 400, and the samples respectively y((k.sub.N+r.Math.N+u.Math.M).Math.Te) or y((k.sub.N+r.Math.N+u.Math.M+).Math.Te) coming from the output of the two delay lines. Because the delay lines are looped with the function for determining the maximum amplitude disposed at their input and their clocking at F.sub.e=1/T.sub.e, their N intermediate outputs deliver the elements of the peak vector as defined here above at the end of the processing of the u-th block of M source symbols.

(56) On the basis of this peak vector associated with a u-th block of M source symbols, the PCTS method implemented in the attenuator 303 successively processes the R sub-blocks of N interleaved symbols delivered by the interleaver INTERL. 300 on the basis of the block of M source symbols formed by the symbols Y.sub.(u.Math.M), Y.sub.(u.Math.M+1), . . . , Y.sub.(u.Math.M+M1).

(57) Thus, for an r-th sub-block of N interleaved symbols composed of symbols denoted as X.sub.(u.Math.M+r.Math.N), X.sub.(u.Math.M+r.Math.N+1), . . . , X.sub.(u.Math.M+r.Math.N+N1), a time signal corresponding to the inverse discrete Fourier transform of the r-th sub-block of N interleaved symbols is delivered by a Fourier transform module sized N TF N 401. This signal can be expressed at instants that are integer multiples of the sampling period T.sub.e as:

(58) $x\left(\left(u.Math.M+r.Math.N+k_M\right).Math.\mathrm{Te}\right)=\underset{n=0}{\overset{N-1}{.Math.}}{X}_{\left(u.Math.M+r.Math.N+n\right)}.Math.e^{2.Math.j.Math..Math.\frac{n}{N}.Math.k_N}$

(59) In one variant, N is expressed as a power of two enabling this discrete Fourier transform to be implemented as a fast Fourier transform.

(60) The digital signal x(.Math.) is then over-sampled by a factor of two in the time domain by a second over-sampler 402 which delivers a vector INITN[ ] of 2*N samples corresponding to the r-th sub-block of N interleaved symbols of the u-th OFDM block taken at instants that are integer and semi-integer multiples of the sampling period T.sub.e which is expressed as:

(61) 0${\mathrm{INITN}}_{\left(u.Math.M+r.Math.N\right)}\left[\right]=\left[\begin{array}{c}\underset{n=0}{\overset{N-1}{.Math.}}{X}_{\left(u.Math.M+r.Math.N+n\right)}\\ \underset{n=0}{\overset{N-1}{.Math.}}{X}_{\left(u.Math.M+r.Math.N+n\right)}.Math.e^{2.Math.j.Math..Math.\frac{n}{N}.Math.\frac{1}{2}}\\ \mathrm{.Math.}\\ \underset{n=0}{\overset{N-1}{.Math.}}{X}_{\left(u.Math.M+r.Math.N+n\right)}.Math.e^{2.Math.j.Math..Math.\frac{n}{N}.Math.l}\\ \mathrm{.Math.}\\ \underset{n=0}{\overset{N-1}{.Math.}}{X}_{\left(u.Math.M+r.Math.N+n\right)}.Math.e^{2.Math.j.Math..Math.\frac{n}{N}.Math.\frac{2.Math.N-2}{2}}\\ \underset{n=0}{\overset{N-1}{.Math.}}{X}_{\left(u.Math.M+r.Math.N+n\right)}.Math.e^{2.Math.j.Math..Math.\frac{n}{N}.Math.\frac{2.Math.N-1}{2}}\end{array}\right]$

(62) In one variant, the second over-sampling 402 is implemented in the form of a linear-phase finite impulse response filter, or linear-phase FIR filter so as to obtain an implementation that is simple and does not induce any phase distortion of the over-sampled signal.

(63) The vector INITN[ ], which represents the temporal signal corresponding to the r-th sub-block of N interleaved signal, is then delivered to the accumulation module ACCU 403 in such a way as to gradually pre-build the sequence of 2*N samples of the temporal signal corresponding to the signal obtained after correction of the N interleaved symbols and the application of an inverse discrete Fourier transform on the N corresponding corrected interleaved symbols with interpolation by a factor of two.

