Design and Analysis of a Novel MEMS Dual Axis Accelerometer

Transcription

1 Design an Analysis of a Novel ES Dual Axis Accelerometer Zijun He, Xingguo Xiong, Wei Quan Abstract Due to their small size, low weight, low cost an low energy consumption, ES (icroelectromechanical Systems) evices have achieve great commercial success in recent ecaes. ES accelerometers have been wiely use in automobile airbag eployment systems, inertial navigations, etc. In this paper, the esign an simulation of a novel bulk-micromachine capacitive ES ual axis accelerometer base on Silicon-on-Glass (SoG) structure is propose. Theoretical moel is use to analyze the working principle of the accelerometer. Base on analysis, a set of optimize esign parameters are suggeste. ANSYS simulation is use to verify the function of the evice. The evice has Silicon-on-Glass compoun structure base on silicon-glass anoic boning, which can reuce parasitic capacitance an ease the signal etection. DRIE is use to pattern the silicon evice structure. The fabrication flow of the ES accelerometer is also suggeste. The propose accelerometer can be use for -axis inertial navigation applications. eywors: icroelectromechanical Systems (ES), Accelerometer, Dual-axis Accelerometer, Silicon-on-Glass (SoG), Inertial Navigation. INTRODUCTION icro-electro-mechanical systems (ES) are evices an systems integrate with mechanical elements, sensors, actuators, an electronic circuits on a common silicon substrate through microfabrication technology. As a newly evelope interisciplinary fiel, ES have achieve tremenous progress in recent ecaes. Due to their small size, low cost, low energy consumption an high resolution, ES have foun applications in many ifferent fiels. As an example, ES accelerometers have been use in airbag eployment systems in automobiles, inertial navigation system, consumer proucts, etc. ES accelerometers base on ifferent working principles have been reporte, such as piezoresistive accelerometers [], piezoelectric accelerometers [], capacitive accelerometers [], an optical accelerometers [4], etc. ES accelerometers are generally esigne to sense acceleration only along one certain sensitive irection. If acceleration sensing along multiple axes is neee, several ES accelerometers with ifferent sensitive irections are integrate. However, a more efficient solution is to fabricate a single ES accelerometer which can simultaneously sense accelerations along multiple axes (e.g. -axis or -axis accelerometers). Such multiple-axes accelerometers lea to more compact esign an better signal integration. However, the esign of multiple-axis accelerometers is more challenging ue to signal coupling issue among ifferent axes in the working moe. Some research work on multiple-axis ES accelerometers has been reporte [5]-[]. As shown in Fig. (a), a ES ual-axis accelerometer integrating two iniviual ES single-axis accelerometers in a single chip is reporte [9]. It simply integrates two single-axis ES comb accelerometers in perpenicular to each other, with one sensing the acceleration along X irection an another one sensing the acceleration along Y irection. The two evices are inepenent from each other with separate beam, mass an comb fingers. As a result, the evice esign is simple, an there is no signal coupling issue Department of Electrical Engineering, University of Brigeport, CT 664, hzj4@hotmail.com Department of Electrical Engineering, University of Brigeport, University Avenue, Brigeport, CT 664, xxiong@brigeport.eu Department of Electrical Engineering, University of Brigeport, Brigeport, CT 664, weiquan@brigeport.eu ASEE Northeast Section Conference University of assachusetts Lowell Reviewe Paper April 7-8,

