MEMS for Distributed Wireless Sensor Networks

Brett A . Wurneke, Krisiofer S.J. Pister

Berkeley Sensor and Actuator Center, University of California at Berkeley

ABSTRACT

MEMS technology is enabling the development of

inexpensive, autonomous wireless sensor nodes with
volumes ranging from cubic mm to several cubic cm. These
tiny sensor nodes can form rapidly deployed, massive
distributed networks to allow unobtrusive, spatially dense,
sensing and communication. MEMS enable these devices
by reducing both the volume and energy consumption of
various components. This paper will review some of the
wireless sensor nodes under development and applicable
MEMS devices for small and efficient optical
communication, micropower generation, and sensing. In
addition, CMOS post-process micromachining will be
discussed as a method of achieving low cost and high
integration.
Figure 1: Conceptual Diagram of a Smart Dust mote, one
example of a tiny, autonomous, wireless sensor node.

microelectronics is also a major contributor to the realization

1. INTRODUCTION of tiny wireless sensor nodes by increasing the amount of
functionality that can be packed into a given area, by
Over the course of the last several decades, technological enhancing levels of circuit integration with mixed-signal
advances have promoted exponential decreases in the size of design, and by reducing circuit power consumption.
computational functionality, rapid growth in the numbcr of However, this paper will focus on the contribution of
networked devices, escalating numbers of wireless devices, MEMS to wireless sensor nodes.
diminishing wireless device sizes, enhanced battery life, This paper will begin with a description of some
brisk reduction in sensor sizes, and a continual decline in microfabrication processes that provide high levels of
cost. The convergence of these trends is expected to result integration, which is beneficial to building small systems,
in “ubiquitous computing” [I]. One manifestation of then discuss the primary components of these nodes:
ubiquitous computing will be small, inexpensive, sensors, communication, and micropower sources. Finally,
autonomous, wireless sensor nodes (Figure 1) that can be some example sensor systems will be reviewed, with
rapidly deployed in vast numbers throughout the particular emphasis on the authors’ own work on the Smart
environment. Because of the discreet size, substantial Dust project, which has achieved a 16 mm3 node.
functionality, connectivity, and low cost, these devices will
enable completely new methods of interacting with the
environment, providing more information from more places
2. HIGHLY INTEGRATED PROCESSES
in a less intrusive way than before. Application areas
Shrinking an entire autonomous system into volumes
include defense networks that could be rapidly deployed by
unmanned aerial vehicles (UAV), tracking the movements of between a cubic mm and a cubic cm can be assisted by
birds, small animals, and even insects, fingertip utilizing highly integrated processes that allow disparate
accelerometer virtual keyboards, monitoring environmental components to be fabricated on a common substrate.
conditions that affect crops and livestock, inventory control, Frequently, such a process would combine CMOS
product quality monitoring, smart office spaces, and transistors with some micromachining capabilities. One
interfaces for the disabled. approach to this that has been successfully commercialized
The primary facilitators for the aforementioned trends by Analog Devices [Z] begins with a standard BiCMOS
are advances in microfabrication technology, both in process, then a 4pm low-stress polysilicon structural layer is
inserted into the process before the interconnect metakation
microelectronics and microelectromechanical systems
(MEMS). MEMS technology in particular and more is deposited. An extra mask at the end protects the oxide
recently is allowing a dramatic reduction in the size and over the circuits during the sacrificial oxide release etch.
power consumption of sensors, actuators, communication A similar approach developed by Sandia National
components, and power systems. Progress in Laboratory [3] forms the microstructures before the CMOS
by fabricating the polysilicon structures in a trench in the

