Method for operating a digital mobile radio network with space-time block codes

Abstract

The invention relates to the preparation and numerical optimization of non-linear space-time block codes for application in a digital mobile radio system, with a maximum transmission diversity for “Rate 1,” transmission systems in the case of two or more transmitter antennae and with complex symbols.

Description

CROSS REFERENCE TO RELATED APPLICATIONS

This application is based on and hereby claims priority to PCT Application No. PCT/EP02/02330 filed on 4 Mar. 2002 and European Application No. 01107920.9 filed on 28 Mar. 2001 and German Application No. 101 15 261.2 filed on 28 Mar. 2001, the contents of which are hereby incorporated by reference.

BACKGROUND OF THE INVENTION

The present invention relates to a method for operating a digital mobile radio network using space-time block codes (ST codes).

Such methods are for example disclosed in “Space-Time Codes for High Data Rate Wireless Communication Performance Criterion and Code Construction” by V. Tarokh et al. in IEEE Transactions on Information Theory, vol. 44, No. 2 Mar. 1998, pages 744-765 or in “Space-Time Block Codes from Orthogonal Designs”, IEEE Transactions on Information Theory, vol. 45, No. 5, July 1999, pages 1456-1467 by V. Tarokh et al.

Use of the time-space block codes disclosed there is subject to the following physical problem in mobile radio networks: the high level of attenuation or distortion caused by the transmission path of a transmission signal transmitted in a wireless manner from a transmitting station (e.g. a fixed base station in a mobile radio network) makes it very difficult for a receiving station (e.g. a mobile station in a mobile radio network) to identify the originally transmitted transmission signal correctly. This applies in particular, when the receiving mobile station is located in a multipath wireless environment, where it only receives a plurality of highly attenuated echoes of the original transmission signal due to multiple reflections of the original transmission signal off the walls of surrounding buildings. To rectify this, a certain “diversity” of the received signal has to be ensured. This is achieved by providing the receiving station with one or more additional, less highly attenuated “images” of the transmitted transmission signal. What is known as the “diversity order” here is a measure of the number of statistically independent “images” of the transmission signals transmitted by the transmitter received at the receiver.

In practice, this necessary diversity is achieved, for example, by using a plurality of transmitter or receiver antennas, each positioned spatially apart from each other or with different polarities on the transmitter and/or receiver side, each of which transmits or receives an “image” of a signal to be transmitted. “Image” here does not necessarily mean that two or more exactly identical copies of the same signal are transmitted.

Instead, even with the known space-time block codes, different signals can be transmitted at a defined time from the different transmitter antennas, each of the signals being created by a specific algorithm from the data bits of the original signal. Each receiver antenna then receives the sum of the transmission signals, modified due to the transmission path, at a defined (later due to the signal runtime) time, from which signals those data bits, which correspond with the greatest likelihood to the data bits of the original signal, are reconstructed on the receiver side using MLD (maximum likelihood detection) estimate algorithms.

See FIGS. 1 and 2 for a clearer explanation of the problem.

FIG. 1 shows a diagrammatic illustration of a known single antenna to single antenna radio transmission link between a transmitter Tx and a receiver Rx in a digital mobile radio network.

Data bits are fed into the transmitter, comprising, for example, a series of ones and zeros. Let it be assumed that a bit vector (1,0,1) of length l=3 is fed in, comprising the series 1-0-1. Depending on the coding method used, this bit vector is transformed on the transmitter side with one-to-one correspondence (i.e. with reversible uniqueness) into a symbol vector of length n. As shown in FIG. 1 by different shades of gray, the figures occurring in the symbol vector do not have to be 1 and 0. In particular, the figures occurring in the symbol vector can also be complex figures, with which for example the signals corresponding to the real component and the imaginary component are sent with a phase offset of 90° in respect of each other.

FIG. 1 shows the case where n=l=3, i.e., each individual bit is coded into an individual symbol (BPSK modulation—binary phase shift keying; with QPSK—quadrature phase shift keying n=l/2 would apply).

A symbol vector comprising a plurality of individual symbols is then transmitted by an antenna in the transmitter Tx and received by a receiver Rx. The symbol vector of length n=3 is reconstructed there by a reverse transformation into the original bit vector (here (1,0,1)) of length l=3.

FIG. 2 shows a diagrammatic illustration of a radio transmission link between a transmitter unit with n=3 transmitter antennas Tx1, Tx2, Tx3 and a receiver Rx in a digital mobile radio network, with which space-time block codes already known per se are used. A bit vector of length l=3 is fed to a space-time block coding device (ST coder=space time coder). This maps an incoming bit vector onto an n×n matrix.

Here, n corresponds to the number of transmitter antennas used in the transmitter unit. In a defined time slot j an antenna i sends a signal, which corresponds to the matrix element c_ijk of a 3×3 matrix C_k, which was coded by the ST coder from the incoming vector of length l=3. k here is an index, which differentiates individual matrices, which in turn correspond for their part to k different bit vectors. These relationships are described in more detail below.

As a result of the signals transmitted via the three transmitter antennas Tx1, Tx2 and Tx3, signals arrive at the receiver antenna Rx at a receiving time corresponding to the transmission time slot j (later due to runtime), where the signals correspond to the matrix elements c_1jk, c_2jk, c_3jk. On the receiver side, a space-time block matrix C_k is reconstructed from the sum of the incoming signals in a space-time block decoding device (ST decoder) by MLD algorithms and translated back by reverse mapping into the corresponding original bit vector (here (1,0,1)).

The general problem here is maximizing diversity at the mobile station of a digital mobile radio network by using a plurality of transmitter antennas at the base station. No specific prior knowledge about the downlink channel (from a base station to a mobile station), which changes over time, should be assumed.

In the case of linear space-time block codes with two antennas, the result is known (it is optimal in the sense that it doubles the diversity for two antennas) and it has become part of the UMTS standard (3GPP TS 25.211 V3.4.0: Physical channels and mapping of transport channels onto physical channels (FDD) (1999 edition), September 2000). This known space-time block code scheme satisfies the “rate 1” requirement.

A “rate 1” space-time block code scheme can be described as a system, in which, for each time interval considered, precisely the same number of data bits can be sent through effectively from a transmitter to a receiver as with the reference system shown in FIG. 1 with only one transmitter and one receiver antenna. In other words, a “rate 1” ST code has the same transmission rate compared with a basic system with only one transmitter and one receiver antenna. This is for example a system in which a block of two code words is transmitted to the receiver at the same time in two successive time windows via two different transmission channels; the receiver thereby receives precisely the same amount of information per unit of time as if two corresponding individual bits had been transmitted in two successive time windows via a single transmission channel. Provision is therefore made in the W-CDMA mode of the UMTS standard for future mobile radio systems (see for example www.3GPP.org) to transmit the standard mapping of four bits onto 2⁴=16 ST symbols in two time stages via two transmitter antennas. A “rate 1” system for external blocks therefore behaves in exactly the same way as the basic system with only one transmitter and one receiver antenna. This characteristic is decisive when upgrading a basic system to a system with a plurality of transmitter antennas and space-time block codes.

