EP1425891A1

EP1425891A1 - Multicarrier signal, method for tracking a transmission channel based on such a signal and device therefor

Info

Publication number: EP1425891A1
Application number: EP02779628A
Authority: EP
Inventors: Christophe Brutel; Philippe Mege
Original assignee: EADS Telecom SAS
Current assignee: EADS Secure Networks SAS
Priority date: 2001-09-12
Filing date: 2002-08-09
Publication date: 2004-06-09
Also published as: FR2829642A1; FR2829642B1; US7471728B2; CA2460504A1; US20040240572A1; WO2003024041A1

Abstract

The invention provides a multicarrier system constructed on a time-frequency lattice comprising frames having MxN symbols distributed over M subcarriers each of which is divided into N specific symbol times, each frame comprising P pilot symbols (21) distributed time-wise and frequency-wise so as to cover the frame in a meshed structure so that, in particular, some at least of the M subcarriers contain no pilot symbols and/or no pilot symbol should be transmitted to some at least of the N symbol times. The invention also concerns a method for tracking a transmission channel using such a signal and a device for implementing said method.

Description

MULTI-CARRIER SIGNAL, METHOD FOR TRACKING A TRANSMISSION CHANNEL FROM SUCH A SIGNAL AND DEVICE FOR THE SAME

IMPLEMENTATION

The present invention relates to a multicarrier signal, a method of tracking a transmission channel from such a signal, as well as a device for implementing the method.

It relates to the field of digital transmissions (digital radio transmissions), and finds applications in particular in receivers of digital radiocommunication systems with mobiles, for example professional radiocommunication systems.

(PMR systems, from English "Professional Mobile Radio").

In these systems, digital data is transmitted by modulation of a radio frequency carrier wave. In other words, a radio signal is transmitted on the transmission channel, this signal being modulated to carry the digital information to be transmitted.

We seek to implement modulation techniques which offer better resistance to disturbances suffered by the radio signal during its transmission through the transmission channel. For the most part, these disturbances come from:

- on the one hand the phenomenon of fading ("Fading", in English), which is frequency selective as soon as one exceeds the band of coherence (one speaks in this first case of selective fading ("selective fading""In English)), but which is not frequency selective as soon as the width of the channel is less than the coherence band (in the latter case we speak of flat fading (" flat fading "in English)) . This phenomenon of fading is due to the multiple paths of propagation ("Multipaths", in English) which generate inter-symbol interference (ISI, from English "Intersymbol Interference,") also known by the term of intersymbol distortion. Selective fading is the frequency domain image of multiple propagation paths with long delays between paths (the maximum delay between paths is such that the inverse of this delay is less than the width of the signal band). Flat fading occurs when the delay between multiple propagation paths is small and the maximum delay between paths is such that the inverse of this delay is greater than the bandwidth of the signal. - on the other hand, the amplitude and the phase of or each of the propagation paths can be static (in the sense that they do not vary over time) or on the contrary dynamic (when the propagation conditions vary at course of time). In the dynamic case, the frequency of this phenomenon (also called frequency of fading) and, more generally, the frequency spectrum of fading are related to the speed of the mobile and to the carrier frequency of the transmitted signal. The classic model chosen for the power spectrum of fading is described in the book "Microwave Mobile Communications", by William C. Jakes, Jr., Editions John Wiley & Sons, 1974, pp 19-25 ”), and involves the Doppler frequency fo given by:

where V is the speed of the mobile, c is the speed of light, and f _c the frequency of the carrier.

The power spectral density of the fading is therefore:

with Pfading (f) ^has power of the fading, i.e. the power of the received signal.

We are currently seeking to implement a multicarrier modulation called OFDM (from the English "Orthogonal Frequency Multiplexing"). This modulation technique was chosen for the European standard on digital audio broadcasting systems (DAB systems, from the English "Digital Audio Broadcasting"). It consists in distributing the data to be transmitted over a set of subcarriers ("subcarriers" in English) transmitted in parallel in the radio signal. This results in a flat fading effect vis-à-vis each subcarrier because the bandwidth of each subcarrier is less than the coherence band. In addition, this results in a reduction in the sensitivity of the transmission to the multipath phenomenon.

However, the OFDM technique presents certain constraints in applications of the type envisaged above, in which the spectral efficiency of the transmission is a key characteristic. Indeed, the signal to be transmitted is built on a time-frequency network (also called "lattice" in English). The signal is structured in a frame which is transmitted successively through the transmission channel. Each frame comprises a number M of adjacent subcarriers inside a channel of determined spectral width, each of these subcarriers being divided into N time intervals, called symbol times. The duration of a symbol time corresponds to the duration of transmission of a symbol. A signal frame therefore includes MxN symbols. It is recalled that a symbol corresponds to a determined number of bits of information, for example eight bits, which takes a determined value in an ad-hoc alphabet.

However, it is necessary to introduce pilot symbols in the frame in order to allow the continuation of the transmission channel ("Channel Tracking", in English) by the receiver. It is recalled that a pilot symbol is a symbol introduced into the frame by the transmitter, at a location and with a value which are known to the receiver. Channel tracking is carried out by a sequence of steps implemented by the receiver before the demodulation of the actual signal received. This sequence of steps comprises, on the one hand, the estimation of the value of the channel (that is to say the value of the amplitude and of the phase of the channel) for the pilot symbols of the frame, which produces estimated channel values for all pilot symbols in the frame. It also includes an interpolation of the value of the channel for the other symbols of the frame, which produces MxN values of the channel interpolated for all the symbols of the frame. According to a usual interpolation method, the interpolated channel values for the pilot symbols of the frame correspond to the estimated channel values for these symbols (in other words, the interpolation step keeps, for the pilot symbols, the values produced by the estimation stage).

