This application is the U.S. national phase of the International Patent Application No. PCT/FR2007/052363 filed Nov. 19, 2007, which claims the benefit of French Application No. 06 55106 filed Nov. 24, 2006, the entire content of which is incorporated herein by reference.
The field of the present invention is telecommunications. Within that field, the invention relates more particularly to digital communication. Digital communication includes wireless radio communication and cable communication. The communication transmission medium is commonly called the transmission or propagation channel, originally with reference to an aerial channel and by extension with reference to any channel.
The invention relates to transmission and reception techniques specifically adapted to estimate transmission channels in a multi-antenna system of the MIMO (multiple-input multiple-output) or MISO (multiple-input single-output) type using multiple carriers and equalization in the frequency domain or the time domain. In a system with multiple transmit or receive antennas, there are as many transmission channels as there are transmit antenna/receive antenna pairs; these channels are sometimes called sub-channels. Estimation consists in estimating the impulse response of each of these channels. The invention applies to multiple-carrier multiple-antenna systems typically using at least two transmit antennas. However, the invention applies equally to a system using only one transmit antenna.
These systems use a frame of content, at the input of the transmitter, that consists of payload data symbols, i.e. symbols that code the information of an input signal, and at the output of the same device, that is divided in time and in frequency for transmission on a plurality of carriers.
The invention can be applied to uplink communication (from a terminal to a base station) and to downlink communication (from a base station to a terminal).
One example of application of the invention is the field of fixed or mobile radio communication, notably for fourth generation and later systems typically referred to as B3G (Beyond 3rd Generation) systems. These systems include MC-CDMA (multi-carrier code division multiple access) systems and OFDMA (orthogonal frequency division multiple access) downlink or uplink systems using an MIMO (multiple-input multiple-output) transmission scheme in which data to be transmitted is divided between time-frequency frames including pilot symbols and possibly null carriers. The invention applies in particular to any type of system using OFDM modulation, for example OFDMA or LP-OFDM systems, and to IFDMA systems.
Standard transmission methods include a modulation step. If differential modulation (non-coherent system) is not used, it is imperative for the receiver to estimate the propagation channel (coherent system) in order to be able to equalize the received signal and detect the bits transmitted. Differential modulation as applied to multiple-antenna systems is not at present considered a promising candidate for high bit rate communications systems. It doubles the noise level, which degrades performance by around 3 decibels (dB).
The channel estimation function is therefore particularly important for multi-antenna systems because the performance of these systems is directly linked to channel estimation in the receiver. In multiple-antenna systems, this function must estimate the various channels connecting a transmit antenna to a receive antenna independently of each other.
Several transmission techniques are already known that are suitable for transmission channel estimation in a multiple-antenna system including a plurality of transmit antennas. These techniques include those that rely on reference symbols, called pilot symbols, inserted into the time-frequency frame on transmission. These pilot symbols are known to the receiver and they enable it to estimate the transmission channels corresponding to each transmit antenna. Theoretically, the capacity of MIMO systems increases linearly with whichever is the lower: the number of transmit antennas and the number of receive antennas. In practice, the usable spectral efficiency falls as the number of antennas increases because of the necessary presence of pilot symbols in the frame.
Various techniques exist for inserting pilot symbols into the time-frequency frame transmitted by an antenna. The set of all pilot symbols for the same time-frequency frame forms a training sequence. In a multiple-antenna system there are as many training sequences as there are transmit antennas.
A known technique for inserting pilot symbols and shown in
where Δt is the time offset between the pulses transmitted by the various antennas. For the impulse responses of the various sub-channels not to interfere with each other, the condition L≦Δt must apply, where L is the maximum spread of the delays of the various sub-channels. Moreover, a condition for determining all impulse responses is Nt×L≦Np, where Nt is the number of transmit antennas TXi. To meet this latter condition, the time offset Δt is chosen to satisfy the condition Nt×Δt≦Np.
