This article has been accepted for publication in a future issue of this journal, but has not been

fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2019.2944346, IEEE
Sensors Journal

> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 1

Long-Term Reliability of Fiber-Optic Current

Sensors
Miklós Lenner, Andreas Frank, Lin Yang, Tomas Roininen, and Klaus Bohnert

such as circuit breakers leads to a substantial substation

Abstract—We present studies the long-term reliability of footprint reduction [7]. In addition, FOCS provide a more
interferometric fiber-optic current sensors (FOCS) for use in accurate image of the primary current waveform due larger
electric power transmission systems. Accelerated ageing tests are bandwidth and absence of magnetic saturation at fault currents.
performed on crucial optical sensor components and a three-phase
sensor system is subjected to an extended field trial. The sensor
Another important motivation for FOCS has been the transition
components under test include the sensor’s superluminescent light from analog to digital control and communication in electric
emitting diode light source, the integrated-optic phase modulator power substations.
and various passive components such as fiber couplers, fiber On the other hand, fiber-optic current sensors represent a new
polarizers, polarization-maintaining fiber connectors, and fiber technology in a conservative industry. Typically, conventional
coatings. The components are exposed to accelerated ageing current transformers have lifetimes in excess of 30 years.
conditions for extended periods of time, i.e., temperature cycling
(between -25 °C and 65 °C for up to 15000 cycles), constant high
Naturally, experience on the long-term performance of FOCS
temperature at dry conditions (up to 115 °C for up to 20000 h), is still limited. For use in electric power transmission, the
and damp heat (85 % relative humidity at 85 °C for up to 7900 h). sensors must be accurate to within ±0.2 % over wide
Crucial component parameters such as the source wavelength and temperature ranges, e.g., from -40 °C to 85 °C and ideally
polarization extinction ratios are repeatedly measured as a should maintain this accuracy over the entire operation time
function of temperature at defined intervals during the ageing without need of recalibration [10]. Also, sensor failures in the
periods and examined for potential drift of component failures.
The field trial is carried out for a three-phase FOCS system
field be avoided.
integrated into 420 kV double-chamber circuit breakers over a In this context, we carried out extensive tests on both
period of more than three years. The sensor signals are compared component and system level over a period of several years to
to the signals of conventional current transformers. In addition, prove the long-term reliability of modern FOCS under harsh
the evolution of various operational parameters such as the light environmental conditions. Most sensor components are
source power is continuously recorded. The results proof a high commercially available, some of them adapted for use in FOCS.
degree of reliability of modern FOCS systems.
Typically, the components underwent qualification tests by the
manufacturer with regard to the requirements of the
Index Terms—Current measurement, fiber-optic current telecommunication industry or other industries. Commonly,
sensor, magneto-optic effect, optical fiber sensors, reliability, those tests, however, only partially address the requirements of
thermal stability, accelerated ageing, system testing. FOCS. For example, the long-term stability of the source
wavelength may not be particularly critical in some applications
such as optical coherence tomography but is essential to FOCS.
I. INTRODUCTION Similarly, a polarization extinction ratio of the phase modulator
of 20-25 dB in a limited temperature range may be sufficient
I N RECENT years, fiber-optic current sensors (FOCS)
employing the Faraday effect in a coil of sensing fiber have
reached a high degree of maturity. Their main application is in
for use, e.g., in cable television applications but would be
inadequate for FOCS. In addition, results of simple pass/fail-
high voltage (HV) electric power transmission, where the new type qualification test do not deliver information on the
technology represents an alternative to the traditional inductive component’s lifetime. The present work is, to our knowledge,
instrument transformers [1]-[8]. Another important application the first comprehensive report on FOCS reliability and
is in the electro-winning of metals, e.g., at aluminum smelters corroborates the high maturity the technology has reached
or copper refining [9]. Here, the sensors measure direct currents today. We should mention that parts of the work have been
(DC) up to several hundred kiloamps. Important advantages of presented at various technical conferences in recent years as
FOCS include small size and weight and hence high application indicated by corresponding references [8], [11]-[15]. Here, we
flexibility (by contrast, traditional instruments can weigh up to present a comprehensive aggregation of our final findings with
a few tons). The integration of FOCS in other HV equipment a summary of insights and conclusions.

M. Lenner, A. Frank, L. Yang, and K. Bohnert are with ABB Switzerland Ltd, Voltage Products, 77180 Ludvika, Sweden (e-mail:
Corporate Research Center, 5405 Baden, Switzerland (e-mail: tomas.roininen@se.abb.com.
miklos.lenner@fhnw.ch, andreas.frank@ch.abb.com, lin.yang@ch.abb.com,
klaus.bohnert@ch.abb.com). T. Roininen is with ABB Sweden Ltd, High

1558-1748 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2019.2944346, IEEE
Sensors Journal

> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 2

The paper is organized as follows: Section II briefly presents high temperature at dry conditions and damp heat. The
the sensor principle and sensor configuration. Section III temperature cycles were performed between -25 °C and 65 °C
describes the employed test procedures. Section IV presents the (specified temperature range of operation of the opto-
results of the component tests with particular focus on the light electronics unit) at slopes of ±1.5 °C/min and included 20 min
source and the modulator as the two most critical sensor dwelling periods at the extreme temperatures, i.e., a full cycle
components. Other investigated components include fiber lasted 160 minutes. Thermal cycling serves to discover
couplers, fiber polarizers, polarization-maintaining fiber potential failure modes associated with thermo-mechanical
connectors, and fiber coatings. Section V shows data from a stress or fatigue at repeated differential thermal expansion of
field trial of a three-phase sensor system in a 420-kV-substation the involved materials. It is important to note that the
over a period of more than three years. Finally, Section VI temperature range for the accelerated tests was within the
draws some conclusions. components’ specifications. This low-acceleration condition
avoids triggering “trivial” failures that are associated with
II. SENSOR CONFIGURATION inappropriate operation. Tests at elevated (constant)
Fig. 1 shows the sensor configuration [16]-[18]. It is temperatures are well suited to address material ageing, e.g., of
worthwhile to note that similar schemes are being used by the fiber coatings and glues. Damp heat accelerates diffusion and
majority of commercial FOCS suppliers. A superluminescent oxidation processes and is known to trigger packaging and
diode light source (1310 nm) sends two light waves with material related failure modes [21]-[23].
orthogonal polarization directions through a polarization- Tested sensor components included light sources, phase
maintaining (PM) fiber link (Panda fiber) to a coil of sensing modulators, fiber polarizers, fiber couplers, photodiodes, PM
fiber operated in reflection mode. A fiber-optic quarter-wave fiber connectors, and fiber coatings. Typically, the tests
retarder made from elliptical-core fiber at the coil entrance included several samples of each component, sometimes from
generates left and right circularly polarized light waves. As a several suppliers. An initial characterization of the components
result of the Faraday effect, the circular waves accumulate provided reference values for the crucial component parameters
during their roundtrip through the coil a differential magneto- such as light source wavelengths, polarization extinction ratios,
optic phase shift Df = 4VNI, where V is the Verdet constant of insertion losses, etc. During the accelerated ageing, the
the fused silica fiber (1.0×10-6 mrad/A at 1310 nm), N is the components were repeatedly recharacterized at defined
number of fiber windings around the current conductor, and I is intervals (interval censoring) in order to determine drifts, if any,
the current to be measured. The two light waves return again as or, in the worst case, component failure.
orthogonal polarization states to the opto-electronics unit but In addition, a complete three-phase sensor system was
subjected to a long-term field trial. During the trial, the sensor
now have swapped polarization directions. A closed-loop
detection circuit with an integrated-optic birefringence signals, various auxiliary sensor parameters, the signals from
modulator recovers the magneto-optic phase shift, similarly as corresponding conventional current transformers, and
rotation-induced Sagnac phase shifts are measured in fiber- temperatures at various locations were recorded at periodic
optic gyroscopes [19]. intervals.
Orthogonal linear light waves
y x IV. ACCELERATED COMPONENT AGEING
Left and right
circular waves
A. Light source
Opto-electronics module

Light Phase modulator Fiber quarter-wave The FOCS light source is a 5-mW superluminescent light
retarder
source Polarizer emitting diode (SLED), with a center wavelength near 1310 nm
PM fiber
and a spectral width of about 40 nm (full width at half
45°-
Reflector maximum, FWHM). The SLED resides in a butterfly package
splice
Current with an integrated thermo-electric cooler (TEC). The TEC
Signal
conductor maintains a constant SLED chip temperature during operation,
Photo processor
diode here 25 °C. Of particular importance is the long-term stability
Sensing fiber of SLED center wavelength l, since the Faraday effect roughly
Df (~ current) coil scales in inverse proportion to l2 [24]. Hence, accuracy to
Fig. 1. Sensor configuration. ±0.2 % requires that the wavelength remains stable within
±1.3 nm, if one disregards other error sources. In practice, the
wavelength should remain in a significantly narrower band,
III. SENSOR COMPONENTS AND TEST PROCEDURES since wavelength related errors should consume only a fraction
of the total error budget. In addition, it is of importance that the
As part of our reliability program, we initially performed an
set chip temperature does not drift over time, as the wavelength
extensive Failure Mode and Effects Analysis (FMEA) [20] that
is rather sensitive to temperature (»0.6 nm/°C). Assuming a
included all essential sensor components. The FMEA helped to
stable drive current, important potential sources of wavelength
define the tests considered in the following. For reliability
shifts are ageing of the SLED chip and degradation of the
testing, the components were subjected to temperature cycles thermal contacts associated with the thermo-electric cooler and
(in some cases as many as 15000 cycles) as well as constant thermistor.

1558-1748 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2019.2944346, IEEE
Sensors Journal

> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 3

Typically, qualification tests by SLED manufacturers are

performed for chips on submount at constant elevated
temperature (e.g., 85 °C) and focus on the power stability versus
time, whereas little information on the wavelength evolution is
available. The same is true for the long-term reliability of the
chip temperature control.
Therefore, the accelerated life tests were conducted in order
to verify the SLED long-term performance, in particular the
wavelength stability, under field conditions; i.e., the SLED chip
is kept at nominally 25 °C, while the environmental
temperature can exhibit large variations. Note, that even if the
SLED chip is kept at a constant and moderate temperature, the
electric stress caused by the driving current potentially
contributes to ageing effects.

