Methods and Apparatuses for reducing feedback overhead

TECHNICAL FIELD

The present disclosure relates to the field of wireless communications, and in particular to methods and apparatuses for reducing feedback overhead (e.g. CSI feedback) by employing efficient amplitude and phase quantization and reporting of coefficients in a communications network.

BACKGROUND

Beamforming is a crucial part of the third Generation Partnership Project (3GPP) Release (Rel.) 15 which define a New Radio (NR) access technology that enables a radio base station (also denote herein gNB) and a User Equipment (UE) to establish and adapt communication links using spatially precoded pilot signals. Important information in 5G to improve communication links and to efficiently adapt the beamforming technique is feedback reported by a gNB and/or a UE regarding Channel State Information or CSI feedback reporting.

In Rel.-15 Type-II CSI reporting, it is assumed that dual-stage precoding is performed in the frequency domain on a per subband basis, i.e., a single precoder is calculated for a group of adjacent Physical Resource Blocks (PRBs), referred to as a ‘subband’. The Rel.-15 Type-II dual-stage precoder comprises two components: A first-stage precoder denoted F1 that is identical for all subbands and which contains the selected entries/beams selected from a Discrete Fourier Transform - based codebook (DFT-based codebook), and a second stage precoder denoted F2 which contains the subband-dependent beam-combining coefficients of all subbands.

The feedback overhead for reporting the beam-combining coefficients increases approximately linearly with the number of subbands, and it

becomes considerably large for large numbers of subbands. To overcome the high feedback overhead of the Rel.-15Type-ll CSI reporting, it has been decided in 3GPP-Radio Access Network (RAN) standardization meeting, 3GPP RAN#81 [1]to study feedback compression schemes for the second stage precoder F2. In several contributions [2 ]-[7], it has been demonstrated that the number of beam-combining coefficients in F2 may be drastically reduced when transforming F2 using a small set of DFT or Discrete Cosine Transform (DCT) basis vectors into the delay domain.

Rel.-15 dual-stage precodinq and CSI reporting:

Assuming a rank-R transmission and a dual-polarized antenna array at the gNB with configuration (N1, N2, 2), the Rel.-15 double-stage precoder disclosed in [8] for the s-th subband and r-th transmission layer is given by

where the precoder matrix F(r)(s) has 2N1N2 rows corresponding to the number of ports, and S columns for the reporting subbands/PRBs. The matrix
is the wideband first-stage precoder containing 2L spatial beams for both polarizations, which are identical for all S subbands, and FA is a diagonal matrix containing 2L wideband amplitudes associated with the 2L spatial beams, and is the second-stage precoder

containing 2L subband (subband amplitude and phase) complex frequency-domain combining-coefficients associated with the 2L spatial beams for the s-th subband.

According to [8], the reporting and quantization of wideband amplitude matrix FA and subband combining coefficients in are quantized and

reported as follows:

Quantization and reporting of wideband amplitudes of Matrix FA

- The wideband amplitude corresponding to the strongest beam which has an amplitude value of 1 is not reported. The wideband amplitude values associated with the remaining 2L - 1 beams are reported by quantizing each amplitude value with 3 bits.

Quantization and reporting of amplitude and phase values of subband precoder

- The amplitudes and phase values of the coefficients associated with the first leading beam are not reported (they are assumed to be equal to 1 and 0).

- For each subband, the amplitudes of the 5 coefficients associated with the first 5 - 1 leading beams (other than the first leading beam) are quantized with 1 bit

- The amplitude values of the remaining 2 L - B beams are not reported (they are assumed to be equal to 1 ).

- For each subband, the phase values of the 5 - 1 coefficients associated with the first 5 - 1 leading beams (other than the first leading beam) are quantized with 3 bits.

- The phase values of the remaining 2 L - B beams are quantized with 2 bits.

- The number of leading beams for which the subband amplitude is reported is given by 5 = 4, 4 or 6 when the total number of configured spatial beams L = 2, 3, or 4, respectively.

