Compensation of parasitic losses in an extrinsic fiber-optic temperature sensor based on intensity measurement

Sensors and Actuators A 173 (2012) 49–54

Contents lists available at SciVerse ScienceDirect

Sensors and Actuators A: Physical journal homepage: www.elsevier.com/locate/sna

Compensation of parasitic losses in an extrinsic fiber-optic temperature sensor based on intensity measurement Andreas Apelsmeier a,1 , Bernhard Schmauss b , Mikhail Shamonin a,∗ a b

Laboratory for Sensor Technology, University of Applied Sciences Regensburg, Seybothstr. 2, 93053 Regensburg, Germany Lehrstuhl für Hochfrequenztechnik, Universität Erlangen-Nürnberg, Cauerstraße 9, 91058 Erlangen, Germany

a r t i c l e

i n f o

Article history: Received 24 June 2011 Received in revised form 12 October 2011 Accepted 12 October 2011 Available online 19 October 2011 Keywords: Extrinsic optical fiber sensor Temperature measurement Intensity measurement Intensity referencing Indium phosphide Transient operation

a b s t r a c t A method of referencing in an extrinsic optical fiber sensor system utilizing temperature dependence of the absorption edge in a semiconductor crystal (semi-insulating iron-doped indium phosphide) is demonstrated. The intensity reference is provided by controlling the temperature of an LED source and transmission measurements with different emission spectra. A transient operation regime is introduced. The entire process is controlled by a microprocessor unit. The performance of the sensor system is investigated and it is shown that the connector losses may be compensated for. Contrary to the published works performed with GaAs crystals it was not observed that the absorption coefficient of the semiconductor follows the law for idealized direct-gap semiconductor but can be described by the so-called Urbach tail. Since the proposed sensor system comprises a single LED source, simple electronics and no optical fiber couplers it is promising for realization of low-cost fiber-optic temperature sensors, e.g. for power transformer monitoring or magnetic resonance imaging applications. © 2011 Elsevier B.V. All rights reserved.

1. Introduction In some specific applications, e.g. power transformer monitoring [1] or magnetic resonance imaging [2] fiber-optic sensors are used for temperature measurements since they do not interfere with close proximity electromagnetic fields. Conventional fiberoptic temperature sensors for these applications are based on one of three methods: fluorescence decay time, Fabry–Perot interferometry or the shift of the absorption edge in semiconductor crystals [3–5]. These systems are not low cost, since they comprise relatively complicated optical components or subsystems such as interferometer, couplers, spectrometer, filters, etc. (see [6] and references cited therein). The costs of fiber-optic Bragg-grating temperature sensors remain very high and they suffer from the cross sensitivity to other parameters [6]. In the present paper a low-cost fiber-optic temperature sensor system for the above mentioned applications based on intensity measurement is demonstrated. Like all optical fiber sensors relying upon intensity measurement it requires some form of intensity referencing to avoid errors arising from parasitic losses [7]. An idea for such a low-cost sensor system was patented almost 30 years ago [8] but to the best of our knowledge was neither

∗ Corresponding author. Tel.: +49 941 9431105; fax: +49 941 9431424. E-mail addresses: [email protected] (A. Apelsmeier), [email protected] (M. Shamonin). 1 Present address: Audi AG, 85045 Ingolstadt, Germany. Tel.: +49 841 575966. 0924-4247/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.sna.2011.10.015

experimentally verified nor developed further. The proposed sensor system comprises a single LED source, simple electronics and no optical fiber couplers. We concentrate on a method of referencing in an extrinsic optical fiber sensor system utilizing temperature dependence of absorption edge in a semiconductor crystal. To the best of our knowledge this issue has been addressed hitherto in a single publication [9] where a common-path reference method (requiring three fiber couplers, an optical isolator and an optical switch) combined with node-type error compensation technology has been proposed. In the context of power transformer monitoring “low cost” means that the market price of the sensor system should be below 1000 D per measurement channel. In the presented laboratory system the overall costs of components is about 100 D thus leaving enough free space to meet the price requirements for commercial applications. Known techniques of referencing such as balanced bridge, divided beam systems or two-wavelength referencing do not possess more simplicity than the proposed system since they use multiple LED sources, fiber couplers, filters, etc. [10]. 2. Working principle The concept of a sensor [11] is shown in Fig. 1. Light from an LED (peak wavelength P ≈ 950 nm at room temperature) is guided through the input optical fiber into the sensing head. There it is coupled into a semi-insulating iron-doped indium phosphide (InP) prism (material supplied by Wafer Technology Ltd., Milton

50

A. Apelsmeier et al. / Sensors and Actuators A 173 (2012) 49–54

Fig. 1. Working principle of sensor head.

