Digital resonance for MEMS

Proceedings of the 2009 Spanish Conference on Electron Devices - Feb 11-13, 2009. Santiago de Compostela, Spain. Digital Resonance for MEMS J. Ricart, J. Pons, S. Gorreta and M. Dominguez Micro and Nano Technologies Research Group (MNT), Technical University of Catalonia (UPC). Address: Jordi Girona 1-3, m6dul C4, CP:08034, Barcelona, Spain. First author's e-mail: [email protected] 1 k Abstract-This work shows the application of Pulsed Digital f = | (1), Oscillators to the detection of physical changes in MEMS devices 2TZ jmMEMS that cause small shifts in their resonance frequencies. Such devices can be used as resonators in several sensor applications. According to this, a case study using a PDO structure to measure being fo the mechanical resonant frequency, k the spring small concentrations of volatile organic gas compounds (VOCs) is constant and mMEMS the total mass of the moveable plate. introduced here. The MEMS devices used are cantilevers with a Therefore, in a PDO system with such device the resonance thin layer of polymer sensitive to the VOC concentration. Such frequency will inversely vary with the mass of the MEMS devices have been simulated with the Coventor software to see the device. This principle can be applied to the sensing of influence of the polymer layers on their mechanical responses. Finally, experimental measurements with various VOCs have dangerous compounds, such as VOCs, if we had MEMS been done, and results extracted from two PDO system sources, devices where a change of the compound concentration would digital and analog, have been analyzed and compared. lead to a mass variation. For VOCs like toluene or benzene, polymers like PDMS or PECH can be used for this purpose. I. INTRODUCTION Finally, let us note that the use of layers of polymers Pulsed Digital Oscillators (PDOs) are simple circuits sensitive to VOCs is widely known and extensive bibliography initially designed to overcome some of the usual non linearity on the subject [3, 5, 6] is available. problems of MEMS actuation and sensing [1,2]. PDOs are sampled structures which, at each sampling time, detect if aP0 MEMS device (i.e. a two-parallel plate resonator) is above or DE2 da below its rest position and, through a 1-bit quantifier followed boad by a digital feedback loop, they generate series of short force pulses that actuate the MEMS resonator. So, in order to work PD0 FDOg PO ital properly, PDO structures only need to sense the sign of the anEai i 6dback loop position of the MEMS device. out Apart from their simplicity, another key feature of PDO structures is that the oscillation frequency can be extracted Frequenc meoter from the bit stream that actuates the MEMS device (digital output in Figure 1). Therefore working in the digital domain is straightforward when using PDOs, thus all digital facilities for A&D osllocope MEMS & PDO data storing, data processing, etc. become easily available. On analog sedclon the other hand, one can also sense the MEMS position Figure 1. PDO-based experimental setup. Two simultaneous frequency behavior in PDOs (i.e. analog output in Figure 1) and use it as measurements can be done, one extracted from the MEMS position sensing an additional way to obtain the oscillation frequency, but the (analog output), and a second one extracted from the PDO bit stream (digital output). prize to pay is that the sensing requirements may increase and that some extra complex equipment for analog measurements must be used. Anyway and only for comparison purposes both ways, digital and analog, to extract the oscillation frequency II. MECHANICAL SIMULATIONS have been used in this work. The MEMS devices used are similar to those used in some On the other hand, a type of application where PDO systems previous works [1-3]: silicon cantilevers with thermoelectric can be rather advantageous is the sensing of phenomena that actuation and piezoresistive position sensing through a cause changes in resonance frequency. It is well known that Wheatstone bridge. Each cantilever is a moveable squared the mechanical frequency of MEMS devices depends on the plate held by three arms of 100 pim long. Plate dimensions are mass of their moveable parts. In example, if we take the usual 300 pim long and 300 pim wide, while thickness is 15 pim. In iD mass-spring model for a two parallel plate MEMS device, order to allow VOC sensing, a micropipette was used to then we have deposit small polymer layers on top of the moveable plates. 