DE102008057059B3 - Receipt device for receiving transmitted signals from overlay impaired, channel-coded transmitted signals, has branch of receipt for receiving overlay of impaired, channel-coded transmitted signals - Google Patents

Abstract

The receipt device has a branch of receipt for receiving overlay of impaired, channel-coded transmitted signals and a linear detector unit coupling to branch of receipt, whose entrance is coupled laterally with the branch of receipt. The detector output signal has probability information over potentially sent transmission symbols of the transmitted signals. An independent claim is also included for a method for receiving transmitted signals.

Description

The The present invention relates to devices and methods for receiving an interference-disturbed Signals, such as in a multi-user communication system or to receive one over a multipath channel transmitted Signals can be used.

In a multiuser communication system are used by a plurality by users independently simultaneously broadcasting broadcast signals. In the transmission of the transmission signals can Problems such as multipath propagation, co-channel interference, Signal fading and scattering occur, which is the reception of the signals difficult. Communicate with multi-user communication system the users in the same frequency band and within the same time span, resulting in some significant interference for a single user. By estimates for parameters, such as B. Estimates for channel coefficients and intoxication, can Receive signals receiver side be processed to estimates for the to receive transmitted signals. This is used in modern day Mobile communication systems by the recipient increases several receiving antennas, about space diversity exploit what leads to so-called MIMO transmission channels (MIMO = Multiple-Input Multiple Output).

interference can also arise with a single user, that over a Multipath channel transmitted will, so that at the receiver Overlay signal echoes. One speaks here of so-called Intersymbolinterferenzen. In this case the signal must be rectified on the receiving side, d. H. the influences of Channels need preferably reversed become.

The Principles for receiving signal detection in multi-user communication systems and / or used in intersymbol interference-prone received signals will resemble mostly very. Therefore, you can embodiments the invention described below for both multi-user detection as well as for single-user detection in multi-path channel transmission be used. Nevertheless, the following is an example in most cases talk about multi-user systems, but this is not restrictive should be.

in the State of the art are a variety of multi-user detectors (MUD = Multi-User Detector) known. An optimal multi-user detector based on the so-called maximum likelihood sequence estimator (MLSE) and compares one or more receive signals with all options for the corresponding transmission signals - and though for every send symbol. The possible Signal paths can be represented in a so-called trellis diagram. From an optimal MLSE, however, results in a computational effort, the currently for Real-time signal processing for many users and / or large symbol alphabets is not feasible.

Of Furthermore, there are some suboptimal multi-user detectors that use less Need computational effort. Examples of this are so-called linear detectors, such as decorrelators, MMSE detectors (MMSE = minimum-mean-squared-error) and ZF detectors (ZF = zero-forcing). However, such linear detectors exhibit higher signal-to-noise power ratios and with an increasing number of users compared to one optimal MLSE multi-user detector significantly worse performance.

One another approach for low-cost multi-user detection is the use of suboptimal MLSE detectors that have paths dropped with a bad metric and thus in trellis not be followed up. Examples of such algorithms are the known M and T algorithms. In the M algorithm, only the best metric paths are traced. In the T-algorithm only the paths are pursued, the do not exceed a certain metric threshold.

at a so-called turbo receiver becomes the entire transmission system considered as a serial concatenation of two convolutional codes - the actual one Error correction code (channel code), which is used on the transmitter side, and the effects of modulation and multiuser interference (and / or Inter-symbol interference). A multi-user detector as part of a turbo-receiver leads first one Interference reduction by and determines estimates of the transmitted transmit symbols for each Users. These symbol estimates then, for each user, supplied to a channel decoder, for estimating the transmitter-side channel decoder input bits, d. H. the actual Information bits to determine.

The turbo receivers approach the APP (A Posteriori Probability) optimal receiver by iteratively performing MIMO and / or multi-user detection and channel decoding (eg Viterbi decoder, turbo decoder, etc.), taking into account a priori information of the other block. For this purpose, turbo receivers are conventionally proposed both with an MMSE detector and with a suboptimal tree search (eg QRD-M algorithm) as MIMO and / or multi-user detector. MMSE has the disadvantage over the approach with tree search that the iteratively fed back a priori information of the channel decoder is used only very inadequate. On the other hand, non-linear MIMO detectors (such as tree search) have a significantly higher computational cost, which is usually a multiple of that of MMSE.

The publication WO 2007/077299 A1 describes a turbo equalizer with an inner loop for an equalization process and an outer loop for a decoding process. The equalization process provides an equalized received signal and additional soft information for the subsequent decoding process. The decoding may be based on a max log MAP (maximum a posteriori probability) decoder.

The publication WO 2006/052156 A1 deals with a feedback MIMO equalizer decoder. In this case, a received MIMO signal is at least partially equalized by a multi-dimensional forward filter. Remaining so-called post-cursor intersymbol interference cancellation and trellis decoding can be done subsequently.

In D. Karakolah et al .: "Architecture dedicated to the MMSE equalizer of iterative receiver for linearly precoded MIMO systems ", becomes an MMSE equalizer of an iterative receiver for linear precoded MIMO systems described. The receiver structure shown is divided into two main elements: a MIMO equalizer and a Channel decoder. The iterative process is based on an exchange soft information between these two elements.

