Transcript Digital Micromirror Device

Reliability of MEMS: Case study
January 30th 2007
Digital Micromirror Device
Begon Martin
Ciapala Richard
Deaki Zoltan
What is a DMD ?
• Matrix of micromirrors
1024 x 768 mirrors for example
• Size of the mirrors: 16 x 16 um
What is a DMD ?
What is a DMD ?
Applications
• Mainly projection systems (Digital Light
Processing)
• Other emerging applications such as 3D
metrology, confocal microscopy,digital TV
Hinge fatigue
• Fatigue: slow growth of a crack driven
by repeated plastic deformation
• Mirror in normal operating mode
switches every 200 microseconds
• 5 years use with 1000 operating hours a
year  mirrors switch 90x109 times
Hinge fatigue
• First approach: anaylsis using bulk
properties of the hinge material 
showed that fatigue would be a big
problem
• However… accelerated tests proved
that wrong, samples easily exceeding
100x109 switches showing no fatigue.
• Explanation: hinge so thin  governed
by thin film properties!
Hinge memory
• Most significant mode of failure
• Occurs when a mirror operates in the
same direction for a long period of time
• Main factors are the duty cycle and the
operating temperature
Duty cycle: percentage of time a mirror is addressed to one
side.(95/5)
Temperature is the dominant factor for hinge memory lifetime
Hinge memory
• Life test under standard condition of 65°C
and 5/95 duty cycle
– Bias voltage has to be increased to annihilate
residual tilt angle.
Evolution of the bias voltage through
the time reported to the number of non
functional micromirros
Micromirrors in the back have a
residual tilt angle compared to the ones
in the front due to the hinge fatigue
Hinge memory
• Cause:
Metal creep of the hinge material
Solutions:
Selection of a new material with lower
degree of metal creep to replace aluminium
Improvemed lifetime by a factor of 5 (1000 hours worst-case
→not good enough).
Implementation of stepped VDD and a
“bipolar reset”
Allowed mirrors to be efficiently controlled over a wider range
of hinge memory. Increased lifetime by a factor of 5 (5000
hours worst-case situation).
Hinge memory
Thermal management of the DMD
• Several sources of heat contribute to
hinge memory:
– Radiant energy from the light source
– Equipment composing the DLP projector and
surrounding the DMD
• Solution:
– Efficient thermal management design required
– Heatsinks on the back of most packages to keep the
temperature as low as possible
– DMD operates at temperatures only 7 to 10 °C above
the projector ambient
Hinge memory
Efficient heat management added to the previously cited
improvements can ensure a lifetime greater than 40000
hours.
Hinge memory mean lifetime estimates over testing time
Stiction
• Induced by an excessive adhesive force
between the landing tip and its landing
site
• Adhesive forces can be induced by:
– Surface contamination
– Capillary condensation
– CMOS defects
– Van der Walls forces
Stiction
• Reliability testing can be done to measure the
distribution of surface adhesion across the
device to determine the number of operating
devices under different switching voltages
Solution to stiction
Environment robustness
Based on standard semiconductor
tests requirement
• Capillarity force
– Humidity everywhere
• UV light exposure
• Thermal testing
• Surface contamination
– During production process
Size vs Robustness
• Small size enable robustness to
mechanical shocks
– Lowest resonant frequency in KHz
– Test :1500G and 20G in vibration with no
mirror breaking
– Weaknesses on the package identified and
annihilated
Optical properties
• Optical properties via glass window
– Enable high quality image
Summary of DMD reliability
• Hinge memory lifetime
>100’000 hours at normal operating
conditions
• Random defects
>650’000 hours MTBF (<1500 FIT)
• Hinge fatigue lifetime
>3.67 trillion cycle or >250’000 hours
• Environmentally robust
Conclusion
• Misleading apparence
– Experience plan must be done to find
critical failure modes
• Concern to reliability
– The reliability of the DMD has been
exemplary and should be considered as a
reference for development of other MEMS