The space-time block code used in W-CDMA mode corresponds to what is known as the Alamouti code. This is a code that is very simple to reconstruct on the receiver side to increase diversity in a digital mobile radio network with a transmitting station with two (n=2) transmitter antennas. The Alamouti code is for example disclosed in “A simple Transmitter Diversity Scheme for Wireless Communications”, IEEE J. Select Areas Commun, vol. 16, pages 1451-1458, October 1998 by S. M. Alamouti or in the two publications by V. Tarokh referred to above.

In order to increase diversity in a mobile radio network further in the future both in the uplink direction (from a mobile station to a base station) and in the downlink direction (from a base station to a mobile station), the number of antennas per sector of a base station should be greater than two. It is therefore clear that space-time codes are required, which can be used in the case of three, four or more transmitter antennas. An increase in diversity with the same transmission power results in an increase in receiving quality. Or, looked at another way, an increase in diversity with the same receiving quality means a reduction in transmission power. The transmission power then not used up in a transmitter can in turn be used to supply more users.

The performance of a space-time block code is also influenced by the intervals between the code words. Observations on this are contained in the two articles by V. Tarokh referred to above.

In the transmission shown diagrammatically in FIG. 2 the matrix shown with the elements c_ijk corresponds to a code word with the number k. The “interval” between code words, i.e. the “interval” between two matrices C_k1 and C_k2 is ultimately a measure of the quality of a transmission code, i.e., the likelihood with which the originally transmitted code words can be reconstructed as uniquely as possible from a sequence of code words received at the receiver with distortion due to the transmission path. In this respect the known space-time blocks in the case of two transmitter antennas can be further improved.

In the case of three, four or more antennas (n>2) it is shown in “Space-time Block Codes from Orthogonal Designs”, IEEE Transactions on Information Theory, vol. 45, No. 5, July 1999, pages 1456-1467 by V. Tarokh et al. that linear codes with “rate 1” and complex symbols cannot exist.

In a presentation given at Globecom 2000 in December 2000 on “Complex Space-Time Block Codes for Four Tx Antennas” by Olav Tirkkonen and Ari Hottinen, the case of n=4 transmitter antennas was examined and complex space-time block codes specified with a rate of ¾.

The search for space-time codes, in particular for a number n>2 of transmitter antennas, has therefore taken two directions:

1. In “Space-time Block Codes from Orthogonal Designs”, IEEE Transactions on Information Theory, vol. 45, No. 5, July 1999, pages 1456-1467 by V. Tarokh et al., linear space-time codes with a rate less than 1 are constructed for more than two transmitter antennas. Here the space-time symbols are linear combinations of the original signals. These constructions are not available for external coding, because they do not have the “rate 1” characteristic. This means that they cannot be integrated as simple additional features in an existing mobile radio system without space-time coding.

2. Space-time trellis codes are constructed by combining space-time codes with external error correction codes. Combining space-time mapping with external coding however makes it impossible to integrate space-time codes as a simple additional feature in an existing system. Also in such a case proven external coding techniques such as turbocodes (see C. Berron, A. Glavieux and P. Thitimajshima: Near Shannon limit error correcting codes and decoding: turbo codes” in Proc. IEEE ICC, Geneva, May 1993, pages 1064-1070)) or trellis codes must be modified.

This reveals a problem in that the digital code words to be transmitted between the transmitting and receiving stations in a digital mobile radio network should be optimized in the case of two or more transmitter antennas with regard to the following:

1. There should be a “rate 1” code if possible.

This is a mandatory requirement in order to be able to upgrade mobile radio networks already commercially available with the lowest possible upgrade costs for the use of the new space-time block codes. When “rate 1” is used when upgrading from a known single antenna to single antenna system shown in FIG. 1 to a multi-antenna to single antenna system as shown in FIG. 2 (or even to multi-antenna to multi-antenna systems), the assemblies (not shown in FIGS. 1 and 2) required on the transmitter side to generate the bit vector to be fed in and the assemblies (also not shown) required on the receiver side to further process the reconstructed bit vectors output by the receiver unit do not have to be changed. “Rate 1” codes thus guarantee “downward compatibility” of digital mobile radio stations operated with multi-antenna units and corresponding space-time block codes with other already existing system components, thereby eliminating a significant obstacle to investment impeding the practical introduction of “rate≠1” systems, as with these the assemblies for transmitter-side bit vector generation and for receiver-side bit vector reconstruction must also be changed.

The introduction of “rate 1” space-time blocks means that it is also possible to leave the other parameters of a transmitter code scheme (such as channel coding, interleaving, service multiplexing, etc.) unchanged.

2. The code should be simple to construct on the transmitter side and simple to reconstruct on the receiver side.

3. The code words should have the “biggest interval” possible between them. This means that a set of code words should be structured so that from the signals received on the receiver side with noise and/or distortion, each of which comprises the original signal transmitted on the transmitter side times a “fading factor” describing the fading of intensity associated with increasing distance plus noise (thermal noise at the input amplifier of the receiver plus interference noise due to disruptive signals from other users of the mobile radio network), the original signal can be reconstructed even with a relatively high level of interference in the most error-free possible manner as the signals transmitted on the transmitter side (i.e. without confusion between individual code words).

4. The space-time block code used should maximize diversity, i.e., for two transmitter antennas the theoretically maximum degree of diversity 2 should be achieved if possible, for three transmitter antennas the theoretically maximum degree of diversity 3, etc.

5. The space-time block code used should allow complex transmission symbols, in order for example that it can be used with UMTS where QPSK (quadrature phase shift keying) modulation is used. Complex symbols also allow 8-PSK (8-phase shift keying) or M-QAM (M-fold quadrature amplitude modulation).

SUMMARY OF THE INVENTION

One of the potential objects of the present invention is to optimize the known methods for operating a digital mobile radio network as far as possible by using alternative space-time block codes in respect of the above criteria 1 to 5. In particular “rate 1” space-time block codes with maximum transmission diversity should be prepared in the case of three or more transmitter antennas (n>2). Space-time block codes with interval optimization should also be prepared in particular in the case n≧2.

