DSL TECHNOLOGIES
In this chapter we briefly give an introduction to the world of the DSL technologies with a special emphasis on the DSL environment. We present well known formulas on which we will refer throughout all the project: crosstalk characterisation and Shannon formula.
2.1 Introduction
Digital Subscriber Line technologies provide transport of high-speed digital information trough the installed telephone lines that connect the end users at the customer premises to the telephone central office.
The existing telephone lines in the Public Switched Telephone Network were designed and built to carry analogue voice, or Plain Old Telephone Service, with a 3,4 kHz bandwidth channel. That is why a cheap narrow-band medium, the twisted pair cable, was chosen to support this service. Furthermore the link supported by the twisted pair, also called local loop, was optimised for analogue transmission with 4 kHz bandwidth by using low-pass filters to limit the bandwidth.
With new high-speed data services being developed, the local loop needs to evolve to a broadband digital network capable of delivering voice, high-speed data and other added value communication services. DSL technologies are the solution to support and assure this broadband digital network on common telephone lines.
DSL technologies use advanced and sophisticated digital signal processing
line due to attenuation and reflection of the signal and noise sources such as the crosstalk noise, the radio-frequency noise and the impulse noise.
With DSL technologies integrated into the local loop, information is sent as digital signals at much higher frequencies than for traditional analogue telephony transmission. This enables more data traffic to be delivered over an existing twisted pair.
Next figure shows actual coverage in ADSL:
Figure 2.1: ADSL Coverage
Since more and more customers are asking for high-speed connections and since new services are requiring even higher data rate, there is an actual need in DSL to extend rate and reach.
Possible solutions are either the introduction of new DSL flavours (ADSL2,
ReADSL, VDSL) , either the introduction of new bitallocation techniques.
Tier 1: 1.5 Mb/s Tier 2: 3.5 Mb/s Tier 3: 5.5 Mb/s Tier 4: 7.5 Mb/s Tier 5: 10 Mb/s
The Access Challenge: Coverage and Bandwidth
Green Zone Red Zone (tier 1
reach from CO)
C Grey Zone (Tier 2
reach from CO)
Covera ge
Bandwi Services dth ADSL coverage
from CO
Figure 2.2: Access Challenge: extend coverage and bandwidth
Our analysis will focus on a new bitallocation technique (DSM level-1) and its gains with respect to the reference ADSL.
In order to achieve it, we now briefly introduce the DSL environment and its well known characteristics in the following paragraphs. Then we will start our analysis beginning in chapter 3.
2.2 DSL environment
DSL modems use frequencies above the traditional voice band to carry high-speed
data. The telephone channels are severely frequency selective. One way to combat
intersymbol interference (ISI) is to use Discrete Multitone (DMT) modulation, which
divides the frequency band into a large number of ISI-free sub-channels and lets each
sub-channel carry a separate data stream. DMT modulation scheme is standardized
for ADSL and VDSL. The use of DMT modulation allows arbitrary power
assignment in each frequency tone, thus making spectral shaping easy to realize.
A DSL binder can consist of up to 100 subscriber lines bundled together. Because of their close proximity, the lines create electromagnetic interference into each other, thus causing crosstalk noise (see Figure 4). Near-end crosstalk (NEXT) refers to crosstalk created by transmitters located on the same side as the receiver. Far-end crosstalk (FEXT) refers to crosstalk created by transmitters located on the opposite end of the line. NEXT is usually much stronger than FEXT. To avoid NEXT, DSL transmission uses either frequency-division duplex (FDD), where all loops transmit in the same direction in every frequency band, or time-division duplex (TDD), where all loops transmit in the same direction in every time slot. North American and European standards use FDD, while TDD is used in Japan. In our project we consider frequency-division duplex system, where the entire frequency band is divided into upstream and downstream bands.
Upstream Downstream
POTS
Figure 2.3: ADSL BAND.
The DSL environment is a Multi-user environment, because the background noise in the loop is typically small, and the system performance is limited by crosstalk. The DSL environment consists of multiple transmitters and multiple receivers interfering into each other. This model is usually referred to as an interference channel, and it is different from a multiple access or a broadcast channel which models many-to-one or one-to-many transmissions. The multiple access and broadcast models are not suitable for the DSL environment considered here, because data coordination among the transmitters or the receivers is generally not possible. Coordination is not
300Hz 3400Hz
30kHz 125 kHz 164 kHz 1,1 MHz
138 kHz
possible at the remote terminals, because they are located in different geographical locations.
Coordination at the central office side is feasible only if all loops are controlled by the same service provider, which might not be the case in an unbundled environment.
2.3 Crosstalk
For the DSL family of transmission systems the limiting factor on loop range has been crosstalk coupling of signal energy from similar or different transmission systems on other pairs in the cable and not from the end-to-end attenuation of the signal.
The current crosstalk models were developed in the 1980s based on computer simulations of the physical structure of the cables and later compared with measurements.
Figure 2.4: DSL crosstalk environment.
2.3.1 Near end crosstalk, NEXT
Crosstalk noise that occurs when a receiver on a disturbed pair is located at the same
end of the cable as the transmitter of a disturbing pair is called Near-End-Crosstalk
(NEXT).
The simplified NEXT model is expressed by
[ ]
f,n S( f ) X n06 f3 2(
1 H( f,L)4)
NEXT = ⋅ N ⋅ . ⋅ / ⋅ −
where
XN =8.536×10−15,
n=number of disturbers, f = frequency in Hz, NEXT coupling length,
=
L
H( f,L)is the magnitude of the insertion gain transfer function for a loop of length , and is the power spectrum of the interfering system at the initial point of coupling with the interfered system.
