A MIMO receiver and a method for performing a MIMO channel equalization

Abstract

The present disclosure relates to a method (10) for performing a MIMO channel equalization. The method (10) comprises: receiving (11) a communication signal which is transmitted over a MIMO channel; executing (12) a data-aided, DA, channel estimation algorithm to calculate a first equalization matrix based on the received communication signal; executing (13) a blind channel estimation algorithm to calculate a second equalization matrix based on the received communication signal; combining (14) the first and the second equalization matrix to receive a combined equalization matrix (42); and applying (15) the combined equalization matrix (42) on the received communication signal to perform a channel equalization.

Description

[0001] A M IMO RECEIVER AND A METHOD FOR PERFORMING A M IMO CHANNEL EQUALIZATION

[0002] TECHNICAL FIELD

[0003] The present disclosure relates to the field of equalization in telecommunication channels. More specifically, the present disclosure relates to a method for performing a MIMO channel equalization and to a MIMO receiver capable of performing a MIMO channel equalization.

[0004] BACKGROUND

[0005] Channel equalization is a technique to compensate distortions of signals which are transmitted through a communication channel. This is especially important for multiple-input-multiple-output (MIMO) channels, where the presence of multiple transmitting and receiving elements (e.g., antennas) can introduce spatial distortions and interferences that can affect the received signals. By adjusting the amplitude and phase of the received signals, a MIMO equalizer aims to recover the transmitted symbols accurately and improve the overall system performance.

[0006] For example, in a fiber which is deployed in optical ground wire (OPGW) cables, lightning strikes can cause ultra- fast changes of the state of polarization (SOP) of a transmitted signal. In a coherent optical transceiver, such SOP changes require fast adaptation of the MIMO equalizer to compensate such distortions.

[0007] Conventional fast equalization techniques, which are purely data-aided (DA), require frequent and long training sequences in the transmitted data stream, which results in a waste of bandwidth. For instance, depending on the symbol rate, a 20% training overhead is required for a 0.6 dB optical signal-to-noise ratio (OSNR) penalty at 20 Mrad / s, and a 11% training overhead is required for a 1.2 dB OSNR penalty at 20 Mrad / s.

[0008] In contrast, conventional blind equalizers do not require such training data, but are often too slow to compensate fast changes in the signal. In a blind equalizer, the so-called Teaming rate’ determines how much each error vector is weighted. Low learning rates result in stable and accurate equalization but can track only slowly changing channels (e.g., static changes, temperature effects, etc.). Dynamic systems, e.g. with fast SOP changes, need a higher learning rate, which can result in instable behavior. The maximum learning rate can be limited by a degree of parallelization in an ASIC which calculates the blind equalization. This is especially problematic in fiber-optic communications, where the disproportion between the signaling rate and the ASIC frequency asks for massive parallel processing.

[0009] SUMMARY

[0010] In view of the above, this disclosure aims to provide an improved method for performing a MIMO channel equalization and an improved MIMO receiver, which overcome the above mentioned limitations and disadvantages.

[0011] These and other objectives are achieved by the solution of this disclosure as described in the independent claims. Advantageous implementations are further defined in the dependent claims.

[0012] A first aspect of this disclosure provides a method for performing a multiple-input-multiple-output (MIMO) channel equalization. The method comprises: receiving a communication signal which is transmitted over a MIMO channel; executing a data-aided (DA) channel estimation algorithm to calculate a first equalization matrix based on the received communication signal; executing a blind channel estimation algorithm to calculate a second equalization matrix based on the received communication signal; combining the first and the second equalization matrix to receive a combined equalization matrix; and applying the combined equalization matrix on the received communication signal to perform a channel equalization.

[0013] This achieves the advantages that a channel equalization can be carried out which efficiently combines the effects of a fast DA equalization and a slower blind equalization. For instance, the combined equalization matrix can track and reverse fast channel changes (e.g., changes in polarization), as well as slow channel changes (e.g., static changes, temperature effects, etc.).

[0014] The first equalization matrix can represent a MIMO response of the communication channel, and the second equalization matrix can represent a scalar response of the communication channel.

[0015] The DA channel estimation algorithm can use training data in the communication signal to determine the first equalization matrix. The training data can comprise a number of training symbols which are transmitted in a known pattern. Due to the combination of DA channel estimation and blind channel estimation, a lower number of training symbols is required for the DA estimation. This reduces the training overhead, thus increasing the usable bandwidth.