Keynes, UK), deflected twice and coupled into the output fiber. The transmitted light power is reduced with the growing temperature of the InP-prism Ta . Due to the tiny prism dimensions a novel manufacturing method bridging fine mechanics and microsystems technology is required [12]. In the present paper the sensing head described in [12] is utilized. The block diagram of the sensor system is shown in Fig. 2. The temperature of the diode function block is maintained at constant level TB above the surrounding temperature. This is achieved by heating the diode function block with a power transistor. The monitoring photodiode controls the intensity emitted by the LED. In the envisaged applications usually the temperature range −50 ◦ C ≤ Ta ≤ 150 ◦ C is stipulated. In a practical implementation the sensitive element is embedded into the technical equipment (e.g. into windings of a power transformer) and connected with the diode function block by optical fibers. The diode function block is outside of the technical equipment. Fig. 3 shows the explosive view drawing of the diode function block.

Fig. 2. Block diagram of sensor system.

The photocurrent Iph is transformed into the output voltage Uph using a transimpedance amplifier. In the stationary regime the LED temperature TLED (strictly speaking the LED junction temperature) depends on TB and I0 . When calculating the sensor current Iph the following striking features of the sensor system must be taken into account: (i) the emission spectrum of the employed LED (GaAs infrared emitter LD242-2, Osram Optosemiconductors, Regensburg, Germany) S(,TLED ,I0 ) is not a simple symmetric Gaussian function of  (cf. the data sheet of this LED). (ii) With rising temperature TLED the total emitted optical power decreases and the spectral maximum shifts to larger wavelengths [13–17]. (iii) Contrary to the previous works performed on the GaAs crystals (see,

3. Physical modeling of the sensor system Correct modeling of the sensor system is a prerequisite for its effective design. The current Iph received by the photodiode 1 (receiver) can be mathematically expressed as



Iph (Ta ) = d

∞

R(, Ta )tInP (, Ta )S(, TLED , I0 ) d,

(1)

0

where  is the wavelength, R(,TB ) is the responsivity of the photodetector, S(␭,TLED ,I0 ) is the spectral power distribution of the LED and I0 is the forward drive current of the LED. It is assumed that the losses in the fiber-optic system (coupling loss, bending loss, fiber attenuation, etc.) are not significantly wavelength-dependent and may be accounted for by an attenuation coefficient 0 < d < 1. The transmission factor tInp of the sensing element is given by tInp = exp(−˛l), where ˛l is the product of the optical absorption coefficient ˛ and the effective path length of light l in the semiconductor material. The derivation of Eq. (1) is given in Appendix A.

Fig. 3. Exploded view drawing of diode function block.

A. Apelsmeier et al. / Sensors and Actuators A 173 (2012) 49–54 Table 1

1.5 E 320 K

V 230 K

S0 0.23

SV 2.6

X 1.2

˛g l 1200

e.g. [6,9,18]), we did not observe that the absorption coefficient ˛ versus photon energy hfollows the law for an idealized directhv − EG (Ta ) but it can be described by gap semiconductor ˛ ∝ the so-called Urbach tail [19,20] ˛(hv) = ˛g exp

 (hv − E )  G

E0

,

(2)

where E0 is the characteristic energy of the Urbach edge, EG is the extrapolated optical band-gap energy, and ˛g is the optical absorption coefficient at the band-gap energy. In the Einstein model, the width of the Urbach edge is E0 (Ta ) = S0 kE

1 + X 2

+



1 , exp(E /Ta ) − 1

(3)

where the dimensionless parameter X is a measure of the structural disorder,  E is the Einstein temperature, S0 is a dimensionless constant related to the electron–phonon coupling, and k is the Boltzmann constant. The temperature dependence of the band gap is given by



EG (Ta ) = EG (0) − SV kV



Ta2 (V2 + V Ta )