(64) To this end, the accumulation module ACCU 403 uses the complex vectors GEN.sub.k.sub.N [ ] delivered by the generation module 404 and representing the time domain samples corresponding to the k.sub.N order carrier, k.sub.N being an integer ranging from 0 to N1 at output of the Fourier transform sized N. Thus, the real and imaginary components of GEN.sub.k.sub.N [ ] are the two vectors COS.sub.k.sub.N [ ] and SIN.sub.k.sub.N [ ] expressed, in taking account of the temporal over-sampling by a factor of two, as:

(65) ${\mathrm{COS}}_{k_N}\left[\right]=\left[\begin{array}{c}1\\ \cos \left(.Math.\frac{k_N}{N}\right)\\ \mathrm{.Math.}\\ \cos \left(l.Math..Math.\frac{k_N}{N}\right)\\ \mathrm{.Math.}\\ \cos \left(\left(2.Math.N-2\right).Math..Math.\frac{k_N}{N}\right)\\ \cos \left(\left(2.Math.N-1\right).Math..Math.\frac{k_N}{N}\right)\end{array}\right]\mathrm{and}$${\mathrm{SIN}}_{k_N}\left[\right]=\left[\begin{array}{c}0\\ \sin \left(.Math.\frac{k_N}{N}\right)\\ \mathrm{.Math.}\\ \sin \left(l.Math..Math.\frac{k_N}{N}\right)\\ \mathrm{.Math.}\\ \sin \left(\left(2.Math.N-2\right).Math..Math.\frac{k_N}{N}\right)\\ \sin \left(\left(2.Math.N-1\right).Math..Math.\frac{k_N}{N}\right)\end{array}\right]$

(66) On one variant, the elements of the two vectors COS.sub.k.sub.N [ ] and SIN.sub.k.sub.N[ ] are pre-computed and stored in a memory. In another variant, these values are computed on the fly, for example by means of a CORDIC type algorithm well known to those skilled in the art.

(67) The accumulation module ACCU 403 then also takes account of the complex correction shifts dX.sub.u.Math.M+r.Math.N+k.sub.N=dA.sub.u.Math.M+r.Math.N+k .sub.N+j.Math.dB.sub.u.Math.M+r.Math.N+k.sub.N, delivered by the PCTS correction module CORR. SYMB. 406 and to be applied to the symbol X.sub.u.Math.M+r.Math.N+k.sub.N that are intended for modulating the corresponding k.sub.N order carriers of the r-th sub-block of the u-th OFDM block. A signal vector dN.sub.(u.Math.M+r.Math.N+k.sub.N.sub.)[ ], associated with the k.sub.N order carrier modulated by the shift dX.sub.u.Math.M+r.Math.N+k.sub.N is then determined with the following expression:
dN.sub.(u.Math.M+r.Math.N+k.sub.N.sub.)[ ]=(dA.sub.u.Math.M+r.Math.N+k.sub.N+j.Math.dB.sub.u.Math.M+r.Math.N+k.sub.N.sub.).Math.GEN.sub.k.sub.N[ ]

(68) A complex signal vector ACC.sub.(u.Math.M+r.Math.N+k.sub.N.sub.)[ ] corresponding to the partial sum of the carriers modulated by the corrected symbols up to the k.sub.N1 order is thus obtained with the following expression for k.sub.N0:

(69) ${\mathrm{ACC}}_{\left(u.Math.M+r.Math.N+k_N\right)}\left[\right]={\mathrm{INITN}}_{\left(u.Math.M+r.Math.N\right)}\left[\right]+\underset{n=0}{\overset{k_N-1}{.Math.}}{\mathrm{dN}}_{\left(u.Math.M+r.Math.N+n\right)}\left[\right]$

(70) From this we deduce the vector PREC.sub.(u.Math.M+r.Math.N+k.sub.N.sub.)[ ] that is delivered by the accumulation module ACCU 403 to the PCTS extrema detection module DETECT. EXTREMA 405 and is expressed as follows:

