Dynamic Spectrum Management

1
Dynamic Spectrum Management
July 2002

• John M. Cioffi
• Prof. EE, Stanford U.
• Cioffi_at_stanford.edu
• And all the work by students
• Taek Chung, George Ginis, Jeannie Fang, Jungwon
  Lee, Dimitris Toumpakaris, Wei Yu

2
Copper has more bw than fiber?
fiber
Cable of copper pair

• DSL will deliver 100 Mbps symmetric to all
• 50 line bundle (500 meters of cat 3 tp)
• 50 lines (100 Mbps/line) 5 Gbps
• FTTH shares 622 Mbps among 50 or more homes in
  PON architecture (even future systems are 2.5
  Gbps)
• So, where is DSL today?
• Where is DSL going?

3
DSL 200X- (V/M/ADSL)
10/100 Mbps

• Data , Voice , Video

4
DSMDynamic Spectrum Management
EFM
10MDSL
SM (static)
VDSL
100MDSL
ADSL
VDSL-2

• Standard in progress USA (ANSI / T1E1.4)
• Dynamic loops, xtalk, rates very
• Not limited by worst-case static situations
• Extension to existing Spec Management
• T1.417-2002
• Improve considerably reliability and performance
• Cannot have spectrum compatibility at higher
  speeds without DSM
• Crosstalk effect is too strong for static
  regulation
• THIS IS A CRUCIAL turning point for DSL
• Goals 100 Mbps symmetric everywhere
• Improve the range of every rate lt 100 Mbps

5
Outline

• Crosstalk and Management
• Autonomous (de-regulated) DSL
• Bonded/Vectored DSL

6
DSLs Challenge Crosstalk
user
Phone CoBase (DSLAM)
user

• Same service provider
• Different service provider
• Management?
• Spectrum fixed or static
• Ok, but performance and reliability severely
  compromised
• Symmetric versus asymmetric

7
Some Crosstalk
FEXT British Telecom (gain)
NEXT - Bellcore (attenuation)

• Xtalk noise increases with frequency

8
Loop Topology

• Many other configurations
• symmetric/asymmetric mixtures
• Data rate mixtures
• Shorter lines
• Bonding of lines

9
ADSL DMT Loading Basics(adapts to each line)
Frequency
800 Million phone lines and growing fast!
Bellcore tests, 1993 (ANSI Standard) GTE tests,
1996 NSTL, 1996
10
Evolution of Crosstalk / management
time
11
Outline

• Crosstalk and Management
• Autonomous (de-regulated) DSL
• Bonded/Vectored DSL

bi n d e r
ILEC DSLAM 1
Central Office
1-20 Mbps (asymmetric/ symmetric)
Present DSL 2002 autonomous (no
coordination) SMC for provisioning, Fault
isolation
Network
ILEC DSLAM 2
SMC1
CLEC DSLAM
SMC2
12
Model Autonomous Case
. . .

• Crosstalk between users is matrix channel
• Interference channel in information theory
• Can users be compatible without large loss? (SSM
  no, DSM - yes)

13
DSM Multi-user Rate REGIONS
Rlong
Rshort

• Plot of all possible rates of users
• ADSL some VDSL use DSM any point is possible
• More than 2 users (vector of possible
  rate-tuples)
• Regions size (volume) can vary with
• each binder
• transmission methods (i.e., DMT has much larger
  regions)

14
ANSI DSM ADSL Blueprint Example(T1E1.4/2001-
273, 278)
ADSL CO
ADSL rcvr
10 X kft
Line 1
ADSL RT
10 kft
5 kft
ADSL rcvr
Line 2
fiber
CO
Same binder FEXT

• CLASSIC field problem with ADSL present-day
  deployments
• Long-line often does not work and field personal
  have to be dispatched
• Improve the performance of line 1 AUTONOMOUSLY
  (no coordination) using DSM-mode in ADSL modems
• Eliminate problem automatically
• Greatly extend ADSL range

