Distributed Fiber Optic Sensing: Temperature Coefficient for Polyimide Coated Low Bend Loss Fiber, in the -40°C to 200°C Range

www.lunainc.com
1 Riverside Circle, Suite 400 | Roanoke, VA 24016 solutions@lunainc.com
1.540.769.8400 Technical Note EN-FY1403
Revision 1 July 23, 2014
Distributed Fiber Optic Sensing: Temperature Coefficient for
Polyimide Coated Low Bend Loss Fiber, in the -40°C to 200°C
Range
Contents
Introduction …………………………………………………………………………………………………………………………………………………………. 1
Theory …………………………………………………………………………………………………………………………………………………………………… 1
Test Setup …………………………………………………………………………………………………………………………………………………………….. 3
Aluminum Enclosure …………………………………………………………………………………………………………………………………….. 3
RTDs ………………………………………………………………………………………………………………………………………………………………….. 3
Fiber Layout ……………………………………………………………………………………………………………………………………………………. 3
Temperature Profile ………………………………………………………………………………………………………………………………………. 5
Results …………………………………………………………………………………………………………………………………………………………………… 5
RTD Calibration ………………………………………………………………………………………………………………………………………………. 5
Temperature Coefficient ………………………………………………………………………………………………………………………………. 6
Segmented Linear Fit …………………………………………………………………………………………………………………………………… 11
Exercise Cycle ……………………………………………………………………………………………………………………………………………….. 13
Coefficient Verification ……………………………………………………………………………………………………………………………….. 13
Summary …………………………………………………………………………………………………………………………………………………………….. 15
References ………………………………………………………………………………………………………………………………………………………….. 15
Appendix …………………………………………………………………………………………………………………………………………………………….. 16
Product Support Contact Information ………………………………………………………………………………………………………….. 17
Introduction
This Technical Note describes the methods employed to obtain a temperature coefficient for
polyimide coated low bend loss fiber, in the -40°C – 200°C range. Results indicate that a linear fit
with coefficient -6.38E-1 °C /GHz results in a maximum deviation of 10.63°C within this
temperature range, while a quartic fit with coefficients of -7.80E-1°C /GHz, -7.96E-4 °C /GHz2
,
-1.57E-6°C /GHz3
, and -1.57E-9°C /GHz4
, (calculated for a tare at 0°C) results in a much reduced
maximum deviation of 0.28°C within this temperature range. A piece-wise linear fit of the data also
results in reduced deviation. Humidity effects are a significant factor in temperature measurements
and should be controlled for best measurement accuracy.
Theory
Luna utilizes swept-wavelength interferometry to interrogate fiber optic sensors. Physical changes
in the sensors create a measurable change to the light that is scattered in the fiber (Rayleigh
scatter). By comparing locally-reflected spectra between two measurements of the same fiber optic
sensor, the local spectral shift may be deduced and calibrated to an external stimulus (e.g. strain,
temperature, etc.)
© 2014 Luna Innovations Incorporated. All rights reserved. Page 2 of 17
EN-FY1403 The physical length and index of refraction of the fiber are intrinsically sensitive to environmental
parameters: temperature and strain, and to a lesser extent, pressure, humidity (if the fiber coating
is hydroscopic), electromagnetic fields, etc. In most practical cases the effects of temperature and
strain will dominate the spectral response of the Rayleigh backscatter. Changes in the local period
of the Rayleigh scatter cause temporal and frequency shifts in the locally-reflected spectrum. These
shifts can be scaled to form a distributed sensor.
A change in temperature or strain from the baseline condition results in a shift in the spectrum of
light scattered in the fiber. The strain response arises due to both the physical elongation of the
sensor, and the change in fiber index due to photoelastic effects. The thermal response arises due to
the inherent thermal expansion of the fiber material and the temperature dependence of the
refractive index, n. The thermal response is dominated by the dn /dT effect, which accounts for
~95% of the observed shift. [1]
The shift in the spectrum of light scattered in the fiber in response to strain or temperature is
IK
IST and KB are the temperature
and strain calibration constants, respectively. Common values for most germanosilicate core fibers
are KT = 6.45 x 10-6 °C-1 and KB = 0.780. The values for KT and KB are somewhat dependent on the
dopant species and concentration in the core of the fiber, but also to a lesser extent on the
composition of the cladding and coating. Variations of 10% in KT and KB between standard telecom
fibers are common. [2,3]
In the absence of strain, the temperature change can be written as:
where is the center wavelength of the scan and c is the speed of light.
