ISSN 1975-8359(Print) / ISSN 2287-4364(Online)

The Transactions of the Korean Institute of Electrical Engineers Vol. 65, No. 9, pp. 1583 1589, 2016
http://dx.doi.org/10.5370/KIEE.2016.65.9.1583

개선된 SSTDR을 이용한 케이블 고장 검출과 위치 계산

Detection and Location of Cable Fault Using Improved SSTDR

전 정 채* ․ 김 재 진 * ․ 최 명 일 *
(Jeong-Chay Jeon ․ Jae-Jin Kim ․ Myeong-Il Choi)

Abstract - This paper proposes an improved spread spectrum time domain reflectometry (ISSTDR) using time-frequency
correlation and reference signal elimination method in order to have more accurate fault determination and location detection
than conventional (SSTDR) despite increased signal attenuation due to the long distance to cable fault location. The proposed
method has a two-step process: the first step is to detect a peak location of the reference signal using time-frequency
correlation analysis, and the second step is to detect a peak location of the correlation coefficient of the reflected signal by
removing the reference signal. The proposed method was validated through comparison with existing SSTDR methods in open-
and short-circuit fault detection experiments of low voltage power cables. The experimental results showed that the proposed
method can detect correlation coefficients at fault locations accurately despite reflected signal attenuation so that cable faults
can be detected more accurately and clearly in comparison to existing methods.

Key Words : SSTDR, Cable fault location, Time-frequency correlation, Reference signal elimination

1. Introduction such as radar is injected into a cable, thereby measuring a

reflected signal produced due to the mismatch of characteristic
Power cables are susceptible to many faults such as impedance from fault location.
insulation damage, open fault, or short circuit due to Reflectometry utilizes the principle of reflection of electro-
inappropriate installation and other various physical, electrical, magnetic waves flowing through a cable when characteristic
or environmental factors, which could lead to electrical fires. impedance changes. A fault caused by open fault can be
According to the electrical accident statistics published by the identified due to the fact that a generated reflected wave has
Korea Electrical Safety Corporation, 20% of electric facility in-phase with the injected signal (or reference signal) due to
accidents are caused by cables, and most electrical fires are larger impedance than the characteristic impedance of the
related to cables[1]. transmission line. On the other hand, out-of-phase is generated
Cable fault can result in power outage and fire, causing a in case of a fault caused by short circuit due to smaller
number of problems such as property damage, information impedance than characteristic impedance. The location to the
loss, or production delay. Therefore, it is necessary to prevent fault can be calculated by measuring the time elapsed to detect
such accidents through accurate detection of fault locations a reflected signal.
and rapid restoration as soon as a cable fault is detected. Typically, reflectometry can be divided into two categories:
Furthermore, it is also important to reduce cost and shorten TDR (Time Domain Reflectometry), in which signals reflected
the restoration time through accurate location measurement. using a pulse after injected signals are analyzed in the time
Many methods have been proposed to diagnose cable domain, and FDR (Frequency Domain Reflectometry), in
faults and detect their locations such as partial discharge which frequency signals are employed and reflected signals
measurement. Among them, reflectometry has been most are analyzed in the frequency domain[2-3].
commonly used for fault detection, in which a specific pulse As for the TDR, it is difficult to extract reflected waves
accurately in a noisy environment, and its performance is
Corresponding Author : Electrical safety Research Institute, dependent on the resolution in the time domain of related
Korea Electrical Safety Co., Korea equipment. As for the FDR, its performance is dependent on
E-mail: [email protected] the size of measurable frequency bandwidth. In addition, the
* Electrical Safety Research Institute, Korea Electrical Safety
disadvantage is the high error rate in phase measurement
Co., Korea
Received : July 28, 2016; Accepted : August 8, 2016 when there is noise[4].