0-7803-7596-3/02/$17.00 02002 IEEE 29 1

Metal 1 Metal 2 different process. These die are presently hand-assembled,
hut production would utilize microassembly techniques.
Additionally, the next generation device will pack the same
functionality into two die: a CMOS ASIC and a highly
integrated custom process that yields solar cells, high
voltage FETs, and micromachined structures.
Figure 2: Cross-sectional diagram of an etch hole in a
standard CMOS process that h a been post-processed with 3. SENSORS
an isotropic etchant. Contact, via, and overglass cut layers
are stacked to yield the oxide hole to the substrate.
-
Standard intearated circuit Drocesses can inherentlv sense a
few phenomenon, such as light and temperature, in a
silicon substrate. The trench is refilled with sacrificial oxide
compact area, but micromachining has allowed researchers
then the CMOS process is run. The main advantages of this
to shrink many types of sensors into small volumes while
technique are that the CMOS process does not need
often maintaining similar, or even exceeding, performance
substantial modification and its thermal budget is unaffected
levels of conventional transducers [16],[17]. Examples
by the polysilicon deposition.
include thermal sensors [I I], accelerometers [I 81,
Despite being based on standard CMOS, both of these
integrated processes require control over the fabrication flow gyroscopes [2], pressure sensors [19], microphones [20],
radiation detectors, magnetic sensors, flow sensors, and
to add the micromechanical layers. A number of approaches
have been developed to do post-process micromachining on chemical and biological sensors. Several factors need to he
considered when selecting sensors for use in tiny wireless
foundry CMOS. Franke, et al. [4] demonstrated post-CMOS
modular integration of poly-SiGe microstructures with poly- sensor nodes: volume; power consumption; suitableness to
power cycling, as this is a primary means of reducing the
Ge sacrificial layers. LPCVD poly-Ge and poly-SiGe can be
energy consumption, hut some sensors require more time to
deposited at such a low temperature that the CMOS
generate a stable measurement than they would he powered
aluminum interconnects are not damaged. Furthermore,
hydrogen peroxide etches poly-Ge without attacking the off; fabrication and assembly compatibility with other
CMOS layers, which simplifies the process. components of the system; and packaging needs, as sensors
that require contact with the environment, such as chemicals,
An even simpler method allows micromachining to he
performed with a single maskless post-processing step [5] add significant packaging considerations.
on foundry CMOS wafers or die, thus requiring very little
extra equipment. This technique relies on the ability to stack 4. COMMUNICATION
the contact, via, and overglass cut layers (Figure 2) to yield a
‘pit’ of exposed silicon when the chip returns from the Certainly, the most common form of wireless
foundry, which can then he sacrificially etched by hulk Si communication in use today is RFlmicrowavelmillimeter
etchants such as xenon difluoride (XeF2) [ 6 ] , wave radiation. While developments in MEMS [21] are
tetramethylammonium hydroxide (TMAH), and reducing the size of RF components and allowing tighter
ethylenediamine-pyrocatechol (EDP) [7] with the oxide circuit integration, there are certain limitations that make it
acting as the mask and structural layer. Each of these etch less efficient in tiny, energy-constrained devices. First,
methods require relatively inexpensive equipment and antennas need to be a significant fraction of a wavelength to
infrastructure, making this a viable option for designers he at all efficient, so at RF frequencies antennas are
without a fabrication facility. This technique has been used relatively long. A second issue is that a small RF antenna
to create a wide variety of devices including accelerometers will have very low antenna gain, as beam divergence is
[SI,microwave power sensors [7], higher-Q spiral inductors fundamentally limited by diffraction, which is dependent on
[SI, heart-cell transducers [lo], thermal sensors [ I l l , and wavelength. To achieve the same milliradian collimation of
thermally isolated circuits [12]. Although this method does an inexpensive laser pointer at IGHz would require a lOOm
not work in sub-micron processes that use tungsten plugs in antenna. A third problem with RF communication is that the
the vias, it is still useful in the older, less expensive transceivers have poor overall efficiency because there is
processes. usually I-100mW of overhead due to the mixers, biasing,
Another maskless post-CMOS process that does work etc.; however, researchers are working to improve these
in sub-micron processes uses the top metal layer as the mask efficiencies and build IOOpW radios [22]. A fourth
[13]. A high aspect ratio reactive ion etch (RIE) removes disadvantage of RF communication is that the received
the oxide not protected by metal, leaving near vertical power is = lid2+’ due to multi-path fading. For ground-
sidewalls. Next, a plasma etch removes the silicon substrate based communication, such as mobile telephone, the average
from beneath the microstructures. This approach has the factor is four. These reasons combine to make RF
advantage of providing high-aspect ratio structures, but it communication unattractive for tiny wireless nodes. For
requires expensive processing equipment. example, the Bluetooth radio standard, which was designed
An additional technology area that can help shrink an for relatively low power consumption devices, costs about
autonomous system is microassembly [14]. Pick and place IOOnJibit to communicate over 10s of meters. Nevertheless,
methods for the microdomain, microtweezers [IS], hatch since RF radiation can propagate through a wide variety of
transfer, fluidic microassembly, flip-chip bonding, and many materials, thus not requiring line-of-sight, wireless nodes
other techniques are facilitating the compact assembly of designed for certain applications will utilize it. One notable
heterogeneous dies. For instance, the current Smart Dust point of correspondence between RF MEMS components,
mote, shown in Figure 3, contains four die, each from a particularly resonators, and tiny wireless nodes is that these