The paper by M-S. Baek, H-J. Kook, M-J. Kim, Y-H. You, and H-S. Song, “Multi-Antenna Scheme for High Capacity Transmission in the Digital Audio Broadcasting”, IEEE Transactions on Broadcasting, Vol. 51, No. 4, December 2005 and the paper by I. Barhumi, G. Leus, and M. Moonen, “Optimal Training Design for MIMO OFDM Systems in Mobile Wireless Channels”, IEEE Transactions on Signal Processing, Vol. 51, No. 6, June 2003 rely on this technique. In the first paper, all the sub-carriers of an OFDM symbol are dedicated to channel estimation, which enables the receiver to recover the various impulse responses ahead of OFDM demodulation. In contrast, in the second paper, multiplexing the payload data symbols and the pilot symbols, with the pilot systems shared over one or more OFDM symbols to form a training sequence, implies that the operation of recovering the impulse responses on reception is effected after OFDM demodulation. This operation employs a matrix A constructed from the training sequence and a Fourier matrix of appropriate dimension. Estimating the coefficients of the various impulse responses entails multiplying the received signal demodulated by the pseudo-inverse matrix of this matrix A. The methods disclosed have the advantage of preserving the disposition of the pilot symbols of a single antenna-frame between the various transmit antennas, using the same set of sub-carriers for all pilot frequencies, and avoiding the mandatory presence of null pilot symbols. They therefore offer high spectral efficiency. Their advantage is that they estimate the channel for all modulated carriers.
An objective of the invention is to be even more efficient than known methods of transmitting time-frequency frames with inserted pilot symbols forming a training sequence when used in a system with at least one antenna, in order to improve performance.
To this end, the present invention consists in a transmission method, a transmission module, a channel estimation method, and a channel estimator module.
A transmission method according to the invention is suitable for a system with Nt transmit antennas TX1 to TXNt and Nr receive antennas, where Nt and Nr are greater than or equal to 1, a transmit antenna being separated from a receive antenna by a transmission sub-channel, and is implemented by Nt multiple-carrier transmitters each including a multiplexing and modulation module using NFFT orthogonal functions forming orthogonal symbols intended to be transmitted by the Nt transmit antennas. The method is noteworthy in that it includes the steps of:
replicas of the basic training sequence {tilde over (c)}(p) such that at least one replica is time offset from the basic sequence, with Ki≧1 and at least one Ki≧2, subject to the constraints that the extreme time offset between replicas must be less than the number Np of pilot carriers and that the minimum time offset between replicas of two training sequences is greater than or equal to the maximum spread of the delays of the sub-channels; and
A transmission module according to the invention includes a multiplexing and modulation module using NFFT orthogonal functions, suitable for a multiple-antenna system using Nt transmit antennas TX1 to TXNt, Nr receive antennas, where Nt and Nr are greater than or equal to 1, and a time-frequency frame for each transmit antenna including pilot symbols and data symbols of a payload signal, the data symbols and the pilot symbols being frequency-modulated by the multiplexing and modulation module to form orthogonal symbols that are transmitted by a transmit antenna in the form of a multiple-carrier signal using NFFT carriers and using Np pilot carriers. This module is noteworthy in that it includes:
replicas of the basic training sequence such that at least one replica is time offset from the basic sequence, with Ki≧1 and at least one Ki≧2, subject to the constraints that the extreme time difference between replicas is less than the number Np of pilot carriers and that the minimum time offset between replicas of two training sequences is greater than or equal to the maximum spread of the delays of the sub-channels connecting a transmit antenna to a receive antenna; and
A method according to the invention for estimating transmission sub-channels in a multiple-antenna system using Nt transmit antennas, where Nt is greater than or equal to 1, at least one receive antenna, and a time-frequency frame for each transmit antenna including Np pilot symbols forming a training sequence and data symbols of a payload signal, each of the Nt training sequences being the sum of Ki replicas of a basic training sequence, either time offset or not, such that the values Ki are greater than or equal to 1 and at least one value Ki is greater than or equal to 2, being known to the receiver and enabling the receiver to estimate an impulse response corresponding to the transmission channel separating a transmit antenna from a receive antenna concerned, the data symbols and the pilot symbols being frequency-modulated by a multiplexing and modulation module using NFFT orthogonal functions to form orthogonal symbols that are transmitted by the transmit antennas in the form of a multiple-carrier signal with NFFT carriers. This method is noteworthy in that it includes:
A module according to the invention for estimating transmission channels in a multiple-antenna system using Nt transmit antennas, where Nt is greater than or equal to 1, at least one receive antenna and a time-frequency frame for each transmit antenna including pilot symbols forming a training sequence and data symbols of a payload signal, the Nt training sequences being known to the receiver and enabling the receiver to estimate Nt impulse responses corresponding to the Nt transmission channels respectively separating one of the transmit antennas from the receive antenna concerned, the data symbols and the pilot symbols being frequency-modulated by a multiplexing and modulation module using NFFT orthogonal functions to form orthogonal symbols that are transmitted by the transmit antennas in the form of a multiple-carrier signal using NFFT carriers with Np pilot carriers. This module is noteworthy in that it includes:
The transmission method, the estimation method, and the associated modules and devices of the invention solve the stated problem. Transmitting at least one training sequence including at least two replicas of a basic training sequence at least one of which is time offset produces at least two estimates of the same impulse response and so an average of those estimates can be computed. This improves the estimate of the impulse response of a channel.