In front This is the background of the object of the present invention in it, a turbo receiver to provide the bit error rates of conventional turbo receivers In contrast, reduced computational effort or improved at the same computational cost Can provide bit error rates.

These The object is achieved by a receiving device according to claim 1 and a receiving method according to claim 11 solved.

The Recognition of the present invention is a linear Detector and a nonlinear detector in a hybrid turbo receiver concept combine to exploit the respective advantages of both detectors, and thus to allow a reduction of the computational effort - at the same Bit error rates as conventional Turbo receiver.

To takes place in a first detector stage, if no a priori Probability information is available, a linear detection, at one or more received signals a first interference reduction perform. One or more output signals of the linear detector are subsequently iterated each postprocessed by one channel decoder to one or more Channel decoder output signals with probability information about the to receive potentially transmitted transmission signals. At the first detection pass becomes a good interference reduction by using the linear detector achieved with minimal computational effort. Once after one or more Channel decoder runs extrinsic information is present, a renewed, but this time Non-linear detection for further interference reduction. By the quality of extrinsic information achieved by the channel decoder can the non-linear detection with relatively little Re chenaufwand respectively. Becomes a non-linear detector a suboptimal MLSE algorithm, such as As the M-algorithm, so can by the inventive concept the number M of the traced metrics best paths are reduced. In a so-called sphere decoder as a nonlinear detector this corresponds to a smaller search radius. Overall complexity and reception quality of the receiver concept according to the invention can be, for example, by the number of channel decoder and nonlinear detector iterations and parameter selection of these signal processing blocks themselves adjust the SNR requirements.

Exemplary embodiments of the present invention provide a receiving device for receiving a channel-coded received signal, having a receiving branch on which a superimposition of a plurality of received signals can be received, and a linear detector device, which is coupled on the input side to the receiving branch, in a first detection step, on the output side, by a linear detection method to provide a first detector output, wherein the first detector output includes first likelihood information about potentially transmitted transmit symbols. Furthermore, the receiving device comprises a channel decoder device for decoding the first detector output signal and for providing a channel decoder output signal including second probability information about the potentially transmitted transmit symbols that is more reliable than the first probability information. A non-linear detector means is coupled on the input side to the receiving branch and the channel decoder output signal for outputting, by a non-linear detection method, a second detector output signal comprising third probability information about the potentially transmitted transmission symbols, which is more reliable than the second probability information. The second detector output signal is fed back to an input of the channel decoder means to replace the first detector output signal in a first subsequent detection step.

According to one preferred embodiment According to the present invention, the linear detector device is a linear MIMO detector device, the input side with a plurality is coupled by receiving branches, wherein at each of the receiving branches an overlay a plurality of received signals can be received. On the output side provides the linear MIMO detector device, by a linear MIMO detection method a plurality of first detector output signals, a plurality of channel encoded transmit signals, each of the first detector output signals a first probability information about potentially transmitted transmission symbols includes. Preferably, the linear MIMO detector means a linear MMSE detector on to separate the plurality of received signals, so that each of the plurality of first detector output signals is a transmit signal is assignable. In this case, the linear MIMO detector device is adapted, around the first detector output signals in the form of so-called log likelihood quotients issue.

The Channel decoder means is adapted to accept the plurality of first ones To decode detector output signals and a plurality of channel decoder outputs wherein each of the channel decoder outputs is a second one Probability information about includes the potentially transmitted transmit symbols, which is more reliable as the first probability information. In this case, the channel decoder device adapted to a transmitter-side channel encoder. For this purpose, the channel decoder device a block decoder for Block codes, a convolutional decoder for convolutional codes or a Turbo decoder for serially or parallel concatenated convolutional codes. On the input side can before the channel decoder Entschachteler (deinterleaver) and on the output side interleavers can be used to create a transmitter-side interleaving process to meet.

According to one preferred embodiment The present invention is the non-linear detector device a non-linear MIMO detector device, the input side with the plurality of receive branches and the channel decoder output signals is coupled. On the output side, the non-linear MIMO detector device provides by a non-linear MIMO detection method, a plurality of second detector output signals, the majority of channel-coded Transmit signals correspond to each of the second detector output signals comprises third probability information about potentially transmitted transmission symbols, the more reliable is the first and second probability information. The non-linear MIMO detector device comprises a trellis-based MIMO detector to separate the majority of the received signals, such that each of the plurality of second detector output signals corresponds to a Transmission signal can be assigned. For a suboptimal use For example, the MLSE algorithm may be variants of the T algorithm or the M-algorithm.

at the channel decoder output signals and the detector output signals it is according to embodiments the present invention to so-called soft symbols, in the form of log-likelihood ratios as extrinsic information become.

