2012 06 04 dsl seminar tno brink v08 - joepeesoft.com€¦ · 2 progressing g.fast -dr. rob f.m....
Post on 16-Jul-2020
2 Views
Preview:
TRANSCRIPT
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium1
Progressing G.fastcontributions from 4GBB consortium to standards
Rob F.M. van den Brink – TNO
Chenguang Lu – Ericsson
Rob.vandenBrink@tno.nl
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium2
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
MUSE Consortium
1. Progressing G.fast – how did it originate?
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium3
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
4GBB ConsortiumCELTIC
- show feasibility- initiate a standard- bring the industry into motion
Aims:
MUSE
1. Progressing G.fast – how did it originate?
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium4
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
MUSE 4GBB ConsortiumCELTIC
2009 2010 2011 2012 2013 2014 2015
DSL seminar
San Francisco meeting
4GBB (up to 1 Gb/s)
FttH
Full
Hybrid
20-200m
Curb
Basement
Meter closet
FttH
FttH
(c) TNO
missing
1. Progressing G.fast – how did it originate?
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium5
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
MUSE 4GBB ConsortiumCELTIC
2009 2010 2011 2012 2013 2014 2015
2011 2012 2013 2014 2015
start of G.fast consent of G.fast
DSL seminar
San Francisco meeting
4GBB (up to 1 Gb/s)
FttH
Full
Hybrid
20-200m
Curb
Basement
Meter closet
FttH
FttH
(c) TNO
G.fast
1. Progressing G.fast – how did it originate?
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium6
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
MUSE 4GBB ConsortiumCELTIC
HFCC/G.fastCELTIC
2009 2010 2011 2012 2013 2014 2015
2011 2012 2013 2014 2015
Phase 1start of G.fast
DSL seminar
San Francisco meeting
Phase 2 - vectoring?Phase 3 - integration?
4GBB (up to 1 Gb/s)
FttH
Full
Hybrid
20-200m
Curb
Basement
Meter closet
FttH
FttH
(c) TNO
G.fast
1. Progressing G.fast – how will it continue?
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium7
1. Progressing G.fast – on what topics?
• Chapter 1 – Progressing G.fast• Chapter 2 – Progress on cable transmission modeling• Chapter 3 – Progress on crosstalk modeling• Chapter 4 – Progress on performance studies
e.g. by developing knowledge on cable modelling & simulations
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium8
1. Progressing G.fast – on what topics?
aimed service rates (US+DS rates) under noisy conditions from 8 disturbersnote: different service rate definitions are under discussion • agreed: at least 10Mb/s when deployed on loops with exceptional length and noise• agreed: 150 Mb/s @ 250m• agreed: 200 Mb/s @ 200m• agreed: 500 Mb/s @ 100m (typically a FTTB deployment) à 2×500 Mb/s with bonding• agreed: 500 Mb/s @ 50m when above 17.7MHz, and not too bad noise• agreed: up to 1Gb/s on shorter loops (protocol should not prevent 1Gb/s)