Copyright ⓒ The Korean Institute of Electrical Engineers 1583

This is an Open-Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/
licenses/by-nc/3.0/)which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
전기학회논문지 65권 9호 2016년 9월

TFDR (Time Frequency Domain Reflecometry), STDR maximum length is     , which can be produced by
(Sequence Time Domain Reflectometry) and SSTDR (Spread using a primitive polynomial whose maximum degree is  .
Spectrum Time Domain Reflectometry) have been studied to When the auto-correlation value at a periodically repeated
minimize measurement error and facilitate easy fault detection sequence is ≠ , it has a value below  .
[5]-[10]. SSTDR among these methods is known to be robust
to noisy environments and can detect intermittent fault
including open fault and short circuits in dead and live wires
[4]. However, if a sequence of specific length is used in the
SSTDR, the attenuation of injected signal becomes larger
thereby reducing reflected signals as the fault location
becomes more remote, which makes fault detection difficult
and measurement error larger.
Therefore, this paper proposes an improved SSTDR technique
consisting of two steps: peak value of the correlation coefficient
of the reference signal is detected using time-frequency
correlation analysis, and then a peak value of the correlation
coefficient of the reflected signal is detected after removing the
reference signal to solve the problem of inaccurate fault
detection due to signal attenuation. The performance of the
proposed method was evaluated via comparison of existing Fig. 1 The principle of SSTDR
methods during an experiment involving low-voltage cables.
The performance evaluation showed that the proposed method
In the SSTDR, the reference signal  , which is applied to
can identify whether a fault occurred more accurately and can
a cable, is produced by a product of MLS c    ⋯   
track fault locations better than existing methods despite signal
 ∈   whose length is and carrier signal. So,  can be
attenuation.
expressed by



2. Conventional SSTDR         


   (1)

The SSTDR was developed by P. Smith and Professor where  is the sequence length K containing the amplitude
Cynthia Furse at the University of Utah in the USA, as the of +1 and -1 Loop Linear Recursive Sequence (RLS) and
Direct Sequence Spread Spectrum (DSSS) technique was used carrier signal    can be expressed by
in digital communication cables for the purpose of fault
location detection in communication lines[7]. Here, the Spread
Spectrum (SS) technique has mainly been used in
    cos

  ≦   



(2)

communication systems, which has an advantage of improving

resolution or suppressing interference using a wider Therefore, the cycle of  is a recursive linear signal
bandwidth of actual transferred signals than that which is (RLS) of period    consisting of 1s and -1s,  is the
required for information transmission. minimum duration of a 1 or -1 known as a “chip”, and
SSTDR determines fault location and type by dispersing a

band using a pseudo-noise sequence (PN sequence), which    .

has reasonable auto-correlation properties, applying
phase-shift keying signals into power lines, and detecting an If a signal in Eq. (1) is applied to a cable, reflection occurs
arrival time and phase of the signal returned via reflection at the change point of cable impedance. The reflected signal
from the fault location as shown in Fig. 1. experiences a certain time delay, so the reflected signal 
A sequence used to detect a cable fault in the SSTDR is can be expressed as
mostly a Maximum Length Sequence (MLS), which has been

known to have superior detection performance of fault          
  (3)
location[8]. An MLS is a binary pseudo-noise sequence whose 

1584
 Trans. KIEE. Vol. 65, No. 9, SEP, 2016

where  is the amplitude of reflected signal     ,  is value of the correlation function of the reflected signal is
the time delay before receiving reflection  and  is a found via the second time-frequency correlation function
noise signal.    of  and  . Finally, a time difference     
To calculate the distance to the cable fault location in the between the peak values is calculated to obtain the distance
SSTDR, a time difference of maximum values of the time to the cable fault location.
cross-correlation function   is calculated first using an
applied signal  and a reflected signal  . So, the time
cross-correlation output is

 

   
 
   
 

          
  (4)
   
 

  
  

where * refers to a complex conjugate.

Cable fault location using a time difference obtained via
Eq. (4) and VOP (Velocity of Propagation) which is a variable
that represents the propagation velocity of the electro-
magnetic wave in a corresponding cable is given by Fig. 2 The proposed SSTDR

 ×    The time-frequency correlation analysis in the SSTDR

   (5)
 proposed in this study employs the Wigner Ville Distribution
(WVD) to analyze the reference and reflected signals in the
where  is VOP,  is detection time of reflection signal, and
time-frequency domain[5], [10]. First, the reference signal (1)
 is start time of reference signal.
and reflected signal (3) are transformed into the WVD for the
time-frequency correlation analysis, thereby producing Eq. (6)
and Eq. (7).
3. Proposed Method


In general, a time difference is obtained via time

    

 

 
 exp   
  (6)

correlation functions, thereby calculating a fault location in 

exp   

the SSTDR. Since it analyzes signals in the time domain in
    


 

(7)

the time correlation function, it has difficulties in detecting

the accurate location of reflected signals because a side lobe where * refers to a complex conjugate, and it satisfies
in the correlation function with respect to applied signals is                    min     
larger than the main lobe in the correlation function with  
and   ⌊   ⌋   ⌊   ⌋ where ⌊⌋ refers to
respect to reflected signals when a fault location is far,  
thereby leading to large signal attenuation or to weakness in the largest integer that does not exceed  .
reflected signal due to a minor fault[11]. Next, the time-frequency correlation analysis of the
To solve this problem, this paper aims to improve cable reference and reflected signals with respect to the WVD can
fault distance detection performance in the conventional be calculated via the following equation to identify the cable
SSTDR by using time-frequency correlation analysis and fault location[5].
removing the reference signals as shown in Fig. 2.     