According to one aspect of the invention, space-time block codes can be constructed from unitary n×n matrices.

In particular in the case of n>2, i.e., three and more transmitter antennas, it is demonstrated that space-time block codes for mobile radio systems with n transmitter antennas and m receiver antennas with a complex ST modulation scheme and with the degree of diversity n×m and “rate 1” essentially exist, as they can be constructed from unitary n×n matrices.

Further embodiments relate to operating a digital mobile radio network with orthogonally structured space-time block transmission codes with maximum diversity. In the case of two or more (n≧2) transmitter antennas, optimized space-time block codes with a complex ST modulation scheme with the degree of diversity n×m and “rate 1” are determined numerically, with which the code symbols have the greatest possible “interval” in the sense of a metric specified in each instance.

Further embodiments pertaining to the matter relate to base and mobile stations in a digital mobile radio network in which reference tables, which contain the matrix elements of code words used in the methods, are stored as well as computer program products in which corresponding reference tables are implemented.

According to the applicant's proposal, measures are specified in particular for the construction of “rate 1” space-time block codes for three, four or more antennas, with which a maximum diversity amplification is achieved.

This is considered to be a significant theoretical and practical breakthrough and is based on two changes compared with the procedures known from the related art:

1. Non-linear codes are used. This does not represent a problem, as the number of code words in space-time blocks is 2″ for BPSK (binary phase shift keying); digital frequency modulation technique for sending data via a coaxial cable network: this type of modulation is less efficient but also less subject to noise than similar modulation techniques, such as for example QPSK (quadrature phase shift keying) and 4″ for QPSK modulation. In other words, even for n=4, when a QPSK modulation is used, only 256 code words are used and these can be stored easily in a table.

2. Neither the transmitted energy per antenna and time unit nor the energy conveyed via all the antennas at a predetermined time is kept constant. This is not critical with W-CDMA in particular, as greater fluctuations can also occur in the emitted energy in a standard system due to the superimposition of different user signals. The additional fluctuations in power, introduced by the use of new space-time codes, are negligible in comparison.

BRIEF DESCRIPTION OF THE DRAWINGS

These and other objects and advantages of the present invention will become more apparent and more readily appreciated from the following description of the preferred embodiments, taken in conjunction with the accompanying drawings of which:

FIG. 1 shows a diagrammatic illustration of a radio transmission link and a corresponding symbol vector coding and decoding for a known single antenna to single antenna arrangement in a digital mobile radio network;

FIG. 2 shows a diagrammatic illustration of a radio transmission link and a corresponding symbol coding and decoding for a space-time block coding known per se in a multi-antenna to single antenna arrangement in a digital mobile radio network;

FIG. 3 shows in table form a specific example of a BPSK-compliant coding of eight bit vectors into the matrix elements of eight complex unitary matrices (ST symbols), which are used as the basis for a space-time block coding during the implementation of a method according to one aspect of the invention for n=3 transmitter antennas;

FIG. 4 shows the intervals between eight ST symbols in the case of a space-time block coding discussed in relation to FIG. 3.

FIG. 5 shows the comparison in the code spectrum for an inventively constructed interval-optimized space-time block code for the case SU(2) in the case of n=2 transmitter antennas and a space-time code constructed according to the known Alamouti scheme, which is not interval-optimized;

FIG. 6 shows a code spectrum for an inventively constructed interval-optimized space-time code for the case SU(3) in the case of n=3 transmitter antennas and QPSK modulation using a minimum standard;

FIG. 7 shows a code spectrum for space-time block codes constructed by inventive numerical optimization methods in the case of n=3 transmitter antennas and BPSK modulation using different standards;

FIG. 8 shows a BPSK simulation for two and three antennas using an L_min code;

FIG. 9 shows the spectrum of minimum eigenvalues with different inventive codes for n=3 transmitter antennas and BPSK modulation; and

FIG. 10 shows a BPSK simulation for n=3 transmitter antennas using an L₋₁ code.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT

Reference will now be made in detail to the preferred embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to like elements throughout.

To give a better understanding, constructive proof is first provided to show that complex space-time block codes with maximum diversity exist not only for the case n=2 transmitter antennas (as shown for example by the known Alamouti scheme) but that complex space-time block codes with maximum diversity also exist for the case with n>2 transmitter antennas.

The mathematical structure is relatively complicated. In practice however, the method can be reduced to the use of a reference table for space-time code words, which are stored in the receiver and transmitter. FIG. 3 shows an example of such a reference table. Its structure and method of use are explained in more detail below. These tables can then be used without a deeper knowledge of their without a deeper knowledge of their derivation being required.

The mathematical notation used in the present application is for example explained in Simon Barry, “Representation of Finite and Compact Groups”, Graduate Studies in Mathematics, Volume 10, American Mathematical Society or in Bronstein, Semendjajew: “Taschenbuch der Mathematik” (Pocket Book of Mathematics), p. 155. published by Harri Deutsch, Thun and Frankfurt/Main, ISBN 3 87144 492 8.

The inventors propose a conceptually new approach is defined for space-time codes. First the transmitter units are disclosed. Next a standard multi-input-multi-output fading channel is disclosed. Finally, the process for recording space-time symbols is disclosed. On this basis, constructive proof of the existence of space-time codes with maximum diversity is given for mobile radio systems with n≧2 transmitter antennas. This proof covers cases with complex and real space-time symbols.

Although the proof of existence is constructive, it does not provide an optimum space-time code, i.e., a space-time code with optimized “intervals” between the code words. Therefore practical methods for constructing codes based on numerical optimization are defined. Finally the results are also confirmed by simulations.

Transmitter

Let us assume first that there are n transmitter antennas and one receiver antenna.

Extension to n transmitter antennas and m receiver antennas looks like this:

In the case of m>1 receiver antennas, the codes and detection methods disclosed below can be implemented separately in each instance for all the m receiver antennas. The results are then combined in what is known as a “maximum ratio combining” method. Such a “maximum ratio combining” method is for example described in J. Proakis, M. Salehi: Communication Systems Engineering, Prentice Hall, 1994, ISBN: 0-13-306625-5.

This then gives a maximum degree of diversity of n×m.

In particular the extension to m>1 receiver antennas does not influence optimization of the transmitter side. This can be identified for example from the equation (8) in Tarokh et al., “Space-Time Codes for high data wireless communication: performance criterion and code construction”. Here it is shown that the likelihood of mutilation between two sequences is equal to the product of the likelihood of mutilation for only one receiver antenna in each instance and all product components are of equal size.

