DSL technology provides high bit rate digital services over the existing telephone lines without the need for modifications to the customers facilities. Since the first field test of DSL technology, it has been recognized that Impulse Noise along with other impairments like attenuation and crosstalk, can affect the achievable bit rate of DSL systems. [1]
The purpose of this assignment is to analyze and simulate the impact of impulse noise on a DSL signal and the Impulse Noise protection's impact on the achievable Bit-rate. How ISPs use different "DSL profiles" on a customer's line to achieve better data-rates and synchronization stability will also be presented. A theoretical approach of the subject will be studied along with a practical one with the use of Matlab and the internal DSLAM Monitoring tool that a greek ISP company uses. Different "DSL profiles" will be tested on a customer's line to extract conclusions.
1. FACULTY OF ENGINEERING
DEPARTMENTS OF ELECTRONICS AND AUTOMATION
TECHNOLOGICAL EDUCATIONAL INSTITUTE OF PIRAEUS
MSc IN NETWORKING AND DATA COMMUNICATIONS
COURSEWORK
MODULE:
CI 7110 - Digital Communications
ID: K1465156
Module Coordinator:
Dr. H. Simos & Dr. Ch. Patrikakis
Date of Module:
14/5/2016
Name of Student:
Grigoropoulos Michail
Kingston University London
2. FACULTY OF ENGINEERING
DEPARTMENTS OF ELECTRONICS AND AUTOMATION
TECHNOLOGICAL EDUCATIONAL INSTITUTE OF PIRAEUS
Subject: Impulse Noise over xDSL Systems
Submission Date: 14/5/2016
Grade (%): ________________________________________________
% Grade reduction because of submission delay: _____
(5% Grade reduction per every day of Cwk delay).
Final Grade (%): ________________________________________
Kingston University London
3. Impulse Noise over xDSL Systems – MSc Networks & Data Communications - Kingston University London 3
Abstract - DSL technology provides high bit rate digital
services over the existing telephone lines without the need for
modifications to the customers facilities. Since the first field test
of DSL technology, it has been recognized that Impulse Noise
along with other impairments like attenuation and crosstalk, can
affect the achievable bit rate of DSL systems. [1]
The purpose of this assignment is to analyze and simulate the
impact of impulse noise on a DSL signal and the Impulse Noise
protection's impact on the achievable Bit-rate. How ISPs use
different "DSL profiles" on a customer's line to achieve better
data-rates and synchronization stability will also be presented. A
theoretical approach of the subject will be studied along with a
practical one with the use of Matlab and the internal DSLAM
Monitoring tool that a greek ISP company uses. Different "DSL
profiles" will be tested on a customer's line to extract
conclusions.
Keywords: Impulse Noise, Protection, INP, SNR Targeting, xDSL,
ISP, DSL Profiles, Data Rate, Bit-Rate
I. INTRODUCTION
DSL technology was developed in 1989 over the existing
widely spread telephone network so that no modifications to
the outer plant facilities would have to be made. This fact
would make it easier for the ISP to provide services and for
the customers to buy them but it would also import problems
to the venture because of the telephones network age,
condition and topology. [1]
Impairments such as crosstalk, attenuation, low SNR and
Impulse Noise are major afflictions to the transmission rate
and the stability of the synchronization. [2] Applications like
streaming video, Voice-over IP (VoIP) and video
conferencing are now more popular than ever. This type of
application is often referred to as 'time critical'. To work at
their best, time-critical applications require a stable connection
so data packets arrive at their destination consistently,
ensuring sounds and images flow without interruptions or
pauses. To mitigate the effects of Impulse Noise and low
SNR, ISPs use standard line profiles that suit each customer's
line to achieve higher data rates and service stability.
ADSL line profiles provide a range of pre-set line
configurations which allows the Internet Service Provider
(ISP) to fine tune the performance of its DSL service. They
are particularly useful if the customer experience drop outs or
data errors with their Internet connection.
II. BACKGROUND
A. Impulse Noise
Impulse Noise is a non-stationary, unpredictable burst noise
phenomenon that occurs in semiconductors and consists of
sudden step-like transitions between two or more discrete
voltage or current levels, as high as several
hundred microvolts, at random times. Impulse noise can be
caused from several sources, like on-off telephone hook
events, electrical appliances, transport vehicles or atmospheric
noise from electrical discharges and it usually creates an
electric Noise of 5-10 db.
