Position encoding and closed loop control of MOEMS translatory actuators

Transcription

1 Position encoding and closed loop control of MOEMS translatory actuators M. Lenzhofer 1, A. Tortschanoff 1, A. Frank 1, T. Sandner 2, H. Schenk 2, M. Kraft 1, A. Kenda 1, 1 Carinthian Tech Research AG Austria, Europastraße 4/1, 9524 Villach, Austria 2 Fraunhofer Institute for Photonic Microsystems, Maria-Reiche-Str. 2, Dresden, Germany ABSTRACT We present a new method for detecting the accurate position of micro-electro-opto-mechanical system (MOEMS) devices, thus enabling the implementation of closed-loop controls. The ensuing control mechanism allows building robust MOEMS-based Fourier-transform infrared (FTIR) spectrometers with large mechanical amplitudes and thus good spectral resolutions. The MOEMS mirror device, a rectangular 1.65 mm² metalized plate mirror suspended on bearing springs and driven by comb-structured electrodes, is driven by a rectangular signal with a duty cycle of 50% and high voltage levels up to 140 V at a frequency near twice its mechanical resonance frequency. Out-of-plane mirror displacements of up to ±100 µm have thus been achieved. To handle the high bandwidth of the sinusoidal mirror position reference signal, which is generated by a laser reference interferometer, an analog position detection circuit is necessary. This dedicated circuit demodulates the reference signal and generates a highly accurate control signal returning the zerocrossing position of the mirror. This permits the implementation of a closed-loop control, which ensures optimally stable MOEMS mirror movement and maximal mechanical amplitude, even under varying environmental conditions. While this solution has been developed for a specific MOEMS device, the principle is widely applicable to related components. Keywords: MOEMS, translatory micro-mirror drive, FTIR spectrometer, Michelson interferometer, closed-loop control 1. INTRODUCTION Infrared (IR) spectroscopy is used for the chemical analysis of many (organic) substances, mainly because it is a noninvasive, non-destructive measurement method with high information content [1,2]. With their ability to cover wide spectral ranges and their superior sensitivity in comparison to other IR principles, Fourier-transform infrared (FTIR) spectrometers are now the instruments of choice for most such applications. FTIR systems are based on Michelson interferometers, and, as such, require precisely and reproducibly aligned and driven mirrors [3]. With standard mirrors, this necessarily requires complex and expensive mirror drives. To overcome the related limitations, especially such related to acquisition speed, size and instrumentation cost, we recently developed a MOEMS device-based FTIR spectrometer [4,5]. Based on a MOEMS translatory mirror [6], these compact spectrometers include two interferometers, one to detect the position of the MOEMS device and consequently control the stable oscillation of the modulating mirror, and the other one for the actual spectral measurement. To guarantee a stable mirror movement, and thus a well-defined and constant spectral bandwidth even under varying environmental conditions, a closed-loop control of the MOEMS device is highly advantageous. The practical approach presented here consists of an analog circuit that first demodulates the acquired reference interferometer signal. This information is then used in real-time to determine the zero-crossing position and actuate the driving voltages with optimal timing. The following focuses on the electronic integration of the position detection circuit, the closed-loop control and the thus achieved advanced. This work thus deals with a vital detail for optimized driving and control of MOEMS components and their application in real-world systems. Smart Sensors, Actuators, and MEMS IV, edited by Ulrich Schmid, Carles Cané, Herbert Shea Proc. of SPIE Vol. 7362, SPIE CCC code: X/09/$18 doi: / Proc. of SPIE Vol

2 2. TRANSLATORY MOEMS MIRROR DEVICE 2.1. MOEMS Core Component The MOEMS device used for this study is produced from of highly doped single crystalline silicon using standard semiconductor processes. The 1.5 x 1.1 mm² mirror plate is front side-coated with an aluminum layer to make it reflective in the infrared [7]. Front side and backside of the MOEMS device are shown in Figure 1 (a) and (b), respectively. The element is driven by applying modulated electrostatic voltages between the interlocking comb-like drive electrodes located along the edges between the moving part (mirror and springs) and the stationary frame. This arrangement allows the mirror plate to oscillate orthogonally to the image plane. It is worth noting here that the MOEMS component is a nonlinear system with parametric resonances. contact pads spring mirror drive electrodes -rrrrl..l L_LJ1 Figure 1: Photograph of an 8 khz device: front side (a) and backside (b). [6] 500µm a b The implementation of the control circuit is based on the mirror devices characterized in [6]. For device characterization, the device to be measured was housed into a ceramic controllable vacuum packaging and introduced into a Michelson interferometer setup as the moving mirror. Pressure and driving voltages were varied systematically while the oscillation amplitude was determined from the interferogram of a nm HeNe laser source. Figure 2 shows parameter dependencies for a 5 khz device, giving the dependence of amplitude and phase position on the actuation frequency at constant driving voltage and constant pressure (a) and the achievable frequency/amplitude correlations for different driving voltages (b). Amplitude / µm Δϕ / π Pa, 20 V Oscillation frequency / Hz (b) Oscillation frequency / Hz (a) Figure 2: (a) Amplitude and phase versus frequency at a pressure of 6 Pa and a constant driving voltage level of 20 V. (b) Amplitude versus frequency of a 5 khz device and a constant pressure of 10 Pa. Amplitude / µm V 16 V 18 V 19 V 20 V 21 V 22 V Proc. of SPIE Vol