292
 requires the generation of radiation, such as lasers or RF, but
it does not facilitate peer-to-peer communication. In such a
network the nodes are unlikely to talk to each other,
although distributed compression algorithms are being
developed to take advantage of such a network. Because a
single base-station will simultaneously communicate with
potentially thousands of devices, synchronous
communication can be used to lock the data sent by each
node to a master clock signal from the interrogator.
Active-steered laser communication is currently under
development. It would utilize an onboard light source, such
as a VCSEL, a collimating lens, and MEMS beam-steering
optics [26],[27] to send a tightly collimated light beam
toward an intended receiver, thus facilitating peer-to-peer
communication.

Figure 3: 16mm3 mock-up of the current Smart Dust 5. MICROPOWER SOURCES

system, showing a 0.25pm CMOS ASIC, solar power array,
accelerometer and CCR, each on separate die. The Wireless sensor nodes will likely utilize a combination of
functional device does not have the accelerometer. energy storage and energy scavenging devices. A number of
micromachined microbatteries have been reported [28],
components often have relatively low power handling including some that achieve an energy density of 5.6 Umm’
capabilities that render them unsuitable for many RF [29], which is very competitive with macro-scale batteries.
applications; however, these limits are high enough to be Scavenging energy from the environment will allow the
applicable to distributed wireless sensor networks that utilize wireless sensor nodes to operate nearly indefinitely, without
short-range, multi-hop communication links. their battery dying. Solar radiation is the most abundant
Free-space optical communication provides an energy source and yields around 1 mWlmm’ (1 Jldaylmm’)
attractive wireless communications technique for minute in full sunlight or 1 pW1mm’ under bright indoor
sensor nodes (Figure 3). First, optical radiators such as laser illumination. Solar cells have conversion efficiencies up to
diodes and mirrors can be made extremely small - 0.03pm’ 30% and are a well-established technology, making them
lasers have been demonstrated [23]. As mentioned above, attractive for early use in sensor nodes. Vibration harvesting
optical communication also provides extremely high antenna [30] is another potential energy source, scavenging energy
gain, which yields higher transmission efficiencies. Thirdly, from the vibrations of copy machines, ventilation systems,
while laser output slope efficiencies are only about 25%, the etc. More exotic energy sources include utilizing the excess
overhead (diode turn-on current) can be as low as IpW for heat from micro rocket engine [31] combustion and micro
vertical cavity surface emitting lasers (VCSELs), so the radioactive sources.
overall output efficiency can be much higher than RF power
amplifiers. A fourth advantage of optical communication is 6. SYSTEMS
that the received power only decays as the inverse of
distance squared, assuming line-of-sight communication. A number of small, MEMS-based wireless sensor node
Finally, optical communication enables the use of spatial systems have been developed. Mason, et a/. [32] at the
division multiple access (SDMA) [24]. lhis is a simple University of Michigan built a multisensor microcluster that
technique wherein an imaging receiver can separately measures pressure, temperature, humidity, and
process simultaneous transmissions from different angles. vibratiodposition, includes a microcomputer, and has a 50m
As such, SDMA requires no communication overhead and RF link. The device is less than IO cm3 and consumes
thus has the potential to be more energy-efficient than the 530pW average power and lOmW while transmitting.
RF media access methods such as frequency, time, and code Asada, et a/. [33] at UCLA developed wireless integrated
division multiple access (FDMA, TDMA, CDMA). network sensors (WINS) that included an infrared imager,
We have explored two approaches to free-space optical seismometer, spectrum analyzer, RF transceiver, and lithium
transmission: passive reflective systems and active-steered coin cells in a volume on the order of tens of cubic inches.
laser systems. The passive reflective system consists of Rabaey, et al. [22] at UC Berkeley are researching RF data
three mutually orthogonal mirrors that form the corner of a nodes with less than 5nJibit communication energy in a
cube (Figure 4), hence the name comer cube retroreflector cubic centimeter volume.
(CCR). Light entering the CCR bounces off each of the The authors’ own Smart Dust project [34] seeks to
mirrors and is reflected back parallel to the incident beam. aggressively push the volume of wireless sensor nodes down
By electrostatically actuating the bottom mirror, the to a cubic millimeter (Figure I), which also puts extreme
orthogonality can be broken, causing less light to return to constraints on the energy consumption. The current device
the sender. The CCR can thus communicate with an [35] is a 16mm’ autonomous solar-powered sensor node
interrogator by modulating the reflected light, with the only with bi-directional optical communication. The device
energy consumption being the charging of about 3pF in the digitizes integrated sensor signals and transmits and receives
actuator and a demonstrated range of 180m [25]. This data optically. The system consists of three die-a 0.25pm
technique consumes much less power than an approach that CMOS ASIC, a trench-isolation SO1 solar cell array, and a