Fig 1. Tx Waveform and Burst Noise
Even though due to its nature, Impulse Noise is very
difficult to be described statistically, several studies have been
conducted on the subject [3].
The impulse amplitude model is based on an approach
originally proposed by Henkel and Kessler, which consists of
approximating the voltage histograms with a generalized
exponential distribution of the form:
(1)
where u is the voltage and u0 a scaling parameter. This
model reflects well the fact that voltage distributions are
heavy-tailed and offers a good approximation for all measured
impulse noise voltage amplitude distributions collected in the
networks of both DT and BT. [4]
B. Impact of impulse noise on a DSL System
Impulse noise is an additive source that is only active for
very short intervals in time. If there is no interleaving, the FEC
cannot cope with the burst errors that the arrival of noise
impulses can produce. This can cause from noticeable
corruption of images or video services to loss of
synchronization. With the presence of interleaving and as its
depth rises, some errors can be corrected but still the effects of
the burst noise will be noticeable. Only with 64 step
interleaving almost all errors can be corrected but most
applications cannot tolerate the latency issues this depth of
interleaving causes.
Impulse Noise over xDSL Systems
Michail Grigoropoulos, MSc Networks and Data Communications, Kingston University
4. Impulse Noise over xDSL Systems – MSc Networks & Data Communications - Kingston University London 4
Fig 2. Error scatter plot. In the center we can see the effects of gaussion noise,
while on the outside circle, shows the impact of impulse noise.
Although the statistics of error free seconds do not change
significantly with various user data rates, the distribution of
the number of error free cells is. For different bit rates one unit
of time contains a different number of cells. At higher data
rates the distribution shifts towards larger number of
consecutive error free cells. This also means that Frame Error
Rate probability will be higher than Cell Error probability at
lower bit-rates with low depth interleaving, while on higher
depth interleaving, FER will be lower than CER. [5]
Fig 3. FER Probability Vs CER Probability
C. Impulse Noise Protection
INP parameter is a measure of the amount of protected /
recovered DMT symbols after a noise burst occurs. Various
INP techniques have been proposed.
RS-Erasure Decoding
Retransmission
Frame Blanking / Repetition.
An interleaver combined with an RS Decoder is a popular
scheme that is used widely by ISPs. This scheme ingresses
impulse noise with different code words sizes and variable
interleave depth.
The Reed-Solomon (N, N-R) is a linear block code which is
capable of correcting up to R/2 errors (R is no. of redundancy
bytes) within a codeword. An Interleaver and de-Interleaver
combination is used to spread the burst errors across several
codewords incurred by the Impulse noise. By increasing the
redundancy bits and interleaving dept, the coding gain in the
system increases and the effect of impulse noise on the signal,
is mitigated. Although, lower data-rates and higher delay
occurs by the process. The higher the INP requirement is, the
higher memory is needed and increase in delay occurs. [7]
Table 1. Rs Encoder Parameters
Fig 4. RS-Erasure Decoding
At the transmission layer, DMT symbols are of fixed
duration of 250 microseconds. The INPMin parameter defines
the minimum number of DMT symbols that will be protected
from impulse noise and thus the minimum duration of impulse
noise from which error correction should be able to recover.
For example, INP=2 can correct up to two successive
corrupted DMT symbols during one period of 250μs. To
provide maximum error protection, INPMin should be set as
high as possible without unduly compromising bitrates and
latency.
The INPmin setting is directly related with the symbol rate.
At higher values of INPmin, the circuit will be restricted to
lower maximum bit-rates. INP setting is related with FEC
parameters (interleaving depth and delay). With a good
combination of low delay and high INP, high stability of a
DSL line against burst noises, can be achieved. However, such
settings, forces the FEC parity ratio to 1/3 or 1/2 at the cost of
maximum data-rate and delay due to memory capacity and
processing speed requirements. More specifically, FEN, delay
and INP are related by the following equation. [8]
FEC parity ratio = ½ * (INP/Delay) (2)
Table 2. Max downstream related with INPmin and delay
5. Impulse Noise over xDSL Systems – MSc Networks & Data Communications - Kingston University London 5
III. SIMULATION OF IMPULSE NOISE
A. Simulink model
To simulate a burst noise and its effects on a DSL system,
the standard simulink's 256 ADSL ITU - T G.992.1 demo was
used. The original model was altered to simulate a burst
additive noise to the signal and measure the increase in BER
and FER.