[0016] The blind channel estimation algorithm can use signal statistics of the communication signal to determine the second equalization matrix and does not require dedicated training data in the communication signal.

[0017] For instance, the method can be used for optical communication with fast channel changes (e.g. fast SOP) and short training sequences.

[0018] The communication signal can be transmitted by a least two transmitter elements on the transmitter side and received by at least two receiver elements on the receiver side. The communication signal can be an optical signal, which is transmitted via a fiber-optical cable.

[0019] In case of a frequency-domain implementation, applying the combined equalization matrix on the received communication signal may refer to multiplying the combined equalization matrix with the received communication signal.

[0020] In an implementation form of the first aspect, the blind channel estimation algorithm is a decision-directed algorithm.

[0021] In an implementation form of the first aspect, the decision-directed algorithm is a least mean squares (LMS) algorithm.

[0022] The LMS algorithm can efficiently track slowly varying effects (like residual CD, bandwidth limitations), which are not measured / tracked by the faster DA channel estimation.

[0023] Alternatively, the decision-directed algorithm can be a constant-modulus algorithm (CMA).

[0024] In an implementation form of the first aspect, the DA channel estimation algorithm is based on constant-amplitude zeroautocorrelation (CAZAC) training sequences.

[0025] Due to the presence of the blind channel estimation algorithm, shorter CAZAC training sequences can be used (i.e. , less training data is required) which saves training overhead.

[0026] In an implementation form of the first aspect, the first and the second equalization matrix are combined by matrix multiplication. In an implementation form of the first aspect, the second equalization matrix is a product of a scalar function and an identity matrix.

[0027] The scalar function can be a function of frequency and / or time.

[0028] In an implementation form of the first aspect, the step of executing the DA channel estimation algorithm comprises: applying a previously determined version of the second equalization matrix on a first copy of the received communication signal to generate an adapted communication signal, and subsequently calculating the first equalization matrix based on the adapted communication signal.

[0029] For instance, using a copy of the received signal to generate the adapted communication signal allows to separate the (slower) calculations for the signal adaption from a signal path of the received signal. For instance, the calculations take place in a separate control path which is coupled out of the signal path.

[0030] The previously determined version of the second equalization matrix can be multiplied with the received signal to form the adapted communication signal.

[0031] In an implementation form of the first aspect, the first equalization matrix is calculated by: calculating a channel estimate based on training data in the adapted communication signal, and calculating the first equalization matrix by inversion of the channel estimate.

[0032] In an implementation form of the first aspect, the step of executing the blind channel estimation algorithm comprises: applying a previously determined version of the first equalization matrix on a second copy of the received communication signal to generate a further adapted communication signal, and subsequently calculating the second equalization matrix based on the further adapted communication signal.

[0033] For instance, using a copy of the received communication signal to generate the further adapted communication signal allows to separate the (slower) calculations for the signal adaption from a signal path of the received signal. For instance, the calculations take place in a separate control path which is coupled out of the signal path.

[0034] The previously determined version of the first equalization matrix can be multiplied with the received signal to form the further adapted communication signal.

[0035] In an implementation form of the first aspect, the second equalization matrix is calculated by: correlating the further adapted communication signal with an error signal, updating a scalar function of an adaptive filter based on said correlation, and multiplying the updated scalar function with an identity matrix.

[0036] In the case of decision-directed LMS, the error can be computed based on the output of the equalizer and the symbol decisions. Other error signals, like a CMA error, can be alternatively used.

[0037] The scalar function of an adaptive filter can be a filter parameter of the adapted filter (e.g., an LMS filter) which is updated based on the correlation.

[0038] In an implementation form of the first aspect, the step of executing the blind channel estimation algorithm comprises: correlating the received communication signal with an error signal, updating entries of a matrix of an adaptive filter based on said correlation, weighting the updated entries of the matrix based on entries of a previously determined version of the first matrix to obtain a weighted scalar function, and multiplying the weighted scalar function with an identity matrix.

[0039] In an implementation form of the first aspect, the updated entries of the matrix are weighted based on signs of the real and / or imaginary parts of the entries of the previously determined version of the first matrix.

[0040] In an implementation form of the first aspect the combined equalization matrix is applied on the received communication signal by a single MIMO equalizer on the receiver side.

