NON-LINEAR PRECODING WITH A MIX OF NLP CAPABLE AND NLP NON-CAPABLE LINES

Abstract

The method includes organizing a plurality of subscriber lines into a first group of subscriber lines and a second group of subscriber lines, the first group of subscriber lines at least including all the subscriber lines of the plurality of subscriber lines that do not support non-linear precoding operation and the second group of subscriber lines including the remaining subscriber lines of the plurality of subscriber lines; scaling first signals to be transmitted over respective ones of the first group of subscriber lines to confine respective intermediate transmit power levels at the input of a modulo unit and further to bypass or make ineffective the operation of the modulo unit; and processing the so scaled first signals and second signals to be transmitted over respective ones of the second group of subscriber lines through the first and second precoding stages.

Claims

1. A method for jointly processing signals to be transmitted over respective ones of a plurality of subscriber lines through a non-linear precoder comprising a first non-linear precoding stage configured to operate according to a first triangular precoding matrix and including a modulo unit, followed by a second linear precoding stage configured to operate according to a second precoding matrix, wherein the method comprises organizing the plurality of subscriber lines into a first group of subscriber lines and a second group of subscriber lines, the first group of subscriber lines at least comprising the subscriber lines of the plurality of subscriber lines that do not support non-linear precoding operation, and the second group of subscriber lines comprising the remaining subscriber lines of the plurality of subscriber lines, scaling first signals to be transmitted over respective ones of the first group of subscriber lines to confine respective intermediate transmit power levels at the input of the modulo unit and further to bypass or make ineffective the operation of the modulo unit, and processing the so scaled first signals and second signals to be transmitted over respective ones of the second group of subscriber lines through the first and second precoding stages.

2. A method according to claim 1, wherein the first group of subscriber lines is assigned the first precoding positions in the first and second precoding matrices while the second group of subscriber lines is assigned the last precoding positions in the first and second precoding matrices.

3. A method according to claim 1, wherein the intermediate power levels are confined within a transmit power mask applicable to transmission of signals over the plurality of subscriber lines.

4. A method according to claim 1, wherein the first group of subscriber lines further comprises a subscriber line of the plurality of subscriber lines that supports non-linear precoding operation.

5. A method according to claim 4, wherein the addition of the subscriber line to the first group of subscriber lines is restricted to given carriers for which non-linear precoding induces a net penalty compared to linear precoding.

6. A method according to claim 5, wherein the method further comprises sending information indicative of the identity of the given carriers to a remote transceiver coupled to the subscriber line.

7. A method according to claim 1, wherein the scaling of the first signals is applied upfront before the first precoding stage.

8. A method according to claim 1, wherein the scaling of the first signals and the processing of the first signals through the first precoding stage are performed by means of a single matrix multiplication stage.

9. A method according to claim 1, wherein the scaling of the first signals and the processing of the first signals through the first and second precoding stages are performed by means of a single matrix multiplication stage.

10. A method according to claim 1, wherein the first signals are processed en bloc through the non-linear precoder.

11. A method according to claim 1, wherein the second precoding matrix is a unitary matrix.

12. A method according to claim 1, wherein the second precoding matrix is a full precoding matrix.

13. A non-linear precoder for jointly processing signals to be transmitted over respective ones of a plurality of subscriber lines, and comprising a first non-linear precoding stage configured to operate according to a first triangular precoding matrix and including a modulo unit, followed by a second linear precoding stage configured to operate according to a second precoding matrix, wherein the plurality of subscriber lines is organized into a first group of subscriber lines and a second group of subscriber lines, the first group of subscriber lines at least comprising the subscriber lines of the plurality of subscriber lines that do not support non-linear precoding operation, and the second group of subscriber lines comprising the remaining subscriber lines of the plurality of subscriber lines, wherein the non-linear precoder is further configured to scale first signals to be transmitted over respective ones of the first group of subscriber lines to confine respective first intermediate transmit power levels at the input of the modulo unit and further to bypass or make ineffective the operation of the modulo unit, and to process the so scaled first signals and second signals to be transmitted over respective ones of the second group of subscriber lines through the first and second precoding stages.