978- 1-4244-2839-7/09/$25.00 (C)2009 IEEE 312 Authorized licensed use limited to: UNIVERSITAT POLITÈCNICA DE CATALUNYA. Downloaded on March 16,2010 at 09:16:00 EDT from IEEE Xplore. Restrictions apply. Proceedings of the 2009 Spanish Conference on Electron Devices - Feb 1 1-13, 2009. Santiago de Compostela, Spain. The first objective was to analyze, through simulations, the the device of Figure 2a, a 20 tm thick layer of PDMS with a mechanical properties (including vibration modes and the main density of 9.7x 1o-17 kg/gm2 was added at this stage, resulting resonance frequency) of the devices before and after the in a new device with the structure and behavior shown in deposition of the polymer layer. Figure 2b. Obviously, this new device exhibits a lower To a priori characterize such mechanical properties of our resonance frequency, as its mass has been increased by the MEMS resonators the Coventor software environment was PDMS layer. In particular, the value of the mechanical used. According to equation 1, the expected result is a decrease resonance frequency decreases to 98.906 kHz. of the MEMS resonance frequency, since by adding a polymer layer the total mass of the device increases. III. POLYMER LAYER CHARACTERIZATION The detailed description of the making of the MEMS Figure 3 shows a 3D image, obtained with a Wyko NT9300 resonator, a bulk micromachining process from a SOI wafer, interferometric microscope, of a MEMS resonator. The image was introduced into the Coventor environment. From this clearly shows the deposited PDMS. Since good values of information we obtained the 3D models that allowed us to height and perimeter of the PDMS layers are easy to obtain, perform mechanical and thermal simulations. As an example, one can estimate the PDMS volumes, and using the specific Figure 2.a illustrates the first vibration mode of a specific gravity provided by the manufacturer (1.03xli0 kg/in), an MEMS cantilever, which exhibits a mechanical resonance approximated value of the deposited PDMS mass can be frequency of 100.44 kHz. calculated. In the case of Figure 3, such extra mass is found to be 1.02x10-9Kg. ModalQDisplacemeniMOttvlag ~ SSE;00 2 SE-0i 5 SE-0 1 5E-01 I 1 ,E00 Li n S100443 Hz COVENTOR w Eigure 2a. Coventor mechanical simulation of the first vibration mode of a MEMS resonator without polymer layer. Figure 3. Interferometric microscope caption of a MEMS resonator with a PDMS layer deposited on its top. Table I shows a comparison of some frequencies obtained from simulations and their corresponding experimental ones, before and after PDMS deposition. Table I also contains the estimation values of the mass of the polymer deposited and the effective spring constant k obtained by reversing equation 1. TABLE I COMPARISON BETWEEN SIMULATION AND EXPERIMENTAL RESULTS Parameter Coventor Experimental Simulation Results ........l....... Initial f0 (kHz) 100.443 96.278 Y1 Mddal Disp!acernent Mag. 1M E+QO 25SE-01 5 OE-01 T5 E-01 1 SOE+00fO ihpoy x rn S7 SE-aSS 2SE-ES SUE-Si H 'OETOR f0 with polymer 98.906 87.331 Figure 2b. Coventor mechanical simulation of the first vibration mode of ( f 'kHz) 1.537 9.397 a MEMS resonator with a PDMS layer deposited on top of the plate. A V z) m (kg) 1.03x10-1 1.02X10-9 In our case, PDMS Sylgard 184 polymer layers have been k (N/in) 398.3 1548.5 deposited on the surface of the MEMS cantilevers. Then, an additional