2008

Université Catholique de
Louvain

Muhammad Hanif

[
Generalizing the SNR method for
]
A POSITIONING SYSTEM APPROACH
Several Methods to estimate the crosstalk channels have been proposed. Among them is the
method based on SNR measurements on each active line. This method uses perturbations
from a given line to all the other lines. The slightly modified SNR values are used to estimate
the crosstalk channel coefficients. In this report we pose the question of crosstalk channel
estimation utilizing SNR measurements as a positioning problem in a multi-dimensional space.
The solution of the system of simultaneous equations gives us the estimate of the precoder
matrix coefficients.
 Table of Contents
Abstract..............................................................................................................................3
Introduction......................................................................................................................3
Conclusions.........................................................................................................................3
References:.........................................................................................................................3

2
 Abstract
Multiuser coordination, or vectoring, has attracted a lot of research for the development
of next generation of DSL systems. Such systems possibly can transmit at hundreds of
megabits per second.
However, these coordination techniques require knowledge of direct as well as crosstalk
channels at the co-located modems usually located in the equipment called DSLAM.
These DSLAMs can be located inside the central office (CO) of the telephone company,
or in the street at a few hundred meters from the customer premises.

3
Introduction
Consider a DSL network of N lines. The channel matrix being C and precoder matrix
F, as usual. The total channel matrix will be,

( )( )
p p
c 11 ⋯ c1 N f 11 ⋯ f 1 N
C F p= ⋮ ⋱ ⋮ ⋮ ⋱ ⋮
c N 1 ⋯ c NN f Np 1 ⋯ f NN
p

( )
N N

∑ c1 x f p
x1 ⋯ ∑ c1 x f xNp
x=0 x=0
¿ ⋮ ⋱ ⋮
N N

∑ c Nx f px1 ⋯ ∑ c Nx f xNp
x=0 x=0

Now we can calculate the SNR on line j as,

|∑ |
n 2
p
c jx f xj σ 2u j
x=0
SNR pj =
∑ |∑ |
N n 2
p 2 2
c jx f xk σu + σn
k j
k=1 x=0
k≠ j
Or,

| |
N N 2

∑ ∑ c jx f p
xk σ 2u +σ 2n
k j
k=1 x=0
1
NSR pj = = k≠ j

|∑ |
p 2
SNR j N
p 2
c jx f xj σu j
x=0
N
2
∑ |c jj f pjk + c jk f pjj| σ 2u +σ 2n
k j
k=1
k≠ j
≈ 2
|c jj f pjj| σ 2u j

Here we have assumed the usual properties of the diagonal and off-diagonal elements of
the precoder and channel matrix, i.e. ,

f jj ≈ 1
c jk f jj ≈ c jk
c jk f jk ≪ c jk f jj +c jj f jk (used ∈numerator )
n

∑ c jk f kj ≪ c jj f jj (used ∈denominator)
k=0
k≠ j

The expression can be further simplified as,

4
 N
2
∑|c jj f pjk +c jk| σ 2u +σ 2n k j
k=1
p k≠ j
NSR ≈ j 2
|c jj| σ 2u j

| |
N 2 2 2
c jk σ u σ n
≈∑
p
f jk + + k j

k=1 c jj σ 2u σ 2u j j
k≠j
N 2 2
2 σ σ
≈ ∑ |f jk + f jk|
p uk nj
2
+ 2
k=1 σ uj σu j
k≠j
We see that this is a non-linear equation in 2(N-1) + 1 variables for each row of the
precoder F (that is, the real and imaginary part of the off-diagonal precoder coefficients
2
on each line, f jk , and noise term, σ n ). j

Suppose we have p = 0:1:P SNR measurements for different test precoders f pjk . We can
get a system of linear equations, for each row of precoder, by first writing the P+1
equations as,

n 2 2
σu σn
NSR =∑ |f
0
j
i=0
( | +|f
0 2
jk jk
2
| −2 f 0
jk , R f jk , R −2 f
0
jk , I f jk , I )σ 2
k
+
σ
2
j

uj uj
.
.
.