The purpose of tracking the channel is to estimate the disturbances undergone by the symbols during the transmission of the frame through the transmission channel, which result in particular from the two aforementioned phenomena. This generates a matrix of interpolated values, which is an MxN matrix, which is used for the actual demodulation of the received frame. The presence of the pilot symbols in the frame generates an overhead ("overhead" in English), which penalizes the useful bit rate of the transmission (generally expressed in number of symbols per second).

In the state of the art, the pursuit of the channel is carried out under-carrier ar-sΘUs-p © rteuse .-- A-and and effect-each subcarrier contains pilot symbols, and a temporal interpolation is carried out for each sub -carrier from the channel values estimated from the pilot symbols it contains.

The object of the present invention is to provide an alternative to this state of the art, which in certain cases makes it possible to limit the overhead of the transmission resulting from the pilot symbols.

The invention indeed provides a multicarrier signal built on a time-frequency network defined by a frequency axis and a time axis, the signal comprising ^ frames having MxN symbols distributed over M subcarriers ^{"" of} which chàcϋnë ^~ is divided in N time determined symbols. Each frame - includes P pilot symbols distributed in time and frequency so as to cover the frame according to a mesh structure defined by first and second determined directions corresponding to non-linear vectors, where the numbers M, N and P are non-zero whole numbers, such that on the one hand the projection on the frequency axis of the maximum spacing between two pilot symbols or blocks of adjacent pilot symbols following both said first direction and said second

-direction -or less than-half of the inverse of the maximum delay between the multiple paths at ^~ frâvêrs ^_ re ^" channel of ^{" _} propagatibn, ^" ëFq! re, on the other hand, the projection on the time axis of the maximum spacing between two adjacent pilot symbols or blocks of pilot symbols following both said first direction and said second direction is less than half the inverse of the fading frequency through the transmission channel, and in such a way, moreover, that at least some of the M subcarriers do not contain any pilot symbol and / or that no pilot symbol is transmitted to at least some of the N symbol times. In addition, the frame comprises blocks of pilot symbols, a block of pilot symbols being a group of symbols pilots for which a time stationary condition and a frequency stationary condition of the transmission channel are satisfied.

The fact that at least some of the M subcarriers do not contain any pilot symbol and / or that no pilot symbol is transmitted to some - at least of the N-time-symbols _{r -} allows in certain cases (i.e. -to say -with

- certain -frame-structures and -sub certain assumptions concerning the propagation through the transmission channel which are taken into account) to reduce the overhead of the transmission.

The blocks of pilot symbols allow an improved estimation of the values of the transmission channel, which takes advantage of the diversity of the pilot symbols contained in each block.

The invention also provides a method of tracking a transmission channel from a multicarrier signal as defined above,

- - comprising the following-steps: a) estimation of the channel from the pilot symbols, so as to

-produce-estimated-channel-values-for-at-least-some-of-the-frame symbols which correspond to pilot symbols; b) first interpolations, from said estimated values, according to

- all— the— directions— ar-alleles-to-said- first direction -determined and passing through at least two different pilot symbols for which there is an estimated channel value, so as to produce first interpolated channel values for the symbols of the frame following all these directions;

---- c) seconds- - interpolations ^ - from - said first interpolated values, in all directions parallel to said second determined direction and passing through at least two symbols for which there is a first interpolated value, so as to producing second interpolated channel values for substantially all of the symbols in the frame; d) if applicable, assignment to each symbol of the frame for which there is no second value of the interpolated channel, respectively of the value of the interpolated channel for the symbol of the frame which is closest to it and for which there is a second value of the interpolated channel.

The channel estimation step is performed by a joint estimation from the pilot symbols of blocks of pilot symbols, a block of symbols pilot being a group ^" of pilot symbols for which a stationary condition in time and a stationary condition in frequency of the transmission channel is satisfied.

Such a method makes it possible to reconstruct all of the symbols of the frame, - despite the fact "-that some- at least of the -M subcarriers do not contain any pilot symbol and / or that no pilot symbol is transmitted to at least some of the symbol times N. The pilot symbol blocks allow an improved estimation of the values of the transmission channel, for the reasons given above. The invention also relates to a device comprising means for implementing the The device comprises:

means for estimating the channel from the pilot symbols, making it possible to produce estimated channel values for at least some of the symbols of the frame which correspond to pilot symbols; - first interpolation means ^" for carrying out first interpolations -from r-of-said-estimated-values, -as all -the -directions parallel to said first determined direction and passing through at least two different pilot symbols for which there are an estimated channel value,

^~ of -management-à- produce -premières-values- of the interpolated channel for the symbols of the frame following all these directions; ^" - second interpolation means, for performing second interpolations, from said first interpolated values, according to all

- directions parallel to - said second determined direction - and passing through at least two symbols for which there is a first interpolated value, so as to produce second interpolated channel values for substantially all of the symbols of the frame.