(71) ${\mathrm{PREC}}_{\left(u.Math.M+r.Math.N+k_N\right)}\left[\right]=\{\begin{array}{c}{\mathrm{INITN}}_{\left(u.Math.M+r.Math.N\right)}\left[\right]\mathrm{when}{k}_N=0\\ {\mathrm{ACC}}_{\left(u.Math.M+r.Math.N+k_N\right)}\left[\right]\mathrm{else}\end{array}$

(72) Thus, the vector PREC.sub.(u.Math.M+r.Math.N+k.sub.N.sub.)[ ] represents the partial sum of the k.sub.N first carriers of the r-th sub-block of the u-th OFDM block, these carriers being modulated by the corrections of symbols preliminarily determined during the previous iterations of orders below k.sub.N.

(73) It can be seen that the vector PREC.sub.(u.Math.M+r.Math.N+k.sub.N.sub.)[ ] thus defined represents the sub-block of N time domain samples corresponding to the inverse Fourier transform of the sub-block of N corrected symbols? determined by the application of the PCTS method for reducing extrema TO the r-th sub-block of interleaved symbols when k.sub.N=N1, i.e. after the last iteration of the PCTS algorithm applied to the r-th sub-block of N interleaved symbols considered.

(74) Besides, according to the principle of the PCTS algorithm as described in the French patent application FR 3 003 107, the partial sum defining the vector PREC.sub.(u.Math.M+r.Math.N+k.sub.N.sub.)[ ] makes it possible to take account of the effects of the corrections made to the symbols modulating the carriers of an order below a given order to determine the correction to be made to the symbol modulating the carrier at this same given order.

(75) To this end, the PCTS extrema detection module DETECT. EXTREMA 405 carries out a thresholding operation on the vector PREC.sub.(u.Math.M+r.Math.N+k.sub.N.sub.)[ ] so as to deliver a vector representing extrema present in the signal corresponding to the partial sum described here above to enable the PCTS correction module CORR. SYMB. 406 to determine the correction to be made to the symbol that is meant to modulate the next-order carrier.

(76) More specifically, in this embodiment, the PCTS extrema detection module DETECT. EXTREMA 405 delivers a vector DETECT.sub.(u.Math.M+r.Math.N+k.sub.N.sub.)[ ] to the PCTS correction module CORR. SYMB. 406. The element indexed l of this vector DETECT.sub.(u.Math.M+r.Math.N+k.sub.N.sub.)[ ] is expressed through the equation (Eq2):

(77) ${\mathrm{DETECT}}_{\left(u.Math.M+r.Math.N+k_N,l\right)}=\{\begin{array}{cc}\left({}_{\left(u.Math.M,l\right)}.Math.\frac{.Math.{\mathrm{PREC}}_{\left(u.Math.M+r.Math.N+k_N,l\right)}.Math.}{.Math.{\mathrm{INITN}}_{\left(u.Math.M+r.Math.N,1\right)}.Math.}-1\right).Math.{\mathrm{PREC}}_{\left(u.Math.M+r.Math.N+k_{N,l}\right)}& \\ & \mathrm{when}{}_{\left(u.Math.M,l\right)}.Math.\frac{.Math.{\mathrm{PREC}}_{\left(u.Math.M+r.Math.N+k_N,l\right)}.Math.}{.Math.{\mathrm{INITN}}_{\left(u.Math.M+r.Math.N,l\right)}.Math.}1\\ 0& \mathrm{else}\end{array}\mathrm{with}\text{:}{}_{\left(u.Math.M,l\right)}=\left(\frac{.Math.\mathrm{MAX}{M}_{\left(u.Math.M,l\right)}.Math.}{}\right)$

