Sonic Nirvana: MEMS Accelerometers as Acoustic Pickups in Musical Instruments | Sensors Magazine

Sonic Nirvana: MEMS Accelerometers as Acoustic Pickups in Musical Instruments
June 1, 2009By:Alex Khenkin,Kieran Harney,Rob O'Reilly SensorsMEMSsensors are in our cars, phones, PCs, cameras, and a variety of otherapplications but until now, they've never been attached to a guitar.The authors seek to answer the question: Can you use a MEMSaccelerometer as an acoustic transducer?
MEMStechnology builds on the core fabrication infrastructure developed forsilicon integrated circuits. Pressure sensors, one of the firsthigh-volume MEMS applications, now monitor pressure in hundreds ofmillions of engine manifolds and tires; and MEMS accelerometers havebeen used for more than 15 years for airbag deployment, rolloverdetection, and automotive alarm systems.
MEMS accelerometers are used for motion sensing in consumerapplications, such as video games and cell phones while MEMSmicromirror optical actuators are used in overhead projectors, HDTVs,and digital theater presentations. In recent years, MEMS microphoneshave begun to proliferate the broad consumer market, including cellphones, Bluetooth headsets, personal computers, and digital cameras.
This article describes some of the key technologies deployed in MEMSaccelerometer products and discusses how this technology can bring anew dimension to acoustic transducers.
MEMS Accelerometer Technology
The core element of atypical MEMS accelerometer is a moving beam structure composed of twosets of fingers: one set is fixed to a solid ground plane on asubstrate; the other set is attached to a known mass mounted on springsthat can move in response to an applied acceleration. This appliedacceleration (Figure 1) changes the capacitance between the fixed and moving beam fingers.

Figure 1. The structure of a MEMS accelerometer
MEMS structures (Figure 2) are typically formed fromsingle-crystal silicon, or from polysilicon deposited at very hightemperatures on the surface of a single-crystal silicon wafer.Structures with very different mechanical characteristics can becreated; spring stiffness, the mass of the sense element, and thedamping of the structure can all be controlled and varied by designresulting in sensors that can measure fractions of one g or hundreds ofg's with bandwidths as high as 20 kHz.

Figure 2. Micrograph of the ADXL50 MEMS accelerometer's structure
The MEMS sensing element can be connected to the conditioning electronics on the same chip (Figure 3) or on a separate chip (Figure 4).For a single-chip solution, the capacitance of the sense element can beas low as 1–2 fF/g, equating to measurement resolution in the attofarad(aF) range. In a two-chip structure, the capacitance of the MEMSelement must be high enough to overcome the parasitic capacitanceeffects of the bond wires between the MEMS and the conditioning ASIC.
Figure 3. The ADXL202 ±2 g accelerometer  (Click image for larger version)

Figure 4. Cross-section of a typical two-chip accelerometer
Accelerometers as Vibration Measurement Sensors
Theconcept of using vibration sensing transducers as acoustic pickups inmusical instruments is not new. Piezo- and electromagnetic transducersare the basis for many of today's acoustic pickup applications. TinyMEMS accelerometers are so small and low in mass that they have nomechanical or mass loading effect on the instrument, making themattractive for these applications; but to date their use has beenlimited due to the narrow bandwidth of commercially availableacceleration sensors.
Some recent breakthroughs in accelerometer technology have enabledthe production of very small accelerometers with very wide bandwidth. Ahigh-g (±70 g to ±500 g), single-axis accelerometer, such as AnalogDevices' ADXL001 (Figure 5), has 22 kHz bandwidth and comes in a5 × 5 × 2 mm package. This is ideal for monitoring vibration todetermine the state of health of motors and other industrial equipmentby detecting changes in their acoustic characteristics. This particularsensor is not sensitive enough for use as an acoustical vibrationsensor for musical instruments. Also, it only senses along one axis ofmotion, while an ideal acoustic sensor will measure the response alongall three axes. It does demonstrate, however, that full audio bandwidthacceleration transducers can be produced using MEMS technology.

