ADSL System Enhancement with Multiuser Detection

1
ADSL System Enhancement with Multiuser Detection

• Liang C. Chu
• School of Electrical Engineering
• Georgia Institute of Technology
• Atlanta, GA 30332

2
Table of Contents

• Introduction
• Background History of the Problem.
• Crosstalk
• ADSL and SDSL in a binder.
• DMT-ADSL Channel Characteristics
• DMT
• DMT-ADSL Standard
• Multiuser Transmission
• Telephone Channel
• Multiuser Transmission Systems

3

• ADSL System Enhancement
• Multiuser Detection on DMT-ADSL
• Channel Capacity Studies
• Joint MLSE
• Performance Studies
• Low Complexity Enhancement on ADSL Receiver
• Tone-zeroing
• Multi-stage JMLSE
• Simulation Studies and Results
• Conclusions
• Recommendations

4
Introduction

• An enhancement approach on the DMT-ADSL system.
• Goal spectral compatibility better capacity
  utilization support fast Internet services.
• Core method either increasing signal
  constellation sizes / per sub-channel, or
  extending the deployment ranges with a fixed
  transmission rate, or compensating on a poor BER
  channel in achieving better throughput.
• ADSL service
• Telephone channel, high-bandwidth services.
• New infrastructure for multimedia service.
• Economical and less time to launch service.

5

• Physical channel medium unshielded twisted pair
  line.
• Co-channel interference (crosstalk).
• TPC model and proposed multiuser model.
• Sub-optimal approach on receiver enhancement.

6
Background Problems

• Major threat spectral compatibility.
• Signals coupling in same binder
• crosstalk
• NEXT
• Near-end crosstalk
• FEXT
• Far-end crosstalk

7
Crosstalk Comes From

• Environmental
• Physical media unshielded twisted pair.
• Bandwidth-efficient digital transmission system.
• Different kinds of DSL services in same binder.

8
Near-End Crosstalk
9
Far-End Crossatlk
10
Crosstalk Characteristics

• NEXT dependent on frequency.
• FEXT dependent on frequency, but attenuated by
  twisted cable length.

11
Example on NEXT and FEXT

• Resultsmaximum theoretical data rate.
• NEXT and FEXT limited operation on ADSL.
• ANSI ADSL, 256 channels from DC to 1.104MHz
• Tones 7 to 255 for data transmission.
• Each tone QAM at 0 to 15 bits/Hz based on SNR
• AWGN at 140dbm/Hz, no ISI assumed
• NEXT is the dominated crosstalk.

12
NEXT Coupling Characteristics
13
Discussions

• NEXT increases as f1.5 with frequency, but with
  significant variation in coupling function.
• Any given frequency, only few other pairs may
  contribute significantly to crosstalk, but over
  all frequencies, many wire lines contribute
  randomly.
• Challenge hard to detect in single-user
  detection.
• Solution modify receiver.

14
Current Crosstalk Distribution

• Gaussian Distribution.
• Random interferes, central limit theorem.
• Practical interests and only accurate on single
  type of crosstalk.
• Drawback dependent on error size of Gaussian and
  true distribution.
• Pessimistic on channel capacity especially on
  multiple DSL services.
• New area on multiple DSL services crosstalk
  models.

15
SDSL to ADSL (Multiple DSLs)

• SDSL symmetric DSL
• 2B1Q modulation - 4-level baseband pulse
  amplitude modulation signals
• same data rate in the upstream and downstream
  directions
• same transmit PSD in the upstream and downstream
  directions
• Focus studies on SDSL crosstalk to ADSL
• SDSL services in high demand, together exiting
  with ADSL service.

16
PSD of 2B1Q SDSL

• Spectral compatibility problem with ADSL
• 
  overlap in psd

17
SDSL with T1.413 ADSL

• Results are calculated for same-binder NEXT with
  the standard Unger 1 NEXT model.
• T1.413 full-rate DMT ADSL in the presence of
  NEXT from SDSL (1552 kbps and 2320 kbps).
• DMT tones separated by 4.3125 kHz.
• each tone carries with a 6dB SNR margin.
• Downstream ADSL transmits from 160 kHz to 1104
  kHz.

18
SDSL Crosstalk to ADSL
19
Current Mitigation Plan

• Loop plan
• Testing estimating deployment loops.
• Limiting coverage area and customers.
• Limiting on deployed data rate.
• Drawback
• Inconvenience.
• Capacity waste.

