- Email: [email protected]
A spectral vibration detection system based on tunable micromechanical resonators
Sensors and Actuators A 123–124 (2005) 63–72
A spectral vibration detection system based on tunable micromechanical resonators D. Scheibner a,∗ , J. Mehner c , D. Reuter b , T. Gessner b , W. D¨otzel b b
a SIEMENS AG, A&D ATS2, Nuremberg, Germany Chemnitz University of Technology, Chemnitz, Germany c Fhg IZM, Berlin, Germany
Received 13 September 2004; received in revised form 7 March 2005; accepted 17 March 2005 Available online 25 April 2005
Abstract We present a frequency selective vibration detection system based on tunable micromechanical resonators for the frequency range 1–10 kHz. The Single Crystal Reactive Etching and Metallization (SCREAM)-fabricated sensor structure consists of an array of eight resonators with stepped base frequencies between 2.8 and 10.2 kHz. The resonators provide a resonance frequency tuning capability by electrostatic forces to set the detection frequency to the desired value. Experimental results of the fabricated resonator arrays are presented. A maximum tuning voltage of 35 V is required for continuous resonance frequency tuning from 1 to 10 kHz. The amplitude range of the resonators amounts to 10 m. Over the whole amplitude range, a maximum resonance frequency shift of 0.7% was measured. An increase of ambient temperature from 30 to 150 ◦ C results in a small resonance shift of 0.5% and to an increase of resonance peak bandwidth of 23%. The sensor structure is part of a vibration detection system, which is completed with analogue and digital circuitry. Test measurements with the system demonstrate its functionality and applicability in an industrial environment. © 2005 Elsevier B.V. All rights reserved. Keywords: Resonant; Tuning; SCREAM; Vibration
1. Introduction Vibration monitoring has become an important means for wear state recognition of industrial machinery such as cutting tools, bearings, gears, pumps or engines [1,2]. The majority of mechanical vibration used to identify the wear state is found in the frequency range from a few hertz to 10 kHz. Currently, piezoelectric wide-band transducers combined with signal analysers are usually used to obtain the spectrum. The equipment is expensive and requires trained operators. Therefore, permanent monitoring is limited to extremely expensive machinery or safety related applications. For the characterization of the wear state, the observation of a few spectral lines is normally sufficient. This fact sug∗ Corresponding author. Tel.: +49 911 895 3020; fax: +49 911 895 3762. E-mail address: [email protected] (D. Scheibner).
0924-4247/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.sna.2005.03.034
gests a narrow-band resonance operation of the sensor structure. It amplifies the vibration signal in a small band around its resonance frequency and eliminates other spectral ranges. Advantages of this frequency selective approach are the improvement of the signal-to-noise ratio and simplifications in the signal conditioning circuitry without Fourier transformation. This concept has already been implemented with MEMS technology [3–5]. The fixed resonance frequency of the described sensors limits their employment in applications with well-known and constant measurement frequencies. To overcome this restriction, resonance tuning mechanisms are used. Several publications describe resonators using stressstiffening [6,7] or electrostatic-softening [8–10] effects to tune the resonance frequency of micromechanical resonators by electrostatic forces. In this paper, we present a vibration measurement system based on micromechanical tunable resonators. As sensor structure, a single crystal reactive etching and metallization
D. Scheibner et al. / Sensors and Actuators A 123–124 (2005) 63–72
64
Fig. 2. Resonance tuning by electrostatic-softening. Fig. 1. Principle of frequency selective vibration detection.
(SCREAM)-fabricated array of tunable resonators is used. Resonance tuning is achieved by electrostatic-softening. The sensor array is completed with analogue signal detection and digital control circuitry in a compact measurement system with fully digital interface. Experimental characterizations of the sensor structure and measurement results of the complete system are presented.
2. Frequency selective principle Vibration sensors usually work as wide-band transducers far below their resonance frequency. In this range, the sensor output is proportional to the acceleration acting on the sensor. For the characterization of the wear state, usually the observation of only a few spectral lines is sufficient. This fact suggests a narrow-band resonance operation of the sensor structure. A small band of the incoming spectrum determined by the bandwidth of the resonance peak is amplified at the resonance frequency of the sensor structure (Fig. 1). Other spectral ranges are suppressed. Advantages of this frequency selective principle are the direct extraction of spectral information without Fourier transformation and the improvement of the signal-to-noise ratio by the quality factor. Furthermore, such a sensor is insensitive to large interfering signals, which would overdrive wide-band transducers.
3. Resonance frequency tuning Sensors working in a frequency selective mode are limited to one-single measurement frequency, i.e. its resonance frequency. To extend the measurement range of such a sensor, resonance frequency tuning is used. The structures presented in this paper implement electrostatic-softening to vary the resonance frequency. The principle is based on electrostatically generated, amplitude-dependent forces acting on the seismic mass (Fig. 2). Therefore, the resonator structure contains an extra electrode system for resonance tuning at which a dc voltage, the tuning voltage Vtun is applied.
The total stiffness ktotal of the tuned resonator is calculated as follows: ktotal = k0 −
2 d2 C(x) Vtun 2 dx2
(1)
where Vtun is the tuning voltage, k0 the mechanical stiffness and C is the total capacitance between stator and seismic mass. The forces lead to a softening of the system and therewith to a lowering of the resonance frequency. In this way, the resonance frequency and thus the sensitive frequency band are set by the tuning voltage. To achieve linear sensor characteristics, it is essential that the second derivative of the capacitance function d2 C(x)/dx2 is constant over the amplitude range. Otherwise the resonance frequency depends on the signal amplitude. Such non-linearities implicate large amplitude errors. A quadratic capacitance function implemented by a comb system of linearly varied finger length using poly-Si technology is described in ref. [8]. A SCREAM-technology compatible approach using non-overlapping comb fingers is presented in ref. [10]. 4. Sensor structure 4.1. Fabrication A near-surface silicon bulk technology represents the technological basis for the sensor structures. The SCREAM technology has been previously described in several publications [11,12]. In a single-mask process laterally movable grid and beam structures with a cross-section as sketched in Fig. 3 are fabricated. The aspect ratio typically amounts to about 2–20 m. Top metal layers insulated from the bulk substrate by a SiO2 layer are used as electrodes for electrostatic force generation and capacitive signal detection. Individual electrodes are separated by trenches. Due to the reactive ion etching (RIE) lag, SCREAM-fabricated structures are characterized by a sensitivity of the produced trench depth to the local width of the mask gap. Overlapping comb structures implicate different trench widths in the overlapping and nonoverlapping areas. Since this leads to a varying cross-section of the beams, comb designs should use non-overlapping fingers.
D. Scheibner et al. / Sensors and Actuators A 123–124 (2005) 63–72
65
Fig. 3. Cross-section of a SCREAM-fabricated structure. (a) Schematic and (b) SEM-view.
