WO1999062197A1

WO1999062197A1 - Communications receiver and method of detecting data from received signals

Info

Publication number: WO1999062197A1
Application number: PCT/EP1999/002814
Authority: WO
Inventors: Leo Rademacher
Original assignee: Siemens Aktiengesellschaft
Priority date: 1998-05-22
Filing date: 1999-04-26
Publication date: 1999-12-02
Also published as: AU4033799A; GB2337673A; GB9810987D0

Abstract

A radio communications receiver (4), comprising means (6, 8, 10) for detecting radio signals representative of data and for generating, from the detected radio signals digital signal samples, a plurality of the signal samples being representative of a datum of the data. The receiver (4) further comprises at least one data detector (14) having a scaling filter (22) coupled to the means (6, 8, 10) for detecting the radio signals and arranged to scale the plurality of signal samples ri with a plurality of scaling coefficients gi and to combine the scaled plurality of signal samples into a composite signal sample zi, from which the datum is re-generated. The data detector may further include a data detector controller (36) coupled to the scaling filter (22) which operates to adapt the scaling coefficients gi consequent upon an error signal ei derived from the composite signal. The receiver may be arranged to detect CDMA signals, and in particular TD-CDMA signals. The receiver can be developed in a modular fashion by providing a separate data detector for each of the transmitted signals to be detected.

Description

Description of Invention

Communications receiver and method of detecting data from received signals

The present invention relates to radio communications receivers which operate to detect radio signals and to generate data communicated by the radio signals. Furthermore the present invention relates to methods of detecting data from received signals. More particularly, but not exclusively, the present invention, relates to mobile radio communications apparatus.

Digital communications systems operate to communicate data by transmitting signals representative of the data between a transmitter and a receiver. Such digital communications systems often include means for representing each datum to be communicated as a plurality of symbols, which are arranged to modulate a carrier signal. A receiver of the communications system operates to detect the radio signals and to process the detected signals in order re-generate each datum which the modulated symbols represent.

An example of a digital communications system in which data is communicated as a plurality of modulating symbols is a mobile radio communications system which operates to communicate data via radio signals generated in accordance with a code division multiple access (CDMA) radio access technique. Code division multiple access is a radio access technique in which data to be communicated is arranged to be represented by a spreading code or sequence in some way. The spreading code is then arranged to modulate a radio frequency carrier signal. As such, each datum is represented as a plurality of modulating symbols corresponding to the symbols of the spreading code, which modulate the radio frequency carrier signal. The modulated symbols of the code are known as chips. At a receiver, the radio signal is detected and compared with a local version of the spreading code or sequence in some way, in order to recover the modulating data. The spreading code or sequence thereby provides a gain in the power of the received signal with respect to noise for the code associated with a particular transmitter. This gain is known as a spreading gain. Furthermore the spreading code or sequence has an effect of increasing the bandwidth of the transmitted radio signals thereby providing substantial resistance to the effects of time and frequency selective fading. A particular example of a CDMA-radio access technique is time-division CDMA in which each mobile station is provided with a time-slot and a predetermined spreading sequence which is convolved with the data to be communicated. A plurality of mobile stations operating in the system are arranged to transmit in the same time-slots. Other mobile stations are allocated to further time slots, to the effect that a substantial increase in capacity in terms of a number of data communications channels that can be provided contemporaneously is effected, within a portion of the radio frequency spectrum allocated to the mobile radio system.

An inherent characteristic of a communications system, which operates in accordance with TD-CDMA, is that because signals are arranged to be contemporaneously transmitted, a receiver must operate to separate the contemporaneously transmitted wanted data from contemporaneously transmitted unwanted data from other mobile stations using the same time-slot. Known radio communications receivers for time-division CDMA-signals must operate to first estimate an impulse response of the communications channel via which both wanted and interfering signals are communicated and second cancel the unwanted signals from the wanted signal in order to re-generate the communicated data.

Time-division CDMA is therefore representative of an example of a communications system in which data is represented as a plurality of modulating symbols and in which the data must be recovered from the radio frequency carrier signals in the presence of unwanted interfering signals.

It is therefore a technical problem to provide a receiver which can re-generate data from detected radio signals comprising a plurality of symbols representative of a communicated datum in the presence of unwanted interfering signals .

The technical problem is addressed generally by providing a radio receiver with a scaling filter arranged to scale each of the plurality of symbols representative of each communicated datum with a plurality of scaling coefficients, and to adjust the scaling coefficients consequent upon an error signal derived from the radio signals. As such, the solution is provided by treating the symbols as if they had been generated by a corresponding plurality of antennas.

According to the present invention there is provided a radio communications receiver comprising,

- means for detecting radio signals representative of data and for generating, from said detected radio signals a plurality of digital signal samples, a plurality of said signal samples being representative of a datum of said data, and at least one data detector having

- a scaling filter coupled to said means and arranged to scale said plurality of signal samples with a plurality of scaling coefficients and to combine the scaled plurality of signal samples into a composite signal sample, from which composite signal said datum is re-generated.

The term scaling filter as used herein includes any type of filter which operates to scale each of a plurality of signal samples and form a composite signal sample by summing the scaled signal samples. By adapting the scaling-coefficients of the scaling filter, the radio communications receiver operates to detect the wanted signal and re-generate the communicated datum, without a requirement for detecting and regenerating unwanted signal samples which may be communicated with the wanted signal sample contemporaneously.