20
Observation and Plan

• Crosstalk channel characteristics change very
  slowly over the time.
• Modeled as static and time invariant.
• Types of crosstalk on practical loops does not
  change.
• Normally fixed DSL services in the same binder
  from the CO to CPE sides.
• Plan on mitigate crosstalk
• Enhance the ADSL receiver, filters the
  crosstalk noise.
• Multiaccess ADSL channel model

21
Multiaccess ADSL Channel Model
hk is the channel impulse response when k1, and
sum together with crosstalk coupling function
when kgt1
22
Discussions

• Background noise is Gaussian.
• DSL Gaussian channel
• Crosstalk is not Gaussian distribution.
• Sum of several filtered discrete data signals
  ADSL (desired) and SDSL (crosstalk).
• Channel model multiple input and single (vector)
  output.

23
Brief on DMT

• Basic Principle
• Split available BW into a large number of
  subchannels.
• Motivation
• Make BW of each the sub-channel sufficiently
  narrow, then no ISI occurs on any sub-channel.
• Technique method
• Transmits many parallel data-streams concurrently
  over the transmission channel.

24
DMT-ADSL (ANSI)

• Two traffic channels
• downstream transmission
• sampling rate of 2.208 MHz, a block size of 512
  (FFT), meaning 256 tones from 0 to 1.104MHz.
• symbol rate is 4 kHz and the width of a tone is
  4.3125 kHz. Average downstream PSD is 40 dBm/Hz.
  
• Upstream transmission
• sampling rate of 276 kHz, a block size 64,
  meaning 32 tones from 0 to 138 kHz.
• symbol rate is 4 kHz and the width of the tone
  remains 4.3125 kHz. Average upstream PSD is 38
  dBm/Hz

25
DMT-ADSL Spectrum
26
Loading Algorithm
27
Physical Channel

• Unshielded twisted pairs
• does not change its physical behavior
  significantly with time and considered a
  stationary channel.
• The transfer function
• The sources of noise in the telephone channel
• digital quantization noise, thermal noise in
  detectors,
• impulse noise and crosstalk.
• Telephone channel is normally treated as a
  Gaussian channel.

28
Multiuser Transmissions

• The fundamental limit of multiuser detection
• mitigate the interference among different
  modulated signals.
• Basic model
• 
  (4.2.1.1)

29

• Multiuser channel is described by the conditional
  probability distribution
• Normally, many channels fit in the linear AWGN
  model, shown in Eq. (4.2.1.1).
• Optimum multiuser detection
• a generalization form of the optimum single-user
  channel detector - maximum likelihood multiuser
  detector.

30

• Linear multiuser detection in AWGN channel
• As Eq. (4.2.1.1)
• detection of desired input user xl, it may be
  that the overall minimum distance is too small
• a single fixed value for xl may corresponding to
  the two multiuser codewords that determine the
  overall dmin.
• defined as
  (4.2.2.1.1)
• Results
  (4.2.2.1.2)
• it is possible for a detector extracting a single
  user to have better performance on one that
  extracts all other users.

31
Channel Capacity

• Conventional single-user ADSL receiver
• Sum all the crosstalk signals and background
  noise together as AWGN.
• 
  (5.1.2.5)
• Enhanced multiuser ADSL receiver
• JMLSE selects all possible inputs, min.
  distance on output.
  
• 
  (5.1.2.8)

32
Two Users

• Consider the two user case
• where, N is AWGN,
• , the desired signal and ,an
  interfered signal.
• Capacity for user 1
  (5.1.2.10)
• Capacity for user 2
• 
  (5.1.2.11)

33

• Jointly detect, then the achievable capacity
  
• (5.1.2.12)
• Considerable capacity improvement when the
  interference structure is taken into account.
• (5.1.2.13)

34
Single vs. Multiuser Channels
Rate (User 1)
C1
Multiuser
C1
Single User
Rate (User 2)
C2
C2
35
Alternative Viewpoint

• Multiple input x.
• Mutual information I(x,y), and I(r,y).
• Data rate individual input.
• Aggregate data rate .
• Shannon theorem upbounded by I(r,y).

36

• Achievable data rate on desired channel
• (5.1.2.15)
• Discussion
• Limit on (5.1.2.15) can be much larger than the
  data rate based on Gaussian crosstalk
  assumptions.
• The sum on right can be much smaller number, due
  to frequency-selective crosstalk coupling
  function.

