International Journal of Computer and Electrical Engineering, Vol. 5, No. 1, February 2013
PLC Impulsive Noise in Industrial Zone:
Measurement and Characterization
Trung H. Tran, Dung D. Do, and Tue H. Huynh
is evident. Due to this conclusive observation, we propose to
use the alpha stable distribution as model for impulse
amplitude. Our approach is different from the known ones
which often use Gaussian mixture to model this kind of noise
[8]. In our work, we study the time varying characteristics of
noise parameters, including amplitude statistic parameters,
duration and inter-arrival time.
Abstract—Noise in PLC is of relatively complex structure of
which the most important component is the asynchronous
impulsive noise. Based on measurements realized in an
industrial zone, this paper shows the heavy tail phenomenon
observed in experimental measures. Consequently, the alpha
stable model is the most natural to be used in describing the
statistics of PLC impulsive noise in industrial zones.
Furthermore, the inter-arrival time is of Pareto distribution
while the duration of impulses follows a mixed exponential
distribution. All parameters of the alpha stable model as well as
the Pareto parameters and the mixed exponents are statistically
estimated. The variation of these parameters with time is also
discussed.
II. MEASUREMENT
Index Terms—Power line communication, impulsive noise,
measurement and characterization.
I. INTRODUCTION
Fig. 1. Measurement set-up
To design a good performance data transmission system
that operates on the power line networks (PLC), it is
necessary to know all impairments induced by the PLC
networks used as communication channels. It is well known
that PLC networks are complicated communication channels
[1]. Due to many interconnections and taps, it is quite natural
to model this kind of channels as random multipaths [2]-[4].
Furthermore, the mechanism generating additive noise is
much more complex; based on experimental measurements
as well as on physical modeling, additive noise on PLC is
divided into different categories of which the most important
component is the asynchronous impulsive noise that causes
serious flaw to the data transmission systems that employ
PLC as communication medium. The last few decades, many
researchers have given many tries to model this noise
component [5]-[7]. In our work, we are interested in
measuring, analyzing and modeling this impulsive noise
component in an industrial zone. Our measurements have
shown that in such environment, the general characteristics
are still observed (high amplitude, bursty) but in industrial
zone, impulsive noise is of heavy tail. We organized a very
intensive and complete campaign of measurement in order to
obtain a very rich set of quite representative noise samples.
Based on this experimental result, the heavy tail phenomenon
We have conducted a measurement campaign in Do Son
Industrial Park for 2 weeks, three times per day (i) from 8AM
to 10AM; (ii) from 11AM to 1PM, and (iii) from 2PM to
4PM. Firstly, the electrical signal is extracted from the AC
220V 50Hz power line by an isolated coupling circuit and
sampled at the rate of 500MHz by DSO8502, which can store
1,045,487 samples in its 2MB RAM. To avoid recording too
much data, the trigger level is set to 960mV and the recording
time of each measurement is 524µs. Results are then
transferred to a PC and processed by a Matlab program.
III. MEASUREMENT RESULTS
A. Patterns of Measured Pulses
30
25
20
Magnitude
15
10
5
0
-5
-10
Manuscript received October 10, 2012; revised November 29, 2012.
Tran Huu Trung is with the Haiphong Private University, Vietnam
(e-mail: trungth@hpu.edu.vn).
Do Duc Dung was with the Bacha International University, now with the
Samsung Electronics Vietnam (e-mail: dddo@bhiu.edu.vn).
Huynh Huu Tue was with the Department of Electrical and Computer
Engineering, Laval University, Canada, now with the School of Electrical
Engineering, International University, HCM, Vietnam (e-mail:
hhtue@hcmiu.edu.vn).