15
DSM ADSL basic spectrum result
Downstream Spectra (2 lines)
Initial Short-line spectra, Long-line turns on
Long Line spectra
Intermediate Short-line spectra,
Intermediate Short-line spectra,
Short-line spectra
f

• Short line yields to long
• No DSL hogging
• Enormous improvement on long line
• At expense of reduced rate on short line
• Do so autonomously

16
Verizon Experimental Loop Configuration(T1E1.4/20
02-069 Bell Atlantic Test Lab, Maryland)
Simulation of CO-ADSL fed through RT binder with
RT-ADSL
14 kft
6 kft
Cable Vault
Cable Vault
Mainframe
Mainframe
Pair 1 of 14 kft, 24 gauge 25-pair binder
Pair 16 of 6 kft, 26 gauge 25-pair
binder (Simulates RT portion of loop)
Pair 17 of 6 kft, 26 gauge 25-pair
binder (Simulates RT-ADSL loop)
Remote Side
Remote Side
CO side
Remote Side
CO side
ATU-C
ATU-R
ATU-R
ATU-C
CO ADSL
RT ADSL
Verizon Broadband Integration Lab
17
Crosstalk into pair 17
Pair  Noise (dB)        Pair  Noise (dB) 1    
-82.5                   13    -84.4 2    
-78.1                   14    -80.7 3    
-74.9                   15    -77.7 4    
-82.3                   16    -63.2 5    
-83.0                   18    -67.7 6    
-84.2                   19    -67.5 7    
-78.0                   20    -76.9 8    
-81.0                   21    -80.8 9    
-75.3                   22    -78.7 10   
-73.3                   23    -78.0 11   
-79.6                   24    -73.4 12   
-82.7                   25    -80.7

• Used worst-case FEXT coupling pairs for
  experiment

18
DSM-ADSL Rate Region (Verizon loops)

• Rate-Adaptive produces (9,.1)
• Mode used by Verizon nominally in lab
• Instead, DSM-mode provides (Short, Long)
• (2, 1.8), (4, 1.4), (6, .9) , and (8, .6)

19
Telcordia (Bellcore) Results Blueprint
Test(T1E1.4/2002-063)
ADSL CO
ADSL rcvr
ADSL rcvr
ADSL RT
ADSL rcvr
ADSL RT

• Improve the performance of lines 1 and 3 (which
  is bad under RA operation)
• Line 1 only operates at 300 kbps in normal mode

20
Telcordia DSM-ADSL Rate Regionst1e1.4/2002-063
lines 1 3 (2 held at 1.6 Mbps)

• No coordination (line 2 held at 6 dB margin, 1.6
  Mbps
• 4x improvement on Line 1 (was at 300 kbps if line
  2 does RA or margin-max training) like
  Verizon expt.
• 4x because 15kft, not 20 kft like Verizon

21
DSM Algorithm (Iterative-Water-filling)
NSR(f)
S(f)

• Water-filling is known optimum on single-user
  channel
• Use DMT, and MINIMIZE POWER for given rate/margin
• Best autonomous solution to interference
  problem in binder
• The more lines that do DSM, the better they all
  work
• NEVER any worse than existing fixed/static
  Spectrum Management
• AUTONOMOUS no coorination necessary, procedure
  assures overall best mutual spectra (least
  harmful xtalk) of all DSLS in distributed fashion

22
3-steps to Broadband DSL
ADSL to D 7 Mbps U .3 Mbps 1 line 0-6 km

• DSM
• ADSL -Software option in current ADSL modems
• ADSL 10M-DSL (DMT only) iterative wf
• 100M-DSL (DMT only)

23
10MDSL (EFM) Range/Rate Goals 1
rate lines Per line Length (kft) Goal (kft)
10 M 1 10 2.5 gt3.5
10M 2 5 3.5 gt4.5
10M 3 3.33 5 gt6
10M 4 2.5 7 gt8
10M 5 2 8 gt9
4M 4 1 12 gt12
2M 4 .5 15 gt15
24
25 10MDSLs in same binder (same length)
Verified By Voyan
Verified By Telcordia