e.g. by developing knowledge on cable modelling & simulations
Bitrate targets for G.fast, as agreed within ITU
• Chapter 1 – Progressing G.fast• Chapter 2 – Progress on cable transmission modeling• Chapter 3 – Progress on crosstalk modeling• Chapter 4 – Progress on performance studies
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium9
2. Progress on cable transmission modeling
Z ×dxs
Y ×dxp
dx
( )( ))()(
)()(ωωωω
jYrealjGjZrealjR
pp
ss==
( )( ) ωωω
ωωω/)()(/)()(
jYimagjCjZimagjL
pp
ss==
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium10
2. Progress on cable transmission modeling
Z ×dxs
Y ×dxp
dx
( )( ))()(
)()(ωωωω
jYrealjGjZrealjR
pp
ss==
( )( ) ωωω
ωωω/)()(/)()(
jYimagjCjZimagjL
pp
ss==
1k 10k 100k 1M 10M 100M0
0.5
1
1.5
2
[freq]
[ohm/m]
series resistance: Rs (per unit length)
(c) TNO 2011
1k 10k 100k 1M 10M 100M0
200
400
600
800
1k
[freq]
[nH/m]
series inductance: Ls (per unit length)
(c) TNO 2011
1k 10k 100k 1M 10M 100M0
20
40
60
80
100
[freq]
[pF/m]
shunt capacitance: Cp (per unit length)
(c) TNO 2011
1k 10k 100k 1M 10M 100M0
5u
10u
15u
20u
[freq]
[mho/m]
shunt conductance: Gp (per unit length)
(c) TNO 2011
Rs
Cp Gp
Ls
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium11
Z ×dxs
Y ×dxp
dx
( )( ))()(
)()(ωωωω
jYrealjGjZrealjR
pp
ss==
( )( ) ωωω
ωωω/)()(/)()(
jYimagjCjZimagjL
pp
ss==
1k 10k 100k 1M 10M 100M0
0.5
1
1.5
2
[freq]
[ohm/m]
series resistance: Rs (per unit length)
(c) TNO 2011
1k 10k 100k 1M 10M 100M0
200
400
600
800
1k
[freq]
[nH/m]
series inductance: Ls (per unit length)
(c) TNO 2011
1k 10k 100k 1M 10M 100M0
20
40
60
80
100
[freq]
[pF/m]
shunt capacitance: Cp (per unit length)
(c) TNO 2011
1k 10k 100k 1M 10M 100M0
5u
10u
15u
20u
[freq]
[mho/m]
shunt conductance: Gp (per unit length)
(c) TNO 2011
Rs
Cp Gp
Ls
Legacy approach (see VDSL2 standard)Model Rs, Ls, Cp, Gp independently, using 4 functions of “w”(BT Model (1996): 13 parameters, KPN Model (1997), etc)modification needed to make impulse response causalproblem: this troubles time-domain simulations
Improved approach (TNO model, 2009/2011):Model (Rs+jw.Ls) and (Gp+jw.Cp), using 2 functions of “jw”(7+1 parameters)you get causality almost for “free” (if done in a clever way)
2. Progress on cable transmission modeling
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium12
2. Progress on cable transmission modeling
Z ×dxs
Y ×dxp
dx
( )( ))()(
)()(ωωωω
jYrealjGjZrealjR
pp
ss==
( )( ) ωωω
ωωω/)()(/)()(
jYimagjCjZimagjL
pp
ss==
1k 10k 100k 1M 10M 100M0
0.5
1
1.5
2
[freq]
[ohm/m]
series resistance: Rs (per unit length)
(c) TNO 2011
1k 10k 100k 1M 10M 100M0
200
400
600
800
1k
[freq]
[nH/m]
series inductance: Ls (per unit length)
(c) TNO 2011
1k 10k 100k 1M 10M 100M0
20
40
60
80
100
[freq]
[pF/m]
shunt capacitance: Cp (per unit length)
(c) TNO 2011