In one particular embodiment, a guard interval is added before transmitting an orthogonal symbol and the constraint on the minimum value of the time offset is complied with by choosing a minimum time offset equal to the value of the guard interval.
In one particular embodiment, the replicas are advantageously determined in the frequency domain and the time offsets for the replicas are obtained by inserting a phase offset that is easy to compute. In a first embodiment, the phase offset values are chosen in a particular order, which is known to at least one receiver associated with a receive antenna of the system. In a second embodiment, the phase offset is a function of a carrier index.
In one particular embodiment, the basic training sequence is spread over a plurality of orthogonal symbols.
In one particular embodiment, each of the impulse response estimates is computed from the Np pilot carriers over a time interval Δt. This two-fold limitation advantageously reduces the computations required and therefore reduces the computation power necessary to implement the method.
In one particular embodiment, computation of the impulse responses is repeated for each receive antenna concerned of the multiple-antenna system. This embodiment is suitable for systems including a plurality of receive antennas.
The invention consists further in a transmitter for a multiple-antenna system. The transmitter includes at least one transmission module as defined above.
The invention consists further in a receiver for a multiple-antenna system. The receiver includes at least one transmission channel estimator module as defined above.
The invention consists further in a multiple-antenna system including at least one receiver and/or one transmitter as defined above.
In a preferred embodiment, the multiple-antenna system typically has at least two transmit antennas and is a MIMO system or a MISO system. In the remainder of this document, a MIMO system is to be understood as a system with a plurality of transmit antennas, regardless of the number of receive antennas (Nr≧1).
In a preferred embodiment, the steps of the method are determined by instructions of a transmission, respectively estimation, program in an electronic circuit such as a microchip that can be included in an electronic device such as a transmitter, respectively, a receiver. The transmission, respectively estimation, method of the invention can equally well be executed when this program is loaded into a computation unit such as a processor or the like the operation of which is then controlled by the execution of the program.
Consequently, the invention applies equally to a computer program, notably a computer program on or in an information medium, adapted to implement the invention. This program can use any programming language and take the form of source code, object code or a code intermediate between source code and object code, such as a partially-compiled form, or any other form desirable for implementing a method of the invention.
The information medium can be any entity or device capable of storing the program. For example, the medium can include storage means such as a ROM, for example a CD ROM or a microelectronic circuit ROM, or magnetic storage means, for example a floppy disk or a hard disk.
Alternatively, the information medium can be an integrated circuit incorporating the program and adapted to execute the method in question or to be used in its execution.
Moreover, the transmission or estimation program can be translated into a transmissible form such as an electrical or optical signal, which can be routed via an electrical or optical cable, by radio or by other means. The program of the invention can in particular be downloaded over an Internet-type network or the like.
Other features and advantages of the invention become apparent in the course of the following description given with reference to the appended figures provided by way of non-limiting example:
The situation shown is one with one transmit antenna and one receive antenna for transmitting time-frequency frames using training sequences. According to the invention, the training sequence associated with the transmit antenna consists of at least two replicas of a basic training sequence, that are time offset relative to each other. In this illustration, the basic training sequence is a time pulse. Each replica is used to determine an estimate of the response of the transmission channel between the transmit antenna and the receive antenna. According to the invention, the impulse response of the channel is then determined as the average of those two estimates.
According to the invention, in a multiple-antenna system using a plurality of transmit antennas, the training sequence of at least one transmit antenna consists of at least two replicas of a basic training sequence that are time offset relative to each other. In this illustration, the basic training sequence is a time pulse and the training sequence for the antenna TX1 consists of two replicas that are time offset relative to each other. Each of these replicas is used to determine an estimate of the response of the transmission channel between the transmit antenna TX1 and the receive antenna RXj concerned. According to the invention, the response of the transmission channel between the transmit antenna TX1 and the receive antenna RXj is then determined as the average of these two estimates.