4.2. Layout
Table 1 Design parameters of the sensor array
The sensor structures consist of a laterally movable mass supported by four-folded flexures (Fig. 4). The thickness of the mass is 20 m. Two-different comb systems at the seismic mass are designed either for capacitive signal detection (const. dC(x)/dx) or for resonance frequency tuning (const. d2 C(x)/dx2 ). The gap spacing between opposite fingers amounts to 2 m. The maximum resonance amplitude is limited to 10 m. The tuning range of a single tunable resonator is limited by the maximum tuning voltage. By grouping eight resonators with stepped base frequencies and overlapping tuning ranges into an array, the frequency range is widely extended. Lateral dimensions of the seismic mass range from 630 m × 660 m (cell 5) to 710 m × 840 m (cell 1). Flexures are between 146 m (cell 8) and 301 m (cell 1) long. The total size of the sensor chip is 7 mm × 10 mm (Fig. 5). Table 1 gives an overview of the design parameters.
Cell 1 2 3 4 5 6 7 8
m (g)
k0 (N m−1 )
f0 (Hz)
dC(x)/dx (F m−1 )
d2 C(x)/dx2 (F m−2 )
6.71 6.69 4.14 4.13 4.31 4.30 4.30 4.29
2.09 3.85 4.71 7.00 9.86 12.41 14.97 17.59
2811 3820 5369 6553 7611 8550 9396 10186
2.4e−3 2.4e−3 3.2e−3 3.2e−3 3.6e−3 3.6e−3 3.6e−3 3.6e−3
2.65e−8 2.65e−8 1.33e−8 1.33e−8 0.884e−8 0.884e−8 0.884e−8 0.884e−8
4.3. Experimental characterization 4.3.1. Measurement setup The sensor structures were experimentally characterized with regard to resonance frequency and bandwidth, extracted from the frequency response function. The frequency response of the sensor structures was obtained by sweep measurements using a lock-in-amplifier EG&G 7265. The
Fig. 4. Tunable microresonator: (a) schematic and (b) SEM view.
D. Scheibner et al. / Sensors and Actuators A 123–124 (2005) 63–72
66
Table 2 Measured parameters of the sensor array (11-113-7 Chip 36), f0tuned at Vtun = 35 V Cell
f0 (Hz)
1 2 3 4 5 6 7 8
2888 3957 5473 6605 7749 8656 9528 10271
f0tuned (Hz) 901 2841 3741 5277 6545 7578 8565 9385
Quality factor at f0
Bandwidth (Hz)
14.1 19.3 21.1 25.4 32.8 37.3 40.8 43.7
202 204 259 260 236 232 228 226
amplifier generates the driving signal and extracts the according frequency component out of the sensor response. Structures were excited mechanically by an electrodynamic shaker, the amplitude of which was controlled by laser vibrometer reference measurement. Vibration signal was detected capacitively by I/V (current-to-voltage) converters as intended in the later application (cf. Section 5.1), whereas temperature characteristics were analysed in a heat chamber using electrostatic actuation as described in ref. [13]. 4.3.2. Resonance characteristics Table 2 gives an overview of the measured characteristics. Due to technological tolerances of the fabrication process the untuned resonance frequencies of the characterized arrays (Wafer 11-113-7) are up to 3% higher than expected (Fig. 6). The tolerances arise from lithography (resist thickness, exposure time), etch (depth, profile) and coating (thickness) steps affecting geometric values such as structure height, width and beam profile. For the bandwidth of the resonance peak values between 200 and 260 Hz were measured. This is small enough to demonstrate the working principle, but for practical frequency selective measurements an electronic postfiltering with a smaller bandwidth (e.g. 50 Hz) is necessary (cf. Section 5.2). An essential feature of the sensor array is the overlapping of the tuning ranges. Experimental characterization shows that the desired tuning range is covered by a maximum tuning voltage of 35 V (Fig. 7). Therewith the fabricated array allows
Fig. 5. Photo of the sensor array.
Fig. 6. Resonance characteristics of untuned array cells (measured data).
continuous measurements in the frequency range from 1 to 10 kHz as required for the application in the measurement system. Some wafers, e.g. (11-113-2), which was used for temperature tests, showed higher resonance frequencies (up to +11% deviation from theoretical value). The reason was an exposure problem in the lithography step leading to wider structures, and thus stiffer flexures. Nevertheless, these arrays allowed continuous tuning down to 1 kHz. 4.3.3. Non-linearities The resonator array is used as a frequency selective vibration sensor. Therefore, it has to satisfy high requirements with regard to the linearity of the output signal. In theory, the bandwidth of the resonance peak is only dependent on damping effects and not on the actual resonance frequency. Therefore, the bandwidth is in contrast to the quality factor not influenced by the resonance tuning. This is an
Fig. 7. Resonance tuning of the array (measured data).
D. Scheibner et al. / Sensors and Actuators A 123–124 (2005) 63–72
67
Fig. 8. Stability of bandwidth over tuning range, measured data of cell 2.
important condition to correctly recalculate to the incoming acceleration by a frequency-independent transfer factor. Experiments confirm that the bandwidth of the resonance peak stays constant over the tuning range of the resonators (Fig. 8). The comb systems at the seismic mass generate electrostatic forces, which influence the resonance frequency. The impact on the sensor characteristics was analysed by monitoring the resonance frequency at different amplitude levels. Signal detection is carried out capacitively by the detection of reload currents using I/V converters. This requires a bias voltage Vbias at the detection combs (cf. Section 5.1). The detection comb system is designed for dC(x)/dx = const. and thus d2 C(x)/dx2 = 0. In practice, remaining non-linearities of the capacitance function were experimentally verified. This implicates a non-zero d2 C(x)/dx2 and leads to nonlinear amplitude-dependent forces. As a result, the resonance
Fig. 9. Impact of bias voltage on linearity, experimental data at tuning voltage Vtun = 0 V.
Fig. 10. Impact of bias voltage on linearity, experimental data at tuning voltage Vtun = 35.5 V.
frequency shifts to higher values with rising sensor amplitude. This undesirable tuning effect increases with rising bias voltage (Fig. 9). The tuning comb system is optimized for d2 C(x)/dx2 = const., but experimental characterization also revealed a non-ideal capacitance function. Measurements of transfer functions with maximum tuning voltage at different amplitude levels show the superimposition of non-linear effects of detection and tuning combs. In this case, the resonance frequency shifts to lower values with rising sensor amplitude (Fig. 10). This effect decreases with rising bias voltage. An optimum value for the bias voltage is found at Vbias = 4 V. The maximum resonance frequency shift over the amplitude range now amounts to 0.68% (Fig. 11).
Fig. 11. Frequency response of an array cell at different excitation amplitudes (measured data at Vbias = 4 V).