Figure 5. The frequency response curve of the ADXL001
Low-g accelerometers can measure acceleration down to milli g's, butare typically bandwidth-limited around 5 kHz. This limitation may beassociated with the fact that few commercial applications requiresignificant bandwidth (the primary applications involve the detectionof human motion or gravity-driven acceleration), so there has beenlittle motivation to develop sensors suited specifically for audio-bandmeasurement.
A 3-axis accelerometer has three separate outputs that measureacceleration along the Cartesian X, Y, and Z axes. As an example,Analog Devices' ADXL330 3-axis, low-g accelerometer has a bandwidth upto 6 kHz on the X and Y axes, and around 1 kHz on the Z axis. While notideal, this expanded bandwidth allows the part to gather usefulinformation in the audio band. The output is analog, so it can beeasily instrumented and used with standard audio recording equipment.Because its size is less than 4 × 4 × 1.45 mm (Figure 6), thesensor can fit into very small places and it does not cause massloading or other changes in the response of the system being measured.Later we will explore how this low-g accelerometer can be applied as anacoustic pickup for a guitar.

Figure 6. The ADXL330 MEMS accelerometer measures 4 mm × 4 mm × 1.45 mm
Acoustic Feedback
Beginning with the introduction ofomnidirectional condenser and dynamic microphones in the mid-1920s bySøren Larsen, the Danish scientist who first discovered the principlesof audio feedback (known as the Larsen effect), acoustic feedback hasbeen a demon few audio engineers are able to totally control, making itunavoidable in live sound. The Beatles experimented with this audioartifact, then decided to add it to their memorable introduction to "IFeel Fine" in 1964. Rock 'n Roll then set out to tame the beast byembracing it, making acoustic feedback a striking characteristic ofrock music. Electric guitar players such as Pete Townshend and JimiHendrix deliberately induced feedback by holding their guitars close tothe amplifier. As the fad waned, audio engineers continued theirstruggle with acoustic feedback's undesirable ear-shattering effects,particularly in live sound applications. In the perfect world of awell-appointed and acoustically treated recording studio, a high-endomnidirectional microphone will record instruments with an astonishingdegree of realism and fidelity. Artists who know and cherish this soundhave long sought the ability to reproduce it on stage. Althoughrecording a live show with studio sound quality is every musician'sdream, it has been virtually impossible. Even if sound reinforcementrigs sounded good, arenas had excellent acoustics, and sound engineersknew everything there was to know about mixing sound and had the bestgear available, there would still remain one obstacle on the road tosonic nirvana: acoustic feedback.
Acoustic Pickups
Acoustic feedback is typically minimizedby using directional microphones. This works to a certain extent, butrequires constant management by sound engineers to adapt to thechanging characteristics of a stage venue.
Musical instruments can be amplified using pickups. The technologiesvary, but the basic idea is to sense the vibrations of the instrument'sbody directly, rather than the sound it produces in the air. Thepickups generate almost no acoustic feedback, as they are not sensitiveto airborne sound. However, finding a good-sounding location on aninstrument body is notoriously difficult, the sonic characteristics ofpiezo pickups are far from perfect, and their high output impedancerequires special instrument inputs or direct boxes. In addition, theycan be large and can interfere with the natural acoustic behavior ofthe instrument.
This leads to the idea of a low-mass contact microphone. Supposethat we use a surface transducer that measures the acceleration of theinstrument's body, preferably on more than one axis. This transducerwould have good linearity and be so lightweight that it would notacoustically affect the instrument being measured. Suppose further thatthe transducer has similar output level, output impedance, and powerrequirements to a traditional microphone. In short, suppose that amusician could just plug this transducer into a microphone preamp ormixer input, just like any other microphone.
Contact Microphones
An attentive reader will notice themention of acceleration in the preceding paragraph. Our ears respond tosound pressure, so microphones are designed to sense sound pressure. Tosimplify matters greatly, the sound pressure in the immediate vicinityof a vibrating body is proportional to acceleration. What if anaccelerometer had enough bandwidth to be used as a contact microphone?
To explore this concept, we mounted a 3-axis accelerometer on anacoustic guitar to act as a pickup. The vibration of the instrument wasmeasured and compared to the built-in piezo pickup and to a MEMSmicrophone mounted near the guitar. The guitar used was a FenderStratacoustic acoustic with a built-in Fender pickup. An analog outputMEMS accelerometer was mounted on a lightweight flex circuit andattached to the guitar body using beeswax at the bridge location, asshown in Figure 7. The X-axis of the accelerometer was orientedalong the axis of the strings, the Y-axis was perpendicular to thestrings, and the Z-axis was normal to the surface of the guitar. A MEMSmicrophone with a flat frequency response out to 15 kHz was mounted 3in. from the strings for use as a reference.