2 between two evices. However, such a simple esign is not efficient an compact. As shown in Figure (b), a ES ual-axis accelerometer base on a single evice structure is reporte []. It utilizes ifferential capacitive sensing of comb finger groups attache to a single proof mass. The proof mass is supporte by 8 fole beams properly arrange so that the mass can move along both X an Y irections. The 8 fole beams are locate in the four corners of the square mass, with each corner containing fole beams. If there is acceleration inputs along X an Y irections, the resulte inertial forces lea to bening of the fole beams, hence the proof mass move for a certain isplacement along X an Y irections. There are four set of movable fingers extruing from the top, bottom, left an right sies of the square mass. For each movable finger, there are fixe fingers in its left/right (or top/bottom) sies, an together they constitute ifferential capacitance. When there is no acceleration input, the movable fingers stay in the mile between left/right (or top/bottom) fixe fingers, hence left/right (or top/bottom) capacitances are equal. If the movable fingers move ue to inertial force, the capacitance gaps change an the resulte ifferential capacitance change can be measure, hence the input acceleration along X an Y irections can be etermine. Various other ual-axis ES accelerometers have been reporte. Due to their functional efficiency an compact structure esign, ES ual-axis accelerometers have attracte interest from researchers an will continue to improve in the future. (a) Integration of separate evices in a single chip (b) Dual-axis accelerometer with single evice structure Figure. Reporte research work of ES ual axis accelerometers [9][] In this paper, a ual-axis ES accelerometer base on ifferential comb capacitance sensing is reporte. The evice utilizes silicon DRIE (Deep Reactive Ion Etching) bulk-micromachining technique to increase the evice thickness to more than µm. Compare to poly-si surface-micromachine evice, the evice capacitance can be greatly increase, which makes the ifferential capacitance signal sensing easier. Further, the evice utilize siliconon-glass (SoG) compoun structure. By using glass (insulator) as substrate, the parasitic capacitance can be greatly reuce compare to accelerometers using silicon as substrate. The evice utilizes a single proof-mass for sensing of acceleration along both X an Y irections, which greatly improves the functional efficiency an leas to more compact structure esign. The propose ES ual-axis accelerometer can be use for inertial navigation system where acceleration sensing along two ifferent irections is requie. DUAL-AXIS ACCELEROETER DESIGN The propose bulk-micromachine ES capacitive ual-axis accelerometer structure is shown in Fig.. As seen in Figure, an H-shape central mass is connecte to four vertical fole beams, which are in turn connecte to two horizontal straight beams with one en anchore to substrate. There are horizontal an vertical movable fingers extruing from sie an central mass respectively. Vertical movable fingers constitute ifferential capacitance with left/right fixe fingers on the top/bottom, an horizontal movable fingers constitute ifferential capacitance with top/bottom fixe fingers. The esigne vertical fole beams an horizontal straight beams can ben along horizontal an vertical irections separately. As a result, if there is a horizontal acceleration, the resulte ASEE Northeast Section Conference University of assachusetts Lowell Reviewe Paper April 7-8,

3 inertial force on central mass causes the vertical fole-beams to ben. The vertical movable fingers move an by measuring the resulte ifferential capacitance change of vertical fingers, the input horizontal acceleration can be erive. If there is vertical acceleration input, the resulte inertial force on central mass causes the horizontal straight beams to ben. The horizontal movable fingers move an by measuring the resulte ifferential capacitance change of horizontal fingers, the input vertical acceleration can be erive. This is the working principle of the ualaxis ES accelerometer. Figure. Structure esign of the ES ual-axis accelerometer The ifferential capacitance sensing for ES comb accelerometer is shown in Figure. For simplicity, only a single capacitance group with one movable finger an its left/right fixe fingers is shown in the figure. The movable finger constitutes ifferential capacitance c an c with its left an right fixe fingers. When there is no acceleration (a=), the movable fingers stay in mile between left an right fixe fingers. As shown in Figure (a), left capacitance C equals to right capacitance C an it can be expresse as C N S N tl r r f C C () where N is the number of comb finger groups, ε r is the relative ielectric constant of air, ε is the ielectric constant of vacuum (ε = F/m), S is the overlap area between movable finger an fixe finger, t is evice thickness, L f is the length of movable fingers, an is static capacitance gap between movable an fixe finger (when a=). (a) Without acceleration (a=) (b)with acceleration (a ) Figure. Differential capacitance sensing in ES accelerometer When there is acceleration input along horizontal irection, the movable fingers move towar left or right ue to the inertial force (see Figure.b). As a result, the ifferential capacitance changes. If input acceleration is towar left irection, the movable fingers move towar right with isplacement x along right irection, uner small eflection approximation (assume x<< ), then ASEE Northeast Section Conference University of assachusetts Lowell Reviewe Paper April 7-8,

4 C C N N tl r S r ( x) x x C C C f N N tl r S r ( x) x x C C C f In orer to sense the ifferential capacitance change ΔC=C (x/ ), moulation voltages +Vs an Vs are applie to the left an right (or top an bottom) fixe fingers separately, as shown in Figure 4, which has been iscusse in many reference books. () () Figure 4. Differential capacitance sensing Assume the voltage output in the movable finger is V o. Accoring to charge conservation law, the charge store in capacitance C an C shoul be equal. That is, V V ) C ( V V ) C () Solving above equation, voltage output V o in movable finger can be calculate as Base on Equations () an (), we have: ( o s o s C o V s C C C C C (5) C C C C (6) Substituting Equations (5) an (6) to Equation (4), the output voltage V o can be calculate as C x Vo Vs Vs C V From Equation (7), the output voltage is linearly preoperational to the isplacement of movable finger. Uner small eflection approximation, the fole/straight beams can be treate as springs with effective spring constant k tot. In equilibrium state, the inertial force ue to input acceleration an the restoring force of the beams are equal to each other. That is, F a k x (8) inertial Where s is the total sensing mass (incluing the proof mass, connectors an all the movable fingers), a is the input acceleration. Thus the input acceleration a can be calculate as Then, making use of Equation (7), the acceleration is foun to be proportional to voltage output: ktotv a () svs As a result, the acceleration can be calculate by measuring the voltage output in the movable fingers. This is the working principle of most of the ES capacitive accelerometers. s ktotx a s tot (4) (7) (9) ASEE Northeast Section Conference University of assachusetts Lowell Reviewe Paper April 7-8,