[0041] A second aspect of this disclosure provides a MIMO receiver for performing a MIMO channel equalization. The MIMO receiver comprises: at least two receivers configured to receive different spatial projections of a communication signal which is transmitted over a MIMO channel; and a processor configured to execute a data-aided (DA) channel estimation algorithm to calculate a first equalization matrix based on the received communication signal; wherein the processor is configured to execute a blind channel estimation algorithm to calculate a second equalization matrix based on the received communication signal; and wherein the processor is configured to combine the first and the second equalization matrix to receive a combined equalization matrix. The MIMO receiver further comprises a MIMO equalizer configured to apply the combined equalization matrix on the received communication signal to perform a channel equalization.

[0042] The at least two receivers can be receiving elements of the MIMO receiver (e.g., optical receiver elements or antennas).

[0043] The processor can be an ASIC, a digital signal processor (DSP) or a microprocessor of the MIMO receiver.

[0044] The MIMO receiver according to the second aspect of this disclosure can be configured to carry out the method according to the first aspect of this disclosure.

[0045] BRIEF DESCRIPTION OF DRAWINGS

[0046] The above described aspects and implementation forms will be explained in the following description of specific embodiments in relation to the enclosed drawings, in which:

[0047] FIG. 1 show a flow diagram of a method for performing a MIMO channel equalization according to an embodiment;

[0048] FIG. 2 shows a flow diagram of a method for performing a MIMO channel equalization according to an embodiment;

[0049] FIG. 3 shows a flow diagram of a method for performing a MIMO channel equalization according to an embodiment;

[0050] FIG. 4 shows a flow diagram of a method for performing a MIMO channel equalization according to an embodiment;

[0051] FIGS. 5A-B show flow diagrams of a method for performing a MIMO channel equalization according to an embodiment; and FIG. 6 shows a schematic diagram of a MEMO receiver according to an embodiment.

[0052] DETAILED DESCRIPTION OF EMBODIMENTS

[0053] FIG. 1 shows a flow diagram of a method 10 for performing a multiple- input-multiple-output (MIMO) channel equalization according to an embodiment.

[0054] The method 10 comprises the following steps: receiving Il a communication signal which is transmitted over a MIMO channel; executing 12 a data-aided (DA) channel estimation algorithm to calculate a first equalization matrix based on the received communication signal; executing 13 a blind channel estimation algorithm to calculate a second equalization matrix based on the received communication signal; combining 14 the first and the second equalization matrix to receive a combined equalization matrix; and applying 15 the combined equalization matrix on the received communication signal to perform a channel equalization.

[0055] For instance, the first equalization matrix can be optimized to track and reverse fast channel changes (e.g., changes in polarization), and the second equalization matrix can be optimized to track and reverse slow channel changes (e.g., static changes, temperature effects, etc.).

[0056] The DA channel estimation algorithm can use training data in the communication signal to determine the first equalization matrix. The training data can comprise a number of training symbols which are transmitted in a known pattern. Due to the combination of DA channel estimation and blind channel estimation, a lower number of training symbols is required for the DA estimation. This reduces the training overhead, thus increasing the usable bandwidth. For instance, the method 10 can be used for optical communication with fast channel changes (e.g. fast SOP) and short training sequences.

[0057] The blind channel estimation algorithm can use signal statistics of the communication signal to determine the second equalization matrix and does not require dedicated training data in the communication signal.

[0058] In case of a frequency-domain implementation, applying 15 the combined equalization matrix on the received communication signal may refer to multiplying the combined equalization matrix with the received communication signal.

[0059] For instance, the combined equalization matrix can be applied 15 on the received communication signal by a single MIMO equalizer on the receiver side.

[0060] The communication signal can be transmitted by a least two transmitter elements on the transmitter side and received by at least two receiver elements on the receiver side. The communication signal can be an optical signal, which is transmitted via a fiber-optical cable.

[0061] The DA channel estimation algorithm can be based on constant-amplitude zero-autocorrelation (CAZAC) training sequences. The communication signal can comprise these CAZAC training sequences, e.g. in the form of training symbols. The DA channel estimation algorithm can be optimized to track fast channel changes.

[0062] Due to the presence of the blind channel estimation algorithm, shorter CAZAC training sequences can be used (i.e. , less training data is required) which allows reducing the training overhead in the signal. The blind channel estimation algorithm can be a decision-directed algorithm. For example, the decision-directed algorithm is a least mean squares (LMS) algorithm. Alternatively, the decision-directed algorithm can also be a constant-modulus algorithm (CMA).