A transmission method 1 of the invention is suitable for a system using Nt transmit antennas and Nr receive antennas, where Nt and Nr are both greater than or equal to 1. A transmit antenna is separated from a receive antenna by a transmission sub-channel. The method is implemented by Nt multi-carrier transmitters each including a multiplexing and modulation module using NFFT orthogonal functions forming orthogonal symbols to be transmitted by the Nt transmit antennas TX1 to TXNt. The system uses for each transmit antenna one time-frequency frame containing pilot symbols and payload data symbols. The data symbols and the pilot symbols are frequency-modulated by the multiplexing and modulation modules to form orthogonal symbols that are transmitted by a transmit antenna in the form of a multi-carrier signal using NFFT carriers with Np pilot carriers.
A first step 2 of the method 1 determines a basic training sequence {tilde over (c)}(p), where p is the index of the pilot carrier.
This sequence is typically determined in the frequency domain by determining the Np pilot carriers and sharing them over one or more orthogonal symbols. These orthogonal symbols are OFDM symbols if the transmit orthogonal modulation is OFDM modulation. If the basic training sequence is shared over a plurality of orthogonal symbols, the pilot carriers of the orthogonal symbols can be identical, different, or partially overlapping. For example, one basic training sequence is one time pulse as illustrated by
A second step 3 of the method determines under constraint a number Kn replicas of index n of the basic training sequence such that at least one replica is time offset from the basic training sequence, where:
The values Ki are equal to or greater than 1 and at least one value Ki is greater than or equal to 2. The constrained determination of replicas, whether time offset or not, is typically effected in the frequency domain by introducing a phase offset of value φn by multiplying the basic training sequence with a phase-rotation factor conforming to the expression:
where 1≦n≦Kn and with the following constraints satisfied.
A first constraint is that the extreme time offset between the various replicas of the basic training sequence determined beforehand is less than the number Np of pilot carriers, in other words:
Kn×Δt≦Np (3)
A second constraint is that the minimum time offset between two replicas of two training sequences associated with two transmit antennas i and j, {tilde over (c)}(p)i and {tilde over (c)}(p)j, with i≠j, is greater than or equal to the maximum spread of the delays of the various channels:
L≦Δt (4)
where L is the maximum spread of the delays of the various sub-channels.
Moreover, a condition for determining the set of all impulse responses is
Nt×L≦Np (5)
This condition is automatically satisfied if the time offset Δt chosen satisfies the above two constraints.
In a third step 4, the method determines the training sequence for each transmit antenna TXi.
For at least one transmit antenna TXi the training sequence is the sum of Ki replicas of the basic training sequence, where Ki≧2, with at least one replica from the basic sequence time offset.
In a preferred embodiment of the invention, the training sequences are given by the following expression, on condition that constraints (3), (4), and (5) are satisfied:
In the particular situation where the value Ki≧2 is equal to 2, the training sequence of at least one antenna TXi corresponds to computing the expression:
For each of the other antennas TXi, the training sequence includes at least one replica of the basic training sequence phase offset by φi, with φi≧0, which corresponds to computing the expression:
with Ki≧1 being the number of replicas constituting the training sequence of the antenna TXi and subject to the constraints (3), (4), and (5) being satisfied.
In one particular embodiment, the training sequence associated with each transmit antenna is the sum of the K replicas of the basic training sequence. For any transmit antenna TXi, the training sequence is preferably given by the following expression, on condition that the constraints (3), (4), and (5) are satisfied:
Here Kn=Nt×K.
The next example is given by way of illustration. The parameters of the system take the following values: NFFT=Nmod=Np=8, Δt=2. NFFT, Nmod, and Np are respectively the size of the FFT, the number of modulated carriers, and the number of pilot carriers. The number g of orthogonal symbols over which the training sequence is spread is taken as equal to 1. The system is a multiple-antenna system with two transmit antennas (Nt=2). Each training sequence is constructed from K1=K2=K=2 replicas. This yields:
where {tilde over (c)}(p) is the pilot symbol from the basic training sequence transmitted at the carrier frequency with index p. The constraint Kn×Δt≦Np is satisfied since (2+2)×2=8≦8. The pilot symbol {tilde over (c)}(p) is arbitrarily chosen to be equal to 1 for all pilot carriers. The pattern of the OFDM symbol containing the basic training sequence is represented in Table 1 in Appendix A. The first column of Table 1 represents the frequency indices and the second column the associated data type (P=pilot symbol).
