Communications and Network, 2013, 5, 517-523
http://dx.doi.org/10.4236/cn.2013.53B2095 Published Online September 2013 (http://www.scirp.org/journal/cn)
An Analytical Study of Variable Pr eamble Length-Based
Broadcasting Scheme for WSNs
Arun Kumar 1, Kai-Juan Wong2
1School of Computer Engineering, Nanyang Technological University, Singapore
2Singapore Institute of Technology, Singapore
Received May 2013
ABSTRACT
In this paper, we present a detailed analytical study of previously proposed variable preamble length-based broadcasting
scheme for wireless sensor networks (WSNs). Analytical results show that in dense wireless sensor networks, broad-
casting a packet with a small size preamble can significantly improve the energy conservation on already limited battery
powered sensor nodes. For fairer comparison with variable preamble length-based broadcasting scheme, a comprehen-
sive analytical study for an existing probability-based broadcasting scheme is also presented. Analytically calculated
results show that variable preamble length-based broadcasting scheme is more energy-efficient than the probability-
based br oa dc asti ng scheme .
Keywords: Broadcast; Flooding; Analytical Model; Wireless Sensor Network
1. Introduction
A large-scale wireless sensor network (WSN) can consist
of thousands of sensor nodes and are helpful in many
situations [1], such as disaster relief missions and battle-
field communication facilities, infrastructure protection
and scientific exploration [1-3]. Wireless data transmis-
sion consumes more energy than data processing in a
sensor node. The situation becomes worse in dense wire-
less networks where broadcasting of data is necessary. In
such cases, transmission of a single bit may consume the
same amount of energy as that needed for processing
thousand operations in a typical sensor node.
Wireless sensor networks have broadcast as one of its
most fundamental service. Broadcast provides maximum
message propagation across the whole network and serves
high-level operations, making it critical to the overall
network design. Although broadcasting in wireless sen-
sor networks have many advantages, it can cause serious
problems like “broadcast storm problem” [4] , whi c h could
cause a lot of contention, redundant retransmission, col-
lision and most importantly, waste immense amount of
energy. A possible solution is that the sensor nodes al-
ternate between active and dormant states, which help
them to conserve energy and extend the network lifetime.
Simple broadcasting in this environment, where a node
goes to active and dormant state periodically in asyn-
chronous manner can be very challenging if there are not
enough nodes in the active state when a source wants to
send data.
As wireless communication is the major cause of en er-
gy consumption on sensor nodes, efforts have focused on
how to conserve energy especially at the medium access
control (MAC) layer. While a long preamble is used be-
fore the data packet in BMAC [5], we believe that a
small preamble is enough to achieve sufficient reachabil-
ity and throughput in dense wireless area networks. On
the other hand, it can dramatically save the energy re-
quired to broadcast a packet.
In this paper, we present an analytical study for varia-
ble preamble length-based broadcasting scheme [6]. An
Analytical study for an existing probability-based broad-
casting scheme is also presented and the scheme is im-
plemented to compare with the variable preamble length-
based scheme.
The rest of the paper is organized as follows. Section 2
introduces the motivation behind variable preamble length-
based broadcasting scheme. In Section 3, an analytical
study for variable preamble length-based and probabili-
ty-based broadcasting schemes is given. The parameters
used during the simulations and results are given in Sec-
tion 4. Section 5 concludes the paper and gives some
suggestions for future work.
2. Motivation and Variable Preamble
Length-Based Broadcasting
Wireless sensor networks have an important property that

A. KUMAR, K.-J. WONG
518
many sensor nodes alternate between active and dormant
states, helping them to conserve energy and extend net-
work lifetime. In dense area networks, there is a greater
chance of a node being in an active state to listen to the
channel at a particular time, when compared to a sparse
area network. Simple flooding in dense area networks,
where more nodes are active at a time near the relay node
will only cause energy consumption rather than covering
additional areas in the network, resulting in unnecessary
retransmission of packets and collisions.
