You are on page 1of 7


7, JULY 1995


Historical Review of Microbend Fiber-optic Sensors

John W. Berthold I11
Abstract-This paper traces microbend sensors from the early investigative work into initial applications to measure many different parameters, through advanced prototype development and commercial hardware. Advantages and disadvantages of microbend sensors are discussed.



I. INTRODUCTION HE MICROBEND sensor is one of the earliest of sensors. Microbend losses have always been a curse to the cable designer, but it is this very same microbend loss effect in optical fibers which is exploited by the microbend sensor designer who has adapted the microbend effect to the measurement of many physical parameters and physical variables such as temperature and pressure. Microbend sensors are very interesting sensors with some outstanding performance characteristics that have made them successful in several commercial applications. These characteristics, and applications, and some problems inherent to microbend sensors are discussed in this paper. Sensors based on microbend loss in optical fibers were first proposed and demonstrated in 1980 [l], [2], although careful experiments done earlier had demonstrated vibration-induced intensity modulation of light in bent fibers [3]. Since this early work, much has been done to understand microbend sensors and to investigate how to increase dynamic range and improve sensitivity to the measurement parameter of interest while reducing sensitivity to unwanted variables. The early interest in microbend sensors was for hydrophone applications [4]-[6] and this work was driven by the Navy FOSS ( Sensor Systems) program. Since that time, over 100 different studies on microbend sensors have appeared in the literature, and the sensors have been adapted to many different measurement applications. In addition to the general advantages of sensors so often touted (EM1 resistance, electrical isolation and freedom from ground loops, small size and weight), microbend sensors have been demonstrated to have their own unique set of advantages. These advantages include: mechanical and optical efficiency that leads to low parts count and low cost easy mechanical assembly that does not require fiber bonding to other components and thus avoids differential thermal expansion problems fail-safe fiber sensor which either produces a calibrated output signal or fails to a state of no light output.
Manuscript received December 8, 1994. The author is with Babcock & Wilcox-Research sion, Alliance, OH 44601 USA. IEEE Log Number 9412216.

Fig. 1. Original diagram of microbend sensor.

In addition, microbend sensors have been used in hostile environment applications such as in high temperature zones and explosion hazard areas. 11. BACKGROUND AND PHYSICS MICROBENDER TECHNOLOGY

What is microbending? In simplest terms, the mechanical perturbation of a multimode fiber waveguide causes a redistribution of light power among the many modes in the fiber. The more severe the mechanical perturbation or bending, the more light is coupled to radiation modes and is lost. Thus, the important characteristics of a microbend sensor (microbender) are that it uses a multimode optical fiber, it is a light intensity sensor and the light intensity decreases with mechanical bending. Most of these characteristics are problems for the fiber cable designer, but Fields, Asawa, Ramer, and Barnoski came up with a way to enhance the microbending effect by squeezing the fiber between a set of corrugated plates called deformer plates or tooth blocks [ 11. Their configuration is shown in Fig. 1. They found that by tuning the mechanical bend frequency, the microbending loss could be increased by orders of magnitude. Fig. 2 shows plots of sensor transmission versus tooth spacing, for different tooth block displacements, D ,for both graded and step index fiber. The graded index fiber exhibits a resonant response and the step index fiber a threshold response. A microbender is called an intrinsic sensor because light is not permitted to exit the fiber into free space. The microbender is a displacement sensor. When the separation between the tooth blocks changes (see Fig. l), the sinusoidal amplitude of the clamped fiber changes accordingly. The light transmission through the clamped region is a very sensitive function of the sinusoidal amplitude of the bent fiber. Fields analyzed all of these results [2] and arrived at the basic set of equations used to predict microbend sensor performance. For power law refractive index profile,
n 2 ( r )= n:[l - 2 A ( ~ / a ) ~ ] .

The propagation constant difference

and Development Divi-

0733-8724/95$04.00 0 1995 IEEE



For step index fibers (Y = 00 and there is no resonance condition because the difference SP, between adjacent mode pairs is not a constant. However, it is possible to calculate a threshold value for the period of the tooth spacing, and for A less than the threshold, the microbend loss increases.