37
Analysis and Examples
SDSL Crosstalk
ADSL Channel
Example2
Example1
38
Example 1crosstalk mutual information
1552 kbps SDSL coupling to ADSL
Unger 1 model,
Mutual information of crosstalk on each DMT-ADSL
tone
If silence near 20 tones, fully detected
39
Example 2Throughput Comparison

• Conclusion
• still having enough room for ADSL.
• too pessimistic on current model.

SDSL crosstalk
Theoretic ADSL capacity
ADSL Channel
Gaussian model
40
Joint MLSE

• Principle
• search all possible transmitted signals, find a
  best match signal set on the received signal.
• The best detector, with upper bound on multiuser
  system.
• Drawback large computational complexity.

41
Details on Receiver

• Viterbi decoding engine for MLSE receiver.
• Select the state having the smallest accumulated
  error metric and save the state number of that
  state.
• Iteratively perform the following step until the
  beginning of the trellis is reached working
  backward through the state history table, for the
  selected state, select a new state, which is
  listed in the state history table as being the
  predecessor to that state. Save the state number
  of each selected state. This second step is
  called traceback.

42

• work forward through the list of selected states
  saved in the previous steps. Look up what best
  estimated input bit corresponds to a transition
  from each predecessor state to its successor
  state.
• Use T/2-spaced MLSE Receiver
• eliminate implementation for whitening matched
  filters - with fixed analog filters, not depend
  on unknown channel (pulse shaping filter).
• nearly insensitive to sampling time off-set,
  capable of recoving non synchronized cochannel
  signals more easily.

43
JMLSE ADSL Receiver (Optimal)

• Multiple input and single output model.
• Detect desired ADSL and filtered coupling
  crosstalk signals.
• JMLSE ADSL Receiver
• extension of the single channel MLSE.
• assume Gaussian channel.
• Ex co-channel pairs caseJMLSE selects the ith
  joint symbol sequence that maximizes
  the metric
• (5.2.6.1)
• meaning select a signal set with minimized
  distance from the received signals.
• MethodJoint Viterbi algorithm.

44
Joint Viterbi Algorithm

• Objective determine the pair of sequence
  
• that minimizes the sum of squared
  errors
• defined by the error sequence .

k

r
k
x
i
1
,
Primary Channel

f
(k)


Estimate

1
k
e
k
ˆ
r
j
i
,
j
i
,





-

k
x

j
,
2
Seconda
ry Channel

f
(k)

Estimate


2
45

• Joint VA (JVA) for JMLSE is very similar to the
  standard VA.
• Joint state
• number of states required to implement JVA
• Each joint state at time k-1
• Transition to states at time k.
• Be reached by same number of states from time k-2.

46
Prototype on Modification of Receiver
47
Performance Study (Optimal)
one SDSL disturber NEXT into one T1.413 full rate
DMT-ADSL system
gap of 4 dB ,FIR channel with 256 memory states
48
(No Transcript)
49
Low Complexity Enhancement

• JMLSE is an optimal solution.
• drawback high computational complexity.
• Goal Reduce computational complexity
• Multistage JMLSE multiple MLSE like.
• Tone zeroing use DMT loading algorithm, and
  adaptive decision feedback or echo
  cancellation like.

50
Tone Zeroing Method

• Principle Use loading algorithm to silence some
  selected BW tones with low SNR, then building a
  adaptive cancellation table.

51
SDSL Coupling to ADSL Example

• Adjacent pairs SDSL to ADSL.
• assume ADSL channel is static.
• relative constant on crosstalk profile table
  using LMS algorithm.
• zeroing about 20 tones to build up a NEXT
  cancellation table.
• Result up to 6 dB in margin.
• Discussions
• advantage of mitigate the NEXT and complexity
  reduction (comparing with JMLSE) with asymmetric
  and symmetric services coexist.

52

• key issue for the tone zeroing is necessity of
  accurate modeling of noise (crosstalk).
• feedback section is using some kind of adaptive
  filter technique, and adaptive filter coefficient
  is largely depends on frequency components with
  high power.
• If a frequency band making NEXT noise has small
  power, it can not be modeled correctly due to
  high power frequency component until sufficient
  number of coefficient are used.
• tone zeroing modeling works well for high
  frequency power noise component.

53

• telephone channel, many kinds of random noises
  often occur in any selected frequency band.
• may make an error decision on the cancellation
  table and induce error propagation.
• Proposed multi-stage joint MLSE for ADSL receiver
  (applied to both DMT and non-DMT DSL solutions).