3 2.2. Driving of the MOEMS Devices Translatory MOEMS mirror devices with mechanical resonance frequencies between 5 khz and 10 khz have been designed by the Fraunhofer IPMS. For driving the MOEMS devices, several requirements concerning the operation environment have to be met. This includes the pressure in the vacuum packaging, which has to be below 10 Pa to allow mechanical amplitudes of ±100 µm, the aspired oscillation amplitude and the driving voltage. The micro-mirror device is driven by a rectangular driving signal with a duty cycle of 50% at a modulation frequency near the double mechanical resonance frequency of the respective element [8]. In operation, the driving voltage thus accelerates the mirror towards its rest position (i.e. the neutral in-plane position) and should be switched off instantaneously once the mirror passes that point. The voltage is subsequently turned on again once the oscillating element reaches its maximum deflection, and switched off again when the cross-over point of the mechanical oscillation is reached on the back-travel of the mirror element. Thus, when applying a rectangular voltage signal with a frequency sufficiently near the double resonance frequency of the device, the MOEMS device will start to oscillate. As the maximum energy is coupled in when the comb electrodes are in maximal proximity, which is the case for the nondeflected position where comb-structured electrode fingers are parallel, timing of the switch-off is critical; delays in this may cause the mirror to stall. Therefore, a very precise zero-crossing detection circuit is necessary. In comparison, timing of the switch-on point is much less critical, as both the attractive forces and the mirror speed are significantly lower there. 3. SYSTEM INTEGRATION 3.1. Mechanical and Electrical Setup in the FTIR Spectrometer The closed-loop control was installed into an existing FTIR spectrometer prototype setup (Figure 4) comprising two separate interferometers, one for detecting the position of the MOEMS mirror device, which is required for both signal back-transformation and mirror control, and the other for performing the actual spectral measurements. The Michelson interferometer is one of the most common configurations for optical interferometry [3]. Basically, the interferometer generates a position-variable interference. A beam splitter splits the incoming beam (in the reference laser interferometer a mode-stabilized HeNe laser for the reference measurements and a vertical-cavity surface-emitting laser (VCSEL) with a wavelength of 690 nm for the spectrometer prototype, respectively) into two equal beams. These beams are then reflected back, in one instance by a stationary reference mirror, in the other by the backside of the MOEMS micro mirror, and recombined on the beam splitter; the resulting signal is then transferred to a detector (Figure 3 (a)). A III ui,i!lii! t Laser reference interferometer Reference Signal! V driving voltage! V (a) Figure 3: (a) Reference laser interferometer. The position sensing Michelson setup to the left of the moveable micro-mirror uses the reflection off the micro-mirror s backside. (b) Driving voltage and position signals: The square pulsed driving signal of the MOEMS device and the response signal from the reference interferometer. (b) Proc. of SPIE Vol