(b) Fig. 2. Transmission versus tooth spacing (A) for various tooth block displacements (D). (a) Parabolic-profile fiber. (b) Step-profile fiber.

SP, = ( 2 A ' / 2 / a ) ( m / M ) (5) A 5 A, =na/A'/'. (6) Lagakos er al. [7], confirmed that in a microbender, light is lost from the fiber core via coupling between guided modes and radiation modes. In a multimode fiber, the higher-order modes are those modes that are most easily coupled out of the fiber at small bends. Work done by Horsthuis and Fluitman [8], Diemeer and Trommel [9], and Lagakos er al. [lo], provides the microbender designer with information on how to select the tooth spacing on the microbender blocks to achieve best sensitivity. Others including Mavadaat [l 11 have done ray analyses to predict microbend sensor performance. This work was done in 1984 and produced results that agreed well with the modal analyses. However, both the modal and ray models have had difficulty in predicting the light loss versus displacement of the deformer plates, and the saturation effects that occur at large displacements when the core light is substantially depleted. Analyses are still being performed on microbend sensors to improve predictions of light loss versus displacement of the deformer plates. Recent work combines finite element modeling of the sensor mechanics with numerical models of beam propagation [ 121, and wave optics calculations for all the modes in bent multimode fibers [13] hold great promise for understanding effects in microbend sensors such as changes in sensitivity with temperature. In the clamped region of the microbender, it is important to ensure that light coupled out of the core does not re-enter. Yao and Asawa [14] have shown how radiation modes in the cladding can re-enter the core and reduce the sensitivity of the microbender. This work was presented at the First Optical Fiber Sensors (OFS) Conference held in London in 1983. The most straightforward way to remove light from the cladding is to use an optical fiber with an absorptive buffer coating to perform radiation mode stripping along the sensing region.

Two examples of early applications of microbend sensors For graded index fibers, n(r)is the refractive index which is a are described in h i s section. At the Third OFS conference held function of distance, T , between the center of the core and the in San Diego in 1985, A . L. Harmer presented results of a study cladding, A is the normalized index difference between core and clad, a is the fiber radius, p the propagation constant for on the effects of fiber jacket materials on the performance of the mth mode, m the mode number and M the total number of microbenders [15]. The intent of the study was to optimize modes. For (Y = 2, the refractive index profile is parabolic and the performance of a distributed microbender over the entire the difference & I between , , , propagation constants for adjacent length of the jacketed fiber. This was one of the first reports modes is a constant. This allows calculation of the period of the on the development of distributed sensors. He found that fibers tooth spacing A, on the tooth blocks to achieve the resonance with rubber and teflon jackets exhibited microbend sensitivity condition where the microbend loss is sharply peaked (see similar to fibers with no jacket. Harmer's work led to commercialization of a distributed Fig. 2). microbend sensor by Herga Electric. In this sensor described SP, = 2n/A, (3) by Oscroft [ 161 and illustrated in Fig. 3, the fiber is deformed Ac = 2 ~ a / ( 2 A ) l / ~ . (4) by a spiral deformer such as a stiff spring, where the period



S l R " OIOE


UGH1 O U T /


Fig. 4. microbend sensor configured for longitudinal strain measurement.

Fig. 3. Diagrams of the Herga dismbuted microbend sensor, the first commercialized microbender.

plates, and kf is the effective spring constant of the optical fiber clamped between the deformer plates
I C 1 f

