High Performance Inertial Navigation Grade Sigma-Delta MEMS … · 2016. 2. 15. · The combination...
Transcript of High Performance Inertial Navigation Grade Sigma-Delta MEMS … · 2016. 2. 15. · The combination...
High Performance Inertial
Navigation Grade Sigma-Delta
MEMS AccelerometerP. Zwahlen, Y. Dong, A-M. Nguyen, F.Rudolf, Colibrys SA
P. Ullah, V. Ragot, Sagem
September 18th 2012
2
Why a MEMS accelerometer for high end applications ?
� MEMS accelerometers are widely spread in the automotive, consumer
and industrial markets
� MEMS accelerometers are starting to replace established, expensive
and fragile high-end electromechanical devices
� MEMS accelerometers can offer same or better performance at lower
cost, lower power consumption, smaller size and greater strength
� Some MEMS accelerometers have already penetrated on civilian and
defence programs (Airbus and Boeing aircrafts, gun lauched smart
munitions, *)
� MEMS accelerometers will soon be compatible with high-end inertial navigation systems
3
Objectives
AIDA: a MEMS accelerometer for inertial navigation
� Need
– Lower cost compared to non-MEMS solutions
– High performance� Full scale ~15g
� Bias stability < 1mg
� Rectification < 10 µg/g²
� Scale factor < 1000 ppm
� Bandwidth > 300 Hz
� Low frequency noise ~1 µg/√Hz (equivalent to 18bit resolution for 300Hz bandwidth)
� Architecture
– Bulk silicon pendular MEMS� Robust, stable
– Capacitive sense/actuation� Detector == actuator
4
Accelerometer architecture
� Open-loop limitations
– Noise� Brownian noise due to gas damping
– Linearity� Squeeze-film effect, electrostatic forces (non-linear functions of gap)� Vibration-induced rectification
– Bandwidth� Limited by natural frequency and damping
� Closed-loop solution
– Noise� Operation at high Q for the MEMS possible. Allows reduction of Brownian noise� Position control achieved through electronic regulation
– Linearity� Reduced seismic mass excursion thanks to feedback control. Improved vibration rectification
– Bandwidth� Signal bandwidth extension up to over 10kHz � Precise data time stamp is essential for inertial navigation and guidance applications
555
Part I: System Description
6
Closed-loop sensor System Architecture
� Capacitive bulk micromachined MEMS Chip
� Capacitive position detector including low resolution ADC (7 bit
equivalent)
� Digital loop filter (position control)
� Oversampled Sigma-Delta converter
– Pulse density modulation
– 1-bit comparator � Bitstream output
– 1-bit DAC applying constant voltage to bottom or top electrode
� Acceleration mean value estimated through high linearity 1-bit DAC
7
Sigma-Delta converter for accelerometer
Sigma-Delta principle
� High frequency 1-bit conversion instead of high resolution
� High quantization noise rejected in high frequency by noise shaping
concept
� High linearity achieved by averaging
Low passband noise
Increased computing
complexity
Hardware simplicity
(analog detection,
actuation voltage)
High sampling frequencyHigh linearity
Linearization of
electrostatic forces
Trade-offsAdvantages( )
2
2
0
2
2
1
e
VAFe
ref
r
⋅⋅⋅⋅⋅= εε
8
MEMS sensor and detection
MEMS Sensor
� 3-stack silicon electrode assembly
� High temperature Silicon Fusion Bonding
(SFB) technology resulting in highly
stable assembly
� Time multiplexing concept
– Allows usage of same electrodes for both sense readout and forcing
Capacitive detection
� Voltage amplifier topology
� Analog modulation / demodulation
(Correlated Double Sampling)
� Switched capacitor
– Chosen for its versatility to interface with different size MEMS capacitors
– Dedicated phase for charge injection removal
Mass
SpringTop electrode
Bottom electrode
Middle electrode
Mass
SpringTop electrode
Bottom electrode
Middle electrode
9
Testboard V2 Accelerometer interface
Sigma-Delta board V2
� Test board
