Design of a Shock Immune MEMS Acceleration Sensor and Optimization by Genetic Algorithm

Transcription

1 2012, TextRoad Publication ISSN Journal o Basic and Applied Scientiic Research Design o a Shock Immune MEMS Acceleration Sensor and Optimization by Genetic Algorithm Hamid Reza Sabouhi 1, Masoud Baghelani 2 1 Islamic Azad University, Jirot Branch, Iran 2 Department o Electrical Engineering, Sahand University o Technology, Tabriz, Iran ABSTRACT This paper presents a novel capacitive micromachined MEMS acceleration sensor with normal-to-plane shock immunity. The paper takes advantage o suspended cantilevers over the springs o the structure which shorten the spring length or normal movements o the proo mass ater a speciic distance. Thereore, this makes the springs very much stronger which subsequently prevent more normal movements and catastrophic ailures. Also, as the one o the most important parameters o the on-chip devices, which determines the abrication costs o the devices, consumed area is optimized in this work by genetic algorithm. Modal and several structural analyses are provided in this paper which result very good characteristics or the proposed sensor. KEYWORD: MEMS, acceleration sensor, genetic algorithm, natural requency, shock immunity, surace micromachining, and capacitive sensor 1. INTRODUCTION The state-o-the art navigation systems require high perormance acceleration sensors with high shock immunity. Also, acceleration sensing is important in lots o other applications rom automotive industries (or airbag control, car suspension control, saety belt), or vibration control o generators and bridges, robotics, geosciences, cell-phones, medical devices, hard disk protection o computers, pacemaker control, security devices, headlight leveling, joy sticks, OCR systems and etc [1]. Various methods have been presented or acceleration sensing with dierent unctionalities, where among them Micro-Electro-Mechanical System (MEMS) technology achieves great deal o interest [2]. Due to their low production cost and uniorm production according to batch processing, CMOS compatibility, small size, high sensitivity and resolution, MEMS accelerometers have been successully commercialized and received the largest MEMS market [3]. Thereore, improving their perormance may directly aect their commercial status and open new application(s) or them. One o the most important problems associated with MEMS accelerometers is their low shock immunity [2] or preventing the device damages. This paper presents a method or protection o the device against unwanted normalto-plane accelerations. Another important parameter o all integrated on-chip devices is consumed die area which represents the number o devices on a speciic waer and hence predetermines the cost o the device. For this reason, this paper employs genetic algorithm or minimization o the required die area while considering practical issues o the accelerometer. Many works have been presented in literature or MEMS accelerometers which are mostly include: surace micromachined capacitive accelerometers [4-6], thermal accelerometers employing thermal convection [7-8], resonant accelerometers [9-10], SOI capacitive accelerometers which employ silicon-on-insulator technique [11], piezoelectric based MEMS accelerometers [12-13] and etc. This paper employs surace micromachined capacitive technique due to its high perormances as described in next sections or achieving the best unctionality. The paper is organized as ollows; ater an introduction in section1, design parameters o the MEMS accelerometer are described in section 2. Die area optimization by genetic algorithm is presented in section 3. The proposed method or normal-to-plane shock immunity o the device is discussed in section 4. Proposed abrication processes are detailed in section 5. Analysis and simulation results are achieved in section 6 ollowed by a conclusion in section 7. *Corresponding Author: Hamid Reza Sabouhi, Islamic Azad University, Jirot Branch, Iran 10480

2 Sabouhi and Baghelani, DESIGN PARAMETERS OF THE MEMS ACCELEROMETER There are several abrication methods or MEMS accelerometers such as surace and bulk micromachining, which each has its own advantages and drawbacks. Also several sensing methods such as capacitive, piezoelectric, resonant and thermal methods have been presented. Among them, capacitive surace micromachining technique is employed in this work due to its high sensitivity, CMOS compatibility, low noise and low temperature dependency [14-15]. Figure 1. Schematic view o a typical MEMS capacitive accelerometer Fig.1 illustrates a schematic view o a common capacitive surace micromachined MEMS accelerometer or explains its unctionality. The structure is a proo mass suspended above the substrate through 4 springs. Based on mass-spring-damper model, the output displacement versus input acceleration has a second order transer unction given by: 1 2 r 2 X s A s s s Q Where ω r is the angular resonance requency o the structure, Q is the quality actor o the sensor and is depends on various parameters such as abrication deiciencies, squeeze ilm damping, mode shape and etc. r (1) Figure 2. Form o the spring o the sensor At very low requencies (ω<<ω r ), the transer unction could be simpliied as: X s 1 (2) A s 2 r Since, changes in displacement (X) as the result o acceleration (A) is the sensitivity o the device, it can be seen rom equation (2) that, the sensitivity increases by decreasing the resonance requency with a parabolic 10481