54
Same Example w/Tone-Zeroing
55
Complexity Reduced JMLSE

• Multi-stage JVA
• very similar to conventional VA receiver.
• having multi-stage inputs and outputs.
• Method as adjacent pair-wise case
• the primary (strong) signal r1(k) is estimated
  using low delay decisions from a single-channel
  VA, and is forwarded to the
  second VA section to estimate the co-channel
  signal.
• Advantage this structure is largely reducing the
  complexity on optimal JVA (JMLSE).
• Complexity as a similar range of a conventional
  VA, with just a scale-increasing factor by N.

56
N Co-channel Binder

• Ratio
• Assume equal lengths, L,
• obvious to us R is always (much) lt 1.

Multi-stage JMLSE
JMLSE
57
Two Methods (Pair-wise)
Two-stage JVA ,without Feedback Section
only an additional L tap filter computational
increasing
Two-stage JVA ,with Feedback Section
58
Make Decision
Example on PAM channel, signal-corsstalk-ratio10
dB
T/2-spaced MS-JMLSE-W/FB
59
Performance Simulations

• Test Environment
• SDSL and other DSLs NEXT to ADSL.
• Loop Characteristics

60
Test Loop 1
61
Test Loop 2
62
Test Loop 3
63
Other works on xDSL Crosstalk

• Crosstalk with Gaussian Distribution for DSL
• (1) cook,1999 (2) zimmerman, 1998 (3) kerpez,
  1995 (4) kerpez, 1993.
• Multiuser Detection, but for wireless
  communications
• (5) Verdu, 1998.
• Multiuser detection in VDSL study
• (6)Cioffi, 1998.
• (1) The noise and crosstalk environment for ADSL
  and VDSL systems ,Cook, J.W. Kirkby, R.H.
  Booth, M.G. Foster, K.T. Clarke, D.E.A. Young,
  G.,IEEE Communications Magazine , Volume 37
  Issue 5 , May 1999.
• (2) Achievable rates vs. operating
  characteristics of local loop transmission HDSL,
  HDSL2, ADSL and VDSL , Zimmerman, G.A. ,
  Conference Record of the Thirty-First Asilomar
  Conference on , Volume 1 , 1998.
• (3) High bit rate asymmetric digital
  communications over telephone loops , Kerpez,
  K.J. Sistanizadeh, K.,Communications, IEEE
  Transactions on , Volume 43 Issue 6 , June
  1995.
• (4)Near End Crosstalk is Almost Gaussian, K. J.
  Kerpez, IEEE Transactions on Communications, Vol.
  41, No. 5, May 1993.
• (5) Multiuser Detection, S. Verdu , Cambridge
  Press, 1998.
• (6)Mitigation of DSL Crosstalk via Multiuser
  Detection and CDMA, J. Cioffi , ANSI,
  T1E1.4/98-253, August 1998.

64
Related DSL Publications

• An Enhancement Study on the SDSL Upstream
  Receiver, 2001 IEEE International Symposium on ,
  Volume 4 , 6-9 May 2001, Page(s) 442 445.
• Mitigation of Crosstalk on the SDSL Upstream
  Transmission with Vector Equalization, IEEE
  International Conference on Communications,
  Session AN5 Transmission Systems, Helsinki,
  Finland, June 11-14, 2001 .
• A Study on Multiuser DSL Channel Capacity with
  Crosstalk Environment, 2001 IEEE Pacific Rim
  Conference on Communications, Computers, and
  Signal Processing, Session MP4 DSP for
  Communications, Victoria, BC, Canada, August
  24-28, 2001.
• Performance Enhancement on a Multiuser Detection
  ADSL Modem, In preparation to IEEE Transitions
  on Consumer Electronics.
• Complexity Reduced ADSL System with Multiuser
  Detection , Submitted to 2002 IEEE International
  Conference on Communications.

65
Conclusions

• Overview the problem on xDSL spectral
  compatibility problems.
• Traditional Gaussian crosstalk under-project ADSL
  achievable capacity.
• ADSL system enhancement with multiuser detection.
• a core method on improvements of either
  increasing transmission data rate, or extending
  deployment areas, or compensating in poor BER DSL
  channels, based on different requirements.
• Enhanced ADSL receiver has acceptable
  computational complexity for a chip realization.
  
• Benefit on QoS for last-mile fats Internet
  transmission.

66
Recommendations

• This approach can apply to DMT and non-DMT ADSL,
  HDSL, SDSL and future VDSL studies.
• may extensible to fiber and wireless.
• Other complexity reduction methods for JVA
  decoding can be further studied (this thesis
  gives a kind of beginning point).
• Possible dual-mode DSL transceivers.