Method for Estimating the Symbols of a Digital Signal and Receiver for Implementing Said Method

Abstract

The invention relates to a method of estimating symbols carried by a digital signal that is received by a receiver over a communication channel (5), said symbols being multiplexed on orthogonal frequency sub-carriers. The inventive method comprises the following steps in relation to each symbol carried by the digital signal, consisting in: performing at least two transforms towards the frequency domain (8,9) on a portion of the received signal essentially corresponding to the symbol, said transforms being performed with a determined time lag (10) therebetween; estimating the parameters (r0, r1, . . . , rn, r′0, r′1, . . . , r′n) of the communication channel from pre-determined binary information contained in the digital signal; and estimating the symbol from a combination of the result of each of the transforms performed and the estimated communication channel parameters.

Description

Other particularities and benefits of this invention will appear in the following description of non-limiting examples of implementation, referring to the attached drawings, in which:

FIG. 1, already commented, is a diagram illustrating a signal transmission-reception sequence using OFDM multiplexing technology;

FIG. 2 is a diagram illustrating a receiver according to the invention;

FIG. 3 is a diagram illustrating an example of implementation of the invention.

Consider a receiver capable of receiving digital signals carrying multiplexed symbols according to OFDM technology. Said receiver may, for example, be part of a communication station of a wired communication system. In another embodiment illustrated in FIG. 2, the receiver is part of a mobile communication system, for example a base station or a mobile terminal supporting UMTS technology. In this case, it receives digital radio signals at an antenna or an array of antennas 16.

The stream of data received at the receiver in FIG. 2 contains symbols Sm corresponding to respective OFDM symbols sm carried by portions of the signal transmitted over the communication channel to which the receiver is tuned.

For each OFDM symbol gm received having a duration T_s=T_u+T_g, a first fast Fourier transform, called LFFT 8 (for Left FFT), is performed on a portion of the signal carrying that symbol. The calculation window for LFFT 8, i.e., the time span having a duration T_u over which the LFFT 8 is performed, can be positioned in various ways. Said positioning may, for example, be the result of one of the methods described in the prior art. An advantageous example of positioning such a window will be described in greater detail below.

In addition, a second fast Fourier transform, called RFFT 9 (for Right FFT), is performed on the portion of the signal carrying the same OFDM symbol. The position of the calculation window of the RFFT 9 is established relative to the previously determined position of the LFFT 8. In effect, a time lag 10 is selected, to be observed between the LFFT 8 and the RFFT 9.

This time lag 10 may be fixed. In this case, it will be advantageously selected based on the characteristics of the environment containing the communication channel over which the signal is sent. In effect, some environments, for example relatively dense urban environments, foster multiple reflections and diffractions of radio signals, which favors the dispersal of the paths of propagation over time. However, dispersal of paths of propagation is a factor of inter-symbol interference (ISI), as explained above. In order to limit ISI, a relatively long time lag 10 is selected in this type of environment so that the LFFT 8 and the RFFT 9 calculation windows are spread over a sufficient period of time to cover the principal paths of propagation followed by the signal portion involved.

Inversely, other environments do not foster the dispersal of paths of propagation. These include, for example, rural type environments, where signals are most often propagated in a line of sight. This is also the case in wired systems, where the propagation times generally vary little. For such environments, a shorter time lag 10 can be selected, for example on the order of the duration of just a few binary elements.

It is also possible to perform other FFTs in addition to the LFFT 8 and the RFFT 9. In this case, these FFTs are also differently positioned in time, each exhibiting a certain time lag relative to the LFFT 8, for example.

In one advantageous embodiment of the invention, the LFFT 8 and the RFFT 9, as well as any other FFTs, are dynamically positioned in time based on an analysis of the pulse response 11 of the communication channel traveled by the signal sent. This embodiment is particularly interesting when the communication channel is a radio channel and the communication system to which the receiver belongs is mobile.

The pulse response 11 of the communication channel is calculated by the receiver. Advantageously, such a calculation is performed several times, for example periodically, in order to have an up-to-date estimate that takes account of the changes in the profile of propagation. In this case, the calculation can be performed for each OFDM symbol transmitted, i.e., with a period approximately equal to T,. However, a longer period may also be used, when the characteristics of the channel vary less.

The pulse response 11 of the communication channel is calculated according to a conventional method. For example, the receiver involved is a rake receiver. Such a receiver estimates the pulse response of the radio channel by a series of peaks, each peak appearing with a delay corresponding to the propagation time along a specific path and having a complex amplitude corresponding to the attenuation and dephasing of the signal along that path (instantaneous fading).

A pilot channel can be provided to estimate the pulse response in the form of a succession of peaks. The pulse response is estimated by means of an adapted filter. For example, when a coded distribution access technique is used, such as the CDMA (Code Division Multiple Access) used in the UMTS system, the filter is adapted to a pilot spread code with which the transmitter modulates a sequence of known symbols, for example 1 symbols. The positions of the maximums of the output of this adapted filter give the delays used in the fingers of the rake receiver, and the associated complex amplitudes correspond to the values of said maximums.

