DE10106391A1 - Icon-based JD equalization - Google Patents

Abstract

The invention relates to a method for JD equalization of a spread coded signal wherein a channel estimation is undertaken in order to calculate channel coefficients in relation to several subscribers. Equalization coefficients are subsequently calculated in relation to a specific subscriber, taking a system matrix as a basis wherein only one subset of the data symbols contained in a data block is taken into account. The calculated equalization coefficients are used to perform symbolic equalization of the spread coded signal thus transmitted. .

Description

The invention relates to a method for JD equalization spread-coded signal transmitted over a channel and a JD receiver working according to this procedure.

In the UMTS (Universal Mobile Telecommunications System) standard the TDD (Time Divisi on duplex) fashion for the so-called "unpaired band" (that for the Upward and downward route simultaneously provided Fre quenzband). In contrast to the FDD (Frequency Division duplex) mode is the spreading factor for TDD mode ma ximal is 16. Because of this low spreading factor individual participant detection, such as through adapted filtering (MF: Matched Filtering) can be realized can, too inefficient. For a given quality of service (QoS: The commitment will be able to maintain quality of services) powerful JD (Joint Detection) algorithms required Lich.

For JD algorithms (also in German literature as "Algorithms for common detection" referred to) the receiver sees signals from several active participants same cell. The principle of the JD exists in it, by explicit detection not only of the desired one but also to achieve other subscriber signals that the do not contribute to the disturbance. In this way the In tra cell interference (interference from other active participants) significantly reduced or ideally eliminated.

A disadvantage of the previously known JD algorithms is in that this - due to the signal detection of several or of all active participants - a very high rake require effort. This can be done with the usually in Signal processors used by mobile stations cannot be reached  and would not have been more powerful (and thus more expensive) signal processors can be realized, since in this In the event of excessive power consumption.

The invention has for its object a JD detection procedure to specify that for use in mobile radio systems suitable is. In particular, the process is said to be low Power consumption and low signal processing wall with a sufficiently high quality at the same time enable. The invention further aims to provide one JD receiver for mobile radio that has these properties to create systems.

The problem underlying the invention is solved by the features of the independent claims.

Accordingly, two measures essentially make one clear reduction of the Re due to the JD equalization achieved: First, the calculation of the ent distortion coefficients for a particular participant or groove not all received data symbols of a data blocks, but only a subset of the same. Second, the received signal is so used with the help of this calculated equalizer coefficient data symbol for data symbol detected. The signal processing is carried out by both measures maintenance costs significantly reduced, which means that the use of "Symbol-wise" JD equalization according to the invention in mobile radio systems is made possible.

A ZF (Zero Forcing) equalization or a MMSE (Minimum Mean Square Error) equalization carried out.

Preferably, the set of equalizer coefficients includes QWs Coefficients, where Q is the spread factor used by the transmitter tor and Ws an integer between 1 and 10, especially 1 and 5, is. As explained in more detail below, there may be through and in connection with the symbolic equalization one  Reduction of computing time by more than three orders of magnitude compared to previous JD detection methods be enough.

Further advantageous embodiments of the invention are in specified in the subclaims.

The invention is illustrated below with the aid of an embodiment game explained in more detail with reference to the drawing; in this shows:

Figure 1 is a schematic representation of a JD receiver in the form of a block diagram.

Fig. 2 is a schematic diagram illustrating the operation of a JD equalizer according to the prior art;

Fig. 3 is a schematic diagram illustrating the operation of a JD equalizer according to the inven tion; and

Fig. 4 is a schematic representation to illustrate the correspondence between the index k for the selected user and the index j of the zero-forcing condition.

Hil describes the mathematical description of the invention fe of the matrix vector formalism. In the following denote in Superscript T the transposed matrix or the transposed vector and the sign * the complex conjuga tion. H as a superscript symbol stands for an abbreviation for * T. An underscore under a mathematical size means tet that this is or can be complex.