Assuming a scan center wavelength of 1550 nm, the constant KT can be substituted in to yield the
conversion factor:
such that:
In other words, the distributed temperature and strain curves are merely rescaled copies of the
frequency shift distribution. However, the linear approximation commonly made in the literature
does not fully account for the observed optical frequency response to temperature. In addition to
variation in the linear coefficient with core dopant species and concentration and fiber coating
material and thickness, higher order fitting terms may be needed to fully describe response,
especially over wide temperature ranges.
© 2014 Luna Innovations Incorporated. All rights reserved. Page 3 of 17
EN-FY1403 The temperature coefficient of a particular fiber type may be calibrated in a straightforward
manner by recording the frequency shift for a known applied temperature shift. For this Technical
Note, the temperature coefficient for polyimide coated low bend loss (LBL) fiber is calibrated, in the
-40°C – 200°C range.
Test Setup
Aluminum Enclosure
An enclosure was machined out of solid Aluminum (Figure 1). The mass of the enclosure ensured
that the temperature distribution within the enclosure cavity was uniform throughout the test.
Temperature uniformity within the cavity was verified by looking at the temperature uniformity
along all fiber sensor segments (Figure 5).
Figure 1: Aluminum enclosure
RTDs
Six calibrated Platinum RTDs (resistance temperature detectors) were used for these tests, as the
temperature measurement standard against which the fiber measurements were compared. The
RTDs were 4-wire, class 1 /3B, wire wound. The manufacturer specifies accuracy for these RTDs
from -100°C to 350°C. The RTD tolerances were measured using an in-house metrology well from
50°C to 200°C. An Agilent 34972A electrical readout system was used to log RTD measurements.
Fiber Layout
A single fiber sensor was strung in multiple passes within this enclosure (Figure 2). The sensor
consisted of stripped LBL spliced to polyimide coated LBL fiber. The six stripped LBL passes were
used for comparison with the following seven polyimide coated LBL fibers.
© 2014 Luna Innovations Incorporated. All rights reserved. Page 4 of 17
EN-FY1403
Figure 2: Fiber layout within Aluminum enclosure
Silicone pads were used to hold the fiber ingress and turnarounds. The fiber was instrumented
loosely drooping between the silicone pad strips, without touching the bottom of the cavity. This
droop was necessary to compensate for thermal expansion of the Al enclosure at maximum
temperature, and effectively isolated the fiber from strain.
An Aluminum lid was bolted on the enclosure before installation in a temperature chamber
(Tenney model TJR).
© 2014 Luna Innovations Incorporated. All rights reserved. Page 5 of 17
EN-FY1403
Temperature Profile
The temperature chamber was programmed to ramp over the temperature range of -40°C – 200°C
at a rate of 20°C per hour (Figure 3), with a 5 hour plateau at the limits of this range. Frequency
shift and temperature measurements were continuously recorded. This setup was put through 2
full cycles. Measurements on both the RTDs and the ODiSI B were taken at 1 minute intervals.
Figure 3: Temperature profile of the test
Results
RTD Calibration
The RTDs used in this test were measured against a metrology well (Fluke, model 9144) before the
0.1*(0.3+0.005*Temperature) °
without any further calibration (Figure: 4).
© 2014 Luna Innovations Incorporated. All rights reserved. Page 6 of 17
EN-FY1403
Figure: 4 RTD calibration results
Temperature Coefficient
A representative plot of frequency shift as a function of length along the sensor is shown in Figure
5. The frequency shift along the passes of each fiber type is seen to be uniform, confirming the
temperature uniformity within the enclosure.