The data detector may further include a data detector controller coupled to the scaling filter which operates to adapt the scaling coefficients consequent upon an error signal derived from the composite signal, to the effect of substantially improving a probability of correctly regenerating each datum.

The radio communications receiver may further include a data detector for each of a plurality of transmitters which transmit wanted signals from which data is to be recovered.

The radio signals may be code division multiple access signals including a plurality of wanted radio signals. The wanted radio signal may be convolved with a pre-determined spreading sequence. The radio signal may be a wanted radio signal modulated using a pre-determined spreading code.

By treating the signal samples as having been detected as spatial samples, by for example, an antenna array, rather than sequential samples, and applying an adaptive antenna technique to re-generate the data, the receiver is arranged to detected the wanted signals, and reject the unwanted signals without a requirement for estimating the unwanted signals. Furthermore, a receiver which is required to detect a plurality of wanted signals may be constructed in a modular fashion, with each module being a data detector arranged to detect data from the received radio signals independently of other data detectors. A significant advantage is therefore provided by constructing a receiver in accordance with the present invention, in that the receiver is in a modular form. As such, signals generated from other wanted signal sources may be detected by simply adding further data detectors, and arranging for these data detectors to adapt the scaling coefficients of the scaling filter, to recognise the samples of the radio signal as wanted signal samples. This is achieved in the case of TD-CDMA from the spreading sequence assigned to each transmitter. Furthermore, since detection of data for one wanted signal does not require estimation of contemporaneously transmitted signals from all other interfering signals, a receiver constructed in accordance with an embodiment of the invention, has substantially reduced complexity and therefore cost in comparison with known TD-CDMA receivers. Furthermore, unlike known TD-CDMA receivers, the receiver may be constructed in a modular fashion to the effect that further wanted signals may be detected by simply adding further data detectors adapted to re-generate data from these wanted signals.

According to a first aspect of the present invention there is provided a mobile radio communications apparatus, comprising at least one base station, and at least one mobile station, wherein at least one of said base station and said mobile station includes a receiver as claimed in any preceding claim.

According to a second aspect of the present invention there is provided a method of recovering data from radio signals, which data is represented by a plurality of symbols with which said radio signals have been modulated, said method comprising the steps of;

- detecting said radio signals; - generating a plurality of signal samples representative of the plurality of symbols corresponding to a datum of said data; - scaling said plurality of signal samples with a plurality of scaling coefficients;

- combining said scaled plurality of signal samples to form a composite signal sample; and - adapting said plurality of scaling coefficients so that a probability of correctly regenerating each datum is substantially improved.

The method may further include the step of forming an error signal from the composite signal sample and adapting the scaling coefficients so that the error signal is minimised.

Adaptation of the scaling coefficients may be made in accordance with a process known for use in adaptive antenna systems.

One embodiment of the present invention will now be described by way of example only with reference to the accompanying drawings wherein;

FIGURE 1 is a schematic block diagram of part of a mobile radio telephone system;

FIGURE 2 is a schematic block diagram of a receiver forming part of the base station shown in Figure 1; and

FIGURE 3 is a schematic block diagram of a data detector forming part of the receiver shown in Figure 2.

An example embodiment of the present invention will be described with reference to a mobile radio telephone system operating in accordance with a time-division code division multiple access (TD-CDMA) -system. However, as will be clear from the following description of the example embodiment, the invention finds application in receivers of digital communications systems with other multiple access schemes and modulation techniques. An illustration of part of a mobile radio system is shown in Figure 1. In Figure 1 three base stations BS are shown to be interconnected via a mobile network infra-structure, Net. Data is communicated between mobile stations MS and the base stations BS, by transmitting and receiving radio signals 1, operating within a radio coverage area effected by each of the base stations. The radio coverage area is shown illustrated as a broken line 2, and serves to indicate a boundary within which radio communications can be effected with the mobile stations MS. In the present illustrative embodiment the mobile stations MS communicate with the base stations BS in accordance with a TD-CDMA system. A more detailed explanation of how data is communicated using a time division CDMA system is provided in an article, entitled performance of a Cellular Hybrid C-TDMA Mobile Radio System Applying Joint Detection and Coherent Receiver Antenna Diversity" by G. Blanz, A. Klein, M. Naβhan and A. Steil published in the IEEE Journal on Selected Areas in Communications, Volume 12, no. 4, May 1994 at page 568, the content of which is incorporated herein by reference. However for the present explanation it should be noted that communications using TD-CDMA systems is characterised in that radio frequency carrier signals via which communication of data is effected are divided into a plurality of time-slots. Each of these time-slots is assigned to a plurality of mobile stations which operate to communicate radio signals in the time-slot. In order to separate data communicated by the plurality of mobiles assigned to the same time-slot, each mobile is provided with a user specific code which is convolved with the data to be communicated. A receiver of the radio signals from the mobile stations MS, must therefore operate to separate data communicated by other mobile stations MS contemporaneously.