This finally yields for the training sequences of the two antennas:
The estimation method 10 of the invention estimates transmission channels in a multiple-antenna system using at least one transmit antenna and one receive antenna, Nt≧1 and Nr≧1. A transmission channel separates a transmit antenna TXi from a receive antenna RXj. Each transmit antenna of a multiple-antenna system transmits one time-frequency frame. A time-frequency frame determines the placement in time of the payload data symbols and the pilot symbols on the various carriers. The time-frequency frame can further include null symbols, but as these are not involved in the estimation process they are not referred to in the description. Nor does the description cover guard carriers, because they are not involved in the estimation process either. The pilot symbols for the same transmit antenna form a training sequence. The estimation method of the invention is suitable for signals transmitted by a transmission method conforming to a first embodiment of the invention but can be used for other types of signals, although they would be transmitted with lower efficiency.
The description that follows of the estimation method of the invention assumes that the signals are transmitted by a transmission method conforming to a first embodiment of the invention.
The replicas constituting the Nt training sequences are time offset relative to each other by at least one time slot Δt. The training sequences are known to the receiver and enable it to estimate Nt impulse responses that correspond to the Nt transmission channels. For a given transmit antenna, the data symbols and the pilot symbols are frequency-modulated by a multiplexing and modulation module using NFFT orthogonal functions, commonly called an orthogonal multiplexer, to form orthogonal symbols that are transmitted by the transmit antenna connected to the orthogonal multiplexer in the form of a multiple-carrier signal using NFFT carriers and using Np pilot carriers. Depending on the time-frequency frame concerned, the pilot symbols can be divided over one or more orthogonal symbols. In one particular embodiment the modulation is OFDM modulation.
An estimation method of the invention is applied in the receiver after demodulation of a time signal received by the receive antenna RXj concerned and uses an FFT of a certain size to obtain a frequency-domain signal Rj(n) and extract Np pilot symbols from it. In one particular embodiment, the demodulation is OFDM demodulation. The size of the FFT is typically determined as a function of the size NFFT of the FFT used for transmission. These sizes are typically chosen to be identical. On transmission the FFT is of the inverse type whereas on reception the FFT is of the direct type.
At time n, the signal from the receive antenna RXj can be expressed in the frequency domain, i.e. after suppressing the guard interval and after OFDM demodulation, in the form of a column vector of dimension NFFT:
where Xi(n) is a column vector of dimension NFFT, the OFDM symbol transmitted at the time n by the antenna TXi;
To simplify the computation, which is advantageous, the estimation method uses the demodulated signal according to equation (10) only over the duration Δt:
where F′ is the matrix consisting of the Δt first columns of the Fourier matrix F of size NFFT×NFFT and hj,i′ is a column vector of length Δt of which the row elements beyond L are at zero:
hj,i′=[hj,iT,0Δt-L]T (12)
The notation diag{x} denotes a diagonal matrix having the column vector x on its diagonal:
The Fourier matrix F is square, with dimensions NFFT×NFFT, and of the following form:
The matrix F′ is deduced directly from the matrix F:
The estimation method 10 of the invention includes a step 11 of computing a matrix A and, for a receive antenna RXj, a step 12 of computing in the time domain estimates of the Nt impulse responses of the Nt channels respectively separating a transmit antenna TXi from the receive antenna RXj concerned. The method further includes a step 13 of computing the average of the impulse response estimates.
To obtain the Nr×Nt channels, steps 12 and 13 must be repeated for each of the receive antennas RXj.
The computation step 11 computes a matrix A constructed in the form of blocks from training sequences and the Fourier matrix with dimensions NFFT×NFFT. The number of blocks is equal to the product of the number of orthogonal symbols of the basic training sequence by the sum of the values Ki. Each block is associated with a transmit antenna and an orthogonal symbol over which the training sequence is spread. A block is determined by the product of the diagonal matrix formed by the pilot symbols linked to the block and the Fourier matrix. To limit the computation, which is advantageous, the Fourier matrix is limited to the pilot carriers and to Δt time samples, because Δt is greater than or equal to the maximum spread of the temporal response of the channel.
Each OFDM symbol transmitted at the time n by the transmit antenna TXi, Xi(n) can be broken down into a vector containing the data symbols and a vector containing the pilot symbols:
Xi(n)=Si(n)+Bi(n) (16)
where:
Si(n) is the payload data symbol vector of dimension NFFT and Bi(n) is the pilot symbol vector of the same dimension.