When necessary, that is to say when the frame structure is such that there are symbols in the frame for which there is no second value of the interpolated channel (in particular the symbols at the frame boundary) , the device further comprises allocation means, for assigning to each symbol of the frame for which there is no second value of the interpolated channel, the value of the interpolated channel for the symbol of the - -frame-which is the closest -and- for which there is a second value of the interpolated channel.

Other characteristics and advantages of the invention will become apparent on reading the description which follows. This is purely illustrative and should be read with reference to the appended drawings in which there is shown: in FIG. 1: a diagram illustrating a network-time-frequency on which the signal transmitted on the transmission channel is constructed;

- in Figures 2a to 2c: diagrams illustrating the structure of a frame of a multicarrier signal according to the invention, according to four different examples; - in Figure 3: a diagram illustrating the steps of a method according to the invention;

- in Figures 4a to 4c: tables corresponding to a matrix of values of the channel; - - - - - - -

-in -la- figure-5 — a-diagram- illustrating- the-structure of a frame of a multicarrier signal according to an exemplary embodiment of the invention.

However, according to one advantageous characteristic of the invention, it is noted that in all cases there are subcarriers on which no pilot symbol is emitted. This makes it possible, in certain frame structures, to reduce the over-speed of the transmission. The diagram in Figure 1 illustrates a time-frequency network defined by a frequency axis and a time axis. The network comprises a set of symbols. Symbolically represented by crosses.

The object of this figure is to introduce a certain number of conventions

—Which are used in the following. By convention in effect, the frequency axis is represented vertically and the time axis is represented horizontally. Each symbol is identified by an index m along the frequency axis, and by an index n along the time axis. In the following and in the figures, a symbol whose position along the frequency axis is defined by the index m, and the position along the time axis is defined by the index n is denoted Sm, _n . Finally, the spacing between the symbols along the frequency axis is noted γ ₀ . Similarly, the spacing between the symbols along the time axis is noted τ ₀ . If we note S (t) a signal built on such a network of symbols, we can decompose the signal S (t) in the form:

S (t) = ∑c _rn , n xgm, n (t) = ∑c _m , _n ^χ e ² - ⁱ - ^m T ^o -t ^χ g (t -n -τ ₀ ) m, nm, n 5 where the coefficients c _mn are coefficients corresponding to the value of the symbol Sm _{, n} , and where the functions g _m , _n (t) are prototype functions which define a Hilbert space, and where the function g (t) has the property d 'be invariant by discrete Fourrier transformation (DFT for "Discrete Fourrier Transform") or by inverse Fourrier transformation (IDFT, for "Inverse Discrète Fourrier Transform").

According to intrinsic properties of the time-frequency network, the symbols are orthogonal, which allows simple and efficient demodulation. We then say that the network is orthogonal. By definition, symbols are orthogonal if their scalar product is zero.

This is particularly the case when implementing an OFDM / QAM (from English "OFDM / Quadrature Amplitude Modulation") or OFDM / OQAM (from English "OFDM / Offset Quadrature Amplitude Modulation") modulation. . The prototype functions then define a Hilbert basis of unit dimension, that is to say such that γ ₀ χτ ₀ = 1. Advantageously, the coefficients c _mιn are then complex coefficients. We can write Cm n = a _{rn n} + i ^χ b _mιn , where a _{m> n} and bm.n are real coefficients. This offers the possibility of using amplitude and phase modulation.

However, it is preferable to guarantee a frequency guard band to guarantee orthogonalite in time and in frequency for the OFDM / QAM. Similarly, it is preferable to guarantee a guard interval between the transmission of two consecutive symbols on each subcarrier, so as to limit the intersymbol distortion in the case of OFDM / OQAM. In the case of OFDM / QAM and OFDM / OQAM, these frequency or time guards penalize the transmission rate (expressed in number of symbols per second). When using ^" modulation ΘFDM / IOTA (from- of English" OFDM / lsotropic Orthogonal Transform Algorithm "), orthogonality between symbols is also ensured. With such modulation, the prototype functions define a Hilbert basis of dimension 2, that is to say such that ^"" γôχτcT = ^" 2τ ^" La ^{^} facOn ^{^} dont-One defines u τnetwork = orthogonal time / frequency with tellerie such modulàtio e ^ f ^" p ^ r ^ xêrnple ^" described ^" in ^" the article" Coded Orthogonal Frequency Division Multiplex ", Bernard LE FLOCH et al., Proceedings of the IEEE, Vol. 83, NO. 6, June 1995. The coefficients c _mn are then real or imaginary coefficients pure according to their placement in the frame. They are always mono-dimensional. This only offers the possibility of amplitude modulation. However, it is not necessary to guarantee a guard time between the symbols or between the sub-carriers, which has the advantage of increasing the speed of the transmission.

A frame is defined along the frequency axis and along the time axis, respectively by a frequency band B, and by a duration D. It includes M subcarriers, where M is an integer such that B = Mχγ ₀ . In addition, each subcarrier is divided into N symbol times, where N is an integer such that D = Nxτ ₀ - The frame therefore includes MxN symbols.