Accordingly, the minimization of a product component results in the minimization of the product as a whole. This means that the number of receiver antennas has no influence on the optimality of the transmission symbols.

To this extent all the transmission methods disclosed below can be used without change for any number of receiver antennas.

A space-time block coding (“block ST modulation”) is described as the mapping of a total of 2^l different bit vectors {right arrow over (b)}εB^l (each of which comprise l bits ε{0,1}) onto a set of 2^l space-time symbols C({right arrow over (q)})εU(n), which are described as unitary n×n matrices by the mapping:STM:B^l→U(n){right arrow over (b)}→C({right arrow over (b)})  (1)

The modulation rate is R=1, if the l input bits are arranged in n symbols, with each symbol comprising l/n bits (e.g. l/n=1 for BPSK and l/n=2 for QPSK).

The mapping of bit vectors {right arrow over (b)}_k onto the symbols transmitted in n stages via an antenna (see FIG. 1) is now replaced by the mapping of bit vectors {right arrow over (b)}_k onto a set of space-time symbols ST, all of which are transmitted via n antenna in n stages. The application of these space-time codes is then without influence (transparent) for external transmission blocks, e.g. such as a channel coder.

At the transmitter the matrix element c_ijk of the 2^l unitary n×n matrices C_k are then processed as space-time variables as follows: if the unitary n×n matrix C_k as an ST symbol corresponds to a bit vector {right arrow over (b)}_k to be transmitted, a signal corresponding to the matrix element c_ijk (with line index i=1, . . . , n, column index j=1, . . . , n) is transmitted from the ith antenna in time window j. With this set-up, the transmission time period for a complete ST symbol is then n time units.

As all matrices C_k are unitary, i.e., C_k^†C_k=C_kC_k^\=1, the lines and columns of C_k are orthonormal. This implies that the power transmitted from each antenna (averaged over time) is identical. Also the total symbol energy (added together over all antennas) is constant in each time slot. It should be noted that the transmission power E_b is automatically standardized and independent of n, due to the fact that the C_k are unitary.

Channel Model

A fading channel with simple Rayleigh or Rician distribution is assumed below. Transmission paths (from each antenna) are subject for example to independent Rayleigh fading in each channel state α_ij and it may be assumed that the channel does not change significantly during the transmission time of an ST symbol (corresponding to n time slots). Let the signal to noise ratio in each channel be the same, i.e. γ_b=E_b^|N₀.

Receiver

With the modulation and channel model described above, the receiver will receive a signal at a receiving time, which is offset due to the runtime and corresponds to the transmission time j, the signal resulting from the summation of all matrix elements c_ijk, j=const., in a matrix column (fixed transmission time j), for a specific code word C_k, in each instance weighted by a fading factor specific to the channel state, with noise still having to be taken into account. This means, when transmitting a code symbol C_k, which corresponds to a bit vector {right arrow over (b)}_k, that the signal (the receiving vector)

$r_j^k=\sum _{i=1}^n{c}_{\mathrm{ijk}}{\alpha }_i+\mathrm{noise}$ is present on the receiver side at a receiving time, which is offset due to the runtime. In other words, the channel state characterized by a does not depend on the space-time symbol C_k sent and is constant for all n time slots.

Let the index k be omitted below for reasons of clarity.

If we write the receiving vector as {right arrow over (r)}=(r₁, . . . , r_n)^T and the channel state vector as {right arrow over (a)}=(a₁, . . . , a_n)εCⁿ, this can be written in matrix form as:{right arrow over (r)}({right arrow over (b)})=C({right arrow over (b)}){right arrow over (a)}+{right arrow over (z)},  (2)

where {right arrow over (z)} describes the noise vector distributed as standard. From the point of view of the receiver the receiving symbols {right arrow over (r)}({right arrow over (b)}) are not fixed but are for their part also stochastically distributed variables (i.e. they are dependent on the channel state!). The overall modulation is therefore a mapping of the bits onto the n-dimensional complex number level:B^l→U(n)→Cⁿ  (3){right arrow over (b)}→C({right arrow over (b)})→{right arrow over (r)}({right arrow over (b)}).  (4)

For a good code this mapping must of course be uniquely reversible, for each channel state {right arrow over (a)}≠0. This imposes basic conditions on the set of ST symbols {C({right arrow over (b)})}_{{right arrow over (b)}εB}^l. In fact for any two symbols C({right arrow over (b)}_i) and C({right arrow over (b)}_j), {right arrow over (b)}_i≠{right arrow over (b)}_j the Euclidean interval cannot be zero, i.e. the following must apply:∥C({right arrow over (b)}_i){right arrow over (a)}−C({right arrow over (b)}_j){right arrow over (a)}∥₂≠0  (5)

for any state vector {right arrow over (a)}, which is not equal to the zero vector.

The following must apply in particular:└C({right arrow over (b)}_i)−C({right arrow over (b)}_j)┘·{right arrow over (a)}≠0

for all {right arrow over (a)}≠0 (zero vector).

If this condition is satisfied, a unique solution must be defined for the present linear equation system, i.e. det└C({right arrow over (b)}_i)−C({right arrow over (b)}_j)┘≠0, where the determinate is equal to the product of the eigenvalues of └C({right arrow over (b)}_i)−C({right arrow over (b)}_j) ┘.

Minimizing the equation (4) for all {right arrow over (a)} on condition that ∥{right arrow over (α)}∥=1 gives an eigenvalue equation for the matrix C({right arrow over (b)}_i)−C({right arrow over (b)}_j).

Thus the minimum interval (determined by the “worst case” channel α″ with the greatest signal distortion) for a defined pair of ST codes is equal to the minimum eigenvalue λ_min of C({right arrow over (b)}_i)−C({right arrow over (b)}_j). λ_min must be positive so that there is reversibility.

The following statements are equivalent:

The eigenvalues λ_min of C({right arrow over (b)}_i)−C({right arrow over (b)}_j) are not equal to zero for any pair of ST symbols.

For all i, j(i≠j) the rank of C({right arrow over (b)}_i)−C({right arrow over (b)}_j) is equal to n. For all i, j(i≠j) the determinant det(C({right arrow over (b)}_i)−C({right arrow over (b)}_j)) cannot disappear.