,

(4)

where SV and  V are fitting parameters. The parameters are summarized in Table 1. ˛g l is the product of the the optical absorption coefficient at the band-gap energy and the effective path of light in the semiconductor material. In general the parameters agree well with those reported in [20]. EG (0) is slightly lower than in the cited reference what can be explained by the band-gap narrowing in our sample due to Fe doping [21]. Parameter SV is about 50% of the reported value [20], what from our point of view has no further physical significance. This set of parameters was used in all simulations through the paper. Fig. 4 compares the results of theoretical modeling with the stationary measurements of the output voltage Uph versus ambient temperature Ta at two different temperatures TB . It is seen that a semiquantative agreement between the experimental results and the modeling is achieved. In these and the following measurements

5

Measurement TB=30 °C Measurement TB=60 °C Simulation TB=30 °C Simulation TB=60 °C

3

U

ph

[V]

4

1.3 1.2 1.1 1.0 -60

-40 -20

0

20

40 60 T [°C]

80

100 120 140 160

a

Fig. 5. Comparison between experiment (crosses) and simulation (continuous line) in the stationary regime, I0 = 80 mA = const, TB1 =60 ◦ C, TB2 = 30 ◦ C.

the sensing head has been placed into a conditioning cabinet where the ambient temperature Ta is set. The temperature Ta has been measured by a conventional platinum-based resistance sensor. 4. Compensation of parasitic losses The emission spectrum of an LED depends on its temperature TLED . From the literature it is known that the peak wavelength P increases with the growing temperature TLED [13–17]. The LED used in this work has a measured value of P /TB = 0.22 nm/K. As mentioned above, in stationary regime the temperature of a diode block TB determines the initial LED temperature. The ratio R1/2 (Ta ) = Uph (Ta ,TB1 )/Uph (Ta ,TB2 ) of two output voltages taken at two different temperatures TB1 , TB2 is independent of transmission losses d:

∞ R(, TB1 )tInP (, Ta )S(, TLED1 , I0 ) d . R1/2 (Ta ) = 0∞ R(, TB2 )tInP (, Ta )S(, TLED2 , I0 ) d

0

(5)

Fig. 5 compares the results of measurements of R1/2 with modeling. Notice that the ratio R1/2 for two selected values of TB1 , TB2 is not a monotonic function of Ta in the required range of temperatures. Note that a simplified model with symmetric spectral distribution of S(,TLED ,I0 ) (e.g. Gaussian shape with nominal spectral bandwidth) would overoptimistically predict monotonic behavior of R1/2 on Ta . In this simulation we used experimentally determined dependences of LED spectra on TB at constant driving current I0 . The measurement of R1/2 eliminates only the transmission losses. To compensate for the source fluctuations the monitoring diode is introduced. The ratio RN1/2 of two normalized photocurrents F = Uph (Ta ,TB )/Umon (TB ), where Umon is the output voltage of a monitoring diode, taken at two different temperatures TB1 , TB2 is independent of transmission losses d and the source fluctuations:

∞ R(, TB1 )tInP (, Ta )S(, TLED1 , I0 ) d RN1/2 (Ta ) = 0∞

2

0

R(, TB2 )tInP (, Ta )S(, TLED2 , I0 ) d

∞ Rmon (, TB2 )S(, TLED2 , I0 ) d × 0∞ ,

1 0 -60 -40 -20

1.4

1/2

EG (0) 1.402 eV

Measurement Simulation

R

Parameter Value

51

0

20

40 60 T [°C]

80

100 120 140 160

a

Fig. 4. Dependence of output voltage Uph on ambient temperature Ta in the stationary regime (I0 = 80 mA = const.) for two different values of the diode block temperature TB = 30 ◦ C and 60 ◦ C. Crosses represent experimental points, continuous and dashed lines are the simulated dependences.

0

Rmon (, TB1 )S(, TLED1 , I0 ) d

(6)

where Rmon (,TB ) is the responsivity of the monitoring photodiode. In our experiments measurements of R1/2 or RN1/2 gave the results of the same quality leading to the conclusion that the source fluctuations were not significant in this case. However, we believe that for the long-term operation (e.g. stipulated 30 years in power transformer monitoring) the measurement of RN1/2 (see Eq. (6) or

52

A. Apelsmeier et al. / Sensors and Actuators A 173 (2012) 49–54

2.0

two different LED temperatures (see detailed explanation below). In the envisaged applications long-term aging of optical components and the corresponding slow drift of optical parameters are the challenge for optical intensity-based temperature measurement. In these applications the sensor system should be mainly operated in the stationary intensity-measuring regime at lower temperatures. These measurements can be validated from time to time (e.g. once a day or on demand) by self-recalibration in the transient regime.