1) Thermal cycling
Fig. 2. SLED center wavelengths vs number of temperature cycles (between -
Four SLEDs in fiber-pigtailed butterfly packages were 25 °C and 65 °C). The wavelengths were measured at a drive current of 135 mA
and a TEC temperature of 25 °C. The symbols represent the center wavelengths
subjected to extended thermal cycling as specified above. Two at an ambient temperature of 25 °C. The vertical bands indicate the residual
of the SLEDs from Supplier 1 (SLEDs 1, 2) experienced about wavelength excursions during the temperature cycles. SLED 1 – 4: Supplier 1.
15,530 cycles; the other two SLEDs (SLEDs 3, 4) were added SLED 5, 6: Supplier 2.
later and saw a total of 12,200 and 10,900 cycles. The total test
durations were 1725, 1355 and 1211 days, respectively. In We should also mention that we observed some minor but
addition, two light sources from Supplier 2 (SLEDs 5, 6) were noticeable narrowing of the SLED spectra during the ageing.
tested for about 4650 cycles (515 days). The SLED chip The width reductions were between 0.5 nm and 1.8 nm during
temperature was set to a nominal value of 25 °C by means of the up to 15000 temperature cycles (for comparison, the FHHM
the TEC, which was kept constant during the entire test period. of the spectra was about 40 nm as mentioned above).
A driving current of 170 mA (about 70 % of the rated maximum Considering the entire test duration, the variation of the
current) was applied during all cycles. The SLED center SLED output power versus temperature (-25 °C to 65 °C) was
wavelengths and output powers were measured at periodic between 0.2 dB and 0.65 dB for the SLEDs from Supplier 1 and
intervals with a wavelength-calibrated optical spectrum between 0.1 dB and 0.2 dB for the SLEDs from Supplier 2. The
analyzer (estimated long-term accuracy of about ±150 pm) and numbers essentially reflected the varying insertion losses of the
a power meter, respectively. The recorded center wavelengths fiber pigtails with changing temperature. There were no
corresponded to the center of gravity of the spectra as significant changes of the temperature dependence over time.
determined by the OSA. The OSA resolution was set to 70 pm. However, the absolute output power at room temperature of the
Fig. 2 shows the evolution of the center wavelength versus the SLEDs of Supplier 1 dropped by roughly 1 dB in average. Such
number of temperature cycles [11]. The wavelengths were a drop, while not negligible, is still uncritical for FOCS
measured at a drive current of 135 mA, a typical value for operation, though. For comparison, under the more benign
sensor operation. This value is in a range in which the conditions of the field trial (Section V), the changes in absolute
wavelength changes only little with current (see Fig. 3) and is output power remained insignificant. Other than the SLEDs
deemed least susceptible to drift in case of SLED ageing and/or from Supplier 1, the SLEDs 5, 6 from Supplier 2 did not show
drive current variations. Note that we have deliberately chosen any significant drop in the output power during accelerated
a drive current above the normal operating current but below ageing. (Note however that the test period of those SLEDs was
the maximum rating during the ageing periods (170 mA as shorter, Fig. 2).
mentioned above) in order accelerate degradation, if any.
Overall, the center wavelengths of the four SLEDs did not 2) Damp heat
exhibit obvious long-term drift. The observed variations were
within ±190 pm to ±380 pm. Common trends in the traces if Two additional SLEDs (one from each supplier) were
plotted versus real time let us conclude that in part the variations subjected to damp heat over 7200 hours at 65 °C and 85 %
were associated with limitations of the wavelength relative humidity (Rh). Similar to the test conditions of the
measurement. In any case, the variations remained well below thermal cycling test, the temperature of the SLED chips was set
critical limits for a sensor with ±0.2 % accuracy. The vertical to 25 °C and the devices were operated at a constant drive
bands indicate the residual wavelength variation current of 170 mA [12]. The center wavelengths remained
between -25 °C and 65 °C as a result of residual temperature within ±200 pm of the pristine wavelengths without signs of
variations of the temperature-stabilized SLED chips. It should long-term drift. The inset of Fig. 3 shows the SLED center
be noted that the sensor calibration accounts for such variations. wavelength versus drive current (measured at room
No obvious degradation of the TEC temperature control was temperature, TEC at 25 °C) with the mentioned plateau for
observed. currents above about 120 mA. The three other data sets show
the wavelength shifts versus current measured after different
periods of damp heat treatment. In the flat range, the

1558-1748 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2019.2944346, IEEE
Sensors Journal

> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 4

wavelength changes remained small and insignificant but were The PER is mainly determined by the two fiber-to-waveguide
slightly larger in the slope at small drive currents. Again, no junctions. Typically, the PER of the modulator waveguide is
signs of obvious degradation were discovered. For both SLEDs, much higher and therefore not of a particular concern. A
the measured output power remained within ±5 % and did theoretical analysis of the optical circuit shows that only
exhibit signs of drift. polarization-crosstalk at the coil-side fiber pigtail affects the
sensor scale factor, whereas cross-talk at the source-side pigtail
3) SLED lifetime estimate at field conditions is largely uncritical [27], [28]. The reason is that the waves
resulting from cross-coupling at the source-side modulator end
By making use of the Coffin-Mason equation [25]-[26], we experience a phase modulation that is 180° out of phase with
can translate the thermal strains that the SLEDs accumulated respect to the modulation of the (none cross-coupled) primary
over the test period into periods of time in a field environment waves. This, in combination with opposite Faraday phase shift
over which the SLEDs would accumulate the same level of of the cross-coupled waves, nullifies the cross-coupling effect.
strain [11]. As an example, we assume an average daily In contrast, cross-coupling at the coil-side modulator end
temperature swing of 9.6 °C as observed in the field trial enhances the apparent magneto-optic phase shift Df according
described in Section V. Assuming a Coffin-Manson exponent to
of = 1, the three sets of temperature cycles mentioned above
(15530, 12200 and 10900 cycles, Supplier 1 SLEDs) then ∆ =Δ / cos(2 ) (1)
correspond to about 400, 314 and 280 years of field operation,
respectively. Note, that the assumption of q = 1 represents a Here, Df0 is the magneto-optic phase shift without cross-
conservative estimate which disregards the additional ageing coupling. In (1), we associate the cross-coupling with a
effect due to the larger temperature gradients (±1.5 °C/min) and misalignment between the principal axes of the modulator
temperature excursions in the accelerated ageing tests waveguide and the PM fiber by an angle a. The PER is then
compared to typical field conditions. Following the approach given as
described in [11], we estimate a SLED lifetime based on the
failure-free operation periods of the devices. For the mentioned PER(dB) = −10log (tan ) (2)
field conditions and with a Weibull probability function to
describe the reliability vs time, we obtain a B5%-lifetime (after In practice, fiber stress at the modulator’s V-groove fiber
which 5 % of the components are lost) of 70 years at a 95 % holders also contributes to the cross-coupling. It should be
confidence level, which translates into a mean-time-to-failure noted that only light waves which result from two-fold cross-
(MTTF) of 320 years or a survival probability of 99.5 % after coupling, i.e., both on the forward and return path, will affect
21 years of field operation. the signal. Light waves originating from a single cross-coupling
event accumulate group delays with respect to primary waves
that make them incoherent. For ±0.2 %-sensor accuracy, the
PER at the coil-side modulator pigtail should be at least above
30 dB, preferably above 32 dB, over the entire temperature
range of operation. By contrast, changes in the modulator’s
half-wave voltage, e.g., due to temperature changes or ageing,
are continuously calibrated off by the signal processing and thus
do not affect the scale factor of the sensor [19].
Modulators from several suppliers were analyzed. Both the
integral and spatially resolved cross-coupling of the modulators
were measured between -40 °C and 80 °C. For the spatially
resolved measurements, a home-built white light interferometer
was used. The PER of the initial off-the-shelf modulators that
had been developed for communication applications was often
not sufficient for FOCS applications. Typical integral
extinction ratios were only around 23-26 dB at room
Fig. 3. SLED in damp heat: The inset shows the pristine wavelength-versus- temperature and showed a significant temperature dependence.
current relationship. The three data sets show the wavelength shifts vs drive Design adaptations and improved manufacturing procedures
current measured after different periods of damp heat treatment. consistently resulted in a PER of the two waveguide junctions
above 35 dB over the entire the temperature range of operation
B. Phase modulator (-25 °C to 65 °C). The following long-term test data were
obtained from early generation modulators. Therefore, they do
not necessarily reflect the finally achieved extinction ratios,
The FOCS phase modulator (birefringence modulator) is an nevertheless, provide information on the reliability of the
integrated-optic lithium niobate device with PM fiber pigtails. modulators.
A critical parameter is the polarization extinction ratio (PER),
as a measure for the polarization cross-talk between the two
orthogonal polarization directions, and its variation with 1) Thermal cycling
temperature. Another parameter of interest is the insertion loss.

1558-1748 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2019.2944346, IEEE
Sensors Journal

> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 5

finding that clearly confirms the importance of such tests.

Four modulators from the same supplier were exposed to a Meanwhile, the supplier has remedied the problem.
total of up to 15,000 temperature cycles (over a period of about The modulators of the second supplier did not exhibit any
4.5 years) in a dry atmosphere and repeatedly characterized obvious degradation. The light transmission remained stable
during this time span. One of the modulators failed early in the within a few percent over up to 7900 hours in damp heat. The
test phase (loss of transmission) and was classified as an “infant transmission variation as a function of temperature (-40 °C to
mortality”-type failure. The remaining modulators survived the 80 °C) was even smaller. Fig. 5 shows the evolution of the
test without significant degradation. On the contrary, the PER extinction ratio vs temperature at a waveguide-to-fiber junction
tended to increase over time which we interpreted as stress after increasing periods of time in damp heat for a later
relaxation at the fiber pigtails. Note however, that also an generation modulator [12]. Here, the extinction ratios were
increasing PER will cause inacceptable error, if the initial PER between 37 and 45 dB. The corresponding sensor scale factor
is too low, i.e., a high PER of the pristine modulators remains variations were within about 0.01 % and thus well below critical
essential. limits.
Fig. 4 depicts the evolution of the spatially resolved PER at For the other three modulators the minimum and maximum
a fiber-to-waveguide junction of one of the modulators with room temperature extinction ratios and corresponding scale
increasing number of temperature cycles [13]. It is obvious that, factor variations during the ageing period were as follows:
in average, the PER increased by several dB over time. If we (i) 34.1 dB and 43.0 dB, 0.03 %; (ii) 37.8 dB and 44.1 dB,
assume that a corresponding sensor would have been calibrated 0.01 %; (iii) 36.0 dB and 50.0 dB, 0.02 %.
at room temperature with a pristine modulator, the scale factor
variations between -25 °C and 65 °C (temperature range of
operation) would be within +0.025 % and -0.065 % As noted,
improved design and manufacturing procedures result in higher
initial PER. For the other two surviving modulators the
minimum and maximum room temperature extinction ratios
and the corresponding scale factor variations during the ageing
period were as follows: (i) 31.7 dB and 37.8 dB, 0.05 %;
(ii) 31.0 dB and 39.3 dB, 0.07 %.
The insertion losses at room temperature varied over the test
period between 0.1 dB and 0.3 dB for the three surviving
modulators. Typical variations as a function of temperature
were within about 0.1 dB and well below critical limits.

Fig. 5. Polarization extinction ratio at a fiber-to-waveguide junction vs

modulator temperature at increasing periods of time in damp heat.

