Proceedings of Meetings on Acoustics

Transcription

1 Proceedings of Meetings on Acoustics Volume 19, ICA 2013 Montreal Montreal, Canada 2-7 June 2013 Engineering Acoustics Session 4aEAb: Acoustics for Navigation 4aEAb1. Ultrasonic transducers for navigation Berhard E. Boser*, Richard J. Przybyla, David A. Horsley, Stefon E. Shelton and André Guedes *Corresponding author's address: Berkeley Sensor and Actuator Center, University of California, Berkeley, CA 94709, Free space ultrasonic ranging is attractive for applications such as gesture recognition and robotic navigation. Unlike optical ranging technologies, ultrasound based solutions are insensitive to ambient illumination and can therefore be used in- and outdoors. Using time-offlight, ultrasound rangers work over distances of up to a few meters and achieve sub-mm resolution. Using arrays, objects can be localized in three dimensions. Transducers consist of 400μm Aluminum-Nitride membranes sandwiched between actuation electrodes batch fabricated on Silicon wafers. Unlike capacitive transducers which require actuation voltages in excess of 100V, piezoelectric devices are compatible with lowvoltage actuation. At the 200kHz resonance frequency, the wavelength at atmospheric pressure is 2mm, ideal for compact arrays. The transducers do not dissipate static power and are therefore ideal for battery powered applications. Energy consumption is dominated by the lownoise readout amplifier and is on the order of 1μJ per channel including analog-digital conversion and signal processing, enabling video-rate object tacking at less than 1mW power dissipation. A prototype system consisting of seven transducers on a 1mm grid operates up to a 750mm range and ±35o angle span with ±3.5mm accuracy and ±3o worst case angle error. Published by the Acoustical Society of America through the American Institute of Physics 2013 Acoustical Society of America [DOI: / ] Received 18 Jan 2013; published 2 Jun 2013 Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 1

2 INTRODUCTION The emergence of hand-held battery powered devices such as smart phones creates a pressing need for novel user interfaces. A wide array of technologies for augmenting presently dominant touch interfaces are under consideration. Touch-free 3D gesture recognition [1, 2] in particular is receiving a lot of attention from industry and academia with a number of different technologies begin pursued, including both active and passive optical 3D cameras [3] and electrical field sensors [4]. In this paper we demonstrate gesture recognition based on ultrasonic transducers. Ultrasonic transduction already finds use in a wide variety of applications including medical imaging, non-destructive sample evaluation, echolocation, and proximity detection. Compared to optical techniques, ultrasound has a number of advantages. Independent of ambient illumination, ultrasonic gesture interfaces work both indoors and outdoors or in the dark. Low power dissipation is an enabling advantage of ultrasonic technology: unlike active optical solutions with typical power dissipation of several Watts [5], ultrasonic transducers operate off micro-watts, enabling operation even during standby with negligible battery drain. Thanks to the wavelength of sonic waves in air at the millimeter scale, phased array imaging can be employed to scale the device to the chip-scale while still covering a wide field of view and distances up to at least a meter [6, 7]. ULTRASONIC TRANSDUCERS Conventional technology largely relies on bulk piezoelectric ceramic materials which suffer from poor acoustic coupling to air and are expensive to machine into two-dimensional (2D) transducer arrays. In comparison, micromachined ultrasonic transducers (MUTs) are compliant membrane structures fabricated using integrated circuit (IC) manufacturing technology, allowing compact 2D arrays to be realized and offering the potential for integration with signal processing electronics [?]. Here we describe piezoelectric micromachined ultrasonic transducers (pmuts) fabricated using aluminum nitride (AlN). Compared to capacitive micromachined ultrasonic transducers (cmuts), pmuts have lower electromechanical coupling but do not need the high polarization voltages (approaching 1000 V) and small capacitive gaps required by cmuts. There is a long history of research on piezoelectric micro-electromechanical systems (MEMS) including pmuts composed of zinc oxide (ZnO) and lead zirconate titanate (PZT). However, with the exception of ink-jet print-heads, piezoelectric materials have seen little use in commercial MEMS devices until the recent success of AlN thin-film bulk acoustic wave RF filters. AlN is attractive because it is compatible with standard CMOS technology, allowing monolithic integration of MEMS transducers and circuitry. Although the integration of ZnO MEMS devices with circuitry has been successfully demonstrated, ZnO films are higher in conductivity than AlN, resulting in power loss. Additionally, the fact that Zn is a fast-diffusing ion may result pose contamination issues for CMOS manufacturing. In comparison with PZT, the lower piezoelectric coefficients of AlN are mitigated by a significantly reduced dielectric constant. The reduced capacitance of AlN PMUTs in comparison with earlier PZT pmuts can result in improved signal-to-noise ratio (SNR) [9]. Individual ultrasound transducer elements consist of a unimorph membrane with diameter 400μm composed of an SiO 2 /Pt/AlN/Al sandwich fabricated on a Si handling wafer. Figure 1 shows a cross-section with a trench etched though the wafer to expose both sides of the membrane. The electrical field resulting from a voltage applied between the Al and Pt electrodes results in a transverse stress in the AlN layer and consequent out-of-plane bending of the membrane which produces a pressure wave. Similarly, an incident pressure wave results in Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 2

