Lecture 14: Intelligent Sensor Systems
g Compensation
n Self-diagnostics, self-calibration, adaptation
g Computation
n Signal conditioning, data reduction, detection of trigger events
g Communications
n Network protocol standardization
g Integration
n Coupling of sensing and computation at the chip level
n Micro electro-mechanical systems (MEMS)
g Others
n Multi-modal, multi-dimensional, multi-layer
n Active, autonomous sensing
Intelligent Sensor Systems
Ricardo Gutierrez-Osuna
Wright State University
Compensation
g Self-diagnostics versus self-calibration
n An intelligent sensor should be able to answer the following
g
g
g
g
Is the output a reasonable value?
Does it agree with the result of an adjacent sensor?
Is the rate of change of the output reasonable?
Is the output actually changing?
g Compensation
n Offset compensation
g To fully utilize the dynamic range of ADCs
n Gain
g By means of programmable gain amplifiers
n Linearity
g By means of look-up tables
n Cross-sensitivity
g Temperature control and/or compensation
g Discussion
n How could you incorporate some of these elements into our e-nose?
Intelligent Sensor Systems
Ricardo Gutierrez-Osuna
Wright State University
Computation
g Various degrees of computation
n
n
n
n
n
Signal conditioning (e.g., filtering)
Signal conversion (e.g., analog to digital)
Logic functions (e.g., triggering events)
Data reduction (e.g., feature extraction)
Decision making (e.g., classification)
g Advanced sensing systems have a hierarchical structure with
different abstraction layers
n LOWER LAYER performs Signal processing
g Conditioning, filtering, conversion, contrast enhancement
n MIDDLE LAYER performs Information processing
g Feature generation, sensor signal fusion and parameter tuning
n UPPER LAYER performs Knowledge processing
g Clustering, prediction, classification, decision making, communications
g Can you identify these layers in our e-nose system?
Intelligent Sensor Systems
Ricardo Gutierrez-Osuna
Wright State University
Hierarchical system structure
UPPER LAYER
Knowledge Processing
MIDDLE LAYER
Information Processing
MIDDLE LAYER
Info. Processing
LOWER LAYER
Sensor Sensor Sensor Sensor
Multi-sensor systems
Intelligent Sensor Systems
Ricardo Gutierrez-Osuna
Wright State University
Audio
Video
Others
Man Machine Interface
4
Processing approaches
g Classical
n Statistical signal processing
n Statistical pattern analysis
g Connectionist
n Multilayer feed-forward neural networks
n Unsupervised learning
g Fuzzy logic
n Fuzzy control
n Fuzzy signal processing
g Evolutionary
n Genetic algorithms
n Genetic programming
g Hybrid approaches
n Neuro-fuzzy
n Neuro-genetic
n
Intelligent Sensor Systems
Ricardo Gutierrez-Osuna
Wright State University
Communications
g Traditionally, each sensor system is custom-designed for
specific applications by experience designers
g This approach has several limitations
n Complexity: a limited number of sensors may be installed in each
system, imposed by the level of complexity that human designers can
deal with
n Cost: system is composed of a small number of highly specialized,
relatively expensive sensors
n Flexibility: the resulting system cannot be easily expanded, modified,
maintained or repaired. Highly trained personnel is required for these
functions
g Solution
n Standardization of transducer interfaces
g Electrical, mechanical(?), communications protocol
n Addition of communication capabilities
g The ideal: Plug-and-play sensors
g Autonomous, distributed, re-configurable sensors
Intelligent Sensor Systems
Ricardo Gutierrez-Osuna
Wright State University
Integration
g On-chip signal conditioning and self-diagnostics constitute
smartness in the ADXL series of accelerometer ICs from Analog
Devices
Intelligent Sensor Systems
Ricardo Gutierrez-Osuna
Wright State University
Integration
DATA ACQUISITION
COMPUTING
Embedded
Micro-controllers
Digital Signal Processors
4,8,16,32-bits
Monitoring devices
Volatile memories
Static RAM
Dynamic RAM
Non-volatile memories
ROM
EEPROM
Flash
Disk-on-a-chip
Instrumentation amplifiers
Filters
Sample and Hold
Analog to Digital Converters
Voltage to Frequency Converters
Multiplexers
Oscillators
Voltage references
Sensor-specific devices
Complete DAQ sub-systems
COMMUNICATIONS
Line drivers
Line receivers
Bus transceivers
Bus controllers
POWER SUPPLY
AC/DC converters
DC/DC converters
Intelligent Sensor Systems
Ricardo Gutierrez-Osuna
Wright State University
CONTROL
Digital to Analog Converters
Frequency to Voltage Converters
Switches
Power drivers
Actuator-specific devices
8
Examples
g Microchip temperature to voltage converters (TC1047)
n Sensor provides an output voltage directly proportional to measured
temperature
g Temperature range: -40C to +125C
g Linear temperature slope: 10mV/ C
g Small 3-pin SOT-23B Package
Intelligent Sensor Systems
Ricardo Gutierrez-Osuna
Wright State University
Examples
g Microchip dual-trip temperature
sensor (TC623)
n Integrated temperature sensor and
logic threshold
n 8-pin DIP or SOIC for direct PCB
mounting
n 2 user-programmable temperature
set-points (w/ external resistor)
n 2 independent temperature limit
outputs
g Application
n Low temp reduces CPU CLK
n High temp further reduces CPU CLK
n Control output starts fan
Intelligent Sensor Systems
Ricardo Gutierrez-Osuna
Wright State University
10
Examples
g Microchip 2-wire serial temperature sensor
n Standard 2-wire serial interface
g Programmable trip point and hystheresis
g Digital readout
g Device configuration
n Multiple operation modes
g Comparator
g Interrupt
g Standby (power management)
n Address lines
g Up to 8 devices can share
the 2-wire bus lines
Intelligent Sensor Systems
Ricardo Gutierrez-Osuna
Wright State University
11
Example
g Dallas Semiconductor 1-Wire digital thermometer (DS18B20)
n
n
n
n
One wire interface requires only one communication pin
Can be powered from a data line
Programmable thermometer resolution from 9 to 12 bits
2 and 3 wire versions are also available
Intelligent Sensor Systems
Ricardo Gutierrez-Osuna
Wright State University
12
Example
g Analog Devices 2-axis accelerometer (ADXL202)
n Can measure both dynamic acceleration (e.g., vibration) and static
acceleration (e.g., gravity)
n The outputs are Duty Cycle Modulated (DCM) signals
g Duty cycles (ratio of pulsewidth to period) proportional to the acceleration in
each of the 2 sensitive axe
g These outputs may be measured directly with a microprocessor counter,
requiring no A/D converter or glue logic.
n If an analog output is desired, an analog
output proportional to acceleration is
available from the XFILT and YFILT pins
g or may be reconstructed by filtering the
duty cycle outputs
n Bandwidth may be set from 0.01 Hz to
6 kHz via capacitors CX and CY
Intelligent Sensor Systems
Ricardo Gutierrez-Osuna
Wright State University
13
Analog Devices 2-axis accel. (ADXL202)
Intelligent Sensor Systems
Ricardo Gutierrez-Osuna
Wright State University
14
Example
g Analog Devices 2-axis self-test accelerometer (ADXL250 )
n Low noise (80dB SNR), wide dynamic range (50g), reduced power
consumption
n A Logic 1 applied to the self-test input will cause an electrostatic force
to be applied to the sensor that will cause it to deflect!
Intelligent Sensor Systems
Ricardo Gutierrez-Osuna
Wright State University
15