Part IV – Wireline Multiuser Examples
Prof. John M. Cioffi
Dept of EE
Stanford University
[email protected]
March 31, 2001
April 26 2001
September 9, 2001
1
Multiuser Wonders
Parts 3 and 4: Outline/Schedule
2:00-2:45
 2:45-3:30
 3:30-4:00
 4:00-4:30
 4:30-5:15

MU Theory
channels for wireline
Coffee
DSL and Ethernet arch
Multiuser improvements
2
Multiuser Wonders
Outline – Part 4
DSL and Ethernet Applications
 Evolution of DSL and DSM vs SM

 Unbundling
 Towards
ethernet
Multiuser Spectrum Balancing
 Vectored VDSL
 Vectored Ethernet

3
Multiuser Wonders
DSL in 2001
digital
Internet
Service
Provider
Voice
Service
Provider
G
A
T
E
W
A
Y
S
ADSL
modem
Mux
or
Demux
analog
ADSL
modem
split
.
.
Splitter ?
ADSL
ADSL
modem
analog
0-4 miles
DSLAM
digital
POTS
Network
customer premises
Telephone company office


Data, some voice beginning, to 1.5/.5 Mbps
Conservative deployments (9 Million lines)
4
Multiuser Wonders
DSL Future Applications
To
service
provider
A/VDSL
router
Enet.
100BT
Gigabit E
Home/business wiring
VoDSL
interface
Enet
mfilt
802.11


Multiline voice (VoDSL)
Video (internet, packet based)
 Not

wireless (802.11) or
other distribution possible
within CPE
likely “traditional” TV
Small Businesses – less asymmetric
5
Multiuser Wonders
Part Fiber, Part DSL = VDSL
To 100 Mbps
ONU
fiber
VDSL
POTS
C-Bank
S
p
l
i
t
S
p
l
i
t
.1- 2 km
LT

Only 12,000 of 107 businesses connected by fiber all
the way – essentially 0 residences


VDSL
Rather ONU’s/remote terminals, incrementally with time
How long is fiber? Twisted-Pair?

Incremental trade-off with time/demand
6
Multiuser Wonders
Ethernet – 10BT
Hub
(yellow
coax)
Server
Internet
Router

STAR wiring – collision detection not really used



100’s millions deployed on copper




Hub is electronic version of a “coax-wired” connection
All lines on Hub share the 10 Mbps
Manchester coding
Category 5 tp (20 dB less xtalk than phone lines- cat 3, 24 gauge)
Each line is 2 pairs (so 2 of 4 in cat 5)
100 meters (systems run longer)

Result of delay specification related to days in which HUB was a coax.
7
Multiuser Wonders
Introduction of Bridges
Hub
bridge
Hub
Server
Internet
Router



Little higher level than Hub, which is an electronic
version of a “yellow coax”
Allows each line’s 10 Mbps to be different
Dedicated connection to each user
8
Multiuser Wonders
Ethernet – 100BT
Server
Bridge
(router)
Internet



Each user gets 100 Mbps, dedicated link
MLT3 code (4B5B to ternary) – 125 MHz
100 meter range – back compatible with 10BT




Data sent when packets available
Delay spec really not needed on link anymore
Remains a 2-wire duplex situation
Lead suppliers do 170m range
9
Multiuser Wonders
Ethernet – 1000BT
4
Server
Router
Internet

Continuous, 250 Mbps on each of 4 lines for 1 Gbps
total




Continuous only
All 4 wires in cat 5 used by a single user in duplex, echo-cancelled
fashion
125 MHz 5-level (effective 4, or 2 bits) PAM on each line
100 meters – physical layer constraint (lead suppliers do 160 m)
10
Multiuser Wonders
Ethernet Faster?
VDMT
4
VDMT
Router VDMT
Server
VDMT
VDMT
VDMT
VDMT
VDMT
VDMT
Internet

10,000BT – vectored multiuser, 100
meters
11
Multiuser Wonders
Ethernet – longer (EFM)
Router
DSLAM
VDMT
4
VDMT
Server
VDMT
VDMT
VDMT
VDMT
VDMT
VDMT
VDMT
Sites may now
be at homes/businesses
at end of telco lines
Internet




VDSL=VDMT
100BT (4 wires) > 1 km
Single wire – very high speeds,
 Distance/rate tradeoff
Vectored/spectrally balanced – number of pairs
allocated, etc.
12
Multiuser Wonders
Outline – Part 4
DSL and Ethernet Applications
 Evolution of DSL and DSM vs SM

 Unbundling
 Towards
ethernet
Multiuser Spectrum Balancing
 Vectored VDSL
 Vectored Ethernet

