calibration technique for mems membrane type strain sensors

8/6/2019 Calibration Technique for MEMS Membrane Type Strain Sensors

1/7

Calibration technique for MEMS membrane type strain sensorsLi Cao", Tae Song Kimb, Jia Zhou", Susan C. Mantell"", an d De nnis L. Pollab

" Dept. of Mechanical Engineering, University of Minnesota, Minneapolis, MN 55455Dept. of Electrical Engineering, University of Minnesota, Minneapolis, MN 55455* Corresponding authorAbstract- A MEM S based piezoresistive strain sensor wa sdesigned, fabricated and calibrated. A single strip of doped n-polysilicon sensing material was patterned over a thin S i3N4/Si02membrane. The silicon wafer was etched beneath this thinmem brane. The intent of this design was to fabricate a flexibleMEMS strain sensor. A calibration technique for measuring thestrain sensor performance is described. The sensor calibrationtechnique (to find the relationship between chan ge in resistanceand stra in) entails developing a repeatable relationship betweenthe chan ge in senso r resistance and the strain m easured at thesens or. The sen sor sensitivity is evaluated by embedding thesensor in a vinyl ester epoxy plate and loading the plate. Thiscalibration technique captures the effectsof strain transfer to thestiff silicon w afer.

structure, only 6 0 0 - 1 O O O p occurred in the silicon wafer.These researchers concluded that MEMS strain sensorresponse must be sensitive so as to produce the desiredresolution for sub seque nt signal processing.

In this paper, a method for evaluating strain given thechange in sensor resistance is presented. The method ,referred to as a calibration technique, is based onexperimental data and follows accepted standards for metalfoil strain gage calibration . In the subsequent sections, straingage calibration is describ ed, M EM S strain sensor calibrationis described, and test data for MEMS strain sensor arepresented.

1. Introduction2. Background

In a separate project, a MEMS strain sensor wasfabr ica ted for remotely m onitoring of the structural health of afiber reinforced laminated composite structure. The overallsensor s yste m consists of a strain sensor, a signal conditioningand telem etry circu it, and an antenna. The entire system is tobe embe dde d in the composite structure. This sensor holdstwo distinct advantages over conventional metal foil straingages i n that the ov erall sensor system does not require a w ireconnection to the interrogation systems and is able to becofabricated on a single chip by using MEMS technology.These completely integrated devices are well suited forcomposite aircraft components (door panels, fuselage), bridgeand sh ip structures. Detection of excessive loads can helpavoid ca tastrophic failure of these structures.

Cri tica l to this application is the sensor sensitivity. Th eMEMS strain sensor must be sufficiently sensitive to measurestrain up to 2000pe in the structure. Three different MEMSbased piezoresistive strain sensors were already developed byHautamaki et al at the University of Minnesota [l]: apiezoresistive filament fabricated directly on the wafer, arectangular cantilever beam and a curved cantilever beam.Typical sensor sensitivity to a uniaxial tensile strain ofl 0 O O p ~anged from 1.0 to 1 .5% of the nominal resistancechange (AWR), corresponding to a gage factor of -10 to -15(Negative gage factor indicates that resistance decreases withincreasing strain.). Hautamaki et al [ I ] found that the straintransfer from the polymer composite to the sensor was onlyon the order of 30-50%. That is , for 2 0 0 0 p load on the

Standard calibration for metal foil strain gages isexperimentally and statistically based. Several gages for eachlot are tested to determine the sensitivity for the entire lot.ASTM E251- 92 [4] is the standard test method forperformance characteristics of metallic bonded resistancestrain gages. Th e sensitivity of a strain gage is characterizedby a gage factor. Th e gage factor GF is related to the nominalresistance R, the change in nominal resistance AR and thestrain E as follows

A R / RGF = ~ E

According to A ST M E 25 1-92, there are three separate testmethods for calibration of the gage factor: (1 ) Constantbending moment beam test method; (2) Constant stresscantilever beam test method; and (3) Direct tension orcompression test meth od. Eve ry method has its typicalmechanical system and specimen, but all have the sameprocedure. In each, the strain gage is mounted on a metalspecimen and calibrated by m echanically loading thespecimen. I t is important to recognize that since strain gagesare part of a complex system and cannot be reinstalled, gagefactors are stated only on a statistical basis. Currentcalibration techniques as sume that strain gages from the sam efabrication lot will have the same gage factor as the devicesthat are mounted and tested.