R

N1/2

1.8 1

1.6

2 3

1.4

4

6. Measurements

1.2 1.0 -60 -40 -20

0

20

40 60 T [°C]

80

100 120 140 160

a

Fig. 6. Simulated dependence of RN1/2 on ambient temperature Ta in the stationary regime (I0 = 80 mA = const.) for different combinations of the diode block temperature TB1 , TB2 : 60 ◦ C/30 ◦ C (1), 90 ◦ C/60 ◦ C (2), 120 ◦ C/90 ◦ C (3) and 150 ◦ C/120 ◦ C (4).

The measurements in the transient regime have been performed as follows. The initial LED temperature was set to TB = 70 ◦ C. Then the LED was self-heated by sending a current pulse of I0 = 500 mA during 2 s. The amplified output voltages of a photodiode 1 Uph and a monitoring diode Umon have been determined at definite time points i at the beginning (t1 ) and the end (t2 ) of the current pulse. The output rate fout was 100 Hz. In order to compensate for transient oscillation effects averaged values as given below were calculated:

 1 Uph,i ; × 25 30

Non-monotonic behavior of R1/2 with the temperature Ta is definitely a disadvantage (RN1/2 behaves similarly as well), since it limits the measured temperature range where the temperature may be determined unambiguously. How to extend the temperature range where RN1/2 is a monotonic function of Ta ? The answer is to play with temperatures TLED1 and TLED2 . The performance of the sensor system has been investigated by calculating RN1/2 at constant current I0 from our verified model for various values of TB in the temperature range −50 ◦ C ≤ Ta ≤ 150 ◦ C. In these calculations the LED spectra for TB > 70 ◦ C were extrapolated from the measured spectra. For the sake of brevity the conclusions are summarized here. Let TB1 = TB2 + T. For constant TB2 and increasing T, the shape of RN1/2 (Ta ) remains essentially the same, the range of variation of RN1/2 is increasing and the maximum of RN1/2 shifts slightly to the larger temperatures. For constant T and increasing TB2 the curve RN1/2 shifts strongly to the larger temperatures (see Fig. 6). The full-scale range of RN1/2 (Ta ) remains practically unchanged. To get monotonic behavior of RN1/2 in the required temperature range one has to adapt the temperature of the LED. Continuous operation of LED at elevated temperatures decreases its expected time of operation. To overcome this limitation we introduced a transient operation regime. The temperature of the LED can be increased by increasing the surrounding temperature (TB ) or by self-heating [17]. The LED is loaded by higher current pulse and heated itself due to power dissipation. The temperature of the LED varies during the pulse. By measuring the output signal of a photodiode 1 at two different time points, it is possible to make measurements at

i=6

194 

Uph,i ;

(7)

i=170

 1 Umon,i ; × 25 30

U¯ mon2 =

U¯ mon1 =

1 × 25

i=6

194 

Umon,i ;

(8)

i=170

−1 −1 where Uph,i = Uph (i × fout ), Umon,i = Umon (i × fout ). The measured ratio RN1/2,meas is given by −1 ¯ −1 Umon1 . RN1/2,meas = U¯ out1 U¯ mon2 U¯ out2

(9)

At given time point 1 ≤ i ≤ 200, Uph,i and Umon,i have been calculated as values averaged over 5 measurements performed with the sampling rate fs . This intermediate averaging step is done in order to eliminate the digital noise. The actual sampling rate fs = 5 kHz resulted from the clock frequency of the external quartz oscillator, the prescaler and the number of clock cycles required for the analog-digital conversion. Since fs  5 fout the averaged values Uph,i and Umon,i can be attributed to the particular time point i. The entire process is controlled by a micro-computer unit (ATmega1280), where the output ratio RN1/2 is calculated. Theoretically the measurement of LED temperature at two different time points is sufficient for the compensation method, thus the increase in the output rate did not immediately result in improving the accuracy of measurements. 1.5