As already mentioned at the beginning In the case of a receiver according to the invention, it may also be a SISO turbo receiver (SISO Single Input Single Output) to handle intersymbol interference Receive received signals. In this case one speaks with the linear and nonlinear detectors also of linear and nonlinear Equalizers. An inventive SISO turbo receiver could, for example in existing GSM / EDGE (GSM = Global System for Mobile Communications, EDGE = Enhanced Data Rates for GSM Evolution) Mobile Radio Systems be used.

The reception quality from conventional Turbo receivers let yourself by using a first linear receiver stage before the feedback loop channel decoder and non-linear detector with significantly reduced Achieve computational effort. The combination possibilities of the different ones Signal processing blocks enable a scalable adaptation of a required computing power the respective reception requirements of a receiver or communication standard.

The inventive receiver concept is especially for modern software-defined radio implementations interesting because the different algorithms are not available as dedicated hardware have to, (then would possibly more chip area consumed), but on processors successively different Code executed can be. Necessary is then only a minimally larger code memory.

preferred embodiments The present invention will be described below with reference to FIG the enclosed drawings closer explained. Show it:

1 a flowchart of a method for receiving a plurality of received signals according to an embodiment of the present invention;

2 a block diagram of a receiving device according to an embodiment of the present invention;

3 a block diagram of a linear MIMO detector device according to an embodiment of the present invention;

4 a block diagram of a channel decoder according to an embodiment of the present invention;

5 a block diagram of a non-linear MIMO detection method according to an embodiment of the present invention;

6 a block diagram of an OFDM receiver according to an embodiment of the present invention;

7 a representation of trans information plotted over SNR for various MIMO detectors;

8th EXIT chart showing an iterative reception improvement by a turbo decoder;

9a , b representations of the utilization of the a priori information of the channel decoder by the M-algorithm;

10 a representation of computational effort for different receiver algorithms plotted over target SNR.

wobei c ein Informationsbit bedeutet, das die Werte ±1 annehmen kann. Der inverse Zusammenhang ist:

A channel coded receiver requires knowledge of the channel matrix between the transmit and receive antennas, as well as allowed codewords of the channel code used. The transmission over the transmission channel is accomplished by means of complex modulation symbols, whereas the channel code is applied to information bits before the modulation. The combination of channel state information and allowed codewords uses log likelihood ratios or ratios (LLRs):

where c is an information bit that can take the values ± 1. The inverse context is:

One Detector generates LLRs from a receive symbol vector y and a Channel matrix H, for example by means of known pilot symbols estimated can be. A subsequent soft-in / soft-out channel decoder improved the LLRs by taking advantage of the known channel code structure. Finally, the transmitted information bits are estimated by a binary quantization of the LLRs, only the sign of the LLR is considered.

The transmission over the (MIMO) channel H can be as: y = H * x (c) + n, (3) be modeled. For example, Eq. (3) as a subcarrier model. MIMO-OFDM transmission. According to J. Hagenauer, "The Turbo Principle in Mobile Communications," in International Symposium on Information Theory and Its Applications, 2002, an MLSE detector maximizes

wobei L_E extrinsische Information vom Kanaldecoder bedeutet. Diese Suche über sämtliche möglichen Sendesymbolkombi nationen ist jedoch viel zu komplex für derzeitige Modulationsalphabete und Antennenzahlen (64QAM mit vier Antennen bedeutet ca. 17·10⁶ mögliche Kandidaten).Assuming additive white Gaussian noise, Eq. (4) be written as:

where L _{E means} extrinsic information from the channel decoder. However, this search across all possible transmitter symbol combinations is far too complex for current modulation alphabets and antenna numbers (64QAM with four antennas means about 17 × 10 ⁶ possible candidates).

One practically realizable receiver So it should be a suboptimal estimate which could then be improved by iteration as soon as possible better extrinsic information of a channel decoder is available. A receiver, such an iterative combination of detection and decoding is generally referred to as a so-called turbo receiver.

The Turbo receiving concept according to the invention will now be detailed in the following with reference to the accompanying figures explained become. This is exemplified by a multi-user MIMO scenario gone out, d. H. several independent Data streams, the above several antennas are sent and received. It becomes one Those skilled in the art should be aware that the described concept is equally applicable to single-user SISO scenarios can be applied with ISI.

1 shows a schematic flow diagram of a method 100 for receiving a plurality N of received signals corresponding to N data streams. In this case, the plurality N may, for example, correspond to a number of users to be detected or to a number of subcarriers in an Orthogonal Frequency Division Multiplex (OFDM) system.

In a first step 102 the plurality N of the received signals are received with a plurality R of receiving antennas or receiving branches, wherein at each of the receiving branches an overlay of the plurality N of the received signals is received. Ie. the N receive signals interfere. The plurality N of the received signals corresponds to a plurality of N transmit data streams which are transmitted via a MIMO channel to a receiver according to the invention.

In a second step 104 For example, based on the R receive signals, N first detector output signals are detected by a linear detection method, the N first detector output signals corresponding to the N channel coded transmit signals, and wherein each of the first detector output signals comprises first probability information about potentially transmitted transmit symbols. In other words, this means that in the step 104 from the R receive signals N first detector output signals are determined, which can be assigned to the N transmit signals. That is, by the linear detection method, based on the R reception signals, the N individual data streams are separated. In this case, R <N, R = N or R> N. In an extreme case, R = 1 and N = 1, which then corresponds to only one receiving antenna and one data stream.