(78) The PCTS correction module CORR. SYMB. 406 then determines the correction dX.sub.u.Math.M+r.Math.N+k.sub.N=dA.sub.u.Math.M+r.Math.N+k.sub.N+j.Math.dB.sub.u.Math.M+r.Math.N+k.sub.N, to be applied to the symbol X.sub.u.Math.M+r.Math.N+k.sub.N intended for the modulation of the k.sub.N order carrier of the r-th sub-block of the u-th OFDM block in requiring it to be proportional to the product of correlation between the vector DETECT.sub.(u.Math.M+r.Math.N+k.sub.N.sub.)[ ] delivered by the PCTS extrema detection module 405 and the k.sub.N order carrier, i.e.:
dX.sub.k.sub.N=SIGCORR.sub.(u.Math.M+r.Math.N+k.sub.N.sub.)

(79) where >0 and:

(80) ${\mathrm{SIGCORR}}_{\left(u.Math.M+r.Math.N+k_N\right)}=\underset{l=0}{\overset{2.Math.N-1}{.Math.}}{\mathrm{DETECT}}_{\left(u.Math.M+r.Math.N+k_N,l\right)}.Math.{\mathrm{GEN}}_{\left(k_N,l\right)}^{*}$

(81) Referring now to FIGS. 7a and 7b, we describe the gains obtained in terms of numbers of logic gates and memory size needed to implement a PCTS extrema reducing method by using the technique described.

(82) More particularly, the reduction of complexity of the algorithm implemented will depend on the ratio M/N=R. This reduction will be all the greater as this ratio is high. Nevertheless, this reduction of complexity when M/N increases reaches limits related to a reduction of the efficiency of the technique described. This is because all the N detected maximum values of the peak vector MAXM.sub.u.Math.M[ ] associated with the u-th block of M source symbols, are correspondingly greater at initialization and correspondingly closer to the threshold value imposed. Indeed, each maximum value results from the comparison of R samples with one another and this value becomes, on average, all the higher as R is great. This therefore makes it necessary to raise the threshold value so that the algorithm can continue to work accurately but therefore to the detriment of the reduction of the final PAPR that is obtained.

(83) It is therefore worthwhile assessing the ratio M/N=R that offers the reduction that is the best trade-off. Should M and N be expressed as powers of two, the ratio M/N=R can advantageously be taken into account through the parameter defined by the relationship q=ceil(p) with <<1, M=2.sup.p and N=2.sup.q.

(84) FIGS. 7a and 7b then represent, on a logarithmic scale, the compared complexities, respectively in terms of logic cells and of memory, of the two algorithms, i.e. the PCTS algorithm alone (curves 71a, 71b) and the PCTS algorithm encapsulated in the method described as well as the complexities of the corresponding FFT (curves 70a, 70bFFT size M alone) taken as references for two values of : (curves 72a, 72bPCTS and decimation FFTBeta=) and (curves 73a, 73bPCTS and decimation FFTBeta=).

(85) It can be seen that the reduction of complexity becomes very great when becomes smaller than . The result of this is that a satisfactory trade-off can be obtained when is in the region of ; for this value we then obtain the values of M, N and R as given in Table 1.

(86) TABLE-US-00001 TABLE 1 M 16 64 256 1024 2048 8192 32768 N 8 16 64 128 256 512 1024 R 2 4 4 8 8 16 32

(87) Referring now to FIGS. 5a and 5b, we describe alternative modes of implementation of methods for generating and attenuating extrema according to different embodiments of the invention.

(88) It can be seen indeed, in the embodiment described with reference to FIG. 4, that the equation (Eq2) brings into play divisions and is therefore complex to implement.

(89) FIG. 5a then presents an alternative in which the element indexed l of the vector DETECT.sub.(u.Math.M+r.Math.N+k.sub.N.sub.)[ ] is expressed as the product of the element indexed l of the vector PREC.sub.(u.Math.M+r.Math.N+k.sub.N.sub.)[ ] and of a weighting factor equal to the maximum of 0 and of |MAXM.sub.(u.Math.M,l)|.Math.|PREC.sub.(u.Math.M+r.Math.N+k.sub.N.sub.,l)|.Math.|INITN.sub.(u.Math.M+r.Math.N,l)|.