14. An access node comprising a non-linear precoder according to claim 13.

15. An access node according to claim 14, wherein the access node is a Distribution Point Unit DPU.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0070] The above and other objects and features of the invention will become more apparent and the invention itself will be best understood by referring to the following description of an embodiment taken in conjunction with the accompanying drawings wherein:

[0071] FIG. 1 represents an overview of an access plant;

[0072] FIGS. 2A, 2B and 2C represent different precoding matrices that can be used in a non-linear precoder as per the present invention; and

[0073] FIGS. 3A and 3B represent two different implementations of a non-linear precoder as per the present invention.

DETAILED DESCRIPTION OF THE INVENTION

[0074] There is seen in FIG. 1 an access plant 1 comprising a network unit 10 at a CO, an access node 20 coupled via one or more optical fibers to the network unit 10, and further coupled via a copper loop plant to Customer Premises Equipment (CPE) 30 at various locations. The transmission media is typically composed of copper Unshielded Twisted Pairs (UTP).

[0075] As an illustrative example, the loop plant comprises four subscriber lines L.sub.1 to L.sub.4 sharing a common access segment 40, and then going through dedicated loop segments 50 for final connection to CPEs 30.sub.1 to 30.sub.4 respectively.

[0076] Within the common access segment 40, the subscriber lines L.sub.1 to L.sub.4 are in close vicinity with each other and thus induce crosstalk into each other (see the arrows in FIG. 1 between the respective subscriber lines).

[0077] The access node 20 comprises a Vectoring Processing Unit 21 (or VPU) for jointly processing the data symbols that are being transmitted over, or received from, the loop plant in order to mitigate the crosstalk and to increase the communication data rates achievable over the respective subscriber lines.

[0078] The description will now focus on downstream communications, and further on the precoding of the downstream communication signals.

[0079] The VPU 21 implements NLP in order to enhance the vectoring gains in the presence of strong crosstalk. Yet, the CPEs 30 are not necessarily all NLP capable as this requires the use of a modulo unit within the CPE to recover the original transmit samples, as well as some further scaling logic. So there is assumed a mix of NLP capable and NLP non-capable CPEs. Presently, the CPEs 30.sub.1 and 30.sub.3 are NLP capable CPEs and support both linear and non-linear precoding operation, whereas the CPEs 30.sub.2 and 30.sub.4 are NLP non-capable CPEs supporting linear precoding operation only.

[0080] There is seen in FIG. 2A a functional illustration of the precoding and scaling matrices used in a non-linear precoder as per the present invention.

[0081] The mathematical notations that were introduced for the discussion of the prior art are still adhered to.

[0082] The subscriber lines are organized into a first group of subscriber lines G.sub.A comprising at least all NLP non-capable lines of the vectoring group, presently the subscriber lines L.sub.2 and L.sub.4 with regard to FIG. 1, and a second group of subscriber lines G.sub.B comprising the remaining NLP capable lines of the vectoring group, presently the subscriber lines L.sub.1 and L.sub.3 with regard to FIG. 1.

[0083] The first group G.sub.A may further comprise additional NLP capable lines for all carrier frequencies or for some specific carrier frequencies only. In the latter case, the group definition (and thus the precoding position) is frequency-dependent. This would be the case if NLP operation induces a net penalty compared LP operation on account of the increased modulation gap or on account of the increased average power, which is especially true for low-size constellation grids.

[0084] Let N denote the total number of subscriber lines in the vectoring group, and M the number of subscriber lines in the first group G.sub.A. The number M may be frequency-dependent if the group definition is so.

[0085] The subscriber lines of the first group G.sub.A are precoded first and are assigned the precoding indexes from 1 to M. The subscriber lines of the second group G.sub.B are precoded last and are assigned the precoding indexes from M+1 to N.

[0086] Let the matrices subscripts AA, AB, BA and BB denote the matrix sub-block decompositions with respect to the index ranges [1;M] and [M+1;N], or equivalently with respect to the groups G.sub.A and G.sub.B.

[0087] In a first step, gain scaling is applied to the transmit samples u.sub.1 to u.sub.M in order to confine the respective transmit power levels at the input of the modulo unit within a transmit power mask, and further in order to bypass the modulo operation or to make the modulo operation ineffective (i.e., the output value of the modulo unit matches its input value). The transmit samples u.sub.M+1 to U.sub.N are not scaled and are processed through the non-linear precoder in sequential order.