In a further step 106 For example, the N first detector output signals are channel decoded to take advantage of redundancy information of a transmission side channel code. By channel decoding, N channel decoder output signals are provided, wherein each of the N channel decoder output signals comprises second probability information on the respective potentially transmitted transmit symbols, which by knowing the channel code is more reliable than the first probability information of the first detector output signals.

In one step 108 Now, a nonlinear detection method is used to obtain a plurality N of second detector output signals based on the R reception signals and the N channel decoder output signals. Each of the N second detector output signals has third probability information about the potentially transmitted transmit symbols, which should normally be more reliable than the second (and first) probability information.

In a feedback step 110 For example, the N second detector output signals are input to the channel decode step 106 ie, the second detector output signals (extrinsic information) are used as a priori probability information for decoding.

Here are the steps 106 . 108 and 110 in each case a feedback loop, which can now be run through so often until no significant improvement in the reliability of the second and / or third probability information can be achieved.

A block diagram of a receiving device according to the invention 200 to carry out the process 100 according to 1 is in 2 shown.

wobei h_nr den Kanalkoeffizienten des Übertragungskanals von der n-ten Datenquelle zum r-ten Empfangszweig, x_n das Sendesymbol der n-ten Datenquelle und n_r additives weißes Gauss'sches Rauschen am r-ten Empfangszweig bedeutet.The receiving device 200 has R reception branches 202-r (r = 1, 2, ..., R), wherein at each of the reception branches 202-r (R = 1, 2, ..., R) is a superposition of N by the transmission impaired data streams receivable. That is, the rth receive signal y _r can be represented without considering the (discrete) time

where h _nr denotes the channel coefficient of the transmission channel from the nth data source to the rth reception branch, x _n the transmission symbol of the n th data source, and n _r additive white Gaussian noise at the r th reception branch.

The R received signals y _r and a resulting received signal _vector y = [y ₁ , y ₂ ,..., Y _R ] ^T become a linear MIMO detector device 204 fed, the input side with the plurality of R reception branches 202-r is coupled. In the linear MIMO detector device 204 is applied to the R received signals, in a first Detekti onsschritt, a linear detection method to the output side to obtain N first detector output signals. The N first detector output signals correspond to first estimates for the N channel coded transmission signals x _n .

bestimmt werden, wobei x ^_n (n = 1, 2, ..., N) einen harten Schätzwert für x_n bedeutet und

die Varianz der bedingten Ausgangsverteilung ist.Each of the first N detector output signals comprises first probability information l _n ¹ (n = 1, 2, ..., N) over the potentially transmitted transmission symbols x _n . This first probability information can, for example, according to

be determined, where x ^ _n (n = 1, 2, ..., N) means a hard estimate for x _n and

is the variance of the conditional output distribution.

The first detector output signals in the form of the first probability information l _n ¹ (n = 1, 2,..., N) become a channel decoder device 208 supplied as a priori probability information. Based on this, the channel decoder provides 208 output N channel decoder output signals, each of the channel decoder output signals comprising second probability information l _n ² (n = 1, 2, ..., N) over the potentially transmitted transmit symbols x _n , which is more reliable than the first likelihood information l _n ¹ (n = 1) , 2, ..., N). That is, the channel decoder device 208 In one embodiment, N comprises soft-in / soft-out channel decoders 208-n (n = 1, 2, ..., N), based on that of the linear MIMO detector device 204 delivered N first probability information l _n ¹ (n = 1, 2, ..., N) on the output side estimates for the transmission symbols x _n in the form of log likelihood ratios l _n ² (n = 1, 2, ..., N ).

The channel decoder algorithm of the channel decoder device 208 is adapted to a channel coder used on the transmitting side and can be selected, for example, from the group comprising a maximum a posteriori (MAP) decoder, soft output Viterbi algorithm (SOVA) decoder, and Bahl, Cocke, Jelinek, Raviv (BCJR) De be selected.

The of the channel decoder device 208 provided soft output symbols l _n ² (n = 1, 2, ..., N) are a non-linear MIMO detector device 212 supplied as a priori information. Furthermore, the non-linear MIMO detector device 212 with the majority of reception branches 202-r (n = 1, 2, ..., R), so that the R received signals y = [y ₁ , y ₂ ,..., y _R ] ^T can be supplied to it. Based on the a priori information l _n ² (n = 1, 2, ..., N) and the received signal y = [y ₁ , y ₂ , ..., y _R ] ^T represents the non-linear MIMO detector device 212 by a non-linear detection method on the output side N second detector output signals, each of which includes a third probability information l _n ³ (n = 1, 2, ..., N) on the potentially transmitted transmission symbols x _n . Due to the utilization of the second probability information l _n ² (n = 1, 2, ..., N) as a priori information, the third likelihood information l _n ³ (n = 1, 2, ..., N) should be more reliable than the second (and the first) probability information. The third probability information l _n ³ (n = 1, 2,..., N) is provided according to exemplary embodiments in the form of log likelihood ratios and can, according to Eq. (5) (see above), where L _{E is to be} replaced by l _n ² (n = 1, 2, ..., N).