of the spiral matches the optimum fiber bend period (see (4)). The Herga distributed sensor has been used in a number of commercial applications including tactile sensing in the bumpers of automatic guided vehicles to provide automatic shut-&. A second interesting application of microbenders was investigated by Jenstrom and Chen at Purdue University [ 171. TWO parallel arrays of optical fibers at right angles form an n x n array where the fiber spacing equals the optimum deformer period A. The parallel array of fibers below acts to deform each of the fibers laid across in the right angle parallel array above, so that an array of m x n microbend sensors is created. Each fiber has its own light source and detector. The problem with decoding the light intensity changes is that with a 5 x 6-fiber array for example, there are 11 known variables, the light intensity change through each fiber, but a total of 30 unknown light intensities at each overlap joint in the 5 x 6 array. Jenstrom and Chen arrived at a very clever algorithm to decode the 11 light intensity changes from each fiber to obtain the pressure distribution and contour over the entire grid. An important result from the design of this array is the linearity and low hysteresis of each microbend sensor in the array and the potential low cost of the arrays, all of which would be ideal for robotic tactile sensing. IV. MICROBENDER ENGINEERING A. General In 1987, Lagakos, Cole, and Bucaro published an excellent analysis [18] on the operation of microbend sensors which built on the early work of Fields [ 2 ] .Lagakos et al. derived some very useful formulas for microbender modeling and design. The change in light transmission, AT, through a microbend sensor can be written as

A3 - ___ - 37rYd4r)'

The effective spring constant for the assembly is an important parameter for the designer, and is expressed in terms of A, (deformer tooth spacing) cubed, Young's modulus Y for the glass, the fiber diameter, d, and the number of bends, 77. The critical parameter from the design standpoint is the sensitivity AT/AX in (7). This parameter depends on the properties of the optical fiber including core and clad refractive indices, the change in these indices with temperature, the type of fiber buffer coating, and the optical power distribution among the many modes in a multimode fiber. The engineering of microbenders for specific applications requires a detailed understanding of how all of these factors affect microbender sensitivity and how to trade off variations of these factors in order to optimize AT/AX for the specific application.
B. Microbender Advantages and Disadvantages

where AX is the displacement of the deformer plates, A T / A X is the sensitivity (slope of the transmission versus displacement response curve for a microbender), A F is the applied force, A,, Y, , and e, are the area, Young's modulus, and thickness of the spacer material between the deformer

Shown in Fig. 4 is a microbender configured as a weldable strain gage. It is useful to examine this simple microbender in some detail in order to understand the specific performance advantages and shortcomings of microbenders. The deformer plates in this application are welded in place, and the period of the tooth spacing A is 1.5 mm. The fiber is step index, polyamide buffered 100/140/170. For this specific configuration, the change in the log of transmission versus applied force is given in the plot of Fig. 5. This plot represents hundreds of measured data points, as does the plot in Fig. 6 which shows measured log of transmission versus displacement of the deformer plates as the plates are moved closer together. The characteristics of this typical microbend sensor response function are worth noting. There are three distinct regions of sensitivity. The first nonlinear region (0-15-pm displacement) where the deformer spacing is approximately the fiber diameter, the intermediate linear operating region (15-50-pm displacement), and the final nonlinear region beyond 50 ,urn, where depletion of light from available modes is observed. The mechanical properties of the fiber buffer coating determine the shape of the first region and whether or not hysteresis is present. The slope of the linear region determines the



0.2 0.4 0.6

0.8 1.0 1.2 1.4 1.6 Load (Ibs.)


Fig. 5 . Measured throughput (log transmission) versus load.

diameter, but is generally in the range 500-2000 Ibshn. To ensure linear and repeatable performance, it is important that the microbender stiffness be small compared to the stiffness of other components in the mechanical system. Microbend sensors share disadvantages common to other intensity sensors. The primary disadvantage, which has many manifestations, is sensitivity to optical power level. In a system, changes in optical power level can result from changes in output of the optical source, changes in coupling between fiber connectors, changes in transmission along the fiber link, and changes in transmission or coupling through other components in series with the microbender. There are also issues with fiber strength and the effects of fiber buffer coating on microbender accuracy and repeatability. Progress has been made in all of these areas over the last several years and each area is briefly discussed in the next few paragraphs.
C. Self-Referencing