Given equation (2) an optimum MLD (minimum likelihood detector) minimizes the interval in respect of all possible channel symbols {right arrow over (r)}_j=C({right arrow over (b)}_j)â, where an estimated value â is used for the channel state information (e.g. by pilot sequences):

$\hat{b}=\arg \underset{j}{\min }\overrightarrow{r}-C\left({\overrightarrow{b}}_j\right)\hat{\alpha }.$

A logarithmic likelihood value (“soft symbol”) can be derived for the bit vector {circumflex over (b)} from the interval ∥{right arrow over (r)}−C({right arrow over (b)}_j){circumflex over (α)}∥. If mapping of the bits in symbols (Gray Code or similar) is specified (e.g. by pilot sequences) an LLR (log likelihood ratio) value

$\log \left(L\left(x\right)=\frac{P\left(b=0\right)}{P\left(b=1\right)}\right)$ could also be determined for each bit.

Existence of Time-Space Symbols (ST Symbols) with Maximum Rank

The question is whether there is a set of 2^l ST symbols (matrices C_k) for every number n of antennas, with the characteristic that they have a maximum rank. In other words

$\begin{array}{cc}{d}_{\min }=\underset{i\ne j}{\min }\det \left(C_i-C_j\right)& \left(9\right)\end{array}$

has to be determined for all pairs of C_i, C_j.

In the case of unitary n×n matrices, their rank is always equal to n. The decisive question for the ST codes considered now is whether all the differences between the selected matrices have full rank (n).

If we can construct a set of matrices C={C_i} in such a way that they form a group in respect of standard matrix multiplication, things become simpler:

Lemma 1:

Let C={C_i} be a group. If C_i=1 is the unit matrix of the group, then

$d_{\min }=\underset{i\ne j}{\min }\det \left(1-C_j\right)$ applies.

Proof:

Due to the group characteristic, C_i⁻¹=C_i^†εC. This means that C_jC_i⁻¹=C_kεC and the following applies:min_ij=|det(C_i−C_j)|=min_ij|det(1−C_jC_i^†)|=min_k|det(1−C_k)|.

Use is also made here of the fact that |det(C)|=1.

Example 1

The Alamouti scheme is based onC=1┌1±j1±j┐2└−1±j1∓j┘ etc.

Actually, the Alamouti scheme comprises a subset of the cube group with the order o=24. Eight of the group elements are not used, although they give the same d_min=1. In this scheme in principle therefore Id(24)>4 bits could be transmitted without data loss. (However the symbols not transmitted have a poor crest factor, e.g. C=1).

It is known from the publication “Space Time Codes for High Data Wireless Communication: Performance Criterion and Code Construction” by V. Tarokh, referred to above, that the rank n of n×n code matrices constructed as space-time block symbols is equal to the degree of diversity. It therefore also follows that the rank of the difference between two n×n code matrices is maximum n. This knowledge is used below.

Complex Symbols

Theorem 2:

For any l and n≧2 there is a complex ST modulation scheme with a degree of diversity of n.

Proof: by Specific Construction

Note that, in this case, the ST code {C_k({right arrow over (b)}_k)}_{{right arrow over (b)}εB}¹ comprises a subset of all possible unitary n×n matrices. On the basis of the spectral theorem set out in Simon, Barry: Representations of Finite and Compact Groups, 1996, Graduate Studies in Mathematics, American Mathematical Society, ISBN 0-8218-0453-7, every unitary matrix C can be written as:

$C=V\left(\begin{array}{ccc}{\lambda }_1& & 0\\ & \cdots & \\ 0& & {\lambda }_n\end{array}\right)V^{-1}$

where V is unitary and λ_i=exp(jβ_i) (in the argument for the function exp, j is the imaginary unit). Let us now select

${\beta }_i=\frac{2\pi }{2^l}{q}_i,$ where q_i is any odd whole number. Then C^2l=1 applies and {C_k}_{k=0 . . . 2l−1} is an (Abelian) group, with a generator

$\begin{array}{cc}C=V\left(\begin{array}{ccc}\exp \left(\frac{2\pi j}{2^l}{q}_1\right)& & 0\\ & \cdots & \\ 0& & \exp \frac{2\pi j}{2^l}{q}_n\end{array}\right)V^{-1}& \left(7\right)\end{array}$

for any fixed V. The eigenvalues of C_k are

${\lambda }_i^k=\exp \left(\frac{2\pi j}{2^l}{q}_i^k\right).$

Constructing with k=0 . . . 2^l−1 means that they are clearly different from 1. Based on this it can be shown thatd_min=min_k≠0|det(1−C_k)≠0.

As |det(V)|=1,

$d_{\min }={\min }_{k\ne 0}\det \left(1-{\left(\begin{array}{ccc}{\lambda }_1& & 0\\ & \cdots & \\ 0& & {\lambda }_n\end{array}\right)}^k\right)={\min }_{k\ne 0}\prod _{i-1}^n\left(1-{\lambda }_i^k\right)\ne 0$

Note that the key to the above proof is based on the fact that it is always possible to construct a set of unitary matrices, the eigenvalues of which are all different from 1.

Equation (7) therefore provides the basis for a method for constructing complex unitary n×n code matrices C_k, which can be used as symbol words for space-time block codes for any number of n≧2 transmitter antennas, with the space-time block codes providing a maximum diversity n, as the rank of the complex unitary n×n code matrices C_k is equal to n.

If a group of code matrices is calculated according to the above equation (7), the result is a list of 2^l unitary n×n matrices with complex matrix elements, which can be used as a basis for a non-linear space-time block code, with which a radio transmission link shown diagrammatically in FIG. 2 can be operated in a digital mobile radio network.

In the case of n=3 transmitter antennas and BPSK modulation, i.e., number of transmitter antennas n=length of the symbol words l, for the sake of clarity a specific example of such a list of code matrices C_k is given in the reference table in FIG. 3.

In the reference table shown in FIG. 1 the bits b_iε{0,1} are mapped onto 2³=8 bit vectors. These are specifically the eight possible vectors (0,0,0); (0,0,1); (0,1,1); (0,1,0); (1,1,0); (1,1,1); (1,0,1); (1,0,0) formed by permutation.

These eight bit vectors are each uniquely (reversibly uniquely) associated with one of eight code matrices C_k=C(:;:k). Here the code matrices in table 1 are unitary complex 3×3 code matrices, the matrix elements of which are calculated according to equation (7).

If the association between the eight bit vectors and the eight code matrices C_k, shown as an example in FIG. 3, is used for example with a digital mobile radio system as shown in FIG. 2, and a bit vector (1,0,1) for example arrives there at the ST coder, the coder transmits each column of the code matrix C₇ in three successive time slots, so that the complex matrix element c₁₁₇=0,0000+0.6442i is transmitted via the antenna 1 in the time slot 1, the complex matrix element c₂₁₇=−0.2175−0.0412i via the antenna 2 in the time slot 1, etc. until the complex matrix element c₃₃₇=0.0000−0.50651 is transmitted via the antenna 3 in the time slot 3 (in FIG. 3 i is used for the imaginary unit).