D. Scheibner et al. / Sensors and Actuators A 123–124 (2005) 63–72
68
Table 3 Impact of temperature on resonance frequency (Wafer 11-113-2 Chip 5A) Cell 1 3 8
fres at 30 ◦ C (Hz) 3121.2 5894.6 11088
fres at 150 ◦ C (Hz)
fres (%) −0.5 −0.42 −0.3
3105.2 5870 11055
Table 4 Impact of temperature on bandwidth (Wafer 11-113-2 Chip 5A) Cell
BB3dB at 30 ◦ C (Hz)
BB3dB at 150 ◦ C (Hz)
BB3dB (%)
1 3 8
223.1 280.0 324
270.7 345.0 389
21.3 23.2 20.1
4.3.4. Temperature behaviour SCREAM-fabricated structures consist of a stack of materials with different temperature coefficients. This mismatch results in temperature-dependent behaviour. Measurements in the temperature range from 30 to 150 ◦ C show a small resonance frequency shift of maximum 0.5% to lower values (Table 3) and a large increase of bandwidth of maximum 23% (Table 4). The folded flexures used in the resonator design release thermal tension, and are therefore relatively insensitive to temperature change. But the layers are deposited only on top and on the sidewalls of the silicon substrate (cf. Section 4.1). Thermal mismatch leads to tensions on the top side and bends down the free end of the flexure. This stress is released at higher temperatures. Consequently, stiffness and resonance frequency decrease. The increase of resonance peak bandwidth is probably caused by a lowering of the seismic mass. At room temperature, the mass is lifted out of the wafer surface due to thermal tension. The relaxation of the structure with rising temperature reduces the gap under the structure and slide-film damping effects increase. Therewith the bandwidth of the resonance peak increases. For the application as vibration sensor in a wide-temperature range, the large change of bandwidth has to be considered by in-system temperature calibration.
Fig. 12. Capacitive signal detection using I/V converters.
Fig. 13. Phase selective demodulation circuit.
Fig. 14. Schematic overview of the measurement system.
D. Scheibner et al. / Sensors and Actuators A 123–124 (2005) 63–72
69
5. Signal detection and evaluation 5.1. Capacitive signal detection The capacitive signal detection circuitry is based on the amplification of reload currents by I/V converters. The seismic mass is set to a bias voltage of 4 V. Stator combs are virtually connected to ground via the I/V converter stages. The I/V converters use a feedback resistor of R = 6.8 M. A parallel capacitance of Cp = 1 pF limits the bandwidth to a maximum frequency fmax = (2RCp )−1 and ensures stable operation. The signals of the two-stator combs are combined in a differential amplifier. The stator combs of all array cells are parallel connected (Fig. 12). The active cell is selected by the bias voltage at the seismic mass. All other cells are deactivated by connecting the seismic mass to ground. This concept allows the application of only one detection circuit for the complete array. 5.2. Phase selective demodulation The mechanical bandwidth of the sensor structures amounts to 200–260 Hz. This value is too large for spectral evaluation in the kilohertz range. Therefore, an electronic post-filtering is necessary. The measurement system uses a phase selective demodulation circuit working according to the lock-in principle. The demodulation is carried out by multiplying the raw signal with a reference signal. The reference frequency in the raw signal is rectified to a dc component while all other frequencies remain ac signals. A subsequent low-pass removes the ac signals. The corner frequency of this low-pass – in our case 25 Hz – sets the bandwidth of the demodulation. This process is performed in two-parallel paths with reference signals 90◦ out of phase. As a result, real and imaginary part of the raw signal at the reference frequency are extracted. In a last step, the RMS value is calculated from real and imaginary part. The implemented circuit uses two D-flip-flops to generate the two reference frequencies 90◦
Fig. 15. Photo of the frequency selective measurement system.
Fig. 16. Laboratory test measurement: (a) driving signal with large lowfrequency component, (b) sensor output, time signal and (c) spectrum obtained by frequency sweeping.
70
D. Scheibner et al. / Sensors and Actuators A 123–124 (2005) 63–72
out of phase and AD 630-demodulators (Analog Devices) for multiplication (Fig. 13).
6. Measurement system 6.1. Setup The sensor array has a complex interface. For practical measurements, the appropriate cell and tuning voltage have to be chosen and the result has to be recalculated. This effort is kept away from the user and encapsulated in a measurement system with a fully digital interface. The user sends a measurement command for a certain frequency. Now the system sets the sensor array to this frequency using internal calibration data, evaluates the sensor output and sends back the result (RMS value of acceleration at desired frequency). The measurement system contains the micromechanical structure, analog circuitry (I/V converter, phase selective
Fig. 17. Dynamic test using a 5 Hz modulated 3.9 kHz driving signal: (a) sensor ac output and (b) demodulated measurement result.
demodulation circuit, multiplexer for cell selection, amplifier for tuning voltage), digital control (microcontroller ATmega16) and interface (RS232) (Fig. 14). For low backlash to the object to be measured, the mass of the connected part has to be small. Therefore, the system is split into a small sensor head containing the micromechanics chip and the first analog detection stage and a larger control unit (Fig. 15). 6.2. Measurements For laboratory tests, the sensor head was mounted on an electrodynamical shaker. The driving signal contained several frequency lines in the measurement range of the system and a large low-frequency (75 Hz) component (Fig. 16a). The low-frequency component is far away from the resonance range of the sensor array, and thus filtered out mechanically. In the time signal of the sensor output the 75 Hz is no longer identifiable (Fig. 16b). By sweeping the resonance frequency, the spectrum was obtained and all higher frequencies contained in the driving signal were detected (Fig. 16c). Therewith the insensitivity to large interfering signals as an advantage of frequency selective sensors to wide-band transducers is practically demonstrated. In a second experiment the dynamic behaviour of the system was tested. A 3.9 kHz carrier modulated with 5 Hz was used as driving signal. The system continuously measured at 3.9 kHz. Results show, that the system including digital communication follows the 5 Hz modulation with 35 samples/s (Fig. 17). The sample rate is mainly limited by the settling
Fig. 18. Sensor mounted on the cylinder head of a combustion engine (MZ RT 125).
D. Scheibner et al. / Sensors and Actuators A 123–124 (2005) 63–72
time of the sensor structure and the phase selective demodulation circuitry (cf. Section 5.2). After verifying the functionality in laboratory experiments, the system was used in several practice tests, for instance at a combustion engine. In this measurement, the sensor was mounted on the cylinder head of the engine to detect knocking combustion (Fig. 18). Knocking is uncontrolled combustion, caused by the spontaneous ignition of portions of the mixture, which have not yet been reached by the advancing flame front, triggered by the ignition spark. The consequence of knocking is overheating of parts of the engine, and thus destruction of the engine. Therefore, knocking detection is an important part of current ignition timing systems. On an engine test stand normal and knocking combustion were setup and the vibration spectrums for both cases were obtained with the frequency selective measurement system (Fig. 19). If knocking occurs obvious changes in the spectrum are clearly visible. The overall spectrum increases
71
in amplitude. At knocking combustion, the amplitude of the clear peak found at 5 kHz approximately doubles. The frequency selective vibration sensor can be set to an expressive frequency (e.g. 5 kHz) to permanently monitor the vibration amplitude. Then – in case of exceeding a critical level – the system gives an alert indicating knocking combustion. By these tests the applicability of the system for measurements in industrial environment was demonstrated.