1k 10k 100k 1M 10M 100M0
5u
10u
15u
20u
[freq]
[mho/m]
shunt conductance: Gp (per unit length)
(c) TNO 2011
Rs
Cp Gp
Ls
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium13
2. Progress on cable transmission modeling - expressions for impedance Zs
1k 10k 100k 1M 10M 100M0
0.5
1
1.5
2
[freq]
[ohm/m]
series resistance: Rs (per unit length)
(c) TNO 2011
1k 10k 100k 1M 10M 100M0
200
400
600
800
1k
[freq]
[nH/m]
series inductance: Ls (per unit length)
(c) TNO 2011
Rs Ls
Z ×dxs
Y ×dxp
dx
0sL
∞sL
0sR
ssR ωω×0
( )( ) ωωω
ωωω/)()(/)()(
jYimagjCjZimagjL
pp
ss==
( )( ))()(
)()(ωωωω
jYrealjGjZrealjR
pp
ss==
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium14
2. Progress on cable transmission modeling - expressions for impedance Zs
1k 10k 100k 1M 10M 100M0
0.5
1
1.5
2
[freq]
[ohm/m]
series resistance: Rs (per unit length)
(c) TNO 2011
1k 10k 100k 1M 10M 100M0
200
400
600
800
1k
[freq]
[nH/m]
series inductance: Ls (per unit length)
(c) TNO 2011
Rs Ls
Z ×dxs
Y ×dxp
dx
( ) sssss qRLL //00 ω+= ∞
⋅⋅+−×+⋅= ∞
ssZsssss q
jQqqRLjjZω
ωωω 20 1)(
ssss qRL ,,, 0 ω∞
substitution
parameters
)( ΩjQZ glue (shaping function)→
Approach: take observations as parameters
0sR
ssR ωω×0
0sL
∞sL
Qz is restricted by certain analytical properties
( )( ) ωωω
ωωω/)()(/)()(
jYimagjCjZimagjL
pp
ss==
( )( ))()(
)()(ωωωω
jYrealjGjZrealjR
pp
ss==
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium15
2. Progress on cable transmission modeling - expressions for impedance Zs
1k 10k 100k 1M 10M 100M0
0.5
1
1.5
2
[freq]
[ohm/m]
series resistance: Rs (per unit length)
(c) TNO 2011
1k 10k 100k 1M 10M 100M0
200
400
600
800
1k
[freq]
[nH/m]
series inductance: Ls (per unit length)
(c) TNO 2011
Rs Ls
Z ×dxs
Y ×dxp
dx
( ) sssss qRLL //00 ω+= ∞
⋅⋅+−×+⋅= ∞
ssZsssss q
jQqqRLjjZω
ωωω 20 1)(
ssss qRL ,,, 0 ω∞
substitution
parameters
)( ΩjQZ glue (shaping function)→
Approach: take observations as parameters
0sR
ssR ωω×0
0sL
∞sL
Qz is restricted by certain analytical properties
( )( ) ωωω
ωωω/)()(/)()(
jYimagjCjZimagjL
pp
ss==
( )( ))()(
)()(ωωωω
jYrealjGjZrealjR
pp
ss==
⋅++−×+⋅= ∞
ssssss
jqsqrtqRLjjZωω
ωω 21)( 20The simplest form of Qz is already very good
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium16
2. Progress on cable transmission modeling - expressions for admittance Yp
Z ×dxs
Y ×dxp
dx
( )( ))()(
)()(ωωωω
jYrealjGjZrealjR
pp
ss==
( )( ) ωωω
ωωω/)()(/)()(
jYimagjCjZimagjL
pp
ss==
1k 10k 100k 1M 10M 100M0
20
40
60
80
100
[freq]
[pF/m]
shunt capacitance: Cp (per unit length)
(c) TNO 2011
1k 10k 100k 1M 10M 100M0
5u
10u
15u
20u
[freq]
[mho/m]
shunt conductance: Gp (per unit length)
(c) TNO 2011
Cp Gp
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium17
2. Progress on cable transmission modeling - expressions for admittance Yp
Z ×dxs
Y ×dxp
dx
1k 10k 100k 1M 10M 100M0
20
40
60
80
100
[freq]
[pF/m]
shunt capacitance: Cp (per unit length)