stage was included in the virtual description of the fabrication process mentioned above. In the specific case of 978-1 -4244-2839-7/09/$25 .00 (C)2009 IEEE 3 13 Authorized licensed use limited to: UNIVERSITAT POLITÈCNICA DE CATALUNYA. Downloaded on March 16,2010 at 09:16:00 EDT from IEEE Xplore. Restrictions apply. Proceedings of the 2009 Spanish Conference on Electron Devices - Feb 1 1-13, 2009. Santiago de Compostela, Spain. IV. EXPERIMENTAL RESULTS The setup described in Figure 1, formed by the MEMS resonator and two electronic boards that control the analog and digital parts of the PDO system, was used in the experimental measurements. Let us remember that one of the highlights of PDO systems is a built-in analog to digital conversion, but it is also possible to obtain data in analog format. As discussed above, the difference between these two available output formats is that the analog one requires a continuous position sensing mechanism and some extra equipment to monitor the frequency, while the digital option uses the same board to implement the digital feedback loop and to directly extract the _____l digital data. II.. All digital parts of the measurement system have been written in VHDL and programmed on a Cyclone II FPGA from Altera contained in the digital board used, namely a DE2 Figure 4. Oscilloscope screen captures of resonator position (chAt), input development board from Terasic Technologies. Moreover, the pulses (chA3), quantifier-comparator output (chDI), delayed comparator SRAM available on the board was used to temporarily store output (chD2) and sample clock (chDO), for a PDO topology with m = 1 and a sampling frequencyfs =526 kHz. the data, before sending them through an USB link to a personal computer for processing and visualization. Figure 5 shows an example of temporal behavior of the The measures have been done using MEMS resonators PDO oscillation frequency after some changes on the coated with PDMS introduced in a test chamber. Constant 100 concentration of benzene within the test chamber. At first the ml flows of synthetic air alternating with flows of various chamber contains only ambient air, but after 50 s, 400 ml of concentrations of benzene and toluene were introduced into the synthetic air are introduced. This changes the trend of the chamber. In the specific examples described below the 100 ml response curve due to the dehumidification of the PDMS layer: of synthetic air alternated with 100 ml of benzene. it means loss of mass, so the oscillation frequency increases. During the interval between 200 s and 400 s, 200 ml of A. Analog Results benzene mixed with 200 ml of synthetic air have been introduced. We can see there that the polymer increases its The analog output channel directly shows the PDO mass due to the progressive absorption of benzene molecules, oscillation frequencies measured. It comes from an instrutranslating it as a decrease of the oscillation frequency. After mentation amplifier located after the MEMS position sensor that, both the dehumidification mechanism commented above structure and before the transition to the digital domain done and the temperature drift apply (there is no temperature control by a 1-bit quantifier. An Agilent 53131 universal counter, in the experiment), resulting on a frequency increase that leads which periodically sends the data to the personal computer, has to some saturation. been used to monitor the oscillation frequencies. In addition, a mixed analog & digital oscilloscope has been used to check the time behavior of the most relevant signals of the measurement system. Figure 4 shows an oscilloscope 95.9l screenshot example of those analog and digital signals. Analog 95924 channel 1 (chAl) shows the position of the MEMS resonator, Nl while chA2 is a signal which enables generating the pulses to _L -95.923 drive the MEMS resonator (chA3). Concerning the digital channels, chD0 is the sampling clock D signal, while chDl is the synchronized output of the 1-bit 95.922 quantifier, and channels chD2 to chD4 correspond to the same signal, but after a chain of m delays. This m-delay chain is the I digital feedback loop used in the specific PDO systems of this 95.921 work. For the case shown in Figure 4, only one delay (m=1), which corresponds to the channel chD2, was used. 95.920 0 200 400 600 800 1000 Time (s) Eigure 5 'Analog' oscillation frequency transient obtained from the output of the PDO instrumentation amplifier after a frequency meter. 