Space-delay Precoder:

Collecting the precoders F(r) (s) for all S subbands in a matrix F(r), we obtain [2,3,7]

Then the second-stage precoder can be written as:

whose u-th row contains the complex

combining-coefficients associated with the u-th beam over S subbands,

By considering a transformation on the subband precoder the overall
precoder may be written as

where matrix contains the complex-combining coefficients and

the matrix is composed of a number of basis vectors used to

perform a compression in the frequency domain. In general, when V < S a compression of the combining coefficients
is achieved. Each complex coefficient in
in (3) is associated with a specific delay (in the transformed domain) as each DFT/DCT basis vector models a linear phase increase over the subbands.

The number of spatial beams and indices of spatial beams may be different, identical, partially identical or non-identical over the transmission layers.

In addition, with respect to the spatial beams, the delays may be partially identical or non-identical over the beam. Due to different spatial beam configuration over the layers, the delay configuration may vary over the layers as well. Therefore, multiple configurations of the beam and delay configurations are possible. However, the spatial beam and delay configuration of the space-delay precoder shall be aligned with the physical structure of the radio channel. The radio channel is comprised of a number of clusters of scatterers associated with respective delays (see channel cluster #1, delay #1,.., channel cluster #3, delay #2 in Figure. 1).

In the example of Figure 1, each transmit spatial beam of the gNB (beam#1, beam#2, beam#3 and beam#4 in this example) is associated with a single or few channel clusters with corresponding delays. Beam #1 is associated with cluster #1 and delay #1 . Beam #2 and Beam #3 are associated with the direct Line Of Sight (LOS) channel component and with cluster #3 and delay #3. Beam #4 is associated with cluster #2 and delay #2. As shown, the delays of cluster #1 and cluster #2 are different and longer than the delay of cluster #3 (closest to the UE).

In order to capture a significant portion of the energy of the radio channel at the UE, the spatial DFT/DCT beams of the first stage precoder need to point in the direction of the channel clusters. In a typical channel setting the clusters are uniformly distributed around the gNB, and each transmit spatial beam is associated with a single or few neighbored clusters. Moreover, due to the uniform distribution of the channel clusters, each cluster is associated with a different delay. The number of clusters to which each spatial beam is associated with depends mainly on the beam width (which is related to the aperture size of the antenna array at the gNB). The larger the beam width (i.e., the smaller the aperture size of the antenna array) the more channel clusters are associated with the spatial beam. Therefore, the delay configuration (number of delays per beam and the values of the delays) of each spatial beam depends on the channel cluster(s) to which the spatial beam is associated with.

For the transformed precoder, each spatial beam is associated with either a single or a small set of delays. The transmit beams are hence “delayed” by specific delay(s) before transmission. The delays need to be selected in such a way that all 2 L beams are coherently combined at the UE. Note again that each delay is represented by entries of a DFT/DCT vector which models a linear phase increase over the subbands.

The selection of the delays for a spatial beam is therefore identical to a selection of DFT/DCT vectors. Due to the delay distribution of the channel clusters, it should be understood that the delays associated with one spatial beam might be different to the delays associated with another spatial beam. Similarly, the delay configuration (number of delays and the delay values) may be different for different beams. Different delay configurations are discussed in detail in applicant's documents [2], [3], [7]. Note that document [7] has not yet been published when the present application is filed and hence [7] is not to be considered prior art for the subject matter of the claims and teachings of the present application.

When different DFT/DCT vectors are used per spatial beam, a matrix containing all selected DFT/DCT vectors of all configured spatial beams may be used to form a common matrix for the transformation. The common transformation matrix contains all the selected DFT/DCT vectors of all beams. When such a matrix is used for the transformation, the combining coefficients associated with a beam contain few non-zeros coefficients only for the DFT/DCT vectors it is associated with and zeros elsewhere.

Therefore, the complex combining coefficients in matrix in (3) may
contain a large number of values which are close to zero.

SUMMARY

An object of embodiments herein is to provide methods and apparatuses in the form a User Equipment (UE) and a radio base station or network node or gNB respectively for reducing (CSI) feedback overhead by employing efficient amplitude and phase quantization and reporting of coefficients in a communications network that employs beamforming and/or MIMO operation. Further, the present embodiments address the problem of how to efficiently quantize and report the transformed combining coefficients.