[0088] As Q is unitary and thus preserves norms and powers, applying gain scaling on P=QR*.sup.−1 corresponds with applying gain scaling on R*.sup.−1=L(diag(R*)).sup.−1. As an example of gain scaling, one may apply Column Norm (CN) scaling, whereby each and every column of the matrix R*.sup.−1 has substantially the same norm.

[0089] Owing to the line grouping, gain scaling is applied on the matrix (R*.sup.−1).sub.AA=L.sub.AA(diag(R*.sub.AA)).sup.−1 only, thereby increasing the gain values and thus the achievable data rates over the subscriber lines of the first group G.sub.A. This results in a first diagonal scaling matrix T.sub.AA comprising t.sub.1 to t.sub.M as respective scaling factors, and a second diagonal scaling matrix S.sub.AA.sup.−1=(diag(R*.sub.AA)).sup.−1 before processing through the matrix L.

[0090] The matrix T.sub.AA is such that (R*.sup.−1).sub.AAT.sub.AA=L.sub.AA(diag(R*.sub.AA)).sup.−1T.sub.AA has intermediate output power at or below the transmit power mask, or equivalently:

∥R.sup.*−1.sub.AA(i,*)t.sub.i∥=∥L.sub.AA(i,*)t.sub.i/r.sub.ii∥≦TXPSD for 1≦i≦M  (11),

wherein TXPSD denotes the applicable transmit power mask (omitting the frequency dependence k).

[0091] On account of this gain scaling and corresponding power confinement, the modulo function Γ can be bypassed or is made ineffective when the samples u.sub.1 to u.sub.M are processed through the non-linear precoding stage L+Γ (see hashed area in the modulo function Γ in FIG. 2). These transmit samples can thus be linearly processed en bloc through the successive matrices T.sub.AA, S.sub.AA.sup.−1 and L.sub.AA, thereby yielding partially-processed transmit samples x′.sub.1 to X′.sub.M.

[0092] In a second step, the transmit samples u.sub.M+1 to u.sub.N of the second group G.sub.B are kept unscaled and are processed in sequential order one after the other through the nonlinear precoding stage L+Γ as is typical for a non-linear precoder, thereby yielding partially-processed transmit samples x′.sub.M+1 to X′.sub.N. This second step can alternatively take place before the first step.

[0093] In a third and last step, all the partially-preprocessed transmit samples x′.sub.1 to x′.sub.N are passed through the linear precoding stage Q for further transmission over the channel H.

[0094] Two alternative embodiments are depicted in FIGS. 2B and 2C.

[0095] In the first alternative embodiment, the scaling matrices T.sub.AA and S.sub.AA.sup.−1 are merged with the matrix L, thereby yielding a new lower-triangular precoding matrix L′=[(R*.sup.−1).sub.AAT.sub.AA 0; (R*.sup.−1).sub.BAT.sub.AA L.sub.BB]. It is noteworthy that the M first diagonal elements of the matrix L′ are different from 1 on account of the gain scaling, while the N−M last diagonal elements of the matrix L′ are approximately equal to 1 (in practice, they need to be slightly smaller than 1 due to the fact that the modulo operation adds a little power).

[0096] In the second alternative embodiment, the scaling matrices T.sub.AA and S.sub.AA.sup.−1 and the upper part of the triangular matrix L are all merged with the matrix Q, thereby yielding a new precoding matrix Q′=[(R*.sup.−1).sub.AAT.sub.AAQ.sub.AA Q.sub.AB; (R*.sup.−1).sub.AAT.sub.AAQ.sub.BA Q.sub.BB], and the transmit samples u.sub.1 to u.sub.M are directly input to the linear precoding stage Q′. In this implementation, the transmit samples u.sub.M+1 to U.sub.N need to be processed first through the nonlinear precoding stage L′+Γ, thereby yielding the partially-processed samples x′.sub.M+1 to x′.sub.N, and next the transmit samples u.sub.1 to u.sub.M together with the partially processed samples x′.sub.M+1 to x′.sub.N are processed through the second linear precoding stage Q′.

[0097] This embodiment is particularly advantageous in that only two precoding steps are required, thereby saving substantial precoding resources.

[0098] There is seen in FIG. 3A a first possible implementation for a non-linear precoder as per the present invention.