The proposed method in this study searches the maximum    

   
         
(8)
value location  of the reference signal from the time-
frequency correlation function    of reference and where and are the normalization factors that refer to energies
reflected signals to find a  is removed from the measured of the reference and reflected signals in the time-frequency
signal  to make         , and the peak  domain, respectively, which are expressed below[5]:

개선된 SSTDR을 이용한 케이블 고장 검출과 위치 계산 1585

전기학회논문지 65권 9호 2016년 9월

      coaxial cables have relatively stable VOP. On the other hand,
     
       
(9)
power cables, which are vulnerable to external environments,
      have less uniformed VOP, thereby causing error in distance
      
        
(10) calculation.
In this paper, VOP was measured first with respect to the
The maximum value of the time-frequency correlation experimental target cables prior to the SSTDR experiment. To
function for the reflected signal         that minimize measurement error as much as possible, a pulse
signal of 10, 100, 200, and 1000ns was injected into the
removes reference signal using Eq. (8) has a value 1 because
experimental target cable and averages of tenfold measurement
of normalization and reference signal elimination. A time
were computed. As a result, a VOP of 1.905×108m/s for the
difference between peak values of the correlation function of
target cable was obtained.
the reference and reflected signals is obtained and a distance
to the cable fault location is calculated using Eq. (5).
4.2 Results

This paper select the length of MLS is N=7, modulated by

4. Experimental Results
sinusoidal signal in the same frequency, as the test signal for
SSTDR. The amplitude of the reference signal is set to 5 V
4.1 Experimental Setup
and the sampling rate for reflected signal acquisition of the
SSTDR is 25 MS/s.
To validate the performance of the proposed SSTDR, an
experiment was conducted as shown in Fig. 3. An F-CV2C6SQ
2

cable was used for the experimental target cable because it has 1
reference

been most widely used in low-voltage power systems. The 0

-1

SSTDR experimental setup in Fig. 3 consists of a control unit, -2

0 50 100 150 200

an arbitrary waveform generator, a digital oscilloscope, and “T” sample (n)

connector. To automatically control the Arbitrary Waveform 1

measure

Generator (NI PXI 5422, 16bits, 200 MS/s) that generates a 0

-1

signal injected into a cable and digital oscilloscope (NI -2

0 50 100 150 200
sample (n)
PXIe-5162, 10bits, 5 GS/s, 1.5 GHz) that acquires a signal
40 56(+)
1
reflected from the cable fault point, the NI LabVIEW program 0.5
correlation

0
was developed and MATLAB was used to analyze correlations -0.5

-1

between reference and measured signals. In the experiment, 0 50 100 150 200
sample (n)

the reference and measurement signals were injected and

measured through the T connector and RG58 cables. (a)

1
reference

-1

-2
0 50 100 150 200
sample (n)

2 original
eliminated
1
measure

-1

-2
0 50 100 150 200
sample (n)

40 56(+)
1 1st corr

Fig. 3. Experimental setup for the proposed SSTDR 0.5

2nd corr
correlation

-0.5

-1

As described previously, accurate VOP is needed to 0 50 100 150 200

sample (n)

calculate cable fault location. In reflectometry, calculation of

fault locations can be affected by the measured time as well (b)
as the accuracy of the VOP. In general, communication cables Fig. 4. The results of F-CV2C6SQ 60 m open fault (a) The
that can maintain characteristic impedance uniformly such as conventional SSTDR method (b) The proposed method