Figure 5: Frequency shift along the sensor length at 200°C
The resulting temperature response curve for polyimide coated LBL is shown in Figure 6. The plots
on the top row are the measured response along with a linear fit (left) and a quartic fit (right). A
quartic fit was selected as an alternative to the linear fit as it resulted in much smaller deviations.
The plots on the middle and bottom row are the residuals of the linear fit (left) and quartic fit
(right) against frequency shift and time respectively. RTD Verification
© 2014 Luna Innovations Incorporated. All rights reserved. Page 7 of 17
EN-FY1403
Figure 6: Polyimide coated LBL temperature response. Dash lines are at 0°C and 100°C. Top: Temperature as a function of
frequency shift, with a linear fit (left) and quartic fit (right) applied. Middle: Temperature difference between measured and
fit results, as a function of frequency shift, with a linear fit (left) and quartic fit (right) applied. Bottom: Temperature
difference between measured and fit results, as a function of time, with a linear fit (left) and quartic fit (right) applied.
Up ramp
0 1 R0 1 R
0 0 R0 0 R
A
B
C D
E F
© 2014 Luna Innovations Incorporated. All rights reserved. Page 8 of 17
EN-FY1403 It is clear from Figure 6C that the sensor behavior is different on the up ramp (green) compared to
the down ramp (blue), with the down ramp showing a smoother and more repeatable responseIt is
postulated that this is due to humidity effects on the polyimide coating as it is known that polyimide
may absorb moisture from air and swell in response, causing a humidity-dependent strain on the
optical fiber. We believe that on the down ramp, moisture has been baked out of the environment
and therefore the polyimide coated LBL exhibits a repeatable behavior. The sensor is then held at –
40°C. The time spent at low temperatures then allows polyimide to start absorbing moisture,
causing it to exhibit drastic nonlinearities on the up ramp. At temperatures above 100°C, the
moisture is once again baked out of the polyimide. The effect of nonlinearities of the polyimide
coating can be seen when comparing results with measurements obtained from the stripped fiber
segments. For the stripped fiber, the temperature response is smooth and uniform throughout the
test (Figure 7).
Considering this, fits were carried out using only the measurements from the down ramps, while
deviations were plotted for the entirety of the test. These results are summarized in Table 1. For the
down ramp of temperature tests carried out in the -40°C – 200°C range with polyimide coated LBL
fiber, a linear fit results in a maximum deviation of 10.63°C at a test temperature of
-34.99°C, while a quartic fit results in a much reduced maximum deviation of 0.28°C at a test
temperature of -180.52°C. For the up ramp, a linear fit will result in a maximum deviation of
15.23°C at a test temperature of -40.40°C, while a quartic fit will result in a reduced maximum
deviation of 11.85°C at a test temperature of 9.05 °C.
© 2014 Luna Innovations Incorporated. All rights reserved. Page 9 of 17
EN-FY1403
Figure 7: Stripped LBL temperature response. Dash lines are at 0°C and 100°C. Top: Temperature as a function of frequency
shift, with a linear fit (left) and quartic fit (right) applied. Middle: Temperature difference between measured and fit results,
as a function of frequency shift, with a linear fit (left) and quartic fit (right) applied. Bottom: Temperature difference between
measured and fit results, as a function of time, with a linear fit (left) and quartic fit (right) applied.
Up ramp
0 1 R0 1 R
0 0 R0 0 R
A
B
C D
E F
© 2014 Luna Innovations Incorporated. All rights reserved. Page 10 of 17
EN-FY1403
Polynomial Fit Order Linear Quartic tared at 0°C Quartic tared at 25°C
Linear Coefficient (°C /GHz) -6.38E-1 A4 = -7.80E-1 B4 = -7.32E-1
2nd
order Coefficient (°C /GHz2
) 0 A3 = -7.96E-4 B3 = -6.50E-4
3rd
order Coefficient (°C /GHz3
) 0 A2 = -1.57E-6 B2 = -1.37E-6
4th
order Coefficient (°C /GHz4
) 0 A1 = -1.57E-9 B1 = -1.57E-9
Tare Temperature (°C) Any A5 = 0 B5 = 25
Largest Residual (°C) -10.63 (-15.23) -0.282 (-11.85) -0.282 (-11.85)
Table 1: Coefficients and residuals for polyimide coated LBL fiber. Largest residuals in parentheses are for the up ramp.