By convention, we define a double order relation to identify the location of a symbol- in. the frame along the frequency axis on the one hand, and along the time axis on the other hand. According to this order relation, the symbol_Sj, ι is I symbol which is carried on the first subcarrier (that having the index 1) and which transmitted first on this subcarrier that is to say which is located in the first symbol time (that of index 1). This symbol Sι _, ι is conventionally represented at the bottom left in the figures. Similarly, the symbol SM, N is the symbol which is carried on the last subcarrier of the frame (the one having the index M) and which is transmitted last on this subcarrier, that is to say which is located in the last symbol time (that of index N). This symbol SM, N is conventionally represented at the top right in the figures. Generally, the symbol S _{m, n} is the symbol which is carried on the m-th subcarrier of the frame (the one having the index m) and which transmitted in on this subcarrier in the nth time symbol (that of index N). The diagrams of FIGS. 2a, 2b and 2c represent the structure of a frame of a multicarrier signal according to the invention, in an example where M = 7 and N = 13, and respectively in three distinct cases which will be explained below . The frame is symbolized by a table, the MxN cells of which correspond to the symbols of the frame.

In order to allow continuation of the channel, the frame contains pilot symbols or blocks of pilot symbols, a pilot symbol being a symbol whose location in the frame and whose value is known to the receiver. In the figures, the pilot symbols are represented by black cells, and the other symbols are represented by white cells.

In the case of FIG. 2a, the pilot symbols such as 21 are disjoined inside the frame.

Pilot symbols such as 21 are the symbols S _{m <n} with m = 1 + 3j and n = 1 + 4k, where j is an integer between 0 and 2 and where k is an integer between 0 and 3. These pilot symbols are therefore distributed in time and in frequency so as to cover the frame according to a mesh structure defined by the frequency axis and the time axis. The spacing of pilot symbols along the frequency axis corresponds to twice the spacing between the subcarriers (Le., 2χγ ₀ ), and their spacing along the time axis corresponds to three symbol times (Le., 3χτ ₀ ).

In the case of FIG. 2b, some of the pilot symbols, which are adjacent two by two, form a block of pilot symbols such as 22. We thus say, in the example shown in the figure, four blocks of pilot symbols, each block comprising several pilot symbols. By convention, in the following, the position of a block of pilot symbols in the frame is identified by the position of the pilot symbols of this block which is on the carrier of smallest index, and in time symbol transmitted first (on the figures, it is, for each block, the pilot symbol which is the lowest and the leftmost). Likewise, the size of the block is defined by a dimension along the frequency axis (hereinafter "height", denoted h) expressed in numbers of symbols, and by a dimension along the time axis (hereinafter " length ", noted I), expressed in numbers of symbols. The size of the block is noted hxl, where h denotes the height and I denotes the length. This convention is convenient - in cases - where the blocks - of pilot symbols have regular dimensions (not forming examples of lines, or blocks of pilot symbols, that is to say squares or rectangles). However, it is understood that a block of pilot symbols may have an irregular structure (for example- three pilot symbols - adjacent in pairs but not aligned).

In addition, it is specified that the concept of block of pilot symbols does not necessarily correspond to a concept of adjacency. In reality, the definition of a block of pilot symbols is as follows: the pilot symbols of the same block, which may be adjacent or not, are considered as symbols satisfying a double condition of stationary in frequency and stationary in transmission channel time. We will see later how these parking conditions are defined.

In FIG. 2b, there are thus distinguished six blocks of pilot symbols 22 of dimensions 1x3, which are respectively located on the symbols Sι, _n , and Sι ₂ , _n with n = 1 + 3j where j is an integer between 0 and 2.

In the case of FIG. 2c, certain pilot symbols, although being separate as in the case of FIG. 2a, nevertheless form blocks of pilot symbols such as 23, in the sense defined above. Using the conventions set out above, the dimensions and the position of the blocks 23 in FIG. 2c are identical, respectively, to those of the blocks 22 in FIG. 2b.

Of course, ~ one or more isolated pilot symbols as shown here ^~ figϋré ^~ 2to, ùrTδirplϋsiëu ^'rs ^"blocks ^*""AE piloted adjacent symbols as shown in Figure 2b, and / or one or more blocks of disjoint pilot symbols as shown in Figure 2c, can be simultaneously present and distributed in a frame.

The blocks of pilot symbols can be used for an improved estimation of the values of the transmission channel, which takes advantage of the diversity of the pilot symbols contained in each block.

In the three cases envisaged in the examples of FIGS. 2a to 2c described above, the pilot symbols or blocks of pilot symbols are distributed in frequency and in time so as to cover the frame according to a mesh structure defined by the direction of the frequency axis on the one hand, and by the direction of the time axis on the other hand.

According to a characteristic of the invention, at least some of the M subcarriers do not contain any pilot symbol. Alternatively or additionally, no pilot symbol is transmitted to at least some of the N symbol times.

As a result, interpolated values covering almost all of the symbols in the frame cannot be obtained by frequency interpolations only or by temporal interpolations only, as is the case in the prior art. A more complex channel tracking method is required, as will now be described with reference to the diagram in Figure 3.

Thanks to this characteristic, it is however possible to reduce the number of pilot symbols contained in the frame, in proportions which differ according to the frame structure and according to the characteristics of the propagation-which-are-taken into account during the choice. of the distribution of pilot symbols.

Characteristics of propagation through the transmission channel

—Are-defined-by-the-maximum-frequency-of-the-variations-of-fainting (called "fading frequency" in the jargon of those skilled in the art) and the maximum delay between multiple paths. In an example, for a propagation type HT ("Highly Terrain") which is the most constraining, it is established that the fading frequency is equal to 148.2 Hz (hertz) for a maximum speed of movement of mobiles equal to 200 km / h (kilometer per hour) and for a carrier frequency equal to 400 MHz (Megahertz), on the one hand, and that the maximum delay between the multiple paths corresponds to ± 7.5 μs (microsecond), i.e. one maximum delay between the most advanced path and the most delayed path of 15 μs, on the other hand.

