Real Signal Processing with Wireless Sensor Networks

1
Real Signal Processing with Wireless Sensor
Networks

• György Orosz, László Sujbert, Gábor Péceli
• orosz,sujbert,peceli_at_mit.bme.hu
• Department of Measurement and Information Systems
• Budapest University of Technology and Economics,
  Hungary
• Regional Conference on Embedded and Ambient
  SystemsRCEAS 2007
• Budapest, Hungary, Nov. 22-24, 2007

2
Wireless signal processing

• Real signal processing
• Fast changing signals
• Hard real-time operation
• Advantages of Wireless Sensor Networks (WSNs)
• Easy to install
• Flexible arrangement
• Difficulties of utilization of WSN
• Data loss
• Limit of the network bandwidth
• Lots of autonomous systems
• Sensor network from signal processing aspects
• Topics
• Signal sensing
• Synchronization
• Distributed signal processing

3
ANC a case study

• Plant to be controlled acoustic system
• Noise sensing
• Berkeley
• micaz motes
• Actuators
• active loudspeakers
• Gateway network ? DSP
• Signal processing
• DSP board
• ADSP-21364 32 bit floating point
• 330 MHz
• 8 analog output channels
• Motes
• TinyOS
• ATmega128
• Sensor boards
• Identification

4
Physical arrangement
5
Sampling precision 1.

• Sampling with low priority
• Shared timer

Sampling with high priority Dedicated timer
6
Sampling precision 2.

• ? Middle level timing priority
• ? 25 samples size packets
• ? Effects of disturbances
• Random disturbance contributes to noise
• Periodic disturbance spurious spectrum lines

Increasing deviation (td) from periodic
disturbance
7
Synchronization 1.

• Delay Td Tt dt
• Unsynchronized subsystems
• Changing delay
• Stability problems in feedback systems
• Goal constant delay
• Ttconst. deterministic protocol
• dtconst. synchronization

8
Synchronization 2.

• Physical synchronization
• Sampling frequencies are the same
• Tuning of the timers
• Interpolation Signal value is estimated in
  signal processing points
• Algorithm transformation algorithm parameters
  are transformed into Ta (when data arrived).
• Synchronization in the ANC system
• Motes physical
• Motes ? DSP linear interpolation

tsyst1
Td1
Td2
tsyst2
Tn
Td1Td2const
d1
f(t)
d2
Ti
t
dt
d3
TSmote
Ta arrival time of data
9
Data transmission methods
Data transmission methods

• Transmission of
• row data
• 1.8 kHz sampling frequency on the motes
• Synchronization of WSN?DSP
• LMS and resonator based ANC algorithms
• Bandwidth restriction about 3 sensors

• Transformed domain
• data transmission
• 1.8 kHz sampling frequency on the motes
• Transmission of Fourier-coefficients
• Increased number of sensors8 sensors (expansion
  possible)

10
Distributed ANC system
A(z)
error signals
acoustic plant
mote2
gateway
control signals
synchronization messages data
(Fourier-coefficients) transmission messages

• Fourier analysis on motes
• Control algorithm on DSP
• Synchronization of base functions
• Computational limits

11
Summary and future plans

• Utilization of WSN in closed loop signal
  processing systems
• Importance of signal observation
• Sampling
• Synchronization
• Distributed signal processing
• Searching for possible ways of data reduction