On the receiver side the complete code matrix C₇ is then reconstructed from the signals received by a receiver antenna Rx as described above by an MLD detector and the original bit vector (1,0,1) is associated with this again in a reverse mapping.

For practical purposes only the matrices calculated according to equation (7) should be stored with the respective bit vector association in a reference table in the form of the table shown in FIG. 3 in a storage unit of the ST coder on the transmitter side and correspondingly an identical reference table in a receiver-side ST decoder.

The transmitter-side ST coder can be integrated in a base station of a digital mobile radio network and the receiver-side ST decoder in the mobile station of a digital mobile radio network. In principle this can however be reversed.

The reference tables can be stored as computer program products in machine-readable form, for example on diskette or in the form of machine-readable files, which can be transmitted via the internet or the radio transmission links and if necessary can be input into corresponding storage units of transmitter-side ST coders or receiver-side ST decoders in the base stations or mobile stations in a digital mobile radio network.

Real Symbols

For real symbols the characteristic used above for constructive proof resulting in equation (7), that it is always possible to construct a set of unitary matrices, the eigenvalues of which are all different from 1, no longer applies.

Real ST Codes

If we restrict the ST code matrices to real matrix elements c_ijk, a maximum rank (and therefore a maximum order of diversity of the ST codes) can only be constructed for the case of an even number of antennas.

Theorem 3:

Real ST codes of the order 2n+1 have a non-maximum diversity order.

Proof:Any OεSO(2n+1) can be written as O=VDV⁻¹  (8)

(see Simon, Barry: Representations of Finite and Compact Groups, 1996, Graduate Studies in Mathematics, American Mathematical Society, ISBN 0-8218-0453-7),

where V is orthogonal and

$D=\left(\begin{array}{cccc}\left(\begin{array}{cc}\cos {\Phi }_1& -\sin {\Phi }_1\\ \sin {\Phi }_1& \cos {\Phi }_1\end{array}\right)& & & 0\\ & \left(\begin{array}{cc}\cos {\Phi }_2& -\sin {\Phi }_2\\ \sin {\Phi }_2& \cos {\Phi }_2\end{array}\right)& & \\ & & \cdots & \\ 0& & & 1\end{array}\right)$

(For matrices with det O=−1 the proof is essentially identical).

Let us now consider det(O₁−O₂)=det(1−O₂O_l⁻¹)=det(1−O_2l), where due to the group structure O_2lεSO(2n+1) applies. O_2l then has the structure according to equation (8) and the determinate det(O_l−O₂) disappears.

The reason why SO(2n+1) does not provide a maximum diversity order is based on the fact that every orthogonal matrix with an uneven dimension has (at least) one eigenvalue equal to 1.

For an even number of antennas, the additional 1 does not occur at the position (n,n) of D and a code construction similar to the unitary one is possible.

ST Symbol Optimization

Although the theorem for complex ST modulation is constructive, it does not provide an optimal ST code. The asymptomatic symbol error is of the form

$P_e\sim c{\left(\frac{E_b}{N_0}\right)}^{-n},$

as a result of which an optimum diversity order occurs for

$\frac{E_b}{N_0}.$ However the constant c is not minimal.

Practical methods for code construction based on optimization considerations are therefore set out below. The results are confirmed by simulation.

In the sections below, construction methods are set out for “good” unitary ST codes, i.e. those ST codes with which the intervals between the code symbols are optimized.

Optimization

The idea is to find suitable parameterization for a set of unitary U(n) matrices and then to minimize a suitable metric numerically, the metric representing the intervals between the code words. As the distance measurement D_ev=min{eigenvalues of C({right arrow over (β)}_i)−C{right arrow over (β)}_j} cannot be differentiated, we selectd_ij:=d(C({right arrow over (β)}_i),c({right arrow over (β)}_j))=|det(C({right arrow over (β)}_i)−C({right arrow over (β)}_j))|.

Here {right arrow over (β)}_k stands for the parameter of the kth code matrix C_k. As a target functional

$\begin{array}{cc}{E}_q[{\overrightarrow{\beta }}_1,{\overrightarrow{\beta }}_2,\dots ]={\left(\sum _{i<j}^{2^l}\left[d\left(C\left({\overrightarrow{\beta }}_i\right),C\left({\overrightarrow{\beta }}_j\right)\right)\right]\right)}^{\frac{1}{q}}& \left(6\right)\end{array}$

for global extremal value creation, we can use the L_q standard of all both-way code intervals. For example, q→−∞ gives the minimum standard (a large negative q can be used for numerical optimization); the case where q=−1 can be interpreted as electrical potential. In fact due to the compactness of U(n) the problem is comparable to the minimization of the electrical energy from 2^l equally charged particles moving on a sphere. Positive values for q are not meaningful, as they exclude no distances which could be zero (i.e. two particles, at the same place, do not generate infinite energy, so they would not repel each other).

The Case of the Group SU(2)

This case corresponds to n=2 antennas.

In this case the energy function described above can be specifically constructed, as according to Simon Barry, “Representation of Finite and Compact Groups”, Graduate Studies in Mathematics, Volume 10, American Mathematical Society, every unitary matrix C can be parameterized asC=1β₀+j(β₁σ₁+β₂σ₂+β₃σ₃):={right arrow over (β)}·{right arrow over (σ)}

where {right arrow over (σ)}_i are the known Pauli spin matrices:

${\overrightarrow{\sigma }}_1\left(\begin{array}{cc}0& 1\\ 1& 0\end{array}\right),{\overrightarrow{\sigma }}_2\left(\begin{array}{cc}0& -j\\ j& 0\end{array}\right),{\overrightarrow{\sigma }}_3\left(\begin{array}{cc}1& 0\\ 0& -1\end{array}\right),$ and here j is the imaginary unit.

The real parameters β here are subject to the restriction

$\sum _{i=0}^3{\beta }_i^2=1.$

In fact this ensures that SU(2) is isomorphous in relation to a 3-sphere (a sphere with four dimensions). It can easily be shown that det(C({right arrow over (β)}_i)−C({right arrow over (β)}_j))=2−2{right arrow over (β)}_i·{right arrow over (β)}_j, therefore we defined_ij=√{square root over (1−{right arrow over (β)}_i·{right arrow over (β)}_j)}

as the measure of distance between two code matrices. Note that d_ij is actually a metric in the case of SU(2).

