WO1997048192A1 - Process for separating a received mixture of signals - Google Patents

Abstract

In a multi-subscriber radio system (1) in which the signals are given a subscriber-specific form, after a transmission via one and the same channel (4), the mixture of signals received at a subscriber's station (5) is separated in two steps a manner tailored specifically for the subscriber concerned. In the first step channels are equalised (7). In the subsequent second step the signals are filtered (8), whereby the filter coefficients are determined from the signal form specific to the subscriber concerned. The process of the invention can be used for the down-link of a mobile telephone system operating on CDMA.

Description

 description

METHOD FOR SEPARATING A RECEIVED SIGNAL MIXTURE The invention relates to a method for separating a received signal mix according to the preamble of claim 1.

The reception-side separation of a signal mixture in the context of a multi-subscriber signal transmission system, the transmission of the different subscriber signals taking place via one and the same transmission channel, is required, for example, on the reception side of the downlink of mobile radio systems in which a fixed base station communicates with a large number of mobile stations can.

It can be assumed that the base station wants to communicate with K subscriber stations in a mobile radio system at the same time. It must therefore be ensured that in the downlink the signal coming from the base station and intended for one of the K subscriber stations can be separated from the signals intended for other subscriber stations in the relevant subscriber station. Such separation is made possible by using multiple access methods. A distinction is made between the more classic multiple access methods frequency division multiple access (FDMA) and time division multiple access (TDMA) as well as the modern multiple access method code division multiple access (CDMA).

These three methods can also be used in combination. The multiple access procedure CDMA is based on the principles of the spreading technique. In CDMA systems, the K subscriber signals, which each consist of a data sequence with data symbols strung together, differ in the CDMA spreading codes contained in the data symbols, which are also called signatures. The following considerations gen primarily concern systems using CDMA. However, the applicability of the method according to the invention is not limited to CDMA systems. The separation of a signal mixture, which corresponds to a data estimate in CDMA systems, can be accomplished in a known manner by means of signal-adapted filtering (MF, Matched Filtering), which is adapted to the CDMA spreading code of the subscriber. A receiver performing this signal-adapted filtering can be implemented, for example, as a correlator or as a RAKE receiver, cf. eg the book by Proakis, JG: "Digital Communications", New York, McGrawHill, 1989. Another known possibility for separating an incident signal mixture is the so-called interference cancellation (IC, interference cancellation), which is described, for example, in German patent application 195 09 867.6, in Ewerbring, M .; Gudmundson, B .; Larrson, G .; Teder, P .: "CDMA with interference cancellation: A technique for high capacity wireless Systems", Proc. Int. Conf. Commun., Geneva, 1993, pages 1901 to 1906 and in Dent, P .; Gudmundson, B .; Ewerbring, M .:

"CDMA-IC: A novel code division multiple access scheme based on interference cancellation", Proc. Symp. Personal, Indoor and Mobile Radio Commun., Boston, 1992, pages 98-102.

In addition, a signal mixture can be separated in a known manner with the aid of a so-called joint detection (JD, joint detection), which is described, for example, in DE 41 21 356 C2 and DE 43 29 320 A1.

The signal-adapted filtering results in a very poor performance when separating a signal mixture, e.g. B. in the form of bit error probabilities, with multi-way channels. When using interference elimination (IC) and joint detection (JD) there is the disadvantage that when separating a signal mixture on the receiving side, the spreading codes of all active subscribers that are to be processed with IC or JD must be known in each receiver . In the case of downlinks in mobile radio systems, the spreading codes, for example the CDMA codes, which are used for the transmission from the base station to all mobile stations, must therefore be known in each of the mobile stations and thus transmitted to each mobile station, which is a high additional Effort means. In addition, the computing effort required when using interference elimination (IC) and joint detection (JD) is relatively high.