5 DESIGN OPTIIZATION AND SIULATION Assume the with, length an thickness of each section of the fole beams are W b, L b an t b separately, an the with, length an thickness of each straight beam are W b, L b an t b separately. The mass of H-shape mass is h. Given Young s moulus of Si material as E, an gravity acceleration is g. Each section of the fole beam can be treate as ouble-clampe beam moel. Its spring constant can be calculate as EI b Etb W b b L L () For each fole beam, two sections are connecte in series. Thus, the spring constant of each fole beam is b EI b Etb W b b Lb Lb The four fole beams are connecte in parallel, thus the total spring constant x _ of four fole beams along X tot axis is: 4EI b Etb W b x _ tot 4b Lb Lb () where I b is moment of inertia of one section of the fole beam ( Ib tb W b). The spring constant of one straight beam is b, Each straight beam can be treate as ouble-clampe beam moel, with its spring constant EI b EtbWb b L L () an the total spring constant of two straight beams ( ) along Y axis is 4EI (4) b EWb tb b Lb Lb where I b is moment of inertia of one straight beam Ib Wb tb. The X-axis isplacement sensitivity of the accelerometer is efine as the isplacement of movable fingers along X irection in response to g acceleration along X irection. It can be calculate as x _ inertial x _ tot Corresponingly, the Y-axis isplacement sensitivity of the accelerometer is: The resonant frequency of the accelerometer along X irection is x _ f x The resonant frequency of the accelerometer along Y irection is S S x y b F F f y y _ inertial Base on those equations, we plot the relationship between X-axis (Y-axis) isplacement sensitivity an the with of fole beams (straight beams) respectively, as shown in Fig. 5. As shown in the figures, the X-axis (Y-axis) sensitivity is very sensitive to the with of fole beams (straight beams). If the beam with is reuce, the sensitivity increases very rapily. The curves can guie us in the evice esign optimization. Base on the analysis, b b b h g x _ tot hg h tot h (5) (6) (7) (8) ASEE Northeast Section Conference University of assachusetts Lowell Reviewe Paper April 7-8,

6 Displacement Sensitivity (um/g) Displacement Sensitivity (um/g) we erive a set of optimize esign parameters of the ES accelerometer, as shown in Table. The gap between two neighboring movable fingers is μm, the thickness of the evice is 8μm..5 Relationship between the with of fole beam an Sensitivity 5 Relationship between the with of straight beam an Sensitivity S=.96um per g.5..5 S=.8um per g With of fole Beam (um) With of Straight Beam (um) Figure 5. Displacement sensitivity versus the with of beams Table. The optimize esign parameters of the ES accelerometer Component Amount Length(μm) With(μm) Central mass 5 Sie mass 4 5 ovable Fingers 6 4 Fole beams 4 5 ( section) 4 Straight beams ( section) 4 Connection mass 4 Connector 4 5 Connector 4 All of above analysis an calculation are base on theoretical moel. In orer for more accurate analysis, ANSYS FE simulation is use verify the function of the evice. Note that in ANSYS moel, only the movable parts an the anchors are isplaye. The fixe comb fingers only contribute to the ifferential capacitance of the evice, but o not contribute to the isplacement sensitivity analysis, resonant frequency analysis an stress istribution of the movable parts, thus they are not shown in the simulation. Sensitivity Simulation Displacement sensitivity of the accelerometer is simulate in ANSYS. Input acceleration (g) is applie to the evice in X irection an Y irection separately. Figure 6 shows ANSYS simulation result of isplacement sensitivity when experiencing g (g=9.8m/s ) acceleration along X or Y irections. From ANSYS simulation results, it is shown that the isplacement sensitivity along X irection is S x =.5nm/g, an the isplacement sensitivity along Y axis is Sy=7.64nm/g. This is in goo agreement with our han calculation results. ASEE Northeast Section Conference University of assachusetts Lowell Reviewe Paper April 7-8,

7 Stress Simulation (a). acceleration along X irection (b). acceleration along Y irection Figure 6. ANSYS isplacement sensitivity analysis when acceleration a=g Stress istribution analysis can tell which part of the evice experiences the largest stress when evice is in its working moe. This is very important for the long term reliability of the evice. Stress istribution of the accelerometer uner the eflection in riving an sensing moes is simulate in ANSYS. Figure 7 shows the ANSYS contour plot of stress istribution when the evice experiences acceleration input of 5g along X an Y irections. From ANSYS stress contour plot, we can see that the maximum stress occurs on the beams close to the anchors or close to the connectors of the beam section. As a result, in the evice esign, we intentionally wien these parts so that they can withstan more stress without being broken. (a). acceleration along X irection (b). acceleration along Y irection Figure 7. ANSYS stress istribution analysis when input acceleration a=5g ASEE Northeast Section Conference University of assachusetts Lowell Reviewe Paper April 7-8,