In order to allow the channel to be continued according to the method of the invention, thanks to only the estimated channel values obtained for pilot symbols present in the frame, the distribution of the symbols in the frame must respect two conditions respectively called frequency condition and condition temporal and respectively denoted Cf and Ct below. These conditions- are derived from Shannon's condition for reproducibility of sampled signals. It is expressed here by:

Cf: the maximum spacing between two adjacent pilot symbols (or blocks of pilot symbols) along the frequency axis is less than the inverse of the maximum delay between the multiple paths across the propagation channel; moreover it is preferable to have at least 3 pilots or 3 blocks of pilots placed regularly or not spaced along the frequency axis;

Ct: the maximum spacing between two adjacent pilot symbols (or blocks of pilot symbols) along the time axis must be less than the inverse of the fading frequency through the propagation channel; in addition it is preferable to have at least 3 pilots or 3 blocks of pilots placed regularly or not spaced along the time axis;

Using the example of a HT type propagation considered above, -o can show that the inverse of the maximum delay-delay (1/15-μs) is equal to 66.67 kHz (kilohertz). The maximum spacing entered ^" where rsyrnboles pilots

-adjaeents-FollowingH -frequency axis must then be less than 66.7 kHz.

In the case where the spacing γ ₀ between the subcarriers is equal to 2 kHz, this means that there must be, at a determined symbol time, a pilot symbol

1ransrτîrs ^~ iaα minimum tôutës ^' les 33 ^' subcarrierë "^{""" ~ "~" * "} The preferred additional condition will ideally place at least 3 symbols or 3 blocks of symbols on the frequency axis regardless of the width of the channel and therefore whatever the number of subcarriers.

Furthermore, we can say that the inverse of the ^" fading frequency (which is equal to twice the Doppler frequency), is equal to 6.76 ms (milliseconds). The spacing between pilot symbols along the time axis must therefore be less than 6.76 ms. In the case where a symbol time is equal to 250 μs (ie a flow rate of 4 kilosymbols per subcarrier), this means that there must be, for a specific carrier, at least one pilot symbol every 27 symbols transmitted. The preferred additional condition will ideally place at least 3 symbols or 3 blocks of symbols on the time axis regardless of the duration of the frame and therefore regardless of the number of symbols transported in each frame on each subcarrier. The distribution of the pilot symbols in the frame, as well as the value of these pilot symbols, which are known to both the transmitter and the receiver since it is an intrinsic characteristic of the frame structure, is respected by the transmitter during the construction of the transmitted signal. The diagram of FIG. 3 schematically illustrates the steps of the channel tracking method according to the invention, which are implemented at the level of the receiver. These steps are applied to a frame of the signal as transmitted through the transmission channel and as received by the receiver. They precede the actual demodulation operations. In a step 31, the value of the channel is estimated from the pilot symbols, so as to produce estimated channel values C0 _min for at least some of the symbols of the frame corresponding to pilot symbols. It is possible that estimated values are thus obtained for all the symbols of the frame corresponding to pilot symbols. However, this is not necessarily the case, in particular when the estimation is carried out by joint estimation of the pilot symbols of a block. Indeed, in this case, it may be sufficient to assign the estimated value which results from this joint estimation, only to certain pilot symbols of the block only, the value of the channel corresponding to the other pilot symbols being obtained by one next steps.

The values thus estimated are stored in an ad hoc memory structure, for example a table having MxN cells. The diagram in FIG. 4a shows the content of such a memory structure, for example a table of values, in the case of the frame shown in FIG. 2a. In a step 32, first interpolations are then carried out, from said estimated values stored in the memory structure, in all directions parallel to the first determined direction and passing through at least two different pilot symbols, so as to produce first values of the interpolated channel C1 _{m, n} covering the symbols of the frame in all these directions.

When the mesh structure is defined by the direction of the frequency axis on the one hand, and by the direction of the time axis on the other hand, these first interpolations are, ^" according to a first embodiment, - interpolations along the frequency axis. They are therefore frequency interpolations. They are carried out for all symbol times at which pilot symbols are emitted. second embodiment, these first interpolations are interpolations along the time axis, and are therefore temporal interpolations and are performed for all the subcarriers containing pilot symbols.

The values thus interpolated are stored in the aforementioned memory structure. In the case for example of the aforementioned first embodiment (frequency interpolations) this memory structure then contains values C1 _{m, n} as shown in FIG. 4b, that is to say with m between

1 and M and with n = 1 + 4j, where j is an integer between 0 and 3. In the case, for example, of the aforementioned second embodiment (temporal interpolations) this memory structure then contains values C1 _mn as shown on FIG. 4c, that is to say with m = 1 + 3j, where j is an integer between 0 and 2, and with n between 1 and N.

In a step 33, second interpolations are then carried out, from the first interpolated values stored in the memory structure, in all the directions parallel to said second determined direction, so as to produce second values of the interpolated channel C2 _min to substantially the all symbols of the frame.

The values thus interpolated are stored in the aforementioned memory structure. This memory structure then contains values C2 _mn as shown in FIG. 4d, that is to say with m between 1 and M and with n between 1 and N.

