Transcript ARL Sensors and Electron Devices Directorate MEMS

A No-Power MEMS Shock Sensor
Luke Currano
U.S. Army Research Laboratory
[email protected]
September 12, 2005
Motivation
 Shock monitoring is important for condition-based maintenance
 There are many MEMS accelerometers available, but all require
some constant operating power
– Electrostatic accelerometers work by monitoring capacitance between a
fixed electrode and a spring-mounted electrode
» Some circuitry is required to monitor capacitance changes and convert them
into voltages
– Piezoelectric accelerometers produce a charge as a result of
acceleration-induced deformation
» No power needed to monitor deflections, but conditioning circuits which
consume power are required to use the sensor output
 Eliminating constant power consumption by MEMS accelerometer
could increase battery lifetime significantly
– 8μA constant current draw (100% duty cycle) at ~3V, for 24μW
continuous power draw
 Health monitoring of long-shelflife or long-lifetime systems without
changing batteries is needed
– Some Army systems have 20-year shelflife combined with limited space
Requirements




Very low power/no power sensing of shock events
Non-destructive (i.e. must be reusable/resetable)
3-axis sensing required, bidirectional (+/-) in each axis
5 levels desired over the range of 10g-150g
Major Accomplishments
 Designed and fabricated functional no-power MEMS
shock sensors
– Up to 7 acceleration threshold levels (one axis, bidirectional) per
1cm2 chip
– Latching demonstrated between 25g and 150g
 Designed and fabricated functional thermal reset
actuators
Shock Sensor Design
Resettable latching no-power MEMS shock sensor
Latch and release mechanism closeup.
 Mechanical latching threshhold sensor design approach
 Silicon MEMS fabrication process allows for very small devices
and very tight tolerances
Design Details

Design set 4
– 4-spring design to make stiffer in z-axis
– Narrowed springs to lower spring
constant
– Added anti-stiction bumps to springs
– Version with metal-coated latch to lower
resistivity in process
– Pyrex cap wafer in process (this is main
impediment to getting test data)
Design Set 4 Shock Sensor
(latched)
Platinum coated latch
for lower resistivity
Shock Sensor Usage
 Designed to be used as either:
– Wakeup sensor
» Power supply connected to processor and other sensors through
shock sensor
» Traditional high-resolution accelerometer used to record shock pulse
after shock sensor wakes system up
» One or more trigger levels
– Mechanical memory
» Shock event triggers device, device “remembers” event
» Interrogate sensor periodically or just before use (go / no go)
» The more trigger levels, the better the resolution
 Either way power savings comes from having system off
most of the time
– Shock sensor itself does not draw any power except small
amount when interrogating/waking up system
Fabrication Process Flow for MEMS Shock Sensor
Starting material – SOI wafer with 20μm
thick device layer, 2 μm oxide
1. Pattern and liftoff Cr/Au bondpads
2. Deep reactive ion etch device layer to
define spring, mass, and latches
3. Isotropically etch the oxide layer
to release the mass
Analytical Modeling


Force balance:
m( yr  ym )  ky  0 (1)
r
r
ky2
 yr y 
(3)
2
2m
y m
r
ky2
 2 yr y 
(4)
m
Set v = 0 to find maximum travel:
ky 2
0  2 yr y 
(5)
m


yL  
T


4 2
(7)
tlatch 

2
m 

k 2
yL
2 yr
(8)
Result: response time is dictated solely by latching distance given a threshold
level
–
Caveat: adding damping to the system allows for slowing the response time but not
speeding it up
 d  1  2 

2myr
(6)
k
For a given level of shock, two of three variables (mass spring constant and
desired deflection to latch) picked by designer, the third is solved from (6)
Time to latch due to an impulse is determined by natural frequency of device:
tlatch 

ky
(2)
m
Integrate equation 2, using the fact that: a dy = v dv:
2
y m r

ym  yr 
Response times of ARL designs 2.2ms or lower
Experimental Results - Latch


Centrifuge test of devices designed to latch at static levels of 10G -75G
Visually and electrically confirmed latching during centrifuge tests
Designed Trigger Level
(G’s)
Centrifuge Actual Trigger Level
(G’s)
10
25
25
57
50
95
75
142
– Factor of ~2 between designed trigger level and actual level
– This is attributed to simplification in model – not including interaction of mass and
latch (friction and normal force both contribute to resist motion of mass once in
contact)
– Complete nonlinear model is under development

Shock table tests
– Large amount of out-of-plane vibration
– Out-of-plane vibration caused devices to reset themselves
– Cap chip needed – packaging process under development
Thermal Reset Actuator
• 15V, 125mA currently required to reset the devices
• Pulse duration 10ms
• Vacuum packaging or removing the substrate underneath the device will
decrease the power required by 75%
• 20mA, 10.1V for 15μm deflection in air
• 10mA, 5.1V for 15μm deflection at 250mT
Future Work
 Packaging
– Wafer-level encapsulation of devices is critical to produce chips
with 5 level, three axis shock sensors with a small enough form
factor
– Also critical to produce any reasonable number of testable
devices, since devices often fracture when cleaving wafer
– Lower voltage and power requirements of reset actuators
through vacuum packaging
 Modeling
– Complete contact/friction model to more accurately predict
trigger level
 Refine design for more robust, smaller sensors