(90) The same thresholding effect as the one sought via the equation (Eq2) is thus obtained but without any need to apply functions that are costly in terms of computation load such as division for example.

(91) Besides, it can be seen that the correction of the interleaved symbols obtained by the successive application of the method for attenuating extrema at each sub-block of N interleaved symbols can, in addition to minimizing extrema present in original peak vector, leads to the generation of new extrema of the corresponding OFDM block that are not present in the original peak vector and therefore not taken into account by default during the application of the method for attenuating extrema in each of the sub-blocks of N interleaved symbols.

(92) In the embodiment described with reference to FIGS. 1, 2 and 3, this generation is linked to the coherent summing of the sub-blocks of time domain samples corresponding to the inverse discrete Fourier transforms of the R sub-blocks of N corrected symbols as given by the equation (Eq1). More specifically, there can thus be a refreshing of one or more signal peaks in samples indexed k.sub.M of the OFDM block considered because of the consistent summing of the R samples indexed k.sub.N of these sub-blocks of time domain samples corresponding to the corrected symbols.

(93) Thus, in one embodiment, it is proposed that the peak vector should be updated between each iteration and the next of the method for reducing extrema by shifting of symbols so as to take account of the potential generation of new extrema as and when the sub-blocks of N interleaved symbols are corrected.

(94) To this end, the invention proposes that the method for reducing extrema should be based not directly on the original peak vector m MAXM.sub.u.Math.M[ ] delivered by the module for determining peak vectors 302 but on a peak vector MAXM.sub.u.Math.M+r.Math.N [ ] updated at the end of the processing of the r-th sub-block of N interleaved symbols of the u-th block of M source symbols by the attenuator 303 and proposes to take account of it in the processing of the r+1th following sub-block.

(95) It can be seen from the diagram illustrated in FIG. 5b that when the processing is initialized, i.e. when the method described is applied to a new block of M source symbols, for example the u-th block, the correction made to the original peak vector MAXM.sub.(u.Math.M)[ ] is null. In the embodiment described with reference to FIG. 4, where the method for reducing extrema applied to the sub-blocks of interleaved symbols is a PCTS method, the element indexed l of the vector DETECT.sub.(u.Math.M+r.Math.N+k.sub.N.sub.)[ ] then depends solely on the element indexed l of the original peak vector MAXM.sub.(u.Math.M,l) relative to the imposed threshold (null signals of the arms A and B). Then, during the processing of the first sub-block of N interleaved symbols by the attenuator 303, the variation of amplitude of the sample indexed l is taken into account in the combination of the signal of the arm A with that of the arm C.

(96) At the end of the processing of the first sub-block of N interleaved symbols by the attenuator 303, the variation of the amplitude of the sample indexed l, corresponding to the inverse discrete Fourier transform of the sub-block of N corrected interleaved symbols, relative to its initial state given by the amplitude of the sample indexed l of the vector INITN.sub.(u.Math.M+1.Math.N), is memorized in an accumulator ACCU 500, having been preliminarily weighted differently depending on its sign.

(97) Thus, in alternative embodiments, the variation of the amplitude of the sample relative to its initial value is accounted for with a factor Gup that is equal to 1.0 when it has been amplified because it could potentially be summed constructively with other same-index samples of the R1 other sub-blocks. The variation is accounted for with a factor Gdown that is equal to 0.707 when it is attenuated, corresponding to a root-square summing (square-root contribution of the sum of the squares of the module) with the amplitude of other same-order samples, so as not to overestimate the impact, on the final signal, of the reduction obtained.

(98) At the beginning of the processing of the second sub-block of N interleaved symbols by the attenuator 303, the signal output from the arm B therefore takes account of the previously made correction.

(99) Thus, with the accumulator ACCU 500 being re-updated at the end of the processing of each sub-block of N interleaved symbols, the updated peak vector MAXM.sub.u.Math.M+R.Math.N[ ] resulting from the summing of the arms B and C, gives an estimation, after the processing of the R sub-blocks of the N interleaved symbols, of the amplitude of the samples in the time domain of the u-th OFDM block corresponding to the inverse discrete Fourier transform of the block of M corresponding corrected source symbols.