When the mesh structure is defined by the direction of the frequency axis on the one hand, and by the direction of the time axis on the other hand, these second interpolations are temporal interpolations in the aforementioned first embodiment, and frequency interpolations in the aforementioned second embodiment. These interpolations are performed for all symbol times or for all the subcarriers, respectively in the first or in the second embodiments mentioned above. It will be noted that the -second interpolations are carried out starting from the values obtained by the first interpolations, that is to say from values which can be interpolated values (and not necessarily estimated values). In a manner known per se, -the first interpolations- 32 and / or the

-second interpolations 33 each comprise, successively a Fourrier transformation operation (DFT) or an inverse Fourrier transformation operation (IDFT), then a zero padding operation ("0 padding" in English), and finally a transformation operation of Fourrier- inverse (IDFT) or Fourrier transformation (DFT) respectively.

It is possible, in certain cases, that the second interpolations are not sufficient to reconstruct the values of the channel corresponding to all the symbols of the frame. In particular, the values of the channel corresponding to the symbols situated at the limits of the frame may not be able to be thus obtained.

If necessary, in a step 34 (FIG. 3), we then assign to each value of the channel corresponding to a symbol of the frame not covered by the second interpolation, respectively the value of the interpolated channel corresponding to the symbol of the frame. the closest - which is covered by these second interpolations (Le., for which there is a second value of the interpolated channel).

In the above, it was considered that the mesh structure corresponding to the distribution of the pilot symbols in the frame was defined by the direction of the frequency axis on the one hand and by the direction of the time axis on the other go. This is however only a special case, which is convenient for the description of the invention and for the implementation of the first and second interpolations. However, this is by no means mandatory.

In fact, in the general case, the pilot symbols can be distributed in frequency and in time in the frame so as to cover the halftone according to a mesh structure defined by first and second determined directions corresponding to non-collinear vectors. These directions are - therefore not parallel to each other and inclined with respect to the directions of the frequency axis and the time axis.

If we again consider the case of FIG. 2a, these first and second directions can be non-parallel diagonals passing through pilot symbols such as 21.

- - In this case, the -condition- Cf above is expressed as follows:

Cf: the projection on the frequency axis of the maximum spacing between two adjacent pilot symbols or blocks of pilot symbols following both said first direction and said second direction is less than the inverse of the maximum delay between the multiple paths through the propagation channel;

Likewise; the condition Ct above is expressed as follows: Ct: the projection on the time axis of the maximum spacing between two pilot symbols or blocks of adjacent pilot symbols following both said first direction and said second direction is less than the inverse of the fading frequency through the transmission channel

Although a priori less natural, the case of a mesh structure defined by a first and a second determined direction corresponding to directions different from that of the frequency axis (or the time axis) and l The time axis (or the frequency axis respectively), can in certain cases make it possible to further reduce the overhead of the transmission. A compromise will then have to be found with the increase in the complexity of the

^~ Treatτnent ^" which results from it. ^{" ~ ~} II was mentioned above, on the subject of the concept of pilot symbol block within the meaning of the present description, a double condition of stationary in time and stationarity in frequency of the propagation channel. These two conditions can be expressed in terms of maximum spacing of the pilot symbols, as will now be explained. As was said previously with regard to the example of an HT type propagation (for a maximum mobile speed V of 200 km / h, for a carrier frequency f _c of 400 MHz and for a throughput of 4 kilosymbols by subcarriers), the frame must include pilot symbols (or pilot blocks) with -a spacing on the ^~ time axis- (or a spacing projected on the time axis), called time spacing, which must be less than the inverse of the frequency of the fading, i.e. - say that it must include a pilot symbol every 27 symbols maximum. This- maximum spacing -of -27— symbols— on the time axis

-corresponds to a sampling of the propagation channel (fading) carried out more quickly (even only slightly faster) than the occurrence of successive fading holes (zero crossing of the fading on the time axis). Between two successive fading holes, the fading phase has rotated by π (PI number). On, for example, a tenth of this period between fading holes, i.e. over a period corresponding to 2.7 successive symbols (in the following we will consider for practical reasons a group having 2 successive symbols along the axis times), the fading will have rotated by π / 27. If the fading in the middle ^" of ^~ this period ^~ of 2 symbols has a certain value F _m determined, the fading at the end of this period of 2 symbols will have a value F _f which will be ^~ very close to F _m x ë ^{ιxπx, 54} : So we have a quadratic error given by: ε ² = | ^F _m - F _f = | F _m x [2 ^χ sin (π / (2 ^χ 54))] ² = | F _m ^χ 0.00338

A signal to noise ratio of: 24.71 dB. Similarly if the fading at the beginning of this period of 2 symbols has a certain determined value Fd, in the middle of this period there is a given quadratic error jDar: = ψ _m τF _d f = || F _m fx [2 x sin (π 1 (2 x 54))] ² = || F _m x 0.00338

The same signal-to-noise ratio of 24.71 dB. We can therefore consider without any drawback that the channel is stationary in time over a period corresponding to 2 successive symbol times.

