1

Modeling of an ADSL Transceiver Data Transmission Subsystem

Elmustafa Erwa
ABSTRACT
Recently, there has been an increase in demand for digital services provided over the
public telephone line network. Asymmetric digital subscriber lines (ADSL) transmit high
bit rate data in the forward direction to the subscriber, and lower bit rate data in the
reverse direction to the central office, both on a single copper telephone loop. In this
paper I present a survey on the recent research and standardization of ADSL transceivers.
I propose to model and simulate an ADSL transceiver using Synchronous Data Flow
(SDF).


1. INTRODUCTION
With the emergence of the Internet as the cornerstone to most of the
communications in this age, the demand for high speed Internet access has only been
increasing. Asymmetric digital subscriber lines (ADSL) is one of the technologies that
provide high-speed Internet access in residences and offices [1]. It facilitates the use of
normal telephone services, Integrated Services Digital Network (ISDN), and high-speed
data transmission simultaneously. Hence, bandwidth-demanding technologies, such as
video-conferencing and Video-on-demand, are enabled over ordinary telephone lines.
The European Telecommunications Standards Institute (ETSI) and the American
National Standards Institute (ANSI) have defined standards for ADSL transceivers. Both
of these standards propose the use of discrete multi-tone (DMT) as the modulation
technique for ADSL transceivers. DMT divides the effectively band-limited
communication channel into a larger number of orthogonal narrowband subchannels.
This allows for maximizing the transmitted bit rate and adapting to changing line
conditions.
2
In this literature survey, I briefly discuss the recent research and standardization
of ADSL transceivers. I also propose to model and simulate an ADSL transceiver using
Synchronous Data Flow (SDF).

2. BACKGROUND
2.1. ADSL Architecture
ADSL transmits data over frequencies up to 1.1 MHz. The limit of 1.1 MHz is
due to power constraints imposed by the Federal Communications Commission (FCC).
The 1.1 MHz bandwidth is divided into 256 narrowband subchannels downstream (from
the service provider to the customer) and 32 subchannels upstream (from the customer to
the service provider). Each 4.3 kHz narrowband subchannel has a separate carrier, and
the carriers are harmonically related. This type of data transmission is known as DMT
modulation, which is the type of multicarrier modulation proposed in the ADSL technical
specification.
The first channel in ADSL (0-4.3 kHz) is always dedicated for voice. The rest of
the channels are specified differently according to two different methods: 1) Frequency
Division Multiplexing, where upstream and downstream channels use different
frequencies, and 2) Echo Cancellation, where upstream and downstream channels
overlap. In the latter case, echo cancellation techniques make the channels independent.
2.2. ADSL Impairments
The performance of ADSL systems faces degradations due to severe channel
attenuation, Intersymbol interference (ISI) or Interblock Interference (IBI), and other line
impairments including crosstalk, additive white Gaussian noise and impulse noise. The
effective length of the channel impulse response v will cause IBI as the tail of the
3
previous ADSL block symbol will corrupt the beginning of the current block symbol,
hence the name IBI [2].
Crosstalk is caused by electromagnetic radiation due to other signals that are
traveling in adjacent or nearby cables. Near-end crosstalk (NEXT) is caused by signals
that are traveling in the same direction, while far-end crosstalk is caused by signals
traveling in the opposite direction. Both NEXT and FEXT increase with frequency. Table
1 shows the attenuation of a signal traveling at 1.1 MHz in a 24 gauge wire relative to the
length of the wire.
Table 1. Attenuation of a 24 gauge wire signal at 1.1 MHz.

2.3. Discrete Multi-tone Modulation
The fundamental goal of multicarrier modulation techniques is to partition a data
transmission channel with ISI into a set of orthogonal, memoryless subchannels, each
with its own carrier [2]. Data is transmitted through each subchannel independently of
other subchannels. Previous research has shown that such a system is capable of
transmitting at the highest data rate when allocating more bits and energies to
subchannels with higher signal-to-noise ratio (SNR) [3, 4]. DMT modulation is one of
4
many multicarrier techniques including Vector Coding, Structured Channel Signaling,
and Discrete Wavelet Multi-tone Modulation.
DMT based systems uses Quadrature Amplitude Modulation (QAM) to encode
the input bit stream into a finite set of symbols in a one-to-one mapping of integers in [0,
N-1] to complex numbers, where N is a power of 4. Each subchannel carries one QAM
symbol. The number of bits that can be used to generate the QAM symbol in the ith
subchannel with an estimated SNR
n
is approximately given by
) 1 ( log
2