The following abbreviations are first introduced:
K Number of active participants
N Number of data symbols in a block or burst
Q spreading factor
L Channel length, ie the number of paths taken into account in the time-discrete channel model, in chips
W Number of chips taken into account for the equalization
Ws Number of symbols considered for the equalization, ie Ws = W / Q
Ts symbol duration
Tc chip period; Q.Tc = Ts applies
d ^(k) = ( d k | 1... d k | N) ^T vector which contains the complex-value data symbols d k | 1,... emitted by the k-th participant within a burst. ., d represents k | N. The data symbols are sent in symbol cycle 1 / Ts
c ^(k) = ( c k | 1... c k | Q) ^T vector which represents the (subscriber-specific) CDMA code used by the k-th subscriber; where c k | 1,. ., c k | Q the complex chips of the CDMA code under consideration. The chips are sent in chip cycle 1 / Tc
h ^(k) = ( h k | 1.. h k | L) ^T vector, which represents the time-discrete channel impulse response valid for the k-th subscriber; hk | 1,. ., hk | L the complex weights (so-called channel coefficients) of the 1st,. ., Wth transmission path, with adjacent transmission paths each having a time delay corresponding to the chip duration Tc to one another
b ^(k) = ( b k | 1... b k | Q _{+ L-1} ) ^T vector of the combined channel impulse response, which according to b ^(k) = c ^(k) . h ^(k) results from the CDMA code vector and the channel pulse vector with respect to the kth subscriber. * Denotes the time-discrete convolution of the vectors mentioned
n = ( n _{1 ...} n _{NQ + L-1} ) ^T vector of additive perturbations; the vector n represents both thermal noise and multiple access interference, such as adjacent channel interference or intercell interference. The time discretization is based on the chip clock 1 / Tc
e = ( e _{1 ...} e _{NQ + L-1} ) ^T vector of the received, disturbed data signal; where e ₁ . , , e _{NQ + L-1} the detection results obtained at the receiver in the chip clock 1 / Tc.

The starting point of the following illustration is the (multi-part system equation based on vector matrix description on the discrete-time transmission model. This description a transmission system is known in the prior art and Z. B. in detail in the book "Analyze und Design digita ler mobile radio systems "by P. Jung, B. G. Teubner Verlag Stutt gart, 1997, on pages 188-215. This litera by reference is the subject of the present Font.

In order to keep the mathematical representation clearer, a receiver with only one receiving antenna is considered below and it is further assumed that there is no scrambling of the signal on the transmitter side. The invention is not limited to these assumptions. The system equation is:

e = Ad + n (1)

using the following notation:

A = ( A _{µ, ν} ) with µ = 1,. ., NQ + L - 1 (number of lines)
n = 1. ., KN (number of columns)

The elements of the system matrix A are defined by the components of the vector b ^{(k) of} the combined channel impulse response according to the following relationship:

Thus, the vector e is the output of a transmission system described by the system matrix A , which is fed with the input vector d (going back to all active subscribers) and also has an additive noise contribution described by the vector n .

The system described by the system matrix A includes the channel and possibly also the structure of the receiver (e.g. number of antennas (not taken into account here), oversampling). The channel in the above usage includes the physical channel and the transmitter signal processing (used block structure, spreading codes and scrambling scodes, the latter, as already mentioned, not taken into account here for the sake of simplicity).

The system matrix A is approximately known in the receiver: Because the physical channel is estimated by a channel estimator according to the usual procedure and described by its (estimated) channel coefficients (vectors h ^(k) ), spreading and, if applicable, scrambling codes used are assumed to be known.

The (system) input vector d is unknown in the receiver. The aim of the data detection is to determine one or more estimated vectors ^(k) = (k | 1... K | n) ^T for one or more participants in the receiver, which one or which matches the corresponding one as best as possible transmitted vector or vectors d ^(k) = ( d k | 1 ... d k | n) ^T matches or matches.