The separation and re-generation of data communicated by mobile stations contemporaneously in the same time-slot is achieved with known receivers of TD-CDMA signals by estimating the channel impulse response via which signals transmitted by other unwanted mobiles within the same time- slot, and thereafter eliminating the unwanted signals from the wanted signal. In order to provide an explanation of the operation of a receiver in accordance with an embodiment of the present invention, a brief description of the operation of a known receiver for TD-CDMA signals will be presented in the following paragraphs.

Data communicated from, for example, a mobile station denoted d_u(i) is arranged to generate a spread spectrum radio signal by convolving the data d_u(t) with a spreading code. Individual symbols of the code, known as chips, include pulse shaping to provide appropriate limitation and spectral shaping of the frequency bandwidth occupied by the radio signals. Using the symbol ® for convolution the radio signal communicated in accordance with a TD-CDMA system, after sampling the continuous valued received signal at the chip rate can be expressed as given in equation 1, where d_u(t) is at the symbol rate.

r(t) = ∑c_u(t) ®h_u(t, τ) ®d_u(t) + n(t) (1)

In equation (1) , r(t) is the continuous valued received signal with contributions from u transmitters sampled at the chip rate, ^ t^) the channel impulse response sampled at the chip rate for transmitter u , d_u(f) the data symbols from transmitter u , c_u(t) the spreading code for transmitter u (including the chip modulation) at the chip rate and n(i) is continuous valued additive white Gaussian noise sampled at the chip rate.

Introducing the combined channel impulse response b_u(t, τ) (including code and modulation) yields equation (2) : r(t) = ∑b_u(t, τ) ®d_u(t) + n(t)

In case of time variance the combined channel b_u(t, τ) also becomes time variant. Introducing a matrix notation, r is a vector of received signal samples, ή is a vector of noise samples; the element index of these vectors describes the sample time in chips, i.e. the number of the symbol and the number of chips within a symbol. For clarification the time axis for sampling at the chip rate will be described by their symbol number si, s2 (2 symbols transmitted by each user) and their chip number cl, c2, c3 (3 chips per symbol) , as for example illustrated by equation (3) :

A corresponding data symbol vector d consists of several sub-vectors, each sub-vector contains the data symbols s\,sl for each transmitter u\ , ul which for two trasnmitters is expressed in equation (4):

d(u\,s\) ^' d(u\,s2)

(4) d(u2,s2) d(u2,s2)

It is well known that a convolution can be described by matrix multiplication if one vector is extended to a matrix. When constructing this matrix the column vector becomes the first column of the matrix, with the second column containing the elements of the first column but shifted by one element, the third column the elements of the second column shifted by one element and so on. In the case of CDMA, spreading with the code also introduces temporal over-sampling; therefore the shift from one column to the next is not one but equals the number of chips per symbol, which is the number of symbols representing a datum of the data to be communicated. For each transmitter the combined channel impulse response is now extended to such a matrix A(u\), A(u2) so that the contribution of each transmitter rl and r2 can be described as given by equation (5) :

f\ = A(u_)d(uϊ) and r2 = A(u2)d(u2) (5)

Additionally, the summation over transmitters is accommodated by the matrix multiplication. For this purpose the total combined channel matrix A is formed by the sub-matrices A(u\) and A(u2) , so that the vector of received signal samples may be expressed as given in equation (6) :

The matrix notation has been described for the case that the symbol sub-vectors contain all symbols for one transmitter and the received signal is arranged according to the sampling in the time domain. Obviously, any arrangement of vectors r and d is possible and the structure of the matrix A varies accordingly because the matrix row index must correspond with the index of the vector r and the matrix column index corresponds with the index of the vector d .

The matrix equation (6), can now be solved in the least squares sense. Minimising the residual noise leads to the generalised matrix inverse of the matrix A , known as the Moore-Penrose pseudo-inverse. This is expressed as equation (7), where prime (^λ) denotes conjugate transposition (the

Hermitian) . d denotes the reconstructed continuous-value symbol vector, before the detected data symbols are determined with respect to a threshold of hard decision values which is known as slicing.

d = A⁺r = (A' A)^~l A'r ( 1 )

Equation (7) holds for un-correlated white noise and minimum noise (zero-forcing) criterion. For severe channel distortion this criterion suffers from poor performance because the difference between the actually received signal r - Ad + n and a reference receive signal f = Ad is minimised (zero forcing criterion) . This effect results from noise enhancement and is well-known within the equalisation art. The usual approach there is to use the minimum mean square error (MMSE) criterion that constitutes a compromise between perfect equalisation and noise enhancement by minimising the difference between the estimated data symbols d and the transmitted data symbols d . The matrix formulation according to equation (7) is implemented in a receiver for TD-CDMA signals. In effect the estimated data vector in the presence

2 of noise is described by equation (8), where σ denotes the noise power and I is the identity matrix:

d (A' A + σ²iy ^l A'r

The additional noise term provokes the compromise between perfect transmitter separation and equalisation on one side and noise enhancement on the other. For this purpose the noise must be estimated. A standard worst-case value can also be used but it leads to a degradation in the case of little noise. TD-CDMA does not use the constraints that only certain symbol values are allowed in the optimisation process but performs the respective minimisation under the hypothetical assumption of a Gaussian distribution of symbol values. Finally, a threshold device or slicer confines the detected symbols to predetermined allowed values. This process of neglecting the constraints during optimisation and then slicing the optimisation results according to the constraints constitutes a sub-optimum optimisation procedure.