1586
 Trans. KIEE. Vol. 65, No. 9, SEP, 2016

Fig. 4 shows the measurement result of the open fault of Note that the connector cable length for signal injection
the 60 m cable. Fig. 4 (a) shows the measurement result should be subtracted from the measured length (in this paper,
using the existing SSTDR in which accurate measurement of it was 2 m).
the reflected signal at the open fault point was difficult due Fig. 5 and 6 show the measurement results of the short
to the side lobe of the reference signal. Fig. 4 (b) shows the circuit fault using existing and improved SSTRD methods at
analysis result using the proposed method in this study, in distances of 120m and 153m. As shown in Fig. 5(a) and 6(a),
which accurate fault measurement and distance calculation in the conventional SSTDR, the further the fault location, the
can be achieved, as the correlation coefficient at the fault smaller the reflected signal and, as a result, the correlation
location was close to 1. coefficient became smaller, which made it difficult to
To calculate the distance to the fault location in Fig. 4, a distinguish the side lobe of the reference signal and other
sample difference between peak values of the correlation noises. This inability to discriminate the signal, in turn, led to
coefficient was divided by the sampling rate. That is, a difficulty in fault detection and high error detection rate.
difference of one sample is 1/25MS/s=0.04. In Fig. 3, the On the other hand, the proposed method in this paper
sample difference is 16(56-40), so the time difference can be removes the reference signal and employs the time-frequency
calculated as 16/25MS/s=0.64, and the distance to the fault correlation analysis using a larger correlation coefficient at the
location using Equations (5) can also be computed as fault location as shown in Fig. 5(b) and 6(b), thereby providing
more accurate fault detection and easier distance calculation.
 ×   × ×   The location of the fault can be estimated to be 119.92m and
        (11)
 150.4m (an error is within 1 %) respectively.

2
2
1
1
reference
reference

0
0
-1
-1

-2
-2 0 20 40 60 80 100 120 140
0 20 40 60 80 100 120 sample (n)
sample (n)

2
1

1
measure
measure

0
0

-1
-1

-2 0 20 40 60 80 100 120 140

0 20 40 60 80 100 120 sample (n)
sample (n)

2 7(-)
12 17(-) 1
1
0.5
correlation

0.5
correlation

0
0
-0.5
-0.5
-1
-1
0 20 40 60 80 100 120 140
0 20 40 60 80 100 120 sample (n)
sample (n)

(a) (a)

2
2

1
1
reference
reference

0
0

-1
-1

-2
-2 0 20 40 60 80 100 120 140 160 180
0 20 40 60 80 100 120
sample (n)
sample (n)

2 2
original
original
eliminated
eliminated 1
1
measure
measure

0
0

-1
-1

-2
-2
0 20 40 60 80 100 120 140 160 180
0 20 40 60 80 100 120
sample (n)
sample (n)

12 52(-)
12 44(-) 1 1st corr
1 1st corr
2nd corr
2nd corr
0.5
correlation

0.5
correlation

0
0
-0.5
-0.5
-1
-1

0 20 40 60 80 100 120 140 160 180

0 20 40 60 80 100 120
sample (n)
sample (n)

(b) (b)
Fig. 5. The results of F-CV2C6SQ 120m short fault (a) The Fig. 6. The results of F-CV2C6SQ 153m short fault (a) The
conventional SSTDR method (b) The proposed method conventional SSTDR method (b) The proposed method