Energy efficient MAC protocols are used to conserve
energy of sensor nodes in wireless sensor networks . BMAC
[5] is an energy-saving, distributed, asynchronous and
random-access MAC protocol best suited for wireless
sensor networks. BMAC employs the idea of “preamble
listening” to conserve energy, whereby nodes would du-
ty-cycle the radio receivers by periodically sampling the
channel for activity. In order to transmit a packet, a node
would transmit a long preamble to inform its neighbors
of impending data. This allows B-MAC to support asyn-
chronous and random-access transmission of data pack-
ets. However, a long preamble contains no useful data
and in dense wireless networks it is the major cause of
energy wastage. Preamble length TPreamble in BMAC must
be such that
Preamble Interval
TT>
where TPreamble is the duration of preamble length and
TInterval is the wakeup interval of the nodes.
Variable preamble leng th-based broadcast uses a small
preamble before the data packet in dense wireless sensor
networks [6]. Figure 1 shows the difference between (a)
broadcast using the long preamble before the data pack et
and (b) broadcast using variable preamble length in asyn-
chronous duty cyc led wireless sen sor network. The small
preamble can clearly affect the overall performance of
the broadcasting scheme in dense networks. In case of
small preamble being sent before the data packet, this
preamble wil l only be list ene d by a subs et of nodes present
in the radio range of a relay node or sender node. The
subset of sensor nodes that receives the data packet then
rebroadcasts the packet using the same preamble length.
This practice of sending a small preamble before the data
packet reduces the number of rebroadcasts as well as the
overall energy consumption in the network, as only a
subset of nodes will listen to the preamble and receive
the data packet. Sensor nod es will rebroadcast the packet
only after a random delay.
(a)
(b)
Figure 1. Shows the difference between (a) Broadcast using the long preamble before the data packet and (b) Broadcast using
variable pr eamble length in as ynchronous duty cycled wireless s ensor networks.

A. KUMAR, K.-J. WONG
519
In dense area networks, even small preambles used in
variable preamble length-based broadcast [6] can give
sufficient data dissemination and dramatically save the
energy of the node. In sparse area network, a compara-
tively large preamble will b e required due to the fact tha t
there are lesser chances that a node is active in the neigh-
borhood of the relaying node or the sender node. These
small preambles will save most of the rebroadcasts and
conserve maximum energy in dense area networks. In
case of small preamble being used the preamble length
TPreamble must be such that
Preamble Interval
TT≤
where TPreamble is the duration of the preamble and TInterval
is the wakeup interval of the nodes.
3. Analytical Models
Most of the energy consumption on a node is because of
the node’s radio power consumption. Therefore, analyti-
cal models focus on the radio power consumption of
nodes. Some equations in this section are due to [7].
Table 1 lists the abbreviations and units used in the
equations for the calculation of the energy consumption.
Furthermore, to simplify the comparisons, the analyti-
cal study is based on the following conditions:
• All equations are normalized to one second.
• All the scenarios have same number of nodes and
transmission conditions.
• Both the broadcasting algorithms use preambles be-
fore the data packet but the size of the preamble dif-
fers substantially. The following equation s are used to
calculate the duration of TPreamble and TS_Preamble for the
probability-based and variable preamble length-based
broadcasting schemes respectively:
PreambleIntervalSample Guard
TT TT= ++
(1)
_S PreambleVariableSampleGuard
TTTT= ++
(2)
• To perform channel sampling from an idle state, a
node has to turn its radio on and then sample the
channel. Thus, the time taken to perform channel sam-
pling is calculated to be:
SampleIRx RSSI
T TT= +
(3)
• A node will turn on its receiver and wait for the Radio
Strength Signal Information (RSSI) value to stabilize
before transmitting a packet. Then the node will check
whether the channel is free for the entire duration of
Tcsma. For simplifying the comparison between the
different protocols, the channel is always assumed to
be clear, and therefore no back-offs were considered
during channel contention. At the MAC layer, for
both the broadcasting protocols, the total time, TCSMA,
in which the radio receiver is turned on to perform
CSMA is calculated as follows:
( )
CSMAIRx RSSIcsmaPkts
TTT TN=++ ∗
(4)
Table 1. Notations in the analytical model.