(100) Besides, the accumulator ACCU 500 is then set at zero at the end of the processing by the attenuator 303, of the last sub-block of N interleaved symbols of a block of M source symbols given with a view to processing the next block of M source symbols.

(101) In one variant, the peak vector updating module 501 is implemented directly in the peak vector determining module MAX 302. Indeed, it can be seen that the peak vector updating module 501 requires only the final result of the processing of each of the R sub-blocks of N interleaved symbols by the attenuator 303 to update the peak vector (this result can be obtained through the implementing of any method for attenuating extrema by the correction of symbols). This processing result can then be transmitted by the attenuator 303 to the peak vector determining module MAX 302.

(102) Referring now to FIG. 6, we describe a device 401 for attenuating the extrema of an OFDM type multicarrier signal according to one particular embodiment of the invention.

(103) In this embodiment, the method for reducing extrema implemented by the attenuator 303 is a method using ACE symbol correction.

(104) The ACE method is especially recommended in the DVB-T2 standard and the standardizing document ETSI EN 302 755 V1.3.1 (2012-04) Digital Video Broadcasting (DVB); Frame structure channel coding and modulation for a second generation digital terrestrial television broadcasting system (DVB-T2) which describes the corresponding functions that have to be implemented.

(105) These functions include the two main functions of detecting extrema of the OFDM block to which the method is applied and correcting the corresponding symbols. These functions are implemented respectively in an ACE correction module CORR. SYMB. 600 and in an extrema detection module ACE DETECT. EXTREMA 601.

(106) The ACE method has the particular feature of being iterative, which means that, for a real-time implementation, operation at higher frequency can only very partially be enough to compensate for the time needed to carry out all the indispensable iterations (5 to 20 for significant results) to obtain an acceptable level of extrema reduction. The algorithm must therefore be partly or entirely deployed in the context of a real-time application and the two functions of extrema detection DETECT. EXTREMA and correction CORR. SYMB. are then duplicated and cascaded over multiple levels.

(107) Thus, an example of application of the ACE algorithm with four iterations is represented in FIG. 6 in which the method of processing in sub-blocks of N interleaved symbols, as described in the present patent application, is applied.

(108) The Table 2 then presents the complexity and the compared latency of the direct implementation of the ACE method as recommended in the DVB-T2 and that of the ACE method to which the method of processing in sub-blocks, as illustrated in FIG. 6, is applied.

(109) This complexity is accounted for in terms of numbers of discrete Fourier transforms (implemented in FFT form here) that have to be computed to process an OFDM block of M carriers modulated by M source symbols. This is done for three values of M (M=64, 2048 and 8192). Besides, three values of a decimation factor R used to determine the sub-blocks of N interleaved symbols are also considered (R=4, 8 and 16).

(110) It may be recalled that the computational complexity of an FFT of size M develops according to M.Math.Log(M), and all the interleaving functions and de-interleaving functions or detection functions needed to apply the claimed method as assessed as having a complexity equivalent to 0.5 FFT sized M.

(111) The latency for its part is accounted for relative to the latency L.sub.M, corresponding to the time taken to compute an FFT sized M implemented in a same technology.

(112) TABLE-US-00002 TABLE 2 ACE with ACE with ACE with the claimed the claimed the claimed ACE alone method (4 ACE alone method (4 ACE alone method (4 Algorithm (4 Iterations) Iterations) (4 Iterations) Iterations) (4 Iterations) Iterations) M 64 64 2048 2048 8192 8192 R 1 4 1 8 1 16 N 64 16 2048 256 8192 512 Number of FFT 8 1.5 8 1.5 8 1.5 sized M Number of FFT 8 8 8 sized N Complexity 100% 40% 100% 30% 100% 25% Latency 8 .Math. L.sub.M 5 .Math. L.sub.M 8 .Math. L.sub.M 4 .Math. L.sub.M 8 .Math. L.sub.M 3.5 .Math. L.sub.M

(113) Thus, a very significant reduction of complexity of implementation is noted all the more so as M and R are high, but there is also a substantial reduction of latency of the algorithm.