DOI: 10.7763/IJCEE.2013.V5.660
0
100
200
300
Time us
400
500
600
Fig. 2. A single decayed impulse
Fig. 2 shows a single decayed impulse. It has a vertical
jumping
at
start
and
reduces
by
the
exponentially-decaying-bound sinusoidal function. The
pulse peak measured is up to 25V. We have observed 871
48
International Journal of Computer and Electrical Engineering, Vol. 5, No. 1, February 2013
pulses of this form during an hour and 48 minutes, which
corresponds to a frequency of 0.1344 impulses/minute. The
appearance of 25V-peak pulses at such high frequency can be
seen as of heavy tail.
30
20
10
Magnitude
15
10
0
-10
Magnitude
5
-20
0
-30
0
200
400
600
800
Time us
-5
-10
100
200
300
Time us
400
500
600
1400
Fig. 4 and 5 show bursty impulsive noises whose form is
not clearly determined[9], [10]. It composes of many single
impulses and their magnitude can be high up to 25V. These
high-magnitude impulses can seriously affect the
performance of PLC systems.
An observed sample (see in Fig. 3) shows us 3 impulses
with interval time between each of them is rather small, from
100µs to 200µs. As the results of the measurement, we found
that the industrial zone has affected strongly by impulsive
noise with the high frequency of occurrence and the high
magnitude. It can be considered as of heavy tail.
B. Magnitude of Impulse
As previously discussed, given the heavy tail characteristic
of impulsive noise in our measrement, we propse to use the
alpha model [11] for industrial zone PLC impulsive noise.
Characteristic function φ(t) of alpha stable model is defined
as follow:
12
10
exp .[t ] .[1 i sign(t ).tan 2 ] i t 1
(t )
exp .[t ].[1 i 2 sign(t ).ln(t )] i t 1
8
6
Magnitude
1200
Fig. 6. Occurrence frequency of bursty impulsive noise
0
Fig. 3. Example of frequency of impulse
4
2
where (0, 2]; [1,1]; 0; R;
0
-2
1 t 0
sign(t ) 0 t 0
1 t 0
-4
-6
0
100
200
300
Time us
400
500
600
Fig. 4. Negative impulse
0.35
Fig. 3 shows a negative impulse. This impulse does not
have the form of decreasing-bound impulse as described in
[1], [4]. The existing of negative impulses requires us to
study on Aimp- and Aimp+ seperately while building a
magnitude distribution model.
1
M easured data
Fitstable density
0.3
0.9
E m piricalC D F
E stim ated C D F
0.8
0.25
0.7
0.6
0.2
PD F
CDF
-8
1000
0.15
30
25
0.5
0.4
0.3
0.1
0.2
20
0.05
0.1
Magnitude
15
0
-20
10
-10
0
10
20
0
-20
-10
0
M agnitude (V )
10
5
Fig. 7. PDF and CDF of magnitude of impulse measured from 11:00 to 13:00
0
at 29/7/2011
-5
-10
0
200
400
600
800
Time us
1000
1200
Using the Maximum likelihood technique, we estimated
the impulsive noise PDF function. The obtained values of the
alpha stable distribution are α0=1.8236,
0=0.3109,
=0.78691,
δ
=0.46598.
This
distribution
with
0
0
0 >0
expresses the deviation clearly tended to the positive
1400
Fig. 5. Bursty impulsive noise
49
International Journal of Computer and Electrical Engineering, Vol. 5, No. 1, February 2013
direction.
The parameters of the function y=a×exp(b×t)+c×exp(d×t)
are a=0.6601, b=-0.0004575, c=0.03579, d =-8.866e-006.
TABLE I: THE FITTING RESULTS USING ALPHA DISTRIBUTION FOR DATA
RECORDED DURING 29/JULY/2011 AT DIFFERENT TIME PERIODS.