The object of the invention is to achieve an improvement in performance compared to signal-matched filtering in the separation of a signal mixture in the downlink (downlink) at the receiving end, and a performance similar to that in performing interference elimination (IC) or joint detection (JD). to reach. In contrast to interference elimination and joint detection, the need to transmit the spreading codes (CDMA codes) of all participants to all receivers is to be avoided. In addition, at least in most cases, the computational effort is to be reduced in comparison with interference elimination and for joint detection. According to the invention, this object is achieved in a generic method by the features specified in the characterizing part of claim 1. In principle, any equalizer known from the literature (for example the book by Proakis, JG: "Digital Communications", New York, McGraw-Hill, 1989) mentioned in the literature can be used for the channel equalization carried out in the first step. Appropriate further developments and possible embodiments of the method according to the invention are specified in the subclaims. Some of the subclaims contain particularly advantageous options for channel equalization.

The method according to the invention for separating a signal mixture is explained below using an exemplary system model for the downlink of a digital CDMA mobile radio system. Show it:

Fig. 1 shows the block diagram of this system model for the

 Downlink of a mobile radio system,

2 shows the structure of the covariance matrix RQ,

Fig. 3 shows the structure of the matrix and

Fig. 4 steps to modify the matrix and

 Matrix inversion.

In this context, a data estimation algorithm for single detection, namely a so-called SD (single user detection) algorithm, is being developed. In the following description of the discrete-time system model, vectors and matrices are bolded in small and large letters, complex sizes are underlined. The complex conjugation or transposition is described by (.) ^* Or (.) ^T , the formation of the expected value is described by E (.). As shown in FIG. 1, a base station 1 transmits K data symbol sequences d ^(k) , k = 1. , , K, to K subscriber stations that are operated simultaneously in the same frequency band in the same radio cell. The data symbol sequences d ^(k) , k = 1. , , K, represent a data block either before or after the midamble. each

The data symbol of the data symbol sequence d (k) is generated in the base station 1 using devices 2 for loading the subscriber-specific codes with a subscriber-specific CDMA code c ^(k) , k = 1. , , K, spread. In the base Station 1 is then a sum signal s by means of a summing element 3. educated. The sum signal s results in

8 = (S ₁ , S ₂ ... S _NQ ) ^T = _CD (1) with the matrix

C = (C _{i, j} ); i = 1 ... QN; j = 1 ... KN; Q = number of chips in the code

C _{Q (n-1) + g, k + K (n-1)} =

die die teilnehmerspezifischen CDMA-Codes enthält, und mit einem kombinierten Datenvektor d = (d₁d₂...d_KN)^T. (3)

which contains the subscriber-specific CDMA codes, and with a combined data vector d = (d ₁ d ₂ ... d _KN ) ^T. (3)

The transmission of the sum signal s over a mobile radio channel 4 with the impulse response h can be done by multiplication with a matrix

dargestellt werden. Das Empfangssignal e ergibt sich zu e = F·s+n = F·C·d+n = A d + n (5) mit der additiven Rauschfolge n. Die Systemmatrix F = (F _{i, j} ); i = 1 ... QN + W-1; j = 1 ... QN;

being represented. The received signal e results in e = F · s + n = F · C · d + n = A d + n (5) with the additive noise sequence n. The system matrix

A = F · C (6) to represent the received signal e

A = (A _{i, j} ); i = 1 ... NQ + WQ-1; j = 1 ... KN;

(8) charakterisiert werden. Eine kombinierte Kanalimpulsantwort entsteht hierbei durch diskrete Faltung eines CDMA-Codes c^(k) mit einer zeitdiskreten Kanalimpulsantwort h mit W Kanalparametern. Wie bereits beschrieben, wird das Summensignal s also über einen Mobilfunkkanal 4 mit der Impulsantwort h übertragen. Im Empfänger 5 eines Teilnehmers k, 1. . . K, ergibt sich dann nach der durch ein Summierglied 6 symbolisierten Addition eines Rauschsignals n das Empfangssignal e. Das Empfangssignal e wird in einem ersten Schritt in einem Kanalentzerrer 7 weiterverarbeitet, um einen Schätzwert s für das gesendete Summensignal s zu erhalten. In einem zweiten Schritt wird dann mittels eines signalangepaßten Filters 8, das speziell auf den spezifischen Code c^(k), k=1 . . . K, des interessierenden Teilnehmers k angepaßt ist , eine codeangepaßte Filterung durchgeführt . Durch diese Maßnahme werden wertkontinuierlicheThe K subscriber signals are transmitted from the base station 1 to the individual subscribers via K different channels, which are combined by the K combined impulse responses b ^(k) = (b ₁ ^(k) , b ₂ ^(k) ...