The output of the adapted filter is then analyzed. This analysis consists of statistical calculations on the output of the adapted filter 11 to determine the delays (τ_i)_0≦i≦M-1 associated with the M paths or echoes observed, as well as the average receiving energies (E_i)_0≦i≦M-1 associated with said paths.

The highest-energy paths detected correspond to principal paths dependent upon the environment. The highest-energy path will correspond, for example, to a direct path between the transmitter and the receiver if they are in direct visibility to one another. The other principal echoes will schematically be those that yield the fewest reflections and diffractions between the transmitter and the receiver.

The energy E_i associated with a path is the mathematical expectation of the square of the module of the instantaneous amplitude A_i(t) of reception on the corresponding path of propagation.

FIG. 3 schematically illustrates the principal paths, for example the paths on which an energy higher than a threshold is detected relative to the radio channel involved. Three paths are seen, characterized by their respective delays τ₀, τ₁ and τ₂ with respect to a time reference (which itself may be determined relative to a first detected path) and by their respective energy levels E₀, E₁, and E₂.

In this embodiment, the estimate of the pulse response 11 is used to position the LFFT 8 and RFFT 9 calculation windows. For example, the LFFT 8 window is positioned relative to the first detected path. For example, it begins with a delay approximately equal to To relative to a time reference, so as to cover the energy peak relating to the first detected path. This is what is shown in FIG. 3.

In addition, as was described above, the calculation window of the RFFT 9 is positioned relative to that of the LFFT 8 by a time lag 10. Advantageously, this time lag 10 is adjusted based on the estimated pulse response 11, as illustrated in FIG. 2. For example, this time lag, called D in FIG. 3, is selected so that the RFFT 9 window is positioned relative to the last detected path in the pulse response 11, for example so as to end approximately at the time corresponding to the delay 2 relative to said last path.

In the example illustrated in FIG. 3, it can therefore be seen that the LFFT and RFFT calculation windows are positioned so as to spread over time spans covering the principal paths of propagation characterizing the radio channel then being considered. This ensures that the two FFTs will be calculated over spans where a large quantity of energy from the signal portion involved is received. In addition, there is some degree of overlap between the calculation windows of the two FFTs, which makes it possible to increase the reliability of the calculation, as will be better understood by reading the continuation of the description.

Of course, other positionings of the fast Fourier transform calculation windows can be applied. For example, the RFFT 9 can be calculated on a window that ends after τ₂, so as to ensure that the echo of the signal portion involved corresponding to the last path on the radio channel will actually be received this window.

Referring to FIG. 2, the use of the FFT calculations performed is described below. As indicated above, each FFT makes it possible to estimate data elements {circumflex over (X)}_n,0, {circumflex over (X)}_n,1, . . . {circumflex over (X)}_n,N-1 each corresponding to one of the frequency sub-carriers used, in order to obtain an estimate {circumflex over (X)}n of the transmitted data X_n corresponding to the content of the OFDM symbol ŝ_m received.

Furthermore, for each FFT calculated, an estimate is made of the radio channel over which the signal is sent. So an estimate of channel 12 is made at the output of the calculation module of the LFFT 8 and an estimate of channel 13 is made at the output of the calculation module of the RFFT 9. Numerous known methods can be used to estimate the channel. For example, the least squares method can be used.

According to this last method, the estimate module of channel 12 estimates parameters r₀, r₁, . . . , r_n corresponding to disturbances appearing on the radio channel, such that {circumflex over (X)}_k=r₀X_k+r₁X_k-1+. . . r_nX_k-n+w_k, where k is an integer greater than n, X_i is the data corresponding to the i^th OFDM symbol transmitted (X_i may, for example, be the Xn of FIG. 1), Xk is the estimate corresponding to the k^th OFDM symbol received, which follows the calculation of the LFFT 8 (for example, {circumflex over (X)}_n in FIG. 1), and wk is a first estimate of the noise affecting the radio channel over which the OFDM symbols are being sent. In order to estimate the parameters r′₀, r′₁, . . . , r′_n, pilot bits transmitted over the radio channel and known to the receiver are conventionally used.

Similarly, the estimate module of channel 13 estimates parameters r′₀, r′₁, . . . , r′_n, such that {circumflex over (X)}_k=r′₀X_k+r′₁X_k-1+. . . r′_nX_k-n+w′_k, where k is an integer greater than n, {circumflex over (X)}{circumflex over (′)}_k is the estimate corresponding to the k^th OFDM symbol received, which follows the calculation of the RFFT 9, and w′_k is a second estimate of the noise affecting the radio channel over which the OFDM symbols are being sent.