Fig. 5 256 Channel ADSL - ITU - T G.992.1 Block diagram
For the simulation, the "Telephone Line" block was
changed. Two vector scopes were added to visualize the signal
without the impulse noise addition and the final transmitted
signal with the noise. To simulate the burst noise, a "Gaussian
Noise Generator" was used with 10 dB gain to amplify it and a
delay of -512 samples. A N-Sample Switch was used to adjust
the pulse duration and a Sum block to merge the impulse noise
with the original signal.
Fig. 6 Telephone line block diagram
A discrete-time scatter plot was also used inside the
demodulator block, to visualize the constellation and the noise
effect on it.
Fig. 7 Demodulator with Scatter plot block
B. Simulation
Running the simulation, two vector scopes and one scatter
scope open. The simulation runs with sampling time 0.5 and
2001 frames are transmitted. At first the original model
(without the impulse noise) was simulated.
Fig. 8 - Simulation without impulse noise
Fig. 9 - Scatter scope & Vector scope - No impulse noise
The constellation can be easily distinguished and the
signal's amplitude goes from -2 to 2 dB. BER is 5 x 10-4,
which
is an acceptable measurement.
Then, the model with the impulse noise addition was
simulated.
Fig 10. Simulation with impulse noise addtion
6. Impulse Noise over xDSL Systems – MSc Networks & Data Communications - Kingston University London 6
Fig 11. Scatter scope & Vector scope - No impulse noise
At the moment the impulse noise arrives, the constellation
cannot be distinguished at all and the signal's amplitude goes
from -10 to 10 dB. BER now is 8 x 10-3
. As for BER and FER,
the following results occurred:
No Impulse noise Impulse noise
Non
Interleaved
Interleaved
Non
Interleaved
Interleaved
Total Errors 986 910 1.326 x 10
4
1.346 x 10
4
Total Bits 1553 x 10
6
2.856 x 10
5
1.553 x 10
6
1.552 x 10
6
BER 0.000635 0.000586 0.008541 0.008674
#Error Frames 60 68 0.1111 92
Total Frames 2001 2001 2001 2001
FER 0.02999 0.03398 369 0.04598
Table 3. Simulation results (BER - FER)
IV. HOW INP AFFECTS CURRENT SYNCHRONIZATION
SPEED IN PRACTICE
A. Line Profiles
DSL lines are configured to a profile according to the
operating environment, the loop quality and the service
agreement. The ISPs technicians can configure each line to a
specific profile by accessing the DSLAM remotely through a
centralized system or automatically. [6]
A line profile consists of the following settings both for
upload and download:
Min/Max rate (Kbps)
Min/Max SNR
Target SNR
Impulse Noise Protection
Interleaving (Y/N)
Max Interleave delay
Many DSLAMs also support automated methods of
applying the best matching profile by measuring line's
characteristics based on historical performance. This method
is usually referred as Dynamic Line Management (DLM).
A greek ISPs internal DSLAM management system will be
used in this paper to test a customer's line to different profiles.
B. Applying profiles
For the needs of this assignment, access to a greek ISP's
internal DSLAM management system was granted. The
ADSL Measure tool was used to measure the effect of
interleaving and INP on a customer's line. The customer's
router synchronized at 10.255 Kbps (downstream) and 995
Kbps. So, profiles locked at Max. downstream rate 12.082
Kbps were used. The following line profiles were tested:
12 Mbps with no interleaving (Fast Path)
12 Mbps with Interleaving at 16ms
12 Mbps with Interleaving at 16ms and INP=1
12 Mbps with Interleaving at 16ms and INP=2
12 Mbps with Interleaving at 16ms and INP=4
12 Mbps WDLM no interleaving (Fast Path)
Table 4. Profile characteristics
In Table 4, each profile's configuration characteristics can
be found.