– FPGA for digital filtering, decimation & control sequencing
– Clock oscillator– Power supply decoupling capacitors– Communication interface
� ASIC & MEMS & Temp sensor packaged inside a standard JLCC-44 package
10
System design for high stability
� High stability MEMS sensor
– Careful spring anchoring design
– 3-stack Silicon MEMS technology with Silicon Fusion Bonding technologies for excellent long-term stability and hermeticity
– Die attach stress decoupling technique
� Detection chain offset reduction
– CDS technique
– Charge injection removal
� High stability voltage reference
– Low impedance output up to high frequency
– Low noise
– High stability
� Matching and repeatability of electrode switching operation
111111
Part II: Performance Results
12
Noise Transfer Function
CIC filtered output
Bitstream output
� White noise: 1µµµµg/sqrt(Hz)
� Noise bandwidth: 300 Hz
13
Temperature modeling Bias (K0)
� Bias temperature
slope: Class 200 µµµµg/°C
� Low bias residues:
– < 300µg
14
Temperature modeling Bias (K0): Statistical distribution
� Statistical evaluation over a population of 11 boards
– all components are based on a single MEMS design and technology
� Thermal bias
slopes
distribution is
below 150 µµµµg/°C
15
Temperature modeling Scale Factor (K1)
� Excellent Scale
Factor
repeatability
(device to device)
� Scale factor
temperature slope:
Class 100 ppm/°C
� Low scale factor
residues:
– < 200 ppm
16
Temperature modeling Scale Factor: Statistical distribution
� Scale factor temperature slope distribution below 100 ppm/°C
� Scale factor residues distribution: < 200 ppm
17
Temperature modeling Misalignment (Kp)
� Misalignment
temperature slope:
< 50 µµµµrad/°C
� Misalignment
residues:
– < 60 µrad
18
Temperature modeling Misalignment: Statistical distribution
� Misalignment temperature slope distribution: < 50 µµµµrad/°C
� Misalignment residues distribution: < 60 µµµµrad
19
Short-term bias stability under warm-up
� Controlled within +/-10 µµµµg
under warm-up condition
with a 10g FS sensor
– � 1 ppm bias stability
� Limited warm-up transient
left after thermal
compensation
� Warm-up potential after low-
pass filtering over 160s &
time derivation
– 8µg/mn (obtained over a larger sample)
– Compliant with gyrocompass alignment requirements
data averaged over 1s
20
Bias stability / Allan variance
� 10 sec of data observation is
enough to get micro-g signal
precision
� Signal stability is guaranteed at
observation time of at least up
to 300 sec
external vibrations noises
Bias instability (Random flicker noise)
Bias stability: Temporal, Allan Variance, PSD
21
Vibrations: 2nd order non-linearity (K2)
Non-linearity (K2)
� K2(f) device to device
repeatability
� K2 < 10 µµµµg/g2 (0 to 100 Hz)
� K2 < 20 µµµµg/g2 (up to 1 kHz)
22
Performance review
g4000Shock resistance
[0; 100] Hz / [100; 1000] Hzµµµµg/g210 / 20Vibration (K2)
mW100Power consumption
After 3rd order polynomial curve fittingµµµµrad60Kp residues
After 3rd order polynomial curve fittingppm200K1 residues
ppm/°C100K1 Temperature slope
Scale Factor (K1)
µµµµrad/°C
µµµµg
µµµµg/°C
µg/mn
µµµµg/√√√√Hz
g
Unit
50Kp Temperature slope
Misalignment (Kp)
After 3rd order polynomial curve fitting300Temperature residues
max150Temperature bias slope
8Short-term (under warm-up)
Bias stability (K0)
typ.1Noise
typ.15Full scale range
CommentValue
23
Conclusions
� The combination of MEMS pendular and sigma/delta allows "medium
navigation grade" class performances with gyrocompass functions.
– Bias stability� 8 µg/min bias stability under warm-up condition
� 300 µg of residues
– Scale factor� 200 ppm residues
� MEMS based accelerometer highlights
– High shock tolerance
– Low weight
– Size
– Cost (Batch manufacturing process)
� The next developments will focus on packaging and system integration.
24
Thank you for your attention