α0
1.25924
11:00 - 13:00
14:00 - 16:00
0
δ0
0
0.16
0.344402
0.675493
1.8236
0.758641
0.3109
0.78691
0.46598
1.8284
1.42397
0.455938
0.45939
0.14
0.8
0.7
PD F
0.1
M easured data
Fitstable density
0.9
0.3
0.04
0.2
0.02
0
0.1
0
0.5
E m piricalC D F
E stim ated C D F
E m piricalC D F
E stim ated C D F
0
-5
2
4
Tim e (s)
6
8
x 10
-6
Fig. 10. PDF and CDF of inter-arrival time
0.6
CDF
0
The pdf of the inter-arrival time of the results measured
from 8AM to 10AM, as shown in Figure 10, is appropriate to
the Pareto distribution with K=0.099γ5γ và σ=γ.165βe-7.
0.1
0.08
1
x 10
0.7
0.12
0.5
0.4
0.06
0.8
0.14
PD F
0.6
0.08
1
0.16
0.9
0.12
Table I shows the location parameter δ0 does not change
much while the deviation 0 change values according to
different time periods. The slight change of parameters of α0,
expresses that the distribution shape is not very sensitive to
time frames.
The distribution of total magnitude at different time
periods is shown in Fig. 8.
0.18
1
M easured data
Fitpareto density
CDF
Time
8:00 - 10:00
D. Inter-Arrival Time
0.5
0.4
0.06
0.2
0.02
0
-20
TABLE III: THE PARETO-FITTED RESULTS OF INTER-ARRIVAL TIME OF THE
DATA RECORDED DURING 3 DAYS (28, 29 & 31/JULY/2011).
0.3
0.04
0.1
-10
0
10
0
-20
20
-10
0
M agnitude (V )
10
Fig. 8. Amplitude distribution for data obtained on the 29/7/2011
Date
K
Σ
28/7 (Working day)
29/7 (Working day)
31/7 (Sunday)
0.02112
-0.02759
0.1059
3.9791e-7
3.8172e-7
7.5649e-7
TABLE II: THE FITTING RESULTS WITH DATA RECORDED IN DIFFERENT
DAYS
Date
α0
28/7
29/7
1.11063
1.47117
0.027578
0.895837
0.265596
0.415185
0.285798
0.583091
31/7
0.853063
0.657667
0.113396
-0.382411
0
IV. CONCLUSION
δ0
0
We have organized a very intensive campaign of
measurement of PLC asynchronous impulsive noise in an
industrial zone. Through the experimental measurement
results, we have observed the heavy tail phenomenon for the
impulsive noise amplitude. For this observation, we proposed
to use the alpha stable model for the amplitude distribution
and the Pareto distribution for the inter-arrival time. We have
also proposed a mixed exponential distribution for the
impulse duration. When fitting with estimated parameters,
the concordance between measured data and the theoretical
curves is good
C. Duration of Impulse
Analyzing the behavior of the impulse duration, it is clear
that histogram of the impulse duration follow different
exponential shape. To show that this type of mixed
exponential distribution can be used with confidence, we
have fitted many different situations of which the fitted result
have shown that this model is a good choice. Fig. 9 shown a
sample of the duration of impulse where we have proposed
the sum of two exponential functions to fit the result.
REFERENCES
[1]
Histogram time impulse
0.25
Measured data
Fit type exp2
0.2
[2]
0.15
[3]
0.1
[4]
0.05
0
[5]
0
1
2
3
Time (ns)
4
5
6
5
x 10
Fig. 9. Pdf of impulse duration
50
M. Zimmermann and K. Dostert, “An analysis of the broadband noise
scenario in power-line networks,” in proc. of International Symposium
on Power-Line Communications and Its Applications, Limerick, pp.
131-138. , 5-7 April 2000.
M. Zimmermann and K. Dostert, “A multi-path model for the power
line channel,” IEEE Transactions on Communications, vol. 50, no. 4,
2002, pp. 553-559.