Known receivers for TD-CDMA signals perform an optimisation of the continuous-value data vector d for all transmitters. As such, unwanted interfering signals must also be considered, and their channel estimated, in order to suppress them. All versions can be described with the separation and equalisation matrix M that transforms the received signal vector r into the data signal vector d containing the data symbols for all transmitters. This is represented as equation (9) :

d = Mr (9)

Now the data symbol vector d is assumed to be subdivided into sub-vectors each containing the data symbols for one specific transmitter. For example the case for two transmitters is shown in equation (10) :

d(u\) M(u\) d(u2) M(u2) ;ιo;

Regarding each transmitter separately delivers the individual separation/equalisation equations, dependent on the transmitter u , as determined by equation (11) :

d_u = M_ur (11)

The process of recovering data from signals transmitted by a specific transmitter, can therefore be effected by individual decimating filters, one for each transmitted signal to be detected. However, for the calculation of all filter coefficients, i.e the matrices M_u as sub-matrices of M , all the interferers must be included and no straightforward estimation of the M_u is performed. This results in the non- modular architecture in that for a single transmitter, a calculation of all interferers is required to be made. Therefore known receivers of TD-CDMA signals require a data processor to detect data which must execute calculations representing some considerable complexity. However, a receiver which operates to detect data communicated using TD- CDMA signals and provides a modular architecture, is shown in Figure 2 where parts also appearing in Figure 1 bear identical numerical designations.

In Figure 2 four mobile stations MS are shown to be communicating radio signals contemporaneously to a receiver 4, forming part of the base station BS shown in Figure 1. Each of the four mobile stations MS is arranged to communicate data contemporaneously in the same time-slot of a TD-CDMA time-frame. As already explained each mobile station operates to convolve the data to be communicated with a pre- determined spreading sequence. The radio signals are thereafter communicated via the ether to an antenna 6, of the receiver 4. Effectively, therefore, the data communicated by each of the mobile stations MS is convolved with a transmitter specific spreading code

and an impulse response ₀,h\,h2,h corresponding to the radio communications channel formed by the transmission of radio signals from the mobile stations MS, to the antenna 6 of the base station BS . This is represented mathematically in equations (1) and (2) above .

The detected radio signals are fed from the antenna 6, to a down-converter 8, which operates to convert the radio frequency signals detected by the antenna 6 to a base band representation. The base band signals are thereafter fed to a digital to analogue converter 10 which operates to sample the radio signals substantially at the chip rate (1/Tc) corresponding to the TD-CDMA spread spectrum signals. The digital signal samples are thereafter fed to a serial to parallel converter 12 which operates to feed each of the chip samples corresponding to a modulated symbol to one of a plurality of data detectors 14, via parallel conductors 16. Thus, the sampled input signal is transformed from its usual serial CDMA style form temporally over sampled at the chip rate to a spatially over sampled adaptive antenna style parallel form. Each of the data detectors 14, is provided to detect data communicated from a corresponding mobile station.

An example embodiment of a data detector 14 is shown in Figure 3 where parts also appearing in Figure 2 bear identical numerical designations. In Figure 3 the data detector 14 is shown to be fed with the digital signal samples from the serial to parallel converter 12 by parallel conductors 16. Each of the parallel conductors 16 feeds a corresponding digital signal sample to a first input of one of a corresponding plurality of multipliers 18, which in combination with a first adder 20 form a scaling filter 22. Each of the multipliers 18 is fed on a second input thereof with one of a plurality of scaling coefficients g_j , which form a ,,spatial-style" impulse response.

An output of each of the multipliers 18, feeds the corresponding scaled digital signal sample to a first adder 20. The first adder 20, sums each of the scaled digital signal samples to produce a composite signal sample ZJ which is fed to a first input of a second adder 23. Fed on a second input is a reference signal sample kj . The reference signal sample kj is generated by a reference signal generator 24. The reference signal generator 24, is comprised of a shift register 26, having a plurality of stages 28, which are fed with reference data fed from a reference data store 30. Each of the stages 28, of the shift register 26, has an output which feeds the content of the stage to an input of a corresponding multiplier 32. The multipliers 32, serve to scale the stored data samples from the stages 28, of the shift register 26, by a corresponding coefficient A/ of an impulse response estimate of the communications channel via which the radio signals were communicated as seen after the decimating filter 22. An output of each of the multipliers 32 feeds a third adder 34. The second adder 23, serves to subtract the reference signal sample k , provided at the output of the third adder 34, from the composite signal sample z/ , to generate an error signal e/ which is fed to a data detector controller 36. The detector controller 36, operates to generate a new version of the channel impulse response estimate hj and a new version of the scaling coefficients gj at outputs 38, 40 respectively. The index i corresponds to the i-th datum of the transmitted data being detected. Also connected to the output of the second adder 23 is a data estimator 42. The data estimator 42, operates to regenerate the data communicated by the radio signals, from the error signal ed by choosing such symbols that the resulting overall error formed by the error signal samples is minimised. The data estimator 42, may be, for example a maximum likelihood or near maximum likelihood sequence estimator.