[0063] The LMS algorithm can efficiently track any remaining slowly varying effects (like residual CD, bandwidth limitations, etc.), which are not measured / tracked by the faster DA channel estimation. Thus, the LMS algorithm can compensate distortions in a static part of the communication channel.

[0064] For example, the first equalization matrix forms or represents a MIMO equalizer M, e.g. a 2x2 MEMO equalizer, and the second equalization matrix forms or represents a scalar equalizer S. At step 14, the scalar equalizer and the 2x2 MIMO equalizer are combined into one (combined) 2x2 MIMO equalizer. The scalar equalizer can be adapted at a relatively slow pace using the LMS algorithm, whereas the MIMO equalizer can be adapted at a fast pace by using the training sequences, e.g. the CAZAC sequences.

[0065] The first and the second equalization matrix can be combined 14 by matrix multiplication, in particular in case of a frequencydomain implementation. For example, the 2x2 MIMO equalizer can be decomposed into two components \ / [f] and >S'[ / ] that commute, according to:

[0066] The technical effect of the commutative property (M [ ] • S[ ] = S[ ] • M [ ]) is that a stable decomposition of the equalizer into slow and fast components can be obtained, which is independent of the position of the slow and fast components in the channel. For instance, this decomposition is achieved by the second equalization matrix >S'[ / ] being a product of a scalar function s[ / ] and an identity matrix according to: where I is the identity matrix. The scalar function s [ / ] can be a function of frequency f and / or time.

[0067] The following FIGS. 2 to 5B show exemplary steps of the method 10 for performing the MIMO channel equalization. For instance, these figures show how two separate equalizers respectively equalization matrices can be combined into one. Therefore initially, two distinct and consecutive equalizers are considered (left parts of FIGS. 2 and 3). These equalizers can be adapted and combined into a single MIMO equalizer in a signal path 21 (right parts of FIGS. 2 and 3). This combination is made possible due to the commutative nature of the individual (adapted) equalizers. The two equalizers which are combined comprise: the scalar equalizer S, which is based on the decision-directed LMS algorithm or on a constant-modulus algorithm (CMA), and the MIMO equalizer M, which is based on a feed-forward adaptation via CAZAC sequences.

[0068] FIG. 2 shows an exemplary adaption and / or calculation of the DA MIMO equalizer M which can be represented by the first equalization matrix. For instance, FIG. 2 visualizes a possible implementation of step 12 of the method 10.

[0069] The left part in FIG. 2 shows a conventional approach of applying a decision-directed scalar equalizer S to an incoming signal, followed by a separate feedforward DA MIMO equalizer M. In this approach, after application of the S equalizer, the separate M equalizer accesses the communication signal between S and M to update its own filter coefficients. However, this approach is resource-intensive, as both equalizers are applied one after the other in the signal path 21.

[0070] The right part in FIG. 2 shows an improved modification of the M component. To combine the equalizers S and M, the signal tap between them is avoided. Instead, a copy of the incoming signal is coupled out of a signal path 21 into a control path 22 (also referred to as: DA adaptation path).

[0071] Since the reference signal before the DA MIMO component M is no longer directly accessible, it is reconstructed by an additional scalar equalizer S, 23 in the control path 22. For instance, this scalar equalizer 23 was previously determined (e.g., in a previous equalization step).

[0072] Thus, a previously determined version of the second equalization matrix 23 can be applied on a first copy of the received communication signal. This generates an adapted communication signal.

[0073] For instance, the previously determined version of the second equalization matrix 23 can be multiplied with the received signal to form the adapted communication signal. By moving the scalar equalizer S from the signal path 21 (left part of FIG. 2) to the control path 22 (right part of FIG. 2), a complexity of the equalization can be reduced, because the control path 22 can run only on a small fraction (e.g., 3%) of the data.

[0074] After applying the scalar equalizer 23 to the copy of the communication signal in the control path 22, the first equalization matrix can be calculated based on the thus adapted communication signal. As further shown in the right part of FIG. 2, this can be done by: calculating a channel estimate 24 based on training data in the adapted communication signal, and calculating the first equalization matrix by inversion 25 of the channel estimate.

[0075] FIG. 3 shows an exemplary adaption and / or calculation of the scalar equalizer S which can be represented by the second equalization matrix. For instance, FIG. 3 visualizes a possible implementation of step 13 of the method 10.