A combination module 14 is then used to make an estimate of the data corresponding to the OFDM symbols sent, on the basis of the elements sent by the channel 12 and 13 estimate modules. If the least squares method has been used to estimate the channel, the module 14 can then proceed as follows. A vector {circumflex over (X)} is defined as a concatenation of the 2k+2 estimates {circumflex over (X)}₀, {circumflex over (X)}₁, . . . , {circumflex over (X)}_k and {circumflex over (X)}′₀, {circumflex over (X)}′₁, . . . , {circumflex over (X)}′_k made after the channel estimates made respectively by the modules 12 and 13. In addition, a first convolution matrix R is defined based on the parameters r₀, r₁, . . . , r_n estimated by the channel 12 estimate module. This matrix R, whose dimensions are (k+1)·(k+1), has the following structure:

$R=\left(\begin{array}{cccccccccc}{r}_0& 0& 0& •& •& •& •& •& •& 0\\ r_1& r_0& 0& & & & & & & 0\\ r_2& r_1& r_0& •& & & & & & 0\\ \cdots & r_2& r_1& •& •& & & & & 0\\ r_n& \cdots & r_2& •& •& •& & & & 0\\ 0& r_n& \cdots & •& •& •& •& & & 0\\ 0& 0& r_n& & •& •& •& •& & 0\\ 0& 0& 0& •& & •& •& •& •& \cdots \\ \cdots & \cdots & \cdots & •& •& & •& •& •& 0\\ 0& 0& 0& \cdots & 0& r_n& \cdots & r_2& r_1& r_0\end{array}\right)$

Similarly, a second convolution matrix R′ is defined based on the parameters r′₀, r′₁, . . . , r′_n estimated by the channel 13 estimate module. This matrix R′, whose dimensions are (k+1)·(k+1), has the following structure:

$R=\left(\begin{array}{cccccccccc}{r}_0^{\prime }& 0& 0& •& •& •& •& •& •& 0\\ r_1^{\prime }& r_0^{\prime }& 0& & & & & & & 0\\ r_2^{\prime }& r_1^{\prime }& r_0^{\prime }& •& & & & & & 0\\ \cdots & r_2^{\prime }& r_1^{\prime }& •& •& & & & & 0\\ r_n^{\prime }& \cdots & r_2^{\prime }& •& •& •& & & & 0\\ 0& r_n^{\prime }& \cdots & •& •& •& •& & & 0\\ 0& 0& r_n^{\prime }& & •& •& •& •& & 0\\ 0& 0& 0& •& & •& •& •& •& \cdots \\ \cdots & \cdots & \cdots & •& •& & •& •& •& 0\\ 0& 0& 0& \cdots & 0& r_n^{\prime }& \cdots & r_2^{\prime }& r_1^{\prime }& r_0^{\prime }\end{array}\right)$

Lastly, a vector X is defined, which includes all the k+1 quantities of data X₀, X₁, . . . , X_k corresponding to the OFDM symbols sent, which need to be determined.

It is then possible to write the following equation: {circumflex over (X)}=M·X, where M represents a column concatenation matrix of the block matrices R and R′, i.e.,

$M=\left(\begin{array}{c}R\\ R^{\prime }\end{array}\right).$

This matrix M thus has 2k+2 lines and k+1 columns.

The noises wk and w′_k introduced above are delayed from one another due to the lag 10 existing between the two FFTs performed, so they are independent and do not enter into the equation system to detect the value of the symbols sent.

The vector X can then be estimated as the product of (M^HM)⁻¹·M^H·{circumflex over (X)} where the operator (·)^H is the conjugated transpose.

Of course, many methods other than the least squares method can be used to estimate the channel. In this case, the estimate of the vector X will be adapted to the method used (for example, one of the following methods: a posteriori maximum, MLSE (Maximum Likelihood Sequence Estimator), Viterbi algorithm, etc.).

Once the combination is performed by the module 14, an estimate is obtained of the data X₀, X₁, . . . , X_k corresponding to the signals sent. The reliability of this estimate is greater than that of those obtained by conventional methods because it relies on a combination of estimates made as a result of two different observations. Furthermore, as was indicated above, the time lag between the FFTs makes it possible to achieve some data redundancy which is taken advantage of in the combination 14, all the while ensuring that the final estimate is based on the highest-energy, and therefore the most significant, echoes.

It will be noted that the principles discussed above can be implemented by using a computer program containing the corresponding instructions and run, for example, by a processing unit in the receiver involved.

Claims

1-11. (canceled)

12. A method for estimating symbols carried by a digital signal received by a receiver over a communication channel, the symbols being multiplexed on orthogonal frequency sub-carriers, the method comprising the following steps relative to each symbol carried by the digital signal: executing at least two transforms to the frequency domain on a portion of the received signal that substantially corresponds to said symbol, said transforms being executed one after the other with a predetermined time shift;estimating communication channel parameters (r0, r1, . . . , rn, r′0, r′1, . . . , r′n) from predetermined binary information contained in the digital signal; andestimating said symbol from a combination of the result of each of the executed transforms and the estimated communication channel parameters.