H. Philipps, “Development of a statistical model for power line
communications channels,” International Symposium on Power-Line
Communications and Its Applications, Limerick, pp. 153-160. , 5-7
April 2000
H. Chaouche, F. Gauthier, A. Zeddam, M. Tlich, and M. Machmoum,
“Time domain modeling of powerline impulsive noise at its source,”
Journal of Electromagnetic Analysis and Applications, pp. 359-367. ,
March 2011
V. Degardin, M. Lienard, P. Degauque, A. Zeddam, and F. Gauthier,
“Impulsive noise on indoor power lines: characterization and
mitigation of its effect on PLC systems,” IEEE International
Symposium on Electromagnetic Compatibility, Istanbul, vol. 1, pp.
166-169. 11-16 May 2003,
International Journal of Computer and Electrical Engineering, Vol. 5, No. 1, February 2013
F. Kural and M. Şafak, “An experimental investigation of impulse
noise on low voltage powe rlines,” International Symposium on
Power-Line Communications and Applications, Athens, Greece,
27-29, March 2002.
[7] M. H. L Chan and R. W. Donaldson, “Amplitude, width and
interarrival distributions for noise impulses on intrabuilding power line
communication networks,” IEEE Trans. Elec. Compatibility, vol. 3, no.
3, pp. 320-323, August 1989.
[8] M. Nassar, K. Gulati, Y. Mortazavi, and B. L. Evans, “Statistical
modeling of asynchronous impulsive noise in powerline
communication networks,” in Proc. of IEEE Global Communications
Conference, Houston, USA. , Dec. 5-9, 2011
[9] J. A. Weyepe Tina, A. J. Snyders, and H. C Ferreira, “Implementation
of a gap recorder for measuring impulsive noise error distributions in
power line communications using the fritchman model,” International
Symposium on PowerLine Communications and Applications, pp.
27-30, March 2012.
[10] P. Pagani, R. Hashmat, A. Schwager, D. Schneider, and W. Bäschlin,
“European MIMO PLC field measurements: noise analysis,”
International Symposium on PowerLine Communications and
Applications, 27-30, March 2012.
[11] G. Samorodnitsky and M. S. Taqqu, Stable Non-Gaussian Random
Processes: Stochastic Models with Infinite Variance, Chapman and
Hall, New York, 1994.
[6]
Dr. Do Duc Dung was born in Hanoi, Vietnam in 1979.
He received M.Sc. and Ph.D. degree in information and
computer technology from Chungbuk National
University, Korea in 2004, and 2007, respectively. From
2008 to 2011, he was a lecturer of Bac-Ha International
University, Vietnam. Since 2012, he has been a
researcher in Samsung Electronics Vietnam. His current
research interests include the signal processing, software
solution.
Prof. Huu Tue Huynh received the Sc.D. degree in
1972 from Laval University (Canada) where he had
been a Faculty member of the Department of Electrical
and Computer Engineering since 1969. He was an
Invited Guest at The AT&T Information Systems in
Neptune, N.J. in 1984 and has been invited to give
lectures at several Universities in Europe, America as
well as in Asia. Professor Huynh is author and
coauthor of two books and more than two hundred papers and reports in
Information Processing. He has served as Consultant to a number of
Canadian Government Agencies and Industries. His research interests cover
stochastic simulation techniques, information processing, fast algorithms and
architectures with applications to finance and to communications. In 2005,
he left Laval University to create the Department of “Information
Processing” at the College of Technology, VNU, Hanoi. During the period
2007-2011, he was invited to set up Bac-Ha International University, Hanoi,
as her first President. Professor Huynh is now working as a research
professor at the School of Electrical Engineering of VNU- HCM’s
International University, where his main responsibility is creating a new
research group in “Intelligent Signal Processing”. He is the Technical
Editor-in-Chief of “REV-Journal on Electronics and Communication.
Tran Huu Trung was born in Haiphong, Vietnam in
1977. He received B.Sc. in Electronics Engineering from
Haiphong Private University, Vietnam, in 2001 and
M.Sc. degree from University of Engineering and
Technology, Hanoi, Vietnam in 2005. From 2001 to
2007, he was an lecturer in the Department of
Electronics, Haiphong Private University, Vietnam. His
research interests include signal processing, powerline
communications.
51