(8) can be characterized. A combined channel impulse response arises here by discrete folding of a CDMA code c ^(k) with a time-discrete channel impulse response h with W channel parameters. As already described, the sum signal s is thus transmitted via a mobile radio channel 4 with the impulse response h. In the receiver 5 of a participant k, 1.. , K, the reception signal e then results after the addition of a noise signal n symbolized by a summing element 6. In a first step, the received signal e is processed further in a channel equalizer 7 in order to obtain an estimated value s for the transmitted sum signal s. In a second step, using a signal-adapted filter 8, which is specific to the specific code c ^(k) , k = 1. , , K, the subscriber of interest k is adapted, code-matched filtering is carried out. This measure makes the value continuous

k=1 . . . K, der gesendeten Datensymbolfolgen d^{( k)}, k=1 . . . K, bestimmt. Besonders vorteilhaft ist das Verfahren nach der Erfindung, wenn die teilnehmerspezifischen CDMA- Codes orthogonal sind. Im Empfänger 5 des Teilnehmers k müssen somit im Gegensatz zum JD-Algorithmus und zum IC-Algorithmus nur die Kanalimpulsantwort h und der eigene teilnehmerspezifische CDMA-Code, dagegen nicht der CDMA-Code der übrigen Teilnehmer bekannt sein. Zur Bestimmung des Schätzwertes für die gesendete Symbolfolge s kann im Kanalentzerrer 7 ähnlich wie bei dem im Zusammenhang mit der gemeinsamen Detektion (JD, Joint Detection) eingesetzten Algorithmus das ZF- oder MMSE-Kriterium angewandt werden. Die Schätzwerte

ergeben sich für den Fall unkorrelierten Rauschens n mit der Varianz σ² bei Anwendung des ZF-Kriteriums zu

und bei Anwendung des MMSE-Kriteriums zu

Estimates

k = 1. , , K, the transmitted data symbol sequences d ^(k) , k = 1. , , K, definitely. The method according to the invention is particularly advantageous if the subscriber-specific CDMA codes are orthogonal. In contrast to the JD algorithm and the IC algorithm, only the channel impulse response h and the subscriber-specific CDMA code, but not the CDMA code of the other subscribers, need to be known in the receiver 5 of the subscriber k. To determine the estimated value for the transmitted symbol sequence s, the ZF or MMSE criterion can be used in the channel equalizer 7, similarly to the algorithm used in connection with the joint detection (JD, Joint Detection). The estimates

for the case of uncorrelated noise n with the variance σ ² when using the IF criterion

and when using the MMSE criterion

 The estimated data values after the code-matched filtering in the subscriber-specific signal-matched filter 8 result in

where R _s = E (s * s ^{* T} ) is the covariance matrix of the sum signal s. Contain the data values _{c, ZF} according to the ZF criterion a useful and a noise component, the data estimates) _{c, MMSE} according to the MMSE criterion additionally contain an intersymbol (ISI) and multiple access (MAI) interference component.

In the case of MMSE equalization, the covariance matrix R _s , which depends on how many subscribers are currently active and which CDMA codes are used, must be redetermined with each subscriber who is added or is dropped. The covariance matrix R _s has a band and block structure according to FIG. 2 and is Hermitian. In the event that the amount squares of the components of the data vector d and the subscriber-specific CDMA codes c ^(k) , k = 1. , , K, equal to one, all elements on the main diagonal of R _{s have} the value K. The amounts of the values on the Q-1-occupied secondary diagonals are smaller than K. In order to avoid the computationally inverse inversion of the covariance matrix R _s , this matrix, since it is only sparsely occupied, approximated by R _s ≈ K · I (13) with the (NO) · (NO) unit matrix I. This gives the data estimates in the case of MMSE equalization from equation (12)

ausgeführt. Im folgenden wird am Beispiel des MMSE-Kanalentzerrers nach der Gleichung (14) ein Verfahren zur suboptimalen rechenaufwandgünstigen Bestimmung der Datenschätzwerte im Frequenzbereich hergeleitet. Dazu wird ein bereits bekanntes Verfahren angewandt, das eine aufwandsgünstige Matrixinversion im Frequenzbereich erlaubt, sofern die zu invertierende Matrix rechtszirkulant ist. Die in der Gleichung (14) zu invertierende Matrix X hat die in Fig. 3 dargestellte Struktur