According to an aspect of embodiments herein, there is provided a method performed by a UE for reducing feedback overhead related to CSI in a communications network employing MIMO operation:

- decomposing each entry corresponding to a (i, j)-th combining coefficient of a precoder matrix
into at least two coefficients, wherein r denotes a r-th transmission layer; said (i, j)-th combining coefficient is associated with a i-th beam and a j-th delay, and wherein each combining coefficient is associated with an amplitude and a phase-information;

- quantizing, separately, each of said at least two coefficients with a least one bit, and

- reporting information related to at least one phase value or at least one amplitude value or at least one phase value and an amplitude value of said quantized coefficient.

According to another aspect of embodiments herein, there is provided an apparatus in the form of UE for reducing feedback overhead in a communications network, the UE comprising a processor and a memory, said memory containing instructions executable by said processor whereby said UE is operative to perform any one of the subject matter of method claims 1-29.

There is also provided a computer program comprising instructions which when executed on at least one processor of the UE according to claim 30, cause tat least said one processor to carry out the method according to anyone of claim 1-29.

A carrier containing the computer program, wherein the carrier is one of a computer readable storage medium; an electronic signal, optical signal or a radio signal.

There is also provided a method performed by a radio network node gNB for reducing feedback overhead related to Channel State Information, CSI, in a communications network employing Multi Input Multi Output, MIMO operation, the method comprising: receiving, from a UE, a report including information related to at least one phase value or at least one amplitude value or at least one phase value and an amplitude value each quantized coefficient which is quantized with a least one bit by a UE;

wherein each entry corresponding to a (i, j)-th combining coefficient of a precoder matrix is decomposed by the UE into at least two

coefficients, wherein r denotes a r-th transmission layer; said (i, j)-th combining coefficient is associated with a i-th beam and a j-th delay, and wherein each combining coefficient is associated with an amplitude and a phase-information;

According to another aspect of embodiments herein, there is also provided a radio network node or gNB for reducing feedback overhead in a communications network, the radio network node comprising a processor and a memory, said memory containing instructions executable by said processor whereby said the radio network node is operative to perform the subject-matter of anyone of claims 31 -39.

There is also provided a computer program comprising instructions which when executed on at least one processor of the radio network node

according to claim 40, cause the at least said one processor to carry out the method according to anyone of claims 31-39.

A carrier is also provided containing the computer program wherein the carrier is one of a computer readable storage medium; an electronic signal, optical signal or a radio signal.

Several advantages with the embodiments herein are presented in the detailed part of this disclosure.

BRIEF DESCRIPTION OF THE DRAWINGS

Examples of embodiments and advantages of the embodiments herein are described in more detail with reference to attached drawings in which:

Figure 1 depicts an example network scenario wherein embodiments herein may be employed.

Figure 2 illustrates amplitude distribution of the combining coefficients in matrix

Figure 3 Amplitude distribution of coefficients bi,j for scheme 3 according to an embodiment herein.

Figure 4 illustrates a flowchart of a method performed by a UE according to some exemplary embodiments herein.

Figure 5 is a block diagram depicting a UE according to exemplary embodiments herein.

Figure 6 is a block diagram depicting a radio network node according to exemplary embodiments herein.

DETAILED DESCRIPTION

In the following is presented a detailed description of the exemplary embodiments in conjunction with the drawings, in several scenarios, to enable easier understanding of the solution(s) described herein.

As previously described, in 3GPP new radio system, two types of codebook, namely Type-1 and Type-2 codebook, have been standardized for the CSI feedback in the support of advanced MIMO operation.

The present embodiments address the problem of how to efficiently quantize and report the transformed combining coefficients in order to reduce CSI feedback overhead in a communications network employing beamforming.