3 relation.like any other mechanical mass-spring-damping system, the resonance requency o the structure could be calculated by: k r (3) m I the applied acceleration has the same requency as ω r, the structure starts to oscillate and ails to measure the acceleration. Thereore, ω r needs to be as high as possible or preventing this ailure.hence, there is a trade-o between bandwidth (the spectrum o measurable incoming acceleration) and sensitivity. For overcoming this problem, the sensing element must be critically damped, i.e. b 2 m r (4) Where b is the damping coeicient. For the ease o the abrication process, there must be some holes in the mass o the accelerometer to be discussed in section 5. Thereore the mass o the structure is: m ab 16 N EH Nlw h (5) Where N EH is the number o 4μm by 4μm etch holes, ρ is the density o the structural material, h is the thickness o the structure and the other parameters are shown in Fig.1. Fig.2 shows the spring o the structure. Since 4 springs are employed or suspension o the structure, they could be considered as 4 parallel springs and thereore, the overall spring constant o the accelerometer is: 3 w kx 2 Eh (6) l Where E is the Young modulus o the structural material (here polysilicon), h is the thickness o the structure and the other parameters are shown in Fig.2. Another parameter to be concerned is noise equivalent acceleration given by [1] : 4kBTrBW anoise (7) Qm Where k B is the Boltzmann constant, T is the absolute temperature and BW<ω r /2π, is the required bandwidth. Another important parameter which determines the resolution o the sensor is overall electrodes to sensor capacitance, given by: l h Ct 2 0N (8) d Where N is the number o these small capacitances and other parameters are deined in Fig.1. Fig.3 shows a typical transer unction o the structure and the critical damped region and operating point o the sensor which could be included in succeeding steps. Figure 3. Normalized amplitude versus requency o the system o equation (1); Q=0.5 is related to critically damped system which is the ideal working point or MEMS accelerometers. Since the most important source o damping in this kind o resonators is the air damping between structure and electrode ingers, its equation may be useul or our design, 10482

4 Sabouhi and Baghelani, h b 7.2 l (9) d Where µ is the penetration coeicient in air. Dimension optimization o the sensor is discussed in section DIMENSION OPTIMIZATION BY GENETIC ALGORITHM As mentioned beore, consumed die area directly determines the cost o the device. Thereore, its minimization must be concerned as an important problem. But, there is a trade-o between area minimization, resonance requency and thereore sensitivity. Since, choosing small area means small mass and also shorter and stronger springs which, according to equations (3) and (6), causes increasing the resonance requency and subsequently, based on equation (2), decreasing the sensitivity. Hence, an optimization is required here. Genetic algorithm is a stochastic searching technique to ind an approximate solution or optimization in computer science. It is a sophisticated method or solving minimization problems like this work. It s a special kind o evolutionary algorithms employing biological techniques such as inheritance and mutation. Since genetic algorithm is well-known and several standard books such as [16-17] have been published about, or the sake o abbreviation, we skip describing it here and directly describe utilized inequalities and constraints. The optimization goal (itness unction o the genetic algorithm) is minimizing the die area given by: area b 2l a 10 w (10) Table1. Design summary o the proposed sensor as the result o optimization by genetic algorithm Parameter Value Parameter Value Parameter Value a 308μm w 0.5μm N 75 b 115.2μm h 5.7μm NEH 140 l 44μm l 96μm Area 61976μm 2 E 160GPa ρ 2320Kg/m 3 C t 330F While considering below inequalities and constraints: b 1 m l l 5m 2 0.5m w 5m 2m h 10m 100F C t 10m b 500m 10 m a 600m 1 m l 100 m (11) w 2m d 1m a b r a 3N m r r, n Where ω r,n is the normal-to-plane natural angular requency to be ormulated in next section. Hence the variables are; a, b, l,h and w.results o the genetic algorithm optimization are summarized in table1. 4. NORMAL-TO-PLANE SHOCK IMMUNITY METHOD Although previously mentioned very good speciications, capacitive surace micromachined MEMS accelerometers suer rom high sensitivity to out-o-plane accelerations and shocks. This is as the result o low 10483