The estimated data values according to equations (11) and (14) can be determined either directly by means of matrix inversion or indirectly by means of Cholesky decomposition. In the indirect determination, the estimated values s are first determined using Cholesky decomposition, and then the multiplication with the matrix C ^{* T is used} to determine the estimated value

executed. In the following, using the example of the MMSE channel equalizer according to equation (14), a method for the suboptimal calculation of the data estimates in the frequency domain, which is inexpensive in terms of computing power, is derived. For this purpose, an already known method is used, which allows an inexpensive matrix inversion in the frequency domain, provided that the matrix to be inverted is right-handed. The matrix X to be inverted in equation (14) has the structure shown in FIG. 3

and is therefore only approximately right-wing. In Fig. 3, the same elements are shown by the same hatching.

The matrix is Hermitian, thinly coated and possesses

a band structure, the main and each of the secondary diagonals each consisting of the same values. The matrix

is modified so that the inexpensive

 Inversion in the frequency domain becomes possible, and the resulting inverse is used as an approximation for

used in equation (14). The following modifications to the matrix are made:

The matrix is expanded to the right, i.e. the matrix elements that are missing for the right-hand circle are added. This enables the cost-effective matrix inversion in the frequency domain.

The dimension of the matrix is increased. This reduces the systematic error and thus the additional disruption that the right-hand circular expansion causes, particularly with the first and last data symbols of each participant. It also enables the dimension of the matrix to be a power of two, so that the fast Fourier transform (FFT, Fast Fourier Transform) can be used to solve the problem in the frequency domain. For each element of the main diagonals of the matrix, in addition to the term, see equation (14), which is the

Represents variance of the noise, an additional variance σ ² _{z is} added. This variance σ ² _z represents the additional perturbation caused by the right-hand circular expansion of the originally non-right-hand circulants

 Matrix is inserted. To every element of the

The main diagonals of the matrix are added the same value σ ² _z and not a higher value for the more disturbed first and last data symbols of a participant, so that the right-hand circular structure of the matrix is preserved. The steps to modify the matrix and matrix

inversion are shown in Fig. 4. After calculating the inverse in the frequency domain, a sub-matrix is cut out of the inverse, which has the dimension of the original matrix. This sub-matrix is used in equation (14) instead of the matrix.

Claims

claims 1. Method for separating a received signal mixture e for use in the receivers of the participants of a multi-participant signal transmission system, in which the transmission of the K different participant signals (d ^(k) , k = 1 ... K), each with a different subscriber-specific signal form on the transmission side (c ^(k) , k = 1 ... K,) are applied as a sum signal s over one and the same transmission channel, characterized , that in the receiver of a subscriber k, in a first step, a channel equalization of the received signal mixture e is carried out, which has an estimated value

of the transmitted sum signal s, and that in a subsequent second step a subscriber-specific filtering is carried out, in which the filter coefficients only from the subscriber-specific signal form (c ^(k) , k = 1 ... K,) of the person of interest Participant k are determined and thus have an estimated signal at their output (