For a second detection step, the second detector output signals in the form of the third probability information l _n ³ (n = 1, 2,..., N) are respectively applied to the N inputs of the channel decoder device 208 fed back as a priori information. There are switches 214-n opened to the first detector output signals l _n ¹ (n = 1, 2, ..., N) of the linear detector device 204 from the inputs of the channel decoder device 208 decouple. Thus, for the second detection step, the first quasi Detektoraus output signals l _n ¹ (n = 1, 2, ..., N) by the second detector output signals I _n ³ (n = 1, 2, ..., N) is replaced.

The linear MIMO detector device 204 could work, for example, according to the IF principle or preferably according to the MMSE principle.

In the IF principle, the transmission symbol vector x = [x ₁ , x ₂ ,..., X _N ] ^T becomes x ^ = (H H H) -1 H H · Y (7) (H ^H H) ^-1 H ^{H is} the Moore Penrose inverse of the channel matrix H.

geschätzt, wobei noch die Rauschvarianz σ² des Empfangssignals berücksichtigt wird.In the MMSE principle, the transmission symbol vector x = [x ₁ , x ₂ ,..., X _N ] ^{T is determined} according to

estimated, wherein the noise variance σ ^{2 of} the received signal is taken into account.

Based on 3 is an example of the linear MMSE detection method, which according to an embodiment in the linear detection device 204 expires, be illustrated.

berechnet werden (302). Die daraus resultierende Matrix wird nun mit dem Empfangsvektor y multipliziert (304), um den geschätzten Sendesymbolvektor x ^ zu erhalten. Die geschätzten Sendesymbole x ^_n (n = 1, 2, ..., N) können nun mittels einer Soft-Demodulation (306), beispielsweise gemäß Gl. (7), in erste Wahrscheinlichkeitsinformationen l_n ¹ (n = 1, 2, ..., N) umgerechnet werden, die der nachfolgenden Kanaldecodereinrichtung 208 als a priori Informationen zugeführt werden.Based on the receive vector y and known pilot symbols, first a channel estimation and a noise power estimation can be performed ( 300 ), so that estimates for H and σ ² are obtained. Such estimation methods are known in the art. Based on the estimated channel matrix H and the noise power σ ² , the expression

be calculated ( 302 ). The resulting matrix is then multiplied by the receive vector y ( 304 ) to obtain the estimated transmit symbol vector x ^. The estimated transmission symbols x ^ _n (n = 1, 2,..., N) can now be determined by means of a soft demodulation (FIG. 306 ), for example according to Eq. (7), in first probability information l _n ¹ (n = 1, 2, ..., N) are converted, the subsequent channel decoder device 208 as a priori information is supplied.

Referring to 4 Let's look at an example of a channel decoder 208-n (n = 1, 2, ..., N) of the channel decoder device 208 be explained. 4 shows a block diagram of a so-called turbo decoder 208-n , provided that a corresponding turbo encoder was used for channel coding on the transmitter side.

The turbo decoder 208-n includes a first and a second soft-in / soft-out convolutional decoder 402-1 . 402-2 , a turbo interleaver 404-1 and a turbo-deinterleaver (de-interleaver) 404-2 , On the input side are estimates for transmit symbols x ^ _n (n = 1, 2, ..., N) and associated first parity information 406-1 (corresponding to a first encoder) provided as soft information. The two convolutional decoders 402-1 . 402-2 can z. B. MAP symbol estimator. The first convolutional decoder 402-1 calculates first logarithm reliability data LLR1 in the form of log likelihood ratios based on the estimated transmission symbols in x ^ _n (n = 1, 2, ..., N), the first parity information and a feedback signal LLR2.

The first logarithmic reliability data LLR1, which also includes the systematic data of the transmission signal, is provided by the turbo interleaver 404-1 nested and the second convolutional decoder 402-2 fed. The second convolutional decoder 402-2 calculated from the interleaved first reliability data LLR1, from the estimated interleaved transmission symbols x ^ _n (n = 1, 2, ..., N) and second parity information 406-2 (corresponding to a second encoder) second logarithmic reliability data LLR2 and deinterleaves these second logarithmic reliability data LLR2 with the deinterleaver 404-2 to iteratively the first channel decoder 402-1 as extrinsic information. The recursion loop shown can be traversed several times until finally the second channel decoder 402-2 delivered reliability data LLR2 by the deinterleaver 402-2 are deinterleaved to obtain the channel decoder output signals l _n ² (n = 1, 2, ..., N) in the form of LLRs.

At this point it should be emphasized again that the in 4 illustrated turbo decoder is meant only by way of example. Another soft-in / soft-out channel decoder could also be used, depending on the transmitter-side channel encoder.

An embodiment of the non-linear MIMO detection method in the non-linear MIMO detector 212 will be explained below by means of 5 be explained. In addition shows 5 a schematic representation of a sequence of the so-called SQRD-M algorithm (Sorted QR Decomposition and M-algorithm), which essentially Eq. 5 triggers.