3 FIGURE 1: Cross-section of the pmut. membrane deformation and consequent charge on the electrodes enabling the device to be used both as a transmitter and receiver. FIGURE 2: Equivalent electrical model of the pmut transducer. For small displacements of less than approximately 0.6 μm the membrane behaves like a linear resonator. Figure 2 shows an electrical equivalent model. In this diagram, capacitor C m models the equivalent membrane stiffness, L m the mass, and R m the loss to the substrate. The impedance Z a represents the air surrounding the transducer with the resistive part R a modeling the acoustic power delivered to or received from the air. For good coupling, the values of R m and R a should be comparable. In this design, R a = 69kΩ and R m = 385kΩ. FIGURE 3: Measured and calculated sound pressure level (SPL) versus normalized frequency. Figure 3 shows the measured sound pressure level of the transducer as a function of frequency at 5 mm distance when the transducer is driven by a 12V pp sinusoid. The resonant frequency and quality factor of the device are f o = 200kHz and Q = 20, respectively. The shunt capacitance C o 14pF is dominated by device capacitance and wiring. It should be minimized to avoid signal attenuation when the device is used as a receiver. Although the attenuation can be compensated by electronic amplification, this process invariably elevates the electronic noise floor, resulting in either more stringent amplifier noise specifications and consequent higher power dissipation, or SNR degradation [10]. At f o = 200kHz the acoustic wavelength in air is λ 1.7mm. Phased array beam steering requires an array of individual transducer elements with approximately half wavelength spacing. Figure 4 shows a micrograph of the array used by the ranger presented in the next Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 3

4 FIGURE 4: Micrograph of two-dimensional array of individual transducer elements. section. PHASED ARRAY TRANSDUCER FOR GESTURE RECOGNITION In this section we describe the electronic interface and signal processing used for 2D ultrasonic imaging with the device shown in Figure 4. In this demonstration, a single array element is used as transmitter [11] and a 1D array consisting of seven elements serves as receiver. For each measurement, the system first emits a short pulse and then records the echo at the receiver elements. Distance is extracted from the time-of-flight of the echo. The relative arrival time of the echoes at the individual receiver elements is the basis for estimating the angle of the target relative to the transducer surface, thus facilitating target localization in two dimensions. The system operates at distances up to a 800 mm and ±35 degree target angles with ±3.3mm range and ±2.3 degree angle errors, respectively. Pulsed time-of-flight measurement has been chosen to combat multi-path fading and excessive receiver dynamic range requirements characteristic of continuous time solutions. At 1 m range, the round trip delay of the echo is approximately 5.9 ms and sets a maximum operating frequency of 170 Hz. An ideal system would transmit a narrow Dirac pulses. Short duration permits resolving closely spaced targets, while fast rise time is critical to minimize timing errors due to amplitude fluctuations. Consequently it is desirable to use all of the available bandwidth. In practice, this means the bandwidth of the system should be designed to be limited by the bandwidth of the ultrasound transducer. The ideal transmit signal is approximated in practice with a burst of a sinusoidal signal at the resonant frequency of the transmitter element. The choice of the length T p of the burst is governed by a tradeoff between minimum and maximum target range. Since the receiver is disabled during transmission, the minimum target distance is R min = ct p /2. However, bursts much shorter than Q/(2f o ) 80ms, corresponding to R min 14mm, result in reduced transmit amplitude and hence smaller maximum target range. Dynamic adaptation of T p can be employed to avoid this tradeoff for clustered targets. An amplitude threshold is used in the detector to discriminate echoes from noise. The threshold must be set sufficiently low to capture low amplitude echoes at the maximum range but to reject noise which results in phantom targets. A minimum signal-to-noise ratio of 12 db ensures occurrence of false positives no more often than once every 24 minutes [12], an Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 4