7. Conclusions SCREAM-fabricated resonators are suitable for frequency selective vibration measurements. Electrostatic-softening turned out to be an appropriate means for resonance frequency tuning in the range of 1–10 kHz. A measurement system containing the sensor array, analogue and digital circuitry allows measurements in industrial environment and demonstrates the advantages of the frequency selective approach in practice. Further work concentrates on an optimization of the sensor structure. The aim of a new design is improved linearity. This will be achieved by redesigned comb structures with better capacitance characteristics (dC(x)/dx = const. for signal detection, d2 C(x)/dx2 = const. for resonance tuning). Besides the further improvement of the frequency selective vibration sensor also other applications for such a measurement system such as tool wear at cutting tools and lathes or cavitation sensing at pumps have to be investigated.
Acknowledgement Parts of the work were done within the Collaborative Research Center SFB 379, which is funded by the German Research Association (DFG).
References
Fig. 19. Vibration spectrum of the engine running at 4000 rpm measured using the frequency selective system: (a) normal combustion and (b) knocking.
[1] H. Bloch, F. Geitner, Practical Machinery Management for Process Plants, Machinery Failure Analysis and Troubleshooting, vol. 2, Gulf Publishing, Houston, 1994. [2] V. Wowk, Machinery Vibration: Measurement and Analysis, McGraw-Hill, New York, 1991. [3] W. Benecke, L. Csepregi, A. Heuberger, K. K¨uhle, H. Seidel, A frequency-selective, piezoresistive silicon vibration sensor, in: Proceedings of TRANSDUCERS, Philadelphia, 1985, pp. 105–109. [4] H. Fritsch, R. Lucklum, T. Iwert, P. Hauptmann, D. Scholz, E. Peiner, A. Schlachetzki, A low frequency micromechanical resonant vibration sensor for wear monitoring, Sens. Actuators A 62 (1997) 616–620. [5] E. Peiner, R. Mikuta, T. Iwert, H. Fritsch, P. Hauptmann, K. Fricke, A. Schlachetzki, Micromachined resonator for cavitation sensing, Sens. Actuators A 76 (1999) 266–272. [6] J. Wibbeler, C. Steiniger, M. K¨uchler, K. Wolf, T. Frank, J. Mehner, W. D¨otzel, T. Gessner, Resonant silicon vibration sensors with voltage controlled frequency tuning capability, in: Proceedings of EUROSENSORS XIII, The Hague, September 12–15, 1999, p. 24B3.
72
D. Scheibner et al. / Sensors and Actuators A 123–124 (2005) 63–72
[7] J. Mehner, D. Scheibner, J. Wibbeler, Silicon vibration sensor array with electrically tunable band selectivity, in: Proceedings of MICRO SYSTEM TECHNOL, D¨usseldorf, March 27–29, 2001, pp. 267–272. [8] K. Lee, Y. Cho, A triangular electrostatic comb array for micromechanical resonant frequency tuning, Sens. Actuators A 70 (1998) 112–117. [9] W. Ye, S. Mukherejee, N.C. MacDonald, Optimal shape design of an electrostatic comb drive in microelectromechanical systems, J. Microelectromech. Syst. 7 (1) (1998) 16–25. [10] D. Scheibner, J. Wibbeler, J. Mehner, B. Br¨amer, T. Gessner, W. D¨otzel, A frequency selective silicon vibration sensor with direct electrostatic stiffness modulation, Analog Integr. Circ. Signal Process. 37 (2003) 35–43. [11] K.A. Shaw, Z.L. Zhang, N.C. MacDonald, SCREAM I: a single mask, single-crystal silicon, reactive ion etching process for microelectromechanical structures, Sens. Actuators A 40 (1994) 63–70. [12] T. Gessner, A. Bertz, C. Steiniger, U. Wollmann, Material and technology approaches of surface micromachining, in: Proceedings of MICRO MAT, Berlin, April 16–18, 1997, pp. 90–95. [13] D. Scheibner, J. Mehner, D. Reuter, U. Kotarsky, T. Gessner, W. D¨otzel, Characterization and self-test of electrostatically tunable resonators for frequency selective vibration measurements, Sens. Actuators A 111 (2004) 93–99.
Biographies Dirk Scheibner is a research engineer at SIEMENS A&D responsible for microsystems. He received the Dipl.-Ing. degree in Microsystem and Precision Engineering from Chemnitz University of Technology in 1999. From 2000–2004 he worked as a Scientific Assistant at the Department of Microsystems and Precision Engineering at the Chemintz University of Technology. He received a Dr.-Ing. degree in electrical engineering and information technology from the same University in 2005. His research interests are design and test of accelerometers and vibration sensors fabricated in silicon surface microtechnology. Jan E. Mehner is a scientific assistant at the Department of Microsystems and Precision Engineering at the Chemnitz University of Technology.
He received a Dr-Ing and Dr-Ing habil degree in electrical engineering and information technology from the same University in 1994 and 1999, respectively. From 1998 to 1999, he was a visiting scientist at the Massachusetts Institute of Technology in the field of macromodelling. His research interests include analytical and numerical methods to design microsystems, CAD-tools and computational algorithms for problems with coupled fields. Danny Reuter is a scientific assistant at the Centre for Microtechnologies at the Chemnitz University of Technology. He received his Dipl-Ing degree in microsystem and precision engineering from Chemnitz University of Technology in 2002. His research interests are silicon surface microtechnology and wafer-level packaging of high aspect ratio MEMS. Thomas Gessner received the Dipl-Phys degree in physics in 1979 and the Dr rer nat degree in 1983 from the Faculty of Natural Science at TU Dresden and the Dr-Ing habil degree in 1989 from University of Technology Karl-Marx-Stadt. Professor Gessner had several positions in the industry (ZMD Dresden) at the development of metallization systems for d-RAM integrated circuits. Since 1991, professor Gessner is director of the Center of Microtechnologies at the Chemnitz University of Technology and has a chair for microtechnology at Chemnitz University of Technology in 1993. In the year 1996, he was appointed as member of the Academy of Science in Saxony. Since 1998, professor Gessner is the head of the Department “Microdevices and Equipment” at Berlin Fraunhofer Institute for Reliability and Microintegration (FhG-IZM) as a branch-lab in Chemnitz. In 1998, he became member of the Scientific Advisory Board of the Federal Republic of Germany. Wolfram D¨otzel received the Dipl-Ing degree in electrical and precision engineering from Technical University Dresden in 1966 and the Dr-Ing degree from the Chemnitz University of Technology in 1971. The research work was focused on the design of electromechanical systems and their reliability. Since 1993, Wolfram D¨otzel is professor for microsystems and precision engineering at the Chemnitz University of Technology. His current work is focused on design and simulation of micromechanical components, their characterization by experimental methods and their application in a macro-system. Professor D¨otzel is member of the Academy of Science of Saxony in Leipzig.