The equation (11) can therefore be expressed in the form:
Knowing that Δt≧L, it is possible to impose Δt=L. With this condition, the accumulation of the received vectors corresponding to the g orthogonal symbols transmitted by each transmit antenna containing pilot symbols constituting a training sequence can be expressed in the form:
where:
Rj is a column vector of dimension NFFT×g, T is a matrix of dimensions (NFFT×g)×(Kn×Δt) containing the payload data symbols during the training sequences of the Nt transmit antennas;
A is a matrix of dimensions (NFFT×g)×(Kn×Δt) containing the pilot symbols of the training sequences of the Nt transmit antennas;
hj′ is the column vector of dimension Kn×Δt containing the Kn expressions of the estimates of the Nt impulse responses;
hj′=[hj,1T, . . . , hj,KnT]T (21)
Ξj is the column vector of dimension NFFT×g containing g representations of the Fourier transform of additive Gaussian white noise.
The matrix A determined during the step 11 is therefore expressed in the form:
The step 12 of computing the Kn estimates of the Nt impulse responses in the time domain multiplies the vector Rj, the expression for which in the frequency domain at the output of the demodulation corresponds to the Np pilot carriers, by the pseudo-inverse matrix of the matrix A using the Barhumi technique:
ĥj=A+Rj (23)
in which the symbol + signifies the following “pseudo-inverse” operation:
A+=(AHA)−1AH (24)
To eliminate interference terms caused by the data present in the OFDM symbol, it is necessary to have the product of A+ by T equal to a null matrix, which is made possible by imposing disjoint sets of pilot symbols and data. Moreover, the computation of the impulse responses can be simplified by considering in the equations only the frequencies dedicated to channel estimation, i.e. by limiting the computations to the pilot carriers:
ĥj=Ã+{tilde over (R)}j (25)
where:
{tilde over (R)}j is the matrix corresponding to Rj considering only the pilot carriers, and where:
the expression for the matrix à is as follows:
or:
The vector expression of the first estimate of the impulse response ĥj,i of the channel separating the particular transmit antenna i, for i taking a value between 1 and Nt, from the particular receive antenna j, for j taking a value between 1 and Nr, is obtained by effecting a selection of the Np successive samples of the column vector ĥj computed for the receive antenna j starting with the sample (i−1)Δt. For the second estimate of the impulse response of the channel ĥj,i, selection starts from the sample Nt×Δt+(i−1)Δt=(Nt+i−1)Δt. For the nth estimate of the impulse response of the channel ĥj,i, selection starts from the sample (n×Nt+i)Δt.
The step 13 computes the average in the time domain of the Ki impulse response estimates. Alternatively, this average is determined in the frequency domain. The step is preferably repeated for all the channels for which Ki is strictly greater than 1.
In a method conforming to one particular embodiment of the invention, the computation of the Nt×Nr impulse responses described above for the Np pilot carriers is complemented by interpolation that can be linear in time or in frequency (one dimensional (1D) interpolation) to obtain the coefficients of each channel for all carriers modulated with payload data. Interpolation can be of an order greater than 1. Interpolation can be effected in the frequency domain by multiplying the Nt vectors of Δt points by a matrix of size NFFT×Δt extracted from a Fourier matrix of size NFFT×NFFT.
The transmission system SY includes a multiple-carrier transmitter device EM and a receiver device RE. The transmitter device EM is connected to Nt transmit antennas TX1 to TXNt, where Nt≧2. The receiver device RE is connected to Nr receive antennas RX1 to RXNr, where Nr≧1.
In the example shown, the transmitter EM includes a channel coding module CdC, a bit interleaver EB, a binary symbol coding module CBS, a space-time coding module CET, and as many OFDM multiplexers MX as there are transmit antennas.
The channel coding module CdC codes input source data corresponding to one or more signals, typically telecommunication signals, to provide at the output coded data dc, for example data coded using a convolutional code.
The bit interleaver interleaves the bits of the coded data in accordance with a particular interleaving law to provide interleaved coded data dce.
The binary symbol coder module CBS converts the interleaved coded data dce into complex data symbols sc, for example by BPSK, QPSK or 16QAM modulation.