ML (maximum likelihood) and MAP (maximum a posteriori) criteria due to their excessive computational effort, solutions can be found the JD problem (i.e. to solve the system equation (1)) Not used.

To solve the JD problem, the transmission system, including the receiver, becomes the linear system of equations

= Me (3)

described.

Analogous to the input vector d , the (system) estimate vector relating to all active participants is formed by the sequence of the estimated data symbols:

M is a (KN × NQ + L - 1) matrix and determines the type of data detector. M is referred to below as the estimation matrix.

Fig. 1 illustrates the structure of a JD receiver known as such. The JD receiver comprises a channel estimator CE and a JD equalizer JD-EQ. As already mentioned, the JD receiver receives the time-discretized vector e of the received signal and feeds it to both the channel estimator CE and the JD equalizer JD-EQ. The vectors ^(k) are estimated in the JD receiver according to the prior art for all active participants k = 1,. ., K. In order to determine these estimated vectors, the JD equalizer JD-EQ uses information regarding the channel, namely the number N of data symbols per data block or burst, the spreading codes c ^(k) , k = used by the active participants 1, . ., K, and the channel impulse responses h ( ^k) determined by the channel estimator CE on the basis of the received signal. Statistical information in the form of the covariance matrix R _{n can also be included} in the data detection. The covariance matrix R _n is defined by the expression R _n = E { nn ^H }.

A well-known concept for solving the JD problem is ZF (Zero Forcing). The following estimation matrix is used at ZF:

In the simplest case, without taking any disturbance into account, R _n = Id. Id denotes the identity matrix.

By inserting the estimation matrix M according to equation (4) into equation (3), the system of equations results:

The usual way to solve this system of equations is given below. A detailed presentation of this approach is described in the book by P. Jung mentioned on pages 315-318. The illustration in FIG. 2 serves to illustrate the procedure.

The solution of the system of equations (5) is based on the so-called Cholesky decomposition of the (KN × KN) matrix A ^H R -1 | n A on the left side of the system of equations (5):

The matrix H is a (KN × KN) matrix of the shape

and the matrix Σ is a real (KN × KN) diagonal matrix of the shape

The decomposition yields the matrix H ^H , which is a lower triangular matrix (i.e. a matrix in which all elements in the upper right diagonal half of the matrix are zero), and the matrix Σ ² H , which is an upper triangular matrix (ie a matrix , in which all elements in the lower left diagonal half of the matrix are zero).

Both of the last mentioned matrices are KN × KN matrices. Known sizes in Fig. 2 are shaded, unknown sizes are shown in stripes. It is first of all the computing process in the case of the known block or burst-wise data detection according to the prior art.

The vector e 'results from the matrix vector product of ( A ^H R -1 | n) and e . The unknown vector z , which follows from the matrix-vector product of (Σ ² H ) and, can be determined by recursively solving a trivial system of equations G1, starting with the first component of z . This recursive solving of the system of equations G1 is illustrated in FIG. 2 by an arrow pointing from top to bottom.

With the aid of the now known vector z , the components of the vector are determined from a second trivial system of equations G2, starting with the last component of recursively. The course of the recursive resolution of the second tri vial system of equations G2 is illustrated by an arrow pointing from bottom to top. After the equation system G2 has been dissolved, the (system) estimate vector is calculated.

The procedure for the symbol-by-symbol equalization according to the invention is described below with reference to FIG. 3. This is done in two steps.

In a first step, a reduced system matrix is formed. The reduced system matrix is a (W × K (L + W - 1)) matrix, where W is a number of chips used for the equalization and is called the "equalizer length", and Q ≦ W <NQ. In other words, only a selection or subset of the data symbols contained in a data block or burst are taken into account in this matrix. The reduced system matrix is approximately known to the receiver by estimating the channel coefficients (vectors h ^(k) ) analogously to the system matrix A. (L + W - 1) represents the length of influence of a sent chip on the chip sequence of the received data signal.