(c) TNO 2011
1k 10k 100k 1M 10M 100M0
5u
10u
15u
20u
[freq]
[mho/m]
shunt conductance: Gp (per unit length)
(c) TNO 2011
Cp Gp
0pC
0→pG
( )( ) ωωω
ωωω/)()(/)()(
jYimagjCjZimagjL
pp
ss==
( )( ))()(
)()(ωωωω
jYrealjGjZrealjR
pp
ss==
angle lossconstant ≈=⋅
)tan()(
)(φ
ωω
ω
p
pC
G
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium18
2. Progress on cable transmission modeling - expressions for admittance Yp
Z ×dxs
Y ×dxp
dx
1k 10k 100k 1M 10M 100M0
20
40
60
80
100
[freq]
[pF/m]
shunt capacitance: Cp (per unit length)
(c) TNO 2011
1k 10k 100k 1M 10M 100M0
5u
10u
15u
20u
[freq]
[mho/m]
shunt conductance: Gp (per unit length)
(c) TNO 2011
Cp Gp
parameters
glue (shaping function) →Approach: take observations as parameters
Qy is restricted by certain analytical properties
dpC ωφ,,0
)( ωjQY
φ
ωω
ωω
×⋅=
dYpp
jQCjjY 0)(
0pC
0→pG
angle lossconstant ≈=⋅
)tan()(
)(φ
ωω
ω
p
pC
G
( )( ) ωωω
ωωω/)()(/)()(
jYimagjCjZimagjL
pp
ss==
( )( ))()(
)()(ωωωω
jYrealjGjZrealjR
pp
ss==
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium19
2. Progress on cable transmission modeling - expressions for admittance Yp
Z ×dxs
Y ×dxp
dx
1k 10k 100k 1M 10M 100M0
20
40
60
80
100
[freq]
[pF/m]
shunt capacitance: Cp (per unit length)
(c) TNO 2011
1k 10k 100k 1M 10M 100M0
5u
10u
15u
20u
[freq]
[mho/m]
shunt conductance: Gp (per unit length)
(c) TNO 2011
Cp Gp
parameters
glue (shaping function) →Approach: take observations as parameters
Qy is restricted by certain analytical properties
dcp qC ωφ ,,,0
)( ωjQY
0pC
0→pG
cpcd
Ypp qCjqjQCjjY ×⋅+−×
×⋅= 00 )1()( ω
ωω
ωωφ
Refinement from Ericsson+LTHcontributed to last ITU meeting
( )( ) ωωω
ωωω/)()(/)()(
jYimagjCjZimagjL
pp
ss==
( )( ))()(
)()(ωωωω
jYrealjGjZrealjR
pp
ss==
Relevant when dielectric loss becomes significant
angle lossdependent frequency==⋅
)tan()(
)(φ
ωω
ω
p
pC
G
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium20
2. Progress on cable transmission modeling - expressions for admittance Yp
Z ×dxs
Y ×dxp
dx
1k 10k 100k 1M 10M 100M0
20
40
60
80
100
[freq]
[pF/m]
shunt capacitance: Cp (per unit length)
(c) TNO 2011
1k 10k 100k 1M 10M 100M0
5u
10u
15u
20u
[freq]
[mho/m]
shunt conductance: Gp (per unit length)
(c) TNO 2011
Cp Gp
parameters
glue (shaping function) →Approach: take observations as parameters
Qy is restricted by certain analytical properties
dcp qC ωφ ,,,0
)( ωjQY
0pC
0→pG
( )( ) ωωω
ωωω/)()(/)()(
jYimagjCjZimagjL
pp
ss==
( )( ))()(
)()(ωωωω
jYrealjGjZrealjR
pp
ss==
Relevant when dielectric loss becomes significant
angle lossdependent frequency≈=⋅
)tan()(
)(φ
ωω
ω
p
pC
G
cpcd
Ypp qCjqjQCjjY ×⋅+−×
×⋅= 00 )1()( ω
ωω
ωωφ
Refinement from Ericsson+LTHcontributed to last ITU meeting
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium21
2. Progress on cable transmission modeling - expressions for admittance Yp
Z ×dxs
Y ×dxp
dx
1k 10k 100k 1M 10M 100M0