Figure 7. Accelerometer mounted on a Fender Stratacoustic Acoustic Guitar
A short sound segment was recorded using the accelerometer, thebuilt-in piezo pickup, and the MEMS microphone. The time domainwaveforms for each transducer are shown in Figure 8. No postprocessing was done on any of the audio clips.

Figure 8. Time domain waveforms using different transducers
Figure 9 shows an FFT-based spectrum of the piezo pickupmeasured at one of the peaks in the time domain waveform. This spectrumshows a response with a strong bass component. Indeed, the actual audiofile sounded excessively full, with a lot of bass response. This soundspleasing (depending on your taste) as the cavity resonance creates afuller bass sound than that heard when listening to the instrumentdirectly.

Figure 9. FFT spectrum of piezo pickup
The MEMS microphone output is very flat and reproduces the sound ofthe instrument very well. It sounds very natural, well balanced, andtrue to life. The FFT-based spectrum measured at the same point in timeas the piezo pickup is shown in Figure 10A. The frequency response of the MEMS microphone is shown in Figure 10B for reference.
Figure 10. The FFT-based spectrum of a MEMS microphone (A) and its frequency response (B)  (Click image for larger version)
The output from the MEMS accelerometer is very interesting. Theimmediate weak points are that the noise floor was too high and audibleat the beginning and end of the track, and that the bandwidth of the Zaxis was clearly limited to lower frequencies. The sound reproductionfrom each axis was noticeably different.
The X and Y axes sounded bright and articulate and had clearlydiscernable differences in tonality. As expected the Z axis obviouslysounded bass dominated. Figure 11 shows the X-axis spectrum (A), the Y-axis spectrum (B), and the Z-axis spectrum (C).
Figure 11. MEMS accelerometer output showing the spectrum of the X-axis (A), Y-axis (B), and Z-axis (C)  (Click image for larger version)
The X, Y, and Z axes mixed together produced a fair representationof the instrument with some brightness. By adjusting the mix, avariation in tonal balance can be achieved with natural soundreproduction. The extended upper harmonics are still missing due to thebandwidth limitation of the current accelerometers, but the soundreproduction was still surprisingly true.
Conclusion
Low-g MEMS accelerometers do not suffer fromtraditional feedback problems, and demonstrate clear potential ashigh-quality acoustic pickups for musical instruments. A 3-axisaccelerometer mounted on a Fender Stratacoustic acoustic guitarachieved promising sound reproduction. The three axes have differenttonal characteristics related to the vibration modes of the instrumentin the different directions of the body, however the three outputchannels can be mixed to generate realistic sound reproduction. Inaddition, these channels can be mixed in different ways to producecreative tonal effects.
While the performance of the accelerometer in this experiment isvery promising, there are a few drawbacks. The noise floor of thesensor is audible, a problem that can be minimized using noise gatingor other techniques, but the ideal sensor will have a noise floorcomparable to conventional microphones. The high-frequency response ofthe sensor needs to be extended, ideally up to 20 kHz to capture thefull tonal range of the instrument.
MEMS accelerometer technology has clear potential for acousticpickup applications in musical instruments, especially in liveperformances where acoustic feedback could be a problem. A very small,low-power MEMS device can be mounted unobtrusively anywhere on theinstrument without affecting its natural vibration characteristics. Infact, multiple sensors can be mounted at different points around theinstrument to provide the sound engineer with additional flexibility toreproduce the natural character of the instrument without fear ofacoustic feedback in live sound applications—one step closer to "SonicNirvana."