n 2 2
σu σn
NSR =∑ |f
P
j
i=0
( | +|f
P 2
jk jk
2
| −2 f P
jk ,R f jk , R−2 f
P
jk , I f jk , I )σ 2
k
+
σ
2
j

uj uj

Subtracting N SR 0j from each succeeding equation, gives the desired linear equations,

n 2
σu
ND =NSR −NSR =∑ |f
1
j
1
j
0
j
i=0
( | −|f | −2 f
1 2
jk
0 2
jk jk , R ∆
1
jk , R −2 f jk ,I ∆
1
jk , I )σ 2
k

uj
.
.
.
n 2
σu
ND =NSR −NSR =∑ |f
P
j
P
j
0
j
i=0
( | −|f | −2 f
P 2
jk
0 2
jk jk , R ∆
P
jk , R −2 f jk , I ∆
P
jk , I )σ 2
k

uj

Where,
p p o
∆ jk =f jk −f jk
Such that,
p p p
∆ jk =∆ jk , R +i∗∆ jk , I

5
For ease in computation we may further define,
2 2
FD pjk =|f pjk| −|f 0jk|
We can represent this linear system of equations in matrix form as, for j = 1,

( )(
N

f j2, R ∑ FD 1jk σ 2uk + σ 2uj ND

k=1
f j3, R k≠j

( )
1 2 1 2 1 2 1 2
∆ jN , R σ u ∆ j 2, I σ u ∆ j3, I σu ∆ jN ,I σ u ⋮ N

∑ FD 2jk σ 2uk + σ 2uj ND

1 2 1 2
∆ σ ∆ σ ⋯ P 2
N 2
…¿
3 N
j2, R u2 j 3 ,R u3
∆ j 3 ,R σ ¿…¿¿
u3 ⋮ ⋮ ⋮ ⋮ f jN , R =
⋮ ⋮ ¿ ¿… P k=1
∆ PjN , R σ 2u ∆ Pj 2, I σ 2u P 2
∆ j3, I σu ∆ jN ,I σ 2u f j2, I k≠ j
N 2 3 N
⋮
⋮ N
f jN , I ∑ FD Pjk σ 2uk +σ 2uj ND
k=1
k≠j

That is,
[ A ] P by2( N−1) [ f ] 2 ( N−1) by 1=[ A ] P by 1

Choice of Test precoder matrices

Chapter 2

6
Appendix

Equation 1

( )( )( )
C 11 ⋯ C 1 n F 11 ⋯ F1n C 1 x F x 1 ⋯ C 1 x F xn
⋮ ⋱ ⋮ ⋮ ⋱ ⋮ = ⋮ ⋱ ⋮
Cn1 ⋯ Cnn F n 1 ⋯ Fnn C nx F x1 ⋯ C nx F xn

( )
C11 ⋯ C1 n +C 11 F 1 n
≈ ⋮ ⋱ ⋮
Cn 1 +C nn F n 1 ⋯ C nn

Where we assume, F ii ≈ 1; C ij F ii ≈ Cij

And, C ij Fij ≪ C ii F ij

( )( )( )
C 11 ⋯ C 1 n C 11 ⋯ 0 0 ⋯ F1n
¿ ⋮ ⋱ ⋮ + ⋮ ⋱ ⋮ ⋮ ⋱ ⋮
Cn 1 ⋯ C nn 0 ⋯ C nn Fn 1 ⋯ 0

¿ C+ [ C ] d [ F ]nd

7
 [1].J.M.Cioffi. (June 2007). CuPON: The Copper Alternative to the PON 100 Gbps
DSL Networks. IEEE Communications Magazine .
[2].Louveaux, J., & van der veen, A.-J. (2006). Error Sign Feedback as an alternative
to Pilots for the tracking of FEXT transfer functions in Downstream VDSL.
EURASIP Journal on Applied Signal Processing .

References:
[1].Simon Haykin Communication Systems, 4th edition John Wiley and sons 2001
[2].T. Cover and J. Thomas, Elements of Information Theory ,Wiley, 1991
[3].John A.C. Bingham ADSL, VDSL and Multicarrier Modulation, Wiley, 2000
[4].R. Cendrillon Multi-user Signal and Spectra Co-ordination for Digital Subscriber
Lines, PhD Thesis, Katholieke Universiteit Leuven 2004
[5].R. Cendrillon et. al. A Near Optimal Linear Crosstalk Canceller for Upstream
VDSL IEEE Transactions on Signal Processing August 2006
[6].R.Cendrillon, G.Ginis, M.Moonen, A Near Optimal Linear Crosstalk Precoder
for DSL
[7].M. Moeneclaey, “On the true and the modified Cramer-Rao bounds for the
estimation of a scalar parameter in presence of nuisance parameters,” IEEE
Transactions on Communications, vol. 46, pp. 1536–1544, November 1998.