k = 1 ... K,) of the interested subscriber k 2. The method according to claim 1, characterized , that an equalizer operating according to the maximum likelihood principle is used for channel equalization of the received signal mixture e. 3. The method according to claim 1, characterized , that a suboptimal equalizer is used for channel equalization of the received signal mixture e. 4. The method according to claim 1, characterized , that for channel equalization of the received signal mixture e one according to the zero-forcing principle (IF), ie with expectations Minimum variance estimation is used for working equalizers. 5. The method according to claim 1, characterized , that for channel equalization of the received signal mixture e according to the minimum mean square error principle (MMSE), i.e. equalizer operating with optimal linear estimate is used. 6. The method according to claim 1, characterized , that the subscriber-specific filtering in the second step is a signal-adapted filtering, adapted to the subscriber-specific signal form (c ^(k) , k = 1 ... K,) of the subscriber k in question. 7. The method according to claim 1, characterized , that the participant-specific signal forms (c ^(k) , k = 1 ... K,) of the participants are orthogonal. 8. The method according to claim 1, characterized , that the received signal mixture e from K transmitted simultaneously, each from finitely long data blocks d ^(k) , k = 1 ... K, data sequences composed of digital data symbols (= subscriber signals) exist which are distorted on the transmission path, with each of the data symbols of a data sequence prior to transmission having a subscriber-specific spreading code assigned to the respective data sequence (= subscriber-specific signal form) c ^(k) , k = 1 ... K, (CDMA code) is modulated, and that in the receiver of the kth subscriber the estimated values k = 1 ... K, the data symbol sequence (= subscriber signal ) of the k-th participant by performing the two successive steps, namely the channel equalization of the received signal mixture g and the subscriber-specific filtering. 9. The method according to claim 8, characterized , that an equalizer operating according to the maximum likelihood principle is used for channel equalization of the received signal mixture e. 10. The method according to claim 8, characterized , that a suboptimal equalizer is used for channel equalization of the received signal mixture e. 11. The method according to claim 10, characterized , that a so-called linear zero-forcing block equalizer (ZF-BLE = Zero Forcing Block Linear Equalizer) is used for channel equalization of the received signal mixture e. 12. The method according to claim 10, characterized , that a so-called minimum mean square error block equalizer (MMSE-BLE = minimum mean square error block linear equalizer) is used for channel equalization of the received signal mixture e. 13. The method according to any one of claims 8 to 12, characterized , that the filtering carried out in the second step is adapted to the subscriber-specific CDMA code of the subscriber k of interest. 14. The method according to any one of claims 8 to 13, characterized , that the participants' CDMA codes are orthogonal. 15. Process according to claims 5 and 12, characterized , that the data values

_{c, MMSE} result from the following equation:

where C is a matrix containing the subscriber-specific codes, F is a matrix containing the channel impulse responses h, σ ^{2 is} the variance of the noise n and I is the unit matrix and K · I is the covariance matrix R _s = E (s · s ^{* T} ) of the sum signal s approximates, where vectors and matrices are written in bold or capital letters, complex sizes are underlined, the complex conjugation or transposition is described by (.) ^* or (.) ^T and the formation of the expected value by E (.), that for the suboptimal determination of the data estimates in the frequency domain the matrix

which is hermitian and thin, and has a band structure, the main and each of the secondary diagonals each consisting of the same values, is modified in such a way that an inexpensive inversion in the frequency domain is possible, the inverse resulting from this being used as an approximate solution for X - is that the following Mon

of the matrix:  - The matrix is expanded to the right, i.e. the for  Right-hand circulatory missing matrix elements are added, which enables the cost-effective matrix inversion in the frequency domain; - The dimension of the matrix is increased, whereby the systematic error and thus the additional disturbance, which causes the right-hand circular expansion especially in the first and last data symbols of each participant, is reduced and, moreover, it enables the dimension of the matrix to be a power of two , so that the fast Fourier transform (FFT, Fast Fourier Transform) can be used for the solution in the frequency domain; - For each element of the main diagonals of the matrix, the variance of the noise is represented in addition to the term

sent, an additional variance o * _{t is} added, which represents the additional disturbance term which is inserted by the right-hand circular expansion of the originally non-right-circulating matrix, with for each element the Main diagonals of the matrix the same value σ ² _z and not a higher value for the more disturbed first and last data symbols of a participant is added so that the right-hand circular structure of the matrix is preserved and that after calculating the inverses in the frequency domain, a submatrix from the inverses is cut out, which has the dimension of the original matrix, and this sub-matrix is then used instead of the matrix in the equation given above.

16. The method according to any one of the preceding claims, g e k e n n z e i c h n e t by using it in the downlink Mobile radio system in the receivers of the K subscriber stations.