13
Multiuser Wonders
Unbundling in DSL
ILEC
CLEC

Different service providers can “rent” lines to
customers
 Consequent

Gets worse with wider bandwidths (shorter lines)
 Can


emissions between lines (crosstalk)
be reduced with asymmetric transmission
ILEC – often wants asymmetric
CLEC – often wants symmetric
14
Multiuser Wonders
Spectrum Management

Regulate the spectra of the different service
providers within the cable



Fixed spectra defined for each type of DSL
service, no matter where it is used in the world



Minimize radiation between cables
Balance interest of symmetric/asymmetric (CLEC/ILEC)
STATIC spectrum management
CO-based SM before FCC for approval
ONU/fiber-feed case still under study
15
Multiuser Wonders
DSL Line-Unbundled Evolution?
Network
twisted pair, DSL
Central
Office
SP #1
SP #2
fiber
..
.

LT
SP #2
multiple fibers
in this area lead
to multiple fibers
to home

SP #1
Fiber, FTTH
Fiber for each service provider?
Space at LT for each service provider?
16
Multiuser Wonders
HFC – Hybrid Fiber Coax
satellite
coax
Cable
Head end

Fiber or coax
fiber
split
Fiber from head end to first split



More bandwidth, easier bidirectionally
Coax still shared among 100’s of users (500)
Multiple content providers, unbundling?
17
?
50 MHz
5-40 MHz
500 MHz
Multiuser Wonders
Packet Unbundling
Service 1
Service 2
twisted pair, DSL
Central
Office
fiber
LT
xtalk
Fiber, FTTH

Single fiber/LT controlled by one service provider



SBC example: FCC allowed in Sept 2000 at LT only
Services unbundled at packet level
MAC in LT to control crosstalk problems

Dynamic spectrum management
18
Multiuser Wonders
The Main Technology Issue
(a prime app for MU)
Cross-section of cable
Also, space

Crosstalk – interference between lines
 Increases
with frequency (data rate)
 Largest source of noise and performance loss
 Requires spectrum management
19
Multiuser Wonders
xtalk coupling “resistance”
65
60
crosstalk coupling (dB)
55
50
45
40
35
30
25
20
5
10
6
7
10
10
8
10
Freq (MHz)
20
Multiuser Wonders
Static Spectrum Management
ILEC
CLEC

Up (green) does not like down (yellow)
 Avoid

overlap of yellow with green = static SM
Fix allocations to compromise ILEC/CLEC
 but
xtalk/lines are different as are customers in
each location
21
Multiuser Wonders
DYNAMIC SM = MU Methods

Line spectra/signals varied according to situation
 Data rates/symmetries of customers
 Relative line lengths
 Topology of cable loops

Large Improvements Possible
Migration Strategy to DSL/DSM of future
necessary

 Line
to packet unbundling
 Static to dynamic spectrum management
22
Multiuser Wonders
DSL Evolution Goal
VDSL Evolution
120
100 100
Data Rate (Mbps)
100
80
Down
60
Up
40
25 25
20
6
10
0.6
5
0
4 km
2 km
1 km
.3 km
TP Length
ADSL
VDSL
VDSL- EFM
23
VEFM
Multiuser Wonders
Loop Plant of a Service Provider
Telephone Loop Plant
customer
premises
Inside wire
Central
Office
NID
drop wire
equipment
pedestals
SAI
Main
Distributing
Frame
feeder
distribution
20,000 to
1,500 to
160,000
4,000
number of lines present at a site
22,000 feet
9,000 feet
wire length to customer (90th percentile)

200 to
800
3,000 feet
4 to
12
500 ft
Convenient points
 RT,
SAI (distribution node), pedestal
24
Multiuser Wonders
Dynamic Spectrum Management Steps

Spectrum balancing – line unbundling




For existing “line unbundling” situation where 3rd party can make
recommendations to individual lines, all or some
Evolution to packet unbundling where situations are mixed pac/line
Implementation possible with current ADSL, VDSL systems
Vectoring – packet unbundling



Lines coordinated at ONU side (or CO side) in terms of signals
placed on the line (packet unbunding)
Highest possible performance levels, shortest lines
Add-on (backward compatible) with ADSL, VDSL
25
Multiuser Wonders
Towards Ethernet

DSLAMs
 Move
toward customer
 Speeds go up and DSM used
 Ethernet reused on top of DSL

Multiuser methods
 Spectrum
balancing
 Vectoring
 Combination
of pairs to get 10, 100, or
1000
26
Multiuser Wonders
Outline – Part 4
DSL and Ethernet Applications
 Evolution of DSL and DSM vs SM