2040-7803-5240-81991510.0001999 IEEE


2/7

A similar approach can be applied to experimentallycalibrate MEM S strain sensors. Several ME MS strainsensors from the same lot must be tested and calibrated. Th eMEMS strain sensor is fabricated on a stiff silicon wafer.Like the metal foil strain gage, the strain is transferred to theMEMS strain sensor through an adhesive bond (i.e. the strainsensor is either bonded or embe dded in the structure ). Siliconhas such a high modulus compared to the surroundingadhe sive that strain transfer is po or [11. Near the sensor, thereis a mismatch between the stiffness of the silicon and theadhesive that bonds the sensor to a specimen or encapsulatesthe sensor in an embedded specimen. On e way tocharacterize this effect is to introduc e the terms nearfield andfarfield strain [3]. Nearfield strain is the actual strain which ismeasured at the MEMS strain sensor. Farfield strain is thestrain which would occur at that location had there been nosilicon wafer to disturb the strain field. The silicon introducesnearfield (sensor) strain disturbances. The nearfield straincan be as sma ll as 30%of the farfield strain.

For conventional metal foil strain gages on a flexiblepolyimide backing, the strain transfer between the specimenand sensor is perfect. The ASTM E251-92 standard assumesthat the mechanical (farfield) strain that is determined fromthe applied load is the same as the (nearfield) strain that istransferred to the strain gage. Th is procedure is not valid forthe stiff silicon backed MEM S sensor. Th e calibrationprocedure must be modified to account for the strain transferto the MEMS sensor. The outcom e of sensor calibratio n is agage factor, which is used to relate the change in resistance tothe nearfield strain at the senso r. In ord er to calibrate aMEMS sensor fabricated on a silicon wafer, the nearfieldstrain at the sensor must be know n.

3. Sensor design and fabricationThe MEMS strain sensor resembles a picture in a frame.

The frame is a lOmm long x lOmm wide x 0.5m thick siliconwafer. The picture is a thin Si3N4/Si02 mem brane with n-polysilicon pattern, as shown in Figure 1. The sensingmaterial is patterned over the thin membrane. A llOOpm xIlOOpm window of the silicon wafer is etched beneath thethin membrane . The thin membrane provides a flexiblebacking for the sensing material. Th e intent of this design isto fabricate a flexible MEMS strain sensor, similar in functionto a flexible metal foil strain gage. A scanning electronicmicroscope (SEM) for this single strip sensor is shown inFigure 2. Th e detailed fabrication steps for this sensor aredescribed in [4].

Thin membrane window

Bonding padsSingle strippolysiliconSilicon frame sensingmaterial

Figure 1. Top view of the windowed single strip sensor.

Figure 2 . SEM of the patterned polysilicon in a windowedsingle strip sensor. Four sensors oriented at 90" are shown.

4. Experimental proceduresFor calibration (i.e. determine the gage factor), thewindowed single strip sensor was embedded in a vinyl esterepoxy plate. Vinyl ester epox y was selected because i t is aroom temperature cure, transparent polymer. Roomtemperature adhesive properties of vinyl es ter are comparableto typical adhesives used to mount strain gages. Twospecimens were constructed: i ) a strain g age sp ecime n with anembedded commercial strain gage (EA-00- 125AD-120)mounted on a blank lOmm x lOmm x 0.5mm silicon wafer;and i i ) a M EM S specimen with an embedded M EMS singlestrip sensor of iOmm x lOmm x 0.5mm overall dimensions.The specimen material (vinyl ester epoxy) and geometry(160 mm x 30mm x 6.0mm) are identical. The strain gagemounted on a blank silicon wafer in the strain gage specimenhad the same size and geometry as the MEMS strain sensor.