The space-time coding module CET determines from the complex data symbols sc a two-dimensional matrix of data symbols sd, for example using Alamouti-type coding, and generates pilot symbols.
Each OFDM multiplexer modulates data symbols sd and pilot symbols sp inserted at the input of the multiplexer by sinusoidal sub-carriers whose orthogonal functions are the conjugate Fourier components of an inverse Fourier transform of size NFFT corresponding to the number of carriers of an OFDM multiplex. The NFFT carriers include Nmod modulated data carriers and Np pilot carriers. The set of NFFT carriers is called an OFDM symbol. The OFDM symbol can include null carriers and unmodulated guard carriers. The output of an OFDM multiplexer constitutes the OFDM signal in the time domain with no guard interval.
The transmitter device inserts a guard interval into the OFDM signal before transmission.
The signals transmitted include training sequences. For each transmit antenna concerned, a training sequence includes at least one replica of a basic training sequence, either phase offset or not. This basic training sequence is determined by the positioning of Np pilot symbols contained in one or more OFDM symbols, the number of which OFDM symbols is g. The basic training sequence is therefore determined by the number g, the amplitude of the pilot symbols, and the pilot carriers. Between two training sequences associated with two transmit antennas, the replicas are time offset relative to each other by a minimum of one interval Δt least equal to the maximum length L of the transmission channels, Δt≧L, so that the impulse responses of the various channels do not interfere with each other. The length of a channel is the transmission duration of the channel or the transmission delay introduced by the channel.
Throughout this document, the interval Δt must be understood as being a plurality of time slots separating samples. The time offset is typically applied before modulation, for example OFDM modulation, by phase offsetting the replicas in the frequency domain.
The basic training sequence is expressed in the form of a series of pilot symbols: {tilde over (c)}(p),p=0, . . . , Np−1 where {tilde over (c)}(p) is the pilot symbol transmitted at the carrier frequency of index p. This basic training sequence is spread over g orthogonal symbols of a time-frequency frame. Moreover, the condition for determining the set of impulse responses is:
Kn×Δt≦Np
where:
For example, g is chosen as equal to 1, Nt as equal to 2, NFFT as equal to 8, Np as equal to 8, K1 and K2 as equal to 2, and Δt as equal to 2. These parameters satisfy the above condition because (2+2)×2≦8.
The training sequence transmitted by the transmit antenna TX1 consists of two replicas of the basic training sequence, one of which is phase offset. Its expression is as follows:
The training sequence transmitted by the transmit antenna TX2 also consists of two replicas of the basic training sequence, both of which are phase offset. Its expression is as follows:
In the example illustrated by
Each signal received by a receive antenna is processed by an OFDM demultiplexer. An OFDM demultiplexer performs the inverse function of an OFDM multiplexer. Accordingly an OFDM demultiplexer applies a direct Fourier transform of size NFFT to the received signal after suppression of the guard interval. An OFDM demultiplexer demodulates a time-domain signal received by the receive antenna RXj concerned by means of an FFT of size NFFT to obtain a frequency-domain signal Rj(n) combining the Np pilot symbols of the training sequences of the various transmit antennas.
The outputs of the OFDM demultiplexers provide frequency-domain OFDM signals Rj(n) that are used by the estimator module and the equalizer module.
The channel estimator module includes means for computing the matrix A according to equation (22) or (27). These means are typically program instructions which in a preferred embodiment are incorporated into an electronic circuit, a computation unit such as a processor or the like the operation of which is then controlled by the execution of the program.
The channel estimator module includes means for computing the Ki estimates of each of the Nt impulse responses ĥj according to equation (25). These means are typically program instructions which in a preferred embodiment are incorporated into an electronic circuit, a computation unit such as a processor or the like the operation of which is then controlled by the execution of the program.
Knowing the parameters of the multiple-antenna system (the size NFFT of the FFT, the number Nmod of modulated carriers, the offset Δt between reference sequences, the number g of pilot OFDM symbols forming a reference sequence, the amplitude {tilde over (c)}(p) of the pilot symbols), an estimator method and module of the invention advantageously pre-compute the matrix corresponding to the following expression in which the matrix à is given by equation (26) or (27):
Ã+=(ÃHÃ)−1ÃH (28)
In a preferred embodiment, the estimation method, respectively the estimator module, multiplies this pre-computed matrix product by the demodulated OFDM signal to obtain the Ki estimates of each channel combined in the expression ĥj according to equation (25).