In order to enable detection with a tree search, the estimated channel matrix H is first QR-decomposed ( 502 ) to obtain a unitary matrix Q and an upper triangular matrix R. Before the QR decomposition, the channel matrix H can be sorted so that detection can be started with the signal that is best received. The sorting of the columns of the matrix H is expressed by a permutation matrix P.

The received signal vector y can be multiplied by the matrix Q ^H to obtain a modified received signal vector y '( 504 ).

The modified received signal vector y ', the matrices Q and R, and the channel decoder output signals l _n ² (n = 1, 2, ..., N) are then fed to the M algorithm ( 506 ), with only the m metrikbesten paths being followed up ( 508 ).

geschrieben werden, wobei die Dimensionen von y' und R dem eigentlichen Baum-Such-Level entsprechen. Um aus der Liste möglichst effizient Log-Likelihood-Ratios zu bestimmen (510), wird die bekannte max-log Annäherung verwendet, so dass:

The metric for the tree search of the M-algorithm can be used as

written, with the dimensions of y 'and R correspond to the actual tree search level. To determine log likelihood ratios as efficiently as possible from the list ( 510 ), the known max-log approximation is used so that:

Finally, in order to obtain the second channel decoder output signals in the form of the third probability information l _n ³ (n = 1, 2,..., N), the log-like values determined by the SQRD-M algorithm are obtained lihood ratios and the log-likelihood ratios l _n ² (n = 1, 2, ..., N) are subtracted from each other ( 512 ).

Again, it should be mentioned that through the block 212 quite different nonlinear detection methods can be realized with other algorithms than the one just described. The only important thing is that soft information can be processed.

6 now shows a block diagram of a turbo-receiver according to the invention 600 , which can be used in the context of OFDM transmission with P sub-frequency carriers.

The recipient shown by way of example 600 includes in a receiver frontend 602 by way of example R = 4 reception branches 602-1 to 602-4 , In these reception branches 602-1 to 602-4 The signals received by the four receive antennas are down-converted and analog-to-digital converted. Thereafter, the so-called. Cyclic prefix is processed and the individual received signals are transformed by means of a discrete Fourier transform in the frequency domain.

wobei x_p[k] das Symbol bedeutet, welches auf Sub-Frequenzträger p (p = 1, 2, ..., P) im k-ten OFDM-Symbolintervall übertragen wurde. H(r,p)p bedeutet den entsprechenden Kanalkoeffizienten zur r-ten (r = 1, 2, ..., R) Empfangsantenne auf dem p-ten (p = 1, 2, ..., P) Subträger.Will with y r p [K] the kth sample of the pth (p = 1, 2, ..., P) FFT output of the receiving antenna r (r = 1, 2, ..., R), if the kth OFDM Symbol is processed, so can write:

where x _p [k] is the symbol transmitted on sub-frequency carrier p (p = 1, 2, ..., P) in the k th OFDM symbol interval. H (R, p) p means the corresponding channel coefficient to the rth (r = 1, 2, ..., R) receive antenna on the pth (p = 1, 2, ..., P) subcarrier.

berechnet werden. Die von dem linearen MMSE-Detektor 604-1 geschätzten Sendesymbole x ^'_p werden einem Frequenz-Offset-Kompensationsblock 604-2 zugeführt, um eventuelle Frequenz-Offsets durch Lokaloszillatordrifts oder Dopplerverschiebungen zu kompensieren. Das kann z. B. mit bekannten Autokorrelations- oder Spektralmethoden geschehen. Die daraus resultierenden frequenzoffsetkompensierten Schätzwerte x ^_p werden einem Block 604-3 zugeführt, um daraus a priori Wahrscheinlichkeitsinformationen 11 (p = 1, 2, ..., P) für den nachfolgenden Kanaldecoderblock 608 zu generieren.A linear MIMO detector 604 comprises, for each sub-frequency carrier p (p = 1, 2, ..., P), a linear MMSE detector 604-1 to determine, based on the four input signals y = [y _p ¹ , y _p ² , y _p ³ , y _p ⁴ ] ^T for each sub-frequency carrier p (p = 1, 2, ..., P), a first one Estimate x ^ ' _p for the corresponding transmitted symbol x _p . This requires the linear MMSE detector 604-1 Estimates for the MIMO channel matrix H and the received noise power σ ^{2 obtained} from a parameter estimator 606 to be provided. Basing on the estimated channel matrix H and the noise power σ ² can now be in one block 607 the expression

be calculated. The from the linear MMSE detector 604-1 estimated transmit symbols x ^ ' _p become a frequency offset compensation block 604-2 supplied to compensate for any frequency offsets by Lokaloszillatordrifts or Doppler shifts. This can z. B. done with known autocorrelation or spectral methods. The resulting frequency offset compensated estimates x ^ _p become one block 604-3 fed to it a priori probability information 11 (p = 1, 2, ..., P) for the subsequent channel decoder block 608 to generate.