We use d=Σ_i<jd_ij as the full interval. A 3-sphere is parameterized simply by three angles:

${\overrightarrow{\beta }}_i=\left(\begin{array}{ccc}\sin {\Phi }_{\mathrm{i3}}& \sin {\Phi }_{\mathrm{i2}}& \sin {\Phi }_{\mathrm{il}}\\ \cos {\Phi }_{\mathrm{i3}}& \sin {\Phi }_{\mathrm{i2}}& \sin {\Phi }_{\mathrm{il}}\\ & \cos {\Phi }_{\mathrm{i2}}& \sin {\Phi }_{\mathrm{il}}\\ & & \cos {\Phi }_{\mathrm{il}}\end{array}\right)$

The gradient here is:

$\frac{\partial }{\partial {\Phi }_{\mathrm{kl}}}d=-\left(\frac{\partial }{\partial {\Phi }_{\mathrm{kl}}}{\overrightarrow{\beta }}_k\right)·\sum _{i\ne k}^{}\frac{{\overrightarrow{\beta }}_i}{d_{\mathrm{ik}}}$

Optimization methods based on that of the steepest descent provide fast results for a reasonable l.

The resulting code spectrum for QPSK (l=4) is compared in FIG. 5 with the code spectrum for the Alamouti scheme. As can be seen, the minimum interval with this procedure is greater than with the Alamouti scheme. Therefore in the asymptomatic extreme case it should be anticipated that this non-linear code will show a higher coding gain (same bit error rate for lower signal to noise ratio).

SU(3) and an Implicit Numerical Gradient Method

In cases with more than two transmitter antennas, i.e. SU(n), n>2, the specific calculation of determinants becomes very extensive. For gradient methods (e.g. the conjugate gradient method, see W. Press, B. Flannery, S. Tenolsky, W. Vetterling: “Numerical recipes in C”, Cambridge University Press, ISBN 0-521-35465-X) is it however sufficient to calculate the local gradient.

We define the L_m interval of the ith matrix in respect of all others as:

$\begin{array}{cc}{d}_i^m=\sum _j^{}{\det \left(C_j-C_i\right)}^m=\sum _j^{}{\det \left(1-C_j^{†}{C}_i\right)}^m:=\sum _j^{}{\det \left(1-A_{\mathrm{ij}}\right)}^m& \left(9\right)\end{array}$

If the ith code matrix is varied by an infinitesimal (unitary) rotationC_i→C_iexp(j{right arrow over (σ)}·{right arrow over (δ)}_i)≈C_i+j{right arrow over (δ)}_i·C_i{right arrow over (σ)}

the following results for the gradient

$\begin{array}{cc}{\overrightarrow{\nabla }}_{{\delta }_i}{d}_i^m:={\overrightarrow{g}}_i=n\sum _{j\ne i}^{}{d_{\mathrm{ij}}}^n\mathrm{Re}\mathrm{Tr}\left(A_{\mathrm{ij}}^{-1}{\overrightarrow{B}}_{\mathrm{ij}}\right)& \left(10\right)\end{array}$

where:{right arrow over (B)}_ij=−jC_j^†C_i{right arrow over (σ)}.

The σ_i here are the corresponding Hermitian standard spin matrices. For larger-dimensional spaces (n>2) these are specified for example in the book “Gauge theory of elementary particle physics” by Ta-Pei Cheng and Ling-Fong Li.

A variation with an increment δ is then applied to the ith code word according toC_i→C_iexp(jδ{right arrow over (σ)}·{right arrow over (g)}_i)  (11)

An algorithm with the steepest descent then functions as follows:

Generate a random quantity of 2^l unitary n×n matrices S_k, k=1 . . . 2^l as initial matrices (this can also be done according to equation (7)).

Calculate the gradient vectors according to equation (10).

“Rotate” the matrices according to the equation (11), then iterate according to stage 2).

Naturally a conjugate gradient method can be constructed correspondingly and stochastic gradient methods are also possible to find the global extreme value creator.

An example of 3 antennas and a QPSK modulation, which provides 2⁶=64 ST matrices, is shown in FIG. 6. FIG. 6 shows a spectrum for SU(3), i.e. three transmitter antennas, and QPSK modulation using a minimum standard. If the same method is used for SU(3) but for a BPSK modulation and using different standards (min, L₋₁, L₋₂), the spectra shown in FIG. 7 are obtained.

It seems to be worthy of note that it is possible to find eight code matrices in SU(3), which all show the same mutual interval. This corresponds to a tetrahedron in the standard three-dimensional space.

FIG. 4 shows a two-dimensional illustration of this. The eight bit vectors (0,0,0); (0,0,1); (0,1,1); (0,1,0); (1,1,0); (1,1,1); (1,0,1); (1,0,0) are mapped onto eight code matrices (e.g. the code matrices specifically set out in FIG. 3), which have optimum intervals in respect of each other.

FIG. 8 shows a spectrum for n=3 antennas and BPSK using different optimization criteria.

In addition to the numerical optimization method described above, there are further approaches for code optimization, i.e. the specific construction of ST symbols in the form of unitary n×n matrices with optimized intervals in respect of each other:

Construction of Hyperspheres:

Let a (hyper) sphere with a defined radius be constructed around a defined code symbol (e.g., a unitary n×n matrix C_k), a second code symbol be found on the sphere around this first code symbol and a third symbol be constructed as the point of intersection of the (hyper) spheres and the first and second code symbol. Then let further code symbols be constructed correspondingly and iteratively as the points of intersection of further (hyper) spheres around the code words already found in each instance.

In the event that the code words are unitary n×n matrices with n≧2, a “sphere” with radius r around a code word is defined by S_r{C′|det(C′−C)=r}.

It can be constructed by

$r=\det \left(C-C^{\prime }\right)=\det \left(1-C_1^{†}{C}^{\prime }\right)=\det \left(1-\exp \left(j\overrightarrow{\sigma }\overrightarrow{\beta }\right)\right)=\prod _{i=1}^n\left(1-\exp \left({\lambda }_i\right)\right),$

where λ_l are the eigenvalues of j{right arrow over (σ)}{right arrow over (β)},

For example it is found in SU(2) thatr=4 sin²(1/2√{square root over (β₁²+β₂²+β₃²)}).

This means that such a sphere can be parameterized as C′=Cexp(j{right arrow over (σ)}{right arrow over (β)}) with restriction of the sum of the squared βs.

A sphere with radius 1 is for example defined byS₁(C)={Cexp(j{right arrow over (σ)}{right arrow over (β)})|β₁²+β₁²+β₁²=π²/9,−π/3≦β_i≦π/3}.

Using this idea, it is actually possible to construct the Alamouti scheme (by constructing two spheres with the radius 1 around the elements 1 and −1 (the center of SU(2), . . . Z(SU(2)).