Development of microbender signal processing methods to compensate and correct for all of the potential changes in optical power level has been the object of extensive work over the last several years. Various schemes have been tried to self-reference sensors, and several of these self-referencing methods work well with microbenders [20]-[22]. I I I I 431 Time division normalization is a self-referencing method 0 20 40 60 that has been successfully used with microbend sensors to Displacement (micrometers) significantly reduce the error contribution of light intensity Fig. 6. Measured throughput (log transmission) versus displacement. changes caused by cable bending, light source fluctuations and sensitivity which is 0.15 dB/pm, and the location of the knee connector losses. Spillman [20], and Lindsay and Morton [2 11 in the curve (50 pm in this example) sets the high end of the have developed similar types of time referencing methods. In the example shown in Fig. 7, time delay is introduced into dynamic range. Although not as sensitive as interferometric sensors, mi- the sensing arm of a 2 x 2-splitter so that light pulses in crobenders have respectable sensitivity to small displacements the sensing and reference arms arrive back at the splitter at of the tooth blocks. Displacement sensitivities on the order different times. In the receiver, the ratio of the two pulses is of m/& have been reported [ 181, [ 191. The dynamic calculated. Spurious intensity changes which affect both light range of microbenders is remarkable and one of the major signals in the same proportion cancel in the ratio. An error advantages of these sensors. A practical mode of operation is plot, Fig. 8, shows the effect that a 6-dB power change has on to use a photodiode detector to convert the light intensity to the output signal. The error introduced is about one-quarter of a photocurrent, and then process the photocurrent to obtain a one percent of full scale. A similar approach based on wavelength normalization microbender response function which is the log of transmission versus tooth block displacement (see Fig. 6). In this mode, instead of time has been used by other workers for intensity the microbender response to small displacements exhibits a self-referencing. In the example shown in Fig. 9, a broadband linear (1%) dynamic range greater than lo4. This dynamic light source is divided into two wavelength channels-a sensor displacement range is achievable for dc measurands if 0.001- and reference channel. The two channels are recombined, and dB changes in light power can be repeatably detected. This separated again at the receiver. The signals are processed performance is possible using present-day silicon photodiodes in two steps to eliminate light level changes in the fiber link caused by cable bends or connector losses that result and high quality instrumentation amplifiers. Microbenders are most useful as contact sensors for mea- in proportional signal modulation in both channels. First, a surement of small displacements in mechanical systems. In reference light intensity signal from each LED source in general, these applications include those where the mechanical the electro-optic transmitter is used to compensate the corcomponents include two or more springs either in series or in responding signal channel from the sensor. The compensation parallel. In order to make effective use of microbenders in may be performed using bicell photodetectors and log ratio such measurement applications, it is important to understand current amplifiers. This operation on each sensor wavelength the limitations of the sensors and how to design around channel removes optical signal variations caused by light these limitations. The critical consideration from a mechanical source intensity fluctuations. Next, the resultant signal voltages standpoint is fiber spring constant (see (8)). The fiber clamped are subtracted in a difference amplifier. This operation removes in the tooth blocks acts as a spring. The spring constant optical signal variations caused by cable and connector losses, (stiffness) varies with tooth spacing, number of teeth, and fiber which affect both wavelength channels the same (to first



.... .ELEF!?!E.

SENSOR HEAD .................................. ..... CO"ECTORS : MICROBEND OPTICAL : / \ : DEUY \ : / \ : - '0 I / m R R O R =


' ...................




: :

Fig. 7. Time division normalization.

der becomes a high insertion loss component, 20 dB being typical. This high insertion loss severely limits the number of microbend sensors which can be multiplexed. Typically, under these conditions, only two or three high performance microbend sensors can be multiplexed onto a single fiber.
E. Fiber Strength and Buffer Coating


Fig. 8. Error versus power change measured with time division normalization processor.

order). Ideally then, all resulting changes in the difference amplifier output are proportional only to measurand-induced changes in the microbenders.
D. Modal Sensitivity

A significant disadvantage unique to microbenders is the result of the very phenomenon which makes this sensor work. The modal sensitivity of the microbender makes it also a modal filter. Thus, any modal filtering action by other components in the multimode link is a potential error source. The fiber itself is a modal filter as well as fiber connectors and power splitters. Self-referencing methods help to correct for the modal filtering action of the link and the connectors. Careful system choices (e.g., waveguide coupler versus fused biconic coupler) can help avoid unwanted modal filtering by components, and various proprietary methods can be used to reduce modal sensitivity. One method to reduce modal sensitivity is to perform mode mixing at the microbender input and output.
High NAAow NA fiber junctions may be used for this purpose