A. Quantization and reporting of complex combining coefficients of matrix
which was previously presented in equation (3) and repeated below is

the overall precoder which may be written as

Each complex coefficient in
in (3) is associated with a specific delay (in the transformed domain) as each DFT/DCT basis vector models a linear phase increase over the subbands. contains complex-combining

coefficients. The values of N1 and N2 are design parameters and may be included in a configuration of an antenna array at a gNB, which antenna array may for example be dual-polarized, although the embodiments herein are not restricted to dual-polarized antenna arrays.

An approach for quantizing the amplitude and phase values of the coefficients in according to an exemplary embodiment, is to quantize

directly each amplitude and phase value with N1 and N2 bits, respectively.

For example, assuming that matrix contains UD coefficients, then

UD(N1 + N2) bits are required for reporting the amplitude and phase information of to the gNB. Flowever, as mentioned previously, each of

the U beams are typically associated with only a set of delays and not all D delays. Therefore, the matrix
may be considered as a sparse matrix where a large number of the coefficients are close to zero.

In the following description, the matrix in equation (3) may

contain the complex combining coefficients associated with all (2L) spatial beams (i.e., U = 2 L), or only a subset of spatial beams (e.g., U < 2 L), and/or all ( V) delays/basis vectors (i.e., D = V), or only a subset of delays/basis vectors (e.g., D < V).

1. Selection and reporting of non-zero coefficients by using a bitmap

In order to save feedback overhead for reporting a quantized version of an approach according to an exemplary embodiment is to feedback

only the amplitude and phase-information of the non-zero coefficients of matrix and to indicate by a bitmap the indices of the reported

coefficients. For example, the first bit in the bitmap may be associated with the first coefficient, the second bit with the first coefficient of matrix and
so on. When a bit in the bitmap is set to '1' the corresponding coefficient (amplitude and/or phase) may be reported and otherwise not. In this way, the overhead for reporting the combining coefficients may be largely reduced; however, the number of feedback bits is not fixed and may vary for each reporting instance. (see below how he number of feedback bits may be fixed).

2. Selection and reporting of K strongest coefficients

In order to fix the number of feedback bits for reporting the combining coefficients, the receiver may be configured to feedback the amplitude and/or phase values of the K strongest coefficients of matrix where
the value of the parameter K is configurable by the gNB. The K strongest coefficients may be represented by the K entries having the highest

amplitude (or power) over the elements in
When a bit in the bitmap is set to '1', the UE may be configured to report the phase and/or amplitude values of the associated coefficient bi,j to the gNB. The bitmap may hence contain no more than K '1' s.

In order to increase the flexibility of selecting the coefficients and to improve the system performance, the receiver (e.g. a UE or another gNB) may be configured to select Ku strongest coefficients per row/beam out of matrix where the parameters Ku may be configurable by the gNB

(transmitter). Note that the values of Ku may be identical for a set of rows/beams of matrix In such a case, a single parameter R may be

used to configure multipole parameters Kd.

Similarly, the receiver may be configured to select Kd strongest coefficients per column/delay out of matrix where the parameters Kd may be

configurable by the gNB. Note that the values of Kd may also be identical for a set of columns/delays of matrix In such a case, a single

parameter 0 is used to configure multiple parameters Kd.

3. Selection of submatrices of
for reporting

According to an exemplary embodiment, to reduce the overhead for reporting the coefficients in the receiver may be configured to report

only the amplitude and/or phase information for a subset of the coefficients in The subset of coefficients in may contain the combining

coefficients associated with the “strongest” beams and/or “strongest” delays. In such a case, the rows and/or columns of
may be assumed to be ordered in such a way that the combining-coefficients satisfy:

and/or

In an example, the receiver may be configured to report the amplitude and/or phase information of the coefficients associated with the U' “strongest” beams. The receiver may then report the amplitude and/or phase information of the coefficients:

In another example, the receiver maybe configured to report the amplitude and/or phase information of the coefficients associated with the D “strongest” delays. The receiver may report the amplitude and/or phase information of the coefficients:

In another example, the receiver may be configured to report the amplitude and/or phase information of the coefficients associated with the U' “strongest” beams and D “strongest” delays. The receiver may report the amplitude and/or phase information of the coefficients:

In order to further significantly reduce the feedback overhead for reporting a quantized version of three decomposition and quantization schemes

for are described according to some embodiments herein:

1. Scheme 1

The first scheme decomposes the (i,;)-th combining-coefficient of matrix associated with the i-th beam and ;-th delay into two coefficients, ai and bi,j,

where bi,j is the complex-valued normalized combining-coefficient associated with the i-th beam and j-th delay, and ai is a real-valued coefficient representing a common amplitude for the combining coefficients for all delays associated with the i-th beam. Note that the calculation of the values ai is implementation specific.