Measurement Simulation

1.4

N1/2

5. Transient operation regime

1 × 25

U¯ out1 =

1.3

R

(9)) should be preferred. The theoretical dependence of both ratios on the ambient temperature Ta is the same; they differ only by a factor due to the different LED spectra at TB1 and TB2 . For example, RN1/2 (Ta )/R1/2 (Ta ) ≈ 1.27 with TB1 = 60 ◦ C and TB2 = 30 ◦ C (compare the theoretical curve in Fig. 5 with the curve 1 in Fig. 6). For calculating the ratio RN1/2 , two measured values of normalized photocurrent are required. As was mentioned above, these values must be taken at two different temperatures TB1 and TB2 . For this purpose the temperature of an LED must be set to both values and the corresponding values of F must be calculated. During the measurement of these two values the temperature Ta must not change significantly. Since our temperature sensor is conceived for the power transformer monitoring, where the thermal time constant is in the range of at least several minutes, this is definitely the case.

U¯ out2 =

1.1 1.1 1.0 -60 -40 -20

0

20

40 60 T [°C]

80

100 120 140 160

a

Fig. 7. Comparison between experiment (crosses) in the transient regime and simulation (continuous line). Parameters of measurement and simulation are described in the text.

A. Apelsmeier et al. / Sensors and Actuators A 173 (2012) 49–54

53

Fig. 8. Temperature measurement on the surface of the working 20 kVA transformer (left). The right picture shows the output of the IR camera. The indicated value of 5793 ◦ C and the shadow of the optical fiber are seen.

Note that according to the data sheet this particular LED cannot be used with I0 = 500 mA in the stationary regime (maximum rating for the stationary regime I0 = 375 mA) leading to the necessity of the utilizing the transient regime. The transient LED temperatures TLED1 (t1 ) and TLED1 (t2 ) cannot be measured directly. Fig. 7 compares the measurement results with the simulation based on Eq. (6) with constant current I0 = 80 mA. A semiquantative agreement between the experimental and simulated results is observed. The best fit is achieved for TB1 = 120 ◦ C and TB2 = 100 ◦ C. The obtained dependence R1/2 (Ta ) is monotonic in the entire working range of temperatures. The error bars in Figs. 5 and 7 refer to the estimated maximum deviations of the measured values. The horizontal error bars cannot be distinguished on the scale of figure. Let us now estimate the junction temperatures TLED1 and TLED2 . We can relate our transient measurements to the stationary measurements with I0 = 80 mA. At this current the voltage drop U across the LED is U ≈ 1.275 V giving the total dissipated power Pel ≈ 0.102 W. Taking from the data sheet the nominal value RthJC = 160 K/W and neglecting the total radiant flux in comparison with Pel one can estimate from TLED ≈ TB + RthJC × Pel that TLED1 ≈ 136 ◦ C and TLED2 ≈ 116 ◦ C. The relatively high operation temperatures of LED estimated in this particular realization of sensor system are a disadvantage. They are required in order to achieve monotonic dependence of the measured value on the ambient temperature in a broad temperature range. We expect that this disadvantage can be overcome either by using LEDs with larger temperature dependence of the band gap energy or by using the semiconductor crystals where the dependence of the absorption coefficient on the energy is closer  to the idealized behavior in hv − EG (Ta ) (cf. [6,18]). direct-gap semiconductor ˛ ∝ Performance of the sensor system has been investigated under realistic conditions. The losses into fiber connectors have been mimicked by pulling the connecting ferrule together with the input optical fiber (the fiber is fixed in the ferrule with an adhesive) out of the diode block and fixing it at definite positions with the side screw. In such a way an additional loss of up to 4 dB was introduced. In the entire temperature range −50 ◦ C ≤ Ta ≤ 150 ◦ C and in the entire range of introduced connector losses the maximum measurement error was ±2.1 ◦ C. At room temperature this uncertainty was even lower. For example, at fixed temperature Ta = 22.4 ◦ C the