In addressing the fiber strength issue, remember that deformer plates clamp the fiber. Thus, large stresses can be produced in the fiber. If the plates are brought too close together, the fiber will break. An empirical design rule is to maintain the ratio of maximum applied stress to fiber break stress less than one to four. Under these conditions, the fiber life is effectively unlimited. (One attractive feature for many applications that results from the clamped fiber microbender is the positive "no-light'' indication of a broken fiber.) The 1 :4 stress ratio guideline was confirmed by Miers [23] over ten years ago and was based on extensive cyclic life test data for microbenders. Lifetimes exceeding 10 billion cycles of the deformer plates can be expected for modest fiber stresses of 60000 psi. Since the microbend teeth push into the buffer coating, it is important to understand the mechanical, physical, and chemical interactions among the deformer plate material, the buffer coating material, and the glass optical fiber. Empirical data has been gathered, much of it proprietary. The important issues with any buffer coating material are whether or not it introduces sensor zero shift and span shift. All buffer coating materials introduce these errors to some extent, but the errors can be made negligible by doing a cyclic shake-down of the microbender over the displacement range of the deformer plates. The desirable characteristics for a fiber buffer coating are listed in Table I.
Over the years, microbenders have been configured for

[18]. The trade-off with this approach is that the microben-

measurement of many different parameters including pressure



Fig. 9. Wavelength division normalization (a) and signal processing (b).





Neuual Disadvantages

Wear Resistant

- Repeatability

Easy-to-hold fiber

- Intrinsic - Fail-safe

- Moderate electronics

- Thermal sensitivity - W P C - Insertion loss




I Mining

1 Medical

Chemical Process Environmental

~ ~~



I Power Generation I Robotics I ~tructura~~ivi~


[24]-[27], temperature [28], [29], acceleration [30], [31], flow [32], local strain [33], [34], and speed [35]. Microbender arrays have been used in tactile sensing systems [17], [36] and in distributed sensing systems for temperature [37], strain [38], [39], and structural monitoring [40], [41]. Given in Table I1 is a list of application areas for which microbenders have at least been designed. In some of these areas, prototypes have been built and tested, and in a few of the areas, commercial products have been or are currently offered. For detailed discussions on many of these applications the reader may refer to several excellent references [42]-[44]. It is important to note that in these applications, microbenders have been shown to be stable and repeatable [45], [46]. Accelerated life test data [23] have shown that lifetime of the sensing fiber can exceed 1O1O cycles, and that wear of the fiber buffer coating over the microbend lifetime is not a performance limitation. Table I11 summarizes the performance characteristics of microbenders based on demonstration of prototypes and com-

mercial devices up to the present time. Advantages and disadvantages are listed along with a subjective assessment of the level of electronics complexity needed to do the signal processing functions in order to achieve high levels of performance. Easy-to-hold fiber refers to the mechanically clamped fiber in the microbender configuration. No glass-to-metal adhesives are needed, which is an important feature necessary to achieve repeatable performance. Intrinsic implies resistance to fouling, which is a common problem with sensors that have exposed fiber ends. The failsafe feature refers to the fact that a broken optical fiber fails to a no-light condition. Thermal sensitivity may be improved to <5 x 10-5PC with an independent temperature measurement, but this approach is not always practical. A 20-dB insertion loss for a fully self-referenced and modally conditioned microbender precludes multiplexing of more than just a few sensors on a single fiber. Engineering trade-offs associated with required operating temperature range versus specified dynamic range versus measurement accuracy set the microbender design constraints, which in turn determine the influence that modal filtering has on the overall achievable error limits. What this means in practice is that microbender performance presently achievable is about f0.05% absolute accuracy at constant temperature, about f1% relative accuracy with temperature compensation (0-100C compensated range using a reference microbender as the compensator), and about &OS% absolute accuracy