(114) Table 3 shows the same comparisons made for 8 iterations, this time for the algorithm ACE. It can be seen that, for a larger number of iterations, the gains in reduction of complexity and in latency are even greater.

(115) TABLE-US-00003 TABLE 3 ACE with ACE with ACE with the claimed the claimed the claimed ACE alone method (8 ACE alone method (8 ACE alone method (8 Algorithm (8 Iterations) Iterations) (8 Iterations) Iterations) (8 Iterations) Iterations) M 64 64 2048 2048 8192 8192 R 1 4 1 8 1 16 N 64 16 2048 256 8192 512 Number of FFT 16 1.5 16 1.5 16 1.5 sized M Number of FFT 16 16 16 sized N Complexity 100% 26% 100% 20% 100% 15% Latency 16 .Math. L.sub.M 7 .Math. L.sub.M 16 .Math. L.sub.M 5 .Math. L.sub.M 16 .Math. L.sub.M 4 .Math. L.sub.M

(116) It can thus be seen that, for a broadcasting application according to the DVBT2 standard based on an OFDM modulation using 32768 carriers (M=32768) and for which the factor R can be equal to 32, i.e. with the technique described implementing an ACE sub-module using a Fourier transform on 1024 interleaved symbols (N=1024), the expected gains in complexity and in latency are very great.

(117) In one variant, the technique described is applied not directly to a block of M source symbols but to a block of M source symbols pre-corrected by the pre-correction module PRE-CORR. 307. According to the present embodiment corresponding to an attenuator 303 implementing an ACE method by correction of symbols, this intermediate state corresponds to a shifting of the source symbols towards the exterior of the constellation relative to their initial value, i.e. a shifting of the symbols away from the center of the constellation. Indeed, the ACE algorithm tends to search for the minimum values of power of the modulated signal, the consequence of which is to hamper the shifting of the correction symbols towards the exterior of the constellation during the correction of peaks of greater dynamic range. Thus, the pre-correction of the source symbols by the pre-correction module PRE-CORR. 307 takes account of this phenomenon of resistance to shifts towards the exterior of the constellation by pre-positioning the symbols at a pre-determined position between their initial position and an extreme correction position, thus improving the speed of convergence of the overall method.

(118) FIGS. 8a, 8b and 8C present examples of structures of a device 310 for generating an OFDM type multicarrier signal as well as attenuating the extrema 410, 410 of an OFDM type multicarrier signal enabling the implementing of a method described with reference to FIG. 1.

(119) The generating device 310 as well as the device for attenuating extrema 410, 410 comprise a random-access memory (a RAM) 803, 813, 823, a processing unit 802, 812, 822 equipped for example with a processor and driven by a computer program stored in a read-only memory 801, 811, 821 (a ROM or a hard-disk drive). At initialization, the code instructions of the computer program are for example loaded into the random-access memory 803, 813, 823 and then executed by the processor of the processing unit 802, 812, 822.

(120) These FIGS. 8a, 8b and 8c illustrate only one particular way among several possible ways of making the generating device 310 as well as the device for attenuating extrema 410, 410 so that they can carry out certain steps of the method described in detail here above with reference to FIG. 1 (in any one of the different embodiments). Indeed, these steps can be carried out equally well on a reprogrammable computation machine (a PC computer, a DSP processor or a microcontroller) executing a program comprising a sequence of instructions or on a dedicated computing machine (for example a set of logic gates such as an FPGA or an ASIC or any other hardware module).

(121) Should the generating device 310 and/or the device for attenuating extrema 410, 410 be made with a pre-programmable computing machine, the corresponding program (i.e. the sequence of instructions) could be stored in a detachable storage medium (such as for example a floppy disk, a CD-ROM or a DVD-ROM) or non-detachable storage medium, this storage medium being partially or totally readable by a computer or a processor.