• No coordination
• 10 Mbps single-line range is 2.5x SSM result
  (SHDSL)
• 10 Mbps gt1.5 km on 2 lines

25
Example PSD for 4000

• Overlap (EC) to 800 kHz
• Each has 3 basic bands

26
MDSL with ADSL present

• Range of ADSL and of MDSL unaffected if DSM is
  used
• Not true for Static SM (or SHDSL)

27
ADSL at CO, MDSL at RT
Copper, 4 kft
Fiber, 10 kft
M
M
A
A
Copper, L kft 12, 14, 16

• Cabinet is 10 kft from CO last 2 to 6 kft in
  same binder
• ADSL at CO swamped by 64 PAM
• ADSL at CO largely unaffected by IW

28
VDSL vs MDSL rate regions

• Fixed is 998 on V and 1 MHz on M
• Still exceeds all targets of M and V

29
Table Revisted (extra column)

• Allows mixture of ADSL with MDSL, and VDSL with
  MDSL
• Rate regions do reduce in size for VDSL
• But all VDSL and MDSL goals still exceeded

30
998 Plan with Iter-water (VDSL)

• Rates exceed VDSL objectives
• Note spectra converges to 998 when 998 is present
  (4000 MDSL)

31
Basic Principles - Summary

• No DSL modem should use more transmit power than
  it needs
• No DSL modem should use more bandwidth than it
  needs
• Can be solved adaptively and autonomously
• Stop proliferating DSL standards and incompatible
  spectrum management

32
Basic Solution

• Use/Extend ADSL DMT Frequency Grid
• Already on 24 million lines (gt 95 of market)
• ADSL 512 tones down / 128 up
• 10MDSL 1024 tones down/1024 up
• VDSL 2048 up/down
• 100MDSL 4096 up/down

33
Outline

• Crosstalk and Management
• Autonomous (de-regulated) DSL
• Bonded/Vectored DSL

Evolving DSL 2002 Bonding Vectoring Common
DSLAM (LT or RT)
Central Office
b i n d e r
Content 1
Content 2
ILEC LT
Fiber Or multi-channel DSL
D S L A M
20-100 Mbps (symmetric)
Switch router
Network
SMC
time
twisted pair
34
Coordinated 2-sided Signals (vector bonding)
Shared channel
Bonded lines
Bonded lines
Controller
Controller

• Full Vectoring Problem
• Highest data rates

35
DSM Level 2 Coordinated 1-sided Signals
Shared channel

• broadcast and multiple-access probs in IT
• One-sided vectoring, FDM of up/down
• Bonding (vector broadcast and vector MA problems
  in IT)

36
998 with/without Vectoring

• Enormous gain, especially upstream at shorter
  lengths where FEXT is large

37
Full Vectoring (sum up/down)

• 100 Mbps range 500 meters , single line
• 1 km on 2 lines
• gt1.5 km on 4 lines
• Even when partially vectored, 100 Mbps on 2 lines
  at 500 m, 1 km on 4 lines

38
Broadband DSL
Bit rate of lines Rate/line Range
100 Mbps 1 100 M 300-500 m
100 Mbps 2 50 M 1 1.2 km
100 Mbps 4 25 M 1.5-2 km

• Enables video, data, voice services and packaging

39
3 steps
Dynamic Spectrum Management
BDSL to D 100 Mbps U 100 Mbps 1-4 lines/LT 0-2
km
ADSL to D 7 Mbps U .3 Mbps 1 line 0-6 km
AMDSL to D 25 Mbps U 10 Mbps 1-2 lines
CO/LT 0-4 km

2002
2003
2004

• ANSI 2001-200R5, 2002-018, 057, 059, 127, 129
• http//isl.stanford.edu/cioffi/dsm/index.html
• Looking to collaborate with service/content
  providers, equipment/semi companies