An initialisor 44, operates to generate on two conductors 46, 48, vectors of data representative of an initial estimate of the impulse response coefficients of the scaling filter Si init i ^and an initial estimate of the channel impulse response hj j_njj . The initial estimates of both are derived, for example, from a known training sequence transmitted with the TD-CDMA radio signals. A part of the radio signals which were modulated in accordance with the training sequence are therefore separated from the received signals and fed to the initialisor 44, although this is not actually shown in Figure 3. The scaling coefficients gj of the scaling filter are selected to provide an appropriate initialisation for the scaling multipliers 18, which form the spatial filter.

Conductors 46, 48, are connected to corresponding inputs of the detector controller 36, and serve to convey to the detector controller 36, the initial estimates of the channel impulse response hj j_nj_f and the scaling filter coefficients g_j ,init . The reference data store 36, is arranged to hold predetermined reference data sequences corresponding to all possible combinations of data symbols within a memory length corresponding to the length of inter-symbol interference of the channel impulse response which would have effected the received data signal sample Zj . The data sequences are fed to the reference signal generator 24, and are convolved with the channel impulse response estimate hj to generate the reference signal sample kj as already explained.

Operation of the example embodiment of the invention will now be described. The receiver 4, according to the example embodiment of the present invention, operates to detect radio signals and regenerate wanted data associated with at least one transmitter, without a requirement to re-generate all interfering signals or at least estimate their channel impulse responses, which is required in known receivers of TD-CDMA signals. This is achieved by considering the received signal as having been generated by an antenna array, with each chip sample corresponding to a sample of the signal detected by an antenna. As such the parallel scaling filer 22, represents a quasi-spatial filter. By adapting the scaling coefficients gj , the data detector 14, operates to effectively cancel contemporaneously detected TD-CDMA signals, which as with an adaptive antenna system, are considered as interfering signals for which the antenna array forms a null in a simulated direction in which the interring signals are considered to emanate. Correspondingly, the data detector 14, operates to adapt the scaling coefficients to „steer a beam" formed by the scaling coefficients gj to the effect that the scaled samples of the wanted signal are summed constructively by the scaling filter, as would be the case for a spatial filter of an adaptive antenna system forming a beam in a direction from which received signals add coherently. A significant advantage is therefore provided by constructing a receiver 4, in the form of the example embodiment of the invention, in that the receiver 4, is in a modular form. As such, signals generated from further mobiles may be detected by simply adding further data detectors 14, and arranging for these data detectors to adapt the scaling coefficients of the filter 22, to recognise the samples of the radio signal as samples of the wanted signals. This is achieved in the case of TD-CDMA signals from the spreading sequence assigned to each mobile station. Furthermore, since detection of data for one mobile station does not require estimation of the channel impulse responses of contemporaneously transmitted signals from all other mobile stations, a receiver constructed in accordance with the embodiment of the invention, has substantially reduced complexity and therefore cost in comparison with known TD-CDMA receivers. Also the lack of coupling between detection of several user signals is more advantageous for a base station architecture.

To complete the description of the example embodiment of the present invention, the operation of the data detector 14, will now be described. Generally the operation of the data detector 14, corresponds to that of a receiver incorporating an adaptive antenna system having a spatial filter, as described in the Applicants German patent application serial number 19604772.2 for time varying signals, and in German patent application serial No. 19639414.7 for time invariant signals. Further advantageous features of the receivers are described in United Kingdom patent application serial No. 9804785.5. The content of these co-pending patent applications is incorporated herein by reference.

The serial to parallel converter 12, serves to convey each chip sample of the chips representing a communicated datum to the data detectors 14. In the data detector 14, the scaling filter 22, serves to scale each chip sample with scaling coefficients and to form the composite signal Z . The scaling filter 22, then generates a combined scaled signal sample representative of the received datum ∑j . The initialisor 44, provides an initial set of antenna weight coefficients gf init and an initial channel estimate hj j_njf , (based on the known training sequence) . The composite signal sample is fed from the scaling filter 22, to the second adder 23, which forms an error signal ey as described below. As will be appreciated, various techniques for forming the composite signal sample may used, for example, feeding the received signal samples to a shift register, scaling the signal samples with the scaling coefficients by using an associated plurality of multipliers coupled to the shift register and forming a sample representative of the datum, by decimating the scaled plurality of radio signal samples.

Referring back to Figure 3, the detector controller 36, adapts the scaling coefficients and channel model impulse response coefficients for every data symbol Zj , and feeds these to the scaling filter 22, and to the reference signal generator 34, via conductors 38, 40. These estimates are updated from the initial values, by the detector controller 36. The data estimator 42, forms a maximum likelihood sequence estimator (MLSE) or a near maximum likelihood sequence estimator which operates to compensate for the effects of inter-symbol interference and fading experienced by the received radio signals and generate signals representative of the data conveyed by the radio signals. This is conveyed on conductor 50, and is shown as the symbols dj .