maximum measurement error in the entire range of introduced connector losses (ten measurements at each particular connector loss, total of 60 measurements) was ±0.9 ◦ C. Available resolution of the ambient temperature of 0.2 ◦ C is determined by the analogdigital converter. Long-term measurements (over 2 weeks) in the transient regime at room temperature have been carried out within the above mentioned larger uncertainty. To demonstrate an application of the sensor system under more realistic conditions the temperature measurement on a surface of a standard working 20 kVA transformer available to us has been verified with an infrared (IR) camera (VARIOSCAN compact). Very good agreement has been noticed and no interference with electromagnetic fields has been observed. For example, under short-circuit conditions the fiber-optic sensor has given a value of 56.8 ◦ C while the IR camera has shown the value of 57.9 ◦ C (Fig. 8). However it should be noted that this was not a power transformer and the measured magnetic field density of 11 mT on the surface was quite low. Corrections to the measured temperature caused by the magnetic field H which are comparable to the measurement uncertainty can be expected in fields o H ∼ 3 T [2,22]. The measured accuracy of our sensor system is acceptable for the demonstrator of a low-cost measurement system for envisaged applications.

7. Conclusions and outlook A method of intensity referencing in an extrinsic optical fiber sensor system utilizing temperature dependence of absorption edge in a semiconductor crystal is described. The reference is provided by heating the emitting LED and thus shifting the emission spectrum towards larger wavelengths. The method is experimentally verified. A transient operation regime is introduced. Since the sensor system contains only one LED, simple electronics, no fiberoptic couplers or spectrometers it may be suitable as a low-cost system for temperature monitoring of power transformers, switching cabinets or MRI devices. This method is suitable for referencing in other optical sensor systems. The test of the system on real power transformer would require a manufacturer to embed the sensor into the coil winding during the production process. The approach can be extended to other semiconductor crystals (GaAs or Si) where it might be possible to achieve monotonic dependence of output signal ratio on Ta at moderate LED temperatures.

54

A. Apelsmeier et al. / Sensors and Actuators A 173 (2012) 49–54

Acknowledgement Financial support by the Bavarian State Ministry of Sciences, Research and the Arts within the priority research program “Miniaturized Sensor Systems with Emphasize on Applications in Medical Engineering, Biotechnology, Automotive and Automation Engineering” is gratefully acknowledged. We thank Ramona Gleixner for making some secondary measurements, Igor Stanke for expert advice and Professor Gareth Monkman for critical reading of the manuscript. We are grateful to the anonymous reviewers for their valuable comments. Appendix A. It is assumed that the losses in the fiber-optic system are not significantly wavelength-dependent around P and may be taken into account by an attenuation coefficient 0 < d < 1. The part of the optical spectrum of the LED in the elementary wavelength interval d around the wavelength  carries the optical power dP = S() d. The part of this power arriving at the detector is d × tInP () × dP. The elementary contribution dIph of the photocurrent Iph arising from the optical power dP in the elementary wavelength interval d around the wavelength  is given by dIph = R() × d × tInp () × dP. Integration of the latter expression over the entire emission spectrum of the LED results in Eq. (1). References [1] G. Betta, A. Pietrosanto, A. Scaglione, An enhanced fiber-optic temperature sensor system for power transformer monitoring, IEEE Trans. Instrum. Meas. 50 (2001) 1138–1144. [2] R.W. Martin, K.W. Zilm, Variable temperature system using vortex tube cooling and fiber-optic temperature measurement for low temperature magic angle spinning NMR, J. Magn. Reson. 168 (2004) 202–209. [3] J. Tang, Fiber-optic measurement systems: microwave and radio frequency heating applications, in: D.R. Heldman (Ed.), Encyclopedia of Agricultural, Food, and Biological Engineering, Marcel Dekker, 2006. [4] E. Pinet, S. Ellyson, F. Borne, Temperature fiber-optic point sensors: commercial technologies and industrial applications, Informacije MIDEM, J. Microelectron. Electron. Comp. Mater. 40 (2010) 273–284. [5] L. Rosso, V.C. Fernicola, Time- and frequency-domain analyses of fluorescence lifetime for temperature sensing, Rev. Sci. Instrum. 77 (2006) 034901. [6] Y. Ding, X. Dai, T. Zhang, Low-cost fiber-optic temperature measurement system for high-voltage electrical power equipment, IEEE Trans. Instrum. Meas. 59 (2010) 923–933. [7] B. Culshaw, J. Dakin (Eds.), Optical Fiber Sensors: Systems and Applications, vol. 2, Artech House, Boston, 1989. [8] M. Adolfsson, T. Brogårdh, S. Göransson, C. Ovrén, Optical measurement system for spectral analysis, United States Patent No. US 4,433,238, filed Oct. 19, 1981 (1984).