5 thickness o the structure and its springs. Hence, normal-to-plane shocks can cause catastrophic ailure and damage o the device. This paper proposes a method or overcoming this problem and protecting o the device against normal-toplane shocks and high accelerations. The idea is to put suspended cantileversover the connection points o springs (seefig.4). Figure 4. Isometric view o the structure When the proo mass moves in normal-to-plane direction (due to normal-to-plane acceleration), the spring constant o the structure is relatively low and permits the movement easily. But, as the proo mass comes up, the second part o the spring comes up through with it since touches the cantilever. Now, the contact point acts as a virtual anchor. Hence, the length o the spring becomes very small. This causes a very much stronger spring which won t permit large delections or movements o the proo mass and subsequently prevents the catastrophic ailure. It is useul to say that, the spring constant o the structure beore achieving to the cantilever is: 3 h k z 2Ew 12 l Because o the mode shape and by considering the design parameters rom table 1, this value is about 130 times greater than in-plane vibration mode and thereore, the resonance requency o the structure in this mode expected to be more than 11 times greater than that o in-plane mode. It is obvious that, this high resonance requency cause more immunity to normal shocks itsel. The spring constant o the structure ater achieving to the cantilever becomes much larger and is equal with: 3 ' h k z 4 Ew (13) l ' Where l is indicated in Fig.4 and is very smaller than l.for this design, the spring constant ater achieving to the cantilever becomes more than times greater than the spring constant beore achieving to the cantilever. This very high stiness puts the resonance requency rom this point orward to ar beyond the ordinary applications requencies around 4MHz and hence solves the out-o-plane shock problem. 5. FABRICATION PROCESS Surace micromachined capacitive Micro-Electro-Mechanical Systems abrication process is ully CMOS compatible and hence is an inexpensive technology which takes advantages o ully integration o CMOS transistors read-out circuitry alongside the MEMS sensors. Fabrication process is outlined in Fig.5 and is include: deposition o a thin adhesive silicon-nitride layer over the silicon substrate. A very thin Cr/Au layer to orming the interconnections is then deposited and patterned. Ater that, a 1.5μm sacriicial oxide layer is deposited by PECVD and patterned or anchor location. Subsequently, a 10484

6 Sabouhi and Baghelani, μm polysilicon structural material is deposited. Second sacriicial silicon oxide layer is deposited and patterned to open the cantilevers locations. Ater deposition and patterning o cantilevers, gaps between ingers o the sensors and electrode and also etching holes are opened by Deep Reactive Ion Etching (DRIE). Ater all deposition and patterning processes, the whole structure is released to hydroluoric acid to etch all the sacriicial oxide layers and the inal cross-section shown in Fig.5 is achieved. Figure 5. Fabrication process o the proposed structure 6. SIMULATION RESULTS This section provides several important analysis o the proposed MEMS acceleration sensor. Modal analysis o the structure results natural requencies o the device both in-plane and normal-to-plane directions as shown in Figs. 6 and 7. It can be seen that, the in-plane natural requency is 3300Hz and normal-to-plane natural requency is 34500Hz which has good consistency with discussions in section 4. Figure 6. In-plane resonance mode o the structure; note the bending direction o the springs Figure 7. Normal-to-planeresonance mode o the structure; note the bending direction o the springs 10485

7 Von-Misses stresses are the most important stresses over the structure during its work. They determine the stability o the device. As shown in Fig.8, Von-Misses stress at the in-plane mode o operation or 20g applied acceleration is 0.79Gpa which is 0.49% o the Young modulus o the structural material and thereore has no signiicant eect over the structure. Figure 8. Von-Misses stress o the structure or in-plane acceleration at the springs attachment points (maximum stress points) Normal displacement o the structure as the result o normal-to-plane accelerations or shocks, causes Von- Misses stresses maximized at the attachment locations o the springs to both the proo mass and anchor. For applied accelerations as high as 36g the Von-Misses proile is look like Fig.9. Figure 9. Von-Misses stress o the structure or normal-to-plane acceleration at the springs attachment points (maximum stress points) Normal-to-plane displacement proile o the structure with cantilever contact is shown in Fig.10. In this condition, the Von-Misses stresses proile o the spring is shown in Fig.11 and Von-Misses stresses o the cantilever in Fig.12. Figure 10. Vector plot o the displacement o the structure with cantilevers as the result o normal-to-plane acceleration 10486

