Professional Documents
Culture Documents
Abstract — This paper analyses an example of estimation and However, there are papers which were closer to observing
calculation of channel capacity of twisted pair channel in DSL primary parameters, like modelling the series impedance of a
environment, from point of view of provider, instead of quad cable for DSL Applications [7], or evaluation of transfer
commonly used point of view of manufacturers of terminal function by introducing a method of balancing star-quad [8].
equipment. A brief overview of twisted pair as communication They do not place primary parameters in focus of interest,
channel was given. The channel capacity end-to end with indeed.
primary parameters as variables was introduced. The results of
measurement on real cable in commercial exploitation and its The goal of this paper is try to estimate twisted pair channel
comparison with introduced theoretical approach were shown. capacity related to primary parameters of that pair. This
The 20-pair, 1000-meter-long cable, completely buried, which estimation was done on real cable in commercial exploitation.
had been occupied by 100% by DSL signals and operating in The observed cable was operating in maximum load condition,
rural conditions, was used. i.e. all pairs were occupied by active triple-play ADSL2+
signals. Such cable was an ordinary cable, with no xDSL
Keywords — access network; channel capacity; digital improvements [9].
subscriber line; primary parameters; spectral efficiency; signal to-
noise ratio; transfer function; transmission line; twisted pair. A theoretical analysis is introduced at the beginning.
Simplified mathematical calculations, based mainly on
I. INTRODUCTION Shannon formula were done. Then, the transfer function of
twisted pair related to its primary parameters was introduced.
It is a well-known fact that number of broadband users has By simplification, the spectral efficiency, instead of channel
increased rapidly in recent years. Digital subscriber line capacity, was used. The worst condition case was assumed.
remains the most popular broadband access technique, After theoretical analysis, the results of measurement on
especially in the Europe [1]. DSL technology in distribution observed object were given. The discussion both theoretical
part of access network has been using “old-fashion” copper and practical results, and its comparison was done, followed by
based cables, developed for narrowband technologies only, so the most important conclusions.
it is necessary to identify if a twisted pair is suitable for a
certain DSL applications, with satisfied QoS and QoE levels.
However, channel capacity is global quantity which may be II. THEORETICAL APPROACH
used in order to evaluate such suitability. In order to calculate the capacity of single twisted pair, we
shall consider, at first, a transmission line connected to the
There are many papers that propose various methods of source of electromotive force ES and impedance ZS, loaded
estimation or analysis of twisted pair channel capacity in DSL with impedance ZL, as shown in Figure 1.
environment. For example, capacity of twisted pair channel,
with appropriated simulation was analysed in [2], while [3] IS IL
describes a frequency domain approach of channel capacity
+
+
+
both on source-port and load-port are random processes ξ(t)
and η(t), with appropriate power spectral densities Sξξ(ω) and ZS
Sηη(ω), respectively. In that sense, the noise is also random US Y UL ZL
process n(t) with corresponding power spectral density (PSD)
Snn(ω).
ES
If Ck is capacity of kth (k = 0, 1, 2, …, n-1, n – total number
of subcarriers) subcarrier from observed DMT system, and if
we, for simplification, neglect adjacent-channel interference,
then the total capacity is
Figure 2. Twisted pair end-to-end as two-port network
k
2
Z L
1 S H
Ck log 2 1 k d
Snn Z L 1 Z L Y Z S Z
2 k
k
1
where underbars denote complex quantities, and where we
where ωk1 and ωk2 are the lowest and the highest frequencies on defined the total series impedance and shunt admittance of
kth subcarrier. twisted pair end-to-end,
SNR k
Ck B0 log 2 1 so, after substitution (8) in (4), it follows that
k
6
12
5
10
bitload [b/s/Hz]
4 8
6
3
4
2
150 160 170 180 190 200 210 220 230 240 250
2
channel number
0
Figure 4. Estimated bit load from 151st to 250th channel
0 50 100 150 200 250 300 350 400 450 500
channel number
5
Pair OK
4.5 False G Figure 6. Comparison of estimated and achieved bit load
False C
4
False R Figure 6 shows a graphical comparison between spectral
False G&C
3.5 efficiency estimated by introduced model and obtained by
measuring on real object. As we see from Figure 6, there is
bitload [b/s/Hz]
3
significant mismatch between two graphs. It can be explained,
2.5 at first, by initial assumption that both source power and noise
2
power are constant all over frequency band. Also, we operated
with the worst case – the lowest source power and the highest
1.5 noise power. Since we know how DSL works, and according to
1 the fact that the assumptions do not remain in DSL system, we
could obviously expect noted difference, i.e. mismatch.