293
 micromachined four-quadrant CCR. The ASIC contains an
optical receiver that consumes 69 pJhit, an ADC that uses [I41 M.B. Cohn, et al., “Microassembly technologies for
360 pJl8-hit sample, a photosensor for measuring ambient MEMS,” Proc. SPIE, ~01.351I, Micromachining and
light, a finite state machine to control the system, and a Microfabrication Process techno log^^ JV, Santa Clara, pp.2-I 6,
IpW, 3.9MHz integrated oscillator. The CCR consumes 21-22 Sept. 1998.
about 16 pJibit-transmitted. In a new DRIE SOliCMOS [I51 C . Keller, Microfabricated High Aspect Ratio
process, we have demonstrated working solar cells, CCRs, Silicon Flexures, MEMS Precision Instruments, 1998.
and a capacitive accelerometer on a single chip. Once this [I61 Lj. Ristic [ed], Sensor Technology and Devices, Artech
process is fully developed, the sensor node will shrink to 6.6 House, London, 1994.
mm’. Another forthcoming development is the addition of [I71 G.T.A. Kovacs, Micromachined Transduceers Sourcebook,
an ultra-low power custom microprocessor and SRAM to WCB McGraw-Hill, San Francisco, 1998.
provide greatly enhanced functionality, including laser [I81 N. Yazdi, F. Ayazi, K. Najafi, “Micromachined inertial
reprogramming, extreme power cycling, hardware sensors,”Proc. IEEE, vo1.86, (no.8), Aug. 1998, pp.1640-59.
implementation of common tasks for reduced power [ 191 W.P. Eaton, J.H. Smith, “Micromachined pressure sensors:
review and recent developments,” Smart Materials and
consumption, and special ADC modes. Assembly and
packaging are challenges that remain largely unaddressed. Structures, vol.6, (no.5), Oct. 1997. pp.530-9.
[201 P. Rombach, et al., “The first low voltage, low noise
differential silicon microphone, technology development and
8. CONCLUSION measurement results,” Sensors and Actuators A (Physical),
vol.A95, (no.2-3), I Jan. 2002, pp.196-201.
Technological trends of smaller devices with more [21] C.T.-C. Nguyen, L.P.B. Katehi, G.M. Reheiz,
functionality, greater connectivity, and more sensors are “Micromachined devices for wireless communications,” Proc.
leading to small, inexpensive, autonomous wireless sensor IEEE, vo1.86, (no.8), Aug. 1998,pp.1756-68.
nodes that will form distributed networks. MEMS are an [22] J. Rabaey, et al., “PicoRadios for Wireless Sensor
enabling technology through smaller sensors, Networks: The Next Challenge in Ultra-Low-Power Design”,
communication components, and micropower sources. IEEE ISSCC, San Francisco, Feb. 2002, pp. 200-201.
[23] 0. Painter, et al., “Two-dimensional photonic band-gap
defect mode laser,” Science, vol.284, (no.5421). 1 I Junc 1999,
9. REFERENCES p. 1819-21.
[24] J.M. Kahn, et al., “Imaging Diversity Receivers for High-
[I] M. Weiser, et al., “The Coming Age of Calm Technology,” Speed Infrared Wireless Communication,” IEEE Comm., Dec.
http:il~.ubiq.co~ihypertext/weiserlacm~tu~e2endnote.htm, 1998, pp. 88-94.
1996. [25] L. Zhou, K.S.J. Pister, J.M. Kahn, “Assembled Comer-
[2] J.A. Geen, et al., “Single-Chip Surface-Micromachined cube Retroreflector Quadruplet”, MEMS 2002, Las Vegas,
lntegrated Gyroscope with 5O“/hour Root Allan Variance,” Nevada, 20-24 Jan. 2002, p.556-559.