Likewise, we said earlier that (for a type propagation

HT, which has a maximum delay between paths of 15 μs, and for a spacing between sub-carriers of 2 kHz) the frame must include pilot symbols (or pilot blocks) with spacing along the frequency axis (or a spacing projected on the frequency axis), called spacing ^~ qurdoirêtre frequency ^{^} lower than the inverse of the delay ^"between the maximum multipath, a pilot symbol every 33 subcarriers at the maximum. This spacing of 33 subcarriers on the frequency axis corresponds to a sampling frequency channel more frequently (even only-slightly more frequently) than the occurrence of successive frequency selectivity holes (zero-crossing-level-of the signal received at certain frequencies). Between two successive frequency selectivity holes the phase of the fading has rotated by π. On, for example, one tenth of this space between frequency selectivity holes, ie 3.3 subcarriers (we will consider, for practical reasons, 3 subcarriers), fading will have rotated by πx 2/33. If the fading among these three sub-carriers has a certain value f _m determined for fading sub-carrier in the higher frequency ^"that group of 3 subcarriers will have a value f _f which

- will-be-very-close to ^ -ë - * - ³³ *! - - - - - _.__ _ - ^" _ ^" So we have a quadratic error given by ^• ^{~~ ~ "} ε ² = | F _m - F _f = | F _m fx [2 χ sin (π / (2 χ 33))] ² = | F | ² X 0.00906

Or a signal to noise ratio of 20.43 dB.

Similarly if the fading for the lowest frequency subcarrier of this group of 3 subcarriers has a determined value F _d , in the middle of this period there is a quadratic error given by: ε ² : HF _m - F _d f = | F _m fx [2 x sin (π / (2 x 33))] ² = || F _m | ^{2 χ} 0.00906 Or the same signal to noise ratio of 20.43 dB.

We can therefore consider without any drawback that the channel is stationary in frequency on 3 successive subcarriers. As a consequence of the above, pilot symbols which are not spaced by more than two symbols along the time axis nor by more than three symbols along the frequency axis can be considered to satisfy a double condition of parking in time and in frequency.

Note that the considerations on the stationary fading are to be evaluated according to the problem to be treated, that is to say in particular the characteristics of the propagation envisaged, the speed of the mobile, the carrier frequency. In examples, the frame includes blocks of pilot symbols

(defined in the sense that a block of pilot symbols is a group of adjacent or non-adjacent pilot symbols for which the condition of stationary in time and stationary in frequency of the transmission channel is satisfied.

In particular, it can be provided that the frame comprises at least one block of six pilot symbols whose dimensions along the frequency axis and along the time axis correspond respectively to three symbols and to two symbols. As a variant or in addition, provision may be made for the frame to include at least one block of three pilot symbols whose dimensions along the frequency axis and along the time axis correspond respectively to one symbol and to three symbols.

As a variant or in addition, it can also be provided that the frame comprises at least one block of three pilot symbols whose dimensions along the frequency axis and along the time axis correspond respectively to two symbols and to two symbols.

The diagram in FIG. 5 represents the structure of a frame of a multicarrier signal according to an exemplary embodiment of the invention. In this example, the signal occupies a frequency band B = 44 kHz

(kilohertz) within a 50 kHz wide channel. In addition, the spacing between the subcarriers is γ ₀ = 2 kHz. The frame therefore comprises M = 22 subcarriers.

In addition, the frame duration is D = 20 ms (millisecond). The flow rate on each subcarrier is 4 kilosymbols / s so the time spacing between the symbols is τ ₀ = 250 μs. In other words, the frame has N = 80 symbol times.

The frame therefore comprises MxN = 1760 symbols. Among these, there are 206 symbols which are pilot symbols. In other words, P = 206. According to this example, the frame comprises 32 blocks such as 51, of six (6) pilot symbols each, the dimensions of which along the frequency axis and along the time axis correspond respectively to three (3) symbols and to two ( 2) symbols. Their respective locations in the frame, identified by the location of the piloted symbol ^" of the block considered which is in ^" the lowest frequency subcarrier and in time the symbol transmitted first (Le., the lowest and most left symbol), are the locations of the symbols Sm, n (it is recalled that m and n are integer indices which locate-position-of the symbol along the frequency axis and along the time axis-respectively), with m included in the set {1, 7 , 14,20} and with n - 1 + 11χj, where j is an integer included in the set [0; 7].

These 32 blocks are used for the pσursuite of the channel according to the method which was described above.

In addition, the frame further comprises a first additional block 52 of six pilot symbols whose dimensions along the frequency axis and along the time axis correspond respectively to three symbols and to -two symbols. - It also includes an additional -second- block 53 of eight pilot symbols whose dimensions along the frequency axis and along the time axis correspond respectively to four symbols and to two symbols.

The respective locations of the additional block 52 and the additional block 53 in the frame, identified by the location of the pilot symbol of the block considered which is in the lowest frequency subcarrier and in symbol time transmitted first (Le. , the symbol at the bottom and the left most), are the locations of the symbols Sm _{, n} with the couple (m, n) included ^" da ^" ήs ^"~ set of couples {(4,1), (10, 1)}. In other words, the blocks 52 and 53 are positioned on the symbols S ₄ , ι and Sio.i respectively. The blocks of additional pilot symbols 52 and 52, in combination with the blocks 51 which are adjacent to them, are used by the receiver for frame synchronization.