For comparison purposes, Table 2 summarizes fit results for stripped LBL fiber. Due to the uniform
behavior of stripped LBL, fits were carried out for the whole data set (up and down ramps
combined). A linear fit results in a maximum deviation of 6.73°C, while a quartic fit results in a
reduced maximum deviation of 0.78°C.
Polynomial Fit Order Linear Quartic tared at 0°C Quartic tared at 25°C
Linear Coefficient (°C /GHz) -7.24E-1 A4 = -8.63E-1 B4 = -8.08E-1
2nd
order Coefficient (°C /GHz2
) 0 A3 = -1.03E-3 B3 = -8.36E-4
3rd
order Coefficient (°C /GHz3
) 0 A2 = -2.34E-6 B2 = -2.02E-6
4th
order Coefficient (°C /GHz4
) 0 A1 = -2.67E-9 B1 = -2.67E-9
Tare Temperature (°C) Any A5 = 0 B5 = 25
Largest Residual (°C) 6.73 (-7.52) -0.779 (0.660) -0.779 (0.660)
Table 2: Coefficients and residuals for stripped LBL fiber. Largest residuals in parentheses are for the up ramp.
For measurements taken with a tare at a temperature other than 0°C, the following equation
describes the quartic form of the temperature response curve:
where:
when
and where can be found by solving the standard Quartic Equation:
This is expanded upon in the Appendix. Additionally, the Appendix contains example coefficients
calculated for a range of tare temperatures.
© 2014 Luna Innovations Incorporated. All rights reserved. Page 11 of 17
EN-FY1403 Piece-Wise Linear Fit
The measurement was also fit with a piece-wise linear fit covering three temperature ranges: -40°C
– 25°C, 25°C – 100°C, and 100°C – 200°C. This resulted in reduced deviation for each temperature
range (Figure 8). Fit coefficients and deviations are given in Table 3. Users carrying out tests within
a smaller range will be able to obtain results with smaller deviations using only a linear fit
coefficient.
Temperature Range (°C) -40 to 25 25 to 100 100 to 200
Linear Coefficient (°C /GHz) -7.91E-1 -6.73E-1 -5.83E-1
Largest Residual (°C) -0.868 (-12.38) -1.017 (-6.23) -1.375 (-1.75)
Table 3: Coefficients and residuals for polyimide coated LBL fiber. Largest residuals in parentheses are for the up ramp.
© 2014 Luna Innovations Incorporated. All rights reserved. Page 12 of 17
EN-FY1403
Figure 8: Polyimide coated LBL temperature response, with segmented linear fits applied. Dash lines are at 0°C and 100°C.
Top: Temperature as a function of frequency shift. Middle: Temperature difference between measured and fit results, as a
function of frequency shift. Bottom: Temperature difference between measured and fit results, as a function of time.
Up ramp
0 1 R0 1 R
0 0 R0 0 RA
B
C
© 2014 Luna Innovations Incorporated. All rights reserved. Page 13 of 17
EN-FY1403 Exercise Cycle
In a previous test carried out within the 10°C – 80°C range, it was observed that the polyimide
fiber (Figure 9) shows that the large difference at the start of the test reduces to 23% after 2 cycles
and 13% after 3 cycles. This exercise cycle is required to relax the polyimide coating on the LBL.
Customers carrying out temperature measurements are therefore advised to run their temperature
sensors through an exercise cycle prior to the actual test. The measurement can then be zeroed at a
known constant temperature along the fiber.
Figure 9: Temperature difference between RTD and fiber.
Coefficient Verification
In order to verify the accuracy of these coefficients, a slow ramp test was carried out with the same
test setup. In this test, the temperature chamber was ramped four times slower from -40°C – 200°C
and back down, at a rate of 5°C per hour. As expected, the quartic fit results in smaller residuals.