Abbreviation Description Unit
TTx Time taken to transmit a data frame Sec
TRx Tim e taken to receive a data frame Sec
TS_Preamble T he duration for which the small preamble is sent Sec
TVariable The variable minimum time duration for which the small preamble should be sent Sec
TInterval Time between each sampling of the channel during channel listening Sec
TPreamble The minimum duration for which the preamble should be sent Sec
TGuard A guard time to allow for clock skews between nodes Sec
Tcsma Duration for which the channel has to be clear before the node can assume that the channel is free Sec
NNeigh Number of neighbors per node
NPkts Number of packets transmitted per second
TTxb Time taken to transmit a byte Sec
TRxTx, TITx Time taken for the radio to switch from Receive (Rx)/Idle to Transmit (Tx) mode Sec
TIRx Time taken for the radio to switch from Idle to Receive mode Sec
TRSSI Time for RSSI to provide a reading Sec
TSample Time taken from the idle state to sample the channel for activity Sec
PRx Power consumption of the radio in Receive mode mW
PTx Powe r consumption of the radio in Transmit mode mW
PIdle Power consum ption of the radio in Idle Mode mW
P Predefined probability with which a data packet is broadcasted [0.2, 1]

A. KUMAR, K.-J. WONG
520
• As in probab ility-based schemes nodes do not require
global topological information of the network to make
rebroadcast decisions. Every node is allowed to re-
broadcast a packet based on a predetermine d forward-
ing probability P. A random number between 0 and 1
is generated and if this random number is below or
equal to a predefined probability P then the packet is
rebroadcasted.
Random NumberP≤
(5)
• It is assumed that the transmitted signals will always
be received correctly. So the c hannel no ise i s not mod-
elled in the equations.
• In this analysis, for a fairer comparison of both the
broadcasting schemes, the equations and results are
based on “average case” scenarios for probability based
broadcasting scheme and on the “average or worst
case” scenarios for variable preamble length-based
broadcasting scheme.
3.1. Probability-Based Broadcasting Scheme
The total power consumption of the radio in the case of
probability-based broadcasting scheme, EProbBroadcast, is cal-
culated from the durations for which the radio receiver or
transmitter is turned on to perform the following opera-
tions: CSMA (4), transmission of the data packet (7),
reception of a data packet (8), channel listening (9) and
idling ( 10 ) .
__
__
ProbBroadcastCSMAProbBroadcast TxProbBroadcastRx
ProbBroadcast ListenProbBroadcast Idle
E EEE
EE
=++
++
(6)
where, (ECSMA = TCSMA*PRx), (EProbBroadcast_Tx = TProbBroad-
cast_Tx*PTx), (EProbBroadcast_Rx = TProbBroadcast_Rx*PRx), (EProb-
Broadcast_Listen = TProbBroadcast_Listen*PRx) and (EProbBroadcast_Idle
= TProbBroadcast_Idle*PIdle).
Before transmitting a data packet, a node always sends
a long preamble followed by the data packet. Therefore,
the time, TProbBroadcast_Tx, for which the transmitter will be
turned on can be represented by following equation;
( )
_ProbBroadcast TxRxTxPreambleTxPkts
TTTTNP=++∗ ∗
(7)
To receive a data packet, nodes turn their receiver on if
any channel activity is detected during the periodic lis-
tening. These nodes will keep their receiver turned on
until a packet is received and the total time, TProbBroad-
cast_Rx, for which the node receiver is turned on for the
reception of a data packet is represented by following
equation:
( )
( )
_
0.5 * **
ProbBroadcast Rx
PreambleRxPkts Neigh
T
TTNN P= +∗
(8)
For the remainder of the time, when a node is neither
receiving nor transmitting data, the node performs peri-
odic channel listening. The duration, TProbBroadcast_Listen, for
which the receiver is turned on for channel listening is
calculated as follows:
( ) ( )
( )
()
_
__
1
ProbBroadcast Listen
ProbBroadcast TxProbBroadcastRxCSMAInterval
Sample
T
T TTT
T

=−−−


∗
(9)
Finally, the total time TProbBroadcast_Idle, for which a
node’s radio remains in an idle state, is calculated to be:
( ) ( )
( )
_
__
_
1
ProbBroadcast Idle
ProbBroadcast TxProbBroadcastRxCSMA
ProbBroadcast Listen
T
TTT
T
=−−−
−
(10)
The above equations will be true only if the following
condition is satisfied:
TProbBroadcast_Listen ≥ 0
This is explained by the fact that a negative value for
TProbBroadcast_Listen implies that the bandwidth of the chan-
nel has been exceeded.