Claims

1. Multicarrier signal constructed on a time-frequency network defined by a frequency axis and a time axis, the signal comprising frames having MxN symbols distributed over M subcarriers each of which is divided into N determined time symbols, each frame comprising P pilot symbols distributed in time and frequency so as to cover the frame according to a mesh structure defined by first and second determined directions corresponding to non-collinear vectors, where the numbers M, N and P are non-zero integers , so that on the one hand the projection on the frequency axis of the maximum spacing between two pilot symbols or blocks of adjacent pilot symbols in both said first direction and said second direction is less than half of l inverse of the maximum delay between multiple paths through the propagation channel, and that, on the other hand, the projection on the time axis the maximum spacing between two adjacent pilot symbols or blocks of pilot symbols in both said first direction and said second direction is less than half the inverse of the fading frequency through the transmission channel, and so , in addition, that at least some of the M subcarriers do not contain any pilot symbol and / or that no pilot symbol is transmitted to at least some of the N symbol times, in which the frame comprises blocks of pilot symbols, a pilot symbol block being a group of pilot symbols for which a stationarity condition in time and a stationarity condition in frequency of the transmission channel is satisfied.

2. Signal according to claim 1, wherein said first direction is the direction of the frequency axis and said second direction is the direction of the time axis.

3. Signal according to claim 1 or claim 2, wherein the frame comprises at least three pilot symbols or blocks of pilot symbols spaced along the frequency axis and / or at least three pilot symbols or three blocks of pilot symbols spaced along the time axis.

4. Signal according to any one of the preceding claims, in which the frame comprises at least one block of six pilot symbols, the dimensions of which along the frequency axis and along the time axis correspond respectively to three symbols and to two symbols .

5. Signal according to any one of the preceding claims, in which the frame comprises at least one block of three pilot symbols, the dimensions of which along the frequency axis and along the time axis correspond respectively to one symbol and to three. symbols.

6. Signal according to any one of the preceding claims, in which the frame comprises at least one block of three pilot symbols, the dimensions of which along the frequency axis and along the time axis correspond respectively to two symbols and to two symbols.

7. Signal according to any one of claims 1 to 3, in which M = 22, N = 80 and P = 206.

8. The signal as claimed in claim 7, in which the frame comprises 32 blocks of six pilot symbols each, the dimensions of which along the frequency axis and along the time axis correspond respectively to three symbols and to two symbols, and whose respective locations in the frame, identified by the location of the pilot symbol of the block considered which is in the lowest frequency subcarrier and in the symbol time transmitted first, are the locations of the symbols S _mn , where m and n are integer indices which locate the position of the symbol along the frequency axis and along the time axis respectively, with m included in the set {1, 7,14,20} and with n = 1 + 1 1χ j, where j is an integer included in the set [0; 7].

9. The signal as claimed in claim 8, in which the frame further comprises a first additional block of six pilot symbols whose dimensions along the frequency axis and along the time axis correspond respectively to three symbols and to two symbols, thus a second additional block of eight pilot symbols whose dimensions along the frequency axis and along the time axis correspond respectively to four symbols and to two symbols, the respective locations of these first and second additional blocks in the frame, identified by the location of the pilot symbol of the block considered which is in the lowest frequency subcarrier and in the symbol time transmitted first, are the locations of the symbols Sm, _n with the couple (m, n) included in the set of couples {(4.1), (10.1)}.

10. A method of tracking a transmission channel from a multicarrier signal according to claim 1, comprising the following steps: a) estimation of the channel from the pilot symbols, so as to produce values of the channel estimated for at least some of the symbols of the frame which correspond to pilot symbols; b) first interpolations, from said estimated values, in all the ^" directions parallel to said first determined direction and passing through at least two different pilot symbols for which there is an estimated value of the channel, so as to produce first values of the channel interpolated for the symbols of the frame in all these directions; c) second interpolations, from said first interpolated values, in all directions parallel to said determined second direction and passing through at least two symbols for which there is a first interpolated value, so as to produce second interpolated channel values for substantially all of the symbols in the frame; d) if applicable, assignment to each symbol of the frame for which there is no second value of the interpolated channel, respectively of the value of the interpolated channel for the symbol of the frame which is closest to it and for which there is a second value of the interpolated channel, __ according to which the step of estimation of the carjal is carried out by a joint estimation from the pilot symbols of blocks of pilot symbols, a block of pilot symbols being a group of pilot symbols for which is satisfied a stationary condition in time and a stationary condition in frequency of the transmission channel.

- 11. The method of claim 10, wherein said first direction being the direction of the frequency axis and said second direction being the direction of the time axis, said first interpolations are time interpolations, and said second interpolations are --inter-polations-frequencies.

12. The method of claim 10 or claim 1 1, wherein the first interpolations (32) and / or the second interpolations (33) successively comprise a Fourrier transformation operation or reverse Fourrier transformation, a stuffing operation of zero, and an inverse Fourrier transformation or Fourrier transformation operation respectively.

- - - —13. Device for the implementation of a process according to any one of claims 10 to 12, comprising:

means for estimating the channel from the pilot symbols, making it possible to produce estimated channel values for at least some of the symbols of the frame which correspond to pilot symbols;

first interpolation means, for carrying out first interpolations from said estimated values, in all directions parallel to said first determined direction and passing through at least two different pilot symbols for which there is an estimated channel value, so as to produce first interpolated channel values for the symbols of the frame in all these directions;

- second interpolation means, for performing second interpolations, from said first interpolated values, in all directions parallel to said second determined direction and passing through at least two symbols for which there is a first interpolated value, so as to producing second interpolated channel values for substantially all of the symbols in the frame.

14. Device according to claim 13, further comprising allocation means for assigning to each symbol of the frame for which there is no second value of the interpolated channel, the value of the interpolated channel for the symbol of the frame which is the closest and for which there is a second value of the interpolated channel.