5 FIGURE 5: Target angle estimation based on path length difference. acceptable value for most applications. Two effects contribute to the ranging error. Amplitude noise and timing jitter as well as variations of the speed of sound degrade the accuracy of the time-of-flight measurement. Secondly, the finite pulse duration results in overlap of the transmit pulse and echo and limits the separation of targets at similar distance from the transducer. For this design, the rms ranging error is σ r = 3.3mm at 800 mm target range. The minimum spatial separation at which targets can be resolved individually is 21 mm. FIGURE 6: Block diagram of the phased array receiver. Difference in arrival time of the echo at individual transducer elements is used to determine the target angle. Figure 5 depicts a linear array of receivers used to receive a plane wave returning from a target. An incident angle Θ results in an arrival time difference ΔT = d/c sin(θ) between neighboring elements. As for target range, noise limits the accuracy of the angle estimation, while the number of receiver elements sets the minimum angular separation for targets to be distinguished. For this design, the rms angle error is 2.3 degrees at maximum range and the minimum separation is 14.5 degrees. Figure 6 shows the block diagram of the prototype 2D ultrasonic ranger implemented with COTS electronics. The seven element 1D receiver array captures the echoes which are amplified, digitized, and quadrature down-converted for each channel separately. For each angle Θ, the contributions from each channel k are shifted by a phase shift kφ = 2πnd/λsin(Θ) and summed. The angle is swept over the entire angle range, resulting in an image such as the one shown in Figure 7. This data has been found to be adequate for gesture recognition applications such as advancing and zooming images in a slide show. Power dissipation in this design is dominated by the general purpose COTS DSP. In a customized ASIC implementation digital circuit power dissipation could be reduced substantially resulting in the analog interface dominating dissipation. With power-gated amplifiers in the receiver, the energy required by the readout circuits would be approximately 2 μj per channel and measurement. Analog-to-digital conversion would approximately double Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 5

6 FIGURE 7: Sample 2D image recorded with the ultrasonic transceiver. this figure. Transmit power, dominated by C o V 2 losses, is negligible. These figures correspond to an estimated power dissipation of less than 20 μw for 1 Hz sampling rate, suitable for standby operation e.g. to control the power state (on/standby) of a device. At 30 Hz video rate, the estimated power dissipation increases to about 500 μw still well within the power budget of hand-held devices. CONCLUSION Microfabricated piezoelectric ultrasonic transducers are an attractive option for gesture recognition for example in battery operated hand-held devices or as alternatives to capacitive touch-screen interfaces. Owing to their small size, the transducer can be integrated even in devices such as remote controllers that are too small to accommodate a display. The technology operates at ranges up to at least one meter and dissipates less than 1 mw at video rates and only tens of microwatts at reduced sampling rate, enabling operation even in standby mode, e.g. to control turning a device on or off. Unlike optical gesture recognition, ultrasonic solutions are not affected by ambient light and work equally well indoors and outdoors. REFERENCES [1] ABIresearch, 600 Million Smartphones Will Have Vision-Based Gesture Recognition Features in 2017, London, July [2] Forbes, Texas Instruments Sees Big Market For Smartphone Gesture Recognition, October [3] D-I. Ko and G. Agarwal, Gesture recognition: Enabling natural interactions with electronics, Texas Instruments White Paper, [4] J. Leber, A New Chip to Bring 3-D Gesture Control to Smartphones, MIT Technology Review, November [5] A. Yoon, Kinect teardown: two cameras, four microphones, 12 Watts of power, no controller, MIT Technology Review, November [6] R. Przybyla, A. Flynn, V. Jain, S. Shelton, A. Guedes, I. Izyumin, D. Horsley, and B. Boser, A micromechanical ultrasonic distance sensor with >1 meter range, in TRANSDUCERS th International Solid-State Sensors, Actuators and Microsystems Conference, 2011, pp Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 6

7 [7] R. J. Przybyla, S. E. Shelton, A. Guedes, I. I. Izyumin, M. H. Kline, D. A. Horsley, and B. E. Boser, In-air rangefinding with an aln piezoelectric micromachined ultrasound transducer, IEEE Sensors Journal, vol. 11, no. 11, pp , [8] S. Shelton, C. Mei-Lin, P. Hyunkyu, D. Horsley, B. Boser, I. Izyumin, R. Przybyla, T. Frey, M. Judy, K. Nunan, F. Sammoura, and Y. Ken, CMOS-compatible AlN piezoelectric micromachined ultrasonic transducers, in 2009 IEEE International Ultrasonics Symposium, 2009, pp [9] A. Guedes, S. Shelton, R. Przybyla, I. Izyumin, B. Boser, and D. A. Horsley, Aluminum nitride pmut based on a flexurally-suspended membrane, in TRANSDUCERS th International Solid-State Sensors, Actuators and Microsystems Conference, 2011, pp [10] B. E. Boser and R. T. Howe, Surface micromachined accelerometers, IEEE Journal of Solid-State Circuits, vol. 31, no. 3, pp , [11] R. Przybyla, S. Shelton, A. Guedes, R. Kriegel, D. Horsley, and B. Boser, Automatic mode matching for high-q vibratory gyroscopes, in Technical Digest Solid-State Sensor and Actuator Workshop, Hilton Head Island, Georgia, [12] M. Skolnik, Introduction to Radar Systems. 3 rd edition, McGraw-Hill, Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 7