20
40
60
80
100
[freq]
[pF/m]
shunt capacitance: Cp (per unit length)
(c) TNO 2011
1k 10k 100k 1M 10M 100M0
5u
10u
15u
20u
[freq]
[mho/m]
shunt conductance: Gp (per unit length)
(c) TNO 2011
Cp Gp
parameters
glue (shaping function) →Approach: take observations as parameters
Qy is restricted by certain analytical properties
dcp qC ωφ ,,,0
)( ωjQY
0pC
0→pG
( )( ) ωωω
ωωω/)()(/)()(
jYimagjCjZimagjL
pp
ss==
( )( ))()(
)()(ωωωω
jYrealjGjZrealjR
pp
ss==
Relevant when dielectric loss becomes significant
angle lossdependent frequency≈=⋅
)tan()(
)(φ
ωω
ω
p
pC
G
cpcd
Ypp qCjqjQCjjY ×⋅+−×
×⋅= 00 )1()( ω
ωω
ωωφ
cpcd
pp qCjqjCjjY ×⋅+−×
+×⋅=
⋅−
0
/2
0 )1(1)( ωω
ωωω
πφ
The simplest form of Qy is already very good
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium22
2. Progress on cable transmission modeling - results
1k 10k 100k 1M 10M 100M 1G-30
-25
-20
-15
-10
-5
0
[freq]
[dB]:100Ω
abs(S21), len=61.15m, model=[ASYMP.sqrt], WirePair=(1-5|2-6)
(c) TNO 2011
Measured S21
Measured S12
modelled v alue
1k 10k 100k 1M 10M 100M 1G-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0.2
[freq]
[dB]
Relative error in S21: dB(S21./S21m), model=[ASYMP.sqrt], WirePair=(1-5|2-6)
(c) TNO 2011
CAT5e cabling
61m
232m
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium23
CAT5e cabling
61m
232m
1k 10k 100k 1M 10M 100M0
200
400
600
800
1k
[freq]
[ohm]
Characteristic impedance Z0, model=[ASYMP.sqrt], WirePair=(1-5|2-6)
(c) TNO 2011
extracted v alue (ignored)
extracted v alue (used)
modelled v alue
1k 10k 100k 1M 10M 100M 1G
-80
-60
-40
-20
0
20
[freq]
[degree]
Characteristic impedance Z0, model=[ASYMP.sqrt], WirePair=(1-5|2-6)
(c) TNO 2011
extracted v alue (ignored)
extracted v alue (used)
modelled v alue
2. Progress on cable transmission modeling - results
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium24
PVC insulated drop cable
Dielectric losses become significant;
Use of proposed refinement from Ericsson and LTH offer valuable improvements
2. Progress on cable transmission modeling - results
24m
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium25
2. Progress on cable transmission modeling - results
New approach
has gained acceptance in ITU
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium26
3. Progress on crosstalk modeling - EL-FEXT, legacy approach
Signal
Noise (FEXT)
EL - FEXT = Signal
Noise (FEXT)S/N ratio
= 1
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium27
3. Progress on crosstalk modeling - EL-FEXT, legacy approach
10-2
10-1
100
101
102
103
-120
-100
-80
-60
-40
-20
0
20
freq:[MHz]
[dB]
Wire-pair EL-FEXT crosstalk (diff mode, CKT1:ETSI model) - (KPN, 100m, in-quad #2)
(c) TNO 2011
Signal
Noise (FEXT)
EL - FEXT = Signal
Noise (FEXT)S/N ratio
= 1
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium28
Legacy model: (FSAN)EL-FEXT is linear proportional with the frequency.
