15 November 2005
doc.: IEEE 802.15-<15-05-0682-00-004b>
Analysis of CAP of IEEE 802.15.4
Superframe
Iyappan Ramachandran
University of Washington
November 15, 2005
Submission
15 November 2005
doc.: IEEE 802.15-<15-05-0682-00-004b>
Model Assumptions
• Beacon-enabled Star
– M nodes attached to a coordinator
– All nodes within the carrier sensing range of each other
•
•
•
•
No inactive period in the superframe, i.e. BO=SO
Contention access part (CAP) occupies active period fully
No acknowledgements
Poisson arrival of packets, i.e. probability p of packet arrival
every backoff slot.
• Packet length is fixed and equal to N backoff slots
• No buffering at nodes
• Only Uplink
Submission
15 November 2005
doc.: IEEE 802.15-<15-05-0682-00-004b>
Approximations to simplify analysis
• Presence of beacons and beacon boundaries have
negligible effect
• Every node sees a probability pic that channel is idle
in the first of two CCA backoff slots
– Not slot-to-slot independence; probability that channel is idle
in the second CCA backoff slot is pci|i
– Independence for backoff slots separated by a backoff
• Channel sees a probability ptn that a node begins
transmission in any generic slot
• Geometrically distributed backoff durations with same
mean as original uniform distribution
• Validity of approximations will be verified by
simulations
Submission
15 November 2005
doc.: IEEE 802.15-<15-05-0682-00-004b>
Consequences of approximations
• CAP can be simply analyzed as nonpersistent CSMA
• Channel and nodes have been virtually
decoupled
– Each node can be analyzed independent
of the others
• Probability pkn that node will get out of
kth backoff stage
Submission
15 November 2005
doc.: IEEE 802.15-<15-05-0682-00-004b>
Node state model (see handout #1)
• Node stays in IDLE state with prob. (1-p) and goes to
BO1 with prob. p
• BO1  CS11 with prob. p1n
• CS11CS12 with prob. pic and BO2 with prob. (1-pic)
• CS12TX with prob. pi|ic and CS12 with prob. (1-pi|ic)
…
• … and so on
• TXIDLE with prob. 1 after N backoff slots
• CS51IDLE and CS52 IDLE with probabilities (1-pic)
and (1-pi|ic) respectively
Submission
15 November 2005
doc.: IEEE 802.15-<15-05-0682-00-004b>
Channel state model (handout #2)
• Channel stays in (IDLE, IDLE) state when no
node begins transmission (prob. α=(1-pt|iin))
• (IDLE, IDLE)SUCCESS when exactly one
node transmits (prob. β=Mpt|iin(1-pt|iin)M-1)
• (IDLE, IDLE)FAILURE when more than one
node transmit (prob. δ=1-α-β)
• Channel stays in SUCCESS/FAILURE state
for N backoff slots
• SUCCESS(IDLE,IDLE) and
FAILURE(IDLE,IDLE) with probability 1
Submission
15 November 2005
doc.: IEEE 802.15-<15-05-0682-00-004b>
Calculation of channel throughput
• Approximations have led to virtual decoupling
of nodes’ activities
– Solve node state chain to find ptn in terms of pic 
(1)
– Solve channel state chain to find pic in terms of ptn
 (2)
• Solve (1) and (2) numerically to find pic and
ptn
• Aggregate channel throughput, S is the
fraction of time spent in SUCCESS state
Submission
15 November 2005
doc.: IEEE 802.15-<15-05-0682-00-004b>
Calculation of average power consumption
• Chipcon CC2420 radio for illustration (see handout
#3)
– Four energy states: shutdown, idle, transmit, receive
• Included beacon receptions
• Considered two cases
– Stay in idle state if no packet is waiting  included idle-toreceive ramp-up for beacon reception and CCA
– Shutdown node if no packet is waiting  included shutdownidle-receive ramp-up for beacon reception and CCA
Submission
15 November 2005
doc.: IEEE 802.15-<15-05-0682-00-004b>
Simulations
• All simulations were run in NS-2; used IEEE
802.15.4 module developed by J. Zheng and
M. J. Lee, CUNY
• Same model assumptions, but NO
approximations
• No. of nodes, M=12; Packet length, N=10
backoff slots
• BO=6  Beacon Interval=3072 backoff
slots=0.983 sec; Beacon length=2 backoff
slots
Submission
15 November 2005
doc.: IEEE 802.15-<15-05-0682-00-004b>
Simulations (cont.)
0
Channel
throughput
10
Without shutdown-Analysis
With shutdown-Analysis
Without shutdown-Simulations
With shutdown-Simulations
-1
10
-3
Ave. Power Consumption
per-node [mW]
10
-2
10
-1
10
0
10
0
10
-3
10
-2
10
-1
10
0
10
bytes-per-Joule
capacity [KB/J]
800
600
400
200
0
-3
10
-2
10
-1
10
Per-node packet arrival rate,  [pkts/pkt duration]
Submission
0
10
15 November 2005
doc.: IEEE 802.15-<15-05-0682-00-004b>
Conclusions
• Analysis predicts very accurate throughput
and power consumption estimates
• Although shutting down has the ramp-up
overhead time, it saves considerable energy
at low traffic
• Analysis can be extended
– Easily to include acknowledgements
– With some effort to include inactive part
Submission
15 November 2005
doc.: IEEE 802.15-<15-05-0682-00-004b>
Thank you!
Submission