This gives the following formula for the code words:

$C=\exp \left(j\frac{\pi }{3\sqrt{3}}\left(1+b_0\right)\sum _{i=1}^3\left(1-2b_i\right){\sigma }_i\right),$

where b_iε{0,1) are the bits and σ_l the Hermitian standard spin matrices.

Specific calculation leads back to the already known Alamouti scheme, thereby proving the correctness of the approach of construction hyperspheres.

As the example of the Alamouti scheme in conjunction with FIG. 5 shows, a local optimum can be found using the hypersphere method but not necessarily a global optimum.

Nevertheless, for n>2 the eigenvalues of j{right arrow over (σ)}{right arrow over (β)} are (analytically) very extensive.

Therefore a different parameterization seems more successful here for n>2:r=det(1−C^†C′)=det(1−VDV^†)=det(1−D)=Π_i(1−exp(jλ_i))

If the eigenvalues λ_i are kept constant, the result is a sphere around the code symbol C withC′=CVDV^†.

V can be parameterized here as V=exp(j{right arrow over (ρ)}{right arrow over (β)}).

Unitary Representations of Finite Groups

A further method of code construction is based on the use of finite groups. For this, see also Simon Barry, “Representation of Finite and Compact Groups”, Graduate Studies in Mathematics, Volume 10, American Mathematical Society.

The combination of two elements in a finite group leads back to one element of the group, as the group is closed after a group axiom. Also the multiplication of two unitary matrices again provides a unitary matrix. There are correspondingly numerous representations of finite groups, in which a unitary matrix is associated with each group element. If such a representation of a finite group, in which the number of group elements is greater than 2^l, is selected, good initial values are obtained for the optimization methods referred to above.

For this, it is necessary to try to find the unitary representations of finite groups (with the dimension n), in which o(G)≧2^l. There is no guarantee that this produces optimum results.

Some Simulation Results:

Simulations are carried out for the instances with n=2 and n=3 antennas, in which only BPSK has (currently) been used. Theoretical limits for antenna diversity can be derived in closed form (see also for example J. Proakis, M. Salehi: “Communications Systems Engineering”, Prentice Hall Int., ISBN 0-13-300625-5, 1994). These can be seen as continuous lines in FIG. 9. Simulation results are shown together with 70% confidence intervals.

As shown in FIG. 10, the theoretical limit for high E_b/N₀ in the case of three antennas is reached for the L_min code. This is due to the fact that L_min is actually the maximum possible smallest interval.

For a good signal to noise ratio in practice only the errors at the smallest intervals make a contribution.

However with a low signal to noise ratio the characteristics of the code deteriorate and they can even be worse than with a diversity of two antennas. In fact ST codes can be seen as a higher modulation scheme (resulting in expansion of the symbol space). It is of course not possible to increase the number of symbols without reducing the intervals in a compact space. This problem becomes even more significant for QPSK (with 64 ST symbols).

For the L₁ code performance is better in the areas with a low signal to noise ratio. Gray coding was used for this code (which has two different intervals in the minimum eigenvalue standard), in order to avoid multiple bit errors in the event of a symbol error. As can be seen in the spectrum below, each code matrix has precisely two immediate neighbors. All other code words have a larger interval. As can be seen, the code is close to the theoretical limit for E_b/N₀>4 dB and it outperforms the L_min code in the area with low E_b/N₀.

The invention has been described in detail with particular reference to preferred embodiments thereof and examples, but it will be understood that variations and modifications can be effected within the spirit and scope of the invention.

Claims

1. A method for operating a digital mobile radio network with orthogonally structured space-time block transmission codes with maximum diversity n×m for a transmitting station having n transmitter antennas and a receiving station having m receiver antennas, where m and n are each an integer greater than or equal to 2, comprising: transmitting a set of 2l data bit vectors {right arrow over (b)}=(b1, b2, . . . bl)εBl with l bits biε{0,1} from the transmitting station, the data bit vectors {right arrow over (b)} being mapped by a one-to-one mapping

2. The method according to claim 1, wherein n is even and cijk are real.

3. A method for operating a digital mobile radio network with orthogonally structured space-time block transmission codes with maximum diversity n×m for a transmitting station having n transmitter antennas and a receiving station having m receiver antennas, where m and n are each an integer greater than or equal to 2, comprising: transmitting a set of 2l data bit vectors b=(b1, b2, . . . bl)εBl with l bits biε{0,1} from the transmitting station, the data bit vectors b being mapped by a one-to-one mapping

4. The method according to claim 3, wherein a) n=2,b) every matrix of the initial set of 2′ unitary n×n initial matrices Sk is parameterized as S=1β0+i(β1σ1+β2σ2+β3σ3)={right arrow over (β)}·σ

5. The method according to claim 3, wherein an Lm interval of the ith initial matrix in respect of all others is defined as:

6. The method according to claim 3, wherein the initial matrices Sk are calculated according to the following specification:

7. The method according to claim 3, wherein a finite group G with dimension n and order o(G)≧2l is mapped onto unitary n×n matrices, which are used as initial matrices Sk.

8. A method for operating a digital mobile radio network with orthogonally structured space-time block transmission codes with maximum diversity n×m for a transmitting station having n transmitter antennas and a receiving station having m receiver antennas, where m and n are each an integer greater than or equal to 2, comprising: transmitting a set of 2l data bit vectors b=(b1, b2, . . . bl)εBl with l bits biε{0,1} from the transmitting station, the data bit vectors b being mapped by a one-to-one mapping

9. A base station, comprising: a storage unit, containing an association table, wherein the association stores an association of individual bit vectors with space-time code symbols, andthe association table is in the form of a matrix that contains the matrix elements cijk association of individual bit vectors with space-time symbols association matrix elements cijk used in one of the methods according to any claims 1-8.

10. A mobile station, comprising: a storage unit, containing an association table, wherein the association stores an association of individual bit vectors with space-time code symbols, andthe association table is in the form of a matrix that contains the matrix elements cijk association of individual bit vectors with space-time symbols association matrix elements cijk used in one of the methods according to any claims 1-8.

11. A computer readable storage medium storing a program to control a processor to perform a method for operating a digital mobile radio network with orthogonally structured space-time block transmission codes with maximum diversity n×m for a transmitting station having n transmitter antennas and a receiving station having m receiver antennas, where m and n are each an integer greater than or equal to 2, comprising: transmitting a set of 2 data bit vectors {right arrow over (b)}=(b1, b2, . . . bl) E Bl with l bits biε{0,1} from the transmitting station, the data bit vectors b being mapped by a one-to-one mapping