[0099] A transmit vector U, which comprises the transmit frequency samples U.sub.A of the first group G.sub.A as top vector coefficients and the transmit frequency samples U.sub.B of the second group G.sub.B as bottom vector coefficients, is first input to a scaling stage 308 for multiplication with the scaling matrix [S.sub.AA.sup.−1T.sub.AA 0; 0 I]. The scaled transmit vector U′=[U.sub.A′; U.sub.B], wherein U.sub.A′=S.sub.AA.sup.−1T.sub.AAU.sub.A, is then fed to a first non-linear precoding stage 301.

[0100] The first non-linear precoding stage 301 uses the modulo function Γ, and a first lower-triangular precoding matrix M in the feedback loop. The matrix M is given by M=I−L.sup.−1=I−S.sup.−1R*, and has zero coefficients along its diagonal. The first precoding stage 301 implements non-linear precoding as per equations (7) and (8), and outputs a partially-precoded vector X′=[X.sub.A′; X.sub.B′] to the second linear precoding stage 302.

[0101] The second linear precoding stage 302 uses a second unitary or almost-unitary matrix Q, and outputs a fully-precoded vector X for further transmission over the MIMO channel 303, which is represented by a channel matrix H=DG=DR*Q*.

[0102] The noisy received vector Y comprises the frequency samples received through the respective communication channels corrupted by some additive white Gaussian noise source Z. The respective coefficients of the vector Y are either processed by NLP non-capable CPEs or by NLP capable CPEs.

[0103] In case of LP operation (see the top processing branch termed LP in FIG. 3A), the receive sample y.sub.i goes through the equalization stages 304 and 305. The first equalization stage 304, which corresponds to the typical FEQ, uses the respective diagonal coefficients h.sub.ii.sup.−1 of the diagonal matrix D.sup.−1 as determined during the training phase, and outputs a partially-equalized frequency sample y.sub.i′. The second equalization stage 305 uses the respective diagonal coefficients t.sub.i.sup.−1 of the diagonal matrix T.sub.AA.sup.−1 as communicated by the remote DPU by means of Transmitter Initiated Gain Adaptation (TIGA) command (see t.sub.i.sup.−1 arrow in FIG. 3A), and outputs a fully-equalized frequency sample y.sub.i″. The first and second equalization stages 304 and 305 can be merged into one single equalization stage. The properly equalized frequency sample y.sub.i″ is then fed to a decision stage 307 (or demodulator) to yield an estimate of the transmit sample u.sub.i.

[0104] In case of NLP operation (see the bottom processing branch termed NLP in FIG. 3A), the receive sample y.sub.i goes through the equalization stage 304 and another equalization stage 305′. The equalization stage 305′ uses the respective diagonal coefficients r.sub.jj.sup.−1 of the diagonal matrix S.sub.BB.sup.−1. Again the equalization stages 304 and 305′ can be merged into one single equalization stage. The properly equalized frequency sample y.sub.i″ goes through the modulo function Γ 306 to yield a possibly-modified frequency sample ŷ.sub.i for detection by the decision stage 307. The decision stage 307 outputs an estimate û.sub.i of the transmit sample u.sub.i.

[0105] An NLP capable CPE may follow the top processing branch for the carriers that have been precoded as forming part of the first group G.sub.A, or may follow the bottom processing branch provided t.sub.i.sup.−1 substitutes for r.sub.ii.sup.−1 in the block 305′ (then the modulo operation is expected to be transparent). The identity of those carriers can be obtained from the remote DPU, thereby allowing the CPE to apply specific modulation gaps and obtain more accurate bit loading values for those specific carriers.

[0106] There is seen in FIG. 3B a second possible implementation for a non-linear precoder as per the present invention and in line with the technical teaching of the aforementioned patent application EP2800283.

[0107] In this implementation, two new precoding blocks 301′ and 302′ substitute for the former precoding blocks 301 and 302 respectively.

[0108] The precoding stage 301′ is a modulo shift unit that adds a shift vector Δ to the scaled transmit vector U′ without any further signal precoding. The shift vector Δ is designed to keep the fully-precoded signal within the allowed power bound. The first precoding stage 301′ outputs the vector X′=U′+Δ to the second precoding stage 302′.