As already explained, each combined scaled sample representing an unknown datum is fed to the first input of the second adder 23. A second input to the adder 23, is fed with a reference signal generated by the reference signal generator 24. The data store 30, is coupled to the reference signal generator 24, and operates to feed locally generated reference data symbols corresponding to each possible set of symbols which may have modulated the radio signals, and which are effected by the memory of the channel to the input of the finite impulse response filter formed by the shift register 26 and multipliers 32. The reference signal generator 24, therefore operates to convolve the local reference data sequence corresponding to the memory of the channel with the channel impulse response estimate hj . The convolution is completed by the third adder 34 and fed to the second input of the second adder 23. By subtracting the reference signal k_j from the received composite signals ZJ the error signal β , is formed which is fed to the detector controller 36, and to the data estimator 42. The detector controller 36, determines subsequent estimates of scaling filter coefficients gj and channel impulse response estimate coefficients hj from the latest error signal e . The data estimator 42 adds the squared error signal to a path metric determined for the state of the channel, in accordance with the maximum likelihood sequence estimation algorithm, or any other related algorithm which is able to re-generate an estimate of the data from the error signal e; , for example, a symbol-by-symbol detector.

The model variable corresponding to the signal formed by the reference signal generator kj is representative of one possible sequence of data convolved with the channel impulse response estimate. Thus the residual multi-path effects caused by the propagation of the radio signals that remain after the decimating filter 22, are simulated with successively arriving signal components being superimposed to form a common signal. The data detector controller 42, operates to generate the antenna coefficients gj and the channel model impulse response estimate coefficients hj in accordance with a least squares error for each received signal sample ZJ . The data detector controller 36, operates to adapt the antenna coefficients gj such that the least square error e\ is maintained to be a minimum. A further condition for the solution of the square error is that of a constraint in that not all antenna weighting factors gj and coefficients of the channel impulse response hj are equal to 0. Therefore, and in order to achieve an unambiguous solution, an additional constraint is introduced, in that the sum of the squared channel coefficients is forced to 1 by appropriate scaling.

As will be appreciated, instead of a solution to a problem of least square error, other algorithms may be used to the effect of minimising deviation of the signal e₇- . Instead of joint optimisation of antenna and channel coefficients, a separate optimisation may be employed and a different normalisation constraint may be selected.

In order to demonstrate that the receiver 4, herein before described can be used to detect and recover data communicated in accordance with TD-CDMA signals, and detected in the presence of like-modulated unwanted TD-CDMA signals, an analysis is presented in the following paragraphs showing that the recovery of data from TD-CDMA radio signals, and the detection of radio signals by a receiver incorporating an adaptive antenna system is substantially the same.

Using the symbol ® for convolution radio signals received at each individual antenna of an antenna system having a plurality of antenna branches p can be described by equation

(12) :

r_p(t) = ∑h_p „(t, τ) ® x_u(t) + n(t) (12) u In equation (12), x_u(t) denotes data transmitted by a transmitter u . The channel impulse responses now describe the channels between the transmitter and the individual antenna branches. In case of time variance the channels hp _U(t) also become time variant. Adopting a matrix notation r is a vector of received signal samples, is a vector of noise samples, wherein the element index of these vectors describes the sample time in symbols and the antenna branch. Thus each element of the vector is determined by the number of the symbol and the number of the antenna element. For clarification of the vector elements, the time axis for sampling with the symbol rate and the branch number will be described by the symbol number si, s2 (2 symbols transmitted by each transmitter) with the antenna branch number being bl, b2, b3 (3-branch adaptive antenna) , as presented in equation (13) :

In this manner the vector elements are arranged in order of the sampling time. If equation (13) is considered in terms of the contribution from each transmitter ul, u2, to the transmitted signal vector detected at the receiver, the transmitted signal vector may be considered to consists of several sub-vectors, each sub-vector containing the transmitted signal samples si, s2 for each transmitter ul, u2, as expressed by equation (14) for two transmitters:

x(u\,s\) x(u\,s2)

: i 4 ) x(u2,sϊ) x(u2,s2)

In the case of an adaptive antenna the existence of several branches introduces spatial over-sampling. Therefore, in matrix form the shift from one column to the next column of the matrix is represented as not one sample, but a number of samples corresponding to the number of antenna elements. For each transmitter the combined channel impulse response can be expressed as a matrix H(u\) , H(u2) , so that the contribution of each transmitter rl and r2 is described by equation (15):

r\ = H(u\)x(ul) and r2 = H(u2)x(u2) (15)

For the signal samples received at the receiver, a total combined channel matrix H is formed by the sub-matrices H(u\) and H(u2) , as given by equation (16):

The matrix notation used in equation (16), has been described for the case that the symbol sub-vectors contain all symbols for one transmitter and the received signal is arranged in sub-vectors of equal sampling time, in which each sub-vector the sub-index describes the branch number.

Obviously, any arrangement of vectors r and x is possible and the structure of the matrix H varies accordingly because the matrix row index must correspond with the index of the vector f and the matrix column index corresponds with the index of the vector x . The time variance destroys the symmetry of the channel matrix H that holds in the time- invariant case. Hence in matrix notation of the received signal r = Hx + , the vector x is the same as d in the TD- CDMA case (equation (6)) . The vector r is composed of individual sub-vectors (one for every antenna element) , with each sub-vector consisting of symbol-rate samples. Also for the vector r in the CDMA case, the same structure of the vector can be achieved by rearrangement as shown above in equations (8) and (9) . The time sample index of f consists of the symbol index and the chip-index. Therefore, a result of introducing a sub-vector for each chip index leads to an identical structure, as that developed for TD-CDMA signals, as expressed in equation (9) .

The matrix description for signals expressed in accordance with TD-CDMA and those expressed for signals detected by an adaptive antenna system is identical. In fact, the different signals can be converted from one to the other by a serial to parallel conversion. This serial to parallel conversion is performed by the serial to parallel converter 12, shown in Figure 2.