+ =
n
n
SNR
b (1)
where the gap I 9.8 dB ,
m
,
c
for a bit error rate of 10
-7
[4]. The quantity ,
c
is coding
gain of the applied code and ,
m
is the margin. The optimum assignment of bits to each
subchannel, which is not necessarily an integer value, is found using bit loading
algorithms [4]. In the case of non integer assignment, bits are truncated or rounded to the
nearest fraction that can be implemented by multidimensional trellis codes.
DMT modulation and demodulation is implemented using the Inverse Discrete
Fourier Transform (IDFT) and Discrete Fourier Transform (DFT). To implement an N/2
subchannel DMT system, an N size IDFT/DFT is required. The size is doubled by
mirroring the data to impose conjugate symmetry in the frequency domain, which results
in real-valued signal in the time domain after applying the IDFT. The general structure of
a DMT system is illustrated in Figure 1, where {X
0
, X
1
, , X
N-1
} are the original
complex QAM symbols, {x
n
} is the modulated data sequence, {y
n
} is the received
sequence, and {Y
0
, Y
1
, , Y
N-1
} are the decoded complex QAM symbols. The IDFT
and DFT are implemented very efficiently using the well known Inverse Fast Fourier
Transform (IFFT) and Fast Fourier Transform (FFT) algorithms.
5

Figure 1. Basic Discrete Multi-tone scheme.

2.4. Initialization and Channel Identification
In order to optimize the bit allocation in the subchannels, the transmitter acquires
an estimate of the channels impulse response and crosstalk noise spectrum before the
data transmission begins. This process is known as the initialization phase if it was before
the first ADSL block symbol. Otherwise, if the estimates are acquired in the middle of
the communication process between the transmitter and receiver, the process is known as
channel identification phase. During these two phases the transmitter and receiver do the
following: 1) Define a common mode of operation and clock and symbol
synchronization, 2) identify the channel, 3) calculate the optimal bit and energy
allocations for each subchannel, and 4) exchange the bit and energy allocation tables [5].

6
Figure 2. A block diagram of an ADSL Transceiver [6]
3. MODELING AND SIMULATION
3.1. ADSL Transmitter
In the ADSL transmitter shown in Figure 2, an input bit stream is first partitioned
into substreams using a serial-to-parallel (S/P) converter. Each substream is then
encoded using quadrature amplitude modulation (QAM), which produces a complex
number representing each encoded bit substream as mentioned in an earlier section. The
outputs of the QAM encoder are mirrored and conjugated before they enter an N point
IFFT, where N/2 is the number of subchannels. The mirroring creates real values at the
output of the IFFT. To mirror the data for M subchannels, N = 2M, the QAM symbols Xi
are given by [1] as
.
parallel to
serial

QAM
decoder

Freq.
domain
EQ

serial to
parallel

QAM
encoder

Input
bit
stream

mirror
data
and
N-IFFT

add
cyclic
prefix

parallel
to serial

DAC and
transmit
filter

N-FFT
and
remove
mirrored
data

serial
to
parallel

remove
cyclic
prefix

ADC, time
domain
EQ, and
receive
filter

channel

Detected
bit
stream

TRANSMITTER

RECEIVER

N/2 channels

N channels

N/2 channels

N channels





7
The IFFT then maps each QAM symbol into orthogonal frequency bins producing a
discrete multitone symbol of N samples. To form a frame, the last v samples of the
symbol, known as the cyclic prefix (CP) are copied and prepended to the symbol. This
provides a buffer against Interblock Interference (IBI), which was discussed earlier.
Unfortunately the addition of the CP decreases the transceiver power efficiency by a
factor of N/(N+ ) [7]. The final two stages serialize the data and convert it to analog via
the parallel-to-serial (P/S) converter and digital-to-analog (DAC) converter respectively.
3.2 ADSL Receiver
An ADSL receiver receives the data through the channel, which is modeled as an
FIR filter. The operation of the receiver is the dual of that of the transmitter, plus the
addition of an equalizer. The equalizer has two tasks: 1) reduce ISI in the time domain
and shorten the channel impulse response to the CP limit, and 2) compensate for
magnitude and phase distortion in the frequency domain [7]. The first task is done by the
time-domain equalizer (TEQ), while the latter is performed by the frequency-domain
equalizer (FEQ).