A spectral vibration detection system based on tunable micromechanical resonators D. Scheibner a,∗ , J. Mehner c , D. Reuter b , T. Gessner b , W. D¨otzel b b
a SIEMENS AG, A&D ATS2, Nuremberg, Germany Chemnitz University of Technology, Chemnitz, Germany c Fhg IZM, Berlin, Germany
Received 13 September 2004; received in revised form 7 March 2005; accepted 17 March 2005 Available online 25 April 2005
Abstract We present a frequency selective vibration detection system based on tunable micromechanical resonators for the frequency range 1–10 kHz. The Single Crystal Reactive Etching and Metallization (SCREAM)-fabricated sensor structure consists of an array of eight resonators with stepped base frequencies between 2.8 and 10.2 kHz. The resonators provide a resonance frequency tuning capability by electrostatic forces to set the detection frequency to the desired value. Experimental results of the fabricated resonator arrays are presented. A maximum tuning voltage of 35 V is required for continuous resonance frequency tuning from 1 to 10 kHz. The amplitude range of the resonators amounts to 10 m. Over the whole amplitude range, a maximum resonance frequency shift of 0.7% was measured. An increase of ambient temperature from 30 to 150 ◦ C results in a small resonance shift of 0.5% and to an increase of resonance peak bandwidth of 23%. The sensor structure is part of a vibration detection system, which is completed with analogue and digital circuitry. Test measurements with the system demonstrate its functionality and applicability in an industrial environment. © 2005 Elsevier B.V. All rights reserved. Keywords: Resonant; Tuning; SCREAM; Vibration
1. Introduction Vibration monitoring has become an important means for wear state recognition of industrial machinery such as cutting tools, bearings, gears, pumps or engines [1,2]. The majority of mechanical vibration used to identify the wear state is found in the frequency range from a few hertz to 10 kHz. Currently, piezoelectric wide-band transducers combined with signal analysers are usually used to obtain the spectrum. The equipment is expensive and requires trained operators. Therefore, permanent monitoring is limited to extremely expensive machinery or safety related applications. For the characterization of the wear state, the observation of a few spectral lines is normally sufficient. This fact sug∗ Corresponding author. Tel.: +49 911 895 3020; fax: +49 911 895 3762. E-mail address: [email protected] (D. Scheibner).
0924-4247/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.sna.2005.03.034
gests a narrow-band resonance operation of the sensor structure. It amplifies the vibration signal in a small band around its resonance frequency and eliminates other spectral ranges. Advantages of this frequency selective approach are the improvement of the signal-to-noise ratio and simplifications in the signal conditioning circuitry without Fourier transformation. This concept has already been implemented with MEMS technology [3–5]. The fixed resonance frequency of the described sensors limits their employment in applications with well-known and constant measurement frequencies. To overcome this restriction, resonance tuning mechanisms are used. Several publications describe resonators using stressstiffening [6,7] or electrostatic-softening [8–10] effects to tune the resonance frequency of micromechanical resonators by electrostatic forces. In this paper, we present a vibration measurement system based on micromechanical tunable resonators. As sensor structure, a single crystal reactive etching and metallization
D. Scheibner et al. / Sensors and Actuators A 123–124 (2005) 63–72
64
Fig. 2. Resonance tuning by electrostatic-softening. Fig. 1. Principle of frequency selective vibration detection.
(SCREAM)-fabricated array of tunable resonators is used. Resonance tuning is achieved by electrostatic-softening. The sensor array is completed with analogue signal detection and digital control circuitry in a compact measurement system with fully digital interface. Experimental characterizations of the sensor structure and measurement results of the complete system are presented.
2. Frequency selective principle Vibration sensors usually work as wide-band transducers far below their resonance frequency. In this range, the sensor output is proportional to the acceleration acting on the sensor. For the characterization of the wear state, usually the observation of only a few spectral lines is sufficient. This fact suggests a narrow-band resonance operation of the sensor structure. A small band of the incoming spectrum determined by the bandwidth of the resonance peak is amplified at the resonance frequency of the sensor structure (Fig. 1). Other spectral ranges are suppressed. Advantages of this frequency selective principle are the direct extraction of spectral information without Fourier transformation and the improvement of the signal-to-noise ratio by the quality factor. Furthermore, such a sensor is insensitive to large interfering signals, which would overdrive wide-band transducers.
3. Resonance frequency tuning Sensors working in a frequency selective mode are limited to one-single measurement frequency, i.e. its resonance frequency. To extend the measurement range of such a sensor, resonance frequency tuning is used. The structures presented in this paper implement electrostatic-softening to vary the resonance frequency. The principle is based on electrostatically generated, amplitude-dependent forces acting on the seismic mass (Fig. 2). Therefore, the resonator structure contains an extra electrode system for resonance tuning at which a dc voltage, the tuning voltage Vtun is applied.
The total stiffness ktotal of the tuned resonator is calculated as follows: ktotal = k0 −
2 d2 C(x) Vtun 2 dx2
(1)
where Vtun is the tuning voltage, k0 the mechanical stiffness and C is the total capacitance between stator and seismic mass. The forces lead to a softening of the system and therewith to a lowering of the resonance frequency. In this way, the resonance frequency and thus the sensitive frequency band are set by the tuning voltage. To achieve linear sensor characteristics, it is essential that the second derivative of the capacitance function d2 C(x)/dx2 is constant over the amplitude range. Otherwise the resonance frequency depends on the signal amplitude. Such non-linearities implicate large amplitude errors. A quadratic capacitance function implemented by a comb system of linearly varied finger length using poly-Si technology is described in ref. [8]. A SCREAM-technology compatible approach using non-overlapping comb fingers is presented in ref. [10]. 4. Sensor structure 4.1. Fabrication A near-surface silicon bulk technology represents the technological basis for the sensor structures. The SCREAM technology has been previously described in several publications [11,12]. In a single-mask process laterally movable grid and beam structures with a cross-section as sketched in Fig. 3 are fabricated. The aspect ratio typically amounts to about 2–20 m. Top metal layers insulated from the bulk substrate by a SiO2 layer are used as electrodes for electrostatic force generation and capacitive signal detection. Individual electrodes are separated by trenches. Due to the reactive ion etching (RIE) lag, SCREAM-fabricated structures are characterized by a sensitivity of the produced trench depth to the local width of the mask gap. Overlapping comb structures implicate different trench widths in the overlapping and nonoverlapping areas. Since this leads to a varying cross-section of the beams, comb designs should use non-overlapping fingers.
D. Scheibner et al. / Sensors and Actuators A 123–124 (2005) 63–72
65
Fig. 3. Cross-section of a SCREAM-fabricated structure. (a) Schematic and (b) SEM-view.