3.2. Variable Preamble Length-Based
Broadcasting Scheme
The total power consumption of a node in the case of
variable preamble length-based broadcasting scheme can
be calculated from the durations for which the node’s
radio receiver or transmitter is turned on to perform the
following functions: CSMA (4), transmission of the data
packet (12), reception of a data packet (13) or (14), chan-
nel listening (15) and idling (16).
__
__
PreaBroadcastCSMAPreaBroadcast TxPreaBroadcastRx
PreaBroadcast ListenPreaBroadcast Idle
E EEE
EE
=++
++
(11)
where, (ECSMA = TCSMA*PRx), (EPreaBroadcast_Tx = TPreaBroad-
cast_Tx*PTx), (EPreaBroadcast_Rx = TPreaBroadcast_Rx*PRx), (EPrea-
Broadcast_Listen = TPreaBroadcast_Listen*PRx) and (EPreaBroadcast_Idle
= TPreaBroadcast_Idle*PIdle).
Before transmitting a data packet, a node will transmit
a small preamble of fixed size followed by a data packet.
This small preamble will always be less than the regular
preamble sent in probability based broadcasting scheme
as shown in Figure 1. The time which a node takes to
transmit a packet can be represented by following equa-
tion;
( )
__
*
PreaBroadcast TxRxTxSPreambleTxPkts
TT TTN=++
(12)
To receive a data packet, nodes will turn their receiver
on if any channel activity is detected during periodic lis-
tening. These nodes will keep their receiver turned on
until a packet is received and the total time, TPreaBroad-

A. KUMAR, K.-J. WONG
521
cast_Rx, for which the node receiver is turned on for the
reception of a data packet is represented by following
equation:
( )
( )
__
0.5** *
PreaBroadcast RxS PreambleRxPktsNeigh
TT TNN= +
(13)
In the worst case (when the receiver checks the chan-
nel just when the transmitting node starts sending the
preamble) the total time, TPreaBroadcast_Rx, for which the
node receiver is turned on for the reception of a data
packet is represented by following equation:
( )
( )
__
**
PreaBroadcast RxS PreambleRxPktsNeigh
TT TNN= +
(14)
For the remainder of the time, when a node is neither
receiving nor transmitting data, it performs periodic chan-
nel listening. The duration, TPreaBroadcast_Listen, for which
the receiver is turned on for channel listening is calcu-
lated as follows:
( )
( )
( )
()
_
__
1
*
PreaBroadcast Listen
PreaBroadcast TxPreaBroadcastRxCSMAInterval
Sample
T
T TTT
T

=−−−


(15)
Finally, the total time, TPreaBroadcast_Idle, for which a
node’s ra dio remains in the idle state, is calculated to be:
( ) ( )
( )
_
__
_
1
PreaBroadcast Idle
PreaBroadcast TxPreaBroadcastRxCSMA
PreaBroadcast Listen
T
TTT
T
=−−−
−
(16)
As worst case assumptions are being used, the above
equations will be true only if the follow ing cond itions are
satisfied:
TInterval ≥ TS_Preamble + TIRx
TPreaBroadcast_Listen ≥ 0
The above mentioned conditions can be explained by
the fact that a negative value for TPreaBroadcast_Listen implies
that the bandwidth of the channel has been exceeded.