The channel decoder 608 works based on the 5 described turbo-decoder principle. It supplies at its output a priori information l _p ² (p = 1, 2,..., P) for a non-linear MIMO detector 612 which, as of a second detection step, the channel decoder 608 again provides second detector output signals in the form of a priori probabilities l _p ³ (p = 1, 2, ..., P), which the first detector output signals l _p ¹ (p = 1, 2, ..., P) from the second Replace detection step. After a desired number of iterations, after a sign evaluation of the LLRs, estimated values for the transmitted symbols or bits can then be output.

After this in the foregoing, the inventive concept is described in detail was, will be in the following on the performance of the receiver concept of the invention To be received.

wobei p(l, c) die zweidimensionale Verteilung der gesendeten Bits c und der Empfänger-LLRs l bedeutet, und p_c(c) und p_l(l) die jeweiligen Randverteilungen bedeuten. Die Transinformation I ist dann ein Wert zwischen 0 und 1, wobei eine 1 eine richtige Detektion bedeutet.The accuracy of the LLRs at any stage of a receiver according to the invention can be measured as so-called transinformation or mutual information I (Mutual Information) with the transmitted bits:

where p (l, c) denotes the two-dimensional distribution of the transmitted bits c and the receiver LLRs l, and p _c (c) and p _l (l) denote the respective edge distributions. The transinformation I is then a value between 0 and 1, with a 1 meaning a correct detection.

7 Figure 12 shows the mutual trans-information I plotted against the signal-to-noise power ratio (SNR) for various MIMO detectors (IF, MMSE, M algorithm) in a 4x4 QPSK transmission. The MIMO detectors do not process any probability information from the channel decoder. That is, there is no feedback from extrinsic information.

For low SNR achieves MMSE as the first detection stage a good reception quality with minimal Computing overhead compared to the QRD-M algorithm for different M.

A turbo decoder comprises two soft-in / soft-out decoders which are operated iteratively with an interleaver and a de-interleaver in between. A so-called EXIT diagram (EXtrinsic Information Transfer) with LTE (Long Term Evolution) turbo code parameters is in 8th shown. The diagram assumes Gaussian LLR distributions and illustrates the exchange of extrinsic information between the two BCJR decoders of a turbo decoder. The different curves in the diagram correspond to different channel qualities. For a noisy channel, the EXIT chart has no aperture and the turbo decoder does not converge to I = 1. Once an opening in the chart is apparent, the point I = 1 can be reached, at least with a large number of turbo decoder iterations become. The better the channel becomes, the further to the opening of the EXIT chart and the turbo decoder converges faster.

The Feedback loop between channel decoder and detector has in a turbo receiver the potential, an opening in the EXIT chart for to create very low SNR and thereby the operating area Recipient to enlarge. If already an opening exists in the EXIT chart, so the feedback loop can be the number the turbo decoder iterations to reduce.

The 9a and b show how the M detector processes extrinsic information from the channel decoder.

Of the M-detector already uses a priori information of the channel decoder good at small M.

For a receiver, defined as a sequence of signal processing blocks, the performance in terms of transinformation I at a particular SNR can be estimated by concatenating EXIT charts. The total computational effort is obtained by adding up the individual computational requirements per LLR per block. The optimal combination of receive algorithms is that which meets the requirements for target trans information at a target SNR with the least amount of computation. An exemplary graph is in FIG 10 shown.

ZF, MMSE and M detectors with M ≦ 16 (with and without extrinsic information) and turbo decoders with up to 20 iterations are considered. The graph in 10 allows only changing sequences between detectors and channel decoders. For the channel decoder, complete iterations of both BCJR decoders and (de) interleavers are assumed. The evaluation was made for a target transinformation of 0.99, for 4 x 4 QPSK transmission and varying SNR.

Linear IF detection and turbo decoder (without feedback loop) leads to the curve 1010 , For a target SNR of 10 dB, two turbo decoder iterations are needed, while for a target SNR of 6 dB, nine turbo decoder iterations are needed.

The curve 1020 stands for M detector and turbo decoder (without feedback loop). This allows a compromise between detector and channel decoder complexity to be found.

The curve 1030 describes a linear MMSE detector and a turbo decoder (without feedback loop).

The curve 1040 describes a M-detector-based turbo receiver with feedback loop, ie extrinsic information is fed from the turbo decoder output back into the M-detector.

The curve 1050 describes a hybrid MMSE turbo receiver according to the invention, in which the first detector is a linear MMSE detector and thereafter follows a feedback loop between channel decoder and M detector.

Out 10 It can be seen that the hybrid turbo receiver according to the invention in an SNR range of 1 dB to 4 dB can be implemented more cost-effectively than all considered conventional receiver concepts. Compared with simple MMSE (curve 1030 ) it has a lower complexity by the number of required turbo decoder iterations can be reduced. For 1 dB target SNR, MMSE detection followed by two turbo decoder iterations, M detection with M = 3 and five turbo decoder iterations, the 260-per-LLR receive concept of the present invention meets the target requirements while being complex below that of receiver concepts using a combination of M-detector-based turbo receiver (curve 1040 ) and M detector / turbo decoder (curve 1020 ).