4.2. Layout
Table 1 Design parameters of the sensor array
The sensor structures consist of a laterally movable mass supported by four-folded flexures (Fig. 4). The thickness of the mass is 20 m. Two-different comb systems at the seismic mass are designed either for capacitive signal detection (const. dC(x)/dx) or for resonance frequency tuning (const. d2 C(x)/dx2 ). The gap spacing between opposite fingers amounts to 2 m. The maximum resonance amplitude is limited to 10 m. The tuning range of a single tunable resonator is limited by the maximum tuning voltage. By grouping eight resonators with stepped base frequencies and overlapping tuning ranges into an array, the frequency range is widely extended. Lateral dimensions of the seismic mass range from 630 m × 660 m (cell 5) to 710 m × 840 m (cell 1). Flexures are between 146 m (cell 8) and 301 m (cell 1) long. The total size of the sensor chip is 7 mm × 10 mm (Fig. 5). Table 1 gives an overview of the design parameters.
Cell 1 2 3 4 5 6 7 8
m (g)
k0 (N m−1 )
f0 (Hz)
dC(x)/dx (F m−1 )
d2 C(x)/dx2 (F m−2 )
6.71 6.69 4.14 4.13 4.31 4.30 4.30 4.29
2.09 3.85 4.71 7.00 9.86 12.41 14.97 17.59
2811 3820 5369 6553 7611 8550 9396 10186
2.4e−3 2.4e−3 3.2e−3 3.2e−3 3.6e−3 3.6e−3 3.6e−3 3.6e−3
2.65e−8 2.65e−8 1.33e−8 1.33e−8 0.884e−8 0.884e−8 0.884e−8 0.884e−8
4.3. Experimental characterization 4.3.1. Measurement setup The sensor structures were experimentally characterized with regard to resonance frequency and bandwidth, extracted from the frequency response function. The frequency response of the sensor structures was obtained by sweep measurements using a lock-in-amplifier EG&G 7265. The
Fig. 4. Tunable microresonator: (a) schematic and (b) SEM view.
D. Scheibner et al. / Sensors and Actuators A 123–124 (2005) 63–72
66
Table 2 Measured parameters of the sensor array (11-113-7 Chip 36), f0tuned at Vtun = 35 V Cell
f0 (Hz)
1 2 3 4 5 6 7 8
2888 3957 5473 6605 7749 8656 9528 10271
f0tuned (Hz) 901 2841 3741 5277 6545 7578 8565 9385
Quality factor at f0
Bandwidth (Hz)
14.1 19.3 21.1 25.4 32.8 37.3 40.8 43.7
202 204 259 260 236 232 228 226
amplifier generates the driving signal and extracts the according frequency component out of the sensor response. Structures were excited mechanically by an electrodynamic shaker, the amplitude of which was controlled by laser vibrometer reference measurement. Vibration signal was detected capacitively by I/V (current-to-voltage) converters as intended in the later application (cf. Section 5.1), whereas temperature characteristics were analysed in a heat chamber using electrostatic actuation as described in ref. [13]. 4.3.2. Resonance characteristics Table 2 gives an overview of the measured characteristics. Due to technological tolerances of the fabrication process the untuned resonance frequencies of the characterized arrays (Wafer 11-113-7) are up to 3% higher than expected (Fig. 6). The tolerances arise from lithography (resist thickness, exposure time), etch (depth, profile) and coating (thickness) steps affecting geometric values such as structure height, width and beam profile. For the bandwidth of the resonance peak values between 200 and 260 Hz were measured. This is small enough to demonstrate the working principle, but for practical frequency selective measurements an electronic postfiltering with a smaller bandwidth (e.g. 50 Hz) is necessary (cf. Section 5.2). An essential feature of the sensor array is the overlapping of the tuning ranges. Experimental characterization shows that the desired tuning range is covered by a maximum tuning voltage of 35 V (Fig. 7). Therewith the fabricated array allows
Fig. 5. Photo of the sensor array.
Fig. 6. Resonance characteristics of untuned array cells (measured data).
continuous measurements in the frequency range from 1 to 10 kHz as required for the application in the measurement system. Some wafers, e.g. (11-113-2), which was used for temperature tests, showed higher resonance frequencies (up to +11% deviation from theoretical value). The reason was an exposure problem in the lithography step leading to wider structures, and thus stiffer flexures. Nevertheless, these arrays allowed continuous tuning down to 1 kHz. 4.3.3. Non-linearities The resonator array is used as a frequency selective vibration sensor. Therefore, it has to satisfy high requirements with regard to the linearity of the output signal. In theory, the bandwidth of the resonance peak is only dependent on damping effects and not on the actual resonance frequency. Therefore, the bandwidth is in contrast to the quality factor not influenced by the resonance tuning. This is an
Fig. 7. Resonance tuning of the array (measured data).
D. Scheibner et al. / Sensors and Actuators A 123–124 (2005) 63–72
67
Fig. 8. Stability of bandwidth over tuning range, measured data of cell 2.
important condition to correctly recalculate to the incoming acceleration by a frequency-independent transfer factor. Experiments confirm that the bandwidth of the resonance peak stays constant over the tuning range of the resonators (Fig. 8). The comb systems at the seismic mass generate electrostatic forces, which influence the resonance frequency. The impact on the sensor characteristics was analysed by monitoring the resonance frequency at different amplitude levels. Signal detection is carried out capacitively by the detection of reload currents using I/V converters. This requires a bias voltage Vbias at the detection combs (cf. Section 5.1). The detection comb system is designed for dC(x)/dx = const. and thus d2 C(x)/dx2 = 0. In practice, remaining non-linearities of the capacitance function were experimentally verified. This implicates a non-zero d2 C(x)/dx2 and leads to nonlinear amplitude-dependent forces. As a result, the resonance
Fig. 9. Impact of bias voltage on linearity, experimental data at tuning voltage Vtun = 0 V.
Fig. 10. Impact of bias voltage on linearity, experimental data at tuning voltage Vtun = 35.5 V.
frequency shifts to higher values with rising sensor amplitude. This undesirable tuning effect increases with rising bias voltage (Fig. 9). The tuning comb system is optimized for d2 C(x)/dx2 = const., but experimental characterization also revealed a non-ideal capacitance function. Measurements of transfer functions with maximum tuning voltage at different amplitude levels show the superimposition of non-linear effects of detection and tuning combs. In this case, the resonance frequency shifts to lower values with rising sensor amplitude (Fig. 10). This effect decreases with rising bias voltage. An optimum value for the bias voltage is found at Vbias = 4 V. The maximum resonance frequency shift over the amplitude range now amounts to 0.68% (Fig. 11).
Fig. 11. Frequency response of an array cell at different excitation amplitudes (measured data at Vbias = 4 V).