4 Depending on the optical path difference, which varies over time with the movement of the driven MOEMS device, the recombined light will interfere accordingly and produce mirror-position dependent interference. If the optical path in the two arms of a Michelson differs by an integer number of wavelengths, constructive interference leads to a strong signal at the detector, while at every half wavelength the two beams interfere destructively, resulting in a weak signal at the photodiode. Consequently, the laser interferometer generates a frequency-modulated sinusoidal signal at the photo detector, i.e. the interferogram of the essentially monochromatic laser (Figure 3(b)). This signal now allows accurately measuring the deflection of the MOEMS device over time. The laser interferometer signal thus serves two purposes. First, it is used to act as the sampling clock of the analog to digital converter [6], thus compensating for the massively non-linear movement of the mirror. Second, it serves as input for the closed-loop control circuit. Interrerometer for Measurements MOEMS Device in Vacuum Packaging Beam Splitter Reference I nterferometer Photodiode Mounting Reference Laser Fixed Mirror Figure 4: Optical bench of the FTIR spectrometer with the measurement and reference Michelson interferometer. The effective accuracy of the interferometric measurements is critically dependent on the wavelength stability of the laser. Therefore, in the FTIR prototype device a temperature-controlled VCSEL laser diode is used. While having worse wavelength stability than the single-mode HeNe laser using for system characterization, this solution proved to be the best compromise between performance, cost and size. The VCSEL diode was controlled by a discretely built-up constant current source. Additionally, a second current source was designed to act as temperature control circuit of the thermo electrical cooler (TEC) located in the housing of the laser diode. For the semiconductor photodiode of the reference interferometer path, a dedicated transimpedance amplifier was developed that converts the photo-current into a processible voltage signal. Integration of closed-loop electronics implies three steps. First, the control electronics have to precisely determine the zero transition of the MEMS mirror and generate a corresponding trigger signal. This signal is then used to synchronize the driving voltage of the mirror. In the test setup, this was done using an external function generator. Finally, the control algorithm has to be implemented in a suitable microcontroller. Additionally, the electronics have to allow to still use the laser signal for position encoding of the MOEMS device, a procedure that is essential to compensate for the non-linear movement of the mirror. Proc. of SPIE Vol

5 Hardware Development Signal Evaluation Approach and Circuit Design As a first step, the relevant signals for driving voltage switching were studied in detail. Figure 5 shows the corresponding characteristic behavior for the free-running (i.e. open-loop) mode, which was used as the standard driving principle prior to the development described here. SOOV Ch2 O.OV H1O.OJs A J 11.2V 19 Mar 2009 U' 291.SOOs 13:05:41 Figure 5: Driving voltage and position signal of the MOEMS mirror device in open-loop mode over time. The cyan curve represents the effective driving voltage applied at the MOEMS mirror device, while the yellow curve is the modulated HeNe laser intensity signal, used also as position encoding signal. To realize a closed-loop control it is essential to switch off the driving voltage as precisely as possible at the mechanical crossing point of the mirror through the neutral plane. The precision of that switching in this point is essential, as otherwise the mirror may stall. To enable this high-precision switching, a dedicated position detection circuit was developed. This device identifies the optimal switch-off point by finding the maximum frequency of the mirror s position decoding signal, corresponding to maximum speed of the mirror. The blue curve in Figure 6 displays the derivative of the frequency modulation of the position encoding signal, calculated from the delay between successive fringes. The zero-crossing of this curve corresponds to the point of maximum speed and thus to the zero-crossing of the MEMS mirror. The applied driving voltage (black curve in the Figure 6) is then controlled based on this information. Intensity / a.u Intensity / a.u T im e / μ s T im e / μ s Figure 6: Overlay of driving voltage, reference signal and zero-crossing detection signal. The red, rapidly oscillating curve is the actual signal from the laser reference interferometer. The black curve is the rectangular driving signal generated by the position detection circuit and the blue curve represents the derivative of the frequency modulation. Inset is a zoom of the zero position, highlighting the achievable timing precision. Proc. of SPIE Vol

6 The circuit thus directly evaluated the frequency modulation of the reference signal and derives the control parameters from this. One key advantage of this approach is that the signal amplitude variation (see Figure 3 (b)), which is caused by a dynamic deformation of the mirror during its oscillation [6], does not interfere. A further requirement is that the circuit has to be able to deal with dynamically changing modulation frequencies between 1 MHz at the reversal points of the MOEMS device and 8 MHz at the zero-crossings. In total, the approach employed here demodulates the reference interferometer signal and generates the corresponding position signal, which is accurate enough to switch the driving voltage on and off at the proper time stamp, thus ensuring an optimally stable amplitude. Figure 7 shows the block diagram of the underlying demodulation circuit. Figure 7: Block diagram of the position detection circuit to switch on and off the driving voltage. The signal chain contains an amplifier, a comparator, a monostable multivibrator and a second comparator stage. As the information of the position is stored within the frequency, as a first stage the constant component of the analog signal is blocked by a capacitor. The residual system is subsequently digitalized using a zero-crossing detector, using a simple comparator stage. The resulting intermediate signal is shown in Figure 8 (cyan curve). The pure frequency modulated signal is then transformed into a pulse-width modulated signal using a pulse-count detector, essentially a signal triggered monostable multivibrator stage. The time of the fixed pulse width is matched to the maximally occurring frequency of the MOEMS mirror, i.e. the highest frequency the duty-cycle is exactly 50%. This ensures maximal voltage levels after the filter even for the highest frequencies. The resulting pulse-width modulated signal with fixed pulse width is then filtered again, yielding an analog voltage with low intermediate frequency, shown as purple signal in Figure 8. To extract the switching signal for the high voltage the signal is connected to a comparator stage again. In contrary to a zerocrossing detector, this comparator stage switches the output, if the input signal changes its direction. This means that it works similar to a peak detector. The maximum of the input signal appears at the highest frequency, so that after the peak detection the high voltage is switched off. The resulting control signal (green) is also shown in Figure 8. Stop - 400mV 4.O V 4O.ps 297p Ji Chi 5OOmV 1Ch2 S.00V H20.OpsA Ch4 J 2.80V Ch2 Freq 3.406M Hz N led rige Autlösung 19 Mar :43: 02 Figure 8: Position encoding signal measured by a photo detector (yellow), digitalized signal (cyan), signal after filter stage (purple) and the resulting control signal to switch the driving voltage (green). Proc. of SPIE Vol

