Link Range and (In)Stability on Sensor Network Architecture

1
Link Range and (In)Stability onSensor Network
Architecture

• Bhaskaran Raman, Kameswari Chebrolu, Naveen
  Madabhushi, Dattatraya Y. Gokhale, Phani K.
  Valiveti, Dheeraj Jain
• Indian Institute of Technology, Kanpur
• The First ACM International Workshop on Wireless
  Network Testbeds, Experimental evaluation and
  CHaracterization (WiNTECH 2006), A MOBICOM 2006
  Workshop
• Presenter Ahey
• Date 2007/05/07

2
Outline

• Motivation
• External antenna communication range study
• Link stability measurements
• Conclusions
• Discussions

3
Motivation (1/2)

• The authors have two claims here
• (1) Use of external antenna
• improve communication range
• more single hop links
• simplify protocol design
• energy efficient
• (2) Dynamic metric based routing questionable due
  to the high variability of RSSI

4
Motivation(2/2)

• What should be the network architecture?
• What is the radio communication range?
• Study communication range, with the use of
  external antennae
• Expected number of hops from/to base node
• Temporal stability of links error rate, RSSI,
  LQI
• Does dynamic distributed routing make sense?
• If so, at what time scale? If not, what else?

5
Outline

• Motivation
• External antenna communication range study
• Link stability measurements
• Conclusions
• Discussions

6
External Antenna Preliminaries

• Cost 50-120
• Form factor 0.5-1m, 0.5- 5kg
• dBi (decibels relative to isotropic)
• G is said to be the gain of the antenna
• When we want to compare antenna gain to an
  absolute level, we use dBi 10log10(G)
• The antenna gain is referenced to an isotropic
  antenna
• Link symmetry is not affected by antenna type
  because transmit gain receive gain

7
Experimental Setup (1/2)

• Tmote sky with CC2420 (802.15.4 compliant)
• Internal antenna 3.1dBi gain
• External connector SMA
• Grid (24dBi, 8o), sector (17dBi, 90o),
  omni (8dBi) ? beam width
• Several antenna combinations
• (1) internal-internal, (2) omni-internal,
• (3) sector-internal, (4) grid-internal, (5)
  omni-omni,
• (6) sector-omni, (7) grid-omni.

8
Experimental Setup (2/2)

• Transmitter
• 6000 packets, about 2 minutes, with one 24 byte
  packet sent every 20ms
• Transmit power 0dBm (max. possible)
• Receiver
• TOSBase, connected to laptop
• Collect RSSI, LQI values
• Environment
• Dense foliage (with heavy path loss)
• Narrow road (with mostly line-of-sight path)

9
Range in Dense Foliage

• Receiver
• on tripod (1.5m)
• Internal ant. only
• Transmitter
• 1.5m
• 6000 packets
• (60 bins x 100 pkts)
• Computer avg. pkt.
• error rate and avg. RSSI
• over 100 packet bins

10
Range in Narrow Road (1/2)

• Transmitter
• at 3.8m for sector/grid
• antennas
• Rows with bold font
• Indicate the approximate
• range of the particular
• transmitter antenna
• Not always the
• farther, the higher

11
Range in Narrow Road (2/2)
12
Implications of Link Range

• More one-hop nodes, lesser hops
• gt better network lifetime
• Foliage range of about 90m
• Useful for applications such as the redwood
  study
• Can have just a single-hop network!
• Volcano monitoring 200-400m range reported
• Range of about 800m with grid antenna
• Useful in situations like Volcano monitor,
  BriMon

13
Outline

• Motivation
• External antenna communication range study
• Link stability measurements
• Conclusions
• Discussions

14
Controlled Calibration Setup

• Step attenuator varied from 0dB to 93dB
• 5000 packet 50 bins x 100 packets
• For each bin error rate, and avg. RSSI
• The receive sensitivity can be found

15
Receive Sensitivity around 90dB
16
RSSI Variability in Other Environment
17
RSSI Variability in Other Environment

• Results similar to Emnet 2006 RSSI is
  underappreciated
• Good correlation between RSSI and error rate
  above some threshold
• But larger spread region (large variability) for
  multi-path prone environment (foliage)
• LQI variability similar to RSSI variability
• 1/LQI, 1/PSR metrics would be unstable

18
Temporal variability in RSSI

• Omni-50m, foliage, avg. pkt. error rate is 7.2
  (neither close to 0 or 100) , without binning
• Large variation (about 15dB)

19
unstable
good
unstable
20
Outline

• Motivation
• External antenna communication range study
• Link stability measurements
• Conclusions
• Discussions

21
Conclusions

• Range 500-800m
• Number of one-hop nodes can be increased
• Better life-time
• Variability in time-scales of 2s, 20s, hours
• Dynamic metric-based routing may not be useful
• Plan for good links, use centralized routing
• Problems arise when RSSI window overlaps with the
  steep region
• Provide sufficient link margin to plan the
  deployment to have good links at the beginning
• Dynamic routing metrics unnecessary

22
Conclusions

• Base node is a powerful node anyway
• Can do centralized routing
• Design, implementation, network might easier
• Think real hard before falling for randomly
  deployed sensor nodes, self-organizing,
  distributed dynamic routing
• Good for solving nice problems on paper
• Practical value questionable

23
Outline

• Motivation
• External antenna communication range study
• Link stability measurements
• Conclusions
• Discussions

24
Discussions

• Contributions Strength
• A good survey about using external antenna to
  improve radio range
• Give some insight about the possibility of
  one-hop sensor network and dynamic routing (for
  outdoor application)
• Provide temporal variation of RSSI within one
  node (which is not discussed in the previous
  Emnet paper)

25
Discussions

• Weakness
• Though RSSI variation is large, it can be a not
  bad link quality predictor as the experiment
  results show. The conclusion here is too strong
  but not totally convincing.
• The duration of the experiment is not long enough
  ( about 2 hours).

26
Relevance to our research

• Han got similar conclusion for our testbed.
  Static routing metrics (to decide whether it is a
  stable link by RSSI from the very beginning) may
  be enough.
• However, for the medium quality links, more
  experiments can be done to quantify the accuracy
  using RSSI, LQI as the predictor of PRR
  dynamically.
• Is one-hop or multi-hop better? Need more network
  overhead v.s reachability analysis.