 Unbundling
 Towards
ethernet
Multiuser Spectrum Balancing
 Vectored VDSL
 Vectored Ethernet

27
Multiuser Wonders
Spectrum Balancing
Central DSM/DSL Maintenance
CLEC
ILEC

Lines report information
 Helps
with deployment, problem isolation
 Can be used for DSM
 Cent
main recommends line spectra
28
Multiuser Wonders
Translation to Telco Terms
Problem?
Data Rate?
Cheap service
Premium Service Need Fiber
Or
Maintenance necessary
time of day
 as available
 guaranteed

29
Multiuser Wonders
Network Maintenance
Dynamic
Spectrum
Management
DSLAM
DSLT
POTS
Switch
MDF
&/or
splitter(s)

>$1B/rboc-ptt/yr !
 Spec
30
ADSL
Telco’s want to know
 prevent problems
 make $

MLT
ADSL
ADSL
ADSL
Management
Multiuser Wonders
Rlong
Rate REGIONS
Spectral pair 1
Spectral pair 2
Rshort


Plot of all possible rates of lines
 Upstream
 Downstream
Any point in region is possible, but each with different spectra


Varies for each cable and loop topology
Varies for each combination of desired (allowed) rates
31
Multiuser Wonders
Simple Example –PBO
4 lines at 3000’ (7.8 Mbps upstream)
 Line at 500, 1000, 1500, 2000, 2500’
 Compare against best SSM

 Very
little coordination (power of line, rate)
500
1000
1500
2000
2500
Ref length(SSM)
14.0
11.5
10.0
9.0
8.5
32
iterwater.
28.5
24.0
19.5
15.5
11.0
Multiuser Wonders
So far, Static SM

Plans
 998
(USA) – more asymmetric
 997
(Europe) – more symmetric
 Flex
plan
Number of bands programmable
 Start/stop frequencies programmable
 USA VDSL Standard – part 3 (allows DSM)

33
Multiuser Wonders
998 with spectrum balancing
Up
Down
Rate Region: 3000ft loops vs 1000ft loops: Upstream, Plan 998
Rate Region: 3000ft loops vs 1000ft loops: Downstream, Plan 998
9
26
8
25
7
24
Mbps
Mbps
6
5
23
4
22
3
21
2
1
4
6



8
10
12
14
Mbps
16
18
20
22
24
20
0
5
10
15
20
25
Mbps
30
35
40
45
50
Line1500 ft versus 3000 ft
26/6 on 3000’ while 30/22 on 1500’
Static SM only 18/1.5 and 6/6
34
Multiuser Wonders
flex with spectrum balancing
Down
Up
Rate Region: 3000ft loops vs 1000ft loops: Upstream, Flexible Plan
Rate Region: 3000ft loops vs 1000ft loops: Downstream, Flexible Plan
14
26
12
25
10
24
Mbps
27
Mbps
16
8
23
6
22
4
21
2
5
10


15
20
Mbps
25
30
35
20
0
10
20
30
Mbps
40
50
60
Line1500 ft versus 3000 ft.
26/13 Mbps on 3000’, 52/26 Mbps on 1500’
35
Multiuser Wonders
VDSL and ADSL
Down 3000’
Down 4500’
ADSL vs VDSL: co-location at CP
ADSL vs VDSL: co-location at CP
30
18
17
28
4500ft VDSL downstream rate (Mbps)
3000ft VDSL downstream rate (Mbps)
29
27
26
25
24
23
22
21
20
16
15
14
13
12
11
10
1
1.5



2
2.5
3
3.5
4
4.5
12000ft ADSL downstream rate (Mbps)
5
5.5
9
1
1.5
2
2.5
3
3.5
4
4.5
12000ft ADSL downstream rate (Mbps)
5
5.5
Yellow on short line acts like green
9000’ ADSL (fig 6)
26/3 possible on 3000’ VDSL while ADSL runs
5/.5.
36
Multiuser Wonders
VDSL and ADSL with Hdsl/Idsn
Down 3000’
Down 4500’
ADSL vs VDSL: co-location at CP
18
29
17
28
16
4500ft VDSL downstream rate (Mbps)
3000ft VDSL downstream rate (Mbps)
ADSL vs VDSL: co-location at CP
30
27
26
25
24
23
22
21
20
15
14
13
12
11
10
9
0

0.5
1
1.5
2
12000ft ADSL downstream rate (Mbps)
2.5
3
8
0
0.5
1
1.5
2
12000ft ADSL downstream rate (Mbps)
2.5
3
9000’ ADSL
37
Multiuser Wonders
Outline – Part 4
DSL and Ethernet Applications
 Evolution of DSL and DSM vs SM