L S ( f )
2.3.2 Far end crosstalk, FEXT
Crosstalk noise that occurs when a receiver on a disturbed pair is located at the other end of the cable as the transmitter of the disturbing pair is called Far-End-Crosstalk (FEXT).
The FEXT model is expressed by
2 6 . 2 0
) ( ) ( ] , ,
[f n l S f H f X n l f
FEXT = ⋅ ⋅ F ⋅ ⋅ ⋅
where
H( f)is the magnitude of the insertion gain transfer function affecting the disturber signal, X
F= 7 . 74 × 10
−21,
n=number of disturbers,
l=the FEXT coupling path length in feet, frequency in Hz, and is the power spectrum of the interfering system at the initial point of coupling with the interfered system.
=
f
S( f)The FEXT model assumes the insertion gain transfer function is computed for the total cable path located between the interfering transmitter and the victim receiver.
On the other hand, the coupling loss is computed only over the coupling path length . The coupling path length is the length of cable over which the victim receiver and far-end disturbing transmitter have a common cable path.
l
2.3.3 FSAN method for combining crosstalk contributions from different types of disturbers
Instead of directly adding the crosstalk power terms, each term is first arbitrarily raised to the power 1/0.6 before carrying out the summation. Then, after the summation, the resultant expression is raised to the power 0.6. This can be expressed as:
( )
0.61
6 . 0 / 1 1
,
, ⎟⎟
⎠
⎞
⎜⎜
⎝
= ⎛
⎟⎟
⎠
⎞
⎜⎜
⎝
⎛ =
∑ ∑
=
=
N i
i N
i
i Xtalk f n
n n f
Xtalk
,
where
Xtalkis either NEXT or FEXT, is the total number of crosstalk disturbers,
nN is the number of types of different disturbers and is the number of each type of disturber.
n
jTotal noise power spectral density
The NEXT term and the FEXT term are computed to arrive at separate NEXT and FEXT disturbance power spectra. These power spectra should then be summed with the background noise to determine the total disturbance seen by the victim receiver.
The total noise seen by the victim receiver is given by:
[ ]
, [ , , ] 10 14 )(f = NEXT f n +FEXT f n l + −
N
mW/Hz
2.4 Shannon formula
xTU-R1
oTU-R
xTU-C1
xTU-C2 xTU-R2
G1
G2
DSLAM
)
2(f S
)
1(f S
)
1(f N
)
2(f N )
11(f H
)
22(f H
)
12(f H
)
21(f H
R1
R2
Fig. 2.5. 2-User DSL (downstream) transmission environment.
Fig. 5 shows the DSL transmission environment for the 2-user case and depicts only the variables relevant for downstream transmission for sake of simplicity. The CO (central office) transceiver xTU-C
1of user 1 sends a transmit PSD (power spectral density) over the (direct) channel with transfer function to the CP (customer premises) transceiver xTU-R
)
1(f
S H11(f)
1
. xTU-R
1has an SNR (signal-to-noise ratio) gap plus noise margin of
Γ1(explanation given below). xTU-R
1is disturbed by the CO transceiver of user 2, xTU-C
2, which has a transmit PSD , via the (far-end) crosstalk channel with transfer function . All other disturbances are grouped into the noise PSD . It is assumed that upstream and downstream transmission are multiplexed by means of FDD (frequency division duplexing), which means that the near-end crosstalk can be neglected. The downstream bit rate for user 1 is given by equation (1). For user 2 the equation is similar.
)
2(f S )
12(f H )
1(f N
R1
As indicated by equation (1), the SNR gap plus noise margin
Γ1expresses how much of the signal-to-noise ratio actually can be exploited for data communication by a particular transceiver under the constraint of a maximum BER (bit error rate).
SNR1
( )
1
1 2
0 1
2
11 1
2 2
0 1 12 2 1
log 1 ( )
( ) ( ) log 1
( ) ( ) ( )
B
B
SNR f
R df
H f S f
df
H f S f N f
⎛ ⎞
= ⎜ ⎝ + Γ ⎟ ⎠
⎛ ⎞
⎜ ⎟
= ⎜ ⎜ ⎝ + Γ + ⎟ ⎟ ⎠
∫
∫
(1)
The smaller
Γ1, the larger the bit rate
R1. On the one hand the SNR gap originates
from the fact that the complexity of a real transceiver is limited. On the other hand the noise margin ensures that the transceiver can maintain a certain bit rate with a certain maximum BER, when the SNR deteriorates. A perfect transceiver with infinite complexity could operate at a BER of 0 with a SNR gap of 0 dB. The bit rate reached by such an ideal transceiver is called the Shannon limit, which defines the maximum achievable bit rate given a certain transmission environment specified by all the remaining variables in (1). The SNR gap depends on the required bit error rate (BER), which is normally set to 10
-7, and the DSL transceiver technology. For example for ADSL with Trellis coding the SNR gap is 5.7 dB. The SNR gap can be decreased by implementing more advanced coding techniques in the DSL transceiver. For example turbo coding allows decreasing the SNR gap to as low as 1 dB, but at the (high) cost of increased DSL transceiver complexity.
We have included in also the noise margin, which is the part of the SNR deliberately not used for the transmission of bits, although this would be possible.
The noise margin is typically set to 6 dB in order to protect the line against increases in noise level for example caused by impulse noise (see paragraph 3.1.3).
Γ1
Equation (1) tells us that the bit rate can be augmented by increasing the variables in the numerator on the one hand: the signal-to-noise ratio , the transmit PSD
and the (direct) channel power transfer function
)
1(f SNR )
1(f
S H11( f)2