8 Sabouhi and Baghelani, 2012 Figure 11. Von-Misses stresses proile or normal-to-plane stresses over the spring Figure 12. Von-Misses stresses proile or normal-to-plane stresses over the cantilever The overall data o the designed structure are summarized in table.2. The proposed MEMS surace micromachined capacitive sensor has an optimized area and is very immune to normal-to-plane shocks which is the most important problem associated with this kind o MEMS accelerometer. Table2. Summary o characteristics o the proposed MEMS accelerometer Parameter Value Parameter Value m 614.1ng k 0.257N/m k z 33.49N/m k' z 2.194MN/m ω r(theory) 20457rad/sec ω r(fea) rad/sec ω r,n(theory) 233.5Krad/sec ω r,n(fea) 212.2Krad/sec Sensitivity 7.6nm/g Maximum detectable acceleration 40g Minimum detectable acceleration 0.02g Maximum tolerable normal acceleration without 540g cantilevers Maximum tolerable normal acceleration with cantilevers 2410g Acceleration noise 0.509mg 10487

9 7. CONCLUSION A novel method or improving shock immunity in capacitive surace micromachined MEMS accelerometers have been introduced by employing suspended cantilevers over the structure. The cantilevers were so designed to act as secondary anchors or springs and null a signiicant part o their lengths. Hence, as springs move upward as the result o normal-to-plane accelerations or shocks, ater achieving springs to cantilevers, the spring constant o the structure became thousands times larger which prevented catastrophic ailures. Also, the very important die area o the sensor has been optimized by the genetic algorithm to decreasing the abrication costs o the device. Our researches on perormance improvement o MEMS accelerometers are on-going. REFERENCES 1. Malu N. and Williams K., 2004, An introduction to microelectromechanical systems engineering. Artech house Boston, pp: Beeby S., Ensell G., Krat M. and White N., 2004, MEMS mechanical sensors. Artech house Boston, pp: Adrian, P., Sensor Business, Marketing and Technology Developments. Sensor Business Digest (News Letter), November. 4. Bruschi P., A. Nannini, D. Paci and F. Pieri, A method or cross-sensitivity and pull-in voltage measurement o MEMS two-axis accelerometers. Sensors and Actuators A, 123,: Zhou F. X., L. Che, B. Xiong, X. Li, J. Wu and Y. Wang, A novel capacitive accelerometer with a highly symmetrical double-sided beam-mass structure. Sensors and Actuators A, 179,: Dong Y., M. Krat and W. R. White, Higher order noise-shaping ilters or high-perormance micromachined accelerometers. IEEE Transactions on instrumentation and measurement, 56(5): Garraud A., A. Giani, P. Combette, B. Charlot and M. Richard, A dual axis CMOS micromachined convective thermal accelerometer. Sensors and Actuators A, 170,: Liao K.-M., R. Chen and B. C.S. Chou, A novel thermal-bubble-based micromachined accelerometer. Sensors and Actuators A, 130,: Su X. P. S., H. S. Yang and A. M. Agogino, A resonant accelerometer with two-stage microleverage mechanisms abricated by SOI-MEMS technology. IEEE Sensors journal, 5(6),: Tocchino A., A. Caspani and G. Langelder, Mechanical and electronic amplitude limiting techniques in a MEMS resonant accelerometer. IEEE Sensors journal, 12(6),: Amini B. V., and F. Ayazi, A 2.5V 14 bit ΣΔ CMOS-SOI capacitive accelerometer. IEEE journal o solidstate circuits, 39(12),: Yu J.-C., C. Lee, W. Kuo, and C. Chang, Modeling analysis o a triaxialmicroaccelerometer with piezoelectric thin-ilm sensing using energy method. Microsystem Technologies, 17(2),: Sankar A. R., J. G. Jency, and S. Das, Design, Fabrication and testing o a high perormance silicon piezoresistive Z-axis accelerometer with proo mass-edge-aligned-lexures. Microsystem Technologies, 18(1),: Abdolvand R., and B.V. Amini, Sub-micro-gravity in-plane accelerometers with reduced capacitive gaps and extra seismic mass. IEEE/ASME Journal o Microelectromechanical Systems, 16(5),: Ki-Ho H., and C. Young-Ho, Sel-balanced navigation-grade capacitive microaccelerometers using branched inger electrodes and their perormance or varying sense voltage and pressure. IEEE/ASME Journal o Microelectromechanical Systems, 12(1),: Edward Goldberg D., 1989, Genetic algorithms in search, optimization, and machine learning. Addison-Wesley publishers. 17. Man K. F., Tang K. S., and Kwong S., 2001, Genetic algorithms. Springer Verlag