 Unbundling
 Towards
ethernet
Multiuser Spectrum Balancing
 Vectored VDSL
 Vectored Ethernet

38
Multiuser Wonders
Vectoring Results
D o w n s t re a m w it h N o is e A
U p s t re a m w it h N o is e A
90
80
997
997
998
998
70
V e c t o re d -V D S L
70
60
60
50
Data rate (Mbps)
Data rate (Mbps)
80
50
40
40
30
30
20
20
10
10
500
1000
1500
2000
2500
3000
3500
4000
4500
L o o p le n g t h (ft )
V e c t o re d -V D S L
0
500
1000
1500
2000
2500
3000
3500
4000
4500
L o o p le n g t h (ft )
Up (see correction)
Down
39
Multiuser Wonders
Rate Regions

Last slide had all lines same length
Rate Region allows tradeoffs between lines (1500, 1000)
R a t e re g io n s (u s e r 1 a t 1 5 0 0 ft , u s e r 2 a t 1 0 0 0 ft )
150
N o D a t a -C o o rd in a t io n
V e c t o re d -V D S L
Data rate of user 2 (Mbps)

100
50
0
0
50
100
150
D a t a ra t e o f u s e r 1 (M b p s )
40
Multiuser Wonders
How much better can we do?
(ADSL)
x 10
7
A D S L e vo lu t io n
7
c u rre n t
c a n c e l s e lf-x t a lk
6
c a n c e l a ll x t a lk
in c re a s e b it c a p
im p ro ve c o d in g
Total rate
5
4
3
2
1
0
0
2000
4000
6000
8000
10000
12000
14000
16000
L e n g t h (ft )
41
Multiuser Wonders
Ultimate VDSL
x 10
8
V D S L e vo lu t io n
3
c u rre n t
c a n c e l s e lf-x t a lk
c a n c e l a ll x t a lk
2.5
in c re a s e b it c a p
im p ro ve c o d in g
u lt ra -D S L
Total rate
2
1.5
1
0.5
0
0
2000
4000
6000
8000
10000
12000
14000
16000
L e n g t h (ft )
42
Multiuser Wonders
Outline – Part 4
DSL and Ethernet Applications
 Evolution of DSL and DSM vs SM

 Unbundling
 Towards
ethernet
Multiuser Spectrum Balancing
 Vectored VDSL
 Vectored Ethernet

43
Multiuser Wonders
EFM Examples
Data Rates achievable over a CAT-5 private network
350
300
Data Rate (Mbps)
250
200
Vectoring within quad-TDD
Vectoring within quad
Vectoring within bundle
150
100
50
0
50
100
150
200
250
300
Wire Length (m)
44
350
400
450
500
Multiuser Wonders
EFM Range
Data Rates achievable over a CAT-5 private network
200
180
160
Vectoring within bundle with 20dBm power
Vectoring within bundle with 11.5dBm power
Vectoring within quad with 20 dBm power
Vectoring within quad with 11.5 dBm power
Data Rate (Mbps)
140
120
100
80
60
40
20
0
500



1000
Wire Length (m)
1500
2 lines, 100BT at 1 km
10BT at 1 km on one line easily
4 lines, 100BT at 2 km
45
Multiuser Wonders
Ethernet Examples
Data Rates achievable over a CAT-5 private network
20
VDSL
Quad - 25 MHz
Quad - 50 MHz
Quad - 100 MHz
Quad - 200 MHz
Quad - 400 MHz
18
16
Data Rate (Gbps)
14
12
10
8
6
4
2
0
50
100
150
200
250
300
Wire Length (m)
46
350
400
450
500
Multiuser Wonders
Copper has more bw than fiber?

50 line bundle in last segment of phone
network
 50
lines (200 Mbps/line) = 10 Gbps
 FTTH shares 2.5 Gbps among several homes in
PON architecture

Get bandwidth up in fiber connections to and
within network
 Copper
in last mile has more BW than system can
handle

100BT/100 Mbps to everyone, everywhere a
phone line goes, is possible in the next
decade.
47
Multiuser Wonders
Conclusions

Enormous wireline opportunity for multiuser
 Gains
may be even larger than for wireless
 Relatively stationary environment

The real broadband
 At
least 100 BT to everyone anywhere over a
twisted pair
Data
 Voice, voice, voice
 Video

 Welcome
to the broadband age in this century
48
Multiuser Wonders
Parts 3 and 4: Outline/Schedule
2:00-2:45
 2:45-3:30
 3:30-4:00
 4:00-4:30
 4:30-5:15