Screenshots from the ADSL Measure system:
Fig 12. 12 Mbps Fast path - No INP
7. Impulse Noise over xDSL Systems – MSc Networks & Data Communications - Kingston University London 7
Fig. 13. 12 Mbps with Interleaving - No INP
Fig 14. 12 Mbps with Interleaving - INP = 4
During the procedure, the following measurements were
taken:
Table 5. Live Measurements
V. RESULTS & DISCUSSION
Impulse noise is intermittent by nature, and much harder to
detect and analyze than traditional interference. It is a major
impairment for xDSL systems and can cause instability, loss
of synchronization, jitter and delay to services such as IPTV
and NoIP. Table 3 shows that a burst noise added to the signal,
can cause dramatic increase in BER. More corrupted data will
cause de-synch or corruption of image and sound that is
broadcasted through the xDSL system. Figure 11 confirms that
at the time an impulse noise occurs, data are completely
scattered and corrupted. Interleaving alone is not enough to
mitigate the effect.
ISPs use different INP methods to mitigate the effect of
burst noises. Choosing the right profile for each line is
significant for the line's stability and the ability to stream
video and voice services. The right profile would be the
profile with higher INP, lower delay, and with as high as
possible Maximum bit rate.
In Table 4, the tested profiles' characteristics are shown.
Choosing to deploy one of these profiles, sets the profiles
parameters in effect on the DSL system. The downstream
direction carries the IPTV picture and the VoIP sound and so
it is more important to protect against impulse noise than the
upstream direction. Higher INP settings will provide better
error protection if supported but may adversely impact
achievable bit rates and latency. Table 5 proves that as INP
rises, the Current Synchronization speed gets lower. Using the
WDLM Fast-path profile gave a higher data rate but left the
system exposed to impulse noises. The reason this profile
gives higher data-rates in comparison with the 12M_FAST
profile is that the 12M_FAST profile, has target SNR of 10 dB
while the WDLM Fast-path profile has target SNR of 8dB.
The WDLM profile will change automatically the target SNR
and enable interleaving when it will monitor the line's de-
synchronizations. This specific line seems to have de-synch
problems and with Interleaving enabled at the same target
SNR with the 12M FAST profile it seems to have a better
current synchronization data rate with INP of 1.5 symbols
(downstream) only from the interleaving factor. Since this line
is susceptible to outer burst noises, the best setting for it would
be the 12M INP 2 0 profile which would give 11344 Kbps
with 2 symbols protection and less delay than INP 4 but not
much more than INP 1.
Service providers and vendors thoroughly research their
environment in order to establish the appropriate profiles for
their own specific deployment. They test the performance of
any combination of parameters used in their profiles in order
to identify any impairments and adjust their profiles
accordingly.
8. Impulse Noise over xDSL Systems – MSc Networks & Data Communications - Kingston University London 8
REFERENCES
[1] J.-J. Werner, “Impulse noise in the loop plant,” in Proceedings of the
IEEE International
[2] W. Henkel, T. Kessler, “An Impulse-noise model - a proposal for
SDSL”, ETSI WG TM6 TD45 992T45A0, 1999
[3] D.B. Levey, S. McLaughlin, “Statistics of impulse noise: Interarrival
times”, ETSI WG TM6 TD19 993T19A0, 1999
[4] W. Henkel and T. Kessler, “Statistical description and modeling of
impulsive noise on the German telephone network,” IEE Electronics
Letters, vol. 30, pp. 935 – 936, June 1994
[5] Nedko H. Nedev, "Analysis of the Impact of Impulse Noise in Digital
Subscriber Line Systems", The University of Edinburgh
[6] Baofeng Frank Jiang, "Automated DSL Performance Adjustment"
United States Patent, Patent N0.: US 7,272,209 B2, Sep. 18, 2007
[7] Rahul Garg, Sunita Meena, Hemant Samdani and Patrick Duvaut,
"Impulse Noise Protection Initiatives in VDSL2 Systems, Conexant
Systems Inc., Feb. 2008
[8] Gavin Young, Dov Zimring, "TR-176 ADSL2Plus Configuration
Guidelines for IPTV", Broadband Forum - Technical report, Sep. 2008