[9] Y. Zhao, M. Rong, Y. Liao, Fiber-optic temperature sensor used for oil well based on semiconductor optical absorption, IEEE Sens. J. 3 (2003) 400–403. [10] G. Murtaza, J.M. Senior, Referencing strategies for intensity modulated optical fibre sensors: a review, Opt. Laser Technol. 25 (1993) 235–245. [11] B. Schmauß, M. März, J. Ernst, A fiber-optic sensor for microwave field measurements, Rev. Sci. Instrum. 66 (1995) 4031–4033. [12] R. Trautner, B. Schmauss, M. Shamonin, Manufacture and characterization of an extrinsic elementary fiber-optical sensor for temperature measurement, tm – Technisches Messen 75 (2008) 565–570. [13] J. Plamper, J. Engl, Process for stabilizing the optical output power of lightemitting diodes and laser diodes, United States Patent No. US 6,807,202, B1, filed Mar. 17, 2000 (2004). [14] J.M. Senior, G. Murtaza, A.I. Stirling, G.H. Wainwright, Single LED based dual wavelength referenced optical fibre sensor system using intensity modulation, Opt. Laser Technol. 24 (1992) 187–192. [15] H.-Y. Chou, T.-H. Yang, Dependence of emission spectra of LEDs upon junction temperature and driving current, J. Light Vis. Environ. 32 (2008) 183–186. [16] F. Reifegerste, J. Lienig, Modelling of the temperature and current dependence of LED spectra, J. Light Vis. Environ. 32 (2008) 288–294. [17] Y. Atsuoka, M. Konishi, K. Takahashi, Temperature compensation for a semiconductor light source used for exposure of light sensitive material, US Patent No. 4,170,631, filed August 27, 1985 (1987). [18] T. Wang, C. Luo, Fiber-optic semiconductor absorption temperature sensor for electrical power system applications, in: Proceedings of the 21st IEEE Instrumentation and Measurement Technology Conference, 2004, pp. 2375–2379. [19] E.F. Schubert, Light-Emitting Diodes, Cambridge University Press, Cambridge, 2005. [20] M. Beaudoin, A.J.G. DeVries, S.R. Johnson, H. Laman, T. Tiedje, Optical absorption edge of semi-insulating GaAs and InP at high temperatures, Appl. Phys. Lett. 70 (1997) 3540–3542. [21] S. Do˘gan, S. Tüzemen, B. Gürbulak, A. Ates¸, M. Yildirim, Optical absorption and photoluminescence measurements in InP and InP:Fe bulk crystals and inspection of the relations between these measurements, Turk. J. Phys. 25 (2001) 551–556. [22] L.P. Zverev, R.V. Emlin, O.E. Rut, Investigations of the absorption edge of InP in a strong magnetic field, Phys. Status Solidi (b) 68 (1975) 93–97.

Biographies Andreas Apelsmeier received the M.Eng. degree in Electrical and Microsystems Engineering from the University of Applied Sciences Regensburg in 2009. He is currently with the Audi AG in Ingolstadt, Germany. Bernhard Schmauss received the Dipl. Ing. and Dr. Ing. degrees in Electrical Engineering from the University Erlangen-Nuremberg in 1989 and 1995, respectively. In 1995, he joined Lucent Technologies, in Nuremberg, Germany. From 2003 to 2005 he was professor at the University of Applied Sciences in Regensburg, Germany. Since October 2005 he is professor for Optical High Frequency Technology and Photonics at the University Erlangen-Nuremberg. His research interests are fiber lasers, medical application of photonics, various aspects of optical transmission systems and optical sensors. He is principal investigator of the Erlangen Graduate School in Advanced Optical Technologies. Mikhail Shamonin received the Diplom degree in physics from the Moscow M.V. Lomonosov State University, Russia in 1989 and the Dr.rer.nat degree in Physics from the University of Osnabrück, Germany in 1995. Since 2002 he is professor for Sensor Technology at the University of Applied Sciences in Regensburg, Germany. His research interests are sensor technology and smart materials.