C. Polarization-maintaining fiber connectors

Polarization-maintaining fiber connects the FOCS fiber coil

with the opto-electronics unit (Fig. 1). Sensor installation in the
field requires fiber splicing, typically at the opto-electronics end
of the fiber cable. The procedure necessitates a PM fiber splicer
and accordingly trained personnel. PM fiber connectors are an
alternative to splicing and can significantly facilitate the
installation procedures. However, for sensor accuracy within
±0.2 %, the polarization-extinction ratio of the connectors
Fig. 4. Polarization extinction ratio at a fiber-to-waveguide junction vs should again exceed 30 – 32 dB at all temperatures and remain
modulator temperature at increasing number of temperature cycles.
at that level over the entire operation time. The connector PER
2) Damp heat is primarily determined by the accuracy of the relative angular
alignment of the principal axes of the two connected fiber
Eight modulators from two suppliers (four samples per sections and fiber stress in the connector ferrules. The latter
supplier) were exposed to damp heat (65 °C, 85 % relative may vary with temperature. Repeated connector open/close
humidity). All four modulators from the first supplier failed the operations can affect both the relative angular alignment and
test. Root cause analysis revealed that the fiber pigtail adhesion state of stress (as well as its temperature dependence). The PER
degraded due to moisture ingress and thus the optical of commercial PM connectors often does not exceed 25 dB and
transmission was lost over periods of a few 1000 hours - a is therefore inadequate for demanding FOCS applications.

1558-1748 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2019.2944346, IEEE
Sensors Journal

> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 6

Here, we present accelerated ageing tests on FC/PC PM PER was observed. However, the variation of the PER vs
connectors pairs that were modified for reduced angular temperature changed significantly during the ageing process.
tolerance. Additional alignment pins and notches reduced the The PER variations observable in Fig. 8 correspond to scale
alignment error from about ±3° of the unmodified pairs to ±1.8° factor variations of -0.03 % at low temperature and +0.12 % at
of the modified pairs. The corresponding PER are 25 dB and 30 high temperature. Damp heat (85 °C, 85 % Rh) tended to
dB, respectively (not taking cross-coupling by fiber stress into increase the PER over time which is indicative of stress
account). After initial characterization, the connector pairs were relaxation (Fig. 9). In all tests no particular changes in the
exposed to temperature cycling and damp heat tests. connector insertion loss were observed.
Fig. 6 depicts the PER of a modified connector pair as a In conclusion, the temperature cycling and damp heat tests
function of the connector temperature (black squares). The PER showed that the upgraded PM connectors meet the requirement
is well repeatable over several temperature cycles. The of substation protection (typically ±1 % accuracy) but would
variation is attributed to varying fiber stress in the ferrules. consume too much of the error budget in revenue metering with
The (red) triangles show the sensor signal at a constant ±0.2 % accuracy. Hence, for high-end applications fiber
applied current as a function of the connector temperature with splicing is recommended. Another alternative is a modified
the same connector now spliced into a sensor. The signal optical circuit that filters out cross-coupled light. In that case,
variation (within ±0.05 %) reflects reasonably well the variation connector PER well below 20 dB are still uncritical, as we have
in PER. (Note that a PER increase reduces the signal, see (1), shown in [28].
(2) [14]). Fig. 7 shows the variation in the sensor signal at 15
consecutive connector open/close operations. The signal
repeatability is within ±0.05 % and consistent with the reduced
angular tolerances of the connectors.

Fig. 8. PER of a PM connector pair vs temperature after exposure to an

increasing number of temperature cycles.
Fig. 6. PER (black squares, left axis) and normalized sensor signal (red
triangles, right axis) vs connector temperature [14].

Fig. 7. Normalized sensor signal vs connector mating cycles [14].

After their initial characterization, several connector pairs

were exposed to more than 13000 temperature cycles as
specified further above and repeatedly recharacterized during
the ageing period. The connections remained closed at all times.
Fig. 9. Polarization extinction ratio of a PM connector pair vs temperature
The total ageing period was about four years. The room- after exposure to damp heat over increasing periods of time.
temperature PER of the pristine connector pairs were between
28 dB and 34 dB; the insertion losses were less than 0.02 dB.
Fig. 8 shows the PER vs temperature of one of the better D. Other discrete components
connector pairs before the test and after 7570 and 13090 Passive fiber components subjected to accelerated ageing by
temperature cycles. In general, no particular degradation in the temperature cycling and damp heat included in-line fiber

1558-1748 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2019.2944346, IEEE
Sensors Journal

> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 7

polarizers, 1×3 fiber couplers (used in 3-phase sensors), and developed between 1600 h and 4600 h of ageing (Fig. 11). A
2×2 fiber couplers. The components were exposed to about spatially resolved measurement of the PER by means of white
16000 temperature cycles over roughly five years and 3168 light interferometry showed that the degradation was not
hours in damp heat at 85 °C, 8 5 % Rh. We tested three pristine distributed homogeneously along the fiber. Rather, increased
samples of each component in both types of tests. Two sets of polarization cross-coupling occurred at discrete points instead
two photodiodes each were subjected to 12300 temperature [15]. A visual inspection of the fiber showed a deformation of
cycles and 3168 h of damp heat. The components were the coating (deviation from circular cross-section) at those
characterized between -40 °C and 80 °C except the photodiodes points accompanied by a slight color change from clear to a
that were characterized only at room temperature. Here we yellow-brownish). With increasing ageing time both the
briefly summarize our observations: number of cross-coupling points and the coupling strength at
One out of the three 1×3 couplers failed after 7400 cycles. those points increased. At 105 °C, the same phenomena were
Obviously due to a fiber breakage inside the coupler sleeve, one observed, only their onset was delayed by a few thousand hours.
of the ports lost the optical transmission. All other fiber At 90 °C, only one of the three Type 1 samples exhibited PER
couplers survived both the temperature cycling and damp heat degradation at low temperatures, which started after 7500 h.
without significant degradation. The same was true for the fiber The other two samples remained stable until the end of the test
polarizers and photodiodes. at 20280 h. No PER degradation was observed for the Panda
fiber sample that was exposed to 105 °C.
Hence, the results indicate clear differences in the high
E. Fiber coatings
temperature behavior of the two e-core fiber types and again
Elevated temperatures can accelerate thermo-oxidation of the emphasize the importance of such investigations. Note
coating accompanied by loss of volatile by-products (coating however, that both fiber types should be suited for the vast
weight loss) and reduced mechanical strength of the fiber [29]. majority of applications, since extended periods of high
In addition, hardening of the fiber coating can lead to increased temperature operation are commonly rare.
fiber stress, especially at low temperatures, and thus to
polarization cross-coupling. Our studies focused on the
polarization maintaining properties of PM fiber samples
between -40 °C and 80 °C that were aged at elevated
temperatures [15]. The fiber retarder at the coil entrance and its
pigtail are made from e-core fiber (due to the relatively small
temperature dependence of the fiber birefringence) and are
exposed to the same environmental temperatures as the fiber
coil. A degradation of the e-core fiber performance by longtime
exposure of the coating to elevated temperature should be
avoided. Note that the sensing fiber itself is a low birefringence
fiber and resides without coating in a thin fused silica capillary
in order to exclude any coating- and packaging related stresses
[9]. Many dozens of fiber coils were subjected to temperature
cycling tests during the sensor development without failures.
Therefore, the coils were not included in the long-term
component ageing tests. Note however that their reliability was
explored in the field trial described in section V. Fig. 10. Polarization extinction ratio vs temperature after different periods of
Fiber samples from two suppliers (Type 1 and Type 2 with ageing at 115 °C (e-core fiber Type 2).
similar (phase) beat lengths of 5.3 mm and 6.7 mm,
respectively) were aged in an air atmosphere at 90 °C, 105 °C,
and 115 °C. The specified maximum operating temperature of
the fibers was 85 °C. Per fiber type and ageing condition we
prepared three samples; all with a length of 6 m, i.e., there were
18 samples in total. The total ageing periods lasted up to
20280 h at 90°C, 14940 h at 105 °C, and 13800 h at 115 °C. For
comparison, a single Panda PM fiber sample with a beat length
of 2.5 mm was exposed to 105 °C. The polarization extinction
ratio of fibers was periodically measured between -40 °C and
80 °C.
The PER of the Type 2 fiber samples did not show any
obvious degradation, not even at the highest ageing temperature
(115 °C, Fig. 10). Note that a few data sets at some intermediate
ageing times were omitted in order not to over-crowd the graph.
By contrast, ageing at 115 °C caused severe degradation of the
Type 1 samples. The degradation showed up as a significant
Fig. 11. Polarization extinction ratio vs temperature after different periods of
reduction of the PER below room temperature, which ageing at 115 °C (e-core fiber Type 1).

1558-1748 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2019.2944346, IEEE
Sensors Journal

> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 8

V. FIELD TRIAL conclusion, no particular degradation of light sources,

A three-phase sensor system was subjected to an extended modulators or any other component was encountered.
field trial to prove the long-term reliability under field Meanwhile the sensors have been in operation for nine years
conditions. The system consisted of two fiber coils per phase without any incidents or failures; systematic data recording and
for redundancy and two 3-phase opto-electronics units. The evaluation has been halted, though.
sensors were integrated in 420 kV double-chamber circuit
breakers (Fig. 12) [7], [8]. The fiber coil housings were
positioned at the high voltage end of the corresponding breaker
pole inside the breaker’s SF6 gas volume (indicated by arrows
in Fig. 12). Appropriate absorber materials inside gas volume
prevent potential corrosive SF6 decomposition by-products
such SF4 and HF to reach the sensor. Such by-products can be
generated by arcing in the breaker chambers during current
interruption. The coils for a given electric phase were placed in
a common housing. The current path was modified to run
through the aperture of the coil housings. The opto-electronics
units were placed into the breaker’s drive cubicle of the central
breaker at the bottom end of that pole. (Note: in our regular
commercial installations the opto-electronics units reside in a
separate cubicle). Fiber cables ran from the opto-electronics
through the hollow interior of the breaker poles to the coils. A
digital process bus (IEC61850-9-2LE standard) connected the
sensors with digital protection relays in the substation control
house and dedicated data acquisition equipment. The signals of
the optical sensors and corresponding conventional current
transformers were recorded every 30 min for a long-term
comparison. In addition, the ambient air temperature, the
temperature inside the cubicle and opto-electronics housings as
well as various operational parameters were recorded.
Fig. 13 shows the relative signal difference between the
optical sensors (of the first 3-phase system) and the Fig. 12. 420 kV double-chamber circuit breaker with integrated FOCS (top);
corresponding conventional CT together with the ambient 3-phase FOCS system (redundant, bottom).
temperature over a period of more than three years. The
deviations of the optical sensors from the conventional CTs 0.4%
Signal difference