The definition of the reduced system matrix is:

As already mentioned, the equalizer length W in becomes Ws = W / Q Units of data symbols. It will depend on it indicated that the block size reduced N in the definition of the System matrix not received.

The system equation corresponding to equation (1) is

= + (10)

where and each W × 1 (column) vectors with one of the Equalizer length are corresponding number of lines.

In the following, a noise-free channel, ie = (0. .0) ^T , is assumed to simplify the illustration. The invention is not limited to this special case.

In the first step, the coefficients of an equalizer, shown here in the form of an equalizer vector m , are calculated using the reduced system matrix and the Cholesky decomposition. The ZF condition is used to calculate the equalizer coefficients

m = _j (11)

based. Here, m is a 1 × W (line) vector, which contains the equalizer coefficients and is referred to as the equalizer vector, and _j is a 1 × K (L + W - 1) (line) vector, which relates to the IF condition of a certain (k-th) participant.

The ZF vector can be represented as follows:

_j = (0 ... 0 1 0 ... 0) (12)

where 1 is at the j-th position of the vector. The jth Position is on the one hand directly a data to be detected symbol assigned and chooses on the other hand, as in the following explained in more detail, a specific user (hereinafter referred to as the kth user) from the majority of users.

To solve the system of equations (11) it is transformed into:

The (line) vector a H | j has the dimension 1 × W and is defined by the expression _j ^H.

The system of equations (13) is solved by Cholesky decomposition of the (W × W) matrix ^H. Either the Choles ky decomposition is carried out directly using equation (13), ie using a line vector equalizer m , or, as illustrated in FIG. 3, the transformed relationship is used for this

used. Here (analogous to the procedure for burst-wise detection), a vector is first calculated by solving a first trivial system of equations G1 'and then the vector m of the equalizer coefficients is calculated by solving a second trivial system of equations G2'. The matrices H ^H (lower triangular matrix) and Σ ² H (upper triangular matrix) obtained during the decomposition are defined analogously to equations (7) and (8), but instead of the quadratic dimension KN here the quadratic dimension W occurs.

When solving the system of equations (14), the IF condition _j forces the equalizer vector to be calculated for the kth subscriber. Here, j is to be selected such that a _j corresponds to the combined channel impulse response b k | 1 of the selected user k. The correspondence between k and the index j of the zero forcing condition is illustrated by the representation shown in FIG. 4.

The system of equations (11) can be solved for each of the K participants with the corresponding IF condition, given by one of the IF vectors _j according to equation (12). The equalizer vector calculated for the k-th subscriber is referred to below as m ^(k) .

In a second step, the received data stream is detected symbol by symbol for the k subscriber with the aid of the previously calculated equalizer vector m ^(k) . The second step is illustrated in the lower part of FIG. 3. The detection is carried out by cross-correlating the chips of the received data signal with the equalizer coefficients calculated for the k-th user according to the equation

E is a Q (Ws + N - 1) × 1 column vector of the received chip sequence, ' ^(k) denotes the estimate vector for d ^(k) calculated according to the invention and the matrix M ^(k) is an N × Q (Ws + N - 1) Matrix, which is built up from the equalizer vector m ^(k) in the following way:

That is, the elements of the equalizer vector m ^{(k) are} entered in each row of the matrix M ^{(k) and} are offset from one another by Q positions with respect to adjacent rows. The remaining elements of the matrix are zeros.

From equation (15) it follows that for the calculation of a Data symbols W complex multiplications are required.

With the help of the following table the blockwei detection (state of the art, first column of the table)  required computational effort with the inventive Computational effort for a Participants (1st example of invention, second column of the table) or all participants (2nd invention example, third Compare column of the table).