978- 1-4244-2839-7/09/$25.00 (C)2009 IEEE 314 Authorized licensed use limited to: UNIVERSITAT POLITÈCNICA DE CATALUNYA. Downloaded on March 16,2010 at 09:16:00 EDT from IEEE Xplore. Restrictions apply. Proceedings of the 2009 Spanish Conference on Electron Devices - Feb 11-13, 2009. Santiago de Compostela, Spain. B. Digital Results x lo' 9.5869 The PDO digital channel allows obtaining measurements directly from the control circuitry through the bit stream 95869 -.... transmitted by the digital feedback loop. Figure 6 shows the 95868 results obtained from the output of the 1-bit quantifier for the same experimental case as in Figure 5. A sampling frequency 9.5868 of 526 kHz was used and 9000 samples were captured. As it 8 :, can be seen, this 'raw' digital response exhibits a strong quantification noise, but its shape is rather similar to the 95867 analog transient in Figure 5. 95866. X 10V 9,5869 ~956 9.5865.4 ............. 9.5868 <t- <; 4 v ! j | j S - 958840 200 400 600 800 1009 1200 1400 Figure 7 'Filtered digital' oscillation frequency transient obtained from 95867 ..... , ,., the PDO feedback loop samples. Frequencies in Hz, time in s. 9.5866 ACKNOWLEDGMENTS This work was supported by the Spanish government I fj igxi< 2 t 2 g0t< 2 g- 0i g g-0v g Sv -S- l-through a TEC2007-6795 1/MIC project. 9,5864 -L KREFERENCES 0 200 400 600 800 1000 [1] M. Dominguez, J. Pons, J. Ricart, A. Bermejo, E. Figueras, "A Novel Figure 6. 'Raw digital' oscillation frequency transient obtained directly SigmaDelta Pulsed Digital Oscillator (PDO) for MEMS",IEEE Sensors from the PDO feedback loop samples. Frequencies in Hz, time in s. Journal, Vol. 5, pp. 13791388, 2005. [2] M. Dominguez, J. Pons, J. Ricart, A. Bermejo, E. Figueras, "Analysis of the Pulsed Digital Oscillator (PDO) for MEMS", IEEE Trans. on Circuits A well-known advantage of digital formats is the data post- and Systems I, Vol. 52, pp. 22862297, 2005. processing facility. Thus, in order to reduce the quantification [3] J. Ricart, J. Pons, M. Domiinguez, A. Rodriiguez, E. Figueras, M.C. noise of the 'raw' digital data, a raised cosine filter was Horrillo, J. Gutierrez, I. Sayago, "Application of pulsed digital oscillators defined and applied using the Matlab environment. The new to volatile organic compounds sensing", Sensors and Actuators B: 'filtered' digital result is shown in Figure 7. Let us note that Chemical, Vol. 134, pp. 773-779, 2008. the improvement between Figures 6 and 7 is more than [4] M. Domiinguez, J. Pons, J. Ricart, "General Dynamics of Pulsed Digital Oscillators", IEEE Trans. on Circuits and Systems I, Vol. 55, pp.2038 obvious. Moreover, it demonstrates the equivalence between 2050,2008. results in PDOs obtained easily by digital means (Figure 7) [5] V. Raimbault, D. Rebiere, C. Dejous, M. Guirardel, V. Conedera, and the real behavior of the system obtained by using "Acoustic Love wave platform with PDMS microfluidic chip", Sensors stfrom the analog channel (Figure 5) and Actuators A: Physical, Vol. 142, pp. 160-165, 2008. [6] Byung-Su Joo, Jeung-Soo Huh, Duk-Dong Lee, "Fabrication of polymer SAW sensor array to classify chemical warfare agents", Sensors and Actuators B: Chemical, Vol. 121, pp. 47-53, 2007. 978- 1-4244-2839-7/09/$25.00 (C)2009 IEEE 315 Authorized licensed use limited to: UNIVERSITAT POLITÈCNICA DE CATALUNYA. Downloaded on March 16,2010 at 09:16:00 EDT from IEEE Xplore. Restrictions apply.