Phase A Phase B Phase C

remained well within ±0.2 % for all six sensors (error limits of 0.2%
the IEC metering class 0.2), independent of the ambient
0.0%
temperature.
Further parameters that were monitored included the light -0.2%

source power (constant drive current), variation in the loop gain -0.4%
of the detection circuits (as a measure for changes in roundtrip 0 1 2 3
insertion loss of the optical circuits), and the p-voltage of the
Ambient temperature (°C)

modulators. 40
summer
30
The optical power of the SLED of the first opto-electronics 20
unit as recorded by the SLED’s internal monitoring diode 10
varied within 0.3 %, i.e., the variation was negligible. The 0
-10
SLED power of the second unit was slightly larger (0.8 %) but -20 winter
again far from critical limits. -30
0 1 2 3
Insertion loss variations were consistent with the temperature
Time (years)
dependence of the modulator insertion losses as determined
during the pre-characterization of the devices. No particular Fig. 13 Difference between optical and conventional current measurement
long-term drift was encountered. signals for the three phases vs time (top); ambient temperature vs. time
(bottom).
The modulator p-voltage changes (after correction for the p-
voltage temperature dependence) were insignificant (<±0.1 %)
for the first opto-electronics unit (modulators from Supplier 1)
VI. CONCLUSIONS
and small (<0.3 %) for the second unit (modulators from
Supplier 2). Note that p-voltage changes do not immediately We investigated the long-term reliability of a state-of-the-art
affect the sensor output due to the self-calibration of the closed interferometric fiber-optic current sensor on component- and
loop circuit as already mentioned in Section IV.B [19]. In system level. Crucial optical sensor components which
included the Peltier-cooled superluminescent light emitting

1558-1748 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2019.2944346, IEEE
Sensors Journal

> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 9

diode source in a butterfly package, the phase modulator, and [4] K. Bohnert, P. Gabus, J. Kostovic, and H. Brändle, “Optical fiber sensors
for the electric power industry,” Optics and Lasers in Engineering,
various passive components were subjected to accelerated
Special Issue “Optics in Switzerland”, Editor Erwin Hack, vol. 43, no. 3-
ageing by extended temperature cycling (up to 15000 cycles 5, pp. 511-526, 2005.
between -25 °C and 65 °C over several years), constant high [5] J. Wang, Z. Guo, G. Zhang, “Study on longterm stability of 110 kV
temperature at dry conditions (up to 115 °C) and damp heat power optical sensor solution - optical current transducer,” in Proc.
APPEEC 2010 (Asia-Pacific Power & Energy Engineering Conference
(85 % Rh at 65 °C and 85 °C). During the ageing periods, the
2010), Chengdu, China, 28-31 March 2010, pp. 1178-1179.
components were characterized repeatedly over the [6] M. Takahashi, K. Sasaki, Y. Hirata, T. Murao, H. Takeda, Y. Nakamura,
corresponding temperature range of operation. Furthermore, a T. Ohtsuka, T. Sakai, and N. Nosaka, “Field test of DC optical current
field trial of a complete three-phase sensor system integrated transformer for HVDC link,” in Technical Digest IEEE Power & Energy
Society General Meeting, 2010, pp. 1-6.
into 420 kV circuit breakers was carried out over several years.
[7] K. Bohnert, A. Frank, T. Roininen, B. Gülenaltin, P. Gabus, A. Vujanic,
Key results are summarized as follows: The long-term “Fiber-optic current and voltage sensors as modern alternatives to
variations in the center wavelength of the SLEDs from two conventional instrument transformers,” in Technical Digest, IEEMA
suppliers were within ±190 pm to ±380 pm. Changes in the TECH-IT 2014, Third International Conference on Instrument
Transformers, New Delhi, India, 2014, pp. I1 – I7.
output power remained uncritical for the FOCS application. The [8] K. Bohnert, A. Frank, G. M. Müller, L. Yang, M. Lenner, P. Gabus, X.
polarization extinction ratio (PER) of the modulators, again Gu, S. V. Marchese, “ Fiber optic current and voltage sensors for electric
from two suppliers, tended to improve during the ageing power transmission systems,“ in SPIE Commercial + Scientific Sensing
periods both in the temperature cycling and damp heat tests. All and Imaging, Fiber Optic Sensors and Applications XV, Proc. of SPIE,
vol. 10654, 1065402; doi: 10.1117/12.2303945, Orlando, FL, 15-19
modulators (four samples) of one supplier failed the damp heat
April, 2018.
tests, though, as a result of inadequate fiber pigtail adhesion. [9] K. Bohnert, P. Gabus, J. Nehring, H. Brändle, and M. G. Brunzel, “Fiber-
(The supplier subsequently remedied the problem). The optic current sensor for electrowinning of metals, “J. Lightw. Technol.,
observed variations in the center wavelengths and polarization vol. 25, no. 11, pp. 3602-3609, Nov. 2007.
[10] Standard of the International Electrotechnical Commission,
extinction ratios remained well within the allowed margins for
IEC60044-8, Instrument transformers – Part 8: Electronic current
sensor accuracy to within ±0.2 %. Likewise, no noteworthy transformers.
degradation of the fiber polarizers and fiber couplers was [11] M. Lenner, L. Yang, A. Frank, and K. Bohnert, “Long-term reliability of
encountered, with exception of one (out of three) 1x3 couplers semiconductor light sources for fiber-optic current sensors,” Topical
Meeting Advanced Photonics Congress, Optical Sensors (Sensors),
which had failed after 7400 temperature cycles. On the other
Optical Society of America, Washington, DC, Barcelona, Spain, 2014,
hand, the PER stability of FC/PC polarization-maintaining fiber paper SeW3C.6.
connectors was found to be insufficient for 0.2 % accuracy, [12] M. Lenner, L. Yang, A. Frank, and K. Bohnert, “Long-term reliability of
even though they were custom-adapted for high PER. Hence, the fiber-optic current sensors in damp heat,“ AMA Conference
Sensor+Test, Nuremberg, Germany, 2017, paper B3.2.
we propose fiber splicing or using an alternative optical circuit
[13] M. Lenner, L. Yang, A. Frank, and K. Bohnert, “Long-term reliability of
[28] instead of the PM connectors. fiber-optic sensor components in harsh industrial environments,” OSA
The long-term exposure to constant high temperature Advanced Photonics Congress, Optical Sensors (Optical Society of
addressed the potential degradation of the coating and thus the America), New Orleans, Louisiana, USA, 2017, paper SeW1E.3.
[14] M. Lenner, C.-P. Hsu, L. Yang, A. Frank, and K. Bohnert, “Performance
PER of elliptical-core fibers that are used for the sensor’s fiber
and long-term reliability of polarization-maintaining fiber connectors for
retarder. Several samples of a fiber from one supplier did not fiber-optic current sensors”, IEEE Sensors 2018, New Delhi. India, 2018,
show any PER degradation even after 13800 hours of ageing at paper ID 1091.
115 °C. By contrast, for fibers from a second supplier, the PER [15] M. Lenner, L. Yang, A. Frank, and K. Bohnert, “Influence of optical
fiber coatings on the long-term accuracy of interferometric fiber-optic
below room temperature strongly degraded after a few thousand
current sensors,” Eurosensors XXX, Budapest, Hungary, Procedia
hours of ageing at 115 °C. The degradation was much less Engineering, Vol. 168, 1735-1738, 2016.
severe when aged at 90 °C, though. [16] G. Frosio and R. Dändliker, “Reciprocal reflection interferometer for a
In the field trial the sensor accuracy was maintained within fiber-optic Faraday current sensor,” Appl. Opt., vol. 33, no. 25, pp. 6111-
2122, Sept. 1994.
±0.1 % over an observation period of more than three years.
[17] J. Blake, P. Tantaswadi, and R.T. de Carvalho, “In-line Sagnac
Meanwhile, the sensors have been in operation for nine years interferometer current sensor,” IEEE Trans. Power Delivery, vol. 11, no.
without incidents or failures. In conclusion, the investigations 1, 116-121, Jan. 1996.
serve well to enhance the confidence in optical current [18] K. Bohnert, P. Gabus, J. Nehring, and H. Brändle,”Temperature and
vibration insensitive fiber-optic current sensor,” J. Lightw. Technol., vol.
transformers even under the harsh field conditions in the
20, no. 2, pp 267-276, Feb. 2002.
electric power industry. [19] H. Lefevre, “The fiber-optic gyroscope,” Boston, London, Artech House,
1993.
REFERENCES [20] D. H. Stamatis, “Failure Mode and Effect Analysis: FMEA from
Theory to Execution,” Milwaukee, WI, USA, ASQ Quality Press,
[1] C. P. Yakymyshyn, M. A. Brubaker, P. M. Johnston, C. Reinbold, 2003.
“Manufacturing challenges of optical current and voltage sensors [21] A. Birolini, “Reliability Engineering,” Berlin, Heidelberg, Germany,
for utility applications,” in Proc. of SPIE 3201: Sensors and Springer Verlag, 2007, p 99.
Controls for Advanced Manufacturing, Pittsburgh, PA, USA, 1997 [22] J. Park, D. S. Shin, “Degradation of fiber optical communication devices
pp. 2-19. under damp-heat aging,” IEEE Phot. Technology Lett., vol. 15, no. 8, pp.
[2] M. Willsch and T. Bosselmann, „Optical current sensor application 1106-1108, 2003.
in the harsh environment of a 120 MVA power generator,“ in [23] H. Nagata, Y. Li, K. R. Voisine, W. R. Bosenberg, “Reliability of
Technical Digest 15 th Optical Fiber Sensors Conference, Portland, nonhermetic bias-free LiNbO3 modulators,” IEEE Phot. Technology
OR, USA, 2002, paper ThD2, pp. 407-409. Lett., vol. 16, no. 11, pp. 2457-2459, 2004
[3] F. Rahmatian, G. Polovick, B. Hughes, and V. Aresteanu, “Field [24] A. H. Rose, S. M. Etzel, and C. M. Wang, “Verdet constant dispersion
experinence with high-voltage combined optical current and voltage in annealed optical fiber current sensors,” J. Lightw. Technol., vol. 15,
transducers,” in Proc. Cigre (International Council on Large Electric no. 7, 803-807, 1997.
Systems) Session 2004, Paris, France, paper A3-111.