The comparison values are the example values K = 9, N = 69, Q = 16 and Ws = 3 based. The table shows the number of complex multiplications, which in each case in the Rows of the table require arithmetic operations are. With block-wise data detection according to the status the technique is the Cholesky decomposition in this example on a 621 × 621 matrix. In contrast, the Cholesky Disassembly in the symbolic detection on a 48 × 48 Matrix.

When solving the first and the second system of equations G1 ', G2' (first step) are in front of the invention way (Examples 1 and 2) much less complex Multiplications than in the prior art, where  the number of complex multiplications required determined by the spreading factor Q and the equalizer length Ws is. The smaller Ws, the less computing effort.

With equalization using the previously calculated coefficients th of the equalizer are used to calculate a data symbol in the second step only W = QWs = 48 complex multiplication NEN is required, which means that the calculation of the 69th Data symbols of a data block only 3312 complex multi complications arise.

Steps 1 and 2 can only be used for one or for several sending participants. In two In these cases, these steps can be carried out participant by participant be carried out sequentially, d. H. in the sequence step 1 (Participant 1), Step 2 (Participant 1), Step 1 (Part Participant 2), step 2 (participant 2),. , ., Step 1 (part participant K), step 2 (participant K).

This is due to the comparatively low computing effort Method according to the invention especially for equalization of signals in the TDD-unpaired band of the UMTS standard for Mo bilfunk can be used.

The JD receiver structure shown in FIG. 1, usually implemented by a DSP (DSP: digital signal processor), can also be used for the equalization according to the invention. Instead of the system matrix A , the reduced system matrix is then used, and instead of the block-wise equalization, symbol-wise equalization is carried out as described above. In the second step (ie in the case of symbolic equalization based on the previously calculated equalizer coefficient), no additional DSP capacity is generally required, since a suitable multiplication field is usually already present in the DSP.

Claims (8)

1. A method for JD equalization of a spread-coded signal transmitted over a channel, comprising the steps:

• Performing a channel estimation to calculate channel coefficients ( h ^(k) ) with respect to several participants,
• Calculating a set of equalizer ^coefficients ( m ^(k) ) with respect to a specific subscriber (k) on the basis of a reduced system matrix (), in which only a subset of the data symbols contained in a data block of the transmitted signal is taken into account,
• - symbol-wise equalization of the transmitted spread-coded signal using the calculated equalizer coefficient set ( m ^(k) ).

2. The method according to claim 1, characterized, that ZF equalization or MMSE equalization is carried out leads. 3. The method according to claim 1 or 2, characterized in that a Cholesky decomposition of a matrix is used to calculate the set of equalizer ^coefficients ( m ^(k) ), which is based on the reduced system matrix (). 4. The method according to any one of the preceding claims, characterized in

• - That the set of equalizer ^coefficients ( m ^(k) ) contains QWs coefficients, where Q is the spreading factor used on the transmitter side and Ws is an integer between 1 and 10, in particular 1 and 5.

5. The method according to any one of the preceding claims, characterized,  that the procedure for equalization of signals in TDD unpaired band of the UMTS standard used for mobile communications becomes. 6. The method according to any one of the preceding claims, characterized in that equalizer coefficient sets ( m ^(k) ) for several participants mer (k) are calculated and the symbolic equalization is carried out for several participants (k). 7. The method according to claim 6, characterized, that the symbolic equalization for the participants (k) se is carried out sequentially. 8. JD receiver for equalizing a spread-coded signal transmitted over a channel,
with a channel estimator (CE), which carries out a channel estimation for the calculation of channel coefficients ( h ( ^k) ) with respect to several participants (k) on the basis of the spread-coded signal transmitted over the channel,
with a JD equalizer (JD-EQ) for calculating a set of equalizer ^coefficients ( m ^(k) ) with respect to a specific subscriber (k) on the basis of a reduced system matrix () in which only a subset of those in a data block of the transmitted signal contained data symbols is taken into account, and for symbol-wise equalization of the transmitted spread-coded signal using the calculated equalizer coefficient set ( m ^(k) ).