Too low for higher frequencies
3. Progress on crosstalk modeling - EL-FEXT, legacy approach
Performance studies for G.fast will be too optimistic when this is ignored
10-2
10-1
100
101
102
103
-120
-100
-80
-60
-40
-20
0
20
freq:[MHz]
[dB]
Wire-pair EL-FEXT crosstalk (diff mode, CKT1:ETSI model) - (KPN, 100m, in-quad #2)
(c) TNO 2011
ETSI, Near worst case level
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium29
KPN Access cable, 100m, in-quad, group 2
3. Progress on crosstalk modeling - EL-FEXT with dual slope
10-2
10-1
100
101
102
103
-140
-120
-100
-80
-60
-40
-20
0
20
freq:[MHz]
[dB]
Wire-pair EL-FEXT crosstalk (diff mode) - (KPN, 100m, in-quad #2)
(c) TNO 2012
10-2
10-1
100
101
102
-140
-120
-100
-80
-60
-40
-20
0
20
freq:[MHz]
[dB]
Wire-pair FEXT crosstalk (diff mode, transfer) - (KPN, 100m, in-quad #2)
(c) TNO 2012
dif f mode FEXTdif f mode transf er
fq ×1
22 fq ×
MHz20
104m
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium30
KPN Access cable, 100m, out-of-quad
3. Progress on crosstalk modeling - EL-FEXT with dual slope
10-2
10-1
100
101
102
103
-140
-120
-100
-80
-60
-40
-20
0
20
freq:[MHz]
[dB]
Wire-pair EL-FEXT crosstalk (diff mode) - (KPN, 100m, out of quad)
(c) TNO 2012
10-2
10-1
100
101
102
103
-140
-120
-100
-80
-60
-40
-20
0
20
freq:[MHz]
[dB]
Wire-pair FEXT crosstalk (diff mode, transfer) - (KPN, 100m, out of quad)
(c) TNO 2012
dif f mode FEXTdif f mode transf er
MHz12
fq ×1
22 fq ×
104m
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium31
Multi quad telephony cabling, 378m, in-quad, group 1
3. Progress on crosstalk modeling - EL-FEXT with dual slope
10-3
10-2
10-1
100
101
102
-140
-120
-100
-80
-60
-40
-20
0
20
freq:[MHz]
[dB]
Wire-pair EL-FEXT crosstalk (diff mode) - (TelCable/30q, 378m, in-quad, gw1|br1)
(c) TNO 2012
10-3
10-2
10-1
100
101
102
-140
-120
-100
-80
-60
-40
-20
0
20
freq:[MHz]
[dB]
Wire-pair FEXT crosstalk (diff mode, transfer) - (TelCable/30q, 378m, in-quad, gw1|br1)
(c) TNO 2012
dif f mode FEXTdif f mode transf er
This cannot be ignored for even VDSL frequencies!(roughly 1.5 MHz)
MHz5.1
378 m
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium32
Low quality telephony cable found in consumer shop (Gamma), 50m
3. Progress on crosstalk modeling - EL-FEXT with dual slope
10-3
10-2
10-1
100
101
102
103
-120
-100
-80
-60
-40
-20
0
20
freq:[MHz]
[dB]
Wire-pair EL-FEXT crosstalk (diff mode) - (Unshielded "Gamma", 50m)
(c) TNO 2012
10-3
10-2
10-1
100
101
102
103
-120
-100
-80
-60
-40
-20
0
20
freq:[MHz]
[dB]
Wire-pair FEXT crosstalk (diff mode, transfer) - (Unshielded "Gamma", 50m)
(c) TNO 2012
dif f mode FEXTdif f mode transf erCrosstalk is soo
high, dual slope effect not present
51m
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium33
3. Progress on crosstalk modeling - EL-FEXT via 2nd-order transfer
10-3
10-2
10-1
100
101
102
-140
-120
-100
-80
-60
-40
-20
0
20
40
freq:[MHz]
[dB]
Wire-pair EL-FEXT crosstalk (diff mode) - (TelCable/30q, 378m, in-quad, gw1|br1)
(c) TNO 2012
Models dual-slope effect very well, …
… models even the resonation peak, …
How to model this?