8 DEVICE FABRICATION The fabrication sequence of the bulk-micromachine comb accelerometer is shown in Figure 8(a-i). The evice is base on Silicon-on-Glass (SoG) compoun structure. Silicon-glass anoic boning an DRIE etching are use in the fabrication. Figure 8. The fabrication sequence of the accelerometer CONCLUSIONS AND FUTURE WOR In this paper, the esign an simulation of a novel bulk-micromachine capacitive ES ual axis accelerometer is propose. Due to the esign of beams, the H-shape mass can move along X an Y irection separately uner corresponing inertial forces. Hence, the input acceleration along X an Y irections can be measure by ifferential capacitance sensing. Base on the analysis, an optimize esign is suggeste. ANSYS simulation is use to verify the evice performance. The fabrication flow of the evice is also propose. In the future, we will look into the signal coupling between X an Y axis of the evice, an minimize its influence by improving the evice esign. REFERENCES [] Haris,. an Hongwei Qu, "A COS-ES piezoresistive accelerometer with large proof mass," 5th IEEE International Conference on Nano/icro Engineere an olecular Systems (NES), pp.9-, Jan. -,. [] C.C. Hinrichsen, J. Larsen, E.V. Thomsen,. Hansen an R. Lou-ller, "Circular piezoelectric accelerometer for high ban with application," 9 IEEE Sensors, pp , Oct. 5-8, 9. [] C.H. Je, S. Lee,.L. Lee, J. Lee, W.S. Yang, C.A. Choi, "Z-axis capacitive ES accelerometer with moving groun masses," IEEE Sensors, pp.65-68, Nov. -4,. [4]. Zani, J. Zou, B. Wong, R.V. ruzelecky, Y. Peter, "VOA-base optical ES accelerometer," International Conference on Optical ES an Nanophotonics (ON), pp.5-6, Aug. 8-,. [5] H. Qu, D. Fang, H. Xie, "A onolithic COS-ES -Axis Accelerometer With a Low-Noise, Low-Power Dual-Chopper Amplifier," IEEE Sensors Journal, Vol.8, No.9, pp.5-58, Sept. 8. [6] T.-C. uo, J.-F. iang, "Design of ual-axis accelerometer with COS-ES," International Conference on Applications of Electromagnetism an Stuent Innovation Competition Awars (AEC), pp.89-9, Aug. -,. [7] W. Chen, J. Ding, X. Liu, C. Wang, "Design an system-level simulation of a capacitive ual axis accelerometer," 7 n IEEE International Conference on Nano/icro Engineere an olecular Systems, (NES'7), pp.64-67, Jan. 6-9, 7. ASEE Northeast Section Conference University of assachusetts Lowell Reviewe Paper April 7-8,

9 [8] A. Baschirotto, A. Gola, E. Chiesa, E. Lasalanra, F. Pasolini,. Tronconi, T. Ungaretti, "A ±-g ual-axis linea accelerometer in a stanar.5-μm COS technology for high-sensitivity applications," IEEE Journal of Soli-State Circuits, Vol.8, No.7, pp. 9-97, July. [9] S. Lee, S. im, S.C. Lee, H. o, Y. Park, N- im, an D.D. Cho, Wafer-level Hermetic Package Dual-axis Digital icroaccelerometer, 6 SICE-ICASE International Joint Conference, Oct. 8-, 6, Bexco, Busan, orea, pp [] D. Chattaraj,.B.. Swamy an S. Sen, Design an Analysis of Dual Axis ES Accelerometer, 7 International Workshop on Physics of Semiconuctor Devices (IWPSD 7), pp. 78 7, Dec. 6-, 7. BIOGRAPHIES Zijun He is a master stuent in Department of Electrical an Computer Engineering at University of Brigeport, CT. He obtaine his B.S. egree in Electrical Engineering in Ocean University of China, Shanong, China in 6. His research interests inclue VLSI an ES. Xingguo Xiong is an associate professor in Department of Electrical an Computer Engineering at University of Brigeport, CT. His research interests inclue ES (icroelectromechanical Systems), nanotechnology, as well as VLSI esign an testing. Wei Quan is a master stuent in Department of Electrical an Computer Engineering at University of Brigeport, CT. He obtaine his B.S. egree in Electrical Engineering in Shanong University, China in. His research interests inclue VLSI an ES. ASEE Northeast Section Conference University of assachusetts Lowell Reviewe Paper April 7-8,