7 For obvious reasons, the time constant of the monostable multivibrator has to be adjusted to the mirror parameters. When varying the parameters, also the resistor and capacitor values of the RC elements have to be adopted, as they define the timing of the monostable multivibrator. Clearly, this is a disadvantage of this analog concept. Still, altogether this approach allows to detect and actuate the optimal switching point in real-time and with unparalleled precision and reliability. As shown in the inset in Figure 6, the absolute error between theoretical optimum and actual switch point is less than one period of the reference signal, i.e. the maximal achievable precision Laboratory Testing Before incorporating the resulting control algorithms in a microcontroller-controlled closed-loop arrangement, the position detection circuit was tested under laboratory conditions in a dedicated test configuration. The position encoding circuit was connected to a Thorlabs photo detector and a stabilized HeNe laser acted as laser source. The Michelson interferometer used, however, was identical to the ones used in the FT-prototype (see Figure 4). For mirror actuation, a function generator connected to a high voltage driver was used. Figure 9 shows the total arrangement of the test equipment. 0[i Photodetector Control Circuit Function Generator HeNe Laser HV Driver Stage MOEMS device Figure 9: Arrangement of the test equipment used for verifying the function of the position detection circuit and testing the closed-loop principle. To test the principle of the closed-loop control, the reference signal of the detector together with the driving signal at the output of the position decoding circuit and the function generator output were scoped. The signal from the position decoding circuit was wired directly to the external input of the function generator to trigger the system. The pulse width modulated control signal of the generator was thus tied to the high voltage stage, switching on and off the driving voltage according to the output of the control circuit. Before switching to closed-loop control mode, the MOEMS device has to be started up by slowly increasing the frequency of the generator towards the resonance frequency of the mirror device. This was done manually. Given a drive working frequency of the mirror of khz, the start-up sweep was started at khz with 1 Hz steps, until a stable open-loop mirror oscillator close to the resonance frequency was achieved. When the frequency got near the resonance of the mirror, the generator was switched to the external trigger mode. For this switch-over of operation modes it is important that the phase difference is not too large at the time of the take-over of the external trigger, as otherwise the mirror will stall. This switching to the external trigger mode of the generator now means a change from open-loop to the desired closed-loop control mode, as the trigger signal is generated by the control electronics switching off the driving voltage precisely at the position of the highest frequency, i.e. the mechanical zerocrossing point of the mirror device. To test the operation-stabilizing function of the closed-loop control, various parameters influencing the resonance frequency of the device were systematically changed. This was additionally monitored by counting detector pulses with a scope card. A LABVIEW test program calculated the corresponding mechanical amplitude of the mirror. The parameter that can be varied most conveniently in a controlled was is the driving voltage, which was varied from 25 V to 34 V in 1 V steps. Proc. of SPIE Vol