The reception quality known turbo receiver let yourself that is, by using a first linear receiver stage achieve significantly reduced computational effort. The combination possibilities of Signal processing blocks allows a scalable adaptation of the computing power to be integrated into the respective receiving requirements of a recipient. The receiver concept according to the invention is thus in particular for modern software-defined radio implementations interesting, because the different algorithms are not as dedicated Hardware must be present but on processors successively simply different Code executed can be.

Depending on the circumstances can the methods of the invention So be implemented in hardware or in software. The implementation can be on a digital storage medium, especially a floppy disk or a CD with electronically readable control signals, which interact with a programmable computer system can, that the inventive method accomplished becomes. Generally, the invention thus also consists in a computer program product with a program code stored on a machine-readable carrier to carry out of the method according to the invention, if the computer program product runs on a computer. In other words The invention can thus be used as a computer program with a program code for execution the process can be realized when the computer program is up a computer or microcontroller expires.

Claims (12)

Receiving device ( 200 ; 600 ) for receiving a transmission signal from a superposition of impaired, channel-coded transmission signals, having the following features: a reception branch ( 202 ; 602 ), at which the superposition of the impaired, channel-coded transmission signals is receivable; a linear detector device ( 204 ; 604 ), the input side with the receiving branch ( 202 ; 602 ), for outputting, in a first detection step, on the output side, by a linear detection method, a first detector output signal comprising a first probability information (l _n ¹ ; l _p ¹ ) about potentially transmitted transmission symbols of the transmission signal; a channel decoder device ( 208 ; 608 ) for decoding the first detector output signal and for providing a channel decoder output signal comprising second probability information (l _n ² ; l _p ² ) over the potentially transmitted transmission symbols of the transmission signal which is more reliable than the first likelihood information (l _n ¹ ; l _p ¹ ) ; and a non-linear detector device ( 212 ; 612 ), the input side with the receiving branch ( 202 ; 602 ) and the channel decoder output signal for outputting, by a non-linear detection method, a second detector output signal comprising third probability information (l _n ³ ; l _p ³ ) about the potentially transmitted transmit symbols of the transmit signal that is more reliable than the second probability information wherein the second detector output signal is fed back to an input of the channel decoder means to replace the first detector output signal in a subsequent detection step. Receiving device according to Claim 1, in which the linear detector device ( 204 ; 604 ) comprises a linear MMSE detector for separating the corrupted channel coded transmit signals so that the first detector output signal is assignable to the transmit signal. Receiving device according to one of the preceding claims, in which the linear detector device ( 204 ; 604 ) is adapted to output the first detector output signal in the form of log likelihood quotients (l _n ¹ ; l _p ¹ ). Receiving device according to one of the preceding claims, in which the channel decoder device ( 208 ; 608 ) has a deinterleaver on the input side to deinterleave interleaved first detector output signals prior to decoding, and wherein the channel decoder means has an interleaver on the output side to re-interleave the decoded first detector output signals. Receiving device according to one of the preceding claims, in which the channel decoder device ( 208 ; 608 ) is adapted to output the channel decoder output signal in the form of log likelihood quotients (l _n ² ; l _p ² ). Receiving device according to one of the preceding claims, in which the channel decoder device ( 208 ; 608 ) has a trellis-based channel decoder. Receiving device according to one of the preceding claims, in which the non-linear detector device ( 212 ; 612 ) comprises a trellis-based detector to separate the impaired channel-coded transmit signals so that the second detector output signal is assignable to the transmit signal. The receiving device of claim 7, wherein the trellis-based Detector is a reduced-cost detector, such that the Reduction of costs by a state reduction and / or by a Path selection is accomplished by the trellis. Receiving device according to one of the preceding Claims, which is adapted to the plurality of impaired, channel-coded transmission signals in at least two detection steps, such that in the first step of the channel decoder means, the first detector output signal supplied is the channel decoder output signal and, based on this, the receive second detector output signal, and that in the second Step of the channel decoder device, the second detector output signal is fed for more channel decoder outputs and, based on that, more to obtain second detector output signals. Receiving device according to one of the preceding Claims, in a MIMO-OFDM receiver is implemented. Procedure ( 100 ) for receiving a transmission signal from a superimposed impaired, channel-coded transmission signals, comprising the following steps: receiving the superimposition of impaired, channel-coded transmission signals with a reception branch; Detecting, by a linear detection method in a first detection step, a first detector output based on the overlay, wherein the first detector output includes first likelihood information about potentially transmitted transmit symbols of the transmit signal; Decoding the first detector output signal and providing a channel decoder output signal including second probability information about the potentially transmitted transmit symbols of the transmit signal that is more reliable than the first probability information; Detecting, by a non-linear detection method, a second detector output based on the overlay and the channel decoder output, the second detector output comprising third probability information about the potentially transmitted transmit symbols of the transmit signal; and replacing, in a subsequent detection step, the first detector output signal by the second detector output signal. Computer program for performing the steps of the method according to claim 11, when the computer program is on a computer and / or Microcontroller expires.