4. Simulation Parameters Used and Results
As the previously proposed variable preamble length-
based broadcasting scheme [6] has been experimentally
evaluated, we now present the preliminary results of the
broadcasting schemes based on analytical study.
Table 2 shows the parameters used for the calculation
of the energy consumption for the broadcasting schemes.
These parameters are determined based on the CC2420
transceiver [8].
Packet broadcasting rate is 1 packet/sec. A d ata packet
is of 32 bytes, time to transmit a byte i s 0 .000 032 sec and
the data rate is 250 kbps. For the evaluation of probabili-
ty-based broadcast the values of TPreamble are recalculated
according to the TInterval (0.01 - 0.05 sec) based on for-
mula (1).
Based on the parameters, the equations are solved to
determine the energy consumption of the broadcasting
nodes un de r different broadc asting c onditions.
4.1. Scenario 1
Figures 2-4 show the energy consumption by the node’s
radios to communicate a packet between the nodes, while
Table 2. Parameters Used To Emulate CC2420.
Abbreviation Unit
TRxTx 0.000192 Sec
TIRx 0.000192 Sec
TRSSI 0.000128 Sec
TSample 0.00032 Sec
TGuard 0.00068 Sec
Tcsma 0.001 Sec
PRx 62.1 mW
PTx 57.4 mW
PIdle 1.41 mW
Figure 2. Energy Consumption (mW) Vs TInterval (sec).
Figure 3. Energy Consumption (mW) Vs TInterval (sec).

A. KUMAR, K.-J. WONG
522
Figure 4. Energy Consumption (mW) Vs TInterval (sec).
the broadcasting node considers 10 nodes in the neigh-
bourhood. The power consumptions by the radios is cal-
culated for different sets of TInterval (0.01 - 0.05 sec).
Figure 2 shows the energy consumption (mW) versus
TInterval (sec). It is observed that as the value of TInterval
increases, the en ergy consumption increases linear ly. This
increase in the energy consumption is due to the long
preamble used f or increased TInterval. Broadcasting a packet
with a low probability shows less energy consumption as
compared to broadcasting a packet with high probability.
The difference in energy consumption of different prob-
abilities of broadcasting is clearly visib le at higher values
of TInterval.
Figures 3 and 4 show the energy consumption (mW)
versus TInterval (sec) for average and worst cases respec-
tively. The value of TInterval varies and the packets are
broadcasted with different preamble length (4 ms to 21
ms). It is observed that as the value of TInterval increases
the energy consumption decreases. Broadcasting a packet
with a small preamble shows less power consumption
compared to broadcasting a packet with a long preamble.
A long preamble always results in high transmission and
reception power consumption. The energy consumption
of the radios in both cases decreases linearly as the TInter-
val increases. This energy saving in case of high TInterval is
because of the small preambles used for broadcasting
packets.
4.2. Scenario 2
Figures 5-7 show the energy consumption by the node’s
radios to communicate a packet between the nod es, wh ile
TInterval is considered as 0.02 sec. The power consump-
tions by the radios is calculated for a range of neighbor-
ing nodes (1 - 10 nodes).
Figure 5 shows the energy consumption (mW) versus
number of neighbors. It is observed that as the number of
neighbours of a broadcasting node increases, the energy
consumption increases linearly. Broadcasting a packet
with a low probability shows less energy consumption as
Figure 5. Energy Consumption (mW) Vs Number of Neigh-
bors.
Figure 6. Energy Consumption (mW) Vs Number of Neigh-
bors.
Figure 7. Energy Consumption (mW) Vs Number of Neigh-
bors.
compared to broadcasting a packet with hig h pr obabili t y.