0.5
0
350 360 370 380 390 400 410 420 430 440 450 IV. CONCLUSION
channel number
Based on introduced theoretical approach and measurement
st th on real system in commercial exploitation, we can conclude
Figure 5. Estimated bit load from 351 to 450 channel
that the presented analysis may not be usable for estimation of
channel capacity in DSL systems indeed, but it may certainly [3] C. Neus, P. Boets, L. V. Biesen, “Channel capacity estimation of digital
be usable for analysis of impact of twisted pair primary subscriber lines: a frequency domain approach,” International
conference on communications (ICC), Glasgow, UK, 2007, pp. 2676-
parameters on channel capacity in observed system. The 2681,
providers can take advantage of obtained results and [4] P. Chambers, C. Downing, H. Baher, “Bandwidth, spectral efficiency
conclusions in order to evaluate ability of its copper-based and capacity variation in twisted-pair cable”, Irish signals and systems
access networks to provide broadband access with satisfied conference (ISSC), Belfast, UK, 2004, pp. 77-82,
level of QoS and QoE [17]. [5] W. Yu, J. Cioffi: “FDMA Capacity of Gaussian multiple-access
channels with ISI”, IEEE Transactions on communications, vol. 50, №
The primary parameter that mostly impacts on channel 1, 2002, pp. 102-111,
capacity is shunt capacitance. Series resistance follows, and [6] K. Amutha, G. Kumar, R. Thirumurthy, J. Purohit, “Loop diagnostics
finally shunt conductance. For example, derogation of in DSL,” National conference on communications, Mumbai, India, 2008,
capacitance of twice decreases spectral efficiency like pp. 486-490,
derogation of conductance of several orders of magnitude. The [7] W. Fouber, C. Neus, P. Boets, L. van Biesen, “Modeling the series
providers can and need to take into consideration these facts impedance of a quad cable for common mode DSL applications”,
International Instrumentation and Measurement Technology Conference
when they perform operations on their access network (I2MTC), Victoria, Canada, 2008, pp.250-253,
structure, e.g. maintenance and/or fault repair, which means [8] V. Šćepanović, “Оценка качества многопарных кабелей в сетях DSL-
mounting, replacing or removing network elements, like cable доступа; Площади референтного и эксплуатационного
sections, joints, connection boxes, line sockets, etc. треугольников (критерий Шчепановича),” Электросвязь, № 8,
Moscow, Russia, 2010, pp. 1-6,
The used cable type and length (20-pair, approximately 1 [9] Đ. Kopitović, V. Šćepanović, “DSL cables in access networks”,
km) suggest that introduced analysis and results can be used at Telecommunications Forum TELFOR, Belgrade, Serbia, 2006, pp. 135-
least on cables up to 20-30 pairs and up to 1 km, which operate 138,
in ADSL2+ systems. Such conditions are satisfied commonly [10] P. Golden, H. Dedieu, K. S. Jacobsen, “Fundamentals of DSL
at rural and sub-urban areas, where connection points more technology,” CRC Press., ISBN 0-8493-1913-7, 2006,
than 20-30 pairs and longer than 1 km distance from MDF are [11] P. Golden, H. Dedieu, K. Jacobsen, “Fundamentals of DSL technology”,
rarity. Auerbach Publications, Taylor & Francis Group, Boca Raton, FL, USA,
2006,
This paper did not take into consideration common [12] A. Begović, N. Škaljo, N. Goran, “An Example of Modelling of Twisted
situation at urban areas – 50-pair and more cables and/or longer Pair Channel in DSL Environment”, unpublished,
than 1 km cables. Also, it did not consider a frequency band [13] T. Starr, M. Sorbara, J. Cioffi, P. J. Silverman, “DSL Advances”,
higher than ADSL2+ one. Further work may observe noted Pretince Hall PTR, 2002,
tasks. In addition, different real cable types (e.g. insulation [14] Telecommunications cables, Catalogue B 02 05 08 07, ELKA kabeli,
types, stranding elements, core bounding types, etc.) could be d.o.o., Zagreb, Croatia, pp. 38-65,
analysing. [15] D. Würtz, A. Klein, M. Kuipers, “Distributed Margin Optimization
Using Spectrum Balancing in Multi-user DSL Systems”, European
Signal Processing Conference (EUSIPCO), Bucharest, Romania, 2012,
REFERENCES pp. 1593-1597,
[1] “World broadband statistic: Q1/2012”, Point Topic, Ltd, London, UK, [16] A. Begović, N. Behlilović, N.Škaljo, “Methods of some Important
June 2012, ADSL Signal Parameters Estimate”, 31st International Convention
MIPRO, Opatija, Croatia, 2008, vol 3, Telecommunications and
[2] S. Huberman, “Ergodic capacity of a symmetric DSL channel”, Information (CTI), pp. 102-106,
technical report, McGill University, Electrical and computer
engineering, Montréal, Canada, 2010, [17] DSL Forum Technical Report 126: Triple-play Service Quality of
Expirience (QoE) Requirements, 2006.