[0076] The left part in FIG. 3 shows a conventional approach of applying a feedforward DA MIMO equalizer M followed by a separate decision-directed scalar equalizer S, e.g. a decision-directed adaptation algorithm (e.g., DD-LMS). In this approach, after application of the M equalizer, the separate S equalizer accesses the communication signal between S and M via an intermediate signal tap to update its own filter coefficients The right part in FIG. 3 shows an improved modification of the S component. To combine the equalizers S and M, the signal tap between them is avoided. Instead, a further copy of the incoming signal is coupled out of a signal path 21 into a further control path 31 (also referred to as: blind adaptation path).

[0077] Since the reference signal before the MIMO component S is no longer directly accessible, it is reconstructed by an additional DA MIMO equalizer M, 32 in the control path 31. For instance, this DA MIMO equalizer 32 was previously determined (e.g., in a previous equalization step.

[0078] Thus, a previously determined version of the first equalization matrix 32 (e.g., DA MIMO equalizer M) can be applied on a second copy of the received communication signal. This can generate a further adapted communication signal. For instance, the previously determined version of the first equalization matrix 32 can be multiplied with the received signal to form the further adapted communication signal.

[0079] After applying the DA MIMO equalizer 32 to the copy of the communication signal in the control path 31, the second equalization matrix S can be calculated based on the thus adapted communication signal.

[0080] As further shown in the right part of FIG. 3, this can be done by: correlating 33 the further adapted communication signal with an error signal, and updating 34 the scalar function s[ / ] of an adaptive filter based on said correlation. The thus updated scalar function s[ / ] can be multiplied with an identity matrix, according to >S'[ / ] = s[ / ]7, in order to receive the second equalization matrix S(f).

[0081] In the case of decision-directed LMS, the error can be computed based on the output of the equalizer and the symbol decisions. Other error signals, like a CMA error, can be alternatively used.

[0082] The scalar function of an adaptive filter can be a filter parameter of the adapted filter (e.g., an LMS filter) which is updated based on the correlation. This can be carried out by an LMS correlator.

[0083] In case of the LMS algorithm, the correlation can be an LMS correlation, and the updating can be carried out by an LMS updater.

[0084] For instance, FIGS. 2 and 3 mimic cascaded equalizers that have a specific order concerning which filter (e.g., equalizer) is applied first, wherein in each case one respective component equalizer is adapted.

[0085] FIG. 4 shows an exemplary joint adaptation (respectively calculation) of the M and the S components. Due to the commutative property of the equalizers S and M, according to

[0086] MS = SM, the two adaptation algorithms of FIGS. 2 and 3 can be combined into a single structure.

[0087] As shown in Fig. 4, a first copy of the communication signal can be coupled into the DA adaptation path 22 to generate the first equalization matrix M, analogous to FIG. 2: and a second copy of the communication signal can be coupled into the blind adaption path 31 to generate the second equalization matrix S, analogous to FIG, 3. Then the two matrices can be combined 14 via matrix multiplication, wherein MS = SM. The resulting combined equalization matrix 42 can then be used in the signal path 21 to perform the channel equalization. Thereby, a previously determined version of the second equalization matrix 23 can be applied to the first signal copy in the DA adaptation path 22, and a previously determined version of the second equalization matrix 32 can be applied to the second signal copy in the blind adaptation path 31. These previously determined versions of the equalizers 23, 23 which are used in each path 22, 31 can be the resulting matrices S and M from the respective other path 22, 31 from an immediately preceding equalization step.

[0088] The approach according to FIG. 4 provides the advantages that a stable decomposition of slow and fast components can be achieved and a single MIMO equalizer 42 can be used in the signal path 21.

[0089] FIGS. 5A and 5B show a joint adaptation of the M and the S components according to a further example. In this example, the LMS equations for the S equalizer component (e.g., the blind channel estimation algorithm) are further simplified compared to the example of FIG. 4.

[0090] For instance, the combined equalizer 42 can be described by:

[0091] &] - ]

[0092] The LMS algorithm can be a gradient descent algorithm for the quadratic error: is the dual-polarization desired output.

[0093] The gradient of e with respect to , can be computed as: where Ex= Xd— Xoand Ey= Yd— Y0.

[0094] Thus, the received communication signal j can be correlated with an error signal t to update an adaptive LMS filter or more specifically entries of a matrix of an adaptive LMS filter.