à Dual slope = 2nd order transfer function
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium34
3. Progress on crosstalk modeling - EL-FEXT via 2nd-order transfer
10-3
10-2
10-1
100
101
102
-140
-120
-100
-80
-60
-40
-20
0
20
40
freq:[MHz]
[dB]
Wire-pair EL-FEXT crosstalk (diff mode) - (TelCable/30q, 378m, in-quad, gw1|br1)
(c) TNO 2012
+
⋅+
+×
=⋅×
− 2
22
22
1
1
/21
1)(
ωω
ωω
ωωωω
ωω
ωjj
jj
FEXTELq
jH
Models dual-slope effect very well, …
… models even the resonation peak, …
10-3
10-2
10-1
100
101
102
-100
-50
0
50
100
150
200
250
300
freq:[MHz]
[deg]
Wire-pair EL-FEXT crosstalk (diff mode) - (TelCable/30q, 378m, in-quad, gw5|br5)
(c) TNO 2012
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium35
3. Progress on crosstalk modeling - EL-FEXT via 2nd-order transfer
Models dual-slope effect sometimesvery well, …
… and also the phase (in quad cables)
+
⋅+
+×
=⋅×
− 2
22
22
1
1
/21
1)(
ωω
ωω
ωωωω
ωω
ωjj
jj
FEXTELq
jH
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium36
3. Progress on crosstalk modeling - EL-FEXT via 2nd-order transfer
New approach
is gaining acceptance in ITU
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium37
4. Progress on performance studies – approach
Traditional approach: Frequency domain calculations› How:
– Calculate SNR from cable models and noise assumptions (from frequency domain calculations)
– Run bit loading algorithm with the SNR– Remove overhead of Cyclic Extension (prefix + suffix) and coding.
› Advantages– Simple – Adequate for channels dominated with additive white Gaussian
noise (AWGN)
GOAL: Predict attainable bitrateswithin a given scenario(Important in system design phase)
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium38
4. Progress on performance studies – approach
Advanced approach: Time domain calculations› Why: G.fast channel can be more ISI limited, especially with short
cyclic prefix,• High tone spacing -> short symbol period• Long dispersion due to bridge taps• CP may not cover the whole channel dispersion.
› How: Take inter-symbol interference (ISI) into consideration for estimating Signal to Noise and Interference Ratio estimation (SINR)
– Calculate that via impulse response of (improved) cable model• Estimate SINR from a time domain simulation from training symbols
Implementation of symbol timing detection (or frame detection) and Frequency domain equalizer (FEQ) are needed.
• Upsampling is needed to reduce windowing effects of Inverse Discrete Fourier Transform (IDFT) when generating impulse noise
GOAL: Predict attainable bitrateswithin a given scenario(Important in system design phase)
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium39
4. Progress on performance studies – approach
0 20 40 60 80 10010
15
20
25
30
35
40
45
50
55
60
Frequency (MHz)
dB
Time-domain: AWG 24, 100 m, CP=0.8 µs Frequency-domain: AWG 24, 100 mTime-domain: AWG 24, 200 m, CP=0.8 µs Frequency-domain: AWG 24, 200 m
SNR comparison: time vs. frequency domain approach
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium40
4. Progress on performance studies – initial studies
Term.DP
5 m
CA
D55
10 mCAD55
≤ 240 mCAD55
Term.DP30
mAW
G 2
6
45 mAWG 26
≤ 205 mAWG 26
22.5
mAW
G 2
6
15 m
AWG
26
7.5
mAW
G 2
6
3.8
mAW
G 2
6
0 m 0 m 0 m 0 m
Term.DP ≤ 250 mCAD55
Term.DP ≤ 250 mAWG 26
Topology 1
Topology 2
Topology 3
Topology 4
Zs = 100 ohmZt = 100 ohm
Initial studies: Selected loop topologies
References:• Loop topology: “G.fast: Wiring topologies and reference loops”, (ITU 11GS3-100)• CAD55 model: BT, “G.Fast: Cable Models ”, (ITU 11RV-026)• AWG 26 model: ANSI TP1 (ANSI T1.417-2003)
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium41
4. Progress on performance studies – initial studies
Initial studies: Simulation assumptions
• Channel bandwidth: 200 MHz• Channel is filtered with 53-tap FIR filter to reduce ringing
from windowing effects of IDFT.• Tone spacing: 48.84 KHz• Data Bandwidth: 3.5 - 100 MHz• Number of tones: 2048
• Tone 0 is DC. Tones below 3.5 MHz are not loaded.• FFT size: 8192. FFT oversampling is used for fitting
channel bandwidth. • Signal PSD: -76 dBm/Hz• AWGN noise: -140 dBm/Hz• Coding: RS + Trellis
• RS: N/R = 255/16.• Trellis: 0.5 bit/tone overhead.• Total coding gain = 5 dB.