Adaptive antenna algorithms estimate a plurality of scaling coefficients, with which the signal detected by each branch of the antenna system is scaled using a scaling filter. Furthermore, since the received signals are comprised of signals transmitted from a plurality of mobile stations, a data detector for each mobile station is provided with a scaling filter and means to adapt the scaling coefficients.

All scaling coefficients for all branches are described by a vector g_u , which consists of sub-vectors for each transmitter. The convolution with this coefficient sub-vector g_u and the received signal vector f results in a separated signal vector z_u which no longer contains interference from other transmitters but may still contain inter-symbol interference. The inter-symbol interference is removed by a suitable means such as, for example, an equaliser as effected by the data estimator 42, in the data detector 14, shown in Fig. 3. Introducing the convolution matrix G_u for the vector g_u and R for the vector f this convolution can be written in matrix form as expressed by equation (17) :

W = SU^R or u = ^Gfu ^'17' The simplest adaptive antenna incorporates a data processor which operates to adapt the antenna coefficients is the well- known LMS array, which operates to adapt the coefficients to the effect of minimising a difference between a combined signal produced by the adaptive antenna, and a reference signal according to the LMS algorithm. If the transmitted symbols are used as the reference signal the adaptive antenna performs separation and equalisation contemporaneously as with TD-CDMA reception. The adaptation criterion demands that z_u match x_u as closely as possible. As such, equation (18) delivers a steady-state solution of the IMS algorithm.

g_u = R⁺x_u (18)

Now the coefficients for the individual adaptive antennas, which can be regarded as spatial decimating filters, are known and the individual transmitter signals can be estimated, in accordance with equation (19) :

^u = G_ur_u (19)

The adaptive antenna performs an optimisation of the continuous-valued data vector x_u for only one transmitter.

A separate data detector is used for each transmitter to be detected. Each set of scaling coefficients is adapted independently of the others and no knowledge is required about the other transmitters for which data is detected. Thus, no other interferers need to be considered or channel impulse responses estimated for the interferers, in order to suppress them. This facilitates construction of a receiver with a substantially modular architecture.

In the matrix representation as previously developed, a set of adaptive antenna coefficients for the receiver illustrated in Figure 3, embodying the present invention, can be described with an overall adaptive antenna matrix G , which transforms the received signal vector f to the data signal vector x for data communicated by all mobile stations which is expressed in accordance with equation (20) :

(«l) R(u_) x(u2) R(u2) g or Rg :2o:

Equation (20) provides a general solution to a problem of recovering data which is temporally over-sampled. In the case where signals are represented as TD-CDMA signals, the temporal over-sampling is formed from convolution with the spreading sequence.

In accordance with the above analysis it will be apparent to those skilled in the art that detecting convolutionally spread, or temporally over-sampled CDMA signals using a temporal decimating filter, is equivalent to the problem of recovering data from a plurality of versions of a radio signals provided by an antenna array and combined by a spatial decimating filter.

In the case of conventional TD-CDMA reception, the filter coefficient estimation cannot be performed in a modular fashion. This is because even if only one wanted signal from a transmitter is to be received the channels of all other transmitters and additional interferers to be cancelled must be estimated and included in the calculation of the pseudo- inverse or the modified versions.

A substantial advantage of the adaptive antenna approach is that the scaling coefficients required in the receiver are estimated directly without explicit knowledge of unwanted interferers or their channel impulse responses, which still have a great implicit influence on the coefficients which converge in such a way that interferers are nulled out. It is well-known from adaptive antennas that an adaptive antenna with N elements can at most null out N-l interferers. An adaptive antenna system usually requires a reference for the wanted signal. This is provided by estimating the channel impulse response of the transmitter concerned as described for the receiver embodying the present invention. No channel estimation of other transmitters and interferers is required, however, so that the entire reception can be implemented in a modular fashion. For the base station BS, shown in Figure 1, a modular architecture is facilitated, which is well suited for integration into ASICs or partitioning onto individual signal processors, which may be implemented as ASICs (one per transmitter) as shown in Fig 3.

The example embodiment of the present invention, can be seen to provide a particular advantage for use in detecting TD- CDMA signals. This is because the receiver is not required to estimate the channel impulse responses of all interfering signals, in order to detect and re-generate the wanted signal. As such a receiver incorporated into a base station of for example a mobile radio system can be implemented in a modular way, in that wanted signals from different mobile stations can be detected and re-generated by simply adding a number of data detectors 14 to the receiver in accordance with the number of mobile stations.

As will be appreciated by those skilled in that art various modifications may be made to the example embodiment without departing from the scope of the present invention. In particular, other forms of data detector may be used to separate the wanted signal from unwanted interfering signals. Furthermore the invention finds application in detecting other forms of signals and multiple access schemes. For example the present invention can also be used to detect Orthogonal Frequency Division Multiplexed signals wherein, a symbol is transmitted over sampled on a plurality of contemporaneously modulated carrier signals. Furthermore, the present invention is not limited to use with TD-CDMA signals, but finds application with direct sequence CDMA signals. However, for the receiver to operate with direct sequence CDMA signals, the spreading code which is modulated by the data symbols to be communicated, should be arranged to be time invariant with respect to the modulating data symbols .