[0095] Assuming that the correlations (X'Ex) (Y*EX') , (X*Ey) , and Y*Ey) can be provided by a conventional MIMO LMS correlator 52 as shown in FIG. 5B, only a modified scalar LMS updater 51 is required, which weights the updates with the entries of M and sums them. Thereby, M can be a previously determined version of the first matrix obtained from the DA adaption path 22.

[0096] Thus, under the assumption that a conventional LMS update is already available in the receiver, the MIMO equalizer 32 in the adaptation path 31 can be dispensed with, as illustrated in FIG. 5A.

[0097] In the solution of FIG. 5A, to avoid additional multiplications, the weights M*qfor p, q E {x, y can be approximated as: M*qsign(Re{Mp() - j ■ sign(lm{Mpf).

[0098] Thus, the modified LMS updater 51 can approximate the actual DA MIMO coefficients of M by the sign of their real part plus the imaginary unity by the sign of their imaginary part and use these for weighting the LMS filter.

[0099] In other words: the conventional updates of the LMS filter are weighted based on signs of the real and / or imaginary parts of the entries of the previously determined version of the first matrix M. Thereby, for example, updated entries of a matrix of the LMS filter can be weighted based on said real and / or imaginary parts, wherein this matrix of the LMS filter may contain the correlations (X'Ex). (Y*EX), (X'Ey)'. and (Y*Ey) multiplied by the learning rate.

[0100] In this way, a weighted scalar function s[ / ] can be determined which can be multiplied with an identity matrix to receive / >[ / ]:

[0101] The approach according to FIGS. 5A and 5B provides the advantage that an available conventional LMS update can be reused, which reduces calculation efforts.

[0102] Furthermore, in any of the previous solutions, in the DA adaptation path 21, the channel estimation 24 and channel inversion 25 may include: channel averaging, time-domain and / or frequency-domain smoothing of the estimated channel response, and / or normalization of the amplitude response.

[0103] The scalar equalizer S in the DA adaptation path 21 may be approximated by the identity matrix.

[0104] Furthermore, the scalar LMS algorithm can be replaced by a CMA or other decision-directed algorithms.

[0105] The method 10 or more specifically the combined equalization matrix 42 combines the advantages of blind and data-aided channel estimation. For instance, the method 10 can be used for fast SOP tracking, and accurate channel modeling and equalization. At the same time, the required training overhead is limited, especially compared to a pure DA approach.

[0106] For instance, the combined equalization matrix 42 can track SOP rotation speeds in excess of 20 Mrad / s, compared to a conventional (legacy) LMS-based MIMO equalizer which can track a maximum SOP rotation speed of e.g. 600krad / s.

[0107] FIG. 6 shows a schematic diagram of a MIMO receiver 60 according to an embodiment.

[0108] The MIMO receiver 60 comprises at least two receivers 61, 62 configured to receive different spatial projections of a communication signal which is transmitted over a MIMO channel; and a processor 63 configured to execute a DA channel estimation algorithm to calculate a first equalization matrix M based on the received communication signal; wherein the processor 63 is configured to execute a blind channel estimation algorithm to calculate a second equalization matrix S based on the received communication signal. The processor 63 is further configured to combine the first and the second equalization matrix S, M to receive a combined equalization matrix 42.

[0109] The MIMO receiver 60 further comprises a MIMO equalizer 64 configured to apply the combined equalization matrix on the received communication signal to perform a channel equalization. The at least two receivers 61, 62 can be receiving elements of the MIMO receiver (e.g., optical receiver elements or antennas).

[0110] The MIMO receiver 60 can be a coherent optical receiver. The communications signal can be an optical signal.

[0111] The MIMO receiver 60 can be configured to carry out the method 10 as shown in any one of FIGS. 1 to 5B. For instance, the processor 63 can be configured to carry out the steps 12 to 14 of the method 10.

[0112] The processor 63 can comprise a processing circuitry (not shown) configured to perform, conduct or initiate various operations. The processing circuitry may comprise hardware and / or the processing circuitry may be controlled by software. The hardware may comprise analog circuitry or digital circuitry, or both analog and digital circuitry. The digital circuitry may comprise components such as application-specific integrated circuits (ASICs), field-programmable arrays (FPGAs), digital signal processors (DSPs), or multi-purpose processors. The device may further comprise memory circuitry, which stores one or more instruction(s) that can be executed by the processor or by the processing circuitry, in particular under control of the software. For instance, the memory circuitry may comprise a non- transitory storage medium storing executable software code which, when executed by the processor or the processing circuitry, causes the various operations of the device to be performed. In one embodiment, the processing circuitry comprises one or more processors and a non-transitory memory connected to the one or more processors.