2. Scheme 2

The second scheme decomposes the (/,y)-th combining-coefficient of matrix
associated with the i-th beam and y-th delay into two coefficients, dj and bi,j,

where bi,j is the complex-valued normalized combining-coefficient associated with the i-th beam and j-th delay, and dj is a real-valued coefficient representing a common amplitude for the combining coefficients for all beams associated with the y-th delay. Note that the calculation of the values dj is implementation specific.

3. Scheme 3

The third scheme decomposes the (i,j)-th combining-coefficient of matrix associated with the i-th beam and j-th delay into three coefficients, aj

and bi,j,

where bi,j is the complex-valued normalized combining-coefficient associated with the i-th beam and j-th delay, dj is a real-valued coefficient representing a common amplitude for the combining coefficients for all beams associated with the j-th delay, and ai is a real-valued coefficient representing a common amplitude for the combining coefficients for all

delays associated with the i-th beam. Note that the calculation of the values ai and dj are implementation specific.

CLAIMS

1. A method performed by a User Equipment, UE, (500) for reducing feedback overhead related to Channel State Information, CSI, in a communications network employing Multi Input Multi Output, MIMO operation, the method comprising:

-(401 ) decomposing each entry corresponding to a (i, j)-th combining coefficient of a precoder matrix
into at least two coefficients, wherein r denotes a r-th transmission layer; said (i, j)-th combining coefficient is associated with a i-th beam and a j-th delay, and wherein each combining coefficient is associated with an amplitude and a phase-information;

-(402) quantizing, separately, each of said at least two coefficients with a least one bit, and

-(403) reporting information related to at least one phase value or at least one amplitude value or at least one phase value and an amplitude value of said quantized coefficient.

2. The method according to claim 1 comprising decomposing (401), using a first scheme, the (i,j)-th combining-coefficient of said matrix associated with the i-th beam and j-th delay into two

coefficients, ai and bi,j,

where bi,j is complex-valued normalized combining-coefficient associated with the i-th beam and j-th delay, and ai is a real-valued coefficient representing a common amplitude for the combining coefficients for all delays associated with the i-th beam.

3. The method according to claim 1 comprises decomposing (401), using a second scheme, the (i ,j)-th combining-coefficient of matrix
associated with the i-th beam and j-th delay into two coefficients, dj and bi,j,

where bi,j is the complex-valued normalized combining-coefficient associated with the i-th beam and j-th delay, and dj is a real-valued coefficient representing a common amplitude for the combining coefficients for all beams associated with the j-th delay.

4. The method according to claim 1 comprises decomposing (401), using a third scheme the (i,j)-th combining-coefficient of matrix
associated with the i-th beam and j-th delay into three coefficients, aj, dj and bi,j,

where bi,j is the complex-valued normalized combining-coefficient associated with the i-th beam and j-th delay, dj is a real-valued coefficient representing a common amplitude for the combining coefficients for all beams associated with the j-th delay, and ai is a real-valued coefficient representing a common amplitude for the combining coefficients for all delays associated with the i-th beam.

5. The method according to anyone of claim 1-4 comprises representing the combining coefficients or only a set of the combining coefficients in by the first scheme, the second scheme, or the

third scheme.

6. The method according to anyone of claims 1-4 comprises representing a first set of combining coefficient of by the first
scheme or the second scheme, and a second set of the combining coefficients of by the third scheme.

7. The method according to anyone of claims 2, 3, 4, comprising quantizing (402) the real-valued coefficients ai (and/or dj) equally with Na(and/or Nd) bits, wherein each complex-valued coefficient bi,j may be quantized with Nb,1 and Nb,2 bits for the amplitude and phase, respectively, where Nb,1 may be lower than Nb,2.