D. Scheibner et al. / Sensors and Actuators A 123–124 (2005) 63–72
68
Table 3 Impact of temperature on resonance frequency (Wafer 11-113-2 Chip 5A) Cell 1 3 8
fres at 30 ◦ C (Hz) 3121.2 5894.6 11088
fres at 150 ◦ C (Hz)
fres (%) −0.5 −0.42 −0.3
3105.2 5870 11055
Table 4 Impact of temperature on bandwidth (Wafer 11-113-2 Chip 5A) Cell
BB3dB at 30 ◦ C (Hz)
BB3dB at 150 ◦ C (Hz)
BB3dB (%)
1 3 8
223.1 280.0 324
270.7 345.0 389
21.3 23.2 20.1
4.3.4. Temperature behaviour SCREAM-fabricated structures consist of a stack of materials with different temperature coefficients. This mismatch results in temperature-dependent behaviour. Measurements in the temperature range from 30 to 150 ◦ C show a small resonance frequency shift of maximum 0.5% to lower values (Table 3) and a large increase of bandwidth of maximum 23% (Table 4). The folded flexures used in the resonator design release thermal tension, and are therefore relatively insensitive to temperature change. But the layers are deposited only on top and on the sidewalls of the silicon substrate (cf. Section 4.1). Thermal mismatch leads to tensions on the top side and bends down the free end of the flexure. This stress is released at higher temperatures. Consequently, stiffness and resonance frequency decrease. The increase of resonance peak bandwidth is probably caused by a lowering of the seismic mass. At room temperature, the mass is lifted out of the wafer surface due to thermal tension. The relaxation of the structure with rising temperature reduces the gap under the structure and slide-film damping effects increase. Therewith the bandwidth of the resonance peak increases. For the application as vibration sensor in a wide-temperature range, the large change of bandwidth has to be considered by in-system temperature calibration.
Fig. 12. Capacitive signal detection using I/V converters.
Fig. 13. Phase selective demodulation circuit.
Fig. 14. Schematic overview of the measurement system.
D. Scheibner et al. / Sensors and Actuators A 123–124 (2005) 63–72
69
5. Signal detection and evaluation 5.1. Capacitive signal detection The capacitive signal detection circuitry is based on the amplification of reload currents by I/V converters. The seismic mass is set to a bias voltage of 4 V. Stator combs are virtually connected to ground via the I/V converter stages. The I/V converters use a feedback resistor of R = 6.8 M. A parallel capacitance of Cp = 1 pF limits the bandwidth to a maximum frequency fmax = (2RCp )−1 and ensures stable operation. The signals of the two-stator combs are combined in a differential amplifier. The stator combs of all array cells are parallel connected (Fig. 12). The active cell is selected by the bias voltage at the seismic mass. All other cells are deactivated by connecting the seismic mass to ground. This concept allows the application of only one detection circuit for the complete array. 5.2. Phase selective demodulation The mechanical bandwidth of the sensor structures amounts to 200–260 Hz. This value is too large for spectral evaluation in the kilohertz range. Therefore, an electronic post-filtering is necessary. The measurement system uses a phase selective demodulation circuit working according to the lock-in principle. The demodulation is carried out by multiplying the raw signal with a reference signal. The reference frequency in the raw signal is rectified to a dc component while all other frequencies remain ac signals. A subsequent low-pass removes the ac signals. The corner frequency of this low-pass – in our case 25 Hz – sets the bandwidth of the demodulation. This process is performed in two-parallel paths with reference signals 90◦ out of phase. As a result, real and imaginary part of the raw signal at the reference frequency are extracted. In a last step, the RMS value is calculated from real and imaginary part. The implemented circuit uses two D-flip-flops to generate the two reference frequencies 90◦
Fig. 15. Photo of the frequency selective measurement system.
Fig. 16. Laboratory test measurement: (a) driving signal with large lowfrequency component, (b) sensor output, time signal and (c) spectrum obtained by frequency sweeping.
70
D. Scheibner et al. / Sensors and Actuators A 123–124 (2005) 63–72
out of phase and AD 630-demodulators (Analog Devices) for multiplication (Fig. 13).
6. Measurement system 6.1. Setup The sensor array has a complex interface. For practical measurements, the appropriate cell and tuning voltage have to be chosen and the result has to be recalculated. This effort is kept away from the user and encapsulated in a measurement system with a fully digital interface. The user sends a measurement command for a certain frequency. Now the system sets the sensor array to this frequency using internal calibration data, evaluates the sensor output and sends back the result (RMS value of acceleration at desired frequency). The measurement system contains the micromechanical structure, analog circuitry (I/V converter, phase selective
Fig. 17. Dynamic test using a 5 Hz modulated 3.9 kHz driving signal: (a) sensor ac output and (b) demodulated measurement result.
demodulation circuit, multiplexer for cell selection, amplifier for tuning voltage), digital control (microcontroller ATmega16) and interface (RS232) (Fig. 14). For low backlash to the object to be measured, the mass of the connected part has to be small. Therefore, the system is split into a small sensor head containing the micromechanics chip and the first analog detection stage and a larger control unit (Fig. 15). 6.2. Measurements For laboratory tests, the sensor head was mounted on an electrodynamical shaker. The driving signal contained several frequency lines in the measurement range of the system and a large low-frequency (75 Hz) component (Fig. 16a). The low-frequency component is far away from the resonance range of the sensor array, and thus filtered out mechanically. In the time signal of the sensor output the 75 Hz is no longer identifiable (Fig. 16b). By sweeping the resonance frequency, the spectrum was obtained and all higher frequencies contained in the driving signal were detected (Fig. 16c). Therewith the insensitivity to large interfering signals as an advantage of frequency selective sensors to wide-band transducers is practically demonstrated. In a second experiment the dynamic behaviour of the system was tested. A 3.9 kHz carrier modulated with 5 Hz was used as driving signal. The system continuously measured at 3.9 kHz. Results show, that the system including digital communication follows the 5 Hz modulation with 35 samples/s (Fig. 17). The sample rate is mainly limited by the settling
Fig. 18. Sensor mounted on the cylinder head of a combustion engine (MZ RT 125).
D. Scheibner et al. / Sensors and Actuators A 123–124 (2005) 63–72
time of the sensor structure and the phase selective demodulation circuitry (cf. Section 5.2). After verifying the functionality in laboratory experiments, the system was used in several practice tests, for instance at a combustion engine. In this measurement, the sensor was mounted on the cylinder head of the engine to detect knocking combustion (Fig. 18). Knocking is uncontrolled combustion, caused by the spontaneous ignition of portions of the mixture, which have not yet been reached by the advancing flame front, triggered by the ignition spark. The consequence of knocking is overheating of parts of the engine, and thus destruction of the engine. Therefore, knocking detection is an important part of current ignition timing systems. On an engine test stand normal and knocking combustion were setup and the vibration spectrums for both cases were obtained with the frequency selective measurement system (Fig. 19). If knocking occurs obvious changes in the spectrum are clearly visible. The overall spectrum increases
71
in amplitude. At knocking combustion, the amplitude of the clear peak found at 5 kHz approximately doubles. The frequency selective vibration sensor can be set to an expressive frequency (e.g. 5 kHz) to permanently monitor the vibration amplitude. Then – in case of exceeding a critical level – the system gives an alert indicating knocking combustion. By these tests the applicability of the system for measurements in industrial environment was demonstrated.