over the temperature range using an independent temperature measurement [46]. VI. CONCLUSION Microbend sensors have found many applications in metrology, and have been used in several commercial products. Significant improvements in the performance of microbenders have been made in parallel with gains in practical understanding and with the introduction of effective self-referencing methods to correct for changes in optical power level. Continued improvements in performance of microbenders may be expected as the modal sensitivity of components is reduced, and as improved methods of modal mixing and temperature compensation are developed. Even at the present time, however, the overall performance of microbenders is more than adequate for many industrial and military applications, especially in hostile environments. Furthermore, microbenders are readily adapted to distributed sensing. The relatively low cost of the microbender compared to other types of sensors ensures that it will be used in many applications and systems over the next few years.
REFERENCES [I] J. N. Fields et al., pressure sensor, J. Acoust. Soc. Am., vol. 67, pp. 8 1 W 1 8 , 1980. [2] J. N. Fields, Attenuation of a parabolic-index fiber with periodic bends, Appl. Phys. Lett., vol. 36, pp. 799-801, 1980. [3] D. F. Nelson et al., Vibration-induced modulation of fiberguide transmission, in Proc. Topical Meet. on Optical Fiber Transmission TU-E7-I to TU-E7-4, 1977. [4] J. N. Fields and J. H. Cole, Fiber microbend acoustic sensor, Appl. Opt., vol. 19, pp. 3265-3267, 1980. [5] N. Lagakos er al., Microbending sensor design optimization, in Proc. CLEOBI, vol. 100, 1981. [6] ~, Microbend sensor as extended hydrophone, IEEE J. Quantum Electron., vol. QE-18, pp. 1633-1638, 1982. [7] -, Multimode optical fiber displacement sensor, Appl. Opt., vol. 20, pp. 167-168, 1981. [8] W. H. G. Horsthuis and J. H J. Fluitman, Sensitivity dependence on number of bends in a microbend pressure sensor, NTG-Fachberichte 79, pp. 147-152,1982; Development of fibre optic microbend sensors. Sensors and Actuators, vol. 3 pp. 99-1 10, 1983. [9] M. B. J. Diemeer and E. S. Trommel, Microbend sensors: Sensitivity as a function of distortion wavelength, Opt. Lett., vol. 9, pp. 266262, 1984. [LO] N. Lagakos and J. A. Bucaro, Optimizing microbend sensor, Proc. SPIE, vol. 718, pp. 12-20, 1987. [ l 11 R. Mavadaat, Ray analysis of microbend fiber sensors, Sensors and Actuators, vol. 6, pp. 289-295, 1984. [12] M. C. Hastings and D. F. Nippa, Integration of the finite element and beam propagation methods to determine performance of microbend sensors, in Proc. OFS, no. 10, 1994, pp. 376-379. [13] K. H. Wanser et al., Novel fiber devices and sensors based on multimode fiber Bragg gratings, in Proc. OFS, no. 10, 1994, pp. 265-268. S. K. Yao and C. K. Asawa, Microbending sensing, Proc. SPIE, vol. 412, 1983, pp. 9-13. A. L. Harmer, Distributed microbending sensor, in Proc. OFS, no. 3, 1985. pp. 126-128. G. Oscroft, Intrinsic sensor; Proc. SPIE, vol. 734, pp. 207-213, 1987. D. T. Jenstrom and C. L. Chen, A microbend tactile sensor array, Sensors and Actuators, vol. 20, pp. 239-248, 1989. N. Lagakos et al., Microbend sensor, Appl. Opt., vol. 26, pp. 217 1-2180, 1987. J. W. Berthold, High temperature pressure sensor, in Proc. ISA, vol. 30, 1984, pp. 85-94. W. B. Spillman, Jr., and J. R. Lord, Self-referencing multiplexing technique for intensity modulating sensors, Proc. SPIE, vol. 718, pp. 182-191, 1987.