Claims

CLAIMS :

1. A radio communications receiver comprising

- means (6, 8, 10) for detecting radio signals (1) representative of data and for generating, from said detected radio signals a plurality of digital signal samples, a plurality of said signal samples being representative of a datum of said data, and at least one data detector (14) having - a scaling filter (22) coupled to said means (6, 8, 10) and arranged to scale said plurality of signal samples with a plurality of scaling coefficients [ gj ) and to combine the scaled plurality of signal samples into a composite signal sample [ Γêæj ) , from which composite signal said datum is re- generated.

2. A radio communications receiver as claimed in Claim 1, wherein said data detector (14) further includes

- a data detector controller (36) coupled to said scaling filter (22) which operates to adapt said scaling coefficients [ gj ) consequent upon an error signal ( ╬▓j ) derived from said composite signal sample ( Zj ) , to the effect of substantially improving a probability of correctly re-generating said datum.

3. A radio communications receiver as claimed in Claims 1 or 2, and further including a data detector (14) for each of a plurality of transmitters which transmit wanted signals from which data is to be recovered.

4. A radio communications receiver as claimed in any preceding Claim, wherein said means (6, 8, 10) for detecting said radio signals and for generating said plurality of radio signal samples comprises - an antenna (6) coupled to a down converter means (8) which operates to convert the radio signals detected by the antenna (6) into base band form, - and an analogue to digital converter (10) coupled to said down converter means (8) and arranged to generate said plurality of signal samples.

5. A radio communications receiver as claimed in any preceding claim, wherein said data detector (14) further includes

- a serial to parallel converter means (12) arranged to convey to said data detector said plurality of signal samples which are representative of said datum.

6. A radio communications receiver as claimed in any of claims 1 to 4, wherein said scaling filter comprises,

- a shift register to which said radio signal samples are fed,

- an associated plurality of multipliers coupled to said shifter register and arranged to scale said radio signal samples with said scaling coefficients, and

- a decimating filter coupled to said multipliers and arranged to form a sample representative of said datum, from said scaled plurality of radio signal samples.

7. A radio communications receiver as claimed in any of Claims 2 to 6, wherein said data detector further includes - means for generating a reference signal ( kj ) , and means for forming said error signal by subtracting said reference signal from said composite signal sample.

8. A radio communications receiver as claimed in Claim 7, wherein said means for generating said reference signal comprises

- a reference signal former (24) being fed from a data store (30) with a plurality of symbols which could have effected the datum being detected by said data detector and a channel impulse response estimate, said reference signal former operating to generate said reference signal sample by convolving said plurality of symbols with said channel impulse response estimate.

9. A radio communications receiver as claimed in Claim 8, wherein said channel impulse response estimate is adapted by said data detector controller in accordance with said error signal .

10. A radio communications receiver as claimed in Claim 9, wherein said data detector further includes

- a data estimator (42) coupled to said means for forming said error signal, which operates to re-generate said data from said error signal.

11. A radio communications receiver as claimed in any preceding Claim, wherein said radio signal is modulated in accordance with a code division multiple access.

12. A radio communications receiver as claimed in any preceding Claim, wherein said code division multiple access is time division-code division multiple access, said radio signals being generated by convolving said data to be communicated with a pre-determined spreading sequence.

13. A mobile radio communications apparatus, comprising at least one base station, and at least one mobile station, wherein at least one of said base station and said mobile station includes a receiver as claimed in any preceding claim.

14. A method of recovering data from radio signals, which data is represented by a plurality of symbols with which said radio signals have been modulated, said method comprising the steps of; - detecting said radio signals; - generating a plurality of signal samples representative of the plurality of symbols corresponding to a datum of said data;

- scaling said plurality of signal samples with a plurality of scaling coefficients;

- combining said scaled plurality of signal samples to form a composite signal sample; and

- adapting said plurality of scaling coefficients so that a probability of correctly regenerating each datum is substantially improved.

15. A method of recovering data from radio signals as claimed in Claim 14, and further including the step of,

- forming an error signal from said composite signal sample and adapting said scaling coefficients so that said error signal is minimised.

16. A method of recovering data from radio signals as claimed in Claim 15, and further including the step of, - forming reference signals from a plurality of reference symbols representative of a possible combination of symbols which have an influence on said composite signal sample;

- generating said error signal by subtracting said reference signals from said composite signal.

17. A method of recovering data from radio signals as claimed in Claims 16, wherein the step of generating said reference signals further includes the steps of;

- generating an estimate of the impulse response of the communications channel through which the received radio signals have passed;

- convolving said channel impulse response with said test symbols to form said reference signals.

18. A method of recovering data from radio signals as claimed in any of Claims 14, to 17, wherein the step of generating said plurality of samples of said plurality of symbols representative of a datum, includes the steps of;

- sampling said detected radio signals;

- feeding said signal samples into a shift register; and - reading said samples out from said shift register in parallel to be scaled by said scaling coefficients, when said signal samples correspond to a datum which is to be recovered.

19. A method of recovering data from radio signals as claimed in any of Claims 14 to 18, wherein said radio signals are code division multiple access radio signals.

20. A radio communications receiver as herein before described with reference to Figures 2 and 3 of the accompanying drawings .