The reduction in ramp rate however, results in more obvious humidity effects even on the down
ramp, due to the increased length of time that the sensor spends at low temperatures (< 100°C).
© 2014 Luna Innovations Incorporated. All rights reserved. Page 14 of 17
EN-FY1403
Figure 10: Top: Polyimide coated LBL temperature response slow ramp. Dash lines are at 0°C and 100°C. Top: Temperature
as a function of frequency shift, with a linear fit (left) and quartic fit (right) applied. Middle: Temperature difference between
measured and fit results, as a function of frequency shift, with a linear fit (left) and quartic fit (right) applied. Bottom:
Temperature difference between measured and fit results, as a function of time, with a linear fit (left) and quartic fit (right)
applied.
Up ramp
0 1 R0 1 R
0 0 R0 0 R
A
B
C D
E F
© 2014 Luna Innovations Incorporated. All rights reserved. Page 15 of 17
EN-FY1403
Summary
These tests result in the accurate calculation of temperature coefficients for polyimide coated LBL
within the temperature range -40°C – 200°C. For a linear fit, a coefficient of -6.38E-1 °C /GHz is valid
at any tare temperature, and for a quartic fit tared at 0°C, the first through fourth order coefficients
are -7.80E-1°C /GHz, -7.96E-4 °C /GHz2
, -1.57E-6°C /GHz3
, and -1.57E-9°C /GHz4
respectively.
As was demonstrated, the presence of humidity has a substantial effect on the error from the fit.
Therefore, the user will obtain best accuracy if temperature testing is conducted in an environment
with very well-controlled humidity, or if the sensor is mounted in a hermetically sealed tube.
may vary from these observations
depending on the degree of humidity change.
Please contact Luna for further technical assistance related to the content discussed in this
Technical Note.
References
sJournal of Lightwave Technology, Vol. 15, No. 8. 1997
t-Optic Strain and Temperature
3 OBR 4600 User Guide
© 2014 Luna Innovations Incorporated. All rights reserved. Page 16 of 17
EN-FY1403 Appendix
can be found by solving the standard Quartic Equation:
as demonstrated below:
such that is real
such that is positive
such that is negative
Coefficients for Quartic Fits of Polyimide Coated LBL Fiber
Linear (°C /GHz) 2nd
order coeff.
(°C /GHz2
) 3rd
order coeff.
(°C /GHz3
) 4th
order coeff.
(°C /GHz4
) Tare Temp. (°C)
-5.50E-001 -2.45E-004 4.05E-007 -1.57E-009 200
-5.70E-001 -2.09E-004 1.25E-007 -1.57E-009 175
-5.88E-001 -2.11E-004 -1.47E-007 -1.57E-009 150
-6.07E-001 -2.46E-004 -4.11E-007 -1.57E-009 125
-6.29E-001 -3.11E-004 -6.65E-007 -1.57E-009 100
-6.57E-001 -4.03E-004 -9.10E-007 -1.57E-009 75
-6.91E-001 -5.17E-004 -1.14E-006 -1.57E-009 50
-7.32E-001 -6.50E-004 -1.37E-006 -1.57E-009 25
-7.80E-001 -7.96E-004 -1.57E-006 -1.57E-009 0
-8.34E-001 -9.51E-004 -1.77E-006 -1.57E-009 -25
© 2014 Luna Innovations Incorporated. All rights reserved. Page 17 of 17
EN-FY1403 Product Support Contact Information
Headquarters: 3157 State Street
Blacksburg, VA 24060
Main Phone: 1.540.961.5190
Toll-Free Support: 1.866.586.2682
Fax: 1.540.961.5191
Email: 0 1 R0 1 R
Website: 0 1 R
Specifications of products discussed in this document are subject to change
]Z}v}]X&}Zo}](]]}vU]]>v[]
0 1 R.
© 2014 Luna Innovations Incorporated. All rights reserved.
Technical Note EN-FY1403