• SNR gap: 9.75 dB• SNR margin: 6 dB• Bit loading: 1-12 bits• Cyclic Extension: 0.4 µs, 0.8 µs, 1.2 µs, 1.6 µs, 2.0 µs, 2.4 µs
(Prefix+Suffix)• Loop length (DP to terminal): 50 m, 100 m, 150 m, 200 m
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium42
4. Progress on performance studies – initial studies
Term.DP
5 m
CA
D55
10 mCAD55
≤ 240 mCAD55
Term.DP ≤ 250 mCAD55Topology 1
Topology 2
Initial studies: Bitrate results
• Topology 1&2: Optimal CP = 0.8 µs, CP overhead = 3.76%• Up to about 100 Mbps decrease due to a 5-meter bridge tap.
Topology 1 Topology 2
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium43
4. Progress on performance studies – initial studies
Initial studies: Bitrate results
• Topology 3: optimal CP = 0.4 or 0.8 µs, CP overhead = 1.88% or 3.76%.• Topology 4: optimal CP = 2 µs, CP overhead = 8.9%.• More than 300 Mbps rate loss can be seen due to multiple bridge taps of star topology.
Topology 3 Topology 4
Term.DP
30 m
AW
G 2
6
45 mAWG 26
≤ 205 mAWG 26
22.5
mA
WG
26
15 m
AW
G 2
6
7.5
mA
WG
26
3.8
mA
WG
26
0 m 0 m 0 m 0 m
Term.DP ≤ 250 mAWG 26Topology 3
Topology 4
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium44
5. Conclusions
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium45
G.fast standardization is progressing very well• Concept of Hybrid FttH via G.fast Initiated by 4GBB consortium• global industry is now active in ITU to develop G.fast• Aimed ITU deadline (dec 2012: consent on G.fast) is just a first phase
5. Conclusions
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium46
G.fast standardization is progressing very well• Concept of Hybrid FttH via G.fast Initiated by 4GBB consortium• global industry is now active in ITU to develop G.fast• Aimed ITU deadline (dec 2012: consent on G.fast) is just a first phase
Improved models for performance simulations under development• New cable model gives causality “for free”• Dual slope FEXT model is very promising• ITU is adopting these models for performance studies
5. Conclusions
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium47
G.fast standardization is progressing very well• Concept of Hybrid FttH via G.fast Initiated by 4GBB consortium• global industry is now active in ITU to develop G.fast• Aimed ITU deadline (dec 2012: consent on G.fast) is just a first phase
Improved models for performance simulations under development• New cable model gives causality “for free”• Dual slope FEXT model is very promising• ITU is adopting these models for performance studies
Initial performance studies are demonstrating the potentials for G.fast• Bitrates above 800 Mb/s over 100m cable• Even bridge taps can be handled, albeit with performance loss
5. Conclusions
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium48
G.fast standardization is progressing very well• Concept of Hybrid FttH via G.fast Initiated by 4GBB consortium• global industry is now active in ITU to develop G.fast• Aimed ITU deadline (dec 2012: consent on G.fast) is just a first phase
Improved models for performance simulations under development• New cable model gives causality “for free”• Dual slope FEXT model is very promising• ITU is adopting these models for performance studies
Initial performance studies are demonstrating the potentials for G.fast• Bitrates above 800 Mb/s over 100m cable• Even bridge taps can be handled, albeit with performance loss
5. Conclusions
Cable models have been developed for hundreds of MHz,so use them to describe your cable plant
DSL Seminar, June 4-6, 2012, The NetherlandsProgressing G.fast - dr. Rob F.M. van den Brink - TNO / 4GBB consortium49
top related