MU Theory
channels for wireline
Coffee
DSL and Ethernet arch
Multiuser improvements
49
Multiuser Wonders
References
1)
2)
3)
4)
5)
6)
7)
8)
9)
10)
T. Starr, J. Cioffi, and P. Silverman, Understanding Digital Subscriber Line Technology, Prentice-Hall,
Upper Saddle River, NJ, 1999.
S. Verdu, Multiuser Detection, Cambridge Press, UK, 1998.
C. Aldana and J. Cioffi, “Channel Tracking for MISO Systems using EM Algorithm,” ICC 2001, Helsinki,
Fin, http://cafe.stanford.edu/people/cioffi/dsm/channelpap/icc2001.pdf
C. Zeng, C. Aldana, A. Salvekar, and J. Cioffi, “Crosstalk Identification in xDSL Systems,”
http://cafe.stanford.edu/people/cioffi/dsm/channelpap/jsac01.pdf, August 2001 IEEE JSAC.
G. Ginis and J.M. Cioffi, "Vectored-DMT: A FEXT CancellingModulation Scheme for Coordinating Users,"
ICC 2001, Helsinki, Finland,pp. 305-309.
A. Duel-Hallen, "Equalizers for Multiple Input/Multiple Output Channelsand PAM Systems with
Cyclostationary Input Sequences," IEEE J. Sel. AreasCommun., vol. 10, no. 3, pp. 630-639, April
1992.(Generalizes infinite length MMSE-DFE's derived with spectralfactorization to MIMO case.)
A. Duel-Hallen, "Decorrelating Decision-Feedback Multiuser Detector forSynchronous Code-Division
Multiple-Access Channel," IEEE Trans. Commun.,vol. 41, no. 2, pp.285-290, Feb. 1993.(Zero forcing
DFE solution.)
J. Yang and S. Roy, ``Joint Transmitter-Receiver Optimization forMulti-Input Multi-Output Systems with
Decision Feedback,'' IEEETransactions on Information Theory, vol. 40, no. 5, pp. 1334-1347,September
1994.(Showed that minimizing the decision feedback error is equivalent toachieving the mutual
information.)
G. J. Foschini, G. D. Golden, R. A. Valenzuela and P. W. Wolniansky,``Simplified Processing for High
Spectral Efficiency WirelessCommunication Employing Multi-Element Arrays,'' IEEE Journal onSelected
Areas in Communications, vol. 17, no. 11, pp. 1841-1852,November 1999.(A zero-forcing GDFE
combined with ordering.)
M. K. Varanasi, ``Decision Feedback Multiuser Detection: A SystematicApproach,'' IEEE Transactions on
Information Theory, vol. 45, no. 1,pp. 219-240, January 1999.(Asymptotic analysis (high SNR) of decision
feedback and issues ofordering.
50
Multiuser Wonders
More References
11)
N. Al-Dhahir and A. H. Sayed, ``The Finite-Length Multi-Input Multi-OutputMMSE-DFE,''
IEEE Transactions on Signal Processing, vol. 48, no. 10,pp. 2921-2936, October 2000.
12)
M. L. Honig, P. Crespo, K. Steiglitz, "Suppression of Near- and Far-EndCrosstalk by
Linear Pre- and Post-Filtering," IEEE JSAC, vol. 10, no. 3,April 1992, pp. 614-629.(MIMO
MMSE linear equalizers.)
A. Sendonaris, V. V. Veeravalli, "Joint Signaling Strategies forApproaching the Capacity
of Twisted-Pair Channels," IEEE Tran. Commun.,vol. 46, no. 5, May 1998, pp. 673-685.
R.S. Cheng and S. Verdu, "Gaussian multiaccess channels with ISI: Capacityregion and
multiuser water-filling". IEEE Trans. Info. Th. IT-39, pp773-783, May 1993.
W. Yu, W.Rhee, S. Boyd, and J. Cioffi, "Iterative Water-filling for Vector Multiple Access
Channel," IEEE International Symposium on InformationTheory 2001.
W. Yu, G. Ginis,and J. Cioffi, "An Adaptive Multiuser Power Control Algorithm for VDSL,"
Submitted to JSAC. Also T1E1.4-2001/200R3
W. Yu, G. Ginis, J. Cioffi, “optimum solution of broadcast communications problem,” in
preparation, 2001, [email protected].
K. Cheong, J. Choi, and J. Cioffi, “Multiuser Interference Canceler via Iterative Decoding
for DSL Applications,” IEEE JSAC, Feb 2002, to appear, see also August 1999, ITU
contribution, SG15/Q4-NG-085.
13)
14)
15)
16)
17)
18)
51
Multiuser Wonders