1558-1748 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JSEN.2019.2944346, IEEE
Sensors Journal

> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 10

[25] L. Coffin, Jr, “Low Cycle Fatigue”, ASM Metals Engineering

Quarterly, vol. 3, no. 4, 1963.
[26] S. Manson, “Thermal Stress and low-cycle fatigue”, New York,
NY, USA, McGraw-Hill, 1966.
[27] J. Yu, C. Zhang, C. Li, X. Wang, Y. Li, X. Feng, “Influence of
polarization-dependent cross-talk on scale factor in in-line Sagnac
inerfereometer current sensor,“ Opt. Eng. vol. 52, no. 11, pp.
117101-1 - 117101-10, Nov. 2013.
[28] K. Bohnert, C.-P. Hsu, L. Yang, A. Frank, G. M. Müller, and P.
Gabus, “Fiber-optic current sensor tolerant to imperfections of
polarization-maintaining fiber connectors,“ J. Lightw. Technol.,
vol. 36, no. 11, pp. 2161-2165, June 2018.
[29] A. A. Stolov, D. A. Simoff, and J. Li, “Thermal Stability of Specialty
Optical Fibers,” J. Lightw. Technol., vol. 26, no. 20, pp. 3443-3451, Oct.
2008.

1558-1748 (c) 2019 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.