4. PROPOSED WORK
The three models of computation that can be used to model the ADSL transceiver
are Dynamic Data Flow (DDF), Synchronous Data Flow (SDF), and Timed Synchronous
Dataflow (TSDF). Figure 3 illustrates the model that is used for each of the blocks in the
ADSL transceivers based on the color of the node.
SDF is best for modeling Digital Signal Processing (DSP) communication
systems because of the static number of inputs and outputs in each node, and the
sequential transfer of data from one node to the next in an acyclic graph [8]. DDF is used
8
in this model due to the initialization and channel identification phases, which
dynamically change the bit allocation for each subchannel based on the channel
estimates. TSDF is used to model continuous time after the conversion of data from
digital to analog via the DAC.
My goal for this project is to model and simulate ADSL transceivers using SDF as
the model of computation. If time permits, I will try to extend the modeling to include
DDF and TSDF. In my simulations, I will attempt to evaluate the bit error rate for
different combinations of channel models, TEQs, and bit allocation tables.


Figure 3. Models of Computations
mirrored
data and
N-IFFT

QAM

S
/
P

FEQ

CP

P
/
S

DAC
Channel

TEQ
ADC N-FFT
and
remove
mirrored
data

QAM

S/
P

C
P

P/
S
DDF SDF
TIMED
SDF
9
5. REFERENCES
[1] Thomas Starr, John M. Cioffi, and Peter J. Silverman, Understanding Digital Subscriber Line
Technology (CD-ROM included), Prentice Hall PTR, 1999, ISBN 0-13-780545-4.
[2] P. S. Chow, J. M. Cioffi, and J. A. C. Bingham, "DMT-based ADSL: concept, architecture, and
performance," IEE Colloquium on High speed Access Technology and Services, Oct. 19, 1994.
[3] P. S. Chow, J. M. Cioffi, and J. C. Tu, "A discrete multitone transceiver system for HDSL
applications," IEEE Journal on Selected Areas in Communications, Volume 9, Issue 6, pp. 895-
908, Aug. 1991.
[4] P. S. Chow, J. M. Cioffi, and J. A. C. Bingham, "A practical discrete multitone transceiver loading
algorithm for data transmission over spectrally shaped channels, IEEE Transactions on
Communications, Vol. 43, Issue 2, Part 3, pp. 773-775, Feb.-March-April 1995.
[5] D. Arifler, M. Ding, and Z. Shen, "Modeling and simulation of discretized data transmission in
very high-speed digital subscriber line, Literature Survey for EE 382C Embedded Software
Systems, The University of Texas at Austin, Austin, TX, March 2002.
[6] G. Arslan, "ADSL Transceivers", Presentation for EE 379K-17 Real-Time Digital Signal
Processing Laboratory, The University of Texas at Austin, Austin, TX, November 15, 1999.
[7] P.J.W. Melsa, R.C. Younce, and C.E. Rohrs, Impulse response shortening for discrete multitone
transceivers, IEEE Transactions on Communications, vol. 44, pp. 1662-1672, Dec. 1996.
[8] A. Lee and D. Messerschmitt, "Synchronous data flow", Proceedings of the IEEE, vol. 75, no. 9,
pp. 1235-1245, Sep. 1987.
[9] B. Hirosaki, An orthogonally multiplexed QAM system using the discrete Fourier transform,
IEEE Transactions on Communications, vol. COM-29, pp. 982-989, July 1981.
[10] K. Kerpez and K. Sistanizadeh High bit rate asymmetric digital communications over telephone
loops, IEEE Transactions Communications, vol. 43, June 1995.
[11] B. R. Saltzberg, Performance of an efficient parallel data transmission system, IEEE
Transactions on Communications Technology, vol. COM-15, pp. 805-811, Dec. 1967.
[12] S. Darlington, On digital single-sideband modulators, IEEE Transactions on Circuit Theory,
vol. CT-17, pp. 409-415, Aug. 1970.
[13] J. A. C. Bingham, Multi-carrier Modulation for Data Transmission: An Idea Whose Time Has
Come, IEEE Communication Magazine, vol. 28 no. 5, pp. 5-14, May, 1990.