2002 IEEE ISSCC., vo1.45, San Francisco, Feb. 2002, p. 426-7. [26] M. Last, et al., “An 8 mm’ digitally steered laser beam
[3] J.J. Sniegowski, et al., “IC-Compatible Polysilicon Surface transmitter,” 2000 IEEE/LEOS Inf ’1 Con/: Optical MEMS,
Micromachining,” Annuol Rex Mat. Sci. 2000. 30:299-333. Kauai, HI, p. 69-70.
[4] A.E. Franke, et al., “Integrated MEMS technologies,” MRS [27] V. Milanovic, M. Last, K.S.J. Pister, “Torsional
Bulletin, vo1.26, (110.4). Mater. Res. Soc. April 2001, p.291-5. Micromirrors with Lateral Achlators,I1 Trarducers’Of -
[5] M. Parameswaran, et al., “A new approach for the Eurosensors XV Conference, Muenchen, Germany, Jun. 2001.
fabrication of micromechanical structures,” Sensors and [28] K.B. Lee, L. Lin, “Electrolyte Based On-Demand and
Actuators, 19 (1989) 289-307. Disposable Microbahery,” MEMS 2002, Las Vegas, Nevada,
[61 B. Wameke, K.S.J. Pister, “In situ characterization of 20-24 Jan. 2002, p.236-239.
CMOS post-process micromachining,”Sensors and Actuators A [29] W.C. West, et al., “Fabrication and testing of all solid-state
(Physical), Vol. 89, Issue I-2,20 March 2001, pp. 142-151. microscale lithium batteries for microspacecraft applications,”
[7] N. Tea, et al., “Hybrid Postprocessing Etching for CMOS- J. Micromechanics and Microengineering, 12(2002), p. 58-62.
Compatible M E M S , ” l E E E J. MEMS, 6 (1997) 363-372. [30] S. Meninger, et al., “Vibration-to-electric energy
[XI E.J.J. Kruglick, B.A. Warneke, and K.S.J. Pister, “CMOS 3- conversion,” IEEE Trans. VLSI Systems, vo1.9, (no. I), Feb.
axis accelerometers with integrated amplifier,” MEMS 98, 2001, p.64-76.
Heidelberg, Germany, Jan. 25-29, 1998, pp. 631-636. [31] D. Teasdale, V. Milanovic, P. Chang, K.S.J. Pister,
[9] A. Rofougaran, J. Y.-C. Chang, M. Rofougaran, A.A. Abidi, “Microrockets for Smart Dust,” Smart Muterials and
“A I CHI CMOS RF front-end IC for a direct-conversion Strucfures, vol.10, (no.6), Dec. 2001, pp.1145-55.
wireless receiver,” J. Solid State Circuits, 31(1996) 880-889. [32] A. Mason, et al., “A Generic Multielement Microsystem
[ I O ] G. Lin, R.E. Palmer, K.S.J. Pister, and K.P. Roos, “Single for Portable Wireless Applications”, Proc. IEEE, Vol. 86, No.
heart cell force measured in standard CMOS,” TRANSDUCERS 8, Aug. 1998, pp. 1733-1746.
‘97, Chicago, June 16-19, 1 9 9 7 , ~199-200.
. [33] G . Asada, et a[., “Wireless Integrated Network Sensors:
[I I ] H. Bakes, et al., “Micromachined thermally based CMOS Low Power Systems on a Chip,” Proc. 1998 European Solid
microsensors,” Proc. IEEE, 86 (1998) 1660-1678. State Circuits Conf, The Hague, 22-24 Sept. 1998, p. 9-16.
[I21 E.H. Klaassen, R.J. Reay, C. Stoment, and G.T.A. Kovacs, [34] B. Wameke, M. Last, B. Leibowitz, K.S.J. Pister, “Smart
“Micromachined thermally isolated circuits,” Sensors and Dust: Communicating with a Cubic-Millimeter Computer”,
Actuators A, A58 (1997) pp. 43-50. Computer Magazine, Jan. 2001, pp. 44-51.
[I 31 G.K. Fedder, et al., “Laminated high-aspect-ratio [35] B. Wameke, et al., “An Autonomous 16mm’ Solar-
microshuctures in a conventional CMOS.process,” Sensors and Powered Node for Distributed Wireless Sensor Networks,” io
Actuators A (Physical),vol.AS7, (no.2), Nov. 1996. p.103-IO. bepresented at IEEE Sensors 2002, Orlando, FL, June, 2002.

294