8. The method according to claim 2 comprises quantizing (402) the entries of matrix with Na bits per amplitude and Na bits per

phase, and using 2 UDNa bits for reporting the coefficients of
wherein UD is the number of combining coefficients in matrix

9. The method according to claim 3 wherein quantizing (402) comprises equally quantizing the phase values of bi,j with Nb,2 = Nd bits, and using DNd + UD(Nb,1 + Nb,2) bits for reporting the amplitude and phase information of the coefficients dj and bi,j, wherein UD is the number of combining coefficients in matrix

10. The method according to claim 4 wherein quantizing (402) comprises equally quantizing the real-valued coefficients ai and dj with Na = Nd bits and the phase values of bi,j with Nb,2 Na bits, the amount of feedback that can be saved by scheme 3 is given by UD(Na - Nb,1) - (U + D)Na bits, and using UNa + DNd + UD(Nb,1 + Nb,2) bits for reporting the amplitude and phase information of the coefficients ai, dj and bi,j, wherein UD is the number of combining coefficients in matrix

11.The method according to anyone of the preceding claims comprises reporting only the phase values, only the amplitude values, or the amplitude and phase values of the K strongest coefficients of matrix where the value of the parameter K is configurable by a radio

network node or gNB.

12. The method according to claim 11 comprises selecting Ku strongest coefficients per row/beam out of matrix where the parameters

Ku is configurable by the radio network node or gNB.

13. The method according to claim 11 comprises selecting Kd strongest coefficients per column/delay out of matrix where the
parameters Kd is configurable by the radio network node gNB.

14. The method according to claim 4 comprises representing the coefficients bi,j by only two quantization levels, and quantizing the amplitude information of the coefficients bi,j using only one bit for the amplitude values, and wherein said UE (500) is configured by a radio network node or gNB with Nb,1 = 1, and representing each amplitude value by two quantization levels “a” and “b", where “a” and/or “b" are given by “a = 0” and “b = 1”.

15. The method according to claim 2 or claim 3 comprises, the UE (500) is configurable by the radio base station or gNB with Nb,1 = 1 and representing each amplitude value by two quantization levels “a” and “b", where for example “a” and/or “b" are given by “a = 0” and “b = 1”, when using the first scheme or the second scheme.

16. The method according to claim anyone of claims 2, 3 or 4 comprises applying different quantization levels for the phase values of the coefficients bi,j.

17. The method according to claim 16 comprising using bits for the

phase values associated with the non-zero coefficients and the U’ strongest beams and
bits for the phase values associated with the non-zero coefficients and the remaining beams, where

18. The method according to anyone of claims 1-17 further comprises reporting indices of Discrete Fourier Transform / Discrete Cosine Transform, DFT/DCT vectors associated with the complex combining coefficients of matrix

19. The method according to claim 18 comprises selecting the DFT/DCT vectors from a set of predefined DFT/DCT basis vectors, where each DFT/DCT basis vector is associated with an index.

20. The method according to claim 19 comprises: when there are S DFT/DCT basis vectors, the first DFT/DCT basis vector is associated with a first index (“1”), the second DFT/DCT basis vector is associated with a second index (“1”), and the last DFT/DCT basis vector is associated with the index (“S”), and when reporting D selected DFT/DCT basis vectors,
feedback bits are required.

21. The method according to anyone of claims 1-17 comprises reporting a bitmap, where each bit in the bitmap is associated with an index "d” from the set of Discrete Fourier Transform / Discrete Cosine Transform, DFT/DCT basis vectors associated with the complex combining coefficients of matrix

22. The method according to claim 21 comprises when bitmap consists of a "1” at position "1 then the amplitude and phase values of the leading beam are considered as follows:

- A "1” at position 1 of the bitmap indicates that the amplitude and phase of the combining coefficient of the leading beam associated with index "1” are given by 1 and 0, respectively, and are not reported; and

- the amplitude and phase values of the remaining combining coefficients of the leading beam associated with other indices are given by 0 and 0, respectively, and are not reported.