For each specimen, the embedded device was placed at aheight one quarter the overall thickness and at the end of thespecimen, as shown in Figure 3. This location was selected toensure good sensor response in bending. In addition to the

205


3/7

embedded devices for either MEMS strain sensor or straingage on silicon, a conventional strain gage (EA-00-125AD-120) was bonded to the outside of each specimen, directlyabove the embedded sensor. Th e surface strain from theapplied load to the specimen was measured with the exteriorstrain gage . The farfield strain was calculated from lineartransformation of the surface strain for bendin g by using [5]

where h is the distance from the surface mounted strain gageto the emb edded sensor and T is the overall thickness of thespecimen. Th e farfield strain was calculated from the surfacestrain for tension by using

Two different loadings were applied, bending and uniaxialtension. A cantilever beam m ethod w as used to apply to thebending load [6]. Th e specimen was held rigid at one end. Abending load, ranging from 0 to 11.8kN, was applied to thefree end of the specime n. Ben ding tests were run in bothtension and compression bending by turning over thespecimen. Th e same specimen was loaded in tension on aMTS Test-QT/lO machine. A tension load, ranging from 0 to1.6kN, was applied.The schematic of the calibration procedure is shown in Figure4. Both specimens were loaded under identical loadingconditio ns. Th e strain gage specim en was tested to establisha relationship between the farfield strain and the nearfieldstrain in bending (bending relationship) and tension (tensionrelationship), respectively. Th e M EM S specime n was testedto obtain the farfield strain versus the sensor nominalresistance in bending (bending data) and tension (tensiondata), respectively. In the M EM S specimen, we cannotrecord the nearfield strain, only the nearfield resistance.The farfield strain from the MEMS specimen bending dataand tension data must be converted to the MEMS nearfieldstrain using the two relationships. Th e bendin g data isconverted using the bending relationship and the tension datais converted using the tension relationship. Th e response ofthe sensor in bending was determined by plotting the changein the MEMS nearfield strain as a function of the change inthe nomin al resistance. This step was repeated to obtain theMEMS nearfield tensile strain as a function of change innominal resistance. Th e gag e factor (Equatio n 1) of thesensor is the slope of each respon se.

Side ViewMEM S sensor or strain gage on silicon

\ Exterior strain g ag e

?/4Top view

Fig ure 3. Sketch of Specimen. Length (L) = 160mm, width(W ) = 30mm and thickness (T) = 6.0mm.

5. Results and discussionFigu res 5(a) and 5(b) show the bending relationship andthe ten sion relationship between the farfield and nearfieldstrain in the strain gage specimen, respective ly. The changesin farfield strain with the nearfield strain were linear for bothben ding and tension loads.Figures 6(a) and 6 (b) show the bending data (a) andtension data (b) for the MEM S strain sensor. The change insenso r nominal resistance versus the change in farfield strainis shown. The normalized resistance change increasedlinea rly with the farfield strain as a compression bending loadwa s applied. Similarly, the normalized resistance changedecre ased as a tension bending or a tension load was applied.Th e bending and tension responses for the sensor (afterconv ersion to the nearfield strain) are show n in Figures 7( a)an d 7(b). The nearfield strain was determin ed from the strainga ge specim en data that were collected unde r similar loadingconditions. The normalized resistance change with

206


4/7

Strain gage specimen-Nearfield strainEmbedded strain gage onsilicon

versus

Exterior strain gage

MEMS specimen-ResistanceEmbedded MEMS strainsensor

Figure 4. Schematic of the calibration procedure.

Resistance

207


5/7

the nearfield strain was linear for each design. Th enormalized resistance change decreased linearly withincreasing n earfield strain fo r both bending and tension loads.Gage factors for bending and tension loads are the slopes ofthe data shown in Figures 7(a) and 7(b). The gage factors fo rthe single str ip sensor were -15 in bending (a) and -13 intension (b).

-3C-.-eEa, -4000Y

h

- 1200.- 1000E 8006005 400

Figure 5.

1200 -1000 -800 -

-1200Far FieldStrain (ue)

0 2000-4000

Farfield strain (ue)

(b)Jearfield and farfield strain for

4000

ne strain gage onsilicon specime n in a) bending and b) uniaxial tension.Nearfield strain wa s recorded directly at the embedded straingage. Farfield strain was calculated based on the surface

strain from the e xterior strain gage and the specimenthickness (follow ing equation 2 and equation 3).