개선된 SSTDR을 이용한 케이블 고장 검출과 위치 계산 1587

전기학회논문지 65권 9호 2016년 9월

The experimental results(Table 1) showed that the proposed References

method can identify whether a fault occurred more accurately
and can track fault locations better than conventional SSTDR [1] Korea Electrical Safety Corporation, “A Statistical
despite signal attenuation. Also, an error of automatic fault Analysis on the Electrical Accident,” 2015
type and location determination by detection of phase and [2] N. G. Paulter, “An Assessment on the Accuracy of Time-
peak value because of elimination of the reference signal and Domain Reflectometry for Measuring the Characteristic
normalization of correlation coefficient not occurred. Impedance of Transmission Lines,” IEEE Trans. on
Instrumentation and Measurement, Vol. 50, No. 5, pp.
Table 1. Experimental results of cable fault detection through 1381-1388 ,Oct.2001
SSTDR and ISSTDR [3] C. Furse. Y. C. Chung, R. Danglo, M. Nielsen, G. Mabey
SSTDR ISSTDR
and R. Woodward, “Frequency-domain refelectometry for
Fault on-board testing of aging aircraft wiring,” IEEE Trans.
Cable measurement Error measurement Error
Type
(m) rate(%) (m) rate(%) Electromagn. Compat., Vol. 45, No. 2, pp. 306-315, May.
Open 58.96 1.57 58.96 1.57 2003.
CV2C6SQ_60m
Short 58.96 1.57 58.96 1.57 [4] Cynthia Furse, "A Critical comparison of reflectometry
Open 116.11 3.7 119.92 0.07 methods for location of wiring faults," Smart Structure
CV2C6SQ_120m
Short Error Error 119.92 0.07
and Systems, Vol. 2, No. 1, pp.25-46, 2006
Open Error Error 154.21 1.12
CV2C6SQ_153m [5] Y. J. Shin, E. J. Powers, T. S. Choe, C. Y. Hong, E. S.
Short Error Error 150.4 1.38
CV3C10SQ_200 Open Error Error 198.824 0.59 Song, J. G. Yook and J. B. Park, "Application of Time -
m Short Error Error 202.686 1.34 Frequency Domain Reflectometry for Detection and
Localization of a Fault on a Coaxial Cable," IEEE Trans.
on Instrumentation and Measurement, Vol. 54, No. 6, pp.
5. Conclusion 2493-2500, Dec. 2005
[6] Chirag R. Sharma, Cynthia Furse and Reid R. Harrison,
This paper proposed employed the reference signal "Low-Power STDR CMOS Sensor for Location Faults in
elimination and time-frequency correlation analysis in order Aging Aricraft Wiring," IEEE Sensors Journal, Vol. 7, No.
to reduce the measurement error of cable fault detection and 1, pp. 43-50, Jan. 2007
distance calculation due to signal attenuation in SSTDR. The [7] P. Smith, “Spread spectrum time domain refletometry,”
performance of the proposed method was evaluated through Ph.D. dissertation, Dept. Elect. Comput. Eng., Utah Univ.,
open- and short-circuit fault detection experiments using Logan, 2003
60m, 120m and 153m low-voltage power cables. The results [8] Paul Smith, Cynthia Furse and Jacob Gunther, “Analysis
validated the superior performance of the proposed method. of Spread Spectrum Time Domain Reflectometry for Wire
The proposed method is expected to play an important role in Fault Loaction,” IEEE Sensors Journal, Vol. 5, No. 6, pp.
resolving the difficulties of cable fault detection due to 1469-1478, Dec. 2005.
various noises and reflected waveform distortion in real cable [9] Cynthia Furse, Paul Smith, Mehdi Safavi and Chet Lo,
fault detection sites when SSTDR devices, in addition to other “Feasibility of Spread Specturm Sensors for Location of
reflected wave measurement equipment, are utilized. In the Arcs on Live Wires,” IEEE Sensors Journal, Vol. 5, No. 6,
future, studies on cable fault detection performance according pp. 1445-1450,Dec.2005.
to the type and length of various sequences, such as Baker [10] Boualem Boashash, “Time Frequency Signal Analysis
and Frank code used in digital communication and radar and Processing,” ELSEVIER, 2003
systems, will be conducted in order to develop better [11] Jeong-Chay Jeon, Taek-Hee Kim and Jae-Geun Yoo,
reflectometry that provides easy fault detection and less error. "Fault Detction of Low Voltage Cable Using Time-
Frequency Correlation in SSTDR.", The trans. of the
KIEE, Vol. 64, No. 3, pp. 498-504, 2015.

감사의 글
This study was supported by "2013 Dual Use Techonology
Program".

1588
 Trans. KIEE. Vol. 65, No. 9, SEP, 2016

저 자 소 개

전 정 채(Jeong-Chay Jeon)
1997년 2월 원광대학교 전기공학과 졸업.
1999년 2월 동 대학원 전기공학과 졸업(석사).
2015년 8월 동 대학원 전기공학과 박사(졸
업). 현재 한국전기안전공사 전기안전연구원
책임연구원/팀장
E-mail : [email protected]

김 재 진(Jae-Jin Kim)
2001년 2월 원광대학교 전기공학과 졸업.
2010년 2월 동 대학원 전기공학과 졸업(석사).
현재 한국전기안전공사 전기안전연구원 선임
연구원
E-mail : [email protected]

최 명 일(Myeong-Il Choi)
2002년 2월 창원대학교 전기공학과 졸업.
2004년 2월 동 대학원 전기공학과 졸업(석사).
현재 한국전기안전공사 전기안전연구원 선임
연구원. 2014년 8월 숭실대학교 대학원 전기
공학과 박사(졸업).
E-mail : [email protected]

개선된 SSTDR을 이용한 케이블 고장 검출과 위치 계산 1589