[21] T. A. Lindsay et al., Standard interface for aerospace applications: T h e domain intensity normalization, Proc. SPIE, vol. 989, pp. 45-55, 1989. [22] M. T. Wlodarczyk, Environmentally insensitive commercial pressure sensor, Proc. SFIE, vol. 1368, pp. 121-131, 1991. [23] D. R. Miers etal., Life tests of optical fibers in a microbend configuration, presented at ACerS, 88th Annual Meeting, Glass Div., Chicago, IL, 1986. a r t II. [24] D. A. Krohn et al., Understanding s for automated control: P Sensing techniques, applications, and economics, Pbnt Eng., vol. 37, pp. 56-58, 1983. [25] K. Spenner, Microbending pressure and displacement sensor, in Proc. OFS, no. 3, 1985, pp. 146-148. [26] J. W. Berthold etal., Design and characterizationof a high temperature pressure transducer, J. Lighwave Technol., vol. LT-5, pp. 870-876, 1987. [27] S. E. Reed et al., Total pressure transducer for aircraft applications, P ~ c SPIE, . vol. 2070, pp. 17-23, 1993. [28] 0. Lumholt et al., Simple low-temperature sensor that uses microbending loss, Opt. Len., vol. 16, pp. 1355-1357, 1991. [29] A. Cutolo and M. Gallo, Microbending optoelectronic sensor for on-line t e m p e m measurements in high-power electrical systems, European Trans. on Elect. Power Eng., vol. 1, pp. 281-287, 1991. [30] J. B. Freal er al., A microbend horizontal accelerometer for borehole deployment, J. Lightwave Technol., vol. LT-5, pp. 993-996, 1987. [31] D. R. Miers et al., Design and characterization of accelerometers, Proc. SPIE, vol. 838, pp. 31k317, 1988. [32] R. E. Sovik, Implementation of s in a vortex shedding flowmeter, in Proc. ISA Digitech 85, 1985, pp, 207-21 1. [33] S. E. Reed and J. W. Berthold, Development of a microbend strain gage, in Proc. SEM Fall Con$ on Experimental Mechanics, Keystone, CO, 1986; also see US. Patent 5,020,379. [34] J. D. Weiss. Strain gage, J. Lightwave Technol., vol. 7, pp. 1308-1318, 1989. [35] D. Varshneya et al., Speed sensor for advanced gas turbine engine control, Proc. SFIE, vol. 1367, pp. 181-191, 1991. [36] J. S . Schoenwald and J. F. Martin, Tactile sensor for robot grippers, Soc.Manu5 Eng. MS84-1041,1984. [37] L. Falco and P. Debergh, Bimorphous distributed transducer for temperature threshold sensor, in Proc. SPIE, vol. 1011, 1989. pp. 166-172. [38] C. K. Asawa et al., High sensitivity strain sensors for measuring structural distortion, Electron. Lett., vol. 18, pp. 362-364, 1982. [39] A. J. Rogers, Distributed optical fiber sensing, Proc, SPIE, vol. 1508, pp. 2-24, 1991. [40]E. Udd et al., Microbending sensors for smart structures, Proc. SPIE, vol. 1170, pp. 478-482, 1990. [41] K . H. Wanser, er al., Distributed sensors for civil structures using OTDR, CA State Fullerton Conf. Monograph 9308W3, 1993, pp. 303-327. [42] Optical Fiber Sensors: Systems and Applications, B. Culshaw and J. Dakin, Eds. Norwood, MA: Artech House, 1989. [43] Sensors: An Introduction for Engineers and Scientists, E. Udd, Ed. New York: Wiley, 1991. [U] Handbook of Intelligent Sensors for Industrial Automation, N. Zuech, Ed. Reading, M A Addison-Wesley, 1992. [45] J. W. Berthold, Field test results on pressure transmitter system, Proc. SPIE, vol. 1584, pp. 3 9 4 7 , 1991. [46] J. W. Berthold et al., Flight test results from FOCSI total pressure transducer, Proc. SPIE, vol. 2295, pp. 216-222, 1994.

John W. Berthold In was born in York, PA, on May 24, 1945. He received the B.A. degree in physics from Gettysburg College, and the M.S. and Ph.D. degrees in optical sciences from the University of Arizona. His thesis work was in application of lasers to precision measurement of small length changes. He has performed research in thin film filters and nonlinear optics at Bell Laboratories, and done integrated optics device development for the U.S. Department of Defense. Since joining McDermotVBabcock & Wilcox in 1979, he has been engaged in industrial applications of optical systems, and sensor and transducer development. D r .Berthold is a Fellow of SPIE, a senior member of ISA, a member of OSA. He holds 25 US. patents.