The vector expression of the first estimate ĥj,i of the impulse response of the channel separating the particular transmit antenna i, for i between 1 and Nt, from the particular receive antenna j, for j between 1 and Nr, is obtained by selecting Np successive samples of the column vector ĥj computed for the receive antenna j, starting with the sample (i−1)Δt. For the second estimate ĥj,i of the impulse response of the channel, selection starts from the sample Nt×Δt+(i−1)Δt=(Nt+−1)Δt. For the nth estimate ĥj,i of the impulse response of the channel, selection starts from the sample (n×Nt+i)Δt.
To obtain the estimates for the Nr×Nt channels, the computation of ĥj,i must be repeated for j varying from 1 to Nr.
The channel estimator module includes means for computing an average of the Ki impulse response estimates, thus obtaining Nt impulse responses.
The following two examples illustrate the computation of the expression for Ã.
First example. The transmission system includes Nt=2 transmit antennas and Nr=1 receive antenna. The parameters of the system take the following values:
The condition
is satisfied because (2+2)×2=8≦8.
A single OFDM symbol is assumed necessary for the channel estimation (g=1) and the pattern of this OFDM symbol, which constitutes the basic training sequence, is represented in Table 1. The first column represents the frequency domain indices and the second column the associated data type (P=pilot symbol). Equation (6) gives the expression for the pilot symbol modulating the carrier p where the training sequence transmitted the transmit antenna TXi is concerned, for i=1 or 2:
Thus c(p)1, is the pilot symbol modulating the carrier frequency p for the training sequence transmitted by the transmit antenna TX1 and c(p)2 is the pilot symbol modulating the carrier frequency p for the training sequence transmitted by the transmitter antenna TX2.
For all the pilot carriers of this example c(p)i=1. The vectors of the pilot symbols B1 and B3, respectively B2 and B4, corresponding to the two estimates of the impulse response of the channel connecting the antenna TX1 to the antenna RXj, respectively the antenna TX2 to the antenna RXj, are then expressed in the form:
The Fourier matrix F is as follows:
in which the symbol j represents the solution of the equation j2=−1. The matrix F′ is expressed in the following form:
The matrix à is then expressed in the following form:
i.e.:
Second example. The transmission system includes Nt=3 transmit antennas and Nr=1 receive antenna. The parameters of the system take the following values:
The condition
is satisfied because (2+1+1)×2=8≦8.
A single OFDM symbol is assumed necessary for the channel estimation (g=1), and the pattern of this OFDM symbol, which constitutes the basic training sequence, is represented in Table 1. Equation (6) gives the expression for the pilot symbol modulating the carrier p for the training sequence transmitted by the transmit antenna TXi, for i=1, 2 or 3:
Thus c(p)1 is the pilot symbol modulating the carrier frequency p for the training sequence transmitted by the transmit antenna TX1, c(p)2 is the pilot symbol modulating the carrier frequency p for the training sequence transmitted by the transmitter antenna TX2, and c(p)3 is the pilot symbol modulating the carrier frequency p for the reference sequence transmitted by the transmit antenna TX3.
For example, c(p)=1 for all the pilot carriers. The vectors of the pilot symbols B1 and B4, respectively B2 and B3, corresponding to the two estimates of the impulse response of the channel connecting the antenna TX1 to the antenna RXj, respectively to the estimate of the impulse response of the channel connecting the antenna TX2 to the antenna RXj, and to the estimate of the impulse response of the channel connecting the antenna TX3 to the antenna RXj, are then expressed in the following form:
The Fourier matrix F is as follows:
in which the symbol j represents the solution of the equation j2=−1. The matrix F′ is expressed in the following form:
The matrix à is then expressed in the following form:
i.e.:
Number | Date | Country | Kind |
---|---|---|---|
06 55106 | Nov 2006 | FR | national |
Filing Document | Filing Date | Country | Kind | 371c Date |
---|---|---|---|---|
PCT/FR2007/052363 | 11/19/2007 | WO | 00 | 5/22/2009 |
Publishing Document | Publishing Date | Country | Kind |
---|---|---|---|
WO2008/062131 | 5/29/2008 | WO | A |
Number | Name | Date | Kind |
---|---|---|---|
6665308 | Rakib et al. | Dec 2003 | B1 |
7450489 | Sandhu | Nov 2008 | B2 |
Number | Date | Country | |
---|---|---|---|
20090304104 A1 | Dec 2009 | US |