[0113] The present disclosure has been described in conjunction with various embodiments as examples as well as implementations. However, other variations can be understood and effected by those persons skilled in the art and practicing the claimed matter, from the studies of the drawings, this disclosure and the independent claims. In the claims as well as in the description the word “comprising” does not exclude other elements or steps and the indefinite article “a” or “an” does not exclude a plurality. A single element or other unit may fulfill the functions of several entities or items recited in the claims. The mere fact that certain measures are recited in the mutual different dependent claims does not indicate that a combination of these measures cannot be used in an advantageous implementation.

Claims

CLAIMS1. A method (10) for performing a MIMO channel equalization, comprising: receiving (11) a communication signal which is transmitted over a MIMO channel; executing (12) a data-aided, DA, channel estimation algorithm to calculate a first equalization matrix based on the received communication signal; executing (13) a blind channel estimation algorithm to calculate a second equalization matrix based on the received communication signal; combining (14) the first and the second equalization matrix to receive a combined equalization matrix (42); and applying (15) the combined equalization matrix (42) on the received communication signal to perform a channel equalization.

2. The method (10) of claim 1, wherein the blind channel estimation algorithm is a decision-directed algorithm.

3. The method (10) of claim 2, wherein the decision-directed algorithm is a least mean squares, LMS, algorithm.

4. The method (10) of any one of the preceding claims, wherein the DA channel estimation algorithm is based on constant-amplitude zero-autocorrelation, CAZAC, training sequences.

5. The method (10) of any one of the preceding claims, wherein the first and the second equalization matrix are combined (14) by matrix multiplication.

6. The method (10) of any one of the preceding claims, wherein the second equalization matrix is a product of a scalar function and an identity matrix.

7. The method (10) of any one of the preceding claims, wherein the step of executing (12) the DA channel estimation algorithm comprises: applying a previously determined version of the second equalization matrix (23) on a first copy of the received communication signal to generate an adapted communication signal, and subsequently calculating the first equalization matrix based on the adapted communication signal.

8. The method (10) of claim 7, wherein the first equalization matrix is calculated by: calculating a channel estimate (24) based on training data in the adapted communication signal, and calculating the first equalization matrix by inversion (25) of the channel estimate.

9. The method (10) of any one of the preceding claims, wherein the step of executing (13) the blind channel estimation algorithm comprises: applying a previously determined version of the first equalization matrix (32) on a second copy of the received communication signal to generate a further adapted communication signal, and subsequently calculating the second equalization matrix based on the further adapted communication signal.

10. The method (10) of claim 9, wherein the second equalization matrix is calculated by: correlating (33) the further adapted communication signal with an error signal, updating (34) a scalar function of an adaptive filter based on said correlation, and multiplying the updated scalar function with an identity matrix.

11. The method (10) of any one of claims 1 to 8, wherein the step of executing (13) the blind channel estimation algorithm comprises: correlating the received communication signal with an error signal, updating entries of a matrix of an adaptive filter based on said correlation, weighting the updated entries of the matrix based on entries of a previously determined version of the first matrix to obtain a weighted scalar function, and multiplying the weighted scalar function with an identity matrix.

12. The method (10) of claim 11, wherein the updated entries of the matrix are weighted based on signs of the real and / or imaginary parts of the entries of the previously determined version of the first matrix.

13. The method (10) of any one ofthe preceding claims, wherein the combined equalization matrix (42) is applied (15) on the received communication signal by a single MIMO equalizer on the receiver side.

14. A MIMO receiver (60) for performing a MIMO channel equalization, comprising: at least two receivers (61, 62) configured to receive different spatial projections of a communication signal which is transmitted over a MIMO channel; and a processor (63) configured to execute a data-aided, DA, channel estimation algorithm to calculate a first equalization matrix based on the received communication signal; wherein the processor (63) is configured to execute a blind channel estimation algorithm to calculate a second equalization matrix based on the received communication signal; and wherein the processor (63) is configured to combine the firstand the second equalization matrix to receive a combined equalization matrix; and wherein the MIMO receiver (60) further comprises a MIMO equalizer (64) configured to apply the combined equalization matrix on the received communication signal to perform a channel equalization.

Images