[0109] The precoding stage 301′ includes a modified modulo function γ, and a matrix N=L−I in the feedback loop. The modulo function is given by:

[00008]$\begin{array}{cc}{Y}_{i,k}\left(x_{i,k}\right)=-d.Math.M_{i,k}.Math..Math.\frac{x_{i,k}+d.Math.M_{i,k}/2}{d.Math.M_{i,k}}.Math.={\Gamma }_{i,k}\left(x_{i,k}\right)-x_{i,k},& \left(12\right)\end{array}$

and the shift vector Δ is given by:

[00009]${\delta }_i=Y_{i,k}\left(u_i^{\prime }+\underset{j=1}{\overset{i-1}{.Math.}}.Math..Math.l_{\mathrm{ij}}.Math.\left(u_j^{\prime }+{\delta }_j\right)\right),$

wherein u.sub.i′, δ.sub.i and l.sub.ij denote the coefficients of U′, Δ and L=R*.sup.−1S respectively.

[0110] The second precoding stage 302′ makes use of the full ZF precoding matrix P=QL, and outputs a fully precoded signal X=PX′ for further transmission over the MIMO channel 303.

[0111] The implementations depicted in FIGS. 3A and 3B fully leverage on the NLP architecture without the need for additional hardware or software logic for processing the transmit frequency samples U.sub.A of the first group G.sub.A (except the upfront scaling). Yet the transmit frequency samples U.sub.A can alternatively be processed through a separate branch including one or more feedforward filters without any modulo unit, for instance by using the successive feedforward filters T.sub.AAS.sub.AA.sup.−1, L and Q as per FIG. 2A, or L′ and Q as per FIG. 2B, or simply Q′ as per FIG. 2C. The transmit frequency samples U.sub.B of the second group G.sub.B keep on being processed as usual through the feedback filter M or N and the modulo unit Γ or γ.

[0112] With the proposed precoding scheme, both the LP and NLP lines have a performance exceeding that of a pure LP system. The performance of the LP lines is better than their performance when all lines would be LP as only a subset of the lines need to be scaled down. Because of being encoded last, NLP lines don't get ideal NLP gain but performance is still better than LP (ignoring power penalties inherent to NLP on account of the modulation gap increase or of the power increase due to uniform distribution of the constellation points) as their transmit power budget is larger than the average power budget in gain-scaled LP.

[0113] It is to be noticed that the term ‘comprising’ should not be interpreted as being restricted to the means listed thereafter. Thus, the scope of the expression ‘a device comprising means A and B’ should not be limited to devices consisting only of components A and B. It means that with respect to the present invention, the relevant components of the device are A and B.

[0114] It is to be further noticed that the term ‘coupled’ should not be interpreted as being restricted to direct connections only. Thus, the scope of the expression ‘a device A coupled to a device B’ should not be limited to devices or systems wherein an output of device A is directly connected to an input of device B, and/or vice-versa. It means that there exists a path between an output of A and an input of B, and/or vice-versa, which may be a path including other devices or means.

[0115] The description and drawings merely illustrate the principles of the invention. It will thus be appreciated that those skilled in the art will be able to devise various arrangements that, although not explicitly described or shown herein, embody the principles of the invention. Furthermore, all examples recited herein are principally intended expressly to be only for pedagogical purposes to aid the reader in understanding the principles of the invention and the concepts contributed by the inventor(s) to furthering the art, and are to be construed as being without limitation to such specifically recited examples and conditions. Moreover, all statements herein reciting principles, aspects, and embodiments of the invention, as well as specific examples thereof, are intended to encompass equivalents thereof.

[0116] The functions of the various elements shown in the figures may be provided through the use of dedicated hardware as well as hardware capable of executing software in association with appropriate software. When provided by a processor, the functions may be provided by a single dedicated processor, by a single shared processor, or by a plurality of individual processors, some of which may be shared. Moreover, a processor should not be construed to refer exclusively to hardware capable of executing software, and may implicitly include, without limitation, Digital signal Processor (DSP) hardware, network processor, Application specific Integrated circuit (ASIC), Field Programmable Gate Array (FPGA), etc. Other hardware, conventional and/or custom, such as Read Only Memory (ROM), Random Access Memory (RAM), and non volatile storage, may also be included.