23. The method according to anyone of claims 2, 3 or 4 wherein the number of leading beams, B, for which the amplitude values of bi,j shall be reported to a radio base station or gNB is given by B = 2L or 2L - 1 for reporting using a DFT/DCT transformation, where L is a number of spatial beams configured.

24. The method according to claims 2, 3 or 4 comprises not reporting to a radio base station or a gNB the quantized amplitude and phase values of bi,j associated with the first leading beam.

25. The method according to anyone of claims 10-24 wherein when Na = 3, the amplitude set for quantizing ai is given

26. The method according to anyone of claims 10-24 wherein when Nd = 3, the amplitude set for quantizing dj is given uniform

27. The method according to anyone of claims 10-24 wherein when Nd = 2, the amplitude set for quantizing dj is given

28. The method according to anyone of claims 2-27 wherein the amplitude set for quantization of bi,j is given {0, 1}.

29. The method according to anyone of claims 2-28 wherein the phase set for quantizing bi,j is given by a 8PSK, Phase Shift Keying, constellation or a 16PSK constellation.

30. A User Equipment (500) for reducing feedback overhead related to Channel State Information, CSI, in a communications network employing Multi Input Multi Output, MIMO operation, the UE (500) comprising a processor (510) and a memory (520) containing instructions executable by said processor (510) whereby the UE (500) is operative to perform anyone of the subject-matter of method claims 1-29.

31. A method performed by a radio network node (700) or gNB for reducing feedback overhead related to Channel State Information, CSI, in a communications network employing Multi Input Multi Output, MIMO operation, the method comprising:

(601) receiving, from a UE (500), a report including information related to at least one phase value or at least one amplitude value or at least one phase value and an amplitude value each quantized coefficient which is quantized with a least one bit by a UE (500); wherein each entry corresponding to a (i, j)-th combining coefficient of a precoder matrix
is decomposed by the UE (500) into at least two coefficients, wherein r denotes a r-th transmission layer; said (i, j)-th combining coefficient is associated with a i-th beam and a j-th delay, and wherein each combining coefficient is associated with an amplitude and a phase-information;

32. The method according to claim 31 comprises configuring the UE (500) to feedback the amplitude and/or phase values of K strongest coefficients of matrix
wherein the value of K is configurable by radio network node (700) or the gNB.

33. The method according to claim 31 or claim 32 comprises configuring the UE to select Ku strongest coefficients per row/beam out of matrix where the parameters Ku is configurable by the radio base

station (700) or gNB.

34. The method according to anyone of claims 31-33 comprises configuring the UE (500) to select Kd strongest coefficients per column/delay out of matrix where the parameters Kd is

configurable by the radio base station (700) or gNB.

35. The method according to anyone of claims 31-35 comprises configuring the UE (500) to report only the amplitude and/or phase information for a subset of the coefficients in

36. The method according to anyone of claims 31-35 comprises configuring the UE (500) to represent the combining coefficients or only a set of the combining coefficients in by the first scheme of

claim 2 or the second scheme of claim 3 or the third scheme of claim 4.

37. The method according to anyone of claims 31-36 comprises configuring the UE (500) to quantize the real-valued coefficients ai

(and/or dj) equally with Na (and/or Nd) bits, wherein ai is a real-valued coefficient representing a common amplitude for the combining coefficients for all delays associated with the i-th beam.

38. The method according to anyone of claims 31-37 comprises configuring the UE (500) to report only the phase values, only the amplitude values, or the amplitude and phase values of the quantized non-zero coefficients bi,j.

39. The method according to anyone of claims 31-38 comprises configuring the UE (500) to report a bitmap, where each bit in the bitmap is associated with an index "d" from a set of DFT/DCT basis vectors.

40. A radio network node (700) or a gNB for reducing feedback overhead related to Channel State Information, CSI, in a communications network employing Multi Input Multi Output, MIMO operation, the gNB (700) comprising a processor (710) and a memory (720) containing instructions executable by said processor (710) whereby the gNB (500) is operative to perform anyone of the subject-matter of method claims 31-39.