8 As shown in Figure 10, in closed-loop mode the in this case voltage-level effected change in mirror amplitude is automatically accompanied by an adjustment of the corresponding resonance frequency, thus invariably driving the MOEMS components at its optimal point of operation. 120E1:I 118C'L' - 117EL' Voltage (V) Figure 10: The red curve indicates the amplitude of the mirror in the closed-loop mode at the corresponding driving levels and the black curve shows the adjustment of the frequency Implementation Approach of the Closed-Loop Control In a final implementation, the previously described manual start-up procedure of the mirror, followed by a switch-over to the closed-loop control, should be performed autonomously. This has been achieved by implementing the system into a microcontroller. The position detection circuit is now connected to both an input port of the controller and an XOR gate, which generates output pulses if the signal from the position decoding circuit and the generated, pulse-width modulated mirror driving signal of the microcontroller differ. Mirror H V-Driver PWM Closed-Loop XOR Phase Control f/2 F- LHH T Reference Amplifier Comparator Monoflop Comparator Interferometer Figure 11: Block diagram of the closed-loop control circuit built up by the position detection circuit, reference interferometer and MOEMS mirror device. The frequency of the driving signal is twice the mechanical frequency, which is detected by the position electronic circuit. This means that the controller has to generate a pulse modulated signal with the double frequency to drive the mirror. Before comparing the signal with the detection electronics, the signal has to be halved in frequency again. The control algorithm in the processor now tries to reduce the pulse width, which represents the phase of the mirror, to an absolute minimum to ensure the maximal mechanical amplitude within a stable oscillation. Figure 11 shows the block Proc. of SPIE Vol

9 diagram of the mirror driving system and its control electronics. The microcontroller also stores all other necessary mirror parameters (amplitude, voltage, resonance frequency), making it possible to drive a MOEMS device at its own specific optimum. As each mirror device has slightly different parameters, these parameters have to be updated when changing the mirror module of the FTIR spectrometer system. 4. SUMMARY AND OUTLOOK To conclude, we have successfully demonstrated the feasibility and advantages of closed-loop operation of translational MEMS mirrors, micro-device that are e.g. at the core of our miniaturized FTIR spectrometer. Using demodulation of the reference signal to find the point zero-crossing works very accurately and was integrated in a compact electronic circuit. The exact zero-crossing determination is important for the FTIR measurements, since it increases the precision of the absolute position determination, which is important when averaging multiple scans. It will represent a significant improvement, compared to the current approach, which uses the zero-peak of the IR interferogram. Furthermore, the results of this work enable a reliable closed-loop operation, which will provide the possibility for more robust and stable operation of our FTIR devices even under challenging and varying environmental conditions. Altogether, closed-loop operation is a promising approach to drive sinusoidally oscillating M(O)EMS devices, including new micro-mirror designs [9] that have significantly larger amplitudes but are more delicate to drive because of the novel suspension principle [7] used. Currently this circuit is implemented and tested in our FTIR device. Further improvement will involve the implementation of the whole control circuit into an FPGA, thus reducing the analog part to an absolute minimum. The fact that such a device works at very high clock speeds makes it is possible just to measure the time between the pulses and so find the correct switching point for the high voltage. In this case the adjustment of the circuit even to significantly different mirror characteristics can be done purely software controlled. REFERENCES [1] [2] [3] [4] [5] [6] [7] [8] [9] J. Chalmers (Ed.), Handbook of Vibrational Spectroscopy, Wiley (2002) M. Kraft, Vibrational spectroscopic sensors, in Optical Chemical Sensors, F. Baldini, A. Chester, J. Homola and S. Martellucci, Eds. NATO Science Series II, Vol. 224, Springer (2006) P. Griffiths, J. dehaseth, Fourier Transform Infrared Spectrometry, Wiley Interscience (1986) M. Kraft, A. Kenda, T. Sandner, H. Schenk, M(O)EMS Spectrometers: Making Spectroscopic Optical Chemical Sensors Fast and Mobile, Proc. IEEE Sensors 2008 (2008) M. Kraft, A. Kenda, T. Sandner, H. Schenk, MEMS-based Compact FT-Spectrometers A Platform for Spectroscopic Mid-infrared Sensors, Proc. Mikroelektronik 08, OVE Publ. 50, p (2008) A. Kenda, Ch. Drabe, H. Schenk, A. Frank, M. Lenzhofer, W. Scherf, Proc. SPIE 6186, pp (2006) C. Drabe, T. Klose, H. Schenk, A. Wolter, H. Lakner, Proc. SPIE 6186, , (2006) H. Schenk, P. Dürr, D. Kunze, H. Kück, International Mechanical Engineering Congress and Exposition, MEMS- Vol. 1, pp , Nashville (1999) A. Tortschanoff, A. Kenda, M. Kraft, T. Sandner, H. Schenk, W. Scherf, Improved MOEMS based ultra rapid Fourier transform infrared spectrometer, presented at SPIE - DSS2009 (2009) Proc. of SPIE Vol