Figures 6 and 7 show the energy consumption (mW)
versus number of neighbors for average and worst cases
respectively. The number of neighbors of a broadcasting
node varies and the packets are broadcasted with differ-
ent preamble length (4 ms to 21 ms). It is observed that
as the number of neighbors of a broadcasting node in-
creases, the energy consumption increases linearly. Broad-

A. KUMAR, K.-J. WONG
523
casting a packet with a small preamble shows less power
consumption compared to broadcasting a packet with a
long preamble. A long preamble always results in high
transmission and reception power consumption. The energy
consumption of the radios in both cases increases as the
neighboring nodes increases. But a higher energy con-
sumption in worst case is due to the long preamble lis-
tening.
5. Conclusion and Future Work
In this paper, we have presented a comprehensive analyt-
ical model of the variable preamble length-based broad-
casting scheme for wireless sensor networks. An analyt-
ical model for an ex isting probability-based broadcasting
scheme is also presented. Analytical models do not con-
sider the circuit and CPU power consumption in the pre-
sented equations. Our analytical study emphasizes the
fact that a small preamble during broadcast can mitigate
the need for higher energy consumption. Analytical re-
sults show that variable preamble length-based broad-
casting scheme provides higher energy conservation in
dense area networks, when compared to the probability-
based scheme.
In future, we plan to collect these power consumption
results by implementing these broadcasting algorithms
on a physical test bed. Results obtained from analytical
studies, simulations and physical test bed will be used to
compare the broadcasting algorithms.
REFERENCES
[1] T. He, S. Krishnamurthy, L. Luo, T. Yan, L. Gu, R. Sto-
leru, G. Zhou, Q. Cao, P. Vicaire, J. A. Stankovic, T. F.
Abdelzaher, J. Hui and B. Krogh, “VigilNet: An Inte-
grated Sensor Network System for Energy-Efficient Sur-
veilla nc e,” ACM Transaction on Sensor Network, Vol. 2
2006, pp. 1-38.
http://dx.doi.org/10.1145/1138127.1138128
[2] L. Selavo, A. Wood, Q. Cao, T. Sookoor, H. Liu, A. Sri-
nivasan, Y. Wu, W. Kang, J. Stankovic, D. Young and J.
Porter, “LUSTER: Wireless Sensor Network for Envi-
ronmental Research,” Proceedings of the 5th Internation-
al Conference on Embedded Networked Sensor Systems,
Sydney, 2007, pp. 103-116.
[3] N.Xu, S.Rangwala, K. K. Chintalapudi, D. Ganesan, A.
Broad, R. Govindan and D. Estrin, “A Wireless Sensor
Network for Structural Monitoring,” Proceedings of the
2nd International Conference on Embedded Networked
Sensor Systems, Balt imore, 2004, pp. 13-24.
[4] Y.-C. Tseng, S.-Y. Ni, Y.-S. Chen and J.-P. Sheu, “The
Broadcast Storm Problem in a Mobile Ad-hoc Network,”
Wireless Networ k , Vol. 8, 2002, pp. 153-167.
http://dx.doi.org/10.1023/A:1013763825347
[5] J. Polastre, J. Hill and D. Culler, “Versatile Low Power
Media Access for Wireless Sensor Networks,” Proceed-
ings of the 2nd International Conference on Embedded
Networked Sensor Systems, Baltimore, November 2004.
[6] A. Kumar and K. J. Wong, “A Variable Preamble
Length-Based Broadcasting Scheme for Wireless Sensor
Networks,” 7th International Conference on Wireless
Communications, Networking and Mobile Computing,
Wuhan, 23-25 September 201 1, pp. 1-4.
[7] K. J. Wong and D. K. Arvind, “SpeckMAC: Low-Power
Decentralised MAC Protocols for Low Data rate Trans-
missions in Specknets,” Proceedings of the 2nd Interna-
tional Workshop on Multi-Hop Ad Hoc Networks: From
Theory to Reali ty, Florence, 2006, pp. 71-78.
[8] Texas Instrument, 2.4 GHz IEEE 802.15.4/ZigBee-ready
RF Transceiver.
http://focus.ti.com/lit/ds/symlink/cc2420.pdf.