2*5

K.4000

-1.5

-2.51Far F ield Strain (ue)

0.5 T

-2.5Far Field Strain (ue)

Figure 6 . Data for the window ed single strip sensor in a)bending and b) uniaxial tension. Normalized resistancechange AR/R was recorded at the sensor. Farfield strain wascalculated based on th e surface strain from th e exterior straingage and the specimen thickne ss (following equation 2 an dequation 3).

208


6/7

tension for a monofilament sensor.

2.5 T

0 1.5 -0

I-2-2.5

Near fieldStrain (ue)

2000 4000a

-2-2.5 1

Near Field Strain ue )

Figure 7. Respo nse of the windowed single strip sensor in a)bending and b) uniaxial tension. Nearfield strain was basedon the strain gage specimen data. Th e slope is the gagefactor.The ideal sensor would have a gage factor that is consistentregardless of the loading. The high gage fac tor indicates highsensitivity to changes in strain. The windowed single stripsensor performs consistently. The gage factor is -15 inbending and -13 in tension. The gage factor has a smallvariation in response between bending and tension loads. Thesensitivity in bending is slightly higher than in tension. Theresults are comparable to that measured by Hautamaki et al[2] for a monofilament strain senso r design. Theseresearchers rep orted g age factors of - 17 in bending and - 16 i n

The gage factors of MEMS fabricated piezoresistivesensors on the thin membrane, which were obtained frombending and tension tests, are larger than those ofconventional metal foil strain gages . Gag e factors for metalfoil strain gages are 2.0 [ 7] . The high magnitude of the gagefactor indicates high sensitivity to changes in strain.Moreover, the high gage factor magnitude reduces the signalprocessing requirements and is co mpatible with lower powerelectronics.

6. ConclusionsA MEM S piezoresistive (do ped n-polysilicon) strain

sensors on a 3pm thin Si3N4/SiOr membrane weresuccessfully designed, fabricated and calibrated. Acalibration technique for a MEMS strain sensor wasdescribed. The calibration establish es a relationship betweenthe change in nominal resistance and the change in nearfieldstrain. The calibration technique was based on experimentaldata and followed the standard procedure similar to metal foilstrain gage calibration. Nearfield and farfield strain con cep tswere introduced to account for poor strain transfer betweensilicon and adhesives. Th e gage factors for the M EM Smembrane type strain sensor design were comparable to thosefor the MEMS strain sensors observed by Hautamaki et alresearchers [11.

7. AcknowledgmentThis research is funded by the US Naval ResearchLaboratory (NR L), contract n um ber NOOO14-94-C-223 1.

8. References[l ] Charles Hautamaki, Shay ne Zurn, Susan C. Mantelland Dennis L. Polla, Embedded MicroelectromechanicalSystem (MEM S) as a Strain Sensors in Composites, Journa l

of MEMS, i n press, 1999.[2] American society for Testi ng and Materials, StandardTest Methods for Performance Characteristics of MetallicBonded Resistance Strain G ages, AS TM E25 1-92 standard.[3] Toru Itoh and George S. Springer, StrainMeasurement with Microsensors, Journal of CompositeMaterials, Vol. 31(19), 1944-1984, 1997.[4] Tae Song Kim, Li Cao, Susan C. Mantell and DennisL. Polla, Fabrication of Piezoresistive Membrane TypeStrain Sensors by Using MEMS Technology, submitted toSensors and Actuators, by March 1999.

209


7/7

[ 5 ] Charles Hautamaki, Discrete EmbeddedMicrosensors in Laminated Composites, Dissertation, 1998.[6] A . Higodon, E.H. Ohlson, W. B. Stiles, J. A. Weese,and W. E Riley, Mechanics of Materials, 4h ed., John

Willey & Sons, 242, 1985.[7 ] J.W. Dally and W.F. Riley, Experimental StressAnalysis, 3rd,McGraw-Hill, New York, 166, 1996.

210

calibration technique for mems membrane type strain sensors

Documents