7. Conclusions SCREAM-fabricated resonators are suitable for frequency selective vibration measurements. Electrostatic-softening turned out to be an appropriate means for resonance frequency tuning in the range of 1–10 kHz. A measurement system containing the sensor array, analogue and digital circuitry allows measurements in industrial environment and demonstrates the advantages of the frequency selective approach in practice. Further work concentrates on an optimization of the sensor structure. The aim of a new design is improved linearity. This will be achieved by redesigned comb structures with better capacitance characteristics (dC(x)/dx = const. for signal detection, d2 C(x)/dx2 = const. for resonance tuning). Besides the further improvement of the frequency selective vibration sensor also other applications for such a measurement system such as tool wear at cutting tools and lathes or cavitation sensing at pumps have to be investigated.
Acknowledgement Parts of the work were done within the Collaborative Research Center SFB 379, which is funded by the German Research Association (DFG).
References
Fig. 19. Vibration spectrum of the engine running at 4000 rpm measured using the frequency selective system: (a) normal combustion and (b) knocking.
[1] H. Bloch, F. Geitner, Practical Machinery Management for Process Plants, Machinery Failure Analysis and Troubleshooting, vol. 2, Gulf Publishing, Houston, 1994. [2] V. Wowk, Machinery Vibration: Measurement and Analysis, McGraw-Hill, New York, 1991. [3] W. Benecke, L. Csepregi, A. Heuberger, K. K¨uhle, H. Seidel, A frequency-selective, piezoresistive silicon vibration sensor, in: Proceedings of TRANSDUCERS, Philadelphia, 1985, pp. 105–109. [4] H. Fritsch, R. Lucklum, T. Iwert, P. Hauptmann, D. Scholz, E. Peiner, A. Schlachetzki, A low frequency micromechanical resonant vibration sensor for wear monitoring, Sens. Actuators A 62 (1997) 616–620. [5] E. Peiner, R. Mikuta, T. Iwert, H. Fritsch, P. Hauptmann, K. Fricke, A. Schlachetzki, Micromachined resonator for cavitation sensing, Sens. Actuators A 76 (1999) 266–272. [6] J. Wibbeler, C. Steiniger, M. K¨uchler, K. Wolf, T. Frank, J. Mehner, W. D¨otzel, T. Gessner, Resonant silicon vibration sensors with voltage controlled frequency tuning capability, in: Proceedings of EUROSENSORS XIII, The Hague, September 12–15, 1999, p. 24B3.
72
D. Scheibner et al. / Sensors and Actuators A 123–124 (2005) 63–72
[7] J. Mehner, D. Scheibner, J. Wibbeler, Silicon vibration sensor array with electrically tunable band selectivity, in: Proceedings of MICRO SYSTEM TECHNOL, D¨usseldorf, March 27–29, 2001, pp. 267–272. [8] K. Lee, Y. Cho, A triangular electrostatic comb array for micromechanical resonant frequency tuning, Sens. Actuators A 70 (1998) 112–117. [9] W. Ye, S. Mukherejee, N.C. MacDonald, Optimal shape design of an electrostatic comb drive in microelectromechanical systems, J. Microelectromech. Syst. 7 (1) (1998) 16–25. [10] D. Scheibner, J. Wibbeler, J. Mehner, B. Br¨amer, T. Gessner, W. D¨otzel, A frequency selective silicon vibration sensor with direct electrostatic stiffness modulation, Analog Integr. Circ. Signal Process. 37 (2003) 35–43. [11] K.A. Shaw, Z.L. Zhang, N.C. MacDonald, SCREAM I: a single mask, single-crystal silicon, reactive ion etching process for microelectromechanical structures, Sens. Actuators A 40 (1994) 63–70. [12] T. Gessner, A. Bertz, C. Steiniger, U. Wollmann, Material and technology approaches of surface micromachining, in: Proceedings of MICRO MAT, Berlin, April 16–18, 1997, pp. 90–95. [13] D. Scheibner, J. Mehner, D. Reuter, U. Kotarsky, T. Gessner, W. D¨otzel, Characterization and self-test of electrostatically tunable resonators for frequency selective vibration measurements, Sens. Actuators A 111 (2004) 93–99.
Biographies Dirk Scheibner is a research engineer at SIEMENS A&D responsible for microsystems. He received the Dipl.-Ing. degree in Microsystem and Precision Engineering from Chemnitz University of Technology in 1999. From 2000–2004 he worked as a Scientific Assistant at the Department of Microsystems and Precision Engineering at the Chemintz University of Technology. He received a Dr.-Ing. degree in electrical engineering and information technology from the same University in 2005. His research interests are design and test of accelerometers and vibration sensors fabricated in silicon surface microtechnology. Jan E. Mehner is a scientific assistant at the Department of Microsystems and Precision Engineering at the Chemnitz University of Technology.
He received a Dr-Ing and Dr-Ing habil degree in electrical engineering and information technology from the same University in 1994 and 1999, respectively. From 1998 to 1999, he was a visiting scientist at the Massachusetts Institute of Technology in the field of macromodelling. His research interests include analytical and numerical methods to design microsystems, CAD-tools and computational algorithms for problems with coupled fields. Danny Reuter is a scientific assistant at the Centre for Microtechnologies at the Chemnitz University of Technology. He received his Dipl-Ing degree in microsystem and precision engineering from Chemnitz University of Technology in 2002. His research interests are silicon surface microtechnology and wafer-level packaging of high aspect ratio MEMS. Thomas Gessner received the Dipl-Phys degree in physics in 1979 and the Dr rer nat degree in 1983 from the Faculty of Natural Science at TU Dresden and the Dr-Ing habil degree in 1989 from University of Technology Karl-Marx-Stadt. Professor Gessner had several positions in the industry (ZMD Dresden) at the development of metallization systems for d-RAM integrated circuits. Since 1991, professor Gessner is director of the Center of Microtechnologies at the Chemnitz University of Technology and has a chair for microtechnology at Chemnitz University of Technology in 1993. In the year 1996, he was appointed as member of the Academy of Science in Saxony. Since 1998, professor Gessner is the head of the Department “Microdevices and Equipment” at Berlin Fraunhofer Institute for Reliability and Microintegration (FhG-IZM) as a branch-lab in Chemnitz. In 1998, he became member of the Scientific Advisory Board of the Federal Republic of Germany. Wolfram D¨otzel received the Dipl-Ing degree in electrical and precision engineering from Technical University Dresden in 1966 and the Dr-Ing degree from the Chemnitz University of Technology in 1971. The research work was focused on the design of electromechanical systems and their reliability. Since 1993, Wolfram D¨otzel is professor for microsystems and precision engineering at the Chemnitz University of Technology. His current work is focused on design and simulation of micromechanical components, their characterization by experimental methods and their application in a macro-system. Professor D¨otzel is member of the Academy of Science of Saxony in Leipzig.